Principles of Psychiatric

Genetics

Edited by
John I. Nurnberger Jr. MD, PhD
Joyce and Iver Small Professor of Psychiatry
Professor of Medical and Molecular Genetics and Medical Neuroscience
Director of the Institute of Psychiatric Research
Department of Psychiatry
Indiana University School of Medicine
Indianapolis, IN, USA

Wade H. Berrettini MD, PhD

Karl E. Rickels Professor of Psychiatry
Director, Center for Neurobiology and Behavior
University of Pennsylvania School of Medicine
Philadelphia, PA, USA
CAMBRIDGE UNIVERSITY PRESS
Cambridge, New York, Melbourne, Madrid, Cape Town,
Singapore, São Paulo, Delhi, Mexico City
Cambridge University Press
The Edinburgh Building, Cambridge CB2 8RU, UK

Published in the United States of America

by Cambridge University Press, New York

www.cambridge.org
Information on this title: www.cambridge.org/9780521896498

# Cambridge University Press 2012

This publication is in copyright. Subject to statutory exception

and to the provisions of relevant collective licensing agreements,
no reproduction of any part may take place without
the written permission of Cambridge University Press.

First published 2012

Printed and bound in the United Kingdom by the MPG Books Group

A catalogue record for this publication is available from the British Library

Library of Congress Cataloging-in-Publication Data

Principles of psychiatric genetics / edited by John I. Nurnberger,
Wade H. Berrettini.
p. ; cm.
ISBN 978-0-521-89649-8 (Hardback)
I. Nurnberger, John I., 1946– II. Berrettini, Wade.
[DNLM: 1. Mental Disorders–genetics. 2. Mental Disorders–
epidemiology. WM 140]
616.890 042–dc23
2012000081

ISBN 978-0-521-89649-8 Hardback

Cambridge University Press has no responsibility for the persistence or

accuracy of URLs for external or third-party internet websites referred to
in this publication, and does not guarantee that any content on such
websites is, or will remain, accurate or appropriate.

Every effort has been made in preparing this book to provide

accurate and up-to-date information which is in accord with
accepted standards and practice at the time of publication.
Although case histories are drawn from actual cases, every effort has
been made to disguise the identities of the individuals involved.
Nevertheless, the authors, editors and publishers can make no
warranties that the information contained herein is totally free from
error, not least because clinical standards are constantly changing
through research and regulation. The authors, editors and
publishers therefore disclaim all liability for direct or consequential
damages resulting from the use of material contained in this book.
Readers are strongly advised to pay careful attention to information
provided by the manufacturer of any drugs or equipment that they
plan to use.
Contents
List of contributors vii
Preface xi
1 Contribution of genetic epidemiology to
our understanding of psychiatric
disorders 1
Kathleen Ries Merikangas and Anibal Cravchik
2 A basic overview of contemporary human
genetic analysis strategies 13
Ondrej Libiger and Nicholas J. Schork
3 DNA methods 23
David W. Craig
4 In silico analysis strategies and resources for
psychiatric genetics research 34
Ali Torkamani, Trygve Bakken, and Nicholas
J. Schork
5 Gene expression studies in psychiatric
disorders 49
Alexander B. Niculescu, III
6 Pharmacogenetics in psychiatry 53
Falk W. Lohoff
7 Functional validation of candidate genetic
susceptibility factors for major mental
illnesses 69
Akira Sawa, Wanli W. Smith, Saurav Seshadri,
Akiko Hayashi-Takagi, Hanna Jaaro-Peled,
and Atsushi Kamiya
8 Epigenetic mechanisms in drug addiction
and depression 79
William Renthal and Eric J. Nestler
9 Panic disorder 90
Ardesheer Talati and Myrna M. Weissman
10 The genetics of the phobic disorders
and generalized anxiety disorder 112
Raymond R. Crowe
11 Genetic contributions to obsessive–
compulsive disorder (OCD) and
OCD-related disorders 121
Dennis L. Murphy, Pablo R. Moya,
Jens R. Wendland, and Kiara Timpano
12 Post-traumatic stress disorder 134
Michael J. Lyons, Tyler Zink, and Karestan
C. Koenen
13 Antisocial behavior: gene–environment
interplay 145
Laura A. Baker, Catherine Tuvblad, Serena
Bezdjian, and Adrian Raine
14 Learning disabilities 160
Shelley D. Smith
15 Attention-deficit hyperactivity disorder 168
Josephine Elia, Francesca Lantieri, Toshinobu
Takeda, Xiaowu Gai, Peter S. White, Marcella
Devoto, and Hakon Hakonarson
16 Autism and autism spectrum disorders 183
Daniel H. Geschwind and Maricela Alarcón
17 The genetics of bipolar disorder 196
John R. Kelsoe
18 The genetics of major depression 212
James B. Potash
19 The genetics of schizophrenia 230
Hugh M. D. Gurling and Andrew McQuillin
20 The genetics of anorexia and bulimia
nervosa 262
Andrew W. Bergen, Jennifer Wessel, and Walter
H. Kaye
21 Genetics and common human obesity 272
R. Arlen Price
v
 Contents
22 Alcoholism 279
Howard J. Edenberg
23 Nicotine dependence 287
Sarah M. Hartz and Laura J. Bierut
24 Human molecular genetics of opioid
addiction 297
Mary Jeanne Kreek, Dmitri Proudnikov,
David A. Nielsen, and Vadim Yuferov
25 Genetics of stimulant dependence
Joseph F. Cubells and Yi-Lang Tang
26 Genetics of personality disorders
C. Robert Cloninger
27 Ethical issues in behavioral
genetics 324
Stephen H. Dinwiddie, Jinger Hoop,
and Elliot Gershon
vi
28 Genetics of Tourette syndrome
and related disorders 336
Maria G. Motlagh, Thomas V. Fernandez,
and James F. Leckman
29 Endophenotypes in psychiatric genetics 347
Andrew C. Chen, Madhavi Rangaswamy,
and Bernice Porjesz
30 Developmental disorders 363
Craig A. Erickson, Khendra I. Peay,
306
and Christopher J. McDougle
31 Alzheimer’s disease 371
316
Carlos Cruchaga, John S. K. Kauwe,
and Alison M. Goate
Index 382
Color plate section is between pp. 212 and 213.
Contributors
Maricela Alarcón, PhD
Center for Autism Research and Treatment,
Semel Institute of Neuroscience, Program in
Neurogenetics, Department of Neurology, David
Geffen School of Medicine, University of California
at Los Angeles, Los Angeles, CA, USA
Laura A. Baker, PhD
Department of Psychology, University of Southern
California, Los Angeles, CA, USA
Trygve Bakken
The Scripps Translational Science Institute
and The Department of Molecular and Experimental
Medicine, The Scripps Research Institute;
Graduate Program in Neurosciences and Medical
Scientist Training Program, University of California,
San Diego, La Jolla, CA, USA
Serena Bezdjian
Department of Psychology, University of Southern
California, Los Angeles, CA, USA
Andrew W. Bergen, PhD
Molecular Genetics Program, Center for Health
Sciences, SRI International, Menlo Park,
CA, USA
Laura J. Bierut, MD
Department of Psychiatry, Washington University
School of Medicine, St. Louis, MO, USA
Andrew C. Chen, MD
Department of Psychiatry, Columbia University
Medical Center College of Physicians and Surgeons,
New York, NY, USA
C. Robert Cloninger, MD
Department of Psychiatry, Washington University
School of Medicine, St. Louis, MO, USA
David W. Craig, PhD
The Translational Genomics Research Institute,
Phoenix, AZ, USA
Anibal Cravchik, MD, PhD
Genetic Epidemiology Research Branch,
National Institute of Mental Health Intramural
Research Program, National Institutes of Health,
Bethesda, MD, USA
Raymond R. Crowe, MD
Department of Psychiatry, University of Iowa Carver
College of Medicine, Iowa City, IA, USA
Carlos Cruchaga, PhD
Department of Psychiatry, Washington University
School of Medicine, St. Louis, MO, USA
Joseph F. Cubells, MD, PhD
Departments of Human Genetics and Psychiatry and
Behavioral Sciences, Emory University School of
Medicine, Atlanta, GA, USA
Marcella Devoto, PhD
Division of Genetics, Department of Pediatrics, The
Children’s Hospital of Philadelphia and Department
of Biostatistics and Epidemiology, University of
Pennsylvania School of Medicine, Philadelphia, PA,
USA; Dipartimento di Medicina Sperimentale,
University La Sapienza, Rome, Italy
Stephen H. Dinwiddie, MD
Department of Psychiatry and Behavioral Science,
The University of Chicago Medical Center,
Chicago, IL, USA
Howard J. Edenberg, PhD
Center for Medical Genomics, Department
of Biochemistry and Molecular Biology,
Indiana University School of Medicine,
Indianapolis, IN, USA
vii
 List of contributors
Josephine Elia, MD
Department of Child and Adolescent Psychiatry,
The Children’s Hospital of Philadelphia and
The University of Pennsylvania,
Philadelphia, PA, USA
Craig A. Erickson, MD
Department of Psychiatry, Indiana University School
of Medicine; Christian Sarkine Autism Treatment
Center, James Whitcomb Riley Hospital for Children,
Indianapolis, IN, USA
Thomas V. Fernandez, MD
Child Study Center, Yale University School
of Medicine, New Haven, CT, USA
Xiaowu Gai, PhD
Department of Pharmacology, Loyola University,
Chicago, IL, USA
Elliot Gershon, MD
Department of Psychiatry, The University of Chicago
Medicine, Chicago, IL, USA
Daniel H. Geschwind, MD, PhD
Center for Autism Research and Treatment, Semel
Institute of Neuroscience, Program in Neurogenetics,
Department of Neurology, David Geffen School
of Medicine, and Department of Human Genetics,
University of California at Los Angeles, Los Angeles,
CA, USA
Alison M. Goate, D.Phil
Departments of Psychiatry, Neurology, Alzheimer’s
Disease Research Center, Genetics, and Hope Center
for Neurological Disorders, Washington University
School of Medicine, St. Louis, MO, USA
Hugh M. D. Gurling, MD
Molecular Psychiatry Laboratory, Mental Health
Sciences Unit, University College London,
London, UK
Hakon Hakonarson, MD, PhD
Division of Pulmonary Medicine, Department of
Pediatrics, The Center for Applied Genomics,
The Children’s Hospital of Philadelphia and The
University of Pennsylvania, Philadelphia, PA, USA
Sarah M. Hartz, MD, PhD
Department of Psychiatry, Washington University
School of Medicine, St. Louis, MO, USA
viii
Akiko Hayashi-Takagi, PhD
Department of Psychiatry, Johns Hopkins University
School of Medicine, Baltimore, MD, USA
Jinger Hoop, MD, MFA
Edward Hines Jr. Veteran's Administration Hospital,
Hines, IL, USA
Hanna Jaaro-Peled, PhD
Department of Psychiatry, Johns Hopkins University
School of Medicine, Baltimore, MD, USA
Atsushi Kamiya, MD, PhD
Department of Psychiatry, Johns Hopkins University
School of Medicine, Baltimore, MD, USA
John S. K. Kauwe, PhD
Department of Biology, Brigham Young University,
Provo, UT, USA
Walter H. Kaye, MD
University of California San Diego, Department
of Psychiatry and Eating Disorder Research and
Treatment Program, La Jolla, CA, USA
John R. Kelsoe, MD
Department of Psychiatry and Institute for Genomic
Medicine, University of California San Diego, VA
San Diego Healthcare System, La Jolla, CA, USA
Karestan C. Koenen, PhD
Harvard School of Public Health, Boston,
MA, USA
Mary Jeanne Kreek, MD
The Laboratory of the Biology of Addictive Diseases,
Rockefeller University, New York, NY, USA
Francesca Lantieri, PhD
Department of Child and Adolescent Psychiatry,
The Children’s Hospital of Philadelphia,
Philadelphia, PA, USA
James F. Leckman, MD
Child Study Center and Departments of Psychiatry,
Psychology, and Pediatrics, Yale University School of
Medicine, New Haven, CT, USA
Ondrej Libiger, MA, PhD
The Scripps Translational Science Institute and
Department of Molecular and Experimental
Medicine, The Scripps Research Institute,

La Jolla, CA, USA; Lekarska Fakulta v Hradci
Kralove, Charles University, Czech Republic
Falk W. Lohoff, MD
Department of Psychiatry, University of Pennsylvania
School of Medicine, Philadelphia, PA, USA
Michael J. Lyons, PhD
Department of Psychology, Boston University,
Boston, MA, USA
Christopher J. McDougle, MD
Professor of Psychiatry and Pediatrics, Director of the
Lurie Center for Autism, Massachusetts General
Hospital and MassGeneral Hospital for Children,
Harvard Medical School, Boston, MA, USA
Andrew McQuillin, PhD
Molecular Psychiatry Laboratory, Research
Department of Mental Health Sciences, University
College London, London, UK
Kathleen Ries Merikangas, PhD
Genetic Epidemiology Research Branch, National
Institute of Mental Health Intramural Research
Program, National Institutes of Health,
Bethesda, MD, USA
Maria G. Motlagh, MD
Child Study Center, Yale University School of
Medicine, New Haven, CT, USA
Pablo R. Moya, PhD
Laboratory of Clinical Science, National Institute of
Mental Health Intramural Research Program,
National Institutes of Health, Bethesda, MD, USA
Dennis L. Murphy, MD
Laboratory of Clinical Science, National Institute of
Mental Health Intramural Research Program,
National Institutes of Health, Bethesda, MD, USA
Eric J. Nestler, MD, PhD
Fishberg Department of Neuroscience, Mount Sinai
School of Medicine, New York, NY, USA
Alexander B. Niculescu, III, MD, PhD
Department of Psychiatry, Indiana University School
of Medicine; Indianapolis, IN, USA
David A. Nielsen, PhD
The Laboratory of the Biology of Addictive Diseases,
Rockefeller University, New York, NY, USA
List of contributors
Khendra I. Peay, MD
Department of Psychiatry, Indiana University School
of Medicine; Christian Sarkine Autism Treatment
Center, James Whitcomb Riley Hospital for Children,
Indianapolis, IN, USA
Bernice Porjesz, PhD
Henri Begleiter Neurodynamics Laboratory,
SUNY Downstate Medical Center, Brooklyn,
New York, NY, USA
James B. Potash, MD, MPH
Department of Psychiatry, Carver College of
Medicine, University of Iowa, Iowa City, IA, USA
R. Arlen Price, PhD
University of Pennsylvania, Center for
Neurobiology and Behavior, Translational
Research Laboratories, Philadelphia,
PA, USA
Dmitri Proudnikov, PhD
The Laboratory of the Biology of Addictive Diseases,
Rockefeller University, New York,
NY, USA
Adrian Raine, DPhil
Departments of Criminology, Psychiatry, and
Psychology, Jerry Lee Center of Criminology,
University of Pennsylvania, Philadelphia, PA, USA
Madhavi Rangaswamy, PhD
Henri Begleiter Neurodynamics Laboratory,
SUNY Downstate Medical Center, Brooklyn,
New York, USA
William Renthal, MD, PhD
Medical Scientist Training Program, The University
of Texas Southwestern Medical Center,
Dallas, TX, USA
Akira Sawa, MD, PhD
Departments of Psychiatry, Neuroscience, Cellular
and Molecular Medicine, Human Genetics, Johns
Hopkins University School of Medicine,
Baltimore, MD, USA
Nicholas J. Schork, PhD
The Scripps Translational Science Institute
and The Department of Molecular and Experimental
Medicine, The Scripps Research Institute,
La Jolla, CA, USA
ix
 List of contributors
Saurav Seshadri
Department of Psychiatry, Johns Hopkins University
School of Medicine, Baltimore, MD, USA
Shelley D. Smith, PhD, FACMG
University of Nebraska Medical Center, Munroe
Meyer Institute for Genetics and Rehabilitation,
Omaha, NE, USA
Wanli W. Smith, MD, PhD
Department of Psychiatry, Johns Hopkins University
School of Medicine, Baltimore, MD, USA
Toshinobu Takeda, MD, PhD
Department of Child and Adolescent Psychiatry,
The Children’s Hospital of Philadelphia,
Philadelphia, PA, USA; Department of Psychiatry,
Ryukoku University, Kyoto, Japan
Ardesheer Talati, PhD
Columbia University and New York State Psychiatric
Institute, New York, NY, USA
Yi-Lang Tang, MD, PhD
Department of Psychiatry and Behavioral Sciences,
Emory University School of Medicine, Atlanta, GA, USA
Kiara Timpano, PhD
University of Miami, Coral Gables, FL, USA
Ali Torkamani, PhD
The Scripps Translational Science Institute and The
Department of Molecular and Experimental
Medicine, The Scripps Research Institute,
La Jolla, CA, USA
x
Catherine Tuvblad, PhD
Department of Psychology, University of
Southern California, Los Angeles,
CA, USA
Myrna M. Weissman, PhD
Columbia University and New York State Psychiatric
Institute, New York, NY, USA
Jens R. Wendland, MD
Laboratory of Clinical Science, and Genetic
Basis of Mood and Anxiety Disorders Section,
National Institute of Mental Health Intramural
Research Program, National Institutes of Health,
Bethesda, MD, USA
Jennifer Wessel, PhD
Molecular Genetics Program, Center for Health
Sciences, SRI International, Menlo Park, CA;
Department of Public Health, Indiana
University School of Medicine, Indianapolis,
IN, USA
Peter S. White, PhD
Department of Pediatrics, The Children’s Hospital
of Philadelphia and The University of Pennsylvania,
Philadelphia, PA, USA
Vadim Yuferov, PhD
The Laboratory of the Biology of Addictive Diseases,
Rockefeller University, New York, NY, USA
Tyler Zink
Department of Psychology, Boston University,
Boston, MA, USA
Preface
The major psychiatric disorders are common illnesses
with complex origins in gene–environment inter-
actions. As is the case with most medical diseases,
such as diabetes and hypertension, genetic factors
for psychiatric disorders are composed of many
common alleles, each with a small effect on risk.
Additionally, there are many rare alleles, including
copy number variants (CNVs), with larger effects
on risk.
Despite this complex picture, substantial progress
has been made in identification of individual risk
alleles, through various molecular approaches, includ-
ing linkage and association, studies of epigenetic
factors, and recently direct sequencing of DNA from
affected persons. It is the intent of the contributors
and editors to describe these recent molecular
advances, in the context of the genetic epidemiology
(population, family and twin studies) and our know-
ledge of the phenomenology and course of illness of
psychiatric disorders.
This book is organized in chapters according to
the current nosology of psychiatric disorders, but the
reader should not conclude that this nomenclature is
based on genetic, epigenetic, neurobiologic, or envir-
onmental influences. The current nosology is based
primarily on acute signs and symptoms, few of
which are “pathognomonic”. Individuals with pat-
terns of these signs and symptoms are categorized
as “having” schizophrenia, for example, but the term
“schizophrenia” undoubtedly refers to a very hetero-
geneous group of brain disorders which share some
acute signs and symptoms and course of illness vari-
ables. In the future, it is expected that the current
nosology will be transformed into one which reflects
knowledge of the neurobiological and experiential
origins of these groups of heterogeneous disorders.
It is anticipated that new diagnostic tools and thera-
peutic agents will result from this expanded molecu-
lar knowledge.
While this book is, by necessity, a picture in time
of current knowledge, the rapidly advancing technol-
ogy of DNA sequencing is likely to produce a multi-
tude of new discoveries in the near future. During the
past decade, the cost of sequencing a human genome
has fallen from ~ $1 000 000 000 USD to ~ $5000
USD, and the “thousand dollar genome” is widely
predicted. This will allow for the sequencing of thou-
sands of affected individuals within each category of
psychiatric disorder. Analysis of this sequence infor-
mation will permit the development of a catalogue of
common and rare genetic variants that increase risk
for these diseases. Studies of gene expression in post-
mortem brain samples and in living blood and skin
tissue from affected persons will enable catalogues of
epigenetic events to be developed as well. These
advances in genetics and epigenetics should permit
an explosion of knowledge concerning the genetic and
environmental risks for psychiatric disorders.
It is now possible to develop and culture neurons
in the laboratory, from easily obtained skin or blood
cells of persons with a psychiatric disorder. This
should allow for an unprecedented correlation of
neuronal phenotype to genotype on a scale we can
only imagine. There will be the potential to character-
ize in beautiful detail the electrophysiologic, morpho-
logic, and neurochemical characteristics of these
neurons from genetically defined origins. This should
permit discovery of specific neuronal abnormalities.
That would enable the targeting of therapeutic agents
to pathophysiology, even if that pathophysiology is
unique to an individual person or family
The book you have before you should be seen as a
window to that day in the future when such targeted,
individualized therapies are used by doctors and
patients around the world. We have designed Prin-
ciples of Psychiatric Genetics to be useful to investi-
gators in related areas, including Psychiatry, Human
Genetics, and Neurobiology. However we also expect
that it will be of value to practicing clinicians who
xi
 Preface
wish to understand the sometimes confusing and
contradictory reports of discoveries in the media. Is
there a blood test for bipolar disorder? Has “a gene”
been identified for alcohol addiction? Is there a gen-
etic test for the proper treatment for a patient with
schizophrenia? What new drugs may we expect for
Alzheimer’s disease?
Implicit in the foregoing discussion is a critical
message for every reader of this book: genetic abnor-
malities are not immutable; they are treatable. Each
genome represents a program for the body and the
brain, but it is not destiny. In fact genetic programs
are altered in their expression by the food we eat, the
medicine we take, and by everyday experience. The
finished product of a human life is a massively com-
plex combination of the genetic program (which in
itself does not change, except for mutation) and the
effects of our experience, beginning with the intra-
uterine environment. Given this complexity, it is
remarkable that genetic signals are even detectable
in behavioral disorders, and yet they are. But one
should never expect them to be constant, or
unchangeable.
Principles of Psychiatric Genetics should be of
interest to every mental health professional in
training. All those who wish to become psychiatrists,
psychologists, and counselors during the twenty-first
century should know about the field that we describe,
because it will affect your practice profoundly. Do we
have a blood test for bipolar disorder? No, not today
(despite what you may read on the internet). But such
tests are not far in the future. When they arrive, they
will likely be based on arrays that examine multiple
gene variants and biological pathways in a single test.
Those of you beginning practice today will likely send
your patients for such tests. They will be designed
based on some of the Principles in this book. The
other questions above may be answered similarly.
There are clues to the answers in the appropriate
chapters in this book.
Because the genetic and epigenetic variants
responsible for these disorders are not fully available
today, we are titling this book with the term Prin-
ciples. The details of this expanding field will change
daily over the coming years. The pace of discovery in
the laboratory is daunting, and to remain current
requires monitoring several hundred journals in print
and/or online (consult the bibliographies in this book
for examples). However, the core areas of psychiatry
xii
and human genetics are somewhat more constant.
You will notice that each chapter on a disease refers
to epidemiology, twin and family studies, and linkage
and association. Some disease-centered chapters also
refer to epigenetic studies, bioinformatic studies, and
drug development. There are special chapters in the
first section of the book on each of these methodo-
logical areas. These have been prepared by subspeci-
alty experts in psychiatric genetics and will be useful
in interpreting the disease-centered chapters and also
the journal literature in these areas.
Each disease chapter utilizes the methods
described and then provides an up-to-date summary
of where we stand now in identifying specific genetic
influences on that trait or traits. Disease chapters also
generally provide an overview of the symptoms, signs,
and life course of the condition described. We have
concentrated on those conditions usually treated by
psychiatrists and other mental health professionals,
but we are aware of the similarities in origin and
course of other neuropsychiatric conditions that
may be more frequently seen by pediatricians or
neurologists (i.e. tuberous sclerosis, seizure disorders,
or vascular dementia). For a more general reference
on medical genetics in clinical practice, we would
refer the reader to the online Mendelian Inheritance
in Man by McKusick and collaborators.
As editors of this volume, we are humbled by the
contributions of those who have gone before us in
defining the field of Psychiatric Genetics. We would
like to thank our mentor and longtime collaborator
Elliot Gershon, who decades ago taught us the
Principles in this volume. Others who influenced our
approach to the issues in this book include Theodore
Reich, Seymour Kety, George Winokur, Irving
Gottesmann, Ming Tsuang, and Robert Cloninger,
among others. We thank our contributing authors
for their insight, their industry, their patience, and
their trust that our shared effort would result in a
book that they and others would admire. We appreci-
ate the indulgence of our colleagues and families with
the time this task took away from other activities.
In this regard we thank Patricia Nurnberger and
Christine Berrettini most of all. Finally we thank the
patients and families who continue to teach us in our
clinics every day. By this book, may we provide a
stepping stone to better and more productive lives
for you . . .
 Contribution of genetic epidemiology to
Chapter
1 our understanding of psychiatric disorders
Kathleen Ries Merikangas and Anibal Cravchik
Introduction
This chapter will: (1) provide a summary of the back-
ground disciplines and approaches to understanding
the role of genetic factors in mental disorders;
(2) review the current knowledge in the genetic epi-
demiology of mental disorders; and (3) summarize
the role of epidemiology in the current generation of
genome-wide association studies of mental disorders.
Genetic epidemiology
The pioneering work of Böök, Sjögren, Angst, Perris,
and others in Europe and Kallman, Heston,
Rosenthal, Wender, and Kety in the United States
firmly established the important role of genetic sus-
ceptibility factors in psychiatric disorders. Heston’s
[1] original finding that adult offspring of hospital-
ized schizophrenic mothers had significantly higher
rates of schizophrenia than offspring of parents with
no mental illness was confirmed and extended by Kety,
Rosenthal, and Wender’s [2, 3] studies in a much larger
sample of adopted away offspring of schizophrenics in
Denmark. These studies demonstrated clearly that the
presence of schizophrenia in birth parents, independ-
ent of the rearing environment, significantly increases
offspring’s risk for the development of the disease
[4, 5]. During the last decades of the twentieth century,
the study of genes in psychiatric disorders expanded
beyond hospital settings to outpatient treatment set-
tings, particularly in the United States [6]. With the
introduction of epidemiology to the study of psychiatry,
systematic control groups were included in family
studies and methods for incorporating population
base rates and risk assessment were developed.
The field of genetic epidemiology is defined as the
study of the distribution of, and risk factors for
Principles of Psychiatric Genetics, eds John I. Nurnberger, Jr. and Wade H. Berrettini. Published by Cambridge University
Press. © Cambridge University Press 2012.
diseases and genetic and environmental causes of
familial resemblance. Genetic epidemiology focuses
on how genetic factors and their interactions with
other risk factors increase vulnerability to, or protec-
tion against, disease [7]. Genetic epidemiology
employs traditional epidemiological study designs to
explain aggregation in groups as closely related as
twins or as loosely related as migrant cohorts. Epi-
demiology has developed sophisticated designs and
analytic methods for identifying disease risk factors.
With increasing progress in gene identification, these
methods have been extended to include both genetic
and environmental factors [8, 9]. In general, study
designs in genetic epidemiology either control for
genetic background while letting the environment
vary (e.g. migrant studies, half siblings, separated
twins) or control for the environment while allowing
variance in the genetic background (e.g. siblings,
twins, adoptees nonbiological siblings). Investigations
in genetic epidemiology are typically based on a com-
bination of study designs including family, twin, and
adoption studies.
Family studies
Familial aggregation is generally the first source
of evidence that genetic factors may play a role in a
disorder. The most common indicator of familial
aggregation is the relative risk ratio, computed as
the rate of a disorder in families of affected persons
divided by the corresponding rate in families of con-
trols. The patterns of genetic factors underlying a
disorder can be inferred from the extent to which
patterns of familial resemblance adhere to the expect-
ations of Mendelian laws of inheritance. The degree of
genetic relatedness among relatives is based on the
proportion of shared genes between a particular
1
 Chapter 1: Contribution of genetic epidemiology
relative and an index family member or proband.
First-degree relatives share 50% of their genes in
common; second-degree relatives share 25% of their
genes in common, and third-degree relatives share
12.5% of their genes in common. If familial resem-
blance is wholly attributable to genes, there should be
a 50% decrease in disease risk with each successive
increase in degree of relatedness, from first to second
to third, and so forth. This information can be used to
derive estimates of familial recurrence risk within and
across generations as a function of population preva-
lence (l) [10]. Whereas l tends to exceed 20 for most
autosomal dominant diseases, values of l derived
from family studies of many complex disorders tend
to range from 2 to 5. Diseases with strong genetic
contributions tend to be characterized by 50%
decrease in risk across successive generations.
Decrease in risk according to the degree of genetic
relatedness can also be examined to detect inter-
actions between several loci. If the risk to second-
and third-degree relatives decreases by more than
50% this implies that more than a single locus must
contribute to disease risk and that no single locus can
largely predominate.
The major advantage of studying diseases within
families is that disease manifestations are more
likely to result within families than they are across
families from the same underlying etiological
factors. Family studies are therefore more effective
than between family designs in examining the val-
idity of diagnostic categories because they more
accurately assess the specificity of transmission of
symptom patterns and disorders. Data from family
studies can also provide evidence regarding etio-
logical or phenotypic heterogeneity. Phenotypic het-
erogeneity is suggested by variable expressivity of
symptoms of the same underlying risk factors,
whereas etiological heterogeneity is demonstrated
by common manifestations of expression of differ-
ent etiological factors between families. Moreover,
the family study method permits assessment of
associations between disorders by evaluating spe-
cific patterns of co-segregation of two or more
disorders within families [11].
Twin studies
Twin studies that compare concordance rates for
monozygotic twins (who share the same genotype)
with those of dizygotic twins (who share an average
2
of 50% of their genes) provide estimates of the degree
to which genetic factors contribute to the etiology
of a disease phenotype. A crude estimate of the gen-
etic contribution to risk for a disorder is calculated
by doubling the difference between the concordance
rates for monozygous and dizygous twin pairs.
Modern genetic studies employ path analytic models
to estimate the proportion of variance attributable to
additive genes, common environment, and unique
environment. There are several other applications of
the twin study design that may inform our under-
standing of the roles of genetic and environmental
risk factors for disease. First, twin studies provide
information on the genetic and environmental
sources of sex differences in a disease. Second, envir-
onmental exposures may be identified through com-
parison of discordant monozygotic twins. Third, twin
studies can be used to identify the genetic mode of
transmission of a disease by inspection of the degree of
adherence of the difference in risk between monozygotic
and dizygotic twins to the Mendelian ratio of 50%.
Fourth, twin studies may contribute to enhancing the
validity of a disease through inspection of the com-
ponents of the phenotypes that are most heritable.
The twin family design is one of the most powerful
study designs in genetic epidemiology because it
yields estimates of heritability but also permits evalu-
ation of multigenerational patterns of expression of
genetic and environmental risk factors. Several recent
updates of findings of twin studies of psychiatric
disorders are available [12, 13].
Adoption studies
Adoption studies have been the major source of evi-
dence regarding the joint contribution of genetic and
environmental factors to disease etiology. Adoption
studies either compare the similarity between an
adoptee and his or her biological versus adoptive
relatives, or the similarity between biological relatives
of affected adoptees with those of unaffected, or con-
trol adoptees. The latter approach is more powerful
because it eliminates the potentially confounding
effect of environmental factors. Similar to the familial
recurrence risk, the genetic contribution in adoption
studies is estimated by comparing the risk of disease
to biological versus adoptive relatives, or the risk of
disease in biological relatives of affected versus con-
trol adoptees. These estimates of risk are often
adjusted for sex, age, ethnicity, and other factors that

may confound the links between adoption status and
an index disease.
With the recent trends towards selective adoption
and the diminishing frequency of adoptions in the
United States, adoption studies are becoming less
feasible methods for identifying genetic and environ-
mental sources of disease etiology [14]. However, the
increased rate of reconstituted families (families com-
prised of both siblings and half siblings) may offer a
new way to evaluate the role of genetic factors in the
transmission of complex disorders. Genetic models
predict that half siblings should have a 50% reduction
in disease risk compared to that of full siblings. Devi-
ations from this risk provide evidence for either poly-
genic transmission, gene–environment interaction,
or other complex modes of transmission.
Migration studies
Migrant studies are perhaps the most powerful study
design to identify environmental and cultural risk
factors. When used to study Asian immigrants to
the United States, this study design demonstrated the
significant contribution of the environment to the
development of many forms of cancer and heart dis-
ease [15]. One of the earliest controlled migrant
studies evaluated rates of psychosis among Norwe-
gian immigrants to Minnesota compared to native
Minnesotans and native Norwegians [16]. A higher
rate of psychosis was found among the immigrants
compared to both the native Minnesotans and Nor-
wegians and was attributed to increased susceptibility
to psychosis among the migrants who left Norway.
It was found that migration selection bias was the
major explanatory factor, rather than environmental
exposure in the new culture. The application of
migration studies to the identification of environmen-
tal factors is only valid if potential bias attributed to
selection is considered. Selection bias has been tested
through comparisons of factors that may influence a
particular disease of interest in a migrant sample and
a similar sample that did not migrate.
Genetic epidemiology of psychiatric
disorders
The wealth of data from family, twin, and adoption
studies of the major psychiatric disorders exceeds that
of all other chronic human diseases. The increased
Chapter 1: Contribution of genetic epidemiology
Table 1.1 Risk ratios and heritability estimates for major
mental disorders.
Disorder Risk Heritability
ratios estimates
Mood disorders
Bipolar disorder 7–10 60–70
Major depression 2–3 28–40
Anxiety
All 4–6 30–40
Panic disorder 3–8 50–60
Schizophrenia 8–10 80–84
Substance 4–8 30–50
dependence
recognition of the role of biological and genetic vul-
nerability factors for psychiatric disorders has led
to research with increasing methodological sophisti-
cation over the course of the second half of the
twentieth century. There are numerous comprehen-
sive reviews of genetic research on specific disorders
of interest as well as on psychiatric genetics in general
[12, 17–20].
Table 1.1 presents a summary of the relative risks
(i.e. proportion affected among first-degree relatives
of affected probands versus those of relatives of
controls) derived from controlled family studies of
selected psychiatric disorders. The risk ratios compar-
ing the proportion of affected relatives of cases versus
controls are greatest for autism, bipolar disorder, and
schizophrenia; intermediate for substance depend-
ence and subtypes of anxiety, particularly panic; and
lowest for major depression. The estimates of herit-
ability (i.e. the proportion of variance attributable to
genetic factors) are derived from twin studies, which
compare rates of disorders in monozygotic and dizy-
gotic twins. These findings reinforce the notion that
genes play a major role in the extent to which mental
disorders run in families. The heritability estimates for
specific disorders shown in Table 1.1 are parallel to the
risk ratios derived from family studies. Furthermore,
adoption and half-sibling studies also support a genetic
basis for the observed familial aggregation.
Schizophrenia
More is known about the genetic basis of schizophre-
nia than perhaps any other psychiatric disorder,
with genetically informative studies stemming from
3
 Chapter 1: Contribution of genetic epidemiology
early in the last century. There are numerous reviews
of this extensive body of research [21–24]. Despite
wide differences in methods, samples, and geographic
locations, controlled family studies yield a remarkably
similar average relative risk of 8.9 to first-degree rela-
tives. The four-fold greater proband-wise concordance
rate of schizophrenia in monozygotic compared to
dizygotic twins, found in 12 studies to date, demon-
strates the role of genetic factors in the familial aggre-
gation of schizophrenia. The average heritability in
liability to schizophrenia across 12 studies is 0.81 [25].
Similarly, adoption studies using traditional paradigms
and modern diagnostic criteria (if available) demon-
strates that the average risk to first-degree relatives was
15.5% compared to 3.6% for controls, yielding a relative
risk of 4.3.
Despite evidence regarding the importance of gen-
etic risk factors for schizophrenia, the lack of expected
Mendelian risk ratios in the difference in risk of schizo-
phrenia as a function of genetic similarity suggests that
schizophrenia is a genetically complex disorder [10].
Recent reviews of the genetic epidemiology of schizo-
phrenia also converge in demonstrating the multifac-
torial etiology of this condition [25–29]. The largest
and most recent cross-fostering study of schizophrenia
showed that adoptive family environment was associ-
ated with schizophrenia spectrum disorders among
genetically vulnerable individuals [30], implying the
contributions of nonspecific environmental factors
(i.e. multiple factors that may affect brain develop-
ment) to schizophrenia’s etiology.
Another important clue about potential environ-
mental risk factors is the increased risk for the devel-
opment of schizophrenia among immigrants in
several different countries including East African
immigrants to Sweden [31], Surinamese immigrants
to the Netherlands [32], Afro-Caribbean immigrants
to the UK [33], Finnish immigrants to Sweden [34],
and European immigrants to Canada [35]. Although
selective migration may be one explanation, there is
converging evidence that socially disrupted environ-
ments may trigger the onset of schizophrenia in sus-
ceptible individuals.
Children at high risk for schizophrenia (children
with an affected parent) show an increased incidence
of numerous neurodevelopmental abnormalities as
compared to offspring of parents without schizophre-
nia [36, 37]. This discrepancy has led to a focus on
early developmental factors in the etiology of schizo-
phrenia. Several recent studies have focused on
4
genomic copy number variants (CNVs) potentially
affecting the expression or function of genes that are
relevant to brain development [38]. Of particular
interest is the velo-cardio-facial syndrome caused by
a deletion CNV in chromosome 22q, which confers a
25% risk for schizophrenia [39]. Some of the specific
environmental risk factors currently under investiga-
tion include obstetric complications [40], childhood
trauma [41], prenatal factors such as nutritional defi-
ciencies [42], increased paternal age [43], family inter-
actions [28], maternal infections [44], maternal
cytokines [45], gluten sensitivity [46], and cannabis use
[47, 48]. In summary, schizophrenia is now widely
viewed as a neurodevelopmental disorder comprised of
a confluence of vulnerability genes and environmental
exposures [49].
Mood disorders
A heterogeneous group of syndromes, of which major
depression and bipolar disorder (manic depression)
are major subtypes, comprise mood disorders. Bipo-
lar disorder is one of the psychiatric disorders most
widely studied from a genetic perspective [50, 51].
Both major depression and bipolar disorder have
important genetic components. Controlled family
studies show a five-fold risk to relatives of major
depression, and greater than a ten-fold risk to first-
degree relatives of bipolar patients for developing
these disorders. The concordance rate for bipolar
monozygotic twins is over five times that of dizygotic
twins, and twin concordance for depression shows
less dramatic but still notable differences. A summary
of five methodologically comparable twin studies of
major depression yielded an average estimate of the
heritability of major depression of 0.37, with the
remainder (0.63) nearly totally attributable to envir-
onmental factors unique to the individual [50]. The
relative risks based on the few existing adoption stud-
ies also confirm that the familial recurrence cannot
be attributed solely to environmental factors [51].
The aggregate adoption study data on mood dis-
orders reveal a moderate increase in rates of mood
disorders among the biological compared to adoptive
relatives of adoptees with mood disorders [52]. With
respect to bipolar disorder, there is little evidence for
differential risk among biological compared to adop-
tive relatives of adoptees with bipolar disorder. How-
ever, the small numbers of bipolar adoptees who have
been studied (i.e. less than 50) do not provide an

adequate test of genetic and environmental influences.
The most compelling finding from adoption studies,
however, is the dramatic increase in completed suicide
among biologic, as opposed to adoptive, relatives of
mood disorder probands [2, 53].
Anxiety disorders
At present, relatively few studies have examined
anxiety disorders from the perspective of genetic epi-
demiology, and there is virtually no data from certain
paradigms, such as adoption studies [54, 55]. How-
ever, the existing research indicates that most anxiety
disorders aggregate in families and several investiga-
tions have offered specific support for genetic etiology.
Panic disorder
Panic disorder has the strongest degree of familial
aggregation of any of the anxiety disorder subtypes.
A review of 13 family studies of panic disorder by
Gorwood [56] shows a seven-fold relative risk of
panic among relatives of panic probands compared
to controls. In addition, early-onset panic, panic asso-
ciated with childhood separation anxiety, and panic
associated with respiratory symptoms have each been
shown to have a higher familial loading than other
varieties of panic disorder [57]. Although there has
been some inconsistency reported among twin studies
of panic disorder, recent studies using contemporary
diagnostic criteria show that panic disorder has the
highest heritability of all anxiety disorders (44%) [58].
Phobic states
Although there are far fewer controlled family and
twin studies of the anxiety subtypes other than panic
disorder, all of the phobic states (i.e. specific phobia,
agoraphobia) have also been shown to be familial [59,
60]. The average relative risk of phobic disorders in
the relatives of phobics is 3.1, with greater familial
aggregation for the generalized subtype of social
phobia. The heritability of phobias according to twin
studies is about 0.35% [61].
Generalized anxiety disorder
A limited number of studies also provide evidence for
both the familial aggregation and heritability of gener-
alized anxiety disorder [62]. The average familial odds
ratio for the disorder is approximately 5 [60, 63] and
the heritability is 0.32 among female twin pairs [64].
Chapter 1: Contribution of genetic epidemiology
Obsessive–compulsive disorder
Controlled family studies of obsessive–compulsive
disorder reveal an elevated familial risk in probands
with obsessive–compulsive disorder compared to con-
trols, with greater familial aggregation associated with
early age of onset and obsessions rather than compul-
sions [65–67]. Twin studies of obsessive–compulsive
disorder, however, have yielded only weak evidence for
heritability [54, 68, 69].
Substance use disorders
A positive family history of a substance use disorder
is a consistent and robust risk factor for substance
use in first-degree relatives (for comprehensive
reviews of alcoholism see [70–73]). Controlled family
studies of alcohol use disorder reveal a three-fold
increased risk of alcoholism and two-fold increased
risk of drug abuse among the relatives of probands
with alcoholism as compared to those of controls
[74]. Both alcohol abuse and dependence appear to
be familial among females, whereas only dependence
aggregates among the relatives of males with alcohol
dependence [75].
Twin studies have demonstrated the contribution
of both genetic and environmental risk factors to both
alcoholism and drug abuse [76]. Heritability of alco-
holism (narrowly defined) has been estimated at
59% by some researchers [77], while the heritability
of problem drinking (using broad definitions) has
been estimated at 8–44% in females and 10–50% in
males [70]. Several adoption studies conducted in
Scandinavia demonstrated the importance of genetic
factors underlying alcoholism [78–80]. Adoption
study paradigms have shown not only that a disturbed
adoptive family environment interacts with a genetic
predisposition for alcoholism to increase the risk for
the disorder [81], but also that the adoptive family
environment can predict alcohol abuse or dependence
independent of genetic vulnerability [82]. A recent
“quasi-adoption” study that investigated the associ-
ation between the biological family background (gene-
tic factors), and a history of exposure to alcoholism
during childhood (environmental factors) revealed
greater effect of genetic risk factors among men than
among women. The study also showed common gen-
etic and environmental risk factors contributing to
alcohol dependence in both men and women [83].
The importance of the environment as a mediating
factor in the transmission of substance use disorders
5
 Chapter 1: Contribution of genetic epidemiology
was demonstrated in a recent study of adoptive and
step families [84].
Although there has been less systematic research
on the familial aggregation of drug use disorders,
numerous family history studies and uncontrolled
and controlled family studies have demonstrated that
rates of substance use disorders are elevated among
relatives of drug abusers compared to those of con-
trols and compared to population expectations [85,
86]. One controlled family study of drug use disorders
using contemporary family study data [75] showed an
eight-fold increased risk of substance use disorders
(opioids, cocaine, cannabis, and alcohol) among rela-
tives of probands with drug disorders compared with
relatives of people with psychiatric disorders and
normal controls. Family, twin, and adoption studies
have also demonstrated common genetic and envir-
onmental factors that contribute to cannabis use dis-
orders and other drug use disorders [87]. The results
of numerous twin studies of substance use disorders
in general as well as those of specific drugs have
shown that there are genetic factors underlying drug
abuse in general [88], as well as unique genetic factors
associated with specific drugs of abuse including nico-
tine and cannabis [72, 77, 89, 90].
Sources of complexity in mental
disorders
Two factors which contribute to the complexity of the
patterns of inheritance of psychiatric disorders are
the lack of validity of the classification of psychiatric
disorders (e.g. phenotypes, or observable aspects of
diseases) and the complexity of the pathways from geno-
types to psychiatric phenotypes (i.e. heterogeneity).
Lack of validity of the classification system
The development of structured interviews has enhanced
comparability of diagnostic methods within the United
States and worldwide. There is now an exciting venture
designed to collect information on the prevalence of
mental disorders using comparable diagnostic tools in
more than 30 countries under the auspices of the World
Health Organization and Harvard University [91]. The
lack of conclusive evidence for the validity of classifica-
tions of psychiatric disorder phenotypes, because they
are based solely on clinical manifestations without path-
ognomonic markers, continues to impede advances in
psychiatry [92, 93]. Growing research on the
6
dimensional classification of disorders further demon-
strates the difficulties in creating a valid classification
system for psychiatric disorders because of the preva-
lence of subthreshold diagnostic categories and diag-
nostic spectra, and the pervasive comorbidity between
purportedly distinct diagnostic entities; there is wide-
spread agreement that the categorical classification
system in psychiatry lack validity [27, 94, 95].
The greater complexity of psychiatric disease, as
compared to other types of disease explains the con-
tinued reliance on the descriptive approach as the sole
basis for diagnosis in psychiatry. The difficulty in
classifying human cognition, behavior, and emotion
is understandable in light of the complex psychological
and physiological states underlying mental function,
which is the culmination of human adaptation to the
environment up to the current point in time. Progress
in neuroscience that reveals information about the
biological pathways underlying psychiatric disorders
should also advance our understanding of the classifi-
cation of psychiatric phenotypes.
Complex patterns of transmission
The application of advances in genomics to mental
disorders is still limited by the complexity of the
process through which genes influence the develop-
ment and progression of mental disorders. There is
substantial evidence that a lack of one-to-one corres-
pondence between the genotype and phenotype
exists for most of the major psychiatric disorders.
Psychiatric disorders, like numerous other complex
disorders for which susceptibility alleles have been
identified, are characterized by phenomena such as
incomplete penetrance (i.e. probability of phenotypic
expression among individuals with susceptibility
gene), variable expressivity (i.e. variation in clinical
expression associated with a particular gene), gene–
environment interaction (i.e. expression of genotype
only in the presence of particular environmental
exposures), pleiotropy (i.e. capacity of genes to mani-
fest several different phenotypes simultaneously),
genetic heterogeneity (i.e. different genes leading to
indistinguishable phenotypes), gene–environment
correlation [96] and polygenic and oligogenic modes
of inheritance (i.e. simultaneous contributions of mul-
tiple genes rather than Mendelian single gene models)
[10, 97]. Other proposed mechanisms of transmission
include mitochondrial inheritance, imprinting, and
epigenetic phenomena [98].

Comorbidity
The high magnitude of comorbidity and co-aggrega-
tion of index disorders with other major psychiatric
disorders (i.e. bipolar disorder and alcoholism, major
depression and anxiety disorders, schizophrenia and
drug dependence), in part induced by the classification
system, has been demonstrated in both clinical and
community studies [11, 86, 99, 100]. For example,
alcoholism, a well-established complication of bipolar
illness, may mask the underlying features of bipolarity,
leading to phenotypic misclassification in genetic
studies [101]. Nonrandom mating is also a common
phenomenon in mental disorders that impedes evalu-
ation of patterns of familial transmission [102].
Assortative mating is particularly pronounced for
substance use disorders for which substance depend-
ence among spouses of substance dependent probands
may be as high as 90% [103].
These phenomena serve to decrease the signal to
noise ratio in defining psychiatric disorders for gen-
etic studies. Studies that attempt to identify the
impact of these phenomena on phenotypic expression
in individuals and families will bring us closer to
understanding the role of the underlying genes on
the components of psychiatric disorders.
Applications of genetic epidemiology
to gene identification
There is a widespread consensus among geneticists
and epidemiologists on the importance of epidemi-
ology to the future of genetics and on the conclusion
that the best strategy for susceptibility risk factor
identification for common and complex disorders will
ultimately involve large epidemiological studies from
diverse populations [73, 104–108]. It is likely that
population-based association studies will assume
increasing importance in translating the products of
genomics to public health. There are several reasons
that population-based studies are critical to current
studies seeking to identify genes underlying psychi-
atric disorders. First, the frequency of newly identified
polymorphisms, whether single nucleotide poly-
morphisms (SNPs) or other variants such as copy
number variation (CNVs), especially in particular
population subgroups, is not known. Second, current
knowledge of genes as risk factors is based nearly
exclusively on clinical and nonsystematic samples.
Hence, the significance of the susceptibility alleles that
Chapter 1: Contribution of genetic epidemiology
have been identified for cancer, heart disease, dia-
betes, and other common disorders is unknown in
the population at large. In order to provide accurate
risk estimates, the next stage of research needs to
move beyond samples identified through affected
individuals to the population as a whole. Third, iden-
tification of risk profiles will require large samples to
assess the significance of vulnerability genes with
relatively low expected population frequencies. Fourth,
similar to the role of epidemiology in quantifying risk
associated with traditional disease risk factors, applica-
tions of human genome epidemiology can provide
information on the specificity, sensitivity, and impact
of genetic tests to inform science and individuals [105].
Below we review the role of the tools of epidemiology
in ongoing and future studies designed to identify
genes underlying mental disorders.
Samples
The shift from systematic large-scale family studies
to linkage studies in psychiatry has led to the collec-
tion of families according to very specific sampling
strategies (e.g. many affected relatives, affected sib-
ling pairs, affected relatives on one side of the family
only, availability of parents for study, etc.) in order
to maximize the power of detecting genes according
to the assumed model of familial transmission. Des-
pite the increase in power for detecting genes, these
sampling approaches have diminished the general-
izability of the study findings, and contribute little
else to the knowledge base if genes are not dis-
covered. Recent genome-wide association studies of
psychiatric disorders have included probands from
families previously collected for linkage studies and
single cases collected more recently from hospital
admissions, almost all of self-reported European
descent. Future studies will attempt to collect both
families and controls from representative samples of
the population so that results can be used to esti-
mate population risk parameters, to examine the
specificity of endophenotypic transmission and so
results can be generalized to whole populations.
Selection of controls
The most serious problem in the design of association
studies is the difficulty of selecting controls that are
comparable to the cases on all factors except the
disease of interest [109, 110]. Controls should be
7
 Chapter 1: Contribution of genetic epidemiology
drawn from the same population as cases, and must
have the same probability of exposure (i.e. genes)
as cases. Controls should be selected to ensure the
validity rather than the representativeness of a study.
Failure to equate cases and controls may lead to
confounding (i.e. a spurious association due to an
unmeasured factor that is associated with both the
candidate gene and the disease). In genetic case-
control studies, the most likely source of confounding
is ethnicity because of differential gene and disease
frequencies in different ethnic subgroups. Recent
genome-wide association studies of psychiatric dis-
orders have included control samples recruited from
the general population using self-administered psy-
chiatric screens and from blood bank samples that
exclude donors reporting major psychiatric diagnoses
or taking psychiatric medications. The matching of
controls to cases on ethnic background is largely
based on self-report; several methods are used to
screen for and exclude subjects with substantial
differences in ancestry.
Risk estimation
Because genetic polymorphisms involved in complex
diseases are likely to be nondeterministic (i.e. the
marker neither predicts disease nor nondisease with
certainty), traditional epidemiological risk factor
designs can be used to estimate the impact of these
genetic polymorphisms. Increased attention to alleles
as a part of risk equations in epidemiology will likely
resolve the contradictory findings from studies that
have generally employed solely environmental risk
factors, such as diet, smoking, alcohol use, etc. Like-
wise, the studies that seek solely to identify small risk
alleles will continue to be inconsistent because they
do not consider the effects of nongenetic biological
parameters or environmental factors that contribute
to the diseases of interest.
There are several types of risk estimates that are
used in public health. The most common is relative
risk, defined as the magnitude of the association
between an exposure and disease. It is independent
of the prevalence of the exposure. The absolute risk
is the overall probability of developing a disease in
an individual or in a particular population [111].
The attributable risk is the difference in the risk of
the disease in those exposed to a particular risk
factor compared to the background risk of a disease
in a population (i.e. in the unexposed). The
8
population attributable risk relates to the risk of a
disease in a total population (exposed and unex-
posed) and indicates the amount the disease can
be reduced in a population if an exposure is elim-
inated. The population attributable risk depends on
the prevalence of the exposure, or in the case of risk
alleles, the allele frequency. Genetic attributable risk
would indicate the proportion of a particular dis-
ease that would be eliminated if a particular gene or
genes were not involved in the disease. For example,
the two vulnerability alleles for Alzheimer’s disease
include the very rare, but deterministic alleles in the
b-amyloid precursor, presenilin-1, and –2 genes,
which signal a very high probability of the develop-
ment of Alzheimer’s disease, particularly at a young
age, and the susceptibility allele ε4 in the apolipo-
protein-E gene (APOE ε4) [112]. The apolipopro-
tein-Ε ε4 (APOE ε4) allele has been shown to
increase the risk of Alzheimer’s disease in a dose-
dependent fashion. Using data from a large multi-
ethnic sample collected by more than 40 research
teams, Farrer [113] reported a 2.6–3.2 greater odds
of Alzheimer’s disease among those with one copy,
and 14.9 odds of Alzheimer’s disease among those
with two copies of the APOE ε4 allele. Moreover,
there was a significant protective effect among those
with the ε2/ε3 genotype. As opposed to the deter-
ministic mutations, the APOE ε4 allele has a very
high population attributable risk because of its high
frequency in the population. The APOE ε4 allele is
likely to interact with environmental risk
and protective factors [114, 115]. The population
risk attributable to these mutations is quite low
because of the very low population prevalence of
disease associated with these alleles. This model of
combination of several rare deterministic alleles in a
small subset of families and common alleles with
lower relative risk to individuals but high popula-
tion attributable risk is likely to apply to many of
the psychiatric disorders as well, and may in part
explain some of the discrepancies in findings across
studies to date.
Recent genome-wide association studies have
uncovered risk alleles associated with coronary artery
disease, Crohn’s disease, rheumatoid arthritis, type 1
and type 2 diabetes [116], and schizophrenia. Those
genetic variants appear to confer only modest increases
in disease risk (odd ratios [ORs] between 1.2 and 1.5)
compared with other established risk factors for
common chronic diseases.

Use of endophenotypes for classification
Numerous studies have begun to deconstruct psychi-
atric phenotypes by their component features or sub-
types including bipolar disorder [117, 118], general
anxiety disorder [119], obsessive–compulsive dis-
order [120], schizophrenia [121], and panic disorder
[122]. Identification of phenotypic traits or markers,
which are themselves heritable, and which may repre-
sent intermediate forms of expression between the
output of underlying genes and the broader disease
phenotype, have been termed “endophenotypes”
[123]. Studies of the role of genetic factors involved
in these systems may be more informative than stud-
ies of the aggregate psychiatric phenotypes since they
may more closely represent the expression of under-
lying biological systems. To the extent that particular
endophenotypes more clearly represent expression of
genotypes, they may help to unravel the complexity of
transmission of the mental disorders. For example,
some of the endophenotypes that may underlie mood
disorders include circadian rhythm, stress reactivity, and
mood, sleep and appetite regulation [95]. Cognitive,
neurophysiologic, and structural measures continue
to be tested as potential schizophrenia endopheno-
types [124, 125]. However, before applying endophe-
notypes in gene identification studies, there should
be evidence that the endophenotype has a stronger
genetic signal than the broader phenotype. A recent
meta-analysis of psychiatric endophenotypes [126]
and a review of the genetic architecture of traits in
model organisms do not provide evidence that endo-
phenotypes are superior to current phenotypic disease
definitions [127].
Identification of environmental factors
In parallel with the identification of susceptibility
alleles, it is important to identify environmental
factors that operate either specifically or nonspecifi-
cally on those with susceptibility to psychiatric dis-
orders in order to develop effective prevention and
intervention efforts. Langholz et al. [128] describes
some of the world’s prospective cohort studies that
may serve as a basis for studies of gene–disease asso-
ciations or gene–environment interactions. There
is increasing evidence that gene–environment inter-
action will underlie many of the complex human
diseases. Some examples include inborn errors of
metabolism, individual variation in response to drugs
Chapter 1: Contribution of genetic epidemiology
[129], substance use disorders [71, 130] and the pro-
tective influence of a deletion in the CCR5 gene
on exposure to the human immunodeficiency virus
(HIV) [131].
In prospective studies, however, few environmen-
tal exposures have been shown to have an etiological
role in psychiatric disorders [132]. Over the next
decades, it will be important to identify and evaluate
the effects of specific environmental factors on disease
outcomes and to refine measurement of environ-
mental exposures to evaluate specificity of effects.
Study designs and statistical methods should focus
increasingly on gene–environment interaction [106,
133, 134]. Although numerous recent studies have
reported gene–environment interaction between sev-
eral genes that interact with nonspecific environmen-
tal exposures such as life stress or childhood adversity
and a range of outcomes including depression, can-
nabis dependence, and conduct disorder [135], repli-
cation of these findings has not been consistent [136].
Increased knowledge of the developmental pathways
of emotion, cognition, and behavior will expand our
ability to identify specific environmental factors such
as infection, poor diet, prenatal environment, and
early life experiences that interact with the genetic
architecture of mood regulation and cognition [137].
Impact of genomics on psychiatric
science and practice
Progress in genomics has far outstripped advances
in our understanding of psychiatric disorders and
their etiologies. Technical advances and availability
of rapidly expanding genetic databases provide extra-
ordinary opportunities for understanding disease
pathogenesis. However, the application of psychiatric
genetic research to study diagnostic heterogeneity,
course and/or treatment outcome is still limited due
to the lack of consistent genetic findings to date. Over
the next decade increasing understanding of the com-
plex mechanisms through which genetic risk factors
influence disease should enhance the clinical utility of
psychiatric genetics.
The goal of genomics research is ultimately
prevention, the cornerstone of public health. An
understanding of the significance of genetic risk
factors and proper interpretations of their meaning
for patients and their families will ultimately become
part of clinical practice. Clinicians will become
9
 Chapter 1: Contribution of genetic epidemiology
increasingly involved in helping patients to compre-
hend the meaning and potential impact of genetic risk
for both psychiatric and nonpsychiatric disorders.
As our knowledge of the role of genetic risk factors
in psychiatric disorders advances, it will be incumbent
References
1. Heston LL. Br J Psychiatry 1966;
112(489):819–825.
2. Wender PH, et al. Arch Gen
Psychiatry 1986;43(10):923–929.
3. Wender PH. Annu Rev Med
1972;23:355–371.
4. Kety SS, et al. Schizophr Bull
1976;2(3):413–428.
5. Kety SS, et al. Genetic Relationships
within the Schizophrenia Spectrum:
Evidence from Adoption Studies.
New York: Raven Press; 1978.
6. Gershon ES, et al. (eds.). Genetic
Approaches to Mental Disorders.
New York: American
Psychopathological Association
Press; 1994.
7. Beaty TH, et al. Epidemiol Rev
2000;22(1):120–125.
8. MacMahon B, et al. Epidemiology:
Principles and Methods. Boston:
Little Brown and Company; 1996.
9. Kuller LH. Prog Clin Biol Res
1979;32:489–495.
10. Risch N. Am J Hum Genet 1990;
46(2):222–228.
11. Merikangas KR. Comorbity for
Anxiety and Depression: Review of
Family and Genetic Studies.
Washington, DC: American
Psychiatric Press; 1990.
12. Shih RA, et al. Int Rev Psychiatry
2004;16(4):260–283.
13. Kendler KS, et al. Psychol Med
2007;37(5):615–626.
14. National Adoption Information
Clearinghouse. Families 2007
Edition. Washington, DC: US
Department of Health and Social
Services; 2007.
15. Kolonel LN, et al. Nat Rev Cancer
2004;4(7):519–527.
10
upon clinicians to become familiar with knowledge
gleaned from genetic epidemiological and genomics
research. In the meanwhile, use of recurrence risk
estimates from family studies best predicts the risk
of the development of mental disorders.
16. Ödegaard Ö. Emigration and
Insanity: A Study of Mental
Disorders among the Norwegian
Born Population of Minnesota.
Copenhagen: Levin and
Munksgaards; 1932.
17. Merikangas K, et al. Epidemiol Rev
1996;19:1–12.
18. Kendler KS. Arch Gen Psychiatry
1995;52(11):895–899.
19. Tandon K, et al. Eur J Neurosci
2002;16(3):403–407.
20. Tsuang MT, et al. Methods Mol
Med 2003;77:251–265.
21. Zerbin-Rudin E. Nervenarzt
1980;51(7):379–391.
22. Gottesman, II. Acta Psychiatr
Scand Suppl 1994;384:26–33.
23. Kendler KS, et al. Schizophr Bull
1993;19(2):261–285.
24. Sullivan PF. PLoS Med 2005;
2(7):e212.
25. Sullivan PF, et al. Arch Gen
Psychiatry 2003;60(12):
1187–1192.
26. Hallmayer J. Aust N Z J Psychiatry
2000;34 Suppl:S47–55; discussion
S56–47.
27. Tsuang MT. Am J Med Genet
2001;105(1):8–10.
28. McGuffin P. Psychiatr
Prax 2004;31 Suppl 2:
S189–193.
29. Portin P, et al. Acta Psychiatr
Scand 1997;95(1):1–5.
30. Tienari P, et al. Br J Psychiatry
2004;184:216–222.
31. Selten JP, et al. Am J Psychiatry
2002;159(4):669–671.
32. Hanoeman M, et al. Schizophr Res
2002;54(3):219–221.
33. Cooper B. Epidemiol Psichiatr Soc
2005;14(3):137–144.
34. Leao TS, et al. J Nerv Ment Dis
2006;194(1):27–33.
35. Smith GN, et al. Schizophr Res
2006;87(1–3):205–211.
36. Owens DG, et al. Psychol Med
2006;36(11):1501–1514.
37. Brewer WJ, et al. Schizophr Bull
2006;32(3):538–555.
38. Cook EH, Jr., et al. Nature
2008;455(7215):919–923.
39. Bassett AS, et al. Hum Mol Genet
2008;17(24):4045–4053.
40. Clarke MC, et al. Schizophr Bull
2006;32(1):3–8.
41. Morgan C, et al. Schizophr Bull
2007;33(1):3–10.
42. Ludvigsson JF, et al. Scand
J Gastroenterol 2007;49:179.
43. Malaspina D, et al. CNS Spectr
2002;7(1):26–29.
44. Buka SL, et al. Arch Gen Psychiatry
2001;58(11):1032–1037.
45. Buka SL, et al. Brain Behav
Immun 2001;15(4):411–420.
46. Kalaydjian AE, et al. Acta Psychiatr
Scand 2006;113(2):82–90.
47. Arseneault L, et al. BMJ 2002;325
(7374):1212–1213.
48. Dean K, et al. Dialogues Clin
Neurosci 2005;7(1):69–80.
49. Dealberto MJ. Med Hypotheses
2007;68(2):259–267.
50. Sullivan PF, et al. Am J Psychiatry
2000;157(10):1552–1562.
51. Merikangas KR, et al. Clin
Neurosci Res 2002;2:127.
52. Faraone SV, et al. Psychol Bull
1990;108(1):109–127.
53. Mendlewicz J, et al. Nature
1977;268(5618):327–329.
54. Carey G, et al. In Rabkin J (ed.).
Anxiety: New Research and

Changing Concepts. Vol. 117.
New York: Raven Press; 1981.
55. Hettema JM, et al. Am J Psychiatry
2001;158(10):1568–1578.
56. Gorwood P, et al. Encephale
1999;25(1):21–29.
57. Goldstein RB, et al. Arch
Gen Psychiatry 1997;54(3):
271–278.
58. Kendler KS, et al. Psychol Med
2001;31(6):989–1000.
59. Fyer AJ, et al. Arch Gen Psychiatry
1995;52(7):564–573.
60. Noyes R, Jr., et al. Am J Psychiatry
1987;144(8):1019–1024.
61. Kendler KS, et al. Arch Gen
Psychiatry 1992;49(4):273–281.
62. Newman SC, et al. Psychol Med
2006;36(9):1275–1281.
63. Mendlewicz J, et al. Psych Genet
1993;3(2):73–78.
64. Kendler KS, et al. Arch Gen
Psychiatry 1992;49(4):267–272.
65. Nestadt G, et al. Arch Gen
Psychiatry 2000;57(4):358–363.
66. Grabe HJ, et al. Am J Psychiatry
2006;163(11):1986–1992.
67. Pauls DL, et al. Am J Psychiatry
1995;152(1):76–84.
68. Bellodi L, et al. Psychiatry Res
1992;42(2):111–120.
69. Lenane MC, et al. J Am Acad Child
Adolesc Psychiatry 1990;29
(3):407–412.
70. Heath AC, et al. Psychol Med
1997;27(6):1381–1396.
71. Heath AC, et al. Br J Psychiatry
Suppl 2001;40:S33–S40.
72. Tsuang MT, et al. Harv Rev
Psychiatry 2001;9(6):267–279.
73. Merikangas KR. In D’haenen
HAH, et al. Biological Psychiatry.
Vol. 2. Chichester: John Wiley and
Sons; 2002.
74. Nurnberger JI, Jr., et al. Arch Gen
Psychiatry 2004;61(12):1246–1256.
75. Merikangas KR, et al.
Arch Gen Psychiatry
1998;55(11):973–979.
Chapter 1: Contribution of genetic epidemiology
76. Kendler KS, et al. Am J Psychiatry
1994;151(5):707–715.
77. Kendler KS, et al. Arch Gen
Psychiatry 2000;57(3):261–269.
78. Cloninger CR, et al. Recent Dev
Alcohol 1985;3:37–51.
79. Sigvardsson S, et al. Arch Gen
Psychiatry 1996;53(8):681–687.
80. Goodwin DW. Arch Gen
Psychiatry 1985;42(2):
171–174.
81. Cutrona CE, et al. Compr
Psychiatry 1994;35(3):171–179.
82. Cadoret RJ, et al. Arch Gen
Psychiatry 1995;52(1):42–52.
83. Light JM, et al. J Stud Alcohol
1996;57(5):507–520.
84. Newlin DB, et al. Alcohol Clin Exp
Res 2000;24(12):1785–1794.
85. Bierut LJ, et al. Arch Gen
Psychiatry 1998;55(11):982–988.
86. Merikangas KR, et al. Psychol Med
1998;28(4):773–788.
87. Agrawal A, et al. Addiction
2006;101(6):801–812.
88. Kendler KS, et al. Acta Psychiatr
Scand 1999;99(5):368–376.
89. Kendler KS, et al. Am J Psychiatry
1998;155(8):1016–1022.
90. Kendler KS, et al. Br J Psychiatry
1998;173:345–350.
91. Kessler R, et al. Hosp Manage
Internat 2000:195–196.
92. Kendell RE. Psychol Med 1989;
19(1):45–55.
93. Angst J. World Psychiatry 2007;
6(2):94–95.
94. Angst J, et al. J Affect Disord
2003;73(1–2):133–146.
95. Lenox RH, et al. Am J Med Genet
2002;114(4):391–406.
96. Dick DM, et al. Ann Clin
Psychiatry 2006;18(4):223–231.
97. Gottesman II, et al. Hum Hered
1971;21(6):517–522.
98. Guttmacher AE, et al. N Engl
J Med 2002;347(19):1512–1520.
99. Maier W, et al. Schizophr Res
2002;57(2–3):259–266.
100. Maier W, et al. Eur Arch
Psychiatry Clin Neurosci 1993;
243(3–4):205–211.
101. Merikangas KR, et al.
Arch Gen Psychiatry
2007;64(5):543–552.
102. Merikangas KR. Arch Gen
Psychiatry 1982;39(10):1173–1180.
103. Galbaud du Fort G, et al. Psychol
Med 1998;28(4):789–802.
104. Peltonen L, et al. Science 2001;
291(5507):1224–1229.
105. Khoury MJ, et al. Am J Epidemiol
2000;151(1):2–3.
106. Yang Q, et al. Epidemiol Rev
1997;19(1):33–43.
107. Merikangas KR, et al. Science
2003;302(5645):599–601.
108. Risch N. Genet Epidemiol
1990;7(1):3–16; discussion
17–45.
109. Wacholder S, et al. J Natl Cancer
Inst 2000;92(14):1151–1158.
110. Ott J. Neurology 2004;63(6):
955–958.
111. Gordis L. Epidemiology.
Philadelphia: W B Saunders;
2000.
112. Tol J, et al. Rev Neurol (Paris)
1999;155(Suppl 4):S10–S16.
113. Farrer LA, et al. JAMA 1997;
278(16):1349–1356.
114. Kivipelto M, et al. Neurology
2001;56(12):1683–1689.
115. Kivipelto M, et al. Ann Intern
Med. 2002;137(3):149–155.
116. Wellcome Trust Case Control
Consortium. Nature 2007;
447(7145):661–678.
117. Benazzi F. Lancet 2007;
369(9565):935–945.
118. Angst J. Br J Psychiatry
1997;190:189.
119. Angst J, et al. Psychol Med 2006;
36(9):1283–1292.
120. Eapen V, et al. J Psychosom Res
2006;61(3):359–364.
121. Braff DL, et al. Schizophr Bull
2007;33(1):21–32.
11
 Chapter 1: Contribution of genetic epidemiology
122. Smoller JW, et al. Am J Psychiatry
1998;155(9):1152–1162.
123. Gottesman II, et al. Am
J Psychiatry 2003;160(4):636–645.
124. Horan WP, et al. Schizophr Res
2008;103(1–3):218–228.
125. Radant AD, et al. Schizophr Res
2007;89(1–3):320–329.
126. Flint J, et al. Psychol Med 2007;
37(2):163–180.
127. Valdar W, et al. Nat Genet
2006;38(8):879–887.
12
128. Langholz B, et al. J Natl Cancer
Inst Monogr 1999(26):39–42.
129. Nebert DW. Clin Genet 1999;
56(4):247–258.
130. Rose RJ, et al. Alcohol Clin Exp Res
2000;25:637.
131. Michael NL. Curr Opin Immunol
1999;11(4):466–474.
132. Eaton WW. Risk Factors for
Mental Health Disorders.
Bethesda, MD: National Institute
of Mental Health; 2004.
133. Ottman R. Genet Epidemiol 1990;
7(3):177–185.
134. Beaty TH. Epidemiol Rev 1997;
19(1):14–23.
135. Caspi A, et al. Science 2003;
301(5631):386–389.
136. Zammit S, et al. Br J Psychiatry
2006;188:199–201.
137. Meaney MJ. Annu Rev Neurosci
2001;24:1161–1192.
 A basic overview of contemporary human
Chapter
2 genetic analysis strategies
Ondrej Libiger and Nicholas J. Schork
Abstract
Human genetics research has received considerable
recent attention as a result of major developments in
molecular genetic technologies, large-scale coordin-
ated research initiatives such as the Human Genome
Project and the International HapMap Project, and
major discoveries concerning the genetic basis of dis-
ease. The strategies that human geneticists exploit
to identify DNA sequence variations that influence
disease susceptibility leverage an understanding of the
behavior of sequence variations when transmitted
from parents to offspring across generations either
within a specific family, among a set of families, or
in the population at large. In this brief review we
describe the basic principles behind the most widely
used human genetic strategies to identify inherited
disease susceptibility factors. We also point out the
limitations and issues that plague these strategies as
well as areas for further study.
Introduction
Human genetic research has a long and illustrious
history, but has received considerable recent attention
as a result of major developments in DNA sequencing
and genotyping technologies as well as the application
of those technologies to the identification of inherited
DNA sequence variations (or ‘variants’) that contribute
to disease susceptibility. A human haploid genome
is over 3 billion nucleotides in length and because
humans are diploid (i.e. inherit one of each of the
22 pairs of autosomal chromosomes as well as the sex
chromosomes comprising the complete 3 billion nucle-
otide genome from a mother and a father) they actually
possess over 6 billion nucleotides. A large number of
these nucleotides (10–20 million) vary frequently
from person to person and other nucleotides differ in
Principles of Psychiatric Genetics, eds John I. Nurnberger, Jr. and Wade H. Berrettini. Published by Cambridge University
Press. © Cambridge University Press 2012.
only a few people or even a single individual. The types
of variation that populate the human genome and
differentiate disease susceptible from nondisease sus-
ceptible individuals range from simple single nucleot-
ide polymorphisms (SNPs) that involve a difference in
a single nucleotide, to large stretches of sequence that
are deleted or even repeated some number of times [1].
If these variants influence gene function in important
ways they could perturb normal physiological function
and lead to disease.
Identifying the specific variants that influence
particular diseases is not trivial given the total
number of variations that might be responsible for
those diseases. However, as daunting as the task of
sifting through the tens of millions of nucleotides that
a group of susceptible individuals possess may seem,
this task is made possible through the exploitation of
a few fundamental genetic phenomena: segregation,
linkage, linkage disequilibrium (LD), and (causal)
gene perturbation. In this brief review we describe
these basic phenomena and how they can be leveraged
to “map” the locations of disease susceptibility vari-
ants in family and population-based genetic studies.
We also describe the problems that plague genetic
approaches to susceptibility variant identification as
well as human genetic research areas that will likely
receive greater attention in the future.
Establishing heritability
Prior to embarking on a study to identify specific
genetic variants that influence susceptibility to a
particular disease or trait, many researchers try to
establish the degree to which the disease or trait is
“heritable” or attributable to inherited factors.
Although there are many strategies for estimating
the overall genetic contribution to a disease or trait,
13
 Chapter 2: Overview of human genetic analysis strategies
virtually all of them rest on the principle that individ-
uals that are more closely related – and hence share
more genes – should exhibit greater phenotypic simi-
larity with respect to a trait of interest than more
distantly related individuals, if genetic factors do, in
fact, influence that trait. For example, if one were to
collect data on pairs of individuals who had varying
degrees of biological relatedness (e.g. monozygotic
[MZ] twins, dizygotic [DZ] twins, cousin pairs, second
cousin pairs, etc.), the concordance rates for the dis-
ease or trait of interest should be higher for the pairs
that were more closely related, if the trait had a
genetic component. From an analysis of the correl-
ation between the degree of relatedness and degree
of phenotypic similarity, one can estimate the degree
to which a trait is heritable [2, 3].
Heritability is often estimated by comparing MZ
twins, whose genomes are identical, to DZ twins,
whose genomes, like all sibling pairs, share on average
half of their content. Depending on how divergent
the concordance rates among MZ and DZ twins are,
one can infer that the trait has a heritability that is low
or high [2–4]. However, there are many assumptions
behind heritability estimates, including those based
on twins [4] that are nontrivial to deal with. One of
the most vexing has to do with distinguishing the
effects of shared environment from shared genes,
since individuals that are more closely related tend
to live in similar environments.
Many ingenious strategies have been devised to
help disentangle shared environment from shared gen-
etic influences on a trait. For example, migration studies
consider the differences in rates of a particular disease
among individuals living in one environment to rates
achieved after those individuals have moved to a differ-
ent environment [5, 6]. If the rates change substantially,
then a major environmental component to the disease
can be inferred. More compelling strategies involve the
study of twin pairs in which one of the twins was
brought up in a different environment (say after adop-
tion) than the other twin. The use of combinations of
twins, some reared apart and some together, can lead to
more precise estimates of the genetic contribution to
a trait, but will never be free from concerns about the
context-dependency of the findings [7].
Genetic variants: linkage and LD
Recombination and linkage. As noted, there are so many
genetic variants that populate the human genome that
14
identifying which subset of those variants is respon-
sible for a particular phenotype is nontrivial. The fact
that these variants differ in frequency both within
and across different populations further complicates
things. However, the fundamental phenomenon
of recombination during meiosis can facilitate the
search for variants influencing causative variants
for a disease. During meiosis homologous chromo-
somes (i.e. the chromosomes pairs, one inherited
from the mother and one from the father) pair up
and exchange chromosomal segments (i.e. ‘‘recom-
bine’’) before a single recombined chromosome
from every chromosome pair arises, contributing to
a haploid ‘‘gamete” which, when combined with a
gamete resulting from the same process from the
other parent, leads to fertilization and a resulting
diploid embryo. Recombination results in sequence
variants neighboring each other on a single parental
chromosome to be inherited together except at the
sites of recombination (i.e. the “breakpoints”). The
closer two positions in the DNA sequence (referred
to as “loci”) harboring variant alleles are to each
other, the more likely they will be transmitted
together, since it is unlikely that a recombination
breakpoint will occur between them (on average there
are only 1–2 recombination events that occur on any
chromosome although there are likely to be more
recombinations on larger chromosomes).
Two loci are said to be “linked” if they are close
enough in proximity such that any (parental) chromo-
some harboring specific nucleotides or variants at
those loci will be transmitted on a single gamete
more often than expected if they were not near each
other; i.e. if two loci are on different chromosomes or
very far from each other such that recombination
breakpoints are likely to occur between them, then
the nucleotides or variants at those loci will be trans-
mitted or “segregate” independently of one another.
Consider the members of the family depicted
in Figure 2.1. The mother, denoted as person X-1,
has a stretch of DNA sequence with the nucleotides
A-C-G-G-G on one of her chromosomes and on the
homologous chromosome the nucleotides A-A-G-G-C.
The father, Y-1, has the nucleotides A-C-G-C-G at the
same positions as the mother on one of his chromo-
somes and on his homologous chromosome has the
nucleotides G-A-T-C-C. The nucleotide T at the third
position is a dominant acting (i.e. only one copy of
the variant is needed in order to get the disease, as
opposed to “recessive” situations in which two copies

A A
C A
G G
G G
G C
X-1
A A A A A G
C C A C A A
G T+ G G G T+
G C G C G C
G C C C C C
C-1 C-2 C-3
are needed, and “additive” or “co-dominant” situ-
ations in which having two copies is more likely to
result in disease than having only one copy) and
highly penetrant variant (i.e. the probability of getting
the disease given that one has the variant is close
to 1.0, as opposed to situations involving “reduced
penetrance” in which this probability is much less
than 1.0) that causes a disease, and hence the father
has the disease (denoted by the “+” by the T allele or
variant). Note that the mother has a homozygous AA
genotype at the first locus, the heterozygous CA geno-
type at the second locus, the homozygous GG genotype
at the third locus, the homozygous GG genotype at the
fourth locus, and the heterozygous GC genotype at
the fifth locus.
In Figure 2.1, the parents have five children and
three have been transmitted the T allele and hence
have the disease (i.e. the children denoted C-1, C-3,
and C-4). Note also that affected children C-3 and
C-4 have inherited the entire paternal haplotype G-A-T-
C-C harboring the disease causing allele T. Individual
C-1 inherited the “T-C-C” component of this paternal
haplotype, but because of a recombination event in the
meiosis leading up to the formation of the gamete
contributing to the fertilized embryo of individual
C-1, the “G-A” variants were recombined with the
“A-G” nucleotides on the paternal homolog. Thus,
the affected offspring not only share the T allele
inherited from their father, but also the “C-C” alleles
at loci linked to the T allele and could have also
all inherited the neighboring “G-A” variants had a
Chapter 2: Overview of human genetic analysis strategies
A G Figure 2.1 A nuclear family with five
C A offspring. The father, individual Y-1,
G T+ possesses a “T” allele on a haplotype,
C C G-A-T-C-C, which causes a disease,
G C denoted by the plus sign. Three of the
offspring inherited the “T” variant (i.e.
Y-1 individuals C-1, C-3, and C-4) and hence
have the disease but only one, C-4,
inherited the entire “G-A-T-C-C”
haplotype due to recombination
(instances of which are denoted by
the light shading).
A G A A
C A A C
G T+ G G
G C G C
G C C G
C-4 C-5
recombination event not occurred in the formation
of the gamete contributing to individual C-1.
LD and variant segregation in populations. Figure
2.1 provided a simple example of the impact of
recombination on the transmission of variant nucle-
otides possessed by a mother and father to their
offspring over a single generation. Figure 2.2 depicts
a hypothetical situation in which a large number
of generations are considered over which particular
variants have been transmitted. Individual A-1
possesses a haplotype with nucleotides G-A-T+-C-C.
The “+” sign next to the T variant again denotes a
disease causing variant. Note that unlike Figure 2.1,
both homologous chromosomal haplotypes are not
depicted since focus is on the transmission of the
“ancestral” G-A-T+-C-C nucleotides from the
“founder” individual A-1. Each broken line denotes
a line of descent between individuals in the latest
generation (e.g. individuals starting with the letters
E, F, G, H, and I) and ancestors in previous gener-
ations. The actual number of generations separating
the individuals is arbitrary. Individual C-1 is a des-
cendent of individual A-1 but, because of a recombin-
ation event in one of his parents (who was also
a descendent of individual A-1), was transmitted
only variants T+-C-C from the original G-A-T+-C-
C haplotype possessed by individual A-1. Thus, all the
descendants of individual C-1 who receive the T+ dis-
ease causing variant (i.e. individuals E-1 to E-N) will
be transmitted, barring further recombination or
mutation events, the A-C-T+-C-C haplotype rather
15
 Chapter 2: Overview of human genetic analysis strategies

G
A
T+
C
C
A
A-1 C
+T
C
G

A B-1
C
T+ G
C A
C T+
G
C-1 C

D-1

A A A A G G G G G G G G G G A A G
C C C C A A A A A A A A A A C C C
T+ T T+ T+ T+ T+ T+ T T T+ T+ T T+ T+ +T +T G+
C C C C C C C C C C G G G G C C C
C C C C C C C C C C C C C C G G C
E-1 E-2 E-3 E-N F-1 F-2 F-3 F-4 F-5 F-N G-1 G-2 G-3 G-N H-1 H-N I-1
Figure 2.2 Chromosomal haplotypes and their histories back some number of generations. Each dashed line represents a line of descent
from a “founder” individual who introduced the chromosomal segment into the population. Only one homologous chromosome is drawn for
each person for simplicity and mate chromosomes are also not depicted. A “+” denotes the presence of a disease phenotype for the individual
possessing the haplotye. Generations and individuals are labeled as groups A-I, and individuals within those groups numbered 1-N.

than the G-A-T+-C-C haplotype possessed by the
original founder A-1.
Individual D-1 is also a descendant of individual
A-1. Because of a mutation event changing the fourth
nucleotide, C, to a G nucleotide (denoted by the
underline) this individual possesses a G-A-T+-G-C
haplotype. Descendants of individual D-1 who receive
the T+ disease causing variant (i.e. individuals G-1 to
G-N) will be transmitted, barring further recombin-
ation or mutation events, the G-A-T+-G-C haplotype
rather than the G-A-T+-C-C haplotype. Note that
other descendants of individual A-1 received the
G-A-T+-C-C haplotype in tact (i.e. individuals F-1
to F-N). Also note that because of reduced or “incom-
plete” penetrance of the T disease-causing nucleotide,
some individuals with the T nucleotide do not mani-
fest the disease (e.g. individuals E-2, F-4, F-5, and
G-2). Environmental factors or additional “protective”
variants may explain why the individuals carrying the
T variant do not manifest the disease.
Individual B-1 developed a T nucleotide mutation
at position 3 in the sequence de novo (i.e. independ-
ently of the event introducing the mutation to indi-
vidual A-1). Individual B-1’s disease causing mutation
is denoted +T to distinguish it from individual A-1’s
T+ mutation. Individual B-1 had haplotype A-C-+T-
C-G. All descendents of individual B-1, who inherit
the +T variant would, again barring recombination
and mutation events, also inherit the A-C-+T-C-G
haplotype (individuals H-1 to H-N). Finally, individ-
ual I-1 does manifest the disease but does not carry
the T variant at the third position. This could reflect
“allelic” or “locus” heterogeneity in which different
variants can cause a disease. Individual I-1 could be
referred to as a “phenocopy” relative to individuals
with the disease who possess the T variant.
Ultimately, Figure 2.2 suggests that there are
groups of individuals in the latest generation who
have the disease for different reasons, some individ-
uals harboring a disease causing variant do not

16
 Chapter 2: Overview of human genetic analysis strategies
manifest the disease due to incomplete penetrance of
the T variant, and among the individuals who have
the disease because they possess the T variant, there
are clusters of individuals who are descendents of
particular individuals who have different haplotypes
or nucleotides surrounding the T variant.
Linkage and LD mapping
Linkage analysis. Linkage analysis is a statistical
method for identifying chromosomal regions
harboring variants that influence a particular pheno-
type. It works by examining the consistency with
which certain variants are transmitted from parents
to offspring within different families. Consider Figure
2.2 and families defined by individuals E-1 to E-N
(the “E” family), individuals F-1 to F-N (the “F”
family), individuals G-1 to G-N (the “G” family),
and individuals H-1 to H-N (the “H” family). The
individuals within each family with the disease share a
certain set of variants. For the E family these variants
are A-C-T+-C-C and for the G family these variants
are G-A-T+-G-C. Note that many of the variants
shared within each family are different from family
to family. Despite this, there is consistency of the
affected family members to share variants at particu-
lar positions in the genome within those families.
Thus, if a researcher genotyped individuals in the
families at the first position, he or she would see that
affected individuals in family E consistently share
an A variant, affected family members in family F
consistently share a G variant, affected family
members in family G consistently share a G variant,
and affected family members in family H consistently
share an A variant. Knowing the position of these
variants in the genome, the researcher would then
examine other variations in the vicinity of the pos-
ition harboring the shared A and G variants and
hopefully identify the T+ variant that causes the dis-
eases. This strategy of identifying variants that are
linked to a genomic position harboring a disease
causing variant by studying families and then refining
such linkages until a causative mutation is found has
often been referred to as “positional cloning” [8].
Many clever statistical analysis methodologies have
been devised to assess evidence for within-family
consistency of shared variants in the linkage analysis
paradigm that account for, e.g. periodic recombin-
ation events within a family, incomplete penetrance,
the frequency of variants in the population, and
phenocopies. These statistical analysis methodologies
are beyond the scope of this chapter; however, the
reader is referred to excellent books on the statistical
methods behind them as well as the results of many
family-based linkage studies [9, 10].
LD mapping. A drawback of the linkage strategy
described above is that one needs families to embark
on the analysis. Population-based LD mapping does
not require families, but does have a few drawbacks of
its own. Consider Figure 2.2, Although individuals
in families E, F, and G who have the disease because
they possess the T+ variant do not share a common
haplotype of which the T+ variant is a component,
they do all share the C variant at the fifth position due
to the historical haplotype transmitted to them over
the generations from individual A-1. Thus, if a
researcher was to genotype all E, F, and G affected
individuals at the fifth position he or she would see
that they share a C variant. If unaffected individuals
did not possess this variant (or at least not with the
same frequency as affected individuals) then the
researcher might infer that the C variant was in LD
with a disease causing variant. He or she could then
genotype the affected individuals at other variant sites
around the site harboring the C variant and hopefully
identify the T+ causative variant.
The definition of a family or population and the
requisite strength of the LD providing evidence of a
region harboring a putative disease causing variant in
the context of LD mapping is somewhat arbitrary. For
example, if one confined attention to individuals in
families F and G, then the G, A, and C nucleotides at
the first, second, and fifth positions would be in LD
with the disease causing variant. In addition, if all the
affected individuals in families E-I were genotyped at
the fifth locus and there were only a few affected
individuals in the sample from the H and I families,
then despite the fact the individuals in the H and
I families do not possess the C variant, there still
might be a very strong association between the
C variant and the disease. Suffice it to say that the
word family or pedigree is typically used to refer to
individuals related to each other over a small number
of generations, whereas a population (or subpopula-
tion) refers to individuals related to each other pos-
sibly over many generations. In addition, in terms of
LD strength, it is obvious that the more affected
individuals possessing a particular variant and the
fewer unaffected individuals not possessing the vari-
ant, the greater the statistical or probabilistic evidence
17
 Chapter 2: Overview of human genetic analysis strategies
that the variant is likely either itself causing the dis-
ease or in LD with something that is.
In order to facilitate population-based LD map-
ping that does not require families but only a broad
definition of individuals sampled from a particular
population, the International HapMap Project was
initiated, with the goal of identifying as many genetic
variations as possible that could be interrogated in
studies seeking to determine if affected individuals
share a particular variant more often than not. In
addition, the International HapMap Project also
sought to characterize the strength of the LD that
neighboring variations exhibit so that researchers
could realistically estimate the likelihood that any
one variant may mark a position in the genome that
is harboring a disease causing variant when geno-
typed in a study [11]. To date hundreds of LD map-
ping studies have been pursued [12]. The results of
these studies are listed in many publicly accessible
databases (http://www.genome.gov/26525384).
Genome-wide versus candidate gene studies. Many
linkage and LD mapping studies focus on particular
genomic regions for which some biological evidence
exists that the region may harbor a gene or variants
that influence disease susceptibility. Such “candidate
gene” or “candidate region” studies are really only as
reliable as the a priori evidence linking the region to
the disease. Alternatively, researchers can merely test
variants located throughout the genome for linkage or
LD with a disease and thereby identify regions
harboring putative susceptibility variants. For linkage
analyses, this would involve looking for patterns of
variant-phenotype “co-segregation” within families
[9]. A traditional measure, known as the “logarithm
of odds” or “LOD” score, is typically used to quantify
evidence for a co-segregation pattern. For genomic
regions in which this LOD score has a value with a
very low probability of occurring purely by chance, a
researcher can infer the existence of a susceptibility
variant or variants. For LD mapping studies, variants
are simply tested in turn for association with a disease
(for example, by contrasting the frequencies of the
variant among individuals with the disease [cases]
and without the disease [controls] using contingency
table methods). For variants that exhibit association
strength (e.g. based on the odds ratio or p-value
obtained from the contingency table analysis) that is
unlikely to occur purely by chance, a researcher can
infer that either that variant in question directly influ-
ences the disease (e.g. T+ in Figure 2.2) or is simply in
18
LD with a variant that does (e.g. the C variant in
relation to the T+ variant in Figure 2.2). As straight-
forward as linkage and LD mapping may seem, how-
ever, there are many complicating factors that are
discussed in the sections below.
Special designs
There are many extensions and modifications to the
basis linkage and LD mapping approaches described
above. We describe the intuitions and basic strategies
behind a few of the more widely used extensions in
the following.
Affected sibling pair tests. Instead of collecting
entire families or large pedigrees with genealogical
information – which can be costly and problematic
if many individuals in the earlier generations have
died – linkage tests that only require affected relative
pairs such as siblings have been proposed. The basic
intuition behind these tests is to identify genetic
regions where the affected pairs share genetic variants
likely to have been transmitted from one of their
parents to a greater degree than expected by chance.
The variants in these regions that are shared by
the relative pairs are thus likely to be “identical-by
descent” from their parents and hence suggest that
variants in those regions are responsible for the
phenotype that all the relative pairs have in common
[13, 14]. The affected relative pairs strategy is intui-
tive, involves a sampling unit, affected relative pairs,
that might be easy to collect and easy to perform
relevant calculations with, and has been applied to a
number of diseases. However, the power to detect
sharing of variants within the relative pair units that
is consistent with a common disease causing variant
across the pairs is notoriously low, and thus the
strategy requires an enormous number of pairs to
identify a true effect in most realistic complex,
multifactorial disease settings [15, 16].
The transmission-disequilibrium test (TDT). A
clever test for both linkage and LD involves the
assessment of the consistent transmission of a par-
ticular variant from a parent who is heterozygous
at the locus harboring that variant to affected off-
spring [17]. Figure 2.3 provides a graphical depiction
of the necessary setting: Each of five parent-affected
offspring trios is portrayed. Affection status is
denoted by the “+” sign. For each trio one of the
parents is heterozygous for the T variant in the second
position of a haplotype (i.e. they all possess the GT

CA CA CA CA CA CA
GG GT GG G T+ GG GT
GG CC GG CC GG CC
AA CA CA
G T+ G T+ G T+
GC GC GC
Figure 2.3 Depiction of the transmission of a particular variant, T, and surrounding variants, A and C, to offspring affected (denoted by
the “+” sign) across five different matings involving a parent heterozygous for the T allele. The assessment of the transmission of a particular
allele from heterozygous parents to affected offspring forms the basis for the “transmission disequilibrium test” (TDT; see text for details).
genotype at the second position). The other parent is
homozygous at this position with the GG genotype).
Four of the five affected offspring have been transmit-
ted the T variant. Under Mendel’s laws, a variant from
a heterozygous parent should, on average, be trans-
mitted to offspring half of the time. The TDT statistic
would thus be consistent with linkage if the frequency
of the transmission of a particular variant from
heterozygous parents to affected offspring departed
significantly from the expected value of half. Note
that the TDT does not involve controls or individuals
without the disease; rather, the “control” in the TDT
setting is the variant associated with the heterozygous
genotype that is not transmitted. This is important
in diseases for which the definition of a control is
problematic (e.g. neuropsychiatric and behavioral dis-
orders; diseases that only manifest when other condi-
tions are present such as an environmental factor;
or age-related diseases). Also, the TDT test does not
make assumptions about the consistency of the fre-
quency of the test variant across different populations
and thus avoids issues associated with the problem
of stratification described below [13, 17]. However,
the TDT strategy does require parental genotype
information and heterozygous parents which may
limit its applicability. In addition, because parental
information is exploited in the TDT setting, one can
test specific hypotheses about parent-of-origin effects
due to, e.g. imprinting.
Admixture mapping. Another clever approach to
linkage and LD mapping leverages populations
known to have arisen from the admixture of two
parental populations with different variant frequency
profiles and different rates of particular diseases. The
individuals in the admixed population can be con-
sidered “hybrid” individuals of the two parental
populations since they will possess variants of greater
frequency in parental populations. Consider, e.g. His-
panic populations as admixed between European and
Native American populations where it is known that
Chapter 2: Overview of human genetic analysis strategies
CA CA CA CA
GG G T+ GG G T+
GG CC GG CC
CC AA
G G+ G T+
GC GC
diabetes has a higher frequency among Native Ameri-
cans, and African-Americans as admixed between
European and African populations where it is known
that certain cancers have a higher frequency among
African individuals. If particular variants that cause a
disease are likely to be more frequent in the parental
populations for which the frequency of that disease
is greater, then one can search for variants possessed
by affected admixed individuals that are known to
be of higher frequency in that parental population.
Although a gross simplification, admixture mapping
along these lines has been used to great advantage for
a few diseases for which an appropriate admixed
population has been identified [18].
Issues in genetic analysis
There are a number of issues that plague linkage
and LD mapping strategies for identifying genetic
variants that influence disease susceptibility. We
briefly describe a few of these issues below and note
that all of them are well-recognized by the genetics
community and that concerted efforts to deal with
them in various ways have been made.
Mutation and nonuniform recombination rates.
It is known that mutation and recombination rates
differ across the genome [11]. This creates interpret-
ive issues for linkage and LD mapping studies since it
might be the case that, despite the proximity of a
genetic variant to an actual disease causing variant it
may not actually exhibit linkage or LD with a causal
variant. To overcome this problem, in-depth studies
of the mutation and recombination rates in the
genome have been pursued as well as detailed studies
of the actual LD relationships of genetic variants in
different populations [11, 12].
Multiple comparisons. In searching for genetic
variants throughout the genome, small subsets of
which may exhibit linkage or LD with a disease,
researchers must be sensitive to the potential for false
19
 Chapter 2: Overview of human genetic analysis strategies
positive findings due to the massive number of statis-
tical tests performed. For example, in a typical
genome-wide association study (GWAS) leveraging
LD mapping, as many as 500 000 to 1 000 000 variants
may be tested for association with a disease. To avoid
false positive associations conservative type I error
rates, on the order of 10–7 or 10–8 are required when
declaring significance for a statistical test involving
any one variant [16].
Genetic heterogeneity. As noted previously, it is
likely that most diseases of contemporary public
health concern have many genetic and nongenetic
determinants. Thus, there may be many different sets
of factors that can cause a disease. Such heterogeneity
can reduce the marginal effect of any one factor on
disease susceptibility and hence create power issues
to detect the effect of that factor in any one study.
To overcome this problem, researchers typically try
to identify individuals with more homogenous
phenotypic profiles or “endophenotypic” profiles on
the assumption that these individuals likely have the
disease due to the same genetic and nongenetic
factors [19, 20].
Interactions. Genetic factors do not typically work
independently of nongenetic factors to mediate dis-
ease susceptibility. Rather, genetic factors often
“interact” with environmental factors such that the
effect of any one genetic factor could be exacerbated
or reduced in the presence of a particular environ-
mental factor or vice versa. Such interactions have
been documented in the literature [21] and statistical
methods for identifying and accounting for such
interactions have been devised [22].
Phenotypic heterogeneity. In addition to genetic
heterogeneity, phenotypic heterogeneity may plague
linkage and LD mapping studies. Phenotypic hetero-
geneity typically arises when genetic and nongenetic
factors influence diseases with similar manifestations,
such as different forms of autism spectrum disorder
or Alzheimer’s disease. Better clinical characterization
and an understanding of the potential pleiotropic
effects of genes (i.e. an understanding of how perturb-
ations in a single gene may influence different bio-
logical and physiological processes) can help mitigate
the untoward effects of phenotypic heterogeneity.
Stratification. One of the most vexing problems
plaguing LD mapping studies is rooted in the fact that
genetic variants typically exhibit different frequencies
in different populations. Thus, for example, a typical
variant may be more or less frequent in African
20
populations rather than European populations. If a
researcher sampled individuals with a disease (cases)
from one population (e.g. Africa) and individuals
without the disease (controls) from another (e.g.
Europe) then any variant with different frequencies
in these two populations could exhibit an association
with the disease in the sample of cases and controls,
resulting in a noncausal, false positive association
[13]. To avoid this genetic background associated
population “stratification” problem, researchers typ-
ically either sample cases and controls from a single
population, use TDT analysis strategies, or use appro-
priate statistical methods to adjust for population
genetic background differences between the cases
and controls.
Biological significance and functional assessments.
Identifying variants that are linked or exhibiting LD
with a particular disease can create interpretive issues
as to the biological significance of the linkage or LD.
This is especially the case if many variants in a par-
ticular region exhibit linkage or LD with a disease,
since it will not often be obvious which variant or
subset of variants are causally influencing the disease
and which are merely in LD with the causal variant(s).
To identify the causal variants, sophisticated labora-
tory-based functional assays must be exploited,
although large-scale functional variant characteriza-
tion projects such as the ENCODE (ENCyclopedia
Of DNA Elements) project have been initiated to
facilitate biological understanding of the role of gen-
etic variants in mediating disease susceptibility [23].
DNA sequencing and genetic studies
The frequency of genetic variants strongly influences
the power studies have to detect their effect on a
disease in a linkage or LD mapping study. As a result,
most linkage and LD mapping studies performed to
date have focused on the detection of common disease
predisposing variants. However, it is now well-recog-
nized that common variants explain only a small to
moderate fraction of the heritability of most diseases
[1, 24, 25]. Strategies that can identify the effects of
collections of rare variants – any one of which if
possessed by an individual could cause disease – are
currently receiving a great deal of attention [1, 24, 25].
Advances in DNA sequencing technologies have
facilitated studies of rare variants, as these technolo-
gies can be used to exhaustively identify all forms of
variation, common and rare, in a genomic region for
 Chapter 2: Overview of human genetic analysis strategies
ACGG
T AGTAGAGTAGAGTCCTAGATCGAAATCGATAGCTAGATAGCAC
ACGG
T AGTAGAGTAGAGTCCTAGATCGAAATCGATAGCTAGATAGCAC
ACGG
T AGTAGAGTAGTGTCCTAGATCGAAATCGATAGCTAGATAGCAC
Cases
ACGG
T AGTAGAGTAGAGTCCTAGATCGAAATCGATAGCTAGATAGTAC
ACCTGAGTAGAGTAGAGTCCTAGATCGAAATCGATAGCTAGATAGCAC
ACGTGAGTAGAGTAGACTCCTAGATCGAAATCGATAGCTAGATAGCAC
ACGTGAGTAGAGTAGAGTCCTAGATCGAAATCGATAGCTAGATAGCAC
ACGTGAGTAGAGTAGAGTCCTAGATCGAAATCGATAGCTAGATAGTAC
ACCTGAGTAGAGTAAAGTCCTAGATCGAAATCGATAGCTAGATAGCAC
ACCTGAGTAGAGTAGAGTCCTAGATCTAAATCGATAGCTAGATAGCAC
ACCTGAGTAGAGTAGAGTCCTAGATCGAAATCGATAGCTAGATAGCAC
ACCTGAGTAGAGTAGAGTCCTAGATCGAAATCGATAGCTAGATAGCAC
Controls
ACCTGAGTAGAGTAGAGTCCTAGTTCGAAATCGATAGCTAGATAGCAC
ACGTGAGTAGAGTAGAGTCCTAGATCGAAATCGATAGCTAGATAGCAC
ACCTGAGTAGAGTAGAGTCCTAGATCGAAATCGATAGCTAGATAGCAC
ACGTGAGTAGAGTAGAGTCCTAGATCGAAATCGATAGCTAGATAGCAC
ACCTGAGTAGAGTAGAGTCCTAGATCGAAATCGATAGCTAGATAGCAC
ACCTGAGTAGAGTAGAGTCCTAGATCGAACTCGATAGCTAGATAGCAC
Figure 2.4 DNA sequences among a group of cases and controls.
The darkened nucleotides denote variant nucleotides. The dashed
boxes denote “functional” sites or elements in the sequence. Note
that the left-most functional site harbors a variant “T” that is more
frequent in cases than controls and the three right-most functional
regions harbor variants that only a few cases and no controls
possess. Also note that there are few variants that only controls
possess but they are not in functional elements.
a group of individuals. Statistical analysis methods for
such studies are not trivial since the relevant tests
must focus on the collective effect of the variants,
rather than any single variant. Figure 2.4 depicts a
setting in which sequences for a genomic region have
been obtained on a number of individuals with and
without a disease. The darkened nucleotides are vari-
ant nucleotides. The dashed boxes indicate regions
in the sequence that contain functional elements.
The first box includes a common variant. The next
three boxes contain rare variants possessed by only a
few individuals. However, these variants are collect-
ively more frequent among the cases rather than the
controls. By leveraging functional annotations associ-
ated with regions of the genome, researchers can
“collapse” rare variants in intuitive and biologically
References San Francisco: Benjamin
Cummings; 1996.
1. Frazer KA, et al. Nat Rev Genet
2009;10(4):241–251. 3. Lynch M, et al. Genetics and Analysis
of Quantitative Traits. Boston:
2. Falconer DS, et al. Introduction
Sinauer and Associates; 1998.
to Quantitative Genetics.
meaningful groups and test their collective frequency
differences between individuals with and without
disease [26, 27].
Conclusions
Research investigating the genetic determinants of
disease susceptibility will continue to receive a great
deal of attention, especially since many contemporary
studies have been successful in identifying suscepti-
bility variants but have clearly not discovered the vast
majority of such variants. Future studies will clearly
take advantage of more sophisticated sequencing and
genotyping technologies, but will also need to leverage
better statistical and computational methods as well
as better ways of integrating functional annotations of
the genome in relevant analyses. In addition, although
identifying susceptibility variants is an important first
step in understanding disease pathogenesis, further
“translational” genomic studies that either consider
the identified susceptibility variants and the genomic
regions they reside in as drug targets or as potential
prognostic or risk factors are needed to fully capitalize
on the information that variant discovery studies
can provide health researchers, clinicians and public
health workers.
Acknowledgements
This work was supported in part by the following
research grants: The National Institute on Aging
Longevity Consortium (grant number U19
AG023122–05); The National Institute of Mental
Health (NIMH) -funded Genetic Association Infor-
mation Network Study of Bipolar Disorder National
(grant number R01 MH078151–03); National Institutes
of Health grants: N01 MH22005, U01 DA024417–01,
P50 MH081755–01, R01 AG030474–02, N01
MH022005, R01 HL089655–02, R01 MH080134–03,
U54 CA143906–01; Scripps Translational Sciences
Institute Clinical Translational Science Award (grant
number UL1 RR025774–03), the Price Foundation
and Scripps Genomic Medicine. Ondrej Libiger is also
supported by a grant from Charles University (GAUK
number 134609).
4. Martin N, et al. Nat Genet 1997;
17(4):387–392.
5. Elford J, Ben-Shlomo Y. In
Kuh D, et al. (eds.). A Life Course
Approach to Chronic Disease
21
 Chapter 2: Overview of human genetic analysis strategies
Epidemiology. Oxford: Oxford
University Press; 2004.
6. Schooling M, et al. Int J Epidemiol
2004;33(6):1219–1226.
7. Lewontin RC. Am J Hum Genet
1974;26(3):400–411.
8. Collins FS. Nat Genet 1995;
9(4):347–350.
9. Ott J. Nat Genet 1991;38(8):
904–909.
10. Sham P. Statistics in Human
Genetics. New York: John Wiley
and Sons; 1997.
11. Frazer KA, et al. Nature 2007;
449(7164):851–861.
12. Manolio TA, et al. Nature
2009;461(7265):747–753.
22
13. Lander ES, et al. Science 1994;265
(5181):2037–2048.
14. Terwilliger JT, et al. Handbook of
Human Genetic Linkage. Baltimore:
Johns Hopkins Press; 1994.
15. Kruglyak L, et al. Am J Hum Genet
1996;58(6):1347–1363.
16. Risch N, et al. Science 1996;273
(5281):1516–1517.
17. Spielman RS, et al. Am J Hum
Genet 1994;54(3):559–560; author
reply 560–553.
18. Freedman ML, et al. Proc Natl
Acad Sci U S A 2006;103
(38):14068–14073.
19. Burmeister M, et al. Nat Rev Genet
2008;9(7):527–540.
20. Morris AP, et al. Genet Epidemiol
2010;34(2):188–193.
21. Caspi A, et al. Science 2003;
301(5631):386–389.
22. Thomas D. Nat Rev Genet 2010;
11(4):259–272.
23. Birney E, et al. Nature 2007;
447(7146):799–816.
24. Bodmer W, et al. Nat Genet
2008;40(6):695–701.
25. Schork NJ, et al. Curr Opin Genet
Dev 2009;19(3):212–219.
26. Li BS, et al. Am J Human Genetics
2008;83(3):311–321.
27. Madsen BE, et al. PLoS Genet
2009;5(2):e1000384.
 DNA methods
Chapter
3 David W. Craig
Introduction
In this chapter, we outline methods for studying the
genetic basis of psychiatric disorders. We focus on
methods and study designs emerging within the past
decade, and those likely to be influential in the near-
term. We caveat that numerous pivotal findings and
methods emerged throughout the 1990s and in the
early part of the 2000s that are still relevant today, and
excellent review articles are available.
Before beginning our formal overview of DNA
methods, it is important to highlight what we will
and won’t be covering in this chapter. This chapter
will specifically be limited to two types of technologies,
high-density single nucleotide polymorphism (SNP)
genotyping arrays and so-called “next-generation
sequencing” (also referred to as second-generation
sequencing). We focus on the types of study designs
these technologies enable and how they have been
used to discover the genetic basis of psychiatric dis-
orders. We do not review extensively the large
breadth of genetics findings made within the last
decade. Since we are focusing on DNA in this chapter,
we limit our discussion to identifying germ-line gen-
etic variants predisposing to disease. This means that
we do not give a full discussion of study designs that
are particularly effective at identifying somatic vari-
ation. Mainly, in the case of psychiatric disorders the
ability to identify somatic DNA variants involved in
disease pathogenesis is limited by either knowing
what tissue to investigate or gaining access to that
tissue. Simply, most DNA samples for psychiatric
disorders come from a whole-blood sample – only a
small number of studies focus on analyzing DNA
within brain tissue for example. Related, we do not
provide an extensive discussion of epigenetic DNA
changes and how they may impact disease progression.
Principles of Psychiatric Genetics, eds John I. Nurnberger, Jr. and Wade H. Berrettini. Published by Cambridge University
Press. © Cambridge University Press 2012.
Taken as a whole, this chapter is thus focused on tools
and study designs for identifying genetic variants to
explain the heritable component of psychiatric disease.
The two types of methods we describe, SNP geno-
typing and sequencing, are focused on characterizing
two different types of variants predisposing to disease:
common and rare. Common genetic variants predis-
posing to disease are those frequently found within a
population and give a person an increased “relative
risk” for developing a disease. As will be discussed
later, SNP genotyping arrays provide a mechanism
for testing association of common SNPs within a
region for association to a heritable disorder. Rare
variants increasing one’s relative risk for developing
a disorder are primarily accessible through sequencing-
based studies (noting the exception that SNP arrays
can identify large rare chromosomal abnormalities).
Over the past several years, there have been multiple
studies investigating the majority of common variation
within various psychiatric disorders at a genome-wide
level. These are termed genome-wide association stud-
ies (GWAS). In the next several years, we can expect
that genome-wide sequencing will become realistic,
enabling a genome-wide search for both common
and rare-variants contributing to psychiatric disorders.
We now expand on these topics in greater detail.
SNP genotyping
SNPs as a tool for mapping diseases
across the genome
While 99.9% of the human genome is identical
between individuals, it is estimated that there is on
the order of 10 million SNPs that differentiate indi-
viduals [1]. SNPs are nucleotide variants at specific
23
 Chapter 3: DNA methods
positions in the human genome, such as from an
adenosine to thymine (A!T) or a guanine to a cyto-
sine (C!T). By definition, they are commonly found
when sampling a population, usually when the minor
allele is found at a frequency above 1% when
sampling a population.
There are three general classes of SNPs when it
comes to function: (1) those with strong functional
significance that dramatically alter a gene’s behavior
(classic single nucleotide mutations); (2) those with
more subtle functional effects that predispose an indi-
vidual to disease in concert with genetic background
or environment (functional SNPs); (3) those that are
completely silent with respect to function (nonfunc-
tional SNPs). While SNPs can contribute directly to
disease predisposition by modifying a gene’s function,
it is really their ability to be used as genetic markers
to detect nearby disease-causing mutations through
association or family-based linkage studies. As
markers, they specifically allow the ability to track
how certain regions of the genome co-segregate with
disease [2].
The immediate importance of genotyping a high
density of SNPs does not necessarily come from their
functional relevance, but rather the proximity of the
SNP to the disease-causing genetic mutation or func-
tional SNP [3]. By mapping thousands of SNPs across
the genome, it is possible to determine common
regions between individuals, either ancestral or famil-
ial. In familial linkage studies, SNPs can be used to
track sections of chromosomes that co-segregate
through a pedigree with the trait. In case-control
association studies, SNPs can be identified that statis-
tically associate to complex, multigenic diseases due
to their close proximity to other, possibly unmapped
polymorphisms. In this review, several predominant
technologies that allow for rapid, low-cost genotyping
of tens to hundreds of thousands of SNPs across the
genome will first be described. The basis of how, with
increasing SNP density, SNP mapping can be used to
determine the genetic basis of more complex diseases
will also be addressed.
Of particular importance is that many SNPs are
in significant linkage disequilibrium (LD) [4, 5]. LD
refers to when two polymorphisms are inherited
together more often than they would be expected to
by random chance alone due to their physical
proximity to one another on a chromosome. In
high-density SNP data, some SNPs are expected to
be in LD due to their close proximity, in some cases
24
within hundreds of base pairs. Unfortunately, multi-
point logarithm of odds (LOD) score calculations
assume independence between SNP markers, which
is not the case for neighboring SNPs in significant
LD. In this case, LOD scores will be biased, likely
inflated. This finding is concerning since the biasing
of a LOD score may be uneven throughout the
genome due to variable meiotic recombination rates
at different positions within the genome and the
associated uneven physical distribution of the
markers (on most genotyping panels) throughout
the genome. Therefore, without correct attention to
LD, one may falsely identify a genetic locus predis-
posing to a disease. As an alternative, one can con-
struct virtual multiallelic markers in the form of
haplotypes of markers that always travel together
through a pedigree [6]. These virtual markers can
be used for a two-point LOD score calculation with-
out introducing a bias to the LOD score. While
analysis of high-density SNP data for linkage studies
is still being optimized, it is encouraging to see the
development of software that can handle the large
amount of SNP data that is already being utilized to
identify disease-causing genes. In the GWAS section
of this chapter we further describe LD in mathemat-
ical terms.
Technologies for high-density SNP
genotyping
SNPs have been used since the mid 1990s as markers
for identifying the genetic basis of disease. The use
was often limited to studying candidate gene studies,
since sequencing large numbers of SNPs is costly. In
the early part of the 2000s researchers began working
on the International HapMap Project that had the
goal of assessing common variation across multiple
ethnicities by genotyping millions of SNPs. One part
of the project was identifying where common SNPs
were in the human population and the other part was
genotyping those SNPs across 240 individuals. In part
driven by this project, methodologies emerged for
inexpensively and rapidly genotyping tens of thou-
sands of SNPs across multiple individuals. Some
methods were based on sequencing, some based on
mass spectrometry (Sequenom), whereas others were
based on single nucleotide extension of oligionucleo-
tide microarrays to allow differentiating alleles. These
arrays were already widely used for quantifying RNA
expression, and a clear path was already in place to

adapt them to classifying whether a person was AA,
AB, or BB (“B” being the reference or ancestral allele
and “A” being the alternative allele). By October 2005,
the HapMap project’s first phase was published with
the genotypes of approximately 2.4 million SNPs on 4
populations and 240 individuals. A planned outcome
of the HapMap project was an assessment of the
independence between two adjacent SNPs [7]. The
fact that SNPs in close proximity are not independent,
but rather in LD, meant that only a subset of SNPs
need to be genotyped in order to assess common
variation. By 2005, two companies had developed
products capable of genotyping a half million SNPs,
and population-based study designs became possible
and termed GWAS [8]. Using these genome-wide
high-density SNP genotyping arrays, it was possible
to complete genome-wide association tests [9].
Simply, by genotyping hundreds to thousands of
cases and controls, one tested to see if a particular
SNP allele was associated with an increased risk of
a disorder.
We provide a technical overview of two technolo-
gies that have been used in the majority of linkage
studies and population-based GWAS: Illumina
BeadArray Mapping arrays and Affymertrix Gene-
Chip Arrays. Of course, other genotyping platforms
exist. Sequenom based on mass-spectroscopy differ-
entiation is a highly flexible platform for custom
genotyping of a few thousand SNPs across a large
number of people. Likewise, Applied Biosystem’s
TaqMan and SNPplex assays allow SNP genotyping
in hundreds of SNPs across thousands of samples.
Services by Perlegen using Affymetrix technology
provided millions of genotypes for hundreds of indi-
viduals. We focus on the Illumina and Affymetrix
technology because they are typically used within
many laboratories, and are the front-end discovery
engine.
®
The Affymetrix GeneChip Mapping Arrays are a
direct extension of the more commonly used expres-
sion profiling GeneChip Assays [10, 11]. In the SNP
genotyping assay, fluorescently labeled fragments of
DNA containing specific SNPs are genotyped by
whether or not they hybridize to a sequence-specific
oligomer probe set tiled on a silicon wafer [12]. The
Gene-Chip Mapping Array assay uses a whole
genome polymerase chain reaction (PCR) amplifica-
tion of digested genomic DNA ligated to universal
adapters. DNA is purified, biotin labeled, fragmented,
and hybridized to a microarray. For each SNP, tiled
Chapter 3: DNA methods
oligomer probes interrogate each of two possible SNP
alleles. There have been several different designs of
Affymetrix SNP genotyping arrays, and a variety of
SNP genotyping analysis strategies. The number
of SNPs genotyped in the Affymetrix 6.0 exceeds over
1 million SNPs and includes additional markers
specific for copy number.
The BeadArrayTM platform, manufactured by
Illumina, also allows for high-density genotyping.
In this assay, hybridization is only used for detec-
tion, and SNP allele discrimination is enzymatic
[13–15]. In the enzymatic allele discrimination part
of the assay, genomic DNA is attached to a solid
support to better facilitate PCR amplification. Three
oligomers, specific for each SNP, are used for exten-
sion of the sequence containing the SNP. The first
two oligomers are specific at the 3´ end for the two
different SNP allele types. Only one oligomer will
anneal depending on the particular SNP allele. The
third oligonucleotide anneals in the reverse direction
downstream and also contains an addressing
sequence. All three oligomers contain a universal
primer sequence at the 5´ end for subsequent PCR
amplification. Oligomers are annealed and an allele-
specific primer extension is carried out, joining the
two annealed oligomers. Depending on the SNP
allele, one of two possible nucleotides will be incorp-
orated. In a new reaction with the extended product,
one of two allele-specific primers containing differ-
ent fluorescent dyes will anneal. A PCR reaction is
carried out and amplicons are hybridized to micro-
scopic beads at the end of a fiberoptic bundle.
Unique to the BeadArray, the address sequence,
and not the SNP, hybridizes to the array. These
address sequences are not public but they do provide
an important advantage, since each bead is not spe-
cific to a particular SNP. Consequentially, probing a
custom set of SNPs does not require a custom array.
Genotype calls are made from transforming
trimmed mean intensities of Cy3 and Cy5, calculated
for each bead type, into a polar coordinate system
where outliers are rejected. For large datasets, quality
scores for each SNP can be calculated using a separ-
ate software package that compares individual bead
intensity distributions with their standard distribu-
tions. Like Affymetrix, current arrays by Illumina
can assay over a million SNPs. While SNP genotyp-
ing algorithms are largely static with Illumina, the
technology has evolved to assay multiple samples on
a single slide.
25
 Chapter 3: DNA methods
Copy-number analysis
Recently the impact of human genomic variation in
copy number and the link to disease risk has been
demonstrated in studies of autism, schizophrenia, and
amyotrophic lateral sclerosis (ALS) among others.
Interestingly, copy number variant (CNV) findings
were often made with high-density SNP genotyping
arrays, which originally were not designed for CNV
detection. As with SNPs, CNVs can be both patho-
genic or be part of normal genetic variation. Indeed
identifying CNVs in individuals with no overt symp-
tomology of disease, laid the groundwork for the
study of CNVs in disease that we are seeing emerge
in the literature today. Importantly these initial
manuscripts showed that these so-called structural
variations can exist as both common and rare vari-
ants, and due to this finding it is now commonplace
to perform both SNP and CNV association analysis
in concert for every whole genome association study.
To facilitate this, both of the major SNP chip manu-
facturers have more recently built in probes specific-
ally designed to facilitate CNV measurement.
Much of the work in the CNV field has primarily
revolved around the accurate identification of CNVs
within each assayed sample. The testing of CNVs for
association with a heritable trait or disease is then
typically performed using the standard statistical tools
for SNP association analysis and, as with SNPs, these
tools may vary based upon the frequency of the
detected CNV (e.g. common versus rare).
A CNV that occurs in greater than 1% of the
general population is termed a CNP – copy number
polymorphism. The association analysis of CNPs util-
izes the typical standard statistical tests (e.g. allele
frequency comparisons) used for SNP associations.
Typically CNPs are di-allelic and can therefore be
incorporated as if they were simply SNP genotypes
(e.g. A/A = no copies, A/B = one copy, and so on).
These will then be assessed using standard quality
control (Mendelian inheritance, duplicate samples,
and Hardy–Weinberg equilibrium [HWE]) and asso-
ciation (allelic Chi-square) tools. This is advantageous
as these methodologies for statistical analysis already
exist due to work in the SNP genotyping field. In
some cases CNPs exhibit three of more classes. Rare
CNVs, which we may be powered to identify even in
the typically “common variant” powered GWA study
design as we are able to use multiple SNP probe
fluorescent values to identify rare CNVs, are less
26
constrained when it comes to their analysis. There-
fore, for these rare CNVs it is critical to demonstrate
their true apparent frequency in large numbers of
control individuals. Additionally, the discovery of a
rare CNV with a putative association with disease must
result from an equivalent identification strategy in the
cases and controls – in other words, the rare CNVs will
not be identified simply by examining the array data
from the cases (to avoid “ascertainment bias”).
Linkage or family-based study designs
In linkage studies, surrogate markers are used to track
genomic regions as they co-segregate with a disease
through a family structure. While linkage studies have
historically relied on multiallelic microsatellite markers,
it is now possible to use thousands of SNP markers
for this purpose. High-density SNP genotyping offers
an alternative to microsatellites in that while they are
biallelic markers, the shear number of markers typed
provides higher overall information that is more evenly
distributed. Indeed, in one of the first high-density
whole genome studies, Matise and coworkers demon-
strated that a carefully selected set of 2988 SNPs is
more informative than the Marshfield Clinic screening
microsatellite version 10 [16]. Perhaps the largest
family-based linkage study using SNPs was performed
in autism, with nearly 7500 Affymetrix 10 000 arrays
across 1500 families. New linkage peaks were found
with evidence for the neurexin pathway – perhaps most
interesting were observations of large copy number
changes (discussed in the previous section) [17].
Analysis of linkage studies is well established in
the literature and excellent texts are available on this
topic [18]. For the purposes of this review, the pri-
mary statistical output at each marker is a LOD score,
which can be simplified as the likelihood that the
underlying genetic mutation for a disease is associated
with a particular marker. Generally, a LOD score
above 3 (1024 to 1 odds of a disease gene residing at
that location in the genome) is considered significant.
While analysis of microsatellite genotype data are well
established, there are a number of practical difficulties
in analyzing high-density SNP data in linkage studies.
Both Merlin [19] and GeneHunter [20] programs
are based on the Lander–Green algorithm [21] and
have been used in whole genome, high-density link-
age studies with SNP data. Programs relying on the
Lander–Green algorithm are generally able to handle
large numbers of SNP markers since computational

time with this algorithm scales linearly with the
number of markers, and exponentially with people.
Conversely, other algorithms (e.g. the Elston–Stewart
algorithm [22]) used in some linkage software pack-
ages are not as appropriate for high-density SNP data
since they scale exponentially with marker number.
GWA studies
Most diseases fall under the category of complex
diseases – those that are influenced by multiple envir-
onmental and genetic factors [23]. As complex dis-
eases are not typically inherited in simple Mendelian
fashion, traditional linkage studies are not always
appropriate or even possible. In 1996 Kathleen
Merikangas and Neil Risch predicted that if research-
ers could assemble 1000 well-characterized cases and
1000 properly matched controls, these studies could
find disease variants with moderate risks [3]. Recent
advances in genotyping technology and analytical
tools have now made it possible to perform GWAS
using thousands of samples from well-characterized
case and control populations. These studies assay
hundreds of thousands of markers across the genome
and have elucidated the genetic underpinnings for
many diseases and disorders, including type 2 dia-
betes, breast cancer [24], prostate cancer, rheumatoid
arthritis, Crohn’s disease, autism [25], bipolar dis-
order [26], and resulted in hundreds of publications
in the last four years. Association studies based on
high-density SNP panels are termed indirect associ-
ation studies, as the interrogated SNPs are not neces-
sarily expected to contribute to disease susceptibility.
Rather, genotyped SNPs are expected to be in close
proximity to the true disease-causing polymorphism
such that the genotyped SNP statistically associates
with the disease in a case-control manner.
Case-control GWAS are fundamentally based on
the concept of LD. As previously discussed, LD is
when two polymorphisms are inherited together
more often than they would be expected to by random
chance alone. This will most likely be the case with
two SNPs in very close physical proximity. In an
association study, the SNP being genotyped is
expected to be in LD with the disease-causing genetic
variant. The amount of LD between any two markers
is a function not only of meiotic recombination but
also natural selection, mutation, genetic drift, ances-
tral population, demographics, and mating patterns
[1]. From a practical perspective, it is useful to
Chapter 3: DNA methods
understand how LD is analytically characterized. LD
is predominantly measured by D´ and R2. D is the
difference in probability of observing two marker
alleles on the same haplotype versus observing them
in an independent population: D = pAB-pA*pB, where
pAB is the frequency of observing the two calls, A and
B, on the same haplotype versus the expected prob-
ability, pA*pB, if the two calls were completely inde-
pendent. A normalized version of D, D´, is commonly
used so that markers with different allele frequencies
can be compared. A measurement of D´ ¼ 0 implies
complete independence, and a value of one implies
that the lower frequency allele is consistently found
with the higher frequency allele. Unfortunately,
intermediate values of D´ are difficult to interpret.
The second value, R2, is emerging as a more often
employed metric, since it allows for comparison of R2
measurements from two distinct SNP pairs. An R2
of 0 implies independence and a value of one implies
that the marker alleles at one marker are completely
predictive of what allelic genotype exists at the other
marker. Conceptually, R2 indicates the amount of
information one SNP provides towards the other
SNP marker. The R2 value, including its intermediate
values, can be used to approximate sample size.
The primary measure of effect in case-control stud-
ies is calculation of an odds ratio (OR). An OR is the
odds of a person having a disease given that they have a
particular SNP [27]. An OR for a SNP of greater than
one suggests an increased risk, while an OR of less than
one implies a protective effect. An OR of one indicates
the SNP has no impact on whether a person will have a
disease. OR is dependent on four parameters: OR of
the true disease-causing genetic variation, extent of LD
between the two markers, marker frequency, and the
disease frequency. Statistical significance is typically
determined by an allelic test of association, though
various disease model tests can be implemented. One
of the most common software packages for analyzing
and quality control of GWA data is PLINK developed
by Shawn Purcell and colleagues [28].
In combination between linkage and case-control
association studies are a number of family-based
approaches largely falling under the category of trans-
mission/disequilibrium tests (TDTs) [29]. In-depth
reviews of TDT and their variants are available else-
where, thus the major points of TDT as they relate
to high-density SNP data will only be mentioned.
The rationale behind TDT is that in the absence of
linkage and association between markers, alleles will
27
 Chapter 3: DNA methods
be transmitted randomly from parents to children
[30]. TDTs have a number of advantages over case-
control association studies. First, TDT eliminates
affects of population stratification. Second, having
parental genotype information allows for more accur-
ate construction of haplotypes. For example, if a child
has the heterozygous genotype of C/T at a SNP,
the father is T/T and the mother is C/C, it is possible
to deduce that the C is from the maternal and the
T is from the paternal chromosome. Construction
of virtual multiallelic markers crossing several neigh-
boring SNPs becomes more reliable with parental
information. Practically, TDTs are often difficult to
conduct since obtaining genotype information on
parents is clinically overwhelming. This problem
is significant for high-density SNP platforms as it
has been shown that (compared with multiallelic
markers) biases can emerge if both parents are not
genotyped for biallelic SNPs [31].
When designing an association study, there are
many possible sources of error. These include obvious
technical errors such as an underpowered study or
statistically expected false positives, but also includes
unseen sampling errors resulting from population
stratification and population admixture [32]. The dif-
ficulty in identifying truly associated SNPs is already
evident in candidate gene association studies that
genotype a significantly fewer number of densely
populated SNPs. Indeed, some reviews have found
that as few as 5–30% of findings from association
scan are reproduced [33]. False associations from
population stratification occur when a regionally or
ethnically defined group with different risks of disease
is unknowingly sampled. Beyond a high rate of statis-
tical false positives and population stratification,
population admixture can also lead to false associ-
ations. Lastly, in assessing the significance of a test,
it is preferable to determine p-values empirically by
comparison with large sets of random permutations;
that is, sets of randomized data without association.
Permutational analysis is more robust to false
assumptions about the distribution, such as from
stratified populations or admixed individuals, than
standard distributions.
Sequencing
We are at an inflection point in the history of geno-
mic sequencing; a moment every bit as disruptive as
the introduction of commercial capillary sequencers
28
in 1999. Recently developed massively parallel “next
generation” sequencing technologies have begun to
replace the previously dominant Sanger sequencing
technology [34] for large-scale sequencing projects
[35]. Technologies like Illumina’s Genome AnalyzerTM
(GA) [36], Roche’s 454TM Sequencing System (454)
[37], and Applied Biosystems” SOLiDTM [37] are able
to generate billions of bases of raw sequence in a
matter of days. These technologies generate relatively
short reads, typically from a few tens to a few hun-
dreds of bases in length, with a general inverse relation
between the total number of reads and the read length.
In the context of whole human genome resequencing,
on the order of a billion short reads are required to
accurately resequence an individual genome (10–20 
depth coverage) using these technologies. In the con-
text of sequencing a specific region across a large
number of individuals, it is possible to sequence
multiple individuals within the same sequencing run
using bar coding methods leveraging the pseudo
single-molecule capabilities of next-generation tech-
nologies. While there are many potential applications
of these technologies, we focus on those involving
sequencing DNA for the purpose of identifying the
genetic basis of disease, specifically dividing this
section into targeted methods and genome-wide
approaches.
A key feature of all these platforms is the concept of
“pseudo single-molecule sequencing”, where each read
is derived from one molecule rather than an aggregate
measure of all reads. Consequently, if an individual is
heterozygote at a base having both A and T nucleotides,
reads at that base will have discrete values of either A or
T rather than a mixture of signals from both alleles.
The term “pseudo” is used only to indicate that some
sort of clonal amplification has been used for a single
molecule. Resequencing a base multiple times using
pseudo single-molecule sequencing opens the window
for distribution sampling algorithms for polymorphism
discovery and genotype calling algorithms.
Next-generation sequencing technologies
Until recently, widespread sequencing of DNA has
largely been built around modifications to the Sanger
method developed in the 1970s [34]. Key technological
advances permitted scaling of Sanger sequencing
including capillary electrophoresis, automation, robot-
ics, and fluorescent-labeled nucleotides. Over the last
five years, most sequencing has been completed using

capillary Sanger sequencing through machines such
as the ABI 3730. More recent efforts to develop
sequencing platforms with fundamentally different
chemistries have matured. The result is a series of
next-generation sequencing technologies with the
implied goal of making whole genome resequencing
rapid and inexpensive. Some of the technological
advances behind the current wave of commercially
available next-generation sequencers (often referred
to as second generation) are pyrosequencing, revers-
ible terminators, and advanced optics. Integrating
these advances with oligo arrays and beads has
allowed massively parallel sequencing. Next-gener-
ation sequencing largely is also about sequencing
one molecule at a time, whether by direct measure-
ment or indirectly through amplification.
Additional features include mate-pair sequencing,
where bases from both ends of a larger DNA frag-
ment are sequenced such that the two resultant
reads are paired. Since both the SOLiD and GA have
relatively short reads (< 150 bp), paired data is highly
useful for mapping reads to the genome. We provide
an overview of features for all three commercially
available platforms (SOLiD, GA and 454) below.
SOLiD. The Applied Biosystems SOLiD platform
enables massively parallel sequencing of clonally amp-
lified DNA fragments linked to beads. Clonal ampli-
fication is accomplished through emulsion-based
PCR. The sequencing chemistry is based on sequen-
tial ligation of dye-labeled oligonucleotide octamer
probes that are hybridized, ligated, and imaged in
consecutive reactions for generation of paired bases,
skipping five bases. Skipped bases are sequenced in
five additional rounds of ligation-based sequencing,
utilizing universal sequencing primers with different
start sites [37]. Each probe assays two-base positions
at a time and each base is assayed in two independent
extension reactions with overlapping probes. The
system uses four fluorescent dyes to encode for the
sixteen possible dibase (adjacent pairs of bases) com-
binations using a degenerate coding scheme that sat-
isfies a number of rules (H. Breu, Applied Biosystems,
pers. comm., 2008). A single color in the read can
represent any of four dibase combinations, but the
overlapping properties of the dibase probes and the
nature of the color code (so-called “color space”)
allow for error-correcting properties.
Illumina sequencing by synthesis. Illumina’s Genome
Analyzer and HiSeq are built around single-base exten-
sion and reversible terminators. Clonal production is
Chapter 3: DNA methods
accomplished by bridged PCR-based amplification
to fragmented genomic DNA tethered to a surface
by a common oligo sequence ligated during sample
preparation. Sequencing consists of single-base
extension using fluorescently labeled nucleotides that
have reversible terminators. Following imaging, the
terminators are cleaved and the process is repeated.
Important features include phasing due to incomplete
cleaving of the terminator and pre-phasing due to
incorporation of unterminated nucleotides, both of
which result in an error profile that increases at later
bases; and use of two lasers with four filters to detect
four fluorophores leading to a bias towards over-
calling C/T bases and undercalling A/G.
Roche 454. The Roche GS FLX next-generation
sequencer is based on technology acquired by Roche
as part of their 2007 purchase of 454 Life Sciences
and consequently these sequencers are commonly
referred to as “454 sequencers”. This 454 sequencing
is based on massively parallel pyrosequencing of
DNA fragments adapter-ligated to small beads which
are then subjected to PCR and washed into a small
well on a PicoTiterPlate where the well contains the
enzymes required for the sequencing chemistry.
The 454 platform is somewhat different from the
other 2 platforms in that it is less parallel (approxi-
mately 1 million reads per run), the reads are longer
(400–600 bases), and runs are shorter (hours rather
than days). Sequencing is based on quantifying light
released as bases complementary to the bound tem-
plate strand are ligated. Unterminated bases are
washed over the slide one at a time so the primary
failure mode of the 454 is that homopolymer runs
result in multiple ligation events that can only be
recognized by quantifying the amount of light released.
Emerging platforms. A number of emerging
platforms are becoming more widespread in their
use. Complete Genomics now offers whole genome
sequencing through a service center utilizing unchained
reads on self-assembled DNA nanoarrays as a series
of short ( 10–20 bp) reads connected by a defined
distance [38]. Ion Torrent (Life Technologies) utilizes
a variation of pyrosequencing that differs in their
detection method utilizing direct measurement of
hydrogen incorporation on a semiconductor array.
Over the next few years, undoubtedly many new
methods will emerge focused on the task of sequen-
cing DNA (and RNA) molecules. The major challenge
will likely not be generation of data, but rather analy-
sis of data and that is the focus of the next section.
29
 Chapter 3: DNA methods
Data analysis and standards for
next-generation sequencing data
Analysis of next-generation sequencing data falls into
three categories: (1) alignment/assembly; (2) quality
control; and (3) variant calling.
Alignment of short-reads is not surprisingly one
of the first areas to see various analysis tools emerge.
Excellent reviews exist and we highlight a few aligners
[39]. Jeck et al. [40] have described an aligner that
uses an extension of an aligned shorter read to com-
pensate for the fact that accuracy decreases as the read
gets longer. Cokus et al. [41] have presented a prob-
abilistic aligner, along with a pipeline of analysis tools
for Illumina’s Genome Analyzer and HiSeq. Schatz
et al. [42] describe an excellent use of graphic cards to
accelerate the assembly and alignment of sequence
reads. Warren et al. [43] describe a program, SSAKE,
for stringent assembly of nearly identical sequences
that uses a prefix tree under specific search strategies.
Addressing data management and analysis, Trombetti
et al. [44] describe a computational pipeline with
integrated data storage for analyzing data from
454 sequencers. Receiving rather quick adoption for
assembly is Velvet developed by Daniel Zerbino at
EMBL-EBI [45]. A few mature nonvendor analysis
suites to emerge are: BWA developed by Heng Li and
colleagues at the Wellcome Trust Sanger Institute [46]
(WTSI) and Mosaik/GigaBayes [47]/EagleView [48] by
Gabor Marth and colleagues at Boston College (BC).
Quality control steps are an obvious part of any
analysis pipeline. For next-generation sequencing,
quality control is less about filtering (as was the case
for SNP genotyping) and more about quantifying
quality. The first step this occurs is post-alignment
recalibration of sequence quality data. Sequence data
is typically quantified using a PHRED scale for each
base, where a value typically from 0 to 60 is a loga-
rithmic description of the probability of the assigned
base. While the sequencing software typically pro-
vides an estimate of this value, this value is often
systematically off. By inspecting alignments at pos-
itions believed to be nonpolymorphic one can assess
the mismatch rate under various parameters, such as
base-position. Recalibration, thus, corrects the quality
scores to be more consistent with a logarithmic prob-
abilistic framework that feeds directly into variant
calling. Additional quality control values can include
mapping quality that also provides a logarithmic
approximation of the probability that the alignment
30
is correct. Alignments can be further refined through
a micro or local realignment step, where evidence
from multiple reads is used to assess for more com-
plex variants. For example, three neighboring SNPs
might be better explained using an 1 bp insertion.
Finally, a process of marking or removing duplicates
is typically completed prior to base-calling. Essen-
tially, repeated measures of the same fragment do
not provide additional information about a variant
and can disrupt some variant calling software.
Calling of genetic variants begins first with those
variants smaller than the read, before proceeding to
structural variants. These variants are typically SNPs
and short insertion/deletions (indels). While almost
all variant calling software has some heuristic com-
ponent, at some level variant calling relies on: (1) the
quality of bases predicting a variant; (2) the quality
of mapping of the read to the reference genome; and
(3) the proportion of reads predicting a variant at
a given position. Additionally, some variant calling
approaches utilize additional information such as the
directionality of read or quality of bases immediately
neighboring the predicted variant. In practice, most
variant calling software can yield fewer than 10% false
positives when read depth is greater than 10  cover-
age per ploidy. The most difficult to call genetic
variants are generally heterozygotes since they depend
not only on error rate, but also on binomial sampling
of the variant. In other words, if three of three reads
contain the variant, it is still reasonable that the vari-
ant position is an undersampled heterozygote.
Structural variants are typically ascertained by
three separate approaches involving analysis of read-
pairs or fragments: abnormal read pairs (RP) that
differed in orientation or separation from expected
values; discrepant read depth (RD) in comparative or
absolute terms; and alignment/assembly issues in
terms of split reads, directed de novo assembly
(microassembly) or global de novo assembly (macro-
assembly). Types of structural variants assessable
by paired data include:
 Large deletions – Evident by a shorter than
expected distance between read pairs, indicating
that a deletion has occurred between read pairs.
The region containing the deletion will be missing
alignments.
 Inversions – Evident by inconsistent direction
between read pairs, indicating that one of the read
pairs is a section that has been inverted. The region

at the point of the inversion will also be missing
alignments
 Copy number variants – Evident by a large
increase or decrease in alignments across a region
of the genome, outside of the expected Poisson
distribution and inconsistent with repeats from
other regions of the genome.
 Large insertions – Evident by systematic mapping
of only one of the two read pairs, along both
read direction. Nonmapping read pairs may also
be assembled into a contig.
 Translocations – mapping of read pairs to
separate chromosomes in a manner that is
inconsistent with mapping to repeat regions.
Data standards
Next-generation sequencing produces a massive
amount of data. Sequencing a genome 30  coverage
requires nearly 100 billion bases, each containing
value describing quality. Typically, researchers retain
both the raw-sequence data and the aligned data. If
not careful, it is quite easy to end up with terabytes of
data describing the genome of one individual.
The complexity and changeability of the data
flows is a powerful argument in favor of a modular
design based on specified data input and output for-
mats. The data formats will not only ease integration
of modular tools developed, but it will also help
establish the necessary standards for catalyzing devel-
opment of third party software that is interchangeable
and platform independent. The key data formats
currently in use are short read format (SRF), FASTA
with quality (FASTQ), sequence alignment/map (SAM),
binary SAM (BAM), and genotype likelihood format
(GLF), described briefly below.
 Sequence read format (SRF). A platform agnostic
generic format for DNA sequence data and
capable of including platform specific markup
language.
 Fastq. A FASTA-like format that includes two
records for each read, one with sequence and one
with quality scores. Quality scores are typically
transformed so each score can be represented
as a single printable ASCII character [33].
 Sequence alignment/map (SAM). A simple,
compact, plain-text format for describing the
alignment of a query sequence to a reference
sequence or assembly developed through the
1000 Genomes Project [49]. SAM defines header
Chapter 3: DNA methods
records that can include information about the
reference sequence, the platform, and the run.
SAM also has alignment records that include
the query sequence, quality scores and alignment
information along with the ability to add
user-defined attributes to each alignment record.
 BAM. Binary equivalent to SAM that can
optionally be indexed for fast nonsequential access.
 Genotype likelihood format (GLF). A format for
describing evidence for variants at a position
within a reference genome.
 Variant consensus format (VCF). A format for
describing genetic variants and their respective
alleles emerging out of the 1000 Genomes Project.
Sequencing
Current next-generation sequencing technologies can
generally provide gigabases of sequence per day at
relatively low cost. At this point, it is still quite expen-
sive to sequence the genomes of thousands of individ-
uals. Moreover, often evidence from GWAS or other
family-based linkage studies may point to a specific
region of the genome for sequencing.
The pseudo-single molecule nature of next-
generation sequencing allows an elegant solution
whereby short 3–10 bp DNA barcodes are ligated to
each genomic fragment for multiple individuals and
pooled prior to sequencing. Essentially, the first
several bases or an indexing read allows the researcher
to determine from which individual each read is
derived. Multiplex sequencing by DNA barcoding can
significantly reduce the risk of false positives and nega-
tives when resequencing large genomic regions because
coverage can be fine-tuned to ensure that each base
is sequenced many times, making next-generation
sequencing a practical method for mutation/poly-
morphism discovery. Beyond adding the obvious
advantage of multiplexing large numbers of samples
within a run, DNA barcoding offers two additional key
advantages: direct measure of base-by-base error rate
and reduction of day-to-day variability.
The next hurdle is to amplify or enrich for only
the genomic regions of interest. Tiling across a region
using PCR and then pooling is one option. If long-
range PCR is used it can be practical for sequencing
tens of kilobases. More recently, multiple groups have
demonstrated the use of various targeting strategies
using oligo microarrays for enrichment or for multi-
plexed amplification [50–52]. These efforts build on
31
 Chapter 3: DNA methods

existing targeting strategies including molecular
inversion probes and multiplexed PCR. These recent
papers are exciting although they have also described
a number of unresolved problems and challenges with
capturing a specific portion of the genome.
Hybridization-based target enrichment. Targets for
resequencing must be isolated or enriched. However,
long-range PCR is less practical for targeting hun-
dreds of exons disparately distributed throughout
the genome. Methods have emerged usually utilizing
custom oligonucleotide microarrays for hybridizing
and capture fragmented genomic DNA, while nontar-
geted DNA is washed away. Essentially, these methods
utilize biotinylated RNA “baits” to fish out targeted
regions from a “pond” of fragmented DNA. Magnetic
streptavidin beads are used to bind and pull down
baits, as nontargeted genomic DNA is washed away.
Targeted sequencing of coding sequencing or the
exome (exome-sequencing) is one of the more obvi-
ous regions of the genome to specifically capture for
sequencing. Indeed, multiple groups now report
exome sequencing using various hybridization-based
capture methods, and one of the first applied
examples of identifying a genetic variant of large
effect using exome-sequencing has been shown for
Freeman–Sheldon syndrome [53, 54]. Practically,
targeted sequencing on an exome-wide scale is not
100%, though it remains remarkably efficient. While
numbers vary, typically about 90% of the targeted
bases are captured at sufficient depth to allow for
heterozygote calling. About 80–90% of exonic bases
are targeted. While subjective, it is frequently viewed
that despite the inefficiencies of exon-sequencing
from capture, it is possible to get 80% of the bases
sequenced at about 1/10th the cost of whole genome
sequencing. The primary sources of variability in
these early assays are differences in hybridization
efficiencies between probes and off-target capture
(e.g. immediately neighboring the region or com-
pletely off-target). Despite these inefficiencies, it is
clear at this point that many exon-sequencing studies
will be attempted in the next several years. The alter-
native is whole genome sequencing. At the time
of writing whole genome sequencing is somewhat
cost-prohibitive for sequencing hundreds to thousands
of individuals ($6000–30 000 for a 30  genome
depending on assumptions). One appeal of exon
sequencing is that it is more readily easy to interpret
certain types of changes, such as a one base deletion
causing a premature truncation or nonsynonymous
amino acid change.
To a certain extent, the ability to meaningfully dis-
tinguish between benign and functional mutations will
be a major challenge for making practical discoveries –
both at the exome level and the genome-wide level.
Data collection efforts such as the 1000 Genomes Pro-
ject [38], which aims to characterize common human
variation, or efforts such as the ENCODE (ENCyclo-
pedia Of DNA Elements) project [55], which aims to
guide our understanding of how genetic variation can
impact function will be critical, allowing researchers
to hopefully move beyond genetic associations to
understanding the functional mechanism of specific
variants causing or contributing to psychiatric disease.
One of the major challenges for any study is being
able to sift through the large number of germline
genetic variations in any one individual. Based on
the 1000 Genomes Project [38], a typical person will
have 3–4 million genetic variants mostly composed
of SNPs. For SNPs, typically 10 000–12 000 will be
nonsynonymous. Of these a person will carry 250 to
300 potential loss-of-function genetic variants and
50–100 variants linked to disease. Clearly differentiat-
ing carrier mutations that don’t lead to disease in a
person with a psychiatric illness from the true disease-
causing mutation(s) in that person will rely heavily
on the sample set and the simple observation of a loss-
of-function does not imply a person’s psychiatric
illness is due to that variant. In analysis of families,
co-segregation can be used to filter to limited regions
of the genome. Identifying enrichment of one variant
class in a group of cases versus controls is another
possibility. Likely, use of any of these methods may only
identify a candidate list of potentially causative variants
and further functional characterization of the mutation
by other experimental means will be warranted.

References 3. Risch N, et al. Science

1. Carlson CS, et al. Am J Hum
Genet 2004;74(1):106–120.
2. Kruglyak L, et al. Nat Genet 2001;
27(3):234–236.
1996;273(5281):1516–1517.
4. Pritchard JK, et al. Am J Hum
Genet 2001;69(1):1–14.
5. Reich DE, et al. Nature 2001;
411(6834):199–204.
6. Schaid DJ, et al. Am J Hum Genet
2004;75(6):948–965.
7. Altshuler D, et al. Nature
2005;437(7063):1299–1320.
8. Craig DW, et al. Expert Rev Mol
Diagn 2005;5(2):159–170.

32

9. Hirschhorn JN, et al. Nat Rev
Genet 2005;6(2):95–108.
10. Chee M, et al. Science 1996;
274(5287):610–614.
11. Fodor SP, et al. Nature 1993;
364(6437):555–556.
12. Kennedy GC, et al. Nat Biotechnol
2003;21(10):1233–1237.
13. Fan JB, et al. Genome Res 2004;
14(5):878–885.
14. Gunderson KL, et al. Genome Res
2004;14(5):870–877.
15. Oliphant A, et al. Biotechniques
Suppl 2002;56–58:60–51.
16. Matise TC, et al. Am J Hum Genet
2003;73(2):271–284.
17. Hu-Lince D, et al. Am
J Pharmacogenomics 2005;
5(4):233–246.
18. Ott J, et al. Eur Child Adolesc
Psychiatry 1999;8 Suppl 3:43–46.
19. Abecasis GR, et al. Nat Genet
2002;30(1):97–101.
20. Kruglyak L, et al. Am J Hum Genet
1996;58(6):1347–1363.
21. Lander ES, et al. Proc Natl Acad
Sci U S A 1987;84(8):2363–2367.
22. Elston RC, et al. Hum Hered
1987;21(6):523–542.
23. Carlson CS, et al. Nature 2004;
429(6990):446–452.
24. Gudmundsson J, et al. Nat Genet
2007;39(5):631–637.
25. Weiss LA, et al. Nature 2009;
461(7265):802–808.
26. Purcell SM, et al. Nature 2009;
460(7256):748–752.
27. Teng J, et al. Genome Res 1999;
9(3):234–241.
28. Purcell S, et al. Am J Hum Genet
2007;81(3):559–575.
29. Spielman RS, et al. Am J
Hum Genet 1993;52(3):
506–516.
30. Schulze TG, et al. Am J Med Genet
2002;114(1):1–11.
31. Sun F, et al. Am J Epidemiol
1999:150(1):97–104.
32. Marchini J, et al. Nat Genet
2004;36(5):512–517.
33. Neale BM, et al. Am J Hum Genet
2004;75(3):353–362.
34. Sanger F, et al. Nature 1977;
265(5596):687–695.
35. Ley TJ, et al. Nature 2008:
456(7218):66–72.
36. Bentley DR. Curr Opin Genet Dev
2006;16(6):545–552.
37. Margulies M, et al. Nature
2005;437(7057):376–380.
38. Durbin RM, et al. Nature
2010;467(7319):1061–1073.
39. Li H, et al. Brief Bioinform 2010;
11(5):473–483.
40. Jeck WR, et al. Bioinformatics
2007;23(21):2942–2944.
Chapter 3: DNA methods
41. Cokus SJ, et al. Nature 2008;
452(7184):215–219.
42. Schatz MC, et al. BMC
Bioinformatics 2007;8(1):474.
43. Warren RL, et al. Bioinformatics
2007;23(4):500–501.
44. Trombetti GA, et al. BMC
Bioinformatics 2007;8(Suppl 1):
S22.
45. Zerbino DR, et al. Genome Res
2008;18(5):821–829.
46. Bentley DR, et al. Nature 2008;
456(7218):53–59.
47. Marth GT, et al. Nat Genet
1999;23(4):452–456.
48. Huang W, et al. Genome Res
2008;18(9):1538–1543.
49. Li H, et al. Bioinformatics 2009;25
(16):2078–2079.
50. Albert TJ, et al. Nat Methods
2007;4(11):903–905.
51. Okou DT, et al. Nat Methods
2007;4(11):907–909.
52. Porreca GJ, et al. Nat Methods
2007;4(11):931–936.
53. Ng SB, et al. Nat Genet 42(9):
790–793.
54. Ng SB, et al. Nature 2009;461
(7261):272–276.
55. The ENCODE (ENCyclopedia
Of DNA Elements) Project.
Science 2004;306(5696):
636–640.
33
 In silico analysis strategies and resources
Chapter
4 for psychiatric genetics research
Ali Torkamani, Trygve Bakken, and Nicholas J. Schork
Abstract
Although a number of strategies for identifying
genetic variations that influence common complex
neuropsychiatric diseases have been proposed, imple-
mented, and pursued, many of these strategies have
not been able to yield compelling insights into disease
pathogenesis. The reasons for this are themselves com-
plex, but it is arguable that extending and integrating
available strategies to include detailed biological infor-
mation can only improve their yield. In this review
we consider computational methods, databases, and
related resources that can help put into perspective
the biological and functional significance of genes
and genetic variations interrogated in contemporary
gene mapping strategies for neuropsychiatric diseases.
Computational methods for evaluating the biological
significance of genetic variations, as opposed to labora-
tory assays, are quick and, for the most part, easy to
use and are growing in sophistication. We provide a
discussion of the limitations of available resources, but
ultimately argue that more integrated approaches to
the genetic dissection of complex neuropsychiatric
conditions are necessary and likely to be the rule rather
than the exception in future investigations.
Introduction
Geneticists often consider searching the entire human
genome for variations that influence susceptibility to
disease [1–3]. Strategies for searching the human
genome for disease-predisposing variations are
varied, each having certain advantages and dis-
advantages [2]. The earliest methods focused on the
analysis of families and pedigrees that included a
number of members who had the disease of interest
and a number of members who did not have the
disease of interest. The members of these families
Principles of Psychiatric Genetics, eds John I. Nurnberger, Jr. and Wade H. Berrettini. Published by Cambridge University
Press. © Cambridge University Press 2012.
34
were then genotyped at “marker” loci throughout the
genome and statistical evidence for co-segregation, or
“linkage”, between a genetic marker (whose position
in the genome was known) and unobserved variants
that were likely responsible for the disease was then
assessed. The number of markers that could be inter-
rogated in early linkage studies was small due to
technical limitations with available genotyping assays.
In addition, the results of linkage studies often sug-
gested that a large genomic region was likely to
harbor predisposing variations, raising questions not
only about how one could identify the causative vari-
ations in the region, but also about the biological
meaningfulness of the result. Extensions and adapta-
tions to the basic family-based approaches involving
analyses limited to sibling pairs and mother–father–
affected offspring trios were introduced in order to
avoid the need for large families [2, 4]. Although
linkage analysis strategies did produce notable suc-
cesses in the identification of disease-predisposing
DNA sequence variations, most of these successes
were confined to monogenic and overtly Mendelian
diseases [2]. When early linkage-based methods were
applied to major neuropsychiatric conditions, the
results were largely equivocal, raising concerns about
the amenability of such diseases to linkage-based
strategies for gene discovery (see, e.g. [5]).
In the wake of the lack of overt, consistent success
of the application of traditional linkage-based methods
to common complex diseases, approaches that con-
sidered the interrogation of a larger set of markers
in population-based samples were proposed [6]. The
strategies involve testing the association between
hundreds-of-thousands to a few million genetic vari-
ations characterized on individuals, e.g. with and
without a specific phenotype. If evidence for associ-
ation between the variation and the phenotype was

strong enough to overcome the inevitable multiple
comparisons that must be made, then one could infer
that either the genetic variation is likely to be causally
related to the phenotype or in linkage disequilibrium
with a variation that is causally related to the pheno-
type. Such genome-wide association studies (GWAS)
have been pursued for over 300 different disease
phenotypes [3]. Queriable databases cataloging the
results of these studies have been devised to facilitate
the dissemination of information to the genetics
community (https://gwas.lifesciencedb.jp/cgi-bin/gwasdb/
gwas_top.cgi and http://www.genome.gov/gwastudies/).
Although GWAS have produced many unequivocal
findings, the number of variations found to be com-
pellingly associated with disease is small and collect-
ively the variations do not explain a large fraction of the
risk of most diseases. In addition, GWAS for neurop-
sychiatric conditions have been largely negative and not
replicable (though meta-analyses have been somewhat
more positive and consistent).
Despite the unmitigated successes of the GWAS
approach to many diseases, there is now general
acceptance that the primary focus of GWAS on
common genetic variations that contribute to disease
is itself limited and that other approaches are needed
[7–9]. In fact, there are many researchers who have
concluded that in order to exhaustively identify gen-
etic variations that contribute to disease predispos-
ition, one must exhaustively study the genomes of
individuals, either through focused DNA sequencing
strategies, assays that could reveal rare structural vari-
ations or polygenic factors that contribute to disease,
or complete genome sequencing [10–13].
No matter what genetic strategy researchers have
ultimately chosen in the identification of genetic vari-
ations that predispose to disease (e.g. family-based
linkage analysis or association analyses involving can-
didate genes, GWA study paradigms, and/or exhaust-
ive DNA sequencing protocols) they have largely been
pursued as a statistical exercise, with evidence impli-
cating a genomic region, gene, or specific genetic vari-
ations being more or less based on the probability
that a linkage or association signal could have occurred
purely by chance. It is now generally accepted that the
more biological information one can leverage in deter-
mining the significance of a linkage or association the
better. However, what biological information should
be leveraged – or is even feasible both logistically and
financially to leverage – is an open question. The use of
databases and computational methods to sort out the
Chapter 4: In silico analysis strategies and resources
potential functional significance of a gene or variation
implicated in a linkage or association is now relatively
easy and is becoming more commonplace. In the
following we describe databases and available compu-
tational-based functional analysis tools that can facili-
tate understanding of the biological meaningfulness of
an association involving a specific genomic region,
gene, or specific variation. We note that there are some
excellent reviews of available resources that provide an
extensive list of references not discussed here [14–16].
We also note that Tables 4.1, 4.2, and 4.3 list websites
describing the resources we consider.
Sequence analysis
DNA sequence data must be dealt with at some level
in any genetic investigation. It is therefore important
to know what tools might be available for manipulat-
ing such data. For example, translating DNA or RNA
sequences in protein sequences is a basic task in
sequence analysis. Depending upon the aims of the
investigator, the resultant protein sequence may be
used to search for homologous proteins, be aligned
with other proteins of interest, or the protein sequence
may be queried for functional domains to gain insights
into the function of the protein. All these tasks will be
described in the following sections. Since nucleotides
code for amino acids in sets of three (codons), there are
three distinct translations for any nucleotide sequence
depending upon the choice of a starting nucleotide.
Automated tools for translation, for example, the
ExPasy Translate tool (http://www.expasy.ch/tools/
dna.html) will automatically generate all three protein
sequences in the forward direction, as well as three
protein sequences reading the nucleotides in the
reverse direction. However, the correct translation
initiation site must usually be known beforehand.
In other cases, an investigator may have a DNA
sequence as well as a corresponding amino acid
sequence, but be interested in determining which
specific codons correspond to the amino acids of the
protein sequence. For example, an investigator may
want to determine where or how a disease-causing
mutation in the DNA sequence affects the resultant
protein sequence. Specific web-based tools have been
developed to align a nucleotide sequence with a pro-
tein sequence. The EBI tool, Genewise (http://www.
ebi.ac.uk/Tools/Wise2/index.html) [17], can be used
to automatically generate an alignment for a given
input protein and nucleotide sequence.
35
 Chapter 4: In silico analysis strategies and resources

Table 4.1 DNA sequence manipulation tools and resources.

Analysis type Function Website (name of resource)

General Genome browser http://genome.ucsc.edu/ (UCSC Genome Browser)
General Aggregate information http://www.genecards.org/ (GeneCards)
Sequence analysis Translation http://www.expasy.ch/tools/dna.html (Expasy Translate Tool)
Sequence analysis Alignment http://www.ebi.ac.uk/Tools/Wise2/index.html (Genewise)
Sequence motifs Motifs http://meme.sdsc.edu/meme/intro.html (MEME/MAST)
Sequence motifs Motifs http://genie.dartmouth.edu/scope/ (SCOPE)
Sequence motifs Motifs http://zlab.bu.edu/cluster-buster/ (Cluster-Buster)
Sequence motifs Motifs http://159.149.109.9/modtools/ (MoD Tools)
Sequence motifs Logos http://weblogo.berkeley.edu/ (WebLogo)
Alignments General alignment http://jaligner.sourceforge.net/ (JAligner)
Alignments Similarity search http://blast.ncbi.nlm.nih.gov/Blast.cgi (Blast)
Alignments General alignment http://www.ebi.ac.uk/Tools/fasta33/index.html (FASTA)
Alignments General alignment http://www.ebi.ac.uk/Tools/clustalw2/ (ClustalW)
Alignments General alignment http://www.ebi.ac.uk/Tools/t-coffee/ (T-Coffee)
Alignments General alignment http://www.ebi.ac.uk/Tools/muscle/ (MUSCLE)
Alignments General alignment http://www.ebi.ac.uk/Tools/sequence.html (EBI)
Alignments Local alignment http://www.ebi.ac.uk/Tools/emboss/align/index.html (EMBOSS)
Alignments Structural alignment http://www2.ebi.ac.uk/dali/ (Dali)
Alignments Structural alignment http://www.ncbi.nlm.nih.gov/Structure/VAST/vastsearch.html (VAST)

compilation of disease-causing mutations. A major

Sequence variation
Obviously, the goal of genetic studies is to identify
DNA sequence variations that contribute to vari-
ations in phenotypic expression. Therefore, under-
standing the biological contexts in which sequence
variations might influence a phenotype is crucial to
genetic studies. There are a number of databases and
tools that can be queried or studied to obtain bio-
logical context. Some of these databases and tools
focus on the general properties of types of genetic
variations, such as single nucleotide polymorphisms
(SNPs), such as PupaSuite (http://pupasuite.bioinfo.
cipf.es/), TAMAL (http://neoref.ils.unc.edu/tamal/),
and BrainArray (http://brainarray.mbni.med.umich.
edu/Brainarray/Database/SearchSNP/snpfunc.aspx), or
structural variations, such as DGV (http://projects.tcag.
ca/variation/) and dbVAR (http://www.ncbi.nlm.nih.
gov/projects/dbvar/). However, there are a large number
of databases dedicated to the mere collection and
repository of information is the Online Mendelian
Inheritance in Man (http://www.ncbi.nlm.nih.gov/sites/
entrez?db=omim) database (OMIM) [18]. The OMIM
database has been maintained electronically by the
late Victor McKusick and colleagues for over 40 years.
The database has two basic types of entries, gene-
centric entries and disease-centric entries. Disease-
centric entries generally begin with a basic phenotypic
description of the disease, describe clinical features,
inheritance patterns, biochemical features, initial
mapping of risk loci, molecular genetics, and animal
model information. In some cases, information about
clinical management or other clinical or molecular
features is also available. Gene-centric entries also
begin with a general description, describe initial
cloning, gene structure, mapping, biological func-
tions, early studies and history, population genetics,
and eventually end in specific mutational information
regarding the gene or disease. Note that each entry is

36
 Chapter 4: In silico analysis strategies and resources

Table 4.2 DNA sequence variation, mutation, and polymorphism analysis tools.

Analysis type Function Website (name of resource)

Sequence variation Disease mutations
Sequence variation Disease mutations
Sequence variation Disease mutations
Sequence variation Disease mutations
Sequence variation Alzheimer’s mutations
Sequence variation Parkinson’s mutations
Sequence variation SNPs
Sequence variation SNPs
Sequence variation Disease associations
Coding SNPs Prediction
Coding SNPs Prediction
Coding SNPs Prediction
Coding SNPs Prediction
Coding SNPs Prediction
Coding SNPs Machine learning
Noncoding SNPs Regulatory SNPs
Noncoding SNPs Transcription factor
binding sites
Noncoding SNPs Regulatory regions
Noncoding SNPs Regulatory regions
Noncoding SNPs Enhancers
Structural variations General information
Structural variations General information
http://www.ncbi.nlm.nih.gov/sites/entrez?db=omim (OMIM)
http://www.hgmd.cf.ac.uk/ac/index.php (HGMD)
http://www.hgvbaseg2p.org/index (HGVBase)
http://www.mutationdiscovery.com/md/MD.com/home_page.jsp
(Mutation)
http://www.alzforum.org/res/com/gen/alzgene/default.asp
(AlzGene)
http://www.pdgene.org/ (PDGene)
http://www.ncbi.nlm.nih.gov/projects/SNP/ (dbSNP)
http://www.hapmap.org/ (Hapmap)
http://www.ncbi.nlm.nih.gov/sites/entrez?db=gap (dbGAP)
http://sift.bii.a-star.edu.sg/ (SIFT)
http://www.pantherdb.org/tools/csnpScoreForm.jsp (subPSEC)
http://genetics.bwh.harvard.edu/pph/ (Polyphen)
http://mmb2.pcb.ub.es:8080/PMut/ (PMut)
http://www.snps3d.org/ (SNPs3D)
http://www.cs.waikato.ac.nz/ml/weka/ (Weka)
http://burgundy.cmmt.ubc.ca/cgi-bin/RAVEN/a?rm=home (RAVEN)
http://www.biobase-international.com/gene-regulation/ (BIOBASE)
http://wwwmgs.bionet.nsc.ru/mgs/gnw/trrd/ (TRRD)
http://www.oreganno.org/oregano/Index.jsp (ORegAnno)
http://www.oreganno.org/oregano/Index.jsp (VISTA Enhancer)
http://projects.tcag.ca/variation/ (DGV)
http://www.ncbi.nlm.nih.gov/dbvar/ (DBVar)

not limited to the above information, nor is the entir-
ety of this information available for all diseases/genes.
The database is easily searchable by keywords and
is thoroughly crosslinked to other related disease or
gene entries as well as references to primary literature.
OMIM entries are not restricted to genes with vari-
ation information. Suggestive disease links without
mutational evidence are also included. The OMIM
database is an excellent resource for a disease overview
and history, however, the mutational information is
not comprehensive, and many of the entries contain a
select set of representative or classical variations.
Another, more comprehensive, mutational resource
is the Human Gene Mutation Database (http://www.
hgmd.cf.ac.uk/ac/index.php) (HGMD) [19]. This
database is less descriptive but represents a more
comprehensive catalog of disease-causing mutations.
Mutations are broken down into categories, such as
missense, splicing, or regulatory variations, and are
searchable by a number of specific identifiers such as
gene symbol, OMIM number, or disease name. The
database contains both publicly available data and
proprietary information, which can be accessed by
paying a subscription fee. Note that both HGMD
and OMIM are hand curated, using a different set
of evidence requirements for entry into the database.
Some mutations reported in the literature may not
pass these quality requirements and thus, may not be

37
 Chapter 4: In silico analysis strategies and resources

Table 4.3 Functional assessment of genetic variations, genes, and sets of genes.

Analysis type Function Website (name of resource)

SNP function General http://pupasuite.bioinfo.cipf.es/ (PupaSuite)
SNP function General http://neoref.ils.unc.edu/tamal/ (TAMAL)
SNP function General http://brainarray.mbni.med.umich.edu/Brainarray/ (BrainArray)
Protein features Structure http://www.wwpdb.org/ (PDB)
Protein features Structure http://www.rcsb.org/pdb/home/home.do (PDB)
Protein features Structure http://www.ebi.ac.uk/pdbe/ (PDBE)
Protein features Motifs http://prosite.expasy.org/ (PROSITE)
Protein features Domains http://pfam.sanger.ac.uk/ (Pfam)
Protein features Domains http://www.ebi.ac.uk/Tools/pfa/iprscan/ (InterProScan)
Protein features Signal sequences http://www.cbs.dtu.dk/services/ (Various)
Gene expression Expression http://www.ncbi.nlm.nih.gov/geo/ (GEO)
Gene expression Expression http://www.ebi.ac.uk/arrayexpress (ArrayExpress)
Gene expression Expression http://ecoliwiki.net/colipedia/index.php/Center_for_Information_
Biology_Gene_Expression_Database_(CIBEX) (CIBEX)
Gene expression eQTL analysis http://compute1.lsrc.duke.edu/softwares/SNPExpress/ (SNP Express)
Gene expression Expression patterns http://serpanalytics.com/site/tstag.molgen.mpg.de (TStag)
Gene expression Expression patterns http://biogps.gnf.org/#goto=welcome (BioGPS)
Gene expression Brain expression http://www.brain-map.org/ (Allen Brain Atlas)
Pathway analysis Analysis http://www.pathguide.org/ (Pathguide)
Pathway analysis Analysis http://www.geneontology.org/ (Gene Ontology)
Pathway analysis Analysis http://bond.unleashedinformatics.com/ (BOND)
Pathway analysis Analysis http://www.reactome.org/ (Reactome)
Pathway analysis Analysis http://www.kegg.jp/ (KEGG)
Pathway analysis Visualization http://www.cytoscape.org/ (Cytoscape)
Pathway analysis Visualization http://www.biologicalnetworks.net/ (BiologicalNetworks)
Pathway analysis Visualization http://visant.bu.edu/ (VisANT)
Pathway analysis Analysis http://www.genmapp.org/ (GenMAPP)
Pathway analysis Visualization http://www.cs.bilkent.edu.tr/patikaweb/ (PATIKAweb)
Pathway analysis Analysis http://www.ariadnegenomics.com/ (Genomics Pathway Studio)
Pathway analysis Analysis http://www.ingenuity.com/ (Ingenuity)
Pathway analysis Analysis http://www.genego.com/ (GeneGO)

included in one, or both, of the databases. Other
mutational databases exist, such as Mutation Discov-
ery (http://www.mutationdiscovery.com/md/MD.com/
home_page.jsp) and HGVBase (http://www.hgvbaseg2p.
org/index), however, they do not appear to be nearly
as comprehensive as OMIM or HGMD. Disease spe-
cific databases exist as well, for example AlzGene
(http://www.alzforum.org/res/com/gen/alzgene/default.
asp) and PDGene (http://www.pdgene.org/) dedicated
to Alzheimer’s and Parkinson’s disease respectively.

38

General mutation databases containing commonly
occurring, and not necessarily disease-associated,
mutations also exist. The most useful and compre-
hensive of these is the Single Nucleotide Polymor-
phism database (dbSNP), maintained at the NCBI
website (http://www.ncbi.nlm.nih.gov/projects/SNP/)
[20]. Investigators may submit any sequence vari-
ations they discover to this database, which aims to
catalogue all known DNA variations, including those
falling into regions of the genome not mapping to any
particular gene. The interface of this database is simi-
lar to that of OMIM, and is searchable by simple
keyword entries. Additionally, each SNP is designated
an identifier, which can be used to query a particular
SNP in other databases or search for phenotypic infor-
mation in primary literature. SNPs in dbSNP are also
accompanied by flanking sequence information, gene
information, population frequencies if they are known,
and validation status, among other pieces of useful
information. dbSNP is particularly useful for identifying
the known genetic variations in your gene of interest.
Another general SNP database is the HapMap data-
base (http://www.hapmap.org/) [21]. All HapMap SNPs
are contained within dbSNP; HapMap contains in-
depth information regarding the population distribu-
tion of SNPs, as well as the linkage structure around
each SNP. Linkage structure reflects the co-occurrence
of SNPs with one another, that is, how well possession
of one SNP predicts the possession of another neigh-
boring SNP. Many SNPs are tightly linked with one
another because DNA is inherited in large blocks,
called haplotype blocks, and thus the genetic vari-
ations within a block travel together from parent to
child. The HapMap project essentially provides a map
of genetic information so that investigators may
pursue studies linking particular diseases or other
phenotypes to inherited genetic information.
This haplotype block information may be lever-
aged to design association studies. In an association
study, SNPs which can be accurately used as surro-
gates for the majority of SNPs within a single haplo-
type block, are used to represent, or tag, a particular
haplotype block, and are thus named tag SNPs. In
carrying out an association study, investigators geno-
type large groups of individuals with or without a
phenotype of interest, and analyze the resultant geno-
type information for frequency differences among the
disease versus control group. Tag SNPs with a statis-
tically elevated frequency within the disease group
are likely to be in close proximity to a linked genetic
Chapter 4: In silico analysis strategies and resources
aberration which contributes to disease risk. The data
collected from such large studies is available in the
primary literature, or it may be deposited in the
dbGAP database (http://www.ncbi.nlm.nih.gov/sites/
entrez?db=gap) [22]. Some information within the
database is available publicly; however, in general, prior
permission or authorization is required to gain access to
the bulk of the information contained within the data-
base. Access can be requested by applying to the National
Institutes of Health (NIH) Data Access committee.
Nonsynonymous polymorphisms
When faced with a series of potential nonsynonymous
polymorphisms that may underlie your phenotype of
interest, it can pay to prioritize mutations before
testing them further. A number of computational
techniques aimed at differentiating neutral from func-
tional nonsynonymous polymorphisms have been
developed [23]. These methods may be applied to
narrow down a series of common polymorphisms in
candidate genes before an association study in order
to increase the power of the study. Alternatively these
methods may be helpful in narrowing down the
number of mutations to be tested for functional sig-
nificance in in vitro or in vivo settings, sparing the
investigator a lot of time, energy, and money.
Most computational methods will take advantage
of sequence conservation in some form. SIFT, argu-
ably the most popular of these methods, takes as its
input your protein of interest, searches for similar
proteins, generates multiple alignments with these
proteins, and uses these alignments to determine the
degree of conservation at any residue, which in turn
is used to output a probability that an amino acid
substitution is deleterious [24]. SIFT is available at:
http://blocks.fhcrc.org/sift/SIFT.html. Other conser-
vation-based methods, such as subPSEC, are also
available online (http://www.pantherdb.org/tools/csnp
ScoreForm.jsp) [25]. Other methods, such as Polyphen
(http://genetics.bwh.harvard.edu/pph/) [26] or PMut
(http://mmb2.pcb.ub.es:8080/PMut/) [27], take advan-
tage of other sequence-based or physiochemical prop-
erties in addition to conservation. Other methods,
such as SNPs3D, (http://www.snps3d.org/) take full
advantage of available crystallographic data when
performing predictions [28]. It should be noted that
PMut and Polyphen also use crystallographic data in
some instances. In some cases, predictors specific
to a protein family, or even a specific protein have
39
 Chapter 4: In silico analysis strategies and resources
been generated. For example, predictors have been
developed for DNA repair genes [29], protein kinases
[30], and G-protein coupled receptors [31].
If training data is plentiful for your protein, or
protein family, of interest, the framework for develop-
ing custom machine learning prediction methods is
available as a Java-based platform, Weka (http://www.
cs.waikato.ac.nz/ml/weka/) [32]. Weka implements a
series of popular data-mining methods such as neural
networks, support vector machines, random forests,
decision tables, and a whole host of other predictive
algorithms. Attribute selection algorithms are also
implemented to help an investigator determine what
features are informative when developing a prediction
model. Custom predictive methods may not be feas-
ible in many cases because of the requirement for a
large training set to form predictions with confidence.
In such cases, a general predictor is sufficient. How-
ever, caution should be taken in interpreting the
results of any of the above prediction methods. It is
important to understand how the predictions are
being formed, and how the predictive attributes may
apply to your gene of interest. For example, mutations
occurring at relatively unconserved amino acids on
the surface of a protein may be considered neutral by
structural or evolutionary predictors, even though as in
the case of sickle cell anemia, these mutations can result
in disease [33]. The training data used to train any
predictor can greatly influence its predictive accuracy
on a subset of proteins. For example, structure-based
predictors trained solely upon DNA repair genes would
not be of great predictive value for other types of pro-
teins. Though this is an extreme example, more subtle
biases in training data could result in inaccuracies when
the predictive algorithm is applied to an underrepre-
sented protein family [34]. When used appropriately,
these predictive algorithms can be powerful tools in
the hunt for disease-causing polymorphisms.
Regulatory polymorphisms
Predictive algorithms have been developed less suc-
cessfully for noncoding polymorphisms. The reasons
for the lack of predictive algorithms for identifying
regulatory noncoding polymorphisms are the relative
scarcity of training data for disease-associated SNPs
falling outside protein coding regions, and the relative
ease of assigning predictive attributes, such as amino
acid conservation and structural features of proteins,
to protein coding SNPs as compared to noncoding
40
SNPs. However, some computational strategies
focused on the identification and prediction of the
functional effects of nucleic acid substitutions within
transcription factor binding sites or other regulatory
sites have been developed. A recent example is RAVEN
(Regulatory Analysis of Variation in Enhancers), which
uses a combination of features, including known tran-
scription factor binding sites, to predict which SNPs
may fall into functionally important regulatory regions
(http://burgundy.cmmt.ubc.ca/cgi-bin/RAVEN/a?rm=
home) [35]. A series of tools are available from BIOBASE
(http://www.gene-regulation.com/index.html), such as
MATCH [36], P-MATCH [37], and AliBaba2 [38], all
of which identify transcription factor binding sites
in a variety of ways. These methods are based upon
information in the TRANSFAC database [39], a
proprietary database of regulatory regions. Another
database containing information about transcription
regulatory regions, rather than individual transcrip-
tion factor binding sites, is the TRRD database
(http://wwwmgs.bionet.nsc.ru/mgs/gnw/trrd/) [40].
Many of the previous tools rely upon identifying
known regulatory factor binding sites or regions.
However, in some cases the investigator may be inter-
ested in identifying novel regulatory motifs which
may control a set of genes of interest. In this case,
motif search tools have been made available to iden-
tify de novo motifs within regions of interest. Two
tools, MEME [41] and MAST [42] (http://meme.sdsc.
edu/meme/intro.html), can either be used to search
for a known motif in sequence databases, or discover
motifs within a specific set of sequences. Similar web-
based tools are available, such as SCOPE (http://genie.
dartmouth.edu/scope/) [43], Cluster-Buster (http://
zlab.bu.edu/cluster-buster/) [44], and MoD Tools
(http://159.149.109.9/modtools/) [45]. This list does
not represent a comprehensive set of the available
tools by any means.
Much, if not all, of the publicly available infor-
mation on transcription factor binding sites, and
other regulatory regions, is available on the University
of California, Santa Cruz (UCSC) Genome Browser
(http://genome.ucsc.edu/; [46]) (Figure 4.1). A number
of different tracks relevant to gene regulation are avail-
able. Information from the ENCODE regions, regions
of the genome which have been heavily characterized
in terms of transcription factor binding sites, DNA
hypersensitivity, and epigenetic modifications, among
other characteristics, is available at the UCSC Genome
Browser [47]. Some of the elements within the

Export
sequences
Switch
genomes
Search
sequences
Export
data
Rearrange
features
Figure 4.1 University of California, Santa Cruz (UCSC) genome browser layout with key features highlighted.
ENCODE regions have been expanded beyond those
regions to the whole genome. For example, transcrip-
tion start sites as determined by luciferase promoting
activity have been identified for nearly the whole
genome [48]. Other tracks on the UCSC Genome
Browser include but are not limited to the identification
of CpG islands, conserved transcription factor
binding sites, microRNA sites, sites with regulatory
potential, and regulatory factor binding sites as deter-
mined by chromatin immunoprecipitation. A track of
note, which combines many of the previous tracks
and provides the type of evidence used for each assign-
ment is the ORegAnno track (http://www.oreganno.
org/oregano/Index.jsp) [49]. ORegAnno is an open
access community driven resource for the annotation
of regulatory elements which is consistently updated
with new findings by the scientific community.
Another track of note, which is available at the UCSC
Genome Browser as well as at a standalone site, is the
VISTA Enhancer Browser (http://enhancer.lbl.gov/)
[50]. This resource contains experimentally validated
human noncoding fragments with gene enhancer
activity as assessed in transgenic mice. These enhancers
correspond to regions of extreme conservation across
vertebrates.
Conservation can be a useful tool in identifying
many different types of functional elements in the
human genome. The UCSC Genome Browser
Chapter 4: In silico analysis strategies and resources
Navigate the genome
View
genomic
features
Select
features to
view
contains a number of tracks which describe sequence
conservation at a variety of levels. Sequence conservation
can be queried across 28 vertebrate species or across
17 placental mammal species. Regions are denoted by a
conservation percentage and ultra-conserved elem-
ents are highlighted in a separate track.
Motifs, conservation, and regulatory polymorph-
isms are all interrelated. When a regulatory motif,
or simply a motif of interest, is identified, an investi-
gator may be interested in determining what nucleo-
tides of the motif are conserved across all instances
of that motif in a set of sequences of interest.
Nucleotides conserved across motifs are likely to be
important in mediating the function of that motif.
A useful visualization tool for sequence motifs of
any sort, nucleotide or protein, is WebLogo (http://
weblogo.berkeley.edu/) [51]). WebLogo takes as its
input a series of alignments and generates a graphic
where the heights of each character at each position
represent its preponderance at that position. This
provides a very simple and powerful tool for identi-
fying functionally important nucleotides or amino
acids within a motif.
Alignments
Sequence alignments may be generated by many of
the above tools; however they are not generated in all
41
 Chapter 4: In silico analysis strategies and resources
cases. Sequence alignments are a robust tool used for
many purposes, the most common of which is the
identification of similar stretches of sequence across
DNA or protein regions. Similar stretches of sequence
are identified by arranging sequences in such a way
that equivalent characters, single letter representa-
tions of nucleotides or amino acids, are aligned.
Gaps, or missing chunks of sequence, are generally
represented by a “-” character. Alignment allows the
investigator to compare equivalent positions in differ-
ent sequences, to identify patterns occurring across
the sequences.
Alignments fall into two general categories: global
alignments and local alignments. Global alignments
are used to determine the best overall alignment for
the entirety of a set of sequences. This approach is
generally applicable when an investigator knows of an
a priori similarity between a set of sequences, and
would like to identify corresponding residues across
sequences. For example, global alignments would be
useful when aligning sequences from different indi-
viduals but from the same gene in order to identify
sequence variations between the individuals. Local
alignments are used to find regional similarities
between two or more sets of sequences. For example,
a set of sequences that share a protein domain but
contain different accessory domains would be subject
to local alignment.
Pairwise alignments are generally local alignments
that span only two sequences. These alignments are
efficient and quick to calculate, given that the two
input sequences are of sufficient similarity. They are
calculated in two major ways, dynamic programming
algorithms or word methods. Dynamic programming
algorithms, the Needleman–Wunsch algorithm or
Smith–Waterman algorithm, identify the optimal
alignment by calculating scores for all possible align-
ments and retracing a path which selects the highest
scoring alignment. Though dynamic programming
algorithms are more exact, they are also very compu-
tationally intensive and excessive for general pur-
poses. Implementations of these algorithms are
available as software packages, such as JAligner
(http://jaligner.sourceforge.net/). Word methods, on
the other hand, are extremely efficient alignment tools,
but are not guaranteed to find the optimal solution.
Word methods search the input sequences for short
subsequences of high similarity, and evaluate only
matched pairs of sequences with high similarity. Two
major implementations of word methods are NCBI
42
Blast (http://blast.ncbi.nlm.nih.gov/Blast.cgi) [52], and
the EBI tool, FASTA (http://www.ebi.ac.uk/Tools/
fasta33/index.html) [53]. For most purposes, word
methods are sufficiently accurate. NCBI Blast is also
useful for finding similar genes, or stretches of DNA,
to a sequence of interest. The investigators’ sequence of
interest can be used to query databases for similar
sequences contained within large databases.
Multiple alignments present a much more chal-
lenging problem, in which one attempts to find the
optimal alignment for numerous sequences. Multiple
alignments are useful for discovering highly con-
served elements, for example catalytic site residues
or other types of conserved sequence motifs. Dynamic
programming algorithms can be used to tackle this
problem; however, because of the exponentially
increasing complexity as the number of sequences is
increased, such approaches are generally limited to a
very small number of sequences. NCBI has a multiple
sequence alignment package available [54]. The major
approaches to multiple sequence alignments include
the progressive approach or motif finding approach.
In the progressive approach, the two most related
sequences are aligned, and the resultant alignment is
used as a sequence for the next pairwise alignment.
This process is iterated until all sequences are
aligned. A number of different implementations
using slightly different approaches exist, including
ClustalW (http://www.ebi.ac.uk/Tools/clustalw2/) [55],
T-Coffee (http://www.ebi.ac.uk/Tools/t-coffee/) [56],
and MUSCLE (http://www.ebi.ac.uk/Tools/muscle/)
[57]. Motif-based alignments attempt to discover
short sequences of high similarity that are common
elements of all the sequences being aligned. Pairwise
alignments guided by these common motifs are exe-
cuted. Then the pairwise alignments to the common
motifs are combined to provide the resultant overall
alignment. One example uses Gibbs sampling to gen-
erate motifs for pairwise alignments [58]. A large
number of tools exist for generating pairwise or
multiple alignments, some of which are provided in
a web interface by EBI (http://www.ebi.ac.uk/Tools/
sequence.html).
Local alignments, in contrast, are used to identify
short spans of similarity shared by two or more
sequences. This approach is useful when a shared
short element is suspected to be contained within
a set of sequences. This shorter element may be
surrounded by stretches of sequence which differ
greatly across all the sequences being compared.

For example, an investigator may use a local align-
ment search to attempt to find a short transcription
factor binding site motif within the long upstream
promoter sequence of a gene of interest. The
EMBOSS pairwise alignment tool at EBI can be set
to generate local alignments (http://www.ebi.ac.uk/
Tools/psa).
Though sequence similarity is indicative of func-
tional similarity, it is still possible that disparate pro-
tein sequences may result in similar protein folds and
similar structure and function. The so-called “twilight
zone” involves comparison of sequences with less
than 25% sequence identity. At this point, sequence-
based alignments are extremely difficult to generate
with high confidence. There are numerous examples
of proteins with similar, or the same, function, with
low sequence identity. In this case, structural align-
ments become useful tools. Though the sequences of
two related proteins may differ to the point where
their similarities are almost unrecognizable, proteins
of similar function tend to fold into a very similar
three-dimensional shape. Structural alignment algo-
rithms take advantage of solved three dimensional
structures to find the closest structural superposition
of the two proteins. That is, the two structures are
superimposed so that the distance between “equiva-
lent” amino acids in the two structures is minimized.
Corresponding amino acids can be assigned based
upon the closest amino acid pairs across the two
structures to define a sequence alignment. Numerous
structural alignment tools are available, including
EBI Dali (http://www2.ebi.ac.uk/dali/) [59] and NCBI
VAST (http://www.ncbi.nlm.nih.gov/Structure/VAST/
vastsearch.html) [60].
Protein structures
Structural alignments require as their input, protein
structure files. The WorldWide Protein Data Bank
(http://www.wwpdb.org/) is a federation of organ-
izations which act as deposition, data processing,
and distribution centers for protein structural infor-
mation. The RCSB PDB (http://www.rcsb.org/pdb/
home/home.do) contains a simple web interface for
performing searches for an investigator’s protein of
interest. More advanced searches by sequence,
motifs, structural domains, and a whole host of other
filtering parameters are also possible. PDB informa-
tion can be accessed by numerous web portals.
Another example is the PDBe by EBI (http://www.
Chapter 4: In silico analysis strategies and resources
ebi.ac.uk/pdbe/) [61]. A tutorial for E-MSD is avail-
able on the web (http://www.ebi.ac.uk/pdbe/docs/
roadshow_tutorial/).
Protein functional sites and domains
Protein structures are not necessary for the identifica-
tion of functional sites, such as catalytic sites, or
functional domains. Proteins contain independently
folding functional regions, called functional domains,
which perform specific biochemical functions as inde-
pendent subunits of a whole protein. Domains of
importance can be identified by patterns of physio-
chemical properties; for example, transmembrane
domains tend to contain long stretches of hydropho-
bic residues. Important functional sites can also be
identified by particular amino acid patterns; for
example, post-translational modification sites tend
to be defined by specific amino acid motifs. These
functional domains or sites can be identified by
searching the amino acid sequence for particular pat-
terns indicative of the presence of a functional
domain. These patterns are defined in a number of
different ways including basic amino acid patterns in
PROSITE (http://prosite.expasy.org/) [62], groups of
motifs, or fingerprints, in PRINTS [63], or hidden
markov models in Pfam (http://pfam.sanger.ac.uk/) [64],
among others. EBI provides an integrated tool, Inter-
ProScan (http://www.ebi.ac.uk/Tools/pfa/iprscan/)
[65], which provides the above tools as well as numer-
ous other functional annotation tools in a single
web-based query system.
In additional to functional sites, proteins may
contain targeting signals which determine the cellular
distribution of some proteins. These signal peptides
tend to be the most N-terminal amino acids and
contain specific amino acid patterns, patterns of
physiochemical properties, or a combination of the
two, which determine what cellular compartment the
proteins should be sorted to. The Center for
Biological Sequence Analysis, at the Technical Uni-
versity of Denmark, contains a suite of prediction
servers for bacterial, plant and animal signal
sequences (http://www.cbs.dtu.dk/services/). In add-
ition, the DTU prediction servers contain a series of
post-translational modification site prediction
algorithms that can perform specific post-transla-
tional modification predictions (e.g. phosphorylation,
O-glycosylation and N-glycosylation) similar to some
of the predictions provided by the PROSITE tool.
43
 Chapter 4: In silico analysis strategies and resources
Expression
If a gene or genetic variation within a gene is identified
via a linkage or association study, a logical question
might be whether that gene is expressed in a tissue of
relevance to the disease under study (e.g. the brain, in
neuropsychiatric disease). Gene expression datasets are
scattered widely throughout a large number of gene
expression data repositories. There are databases and
resources that provide information on the expression
patterns of genes, but these are limited by the tissues
and assays studied. Some of the most relevant data-
bases for neuropsychiatric disease include TStag
(http://tstag.molgen.mpg.de/), BioGPS (http://biogps.
gnf.org/#goto=welcome), and the evolving Allen Brain
Atlas (http://www.brain-map.org/).
Other gene expression resources harbor actual
gene expression datasets that can be mined and used
in meta-analyses. The major gene expression reposi-
tories are the Gene Expression Omnibus (http://www.
ncbi.nlm.nih.gov/geo/) (GEO) [66], maintained by
NCBI, ArrayExpress (http://www.ebi.ac.uk/arrayex-
press) [67], and the Center for Information Biology
Gene Expression Database (http://ecoliwiki.net/
colipedia/index.php/Center_for_Information_Biology_
Gene_Expression_Database_(CIBEX)) [68]. Data
adheres to the “Minimum Information about a Micro-
array Experiment” guidelines, which should guarantee
access to adequate information to conduct additional
analyses using third party generated data. The infor-
mation generally presented is the experimental design,
array design, samples, hybridization protocol, measure-
ments, and normalization controls [69]. The GEO data-
base allows you to either browse through the available
information, or search it using keywords, experiment
identifiers (known as the GEO accession, which is
generally published within the original manuscript
describing the experiment), or platform identifiers cor-
responding to the type of microarray used.
An emerging set of resources and databases
for gene expression analysis of relevance to genetic
studies of neuropsychiatric diseases are “expression
quantitative trait locus” or “eQTL” resources. These
databases harbor information on the results of associ-
ation studies that identify genetic variations that
influence the expression levels of particular genes; this
information may be useful in arguing that a particular
variation influences the regulation of a particular
protein rather than the protein’s structure. Many of
these databases are severely limited by the tissues for
44
which the expression assays have been performed
as well as sample size. One of the best, however, is
the SNPExpress database (http://people.genome.duke.
edu/dg48/SNPExpress).
Pathway analysis
Genes and proteins do not act in isolation within a
cell, and one can use information about their inter-
actions to choose candidate SNPs for a study and to
help interpret the significance of a study’s results. For
example, one can apply gene set expression analysis
(GSEA) [70, 71] to microarray data to determine the
significance of perturbations to an entire pathway
of interest, which can be more powerful than com-
paring expression levels of individual genes. Pathway
analysis may also help explain the connections between
significant results in GWAS. Torkamani et al. showed
that SNPs in the Wellcome Trust study that did not
meet genome-wide significance for disease association
were disproportionately found within particular path-
ways [72]. Some of these results were expected, such as
that SNPs associated with type I diabetes were in
immune response pathways, but other pathways may
be novel targets for experimental validation.
There are three major sources of information
about biological pathways in the cell – single experi-
ments, high-throughput assays, and bioinformatics –
and each technique has its trade-offs. First, cell and
molecular biological experiments over the past few
decades have provided detailed data about many meta-
bolic pathways. These studies provide a gold standard
for accuracy, but they are time-consuming and hence
are not comprehensive. More recently, high through-
put experiments have captured a more global picture
of gene and protein interactions within the cell at the
cost of some false positives. These data include, for
example, binding affinities of thousands of pairs of
proteins, enhancer and transcription factor binding
sites, and the effects of noncoding RNAs such as
microRNAs. Bioinformatic tools can identify patterns
in this high-throughput data in order to predict add-
itional interactions. For example, a whole genome
search for microRNA nucleotide binding motifs can
help predict which gene transcripts will be degraded by
these microRNAs. Bioinformatic predictions are the
least reliable sources of biological interactions, and
they must eventually be verified in living cells, but they
can help guide experiments to elucidate metabolic
pathways about which we know surprisingly little [73].

Pathway data repositories
While there are many online repositories of specific
types of interaction data (e.g. PIPs [74], UniProbe
[75, 76], VISTA), there are relatively few sites that
attempt to combine this information together into
pathways. This merging has been facilitated by the
adoption of standard data formats. Three of the most
popular formats are BioPAX, SBML and PSI-MI,
which are appropriate for different types of pathway
data [77]. Pathguide (http://www.pathguide.org/)
lists many pathway data sites and the formats that
they support, and it is a good place to look for
specific datasets, such as kinase pathways or HIV/
host interactions [78]. However, it may make sense to
start with a more comprehensive database of pathway
information.
The Gene Ontology (GO) (http://www.geneontology.
org/) database describes the cellular components,
molecular functions, and biological processes of gene
products with a species-independent vocabulary (The
Gene Ontology Project, 2006). For example, a search
for the neuron receptor GluR2 would locate it in
the plasma membrane, and associate it with AMPA/
kainate receptor activity, signal transduction, and syn-
aptic transmission. These terms exist in a hierarchical
tree, and so, for example, synaptic transmission is a
type of cell-cell signaling, which is a type of cell
communication, which is a cellular process, and so
on. One can easily search for all genes associated with
GO terms across species or vice versa, which is a
powerful way to look for common functional themes
in a set of SNPs and their associated genes.
The Biomolecular Object Network Databank (BOND)
(http://bond.unleashedinformatics.com/) includes path-
way information from the Biomolecular Interaction
Network Databank (BIND) [79] and a comprehensive,
annotated list of patented DNA sequences from
GENESEQ. BIND has information about more than
200 000 interactions from over 1500 organisms, many
of which have been manually curated from the litera-
ture. GENESEQ allows one to screen, for example, drug
targets for potential patent infringements world-wide.
Thomson Reuters also offers a pay version called
BONDplus that is updated more frequently than BIND
(http://bond.unleashedinformatics.com/).
Reactome (http://www.reactome.org/) makes it
easy to browse for specific pathways, although it has
a less comprehensive database of interactions than
BOND [80]. Each pathway is contributed by experts
Chapter 4: In silico analysis strategies and resources
in their fields, cross-referenced to sequence databases
such as NCBI and Ensembl, and finally checked by
the editorial board. Contributing authors and review-
ers are listed with each pathway, which may increase
the accuracy of the entries. Pathways also include a
short description of their biological role as well as the
interacting components.
Kyoto Encyclopedia of Genes and Genomes
(KEGG) (http://www.kegg.jp/) is perhaps the most
comprehensive free resource of pathway information
available [81, 82]. KEGG was started in 1995 and is
now composed of 19 interlinked databases in three
categories: genomic, chemical, and systems. The
GENE database contains genomic information about
the building blocks of the cell – genes and proteins.
The COMPOUND and DRUG databases contain
chemical information about endogenous and exogen-
ous substances, such as enzymes and pharmaceuti-
cals approved in the United States and Japan.
PATHWAY has interactions between these genetic
objects that form the molecular wiring diagram
of a cell. Finally, BRITE ties together the other data-
bases and gives higher level information about
interactions between genes, proteins and the environ-
ment. Much of this information is manually curated
or annotated data from other public databases such
as NCBI’s RefSeq.
Pathway visualization and analysis
Pathways in a cell are better visualized as a network
rather than a list of parts. While the aforementioned
sites have some visualization features, there are a
plethora of more powerful tools available. Suderman
and Hallett recently compared the features of 35 of
these tools [83], and we will highlight a couple that
balance power with flexibility and ease of use.
Cytoscape (http://www.cytoscape.org/) is a free tool
that can display and analyze large (> 100 000 nodes)
molecular interaction networks derived from many of
the above resources, including KEGG and Reactome
[84, 85]. Researchers have written almost 100 plugins
for it that extend its functionality, and many of the
network display options are user customizable. Cytos-
cape allows one to connect gene nodes into a network
based on multiple interaction types, including pro-
tein–protein and protein–DNA. Gene expression data
can be overlaid on this network by coloring genes
based on their fold change in expression. Visual
inspection of the network can reveal clusters of
45
 Chapter 4: In silico analysis strategies and resources
interconnected genes that show similar expression
changes, and the jActiveModules plugin automates this
process. A search within these sub-networks may reveal
hub genes that regulate the function of neighboring
nodes. The BiNGO plugin lists biological processes
from the Gene Ontology database that are significantly
over-represented within the network [86]. Another
useful plugin, Cerebral, will alter the network layout
so that each gene node is in the appropriate sub-cellular
location [87]. This makes the co-regulation of sub-
networks more plausible by showing that the gene
products could physically interact within the cell.
Many of these network analyses can be accom-
plished in other tools. BiologicalNetworks (http://
www.biologicalnetworks.net/) is the visualization inter-
face for the PathSys database, which contains much of
the same information as KEGG. PathSys has a more
flexible architecture than Cytoscape and can incorpor-
ate a broader range of molecular data and interactions
[88]. It also has a powerful query language called
BioNetSQL that can be used for advanced analysis of
a network. Another tool worth trying is VisANT
(http://visant.bu.edu/), which supports large networks
and plugins like Cytoscape, but it makes it easier
to integrate data from different sources [89, 90]. Gen-
MAPP (http://www.genmapp.org/) includes thousands
of hand curated pathway MAPPs and will soon use
Cytoscape for network visualization [91]. PATIKAweb
(http://www.cs.bilkent.edu.tr/patikaweb/) allows one
to quickly visualize a pathway using a pre-built data-
base of interactions derived from numerous free
sources such as Reactome, NCBI, and GO [92].
Commercial pathway analysis tools benefit from
access to high-quality databases covering a wide range
of molecular interactions, but their lack of plugin
support makes them less flexible. Ariadne Genomics
Pathway Studio (http://www.ariadnegenomics.com/
products/pathway-studio/) combines data from public
resources such as KEGG with data mined from current
biomedical literature [93]. Their MedScan technology
applies natural language processing to the full text of
journal articles retrieved from PubMed in order to
extract interactions between biological objects, such
as genes, proteins, or drugs [94]. MedScan allows
researchers to easily build targeted data content for a
specialized field that may not be available in conven-
tional pathway databases.
Ingenuity Pathway Analysis (http://www.ingenuity.
com/) and GeneGo MetaCore (http://www.genego.
com/) are two commercial tools that have generated
46
proprietary pathway databases based on manual
curation of full text journal articles. This keeps the
information current and automatically provides ref-
erences for all the interactions. Like PATIKAweb,
MetaCore is a web application, and so it uses the most
current biological knowledge. MetaCore has a higher
quality database than free tools such as Cytoscape,
and it provides many of the same visualization and
analysis features. In addition, it facilitates import of
many experimental data formats, has flexible options
for network generation, and allows functional annota-
tion based on a customized version of the free GO
database.
Pathway-based analysis is already a useful biological
tool, and it will become much more powerful as more
interaction data is collected between a broad range of
objects within a cell, including proteins, DNA, RNA,
and small molecules. These interactions are dynamic
processes that should be classified by direction, stabil-
ity, sub-cellular compartment, developmental time
point, and many other attributes [95]. Visualization
tools must be developed that can display this context
dependence of pathways in a way that highlights the
biological significance of experimental data.
Conclusions
As the costs and labor associated with genotyping
and sequencing are reduced through technological
developments, more and more research investigating
the role of inherited DNA sequence variations in
disease pathogenesis will be pursued. It is likely that
these investigations will rely on family-based linkage
strategies, genotyping-based GWA study strategies,
and sequencing-based strategies, but will leverage a
more complete understanding of the full array of
variations that populate the genome. Given that more
variations will be interrogated in future investiga-
tions, the need to leverage biological and functional
annotations of genes and genetic variations will
become more pronounced in order to make sense of
potential linkages and associations. There is growing
precedent with such approaches. For example, The
International Schizophrenia Consortium recently
pursued a study in which they found a greater
number of deletions in the genomes of individuals
with schizophrenia than in the genomes of controls
[13]. However, since the same deletions were not seen
across all schizophrenic individuals, the researchers
needed to determine whether the various deletions
 Chapter 4: In silico analysis strategies and resources

possessed by the schizophrenic subjects were likely
to disrupt genes with similar biological effect. With-
out this link between the deletions, the association
involving a greater number of deletions among
schizophrenics would not have been compelling from
a biological standpoint, and might have been dis-
missed as a false positive. In addition, Torkamani
et al. [72] examined the biological pathways influ-
enced by variations exhibiting weak associations with
a number of traits and showed that the genes and
pathways harboring weakly associated variations
were biologically meaningful and not likely to have
occurred by chance.
Although the pursuit of genetic studies exploiting
as much biological information about genes and
genetic variations as possible are logical, they are
obviously limited by the community’s incomplete
knowledge of gene function and genome biology.
However, this limitation will be overcome over time
in much the way inefficiencies in genotyping and
sequencing technologies were overcome via techno-
logical developments. Large-scale efforts to disclose
functional elements in the genome, such as the
ENCODE project [96] and the Human Epigenome
Project [97], along with attempts to develop uniform
standards in referring to and assessing gene function
and nomenclature [98], will lead to resources that will
facilitate more integrated approaches to the genetic
analysis of human neuropsychiatric diseases.
Acknowledgements
The authors are supported in part by the following
research grants: the National Institute on Aging
Longevity Consortium (grant number U19 AG023122–
01); the National Institute of Mental Health (NIMH)
-funded Genetic Association Information Network
Study of Bipolar Disorder National (grant number
1 R01 MH078151–01A1); National Institutes of
Health grants (grant numbers N01 MH22005, U01
DA024417–01, P50 MH081755–01); the Scripps
Translational Sciences Institute Clinical Translational
Science Award (grant number U54 RR0252204–01),
the Price Foundation and Scripps Genomic Medicine.

References 13. International Schizophrenia

1. Carlson CS, et al. Nature 2004;
429(6990):446–452.
2. Lander ES, et al. Science 1994;
265(5181):2037–2048.
3. Manolio TA, et al. J Clin Invest
2008;118(5):1590–1605.
4. Elston RC. Stat Methods Med Res
2000;9(6):527–541.
5. Risch N, et al. Nat Genet 1996;
12(4):351–353.
6. Risch N, et al. Science 1996;
273(5281):1516–1517.
7. Frazer KA, et al. Nat Rev Genet
2009;10(4):241–251.
8. Manolio TA, et al. Nature 2009;
461(7265):747–753.
9. Schork NJ, et al. Curr Opin Genet
Dev 2009;19(3):212–219.
10. Ng SB, et al. Nature 2009;
461(7261):272–276.
11. Ng SB, et al. Nat Genet 2010;
42(1):30–35.
12. International Schizophrenia
Consortium. Nature 2008;
455(7210):237–241.
Consortium. Nature 2009;
460(7256):748–752.
14. Baxevanis AD. Curr Protoc
Bioinformatics; 2006:Chapter 1:
Unit 1.1.
15. Plumpton M, et al. In Barnes
MR (ed.). A Bioinformatics Primer
for the Analysis of Genetic Data,
2nd edition. New York: John
Wiley and Sons; 2007.
16. Zweig AS. Genomics 2008;
92(2):75–84.
17. Birney E, et al. Genome Res
2004;14:988–995.
18. McKusick VA. Am J Hum Genet
2007;80:588–604.
19. Stenson PD, et al. Hum Mutat
2003;21:577–581.
20. Sherry ST, et al. Nucleic Acids Res
2001;29:308–311.
21. The International HapMap
Consortium. Nature 2003;
426(6968):789–796.
22. Mailman MD, et al. Nat Genet
2007;39:1181–1186.
23. Ng PC, et al. Annu Rev Genomics
Hum Genet 2006;7:61–80.
24. Ng PC, et al. Genome Res 2002;
12(3):436–446.
25. Thomas PD, et al. Proc Natl
Acad Sci U S A 2004;101(43):
15398–15403.
26. Sunyaev S, et al. Hum Mol Genet
2001;10(6):591–597.
27. Ferrer-Costa C, et al.
Bioinformatics 2005;21(14):
3176–3178.
28. Yue P, et al. BMC Bioinformatics
2006;7:166.
29. Nakken S, et al. Neuroscience
2007;145(4):1273–1279.
30. Torkamani A, et al. Bioinformatics
2007;23(21):2918–2925.
31. Xue D, et al. J Hum Genet 2008;
53(5):379–389.
32. Witten IH, et al. Data Mining:
Practical Machine Learning Tools
and Techniques, 2nd edition. San
Francisco: Morgan Kaufmann; 2005.
33. Tchernitchko D, et al. Clin Chem
2004;50(11):1974–1978.

47
 Chapter 4: In silico analysis strategies and resources
34. Care MA, et al. Bioinformatics
2007;23(6):664–672.
35. Andersen MC, et al. PLoS Comput
Biol 2008;4(1):e5.
36. Kel AE, et al. Nucleic Acids Res
2003;31(13):3576–3579.
37. Chekmenev DS, et al. Nucleic
Acids Res 2005;33(Web Server
issue):W432–W437.
38. Grabe N. In Silico Biol 2002;
2(1):S1–S15.
39. Wingender E, et al. Nucleic Acids
Res 1996;24(1):238–241.
40. Wingender E, et al. Nucleic Acids
Res 1997;25(1):265–268.
41. Timothy L. Bailey, et al. Nucleic
Acids Res 2006;34:W369–W373.
42. Timothy L. Bailey, et al.
Bioinformatics 1998;14:48–54.
43. Carlson JM, et al. Nucleic Acids
Res 2007;35(Web Server issue):
W259–W264.
44. Frith MC, et al. Nucleic Acids Res
2003;31(13):3666–3668.
45. Pavesi G, et al. Nucleic Acids Res
2006;34(Web Server issue):
W566–W570.
46. Kent WJ, et al. Genome Res
2002;12(6):996–1006.
47. Thomas DJ, et al. Nucleic Acids
Res 2007;35(Database issue):
D663–D667.
48. Kim TH, et al. Genome Res
2005;15(6):830–839.
49. Montgomery SB, et al.
Bioinformatics 2006;22(5):
637–640.
50. Visel A, et al. Nucleic Acids Res
2007;35(Database issue):D88–D92.
51. Crooks GE, et al. Genome Res
2004;14(6):1188–1190.
52. Altschul SF, et al. J Mol Biol
1990;215:403–410.
53. Pearson WR. Methods Mol Biol
1994;25:365–389.
54. Lipman DJ, et al. Proc Natl Acad
Sci U S A 1989;86:4412–4415.
55. Edgar RC. BMC Bioinform
2004;5:113.
48
56. Notredame C, et al. J Mol Biol
2000;302(1):205–217.
57. Higgins D, et al. Nucleic Acids
Res 1994;22:4673–4680.
58. Neuwald AF, et al. Protein Sci
1995;4:1618–1632.
59. Holm L, et al. Science
1996;273:595–603
60. Madej T, et al. Proteins
1995;23:356–369.
61. Golovin A, et al. Nucleic Acids
Res 2004;32(Database issue):
D211–D216.
62. Sigrist CJA, et al. Brief Bioinform
2002;3:265–274.
63. Attwood TK. Brief Bioinform
2002;3:252–263.
64. Finn RD, et al. Nucleic Acids Res
2006;34:D247–D251.
65. Zdobnov EM, et al. Bioinformatics
2001;17:847–848.
66. Edgar R, et al. Nucleic Acids Res
2002;30:207–210.
67. Parkinson H, et al. Nucleic Acids
Res 2007;35:D747–D750.
68. Ikeo K, et al. C R Biol
2003;326:1079–1082.
69. Brazma A, et al. Nat Genet 2001;
29:365–371.
70. Subramanian A, et al. Proc Natl
Acad Sci U S A 2005;102(43):
15545–15550.
71. Mootha VK, et al. Nat Genet
2003;34(3):267–273.
72. Torkamani A, et al. Genomics
2008;92(5):265–272.
73. Duarte NC, et al. Proc Natl Acad
Sci U S A 2007;104(6):1777–1782.
74. McDowall MD, et al. Nucleic Acids
Res 2008;6:6.
75. Philippakis AA, et al. J Comput
Biol 2008;15(7):655–665.
76. Berger MF, et al. Nat Biotechnol
2006;24(11):1429–1435.
77. Stromback L, et al. Bioinformatics
2005;21(24):4401–4407.
78. Bader GD, et al. Nucleic Acids
Res 2006;34(Database issue):
D504–D506.
79. Bader GD, et al. Nucleic Acids Res
2001;29(1):242–245.
80. Joshi-Tope G, et al. Nucleic Acids
Res 2005;33(Database issue):
D428–D432.
81. Aoki-Kinoshita KF, et al. Methods
Mol Biol 2007;396:71–91.
82. Kanehisa M, et al. Nucl Acids
Res 2008;36(Database issue):
D480–D484.
83. Suderman M, et al. Bioinformatics
2007;23(20):2651–2659.
84. Cline MS, et al. Nat Protoc 2007;
2(10):2366–2382.
85. Shannon P, et al. Genome Res
2003;13(11):2498–2504.
86. Maere S, et al. Bioinformatics
2005;21(16):3448–3449.
87. Barsky A, et al. Bioinformatics
2007;23(8):1040–1042.
88. Baitaluk M, et al. BMC Bioinform
2006;7:55.
89. Hu Z, et al. Brief Bioinform 2008;9
(4):317–325.
90. Hu Z, et al. Nucleic Acids Res
2005;33(Web Server issue):
W352–W357.
91. Dahlquist KD, et al. Nat Genet
2002;31(1):19–20.
92. Demir E, et al. Bioinformatics
2002;18(7):996–1003.
93. Nikitin A, et al. Bioinformatics
2003;19(16):2155–2157.
94. Novichkova S, et al.
Bioinformatics 2003;19(13):
1699–1706.
95. Beyer A, et al. Nat Rev Genet
2007;8(9):699–710.
96. ENCODE Project Consortium,
et al. Nature 2007;447(7146):
799–816.
97. American Association for Cancer
Research Human Epigenome
Task Force, et al. Nature 2008;
454(7205):711–715.
98. Genome Group of the Gene
Ontology Consortium. PLoS
Comput Biol 2009;5(7):
e1000431.
 Gene expression studies in psychiatric
Chapter
5 disorders
Alexander B. Niculescu, III
Identifying genes for psychiatric disorders through
classic genetic approaches has proven arduous, des-
pite some recent successes mentioned in this book.
This is due to the likely complex, polygenic and het-
erogeneous nature of these disorders – multiple genes,
multiple polymorphisms in those genes, in different
combinations of variable penetrance – involved in
different subtypes of the illnesses. As a consequence,
most genome-wide association studies to date have
been underpowered to detect the full complement of
genes and polymorphisms, or even a small portion of
it [1–3]. The imprecise nature of broad psychiatric
phenotypes has also been a major rate limiting step
[4–6], and is the subject of study of a new field,
psychiatric phenomics [7, 8]. In addition to complex-
ity and heterogeneity, there is also a growing appreci-
ation of the genetic, neurobiologic, and phenotypic
overlap and interdependence of various major neuro-
psychiatric disorders [6, 9–12]. The use of case–case
designs, subphenotypes, and boot-strapping with other
lines of work – neurophysiology [13], imaging [14],
and animal models [15–17], may provide for an accel-
erated pace of gene identification in the years to come.
The completion of the sequencing of the human
genome and that of other model organisms, coupled
with the advent of microarray technology over the last
decade, have made large-scale gene expression studies
scientifically and economically feasible. After some
initial debate about different microarray platforms
[18], there is an emerging consensus that different
platforms perform with similar accuracy and reliabil-
ity if employed well [19]. The main room for
improvement has been and is at the level of designing
appropriate biological experiments, and integrating
multiple independent lines of evidence in a Bayesian
fashion. Human postmortem brain gene expression
profiling studies from subjects with neuropsychiatric
Principles of Psychiatric Genetics, eds John I. Nurnberger, Jr. and Wade H. Berrettini. Published by Cambridge University
Press. © Cambridge University Press 2012.
disorders have produced interesting leads [20, 21].
However, this important line of work, if pursued by
itself, suffers from multiple caveats [22] – genetic
variability, difficulty of building large enough cohorts,
uncertainty about exact pre-mortem diagnosis, agonal
artifacts [23], impact of comorbid medical conditions,
and the potential effects of environmental variables
(medications, drugs of abuse, stress, nutrition) on
brain gene expression changes. Animal model gene
expression studies avoid these caveats, but suffer from
the potential limited relevance of the animal model
used to the human condition [24].
A combined approach, termed convergent func-
tional genomics (CFG) (Figure 5.1), which integrates
genetic and gene expression data, in humans and
animal models, has been developed as a way of
avoiding the limitations of the individual approaches
mentioned, and reinforcing their strengths in a Baye-
sian fashion [25]. This approach has been applied
with some success to bipolar disorder [16, 17], alco-
holism [26], and schizophrenia [11]. Candidate genes
identified by such an approach can be pursued in
a prioritized fashion to obtain additional unambigu-
ous evidence for involvement in the illness, through
human candidate gene association studies and human
transgenic mouse studies. The list of prioritized genes
identified by approaches such as CFG also provides
testable hypotheses for epistatic interactions among
the co-expressed genes [16, 25]. Last, but not least, a
combination of CFG with the emerging and growing
large genome-wide association datasets may be
particularly powerful in breaking the genetic code of
psychiatric disorders [27, 28].
More recently, there has been renewed interest in
identifying peripheral correlates of neuropsychiatric
disorders, termed biomarkers. There are to date no
well established, specific clinical laboratory blood tests
49
 Chapter 5: Gene expression studies
Expanded convergent functional genomics:
Multiple independent lines of evidence for Bayesian cross-validation
Animal model studies
pharmacogenomic, transgenic, inbred strains
Animal model genetic evidence
QTL or transgenics
Animal model brain Candidate
evidence gene/
biomarker
Animal model blood
evidence
Sensitivity
for psychiatric disorders. Given the complex nature of
psychiatric disorders, the current reliance on patient
self-report of symptoms and the clinician’s impres-
sion on interview of patient is a rate limiting step in
delivering the best possible care with existing treat-
ment modalities, as well as developing new and
improved treatment approaches, including new medi-
cations. Identifying molecules in the blood that reflect
illness in the brain would be a major advance. These
molecules could be used to develop clinical laboratory
tests to aid: (1) diagnosis of illness; (2) early interven-
tion and prevention efforts, as well as; (3) prognosis
of course of illness; and (4) monitoring response to
various treatments, including medications. In con-
junction with other clinical information, such tests
will play an important part of personalizing treatment
to increase effectiveness and avoid adverse reactions –
personalized medicine in psychiatry. Moreover, they
will be of immediate use to pharmaceutical com-
panies engaged in new neuropsychiatric drug devel-
opment efforts, at both a pre-clinical and clinical
(Phase I, II, and III) stages of the process.
Lymphocyte protein studies [16] and gene expres-
sion profiling [29–33] have emerged as a particularly
interesting area of research in the search for periph-
eral biomarkers. Most early studies suffer from one or
more of the following limitations: (1) The sample size
used in most reports so far is small. Given the genetic
heterogeneity in human samples and the effects of
50
Figure 5.1 Convergent functional
genomics. QTL, quantitative trait loci.
Human studies
Human
genetic
linkage or
association
Human
postmortem brain
evidence
Human blood
evidence
Specificity
illness state and environmental history, including
medications and street drugs, on gene expression, it
is questionable if they have sufficient power to extract
bona fide findings about trait and diagnosis, despite
the variety of sophisticated statistical methodologies
used. Combined approaches, such as CFG, may be
useful in terms of overcoming current limitations
[34]. (2) Use of lymphoblastoid cell lines. Fresh blood,
with phenotypic state information gathered at time
of harvesting, may be more informative than immor-
talized lymphocytes, and avoid some of the caveats of
Epstein–Barr virus (EBV) immortalization and cell
culture passaging [35].
The key litmus test for any biomarker (or genetic)
study should be whether the panels of markers show
predictive ability in independent cohorts. A series of
recent studies looking in whole blood at state gene
expression biomarkers for discrete phenes (mood,
hallucinations, delusions) [34, 36], using a CFG meth-
odology, have provided reasons for optimism. Panels
of DNA markers in genes prioritized by CFG have
also shown predictive ability in independent cohorts
in terms of distinguishing between bipolar disorder
and controls, and between more severe and less severe
forms of bipolar disorder [28].
P-values at best permit the precise ranking of find-
ings within a study. At worst, they are over-interpreted
and a source of frustration in terms of reproducibility
of findings across studies. This “p-value illusion” in
 Chapter 5: Gene expression studies

genetics [37], that a significant p-value in one study
should necessarily reproduce in another independent
study, is based on an under-appreciation of two
factors. The first one is the fit-to-cohort effect of
classic statistical analyses of genetic studies, the
second one is the above discussed complexity and
heterogeneity of most disorders. In essence, genes
are identified from the data of a single cohort. This
guarantees that best findings will not be at the top of
the list in a study conducted in a different cohort,
since complexity and heterogeneity will ensure that
no two cohorts are alike. This phenomenon is some-
times described as the “winner’s curse”, a strong ini-
tial finding not being as strong in subsequent
independent cohorts.
A solution is to use a fit-to-disease approach, like
CFG [11, 17, 26, 27, 34, 36, 38]. Such an approach, in
addition to p-values, uses multiple independent lines
of evidence related to illness, including gene expres-
sion studies, as a way of prioritizing findings within a
cohort, similar to a Google PageRank algorithm.
Genes prioritized in such a way may not have the
highest p-values, but will generalize and reproduce
well in independent cohorts. Since this approach is
based on a gene level integration of data rather than a
single nucleotide polymorphism (SNP) level analysis,
it reduces heterogeneity. It can and has been used
profitably to mine genome-wide association studies
[27, 28] and biomarker [34, 36] datasets, and to
extract panels of top genes or biomarkers that repro-
duce well in independent cohorts. Studies that by
themselves are relatively under-powered can be
mined and made to yield results using CFG, by
bringing to bear other large datasets and databases
relevant to that disease, resulting in essence in a field-
wide collaboration. Cross-validating signals from
other sources, with different noise factors, can
increase the signal–noise ratio and decrease the
required sample size.
A relatively recent development has been the focus
on regulatory RNAs [39], and on epigenetic modifi-
cations [40]. It is clear that gene expression studies
will play a major role in understanding how these
mechanisms come into play, and what impact the
environment has. After all, Gene  Environment =
Gene Expression!
In conclusion, gene expression studies have
proven to be a useful partner to classic genetics
approaches, and combined approaches may provide
shortcuts to discovery of genes and overall under-
standing of the neurobiology involved. For a complete
understanding of the illness, the analyses then need to
be pursued at a biological pathway and mechanistic
level, integrating environmental effects as key modu-
lators of gene expression and phenotype manifest-
ation. More progress in quantitative profiling of
psychiatric phenotypes, and borrowing of concepts
and paradigms from other medical fields that are
farther along, such as cancer genetics and genomics,
are exciting areas of advance for the near future.
A (r)evolution in medical nosology in general, and
psychiatric nosology in particular, will occur as a
result of such studies. It is hoped that together, all
these approaches will provide in the long term a
sound scientific basis for the development of person-
alized medicine in psychiatry [41–43].

References 7. Freimer N, et al. Nat Genet

1. Baum AE, et al. Mol Psychiatry
2008;13:197–207.
2. Smith EN, et al. Mol Psychiatry
2009;14:755–763.
3. Wellcome Trust Case Control
Consortium. Nature 2007;447:
661–678.
4. Schulze TG, et al. Hum Hered
2004;58:131–138.
5. Bearden CE, et al. Trends Genet
2006;22:306–313.
6. Niculescu AB, III, et al. Am J Med
Genet B Neuropsychiatr Genet
2006;141:653–662.
2003;34:15–21.
8. Kelsoe J, et al. CNS Spectrums
2002;7:215–226.
9. Berrettini W. Am J Med Genet C
Semin Med Genet 2003;123:59–64.
10. Craddock N, et al. Schizophr Bull
2006;32:9–16.
11. Le-Niculescu H, et al. Am J Med
Genet B Neuropsychiatr Genet
2007;144:129–158.
12. Purcell SM, et al. Nature
2009;460:748–752.
13. Dick DM, et al. Behav Genet
2006;36:112–126.
14. Meyer-Lindenberg A, et al.
Nat Rev Neurosci 2006;7:
818–827.
15. Le-Niculescu H, et al. Am J Med
Genet B 2008;147B:134–166.
16. Niculescu AB, III, et al. Physiol
Genomics 2000;4:83–91.
17. Ogden CA, et al. Mol Psychiatry
2004;9:1007–1029.
18. Tan PK, et al. Nucleic Acids Res
2003;31:5676–5684.
19. Shi L, et al. Nat Biotechnol
2006;24:1151–1161.
20. Mirnics K, et al. Neuron 2000;
28:53–67.

51
 Chapter 5: Gene expression studies
21. Choudary PV, et al. Proc Natl
Acad Sci U S A 2005;102:
15653–15658.
22. Niculescu AB, III. Genome Biol
2005;6:215.
23. Vawter MP, et al. Mol Psychiatry
2006;11:615.
24. Gould TD, et al. Neurosci
Biobehav Rev 2007;31:825–831.
25. Bertsch B, et al. Methods 2005;
37:274–279.
26. Rodd ZA, et al. Pharmacogenomics
J 2007;7:222–256.
27. Le-Niculescu H, et al. Am J Med
Genet B Neuropsychiatr Genet
2009;150B:155–181.
28. Patel SD, et al. Am J Med Genet B
Neuropsychiatr Genet 2010;
153B:850–877.
52
29. Glatt SJ, et al. Proc Natl
Acad Sci U S A 2005;102:
15533–15538.
30. Middleton FA, et al. Am J Med
Genet B Neuropsychiatr Genet
2005;136:12–25.
31. Segman RH, et al. Mol Psychiatry
2005;10:500–513.
32. Tsuang MT, et al. Am J Med
Genet B Neuropsychiatr Genet
2005;133:1–5.
33. Vawter MP, et al. Schizophr Res
2004;67:41–52.
34. Le-Niculescu H, et al.
Mol Psychiatry 2009;14:
156–174.
35. Rollins B, et al. Am J Med Genet
B Neuropsychiatr Genet
2010;153B:919–936.
36. Kurian SM, et al. Mol Psychiatry
2011;16(1):37–58.
37. Niculescu AB, III, et al. Am J Med
Genet B Neuropsychiatr Genet
2010;153B:847–849.
38. Niculescu AB, III, et al. Physiol
Genomics 2000;4:83–91.
39. St Laurent G, III, et al. Neurosci
Lett 2009;466:81–88.
40. Murgatroyd C, et al. Nat Neurosci
2009;12:1559–1566.
41. Niculescu AB III. Psychiatr Genet
2006;16:241–244.
42. Niculescu AB, III, et al.
J Affect Disord 2010;123(1–3):
1–8.
43. Niculescu AB, III, et al.
Neuropsychopharmacology
2010;35:355–356.
 Pharmacogenetics in psychiatry
Chapter
6 Falk W. Lohoff
Introduction
The introduction of medication for the treatment of
psychiatric disorders in the 1950s marked the beginning
of a new era in psychiatry. For the first time psychiatric
symptoms could be treated and alleviated with medica-
tion. This paradigm shift in psychiatry brought with it a
new focus on research, in particular investigation of
biological and neurochemical aspects in the pathophy-
siology of psychiatric disorders. The field of neuropsy-
chopharmacology was born and subsequently many
new compounds were developed for the treatment of
psychiatric symptoms. However, despite the continued
effort to optimize chemical compounds and the effort
to selectively target specific neurotransmitter systems,
response to and tolerability of medication remains
highly variable with some patients responding to one
treatment but not another. There are several potential
explanations for the variability in drug response rates,
including clinical heterogeneity and diagnostic uncer-
tainty, environmental and social factors, and genetic
factors. This chapter will describe the recent develop-
ments in the field of psychiatric pharmacogenetics.
Definition of pharmacogenetics
Pharmacogenetics is the term used to describe the
phenomenon that genes influence drug response and
side effects. Drug response and tolerability are influ-
enced by two main processes referred to as pharma-
cokinetics and pharmacodynamics. Pharmacokinetics
involves the absorption, distribution, metabolism,
and excretion of a drug while pharmacodynamics
refers to the effects of the compound on receptors,
transporters, and other downstream targets. Both of
these systems are influenced by heritable variation in
genes. The complex interaction of multiple genes
involved in pharmacokinetic and dynamic processes
Principles of Psychiatric Genetics, eds John I. Nurnberger, Jr. and Wade H. Berrettini. Published by Cambridge University
Press. © Cambridge University Press 2012.
in the context of environmental influences comprises
essentially the action of a drug in an individual. By
identifying genetic components implicated in drug
response and side effects, the hope is to precisely
match a medication to the patients’ genetic makeup
and thus maximize treatment response while minimiz-
ing potential adverse events. The environment of
a patient will always remain unique and “personal”;
however, advances in genomic medicine promise a
more comprehensive “personalized” pharmacotherapy.
Pharmacogenetics of antidepressants
Major depressive disorder (MDD) is a common psy-
chiatric illness with high levels of morbidity and mor-
tality. It is estimated that 10–15% of the general
population will experience clinical depression during
their lifetime and 5% of men and 9% of women will
experience a depressive disorder in a given year [1].
The introduction of tricyclic antidepressants (TCAs) in
the 1950s represented a great advance in the treatment
of depression; however, it soon became clear that ser-
ious side effects and toxicity limited their use and also
varied substantially between patients. Although there
are multiple new pharmacological treatments available
with better side effect profiles, response and tolerability
to medication continue to be highly variable and in
many cases poor. In general, only one-third of treated
MDD patients respond to pharmacological treatment
and achieve remission of symptoms [2, 3]. Based on
these circumstances, the search for biomarkers that
predict treatment response and efficacy in depression
is of high interest to the field.
Genetic influences on antidepressant
drug pharmacokinetics
Pharmacokinetic factors in antidepressant response
have been subject to extensive research over the last
53
 Chapter 6: Pharmacogenetics in psychiatry
several decades [4, 5]. Most attention has been given
to the cytochrome P450 systems, involved in phase 1
oxidation. Cyptochrome P450 (CYP450) enzymes are
drug-metabolizing hemoproteins present in multiple
tissues with predominance in liver. Over 50 enzymes
have been described which are encoded by over
63 CYP450 genes. The main P450 enzymes involved
in antidepressant drug metabolism are CYP2D6,
CYP2C19, CYP3A4, and CYP1A2 [6]. The CYP2D6
system has been studied extensively and is so far the
best characterized [7]. Early experiments with debri-
soquin and nortriptyline documented that patients
fall into different categories, including poor (PM),
intermediate (IM), extensive (EM), and ultra-rapid
metabolizers (UM). Cloning and characterization
of the CYP2D6 gene led to the identification of over
75 CYP2D6 alleles which determine metabolizer
status. For example some individuals carry alleles that
encode for an inactive enzyme or no enzyme at all,
while others have gene duplications resulting in UM
of CYP2D6 substrates. Metabolizer status is also
influenced by ethnicity, with 7% of Caucasians being
PM compared to 1% in the Asian population. Certain
African populations have higher proportions of UM
resulting from multiple copies of the CYP2D6 gene.
Such individuals can have an inadequate therapeutic
response to standard doses of the drugs metabolized
by CYP2D6. In fact, plasma levels of the TCA nor-
triptyline and imipramine were closely related to the
number of functional CYP2D6 gene copies [8, 9].
Another study showed that patients who lack either
one or both functional copies of the gene reach thera-
peutic plasma levels with starting doses of nor-
triptyline and reach potentially toxic plasma levels
with high-normal doses. Patients with two to four
copies on the other hand require high-normal doses
just to reach therapeutic plasma levels [10]. Similar
CYP2D6-allele/plasma level concentration correl-
ations have been reported for selective serotonin
reuptake inhibitors (SSRIs), such as fluoxetine and
paroxetine [11], and serotonin norepinephrine reup-
take inhibitors (SNRIs), such as venlafaxine [12, 13].
While a clear correlation between drug plasma levels
and therapeutic antidepressant efficacy has been
described for TCA [14], it is less apparent for SSRIs
and SNRIs [15]. Knowledge of the genetic metabolizer
status of a patient is thus helpful to the clinician in
order to potentially avoid side effects, and might
further help to reach therapeutic levels faster; never-
theless, the overall effects on improved efficacy are
54
still unclear and need to be established. The idealistic
vision of predicting side effects based on genotypes
becomes complicated by the reality of environmental
factors and gene–environment interactions. For
example dietary factors (grapefruit juice, caffeine,
nicotine) have known influences on drug metabolism
[16]; however, perhaps the most common influencing
factor is taking another medication. A drug that
strongly inhibits CYP2D6, for example cimetidine
or amiodarone, may make an individual who is other-
wise a normal metabolizer, appear like a poor meta-
bolizer with increased side effects. Future studies will
be necessary to dissect the complicated interaction
between clinical presentation, environmental factors,
and CYP450 genotype.
Genetic influences on antidepressant
drug pharmacodynamics
The term pharmacodynamics is used to describe the
effects of a drug on the body. Pharmacodynamic
aspects thus include interactions of a drug with
neurotransmitter receptors, transporters and other
downstream targets. Although the primary mechan-
ism of action is known for most antidepressant drugs,
and thought to involve predominantly monoaminer-
gic neurotransmitter systems, the exact mechanisms
by which antidepressant medications work remain
elusive. Most pharmacogenetic studies in MDD so
far have focused on candidate genes involved in
monoaminergic neurotransmission. Some of the
most obvious targets that have been studied include
genes encoding the serotonin transporter and recep-
tor, norepinephrine transporter, dopamine receptors,
monoamine oxidase A, tryptophan hydroxylase, and
cathechol-O-methyltransferase [17]. It should be noted
that candidate gene association studies in pharmaco-
genetics face similar conceptual issues, as do genetic
association studies in general. Many studies have been
limited by small sample sizes, clinical and diagnostic
heterogeneity, incomplete clinical data, differences in
definition of treatment response, genetic complexity,
unknown functional relevance of single nucleotide
polymorphisms (SNPs), and limited biological evi-
dence for candidate gene selection. Despite these
obstacles, several candidate genes deserve mention as
they have been suggested repeatedly to be implicated
in treatment response to antidepressant drugs.
One of the most widely studied genes in pharma-
cogenetic studies of depression is the serotonin

transporter gene (SLC6A4). This gene is located on
chromosome 17q and several polymorphisms have
been reported in association studies. One polymorph-
ism in the promoter region of the gene (5HT trans-
porter gene linked polymorphic region: 5HTTLPR)
consists of an insertion or deletion of a repetitive
sequence, producing a short (S) allele or a long (L)
allele [18]. Although the 5HTTLPR was originally
described as bi-allelic, rare very-long and extra-long
alleles have been described in Japanese and African-
Americans [19]. The 5HTTLPR L-allele has been
shown to affect transporter function, resulting in
higher rates of serotonin reuptake by the transporter.
Based on this observation, several pharmacogenetic
studies have investigated this polymorphism with
regards to antidepressant drug response [20–22].
Results have been mixed, with some studies showing
an effect of the L-version while others failed to dem-
onstrate an effect. A recent meta-analysis of 15 pub-
lished studies however indicated that there was a
significant association between the L-allele and better
treatment response to SSRIs [23]. Interestingly, the
association between the L-allele and antidepressant
treatment response within four weeks was the most
robust finding in this meta-analysis, suggesting that
the 5HTTLPR might predict not only treatment
response but also the time-course of response and
remission. Despite these interesting findings, it is still
premature to recommend 5HTTLPR genetic testing
on a widespread level for patient with MDD. Future
large-scale prospective clinical studies are needed, to
confirm the effect of the 5HTTLPR on antidepressant
treatment response and to identify potential other gen-
etic variants that in combination with the 5HTTLPR
will predict treatment outcome.
While several other genes have been investigated
as candidates for antidepressant drug response, the
results are ambiguous and a clear established relation-
ship between genetic variants and antidepressant drug
treatment outcome has not yet been demonstrated
[4, 5, 24, 25]. Recently, results from the Sequenced
Treatment Alternatives to Relieve Depression (STAR*D)
trial became available for review. This multicenter
trial included 3671 patients with MDD who were
treated with the SSRI citalopram as a first-line agent
for 12 weeks [26]. Main outcome measures were cat-
egorical response, remission, tolerance, and adverse
effect burden. Analyses of the first phase of the trial
showed that only 30% of treated patients reached
remission after adequate citalopram treatment [3].
Chapter 6: Pharmacogenetics in psychiatry
As part of the trial, 1953 patients gave blood for
pharmacogenetic testing.
Several groups have investigated whether genetic
factors might predict clinical outcome in the STAR*D
sample. McMahon et al. [27] investigated 68 candi-
date genes and genotyped 768 SNPs in this sample.
They detected an association between the serotonin
receptor gene HTR2A and treatment outcome [27].
Although citalopram does not bind directly to
5HT2A receptors, citalopram downregulates 5HT2A
receptors in animal models [28, 29], which could
play a role in differences in treatment response. Other
candidate genes implicated in antidepressant drug
efficacy in the STAR*D trial are FKBP5 and GRIK4
[30, 31]; however, all of these findings require inde-
pendent replication.
Given the fact that response and remission are
complex and influenced by multiple factors, perhaps
focusing on the side effect profile of antidepressants
might reveal a closer link to genetics. For example,
while an effect of the 5HTTLPR on primary efficacy
could not be established in the STAR*D sample [32],
this polymorphism has been reported to be associated
with side effects [33]. In addition, several recent stud-
ies have suggested an association between genetic
markers and treatment emergent suicidal ideation
using the STAR*D sample [25, 34].
All of the abovementioned studies using the
STAR*D sample were candidate gene studies. Genes
and polymorphisms were selected based on the
current understanding of the neurobiology of mood
disorders. With the advance of technology, it is now
possible to investigate over a million SNPs in an
individual simultaneously without a priori SNP selec-
tion based on biological plausibility. Such studies,
called genome-wide association studies (GWAS),
are currently underway in the STAR*D and other
samples. Recently, results from the first GWAS of
antidepressant response became available using a
sample from Germany [35]. Despite the inclusion of
more than 1500 patients with depression, 700 of them
with genome-wide genotyping, the study failed to
identify single SNP signals that satisfied the criteria
for genome-wide statistical significance, suggesting
that multiple loci are involved with only modest effect
sizes. Even though these promising new approaches
to investigate genetic variants at a large scale are
intriguing, caution should be used in the interpret-
ation of findings. While the STAR*D sample has the
advantages of large sample size and a consistent
55
 Chapter 6: Pharmacogenetics in psychiatry
design in which all patients received citalopram, it is
important to note that the STAR*D trial was not
designed to answer pharmacogenetic questions. There
are several concerns regarding this sample from a
pharmacogenetic perspective, including heterogeneity
in the severity of depression, multicenter design,
inclusion of different ethnic populations, and lack of
data on drug levels. In order for pharmacogenetic
studies in MDD to succeed, there is a clear need for
well-designed, large prospective trials with a focus on
the genetic component [36]. One such attempt is cur-
rently being undertaken by the European Consortium
project, Genome-Based Therapeutic Drugs for Depres-
sion (GENDEP). This is the first large-scale multicen-
ter human pharmacogenomics study focused on the
prediction of therapeutic response to antidepressants
and adverse effects [37]. This open-label, flexible dose,
multicenter trial included 760 patients with MDD that
were treated with either citalopram or nortriptyline for
12 weeks. Initial analysis of 10 candidate genes
involved in serotonin, norepinephrine, neurotrophic,
and glucocorticoid signaling revealed an association
between treatment response to escitalopram and sev-
eral variants in the serotonin receptor gene (HTR2A)
with one marker (rs9316233) explaining 1.1% of the
response variance. SNPs in the norepinephrine trans-
porter gene (SLC6A2) predicted response to nortripty-
line, and variants in the glucocorticoid receptor gene
(NR3C1) predicted response to both antidepressants
[38]. These data further support a role for the influence
of genetic variants on treatment response to anti-
depressant drugs. Since single marker analysis only
explains a small fraction of the variance, future studies
will have to use a multiple variant approach in order to
find clinically meaningful genetic prediction
algorithms.
Pharmacogenetic studies of antidepressant-
induced adverse events
Prediction and prevention of adverse events are
important goals of pharmacogenetics and personal-
ized medicine. Most pharmacogenetic studies of anti-
depressants have focused predominantly on treatment
response, perhaps also due to the fact that most cur-
rent antidepressants are well tolerated and fairly safe.
Nevertheless, SSRIs and SNRIs can also cause signifi-
cant side effects including sexual dysfunction, gastro-
intestinal disturbances, and weight gain, all of which
might influence compliance with treatment [39].
56
A recent study investigated the serotonin transporter
gene promoter polymorphism in relation to mirtaza-
pine and paroxetine efficacy and adverse events in
geriatric major depression. Results showed that the
short allele of the 5HTTLPR polymorphism had a
significant effect on adverse events among paroxe-
tine-treated subjects [40]. Similarly, another study
evaluated the association between adverse events
during SSRI treatment in 214 MDD patients and 2
polymorphisms in the serotonin transporter gene and
showed that patients with the 5HTTLPR S/S or S/L
genotype appeared to have an increased risk of
adverse events, especially general adverse events such
as dermatological reactions, weight change, and
fatigue [41]. Hu et al. [33] investigated the 5HTTLPR
polymorphism in the STAR*D sample and showed
that the S-allele is associated with citalopram adverse
effects, in particular diarrhea. These data suggest that
the short allele of the 5HTTLPR polymorphism pre-
dicts adverse effects of SSRI therapy.
Another serious adverse event of antidepressant
therapy has recently gained much attention. Treat-
ment emergent suicidal ideations (TESI) are serious
adverse events in the management of psychiatric
disorders with antidepressants. It is estimated that
approximately 4% of patients treated with anti-
depressants develop TESI compared to 2% of
patients treated with placebo [42–44]. Since the US
Food and Drug Administration (FDA) -mandated
black box warning on antidepressant-induced sui-
cidal ideations, the number of antidepressant pre-
scriptions has significantly decreased. The recent rise
in the number of suicides and drop in the number of
antidepressant prescriptions [45], possibly related to
the TESI black box warning, mandates thorough
investigation of TESI both clinically and preclinically.
Two genetic association studies have investigated
whether genetic variants contribute to TESI in the
STAR*D sample [34, 46]. Laje et al. [34] investigated
768 SNPs in 68 candidate genes in the STAR*D sample
with respect to TESI. They reported two markers within
GRIK2 and GRIA3 that were associated with TESI
during citalopram therapy. Perlis et al. [46] showed that
polymorphisms in the CREB1 gene were associated with
TESI among men with depression. These studies await
replication and in both cases interpretation of results is
complicated by several factors including the a priori
selection of candidate genes (Laje et al. [34] 68
candidate genes; Perlis et al. [46] one candidate gene),
the heterogeneous clinical characteristics of the

STAR*D population (see above) and the lack of bio-
logically functional assessment of genetic variants on
the cellular level. Results of the first GWAS for TESI
have been recently reported for the STAR*D sample,
suggesting a potential role of variants in the PAPLN
and IL28RA genes. PAPLN encodes papilin, a proto-
glycan-like sulfated glycoprotein. IL28RA encodes an
interleukin receptor [47]. Another study showed that
polymorphisms in the BDNF gene were significantly
associated with an increase in suicidal ideation in the
GENDEP cohort [48]. Interestingly, a significant
interaction between variants in BDNF and NTRK2,
the gene encoding the BNDF receptor, could be dem-
onstrated, suggesting the importance of genetic vari-
ants in biological pathways and risk to TESI.
Pharmacogenetics of antipsychotic
drugs
Pharmacotherapy is the treatment of choice for
psychotic symptoms of mental conditions such as
schizophrenia, bipolar disorder, and psychotic
depression. Antipsychotic drugs are traditionally
divided into two groups: typical (first-generation)
antipsychotics, with strong affinity for the dopamine
receptor, and atypical (second-generation) antipsy-
chotics, with multiple receptor targets. Antipsycho-
tic drugs have significantly improved the clinical
outcome for schizophrenia patients, although several
serious side effects remain important limitations,
including metabolic abnormalities [49, 50], cardio-
vascular events [51], and movement disorders [52].
Despite their widespread use, about 70% of patients
with chronic schizophrenia discontinued their anti-
psychotics drugs in a recent multicenter study due to
poor effectiveness or tolerability [53]. Antipsychotic
pharmacotherapy achieves clinical improvement in
the treatment of psychosis in about 50% of patients
[53, 54]. This high degree of response variability and
subsequent treatment failure can be attributed to
multiple factors including clinical heterogeneity,
environmental factors, and genetic variations. While
epidemiological studies of schizophrenia suggest a
strong genetic component, no similar studies exist
in relation to response to antipsychotics. Neverthe-
less, the search for genetic variation that predicts
treatment response and occurrence of adverse events
is the subject of current research and might, in the
near future, alter the way antipsychotic medications
are prescribed.
Chapter 6: Pharmacogenetics in psychiatry
Genetic influences on antipsychotic drug
pharmacokinetics
As discussed above for antidepressant drugs, recent
pharmacokinetic genetic studies of antipsychotics
have mostly focused on the cytochrome P450 system;
however, other areas in which genetic variation may
impact pharmacokinetics include blood–brain and
blood–intestine barrier systems [55, 56]. The cyto-
chrome P450 (CYP450) enzymes mediate phase 1
oxidation of many antipsychotic drugs. Here again,
patients can be grouped into different phenotypic
metabolizer categories, including PM, IM, EM, and
UM. Several genetic polymorphisms in genes encod-
ing CYP450 enzymes have been associated with these
phenotypes and are predictive of metabolizer status.
Although antipsychotic drugs are usually metabolized
by a variety of different CYP450 enzymes, major
metabolic pathways that are clinically relevant have
been identified for most commonly prescribed drugs
[57, 58]. Detailed description of individual antipsy-
chotic compounds and their metabolism can be found
elsewhere [59].
CYP2D6 is the major metabolic enzyme for classic
antipsychotic medications such as chlorpromazine
and haloperidol but also for the atypical risperidone
[59]. Several polymorphisms in the CYP2D6 gene are
responsible for PM status and gene duplication can
lead to UM status. As mentioned earlier, the frequen-
cies of these phenotypes vary between ethnic groups,
with 7–10% of Caucasians being PM in contrast to
only 1–2% of Asians. Recent data show that the meta-
bolic ratio of antipsychotic substrates for CYP2D6
can be affected by genetic variants. For example, the
metabolism of haloperidol is severely reduced in PM
and dose adjustment is recommended [60]. On the
other hand, for patients that are UM, higher thera-
peutic doses are necessary to compensate for their
rapid elimination of drugs.
Another important CYP450 enzyme involved in
antipsychotic pharmacokinetics is CYP1A2, for which
clozapine and olanzapine are substrates [61, 62]. Sev-
eral variations in the CYP1A2 gene result in decreased
enzyme activity [63, 64] although the clinical rele-
vance remains to be determined. Some reports indi-
cate UM status for CYP1A2 is associated with delayed
response to clozapine [65, 66]. While CYP1A2
enzyme activity is in part genetically determined, out-
side factors are also involved. This is an important
consideration in particular for the treatment of
57
 Chapter 6: Pharmacogenetics in psychiatry
patients with schizophrenia, since 80–90% are nico-
tine dependent and smoking significantly increases
the activity of CYP1A2 resulting in higher metabolic
ratios, reduced drug levels, and subsequent lack of
treatment response [67]. Genotyping and identifica-
tion of CYP1A2 UM status, in particular in patients
who smoke, is recommended for accurate dose
adjustment. Numerous polymorphisms have been
discovered for CYP3A4, CYP2C9, CYP2C19, and
CYP2A5; however; no clear connection between vari-
ants and level of response to antipsychotic medication
has yet been established.
Genetic influences on antipsychotic drug
pharmacodynamics
Antipsychotic drugs display a wide range of affinities
for different neurotransmitter receptors, including
dopaminergic, serotoninergic, histaminergic, muscar-
inergic, glutaminergic, and adrenergic receptors [68,
69]. The dopaminergic system is thought to play a
major role in the pathophysiology of psychosis and
dopamine receptor blockade has been associated with
amelioration of psychotic symptoms [70]. Positron
emission tomography (PET) studies show that 60%
minimum occupancy of D2-like receptors is necessary
to obtain therapeutic response and occupancy above
80% may lead to extrapyramidal side effects [71, 72].
Variants of the genes encoding receptors targeted by
antipsychotic medications are logical candidates for
pharmacogenetic investigations. Several studies have
investigated polymorphisms in the DRD2 gene and
response to antipsychotic treatment. The promoter
insertion/deletion polymorphism -141-C Ins/Del was
associated with lower treatment response [73–75].
Although the functional relevance of this polymorph-
ism remains unclear, some evidence suggests an asso-
ciation of the Del allele with higher striatal D2 receptor
density [76].
The DRD3 gene has also been investigated with
respect to antipsychotic treatment response. The
Ser9Gly polymorphism in the DRD3 gene has been
associated with effects on dopamine binding affinity
[77]. Several studies suggest that the Gly allele is associ-
ated with good treatment response [78–80]; however,
one study in patients of Chinese descent indicated pre-
diction of poor treatment response [81]. Several other
genes involved in dopaminergic neurotransmission
have been investigated with regards to antipsychotic
58
treatment response but no clear association could be
documented [57].
The serotonergic system has been also the focus of
pharmacogenetic studies of antipsychotics, in part
driven by the high occupancy of 5HT2 receptors by
atypical antipsychotics [82]. Several groups have
investigated the relationship between the 102-T/C
substitution polymorphism in the 5HT2A gene and
response to clozapine and risperidone with both posi-
tive and negative results reported [83–88]. Another
variant in this gene, the amino-acid substitution poly-
morphism His452Tyr, was also found to be associated
with clozapine treatment response in several datasets
[89, 90]. A recent meta-analysis further supports an
association of the 102-T/C and His452Tyr variants with
clozapine response, reinforcing the role of the 5HT2A
receptor as an important therapeutic target [84].
A variety of other serotonin receptor subunit genes
have been investigated as well with respect to antipsy-
chotic treatment response, including 5HT1A, 5HT2C,
5HT3A, 5HT3B, 5HT5A, and 5HT6 [57]. Results of
these studies are mixed and follow-up studies will have
to use larger samples in randomized prospective treat-
ment trials. Future studies using GWAS designs are
also likely to be used for investigations of pharmaco-
genetic aspects of antipsychotic treatment response;
however, methodological issues regarding sample size
and multiple testing are significant obstacles that
impact the feasibility of such endeavors.
Pharmacogenetic studies of antipsychotic
drug-induced adverse events
The side effects associated with antipsychotic treatment
often contribute to poor compliance and treatment
failure. Drug-induced weight gain and tardive dis-
kinesia (TD) have been the primary focus of pharma-
cogenetic studies in the recent past, although
accumulating data are also available for less common
side effects like drug-induced agranulocytosis and
neuroleptic malignant syndrome.
Antipsychotic-induced weight gain is a serious
side effect that is particularly high among patients
receiving multiple antipsychotic drugs simultaneously
[91]. Evidence from twin studies suggests a genetic
component to this phenomenon [92, 93]. Since sero-
tonin and histamine receptors play important roles
in eating behavior, genes encoding these receptors
are logical candidates for pharmacogenetic inquiry.
Numerous studies have investigated the 5HT2C

receptor gene, and although some studies did not find
an association, accumulating evidence suggests a role
for the functional -759-T/C polymorphism in medi-
ating weight gain [94–98]. Interestingly, the T-allele
has strong protective properties against antipsychotic-
induced weight gain (odds ratio [OR] = 6) [99].
Another candidate that has been investigated with
regards to weight gain is the leptin gene. Results from
a nine-month antipsychotic treatment trial suggest a
significant association between the 2548-A/G poly-
morphism and weight gain [98]; however, no biological
functional data on this SNP has been reported. Several
other genes have also investigated with regards to anti-
psychotic weight gain; however, results await con-
firmation and are reviewed in detail elsewhere [100].
TD and abnormal extra pyramidal movements are
serious side effects observed in some patients treated
chronically with antipsychotic drugs. Although the
mechanism of TD remains unknown, symptom
occurrence has been directly associated with drug
dosage and plasma levels [52]. Several genes have
been investigated as potential candidates for antipsy-
chotic-induced TD [57, 101]. One of the most prom-
ising findings has been the association between the
Ser9Gly variant in the dopamine D3 receptor gene
(DRD3) and antipsychotic-induced movement dis-
orders [102–104]. A combined analysis of 780 patients
from different ethnic groups showed that the DRD3
Gly allele increased susceptibility to TD with a pooled
OR of 1.33 [105]. A recent meta-analysis further
confirmed a significant contribution of the DRD3
Gly allele with an increased risk of developing TD with
an OR of 1.17 [106]. While these data support the
DRD3 Ser9Gly polymorphism as a pharmacogenetic
marker for TD, the modest ORs indicate a weak effect
and it is likely that other genetic variants are involved
in the complex pathophysiology of TD.
Antipsychotic-induced agranulocytosis is a rare
(0.7–3.0%) but severe adverse event, most often
observed in patients treated with clozapine [107].
While the etiology of this severe side effect remains
elusive, some reports have indicated an involvement
of immune mechanisms. In particular, genes coding
for human leukocyte antigens (HLA) have been asso-
ciated with clozapine-induced agranulocytosis [108–
112]. However, as with other antipsychotic-induced
adverse events, the genetic architecture of drug-
induced agranulocytosis is complex, likely involving
multiple genes, gene–gene, and gene–environment–
drug interactions.
Chapter 6: Pharmacogenetics in psychiatry
Pharmacogenetics of anticonvulsants
and mood stabilizers
Anticonvulsant drugs are widely used in the manage-
ment of behavioral disorders, including bipolar dis-
order, mood disorders, and impulse control disorders
[113–115]. Similar to other drugs, inter-individual
differences exist in drug response and tolerability,
at least partially related to genetic factors [116].
Analogous to the pharamacokinetic pharmacoge-
netics of antidepressants and antipsychotics, variation
in metabolizing enzymes play an important role in
determining successful treatment with certain anti-
convulsants, including phenytoin, carbamazepine,
and benzodiazepines. In particular genetic variants
in the CYP2C19 and CYP3A4/5 pathway have been
documented to influence benzodiazepine plasma
levels and related adverse events [117]. While the extent
of pharmacogenetic factors for other anticonvulsants,
such as valproate, lamotrigine, gabapentin, pregabalin,
and topiramate remains to be determined, accumu-
lating evidence suggests genetic factors contribute
to lithium response and carbamazepine-induced
Stevens–Johnson syndrome.
Pharmacogenetic studies of lithium
Lithium is an alkali metal that is used in salt form,
such as lithium carbonate and lithium citrate, for
the treatment of bipolar disorder. The therapeutic
efficacy of lithium for bipolar illness was first dis-
covered by John Cade in 1949 [118] and further
systematically evaluated by Mogens Schou in a series
of clinical trials [119]. Lithium is currently considered
first-line treatment for bipolar disorder, in particular
for patients with euphoric mania. Epidemiological
studies in bipolar disorder have consistently docu-
mented a strong genetic component to its etiology,
with heritability estimates ranging from 65 to 80%
[120]. Despite these strong genetic predispositions,
identification of susceptibility alleles for bipolar dis-
order has been difficult due to the complex mode
of inheritance, moderate effect sizes and genetic and
clinical heterogeneity. Clinical heterogeneity is par-
ticularly apparent with regards to treatment response.
Some patients respond remarkably well on lithium
monotherapy, while for others the medication has
no benefit [121]. The “endophenotype” of lithium
responsiveness has been used diagnostically in clinical
59
 Chapter 6: Pharmacogenetics in psychiatry
practice and several family studies have suggested that
lithium response is a heritable trait. Mendlewicz
carried out the first family study of lithium response
and showed a higher risk for bipolar disorder in first-
and second-degree relatives of responders compared
to nonresponders [122]. Since then, several other
studies replicated this finding [123–125]. Based on
the high degree of heritability of bipolar disorder,
and more so on evidence for a role of genetics on
lithium responsiveness, pharmacogenetic studies are
logical steps in the dissection of bipolar disorder
pathophysiology. Similar to the early genetic studies
of bipolar disorder investigating disease susceptibility
loci via molecular linkage methods [120], the earliest
pharmacogenetic studies of lithium response were
pedigree linkage studies. A genome-wide linkage
study of bipolar families from Quebec with positive
lithium response found a major locus on chromo-
some 12q23-q24 [126]. Another linkage study of
31 families with lithium responsiveness reported sig-
nificant results for chromosome 15q14 and suggestive
results for 7q11 [127]. Despite these interesting data,
none of the linkage regions has been confirmed
independently.
The search for genetic factors involved in lithium
response has also used a candidate gene approach.
This methodology depends upon biological collateral
data to inform potential genetic targets of investiga-
tion. Several biological pathways have been suggested
to play an important role in the mechanism of action
of lithium [128]. The phosphoinositide pathway has
been shown to be affected by lithium, which causes
inhibition of inositol monophosphatase and inositol
polyphosphate 1-polyphosphatase [129]. Inhibition of
these enzymes by lithium causes a reduction in the
amount of free inositol available for regeneration of
phosphatidylinositol biphosphate, ultimately leading
to diminishing cellular levels of inositol triphosphate
and diacylglycerol. In addition, lithium’s inhibitory
effects on glycogen synthase kinase 3 [130], with
downstream effects on transcription, are thought to
be involved in its clinical effectiveness. Based on these
data, recent studies have investigated an association
between the promoter -50T/C polymorphism in the
GSK3-b gene and lithium response. While two associ-
ation studies documented a relationship between the
promoter variant and lithium prophylaxis [131, 132],
another group failed to replicate the finding [133].
Interestingly, a more recent study documented an
association between the GSK3-b -50T/C SNP and
60
response to lithium augmentation in acutely depressed
antidepressant nonresponders, expanding the potential
role of lithium responsiveness across diagnostic
categories [134]. Several other candidate genes have
been investigated with regards to pharmacogenetics
of lithium response [135, 136], most of them involved
in monoamine neurotransmission, including COMT,
DRD2, DRD3, MAOA, serotonin transporter, and
receptor genes [137–139]. So far no clear correlation
between variants in these genes and lithium response
has been established and caution should be used in
interpretation of the results given the significant clin-
ical heterogeneity and differences in treatment
response criteria between studies [17, 137, 140–158].
While the field of psychiatric genetics has advanced
to GWAS, no data exist for lithium response. However,
a recent GWAS on bipolar disorder [159] genotyped
over 550 000 SNPs in two independent case-control
samples of European origin. Several genes were iden-
tified, with rather modest effects. Nevertheless, the
strongest result of this study was related to variation
in the diacylglycerol kinase, eta (DGKH) gene,
involved in the lithium-sensitive phosphatidyl inositol
pathway. Although the field of lithium response
genetics is mostly unexplored, results obtained so far
are encouraging. Future studies will have to include
careful characterization of the bipolar phenotype,
improved methodological quality and standardized
criteria for lithium response.
Pharmacogenetic studies of carbamazepine-
induced Stevens–Johnson syndrome
Carbamazepine (CBZ) is an important treatment for
seizure disorders, bipolar disorder and chronic pain.
Tolerability varies among patients and common side
effects include drowsiness, headaches, motor coordin-
ation impairment and/or upset stomach. Beside these
common side effects, several serious adverse events
have been reported under CBZ therapy, such as aplas-
tic anemia, fatal arrhythmias, and life-threatening
cutaneous disorders. Stevens–Johnson syndrome
(SJS) and toxic epidermal necrolysis (TEN) have both
been associated with CBZ use. The incidence of CBZ-
induced SJS/TEN is approximately 1–6 per 10 000
of new users in mainly Caucasian countries and is
10-fold higher in some Asian countries [160]. The
skin lesions of SJS and TEN consist of blisters that
arise on erythematous or purpuric macules and
involve two or more mucosal surfaces. Both disorders

carry a mortality of almost 30% [161, 162]. While
these two disorders are considered distinct entities,
they lie on a continuous clinical spectrum, with TEN
being considered the most severe form [163]. The
pathophysiological mechanisms for the development
of SJS/TEN remain elusive, although current evidence
suggests an immune-complex-mediated hypersensi-
tivity with subsequent keratinocyte cell apoptosis.
Recently, a strong association between the major
histocompatibility complex HLA-B*1502 allele and
CBZ-induced SJS among Han Chinese has been iden-
tified [164]. The initial study included 44 patients that
developed SJS under CBZ and 101 controls that toler-
ated CBZ well. All subjects were Han Chinese patients
living in Taiwan. Genotyping of 157 SNPs in candi-
date genes was carried out and results showed that
all 44 patients with SJS were positive for the HLA-
B*1502 but only 3% of CBZ-treated patients without
SJS had the same allele. The allele was present in 8.6%
of normal controls from the general population.
The OR for developing CBZ-induced SJS if positive
for HLA-B*1502 was  2500 with a positive predictive
value of 93.6% [164]. A follow-up study by the same
research group included 16 new CBZ-induced SJS
cases of Chinese descent, again demonstrating a
strong association of the HLA-B*1502 allele [165].
Another smaller study from Hong Kong included
eight Han Chinese patients with CBZ-induced SJS
and reported that six of them had the HLA-B*1502
allele [166]. Furthermore, Locharernkul et al. [167]
showed for the first time a strong association between
the HLA-B*1502 allele and CBZ-induced SJS in the
Thai population, demonstrating an effect of the allele
in a non-Chinese population. In contrast to these
robust positive reports, analysis of European patients
with CBZ-induced SJS failed so far to identify a clear
relationship between HLA-B*1502 and development
of this severe adverse event [168, 169]. One possible
reason that no clear association was found between
HLA status and phenotype might be due to the low
HLA-B*1502 allele frequency in patients of European
descent. In fact, the frequency of HLA-B*1502 varies
greatly across ethnic groups. Individuals of Han
Chinese ancestry show allele frequencies of greater
than 15% in individuals from Hong Kong, Thailand,
Malaysia, and parts of the Philippines. Other Asian
ethnic groups show less frequent occurrence of the
HLA-B*1502 allele with estimates of 9–10% in Taiwan-
ese individuals, 4% in North Chinese, 2–4% in South
Asians and Indians, and less than 1% in Japanese and
Chapter 6: Pharmacogenetics in psychiatry
Koreans. In Caucasians, African-Americans, Hispanics,
and Native Americans, the HLA-B*1502 allele is rare
and below 2%; however, limited data exist to estimate
allele frequencies accurately in these populations
[170, 171].
Based on these compelling data, the FDA released
a warning in 2007 stating that serious and potentially
fatal skin reactions may occur with the administration
of CBZ in patients positive for the HLA-B*1502 allele.
In addition the FDA recommended genotyping
patients of Asian descent before initiation of CBZ
treatment [172]. While the strong link between the
HLA-B*1502 marker and CBZ-induced SJS is the first
robust and clinically relevant example of the use of
pharmacogenetics, future studies not only have to
replicate above findings in large cohorts of different
ethnic background, they also have to carefully exam-
ine potential class effects of compounds with similar
chemical structures. It will be important for future
studies to investigate the extent of the risk for a severe
skin reaction in HLA-B*1502 carrier of non-Asian
descent.
Future directions and implications
for clinical practice
The field of psychiatric pharmacogenetics is develop-
ing rapidly and there is hope for identifying genetic
variants that are clinically meaningful. This expect-
ation is in particular supported by the strong associ-
ation between the HLA-B*1502 allele and CBZ-
induced SJS, which not only showed a large effect in
a subset of patients with a very specific adverse event,
but also demonstrated that it is possible to identify
clinically relevant pharmacogenetic associations in small
to moderate-sized groups of patients. “Pharmaco-
genetics” by definition has the conceptual problems
of both disciplines. Psychopharmacology has been
plagued traditionally by high placebo response rates,
poor treatment response rates, and low remission
rates. For example, pharmaceutical trials in depres-
sion show response rates of about 50%, remission
rates of about 30% [26], and a placebo response rate
of 30–60% [173–175]. Furthermore, the influence of
environmental factors on drug response is poorly
understood and thus difficult to control. The field
of genetics has been challenged by lack of power in
samples studied, clinical and genetic heterogeneity,
and lack of biological functional data on genetic vari-
ants. Genetic studies can only suggest an association
61
Table 6.1 Pharmacogenetic tests in clinical practice.

Genetic test Gene Drug Useful for Commercially Notes

available
Cytochrome CYP450 2D6 Antidepressants  Identifying patients who are poor, Roche AmpliChip; FDA approved
P450 Drug intermediate, extensive, or ultrarapid available in special
Mainly metabolized:
metabolizer metabolizers of psychotropic drugs laboratories
desipramine
status metabolized by CYP 2D6 and 2C19
fluoxetine
 Adjusting dosages for psychotropic
nortriptyline
drugs that are metabolized by
paroxetine CYP 2D6, 2C19
Partly metabolized:
amitriptyline
bupropion
citalopram
duloxetine HCL
escitalopram
fluvoxamine
maprotiline
mirtazepine
venlafaxine

Antipsychotics
Mainly metabolized:
aripiprazole
fluphenazine
perphenazine
risperidone
thioridazine

Partly metabolized:
chlorpromazine
haloperidol
olanzapine
 Stimulants
Mainly metabolized:
atomoxetine

Partly metabolized:
amphetamine/
dextroamphetamine
dextroamphetamine
Cytochrome CYP 450 amitriptyline  Identifying patients who are poor, Roche AmpliChip; FDA approved
P450 Drug 2C19 citalopram intermediate, extensive, or ultrarapid Available in special
metabolizer clomipramine metabolizers of psychotropic drugs laboratories
status escitalopram metabolized by CYP 2D6 and 2C19
imipramine  Adjusting dosages for psychotropic
diazepam drugs that are metabolized by CYP
2D6, 2C19
CBZ-induced HLA-B*1502 carbamazepine  Identifying patients who are at Available in special FDA recommended for
SJS increased risk for CBZ-induced SJS laboratories patients of Asian
 Patients who test positive for descent started on CBZ
HLA-B*1502 may be at increased risk
of SJS/TEN from other antiepileptic
drugs that have been associated with
SJS/TEN. Therefore, in HLA-B*1502-
positive patients, consider avoiding use
of other antiepileptic drugs associated
with SJS/TEN when alternative therapies
are equally acceptable
Serotonin SLC6A4 fluoxetine  Predicting response time to Available in special
transporter fluvoxamine improvement with SSRIs (L-allele) laboratories
escitalopram  Identifying patients who have increased
sertraline risk of side effects under SSRI treatment
citalopram (S/S, S/L genotype)
paroxetine  Identifying patients who have reduced
amounts of the serotonin transporter
and thus might have an altered
response to SSRIs
 Evaluating patients who have failed
therapy with SSRIs and
who might respond favorably to
nonselective antidepressants (S-allele)
 Table 6.1 (cont.)

Genetic test Gene Drug Useful for Commercially Notes

available
Serotonin HTR2A fluoxetine  Guiding treatment choice of an SSRI or Available in special
receptor HTR2C fluvoxamine non SSRI antidepressant laboratories
escitalopram  Guiding treatment choice in individuals
sertraline who have drug-metabolizer
citalopram phenotypes discordant with CYP450
paroxetine genotypes
 Identifying patients who may benefit
from treatment with the atypical
antipsychotic clozapine
Clozapine- HLA-DQB1 Clozapine  Identifying patients who are at PGxPredict:Clozapine
induced increased risk for clozapine-induced
Agranulocytosis agranulocytosis
 guiding decisions about the frequency
of hematological monitoring, and about
treatment decisions in the face of falling
white blood cell counts
 a negative test does not eliminate the
need for blood monitoring
Abbreviations: CBZ, carbamazepine; FDA, US Food and Drug Administration; SJS, Stevens–Johnson syndrome; SSRI, selective serotonin reuptake inhibitor; TEN, toxic epidermal necrolysis.
 Chapter 6: Pharmacogenetics in psychiatry

between a polymorphism and treatment. Currently,
very little is known about the biological relevance
of most genetic polymorphisms and how genetic vari-
ation influences treatment response on the cellular
level. In the future, SNPs associated with drug treat-
ment response might lead to promising new drug
development, but first these findings must undergo
extensive validation and neurobiological investiga-
tion. To address these issues, there is a strong need
to develop appropriate and standardized methodolo-
gies for pharmacogenetic studies, as proposed by
some groups [36]. It is necessary to develop prospect-
ive large pharmacogenetic clinical trials to evaluate
the effects of genetic variants on treatment outcome
comprehensively. In addition, other questions should
be addressed, including investigation of the genetics
of placebo response, genetics of psychotherapy response,
or the genetics of treatment response in general.
Despite many obstacles, it is likely that genetic
patient information will influence clinical practice
in the very near future. Currently, there are only a
few commercial pharmacogenetic tests available
(Table 6.1) which can be ordered through a few com-
mercial and academic laboratories. The Roche Diag-
nostic AmpliChip CYP450 test was FDA approved in
2005 and provides genotypes for the two cytochrome
P450 genes CYP2D6 and CYP2C19. Theoretically,
by genotyping patients for variation in these genes,
the metabolizer status of a patient may be predicted
and this can be used to help guide medication choice
and dosing. This might be useful for example in a
depressed patient who has a history of being very
sensitive to antidepressant medication preferentially
metabolized through CYP2D6, or in a depressed
patient who appears treatment resistant despite
adequate dosing of antidepressant drugs. Currently
there are no universal recommendations regarding
which patients should get tested, but several sugges-
tions have been published [176, 177]. Limitations of
the AmpliChip test include false positive and false
negative results, an issue with every laboratory test,
but more importantly the test does not identify all
poor or ultra-rapid metabolizers, since some individ-
uals might have rare, previously unknown, gene
variants that drive their metabolizer status. In addi-
tion, cost effectiveness and long-term benefits for
patients taking antidepressant drugs have yet to be
established. Several laboratories now also offer the
HLA-B*1502 allele test for CBZ-induced SJS, and
additional tests are anticipated to become available
in the near future.
Although pharmacogenetics in clinical practice is
currently limited to “side effects” and “metabolism”,
comprehensive pharmacogenetic profiling is already a
reality in many pharmaceutical companies. Many
phase II and phase III trials now have genetic com-
ponents and the FDA has recently issued the first
genotype-based indications for warfarin [178]. This
paradigm shift offers great opportunities in identify-
ing particular subgroups of patients for which a
compound works especially well or causes severe side
effects. On the other hand, pharmaceutical companies
might limit their phase II and III trials to certain
“genetic” populations, in order to document safety
and efficacy, leaving patients with “complicated risk
genetics” out of trials and drug development. It is
thus important to develop comprehensive pharmaco-
genetic policies and regulations in order to avoid
misuse of genetic information [179]. While psychiatry
has entered the new area of pharmacogenetics, it is
important to remember that this new technology will
only provide additional information on one aspect
of the complex and personal history of psychiatric
patients. It is the sum of inside and outside factors
that contribute and influence mental pathology and
well-being.
Acknowledgements
This work was supported by the Center for Neuro-
biology and Behavior, Department of Psychiatry,
University of Pennsylvania. The author would like
to thank Thomas Ferraro for very helpful comments,
suggestions, improvements, and corrections.

References 4. Binder EB, et al. Ann Med

1. Kessler RC, et al. Arch Gen
Psychiatry 2005;62:617–627.
2. Rush AJ, et al. Am J Psychiatry
2006;163:1905–1917.
3. Trivedi MH, et al. Am J Psychiatry
2006;163:28–40.
2006;38:82–94.
5. Serretti A, et al.
Psychopharmacology (Berl)
2004;174:490–503.
6. Staddon S, et al.
Psychopharmacology (Berl)
2002;162:18–23.
7. Weinshilboum RM, et al. Annu
Rev Genomics Hum Genet
2006;7:223–245.
8. Dalen P, et al. Clin Pharmacol
Ther 1998;63:444–452.
9. Schenk PW, et al. Mol Psychiatry
2008;13:597–605.

65
 Chapter 6: Pharmacogenetics in psychiatry
10. Bertilsson L, et al. Br J Clin
Pharmacol 2002;53:111–122.
11. Charlier C, et al. Ther Drug Monit
2003;25:738–742.
12. Shams ME, et al. J Clin Pharm
Ther 2006;31:493–502.
13. Veefkind AH, et al. Ther Drug
Monit 2000;22:202–208.
14. Perry PJ, et al. J Clin
Psychopharmacol 1994;14:
230–240.
15. Gex-Fabry M, et al. Ther Drug
Monit 2008;30:474–482.
16. Alvares AP, et al. Drug Metab Rev
1979;9:185–205.
17. Serretti A, et al. Pharmacogenomics
J 2004;4:267–273.
18. Lesch KP, et al. Science
1996;274:1527–1531.
19. Gelernter J, et al. Am J Med Genet
1999;88:61–66.
20. Baune BT, et al. Depress Anxiety
2008;25:920–925.
21. Mrazek DA, et al. Am J Med Genet
B Neuropsychiatr Genet
2008;147B:1337–1344.
22. Serretti A, et al. Prog
Neuropsychopharmacol Biol
Psychiatry 2005;29:1074–1084.
23. Serretti A, et al. Mol Psychiatry
2007;12:247–257.
24. Lin E, et al. Pharmacogenomics
2008;9:935–946.
25. Perlis RH, et al. Arch Gen
Psychiatry 2007;64:689–697.
26. Rush AJ, et al. Control Clin Trials
2004;25:119–142.
27. McMahon FJ, et al. Am J Hum
Genet 2006;78:804–814.
28. Manji HK, et al. Nat Med
2001;7:541–547.
29. Strome EM, et al. Biol Psychiatry
2005;57:1004–1010.
30. Lekman M, et al. Biol Psychiatry
2008;63:1103–1110.
31. Paddock S, et al. Am J Psychiatry
2007;164:1181–1188.
32. Kraft JB, et al. Biol Psychiatry
2007;61:734–742.
66
33. Hu XZ, et al. Arch Gen Psychiatry
2007;64:783–792.
34. Laje G, et al. Am J Psychiatry
2007;164:1530–1538.
35. Ising M, et al. Arch Gen Psychiatry
2009;66:966–975.
36. Serretti A, et al. Pharmacogenomics
J 2008;8:90–100.
37. Uher R, et al. Psychol Med
2008;38:289–300.
38. Uher R, et al. Pharmacogenomics J
2009;9:225–233.
39. Murphy GM, Jr., et al. Am
J Psychiatry 2003;160:1830–1835.
40. Murphy GM, Jr., et al. Arch Gen
Psychiatry 2004;61:1163–1169.
41. Smits K, et al. Int Clin
Psychopharmacol 2007;22:
137–143.
42. FDA. http://www.fda.gov/Drugs/
DrugSafety/PostmarketDrugSafety
InformationforParients and
Providers/DrugSafetyInformation
forHealthcareProfessionals/Public
HealthAdvisories/ucm161679.htm;
2004.
43. Jick H, et al. JAMA 2004;292:
338–343.
44. Licinio J, et al. Nat Rev Drug
Discov 2005;4:165–171.
45. Gibbons RD, et al. Am J Psychiatry
2007;164:1356–1363.
46. Perlis RH, et al. Arch Gen
Psychiatry 2008;65:882–892.
47. Laje G, et al. Pharmacogenet
Genomics 2009;19:666–674.
48. Perroud N, et al.
Neuropsychopharmacology
2009;34:2517–2528.
49. Henderson DC. J Clin Psychiatry
2008;69:e04.
50. Scheen AJ, et al. Diabetes Metab
2007;33:169–175.
51. Meltzer HY, et al. J Clin Psychiatry
2002;63(Suppl 9):25–29.
52. Kane JM. Am J Psychiatry
2006;163:1316–1318.
53. Lieberman JA, et al. N Engl J Med
2005;353:1209–1223.
54. Miyamoto S, et al. Mol Psychiatry
2005;10:79–104.
55. Cascorbi I. Pharmacol Ther
2006;112:457–473.
56. Linnet K, et al. Eur
Neuropsychopharmacol 2008;
18:157–169.
57. Arranz MJ, et al. Mol Psychiatry
2007;12:707–747.
58. Spina E, et al. Basic Clin
Pharmacol Toxicol 2007;
100:4–22.
59. Urichuk L, et al. Curr Drug Metab
2008;9:410–418.
60. Kirchheiner J, et al. J Clin
Psychopharmacol 2006;26:
440–442.
61. Eiermann B, et al. Br J Clin
Pharmacol 1997;44:439–446.
62. Ring BJ, et al. J Pharmacol Exp
Ther 1996;276:658–666.
63. Murayama N, et al. J Pharmacol
Exp Ther 2004;308:300–306.
64. Sachse C, et al. Br J Clin
Pharmacol 1999;47:445–449.
65. Eap CB, et al. J Clin
Psychopharmacol 2004;24:214–219.
66. Ozdemir V, et al. J Clin
Psychopharmacol 2001;21:603–607.
67. Bondolfi G, et al. Ther Drug Monit
2005;27:539–543.
68. Buckley PF. J Clin Psychiatry
2007;68(Suppl 6):5–9.
69. Gardner DM, et al. CMAJ
2005;172:1703–1711.
70. Tamminga CA, et al. Mol
Psychiatry 2005;10:27–39.
71. Kapur S, et al. Am J Psychiatry
2000;157:514–520.
72. Mamo D, et al. Am J Psychiatry
2004;161:818–825.
73. Himei A, et al. Psychiatry Clin
Neurosci 2002;56:97–102.
74. Lencz T, et al. Am J Psychiatry
2006;163:529–531.
75. Wu S, et al. Neurosci Lett
2005;376:1–4.
76. Jonsson EG, et al. Mol Psychiatry
1999;4:290–296.

77. Lundstrom K, et al. Biochem
Biophys Res Commun
1996;225:1068–1072.
78. Joober R, et al. J Psychiatr Res
2000;34:285–291.
79. Scharfetter J, et al. Eur
Neuropsychopharmacol
1999;10:17–20.
80. Szekeres G, et al. Am J Med Genet
B Neuropsychiatr Genet
2004;124B:1–5.
81. Reynolds GP, et al. Eur
Neuropsychopharmacol
2005;15:143–151.
82. Meltzer HY, et al.
Arzneimittelforschung
1992;42:268–272.
83. Arranz M, et al. Lancet
1995;346:281–282.
84. Arranz MJ, et al. Schizophr Res
1998;32:93–99.
85. Joober R, et al. J Psychiatry
Neurosci 1999;24:141–146.
86. Lane HY, et al. Am J Psychiatry
2002;159:1593–1595.
87. Lin CH, et al. Neuroreport
1999;10:57–60.
88. Yu YW, et al. Neuropsychobiology
2001;43:79–82.
89. Arranz MJ, et al. Neurosci Lett
1996;217:177–178.
90. Masellis M, et al.
Neuropsychopharmacology
1998;19:123–132.
91. Correll CU, et al. Schizophr Res
2007;89:91–100.
92. Theisen FM, et al. Psychiatr Genet
2005;15:285–289.
93. Wehmeier PM, et al. Psychiatry
Res 2005;133:273–276.
94. Ellingrod VL, et al. Am J Med
Genet B Neuropsychiatr Genet
2005;134B:76–78.
95. Miller DD, et al. Am J Med Genet
B Neuropsychiatr Genet
2005;133B:97–100.
96. Reynolds GP, et al.
J Psychopharmacol 2006;20:15–18.
97. Reynolds GP, et al. Am
J Psychiatry 2003;160:677–679.
Chapter 6: Pharmacogenetics in psychiatry
98. Templeman LA, et al.
Pharmacogenet Genomics
2005;15:195–200.
99. Reynolds GP, et al. Lancet
2002;359:2086–2087.
100. Muller DJ, et al. Pharmacogenomics
2006;7:863–887.
101. Muller DJ, et al. Pharmacogenomics
J 2004;4:77–87.
102. Liao DL, et al. Neuropsychobiology
2001;44:95–98.
103. Steen VM, et al. Mol Psychiatry
1997;2:139–145.
104. Segman R, et al. Mol Psychiatry
1999;4:247–253.
105. Lerer B, et al.
Neuropsychopharmacology
2002;27:105–119.
106. Bakker PR, et al. Schizophr Res
2006;83:185–192.
107. Munro J, et al. Br J Psychiatry
1999;175:576–580.
108. Amar A, et al. Int
J Neuropsychopharmacol
1998;1:41–44.
109. Dettling M, et al. Pharmacogenomics
J 2007;7:325–332.
110. Dettling M, et al. Pharmacogenetics
2001;11:135–141.
111. Lieberman JA, et al.
Arch Gen Psychiatry
1990;47:945–948.
112. Valevski A, et al. Eur
J Immunogenet 1998;
25:11–13.
113. Bowden CL. Expert Rev Neurother
2007;7:9–16.
114. Ettinger AB. Neurology
2006;67:1916–1925.
115. Post RM, et al. Neuropsychobiology
1998;38:152–166.
116. Ferraro TN, et al. Epilepsy Behav
2005;7:18–36.
117. Fukasawa T, et al. J Clin Pharm
Ther 2007;32:333–341.
118. Cade JF. Med J Aust 1949;
2:349–352.
119. Schou M. Nervenarzt
1983;54:331–339.
120. Lohoff FW, et al. In Charney D,
et al. (eds.). Neurobiology of
Mental Illness. Oxford: Oxford
University Press; 2008.
121. Fountoulakis KN, et al. Int
J Neuropsychopharmacol
2008;11:999–1029.
122. Mendlewicz J, et al. Am
J Psychiatry 1973;130:1011–1013.
123. Grof P, et al. J Affect Disord
1994;32:85–95.
124. Prien RF, et al. Arch Gen
Psychiatry 1974;31:189–192.
125. Smeraldi E, et al. J Affect Disord
1984;6:139–151.
126. Morissette J, et al. Am J Med Genet
1999;88:567–587.
127. Turecki G, et al. Mol Psychiatry
2001;6:570–578.
128. Coyle JT, et al. Nat Med
2002;8:557–558.
129. Berridge MJ, et al. Cell
1989;59:411–419.
130. Klein PS, et al. Proc Natl Acad Sci
U S A 1996;93:8455–8459.
131. Benedetti F, et al. Neurosci Lett
2004;355:37–40.
132. Benedetti F, et al. Neurosci Lett
2005;376:51–55.
133. Szczepankiewicz A, et al. World
J Biol Psychiatry 2006;7:158–161.
134. Adli M, et al. Biol Psychiatry
2007;62:1295–1302.
135. Mamdani F, et al.
Pharmacogenomics J 2004;
4:161–170.
136. Serretti A, et al. Am
J Pharmacogenomics 2003;3:
17–30.
137. Serretti A, et al. Psychiatry Res
1999;87:7–19.
138. Serretti A, et al. Am J Med Genet
2002;114:370–379.
139. Serretti A, et al. J Psychiatr Res
2000;34:89–98.
140. Benmessaoud D, et al. BMC
Psychiatry 2008;8:40.
141. Bremer T, et al. Mol Diagn Ther
2007;11:161–170.
67
 Chapter 6: Pharmacogenetics in psychiatry
142. Cavazzoni P, et al. Psychiatry Res
1996;64:91–96.
143. Dmitrzak-Weglarz M, et al.
Pharmacogenomics 2008;9:
1595–1603.
144. Dmitrzak-Weglarz M, et al.
Pharmacol Rep 2005;57:761–765.
145. Duffy A, et al. J Psychiatry
Neurosci 2000;25:353–358.
146. Mamdani F, et al. Am J Med Genet
B Neuropsychiatr Genet
2008;147B:500–504.
147. Masoliver E, et al.
Psychiatr Genet 2006;
16:25–29.
148. Masui T, et al. Int
J Neuropsychopharmacol
2006;9:83–88.
149. Masui T, et al. Psychiatr Genet
2006;16:49–50.
150. Masui T, et al. Prog
Neuropsychopharmacol Biol
Psychiatry 2008;32:204–208.
151. Michelon L, et al. Neurosci Lett
2006;403:288–293.
152. Rybakowski JK, et al. Pharmacol
Rep 2005;57:124–127.
153. Rybakowski JK, et al.
Pharmacopsychiatry 2005;
38:166–170.
68
154. Serretti A, et al. J Psychiatr Res
1999;33:371–377.
155. Serretti A, et al.
Pharmacogenomics J 2001;
1:71–77.
156. Squassina A, et al. Pharmacol Res
2008;57:369–373.
157. Turecki G, et al. Psychiatry Res
1996;63:17–23.
158. Zill P, et al. Psychiatr Genet 2003;
13:65–69.
159. Baum AE, et al. Mol Psychiatry
2008;13:197–207.
160. Mockenhaupt M, et al. Neurology
2005;64:1134–1138.
161. Rzany B, et al. Lancet
1999;353:2190–2194.
162. Tennis P, et al. Neurology
1997;49:542–546.
163. Bastuji-Garin S, et al. Arch
Dermatol 1993;129:92–96.
164. Chung WH, et al. Nature
2004;428:486.
165. Hung SI, et al. Pharmacogenet
Genomics 2006;16:297–306.
166. Man CB, et al. Epilepsia
2007;48:1015–1018.
167. Locharernkul C, et al. Epilepsia
2008;49:2087–2091.
168. Alfirevic A, et al. Pharmacogenomics
2006;7:813–818.
169. Lonjou C, et al. Pharmacogenet
Genomics 2008;18:99–107.
170. Ferrell PB, Jr., et al.
Pharmacogenomics 2008;9:
1543–1546.
171. Lonjou C, et al. Pharmacogenomics
J 2006;6:265–268.
172. FDA. Safety Alerts for Drugs,
Biologics, Medical Devices, and
Dietary Supplements. Silver
Spring, MD: US Food and Drug
Administration; 2008.
173. Charney DS, et al. Arch Gen
Psychiatry 2002;59:262–270.
174. Khan A, et al. J Psychiatr Res
2008;42:791–796.
175. Montgomery SA. Eur
Neuropsychopharmacol
1999;9:265–269.
176. de Leon J, et al. Psychosomatics
2006;47:75–85.
177. Mrazek DA. CNS Spectr
2006;11:3–4.
178. Tan GM, et al. Pharmacogenomics
2010;11:439–448.
179. Katsanis SH, et al. Science
2008;320:53–54.
 Functional validation of candidate
Chapter
7 genetic susceptibility factors for major
mental illnesses
From protein chemistry, cell biology,
and animal studies, to human brain imaging
Akira Sawa, Wanli W. Smith, Saurav Seshadri, Akiko Hayashi-Takagi,
Hanna Jaaro-Peled, and Atsushi Kamiya
Introduction
Genetic susceptibility factors for major mental illnesses,
such as schizophrenia (SZ) and bipolar disorder, have
now become available due to the long-term efforts of
psychiatric genetics [1, 2]. Recent progress in whole
genome association studies, and work on copy number
variations have further enhanced our knowledge of
such genetic factors [3–5]. In contrast to Huntington’s
disease (HD) in which a specific mutation in a single
gene, huntingtin (htt), causes the disease [6], none of the
susceptibility genes for major mental illnesses available
thus far is such a direct causal factor. Instead, com-
binations of genetic factors together with environ-
mental stressors underlie the pathogenesis of major
mental illnesses [7]. Genetic variants in some of these
factors occur commonly and result in mild risk in the
functional context; whereas other genetic variations
occur very rarely or de novo, and result in greater
impact on biological processes underlying disease
pathology, though they are not causal per se [8].
How can we utilize such genetic information for a
better understanding of the biology underlying disease
pathology? The goal of this chapter is to address this
important question. Before such genes became available,
molecular studies for major mental illnesses had been
limited to pharmacological approaches, in which cellular
responses to drugs used in clinical settings were studied.
We will review functional studies based on genetic
information for better understanding of major mental
illnesses, such as SZ, by using the many biological
Principles of Psychiatric Genetics, eds John I. Nurnberger, Jr. and Wade H. Berrettini. Published by Cambridge University
Press. © Cambridge University Press 2012.
studies on disrupted in schizophrenia 1 (DISC1) over
the past decade as examples [7, 9]. The gene coding
for DISC1 was originally identified at the breakpoint
of a balanced chromosomal translocation in a Scottish
pedigree in which many family members suffer from
mental illnesses, such as major depression, SZ, and
bipolar disorder [10]. Genetic studies in the general
population have indicated that DISC1 affects endophe-
notypes that underlie several major mental illnesses
[7, 9]. Consequently, we can find genetic associations of
DISC1 with a wide range of psychiatric disorders [7, 9].
Functional studies examining how such genetic
factors influence the pathology of major mental illnesses
are conducted at many levels. Most fundamentally, bio-
chemical studies on protein interaction among these
factors are performed. Such molecular mechanisms are
first validated and characterized in cell models. Roles for
such molecular mechanisms in the neuronal circuitry
are then tested in animal models. Finally, brain imaging
studies allow us to examine how genetic variations are
correlated with human brain functions in vivo.
Protein chemistry
Once genetic studies identify candidate susceptibility
genes, the first step of functional studies is to charac-
terize the proteins that are encoded by these genes.
As described above, no one factor by itself can cause
major mental illnesses. Instead, combinations of more
than one factor result in the diseases. Thus, a key
question in psychiatric genetics is how such genetic
69
 Chapter 7: Validation of candidate genetic susceptibility factors

Neurotrophic factors
BDNF EGFR
NT-3
NT-4
IGF-1
TRK EGFR AMPAR LICAM NMDAR EPHB2 DOPAMINE RECEPTORS GPCR

AKAP9
ADENYLATE CYALASE
KALRN
INTEGRIN CAMKK2 SNAP91 CDK5RAP3
RAC
MAP3K71P2

AB12 FLJ13386 CDKS cAMP

SPTBN1
NEF3 NCK
SOS1

SPTAN1 SNX6 PDE4B

TNIK
NDEL1 GNB1
TRAF3IP1
RAS SH3BP5 CDCSL
PAFAH1B1
ANK2
CLU

JNK2 JNK3 JNK1 DST

SEC3L1 DISC1 SRGAP3

TRIM9
MACF1
RHO GTPASES
TRI0
SNAP25

DMD
RHO GTPASES

Figure 7.1 Protein interactome simulated by pathway analyses. Iterative yeast-two hybrid screens, combined with detailed pathway, and
functional analyses suggest DISC1 interacts with proteins involved in key processes involved in neurodevelopment. Deficits in these processes
may underlie decreased dendritic branching, arborisations and neuropil size observed pathologically. (Adapted from [16], with permission.)
See plate section for color version.

susceptibility factors may have molecular and/or
functional interactions, which may result in the
pathology. To explore possible protein interactions,
a yeast two-hybrid assay is frequently used [11]. This
assay is especially powerful in detecting intracellular
protein interactions. Protein interactions suggested by
yeast two-hybrid need to be further confirmed by
biochemical approaches, such as in vitro protein
binding assays with purified recombinant proteins
and co-immunoprecipitation. The former informs us
whether such protein interactions are direct or indirect.
By utilizing site-directed mutagenesis as well as gener-
ating several deletion mutants, we can further deter-
mine which small domains of the proteins are required
for their interaction. This information is very useful in
addressing the functional significance of the protein
interactions, which are validated in cell and in vivo
animal models, as described in the following subsections.
In addition to biochemical approaches, protein inter-
actions in the cellular context are assessed by their
co-localization by using confocal microscopy and
dual-label electron microscopy.
DISC1 is an intracellular protein that contains
several coiled-coil domains. This structure suggested
that DISC1 might be an anchoring molecule that
would interact with many proteins, but at that time
no concrete indication of its function was available.
Therefore, several groups conducted yeast two-hybrid
screening, and have identified many interesting inter-
actors, such as nuclear distribution element-like
(NDEL1), phosphodiesterase 4 (PDE4), and pericen-
triolar material 1 (PCM1) [12–15]. One of the most
extensive studies using this approach is the work by
Camargo et al. in which the possible interactome of
DISC1 is suggested [16] (Figure 7.1). Importantly,
many of these protein interactors are also genetic
susceptibility factors for major mental illnesses pro-
posed by independent genetic studies [17–20].
At present, many investigators agree on a general
concept that genetic susceptibility factors for mental

70
 Chapter 7: Validation of candidate genetic susceptibility factors
illnesses frequently interact with each other, forming
meaningful “pathways” [7, 21].
Cell biology
Once genetic studies identify candidate susceptibility
factors for the diseases, functions of such proteins can
be tested in cells by modulating expression of the target
molecules or by expressing their genetic variants.
Molecular manipulations
There are three types of molecular manipulations
in cells. First, we can simply overexpress molecules
of interest exogenously. Overexpression of wild-type
proteins may be useful in addressing the cellular
effects of genomic duplication. If nonsynonymous
genetic variations are available, overexpression of
such mutants may address the effects of the genetic
variations. Especially if biological significance of such
genetic variants is the result of a dominant-negative
mechanism, we frequently see meaningful phenotypic
changes following overexpression. Second, we can
knockdown expression of target molecules by RNA
interference (RNAi) or other approaches. This knock-
down approach is regarded to be the best way to
address the physiological roles of target molecules.
As described below, if knockout mice are available,
cells primarily prepared from such animals can be
used for this aim. A knock-in approach, in which
endogenous wild-type molecules are replaced by gen-
etic variants, is the third strategy. To address subtle
functional differences between wild-type mice and
mutants, this approach is more sensitive and more
closely reflects endogenous conditions compared with
mere overexpression of genetic variants. A knock-in
approach can be used to address the significance of
protein interactions. Phenotypes elicited by knock-
down expression of a target protein (protein X) can
be rescued (normalized) by overexpressing wild-type
protein X. If interaction of proteins X and Y is crucial,
mutant protein X that lacks the domain required for
binding with protein Y may not rescue the phenotypes
elicited by knockdown of endogenous protein X.
Such molecular manipulations are usually per-
formed by transfection of plasmids or infection of
viral vectors, and their effects are assayed within
several days. In contrast to this type of “transient”
manipulation, some studies utilize “stable” expression
cell models. Although lower throughput and labor-
intensive, these stable models have several advantages.
Because these cells are selected from homogenous
colonies with ectopic plasmids (either for overexpres-
sion or knocking down the targets), the potential to
expand the manipulated cells in a large scale and
homogeneous manner is an advantage of this system.
Thus, stable expression cell models can be effectively
used for compound screening.
Types of cells used for assays
For molecular manipulations, cell lines are frequently
used. In general, we can introduce plasmids in cell
lines by transfection with a high efficiency. For
example, PC12 cells, which can be differentiated from
a proliferating to postmitotic status by reducing
serum concentration and adding nerve growth factor
in media, are frequently used to address questions of
how target molecules may play roles in neurite out-
growth. However, while cell lines are easy to use, they
may not truly reflect endogenous conditions, as the
artificial processes used in their production, such as
immortalization, may affect cellular signaling.
To address this concern with cell lines, primary
cell cultures are useful. Established protocols for
neuron, astrocyte, microglia, and oligodendrocyte
cultures are available. If necessary, mixed culture with
different types of cells is also possible. In recent years,
investigators have paid more attention to the use of
stem cells. A technical difficulty in using these cells
(primary and stem cells) in molecular manipulations is
the relatively lower transfection efficiency compared
with that for cell lines. To overcome this issue, virus-
mediated gene transfer with vectors of lentivirus,
retrovirus, and adeno-associated virus (AAV) is used.
In addition, in molecular psychiatry, it is espe-
cially important to consider using cells derived from
patients and controls. Many studies have used periph-
eral cells from human subjects, such as lymphocytes,
lymphoblasts, and fibroblasts. To obtain neuronal-
like cells in a minimally invasive manner, olfactory
neurons obtained via nasal biopsy can be useful [22].
Furthermore, recent progress in stem cell biology
allows us to establish induced pluripotent stem cells
(iPS cells) from peripheral cells, such as fibroblasts
[23]. Although there are still many technical barriers,
this iPS cell approach is regarded as the most promis-
ing strategy to obtain any type of cells in the central
nervous system in the near future. A derivative of this
approach is the direct transformation of fibroblasts
into neurons [24].
71
 Chapter 7: Validation of candidate genetic susceptibility factors
Table 7.1 In vitro cellular assays relevant to mental illnesses. Examples of possibly altered cellular events in individuals with mental
disorders, and assays used for detection of those deficits in cellular model system in vitro are shown.
Cellular event
Neurotransmitters Production
Release
Uptake
Neurodevelopment Proliferation
Migration
Neurite outgrowth
Dendritic arborization
Synaptogenesis/reorganisation
Metabolism Oxidative stress
Readouts relevant to mental disorders
(Table 7.1)
What are the relevant cellular assays associated with
mental illnesses? Mechanisms associated with neuro-
transmitter production, release, and uptake may be
the first category of candidates. Recent studies of
brain imaging and neuropathology in patients
suffering from mental illnesses have suggested that
there are anatomical and histological abnormalities
originating from neurodevelopment. Thus, the key
components underlying these changes, such as cell
proliferation, cell migration, neurite outgrowth, den-
dritic arborization, and synaptic spine formation, are
the targets to be examined. Accumulating evidence in
clinical studies has also pointed out that mental ill-
nesses, especially schizophrenia, are associated with
intrinsic abnormalities in metabolic signaling [25]. In
accordance with this notion, some studies have pro-
posed a role for oxidative stress in mental illnesses
[26]. Thus, effects of target molecules on cellular
responses to metabolic and oxidative stress may be
candidate readouts relevant to mental disorders.
Cell biology in addressing roles
of genetic factors for mental illnesses:
DISC1 studies as examples
When the gene coding for DISC1 was first reported
from a Scottish pedigree [10], nobody knew its cellu-
lar function. As described above, structural prediction
from the deduced amino acids suggested that it is an
intracellular anchoring molecule that interacts with
72
Assay
Concentration measurement in cell lysate
Concentration measurement in culture media
Fluorescence-based assay, Synaptosomal assay
BrdU cell proliferation assay
Wound healing assay, Boyden chamber assay
Measurement of neurite length (neuroblastoma such as PC12,
SHSY-5Y cells)
Scholl assay (primary neuron culture)
Morphometry analysis (primary neuron culture)
Measurement for DNA/RNA damage, oxidative protein, ROS,
and lipid peroxidation
many proteins. Indeed, yeast two-hybrid screening
and biochemical validations have identified many
proteins associated with DISC1. Although no func-
tion of DISC1 was originally available, cellular roles
for these protein interactors were known. Thus, to
address the possible function of DISC1, investigators
used assays that were relevant to the interactors. For
example, neurite outgrowth was used based on the
observation that DISC1 binds with cytoskeletal regula-
tors, such as fasciculation and elongation protein zeta-1
(FEZ1), lissencephaly 1 (LIS1), and NDEL1 [12, 27–29].
Interaction of PDE4 with DISC1 was the rationale
that investigators used in choosing assays of cellular
response to cyclic adenosine monophosphate (cAMP)
signaling [20, 30]. Likewise, assays of cell proliferation
and Wnt signaling were based on interaction of GSK3b
with DISC1 [31]. Assays of cAMP and GSK3b/Wnt
signaling are also relevant in the light of clinical
pharmacology. Aberrant connectivity of neuron net-
works is believed to underlie the pathology of major
mental illnesses; especially in schizophrenia, decrease in
synaptic spine density is reproducibly reported [32].
A role for DISC1 in spine reorganization has been
established in primary neuron cultures [33]. In addi-
tion, cell models can be used to address the functional
relationships of genetic susceptibility factors, even if
they do not directly bind at molecular levels: one
example is a study that demonstrates the effects of
Neuregulin-1 on expression of a DISC1 isoform [34].
Animal models
Although cell models are useful to identify the signal
transduction pathways involving genetic susceptibility
 Chapter 7: Validation of candidate genetic susceptibility factors
factors that result in cellular manifestations, such
as neurite outgrowth, cell proliferation, and synaptic
spine reorganization, they cannot address the most
important question of psychiatry directly; that is, how
do genetic susceptibility factors affect formation and
functions of neuronal circuitry in the brain in vivo?
Therefore, we need to generate animal models based
on information from psychiatric genetics, in which
knowledge from cell biology should be fully utilized.
Rodents, especially mice, are the species in which
genetic engineering technology is the most advanced.
Thus, in testing brain functions of genetic suscepti-
bility factors for major mental illnesses, mice are the
first choice.
The onset of major mental illnesses (addictions,
SZ, mood disorders, and related conditions) occurs
in adolescence or young adulthood. However, accu-
mulating evidence has suggested neurodevelopmen-
tal etiology of these disorders, especially in SZ [35,
36]. Recent genetic studies indicate that some of the
most promising susceptibility factors of these dis-
eases, such as Neuregulin-1, ErbB4, and DISC1, play
important roles in development [37, 38]. Thus, gen-
etic model mice in which such development-related
disease-associated factors are altered may be useful in
characterizing how the disease etiologies develop over
time until development of full-blown disease [7]. It is
likely that nonlethal but significant insults in early
development affect postnatal brain maturation, which,
in turn, results in manifestation of psychiatric symp-
toms after puberty. Early intervention in these diseases,
including preventive medication to prodromal subjects
is a hot topic in translational research [39]. To address
these questions, although we could design longitudinal
studies with human subjects, research with genetic
mouse models has significant advantages.
There is still debate about whether it is possible to
use rodents to model psychiatric disorders in which
high brain functions, that are probably in part unique
to humans, are impaired [40]. Thus, genetic manipu-
lations in primates by using AAV technology com-
bined with stereotaxic injection into targeted brain
regions are attempted [41]. Although the primate
model has high potential, we should also acknowledge
its current limitations, including high costs, shortage
of experimental tools, and ethical concerns.
On the other hand, investigators have also paid
attention to zebrafish (Danio rerio), a vertebrate lower
than rodents, as a model for functional studies due to
its advantages of low cost maintenance, rapid life
cycle, optical clarity, rapid external embryonic devel-
opment, and multiple molecular genetic techniques.
A recent paper by Wood [42] studied the function
of DISC1 and Neuregulin-1 in zebrafish, suggesting
that these factors affect similar neurodevelopmental
processes. With similar reasons, fruit fly Drosophila
melanogaster may be useful in testing functions of
genetic susceptibility factors for mental illnesses.
In particular, many outstanding studies in memory,
sleep, and circadian cycle have been done using the
fruit fly. For example, Swamura et al. [43] reported
that transgenic files overexpressing DISCI display
disturbance homeostasis.
In this chapter, we will describe rodent models
with manipulations for genetic susceptibility factors
of mental illnesses in greater detail.
Molecular manipulations in rodents
As in cell models, there are three types of molecular
manipulations in animal models. The first strategy is
to establish gain-of-function models by exogenously
overexpressing wild-type or mutant proteins under
the control of appropriate promoters. The second
one is to establish loss-of-function models in which
target genes are knocked out or their expressions are
knocked down. The third strategy is knock-in, in
which endogenous wild-type molecules are replaced
by genetic variants.
Mouse genetic engineering techniques allow us to
generate hereditary mouse models, such as transgenic,
knockout, and knock-in mice in which these three
types of molecular manipulations are performed,
respectively. In transgenic mice, exogenous protein
expression can be spatially and temporally controlled
by choosing an appropriate promoter. Such pro-
moters include the a-CaMKII promoter mainly for
the pyramidal neurons and the GAD67 promoter for
interneurons. Furthermore, by combining this with a
tetracycline-inducible system, an additional temporal
control in expression is available [44]. In knockout
mice, spatial and temporal control of knockout
becomes possible by utilizing the Cre-loxP system
[45]. Such spatial and temporal expression control is
very important in considering neural circuitry-
dependent functional validation.
In addition, by injecting plasmids or virus-mediated
constructs, expression of target molecules can be modu-
lated (Figure 7.2) [46, 47]. First, in utero gene transfer
is a technique to modulate expression of target genes
73
 Chapter 7: Validation of candidate genetic susceptibility factors

(a) Embryo Figure 7.2 Brain region specific

gene manipulations. (a) Schematic
representation of bilateral in utero
injection of constructs followed by their
E14 incorporation by electroporation into
Injection progenitor cells in the ventricular zone at
– + embryonic day 14 (E14). Migrating cells
+ + with green fluorescent protein (GFP) are
visualized at E18 after injection of a GFP
expression construct. (Adapted from [46],
with permission.) (b) Stereotaxic
coordinates and actual injection of
lentivirus-based enhanced GFP (EGFP).
– – Stereotaxic coordinates were determined
from the rat brain in stereotaxic
coordinates. The coordination in
E18 Sprague–Dawley rat at postnatal day 21
was stereotaxically injected with lentivirus
+ – containing EGFP at the coordinates of
AP = +2.2; ML = +0.9; DV = +2.0, +1.5,
+1.0 from the bregma. The crosses
indicate injected sites. Cx, cerebral
cortex; Hip, hippocampus; Th, thalamus;
Str, striatum; Amy, amygdala; Hypo,
hypothalamus; VTA, ventral tegmental
area; P, pontine tegmentum. (Adapted
from [47], with permission.) See plate
section for color version.

(b) Adult brain

medial lateral
0 1 2 3 4 mm

Cx
Cx Hip

Str Th
Amy

AP +2.2 mm AP –1.34 mm
anterior bregma posterior
3 2 1 0 –1 –2 –3 –4 –5 –6 –7 –8 mm
0 dorsal
Cx 1
Hip 2
Str 3
Th
VTA 4
Hypo 5
6 ventral

by introducing expression or RNAi plasmid(s) in the
developing brain [48]. Expression of more than one
gene can be simultaneously modulated with this tech-
nique. This feature is particularly advantageous in
studying genetic roles for mental disorders, in which
a combination of multiple genes plays a role in the
etiology. Temporal expression control again becomes
possible by using an inducible system [49]. Regional
specific expression can be controlled by changing the
orientation and position of the electrodes used in the
procedures. Second, sterotaxic injection of virus-
mediated expression and/or RNAi constructs in the
brain is another approach to modulate expression of
target molecules in a temporally and spatially specific
manner [47, 50].
Readouts relevant to mental disorders
Functional outcomes in the brain elicited by genetic
susceptibility factors for major mental illnesses are

74
 Chapter 7: Validation of candidate genetic susceptibility factors

Table 7.2 Behavioral assays relevant for different assayed in several ways. As previously published,
neuropsychological domains, which are used for testing
of mouse models of psychiatric diseases.
both histological/anatomical and behavioral assays
are utilized (Table 7.2). When genetic factor(s)
Putative Characteristic Typical test are modulated, it is important to address their
domain causal links to the observed phenotypes by both
Cognitive Working memory Delayed nonmatch cell autonomous and noncell autonomous/circuitry
to place mechanisms.
Behavioral Reversal learning Some of the most promising genetic factors for
flexibility major mental illnesses most likely play roles during
Learning and Water maze neurodevelopment. To address how the effects of
memory genetic factors in early development result in adult
Attention Latent inhibition phenotypes, we may need to consider postnatal
Cognitive/ Sensorimotor Prepulse inhibition brain maturation [7]. Especially in adolescence, sev-
positive gating eral key characteristics in the brain are dynamically
Positive Novelty induced Open field changed (Figure 7.3). Thus, when we consider read-
hyperactivity outs, their age-dependent progression should be
Hypersensitivity to Methamphetamine examined.
psychostimulants induced hyperactivity Molecular profiling studies with human cells
in the open field and tissues have identified genes and proteins that
Negative Social interaction Three chamber social are differentially expressed between normal controls
interaction test and patients. It is important to examine whether
Anhedonia Sucrose consumption such molecules are also differentially expressed
when genetic factors are modulated in animals. Such
Depressive Behavioral despair Forced swim test
study design opens a window for translational
Anxiety Anxiety Elevated plus maze research.

Figure 7.3 Dynamic changes in the adolescent brain. Disturbances generated by susceptibility genes and environmental insults during early
development (three stars on the left-hand side) may impair some of the crucial processes in early development, including progenitor cell
proliferation, neuronal migration and dendritic arborization and outgrowth. Independent of such initial risks/insults, intrinsic disease-associated
factors might also directly affect postnatal brain maturation (two central stars) contributing to the emergence of schizophrenia (SZ) in
young adulthood. (Adapted from [7], with permission.) See plate section for color version.

75
 Chapter 7: Validation of candidate genetic susceptibility factors

(a) Human Mouse

Control SZ 5

Ventricle volume (mm3)

4

0
WT Tg

100
WT
(b)
WT Tg Tg
80

Cells/section
Control SZ
nCI/g
2 080
60
1 200 PV
860
620
390 40
210
90
0 1 mm 1 mm
0 PV mRNA 20
CB
0
PV CB
Figure 7.4 Anatomical and histological abnormalities in schizophrenia patients and in mouse models. (a) Enlarged lateral ventricles as
detected by in vivo magnetic resonance imaging (MRI) in schizophrenia patients (left) and in a DISC1 mouse model (right). (b) Decreased
parvalbumin (PV) expression in the prefrontal cortex of schizophrenia patients (mRNA, left) and a DISC1 mouse model (right).
(Adapted from [60], with permission.) See plate section for color version.

Animal models in addressing roles
of genetic factors for mental illnesses:
DISC1 studies as examples
Mice with eight distinct genetic manipulations on
DISC1 have been published. Gogos and colleagues
[51] published a model in which a spontaneous muta-
tion in the DISC1 gene is utilized in combination with
genetic manipulation, achieving knockdown of some
types of DISC1 isoform. Clapcote and colleagues [52]
selected two lines with point mutations in the DISC1
gene from ENU mutagen-treated animals. Four
groups have generated transgenic mice expressing
dominant-negative mutants of DISC1, two of which
are under an inducible expression system [53, 54].
Most recently, a model generated by in utero gene
transfer in which transient knockdown of DISC1 in
the pre- and perinatal stages, specifically in a lineage
of pyramidal neurons mainly in the prefrontal cortex,
was reported [46].
In these animals, many behavioral assays have been
conducted. As far as reported, genetic manipulations
of DISC1 seem to result in behavioral deficits, which
include endophenotypes associated with both schizo-
phrenia and mood disorders [55]. In SZ, several histo-
logical and anatomical changes are reported [56].
These include enlargement of lateral ventricles [57]
and interneuron deficits, which is easily detected by a
decrease in the immunoreactivity of parvalbumin [58].
By using in vivo or ex vivo MRI scans, enlarged lateral
ventricles or shrinkage of some brain regions, consist-
ent with reports from patients with SZ, were observed
in several DISC1 mice [52, 53, 59, 60] (Figure 7.4a).
Moreover, immunohistochemistry revealed that there
was a selective decrease in the immunoreactivity of par-
valbumin in different types of DISC1 mice [52, 59, 61]
(Figure 7.4b).
The DISC1 model generated by in utero gene
transfer is a good example to present the importance
of assays in different developmental ages [54]. In this
model, although dendritic pathology exists at post-
natal day 14 (P14) due to transient knockdown of
DISC1 in early development, no robust changes in
neurochemistry and behavior are observed before

76
 Chapter 7: Validation of candidate genetic susceptibility factors

puberty (at P28). Interestingly, selective disturbance
of dopaminergic neurotransmission and associated
behavioral alterations become prominent after puberty
(at P56 and later).
Human studies with brain imaging
A series of studies by Weinberger and associates has
pioneered the possible correlation of brain dysfunc-
tion with genetic variations in susceptibility factors
associated with mental illnesses [62, 63]. This approach
is very important in addressing genetic effects on
human functions directly. Nonetheless, these studies
are limited to descriptively presenting correlative
relationships. To identify mechanistic links from
genetic factors to the phenotypes, especially those
observed during brain development and maturation,
a combination of human studies with animal
experiments is expected.
Concluding remarks
In this chapter, we describe how functions of genetic
susceptibility factors can be validated, specifically
using DISC1 as an example. Studies at multiple levels,
from protein chemistry, cell biology, animal study, to
clinical work will provide comprehensive understand-
ing of the functions of susceptibility factors.

References 16. Camargo LM, et al. Mol Psychiatry

1. Allen NC, et al. Nat Genet 2008;
40(7):827–834.
2. Sun J, et al. Am J Med Genet
B Neuropsychiatr Genet 2008;
147B(7):1173–1181.
3. Cook EH, Jr., et al. Nature 2008;
455(7215):919–923.
4. Sebat J, et al. Trends Genet 2009;
25(12):528–535.
5. Cichon S, et al. Am J Psychiatry
2009;166(5):540–556.
6. The Huntington Disease
Collaborative Research Group.
Cell 1993;72(6):971–983.
7. Jaaro-Peled H, et al. Trends
Neurosci 2009;32(9):485–495.
8. Manolio TA, et al. Nature
2009;461(7265):747–753.
9. Chubb JE, et al. Mol Psychiatry
2008;13(1):36–64.
10. Millar JK, et al. Hum Mol Genet
2000;9(9):1415–1423.
11. Fields S, et al. Nature 1989;
340(6230):245–246.
12. Ozeki Y, et al. Proc Nat Acad Sci
U S A 2003;100(1):289–294.
13. Morris JA, et al. Hum Mol Genet
2003;12(13):1591–1608.
14. Miyoshi K, et al. Biochem Biophys
Res Commun 2004;317(4):
1195–1199.
15. Millar JK, et al. Biochem Biophys
Res Commun 2003;311(4):
1019–1025.
2007;12(1):74–86.
17. Gurling HM, et al. Arch Gen
Psychiatry 2006;63(8):844–854.
18. Burdick KE, et al. Hum Mol Genet
2008;17(16):2462–2473.
19. Tomppo L, et al. Biol Psychiatry
2009;65(12):1055–1062.
20. Millar JK, et al. Science 2005;
310(5751):1187–1191.
21. Harrison PJ, et al. Mol Psychiatry
2005;10(1):40–68.
22. Tajinda K, et al. Mol Psychiatry
2010;15(3):231–232.
23. Takahashi K, et al. Cell 2006;
126(4):663–676.
24. Vierbuchen T, et al. Nature
2010;463(7284):1035–1041.
25. Holmes E, et al. PLoS Med 2006;
3(8):e327.
26. Do KQ, et al. Curr Opin Neurobiol
2009;19(2):220–230.
27. Kamiya A, et al. Nat Cell Biol
2005;7(12):1167–1178.
28. Kamiya A, et al. Hum Mol Genet
2006;15(22):3313–3323.
29. Miyoshi K, et al. Mol Psychiatry
2003;8(7):685–694.
30. Murdoch H, et al. J Neurosci
2007;27(35):9513–9524.
31. Mao Y, et al. Cell 2009;136(6):
1017–1031.
32. Glantz LA, et al. Arch Gen
Psychiatry 2000;57(1):
65–73.
33. Hayashi-Takagi A, et al. Nat
Neurosci 2010;13(3):327–332.
34. Seshadri S, et al. Proc Natl
Acad Sci U S A 2010;107(2):
5622–5627.
35. Lewis DA, et al. Ann Rev Neurosci
2002;25:409–432.
36. Rapoport JL, et al. Mol Psychiatry
2005;10(5):434–449.
37. Mei L, et al. Nat Rev Neurosci
2008;9(6):437–452.
38. Ishizuka K, et al. Biol Psychiatry
2006;59(12):1189–1197.
39. White T, et al. Am J Psychiatry
2006;163(3):376–380.
40. Arguello PA, et al. Neuron
2006;52(1):179–196.
41. Sanftner LM, et al. Exp Neurol
2005;194(2):476–483.
42. Wood JD, et al. Hum Mol Genet
2009;18(3):391–404.
43. Sawamura N, et al. Mol Psychiatry
2008;13(12):1138–1148, 1069.
44. Bockamp E, et al. Physiol
Genomics 2002;11(3):115–132.
45. Kos CH. Nutr Rev 2004;62
(6 Pt 1):243–246.
46. Niwa M, et al. Neuron 2010;
65(4):480–489.
47. Seshadri AJ, et al. Prog Brain Res
2009;179:17–27
48. LoTurco J, et al. Cereb Cortex
2009;19 Suppl 1:i120–125.
49. Matsuda T, et al. Proc Natl
Acad Sci U S A 2007;104(3):
1027–1032.

77
 Chapter 7: Validation of candidate genetic susceptibility factors
50. Cetin A, et al. Nat Protoc 2006;
1(6):3166–3173.
51. Koike H, et al. Proc Natl Acad Sci
U S A 2006;103(10):3693–3697.
52. Clapcote SJ, et al. Neuron 2007;
54(3):387–402.
53. Hikida T, et al. Proc Natl Acad Sci
U S A 2007;104(36):14501–14506.
54. Pletnikov MV, et al. Mol Psychiatry
2008;13(2):173–186, 115.
78
55. Jaaro-Peled H. Prog Brain Res
2009;179:75–86.
56. Harrison PJ. Brain 1999;122
(Pt 4):593–624.
57. Shenton ME, et al. Schizophr Res
2001;49(1–2):1–52.
58. Lewis DA, et al. Am J Psychiatry
2001;158(9):1411–1422.
59. Shen S, et al. J Neurosci 2008;
28(43):10893–10904.
60. Jaaro-Peled, et al. Schizophr Bull
2009;35:865–873.
61. Ayhan Y, et al. Mol Psychiatry
2011;16(3):293–306.
62. Meyer-Lindenberg A, et al.
Nat Rev Neurosci 2006;7(10):
818–827.
63. McGuire P, et al. Trends
Pharmacol Sci 2008;29(2):
91–98.
 Epigenetic mechanisms in drug
Chapter
8 addiction and depression
William Renthal and Eric J. Nestler
Introduction
Epigenetics is classically defined as the interaction
between genes and the environment that gives rise
to a specific phenotype. An example of this process
is observed in cellular differentiation where chemical
signals induce totipotent stem cells to differentiate
into genetically identical cell types with vastly differ-
ent functions. This is easily apparent when comparing
the brain, for example, which is extremely sensitive
to tissue damage and unable to regenerate, to other
organs (e.g. liver) that can regenerate rapidly. This
is due in part to the vastly different sets of genes
expressed between neurons and hepatocytes, despite
their identical DNA templates. Mechanistic insight
into this process has been gained over the past
20 years and involves the transduction of unique
environmental signals into precise and highly stable
alterations in chromatin structure that ultimately gate
access of transcriptional machinery to specific gene
programs, thereby providing unique gene expression
profiles in response to specific environmental cues
[1]. Importantly, many of these chromatin remodel-
ing mechanisms are highly stable, which contributes
to the maintenance of specific gene expression pro-
grams in the correct tissues throughout the life of an
individual.
The strong control epigenetic mechanisms exert
on gene expression and their potential stability over
time suggests a potential role in mediating aspects of
the long-lasting neural plasticity that ultimately result
in psychiatric syndromes such as drug addiction or
depression, as well as their reversal during effective
treatment. Thus, epigenetic research in psychiatry
attempts to identify whether environmental stimuli
(e.g. drugs of abuse, stress) induce certain changes
in chromatin structure that mediate an “addicted” or
Principles of Psychiatric Genetics, eds John I. Nurnberger, Jr. and Wade H. Berrettini. Published by Cambridge University
Press. © Cambridge University Press 2012.
“depressed” transcriptional program in neurons,
much in the same way environmental cues differen-
tiate a stem cell into specific lineages. While this field
is still in its infancy, great progress is being made in
identifying epigenetic alterations in drug addiction
and depression, as well as in several other neuropsy-
chiatric syndromes such as schizophrenia, Alzheimer’s
disease, and Rett syndrome. Focusing on drug addic-
tion and depression, this chapter will briefly review
the molecular machinery underlying epigenetic mech-
anisms and discuss how their dysregulation may con-
tribute to these chronic psychiatric disorders.
Epigenetic mechanisms
Chromatin is the complex of DNA, histones, and
associated nonhistone proteins in the cell nucleus.
DNA wraps around histone octamers made up of
two copies of histone H2A, H2B, H3, and H4 [2],
which then undergo a complex supercoiling process
to form a highly condensed structure (Figure 8.1) [3].
Initially, it was thought that this elaborate chromatin
structure only functioned to condense meters of DNA
into the microscopic cell nucleus, but it is now known
to participate in the transcriptional status of nearly
every eukaryotic gene. Because DNA is tightly associ-
ated with histones and often embedded deep within
chromatin supercoils that are structurally inaccessible
to transcriptional activators [4, 5], cellular mechan-
isms exist to modify and remodel chromatin structure
to allow for the coordinated expression of specific
transcriptional programs and the silencing of others
[6]. Such modifications typically occur on N-terminal
histone tails and include acetylation, phosphorylation,
and methylation of histones, methylation of DNA, and
many others, with each modification either directly
altering histone/DNA interactions or serving as a mark
79
 Chapter 8: Epigenetic mechanisms in drug addiction and depression

(a) Histone tail Figure 8.1 Chromatin remodeling. (a) Picture of a

nucleosome showing a DNA strand wrapped around
a histone octamer composed of two copies each of
the histones H2A, H2B, H3, and H4. The amino (N) termini
of the histones face outward from the nucleosome
DNA
complex. (b) Chromatin can be conceptualized as
H2B existing in two primary structural states: as active, or
H4
open, euchromatin (top left) in which histone acetylation
(A) is associated with opening the nucleosome to allow
H3
binding of the basal transcriptional complex and other
H2A activators of transcription; or as inactive, or condensed,
heterochromatin where all gene activity is permanently
silenced (bottom left). In reality, chromatin exists in a
A Acetylation continuum of several functional states (active; permissive
Histone [top right]; repressed [bottom right]; and inactive).
M Methylation Enrichment of histone modifications such as acetylation
(b) Active P Phosphorylation and methylation (M) at histone N-terminal tails and
Histones Histone Transcription factor
related binding of transcription factors and co-activators
tail +
Permissive
(Co-Act) or repressors (Rep) to chromatin modulates the
A A A M M transcriptional state of the nucleosome. Recent evidence
P A A suggests that inactivated chromatin may in some cases
be subject to reactivation in adult nerve cells, although
this remains uncertain. (c) Summary of common
covalent modifications of H3, which include acetylation,
A Co-Act A
methylation and phosphorylation (P) at several amino
A M A
M acid residues. H3 phosphoacetylation commonly
DNA Basal transcription
complex
involves phosphorylation of S10 and acetylation of K14.
Acetylation is catalysed by histone acetyltransferases
Inactive (HATs) and reversed by histone deacetylases (HDACs);
M M
P P lysine methylation (which can be either activating or
repressing) is catalysed by histone methyltransferases
Repressed
Rep Rep (HMTs) and reversed by histone demethylases (HDMs);
Rep
M M M
M and phosphorylation is catalysed by protein kinases (PK)
Rep M
A M M and reversed by protein phosphatases (PP), which have
Rep
A M not yet been identified with certainty. K, lysine residue;
M
M
? S, serine residue. (From [3].) See plate section for
color version.
Rep M
M M
M
M M

(c)
Demethylation Demethylation

HDM HMT HDM HMT

Methylation Methylation
(activating) (repressing)

M M
M M
H3 K4 K27 S28 M
K9 K23 K36
S10 K14 K18 K79
P
A A
P A A
Histone
tail
Acetylation Phosphorylation
(activating) (activating)

HAT HDAC PK PP

Deacetylation Dephosphorylation

80
 Chapter 8: Epigenetic mechanisms in drug addiction and depression
that recruits specific transcriptional co-regulators to
positively or negatively regulate the underlying gene’s
activity. Ultimately, dozens of potential modifications
that occur at many distinct histone residues summate
to determine the final transcriptional output of a
given gene [7]. An especially important aspect of
certain chromatin modifications is their apparent
stability, as is seen with maternal imprinting in
Prader–Willi syndrome or X-inactivation, where
DNA methylation contributes to life-long gene silen-
cing [3]. However, despite the apparent stability of
epigenetic mechanisms in vivo, all types of chromatin
modifications identified to date are potentially revers-
ible and have specific enzymes or processes which
mediate the addition or removal of each mark [6].
The reasons for this apparent biochemical and
physiological disconnect are currently not clear.
Histone acetylation
Acetylation of histone lysine residues reduces the
electrostatic interaction between histone proteins and
DNA, which is thought to relax chromatin structure
and make DNA more accessible to transcriptional
regulators (Figure 8.1) [6]. Genome-wide studies indi-
cate that the presence of high levels of histone acetyla-
tion in gene promoter regions is almost invariably
associated with transcriptional activation while low
levels of acetylation correlate with less gene activity
[8, 9]. Most genome-wide studies of histone acetyla-
tion have focused on acetylation of histones H3
and H4, but histone acetylation can occur on other
histone proteins as well.
Histone acetylation is a dynamic process, con-
trolled by specific enzymes which either add or
remove the acetyl mark. Histone acetyltransferases
(HATs) catalyze the addition of acetyl groups onto
lysine residues of histone proteins. There are over a
dozen known HATs that have varying degrees of
specificity for different lysine residues and histone
proteins and, importantly, many of them can also
acetylate nonhistone proteins such as transcription
factors (e.g. p53). Certain HATs, such as CREB-binding
protein (CBP) and p300/CBP-associated factor
(PCAF), have been implicated in behavioral responses
to drugs of abuse, stress, and learning and memory
[10–12]. Recently, several transcription factors
already implicated in behavioral responses to emo-
tional stimuli (e.g. ATF2 [activating transcription
factor 2], CLOCK) have also been shown to possess
intrinsic HAT activity, suggesting that part of their
function in regulating emotional behaviors could be
a result of histone acetylation [13, 14].
Histone deacetylases (HDACs), which remove
acetyl groups from histones, are divided into four
classes. Class I HDACs (e.g. HDAC1, 2, 3, and 8)
are ubiquitously expressed and likely mediate the
majority of deacetylase activity within cells. Recent
evidence points to HDAC2 in the hippocampus as a
key regulator of learning and memory [15]. Class II
HDACs (e.g. HDAC4, 5, 9) are only expressed in
specific tissues such as heart and brain and are much
larger enzymes that also contain an N-terminal regu-
latory domain that enables them to be shuttled in
and out of the nucleus in a neural activity-dependent
manner [16]. While Class II HDACs can deacetylate
histones, they are far less efficient enzymes than Class
I HDACs [17]. There is currently one Class IV
HDAC, HDAC11, and it has characteristics of both
Class I and Class II enzymes [18]. Class III HDACs
(Sirtuins) are mechanistically distinct from the other
HDACs and have been implicated in the regulation of
life span and metabolism [19]. While Classes I–III are
evolutionarily conserved in yeast, there are far more
members in each class in mammals, with each indi-
vidual HDAC performing a unique function. Perhaps
the best studied example of this is HDAC6, which is a
Class II HDAC that deacetylates tubulin and is not
even present in the nucleus. The individual functions
of other HDACs remain an active topic of investigation.
Histone phosphorylation
Histone phosphorylation is generally associated with
transcriptional activation; it can be observed on the
promoters of immediate early genes such as c-fos
when they are induced after cyclic adenosine mono-
phosphate (cAMP) stimulation or glutamate treat-
ment in cultured striatal neurons [20, 21]. One of
the best characterized histone phosphorylation sites
is serine 10 on histone H3 (H3S10). This modification
stabilizes the HAT, GCN5, on gene promoters while
antagonizing the repressive modification – methylation
of lysine 9 on histone H3 (H3K9) and its subsequent
recruitment of HP1 (heterochromatin protein 1) (see
below) [6]. Since phosphorylation at H3S10 recruits a
HAT, the neighboring lysine residue at H3K9 is often
acetylated in concert with phosphorylation, a process
called phospho-acetylation that further potentiates
gene activation.
81
 Chapter 8: Epigenetic mechanisms in drug addiction and depression
There are several nuclear protein kinases and
protein phosphatases known to regulate histone
phosphorylation [6]. The mitogen-activated protein
kinase, MSK1, and the dopamine and cAMP regu-
lated protein phosphatase inhibitor, DARRP-32, are
elegant examples shown to regulate H3S10 phosphor-
ylation in the adult brain in response to cocaine expos-
ure [22, 23]. Furthermore, genetic disruption of the
histone modifying ability of MSK1 or DARRP-32
in vivo has dramatic effects on behavioral responses
to cocaine. Thus, histone phosphorylation likely plays
an important role in the regulation of brain function.
Histone methylation
Histone methylation is thought to function primarily
by generating docking sites that recruit transcrip-
tional co-regulators to specific gene loci. Histone
methylation occurs on lysine residues and can exist
in mono-, di-, or tri-methylated states, enabling each
state to recruit unique co-regulators and exert distinct
effects on transcriptional activity [6]. Additionally,
methylation of different histone lysine residues can
result in distinct or even opposite effects on transcrip-
tion. For example, tri-methylation of H3K4 is highly
associated with gene activation, while tri-methylation
of H3K9 or H3K27 are usually associated with gene
repression [5]. The repression caused by tri-methyla-
tion of H3K9 is mediated in part via the recruitment
of co-repressors, such as HP1, as stated earlier. How-
ever, even this is an oversimplification, as methylated
H3K9 is often found in the coding region down-
stream of a gene promoter and may be involved in
transcriptional elongation [6, 24]. Thus, through a
vast array of combinatorial options, histone methyla-
tion provides each cell with exquisite control over an
individual gene’s activity.
Like histone acetylation and phosphorylation, the
enzymes which regulate histone methylation can
be divided into two main groups: those which add
the mark, histone methyltransferases (HMTs), and
enzymes which remove it, histone demethylases
(HDMs) (Figure 8.1). HMTs and HDMs not only
discriminate between specific histone lysine residues,
but each enzyme is also unique in its ability to cata-
lyze mono-, di-, or tri-methylation or demethylation
at that site [6]. For example, the HMT, KMT1C
(G9a), is specific for histone H3K9 but only adds
1 or 2 methyl groups, with the distinct HMT, KMT1A
(SUV39H1), catalyzing trimethylation of this site.
82
Similarly, the HDM, KDM3A (JHDM2a), can
demethylate one or two methyl groups on H3K9,
requiring a distinct demethylase (e.g. KDM4D
[JMJD2D]) to fully demethylate the trimethylated state.
Thus, many enzymes, which are often found in large
complexes with other transcriptional co-regulators, are
required to move between the unmethylated and fully
trimethylated states. Proper balance of histone methyl-
ation has already been strongly implicated in normal
brain function, as the HDM, KMT5C (SMCX), con-
trols dendritic spine density and is mutated in patients
with intellectual disability [25, 26].
DNA methylation
DNA methylation refers to the enzymatic methylation
of cytosine bases; this is, a fundamental cellular pro-
cess required for development, tissue-specific gene
expression, X-inactivation, and genetic imprinting,
to name a few examples [27]. DNA methylation is
thought to repress gene expression by interfering with
transcription factors binding to their target sequences
or by initiating the recruitment of co-repressors.
For example, the cAMP-response element (CRE) con-
tains a cytosine-guanine dinucleotide in the middle
of its consensus sequence, which, when methylated,
prevents the transcription factor CRE-binding protein
(CREB) from binding [28]. Thus, for genes at which
CREB is necessary to initiate transcription, methyla-
tion at this site is repressive. Methylated DNA can
also recruit methyl-binding domain-containing pro-
teins, such as MeCP2, which can then recruit and
stabilize transcriptional co-repressors such as HDACs
on specific gene promoters. Mutations in MeCP2
cause the autistic spectrum disorder, Rett syndrome,
illustrating the importance of DNA methylation in
normal brain development [29]. While there is a
strong correlation between methylated DNA and
repressed gene activity, recent studies of MeCP2 indi-
cate it may also serve to activate gene activity under
some circumstances [30], suggesting that the context
in which DNA methylation occurs is an important
factor in its ultimate effect on transcription.
There are three known enzymes which catalyze
DNA cytosine methylation: DNMT1, DNMT3a, and
DNMT3b. DMNT2 was recently shown to methylate
RNA rather than DNA [31]. Together, these enzymes
establish and maintain the unique methylation
patterns that exist within each cell type. While the
regulation of these enzymes in brain remains unclear,
 Chapter 8: Epigenetic mechanisms in drug addiction and depression
pharmacological inhibition of DNA methylation in
the brain in vivo results in rapid demethylation of
specific gene targets and severe deficits in learning
and memory [32]. The mechanism by which this
occurs, however, remains unclear because, unlike
other chromatin modifications, the existence of DNA
demethylases remains highly controversial [33]. Never-
theless, regulation of DNA methylation by environ-
mental stimuli remains an attractive mediator of
long-lasting changes in transcription in adult neurons.
Epigenetic mechanisms in drug
addiction
Drug addiction is a chronic relapsing disorder
whereby behavior related to seeking and taking drugs
of abuse becomes compulsive and pathological [34].
The process by which repeated drug experimentation
transitions into a chronically addicted state is the
focus of intense research, as clues into these mechan-
isms may help better manage or perhaps fully treat
addicted patients. Currently, treatment of drug addic-
tion is a major clinical challenge because of high rates
of relapse, which are observed even after long periods
of drug abstinence. Thus, another major research goal
is to identify the mechanisms underlying drug relapse
and target them with novel therapeutics that improve
treatment outcomes.
One such mechanism involves long-lasting alter-
ations in gene expression in brain reward regions that
contribute to both the pathogenesis and persistence of
drug addiction. For example, the transcription factor,
ΔFosB, a splice product of the immediate early gene
fosB, accumulates several-fold uniquely in brain
reward regions after repeated drug exposure [35].
This establishes a positive-feedback loop whereby
higher levels of ΔFosB facilitate greater drug-seeking
behavior, which in turn induce higher levels of
ΔFosB. Thus, ΔFosB is one molecule that may con-
tribute to the transition to an addicted state. Activator
of G-protein signaling 3 (AGS3) is an example of a
protein that may contribute to drug relapse, since it
remains elevated after weeks of drug withdrawal and,
when genetically manipulated in brain, alters relapse
behavior [36]. ΔFosB and AGS3 not only exemplify
how drug-induced gene expression changes in brain
can contribute to clinically important aspects of the
addicted state, they also illustrate the potential for
new therapeutic targets that could prevent or reverse
the stable effects of chronic drug exposure. Over the
past decade, as the knowledge of molecular/epigenetic
mechanisms by which individual genes are activated
or silenced in cell culture has exploded, the question
of how drugs of abuse can mediate long-lasting
changes in the activity of specific genes is now a major
focus in addiction research. While these epigenetic
studies are ongoing for a variety of substance abuse
models, we focus here on the psychostimulants
cocaine and amphetamine because these studies are
more mature.
Both cocaine and amphetamine are frequently
studied in rodents using forced investigator adminis-
tration of the drug. The molecular changes in brain
can then be studied after a single injection or repeated
injections to provide insight into changes that occur
initially in response to the drug versus changes which
require chronic exposure. While this model of forced
drug administration conveniently analyzes the acute
and chronic pharmacological effects of drugs, it only
permits an elementary assessment of addictive-like
behaviors. Thus, rodent self-administration para-
digms, in which animals are trained to push a lever
to obtain drug volitionally, while much lower in
throughput, better model human addiction. Both
forced administration and self-administration models
have been used to study epigenetic changes in
response to psychostimulants.
The first studies to identify changes in chromatin
structure in response to drugs of abuse focused on
histone acetylation on the promoters of genes previ-
ously implicated in cocaine action (e.g. c-fos, fosB,
cdk5) [37]. These studies found that acute adminis-
tration of cocaine significantly increased histone H4
acetylation on the immediate early genes c-fos and
fosB in striatum with a rapid time course consistent
with the induction time of their mRNAs (Figure 8.2).
The histone acetyltransferase CBP appears to be
required for the drug-induced acetylation of the fosB
promoter [10]. Interestingly, an acute cocaine dose
increases total levels of histone H4 acetylation, and
histone H3 phospho-acetylation in striatum, while
cocaine does not induce histone acetylation at several
control gene promoters [22, 37]. These global increases
in histone acetylation, which are also observed in
response to environmental enrichment and tests of
learning and memory [38, 39], may be accounted for
by high levels of acetylation on specific subsets of genes.
The promoters of certain genes induced by chronic
cocaine exposure are hyperacetylated for days to weeks
after the last drug exposure (Figure 8.2). For example,
83
 Chapter 8: Epigenetic mechanisms in drug addiction and depression

Figure 8.2 Regulation of chromatin structure by drugs of abuse. Drug-induced signaling events are depicted for cocaine and amphetamine.
Cocaine and amphetamine can increase cyclic adenosine monophosphate (cAMP) levels in striatum, which activates protein kinase A
(PKA) and leads to phosphorylation of its targets. This includes the cAMP response element binding protein (CREB), the phosphorylation of
which induces its association with the histone acetyltransferase, CREB binding protein (CBP) to acetylate histones and facilitate gene activation.
This is known to occur on many genes including fosB and c-fos in response to psychostimulant exposure. ΔFosB is also upregulated by chronic
psychostimulant treatments, and is known to activate certain genes (e.g. cdk5) and represses others (e.g. c-fos) where it recruits HDAC1.
This repression of c-fos also involves increased repressive histone methylation, which is thought to occur via the induction of specific histone
methyltransferases. It is not yet known how cocaine regulates histone demethylases (HDM) or DNA methyltransferases (DNMTs). Cocaine
also activates the mitogen activated protein kinase (MAPK) cascade, which through MSK1 can phosphorylate CREB and histone H3 at serine 10.
Cocaine promotes H3 phosphorylation via a distinct pathway, whereby PKA activates protein phosphatase 2A, leading to the
dephosphorylation of serine 97 of DARPP32. This causes DARPP32 to accumulate in the nucleus and inhibit protein phosphatase 1 (PP1) which
normally dephosphorylates H3. Chronic exposure to psychostimulants is also known to increase glutamatergic signaling from the prefrontal
cortex to the NAc. Glutamatergic signaling elevates Ca2+ levels in NAc synapses and activates CaMK (calcium/calmodulin protein kinases)
signaling, which, in addition to phosphorylating CREB, also phosphorylates HDAC5. This results in nuclear export of HDAC5 and increased
histone acetylation on its target genes (e.g. NK1R [NK1 or substance P receptor). (From [3].) See plate section for color version.

bdnf (brain-derived neurotrophic factor) and npy
(neuropeptide Y) expression were found to be upre-
gulated after cocaine self-administration and their
gene promoters were hyperacetylated, while egr-1
(early growth response 1) was found to be downregu-
lated and hypoacetylated after cocaine withdrawal
[40]. These findings suggest a role of histone acetyl-
ation in the maintenance of gene expression involved
in drug withdrawal and relapse.
Cocaine-induced alterations in chromatin struc-
ture in the nucleus accumbens (NAc), the ventral
portion of striatum heavily implicated as a brain
reward region, have been shown to regulate behav-
ioral responses to drugs of abuse. Pharmacological
inhibition of HDACs in the NAc, which increases
histone acetylation in this brain region, significantly
potentiates the locomotor-activating and rewarding
responses to cocaine and amphetamine [37, 41, 42].
Conversely, reducing histone acetylation by overex-
pressing certain HDACs, or knockdown of the HAT,
CBP, results in less sensitivity to cocaine [10, 36, 41].
Cocaine alters histone acetylation through many
enzymes in the NAc, but one particular HDAC,
HDAC5, responds uniquely to chronic cocaine
administration, raising the interesting possibility that
this HDAC is involved in the behavioral transitions
which occur between acute and chronic cocaine
exposure (e.g. drug experimentation to compulsive
drug use). Chronic cocaine administration increases
the phosphorylation of HDAC5 and shuttles it out of
the nucleus, permitting hyperacetylation of HDAC5
target genes (Figure 8.2) [41]. This phosphorylation

84
 Chapter 8: Epigenetic mechanisms in drug addiction and depression
reaction may be mediated via Ca2+/calmodulin-
dependent protein kinase II (CaMKII). Consistent
with its regulation by cocaine, mice deficient for
HDAC5 display normal rewarding responses to initial
cocaine exposure, but become hypersensitive when
treated with a chronic course of cocaine [41]. Thus,
pharmacological and genetic manipulations that
increase histone acetylation appear to potentiate behav-
ioral responses to cocaine and suggest that altered
histone acetylation may contribute to establishment
of an addicted state.
Histone H3 phosphorylation and phospho-acety-
lation also appear to play key roles in drug-regulated
behaviors. Global levels of histone H3 phosphoryl-
ation at serine 10 are induced by acute cocaine in
striatum, a process which requires MSK1 [22, 37].
The function of MSK1 is behaviorally important, as
mice lacking this kinase have attenuated locomotor
responses to cocaine. The genes at which histone
phosphorylation is occurring in response to cocaine
remain mostly undefined, with the exception of c-fos,
where histone phosphorylation occurs in conjunction
with acetylation (phospho-acetylation).
Preliminary findings suggest that histone methyl-
ation is also regulated by cocaine and, in turn, alters
behavioral responses to the drug. For example, inhib-
ition of a particular H3K9 histone methyltransferase,
KMT1C (G9a), whose expression is regulated in the
NAc by chronic cocaine administration, potentiates
behavioral responses to the drug [43]. These findings
are consistent with histone acetylation findings,
since inhibition of H3K9 methylation would also be
expected to enhance gene activity. Together, these data
suggest that, in general, increases in gene expression
potentiate behavioral sensitivity to drugs of abuse.
Overall, these findings implicate changes in his-
tone acetylation, phosphorylation, and methylation in
mediating expression changes in specific sets of genes
that are crucial for controlling behavioral responses
to drugs of abuse.
Epigenetic mechanisms in depression
Depression is a chronic disorder characterized by
many debilitating symptoms including dysphoria,
anhedonia, sleep disturbances, and weight changes.
Most people diagnosed with depression are prescribed
some type of antidepressant treatment, of which
selective serotonin reuptake inhibitors (SSRIs) or
mixed serotonin–norepinephrine reuptake inhibitors
(SNRIs) are the most common. Unfortunately, less
than 50% of patients exhibit a complete response to
SSRIs, SNRIs, and related antidepressants, thus leav-
ing a substantial portion of depressed patients with a
treatment-resistant depression, which may become a
chronic condition [see 44]. Psychiatric research is
thus focused on identifying new mechanisms that
are involved in the pathogenesis and maintenance of
depression, which may serve as new targets for more
effective therapeutics.
One of the most challenging obstacles for depres-
sion research has been the development of an animal
model that accurately recapitulates the symptoms of
human depression. While no model can effectively
model all aspects of human depression (e.g. suicide),
some of the major symptoms such as anhedonia and
sleep and weight disturbances, and their reversal by
antidepressant treatment, can be studied in rodents.
The pathogenesis of depressed-like states is typically
modeled in rodents by chronic exposure to stress [44].
One such model, chronic social defeat stress, involves
the repeated exposure of an experimental mouse to a
series of aggressive mice over 10 days. Each day the
stress begins as a brief physical encounter (typically
5–10 minutes) followed by a full day of sensory con-
tact (e.g. smell, sight) as the mice are separated by a
screen. After 10 days of social defeat, the experimental
mice develop a chronic syndrome (lasting more than
a month) that is characterized by anhedonia, anxiety-
like symptoms, weight loss, and loss of interest in
social interaction. Importantly, SSRIs or SNRIs
reverse most of these behavioral end points, making
chronic social defeat stress an attractive model in
which to study the molecular adaptations associated
with a depressed-like state and those involved with
antidepressant action [45, 46].
BDNF plays a critical role in the development of
the social defeat phenotype and its reversal by anti-
depressant treatment. It was observed that BDNF in
the hippocampus is downregulated for at least one
month after chronic social defeat stress, and that
chronic antidepressant treatment reversed this down-
regulation [46]. A mechanism for this long-lasting
regulation of gene expression was identified as meth-
ylation of H3K27, a repressive histone modification,
that remains hypermethylated on the bdnf promoter
within hippocampus for at least a month after defeat
stress (Figure 8.3 [3]). While chronic antidepressant
treatment of mice exposed to chronic social defeat
ameliorates many of the behavioral deficits and
85
 Chapter 8: Epigenetic mechanisms in drug addiction and depression

(a) Figure 8.3 Regulation of the bdnf gene

HDAC5 HDAC5 by social defeat. (a) In the absence of
stress, the chromatin state of brain-
A A A derived neurotrophic factor (Bdnf) is at
a basal level, characterized by moderate
+ levels of histone H3 acetylation and
Bdnf expression
virtually no H3K27 dimethylation. In this
state, histone deacetylase 5 (HDAC5)
A A might repress unnecessary activation of
M
M BDNF and maintain a chromatin balance.
(b) Chronic defeat stress induces the
specific and prolonged dimethylation
(b) HDAC5 of histone H3K27. This induces a more
HDAC5
M M M M M “closed” chromatin state at bdnf promoters
A M M A M M A M P3 and P4, and a corresponding repression
of Bdnf transcripts III and IV expression.
H3 acetylation and HDAC5 regulation are
–
Bdnf expression not affected after chronic defeat stress
alone, corroborating the idea that the
main repressive marker after chronic stress
A A
M
is histone methylation. (c) Chronic
M M M M
M M M M M imipramine (antidepressant) treatment
after defeat stress downregulates Hdac5
(c) expression and increases H3 acetylation,
M M M M
M M M M with little if any change in H3K27
A A A A A dimethylation. Imipramine-dependent H3
hyperacetylation at the bdnf promoters
+ P3 and P4 allows partial “reopening” of
Bdnf expression
the repressed chromatin state caused by
defeat stress, and results in transcriptional
M
A A A A A M H3–K27 dimethylation reactivation of the bdnf gene. K, lysine
M M M M residue. (From [3].) See plate section
M M M M A H3 acetylation for color version.

restores Bdnf mRNA to normal levels, H3K27 remains
hypermethylated. The maintenance of H3K27 methyl-
ation even after chronic antidepressant treatment sug-
gests that BDNF expression might revert to a repressed
state if drug administration were stopped. This novel
epigenetic mechanism, which was proposed as a form
of “molecular scar”, may describe a potential mech-
anism by which the symptoms of depressed patients
reappear after cessation of antidepressant treatment;
however, this remains speculative and further research
is needed.
The recovery of Bdnf expression after antidepres-
sant treatment is likely mediated by the antidepressant-
induced increase in histone H3K4 methylation and
H3 poly-acetylation in hippocampus, which are associ-
ated with gene activation [46]. Interestingly, tranylcy-
promine, which inhibits monoamine oxidases and is
used as an antidepressant, is actually a much stronger
inhibitor of the histone H3K4 demethylase KMT1A
(formerly, LSD1) than it is of either monamine oxidase
A or B [47]. Thus, it would be interesting to determine
whether any of the antidepressant properties of
tranylcypromine derive from its blockade of KMT1A
and the subsequent facilitation of H3K4 methylation.
Arguing against this interpretation is the knowledge
that several structurally unrelated monoamine oxidase
inhibitors, which have not been shown to inhibit his-
tone demethylases, are still effective antidepressants.
The increase in H3 acetylation by antidepressant
treatment suggested that HDAC inhibitors may also
have antidepressant-like effects. Indeed, in both the
chronic social defeat model and in the forced swim
test, HDAC inhibitors demonstrated antidepressant-
like prosperities [46, 48]. This was especially apparent
when an HDAC inhibitor was administered in addi-
tion to the SSRI fluoxetine. While these inhibitors target
numerous HDACs, one specific isoform, HDAC5,
stood out because it was oppositely regulated by stress
and antidepressant treatment [46]. Indeed, overexpres-
sion of HDAC5 in the hippocampus blocks the behav-
ioral effects of chronic antidepressant treatment,
suggesting that increased histone acetylation on the
bdnf promoter is a key mechanism to overcome the
repressive effects of H3K27 methylation.

86
 Chapter 8: Epigenetic mechanisms in drug addiction and depression
Another intriguing aspect of chronic social defeat
stress is that the severity of the depressed-like pheno-
type varies within a cohort of inbred (i.e. nearly
genetically identical) mice. It was observed that more
susceptible mice in order to defeat stress show signifi-
cantly higher firing rates of dopaminergic neurons in
the ventral tegmental area (VTA). However, less sus-
ceptible mice had normal VTA firing rates because of
an upregulation of potassium channels in this brain
region [44]. Why do certain mice upregulate protect-
ive potassium channels in the VTA while others
fail to do this and become “depressed?” Perhaps
an epigenetic mechanism is involved in altering the
promoters of certain potassium channels to ultimately
determine if the gene will be induced in response to
chronic stress. If so, which life experiences trigger
these chromatin remodeling events? These are
important questions that may shed new light into an
extraordinarily complex syndrome and provide new
avenues for the development of more effective
antidepressants.
Another important epigenetic mechanism that
may contribute to long-lasting changes in neural
function and behavior is DNA methylation. Early
insight into the role of DNA methylation in behavior
followed from studies of maternal care. Rats that
receive poor maternal care as pups grow up to deliver
poor maternal care to their pups (as defined by licking
and grooming). In addition to delivering poor mater-
nal care, these rats also develop long-lasting heightened
anxiety and stress responses. Meaney and colleagues
identified a region of the glucocorticoid receptor (GR)
gene, which was hypermethylated throughout adult-
hood in rats that received poor maternal care. Treat-
ment with an HDAC inhibitor not only reduced DNA
methylation on the GR receptor gene but also improved
anxiety and stress responses in these rats [49]. While
these data are correlative, they suggest an important
role for epigenetic mechanisms in anxiety and stress
and suggest that DNA methylation at the GR gene
promoter (and probably other genes) may contribute
to this phenomenon.
Taken together, these studies demonstrate that
chromatin structure is an important substrate for
long-lasting changes in behavioral responses to stress
and antidepressant action. While the precise signaling
mechanisms by which environmental stresses con-
verge on chromatin remain unclear (e.g. Figure 8.2),
these early studies suggest the exciting possibility
that pharmacological manipulation of chromatin
remodeling pathways could be a novel approach to
new antidepressant development.
Challenges of epigenetics research
in psychiatry
The study of gene expression and chromatin remodel-
ing in adult brain is made difficult by two major
technical challenges. The first challenge is that brain
tissue is highly heterogeneous and contains many
subtypes of neurons and glia, each of which is further
distinguished based on connectivity and function.
Thus, the cellular and molecular responses to cocaine
can vary tremendously even within a small subregion
of the NAc. In response to cocaine, for example,
neurons which express Gs-coupled dopamine D1
receptors in the NAc show higher levels of cAMP
and also show its downstream consequences (such
as phosphorylation of CREB), while neurons express-
ing Gi-coupled D2 receptors respond oppositely.
There is even a small population of striatal neurons
which express both D1 and D2 receptors, which seem
to form hetero-oligomers that couple to Gq signaling
[50], leading to more complex cellular responses to
cocaine. The end result of studying tissue lysates from
such a heterogeneous structure is that most of the
observed effects are very small since they are averaged
together with several other, differently responding,
cell types. This is particularly problematic for micro-
array analyses, which depend on strong effects for
statistical significance. One solution that is being cur-
rently developed is fluorescence-activated cell sorting
(FACS), which may be used in conjunction with
BAC-transgenic mice that express green fluorescent
protein (GFP) in specific cell types (e.g. dopamine D1
versus D2 neurons) [51]. This new technology will
permit the study of the mechanism by which cocaine
alters gene expression and histone modifications in
isolated cell populations.
The second major challenge is distinguishing
causality from correlation. How does one determine
whether histone acetylation at the bdnf promoter
causes a transcriptional or behavioral response
in vivo? To address this question, one must induce
or prevent histone acetylation on the bdnf gene spe-
cifically. Simply overexpressing a HAT or inhibiting
HDACs would likely hyperacetylate numerous other
target genes in addition to bndf, thus confounding the
transcriptional and behavioral interpretation of the
experiment. Recently, an exciting breakthrough using
87
 Chapter 8: Epigenetic mechanisms in drug addiction and depression

zinc finger proteins may enable investigators to dir-
ectly address this very challenging question. Zinc
finger peptides can be designed or screened for highly
sequence-specific DNA binding properties. These
zinc finger peptides can then be fused to a chromatin
remodeling enzyme, which effectively targets that
enzyme to the promoter of a specific gene. This was
accomplished for the first time using a DNA methyl-
transferase in cell culture [52, 53], and may now
permit behavioral neuroscientists, using viral-medi-
ated gene transfer or genetic mutant mice, to ask
whether histone modifications at specific genes are
indeed causally linked to the transcriptional and
behavioral phenotypes observed.
Concluding remarks
There is now growing evidence that epigenetic mech-
anisms, such as histone acetylation and methylation,
are involved in regulating the salience of environmen-
tal stimuli [3, 10, 37–39, 41, 54]. In most studies to
date, including learning and memory, chronic pain,
addiction, and depression models, manipulations that
elevate histone acetylation potentiate the behavioral
effects of the associated environmental stimulus. This
has important implications for the pathogenesis of
drug addiction and depression, since novel therapeut-
ics could target such mechanisms to ultimately block
or even reverse a chronically addicted or depressed
state. As illustrated by the persistent hypermethyla-
tion in mice exposed to chronic social defeat stress
(H3K27 methylation at the bdnf gene) and rats sub-
jected to poor maternal care (DNA methylation of the
GR gene), epigenetic mechanisms are also attractive
candidates for novel molecular substrates that medi-
ate long-lived changes in brain. Ultimately, the key
function of chromatin remodeling is to alter the
transcription or the transcriptional potential of genes
which eventually affect neural function [37]. Thus,
any study of chromatin regulation is inexorably
linked with the study of the underlying gene activity.
This is an especially important point for the purposes
of treating chronic psychiatric diseases, because a
drug that can permanently reverse the underlying
transcriptional defect may circumvent the side effects
and compliance issues that plague chronic therapies
based on antagonizing the dysregulated gene targets.
While extremely exciting, epigenetic research in
psychiatry is still in its infancy, and far more research
is needed to identify both the dysregulated genes and
chromatin modifications responsible for individual
psychiatric diseases and their improvement during
effective therapy. Work is needed to determine
whether meaningful epigenetic modifications can be
detected in human peripheral tissues (blood, epithelial
cells) that are useful in establishing a diagnosis or
tracking treatment. Consideration of epigenetic
mechanisms should inform and eventually facilitate
the identification of genetic factors that contribute to
psychiatric syndromes. Future advances might
include the development of novel positron emission
tomography (PET) or other brain imaging ligands
that would make it possible to assess global levels of
chromatin modifications in a patient’s brain. Finally,
it will be important to determine whether small mol-
ecules that target chromatin modifying mechanisms
might be useful in the treatment of mental illness.
Acknowledgements
Parts of this chapter were based on Tsankova et al. [3],
no permission needed, and Renthal et al. [55], with
permission.

References 7. Strahl BD, et al. Nature 2000;

403(6765):41–45.
12. Oliveira AM, et al. Learn Mem
2007;14(9):564–572.
1. Surani MA, et al. Cell 2007;
128(4):747–762. 8. Kurdistani SK, et al. Cell 2004; 13. Doi M, et al. Cell 2006;125(3):
117(6):721–733. 497–508.
2. Luger K, et al. Curr Opin Genet Dev
1998;8(2):140–146. 9. Pokholok DK, et al. Cell 2005; 14. Kawasaki H, et al. Nature 2000;
122(4):517–527. 405(6783):195–200.
3. Tsankova N, et al. Nat Rev Neurosci
2007;8(5):355–367. 10. Levine AA, et al. Proc 15. Guan JS, et al. Nature 2009;
4. Felsenfeld G, et al. Nature 2003; Natl Acad Sci U S A 2005; 459(7243):55–60.
421(6921):448–453. 102(52):19186–19191. 16. Chawla S, et al. J Neurochem 2003;
5. Li B, et al. Cell 2007;128(4):707–719. 11. Maurice T, et al. 85(1):151–159.
6. Kouzarides T. Cell 2007;128(4): Neuropsychopharmacology 17. Fischl W, et al. Mol Cell 2002;
693–705. 2007;33(7):1584–1602. 9(1):45–57.

88

18. Yang XJ, et al. Nat Rev Mol Cell
Biol 2008;9(3):206–218.
19. Haigis MC, et al. Genes Dev
2006;20(21):2913–2921.
20. Brami-Cherrier K, et al. J
Neurochem 2007;101(3):
697–708.
21. Li J, et al. J Neurochem 2004;
90(5):1117–1131.
22. Brami-Cherrier K, et al. J Neurosci
2005;25(49):11444–11454.
23. Stipanovich A, et al. Nature
2008;453(7197):879–884.
24. Vakoc CR, et al. Mol Cell 2005;
19(3):381–391.
25. Iwase S, et al. Cell 2007;128(6):
1077–1088.
26. Jensen LR, et al. Am J Hum Genet
2005;76(2):227–236.
27. Suzuki MM, et al. Nat Rev Genet
2008;9(6):465–476.
28. Zhang X, et al. Proc Natl Acad Sci
U S A 2005;102(12):4459–4464.
29. Ramocki MB, et al. Nature
2008;455(7215):912–918.
30. Chahrour M, et al. Science
2008;320(5880):1224–1229.
Chapter 8: Epigenetic mechanisms in drug addiction and depression
31. Goll MG, et al. Science 2006;
311(5759):395–398.
32. Miller CA, et al. Neuron 2007;
53(6):857–869.
33. Ooi SK, et al. Cell 2008;133(7):
1145–1148.
34. Hyman S, et al. Annu Rev Neurosci
2006;29:565–598.
35. Nestler EJ. Philos Trans R Soc
Lond B Biol Sci 2008;363
(1507):3245–3255.
36. Bowers MS, et al. Neuron 2004;
42(2):269–281.
37. Kumar A, et al. Neuron 2005;
48(2):303–314.
38. Fischer A, et al. Nature 2007;
447(7141):178–182.
39. Levenson JM, et al. J Biol Chem
2004;279(39):40545–40559.
40. Freeman WM, et al.
Neuropsychopharmacology
2008;33(8):1807–1817.
41. Renthal W, et al. Neuron 2007;
56(3):517–529.
42. Schroeder FA, et al.
Neuropsychopharmacology
2008;33(12):2981–2992.
43. Maze I, et al. Science 2011;
327:213–216.
44. Krishnan V, et al. Nature
2008;455(7215):894–902.
45. Berton O, et al. Science 2006;
311(5762):864–868.
46. Tsankova N, et al. Nat Neurosci
2006;9(4):519–525.
47. Lee MG, et al. Chem Biol 2006;
13(6):563–567.
48. Schroeder FA, et al. Biol
Psychiatry 2007;62(1):55–64.
49. Weaver IC, et al. Nat Neurosci
2004;7(8):847–854.
50. Rashid AJ, et al. Proc Natl Acad Sci
U S A 2007;104(2):654–659.
51. Lobo MK, et al. Nat Neurosci
2006;9(3):443–452.
52. Li F, et al. Nucleic Acids Res
2007;35(1):100–112.
53. Smith AE, et al. J Biol Chem
2008;283(15):9878–9885.
54. Zhang C, et al. Cell 2002;110(4):
479–488.
55. Renthal W, et al. Trends Mol Med
2008;14:341–350.
89
 Panic disorder
Chapter
9 Ardesheer Talati and Myrna M. Weissman
Panic disorder (PD) is a complex genetic anxiety
disorder. Although data from family and twin studies
suggest familial heritability, genetic studies have been
to date only partially successful in identifying putative
genetic regions underlying the disorder. This chapter
provides a broad overview of the clinical and epidemi-
ological profile of PD, and a review of genetic studies,
including promising areas of research and ways of
subtyping the disorder for greater possible genetic
tractability. We conclude with a review of neuroima-
ging studies of PD to show how neurobiological data
may be coupled with genetics to more comprehen-
sively probe the underlying mechanisms.
Diagnosis and epidemiology
PD is a complex anxiety disorder characterized by
recurrent attacks of sudden and uncontrollable fear.
The attacks typically come out of the blue, peak
rapidly, and are accompanied by a wide range of
autonomic, cardiovascular and gastrointestinal symp-
toms, as illustrated in Figure 9.1. DSM-IV criteria [1]
require that the attacks be both unexpected and recur-
rent, and that they be followed by at least one month
of either the fear of having additional attacks, or the
development of phobic avoidance or other significant
changes in behavior. The attacks must also not be
explainable by neurological or general medical condi-
tions, medication, or substance use.
The lifetime prevalence of PD is approximately
1–5% [2–4], although isolated panic attacks occur
more frequently [5]. The epidemiology is similar across
cultures, with first onset typically in early adulthood
(though it can occur in children), and higher rates
among women [4, 6]. Because the symptoms of a panic
attack can mimic serious medical conditions (such as a
heart attack), the disorder is associated with high use of
Principles of Psychiatric Genetics, eds John I. Nurnberger, Jr. and Wade H. Berrettini. Published by Cambridge University
Press. © Cambridge University Press 2012.
90
medical services – notably emergency rooms [7–9] –
and is therefore costly not only to the individual
patient but to society at large. PD is frequently comor-
bid with depression, other anxiety disorders, and sub-
stance use and abuse [6, 10, 11]. Agoraphobia (AG)
(the fear of having attacks in public or unfamiliar
settings, in extreme cases resulting in complete con-
finement to the home) is also very common [4]. It
should be noted however that even though AG most
frequently presents as a complication to the primary
symptoms, not all agoraphobics meet criteria for PD
[5]; furthermore, of those that do, a third experience
AG symptoms prior to panic onset [12], and of those
that do not, some may fail to express overt panic
symptoms due to successful avoidance rather than a
true lack of vulnerability to the disorder itself. The
epidemiology of PD and AG relative to each other is
therefore not clear cut, and consensus on whether AG
should be classified independently of, or as part of, PD,
has been the subject of an enduring debate.
There is a confluence of evidence from family,
twin, and molecular genetic studies, suggesting a gen-
etic contribution to PD. If genes influence the suscep-
tibility to PD, biological relatives of individuals with
the disorder should have a higher prevalence of PD
than family members of controls without the dis-
order. Twin studies further examine the relative con-
tribution of genetic and environmental influences on
the variance in PD risk. Finally, molecular genetic
studies seek to identify specific genes that increase
risk of developing the disorder. These studies are
reviewed respectively in the following sections.
Family and twin studies
A number of family studies have reported familial
co-aggregation of the disorder, with first-degree
relatives (FDRs) of persons with PD, as compared to
 Chapter 9: Panic disorder

Figure 9.1 Panic attacks and panic

Increased disorder. The panic attack is the central
Sweating feature of the disorder, but is typically
heart rate
accompanied by a range of
cardiovascular, gastrointestinal, and
Derealization or Chest pain neurological symptoms. Not all panic
depersonalization attacks lead to, or are associated with
the full panic disorder syndrome, which
require repeating panic attacks with
Fear of going crazy
Panic Shortness of breath significant and lasting impairment.
attack
or dying
Requires ≥ 4 of the
surrounding symptoms
Stomach/abdominal Dizziness
distress

Numbing or tingling
Hot /cold flashes

Attacks are recurrent and unexpected

And accompanied by:

(at least one or more, for one or more month)

Worry about Concern about Modification in

consequences additional attacks behavior

And not better accounted for by:

Other anxiety General medical

Substance use
disorder condition

Panic disorder

those of persons without, at several-fold greater risk
for PD themselves [13–17] (A recent weighted meta-
analysis reported an adjusted mean seven-fold
increase [18].) Despite the frequent comorbidity with
depression and other anxiety disorders such as gener-
alized anxiety disorder (GAD), at least some of the
transmission appears specific to PD [14, 19]. Though
familial transmission does not appear to particularly
vary by gender or parent-of-origin [20, 21], earlier
onset has been associated with greater familial risk.
Goldstein et al. for example reported that whereas
FDRs of PD probands with onset after age 20 were
at 6-fold higher risk than those of controls, those of
probands with onset under age 20 were at 17-fold risk

91
 Chapter 9: Panic disorder
100
Percent of variance due to heritable factors
50
0 The authors identified a genetic factor shared with
sm
tici
ro
u en is
Ne G ty d ani
xie
an
Figure 9.2 Comparison of heritability estimates for panic disorder
with those of other major psychiatric disorders. Comparison of the
occurrence of mental illness in monozygotic and dizygotic twins
reveals the influence of genetic factors for major psychiatric
disorders. Heritability for panic disorder is higher than for
generalized and social anxiety disorders, but lower than for
alcoholism and schizophrenia. (From [9], with permission from
Macmillan Publishers Ltd.)
[22], suggesting that early-onset PD may represent a
phenotypic subtype with greater genetic loading, and
thereby be of particular value in genetic studies.
Family studies alone however cannot tell us
whether a disorder runs in families for genetic or
environmental reasons, as family members are not
only more genetically similar to each other but are
also more likely to share the same environmental
exposures. Twin studies have the advantage that they
allow us to further disentangle the relative contribu-
tions of genetic and environmental influences. For
PD, twin studies suggest a significant genetic contri-
bution, with heritability estimates typically ranging
between 0.3 and 0.4, and as high as 0.48 [23–28].
(See Figure 9.2 for a comparison with estimates of
other major disorders.) Precise estimates vary by
ascertainment criteria, a point we will revisit later in
this chapter. Inclusion of AG with panic attacks [28]
or phobic avoidance [24] for instance increases the
heritability estimate; inclusion of sporadic panic
attacks on the other hand diminishes it, as illustrated
by a study by Perna et al. [25] showing that whereas
73% of monozygotic (MZ) and no dizygotic (DZ)
twin pairs were concordant for PD, 57 and 43% were
concordant respectively for panic attacks. Finally,
twin studies report significantly greater concordance
92
of induction of panic attacks using 35% carbon dioxide
among healthy MZ (56% concordant) than DZ (12%)
twins [29].
Even though twin studies clearly document herit-
ability for PD, they also suggest that much of this
heritability may be shared with other mental dis-
orders. In a large study of more than 5600 twin pairs
drawn from the population-based Virginia Twin
Registry, Kendler et al. [30] examined the underlying
structure of risk factors for five internalizing dis-
orders, including PD, major depressive disorder
(MDD), GAD, animal phobia, and situational phobia.
d
ize r rd
er bi
a
ism ni
a n’s MDD and GAD that explained
10% of the total
ral rde iso p ho h ol h re g to se
a
e o d l co p n
ti se variance in liability to PD; a separate genetic factor
c cia Al zo un di
o c hi H
P S S with strong loadings on animal and situational
phobias that explained 2% of the liability, and about
4% of variance that was specific to PD. In a subse-
quent examination of anxiety disorders within the
same sample, Hettema et al. [31] identified another
genetic factor that explained 25% of the total variance
for PD and GAD. Furthermore, the underlying liabil-
ity structure appeared similar across gender, despite
the higher prevalence of anxiety disorders among
women. The authors hypothesized that rather than
individual vulnerability for each disorder, a set of
genes may instead increase vulnerability to some
intermediate phenotype (such as an anxious person-
ality trait like neuroticism), which in interaction with
other genes and/or stressful life events may increase
the risk of developing one or more anxiety disorders.
Molecular genetic studies
Whereas family and twin studies can identify the
presence of genetic risk, molecular genetic studies
allow us to further probe the actual genetic markers
conferring that risk. The most common types of
molecular genetic approaches are linkage and associ-
ation studies. Linkage studies investigate the entire
genomes of individuals, using DNA markers to locate
chromosomal areas that are passed on together with a
particular disorder. Linkage studies have implicated
a number of chromosomal regions in PD, including
1q [32, 33], 2p,q [34], 4q [35], 7p [32, 36, 37], 9p,q
[34, 38], 11p [33], 12q [39], 13q [40, 41], 15q [34], and
20p [36]. However, these studies have failed to iden-
tify major gene loci, underscoring the genetic com-
plexity and likely multifactorial inheritance patterns
of the disorder.

Association studies instead test whether specific
genetic variants (polymorphisms) of interest are asso-
ciated with a given disorder, typically by comparing
affected cases with unrelated controls. Genome-wide
association studies (GWAS), which allow testing of
entire genomes without requiring a priori hypotheses,
can play an important first-step to identify candidates
which can then be more narrowly targeted in associ-
ation studies. There are however no published (or
to our knowledge, ongoing) GWAS of PD; rather,
the genetic research here has largely focused on
hypotheses within specific neurotransmitter systems,
suggested by animal models or clinical data.
Genes involved in the serotonin system
Given that drugs modulating serotonin levels are the
most efficacious and widely used line of treatment
[42, 43], the greatest convergence of genetic studies
of PD has also not surprisingly been on genes involv-
ing serotonergic transmission. Of these, the most
targeted is the gene encoding the serotonin trans-
porter protein (SLC6A4), the binding site of the
selective serotonin reuptake inhibitors (SSRIs) class
of drugs (e.g. fluoxetine, paroxetine, sertraline, citalo-
pram, escitalopram), and particularly a functional
polymorphism within the promoter known as
5HTTLPR (serotonin transporter linked polymorph-
ism region). This polymorphism was originally iden-
tified as a 14 (short) or 16 (long) base pair repeat
sequence (additional functional variants have since
been identified) that determines the amount of sero-
tonin and is related to depression and anxiety related
traits [44–46]. But numerous investigative groups
have failed to find any association [44–57] as elegantly
summarized in a recent meta-analysis [47] revealing
a null overall odds ratio of 0.9. We also found no
evidence of association within the promoter in a
recent case-control study of PD [48]; a single nucleo-
tide polymorphism (SNP) within an intron however
was associated with the panic phenotype (but not
social anxiety), and we are currently investigating its
functional significance. Though our observations of
course require replication, they nevertheless under-
score the importance of incorporating the entire gen-
etic region of the transporter rather than zeroing in
on the promoter region only.
If selective serotonin reuptake inhibitors (SSRIs)
are the first line of treatment for PD, and if they work
by targeting the serotonin transporter, can genetic
Chapter 9: Panic disorder
variation within the transporter predict who would
respond to treatment and who would not? To test this
question, Perna et al. [49] employed a 12-week study
of 92 PD patients being treated with the SSRI parox-
etine and measured whether the polymorphism was
associated with treatment response or remission.
They found that female patients with one or more
long variants of the transporter promoter (i.e. “ll” or
“sl”) had a significantly greater reduction in panic
symptoms than those with only the short variant
(“ss”), suggesting that bioavailability of serotonin
may be a primary determinant of the panicolytic
effect. The specificity to females may reflect inter-
action with sex hormones, a notion supported by
evidence that progesterone can protect against panic
attacks [50], and estrogen exacerbate them [50, 51].
A further study compared serotonin transporter bind-
ing within current and remitted PD patients to study
the effect of treatment on receptor binding [52].
Patients with PD exhibited a significant reduction in
binding within midbrain raphe nucleus, temporal and
thalamic regions of the brain. Binding within the
raphe and temporal regions was dose dependent, with
symptom severity inversely related to binding levels.
Interestingly, remitted patients still had lower binding
in the thalamus, but in all other regions they were
indistinguishable from controls. The authors posit
that the reduced binding in PD patients may reflect
either a deficit of the transmitter or some compen-
satory mechanism to increase transmitter levels, and
that treatment may serve to re-normalize transporter
binding within the temporal and thalamic regions.
Because thalamic binding does not change based on
remission, the authors advance that this may reflect a
trait marker that is independent of symptom severity.
SLC6A4 is only one of a number of genes that can
modulate serotonin levels. The serotonin receptor
family includes a heterogeneous class of G-protein
coupled receptors (except for 5-HT3, which activates
ion channels) that serve multiple pre- and post-synaptic
roles in serotonin neurotransmission [53]. A number of
polymorphisms within the receptor family have been
tested, but with inconclusive results. The 5-HT1A
receptor is the binding site for serotonergic agonists
such as buspirone, and animal studies have suggested
a role for this subtype in anxiety and depression.
Klemenhagan et al. [54] found that 5-HT1A knockout
(KO) mice freeze to a larger array of familiar and
unfamiliar cues following contextual fear condition-
ing than their wild type counterparts, who only freeze
93
 Chapter 9: Panic disorder
when presented with specific preconditioned threat-
associated stimuli. The behavior by the KO mice
parallels that of PD and post-traumatic stress disorder
(PTSD) patients, who tend to overly generalize threat
and therefore may respond with fear even to neutral
situations. In human studies, PD patients have also
been shown to have reduced 1A binding in the raphe
nucleus (the primary site of serotonergic neurons),
cingulate cortex, orbitofrontal cortex, and amygdale
[55, 56]. Interestingly, remitted patients showed
reduction in presynaptic binding but unimpaired
post-synaptic binding, suggesting that post-synaptic
receptor availability may modulate vulnerability to
PD [55]. There is also some genetic evidence that a
cytosine-guanine substitution at position 1019 of the
5-HT1A receptor may predispose to anxiety, but spe-
cificity to PD is questionable, as in one study only PD
cases with comorbid AG showed any association [57],
while in another, there was an association with panic
attacks but not PD per se [58].
For the class 2 receptors, findings are inconsistent.
Selective ablation of 5-HTR2A in mice reduces con-
flict anxiety but not depressive or fear-conditioning
related behaviors [59]. But PD patients typically do
not respond to 5-HT2A agonists [60]. A T102C poly-
morphism within 5-HT2A was associated with the
pure panic phenotype in one sample [61], but in
another, was only associated among the AG subtype
[62], and in a third there was no association even if
AG was present [63]. A fourth case-control study also
did not identify any genetic differences between cases
and controls, but within the group of PD cases, a
number of SNPs predicted symptom severity [64].
Animal models have also implicated 5-HT2C in
panic-like behaviors [65], but the human receptor,
though associated with anxiety and depressive pheno-
types [66] is not associated with panic [60, 62, 67].
Finally, polymorphisms within 5-HT1B [60, 61, 68]
and 5-HT3A [61, 69] have yielded exclusively negative
findings.
Monoamine oxidase A (MAOA) is another
important candidate as it plays a central role in the
degradation of serotonin as well as norepinephrine
(NE). A functional polymorphism upstream of the
gene that consists of a 30 base pair variable number
tandem repeat (VNTR) sequence [70] has been
targeted as it alters transcription levels of the gene,
resulting in the longer version having greater enzym-
atic activity. The VNTR polymorphism has been asso-
ciated with the panic phenotype in at least three
94
association studies [68, 71, 72], each of which showed
an association among females only. Although a link-
age study did not replicate these findings [73] that
may have been because power was limited, and only
half of the families in that study stemmed from female
probands. Regardless, the sex specificity is interesting,
especially given the location of the MAOA gene is
located on the X-chromosome, and may reflect inter-
actions with sex hormones or other X-linked genes.
A separate polymorphism (T941G) within the MAOA
gene itself has been associated with generalized anx-
iety but not PD [74], and a study testing synergy
between MAOA, VNTR, and the serotonin trans-
porter promoter polymorphism (5-HTTLPR) did
not find any interactions.
Finally, whereas MAOA degrades serotonin, tryp-
tophan hydroxylase (TPH) serves as the rate-limiting
enzymatic step in it synthesis. Unlike MAOA, how-
ever, reports of TPH have been mostly negative [68,
75–78] including tests of a recently recognized variant
preferentially expressed in neurons [79].
Genes involved in the dopamine system
Of the genes related to dopaminergic function,
catechol-O-methyltransferase (COMT), a methylation
enzyme that degrades catecholamines (such as dopa-
mine, epinephrine, and NE) and is the primary deter-
minant of cortical dopamine levels, has been the most
encouraging. A conversion of a valine to a methionine
at position 158 has been much focused on due its
functional significance: the valine variant results in
three to four times as much enzymatic activity as the
methionine. This polymorphism, commonly known
as val158met (dbSNP rs4680) has been associated
with PD in a number of studies, particularly among
women [80–85], but these are tempered by some
failures to replicate [72, 86, 87] or reports of nonspe-
cificity [23]. Even within the positive studies, some
report the valine encoding allele (i.e. an excess of
COMT) associated with presence of PD, but the stud-
ies by Woo et al. [84, 85] finding the reverse.
This apparent discrepancy might perhaps be
explained due to ethnic heterogeneity, given the find-
ings of a recent meta-analysis showing no overall
association between COMT val158met and PD, but a
val encoding allele effect on PD among Caucasians,
and a met effect in Asian populations [80]. In both
cases, though, the association was predominantly
among females, paralleling gender-specific reports of

COMT in obsessive–compulsive disorder [88] and
schizophrenia [89]. Though the mechanisms of this
gender specificity are not entirely clear, estrogen may
play a modulating role, as the COMT gene contains
an estrogen response element within its promoter
region [90]. Interestingly, the COMT enzyme also
plays a role in the conjugation of estrogens, so the
two products may cross-regulate each other [91].
Freitag et al. [92] further reported an interaction
between the aforementioned COMT val158met and
5-HT1A 1019C/G polymorphisms among PD cases.
Interestingly, they did not find any additive effects:
instead, the risk for PD associated with each variant
was highest in the presence of the low risk allele of
the other polymorphism, suggesting “a ceiling effect
at the molecular level . . . that if one element of mono-
aminergic transmission is dysfunctional, dysfunction
of other elements does not further increase the risk
for PD” [92]. Even though this observation should be
treated cautiously given the small sample, it serves to
remind us of the complexity and interdependency of
genetic risk factors. A separate study similarly tested
for interactions between COMT and the MAOA,
but did not find any [81]. Finally, dopamine receptor
1 (DRD1) has been associated with PD [61], but
studies of DRD2 and 4, and the dopamine transporter
(DAT) have yielded mostly negative results [93, 94].
Genes involved in the adenosine system
Though not a neurotransmitter in the classical sense
(it is a purine nucleoside), adenosine can alter levels
of several other neurotransmitters and neuromodula-
tors [95]. There are four receptor subtypes, but the
adenosine receptor 2A (ADORA2A) on chromosome
22, is the most provoking candidate as it is preferen-
tially expressed in the brain, and excess caffeine – a
potent 2A antagonist – can trigger panic attacks [96].
The subtype has been associated with a number of
psychiatric phenotypes in mice [97]. There is now
evidence that exons within the ADOR2A but not
other receptor subtypes may be involved in human
PDs as well, with two Caucasian populations showing
a significant association [98, 99].
Genes involved in neuropeptides systems
A number of neuropeptides have also been associated
with PD including neuropeptide Y Y5 receptor [100],
tachykinin receptor 1 (TACR1) [101], and gastrin
releasing peptide (GRP) [101]. Cholecystokinin
Chapter 9: Panic disorder
(CCK) and its receptors have been of particular inter-
est as administration of CCK can induce panic attacks
in healthy individuals, PD patients are more sensitive
to CCK than controls, and anti-panic medications
can alter the CCK gene [102, 103]. SNPs within the
CCK gene itself have been linked to PD [104–106];
polymorphisms within the receptor however have
yielded both positive and negative findings for CCK-A
[107–109] and largely negative findings for CCK-B
[106, 108, 110–112].
Keck et al. [113] examined the relationship
between the vasopressin (AVP) neuropeptide system
and the corticotropin releasing hormone (CRH), gen-
otyping 71 SNPs within the CRH, CRHR1, CRHR2,
AVP, AVPR1A, and AVPR1B genes. A number of
polymorphisms within CHRH1 and AVPR1B were
nominally associated with PD, and 15 of them showed
significant gene-by-gene (epistatic) interactions.
These findings are provoking given that the expres-
sion of CRHR1 and AVPR1B has been shown to most
strongly co-localize in the hippocampus, amygdala,
and pituitary gland, all brain regions relevant to anx-
iety. Furthermore, unlike so many other polymorph-
isms identified, the most significant SNPs lie within
genetically consequential exon (AVPR1B) and 3’ UTR
(CRHR1) regions, allowing for the genetic variation
to have a direct functional impact. Finally, the associ-
ations are concordant both with murine models asso-
ciating these hormone receptors with anxiety related
behaviors [114], and with human linkage studies
of PD implicating regions of chromosome 1q that
contains the human AVPR1B gene [32].
Genes involved in GABA-ergic system
The efficacy of benzodiazepines in treating panic
symptoms spotlighted the role of the g-aminobutyric
acid (GABA) system. No significant associations to
our knowledge have been found thus far for any of
the GABA receptor subtypes [93], although the b3
subunit has been implicated in PTSD [115]. However,
lower levels of glutamic acid decarboxylase (GAD), a
key enzyme in the synthesis of GABA, has been asso-
ciated with depression and bipolar disorder, as well as
in neurotic patients [116]. In humans, GAD exists as
one of two isoforms, encoded by the GAD1 and GAD2
genes. Though there have been no concrete findings
for PD, variation within the GAD2 gene has been
related to behavioral inhibition, and is anxiety related
[117], whereas that within GAD1 has been linked to a
95
 Chapter 9: Panic disorder
genetic risk commonly shared across multiple anxiety
and depressive phenotypes [118].
Genes involved in the regulators
of G-protein signaling family
Finally, associations with PD have been found for a
family of proteins known as the regulator of G-pro-
tein signaling (RGS). The RGS proteins can decrease
G-protein function via GTPase activity, and have
been shown to decouple autoreceptors from noradre-
nergic neurons in the locus coeruleus [119]. Two
members of the large RGS family have been involved
in the etiology of PD; these two genes lie on chromo-
some 1q, one of the risk loci suggested for PD [33].
The RGS2 variant has been linked to PD in males
[120] and RGS7 [121] in females.
Genes related to carbon dioxide
hypersensitivity
PD is unique within the anxiety spectrum in that its
central manifestation, the panic attack (PA), can be
incited in laboratory settings by specific chemical
challenges such carbon dioxide (CO2) inhalation.
The recognition that an increase in brain CO2 levels
is an indicator of potential suffocation, coupled with
the ability of CO2 inhalation to generate panic attacks,
has promoted the false suffocation alarm theory of
PD [122], in which a hypersensitive suffocation alarm
system triggers the suffocation alarm erroneously (i.e.
in the absence of any real danger) producing respira-
tory distress, hyperventilation, and panic. In that
sense, initial attacks may arise from an unconditioned
biological origin, but among those who go on to
develop the full disorder, successive associative learn-
ing or classical conditioning may continue to generate
learned alarm responses to internal or external cues.
There is a body of converging evidence for a
genetic basis to CO2 sensitivity. Unaffected FDRs of
PD patients have increased hypersensitivity to CO2
[123, 124], and MZ twins are substantially more con-
cordant on CO2 induced panic induction than DZ
twins [29]. Genetic modeling suggests that common
genetic factors may account for trait sensitivity to
CO2 challenge and naturally occurring PAs [125],
and that transmission of PD is higher among families
in which the proband is responsive to the CO2 chal-
lenge [126]. Given these observations, the discovery of
genetic variants associated with individual differences
96
in sensitivity to hypercapnia may also serve as useful
markers to study the etiology of human PD, especially
given that this phenotype appears selective to panic.
Serotonin is thought to play an important role in
hypersensitivity: serotonergic neurons in the medulla
are central respiratory chemoreceptors, and seroto-
nergic neurons in the midbrain of rats are also highly
chemosensitive to small changes in blood CO2, indu-
cing changes in arousal, anxiety, and cerebrovascular
tone [127, 128]. In a genetic study of the serotonin
transporter gene, Schmidt et al. found that healthy
subjects homozygous for the long allele of the 5-
HTTLPR promoter polymorphism showed higher
sensitivity and a stronger fear response to 35% CO2
[129] than subjects with the heterozygous or homo-
zygous for the short allele. But a subsequent extension
of this hypothesis to study a population of PD
patients [130] failed to find any differences. Philbert
et al. [131] examined lactate dehydrogenase (LDH),
the enzyme that converts pyruvate (the final product
of glycolysis) to lactic acid under anaerobic condi-
tions. LDH has two separately encoded variants,
LDH-A and LDH-B. The authors found that a haplo-
type within the fifth exon of the LDH-A gene predicted
ventilation response to CO2 inhalation that could
be differentiated between subjects at high and low risk
for PD. Finally, the cholinergic system has also been
postulated to play an evolutionary role in CO2 sensi-
tivity [132], although there is no molecular evidence
implicating specific gene variants to our knowledge.
Genetic studies: conclusions
If there is a single conclusion from the above studies,
it is one of inconsistency. For many of the reported
associations, replication in independent samples – so
vital to psychiatric genetics [133] – has either not
been attempted, or has failed. Some failures to repli-
cate might be method-related, such as small sample
sizes or variation in how the panic phenotype is
operationalized. But other failures could represent
true negatives, with the originally reported associ-
ations having being the false-positive victims of popu-
lation stratification or use of overly liberal statistical
tests, etc. Regardless, given that negative findings have
a lower overall likelihood of being published, true
replication rates are likely even lower than what a
literature search would bear forth.
Given these concerns, it is imperative that investi-
gators rigorously document their study characteristics

so that when discrepant findings emerge, they can be
probed analytically rather than simply descriptively.
Some characteristics such as gender, age, and ethni-
city, are obvious, yet critical, as they can substantially
impact the findings. (Consider the earlier discussion
of COMT: not only were associations gender-specific,
but they veered in opposite directions across different
ethnic groups.) Others, such as onset age or family
history, however are less frequently considered even
though these may impact genetic loading. Interview-
ing relatives directly may be prohibitive, but there are
a number of standardized instruments such as the
Family History Screen (FHS) [134] and the Family
Interview for Genetic Studies (FIGS) [135] through
which family history can be rapidly and reliably ascer-
tained. Finally, the importance of selecting appropri-
ate controls cannot be understated, as illustrated in a
recently published analysis comparing effect sizes in
PD studies using different control populations [136].
Controls should be recruited and assessed using the
same methods and instruments as cases. Ideally, they
should be free of all anxiety disorders, depression, and
substance use, since these are frequently comorbid
with PD and could fuzzy the diagnosis. They should
not only be matched on age, gender, and ethnicity,
but also be selected to ensure that they have passed
the age of greatest risk for the disorder of interest. PD
for instance onsets in early adulthood; inclusion of
teenagers as controls might contaminate the sample
with some subjects who though not symptomatic,
may still be at the same genetic risk as persons with
the disorder. Having said that, the more restrictive the
selection criteria, the smaller will be the available pool
of subjects to recruit from, and the trade-offs between
sample size and sample homogeneity will need to be
carefully evaluated for each context.
Genetic studies, while instrumental, cannot alone
address the etiological complexities of most psychiatric
disorders. In the remaining sections of this chapter, we
turn to two integrative approaches that combine gen-
etics with other clinical or biological methods to target
the underlying mechanisms. In the first, we discuss
exploiting the relationship between psychiatric and
nonpsychiatric medical manifestations (the “expanded
spectrum” approach). This approach is particularly
relevant to PD, where the panic attacks are accompan-
ied by a range of physiological responses that may be
central to the etiology. Then, we focus on neurobio-
logical phenotypes, and in particular, on using meas-
ures of brain structure and function to identify genetic
Chapter 9: Panic disorder
variation and then study the mechanisms via which
genes can impact behavior.
The “expanded spectrum” approach
and a possible PD syndrome
In this section, we describe a series of studies on
a potential PD syndrome conducted by our group,
which grew out of a single initial observation: specific
medical conditions tended to cluster nonrandomly
within some PD families but not others. The objective
of this section is not only to provide an overview of
this syndrome, but also to serve as an example of an
investigative approach that can be used to refine the
clinical phenotype in ways that may be useful to
genetic investigation. The overall arc of these studies
is conceptualized in Figure 9.3; details of individual
studies are summarized in Table 9.1 [48, 136–139].
A distinguishing feature of PD is that is frequently
accompanied by a range of medical symptoms,
including cardiovascular (e.g. shortness of breath,
palpitations), gastrointestinal (nausea, abdominal
distress), urinary, and neurological (dizziness, trem-
bling) [140–143] (see also Figure 9.1). It is in fact
often these clinical manifestations that drive individ-
uals to their doctor or emergency room, where their
PD is first identified [7–9]. In the course of a previous
genetic study of multiplex families with PD, we
observed that some medical problems – particularly
those related to bladder, thyroid, and mitral valve
function – were not randomly distributed. Instead,
they tended to cluster within some families with PD
but not others. We hypothesized that families who
had this clustering of conditions might perhaps be
genetically different from those that did not, and
examined the linkage patterns after reclassifying the
families based on the presence or absence of these
medical conditions. Indeed, reclassification revealed a
strong linkage peak (logarithm of odds [LOD] score:
3.6) on marker D13S779 of chromosome 13q, raising
the hypothesis of a possible syndrome wherein these
diverse symptoms could be sub-served by a common
genetic mechanism related to this chromosome
(pleiotropy) [41, 136]. These original findings were
subsequently replicated in a larger sample of 587
individuals from 60 multiplex pedigrees (including
the original 19 families), that reported a linkage peak
on marker D13S793, only 7 cMs from the original
peak [40].
97
 (a)
PD probands First-degree relatives
80

70

60
Frequency (%)

50

40

30 +

20

10

0
IC

MVP

MIG

TH

ANY

IC

MVP

MIG

TH

ANY
Medical conditions

(b)
IC probands First-degree relatives
80
+
70

60
Frequency (%)

50

40

30

20

10

0
PD

MVP

MIG

TH

ANY

PD

MVP

MIG

TH

ANY

Medical conditions
Figure 9.3 Case-control studies of panic disorder (PD) and interstitial cystitis (IC). (a) Rates of IC and other hypothesized medical
conditions among probands with and without PD (left) and their first-degree relatives (FDRs) (right). PD probands were at significantly
increased risk for IC, mitral valve prolapse (MVP), and migraines than controls; FDRs of PD probands were also at increased risk for IC, MVP,
migraines, and thyroid problems than FDRs of controls. (b) Rates of PD and other hypothesized medical conditions among probands with and
without IC (left) and their FDRs (right). IC probands were at significantly increased risk for PD and thyroid problems; FDRs of IC probands were
at increased risk for PD and migraines than FDRs of controls. ANY, any syndrome condition (that is, PD, IC, MVP, MIG, or TH); IC, interstitial
cystitis; MIG, migraine headaches; MVP, mitral valve prolapse; PD, panic disorder; TH, thyroid problems.
**** p < 0.001; *** p < 0.005; ** p < 0.01; * p < 0.05; + p < 0.1
P-values reflect odds ratio, adjusted for age and gender. For data on FDRs, the odds ratios are further adjusted for whether or not the
proband reported the same condition. For example, the odds ratio for whether a FDR of a panic proband was more likely to have mitral valve
prolapse than a FDR of a control proband, were adjusted for the FDRs age, gender, and whether or not the corresponding proband had
mitral valve prolapse. Darker bars indicate probands (i.e. PD cases in part (a); IC cases in part (b)); lighter gray bars indicate controls. In each
figure, the left panel shows data for probands versus controls; the right panel shows data for FDRs of probands versus FDRs of controls.
PD was diagnosed based on DSM-IV criteria using the Schedule for Affective Disorders-Lifetime version (SADS-LA-IV). Medical conditions in the
proband were ascertained by a medical checklist at the time of the interview; IC was also assessed using screening criteria developed by
the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) for use in genetic studies of IC. Medical conditions in the FDRs
were assessed using the Family History Screen, with the proband as informant.

98
 Chapter 9: Panic disorder

Table 9.1 Potential PD syndrome: study characteristics.

Study Sample Assessments Ref.

1. Family study of PD 34 families Schedule for Affective Disorders and Fyer et al., 1999;
Schizophrenia – Lifetime Version IV Weissman et al., 2000
2. Replication of (1) 60 families Schedule for Affective Disorders and Hamilton et al., 2003
Schizophrenia – Lifetime Version IV
3. Case-control study 67 IC cases; 79 non-IC Assessment of IC: Urodynamics and/or Weissman et al., 2004
of IC urological controls cystoscopy with bladder distention
Assessment of PD and other psychiatric
disorders: Schedule for Affective Disorders
and Schizophrenia – Lifetime Version IV
4. Clinical case- Cases: 219 (PD); 199 Assessment of IC: Medical history; NIDDK Talati et al., 2008
control (SAD); 173 (PD + SAD); screen for IC
study of PD Controls: 102a. Assessment of psychiatric disorders: Schedule
for Affective Disorders and Schizophrenia –
Lifetime Version IV
Required onset by age 30 and family history
of anxiety in  1 FDR
5. Genetic case- Casesb: 179 (PD); 161 Schedule for Affective Disorders and Strug et al., 2008
control (SAD); 140 (PD + SAD); Schizophrenia – Lifetime Version IV
study of PD Controls: 470c.
6. Replication of IC Casesb: 179 (PD); 161 Assessment of IC: Medical history; NIDDK Underway
linkage findings (SAD); 140 (PD + SAD); screen for IC
from (1) and Controls: 574d. Assessment of psychiatric disorders: Schedule
(2) using case- for Affective Disorders and Schizophrenia –
control sample (4) Lifetime Version IV
a
These controls were collected as part of our study, and were required to have no lifetime personal or family history of any psychiatric disorder.
b
These are the same cases as those in the clinical study (no. 4). The sample is smaller here due to subject attrition (some subjects
completed the interview but did not give blood), or genotyping failure.
c
The controls for this analysis were not originally collected as part of our sample but rather obtained from a national repository (the Rutgers
University Cell and DNA repository [RUCDR]) that had been set up by the National Institute of Mental Health’s (NIMH’s) Human Genetics
Initiative to aid genetic studies. Subjects were assessed using an on-line self-report based on the Composite International Diagnostic
Interview–Short, assessing eight syndromes: major depressive episodes, panic attacks, agoraphobia, social phobia, specific phobia, generalized
anxiety disorder, obsessive–compulsive disorder, and alcohol and drug dependence (see Talati et al. 2008 [136] for more information on
the NIMH controls). From the available sample of more than 3000 subjects, we selected the 480 subjects who reported no history of any
anxiety, depressive or substance use disorders, and were in the same age and gender range as our cases.
Abbreviations: FDR, first-degree relative; IC, interstitial cystitis; NIDDK, National Institute of Diabetes and Digestive and Kidney Disease;
PD, panic disorder; SAD, severe anxiety disorder.

Of the various conditions clustering with PD, the
strongest linkage evidence was for kidney and bladder
problems (LOD scores > 3). It was unclear at the time
what these symptoms represented, so we presented
the findings to expert urologists, who suggested these
problems may be related to interstitial cystitis (IC). IC
is a low-prevalence (
0.5%, lifetime) but chronic and
debilitating syndrome characterized by frequent or
urgent need to urinate, and pain and discomfort on
bladder filling, not accounted for by a urinary tract
infection [144]. Like PD, IC is more common among
women than men, although with later onset at around
40 years [145], and is accompanied by neurological,
gastrointestinal, and musculoskeletal distress [145–147].
Data from twin studies allude to possible genetic
underpinnings [148, 149], and a large genetic study
of IC (Maryland Genetics of Interstitial Cystitis
[MAGIC]) is currently underway.
To follow up on the linkage patterns from our
family studies, we designed a case-control study of
IC, followed by a larger case-control study of PD.
For the former, we recruited 67 IC cases who were

99
 Chapter 9: Panic disorder

(a) Figure 9.4 Conceptualization of the

studies to test the panic disorder
Initial observation during a genetic linkage study of PD that
syndrome. FDR, first-degree relatives; IC,
specific medical problems appear to cluster nonrandomly among
interstitial cystitis; LOD, logarithm of odds;
some families but not others
NIDDK, National Institute of Diabetes and
Linkage study of PD

Digestive and Kidney Diseases; PD, panic

Reclassification of families based on presence/absence disorder; SAD, social anxiety disorder.
of medical conditions and reanalyze data

Linkage peak on chromosome13q, marker D13S779

Replication of above finding in a larger sample of 60 families, with

linkage peak on marker D13S793, only 7 cM from first peak

(b)
Association study of IC

Largest LOD score for bladder problems; urologist consultation

suggests symptoms represent interstitial cystitis (IC)

Design new case-control study of IC versus non-IC urological

patients

Findings: IC patients are at increased risk for PD; so are their first
degree relatives, regardless of whether or not the proband had PD

(c)
Case-control study of PD and SAD using revised criteria to screen
Association study of PD

for IC developed by the NIDDK for use in genetic studies

Panic probands at five-fold higher risk for IC, and two-fold or

greater risk for mitral valve, migraines. FDRs also at higher risk for
these disorders regardless of proband having these conditions

Fine mapping to chromosome 13q in the area originally

identified by the linkage analysis to hone in on the functional
genetic region that may be involved in this pleiotropic syndrome

well-documented by urologists using urodynamics PD status, consistent with a pleiotropic model of

and/or cystoscopy with bladder distention [138]; con-
trols (N = 79) were drawn from patients with other
urological disorders but not IC. A total of 815 FDRs
were also assessed using the family history screen
[134]. We found that 27% of IC patients in the
sample, as compared to only 8% of controls, had been
diagnosed with PD, an adjusted 4-fold increase. Very
similar patterns have since been independently
observed in the Events Preceding Interstitial Cystitis
(EPIC) study, where 27% of IC cases but 3% of con-
trols had a PD diagnosis [Warren et al., unpublished].
Strikingly, in our study, the FDRs of probands with
IC were also three-fold more likely to have PD than
the FDRs of controls, regardless of the proband’s own
transmission that extends beyond either PD or IC
alone. The results are summarized in Figure 9.4b.
The second case-control study was designed to
examine the same question from the opposite per-
spective: i.e. patterns of IC within PD cases (Figure
9.4c). In order to test for specificity, we also included
an additional group of persons with social anxiety
disorder (SAD). IC was now assessed using revised
criteria developed by the National Institute of Dia-
betes and Digestive and Kidney Diseases (NIDDK)
for use in genetic studies [150]. To increase the likeli-
hood of capturing a genetic phenotype (see earlier
discussion) cases were required to have onset prior
to age 30, and a documented history of anxiety in

100

at least one FDR. Controls, by contrast, were required
to have no evidence of any psychiatric disorder over
their lifetime, have no family history of anxiety, and
be at least 30 years old at the time of assessment
(the last criteria serving to minimize capturing those
still at risk).
The findings in this latest sample were even more
robust than our previous studies. Here, PD probands,
as compared to controls, were more than nine-fold as
likely as controls to report IC symptoms (but not any
other genitourinary problems), and more than 4% of
them had been formally diagnosed by a physician or
urologist with IC [150]. Probands with PD were also
at two-fold or greater risk for mitral valve prolapse,
migraines, hypercholesterolemia, colitis and ulcers.
Paralleling the observations of the case-control study
of IC, the FDRs of PD probands were also at increased
risk for these disorders vis-á-vis those of controls,
regardless of the proband’s status.
Potential mechanisms
Why is it that some medical conditions, but not
others, cluster with PD, and what mechanisms could
explain the clustering patterns? One possibility is that
the medical conditions are themselves stress-provoking,
and lead to panic symptoms. Though this may be true
in some cases, the relative onset ages of the medical
conditions vis-á-vis PD make it an unlikely primary
explanation: PD onsets in early adulthood whereas IC
does not typically until well into the forties. An alter-
nate explanation may stem from the observation that
a number of the clustering medical conditions involve
smooth muscle functions, suggesting possible dysre-
gulation of the autonomic nervous system (ANS). PD
results in part from a failure to regulate the ANS,
which results in increased reactivity and an excess of
NE release from the brainstem, and particularly the
locus coeruleus (LC) (which comprises more than
70% of all NE projections to the forebrain) [151–
153]. In asymptomatic individuals, the NE activity
in the LC is cross-regulated by serotonin projections
from the raphe nuclei and by neuromodulators such
as CRF, maintaining a homeostatic balance [154].
Breakdown of this regulation results in the increased
reactivity and stress associated with many of the anx-
iety phenotypes. But the same locus coeruleus–nore-
pinepherine (LC-NE) system also activates the
hypothalamus–pituitary–adrenal axis (HPA) and the
sympathetic branches of the ANS , which subsequently
Chapter 9: Panic disorder
innervate smooth muscles throughout the body. The
dysregulation leading to the increased fear response
could therefore couple with the excess sympathetic
projections to muscles in pulmonary, cardiovascular,
genitourinary, and gastrointestinal organs, accounting
for the coexistence of PD and the syndrome’s other
medical conditions.
Though not the only explanation, a number of the
observed conditions fit nicely within the autonomic
dysregulation model. For example, autonomic nuclei
are the major source of innervation to the bladder
and urethra [155, 156], and IC patients demonstrate
abnormal vasomotor tone, increased density of sym-
pathetic neurons in their bladder, and increased
excretion of NE, all suggestive of autonomic dysfunc-
tion [145, 157, 158]. Animal models also show a
marked decrease in cortisol response to stress in cats
with feline IC [159]. Patients with mitral valve pro-
lapse (MVP) similarly manifest elevated NE levels and
increased vagal tone [160]. Strikingly, MVP in panic
patients tends to be disproportionately of the noncal-
cified type, whose etiology has been linked to auto-
nomic function [161], and MVP symptoms can be
ameliorated by treating with PD medication only
[162]. Autonomic abnormalities have also been
reported in individuals with migraine [163] and
migraine pathophysiology involves inappropriate dila-
tion of cerebral blood vessels, an action which is under
autonomic control [164]. Similar links have been
reported for gastrointestinal disorders [165, 166] and
hypercholesterolemia [167, 168].
Collectively, our observations are consistent with
a pleiotropic model in which vulnerability conferred
by a genetic change (possibly on chromosome 13q)
gives rise to dysfunction in multiple autonomically
regulated sites, resulting in the host of seemingly
disparate clinical phenotypes. The previously identi-
fied region of chromosome 13q encompasses, among
others, genes encoding the endothelin B receptor
(a promising candidate, related to vasoconstriction),
the serotonin 2A (HTR2A) receptor, and multiple
transcription factors [40, 41], and further genotyping
is currently underway to more specifically identify
the susceptibility locus for this interesting clinical
syndrome.
The coupling of psychiatric and nonpsychiatric
symptoms we found is hardly novel, having been
previously reported for disorders such as Huntington’s
disease [169], Wolfram’s syndrome [170], and rheuma-
toid fever [171, 172]. But few psychiatric genetic studies
101
 Chapter 9: Panic disorder
have exploited the links between psychiatric and
nonpsychiatric manifestations to explore underlying
mechanisms, instead typically relegating the latter
to comorbidities or confounders [173]. Contrast
this with diabetes research, for example, where the
study of glucose metabolism has informed our under-
standing of the associated neuropathy, retinopathy,
and nephropathy [174]. Our observations support
using “expanded spectrum” approaches wherein psy-
chiatric phenotypes can be refined or morphed with
other associated clinical symptoms in ways that may
provide more tractable substrates at the biological level.
Neuroimaging studies of PD
In the course of the last decade, neuroimaging has
emerged as a potent tool via which to address the
machinery of the brain and the mind. Understanding
the neural circuitry of normal brain functioning
allows us to predict and then test how breakdown
of that circuitry might lead to psychopathological
phenotypes of interest. In this section we discuss some
of the recent advancements in neuroimaging of PD.
Although we review both structural and imaging stud-
ies, we especially focus on imaging-genetics approaches,
as integration of genetics with neuroimaging data
may ultimately allow us to most comprehensively
probe the underlying relationship between genes and
psychopathology.
Structural studies
Structural imaging studies have identified changes in
a number of cortical and sub-cortical brain regions of
PD patients, including volume reductions in the right
anterior cingulate cortex [175, 176], bilateral amyg-
dala [177, 178], putamen [179], left hippocampus
[178] and parahippocampal gyrus [180], and volu-
metric increases within the rostral pons [181] and left
insula [176]. Both decreases [178, 182, 183] and
increases [176] of gray-matter volume have been
reported within the temporal cortex, and temporal
lobe abnormalities have been associated with earlier
age of onset and higher rates of panic attacks [183].
Oddly, in the hippocampus, later onset has been asso-
ciated with greater volumetric reduction [178],
though this has not been replicated. These findings
are summarized in Table 9.2 [175–183]. Consistent with
the structural studies, resting state positron emission
tomography (PET) studies have also reported higher
glucose uptake (indicative of greater regional activity)
102
within the amygdala, hippocampus, thalamus, and
midbrain of PD patients [184].
Although an exhaustive neuroanatomical review is
beyond the scope of this chapter, it bears note these
regional findings are consistent with the proposed
cortical circuitry of fear and anxiety. The anterior
cingulate is centrally involved in the processing of
emotional (especially negatively valenced) stimuli,
and volume reduction may diminish the ability to
ascribe the appropriate emotional weight to incoming
information, leading to even neutral stimuli being
perceived as threatening. The anterior cingulate
cortex also downregulates sub-cortical structures such
as the amygdala that generate the fear response;
reduced volume may therefore reflect a decoupling
of this regulation. The insula is additionally involved
in integration of afferent and efferent cortical connec-
tions to the brain, and PD patients have reduced
GABA binding here [185], suggesting that inhibition
by the insula may protect against PD. On the other
hand, the increase in midbrain nuclei is consistent
with their role in unconditioned fear responses [186].
Furthermore, the reduced serotonin transporter bind-
ing found in the midbrain of PD patients [52], and the
observation that treated PD patients show response-
related changes in glucose binding [187] allude to a
functional role for a lack of serotonin inhibition
within this region.
Functional studies
There are few functional imaging studies of PD.
Most studies have relied on experimental proxies
for the panic phenotype, as directly imaging panic
attacks is, for a variety of ethical and logistic reasons,
unfeasible. There is a case report of a 26-year-old
woman who experienced a spontaneous attack while
being scanned. Her scan shows increased amygdala
firing during the time period corresponding to the
attack [188] but because this was a single scan, it is
difficult to dissect causal from compensatory or medi-
cation-related (she was on an SSRI) elements. In an
unrelated PET study, another healthy female subject
experienced an unexpected panic attack in the scanner
while being subject to electric shocks as part of a fear-
conditioning paradigm. Here, the panic attack was
associated with reduced blood flow in right orbito-
frontal, prelimbic, and anterior cingulate regions
but no changes in the amygdala [189]. Induction of
panic attacks among healthy volunteers using sodium
 Chapter 9: Panic disorder

Table 9.2 Structural brain changes associated with PD.

Anatomical Source Sample Hemisphere Effect (PD versus controls)

region (Cases/
Controls)
Amygdala Massana et al., 2003 12/12 L and R Decrease in volume
Uchida et al., 2003 11/11 L and R Decrease in volume
Anterior cingulate Asami et al., 2008 26/26 R Decrease in volume
Uchida et al., 2008 19/20 R Decrease in volume
Hippocampus Uchida et al., 2003 11/11 L Decrease in volume
Parahippocampal Massana et al., 2003 18/18 L Decreased gray matter intensity
gyrus
Insula Uchida et al., 2008 19/20 L Increase in volume
Midbrain/pons Protopopescu et al., 10/23 L Increase in volume
2006
Uchida et al., 2008 19/20 L Increase in volume
Putamen Yoo et al., 2005 18/18 L and R Decrease in volume
Temporal lobe Fontaine et al., 1990 31/20 R Decrease in volume
Ontiveros et al., 1989 30/20 R Greater MRI abnormalities, correlated with
younger age of onset
Uchida et al., 2003 11/11 L and R Decrease in volume; preferentially in the
left hemisphere
Uchida et al., 2008 19/20 L Increase in volume
Abbreviations: L, left; MRI, magnetic resonance imaging; PD, panic disorder; R, right.

lactate has been reported to increase activation in the
insula, temporal lobe and the periacqueductal gray
region (PAG) [190], whereas induction by CCK-4
reported increased activity within hypothalamus,
insula, and PAG [191]. The activation of the PAG is
important as this region plays a critical role in the
processing of unconditioned fear, a hallmark of panic
attacks [192]. But because these are single studies
or case reports, the findings should be interpreted
cautiously. Additionally, data from panic induction
studies should not be overly generalized, as experi-
mentally induced panic attacks do not evoke all of the
symptoms of the disorder with equal reliability and
may represent a more restrictive or etiologically
biased phenotype [193].
Most functional imaging studies of PD therefore
do not target “active” panic symptoms, but rather,
underlying trait differences that distinguish those
with and without the disorder. This of course is
contingent upon the differences of interest being
detectable in the scanner by appropriately targeted
experimental paradigms, even if the subject is currently
asymptomatic. But current symptom levels at the time
of the scan will also vary across patients, and the
scanner can itself be a panicogenic agent. Current
symptoms can greatly influence task performance
and outcome; therefore, even if the target is under-
lying trait differences, it is imperative that state
differences be accounted for. One approach is via
quantitative symptom measures such as the Spielber-
ger State Trait Anxiety Inventory (STAI), Acute Panic
Inventory, or the Hamilton Rating Scale for Anxiety
(HRSA), which can be administered at the time of
the scan (ideally at multiple points), and can then be
subsequently modeled into the statistical analysis.
The most widely used functional paradigms to
study anxiety disorders involve testing brain
responses to fear, based on the idea that anxiety states
may arise from a failure of the brain to regulate fear
processing. The model of anxiety disorders as an
exaggerated extension of the natural fear system is
an attractive one to study, as the neural circuitry of
fear has been well worked out in both animal models
and humans [194]. Most of the proposed pathways
have implicated a top-down control mechanism,
wherein various frontal structures downregulate

103
 Chapter 9: Panic disorder
activity sub-cortical structures (notably the amyg-
dala), resulting in the generation and extinguishing
cycle of the fear response [195–197]. Anxiety disorders
may result from a breakdown of this regulation,
with the precise mechanisms varying by phenotype.
A recent meta-analysis [197] reported that the specific
and social phobias, where exaggerated fear was the
primary feature, invoked greater activation in the
amygdala and insula in response to viewing negatively
valenced emotional stimuli; PTSD, in contrast, wherein
fear was part of a larger emotional disturbance, showed
reduced activation within cingulate cortex. In PD
patients, symptom-related reductions in cingulate activ-
ity have been reported while viewing fearful faces [198]
(as well as a converse increases while viewing happy
faces [199]); however, the above meta-analysis did
not consider PD owing to an overall paucity of studies.
We currently have an ongoing study to test where
PD falls along this spectrum. Subjects with PD, along
with a comparison group of those with social anxiety
disorder (SAD) and a healthy control group, are
scanned while performing experimental paradigms
that target facets of emotional processing. In the first
paradigm, illustrated in Figure 9.5 [200], subjects are
imaged while passively viewing masked (uncon-
sciously perceived) or unmasked (consciously per-
ceived) fearful faces. (The masked conditions are
generated by presenting a fearful face first briefly
[in this paradigm, 33 ms], followed by a matched
neutral face for the remaining duration of the stimu-
lus [167 ms], such that the fearful face is not con-
sciously perceived by the subject.) Brain scans on
healthy volunteers have found that the conscious
and unconscious fear processing activate different
subregions of the amygdala [200] (Figure 9.6): uncon-
scious fear (experimentally speaking, [activity masked
fearful] – [activity masked neutral]) activates the basolateral
complex, whereas conscious fear [activity unmasked fear-
ful] – [activity unmasked neutral]) activates the dorsal
region. Importantly, unconscious, but not conscious,
processing in the amygdala correlates with trait anx-
iety. This makes sense behaviorally: when faced with
unknown or unrecognized threats, subjects are more
likely to respond based on their pre-existing individual
proclivities to anxiety (trait). But when the context
of the threat is clarified (unmasked), these differences
converge as subjects respond more uniformly. The above
data raise the question of whether (and if so, which?)
anxiety phenotypes are disorders of extreme uncon-
scious emotional vigilance. Our study allows us to
104
Run
15s “on” 15s “off”
Epoch
Event
200 ms 400 ms 200 ms 1200 ms
Nonmasked:
Fearful (F)
or
Neutral (N)
or
200 ms 400 ms 33 ms 167 ms 1200 ms
Masked:
Masked
Fearful (FN)
or
Masked
Neutral (NN)
Figure 9.5 Experimental paradigm to test brain responses to
masked versus unmasked fearful stimuli. Stimuli were faces with
either fearful (F) or neutral (N) expressions, which were pseudo-
colored in red, yellow, or blue. The faces were presented in either
masked or unmasked conditions. For the unmasked condition, the
F or N face was shown for the full stimulus duration (200 ms).
For the masked conditions (FN and NN), a fearful or neutral face
was presented for the first 33 ms, followed by a neutral face of
the same gender and color for the remaining 167 ms. Activity
attributable to unconscious fear processing was calculated as
[activity masked fearful] – [activity masked neutral]); conscious fear
processing was [activity unmasked fearful] – [activity unmasked neutral]).
(From [200], with permission from Elsevier.)
compare the processing of consciously and uncon-
sciously perceived fear and address this question.
Imaging genetics
We conclude with an overview of imaging genetic
studies of PD, and particularly, of how data from
imaging studies can be used to enhance the tractabil-
ity of genetic targets. Brain imaging, and particularly
functional imaging, can provide invaluable insights to
genetic studies, as it allows identification of specific
neural mechanisms that may mediate the effects of
genes on psychiatric disorders. Additionally, because
variation in brain structure or function may be more
objectively indexed to the underlying genetic architec-
ture than behavioral outcomes, neural phenotypes
may provide more penetrant targets through which
to identify genes related to clinical psychopathology.

Nonmasked Masked fear
fear (F-N) (FN-NN) vs. STAI-T
T processing. Twenty PD patients were scanned while
value
6
5
5
4
4 sions, and their DNA was genotyped on the val158met
3 3
2 2
1 1
0 0
Figure 9.6 Amygdala activity related to masked and unmasked
fear processing in healthy subjects. Enlarged views of the right
amygdala illustrating: (1) the dorsal amygdalar cluster from the
nonmasked fear (F-N) comparison (coronal view at Y ¼ 8 [A] and
axial view at Z ¼ 16 [B]); and (2) the basolateral amygdalar cluster
from the correlation of masked fear-induced activity (FN-NN) with
trait anxiety (coronal view at Y ¼ 8 [C] and axial view at Z ¼ 28
[D]). The color bar indicates the significance, with lighter colors
indicating a greater difference between the respective fearful and
neutral conditions. (From [200], with permission from Elsevier.)
See plate section for color version.
In other words, imaging can help both in identifying
genetic risk and in exploring the mechanisms of that
genetic risk [201]. Such approaches have already been
effectively applied to healthy subjects. In an elegant
and pioneering study, Hariri et al. [202] found that
individuals with one or more short alleles of the
serotonin transporter had greater amygdala activation
in response to viewing of fearful faces. Following
up on these results, the same group [203] found that
the regulatory circuit between the amygdala and the
cingulate cortex was uncoupled among individuals
with the short allele of the serotonin transporter.
Furthermore, this uncoupling accounted for a third
of variation in temperamental anxiety, suggesting that
anxiety disorders may be an extreme form of this
decoupling. Subsequently, variation in cortical
Chapter 9: Panic disorder
connectivity resulting from the low functioning vari-
ant of the MAOA gene has been found to predict
propensity to violent behaviors, harm avoidance,
and reward dependence [204–206], demonstrating
that genetic predispositions can indeed impact behav-
ioral outcomes via changes in brain functioning.
But can such models predict psychopathology?
The first study [81] to apply such an approach to
PD focused on the COMT gene, and in particular
the aforementioned “val158met” polymorphism. This
polymorphism is relevant not only because it func-
tionally reduces COMT levels [206] but it also reduces
cortical activity during emotional processing tasks
among healthy volunteers [207, 208]. The investiga-
tors tested whether the association between COMT
and PD was mediated via these changes in emotional
being presented with a series of standardized faces
[209] bearing fearful, angry, happy, or neutral expres-
polymorphism. The investigators reported that
presentation of fearful faces resulted in an increased
activation in the right amygdala as well as the left
orbitofrontal cortex – but this was only observed
among patients carrying at least one val encoding
allele; patients with both copies of the met encoding
allele did not show these changes (Figure 9.7). Angry
or happy faces activated the ventromedial cortex but
not either the amygdala or the orbitofrontal cortex.
The authors speculate that the excessive orbitofrontal
activation in their study might reflect a failure to
appropriately inhibit the processing of anxiety related
emotional cues by the amygdala, and that this may be
an outcome of reduced dopamine levels conferred by
the val allele (val allele ! higher COMT levels !
more breakdown of dopamine ! lower cortical dopa-
mine). This explanation parallels data from animal
models showing dopamine loss in the prefrontal
cortex to result in delayed fear extinction [210].
The investigators then extended this strategy to the
serotonergic system, focusing on both the promoter
polymorphism within the transporter (5-HTTLPR) as
well as a polymorphism involving a cytosine to guanine
switch at position 1019 of the serotonin receptor 1A,
that had been previously shown to result in lower
serotonin transmission [211] and had been associated
with anxiety-related traits in healthy volunteers [212].
They found that patients homozygotes for the g alleles
in the 1A receptor had reduced activation in orbito-
frontal, ventromedial prefrontal, and cingulate cortices
105
 Chapter 9: Panic disorder
when viewing unmasked fearful faces, as compared to
patients with at least one c allele (Figure 9.8) [213].
These observed reductions were also largely within
the right hemisphere, consistent with the role of that
hemisphere in processing emotional faces [214]. The
authors postulate that the genetic variation within the
receptor may increase the odds of psychopathology by
impairing processing within specific cortical sites
related to emotion and anxiety, including the orbito-
frontal and cingulate cortices. Even though the samples
in these two studies were relatively small and bereft of a
true control group of persons without PD, the findings
are important as they are the first to submit that that
genetic variation may alter vulnerability to PD via
specific, behaviorally relevant, brain pathways.
Conclusions
There is ample evidence that PD runs in families and
is heritable; yet the search for specific genes under-
lying the disorder has only been partially informative.
A number of factors may contribute to this: Many of
the individual studies have been jeopardized by rather
small sample sizes, and the substantial phenotypic
heterogeneity across studies has further impeded the
106
Figure 9.7 Amygdala activation in
response to fearful faces is modulated by
catechol-O-methyltransferase (COMT)
genotype. Whole brain voxel-wise t-map
(uncorrected, P > 0.001) of a single panic
disorder patient with a genotype 472G/A
(i.e. val-met heterozygous) comparing
activation for the fearful versus the
no-face condition. Note that there is
significantly greater amygdala activation
(center of cross hair) in the fearful
condition. This increase was not observed
among patients homozygous for the met
encoding allele. The transverse plane is
the original one; the coronal and sagittal
planes are planar reconstructions
orthogonal to the original image.
Reader’s right is subject’s right. Emotional
face stimuli and no-face control stimuli
were controlled for dynamics and
luminance. (From [81], with permission
from Cambridge University Press.)
See plate section for color version.
ability to zero in on the most salient candidates.
Another limitation is the lack of GWAS, which
are particularly valuable for candidate gene identifi-
cation as they allow probing of entire genomes at
relatively high resolution without requiring a priori
hypotheses [215]. GWAS are opening doors in genetic
research on major depressive disorder, autism,
schizophrenia, and bipolar disorder; yet, they have
been largely overlooked for most anxiety phenotypes,
including PD.
In the course of this chapter, we touched upon a
number of approaches aimed at refining the panic
phenotype to increase its tractability. We saw that
early onset substantially magnified familial loading
of PD, and may therefore represent a particularly
valuable subtype. Requiring cases with family history
of anxiety may similarly increase the genetic loading
while minimizing sporadic cases (by corollary, con-
trols should be selected to be free not just of personal
but of family history as well). Other fine-tuning could
include restricting to cases with AG, those showing
CO2 hypersensitivity, or those lying within extreme
ranges of quantitative anxiety measures such as neur-
oticism. Finally, operationalization need not be
restricted to the psychiatric domain, as illustrated by
 Chapter 9: Panic disorder

further as gene products regularly cross-talk. For

Figure 9.8 Serotonin 1A receptor genotype moderates fear
processing in panic disorder patients. Random effects statistical
parametric map for the fearful versus neutral faces contrast overlaid
on a three-dimensional canonical Montreal Neurological Institute
brain showing right-lateralized activity differences in the prefrontal
cortex between the two patient groups (5HT1A1019GG versus
CC/CG; p < 0.001, uncorrected). Patients homozygotes for the
g alleles had reduced activation in orbitofrontal, ventromedial
prefrontal, and cingulate cortices when viewing unmasked fearful
faces, as compared to patients with the CG or CC genotype.
(From [213], with permission from Cambridge University Press.)
See plate section for color version.
our observations that transmission was higher among
families in which the panic clustered with other blad-
der and cardiovascular problems.
But these considerations aside, some of the diffi-
culty in uncovering genetic associations may also lie
in the very complexity of the disorder and in the
expectation implicit in many study designs that there
is a direct or linear relationship between gene and
disease. Complete penetrance is rare, and anxiety
disorders, at least, more likely arise from complex
interplay between multiple genetic and environmen-
tal variables, each with only a marginal role [201,
216]. Any given genetic variation is unlikely to be
either specific or sufficient, and in the case of neuro-
transmitter-encoding genes, specificity is muddied
example, successful treatment of PD patients with
the SSRI fluoxetine, increases plasma levels not only
of serotonin metabolites, but of NE too [154]. Finally,
the detection of certain genetic effects may
be conditional on other genes or environmental
factors, as so elegantly exemplified by the work
of Caspi and colleagues [44] who found a dose-
dependent association between the serotonin trans-
porter polymorphism 5-HTTLPR and depression,
but only among offspring subjected to stressful life
events. There were no main effects, and had the
authors not tested for interaction, they would have
obtained null findings. Yet few of the PD studies
have searched for gene-by-gene, and even fewer,
for gene-by-environment, effects.
Given these layers of complexity, studies relying
on 1 : 1 gene-to-disorder approaches will likely be
of increasingly limited utility. On the one hand, as
is already clear, genetic findings need to be coupled
with epistatic, epigenetic, gene expression, and
translational studies in order to probe the func-
tional consequences of the genetic variation. (Many
of the variants identified thus far lie within introns or
do not alter expression, so their significance remains
unclear.) But the amassing genetic data will also need
to be coupled with other downstream functional
phenotypes (such as the brain structure and function
measures provided by neuroimaging techniques)
that can help shed light on how the genes and their
products translate into the psychiatric outcomes of
interest.
As the field moves toward DSM5, the classifica-
tion of psychiatric disorders also appears to be evolv-
ing toward a more etiological approach integrating
knowledge from genes, environments, and basic
neuroscience [217]. Functional imaging can play an
important guiding role in this evolution: by identify-
ing specific cortical pathways relevant to fear and
anxiety, we can generate more targeted neurobio-
logical markers which can then be used both as a tool
to identify novel genetic variation, and to study the
neurobiological pathways through which genes might
ultimately influence psychopathology.

References IV). Washington, DC: American

1. American Psychiatric Association.
Diagnostic and Statistical Manual of
Mental Disorders, Version IV (DSM-
Psychiatric Association; 1994.
2. Kessler RC, et al. Arch Gen
Psychiatry 2005;62(6):593–602.
3. Kessler RC, et al. Arch Gen
Psychiatry 2005;62(6):617–627.
4. Weissman MM, et al. Arch Gen
Psychiatry 1997;54(4):305–309.

107
 Chapter 9: Panic disorder
5. Kessler RC, et al. Arch Gen
Psychiatry 2006;63(4):415–424.
6. Kessler RC, et al. Arch Gen
Psychiatry 1998;55(9):801–808.
7. Demiryoguran NS, et al. Emerg
Med J 2006;23(2):99–102.
8. Marchesi C, et al. Emerg Med J
2004;21(2):175–179.
9. Marchesi C, et al. Int J Psychiatry
Med 2001;31(3):265–275.
10. Goisman RM, et al. Depress
Anxiety 1998;7(3):105–112.
11. Zimmermann P, et al. Psychol
Med 2003;33(7):1211–1222.
12. Bienvenu OJ, et al. Br J Psychiatry
2006;188:432–438.
13. Fyer AJ, et al. Anxiety 1996;
2(4):173–178.
14. Goldstein RB, et al. Arch Gen
Psychiatry 1994;51(5):383–394.
15. Horwath E, et al. Arch Gen
Psychiatry 1995;52(7):574–582.
16. Maier W, et al. J Psychiatr Res
1995;29(5):375–388.
17. Noyes R, Jr., et al. Arch Gen
Psychiatry 1986;43(3):227–232.
18. Smoller JW, et al. Am J Med
Genet C Semin Med Genet
2008;148(2):118–126.
19. Weissman MM, et al. Arch Gen
Psychiatry 1993;50(10):767–780.
20. Battaglia M, et al. J Psychiatr Res
1999;33(1):37–39.
21. Haghighi F, et al. Am J Med Genet
1999;88(2):131–135.
22. Goldstein RB, et al. Arch
Gen Psychiatry 1997;54(3):
271–278.
23. Hettema JM, et al. Am J Psychiatry
2001;158(10):1568–1578.
24. Kendler KS, et al. Psychol Med
1993;23(2):397–406.
25. Perna G, et al. Psychiatry Res
1997;66(1):69–71.
26. Scherrer JF, et al. J Affect Disord
2000;57(1–3):25–35.
27. Skre I, et al. Acta Psychiatr Scand
1993;88(2):85–92.
108
28. Torgersen S. Arch Gen Psychiatry
1983;40(10):1085–1089.
29. Bellodi L, et al. Am J Psychiatry
1998;155(9):1184–1188.
30. Kendler KS, et al. Arch Gen
Psychiatry 2003;60(9):929–937.
31. Hettema JM, et al. Arch
Gen Psychiatry 2005;62(2):
182–189.
32. Crowe RR, et al. Am J Med Genet
2001;105(1):105–109.
33. Gelernter J, et al. Am J Med Genet
2001;105(6):548–557.
34. Fyer AJ, et al. Biol Psychiatry
2006;60(4):388–401.
35. Kaabi B, et al. Am J Hum Genet
2006;78(4):543–553.
36. Knowles JA, et al. Am J Med Genet
1998;81(2):139–147.
37. Logue MW, et al. Am J Med
Genet B Neuropsychiatr Genet
2003;121B(1):95–99.
38. Thorgeirsson TE, et al. Am J Hum
Genet 2003;72(5):1221–1230.
39. Smoller JW, et al. Am J Med Genet
2001;105(2):195–206.
40. Hamilton SP, et al. Proc Natl
Acad Sci U S A 2003;100(5):
2550–2555.
41. Weissman MM, et al. Am J Med
Genet 2000;96(1):24–35.
42. Kent JM, et al. Biol Psychiatry
1998;44(9):812–824.
43. Nutt DJ. J Clin Psychiatry 1998;
59(S8):24–28; discussion 29.
44. Caspi A, et al. Science 2003;
301(5631):386–389.
45. Heils A, et al. J Neural Transm
1997;104(10):1005–1014.
46. Lesch KP, et al. Science 1996;
274(5292):1527–1531.
47. Blaya C, et al. Behav Brain Funct
2007;3:41.
48. Strug LJ, et al. Mol Psychiatry
2010;15(2):166–176.
49. Perna G, et al. Neuropsycho-
pharmacology 2005;30(12):
2230–2235.
50. Perna G, et al. Biol Psychiatry
1995;37(8):528–532.
51. Price WA, et al. Psychosomatics
1988;29(4):433–435.
52. Maron E, et al. Psychiatry Res
2004;132(2):173–181.
53. Olivier B, et al. Eur J Pharmacol
2005;526(1–3):207–217.
54. Klemenhagen KC, et al.
Neuropsychopharmacology
2006;31(1):101–111.
55. Nash JR, et al. Br J Psychiatry
2008;193(3):229–234.
56. Neumeister A, et al. J Neurosci
2004;24(3):589–591.
57. Rothe C, et al. Int J
Neuropsychopharmacol 2004;
7(2):189–192.
58. Huang YY, et al. Int J
Neuropsychopharmacol 2004;
7(4):441–451.
59. Weisstaub NV, et al. Science
2006;313(5786):536–540.
60. Fehr C, et al. Psychiatry Res
2000;97(1):1–10.
61. Maron E, et al. Psychiatr Genet
2005;15(1):17–24.
62. Inada Y, et al. Psychiatry Res
2003;118(1):25–31.
63. Rothe C, et al. Neurosci Lett
2004;363(3):276–279.
64. Yoon HK, et al. J Anxiety
Disord 2008;22(8):
1529–1534.
65. Jenck F, et al. Eur Neuro-
psychopharmacol 1998;8(3):
161–168.
66. Ribases M, et al. J Psychiatr Res
2008;42(1):50–57.
67. Deckert J, et al. Int J
Neuropsychopharmacol 2000;
3(4):321–325.
68. Maron E, et al. Int J
Neuropsychopharmacol 2005;
8(2):261–266.
69. Mizuta N, et al. Psychiatr Genet
2008;18(1):44.
70. Sabol SZ, et al. Hum Genet
1998;103(3):273–279.

71. Deckert J, et al. Hum Mol Genet
1999;8(4):621–624.
72. Samochowiec J, et al. Psychiatry
Res 2004;128(1):21–26.
73. Hamilton SP, et al. Mol Psychiatry
2000;5(5):465–466.
74. Tadic A, et al. Am J Med
Genet B Neuropsychiatr Genet
2003;117B(1):1–6.
75. Fehr C, et al. Prog Neuro-
psychopharmacol Biol Psychiatry
2001;25(5):965–982.
76. Kim W, et al. Prog Neuro-
psychopharmacol Biol Psychiatry
2006;30(8):1413–1418.
77. Maron E, et al. Neurosci Lett
2007;411(3):180–184.
78. Unschuld PG, et al. Am J Med
Genet B Neuropsychiatr Genet
2007;144B(4):424–429.
79. Zhang X, et al. Science 2004;
305(5681):217.
80. Domschke K, et al. Am J Med
Genet B Neuropsychiatr Genet
2007;144B(5):667–673.
81. Domschke K, et al. Int J
Neuropsychopharmacol 2004;
7(2):183–188.
82. Hamilton SP, et al. Biol Psychiatry
2002;51(7):591–601.
83. Rothe C, et al. Neuro-
psychopharmacology 2006;
31(10):2237–2242.
84. Woo JM, et al. J Psychiatr Res
2004;38(4):365–370.
85. Woo JM, et al. Am J Psychiatry
2002;159(10):1785–1787.
86. Ohara K, et al. Psychiatry Res
1998;80(2):145–148.
87. Zintzaras E, et al. Psychiatr Genet
2007;17(5):267–273.
88. Alsobrook JP, et al. Am J Med
Genet 2002;114(1):116–120.
89. Lee SG, et al. Hum Genet 2005;
116(4):319–328.
90. Bourdeau V, et al. Mol Endocrinol
2004;18(6):1411–1427.
91. Liehr JG, et al. Free Radic Biol Med
1990;8(4):415–423.
92. Freitag CM, et al. Psychiatr Genet
2006;16(2):59–65.
93. Gratacos M, et al. Genes Brain
Behav 2007;6(S1):2–23.
94. Hamilton SP, et al. Am J Med
Genet 2000;96(3):324–330.
95. Sebastiao AM, et al. Trends
Pharmacol Sci 2000;21(9):341–346.
96. Charney DS, et al. Arch Gen
Psychiatry 1985;42(3):233–243.
97. Moreau JL, et al. Brain Res Brain
Res Rev 1999;31(1):65–82.
98. Deckert J, et al. Mol Psychiatry
1998;3(1):81–85.
99. Hamilton SP, et al. Neuro-
psychopharmacology 2004;
29(3):558–565.
100. Domschke K, et al. Am J Med
Genet B Neuropsychiatr Genet
2008;147B(4):510–516.
101. Hodges LM, et al. Am J Med
Genet B Neuropsychiatr Genet
2008;147B(8):1476–1480.
102. Charney DS. Acta Psychiatr Scand
Suppl 2003(417):38–50.
103. Hamilton SP, et al. Am J Med
Genet B Neuropsychiatr Genet
2004;126B(1):111–115.
104. Ebihara M, et al. Am J Med
Genet B Neuropsychiatr Genet
2003;118B(1):32–35.
105. Hattori E, et al. Mol Psychiatry
2001;6(4):465–470.
106. Hosing VG, et al. J Neural Transm
Suppl 2004(68):147–156.
107. Ise K, et al. Am J Med
Genet B Neuropsychiatr Genet
2003;118B(1):29–31.
108. Kennedy JL, et al. Mol Psychiatry
1999;4(3):284–285.
109. Miyasaka K, et al. Am J Med
Genet B Neuropsychiatr Genet
2004;127B(1):78–80.
110. Hamilton SP, et al. Mol Psychiatry
2001;6(1):59–65.
111. Hattori E, et al. Am J Med Genet
2001;105(8):779–782.
112. Kato T, et al. Am J Med Genet
1996;67(4):401–405.
Chapter 9: Panic disorder
113. Keck ME, et al. Am J Med Genet B
Neuropsychiatr Genet 2008;
147B(7):1196–1204.
114. Griebel G, et al. Curr Drug
Targets CNS Neurol Disord 2003;
2(3):191–200.
115. Feusner J, et al. Psychiatry Res
2001;104(2):109–117.
116. Kaiya H, et al. Psychiatry Res
1982;6(3):335–343.
117. Smoller JW, et al. Am J Med Genet
2001;105(3):226–235.
118. Hettema JM, et al. Mol Psychiatry
2006;11(8):752–762.
119. Jedema HP, et al. Eur J Neurosci
2008;27(9):2433–2443.
120. Leygraf A, et al. J Neural Transm
2006;113(12):1921–1925.
121. Hohoff C, et al. J Neural Transm
2009;116(111):1523–1528.
122. Klein DF. Arch Gen Psychiatry
1993;50(4):306–317.
123. Coryell W, et al. J Affect Disord
2006;92(1):63–70.
124. van Beek N, et al. Biol Psychiatry
2000;47(9):830–835.
125. Battaglia M, et al. Am J Med
Genet B Neuropsychiatr Genet
2008;147B(5):586–593.
126. Cavallini MC, et al. Biol Psychiatry
1999;46(6):815–820.
127. Bradley SR, et al. Nat Neurosci
2002;5(5):401–402.
128. Severson CA, et al. Nat Neurosci
2003;6(11):1139–1140.
129. Schmidt NB, et al. J Abnorm
Psychol 2000;109(2):308–320.
130. Perna G, et al. Am J Med Genet B
Neuropsychiatr Genet 2004;
129B(1):41–43.
131. Philibert RA, et al. Am J Med
Genet B Neuropsychiatr Genet
2003;117B(1):11–17.
132. Battaglia M. Mol Psychiatry 2002;
7(3):239–246.
133. Chanock SJ, et al. Nature 2007;
447(7145):655–660.
134. Weissman MM, et al. Arch Gen
Psychiatry 2000;57(7):675–682.
109
 Chapter 9: Panic disorder
135. Maxwell ME. Manual for the
FIGS. Clinical Neurogenetics
Branch, Intramural Research
Program. National Institute of
Mental Health 1992.
136. Talati A, et al. Mol Psychiatry
2008;13(2):122–130.
137. Fyer AJ, et al. Am J Med Genet
1999;88(2):173–181.
138. Weissman MM, et al. Arch Gen
Psychiatry 2004;61(3):273–279.
139. Hamilton SP, et al. PNAS 2003;
5(100):2550–2555.
140. Gorman JM, et al. Psychosom Med
1988;50(2):114–122.
141. Lydiard RB, et al. Am J Psychiatry
1994;151(1):64–70.
142. Placidi GP, et al. Neuropsychobiology
1998;38(4):222–225.
143. Stewart W, et al. Neurology 1994;
44(10/S7):S23–S27.
144. Hanno PM. Urol Clin North Am
1994;21(1):63–66.
145. Curhan GC, et al. J Urol 1999;
161(2):549–552.
146. Buffington CA. J Urol 2004;
172(4 Pt 1):1242–1248.
147. Alagiri M, et al. Urology 1997;
49(5A Suppl):52–57.
148. Warren JW, et al. Urology 2001;
57(6 Suppl 1):22–25.
149. Zondervan KT, et al. Behav Genet
2005;35(2):177–188.
150. Talati A, et al. Biol Psychiatry
2008;63(6):594–601.
151. Chrousos GP, et al. JAMA
1992;267(9):1244–1252.
152. Gorman JM, et al. Am J Psychiatry
2000;157(4):493–505.
153. Sullivan GM, et al. Biol Psychiatry
1999;46(9):1205–1218.
154. Coplan JD, et al. Arch Gen
Psychiatry 1997;54(7):643–648.
155. Blok BF, et al. Brain 1998;121
(Pt 11):2033–2042.
156. Gosling JA, et al. J Urol 1977;
118(2):302–305.
157. Hohenfellner M, et al. J Urol
1992;147(3):587–591.
110
158. Stein PC, et al. Urology 1999;
53(6):1140–1143.
159. Westropp JL, et al. J Urol 2003;
170(6 Pt 1):2494–2497.
160. Pasternac A, et al. Am J Med
1982;73(6):783–790.
161. Gorman RC, et al. J Thorac
Cardiovasc Surg 1995;109(4):
684–693.
162. Coplan JD, et al. Am J Psychiatry
1992;149(11):1587–1588.
163. Zigelman M, et al. Headache
1994;34(10):569–577.
164. Welch KM. Semin Neurol 1997;
17(4):335–341.
165. Goddard E, et al. Primary
Psychiatry 2007;14(4):69–73.
166. Drossman DA, et al. Gut 1999;
45 S2:1125–1130.
167. Agargun MY, et al. Can J
Psychiatry 2004;49(11):776–778.
168. Bajwa WK, et al. Am J Psychiatry
1992;149(3):376–378.
169. Wexler NS, et al. Nature 1987;
326(6109):194–197.
170. Swift RG, et al. Mol Psychiatry
1998;3(1):86–91.
171. Herdy GV, et al. Braz J Med Biol
Res 1992;25(8):789–794.
172. Swedo SE, et al. Am J Psychiatry
1997;154(1):110–112.
173. Frank ECG, et al. CNS Spectrums
1998;3(4):23–34.
174. Mahtani MM, et al. Nat Genet
1996;14(1):90–94.
175. Asami T, et al. Psychiatry Clin
Neurosci 2008;62(3):322–330.
176. Uchida RR, et al. Psychiatry Res
2008;163(1):21–29.
177. Massana G, et al. Neuroimage
2003;19(1):80–90.
178. Uchida RR, et al. Braz J Med Biol
Res 2003;36(7):925–929.
179. Yoo HK, et al. Eur J Neurosci
2005;22(8):2089–2094.
180. Massana G, et al. Am J Psychiatry
2003;160(3):566–568.
181. Protopopescu X, et al. Neuroreport
2006;17(4):361–363.
182. Fontaine R, et al. Biol Psychiatry
1990;27(3):304–310.
183. Ontiveros A, et al. J
Neuropsychiatry Clin Neurosci
1989;1(4):404–408.
184. Sakai Y, et al. Neuroreport 2005;
16(9):927–931.
185. Cameron OG, et al. Arch
Gen Psychiatry 2007;64(7):
793–800.
186. Coplan JD, et al. Biol Psychiatry
1998;44(12):1264–1276.
187. Sakai Y, et al. Neuroimage 2006;
33(1):218–226.
188. Pfleiderer B, et al. World J Biol
Psychiatry 2007;8(4):269–272.
189. Fischer H, et al. Neurosci Lett
1998;251(2):137–140.
190. Reiman EM. J Clin Psychiatry
1997;58(S16):4–12.
191. Javanmard M, et al. Biol
Psychiatry 1999;45(7):872–882.
192. Graeff FG. Neurosci Biobehav Rev
2004;28(3):239–259.
193. Perna G, et al. Psychiatry Res
1994;52(2):159–171.
194. LeDoux JE. Annu Rev Neurosci
2000;23:155–184.
195. Bush G, et al. Trends Cogn Sci
2000;4(6):215–222.
196. Etkin A, et al. Neuron 2006;51(6):
871–882.
197. Etkin A, et al. Am J Psychiatry
2007;164(10):1476–1488.
198. Pillay SS, et al. J Affect Disord
2006;94(1–3):173–181.
199. Pillay SS, et al. J Anxiety Disord
2007;21(3):381–393.
200. Etkin A, et al. Neuron 2004;
44(6):1043–1055.
201. Meyer-Lindenberg A, et al.
Nat Rev Neurosci 2006;7(10):
818–827.
202. Hariri AR, et al. Science 2002;
297(5580):400–403.
203. Pezawas L, et al. Nat Neurosci
2005;8(6):828–834.
204. Buckholtz JW, et al. Mol
Psychiatry 2008;13(3):313–324.

205. Meyer-Lindenberg A, et al. Proc
Natl Acad Sci U S A 2006;
103(16):6269–6274.
206. Chen J, et al. Am J Hum Genet
2004;75(5):807–821.
207. Egan MF, et al. Proc Natl Acad Sci
U S A 2001;98(12):6917–6922.
208. Smolka MN, et al. J Neurosci
2005;25(4):836–842.
209. Ekman PFW. Pictures of
Facial Affect. Palo Alto,
CA, Consulting Psychologists,
1976.
210. Morrow BA, et al.
Neuroscience 1999;92(2):
553–564.
211. Lemonde S, et al.
J Neurosci 2003;23(25):
8788–8799.
212. Strobel A, et al. J Neural
Transm 2003;110(12):
1445–1453.
Chapter 9: Panic disorder
213. Domschke K, et al. Int J
Neuropsychopharmacol 2006;
9(3):349–355.
214. Schmitt JJ, et al. Cortex 1997;
33(1):65–81.
215. Pearson TA, et al. Jama 2008;
299(11):1335–1344.
216. Risch N. Genet Epidemiol 1990;
7(1):3–16; discussion 17–45.
217. Watson D. J Abnorm Psychol
2005;114(4):522–536.
111
 The genetics of the phobic disorders
Chapter
10 and generalized anxiety disorder
Raymond R. Crowe
With the phobic disorders genetics returns to its
origin because Mendel suffered from a phobia of
disease [1]. His symptoms would satisfy modern diag-
nostic criteria for a phobic disorder because their
severity made it impossible for him to attend to ill
or dying parishioners. It was for that reason that he
could not be assigned to a parish church so he was
assigned instead to the monastery at Altbrun where
he did his remarkable work on heredity. With his
broad-ranging interests and mathematical mind, no
doubt he would be fascinated to learn of modern
genetic analyses of phobias: that the magnitude of
genetic and environmental effects can be separated
and measured; that environmental influences can be
separated into family environment and nonfamilial
perturbations; that some of the genetic liability is
shared with other anxiety disorders; and most of all
that attempts are being made to find the genes
responsible for phobias.
Diagnosis and classification
The phobias fall into three major groups with sub-
types within two of these groups. Agoraphobia is
characterized by phobic anxiety over being in situ-
ations from which escape might be difficult or embar-
rassing or in which help might not be available; for
example, sitting in large audiences. In order to be
diagnosed as agoraphobia the feared situations must
be either avoided or endured with marked distress.
Agoraphobia is further classified as occurring with or
without panic disorder. Social phobia is characterized
by a fear of embarrassing oneself in social or perform-
ance situations. For example, persons with social
phobia might be unable to make presentations before
groups for fear of fainting. The diagnosis of social
phobia requires that the feared situations be avoided
Principles of Psychiatric Genetics, eds John I. Nurnberger, Jr. and Wade H. Berrettini. Published by Cambridge University
Press. © Cambridge University Press 2012.
112
or endured with marked distress; in addition, they
must predictably provoke anxiety, the phobic individ-
ual must recognize that the fear is excessive and
unreasonable (in adults), and the avoidance or attend-
ant anxiety must interfere with some realm of life.
Social phobia is further subclassified as generalized
social phobia when the fear and anxiety is not limited
to a few situations but is caused by many social
encounters. Specific phobia, previously termed “simple
phobia”, is characterized by persistent fear of discrete
objects and situations. Animal phobias are examples
of feared objects, and heights, storms, and medical
encounters are examples of feared situations. The diag-
nosis requires that, in addition to the phobic fear,
the object or situation predictably elicit the feared
response, the person recognize that the fear is excessive
and unreasonable, that the stimulus be avoided or
endured with marked distress, and that the avoidance
or phobic anxiety interfere with some realm of the
person’s life. Specific phobia is further subclassified,
depending on the feared stimulus, as animal type,
natural environment type, blood-injection-injury type,
situational type, and other type (not fitting any of the
previous subtypes).
Epidemiology
Unreasonable fears and phobias are quite prevalent in
the general population. A representative population
survey using DSM-III-R criteria (Table 10.1) found
the lifetime prevalence of social phobia to be 13.3%,
specific phobia 11.3%, and agoraphobia 5.3%, with
females affected more frequently in each case [2].
These findings have now been updated with DSM-IV
criteria and Table 10.1 shows that the results for
the combined-sex totals are essentially unchanged
for social and specific phobia but the prevalence of
 Chapter 10: Genetics of the phobic disorders and GAD
Table 10.1 Lifetime prevalence and age at onset of phobic
disorders.
Phobia DSM-III-R DSM-IV
Male Female Total
Agoraphobia 3.5 7.0 5.3 1.4
w/o panic
Specific 6.7 15.7 11.3 12.5
phobia
Social phobia 11.1 15.5 13.3 12.1
Prevalences are expressed in percentages.
Data from Kessler et al. [2, 3].
agoraphobia was only 1.4% [2, 3]. Phobic disorders
develop early in life; the same survey found the
median age of onset to be 7 years for specific phobia,
13 years for social phobia, and 20 years for agorapho-
bia. Among specific phobias, animal phobia develops
at an earlier age than situational phobias [4, 5]. Table
10.2 shows the relationship of irrational fears to phobic
disorders in a population-based sample of twins [6, 7].
In both sexes irrational fears are substantially more
prevalent than the corresponding phobia, the ratio
being more pronounced in social phobia and the spe-
cific phobias. Table 10.2 also presents the prevalence
of the major subclasses of specific phobia, which were
not included in the survey presented in Table 10.1.
The prevalence of phobias is greater in females (with
the possible exception of illness-injury phobia), but
the prevalence of irrational fears is similar in both
sexes (with the possible exception of agoraphobia).
Comorbidity with multiple phobias is common; the
tetrachoreic correlation among phobias within indi-
viduals ranges from 0.29 to 0.54 [4, 5]. The reliability
and long-term stability of the diagnoses of phobias
was examined with repeated interviews after one
month and again after eight years [6]. Kappas ranged
from 0.48 to 0.52 for all phobias but blood-injury, for
which the kappa was 0.29. Long-term stability was
even lower, with kappas ranging from 0.24 to 0.38.
These findings speak for the importance of follow-up
interviews in phenotyping the phobic disorders.
Genetic epidemiology
Genetic epidemiology can address a number of ques-
tions of practical importance for a search for genes
for phobic disorders. The fundamental question is
whether additive genes predispose to phobic disorders.
Table 10.2 Prevalence of fears and phobias in male and
female twins.
Prevalence Male twins Female twins
Fear Phobia Fear Phobia
Agoraphobia 8.6 4.0 12.7 9.1
Animal 28.3 5.2 34.5 10.5
Blood-injury 23.3 5.7 20.2 6.6
Situational 31.2 9.4 30.4 11.3
Social 32.7 6.3 27.5 14.1
Prevalences are expressed as percentages.
Data from Kendler et al. [6, 7].
The magnitude of the additive genetic variance to the
total variance measures the heritability of a trait.
A related question is whether the current classification
of phobias has genetic validity. If all phobias are caused
by the same genetic predisposition but the subclasses are
shaped by different environmental influences, then the
phobias as a group have genetic validity but the sub-
classes do not. This has obvious implications for phe-
notyping subjects for gene searches. Understanding the
genetic basis for sex differences in prevalence rates of the
phobias is important in modeling how genes cause the
phenotype. Since phobias and unreasonable fears are
both highly prevalent, it will be important to know
whether phobias represent the severe end of a con-
tinuum of irrational fears or whether they are separate
disorders. If they are genetically distinct, then phobic
disorders must be diagnosed carefully in order to maxi-
mize the homogeneity of samples used in genetic studies.
The familial transmission of several phobia sub-
types demonstrates that these phobias are familial and
implies that the transmission is specific for each of the
subclasses examined [8]. The findings in Table 10.3
reflect rates of phobias in interviewed relatives of
probands with specific phobia, social phobia, and
agoraphobia. The rates of all three phobias are
increased in the relatives, but only in relatives of
probands with each respective phobia. The magnitude
of the increase is moderate since the relative risks
were 3.4 for specific phobia, 2.5 for social phobia,
and 3.3 for agoraphobia. It might be assumed that
agoraphobia with panic would increase the familial
risk for agoraphobia without panic but no trend in that
direction was found. Familial transmission patterns,
therefore, are consistent with genetic transmission of
these phobias in a subtype-specific manner.
113
 Chapter 10: Genetics of the phobic disorders and GAD

Table 10.3 Familial specificity of phobias.

Specific Social Agoraphobia Unaffected

phobiaa phobia with panic controls
No. of probands 15 39 49 77
No. of interviewed relatives 49 105 131 231
Phobia types (%)
Specifica 31 13 10 9
Social 10 15 8 6
Agora with panic 0 2 10 3
a
Specific phobia was referred to as “simple phobia” in the reference.
Rates in bold are statistically significant at p < 0.05 or less.
From Fyer et al. [8], with permission.

Table 10.4 Genetic and environmental variance in twin by both twins such as family environment, and specific
concordance. environment refers to environmental influences that
Phobia a2 c2 e2 affect one twin and not the other. In extrapolating these
effects from twins to the general population it is easily
Agoraphobia w/o panic seen that the same concepts apply equally to families.
Females 0.39 – 0.61
The principal finding is that additive genes make a
Males 0.37 – 0.63
relatively small, though not an inconsiderable contri-
Social phobia bution to all four phobias, the heritability ranging
Females 0.30 – 0.70 from 20 to 39%. Differences in heritability between
Males 0.20 – 0.80 the sexes are not large, arguing against sex-specific
Animal phobia genetic effects that might be expected, considering
Females 0.32 – 0.68 that all phobia subtypes except the blood-injury group
Males 0.35 – 0.65 are more common among women. Although the
Situational phobia preponderance of the variance was environmental,
Females – 0.27 0.73 no common environmental influence was found, the
Males 0.25 – 0.75 only exception being in women with situational
a2 ¼ additive genetic variance; c2 ¼ common environmental phobia where all of the variance was environmental
variance; e2 ¼ specific environmental variance. The variance with a minor proportion of it belonging to common
estimates are from the best fitting model from each analysis. environment. The authors were skeptical of the find-
The data on males are for irrational fears plus phobias; figures
on females are for phobias. From Kendler et al. [4, 7], with ing, and indeed, the profile of male twins resembles
permission. that of the other three phobias. As noted above, the
phobias have rather modest diagnostic reliability,
and this has the effect of underestimating the genetic
variance and overestimating the environmental vari-
Twin studies are able to separate genetic and envir- ance. In this regard, it is reassuring that when test–retest
onmental components of transmission and quantitate reliability was maximized with repeated interviews,
the magnitude of each liability. Table 10.4 presents the the genetic variances were found to be substantially
results of a population-based study of male and female greater: agoraphobia 0.67, animal phobia 0.47, situ-
twins [4, 5, 7]. The variance of each subclass of phobic ational phobia 0.46, social phobia 0.51 [6].
disorder is partitioned into additive genes, common When the phobic themes of blood, needles, hos-
environment and individual-specific environment in pitals, and disease were examined separately, irrational
male and female twins separately. Common environ- fear or phobia of the first three themes had additive
ment refers to environmental influences that are shared genetic influences accounting for 34–56% of the

114
 Chapter 10: Genetics of the phobic disorders and GAD
variance in concordance, with unshared environment
accounting for the remainder [9]. Fear and phobia of
disease, however, had no genetic component, the vari-
ance being fully accounted for by specific (70%) and
common environment (30%).
Do irrational fears lie on a genetic continuum
with phobias or are they genetically distinct dis-
orders? This question was examined by applying the
multiple threshold model to twin data; the broad
threshold included fears and phobias and the narrow
one only phobias [6]. The phobic disorders examined
in the analysis included agoraphobia, social, situ-
ational, animal, and blood-injury phobias. Multiple
threshold models fit each of the individual phobias,
consistent with the conclusion that irrational fears
and phobias do indeed lie on a genetic continuum.
Similar to the relationship of fears to phobias,
the greater prevalence of phobias in women might
be explained by two genetic hypotheses. Women
might have a lower threshold for manifestation of
the phenotype; or alternately, the same liability genes
could differ between the sexes. Here again, the two
hypotheses can be tested with the multiple threshold
model, assuming that women have a broader thresh-
old than men to account for their greater prevalence.
For agoraphobia, situational, and blood-injury
phobias the findings indicate similar heritabilities
between the sexes but some differences in the liability
genes [10]. Animal phobias have no differences in
either parameter, implying the same heritability and
the same genes for men and women. In this study, the
best-fitting model estimated a heritability in women
for social phobia of zero. The investigators had previ-
ously found the heritabilities of men and women to be
similar (for example, see Table 10.4) and suggested
that the discrepant outcome may have resulted from
stochastic variation in the data.
Summarizing, the phobias have an additive gen-
etic liability of small to moderate magnitude and a
moderate-to-large specific environmental compon-
ent but no common environmental influences. Do
all of the phobias share the same genetic liability
with the type of phobia being determined by the
environment; alternatively, is the genetic liability
for each phobia specific, or does the truth lie some-
where in between? This question can be broadened
to ask whether the phobic disorders as a group have
specific “phobia” genes or whether some of the
liability genes are shared with other anxiety dis-
orders, and even with nonanxiety disorders.
With regard to genes shared among the phobias,
twins can be used to partition the variance into com-
ponents that are shared by two or more disorders and
unshared (i.e. specific to each disorder). Likewise, the
specific environmental variance can be partitioned
into shared and unshared components. (N.B. the
terms shared and unshared are used in place of the
terms “common” and “specific”, used in the publica-
tions cited to avoid confusion with common and
specific environmental variance as those terms are
used elsewhere in the chapter.) The results of such
an analysis are presented in Table 10.5 [7]. Shared
additive genes contributed 5–35% of the total variance
for four phobias; animal phobias had the greatest
proportion of shared genes (21–35%); and social
phobia the least (5–10%). Unshared genetic influences
ranged from 2 to 29%, with social, animal and situ-
ational phobias having the largest proportion at
13–18%. Specific environment accounted for most
of the residual variance, which was also composed of
shared and unshared components. The question that
is likely of most interest to psychiatric geneticists
is what proportion of the total genetic variance is
accounted for by shared versus unshared genes. Aver-
aging the sexes in Table 10.5 and expressing shared
genetic variance as a percent gives the following
proportions: agoraphobia 52%, social phobia 30%,
animal phobia 79% (with a large difference between
the sexes), and situational phobia 36%.
The anxiety disorders could also share common
genes as an explanation for the observed comorbidity
among these conditions, though there are other
explanations. For six anxiety disorders (generalized
anxiety disorder, panic disorder, agoraphobia, social,
animal, and situational phobias) the odds ratios for
comorbidity between all possible pairs of disorders
ranged from 1.5 to 6.3 [11]. To what extent could this
comorbidity be due to shared genes? The best fitting
genetic model in twins, shown in Table 10.6, found
two additive genetic factors contributing to the six
disorders: one factor loaded on generalized anxiety
disorder, panic disorder, and agoraphobia; the other
loaded on the animal and situational phobias, leaving
social phobia intermediate between the two [11].
Although each shared genetic factor accounted for
only 20–27% of the total variance (genetic plus envir-
onmental) of the disorders on which they loaded,
together they accounted for all of the genetic variance
for each phobia, with the exception of agoraphobia
where they accounted for 58% (with unshared genetic
115
 Chapter 10: Genetics of the phobic disorders and GAD

Table 10.5 Proportion of variance.

Phobia Genetic variance Environmental variance

Specific environment Common

environment
Shared Unshared Shared Unshared
Agoraphobia
Female 0.07 0.29 0.64 0.00 0.00
Male 0.11 0.02 0.25 0.39 0.23
Social
Female 0.10 0.21 0.32 0.36 0.00
Male 0.05 0.13 0.42 0.35 0.05
Animal
Female 0.35 0.00 0.05 0.59 0.00
Male 0.21 0.15 0.15 0.49 0.00
Situational
Female 0.09 0.20 0.17 0.53 0.00
Male 0.13 0.18 0.12 0.57 0.01
In this table the total variance is partitioned into genetic, common environmental, and specific environmental components as in the text
and in Table 10.4. The genetic and specific environmental variance are further partitioned into shared and unshared components. Shared
variance consists of genetic and environmental liabilities that influence two or more disorders. Specific variance is comprised of genetic
and environmental liabilities that influence only that disorder. The terms “shared” and “unshared” in the table replace the terms “common”
and “specific” in the publication cited in order to avoid confusion with the way the latter two terms are used elsewhere in the chapter and
in Table 10.4. From Kendler et al. [7], with permission.

Table 10.6 Components of the genetic variance for six anxiety disorders.

Disorder Shared genetic Unshared genetic Total genetic

First Second
General anxiety disorder 20 3 0 23
Panic disorder 27 1 0 28
Agoraphobia 20 1 15 36
Social phobia 8 2 0 10
Animal phobia 1 23 0 24
Situational phobia 1 23 0 24
The total genetic variance can be partitioned into shared and unshared components. Shared genetic variance is common to all disorders
and was further partitioned into two components labeled as the first and second common genetic component. Unshared genetic variance
refers to additive genes that contribute only to that disorder. The terms “shared” and “unshared” in the table replace the terms “common”
and “specific” in the publication cited in order to avoid confusion with the way the latter two terms are used elsewhere in the chapter
and in Table 10.4. From Hettema et al. [11], with permission.

factors explaining the remaining 42%). Environmental
variance was predominantly specific and composed of
shared and unshared sources.
Further evidence for shared liability with the
phobic disorders comes from a study of internalizing
and externalizing disorders in twins [3]. Internalizing
disorders included major depressive disorder, gener-
alized anxiety disorder, and phobic disorder; the
externalizing disorders were substance dependence,
antisocial behavior, and conduct disorder. The results

116
 Chapter 10: Genetics of the phobic disorders and GAD
supported separate additive genetic liabilities for
internalizing and externalizing disorders, and each
genetic factor was shared by all disorders within its
respective group. The shared genetic liability to the
internalizing disorders was further subdivided into
two liabilities: one shared by major depression and
generalized anxiety disorder, the other by the phobic
disorders (panic disorder correlated weakly with the
first genetic liability). The genetic liability of the
internalizing disorders was virtually entirely shared,
panic disorder being the only exception with
unshared genetic liability. The results were summar-
ized by suggesting that the internalizing disorders
could be grouped into those characterized by “anx-
ious misery” (depression and generalized anxiety) and
those characterized “fear” (situational and animal
phobias), with panic disorder not falling cleanly into
either group. Agoraphobia, social, specific, and
blood-injury phobias were not examined and whether
they would also cluster into the “fear” group remains
unresolved.
Social phobia can be further specified as general-
ized or not, although generalized social phobia is not
considered as a separate subclass in DSM-IV. First-
degree relatives of probands with generalized social
phobia have an increased risk for that phobia com-
pared with families of unaffected probands at an
odds ratio of 9.7 [12]. The transmission was limited
to generalized social phobia as the relatives did not
have an increased risk of nongeneralized social
phobia. The rate of avoidant personality disorder
was also increased in the families of generalized social
phobia probands, being diagnosed in 19.8% com-
pared with 0% of controls.
The observed comorbidity between generalized
social phobia and generalized anxiety disorder was
examined by comparing the rates of both disorders
in relatives of probands with each diagnosis, probands
comorbid for both disorders, and unaffected controls
[13]. The rate of generalized social phobia was
increased in family members of probands with gener-
alized social phobia (with or without comorbid gen-
eralized anxiety disorder) at an odds ratios greater
than 3. These results suggest that generalized social
phobia is familial and its transmission is limited to
generalized social phobia.
Social phobia may overlap to a substantial extent
with avoidant personality disorder. The cardinal fea-
ture of that disorder according to DSM-IV is social
inhibition due to pervasive feelings of inadequacy and
hypersensitivity whereas the key feature of social
phobia is an unreasonable fear of embarrassment.
Are these two conditions unrelated or are they different
manifestations of the same disorder? A study of female
twins from the Norwegian population found that
the same additive genetic component contributed to
both avoidant personality disorder and social phobia,
accounting for 37 and 39% of the variance, respectively
[14]. The remainder of the variance for social phobia
was accounted for by specific environment. These find-
ings suggest that social phobia and avoidant personality
disorder share their genetic liability with a phobic
outcome determined by the environment.
With the genetics of the phobic disorders estab-
lished, studies are beginning to ask more specifically
what constitutes the additive genetic component.
Extraversion and neuroticism are recognized person-
ality traits that are reliably measurable and known
to be genetic. Could high neuroticism and low extra-
version be genetic liabilities to phobic disorders,
particularly social phobia? This hypothesis was tested
in twins by examining correlations between the two
personality traits and individual phobic disorders,
with differences between monozygotic and dizygotic
twin pairs estimating the genetic influences under-
lying the correlations [15]. Each personality trait
and three phobias (agoraphobia, social, and animal
phobia) had genetic predispositions, as evidenced by
greater monozygotic than dizygotic twin concord-
ance. High neuroticism and low extraversion correl-
ated moderately with agoraphobia and social phobia;
but animal phobia was uncorrelated with extraversion
and only weakly correlated with neuroticism. When
correlations between the two personality traits and the
three phobias were included in genetic models the
personality traits accounted entirely for the genetic
liability to agoraphobia and social phobia but not to
animal phobia. These findings would imply that the
genes responsible for extraversion and neuroticism
fully account for the genetic liability to agoraphobia
and social phobia.
Further traits that could predispose to social phobia
are social anxiety, trait anxiety, and personality traits.
These traits were examined in a family study by measur-
ing social anxiety with assessments of fear: negative
evaluation, social performance and social interaction;
trait anxiety: the State-Trait Anxiety Inventory; and per-
sonality: the Tridimensional Personality Questionnaire
[16]. First-degree relatives of probands with general-
ized social phobia from an earlier study [12] had
117
 Chapter 10: Genetics of the phobic disorders and GAD
higher social anxiety, higher trait anxiety, and greater
harm avoidance, compared with relatives of control
probands who were free of social phobia. Since these
measures were strongly intercorrelated, a principal
components analysis was performed and resulted in
a single component that accounted for 82% of the
variance that loaded on all of the measures enumer-
ated above. These findings are correlative and cannot
address whether the traits predisposed to or were
influenced by the phobic disorders, but the analyses
do identify a number of potential endophenotypes
for further genetic studies of social phobia.
Apart from personality traits, another possible
endophenotype is susceptibility to fear conditioning,
which has been shown to be genetic in humans [17].
Fear-relevant images (snakes and spiders) and fear-
irrelevant images (geometric shapes) were paired with
a mild electric shock, and galvinic skin response was
measured as the conditioned response in twins. One
image of each pair (fear-relevant and fear-irrelevant)
was presented without shock, providing a measure
of generalization of the conditioned response to an
unconditioned stimulus. Three phases of condition-
ing were examined: habituation (presentation of
images without shock), acquisition (presentation of
images with shock), and extinction (re-presentation
of images without shock). The conditioning process
was found to be moderately heritable: the heritability
of habituation was 35%, acquisition 35–43%, and
extinction 36%. The heritability of acquisition is similar
to the heritability found in animal phobias, 32 and 35%
respectively (Table 10.4). Genetic modeling implied
different additive genes for the conditioned response
to fear-relevant and fear-irrelevant stimuli. This finding
is consistent with the theory that phobias evolved to
protect against environmental threats.
Genomic studies of phobic disorders
Genome searches have been reported for agorapho-
bia, simple phobia, social phobia, and for a broad
phenotype that includes these three phobias in addi-
tion to other anxiety disorders [18–21]. All four stud-
ies are based on the same sample of families selected
through probands with panic disorder and genotyped
with a 10 cM marker map.
A genomic search of agoraphobia was conducted
in 153 family members of panic disorder probands in
20 American families. Approximately two-thirds of the
61 agoraphobics had panic disorder. The families
118
analyzed for logarithm of odds (LOD) and heterogen-
eity LOD (HLOD) scores, as well as nonparametric
linkage (NPL) scores. The parametric analyses were
conducted with dominant and recessive genetic
models; both a broad (definite and probable diag-
noses) and a narrow (definite diagnoses only) pheno-
type were analyzed. The strongest evidence for linkage
to agoraphobia appeared on chromosome 3 at pos-
ition 167.5, where an NPL score of 2.75 was associated
with a p-level of 0.005. None of the parametric ana-
lyses in this region supported linkage. On inspection
of the individual pedigrees one family generated an
NPL score of 10 and was responsible for the NPL peak
in the full pedigree set.
The same pedigree set was analyzed for linkage
to specific phobia [19]. In these analyses 14 pedigrees
were informative for linkage. Fifty-seven family
members had simple phobia, 66% of which were
phobias of the natural environment, with the remain-
der phobic for animals, blood-injury, situations, and
other conditions. The linkage scan was strongly indi-
cative of linkage to a locus on chromosome 14 at
37 cM. A marker at that locus generated a LOD score
of 3.17 with the narrow-phenotype, dominant model,
and a Zlr score of 3.93, associated with a p-value of
4  10–5. This result is particularly interesting because
the provisionally linked region is homologous to a
region of the mouse genome that has been linked to
an anxiety model. Furthermore, a candidate gene for
anxiety, human somatostatin receptor-1, is located in
that region as well.
Genes for social phobia were also sought in these
pedigrees [20]. In 17 informative pedigrees, the 163
relatives included 56 with social phobia in addition
to 4 probable cases. Parametric analyses of a broad
phenotype (definite and probable social phobia) and
a recessive genetic model resulted in a LOD score
of 2.22 on chromosome 16 at 71.1 cM. A nonpara-
metric linkage analysis at the locus resulted in a
p-value of 0.002. Thus, searches for genes for three
of the phobic disorders in these pedigrees found
separate regions of interest for each disorder with
no overlapping regions among them. This outcome
is consistent with evidence reviewed earlier that
agoraphobia, social phobia, and situational phobias
have different genetic predispositions.
A final analysis of this pedigree set included the
above three phobias (agoraphobia, social, and simple
phobia) and panic disorder as the affected phenotype
[21]. Cluster analysis created two overlaping groups
 Chapter 10: Genetics of the phobic disorders and GAD
by scoring each diagnostic assessment on a three-
point scale (absent, probable, or definite diagnosis).
The phenotype used in the linkage analysis was a
“group of membership” score that was analyzed by
the Haseman–Elston linkage algorithm. The analysis
identified a locus at chromosome 4q31.21–32.3
(D4S413) with nominal, 2-point and multipoint p-
values less than 10–5, reaching the accepted signifi-
cance level for linkage. The genome-wide, empirical
multipoint p-value for linkage was 5.6  10–4. The
region is of interest because it contains the gene for
neuropeptide Y receptor, which has been implicated
in anxiety in rodent models. This analysis was par-
ticularly interesting because of the novel way it com-
bined multiple disorders into an affected phenotype,
thus capitalizing on findings that many of the anxiety
disorders share additive liability genes.
Conclusion
Genetic epidemiology indicates that finding genes for
the phobic disorders will require samples with con-
siderable statistical power since the heritability’s of
these disorders are modest. Thus, it is encouraging
that heritabilities can be increased substantially by
repeated assessments. It will be critical to choose
phenotypes wisely to avoid degrading statistical
power with genetic heterogeneity. A number of the
phobias share additive genes and understanding this
will lead to more informed phenotype definitions.
Well-validated endophenotypes could provide an
index of gene expression that is closer to the genome
than clinical diagnosis. All of these considerations will
require careful balancing since the large samples
needed for genome-wide association studies will place
limits on the amount of time and resources that can
be applied to phenotype specification.
Generalized anxiety disorder
The cardinal feature of generalized anxiety disorder
(GAD) is at least six months of apprehensive expect-
ation, characterized by excessive anxiety and worry;
the apprehension is manifest through such symptoms
of anxiety as restlessness, fatigue, poor concentration,
irritability, muscle tension, and insomnia [22]. This is
a frequent condition, affecting 5.7% of the population
at some time during their lives [3].
Hettema et al. [23] performed a meta-analysis
of two family studies of generalized anxiety disorder
[24, 25] and found the odds ratio for that disorder
among first-degree relatives to be 6.1, compared with
relatives of unaffected probands. The study by Noyes
et al. [24] found a rate of 19.5% in family members of
probands with GAD but the rate was not increased in
families of probands with panic disorder or agora-
phobia, supporting a hypothesis of genetic specificity
for GAD. An additional family study published since
the meta-analysis is important because, in contrast
to the above studies, the families were ascertained
from a population-based sample [26]. The analyses
controlled for potential confounding by comorbid
depression and other anxiety disorders, and found
that the odds ratios for first-degree relatives ranged
from 1.4 to 1.8. These odds ratios are lower than the
estimate of 6.1 from the meta-analysis, as might be
expected since the studies analyzed were from clinic-
based samples.
Twin studies indicate that the familial transmis-
sion of generalized anxiety disorder is due to genes.
A population-based sample of 20 monozygotic and
29 dyzygotic twins with GAD from Norway found
40% of monozygotic twins concordant compared
with 10% of dizygotes [27]. Kendler et al. [4, 5]
studied DSM-III-R diagnoses in a population-based
sample of female twins. Cases with a minimum of
one-month and six-months duration were analyzed
together with a multiple threshold model, the former
being the broad and the latter the narrow threshold.
This analysis found 34% concordance in monozygo-
tic compared with 13% in dyzgotic twins. Genetic
factors accounted for 28% of the variance with spe-
cific environment accounting for the remainder.
Like the phobias, GAD appears to have a definite
but modest genetic predisposition. When cases of
GAD that only occurred during episodes of depres-
sion or panic disorder were excluded, the results
were similar to those found without a diagnostic
hierarchy, indicating that the concordance was
indeed due to GAD and not comorbid conditions.
In a similar analysis of these data, additive genes
contributed 14% of the variance in both males and
females, with specific environment contributing the
remaining 86% [28]. If depression excluded GAD in
hierarchical diagnoses, the best-fitting model esti-
mated the additive genetic variance at 21%, some-
what closer to the 28% found in the Kendler et al.’s
analyses [4, 5].
A twin panel of Vietnam-era veterans provides
additional evidence on the genetics of GAD [29, 30].
119
 Chapter 10: Genetics of the phobic disorders and GAD

Diagnoses were based on DSM-III-R criteria except
that the illness duration was one month (as in DSM-
III) rather than six months (as in DSM-III-R and
DSM-IV). The data were analyzed for GAD and panic
disorder jointly. Additive genes were found to have a
small but appreciable influence, accounting for 38%
of the variance of GAD with unshared environment
contributing the remaining 62%. The additive genetic
component of GAD also accounted for 22% of the total
variance of panic disorder and was the sole additive
genetic factor for panic.
The extent to which the anxiety disorders share
some of their genetic predisposition has been con-
sidered from the perspective of the phobias in that
section [11]. With regard to GAD, additive genes
shared with panic disorder and agoraphobia
accounted for 20% of the total variance of GAD;
indeed, this shared variance component made up
87% of its additive genetic variance. The genetic pre-
disposition to generalized anxiety disorder is also
shared with major depressive disorder as well [4, 5].
GAD in female twins was defined by three durations
of symptoms between one and six months. With each of
the three durations of GAD symptoms the best-fitting
genetic model predicted that the genetic predisposition
was completely shared between the two disorders, and
the remaining variance was predominantly specific
environment.

References 11. Hettema JM, et al. Arch Gen

1. Orel V. Gregor Mendel – The First
Geneticist. Translator: Finn S.
Oxford: Oxford University Press;
1996.
2. Kessler RC, et al. Arch Gen
Psychiat 1994;51:8–19.
3. Kessler RC, et al. Arch Gen
Psychiat 2005;62(6):593–602.
4. Kendler KS, et al. Arch Gen
Psychiatr 1992;49:273–281.
5. Kendler KS, et al. Arch Gen
Psychiatr 1992;49:716–722.
6. Kendler KS, et al. Psychol Med
1999;29:539–553.
7. Kendler KS, et al. Arch Gen
Psychiat 2001;58:257–265.
8. Fyer AJ, et al. Arch Gen Psychiat
1995;52:564–573.
9. Neale MC, et al. Am J Med Genet
1994;54:326–334.
10. Kendler KS, et al. Psychol Med
2002;32:209–217.
Psychiat 2005;62:182–189.
12. Stein MB, et al. Am J Psychiat
1998;155:90–97.
13. Coelho HF, et al. J Affect Dis
2007;100:103–113.
14. Reichborn-Kjennerud T, et al.
Am J Psychiat 2007;164:
1722–1728.
15. Bienvenu OJ, et al. Am J Psychiat
2007;164:1714–1721.
16. Stein MB, et al. Am J Med Genet
2001;105:79–83.
17. Hettema JM, et al. Arch Gen
Psychiat 2003;60:702–708.
18. Gelernter J, et al. Am J Med Genet
2001;105:548–557.
19. Gelernter J, et al. Molecular Genet
2003;8:71–82.
20. Gelernter J, et al. Am J Psychiat
2004;161:59–66.
21. Kaabi B, et al. Am J Hum Genet
2006;78:543–553.
22. American Psychiatric Association.
Diagnostic and Statistical Manual
of Mental Disorders, Fourth
Edition, Text Revision (DSM-IV-
TR). Washington, DC: American
Psychiatric Association; 1994;
pp. 441–456.
23. Hettema JM, et al. Am J Psychiatry
2001;158:1568–1578.
24. Noyes R. Jr., et al. Am J Psychiatry
1987;144:1019–1024.
25. Mendlewicz J, et al. Psych Genet
1993;3:73–78.
26. Newman SC, et al. Psychol Med
2006;36:1275–1281.
27. Skrye OI, et al. Acta Psychiatr
Scand 1993;88:85–92.
28. Hettema JM, et al. J Nerv Ment Dis
2001;189:413–420.
29. Scherrer JF, et al. J Affect Dis
2000;57:25–35.
30. Chantarujikapong SI, et al.
Psychiatry Res 2001;103:
133–145.

120
 Genetic contributions to obsessive–
Chapter
11 compulsive disorder (OCD) and
OCD-related disorders
Dennis L. Murphy, Pablo R. Moya, Jens R. Wendland, and Kiara Timpano
Introduction
Obsessive–compulsive disorder (OCD) is a major,
severe neuropsychiatric disorder, with an estimated
lifetime prevalence based on population surveys con-
ducted in many communities nationally and inter-
nationally of 2–3% with a > 2-fold prevalence than
schizophrenia and an approximately similar prevalence
as bipolar disorder [1].
Onset of OCD can be in the pre-teen years as
young as age three, although most commonly, the
period of greatest risk is from adolescence to early
adulthood; less commonly the disorder may first
occur in later decades. Most patients experience a
chronic course, while some have an episodic cyclic
disorder, with exacerbations and remissions. In either
case, OCD symptoms usually are only partially
responsive to various pharmacological and behavior-
ally based treatments, being persistent and lifelong,
substantially impairing function.
There is now compelling evidence for biological con-
tributions to OCD, primarily based on brain imaging
studies that have defined likely brain circuitry involved in
OCD symptoms plus increasing contributions from gen-
etic studies. Early studies had suggested a role for hered-
ity in OCD. More recently, investigations have indicated
likely contributions of chromosomal risk regions, spe-
cific candidate genes and gene pathway networks to
OCD etiology, as will be reviewed in this chapter.
Family and twin studies of OCD
and OCD-related disorders
Early family studies of OCD were conducted with vary-
ing methodological rigor [2–5]. More recent family
studies, using quantitative diagnostic methods, have
found substantially higher risks for OCD and OC
Principles of Psychiatric Genetics, eds John I. Nurnberger, Jr. and Wade H. Berrettini. Published by Cambridge University
Press. © Cambridge University Press 2012.
symptoms in first-degree relatives of OCD probands
when compared to first-degree relatives or
psychiatrically healthy controls. In one of the first of
these, the age-corrected risk of “broadly defined OCD”
was higher in first-degree relatives of OCD subjects
compared to relatives of psychiatrically normal con-
trols (10% versus 1.9%) [6]. More recently, an increased
risk of definite OCD in first-degree relatives of OCD
probands was substantiated when compared to first-
degree relatives of matched community controls, an
excess in the range of four- to eight-fold [4, 7, 8].
Some segregation analyses of OCD families have
suggested a dominant major gene [9–12]. Validation
with larger samples which focused solely on OCD
diagnosis rather than OCD spectrum symptoms
found results consistent with a complex genetic model
including a possible single major locus, which was
neither dominant nor recessive, in combination with
multiple other contributing loci – suggestive of a mixed
model of inheritance [13, 14].
In a still-limited number of studies, monozygotic
twins have been reported as concordant for obsessive–
compulsive symptoms [15–18]. Although reports con-
trasting the rates of the disorder in monozygotic
versus dizygotic twins with OCD are few in number,
a Japanese series found concordance for obsessive–
compulsive symptoms in 80% of monozygotic twins,
compared to 50% of dizygotic twin pairs [19]. In the
Maudsley Twin Register, the concordance rates in
monozygotic and dizygotic twin pairs were 87% and
47%, respectively, giving a heritability estimate of 80%
[20]. No adoption or separation studies, comparing
the rates of OCD in twins raised together or apart,
have been reported. The most recent reviews of OCD
or OCD symptoms in twins found intermediate esti-
mates of heritability which were higher in pediatric
than adult OCD probands [21, 22].
121
 Chapter 11: Genetic contributions to OCD and related disorders
Formal genetics of OCD
Linkage approaches to the genetics of OCD
The first genome linkage scan for well-diagnosed
OCD in seven extended pedigrees identified one can-
didate region on 9p24 that met criteria for suggestive
significance (logarithm of odds [LOD] 1.97) [23]. In
an attempt to replicate this finding, a second study of
50 OCD families focused on microsatellite markers
spanning the 9p24 candidate region. In support of the
original report, a nonparametric analysis identified
a linkage signal at marker D9S1813 with an nonpara-
metric linkage (NPL) of 2.52 (p ¼ 0.006); this peak
lies within 0.5 cM of the original 9p24 linkage signal
[23, 24]. A follow-up linkage analysis by authors of
the first scan included 121 subjects from 26 families
and reported a maximum NPL score of 2.43 at
chromosome 10p15, considered “suggestive” only [25].
In a second genome-wide linkage study, OCD pro-
bands and their sibling pairs and some family members
(N ¼ 1008 total) were evaluated using a NPL method.
Both Kong and Cox LOD all and Kong and Cox LOD
pair statistics were computed and empirical p-values
for all “significant signals” were evaluated with Merlin,
using 10 000 replicates. Evidence for susceptibility loci
was found on chromosomes 1q, 3q, 6q, 7p, and 15q
[26] and were genotyped at the Center for Inherited
Disease Research (CIDR) using 386 microsatellite
markers with an average spacing of 9 cM. Of interest,
no peak in chromosome 9 exceeded a LOD p-value of
1.0, thus not replicating the two prior studies that had
pointed to a 9p24 region of interest [26]. Rather, the
multipoint nonparametric analyses found several sug-
gestive linkage regions that included chromosome 1 (p-
value ¼ 0.003) and chromosome 3 (p-value ¼ 0.0002).
The highest Kong and Cox LOD score (2.67, p ¼ 0.0002)
was obtained at the marker D3S2398 located at 209 cM on
chromosome 3. None of these linkage peaks exceeded
values needed for more than “suggestive” linkage with
OCD.
Fine mapping of chromosome 3: 3q27–28
The strongest linkage result with the second linkage
study near marker D3S2398 was further assessed by
fine mapping using a combined set of microsatellite
and SNP markers to obtain an average marker density
of 360 kb [26]. This follow-up study narrowed the
linkage region from 37 to 14 cM. When family-based
association analyses were performed using the same
122
probands, a relatively strong association peak was
detected nearby at allele D3S3600 (p ¼ 0.002).
Age of OCD onset as a secondary
phenotype in linkage studies
Age at onset, used as a secondary phenotypic approach
to investigate linkage regions in this second genome
linkage study, led to additional results of interest [26].
The median age of onset in this sample of OCD pro-
bands and families was 7 years, reflecting in part, that a
criterion for entry into this study was onset before age
18. In the analyses using MERLIN, suggestive linkage
for the younger age of onset group (LOD ¼ 3.21; p ¼
0.0001) was found at 1q23–24; a smaller shift in LOD
score peaks was also observed in one of the chromo-
some 3 peaks when the sample was divided into groups
with age of onset before or after 7 years [26]. This was
confirmed when families with two or more young age of
onset OCD relatives were evaluated as an additional
subgroup. When the sample was divided based on the
proband’s gender (78 male proband families; 141 female
proband families), a greater increase in the linkage
signal at 11p15 was observed [27]. After genotyping
additional microsatellite markers, the gender-stratified
analysis using MERLIN revealed a LOD ¼ 3.02 (p ¼
0.0001) in the male group. This region contains plaus-
ible candidate genes, such as tyrosine hydroxylase (TH)
and the dopamine D4 receptor (DRD4).
Hoarding among probands with OCD as
a secondary phenotype in linkage studies
When this sample was stratified based on the presence of
two or more relatives with OCD plus hoarding symp-
toms (74 hoarding families; 145 nonhoarding families), a
suggestive linkage signal was found at 14q31–32 (LOD ¼
2.99; p ¼ 0.0001). Candidate genes in this region include
three serotonin receptor genes (3C, 3D, and 3E),
although only 3C is known to be expressed in brain [28].
Genome-wide association analysis
As with other genetically complex medical disorders,
methodological approaches in psychiatric genetics are
shifting to genome-wide association studies (GWAS),
which make more comprehensive single nucleotide
polymorphism (SNP) -based assessments possible
(e.g. [29, 30]). One GWAS study investigating over
1000 OCD probands and over 1000 controls has been
 Chapter 11: Genetic contributions to OCD and related disorders
partially analyzed and preliminary reports have been
reported in abstract form only [31]. The highest
genome-wide p-values point towards glutamate
system gene involvement [31].
Candidate chromosomal regions
and candidate genes for OCD
Potential candidate genes for OCD based on
functional studies and prior investigations
Numerous gene products seem highly relevant to
neurotransmitter system pathways and developmen-
tal sequences important in OCD, but only relatively
few have been investigated. These include glutamate,
dopamine, serotonin, and other systems, neuro-
trophic factor genes and their affiliated receptors,
and genes indirectly implicated via comorbid dis-
orders or suggested from animal models of OCD
and OCD-related behaviors (such as perseverative
behaviors, over-grooming, and hoarding).
Candidate genes for OCD
SLC1A1 and other glutamate system genes
SLC1A1 encodes the only neuronal glutamate trans-
porter and is an attractive OCD candidate gene for
multiple reasons. As a positional candidate, it is the
most evident brain gene of interest located in the
chromosomal region 9p24, the region identified in
the first genome-wide linkage study of mixed large
and small families with OCD [23]; an adapted version
of the four-generation pedigree of one large family
from this study is reproduced in Figure 11.1. As noted
above, this linkage result was supported by a second,
more directed study of this region [24], but was not
replicated as a candidate chromosomal region in the
subsequent largest whole genome linkage study [26].
Figure 11.1 An extended
multiply-affected family. Obsessive–
compulsive disorder (OCD) -affected
males (shaded squares) and females
(shaded circles); nonaffected family
members shown by non-shaded squares
and circles. (Modified from [23].)
There are five major studies of SLC1A1 in OCD,
all of which reported significant association of the
gene with OCD. Four of these were family-based
studies, primarily using trio samples [32–35]; these
were supported by a fifth large case-control study
(Table 11.1) [36].
The glutamate transporter gene SLC1A1 (also
known as EAAC1), is located 350 kb centromeric to
the linkage peak, 9p24, that was described in the first
genome linkage scan of OCD. Its protein product, the
neuronal transporter gene, seems critically important
based on the central role of glutamate in circuits con-
sistently implicated in OCD. Although SLC1A1 was
not specifically identified in the first genome scans, nor
in a follow-up study of the 9p24 region where SLC1A1
is located [23, 24], and no signal was detected in the
largest genome scan, the series of direct SNP-based
studies finding significant associations with this neur-
onal glutamate transporter gene, and other studies
suggesting additional glutamate system candidates,
such as GRIN-2B and GRIK-2, sustain continued inter-
est in the brain glutamate system in OCD [36–38].
Biological and neurodevelopmental studies also
support the glutamate system as being involved in
the neurocircuitry of OCD, including brain imaging
studies and direct magnetic resonance study (MRS)
-based evaluations of brain glutamate concentrations
in OCD treatment studies [39]. In addition, riluzole,
a glutamate antagonist has been evaluated in several
treatment trials of OCD patients, with positive
results, although replicated, placebo-controlled trials
of this agent have not yet been reported [40–42].
SLC6A4, the serotonin transporter (SERT, 5-
HTT) gene and other serotonin system genes
SLC6A4 was among the very first brain genes associ-
ated with an OCD diagnosis. In particular it is the
insertion/deletion variant (the long L 5-HTTLPR) in
123
 Table 11.1 Compilation of five studies investigating SLC1A1 and obsessive–compulsive disorder (OCD) [36].

chr9: 4400000 4450000 4500000 4550000

13–1518,19
Markers 7 9 11 21
Markers 1 2 3 4 5 6
RefSeq Genes 8 10 12 16,’7 20
SLC1A1
C9orf68
A’nold et al, 2006 Dickel et al, 2006 Stewart et al, 2007 Wendland et al, 2009 Shugart et al, 2009
Polymorphism HG18 position Single locus Haplotype Single locus Haplotype Single locus Haplotype Single locus Haplotype Single locus Haplotype
1 rs3933331 (LCL eQTL) 4,379,941 not tested not tested not tested ns not tested
2 rs10814987 4,473,303 not tested ns not tested not tested 0
3 rs10739062 4,492,848 not tested ns not tested not tested 0 P = .046
4 rs1980943 4,504,246 ns ns not tested not tested 0 P = .046
5 rs10115600 4,509,369 not tested ns not tested not tested 0
6 rs3780415 4,545,573 ns not tested not tested not tested 0
7 rs7858819 (LCL & DLPFC eQTL) 4,549,892 not tested not tested not tested ns P =.0002 not tested
8 rs10974625 4,551,596 not tested not tested ns not tested not tested
9 rs7856209 (syn cSNP) 4,554,432 ns not tested ns not tested PBAT dominant: P = .04
10 rs3780413 4,557,353 not tested not tested ns not tested not tested
11 rs3780412 4,562,480 ns P =.04 ns ns 0
12 rs12682807 4,564,022 not tested not tested ns P =.004 not tested not tested
13 rs2072657 4,566,451 not tested not tested ns P =.004 not tested not tested
14 rs301430 (syn cSNP, LCL & DLPFC eQTL, luciferase) 4,566,680 ns P =.03 P =.03 ns P =.004 ns P =.0002 0
15 rs301979 4,566,851 ns ns P =.03 ns not tested 0
16 rs301434 4,572,082 P = .0007 P = . 003 ns ns ns 0
17 rs301435 4,572,843 P = .0009 not tested ns not tested 0
18 rs3087879 (3*.-UTR) 4,576,808 P = .006 P = . 003 not tested ns P =.02 P =.0002 0 P = .02
19 rs301437 4,577,560 not tested not tested ns not tested not tested
20 rs301440 4,580,305 not tested not tested ns not tested not tested
21 rs301443 4,584,919 not tested rs not tested not tested PBAT dominant: P= .000007 P = .02
 Chapter 11: Genetic contributions to OCD and related disorders

STin2 IIe425Val
SNPs VNTR
rs25531,
67 Alternate
(a) Gene rs25532
2 34 5 8 9
1011
1A 1C 1B 12 14 polyadenylation sites
13
5’

3’

5HTTLPR
(LA,LG,SA,SG)
G56A

Alternative
splicing
34

(b) SERT SNPs: 8 Extracellular

11 12
6 7
1. T4A1 9. P339L TMS
9 1013
2. G56A2 10. L362M 2 14 15 Intracellular
16
3. S214S 11. L383L
COOH
NH2
4. E215K1 12. A419A 1
5. H235H 13. 2X1425V(/L)
6. L255M 14. T439T
7. S293F 15. K605N1 or 5HT transport in transfected cells
1
No response to PKG/p38 MAPK activation
8. G308G 16. P621S1 2
No response to 8BrcGMP
Figure 11.2 Human SERT gene organization, with multiple functional variants. See plate section for color version.

the promoter region which is associated with greater
transporter expression that has been associated with
OCD in case-control and family-based studies [43–
45]. While a series of nonreplications were reported
in different countries and ethnic groups, recent
reviews and a recent meta-analysis found that the
L allele was significantly associated with OCD in
family-based studies [46, 47]. Further interest in
SLC6A4 arose when SNPs in the 5-HTTLPR region
were discovered and found to differentially affect
expression, with the LG variant found to convert the
L allele into the equivalent of an S allele [48]. Prior
studies that did not separate the LG and LA alleles may
have introduced a genotyping error of 1–24%,
depending upon ethnicity [48].
In this study by Hu and colleagues [48], associ-
ations between the LA allele and LALA genotype and
OCD diagnosis were found in a family-based study of
175 trios and also in a replicate case-control study of
169 OCD probands and 253 controls. A following
case-control study did not find significant associ-
ations of the triallelic 5-HTTLPR (5-HTTLPR plus
rs25531) with OCD, instead discovering only a nom-
inal association with rs25531 alone that did not sur-
vive correction for multiple testing [49]. The most
recent study reported another new variant, rs25532,
also in the 5-HTTLPR region (Figure 11.2) and also
affecting luciferase reporter gene expression by 15–
80%, (depending on combinations of variants and
cells chosen to evaluate expression). Association with
OCD of a novel haplotype that included this variant
together with the triallelic 5-HTTLPR plus another
SNP, rs16965628 (which is located in intron1) was
found in a sample of 295 OCD probands and 657
controls [50]. Of great interest, it was the higher
expressing allele at each locus that was associated with
OCD diagnosis. Functional variation in the 3’
untranslated region of SLC6A4 was recently reported,
but has not yet been evaluated in OCD [51]. Thus,
from the very earliest studies, which genotyped only

125
 Chapter 11: Genetic contributions to OCD and related disorders
the 5-HTTLPR alone, to the most recent studies that
investigated rs25531 and rs25532 alone or together,
all supported the nation of increased serotonin trans-
porter functioning with decreased extracellular fluid
serotonin but increased serotonin in vesicles available
to be released as possibly contributing to OCD.
Additionally, re-interest in SLC6A4 was also
stimulated by observation of a rare coding region
functional variant, SERT I425V, found associated
with OCD as a complex phenotype [52], a finding
subsequently replicated in a larger study [53], as dis-
cussed in greater detail below. Functional studies of
this variant found altered regulation of the SERT
protein in cell expression systems. An enhanced basal
serotonin transporter function could not be further
stimulated by nitric oxide precursors, unlike the
common 425I allele; however a nearly two-fold
greater basal expression was found with I425V [54].
A follow-up evaluation of the I425V literature in
OCD cases and controls as well as autism subjects
led to the conclusion that OCD was most consistently
associated with this mutation (p ¼ 0.004, odds ratio
[OR] ¼ 6.54; Fisher’s Exact Test, corrected by family
coefficient of identity), with SLC6A4 I425V found in
1.5% of 530 individuals with OCD and 0.23% of
1300 controls [50]. SLC6A4 I425V has now been
designated “OCD 1” in OMIM (Online Mutations
in Man in Pubmed, http://www.ncbi.nlm.nih.gov/
pubmed), in recognition of its high penetrance and
replicated associations with OCD.
SLC6A4 was initially considered a prime candidate
for investigation in OCD because the only well-valid-
ated drug treatments for OCD are the selective sero-
tonin reuptake inhibitors [43, 45, 55–59]. While
depressive disorders seem to be equally well treated
with tricyclic and MAO-inhibiting antidepressants as
well as selective serotonin reuptake inhibitors (SSRIs),
these other antidepressants do not seem of benefit in
OCD in placebo-controlled trials [43, 45, 55–59].
SERT and SLC6A4 are also associated with other neu-
ropsychiatric disorders and with diverse medical dis-
orders in humans [58], and are likewise associated with
multiple behavioral and neurochemical alterations in
mouse and rat models of serotonin transporter defi-
ciency and over-expression [60–66]. Endophenotypes
or changes related to SLC6A4 genotypes include alter-
ations in amygdala and cortico-amygdala function in
brain imaging studies of healthy humans, human and
mouse anxiety and stress responses, and serotonin
receptor density and function [44, 61, 67–69].
126
Dopamine system candidate genes in OCD
Dopamine D2, D3, and most frequently, the D4
receptor gene plus the dopamine transporter (DAT)
gene, as well as genes in the catabolic pathways for
dopamine such as catechol-O-methyltransferase
(COMT) and other catecholamines, have yielded
some results that have been replicated, although con-
flicting gender-related differences were also found
[70–76]. Of note, a low activity allele of the COMT
gene was found to be associated with OCD in females
and subsequent studies also observed positive associ-
ations, although only in specific subgroups or in
males only [77–81]. Likewise, the monoamine oxidase
isoform A (MAOA) gene which is involved in both
catecholamine and serotonin metabolism, was found
to be associated with OCD in two studies, although
gender differences were again found [82, 83].
Other candidate genes in OCD
Multiple other candidate genes for OCD and OCD-
related disorders have been reported but most thus far
have not been replicated. Several have been fairly
comprehensively evaluated and perhaps the strongest
data suggest that a variant in the neurotrophin gene,
BDNF (e.g. V66M), is strongly associated with OCD
[84]. Other candidates include transcription factor
genes such as Hoxb-8 (grooming [85, 86]; MOG [87,
88]), neuropeptide genes and their receptors (e.g.
BDKRB27 [89]), modulators of genes otherwise
implicated in OCD, and, of course, genes already
associated with neuropsychiatric disorders with
common comorbidity with OCD.
Chromosomal anomalies and rare
gene involvement in OCD
Additional uncommon chromosomal anomalies or
gene variants have been investigated in OCD and
OCD-related or OCD-comorbid disorders, especially
Tourette syndrome (TS) families. As OCD is often
found as a triad together with tic disorders plus atten-
tion-deficit hyperactivity disorder (ADHD) in TS, for
the purposes of this review we will consider chromo-
some regions and genes associated with TS as possible
OCD-related. The two findings with the most
common TS–OCD comorbidity are associated with
the 22q11 microdeletion syndrome [90–93] and OCD
associated with the dystonia myoclonus syndrome
[94–97]. The latter is of note because its 7q21-q31
 Chapter 11: Genetic contributions to OCD and related disorders
region is near that of the chromosomal anomalies
described in cases at 7q31 and 7q35–q36 (reported
to be associated with OCD and TS) [98–100]. Add-
itionally, a family-based association study using
markers in the 7q31 region demonstrated biased
transmission of marker alleles in individuals with
comorbid TS, OCD, and ADHD [101].
OCD and the dystonia myoclonus syndrome
In one study of three extended dystonia myoclonus
syndrome (DMS) families, OCD was present in 25% of
symptomatic DMS carriers with the 7q21 haplotype,
but in only 9% (1/11) of nonsymptomatic carriers and
0% (0/28) of nonhaploytpe carriers [96]. OCD comor-
bid with generalized anxiety disorder and major
depressive disorder was also significantly increased in
these three families (but neither generalized anxiety
disorder nor major depression was increased with
DMS alone) [96]. In another study of three extended
families with DMS, OCD was diagnosed in three
members of one of the families; all three were symp-
tomatic DMS carriers; OCD was not present in any of
the other 10 members of this family nor in 14 members
of another 2 families from which psychiatric profiles
were obtained [94]. Individuals in the family with
DMS and OCD had a 7q21 deletion mutation shown
to truncate the DYT1 (e-Sarcoglycan) locus.
In another single family with 14 members, 3 indi-
viduals with DMS plus OCD (together with diagnoses
of depression) among 6 total had DMS attributed to a
truncating mutation within the SGCE gene [95, 97].
DMS has also been found linked to two other loci
besides DYT1 at 7q21: a 16.9 cM region between
DISS1132 and D183843 on 18p11 [102]. However, a
direct examination of the SGCE gene in 32 TS patients
with OCD using WAVE DHPLC analysis (plus direct
sequencing of 5 of the patients) detected no abnor-
malities in this gene in comparison to 60 CEPH
controls [103]; nor was OCD increased in mutation
carriers of the DYT1 gene among family members of
dystonia probands [104].
Some prior evidence of a dystonia–OCD connection
had been suggested on the basis of an elevated frequency of
OCD symptoms measured by the Maudsley Obsessive–
Compulsive Disorder Scale and the Yale–Brown
Obsessive–Compulsive Scale (YBOCS) in individuals
with two dystonic syndromes, spasmodic torticollis
and blepharospasm [105]. Also, a 14–20% incidence
of OCD has been found in 2 studies of over 100
patients with focal dystonias (relative to the 2.5%
OCD prevalence in the general population) [106,
107]. All of these disorders, including TS, appear to
share alterations in cortico-striato-thalamocortical
(CSTC) pathways, although some differences in
pathway involvement require further elucidation via
ongoing brain imaging, electrophysiological, and
postmortem brain studies [106, 108–110].
OCD and the chromosome 22q deletion
syndrome.
The 22q11.2 contiguous microdeletion syndrome
occurs in approximately 0.3% of live births, generally
in sporadic cases, although a minority of cases
( 10%) may follow an autosomal dominant inherit-
ance pattern. The size of the deletion is 1.5–3.0 mb,
involving approximately 25 genes. Initially recognized
because of facial and cardiac malformations associ-
ated with learning disabilities, and therefore named
the velo-cardio-facial syndrome (VCFS), more exten-
sive evaluation of children and adults with this dele-
tion uncovered diverse psychopathology, ranging
from ADHD, pervasive developmental disorder
(PDD), and anxiety disorders – including OCD – to
schizophrenia and bipolar disorder [111].
In an evaluation of children with VCFS, obses-
sive–compulsive behaviors were reported with the
22q11 deletion syndrome although none in this early
study received an OCD diagnosis [112]. Ascertain-
ment strategies and diagnostic evaluations have been
quite varied in these VCFS studies, and in one of the
few investigations which compared children with
VCFS to children with similar cognitive impairments,
a high, but equal number of behavior problems and
psychiatric disorders were found in the two groups
[90]. For OCD specifically, 11% of the children with
VCFS received an OCD diagnosis, as did 14% of the
cognitive ability-matched control [90].
In more detailed studies, groups of adult and
mixed-age clinic subjects meeting diagnostic criteria
for OCD have been noted in four studies of individ-
uals with 22q11 deletions, although all but one of
these reports were primarily focused on the evalu-
ation of schizophrenia or affective disorders [91–93,
113]. Psychiatric evaluation of 1 cohort of 14 VCSF
patients over age 15 revealed 4 with schizophrenia or
schizo-affective disorder, 2 of whom also had OCD
[93]. A similar cohort of patients with VCFS who
were psychiatrically evaluated found that most of the
127
 Chapter 11: Genetic contributions to OCD and related disorders
patients shared common mood, anxiety, and obses-
sive–compulsive symptoms, but the majority received
bipolar, schizoaffective, and ADHD diagnoses (64%),
while only 8% met OCD criteria [92].
In the only more comprehensive study that used the
YBOCS scale to evaluate OCD symptomatology in psy-
chiatric interviews of a VCSF clinic sample, 33%
received primary OCD diagnoses [91]. The most
common comorbid diagnoses in those with OCD in
this study were ADHD, simple phobia, and social
phobia. The highly varied forms of psychopathology
found in VCFS probands and the large range of 8–
33% occurrence of OCD in the 4 studies with a total
of 120 VCFS probands clearly requires further evalu-
ation. This will be of interest, as the COMT locus
previously nominated as an OCD candidate gene, lies
within the 22q deletion and several studies found that
the V158M COMT variant was associated with OCD
and other disorders, as noted above and as recently
reviewed [78]. Changes in the forebrain metabolism
of dopamine have been shown in individuals with this
functional COMT V158M variant [114]. Individuals
with VCSF and the 22q microdeletion have not yet
been evaluated in similar magnetic resonance imaging
(MRI) studies, or positron emission tomography
(PET) studies used in these studies of COMT variants.
Study of such variants may provide clues about more
common variants nearby. The study of rare variants
represents a complementary approach to the identifi-
cation of disease genes and ultimately drug targets
for disease interventions, as increasingly discussed
[115, 116].
Animal models of OCD-like features
that suggestively implicate specific
genes in behavior
Many models of behavioral changes resembling
OCD-related features such as perseveration, “compul-
sive” grooming, food-restriction-induced compulsive
wheel running, or drinking have been reported [27,
55, 117–121]. However, despite “face validity”, and in
some instances, apparent validity based upon bio-
logical neurocircuitry or drug response similarities
to human OCD, relatively few of these models have
implicated specific genes. Most of the models which
have done so have utilized knockout or transgenic
mouse models or evaluations of behavioral features
in mice of different background strains.
128
Given the current state of the evidence that OCD
is a genetically and phenotypically complex and het-
erogeneous disorder, it is not surprising that some of
the most interesting and persuasive animal models
have focused on specific OCD sub-phenotypes. Sev-
eral recent reviews have been directed to animal
models of OCD features [27, 117–121] and this
review will primarily be directed to the most recent
findings reported.
Over-grooming behavior in mice with
SAPAP3 deficiences, and two initial studies
of SAPAP3 in humans with OCD,
trichotillomania, and grooming disorders
Knockout of the excitatory glutamate-system-related
post-synaptic density gene, Sapap3 was recently
reported to be associated with excessive facial self-
grooming [122]. Facial hair and some skin were
essentially denuded in mice lacking Sapap3, a
DLGAP3-related gene located on human chromo-
some 1p35 that encodes a protein localized to den-
drites and cell bodies and is enriched in synaptosomal
membrane fractions. Rescue of this prominent
phenotype was essentially complete following chronic
treatment with the serotonin reuptake inhibitor (SRI),
fluoxetine, as well as by replacement expression of
SAPAP3 in the striatum using a lentiviral vector.
In follow-up studies in humans, resequencing of
SAPAP3 was carried out in individuals with OCD
alone, with OCD plus trichotillomania (TTM, pri-
marily characterized by self-removal of body hair),
TTM together with a control group [123]. The major
findings were an excess of rare SAPAP3 variants
(Table 11.2), some with estimated likelihood of dam-
aging effects, in the OCD/TTM groups from the two
sites. The incidence of these rare variants in cases was
2.1% (7/330) versus 0.56% (2/356) in controls. As
noted in Figure 11.3b, SAPAP3 SNPs are highly con-
served across mammalian species including the rare
variant SNPS found in greater proportions in patients
than controls. It is noteworthy that conservation
of these SNPs is  97% identical among specific
amino acids.
A second study of SAPAP3 in sibling pairs with
“grooming disorders” found that 32% of 1638 study
participants (65% of whom met lifetime criteria for
OCD) met defined criteria for “grooming disorders”
[124]. Six SNPs, constituting three haplotype blocks,
 Chapter 11: Genetic contributions to OCD and related disorders

Table 11.2 Identified rare variants in SAPAP3 and predicted functional relevance.

Detected nonsynonymous Analyzed samples Prediction algorithms

variants
121 Duke TTM 44 NIMH 48 130 PMuta PolyPhena
and NIMH OCD wlo NIMH Duke
OCD w. TTM TTM controls controls
R13C; c.38C>T 1 0 0 0 Pathological Possibly
damaging
A148insGPAGA 0 1 0 0 NAb NAb
c.441_442insGGGCCAGCAGGGGCA
T156M; c.467C>T 0 0 0 1 Neutral Benign
A189V; c.566C>T 1 1 0 1 Neutral Benign
T523K; c.1569-70CC>AA 1 0 0 0 Pathological Possibly
damaging
P606T; c.1816C>A 0 1 0 0 Neutral Possibly
damaging
K910R; c.2728A>G 1 0 0 0 Neutral Benign
Combined allele frequencies 7/330TTM and OCD (2.1%) 2/356 controls
(0.56%)
a
Pmut: http://mmb2.pcb.ub.es:8080/Pmut; PolyPhen: http://genetics.bwh.harvard.edu/pph.
b
Prediction of the effects of in/dels is not possible with PMut and PolyPhen, but a functional effects is likely for a five amino acid insertion.
Reproduced from [141] with permission from Nature Publishing group.
Abbreviations: NA, not applicable; NIMH, National Institute of Mental Health; OCD, obsessive–compulsive disorder; TTM, trichotillomania.

(a)
A189V

Variations in OCD/ R13C A148insGPAGA T523K P606T K910R

TTm patients

Guanylate-Kinase-associated protein (GKAP) domain

Variations in
T156M A189V
controls subjects

(b) Human MRGYHGDRGSHPRPARFADCQ TLPYQRGPAGAGPGPAPGTGTAPEPRSE SHSLBAPGKRDY OLPLLAAPAAVSGRP VPPRASPKPPT LLEPKEEXKVPPPIP

Rhesus MRGYHGDRGSHPRPARFADCQ TLPYQRGPAGAGPGPAPGTGTAPEPRSE SHSLBAPGKRDY OLPLLATPAAVSGRP VPPRASPKPPT LLEPKEEXKVPPPIP
Cow MRGYHGDRGSHPRPARFADCQ TLPYQRGPAGAGPGPAPGTGTAPEARSE SHSLBAPGKRDY OLPLLAAPASVSGRP VPPRASPKPPT LLEPKEEXKVPPPIP
Horse MRGYHGDRGSHPRPARFADCQ TLPYQRGPAGAGPGPAPGTGTAPEARSE SHSLBAPGKRDY OLPLLAAPASVSGRP VPPRASPKPPT LLEPKEEXKVPPPIP
Mouse MRGYHGDRGSHPRPARFADCQ TLPYQRGPAGFGPGP--GSGAAPEARSE SHSLBAPGKRDY OLPLLAAPASVSGRP VPPRASPKPPT LLEPKEEXKVPPPIP
Rat MRGYHGDRGSHPRPARFADCQ TLPYQRGPAGFGPGP--GSGAAPEARSE SHSLBAPGKRDY OLPLLAAPASVSGRP VPPRASPKPPT LLEPKEEXKVPPPIP

R13C A148insGPAGA T156M A189V T523K P606T K910R

Figure 11.3 Identified rare nonsynonymous polymorphisms in synapse-associated protein 90/postsynaptic density-95-associated protein 3
(SAPAP3). (a) Schematic of SAPAP3, which consists of 10 coding exons (boxes). Seven rare changes were identified in trichotillomania and
obsessive–compulsive disorder (OCD) patients (a), but only two in controls (b). Most mutations fell into exon 1; however, three changes
affected the conserved guanylate-kinase-associated protein (GKAP) domain. (b) SAPAP3 is highly conserved between species ( 97% identical
amino acids between human and mouse). Accordingly the identified rare changes affected conserved residues. See plate section for color
version.

129
 Chapter 11: Genetic contributions to OCD and related disorders
were examined. OCD diagnosis itself was not signifi-
cantly associated with any of the six SNPs or three
haplotypes, but the “grooming disorder” subgroup
showed a nominal association of at least one
“grooming disorder” (p < 0.05) with four of the six
SNPs genotyped and with all three haplotypes [124].
These results were not corrected for multiple testing
and must be considered preliminary. However, rela-
tively few groups studying OCD have been examining
the “grooming disorder” subgroup.
Environmental contributions to OCD,
including traumatic life events and
possible contributions from immune
system challenges such as streptococcal
infection as postulated in the PANDAS
syndrome
Environmental and gene  environment
interactional contributions to OCD
There is some evidence for contributions to OCD
onset, OCD severity, and other features of OCD from
diverse environmental factors, from psychological
trauma, head trauma, and autoimmune reactions
[125–130]. Interpretation on these studies is limited
by study design differences, small subject numbers,
and the paucity of studies examining interactions
between any of these factors and gene variants.
Comments and conclusions
Except for candidate gene studies, OCD has been
neglected as a disorder of interest to geneticists until
recently. Although it is a chronic, only partially treat-
able disorder of higher frequency than schizophrenia
and of similar frequency as bipolar disorder (both of
which have been the object of 20+ genome-wide
scans, recent GWAS reports, and many other studies
of twins, families, and gene-associated neurobio-
logical studies), OCD has only very recently been
the subject of three small genome-wide linkage stud-
ies. One of these was based on only seven families, the
second was primarily a sibling-pair based study, and
the third was an extension of the first genome-wide
linkage study [23, 26, 131]. A first GWAS of OCD
probands has been accomplished, but as of 2009, only
preliminary analyses and results have been reported
in abstract form [31]. However, the 9p24 region (of
130
highest importance in the first genome-wide scan
[23]) was supported by a study using more dense
targeted markers [24]. This region includes SLC1A1,
thus far the most-replicated candidate gene in OCD,
with positive findings in five studies. A single major
functional gene variant has not yet been identified,
although some of the SNPs associated with OCD
show evidence of functional changes in HapMap gene
expression data, in human brain expression data, and
in studies of luciferase-measured gene expression in
cultured cell systems [50].
Several other uncommon, less well quantitated
genetic variations occur with an OCD phenotype,
including chromosomal anomalies and some other
rare gene variants, such as SLC6A4 I425V, SGCE,
GCH1, and SAPAP3. A tentative conclusion is that
OCD resembles other complex disorders in being
etiologically heterogeneous and in having both highly
penetrant familial subtypes associated with rare alleles
or chromosomal anomalies [115, 116, 132], as well as
having relatively more common, polygenic contribu-
tions that may involve polymorphisms in such genes
as SLC1A1, SLC6A4, BDNF, and perhaps COMT,
DRD4, and HTR2A.
OCD seems likely to be genotypically and pheno-
typically heterogeneous. Phenotypic variables such as
compulsive hoarding behavior, age of onset, associ-
ated comorbid disorders, and gender, have shown
influences in genome-wide linkage analyses and other
studies. This has primarily been due to the large-scale
quantitative phenotypical variable assessment built
into the second GWAS of OCD and other studies
from the OCD Collaborative Genetics Study (OCGS)
group [28, 131, 133–137] and also in the pending
GWAS study from another consortium [31].
Compulsive hoarding behavior as a subtype of
OCD or an entity of its own has emerged as a major
target in genetic studies [28, 135, 138, 139]. An earlier
study also found important genome-wide linkage
regions associated with hoarding in a TS cohort
[140]. Compulsive hoarding and related perseverative
behaviors have also been found in other human dis-
orders with repetitive behaviors or obsessions
(autism, schizophrenia, “punding” in Parkinson’s dis-
ease, “obsessive flashbacks” in post-traumatic stress
disorder) as well as in animal models of compulsive
behaviors such as over-grooming [122, 124, 141],
canine acral lick and other behaviors [142], feline
wool chewing, and additional behaviors in other
species [143–145]. There are many potentially fruitful
 Chapter 11: Genetic contributions to OCD and related disorders

network and pathway hypotheses, including those
involving serotonergic, dopaminergic, glutamatergic
neuropeptide, neurodevelopmental, as well as genes
related to frequent OCD–comorbid disorders. We
may most likely anticipate a combination of rare
genes ( 0.5%), together with a more extensive poly-
genic group contributing to the OCD subphenotypes.
Other phenotypic subgroups such as those associated
with comorbid disorders may be partially genetically
based. Among these, gene variants may interact with
traumatic events such as drug exposure in utero,
streptococcal infections, and personal traumatic events.
Acknowledgements
This research was supported by the Intramural
Research Program of the National Institute of Mental
Health (NIMH), National Institutes of Health (NIH).
The authors are grateful to Theresa B. DeGuzman for
her editorial and artwork assistance.

References 17. Mcguffin P, et al. Br J Psychiatry

1. Kessler RC, et al. Arch Gen
Psychiatry 2005;62:593–602.
2. Bellodi L, et al. Psychiatry Res
1992;42:111–120.
3. Lenane MC, et al. J Am Acad Child
Adolesc Psychiatry 1990;
29:407–412.
4. Pauls DL, et al. Am J Psychiatry
1995;152:76–84.
5. Riddle MA, et al. J Am Acad
Child Adolesc Psychiatry
1990;29:45–48.
6. Black DW, et al. Arch Gen
Psychiatry 1992;49:362–368.
7. Hettema JM, et al. Am J
Psychiatry 2001;158:
1568–1578.
8. Nestadt G, et al. Arch Gen
Psychiatry 2000;57:358–363.
9. Cavallini MC, et al. Am J Med
Genet 1999;88:38–43.
10. Hanna GL, et al. Am J Med Genet
B Neuropsychiatr Genet 2005;
134B:13–19.
11. Nestadt G, et al. Am J Hum Genet
2000;67:1611–1616.
12. Nicolini H, et al. Ursus Medicus
1991;1:25–28.
13. Alsobrook JP, et al. Am J Med
Genet B Neuropsychiatr Genet
1999;88:669–675.
14. Eapen V, et al. J Psychosom Res
2006;61:359–364.
15. Cryan EM, et al. Br J Psychiatry
1992;161:694–698.
16. Marks IM, et al. Br J Psychiatry
1969;115:991–998.
1980;137:285–287.
18. Woodruff R, et al. Am J Psychiatry
1964;120:1075–1080.
19. Inouye E. Am J Psychiatry
1965;121:1171–1175.
20. Carey G, et al. In Klein DF et al.
(eds.). Anxiety: New Research and
Changing Concepts. New York:
Raven Press; 1981.
21. Jonnal AH, et al. Am J Med Genet
2000;96:791–796.
22. Van Grootheest DS, et al. Twin
Res Hum Genet 2005;8:450–458.
23. Hanna GL, et al. Am J Med Genet
2002;114:541–552.
24. Willour VL, et al. Am J Hum
Genet 2004;75:508–513.
25. Hanna GL, et al. Biol Psychiatry
2007;62:856–862.
26. Shugart YY, et al. Mol Psychiatry
2006;11:763–770.
27. Wang Y, et al. Am J Hum Genet
2009;84:52–59.
28. Samuels J, et al. Am J Psychiatry
2007;164:493–499.
29. Sklar P, et al. Mol Psychiatry
2008;13:558–569.
30. The Psychiatric GWAS
Consortium Steering (TPGCS)
Committee. Mol Psychiatry
2009;14:10–17.
31. Stewart E, et al. ACNP Meeting
Miami, FL; 2010.
32. Arnold PD, et al. Arch Gen
Psychiatry 2006;63:769–776.
33. Dickel DE, et al. Arch Gen
Psychiatry 2006;63:778–785.
34. Shugart YY, et al. Am J Med Genet
B Neuropsychiatr Genet
2009;150B(6):886–892.
35. Stewart SE, et al. Am J Med Genet
B Neuropsychiatr Genet 2007;
144B:1027–1033.
36. Wendland JR, et al. Arch Gen
Psychiatry 2009;66:408–416.
37. Arnold PD, et al. Psychiatry Res
2009;172:136–139.
38. Sampaio AS, et al. CNS Neurosci
Ther 2011;17(3):141–147.
39. Moore GJ, et al. J Am Acad Child
Adolesc Psychiatry 1998;
37:663–667.
40. Coric V, et al. Biol Psychiatry
2005;58:424–428.
41. Grant P, et al. J Child Adolesc
Psychopharmacol 2007;
17:761–767.
42. Schwenkreis P, et al. Exp Brain Res
2000;135:293–299.
43. Bengel D, et al. Mol Psychiatry
1999;4:463–466.
44. Lesch K-P, et al. Science 1996;
274:1527–1531.
45. Mcdougle CJ, et al. Arch Gen
Psychiatry 1994;51:302–308.
46. Bloch MH, et al. Am J Med Genet
B Neuropsychiatr Genet
2008;147B:850–858.
47. Dickel DE, et al. Biol Psychiatry
2007;61:322–329.
48. Hu XZ, et al. Am J Hum Genet
2006;78:815–826.
49. Wendland JR, et al.
Neuropsychopharmacology
2007;32(12):2543–2551.

131
 Chapter 11: Genetic contributions to OCD and related disorders
50. Wendland JR, et al. Psychiatr
Genet 2008;18:31–39.
51. Vallender EJ, et al. Genes Brain
Behav 2008;7:690–697.
52. Ozaki N, et al. Mol Psychiatry
2003;8(895):933–936.
53. Delorme R, et al. Mol Psychiatry
2005;10(12):1059–1061.
54. Kilic F, et al. Mol Pharmacol
2003;64:440–446.
55. Altemus M, et al. Am J Med Genet
1996;67:409–411.
56. Insel TR, et al. Biol Psychiatry
1985;20:1174–1188.
57. Mcdougle CJ, et al. Mol Psychiatry
1998;3:270–273.
58. Murphy DL, et al. Mol Interv
2004;4:109–123.
59. Torres GE, et al. Mol Pharmacol
2003;64:196–198.
60. Fox MA, et al. Psycho-
pharmacology (Berl)
2008;201:203–218.
61. Holmes A. Neurosci Biobehav Rev
2008;32:1293–1314.
62. Jennings KA, et al. J Neurosci
2006;26:8955–8964.
63. Murphy DL, et al. Neuro-
pharmacology 2008;55:932–960.
64. Murphy DL, et al. Nat
Rev Neurosci 2008;9:
85–96.
65. Homberg J, et al. BMC Genet
2010;11:37.
66. Homberg JR, et al. Neuroscience
2007;146:1662–1676.
67. Greenberg BD, et al. Am J Med
Genet 1999;88:83–87.
68. Hariri AR, et al. Arch Gen
Psychiatry 2005;62:146–152.
69. Lothe A, et al. Neuroimage
2009;47:482–492.
70. Billett EA, et al. Psychiatr Genet
1998;8:163–169.
71. Catalano M, et al. Am J Med Genet
1994;54:253–255.
72. Cruz C, et al. Neurosci Lett
1997;231:1–4.
132
73. Hemmings SM, et al. Eur Neuro-
psychopharmacol 2003;13:
93–98.
74. Millet B, et al. Am J Med Genet
B Neuropsychiatr Genet
2003;116:55–59.
75. Nicolini H, et al. Mol Psychiatry
1996;1:461–465.
76. Walitza S, et al. J Neural Transm
2008;115:1071–1078.
77. Alsobrook JP, 2nd, et al. Am J Med
Genet 2002;114:116–120.
78. Azzam A, et al. Am J Med Genet
B Neuropsychiatr Genet
2003;123:64–69.
79. Erdal ME, et al. Depress Anxiety
2003;18:41–45.
80. Niehaus DJ, et al. J Affect Disord
2001;65:61–65.
81. Schindler KM, et al. Am J Med
Genet 2000;96:721–724.
82. Camarena B, et al. Int J
Neuropsychopharmacol
2001;4:269–272.
83. Karayiorgou M, et al. Biol
Psychiatry 1999;45:
1178–1189.
84. Hall D, et al. Am J Hum Genet
2003;73:370–376.
85. Chen SK, et al. Cell 2010;
141:775–785.
86. Greer JM, et al. Neuron 2002;
33:23–34.
87. Atmaca M, et al. J Affect Disord
2010;123:258–263.
88. Zai G, et al. Am J Med Genet
B Neuropsychiatr Genet
2004;129B:64–68.
89. Gratacos M, et al. Am J Med Genet
B Neuropsychiatr Genet 2009;
150B:808–816.
90. Feinstein C, et al. Biol Psychiatry
2002;51:312–318.
91. Gothelf D, et al. Am J Med Genet
B Neuropsychiatr Genet 2004;
126:99–105.
92. Papolos DF, et al. Am J Psychiatry
1996;153:1541–1547.
93. Pulver AE, et al. J Nerv Ment Dis
1994;182:476–478.
94. Doheny D, et al. Neurology 2002;
59:1244–1246.
95. Marechal L, et al. Am J Med Genet
B Neuropsychiatr Genet 2003;
119:114–117.
96. Saunders-Pullman R, et al.
Neurology 2002;58:242–245.
97. Zimprich A, et al. Nat Genet
2001;29:66–69.
98. Boghosian-Sell L, et al. Am J Hum
Genet 1996;59:999–1005.
99. Petek E, et al. Am J Hum Genet
2001;68:848–858.
100. Verkerk AJ, et al. Genomics 2003;
82:1–9.
101. Diaz-Anzaldua A, et al. Am J Med
Genet A 2004;127:17–20.
102. Grimes DA, et al. Neurology
2002;59:1183–1186.
103. De Carvalho Aguiar P, et al. Mov
Disord 2004;19:1237–1238.
104. Heiman GA, et al. Am J Med
Genet Part B Neuropsychiatr
Genet 2007;144B(3):361–364.
105. Bihari K, et al. Psychiatry Res
1992;42:267–272.
106. Cavallaro R, et al. Biol Psychiatry
2002;52:356–361.
107. Voon V, et al. Mov Disord
2010;25:2249–2252.
108. Greenberg BD, et al. Am J Med
Genet 2000;96:202–216.
109. Rauch SL, et al. Adv Neurol
2001;85:207–224.
110. Lerner A, et al. Cereb Cortex
2009;19:218–223.
111. Antshel KM, et al. Neuropsychol
Dev Cogn C Child Neuropsychol
2005;11:5–19.
112. Eichstedt JA, et al. Clin Psychol
Rev 2001;21:137–157.
113. Fisher SE, et al. Am J Hum Genet
2002;70:1183–1196.
114. Meyer-Lindenberg A, et al. Nat
Neurosci 2005;8:594–596.
115. Asimit J, et al. Annu Rev Genet
2010;44:293–308.
116. Cirulli ET, et al. Nat Rev Genet
2010;11:415–425.

117. Joel D. Prog Neuro-
psychopharmacol Biol Psychiatry
2006;30:374–388.
118. Joel D, et al. Behav Brain Res
2005;157:253–263.
119. Korff S, et al. Psychiatr Clin North
Am 2006;29:371–390.
120. Leonard HL, et al. Int J
Neuropsychopharmacol
2001;4:191–198.
121. Wang L, et al. Behav Pharmacol
2009;20:119–133.
122. Welch JM, et al. Nature
2007;448:894–900.
123. Zuchner S, et al. Ann Hum Genet
2008;72:725–731.
124. Bienvenu OJ, et al. Am J
Med Genet B Neuropsychiatr
Genet 2008;150B:710–720.
125. Grisham JR, et al. Eur Arch
Psychiatry Clin Neurosci
2008;258:107–116.
Chapter 11: Genetic contributions to OCD and related disorders
126. Snider LA, et al. Mol Psychiatry
2004;9:900–907.
127. Gershuny BS, et al. Depress
Anxiety 2008;25:69–71.
128. Gershuny BS, et al. Behav Res Ther
2003;41:1029–1041.
129. Cromer KR, et al. Behav Res Ther
2007;45:1683–1691.
130. Sasson Y, et al. Psychiatry Res
2005;135:145–152.
131. Samuels JF, et al. Am J Med Genet
B Neuropsychiatr Genet 2006;
141B:201–207.
132. Gorlov IP, et al. Am J Hum Genet
2008;82:100–112.
133. Hasler G, et al. Biol Psychiatry
2007;61:617–625.
134. Samuels J, et al. Behav Res Ther
2002;40:517–528.
135. Samuels JF, et al. Behav Res Ther
2007;45:673–686.
136. Samuels JF, et al. Behav Res Ther
2008;46:836–844.
137. Samuels JF, et al. Behav Res Ther
2008;46:1040–1046.
138. Lasalle VH, et al. Depress Anxiety
2004;19:163–173.
139. Wheaton M, et al. J Anxiety Disord
2008;22:243–252.
140. Zhang H, et al. Am J Hum Genet
2002;70:896–904.
141. Zuchner S, et al. Mol Psychiatry
2009;14:6–9.
142. Dodman NH, et al. Mol Psychiatry
2010;15:8–10.
143. Luescher UA, et al. Vet Clin North
Am Small Anim Pract 1991;
21:401–413.
144. Rapoport JL, et al. Arch Gen
Psychiatry 1992;49:517–521.
145. Voon V. Mov Disord 2004;
19:367–370.
133
 Post-traumatic stress disorder
Chapter
12 Michael J. Lyons, Tyler Zink, and Karestan C. Koenen
The “diathesis–stress” model [1] has been an influen-
tial approach for conceptualizing mental disorders.
It reconciles the roles of nature and nurture in the
etiology of a disorder. “Diathesis” refers to a predis-
position, tendency, or pre-existing vulnerability to
develop a disorder when the individual is exposed to
an environmental stressor. Post-traumatic stress dis-
order (PTSD) is a quintessential example of a diathesis–
stress phenomenon because it describes a process in
which an individual reacts to an environmental stres-
sor (the “trauma”) by developing a mental disorder
(PTSD). This chapter will provide some background
about PTSD and review the evidence for the role of
genetic factors as a diathesis for PTSD.
Overview of diagnosis
According to the DSM-IV-TR, PTSD is an anxiety
disorder characterized by a series of symptoms dir-
ectly related to the experience of a traumatic event [2].
In fact, PTSD is only one of two mental disorders in
DSM-IV-TR for which a precipitating event is neces-
sary, although not sufficient, for diagnosis. In order
for an event to be considered a potential antecedent to
PTSD, the individual must: (1) experience, witness, or
be confronted by actual or threatened death or serious
injury or a threat to the physical integrity of the self or
others; and (2) experience intense fear, helplessness,
or horror as a result of the traumatic event. Once such
a traumatic event has been established as a possible
antecedent, an individual must re-experience the event,
avoid stimuli related to it, and experience increased
arousal as a result of it in order to qualify for a
diagnosis of PTSD. Additionally, such symptoms must
persist for at least one month and significantly impair
daily functioning [2].
Principles of Psychiatric Genetics, eds John I. Nurnberger, Jr. and Wade H. Berrettini. Published by Cambridge University
Press. © Cambridge University Press 2012.
134
History of the construct
The definition of PTSD as a mental disorder, how-
ever, has evolved over time. While ideas about the
negative psychological impact of trauma have been
accepted for centuries [3], PTSD did not have a
formal psychiatric definition until 1980 in DSM-III
[4]. Inclusion of PTSD in DSM-III marked formal
acknowledgement from the psychiatric community
that the effects of trauma should be considered from
a mental health perspective, that inherent personal
weakness does not drive traumatic sequelae, and that
the negative experiences of those having suffered
trauma are legitimate [5]. However, before this shift,
recognition of the psychological importance of trauma
was not universal. In fact, psychological sequelae of
traumatic events only became an important focus of
medicine during the American Civil War. Pizarro et al.
[6] reported that 44% of soldiers reported signs of
mental or “nervous” disease after the Civil War, which
was often called “irritable heart” by nineteenth-century
physicians. Also, individuals suffering from mental and
physical reactions to train accidents spurred the notion
of “railway spine” or “postconcussion syndrome” ([79],
as cited in [5]). Considering that the symptoms of
these ailments included sleep disruptions, nightmares
about train accidents, avoidance of train travel, and
chronic pain, Oppenheim renamed the phenomenon
“traumatic neurosis” ([78] as cited in [5]).
Reference to psychological reaction to trauma
appeared in DSM-I under the label of traumatic neur-
osis [7]. However, this disorder was not present in
DSM-II [8]. The understanding of psychological reac-
tion to traumatic events underwent a metamorphosis
throughout this time period, leading up to the inclu-
sion of PTSD as a mental disorder in DSM-III [4].

Before World War I, traumatic neurosis was explored
through the uncovering of dissociated experiences via
hypnosis [9]. Ideas developed from Charles Samuel
Myers’s concept of World War I trench warfare vet-
erans experiencing “shell shock” to Abram Kardiner’s
notion of “war neurosis”, which was lumped together
with hysteria under the general umbrella of “psychic
trauma” from nonspecific damage to the nervous
system [10]. The return of Vietnam War veterans to
the United States and the feminist movement of the
1960s, served as an impetus to combine the under-
standing of psychological reaction to traumatic events
across several categories (combat exposure, sexual
abuse, natural disaster survival, and so forth) for
inclusion as one disorder in DSM-III [11].
Epidemiology of trauma exposure
While PTSD earned formal psychiatric recognition
due to the large number of suffering war veterans
[5, 11], many different forms of trauma affecting
different groups of people have been demonstrated
to be antecedents to the development of PTSD [2].
Given that researchers differ in their methods of
assessing traumatic events and the definition of what
constitutes a traumatic event has changed with each
revision of the DSM [12], there is a wide range of
estimates for the prevalence of trauma exposure. For
example, some researchers combine all combat
related traumatic events [13, 14], while others specify
two types [15] or three types [16] of combat-related
PTSD antecedents. Keeping potential definitional and
methodological differences in mind, estimates of the
prevalence of exposure to traumatic events typically
range from 40% (using DSM-III/III-R criteria for
trauma exposure) to 85% (using DSM-IV criteria for
trauma exposure) in community samples [17–22]. More
recent studies using DSM-IV criteria show much higher
estimates of trauma exposure [23, 24].
Epidemiology of diagnosis
Many individuals experience traumatic events, how-
ever, not all develop PTSD. In fact, as compared to
estimates of trauma exposure, PTSD develops in only
about 2–15% of the general population [15, 17, 20, 22,
25–27]. This wide range of estimates reflects differ-
ences in methodology and changes in diagnostic
criteria. Earlier studies, including the Epidemiologic
Catchment Area (ECA) study and those using similar
methodology, typically provided lower prevalence
Chapter 12: Post-traumatic stress disorder
estimates than later studies because the DSM-III and
DSM-III-R employed a more narrow definition of
trauma exposure than that of the DSM-IV. Later
surveys also queried participants about a wider
range of traumatic events using interview probes
designed to increase the likelihood that respondents
would provide sensitive and personal information.
For example, in the National Comorbidity Survey-
Replication, participants were specifically asked
whether they were physically abused as children. This
question was not asked of participants in the ECA
survey [12, 28].
Because individuals exposed to trauma do not
always develop PTSD, it is important to identify
factors associated with the probability of developing
PTSD. Many researchers have reported that more
women receive a diagnosis of PTSD than do men
[12, 29, 30]. A recent meta-analysis [31] suggested
that women are at greater risk for developing PTSD
across all types of trauma except for assaultive vio-
lence. However, this difference may reflect gender
differences in type of trauma exposure. Specifically,
women have a higher conditional risk for developing
PTSD because of their increased exposure to assault-
ive violence [29, 30]. Some of the other factors con-
tributing to the development of PTSD among those
exposed to trauma include being physically wounded
during the event [15], having lower cognitive ability
[32–34], having behavioral problems in adolescence
[15, 33, 35], being of lower socioeconomic status [14],
being sexually assaulted by an acquaintance [36, 37],
being high in trait neuroticism or introversion [38],
living in a country in the midst of political or ethnic
violence [28], and prior trauma exposure [39, 40].
Genetic studies of PTSD
Quantitative genetic studies
There are three approaches that have traditionally
been used to investigate the presence of genetic
influences on psychopathology: family studies, twin
studies, and adoption studies. In general, the logic of
a family study is that if a disorder is influenced by
genetic factors, relatives who share their genes with an
individual with the disorder of interest (a “proband”)
should have a higher risk for the disorder than indi-
viduals who are not related to a proband. Davidson
et al. [41] investigated female rape victims and found
that subjects with a family history of major depression
135
 Chapter 12: Post-traumatic stress disorder
were more likely to develop PTSD following rape than
individuals without such a family history. Yehuda and
her colleagues [42, 43] have investigated the relation-
ship of PTSD in Holocaust survivors and PTSD
in their offspring. In one study [42] they found that
offspring who had experienced traumatic events
themselves were more likely to develop PTSD if their
Holocaust-surviving parents had chronic PTSD. In a
second study that used some of the subjects from their
1998 study, but added additional subjects, Yehuda
et al. [43] found that the offspring of Holocaust sur-
vivors with PTSD had a higher risk of PTSD than
offspring of Holocaust survivors without PTSD. They
also observed that maternal PTSD was a stronger
predictor of offspring PTSD that paternal PTSD.
These observations are consistent with a familially
transmitted genetic vulnerability to the development
of PTSD following exposure to trauma, but are
equally compatible with a socially transmitted vulner-
ability. A weakness of family studies is the inability
to distinguish between genetic and environmental
factors as the mechanism of transmission.
Twin studies can avoid a shortcoming of family
studies by distinguishing genetic influences from
influences due to the family environment. This is
accomplished by capitalizing on a naturally occurring
experiment, that is, the difference in the genetic simi-
larity between monozygotic (MZ) and dizygotic (DZ)
twins. MZ twins share 100% of their genes, while DZ
twins share, on average, 50%. To the extent that MZ
twins are more similar to one another on some trait, it
can be inferred that genetic factors have some influ-
ence on the trait. Twin studies can decompose the
sources of individual differences into those that reflect
genetic differences among individuals, those that
reflect the aspects of the environment that are shared
by twins, such as neighborhood and parental educa-
tion, and environmental features that are not shared
by both members of a twin pair.
True and colleagues [44] studied over 4000 pairs
of male twins who had served in the US military
between 1965 and 1975 comprising the Vietnam Era
Twin Registry. Data on PTSD symptomatology were
collected by questionnaire. Genetic factors contrib-
uted to susceptibility for nearly all symptoms of PTSD
with heritabilities ranging from 13 to 34%. Combat
exposure was a strong predictor only of the
“re-experiencing” symptom cluster and the “avoided
activities” symptom. There was no detectable influ-
ence from environmental factors shared by twins on
136
susceptibility to PTSD symptoms. In another twin
study, Stein et al. [45] reported heritabilities ranging
from 28 to 38% for the individual PTSD symptom
clusters and no influence from environmental factors
shared by twins. Jang et al. [46] conducted a twin study
investigating genetic and environmental contributions
to susceptibility for developing post-traumatic stress
symptoms (PTSS) following exposure to traumatic
events. They studied over 400 twin pairs and deter-
mined lifetime frequency of exposure to assaultive and
nonassaultive trauma. They found that in the case
of nonassaultive trauma, PTSS were directly affected
by environmental factors that also influence exposure
to nonassaultive trauma. Both genetic and nonshared
environmental influences jointly affected PTSS for
assaultive trauma, and the number of traumatic events
moderated the severity of PTSS. The influence of gen-
etic factors became less important beyond some
threshold number of traumatic events.
Adoption studies are another method for disen-
tangling genetic factors from family environmental
factors, but we know of no adoption studies of PTSD.
One of the factors that contributes to the complexity
of conducting research on genetic factors in PTSD is
the role that trauma exposure plays in the disorder.
Exposure to a qualifying trauma is necessary, but not
sufficient for the diagnosis of PTSD. Therefore, it is
not possible to assess an individual’s vulnerability to
PTSD in the absence of exposure to trauma. These
considerations make the logistics of conducting an
adoption study of PTSD extremely difficult.
Genetic influence in exposure to trauma
Stein and colleagues [45], using the same sample
utilized in the report by Jang et al. [46] described
above, found that exposure to assaultive trauma, such
as robbery and sexual assault, was influenced signifi-
cantly by genetic factors, but not by shared environ-
mental factors. Interestingly, exposure to nonassaultive
trauma, such as natural disasters and motor vehicle
accidents, was significantly influenced by the shared
environment, but not genetic factors. They also found
that to some extent genetic factors that influenced
exposure to assaultive traumatic events also influ-
enced vulnerability to PTSD symptoms. Certainly
many traumatic events are fateful and befall an indi-
vidual regardless of his or her behavior. However, it
may be that personality traits, such as high sensation
seeking and/or low harm avoidance, make it more

likely that an individual will seek out environments in
which he or she will be exposed to trauma. This
reflects the phenomenon of “gene–environment cor-
relation” in which to some extent an individual’s
genetically influenced characteristics affect the type
of environment that the individual seeks out.
Lyons and colleagues [47] studied members of
the Vietnam Era Twin Registry to examine genetic
and nongenetic factors that influence wartime expos-
ure to traumatic events. Specific events examined
were volunteering for service in Vietnam, actual ser-
vice in Southeast Asia, a composite index of 18
combat experiences, and information from military
records about being awarded combat decorations.
There was a significant genetic influence on volun-
teering for service in Vietnam. Among twin pairs in
which both siblings served in Southeast Asia there
was a significant genetic influence on self-reported
combat experiences. The family environment did not
have a significant effect on any of the variables.
Analyses of data from military records regarding
being awarded a combat decoration provided very
similar results to those found for self-reported
combat experiences.
Molecular genetic studies
There have been a number of case-control association
studies conducted to try to identify genetic variants
that are related to a vulnerability to develop PTSD
following exposure to trauma. The strategy that has
been used thus far in this research has been the
“candidate gene” approach because of the very large
number of genes in the human genome and the tech-
nical limitations that existed until recently in terms of
genotyping very large numbers of markers simultan-
eously. In this approach genes are selected for interro-
gation based on the plausibility that they play a role in
the phenotype of interest. For behavioral and psychi-
atric phenotypes, such as PTSD, the genes with the
greatest biological plausibility are those that operate
in the central nervous system. However, there are
thousands of genes that influence the brain, so it has
been helpful to narrow the field of candidates further.
Information that we know about the neurobiology of
PTSD can inform decisions about where to look for
relevant genes. Genes that influence the hypothalamic-
pituitary-adrenal (HPA) axis, genes that influence the
locus cerulius/noradrenergic system, and genes that
influence limbic-frontal brain systems are all very
Chapter 12: Post-traumatic stress disorder
reasonable candidates for influencing the risk of
developing PTSD following exposure to trauma.
The most common form of association study is
very similar to the traditional case-control study.
A sample of unrelated individuals with the disorder
of interest is compared to a matched control sample of
unrelated individuals without the disorder in terms of
the presence of a risk factor. For example, a sample
of individuals with emphysema (cases) might be com-
pared to a matched sample of individuals without
emphysema (controls) for the putative risk factor
of smoking. If a significantly greater proportion of
individuals with emphysema were found to be
smokers than individuals without emphysema, then
an association between smoking and emphysema
would be inferred. In the genetic case-control associ-
ation study, samples of cases and controls are estab-
lished and the risk factor that is examined is a genetic
polymorphism. Genetic polymorphisms refer to two
or more genetic markers or alleles at the same genetic
locus, each of which occurs more often than could
be explained by mutation. The basic logic of the case-
control candidate gene study is that if it is determined
that one (or more) of the alleles is found more fre-
quently among cases than it is among controls, that
allele is associated with the phenotype. There are
a number of factors that go into determining the
statistical power (the probability of detecting a true
genetic effect) of association studies. As with other
types of studies, the size of the sample and the thresh-
old set for significance influence the probability of
detecting a significant effect. The power is also influ-
enced by the prevalence of the “risk” genotype and
its effect size (i.e. how much is risk of the disorder
increased in the presence of the specified marker).
Table 12.1 includes information about the 18 asso-
ciation studies of PTSD of which we are aware
(although several of these studies investigated clinical
features of subjects with PTSD, rather than the risk of
PTSD, per se). Eight of the studies investigated genes
in the dopaminergic system; six looked at the D2
dopamine receptor gene, one at the dopamine trans-
porter gene, and one at the dopamine beta-hydroxylase
(DBH) gene. Comings et al. [48] reported a greater
frequency of the A1 allele of the D2 dopamine recep-
tor among 35 patients with PTSD. However, after
correcting for multiple comparisons, this was not
significant. In 1996, Comings et al. [49] reported the
results from a study of Vietnam veterans who were
patients on an addiction treatment unit. There was
137
 Chapter 12: Post-traumatic stress disorder

Table 12.1 Candidate gene association studies of post-traumatic stress disorder (PTSD).

First Year Trauma Trauma Gene name Finding

author exposed type
controls?
Comings 1991 No Combat Dopamine receptor Excess D2A1 allele in PTSD cases p ¼ 0.007
D2 (DRD2)
Comings 1996 Yes Combat Dopamine receptor (Two samples) Excess D2A1 allele in PTSD
D2 (DRD2) cases p ¼ 0.041
Excess D2A1 allele in PTSD cases p ¼ 0.002
Gelernter 1999 No Combat Dopamine receptor No significant association between D2A1
D2 (DRD2) allele/DRD2 haplotypes and PTSD
Young 2002 No Combat Dopamine receptor Excess D2A1 allele only in PTSD cases with
D2 (DRD2) harmful drinking p < 0.001
Mustapić 2007 Yes Combat Dopamine No main effect for DBH gene on risk of PTSD
β-hydroxylase (DBH)
Segman 2002 Yes Various Dopamine Excess 9-repeat allele in PTSD cases p ¼ 0.012
transporter (DAT1)
Bachman 2005 Yes Combat Glucocorticoid No significant association between GCCR
receptor (GCCR) polymorphisms and PTSD
Lappalainen 2002 No Combat Neuropeptide Y No significant association between Leu7Pro
(NPY) polymorphism and PTSD
Lu 2008 No Various Cannabinoid significant association between a CNR1
receptor (CNR1) haplotype and PTSD, p < 0.04
Freeman 2005 No Combat Apolipoprotein E Association between APOE ε2 and poorer
allele (APOE) memory scores as well as more severe trauma
re-experiencing
Koenen 2005 No Acute Glucocorticoid Two FKBP5 SNPs are significantly associated
injury receptor-regulating with dissociation during (p < 0.05) and after
cochaperone (FKBP5) ( p < 0.03) traumatic accidents in children
Binder 2008 No Various Glucocorticoid Genetic variation in the FKBP5 gene may
receptor-regulating place individuals with significant past child
cochaperone (FKBP5) abuse at significant risk for PTSD, p < 0.0004
Lee 2005 No Various Serotonin transporter Excess s allele in PTSD cases p ¼ 0.04
(SLC6A4)
Kilpatrick 2007 Yes Hurricane Serotonin transporter Significant association between s/s genotype
(SLC6A4) and PTSD in adults with high hurricane
exposure and low social support, p < 0.03
Lee 2006 No Not Brain-derived No association between BDNF Val66Met
specified neurotrophic factor and PTSD
(BDNF)
Zhang 2006 Not Not Brain-derived No significant association between three
specified specified neurotrophic factor BDNF variants and PTSD
(BDNF)

138

Table 12.1 (cont.)
First Year Trauma Trauma Gene name
author exposed type
controls?
Lawford 2003 No Combat Dopamine receptor
D2 (DRD2)
Lawford 2006 No Combat Dopamine receptor
D2 (DRD2)
SNPs, single nucleotide polymorphisms.
a significantly greater frequency of the A1 allele of the
D2 dopamine receptor among the 37 subjects with
PTSD compared to the 19 controls who had been
exposed to high levels of combat but did not develop
PTSD. Gelernter et al. [50] attempted to replicate
Comings’ findings using a sample of 52 Vietnam
veterans with PTSD and 87 controls without PTSD.
They found no association between the A1 allele of
the D2 dopamine receptor and PTSD. Young et al.
[51] examined the frequency of the A1 allele among
91 patients with PTSD versus 51 controls and found
that A1 allelic frequency was significantly higher
among the PTSD patients. They also found that PTSD
patients with the A1 allele were more likely to have
alcohol problems than PTSD patients without the
A1 allele. Mustapić et al. [52] investigated the DBH
gene which converts dopamine to norepinepharine.
Among a sample of individuals exposed to combat
trauma, they did not find a relationship between
PTSD and the polymorphism in the DBH gene that
they investigated. There was, however, a trend for an
interaction between PTSD and DBH genotype; PTSD
subjects with one version of the DBH polymorphism
(the CC genotype) had lower levels of plasma DBH
activity than subjects without PTSD, but in subjects
with other genotypes, there was no relationship
between PTSD and plasma DBH activity.
In another study of the dopaminergic system,
Segman et al. [53] examined the dopamine trans-
porter gene (DAT). Specifically, they investigated the
association of the nine repeat allele of the variable
number tandem repeat in the dopamine transporter
gene. There was a significantly higher frequency of
the 9 repeat allele among the 102 chronic PTSD
patients compared to 104 controls who had been
exposed to trauma, but did not develop PTSD.
Bachmann et al. [54] investigated two common
glucocorticoid receptor (GR) polymorphisms (N363S
Chapter 12: Post-traumatic stress disorder
Finding
Possible association between D2A1 allele
and response to SSRI paroxetine, p ¼ 0.03
DRD2 gene is associated with comorbid
depression, anxiety, and social dysfunction
in untreated veterans with PTSD, p < 0.05
and Bcll) in 118 Vietnam veterans with PTSD and
42 controls who were Vietnam veterans with combat
exposure, but without PTSD. The frequency of the
N363S and Bcl1 GR polymorphisms were not greater
in PTSD patients than controls and did not differ
from reported population frequencies.
In a study that was focused on alcohol depend-
ence, Lappalainen, Kranzler, and Malison [55]
included a group (n = 77) of Vietnam-era combat
veterans with PTSD who were compared to 202 com-
munity controls. They found that the groups did
not differ in the frequency of the neuropeptide
Y Leu7Pro polymorphism. Lu and colleagues [56]
conducted an association study focused on atten-
tion-deficit hyperactivity disorder (ADHD) that also
assessed PTSD and several other disorders that are
potentially comorbid with PTSD. The candidate gene
that they investigated was the cannabinoid receptor
gene (CNR1). They found a significant association
between a single nucleotide polymorphism (SNP)
haplotype and PTSD. While suggesting that their
finding was preliminary, they concluded that the
CNR1 gene warrants further investigation for a role
in psychiatric disorders, including PTSD.
Some studies have looked at the association of
candidate genes with characteristics among individ-
uals with PTSD, rather than looking at the genotype
as a risk factor for PTSD. Freeman et al. [57] looked
at the relationship of APOE genotype (which has been
implicated in the risk of Alzheimer’s disease) among
54 veterans with PTSD. They found that the APOE
ε2 allele was associated with significantly worse
re-experiencing symptoms and impaired memory
function in this population. Koenen and colleagues
[58] examined the FKBP5 gene, which is a gluco-
corticoid receptor-regulating cochaperone of stress
proteins. They studied 46 children following their
admission to a hospital for an acute medical injury.
139
 Chapter 12: Post-traumatic stress disorder
They found that two SNPs in the FKBP5 gene were
significantly associated with dissociation during and
since the injury.
Binder et al. [59] reported a cross-sectional study
examining genetic and psychological risk factors
in 900 nonpsychiatric clinic patients (762 included
for all genotype studies) with significant levels of
childhood abuse as well as nonchild abuse trauma.
Eight SNPs spanning the FKBP5 locus were utilized.
Although FKBP5 SNPs did not directly predict PTSD
symptom outcome or interact with level of nonchild
abuse trauma to predict PTSD symptom severity, four
SNPs in the FKBP5 locus significantly interacted with
the severity of child abuse to predict level of adult
PTSD symptoms.
Several studies have investigated the serotonin
transporter gene (SLC6A4; SERT). Lee and colleagues
[60] examined the serotonin-transporter-linked poly-
morphic region (SERTPR or 5HTTLPR). They stud-
ied 100 PTSD patients and 197 unrelated healthy
controls using a case-control design. PTSD patients
had a higher frequency of the low expression (s/s)
genotype than normal controls. Kilpatrick et al. [61]
also investigated whether the same polymorphism
influenced the risk of PTSD following exposure to
a hurricane. Subjects were 589 adults from Florida.
The low-expression variant of the 5HTTLPR poly-
morphism was associated with risk for PTSD following
high exposure to hurricane, but only if there was low
social support. High risk individuals (high hurricane
exposure, the low-expression 5HTTLPR variant, low
social support) had 4.5 times the risk of developing
PTSD compared to low risk individuals.
Lee et al. [62] studied the genotype and allele
frequencies of the brain-derived neurotrophic factor
(BDNF) gene Val66Met polymorphism in 106 PTSD
patients and 161 unrelated healthy controls using
a case-control design. The genotype and allele fre-
quencies for the BDNF gene polymorphism did not
differ between the two groups. Zhang et al. [63] also
investigated the relationship of the BDNF Val66Met
polymorphism to PTSD along with two other BDNF
polymorphisms. They studied 96 patients with PTSD
(as well as a number of other psychiatric disorders)
and 250 normal controls. No association was found
between any of the three included BDNF gene variants
and PTSD.
Two studies by Lawford and colleagues [64, 65]
did not look at the association between genotype and
risk for PTSD, but investigated the possible effects of
140
D2 dopamine receptor genotype on features related to
PTSD. They found that among their subjects with
PTSD, those with the A1 allele of the D2 dopamine
receptor gene had higher scores for anxiety/insomnia,
social dysfunction, and depression [65]. PTSD sub-
jects with the A1 allele had more severe comorbid
psychopathology. In an earlier study, Lawford et al.
[64] found that PTSD subjects with the A1 allele of
the D2 dopamine receptor gene showed significant
improvement in social functioning when treated with
paroxetine compared to PTSD subjects without the
A1 allele.
Genetic influences on comorbidity
There have been a number of twin studies that have
examined possible genetic influences that contribute
to the comorbidity of PTSD with other mental dis-
orders. Koenen et al. [66] found that a genetic liability
that influenced both PTSD and major depression was
the primary reason that the two disorders co-occur. In
2003, Koenen and colleagues reported that comorbid-
ity between PTSD and depression and PTSD and
dysthymia reflect, in part, genetic factors that impart
a risk for all three disorders. Xian et al. [67] demon-
strated that genetic factors contributed to the comor-
bidity observed among PTSD, depression, and alcohol
dependence. McLeod et al. [68] found that genetic
factors contribute significantly to the frequently
observed comorbidity between PTSD and alcohol
dependence. Chantarujikapong et al. [69] found that
some of the genetic vulnerability to PTSD also confers
vulnerability to generalized anxiety disorder and
panic disorder. Koenen et al. [58] found that most
of the association between PTSD and nicotine
dependence reflects shared genetic influences, but
the relationship between late-onset smoking and
PTSD is not influenced by genetic factors. Overall,
these various studies indicate that at least some of the
genetic factors that impart a risk for PTSD also
impart some risk for a number of other disorders.
Future directions
Epigenetics
An area that is likely to play an increasingly important
role in understanding how genes influence psycho-
pathology in general and PTSD in particular is epige-
netics. Epigenetics refers to changes in gene expression
that are not caused by the DNA sequences, but are

instead caused by other mechanisms. Perhaps more
than any other psychiatric disorder, PTSD is amenable
to this type of approach because trauma might plaus-
ibly be associated with changing gene expression.
At least two studies of PTSD have utilized this
approach. Segman and colleagues [70] used oligonu-
cleotide microarrays to measure peripheral blood
mononuclear cell (PBMC) gene expression among
survivors of trauma at an emergency room and four
months later. The profile of gene expression signa-
tures at both times distinguished survivors with PTSD
at one and four months, from those who met no
criteria for PTSD. Results demonstrated a general
reduction in PBMC ’s expression of transcription acti-
vators among trauma survivors affected by PTSD.
Several genes that differentiated the groups had pre-
viously been found to have a role in stress response.
Zieker et al. [71] studied eight individuals who
had been exposed to the Ramstein air show catas-
trophe in 1989 and eight controls using cDNA micro-
arrays. The genes that were selected for inclusion on
the custom chip used for the molecular analyses were
those that played a role in stress and immune
response (inflammation, apoptosis, stress response
and related pathways). Their statistical analysis
identified 4 upregulated genes and 14 downregulated
genes in the PTSD group versus controls. Most of the
genes that were downregulated were associated with
immune functions or with reactive oxygen species.
The authors concluded that transcript differences that
they observed suggest a temporary perturbance of the
oxidative stress system and specific immune parameters
in patients with PTSD.
Animal models
There have been a number of studies that have used
animal models to investigate the role of genetic
factors in PTSD. Clearly one cannot examine most
of the psychological and behavioral features of human
PTSD in animals. However, plausible animal models
can be proposed because traumatic stress can be
delivered very reliably and various species display
behavioral and physiological characteristics that rep-
resent a reasonable analogue to PTSD. Kaufer et al.
[72] investigated long-lasting changes in cholinergic
gene expression in mice as a possible mechanism by
which exposure to trauma is transduced into a rela-
tively stable phenotype such as PTSD. They found
that modulated cholinergic gene expression reduced
Chapter 12: Post-traumatic stress disorder
available acetylcholine and depressed cholinergic neu-
rotransmission following exposure to stress. They
suggested that while this mechanism might have the
beneficial short-term effect of quieting brain activity
following trauma, the long-term effects could be
damaging.
Several animal model studies have investigated the
promoter polymorphism of the serotonin transporter
(5-HTTLPR) gene that is associated with increased or
decreased transcription (several human case-control
candidate gene studies of the 5-HTTLPR are
described above). Barr et al. [73] examined macaques
who had either been raised with their mothers or in
peer-only groups. They examined adrenocortico-
tropic hormone (ACTH) and cortisol levels during
separation stress. Macaques with the l/s allele of
5-HTTLPR had higher ACTH levels during stress
and there was a significant interaction of separation
stress  rearing  5-HTTLPR genotype, which
indicated that the effect of 5-HTTLPR genotype on
hormonal response during stress is modulated by
early experience. Bennett et al. [74] examined the
effect of the 5HTTLPR promoter polymorphism in
rhesus monkeys. They found that the 5HTTLPR poly-
morphism interacted with deleterious early rearing
experiences. Specifically, monkeys with the genotype
associated with less transcription had lower 5-hydro-
xyindoleactic acid (a 5HT metabolite) if they had been
reared in deleterious conditions, but not if they had
been raised in more favorable conditions. Rasmusson
et al. [75] studied the effects of psychological stress
(cues conditioned to foot-shock) on the expression of
BDNF in male rats. Tones that had previously been
paired with foot-shock downregulated BDNF; BDNF
mRNA was significantly decreased in the hippocam-
pus following exposure to the aversive conditioned
stimulus.
Animal models are very valuable because more
invasive techniques can be utilized (e.g. measuring
mRNA in the hippocampus) and controlled exposure
to stressors is possible. It is not clear how precise an
analogue to PTSD one can create in an animal model,
but it seems likely that many of the same genetic and
physiological phenomena are operating.
Genome-wide association studies
The cutting-edge research design for gene discovery
is currently the genome-wide association studies
(GWAS). In the most typical case, a large number of
141
 Chapter 12: Post-traumatic stress disorder
unrelated individuals with the disorder are identified
and compared to a large number of well-matched
individuals without the disorder. The density of gen-
etic markers and the extent of linkage disequilibrium
must be adequate to capture a large proportion of
the common variation in the human genome. In
order to identify variants of modest effect, which
will probably include most if not all relevant to psy-
chiatric disorders, a large number of subjects for
genotyping is necessary. At this writing, no GWAS
studies of PTSD have been conducted.
Methodological challenges
As with most psychiatric disorders, it seems likely
that more than one gene, and quite possibly a large
number of genes, play a role in imparting vulnerabil-
ity to PTSD. There may also be genetic heterogeneity,
that is, there may be a number of different genetic
pathways to PTSD, with none being necessary or
sufficient. There may be incomplete penetrance (i.e.
the phenotype may not appear in all individuals with
the relevant genotype) and variable expressivity
(the extent to which the genotype is expressed in an
individual’s phenotype may vary). The situation may
be further complicated by interactions among genes.
The interaction of PTSD relevant genes with the
environment (G  E) is assured by the criteria for
PTSD which require that an individual be exposed to
a traumatic event. By this definition, genetic factors
cannot produce PTSD in the absence of a stressor.
The definition of a G  E interaction is that genetic
factors determine sensitivity to the environment.
In the case of PTSD we are searching for genes that
determine whether an individual develops the charac-
teristic signs and symptoms of PTSD after he or she is
exposed to an environmental stimulus (the stressor).
An important methodological issue in the design
of studies to identify genes that influence PTSD is the
selection of cases and controls. The question is “who
should be considered to be a PTSD case?” The selec-
tion of cases for this type of study is considered
“phenotype definition.” Researchers must decide
whether to select on the basis of current PTSD or
PTSD at any time during the individual’s lifetime.
Since the primary sequence of DNA does not change,
it seems feasible to utilize individuals who have ever
had PTSD, whether it is current or not. (This
reasoning does not apply to epigenetic studies because
phenomena-like gene expression do change and
142
effects that are present in current cases of PTSD
might not be present in individuals who have
recovered.) A simple and reasonable answer to defin-
ing a case is to apply the DSM criteria for the dis-
order. However, evidence suggests that PTSD, as
defined by the DSM criteria, is a heterogeneous dis-
order composed of at least two subtypes, internalizing
and externalizing [76, 77]. Future genetic studies may
benefit from a more homogenous definition of cases
using these subtypes. An important consideration in
selecting controls is whether or not they have been
exposed to a traumatic event. If an individual has
not been exposed to a traumatic event, he or she
cannot manifest PTSD regardless of the genetic sus-
ceptibility that he or she might have. For the purpose
of studies such as association studies, it is probably
best to classify individuals without trauma exposure
as “phenotype unknown” and exclude them from the
study. It might be that putative PTSD endopheno-
types, such as hypothalamic-pituitary-adrenal axis
functioning, acoustic startle response, and physio-
logical markers of increased arousal might improve
the ability to detect genes beyond sole reliance on
the clinical phenotype.
Population stratification is another issue that
must be considered in designing association studies
of PTSD. This refers to a situation in which groups
may differ in the frequency of an allele and may differ
in some other, unrelated characteristic that is relevant
to the phenotype of interest. For example if Northern
Europeans and South Asians differ in the frequency of
some genetic marker and they also differ in social
support, which has a demonstrated influence on the
risk of PTSD, one might draw the spurious conclu-
sion that the marker was associated with PTSD if the
two ethnic groups were differentially represented
among cases and controls.
Conclusions
Evidence from twin and family studies clearly
demonstrates that genetic factors exert a significant
influence on the risk of PTSD, but at the level of
identifying specific genes that contribute to risk, the
evidence is less clear. There have been a number of
positive findings with regard to genes that play a role
in the dopaminergic system, but none have a compel-
ling record of replication. Similarly, some findings
point to the relevance of genes related to serotonin
and glucocorticoid functioning, but replication will be

required. It seems that we are only at the beginning of
explicating what is likely to be a very complex rela-
tionship between genes and PTSD. However, if the
dramatic advances that are taking place in technology
at the molecular level and at the statistical level are
References
1. Zubin J, et al. J Abnorm Psychol
1977;86:103–126.
2. American Psychiatric Association.
Diagnostic and Statistical
Manual of Mental Disorders,
Fourth Edition, Text Revision
(DSM-IV-TR). Washington,
DC: American Psychiatric
Association; 2000.
3. Tomb DA. Psychiatr Clin North
Am 1994;17:237–250.
4. American Psychiatric Association.
Diagnostic and Statistical Manual
of Mental Disorders, Third Edition
(DSM-III). Washington, DC:
American Psychiatric Association;
1980.
5. Lasiuk GC, et al. Perspect Psychiatr
Care 2006;42:13–20.
6. Pizarro J, et al. Arch Gen
Psychiatry 2006;63:193–200.
7. American Psychiatric Association.
Diagnostic and Statistical Manual
of Mental Disorders, First Edition
(DSM-I). Washington, DC:
American Psychiatric Association;
1952.
8. American Psychiatric Association.
Diagnostic and Statistical Manual
of Mental Disorders, Second
Edition (DSM-II). Washington,
DC: American Psychiatric
Association; 1968.
9. Herman JL. Trauma and Recovery:
The Aftermath of Violence –
From Domestic Abuse to Political
Terror. New York: Basic Books;
1997.
10. Lamprecht F, et al. Psychosom
Med 2002;64:222–237.
11. Paige SR. Integr Physiol Behav Sci
1997;32:75–83.
12. Breslau N. Can J Psychol 2002;47:
923–929.
Chapter 12: Post-traumatic stress disorder
matched by the commitment of clinical and epidemi-
ological researchers, it is not unreasonable to look
forward to exciting discoveries that eventually will
help to reduce or eliminate the suffering associated
with PTSD.
13. Gallers J, et al. J Trauma Stress
1988;1:181–192.
14. Sutker PB, et al. Psychol Rep
1990;66:912–914.
15. Helzer JE, et al. N Engl J Med
1987;317:1630–1634.
16. Delimar D, et al. Mil Med
1995;160:635–639.
17. Breslau N, et al. Arch Gen
Psychiatry 1991;48:216–222.
18. Breslau N, et al. Arch Gen
Psychiatry 1997;54:81–87.
19. Creamer M, et al. Psychol Med
2001;31;1237–1247.
20. Kessler RC, et al. Arch Gen
Psychiatry 1995;52:1048–1060.
21. Norris FH. J Consult Clin Psychol
1992;60:409–418.
22. Resnick HS, et al. J Consult Clin
Psychol 1993;61:984–991.
23. Breslau N, et al. Psychol Med
2004;31:889–898.
24. Breslau N, et al. J Urban Health
2004;8:530–544.
25. Shore JH, et al. J Nerv Ment Dis
1989;177:681–685.
26. Davidson JRT, et al. Psychol Med
1991;21:1–19.
27. Kessler RC, et al. Arch Gen
Psychiatry 2005;62:593–602.
28. Kessler RC. J Clin Psychiatry
2000;61:4–14.
29. Breslau N, et al. Psychol Med
1999;29:813–821.
30. Breslau N, et al. Arch Gen
Psychiatry 1998;55:626–632.
31. Tolin DF, et al. Psychol Bull
2006;132:959–992.
32. Breslau N, et al. Arch Gen
Psychiatry 2006;63(11):
1238–1245.
33. Koenen KC, et al. Psychol Med
2007;37:181–192.
34. Kremen WS, et al. Arch Gen
Psychiatry 2007;64:361–368.
35. Storr CL, et al. Am J Psychiatry
2007;164:119–125.
36. McLeer SV, et al. J Am Acad
Child Adolesc Psychiatry 1988;
27:650–654.
37. Arata CM, et al. J Psychol Hum Sex
1995;8:79–92.
38. McFarlane AC. Br J Psychiatry
1988;152:116–121.
39. Brewin CR, et al. J Consult Clin
Psychol 2000;68:317–36.
40. Ozer EJ, et al. Psych Bull
2003;129:52–73.
41. Davidson JRT, et al. J Psychiatr
Res 1998;32:301–309.
42. Yehuda R, et al. Am J Psychiatry
1998;155:841–843.
43. Yehuda R, et al. J Psychiatr Res
2001;35:261–270.
44. True WR, et al. Arch Gen
Psychiatry 1993;50:257–264.
45. Stein MB, et al. Am J Psychiatry
2002;159:1675–1681.
46. Jang KL, et al. Twin Res Hum
Genet 2007;10:564–572.
47. Lyons MJ, et al. Am J Med
Genet B Neuropsych Genet 1993;
48:22–27.
48. Comings DE, et al. JAMA
1991;266:1793–1800.
49. Comings DE, et al. Biol Psychiatry
1996;40:368–372.
50. Gelernter J, et al. Biol Psychiatry
1999;45:620–625.
51. Young BR, et al. Alcohol
Alcoholism 2002;37:451–456.
52. Mustapić M, et al. Am J Med
Genet B Neuropsych Genet
2007;144:1087–1089.
53. Segman RH, et al. Mol Psychiatry
2002;7:903–907.
143
 Chapter 12: Post-traumatic stress disorder
54. Bachmann AW, et al. Psycho-
neuroendocrinology 2005;30:
297–306.
55. Lappalainen J, et al. Arch Gen
Psychiatry 2002;59:825–831.
56. Lu AT, et al. Am J Med Genet B
Neuropsych Genet 2008;147:
1488–94.
57. Freeman T, et al. J Neuropsychiatry
Clin Neurosci 2005;17:541–543.
58. Koenen KC, et al. Arch Gen
Psychiatry 2005;62:1258–1265.
59. Binder EB, et al. JAMA
2008;299:1291–1305.
60. Lee HJ, et al. Depress Anxiety
2005;21:135–139.
61. Kilpatrick DG, et al. Am
J Psychiatry 2007;164:1693–1699.
62. Lee HJ, et al. Stress Health
2006;22:115–119.
144
63. Zhang H, et al. Am J Med Genet B
Neuropsych Genet 2006;141:
387–393.
64. Lawford BR, et al. Eur
Neuropsychopharmacol 2003;
13:313–320.
65. Lawford BR, et al. Eur
Psychiatry 2006;21:
180–185.
66. Koenen KC, et al. J Affect Disord
2008;105:109–115.
67. Xian H, et al. Drug Alcohol
Depend 2000;61:95–102.
68. McLeod DS, et al. J Trauma Stress
2001;14:259–275.
69. Chantarujikapong SI, et al.
Psychiatry Res 2001;
103:133–146.
70. Segman RH, et al. Mol Psychiatry
2005;10:500–513.
71. Zieker J, et al. Mol Psychiatry
2007;12:116–119.
72. Kaufer D, et al. Nature 1998;
393:373–377.
73. Barr CS, et al. Biol Psychiatry
2004;55:733–738.
74. Bennett AJ, et al. Mol Psychiatry
2002;7:118–122.
75. Rasmusson AM, et al. Neuro-
psychopharmacology 2002;
27:133–142.
76. Miller MW, et al. Psychol Assess
2003;15:205–215.
77. Miller MW, et al. Behavior Ther
2007;38:58–71.
78. Weisaeth L, et al. PTSD Res Q
1991;2(2):1– 7.
79. Cohen ML, et al. Pain Rev 1996;
3:181–202.
 Antisocial behavior: gene–environment
Chapter
13 interplay
Laura A. Baker, Catherine Tuvblad, Serena Bezdjian, and Adrian Raine
The term “antisocial behavior” is a broad one that
encompasses many facets of deviance, most of which
bring harm to another person or involve the violation
of the rights of others in some way. Violence and
aggression bring physical or psychological harm to
other individuals, while property destruction and
theft show disregard and possible damage to another
person by way of their possessions. Antisocial behav-
ior often involves breaking the law, although less
serious or more subtle forms of rule violations such
as skipping classes, disruptive behavior, lying, and
deception are also considered to be antisocial in the
broadest definition.
As such, antisocial behavior in various forms is of
key importance in several psychiatric disorders, both
in children and adults. The American Psychiatric
Association [1] definitions for oppositional defiant
disorder and conduct disorder in children, as well as
antisocial personality disorder in adults, all include
some form of antisocial behavior in their symptoms,
while aggression features significantly in conduct dis-
order and antisocial personality disorder. Further-
more, these primary disorders for which antisocial
behavior is symptomatic are often comorbid with
other disorders, including attention-deficit hyper-
activity disorder (ADHD) and substance dependence,
while aggression is either a comorbid condition of
(or is central to) many psychiatric disorders, includ-
ing autism, schizophrenia, depression, dementia, and
intermittent explosive disorder. Alcohol and other
substance abuse in particular are often considered
forms of antisocial behavior, because of the harm they
bring to family members, friends and neighbors, or
society as a whole. Although not currently defined in
the DSM-IV-TR, psychopathy is another disorder for
which antisocial behavior plays a major role. Indeed,
some clinicians and researchers cast an even broader
Principles of Psychiatric Genetics, eds John I. Nurnberger, Jr. and Wade H. Berrettini. Published by Cambridge University
Press. © Cambridge University Press 2012.
net around antisocial behavior and these related
behaviors under the term “externalizing disorders”.
Externalizing disorders include different disruptive
and problem behaviors within the disinhibitory spec-
trum such as substance use, antisocial behavior,
ADHD, conduct disorder, as well as personality traits
such as impulsivity and sensation seeking. Several
studies have shown that these traits and behaviors
are related on a phenotypic level [2, 3]. Twin studies
have shown that they also share a common genetic
component [4].
Genetic influences are well evident for antisocial
behavior, both in the broadest sense of externalizing
behaviors as well as in various narrower forms such as
aggression, law-breaking, and psychopathic personal-
ity traits. This research stems largely from over
three decades of twin, family, and adoption studies
in which concordance for disorders or trait covaria-
tion among relatives has been amply demonstrated to
vary according to their genetic relatedness. That is,
monozygotic (MZ) or identical twins show greater
similarity than dizygotic (DZ) or fraternal twins and
nontwin siblings, who in turn are more similar than
unrelated siblings for individual and composite meas-
ures of antisocial behavior. Although genetic influ-
ences on antisocial behavior appear to hold across
both sexes, their effects to some extent depend on
age, and the ways in which antisocial behavior
is defined and measured. Most importantly, genetic
predispositions, though important, often appear more
deleterious in the presence of adverse environments.
Indeed, antisocial behavior is one domain of human
behavior in which gene  environment interactions
have been found repeatedly.
In spite of the overwhelming evidence for genetic
contributions to the risk for engaging in antisocial
behavior, the specific nature of the genetic risk is only
145
 Chapter 13: Antisocial behavior: gene–environment interplay
beginning to be understood. For the most part, the
genetic effects in antisocial behavior remain unspeci-
fied, and heritable influences are considered largely as
a black box. This lack of specificity is also true for
environmental influences in behavioral genetic stud-
ies of antisocial behavior, in that these effects are
“anonymous” and we have little certainty about which
experiences or circumstances exert true causal effects
on deviant outcomes. Advances in molecular genetics
and multivariate genetic models, however, have
provided methods for unpacking both the genetic
and environmental black boxes for antisocial behav-
ior. Based on molecular genetic techniques that are
now widely accessible to social scientists, we now have
opportunities for identifying specific genes, and for
identifying environmental factors that limit or
enhance genetic expression in antisocial behavior.
Even without specific DNA measures, multivariate
genetic models help unpack the genetic and environ-
mental effects by examining covariation between
antisocial behavior and its various social (e.g. socioeco-
nomic status, peer characteristics, parental monitoring,
and discipline) and biological (e.g. neurotransmitters,
physiological arousal, frontal lobe function, and hor-
mones) risk factors and how they might be genetically
and environmentally mediated.
This chapter reviews the evidence for genetic and
environmental influences – both specified and
unspecified – in antisocial behavior. We discuss her-
itability of both adult and child mental disorders in
DSM-IV-TR, for which antisocial behavior is central
to their diagnosis, but also review heritability of the
related externalizing disorders. Genetic studies of
crime in particular also illustrate the importance of
genotype  environment interactions, which are now
beginning to be understood at a more molecular level.
We highlight some of the most exciting new direc-
tions in this field, which aim to unpack the genetic and
environmental black boxes in antisocial behavior, and
understand the complexities of the gene–environment
interplay in antisocial development.
Heritability of the broader construct
of externalizing problems
One approach for understanding the role of genetic
influences in the wider realm of externalizing behav-
ior problems is through meta-analysis of behavior
genetic studies, in which heritability estimates for
the varying forms of antisocial behavior can be
146
compared and combined. One recent meta-analysis
[5] included 51 distinct twin and adoption studies
that focused primarily on some dimension of anti-
social behavior in children or adults. These included
studies of trait aggression, criminal offending, and
symptoms for the major psychiatric disorders involv-
ing antisocial behavior (i.e. conduct disorder, opposi-
tional defiant disorder, antisocial personality disorder),
and involved a variety of methods of assessment
(e.g. self-report, parent ratings, and official records).
Heritable effects, including additive genetic influ-
ences (0.32), nonadditive genetic influences (0.09),
explained in total 41% of the variance in antisocial
behavior when combining across studies, and these
effects did not differ for males and females. Environ-
mental effects shared by relatives accounted for less
than half as much variance (16%), with the remaining
variance (43%) being explained by nonshared envir-
onment. These effects were comparable for males
and females, although they differed significantly
according to definition, method of assessment of
antisocial behavior, and age of the subjects. Criminal
convictions appear more influenced by nonadditive
genetic effects (i.e. due to genetic dominance or epis-
tasis), while parental ratings and younger samples are
more influenced by shared environmental factors.
Given the confound between age and method of
assessment across studies – studies of children tend
to rely on parental ratings – it is difficult to know
whether the greater effect of shared environment in
children may simply be an artifact of rater bias.
It should also be noted that error variance and the
variance due to gene  environmental interactions
are included in the component attributed to non-
shared environmental influences, thus tending to
underestimate genetic contributions and overestimate
environmental contributions.
Individual studies have also examined the genetic
influence in the broad externalizing factor by com-
bining various antisocial behavior measures from the
same subjects into one model (Figure 13.1). A recent
study of antisocial behavior in 9–10-year-old twins
attempted to evaluate the effects of raters’ identity
on estimates of genetic and environmental influences
during childhood [6]. The study was based on a socio-
economically and ethnically diverse sample of 605
pairs of twins (MZ, DZ same sex, and DZ opposite
sex) and their caregivers who participated in a com-
prehensive assessment of the twins’ antisocial behav-
ior and related risk factors. Both the child and his/her
 Chapter 13: Antisocial behavior: gene–environment interplay

Figure 13.1 Standardized parameter

AC CC EC estimates for common pathways model
of antisocial behavior in multiple
informants. (From [111], with permission.
Copyright © 2007 by the American
0.00 Psychological Association.)
0.98* 0.19

Shared
view

0.67* 0.55*
0.42*

Caregiver Child Teacher

report report report

0.24 0.38* 0.58* 0.55* 0.15 0.71* 0.33 0.53* 0.45* 0.33*

AM RM EM AK CK EK AT RT CT ET

caregiver provided reports of the child’s antisocial Table 13.1 Proportions of variance in antisocial behavior
behavior, in addition to teacher ratings of each child. explained by genetic, shared and nonshared environment in
different informants.
Composite measures of antisocial behavior were com-
puted for each rater – parent, teacher, and child – Source of influence Caregiver Child Teacher
based on several standardized instruments measuring
Genetic
rule-breaking behaviors, including theft and violence,
as well as reactive and proactive aggression and con- Common factor 0.436 0.168 0.289
duct disorder symptoms, as rated by each informant. Informant specific 0.059 0.304 0.108
The pattern of genetic and environmental influences
Total 0.495 0.472 0.397
in the composite antisocial behavior measures for
these pre-adolescent children, as summarized in Shared environment/
Figure 13.1 and Table 13.1, was similar to that found rater effects
in Rhee and Waldman’s meta-analytic review, in that: Common factor 0.000 0.000 0.000
(1) genetic effects were significant for antisocial a
Informant specific 0.146 0.021 0.203
behavior as assessed by each of the three informants;
(2) shared environmental effects were larger for Rater effects 0.281
parent and especially for teacher ratings of antisocial (correlated errors)b
behavior as compared to the children’s self-report; Total 0.146 0.021 0.484
and (3) the respective magnitudes of genetic and Nonshared
environmental effects were comparable for males environment
and females. A larger effect of shared environment
for childhood antisocial behavior was exhibited, as Common factor 0.016 0.006 0.011
rated by parents and teachers, although genetic influ- Informant specific 0.344 0.500 0.108
ences are still significant at this early age. The larger Total 0.360 0.506 0.119
shared environmental effect estimated in twin studies a
Informant-specific shared environment and rater effects cannot
relying on parent or teacher ratings of antisocial be differentiated in caregiver reports.
b
behavior may thus be due in part to a form of rater Rater effects are not applicable in child reports.
bias, rather than to a true, shared environmental From Baker et al. [111]. Copyright © 2007 by the American
Psychological Association. Reproduced with permission.
effect. Perhaps most compelling is the finding that

147
 Chapter 13: Antisocial behavior: gene–environment interplay

the latent common factor underlying the three differ- broader or secondary ones (ADHD and substance
ent raters of the child’s antisocial behavior was almost use disorders) are summarized below.
completely explained by genetic factors (96%). This
very high heritability of the latent common factor Antisocial personality disorder
is due to less error of measurement in a construct
Antisocial personality disorder is probably one of the
derived from three raters, and as such could be con-
most extensively researched personality disorders. It
sidered as the relative importance of genetic factors on
is characterized as a pervasive pattern of disregard for
the true score variance underlying antisocial behavior
and violation of the rights of others occurring since
measures across informants. When such true score
childhood or early adolescence. Individuals diagnosed
models are used to assess the underlying externalizing
with antisocial personality disorder must be at least 18
problem factor in a more reliable manner, genetic
years old, and have had a history of conduct disorder
influences generally become most evident.
before age 15. To be diagnosed, symptoms or behav-
iors of the disorder must have been occurring since
Heritability of DSM-IV disorders childhood or early adolescence. These behaviors
related to antisocial behavior include fighting, setting fires, running away from
home, cruelty to animals, and conflicts with authority
As noted earlier, antisocial and aggressive behaviors
figures [1].
are fundamental in the diagnoses of several psychi-
Estimates of antisocial personality disorder in the
atric disorders. At the symptom level, antisocial per-
community are thought to be 3% [1]. In forensic
sonality disorder in adulthood includes behaviors
settings, the prevalence ranged from about 3 to 30%.
such as aggression, impulsivity, irresponsibility, and
The disorder is more prevalent in males compared to
recklessness. Consequently, it becomes difficult for
females [1]. Having a relative that is affected increases
affected individuals to conform to social or cultural
the risk of developing the disorder [7]. In addition,
norms, that is, to keep a job, complete an education,
despite the prevalence of antisocial personality disorder
and have a long-term romantic relationship. During
being lower in females, affected female probands
childhood and adolescence conduct disorder involves
normally have more affected relatives than male
externalizing behaviors including aggression towards
probands. This indicates that females may require
people and animals, destruction of property, dis-
stronger environmental and genetic influences than
honesty, theft, and other serious violations of age-
males to develop antisocial personality disorder [7].
appropriate rules such as truancy. Conduct disorder
The early adoption studies from the 1970s and
tends to be preceded by oppositional defiant disorder
1980s probably provide the most important findings
which is characterized by overly aggressive and defi-
that heritable influences account partly for the devel-
ant behaviors, and a negative and hostile behavior
opment of antisocial and criminal behavior. These
towards authority figures. These disorders are typic-
studies demonstrated that having a criminal birth
ally measured as diagnostic criteria categories; either
parent increased one’s own risk of having a criminal
the individual meets the criteria for the disorder, or
conviction as an adult, regardless of if the person was
not. In addition to these disorders for which antisocial
reared by pro-social and law-abiding but genetically
behavior is symptomatic, there are several other DSM
unrelated foster parents. This finding was replicated
disorders which may be considered as antisocial in
in adoptive cohorts across different populations, includ-
the broader sense (e.g. drug use), or which are often
ing Scandinavia [8, 9] and the United States [10]. Based
comorbid with conduct disorder and oppositional
on these adoption studies, the genetic effect on crim-
defiant disorder in children, and antisocial personality
inal outcomes appears important for both sexes,
disorder in adults. These additional disorders include
although individual genetic risk is typically more
substance use disorders and ADHD. The influence of
extreme for female than male offenders [11].
genetic and environmental factors in these different
disorders has been examined in both twin and adop-
tion studies. Behavior genetic studies of the three Conduct disorder
primary disorders related to antisocial behavior (anti- The characteristic feature of conduct disorder in chil-
social personality disorder, conduct disorder, and dren and adolescents is a repetitive and persistent
oppositional defiant disorder), as well as the other pattern of behaviors including aggressive behaviors

148
 Chapter 13: Antisocial behavior: gene–environment interplay
causing physical harm to people and/or animals, non-
aggressive behaviors causing property loss or damage,
deceitfulness, theft, and violation of rules. Conduct
disorder is mainly diagnosed in individuals 18 years
or younger. Conduct disorder can be further divided
into two subtypes: childhood-onset type and adoles-
cent-onset type. The childhood-onset type requires
an onset prior to the age of 10, and these individuals
are generally characterized as frequently being aggres-
sive towards others and having disturbed peer
relationships. Compared to the childhood-onset type,
adolescent-onset individuals tend to be less aggressive
and to have more normative peer relationships [1].
The prevalence of conduct disorder for children
and adolescents ranges between 1 and 15% in commu-
nity samples [12, 13] and makes it one of the most
common child psychiatric disorders. A general finding
is that conduct disorder is more common among boys
[14, 15]. There is also an increasing prevalence with
age, in both sexes, especially in the mid-teens [16].
Conduct disorder is also associated with drug
abuse and dependence [17, 18] and children with
conduct disorder are more likely to later develop
other psychiatric disorders, including depression and
anxiety [19]. Most importantly, numerous studies
have shown that conduct disorder is one of the
strongest predictors of antisocial personality disorder
in adulthood [20, 21]. This association is particularly
strong in combination with having a biological parent
with antisocial personality disorder [22].
Results from twin studies suggest genetic factors
contribute to the development of conduct problems
in children [23]. A recent study showed that conduct
disorder symptoms are fairly heritable, with estimates
ranging between 27 and 78% [24]. Further, shared
environmental factors also seem to be of importance
in explaining the development of conduct disorder
[25, 26]. Shared environmental influences refer to
nongenetic influences that contribute to similarity
within pairs of twins. Shared environmental risk
factors for conduct disorder may for example include
family-related factors (e.g. poor child rearing practices,
maltreatment [27–29]) and contextual factors (e.g.
neighborhood disadvantage and poverty [30–32]).
Oppositional defiant disorder
Oppositional defiant disorder is also a childhood-
onset psychiatric disorder, and is characterized by a
pattern of disobedient and hostile behavior towards
authority figures. Oppositional defiant disorder includes
behaviors such as lashing out at adults, losing one’s
temper, and intentionally annoying others [1]. The
prevalence of oppositional defiant disorder varies
depending on age and tend to decrease across age: in
children 6–8 years of age the prevalence ranges from
1.1 to 13.3%, in children 8–14 years of age the preva-
lence ranges from 1.7 to 6.9%, and in children 14–17
years of age the prevalence ranges from 1.0 to 5.8%
[16]. Results from twin studies have shown that
genetic, as well as shared environmental influences
are important in the development of oppositional
defiant disorder symptoms, explaining approximately
one-third each of the total variance [25, 26, 33, 34].
Attention-deficit hyperactivity disorder
ADHD is one of the most common childhood psychi-
atric disorders. The worldwide prevalence of ADHD
was recently estimated in a meta-analysis to be around
5.29% in children and adolescents [35]. It is a disorder
with two separate clusters of symptoms: hyperactivity–
impulsivity, including excessive activity and impulsiv-
ity; and inattention including difficulties in sustaining
attention, distractibility, lack of task persistence, and
disorganization [1]. Although DSM-IV recognizes
either high levels of hyperactivity–impulsivity or high
levels of inattention, most affected individuals exhibit
high levels of both types of symptoms. ADHD tends
to persist across development, but whereas symptoms
of hyperactivity and impulsivity decrease with age,
symptoms of inattention tend to persist. The preva-
lence of ADHD in adulthood has been estimated to be
4.4% [35]. The disorder is associated with academic
underachievement, substance use and dependence,
social maladjustment, and antisocial behavior [36].
Childhood ADHD is related to antisocial behavior
in two different ways. First, several studies have
shown childhood ADHD is associated with a later
onset of antisocial behavior in adolescence and adult-
hood [37, 38]. Second, individuals having a combin-
ation of ADHD and conduct disorder have been
found to have a worse outcome, compared with
individuals having either ADHD or conduct disorder
[39]. It has also been suggested that children with
both types of disorders manifest more severe forms
of antisocial and aggressive behaviors [39].
Results from twin and adoption studies provide
strong evidence suggesting a genetic basis for ADHD
symptoms [40]. The heritability of ADHD symptoms
149
 Chapter 13: Antisocial behavior: gene–environment interplay
have been found to be approximately in the range
of 55–80% [41–43]. A general finding is that shared
environmental factors are of minor importance for
the underlying symptoms of ADHD, whereas the
nonshared environment explains some part of the
variance, around 15–25% [42]. The relative lack of
shared environmental influences is consistent with
findings for antisocial behavior in general [5]. Non-
shared environmental factors may include unique
individual experiences, and possibly biological factors
that may not be genetic in origin. Such nonshared
environmental factors refer to experiences that may
affect one twin in a pair but not the other, for example
head injury or differences in parental treatment.
In sum a vast majority of twin and adoption
studies provide consistent evidence for a strong gen-
etic contribution to ADHD; environmental factors
seem to be largely nonshared. In addition, several
studies have reported comorbidity among ADHD,
conduct disorder and oppositional defiant disorder
in both epidemiological and clinical samples [16,
44–46]. Studies examining the co-occurrence among
these disorders generally show that these different
disruptive and problem behaviors can be united by
an externalizing higher-order factor [2, 3]. This
higher-order externalizing factor has been found to
be highly heritable [47], explaining about 60% of the
total variance [34].
Substance use and abuse
The relationship of antisocial behavior (including
violence and aggression) with alcohol and other drug
use is well established [18]. Family and twin studies
can be used to understand further the relationship
between antisocial behavior and substance use. The
co-occurrence of substance use and aggression may
be due to shared common risk factors, such as genetic
or temperamental traits, antisocial personality dis-
order, or parental substance use and violence. As with
criminal and antisocial behavior [48], substance
dependence tends to run in families. Most of this
family resemblance is explained by genetic influences
[49]. Additionally, the relative influence of genetic
and environmental influences on substance depend-
ence tends to be the same in both men and women.
However, as with antisocial and criminal behavior [5],
it has also been indicated that women require greater
familial loading, that is, greater genetic propensity
or liability, to express these disorders [50].
150
It should be mentioned that although genetic
influences partly explain the risk for disorders involv-
ing antisocial behavior as well as those involving
substance dependence; this does not necessarily mean,
however, that the same genetic and/or environmental
factors influence both types of disorders. To investi-
gate the genetic and environmental overlap between
substance abuse and other antisocial behavior related
disorders, a genetic informative design is required.
Several twin studies have in fact investigated this.
Different disruptive and problem behaviors within
the disinhibitory spectrum such as substance use,
antisocial behavior, ADHD, conduct disorder, and
impulsive and sensation-seeking personality traits
can on a phenotypic level be united by a common
higher-order externalizing factor [2, 3]. The genetic
influences on a common externalizing factor describ-
ing conduct disorder, substance use, ADHD, and
novelty seeking was found to account for more than
80% of the variation in an adolescent sample [4].
Similar findings were reported in a study linking
antisocial personality disorder, conduct disorder,
alcohol and drug dependence and unconstrained
personality style [47]. Strong heritable influences on
an externalizing factor of antisocial behavior, sub-
stance abuse and conduct disorder were also found
in an adult sample [51]. Consequently, the broader
construct of externalizing behaviors based on mul-
tiple measures shows approximately the same high
level of heritability that has been demonstrated for
a multiple-measure, multiple-informant construct of
broad antisocial behavior [6].
Importantly, it does appear that the same genetic
risk factors leading to substance dependence may in fact
provide risk for a wider range of antisocial disorders,
both in adolescent and adult samples. Given the
common genetic link across these various DSM-IV
disorders pertaining to antisocial behavior and sub-
stance use, it is also important to examine the genetic
underpinnings of some of the personality traits that may
underlie antisocial behavior. We thus turn attention to
the heritability psychopathic personality traits, which
have been studied extensively in twins in recent years.
Heritability of psychopathic traits
Psychopathy is, in its adult and full manifestation,
considered to be a serious personality disorder that
is associated with severe and persistent antisocial and
criminal behavior [52]. The principal view of
 Chapter 13: Antisocial behavior: gene–environment interplay
psychopathy originates from the personality-based
clinical conception of the syndrome in which core
interpersonal affective deficits are the focus, e.g.
superficial charm, lack of remorse, unemotionality
[53]. Its base rate in the general population is not
known, but has been estimated at between 0.5 and
1.0% [54]. Psychopathy is today generally conceptual-
ized as a disorder defined by callous and unemotional
affects, a grandiose and manipulative interpersonal
style, pervasive impulsive and irresponsible behavior,
and criminal offending [54]. The most common
measure used to assess adults is the Psychopathy
Checklist – Revised (PCL-R) [54, 55]. The PCL-R is
a semi-structured interview that consists of 20 items
based on Cleckley’s criteria [53]. The measure was
initially designed to assess the traits and behaviors
of psychopathy in incarcerated settings [54, 56].
Psychopathy is distinct, but related to and overlaps
with the more behavioral-based diagnosis of anti-
social personality disorder [1]. It should be noted that
individuals identified as psychopathic very often
receive a diagnosis of antisocial personality disorder,
but the reverse is not true.
In recent years there has been a growing interest
to study the developmental aspects of this severe
personality disorder [57]. Researchers have attempted
to extend the construct of psychopathy downward
to children and adolescents [58, 59]. So far, the vast
majority of research on psychopathy has been
conducted on male incarcerated samples. Though
important, such research does not provide any infor-
mation regarding the development of psychopathic
traits. To understand the developmental origins of
adult psychopathy it is necessary to use samples
including younger individuals. Further, to understand
genetic and environmental contributions to the devel-
opment of psychopathy genetically informative
samples are required. Male incarcerated samples are
probably also biased in that they include comorbid
cases and only identify the “unsuccessful” (caught)
psychopath. Recently several studies have shown that
psychopathic personality traits are present and can be
meaningfully assessed during adolescence [60, 61]. It is
however important to stress that the purpose of assess-
ing psychopathic personality traits in youths is not to
assign a formal diagnosis of psychopathy, but rather
for prevention purposes to identify at-risk adolescents.
There are a few twin studies that have examined
the relative influence of genetic and environmental
effects on psychopathic personality traits in children
and adolescents. These studies together suggest that
psychopathic personality shows significant heritability
[62–64]. For example, a study conducted by Larsson
and colleagues [62] using a set of 16–17-year-old
twins, showed that psychopathic personality dimen-
sions, callous–unemotional, grandiose–manipulative,
and impulsive–irresponsible, could be explained by a
common higher-order “psychopathic personality”
factor. Genetic or heritable influences explained as
much as 63% of the total variance in a latent “psycho-
pathic personality” factor, while nonshared environ-
mental (i.e. refers to experiences unique to the
individual) factors explained the remaining 37% of
the variance. Shared environmental influences (i.e.
nongenetic influences that contribute to similarity
within pairs of twins) did not contribute to the
explanation of “psychopathic personality”.
Additional twin studies have shown that a
common genetic factor largely accounts for the pat-
tern of covariation between psychopathic personality
traits and antisocial and delinquent behavior [65].
Blonigen and colleagues conducted a study in 2005
on a set of 17-year-old twins. They found that the
overlap between self-reported psychopathic personality
traits (Fearless Dominance and Impulsive Antisociality
measured using the Multidimensional Personality
Questionnaire) and externalizing psychopathology was
due to common genetic influences [66].
There are also a few longitudinal twin studies
investigating stability of and change in genetic and
environmental influences over time [67]. A recent
study conducted by Forsman et al. [68] used a longi-
tudinal sample including 16–17 and 19–20-year-old
twins. Results showed that three psychopathic person-
ality dimensions, callous–unemotional, grandiose–
manipulative, and impulsive–irresponsible, were highly
stable across time and linked to a higher-order latent
factor (i.e. psychopathic personality factor). Genetic
factors contributed substantially to the stability of this
general higher-order factor, whereas environmental
factors were of little importance [68].
Gene × environment interactions
Gene–environment (G  E) interaction is defined as
a genetic influence on a given phenotype that depends
upon certain environmental factors, or vice versa.
A genetic influence on a specific phenotype can be
moderated by an individual’s experiences or exposure
to certain environments. Likewise, various individuals
151
 Chapter 13: Antisocial behavior: gene–environment interplay
may respond differently to the same environmental
exposure because they have different genotypes. Sev-
eral studies, using different research designs, including
molecular genetic, twin and adoption designs have
shown that genetic influences and environmental risk
factors interact in the development of antisocial and
criminal behaviors.
Evidence of G  E interaction for antisocial and
delinquent behavior has recently been demonstrated
in molecular genetic studies [27]. Caspi et al. [27]
reported that a functional polymorphism in the gene
encoding the neurotransmitter-metabolizing enzyme
monoamine oxidase A (MAOA) moderated the impact
of early childhood maltreatment on the development
of antisocial behavior in males. Maltreated boys
with a genotype conferring high levels of MAOA
expression were found to be less likely to develop
antisocial problems than maltreated boys who had a
genotype conferring low levels of MAOA expression.
The authors concluded that their findings may partly
explain why not all victims of maltreatment grow
up to victimize others. There has so far been a few
replications of this interesting finding [69–71] and
one published failure to replicate [72].
An indirect way to test for G  E interaction is to
use data from adoption studies. The coherence of
research based on adoption samples showed that the
combination of a genetic predisposition (i.e. psycho-
pathology in biological parents) with a high risk
environment (i.e. adverse adoptive home environ-
ment) lead to greater pathology than what would be
expected from either factor acting alone or both in an
additive combination [9, 73, 74]. This was demon-
strated in a study by Cloninger et al. [8]. They studied
the interaction of a genetic predisposition (i.e.
whether the biological parents were criminal) and a
high risk environment (i.e. adverse rearing experi-
ences and adoptive placement) in 862 Swedish men
adopted at an early age. When both hereditary and
adverse environmental factors were present, 40% of
the males were found to be criminal; if only genetic
factors were present 12.1% became criminal; if only
environmental factors were present 6.7% were crim-
inal; and when neither hereditary nor adverse envir-
onmental factors were present 2.7% were criminal.
The fact that the combination of both hereditary
and environmental factors (40%) was higher than
either hereditary (12.1%) or environmental (6.7%)
factors, or hereditary and environmental factors
together (18.8%), indicated that hereditary and
152
environmental factors interact in the development of
antisocial and criminal behavior.
Several twin studies have also reported G  E
interaction [75]. Tuvblad et al. [76] showed that socio-
economic status moderated the influence of genetic and
environmental factors on antisocial and delinquent
behavior. Using a sample of Swedish 16–17-year-old
twins, heritability for antisocial and delinquent
behavior was found to be higher in more advantaged
neighborhoods (37%) compared to the less advan-
taged neighborhoods (1%). In contrast, the shared
environment was higher in the less advantaged neigh-
borhoods (69%) compared to more advantaged
neighborhoods (13%). A similar, but less evident
pattern was found for girls. Raine [77] has argued
for a “social push hypothesis” in which the influence
of biological risk factors are more likely to become
expressed in an environment where an antisocial
individual lacks the environmental risk factors that
push or predispose him/her to behave antisocially.
In contrast, the relationship between antisocial behavior
and biological risk factors will be weaker in adolescents
from disadvantaged backgrounds because the envir-
onmental determinants of antisocial behavior will
camouflage the genetic contribution [78]. An inter-
pretation of these results in view of the “social push
perspective” would suggest that the influences of
genes on antisocial behavior are more expressed in a
socioeconomically advantaged environment where
the environmental risk factors are absent. On the
contrary, genetic factors for antisocial behavior will
be weaker and the shared environment will be more
important in a socioeconomically less advantaged
environment because the environmental risk factors
will conceal the genetic contribution [76].
Unpacking the genetic and
environmental “black boxes”
In spite of the extensive evidence for genetic influ-
ences in antisocial behavior from twin and adoption
studies, the exact nature of either genetic or environ-
mental effects remains largely unknown. As in other
domains of human behavior, the behavior genetic
studies that focus on the heritability of antisocial
behavior are often criticized for their “black box”
approach, since both genes and environment remain
anonymous. Moreover, early behavior genetic studies
failed to specify the exact biological or social mechanisms
that underlie these global genetic and environmental
 Chapter 13: Antisocial behavior: gene–environment interplay
influences. There is a growing body of research,
however, which attempts to unpack the genetic and
environmental black boxes, and understand the
physiological pathways and the specific nature of
individual circumstances or experiences that may dir-
ectly influence antisocial outcomes, or moderate the
genetic influences themselves. We describe here two
primary approaches being used to unpack these black
boxes and understand the biosocial pathways for
antisocial behavior development: These include: (1)
a measured gene/measured environment approach,
in which researchers are attempting to identify spe-
cific genes that increase risk for antisocial outcomes,
along with narrowly defined measures of individual
circumstances and experiences that attempt to flesh
out the nature of environmental influences; and
(2) a measured risk factor approach, in which various
biological and social risk factors for antisocial behav-
ior are studied in a genetically informative design.
The measured gene/measured environment approach
may be used to investigate main effects of both genes
and environment, or to understand more complex
interactions whereby environments may moderate
genetic effects or vice versa. The measured risk factor
approach includes the identification of endopheno-
types for antisocial behavior, and relies extensively
on multivariate genetic models that aim to under-
stand the genetic and environmental mediation of
relationships between measured risk factors and
antisocial behavior.
Gene identification
Rapid advances in molecular genetic techniques have
provided increased interest and practical opportun-
ities to search for specific genes that contribute to
the risk for antisocial behavior and its correlates,
such as impulsivity. This research, nonetheless, is still
in its infancy and there are only a handful of studies
to date that have evidenced associations between high
risk alleles and antisocial outcomes, particularly using
narrower definitions such as criminal offending or
psychiatric disorders for which antisocial behavior
is primary. There have been no genome-wide linkage
studies of antisocial behavior, in particular, although
if one widens the scope of antisocial behavior to
include substance abuse and other correlates such
as impulsivity, molecular genetic studies are much
more extensive and include both association and
linkage studies.
One important and widely cited study demon-
strating a link between a specific genotype and anti-
social behavior is the large Dutch pedigree study by
Han Brunner and colleagues [79]. Several family
members (especially males) who showed a history of
violence and impulsive aggression were found to share
a mutant form of the gene on the X-chromosome that
codes for MAOA. Specifically, MAOA is an enzyme
that metabolizes several neurotransmitters involved
in impulse control, attention, and other cognitive
functions, including dopamine, norephinephrine,
and serotonin [79]. Mutations in the normal MAOA
gene lead to deficient production of the MAOA
enzyme, which in turn disrupts the normal function
of these neurotransmitters, resulting in a wide range of
disorders, including ADHD, alcoholism, drug abuse,
impulsivity, and other risky behaviors [80–83]. Other
studies have found significant associations between
X-linked MAOA alleles with greater numbers of
dinucleotide repeats, and various behavior disorders,
including ADHD, conduct disorder, major depressive
disorder, drug abuse, alcoholism, reward dependence,
and learning disabilities (see [84]). One hypothesis is
that longer repeat allele variations result in greater
disruptions in gene regulation, leading to variation
in MAO function and hence individual differences
in these related phenotypic disorders [85].
This finding of increased aggression being associ-
ated with MAO deficiency produced by a genetic
mutation in the MAOA allele in the Brunner et al.
[79] pedigree coincides with animal research using
knockout strains of mice [86], where the same associ-
ation between MAOA and violent aggression has been
repeatedly demonstrated. Although the main effects
of the MAOA mutation on aggression and violence
have not been replicated yet in any other large human
pedigrees, this genetic defect remains the first such
link to antisocial behavior in humans. One paradox is
that MAOA deficiency is associated with particularly
high levels of the neurotransmitter serotonin, and yet
low serotonin is a very well-replicated neurochemical
correlate of antisocial and aggressive behavior [87].
Furthermore, MAOA-deficient mice show significantly
better fear conditioning [88], yet poor fear condition-
ing is a well-replicated correlate of criminal and psy-
chopathic behavior [87]. These conflicting findings
clearly require further investigation.
One reason for the failure to find main effects
of the MAOA mutation may be due to G  E inter-
actions, whereby the deleterious effects of the mutant
153
 Chapter 13: Antisocial behavior: gene–environment interplay
gene are only observed under certain environmental
circumstances. The idea that environmental factors
may moderate the effects of the MAOA mutations
on aggression could in general be predicted from the
well-replicated finding of G  E interactions in crim-
inal behavior in adults and conduct problems and
antisocial behavior in youth. Caspi et al. [27] did in
fact demonstrate such a G  E interaction, whereby
the functional polymorphism in the MAOA gene was
found to increase the risk for antisocial behavior in
males, but only for those who experienced early child-
hood maltreatment. The fact that the main effect of
the MAOA mutation as found in Brunner et al. [79]
was not found in the Caspi et al. [27] study under-
scores the importance of investigating specific
genetic effects under a variety of environmental cir-
cumstances in order to fully understand the risk for
criminal offending and other antisocial behavior. The
findings of Caspi et al. are particularly intriguing
since it was one of the first studies to illustrate the
well-replicated G  E interaction in criminal behavior
using a measured gene/measured environment
approach [27]. At the same time, one study has shown
that while the MAOA  abuse interaction holds for
whites, it is not observed in African-Americans [89]
and consequently caution should be exercised as
interaction effects may not apply to all subgroups.
Another study using a measured gene/measured
environment approach to understand antisocial out-
comes focused more specifically on the age of onset
for criminal offending. Among those adolescents with
a history of criminal offending, polymorphisms in
genes related to the neurotransmitter dopamine (also
discussed further below) were shown to be associated
with age of first police contact and arrests, but only
for youth from low risk family environments [90].
That is, the individuals at greatest risk for later onset
criminal offending were those with the A1 allelic form
of the DRD2 gene, in combination with favorable
home environments as defined by maternal attach-
ment, involvement, and engagement. Unlike the
Brunner et al. [79] and Caspi et al. [27] studies which
involved the overall risk for engaging in antisocial
behavior, the DeLisi et al. [90] finding involves the
age of onset of first police contact, and not the overall
risk for offending versus not offending. Nonetheless,
other studies have also shown that different forms
of the DRD2 allele show associations with criminal
victimization [91] and age of first sexual intercourse
[92], as well as normal personality variation [93].
154
A gene–gene interaction between DRD2 and DRD4
in predicting conduct disorder in childhood and
criminal offending in adults has also been shown
[91]. Still, the finding of enhanced risk for later onset
criminal offending as a function of high genetic risk
combined with low environmental risk is contrary
to predictions from other developmental models of
antisocial behavior. Moffitt [94] has suggested that
persistent and early-onset antisocial behavior would
be more influenced by genetic factors than late onset,
transient forms of antisocial behavior, yet the converse
pattern was found by DeLisi et al. [90]. Nonetheless,
the DeLisi et al. findings are particularly interesting
in that they also demonstrate a G  E interaction in
criminal offending using specific genetic markers and
well-defined measures of the environment [90].
Gene identification studies showing direct associ-
ations with more narrowly defined antisocial behavior
are still rare. Dick et al. [95] reported a linkage to a
region of chromosome 7, which appears to contain
genes conferring risk to the externalizing spectrum
(including alcohol, drug dependence, conduct disorder,
antisocial personality disorder, novelty, and sensation
seeking). Specifically, the CHRM2 gene could be asso-
ciated with the spectrum of externalizing psycho-
pathology [95].
There is however a more extensive literature
examining specific genetic effects in behavioral and
neurochemical, and other physiological correlates
of antisocial behavior, and for psychiatric disorders
for which antisocial behavior is symptomatic. For
example, numerous studies in both mice and humans
have identified specific genes involved in the produc-
tion and function of the neurotransmitter serotonin
(5-hydroxytryptamine, or 5-HT), which has been
shown to play a role in impulsive and other risky
behaviors such as drug abuse, gambling, and suicide
[96–99]. A recent study reported an association
between genes related to 5-HT function and impulsivity
in children with ADHD [100]. Evidence of an associ-
ation between catechol-O-methyltransferase (COMT)
valine/methionine polymorphism and the development
of antisocial behavior among children with ADHD
has also been found [101] and replicated [102].
Measured risk factors (endophenotypes)
In the absence of measured genes, it is still possible to
unpack the genetic and environmental black boxes
using multivariate studies in a genetically informative
 Chapter 13: Antisocial behavior: gene–environment interplay
A C E
Antisocial
Risk
behavior factor
a c e a c e
Figure 13.2 Multivariate genetic model for antisocial behavior
and a measured risk factor.
design such as a classical twin or adoption study.
More specifically, it is possible to examine the genetic
and environmental influences between different traits
and behaviors that are known to correlate with the
risk of antisocial behavior. In this “measured risk
factor” approach multivariate genetic models are used
to explain sources of genetic and/or environmental
covariance that underlie the associations between risk
factors and antisocial behavior [103]. This is similar
to the “endophenotype approach”, in which research-
ers identify highly heritable traits that show associ-
ations with antisocial behavior [104]. In addition to
estimating the components of genetic and or environ-
mental variance common to both antisocial behavior
and a specific risk factor, multivariate genetic twin
models can also be used to assess the correlation
between genes influencing antisocial behavior and
genes influencing the risk factor or endophenotype.
A large genetic correlation would result if the same
gene(s) influences the risk factor and antisocial behav-
ior; this is referred to as pleiotropy. A multivariate
genetic model involving antisocial behavior and one
risk factor is presented in Figure 13.2. Common gen-
etic and environmental effects (A, C, and E) influence
both the risk factor and antisocial behavior.
To the extent that genes have pleiotropic effects,
that is, to the extent that they influence more than
one phenotype, common genetic influences on the
risk factor and antisocial behavior should result. Close
proximity – “genetic linkage” – of the genes influ-
encing antisocial behavior to the risk factor, or certain
patterns of assortative mating, such as antisocial indi-
viduals paired systematically to mates with extreme
values of the risk factor, or both, can also contribute
to the common genetic factors in Figure 13.2 [105,
106]. Additional genetic influences that are unique to
either antisocial behavior or the risk factor are also
indicated in this model (A: genetic effects; C: shared
environmental effects; E: nonshared environmental
effects for each phenotype). Thus, the total heritability
for a given trait such as antisocial behavior may be
parsed into components that are shared and not
shared with the risk factor, in order to specify more
exactly the nature of the total genetic influences in
antisocial behavior.
Many correlates of antisocial behavior may be
considered as possible endophenotypes, in particular
personality traits such as impulsivity and attention
deficits. Other correlates include both biological and
social risk factors. Key biological risk factors include
hormones, physiological (autonomic) underarousal,
frontal lobe function (and dysfunction), and neuro-
transmitters. Any of these traits may be examined for
common genetic relationships with antisocial behav-
ior, although little work has been done using this
approach to date.
One example of a biological risk factor for antisocial
behavior is autonomic underarousal in antisocial –
including violent – individuals. Antisocial individuals
and criminal offenders have shown a lower resting
heart rate and lower electrodermal (skin conductance)
responsivity both in orienting and fear conditioning
paradigms [107, 108]. The connections of these find-
ings to antisocial behavior have been interpreted in
several ways. For example, low heart rate may reflect
fearlessness, or reduced anxiety [109], lack of social-
ization, poor learning abilities due to cognitive def-
icits or emotional withdrawal or both, or reduced
brain functioning in areas involved in mediating psy-
chophysiological responding [77]. Another explan-
ation is that low arousal may lead to stimulation-
seeking behaviors, including violence, in an attempt
to raise autonomic arousal to optimal levels.
Although low resting heart rate is the best-replicated
biological correlate of antisocial behavior in child and
adolescent populations [110], and while heart rate
has been shown to be at least partially heritable, it
is unclear to what extent cardiovascular underarousal
may be related to antisocial behavior due to a
common genetic link. This question can be addressed
using multivariate genetic analyses as indicated in
Figure 13.2. In our own longitudinal twin study of
antisocial behavior, we found using multivariate ana-
lyses a significant genetic correlation did in fact exist
between heart rate and antisocial behavior. Analyses
155
 Chapter 13: Antisocial behavior: gene–environment interplay
showed that the relationship between low resting
heart rate and antisocial behavior was significantly
and entirely explained by common genetic influences,
although the heritable component of heart rate
explained only a small portion (1–4%) of the substan-
tial genetic variance in antisocial behavior. Despite the
effect size being small, children with low resting heart
rate appear to be genetically predisposed towards exter-
nalizing behavior problems as early as age nine [111].
Environmental mechanisms may also be investi-
gated using the same “measured risk factor” approach
in multivariate genetic models. As shown in Figure
13.2, the common and specific environmental influ-
ences for the risk factor and antisocial behavior may
also be estimated. This approach may be used to help
elucidate the nature of environmental influences
important to antisocial behavior by determining, for
example, the extent to which certain measured social
risk factors and antisocial behavior may have correl-
ated etiologies. Individual social risk factors that have
been identified as being important to antisocial
behavior include various aspects of parenting, such
as harsh discipline and monitoring or awareness of
children’s activities and behaviors. These and other
environmental factors need to be investigated in gen-
etically informative designs to determine the extent to
which their effects may be moderated by individual
genetic predispositions.
Conclusions
Compelling evidence from twin and adoption
research show that at least half of the total variance
in antisocial behavior is due to heritable factors
[5, 94], and the heritability may be as high as 90%
for more refined, true score estimates of antisocial
behavior (e.g. Baker et al. [111]). Nonetheless,
molecular genetic research has identified only a few
specific genes associated with antisocial behavior or
its correlates. Clearly more research is needed to
identify which genes are involved in these heritable
influences. Our understanding of the link between
genes and antisocial behavior also requires further
investigation of the brain processes that underlie
antisocial behavior throughout the lifespan.
The “genes to brain to antisocial behavior” model
hypothesizes that specific genes result in structural and
functional brain alterations which in turn predispose
to antisocial behavior. The environment is also con-
sidered to play an important role in the development
156
of antisocial behavior [112]. The model proposes that
gene abnormalities lead to structural brain abnormal-
ities which result in emotional/cognitive/behavioral
abnormalities, which in turn predispose to antisocial
behavior. There is increasing evidence for brain
impairments in antisocial individuals, with particu-
larly strong evidence for the prefrontal cortex [113].
For example, neurological patients suffering damage
to the ventral prefrontal cortex exhibit psychopathic-
like, disinhibited behavior, autonomic and emotional
blunting, and bad decision-making [114]. This indi-
cates that there is a significant brain basis to antisocial
behavior, and that these neurobehavioral processes
are relevant to understanding the development of anti-
social behavior. The next step is to understand how
brain structural/functional impairments translate into
the cognitive, emotional, and behavior risk factors
predisposing to antisocial behavior. For example, the
amygdala is centrally involved in fear conditioning.
Poor fear conditioning may result in a failure to fully
develop a conscience – a set of conditioned emotional
responses that motivate individuals to desist from
previously punished behavior. Poor conscience devel-
opment is in turn viewed as a predisposition to anti-
social behavior. Similarly, ventral prefrontal damage
results in disinhibited behavior that predispose to
lawless behavior.
Even though the “genes to brain to antisocial
behavior” model argues for a direct causal pathway
from genes to brain to antisocial behavior, the
importance of environmental influences must be
stressed. This is well in line with results from twin
and adoption studies, which consistently show that
environmental influences explain a substantial part of
the total variance in antisocial behavior, explaining
approximately more than 50% [5]. Environmental
influences early in development could directly change
gene expression, in turn altering brain functioning
and resulting in antisocial behavior. Social environ-
ment can interact with genetics and biological risk
factors for antisocial behavior in different ways [77].
Results from adoption studies indicate that antisocial
behavior increases, perhaps even synergistically, when
social and biological risk factors are combined. There
is also replicated evidence that an abnormality in
the MAOA gene interacts with early child abuse in
predisposing to adult antisocial behavior [27].
Evidence of genetic influences on antisocial
behavior does not implicate that individuals exhibit-
ing antisocial behavior are immune or resistant to
 Chapter 13: Antisocial behavior: gene–environment interplay

interventions. The importance of genetic influences
implies that biological processes are involved in the
etiology of antisocial behavior. However, it does not
imply that a genetic influence on antisocial behavior
requires a biological intervention, but instead it may
require an environmental intervention. For example,
a genetic liability to antisocial behavior may be best
prevented or treated through, for example, parental
and/or teacher training. Further, poor nutrition in the
first three years of life has been associated with long-
term antisocial behavior throughout childhood and
late adolescence [115]. Simple prevention programs
that manipulate nutrition early in life have resulted
in reduced delinquency [116] and criminality [117].
Environmental manipulations can in theory reverse
brain risk factors for antisocial behavior.
Understanding the complex interactions between
specific genes and measured environments will pre-
sumably give rise to the most effective prevention and
treatment courses for these antisocial disorders. As in
other genetic disorders – such as phenyletonuria
(PKU), which is entirely treated through an environ-
mental (dietary) intervention – it is entirely possible
that treatment avenues for genetically based psychi-
atric disorders could potentially involve noninvasive,
environmentally based methods which are tailored to
individuals based on their measured genetic risk.
In summary, antisocial behavior underlies the
diagnosis of several psychiatric disorders, for which
the genetic effects are clearly established. The substan-
tial genetic influence on antisocial and aggressive
behavior itself may be at the root of the comorbidity
among various externalizing disorders, including
conduct and oppositional disorders and ADHD in
children, and antisocial personality disorder and sub-
stance abuse in adults. Future research with combined
approaches from behavior genetics and neuroscience
will lead to better understanding of specific genes that
result in structural and functional brain impairments
that in turn give rise to antisocial, violent, and psy-
chopathic behavior. As in other domains of human
behavior, a major challenge will be to understand the
processes involved in gene expression, including ways
the environment may exacerbate or protect individ-
uals from risk. Longitudinal, genetically informative
studies that involve sophisticated neuroscience
methods such as brain imaging will undoubtedly be
the key to understanding individual pathways toward
psychiatric outcomes for which antisocial behavior is
involved.

References 9. Mednick SA, et al. Science

1. American Psychiatric Association.
Diagnostic and Statistical Manual of
Mental Disorders, Fourth Edition,
Text Revision (DSM-IV-TR).
Washington, DC: American
Psychiatric Association; 2004.
2. Krueger RF, et al. J Abn Psychology
2007;116(4):645–666.
3. Krueger RF, et al. J Abn Psychology
2005;114(4):537–550.
4. Young SE, et al. Am J Med Genet B
Neuropsych Genet 2000;96:684–695.
5. Rhee SH, et al. Psychol Bull
2002;128:490–529.
6. Baker LA, et al. J Abn Psychology
2007;116(2):219–235.
7. Cloninger CR, et al. In Mednick
et al. (eds.). The Causes of Crime:
New Biological Approaches.
Cambridge: Cambridge University
Press; 1987.
8. Cloninger CR, et al. Arch Gen
Psychiatry 1982;39:1242–1247.
1984;224:891–894.
10. Cadoret RJ, et al. Behav Genet
1983;13:301–310.
11. Baker LA, et al. Behav Genet
1989;19:355–370.
12. Costello EJ, et al. Arch Gen
Psychiatry 2003;60(8):837–844.
13. Loeber R, et al. J Am Acad Child
Adolesc Psychiatry 2000;39(12):
1468–1484.
14. Costello EJ, et al. Arch Gen
Psychiatry 1996;53:1129–1136.
15. Hipwell AE, et al. Crim Behav
Mental Health 2002;12(1):99–118.
16. Maughan B, et al. J Child Psychol
Psychiatry 2004;45(3):609–621.
17. Miles DR, et al. Am J Med Genet
2002;114:159–618.
18. White HR, et al. Psychol Addict
Behav 2001;15(3):210–218.
19. Kim-Cohen J, et al. Arch Gen
Psychiatr 2003;60(7):709–717.
20. Loeber R, et al. In Loeber R and
Farrington DP (eds.). Serious
and Violent Juvenile Offenders.
Risk Factors and Successful
Interventions. Thousands Oaks,
CA: Sage Publications; 1998.
21. Rutter M, et al. Antisocial
Behavior by Young People.
Cambridge: Cambridge
University Press; 1998.
22. Robins LN. Deviant Children
Grown Up. Baltimore: Williams
and Wilkins; 1966.
23. Scourfield J, et al. Arch Gen
Psychiatry 2004;61:489–496.
24. Gelhorn HL, et al. J Child
Psychol Psychiatry 2005;46(6):
580–591.
25. Burt AS, et al. J Abn Psychology
2001;110(4):516–525.
26. Dick DM, et al. J Abn Child
Psychology 2005;33(2):219–229.
27. Caspi A, et al. Science 2002;
297(5582):851–854.

157
 Chapter 13: Antisocial behavior: gene–environment interplay
28. Cottle CC, et al. Crim Justice
Behav 2001;28(3):367–394.
29. Lipsey MW et al. In Loeber R
and Farrington DP (eds.). Serious
and Violent Juvenile Offenders.
Risk Factors and Successful
Interventions. Thousands Oaks,
CA: Sage Publications; 1998.
30. Beyers JM, et al. J Abn
Child Psychology 2001;29(5):
379–381.
31. Lynam DR. J Abn Psychology
2000;109(4);563–574.
32. Sampson RJ, et al. Science
1997;277:918–924.
33. Eaves L, et al. J Child Psychol
Psychiatry 1997;38:965–980.
34. Tuvblad C, et al. J Abn Child
Psychology 2009;37(2):153–167.
35. Polanczyk G, et al. Am J Psychiatry
2007;164(6):856–858.
36. Spencer TJ, et al. J Pediatric
Psychology 2007;32(6):
631–642.
37. Hechtman L, et al. Am J
Orthopsychiatry 1984;54:415–425
38. Satterfield R, et al. Am J Psychiatry
1982;139:795–798.
39. Lahey BB, et al. J Nerv Ment Dis
1979;167:734–741.
40. Barkley RA. ADHD and the
Nature of Self-control. New York:
Guilford Press; 1997.
41. Rhee SH, et al. J Abn Psychology
1999;108(1):24–41.
42. Rietveld MJ, et al. Am J Med
Genet B Neuropsych Genet 2003;
117(1):102–113.
43. Thapar A, et al. J Am Acad
Child Adolesc Psychiatry 2000;
39:1528–1536.
44. Angold A, et al. J Child Psychol
Psychiatry 1999;40:57–87.
45. Biederman J, et al. Am J Psychiatry
1991;148(5):564–577.
46. Faraone SV, et al. J Am Acad Child
Adolesc Psychiatry 1998;37(2):
185–193.
47. Krueger RF, et al. J Abn
Psychology 2002;111:411–424.
158
48. Farrington DP, et al. Legal
Criminol Psychol 1996;1:
47–63.
49. Hicks BM, et al. Arch Gen
Psychiatry 2004;61:922–923.
50. Merikangas KR, et al. Arch Gen
Psychiatry 1998;55(11):973–979.
51. Kendler K, et al. Arch Gen
Psychiatry 2003;60:929–937.
52. Hare RD. In Gray J, et al. (eds.).
Criminal Justice, Mental Health,
and Politics of Risk. London:
Cavendish Publishing; 2002.
53. Cleckley HJ. The Mask of Sanity.
St. Louis: Mosby; 1941.
54. Hare RD. The Hare Psychopathy
Checklist-Revised (PCL-R), 2nd
edn. Toronto: Multi-Health
Systems; 2003.
55. Hare RD. The Hare Psychopathy
Checklist-Revised Manual.
Toronto: Multi-Health Systems;
1991.
56. Cooke DJ, et al. Psychol Assess
2001;13(2):171–188.
57. Kotler JS, et al. Clin Child Fam
Psychol Rev 2005;8(4):291–325.
58. Andershed H, et al. In Blaauw E
and Sheridan L (eds.).
Psychopaths: Current
International Perspectives. The
Hague: Elsevier; 2002.
59. Forth AE, et al. The Psychopathy
Checklist: Youth Version. Manual.
North Tonawanda, NY: Multi-
Health Systems, Inc.; 2003.
60. Lynam DR, et al. Ann Rev Clin
Psychol 2005;1:381–407.
61. Vitacco MJ, et al. Assessment
2003;10(2):143–150.
62. Larsson H, et al. J Abn Psychology
2006;115(2):221–230.
63. Taylor A, et al. J Abn Child
Psychology 2003;31(6):633–645.
64. Viding E, et al. J Child Psychol
Psychiatry 2005;46(6):592–597.
65. Larsson H, et al. Psychol Med
2007;37(1):15–26.
66. Blonigen DM, et al. Psychol Med
2005;35:637–648.
67. Blonigen DM, et al. J Abn
Psychology 2006;115(1):85–95.
68. Forsman M, et al. J Abn
Psychology 2008;117(3):606–617.
69. Foley DL, et al. Arch Gen
Psychiatry 2004;61:738–744.
70. Kim-Cohen J, et al. Mol Psychiatry
2006;11:903–913.
71. Nilsson KW, et al. Mol Psychiatry
2006;59:121–127.
72. Haberstick BC, et al. Am J Med
Genet B Neuropsych Genet
2005;135B(1):59–64.
73. Bohman M, et al. Arch Gen
Psychiatry 1982;39:1233–1241.
74. Cadoret RJ, et al. Arch Gen
Psychiatry 1995;52(11):916–924.
75. Rowe DC. In Plomin R, et al.
(eds.). Behavioral Genetics in the
Postgenomic Era. Washington,
DC: American Psychological
Association; 2003.
76. Tuvblad C, et al. J Child Psychol
Psychiatry 2006;47(7):734–743.
77. Raine A. J Child Psychol Psychiatry
2002;43(4):417–434.
78. Raine A. J Abn Child Psychology
2002;30(4):311–326.
79. Brunner HG, et al. Science
1993;262(5133):578–580.
80. Von Knorring A-L, et al. Alcohol
Alcohol 1991;26:409–416.
81. Shekim WO, et al. Psychiatry Res
1989;27:81–88.
82. Buchsbaum MS, et al. Science
1976;194:339–341.
83. Gottfries CG, et al. J Neurochem
1975;25:667–673.
84. Comings DE. In Fishbein DH
(ed.). The Science, Treatment, and
Prevention of Antisocial Behaviors:
Application to the Criminal Justice
System Kingston, NJ: Civic
Research Institute; 2000.
85. Gade R, et al. Mol Psychiatry 1998;
3:50–60.
86. Shih JC. Neurotoxicology 82;
25(1–2):21–30.
87. Raine A. The Psychopathology of
Crime: Criminal Behavior as a

Clinical Disorder. San Diego:
Academic Press; 1993.
88. Kim JJ, et al. Proc Natl Acad Sci
U S A 1997;94(11):5929–5933.
89. Widom CS, et al. Biol Psychiatry
2006;1(1):60.
90. DeLisi M, et al. J Crim Justice
2008;36:217–223.
91. Beaver KM, et al. Int J Offender
Ther Comp Criminol. 2007;
51(6):620–645.
92. Miller WB, et al. J Biosoc Sci
1999;31(1):43–54.
93. Munafò MR, et al. Mol Psychiatry
2003;8(5):471–484.
94. Moffitt TE. Psychol Bull 2005;
131(4):533–554.
95. Dick DM, et al. Arch Gen
Psychiatry 2008;65(3):310–318.
96. Åsberg M, et al. In Meltzer HY
(ed.). Psychopharmacology:
The Third Generation of Progress.
New York: Raven Press; 1987.
97. Moore TM, et al. Aggressive Behav
2002;28:299–316.
Chapter 13: Antisocial behavior: gene–environment interplay
98. Roy A, et al. In Emil F, et al. (eds.).
Serotonin in Major Psychiatric
Disorders. Washington,
DC: American Psychiatric
Press; 1990.
99. Virkkunen M, et al. Archiv Gen
Psychiatry 1989;46:600–601.
100. Oades RD, et al. Behav Brain
Funct 2008;4:48.
101. Thapar A, et al. Hum Mol Genet
2005;14:R275–R282.
102. Caspi A, et al. Arch Gen Psychiatry
2008;65(2):203–210.
103. Evans DM, et al. Biol Psychology
2002;61:33–51.
104. Gottesman II, et al. Am J
Psychiatry 2003;1160:
636–645.
105. Krueger RF, et al. Behav Genet
1998;28(3):173–186.
106. Maes HH, et al. Twin Res
Hum Genet 2007;10(1):
136–150.
107. Raine A, et al. Am J Psychiatry
1990;147:933–937.
108. Raine A, et al. Dev Psychol 1996;
32:624–630.
109. Farrington DP. In Raine A,
et al. (eds.). Biosocial Bases of
Violence. New York: Springer-
Verlag; 1997.
110. Ortiz J, et al. J Am Acad Child
Adolesc Psychiatry 2004;43(2):
154–162.
111. Baker LA, et al. Dev Psychopathol
2009;21(3):939–960.
112. Raine A. Curr Direct Psychol Sci
2008;17(50):323–328.
113. Raine A, et al. Soc Cogn Affect
Neurosci 2006;1:203–213.
114. Damasio A. Descartes’ Error:
Emotion, Reason, and the Human
Brain. New York: G P Putnam’s
Sons; 1994.
115. Liu JH, et al. Am J Psychiatry
2004;161:2005–2013.
116. Olds D. JAMA 1999;281:
1377–1378.
117. Raine A, et al. Am J Psychiatry
2003;160:1627–1635.
159
 Learning disabilities
Chapter
14 Shelley D. Smith
Learning disabilities: definitions
and prevalence
Clinically, learning disabilities are generally defined as
a deficit in a given cognitive ability compared to the
“expected” ability based on intelligence, abilities in
other areas, and opportunity to learn. Specific reading
disability (SRD or dyslexia) is the most common
learning disability with a prevalence of about 9% in
American school children [1], and has been defined as
a specific learning disability that is neurobiological in
origin. It is characterized by difficulties with accurate
and/or fluent word recognition and by poor spelling
and decoding abilities. These difficulties typically
result from a deficit in the phonological component
of language that is often unexpected in relation to
other cognitive abilities and the provision of effective
classroom instruction. Secondary consequences may
include problems in reading comprehension and
reduced reading experience that can impede growth
of vocabulary and background knowledge [2].
Although the popular idea of dyslexia is that of visual
confusion of letters, this definition emphasizes that
the auditory (phonologic) component is the primary
cause of reading disability. Furthermore, genetic stud-
ies have suggested that the auditory and visual routes
to reading are not totally distinct pathways since they
show association with the same genes [3, 4]. A number
of studies have established the necessity for phonemic
awareness in reading ability, and recent studies have
also demonstrated that rapid serial naming is also an
important skill for reading (reviewed in [1]). These
endophenotypes, presumably closer to the etiology of
reading problems, appear to be better measures for
genetic studies of reading.
Similarly, specific language impairment (SLI), the
second most common learning disability at about 7%
Principles of Psychiatric Genetics, eds John I. Nurnberger, Jr. and Wade H. Berrettini. Published by Cambridge University
Press. © Cambridge University Press 2012.
160
of 6 year olds [5], is defined as unexplained deficits in
receptive and/or expressive language skills, without
any evidence of deficits in nonverbal IQ, neurological
impairment, or environmental or emotional prob-
lems that could explain the language delays. These
deficits can be seen in some or all of five global
language domains: phonology, morphology, syntax,
semantics, and pragmatics [5–7] (see [8] for review).
There have been several theories regarding the com-
ponent skills necessary for language, and this has
affected the endophenotypes used in genetic studies.
There is growing consensus that nonword repetition
is important, and this has been used as the diagnostic
phenotype in some studies such as those by the SLI
Consortium [9, 10]. Mastering the morphology of
grammar is also a core skill [11], and at least one
study has shown that both may be needed to produce
SLI [12].
An additional language disability, speech sound
disorder (SSD), concerns developmentally inappro-
priate articulation problems that interfere with intel-
ligibility. The prevalence is about 8% [13]. SSD is
sometimes included with SLI, and has also been
divided into a phonologic disorder, considered to be
a central deficit, and developmental dyspraxia (or
childhood apraxia of speech), which is a motor dis-
order confined to articulation [14]. It is not clear that
these are etiologically distinct disorders, however.
Overall, SSD shares a phonological component with
reading disability, but also includes an articulation
component.
These definitions and subtypes raise a number of
questions. All of these disorders occur on a con-
tinuum of severity, so the determination of an appro-
priate cut-off for diagnosis is arbitrary and may be
tied more to the requirements for eligibility for ser-
vices than to qualitative difference between affected

and unaffected children. Similarly, the designation of
“specific” reading or language disability may not
reflect a qualitative difference in deficits; for example,
the nature of the reading problems of children with
IQs in the normal range may not be different than
those of children whose (lower) IQs are similar to
their reading achievement. There may be etiological
differences across IQ, however, especially when gen-
etic causes are considered. Another question regards
the use of diagnostic categories to determine under-
lying disability. Diagnosis of a particular learning
disability or its subtype may be helpful in classifying
deficits at a point in time, targeting them for reme-
diation, but a diagnostic category may not stay the
same over time, and thus may not reflect the under-
lying etiology. Rather than being separate and dis-
tinct, learning disabilities are often comorbid in
children, suggesting that there may be overlaps in
deficits and etiologies. Through the identification of
genes influencing processes that are important to
learning, new developments in genetic analysis may
help define alternate ways of conceptualizing different
types of learning disabilities based on the genes and
endophenotypes that are involved in each one. Since
reading disability has been studied in depth, it can be
used as an illustration of what is known about the
phenotypic, neurodevelopmental, and genetic aspects
of learning disability.
Reading disability
Gender differences: Studies of clinical populations
have consistently shown male : female sex ratios of 3–
5 : 1, but at least some of the increase in males appears
to be bias of ascertainment since population-based
studies showed sex ratios closer to 1–2 : 1 [15], leading
to the suggestion that boys tended to be more disrup-
tive and thus more often referred for treatment. Stud-
ies of a large twin population indicated that heritability
was different for males than for females [16], but a twin
population with more detailed behavioral testing did
not support this conclusion [17]; rather, this study has
found evidence that males showed greater variation in
severity than females [18]. Since sex-linked genes do
not appear to play a prominent role in reading disabil-
ity (RD), these differences could be due to some sex-
influenced effects of autosomal genes.
Genetics: There has been evidence since the turn
of the last century that RD occurs in families, and
twin studies have consistently shown heritabilities
Chapter 14: Learning disabilities
around 0.56 [19–20]. Some authors have suggested
that this is due to a combination of many genes; in
particular, Plomin and Kovas [21] have developed a
“generalist gene” hypothesis, which proposes that
there are many genes that influence overall cognitive
ability and that learning disabilities result when
enough of these genes are disrupted in their function.
The exact nature of the learning disability is then
determined by environmental factors acting in com-
bination with genetic susceptibility. In contrast, other
studies have suggested that there is a finite number of
genes involved, and that different genes may have
different phenotypic effects. Several types of studies
have supported the idea that there are a smaller
number of genes influencing RD. Segregation ana-
lyses have been consistent with the existence of a
single major gene effect with reduced penetrance,
although this would not necessarily be the same gene
in different families, and the decrease in penetrance
would presumably be due to interaction with add-
itional genes [22]. Another type of segregation analy-
sis estimated 2–5 genes affecting different aspects of
reading [23, 24]. This distinction between a few or a
large number of genes has strong implications for
studies to try to identify these genes; if there are very
many genes with individually small effects, very large
study populations with thousands of cases will be
needed to detect any one of the genes, but if there
are a smaller number of genes, smaller studies will be
sufficient.
Gene localizations: The results of linkage and
association analyses support the findings of segrega-
tion analysis, in that studies of relatively small popu-
lations of families have resulted in linkages and
associations implicating at least nine loci. These loci
are indicated by the prefix DYX and are numbered in
the order of their publication: DYX1, on chromosome
15q21; DYX2 at 6p22; DYX3 at 2p16-p15; DYX4 at
6q13-q16.2; DYX5 at 3p12-q13; DYX6 at 18p11.2;
DYX7 at 11p15.5; DYX8 at 1p36; and DYX9 at
Xq27.3. In addition, several other loci have been pro-
posed: 2p11, 2q22.3, 12p13.3, 13q12, and 13q21. Some
of these loci have been replicated in independent stud-
ies, particularly DYX1, DDYX2, DYX3, DYX5, and
DYX8.
Candidate genes for RD
Several candidate genes have been identified within
DYX loci. While there have been conflicting results
161
 Chapter 14: Learning disabilities
for nearly all of the candidates, with some studies
replicating an association and other studies finding
lack of support, several genes have consistent support
which is backed up by functional analyses.
DYX1: 15q21, DYX1C1: The first region which
showed linkage to RD phenotypes was on chromo-
some 15 [25–27]. Subsequently, a family was identi-
fied in which RD segregated with a translocation with
breakpoints in the linkage region. One breakpoint
disrupted a gene of unknown function, and single
nucleotide polymorphisms (SNPs) within the gene
showed association in a separate population [28]. This
gene has been given the name DYX1C1 (DYX1 Can-
didate 1), and has also been called EKN1. Independent
follow-up studies initially failed to replicate the asso-
ciation of RD with the same SNPs [29–31], but others
found association with different SNPs in the gene [32,
33]. Subsequently, additional studies replicated the
association [34–36]. In the Italian population that
initially rejected association with RD, later studies
by Marino et al. [37] found association with DYX1C1
with short-term memory.
One of the early questions in the genetic studies of
RD was whether different genes would affect different
component phenotypes. Examples of these phenotypes
were phonemic awareness, phonological coding, ortho-
graphic coding, and word recognition. One linkage
study of DYX1 suggested that the primary influence
of this locus was on word recognition, but an editorial
accompanying that paper cautioned against conceptu-
alizing these components as separate brain functions,
particularly since the abilities were highly correlated
[38]. Subsequent studies verified this caution, finding
that the linkage distinctions between component
phenotypes of reading could not be made [39]. On
the other hand, the finding of linkage of short-term
memory to DYX1C1 may indicate that there could be
differences in effects on related cognitive abilities.
The first functional study of DYX1C1 examined
the effects of knockdown of gene expression in brain
development in the rat and demonstrated delays in
neuronal migration [40]. These results are reminis-
cent of the early anatomic studies of postmortem
human brains which showed cortical heterotopias
and microgyri indicative of migrational abnormalities
[41]. Interestingly, when treated rats were examined
postnatally, juvenile and adult rats showed cortical
microgyri. Depending on the location of the micro-
gyri, they also showed deficits in auditory processing
and spatial learning [42].
162
Certain motifs in the sequence of DYX1C1 are
consistent with this involvement in neuronal migra-
tion, but another area in the DYX1C1 protein binds
to heat-shock proteins involved in cancers, and the
gene was found to be upregulated in tumors [43, 44].
These apparently unrelated functions may be recon-
ciled by a recent report that the gene has estrogen-
receptor binding sites [45]. Since hormone receptors
are known to influence brain development, it is
tempting to hypothesize that this gene could influence
the gender differences that have been observed in RD.
DYX2: 6p22, DCDC2, KIAA0319: Linkage of RD
to the 6p22 region was one of the first replicated
linkages of a behavioral trait, and was a strong
impetus to the expansion of molecular genetic studies
of RD [27, 46–49]. Subsequent association analyses
narrowed the region to a set of about five genes; VMP,
DCDC2, KIAA0319, THEM2, and TTRAP [3, 4] with
most results centering around DCDC2 and KIAA0319
[50–53]. The functions of these two genes are not
clearly established, but the homology of DCDC2
(doublecortin 2) to the X-linked lissencephaly gene
DCX (doublecortin) made it an obvious candidate. In
the X-linked recessive disorder, males who are hemi-
zygous for mutations in the DCX gene have a severe
disruption of neuronal migration with an essentially
smooth cortex or sub-cortical heterotopias resulting
from a defect in microtubule stabilization [54, 55;
reviewed in 56]. This is associated with significant
cognitive deficits and a severe seizure disorder. As
would be expected, heterozygous females have a
somewhat milder phenotype, but still show cortical
abnormalities and cognitive deficits. Based on this,
one could postulate that less severe mutations involv-
ing the autosomal DCDC2 gene might result in less
severe neuronal migration defects and a less severe
cognitive outcome. This has been supported by the
finding that miRNA knockdown of DCDC2 in develop-
ing rat brain results in delays in neuronal migration,
similar to that seen in DYX1C1 [52]. In addition,
sequencing of DCDC2 in disabled readers has not
produced obvious causal disruptions in the coding
sequence, but a set of variants and a deletion in the
second intron has been associated with RD [52]. At
least one of these includes a transcription factor binding
site. Support for association of the intronic deletion
with RD has been conflicting, however [34, 57, 58].
Support for association of KIAA0319 with RD
has been a little more consistent, particularly with
SNPs in the 5’ and first intronic regions of the gene

[4, 50, 51, 57, 59]. Association has also been found
with SNPs in KIAA0319 and variation in reading abil-
ity in the normal range [60]. As with DCDC2, sequen-
cing has not found mutations in the coding region,
and the positions of the associated SNPs in the intronic
and 5’ regions suggest that the effect is in gene regula-
tion [61]. Indeed, one study showed that individuals
with a “risk haplotype” of SNP alleles had decreased
expression of KIAA0319 in lymphoblastoid cells [51],
and more recently a particular SNP in a regulatory
region of the gene has been associated with decreased
gene expression [59]. As with DYX1C1 and DCDC2,
miRNA knockdown experiments in developing rat
brain have shown that decreased expression of this
gene also produces delays in neuronal migration [51].
Some studies of RD find association primarily with
DCDC2 and others primarily with KIAA0319. If both
are real, the differences between studies could be due to
population stratification or in diagnostic differences. It
seems most likely that the small populations in most of
the studies have magnified nonsignificant differences
in association, but it is interesting that the studies that
find evidence for DCDC2 and not KIAA0319 have
tended to be the studies from Germany, with a focus
on spelling ability, and those with stronger evidence
for KIAA0319 have tended to be English and focused
on reading phenotypes. An exception is the Colorado
RD twin population, which found associations in both
genes [3]. Recently, in a study of children ascertained
for attention-deficit hyperactivity disorder (ADHD),
association of ADHD was found with DCDC2 and the
adjacent VMP gene but not with KIAA0319. This
population did not show association of RD phenotypes
to either gene [62]. Interestingly, Pennington and
Bishop [1] have postulated that rapid serialized naming
may be the common deficit explaining comorbidity
between RD and ADHD, so perhaps this endopheno-
type is a more accurate measure of the phenotype
of the gene. Further studies are going to be required
to determine whether these two genes actually have
different phenotypic effects. Since these differences
may be more quantitative than qualitative and the
phenotypes are correlated, it may be very difficult to
establish differential influences of the two genes with-
out larger populations. A further complicating factor
is that the two genes may interact and may have a
common regulatory system [57]; in that case, mutations
affecting either gene should have similar outcomes.
DYX5: 3p12.3, ROBO1: Linkage of RD to the
centromeric region of chromosome 3 was first reported
Chapter 14: Learning disabilities
in a large family showing an autosomal dominant pat-
tern of inheritance of severe reading disability. Subse-
quently, an adult male was found with reading disability
and a translocation in that region of chromosome 3.
Although his sister had also been diagnosed with dys-
lexia and did not share the translocation, testing of both
individuals was not possible, so the authors considered
it likely that there could be another cause for learning
problems in the sister. The translocation disrupted the
ROBO1 gene, which is homologous to the Robo (round-
about) gene in the fruitfly, mouse and rat and is
involved in the migration of axons across the midline
of the brain and spinal cord. The ROBO1 gene was
sequenced in affected individuals from the large linkage
family, but no mutations were found in the coding
region. A SNP haplotype was identified which marked
the chromosome segregating with RD in the family, and
analysis of allelic expression in lymphoblasts from sev-
eral affected individuals indicated that the risk haplo-
type was underexpressed. Allelic expression of nearby
genes was normal, leading to the identification of
ROBO1 as a candidate gene [63]. This appears to be a
rare form of reading disability, however, since the local-
ization has not been replicated in other families.
Overall, there is evidence that disorders of neuronal
migration affect RD, and Galaburda et al. [64] have
suggested a mechanism whereby DYX1C1, DCDC2,
KIAA0319, and ROBO1 proteins interact in neuron
and axon guidance. Several other candidate genes have
also been suggested for reading disability or related
phenotypes. MRPL19 and C2ORF3 are in the region
2p12, which may be the locus responsible for the linkage
signal for DYX3. The original linkage was to markers
around 2p16–15, however, so it is possible that these are
actually separate loci. Association was found to SNPs
very close to these two genes in Finnish and German
families, and it was hypothesized that they are jointly
regulated [65]. It is not known how these genes might
influence RD; MRPL19 is a mitochondrial ribosomal
protein and the function of C2ORF3 is unknown. In
another study, SNPs in SLC2A3 (GLUT3) at 12p13.3
was found to be associated with mismatch negativity in
dyslexic children in a genome-wide association study
(GWAS) of 200 German families. This region had not
been detected in previous linkage studies [66]. The gene
product is known to be a glucose transporter, and thus
could represent a mechanism besides neuronal migra-
tion that can affect RD. The use of an electrophysiologic
endophenotype is also an important development in the
investigation of the causal mechanisms.
163
 Chapter 14: Learning disabilities
Language impairment
Although the definitions of RD and language impair-
ment (LI) are based on measures that are very differ-
ent, the disorders share some similarities; the
heritabilities for deficits are similar [67], males are
affected more often than females, and young children
with LI are at greater risk for RD [1, 68]. Although
limited studies of gene segregation have been done,
analyses have not found the evidence for major genes
[69] that have been seen for RD. Despite this,
genome-wide linkage studies were done and resulted
in localizations of gene effects. Those loci were not in
the same regions as RD loci, suggesting that that
language impairment may have a separate etiology.
The largest linkage studies were done by the SLI
Consortium [9, 10], and resulted in two loci, SLI1
on 16q24 and SLI2 on 19q13. Follow-up studies
narrowed the regions on each chromosome and char-
acterized the phenotypes [70]. The SLI1 locus showed
association with phonological short-term memory,
while the SLI2 locus showed association with expres-
sive language, a grammatical measure of the ability to
use past tense, and literacy. Phonological short-term
memory, often measured by nonword repetition, and
grammatical measures such as tense marking [11]
have been used as primary component phenotypes
for LI, and earlier genetic studies indicated that they
might be under separate genetic control [12]. More-
over, based on their analysis of the published litera-
ture, Pennington and Bishop [1] postulated that
deficits in both abilities are necessary for SLI.
The association of reading measures to the
chromosome 19 SLI2 locus also demonstrates that
LI and RD can overlap in etiology, despite the fact
that these overlaps were not seen in the genome
screens for each disorder. Evidence for overlapping
genetic influences was also seen in a smaller linkage
study of LI which found linkage to chromosome
13q21 in 5 extended families with LI, but the linked
phenotype was a reading measure rather than a lan-
guage measure [71]. Finally, linkage studies that
targeted RD loci in families ascertained for LI repli-
cated the linkage to the DYX8 and DYX2 regions, and
association analysis in these LI families replicated
association of language and reading phenotypes to
several of the same SNPs in KIAA0319 that showed
association with RD [72]. It appears, then, that there
are loci that affect both RD and SLI, but the magni-
tude of the effect differs; the DYX loci have greater
164
influence on reading with lesser influence on LI, so
that their association was missed in genome-wide
studies for LI. Similarly, the SLI1 and SLI2 loci have
primary effects on LI phenotypes, but the SLI2 locus
also has a lesser effect on reading which was not
detectable in genome scans for RD.
Candidate genes for LI
SLI1: 16q23–24, CMIP; 16q24.1, ATP2C2: Two
genes have been cited as independent candidates for
the SLI1 locus, CMIP and ATP2C2 [73]. Consistent
with the linkage results, the primary phenotype that
showed association with SNPs in those genes was
phonological short-term memory (nonword repeti-
tion), and there was no evidence of association with
reading measures. Association was found in two inde-
pendent populations selected for LI, but was not
found in an unselected population, indicating that
the genes may not affect variation of nonword repeti-
tion in the normal range. The mechanism of effect on
LI for these genes is unknown, although the functions
of these genes allow some speculation. Mutation of
the CMIP gene can cause a renal disorder called
minimal change nephritic syndrome which results
from truncation of the gene product [74], but the
gene is also known to be expressed in the brain and
additional functions are being discovered. The gene
product has been found to interact with filamin-A
[75], which suggests a role in the formation of den-
dritic spines, and is involved in the control of NF-kB
[76], which is also important in synaptic activity [73].
The protein produced by ATP2C2 is a transporter of
calcium and manganese ions and localizes to intracel-
lular vesicles, where it is thought to affect calcium
homeostasis necessary for neuronal and synaptic
function. Abnormalities in magnesium ion homeo-
stasis have been associated with neurodegenerative
disorders and memory problems [73].
SPCH1: 7q31, FOXP2: The influence of the
FOXP2 gene on speech and language was discovered
through a unique kindred known as the KE family in
which severe language impairment and oromotor
apraxia segregated as a dominant trait, found to be
due to a missense mutation in the FOXP2 gene [77].
A few additional children with developmental verbal
dyspraxia have been identified with disruptions (either
translocations or mutations) of FOXP2 [77, 78], but
studies of larger LI populations have indicated that
mutations in the gene are not a common cause of LI.

Most studies have looked within the coding region,
however, and some evidence has been found for link-
age or association near the gene [72, 79], so changes in
regulatory regions cannot be ruled out. More import-
antly, as a gene regulator itself (forkhead-winged helix
transcription factor), it can be a key to a network of
genes that affect language [80]. The findings with the
gene CNTNAP2, discussed below, are an example.
Structural MRI characterization of the KE family
demonstrated that the gene mutation resulted in a
25% decrease in volume of the caudate nuclei, and
the extent of decrease correlated with the degree of
oral dyspraxia [81]. Abnormalities were also noted in
cortical regions. Language and oral motor areas,
including Broca’s area and the ventral portion of the
precentral gyrus, showed decreased gray matter, and
these differences were reflected in decreased activation
in these areas seen with fMRI. Overall, the deficits
appeared to be in frontostriatal and frontocerebellar
networks subserving language and oromotor skills.
Studies of the function of FOXP2 also illustrate an
important evolutionary principle: language ability is
likely to be built on existing neurodevelopmental
systems, rather than on new genes that are unique to
humans [80]. The FOXP2 gene is found to be conserved
across species; for example, the homologous gene in
zebra finches is expressed in a similar pattern in the
brain and is involved in the learning of song [82].
Furthermore, decreased expression of the gene caused
deficits in song learning [82]. In mice, knockouts of
FoxP2 or introduction of the human KE family muta-
tion caused severe motor problems and decreased
cerebellar size in homozygotes. Phenotypic effects in
heterozygotes were less clear, possibly depending upon
the genetic background of the mice, indicating that
interaction with other genes is important. Some reports
had indicated that vocalizations of heterozygous pups
were abnormal, but in an extensive review, Fisher and
Scharff [80] have noted that these may be secondary to
other developmental problems. They argue that, rather
than assuming that the gene is important for communi-
cation across species, the evolutionary function of
FOXP2 could be conceptualized as facilitating synaptic
plasticity for planning and coordination of very rapid
movements. Such abilities would be fundamental to the
production of words, and changes affecting this type of
gene could be the basis for the evolution of language.
While such changes have not been identified yet, two
“human specific” amino acid substitutions have been
noted in FOXP2, and when introduced into mice, these
Chapter 14: Learning disabilities
substitutions resulted in changes in the basal ganglia
and increased synaptic plasticity as well as some changes
in ultrasonic vocalizations [83].
CNTNAP2: The CNTNAP2 gene is located distal
to FOXP2 on 7q35. It was first identified as candidate
for autism through genome screens and expression
studies [84, 85]. In the study by Alarcón et al., age
at first word was used as a quantitative phenotype.
The gene has several possible functions in neuronal
migration and synaptic function, and mutations in
CNTNAP2 have been associated with mental retard-
ation, severe epilepsy, and cortical dysplasia [86, 87].
In an independent study designed to identify a genetic
network influencing LI, Vernes et al. [88] identified
CNTNAP2 as a downstream target of FOXP2. They
demonstrated significant association between nine
intronic SNPs in the gene and performance in a non-
word repetition task in families with a child with LI.
Further studies will undoubtedly expand upon the
function of the FOXP2–CNTNAP2 network.
Speech sound disorder
Speech sound disorder (SSD) involves both language
and articulation deficits and is often grouped with LI.
It is also often comorbid with RD. Pennington and
Bishop [1] have pointed out that the comorbidity
between SSD and RD appears primarily when SSD is
included with LI. In their analysis of published results,
the relative risk for RD in children with earlier SSD þ
LI ranged from 4.6 to 8.9, but the rate of RD in
children with SSD alone was negligible. Unfortu-
nately, published data for RD in children with LI
without SSD was inconsistent; in one study, the rate
of RD was 3.2–3.6, while the other LI study had a rate
of RD of only 0.5. Thus, SSD without LI may be
separate from SSD + LI. This is supported by a study
which showed that the heritability of LI was largely
driven by the heritability of LI + SSD [89].
Molecular genetic studies support common gen-
etic influences on RD and SSD. Linkage analyses
have demonstrated linkage of SSD phenotypes to
the DYX5/ROBO1 region on chromosome 3 [90],
to the DYX2 region on chromosome 6 and to the
DYX1C1 region on chromosome 15 [91] as well as
to a more proximal region on 15 [92], and to the
DYX8 region on chromosome 1 [93]. These linkages
were found with articulation phenotypes and phono-
logical phenotypes, but it is unknown whether these
results were driven primarily by children who later
165
 Chapter 14: Learning disabilities

developed RD. More fine-grained association studies Such studies could have a high payoff. As the genes
have not yet been done for most of these regions, so it is influencing learning disabilities are defined, they will
not certain that the localizations to RD regions mean lead to the neurodevelopmental pathways, such as
that the same genes that affect RD also affect SSD. neuronal migration, involved in reading, language, or
articulation, and may demonstrate where the pathways
Conclusions for these disorders converge and diverge. Genetic stud-
ies also suggest that endophenotypes are more valuable
Studies of RD, LI, and SSD have demonstrated that
than clinical definitions in determining the etiology of
subjects with these conditions have overlapping cog-
these disorders. As the cognitive overlaps between
nitive deficits, and the distinctions between them may
disorders are worked out, e.g. the relative contribu-
depend upon particular combinations of deficits (as
tions of phonology, rapid naming, grammar, or articu-
measured by endophenotypes) that produce greater
lation to each disorder (e.g. [1]), concurrent genetic
liability for one disorder over another [94]. Molecular
studies can confirm which endophenotypes have
genetic studies support this concept, with some genes
common genetic effects and which are distinct. By
appearing to primarily affect one endophenotype,
starting with etiology and working up to clinical
such as the linkage of nonword repetition to SLI1,
phenotype, it is likely that the current diagnostic cri-
some affecting just a few, such as the effects of SLI2
teria may be modified. For example, is LI + SSD etio-
on nonword repetition and literacy, and other genes
logically distinct from LI alone or SSD alone, and
appearing to affect several endophenotypes and sev-
should diagnostic criteria take this into account?
eral disorders, such as the association of KIAA0319 to
Should LI be diagnosed based on nonword repetition
reading and language measures in RD and LI popula-
alone, or must measures of grammar also be included?
tions. To characterize the effects of individual genes
Can RD be diagnosed solely by assessment of phon-
on the endophenotypes and their resulting clinical
emic awareness or decoding, or will the inclusion of
manifestations, studies will need to be done with the
rapid serial naming result in dignosis of a more com-
same phenotypes and genotypes in all three disorders,
prehensive disorder with a distinct etiological path-
with large enough populations to reliably distinguish
way? Better definition of the cognitive components of
any qualitative influences of the candidate genes on
the three disorders should lead to optimal diagnostic
different phenotypes. Given the differences in the ages
and treatment procedures based on deficits that are
at which RD, LI, and SSD are usually diagnosed,
closer to etiology than clinical symptoms alone.
longitudinal studies would be especially valuable.

References Genetics, 4th edn. New York:

1. Pennington BF, et al. Annu Rev
Psychol 2009;60:283–306.
2. Lyon GR, et al. Ann Dyslexia
2003;53:1–14.
3. Deffenbacher KE, et al. Hum Genet
2004;115(2):128–138.
4. Francks C, et al. Am J Hum Genet
2004;75(6):1046–1058.
5. Tomblin JB, et al. J Speech Lang
Hear Res 1997;40(6):1245–1260.
6. Tallal P. In Adelman G (ed.).
Encyclopedia of Neuroscience.
Vol. 1. Boston: Birkhauser; 1987.
7. Leonard LB. Appl Psycholinguistics
1989;10:179–202.
8. Smith SD, et al. In Rimoin DL, et al.
(eds.). Emery and Rimoin’s
Principles and Practice of Medical
Elsevier; 2007.
9. SLI Consortium. Am J Hum Genet
2002;70(2):384–398.
10. SLI Consortium. Am J Hum Genet
2004;74(6):1225–1238.
11. Rice ML, et al. J Speech Hear Res
1996;39(6):1239–1257.
12. Bishop DV, et al. Genes Brain
Behav 2006;5(2):158–169.
13. Shriberg LD, et al. J Speech Lang
Hear Res 1999;42:1461–81.
14. Sanger TD, et al. Pediatrics
2006;118(5):2159–2167.
15. Shaywitz SE, et al. JAMA 1990;264
(8):998–1002.
16. Harlaar N, et al. J Child Psychol
Psychiatry 2005;
46(4):373–384.
17. Hawke JL, et al. Dyslexia 2006;
12(1):21–29.
18. Hawke JL, et al. Dyslexia 2009;
15(3):239–242.
19. DeFries JC, et al. Mental Retard
Devel Disab Res Rev 1996;
2:39–47.
20. DeFries JC, et al. In Plomin R,
et al. (eds.). Nature, Nature, and
Psychology. Washington, DC:
APA Press; 1993.
21. Plomin R, et al. Psychol Bull
2005;131(4):592–617.
22. Pennington BF, et al. JAMA
1991;266(11):1527–1534.
23. Wijsman EM, et al. Am J Hum
Genet 2000;67(3):631–646.
24. Chapman NH, et al. Am J Med
Genet B Neuropsychiatr Genet
2003;121B(1):60–70.

166

25. Smith SD, et al. Science 1983;
219(4590):1345–1347.
26. Smith SD, et al. Reading Writing:
Interdis J 1991;3:285–298.
27. Grigorenko EL, et al. Am J Hum
Genet 1997;60(1):27–39.
28. Taipale M, et al. Proc Natl
Acad Sci U S A 2003;100
(20):11553–11558.
29. Marino C, et al. Eur J Hum Genet
2005;13(4):491–499.
30. Meng H, et al. Hum Genet
2005;118(1):87–90.
31. Bellini G, et al. J Mol Neurosci
2005;27(3):311–314.
32. Wigg KG, et al. Mol Psychiatry
2004;9(12):1111–1121.
33. Scerri TS, et al. J Med Genet
2004;41(11):853–857.
34. Brkanac Z, et al. Am J Med
Genet B Neuropsychiatr Genet
2007;144B(4):556–560.
35. Bates TC, et al. Mol Psychiatry
2010;15(12):1190–1196.
36. Dahdouh F, et al. Psychiatr Genet
2009;19(2):59–63.
37. Marino C, et al. Genes Brain
Behav 2007;6(7):640–646.
38. Pennington BF. Am J Hum Genet
1997;60(1):13–16.
39. Grigorenko EL, et al. Am J Hum
Genet 2000;66(2):715–723.
40. Wang Y, et al. Neuroscience
2006;143(2):515–522.
41. Galaburda AM. Neurol Clin
1993;11(1):161–173.
42. Threlkeld SW, et al. Brain Res Bull
2007;71(5):508–514.
43. Chen Y, et al. J Cancer Res Clin
Oncol 2009;135(9):1265–1276.
44. Kim YJ, et al. J Cancer Res Clin
Oncol 2009;135(2):265–270.
45. Massinen S, et al. Hum Mol Genet
2009;18(15):2802–2812.
46. Cardon LR, et al. Am J Hum Genet
1994;55(4):825–833.
47. Cardon LR, et al. Science 1995;
268(5217):1553.
48. Gayán J, et al. Am J Hum Genet
1999;64(1):157–64.
49. Fisher SE, et al. Am J Hum Genet
1999;64(1):146–56.
50. Cope N, et al. Am J Hum Genet
2005;76(4):581–591.
51. Paracchini S, et al. Hum Mol
Genet 2006;15(10):1659–1666.
52. Meng H, et al. Proc Natl Acad Sci
U S A 2005;102(47):17053–17058.
53. Schumacher J, et al. Am J Hum
Genet 2006;78(1):52–62.
54. des Portes V, et al. Hum Mol
Genet 1998;7(7):1063–1070.
55. Sossey-Alaoui K, et al. Hum Mol
Genet 1998;7(8):1327–1332.
56. Leventer RJ. J Child Neurol
2005;20(4):307–312.
57. Harold D, et al. Mol Psychiatry
2006;11(12):1085–1091.
58. Ludwig KU, et al. Psychiatr Genet
2008;18(6):310–312.
59. Dennis MY, et al. PLoS Genet
2009;5(3):e1000436.
60. Luciano M, et al. Biol Psychiatry
2007;62(7):811–817.
61. Couto JM, et al. Am J Med Genet
B Neuropsychiatr Genet
2010;153B(2):447–462.
62. Couto JM, et al. Biol Psychiatry
2009;66(4):368–75.
63. Hannula-Jouppi K, et al. PLoS
Genet 2005;1(4):e50.
64. Galaburda AM, et al. Nat Neurosci
2006;9(10):1213–1217.
65. Anthoni H, et al. Hum Mol Genet
2007;16(6):667–677.
66. Roeske D, et al. Mol Psychiatry
2011;16(1):97–107.
67. Spinath FM, et al. Child Dev 2004;
75(2):445–454.
68. Catts HW, et al. J Speech
Lang Hear Res 2002;45(6):
1142–1157.
69. Lewis BA, et al. Behav Genet
1993;23(3):291–297.
70. Falcaro M, et al. Genes Brain
Behav 2008;7(4):393–402.
Chapter 14: Learning disabilities
71. Bartlett CW, et al. Am J Hum
Genet 2002;71(1):45–55.
72. Rice ML, et al. J Neurodev Disord
[epub ahead of print].
73. Newbury DF, et al. Am J Hum
Genet 2009;85(2):264–272.
74. Grimbert P, et al. J Exp Med
2003;198(5):797–807.
75. Grimbert P, et al. Mol Immunol
2004;40(17):1257–1261.
76. Kamal M, et al. Mol Immunol
2009;46(5):991–998.
77. Lai CS, et al. Nature 2001;
413(6855):519–523.
78. MacDermot KD, et al. Am J Hum
Genet 2005;76(6):1074–1080.
79. O’Brien EK, et al. Am J Hum
Genet 2003;72(6):1536–1543.
80. Fisher SE, et al. Trends Genet
2009;25(4):166–177.
81. Vargha-Khadem F, et al. Nat Rev
Neurosci 2005;6(2):131–138.
82. Haesler S, et al. J Neurosci 2004;
24(13):3164–3175.
83. Enard W, et al. Cell 2009;137(5):
961–971.
84. Alarcón M, et al. Am J Hum Genet
2008;82(1):150–159.
85. Arking DE, et al. Am J Hum Genet
2008;82(1):160–164.
86. Strauss KA, et al. N Engl J Med
2006;354(13):1370–1377.
87. Zweier C, et al. Am J Hum Genet
2009;85(5):655–666.
88. Vernes SC, et al. N Engl J Med
2008;359(22):2337–2345.
89. Bishop DVM, et al. Genes Brain
Behav 2008;7:365–372.
90. Stein CM, et al. Am J Hum
Genet 2004;74(2):283–297.
91. Smith SD. J Child Psychol
Psychiatry 2005;46(10):1057–1066.
92. Stein CM, et al. Behav Genet
2006;36(6):858–868.
93. Miscimarra L, et al. Hum Hered
2007;63(1):47–58.
94. Pennington BF. Cognition 2006;
101:385–413.
167
 Attention-deficit hyperactivity disorder
Chapter
15 Josephine Elia, Francesca Lantieri, Toshinobu Takeda, Xiaowu Gai,
Peter S. White, Marcella Devoto, and Hakon Hakonarson
Introduction
Attention-deficit hyperactivity disorder (ADHD) is
one of the most common neuropsychiatric disorders
with a best estimate prevalence of 5–10% in school
age children [1], 4% in college students [2] and
 2.5% in adulthood [2, 3]. Relatively consistent
prevalence rates have been reported for North
America and Europe [4].
ADHD is characterized by developmentally
inappropriate levels of hyperactivity, impulsivity and
inattention as well as executive function deficits
leading to significant learning, behavioral and social
impairments throughout the lifespan [5]. Children
with ADHD are reported to have higher rates of
academic, emotional, conduct, and social difficulties
[6]. Compared to non-ADHD peers, young adults
with ADHD have poorer academic outcomes [5, 7],
work performance [8], and self-esteem [9]. In add-
ition they have increased rates of smoking [10], drug
use [11], and automobile accidents [12]. They also
have more social problems including fewer close
friends, more dating partners and shorter duration
of dating relationships, more sex partners; they
become parents earlier (38% versus 4% of controls
by age 25), and have higher rates of sexually
transmitted diseases (16% versus 4%) [5].
Family and twin studies
Family, twin, and adoption studies have unequivo-
cally demonstrated that ADHD is influenced by gen-
etic factors. A higher risk for ADHD has been
reported in siblings of ADHD probands (20.8%
versus 5.6% in controls) [13], among first-degree
family members of ADHD male [14] and female
probands [15, 16], and in second-degree relatives
[17]. Shared genes, rather than shared environment,
Principles of Psychiatric Genetics, eds John I. Nurnberger, Jr. and Wade H. Berrettini. Published by Cambridge University
Press. © Cambridge University Press 2012.
168
were suggested to be responsible for the increased
familial aggregation by adoption studies, reporting
18% ADHD in biological versus 6% in adoptive
parents [18].
Twin studies have attempted to distinguish
between genetic and environmental influences with
reported heritabilities ranging between 60 and 100%
for an overall heritability of 0.76 reported in a
review of earlier studies [19]. Additional recent
studies, summarized in Figure 15.1, estimate herit-
ability at 29–90% [20–27]. The lower heritability
ranging between 29 and 37% have been reported
primarily in studies that included older subjects
[28–33]. Parental rating bias [34, 35] can mask
other nonadditive genetic contributions, leading to
over-estimation of the heritability in younger
cohorts [34, 36].
The rest of the variance is explained by shared
and nonshared environmental influences [20, 23,
37, 38] and non-additive genetic influences, such
as a dominant effects [25] resulting from epigenetic
factors [39]. Environmental effects [40] may
also play a role during the different developmental
stages [41]; therefore, a cohort effect may also be
important.
Genome-wide linkage studies
Summarized in Table 15.1 are the ADHD linkage
studies with logarithm of odds (LOD) scores > 1.5
conducted in sibling pairs [33, 42–48], isolated popu-
lations [49, 50], unrelated extended pedigrees [51, 52],
and affected members of single families [53]. Several
chromosomal regions with potential linkage, some
overlapping in two or more studies have been identi-
fied including 2p, 5p, 7p, 16p, 17p, 5q, 9q, 11q, 12q,
14q, 16q, and 18q. In particular, a pooled analysis of
 Chapter 15: Attention-deficit hyperactivity disorder

Rietveld 2004 3–12y

Laarson et al. 2004 13–14y

Saudino et al. 2005 7y

Price et al. 2005 2y *1

Dick et al. 2005 14y *1

Kuntsi et al. 2005 8y *1

Knopik et al. 2005 11–23y

Ehringer et al. 2006 12–19y

Schultz et al. 2006 middle-age

Haberstick et al. 2008 22y

Ouellet-Morin et al. 2008 6–7y

Derks et al. 2008 12y

*2
Young et al. 2009 12y

Young et al. 2009 17y *2

0.00 0.20 0.40 0.60 0.80 1.00

Heritability
Figure 15.1 Heritabilities reported from 2004–09 for attention-deficit hyperactivity disorder (ADHD). The estimates reported here regard
the ADHD diagnosis, with distinguishing for subtypes. The darker bars refer to parental or teacher reports, while the lighter bars refer to
self-reports. *1 and *2 indicate that the same sample has been investigated at successive follow up (longitudinal studies).

the US and Dutch data identified a single region of Dopaminergic pathways

overlap at 5p13 [54] and a meta-analysis identified a
region with genome-wide significance on chromo-
some 16 between 64 and 83 Mb [55].
Overall, the linkage analysis studies conducted to
date have achieved only limited success in identifying
genetic determinants of common complex diseases;
this may be due to the generic problem that the
linkage analysis approach has generally low power in
identifying common genetic variants that have
modest effects [56, 57].
Selected candidate gene studies
The initial search for ADHD genes was hypothesis-
driven and focused on genes involved in neurotrans-
mission, based on evidence from effective pharma-
cotherapeutic agents [58], animal models [59, 60],
and neuroimaging studies [61]. Numerous studies
have thus far been conducted and a summary of
candidate gene polymorphisms associated with ADHD
identified in the most recent meta-analysis is presented
in Table 15.2.
We will review candidate gene-specific investigations
beginning in 1995 when Cook and colleagues [62]
reported a significant association of ADHD with
SLC6A3–10R (DAT1), a variant of the dopamine
transporter gene, located in the perisynaptic regions
of neurons [63, 64] and involved in recycling
dopamine back into the releasing neurons. SLC6A3
is expressed predominantly in basal ganglia where
it preferentially influences caudate volume [42].
In humans, the 3’ untranslated region (UTR) of the
gene has a 40 base pair tandem repeat polymorphism
(VNTR) consisting of 3–11 copies [65]. As reviewed
by Barr and Misener, [66] and Gizer et al. [67] since
Cook’s initial report, replication studies of SLC6A3–
10R have been inconsistent, and meta-analyses and
analyses of pooled odds ratios are in the range
of 1.03–1.27 suggesting a modest contribution to
disease risk.
Additional polymorphisms that may confer risk
include single nucleotide polymorphisms (SNPs)
rs27072, a VNTR in intron 8 [66, 67]. Recently, a

169
 Chapter 15: Attention-deficit hyperactivity disorder

Table 15.1 Genome-wide linkage studies.

Studies Sample Chromo- ADHD Closest Position LOD Results

(Year) some candidate marker (cM)
gene
Fisher et al. 126 ASP 5p12 SLC6A3 D5S418 68 1.04 None exceeded
(2000) 104 families (USA) 10q26 D10S212 193 1.66 genome-wide
12q24 D12S79– 131 1.16 significance
D12S324 165 1.099 thresholds
12p13 SLC2A1 D12S1725– 10 1.51
D12S336 22 2.60
16p13 D16S3075 14 1.51
Ogdie et al. 270 ASP 5p13 SLC6A3 16p13 (GRIN2A)
(2003, 2004) 204 families 6q14 (GRIN2A) was significant
(160 from Fisher 11q25
et al. [2000]) 16p13
17p11
20q13
Bakker et al. 238 ASP 5p13 SLC6A3 D5S2500 10 1.43
(2003) (106 families; 7p13 DDC D7S1818 3.04
the Netherlands) 9q33 DBH D9S1825 2.05
13q33 D13S796 1.91
15q15 GATA50C03 3.54
Loo et al. 283 ASP 16p D16S3046 18 3.73 Reading disability
(2004) (expanded 17q D17S787 46 2.98 and ADHD
sample from 10q D10S196 59 1.26
Ogdie et al. 2004)
Ogdie et al. 308 6q12 D6S430 83 3.30 Suggestive
(2004) (269 from Bakker 16p13 D16S3114 23 3.73 evidence for
et al. 2003) 17p11 D17S947 38 3.63 5p13
Arcos-Burgos 375 4q13.2 D4S409 81.6 2.4 Significant linkage
et al. (2004) (Columbian 5q33.3 D5S2117 138.5 1.5 for all families
genetic isolate 8q11.23 D8S1110 67 3.2 together at these
community: 16 11q.22 D11S1377 129 2.4 sites. However,
multigenerational, 17p11 D17S1876 11.9 3.4 individual families
extended D17S678 16.9 3.5 had significant
pedigrees) D17S1881 17.9 3.9 linkages at
D17S1844 22.8 3.8 different sites
D17S1791 25.1 3.7 (Family 9 at
4q13.2, 5q33.3,
8q11.23; Family 8
at 11q22; Family
14 at 17p11.)
Hebebrand 229 ASP 6q D6S1053 75 0.58 No linkage for
et al. (2006); 102 families 7p D7S2490 88 0.92 SLC6A3 VNTR;
Valera et al. (Germany) 9q D9S1851 104 0.68 Inattentive
(2007) 11q D11S4176 94 94 subtype more
12q D12S392 166 2.10 strongly
17p D17S1308 1 1.39 influenced by
gene in 5p region

170
 Chapter 15: Attention-deficit hyperactivity disorder

Table 15.1 (cont.)

Studies Sample Chromo- ADHD Closest Position LOD Results

(Year) some candidate marker (cM)
gene
Faraone et al. 601 ASP 8 54.2 1.85 No significant
(2008) (217 families) regions
1000 sibs 8 93.4 0.8
(260 families)
Doyle et al. 1212 individuals 3q13 RS2062834 115.4 2.11 Suggested
(2008) (271 families) linkage with
neurocognitive
tasks and
inattentive
symptoms
Zhou et al. (2545 ASP 1p36 45 3.2 Significant
(2008) (1094 families) 9p23 27 2.2 linkage at 1p36
11q21 100 2.6
Abbreviations: ASP, affected sibling pair; cM, centiMorgan; VNTR: variable number of tandem repeats.

Table 15.2 Candidate attention-deficit hyperactivity disorder (ADHD) genes and risk for structural and functional neuroimaging
intermediate ADHD phenotypes.

Gene variant Brain area

Structural studies (Year)
Durston, S. (2005) SLC6A3-10/10 Reduced caudate volume
Monuteaux, M. (2008) DRD47R Reduced superior frontal and cerebellar cortex
Shaw, P. ( 2007) DRD47R Cortex (thickness).
Functional studies (Year)
Durston, S. (2008) SLC6A3–10R Increased striatal activity
Durston, S. (2008) SLC6A3–10R Decreased cerebellar activity
Fan, J. (2003) DRD4, MAOA Activation in anterior cingulated gyrus

rare SLC6A3 coding variant, Ala559Val has also
been identified in two male children with ADHD.
This variant, which results in a normal SLC6A3
protein and dopamine uptake, also enhances non-
vesicular SLC6A3-dependent DA release which is
interestingly also blocked by effective ADHD
medications [68].
Gene variants in dopamine receptors (DRD1,
DRD2, DRD4, DRD5) have also been reported to
be associated with ADHD. The most widely
studied has been DRD4, a gene that has a poly-
morphic region of 2–10 repeats. A positive associ-
ation of the seven-repeat, a variant considered to be
less responsive to dopamine, has been reported in
several meta-analyses indicating moderate risk [66,
67, 69, 70].
Hetereogeneity in effect size was noted across
studies [67]. It is of interest that even 1-year-old
infants with this allele have been reported to show
less sustained attention than those without it. Infants
that also carried the seven repeat in addition to the
homozygous form of 5HTTLPR had the shortest
attention span [71]. Promoter variants of DRD4 have
also been explored in ADHD and a modest associ-
ation was reported in the meta-analyses for the “T” of
SNP 521 (rs1800955) [67].
Other dopaminergic genes conferring risk for
ADHD include a 148 bp allele of a microsatellite
repeat polymorphism on DRD5 while the 136bp
allele confers a protective effect. DRD1, DRD2 and
DRD3 variants have not been shown to confer risk
[66, 67].

171
 Chapter 15: Attention-deficit hyperactivity disorder
Catecholaminergic pathways
ADHD children homozygous for the the catechol-
O-methyltransferase (COMT) Val allele have been
reported to have reduced prefrontal cortex dopamine
neurotransmission [72], and ADHD adult subjects with
the Val/Val genotype have slower reaction times [73].
However this polymorphism was not found to modulate
executive function [74, 75] and several meta-analyses
have provided no evidence for association between
ADHD and the Val allele polymorphism [67, 76].
Several polymorphisms (1021C/T, 2*rrrG/A,
GT microsatellite, TaqI with an undigested band
of 464 base pair (A1) and two bands of 300 bp and
164 bp (A2)) in dopamine beta-hydroxylase (DBH), the
enzyme that catalyzes the conversion of dopamine to
noradrenaline, have been associated with ADHD
in some but not all studies [77]. A recent meta-
analyses showed a trend towards association only
for TaqI [67] and a recent study found an impaired rate
of perceptual processing for rapidly presented visual
events in ADHD subjects carrying the A2 allele [78].
Meta-analyses results did not detect association for
variants of the norepinephrine transporter (NET) gene
(SLC6A2) or alpha-2A adrenergic receptor gene
(ADRA2A) [67]. However, NET deserves further inves-
tigation given the results from a study using a dense-
mapping strategy that captured all genetic variations
within the SLA6A2 that found an association between
ADHD and two SNPs (rs3785143 and rs11568324)
that had also been identified in the IMAGE study [79].
Serotonergic pathways
Results from candidate gene studies investigating gene
variants in serotonin receptors (HTR1B, HTR2A), and
in the serotonin transporter (5-HTT) in ADHD
cohorts have been inconsistent [77]. A population-based
association study that investigated 19 genes encoding
serotonin receptors, transporters, and enzymes involved
in serotonin synthesis and degradation reported an asso-
ciation with polymorphisms of DDC (involved in
serotonin synthesis), MAOB (serotonin degradation),
and 5-HTR2A [80]. A meta-analysis has detected a sig-
nificant association only for an HTR1B variant [67].
Association studies for monoamine oxidase A
(MAOA), which also codes for an enzyme involved
in metabolism of norepinephrine, dopamine, and
serotonin, show mixed results [77]. Results from
meta-analyses indicate a weak associaton for a VNTR
in the promoter region [67].
172
Cholinergic pathways
CHRNA4, the nicotinic acetylcholine receptor a subunit
gene, is also considered important in ADHD; however,
it has not been widely studied. A meta-analyses detected
a trend for association of the “T” allele in exon 2
(rs2273506) and in intron 2 (rs6090384) [67]. The
pedigree disequilibrium test in a genetic isolate found
the T allele of rs6090384 to be undertransmitted in
affected individuals [81]. Individual variations in
spatial attention have been reported to be associated
with variants in CHRNA4 [82].
Glutaminergic neurotransmission
Altered glutamatergic function in frontal and striatal
regions in ADHD has been reported [83–85], and this
appears to be normalized with medications used to treat
ADHD such as the stimulants and atomoxetine [86].
SLC1A3 is located on 5p13 and encodes for a high-affinity
glutamate transporter (GLAST; EAAT1). An association
between several markers (rs 2269272 and haplotypes
rs2269272/rs3776581 and rs2269272/rs2032893) and
ADHD was reported by Turic et al. [87] but not replicated
for the same markers and haplotypes in 257 nuclear
families by Laurin and colleagues [88]. Results from the
CHOP gene-wide study of SLC1A3 [89] found a nominal
association for one intron SNP (rs3776571) and two
downstream SNPs (rs1529461; rs 6863386) but did not
replicate Turic’s findings. Reports for the glutamate
receptor, ionotropic, N-methyl D-aspartate 2A (GRIN2A)
have also been mixed [77].
SNAP-25 and other genes regulating
vesicular neurotransmission
The SNAP-25 gene, on human chromosome 20p11.2,
encodes for a presynaptic membrane protein (synap-
tosomal associated protein 25kDA), that in combin-
ation with synaptobrevin and syntaxin, forms the
SNARE (soluble N-ethylmalemide-sensitive factor
attachment protein receptors) complex, which is
responsible for binding and fusion of neurotransmit-
ter vesicles to the plasma membrane [90]. Barr et al.
[91] first reported a haplotype defined by two bi-
allelic SNPs (T87610G, T87614C), MnlI (rs3746544)
and the DdeI polymorphism (rs1051312), associated
with ADHD in the 3’UTR of SNAP25. A modest
significant association has also been reported for
s3746544 in a study that included pooled data [92]
as well as in meta-analyses [67].

Gene expression may also be playing a role. An
example, not yet explored in ADHD is the cis-acting
element, SP1, in the promoter region of SNAP-25; SP1
has been reported to increase SNAP-25 expression while
inhibition of SP1 results in reduced expression [93].
In summary, several gene variants in genes primar-
ily involved in neurotransmission have been replicated
in several studies; however, their contribution to ADHD
risk is modest. This may be in part due to phenotypic
and genotypic heteregenity and epigenetic factors.
Genome-wide association studies
Genome-wide analyses (GWA) allow investigation for
association between a disorder and a marker virtually
anywhere in the genome. Results of several GWA stud-
ies (GWAS) conducted on data from the International
Multicenter ADHD Genetics (IMAGE I) samples, based
on ADHD diagnoses and related phenotypes, did not
identify any markers with genome-wide significance
[94, 95]. One IMAGE I GWAS, that used quantitative
ADHD phenotypes, identified two SNPs (rs552655 and
rs6565113) with nominal significance, and one of these
SNPs, rs6565113, is located in an intron of CDH13 that
codes for cadherin 13, a member of a family of cell–cell
adhesion proteins [96]. This same SNP was also
reported to have nominal significance and was in the
top 25 SNPs reported by Neale and colleagues in the
same sample and also in a GWAS of an independent
sample of adults with ADHD [51, 97]. However, a meta-
analysis that included samples from IMAGE I, IMAGE
II, The Children’s Hospital of Philadelphia, and the
Pfizer-funded study from the University of California
with a combined sample size of 2064 trios, 896 cases,
and 2455 controls did not detect any genome-wide
significant associations [98].
Structural variants
In addition to single nucleotide variations, larger and
more complex variations have been recently identified in
the human genome, referred to as copy number variants
(CNVs) [99, 100]. These duplications and deletions can
result in changes in gene dosage that may contribute to
genomic instability [101] and phenotypic variations
[102, 103] in complex disorders such as ADHD.
The lack of success in identifying common vari-
ants contributing substantial risk for ADHD in the
European population also led to an alternative
hypothesis, namely that ADHD genetic risk is trans-
mitted largely by rare variants that collectively disrupt
Chapter 15: Attention-deficit hyperactivity disorder
a sizable constellation of genes, presumably with
related functions. Given the high prevalence and her-
itability of ADHD, a corollary of the rare variant
hypothesis would be that most risk variants arose as
relatively recent events, which would considerably
lessen the ability of association studies to detect them.
CNVs in genes or sets of genes related through the
same biological or clinical concept may be more likely
to explain most forms of ADHD.
In the first ADHD study investigating structural
variation in ADHD, 222 CNVs were identified in 335
ADHD patients and their parents that were not detected
in 2026 unrelated healthy controls. The ADHD CNV-
associated gene set was significantly enriched for genes
important for psychological and neurological functions,
including learning, behavior, synaptic transmission,
and central nervous system development. It was also
significant enriched for genes reported in autism,
schizophrenia, and Tourette syndrome [104]. Four
other subsequent studies also identified a number of
variants [105–107]. As expected, different variants were
identified by the four different studies suggesting that
there could be hundreds or even thousands of variants
disrupting genes in neuronal pathways that could
lead to similar phenotypic manifestations. The genes
impacted by structural variants reported in one or more
studies are listed in Table 15.3.
Gene–environment interactions
Heritability studies also suggest that epigenetic factors
play a role in ADHD. As reviewed by Elia et al. [39],
animal studies of malnutrition, maternal stress, infec-
tion, and toxic compounds have been shown to affect
prenatal development of the brain circuitry relevant
to ADHD. Human studies have reported an associ-
ation between ADHD and low birth weight [108],
maternal smoking [109], and psychosocial adversity
[110], but not prenatal alcohol exposure [109, 111].
The dopamine transporter gene (DAT; SLC6A3) has
also been reported to be affected by epigenetic factors
that impact on chromatin structure, DNA methylation,
and transcriptional regulation [112]. DAT mRNA
expression has also been reported to decrease with age
[113], in response to medication [114], environmental
factors [114], and pathogens [115].
ADHD: a complex phenotype
ADHD is highly heteregenous. Symtoms such as
hyperactivity vary in individual patients over
173
 Chapter 15: Attention-deficit hyperactivity disorder

Table 15.3 ADHD genes impacted by CNVS identified in two or more studies.

Elia et al. Williams et al. Lesch et al. Lionel et al. Elia et al. 2011
2009 2011 2011
Cases/ 335/2026 366/1047 UK 99/2026 248/2353 CHOP 1013/4105
Controls US 825/35243 Iceland Germany Canada Replic 2493/9222
CNV size 7.75–1775 kb > 500 kb 110 kb-3 Mb >20 kb CHOP 62.2 kb
Repl 53.6 kg
Genotyping Illumina Human aCGH Affy 6.0
Platform 550 K 660 W-Quad
Cases/controls Illumina 550 K

Chromosome Gene(s)
1p31.1 NEGR1 Dup
1p32.3 UPS24 Del
3p26.1 GRM7 Del
3p26.3 CHL1 dup dup
CNTN6 dup del
3q26.1 ZZBX Dup Del Del
Serpini 12 Dup Del
WDR49
PDC10
3q11.2 EPHA6 Dup del
4p15.2 DHX15 Dup dup
4q22.1 PPM1K Del dup
5p15.2 CTNND2 Del Dup
5q12.3 SGTB/NLN Del
5q13.3 SV2C Dup Dup
IQGAP2 Dup
5q35.2 CPLX2 Del dup
6q22.33 PTPRK Dup del
6q24.3 GRM1 Dup
7p22.2 SDK1 Del/dup del
7q31.33 GRM8 Del
7q32.3 CHCHD3 Dup Dup
7q35 CNTNAP2 Del dup
7q36.2 DPP6 dup
7q.11.22 AUTS2 Del/dup Dup
7q.32 CHCHD3 dup dup
8p23.2 CSMD1 Del del
8p22 SGCZ Dup del
8q21.3 CNBD1 Del del

174
 Chapter 15: Attention-deficit hyperactivity disorder

Table 15.3 (cont.)

Elia et al. Williams et al. Lesch et al. Lionel et al. Elia et al. 2011
2009 2011 2011
11q13.4 DNAJB13 dup dup
CHCHD8
MRPL48
PAAF1
UCP2
11q14.3 GRM5 Del
15q13.3 FMN1 Dup dup
16p11.2 40 genes Del/dup Dup
16q23.3 CHD13 Del del
16q24.1 ATP2C2 Del del
20p12.2 PAK7 Dup dup
20p12.1 MACROD2 del Del/dup del
Del-deletion; Dup-duplication

time [116]. With the exception of the inattentive sub-
type which has strong familial clustering in an isol-
ated Dutch population [117], family studies reported
a lack of similarity for ADHD subtypes in first-degree
relatives of probands with ADHD [17, 118, 119].
Comorbid conditions occurring in approximately
33% add to the heterogeneity of ADHD [120, 121].
These include oppositional defiant disorder (35%),
conduct disorder (30–50%), anxiety disorders (25%),
mood disorders (15–75%), and learning disabilities
(25%) [122–127]. Epidemiological samples of chil-
dren and adolescents also report significant comor-
bidity suggesting that this is not due to referral bias
that might be expected in clinical samples [128–130].
It is not known whether these associated conditions
modify the ADHD phenotype or whether single genes
influence multiple phenotypic traits, as suggested in
a study by Jain et al. [131] where ADHD was found
to co-segregate with oppositional defiant disorder
(ODD) and conduct disorder (CD).
To address this phenotypic heterogeneity, neurocog-
nitive tests [82] and statistical methods used to identify
distinct homogenous ADHD subgroups [132] are being
used to further define ADHD phenotypes. Structural and
functional neuroimaging are additional measures util-
ized as intermediate phenotypes for the investigation of
candidate genes conferring risk for ADHD (Table 15.2).
Thus far, SLC6A3 variants have been associated with
reduced caudate volume [42], increased striatal brain
activation [133], and decreased cerebellar activity [133].
The 10 repeat (10R) homozygous variant of SLC6A3, a
gene expressed primarily in the striatum [134], was
shown to be associated with decreased caudate volume
in subjects with ADHD, their unaffected siblings, and
controls [42]. The 10R allele, which is associated with
decreased dopamine transporter activity [135], may also
be related to reduced gene expression.
Reduced prefrontal gray matter volumes have
been reported for DRD4-R4/4; however there was no
genotype effect detected in ADHD children versus
controls [42], which was not surprising given that this
is not the variant that has been associated with
ADHD. In contrast, the DRD4-7R variant, associated
with ADHD, has been associated with decreased cor-
tical thickness in ADHD children [136] and ADHD
adults [137], where it has also been associated with
decreased volume in the cerebellar cortex [137].
The differences in brain structure in ADHD do not
appear to be static. As reported by Shaw et al. [136],
delays in prefrontal cortical maturation of  3 years
were detected in ADHD children compared to con-
trols. ADHD subjects with the DRD47R genotype had
the thinnest cortex and also had better clinical out-
comes suggesting that this neuroanatomical correlate
of DRD4 genotype resolved by late adolescence [136].
Recent neuroimaging genetic studies that also
included unaffected siblings of ADHD children
are allowing the assessment of familial risk for

175
 Chapter 15: Attention-deficit hyperactivity disorder
candidate genes. The SLC6A3 10R allele was found to
be associated with increased striatal activation in
ADHD children and their nonaffected siblings but
not in controls, indicating that this variant alone does
not result in clinical symptoms [133]. In this same
study, decreased activation of the cerebellar vermis,
which was found in ADHD subjects, siblings, and
healthy controls with the SLC6A3 10R homozygous
genotype, suggested a lack of familial risk [133]. How-
ever, in a subsequent study that did not investigate
genotype, and utilized the same basic go/no-go para-
digm (but had the stimuli occurring at both expected
and unexpected time intervals), unaffected siblings
showed similar anomalous activation to that of
ADHD affected subjects when processing a stimulus
presented at unexpected intervals [138].
Brain imaging studies in twins are being used
to decipher changes that could be attributed to genetic
or environmental factors. A structural magnetic
resonance imaging (MRI) study in monozygotic twins
concordant or discordant for ADHD symptoms
showed reduced function of attentional networks
of cerebellar, occipital/parietal and temporal brain
regions interacting with the prefrontal cortex. In
ADHD of genetic origin (defined as brain differences
between concordant high risk twin pairs for ADHD
and a low risk twin comparison group), the medial
orbitofrontal subdivisions were abnormal; while in
ADHD of environmental origin (brain differences
between the low risk and high risk twins from dis-
cordant MZ twin pairs), the right inferior dorsolateral
prefrontal cortex was abnormal [139]. Functional
MRI studies indicated decreased activation of the left
dorsolateral prefrontal cortex and right parietal lobe
for ADHD of genetic origin, and decreased activation
in left and right temporal lobe areas in ADHD of
environmental origin [140].
Pharmacogenetics
As reviewed by Froehlich et al. [141], pharmacogenetic
studies have for the most part focused on ADHD can-
didate gene variants involved in neurotransmission and
methylphenidate reponse with mixed results. A GWAS
investigating response to a methylphenidate transder-
mal system reported nominal evidence for SLC6A2
SNPs (rs17841329 and rs192303) as well as for GRM7,
a metabotropic glutamate receptor gene (rs3792452)
[142]. Amphetamine compounds have not been as thor-
oughly studied in ADHD subjects. In one study in
176
normal adult volunteers with homozygous (COMT)
Val/Val genotype, improvement with amphetamine
treatment was demonstrated, whereas the homozygous
Met/Met subjects performed more poorly after medica-
tion in working memory tasks [143].
Thus pharmacogenetic studies, considered as com-
plementary approaches in the search for biologically
relevant disease genes, have not yet been helpful in the
quest for the identification of causal genes. This may be
due in part to the fact that our pharmacotherapeutic
probes are nonspecific, as shown in one study indicat-
ing changes in gene expression in over 700 genes in the
striatum of rats treated with methylphenidate [144]. In
addition, as Levy [145] recently hypothesized, genes
conferring risks for ADHD may be different from
genes involved in medication response.
Genetic variants that influence the pharmacoki-
netics of ADHD medications may be more clinically
useful. A gene variant for caboxylesterase 1, the enzyme
used to esterify methylphenidate to d, l-ritalinic acid
and l-ethylphenidate has been reported in one subject
identified as a poor methylphenidate metabolizer [146].
Amphetamine compounds are metabolized through
the hepatic CYP450 system, primarily through CYP3A4
and to a lesser extent through CYP2D6 [147], while
atomoxetine is metabolized primarily through CYP2D6.
Poor metabolizers of atomoxetine were shown to
have greater symptom improvement [148].
ADHD animal models
ADHD animal models have been developed through neu-
rotoxic brain lesions [149, 150], neonatal anoxia [151,
152], toxin exposure such as lead [153], polychlorinated
biphenyls [154], and X-irradiation [155]. More recently, as
reported in several excellent reviews [156–159] and sum-
marized in Figure 15.2, the development of knockout
(KO), knockdown (KD), selected inbred strains, trans-
genic strains, and ENU mutants [160] are providing
insight into the complex genetic components of ADHD.
KO animal models
KO animals, allowing the exploration of single-
ADHD candidate genes, indicate that a number of
different genes, in the same as well as in different
neurotransmitter pathways, may be responsible for
similar behaviors. As reviewed by Viggiano and col-
leagues [59], in the dopaminergic pathway, the homo-
zygous SLC6A3 (DAT1) KO mouse, which clears
synaptic dopamine at a slower rate than the hemizygous

MAO-B MAO-A
DAT1
DRD1
K-O DRD2
DRD3
H1R
K-D DAT1
Hemizygous Deletions SNAP-25
SHR
WK/HA WKY Hyposexual
Selected In Bred
Strains
α2c++
Transgenic α1b++
ENU Models Mutations can occur anywhere in the genome and can affect different alleles
Figure 15.2 Attention-deficit hyperactivity disorder (ADHD): genetic animal models.
and the wild type (WT), is also more hyperactive than
these two strains [161] as is the DRD3 KO model [162].
Decreased locomotor activity occurs in the DA [163],
tyrosine hydroxylase (TH) [164], D2 [165], DARPP-32
[166], and DRD4 KO [167]. D5-KO had normal activity
level [168] while results for the D1-KO animal show
both normal and elevated locomotor activity [169, 170].
Double KO strains result in increased levels for D1 + D3
[171] and decreased levels for D2 + D3 [59].
A review of the noradrenergic pathway by Vig-
giano et al. [60] indicates that locomotor activity is
decreased in SLA6A2 [172] and a2a [173] KO animals
and in strains where a2c [174] and a1b [175] are
over-expressed. Locomotor activity is increased in
KO a1b [176] and a2c [174] models. TH is necessary
for the synthesis of both dopamine (DA) and nor-
epinephrine (NE) [163, 164], since 3,4-dihydroxyphe-
nilalanine (l-DOPA) is converted to DA (by TH)
which is then converted to NE (by DBH). Pure DA
KO mice can be generated by normalizing expression
of TH and these mice are severely hypoactive [163].
Animals with a TH mutation resulting in decreased
Chapter 15: Attention-deficit hyperactivity disorder
COMT
NET
α2c
α2a α1b
TH
DRD4 5H1B DBH
DA
TH
DARPP-32
Rat NHE
NLE C57BL/6
TRβ1 Caly
TRα1R348C
spatial learning and long-term potentiation (LTP) are
normal but perform less well on behavioral tasks and
have lower activity levels which are restored by
increasing NE activity [177]. Interestingly, DBH KO
mice which lack NE have normal locomotor activity.
Activity level changes are not limited to the dopa-
minergic and noradrenergic systems, as hyperactivity
has also been reported in mice lacking the 5-HT1B
receptor [178], while reduced locomotor activity has
been noted in both histamine-1-receptor (H1R) -KO
and WT mice [179].
Toxic lesions in KO animals are also elucidating
signaling necessary for locomotor activity changes.
An example are rat pups injected with intracisternal
6-hydroxydopamine (6-OHDA), one of the earliest
animal models for ADHD [149]. Compared to control
littermates, they presented with hyperactivity and
deficits in brain dopamine that was reversed by
methylphenidate and d-amphetamine [180]. 6-OHDA
injection in DRD4 KO mice, however, prevented the
development of hyperactivity, showing that D4 recep-
tor signaling is essential for hyperactivity [181].
177
 Chapter 15: Attention-deficit hyperactivity disorder
KD animal models
DAT-KD mice have DAT expression reduced to 10%
compared to WT animals, most likely due to reduced
clearance [182]; these animals provide a model of a
chronic hyper-dopaminergic state. These animals are
hyperactive and have alterations in habituation. Elec-
trophysiological recordings from medium-sized spiny
neurons in the dorsal striatum of KD mice showed
alterations in both amplitude and frequency of spon-
taneous glutamate-receptor-mediated synaptic cur-
rents [183]. Since D2 receptors normally dampen
glutamate release signaling [184, 185], it is hypothe-
sized that in a chronic hyperdopaminergic state there
may be alterations in presynaptic D2 receptor function.
Hemizygous deletions
The Coloboma mutant mouse, characterized by
hyperactivity, head bobbing, eye dysmorphology,
and neurobehavioral delays [186] has a deletion in
chromosome 2 that includes the gene for synaptoso-
mal associated protein 25 (SNAP-25). These animals
respond to amphetamine but not to methylphenidate
and this effect has been shown to be mediated
through D2 dopamine receptors [187]. Hyperactivity
in this mouse model is also mediated through a2C-
adrenergic receptors [188] suggesting that both nor-
adrenergic and dopaminergic systems may be involved.
Selected inbred strains
Inbred strains have included the spontaneously hyper-
tensive rats (SHR) [189], the hyposexual rat [190], and
the Naples high- and low-excitability (NHE and NLE)
rats [191].
One of the most frequently studied inbred models
is the SHR, one of the few animal models known to
display several ADHD features including hyperactiv-
ity, impulsivity, and difficulties with attention [158].
By selective inbreeding of this strain with a progeni-
tor, the Wistar Kyoto strain (WKY), a hyperactive but
not hypertensive model was developed [192]. Sequen-
cing studies of candidate dopaminergic genes DRD2,
DRD4, and SLC6A3 in the SHR strain compared to
the WKY strain (nonhyperactive) showed variations
only in the SLC6A3 gene and not in DRD2 or DRD4
[193], as well as reduced SLC6A3 gene expression
in the first postnatal month and increased SLC6A3
gene expression in adulthood [194, 195]. Mesocortical
SLC6A3 expression, increased in SHR strains [193, 196]
178
may be compensatory to the excessive dopamine
present during early development [196].
Naples high-excitability (NHE) rats have high activ-
ity while the low-excitability strain (NLE) has reduced
locomotor activity; both strains have deficits of atten-
tion [191, 197]. In the NHE animals, SLC6A3 and tyro-
sine hydroxylase expression are increased in prefrontal
cortex while D1 receptor expression is decreased [197];
they also have an imbalance between NMDA and non-
NMDA sensitive [L-3H] glutamate receptors [198].
The C57BL/6 is another inbred strain showing
hyperactivity [199] and has low expression of D2-R
autoreceptors in the ventral tegmental area (VTA) [200].
Transgenic animals
Transgenic male mice with a human mutant thyroid
receptor (TRb1) are hyperactive, inattentive, and
impulsive, and with the exception of a brief period
during postnatal development they are euthyroid [201].
Rats with a knock-in TRa1R384C, a mutation of
the TRa1 gene that lowers affinity to thyroid hormone,
were noted to have extreme anxiety, reduced memory
and locomotor activity [202]. A mouse engineered for
the over-expression of the Calcyon gene, a vesicular
protein involved in endocytosis which is important in
the recycling of neurotransmitters, is hyperactive with
reduced anxiety [203].
ENU mutants
Novel mutant animals, produced with chemicals such
as N-ethyl-N-nitrosurea (ENU), are expected to be
better models of the human disease. These mutants
are not limited to loss of single gene function since
mutations can occur anywhere in the genome and can
affect different alleles [160].
Gene–environmental interactions:
animal models
Animal models will also be invaluable in deciphering
gene and environmental interactions. Studies with
the NHE rat have shown that increased maternal
care and high fat diet (induced by small litter size)
resulted in decreased activity and longer scanning times
[204]. Prepubertal handling of NHE animals also
resulted in decreased expression of excitatory amino
acids (L-Glu, L-Asp, and D-asp) and decreased activity
levels compared to nonstimulated animals [205].
Increased stress in the C57BL/6 strain has been shown
 Chapter 15: Attention-deficit hyperactivity disorder

to result in comorbid depressive behaviors [200]. A pilot
study that examined brain RNA from rats exposed to
polychlorinated biphenyl (PCB) compared to SHR and
control rats, found different sets of genes expressed
[206]. A rat model exposed to nitrogen during the first
seven days of postnatal life showed increased anxiety.
While MRI and histological analysis did not detect any
brain damage, expression studies detected decreased
expression of group-I metabotropic glutamate receptors
[207]. In contrast, mice lacking the 5-HT1B receptor are
reported to be hyperactive throughout the lifespan but
show reduced anxiety [178].
Pharmacogenetics of animal models
Animal models such as the DAT1 KO mouse,
remains responsive to methylphenidate in spite of
the lack of a dopamine transporter [208]. SLC6A3
KO animals also respond to fluoxetene [208]. Hyper-
activity in these mice can be increased by NMDA-
receptor blockers and suppressed by drugs that
increase glutamatergic transmission [209].
Methylphenidate and d-amphetamine decreased
symptoms in the SHR compared to WKY controls
[210, 211]. In the SHR model, reduced a2 adrenocep-
tor-mediated inhibition of NE release mediates hyper-
activity [212], pointing to a noradrenergic focus that
has been bolstered by evidence showing the efficacy of
the selective NE transporter (NET) inhibitor, atomox-
etine, in human ADHD studies [213, 214]. However,
the dysfunction may be resulting from a defect in
glutamate-stimulated release of dopamine [215].
Interestingly, in SHR pups, performance in a training
task involving cognition and attention was signifi-
cantly enhanced with methylphenidate and also with
H3 receptor antagonists [216].
Summary
Candidate genes, linkage, and GWAS have identified
several gene variants involved in neurotransmission
which confer a modest risk for ADHD. In part, this
may be due to phenotypic heterogeneity as well as
environmental factors, gene–environment inter-
actions, or multiple variants within genes conferring
risk for ADHD. It is also possible that few common
variants conferring risk for ADHD exist in the European
population and that very large samples will be necessary
to identify them.
Structural variant studies are indicating that ADHD
genetic risk is likely to be transmitted largely by rare
variants that collectively disrupt a sizable constellation
of genes, presumably with related functions. This is
supported by animal genetic models which also impli-
cate numerous genes involved in complex interactions
between neural pathways. Medication effects also sup-
port this, given that over 700 genes involved in the
formation, maturation, and stability of neural connec-
tions were reported to have increased expression in the
striatum of rats treated with methylphenidate [144].
Epigenetic factors, environmental factors, and
gene regulatory elements most likely also play a role
in ADHD genetics, and these are just beginning to be
explored.

References 7. Barkley R. Attention-Defict/

1. Scahill L, et al. Child Adolesc
Psychiatr Clin N Am 2000;9
(3):541–555.
2. Heiligenstein E, et al.
J Am Coll Health 1998;46(4):
185–188.
3. Kooij J, et al. Psychol Med 2005;
35(6):817–827.
4. Polanczyk G, et al. Curr Opin
Psychiatry 2007;20(4):386–392.
5. Barkley R, et al. J Am Acad Child
Adolesc Psychiatry 2006;45(2):
192–202.
6. Strine TW, et al. Prev Chronic Dis
2006;3(2):A52.
Hyperactivity Disorder.
A Clinical Workbook.
New York: Guilford Press;
1998.
8. Kessler RC, et al. J Occup Environ
Med 2005;47(6):565–572.
9. Weiss G, et al. Hyperactive
Children Grown. New York:
Guilford Press; 1993.
10. Pomerleau OF, et al. J Subst Abuse
1995;7(3):373–378.
11. Biederman J, et al. Biol Psychiatry
1998;44(4):269–273.
12. Barkley RA, et al. Pediatrics
1996;98(6 Pt 1):1089–1095.
13. Biederman J, et al. Arch
Gen Psychiatry 1992;49(9):728–738.
14. Lombroso PJ, et al. J Am Acad
Child Adolesc Psychiatry 1994;
33(7):921–938.
15. Faraone SV, et al. Am J Psychiatry
1991;148(1):112–117.
16. Faraone SV, et al. J Abnorm
Psychol 1995;104(2):334–345.
17. Faraone SV, et al. Biol Psychiatry
1994;35(6):398–402.
18. Sprich S, et al. J Am Acad Child
Adolesc Psychiatry 2000;39(11):
1432–1437.
19. Biederman J, et al. Lancet 2005;
366(9481):237–248.

179
 Chapter 15: Attention-deficit hyperactivity disorder
20. Kuntsi J, et al. Biol Psychiatry
2005;57(6):647–654.
21. Price TS, et al. Behav Genet 2005;
35(2):121–132.
22. Saudino KJ, et al. J Abnorm Child
Psychol 2005;33(1):113–130.
23. Hay DA, et al. Biol Psychiatry
2007;61(5):700–705.
24. Polderman TJ, et al. J Child Psychol
Psychiatry 2007;48(11):1080–1087.
25. Derks EM, et al. Behav Genet
2008;38(1):11–23.
26. Ouellet-Morin I, et al. Am J Med
Genet B Neuropsychiatr Genet
2008;147B(8):1442–1449.
27. Wood AC, et al. Behav Genet
2008;38(3):266–276.
28. Ehringer MA, et al. J Abnorm
Child Psychol 2006;34(1):1–17.
29. Schultz MR, et al. Twin Res Hum
Genet 2006;9(2):220–232.
30. van den Berg SM, et al. Am J Med
Genet B Neuropsychiatr Genet
2006;141B(1):55–60.
31. Haberstick BC, et al. Psychol Med
2008;38(7):1057–1066.
32. Young SE, et al. J Abnorm Psychol
2009;118(1):117–130.
33. Saviouk V, et al. Am J Med Genet
B Neuropsychiatr Genet
2011;156B(3):352–362.
34. Simonoff E, et al. Psychol Med
1998;28(4):825–837.
35. Nadder TS, et al. J Child Psychol
Psychiatry 2001;42(4):475–486.
36. Sherman DK, et al. Am J
Psychiatry 1997;154(4):532–535.
37. Hay D, et al. Aust J Psychology
2004;56(2):99–107.
38. Larsson JO, et al. J Am Acad Child
Adolesc Psychiatry 2004;
43(10):1267–1275.
39. Elia J, et al. Curr Top Behav
Neurosci 2011 [epub ahead of
print].
40. Lenroot RK, et al. J Child Psychol
Psychiatry 2011;52(4):429–441.
41. Roth TL, et al. J Child Psychol
Psychiatry 2011;52(4):398–408.
180
42. Fisher SE, et al. Am J Hum Genet
2002;70(5):1183–1196.
43. Smalley S, et al. Am J Hum Genet
2002;71(4):959–963.
44. Bakker SC, et al. Am J Hum Genet
2003;72(5):1251–1260.
45. Ogdie M, et al. Am J Hum Genet
2003;73(5):493.
46. Hebebrand J, et al. Mol Psychiatry
2006;11(2):196–205.
47. Asherson P, et al. Mol Psychiatry
2008;13(5):514–521.
48. Zhou K, et al. Biol Psychiatry
2008;64(7):571–576.
49. Arcos-Burgos M, et al. Am J Hum
Genet 2004;75(6):998–1014.
50. Amin N, et al. Eur J Hum Genet
2009;17(7):958–966.
51. Lesch KP, et al. J Neural Transm
2008;115(11):1573–1585.
52. Romanos M, et al. Mol Psychiatry
2008;13(5):522–530.
53. Vegt R, et al. Eur J Hum Genet
2010;18(2):206–211.
54. Ogdie MN, et al. Mol Psychiatry
2006;11(1):5–8.
55. Zhou K, et al. Am J Med
Genet B Neuropsychiatr Genet
2008;147B(8):1392–1398.
56. Weiss KB, et al. J Allergy Clin
Immunol 2000;106(3):493–499.
57. Mannino DM, et al. MMWR
Surveill Summ 2002;51(1):1–13.
58. Solanto MV. Behav Brain Res
1998;94(1):127–152.
59. Viggiano D, et al. Neurosci
Biobehav Rev 2003;27(7):
623–637.
60. Viggiano D, et al. Neural Plast
2004;11(1–2):133–149.
61. Volkow ND, et al. Neuroimage
2007;34(3):1182–1190.
62. Cook EH Jr., et al. Am J Hum
Genet 1995;56(4):993–998.
63. Hersch SM, et al. J Comp Neurol
1997;388(2):211–227.
64. Nirenberg MJ, et al. J Neurosci
1997;17(18):6899–6907.
65. Vandenbergh DJ, et al. Genomics
1992;14(4):1104–1106.
66. Barr CL, et al. Future Neurology
2008;3(6):705–728.
67. Gizer IR, et al. Hum Genet
2009;126(1):51–90.
68. Mazei-Robison MS, et al.
J Neurosci 2008;28:7040–7046.
69. Faraone SV, et al. Am J Psychiatry
2001;158(7):1052–1057.
70. Li D, et al. Hum Mol Genet
2006;15(14):2276–2284.
71. Auerbach JG, et al. Psychiatr
Genet 2001;11(1):31–35.
72. Lachman HM, et al.
Pharmacogenetics 1996;6(3):
243–250.
73. Boonstra AM, et al. Am J Med
Genet B Neuropsychiatr Genet
2008;147(3):397–402.
74. Mills S, et al. BMC Psychiatry
2004;4:15.
75. Taerk E, et al. BMC Med Genet
2004;5:30.
76. Cheuk DK, et al. Behav Genet
2006;36(5):651–659.
77. Faraone SV, et al. Psychiatr
Clin North Am 2010;33(1):
159–180.
78. Bellgrove MA, et al. Biol
Psychiatry 2006;60
(10):1039–1045.
79. Kim JW, et al. Am J Med Genet
B Neuropsychiatr Genet 2007;
144(6):781–790.
80. Ribases M, et al. Mol Psychiatry
2009;14(1):71–85.
81. Wallis D, et al. Atten Defic
Hyperact Disord 2009;1(1):
19–24.
82. Bellgrove MA, et al. Ann N Y Acad
Sci 2008;1129:200–212.
83. Jin Z, et al. Neurosci Lett 2001;
315(1–2):45–48.
84. MacMaster FP, et al. Biol
Psychiatry 2003;53(2):184–187.
85. Courvoisie H, et al. J Neuro-
psychiatry Clin Neurosci 2004;
16(1):63–69.

86. Carrey N, et al. Clin
Neuropharmacol 2003;26(4):
218–221.
87. Turic D, et al. Biol Psychiatry
2005;57(11):1461–1466.
88. Laurin N, et al. Biol Psychiatry
2006;59(11S):115S.
89. Elia J, et al. Psychiatr Genet
2009;19(3):134–141.
90. Sorensen JB. Trends Neurosci
2005;28(9):453–455.
91. Barr CL, et al. Mol Psychiatry
2000;5(4):405–409.
92. Kim JW, et al. Mol Psychiatry
2008;13(6):624–630.
93. Cai F, et al. J Neurochem 2008;
105(2):512–523.
94. Franke B, et al. Hum Genet
2009;126(1):13–50.
95. Neale BM, et al. J Am Acad Child
Adolesc Psychiatry 2010;49(9):
906–920.
96. Patel SD, et al. Curr Opin Struct
Biol 2003;13(6):690–698.
97. Neale BM, et al. Am J Med Genet B
Neuropsychiatr Genet 2008;
147B(8):1337–1344.
98. Neale BM, et al. J Am Acad Child
Adolesc Psychiatry 2010;49:
884–897.
99. Eichler EE. Nat Genet 2006;
38(1):9–11.
100. Sharp AJ, et al. Nat Genet 2006;
38(9):1038–1042.
101. Emanuel BS, et al. Nat Rev Genet
2001;2(10):791–800.
102. Iafrate AJ, et al. Nat Genet 2004;
36(9):949–951.
103. Sebat J, et al. Science 2004;
305(5683):525–528.
104. Elia J, et al. Mol Psychiatry
2010;15(6):637–646.
105. Williams NM, et al.
Lancet 2010;376
(9750):1401–1408.
106. Lesch KP, et al. Mol Psychiatry
2011;16(5):491–503.
107. Lionel AC, et al. Sci Transl Med
2011;3(95):75–95.
Chapter 15: Attention-deficit hyperactivity disorder
108. Hultman CM, et al. J Am Acad
Child Adolesc Psychiatry 2007;
46(3):370–377.
109. Knopik VS, et al. Psychol Med
2006;36(10):1461–1471.
110. Laucht M, et al. Arch
Gen Psychiatry 2007;64
(5):585–590.
111. Neuman RJ, et al. Biol Psychiatry
2007;61(12):1320–1328.
112. Shumay E, et al. PLoS One 2010;
5(6):e11067.
113. Bannon MJ, et al. Neurology
1997;48(4):969–977.
114. Volkow ND, et al. Biol Psychiatry
2005;57(11):1410–1415.
115. Wang GJ, et al. Brain 2004;
127(Pt 11):2452–2458.
116. Lahey BB, et al. Arch Gen
Psychiatry 2005;62(8):896–902.
117. Croes EA, et al. Eur J Epidemiol
2005;20(9):789–794.
118. Faraone SV, et al. Am J Psychiatry
2000;157(7):1077–1083.
119. Smalley SL, et al. J Am Acad Child
Adolesc Psychiatry 2000;39(9):
1135–1143.
120. Jensen PS, et al. J Am Acad Child
Adolesc Psychiatry 2001;40(2):
147–158.
121. Elia J, et al. Child Adolesc
Psychiatry Ment Health 2008;
2(1):15.
122. Biederman J, et al. Am J Psychiatry
1991;148(5):564–577.
123. Hinshaw SP. J Consult Clin
Psychol 1992;60(6):893–903.
124. Cantwell DP. J Am Acad Child
Adolesc Psychiatry 1996;
35(8):978–987.
125. Jensen PS, et al. J Am Acad Child
Adolesc Psychiatry 1997;36(8):
1065–1079.
126. Brown RT, et al. Pediatrics
2001;107(3):E43.
127. Spencer TJ. J Clin Psychiatry
2006;67(Suppl 8):27–31.
128. Caron C, et al. J Child Psychol
Psychiatry 1991;32(7):1063–1080.
129. Angold A, et al. J Child Psychol
Psychiatry 1999;40(1):57–87.
130. Costello EJ, et al. J Am Acad Child
Adolesc Psychiatry 2006;45(1):8–25.
131. Jain M, et al. Biol Psychiatry
2007;61(12):1329–1339.
132. Elia J, et al. Psychiatry Res
2009;170(2–3):192–198.
133. Durston S, et al. J Am Acad
Child Adolesc Psychiatry
2008;47:61–67.
134. Heinz A, et al. Neuro-
psychopharmacology 2000;22:
133–139.
135. Jacobsen LK, et al. Am J Psychiatry
2000;157:1700–1703.
136. Shaw P, et al. Proc Natl Acad Sci
U S A 2007;104:19649–19654.
137. Monuteaux MC, et al. J Med Genet
B Neuropsychiatr Genet 2008;
147B:1436–1441.
138. Mulder MJ, et al. J Am Acad
Child Adolesc Psychiatry 2008;
47(1):68–75.
139. Van ‘t Ent D, et al. Neuroimage
2007;35(3):1004–1020.
140. Van ‘t Ent D, et al. Neuroscience
2009;164:16–29.
141. Froehlich TE, et al. CNS Drugs
2010;24(2):99–117.
142. Mick E, et al. Am J Med Genet
B Neuropsychiatr Genet
2008;147B(8):1412–1418.
143. Mattay VS, et al. Proc Natl
Acad Sci U S A 2003;100(10):
6186–6191.
144. Adriani W, et al. Ann N Y Acad
Sci 2006;1074:52–73.
145. Levy F. Aust N Z J Psychiatry
2007;41(1):10–16.
146. Patrick KS, et al. Clin Pharmacol
Ther 2007;81(3):346–353.
147. Markowitz JS, et al. Clin
Pharmacokinet 2001;40(10):
753–772.
148. Michelson D, et al. J Am Acad
Child Adolesc Psychiatry 2007;
46(2):242–251.
149. Shaywitz BA, et al. Science
1976;191(4224):305–308.
181
 Chapter 15: Attention-deficit hyperactivity disorder
150. Cardinal RN, et al. Science
2001;292(5526):2499–2501.
151. Dell’Anna ME, et al. Behav Brain
Res 1991;45(2):125–134.
152. Dell’Anna ME, et al. Brain Res
Bull 1993;32(2):159–170.
153. Silbergeld EK, et al. Exp Neurol
1974;42(1):146–157.
154. Holene E, et al. Behav Brain Res
1998;94(1):213–224.
155. Diaz-Granados JL, et al. Behav
Neural Biol 1994;61(3):251–259.
156. Davids E, et al. Brain Res Brain
Res Rev 2003;42(1):1–21.
157. Russell VA, et al. Behav Brain
Funct 2005;1:9.
158. Sagvolden T, et al. Biol Psychiatry
2005;57(11):1239–1247.
159. Mill J. J Neurosci Methods
2007;166(2):294–305.
160. Godinho SI, et al. Eur J Hum
Genet 2006;14(6):651–659.
161. Giros B, et al. Nature 1996;
379(6566):606–612.
162. Accili D, et al. Proc Natl Acad Sci
U S A 1996;93(5):1945–1949.
163. Zhou QY, et al. Cell 1995;
83(7):1197–1209.
164. Kobayashi K, et al. J Biol Chem
1995;270(45):27235–27243.
165. Vallone D, et al. Behav Brain Res
2002;130(1–2):141–148.
166. Nally RE, et al. Neuro-
psychopharmacology 2003;
28(12):2055–2063.
167. Rubinstein M, et al. Cell 1997;
90(6):991–1001.
168. Hollon TR, et al. J Neurosci
2002;22(24):10801–10810.
169. Drago J, et al. Proc Natl Acad Sci
U S A 1994;91(26):12564–12568.
170. Xu M, et al. Cell 1994;79(4):
729–742.
171. Wong JY, et al. Psycho-
pharmacology (Berl) 2003;
167(2):167–173.
172. Xu F, et al. Nat Neurosci 2000;
3(5):465–471.
182
173. Lahdesmaki J, et al. Neuroscience
2002;113(2):289–299.
174. Sallinen J, et al. Mol Psychiatry
1999;4(5):443–452.
175. Zuscik MJ, et al. Nat Med 2000;
6(12):1388–1394.
176. Drouin C, et al. J Neurosci 2002;
22(7):2873–2884.
177. Kobayashi K, et al. J Neurosci
2000;20(6):2418–2426.
178. Brunner D, et al. Behav Neurosci
1999;113(3):587–601.
179. Zlomuzica A, et al. Eur J Neurosci
2008;27(6):1461–1474.
180. Luthman J, et al. Psycho-
pharmacology (Berl) 1989;
99(4):550–557.
181. Avale ME, et al. Mol Psychiatry
2004;9(7):718–726.
182. Zhuang X, et al. Proc Natl
Acad Sci U S A 2001;
98(4):1982–1987.
183. Wu N, et al. J Neurophysiol
2007;98(1):423–432.
184. Cepeda C, et al. J Neurophysiol
2001;85(2):659–670.
185. Bamford NS, et al. J Neurosci
2004;24(43):9541–9552.
186. Heyser CJ, et al. Brain Res
Dev Brain Res 1995;89(2):
264–269.
187. Fan X, et al. Neurobiol Dis 2007;
26(1):201–211.
188. Bruno KJ, et al. Neurobiol Dis
2006;23(3):679–688.
189. Okamoto K, et al. Jpn Circ J
1963;27:282–293.
190. Kohlert JG, et al. Physiol Behav
1993;53(6):1215–1218.
191. Gonzalez-Lima F, et al. Neurosci
Biobehav Rev 2000;24(1):
157–160.
192. Hendley ED, et al. Behav Neural
Biol 1986;45(1):1–16.
193. Mill J, et al. Behav Brain Funct
2005;1:24.
194. Watanabe Y, et al. J Nucl Med
1997;38(3):470–474.
195. Leo D, et al. Neurosci Biobehav
Rev 2003;27(7):661–669.
196. Viggiano D, et al. Behav Brain Res
2002;130(1–2):181–189.
197. Viggiano D, et al. Behav Genet
2002;32(5):315–333.
198. Sadile AG, et al. Behav Brain Res
1996;78(2):163–174.
199. Cabib S, et al. Physiol Behav
1990;47(4):749–753.
200. Cabib S, et al. Behav Brain Res
2002;130(1–2):103–109.
201. Siesser WB, et al. Genes Brain
Behav 2006;5(3):282–297.
202. Venero C, et al. Genes Dev
2005;19(18):2152–2163.
203. Trantham-Davidson H, et al.
Behav Brain Res 2008;189
(2):244–249.
204. Fresiello A, et al. Behav Brain Res
2002;130(1–2):111–115.
205. Ruocco LA, et al. Behav Brain Res
2009;198(1):29–36.
206. DasBanerjee T, et al. Am J
Med Genet B Neuropsychiatr
Genet 2008;147B(8):1554–1563.
207. Casolini P, et al. J Neurochem
2005;95(1):137–145.
208. Gainetdinov RR, et al. Science
1999;283(5400):397–401.
209. Gainetdinov RR, et al. Proc Natl
Acad Sci U S A 2001;98(20):
11047–11054.
210. Wultz B, et al. Behav Neural Biol
1990;53(1):88–102.
211. Sagvolden T, et al. Behav Neural
Biol 1992;58(2):103–112.
212. Russell V, et al. Behav Brain Res
2000;117(1–2):69–74.
213. Michelson D, et al. Am J Psychiatry
2002;159(11):1896–1901.
214. Spencer T, et al. J Clin Psychiatry
2002;63(12):1140–1147.
215. Russell VA. Neurosci
Biobehav Rev 2003;27(7):671–6,
et al.82.
216. Fox GB, et al. Behav Brain Res
2002;131(1–2):151–161.
 Autism and autism spectrum disorders
Chapter
16 Daniel H. Geschwind and Maricela Alarcón
Abstract
Autism is a heterogeneous and broadly defined neu-
rodevelopmental disorder. After a decade or so of
intense genetic investigations, diverse approaches,
from genetic linkage and whole genome association,
to comparative genomic hybridization have begun to
yield significant dividends. These genetic findings
demonstrate more heterogeneity than initially antici-
pated, but have also begun to shed light on potential
neurobiological mechanisms of this vexing spectrum
of conditions.
Introduction
Like other psychiatric conditions, autism is a syn-
drome that is defined by observed behavior and cog-
nition, not etiology. As is described in the DSM-IV-
TR, autism is characterized at its core by deficits in
social communication, language, and the presence of
repetitive, restrictive behaviors, all with onset prior to
age three [1]. Although in the original description of
11 boys with autism in 1943 [2], Kanner clearly high-
lighted what he thought to be its biological etiology,
for many decades subsequently, autism was con-
sidered to be primarily caused by certain aspects of
family environment, including a lack of parental
warmth or “cold mothering”. This sentiment was
widely held in spite of the recognition that autism
had a significant association with many medical con-
ditions having genetic or biochemical etiologies [3, 4].
Rather, many of the rare conditions identified with
autism were thought to be relatively unique circum-
stances, and it was not until the last three decades that
the true nature of its genetic etiology was properly
recognized [5–7].
Subsequently it has become widely accepted that
autism has among the highest genetic liability of any
Principles of Psychiatric Genetics, eds John I. Nurnberger, Jr. and Wade H. Berrettini. Published by Cambridge University
Press. © Cambridge University Press 2012.
neuropsychiatric or neurodevelopmental syndrome
(Table 16.1; Figure 16.1). The evidence for this comes
from three major areas: (1) the existence of known
chromosomal or genetic disorders in which a high
proportion of children have a diagnosis of an autism
spectrum disorder; (2) twin studies clearly show
markedly increased risk for the second twin among
monozygotic twins or a concordance rate of > 70%
for monozygotic twins versus 10% for dizygotic twins;
and (3) the increased risk to first-degree relatives of
autistic probands; that is, the risk to a sibling of an
autistic proband is at least 10–25-fold greater than
other members of the general population [8]. Unlike
schizophrenia, there does not seem to be a large
distinction between the concordance rate among non-
twin siblings and dizogotic twins. Thus, although
there may be a role for shared in utero environment,
it is not as significant a risk factor for autism spec-
trum disorders (ASD) as it may be in schizophrenia
[9]. All of these data clearly indicate that studies that
focus on genetic etiology would prove fruitful in
advancing our understanding of autism.
A brief history of genetic linkage studies
Given the high genetic risk for autism [8], and a
paucity of multigenerational families, linkage studies
using a sibling pair design represented the first
attempts to screen genome-wide for major autism risk
loci (see Table 16.1, which provides a brief history of
genetic studies in autism). Such studies began in
earnest in the 1990s and have culminated in at least
a dozen linkage studies [10–21].
The first linkage study was performed by a collab-
orative group, the International Molecular Genetic
Study of Autism Consortium (IMGSAC) and pub-
lished in 1998. This study, based on a multinational
183
 Chapter 16: Autism and autism spectrum disorders

Table 16.1 Selected historical highlights of genetic studies of autism.

Year Study Contribution

1943 Kanner Described 11 cases of individuals with autism
1977 Folstein and Rutter First twin study of autism
1989 Le Couteur et al. First published diagnostic instrument for autism. Currently
gold standard for research studies
1985/1992 Gillberg and Wahlsrom, 1985; Affiliation of autism with medical conditions and chromosomal
Gillberg and Coleman, 1992 disorders
1994 Bolton et al. First-degree relatives of individuals with autism have mild
cognitive and behavioral traits characteristic of disorder
1995 Bailey et al. Expanded twin study that considered broader phenotypes
of autism
1997 Cook et al. 15q maternally derived duplication causes autism
1997 Geschwind et al. Cure Autism Now established first collaborative gene bank for
the study of autism – Autism Genetic Resource Exchange (AGRE)
is a publicly available database: http://agre.autismspeaks.org
1998 International Molecular Genetic First genome-wide linkage scan for autism
Study of Autism Consortium
(IMGSAC)
1999 Risch et al. Early linkage study that interprets evidence in favor of the
existence of multiple autism genes, each with a small effect
2001 Wassink et al. Documented that up to 9% of cases with autism are due to
chromosomal abnormalities
2002 Alarcón et al. First genome-wide quantitative trait loci scan in autism
2005 Cantor et al. First replication of linkage finding in autism; confirmed 17q21
peak that was attributed to families with male-only affecteds
2006 Jacquemont et al. Microarray comparative genomic hybridization identifies
chromosomal rearrangements in significant proportion of
affected cases and thereby illustrates new methodology for
autism gene search
2007 AGP Consortium; Szatmari et al. Largest linkage study to date. International collaborative effort
of 50 centers from United States and Europe with 1168 families.
Typed for 10 000 markers
2007 Nishimura et al. First blood genomics study with independently validated
results in neural tissue
2007 Sebat et al. Evidence that copy number variants (CNVs) are associated
with autism
2008 Morrow et al. Used homozygosity mapping to detect inherited causal
mutations for autism. Provided further evidence for extreme
heterogeneity, as no loci were shared
2008 Kumar et al.; Weiss et al. 16p11.2 CNV associated with autism
2009 Wang et al. First genome-wide association study that identified significant
common variants

184
 Chapter 16: Autism and autism spectrum disorders

Chr 1 Chr 2 Chr 3 Chr 4 Chr 5 Chr 6 Chr 7 Chr 8 Chr 9 Chr 10 Chr 11 Chr X
1.1 5.1 8.1
3.1 10.1 X.1
9.1
7.1

5.2 11.1
2.1
10.2
3.2 5.3
3.3 7.2 9.2 11.2 X.2
4.1
11.3
4.2 10.3
4.3 6.1 7.4 7.3
X.3
7.5
7.6 9.3 X.4
7.7
3.4 7.8 9.4
6.2 9.5
7.10 7.9 X.5
1.2 7.11 X.6
2.2 2.3
2.4
3.5
3.6 4.4 12.1

1.3 16.1 15.1

2.5 15.2
17.2 12.2
17.1 15.3 13.1
20.1 16.2
19.1 17.3
21.1 20.2 16.3 14.1
17.4
17.5
22.1 16.4
21.2 15.4
22.2

Chr 22 Chr 21 Chr 20 Chr 19 Chr 18 Chr 17 Chr 16 Chr 15 Chr 14 Chr 13 Chr 12 Chr Y

ID Feature Position Refs ID Feature Position Refs ID Feature Position Refs

1.1 Loss 1p36 3 7.4 RELN 7q22 117–121 15.3 Gain 15q11–15q13 4
1.2 Linkage 1q21–1q23 116 7.5 MET 7q31 70,123 15.4 Linkage 15q22–15q26 5
1.3 DISCI 1q42 122 7.6 Loss 7q31 5 16.1 TSC2 16p13 110
2.1 NRXNI 2p16 5,34 7.7 Linkage 7q32–7q34 15 16.2 Loss 16p11 4,20,35,44
2.2 Loss 2p24 4 7.8 CADPS2 7q31 5,45 16.3 Gain 16p11 20,35,44
2.3 Linkage 2p24–2q31 57,58,112 7.9 Linkage 7q34–7q36 14,52,68 16.4 Loss 16q21 5
2.4 SLC 25A12 2p24 124–126 7.10 CNTNAP2 7q35–7q36 37–40 17.1 Loss 17p12 5
2.5 Loss 2q37 4 7.11 EN2 7q36 129,130 17.2 Gain 17p12 107
3.1 OTXR 3p25 127,128 8.1 Gain 8p23 5 17.3 SLC6A4 17q11 131–134
3.2 Loss 3p14 4 9.1 Linkage 9p24 5 17.4 Linkage 17q11–17q21 51,54,135
3.3 Gain 3p14 4 9.2 Loss 9q12 5 17.5 ITGB3 17q21 136,137
3.4 Linkage 3q22 15 9.3 Linkage 9q33 5 19.1 Linkage 19p13 140
3.5 Linkage 3q25–3q27 138,139 9.4 Linkage 9q34 15 20.1 Loss 20p13 5
3.6 Loss 3q27–3q28 3 9.5 TSCI 9q34 110 20.2 Loss 20p13 5

4.1 Loss 4q21 3 10.1 Loss 10q14–10p15 4 21.1 Linkage 21q11 55

4.2 Loss 4q21–23 3 10.2 Gain 10q11–10q21 4 21.2 Loss 21q22 3
4.3 Linkage 4q22–4q25 15 10.3 PTEN 10q23 141 22.1 Loss 22q13 4
4.4 Loss 4q35 5 11.1 Linkage 11q12–11p13 5 22.2 SHANK3 22q13 21,22,142,143
5.1 Linkage 5p15 5 11.2 DHCR7 11q13 108 X.1 NLGN4X Xp22 28
5.2 Linkage 5p13–5q11 140 11.3 Linkage 11q13–11q14 15 X.2 NLGN3 Xq13 28

5.3 Linkage 5q12 5 12.1 CACNAIC 12p13 24 X.3 Linkage Xq21–Xq25 140
6.1 GRIK2 6q21 144–146 12.2 AVPRIA 12p14–12q15 147 X.4 Gain Xq24 3
6.2 AHII 6q23 106 13.1 Gain 13p14 5 X.5 FMRI Xq27 105,148

7.1 Loss 7p21 4 14.1 Linkage 14q23 149 X.6 MECP2 Xq28 109
7.2 Loss 7q11 3 15.1 UBE3A 15q11 102,103
7.3 Linkage 7q22–7q32 52,111–113 15.2 GABRB3 15q12 25,114,115

185
 Chapter 16: Autism and autism spectrum disorders

collaborative effort, consisted of fewer than 100 fam-
ilies, characteristic of genetic studies of neuropsychia-
tric disease at this time [10]. Despite its size, a few
very interesting findings emerged from this study and
are discussed in more detail below. The second major
linkage study came from the Paris Autism Research
International Sibpair Study (PARIS) group [13] and
involved approximately 60 families, identifying no
loci even suggestive of a genome-wide significance
level [22]. The small study was based on the optimis-
tic assumption that there would be only a few major
loci shared by most children with autism, those with
“idiopathic autism”, who did not have another under-
lying genetic disorder. At the time, most cases were
thought to be idiopathic and not to be syndromic
forms, which were considered rare and atypical, des-
pite evidence to the contrary [23]. These assumptions
were prevalent then and had they been true, a few
hundred families would have been sufficient to detect
the genome-wide significant linkage and identify a
gene within a locus.
Important in this regard, was one of the other
early linkage studies, which involved over 100 families
and was performed by a group from Stanford [14].
These investigators found no loci at a genome-wide
suggestive significance level, which they interpreted as
evidence of extreme heterogeneity and suggested the
interaction of multiple genes was necessary to cause
autism in idiopathic cases. The authors concluded
that “positional cloning of susceptibility loci by link-
age analysis may be a formidable task and other
approaches may be necessary”. These conclusions of
multiple interacting genes causing most autism cases
and the need for alternative strategies to identify
them, contrasted with the prevailing wisdom, includ-
ing that of this chapters’ authors; but they have turned
out to be largely correct. Despite a nearly 10-fold
increase in sample size in the most recent whole
genome linkage study, performed by an international
collaborative group of the Autism Genome Project
(AGP), no region reached clear genome-wide signifi-
cance [24].
This is not to say that linkage has been entirely
without success. In 1998, the IMGSAC published a
two-stage genome scan, in which they studied a total
of 99 families from multiple collaborating institutions
in Europe and the United States. They identified
several regions reaching nominal significance and a
broad region on distal chromosome 7q reaching an
MLS of 3.55 in a smaller subset of families from the
United Kingdom. This same region was identified in
further investigations by this group [25] and a study
with an expanded sample revealed a locus on chromo-
some 2q and another on 17q [15]. The original locus
on chromosome 7q was the second most significant,
reaching a LOD score of 3.2 in a region near that of
the original report in a total of approximately 150
sibling pairs (including the original sample). A third
region on chromosome 17 was also identified at a
suggestive level with a maximum LOD score of 2.3
at about 50 cM.
It was noted early on that a region on distal
chromosome 7q reached nominal significance in
nearly every linkage study performed [26], and Bad-
ner and Gershon identified this as the only region
meeting genome-wide significance for autism using
meta-analysis [27]. This region overlaps with the ori-
ginal IMGSAC linkage region and was recently con-
firmed in an independent sample [21].
Following the initial IMGSAC paper, the PARIS
group [13]) performed a genome-wide linkage scan
at < 10 cM density in a small set of multiplex
families. Although none of the regions yielded even
suggestive linkage, several did overlap with other

Figure 16.1 Loci implicated in autism spectrum disorder (ASD) etiology. Entries in the ID column of the table map are entries to the
ideograms of individual chromosomes. Red and yellow bars correspond to de novo losses and gains, respectively, which are observed in cases
but not in controls. Green bars correspond to genes that are observed to modulate ASD risk (either through a rare syndrome or through
genetic association): light green and dark green bars represent promising or probable candidate genes, respectively, as defined in the
table map. Regions shaded in purple correspond to linkage peaks. Only human data were considered in the assembly of the table. AHI1,
Abelson helper integration site 1; AVPR1A, arginine vasopressin receptor 1A; CACNA1C, calcium channel voltage-dependent L type α 1C
subunit; CADPS2, Ca2+-dependent activator protein for secretion 2; CNTNAP2, contactin associated protein-like 2; DHCR7, 7-dehydrocholesterol
reductase; DISC1, disrupted in schizophrenia 1; EN2, engrailed homeobox 2; FMR1, fragile X mental retardation 1; GABRB3, g-aminobutyric
acid (GABA) A receptor β3; GRIK2, glutamate receptor ionotropic kainate 2; ITGB3, integrin β3; MECP2, methyl CpG binding protein 2; MET,
met proto-oncogene; NLGN3, neuroligin 3; NLGN4X, neuroligin 4 X-linked; NRXN1, neurexin 1; OXTR, oxytocin receptor; PTEN, phosphatase
and tensin homologue; RELN, reelin; SHANK3, SH3 and multiple ankyrin repeat domains protein 3; SLC25A12, solute carrier family 25
(mitochondrial carrier, Aralar) member 12; LC6A4, solute carrier family 6 (neurotransmitter transporter, serotonin) member 4; TSC1, tuberous
sclerosis 1; TSC2, tuberous sclerosis 2; UBE3A, ubiquitin protein ligase E3A. (From [8], with permission from Macmillan Publishing Ltd.)
See plate section for color version.

186

suggestive loci identified by the IMGSAC, including
the peaks on 2q, 7q, 16p, and 19p. Three major
studies were published in 1999, one from the
Cooperative Linkage Study in Autism (CLSA) with
75 families [12], the Stanford study with 139 families
[14], and the PARIS group with 51 families [13].
Subsequent linkage studies involved Finnish cohorts
[17] with 38 families, an American group in 2002
with 90 families [18], and the first paper from the
Autism Genetic Resource Exchange (AGRE; [28])
which contained 110 families [16].
It was not until 2003 that a substantial increase in
the number of AGRE families warranted another
genome-wide linkage investigation [19]. The AGRE
project was prompted by previous studies that had
been clearly underpowered and by the fact that a
three- to five-fold increase in sample size would be
necessary to detect loci at genome-wide significance
[22]. Despite this increase in the sample to 345 fam-
ilies, Yonan and colleagues did not detect any signifi-
cant loci. However, several loci, including those on
chromosomes 11 and 5, did garner additional support
as more markers and families were added relative to
the earlier work [16].
These early results provided more evidence for
heterogeneity in autism, which has been bolstered by
recent studies. The University of Washington group’s
genome-wide scan in 200 families [21] revealed the
strongest linkage signal on chromosome 7q which
overlapped with the same region identified previously
[10, 27]. Even more definitive support for heterogen-
eity came from the largest linkage study to date per-
formed by the AGP, an international collaborative
effort of most of the larger autism genetics groups
[24]. Here, despite over 1000 families each with 2
affected individuals and a 10-fold sample increase
over the originally published linkage scans from
almost a decade earlier, there was not an increase in
power substantial enough to detect linkage signals.
Although no signal reached genome-wide signifi-
cance, the most prominent region on chromosome
11p12–13 did reach the suggestive linkage threshold.
The unexpectedly modest results from this inter-
national collaboration could be due to the merging
of samples from distinct geographic regions; the large
sample size may not have compensated for the
increase in heterogeneity introduced by pooling of
the samples.
A thorough summary of all the major linkage
studies has been published and includes a table of
Chapter 16: Autism and autism spectrum disorders
these findings [8]. To date, a very large region on
chromosome 7q likely represents many distinct loci
due to its size and because it has been observed in
multiple studies with independent samples. Another
region on chromosome 17q was identified in the
AGRE cohort by Stone and colleagues [29] with the
striking result that its effect was predominantly due to
male affectation status. Subsequently, this signal was
confirmed in an independent sample, the first formal
confirmation of an autism-susceptibility region at
genome-wide significance [30]. There are several
other regions that have been identified in more than
one study of suggestive linkage [8], but again, none of
these would meet stringent statistical criteria for
genome-wide significance, including those on
chromosomes 5q and 11.
One of the most recent high-resolution genome-
wide linkage and association studies of autism
included multiplex families ascertained from the
AGRE and from the US National Institute for Mental
Health (NIMH) repositories [31]. Two novel putative
linkage regions were identified: chromosomes 20p13
and 6q27. Although there were no markers with evi-
dence for genome-wide association in families, there
were many that just missed the significance threshold
suggesting that the study lacked power to detect
common variants. Using a case-control approach,
the authors did find nominal evidence for association
at several chromosomes (4q13, 5p15, 6p23, 9p24,
9q21, 10q21, and 11p14), but replication analyses only
identified one significant marker – rs10513025 on
chromosome 5p15, about 80 kb upstream of an axo-
nal guidance gene (SEMA5A). No copy number vari-
ations (CNVs) were detected surrounding this SNP or
in the SEMA5A locus. Brain expression for SEMA5A
was reduced in individuals affected with autism com-
pared with controls, although this was a marginal
effect. The authors propose that the inconsistency in
their linkage and association results could be due to
the limitations of the respective methods, in that
linkage scans detect rare variants with strong effects
(20p13), while association tests detect novel common
variants with more modest effects (rs10513025).
All of these linkage studies relied on the qualita-
tive diagnosis of autism for their analysis. Their
results strongly suggest that individual common
genetic risk factors are not likely to cause the entire
core deficits required for the broad diagnosis of
autism or ASD. Thus, we need to define more precise
intermediate phenotypes or endophenotypes that
187
 Chapter 16: Autism and autism spectrum disorders
comprise components of the disorder that might be
more closely related to a few single genes of small effect
size. Endophenotypes are heritable traits characteristic
of the disorder and are present in relatives of affected
individuals more frequently than in the unrelated gen-
eral population. In support of this view, the psychiatric
genome-wide association studies Consortium Steering
Committee recently stated that “Careful attention to
phenotypic measures (rather than reliance on diag-
noses) could prove to be important for identifying
and replicating susceptibility genes” ([32], p 15).
Endophenotypes in ASD
Although it has taken a long time for the autism
genetics community to widely adopt the concept of
intermediate endophenotypes, several groups have
tried to decrease heterogeneity using such phenotypes
to subset patients into potentially more homogeneous
groups. Among the first of these studies that separ-
ated probands (with their families) into groups of
those with and without language delay, resulted in
the identification of a putative locus reaching a sug-
gestive level of significance on chromosome 2q [33].
Bradford et al. used a similar strategy in the CLSA
cohort, identifying loci on chromosomes 7 and 13
[34]. However, subsequent analyses in a larger sample
using the same method of sample stratification based
on language-related endophenotypes while taking
parental language history into account, was unable
to replicate previously reported linkage [35].
Others have successfully used endophenotypes
related to rigidity or obsessive–compulsive disorder
to identify putative linkage regions on chromosomes
1q [36], 15q [37], and 17q [38]. Regression, which is
observed in up to one-third of autistic probands
[39], has also been used to select families for linkage
in the AGRE cohort [40]. This resulted in two seem-
ingly strong linkage signals on chromosomes 21q
and 7q in a very small subset of families. Since the
21q peak is novel, it may be an artifact due to the
small sample size and thus it needs additional
confirmation.
Perhaps the most remarkable findings have
involved using sex as a phenotype for stratification.
This trait was selected based on the observation that
on average males are four times more likely than
females to have autism. Stone et al. reasoned that
the genetic risk for affected male-only containing
families might be distinct from those that contain an
188
affected female [29]. Using this novel approach of
stratification by sex, they reanalyzed data from the
AGRE cohort that had previously been subjected to a
genome-wide linkage analysis [19]. Stone and col-
leagues showed significant enrichment for loci due
to sex related to both female and male sex. Most
importantly, in the male-only families, they identified
a locus with genome-wide significance on chromo-
some 17q. As mentioned previously, Cantor et al.
replicated this finding in an independent sample from
AGRE and performed further fine mapping to narrow
the region to chromosome 17q11 [30]. Although this
is a formal replication on an independent sample,
other studies [21] have not identified the same locus
on chromosome 17, again likely reflecting the signifi-
cant heterogeneity in autism.
Investigators of autism genetics have also used
behavioral endophenotypes (e.g. language deficits) as
covariates in linkage analyses [41, 42] to increase
phenotypic and genetic homogeneity of the sample.
Shao and colleagues applied the ordered subset analy-
sis (OSA) method to a covariate representing “insist-
ence on sameness” derived from a principal
components analysis, and identified a subset of
homogeneous families with autism that were respon-
sible for linkage to a previously reported region on
15q11–13 [37]. And in a novel application of latent
class analysis to autism, Liu and colleagues [43] used
symptom counts from the Autism Diagnostic Inter-
view, Revised (ADI-R) to compute latent class-
derived phenotypes. Nonparametric linkage analysis
of these phenotypes showed stronger signals in
chromosomes 3q24, 6q14, 8p12, and 16p13 than com-
parable analyses using the diagnosis of autism.
Another way to use endophenotype data is to
identify quantitative intermediate phenotypes or
endophenotypes that one can use directly as linkage
variables [44–46]. In other complex diseases, this
strategy has been used for many metabolic parameters
(such as plasma triglyceride concentration [47],
fasting insulin concentration [48], and even gene
expression, so-called eQTLs [49–52]). The use of
eQTLs for gene identification via linkage methods
has already begun in plant and animal models [53]
and holds much promise for human gene expression
traits. In one recent example, a dominant mutation in
a gene that causes retinitis pigmentosa (PRPF31) was
subjected to eQTL analysis; it was found that the
expression of PRPF31, which is correlated with dis-
ease penetrance, was associated with an eQTL in an

8.2 Mb region on chromosome 14q21–23 in 15 CEPH
families [54]. The authors also showed that a nearby
penetrance factor modulated the expression of both
PRPF31 alleles.
With regards to autism, results of the first quali-
tative trait loci (QTL) analysis were published in 2002
and used language-related endophenotypes for analy-
sis. Alarcón et al. used “age at first word” to identify a
locus on chromosome 7q35 [55], which they later
confirmed via additional linkage analyses [56] and
high-density association [57]. In addition, they also
identified a novel locus on chromosome 3q [56],
which awaits independent replication. SNP mapping
of the entire 7q35 QTL region at 5.6 kb density
revealed a polymorphism in contactin associated pro-
tein-like 2 (CNTNAP2) associated with age at first
word, the first successful linkage-directed association
study in autism [57]. Interestingly, as was the case for
the 17q11 locus, male-only containing families were
mainly responsible for the CNTNAP2 association
signal. To test the hypothesis that this language endo-
phenotype was representative of normal population
variation (e.g. [58]), the relationship of this poly-
morphism with specific language impairment was
investigated: association with the same region was
shown in a sample ascertained for specific language
impairment [59]. Even though this CNTNAP2 asso-
ciation with language performance appears to have
been confirmed, the signal from this one gene does
not entirely explain the strong linkage signal on
chromosome 7q35 in the autism sample, indicating
that additional susceptibility loci for autism must
exist within this region.
Other more recent studies have also used QTL
approaches. But, none have identified a region of
genome-wide significance, although certain loci look
promising [21, 43]. The reason for this is probably
manifold, including the fact that the endophenotypes
being studied represent relatively broad cognitive
domains, such as language, and certainly need further
refinement. The problem is that most data that have
been collected in autism are based on diagnostic
interviews and caregiver questionnaires, which may
not have the optimal properties for accurately meas-
uring the most relevant aspects of cognition and
behavior. More measurable and heritable phenotypes
[60], such as brain structure volumes, brain structural
and functional connectivity, and even more gross
phenotypes such as head size, could be interesting
endophenotypes for genetic studies. Further
Chapter 16: Autism and autism spectrum disorders
stratifying of patients on other biomedical parameters
or other medical conditions that co-occur with
autism, such as specific forms of dysmorphology,
seizures, or gastrointestinal dysfunction, could likely
prove fruitful in the future.
Candidate gene association
The vast majority of association studies have involved
the assessment of single candidate genes whose selec-
tion was based either on biological hypotheses or on
published linkage regions. With few exceptions [57],
these investigations have not relied on dense thor-
ough screening of entire regions of linkage, thus
making it very difficult to rank the importance of
such genes relative to each other. Several detailed
reviews of genetic association studies in autism have
been published recently [8, 61]. Thus here we focus
on only a few of the more interesting genes that were
identified based on biological hypotheses, as well as
regional linkage signals. Please refer to Figure 16.1 for
a more comprehensive representation of candidate
genes for autism.
EN2 1and MET. Two genes, EN2 [62] and MET
[63], were interesting candidate genes based on their
roles in neural development and their presence in the
7q2–3 linkage region. In both cases, genetic associ-
ation was identified first using the AGRE sample and
was later replicated in other samples; EN2 by the same
group who found the initial association [64], and
MET by a different group [65]. Notably, with MET,
the same promoter polymorphism that was initially
identified by Campbell and colleagues was not repli-
cated; however, there was other evidence for associ-
ation of a different SNP in the gene – again, this
may reflect the presence of genetic heterogeneity.
In the case of the promoter variant that was originally
identified, Campbell et al. have shown that it is func-
tional, whereas the role of the other MET variant is
unclear. Recently, this group reported an association
of the MET promoter variant with autism in multi-
plex AGRE families with co-occurring gastrointest-
inal dysfunction [66]. In addition, they found no
association with autism in simplex families without
gastrointestinal conditions. Thus, the authors con-
cluded that a disruption in MET signaling may con-
tribute to an increased risk for autism with
gastrointestinal dysfunction. Rather than focus exclu-
sively on a single gene, this line of investigation has
been extended to include other genes that may
189
 Chapter 16: Autism and autism spectrum disorders
interact with MET (such as PLAUR and SERPINE1,
[67]), thereby shifting to a pathway approach.
OXTR and AVPR1a. Other genes of interest that
have shown at least nominal association with autism
in different studies include oxytocin (OXTR) and
arginine vasopressin 1a receptor (AVPR1a). Oxytocin
and vasopressin are interesting because of their roles
in parent–offspring social behavior and maternal
bonding [68, 69]. Copy number variations (CNVs)
have been detected in 20p13 [70, 71], a region that
includes both the OXTR and AVPR1a genes, in indi-
viduals with autism. Oxytocin has been shown to
decrease activation of the amygdala which affects
circuitry involved in fear and trust [72]; moreover,
variants of OXTR and AVPR1a appear to be involved
in amygdala activation and regulation. Variation in
three microsatellites in AVPR1a has not only been
associated with a diagnostic measure of autism
(ADIR), but also with a measure of social skills (the
Vineland Adaptive Behavior Scales) suggesting that
this gene may contribute to the social deficits charac-
teristic of the disorder [73]. Moreover, two of these
variants were also associated with amygdala activation
and personality in unaffected, healthy individuals
[72]. Taken together these studies suggest that the
neural mechanisms that underlie the disorder may
be influenced by the same genes that contribute to
variability in brain function and behavior in healthy
individuals.
SLC6A4. One of the most persistently studied
genes in autism is the serotonin transporter
(SLC6A4). This gene was initially the research focus
of Ed Cook and colleagues [74]. Since this gene is
notoriously difficult to genotype, and evaluating link-
age and association results that were unexpectedly
positive, investigators noticed that genotypes from a
control group were not in Hardy–Weinberg equilib-
rium and thereby suspected genotype errors [75]. The
authors further reported that variation in conditions
for amplifying the SLC6A4 caused genotype errors
[75]. An analysis of the corrected SLC6A4 genotypes
in two autism samples provided no more evidence for
linkage than was originally reported [19] and no
evidence for association [75]. Despite these results,
serotonin remains a very plausible candidate given
its role in some of the cognitive features found in
autism as well as in modulating sleep, which is a
considerable problem for patients with autism [76].
Also, there is evidence from functional imaging stud-
ies that serotonin synthesis during development is
190
disrupted in children with autism, with different
brain regions showing variable levels of this disrup-
tion [77]. Sutcliffe and colleagues [78] investigated the
serotonin transporter in great detail, performing rese-
quencing and identifying several potential deleterious
variants in the gene. However, once again, these rare
variants that potentially cause functional changes
cannot account for the entire linkage signal. It is
remarkable that the serotonin receptor sits directly
under the main linkage peak on 17q11 that is
genome-wide significant in families with male-only
affected [30]. However, given the data that have so far
been published and the fact that both common and
rare variants in this gene cannot account for the 17q
signal, it is highly likely that other polymorphisms
and other genes also contribute to the linkage signal.
What is becoming clear is that when dense SNP
studies of several suggested linkage regions have been
performed, no single common variant can clearly
account for the linkage signal. This includes regions
on chromosome 17 [79] and on chromosome 5
[Stone, Geschwind, and Nelson, unpublished data].
Therefore, there may be many common variants, each
with small effect size, that underlie the linkage peaks
or, alternatively, rare genetic mutations may be con-
tributing more to autism linkage than was previously
suspected.
Whole genome association
Recently, the first genome-wide association study in
autism was performed on the AGRE cohort of 700
multiplex families, consisting of over 3000 subjects.
Replication was performed in a second cohort of over
1200 subjects with autism and  6500 controls, all of
European background. This study was done using the
Illumina HumanHap550 BeadChip from which geno-
types from 486 000 markers were used (http://www.
AGRE.org). In the combined analysis of these two
cohorts, one family based and one control, a SNP
between cadherin 9 and 10 reached genome-wide
significance [80]. This finding was replicated in
another sample of nearly 500 multiplex autism fam-
ilies, genotyped on the Illumina HapMap 1M Bead-
Chip, and again confirmed in another small case-
controlled cohort of 100 cases and 500 control sub-
jects genotyped on a less-dense array platform. The
combined association from pooled analysis revealed
six SNPs in this region on chromosome 5 between
cadherin 9 and 10 with a p-value < 8  10–8. The

same study also showed significant association when
neurexin and cadherin variants were tested as a
group. To perform preliminary functional analysis,
in situ hybridization in developing human fetal brain
was performed. There was significant enrichment of
mRNA expression in the frontal lobe of the develop-
ing human embryo for cadherin 10, but no significant
expression for cadherin 9 at all in mid-gestation fetal
human brain, making it unlikely that cadherin 9 is
responsible. However, further studies will have to be
performed to determine this and to identify the crit-
ical functional variants. Nevertheless, this work rep-
resents the first successful whole genome association
in autism, and demonstrates that enormous samples
are going to be needed to identify variants with
smaller effect sizes. The implication of cadherin 10
adds support to the role of genes involved in various
forms of cell adhesion, such as CNTNAP2 and
Neurexin 1 in ASD.
To address the need for large sample sizes to
continue with genome-wide association attempts,
the AGP Consortium reported a genome-wide associ-
ation study of 1369 ASD families from North Amer-
ica and Europe and 1880 controls genotyped with the
Illumina 1M-Infinium BeadChip array [81]. For this
study, 595 families from the AGRE were used for
independent replication although they were not gen-
otyped on the same platform as the AGP sample. The
Illumina HapMap550 markers (described above) that
were typed in the AGRE families were included in the
1-M array thereby facilitating a procedure used to
infer the missing genotypes. Despite the effort
involved in characterizing the families, genotyping
the samples, and pooling data across continents, only
one SNP (rs4141463) in MACROD2 met genome-
wide significance in the primary analysis under the
specific condition of pooling the AGP families with
the controls. Interestingly, when the exploratory
analysis focused on verbal individuals from the AGP
and AGRE families, two additional SNPs (rs3784730
in ST8SIA2 and rs2196826 in PLD5) had strong asso-
ciation signals. Results of this study strongly suggest
that endophenotypes should be considered an essen-
tial part of the primary, a priori, hypothesis-driven
investigation rather than a secondary part of the
exploratory, post hoc analysis.
In support of the use of endophenotypes in pri-
mary genome-wide association study analyses, Lu and
Cantor increased the power to detect association by
including sex as a risk factor in their logistic
Chapter 16: Autism and autism spectrum disorders
regression analysis [82]. As described previously,
stratification by sex is an approach that has been used
successfully in linkage studies [29], and, not surpris-
ingly, sex also proved to be useful in detecting associ-
ation. Two candidate genes that surpassed the
genome-wide significance threshold were identified:
the first was a novel gene involved in calcium channel
detects, RyR2 (rs6683048), and the second was previ-
ously identified via linkage and association studies,
UPP2 (rs17420138).
Rare versus common variation
and the role of CNV
Similar to other complex genetic diseases, identifying
significant genome-wide linkage and association sig-
nals in autism has been challenging, as the above
discussion demonstrates. Even when significant link-
age signals are identified and variants underlying
these signals are found, none of the common variants
clearly account for the signal, and their effect sizes are
less than 1.5. This pattern is consistent with two
genetic models: one in which many common variants,
each with small effect size, interact to produce the
phenotype, and the second in which the role of rare
variants are predominant. Of course, these models are
presented as dichotomous as possible, but in many
families, both of these and intermediate models are
likely involved [58].
Support for the rare variant hypothesis has
developed over the last two years with the identifica-
tion of significant contributions from de novo struc-
tural chromosomal or CNVs [70]. Sebat and
colleagues identified rare chromosomal structural
variation in 10% of simplex autism cases (i.e. sporadic
autism) and 3% of the multiplex cohort (i.e. families
including at least 2 affected members). A subsequent
study has identified a similar proportion of de novo
rare variation in large autism cohorts [71], while a
recent report using the Illumina Infinium 1-M
genome-wide microarray did not detect a difference
in de novo CNV rates between simplex and multiplex
families with autism [83]. Moreover, Pinto and col-
leagues found that genes that were previously impli-
cated in ASD were disrupted by rare CNV
significantly more frequently in cases versus controls.
In addition, rare CNV (i.e. those that occurred only
one time) were also present at higher rates in cases
than in controls, although this effect was marginal.
These CNV implicated novel ASD genes including
191
 Chapter 16: Autism and autism spectrum disorders
SHANK2, SYNGAP1, and DLGAP2. Genes that over-
lapped with CNV and that had a similar function or a
common pathway were examined as gene-sets and,
subsequently, many novel ASD pathways such as
GTPase/Ras signaling, cellular proliferation, projec-
tion, and motility were suggested by this approach.
Further work is needed to replicate these novel path-
ways and assess their generalizability.
As studies progress, sample sizes become larger,
and the platform resolution continues to increase,
more and more contribution from rare variation will
be detected [84]. Some models have been proposed to
suggest that the majority of autism can be attributed
to Mendelian inherited factors [85]. However, irre-
spective of the model, these studies continue to iden-
tify dozens of new potential autism candidate genes,
regions, and pathways providing a significant advance
for the field.
In this regard, the recent identification of several
forms of recurrent chromosomal structural variation,
including 16p11 [86] and 22q11–13 [87], in addition
to the 15q11–13 maternally inherited duplications
discovered by Ed Cook and colleagues over a decade
ago [88], also represent important advances. Each of
these recurrent variations accounts for about 1% of
ASD, so genetic testing to identify such variants is
warranted. The challenge of moving to individual
genes from these findings is that most of these dupli-
cations or deletions involve multiple genes and there-
fore identifying which gene is responsible requires
large-scale mutation screening and further study in
larger populations. Again, the use of high resolution
arrays and the focus on deletions or duplications of
specific exons or genes, in addition to resequencing,
may help lead to the culprits in these cases. At 22q11–
13, one of the primary candidates is shank3, a synap-
tic adapter protein. Rare point mutations in shank3
have been identified, but when CNV and point muta-
tions involving shank3 are combined, they are
thought to account for between 0.5 and 1.0% of
ASD [89, 90].
The identification of each of the recurrent muta-
tions that underlie ASD risk is an important advance;
but, it is also becoming clear that no single gene
accounts for a major proportion of autism cases,
let alone even 5%. Furthermore, some of these recur-
rent CNV, such as those at 16p, are risk factors for a
broad range of neurodevelopmental disorders, includ-
ing schizophrenia [84, 91]. This is consistent with the
known contribution of many single gene disorders
192
causing intellectual disability that also lead to autism
in a small percentage of cases, such as fragile X, Joubert
syndrome, or Smith–Lemli–Opitz syndrome. To date,
over 100 disease genes and 44 genomic loci involved
with intellectual disability have been identified and
may also contribute to the heterogeneity of ASD [92].
Although a significant proportion of patients with
each of these disorders have an ASD, each accounts
for no more than 1% of autism cases [8].
Is there a single pathway or coherent
biochemical explanation?
These new genetic findings beg the question: Despite
the enormous heterogeneity of autism, are there pat-
terns of genes or pathways that are implicated in this
disorder? It is probably too early to state whether
these exist or not, but certain patterns are emerging,
such as the role of specific genes involved in neural
development (e.g. neural adhesion molecules, as dis-
cussed above). The genes identified so far relate to
many distinct biochemical and biological functions.
Recently, we performed an Ingenuity network analy-
sis with a subset of the most probable candidate genes
to date. This preliminary analysis demonstrates
remarkable connections among many different
groups of genes [93]; however, these Ingenuity path-
ways rely on interactions in many different cell types
and much further investigation is needed to deter-
mine whether these genes are related to each other.
To consider the existence of genetic pathways we
must bear in mind that there may be only a few ways
in which the brain can respond to a wide variety of
developmental insults. It is also important to note that
the functions of social cognition, language, and
mental flexibility, which likely underlie the repetitive,
restrictive behavior of autism, represent a huge chunk
of the early brain’s repertoire; therefore many distinct
pathways can lead to dysfunction in these areas. Fur-
ther, though many pathways may be involved, it is
also clear that there is a specific anatomy to social
cognition in the human brain, and this relies heavily
on anterior temporal and anterior frontal lobes and
their interconnections with sub-cortical circuits
including the striatum. These same regions are also
important for language and have been implicated in a
variety of imaging studies [94–96]. This, coupled with
the anatomical restriction of several major autism
candidate genes to anterior regions, suggests that
perhaps a critical component of this disorder is
 Chapter 16: Autism and autism spectrum disorders

what is called the “developmental disconnection”
of these important frontal circuits [97]. For example,
not only is contactin associated protein-like 2
(CNTNAP2), strongly enriched in human frontal
cortex [57, 98], but so is cadherin 10 [80] and MET
[99]. Modification of the proper development of
human frontal and anterior temporal lobe circuitry
via mutations or polymorphisms in these genes pro-
vides one potential mechanism by which genetic risk
factors could influence specific brain circuits that are
involved in the cognitive and behavioral phenotype
of autism.
It is notable that the hypothesis of developmental
disconnection represents a convergence of many years
of imaging and electrophysiological studies based on
the pioneering work by Marcel Just and colleagues
[100–102], which has implicated functional connect-
ivity in the brain as a potential cause of the autism
phenotype. Genetic evidence now seems to support
this hypothesis, as deficits in many distinct levels of
molecular function, from synapse to neuronal migra-
tion to neuronal path finding, as well as neuronal
maturation, have been implicated, and all could con-
verge on disconnection [97]. These ideas and data
suggest that in addition to molecular cellular studies
of genes they must also be put into an anatomical
context, since autism likely reflects the disruption of a
variety of specific circuits. Given its clear phenotypic
heterogeneity, it is also likely that there is a wide
variety of distinct brain regions involved; for the
disorder to comprise deficits in social cognition and
language at its core there must be some convergence
in frontal systems. Involvement of these brain regions
is also seen in some of the single gene disorders that
are related to autism, such as fragile X and others
[103]. This may also explain why some of the same
risk factors that underlie autism, such as neurexin 1
and perhaps others, are also observed in other dis-
orders, such as schizophrenia [91, 104], which also
have neurodevelopmental etiologies and also involve
frontal systems. It is further likely that disorders that
phenotypically overlap with autism, such as attention-
deficit hyperactivity disorder (ADHD), share some
genetic risk with autism; one can imagine that there
will be other discrete genetic components that will
distinguish these disorders from each other despite a
potentially overlapping core set of genes affecting
these circuits [91]. In support of this hypothesis,
Ronald and colleagues reported that there are
common genetic influences on ADHD behaviors
and autism traits both in a community sample of
twins and in a subsample of the extreme individuals
that are likely comorbid for these disorders. To con-
tinue to move forward in autism genetics, the field
needs very large samples, some of which should
include undiagnosed family members, that should be
phenotyped using instruments that measure concepts
and constructs related to a broad range of cognitive
functions that overlap with other disorders such as
ADHD, bipolar disorder, and schizophrenia. Func-
tional imaging studies connecting gene to brain in
humans, and the use of animal models to incorporate
molecular and developmental mechanisms, are both
very exciting, and given the number of genes that
have been identified, there will likely be an explosion
of such studies over the next several years.

References 5. Ritvo ER, et al. Am J Psychiatry

1. American Psychiatric Association.
Diagnostic and Statistical
Manual of Mental Disorders,
Fourth Edition, Text Revision
(DSM-IV-TR). Washington, DC:
American Psychiatric
Association; 2000.
2. Kanner L. Nervous Child
1943;10:217–250.
3. Gillberg C. J Intellectual Disabilities
Res 1992;36:201–214.
4. Gillberg C. et al. (eds.). The
Biology of the Autistic Syndromes.
Vol. 126. New York: MacKeith
and Cambridge University
Press; 1992.
1990;147:1614–1621.
6. Rutter M. J Abnorm Child Psychol
2000;28:3–14.
7. Folstein SE, et al. Nat Rev Genet
2001;2:943–955.
8. Abrahams BS, et al. Nat Rev Genet
2008;9:341–355.
9. Patterson PH. Behav Brain Res
2009;204:313–321.
10. IMGSAC. Hum Mol Genet
1998;7:571–578.
11. Ashley-Koch A, et al. Genomics
1999;61:227–236.
12. Barrett S, et al. Am J Med Genet
1999;88:609–615.
13. Philippe A, et al. Hum Mol Genet
1999;8:805–812.
14. Risch N, et al. Am J Hum Genet
1999;65:493–507.
15. IMGSAC. Am J Hum Genet
2001;69:570–581.
16. Liu J, et al. Am J Hum Genet
2001;69:327–340.
17. Auranen M, et al. Am J Hum
Genet 2002;71:777–790.
18. Shao Y, et al. Am J Med Genet
2002;114:99–105.
19. Yonan AL, et al. Am J Hum Genet
2003;73:886–897.
20. McCauley JL, et al. BMC Med
Genet 2005;6:1.

193
 Chapter 16: Autism and autism spectrum disorders
21. Schellenberg GD, et al. Mol
Psychiatry 2006;11:1049–1060.
22. Lander E, et al. Nat Genet
1995;11:241–247.
23. Gillberg C, et al. Dev Med Child
Neurol 1985;27:293–304.
24. Szatmari P, et al. Nat Genet
2007;39:319–328.
25. IMGSAC. Hum Mol Genet
2001;10:973–982.
26. Cook EH Jr. Child Adolesc
Psychiatr Clin N Am 2001;
10:333–350.
27. Badner JA, et al. Mol Psychiatry
2002;7:56–66.
28. Geschwind DH, et al. Am J Hum
Genet 2001;69:463–466.
29. Stone JL, et al. Am J Hum Genet
2004;75:1117–1123.
30. Cantor RM, et al. Am J Hum
Genet 2005;76:1050–1056.
31. Weiss LA, et al. Nature
2009;461:802–808.
32. Psychiatric GWAS Consortium
Steering Committee. Mol
Psychiatry 2009;14:10–17.
33. Buxbaum JD, et al. Am J Hum
Genet 2001;68:1514–1520.
34. Bradford Y, et al. Am J Med Genet
2001;105:539–547.
35. Spence SJ, et al. Am J Med
Genet B: Neuropsych Genet
2006;141B:591–598.
36. Buxbaum JD, et al. Mol Psychiatry
2004;9:144–150.
37. Shao Y, et al. Am J Hum Genet
2003;72:539–548.
38. McCauley JL, et al. Am J Med
Genet 2004;127B:104–112.
39. Richler J, et al. J Autism
Dev Disord 2006;36:
299–316.
40. Molloy CA, et al. Mol Psychiatry
2005;10:741–746.
41. Dementieva Y, et al. Proceedings
of the IMFAR. Orlando: IMFAR;
2002.
42. Shao Y, et al. Am J Hum Genet
2002;70:1058–1061.
194
43. Liu XQ, et al. Biol Psychiatry
2008;64:561–570.
44. Streeten EA, et al.
J Bone Miner Res
2006;21:1433–1442.
45. Doyle AE, et al. Am J Med
Genet B: Neuropsych Genet
2008;147B:1399–1411.
46. Mottl AK, et al. Kidney Int
2008;74:1185–1191.
47. Coletta DK, et al. Diabetes
2009;58:279–284.
48. Duggirala R, et al.
Am J Hum Genet
2001;68:1149–1164.
49. Cheung VG, et al. Nature
2005;437:1365–1369.
50. Evans DM, et al. Trends Genet
2006;22:350–354.
51. Nica AC, et al. Hum Mol Genet
2008;17:R129–R134.
52. Cookson W, et al. Nat Rev Genet
2009;10:184–194.
53. Schadt EE, et al. Nature
2003;422:297–302.
54. Rio Frio T, et al. Hum Mol Genet
2008;17:3154–3165.
55. Alarcón M, et al. Am J Hum Genet
2002;70:60–71.
56. Alarcón M, et al. Mol Psychiatry
2005;10:747–757.
57. Alarcón M, et al. Am J Hum Genet
2008;82:150–159.
58. Geschwind DH. Cell
2008;135:391–395.
59. Vernes SC, et al. N Engl J Med
2008;359:2337–2345.
60. Glahn DC, et al. Hum Brain Mapp
2007;28:488–501.
61. Losh M, et al. J Neuropathol
Exp Neurol 2008;67:
829–837.
62. Gharani N, et al. Mol Psychiatry
2004;9:540.
63. Campbell DB, et al. Proc Natl
Acad Sci U S A 2006;103:
16834–16839.
64. Benayed R, et al. Am J Hum Genet
2005;77:851–868.
65. Sousa I, et al. Eur J Hum Genet
2009;17:749–758.
66. Campbell DB, et al. Pediatrics
2009;123:1018–1024.
67. Campbell DB, et al. Autism Res
2008;1:159–168.
68. Nelson EE, et al. Neurosci
Biobehav Rev 1998;22:437–452.
69. Heinrichs M, et al. Prog Brain Res
2008;170:337–350.
70. Sebat J, et al. Science
2007;316:445–449.
71. Marshall C, et al. Am J Hum Genet
2008;82:477–488.
72. Meyer-Lindenberg A,
et al. Mol Psychiatry
2009;14:968–975.
73. Yirmiya N, et al. Mol Psychiatry
2006;11:488–494.
74. Cook EH Jr., et al. Mol Psychiatry
1997;2:247–250.
75. Yonan AL, et al. Psych Genet
2006;16:31–34.
76. Liu X, et al. Child Psychiatry Hum
Dev 2006;37:179–191.
77. Chugani DC. Mol Psychiatry
2002;7:S16–S17.
78. Sutcliffe JS, et al. Am J Hum Genet
2005;77:265–279.
79. Stone JL, et al. Hum Mol Genet
2007;16:704–715.
80. Wang K, et al. Nature 2009;459:
528–533.
81. Anney R, et al. Hum Mol Genet
2010;19:4072–4082.
82. Lu A, et al. Mol Psychiatry
2012;17:215–222.
83. Pinto D, et al. Nature 2010;466:
368–372.
84. Morrow EM. J Am Acad Child
Adolesc Psychiatry 2010;
49:1091–1104.
85. Zhao X, et al. Proc Natl
Acad Sci U S A
2007;104:12831–12836.
86. Weiss LA, et al. N Engl J Med
2008;358:667–675.
87. Manning MA, et al. Pediatrics
2004;114:451–457.

88. Cook EH Jr., et al. Am J Hum
Genet 1997;60:928–934.
89. Durand CM, et al. Nat Genet
2007;39:25–27.
90. Moessner R, et al. Am J Hum
Genet 2007;81:1289–1297.
91. Walsh T, et al. Science 2008;
320:539–543.
92. Betancur C. Brain Res 2011;
1380:42–77.
93. Bill BR, et al. Curr Opin Genet Dev
2009;19:271–278.
Chapter 16: Autism and autism spectrum disorders
94. Kleinhans NM, et al. Brain Res
2008;1221.
95. Mundy P. J Child Psychol
Psychiatry 2003;44:793–809.
96. Redcay E, et al. Biol Psychiatry
2008;64:589–598.
97. Geschwind DH, et al. Curr
Opin Neurobiol 2007;
17:103–111.
98. Abrahams BS, et al. Proc Natl
Acad Sci U S A 2007;104:
17849–17854.
99. Mukamel Z, et al. J Neurosci,
2011;31:11437–11442.
100. Hughes JR. Epilepsy Behav 2007;
11:20–24.
101. Just MA, et al. Cereb Cortex 2007;
17:951–961.
102. Koshino H, et al. Cereb Cortex
2008;18:289–300.
103. Abrahams BS, et al. Arch Neurol
2010;67:395–399.
104. Cantor RM, et al. Neuron 2008;
58:165–167.
195
 The genetics of bipolar disorder
Chapter
17 John R. Kelsoe
Introduction
The heritable nature of mood disorders has been
observed since ancient times. Over the last century
great advances have been made in understanding the
genetic transmission and more recently identifying
some of the responsible genes. Study of the genetics
of bipolar disorder has great potential to improve our
understanding of etiology and mechanism and may
lead to new treatments and methods of diagnosis. In
this chapter, the family studies supporting the role of
genetics will be reviewed and recent molecular genetic
results discussed. The possible role of this informa-
tion in clinical treatment is also considered.
Family epidemiology
Family studies
Numerous family studies over many years have docu-
mented the familiality of bipolar disorder. In this
design, typically, each proband in a clinic or other
source is systematically queried regarding psychiatric
illness in their first-degree relatives. Though such
studies are limited by recollection of the proband
and knowledge of family members, these studies have
provided much evidence regarding the genetic trans-
mission of bipolar disorder. Table 17.1 summarizes a
selection of 20 of these family studies, including over
11 000 relatives, that together show a risk of approxi-
mately 7% in first-degree relatives. Compared to the
general population rate of approximately 1%, this
indicates a 7-fold increase in risk and a substantial
level of familiality. This increase in risk is consistently
observed despite the variability in diagnostic methods
employed amongst these studies. It is also of interest
that unipolar disorder (UPD) or major depression
(MDD) is observed at roughly twice the rate as in
Principles of Psychiatric Genetics, eds John I. Nurnberger, Jr. and Wade H. Berrettini. Published by Cambridge University
Press. © Cambridge University Press 2012.
196
the general population, suggesting a partially shared
genetic basis for these two mood disorders. The study
by Wozniak et al. further addresses risk in the families
of probands with childhood onset [1], where an 18-
fold elevation in risk is observed compared to the
general population. This risk is 2–3 times higher than
in adult-onset probands and reflects a substantially
higher genetic loading.
Twin studies
Family transmission, however, is not equivalent to
genetic contribution. Families share many factors
besides genes, such as behavioral environment, or
alternatively, diet or infectious factors. Twin studies
have been the preferred strategy for separating nature
and nurture. Typically, monozygotic (MZ) and dizy-
gotic (DZ) twins are ascertained where one twin has
bipolar disorder, and the rate of illness in the co-twin,
or concordance rate, is determined. In the most
common design, twins are selected who were raised
together in the same household in order to control for
environmental effects. Table 17.2 illustrates many of
the twin studies conducted in bipolar disorder or
affective illness over the last 80 years. Though, there
is much methodological and diagnostic variability
between studies, overall they demonstrate an approxi-
mate 70% concordance rate for MZ compared to 25%
for DZ twins, thereby arguing for a substantial herit-
ability of bipolar disorder. That all co-twins are not
affected is evidence for reduced penetrance of bipolar
susceptibility genes, and implies that other nonheri-
table factors also play a role in etiology.
Mode of transmission
Though, in principle, one should be able to examine
the patterns of distribution of illness in families and
 Chapter 17: Genetics of bipolar disorder

Table 17.1 Morbid risk among first-degree relatives of probands with bipolar disorder.

Study Relatives at risk Morbid risk (%)

Bipolar Unipolar
Perris, 1966 [100] 627 10.2 0.5
Winokur et al., 1967 [101] 167 10.2 20.4
Mendlewicz et al., 1974 [102] 606 17.7 22.4
Goetzl et al., 1974 [103] 212 2.8 13.7
Helzer et al., 1974 [104] 151 4.6 10.6
Gershon et al., 1975 [105] 341 3.8 6.8
James et al., 1975 [106] 239 6.4 13.2
Johnson et al., 1977 [107] 126 15.5 19.8
Petterson, 1977 [108] 472 3.6 7.2
Smeraldi et al., 1977 [109] 172 5.8 7.1
Angst, 1980 [110] 400 2.5 7
Dunner et al., 1980 [111] 1199 4.2 8.2
Taylor et al., 1980 [112] 600 4.8 4.2
Fieve et al., 1984 [113] 2171 6.6 9
Gershon et al., 1982 [5] 598 8.0 14.9
Tsuang et al., 1985 [114] 608 3.9 9.1
Rice et al., 1987 [4] 557 5.7 23
Sadovnick et al., 1994 [115] 1102 3.5 5.7
Birmaher et al., 2009 (Offspring) [116] 388 10.6 24.3
Wozniak et al., 2009 (Child probands) [1] 487 18 49
Total 11223 6.9 12.1

Table 17.2 Twin studies of bipolar disorder.

Monozygotic Dizygotic

Concordant Total % Concordant Total %

Luxenberger, 1930 [117] 3 4 75.0 0 13 0.0
Rosanoff et al., 1935 [118] 16 23 69.6 11 67 16.4
Slater, 1953 [119] 4 7 57.1 4 17 23.5
Kallman, 1954 [120] 25 27 92.6 13 55 23.6
Harvald et al., 1965 [121] 10 15 66.7 2 40 5.0
Allen et al., 1974 [122] 5 15 33.3 0 34 0.0
Bertelsen, 1979 [123] 32 55 58.2 9 52 17.3
Kendler, 1993 [124] 107 154 69.5 114 326 35.0
Total 202 300 67.3 153 604 25.3

197
 Chapter 17: Genetics of bipolar disorder
infer mode of transmission, in practice this approach
has yielded limited information. Though such segre-
gation analyses can be quite successful in simple
Mendelian traits, their relative lack of success in bipo-
lar disorder is consistent with a large amount of
epidemiological and molecular data arguing for a
complex form of transmission involving many genes.
Attempts to identify a Mendelian form of transmis-
sion in family data have provided some modest sup-
port for a dominant single major locus effect [2–4].
Alternatively, other investigators have examined
family data for evidence of multifactorial or polygenic
transmission in which numerous genes of small effect
combine to produce liability to illness. Such models
are usually applied to quantitative traits, and may be
consistent with the range of severity of illness seen in
family members. A related model is the multithres-
hold multifactorial model, in which numerous genes
of small effect contribute to a latent liability to
psychiatric illness that is normally distributed in
the population. Individuals carrying more suscepti-
bility alleles are affected with more severe forms of
illness [5].
Family studies have revealed some other interest-
ing aspects of the genetic transmission of bipolar
disorder. The cohort effect describes the observation
that the age of onset of bipolar disorder was earlier for
cohorts born later in the twentieth century than those
born earlier [4]. The cause of this well replicated effect
is unclear. It may simply be differences in recollection
of onset events in older versus younger people. How-
ever, it has also been proposed that this is an indica-
tion of the phenomenon of anticipation. Anticipation
is a hallmark of genetic disorders in which the patho-
genic mutation is an expanding trinucleotide repeat,
such as fragile X syndrome or Huntington’s disease.
McInnis et al. [6] have reported a decrease in the age
of onset and increase in disease severity in successive
generations. Though the family data seemed intri-
guing, molecular studies have not been successful in
clearly identifying a trinucleotide repeat expansion
sequence associated with bipolar disorder [7].
Another complicating phenomenon in the genet-
ics of bipolar disorder is assortative mating [8–10].
This is the tendency for patients with affective dis-
orders to marry spouses who also have affective dis-
orders, thereby, producing bilineal families where
illness comes from both the mother’s and father’s side
of the family. This presumably results in greater gene-
tic loading, and more complex genetics where
198
multiple genes may be contributing to illness. Evi-
dence for a parent-of-origin effect has also been
reported for bipolar disorder. Maternal transmission
has been reported to be more common than paternal
transmission [11]. This would be consistent with two
atypical forms of genetic transmission: genomic
imprinting or mitochondrial transmission. Genomic
imprinting is seen in disorders such as Prader–Willi
or Angelman syndrome where different syndromes
result from transmission of the mutation through
the father or the mother. However, maternal trans-
mission has not been consistently observed to be
increased [12]. Despite continued study, it has been
difficult to obtain molecular evidence for either
imprinting or mitochondrial transmission. However,
the area of epigenetics in general is a very active area
of investigation.
Do subphenotypes breed true?
Bipolar disorder manifests itself in a broad range of
symptoms and presentations. An important question
is whether any of these features or presentations of
bipolar disorder represent distinct gene effects or
genetically distinct forms of illness. For example, do
different genes predispose to bipolar I versus bipolar
II disorders? If bipolar disorder can be broken down
into different subphenotypes that are genetically dis-
tinct, then each would potentially represent a simpler
problem for gene mapping studies. Furthermore,
knowledge of which genes are involved in different
forms of illness might lead to an understanding of the
different biological mechanisms operating in those
forms of illness. Family studies of different types or
features of bipolar disorder have been conducted in
order to identify those forms of illness that might be
genetically distinct. If such a subphenotype is prefer-
entially transmitted in families or breeds true, then it
might result from a distinct set of predisposing genes.
A variety of clinical features of bipolar disorder
have been examined for their validity as subpheno-
types. Studies of the bipolar II phenotype suggest that
it may be partially genetically distinct. Affected family
members of bipolar I probands primarily exhibit an
increase in bipolar I illness, while family members of
bipolar II probands are more likely to have bipolar II
[13]. Psychotic features have been reported to occur
more often in affected relatives of psychotic bipolar
probands than nonpsychotic probands [14, 15].
Familiality has also been reported for bipolar disorder
 Chapter 17: Genetics of bipolar disorder

3.5 Figure 17.1 Familiality of different

Adjusted for sex, age, Dx subphenotypes of bipolar disorder. The
3 odds ratio for different subphenotype are
Adjusted for sex, age, Dx, site/wave presented calculated as the risk to other
2.5 bipolar family members in families in
which the proband has the
Odds ratio

2
subphenotype of interest. Hatched bars
indicate the odds ratio when covaried for
1.5
diagnosis (Dx), age, and sex. The solid
bars indicate a generally lower odds ratio
1
when site and wave are also included
.67
as covariates. Stars indicate
subphenotypes with at least nominal
significance. MDE, major depressive
episode. (Adapted from [19].)
m er is ts pt d s g g E DE
lis rd os gh tem oo ode clin chin DE DE DE DE MD DE
ho di
so ch ou at
m is cy it n
M
n
M
n
M
n
M
in n
M
in
M
lco c Ps
y
lt
h
de b l e e p
pi
d s w i
a
i
a
i
e
i e y
i s
a ni da ita nic d AM ni ni tit tit rg es
bi
d pa ici ici irr a Ra oo in som som ppe pe ha sn
or id Su Su i th d m m s e n r a ap L et l es
rb i d I e t
om o w e p wo
r p ed e d s
C m ia Mix Ra Hy eas as Re
Co an d re
M oo e cr I nc
M D

Subphenotypes

with comorbid panic attacks [16], suicidality, and temperaments have been shown to have different
rapid mood switching [17]. Several studies have courses of illness [22]. Evans et al. [23] examined five
investigated a number of traits systematically. Schulze temperaments (dysthymia, hyperthymia, cyclothymia,
et al. [18] examined 1246 individuals in 172 families irritable, and anxious) as assessed by the TEMPS-A
for 32 clinical variables from the diagnostic interview. and found evidence for familiality for cyclothymia,
Five clinical features were significantly correlated hyperthymia, and anxious temperaments.
between family members after correction for multiple
comparisons. These included: social relations, sub-
stance abuse, alcoholism, psychosis, and suicide Gene mapping methods
attempts. A similar analysis was conducted by Saun- In the last two decades, genetic studies of bipolar
ders et al. [19], who examined 1416 bipolar siblings in disorder, as with other genetic disorders, have focused
589 sibships which were part of the National Institute largely on using molecular genetic mapping methods
of Mental Health (NIMH) Genetics Initiative collec- to identify the specific genes responsible for the herit-
tion. As shown in Figure 17.1, rapid cycling showed ability described above. The human genome project
the strongest evidence for familiality. Other signifi- has vastly accelerated this process by providing power-
cantly familial traits included: comorbid alcoholism, ful tools for gene identification. A detailed review of
comorbid panic disorder, psychosis, suicidal mapping methods can be found elsewhere in this
thoughts, and rapid mood switching. Similar results volume, but these methods will be briefly described
were obtained by Potash et al. in their examination of here. The real power of these mapping methods is: (1)
the Phenome Database. They in particular reported their ability to identify novel genes not previously
strong familiality for missed work, obsessive–compul- suspected; and (2) their comprehensive coverage of
sive disorder (OCD), hospitalization, and age at onset all genes in the genome. This has led in psychiatric
[20]. These studies suggest subphenotypes of bipolar disorders and many other genetic disorders to the
disorder that may be useful in mapping genes and discovery of completely new genes not previously
understanding the pathophysiology. implicated in illness. This in turn spawns new hypoth-
Affective temperament is another feature of eses regarding these genes, and new ideas about mech-
bipolar disorder that has been hypothesized to reflect anism. Prior to this approach, most hypotheses of
genetically distinct subphenotypes. In this model, pathophysiology were based on the mechanism of
affective temperaments are genetically transmitted action of medications effective for the disorders. This
and in turn predispose to episodes of mania or depres- has been useful, but has restricted the range of hypoth-
sion [21]. Bipolar patients with different underlying eses to a few systems affected by these drugs.

199
 Chapter 17: Genetics of bipolar disorder
There are two major approaches to genetic map-
ping that are employed. Genetic linkage examines the
co-segregation of anonymous DNA markers with ill-
ness in families. If an allele of a marker is consistently
transmitted from affected parent to affected offspring
in a family, this suggests that a gene for the disease
maps near that marker on a chromosome. If the
chromosomal location of the marker is known then
the approximate location of the gene is known. Practic-
ally, this has been accomplished by using 400–500 DNA
markers spread evenly across the genome at about
10 million base pair (Mb) intervals. Historically, this
has been done using markers that have a highly variable
number of tandem repeats. More recently, dense maps
of single nucleotide polymorphisms (SNPs), or single
base substitutions have been used as well. Such studies
would approximately localize a gene. More detailed fine
mapping studies were then required to identify the
exact gene in the region. The Human Genome Project
made available dense maps of such markers, and later
maps of the genes within the linkage peaks.
Genetic association is an alternative approach that
compares the frequency of a given marker allele in
cases versus controls. If a specific allele occurs more
often in cases than controls, this suggests that that
marker or a mutation very near it contributes to
susceptibility. A key difference between linkage and
association is the genetic resolution. Linkage can be
detected if the marker is within approximately 10 Mb
of the causative mutation. Association is a much
higher resolution technique and requires generally
that the marker is within 2–100 kilobases (kb) of the
mutation. Association can detect nearby mutations
because of the phenomenon of linkage disequilibrium.
If a new mutation causing a disease occurs on a
chromosome very near a marker SNP, then it will
tend to remain associated with the marker allele as
the population expands. The closer they are together,
the fewer opportunities for recombination or other
events to disconnect them. This is statistically evalu-
ated by observing whether the two alleles (disease and
marker) occur together more often than expected by
chance. Comparing the two approaches, association
generally is able to identify specific genes, while link-
age identifies much broader regions. As a result, the
entire genome can be tested for linkage with several
hundred markers, whereas, many more markers are
required to test the whole genome for association.
This has made genome-wide association studies
(GWAS) impractical until recently.
200
GWAS were proposed over a decade ago [24], but
awaited two developments to become practical. First,
the Human Genome Project identified millions of
SNP markers suitable for covering the genome at high
density. Secondly, microarray-based methods were
developed that enabled the inexpensive and rapid
genotyping of hundreds of thousands of markers
simultaneously. Currently, GWAS typically genotype
approximately 1 million markers at an average inter-
val of 3 kb using 1 of 2 genotyping platforms.
A strength of the association method is its ability to
detect small gene effects, i.e. mutations that contrib-
ute only a small amount to disease risk. While linkage
has been successfully used for Mendelian traits, in
which one gene is primarily responsible for genetic
transmission; association is ideally suited to detect
polygenic transmission, where many genes each con-
tribute a small amount to an overall cumulative sus-
ceptibility. This is also referred to as the common
variant–common disease model. However, an issue
with GWAS is that very large sample sizes may be
required to detect such small gene effects. Also, very
low p-values are required to survive the multiple
comparisons correction for the one million statistical
tests conducted on datasets of one million or more
markers. The standard threshold for genome-wide
significance is p < 5
108, which is roughly the
Bonferroni correction.
The future of genetic mapping is whole genome
sequencing. This will allow for complete detection of
all genomic variation. Rapid and inexpensive “next-
generation” sequencing methods are evolving quickly
and sequencing costs are dropping rapidly. Soon, all
previous methods will be replaced by the collection of
complete sequence information. This approach will
be required to test the multiple rare variant model of
transmission. This model posits that numerous rare
variants of strong effect cause genetic vulnerability.
Many different mutations in a number of different
genes transmit susceptibility, but each mutation may
be so rare as to essentially be a “private mutation” in
that family. Genes involved in illness may be detected
by demonstrating a larger number of rare likely func-
tional mutations in cases as compared to controls.
Such studies are just now beginning.
Linkage studies
Linkage studies in bipolar disorder began in 1987
with the report of linkage to chromosome 11p15 in
 Chapter 17: Genetics of bipolar disorder

1 2 3 4 5 6 7 8 9 10 11
4 8
1
1

1
13
13 12
1
2
11

18
15
14

9
1
5
10 2
16 11 6
9
3 1
6

12 13 14 15 16 17 18 19 20 21 22 X Y

Figure 17.2 Chromosomal regions implicated by linkage studies. Chromosomal regions reported to be linked to bipolar disorder are
illustrated. Circles represent individual study reports, stars represent regions implicated by a meta-analysis or combined analysis. Numbers
in the circles or stars refer to referenced papers as follows: 1 ¼ Nurnberger et al. [180], Detera-Wadleigh et al. [181], Rice et al. [182],
Edenberg et al. [183]; 2 ¼ Kelsoe et al. [184]; 3 ¼ Detera-Wadleigh et al. [185]; 4 ¼ Blackwood et al. [186]; 5 ¼ Berrettini et al. [29]; 6 ¼ Straub
et al. [187]; 6 ¼ Morissette et al. [58]; 7 ¼ Pekkarinen et al. [188]; 8 ¼ Egeland et al. [25]; 9 ¼ Freimer et al. [30]; 10 ¼ Stine et al. [31];
11 ¼ Dick et al. [189]; 12 ¼ Cichon et al. [190]; 13 ¼ McInnis et al. [191]; 14 ¼ Adams et al. [192]; 15 ¼ Coon et al. [193]; 16 ¼ Turecki et al. [73];
17 ¼ Seguardo et al. [33]; 18 ¼ Badenhop et al. [194].

a large pedigree from the Old Order Amish popula-
tion in southeastern Pennsylvania [25]. However,
subsequent studies of an expanded version of the
same family demonstrated much reduced evidence
for linkage [26]. This pattern of inconsistent replica-
tion has plagued linkage studies of bipolar disorder.
Figure 17.2 summarizes the major chromosomal
regions that have been reported to be linked to bipolar
disorder. There have been over 20 linkage genome
scans conducted in bipolar disorder. Considering all
of the regions putatively implicated by these scans, a
substantial portion of the genome shows some evi-
dence of linkage.
Several chromosomal regions have been the focus
of particular attention. The original Amish study was
of particular interest because the gene for tyrosine
hydroxylase (TH) and brain-derived neurotrophic
factor (BDNF) mapped in close proximity to the
linked markers. Both of these genes have also been
reported to be associated with bipolar disorder.
Though the expansion of the family reduced the evi-
dence for linkage (LOD score) from over 4.0 to 2.2,
there still remained interesting support for linkage.
This region, however, has not figured prominently in
subsequent linkage studies. A similar story occurred
in early reports of linkage to the X chromosome.

201
 Chapter 17: Genetics of bipolar disorder
X linkage was reported in a series of Israeli families,
that on re-examination could not be replicated [27,
28]. Chromosome 18 was first implicated by Berret-
tini et al. [29] who found evidence of linkage to a
pericentromeric region. Subsequently, three other
regions on chromosome 18 were reported to be
linked. Two of these came from studies of a Costa
Rican population isolate [30], while one demonstrated
a maternal pattern of transmission [31].
The inconsistency of results between studies has
made the interpretation of these data challenging.
Several attempts have been made to systematically
analyze these combined data. Badner et al. [32] con-
ducted a meta-analysis of 11 bipolar genome scans.
They employed the Multiple Scan Probability (MSP)
which was a modification of Fisher’s meta-analysis
using p-values. Significant evidence of linkage was
obtained for two regions: 13q and 22q. Segurado
et al. [33] subsequently examined 18 bipolar genome
scans using the Genome Scan Meta-Analysis (GSMA)
method. This procedure involved binning the genome
and ranking each study across all bins. Ranks were
then compared across regions and distributions
examined. No region achieved genome-wide signifi-
cance. However, several regions showed interesting
evidence of linkage including: 9p22, 10q11, and
14q24. It is interesting that two of these on chromo-
somes 9 and 14 were never a major linkage peak in
any single study. Rather small contributions from
many studies led to this result. McQueen et al. [34]
conducted an analysis of 11 genome-wide linkage
scans in which they combined the original genotype
data and found genome-wide significant evidence for
linkage on chromosomes 6q and 8q.
There are several possible interpretations of these
data and the inconsistency that characterizes them.
Heterogeneity is frequently invoked as a reason for
nonreplication. It is almost certain that multiple genes
may cause bipolar disorder, but the number of genes
remains unclear. Not only may different genes be
involved, but there may be numerous different muta-
tions in those genes. Linkage is specific to the gene
and linkage tests can detect a gene’s effect even if
different mutations in the gene are operating in dif-
ferent families. Association, however, is specific to the
allele, and if a large number of mutations occur in a
gene that may predispose to bipolar disorder, that
effect may be difficult to detect by association
methods. Early linkage studies focused on large fam-
ilies such that one extended family might have
202
enough statistical power to detect linkage. However,
in many cases, in part due to assortative mating, more
than one gene for bipolar disorder might be segregat-
ing with illness, thereby confounding this strategy.
This was likely the case for the large Amish family
described above. Later linkage studies focused on
affected sibling pairs and much smaller families.
Though this approach may be more robust to intra-
family heterogeneity, it may have less power to detect
rare genes. Therefore, in interpreting these results, it
is important to keep in mind that different sizes of
families represent different linkage strategies that are
powered to detect different kinds of gene effects.
Association results as described below are also best
powered to detect genetic effects in the context of
polygenic transmission, or the common variant
common disease model. Many studies today are based
on the assumption that genes for bipolar disorder are
affected by many different mutations, some of which
are rare and some common. Therefore, some incon-
sistency in findings may result from the fact that
different study designs are best powered to detect
different kinds of gene effects.
Another intriguing element is that of natural
selection. Association relies on the existence of
common functional variants. Though many mechan-
isms may exist to lead to the amplification of an allele
in a population, such as population “bottlenecks” or
random genetic drift, one mechanism may be selec-
tion. If a mutation associated with some aspect of
bipolar disorder is evolutionarily advantageous, then
it will be selected, thereby increasing the frequency of
that mutation in the population. Its ultimate fre-
quency may represent balanced selection, where the
advantages that increase the frequency are balanced
against its disadvantages so as to achieve a specific
allele frequency.
Candidate gene studies
In a candidate gene study, a specific gene is hypothe-
sized to be involved in bipolar disorder because of its
known function in brain physiology. Known markers
or functional variants can then be tested in a case-
control association study to determine if the gene is
involved in illness. The advantage of this approach is
its physiological relevance and limited number of
statistical tests. The disadvantages are its inability to
identify novel previously unsuspected genes and that
it does not comprehensively cover the genome.
 Chapter 17: Genetics of bipolar disorder

Table 17.3 Selected candidate genes reported to be associated to bipolar disorder.

Gene symbol Gene name and function References

SLC6A4 Serotonin transporter, reuptake of serotonin from the synapse [125–139]
BDNF Brain derived neurotrophic factor, neuronal growth and maintenance [42, 140]
COMT Catechol-O-methyltransferase, inactivation of catecholamines [45, 46, 141–143]
TH Tyrosine hydroxylase, rate limiting enzyme in catecholamine biosynthesis [144–150]
NTRK2 Trkb neurotrophin tyrosine kinase receptor for BDNF [43]
ADRBK2 (GRK3) G protein receptor kinase 3, desensitization of G protein coupled receptors [60, 62, 151]
P2RX7 ATP receptor [59, 152–155]
TPH2 Tryptophan hydroxylase 2, rate limiting enzyme in serotonin biosynthesis [156, 157]
GRIK4 Glutatmate receptor, kainate type [158]
DAOA D-amino acid oxidase activator, modulator of glutamate neurotransmission [48, 52]
first identified in schizophrenia
DISC1 Disrupted in schizophrenia, identified in a large Scottish family with a [159–162]
balanced translocation
Sp4 Transcription factor [54, 63, 163, 164]
NRG1 Neuregulin1, signaling molecule originally identified in schizophrenia [165]
CLOCK Clock gene [166, 167]
BMAL1 Clock gene [168]
GNAL G protein in the olfactory bulb [169]
GRIN2B NMDA glutamate receptor subunit [170–172]
SLC6A3 (DAT) Dopamine transporter, reuptake of dopamine [95, 173–177]
CACNG2 Stargazin, AMPA type glutamate receptor associated protein [178]

A large number of such candidate gene association
studies have been conducted. Some of the more
prominent genes and studies are listed in Table 17.3.
See this table for detailed citations.
The serotonin transporter has been one of the
most widely studied genes in psychiatry [35]. The
transporter plays a key role in serotonin neurotrans-
mission by mediating the active reuptake of serotonin
at the synapse. It is also the target of selective sero-
tonin reuptake inhibitor (SSRI) antidepressants that
block its action. Most attention on the transporter has
focused on a variable number of tandem repeats
(VNTR) polymorphism 5’ of transcription initiation.
This variant has two predominant forms, long or
short, based on the inclusion or exclusion of two
repeats. This polymorphism has been shown to influ-
ence transcription of the gene and was initially asso-
ciated with the personality trait of neuroticism [36].
A number of studies have been conducted examining
the role of this gene in bipolar disorder (Table 17.3),
though results have been mixed and inconsistent.
Together, these data do not present a strong case for
the gene’s role in vulnerability to bipolar disorder.
A limitation of these studies, as with several candidate
genes, is that the focus of investigation has almost
entirely been on the VNTR polymorphism while
other functional variants are known [37]. This VNTR
polymorphism, designated 5HTTLPR, has been asso-
ciated with risk for depression in response to stress
[38]. It has also been associated with response to SSRI
antidepressants [39], with poorer response associated
with the s or short allele which results in a lower
transcription of the gene.
BDNF has also been extensively studied. It plays
an important role in brain development and in par-
ticular in growth and development of catecholamine

203
 Chapter 17: Genetics of bipolar disorder
systems implicated in bipolar disorder. BDNF expres-
sion is increased by antidepressants and lithium, and
this increase in expression is necessary for antidepres-
sant-related behavioral responses in animal models
[40]. The gene contains a Val/Met substitution at
amino acid 66 that has been shown to affect transport
of the peptide within the cell and also its release.
Several studies have shown BDNF or its receptor
NTRK2 to be associated with bipolar disorder [41,
42]. Variation in BDNF and NTRK2 may also influ-
ence response to lithium [43].
Several genes related to catecholamine biosyn-
thesis have been examined. TH is the rate limiting
enzyme in catecholamine biosynthesis. This gene
contains a tetranucleotide repeat in intron 1 that
has been shown to regulate transcription and to be
associated with bipolar disorder [44]. Catehol-O-
methyltransferase (COMT) catabolizes norepineph-
rine and dopamine thereby leading to their elimin-
ation from the synapse. A Val/Met substitution in
COMT has been shown to functionally affect enzyme
function and has been variably associated with risk
for bipolar disorder [45, 46].
Some of the strongest evidence for association of
specific genes to bipolar disorder comes from studies
initially conducted in samples of patients with
schizophrenia. D-amino acid oxidase activator
(DAOA), neuregulin (NRG1), and disrupted in
schizophrenia (DISC1) are all genes first identified
by mapping studies in schizophrenia, then later
shown to be associated to bipolar disorder. Linkage
studies of families with schizophrenia first identified
a region on 13q [47]; subsequent fine mapping asso-
ciation studies in this region led to the identification
of the DAOA gene [48]. The same region on 13q has
also been shown to be linked to bipolar disorder in
several studies [49–51]. This led investigators to con-
duct association studies of the DAOA gene in bipolar
disorder and find that the same gene was likely
involved in both bipolar disorder and schizophrenia
[52]. The neuregulin gene (NRG1) was similarly
identified first by linkage to 8p in Icelandic families
with schizophrenia followed by fine mapping by
association [53]. Subsequent studies demonstrated
association of the gene to bipolar disorder as well
[54]. Neuregulin is a novel signaling molecule
that interacts with its receptor ErbB4 to modulate
NMDA receptor mediated glutamate signaling [55].
DISC1 was identified by mapping the breakpoint of a
1 : 11 balanced translocation segregating with both
204
bipolar disorder and schizophrenia in a large Scot-
tish family [56]. The breakpoint in DISC1 provided
an unequivocal causative mutation in this family.
This has emboldened biologists to conduct extensive
studies of DISC1 function. Such studies are very
labor intensive and the presence of a clear mutation
identifying the gene justified such an investment. As
a result, DISC1 is probably one of the biggest suc-
cesses of the whole genetic mapping strategy. In
addition to both disorders being present in the ori-
ginal family, some subsequent studies have also
shown evidence of association to bipolar disorder
providing further support that this gene plays a role
in both disorders [57].
Several genes have been identified by linkage in
bipolar families followed by fine mapping association
studies. Linkage to 12q has been reported in an
isolated population in Quebec [58]. Fine mapping
studies identified nonsynonymous SNPs in the gene
P2RX7 associated with bipolar disorder [59]. P2RX7
is a purine receptor that is primarily expressed in
immune cells. This has suggested a role for inflam-
mation in depression. G protein receptor kinase 3
(GRK3 or ADRBK2) is a member of a family of
kinases that are involved in the homologous desen-
sitization of G protein coupled receptors. This gene
was identified first by linkage to 22q in a set of
extended bipolar families [50]. There appeared to
be two distinct linkage peaks on 22q, one peak
mapped to within 20 kb of the GRK3 gene. Sequen-
cing studies of the coding and promoter regions
revealed a series of SNPs in the proximal promoter.
One of these, showed significant association to bipo-
lar disorder in two independent populations [60].
Functional studies showed that the promoter variant
altered binding of Sp1 family transcription factors
and increased expression of the gene in cultured
mouse cortical neurons [61]. GRK3 may interact
with other possible genes for bipolar disorder. One
such candidate is the gene coding for the transcrip-
tion factor Sp4; involvement of this gene was first
suggested by anatomical and behavioral changes in a
knockdown mouse [62] that were suggestive of
psychotic illness. A SNP in Sp4 was later shown to
be associated with bipolar disorder and schizophre-
nia [63]. The promoter variant in the GRK3 gene
impacts a transcription factor binding site that can
be occupied by Sp1 or Sp4, suggesting a possible
interaction. Furthermore, the P2RX7 gene is phos-
phorylated and desensitized by the GRK3 gene [64].
 Chapter 17: Genetics of bipolar disorder

Table 17.4 Genome-wide association studies of bipolar disorder.

Study Type Sample Ethnicity Sample size Platform Results

(Case/
control)
WTCCC [65] Case WTCCC Caucasian 2000/3000 Affymetrix PALB2, p ¼ 108
control 500 K
Sklar et al. [66] Case STEP-BD Caucasian 1461/2008 Affymetrix MYO5B, p ¼ 107;
control 500 K TSPAN8, p ¼ 107;
CACNA1C, p ¼ 104
Baum et al. Pooling NIMH W1–4 þ Caucasian 461/563 þ Illumina DGKH; p ¼ 108
[69] Bonn/ 772/876 550 K
Mannheim 800/800
Scott et al. Pooling þ NIMH/Pritzker Caucasian 3683/14 507 Illumina 1p31.1, p ¼ 107;
[68] Ind GSK WTCCC 550 K 3p21, p ¼ 107;
genotyping MCTP1, p ¼ 107
Ferreira et al. Meta- WTCCC þ Caucasian 4387/6209 Affymetrix ANK3, p ¼ 109;
[179] analysis STEP-BD 500 K CACNA1AC, p ¼ 108
Smith et al. Case BiGS (NIMH Caucasian 1001/1033 Affymetrix Xq27.1, p ¼ 106;
[67] control W1–5) 6.0 (1 M NAP5, p ¼ 106
SNPs)
African- 345/670 DPY19L3, p ¼ 106;
American NTRK2, p ¼ 106
Hattori et al. Case Hondo area Japanese 107/107 Affymetrix KCNMB2, p ¼ 104;
[70] control 100 K AUTS2, p ¼ 104

Genome-wide association studies architecture and the role of common variants in

GWAS have only recently started in bipolar disorder
and other psychiatric disorders. Published studies to
date are summarized in Table 17.4. In 2007, the
Wellcome Trust Case Control Consortium
(WTCCC) published the results of a large GWAS of
seven disorders involving a total of 14 000 cases and
3000 controls [65]. Analysis of 500 000 SNPs identi-
fied a region on 16p12 showing the strongest evi-
dence for association (p ¼ 108). Several genes are
contained in this region including PALB2, which is a
binding partner for BRCA1, and dynactin 5 (DCTN5)
that is known to interact with DISC1. One of the most
striking observations of this study was the difference
in results for the seven disorders. Some disorders
displayed a few clear strong peaks such as coronary
artery disease (CAD) or Crohn’s disease, indicating a
few common variants that made strong contribu-
tions, while others such as bipolar disorder and
hypertension lacked any such strong loci. This sug-
gests differences in the underlying genetic
these disorders.
Sklar et al. conducted a GWAS using the STEP-BD
sample and a sample from University College, London
[66]. STEP-BD was a large naturalistic treatment study
in bipolar disorder. They reported evidence for myosin
5B (MYO5B) and tetraspanin-8 (TSPAN8); however,
this was not replicated in an independent sample. They
also identified the gene CACNA1C, a member of a
family of L-type calcium channels, as showing evi-
dence for association in both their sample and in the
aforementioned Wellcome Trust study. Such replica-
tion made this gene a good candidate even if it did not
have the strongest signal in either study alone. Ferreira
et al. reported a combined analysis of these 2 datasets
plus additional samples for a total dataset of 4387 cases
and 6209 controls. This analysis replicated the earlier
association to CACNA1C, however, the strongest
result was at the gene ankyrin 3 (ANK3). ANK3 codes
for ankyrin G, which is a brain expressed cytoskeletal
protein involved in attaching sodium channels to the

205
 Chapter 17: Genetics of bipolar disorder
cytoskeleton. ANK3 is expressed primarily at the axo-
nal initial segment where action potentials originate.
As some mood stabilizer medications, lamotrigine in
particular, block sodium channels as their likely mech-
anism of action, this gene is of particular interest.
Smith et al. [67] have reported on a GWAS of the
NIMH Genetics Initiative for Bipolar Disorder
sample as part of the Bipolar Genome Study (BiGS).
In total, 1001 European-American (EA) cases and
1033 EA controls, and 345 African-American (AA)
cases and 670 AA controls were genotyped at 1-M
SNPs. Though no single polymorphism showed
genome-wide significance, suggestive evidence was
obtained for several genes including: NAP5,
DPY19L3, and NTRK2. NTRK2 is an especially good
candidate as it codes for Trkb, which is a tyrosine
kinase receptor for BDNF (see above) that has been
extensively studied in psychiatry. This study is one of
the few to date to examine a non-Caucasian popula-
tion. This enabled a combined analysis and allowed
identification of several genes (ROR1, RGS5, and
BTBD16) that showed evidence of association in both
populations, but to different alleles. This suggests that
different mutations occurred in these genes in the
histories of these two populations, and both predis-
pose to bipolar disorder.
Scott et al. [68] examined three populations for a
combined sample of 3683 Caucasian cases and 14 507
controls. A substantial portion of this sample over-
lapped with the Smith et al. study described above.
Though no single SNP achieved genome-wide signifi-
cance suggestive evidence was obtained for nonsy-
nonymous SNPs in ITIH1, GNL3, NEK4, and ITIH3.
This study also provided support for the ANK3 gene
described above.
Two other studies deserve mention. Baum et al.
conducted a pooled GWAS of the NIMH wave 1–4
sample with replication in a German Bonn/Mann-
heim sample [69]. In a pooling strategy, samples are
pooled across many subjects and genotyped
together. Signal strength for each of the two alleles
is used to estimate control and case allele frequen-
cies. This is economical and faster, but has less
ability to distinguish smaller differences in allele
frequencies. This study identified DGKH, a gene
involved in lithium’s mechanism of action as having
suggestive evidence of association. The only
reported study in an Asian population to date is
by Hattori et al. [70] in a Japanese population. In
total, 107 Japanese cases and 107 controls were
206
genotyped and analyzed. No genome-wide signifi-
cant findings were reported.
Mapping genes for subphenotypes
The family data described above are consistent with
different genes predisposing to different forms of
illness. A better understanding of the mapping from
subphenotype to gene will be invaluable in under-
standing how different genes result in different forms
or courses of illness. Early efforts to do this include a
study by Potash et al. [71] who conducted a linkage
study of psychotic bipolar disorder. These investiga-
tors identified two regions, 22q12 and 13q31, with
suggestive evidence for linkage [72]. Both of these
regions have figured prominently in studies of schizo-
phrenia. Turecki et al. [73] attempted to map genes
for lithium responsive bipolar disorder. They identi-
fied a region on 15q that segregated with lithium
responsive bipolar disorder.
Similar studies have been undertaken using
GWAS data. Greenwood et al. [74] examined asso-
ciations to the phenotypes of irritable and euphoric
mania in the BiGS dataset. Initially a case-case
analysis was conducted comparing subjects with
irritable to elated mania. A large region under a
broad peak on 13q was identified as showing asso-
ciation in this case-case analysis. Subsequent ana-
lyses showed strong association between irritable
mania and controls, but no association between
euphoric patients and controls. This suggests that
whatever gene or regulatory element resides in this
region predisposes specifically to irritable mania.
Similar studies of suicide attempts by Willour
et al. [75] in the BiGS and German datasets showed
evidence of association to a region on 2p25 that
contains the gene ACP1. Age of onset, as a quanti-
tative trait, was also examined in a combined BiGS/
German dataset, and evidence of association
obtained to the oxysterol binding protein-like 1A
(OSBPL1A) and neural cell adhesion molecule 2
(NCAM2) [76]. Alcohol dependence has long been
known to have a four-fold greater prevalence in
bipolar disorder as compared to subjects without a
mood disorder. Nwulia et al. [77] examined the
BiGS GWAS dataset to identify genes associated
with the combined phenotype of bipolar disorder
and alcohol dependence. They found evidence
implicating the genes WDR59 and alpha N-catenin
(CTNNA2). Sleeplessness at initial mania has been

16
14
12
Morbid risk (%)
10
8 Bipolar
6 Schizophrenia
4 elevated in the families of those with
2 overlap of causative genes. (Adapted
0
SZ Mania Depression Control
Proband diagnosis
hypothesized to distinguish different forms of ill-
ness. Analysis of this phenotype in the BiGS dataset
yielded three regions of interest: KIAA0143, a
membrane bound protein; NAP5 which was also
detected in the overall bipolar analysis; and MEP1B
a zinc metalloprotease [78].
Affective temperaments have also been examined
as subphenotypes that may index different forms of
illness. Evans et al. [79] conducted a linkage study in
bipolar families using cyclothymia as a quantitative
trait and found evidence of linkage on 18p11. Cor-
responding analysis of the BiGS GWAS dataset
found associations for hyperthymia on chromosome
12 and dysthymia on chromosome 4 [80]. Harm
avoidance as a dimension of temperament has been
associated with the D3 dopamine receptor in
Han Chinese [81]. Another study of a South
African population of bipolar individuals found the
dopamine transporter to be associated with self-
directedness and the serotonin transporter with
harm avoidance [82].
Overlap with other disorders
Since Emil Kraepelin distinguished manic depressive
illness from dementia praecox [83], the relationship
between bipolar disorder and schizophrenia has been
debated. Though this question has many aspects,
genetics may serve as a proxy for differences in
biology and etiology. As the family studies described
above began to elucidate the familiality of each dis-
order, investigators also questioned whether each of
these “breed true”. That is are different genes and
biological abnormalities operating in these two
Chapter 17: Genetics of bipolar disorder
Figure 17.3 Partial overlap in familial
transmission of bipolar disorder and
schizophrenia. This figure illustrates the
familial risk for bipolar disorder,
schizophrenia, and depression in families
with probands of different diagnoses and
controls. Rates significantly different from
controls are indicated with an *. Rates
of bipolar disorder are elevated above
control rates in families of probands with
schizophrenia, and schizophrenia is
Unipolar bipolar disorder. This suggests a partial
from [84]).
disorders. A number of family studies have
addressed this question. Figure 17.3 illustrates the
results of a study by Tsuang et al. of morbid risk
for bipolar disorder, schizophrenia and depression in
the families of probands with each of these three
disorders and controls [84]. In families of bipolar
probands, bipolar disorder displays the largest sig-
nificant increase in risk. However, the rate of schizo-
phrenia is also elevated above the rate in control
probands. The mirror image of this is seen in fam-
ilies of probands with schizophrenia. Schizophrenia
shows the largest significant increase in risk, but
bipolar disorder also occurs at a rate greater than
in that of control probands, but less than that of
bipolar probands. Together these data suggest that
a portion of the genetic vulnerability for these two
disorders is shared.
Twin data also support this idea of partial overlap
with schizophrenia. Cardno et al. [85] examined the
medical records of the Maudsley Twin Psychosis
Registry and found that over time patients presented
differently and there was not strong longitudinal reli-
ability of the two diagnoses. They further found evi-
dence for a significant portion of shared genetic
factors for these two disorders. Molecular mapping
studies have provided further support. Early linkage
studies detected similar regions such as 13q and 22q
in studies of both schizophrenia and bipolar families
leading to the suggestion of shared genetic factors
[86]. As mentioned above, a large family with a trans-
location breakpoint in the DISC1 was identified in a
family segregating for both bipolar disorder and
schizophrenia [87]. Linkage was also reported in an
Italian population isolate between a region on 15q
207
 Chapter 17: Genetics of bipolar disorder
and both bipolar disorder and schizophrenia [88].
Several specific genes have been reported to be asso-
ciated to both disorders such as neuregulin (NRG1)
and D-amino acid oxidase activator (DAOA) [52, 54].
GWAS of both disorders may help illuminate the
common and distinct genes, but such analyses are just
beginning.
A similar overlap may exist with attention-deficit
hyperactivity disorder (ADHD). Some studies suggest
that probands with ADHD are more likely to have
parents with bipolar disorder and parents with bipo-
lar disorder are more likely to have children with
ADHD [89–91]. Family studies further suggest that
bipolar disorder with ADHD may be a distinct form
of illness [92]. The dopamine transporter which is the
target for stimulants has been implicated in both
disorders [93–95]. A linkage study of ADHD traits
in bipolar families identified a region on 10p14 sig-
nificantly linked [96]. This might reflect a gene spe-
cific for a form of illness that shares bipolar and
ADHD traits.
In addition to other psychiatric disorders, there
is much literature to support genetic overlap
between bipolar disorder and other medical condi-
tions. Such evidence for pleiotropy should not be
surprising and may in fact offer clues as to the
disease mechanism involving such genes. Odergaard
et al. have reported an increase in the rate of
migraine headache in patients with bipolar II dis-
order as compared with the general population [97].
They then went on to attempt to map genes respon-
sible for the combined phenotype of bipolar and
migraine. Linkage studies of bipolar studies using
bipolar and migraine as the phenotype identified a
region on 20p11 [98]. More recently, a similar
approach was applied to GWAS data and evidence
of association to a combined phenotype was
obtained for the gene KIAA0564 [99].
Genetic counseling
Families frequently seek advice regarding genetic risk
for bipolar disorder. It is highly heritable and most
affected individuals have affected family members.
Couples from families that have been severely
affected may have reservations about having children
and require counseling regarding the risks. The
family studies cited above provide a guide for esti-
mates of family risk. It is important to take a careful
family history as the risks from studies must be
208
interpreted in the light of the specific family being
counseled. A high familial load suggests a greater
risk as does bilineal transmission. A variety of
aspects of illness may be heritable as described
above. Such factors may be helpful in counseling
families regarding possible course or severity. It is
important to emphasize the treatable nature of the
disorder, and be sensitive to the way in which each
family will hear the data in the light of their own
personal experience.
Though a number of gene findings have been
replicated and likely reflect real effects, together these
account for only a minority of the genetic variance.
For this reason, DNA testing currently has limited
predictive and clinical value. However, it is likely in
the near future that numerous new mutations will
emerge from sequencing studies and these will form
databases against which an individual’s sequence can
be compared. Analyses of such data may provide a
useful aid in diagnosis, and in selection of optimal
medication. As the heritability of the illness is less
than 100% such tests will always be probabilistic in
nature. Yet in addition to aiding the ill patient, such
tests may indicate risks in asymptomatic individuals.
Such information is at risk of being misused in a
discriminatory fashion and requires careful ethical
consideration. However, it might also afford an
opportunity to prevent or ameliorate the course of
illness for some at risk individuals. Little data pres-
ently exists to guide such presymptomatic treatment,
but it is clearly a needed area of research.
Conclusions
Bipolar disorder is a highly heritable psychiatric
disorder, and the strong role of genes is supported
by family and twin studies. Mapping studies using
linkage and association methods have had modest
success to date despite difficulties in replication
between studies. Linkage studies have shown the
best support for chromosomal regions: 6q, 8q, 9p,
13q, 14q, and 22q. Several candidate genes first
identified in studies of schizophrenia have shown
reproducible association in bipolar disorder (such
as NRG1, DAOA, and DISC1). GWAS have been
successful in identifying a few genes with small
effects on risk. ANK3 and CACNA1C, in particular,
have shown reproducible evidence for association
and may represent the genes with the strongest
evidence to date for involvement in bipolar

disorder. The data overall suggest a high level of
both genic and allelic heterogeneity, as well as, a
complex mode of inheritance. The coming availability
of economical whole genome sequencing promises
References
1. Wozniak J, et al. Psychol Med
2009:1–10 [epub ahead of print].
2. Spence MA, et al. Psychiatr Genet
1993;3:143.
3. Cox N, et al. J Psychiatr Res
1989;23:109–123.
4. Rice J, et al. Arch Gen Psychiatry
1987;44:441–447.
5. Gershon ES, et al. Arch Gen
Psychiatry 1982;39:1157–1167.
6. McInnis MG, et al. Am J Hum
Genet 1993;53:385–390.
7. O’Donovan M, et al. Am J Med
Genet C Semin Med Genet
2003;123C(1):10–17.
8. Gershon ES, et al. Biol Psychiatry
1973;7(1):63–74.
9. Negri F, et al. J Affect Disord
1979;1(4):247–253.
10. Merikangas KR, et al. Psychol Med
1982;12(4):753–764.
11. McMahon FJ, et al. Am J Hum
Genet 1995;56:1277–1286.
12. Kornberg JR, et al. J Affect Disord
2000;59(3):183–192.
13. Coryell W, et al. Br J Psychiatry
1984;145:49–54.
14. Potash JB, et al. Am J Psychiatry
2001;158(8):1258–1264.
15. Potash JB, et al. Am J Med Genet
2003;160:680–686.
16. MacKinnon DF, et al. Am J
Psychiatry 2002;159(1):30–35.
17. MacKinnon DF, et al. Bipolar
Disord 2005;7(5):441–448.
18. Schulze TG, et al. Arch Gen
Psychiatry 2006;
63(12):1368–1376.
19. Saunders EH, et al. Am J Med
Genet B Neuropsychiatr Genet
2008;147B(1):18–26.
20. Potash JB, et al. Am J Psychiatry
2007;164(8):1229–1237.
Chapter 17: Genetics of bipolar disorder
availability of complete genomic information. This,
and large samples now being collected, may provide
the datasets necessary to unravel the genetic complex-
ities of this illness.
21. Kelsoe JR. J Affect Disord 2003;
73(1–2):183–197.
22. Perugi G, et al. J Affect Disord
2012;136(1–2):e41–49.
23. Evans L, et al. J Affect Disord
2005;85(1–2):153–168.
24. Risch N, et al. Science 1996;
273:1516–1517.
25. Egeland JA, et al. Nature 1987;
325:783–787.
26. Kelsoe JR, et al. Nature 1989;
342:238–243.
27. Baron M, et al. Nature 1987;
326:289–292.
28. Baron M, et al. Nat Genet 1993;
3:49–55.
29. Berrettini WH, et al. Proc Natl
Acad Sci U S A 1994;91:
5918–5921.
30. Freimer NB, et al. Nat Genet
1996;12:436–441.
31. Stine OC, et al. Am J Hum Genet
1995;57:1384–1394.
32. Badner JA, et al. Mol Psychiatry
2002;7(4):405–411.
33. Segurado R, et al. Am J Hum
Genet 2003;73(1):49–62.
34. McQueen MB, et al. Am J Hum
Genet 2005;77(4):582–595.
35. Caspi A, et al. Am J Psychiatry
2010;167(5):509–527.
36. Lesch KP, et al. Science
1996;274:1527–1531.
37. Kunugi H, et al. Mol Psychiatry
1997;2:457–462.
38. Caspi A, et al. Science 2003;
301(5631):386–389.
39. Smeraldi E, et al. Mol Psychiatry
1998;3(6):508–511.
40. Nibuya M, et al. J Neurosci
1995;15(11):7539–7547.
41. Fan J, et al. Novartis Found Symp
2008;289:60–72; discussion 72–73,
87–93.
42. Sklar P, et al. Mol Psychiatry
2002;7(6):579–593.
43. Bremer T, et al. Mol Diagn Ther
2007;11(3):161–170.
44. Meloni R, et al. Lancet 1995;
345(8954):932.
45. Lachman HM, et al. Psychiatr
Genet 1997;7:13–17.
46. Biomed European Bipolar
Collaborative Group. Br J
Psychiatry 1997;170:526–528.
47. Blouin JL, et al. Nat Genet
1998;20:70–73.
48. Chumakov I, et al. Proc Natl
Acad Sci U S A 2002;99(21):
13675–13680.
49. Detera-Wadleigh SD, et al. Proc
Natl Acad Sci U S A 1999;96(10):
5604–5609.
50. Kelsoe JR, et al. Proc Natl Acad Sci
U S A 2001;98(2):585–590.
51. Shaw SH, et al. Mol Psychiatry
2003;8(5):558–564.
52. Hattori E, et al. Am J Hum Genet
2003;72(5):1131–1140.
53. Stefansson H, et al. Am J Hum
Genet 2002;71(4):877–892.
54. Green EK, et al. Arch Gen
Psychiatry 2005;62(6):642–648.
55. Buonanno A. Brain Res Bull
2010;83(3–4):122–131.
56. Blackwood DH, et al. Nat Genet
1996;12:427–430.
57. Hodgkinson CA, et al. Am J Hum
Genet 2004;75(5):862–872.
58. Morissette J, et al. Am J Med Genet
1999;88:567–587.
59. Barden N, et al. Am J Med Genet
B Neuropsychiatr Genet 2006;
141(4):374–382.
60. Barrett TB, et al. Mol Psychiatry
2003;8:546–557.
61. Zhou X, et al. Biol Psychiatry
2008;64(2):104–110.
209
 Chapter 17: Genetics of bipolar disorder
62. Zhou X, et al. Mol Psychiatry
2005;10(4):393–406.
63. Zhou X, et al. PLoS One 2009;4(4):
e5196.
64. Feng YH, et al. Am J Physiol Cell
Physiol 2005;288(6):C1342–C1356.
65. Wellcome Trust Case Control.
Nature 2007;447(7145):661–678.
66. Sklar P, et al. Mol Psychiatry
2008;13(6):558–569.
67. Smith EN, et al. Mol Psychiatry
2009;14(8):755–763.
68. Scott LJ, et al. Proc Natl Acad Sci
U S A 2009;106(18):7501–6.
69. Baum AE, et al. Mol Psychiatry
2008;13(2):197–207.
70. Hattori E, et al. Am J Med Genet
B Neuropsychiatr Genet
2009;150B(8):1110–1117.
71. Potash JB, et al. Am J Psychiatry
2003;60(4):680–686.
72. Lander E, et al. Nat Genet
1995;11:241–247.
73. Turecki G, et al. Mol Psychiatry
2001;6(5):570–578.
74. Greenwood T, Jr. In World
Congress on Psychiatric Genetics.
San Diego: WCPG; 2009.
75. Willour VlS, et al. Mol Psychiatry
2012;17(4):433–444.
76. Belmonte PJ, et al. In World
Congress on Psychiatric Genetics.
San Diego: WCPG; 2009.
77. Nwulia EH, et al. In World
Congress on Psychiatric Genetics.
San Diego: WCPG; 2009.
78. Coleman VK, et al. In World
Congress on Psychiatric Genetics.
San Diego: WCPG; 2009.
79. Evans LM, et al. Am J Med Genet
B Neuropsychiatr Genet 2008;
147(3):326–332.
80. Greenwood T, et al. In World
Congress on Psychiatric Genetics.
San Diego: WCPG, 2009.
81. Lin CI, et al. Psychiatry Res
2010;177(3):364–366.
82. Savitz J, et al. Genes Brain Behav
2008;7(8):869–876.
83. Krapelin E. Manic-Depressive
Insanity and Paranoia.
210
Manchester, NH: Ayer and
Co.; 1921.
84. Tsuang MT, et al. Br J Psychiatry
1980;137:497–504.
85. Cardno AG, et al. Arch Gen
Psychiatry 1999;56(2):162–168.
86. Berrettini WH. Biol Psychiatry
2000;48(6):531–538.
87. Blackwood DH, et al. Am J Hum
Genet 2001;69(2):428–433.
88. Vazza G, et al. Mol Psychiatry
2007;12(1):87–93.
89. Kent L, et al. J Affect Disord
2003;73(3):211–221.
90. Sachs GS, et al. Am J Psychiatry
2000;157:466–468.
91. Faraone SV, et al. J Am Acad Child
Adolesc Psychiatry 1997;
36:1378–1387.
92. Faraone SV, et al. J Am Acad Child
Adolesc Psychiatry 1997;36(10):
1378–1387; discussion
1387–1390.
93. Waldman ID, et al. Am J Hum
Genet 1998;63:1767–1776.
94. Cook EH, Jr, et al. Am J Hum
Genet 1995;56:993–998.
95. Greenwood TA, et al. Mol
Psychiatry 2006;11(2):125–133.
96. Joo EJ, et al. Am J Med Genet
B Neuropsychiatr Genet 2010;
153B(1):260–268.
97. Oedegaard KJ, et al. J Affect Disord
2005;84(2–3):233–242.
98. Oedegaard KJ, et al. J Affect Disord
2010;122(1–2):14–26.
99. Oedegaard KG, et al. In World
Congress on Psychiatric Genetics.
San Diego: WCPG; 2009.
100. Perris C. Acta Psychiatr Scand
1966;194:102–117.
101. Winokur G, et al. In Worris J
(ed.). Recent Advances in
Biological Psychiatry. New York:
Plenum; 1967.
102. Mendlewicz J, et al. Am J Hum
Genet 1974;26:692–701.
103. Goetzl U, et al. Arch Gen
Psychiatry 1974;31:665–672.
104. Helzer JE, et al. Arch Gen
Psychiatry 1974;31:73–77.
105. Gershon ES, et al. J Psychiatr Res
1975;12:283–299.
106. James NM, et al. Br J Psychiatry
1975;126:449–456.
107. Johnson GFS, et al. Arch Gen
Psychiatry 1977;34:1074–1083.
108. Petterson U. Acta Psychiatr Scand
1977;269:1–93.
109. Smeraldi E, et al. Acta Psychiatr
Scand 1977;56:382–398.
110. Angst J, et al. Hum Genet 1980;
55:237–254.
111. Dunner DL, et al. Am Coll
Neuropsychopharm 1980; Abs.
112. Taylor MA, et al. J Affect Disord
1980;2:95–109.
113. Fieve RR, et al. J Psychiatr Res
1984;18:425–445.
114. Tsuang MT, et al. J Psychiatr Res
1985;19:23–29.
115. Sadovnick AD, et al. Am J Med
Genet 1994;54(2):132–140.
116. Birmaher B, et al. Arch
Gen Psychiatry 2009;
66(3):287–296.
117. Luxenberger H. Zentralblatt fur
diagensamte Neurologie und
Psychiatrie 1930;14:56–57.
118. Rosanoff AJ, et al. Am J Psychiatry
1935;91:725–762.
119. Slater E. Psychotic and Neurotic
Illness in Twins. London: Medical
Research Council; 1953.
120. Kallman F, et al. In Zubin J et al.
(eds.). Depression. New York:
Grune and Stratton; 1954.
121. Harvald B, et al. Genetics and the
Epidemiology of Chronic Diseases.
Washington, DC: US Department
of Health; 1965.
122. Allen MG, et al. Am J Psychiatry
1974;131:1234–1239.
123. Bertelsen A. Origin, Prevention
and Treatment of Affective
Disorder. London: Academic
Press; 1979.
124. Kendler KS, et al.
Arch Gen Psychiatry
1993;50:699–700.
125. Collier DA, et al. Neuroreport
1996;7:1675–1679.

126. Collier DA, et al. Mol Psychiatry
1996;1:453–460.
127. Kelsoe JR, et al. Am J Med Genet
1996;67(2):215–217.
128. Kunugi H, et al. Lancet 1996;347
(9011):1340.
129. Ogilvie AD, et al. Lancet 1996;347
(9003):731–733.
130. Stober G, et al. Lancet 1996;347
(9011):1340–1341.
131. Bellivier F, et al. Biol Psychiatry
1997;41:750–752.
132. Furlong RA, et al. Am J Med Genet
1998;81:58–63.
133. Gutierrez B, et al. Biol Psychiatry
1998;43:843–847.
134. Liu W, et al. Chung Hua I. Hsueh
I. Chuan Hsueh Tsa Chih
1998;15:345–348.
135. Seretti A, et al. Mol Psychiatry
1999;4:280–283.
136. Oliveira JR, et al. Mol Psychiatry
2000;5(4):348–349.
137. Saleem Q, et al. Am J Med Genet
2000;96(2):170–172.
138. Mundo E, et al. Arch Gen
Psychiatry 2001;58(6):539–544.
139. Lotrich FE, et al. Psychiatr Genet
2004;14(3):121–129.
140. Schumacher J, et al. Biol
Psychiatry 2005;58(4):307–314.
141. Kirov G, et al. Mol Psychiatry
1999;4(6):558–565.
142. Shifman S, et al. Am J Med Genet
B Neuropsychiatr Genet 2004;128
(1):61–64.
143. Prata DP, et al. Psychiatr Genet
2006;16(6):229–230.
144. Leboyer M, et al. Lancet
1990;335:1219–1219.
145. Gill M, et al. J Psychiatr Res
1991;25:179–184.
146. Byerley W, et al. Hum Hered
1992;42:259–263.
147. Inayama Y, et al. Am J Med Genet
1993;48:87–89.
148. Korner J, et al. Psychiatr Genet
1994;4:167–175.
149. Smyth C, et al. Am J Psychiatry
1996;153:271–274.
Chapter 17: Genetics of bipolar disorder
150. Smyth C, et al. Genomics
1997;39:271–278.
151. Barrett TB, et al. Psychiatr Genet
2007;17(6):315–322.
152. Shink E, et al. Mol Psychiatry
2005;10(6):545–552.
153. Lucae S, et al. Hum Mol Genet
2006;15(16):2438–2445.
154. Grigoroiu-Serbanescu M, et al.
Am J Med Genet B Neuropsychiatr
Genet 2009;150B(7):1017–1021.
155. McQuillin A, et al. Mol Psychiatry
2009;14(6):614–620.
156. Zhang X, et al. Neuron 2005;45
(1):11–16.
157. Chen C, et al. Bipolar Disord
2008;10(7):816–821.
158. Pickard BS, et al. Mol Psychiatry
2006;11(9):847–857.
159. Maeda K, et al. Biol Psychiatry
2006;60(9):929–935.
160. Hodgkinson CA, et al. Am J Hum
Genet 2004;75(5):862–872.
161. Macgregor S, et al. Mol Psychiatry
2004;9(12):1083–1090.
162. Devon RS, et al. Psychiatr Genet
2001;11(2):71–78.
163. Maier W. Eur Arch Psychiatry Clin
Neurosci 2008;258(Suppl 2):
37–40.
164. Georgieva L, et al. Biol Psychiatry
2008;64(5):419–427.
165. Thomson PA, et al. Mol Psychiatry
2007;12(1):94–104.
166. Benedetti F, et al. Am J Med Genet
B Neuropsychiatr Genet 2003;
123B(1):23–26.
167. Serretti A, et al. Am J Med Genet
B Neuropsychiatr Genet 2003;
121B(1):35–38.
168. Nievergelt CM, et al. Am J Med
Genet B Neuropsychiatr Genet
2006;141B(3):234–241.
169. Vuoristo JT, et al. Mol Psychiatry
2000;5(5):495–501.
170. Avramopoulos D, et al. Genet Med
2007;9(11):745–751.
171. Szczepankiewicz A, et al. World
J Biol Psychiatry 2009;10(4 Pt 2):
469–473.
172. Martucci L, et al. Schizophr Res
2006;84(2–3):214–221.
173. Mick E, et al. Am J Med Genet
B Neuropsychiatr Genet 2008;
147B(7):1182–1185.
174. Horschitz S, et al. Mol Psychiatry
2005;10(12):1104–1109.
175. Keikhaee MR, et al. Am J Med
Genet B Neuropsychiatr Genet
2005;135B(1):47–49.
176. Kelsoe JR, et al. Am J Med Genet
1996;67(6):533–540.
177. Greenwood TA, et al. Am J Med
Genet 2001;105(2):145–151.
178. Silberberg G, et al. Pharmacogenet
Genomics 2008;18(5):403–412.
179. Ferreira MA, et al. Nat Genet
2008;40(9):1056–1058.
180. Nurnberger JI, Jr, et al. Am J Med
Genet 1997;74:227–237.
181. Detera-Wadleigh SD, et al. Am
J Med Genet 1997;74:254–262.
182. Rice JP, et al. Am J Med Genet
1997;74:247–253.
183. Edenberg HJ, et al. Am J Med
Genet 1997;74:238–246.
184. Kelsoe JR, et al. Am J Med Genet
1998;81:461–462.
185. Detera-Wadleigh SD, et al. Proc
Natl Acad Sci U S A 1999;96
(10):5604–5609.
186. Blackwood DH, et al. Nat Genet
1996;12(4):427–430.
187. Straub RE, et al. Nat Genet 1994;8
(3):291–296.
188. Pekkarinen P, et al. Genome Res
1995;5(2):105–115.
189. Dick DM, et al. Am J Hum Genet
2003;73(1):107–114.
190. Cichon S, et al. Mol Psychiatry
2001;6(3):342–349.
191. McInnis MG, et al. Mol Psychiatry
2003;8(3):288–298.
192. Adams LJ, et al. Am J Hum Genet
1998;62:1084–1091.
193. Coon H, et al. Am J Hum Genet
1993;52:1234–1249.
194. Badenhop RF, et al. Mol
Psychiatry 2002;7(8):851–859.
211
 The genetics of major depression
Chapter
18 James B. Potash
Phenotype definition
The word “depressed” refers to a symptom, not a
diagnosis. A depressed mood generally means a low
mood, and that can occur in the setting of many
differing diagnoses. The DSM-IV major depressive
episode diagnosis requires a total of five symptoms,
one of which must be depressed mood or anhedonia
(loss of interest or pleasure), occurring together over
at least two weeks, to meet criteria for a major depres-
sive episode. Further, there must be significant dis-
tress or impairment associated with these symptoms.
It is not uncommon for patients with depressive
episodes to have occasional manic, hypomanic, or
mixed symptoms. If these symptoms are few or
fleeting then the patient would still meet criteria for
major depressive disorder (MDD). However, if they
reach the threshold for a hypomanic, manic, or mixed
episode, then the patient’s diagnosis becomes bipolar
disorder. The Kraepelinian conceptualization of
“manic-depressive illness” included all recurrent
severe disorders of mood: those characterized by
depression and mania, but also those characterized
by depression alone [1]. In the 1950s, Karl Kleist
proposed a nosological division of mood disorders
into unipolar and bipolar types, and Karl Leonhard’s
1959 classification of psychoses made a clear distinc-
tion between bipolar illness and recurrent unipolar
depression. Carlo Perris reported family findings in a
1966 monograph suggesting that unipolar and bipolar
illness were genetically separate [2]. However, not
long after, Reich et al. [3] reported from their study
of manic probands that there is an excess of unipolar
depressive illness in the families of individuals with
bipolar illness. This has been confirmed in several
studies including a family study of over 500 unipolar
and bipolar probands that was completed a couple of
Principles of Psychiatric Genetics, eds John I. Nurnberger, Jr. and Wade H. Berrettini. Published by Cambridge University
Press. © Cambridge University Press 2012.
212
decades later. This study found a higher incidence of
unipolar depressive illness in the families of bipolar
probands, but no excess of bipolar disorder in the
families of unipolar probands (though subsequent
studies have found a slight excess) [4, 5]. This suggests
both overlapping bipolar/unipolar genes and other
genes that uniquely confer risk for unipolar depres-
sive illness.
A key element of the debate over depression clas-
sification has focused on the issue of etiology. What is
the cause of depression? One of the schemes proposed
divided depression into endogenous and exogenous,
meaning that some depressions were caused by
factors inside the brain of the ill individual, deriving
from genetic predisposition, and some were caused by
factors in the person’s environment. A portion of the
etiology of MDD is clearly genetic, and that forms the
basis for the bulk of the discussion in this chapter.
However, given that MDD is only partly inherited,
variations in the makeup of people’s genes cannot
fully explain the variability in susceptibility to this
illness. There are at least two other major kinds of
explanations for this variation. One of these is that
environmental factors play a significant role in trig-
gering illness, and another is that epigenetic factors
are involved. Epigenetic factors are modifications of
genes and chromosomes that can influence the degree
to which genes are turned on or off. These two factors
may interrelate as the environment can cause epigen-
etic changes.
Genetic epidemiology
The first question in considering the genetics of
depression is whether or not the illness is heritable.
Only when heritability has been firmly established
does it make sense to begin looking for the particular
 Neurotrophic factors
BDNF EGFR
NT-3
NT-4
IGF-1
TRK EGFR AMPAR LICAM NMDAR EPHB2 DOPAMINE RECEPTORS GPCR

AKAP9
ADENYLATE CYALASE
KALRN
INTEGRIN CAMKK2 SNAP91 CDK5RAP3
RAC
MAP3K71P2

AB12 FLJ13386 CDKS cAMP

SPTBN1
NEF3 NCK
SOS1

SPTAN1 SNX6 PDE4B

TNIK
NDEL1 GNB1
SH3BP5 TRAF3IP1
RAS CDCSL
PAFAH1B1
ANK2
CLU

JNK2 JNK3 JNK1 DST

SEC3L1 DISC1 SRGAP3

TRIM9
MACF1
RHO GTPASES
TRI0
SNAP25

DMO
RHO GTPASES

Figure 7.1 Protein interactome simulated by pathway analyses. Iterative yeast-two hybrid screens, combined with detailed pathway, and
functional analyses suggest DISC1 interacts with proteins involved in key processes involved in neurodevelopment. Deficits in these processes
may underlie decreased dendritic branching, arborisations and neuropil size observed pathologically. (Adapted from [16], with permission.)
(a) Embryo Figure 7.2 Brain region specific gene
manipulations. (a) Schematic
representation of bilateral in utero injection
of constructs followed by their
E14 incorporation by electroporation into
Injection progenitor cells in the ventricular zone at
– + embryonic day 14 (E14). Migrating cells
+ + with green fluorescent protein (GFP) are
visualized at E18 after injection of a GFP
expression construct. (Adapted from [46],
with permission.) (b) Stereotaxic
coordinates and actual injection of
lentivirus-based enhanced GFP (EGFP).
– – Stereotaxic coordinates were determined
from the rat brain in stereotaxic
coordinates. The coordination in Sprague–
E18 Dawley rat at postnatal day 21 was
stereotaxically injected with lentivirus
+ – containing EGFP at the coordinates of
AP = þ2.2; ML = þ0.9; DV = þ2.0, þ1.5,
þ1.0 from the bregma. The crosses
indicate injected sites. Cx, cerebral cortex;
Hip, hippocampus; Th, thalamus; Str,
striatum; Amy, amygdala; Hypo,
hypothalamus; VTA, ventral tegmental
area; P, pontine tegmentum. (Adapted
from [47], with permission.)

(b) Adult brain

medial lateral
0 1 2 3 4 mm

Cx
Cx Hip

Str Th
Amy

AP +2.2 mm AP –1.34 mm
anterior bregma posterior
3 2 1 0 –1 –2 –3 –4 –5 –6 –7 –8 mm
0 dorsal
Cx 1
Hip 2
Str 3
Th
200 μm 4
VTA
Hypo 5
6 ventral
Figure 7.3 Dynamic changes in the adolescent brain. Disturbances generated by susceptibility genes and environmental insults during early
development (three stars on the left-hand side) may impair some of the crucial processes in early development, including progenitor cell
proliferation, neuronal migration and dendritic arborization and outgrowth. Independent of such initial risks/insults, intrinsic disease-associated
factors might also directly affect postnatal brain maturation (two central stars) contributing to the emergence of schizophrenia (SZ) in young
adulthood. (Adapted from [7], with permission.)
(a) Human Mouse

Control SZ 5

Ventricle volume (mm3)

4

0
WT Tg

100 WT
(b)
WT Tg Tg
80

Cells/section
Control SZ
60
PV
40

20
CB
0
PV CB
Figure 7.4 Anatomical and histological abnormalities in schizophrenia patients and in mouse models. (a) Enlarged lateral ventricles as
detected by in vivo magnetic resonance imaging (MRI) in schizophrenia patients (left) and in a DISC1 mouse model (right). (b) Decreased
parvalbumin (PV) expression in the prefrontal cortex of schizophrenia patients (mRNA, left) and a DISC1 mouse model (right). (Adapted
from [60], with permission.)
(a) Histone tail

DNA
H 2B

H4
H3

H2A

A Acetylation
Histone
M Methylation

(b) Active P Phosphorylation

Histones Histone Transcription factor Permissive
tail +
A A A M M
P A A

A Co-Act
A M A A
M
DNA Basal transcription
complex

Inactive
M M
P P
Repressed
Rep Rep
M Rep
M M M Rep A M M
M
A Rep
M
M
?
M

Rep M
M M
M
M M

Figure 8.1 Chromatin remodeling. (a) Picture of a nucleosome showing a DNA strand wrapped around a histone octamer composed
of two copies each of the histones H2A, H2B, H3, and H4. The amino (N) termini of the histones face outward from the nucleosome complex.
(b) Chromatin can be conceptualized as existing in two primary structural states: as active, or open, euchromatin (top left) in which histone
acetylation (A) is associated with opening the nucleosome to allow binding of the basal transcriptional complex and other activators of
transcription; or as inactive, or condensed, heterochromatin where all gene activity is permanently silenced (bottom left). In reality, chromatin
exists in a continuum of several functional states (active; permissive [top right]; repressed [bottom right]; and inactive). Enrichment of
histone modifications such as acetylation and methylation (M) at histone N-terminal tails and related binding of transcription factors and
co-activators (Co-Act) or repressors (Rep) to chromatin modulates the transcriptional state of the nucleosome. Recent evidence suggests that
inactivated chromatin may in some cases be subject to reactivation in adult nerve cells, although this remains uncertain.
 (c)
Demethylation Demethylation

HDM HMT HDM HMT

Methylation Methylation
(activating) (repressing)

M M
M M
H3 K4 K27 S28 M
K9 K23 K36
S10 K14 K18 P K79
A A
P A A Histone
tail
Acetylation Phosphorylation
(activating) (activating)

HAT HDAC PK PP

Deacetylation Dephosphorylation

Figure 8.1 (c) Summary of common covalent modifications of H3, which include acetylation, methylation and phosphorylation (P) at several
amino acid residues. H3 phosphoacetylation commonly involves phosphorylation of S10 and acetylation of K14. Acetylation is catalysed by
histone acetyltransferases (HATs) and reversed by histone deacetylases (HDACs); lysine methylation (which can be either activating or
repressing) is catalysed by histone methyltransferases (HMTs) and reversed by histone demethylases (HDMs); and phosphorylation is catalysed
by protein kinases (PK) and reversed by protein phosphatases (PP), which have not yet been identified with certainty. K, lysine residue;
S, serine residue. (From [3].)
Figure 8.2 Regulation of chromatin structure by drugs of abuse. Drug-induced signaling events are depicted for cocaine and amphetamine.
Cocaine and amphetamine can increase cyclic adenosine monophosphate (cAMP) levels in striatum, which activates protein kinase A (PKA)
and leads to phosphorylation of its targets. This includes the cAMP response element binding protein (CREB), the phosphorylation of
which induces its association with the histone acetyltransferase, CREB binding protein (CBP) to acetylate histones and facilitate gene activation.
This is known to occur on many genes including fosB and c-fos in response to psychostimulant exposure. ΔFosB is also upregulated by chronic
psychostimulant treatments, and is known to activate certain genes (e.g. cdk5) and represses others (e.g. c-fos) where it recruits HDAC1.
This repression of c-fos also involves increased repressive histone methylation, which is thought to occur via the induction of specific histone
methyltransferases. It is not yet known how cocaine regulates histone demethylases (HDM) or DNA methyltransferases (DNMTs). Cocaine also
activates the mitogen activated protein kinase (MAPK) cascade, which through MSK1 can phosphorylate CREB and histone H3 at serine 10.
Cocaine promotes H3 phosphorylation via a distinct pathway, whereby PKA activates protein phosphatase 2A, leading to the
dephosphorylation of serine 97 of DARPP32. This causes DARPP32 to accumulate in the nucleus and inhibit protein phosphatase 1 (PP1) which
normally dephosphorylates H3. Chronic exposure to psychostimulants is also known to increase glutamatergic stignaling from the prefrontal
cortex to the NAc. Glutamatergic signaling elevates Ca2þ levels in NAc synapses and activates CaMK (calcium/calmodulin protein kinases)
signaling, which, in addition to phosphorylating CREB, also phosphorylates HDAC5. This results in nuclear export of HDAC5 and increased
histone acetylation on its target genes (e.g. NK1R [NK1 or substance P receptor). (From [3].)
(a) Figure 8.3 Regulation of the bdnf gene
HDAC5 HDAC5 by social defeat. (a) In the absence of
stress, the chromatin state of brain-
A A A derived neurotrophic factor (Bdnf) is at
a basal level, characterized by moderate
+ levels of histone H3 acetylation and
Bdnf expression
virtually no H3K27 dimethylation. In this
state, histone deacetylase 5 (HDAC5)
A A might repress unnecessary activation of
M
M BDNF and maintain a chromatin balance.
(b) Chronic defeat stress induces the
specific and prolonged dimethylation
(b) HDAC5 HDAC5 of histone H3K27. This induces a more
M M M M M “closed” chromatin state at bdnf
A M M A M M A M promoters P3 and P4, and a
corresponding repression of Bdnf
–
transcripts III and IV expression. H3
Bdnf expression acetylation and HDAC5 regulation are not
affected after chronic defeat stress alone,
corroborating the idea that the main
A A
repressive marker after chronic stress is
M M M M M
M M M M M histone methylation. (c) Chronic
imipramine (antidepressant) treatment
(c) after defeat stress downregulates Hdac5
M M M M
M M M M expression and increases H3 acetylation,
A A A A A with little if any change in H3K27
dimethylation. Imipramine-dependent H3
+ hyperacetylation at the bdnf promoters
Bdnf expression
P3 and P4 allows partial “reopening” of
the repressed chromatin state caused by
M
A A A A A M H3–K27 dimethylation defeat stress, and results in transcriptional
M M M M reactivation of the bdnf gene. K, lysine
M M M M A H3 acetylation residue. (From [3].)

Figure 9.6 Amygdala activity related to masked and unmasked

fear processing in healthy subjects. Enlarged views of the right
amygdala illustrating: (1) the dorsal amygdalar cluster from the
nonmasked fear (F-N) comparison (coronal view at Y = 8 [A] and
axial view at Z = 16 [B]); and (2) the basolateral amygdalar cluster
from the correlation of masked fear-induced activity (FN-NN) with
trait anxiety (coronal view at Y = 8 [C] and axial view at Z = 28
[D]). The color bar indicates the significance, with lighter colors
Non-masked Masked fear indicating a greater difference between the respective fearful and
FEAR (F-N) (FN-NN) vs. STAI-T neutral conditions. (From [200], with permission from Elsevier.)

T
value
6
5 5
4 4
3 3
2 2
1 1
0 0
Figure 9.7 Amygdala activation in response to fearful faces is
modulated by catechol-O-methyltransferase (COMT) genotype.
Whole brain voxel-wise t-map (uncorrected, P > 0.001) of a single
panic disorder patient with a genotype 472G/A (i.e. val-met
heterozygous) comparing activation for the fearful versus the
no-face condition. Note that there is significantly greater amygdala
activation (center of cross hair) in the fearful condition. This increase
was not observed among patients homozygous for the met
encoding allele. The transverse plane is the original one; the coronal
and sagittal planes are planar reconstructions orthogonal to the
original image. Reader’s right is subject’s right. Emotional face stimuli
and no-face control stimuli were controlled for dynamics and
luminance. (From [81], with permission from Cambridge
University Press.)
Figure 9.8 Serotonin 1A receptor genotype moderates fear
processing in panic disorder patients. Random effects statistical
parametric map for the fearful versus neutral faces contrast overlaid
on a three-dimensional canonical Montreal Neurological Institute
brain showing right-lateralized activity differences in the prefrontal
cortex between the two patient groups (5HT1A1019GG versus
CC/CG; p < 0.001, uncorrected). Patients homozygotes for the g
alleles had reduced activation in orbitofrontal, ventromedial
prefrontal, and cingulate cortices when viewing unmasked fearful
faces, as compared to patients with the CG or CC genotype.
(From [213], with permission from Cambridge University Press.)
 STin2 IIe425Val
SNPs VNTR
rs25531,
67 Alternate
(a) Gene rs25532 2 34 5 8 9
1011
1A 1C 1B 12 13 14 polyadenylation sites
5’

3’

5HTTLPR
(LA,LG,SA,SG) G56A
Alternative
splicing
34

(b) SERT SNPs: 8 Extracellular

11 12
6 7
1. T4A1 9. P339L TMS
9 1013
2. G56A2 10. L362M 2 14 15 Intracellular
16
3. S214S 11. L383L
COOH
NH2
4. E215K1 12. A419A 1
5. H235H 13. 2X1425V(/L)
6. L255M 14. T439T
7. S293F 15. K605N1 or 5HT transport in transfected cells
1No response to PKG/p38 MAPK activation
8. G308G 16. P621S1 2No response to 8BrcGMP

Figure 11.2 Human SERT gene organization, with multiple functional variants.

(a)
A189V

Variations in OCD/ R13C A148insGPAGA T523K P606T K910R

TTm patients

Guanylate-Kinase-associated protein (GKAP) domain

Variations in
T156M A189V
controls subjects

(b) Human MRGYHGDRGSHPRPARFADCQ TLPYQRGPAGAGPGPAPGTGTAPERSE SHSLBAPGKRDY OLPLLAAPAAVSGRP VPPRASPKPPT LLEPKEEXKVPPPIP

Rhesus MRGYHGDRGSHPRPARFADCQ TLPYQRGPAGAGPGPAPGTGTAPERSE SHSLBAPGKRDY OLPLLAAPAAVSGRP VPPRASPKPPT LLEPKEEXKVPPPIP
Cow MRGYHGDRGSHPRPARFADCQ TLPYQRGPAGAGPGPAPGTGTAPERSE SHSLBAPGKRDY OLPLLAAPAAVSGRP VPPRASPKPPT LLEPKEEXKVPPPIP
Horse MRGYHGDRGSHPRPARFADCQ TLPYQRGPAGAGPGPAPGTGTAPERSE SHSLBAPGKRDY OLPLLAAPAAVSGRP VPPRASPKPPT LLEPKEEXKVPPPIP
Mouse MRGYHGDRGSHPRPARFADCQ TLPYQRGPAGAGPGP--GSGAAPEARSE SHSLBAPGKRDY OLPLLAAPAAVSGRP VPPRASPKPPT LLEPKEEXKVPPPIP
Rat MRGYHGDRGSHPRPARFADCQ TLPYQRGPAGAGPGP--GSGAAPEARSE SHSLBAPGKRDY OLPLLAAPAAVSGRP VPPRASPKPPT LLEPKEEXKVPPPIP

R13C A148insGPAGA T156M A189V T523K P606T K910R

Figure 11.3 Identified rare nonsynonymous polymorphisms in synapse-associated protein 90/postsynaptic density-95-associated protein 3
(SAPAP3). (a) Schematic of SAPAP3, which consists of 10 coding exons (boxes). Seven rare changes were identified in trichotillomania and
obsessive–compulsive disorder (OCD) patients (a), but only two in controls (b). Most mutations fell into exon 1; however, three changes
affected the conserved guanylate-kinase-associated protein (GKAP) domain. (b) SAPAP3 is highly conserved between species ( 97% identical
amino acids between human and mouse). Accordingly the identified rare changes affected conserved residues.
 Chr 1 Chr 2 Chr 3 Chr 4 Chr 5 Chr 6 Chr 7 Chr 8 Chr 9 Chr 10 Chr 11 Chr X
1.1 5.1 8.1
3.1 10.1 X.1
9.1
7.1

5.2 11.1
2.1
10.2
3.2 5.3
3.3 7.2 9.2 11.2 X.2
4.1 11.3
4.2 10.3
4.3 6.1 7.4 7.3
7.5 X.3
7.6 7.7 9.3 X.4
3.4 7.8 9.4
6.2 9.5
7.10 7.9 X.5
1.2 7.11 X.6
2.2 2.3
2A 3.5
3.6 4.4 12.1

1.3 16.1 15.1

2.5 15.2
17.2 12.2
17.1 15.3 13.1
20.1 19.1 16.2
17.3 16.3
21.1 20.2 14.1
17.4
17.5
22.1 16.4
21.2 15.4
22.2

Chr 22 Chr 21 Chr 20 Chr 19 Chr 18 Chr 17 Chr 16 Chr 15 Chr 14 Chr 13 Chr 12 Chr Y

ID Feature Position Refs ID Feature Position Refs ID Feature Position Refs

1.1 Loss 1p36 3 7.4 RELN 7q22 117–121 15.3 Gain 15q11–15q13 4
1.2 Linkage 1q21–1q23 116 7.5 MET 7q31 70,123 15.4 Linkage 15q22–15q26 5
1.3 DISCI 1q42 122 7.6 Loss 7q31 5 16.1 TSC2 16p13 110
2.1 NRXNI 2p16 5,34 7.7 Linkage 7q32–7q34 15 16.2 Loss 16p11 4,20,35,44
2.2 Loss 2p24 4 7.8 CADPS2 7q31 5,45 16.3 Gain 16p11 20,35,44
2.3 Linkage 2p24–2q31 57,58,TT2 7.9 Linkage 7q34–7q36 14,52,68 16.4 Loss 16q21 5
2.4 SLC 25412 2p24 124–126 7.10 CNTNAP2 7q35–7q36 37–40 17.1 Loss 17p12 5
2.5 Loss 2q37 4 7.11 EN2 7q36 129,130 17.2 Gain 17p12 107
3.1 OTXR 3p25 127,128 8.1 Gain 8p23 5 17.3 SLC6A4 17q11 131–134
3.2 Loss 3p14 4 9.1 Linkage 9p24 5 17.4 Linkage 17q11–17q21 51,54,135
3.3 Gain 3p14 4 9.2 Loss 9q12 5 17.5 ITGB3 17q21 136,137
3.4 Linkage 3q22 15 9.3 Linkage 9q33 5 19.1 Linkage 19p13 140
3.5 Linkage 3q25–3q27 138,139 9.4 Linkage 9q34 15 20.1 Loss 20p13 5
3.6 Loss 3q27–3q28 3 9.5 TSCI 9q34 110 20.2 Loss 20p13 5
4.1 Loss 4q21 3 10.1 Loss 10q14–10p15 4 21.1 Linkage 21q11 55
4.2 Loss 4q21–23 3 10.2 Gain 10q11–10q21 4 21.2 Loss 21q22 3
4.3 Linkage 4q22–4q2S 15 10.3 PTEN 10q23 141 22.1 Loss 22q13 4
4.4 Loss 4q35 5 11.1 Linkage 11q12–11p13 5 22.2 SHANK3 22q13 21,22,142,143
5.1 Linkage 5p15 5 11.2 DHCR7 11q13 108 X.1 NLGN4X Xp22 28
5.2 Linkage 5p13–5q11 140 11.3 Linkage 11q13–11q14 15 X.2 NLGN3 Xq13 28
5.3 Linkage 5q12 5 12.1 CACNAIC 12p13 24 X.3 Linkage Xq21–Xq25 140
6.1 GRK2 6q21 144–146 12.2 AVPRIA 12p14–12q15 147 X.4 Gain Xq24 3
6.2 AHII 6q23 106 13.1 Gain 13p14 5 X.5 FMRI Xq27 105,148
7.1 Loss 7p21 4 14.1 Linkage 14q23 149 X.6 MECP2 Xq28 109
7.2 Loss 7q11 3 15.1 UBE3A 15q11 102,103
7.3 Linkage 7q22–7q32 52,111–113 15.2 GABRB3 15q12 25,114,115

Figure 16.1 Loci implicated in autism spectrum disorder (ASD) etiology. Entries in the ID column of the table map are entries to the
ideograms of individual chromosomes. Red and yellow bars correspond to de novo losses and gains, respectively, which are observed in cases
but not in controls. Green bars correspond to genes that are observed to modulate ASD risk (either through a rare syndrome or through
genetic association): light green and dark green bars represent promising or probable candidate genes, respectively, as defined in the table
map. Regions shaded in purple correspond to linkage peaks. Only human data were considered in the assembly of the table. AHI1, Abelson
helper integration site 1; AVPR1A, arginine vasopressin receptor 1A; CACNA1C, calcium channel voltage-dependent L type a 1C subunit;
CADPS2, Ca2þ-dependent activator protein for secretion 2; CNTNAP2, contactin associated protein-like 2; DHCR7, 7-dehydrocholesterol
reductase; DISC1, disrupted-in-schizophrenia 1; EN2, engrailed homeobox 2; FMR1, fragile X mental retardation 1; GABRB3, g-aminobutyric acid
(GABA) A receptor b3; GRIK2, glutamate receptor ionotropic kainate 2; ITGB3, integrin b3; MECP2, methyl CpG binding protein 2; MET, met
proto-oncogene; NLGN3, neuroligin 3; NLGN4X, neuroligin 4 X-linked; NRXN1, neurexin 1; OXTR, oxytocin receptor; PTEN, phosphatase and
tensin homologue; RELN, reelin; SHANK3, SH3 and multiple ankyrin repeat domains protein 3; SLC25A12, solute carrier family 25 (mitochondrial
carrier, Aralar) member 12; LC6A4, solute carrier family 6 (neurotransmitter transporter, serotonin) member 4; TSC1, tuberous sclerosis 1;
TSC2, tuberous sclerosis 2; UBE3A, ubiquitin protein ligase E3A. (From [8], with permission from Macmillan Publishing Ltd.)
 5.5 1.0
5.0 0.9
4.5 0.8
Information content
4.0 Wave 1 0.7
3.5 ZLR= 4 .69, 0.6
Wave 2

Information content
lod= 4 .79,
3.0 p < 0.000002 All famillies 0.5
2.5 HaploSim analysis 0.4
ZLR

2.0 0.3
1.5 0.2
1.0 0.1
0.5 0.0
0.0
–0.5 Mb:
80 85 90 95 99
0 5 10 15 20 25 30 35 40 45
Location (cM)

Figure 18.1 Linkage to chromosome 15q25–26 in the GenRED study. Shown are Z likelihood ratio score statistics from analyses of linkage
of recurrent early-onset major depression to 88 chromosome15q single nucleotide polymorphisms in 631 families of predominantly
European ancestry for wave 1, wave 2, all families, and for information content. Physical locations in megabases from the p-telomere are shown,
with the peak at about 93 Mb. On the deCODE genetic map, the location is from 85.2 to 133.6 cM. Also shown are results of HaploSim analysis
demonstrating the absence of any systematic bias in the multipoint analysis due to linkage disequilibrium. (From [48], with permission.)

M IN TF NR2F2 Cluster 2
IN PIK3R1
IN CUGBP2

IN APOB

M IN PON2

IN C036
M IN GSTP1 M IN EPHX2

IN SPON1

IN ADAMTSL1 M IN RYR2 IN HOMER1

Cluster 1
IN SLC8A1 M IN KCNJ6
IN HAS2

IN EDN1 M IN TNFRSF1A IN DAPK1

M IN PTPRK
M IN PTGER4
M IN SLC1A2
IN TF MYBL2
IN A2BP1 M IN AKR1B1 IN TNFRSP21
IN SLC6A11
IN TF MITF
IN NRG1
IN PTPRS M IN FN1
IN SCNN1A
M IN TF NR2E1 IN CADM1
M IN RPS6KA1 Cluster 3
M IN RAP1B

M IN TGFBR3
IN TPM1
IN CALD1

IN SERPINA4
IN EFNAS

IN EPHAS

Figure 18.3 Results of a literature-based pathway analysis that includes all genes from the Munich Antidepressant Response Signature
project genome-wide association study (GWAS) that correspond to the single nucleotide polymorphisms (SNPs) implicated by the STAR*D
replication GWAS sample. Genes were categorized as related when they were co-cited in the same sentence with a functional descriptor in
between. There are 41 genes that cluster around fibronectin 1 (FN1) (cluster 1), ADAMTS-like 1 (ADAMTSL1) (cluster 2), and endothelin 1 (EDN1)
(cluster 3). Genes with corresponding SNPs that achieved nominal significant replication in the STAR*D sample are shaded in red; green
lines indicate transcription factor (TF) binding site matches in target promoters; the line with the yellow circle indicates annotation by
Molecular Connections experts. IN indicates input gene; M, part of a metabolic pathway. From [105], with permission.
 (a) chr15 (q25.1)

(b) rs16969968

76589924 76711042
1.0 80

Recombination rate (cM/Mb)

0.8
60
R-Squared

0.6
40
0.4

20
0.2

0.0 0
CRAB1 IREB2 LCC123088 ADAMTS7
PSMA4
CHRNA5
CHRNA3
CHRNB4

76420 76545 76670 76795 76920

Chromosome 15 position (hg18) (kb)

Figure 23.3 Graphical representation of chromosome 15q25. (a) Chromosome 15 with the region q25.1 demarked by the red line, the
location of CHRNA5 and CHRNA3. This figure was created with the University of California, Santa Cruz (UCSC) genome browser (http://genome.
ucsc.edu). (b) The 100 kb region surrounding rs16969968 (large diamond). Diamonds indicate single nucleotide polymorphisms (SNPs), and the
size of the diamond is directly proportional to the r2 of the SNP with rs16969968. The dashed lines delineate the boundaries for SNPs that
have r2  0.8 with rs16969968. This figure was created using SNAP (http://www.broad.mit.edu/mpg/snap/).
 12

(a)
9

Frequency (Hz)
Fz

Cz
6

Pz
3

Theta

-216 -31 153 338 522 706 891 1075 1259

(b) Time (ms)

Chromosome 7
3.5 θ
Fz, Max LOD = 3 .16 at 161 cM
3
Cz, Max LOD = 3 .6 at 164 cM

Pz, Max LOD = 2 .29 at 162 cM

2.5

2
LOD

1.5

CHRM2
0.5
GRM8

0
D7S1790

D7S1802

D7S1838

D7S2846

D7S1830

D7S3046
D7S1870

D7S1797

D7S1796

D7S1799

D7S1817

D7S2847
D7S1809

D7S1804

D7S1824

D7S1805
D7S513

D7S629
D7S673

D7S817

D7S521
D7S691
D7S478
D7S679
D7S665

D7S820

D7S821

D7S490

D7S509

D7S794
NPY2

0 20 40 60 80 100 120 140 160 180

Chromosome Position (cM)
CHRM2
exon

exon

exon

exon

exon

(c)
1 2 3 4 5 6
Coding
5’-UTR 3’-UTR
Sequence

(d) SNPs rs6948054 rs1424548

rs1378646 rs1378650
rs1455858
rs7782965
rs978437

Figure 29.1 Illustration of endophenotype strategy: from neurocognitive endophenotypes to genes. (a) Endophenotype (Theta oscillation),
Theta (y, 3–7 Hz) event-related oscillations (EROs) underlying the processing of target stimuli during P3 production (300–700 ms) in a visual
oddball task as seen in this time-frequency representation (right panel). Head plot (left) displays scalp topography of theta ERO power.
(Theta EROs are correlates of impaired cognitive brain processes in alcoholism and risk.). (b) Genetic linkage analysis scans the entire genome
to assess chromosomal regions that contain polymorphic genetic markers that are linked to a quantitative trait (theta ERO power) within
families. This is a logarithm of odds (LOD) score plot of linkage for theta ERO power at frontal (Fz), central (Cz), and parietal (Pz) electrodes on
chromosome 7 with a significant linkage peak that contains 2 candidate genes under the linkage peak – CHRM2 and GRM8. (c) Candidate
gene studies focus on genes underlying significant linkage peaks and/or genes with relevant biological significance. This is a schematic
diagram of the CHRM2 gene indicating its coding region (gray) and exons. (d) Genetic association tests are performed for each single
nucleotide polymorphism (SNP) in the candidate gene and the trait variable. SNPs are identified within and flanking the candidate CHRM2
gene (exons and introns). SNPs with significant associations (gray boxes) with the endophenotype indicate loci of interest. (The right-most
box with two SNPs lies in the region beyond the 3’ UTR of the gene.)
(a) (b)

Figure 29.2 (a) Linkage plots showing maximum logarithm of odds (LOD) scores with significant linkage peaks over the GABAA receptor gene cluster on chromosome 4 with two resting
EEG phenotypes. Red trace: Resting EEG beta band power – beta 2 (16.5–20.0 Hz). The dataset consists of 1553 individuals from 250 families. Green trace: Resting EEG high theta band
(6–7 Hz) interhemispheric coherence at parieto-occipital bipolar pairs of electrodes (P4_O2-P3_O1). The dataset consists of 1312 individuals from 251 families. (b) Linkage plot showing maximum
LOD scores with significant linkage peaks on chromosome 7, over the region harboring two candidate genes: a cholinergic muscarinic receptor gene (CHRM2) and a glutamate receptor gene
(GRM8). Blue trace: the central midline theta (4–5 Hz) ERO band power (between 300–700 ms, P3 latency window for visual target case during visual oddball task) on chromosome 7. The dataset
consists of 1337 individuals from 253 families; Green trace: Resting EEG theta band (6–7 Hz) interhemispheric coherence at centro-parietal bipolar pairs of electrodes (C4_P4-C3_P3) on
chromosome 7. The dataset consists of 1312 individuals from 251 families.
(a) Adult subjects
Control Alcohol dependent
Frequency (Hz)

Frequency (Hz)
10 10

5 5

0 0
0 200 400 600 0 200 400 600
Time Time

1 2 3 4 1 2 3 4
Amplitude(μV) Amplitude(μV)

(b)
Adolescent subjects
Low risk High risk
12

Frequency (Hz)
12
Frequency (Hz)

θ
7
7 θ
θ
δ 0
δ
0
0 200 400 600 0 200 400 600
Time
Time δ
2 4 6 2 4 6
Amplitude(μV) Amplitude(μV)

Figure 29.3 S-transform derived time-frequency representations of the average instantaneous amplitude that are z-scored for each
frequency. Instantaneous amplitudes were averaged across individual trials of subjects so that nonphase locked or imprecise phase locked
oscillatory energy is preserved. (a) Plots for the target condition at central (Cz) electrode in 120 alcoholic (right) and 120 control (left) subjects.
Center panel – Topographic headplots of the time-frequency regions of interest (TFROIs) indicate that the theta band (4–5 Hz) power at
300–500 ms shows frontal maxima while the delta band (1–3 Hz) power at 350–700 ms has posterior maxima. Note that alcoholics have
weaker responses in both theta and delta bands. (b) Plots for the target condition at frontal (Fz) electrode in visual oddball task in 87 high risk
adolescent offspring of alcoholics from Collaborative Study on the Genetics of Alcoholism (COGA) families (right) and 57 matched low risk
offspring of nonalcoholics from control families (left). Center panel – Topographic head plots for the TFROI that extends from 300 to 700 ms
for each band revealing frontal maxima for theta band power and parietal maxima for delta band power. Note that high risk adolescents
have weaker responses in both theta and delta bands to target stimuli, similar to the alcoholics.
 Chapter 18: Genetics of major depression

Table 18.1 Family studies of major depression.

Proband/study Sample size (relatives) Age-adjusted lifetime prevalence in FDRs (%)

BP BP-I BP-II SA MDD

Major depression
Smeraldi et al., 1977 [6] 185 0.6 – – – 8.0
Tsuang et al., 1980 [7] 483 2.0 – – – 13.6
Abrams et al., 1980 [8] 106 4.7 – – – 7.5
Jakimow-Venulet, 1981 [9] 306 0.5 – – – 9.5
Gershon et al., 1982 [10] 166 – 1.5 1.5 0.7 16.6
Baron et al., 1982 [11] 143 2.2 3.0 17.7
Weissman et al., 1984 [12] 810 – 0.9 1.9 0.3 17.6
Andreasen et al., 1987 [4] 1171 – 0.6 2.9 0.2 28.4
McGuffin et al., 1987 [13] 315 – – – – 24.7
Maier et al., 1993 [14] 697 1.8 – – 0.5 21.6
Weissman, 1993 [15] 651 – – – – 24.4
Weighted mean 5033 1.7 0.8 2.4 0.5 20.5
Controls
Tsuang et al., 1980 [7] 541 0.3 – – – 7.5
Gershon et al., 1982 [10] 265 – 0 0.5 0.5 5.8
Weissman et al., 1984 [12] 521 – 0.2 1.1 0.2 5.9
Maier et al., 1993 [14] 419 1.8 – – 0.4 10.6
Weissman et al., 1993 [15] 255 – – – – 5.5
Weighted mean 2001 1.0 0.1 0.9 0.3 7.3
Abbreviations: BP, bipolar disorder; FDR, first-degree relative; MDD, major depressive disorder; SA, schizoaffective disorder; – ¼ not studied
or reported.

genes responsible for this inheritance. Determination
of heritability starts with studies of familial aggrega-
tion and proceeds with twin and adoption studies.
If a disease has a genetic basis, it is expected to run
in families. Table 18.1 shows that the lifetime preva-
lence of depression is elevated in first-degree relatives
of depressed probands as compared to relatives of
controls [4, 6–15]. It appears that rates of bipolar
disorder are slightly elevated as well. A useful way of
estimating the strength of the genetic contribution to
illness is to apply the concept of relative risk to sib-
lings, which entails comparing the rate of illness in
siblings of ill probands with that in the general popu-
lation. The relative risk for MDD is
2.8, suggesting
a modest genetic component for the broad phenotype.
An increased relative risk to siblings is observed
when two additional clinical features of MDD, recur-
rent episodes and an early age at onset, are considered
[16]. Most data suggest that familial risk is greater
when the proband’s age of onset is below 40 and
greater still when it is below 30 or 20. Lyons et al.
[17] reported that in 3372 male veteran twin pairs,
heritability of MDD was 0.47 when the age at onset
was below age 30, and 0.10 when it was above 30, and
Bland et al. [18], in a study of 763 first-degree rela-
tives of 75 probands with MDD, found that relatives
of probands with an early age at onset and recurrent
depression had a 17.4% risk of depression while rela-
tives of probands with single-episode depression and
a late age at onset had only a 3.4% risk. The relative

213
 Chapter 18: Genetics of major depression

Table 18.2 Twin studies of major depression.

Findingsa

Proband/study Sample size (pairs) Concordance (%)

MZ DZ Heritabilityb
Bertelsen et al., 1977 [20] 44 54 24 0.39
Torgersen, 1986 [21] 92 27 12 0.54
Andrews, 1990 [22] 82 7 9 0
McGuffin et al., 1996 [23] 177 46 20 0.48
Lyons et al., 1998 [17] 3372 23 14 0.36
Bierut et al., 1999 [24] 2662 33 26 0.30
Kendler et al., 2000 [25] 3790 44 39 0.34
Weighted mean 10 219 34 26 0.34
a
Only studies in which a majority of subjects were directly interviewed are included.
b
Where authors calculate heritability, these figures are provided. If heritability is not provided in a report, Holzinger’s heritability
(MZ concordance  DZ concordance/100  DZ concordance) has been calculated.
Abbreviations: DZ, dizygotic; MZ, monozygotic.

risk to siblings for early-onset recurrent depression is
probably at least 4–5 [19].
Familial aggregation of disease suggests, but does
not prove, a genetic contribution to disease. Environ-
mental factors, such as emotionally traumatic family
experiences or other stressors, for example, could also
lead to familial aggregation. Twin studies have been
the approach used most widely to attempt to disen-
tangle these contributions. The logic of twin studies is
as follows. (Please also see Chapter 1, which discusses
twin, family, and adoption methods in greater detail.)
Identical or monozygotic (MZ) twins are 100% gen-
etically the same, whereas fraternal or dizygotic (DZ)
twins share just 50% of their genes; yet the two twin
types are assumed to be no different in the degree to
which they share environments. Therefore, any
increased similarity in manifestation of depression
detected in MZ twins as compared with DZ twins
should be due to the greater genetic similarity of the
former. The main measure in these studies is the
concordance rate for illness. That is, starting with an
ill twin as the proband, what is the rate of illness in
the co-twin? As seen in Table 18.2, seven studies [17,
20–25] have assessed MDD using primarily a direct
interview approach and reported on concordance
rates; they found rates of 34% in MZ twins and 26%
in DZ twins, with a heritability of 0.34. The largest
twin study to date looked at 15 493 complete twin
pairs born in Sweden between 1886 and 1958 and
contacted between 1998 and 2003 for an assessment
of the presence or absence of depressive epsiodes.
The investigators reported an overall heritability of
0.38, with a significant difference between the sexes.
The heritability for women was 0.42 while that for
men was 0.29 [26].
Many observers have cited the lack of complete
concordance between MZ twins as evidence that
environmental factors such as stressful life events
must play a role in the etiology of MDD. More
recently, others have suggested that this lack of com-
plete concordance could be due to epigenetic factors –
that is, factors that affect the control of gene expres-
sion – and that these factors may be influenced by
the environment [27]. In support of this hypothesis,
there is evidence that MZ twins, despite having
identical DNA sequence for all their genes, may
differ in the way their genes are expressed [28].
Patterns of concordance can be converted to a herit-
ability value – often thought of as the percentage of
the liability to illness that is genetic – in several ways.
Current twin studies employ a liability threshold
model [29], which simultaneously tests the heritabil-
ity of and environmental contribution to illness.
When individuals cross the liability threshold, they
develop illness. The model requires the specification
of a population prevalence of illness to indicate how

214

many people in the population lie on the ill side of
the threshold. Moreover, heritability estimates will
vary with the prevalence figures used, an issue of
particular importance for MDD, for which popula-
tion estimates have varied dramatically. For example,
one depression study reported that heritability was
48% when one estimate was used and 75% when
another was used [23].
A second paradigm for separating genetic from
environmental effects is the adoption study. The con-
ceptual basis for this approach is that in adoptees, the
genetic inheritance occurs through one set of parents,
while the cultural and environmental experience
occurs through a different set. The two sets of factors
and their potential association with illness can there-
fore be disentangled.
There are several possible ways to conduct an
adoption study. One that has been used for MDD is
the adoptees’ relatives method. In this method, ill and
control adoptees are identified as probands, and rates
of illness are then compared in the biological and
adoptive relatives of each group. If genetics plays a
role, the biological relatives of ill adoptees will have
elevated rates of illness as compared with the other
three relative groups.
Adoption studies are not easy to conduct as they
require access to large databases, and there are bar-
riers of confidentiality involved whose breach can be a
highly sensitive matter. Two such studies that have
been done have used national registries – those of
Sweden and Denmark. In these studies, direct inter-
views were not performed. In these two studies, more-
over, probands had diagnoses including “affect
reaction” and “neurotic depressive reaction”, which
are not readily reconcilable with current diagnostic
nomenclature. The results of one of these studies
supported genetic transmission of depression [30],
while those of the other did not [31]. Another adop-
tion study used relatives as probands, and measured
rates of depression in the adoptees of those relatives
who were affectively ill as compared with those who
were not. The results supported genetic transmission
[32]. Thus of the three adoption studies, two provide
support for a genetic vulnerability, and one does not.
In sum, MDD has a substantial genetic compon-
ent, suggesting that the identification of susceptibility
variants should be possible. This encouraged
researchers, beginning in the early 1980s to initiate
molecular studies to search for these variants. How-
ever, the magnitude of the genetic component appears
Chapter 18: Genetics of major depression
to be less than that for several other major psychiatric
illnesses discussed in this book, such as bipolar dis-
order, schizophrenia, and autism, but comparable to
that for alcohol and other substance dependence.
Gene-mapping
Linkage
Linkage is a method for identifying chromosomal
regions likely to harbor a disease susceptibility gene.
It hinges on the biological process of meiosis and the
recombination of homologous chromosomes that is
central to that process. A putative disease gene is
linked to a marker if it stays together with the marker
in the large chunks of chromosome that are recom-
bined. The resolution of linkage is low, on the order
of 20 Mb, because over the course of a couple of
generations, there is relatively little recombining.
The first linkage studies of depression were done
at the University of Iowa by Tanna, Go, Winokur, and
others. Using blood group antigens and enzymatic
and protein markers, they found several modest link-
age signals that could not be replicated [33–36]. Other
teams were also either unable to replicate these early
linkage findings for depression or obtained conflict-
ing results with evidence of linkage to other markers
[37–39].
These early studies of the genetics of depression
usually assumed that the inheritance of MDD was
fairly simple, that straightforward Mendelian mech-
anisms operated in the transmission of risk genes
and that only a few genes would be found to underlie
the development of depressive illness. It is now clear
that this is almost certainly not the case. MDD is a
genetically complex disease. Genetic mechanisms
appear to affect susceptibility to MDD rather than
causing it more or less directly as is seen in, for
example, sickle cell disease, Huntington’s disease, or
phenylketonuria, where the relative risk to siblings is
in the vicinity of 2500.
Because rates of MDD in women are about twice
those in men, investigators have wondered whether
susceptibility genes might differ to some extent by
gender. Because of the potential for sex-specific gen-
etic effects in MDD, gene-mapping studies have often
incorporated or even focused on sex-specific analyses.
One study found significant parametric evidence of
linkage to markers in 2q33–34 in 170 affected female
sibling pairs, but not in male pairs [40]. The region
215
 Chapter 18: Genetics of major depression
between the markers that yielded the peak logarithm
of odds (LOD) score includes the CREB1 gene, which
encodes a cyclic adenosine monophosphate (cAMP) -
responsive element-binding protein (CREB), an
attractive candidate gene because CREB has been
implicated in depression and antidepressant response.
Levels of CREB have been found to be abnormally low
in people with MDD and in the brain tissue of suicide
victims, and are altered by exposure of rat neurons
to antidepressants [41–43]. A second study found
evidence of linkage uniquely in families with at least
four affected males [44]. This linkage, in chromo-
somal region 12q22–23.2, was detected in a sample
of Mormon families in Utah in which ascertainment
was restricted to families with a minimum of four
affected relatives. In addition to subjects with recur-
rent MDD, individuals with only a single episode of
major depression were considered affected, as were
those with bipolar disorder (who made up about 15%
of the subjects with mood disorder in these families).
McGuffin et al. carried out a genome-wide linkage
scan in a sample of 497 sibling pairs concordant for
recurrent MDD [45]. There was suggestive evidence
for linkage on chromosome 1p36 where the LOD
score for female–female pairs exceeded 3.0 (but
reduced to 2.73 when corrected for multiple testing).
The region includes a gene – MTHFR – that in previ-
ous studies has been associated with depressive symp-
toms. Two other regions, on chromosomes 12q23.3–
24.11 and 13q31.1–31.3, showed evidence for linkage
with a nominal p < 0.01. The 12q peak overlaps with
a region previously implicated by linkage studies of
unipolar and bipolar disorders and contains a gene –
DAO – that has been associated with both bipolar
disorder and schizophrenia. The 13q peak lies within
a region previously linked strongly to panic disorder.
A fourth modest peak with a LOD of greater than 1.0
on chromosome 15q lies within a region that showed
genome-wide significant evidence of a recurrent
depression locus in a previous sibling pair study (the
GenRED study, see below).
In the Genetics of Recurrent Early Onset
(GenRED) project, we studied 297 families initially
(Wave 1) and carried out a linkage analysis. We
observed a genome-wide significant peak on distal
15q (LOD ¼ 3.73) [46], with the peak being even
stronger when the analysis was restricted to the 286
European-ancestry families (LOD ¼ 4.39). Later we
jointly analyzed our Wave 1 plus Wave 2 sample of
656 families, including 1494 informative “all possible”
216
affected relative pairs [47]. Evidence for linkage on
15q25–26 was reduced to the “suggestive” level.
In secondary sex-specific analyses, nearly significant
evidence for linkage was observed on chromosome
17p12 (LOD ¼ 4.77, excess sharing in male–male and
male–female pairs), and suggestive linkage on
chromosome 8p22–21.3 (LOD ¼ 3.49, excess sharing
in male–male pairs). We next genotyped 88 single
nucleotide polymorphisms (SNPs) across our
15q25–26 peak in 631 European-ancestry families
and carried out multipoint allele-sharing linkage ana-
lyses [48]. The maximum evidence for linkage was
LOD ¼ 4.78 at 109.8 cM (Figure 18.1). The exact
p-value (0.0000014) surpassed the genome-wide signifi-
cance threshold. It was estimated that the linked locus
or loci in this region might account for a 20% or less
population-wide increase in risk to siblings of cases.
In sum, as has been the case for other psychiatric
disorders, linkage studies in MDD have not consist-
ently identified the same chromosomal regions,
although there has been some encouraging overlap,
most prominently on chromosome 15q25–26. Unlike
the case for bipolar disorder and schizophrenia, the
number of genome-wide linkage studies done to date
has been small, and no meta-analysis of them has yet
been carried out.
Alternative phenotypes
One approach to potentially improving genetic link-
age signals is to attempt to reduce genetic heterogen-
eity through reducing clinical heterogeneity. We have
referred to several alternative phenotypes intended to
capture more homogeneous subsets of patients,
including using recurrence of episodes, earlier age at
onset, and single sex analysis. There are a number of
other phenotypic distinctions that might help define
more genetically homogenous subsets of patients.
For example, Camp et al. [49] examined recurrent,
early-onset MDD (MDD-RE) and anxiety disorders in
87 large, extended Utah pedigrees to investigate link-
age to three phenotypes: “MDD-RE”, “MDD-RE or
anxiety”, and “MDD-RE and anxiety”, where in the
latter definition the disorders must appear comorbid
within an individual. Pedigrees ranged in size from
2–6 generations and contained 3–42 individuals
affected with MDD or anxiety (718 total). In their
primary analyses, they identified three regions with at
least suggestive genome-wide evidence for linkage on
chromosomes 3centr, 7p, and 18q. The best linkage
 Chapter 18: Genetics of major depression

5.5 1.0
5.0 0.9
4.5 0.8
Information content
4.0 Wave 1 0.7
3.5 ZLR= 4 .69, 0.6
Wave 2

Information content
lod= 4 .79,
3.0 p < 0.000002 All famillies 0.5
2.5 HaploSim analysis 0.4
ZLR

2.0 0.3
1.5 0.2
1.0 0.1
0.5 0.0
0.0
–0.5 Mb:
80 85 90 95 99
0 5 10 15 20 25 30 35 40 45
Location (cM)
Figure 18.1 Linkage to chromosome 15q25–26 in the GenRED study. Shown are Z likelihood ratio score statistics from analyses of linkage of
recurrent early-onset major depression to 88 chromosome15q single nucleotide polymorphisms in 631 families of predominantly European
ancestry for wave 1, wave 2, all families, and for information content. Physical locations in megabases from the p-telomere are shown, with the
peak at about 93 Mb. On the deCODE genetic map, the location is from 85.2 to 133.6 cM. Also shown are results of HaploSim analysis
demonstrating the absence of any systematic bias in the multipoint analysis due to linkage disequilibrium. (From [48], with permission.) See
plate section for color version.

evidence was for a novel locus at 3p12.3–q12.3 (LOD ¼
3.88, “MDD-RE or anxiety”) and 18q21.33–q22.2
(LOD ¼ 3.75, “MDD-RE and anxiety”), an established
susceptibility locus for bipolar disorder.
We have also employed the phenotype of postpar-
tum mood symptoms in analyses that combined the
GenRED depression sample and the National Institute
of Mental Health (NIMH) Genetics Initiative bipolar
disorder sample [50]. We included women with a
history of pregnancy, any mood disorder, and infor-
mation about postpartum symptoms. There were 1210
women who met our criteria (30% of whom were
positive for a history of postpartum mood symptoms).
The maximum linkage peak for postpartum symptoms
occurred on chromosome 1q21.3–32.1, with a
chromosome-wide significant ZLR (an allele-sharing
score much like a LOD score) of 2.93 (permutation
p ¼ 0.02). This was a significant increase over the
baseline ZLR of 0.32 observed at this locus among
all women with a mood disorder (permutation p ¼
0.004). Suggestive linkage was also found on 9p24.3–
p22.3 (ZLR ¼ 2.91). Our results suggest that genetic
variations in these regions might increase susceptibil-
ity to postpartum mood symptoms.
Phenotypic subtypes might shed additional light
on linkage results. Comorbid anxiety and postpartum
onset have been studied, but a variety of others might
be relevant, including chronic course [51], attempted
suicide [52], and presence of alcoholism in a family
along with MDD [53].
Association
Association provides an alternative to the linkage
method. While linkage is a property of genes or loci
and occurs within families, association is a property
of alleles and occurs across a population. An associ-
ation study can be used for two different purposes.
The first is to test directly whether a genetic variant
might be implicated in MDD. An association between
a phenotype and an allele at a locus may mean that
the allele in question leads to susceptibility to the
phenotype. The second use of association studies is
to indirectly test the same thing through linkage dis-
equilibrium (LD) mapping. This indirect approach
can provide information about the location of a dis-
ease allele with resolution that is roughly 1000-fold
greater than that of a linkage study.
There are two types of candidate genes – functional
candidates and positional candidates. Functional can-
didate genes are those coding for a protein thought
to have some biological role in MDD. Positional

217
 Chapter 18: Genetics of major depression
5 Figure 18.2 Meta-analysis of association
GNB3
SLC6A3
2
Pooled odds ratio
MTHFR
1 wide association study data. (From [56],
0.5
APOE
0.2
0 1000 2000 3000 4000 5000
Total number of cases and controls
candidates are those identified through a linkage or
LD mapping study.
Functional candidates
Because of neurobiological data implicating the
neurotransmitters, particularly serotonin, norepin-
ephrine, and dopamine, in MDD, most early studies
focused on functional candidate genes from these
systems. While pathophysiological mechanisms of
MDD susceptibility are still not known, more recent
data suggest additional factors that include changes in
the hypothalamic-pituitary axis leading to dysregula-
tion of cortisol secretion, dysregulation of neuropro-
tective mechanisms, and inflammatory and immune
mechanisms.
The evidence that the monoamine neurotranmsit-
ters play a role in MDD first came from drug studies,
when reserpine, a blood pressure medication, was
found to cause depression [54]. It was later learned
that reserpine inhibits vesicular monoamine trans-
porters, thus depleting the brain of dopamine, nor-
epinephrine, and serotonin. These and other
observations led to the theory that depletion of cat-
echolamines, particularly norepinephrine, in brain
synapses, causes depression [55]. Later, attention
focused on depletion of serotonin, which led to the
development of the serotonin selective reuptake
inhibitors (SSRIs).
218
studies of functional candidate genes in
major depressive disorder (MDD). Pooled
odds ratios are presented in relation to
the total number of cases and controls
included in the meta-analyses. Odds
ratios are reported for the allelic analyses.
The five genes that were significant in
either this or genotypic analyses are
SLC6A4 highlighted. Note that this study was
done before the availability of genome-
with permission from Macmillan
Publishers Ltd.)
8000 10000 12000
A review of association studies [56] found that
of 102 functional candidate genes that had been stud-
ied for association in MDD, there were sufficient
data for meta-analysis on 18 of them, among which
11 involved the monoamine neurotransmitters:
COMT, DRD3, HTR1A, HTR1B, HTR2A, HTR2C,
MAOA, SLC6A2, SLC6A3 (DAT1), SLC6A4 (SERT),
and TPH1. Five of the 18 showed statistically signifi-
cant evidence of association (Figure 18.2). SLC6A3
encodes the dopamine transporter, and this gene has
a 40 bp variable repeat polymorphism, which was
investigated in three studies. The pooled odds ratio
for the “9/10” genotype compared to the 10/10 was
2.06. APOE was examined in seven studies, and the ε2
allele was found to be protective. Guanine nucleotide-
binding protein (GNB3) was tested in three studies
and the T allele of a C825T polymorphism was asso-
ciated with increased risk. The MTHFR C677T poly-
morphism was investigated in six studies and the
T allele was associated with increased risk. SLC6A4
encodes the serotonin transporter, the most heavily
studied gene in psychiatric genetics. The promoter
length polymorphism has received exceptional atten-
tion because it was shown to be functional, with the
short variant being associated with lower levels of
gene transcription [57]. Twenty-four studies with
3752 cases and 5707 controls have examined this
variant and found that the short allele carried a sig-
nificantly increased risk for MDD with an odds ratio

of 1.11, while the homozygous genotype increased
risk by 1.39. We discuss this gene further in the
context of the gene–environment interaction section
below. An important caveat in these results is that
among the five associated genes, for only one of them
(SLC6A4) was the number of MDD cases tested
> 1000. The small sample sizes raise questions about
the robustness of the results.
Functional candidate gene studies have also more
recently emerged from the neurotrophic model of
depression, a newer hypothesis about the nature of
depression [58]. The neurotrophic model stems from
research into the mechanism of action of antidepres-
sants, which has found that these medications stimulate
brain derived neurotrophic factor (BDNF) and also its
cousin neurotrophin 3 (NT3) [59] in the hippocampus,
and that injection of these two neurotrophins into the
hippocampus recapitulates the antidepressant effects
in the rat that the medications themselves induce.
BDNF treatment also reversed two animal paradigms
that have been used as models of depression, namely
learned helplessness following exposure to inescapable
shock and the forced swim test [60]. In humans, serum
BDNF levels were shown to be lower in depressed
subjects compared to controls [61, 62]. Some genetic
association studies have implicated BDNF variants in
MDD [63, 64], although a meta-analysis of 14 studies
found no overall association for the gene [65].
After having reported genome-wide significant
linkage on chromosome 15q25.3–26.2 to recurrent
early-onset MDD, as described above, we followed
up with linkage disequilibrium (LD) fine-mapping of
this signal and sequence analysis of NTRK3, which
encodes the receptor for NT3, given its biological
plausibility as a candidate gene [66]. LD mapping
showed nominally significant association in nine genes,
including NTRK3, with MDD-RE. In NTRK3, five SNPs
had nominally significant p-values. Sequence analysis
revealed 35 variants. While the number of rare variants
did not exceed chance expectation, case-control analysis
of 13 common variants did show modest nominal
association of MDD with 3 SNPs (p ¼ 0.008, 0.048,
and 0.034), which were in partial LD with 4 of 5
associated SNPs from the family based experiment.
We felt that while common variants in NTRK3 or one
of the other genes identified might play a role in
MDD, much larger studies would be required for full
evaluation of this region. Another report has found
an association of NTRK3 SNPs with childhood-onset
mood disorders [67].
Chapter 18: Genetics of major depression
A third area of biological interest in relation to
MDD that has generated candidate genes is the hypo-
thalamic-pituitary-adrenal (HPA) axis. The hypothal-
amus produces corticotrophin releasing hormone
(CRH) which stimulates the pituitary to produce
adrenocorticotropin hormone (ACTH), which in turn
stimulates the adrenals to produce cortisol, the stress
hormone. Neurotransmitter systems, including the
serotonergic system, influence the production of
CRH. Cortisol ultimately docks with the glucocorti-
coid receptor, which forms a complex that goes into
the nucleus of the cell where it binds to specific
DNA sequences and influences gene expression. The
first studies showing a relationship between elevated
cortisol and depression were done in the 1950s [68].
Variations in HPA axis genes may influence the stress
response and susceptibility to MDD. Haplotypes of
the glucocorticoid receptor gene have been associated
with MDD [69] as have variations in the CRH recep-
tor 1 gene, as a moderator of the relationship of child
abuse to MDD (see below) [70]. The FKBP5 protein
forms part of a complex with the glucocorticoid
receptor and can modulate cortisol-binding affinity.
Studies have found association of FKBP5 variation
with MDD [71], as well as the related phenotypes of
differential response to antidepressant drug treatment
[71, 72], number of depressive episodes [72], and
bipolar disorder [73]. Again, an important caveat here
is that there are negative reports for each of these
genes as well, and they have not been robustly repli-
cated in genome-wide association studies data.
Genome-wide association studies
Genome-wide association studies (GWAS) provide
a method for screening the entire genome at high
resolution in an unbiased way, in a search for
disease susceptibility variants. This is an extremely
powerful approach, though because tests of so
many DNA markers are being done, the statistical
bar is raised very high for considering findings true
positives. Because these studies perform roughly
1 million tests across the genome, the statistical
threshold for calling a finding genome-wide signifi-
cant is 5  10–2/1  10–6 ¼ 5  10–8. In addition,
because the effect size for any given variant is
expected to be quite low, the number of subjects
required for adequate power is very large, at least
in the thousands, and perhaps in the tens of
thousands.
219
 Table 18.3 Strongest findings in a meta-analysis of three MDD GWAS. Broad phenotype of all MDD.

Band SNP BP A1/ All GAIN-MDD GenRED STAR*D Meta-analysis Annotation

A2 Frq OR p OR p OR p OR p
8p21.3 rs1106634 20110329 A/G 0.12 1.27 3.93E-03 1.27 5.39E-03 1.35 8.18E-05 1.30 6.78E-07 ATP6V1B2; SLC18A1; LZTS1
7p15.3 rs17144465 21470952 A/G 0.04 1.44 4.82E-03 1.82 5.97E-06 1.32 1.47E-01 1.56 7.68E-07 SP4
3p26.1 rs9870680 7504555 C/T 0.43 1.23 1.23E-04 1.22 4.76E-04 1.10 6.89E-02 1.19 1.11E-06 GRM7
7q32.3 rs10265216 130550661 A/T 0.29 1.23 7.64E-05 1.14 3.12E-02 1.16 2.32E-02 1.19 3.02E-06 mRNA AK294384
3q26.32 rs644695 178774391 A/G 0.87 1.61 3.06E-07 1.26 3.20E-02 1.00 9.85E-01 1.35 4.46E-06
2p14 rs724568 67795984 A/C 0.36 1.20 2.58E-04 1.09 1.51E-01 1.19 1.68E-03 1.17 5.32E-06 mRNA BC043421
Xq21.33 rs5990417 94966526 T/C 0.82 1.20 2.90E-03 1.25 3.70E-03 1.27 5.27E-03 1.23 6.11E-06
6p25.1 rs2326810 6557466 C/G 0.92 1.21 7.29E-02 1.48 1.84E-03 1.75 1.07E-05 1.41 6.52E-06 LY86
10p11.23 rs1612122 29331904 A/T 0.48 1.15 4.51E-03 1.20 2.38E-03 1.18 4.34E-03 1.17 6.98E-06
Abbreviations: GWAS, genome-wide analysis studies; MDD, major depressive disorder; OR, odds ratio; SNP, single nucleotide polymorphism.
Modified from Shyn et al. [78].

The first MDD GWAS was published by Sullivan
et al. [74], in the context of the Genetic Association
Information Network, or GAIN, a collaborative effort
between the National Institutes of Health (NIH) and
several private companies. Using a sample from the
Netherlands, Sullivan conducted a GWAS study of
435 291 SNPs genotyped in 1738 MDD cases and
1802 controls selected to be at low liability for
MDD. The top signals were found in a 167 kb region
overlapping the gene piccolo (PCLO), whose protein
product localizes to the cytomatrix of the presynaptic
active zone and is important in monoaminergic
neurotransmission. P-values did not, however, reach
the threshold for genome-wide significance, as the
strongest one, rs2715148, registered at 7.7  10–7.
The investigators undertook replication of SNPs in
this region in 5 independent samples, totaling 6079
MDD cases and 5893 controls, but no SNP exceeded
the replication significance threshold when all repli-
cation samples were analyzed together.
The second published MDD GWAS report
involved two independent datasets, one German and
one Swiss [75]. The first experiment was performed
on 1022 recurrent MDD patients and 1000 controls
genotyped on the Illumina 550 platform. The second
was conducted on 492 recurrent MDD patients and
1052 controls selected from a population-based col-
lection, genotyped on the Affymetrix 5.0 platform.
Neither GWAS identified any SNP that achieved
genome-wide significance. A meta-analysis of the
two did not yield genome-wide significant results
either. The most significant p-value (5.8  10–6) from
the meta-analysis of the two studies was observed for
a SNP, rs4238010, on 12p13 located > 260 kb from the
closest known gene, cyclin D2 (CCND2). Forty-two
SNPs showed meta-analysis association p-values <
10  10–5.
In the GenRED study, we carried out a GWAS
using 1110 European-ancestry (EA) MDD cases, 655
recruited for our linkage study mentioned earlier,
and 455 recruited for association studies; 1020 cases
were available for analysis after quality control (QC).
The genotyping was performed on an Affymetrix
6.0 array. Of 2653 EA control subjects available from
another project, we selected 1636 after excluding
those who met criteria for lifetime MDD or who
reported recurrent MDD that missed the threshold
by one criterion. Unfortunately, no result met
genome-wide significance criteria. The strongest evi-
dence for association was observed on chromosome
Chapter 18: Genetics of major depression
18q22.1 (rs17077540, P ¼ 1.83  10–7) in a region
that has produced evidence for linkage to bipolar I or
II disorder, within an mRNA detected in human
brain tissue (BC896490) and approximately 75 kb
upstream of DSEL. Another signal was on chromo-
some 7p15.3 within SP4, a signal that comes almost
entirely from females. A member of the Sp family of
transcription factors, Sp4 is specific to neurons, forms
complexes with estrogen receptors that influence
regulation of many genes, and could play a role in
the mediation of neuroprotective enzymes and in
glutamate-induced neurotoxicity. Zhou et al., who
have implicated the gene in bipolar disorder [76],
have also reported that reduced expression of Sp4 in
mice leads to hippocampal vacuolization, age-
dependent reduced expression of neurotrophin 3
and deficits in sensorimotor gating and contextual
memory [77].
We then collaborated with Dr. Steven Hamilton at
the University of California San Francisco (UCSF) to
carry out a meta-analysis of three MDD GWAS
samples: GenRED; Dr. Hamilton’s genotypic data
for EA cases from the National Institute of Mental
Health (NIMH) -sponsored STAR*D antidepressant
effectiveness clinical trials program; and the Nether-
lands sample of Sullivan et al. described above [78].
The top results are summarized in Table 18.3 [78].
Even with this much increased sample size, now
totaling 3980 cases and 3428 controls, no signal
reached genome-wide significance. The strongest
evidence for association was observed for intronic
SNPs in ATP6V1B2 (p ¼ 5.69  10–7), GRM7 (p ¼
7.11  10–7), and, once again, SP4 (p ¼ 8.38  10–7).
The SP4 gene has been discussed above. ATP6V1B2
encodes a subunit for a vacuolar proton pump
ATPase. Our signal falls in a distinct LD block within
ATP6V1B2, but SLC18A1 (previously VMAT1), which
transports monoamines into synaptic vesicles, lies
very near ATP6V1B2, and could conceivably be impli-
cated by this signal.
GRM7 is particularly intriguing as it encodes
metabotropic glutamate receptor 7 (mGluR7), which
prior studies suggest may be involved in mood regu-
lation [79]. For example, an mGluR7 agonist had
antidepressant-like effects in mice that were blocked
by knockout of GRM7 [80]. This is the third GWAS to
report evidence of association to mood disorders in
this long gene (880 kb), the others being the German/
Swiss recurrent MDD GWAS mentioned above [75],
and the Wellcome Trust Case-Control Consortium
221
 Chapter 18: Genetics of major depression
GWAS of bipolar disorder [81], though the strongest
SNPs in those studies were in different parts of the
gene. Larger samples will be required to determine if
genome-wide significant evidence for association can
be detected for any variant in GRM7, but the bio-
logical evidence suggests that this locus merits further
investigation.
Prior functional candidate genes showed little
evidence of association in the meta-analysis. There
were 41 assessed, and as a group the distribution of
gene-wide p-values did not deviate significantly from
the null distribution. Among the genes mentioned
above, no SNPs in DRD4, APOE, BDNF, SLC6A3,
SLC6A4, MTHFR, or GNB3 were significant after
correction for the number of SNPs assayed across
the gene. This suggests that if variation in these genes
plays a role in MDD, the effect sizes would have to be
quite small.
A larger meta-analysis is ongoing through the
Psychiatric GWAS Consortium (PGC). It is hoped
that when the combined sample reaches into the
10 000–20 000 subjects range, sufficient power will
be available to definitively identify MDD-associated
genes and SNPs.
In sum, as of this writing, there are no associations
between specific DNA sequence variants and risk of
MDD that can be considered definitively established.
The gene-mapping approach could produce such
variants as the sample sizes continue to grow. Alterna-
tively, other approaches could bear fruit, such as
approaches that investigate epigenetics or gene expres-
sion, or those that involve study of gene function in cell
culture or animal models.
Gene expression and function
Studies of gene expression in MDD have been carried
out in brain samples and in white blood cell samples
from MDD patients, as well as in experimental cell
lines and in mouse models. Studies focused on the
brain face a number of important difficulties includ-
ing limited access to tissues and variable quality of
available samples.
Studies using white blood cells from patients have
the great advantage that these samples are much more
readily obtainable than brain tissue. Although studies
have indicated that functional abnormalities exist in
the lymphocytes of MDD patients, there have been
few efforts to examine gene expression. The major
weakness of this approach to studying MDD is that it
222
is not clear that expression in these white blood cells
mirrors expression in neurons, as control of expres-
sion may be tissue- and cell-type specific. On the
other hand, some abnormalities that have been
detected in lymphocytes appear to mirror changes
detected in postmortem brain tissue. For example,
Sullivan et al. tested for correlation across the tran-
scriptome of lymphocyte to brain region-specific
expression levels and found substantial correlation
with several regions including prefrontal cortex and
amygdala [82].
A recent review of expression studies in MDD
brain demonstrated replicated changes in levels of
15 genes, including decreased expression for genes
in glutamate transport (e.g. GLUL and SLC1A2/
EAAT2) and metabolism, neurotrophic signaling
(e.g. FGF, BDNF and VGF), and MAP kinase path-
ways (Table 18.4) [83]. The authors point out that a
powerful strategy for drug development is to deter-
mine whether genes identified in the brains of
affected patients are altered in a reciprocal way in
the brains of animals exposed to candidate thera-
peutic compounds.
Functional studies of the impact of an allele asso-
ciated with MDD can be performed using cell culture.
A copy of the gene containing the candidate allele can
be inserted (transfected) into cells in culture. These
cells can then be compared with control cells. Gene
expression in these cells can be measured, as can
relevant consequences of gene function, such as bind-
ing capacity for a receptor. For example, the serotonin
transporter promoter polymorphism was studied in
this way. Lymphoblast cell lines were transfected with
genes encoding the long variant and genes encoding
the short variant. The transcriptional activity of the
long variant was found to be twice that of the short
variant [57]. The observation that this variant is func-
tionally significant has helped make the serotonin
transporter promoter polymorphism the object of
intense study.
An even more powerful approach to the study of
gene function is to create mouse models for the gene
under study. A variety of approaches exist for genetic-
ally altering the mouse. The knockout approach allows
for the creation of a mouse lacking one or both copies
of the gene. Tissue-specific knockout methods allow
even more narrowly defined gene effects to be exam-
ined. In the conditional knockout, the mouse carries
the gene, but investigators can turn it off at will.
Transgenic approaches allow for the introduction of
 Chapter 18: Genetics of major depression

Table 18.4 Replicated gene expression changes in major depressive disorder (MDD) brain.

Gene symbol Gene name Expression change

GLUL Glutamate-ammonia ligase Decreased
SLC1A2/EAA T2 Astrocyte high-affinity glutamate transporter Decreased
AGXT2L1 Alanine-glyoxylate aminotransferase 2-like 1 Decreased
FGF1 Fibroblast growth factor protein 1 (acidic) Decreased
FGF2 Fibroblast growth factor protein 2 (basic) Decreased
FGFR1 Fibroblast growth factor receptor 1 Increased
FGFR2 Fibroblast growth factor receptor 2 Increased
FGFR3 Fibroblast growth factor receptor 3 Increased
NCAM1 Neural cell adhesion molecule Decreased
GPR37 G protein-coupled receptor 37 endothelin receptor type B-like Decreased
GPRC5B G protein-coupled receptor, family C, group 5, member B Decreased
DIMT1L DIM1 dimethyladenosine transferase 1-like Decreased
PRPF19 PRP19/PSO4 pre-mRNA processing factor 19 homolog Decreased
NTRK2 Tropomyosin receptor related kinase 2, TrkB Decreased
AQP4 Aquaporin4 Increased
Adapted from Altar et al. [83].

extra copies of normal genes or copies of altered
genes. The knock-in approach allows for the simul-
taneous knockout of the normal gene and transgenic
insertion of an altered version of the gene. With these
mouse models, gene effects in the brain and on behav-
ior can be studied.
The glucocorticoid receptor (GR) has been much
studied in mouse models of depression and anxiety.
Lines of mice generated include the following: those
with disrupted GR alleles; those with nervous system-
specific knockout of GR; transgenic mice with
increased GR expression; transgenic mice that express
an antisense RNA, which results in decreased GR
expression by binding to and neutralizing the normal
GR RNA; and knock-in mice, in which the normal
gene is knocked out while abnormal GR genes are
introduced and then expressed [84]. Behaviors
thought to model depression – such as performance
on the Porsolt forced swim test, the tail suspension
test, or the learned helplessness paradigm – can be
measured in these genetically altered mice as a way of
gauging the impact of the gene, and these behaviors
may have relevance for MDD. A particularly interest-
ing model is one in which the mouse is missing one
copy of the GR gene, the GR-heterozygous mutant
[85]. These mice exhibit normal baseline behaviors,
but demonstrate increased helplessness only after
stress exposure. Similar to depressed patients, these
mutant mice have a disinhibited HPA system and a
pathological dexamethasone/corticotropin releasing
hormone test. They further have downregulation of
BDNF protein in the hippocampus, consistent with
the neurotrophin hypothesis of depression.
Epigenetics
The modest level of heritability of MDD suggests that
the DNA sequence does not fully explain the variabil-
ity in susceptibility to this illness. There are at least
two other major kinds of explanations for this vari-
ation. One of these is that environmental factors such
as stressful life events play a significant role in trig-
gering MDD [86, 87], and another is that epigenetic
factors are involved. These may be interdependent
as the environment may cause epigenetic changes.
Epigenetic changes, such as those in DNA methyla-
tion (DNAm) and in histone modification, are often
correlated with changes in gene expression.

223
 Chapter 18: Genetics of major depression
In an animal model of early life stress, reduced
maternal care, epigenetic changes, including in DNAm
and in histone modification, were seen in the promoter
region of the glucocorticoid receptor gene, and
these persisted into adulthood, where they correlated
with disruption of the hypothalamic-pituitary-adrenal
(HPA) axis [88]. The same group of investigators
showed that in DNA from postmortem hippocampus
obtained from suicide victims with a history of child-
hood abuse, as compared to that from either suicide
victims with no childhood abuse or from controls,
there was increased DNAm of a glucocorticoid recep-
tor gene promoter [89].
There are several examples of epigenetic variation
in candidate MDD genes and in DNA treated with
medications used for MDD. For example, early life
adversity increased DNAm in Bdnf in rats [90].
Valproate [91], used to treat bipolar depression, and
haloperidol [92], used for psychotic depression, as
well as the antidepressants imipramine [93], tranylcy-
promine [94], and fluoxetine [95] have been shown to
induce epigenetic changes in rodent brain. Further,
administration of a histone deacetylase inhibitor,
sodium butyrate, produces an antidepressant effect
in an animal model [96].
Despite the availability of an essentially complete
genome sequence for several years, understanding the
methylome, the full complement of DNA methylation
across the genome, has progressed more slowly,
largely due to limitations in technology affecting
sensitivity, specificity, throughput, quantitation, and
cost among the previously used detection methods.
Microarray-based methods can interrogate much larger
numbers of CpGs, the sites of DNAm, than other
approaches. But no published study has yet assayed the
genome for DNAm in MDD, though several groups,
including our own, are pursuing this work.
Genes and environment
We have mentioned the connected hypotheses that
gene and environment might interact to cause MDD,
that stress might be the most relevant environmental
factor, and that epigenetics might be a mechanism
that mediates between stress and gene. Some have
hypothesized a continuum, with some depressions
emerging out of a high degree of stress, others occur-
ring in the setting of virtually no stress, and others in
between. Complicating this scenario are the likeli-
hood that the HPA axis, which mediates stress, is
224
itself influenced by genetic factors, and that MDD
itself constitutes a major stressor.
Several studies have found support for the inter-
action of genetic factors and stressful life events in
the MDD etiology. Kendler et al. [87], studying 2164
female twins, found 12 life event predictors of depres-
sion. The four strongest were death of a close relative,
assault, serious marital problems, and divorce/
breakup. In those at lowest genetic risk (identical
twin, co-twin unaffected with MDD), the probability
of MDD onset per month was 0.5% and 6.2%, for
those unexposed and exposed, respectively, to a severe
life event. In those at highest genetic risk (identical
twin, co-twin affected with MDD) the rates were 1.1%
and 14.6%. The 14.6% figure represents the interactive
risk of MDD in those with both genetic vulnerability
and exposure to high stress [87].
The most heavily studied interaction between a
specific genetic variant and an environmental factor
in MDD is that between the promoter length poly-
morphism of the serotonin transporter gene, known
as 5HTTLPR, and stressful life events. A study by
Caspi et al. [97], published by Science in 2003, found
that in a sample of 800 New Zealanders the short
allele of 5HTTLPR predicted depression only in the
setting of substantial life stressors. This study was
called one of the findings of the year across all of
science for 2003, and it has since been cited over
2600 times. There have been many attempts at repli-
cation, with studies reporting mixed results. In 2009, a
meta-analysis of 14 studies on the issue was published
in JAMA and it was disappointing. The study con-
firmed a three-fold increased risk for MDD in those
with a high rate of stressful life events, but it failed to
show any interaction between the 5HTTLPR and these
events [98]. However, a more recent meta-analysis that
was far more inclusive – 54 studies included compared
to 14 for the 2009 report – did find that the short allele
was associated with an increased risk of developing
depression under stress [99]. This is, of course, only
one of a great many potential genetic variants that
might be relevant for such interactions.
Another example of gene–environment interaction
comes from a key gene in the HPA axis, CRHR1, for
which variants have been reported to predispose to
depression through an interaction with childhood
abuse. This was originally reported in two samples
[70], and then replicated in a third, but not in a fourth
sample [100]. Additional work will be needed to clarify
these results.

Pharmacogenetics of MDD
Pharmacogenetics is a field that first developed in the
1950s with clinical observations of inherited differ-
ences in drug effects. In the early 1960s, a few studies
of familial correlation in response to antidepressants
were published. One such study found that of 41 pairs
of relatives treated with the antidepressant imipra-
mine, 38 had a concordant response: both responded
in 34 pairs, neither responded in 4 pairs, and 1
responded in 3 pairs (Angst, 1961 and 1964: cited in
[101]). A second study, by Pare et al. [102], found that
in 8 relative pairs, there was concordance for 6/6 pairs
of tricyclic antidepressant trials and for 6/6 pairs of
monoamine oxidase (MAO) inhibitor antidepressant
trials. Pare and Mack [101] later reported concord-
ance in 10/12 new pairs of related patients treated
with antidepressants from the same class, making
the total 22/24 (92%) for the Pare et al. [102] and
Pare and Mack studies [101]. By contrast, analysis of
relatives’ responses to antidepressants of different
classes from the two studies revealed concordance in
just 7/18 pairs (39%).
Many studies have examined antidepressant
response in relation to gene variants, with most of
them focusing on the same serotonin transporter gene
promoter region variant discussed above. A number
of small studies suggested a better response to select-
ive serotonin reuptake inhibitors (SSRIs) in patients
with long alleles and a slower or worse response in
subjects with short alleles. However, in what is by far
the largest study to date, investigators using the
STAR*D sample did not find such an association [103].
There have been two antidepressant GWAS
reports to date. The STAR*D investigators conducted
a GWAS to systematically assess association with
response to citalopram in a sample of 883 responders
and 608 nonresponders [104]. They employed SNPs
from the Affymetrix 500K and 5.0 Human SNP
Arrays, and association analysis was carried out after
correcting for population stratification. No results
reached genome-wide significance. They identified 3
SNPs associated with response with p-values < 1  10–5
near UBE3C (rs6966038, p ¼ 4.65  10–7), another 100 kb
from BMP7 (rs6127921, p ¼ 3.45  10–6), and a third
in RORA (rs809736, p ¼ 8.19  10–6). The Munich
Antidepressant Response Signature (MARS) project
reported a GWAS on 339 patients with a depressive
episode [105]. Replication was sought in a further 361
German subjects with depression and in 832 subjects
Chapter 18: Genetics of major depression
from the STAR*D sample. The strongest association
was found for rs6989467 in CDH17 (p ¼ 7.6  10–7),
while the strongest effect, by another measure, was
observed for rs1502174 in EPHB1 (p ¼ 8.5  10–5).
No effect withstood correction for multiple testing.
The investigators added two innovative analyses.
First, they defined a binary response allele score based
on 46 of the strongest SNPs from the MARS sample
that also showed nominal significance in the STAR*D
sample. The odds ratio for this response allele score
was 2.31 (p ¼ 5  10–8) in the MARS sample and 1.90
(p ¼ 5  10–9) in the STAR*D sample. They further
performed a pathway analysis, categorizing genes as
related if they were co-cited in the same sentence of
an abstract with a functional descriptor in between.
Gene clusters were identified in accordance with the
number of co-citations of each pair of genes. This
resulted in three clusters that centered on fibronectin 1,
ADAMTS-like 1, and endothelin 1 (Figure 18.3).
As with the genetics of MDD, the genetics of
antidepressant response is likely to involve large
numbers of genetic varations acting jointly to influ-
ence the phenotype. A simple tallying up of alleles
might one day be of value in a clinical setting, while
pathway types of analysis that can provide insight into
the biochemical processes at work in the response
may guide future studies of the physiology down-
stream from the relevant genes.
Conclusions
Although familial patterns of depressive illnesses have
been noted for over 100 years, a precise description
of their genetic mechanisms remains an elusive goal.
We do not yet have a single defintively identified
MDD susceptibility gene. However, the field is now
advancing rapidly and much progress should be
expected within the next couple of years. The Psychi-
atric GWAS Consortium is providing an infrastruc-
ture in which increasingly large sample sizes are being
collected. As the sample size grows into the range of
10 000–20 000 cases and a comparable number of
controls, we will acquire the power to detect the
modest gene effects likely to play a role in suscepti-
bility to MDD. Large samples also provide sufficient
power to employ analytic techniques that can robustly
test for interactions between genes.
While GWAS focuses on identifying common
variation associated with MDD, studies searching
for rare variants involved in MDD etiology are just
225
 Chapter 18: Genetics of major depression

M IN TF NR2F2 Cluster 2
IN PIK3R1
IN CUGBP2

IN APOB

M IN PON2

IN CD36
M IN GSTP1 M IN EPHX2

IN SPON1

IN ADAMTSL1 M IN RYR2 IN HOMER1

Cluster 1
IN SLC8A1 M IN KCNJ6
IN HAS2

IN EDN1 M IN TNFRSF1A IN DAPK1

M IN PTPRK M IN PTGER4
M IN SLC1A2
IN TF MYBL2
IN A2BP1 M IN AKR1B1 IN TNFRSP21
IN SLC6A11
IN TF MITF
IN NRG1
IN PTPRS M IN FN1
IN SCNN1A
M IN TF NR2E1 IN CADM1
M IN RPS6KA1 Cluster 3
M IN RAP1B

M IN TGFBR3 IN TPM1
IN CALD1

IN SERPINA4
IN EFNAS

IN EPHAS

Figure 18.3 Results of a literature-based pathway analysis that includes all genes from the Munich Antidepressant Response Signature
project genome-wide association study (GWAS) that correspond to the single nucleotide polymorphisms (SNPs) implicated by the STAR*D
replication GWAS sample. Genes were categorized as related when they were co-cited in the same sentence with a functional descriptor
in between. There are 41 genes that cluster around fibronectin 1 (FN1) (cluster 1), ADAMTS-like 1 (ADAMTSL1) (cluster 2), and endothelin 1
(EDN1) (cluster 3). Genes with corresponding SNPs that achieved nominal significant replication in the STAR*D sample are shaded in red; green
lines indicate transcription factor (TF) binding site matches in target promoters; the line with the yellow circle indicates annotation by
Molecular Connections experts. IN indicates input gene; M, part of a metabolic pathway. From [105], with permission. See plate section for
color version.

beginning, making use of powerful next-generation
sequencing technology. Such studies can be targeted
to particular linkage regions or particular gene sets.
As the cost of performing such studies is coming
down rapidly, studying the whole exome, or even
the whole genome, in this way will soon be feasible.
Findings of statistical association between a gene
variant and MDD will need to be complemented by
functional studies of the relevant gene and its poten-
tially pathogenic alleles. For example, postmortem
brain samples from MDD subjects are a valuable
resource for testing gene expression levels. Brain
tissue is, of course, not available on subjects who are
studied for association, but white blood cells are.
Although these cells express only about one-half of
all genes, and potential tissue-specific regulation of
gene expression may limit their usefulness, the exist-
ence of such cell lines in the very MDD subjects
who show association can provide investigators with
a useful means of studying the expression of some
candidate genes.
Epigenetic studies should advance our under-
standing of the mechanisms that underlie gene
expression. As these epigenetic marks are likely to
be partly inherited and partly acquired, they might
tell us something about the so-called “missing herit-
ability” in MDD, and also something about the bio-
logical impact of stressful life events. Genome-wide
studies of DNA methylation variation and of histone
modifications in MDD are ongoing.

226
 Chapter 18: Genetics of major depression

Approaches that emphasize clinical insights into
the depressive phenotype may provide increased
power by creating more genetic homogeneity. Clinical
subtypes such as postpartum depression and comor-
bid anxiety, as well as potential endophenotypes
including variations in HPA function and brain
region volumes in patients with MDD and their rela-
tives may allow for useful alternative methods for
identifying genetic effects. It may be possible to study
genotype–endophenotype correlations. If a gene is
involved in MDD, it should be possible to demon-
strate the effect the disease allele has on, for example,
HPA abnormalities and reduced hippocampal size.
Given the likelihood of relatively modest effects for
MDD susceptibility alleles, a conclusive case for the
involvement of an implicated gene may require dem-
onstration of a pathogenic pathway.
There are currently no laboratory methods for
diagnosing MDD; rather, these diagnoses rest solely
on clinical data. Identification of a set of causative
variants could clarify the diagnostic process by provid-
ing physical evidence of the disorder. Presymptomatic
testing could lead to preventive treatments. Greater
precision in prognosis may come from genotype–
phenotype correlations, whereby particular symptom
clusters and natural course are found to be associated
with specific combinations of gene variants.
Pharmacogenetic studies could also provide excit-
ing clinical benefits from genetic investigations of
MDD. If the illness is genetically heterogenous, there
may be some genetic vulnerabilities associated with
good SSRI response, and other genetic vulnerabilities
associated with a better response to noradrenergic
medications such as nortriptyline. These findings
could allow clinicians to optimize the use of already
effective medications by choosing the drug most likely
to work for a given patient with a particular genetic
profile. The work of establishing these predictors of
response will benefit from the existence of large net-
works that can rapidly implement pharmacogenetic
trials of existing and newly emerging antidepressants.
The recently created National Network of Depression
Centers, involving 21 US academic medical centers, is
one such example.
Perhaps the most fundamental benefit will come
from an improved understanding of the pathophy-
siology of MDD. Finding a gene will lead to an exam-
ination of the functions of its protein product in the
neuron. This examination could lead in turn to the
elucidation of a cascade of neuronal events at work in
the disorder. Understanding of these basic processes
would guide the search for novel treatments of MDD.
The gene product or another protein with which the
gene product interacts could be a target for conven-
tional pharmacology, such as receptor blockade or
inactivation of an enzyme. Alternatively, an overex-
pressed gene could be knocked down with siRNA.
Work in this area is advancing rapidly [106].
The next decade should be an exciting one for
the genetics of MDD as the pieces are all in place
for making enormous progress. This will benefit
researchers who should soon have definitive genetic
variants involved in MDD that can provide anchor
points for further work elucidating the role of these
variants in disease. More importantly, it should pro-
vide a starting point for discerning ways in which to
make these results relevant for improving the care of
patients suffering with this dreaded disorder.

References 7. Tsuang MT, et al. Br J Psychiatry

1. Kraepelin E. Manic Depressive
Insanity and Paranoia. Edinburgh:
Livingstone; 1899/1921.
2. Perris C. Acta Psychiatr Scand
Suppl 1966;194:15–44.
3. Reich T, et al. Am J Psychiatry
1969;125:1358–1369.
4. Andreasen NC, et al. Arch
Gen Psychiatry 1987;44:
461–469.
5. Winokur G, et al. Arch Gen
Psychiatry 1995;52:367–373.
6. Smeraldi E, et al. Acta Psychiatr
Scand 1977;56:382–398.
1980;137:497–504.
8. Abrams R, et al. Am J Psychiatry
1980;137:658–661.
9. Jakimow-Venulet B. Br
J Psychiatry 1981;139:450–456.
10. Gershon ES, et al. Arch
Gen Psychiatry 1982;39:
1157–1167.
11. Baron M, et al. Acta Psychiatr
Scand 1982;65:253–262.
12. Weissman MM, et al. Arch Gen
Psychiatry 1984;41:13–21.
13. McGuffin P, et al. J Psychiatr Res
1987;21:365–375.
14. Maier W, et al. Arch Gen
Psychiatry 1993;50:871–883.
15. Weissman MM, et al. Arch
Gen Psychiatry 1993;50:
767–80.
16. Sullivan PF, et al. Am J Psychiatry
2000;157:1552–1562.
17. Lyons MJ, et al. Arch Gen
Psychiatry 1998;55:468–472.
18. Bland RC, et al. Arch Gen
Psychiatry 1986;43:1085–1089.
19. Levinson DF. Biol Psychiatry
2006;60:84–92.
20. Bertelsen A, et al. Br J Psychiatry
1977;130:330–351.

227
 Chapter 18: Genetics of major depression
21. Torgersen S. Arch Gen Psychiatry
1986;43:222–226.
22. Andrews G, et al. J Affect Disord
1990;19:23–29.
23. McGuffin P, et al. Arch Gen
Psychiatry 1996;53:129–136.
24. Bierut LJ, et al. Arch Gen
Psychiatry 1999;56:557–563.
25. Kendler KS, et al. Arch Gen
Psychiatry 2000;57:94–5.
26. Kendler KS, et al. Am J Psychiatry
2006;163:109–114.
27. Mill J, et al. Mol Psychiatry
2007;12:799–814.
28. Kaminsky ZA, et al. Nat Genet
2009;41:240–245.
29. Falconer DS. Ann Hum Genet
1965;29:51–76.
30. Wender PH, et al. Arch Gen
Psychiatry 1986;43:923–929.
31. von Knorring AL, et al. Arch Gen
Psychiatry 1983;40:943–950.
32. Cadoret RJ, et al. J Affect Disord
1985;9:155–164.
33. Tanna VL, et al. Neuro-
psychobiology 1976;2:52–62.
34. Tanna VL, et al. Neuropsychobiology
1979;5:102–113.
35. Tanna VL, et al. J Psychiatr Res
1989;23:99–107.
36. Wilson AF, et al. Biol Psychiatry
1989;26:163–175.
37. Crowe RR, et al. Neuro-
psychobiology 1981;7:20–25.
38. Hill EM, et al. Biol Psychiatry
1988;24:903–917.
39. Cox N, et al. J Psychiatr Res
1989;23:109–123.
40. Zubenko GS, et al. Mol Psychiatry
2002;7:460–467.
41. Nibuya M, et al. J Neurosci
1996;16:2365–2372.
42. Dowlatshahi D, et al. Lancet
1998;352:1754–1755.
43. Dwivedi Y, et al. Arch Gen
Psychiatry 2003;60:273–282.
44. Abkevich V, et al. Am J Hum
Genet 2003;73:1271–1281.
228
45. McGuffin P, et al. Hum Mol Genet
2005;14:3337–3345.
46. Holmans P, et al. Am J Hum Genet
2004;74:1154–1167.
47. Holmans P, et al. Am J Psychiatry
2007;164:248–258.
48. Levinson DF, et al. Am
J Psychiatry 2007;164:259–264.
49. Camp NJ, et al. Am J Med Genet B
Neuropsychiatr Genet 2005;
135B:85–93.
50. Mahon PB, et al. Am J Psychiatry
2009;166:1229–1237.
51. Mondimore FM, et al. Am
J Psychiatry 2006;163:1554–1560.
52. Zubenko GS, et al. Am J Med
Genet B Neuropsychiatr Genet
2004;129B:47–54.
53. Nurnberger JI, Jr., et al. Am
J Psychiatry 2001;158:718–724.
54. Freis ED. N Engl J Med 1954;
251:1006–1008.
55. Schildkraut JJ. Am J Psychiatry
1965;122:509–522.
56. Lopez-Leon S, et al. Mol
Psychiatry 2008;13:722–785.
57. Lesch KP, et al. Science
1996;274:1527–1531.
58. Duman RS, et al. Arch Gen
Psychiatry 1997;54:597–606.
59. Smith MA, et al. Proc Natl Acad
Sci U S A 1995;92:8788–8792.
60. Siuciak JA, et al. Pharmacol
Biochem Behav 1997;56:131–137.
61. Karege F, et al. Psychiatry Res
2002;109:143–148.
62. Brunoni AR, et al. Int J
Neuropsychopharmacol
2008;11:1169–1180.
63. Schumacher J, et al. Biol
Psychiatry 2005;58:307–314.
64. Licinio J, et al. Arch Gen
Psychiatry 2009;66:488–497.
65. Verhagen M, et al. Mol Psychiatry
2010;15:260–271.
66. Verma R, et al. Biol Psychiatry
2008;63:1185–1189.
67. Feng Y, et al. Am J Psychiatry
2008;165:610–616.
68. Board F, et al. Psychosom Med
1956;18:324–333.
69. van West D, et al. Neuro-
psychopharmacology 2006;
31:620–627.
70. Bradley RG, et al. Arch Gen
Psychiatry 2008;65:190–200.
71. Lekman M, et al. Biol Psychiatry
2008;63:1103–1110.
72. Binder EB, et al. Nat Genet
2004;36:1319–1325.
73. Willour VL, et al. Mol Psychiatry
2009;14:261–268.
74. Sullivan PF, et al. Mol Psychiatry
2009;14:359–375.
75. Muglia P, et al. Mol Psychiatry
2010;15:589–601.
76. Zhou X, et al. PLoS One 2009;
4:e5196.
77. Zhou X, et al. Mol Psychiatry
2005;10:393–406.
78. Shyn SI, et al. Mol Psychiatry
2011;16:202–15.
79. Pilc A, et al. Biochem Pharmacol
2008;75:997–1006.
80. Palucha A, et al. Psycho-
pharmacology (Berl) 2007;
194:555–562.
81. Wellcome Trust Case Control
Consortium. Nature 2007;
447:661–678.
82. Sullivan PF, et al. Am J Med Genet
B Neuropsychiatr Genet 2006;
141B:261–268.
83. Altar CA, et al. Neuro-
psychopharmacology 2009;
34:18–54.
84. Chourbaji S, et al. Prog Brain Res
2008;167:65–77.
85. Ridder S, et al. J Neurosci 2005;
25:6243–6250.
86. Paykel ES, et al. Arch Gen
Psychiatry 1969;21:753–760.
87. Kendler KS, et al. Am J Psychiatry
1995;152:833–842.
88. Weaver IC, et al. Nat Neurosci
2004;7:847–54.
89. McGowan PO, et al. Nat Neurosci
2009;12:342–348.

90. Roth TL, et al. Biol Psychiatry
2009;65:760–769.
91. Milutinovic S, et al. Carcinogenesis
2007;28:560–571.
92. Shimabukuro M, et al. Behav
Brain Funct 2006;2:37.
93. Tsankova NM, et al. Nat Neurosci
2006;9:519–525.
94. Lee MG, et al. Chem Biol
2006;13:563–567.
95. Cassel S, et al. Mol Pharmacol
2006;70:487–492.
Chapter 18: Genetics of major depression
96. Schroeder FA, et al.
Biol Psychiatry 2007;
62:55–64.
97. Caspi A, et al. Science
2003;301:386–389.
98. Risch N, et al. JAMA 2009;
301:2462–2471.
99. Karg K, et al. Arch Gen Psychiatry
2011;68(5):444–454.
100. Polanczyk G, et al. Arch
Gen Psychiatry 2009;
66:978–985.
101. Pare CM, et al. J Med Genet
1971;8:306–309.
102. Pare CM, et al. Lancet 1962;
2:1340–1343.
103. Kraft JB, et al. Biol Psychiatry
2007;61:734–742.
104. Garriock HA, et al. Biol Psychiatry
2010;67:133–138.
105. Ising M, et al. Arch Gen Psychiatry
2009;66:966–975.
106. Whitehead KA, et al. Nat Rev
Drug Discov 2009;8:129–138.
229
 The genetics of schizophrenia
Chapter
19 Hugh M. D. Gurling and Andrew McQuillin
Summary
Schizophrenia is a common disorder which has a high
heritability. Incomplete penetrance of schizophrenia
susceptibility alleles has frequently been observed.
Therefore there are likely to be many unaffected
carriers of these alleles in the population. Tests of
genetic linkage and allelic association have proven
that schizophrenia is extremely heterogeneous with
many different low frequency disease alleles causing
susceptibility. Recent genome-wide allelic association
studies have further demonstrated heterogeneity even
when comparing sub-samples within a single ances-
tral population. In addition it has been shown that a
proportion of schizophrenics have a genetic suscepti-
bility from deletion and insertion mutations, called
copy number variants (CNVs), some of which are
de novo and some of which are genetically transmit-
ted. Because of the presence of extreme heterogeneity
it will only be possible to determine which modes of
genetic transmission exist for schizophrenia by iden-
tifying the actual etiological base pair changes in
genes implicated by linkage and association studies
or by defining CNVs. The first few plausible etio-
logical base pair changes affecting the DISC1, PCM1,
ABCA13, and SYNGR1 genes causing schizophrenia
have now been reported. Deletions and duplications
are responsible for a proportion of the genetic suscep-
tibility to schizophrenia and there is striking overlap
for some of these mutations with those implicated in
epilepsy.
Pharmacogenomic strategies for the design of new
drug treatments have now become of great import-
ance because of the identification of susceptibility
genes. New drugs can be designed by targeting the
pathogenetic systems with the directionality of drug
effects determined by how the various brain systems
Principles of Psychiatric Genetics, eds John I. Nurnberger, Jr. and Wade H. Berrettini. Published by Cambridge University
Press. © Cambridge University Press 2012.
230
have been perturbed by the disease mutations. Trans-
genic mice with behavioral phenotypes that mimic
aspects of schizophrenia have been created for the
calcineurin (PPP3CC), proline dehydrogenase 2
(PRODH2), disrupted in schizophrenia (DISC1),
microtubule-associated stable tubule only polypeptide
(STOP), and neuregulin 1 (NRG1) genes. Cell biology
studies to explore pathogenesis have begun for NRG1,
DISC1, and PCM1 using human cell lines and zebra-
fish. Hundreds of other genes and their proteins will
need to be studied using these methods in order to
achieve a substantial understanding of the genetic
etiology of schizophrenia.
Introduction
Traditional approaches used in medical genetics, such
as population prevalence, family, twin, and adoption
studies have shown a very high heritability for schizo-
phrenia with very little or absent role for the family
environment. Studies of sex distribution, age of onset,
and clinical variation have been incorporated into
the family and twin studies. One of these, known
as “anticipation” is where the age of onset becomes
earlier and the severity of symptoms more severe in
the later generations of multiply affected families.
We now know that the previous attempts using family
data and segregation analyses to establish the modes
of transmission for schizophrenia were flawed
because of the underlying heterogeneity. Nevertheless
these older studies are important because they dem-
onstrated the great degree of genetic pleiotropism or
variation in phenotypic expression that can arise from
the same genetic susceptibility within families and
within identical twins. There is still controversy about
whether bipolar disorder and schizophrenia can share
the same susceptibility alleles and to what extent they

are different disorders. Clarity on this issue has been
hampered by the fact that diagnosis of schizophrenia
in early-onset cases is often inexact due to the fact that
severe bipolar mania can present with schizophrenic
symptoms.
Linkage studies have dominated efforts to find
the genes causing schizophrenia for the last 20 years.
The logarithm of odds (LOD) method approach
requires multiply affected families and the genotyping
of DNA markers in all the affected and unaffected
members of families. The sib-pair method does not
require the genotyping of unaffected members. The
LOD method of localizing genes by linkage analysis
in families can narrow down any susceptibility genes
to one of several hundred or several thousand in a
chromosomal region of up to 40 million base pairs
of DNA. The LOD score is a statistical test of linkage
between schizophrenia to a specific chromosomal
region as defined by genetic markers within a sample
of families. A LOD of 3.00 (p ¼ 0.001) in two inde-
pendent samples has been used as the criterion
for acceptance of a confirmed linkage. This approach
was slow to get off the ground in schizophrenia
research because early linkages took many years to
confirm because of the presence of disease locus
heterogeneity. However, these studies eventually
became very successful with many confirmed link-
ages with two LOD scores above 3.00 emerging after
sufficient numbers of family samples had been
studied.
The next phase of identifying susceptibility genes
employed methods to detect the evolutionary phe-
nomenon of linkage disequilibrium (LD) in unrelated
cases of schizophrenia and normal unrelated controls
rather than cases within families. Linkage disequilib-
rium is a generalized finding across the whole human
genome and means that genetic variants or mutations
that are close to each other tend to be inherited by the
same individual over evolutionary time. In practice
this means that specific marker alleles are found in
increased frequency in cases compared to controls.
This phenomenon is called allelic association and is
caused LD. Linkage disequilibrium implies that during
evolution, when a new susceptibility mutation arose,
it would have occurred in an individual who already
had pre-existing genetic variants near the mutation.
From then on as the disease mutation is passed down
the generations it will “hitch hike” or co-segregate
with other genetic variants nearby. Advances in
genetic technology now mean that genetic linkage
Chapter 19: Genetics of schizophrenia
and association studies can be carried out at much
greater speed and schizophrenia has benefitted from
the new microarray genotyping, CNV and sequencing
methods. The traditional medical genetic approaches
of family and twin studies will need to be carried out
again once the actual etiological DNA base pair
changes have been detected. Below, the various trad-
itional and more recent molecular approaches are
described in greater detail.
Age of onset, anticipation, and
epigenetic effects in schizophrenia
Age of onset variation is known to be an important
consequence of underlying genetic pathology. It
appears to be a variable which is well defined, but
age of onset is very prone to selection biases when
comparing cases and valid conclusions may be elusive
[1, 2]. Correlations in age at onset between relatives
can be used to predict risk to relatives at a given age.
Kendler and MacLean [3] corrected for various biases
and found that ages at onset for schizophrenia were
positively correlated in pairs of affected relatives.
An early age at onset was associated with higher risk
of illness in siblings and nieces/nephews but not in
children. Neale et al. [4] studied age of onset in a
schizophrenia twin sample. They found that the age
of onset was correlated within twin pairs. In the
Roscommon schizophrenia family study, age of onset
was earlier in hebephrenic and catatonic subtypes
compared to the later onset paranoid schizophrenic
subtype [3].
The relationship between age of onset in parental
and offspring generations and genetic effects has been
a focus in schizophrenia research for many years.
Penrose found what is referred to as “anticipation”
in schizophrenia by collecting age of onset data from
a large family sample [5]. He found strong evidence
for reduced age of onset in the next generation in
parent–offspring pairs who had both developed
schizophrenia. He also found increasing severity of
the illness in successive generations. Penrose was not
entirely convinced that he had a true finding. How-
ever parent–offspring pairs and aunt/uncle–niece/
nephew schizophrenia pairs both showed strong
evidence for anticipation suggesting that selection
biases for types of families could not easily account
for anticipation. Anticipation effects were greatest in
pairs with parents who had a late age of onset but
anticipation was also significant in early age of onset
231
 Chapter 19: Genetics of schizophrenia
parents as well (p ¼ 0.03). A review of Penrose’s study
[6] concluded that it was a valid finding. The findings
supported further investigations of anticipation
and trinucleotide repeat expansion mutations as well
as investigation of methylation abnormalities in the
genetics of schizophrenia.
Bassett and Honer [7] studied three-generation
affected families selected for unilineal, autosomal
dominant-like inheritance. They found strong evi-
dence that more subjects were hospitalized with
psychotic illness in the youngest generation, that the
age of onset became progressively earlier in the later
generations, and that there was increasing severity
of illness. In the study by Asherson et al. [8] of 29
multiply affected pedigrees, an earlier age of onset in
later generations was found but all selection biases
could not be excluded. Other studies in multiple
ancestral populations also found anticipation [9–19].
McInnis [20] reviewed the field and concluded that it
was a genuine effect. The main uncertainty in the
study of anticipation is that age of onset for schizo-
phrenia may have been getting earlier in the later
decades [12] purely as an artifact of early intervention
and increasing awareness of mental illness. However,
Penrose’s [12] study was carried out in 1945 and
anticipation has been found in a variety of different
cultures.
Familial recurrence of schizophrenia
Behavior genetic approaches have been used to obtain
evidence for the relative importance of cultural,
family environment, and unique (specific) environ-
mental factors compared to the effects of genes in the
etiology of schizophrenia. Several complementary
behavior genetic approaches such as the twin, family,
and adoption methods need to be compared and
contrasted to come to a balanced view. Genetic ana-
lyses which are relevant to plant and animal studies,
rather than human populations, have specific limita-
tions in relation to the genetics of schizophrenia
[21, 22]. The most often cited limitation is the fact
that humans do not mate randomly with each other
and therefore there is no stabilized equilibrium in the
general population for the distribution of human
traits and their genotypes.
Valid determination of the recurrence risk for
psychiatric disorders in the relatives of probands is
dependent on accurate diagnosis. An important find-
ing is that the diagnosis of schizophrenia is more
232
stable within individuals than is bipolar disorder
which is often misdiagnosed as schizophrenia in
early-onset cases particularly in young males [23].
Tsuang [24] found that 92.5% of schizophrenics had
a stable diagnosis over a 30–40 year period whereas
only 78.3% of bipolar cases had stable diagnoses. This
will mean that more cases of bipolar disorder are
likely to be incorrectly identified as cases of schizo-
phrenia than vice versa. All family, twin, and adop-
tion studies have observed that the incidence of the
illness in the relatives of schizophrenic probands is
considerably greater than in suitable family or popu-
lation controls [25–34].
Given that we now know that schizophrenia is
highly heterogeneous an important clinical question
is to ask to what extent different subtypes of schizo-
phrenia can recur within the same families. Family
studies have shown that, within a family, the same
disease alleles can give rise to clinical diversity or
pleiotropism within the schizophrenia spectrum [31,
35]. The corollary to this is that clinical variability
does not imply underlying genetic heterogeneity
within an individual. Although there is little evidence
for the proband schizophrenia subtype to breed true
[36, 37] other work has suggested a tendency towards
homotypia within families for the paranoid and hebe-
phrenic subtypes [38, 39]. Some families have been
reported with multiple cases of catatonic schizo-
phrenia [40]. Others have put forward the idea that
the hebephrenic form of schizophrenia may be more
heritable than the paranoid variety [41, 42]. Tsuang
and Winokur [38] have proposed that there is a
catatonic–hebephrenic–paranoid continuum of illness
with the catatonic type being the most advanced and
severe form.
Family and twin studies[32, 43] show that the risk
for schizophrenia and schizophrenia spectrum dis-
orders was not associated with any particular subtype
of proband. Bassett and coworkers [44] have tested
Liddle and Barnes’s three syndrome schizophrenia
model [45] in families with schizophrenia employing
factor analysis. They were able to confirm the “psycho-
motor poverty” factor but found a combined “reality
distortion/disorganization/inappropriate affect” factor
and a “suspiciousness/stereotyped thinking” factor
instead of Liddle and Barnes’s separate reality distor-
tion and disorganization factors.
Keefe et al. [46] who carried out a family study,
compared severely disabled Kraepelinian type schizo-
phrenics with other chronic, but intermittently

exacerbating, schizophrenics. They found a higher
incidence of schizophrenia spectrum disorders in the
relatives of the Kraepelinian group (19.3%) compared
to the relatives of the intermittently exacerbating
group (10.7%). The Kraepelinian group was also char-
acterized by asymmetry of the lateral ventricles and
a lack of response to neuroleptics.
Potential differences between familial and spor-
adic schizophrenia were reviewed in 1994 [47]. In
69 studies very few replicated differences were found.
One difference was the increased volume of the len-
ticular nuclei and greater ventricular asymmetry in
familial cases compared to sporadic cases and normal
control subjects [48].
Schizophrenia and affective disorders
Slater and Cowie [36] drew attention to the fact that
family studies of schizophrenia were associated with
what they called attenuated cases of schizophrenia
that appeared to be cases of neurotic depression. This
idea was later supported by a family study which
found an increase in depression but not bipolar dis-
order in the relatives of schizophrenics [49]. Kendler
and others found that the risk in all interviewed
relatives of schizophrenic probands was 23.6% for
all types of severe psychiatric disorder [33, 50–53].
The risk for schizophrenia in the parents of probands
was much less than that found in siblings indicating
partial or incomplete penetrance. These studies also
showed that the schizophrenia found in these families
did not have any strong familial relationship with
bipolar affective disorder.
Some reviewers of family studies of schizophrenia
have drawn the conclusion that schizophrenia and
bipolar disorder do not share the same genetic etiol-
ogies [54] whereas others argue that they often do
[55]. More recently in a meta-analysis of family stud-
ies it was found that the increased rate of schizophre-
nia in the first-degree relatives (FDRs) of bipolar
subjects was “somewhat equivocal” being marginally
significant in the primary analysis but not significant
in the unweighted analysis [56]. When the rate of
bipolar disorder in the FDRs of schizophrenic pro-
bands was examined the increased rate was not signifi-
cant (p ¼ 0.06) in both the weighted and unweighted
analyses. Only when studies that did not report
morbid risk data were excluded from the analyses
was it found that the risk of schizophrenia in the
first-degree relatives of bipolar probands was
Chapter 19: Genetics of schizophrenia
increased with a significance of p ¼ 0.02. These data
are not strongly in favor of coaggregation of the two
disorders in the same families and the weak evidence
is explainable by mis-diagnosis and by assortative
mating for schizophrenia and bipolar disorder in the
parental generation.
A very large population register study of the
Swedish population entitled “Common genetic deter-
minants of schizophrenia and bipolar disorder in
Swedish families: a population-based study” claimed
common genetic effects for schizophrenia and bipolar
disorder [57]. Both diseases were diagnosed with a
nonhierarchical diagnostic structure. Thus, an indi-
vidual could have both schizophrenia and bipolar
disorder if the person has been diagnosed with each
disease. There were 2543 such dual diagnosis cases
accounting for 5086 (6.6%) cases out of 76 860 pro-
bands. The chance of an individual having both diag-
noses according to the population prevalence of both
these disorders in Sweden without adjustment for age
is 0.0001%. Thus, in this study the expected rate of
dual diagnosis is 66 000 times greater than chance.
The procedure of counting dual diagnosis probands
twice would inflate the cross correlation between
bipolar disorder and schizophrenia within families.
The authors also did model-fitting analyses with hier-
archical diagnoses, in which individuals with two or
more schizophrenia diagnoses were classified as
having schizophrenia and not bipolar disorder, even
if they have had two or more such diagnoses. How-
ever this does not get rid of the problem of misdiag-
nosis because such cases could equally have been
counted as bipolar probands. It is difficult to explain
the differences between the Swedish study [57] with
the recent family meta-analysis [56] without consider-
ing the possibility that the Swedish registry diagnoses,
which were essentially clinical rather than operation-
ally defined research diagnoses, were prone to
unstable diagnosis between the two psychoses [56].
Many other differences between schizophrenia and
bipolar disorder have been found such as voxel-based
morphometry in schizophrenia compared to bipolar
disorder which found consistent differences with
many bipolar studies showing none of the volumetric
deficits found in schizophrenia [58]. These studies
support the notion that schizophrenia and bipolar
disorder generally have different genetic underpin-
nings. However, there are cases of schizoaffective
disorder with clinical features that do cross boundar-
ies between schizophrenia and bipolar disorder but
233
 Chapter 19: Genetics of schizophrenia
these cases are rare [59–62]. Maj and coworkers [61]
stated that DSM-III-R schizoaffective disorder was
close to schizophrenia and largely corresponds to
“schizoaffective schizophrenic” disorder in the
Research Diagnostic Criteria (RDC) [63], whereas
DSM-III-R mood-incongruent psychotic depression
is heterogeneous and corresponds to what could be
called a true “schizoaffective psychosis” state which is
quite rare. The implication is that the RDC system of
diagnosis using the Schizophrenia and Affective Dis-
orders Schedule (SADS) [63] is superior to DSM-III-
R and DSM-IV because it subdivides schizoaffective
disorders into two types. One is schizoaffective bipo-
lar disorder or schizo-mania which is basically a bipo-
lar manic illness with a schizophrenic coloration
during acute mania and the other is schizoaffective
schizophrenia which is a chronic illness more related
to schizophrenia. This concept gains support from
two familial recurrence studies of the families of
“schizoaffective” psychosis probands. One study
carried out in New York found that schizoaffective
psychosis could be blindly subdivided into either
schizoaffective mania (schizomania or schizoaffective
bipolar disorder, SA-BP) or schizoaffective schizo-
phrenia (SA-SCZ) and that the presence of schizo-
affective bipolar disorder in the proband predicted
the presence of bipolar disorder in the first-degree
relatives whereas schizoaffective schizophrenia in the
proband predicted schizophrenia in the relatives [64].
A second study [65] did not subdivide schizoaffective
disorder into the two subtypes of SA-BP and SA-SCZ
but found either recurrent bipolar or recurrent
schizophrenia cases in the relatives of schizoaffective
probands.
Cognitive deficits and imaging
changes in family genetic studies
X-ray computed tomography (CT) brain imaging
studies have demonstrated enlargement of the lateral
ventricles and the third ventricle, enlargement of the
cortical sulci and cortical atrophy. The relationship
between CT scan appearance and family history of
schizophrenia has been a subject of controversy in the
past [66]. Weinberger et al. [67] demonstrated that
schizophrenics have larger ventricle-brain ratios
(VBRs) than their healthy siblings. DeLisi et al. [68]
investigated families containing two or more siblings
with a history of schizophrenia. A two-way analysis
of variance found significant effects from family
234
membership and illness status on CT scan frontal
horn measurements [69]. Some authors have found
no difference between those with and those without a
positive family history [1]. Others have reported sig-
nificant results but the direction of the relationship
has varied, with some finding larger lateral ventricles
in the nonfamilial disorders [70–72] and others dem-
onstrating the reverse [73]. Owen et al. [74] described
a complex curvilinear relationship between family
history and VBR; the highest frequency of positive
family history scores occurred in those with medium
VBRs. Lewis [66] has commented that the distinction
between familial and sporadic forms of schizophrenia
in terms of CT brain scan appearance has become less
tenable. A family CT scan study by Orlova et al. [75]
estimated heritability for CT parameters. Width of
the frontal horn of the left lateral ventricle, width of
the right lateral ventricle frontal horn, and measures
of the right lateral horn were found to be transmitted
genetically and also to be associated with schizophre-
nia. Palmour et al. [76] reported two cases of DSM-
III-R schizophrenia and schizotypal disorder in a
family who both had enlarged ventricles and widened
sulci in scanning. Each person also had a chromosome
4 inversion (inv 4 [p15.2; q21.3]). The most recent
study of familial versus nonfamilial schizophrenia pre-
sented data which showed that familial schizophrenia
was associated with more severe structural abnormal-
ities than sporadic schizophrenia, especially in the
thalamus [77].
Neurophysiological deficits with
family genetic control
Blackwood [78] and Holzman [79] have argued for
augmenting the phenotype of schizophrenia to include
smooth pursuit eye movement studies, evoked electro-
encephalogram (EEG) response and cognitive changes.
This extended phenotype may then form the basis for
family genetic linkage studies [80]. Holzman et al. [81,
82] examined pursuit eye movement dysfunctions in
schizophrenia and found evidence for familial recur-
rence in specific families. Siegel et al. [83] studied
inhibitory gating of auditory-evoked responses in
schizophrenic patients, their FDRs, and normal sub-
jects. The expected deficit in inhibition in people with
schizophrenia was found. Approximately half the
FDRs and one parent had a similar deficit. Deficit in
the parents was associated with a family history of

schizophrenia. Deficits in P300 and other long latency
event-related potentials were found in 65 schizo-
phrenic subjects and these had no relationship with
family history. Auditory P300 and eye tracking dys-
function in schizophrenic pedigrees have been studied
in 20 high-density schizophrenic families [78, 84].
Abnormalities in one or both measures were found
to be correlated with psychiatric illness in most of the
families. Normal relatives also had abnormalities.
Comparison of auditory sensory gating, hippocampal
volume, and catecholamine metabolism in cases
versus unaffected sibling pairs has been carried out
[85–88]. Abnormal auditory sensory gating was
found in all schizophrenics in the 11 families studied
and in 8 of their 20 siblings. Compared with the
schizophrenics, the clinically unaffected siblings with
abnormal auditory gating had larger hippocampal
volumes. Cannon et al. [89] utilized a similar design
and found that both schizophrenics and their non-
schizophrenic siblings were impaired neuropsycholo-
gically compared to normal controls. These and other
results strongly support the hypothesis that impaired
information processing and neurophysiological def-
icits aggregate in the family members of schizophren-
ics and may serve as an indicator of partial penetrance
to the disorder [90–93].
Other studies have demonstrated a deficit in
early visual processing in clinically unaffected FDRs
of patients with schizophrenia [94, 95]. Functional
magnetic resonance imaging (MRI) studies have been
carried out in cases and their apparently healthy
siblings [96, 97] who showed altered activity and con-
nectivity [98, 99]. Many other endophenotypic deficits
have been found in unaffected siblings [97, 100–112].
Every study finds the same pattern, namely, that normal
siblings and relatives of schizophrenic probands who
are not diagnosable as cases of schizophrenia but likely
carriers of schizophrenia susceptibility genes have def-
icits in multiple imaging, neurophysiological, and psy-
chological modalities [113, 114].
Twin studies
Higher concordance in monozygotic twins as opposed
to dizygotic twins suggests a genetic effect on schizo-
phrenia. Gottesman and Shields used systematic ascer-
tainment by selecting their twin probands from the
Maudsley Twin Register where all patients who were
twins were consecutively ascertained [115]. They found
an monozygotic/dizygotic (MZ/DZ) concordance ratio
Chapter 19: Genetics of schizophrenia
of 58 : 12 and taking a weighted average from their own
and four other methodologically similar studies found
a concordance ratio of 46 : 14. The discordance rate of
over 50% in monozygotic twins supports the import-
ance of epigenetic, random and nongenetic biological
or specific environmental effects but which exclude
the family or common shared twin environment. Simi-
lar MZ/DZ concordance ratios were found by other
workers using the narrow Research Diagnostic Criteria
(RDC), DSM-III and Feighner Criteria [42, 116].
In a study of the Maudsley twin series, Farmer
et al. [116] found that the maximum difference in
MZ/DZ concordance (i.e. the highest index for herit-
ability) was for a definition of being affected which
included all psychoses that exhibited mood incongru-
ent delusions, schizotypal personality and atypical
psychosis, thus suggesting a common genetic origin
for this spectrum of disorders. Onstad et al. [117]
found concordance rates of 48% for MZ twins and
4% for DZ twins when DSM-III-R criteria were used
to diagnose schizophrenia. Nonaffective psychotic
disorder such as schizotypal and paranoid personality
disorders were observed in both MZ and DZ co-twins.
Torgersen et al. [118] studied 176 nonschizophrenic
co-twins and other FDRs of schizophrenic probands
and compared them to control twins. Schizotypal
personality disorders were found to be more common
among the biological relatives of schizophrenic pro-
bands. Odd speech, inappropriate affect, odd behav-
ior, and excessive social anxiety, were significantly
more common among the relatives of schizophrenic
probands. A biometrical approach to the genetic and
environmental effects on schizophrenia using twin
data has shown that the family environment effect
in the susceptibility to schizophrenia was less than
1% and that the genetic component was 93% [119].
Other heritability estimates using twin data were
between 67 and 80% [120, 121]. This implies that
the role of mothering and fathering in causing schizo-
phrenia is virtually absent.
Co-twin control studies
A very sensitive method of demonstrating structural
brain changes, cognitive and neurophysiological def-
icits in schizophrenia is through the use of an MZ
twin as a control. Sophisticated measurements of
brain morphometry can be made with MRI and CT
in MZ twins. This method also controls for age and
for polygenic background effects.
235
 Chapter 19: Genetics of schizophrenia
CT studies performed with this design show
consistent increases in ventricle brain ratio in schizo-
phrenia, and MRI studies document reductions
in temporal lobe structures as well [35, 67, 68, 71,
98, 99, 122, 123]. In twin studies of normal and
schizophrenic twin pairs [124, 125] there was a very
high within-twin pair correlation for ventricle brain
ratio in healthy MZ twins which was nonsignificant in
DZ twins. However, when one MZ twin was schizo-
phrenic and the other normal, the correlation
between VBRs was reduced with the schizophrenic
twins having the larger ventricles. This has been taken
to indicate that, in MZ twins discordant for schizo-
phrenia, other stochastic or environmental forces are
operating to cause both the illness and the increased
ventricular size. Suddath et al. [99] used MRI to
examine MZ twins discordant for schizophrenia and
found that, as a group, the affected twins had larger
ventricles and smaller hippocampi than the
unaffected twins. Goldberg et al. [126–135] studied
discordant MZ twins and found within pair correl-
ations between left hippocampal blood flow and
verbal memory and between Wisconsin card sorting
and prefrontal flow. Co-twin control studies of
obstetric complications in schizophrenia were carried
out by Gottesman and Shields which were not sup-
portive of an effect on schizophrenia [115]. Cantor-
Graae et al. have carried out a similar study in which
they examined the whole neonatal period for neuro-
logical hard and soft signs in 22 MZ twin pairs dis-
cordant for schizophrenia and 7 normal comparison
MZ twin pairs [136]. The neurological impairment
that was found in the healthy discordant MZ twins
was significantly related to history of both neonatal
and total obstetric complications. Family history of
psychosis, history of substance abuse and history of
postnatal cerebral trauma were not related to degree
of neurological impairment in the ill or healthy co-
twins. In the same sample McNeil et al. [137, 138]
investigated obstetric complications (OCs) during
pregnancy, labor and the neonatal period. Significant
differences in OC rates were found in discordant pairs
compared to normal control pairs. Labor complica-
tions were more frequent in discordant than concord-
ant pairs. OC rates were equal in affected compared to
healthy discordant twins. Trauma at the time of labor
and delivery and especially prolonged labor appeared
to be of importance for brain structure anomalies
associated within pairs of twins discordant for
schizophrenia.
236
Suddath et al. [99] studied identical twins and
described reduced temporal lobe volume as well as
reduced temporal lobe gray matter in schizophrenia.
In 12 out of the 15 discordant pairs, the twin with
schizophrenia could be identified by visual inspection.
In two pairs no difference could be seen and in one
twin the schizophrenic was misidentified. Quantita-
tive analysis of the hippocampus showed it to be
smaller on the left in 14 of the 15 affected twins,
as compared with their normal twins, and smaller
on the right in 13 affected twins. Another study
of discordant MZ twins [139] found a focal decrease
in gray matter density accompanied by a focal
increase in white matter density. Another three stud-
ies [140–142] found that variation in hippocampal
and ventricular volumes within discordant monozy-
gotic pairs was observed. Hippocampal volumes of
probands were smaller than those of their nonschizo-
phrenic MZ and DZ co-twins and healthy twins.
Hippocampal volumes of unaffected co-twins were
smaller than those of healthy twins, but those of
unaffected MZ and DZ co-twins were similar.
A longitudinal five-year follow-up in MZ and DZ
twin pairs discordant for schizophrenia and healthy
comparison twin pairs using brain MRI was carried
out by Brans et al. [143]. Significant decreases over
time in whole brain and frontal and temporal lobe
volumes were found in patients with schizophrenia
and their unaffected co-twins compared with control
twins. Sommer et al. [144] studied language laterali-
zation and found that it was decreased in discordant
twin pairs compared with the healthy twin pairs.
Functional imaging for verbal tasks in discordant
twins with schizophrenic probands has been carried
out [145]. All subjects with schizophrenia showed
increased activation in the right homologue of Broca’s
area in contrast to healthy individuals. Ettinger et al.
[146] found that thalamic volume was reduced in cases
compared to normal cotwins and that the reduction in
volume was greater in concordant twins than in dis-
cordant twins. Likewise the unaffected twin had more
volumetric reduction than normal controls. When gray
matter volume was examined in the same sample of
twins; frontal, insular, cingulate, medial, parietal, and
temporal cortical gray matter changes were found that
correlated within pairs with diagnosis [147].
A review and meta-analysis of functional scanning
with MRI positron emission tomography (PET) and
single photon emission computed tomography
(SPECT) in case-control studies, in twins and in cases

with family controls, found that vulnerability to
psychosis was associated with medium to large effect
sizes when prefrontal activation was contrasted with
that in controls [148]. Relatives of patients affected
with psychosis, the co-twins of patients and subjects
with an “at risk mental state” shared similar neuro-
cognitive abnormalities. The prefrontal and anterior
cingulate cortex, the basal ganglia, hippocampus, and
cerebellum were most affected. Other co-twin control
studies [130–133, 149–154] confirm these findings.
Clearly within pair differences need an explanation
and the recent finding of genomic imprinting differ-
ences within MZ twin pairs as reflected in differential
methylation effects on DNA is a possible explanation
[155, 156]. Such effects could combine with other
random stochastic genetic and biological processes
or in utero biological effects to produce discordance.
Adoption studies
Adoption and cross-fostering studies can clarify the
respective contributions of the genes and environ-
ment because they are more clearly separated than
in the twin method. The most prominent adoption
studies come from a Danish–American collaboration
which used the Danish National Adoption Register
[43, 157–160]. The studies have been conducted on
a Copenhagen sample, then on provincial samples,
and finally combined into a national sample. All the
studies have shown that the adopted-away children of
schizophrenics had a significantly higher risk of
developing schizophrenia or schizotypal disorder
and related conditions than the adopted-away chil-
dren of controls. Most of the children were born
before the first episode of illness in the parent and
about a third of the schizophrenic parents were male,
thus weakening arguments that the schizophrenia
may have been caused by early mother–child inter-
action or by an intrauterine event. The data obtained
by Kety et al. [158] early in the Copenhagen data
found an increased incidence of schizophrenia and
related disorders in the biological relatives of schizo-
phrenics who had been adopted (20.3%) compared to
the incidence in adoptive relatives and relatives of
controls (5.8%). The data were reanalyzed using
DSM-III criteria [27] and a 13.5% incidence for
schizophrenia or schizotypal personality in the bio-
logical relatives of schizophrenics was found com-
pared to 1.5% in the controls. In 1994, Kety et al.
[160] reported on the provincial sample and found
Chapter 19: Genetics of schizophrenia
similar results. In the combined national sample of
Copenhagen and provincial adoptees, Kety et al. com-
ment that chronic schizophrenia in adoptees was
found exclusively in the biological relatives of schizo-
phrenics and that the overall prevalence of schizo-
phrenia in adoptees was 10 times greater than the
rate in the biological relatives of controls. Later these
studies were reviewed and it was noted that these
adoption studies showed genetic transmission of
schizophrenia, nonaffective psychosis, and schizoty-
pal rather than any general liability to other forms of
psychopathology [161]. Cardno also reviewed the
adoption studies and found that the risk of chronic
schizophrenia in relatives was predicted by the pres-
ence of the pervasive negative symptoms of social
withdrawal, autistic behavior, poverty of thought/
speech, and flat affect and absence of pervasive posi-
tive symptoms [162].
Tienari et al. [163, 164] have carried out adoption
studies of schizophrenia in Finland. The sample was
obtained from a national register. The children of
schizophrenics given up for adoption were compared
with matched controls who were adopted offspring
of nonschizophrenic biological parents. The adoptive
families and biological parents were investigated with
interviews and psychological tests. Among the 155
index offspring, the percentage of schizophrenia and
other psychoses, borderline syndrome, and severe
personality disorders, was significantly higher than
in the 186 matched control adoptees. In adoptive
families rated as disturbed, genetic effects were more
prominent, thus giving an impression that the envir-
onment interacted with the genetic susceptibility to
increase liability to schizophrenia.
Segregation analysis and variance
component analysis of twin and
family data
As discussed in the twin study section above, data
from these studies gave heritability estimates of
between 66 and 93%. Family data indicate that up to
93% of the variance in the etiology of schizophrenia
may be genetic in origin [119, 120, 165–168]. McGue
et al. [167] estimated the family common environ-
mental effect to be about 19%. McGuffin et al. [168]
suggest that stochastic processes may be partly
responsible for the nonfamilial or specific environ-
mental source of etiological variance. Henderson [22]
237
 Chapter 19: Genetics of schizophrenia
also noted that the specific environmental variance
contains an error term, thus potentially diluting the
amount of true specific environmental variance in
schizophrenia. Nevertheless, the specific environmen-
tal component may include a proportion of schizo-
phrenia being due to viral infection, obstetric
complications or other nongenetic causes of brain
disorder, but it could be argued that the data as a
whole shows that the bulk of the schizophrenias are
genetic in origin.
Because of the strong evidence of the genetic
origins of schizophrenia, attempts have been made
using segregation analysis to determine a mode of
transmission. Baron [169] has reviewed some of this
work and commented that there was considerable
variation in methodology between studies and con-
flicting results. Baron found that the single major
locus model was compatible with the data in 7 out
of the 12 analyses but was rejected by 5 [170–180]. In
contrast, the multifactorial polygenic model was
rejected in only one out of five studies [118, 165,
181–183], while two studies rejected both the single
major locus and polygenic models [184, 185]. Finally
two studies that employed the mixed model con-
cluded that major locus transmission with polygenic
background was compatible with the data [180, 186].
All of these studies were of course conducted under
the assumption that the etiology of schizophrenia was
homogeneous in the samples obtained which is an
assumption which we now know is flawed.
Genetic linkage and association
studies of schizophrenia
Considerable effort has been focused on genetic link-
age analysis of schizophrenia employing genetic
markers in multiply affected families to identify
which chromosomal regions harbor susceptibility
genes. This approach must take into account the
complication of heterogeneity of linkage in which a
number of susceptibility genes localized to different
chromosomes contribute to the development of
schizophrenia. One strategy in linkage studies has
been to investigate favored loci implicated by cyto-
genetic abnormalities. This older literature was
reviewed by Bassett [187, 188], but in the last few
years a large amount of extra data has been accumu-
lated using microarrays. A second approach is to
simply study relevant genes such as neurobiological
238
enzymes, peptides and receptors. The third approach
is to detect functional variations in proteins through
the identification of genetic variation at the DNA level
relevant to the functioning of the resultant protein.
The following sections describe a selection of
genes that have been implicated in susceptibility to
schizophrenia by cytogenetic, linkage and/or associ-
ation studies. They do not represent every gene impli-
cated by these studies. They do however provide the
reader with a flavor of the complexity of the issues
surrounding the field and indicate where progress has
been made.
Chromosome 1: NOS1AP, UHMK1,
RGS4, DISC1
A combined allelic association, cytogenetic and
family linkage study on chromosome 1 using hetero-
chromatic C-band variants in the 1q22.1–23 region
found co-segregation of a 1qH (C-band) variant and
Duffy blood group alleles with schizophrenia in a
single family [189]. Later two linkage studies found
significant evidence for linkage between genetic
markers on chromosome 1q23.3 and schizophrenia
with LODs of 6.48 and 3.2 [190, 191]. A meta-analy-
sis of linkage studies conducted by Lewis et al. [192]
found consistency in linkage studies on chromosome
1p13.3–1q23.3 with a significance for nonrandom
positive LOD scores of p ¼ 0.014. Other linkage
studies from Finland found significant evidence for
linkage to schizophrenia at 1q32.2 but at a distance
from the DISC1 susceptibility locus [193, 194] nearer
the telomere. Two further analyses from Finland in
independent family samples found a LOD of 3.21
[195] and then 2.7 at the more distal position of 1q42
near the DISC1 gene [194]. Other linkage studies
found supportive evidence for linkage to schizophre-
nia at 1q23.3 [191, 196, 197]. The strongly positive
1q23.3 linkage studies and the positive LODs lower
down on the chromosome lead to the conclusion
that there is locus heterogeneity with at least two
loci on chromosome 1. The regions of chromosome
1 showing linkage to schizophrenia are shown in
Figure 19.1.
Claims that the NOS1AP/CAPON and RGS4 genes,
which are 700 kb apart on chromosome 1q23.3, show
genetic association (LD) with schizophrenia have
been published [198–200]. RGS4 has been found
to be associated with schizophrenia in a minority

1p13.3-1q23.3
1q22.1-1q23.3
1q21.1 1q23.3
Deletion CNVs
NOS1AP
UHMK1
RGS4
Figure 19.2 Relative genomic positions of schizophrenia associated genes on 1q23.3. The region shown covers 1.13 Mb.
of studies [201, 202] and some doubt exists about
its validity as a schizophrenia susceptibility gene
[201, 203]. Markers at a third gene called UHMK1
in between NOS1AP and NRG1 showed positive allelic
association with schizophrenia [203]. The marker
D1S1679, which lies between UHMK1 and CAPON
has been implicated in two family-based studies of
schizophrenia using tests of transmission disequilib-
rium (TDT) [204] and the extended transmission
disequilibrium test (ETDT) [205], which are both
methods that attempt to combine linkage and LD
approaches. Further fine mapping of the UHMK1
gene locus was then confirmed in a sample from
Aberdeen [203, 206]. The difficulty in replicating
genetic effects from RGS4 and NOS1AP/CAPON in
schizophrenia may be because there is LD between
these genes and the UHMK1 gene which is in the
middle of RGS4 and NOS1AP/CAPON within a 700
kb region. The relative location of the 1q23.3 genes
with evidence for association with schizophrenia are
shown in Figure 19.2.
DISC1 located at 1q42 (Figure 19.1) was originally
identified at the breakpoint of a balanced t(1;11)
translocation in a large Scottish pedigree multiply
affected with psychiatric illness, including schizophre-
nia, bipolar disorder, and recurrent major depression
Chapter 19: Genetics of schizophrenia
Figure 19.1 Regions of chromosome
1 showing linkage to schizophrenia,
1q42
the position of the 1q42.1 gene DISC1 that is
disrupted by a translocation in a pedigree
with schizophrenia and the location of
1q32.2 DISC1 the 1q21.1 Copy number variation (CNV)
1q42.1 deletions are shown.
[207, 208]. DISC1 is expressed at half the normal
levels in translocation cell lines, indicating that
haplo-insufficiency may be the cause in this family
[209]. Multiple linkage and genetic association
studies over 20 years [194, 210–222] have provided
evidence for linkage and association between DISC1
and schizophrenia or bipolar disorder. Endopheno-
typic studies of cognition [212, 215, 223], memory
function [211, 224], and imaging variables [211, 212,
218] have also provided support for the involvement
of DISC1. A large-scale association study combining
4 schizophrenia cohorts with 1275 cases and 1236
population-matched controls showed evidence of
allelic and haplotypic heterogeneity for association
between markers in DISC1 and schizophrenia and
bipolar disorder [225]. These findings provided fur-
ther support for the role of DISC1 in schizophrenia
and bipolar disorder and found that there were
multiple disease haplotypes having an effect within
European populations. Animal models of DISC1 are
discussed below.
Chromosome 2: ZNF804A
A genome-wide association study (GWAS) found good
evidence of association for the zinc finger 804A
239
 Chapter 19: Genetics of schizophrenia

5q11-5q13 5q22-5q34 Figure 19.3 Regions of chromosome

5 showing linkage to schizophrenia.

5q22-5q31 5q33

CLINT1GABRA1
GABRA2
GABRA6

KIAA0171 EBF1 ADRA1B CCNJL ATP10B GABRB2 GABRG2

CLINT1 EBF1 TTC1 SLU7 LOC285629 GABRB2 GABRG2
CLINT1 EBF1 PWWP2A SLU7 GABRB2 GABRG2
CLINT1 EBF1 PWWP2A SLU7 GABRB2 GABRG2
CLINT1 AK123543H PWWP2A PTTG1 GABRB3
RNF145 FABPB ATP10B GABRB2
RNF145 AK097242 ATP10B GABRA6
RNF145 FABP8 ATP10B GABRA6
RNF145 FABP8 KIAA0715 GABRA1
RNF145 CCNJL ATP10B GABRA1
RNF145 CCNJL GABRA1
UBLCP1 CCNJL GABRA1
IL12B C1QTNF2 GABRA1
AK097548K C5orf54 GABRA1
LOC285627 C5orf54 GABRA1
PTTG1
DO658414 k

Figure 19.4 Relative genomic positions of schizophrenia associated genes on 5q33.3–5q34. The region shown covers 4.4 Mb.

(ZNF804A) gene with schizophrenia in a UK sample
[226]. These findings were supported by analysis of
16 726 cases and controls drawn from independent
European samples and those from the United States,
Australia, Israel, China, and Japan [226]. Independent
replication of these findings with the same marker
and allele (rs1344706 allele T) was found in a schizo-
phrenia sample from Ireland [227] and in a study of
5164 schizophrenia cases and 20 709 controls from
Europe and China [228].
Chromosome 3: DRD3
Crocq et al. [229] and Kennedy et al. [230] reported
increased homozygosity or allelic association of the
dopamine receptor gene DRD3 in schizophrenics com-
pared to controls. The statistical significance of this
result was most marked in positive family history cases
and in those with good treatment response. Mant et al.
[231] attempted a replication but found significance of
only 0.05. Evidence of increased homozygosity was
also found in a UK/Icelandic family sample [232].
The combined sample of Crocq and Mant nevertheless
showed a highly significant increase in homozygosity
with p < 0.00005. Further attempts at replication of
this finding were not successful [233–237]. DiBella
et al. [238] found an association between the Msc1
polymorphism and delusional disorder but no
increased homozygosity with schizophrenia.
Chromosome 5: CLINT1 (EPSIN4),
GABRA1, GABRA6, GABRB2, GABRP
The first evidence for a positive LOD score with
genetic linkage markers was claimed by Sherrington
et al. [239] when investigating a chromosome 5 region
implicated by a cytogenetic abnormality [240] which
co-segregated with schizophrenia. Linkage studies
using probes localized a schizophrenia susceptibility
locus to the chromosomal area 5q11–5q13 in five
large Icelandic and two large British families. Two
entirely new cohorts of families from Iceland and
the United Kingdom excluded linkage over the
5q11–q13 region [241]. This linkage was unreplicated
until 2006 when genetic linkage to schizophrenia with
a LOD of 3.8 was found over the 5q11–13 region
[242]. Genetic association studies in the 5q11–13
region have since implicated the kinesin interacting

240

factor 2A (KIF2A) gene in susceptibility to schizo-
phrenia [133], but further tests of association
between KIF2A and schizophrenia are needed to con-
firm this.
Other linkage studies have implicated the more
telomeric chromosome 5q22–34 region. The first was
in a sample of Irish families [243] with a heterogeneity
LOD of 3.35 near the locus D5S804. Subsequently, a
single large kindred from Palau, Micronesia, gave a
LOD of 3.4 showing linkage to schizophrenia with
markers in the 5q22–31 region [244]. Later a study of
Finnish families also found linkage to schizophrenia
with a LOD of 3.56 at 5q22–31 [245]. A British/
Icelandic study implicated the 5q22–34 region with a
LOD score of 3.6 [246]. A collaborative study of the
5q22–31 region [247] found that 4 out of 8 independ-
ent schizophrenia family linkage samples gave LOD
scores above 1.00 [248]. These studies suggest that
there must be at least 2, and possibly 3 or more,
different regions of chromosome 5 which harbor sus-
ceptibility loci for schizophrenia (Figure 19.3). The
evidence implicating the more distal region of chromo-
some 5 at 5q33 is derived from 4 independent linkage
studies reporting LOD scores above 3.00.
Genetic association studies of the 5q21–22 region
have now shown that several genes show allelic associ-
ation with schizophrenia. Pimm et al. [249] implicated
the clathrin interacting protein 1 (CLINT1 or EPSIN4)
in susceptibility to schizophrenia. This was strongly
replicated with several new markers at the CLINT1
locus in independent samples by the International
Schizophrenic Consortium [250]. Liou et al. [251] also
found some degree of support for involvement of the
CLINT1 genes in schizophrenia by showing positive
haplotypic association in a Chinese sample. Support
for the finding also derives from a family-based asso-
ciation study [252] which was positive in a second Han
Chinese population. The methodology of these studies
has been reviewed [253]. Since then a Latino family-
based association study [254] has also confirmed
involvement of CLINT1 in schizophrenia.
A second study of the 5q23–34 region containing
the GABA receptor subunit genes GABRA1, GABRA6,
GABRB2, and GABRP was carried out [255]. Associ-
ations between single nucleotide polymorphisms
(SNPs) and haplotypes in GABRA1 with schizophrenia
was found as well as associations in the GABRP and
GABRA6 genes. The positive GABRA1 and GABRP
findings were replicated in an independent German
family-based sample [255] and association with
Chapter 19: Genetics of schizophrenia
GABRA1 was found in a Japanese sample [256]. The
chromosomal region implicated shows a low level of
genetic recombination. Therefore markers at the
GABRA1, GABRA6, and GABRG1 genes will show
LD with each other making it uncertain as to which
gene is involved. GABRP is 10 million bases away and
will not show LD with the 3 genes nearer the centro-
mere and will give an independent allelic association
signal. GABRB2 also on 5q34 has been associated with
schizophrenia [255] and this gene is not in LD with the
other 5q34 GABA receptor cluster genes. The involve-
ment of the GABRB2 gene has also been shown in
Japanese and German populations [257] as well as in
an independent Japanese sample [258] and in others
[259, 260]. A meta-analysis of the involvement of
GABRB2 gene [261] confirmed that it was associated
with schizophrenia across several studies (Figure 19.4).
Chromosome 6: HLA, DTNBP1,
TRAR4 (TAAR6)
Early studies of association and linkage between the
HLA genes and schizophrenia produced consistently
negative results [262–264] with the exception of asso-
ciation between HLA A9 and paranoid schizophrenia.
This has been put into reverse with several GWAS,
described below, now implicating at least three haplo-
types in the HLA region. Evidence from linkage analy-
sis in families that there is a susceptibility locus for
schizophrenia on chromosome 6p22.3 was provided
by several research groups [265–268]. However, other
research teams were unable to detect the presence of
this locus in family samples [190, 245, 246, 269, 271].
The evidence pointed strongly to heterogeneity of
linkage and the study by Straub and coworkers [272]
was the first time in psychiatry that an admixture test
rejected homogeneity in favor of a linked 6p22–25
subgroup. Confirmed evidence for linkage at the
6p22.3 locus was followed up with allelic association
studies using the family-based methods implemen-
ted in TRANSMIT [272]. When the most stringent
test of allelic association was carried out evidence
for allelic association with schizophrenia was present
[272].
The haplotype structure in the original Irish fam-
ilies showed that 96% of the variability in the sample
was attributed to 6 haplotypes [273]. Only 1 of the
6 haplotypes was identified as a high risk haplotype
and this had a frequency of 6% in the whole sample
[273, 274]. A subsequent study of German, Hungarian,
241
 Chapter 19: Genetics of schizophrenia
and Israeli families also showed positive tests of allelic
TDT with the DTNBP1 locus but only with a single
SNP and not with any of the SNPs that had been
previously found to be associated in the Irish family
sample [275]. The German sample also showed a
positive haplotypic association composed of three
adjacent SNPs. However, the alleles in the associated
haplotype were different to those found in Straub
et al.’s Irish sample [272].
A positive haplotypic association with DTNBP1
was also found in a study of Han Chinese trios, where
the risk haplotype was composed of the most
common SNP alleles [276]. This was in contrast to
the haplotype associated in the Irish study which was
composed of minor frequency SNP alleles [272].
A genetic association study consisting of three case-
control samples from Germany, Poland, and Sweden
found no evidence for association with any of the
five SNPs previously reported to be positive except
for one SNP (rs1011313) in the Swedish sub-sample.
Haplotypic associations were also negative for the
German and Polish sub-samples. Only the Swedish
schizophrenic sample showed haplotypic association
with schizophrenia. Several, two, three, or four
marker haplotypes and a single five-marker SNP hap-
lotype showed association. The strength of association
was increased in the subgroup of cases with a positive
family history of schizophrenia [277]. The Swedish
haplotype was yet another one found to be associated
with schizophrenia. Van Den Bogaert et al. [277]
suggested that the five marker Swedish haplotype asso-
ciated with schizophrenia is phylogenetically related to
the original haplotype found to be associated with
Irish schizophrenia by Straub et al. [272].
A novel DTNBP1 haplotype incorporating SNP A
(T allele) has been identified in a case-control sample
of 708 cases and 711 controls of British and Irish
descent that showed association with schizophrenia
[278]. This haplotype was also found to be associated
with schizophrenia in an Irish case-control sample that
did not initially show association [279, 280]. The same
research group also reported a positive replication in a
sample of 488 Hungarian schizophrenic trios [280].
Another association study by Funke et al. [281] found
an allelic association that was with the opposite allele
found to be associated by Straub. Hall et al. [282]
found no significant allelic or haplotypic associations
between DTNBP1 and schizophrenia [282]. An Israeli
isolate was used in a genome-wide scan by Kohn et al.
[283] to identify loci predisposing to any diagnosis of
242
psychosis. This isolate was unique in that there was
only one paternal origin for the whole village as a result
of a bottleneck event that occurred a few centuries ago.
Identical-by-descent (IBD) haplotype sharing was
used to carry out a genome-wide scan with 359 micro-
satellite markers. A positive linkage (p < 0.0001) was
observed over the DTNBP1 gene on chromosome 6.
However, such an inbred population cannot distin-
guish between linkage and association.
Numakawa et al. used a case-control sample of
670 patients and 588 controls and found that four
SNPs were associated with schizophrenia [284].
Positive allelic and haplotypic associations were also
reported by Li and colleagues [285] in a Han Chinese
population. Positive haplotypic but not allelic asso-
ciation was reported by Tochigi et al. [286]. How-
ever, as with the some other studies [275, 276], the
associated haplotypes were not the same as reported
in the original study. Li et al. [285] screened a
Scottish sample of 580 schizophrenic cases and 620
controls. They were unable to replicate any previously
identified single marker or haplotypic associations
[278]. However a rare haplotype of SNP A with
P1583 remained significant after permutation testing
(p ¼ 0.03) [285]. No evidence for allelic or haplotypic
association with these markers was found in an inde-
pendent UK schizophrenia sample [287]. It seems that
dysbindin is associated with schizophrenia but that
there are heterogeneous disease haplotypes in different
samples.
DTNBP1 is a 40-kDa coiled-coil-containing pro-
tein that shows cross species evolutionary conser-
vation [288]. DTNBP1 binds to both alpha- and
beta-dystrobrevin in muscle and brain [288]. In the
brain DTNBP1 is found in axons, particularly in the
mossy fiber synaptic terminals of the cerebellum and
hippocampus [288]. Talbot and colleagues [289]
compared presynaptic dysbindin-1 levels at hippo-
campal formation sites lacking neuronal dystro-
brevin between schizophrenics and controls and
found that these levels were significantly reduced in
schizophrenics.
Other genes on chromosome 6 that have been
implicated in schizophrenia by allelic association
include the HLA region genes [250, 290] and the trace
amine receptor 4 (TRAR4 now known as trace amine
associated receptor 6 [TAAR6] on 6q23.2 [291]). Early
studies of association and linkage between the HLA
genes and schizophrenia produce consistently negative
results [262–264] with the exception of association

6p22.2-6p21.32
6p22.3 (HLA region)
6p22-6p25 DTNBP1
(TAAR6)
7q12.3 7q22.1
(ABCA13) (RELN)
7q21.11-7q21.12
(GRM3)
between HLA A9 and paranoid schizophrenia. This
has been put into reverse with several genes now
implicating at least three haplotypes in the HLA
region [262–264]. The low level of recombination
across the HLA region makes it difficult to identify
which gene among hundreds is increasing genetic
susceptibility to schizophrenia. It will require deep
DNA sequencing of genomic DNA from schizophre-
nia cases associated with HLA makers followed by
humanizing mouse strains with genomic DNA from
schizophrenia cases with susceptibility alleles and
haplotypes to identify which part of the HLA region
is implicated. Allelic association between TAAR6
markers and schizophrenia has not been replicated
[292, 293] (Figure 19.5).
Chromosome 7: ABCA13, GRM3, RELN
A chromosome abnormality affecting the lipid trans-
porter gene ABCA13 was found in a person with
schizophrenia. Possible etiological base pair changes
in the form of multiple rare coding variants were
revealed in ABCA13 after sequencing 100 cases
[294]. ABCA proteins shuttle lipid molecules across
cell membranes. Different ABCA proteins have roles
in vesicular trafficking, signal transduction, and tran-
scriptional regulation. ABCA13 is strongly expressed
in the choroid plexus and the dentate gyrus granule
cell layer in the central nervous system (CNS).
ABCA13 mutations were also found to be present in
cases of bipolar disorder, but there were differences
between schizophrenia and bipolar disorder in terms
of heterozygosity and homozygosity for rare variants.
Five cases were found to be compound heterozygotes
Chapter 19: Genetics of schizophrenia
Figure 19.5 Regions of chromosome 6
showing linkage to schizophrenia and
the positions of the schizophrenia
associated genes DTNBP1 and TAAR6.
6q23.2
Figure 19.6 The localization of genes
on chromosome 7 that have been shown
to be associated with schizophrenia.
and one was homozygous for missense mutations
whereas controls were heterozygous. This suggests
recessive inheritance with possible allele dosage
effects.
Several studies have implicated the metabotropic
glutamate receptor gene GRM3 in schizophrenia
[295–308]. MRI changes showing abnormal acti-
vation patterns in both cortical regions in control
subjects homozygous for a genetic variant have been
reported [300]. A transgenic animal for schizophrenia
is described below.
Markers at the reelin gene (RELN) were reported
to be associated with schizophrenia [309] and two
studies find this association only in females with
schizophrenia [310, 311]. Five SNP markers were
nominally associated with schizophrenia in the Inter-
national Schizophrenia Consortium (ISC) GWAS
[250]. However a number of studies have not shown
association with RELN [309, 312, 313] (Figure 19.6).
Chromosome 8: NRG1, PCM1,
PPP3CC, FZD3, DPYSL2
Five genetic linkage analyses of the 8p22–21 region
in independent schizophrenia family samples have
confirmed linkage with LODs above 3.00 [242, 246,
314–316]. Two other linkage studies have provided
further support for linkage at 8p23–21 with LODs of
between 2.00 and 3.00 in Irish and African-American
family samples [317, 318]. Linkage on the chromo-
somal region 8p22–21 is therefore one of the most
well-replicated linkages to schizophrenia. Polymorphic
markers localized within or next to five genes: PCM1,
243
 Chapter 19: Genetics of schizophrenia
PPP3CC, DPYSL2, FZD3, and NRG1 have shown
association with schizophrenia in the 8p22–12 region
[319–320]. Similar allelic and haplotypic associations
at the NRG1 locus were found in Scotland [326]. Tests
of haplotypic association did reveal weak evidence
for association (p ¼ 0.04) for the original HAPICE
haplotype associated with schizophrenia in Iceland
[327]. Seventeen microsatellite markers and three
SNPs were analyzed in an Irish case-control sample
and no allelic associations were detected [328]. How-
ever a novel two marker haplotype named HapBIRE
did show association. This haplotype shared some
alleles also found in “at risk” haplotypes reported in
the original Icelandic study [323]. The association of
markers at NRG1 with schizophrenia has also been
detected in the Chinese Han population [325]. An
independent attempt at replicating Yang’s findings
in different Chinese case-control and trio samples
did not reveal any significant associations [329].
Li [330] could find no evidence for association with
the HAPICE haplotype in either a case-control or
family-based sample. However, three novel associated
haplotypes were identified in these studies. A study
from Holland has also shown support for NRG1
involvement in schizophrenia but when patients were
split into “deficit” (chronic symptoms with idiopathic
negative symptoms) and “nondeficit” subtypes of
schizophrenia [331] a weak positive association was
found. A Chinese study took the LD structure into
account when performing haplotypic association ana-
lyses [332]. Thirteen microsatellite markers were gen-
otyped and the investigators observed four different
LD blocks present along the NRG1 region analyzed.
Significant haplotypic associations for two out of the
four LD blocks were found [332]. A further positive
replication was successful [333] using a case-control
and trio sample of Han Chinese origin. As with the
other studies using Han Chinese populations, the
HAPICE haplotype itself was not associated with
schizophrenia, but a novel haplotype was found to
be associated in both the case-control and the family
sample. A study based in the United States [334] also
found haplotypic association between NRG1 and
schizophrenia but not for the core HAPICE haplotype.
Gene expression studies in schizophrenia also sup-
port a role for NRG1 in schizophrenia susceptibility
[334, 335]. Although there has been substantial sup-
port for the NRG1 association with schizophrenia
there have also been negative studies [298, 336–338],
indicating that the effect of NRG1 is only in a few
244
percentage of cases and that the frequency of disease
alleles is very low and also variable even with similar
ancestral populations.
Tests of extended transmission disequilibrium
(ETDT) in the linked chromosome 8p region in a
sample of large multiplex schizophrenia families
found significant allelic association with schizophre-
nia using the marker D8S261 which is within the
PCM1 gene [320]. The association with D8S261 was
then replicated in a sample of 100 US-parent off-
spring trios [320]. The next significant evidence of
allelic association between schizophrenia and 4 SNPs
as well as with D8S261 was found in a case-control
sample of 450 cases and 450 controls [320]. Some of
these SNPs were also associated with schizophrenia
in a sample of 800 cases and 800 controls from
Aberdeen. In both the Scottish and University College
London (UCL) samples, significant haplotypic associ-
ations were found with the same alleles being impli-
cated [339]. Further evidence for allelic association
between PCM1 markers and schizophrenia at the
PCM1 locus was found independently in the Clinical
Antipsychotic Trials in Intervention Effectiveness
(CATIE) GWAS of schizophrenia [340], the Schizo-
phrenia Genetics Consortium (SGENE) GWAS
[290, 341], and the ISC GWAS [251]. Sixty-nine
markers across the PCM1 locus (the PCM1 gene plus
50 kb either side) have shown allelic association with
schizophrenia in multiple samples. The ISC data
showed that 10 PCM1 SNPs were associated with
schizophrenia after the UCL sub-sample was excluded
with p-values of 0.005–0.02 [250]. Taken together, the
association findings provide evidence for replicated
association of PCM1 markers with schizophrenia.
Abnormal brain morphology was also detected in
PCM1 associated cases of schizophrenia compared
to normal controls and non-PCM1 associated cases
of schizophrenia. Significant reduction in the volume
of orbitofrontal cortex gray matter was observed in
PCM1 associated patients whereas non-PCM1 associ-
ated schizophrenics showed reductions in temporal
pole volumes [320].
Sequencing of genomic DNA from schizophrenia
volunteers who had inherited haplotypes associated
with schizophrenia showed a threonine to isoleucine
missense mutation in exon 24 which was likely
to change the structure and function of PCM1
(rs370429). This mutation was found only as a het-
erozygote in 98 schizophrenic research subjects and
controls out of 2246 case and control research

subjects [339]. Among the 98 carriers of the threonine
to isoleucine amino acid change (rs370429) 67 were
affected with schizophrenia. Another potential etio-
logical base pair change in PCM1 was rs445422,
which altered a splice site signal [339]. A further
mutation, rs208747, was shown by electrophoretic
mobility shift assays to create or destroy a promoter
transcription factor binding site [339]. Five further
nonsynonymous changes in exons were also found.
Genotyping of the new variants discovered in the
UCL case-control sample strengthened the evidence
for allelic and haplotypic association (p ¼ 0.02–
0.0002). Given the number and identity of the
haplotypes associated with schizophrenia, further
etiological base pair changes should exist within and
around the PCM1 gene [339]. A rare PCM1 variant
that co-segregated with schizophrenia in a single
pedigree that was predicted to cause a premature
truncation of the PCM1 peptide was reported in an
US sample [342]. A 3 Mb deletion that includes the
entire PCM1 gene has been reported in a single case
of schizophrenia from Germany [343]. In the ISC
collaboration a duplication in a single case of schizo-
phrenia from Sweden that included the first 13 exons
of the PCM1 gene was discovered [343]. PCM1 has
now been shown to interact directly with the schizo-
phrenia susceptibility gene DISC1 [342]. An amino
acid change in DISC1 changes interaction with PCM1,
changes the region of localization of PCM1 at the
centromere, and effects noradrenaline release in
SHY5Y cells [344]. PCM1 is known to be crucial for
microtubule function and trafficking in neurons and
could cause the molecular pathology of a subtype of
schizophrenia.
The PPP3CC gene at 8p21.3, which encodes the
calcineurin gamma A catalytic subunit (CNAg) of
phosphatase 2B, has been implicated by a genetic
association study in schizophrenia. Gerber and col-
leagues [319] selected SNPs within PPP3CC and gen-
otyped these in a family sample from the United
States and South Africa. Allelic and haplotypic associ-
ation was detected with the two SNP alleles in the
PPP3CC gamma subunit in the US sample. The South
African sample failed to display significant global
transmission distortion, but did show a trend toward
association with the common five SNP risk haplotype
observed in the US sample.
An attempt to replicate these findings was per-
formed in a Japanese case-control sample but was
negative [345]. A study using an Ashkenazi Jewish
Chapter 19: Genetics of schizophrenia
population also failed to find association with any of
the markers found within PPP3CC [346]. PPP3CC is
an attractive candidate gene for schizophrenia suscep-
tibility, not only because it is highly expressed in the
central nervous system but also because it is thought
to have a pivotal role in the regulation of dopamine
signal transduction [347, 348].
Another gene reported to be associated with
schizophrenia in the 8p21.1 region is frizzled 3
(FZD3). FZD3 is a seven-transmembrane receptor
for Wnt glycoproteins in the Wnt signal transduction
cascade reactions which ultimately play an essential
part in neurodevelopment. First reported in a Chinese
Han sample of 246 trios, 3 SNPs were found to be
overtransmitted to cases with p-values ranging
between 0.003 and 0.00007 [324]. This sample also
displays association with markers and haplotypes
within the NRG1 locus [325]. Around the same time
a Japanese study identified variants within FZD3
which had allelic, genotypic, and haplotypic associ-
ation with schizophrenia [321, 324]. A Chinese study
also found association with markers at FZD3 [349].
No association at this locus was detected in a British
sample [350], or in Japanese case-control and family-
based samples [351, 352].
A further gene in the 8p21–22 region, dihydropyr-
imidinase related protein (DPYSL2), 4.1 Mb distal to
PPP3CC, has been found to show association with
schizophrenia [322]. A replication was attempted in
two North American samples, one of Caucasian, and
the other of African-American origin, and the C allele
was associated with schizophrenia only in the Cauca-
sian sample [353] (Figure 19.7).
Chromosome 11: FXYD6, DRD2
Genetic linkage analysis carried out as a result of the
finding of a cytogenetic abnormality in a rare family
has yielded a LOD score of above 3.00 on chromo-
some 11 [208]. The LOD score went to above 4.00
when all individuals with psychiatric illnesses were
counted as cases and showed that schizophrenia and
a number of other psychiatric disorders within a
single large kindred segregated with a chromosomal
translocation, implicating the 11q21–11q22 region on
the long arm of chromosome 11 in the vicinity of the
dopamine D2 neuroreceptor gene (DRD2). Chromo-
some 11q22 was first implicated in schizophrenia in a
linkage study of two Japanese pedigrees in which
positive LOD scores were reported [354]. This was
245
 Chapter 19: Genetics of schizophrenia
8p21.3 (PPP3CC)
8p21.2 (DPYSL2)
8p21.1 (FZD3)
8p12 (NRG1)
8p22 ( PCM1)
8p22-8p21
8p23-8p21
11q22.3-11q24.1
11q22
11q21-11q22 11q23.3 (FXYD6)
11q22-11q24
confirmed in a separate study of a large Canadian
pedigree in which a maximum LOD of 3.4 was
obtained [355]. Further evidence of linkage to schizo-
phrenia at 11q23.3 was found with a LOD score of 3.1
and 1 single family was found to have a LOD score of
3.2 [246]. The chromosome 11q22–24 region has
been shown to be one of the most well-established
linkages to schizophrenia by a meta-analysis of 20
genome scans which employed nonparametric rank
order statistics to show that evidence for linkage at
this region was non-random [191]. This region on
11q was subsequently studied in a group of Welsh/
English families and a LOD of approximately 3.40 was
found with a quasi-recessive model [356]. Other Scot-
tish, Irish, Swedish, Icelandic, and Japanese family-
linkage studies were largely negative [354, 357–359].
Markers in the region implicated by the linkage stud-
ies that have shown LD with schizophrenia in case-
control studies include the FXDY6 gene encoding
phosphohippolin and the DRD2 dopamine receptor
genes. FXYD6 was associated with schizophrenia in
two samples from the United Kingdom [360]. The
gene encodes a sodium ATPase regulator, phospho-
hippolin, which is highly expressed in most parts of
the brain [361–364]. Phosphohippolin may play an
important role in the excitability of neurons in the
246
Figure 19.7 Regions of chromosome 8
showing linkage to schizophrenia and
the positions of the schizophrenia
associated genes in the 8p22–12 region.
Figure 19.8 Regions of chromosome 11
11q23.2 showing linkage to schizophrenia and
(DRD2) the positions of the schizophrenia
associated genes in the 11q23 region.
11q23.3
central nervous system during postnatal develop-
ment, as well as those in the adult brain. Allelic
association between markers at or near DRD2
and schizophrenia has been reported surprisingly
frequently given that initial studies were negative
[233, 365–377]. Other studies in several different
populations were negative [378–381]. Two meta-
analyses of the serine or cysteine at codon 311 of
the gene in 2003 were both positive. One of these
included 9152 cases with significance of p < 0.001;
odds ratio (OR) 1.43 [382]. The other found a pooled
OR of 1.3 for the G allele that encodes the cysteine
residue, which was significant (p ¼ 0.007) [370]. A more
recent meta-analysis of the DRD2 SNPs rs1801028
and rs6277 was significant for association with both
SNPs [261] (Figure 19.8).
Chromosome 13: DAOA
Linkage evidence for a schizophrenia susceptibility
locus mapping to chromosome 13q has been reported
by several groups [196, 247, 314, 315, 383–386]. Other
research groups were unable to detect the presence of
this locus in family-linkage studies [246, 387, 388]
once again providing evidence that variability in the
outcome of linkage studies was due to heterogeneity

13q22-13q32
13q11-13q21 13q22-13q34
of linkage. The distribution of these linked markers
would suggest at least two distinct regions of linkage:
13q11–21 and 13q22–32. The distal 13q22–32 region
is more robustly supported by the linkage data.
Chumakov and colleagues [389] initially investigated
the 13q22–34 region in schizophrenia. Six markers
were found to be associated with schizophrenia in a
French-Canadian sample, two of which were also
associated in a Russian sample [389]. This region
contains two overlapping genes, DAOA (G72) and
G30, which are transcribed in opposite directions
[389]. A number of groups have subsequently
attempted to replicate the associations between
DAOA/G30 and schizophrenia. A meta-analysis of
association findings for the DAOA/G30 locus pub-
lished up until April 2007 found consistent evidence
for association between markers rs947267, rs778293,
and rs1421292 with schizophrenia [390]. Hall et al.
[282], who studied US and Afrikaans trios with
schizophrenia, found association between rs2391191
and schizophrenia in the Afrikaans sample. Koros-
tishevsky et al. [391] found suggestive evidence for
association with schizophrenia at the DAOA locus in
a Palestinian-Arab sample. Fallin et al. [346] used an
Israeli sample of schizophrenia trios and showed
association with markers at DAOA/G30. Shin et al.
[392] found evidence for association in a Korean
schizophrenia case-control sample. Shinkai et al.
[393] report association with schizophrenia in their
case-control sample and overtransmission of alleles to
affected offspring in their family-based sample. Cor-
vin et al. [394] report a positive association in an Irish
schizophrenia case-control sample. A study of Pales-
tinians also confirmed involvement of the DAOA
locus in schizophrenia [391].
A British study of allelic association with DAOA
polymorphisms in schizophrenia could not find asso-
ciation [395]. The authors extended their analyses to
include a subset of schizophrenia patients with major
mood disorders in their medical histories as well as
grouping together those found with psychosis only
[395]. Significance values became more significant
when this regrouping of cases was applied for the
Chapter 19: Genetics of schizophrenia
13q33.2 Figure 19.9 Regions of chromosome 13
(DAOA) showing linkage to schizophrenia and
the position of the schizophrenia
associated gene DAOA in the 13q33.2
region.
mood disorder group alone [395]. A second UK study
found allelic associations with both schizophrenia and
bipolar disorder [396]. Several studies have been pub-
lished where association was not detected between
DAOA and schizophrenia [397, 398]. The DAOA gene
is expressed in the endoplasmic reticulum [389] and
may have a role in glutamate synthesis (Figure 19.9).
Chromosome 14: NPAS3, AKT1
A cytogenetic abnormality at 14q13 in a mother and
daughter diagnosed with schizophrenia and schizo-
phrenia comorbid with mild learning disability con-
sisted of a balanced reciprocal translocation t(9,14)
(q34.2;q13). The 14q13 breakpoint disrupted the
NPAS3 gene which encodes CNS expressed transcrip-
tion factor of the basic helix-loop-helix PAS gene
family [399–402]. Over 20 SNPs within NPAS3 were
associated with schizophrenia in the ISC GWAS
[251], and association studies in Scotland and Canada
were positive for this gene [400, 403].
AKT1 (V-akt murine thymoma viral oncogene
homolog 1) located at 14q32.33 is involved in intra-
cellular signaling pathways thought to be of etio-
logical importance in schizophrenia.
Significant AKT1 SNP and/or haplotype associ-
ations have been reported with schizophrenia in
samples from the United States, Germany, Japan,
China, Ireland, and the United Kingdom [404–410].
In common with many of the other loci discussed, a
number of studies have failed to replicate these
findings [411–416] (Figure 19.10).
Chromosome 22: COMT, VCFS REGION,
PRODH, BRD1
A genome scan reported linkage between schizophre-
nia and markers on chromosome 22q12–13 [417].
Pulver et al. [418, 419] computed linkage analyses
and found a positive LOD of 1.54 over the region
22q12–13.1. Reanalysis by varying parameters in the
dominant model maximized the LOD to 2.82. In a
collaborative analysis in a larger sample from five
247
 Chapter 19: Genetics of schizophrenia
14q13.1
(NPAS3)
22q11.21 22q11.21
(PRODH) (COMT)
22q12-22q13.1
22q11 VCFS 22q12-22q13
deletion region
laboratories [419, 420] this linkage could not be con-
firmed assuming either homogeneity or hetero-
geneity. One of the collaborating laboratories [420]
also independently found a maximum positive
LOD of 1.51 between schizophrenia and the marker
D22S278, assuming homogeneity in their sample.
Kalsi et al. [421] in a UK/Iceland schizophrenia link-
age study could not find any evidence over the rele-
vant region for linkage to schizophrenia assuming
either homogeneity or heterogeneity. Polymeropoulos
et al. [422] found a positive LOD of 0.37 for marker
D22S278 and found that the sibling-pair linkage
method also favored linkage. Schwab et al. [266]
who studied German and Israeli families could not
exclude this region using a sibling-pair linkage
method. In a much larger sample which was, at that
time, almost the entire world’s collection of 500
schizophrenia families found linkage to the relevant
region on chromosome 22 [423]. Several other lines
of evidence support involvement of chromosome 22.
A case-control sample from Scotland found support
for the involvement of the BRD1 gene [424]. A com-
bined UK and Danish sample implicated BRD1 and
the neighboring gene ZBED4 in schizophrenia [425].
BRD1 is a brain-expressed gene, which has a bromo
domain and a FAM domain, and by sequence simi-
larity it has been classified as a potential regulator
of transcription. The function of BRD1 is not yet
understood.
The genetic deletion syndrome, velo-cardio-facial
(VCFS; DiGeorge) syndrome, is known to be associ-
ated with schizophrenia in infrequent cases [419, 426–
430]. The deletions on 22q11 include the catechol-O-
methyl transferase (COMT) gene as well as 26 other
248
14q32.33 Figure 19.10 Regions of chromosome 14
(AKT1) showing the position of the schizophrenia
associated genes NPAS3 and AKT1.
NPAS3 in the 14q13.1 region was
disrupted by balanced reciprocal
translocation in a family with
schizophrenia.
22q13.33 Figure 19.11 Regions of chromosome 22
(BRD1) showing linkage to schizophrenia and the
position of the schizophrenia associated
genes PRODH, COMT, and BRD1. The 3 Mb
typically deleted velo-cardio-facial
syndrome (VCFS) region is also shown.
genes including proline dehydrogenase (PRODH2)
[343]. In the case of PRODH2 evidence of association
with schizophrenia has been published [431, 432] but
several studies were negative [433–436]. A mouse
with a mutation in PRODH2, causing hyperprolinae-
mia, had a deficit in sensorimotor gating accompan-
ied by regional neurochemical alterations in the brain
[437]. Mice genetically changed to have PRODH2 defi-
ciency had increased neurotransmitter release at gluta-
matergic synapses and deficits in associative learning
as well as responsiveness to psychotomimetic drugs
[438]. The PRODH2 gene encodes for proline dehy-
drogenase, a mitochondrial enzyme that converts pro-
line to 1-pyrroline-5-carboxylate and is involved in
transfer of redox potential across the mitochondrial
membrane. It is widely expressed in the human brain.
COMT inactivates catecholamines at postsynaptic
sites in the brain and as such is an ideal candidate
gene for schizophrenia. The most widely studied
COMT polymorphism is rs4680 which codes for a
functional variant at codon 108/158 that results in a
change from a valine to a methionine. The valine
form of COMT has significantly lower enzyme activ-
ity than the methionine [439]. Meta-analyses of pub-
lished studies initially concluded that the valine allele
was associated with schizophrenia [440]. However,
more recent analyses including more stringent quality
control has indicated that this finding may not be
robust [441]. A meta-analysis of two additional COMT
polymorphisms that were reported with highly sig-
nificant association p-values in an Ashkenazi Jewish
sample of schizophrenics [442] were later shown to
have modest association in a meta-analysis [261]
(Figure 19.11).

Cytogenetic abnormalities, deletions
and duplications in schizophrenia
Linkage markers mapping to a favored locus, impli-
cated by co-segregation of a familial genetic disease
with schizophrenia or by cytogenetic abnormalities in
rare cases of schizophrenia, have been used to choose
markers in linkage studies of schizophrenia. Cytogen-
etic abnormalities reported in combination with schizo-
phrenic psychoses include fragile sites [443–445],
deletions, pericentric inversions [446, 447], trisomies
[240, 448, 449], and acentric fragments [450, 451].
Many different genetic deletions and duplications,
also known as CNVs have recently been found to
contribute to the genetic etiology of schizophrenia.
Two general types of microarrays have been used.
One microarray uses the genotyping of SNPs with
special markers incorporated for quantitative assess-
ment of CNV fluorescent intensity differences.
The other uses fixed short oligonucleotides or fixed
genomic DNA clones to hybridize with DNA from a
patient in comparison to DNA from a control subject,
a technique known as comparative genomic hybrid-
ization (CGH). The genotypes and fluorescent signal
intensities are then analyzed to detect insertions and
deletions of DNA. The CNVs cause disease by either
releasing the effect of an unexpressed recessive muta-
tion in the homologous gene on the nondeleted piece
of chromosome (hemizygosity) or by a direct effect of
the CNV on the function of the gene. Because of the
low frequency of deletions it is thought that it is the
actual gene dosage effect from the deletion of a gene
causing hemizygosity which is causing schizophrenia.
In some samples, between 10 and 20% of schizophre-
nia cases were found to be positive for these types of
mutation. At present the current methodology used to
find CNVs is only successful in identifying abnormal-
ities above 50 kb (5000 base pairs). The results of two
large recent CNV studies of schizophrenia revealed
remarkably consistent results. The ISC found that
CNVs of 100 kb or more were 1.15 times more likely
to appear in patients compared to controls [343].
Further, 2 new regions were identified as having dele-
tions of 0.5 Mb or more on chromosome 15q13.1 and
chromosome 1q21.1 (Figure 19.1). Additional dele-
tions in patients with typical schizophrenia without
VCFS were identified in the critical 22q11.2 VCFS
region, as well as in the 16p12.2–p12.2 and 12p11.23
regions. The same two regions (15q13.3 and 1q21.1)
were also identified as having deletions in the SGENE
Chapter 19: Genetics of schizophrenia
collaboration [452]. This group also found an associ-
ation in the 22q11.2 region with a deletion in 0.2%
of cases and absent in any controls. An additional new
deletion was also found on chromosome 1q21.1.
Micro-deletions and micro-duplications of greater
than 0.1 Mb were also found in 15% of patients
suffering with schizophrenia but in only 5% of con-
trols [453]. When the phenotype was altered to
include only patients with early-onset disease the gen-
etic abnormalities were then found in 20% of cases.
Encouragingly, some of the regions identified as
having CNVs using the CGH technology were repli-
cated by the GWAS studies [453–455]. A CNV study
by Xu et al. [455] showed that many CNV mutations
were de novo and not transmitted from parents.
Confirmed de novo mutations were significantly asso-
ciated with schizophrenia (p ¼ 0.00078) and were
8 times greater in cases than in unaffected controls.
The discovery that some CNVs arise de novo in some
individuals with schizophrenia was also strongly sup-
ported in the SGENE collaboration where many of
the 66 de novo CNV mutations found were associated
with schizophrenia using a case-control design [452].
A further study of 2977 schizophrenia cases found a
16-fold increase in exonic deletions of the Neurexin 1
(NRXN1) gene in cases compared to controls [456].
A strong overlap between deletions and duplications
found in idiopathic generalized epilepsy and schizo-
phrenia can be observed when comparing the ISC
study [343] and CNV studies of epilepsy based in
Germany [457, 458] (Table 19.1). The chromosome
15q13.1–13.3 deletions were found in both diseases
with equal frequency and only 1 control out of about
6000 had a deletion. The chromosome 22q11.2 dele-
tions likewise were strongly associated with schizo-
phrenia and epilepsy with no affected controls. This
deletion was commoner in schizophrenia (13 cases)
than in epilepsy (2 cases). The overlap between
schizophrenia and epilepsy shows that CNVs can
have pleiotropic effects on the clinical phenotype.
It also shows that the two disorders can be caused
by a single genetic abnormality. The specific CNVs
overlapping in schizophrenia and epilepsy are shown
in Table 19.1. Although cases with both schizophrenia
and epilepsy are not common the presence of the
same CNVs in both disorders may explain why there
is a complex relationship between the two disorders
when they are comorbid, for example when there
is “forced normalization”. When one contrasts the
high yield of positive results for CNVs in individual
249
 Chapter 19: Genetics of schizophrenia
Table 19.1 Number of research subjects (N) with deletions and duplications shared by schizophrenia (SCZ) and idiopathic
generalised epilepsy (IGE) [344, 458–461].
Chromosome Position (Mb) Size
(Mb)
1q21.1 145.0–146.35 1.35
15q13.3 28.7–30.3 1.5
16p11.2 27.05–45.6 5.6
16p13.1 29.5–30.1 0.6
22q11.2 17.5–20.5 3.0
All
cases of schizophrenia compared to the negligible
findings for MRI or CT scans of schizophrenics
it seems that clinical investigation with microarrays
to detect CNVs will be a much more productive clin-
ical investigation than MRI scans both for diagnosis
and as a guide to more personalized drug and behav-
ioral treatments.
Genetic association and CNV
differences between schizophrenia
and bipolar disorder
Comparison of linkage hot spots from family linkage
studies for bipolar disorder and schizophrenia in
meta-analyses found little overlap for the most sig-
nificantly linked chromosome regions [191, 461–463].
Genetic association studies have been given a big
boost from microarray technology and have enabled
most of the human genes to be investigated in large
samples of schizophrenics and bipolar cases. When a
study of 3322 cases of schizophrenia [250] and 4387
cases of bipolar disorder [464] were compared for the
top 200 most strongly associated SNP markers in each
study only one gene (protocadherin 10, PCDH10) was
genetically implicated in both schizophrenia and
bipolar disorder. CNVs and regions of homozygosity
have also been studied in schizophrenia and bipolar
disorder. Lencz et al. [465] and Vine [466] studied
homozygosity in schizophrenia and bipolar cohorts.
Findings were discrepant with no overlaps between
schizophrenia and bipolar disorder. A CNV study of
bipolar disorder [467] employing microarray methods
250
Total IGE/ CNV CNV CNV Genes affected
SCZ (N) IGE (N) SCZ (N) controls (N)
11 1 10 2 ACP6, GJA5, BCL9
41 31 10 1 TRPM1, CHRNA7
14 1 13 2 DOC2A, KIF22,
MAPK3, etc
43 29 14 2 ABCC6, KIAA0430,
ABCC1, etc
15 2 13 0 COMT, SNAP29,
PRODH, etc
124 64 60 7
able to detect CNVs found that the rate of singleton
CNVs in bipolar disorder was 16.2% compared to
12.3% in controls (p ¼ 0.007). This study however
found no overlap of CNVs between schizophrenia
and bipolar disorder. Two further studies found no
evidence that bipolar disorder was caused by the
larger sized CNVs above 100 kb and both studies even
found a slight excess of deletions and duplications
in the controls when compared to the bipolar cases
[468, 469].
A further study of schizophrenia claims that a third
of the genetic variance of schizophrenia and bipolar
disorder is shared and that it is polygenic. The defin-
ition of polygenic transmission is that two or more
disease mutations or susceptibility variants at two or
more distinct gene loci need to be inherited by one
person for them to become affected. In this study
genotype relative risk was used to generate “polygenic”
scores without any attempt to determine whether
individuals actually had inherited two disease alleles.
Therefore the use of the term “polygenic” was a mis-
nomer and transmission should have been described
as being multigenic and heterogeneous rather than
polygenic. The loci which contributed to the genotype
relative risk “polygenic score” were not localized in
terms of their chromosomal positions. These genetic
markers could therefore have clustered around rela-
tively few susceptibility loci and not necessarily at
“thousands” of loci as stated in the paper. If different
mutations occurred in the same genes with one set
causing bipolar disorder but a different set causing
schizophrenia then similar multigenic heterogeneous
scores for these regions would have been computed

and this would obscure the fact that different muta-
tions were responsible for the two disorders. None of
the heterogeneous multigenic loci showed genome-
wide levels of association with schizophrenia or bipolar
disorder. Paradoxically markers showing an absent or
lower level of allelic association with schizophrenia
contributed more to the overall “polygenic” score.
The polygenic score was absent or very much reduced
in an African-American schizophrenia sub-sample
and this was explained by greater recombination in a
population of African descent. Another possibility is
whether the polygenic score might actually be a very
powerful test of hidden genetic stratification between
cases and controls within each sample. This might have
been created by the tendency of psychotic individuals
to migrate more than controls. On the other hand, the
markers identified as contributing to the polygenic
score were more often within genes than in gene
deserts making the hypothesis of genetic drift caused
by schizophrenia disease alleles unlikely.
The claimed sharing of NRG1 association in both
schizophrenia and bipolar disorder also has compli-
cations when the details are observed. A study of
bipolar disorder [470] found nominal allele-wise sig-
nificant association for rs35753505 with the T allele
being overrepresented in cases. This is the opposite
allelic association to the original association study
where the C allele was associated with schizophrenia.
In another attempt to implicate the NRG1 gene in
bipolar disorder [471], there was little association
using the standard BP or mood-incongruent psych-
otic BP phenotypes and none of the associations
withstood correction for multiple maker testing.
For the DAOA (G72) locus there is evidence for
association with both bipolar disorder and schizo-
phrenia [396, 472–475]. In one of these studies the
same marker and allele (rs3918342 allele T) showed
association with both disorders [396]. There is also
modest evidence for association with bipolar disorder
at the ZNF804A locus [228]. The presence of a pro-
portion of misdiagnosed schizoaffective bipolar males
in the schizophrenia group might be an explanation
for associations that cross-over between schizophre-
nia and bipolar disorder.
Functional polymorphisms and
mutations in other candidate genes
Significant association between a neurotrophin 3 gene
polymorphism and schizophrenia has been reported
Chapter 19: Genetics of schizophrenia
[476–481]. Jones et al. [482] found a family with a
b-amyloid precursor protein mutation that was pre-
sent in the genes of a schizophrenic with abnormal
MRI and EEG findings who might have a form of
angiopathy and dementia causing the schizophrenia.
Other research groups have screened for this muta-
tion in association samples of schizophrenics and
have not been able to find more cases carrying this
mutation [235, 482–490].
Pharmacogenetics of schizophrenia
This field is still in its infancy. Markers at the 5-HT2a
receptor gene have been associated with influencing
response to clozapine [491–493]. This result remains
poorly replicated despite being offered commercially
as a clinical test [494].
Animal and cell biology studies of
suscptibility genes in schizophrenia
DISC1
Disrupted in schizophrenia 1 (DISC1) was identified
as having truncated gene in a large Scottish family. In
the mouse a deletion in Disc1 was found in the 129S6/
SvEv strain but it had a normal phenotype. When the
deletion variant was transferred to C57BL/6J it was
found to induce impairment of working memory and
prepulse inhibition [495]. However, all mice were
subsequently found to have this deletion [496]. An
N-ethyl-N-nitrosourea (ENU) induced mutation in
Exon 2 of mouse Disc1 showed depressive-like behav-
ior with deficits in the forced swim test that were
reversed by an antidepressant. A second mutation
exhibited schizophrenic-like behavior, with obvious
deficits in prepulse inhibition and latent inhibition
which could be reversed by antipsychotic treatment
[497]. Disc1 transgenic mice expressing two copies of
the truncated Disc1 gene which encodes the first eight
exons were found to have enlarged lateral ventricles,
reduced cerebral cortex, partial agenesis of the corpus
callosum, and thinning of layers II/III with reduced
neural proliferation during neurogenesis. Parval-
bumin GABAergic neurons were reduced in the
hippocampus and medial prefrontal cortex, and dis-
placed in the dorsolateral frontal cortex. In culture,
transgenic neurons grew fewer with shorter neurites.
Transgenic mice showed immobility and reduced
vocalization in depression-related tests as well as
impaired conditioning of latent inhibition [498].
251
 Chapter 19: Genetics of schizophrenia
PPP3C
Knockout mice engineered to be deficient in calci-
neurin CaNB1 subunit were found to have deficits in
working memory which is an endophenotype found in
some cases of schizophrenia [499]. A follow-up study
by the same group identified further schizophrenia
related behaviors in these mice such as increased loco-
motor activity and pre-pulse inhibition and concluded
that further investigation of the role of calcineurin in
schizophrenia should be carried out [500].
SLC1A3
SLC1A3 is a gene encoding a member of a family of
high-affinity sodium-dependent transport proteins
that regulate neurotransmitter concentrations at the
excitatory glutamatergic synapses of the mammalian
central nervous system. The SLC1A3 mutant mouse
exhibits behavioral abnormalities consistent with
positive symptoms of schizophrenia [501]. These
symptoms can be reversed in the SLCA13 mouse
model by treatment with haloperidol, the mGlu2/3
agonist, LY379268 or the mGlu2/3 agonist, LY379268.
SLCA13 knockout (KO) mice had poor nesting
behavior and impaired sociability. The knockout
heterozygote was also affected but with lowered pref-
erence for novel social stimuli. Homozygous knock-
outs exhibited a significantly reduced acoustic startle
response and had impaired learning in an instrumen-
tal visual discrimination task [501].
CHNRA7
The region of the gene encoding the central alpha7-
Nicotinic acetylcholine receptor (CHNRA7 or alpha7-
nAChR) has been implicated repeatedly in a range of
cognitive deficits in schizophrenia. CHNRA7 homo-
zygous and heterozygous knockout mice exhibited
significantly higher attentional load omissions and
had impaired odor span ability [502]. Attentional
impairment appeared to be central to the cognitive
deficits observed [502].
NRG1
In the mouse type III Nrg1 is transcribed by a different
promoter than the other Nrg1 isoforms. It is expressed
in the medial prefrontal cortex, ventral hippocampus,
and ventral subiculum. Adult heterozygous mutant
mice with disrupted type III Nrg1 have enlarged lat-
eral ventricles and decreased dendritic spine density
252
in subicular pyramidal neurons [503]. MRI imaging
of heterozygous mice showed hypofunction in the
medial prefrontal cortex, the hippocampal CA1, and
subiculum regions. Type III Nrg1 heterozygous mice
showed deficits for delayed alternation memory tasks
and had deficits in prepulse inhibition. Chronic nico-
tine treatment eliminated differences in between type
III Nrg1 heterozygous mice and wild-type mice [503].
In another animal model of NRG1 schizophrenia
it was found that locomotor activity in the open field
or in photocell cages was significantly enhanced in
Nrg1þ/– mice compared to wild-type littermate con-
trols. Treatment with the 5HT1A receptor agonist
8-hydroxy-dipropylaminotetralin (8-OH-DPAT) showed
a differential effect between genotypes, with a disrup-
tion of PPI occurring in Nrg1þ/– mice compared to no
effect in wild-type controls. There was also a signifi-
cant reduction of startle response [504].
STOP
Microtubule stabilisation within neurones has been
shown to be dependent on the action of stable tubule-
only polypeptide (STOP) proteins. Neurons contain
many highly stable microtubules that resist depoly-
merization after exposure to cold. Stable microtubules
have a vital role in neuronal development, as well as
their maintenance and function. Mice lacking the
STOP gene showed no obvious defects in brain anat-
omy but they had synaptic deficits, depleted synaptic
vesicle pools, and impaired synaptic plasticity
together with abnormal behavior that could be
reversed by antipsychotic drugs [505, 506]. Mutant
STOP mice showed reduced expression of synapto-
physin, VGlut1, GAP-43, and spinophilin mRNAs in
the hippocampus or cerebellum [507]. Transgenic
mice lacking the STOP gene showed hyperlocomotion
which could be reversed by clozapine.
RELN
Heterozygous reelin knockout mice have been pro-
posed as a genetic model for schizophrenia [508, 509].
Reelin is an extracellular matrix protein secreted by
GABAergic interneurons that interacts with pyram-
idal neuron integrin receptors and can alter dendritic
spine plasticity. Heterozygous reeler mice express
only about 50% of the normal amount of reelin
mRNA and protein and may mimic the dendritic
spine, GABAergic and other defects described in
schizophrenia [510, 511]. Hypermethylation of the
 Chapter 19: Genetics of schizophrenia

RELN promoter has been proposed as the cause of
low Reelin mRNA levels in postmortem human tissue
from schizophrenics [512, 513]
GRM3
mGluR3 receptor knockout mice were given the
mGluR2/3 agonist, LY379268 [514]. Phencyclidine
evoked hyperactivity, circling, falling, stereotypy,
and ataxia, as well as amphetamine-induced hyper-
activity were reduced by LY379268. LY379268
reversed phencyclidine induced hyperactivity and
behavioral alterations in wild-type and mGluR3
knockout mice but not in mice lacking mGluR2.
Conclusions
The presence of a strong genetic predisposition is by
far the most well-confirmed etiological effect in
schizophrenia. The most striking aspect of the new
knowledge is that etiological genetic heterogeneity is
very considerable with no single genetic effect being
present in more than a few percent of cases. Despite
this linkage and association results were nevertheless
successful with linkage hot spots predicting where
genes would be fine mapped by allelic LD. There are
now plausible etiological base pair changes that
explain some of the linkage and LD data. Twin and
adoption studies have shown that the family envir-
onment has no influence on the etiology of schizo-
phrenia. Many specific environmental factors that
interact with the etiological base pair changes have
been proposed. These have been reviewed recently
[515]. None have reached the status of being com-
pelling. Viruses and immunological factors need to
be reappraised given that the HLA region is now well
implicated in schizophrenia. However there are
many nonimmune-related genes within the HLA
regions. Cannabis use in schizophrenics has not been
shown to be independent of genetic factors as shown
by equally positive family histories of schizophrenia for
probands who have so-called cannabis psychosis to
those that have schizophrenia uncomplicated by
cannabis smoking [516–518]. One study in genetically
high-risk subjects found no increased cannabis use
associated with development of schizophrenia [519].
Life events as a precipitating factor are no longer being
studied largely because life events are likely to be an
epiphenomena of the schizophrenia susceptibility
genes rather having any causal effect. Attention has
now shifted towards methylation and other epigenetic
effects to explain the obvious variable penetrance
and pleiotropism of the clinical symptoms. Stochastic
biological processes during central nervous system
development alone could explain the variable pene-
trance. Despite many genes having been implicated
the identification of clinical subtypes of schizophrenia
related to specific genes is still in its infancy. DNA
sequencing of etiological base pair changes will solve
this problem and finding them will lead to new indi-
vidualized treatments based on genetic tests. There is
scope for narrow spectrum drugs based on mutations
in an individual and medium spectrum drugs based on
which systems are genetically perturbed and also new
broad spectrum drugs to take over from drugs like
clozapine without the side effect of agranulocytosis.
Although systems biology can generate classifications
of etiological genetic effects on schizophrenia based on
evidence of genome-wide association data these studies
would best be conducted using the etiological base pair
changes discovered by DNA sequencing. The genes
implicated in this review could loosely be classified
into those that are synaptic (NRG1, DAOA, GRM3,
DRD2, SYN2, FXYD6, CLINT1, DTNBP1) and those
that are microtubule related (DISC1, PCM1, UHMK1,
RELN, STOP, NPAS3, KIF2A). Once disease mutations
in these genes are well established, then it will be
possible to identify the true modes of transmission as
well as any gene–environment interactions. Postmor-
tem, psychological, neurophysiological, imaging, and
EEG measures will need to be studied afresh in the
light of the discovery of the actual etiological base
pair changes.

References 4. Neale MC, et al. Am J Hum Genet

1. MacLean CJ, et al. Genet Epidemiol
1990;7:419–426.
2. Chen WJ, et al. Psychiatr Genet
1992;2:219–238.
3. Kendler KS, et al. Genet Epidemiol
1990;7:409–417.
1989;45:226–239.
5. Penrose LS. Eur Arch Psychiatry
Clin Neurosci 1991;240:314–324.
6. Bassett AS, et al. Am J Hum Genet
1997;60:630–637.
7. Bassett AS, et al. Am J Hum Genet
1994;54:864–870.
8. Asherson P, et al. Br J Psychiatry
1994;164:619–624.
9. Beckmann H, et al. Am J Med
Genet 1996;67:289–300.
10. Chotai J, et al. Psychiatr Genet
2005;5:181–186.
11. Blinc-Pesek M, et al. Br J
Psychiatry 2007;191:181.

253
 Chapter 19: Genetics of schizophrenia
12. Di Maggio C, et al. Am J
Psychiatry 2001;158:489–492.
13. Laurent C, et al. Am J Med Genet
1998;81:342–346.
14. Thibaut F, et al. Psychiatry Res
1995;59:25–33.
15. Imamura A, et al. J Hum Genet
1998;43:217–223.
16. Ohara K, et al. Biol Psychiatry
1997;42:760–766.
17. Heiden A, et al. Schizophr Res
1999;35:25–32.
18. Merette C, et al. Am J Med Genet
2000;96:61–68.
19. Husted J, et al. Am J Med Genet
1998;81:156–162.
20. McInnis MG, et al. Am J Med
Genet 1999;88:686–693.
21. Cavalli-Sforza LL, et al. Monogr
Popul Biol 1981;16:1–388.
22. Henderson N. Annual Review of
Psychology 1982;33:403–440.
23. Jakubaschk J, et al. Psychiatr Clin
(Basel) 1979;12:80–91.
24. Tsuang MT, et al. Arch Gen
Psychiatry 1981;38:535–539.
25. Gottesman II, et al. Schizophrenia:
The Epigenetic Puzzle. Cambridge,
UK: Cambridge University Press;
1982.
26. Guze SB, et al. Arch Gen
Psychiatry 1983;40:1273–1276.
27. Kendler KS, et al. Arch Gen
Psychiatry 1984;41:555–564.
28. Baron M, et al. Am J Psychiatry
1985;142:447–455.
29. Kendler KS, et al. Arch Gen
Psychiatry 1985;42:770–779.
30. Gershon ES, et al. Arch Gen
Psychiatry 1988;45:328–336.
31. Kendler KS, et al. Am J Psychiatry
1988;145:57–62.
32. Onstad S, et al. Acta Psychiatr
Scand 1991;83:463–467.
33. Kendler KS, et al. Arch Gen
Psychiatry 1993;50:527–540.
34. Maier W, et al. Schizophr Res
1993;9:71–76.
254
35. DeLisi LE, et al. Arch Gen
Psychiatry 1987;44:891–896.
36. Slater E, et al. The Genetics of
Mental Disorders. London: Oxford
University Press; 1971.
37. Leboyer M, et al. Psychiatry Res
1992;41:107–114.
38. Tsuang MT, et al. Arch Gen
Psychiatry 1974;31:43–47.
39. Scharfetter C, et al. Schizophr Bull
1980;6:586–591.
40. Stober G, et al. Eur Arch
Psychiatry Clin Neurosci
1995;245:135–141.
41. Winokur G, et al. J Nerv Ment
Disord 1974;159:12–19.
42. McGuffin P, et al. Arch Gen
Psychiatry 1984;41:541–545.
43. Kendler KS, et al. Arch Gen
Psychiatry 1994;51:456–468.
44. Bassett AS, et al. Schizophr Res
1994;12:213–221.
45. Liddle PF, et al. Br J Psychiatry
1990;157:558–561.
46. Keefe RS, et al. Am J Psychiatry
1987;144:889–895.
47. Roy MA, et al. Am J Psychiatry
1994;151:805–814.
48. Roy MA, et al. Psychiatry Res
1994;54:25–36.
49. Maier W, et al. Arch Gen
Psychiatry 1993;50:871–883.
50. Kendler KS, et al. Arch Gen
Psychiatry 1983;40:951–955.
51. Kendler KS, et al. Arch Gen
Psychiatry 1993;50:952–960.
52. Kendler KS, et al. Arch Gen
Psychiatry 1993;50:781–788.
53. Kendler KS, et al. Arch Gen
Psychiatry 1993;50:645–652.
54. Lapierre YD. Can J Psychiatry
1994;39:S59–S64.
55. Craddock N, et al. Br J Psychiatry
2005;186:364–366.
56. Van Snellenberg JX, et al. Arch
Gen Psychiatry 2009;66:748–755.
57. Lichtenstein P, et al. Lancet
2009;373:234–239.
58. McDonald C, et al. Am
J Psychiatry 2006;163:478–487.
59. Maj M, et al. J Affect Disord
1990;20:71–77.
60. Maj M, et al. J Affect Disord
2000;57:95–98.
61. Maj M, et al. Am J Psychiatry
1991;148:612–616.
62. Maj M, Acta Psychiatr Scand
1985;72:542–550.
63. Spitzer R, et al. The Schedule for
Affective Disorders and
Schizophrenia, Lifetime Version,
3rd edn. New York: New York
State Psychiatric Institute; 1977.
64. Baron M, et al. Acta Psychiatr
Scand 1982;65:253–262.
65. Mendlewicz J, et al. J Affect Disord
1980;2:289–302.
66. Lewis S. Br J Psychiatry 1991;
159:158–159.
67. Weinberger DR, et al. Psychiatry
Res 1981;4:65–71.
68. DeLisi LE, et al. Arch Gen
Psychiatry 1986;43:148–153.
69. Pearlson GD, et al. J Nerv Ment
Disord 1985;173:42–50.
70. Oxenstierna G, et al. Br J
Psychiatry 1984;144:654–661.
71. Reveley AM, et al. Br J Psychiatry
1984;144:89–93.
72. Turner SW, et al. Psychol Med
1986;16:219–225.
73. Nasrallah HA, et al. J Clin
Psychiatry 1983;44:407–409.
74. Owen MJ, et al. Br J Psychiatry
1989;154:629–634.
75. Orlova VA, et al. Zh Nevrol
Psikhiatr Im S S Korsakova
1994;94:85–90.
76. Palmour RM, et al. J Psychiatry
Neurosci 1994;19:270–277.
77. Lui S, et al. Psychiatry Res 2009;
171:71–81.
78. Blackwood DH, et al. Arch Gen
Psychiatry 1991;48:899–909.
79. Holzman PS. Schizophr Res
1994;13:1–9.

80. Blackwood DH, et al. Br
J Psychiatry Suppl, 1996:85–92.
81. Holzman PS, et al. Arch Gen
Psychiatry 1988;45:641–647.
82. Holzman PS, et al. Arch Gen
Psychiatry 1984;41:136–139.
83. Siegel C, et al. Arch Gen Psychiatry
1984;41:607–612.
84. St Clair D, et al. J Psychiatr Res
1989;23:49–55.
85. Waldo MC, et al. Schizophr Res
1988;1:19–24.
86. Waldo MC, et al. Biol Psychiatry
2000;47:231–239.
87. Waldo MC, et al. Psychiatry Res
1991;39:257–268.
88. Waldo MC, et al. Schizophr Res
1994;12:93–106.
89. Cannon TD, et al. Arch Gen
Psychiatry 1994;51:955–962.
90. Clementz BA, et al. J Abnorm
Psychol 1994;103:400–403.
91. Docherty NM. J Nerv Ment Dis
1993;181:750–756.
92. Docherty NM. J Nerv Ment Dis
1994;182:443–451.
93. Hollister JM, et al. Arch Gen
Psychiatry 1994;51:552–558.
94. Frodl T, et al. Int J Psychophysiol
2002;43:237–246.
95. Yeap S. et al. Arch Gen Psychiatry
2006;63:1180–1188.
96. Woodward ND, et al. Schizophr
Res 2007;94:306–316.
97. Woodward ND, et al. Schizophr
Res 2009;109:182–190.
98. Dauphinais ID, et al. Psychiatry
Res 1990;35:137–147.
99. Suddath RL, et al. N Engl J Med
1990;322:789–794.
100. Alfimova M, et al. Int
J Psychophysiol 2003;49:201–216.
101. Alfimova MV, et al. Span J Psychol
2009;12:46–55.
102. Calkins ME, et al. Brain Cogn
2008;68:436–461.
103. Callicott JH, et al. Biol Psychiatry
1998;44:941–950.
Chapter 19: Genetics of schizophrenia
104. Curtis CE, et al. Am J Psychiatry
2001;158:100–106.
105. Guerra S, et al. Brain Cogn
2009;70:221–230.
106. Kremen WS, et al. Am J Med
Genet B Neuropsychiatr Genet
2007;144B:403–406.
107. Leppanen JM, et al. Schizophr Res
2008;99;270–273.
108. Quinones RM, et al. Eur Arch
Psychiatry Clin Neurosci
2009;259:475–481.
109. Schulze KK, et al. Bipolar Disord
2008;10:377–386.
110. Takahashi S, et al. Psychiatry Clin
Neurosci 2008;62:487–493.
111. Turetsky BI, et al. Biol Psychiatry
2008;64:1051–1059.
112. Turetsky BI, et al. Schizophr Res
2008;102:220–229.
113. Braff DL, et al. Schizophr Bull
2007;33:21–32.
114. Braff DL, et al. World Psychiatry
2008;7:11–18.
115. Gottesman II, et al. Schizophrenia
and Genetics: A Twin Study
Vantage Point. Cambridge, UK:
Academic Press; 1972.
116. Farmer AE, et al. Arch Gen
Psychiatry 1987;44:634–641.
117. Onstad S, et al. Acta Psychiatr
Scand 1991;83:395–401.
118. Torgersen S, et al. Am J Psychiatry
1993;150:1661–1667.
119. Fulker DW. Soc Biol 1973;20:
266–275.
120. McGue M, et al. Am J Hum Genet
1983;35:1161–1178.
121. McGuffin P. Psychiatr Prax
2004;31 (Suppl 2):S189–S193.
122. Reveley MA. Br J Psychiatry
1985;146:367–371.
123. Dalton SO, et al. Am J Psychiatry
2004;161:903–908.
124. Reveley AM, et al. Psychiatry Res
1984;13:261–266.
125. Reveley AM, et al. Lancet
1982;1:540–541.
126. Casanova MF, et al. J Neurol
Neurosurg Psychiatry
1990;53:416–421.
127. Casanova MF, et al. Schizophr Res
1990;3:155–156.
128. Casanova MF, et al. Biol
Psychiatry 1990;28:83–84.
129. Goldberg TE, et al. Arch Gen
Psychiatry 1990;47:1066–1072.
130. Goldberg TE, et al. Psychiatry Res
1994;55:51–61.
131. Goldberg TE, et al. Schizophr Res
1995;17:77–84.
132. Goldberg TE, et al. Psychol Med
1993;23:71–85.
133. Kremen WS, et al. J Clin Exp
Neuropsychol 2006;28:208–224.
134. Lyons MJ, et al. Twin Res
2000;3:28–32.
135. Ragland JD, et al. Schizophr Res
1992;7:177–183.
136. Cantor-Graae E, et al. J Nerv Ment
Dis 1994;182:645–650.
137. McNeil TF, et al. Acta Psychiatr
Scand 1994;89:196–204.
138. McNeil TF, et al. Am J Psychiatry
2000;157:203–212.
139. Hulshoff Pol HE, et al.
Neuroimage 2006;31:482–488.
140. van Erp TG, et al. Arch Gen
Psychiatry 2004;61:346–353.
141. van Erp TG, et al. Psychiatry Res
2008;159:271–280.
142. van Haren NE, et al. Biol
Psychiatry 2004;56:454–461.
143. Brans RG, et al. Arch Gen
Psychiatry 2008;65:1259–1268.
144. Sommer IE, et al. Br J Psychiatry
2004;184:128–135.
145. Spaniel F, et al. Eur Psychiatry
2007;22:319–322.
146. Ettinger U, et al. Arch Gen
Psychiatry 2007;64:401–409.
147. Borgwardt SJ, et al. Biol Psychiatry
2010;67:956–964.
148. Fusar-Poli P, et al. Neurosci
Biobehav Rev 2007;31:
465–484.
255
 Chapter 19: Genetics of schizophrenia
149. Baare WF, et al. V Arch Gen
Psychiatry 2001;58:33–40.
150. Hulshoff Pol HE, et al. Biol
Psychiatry 2004;55:126–130.
151. Hulshoff Pol HE, et al.
Am J Psychiatry 2002;159:
244–250.
152. Bartley AJ, et al. Biol Psychiatry
1993;34:853–863.
153. Weinberger DR, et al. Clin
Neuropharmacol 1992;15
(Suppl 1 Pt A):122A–123A.
154. Noga JT, et al. Schizophr Res
1996;22:27–40.
155. Mill J, et al. Am J Hum Genet
2008;82:696–711.
156. Kaminsky ZA, et al. Nat Genet
2009;41:240–245.
157. Rosenthal D, et al. Am J Psychiatry
1971;128:307–311.
158. Kety SS, et al. The Transmission of
Schizophrenia. Oxford, UK:
Pergamon Press; 1969.
159. Kety SS, et al. Am J Psychiatry
1971;128:302–306.
160. Kety SS, et al. Arch Gen Psychiatry
1994;51:442–455.
161. Ingraham LJ, et al. Am J Med
Genet 2000;97:18–22.
162. Cardno AG, et al. Am J Psychiatry
2002;159:539–545.
163. Tienari P, et al. Acta Psychiatr
Scand Suppl 1985;319:19–30.
164. Tienari P, et al. Br J Psychiatry
Suppl 2004;184:216–222.
165. Rao DC, et al. Hum Hered
1981;31:325–333.
166. Kendler KS. Am J Psychiatry
1983;140:1413–1425.
167. McGue M, et al. Genet Epidemiol
1985;2:99–110.
168. McGuffin P, et al. Br J Psychiatry
1994;164:593–599.
169. Baron M. Biol Psychiatry
1986;21:1051–1066.
170. Book JA. Acta Genetica 1953;
4:345–414.
256
171. Slater E. Acta Genetica et Statistica
Medica 1958;8:50–56.
172. Garrone G. J Hum Genet 1962;
11:89–100.
173. Heston LL. Science 1970;167:
249–256.
174. Elston RC, et al. Behav Genet
1970;1:3–10.
175. Karlsson J. Br J Psychiatry
1988;152:324–329.
176. Kay DWK, et al. Br J Psychiatry
1975;127:109–118.
177. Elston RC, et al. Neuropsychobiology
1978;4:193–206.
178. O’Rourke DH, et al. Am J Hum
Genet 1982;34:630–649.
179. Tsuang MT, et al. Br J Psychiatry
1982;140:595–599.
180. Risch N, et al. Am J Hum Genet
1984;36:1039–1059.
181. Gottesman II, et al. Proc Natl Acad
Science U S A 1967;58:199–205.
182. Tsuang MT, et al. Br J Psychiatry
1983;143:572–577.
183. Ungvari G. Acta Psychiatr Scand
1983;68:287–296.
184. Matthysse SW, et al. Am
J Psychiatry 1976;133:185–191.
185. Baron M. Acta Psychiatr Scand
1982;65:263–275.
186. Carter CL, et al. Hum Hered
1980;30:350–356.
187. Bassett AS. Br J Psychiatry
1992;161:323–334.
188. Bassett AS, et al. Am J Med Genet
2000;97:45–51.
189. Kosower NS, et al. Am J Med
Genet 1995;60:133–138.
190. Brzustowicz LM, et al. Science
2000;288:678–682.
191. Blaveri E, et al. Eur J Hum Genet
2001;9:469–472.
192. Lewis CM, et al. Am J Hum Genet
2003;73:34–48.
193. Hovatta I, et al. Am J Hum Genet
1999;65:1114–1124.
194. Ekelund J, et al. Mol Psychiatry
2004;9:1037–1041.
195. Ekelund J, et al. Hum Mol Genet
2000;9:1049–1057.
196. Cai G, et al. Zhonghua Yi Xue
Yi Chuan Xue Za Zhi 2002;
19:491–494.
197. Shaw SH, et al. Am J Med Genet
1998;81:364–376.
198. Brzustowicz LM, et al. Am J Hum
Genet 2004;74:1057–1063.
199. Zheng Y, et al. Biochem
Biophys Res Commun
2005;328:809–815.
200. Puri V, et al. Biol Psychiatry
2006;59:195–197.
201. Rizig MA, et al. Am J Med Genet
2006;141B:296–300.
202. Talkowski ME, et al. Biol
Psychiatry 2006;60:152–162.
203. Puri V, et al. Biol Psychiatry
2007;61:873–879.
204. Miranda A, et al. Schizophr Res
2006;82:283–285.
205. Addington AM, et al. Mol
Psychiatry 2007;12:195–205.
206. Puri V, et al. Eur J Hum Genet
2008;16:1275–1282.
207. Millar JK, et al. Genomics 2000;
67:69–77.
208. St Clair D, et al. Lancet 1990;
336:13–16.
209. Chubb JE, et al. Mol Psychiatry
2008;13:36–64.
210. Hennah W, et al. Hum Mol Genet
2003;12:3151–3159.
211. Cannon TD. Curr Opin Psychiatry
2005;18:135–140.
212. Callicott JH, et al. Proc Natl
Acad Sci U S A 2005;102:
8627–8632.
213. Hodgkinson CA, et al. Am J Hum
Genet 2004;75:862–872.
214. Maeda K, et al. Biol Psychiatry
2006;60:929–935.
215. Thomson PA, et al. Neurosci Lett
2005;389:41–45.

216. Liu YL, et al. Schizophr Res
2006;87:15–20.
217. Chen QY, et al. J Psychiatr Res
2006;41:428–434.
218. Hashimoto R, et al. Hum Mol
Genet 2006;15:3024–3033.
219. DeRosse P, et al. Biol Psychiatry
2007;61:1208–1210.
220. Kilpinen H, et al. Mol Psychiatry
2008;13:187–196.
221. Wood LS, et al. Biol Psychiatry
2007;61:1195–1199.
222. Palo OM, et al. Hum Mol Genet
2007;16:2517–2528.
223. Burdick KE, et al. Neuroreport
2005;16:1399–1402.
224. Hennah W, et al. Mol Psychiatry
2005;10:1097–1103.
225. Hennah W, et al. Schizophr Bull
2006;32:409–416.
226. O’Donovan MC, et al. Nat Genet
2008;40:1053–1055.
227. Riley B, et al. Mol Psychiatry
2010;15:29–37.
228. Steinberg S, et al. Mol Psychiatry
2011;16:59–66.
229. Crocq MA, et al. J Med Genet
1992;29:858–860.
230. Kennedy JL, et al. Am J Med Genet
1995;60:558–562.
231. Mant R, et al. Am J Med Genet
1994;54:21–26.
232. Kalsi G, et al. Psychiatr Genet
1998;8:187–189.
233. Jonsson EG, et al. Psychiatr Genet
2003;13:1–12.
234. Jonsson EG, et al. Psychiatr Genet
2004;14:9–12.
235. Nothen MM, et al. J Med Genet
1993;30:708.
236. Sabate O, et al. Am J Psychiatry
1994;151:107–111.
237. Wiese C, et al. Psychiatry Res
1993;46:69–78.
238. DiBella D, et al. Psychiatr Genet
1994;4:39–42.
239. Sherrington R, et al. Nature
1988;336:164–167.
Chapter 19: Genetics of schizophrenia
240. Bassett AS, et al. Lancet
1988;1:799–801.
241. Kalsi G, et al. Ann Hum Genet
1999;63:235–247.
242. Suarez BK, et al. Am J Hum Genet
2006;78:315–333.
243. Straub RE, et al. Mol Psychiatry
1997;2:148–155.
244. Devlin B, et al. Mol Psychiatry
2002;7:689–694.
245. Paunio T, et al. Hum Mol Genet
2001;10:3037–3048.
246. Gurling HM, et al. Am J Hum
Genet 2001;68:661–673.
247. Levinson DF, et al. Am J Hum
Genet 2000;67:652–663.
248. Schwab SG, et al. Mol Psychiatry
1997;2:156–160.
249. Pimm J, et al. Am J Hum Genet
2005;76:902–907.
250. ISC. Nature 2009;460:
748–752.
251. Liou YJ, et al. Schizophr Res
2006;84:236–243.
252. Tang RQ, et al. Mol Psychiatry
2006;11:395–399.
253. Gurling H, et al. Schizophr Res
2006;89:357–359.
254. Escamilla M, et al. Schizophr Res
2008;106:253–257.
255. Petryshen TL, et al. Mol Psychiatry
2005;10:1074–1088.
256. Ikeda M, et al. Biol Psychiatry
2005;58:440–445.
257. Lo WS, et al. Biol Psychiatry
2007;61:653–660.
258. Inada T, et al. Pharmacogenet
Genomics 2008;18:317–323.
259. Liu J, et al. Biochem Biophys Res
Commun 2005;334:817–823.
260. Yu Z, et al. Clin Biochem
2006;39:210–218.
261. Allen NC, et al. Nat Genet
2008;40:827–834.
262. McGuffin P, et al. Psychol Med
1983;13:31–43.
263. Andrew B, et al. Psychol Med
1987;17:363–370.
264. Campion D, et al. Psychiatry Res
1992;41:99–105.
265. Hwu HG, et al. Am J Med Genet
2000;96:74–78.
266. Schwab SG, et al. Nat Genet
1995;11:325–327.
267. Straub RE, et al. Nat Genet
1995;11:287–293.
268. Moises HW, et al. Nat Genet
1995;11:321–324.
269. Gurling H, et al. Nat Genet
1995;11:234–235.
270. Coon H, et al. Am J Med Genet
1994;54:59–71.
271. Straub RE, et al. Cold Spring
Harb Symp Quant Biol 1996;
61:823–833.
272. Straub RE, et al. Am J Hum Genet
2002;71:337–348.
273. van den Oord EJ, et al. Mol
Psychiatry 2003;8:499–510.
274. Mutsuddi M, et al. Am J Hum
Genet 2006;79:903–909.
275. Schwab SG, et al. Am J Hum Genet
2003;72:185–190.
276. Tang JX, et al. Mol Psychiatry
2003;8:717–718.
277. Van Den Bogaert A, et al. Am
J Hum Genet 2003;73:1438–1443.
278. Williams NM, et al. Arch Gen
Psychiatry 2004;61:336–344.
279. Morris DW, et al. Schizophr Res
2003;60:167–172.
280. Kirov G, et al. Biol Psychiatry
2004;55:971–975.
281. Funke B, et al. Am J Hum Genet
2004;75;891–898.
282. Hall D, et al. Genes Brain Behav
2004;3:240–248.
283. Kohn Y, et al. Am J Med Genet
B Neuropsychiatr Genet
2004;128:65–70.
284. Numakawa T, et al. Hum Mol
Genet 2004;13:2699–2708.
285. Li T, et al. Mol Psychiatry
2005;10:1037–1044.
286. Tochigi M, et al. Neurosci Res
2006;56:154–158.
257
 Chapter 19: Genetics of schizophrenia
287. Datta SR, et al. Behav Brain Funct
2007;3:50.
288. Benson MA, et al. J Biol Chem
2001;276:24232–24241.
289. Talbot K, et al. J Clin Invest
2004;113:1353–1363.
290. Stefansson H, et al. Nature
2009;460:744–747.
291. Duan J, et al. Am J Hum Genet
2004;75:624–638.
292. Duan S, et al. J Neural Transm
2006;113:381–385.
293. Ikeda M, et al. Schizophr Res
2005;78:127–130.
294. Knight HM, et al. Am J Hum
Genet 2009;85:833–846.
295. Jonsson EG, et al.
Neuropsychobiology 2009;
59:142–150.
296. Albalushi T, et al. Am J Med
Genet B Neuropsychiatr
Genet 2008;147:392–396.
297. Bishop JR, et al. Psychiatr Genet
2007;17:358.
298. Chen Q, et al. Schizophr Res
2005;73:21–26.
299. Egan MF, et al. Proc Natl Acad Sci
U S A 2004;101:12604–12609.
300. Fujii Y, et al. Psychiatr Genet
2003;13:71–76.
301. Harrison PJ, et al.
J Psychopharmacol 2008;
22:308–322.
302. Lyon L, et al. Synapse
2008;62:842–850.
303. Mossner R, et al. Eur
Neuropsychopharmacol 2008;
18:768–772.
304. Norton N, et al. BMC Psychiatry
2005;5:23.
305. Sartorius LJ, et al.
Neuropsychopharmacology
2008;33:2626–2634.
306. Schwab SG, et al. Psychiatr Genet
2008;18:25–30.
307. Tan HY, et al. Proc Natl
Acad Sci U S A 2007;104;
12536–12541.
258
308. Tochigi M, et al. Schizophr Res
2006;88:260–264.
309. Goldberger C, et al. Am J Med
Genet B Neuropsychiatr Genet
2005;137B:51–55.
310. Shifman S, et al. PLoS Genet
2008;4:e28.
311. Pisante A, et al. Psychiatr Genet
2009;19:212.
312. Chen ML, et al. Mol Psychiatry
2002;7:447–448.
313. Huang CH, et al. Psychiatry Res
2006;142:89–92.
314. Blouin JL, et al. Nat Genet
1998;20:70–73.
315. Brzustowicz LM, et al. Am J Hum
Genet 1999;65:1096–1103.
316. Takahashi S, et al. Psychiatry Res
2005;133:111–122.
317. Kendler KS, et al. Am J Psychiatry
1996;153:1534–1540.
318. Lin HF, et al. Psychol Med
2005;35:1589–1598.
319. Gerber DJ, et al. Proc Natl
Acad Sci U S A 2003;100:
8993–8998.
320. Gurling H, et al. Arch
Gen Psychiatry 2006;63:
844–854.
321. Katsu T, et al. Neurosci Lett
2003;353:53–56.
322. Nakata K, et al. Biol Psychiatry
2003;53:571–576.
323. Stefansson H, et al. Am J Hum
Genet 2002;71:877–892.
324. Yang JZ, et al. Biol Psychiatry
2003;54:1298–1301.
325. Yang JZ, et al. Mol Psychiatry
2003;8:706–709.
326. Stefansson H, et al. Am J Hum
Genet 2003;72:83–87.
327. Williams NM, et al.
Mol Psychiatry 2003;8:
485–487.
328. Corvin AP, et al. Mol Psychiatry
2004;9:208–213.
329. Hong CJ, et al. Neurosci Lett
2004;366:158–161.
330. Li T, et al. Am J Med Genet
B Neuropsychiatr Genet
2004;129B:13–15.
331. Bakker SC, et al. Mol Psychiatry
2004;9:1061–1063.
332. Tang JX, et al. Mol Psychiatry
2004;9:11–12.
333. Zhao X, et al. J Med Genet 2004;
41:31–34.
334. Petryshen TL, et al. Mol Psychiatry
2005;10:366–374.
335. Hashimoto R, et al. Mol Psychiatry
2004;9:299–307.
336. Iwata N, et al. Mol Psychiatry
2004;9:126–127.
337. Fallin MD, et al. Am J Hum Genet
2004;75:204–219.
338. Thiselton DL, et al. Mol Psychiatry
2004;9:777–83; image 729.
339. Datta SR, et al. Mol Psychiatry
2010;15:615–628.
340. Sullivan PF, et al. Mol Psychiatry
2008;13:570–584.
341. Need AC, et al. Hum Mol Genet
2009;18:4650–4661.
342. Kamiya A, et al. Arch Gen
Psychiatry 2008;65:996–1006.
343. ISC. Nature 2008;455:237–241.
344. Eastwood SL, et al. Mol Psychiatry
2009;14:556–557.
345. Kinoshita Y, et al.
J Neural Transm 2005;112:
1255–1262.
346. Fallin MD, et al. Am J Hum Genet
2005;77:918–936.
347. Shibasaki F, et al. J Biochem
2002;131:1–15.
348. Klee CB, et al. J Biol Chem
1998;273:13367–13370.
349. Zhang Y, et al. Am J Med Genet
B Neuropsychiatr Genet
2004;129:16–19.
350. Wei J, et al. Neurosci Lett
2004;366:336–338.
351. Ide M, et al. Biol Psychiatry
2004;56:462–465.
352. Hashimoto R, et al. J Neural
Transm 2005;112:303–307.

353. Hong LE, et al. Am J Med
Genet B Neuropsychiatr Genet
2005;136:8–11.
354. Nanko S, et al. Jpn J Psychiatry
Neurol 1992;46:155–159.
355. Maziade M, et al. Am J Med Genet
1995;60:522–528.
356. Gill M, et al. Psychol Med
1993;23:27–44.
357. Barr C, et al. Psychiatr Genet
1991;2:66–72.
358. Muir WJ, et al. Psychiatr Genet
1991;2:18.
359. Su Y, et al. Arch Gen Psychiatry
1993;50:205–211.
360. Choudhury K, et al. Am J Hum
Genet 2007;80:664–672.
361. Beguin P, et al. Embo J
2002;21:3264–3273.
362. Crambert G, et al. Sci STKE
2003;161: RE1.
363. Garty H, et al. Ann Rev Physiology
2006;68:431–459.
364. Kadowaki K, et al. Brain Res Mol
Brain Res 2004;125:105–112.
365. Ambrosio AM, et al. Psychiatry
Res 2004;125:185–191.
366. Cordeiro Q, et al. Arq
Neuropsiquiatr 2009;67:191–194.
367. Dubertret C, et al. Schizophr Res
2004;67:75–85.
368. Fan H, et al. Neurosci Lett
2009;477:53–56.
369. Glatt SJ, et al. Mol Psychiatry
2009;14:885–893.
370. Glatt SJ, et al. Mol Psychiatry
2003;8:911–915.
371. Glatt SJ, et al. Am J Med Genet
B Neuropsychiatr Genet
2006;141B:149–154.
372. Hanninen K, et al. Neurosci Lett
2006;407:195–198.
373. Kukreti R, et al. Neurosci Lett
2006;392:68–71.
374. Monakhov M, et al. Schizophr Res
2008;100:302–307.
375. Parsons MJ, et al. Psychiatr Genet
2007;17:159–163.
Chapter 19: Genetics of schizophrenia
376. Schindler KM, et al. Mol
Psychiatry 2002;7:1002–1005.
377. Wu SN, et al. Acta Pharmacol Sin
2006;27:966–970.
378. Chen CH, et al. Am J Med Genet
1996;67:418–420.
379. Crawford F, et al. Am J Med Genet
1996;67:483–484.
380. Kaneshima M, et al. Psychiatry
Clin Neurosci 1997;51:379–381.
381. Spurlock G, et al. Am J Med Genet
1998;81:24–28.
382. Jonsson EG, et al. Am J Med
Genet B Neuropsychiatr Genet
2003;119B:28–34.
383. Camp NJ, et al. Am J Hum Genet
2001;69:1278–1289.
384. Faraone SV, et al. Expert Rev Mol
Med 2002;4:1–13.
385. Lin MW, et al. Psychiatr Genet
1995;5:117–126.
386. Pulver AE, et al. Mol Psychiatry
2000;5:650–653.
387. DeLisi LE, et al. Psychiatr Genet
2000;10:153–158.
388. Williams NM, et al. Hum Mol
Genet 1999;8:1729–1739.
389. Chumakov I, et al. Proc Natl
Acad Sci U S A 2002;99:
13675–13680.
390. Shi J, et al. Schizophr Res
2008;98:89–97.
391. Korostishevsky M, et al. Am J Med
Genet B Neuropsychiatr Genet
2006;141B:91–95.
392. Shin HD, et al. Schizophr Res
2007;96:119–124.
393. Shinkai T, et al. Neuromolecular
Med 2007;9:169–177.
394. Corvin A, et al. Am J Med Genet
B Neuropsychiatr Genet
2007;144B:949–953.
395. Williams NM, et al. Arch Gen
Psychiatry 2006;63:366–373.
396. Bass NJ, et al. Behav Brain Funct
2009;5:28.
397. Mulle JG, et al. Mol Psychiatry
2005;10:431–433.
398. Goldberg TE, et al.
Neuropsychopharmacology
2006;31:2022–2032.
399. Lavedan C, et al. Mol Psychiatry
2009;14:804–819.
400. Pickard BS, et al. Mol Psychiatry
2009;14:874–884.
401. Pickard BS, et al. Am J Med
Genet B Neuropsychiatr Genet
2005;136B:26–32.
402. Pickard BS, et al. Ann Med
2006;38:439–448.
403. Macintyre G, et al. Schizophr Res
2010;120:143–149.
404. Emamian ES, et al. Nat Genet
2004;36:131–137.
405. Ikeda M, et al. Biol Psychiatry
2004;56:698–700.
406. Schwab SG, et al. Biol Psychiatry
2005;58:446–450.
407. Bajestan SN, et al. Am J Med Genet
B Neuropsychiatr Genet
2006;141B:383–386.
408. Norton N, et al. Schizophr Res
2007;93:58–65.
409. Xu MQ, et al. J Clin Psychiatry
2007;68:1358–1367.
410. Thiselton DL, et al. Biol Psychiatry
2008;63:449–457.
411. Lee KY, et al. Neurosci Res
2010;66:238–245.
412. Turunen JA, et al. Schizophr Res
2007;91:27–36.
413. Ide M, et al. J Neurochem
2006;99:277–287.
414. Liu YL, et al. Psychiatr Genet
2006;16:39–41.
415. Ohtsuki T, et al. Mol Psychiatry
2004;9:981–983.
416. Sanders AR, et al. Am J Psychiatry
2008;165:497–506.
417. Coon H, et al. Am J Med Genet
1994;54:72–79.
418. Pulver AE, et al. Am J Med Genet
1994;54:44–50.
419. Pulver AE, et al. Am J Med Genet
1994;54:36–43.
259
 Chapter 19: Genetics of schizophrenia
420. Vallada H, et al. Psychiatr Genet
1995;5:127–130.
421. Kalsi G, et al. Am J Med Genet
1995;60:298–301.
422. Polymeropoulos MH, et al.
Am J Med Genet 1994;
54:93–99.
423. Gill M, et al. Am J Med Genet
1996;67:40–45.
424. Severinsen JE, et al. Mol Psychiatry
2006;11:1126–1138.
425. Nyegaard M, et al. Am J Med
Genet B Neuropsychiatr Genet
2009;150B:151–152.
426. Bassett AS, et al. Biol Psychiatry
1999;46:882–891.
427. Bassett AS, et al. Curr Psychiatry
Rep 2008;10:148–157.
428. Bassett AS, et al. Am J Psychiatry
2003;160:1580–1586.
429. Bassett AS, et al. Am J Med Genet
2005;138:307–313.
430. Chow EW, et al. Am J Med Genet
1994;54:107–112.
431. Liu H, et al. Proc Natl Acad
Sci U S A 2002;99:
3717–3722.
432. Kempf L, et al. PLoS Genet 2008;
4:e1000252.
433. Fan JB, et al. Neurosci Lett
2003;338:252–254.
434. Glaser B, et al. Schizophr Res
2006;87:21–27.
435. Williams HJ, et al. Mol Psychiatry
2003;8:644–645.
436. Williams HJ, et al. Am J Med
Genet B Neuropsychiatr Genet
2003;120B:42–46.
437. Gogos JA, et al. Nat Genet
1999;21:434–439.
438. Paterlini M, et al. Nat Neurosci
2005;8:1586–1594.
439. Lachman HM, et al.
Pharmacogenetics 1996;
6:243–250.
440. Glatt SJ, et al. Am J Psychiatry
2003;160:469–476.
441. Munafo MR, et al. Mol Psychiatry
2005;10:765–770.
260
442. Shifman S, et al. Am J Hum Genet
2002;71:1296–1302.
443. Rudduck C, et al. Hereditas
1983;98:297–299.
444. Chodirker BN, et al. Clin Genet
1987;31:1–6.
445. Garofalo G, et al. Am J Psychiatry
1992;149:1116.
446. Axelsson R, et al. Hereditas
1981;95:337.
447. Hong ML. Chinese J Neurol
Psychiatry 1986;19:188–191.
448. Turner B, et al. Lancet 1961;
ii:49–50.
449. Sperber MA. Biol Psychiatry
1975;10:27–43.
450. Kaplan AR. Biol Psychiatry
1970;2:89–94.
451. Dasgupta J, et al. Indian J Med Res
1973;61:62–70.
452. Stefansson H, et al. Nature
2008;455:232–236.
453. Walsh T, et al. Science
2008;320:539–543.
454. Kirov G, et al. Hum Mol Genet
2008;17:458–465.
455. Xu B, et al. Nat Genet
2008;40:880–885.
456. Rujescu D, et al. Hum Mol Genet
2009;18:988–996.
457. Helbig I, et al. Nat Genet
2009;41:160–162.
458. de Kovel CG, et al. Brain
2010;133:23–32.
459. Dibbens LM, et al. Hum Mol
Genet 2009;18:3626–3631.
460. Heinzen EL, et al. Am J Hum
Genet 2010;86:707–718.
461. Segurado R, et al. Am J Hum
Genet 2003;73:49–62.
462. McQueen MB, et al. Am J Hum
Genet 2005;77:582–595.
463. Ng MY, et al. Mol Psychiatry
2009;14:774–785.
464. Ferreira MA, et al. Nat Genet
2008;40:1056–1058.
465. Lencz T, et al. Proc Natl Acad Sci
U S A 2007;104;19942–19947.
466. Vine AE, et al. Psychiatr Genet
2009;19:165–170.
467. Zhang D, et al. Mol Psychiatry
2009;14:376–380.
468. McQuillin A, et al. Eur J Hum
Genet 2011;19:588–592.
469. Grozeva D, et al. Arch
Gen Psychiatry 2010;67:
318–327.
470. Prata DP, et al. Psychiatr Genet
2009;19:113–116.
471. Goes FS, et al. Am J Med
Genet B Neuropsychiatr Genet
2009;150B:693–702.
472. Hattori E, et al. Am J Hum Genet
2003;72:1131–1140.
473. Prata D, et al. Am J Med Genet
B Neuropsychiatr Genet
2008;147B:914–917.
474. Zhang Z, et al. Psychiatr Genet
2009;19:151–153.
475. Chen YS, et al. Mol Psychiatry
2004;9:87–92; image 5.
476. Hattori M, et al. Am J Med Genet
2002;114:304–309.
477. Hattori M, et al. Hum Mol Genet
1993;2:1511.
478. Hattori M, et al. Biochem Biophys
Res Commun 1995;209:513–518.
479. Kunugi H, et al. Schizophr Res
1999;37:271–273.
480. Kunugi H, et al. Neurosci Lett
1998;241:65–67.
481. Nanko S, et al. Acta Psychiatr
Scand 1994;89:390–392.
482. Jones CT, et al. Nat Genet
1992;1:306–309.
483. Arnholt JC, et al. Biol Psychiatry
1993;34:739–740.
484. Carter D, et al. Hum Mol Genet
1993;2:321.
485. Coon H, et al. Am J Med Genet
1993;48:36–39.
486. Forsell C, et al. Neurosci Lett
1995;184:90–93.
487. Fukuda R, et al. Jpn J Hum Genet
1993;38:407–411.
488. Jonsson E, et al. Biol Psychiatry
1995;37:135–136.

489. Morris S, et al. Psychiatr Genet
1994;4:23–27.
490. Mortilla M, et al. Neurosci Lett
1994;165:45–47.
491. Arranz M, et al. Lancet 1995;
346:281–282.
492. Arranz MJ, et al. Neurosci Lett
1996;217:177–178.
493. Joober R, et al. J Psychiatry
Neurosci 1999;24:141–146.
494. Lin CH, et al. Neuroreport 1999;
10:57–60.
495. Koike H, et al. Proc Natl Acad Sci
U S A 2006;103:3693–3697.
496. Clapcote SJ, et al. Genetics
2006;173:2407–2410.
497. Clapcote SJ, et al. Neuron
2007;54:387–402.
498. Shen S, et al. J Neurosci
2008;28:10893–10904.
499. Zeng H, et al. Cell 2001;
107:617–629.
Chapter 19: Genetics of schizophrenia
500. Miyakawa T, et al. Proc Natl
Acad Sci U S A 2003;100:
8987–8992.
501. Karlsson RM, et al. Neuro-
psychopharmacology 2009;
34:1578–1589.
502. Young JW, et al. Eur Neuro-
psychopharmacol 2007;
17:145–155.
503. Chen YJ, et al. J Neurosci
2008;28:6872–6883.
504. van den Buuse M, et al. Int
J Neuropsychopharmacol
2009;12:1383–1393.
505. Andrieux A, et al. Genes Dev
2002;16:2350–2364.
506. Fradley RL, et al. Behav Brain Res
2005;163:257–264.
507. Eastwood SL, et al. J
Psychopharmacol 2007;
21:635–644.
508. Brigman JL, et al. Behav Neurosci
2006;120:984–988.
509. D’Arcangelo G. Epilepsy Behav
2006;8:81–90.
510. Costa E, et al. Mol Interv
2002;2:47–57.
511. Eastwood SL, et al. Mol Psychiatry
2003;8:769, 821–831.
512. Costa E, et al. Curr Opin
Pharmacol 2002;2:56–62.
513. Grayson DR, et al. Proc Natl
Acad Sci U S A 2005;102:9341–9346.
514. Woolley ML, et al. Psycho-
pharmacology (Berl) 2008;
196:431–440.
515. Ayhan Y, et al. Behav Brain Res
2009;204:274–281.
516. McGuire PK, et al. Schizophr Res
1995;15:277–281.
517. Arendt M, et al. Arch Gen
Psychiatry 2008;65:1269–1274.
518. Boydell J, et al. Schizophr Res
2007;93:203–210.
519. Phillips LJ, et al. Aust N Z
J Psychiatry 2002;36;800–806.
261
 The genetics of anorexia and
Chapter
20 bulimia nervosa
Andrew W. Bergen, Jennifer Wessel, and Walter H. Kaye
Overview of anorexia nervosa
and bulimia nervosa
Anorexia nervosa (AN) and bulimia nervosa (BN)
are related disorders of unknown etiology that most
commonly begin during adolescence in women.
They are frequently chronic and often disabling con-
ditions that are characterized by aberrant patterns of
feeding behavior and weight regulation, and deviant
attitudes and perceptions toward body weight and
shape. In AN, an inexplicable fear of weight gain
and unrelenting obsession with fatness, even in the
face of increasing cachexia, accounts for a protracted
course, extreme medical and psychological morbid-
ity, and standardized mortality rates exceeding those
of all other psychiatric disorders. BN usually emerges
after a period of food restriction, which may or may
not have been associated with weight loss. Binge
eating is followed by either self-induced vomiting,
or by some other means of compensation for the
excess of food ingested. Although abnormally low
body weight is an exclusion for the diagnosis of
BN, some 25–30% of bulimics have a prior history
of AN.
AN and BN are subdivided by eating behavior
and psychopathological characteristics [1–5]. AN is
characterized by severe emaciation [6]. Two types of
consummatory behavior are seen in AN. Restricting-
type anorexics (“AN”) lose weight purely by restricted
dieting and have no history of binge eating or purging.
Binge-eating/purging-type anorexics (AN-BN) also
restrict their food intake to lose weight, but have a
periodic disinhibition of restraint and engage in binge
eating and/or purging. Individuals with bulimia ner-
vosa (“BN”) do not become emaciated and are able to
maintain an average body weight (ABW) above 85%
Principles of Psychiatric Genetics, eds John I. Nurnberger, Jr. and Wade H. Berrettini. Published by Cambridge University
Press. © Cambridge University Press 2012.
262
[6]. There has been much conjecture regarding the
processes underlying pathological eating, but little is
understood of their biology [7, 8]. Similar to AN,
individuals with AN-BN and BN have a seemingly
relentless drive to restrain food intake, an extreme
fear of weight gain and a distorted view of their body
shape. In contrast to AN, however, AN-BN and BN
suffer recurring disinhibition of dietary restraint,
resulting in cycles of binge eating and compensatory
actions such as self-induced vomiting. Transitions
between AN and BN occur frequently, and these
disorders are often cross-transmitted in families
[9–12]. Consequently it has been argued that AN
and BN share some risk and liability factors.
Because AN and BN present most often during
adolescence in women, they are often theorized to be
caused by cultural pressures for thinness [13] since
dieting and the pursuit of thinness are common in
industrialized countries. Still, AN and BN affect only
an estimated 0.3–0.7% and 1.5–2.5%, respectively,
of females in the general population [14]. This dispar-
ity between the high prevalence of pressures for thin-
ness and the low prevalence of eating disorders (EDs),
combined with clear evidence of AN occurring at least
several centuries ago [15], the stereotypic presenta-
tion, substantial heritability, and developmentally
specific age-of-onset distribution, underscores the
possibility of contributing biological vulnerabilities.
Improvements in the understanding and treat-
ment of EDs are of immense clinical and public health
importance [16] since these are often chronic, relaps-
ing illnesses [17–19] with substantial and costly med-
ical morbidity [20]. Importantly, for AN, there is
no proven treatment that reverses symptoms [16];
consequently it has the highest death rate of any
psychiatric illness [21].

ED symptoms
The DSM-IV diagnostic criteria for AN and BN focus
on eating behavior and body image distortions.
Because of their unusual and prominent nature, these
symptoms tend to capture much attention. The
pathogenesis of the disturbed eating behaviors is
poorly understood [7, 8]. Individuals with AN rarely
have complete suppression of appetite, but rather
exhibit an ego-syntonic resistance to feeding drives
while simultaneously being preoccupied with food
and eating rituals to the point of obsession. Individ-
uals with AN severely restrict food intake, particularly
fats and carbohydrates, but rarely stop eating com-
pletely; rather they restrict their caloric intake to a
few hundred calories a day. They tend to be vegetar-
ians, have monotonous choices in food intake, select
unusual combinations of foods and flavors, and
have ritualized eating behaviors. Similarly, BN is not
associated with a primary, pathological increase in
appetite; rather, like individuals with AN, individuals
with BN have a seemingly relentless drive to restrain
their food intake, an extreme fear of weight gain, and
often have a distorted view of their actual body shape.
Loss of control with overeating in individuals with
BN usually occurs intermittently and typically only
some time after the onset of dieting behavior.
Restrained eating behavior and dysfunctional cogni-
tion relating weight and shape to self-concept are
shared by all types of patients with EDs.
AN and BN individuals commonly have clusters
of other puzzling symptoms. Excessive exercise and
motor restlessness are common in AN [22]. While not
well studied, excessive exercise is thought to be asso-
ciated particularly with the purging subtype of AN, as
well as with a constellation of anxious/obsessional
temperament characteristics. Individuals with AN
often have resistance to treatment [23]. In part this
is due to the ego-syntonic nature of the disorder,
which is demonstrated by the patient’s denial of being
underweight and refusal to accept the seriousness
of the medical consequences of the disorder. Conse-
quently, few controlled trials of any therapy have been
performed, in part, because it has been difficult to
enlist cooperation of individuals with AN, and in part
because psychological and pharmacological strategies
that have been successful in other disorders appear to
be less effective in this illness.
Mood and impulse control: Individuals with AN
and BN have elevated rates of lifetime diagnoses of
Chapter 20: Genetics of anorexia and bulimia nervosa
anxiety and depressive disorders, and obsessive–
compulsive disorder [10, 24–26]. In addition, individ-
uals with AN and BN are both consistently character-
ized by perfectionism, obsessive–compulsiveness,
neuroticism, negative emotionality, harm avoidance,
low self-directedness, low cooperativeness, and traits
associated with avoidant personality disorder. Con-
sistent differences that emerge between ED groups
are high constraint and persistence, low novelty seek-
ing, constriction of affect and emotional expressive-
ness, anhedonia and asceticism, and reduced social
spontaneity in restrictor-type AN. Individuals with
BN are more likely to have high impulsivity, sensation
seeking, novelty seeking, and traits associated with
borderline personality disorder in BN, and substance
abuse [27].
Neurocognition: Individuals with AN have an
obsessive, perseverative, and rigid personality style
and have difficulty shifting sets. While those with
AN do well on goal-directed behavior, they have
difficulties incorporating feedback and modifying
their behavior. For example, they often feel that they
should be able to do things perfectly without making
mistakes, and they have little appreciation for the
fact that mistakes are a normal learning experience.
Moreover, they often fail to accurately recognize
and incorporate affective and social stimuli in the
environment, as confirmed by laboratory tests [28,
29]. Those ill with AN and those recovered (REC)
from AN tend [30] to have delayed setshifting, which
normally allows for the adaptation of behavior in
response to the environment. Furthermore, individ-
uals with AN have enhanced ability to pay attention
to detail or use a logical/analytic approach but exhibit
worse performance with global strategies [29, 31].
State and trait: It has long been debated whether
symptoms in individuals with AN and BN are cause
or consequence of malnutrition. Recent studies show
that certain childhood temperament and personality
traits [32–35] such as negative emotionality, harm
avoidance, perfectionism, inhibition, drive for thin-
ness, altered interoceptive awareness, and obsessive–
compulsive personality, create a vulnerability for
developing AN and BN. In other words, the majority
of people with AN and BN exhibit one or more of
such traits in childhood, and thus these traits pre-date
the onset of AN and BN. Moreover, studies done on
three continents have shown that for AN and BN
individuals with a lifetime history of an anxiety dis-
order diagnosis, the anxiety disorder most often
263
 Chapter 20: Genetics of anorexia and bulimia nervosa
began in childhood before the onset of the ED [36–39].
The most common [26] premorbid childhood dis-
orders were obsessive–compulsive disorder (OCD)
and social phobia. In summary, such symptoms may
be susceptibility factors that make people vulnerable
to developing an ED. Malnutrition tends to exagger-
ate these premorbid behavioral traits [40] after the
onset of the illness, with the addition of other symp-
toms that maintain or accelerate the disease process,
including exaggerated emotional dysregulation and
obsessionality [24, 26].
The process of recovery in AN is poorly under-
stood and, in most cases, protracted. Still, approxi-
mately 50–70% of affected individuals will eventually
have complete or moderate resolution of the illness,
often in the early to mid 20s [41–43]. It is important
to emphasize that temperament and personality traits
such as negative emotionality, harm avoidance and
perfectionism, obsessional behaviors (particularly
symmetry, exactness, and order) persist after recovery
from both AN and BN [41, 43–46] and are similar to
the symptoms described premorbidly in childhood.
Compared to the ill state, symptoms in REC AN and
BN tend to be mild to moderate, including elevated
scores on core ED measures. Interestingly, REC AN
and BN tend to be more alike than different on many
of these measures, although there are some differ-
ences on factors related to impulse control or stimuli
seeking, such as novelty seeking [34, 42–43].
In summary, individuals with restricting type AN
are more likely to have restricted eating, constricted
affect and emotional mood expression, and impulse
over-control, as well as personality traits of marked
rigidity, conformity, and reduced social spontaneity.
Individuals with BN may show similar traits, but in
addition, may exhibit histories of episodic overeating,
extremes of intense affect, and impulse dysregulation.
Thus several domains (eating, affect, impulse control)
are involved in systematic ways, specifically, over
control in AN, and switches between over control
and under control in BN, which raises the question
of whether there is a disturbance of modulation of
multiple systems. AN and BN most commonly
develop during adolescence or young adulthood [19]
in proximity to puberty. Adolescence [47] is a time of
profound biological, psychological, and sociocultural
change, and it demands a considerable degree of
flexibility to successfully manage the transition into
adulthood. Psychologically, change may challenge the
rigidity of those at risk for AN and BN, and thus open
264
a window of vulnerability [47]. Thus these vulnerabil-
ities, which are biologically mediated, may significantly
enhance the risk of onset of an ED, particularly in
women.
Heritability
Considering that transitions between syndromes
occur in many individuals, it has been argued that
AN and BN share at least some risk and liability
factors [7, 9]. In fact, AN and BN are cross-transmitted
in families [10, 11]. Moreover there is an increased
prevalence of AN and BN as well as subthreshold
forms of EDs in relatives, consistent with the possibil-
ity of a continuum of transmitted liability in at risk
families manifesting a broad spectrum of eating dis-
order phenotypes [11]. In fact, AN is highly familial
[10, 11, 48]. The relative risk for AN in family
members of probands with AN is 11.3 [11].
Twin studies of AN and BN suggest there is
approximately a 50–80% genetic contribution to liabil-
ity [9, 49–52] accounted for by additive genetic factors.
These heritability estimates are similar to those found
in schizophrenia and bipolar disorder, suggesting that
AN and BN may be as genetically influenced as dis-
orders traditionally viewed as biological in nature.
Twin studies on European populations have
yielded heritability estimates using various strategies.
First, heritability of AN was estimated to be 58% (95%
CI 0.33–0.84), in the context of a bivariate twin analy-
sis with major depression [52]. Second, twin analyses
were conducted for a single question of “have you
ever had AN?” yielding a heritability estimate of 48%
(95% CI 0.27–0.65) [53]. Third, broadening the def-
inition of AN syndrome, Klump et al. reported the
heritability to be 76% [51]. A Swedish Twin Registry
study of 31 406 twins born between 1935 and 1958
and diagnosed by clinical interview, hospital dis-
charge diagnosis of AN, or cause of death certificate
yielded a heritability estimate of 56% (95% CI 0.00–
0.87) with the remaining variance attributable to
shared environment and unique environment [49].
This latter possibility is supported by twin studies
which found essentially no genetic influence on over-
all levels of ED symptoms in 11-year-old twins, but
significant genetic effects (> 50%) in 17-year-old
twins [54]. These findings collectively imply that
puberty may play a role in the genetic diathesis for
ED symptoms. The changes associated with adoles-
cence differ in males and females and may therefore

contribute to the sexual dimorphism of AN. Menar-
che is associated [47] with a rapid change in body
composition and neuropeptides modulating metabol-
ism. Little is known about whether the rise in estrogen
levels associated with puberty in females is contribu-
tory. Estrogens modulate serotonergic function [55]
as well as stress-related neuropeptides such as cortisol
releasing hormone (CRH) [56] via a variety of mech-
anisms. Moreover, a major phase of synaptogenesis,
pruning and myelination of predominantly frontal
and limbic areas occurs around the time of puberty
and adolescence and is thought to have a functional
role in the integration of emotional processing with
cognition [57].
The transmitted liability to AN and BN may be
mediated by a more diffuse phenotype of continuous,
heritable behavioral traits related to disordered eating
regulation [54, 58–60], often resulting in subthreshold
forms of ED in families [10, 11]. Substantial evidence
supports that many of these traits exist premorbidly,
are heritable, are elevated in unaffected family
members, persist after recovery from the disorder,
and are independent of body weight [32, 43, 44, 49,
61–64]. Therefore, we postulate that these traits, such
as anxiety, obsessions, or perfectionism, confer liabil-
ity to the development of AN.
Linkage studies of EDs
The Price Foundation, a private, European-based
foundation, has supported a multicenter international
collaboration to investigate the genetics of AN and BN.
In the first study, 192 families were ascertained. All
probands met modified DSM-IV criteria for AN; at
least one additional affected first- through fourth-
degree relative met DSM-IV criteria for AN, BN, or
eating disorder not otherwise specified (EDNOS) [65].
Blood for DNA was collected from all affected individ-
uals and available biological parents. Factors poten-
tially affecting susceptibility for AN were assessed with
a battery of standardized and validated instruments.
Using the Weber screening set, version 9 (Center for
Medical Genetics, Marshfield Medical Research Foun-
dation) with markers dispersed across the genome at
approximately 10 cM, and analyzing families in which at
least two affected relative pairs had AN, restricting sub-
type (RAN) (N ¼ 37 families, 32 sibling pairs of which
11 pairs had data for both parents) we found evidence
for linkage, with a nonparametric linkage (NPL) score of
3.03 at marker D1S3721 on chromosome 1p [66].
Chapter 20: Genetics of anorexia and bulimia nervosa
Serotonergic and opioidergic neurotransmitter
system alterations have been observed in people with
eating disorders; the genes for the serotonin 1D recep-
tor (HTR1D) and the opioid delta receptor (OPRD1)
are found on chr1p36.3–34.3. These candidate genes
were resequenced to identify or confirm single nucle-
otide polymorphisms (SNPs), and four HTR1D
and five OPRD1 SNPs were tested for linkage and
association with DSM-IV anorexia diagnosis [67,
68]. Linkage analysis of these candidate gene SNPs
with 33 microsatellite markers in N ¼ 37 families
including relative pairs concordantly affected with
RAN substantially increased the evidence for linkage
of this region to restricting AN to an NPL score
of 3.91 (p ¼ 0.00002), which exceeds the Lander/
Kruglyak threshold for significant linkage [69]. Asso-
ciation to DSM-IV AN diagnosis was performed
using 196 families in a family-based association test
and with 98 controls in a case-control design. Using
genotype data on parents and AN probands, 3 SNPs
at HTR1D were found to exhibit significant transmis-
sion disequilibrium (p < 0.05). Statistically significant
genotypic, allelic, and haplotypic association to AN in
the case-control design was observed at HTR1D and
OPRD1 with allelic effect sizes for individual SNPs of
2.63 (95% CI ¼ 1.21–5.75) for HTR1D rs6300 and 1.46
(95% CI ¼ 1.01–2.13) for OPRD1 rs536706. An inde-
pendent association study [70] of 226 females meeting
DSM-IV criteria for AN, and 678 matched controls
genotyped 4 SNPs in HTR1D and 6 SNPs in OPRD1.
One HTR1D SNP overlapped between the studies
(rs674386 or -1123T > C). Three OPRD1 SNPs were
found to be associated with both RAN and binge-
purge AN (BPAN), and two HTR1D SNPs exhibited
significant association with RAN with allelic effect
sizes for individual SNPs of 1.51 (95% CI ¼ 1.08,
2.10) for HTR1D rs856510 and 1.77 (1.20, 2.62) for
OPRD1 rs569356. These data support the hypothesis
that polymorphisms within HTR1D and OPRD1 form
a component of the genetic basis of susceptibility
to AN.
In an exploration of how behavioral covariates
enhanced the linkage signals, Devlin and colleagues
[71] evaluated features of ED for the following
criteria: (1) consistent relationship to eating path-
ology; (2) heritability; and (3) relationship to sever-
ity of some aspect of the disorder. Two variables,
drive-for-thinness and obsessionality (OBS), each
yielded a cluster of affected sibling pairs who had
high and concordant values for these traits, whereas
265
 Chapter 20: Genetics of anorexia and bulimia nervosa
other sibling pairs were notably discordant. Incorpor-
ation of these traits into covariate-based linkage ana-
lyses [72] yielded a significant additional linkage
signal on 1q, with a LOD score of 3.46, p ¼ 0.00003,
marker D1S1660, as well as two other suggestive link-
age signals, one at 2p (LOD ¼ 2.22, p ¼ 0.0007,
marker D251790), and another at 13q (LOD ¼ 2.50,
p ¼ 0.00035, marker D135894).
In further exploration of this linkage sample (154
affected sibling pairs) and an additional BN linkage
sample (244 affected sibling pairs), Bulik and col-
leagues [58] thoroughly explored eating disorder-
related traits. From more than 100 psychiatric, per-
sonality, and temperament phenotypes, they selected
a parsimonious subset of attributes to incorporate
into linkage analyses. Using a multilayer decision
analysis, they chose variables relevant to eating
disorder pathology with published evidence for
heritability. OBS, age-at-menarche, and a composite
anxiety measure (ANX) displayed features of herit-
able quantitative traits, such as normal distribution
and familial correlation, and thus appeared ideal
for quantitative trait locus linkage analysis. By con-
trast, some families showed highly concordant and
extreme values for three variables – lifetime min-
imum body mass index (BMI) (lowest BMI attained
during the course of illness), concern over mistakes
(CM), and food-related obsessions (OBF). These dis-
tributions were consistent with a mixture of popula-
tions, and thus the variables were matched with
covariate linkage analysis. The most compelling sig-
nals arose from the BN cohort [73]. For the BN
cohort, significant linkage signals arose on 4q21.1
(BMI), 14q21.1 (CM, OBF), 16p13.3 (CM). Suggest-
ive linkages were detected at the following chromo-
somal locations: 1q31.1 (ANX), 3p23 (BMI), 4p15.33
(CM), 4q35.2 (ANX), 5p15.3 (BMI), 8q11.23 (CM,
OBF), 10p11.21 (CM), 10p13.1 (BMI, OBF), 16p13.3
(OBF), and 18p11.31 (OBF). For the AN cohort, the
results for linkage were more modest. No result was
genome-wide significant, although there were some
suggestive linkage findings: 4q13.1 (BMI), 6q21
(OBS), 9p21.3 (OBS), 11p11.2 (CM), 15q26.2
(OBF), and 17q25.1 (CM, OBF). While substantial
linkage signals were not seen in both cohorts, more
modest signals did coincide, defining other areas
of suggestive linkage. These linkage findings are intri-
guing, but they require confirmation before resources
are invested to identify critical genetic variation in the
linkage regions.
266
Candidate gene association
studies in AN
In addition to the candidate gene studies motivated by
linkage analysis findings, the use of the neurobio-
logical approach to identify, genotype, and associate
candidate genes with AN diagnosis has been used
since 1997 to explore the potential involvement of
candidate genes with AN susceptibility. These studies
most often included candidate genes encoding pro-
teins involved in monoaminergic function, including
receptors and interacting proteins (dopamine recep-
tors 2, 3, and 4 and serotonin receptors 1B, 2A,
2C, and 7), transporters (norepinephrine, dopamine,
and serotonin), metabolic enzymes (catecholamine
transferase, monoamine oxidase A, tryptophan
hydroxylase) or neurotrophic factors (brain-derived
neurotrophic factor). These candidate gene studies
are characterized by the selection and genotyping of
a limited number of polymorphisms, in an era of
limited information on candidate gene variation and
linkage disequilibrium, and of smaller sample sizes
that would be considered necessary for discovery of
novel associations today. Some studies evaluating
dopaminergic and serotonergic candidate genes for
association with AN are reviewed below.
Schweiger et al. [7] performed a combined case-
control and family-based association analysis of
DSM-IV AN (excluding binge eating) utilizing 191
probands with a diagnosis of DSM-IV AN from Euro-
pean–American multiplex families recruited using a
sibling-pair strategy for linkage analysis, as described
in Kaye et al. [65]. Untransmitted alleles from family
members comprised the controls for family-based
association analysis. For case-control analyses, con-
trols were European–American females, screened for
Axis I disorders, and with weight for height of
< 120%, as defined by the Metropolitan Life tables
of height and weight, to exclude obese individuals.
Seven DNA polymorphisms at the DRD2 locus
were selected for genotyping: (1) the -141-C or Indel,
rs1799732; (2) IVS2–2730T4C, rs1800498; (3) 932C4G,
exon 7, S311C, rs1801028; (4) 939C4T, exon 7, H313,
rs6275; (5) 957C4T, exon 7, P319, rs6277; (6) Ex8–408
3’UTR, rs6278; and (7) 10620C4>T 3’ of STP or
TaqIA, E713K in ANKK1, rs1800497. The most con-
sistent evidence for association to AN diagnosis is
exhibited in haplotype analyses. Haplotypes com-
posed of the promoter (-141 Indel) polymorphism
and the exon 7 SNPs (939Y and/or 957Y) exhibited

statistically significant association in haplotype case-
control analyses, differences in linkage disequilibrium
estimates, and statistically significant transmission
distortion. Confidence in these associations may be
greater because deletion of a cytosine at -141 has been
reported to reduce transcription by an average of 68%
in vitro [75], and because the 957T allele has been
shown to exhibit decreased stability and reduced
translation of the DRD2 transcript in vitro [76]. The
957T allele exhibits statistically significant under-
transmission in combination with the -141 deletion
allele, suggesting that the 957T allele may contribute
to the association to AN diagnosis via a protective
effect. Functional [11C]raclopride neuroimaging
studies show that individuals who are recovered
from AN have findings consistent with elevated
D2 or D3 receptor activity [77]. We note that these
two observations (association of DRD2 SNP alleles
with diagnosis and with in vitro gene expression, and
elevated dopamine receptor activity in individuals
recovered from AN) are superficially consistent,
however, the comparison is between the behavior of
alleles within a cell line and receptors within the
brain, and any mechanistic link between them is
speculative.
Two serotonin candidate gene polymorphisms
have received the most attention with respect to inves-
tigation of association to DSM-IV anorexia diagnosis,
i.e. a promoter SNP at the 2A serotonin receptor
(5-HT2A) locus (HTR2A –1438G>A) [78–91], and a
promoter polymorphism at the serotonin transporter
(SERT) locus (SLC6A4 5-HTTLPR) [84, 90, 92–95].
The 5HT2A receptor is of interest in AN and BN
because it has been implicated in the modulation of
feeding and mood, as well as selective serotonin reup-
take inhibitors (SSRIs) response [96–100]. Post-synaptic
5HT2A receptors are in high densities in the cerebral
cortex and other regions of rodents and humans [101,
102]. In fact, a number of positron emission tomog-
raphy (PET) imaging studies have found that reduced
binding potential of the 5HT2A receptor persists in
individuals who have recovered from AN and BN [96,
103, 104]. In addition, reduced binding of the 5-HT2A
receptor has been found in ill AN in one study [105],
but not another [106]. In addition, using PET imaging,
recovered restrictor AN had elevated binding potential
of a SERT ligand compared to recovered bulimia AN
[107]. Together these imaging studies support the
possibility that eating disorder individuals have alter-
ations of 5-HT2A and SERT.
Chapter 20: Genetics of anorexia and bulimia nervosa
A fixed effects meta-analysis of the HTR2A and
SLC6A4 polymorphisms was performed, including
studies that sampled cases and controls from European
ancestry populations, and using a dominant model
with each polymorphism (Tables 20.1 and 20.2) [108,
109]. For the HTR2A meta-analysis, we excluded two
family-based association studies, based on 45 sibling
pairs discordant for AN diagnosis and on 313 families
[83, 85], neither of which provided evidence for asso-
ciation with HTR2A, and a case-control association
study that used unscreened control samples from a
DNA bank [87], resulting in a final sample of 796 cases
and 1290 controls from 9 sites (Table 20.1). For the
SLC6A4 meta-analysis, the final sample consisted of
489 DSM-IV AN cases and 563 controls from 5 sites
(Table 20.1). The current literature does not facilitate
meta-analyses of diagnostic subtypes of AN, e.g. for
the five studies with genotype data on the SLC6A4
polymorphism, only three provide genotype data by
diagnostic subtype, reducing the sample sizes for meta-
analysis considerably.
Our meta-analyses of the literature did not iden-
tify a significant association of HTR2A -1438G>A
with AN diagnosis (Table 20.2), although a much
larger sample with a similar odds ratio might be
significant. Our meta-analysis of the literature did
identify a significant association of the SLC6A4
5-HTTLPR short allele with AN diagnosis. Individ-
uals carrying the short allele at the 5-HTTLPR poly-
morphism were 38% (95% CI 6–81%) more likely
to fulfill criteria for a diagnosis of DSM-IV AN. There
was no significant evidence for heterogeneity (Q)
or publication bias (Egger regression) for either the
HTR2A or SLC6A4 polymorphism meta-analysis,
although there was a suggestion of effect size hetero-
geneity and publication bias for the HTR2A literature
as both estimates had significance values between
0.05 and 0.10 (Table 20.2).
Summary and recent findings
The candidate gene literature for AN is modest for a
disorder of such morbidity and mortality, and most
candidate genes studied have utilized a limited
number of polymorphisms in samples sizes that
would now be considered suboptimal for the identifi-
cation of significant associations. There is convergent
evidence from neuroimaging and molecular genetics
for involvement of both dopaminergic and serotoner-
gic neurotransmitter systems in the pathophysiology
267
 Table 20.1 HTR2A –1438 G>A and SLC6A4 5–HTTLPR case-control association studies to DSM-IV anorexia nervosa diagnosis.

Author Year PMID Ancestry Case N Case Case Case Case Control N Control Control Control Control
major heteroz minor maf* major heteroz minor maf
HTR2A GG GA AA GG GA AA
Cambell 1998 9259661 British 152 45 68 39 0.48 150 53 67 30 0.42
Collier 1997 9259661 British 81 23 33 25 0.51 226 75 117 34 0.41
Enoch 1998 9635956 American 68 16 35 17 0.51 69 25 38 6 0.36
Fuentes 2004 15167698 Spanish 95 30 48 17 0.43 107 39 49 19 0.41
Hinney 1997 9357428 German 100 41 39 20 0.40 355 116 177 62 0.42
Kipman 2002 12231269 French 145 49 69 27 0.42 98 26 50 22 0.48
Ricca 2004 15245785 Italian 77 16 46 15 0.49 115 44 59 12 0.36
Rybakowski 2006 16397402 Polish 148 16 60 55 0.57 89 14 49 26 0.57
Ziegler 1999 10523809 German 78 39 32 7 0.29 170 89 60 21 0.30
SLC6A4 LL LS SS LL LS SS
Di Bella 2000 10889521 Italian 56 17 22 17 0.50 120 48 57 15 0.36
Fumeron 2001 11244478 French 67 17 31 19 0.51 148 44 76 28 0.45
Hinney 1997 9395256 German 96 29 51 16 0.43 112 43 55 14 0.37
Rybakowski 2006 16397402 Polish 132 48 62 22 0.40 93 39 40 14 0.37
Sundaramurthy 2000 10686552 British 138 40 63 35 0.48 90 34 40 16 0.40
* maf, minor allele frequency.
 Chapter 20: Genetics of anorexia and bulimia nervosa

Table 20.2 HTR2A –1438 G>A and SLC6A4 5HTTLPR meta-analysis to DSM-IV anorexia nervosa diagnosis.

Author OR 95%CI p Weight Heterogeneity (Q) p Publication bias p

HTR2A
Cambell 1.30 0.80, 2.11 0.288 14.29%
Collier 1.25 0.72, 2.19 0.428 11.18%
Enoch 1.85 0.88, 3.89 0.107 5.08%
Fuentes 1.24 0.69, 2.23 0.467 9.98%
Hinney 0.70 0.44, 1.10 0.123 21.29%
Kipman 0.71 0.40, 1.25 0.230 14.35%
Ricca 2.36 1.21, 4.60 0.012 5.85%
Rybakowski 1.34 0.62, 2.91 0.457 5.39%
Ziegler 1.10 0.64, 1.88 0.731 12.59%
Meta-analysis 1.14 0.95, 1.38 0.166 0.081 0.068
SLC6A4
Di Bella 1.53 0.78, 3.01 0.218 15.07%
Fumeron 1.24 0.65, 2.39 0.512 17.82%
Hinney 1.44 0.81, 2.57 0.217 20.85%
Rybakowski 1.26 0.73, 2.18 0.398 24.97%
Sundaramurthy 1.49 0.85, 2.61 0.167 21.29%
Meta-analysis 1.38 1.06, 1.81 0.017 0.985 0.788

of or susceptibility to AN. However, the diversity of
candidate gene families that has been examined in the
AN literature is limited and many additional families
of genes that function in neuronal and other tissues
remain to be examined. Fortunately, the development
of genomic technologies enabling exhaustive analyses
of both common and rare germline variation in the
genome, and an appreciation for the importance of
larger samples sizes in pursuing association studies
provide opportunities for discovery. Over a period of
10 years, the Price Foundation multinational collabor-
ation has ascertained over 1000 unrelated individuals
with a DSM-IV diagnosis of AN in three separate
studies encompassing multiple ascertainment designs,
and nearly 700 unrelated individuals screened for
DSM-IV diagnoses, with normal adult weight (19 <
BMI < 27 kg/m2) and matched with ED participants
on 4 factors [65, 110, 111]. Up to 1762 female partici-
pants (up to 1085 cases and 677 controls) from this
dataset have recently been analyzed for association of
5151 candidate gene SNPs with diagnosis [111], and
up to 1033 female AN cases and up to 3773 pediatric
control subjects were analyzed genome-wide for asso-
ciation of  598 000 SNPs to diagnosis [112]. Both
analyses utilized extensive quality control, assessment
of case-control matching, the Cochran–Armitage
trend test for association testing, and stringent experi-
ment-wide criteria to define significance thresholds
for individual SNPs (p < 9.7EE-6 and p < 1EE-8 for
the two studies, respectively). Neither study identified
association findings related to categorical diagnoses
that reached a priori defined significance levels,
although there were modest associations at biological
candidate genes in both studies. Additional analyses
of the candidate gene dataset for associations with
eating disorder related traits previously studied via
linkage analyses [113], and for recovery outcomes,
with replication [114], resulted in significant associ-
ations between GABA receptor SNPs and recovery.
Investigators concluded that an intronic GABRG1
SNP and an excess of GABA receptor SNPs were
highly (p = 5EE-6) and nominally (p < 0.05) signifi-
cantly associated with recovery, respectively [114].
Using an independent sample of cases and controls,

269
 Chapter 20: Genetics of anorexia and bulimia nervosa

investigators identified a nominally significant associ-
ation of the intronic SNP with trait anxiety, suggest-
ing a mechanism for the recovery association [114].
In summary, a series of linkage and association stud-
ies identified significant linkage in the chr1p34–36
region and provided evidence for association at the
genes coding for the delta opioid and 1D serotonin
receptors in samples ascertained by this collaboration,
which has been replicated in an independent sample.
Further candidate gene and genome-wide analyses
point to additional candidate genes but also to the
need for larger collaborations and alternative molecu-
lar and statistical genetic methods for gene discovery.
The Price Foundation Collaboration, with the largest
and most comprehensive studies in the field, are now
utilizing next generation sequencing technologies to
enable discovery of novel candidate genes and char-
acterization of existing biological candidates.

References 14. Hoek HW, et al. Am J Psychiatry

1. American Psychiatric Association.
Diagnostic and Statistical Manual
of Mental Disorders, Fourth
Edition, Text Revision (DSM-IV-
TR). Washington, DC: American
Psychiatric Association; 2000.
2. Garner DM, et al. Am J Psychiatry
1985;142:581–587.
3. Halmi KA, et al. Am Acad Child
Psychiatry 1982;21:369–375.
4. Herzog DB, et al. N Engl J Med
1985;313:295–303.
5. Strober M, et al. J Nerv Ment Dis
1982;170:345–351.
6. American Psychiatric Association.
Diagnostic and Statistical
Manual of Mental Disorders,
Fourth Edition. Washington,
DC: American Psychiatric
Association; 1994.
7. Schweiger U, et al. In Jimerson
DC et al. (eds.). Balliere’s Clinical
Psychiatry. London: Balliere’s
Tindall; 1997.
8. Vitousek K, et al. J Abnorm
Psychol 1994;103:137–147.
9. Kendler KS, et al. Am J Psychiatry
1991;148:1627–1637.
10. Lilenfeld LR, et al. Arch Gen
Psychiatry 1998;55:603–610.
11. Strober M, et al. Am J Psychiatry
2000;157:393–401.
12. Walters EE, et al. Am J Psychiatry
1995;152:64–71.
13. Strober M. In Brownell K,
et al. (eds.). Eating Disorders
and Obesity – A Comprehensive
Handbook. New York: Guilford
Press; 1995.
1995;152:1272–1278.
15. Treasure J, et al. Psychol Med
1994;24:3–8.
16. NICE. Core Interventions in
the Treatment and Management
of Anorexia Nervosa, Bulimia
Nervosa and Related Eating
Disorders (Clinical Guideline 9).
London: National Collaborating
Centre for Medical Health; 2004.
17. Herzog DB, et al. J Clin Psychiatry
1992;53:147–152.
18. Keel PK, et al. Arch Gen Psychiatry
1999;56:63–69.
19. Klein D, et al. Int Rev Psychiatry
2003;15:205–216.
20. McKenzie JM, et al. Int J Eat
Disord 1992;11:235–241.
21. Sullivan PF. Am J Psychiatry
1995;152:1073–1074.
22. Shroff H, et al. Int J Eat Disord
2006;39:454–461.
23. Halmi K, et al. Arch Gen
Psychiatry 2005;62:776–781.
24. Godart N, et al. J Affect Disord
2007;97:37–49.
25. Godart NT, et al. Int J Eat Disord
2002;32:253–270.
26. Kaye W, et al. Am J Psychiatry
2004;161:2215–2221.
27. Cassin S, et al. Clin Psychol Rev
2005;25:895–916.
28. Kingston K, et al. Psychol Med
1996;26:15–28.
29. Strupp BJ, et al. Neuropsychobiology
1986;15:89–94.
30. Tchanturia K, et al. Int J Eat
Disord 2005;37:S72–S76.
31. Lopez C, et al. Int J Eat Disord
2008;41:143–152.
32. Anderluh MB. Am J Psychiatry
2003;160:242–247.
33. Fairburn CG, et al. Arch Gen
Psychiatry 1999;56:468–476.
34. Lilenfeld L, et al. Clin Psychol Rev
2006;26:299–320.
35. Stice E. Pychopharm Bull
2002;128:825–848.
36. Bulik CM, et al. J Nerv Ment Dis
1997;185:704–707.
37. Bulik CM, et al. Acta Psychiatr
Scand 1997;96:101–107.
38. Deep AL, et al. Int J Eat Disord
1995;17:291–297.
39. Godart NT, et al. Eur Psychiatry
2000;15:38–45.
40. Pollice C, et al. Int J Eat Disord
1997;21:367–376.
41. Steinhausen HC. Am J Psychiatry
2002;159:1284–1293.
42. Strober M, et al. Int J Eat Disord
1997;22:339–360.
43. Wagner A, et al. Int J Eat Disord
2006;39:276–284.
44. Casper RC. Psychosom Med
1990;52:156–170.
45. Srinivasagam NM, et al. Am
J Psychiatry 1995;152:1630–1634.
46. Strober M. J Psychosom Res
1980;24:353–359.
47. Connan F, et al. Physiol Behav
2003;79:13–24.
48. Strober M, et al. Int J Eat Disord
1990;9:239–253.
49. Bulik C, et al. Arch Gen Psychiatry
2006;63:305–312.

270

50. Bulik C, et al. Biol Psychiatry
1998;44:1210–1218.
51. Klump KL, et al. Psychol Med
2001;31:737–740.
52. Wade TD. Am J Psychiatry
2000;157:469–471.
53. Kortegaard LS, et al. Psychol Med
2001;31:361–365.
54. Klump KL, et al. J Abnormal
Psychology 2000;109:
239–251.
55. Rubinow DR, et al. Biol Psychiatry
1997 8;44:839–850.
56. Torpy D, et al. J Clin Endocrinol
Metab 1997;82:982.
57. Benes F. Am J Psychiatry 1998;
155:1489.
58. Bulik C, et al. Am J Med Genet B,
Neuropsych Genet 2005;139:81–87.
59. Rutherford J, et al. Psychol Med
1993;23:425–436.
60. Wade T, et al. Psychol Med 1998;
28:761–771.
61. Bulik C, et al. Int J Eat Disord 2007
[epub ahead of print].
62. Bulik CM, et al. Int J Eat Disord
2000;28:139–147.
63. Holliday J, et al. Psychol Med
2006;36:529–538.
64. O’Dwyer AM, et al. Psychol Med
1996;26:353–359.
65. Kaye WH, et al. Biol Psychiatry
2000;47:794–803.
66. Grice DE, et al. Am J Hum Genet
2002;70:787–792.
67. Bergen AW, et al. Mol Psychiatry
2003;8:397–406.
68. Bergen AW, et al. Current Drug
Targets: CNS Neurolog Disord
2003;2:41–52.
69. Lander E, et al. Nat Genet
1995;11:241–247.
70. Brown K, et al. Biol Psychiatry
2007;61:367–373.
71. Devlin B, et al. Hum Mol Genet
2002;11:689–696.
Chapter 20: Genetics of anorexia and bulimia nervosa
72. Devlin B, et al. Genet Epidemiol
2002;22:52–65.
73. Bacanu S, et al. Am J Med
Genet B, Neuropsych Genet
2005;139:61–68.
74. Bergen A, et al. Neuropsychopharm
2005;30:1703–1710.
75. Arinami T, et al. Hum Mol Genet
1997;6:577–582.
76. Duan J, et al. Mol Psychiatry
2003;8:901–910.
77. Frank G, et al. Biol Psychiatry
2005;58:908–912.
78. Ando T, et al. Psychiatr Genet
2001;11:157–160.
79. Campbell DA, et al. Lancet
1998;351:499.
80. Collier DA, et al. Lancet 1997;
350:412.
81. Enoch MA, et al. Lancet 1998;
351:1785–1786.
82. Fuentes J, et al. Psychiatr Genet
2004;14:107–109.
83. Gorwood P, et al. Mol Psychiatry
2002;7:90–94.
84. Hinney A, et al. Life Sci 1997;61:
PL 295–303.
85. Karwautz A, et al. Psychol Med
2001;32:317–329.
86. Kipman A, et al. Eur Psychiatry
2002;17:227–229.
87. Nacmias B, et al. Neurosci Lett
1999;277:134–136.
88. Nishiguchi N, et al. Biol Psychiatry
2001;50:123–128.
89. Ricca V, et al. Neurosci Lett
2002;323:105–108.
90. Rybakowski F, et al. Neuro-
psychopharm 2006;53:
33–39.
91. Ziegler A, et al. Lancet 1999;
353:929.
92. Di Bella DD, et al. Mol Psychiatry
2000;5:233–234.
93. Fumeron F, et al. Mol Psychiatry
2001;6:9–10.
94. Matsushita S, et al. Am J Med
Genet B Neuropsych Genet
2004;128B:114–117.
95. Sundaramurthy D, et al. Am J Med
Genet 2004;96:53–55.
96. Bailer UF, et al. Neuro-
psychopharmacology 2004;
29:1143–1155.
97. Bonhomme N, et al. J Clin
Psychopharmacol 1998;
18:447–454.
98. De Vry J, et al. Neurosci Biobehav
Rev 2000;24:341–353.
99. Simansky KJ. Behav Brain Res
1996;73:37–42.
100. Stockmeier CA. Ann NY Acad Sci
1997;836:220–232.
101. Burnet PW, et al. Neurochem Int
1997;30:565–574.
102. Saudou F, et al. Neurochem Int
1994;25:503–532.
103. Frank GK, et al. Biol Psychiatry
2002;52:896–906.
104. Kaye WH, et al. Am J Psychiatry
2001;158:1152–1155.
105. Audenaert K, et al. J Nucl Med
2003;44:163–169.
106. Bailer UF, et al. Biol Psychiatry
2007;61:1090–1099.
107. Bailer UF, et al. Psychopharmacology
2007;195:315–324.
108. Bax L, et al. BMC Med Res
Methodol 2006;6:50.
109. Bax L, et al. BMC Med Res
Methodol 2007;7:40.
110. Kaye WH, et al. Int J Eat Disord
2004;35:556–570.
111. Pinheiro A, et al. Am J Med Genet
B Neuropsychiatr Genet
2010;153B:1070–80.
112. Wang K, et al. Mol Psychiatry
2011;16:945–959.
113. Root TL, et al. Eur Eat Disord Rev
2011;19:487–493.
114. Bloss CS, et al. Neuropsychopharm
2011;36:2222–2232.
271
 Genetics and common human obesity
Chapter
21 R. Arlen Price
The prevalence of obesity has increased dramatically
over the past several decades. Since 1976 the preva-
lence of obesity has more than tripled and by 2008
exceeded one-third of the US population (33.8% with
body mass index [BMI] > 30), while another third
was overweight (34.2% with BMI ¼ 25.0–29.9) [1].
Highest obesity rates are found in the United States
and United Kingdom among developed countries,
and in the Middle East and Pacific Islands in
the developing world (WHO: https://apps.who.int/
infobase/). The increase in obesity rates in developing
countries has coincided with a transition to “West-
ern” diets and lifestyles [2–4]. Life in the developed
and developing world has become more sedentary
just as food has become more widely available. Many
lifestyle factors have been suggested to contribute to
the dramatic obesity increase, but the primary cause
is an obvious one . . . excess caloric intake [5, 6]. The
type of diet does contribute to national differences,
but change within countries appears to have been
driven primarily by overall food availability [7]. As
relatively inexpensive, high caloric foods have become
readily available in much of the world, we are eating
more, the food is higher in caloric content, and we are
gaining weight as a consequence.
Heritability
Obesity has moderate to high heritability, but genetic
variation alone cannot account for the dramatic
increase. First of all, gene mutations are rare and allele
frequencies do not change over short time periods in
large populations. The recent prevalence increases
must have an environmental origin. An environmen-
tal effect of so great a magnitude raises the question of
whether the heritability of obesity has declined during
the same period, but this does not appear to be the
Principles of Psychiatric Genetics, eds John I. Nurnberger, Jr. and Wade H. Berrettini. Published by Cambridge University
Press. © Cambridge University Press 2012.
272
case. Many studies have estimated heritability of BMI
and related variables [8] and they are consistent in
finding moderate to high heritability. Furthermore,
the estimates do not depend on the period of the
study. For example, 2 studies of twins conducted
almost 20 years apart found virtually identical esti-
mates of overall heritability of BMI of about 0.80
[9, 10]. So, while estimates from particular studies vary,
there is no trend toward decreasing (or increasing)
heritability.
Gene–environment interaction
So what role do genes play in the presence of
such large environmental effects? Bouchard and col-
leagues completed a series of landmark studies that
helped to explain the role of inherited variation
in mediating environmentally influenced change.
Bouchard’s research group studied monozygotic
twins exposed to long-term positive or negative
energy balance. There were considerable individual
differences in weight gain or loss under the different
conditions, but changes were similar in the genetically
identical co-twins, both in overall weight and visceral
fat. The common genotype of the twins influenced
their similarity in response to the environment [11],
demonstrating that adaptation to environmental
conditions is a heritable trait.
The major environmental changes that are
credited with causing the obesity pandemic have
occurred at a population level, but, as with the study
of twins, individuals differ in their response. While
as much as two-thirds of populations of developed
countries are overweight or obese, the remaining
third, living in the same environment, are of normal
weight or thin. At the least, this implies a behavioral
interaction, and, given the heritability of obesity and

coordinate changes in twins, gene–environment inter-
action must play a major mediating role. A few stud-
ies have tried to identify environmental interactions
with specific genes, focusing of weight gain or loss
as phenotypes, and diet or exercise as components
of the environment. One review identified some 13
studies that reported associations with some measure
of exercise and 15 with diet and/or exercise [12].
However, most associations have not been replicated.
The interaction most consistently supported was
with the Trp64Arg polymorphism in the adrenergic
receptor beta 3 (ADRB3) gene [12]. Limited power
due to small sample size may in part contribute to
the inconsistency of results. However, in the end most
reported associations will be false positives while
a few failures to replicate could be false negatives.
The pairing of new technologies with larger sample
sizes could prove more robust for examining gene–
environment interactions, but this possible outcome
will depend on the nature and magnitude of the
individual interactions.
Candidate genes
The phenotypic response in susceptible individuals
must be influenced by variability in genes that influ-
ence energy balance. Energy homeostasis requires
the coordination of appetite and satiety with energy
expenditure and storage. A great deal has been
learned about how energy homeostasis is maintained
[13]. It is a complex process involving genes that
regulate appetite, energy metabolism, and fat depos-
ition. Many genes that lie in associated regulatory
pathways have become candidates for weight gain
and obesity. These have included Leptin, Leptin Recep-
tor, MC4R, UCPs, PPARG, NPY, and Ghrelin as well
as genes in signaling pathways.
Candidate gene studies have identified mutations
in humans or introduced them in animal models
[14]. The last comprehensive count of human studies
identified associations with 127 genes, most with at
best mixed records of replication. The positive side of
a candidate gene approach is that the genes derive
from an emerging understanding of biology. Any
associations that are detected with common obesity
fit into a pre-existing framework. Candidate gene
studies have had their successes. Major gene muta-
tions have been associated with obesity. However,
they tend to be rare and account for a few cases of
extreme obesity [14].
Chapter 21: Genetics and common human obesity
Common obesity/rare gene variation
Overall, candidate gene studies have been unsuccess-
ful in explaining common forms of obesity. Genes
central to energy balance tend to have low variability,
presumably because of strong selection pressure.
Even so, some have argued that mutations in a large
number of genes may account for most human
obesity and other common diseases. This view is
sometimes called the common trait rare gene hypoth-
esis (CTRV), [15, 16], as opposed the common trait
common variant (CTCV) hypothesis.
Substantial progress in finding rare variants has
come with a focus on copy number variation (CNV,
a variant in a DNA segment of 1 or more kilobase
in length). While major deletions, duplications and
rearrangements of DNA sequence associated with
rare diseases have been know for some time, the
scale of CNV was not appreciated until the last few
years. One whole genome survey found more than
4000 variants, affecting more than 600 Mb of geno-
mic DNA sequence [17]. Large-scale screening has
identified associations of CNVs with a number of
phenotypes [18] including type 1 diabetes, neuropsy-
chiatric conditions [19], and several other common
disorders [20].
Wang et al. [21] completed a genome-wide CNV
survey focused on extreme phenotypes. Larger CNVs
were over-represented in obese cases compared with
never-overweight controls. CNVs larger than 2 Mb
were present in 1.3% of the cases but absent in con-
trols. The most pronounced effects were associated
with rare deletions that disrupt genes (odds ratio
[OR] ¼ 2.7 for CNVs > 1 Mb). Several CNVs were
found to disrupt known candidate genes for obesity,
such as a 3.3 Mb deletion disrupting NAP1L5 and
a 2.1 Mb duplication disrupting UCP1 and IL15.
Other studies have identified CNVs that are asso-
ciated with genes not previously associated with
obesity. An association between BMI and a chromo-
some 10q11 CNV was reported in a Chinese cohort
[22]. Two genes in this region are GPRIN2 and
PPYR1. In other studies, a deletion on 16p11.2 was
reported to be associated with obesity [23, 24]. More
recent studies have identified larger numbers of
CNVs [25, 26], and, although individually rare, a
few have been found in multiple individuals, some
in independent samples. The functional significance
of most of the genes affected by CNVs is unknown.
As such they become new candidates for obesity.
273
 Chapter 21: Genetics and common human obesity
Linkage studies
One source of motivation for proposing the CTRV
hypothesis was that attempts, through linkage and
association, to identify common genes had been
unsuccessful, at least until recently. The search for
common genes has generally taken a genomic
approach in which the entire genome is screened
without prior hypotheses. Linkage studies were the
first to take a whole genome approach. There have
been more than 60 of them for obesity-related traits
[14], for example, but the results have been disap-
pointing. A meta-analysis of 37 of these studies found
only 2 regions to be significantly supported at the 1%
level [27]: chromosome 13q for BMI and chromo-
some 12q for obesity (BMI  30). The outcome of
the meta-analysis helps explain why most comprehen-
sive searches for gene associations under linkage
peaks have been unsuccessful.
Whole genome association studies
Whole genome association (WGA) studies made it
possible to address the two most serious deficiencies
of previous approaches in that new genotyping
technology has been combined with very large sample
sizes. Moreover, WGA studies have several advan-
tages over whole genome linkage scans. The reso-
lution is 2–3 orders of magnitude greater, 2–5 Mb
in linkage studies compared with 10–100 kb with
association. Cases and controls are much easier to
collect than families, and the sample sizes required
while large are much smaller than those required for
linkage [28] and well within reach for collaborative
groups, if not individual investigators. The advantage
of a WGA approach was recognized some time ago
[29] but the available technology was insufficient at
that time.
The recent spate of GWA studies have depended
upon advances in marker identification and genomic
technology for high throughput genotyping. The
International HapMap Project (www.hapmap.org/)
identified more than 4 million single nucleotide poly-
morphisms (SNPs) and 550 000 of them provide
about 95% coverage of the genome in most popula-
tions, with about double that number needed for
Africans [30]. High throughput technology makes it
possible to type up to 1 million genotypes in a single
pass (Affymetrix and Illumina). Greatly reduced costs
have made the technology widely accessible. Finally,
274
large sample sizes have been developed through the
cooperation of investigators at multiple sites.
WGA studies have become widely available only
within the past four years. The breakthrough for
WGA studies came from the Wellcome Trust Case
Control Consortium (WTCCC) study that included
490 000 SNPs and a total of almost 39 000 individuals,
although the initial phase utilized a much smaller
number of cases and controls, about 2000 and 3000,
respectively [31]. The study was unprecedented in size
and in the strength of the association with FTO. An
association with MC4R has also been reported based
on the WTCCC sample [32]. The association with
FTO has been replicated in most studies that followed.
MC4R has been replicated as well, although not
as consistently. Recent large-scale meta-analysis of
multiple GWA studies identified additional genes
harboring common SNPs that associate with BMI
[33–36]. GWA studies have also found associations
with measures of body fat distribution [33, 37, 38].
By far the largest GWA study to date included almost
250 000 individuals and 2.8 million SNPs [39]. Asso-
ciations of BMI with 28 loci reached genome-wide
significance. Ten had been reported previously and
eighteen were newly identified. Four additional loci
were associated with body fat distribution, all of
which had been identified previously. However, even
this major expansion of sample size has not explained
much variation, 1.39% for BMI and 0.16% for body
fat distribution.
Large sample sizes have helped identify associ-
ations and improved replication, however, the effect
sizes have not grown larger and the increase in
variance accounted for has been minimal. These
findings on obesity are consistent with those for
stature, a complex trait with an even higher herit-
ability of at least 0.80. A large GWA study of stature
involving some 63 000 subjects found 54 associated
genes that accounted for only about 5% of the total
variation in height [40–42]. This finding led to much
discussion and speculation as to what happened to
the so called “missing heritability” [43]. Suggestions
have included gene–environment interaction, as well
as epigenetics. As discussed in an earlier section,
gene–environment interaction can play an important
role in the development of obesity, although it
should be born in mind that this may only compli-
cate things further, as environmental response is
itself heritable. Epigenetics will be discussed later
in this chapter.

Disparate approaches appear to
converge, at least in one aspect
WGA studies results have demonstrated that there are
indeed common variants in genes that increase risk
for obesity. This is particularly true for FTO that has
been widely replicated. However, the proportion of
variance in BMI these common genes account for is
quite small. Major gene mutations such as those in
leptin, leptin receptor, and POMC have dramatic
effects on individuals but are so rare that they account
for essentially no common variance. CNVs are much
more common that major gene mutations, but they
are still relatively rare and account for little variance
overall. While there are marked differences in fre-
quency, each approach has been successful. However,
the identified variants have done very little to reduce
the size of the “missing heritability”.
Taken on face value, the results from the different
approaches suggest polygenic inheritance. The classic
polygenic model was devised by R. A. Fisher as a way
of incorporating Mendelian inheritance into quanti-
tative variation [44]. For convenience he assumed
there were multiple causal genes, each with small
and roughly equal effects. The particulars, however,
give a somewhat different picture. It turns out there
are indeed multiple causal genes, and each variant
accounts for little overall variance. However, the vari-
ants have a wide range of effects on the individuals
that carry them. There is as yet little evidence that the
effects sum to create the phenotype, although one
study found that individuals who had all 32 risk alleles
associated with obesity did have a somewhat higher
BMI [39].
Epigenetic modification
There has been much discussion of late about the
possible effects of epigenetic changes on risk for
common disorders [45]. Epigenetic modification refers
to changes in gene expression that are heritable, that
is, which are maintained during somatic cell division
and may in some cases be passed on to offspring.
Genomic imprinting is the most studied form of
epigenetic modification, and involves the differential
marking of parental chromosomes during gameto-
genesis. Imprinting appears to occur in all marsupial
and placental species, and many of the imprinted
genes are related to body size and/or metabolism
[46–48]. The conflict theory suggests the association
Chapter 21: Genetics and common human obesity
of imprinting with body size arose due to differential
parental investment in offspring in polyandrous
animals. Males are invested in larger body size of their
offspring while females have an equal investment in
all offspring regardless of the father. The theory is
supported by fetus size in deer mice (peromyscus)
hybrids of monogamous and polyandrous species [49].
The best know example relating to obesity is
the Prader–Willi and Angelman syndromes, which
are due to imprinting of the paternal or maternal
chromosome, respectively, of region 15q11–13.
Another imprinted gene is insulin-like growth factor 2,
and paternal expression is strongly related to
several measures of fat deposition in pigs [50]. In
addition, quantitative trait loci (QTL), inferred genes
based on linkage, have been identified in mice.
Imprinting is suggested because linkage depends on
parent of origin. In one study, five QTL were found,
two paternal, two maternal, and one with no parent
of origin effect [51].
Parent of origin effects have also been identified
in humans. A large survey reported parent of origin
dependent associations of variants in known imprinted
regions on chromosomes 7q32 and 11p15 with several
complex disorders, including type 2 diabetes [52].
In our own work, we have found parent of origin
effects on linkage in chromosome regions 10p12 and
12q24, where the linkage signal is due entirely to
maternal transmission [53]. Chromosome 12q24 was
one of the best supported linkage results in a meta-
analysis [27], which seems to indicate that the linkage
signal is detectible even if parent of origin is not
modeled in the analyses. The chromosome 10p12
region (19.4–33.3 Mb) is homologous to a largely over-
lapping segment of mouse chromosome 2A3
(15–23 Mb) that has been predicted to be imprinted
based on a machine learning model [54]. Two genes in
this region have previously been associated
with obesity, glutamate decarboxylase 2 (GAD2) and
G protein receptor 158 (GPR158) [55]. The concord-
ance is intriguing, although imprinting mechanisms
remain to be identified through molecular studies.
A further suggestion of imprinting effects in humans
is our recent finding of a CNV deletion of a region of
chromosome 4 including the NAP1L5 gene [21]. The
gene is normally expressed only on the paternal
chromosome, which is deleted, apparently leading to
an absence of gene expression.
Environmentally induced epigenetic modification
has been recognized in cancer for some time, but
275
 Chapter 21: Genetics and common human obesity

a role in complex disorders such as obesity has only

recently begun to be examined at a genomic level.
Applications: prevention and therapy
One goal of genetic research, whether stated or impli-
However, indirect evidence demonstrating environ-
cit, is that findings will eventually make it possible
mental effects on risk for obesity has been known
to use genotype to make decisions about appropriate
for some time. For example, an early study found
approaches to prevention and therapy. The nature of
increased rates of obesity in men who had been
the genetics of human obesity complicates its applica-
in utero or neonatal during the height of the Dutch
tion, particularly in identifying individuals most at
famine of 1944–45 [56]. Other studies of this type also
risk. Odds ratios for most variants will be even
have found that maternal malnutrition contributes to
smaller than for FTO (about 1.65) and have been
risk for obesity and other aspects of the metabolic
estimated to be 1.2 or lower. Prediction will therefore
syndrome [57]. Another study [58] found that pre-
involve only small increments in risk. In most cases,
natal exposure to maternal diabetes increased the risk
familial obesity will continue to be the best predictor
for obesity in Pima Indians. Animal studies similarly
of risk. This difficulty will not only limit application
have shown that maternal exposure to malnutrition,
but can also raise ethical concerns in providing risk
high fat diets, stress and other factors increase risk
assessments to individuals who may never develop
for obesity and the metabolic syndrome. It is of
obesity or become overweight for different reasons.
some interest that both under- and over-nutrition
While overall heritability is substantial, the contribu-
during fetal development can increase risk [59].
tion of individual genes or genotypes is likely to be
It came as something of a surprise that epigenetic
very small relative to the major environmental influ-
effects induced by prenatal diet can be inherited not
ences of diet and lifestyle.
only in a cell lineage but also across generations, for
The identification of protective genes may have
example, by persistent epigenetic marks in sperm [60]
the earliest application in the form of more individu-
that are associated with body weight and resistance to
alized pharmacological treatment, for example, iden-
dietary obesity.
tifying individuals with resistance to drug-induced
More recent studies of humans have also focused
weight gain. To do so, it is not necessary to identify
on epigenetic changes associated with prenatal expos-
genes involved in etiology, only those genes that dir-
ure. A follow-up study of the Dutch famine cohort,
ectly influence drug effectiveness or side effects.
for example, found that exposure indeed led to
Research in others areas have already made it possible
decreased methylation of the imprinted IGF2 gene
to tailor medication to individual genotype, particu-
[61]. Gene expression differences in monozygotic
larly for cancers. Response to tamoxifen treatment
twins discordant for obesity also suggest the possibil-
for breast cancer, for example, appears to be ineffect-
ity of epigenetic modification [62]. While overall
ive in 5–8% of women with a variant of the CYP2D6
differences in expression could be state dependent,
gene [65]. With regard to obesity, several genes have
mitochondrial DNA copy number differences in adi-
been identified that may influence drug induced
pose tissue of discordant twins are consistent with
weight gain, for example, due to olanzapine, including
epigenetic effects.
PMCH, 5-HT2A, ADRA2A, and PKHD1 [66]. In addi-
The obesity state affects expression of many
tion, SLC6A2 and GRIN1 have been associated with
genes, with perhaps as many as 17 000 transcripts
weight loss in response to norepinephrine/dopamine
related to BMI in adipose tissue according to one
transporter inhibitors [67]. Further research will
estimate [63]. Gene expression in normal weight
be needed before genomic screening is practical on a
animals has also been related to later obesity. Inbred
large scale but applications may be generally available
C57BL/6J mice are susceptible to diet induced
in the not too distant future.
obesity, but there is variation in adiposity from an
early age and the differences are maintained under
both high-fat and restricted low-fat diets [64]. Micro- What lies ahead
array analysis found parallel pre-obesity differences in A focus on clarifying the large discrepancy between
expression of genes in several known metabolic path- apparent heritability and variance accounted for by
ways. The causes of the expression differences are individual gene variants, the so called “missing herit-
unclear but could be due to prenatal or early postnatal ability”, is likely to remain a preoccupation in the
environment. near future. As one might expect, there are differences

276
 Chapter 21: Genetics and common human obesity

Table 21.1 Candidate genes for body mass index (BMI)/obesity

and fat distribution (bold) from genome-wide association TFAP2B 6 50,911,009
(GWA) studies. LRRN6C 9 28,404,339
Gene Chromosome Location RPL27A 11 8,561,169
NEGR1 1 72,585,028 BDNF 11 27,682,562
TNNI3K 1 74,764,232 MTCH2 11 47,607,569
1 96,717,385 FAIM2 12 48,533,735
1 176,156,103 MTIF3 13 26,918,180
TMEM18 2 612,827 PRKD1 14 29,584,863
RBJ NRXN3 14 79,006,717
FANCL MAP2K5 15 65,873,892
LRP1B GPRC5B 16 19,841,101
CADM2 SH2B1 16 28,793,160
ETV5 FTO 16 52,361,075
GNPDA2 4 44,877,284 MC4R 18 55,990,749
SLC39A8 4 103,407,732 KCTD15 19 39,001,372
FLJ35779 5 75,050,998 QPCTL 19 50,894,012
ZNF608 5 124,360,002 TMEM160 19 52,260,843
NUDT3 6 34,410,847 Summarized from [39].

in opinion about causes as well as the best
approaches to identify causal mechanisms [68].
Genomic approaches will surely detect other vari-
ants, both rare and common, that have small, incre-
mental influence on risk. Both targeted and whole
genome exon sequencing will identify rare coding
and splice variants. Whole genome sequencing shows
particular promise because of its comprehensive cov-
erage of noncoding regions. Quantitative methods
are constantly under development as well, for
example, haplotype and shared segment analyses of
concordant and discordant relatives, extreme cases
and controls [69, 70]. These methods are less costly
and in some cases may be applied with existing
GWA studies data. Environmental influences may
be better understood by the identification of inter-
actions with specific, measured genotypes. New
genes could provide additional targets for pharmaco-
logical intervention. Genotypes at these loci may
be used in therapeutic interventions through know-
ledge of their influence on drug effectiveness or side
effects (Table 21.1).

References 5. Bleich S, et al. Annu Rev Public

1. Flegal KM, et al. JAMA
2010;303:235–241.
2. Hodge AM, et al. Int J Obes
Relat Metab Disord 1994;
18:419–428.
3. Price RA, et al. Am J Phys
Anthropol 1993;92:473–479.
4. Sugarman JR, et al. Am J Clin Nutr
1990;52:960–966.
Health 2008;29:273–295.
6. Kumanyika SK. Annu Rev
Public Health 2008;
29:297–302.
7. Silventoinen K, et al. Int J Obes
Relat Metab Disord 2004;
28:710–718.
8. Maes HH, et al. Behav Genet
1997;27:325–351.
9. Malis C, et al. Obes Res
2005;13:2139–2145.
10. Stunkard AJ, et al. JAMA
1986;256:51–54.
11. Bouchard C, et al. J Nutr
1997;127:S943–S947.
12. Lu Qi YAC. Nutrition Reviews
2008;66:684–694.
13. Farooqi S, et al. Endocr Rev 2006;
27:710–718.

277
 Chapter 21: Genetics and common human obesity
14. Rankinen T, et al. Obesity 2006;
14:529–644.
15. Bodmer W, et al. Nat Genet 2008;
40:695–701.
16. Iyengar SK, et al. Methods Mol
Biol 2007;376:71–84.
17. Cooper GM, et al. Nat Genet 2007;
39:S22–S29.
18. Wong KK, et al. Am J Hum Genet
2007;80:91–104.
19. Cook Jr EH, et al. Nature
2008;455:919–923.
20. Estivill X, et al. PLoS Genetics
2007;3:e190.
21. Wang K, et al. Diabetes 2010;
59(10):2690–2694.
22. Sha BY, et al. J Hum Genet
2009;54:199–202.
23. Bochukova EG, et al. Nature
2010;463:666–670.
24. Walters RG, et al. Nature
2010;463:671–675.
25. Glessner JT, et al. Am J Hum
Genet 2010;87:661–666.
26. Jarick I, et al. Human
Molecular Genetics 2011;
20:840–852.
27. Saunders CL, et al. Obesity (Silver
Spring) 2007;15:2263–2275.
28. Sham PC, et al. Am J Human
Genetics 2000;66:
1616–1630.
29. Risch N, et al. Science 1996;
273:1516–1517.
30. International HapMap
Consortium. Nature 2007;
449:851–861.
31. The Wellcome Trust Case
Control Consortium. Nature
2007;447:661–678.
32. Loos RJ, et al. Nat Genet
2008;40:768–775.
278
33. Lindgren CM, et al. PLoS Genet
2009;5:e1000508.
34. Scherag A, et al. PLoS Genet
2010;6:e1000916.
35. Thorleifsson G, et al. Nat Genet
2009;41:18–24.
36. Willer CJ, et al. Nat Genet
2009;41:25–34.
37. Heard-Costa NL, et al.
PLoS Genet 2009;5:
e1000539.
38. Heid IM, et al. Nat Genet
2010;42:949–960.
39. Speliotes EK, et al. Nat Genet
2010;42:937–948.
40. Gudbjartsson DF, et al. Nat
Genet 2008;40:609–615.
41. Lettre G, et al. Nat Genet
2008;40:584–591.
42. Weedon MN, et al. Nat Genet
2008;40:575–583.
43. Visscher PM. Nat Genet
2008;40:489–490.
44. Fisher RA. Transactions of the
Royal Society of Edinburg
1918;52:399–433.
45. Feinberg AP, et al. Nature
2007;447:433–440.
46. Reik W, et al. Nat Rev Genet
2001;2:21–32.
47. Wilkinson LS, et al. Nat Rev
Neurosci 2007;8:
832–843.
48. Wood AJ, et al. PLoS Genet
2006;2:e147.
49. Vrana PB, et al. Nat Genet
1998;20:362–365.
50. Nezer C, et al. Nat Genet
1999;21:155–156.
51. de Koning DJ, et al. Proc
Natl Acad Sci U S A 2000;
97:7947–7950.
52. Kong A, et al. Nature
2009;462:868–874.
53. Dong C, et al. Am J Hum Genet
2005;76:427–437.
54. Luedi PP, et al. Genome Res
2005;15:875–884.
55. Bell CG, et al. Nat Rev Genet
2005;6:221–234.
56. Ravelli GP, et al. N Engl J Med
1976;295:349–353.
57. Hales CN, et al. Br Med Bull
2001;60:5–20.
58. Pettitt DJ, et al. N Engl J Med
1983;308:242–245.
59. Tamashiro KL, et al. Diabetes
2009;58:1116–1125.
60. Yazbek SN, et al. Hum Mol Genet
2010;19:4134–4144.
61. Heijmans BT, et al. Proc Natl
Acad Sci U S A 2008;105:
17046–17049.
62. Pietilainen KH, et al. PLoS Med
2008;5:e51.
63. Chen Y, et al. Nature 2008;
452:429–435.
64. Koza RA, et al. PLoS Genet 2006;
2:e81.
65. Weinshilboum R. Med Biol
2008;630:220–231.
66. Muller DJ, et al. Pharma-
cogenomics 2006;
7:863–887.
67. Spraggs CF, et al. Pharma-
cogenet Genomics 2005;
15:883–889.
68. Eichler EE, et al. Nat Rev Genet
2010;11:446–450.
69. Thomas A, et al. Annals of
Human Genetics 2008;72:
279–287.
70. Zhu X, et al. Genetic Epidemiology
2010;34:171–187.
 Alcoholism
Chapter
22 Howard J. Edenberg
Background
Alcohol use disorders (AUDs) are very common
around the world, and cause a tremendous burden
of disability and death [1]. AUDs can be defined as
“maladaptive patterns of [alcohol] use leading to clin-
ically significant impairment or stress” [2]. Current
diagnostic criteria, both from the Diagnostic and
Statistical Manual on Mental Disorders (DSM-IV)
[3], used in the United States, and from the Inter-
national Classification of Diseases (ICD-10) [4], used
elsewhere, divide AUDs into alcohol dependence and
abuse/harmful use. Criteria for alcohol dependence
(alcoholism) are very similar in the two diagnostic
systems, based on a syndromic definition [5]. The
criteria include withdrawal, tolerance, loss of control
of drinking, impaired work or social activities, and
continued use despite known problems. A diagnosis
of alcohol dependence requires that an individual
endorse at least three out of seven DSM-IV criteria
(Table 22.1) or three out of six comparable ICD-10
criteria. Fewer individuals meet ICD-10 criteria than
meet DSM-IV criteria [6, 7]. In spite of the hetero-
geneity in this syndromic definition, the diagnosis of
alcohol dependence is one of the most reliable in the
DSM-IV [6, 8, 9]. The reliability of an abuse diagnosis
is much lower [6, 9, 10].
This article will focus on alcohol dependence
(alcoholism), because the diagnosis is more reliable
and because most of the genetic data focus on
dependence. US data from the 2001–2002 National
Epidemiological Survey on Alcohol and Related Con-
ditions (NESARC) showed a 12-month prevalence of
3.8% for alcohol dependence defined by DSM-IV:
5.4% in men and 2.3% in women [11]. Lifetime preva-
lence of alcohol dependence is 12.5% in the United
States.
Principles of Psychiatric Genetics, eds John I. Nurnberger, Jr. and Wade H. Berrettini. Published by Cambridge University
Press. © Cambridge University Press 2012.
Table 22.1 DSM-IV diagnostic criteria for alcohol dependence
requires meeting 3 or more of the following criteria during
a 12-month period.
1. Tolerance
2. Withdrawal signs or symptoms
3. Drinking more than intended
4. Unsuccessful attempts to cut down on use
5. Excessive time related to alcohol (obtaining it,
hangover)
6. Impaired social or work activities due to alcohol
7. Use despite physical or psychological consequences
Alcoholism is a genetic disease with
a necessary environmental component
Alcoholism is a complex genetic disease, with convin-
cing evidence that variations in both genes and envir-
onment contribute to differences among individuals
in risk. Multiple lines of evidence converge to support
the idea of a genetic contribution to the risk. These
include adoption studies that demonstrate the adoptees
more closely resemble their biological parents than
their adoptive parents in risk [12–15]. They also
include twin studies, showing greater concordance in
alcoholism between monozygotic twins (MZ) (who
share all of their genes) than between dizigotic twins
(DZ) (who share only half of their genes) [14, 16–18].
The fact that MZ co-twins of alcoholics are not all
alcoholics is a clear demonstration that the disease is
not determined only by genes. A third line of evidence
is that aspects of alcoholism, including strong prefer-
ence for alcohol over water, willingness to work for
alcohol, sensitivity to the hypnotic or activating effects
of alcohol and to withdrawal, and demonstrations that
279
 Chapter 22: Alcoholism

alcohol is rewarding even in the presence of food and (followed up with association studies of variants
water, can be modeled in selectively bred rodent lines within the linked regions), and genome-wide associ-
[19–23]. A fourth line of evidence comes from early ation studies. This chapter cannot discuss all of the
genetic studies in humans that demonstrated genetic rapidly growing literature in the area; illustrative
variations in alcohol metabolism affect risk for examples are chosen, often from the work of the
alcoholism [24–27] (and see below). Although there Collaborative Study on the Genetics of Alcoholism
might be caveats to any one line of evidence, the (COGA).
convergence of data provides overwhelming evidence
that genetic variations contribute to individual vari-
ations in the risk for alcoholism. The lack of a simple
Candidate gene studies
The earliest genetic association studies in alcoholism
pattern of inheritance indicates that the genetic risk
were candidate gene studies targeting coding vari-
results from the combined contributions of many
ations in the genes that metabolize alcohol. Most
genes, and probably in part from gene
gene and
ingested alcohol is metabolized in the liver, primarily
gene
environment interactions.
in a two step reaction: oxidation to acetaldehyde,
There is a necessary environmental component to
which is further oxidized to acetate. The first step,
the disease: consumption of alcohol. Absent that,
oxidation to acetaldehyde, can be catalyzed by alcohol
underlying genetic vulnerabilities may surface in
dehydrogenases, cytochrome P450s, and catalase.
other problems or diseases, but not in alcoholism.
The great majority of the alcohol is oxidized by the
Unlike other psychoactive drugs, ethanol is ingested
alcohol dehydrogenases (ADH), with the accompany-
in large amounts; legal intoxication in the United
ing reduction of NADþ to NADH (Figure 22.1). The
States is defined as a blood alcohol concentration of
second step, oxidation of acetaldehyde to acetate,
0.08%. Drinking that allows blood alcohol to reach
is catalyzed primarily by aldehyde dehydrogenases
0.08% or above has been defined as binge drinking
(ALDH), with reduction of another molecule of
[28]. Although there is significant variation among
NADþ to NADH. Humans have seven ADHs and
individuals, a typical 170 pound man would generally
two ALDHs that catalyze most of the ethanol metab-
reach this level after drinking about 5 standard drinks
olism. Variations in the genes encoding enzymes of
within 2 hours and an average women would reach
alcohol metabolism have long been known to affect
this level after ingesting about 4 drinks, due to differ-
the risk for alcoholism [24–27, 30].
ences in weight and body composition. (A standard
A variation in ALDH2 has dramatic effects on
drink in the United States contains 14 g of pure
alcohol metabolism. The ALDH2*2 allele encodes a
ethanol, and is approximately equivalent to 12 ounces
nearly inactive subunit of the mitochondrial aldehyde
of beer, 5 ounces of table wine, or 1.5 ounces of 80-proof
dehydrogenase 2 that is responsible for much of the
spirits.) Alcohol consumption must generally be in
oxidation of ethanol. Presence of a single ALDH2*2
the binge range and frequently repeated to develop
allele renders the ALDH2 catalytic activity (measured
the problems that result in diagnosis of alcohol
in vitro) below the usual limit of detection [31]. In
dependence. The risk of meeting criteria for alcohol
individuals with a single copy of the inactive ALDH2*2
dependence rises nearly linearly with the frequency of
allele consumption of even small amounts of alcohol
binge drinking [29]. Data from the National Epidemio-
causes a dramatic rise in acetaldehyde in blood, which
logic Survey on Alcohol and Related Conditions
triggers a highly aversive reaction similar to that caused
(NESARC) indicate that the number of criteria for
abuse and dependence that are met rise with the fre-
when a patient taking disulfiram (Antabuse ) drinks
alcohol; the reaction includes flushing, tachycardia,
®
quency of binge drinking [2, 29].
This necessary environmental factor, alcohol, inter-
acts with the underlying genetics to result in the disease Alcohol Acetaldehyde Acetate
of alcoholism. Variations in the environment related
ADH ALDH
to alcohol accessibility, price, and social norms, there-
NAD+ NADH + H+ NAD+ NADH + H+
fore, can affect the prevalence of alcoholism.
There are three main approaches to identification Figure 22.1 The primary pathway of alcohol metabolism in the
liver is oxidation by alcohol dehydrogenases (ADH) and aldehyde
of variations in specific genes that affect alcohol dehydrogenases (ALDH) enzymes, with acetaldehyde as the
dependence: candidate gene studies, linkage studies intermediate.

280

ADH4 ADH1B
ADH5 ADH6 ADH1A ADH1C ADH7
Figure 22.2 Genomic arrangement of human ADH genes on
chromosome 4q.
and nausea. This aversive reaction greatly reduces
their propensity to drink, the amount consumed
per occasion, and their risk for alcoholism [24, 26,
27, 31–35]. It is not totally protective: some individ-
uals continue to drink despite the high acetaldehyde
levels. But for someone with a single ALDH2*2 allele,
the relative risk of alcoholism range from about 0.13
to 0.40 in different Asian populations [35, 36]. In vivo,
having two copies of the ALDH2*2 allele leads to
an even more severe reaction and is nearly totally
protective against alcoholism; among thousands of
cases reported in the literature, there are only three
alcoholics homozygous for ALDH2*2 [35, 36].
Variations in ADHs that catalyze the first step in
ethanol metabolism also strongly affect the risk for
alcoholism. Humans have seven ADHs that arose
from repeated gene duplications; the genes encoding
them are clustered in a small region of chromosome 4
(Figure 22.2). The kinetic properties of these enzymes
suggest that at low levels of alcohol, enzymes encoded
by ADH1A, ADH1B, and ADH1C play the major role
in metabolism; ADH1B is the ADH present at highest
levels in the adult liver and presumably contributes
most. When ethanol is present at higher levels (intoxi-
cating) the enzyme encoded by ADH4 makes an
increasing contribution. ADH7, located in the
esophagus and stomach lining, can contribute to “first
pass” metabolism of ethanol because ethanol is at
very high concentrations in stomach during drinking.
Variants in ADH1B and ADH1C that increase the
rates at which the enzymes they encode oxidize etha-
nol reduce the risk for alcoholism [24, 26, 27, 35–40].
The strongest effects are due to variants in ADH1B
that are relatively common in Asians (ADH1B*2, in
which arginine 48 is replaced with histidine) and in
people of African ancestry (ADH1B*3, in which
arginine 370 is replaced with cysteine). Blood acetal-
dehyde levels do not dramatically rise in individuals
with these variants, and there is no severe flushing
reaction comparable to that in individuals with the
ALDH2*2 allele. Nevertheless, the protection against
alcoholism provided by the ADH1B*2 allele is very
strong, with relative risks approximately 0.18–0.26 for
1 allele and 0.10–0.14 for 2 (in Asian subjects
Chapter 22: Alcoholism
homozygous for ALDH2*1) [35, 36, 40, 41]. Neither
of these alleles are common among individuals of
European ancestry, although ADH1B*2 is found at
moderate frequency in people of Jewish ancestry [42,
43]. The low allele frequency makes studies more diffi-
cult, but ADH1B*2 also appears protective in individ-
uals of European descent (with relative risk about 0.5)
[38, 40, 44]. A recent large-scale study has conclusively
demonstrated protective effects in Europeans [45].
Recent genetic studies have shown that variations in
ADH4 play a significant role in affecting risk for alco-
holism, as do noncoding variations in ADH1A and
ADH1B [37]. Other variations in the ADH region were
also shown to affect alcohol metabolism [46], includ-
ing variations in and near ADH7 [47] ADH1A,
ADH1B, ADH1C, and ADH4 [48]. Clearly there is
more to be learned about the contributions of the
ADH genes to both alcohol metabolism and alcoholism.
Despite the fact that blood acetaldehyde levels
do not rise substantially in individuals with the
ADH1B*2 or ADH1B*3 alleles, it is generally thought
that both the more rapid generation of acetaldehyde
by these ADHs and the reduction in the rate of
elimination of acetaldehyde by the inactive ALDH
affect the risk for alcoholism by at least transiently
increasing acetaldehyde levels in the liver, triggering
aversive reactions that reduce excessive drinking.
A combination of the strongly protective alleles at
both ADH1B and ALDH2 loci is synergistic, and
drops relative risk of alcoholism in Asians to about
1–10% [35, 36]. Thus variations in the ADH and
ALDH genes play strong roles in affecting the risk
for alcoholism. These are among the strongest and
best replicated findings in the genetics of complex
diseases. Despite their strong impact on some popu-
lations, the coding variations in ADH and ALDH
genes do not explain a large fraction of the differences
in risk in populations of European ancestry in which
the alleles with the strongest effects are rare.
There have been many other candidate gene
studies, some with markedly mixed results. A full
description of these is beyond the scope of this article,
but several will be discussed as illustrative of the
field. Perhaps the most famous is the association of
the Taq1A allele of the dopamine D2 receptor gene
(DRD2-Taq1A; rs1800497) with alcoholism, first
reported in a small study in 1990 [49]. There have
been numerous attempts to replicate this, and
although most have been negative, e.g. [50–52], one
recent meta-analysis suggests a small but significant
281
 Chapter 22: Alcoholism
effect that might be due in part to publication bias [53].
The “DRD2-Taq1A” polymorphism actually lies
within an adjacent gene, ANKK1 (ankyrin repeat and
kinase domain containing 1) [54]. A recent family
study showed that although rs1800497 was not signifi-
cant, a different allele at DRD2 (rs6277, a synonymous
single nucleotide polymorphisms [SNPs]) was [55].
Another family-based analysis showed that the associ-
ation in this region with alcohol dependence is
strongest in the region of ANKK1 that is not in linkage
disequilibrium (LD) with DRD2 [50], and another
family and case-control study suggested the evidence
for association with alcoholism in this region is in the
NCAM1, TTC12, and ANKK1 genes [56, 57]. There
remains much complexity in interpreting these data.
The opioid system has been implicated in addic-
tions, not only for opiates but also for alcohol [58,
59]. There are three main opioid receptors and three
main genes encoding endogenous ligands. OPRM1
encodes the mu receptor (MOR), and POMC encodes
its primary ligands; OPRK1 encodes the kappa recep-
tor (KOR) and PDYN its primary ligands; OPRD1
encodes the delta receptor (DOR) and PENK its pri-
mary ligands. The mu receptor has been most widely
studied, particularly a coding variation (Asn40Asp); a
recent meta-analysis of those studies concluded that
there was no significant evidence for association [60].
COGA examined all six genes from this system, and
found that variations in the kappa system were asso-
ciated with alcoholism [61], but variations in the mu
or delta system were not [62], nor were variations in
the related nociceptin system [63]. It was particularly
intriguing that variations in the genes encoding
both the receptor (OPRK1) and its ligand (PDYN)
were associated. Follow-up molecular studies demon-
strated that one associated polymorphism affected the
level of expression of OPRK1 [64].
NPY is a candidate gene based upon studies of a
rat model of alcohol preference [65–67], knockout
mice [68], and some but not all human studies, e.g.
[69–71]. NPY itself is on human chromosome 7 and
was not significantly associated with alcoholism in the
COGA sample [72]. Three NPY-receptor genes on
chromosome 4 were also evaluated, and two of these
gave evidence of association: NPY2R was associated
with alcohol dependence and alcohol withdrawal
symptoms, and NPY5R was associated with alcohol
withdrawal characterized by seizures [72].
There are many other potential candidate genes,
including genes in nearly all neuronal receptor/ligand
282
systems. When variations in one gene is found to affect
the risk for alcoholism, it is reasonable to examine other
genes in the system(s) to which it belongs; this has been
useful, for example, in following up initial findings with
GABRA2 [73] in other GABA-receptor genes [74, 75].
There are, however, advantages to taking an unbiased
approach toward identifying genes that affect risk, such
as genetic linkage or whole genome association studies.
Genetic linkage studies
Genetic linkage studies provide an unbiased approach
toward identifying genetic variations that affect the
risk for a complex genetic disease such as alcoholism.
Linkage studies follow the inheritance of particular
regions of the genome within families. The regions
are tagged by some readily assayable marker; micro-
satellite markers were the most widely used when the
studies on alcoholism began, but SNPs are now easier
to assay. Linkage studies test for significantly excess
(or reduced) transmission of particular regions of the
genome to affected family members. Because nearby
sites along a chromosome are usually transmitted
together, a relatively small number of markers can
report on the genome. This is a technical advantage
for linkage studies, but the drawback is that linkage
studies generally identify relatively large regions, usu-
ally between 20 and 50 million base pairs, that contain
many genes. Because linkage studies look within indi-
vidual families, they can identify a region in which
different rare alleles affect the risk in different families.
The earliest whole genome linkage studies on
alcoholism were reported in 1998 [76, 77]. The Col-
laborative Study on the Genetics of Alcoholism
(COGA) recruited families of individuals in treatment
for alcoholism at six sites across the United States, and
did genetic studies on those families in which at least
three first-degree relatives met criteria for alcohol
dependence [77]. The initial linkage study encom-
passed 987 individuals from 105 families, and reported
suggestive evidence for linkage with alcoholism on
chromosomes 1 and 7, weaker evidence on chromo-
some 2, and linkage with a protective phenotype on
chromosome 4q [77]. Long et al. [76] studied 172
sibling pairs from a southwest Native American popu-
lation, and reported evidence for linkage on chromo-
some 11p, 4p and for several markers in the ADH
region of 4q. A follow-up study of an additional 157
families from the COGA project supported their original
findings on chromosomes 1 and 7 and also revealed

some evidence for linkage on chromosome 3 [78].
A smaller study provided support for the linkage
on chromosome 1 [79]. Analysis combining the diag-
nostic phenotype with an electrophysiological vari-
able (amplitude of the P300 component of the
event-related potential) gave the strongest evidence of
linkage in a broad region of chromosome 4q that
encompassed the ADH genes [80]. A quantitative trait,
maximum drinks in a 24 hour period, also showed
linkage on chromosome 4q [81]. Results from the Irish
Affected Sib Pair Study of Alcohol Dependence again
showed linkage to a broad region of chromosome 4q,
and weaker, suggestive evidence on chromosomes 1q,
13q, and 22q for alcohol dependence, on 2q, 9q, and
18p for symptom count [82], on chromosome 9 for
age at onset, on chromosomes 1 and 11 for initial
response to alcohol, 1, 6, and 22 for tolerance, on
chromosomes 12 and 18 for maximum drinks, and
on chromosome 2 for withdrawal symptoms [83].
A more recent linkage study of alcohol dependence
in a set of African-American families originally ascer-
tained for cocaine or opioid dependence showed evi-
dence for linkage to chromosome 10q [84]. As noted
above, these linkage studies point to chromosomal
regions, rather than specific genes.
Follow-up of linkage studies
A productive approach to identifying genes that affect
the risk for alcoholism is to follow-up linkage studies
with association studies of variants within the linked
regions. One can either analyze variations spanning
the entire linkage region in a systematic manner, based
upon LD, or target candidate genes within the region.
The systematic approach generally requires more gen-
otyping and correction for more multiple testing, but
has the advantage of being relatively unbiased.
A disadvantage is that the location of the peak of
linkage is not necessarily centered on the key gene(s)
contributing to the linkage, so it is possible to miss the
gene contributing to the linkage signal if too narrow a
region is analyzed. The candidate gene approach lever-
ages prior data and biological understanding to reduce
the number of genes that are tested, but might miss
potentially important genes for which there is no prior
evidence or hypothesized role in alcoholism.
These issues will be illustrated by discussing some
results from COGA. There was some evidence from
the initial COGA analysis for linkage to markers on
chromosome 4p around a GABA receptor gene [77],
Chapter 22: Alcoholism
and results from a linkage study in Southwest Native
Americans showed the strongest linkage there [76].
Our interest in the region was greatly enhanced by a
very strong linkage of an electrophysiological pheno-
type believed to be related to susceptibility to alcohol-
ism, the amplitude of the b-EEG [85]. A cluster of 4
GABAA receptor genes was located in the middle of
that linkage peak, and was extensively analyzed. Many
variations and haplotypes in GABRA2, encoding the
a2 subunit of the GABAA receptor, were associated
with alcohol dependence as well as with the b-EEG
phenotype [73]. This has since been confirmed in
many studies of different populations [86–92]. The
association in the COGA sample is with the most
severe half of the alcoholic subjects, as defined by
comorbid drug dependence [93], which also correl-
ates with early onset of alcoholism and many other
measures of severity [94]. Other studies have shown
association also with the GABRG1 gene, adjacent to
GABRA2 and in LD with it [96].
One of the large linkage peaks in the COGA
sample was on chromosome 4q [77, 80]; this region
was also implicated in the Irish study [82]. The ADH
genes lie under this peak. As discussed above, these
were targeted by many SNPs, and associations found
with SNPs in ADH4, ADH1A, and ADH1B [37].
These findings were supported and extended by other
groups (see above). However, we hypothesized that a
broad linkage peak was likely to reflect the combined
contribution of several genes, and continued our ana-
lyses of other genes within this region. COGA has,
thus far, identified several additional genes within
this linkage peak in which variations affect alcohol
dependence or a closely-related phenotype. SNCA,
encoding a-synuclein, was a top candidate gene based
upon studies in a rat model of alcohol preference [65,
96, 97] and on data from monkeys [98] and humans
[98–101]. In the COGA subjects, variations in SNCA
were not significantly associated with dependence per
se, but were associated with craving for alcohol [102].
Continuing our analysis of genes in the chromosome
4 linkage peak, we demonstrated that TACR3, encod-
ing the tachykinin 3 (neurokinin B) receptor, was
associated with alcohol dependence, as was NFKB1
[103], encoding a subunit of the ubiquitous transcrip-
tion factor NF-kB, which regulates many genes in the
brain. The association of both TACR3 and NFKB1
were strongest with the most severely affected subjects
[103, 104]. Thus the hypothesis that multiple genes
contribute to this linkage peak was supported.
283
 Chapter 22: Alcoholism
A second major linkage peak in the COGA
sample was on chromosome 7q [77, 105–107].
A systematic screen across an 18 Mb (2-LOD) inter-
val was carried out, selecting SNPs that efficiently
report on the variations within this region based
upon patterns of LD [108]. Several SNPs gave
significant or suggestive evidence of association
with alcoholism; the most consistent evidence was
for several SNPs in the ACN9 gene [108], related to
gluconeogenesis and the assimilation of ethanol
and acetate. Additional genes within this region are
being examined.
Genome-wide association studies
Recently, genome-wide association studies (GWAS)
have largely replaced linkage studies [109, 110]. These
studies take advantage of modern multiplexed
methods for simultaneously genotyping 1 million
SNPs or more. They allow association testing of a
very large fraction of the genome, and have been
successful for several complex diseases [109]. There
are limitations, however. The most obvious is the risk
of false positives due to the extraordinary amount of
multiple testing; less obvious is that the corrections
used to avoid false positives are likely to lead to many
false negatives. The problem of multiple testing leads
to a requirement for very large sample sizes, with
cases and controls very well matched for ethnicity.
The use of correction factors using ancestry informa-
tive markers can accommodate multiple ethnicities in
the same study. Another limitation is that GWAS are
designed to find relatively common polymorphisms
that contribute to the risk for a disease; if a collection
of rare variants in a gene can independently increase
risk, this will often be missed by GWAS (but can be
captured by linkage studies).
To date, only one GWAS on alcoholism has been
published. Treutlein et al. [111] studied 487 male
alcoholic inpatients from Germany and 1358 controls,
and followed up the more significant findings in
another group of 1024 patients and 996 controls. No
SNP met criteria for genome-wide significance in the
initial study, but two did in the combined sample.
Other SNPs provided consistent evidence in the initial
sample and follow up, at less significant levels. The
need for large samples and replications means that
results expected within the next few years will be
important for assessing which SNPs are truly associ-
ated with alcohol dependence.
284
Gene × environment interaction
A substantial gene
environment interaction was
shown dramatically in the Japanese population by
Higuchi [25]. The degree of protection against alco-
holism afforded by the ALDH2*2 allele changed sig-
nificantly between the years 1979 and 1992: in that
time, the fraction of Japanese alcoholics carrying the
ALDH2*2 allele in heterozygous form increased
from 2.5 to 13% [25]. This time was far too small
for any change in the underlying frequency of the
polymorphism in the Japanese population as a whole,
so the only explanation is that its protective effect was
reduced, presumably by sociological changes leading
toward more alcohol consumption. In that same
study, no alcoholics were found to be homozygous
for ALDH2*2 (although 120 would have been
expected based on the allele frequency), suggesting
that the protection afforded by having two copies of
the ALDH2*2 allele is much stronger and was not as
susceptible to the environmental changes [25].
Genes and environment in treatment
It is likely that particular genes and combinations
of genes will differentially affect risk for different
manifestations of alcoholism, such as co-occurrence
with other disorders, and that different combinations
of genes will also affect the response to particular
treatments. There is already evidence that variations
in some genes primarily affect alcoholism that shows
early onset and comorbidity with other drugs of abuse
(e.g. GABRA2, ADH4, CHRM2, NFKB1) [73, 93, 95,
103, 104]. As we learn more about the genetic under-
pinnings of alcoholism, we are likely to find that differ-
ent combinations of genetic variants lead to different
responses to different treatments, which will improve
our ability to design individualized therapies [112].
Summary
There has been much progress in the genetics of
alcohol dependence. Variations in several genes have
clearly been shown to affect the risk for developing
alcohol dependence. Certain variations in ADH and
ALDH genes have very strong effects on the risk for
alcoholism. Variations in other genes appear to have a
much smaller effect on risk. In populations of Euro-
pean ancestry, in which the coding variations in ADH
and ALDH that have the strongest effects are uncom-
mon, most of the individual difference in risk is still
 Chapter 22: Alcoholism

unexplained, and probably reflects the summation of
many genes of small effect, along with gene
gene
and gene
environment interactions. Linkage studies
and their follow-up, along with candidate gene studies
and GWAS, are beginning to fill the gaps. Initial
findings must be confirmed in independent studies,
and much work remains to elucidate the mechanisms
involved. Nevertheless, with the new technologies and
larger samples being studied, progress should acceler-
ate. The future will involve studies of epigenetic
factors, copy number variants, and gene expression,
as well as tests for rare variants of large effect in
specific families [113].
It should be remembered that although variations in
genes clearly affects an individual’s risk for alcoholism,
the disease is not determined solely by genes. The envir-
onment and individual choices plays a major role.
Understanding the genetic contributions to risk should
lead to better understanding of the disease processes and
assist in tailoring treatments to individuals.
Acknowledgements
Related work in the author’s laboratory is supported
by grants from the National Institute of Alcohol
Abuse and Alcoholism, R37AA006460, U10AA008401,
U01AA016660, R21AA016301, P60AA007611.

References 15. Sigvardsson S, et al. Arch Gen

1. World Health Organization.
The World Health Report 2002:
Reducing Risks, Promoting Healthy
Life. Geneva: WHO; 2002.
2. Li TK, et al. Addiction 2007;
102:1522–1530.
3. American Psychiatric Association.
Diagnostic and Statistical Manual of
Mental Disorders, Fourth Edition.
Washington, DC: American
Psychiatric Association; 1994.
4. World Health Organization.
International Classification of
Disease, Tenth Edition. Geneva:
WHO; 1993.
5. Edwards G, et al. Br Med J 1976;
1(6017):1058–1061.
6. Grant BF. Alcohol Clin Exp Res
1996;20(8):1481–1488.
7. Schuckit MA, et al. Addiction
1994;89:1629–1638.
8. Schuckit MA, et al. Alcohol Clin
Exp Res 2002;26(7):980–987.
9. Hasin D, et al. Addiction 2006;
101 S1:59–75.
10. Grant BF, et al. Drug Alcohol
Depend 2007;86(2–3):154–66.
11. Grant BF, et al. Drug Alcohol
Depend 2004;74(3):223–234.
12. Bohman M, et al. Arch Gen
Psychiatry 1981;38(9):965–969.
13. Cloninger CR, et al. Arch Gen
Psychiatry 1981;38:861–868.
14. Heath AC. Alc Health Res World
1995;19:166–171.
Psychiatry 1996;53(8):681–687.
16. Heath AC, et al. Psychol Med
1997;27(6):1381–1396.
17. Kendler KS, et al. Am J Psychiatry
1994;151:707–715.
18. Pickens RW, et al. Arch Gen
Psychiatry 1991;48:19–28.
19. Buck KJ, et al. J Neurosci 1997;
17(10):3946–3955.
20. Crabbe JC, et al. Trends Neurosci
1999;22(4):173–179.
21. McBride WJ, et al. Crit Rev
Neurobiol 1998;12(4):339–369.
22. Murphy JM, et al. Behav Genet
2002;32(5):363–388.
23. Whatley VJ, et al. Alcohol Clin Exp
Res 1999;23(7):1262–1271.
24. Edenberg HJ. Alcohol Res Health
2007;30(1):5–13.
25. Higuchi S, et al. Lancet 1994;
343:741–742.
26. Hurley TD, et al. Pharma-
cogenomics: The Search
for Individualized Therapies.
New York: Wiley-VCH; 2002.
27. Thomasson HR, et al. Am J Hum
Genet 1991;48(4):677–681.
28. National Advisory Council on
Alcohol Abuse and Alcoholism.
Council Minutes: National
Advisory Council Meeting,
February 4–5; 2004.
29. Dawson DA, et al. Alcohol
Clin Exp Res 2005;29(5):
902–908.
30. Stamatoyannopoulos G, et al.
Am J Hum Genet 1975;27:
789–796.
31. Crabb DW, et al. J Clin Invest
1989;83(1):314–316.
32. Bosron WF, et al. Semin Liver Dis
1981;1:179–188.
33. Chen YC, et al. Pharmacogenet
Genomics 2009;19(8):588–599.
34. Harada S, et al. Lancet 1982;2:827.
35. Luczak SE, et al. Psychol Bull
2006;132(4):607–621.
36. Chen YC, et al. Alcohol Clin Exp
Res 1999;23:1853–1860.
37. Edenberg HJ, et al. Hum Mol
Genet 2006;15(9):1539–1549.
38. Macgregor S, et al. Hum Mol
Genet 2009;18(3):580–593.
39. Sherva R, et al. Alcohol Clin Exp
Res 2009;33(5):848–857.
40. Whitfield JB. Am J Hum Genet
2002;71(5):1247–1250; author
reply 1250–1251.
41. Li D, et al. Biol Psychiatry
2011;70:504–512.
42. Carr LG, et al. Am J Med Genet
2002;112(2):138–143.
43. Neumark YD, et al. J Stud Alcohol
1998;59(2):133–139.
44. Borras E, et al. Hepatol 2000;
31(4):984–989.
45. Bierut LJ, et al. Mol Psychiatry,
in press.
46. Birley AJ, et al. Behav Genet
2005;35(5):509–524.

285
 Chapter 22: Alcoholism
47. Birley AJ, et al. Hum Mol Genet
2008;17(2):179–189.
48. Birley AJ, et al. Hum Mol Genet
2009;18(8):1533–1542.
49. Blum K, et al. JAMA 1990;
263:2055–2060.
50. Dick DM, et al. Alcohol Clin Exp
Res 2007;31(10):1645–1653.
51. Edenberg HJ, et al. Alcohol Clin
Exp Res 1998;22(2):505–512.
52. Gelernter J, et al. JAMA 1993;
269:1673–1677.
53. Munafo MR, et al. Mol Psychiatry
2007;12(5):454–461.
54. Neville MJ, et al. Hum Mutat
2004;23(6):540–545.
55. Hill SY, et al. Am J Med Genet
B Neuropsychiatr Genet 2008;
147B(4):517–526.
56. Yang BZ, et al. Hum Mol Genet
2007;16(23):2844–2853.
57. Yang BZ, et al. Alcohol Clin Exp
Res 2008;32(12):2117–2127.
58. Kreek MJ. Mol Psychiatry 1996;
1(3):232–254.
59. Kreek MJ, et al. Neuromolecular
Med 2004;5(1):85–108.
60. Arias A, et al. Drug Alcohol
Depend 2006;83(3):262–268.
61. Xuei X, et al. Mol Psychiatry
2006;11(11):1016–1024.
62. Xuei X, et al. Am J Med Genet
B Neuropsychiatr Genet 2007;
144(7):877–884.
63. Xuei X, et al. Addict Biol 2008;
13(1):80–87.
64. Edenberg HJ, et al. Hum Mol
Genet 2008;17(12):1783–1789.
65. Carr LG, et al. Alcohol Clin Exp
Res 1998;22:884–887.
66. Hwang BH, et al. Alcohol Clin Exp
Res 1999;23(6):1023–1030.
67. Spence JP, et al. Neuroscience
2005;131(4):871–876.
68. Thiele TE, et al. Nature 1998;
396(6709):366–369.
69. Lappalainen J, et al. Arch Gen
Psychiatry 2002;59(9):825–831.
286
70. Mottagui-Tabar S, et al. Alcohol
Clin Exp Res 2005;29(5):702–707.
71. Zhu G, et al. Alcohol Clin Exp Res
2003;27(1):19–24.
72. Wetherill L, et al. Alcohol Clin Exp
Res 2008;32(12):2031–2040.
73. Edenberg HJ, et al. Am J Hum
Genet 2004;74(4):705–714.
74. Dick DM, et al. Alcohol Clin Exp
Res 2004;28(1):4–9.
75. Dick DM, et al. Alcohol Clin Exp
Res 2006;30(7):1101–1110.
76. Long JC, et al. Am J Med Genet
B Neuropsych Genet 1998;81:
216–221.
77. Reich T, et al. Am J Med Genet
1998;81(3):207–215.
78. Foroud T, et al. Alcohol Clin Exp
Res 2000;24(7):933–945.
79. Lappalainen J, et al. Mol
Psychiatry 2004;9(3):312–319.
80. Williams JT, et al. Am J
Hum Genet 1999;65(4):
1148–1160.
81. Saccone NL, et al. Am J Med Genet
2000;96(5):632–637.
82. Prescott CA, et al. Mol Psychiatry
2006;11(6):603–611.
83. Kuo PH, et al. Alcohol Clin Exp
Res 2006;30(11):1807–1816.
84. Gelernter J, et al. Biol Psychiatry
2009;65(2):111–115.
85. Porjesz B, et al. Proc Natl Acad Sci
U S A 2002;99:3729–3733.
86. Covault J, et al. Am J Med Genet
B Neuropsychiatr Genet 2004;
129(1):104–109.
87. Drgon T, et al. Am J Med Genet
B Neuropsychiatr Genet 2006;
141(8):854–860.
88. Enoch MA, et al. Neuro-
psychopharmacology 2009;34
(5):1245–1254.
89. Enoch MA, et al. Am J Med
Genet B Neuropsychiatr Genet
2006;141(6):599–607.
90. Fehr C, et al. Psychiatr Genet
2006;16(1):9–17.
91. Lappalainen J, et al. Alcohol Clin
Exp Res 2005;29(4):493–498.
92. Soyka M, et al. J Psychiatr Res
2008;42(3):184–191.
93. Agrawal A, et al. Behav Genet
2006;36(5):640–650.
94. Dick DM, et al. Addiction
2007;102(7):1131–1139.
95. Covault J, et al. Neuro-
psychopharmacology 2008;33
(4):837–848.
96. Liang T, et al. J Neurochem
2006;99(2):470–482.
97. Liang T, et al. Proc Natl
Acad Sci U S A 2003;100(8):
4690–4695.
98. Walker SJ, et al. Alcohol 2006;
38(1):1–4.
99. Bonsch D, et al. Alcohol Clin Exp
Res 2005;29(5):763–765.
100. Bonsch D, et al. Biol Psychiatry
2004;56(12):984–986.
101. Zhou W, et al. Brain Res 2002;
926(1–2):42–50.
102. Foroud T, et al. Alcohol Clin Exp
Res 2007;31(4):537–545.
103. Edenberg HJ, et al. Hum Mol
Genet 2008;17(7):963–970.
104. Foroud T, et al. Alcohol Clin Exp
Res 2008;32(6):1023–1030.
105. Dunn G, et al. BMC Genet 2005;
6 S1:S122.
106. Jones KA, et al. Int J Psychophysiol
2004;53(2):75–90.
107. Wang JC, et al. Hum Mol Genet
2004;13(17):1903–1911.
108. Dick DM, et al. Biol Psychiatry
2008;63(11):1047–1053.
109. Manolio TA, et al. Annu Rev Med
2009;60:443–456.
110. Manolio TA, et al. Nat Genet
2007;39(9):1045–1051.
111. Treutlein J, et al. Arch Gen
Psychiatry 2009;66(7):773–784.
112. Edenberg HJ, et al. Pharmacol
Ther 2005;108(1):86–93.
113. Edenberg HJ. Biol Psychiatry
2011;70:498–499.
 Nicotine dependence
Chapter

23 Sarah M. Hartz and Laura J. Bierut

Introduction Nicotine dependence is not only a public health

Cigarette smoking is one of the leading causes of
mortality worldwide. Tobacco use is estimated to cause
12% of vascular disease, 66% of respiratory cancers
(trachea, bronchus, and lung cancers), and 38% of
chronic respiratory disease. Because of this, 8.8% of
deaths (4.9 million annually worldwide) are attributed
to tobacco use [1]. The relationship between lung
cancer deaths and US cigarette consumption over time
is particularly striking (Figure 23.1). Lung cancer
deaths were practically nonexistent in the early 1900s.
As smoking increased, lung cancer increased. In the
United States, an estimated 19.8% of adults are cur-
rently smokers [2]. Due to the high prevalence and
severe health consequences of cigarette smoking, the
United States spends $96 billion annually on smoking-
attributable health-care expenditures [3].
risk due to lung cancer and heart disease, but high
comorbidity with psychiatric illness makes it of par-
ticular importance in mental health. There are mark-
edly increased rates of lifetime smoking in subjects with
mental illness as compared to controls (59% versus
39% respectively, p < 0.001) [4]. The prevalence of
nicotine dependence is 13% in the general population
and 30% and 70% in the presence of other psychiatric
disorders (Table 23.1) [5]. Because of the concentra-
tion of nicotine dependence within the psychiatrically
ill population, individuals with a psychiatric disorder
consume 46% of all cigarettes smoked in the United
States. Although these data indicate the strong need
for smoking cessation in the psychiatrically ill popu-
lation, quit rates are substantially lower for subjects
with active mental illness as compared to smokers
without mental illness (31% versus 43% quit rate) [4].

5000 Figure 23.1 US cigarette consumption

70 and age-adjusted lung cancer death rates
4500 over time. Data is from the public-use
Per capita cigarette consumption

data files, National Vital Statistics System,

4000 60 National Center for Health Statistics,
(age-adjusted, per 100 000)

Centers for Disease Control and

Lung cancer death rate

3500 Per capita

50 Prevention, and the Tobacco Outlook
cigarette
3000 Report, Economic Research Service,
consumption
40 US Department of Agriculture.
2500
Lung cancer
2000 death rate 30
1500
20
1000
10
500
0 0
19 0
19 5
19 0
19 5
19 0
19 5
19 0
19 5
19 0
19 5
19 0
19 5
19 0
19 5
19 0
19 5
19 0
19 5
19 0
95
20 0
05
0
0
1
1
2
2
3
3
4
4
5
5
6
6
7
7
8
8
9

0
19

20

Year

Principles of Psychiatric Genetics, eds John I. Nurnberger, Jr. and Wade H. Berrettini. Published by Cambridge University
Press. © Cambridge University Press 2012.

287
 Chapter 23: Nicotine dependence

Table 23.1 DSM-IV diagnoses and comorbidity with nicotine experiment with smoking; (2) regular smoking,
dependence.
defined as having smoked at least 100 cigarettes during
Sample Prevalence their lifetime; and (3) nicotine dependence, a psychi-
size of nicotine atric disorder defined by symptoms of tolerance,
dependence (%) withdrawal, and loss of control [8]. The criteria for
Entire sample 43 093 12.8
nicotine dependence, specified by the latest version of
the Diagnostic and Statistical Manual for Psychiatric
Alcohol dependence 1484 45.4 Disorders (DSM-IV-TR), parallel the dependence cri-
Drug dependence 549 69.3 teria for other substances including alcohol, cocaine,
Major depression 3119 30.0
and opiate dependence. For research purposes, other
psychometric instruments have been developed to
Mania 724 35.3 measure nicotine dependence. The most commonly
Generalized anxiety 894 32.7 used measure is the Fagerström Test of Nicotine
disorder Dependence (FTND) [9]. Table 23.2 delineates the
Panic disorder with 254 39.8 criteria for the DSM-IV-TR definition of nicotine
agoraphobia dependence as compared to the FTND-based defin-
ition. Since the DSM-IV-TR definition of dependence
Antisocial 1422 42.7
can be applied to any substance, the criteria are less
personality disorder
specific for nicotine dependence. Can something be
Data from Grant et al. [5]. said about the peculiar vulnerability of the adolescent
to nicotine addiction?

Not only is a significant proportion of nicotine

dependence comorbid with psychiatric illness, but a
significant portion of morbidity in psychiatric illness
may be attributed to nicotine dependence. It is
striking to note that the greatest risk of mental illness
is premature death due to heart disease and cancer. In
an 8-state comparison of deaths of clients at public
mental health clinics, public mental health clients
lived 13–30 years less than their general public coun-
terpart [6]. Although suicides and accidental deaths
are elevated in this population as compared to con-
trols, the primary causes of the increase in deaths are
heart disease and cancer. Therefore, the largest modi-
fiable risk factor for heart disease and lung cancer in
the mental health population is cigarette smoking.
In addition, nicotine dependence is associated with
disease-specific poor outcomes. For example, lifetime
smoking in bipolar disorder is associated with earlier
age of onset of symptoms, greater severity of symp-
toms, poorer functioning, history of suicide attempts,
and comorbid anxiety and substance use disorders
[7]. The increased mortality and more severe course
of illness highlights the importance of understanding
the etiology of nicotine dependence in the psychiatric
population.
In order to better understand smoking behavior,
it is compartmentalized into: (1) initiation, the period
during which subjects smoke for the first time and
Genetics of nicotine dependence
The conceptualization of the different stages of
smoking is helpful in understanding the varying
environmental and genetic contributions to each step
in the development of smoking behaviors. The herit-
ability of smoking initiation was estimated to be 44%
[10], with a sex differential of 37% for males and 55%
for females [11]. Conversely, the heritability of nico-
tine dependence was markedly higher at 75% [10]. To
underscore the social influences of smoking initiation
in contrast to nicotine dependence, Saccone et al.
found that twins starting to smoke at exactly the same
time is a good index for shared social influences on
smoking initiation, but it is unrelated to nicotine
dependence as defined by DSM-IV and FTND [12].
This indicates that the significant social influences on
smoking initiation do not appear to influence the
development of nicotine dependence. Multiple twin
studies have found little common environmental
influence in risk for nicotine dependence.
Although smoking initiation and nicotine depend-
ence are related, the genetics of nicotine dependence
are of primary interest due to its public health ramifi-
cations: nicotine dependence predicts difficulty with
cessation and carries most of the morbidity associated
with smoking. Specifically, the quantity of cigarettes
smoked over a lifetime is associated with lung disease

288
 Chapter 23: Nicotine dependence

Table 23.2 Definitions of nicotine dependence using DSM-IV-TR and Fagerström Test of Nicotine Dependence (FTND).

DSM-IV-TR [8] FTND [9]

3 (or more) of the following in a 12 month period: Cigarettes per day:
 Tolerance < 10 0 points
 Withdrawal 11–20 1 point
 Using more than intended 21–30 2 points
 Difficulty controlling use > 30 3 points
 Spending a great deal of time obtaining, Time after waking for first cigarette:
using or recovering from substance < 5 min 3 points
 Giving up activities because of use 6–30 min 2 points
 Use despite harm 31–60 min 1 point
>1h 0 points
Smoked more frequently in AM than PM:
Yes 1 point
No 0 points
Smoked where forbidden:
Yes 1 point
No 0 points
Cigarette most valued:
First cigarette after waking 1 point
Any other cigarette 0 points
Smoked when too ill to get out of bed:
Yes 1 point
No 1 point
Nicotine dependence is defined as FTND score ≥ 4

and heart disease, and smoking cessation is a positive
prognostic factor for both. Nicotine dependence
accounts for 13% of the general population but con-
sumes 58% of the cigarettes smoked in the United
States [5]. Thus, understanding the genetics of nico-
tine dependence can lead to targeted treatments and
ultimately significantly decrease tobacco-associated
morbidity and mortality.
In the study of nicotine dependence, it is import-
ant to understand the behavioral progression to nico-
tine dependence when choosing a control group.
Subjects who have not had adequate nicotine expos-
ure (have not smoked enough cigarettes) have not
had opportunities to become nicotine dependent.
For this reason, genetic studies of nicotine depend-
ence carefully choose their control group as regular
smokers who did not become nicotine dependent.
This minimizes the possibility that the controls would
be cases if they had adequate nicotine exposure.
The strong heritability of nicotine dependence has
led to genetic studies. The initial efforts to identify
susceptibility loci primarily used linkage designs. Unfor-
tunately, no associations were robustly replicated. In
addition to linkage studies, isolated candidate gene
case-control studies targeted various neurotransmitters
and their receptors, but these generally had low
power and insufficient density of genotypes.
The advent of the genome-wide association stud-
ies (GWAS) brought increased genomic density
and large datasets. GWAS relied on the completion
of the International HapMap Project in 2005, where
common single nucleotide polymorphisms (SNPs)
were mapped tightly throughout the genome. Using
a case-control design and hundreds of thousands of
SNPs spanning the entire genome, GWAS look for
association between SNPs and disease. Although there
are more statistical tests of association (one test per
SNP) than there are subjects (typically thousands), the
combination of setting a sufficiently low threshold
for p-value significance and placing a high emphasis
on replication of significant results in independent
datasets has resulted in robust associations.
The first GWAS on nicotine dependence was
conducted by Bierut et al. [13]. The study compared
1050 nicotine dependent cases (FTND
4), and 879
controls (smoked
100 cigarettes in lifetime and

289
 Chapter 23: Nicotine dependence
lifetime FTND ¼ 0). Although no individual SNPs
reached genome-wide significance, the data were
reanalyzed for a targeted association study with 3713
SNPs in 348 candidate genes [14]. Several cholinergic
nicotinic receptor genes dominated the top signals
after controlling for multiple testing.
Moving from association to function
It is biologically relevant that the top genetic associ-
ations for nicotine dependence were in genes encod-
ing for nicotinic receptor subunits. Nicotine produces
its central and peripheral actions by binding to neur-
onal nicotinic acetylcholine receptors (nAChRs), a
class of neuronal ligand-gated ion channels expressed
in the nervous system. nAChRs are made of five
combinations of a and b subunits constructed around
a central pore. The subunits are encoded by nine a
(a2–a10) and three b (b2–b4) subunit genes, named
CHRNA2–CHRNA10 and CHRNB2–CHRNB4 respect-
ively. The expression of the different subunits in spe-
cific anatomical areas leads to hypotheses regarding
their functional relevance. The addiction of nicotine
is thought to arise in part from the interaction
between dopaminergic and nicotinic neurons in the
striatum (Figure 23.2). Multiple nicotinic subunits are
involved in this interaction including a4, a5, a6, b2,
and b3. This region has been implicated in the reward
pathway and is important for the development of
substance dependence.
Although GWAS are a technological break-
through in terms of the density of markers that are
tested across the genome, the genome is not com-
pletely represented by the SNP arrays used in GWAS.
For example, the Affymetrix Genome-Wide Human
SNP Array 6.0 genotypes 906 600 SNPs out of an
estimated 10 000 000 SNPs in the human genome
(approximately 2 000 000 of which are known). The
dense coverage in GWAS, although a vast improve-
ment over previous methodology, is still not dense
enough to ensure that associations are found. None-
theless, a SNP reaching genome-wide association sig-
nificance indicates that functional variation is likely
to occur through this SNP or one of the correlated,
untested SNPs.
The most biologically compelling association with
nicotine dependence was found in rs16969968, a non-
synonymous SNP in the a5 nicotinic receptor subunit
gene CHRNA5 [13, 14]. This association has been
replicated with either rs16969968 or correlated SNPs
290
in many other independent studies. Using the quan-
titative trait of cigarettes per day as a proxy for nico-
tine dependence, Berrettini et al. [15] performed a
GWAS and found an association of cigarettes per
day with SNPs in the region of CHRNA3–CHRNA5.
In order to further characterize the genetic relation-
ships in this region, these associations were mapped
with rs16969968. All the significant SNPs in the
CHRNA3–CHRNA5 region formed a single haplotype
block, with high and low risk alleles. Thus, in inde-
pendent samples, association was seen between nico-
tine dependence and rs16969968 or related SNPs.
CHRNA5 and CHRNA3 are nicotinic receptor
subunit genes on chromosome 15q25 coding for the
a5 nicotinic receptor subunit and the a3 nicotinic
receptor subunit respectively (Figure 23.3a). The coding
sequences are adjacent to one another and SNPs in the
two genes are in high linkage disequilibrium (LD)
(Figure 23.3b). rs16969968 is seen in Figure 23.3b in
the coding region of CHRNA5. Although rs16969968
is in the coding sequence of CHRNA5, and a biologic-
ally plausible relationship exists between CHRNA5
and nicotine dependence, SNPs in high LD with
rs16969968 span a large area encompassing the genes
IREB2, PSMA4, CHRNA5, CHRNA3, and CHRNB4.
Of the SNPs in the region of CHRNA5 and
CHRNA3, rs16969968 is the most studied. Although
other SNPs associated with rs16969968 in the region
have been directly associated with nicotine depend-
ence, rs16969968 is a nonsynonymous SNP associated
with nicotine dependence. Specifically, the minor
variant of rs16969968 results in an amino acid change
of aspartic acid to asparagine. The frequency of the
minor allele varies between populations, as seen in
Figure 23.4. It ranges from 0% in African populations
to 37% in European populations.
The biological importance of rs16969968 has
been shown in several settings. First, the a5 nicotinic
receptor is expressed in the brain. Its expression in the
striatum and direct interaction with the dopaminergic
pathway is particularly relevant to addiction. Second,
the protein sequences of CHRNA5 homologues was
examined in multiple species (human, chimpanzee,
Bolivian squirrel monkey, domestic cow, mouse,
chicken, and African clawed frog) and the aspartic
acid residue was present in all species [16]. This
conservation across species suggests that it has func-
tional importance. Third, an in vitro functional study
found that a4a5b2 receptors containing the aspara-
gine amino acid substitution in a5 exhibited an

(a)
Caudatum-putamen
and
nucleus accumbens
Substantia nigra
and VTA
(b)
α6β2β3
α4β2
α4α5β2 DA cell
α6α4β2β3
GABA cell
altered response to a nicotine agonist than did the
receptors with the aspartic acid variant in a5 [16]. This
suggests that decreased nicotinic receptor function is
associated with increased risk for nicotine dependence.
In-depth analyses of the region led to the discov-
ery of additional SNPs that modify expression of
CHRNA5 [17, 18]. The SNP rs588765 was found to
be associated with nicotine dependence and to modify
expression of CHRNA5. In addition, together with
rs16969968, it forms a haplotype to alter the risk of
Chapter 23: Nicotine dependence
Figure 23.2 Nicotinic acetylcholine
receptors in the striatum of the rodent
central nervous system. (a) Scheme of the
mesostriatal dopaminergic pathway. (b)
Subunit composition of the functional
nicotinic acetylcholine receptors
(nAChRs) expressed by dopaminergic
nerve terminals (a6b2b3, a6a4b2b3,
a4b2, a4a5b2). (From [29], with
permission from Elsevier.)
Glu cell
ACh cell
nicotine dependence. Although rs588765 can lead to
both high expression and low expression of CHRNA5
and rs16969968 can lead to high risk and low risk for
nicotine dependence, only three haplotypes exist in
the population: high risk (rs16969968) and low expres-
sion (rs588765), low risk (rs16969968) and low expres-
sion (rs588765), and low risk (rs16969968) and high
expression (rs588765) (Table 23.3). The lowest risk of
nicotine dependence occurs with the low expression
allele of rs588765 and the low risk allele of rs16969968,
291
 Chapter 23: Nicotine dependence

(a) chr15 (q25.1)

(b) rs16969968

76589924 76711042
1.0 80

Recombination rate (cM/Mb)

0.8
60
R-Squared

0.6
40
0.4

20
0.2

0.0 0
CRABP1 IREB2 LCC123088 ADAMTS7
PSMA4
CHRNA5
CHRNA3
CHRNB4

76420 76545 76670 76795 76920

Chromosome 15 position (hg18) (kb)
Figure 23.3 Graphical representation of chromosome 15q25. (a) Chromosome 15 with the region q25.1 demarked by the red line, the
location of CHRNA5 and CHRNA3. This figure was created with the University of California, Santa Cruz (UCSC) genome browser (http://genome.
ucsc.edu). (b) The 100 kb region surrounding rs16969968 (large diamond). Diamonds indicate single nucleotide polymorphisms (SNPs), and the
size of the diamond is directly proportional to the r2 of the SNP with rs16969968. The dashed lines delineate the boundaries for SNPs that have
r2
0.8 with rs16969968. This figure was created using SNAP (http://www.broad.mit.edu/mpg/snap/). See plate section for color version.

and the highest risk for nicotine dependence occurs

with the low expression allele of rs588765 and the
high risk allele or rs16969968. A similar association
was seen in a population of African-Americans with
the SNP rs555018, a SNP in high LD with rs588765 in
both European-Americans and African-Americans,
suggesting that rs555018 is associated with variable
expression of CHRNA5 [17].
The GWAS results and corresponding biological
experiments suggest that CHRNA5 is involved in the
development of nicotine dependence, but the relation-
ship between changes in structure and expression of
CHRNA5 and the development of smoking behavior
is unknown. Various research groups are evaluating
the relationship between these genes and smoking
initiation, the pleasure associated with smoking, and
age of onset of smoking.
Genetics of lung disease
The CHRNA5/CHRNA3/CHRNB4 region has also been
associated with lung cancer and chronic obstructive
pulmonary disease (COPD). Unfortunately, the fact
that both lung cancer and COPD are nearly completely
attributable to smoking makes it difficult to determine
whether the observed association is independent of
nicotine dependence. The difficulty with interpretation
has been emphasized in multiple studies.
Thorgiersson et al. [19] highlighted the import-
ance of gene–environment interactions with a GWAS
that found associations between a SNP in LD with
rs16969968, rs1051730, lung cancer, and peripheral
artery disease, in addition to confirming the prior
association seen with nicotine dependence. Their
hypothesis was that rs1051730 leads to nicotine

292
 Chapter 23: Nicotine dependence

Table 23.3 Risk of nicotine dependence based on haplotypes constructed from rs16969968 and rs588765. Alleles of rs16969968 are
represented as high risk allele for nicotine dependence and low risk allele for nicotine dependence. Alleles for rs588765 are represented
as high expression of CHRNA5 and low expression of CHRNA5. The haplotype of high risk allele (rs16969968) and high expression allele
(rs588765) is not listed because it is extremely rare in the population.

Haplotype risk of nicotine dependence

High Medium Low

rs16969968 High risk allele Low risk allele Low risk allele
rs588765 Low expression allele High expression allele Low expression allele

8
4
7
6
3 5

1
2 9

Figure 23.4 Allele frequency distribution (%) of cSNP rs16969968 in different ethnic populations. The frequency of the minor allele is
the white segment of the circles in the figure. Populations were grouped together on the basis of their genetic structures. Geographic
regions: 1. America; 2. Africa; 3. North Africa; 4. Europe; 5. Middle East; 6. Central/South Asia; 7. Centra/South Asia; 8. East Asia; 9. Oceania.
(From [16] with permission.)

dependence and subsequent smoking behavior. An
association with lung cancer may be explained by
the relationship with smoking behavior, but there
may also be a component of lung cancer that is driven
independently of nicotine dependence.
Other studies, however, reported evidence that
rs1051730 and related SNPs were associated with lung
cancer independent of smoking behavior. Hung et al.
[20] found an association between lung cancer and
rs1051730 with equal risk seen for former smokers as
compared to current smokers. There was an increased
risk for lung cancer associated with rs1051730 in
people who had never smoked, but the risk was not
statistically significant. This evidence led to the con-
clusion that the association was unrelated to nicotine
dependence.
Another study also found association with lung
cancer independent of cigarette quantity. Amos et al.
[21] found an association between lung cancer and
rs1051730 in three independent samples, robust to
adjustment for pack-years of cigarette exposure. The
association between rs1051730 and lung cancer was
absent in nonsmokers. Ideally, if rs1051730 were
related to lung cancer independent of smoking

293
 Chapter 23: Nicotine dependence
exposure the effect would be present in nonsmokers.
However, the repeated evidence of association of this
variant with lung cancer independent of smoking
exposure argues that there is an association between
rs1051730 (and related SNPs) and lung cancer beyond
the smoking exposure from nicotine dependence.
The relationship between rs1051730, nicotine
dependence, and lung cancer was more specifically
explored by Spitz et al. [22]. This was accomplished
by using three samples: (1) lung cancer cases and
controls with a history of smoking; (2) lung cancer
cases and controls who were lifetime nonsmokers;
and (3) bladder cancer cases and controls with a
history of smoking. The investigators found a statis-
tically significant association of rs1051730 with lung
cancer in subjects with a history of smoking. When
stratified for cigarettes per day, the highest risk is seen
in the lightest smokers, subjects smoking less than
20 cigarettes per day. There was no association
between lung cancer and rs1051730 in never smokers
and no association between rs1051730 and bladder
and renal cancers. The authors concluded that
because the highest risk group was in the subjects that
smoke the least, rs1051730 mediated risk to lung
cancer beyond the cigarette exposure from nicotine
dependence.
A recent GWAS for COPD also found association of
rs1051730 with COPD. This association remained sig-
nificant after adjusting for pack-years (p ¼ 5.7  10–10).
Although suggestive that this is a true association with
COPD, the authors reported that the lack of further
smoking variables lead to difficulty teasing out the full
contribution of nicotine exposure on COPD.
One explanation for the association of rs16969968
with lung disease independent of cigarette quantity is
that the quantity of toxin inhaled when smoking is
not well explained by the cigarettes per day. To inves-
tigate this, Le Marchand et al. [23] conducted a study
to evaluate the amount of nicotine and carcinogens
consumed in a sample of 819 smokers. A positive
association was found between the amount of nico-
tine metabolites seen in urine, a measure of the
amount of nicotine consumed, and the nicotine
addiction risk allele (A) of rs16969968 after adjusting
for cigarettes per day. This indicates more nicotine
and associated carcinogens are absorbed by subjects
with the risk allele than subjects without the risk
allele. Moreover, the differential absorption is not
accounted for by cigarettes per day. Because these
data argue that cigarettes per day is not a complete
294
proxy for nicotine and toxin absorption, the associ-
ation between lung cancer and rs16969968 may still
be confounded by the increased nicotine and toxin
intake in nicotine dependence even after adjusting
for cigarettes per day.
Genetic association of nicotine
dependence with other mental illness
The tendency of subjects with mental illness to smoke
has been extensively discussed in the literature.
Hypotheses regarding causation have been presented
in both directions. Some researchers argue that
smoking is a means of self-medicating and nicotine
dependence is therefore “caused” by mental illness.
These studies present evidence that smokers use nico-
tine for depressive symptoms. More recently, data has
been published that smoking precedes the diagnosis
of mental illness and may contribute to the disorder.
One study examined the association between cigarette
smoking and a subsequent first depression. It found
a four-fold higher risk of first-episode depression for
heavy smokers as compared to never smokers [24].
The current interpretation of these data is that
smoking confers additional risk in the development
of psychiatric disorders.
A third explanation for the comorbidity of nico-
tine dependence and other psychiatric illness is a
shared etiology. The heritability of major psychiatric
illness is high with estimates of 64% in schizophrenia
[25], 59% in bipolar disorder [25], and 42% in major
depressive disorder [26]. The combination of the
heritability of nicotine dependence, the heritability
of psychiatric illnesses, and comorbidity of nicotine
dependence with psychiatric illness indicate that spe-
cific genes may be pleiotropic, i.e. contribute to both
nicotine dependence and other psychiatric illness.
Due to the comorbidity of alcohol dependence
and nicotine dependence, the association between
CHRNA5 and alcohol dependence was evaluated.
Although alcohol dependence was not associated with
rs16969968, it was associated with SNPs related to
rs588765, leading to a decreased expression of CHRNA5
[27]. We expect shared mechanisms between nicotine
dependence and alcohol dependence due to the fact
that they both involve addiction, the molecular prop-
erties of nicotine and alcohol are unrelated. This
suggests that expression of CHRNA5 may function
in addiction at a general level, rather than contrib-
uting exclusively to nicotine dependence.
 Chapter 23: Nicotine dependence

Due to the phenotypic complexity of psychiatric
illnesses, the search for pleiotropic genes may help
clarify psychiatric diagnoses and help understand
biological etiologies of mental illness. Not only is
there a relationship between nicotine dependence
and other psychiatric illness, but there are comorbid-
ities within and between substance abuse, personality
disorders and major psychiatric disorders. Suspicion
of shared genetic susceptibilities with bipolar disorder
and schizophrenia has been the subject of recent
research. A study by Ferreira et al. analyzed both
a bipolar GWA dataset and a schizophrenia GWA
dataset for overlap of top signals, and found evidence
of shared association [28]. These studies suggest
that advances may be made by understanding genetic
pleiotropy and interconnections between psychiatric
illnesses.
Conclusions
Nicotine dependence is a heritable disorder with
major public health ramifications. Nicotine depend-
ence accounts for 58% of all cigarettes smoked in the
United States, and has led to an epidemic of comor-
bidities including lung cancer and heart disease. The
population with psychiatric illness is particularly sus-
ceptible to nicotine dependence. Although psychiatric
illness is present in 28% of the US population,
the mentally ill consume 44% of cigarettes smoked.
In addition, public mental health clients had a
decreased life span of 13–30 years, where the majority
of deaths are due to heart disease and cancer. Cigar-
ette smoking is therefore the largest modifiable risk
factor for morbidity and mortality both in the general
population and in the psychiatric population.
GWAS have found associations between nicotine
dependence and the a5 nicotinic receptor subunit
gene. We postulate that there are at least two distinct
biological mechanisms that alter the risk of nicotine
dependence. The first biological mechanism is caused
by an amino acid change in CHRNA5, in the non-
synonymous SNP rs16969968. Functional studies
have found that expression of the minor variant leads
to reduced response to a nicotinic agonist. The second
mechanism altering risk of nicotine dependence
is through altered expression of the a5 mRNA.
A second group of SNPs in the region, including
rs588765 and rs555018, is associated with variable
expression leading to decreased expression of
CHRNA5. The combination of altered protein and
variable mRNA expression leads to different levels
of addiction risk.
Associations in this region have also been found in
lung disease. Specifically, SNPs correlated to rs16969968
are associated with an increased risk of COPD and lung
cancer, after adjusting for cigarette use. Although the
interpretation of these studies is difficult due to the
confounding between lung disease and cigarette expos-
ure, there is evidence of a genetic interaction between
smoking and rs16969968 that may confer additional
risk for lung disease beyond cigarette exposure.
The comorbidity of nicotine dependence with
other heritable psychiatric disorders leads us to investi-
gate genetic pleiotropy. Association was found between
alcohol dependence and expression of CHRNA5.
Advances in understanding of the genetics in nicotine
dependence may therefore not only impact the diag-
nosis and treatment of nicotine dependence, but also
may help us understand shared genetic susceptibilities
in other mental illnesses.
Acknowledgements
The authors are supported by NIH grants UL1RR024992,
KL2RR024994, K02DA021237, P01CA089392, and
T32MH014677.

References 5. Grant BF, et al. Arch Gen Psychiatry

1. World Health Organization. The
World Health Report 2002. Geneva:
WHO; 2002.
2. Centers for Disease Control and
Prevention. MMWR Weekly
2008;57(45):1221–1226.
3. Centers for Disease Control and
Prevention. MMWR Weekly
2008;57(45):1226–1228.
4. Lasser K, et al. JAMA 2000;284
(20):2606–2610.
2004;61(11):1107–1115.
6. Colton CW, et al. Prev Chronic Dis
2006;3(2):A42.
7. Ostacher MJ, et al. J Clin Psychiatry
2006;67(12):1907–1911.
8. American Psychiatric Association.
Diagnostic and Statistical
Manual of Mental Disorders,
Fourth Edition, Text Revision
(DSM-IV-TR). Washington, DC:
American Psychiatric Association;
2000.
9. Heatherton TF, et al. Br J Addict
1991;86(9):1119–1127.
10. Vink JM, et al. Behav Genet
2005;35(4):397–406.
11. Kafkafi N, et al. Behav Neurosci
2003;117(3):464–477.
12. Saccone NL, et al. Am J Med
Genet B: Neuropsychiatric
Genetics 2009;150B(4):453–466.
13. Bierut LJ, et al. Hum Mol Genet
2007;16(1):24–35.
14. Saccone SF, et al. Hum Mol Genet
2007;16(1):36–49.

295
 Chapter 23: Nicotine dependence
15. Berrettini W, et al. Mol Psychiatry
2008;13(4):368–373.
16. Bierut LJ, et al. Am J Psychiatry
2008;165(9):1163–1171.
17. Saccone N, et al. Cancer Res
2009;69(17):6848–6856.
18. Wang JC, et al. Hum Mol Genet
2009;18(16):3215–3135.
19. Thorgeirsson TE, et al. Nature
2008;452(7187):638–642.
296
20. Hung RJ, et al. Nature 2008;
452(7187):633–637.
21. Amos CI, et al. Nat Genet 2008;
40(5):616–622.
22. Spitz MR, et al. J Natl Cancer Inst
2008;100(21):1552–1556.
23. Le Marchand L, et al. Cancer Res
2008;68(22):9137–9140.
24. Klungsoyr O, et al. Am J
Epidemiol 2006;163(5):421–432.
25. Lichtenstein P, et al. Lancet
2009;373(9659):234–239.
26. Edvardsen J, et al. J Affect Disord
2009;117(1–2):30–41.
27. Wang JC, et al. Mol Psychiatry
2009;14(5):501–510.
28. Ferreira MA, et al. Nat Genet
2008;40(9):1056–1058.
29. Gotti C, et al. Prog Neurobiol
2004;74(6):363–396.
 Human molecular genetics
Chapter
24 of opioid addiction
Mary Jeanne Kreek, Dmitri Proudnikov, David A. Nielsen,
and Vadim Yuferov
Introduction
Addiction to opiates is a chronic, relapsing brain
disease that, if left untreated, can cause major med-
ical, social and economic problems. There are at least
three different categories of factors that contribute to
the vulnerability of developing a specific addiction,
once self-exposed: (1) environmental factors, includ-
ing cues, conditioning, external stressors, and the
stress they cause; (2) drug-induced factors, that lead
to a variety of molecular neurobiological changes
resulting in altered behaviors; (3) genetic factors,
which represent approximately 40–60% of the risk
of developing an addiction to opioids [1]. In addition,
addiction to opioids may sometimes arise from
opioid treatment of chronic pain [2].
In this review we present several experimental
approaches performed in the Laboratory of the Biol-
ogy of Addictive Diseases to characterize the relation-
ship of gene variations with heroin addiction and
pharmacogenomics.
The mu-opioid receptor
The mu-opioid receptor (MOPr) is the primary site of
action of several of the endogenous opioid peptides,
including beta-endorphin. This receptor is also the
major target for clinically important opioid analgesic
agents, including morphine, methadone, fentanyl,
and related drugs, as well as a major site of action
for heroin. MOPr belongs to the seven transmem-
brane receptor family and it is coupled to an inhibi-
tory G protein (Gi) that inhibits adenylate cyclase
activity, resulting in a reduction of intracellular cyclic
adenosine monophosphate (cAMP) production, which
acts as a second messenger within cells. Clinical
observations have shown that individuals have varied
sensitivity to opioids, suggesting potential variability
Principles of Psychiatric Genetics, eds John I. Nurnberger, Jr. and Wade H. Berrettini. Published by Cambridge University
Press. © Cambridge University Press 2012.
in the receptor protein and/or in the gene expression
due to sequence differences. After successful cloning of
the MOPr gene (OPRM1) in 1995 [3], our laboratory,
in collaboration with the laboratory of Dr. Lei Yu,
started studies on functionality of the gene variants in
the coding region. Among several single nucleotide
polymorphisms (SNPs), we found two SNPs of high
allelic frequency, each of which alters an amino acid in
the N-terminus of the receptor. The 17T (rs1799972)
variant results in an amino acid change of alanine to
valine at position 6 (A6V) (overall allelic frequency
6.6%), and the A118G, variant, with allelic frequency
range around 2% in African-Americans to over 40% in
Japanese, which results in amino acid change from
asparagine to asparatic acid at position 40 (N40D).
One of these, the 118G (rs1799971), removes one of
five potential N-glycosylation sites, which may affect
the trafficking of MOPr into the plasma membrane,
and an interaction of ligands with the receptor.
In the first report of in vitro functional studies
of MOPr encoded by either 118A or 118G in the
hamster fibroblastoid tumor cells AV-12, we found
that the MOPr 118G variant or the more common
prototypic 118A receptor variant bound most of
the exogenous and endogenous opioid ligands with
similar affinity [4]. However, when the longest of the
endogenous opioids, beta-endorphin, was studied,
we found striking differences. Beta-endorphin bound
with a three-fold higher affinity to the 118G variant as
compared to the 118A variant. Also, in Xenopus
oocytes co-expressing the 118G OPRM1 variant with
the G protein-activated inwardly rectifying potassium
channel, a three-fold greater channel activation was
found when activated by beta-endorphin compared to
activation of the prototype 118A variant.
In the second study, we assessed the effects of the
A118G SNP of the OPRM1 and cell type on cell-surface
297
 Chapter 24: Human molecular genetics of opioid addiction
expression, agonist binding affinity, and receptor sig-
naling through inhibition of adenylate cyclase [5]. In
cell lines AV-12 and HEK293, which stably expressed
either 118A or 118G variants, the [3H]-DAMGO
whole-cell saturation binding assay showed lower
cell-surface expression of the MOPr 118G variant.
Affinity (KD) of the 118G variant was found to be
significantly greater in HEK293 cells whereas there
was no difference in the binding affinity between the
variant and the prototype MOPr in AV-12 cells. To
characterize the agonist-dependent receptor signaling
properties, a decrease of forskolin-induced cAMP
levels were assayed in AV-12 and HEK293 cell lines
expressing the 118A or 118G variants of MOPr
following stimulation with opioid receptor agonists
morphine, methadone, DAMGO, and beta-endor-
phin. All agonists except beta-endorphin displayed
greater potency (EC50) in both cell lines expressing
the 118G variant. In contrast to stable cell lines, the
prototype MOPr or the 118G variant transiently
expressed in either AV-12 or HEK293 showed no
difference in EC50 of DAMGO or beta-endorphin by
the cAMP assay in both cell lines.
A lower in vitro expression of the OPRM1 118G
variant was also reported by other laboratories in
stably expressed HEK293 cells [6] and transiently
expressed CHO cells [7]. Moreover, Sadee and col-
leagues tested the A118G SNP for allele-specific gene
expression in human postmortem brain tissues from
cortical lobe and pons, and showed two-fold lower
OPRM1 mRNA levels of the G allele than the A allele
[7]. Also, they found that the overall OPRM1 mRNA
expression in pons in subjects homozygous for 118G
was lower compared to the 118A variant. It is very
likely that the 118G variant has altered MOPr expres-
sion at the level of mRNA transcription as well as at
cell surface presentation. In vitro functional analyses
of A118G have shown different results in some stud-
ies [6, 8], probably due to differences in experimental
condition. However, a dramatically different func-
tional effect of the A118G SNP has been demonstrated
by many studies in our laboratory and others. These
studies have demonstrated an association of this
receptor variant with both opioid and alcohol addic-
tions, and, further, documented functional differ-
ences, including pharmacodynamic responses to
mu-opioid receptor antagonist, naltrexone, medica-
tion (pharmacogenetics), and physiological responses,
including stress responsivity (physiogenetics) [reviewed
in 1, 9, 10].
298
Epigenetics of the MOPr
Epigenetics is the study of inherited changes in
gene expression that are not encoded in the sequence
of the DNA. Epigenetic regulation provides a layer of
control that can regulate gene expression. The two
main processes that control epigenetics are DNA
methylation and histone modifications. Alterations
in these processes may disregulate gene expression
with significant clinical outcomes. These alterations
may be passed to daughter cells and also to successive
generations.
Cytosine methylation by DNA methyltransferases
at cytosine:guanine dinucleotide (CpG) sites in geno-
mic DNA represent a key epigenetic pathway modify-
ing gene expression. The methylation of specific CpG
sites in genomic DNA may alter transcription factors
binding to DNA regulatory sequences (e.g. [11, 12]).
DNA methylation of CpG dinucleotide sites in the
promoter regions of genes is usually associated with
downregulation of gene expression [13].
Our laboratory has recently found and reported
that two CpG sites in the promoter region of the
MOPr gene OPRM1, the site of action of opiates and
opioids, have increased levels of methylation in
former severe heroin addicts well stabilized in metha-
done maintenance treatment [14]. Of the 16 CpG sites
examined in the OPRM1 promoter region, the –18
and the þ84 CpG sites had hypermethylation in
former addicts. Both of these sites are located in
potential Sp1 transcription factor-binding sites. It is
possible that methylation of these CpG sites could
lead to reduced OPRM1 expression in former heroin
addicts. The DNA hypermethylation of these specific
CpG sites in the former heroin addicts could be due
to: (1) inheritance through genomic imprinting; (2)
changes occurring earlier in life which predispose to vul-
nerability to develop heroin addiction; (3) chronic heroin
use; or (4) long-term methadone pharmacotherapy.
Other studies have shown that other drugs
of abuse can alter DNA methylation or histone modi-
fications. Alcoholism has been shown to increase
overall DNA methylation [15, 16], decrease DNA
methyltransferase expression [17], and increase spe-
cific DNA methylation at specific genes [15, 18]. In
rodents, cocaine administered on an acute or chronic
basis has been shown to cause chromatin remodeling
through histone modification at the cFos, or at the
bdnf and cdk5 promoters, respectively [19], which
may be due to a decrease in the histone deacetylase
 Chapter 24: Human molecular genetics of opioid addiction
HDAC5 function [20]. Cocaine has also been shown
to decrease histone methylation in the prefrontal
cortex [21], while maternal cocaine administration
increases cytosine methylation in the promoter of
the protein kinase Cepsilon (PKCE) gene in fetal
heart [22].
Several pharmacological agents such as azacitidine
and valproic acid (discussed in [14]) which alter DNA
modification or histone modification have been used
in clinical therapies for cancer treatments. Perhaps
these may be used to modify epigenetic states in
addiction.
The kappa-opioid receptor
Although opioid receptors often subserve similar
physiological functions, activation of the kappa-opioid
receptor (KOPr) by exogenous agonists produces dys-
phoria in humans and aversive effects in experimental
animals, in contrast to activation of MOPr [23]. KOPr
agonists, including dynorphin, the primary endogen-
ous peptide ligand for this receptor, decrease basal
and also drug-induced increases of dopamine levels
in several areas of the dopaminergic nigrostriatal and
mesolimbic–mesocortical systems [24]. The KOPr-
dynorphin system may therefore be considered to be
a part of the counter-modulatory mechanisms of the
brain following direct or indirect drug-induced dopa-
minergic stimulation [1]. The ability of dynorphin
peptides to modulate dopaminergic tone in humans
has been demonstrated in studies conducted in normal
volunteers and also in methadone-maintained former
heroin addicts. These studies have measured prolactin
release, which is under tonic inhibition by dopamine in
the tuberoinfundibular region of the hypothalamus.
Administration of dynorphin A(1–13), a shortened form
of the natural peptide dynorphin A(1–17), in humans
leads to an abrupt rise in serum prolactin levels caused
by a lower dopaminergic tone through activation of
KOPr [25]. Prolactin response to dynorphin A(1–13)
was significantly attenuated in methadone (mu-opioid
agonist)-maintained volunteers compared to control
subjects, suggesting either alterations in the KOPr
system or, more likely, a lowering of tuberoinfundibular
dopaminergic tone in former heroin addicts [26]. Also,
clinical studies have shown that dynorphin A(1–13)
attenuates opioid-withdrawal symptoms in humans [27].
The human OPRK1 gene is located on chromo-
some 8q11.2. The initial cloning of OPRK1 indenti-
fied three exons that contained the complete coding
sequence [28]. Studies of the gene structure in our
laboratory defined an additional 50 exon of the
untranslated region and several transcription initi-
ation sites [29]. The redefinition of the exon/intron
organization of the human OPRK1 gene has facili-
tated the localization of the proximal promoter of
the gene, and potential regulatory polymorphisms.
Two studies have evaluated variants of the OPRK1
gene for possible association with opiate addiction.
Yuferov and colleagues [29] analyzed 12 single
nucleotide polymorphisms (SNPs), including 5 novel
variants in 145 former heroin addicts and 146
control subjects. A logistic regression analysis showed
a point-wise significant association of the synonym-
ous SNP G36T (rs1051660) with opiate addiction
(p ¼ 0.016). The number of haplotypes of eight high
frequency SNPs seen in the three ethnic groups were
nine, six, and five for African-Americans, Caucasians,
and Hispanics, respectively, with corresponding sig-
nificance levels for differences in haplotype frequen-
cies between cases and controls of p ¼ 0.0742, 0.1015,
and 0.0041. In Hispanics, the AGCTCGTC haplotype
is found only in control subjects (frequency of 0.06),
while the GGCGTGCC (frequency of 0.07) is found
only in subjects with opiate addiction. An association
of the SNP G36T with opiate addiction was also
shown in a Caucasian population of heroin addicts
[30]. The function of the SNP G36T is not known,
and it may be in linkage disequilibrium (LD) with
other cis-regulatory polymorphisms in OPRK1 gene.
Recently, a novel high frequency insertion of 830 bp
in a putative promoter region of OPRK1 was reported
[31]. In a reporter gene expression assay, the presence
of the insert reduced transcription activity of the
OPRK1 promoter to 53% of that in its absence.
Melanocortin receptor type 2
The melanocortin receptor type 2 (MC2R or adreno-
corticotropic hormone, ACTH receptor) belongs to
the superfamily of G-protein-coupled receptors and is
involved in regulation of adrenal cortisol secretion,
important in the physiological response to stressors.
Physical and psychological conditions have been found
to be associated with dysregulation of the hypothal-
amic-pituitary-adrenal (HPA) axis: post-traumatic
stress disorder, fibromyalgia, Alzheimer’s disease,
major depression, and specific stressors [32]. We found
that specific addictive diseases also associated with
dysregulation of the HPA axis: hyperresponsivity to
299
 Chapter 24: Human molecular genetics of opioid addiction
removal of glucocorticoid negative feedback was
found in cocaine addicts [33]; HPA hypoactivity was
found in medication-free illicit drug-free former
heroin addicts [34].
Adrenocorticotropic hormone (ACTH) is the major
hormone derived from proopiomelanocortin (POMC),
and regulates adrenal glucocorticoid and androgen syn-
thesis in the zonae fasciculata and reticularis in the
adrenal cortex. ACTH binds to its specific melanocortin
2 receptor (MC2R or ACTH receptor) [35]. MC2R,
which is involved in steroidogenesis, is expressed mostly
in adrenal cortex and was also found in human skin and
ovarian steroid cell tumor [32].
In most human genetics studies, variants in the
MC2R gene have been linked to familial glucocorti-
coid deficiency (e.g. [36, 37]). Recent studies showed
possible involvement of MC2R in stress mechanisms.
Substitution of an A to G in –179A>G (–2T>C) of
the MC2R gene results in lower promoter activity of
this gene in vitro and was associated with impaired
cortisol response to ACTH stimulation in vivo [38].
A clinical study with a six-hour ACTH stimulation
test showed that subjects with the –179AA genotype have
a higher dehydroepiandrosterone (DHEA) response
than –179GG subjects, while baseline DHEA concen-
trations did not differ between groups [39].
In our studies [32], we found an experiment-wise
significant association of the minor A allele of the
polymorphism –184G>A of the MC2R gene and the
haplotype AACT bearing –184A variant with a pro-
tective effect from the development of heroin addic-
tion in Hispanics. We also found a strong effect in
association of individual genotype frequencies of this
polymorphism or combination of GA þ AA geno-
types with a protective effect from heroin addiction,
providing genetic evidence supporting our hypothesis
that dysregulation of the HPA axis contributes to the
development of drug addiction.
Catechol-O-methyltransferase
Catechol-O-methyltransferase (COMT) takes an
important part in metabolism of catechol neurotrans-
mitters including dopamine. Prolonged administration
of drugs of abuse can lead to alterations in the dopami-
nergic system. In animal models, chronic administra-
tion of cocaine results in reduction of striatal dopamine
and dopamine D2 receptors levels [40]. In subjects
addicted to drugs of abuse, brain imaging shows reduc-
tions in striatal dopamine D2 receptors [41].
300
COMT has been found in peripheral and central
tissues [40]. A common SNP 472G>A alters amino
acid Val to Met at position 158 which results in
a four-fold decrease of activity of COMT [42]. In
heterozygous human lymphoblast cell lines and
brains, allele 158Met was overexpressed compared to
158Val in all samples studied [43]. In a functional
magnetic resonance imaging (fMRI) study, amphet-
amine administration enhanced the prefrontal func-
tioning in individuals homozygous for the Val allele
during a working memory task, while those homozy-
gous for the Met allele showed no enhancement of
cortical efficiency at low to moderate working memory
load [44].
Variants of the COMT have been used in genetic
studies of association with various neuropsychiatric
conditions including schizophrenia, panic disorder, sui-
cide, major depression, bipolar disorders, obsessive–
compulsive disorder, attention deficit hyperactivity
disorder (ADHD), and efficacy of response to treat-
ment of Parkinson’s disease [40].
The high-activity 158Val allele variant has been
found to be associated with abuse and addiction to
several drugs in Caucasians [45], abuse of metham-
phetamine in Han Chinese [46], and heroin addiction
in Caucasians [47]. In contrast, the 158Met variant of
COMT has been found to be associated with alcohol-
ism in Caucasians [48]. A number of different specific
haplotypes of COMT have been found to be associ-
ated with nicotine dependence in African-Americans
and Caucasians [1] and with cocaine dependence
in African-Americans [49]. A study of dopamine trans-
porter (DAT)–COMT gene–gene interaction [50]
showed that subjects homozygous for 158Met/Met
with lower COMT activity and greater dopamine tone,
presumably, showed greater activation in both pre-
frontal cortex and ventral putamen measured in an
fMRI guessing task sensitive to reward.
The result of an association study might be
gender-specific: in one study the 158Met allele was
associated with obsessive–compulsive disorder in
males, but not in females [40]; in another study,
the allele 158Val was associated with alcoholism
in American-Indian females, but not males [51].
COMT homozygous knockout female mice develop
increased anxiety in a light/dark model compared to
COMT knockout males; also in male mice only an
increased aggressive behavior in COMT heterozy-
gous knockouts compared to other genotypes was
found [40]. In in vitro cellular studies, physiological
 Chapter 24: Human molecular genetics of opioid addiction
concentrations of 17-beta-estradiol were shown to
downregulate COMT gene transcription and protein
expression [40].
In recent studies performed by our group [40], we
tested four SNPs located in the coding region of the
COMT gene for association with heroin addiction;
an association of the Val158Met SNP in Hispanic
subjects with opiate addiction was found. In this
study, stratification of the data by gender has shown
an association of 158 Val/Met and 158Met/Met
genotypes with opiate addiction in women, but not
men [40].
5-hydroxytryptamine (serotonin)-1B
receptor
The 5-hydroxytryptamine (serotonin)-1B receptor
(HTR1B) is involved in many neuropsychiatric
functions including thermoregulation, locomotion
and feeding [52]. Serotonin receptor knockout mice
showed increased impulsive aggression, decreased
anxiety, increased exploratory activity, increased
special memory performance, increased locomotor
response to cocaine administration, and increased
alcohol consumption [53]. Serotonin 1B receptor
knockout mice showed increased cocaine self-
administration, and also administration of serotonin
1B receptor agonists increased serotonin 1B function
leading to reduced cocaine self-administration in
rats [53].
In human genetics studies, polymorphisms of
HTR1B gene were used in association studies with
feeding disorders, ADHD, childhood-onset mood dis-
order, schizophrenia, bipolar disorder, major depres-
sion, aggressive behavior, and suicide [38]. SNPs of the
HTR1B gene were also tested for an association with
substance abuse and alcoholism [e.g. 53–56].
Gene expression studies reported that the T-161
variant is expressed in a higher level than A-161 in
both BeWo and COLO 320 DM cell lines [55]. The
haplotype variant consisting of –261G and –161A
enhances transcriptional activity 2.3-fold compared
to the haplotype consisting of –261T and –161A.
The substitution of A with T in position –161 reverses
this effect, making transcriptional activity of G-261/
T-161 equal to the major haplotype T-261/A-161 [57].
A common polymorphism rs13212041 in the 3’
untranslated region of the HTR1B gene was found to
attenuate microRNA miR-96 function and was asso-
ciated with aggressive behavior [58].
Recently our group has tested a number of SNPs
of the HTR1B gene for association with heroin addic-
tion [53]. A protective effect from the development of
heroin addiction by the minor allele 1180G in Cauca-
sians was found. Association studies of a haplotype
containing a G at position 1180 showed stronger
association effects with protection from heroin addic-
tion, suggesting an involvement of HTR1B in the
mechanisms that underlie the development of addic-
tion to heroin.
Tryptophan hydroxylase 1 and 2
Various facets of impulsivity and mood are regulated
by serotonin. Levels of cerebral spinal fluid
5-hydroxyindolacetic acid (CSF 5-HIAA), a degrad-
ation product of serotonin, are low in subjects with
behaviors characterized by deficits in impulse control
[59, 60]. The rate-limiting enzyme in the biosynthesis
of serotonin is tryptophan hydroxylase (TPH). There
are two known isozymes of TPH, TPH1 and TPH2,
which are encoded by separate genes, TPH1 and
TPH2, respectively [61]. TPH1 is expressed primarily
in the pineal gland, the developing raphe nuclei
(during the late developmental stage), and the enter-
ochromaffin cells of the gut, while TPH2 is expressed
primarily in the raphe nuclei of the brain [62].
TPH2 is also expressed in the brain in the hippo-
campus, hypothalamus, cortex, thalamus, amygdala,
and cerebellum at levels slightly lower than TPH1
mRNA [63].
Variants in the TPH1 gene have been found to
be associated with CSF 5-HIAA concentrations [64],
linked to alcoholism [65, 66], and associated with
other addictive phenotypes, including age of onset
of alcoholism [67], drinking-related antisocial behav-
ior [68], smoking initiation [69, 70], and nicotine
dependence [71]. However, in our study, variants in
TPH2 have only been found to be associated with one
addictive disease, heroin addiction [72]. In our study,
one TPH1 variant, which had previously been found
to be associated with alcoholism and smoking, and
six common TPH2 variants were genotyped. At the
two-locus genotype pattern level in Hispanics, the
interaction of the TPH1 rs1799913 variant with
the TPH2 rs7963720 variant was found to signifi-
cantly interact and to be associated with heroin
addiction. Also, a second interaction of the TPH1
rs1799913 variant and the TPH2 rs4290270 variant
301
 Chapter 24: Human molecular genetics of opioid addiction
was significantly associated with heroin addiction.
In the African-Americans, a significant association
of a specific TPH2 haplotype with heroin addiction
was found [72]. The haplotype blocks containing the
TPH2 and TPH1 variants may contain other variants
that may differ based on ethnic background. The
expression of the two TPH genes may coordinately
regulate serotonergic functions. This combined expres-
sion may be dissimilar in the different ethnic groups,
and may affect an individual’s vulnerability to develop
heroin addiction with unique ethnic characteristics.
P-glycoprotein gene (ABCB1) variants
and methadone dose requirements
Opiate addiction is a chronic relapsing disorder that
is treated world-wide with methadone. Successful
treatment relies, in part, on individual methadone
dose optimization. The inter-individual variability in
methadone dose-effectiveness may be determined in
part by genetic factors. The ABCB1 gene is highly
polymorphic and a few common variants have been
shown to be associated with P-gp expression, drug
response, and disease susceptibility. Since methadone
is a substrate of P-glycoprotein (P-gp) 170 and few
common ABCB1 variants were associated with pro-
tein expression and drug response, we assessed the
association between ABCB1 SNPs and methadone
dose requirements [73]. P-gp is a member of the
ATP-binding cassette (ABC) superfamily. It is com-
posed of two homologous sequences, each containing
six transmembrane domains and an ATP binding
domain. P-gp is expressed in tissues with barrier
function, including the endothelial cells lining the
brain capillaries. The stabilizing methadone doses
were normally distributed with a mean and median
dose of 160 mg/day (range 30–280 mg/day). Nine ABCB1
common SNPs (rs1045642, rs6949448, rs2235067,
rs2032583, rs2032582, rs1922242, rs1128503, rs2520464,
and rs3789243) were genotyped. A significant differ-
ence in genotype frequencies was found between the
“higher” (> 150 mg/day) and “lower” ( 150 mg/day)
dose groups for the synonymous SNP 1236C>T
(rs1128503). Furthermore, individuals bearing the
three-locus genotype pattern TT-TT-TT (rs1045642,
rs2032582, and rs1128503) required higher doses,
while individuals heterozygous for these three SNPs
stabilized at lower doses.
302
Hypothesis-driven genes
association study
In our recent study [74], we have used a SNP hypoth-
esis-driven array that was designed by the group
of D. Goldman at the National Institute of Alcohol
Abuse and Alcoholism (NIAAA). Variants totalling
1350 in 130 candidate genes were scanned in a
Caucasian population of 412 former severe heroin
addicts in methadone treatment, and 184 healthy
controls. The majority of the subjects were from
the United States (New York City and Las Vegas)
and the minority from Israel. A total of 178 ancestry
informative markers (AIMs) were employed to
exclude population stratification. The nine variants
that showed the most significant associations were
in noncoding regions of the genes encoding the
mu-, kappa-, and delta-opioid receptors; the neuro-
peptide galanin; the serotonin receptor 3B; and the
casein kinase 1 epsilon. The two OPRM1 variants
(rs510769 and rs3778151, intron 1) are in strong
LD and belong to a haplotype block of 32 kb. They
are adjacent to rs1799971 (A118G, exon 1). However,
no evidence for association was found with SNP
A118G in this cohort. Galanin and its receptors have
been shown to be involved in behavioral processes,
morphine withdrawal, and effects of opiates and high
stress response, among other functions. Casein
kinase 1 epsilon regulates the circadian clock gene
PER1, which may be involved in drug dependence and
reward, and participates in important signaling path-
ways including DARPP-32 (the dopamine- and cyclic-
AMP-regulated phosphoprotein) phosphorylation
[74].
Genome-wide association studies
Genome-wide association studies (GWAS) allow
the screening of large numbers of variants in a single
study to identify variants and genes that may be
involved in determining a particular phenotype.
A recent study on vulnerability to develop heroin
addiction was conducted using the Affymetrix 10K
GeneChip, which simultaneously genotyped 10 000
SNPs [75]. In that study, DNA from former severe
heroin addicts (meeting Federal criteria for metha-
done maintenance) and control subjects, all of Cauca-
sian ethnicity, was analyzed. Separate analyses were
done for the autosomal and X chromosomal variants.
 Chapter 24: Human molecular genetics of opioid addiction
The strongest association of allele frequency with
heroin addiction was with the autosomal variant
rs965972, located in a Unigene cluster of unknown
function on chromosome 1q31.2. The variant with
the strongest association by genotype frequency with
heroin addiction was in the transcription factor myo-
cardin MYOCD gene at chromosome 17p12. The two
variants with the next strongest association by geno-
type frequency with heroin addiction were located in
intergenic regions at chromosome 17p12 and 1q31.2,
respectively. A strategy was employed in which the
most significant variants identified by association of
genotype frequency with heroin addiction were ana-
lyzed together to identify common genotype patterns
of unlinked alleles associated with heroin addiction.
When evaluated for genotype patterns using these
three variants that had the strongest significance for
association of genotype frequency with heroin addic-
tion, one genotype pattern was found to be signifi-
cantly associated with vulnerability to develop heroin
addiction. This pattern had an odds ratio (OR) of 6.25
and explained 27% of the population-attributable risk
for heroin addiction. Another genotype pattern of
these variants was found to be significantly associated
with protection from developing heroin addiction
(OR 0.13). Lacking this genotype pattern explained
83% of the population-attributable risk for developing
heroin addiction. Evaluation of 393 genes that were
hypothesized or known to be involved in heroin
addiction or affective disorders identified 5 genes that
may be involved in the development of heroin addic-
tion. These were genes coding for: the mu-opioid recep-
tor; the metabotropic receptors mGluR6 and mGluR8;
nuclear receptor NR4A2; and cryptochrome 1
(photolyase-like).
Other association studies of drug and alcohol
addiction have found evidence of specific genes which
may be involved in vulnerability to develop an addic-
tion. In an early study, the Uhl group [76] identified a
number of variants that were associated with drug
abuse vulnerability using a DNA pooling approach
and a 1494 variant chip. Additional chromosomes
regions were found using a similar pooling approach
with a larger cohort and the 10K GeneChip [77].
Further pooling studies have used the 100K and
500K GeneChips to narrow down the significant asso-
ciations to 89 genes suggestive of playing a role in
addiction vulnerability [78, 79]. Short tandem repeat
(STR) markers have also been used in linkage studies
on vulnerability to develop heroin addiction [80–82].
Two studies by the Tsuang group have found a link-
age peak with point-wise significance on chromosome
4q31 [81, 82]. Two recent linkage studies have exam-
ined opioid dependence using SNPs [83, 84]. Lachman
et al. [83] found a single chromosomal region at
chromosome 14q in their Hispanic group “suggestive”
of genome-wide evidence for linkage. In the other
study, the Gelernter group [84] identified eight
variants with point-wise significance of association
with opiate dependence. These studies have provided
evidence that a number of genes act in a complex
cooperative and synergistic manner to influence the
vulnerability and development of heroin addiction.
It will require additional studies to delineate the
role(s) of these genes and genetic variants.
Summary
Many studies now have been conducted to address a
possible association of specific variants of specific
genes with opioid addiction. The genes studied are
usually those which have been shown to be involved
in some aspects of development of addiction. In some
studies, multiple genes, all hypothesized to be poten-
tially involved in opioid addiction, have been studied.
In other cases, GWAS have been performed to iden-
tify regions of chromosomes which may influence
vulnerability to the development of addiction. These
studies usually do not identify specific genes within
chromosome regions. Here we have reviewed our
recent studies and a few selected other studies which
have used some of these approaches.
Acknowledgements
Funding support for this paper was received from the
National Institutes of Health (NIH) – National Insti-
tute on Drug Abuse (NIDA) Research Center Grant
P60-DA05130; the NIH National Institute of Mental
Health (NIMH) grant MH 79880; the NY State Office
of Substance Alcohol and Substance Abuse Services
(OASAS) grant C001839 (Kreek); and the NIH grant
UL1RR024143 and the National Center for Research
Resources (NCRR) (Coller).
The authors also wish to thank Dr. Orna Levran
for reading this chapter, and Kitt Lavoie and Susan
Russo for help in the preparation of the manuscript.
303
 Chapter 24: Human molecular genetics of opioid addiction
References
1. Kreek MJ, et al. Pharmacol Rev
2005;57:1–26.
2. Savage SR, et al. Addict Sci Clin
Pract 2008;4:4–25.
3. Mestek A, et al. J Neurosci
1995;15:2396–2406.
4. Bond C, et al. Proc Natl Acad Sci
U S A 1998;95:9608–9613.
5. Kroslak T, et al. J Neurochem
2007;103:77–87.
6. Beyer A, et al. J Neurochem
2004;89:553–560.
7. Zhang Y, et al. J Biol Chem
2005;280:32618–32624.
8. Befort K, et al. J Biol Chem
2001;276:3130–3137.
9. Kreek MJ, et al. Mol Interv
2007;7:74–78.
10. Kreek MJ. Clin Pharmacol Ther
2008;83:615–618.
11. Douet V, et al. Biochem Biophys
Res Commun 2007;354:66–71.
12. Michelotti GA, et al. FASEB J
2007;21:1979–1993.
13. Jaenisch R, et al. Nat Genet
2003;33:245–254.
14. Nielsen DA, et al. Neuro-
psychopharmacology 2009;
34:867–873.
15. Bleich S, et al. Alcohol Clin Exp
Res 2006;30:587–591.
16. Bönsch D, et al. J Neural Transm
2004;111:1611–1616.
17. Bönsch D, et al. J Neural Transm
2006;113:1299–1304.
18. Bönsch D, et al. Neuroreport
2005;16:167–170.
19. Kumar A, et al. Neuron
2005;48:303–314.
20. Renthal W, et al. Neuron
2007;56:517–529.
21. Black YD, et al. J Neurosci
2006;26:9656–9665.
22. Zhang H, et al. Mol Pharmacol
2007;71:1319–1328.
304
23. Mucha RF, et al. Psycho-
pharmacology (Berl) 1985;
86:274–280.
24. Zhang Y, et al. Psycho-
pharmacology (Berl) 2004;
172:422–429.
25. Kreek MJ, et al. J Pharmacol Exp
Ther 1999;288:260–269.
26. Bart G, et al. J Pharmacol Exp Ther
2003;306:581–587.
27. Specker S, et al. Psycho-
pharmacology (Berl) 1998;
137:326–332.
28. Simonin F, et al. Proc Natl
Acad Sci U S A 1995;92:
7006–7010.
29. Yuferov V, et al. Pharmacogenetics
2004;14:793–804.
30. Gerra G, et al. Am J Med
Genet B Neuropsychiatr Genet
2007;144B:771–775.
31. Edenberg HJ, et al. Hum Mol
Genet 2008;17:1783–1789.
32. Proudnikov D, et al. Neurosci Lett
2008;435:234–239.
33. Schluger JH, et al. Neuro-
psychopharmacology 2001;
24:568–575.
34. Kreek MJ, et al. Neuropeptides
1984;5:277–278.
35. Mountjoy KG, et al. Science
1992;257:1248–1251.
36. Tsigos C, et al. J Clin Invest
1993;92:2458–2461.
37. Elias LL, et al. J Clin Endocrinol
Metab 1999;84:2766–2770.
38. Slawik M, et al. J Clin Endocrinol
Metab 2004;89:3131–3137.
39. Reisch N, et al. Eur J Endocrinol
2005;153:711–715.
40. Oosterhuis BE, et al. Am J Med
Genet B Neuropsychiatr Genet
2008;147B:793–798.
41. Volkow ND, et al. Mol Psychiatry
2004;9:557–569.
42. Lotta T, et al. Biochemistry
1995;34:4202–4210.
43. Zhu G, et al. Psychopharmacology
(Berl) 2004;177:178–184.
44. Mattay VS, et al. Proc Natl Acad
Sci U S A 2003;100:6186–6191.
45. Vandenbergh DJ, et al. Am J Med
Genet 1997;74:439–442.
46. Li T, et al. Am J Med Genet B
Neuropsychiatr Genet 2004;
129:120–124.
47. Horowitz R, et al. Am J Med Genet
2000;96:599–603.
48. Tiihonen J, et al. Mol Psychiatry
1999;4:286–289.
49. Lohoff FW, et al. Neuro-
psychopharmacology 2008;
33:3078–3084.
50. Yacubian J, et al. Proc Natl
Acad Sci U S A 2007;104:
8125–8130.
51. Enoch MA, et al. Alcohol Clin Exp
Res 2006;30:399–406.
52. Barnes NM, et al. Neuro-
pharmacology 1999;38:
1083–1152.
53. Proudnikov D, et al. Pharmacogenet
Genomics 2006;16:25–36.
54. Huang YY, et al. Neuro-
psychopharmacology 2003;
28:163–169.
55. Sun HF, et al. Biol Psychiatry
2002;51:896–901.
56. Cigler T, et al. Am J Med Genet
B Neuropsychiatr Genet
2001;105B:489–497.
57. Duan J, et al. Mol Psychiatry
2003;8:901–910.
58. Jensen KP, et al. Mol Psychiatry
2009;14:381–389.
59. Brown GL, et al. In Sandler M.
(ed.). Psychopharmacology of
Aggression. New York: Raven
Press, 1979.
60. Linnoila M, et al. Life Sci 1983;
33:2609–2614.
61. Cooper JR, et al. J Pharmacol
Exp Ther 1961;132:265–268.
62. Walther DJ, et al. Science 2003;
299:76.

63. Zill P, et al. J Psychiatr Res 2007;
41:168–173.
64. Nielsen DA, et al. Arch Gen
Psychiatry 1994;51:34–38.
65. Nielsen DA, et al. Arch Gen
Psychiatry 1998;55:593–602.
66. Hsu YP, et al. Mol Psychiatry
1998;3:213–214.
67. Chung IW, et al. Alcohol 2005;
36:1–3.
68. Ishiguro H, et al. J Neural Transm
1999;106:1017–1025.
69. Lerman C, et al. Am J Med Genet
2001;105:518–520.
70. Sullivan PF, et al. Am J Med Genet
2001;105:479–484.
Chapter 24: Human molecular genetics of opioid addiction
71. Reuter M, et al. Am J Med
Genet B Neuropsychiatr Genet
2005;134B:20–24.
72. Nielsen DA, et al. Behav Genet
2008;38:133–150.
73. Levran O, et al. Hum Mol Genet
2008;17:219–227.
74. Levran O, et al. Genes Brain Behav
2008;7:720–729.
75. Nielsen DA, et al. Mol Psychiatry
2008;13:417–428.
76. Uhl GR, et al. Am J Hum Genet
2001;69:1290–1300.
77. Liu QR, et al. Proc Natl Acad
Sci U S A 2005;102:
11864–11869.
78. Johnson C, et al. Am J M
Genet B Neuropsychiatr Genet
2006;141B:844–853.
79. Liu QR, et al. Am J Med Genet B
Neuropsychiatr Genet 2006;
141B:918–925.
80. Gelernter J, et al. Am J Hum
Genet 2006;78:759–769.
81. Glatt SJ, et al. Am J Med Genet B
Neuropsychiatr Genet 2006;141B:
648–652.
82. Glatt SJ, et al. Drug Alcohol
Depend 2008;98:30–34.
83. Lachman HM, et al. Hum Mol
Genet 2007;16:1327–1334.
84. Yu Y, et al. Hum Genet 2008;
123:495–506.
305
 Genetics of stimulant dependence
Chapter
25 Joseph F. Cubells and Yi-Lang Tang
Introduction
Central nervous system (CNS) stimulants, also called
psychostimulants, are pharmacological agents that
temporarily enhance many aspects of physiological
and psychological function. The term psychostimu-
lant encompasses a broad array of substances, includ-
ing those prescribed for medical conditions (e.g.
d-amphetamine and methylphenidate, used in the
treatment of attention-deficit hyperactivity disorder
and narcolepsy); those used primarily illicitly for the
“high” they induce (e.g. cocaine and both legally and
illegally manufactured amphetamine-type stimulants
[ATS]) and those found in over-the-counter decongest-
ants (e.g. psuedoephedrine), coffee and tea (caffeine and
theophylline), and tobacco (nicotine). Repeated admin-
istration of many psychostimulants, including cocaine,
ATS, nicotine, and caffeine, leads to dependence, which
can be broadly defined as the repeated and compulsive
use of a substance despite adverse consequences. This
chapter will focus mainly on genetic mechanisms
underlying liability to dependence on cocaine and
ATS, which are commonly used illicitly in the United
States and in many other parts of the world. The genet-
ics of nicotine dependence is the subject of a separate
chapter (Chapter 23).
Both cocaine and ATS are commonly abused
drugs. According to the World Drug Report 2008 [1],
the numbers of abusers for ATS and cocaine were
25.6 and 16.0 million respectively, accounting for
0.8% and 0.4% of the global population (age 15–64)
respectively. According to estimates published by
the US Substance Abuse and Mental Health Services
Administration (SAMHSA), there were approxi-
mately 1.6 million users of cocaine aged 12 or older,
and an additional 400 000 users of other stimulants
(largely amphetamines and derivatives) in the United
Principles of Psychiatric Genetics, eds John I. Nurnberger, Jr. and Wade H. Berrettini. Published by Cambridge University
Press. © Cambridge University Press 2012.
306
States in 2007 [2]. Research on genetic mechanisms
influencing liability to dependence on psychostimu-
lants will hopefully allow development of better treat-
ments and preventive strategies for addressing those
disturbingly high rates of psychostimulant abuse.
Evidence supporting the importance
of genetic inheritance in stimulant
dependence
Before considering molecular genetic mechanisms, it is
important first to establish that genetic inheritance con-
tributes to risk for a particular phenotype. Approaches
for establishing the role of genes in human use of
psychostimulants have included family, twin, and adop-
tion studies. The most basic requirement for establish-
ing the role of genes in any disorder is to demonstrate
that the disorder is familial, or that it tends to run in
families. Familiality, however, does not imply a genetic
mechanism, as family environment could account for
traits shared among family members. Disentangling the
role of genes from environment relies on twin studies
and adoption studies. To date, the only adoption studies
to focus on the offspring of psychostimulant-using
parents are those examining the neurodevelopment of
young children (e.g. [3]), rather than the risk of abuse or
dependence. The focus here will therefore be on family
and twin studies.
Family studies: Family studies have analyzed
transmission of substance abuse disorders from gen-
eration to generation through families. The basic
approach is to determine if family members of sub-
stance abusers are at greater risk for substance abuse
than individuals in the general population. Family
studies consistently find elevated rates of drug abuse
and dependence (regardless of the specific substance)

in families of drug-dependent probands. For example,
Luthar and Rounsaville studied first-degree relatives
of 298 cocaine abuser probands and found that 40%
of brothers and 22% of sisters of cocaine users, as well
as 5% of parents, displayed drug abuse or dependence
[4]. Similarly, data from the Collaborative Study on
the Genetics of Alcoholism (COGA) [5] showed that
relatives of alcoholics are at substantially higher risk
for cocaine dependence (10.7% of first-degree rela-
tives of alcoholic probands versus 1.1% of first-degree
relatives of control probands) and other stimulant
dependence (5.8% versus 0.8% in those same classes
of relatives). Merikangas and colleagues [6] observed
that among 27 probands with a cocaine use disorder
as the predominant disorder, 14.9%, 18.1%, and
78.3% of 94 first-degree relatives were regular users
of illicit drugs, alcohol, and tobacco, respectively.
Those data demonstrate clear familial clustering of
substance use among relatives of cocaine users. In
addition to familial clustering of all substance use
disorders, some evidence also supports substance-
specific clustering. In the Merikangas study, 7.5% of
the first-degree relatives of cocaine users had a
cocaine use disorder. Family data thus suggest that
substance use disorders generally, as well as cocaine
use disorders specifically, cluster in families. Family
studies, however, cannot separate the influence of
genes from that of family environment. Thus, twin
studies of substance dependence are critical for estab-
lishing the role of genes in substance use disorders.
Twin studies: Twin studies compare the degree to
which traits are shared between dizygotic (DZ) twins
(who share on average 50% of their nuclear genes
identical by descent) to those shared by monozygotic
(MZ) twins (who share 100% of their nuclear genes
identical by descent). To the degree that MZ concord-
ance in a given trait is greater than DZ concordance,
there is evidence for the effect specifically for a role of
genes. Twin studies have consistently supported gen-
etic inheritance as an important explanation of the
risk of substance dependence in general, and psychosti-
mulant dependence in particular. A study of male twins
who served in the US military during the Vietnam era
found higher concordance rates for stimulant depend-
ence (cocaine, amphetamines, and amphetamine-like
drugs) among MZ than DZ twins, with 26.2% and
16.5% of each twin type respectively sharing diag-
noses of stimulant dependence [7]. Further analysis
of the same twin cohort found that the best model for
explaining the data was that a common vulnerability
Chapter 25: Genetics of stimulant dependence
to substance use disorders contributes to dependence
for stimulants and other classes of illicit drug with the
exception of opiates, which appeared to associate with
a substantial substance-specific risk. For stimulant
dependence, 33% of the risk was attributable to gen-
etic factors, and most of that genetic risk was due to
vulnerability to substance dependence, whereas only
approximately 9% of the total variance in risk could
be accounted for by stimulant-specific genetic effects.
This pattern, in which genetic factors common to all
substance dependence disorders accounted for most
of the genetic risk for psychostimulant dependence
was also reported by Kendler and colleagues [8]. In
that study, which examined cocaine dependence and
other stimulant dependence separately in a popula-
tion-based cohort of male twins from the Common-
wealth of Virginia, genetic factors associated with a
general liability to substance dependence accounted
for 63% or 57% of the risk for cocaine or stimulant
dependence, respectively, while an additional term for
substance-specific risk did not improve the model.
Considering only cocaine or stimulant use (but not
abuse/dependence), there appeared to be a small
proportion of risk that was substance-specific (7% or
4% for cocaine or stimulants, respectively), and the
majority of the genetic risk (accounting for 49–51% of
the total variance in risk) was attributable to a genetic
influence on general liability to substance use. How-
ever, one of the limitations of that study was that
there were very few cocaine-addicted individuals
in the total sample, so the power to detect cocaine-
specific effects was quite limited. Kendler and col-
leagues examined substance use disorders in a
broader context of psychiatric disorders, examining
samples of both male and female twins [9]; a set of
genetic factors increasing risk for substance use dis-
orders explains most of the risk for all substance use
disorders in both males and females.
In summary, data from family and twin studies
clearly support an important role for genes in the
risk for substance use disorders. A general risk for
substance dependence, rather then substance-specific
genetic risks, appears to account for most of the
genetic risk for substance dependence. Those findings
make sense given clear epidemiological findings that
the abuse of one drug or class of drugs strongly
associates with abuse of other drugs [10]. As molecu-
lar-genetic studies implicating specific genes and
genomic regions emerge, it may well be useful to
evaluate findings, in part, by their applicability to
307
 Chapter 25: Genetics of stimulant dependence
addiction to more than one class of substance [11].
Molecular studies may also introduce unexpected
complexities into the principle that a common set
of genetic factors modifies risk for multiple substance
use disorders. Thus, a recent report [12] provides
convincing evidence that a particular allele of a non-
synonymous single nucleotide polymorphism (SNP)
in the CHNRA5 gene, encoding the a-5 subunit of the
nicotinic cholinergic receptor, associates negatively
with risk for cocaine dependence, whereas prior stud-
ies by the same investigators clearly support a positive
association of the same variants with nicotine
dependence [13].
Molecular genetic studies of
psychostimulant dependence
in humans
Linkage studies
Linkage analyses examine the co-segregation of
molecular markers with phenotypic traits to identify
regions of the genome causally related to the traits.
Genetic markers at a single gene locus are said to be
genetically linked with a trait if specific molecular
markers (which tag specific places, or loci, in the
genome) are consistently co-inherited within families
together with the trait of interest. Another method for
detecting linkage is to estimate the proportion of
relative pairs with a given trait who share a genetic
marker (or set of markers) identical by descent; if
such sharing occurs to a degree greater than expected
by chance, then positive evidence exists for linkage of
the trait to the genomic region represented by the
markers.
To date, only one genome-wide linkage analysis of
cocaine dependence and related traits has been pub-
lished [14]. This study, of small nuclear families ascer-
tained through polysubstance-using probands who
were dependent on cocaine and identified cocaine as
their drug of choice, is noteworthy for several reasons.
First, rather than focusing only on the presence or
absence of cocaine dependence, a variety of cocaine-
related phenotypes were examined, including CIP,
which occurs in 60–80% of chronic cocaine users
[15–17], as well as a series of traits defined by statistical
analysis of drug-use characteristics such as heaviness of
use, preferred drug, and route of administration. The
clusters derived from those characteristics were
shown to be heritable. For cocaine dependence itself,
308
the investigators observed “suggestive” linkage signals
(i.e. those in which the logarithm of the odds [LOD]
scores were positive, but fell short of genome-wide
significance) on chromosome 10 and two distinct
regions of chromosome 3, in European-American
(EA) families only. For CIP, they observed a genome-
wide-significant LOD score of 3.65 on chromosome 9,
in African-American (AA) families. Their strongest
results were observed for the cluster membership
traits, including a LOD score of 4.66 for membership
in the “Heavy Use, Cocaine Predominant” cluster on
chromosome 12 (in EAs only) and a LOD score of
3.35 for membership in the “Moderate Cocaine and
Opioid Abuse” cluster on chromosome 18 in AA
families only. These findings should set the stage
for fine-mapping studies to identify specific genes
accounting for the observed linkage signals.
Association studies
Association studies seek to identify specific alleles that
consistently co-occur with a trait of interest within a
population. In case-control association analysis, such
co-occurrence is demonstrated by showing a statistic-
ally significant difference in the frequency of a given
allele in a large group of cases (who all exhibit the
trait) as compared to controls. As in all case-control
designs in epidemiology, a fundamental assumption
underlying the approach is that cases and controls are
sampled from the same population. Particularly in
genetic case-control studies it is essential that com-
parison groups be from the same geographical ances-
try. When comparison groups differ in known or
unknown ways with regard to population background
(e.g. if different ethnic groups are represented in the
two groups), and coincidentally, the phenotype being
examined differs in frequency between the population
group(s) represented by cases versus the group(s)
represented by controls, population stratification can
give rise to statistically significant, but biologically
meaningless, results. In family-based association
analysis, differential transmission of an allele from
parents to affected children identifies association in
the presence of linkage. An important advantage of
family-based designs is that they are robust to the
effects of population stratification, as population
background is inherently controlled within each
family, and transmissions within families are the
measure of interest. A detailed treatment of popula-
tion stratification, and strategies for addressing this

potential confound, is beyond the scope of this
chapter. However, Hamer and Sirota provide an
entertaining and easily understood overview of the
problem [18], while Devlin and colleagues provide a
more detailed discussion [19].
It is important to note, for readers who are not
geneticists, that genetic association implies more
than simple statistical association: When evidence
across studies consistently supports association
between a specific marker and a phenotype, we
take that as evidence of linkage disequilibrium (LD)
between the marker and a casual variant(s) nearby in
the genome. LD refers to the correlation between
nearby markers in the genome, such that a causal
allele derived from an ancient common ancestor has
been inherited through generations together with
other nearby (noncausal) alleles that are close
enough to the causal variant that meiotic recombin-
ation between them has occurred only infrequently.
Thus, even over dozens or hundreds of generations
between a mutation event and inheritance of the
disease allele by a given case subject, the alleles of
polymorphisms near the casual allele are likely to be
derived from the same ancestral chromosome, and
therefore “tag” the allele exerting a causative effect on
the phenotype. In the presence of such LD, associ-
ation between an allele at a particular marker and a
phenotype cannot be taken as evidence supporting a
direct biological role of the allele in the phenotype,
without other lines of biological evidence to support
such a conclusion.
A key difference between LD and linkage is that
only a few recombination events are likely to occur in
a given chromosome across the two to four gener-
ations typically examined in linkage studies. Evidence
of linkage to a given marker therefore could be due to
a causal allele quite far (up to many megabases) from
the linked marker. In contrast, LD between a marker
and a phenotype typically implicates a much smaller
candidate region, usually within
100 000 bp in
European populations, although this average “LD
distance” varies widely across the genome, and differs
in populations of differing geographical ancestry.
A key step in evaluating most putative genetic associ-
ations is therefore to evaluate the “LD structure”, or
pattern of marker-to-marker correlation, surrounding
the implicated allele. That process, of evaluating LD
structure of a candidate gene or genomic region has
been greatly facilitated by the HapMap project (http://
www.hapmap.org/), which reports allele frequencies,
Chapter 25: Genetics of stimulant dependence
and the degree of LD to adjacent markers, for
approximately 4 million SNPs, in samples from sev-
eral different populations (Japanese, European-
American, British, Yoruban/Nigerian, and Chinese).
Using software tools such as Haploview (http://www.
broad.mit.edu/mpg/haploview/), researchers can esti-
mate how many SNPs will be required to capture
most of the common variation present in a region
within a population of similar ancestry to the
HapMap reference populations, and which SNPs to
target for genotyping. Thus, for example, if two SNPs
exhibit extremely high LD, as indicated by the square
of their correlation coefficient (r2) approaching unity,
it is only necessary to genotype one of them to capture
the genetic variation represented by both SNPs,
because they yield redundant information.
Numerous published studies have attempted to
evaluate association between molecular polymorphic
markers (usually SNPs) and stimulant abuse or
dependence. Until recently, such association studies
have almost always been guided by candidate-gene
hypotheses generated from knowledge about the
neurobiology of cocaine dependence, with a particu-
lar focus on the role of dopamine as an important
mediator of the reinforcing properties of cocaine and
other substances of abuse. Candidate-gene studies of
cocaine dependence have therefore been dominated by
studies of loci encoding key proteins mediating the
synaptic actions of dopamine (DA). Attention has also
been paid to some genes encoding proteins involved
in serotonin (5HT)- or norepinephrine (NE) -mediated
neurotransmission, based on the fact that cocaine is a
potent inhibitor of the uptake of those monoamines as
well. Finally, a variety of neuropeptides are co-localized
with, or play a role in regulating, DA, NE, and 5HT, so
some studies have examined genes relevant to neuro-
peptides and neurotrophic factors.
As in many genetic-association studies of other
complex disorders, many initial “positive” findings
regarding genetic associations to cocaine dependence
have not been replicated in subsequent studies.
Although the reasons vary and are beyond the scope
of this chapter, one very important factor is that
until very recently, almost all association studies of
cocaine-dependence suffered from extremely low
statistical power, with early studies often having
fewer than 100 cocaine-dependent or control subjects
of a single ethnic group per comparison group, and
many studies examined only a single variant (usually
a SNP). In such studies, even in the context of theoretical
309
 Chapter 25: Genetics of stimulant dependence
considerations that might support the hypothesis
that a given polymorphism associates with cocaine
dependence, the a priori chance of a negative result
is so high that levels of nominal statistical significance
that might be indicative of true positive findings in
simpler settings are likely to be due to chance. The
selected results reviewed below therefore must be
viewed as preliminary and in need of replication.
It seems fair to say at present that no locus has yet
been identified that can be said to show a confirmed
genetic association with cocaine dependence or
cocaine-associated behavioral phenotypes.
Studies of dopamine-related genes
Dopamine transporter gene (SLC6A3): The DA
transporter protein, encoded by the locus SLC6A3,
modulates DAergic tone by transporting DA into
the presynaptic neuron following synaptic release,
thereby terminating its action. Cocaine’s primary
addiction-related pharmacological action is thought
to be inhibition of DA by the DAT. At least three lines
of evidence suggest that variation at SLC6A3 influ-
ences human responses to cocaine: (1) several poly-
morphisms identified in SLC6A3, such as T265C
(Val55Ala) and T1246C (Val382Ala), are associated
with deficient DA uptake in in vitro studies [20];
(2) the SNP, G2319A, and a variable-number of
tandem repeats (VNTR) polymorphism found in the
30 -untranslated region of exon 15 of the SLC6A3 gene
have been reported to associate with differences
in availability of DA transporter binding sites as esti-
mated by single proton emission tomography
imaging (SPECT) [21]; and (3) the 9-repeat allele of
the foregoing VNTR, which associated in the SPECT
study with higher ligand-binding potential (presum-
ably reflecting higher levels of available neuronal DA
transporter molecules) was reported to associate with
CIP [22, 23]. Some, but not all, subsequent studies
have supported those conclusions [23]. Thus, the
9-repeat allele associated with methamphetamine-
induced psychosis in a Japanese sample [24] but not
in Chinese cocaine users [25]. A human-laboratory
study suggested that individuals homozygous for
the 9-repeat allele showed diminished subjective
responsiveness to oral d-amphetamine, as compared
to individuals homozygous for the 10-repeat allele,
or heterozygotes [26]. Although that study did not
specifically assess amphetamine-induced paranoia, it
did show an association to amphetamine-induced
310
anxiety. Given phenomenology similarities between
anxiety and paranoia, it is possible that associations
between genetic variation and the two phenotypes
represent the same biological pathway. Additional
evidence regarding association between variation at
SLC6A3 and human responses to psychostimulants
derives indirectly from genetic studies of attention-
deficit hyperactivity disorder (ADHD). Linkage and
association implicate SLC6A3 in risk for ADHD in
some (but not all) studies [27]. Psychostimulants such
as methylphenidate and d-amphetamine are front-
line treatments for ADHD symptoms, and several
studies have suggested that variation at SLC6A3 asso-
ciates with altered therapeutic response to psychosti-
mulants in ADHD [reviewed in 28]. In summary,
there is no evidence to support a direct association
between variation at SLC6A3 and cocaine or amphet-
amine dependence, but some evidence supports the
hypothesis that variation at that locus associates with
subjective response to stimulants.
DRD2: Because DA receptors mediate many of
the rewarding effects of stimulants, genetic alterations
influencing dopamine receptors are obvious candi-
date genes for cocaine and amphetamine dependence.
Many early genetic association studies of substance-
use-related phenotypes focused on DRD2, the locus
encoding the dopamine D2 receptor, with a particular
focus on a SNP of no known function, located
approximately 10 000 bp 3’ to the DRD2 coding
region. This variant, known as TaqI A is a SNP that
was typed as a restriction fragment length poly-
morphism (RFLP), and was named for the restriction
enzyme, Taq1, used for such assays. Claims of associ-
ation of this SNP have been made for polysubstance
abuse [29], age of onset of substance abuse [30], and
psychostimulant [31] and cocaine dependence [32].
The RFLP TaqI B, another SNP of no clear function,
has also been claimed to associate with polysubstance
abuse in Caucasians [29] and with cocaine depend-
ence [33]. All of those studies were very underpow-
ered, and therefore must be viewed with extreme
caution. More recently, as more detailed knowledge
has emerged on the genomic region surrounding the
Taq1 A polymorphism, evidence suggests that the
variant may actually be more relevant to loci other
than DRD2. Most notably, that Taq1 A SNP is actually
a nonsynonymous variant in the coding region of
ANKK1, encoding the ankrin repeat and kinase
domain-containing protein 1. Gelernter and col-
leagues have conducted several association studies

of the Taq1 A SNP, along with several dozen addi-
tional SNPs, to generate a detailed picture of linkage
and association of that genomic region to substance
dependence phenotypes including nicotine depend-
ence and alcohol dependence [34, 35]. Thus, the evi-
dence for an association to cocaine dependence of
variants in DRD2 remains scant, despite early claims
to the contrary.
DRD3: Dopamine D3 receptor (D3R) has been
reported as a mediator in cocaine use. Some studies
in humans also suggest a potential role of the D3R in
cocaine dependence. For example, the D3R is present
in high density in the nucleus accumbens; the prime
site of the dopamine reward pathway. Its density in
this area is two-fold higher in victims of accidental
cocaine overdose [36, 37]; of note, such decedents
often exhibit psychotic excited delirium immediately
prior to death. These observations suggest that cocaine
use leads to a neuroadaptive elevation in D3R density
in response to increased synaptic dopamine and that
the D3R may be one mediator of both the reinforcing
and psychosis-promoting effects of cocaine [38, 39].
Several early studies tested for associations
between the DRD3 SNP rs6280 (a nonsynonymous
SNP also known as ser9gly because it encodes either
serine or glycine at amino acid position 9) and
cocaine dependence, either by itself [40, 41] or in
the context of comorbid schizophrenia [42]. None of
those studies reached definitive conclusions, all
suffered from extremely low statistical power (fewer
than 100 cocaine-dependent subjects of a single ethnic
group per study), and each study examined only a
single SNP (albeit one of clear biological function, as
it alters the predicted amino acid sequence of the
protein). A more recent larger study [43] of rs6280
(N ¼ 730 hospitalized cocaine addicts; 781 healthy
blood donors who denied drug use) found no associ-
ation between genotypes or alleles at the locus and
cocaine dependence. However, that study can be criti-
cized because its comparison groups were ethnically
mixed, with different proportions of self-identified
racial groups, and the analysis did not control for that
potential source of stratification. Overall, however, the
evidence supporting a clear association between vari-
ation at DRD3 and cocaine dependence is quite weak.
COMT: COMT encodes catechol-O-methyltrans-
ferase, a major enzyme of DA and NE catabolism.
A nonsynonymous SNP at COMT, rs4680, is often
referred to as the “val-met” SNP, owing to the fact
that it encodes either a valine or methionine at amino
Chapter 25: Genetics of stimulant dependence
acid position 158 (in the long, membrane-bound
form of the protein) or 108 (in the shorter, soluble
protein). This SNP has been intensively studied as
it has been shown to alter the thermal stability of
the enzyme, such that the methionine-containing
allelomorph is relatively unstable at physiological
temperature, leading to lower enzyme activity [44].
Lower COMT enzyme activity could lead to slower
rates of DA and NE degradation, thus altering synap-
tic activity of these catecholamines. Lohoff and col-
leagues [45] recently reported an association study of
cocaine dependence examining the val-met SNP,
together with two additional SNPs at COMT, located
at opposite ends of the coding region. The study
examined African-American subjects who were either
dependent on cocaine or free of addictive disorders.
They found evidence for an association between
the val-met SNP and cocaine dependence, with the
low-activity associated met allele occurring more
frequently in dependent individuals than in controls
(35% versus 29%, respectively; Bonferroni-corrected
p ¼ 0.014). A study of the val-met SNP in Chinese
methamphetamine abusers versus healthy controls
found the opposite direction of association, with the
val allele occurring more frequently in abusers than
controls [46]. A study of methamphetamine abuse in
Japan found no association to the val-met SNP [47].
The reasons for the apparent discrepancies among
these three studies are not entirely clear, but notable
differences among the studies include the ethnicities
of the comparison groups and the diagnoses exam-
ined (abuse of methamphetamine versus dependence
on cocaine). The evidence supporting a role for gen-
etic variation at COMT in psychostimulant depend-
ence requires clarification and replication.
Dopamine b-hydroxylase (DbH): Plasma DbH
activity is a genetic trait controlled by only a few genes
[48, 49]. Most of the variance in plasma DbH activity
is explained by variation in plasma levels of DbH
protein, as shown by strong correlations (r2  0.80)
between plasma DbH activity and plasma levels of
DbH immunoreactive protein. DbH activity is a
highly heritable trait. Abundant evidence indicates
that the structural gene encoding the protein, DbH,
regulates much of the genetic variation in the trait
(reviewed in [16]). Prior work has shown several
polymorphisms at DBH to be two polymorphisms,
an insertion/deletion polymorphism (DBH*5’-ins/del
[a biallelic 19-nucleotide insertion/ deletion approxi-
mately 4.7 kb 5’ upstream from the transcription
311
 Chapter 25: Genetics of stimulant dependence
initiation site]) and a SNP (DBH*444g/ a, rs1108580),
respectively, are in strong LD and show similar asso-
ciation with plasma levels of DbH. A haplotype of
these two polymorphisms not only correlated with
low DbH levels but also was associated with CIP in
a sample of American-Caucasians who abuse cocaine
[50]. Interestingly, one polymorphism, –1021C/T
(rs1611115), at the promoter region is found to
be associated with low plasma DbH activity in
European-Americans, African-Americans, and Japanese
[51, 52]. This functional polymorphism may also have
pharmacogenomic significance since disulfiram, which
inhibits DbH, could be more effective in individuals
with this SNP [53], and subjects homozygous for the
“very low activity” T allele at –1021C–>T show an
increased propensity to paranoia over time during
cocaine self-administration [54].
Other candidate genes
Prodynorphin gene: Pharmacological and clinical
studies have implicated the dynorphin peptides and
the kappa-opioid receptor are important in the
rewarding properties of cocaine, heroin, and other
drugs of abuse. Prodynorphin is an opioid peptide
precursor molecule which has been shown to be
associated with cocaine, opiate, and other drug abuse
in animals. A repeat polymorphism in the promoter
region of the preprodynorphin gene may cause
changes in cellular functioning with importance for
cocaine addiction. The polymorphism is a 68-nucle-
otide tandem repeat element in the putative promoter
region of the gene. This polymorphism, which con-
tains a putative AP-1 transcription complex (c-Fos/
c-Jun) binding site, is found in one to four copies.
Some have hypothesized that alleles of this gene may
cause an increase in dynorphin levels and, therefore,
may be protective against cocaine addiction [55]. Sev-
eral studies have examined an association of this
polymorphism with drug dependence with conflicting
results. One study showed that Hispanic individuals
with three or four copies of the repeat have a lower
risk for development of cocaine dependence [56].
However, two subsequent studies using more strin-
gent diagnostic criteria showed increased risk for
cocaine dependence and cocaine/alcohol co-depend-
ence in African-Americans with three or four repeats
[57, 58]. More recently, Yuferov et al. [59] reported a
functional haplotype to be associated with reduced
PDYN expression in human brain.
312
Cannabinoid-related genes: A trinucleotide repeat
polymorphism in the 30 flanking region of the canna-
binoid receptor 1 (CNR1) gene is associated with
intravenous drug abuse (cocaine, amphetamine, or
heroin) [60]. A study of 22 polymorphisms in CNR1
identified a haplotype in an intronic 50 region of the
gene that is associated with substance (cocaine, opiate,
alcohol, or other drug) abuse [61]. Recently, Zuo and
colleagues [62] reported an association between a
two-SNP haplotype at CNR1 and cocaine dependence,
originally observed in a European-American family
sample, and then replicated in a European-American
case-control sample. Fatty amide acid hydrolase,
encoded by the FAAH gene, is an enzyme that metab-
olizes endogenous ligands of the cannabinoid recep-
tors. A functional SNP that alters the sensitivity of the
enzyme to protease in vitro is associated with drug
and alcohol abuse [63]. Taken together, the evidence
supporting a role for variation in genes related to
cannabinoid-mediated neurotransmission in cocaine
dependence appears as strong or stronger than for
any other locus.
Genome-wide association studies
A genome-wide association study (GWAS), also
known as whole genome association study (WGA
study), is defined as an association study that surveys
most of the genome for causal genetic variants, with-
out the requirement of prior pathophysiological
knowledge about the disease or any prediction of
candidate genes. As in candidate gene association
studies, GWAS also requires two groups of partici-
pants: people with the disease and similar people
without the disease. With the advancement of
molecular genetic technologies, systematic associ-
ation study of the whole genome can now be con-
ducted using DNA chips or arrays. After obtaining
samples from each participant, the set of markers
such as SNPs are scanned and data are analyzed. If
genetic variations are more frequent in people with
the disease, the variations are said to be “associated”
with the disease. The associated genetic variations
are then considered pointers to the region of the
human genome where the disease-causing problem
resides. Since the entire genome is analyzed for
the genetic associations of a particular disease, this
technique allows the genetics of a disease to be inves-
tigated in a nonhypothesis-driven manner. When a

GWAS detects association between a SNP and the
disease, this signal usually represents association
with a set of several highly correlated SNPs in
strong LD.
A family-based low-density SNP scan, in which
families ascertained for sibling pairs affected by
either cocaine dependence or opioid dependence
[64], was recently reported. That study included
339 African-American families and 334 European-
American families, and examined a total of 1699
individuals. A variety of phenotypes were investi-
gated, including cocaine dependence (with a total of
1284 individuals affected) and CIP (873 affected).
The investigators made the a priori decision to
focus on polymorphisms that yielded nominally
significant results (i.e. p < 0.05, uncorrected for
multiple testing) in both African-American and
European-American families. Several interesting
findings emerged. A SNP in the MANEA locus,
encoding a-endomannosidase, associated with CIP
(p ¼ 0.00005). Although the finding did not with-
stand strict correction for multiple testing (which
required, in this study, p < 0.00001), such a correc-
tion is arguably overly conservative, due to the cor-
relation among at least some of the SNPs examined.
A follow-up study examined additional SNPs across
the MANEA locus, first in the original family
samples, and then in an independently ascertained
case-control sample. Several SNPs showed nominally
significant association to CIP in the denser SNP
analysis. In the replication sample, several of these
also showed significant association to CIP in African-
Americans, and an additional SNP showed asso-
ciation in European-Americans [65]. While not
definitive, those studies suggest that MANEA and its
product, a-endomannosidase, a lysosomal enzyme
deficiency which results in rare neuronal storage
diseases, merit further investigation with regard to
CIP, even though no obvious mechanism suggests
itself that would explain why variation in MANEA
would alter risk for CIP.
In the original family-based study by Yu et al.
[64], a SNP within the 3’ untranslated region of the
STY13 locus was nominally associated with cocaine
dependence. The gene encodes synaptotagmin VIII,
a protein involved in regulation of synaptic neuro-
transmission by calcium. While the level of statistical
significance of this association to cocaine dependence
(p ¼ 0.003 in the combined samples) was modest,
the fact that the gene product regulates synaptic
Chapter 25: Genetics of stimulant dependence
transmission makes the possible association a very
interesting new lead in understanding cocaine
dependence.
At the time of this writing, there has been one
GWAS published on methamphetamine dependence
[66]. Two independent case-control comparisons were
evaluated by pooled genotyping up to
500 000 SNPs
across the genome. The first compared 140 metham-
phetamine-dependent individuals of Han Chinese
ethnicity to 240 healthy control Han Chinese individ-
uals. The second set compared 100 methampheta-
mine-dependent and 100 healthy control individuals
from Japan. The investigators used a pooled genotyp-
ing strategy, in which microarray-based genotyping
methods are applied to pooled samples of DNA from
20 subjects per sample. The relative signal intensities
representing the respective alleles within each pooled
sample are, in theory, proportional to the frequency
of the alleles. The results of the analysis provided
interesting suggestive evidence for SNPs differing
in allele frequency between cases and controls. The
investigators focused on differences that occurred in
both the Taiwanese and Japanese comparisons. In an
intriguing strategy for statistical analysis, the investi-
gators evaluated not just the degree of allelic signal
difference between cases and controls, but also the
clustering of differences along the genome. Clusters
of multiple SNPs showing differential signals between
cases and controls within specific chromosomal
regions were shown by Monte-Carlo simulation to
be unlikely to be due to chance. The analysis impli-
cated a number of interesting candidate genes as
associated to methamphetamine dependence, includ-
ing genes encoding several enzymes likely to regulate
intracellular signaling (e.g. PDE6C, encoding a cyclic-
GMP phosphodiesterase and PRKG1, encoding a
cGMP-dependent protein kinase), as well as a variety
of transcription factors and cell-adhesion molecules.
The results of the study require replication in larger
independent samples, as the sample sizes used in
the study were small. In addition, the pooled geno-
typing strategy used is subject to errors in estimating
relative “allele frequencies” because even small errors
in relative concentrations of DNA from individual
subjects can distort signal intensities. Nevertheless,
the study represents an important step in association
analysis of stimulant dependence. It is interesting
to note that none of the loci implicated by the study
were “obvious” candidate genes such as those dis-
cussed above.
313
 Chapter 25: Genetics of stimulant dependence

Summary and conclusions dependence comparable in genotyping density to

those performed on other complex disorders has yet
In this chapter, we have reviewed evidence supporting
been performed. A GWAS of methamphetamine
the hypothesis that genetic inheritance plays a sub-
dependence, while yielding some interesting leads,
stantial role in dependence on cocaine and (to a less
requires replication in light of its small size, and reli-
well-studied degree) other illicit psychostimulants.
ance on pooled genotyping. While several intriguing
The role of genes in cocaine dependence, however,
candidate-gene associations between specific loci and
may largely reflect a more general liability to develop
cocaine dependence have been reported, to date there
dependence on a variety of substances. Studies of
has yet to be a definitively replicated result reported.
molecular genetic mechanisms in cocaine dependence
Clearly, more work is required in the human genetics of
remain in an early stage of development. Only one
stimulant dependence, to identify and characterize how
genome-wide linkage study of cocaine dependence
specific genes influence risk for this set of disorders.
has been reported to date, and no GWAS of cocaine

References 15. Brady KT, et al. J Clin Psychiatry

1. United Nations Office for Drug
Control and Crime. World Drug
Report 2008. Vienna: United
Nations; 2008.
2. Substance Abuse and Mental
Health Services Administration
(SAMHSA). Results from the 2007
National Survey on Drug Use and
Health: National Findings.
Rockville, MD: SAMHSA; 2008.
3. Nulman I, et al. Clin Invest Med
2001;24:129–137.
4. Luthar SS, et al. Int J Addict
1993;28:415–434.
5. Nurnberger JI Jr., et al. Arch Gen
Psychiatry 2004;61:1246–1256.
6. Merikangas KR, et al. Arch Gen
Psychiatry 1998;55:973–979.
7. Tsuang MT, et al. Am J Med Genet
1996;67:473–477.
8. Kendler KS, et al. Am J Psychiatry
2003;160:687–695.
9. Kendler KS, et al. Arch Gen
Psychiatry 2003;60:929–937.
10. Kessler RC, et al. Arch Gen
Psychiatry 1997;54:313–321.
11. Li MD, et al. Nat Rev Genet
2009;10:225–231.
12. Grucza RA, et al. Biol Psychiatry
2008;64:922–929.
13. Bierut LJ, et al. Am J Psychiatry
2008;165:1163–1171.
14. Gelernter J, et al. Am J Med
Genet B Neuropsychiatr Genet
2005;136B:45–52.
1991;52:509–512.
16. Cubells JF, et al. Drug Alcohol
Depend 2005;80:23–33.
17. Satel SL, et al. Am J Psychiatry
1991;148:495–498.
18. Hamer D, et al. Mol Psychiatry
2000;5:11–13.
19. Devlin B, et al. Genet Epidemiol
2001;21:273–284.
20. Lin Z, et al. Pharmacogenomics
2003;3:159–168.
21. Jacobsen LK, et al. Am J Psychiatry
2000;157:1700–1703.
22. Gelernter J, et al. Neuro-
psychopharmacology 1994;
11:195–200.
23. Haile CN, et al. Behav Genet
2007;37:119–145.
24. Ujike H, et al. Pharmacogenomics
2003;3:242–247.
25. Hong CJ, et al. J Neural Transm
2003;110:345–351.
26. Lott DC, et al. Neuro-
psychopharmacology 2005;
30:602–609.
27. Friedel S, et al. Mol Psychiatry
2007;12:923–933.
28. McGough JJ. Biol Psychiatry
2005;57:1367–1373.
29. O’Hara BF, et al. Hum Hered
1993;43:209–218.
30. Comings DE, et al. Drug Alcohol
Depend 1994;34:175–180.
31. Persico AM, et al. Biol Psychiatry
1996;40:776–784.
32. Young RM, et al. Addict Behav
2004;29:1275–1294.
33. Noble EP, et al. Drug Alcohol
Depend 1993:33:271–285.
34. Gelernter J, et al. Hum Mol Genet
2006;15:3498–3507.
35. Yang BZ, et al. Hum Mol Genet
2007;16:2844–2853.
36. Segal DM, et al. Brain Res Mol
Brain Res 1997;45:335–339.
37. Staley JK, et al. J Neurosci
1996;16:6100–6106.
38. Caine SB, et al. Science
1993;260:1814–1816.
39. Rodriguez De Fonseca F, et al. Eur
J Pharmacol 1995;274:47–55.
40. Comings DE, et al. Mol Psychiatry
1999;4:484–487.
41. Freimer M, et al. Addict Biol
1996;1:281–287.
42. Krebs MO, et al. Mol Psychiatry
1998;3:337–341.
43. Messas G, et al. Psychiatr Genet
2005;15:171–174.
44. Lachman HM, et al. Pharma-
cogenetics 1996;6:243–250.
45. Lohoff FW, et al. Neuro-
psychopharmacology 2008;
33:3078–3084.
46. Li T, et al. Am J Med Genet B
Neuropsychiatr Genet 2004;
129B:120–124.
47. Suzuki A, et al. Psychiatr Genet
2006;16:133–138.
48. Oxenstierna G, et al. J Psychiatr
Res 1986;20:19–29.

314

49. Weinshilboum RM. Hum
Genet Suppl, 1978;101–112.
50. Cubells JF, et al. Mol Psychiatry
2000;5:56–63.
51. Tang YL, et al. Eur J Hum Genet
2007;15:878–883.
52. Zabetian CP, et al. Am J Hum
Genet 2001;68:515–522.
53. Carroll KM, et al. Arch Gen
Psychiatry 2004;61:
264–272.
54. Kalayasiri R, et al. Biol Psychiatry
2007;61:1310–1313.
Chapter 25: Genetics of stimulant dependence
55. LaForge KS, et al. Eur J Pharmacol
2000;410:249–268.
56. Chen AC, et al. Am J Med Genet
2002;114:429–435.
57. Dahl JP, et al. Am J Med Genet B
Neuropsychiatr Genet 2005;
139B:106–108.
58. Williams TJ, et al. Addict Biol
2007;12:496–502.
59. Yuferov V, et al. Neuropsycho-
pharmacology 2008;34:1185–1197.
60. Comings DE, et al. Mol Psychiatry
1997;2:161–168.
61. Zhang PW, et al. Mol Psychiatry
2004;9:916–931.
62. Zuo L, et al. Neuropsycho-
pharmacology 2009;34:
1504–1513.
63. Sipe JC, et al. Proc Natl Acad Sci
U S A 2002;99:8394–8399.
64. Yu Y, et al. Hum Genet 2008;
123:495–506.
65. Farrer LA, et al. Arch Gen
Psychiatry 2009;66:267–274.
66. Uhl GR, et al. Arch Gen Psychiatry
2008;65:345–355.
315
 Genetics of personality disorders
Chapter
26 C. Robert Cloninger
Personality disorders provide an excellent illustration
of the challenges facing contemporary psychiatric
genetics. In several ways, what we know about the
genetics of personality disorders actually allows us to
distinguish approaches to psychiatric genetics that
will be fruitful more readily than do studies of clinical
syndromes like schizophrenia or mood disorders.
I say that because it was recognized that personality
disorders are complex quantitative traits earlier than
for other psychiatric syndromes [1, 2]. Several lines of
research document the inadequacy of phenotypic
assessment based on categories or linear dimensions
of pathological symptoms for understanding the gen-
etics, development, and evolution of personality dis-
orders. Recognition of the functional complexity of
personality disorders may indicate solutions that will
be fruitful for psychiatric genetics in general.
First I will briefly outline the basic observations
that must be faced in order to understand the genetics
of personality disorders. Then I will suggest a general
functional approach based on the evolution of human
brain functions that may be of general utility [3]. In the
past, psychiatric geneticists have often based their ana-
lyses on unrealistic linear assumptions about pheno-
typic traits, thereby failing to describe or appreciate the
nonlinear dynamics of the evolution and development
of human brain systems [4]. More realistic approaches
are now possible, however, and can be informed by
combining an understanding of the genetics of person-
ality with the evolution of the brain systems that
underlie healthy personality and its disorders.
Structure of abnormal personality is the
same as that of normal personality
Traditionally personality disorders have been
described as a set of discrete categories, as in DSM-
Principles of Psychiatric Genetics, eds John I. Nurnberger, Jr. and Wade H. Berrettini. Published by Cambridge University
Press. © Cambridge University Press 2012.
316
IV [5], but taxonomic analyses have long shown that
such categorical descriptions are inadequate [6].
There is no rigorous evidence for discrete boundaries
between categories or even clusters [4], and most
patients with any personality disorder satisfy criteria
for two or more putative categories [7, 8].
Individual differences in personality can be meas-
ured well in terms of multiple quantitative dimen-
sions of personality [2, 8]. One of the most robust
findings about personality, but one that is surprising
to many psychiatrists, is that the same dimensions of
personality are observed whether one begins with
normal personality variation in the general commu-
nity, with abnormal personality traits, or with symp-
toms of personality disorders in treatment samples.
There are three main lines of evidence indicating
that normal and abnormal personality traits and
symptoms all rest on the same foundation. First,
people with personality disorders or other forms of
psychopathology have extreme values on one or
more personality dimensions, but the dimensional
structure is the same in samples from the general
community and from psychiatric treatment facilities
[7, 9, 10]. Second, the genetic structure of normal
and abnormal personality traits in studies of twins is
indistinguishable, suggesting that influences on
normal and abnormal personality act through
systems common to both, whether the twins are
reared together [10] or apart [11]. Third, variability
in normal personality traits that are descriptive of
the general population are diagnostic of personality
disorders [7, 12]. Therefore, normal personality traits
share a common functional foundation with abnor-
mal personality traits and most, if not all, mental
disorders. The same biopsychosocial systems influ-
ence individual differences in normal personality and
its disorders.

Healthy personality is more than the
absence of personality disorder
Health is a state of physical, mental, social, and spirit-
ual well-being, rather than merely the absence of
disease according to standard definitions [13]. Mental
health is a state of well-being characterized primarily
in terms of a well-integrated personality [14]. A well-
integrated personality is characterized by a person
being self-directed, cooperative, and self-transcendent
[15, 16]. Clearly the absence of particular abnormal
traits, such as anxiousness or callousness, does not
guarantee that personality is well-integrated.
Consequently, a narrow focus on the genetics of
abnormal traits or personality disorders is likely to
ignore the fact that evolution always operates on the
adaptive functioning of whole individuals in popula-
tions, not on particular structures or traits that differ
between individuals. To understand the genetics of
personality it is essential to characterize the dynamics
of adaptive functioning within an individual, which
makes the person more or less reproductively fit [3].
Well-functioning of personality
depends on integration of all aspects
of health
It is surprising to many specialty trained physicians
that health depends on the integration of all aspects of
a human being – sexual, corporeal, emotional, intel-
lectual, and spiritual. Yet it is well established that
there is strong interdependence among physical
health, personality disorder, emotional distress, social
dysfunction, and spiritual dysfunction [14, 15, 17].
Common genetic factors account for 45–60% of the
associations among subjective well-being, perceived
health, and somatic illness [18]. Therefore, the inte-
grated functioning of the whole individual needs be
considered to understand heritable functions, rather
than focusing on particular traits or disorders [3].
Implications for research on the
genetics of personality disorder
The robust observation that personality is a complex
biopsychosocial matrix of functions of whole persons
in populations has strong implications for how to
approach the study of personality and its disorders.
Complex or quantitative traits always involve
Chapter 26: Genetics of personality disorders
extensive gene–gene and gene–environment inter-
action [19]. Nonlinear dynamic systems tend to main-
tain intermediate adaptive optima so that extreme
high or low values on traits are disadvantageous
[15]. The liability to illness is proportional to the
inverse of the prevalence for semi-quantitative traits
(that is, those that can be ranked by severity: rare
traits have greater familial loading than common
traits) [20]. More information is available about com-
plex traits when they are measured quantitatively than
when they are reduced to categorical disorders. In
other words, categorical, semi-quantitative, and quan-
titative measurements provide increasing amounts of
information about the underlying causes of personal-
ity and its disorders. These theoretical expectations
can be and have been tested with available informa-
tion about personality and its disorders.
Evidence of quantitative multifactorial
variability
The personality disorder that has been most exten-
sively studied is antisocial personality. Family, twin,
and adoption studies have been carried out showing
the moderate heritability of antisocial personality dis-
orders. A wide variety of diagnostic criteria have been
used, but regardless of the criteria heritability has
been consistently moderate. Adoption studies of
psychopathy [21], antisocial personality [22, 23], and
criminality [24, 25] were carried out in the United
States and Scandinavia, all showing substantial herit-
able influences. Interactions between genetic predis-
position and childhood rearing in an unstable or
hostile environment were also demonstrated for petty
criminality and antisocial personality. Monozygotic
(MZ) twins were also more often concordant for
criminality than were dizygotic (DZ) twins, as
reviewed in detail elsewhere [26]. A recent meta-
analysis of 51 twin and adoption studies estimated
that there were moderate proportions of variance
due to additive genetic influences (0.32), nonadditive
genetic influences (0.09), shared environmental influ-
ences (0.16), and nonshared environmental influences
(0.43) [27]. Diagnostic criteria, assessment method,
zygosity determination method, and age, but not
gender, were significant moderators of the magnitude
of genetic and environmental influences on antisocial
behavior.
However, when consistent diagnostic criteria and
ascertainment methods are applied, gender differences
317
 Chapter 26: Genetics of personality disorders
Table 26.1 A study of categorical personality disorders assessed by Structured Clinical Interview for DSM-IV-TR Axis II Personality
Disorders (SCID-II) interviews in 92 monozygotic (MZ) twin pairs and 129 dizygotic (DZ) twin pairs in Norway.
Diagnoses MZ correlation 100 (r
SE)
Any personality disorder 58
10
Any Cluster A 37
14
Any Cluster B 60
11
Any Cluster C 61
9
Adapted from Torgersen et al. [29].
have a substantial impact in both twin and family
studies. The concordances for criminality and anti-
social personality were well predicted by differences
in severity of liability by a function proportional to
the inverse of the prevalence in the general population,
indicating that susceptibility was to an underlying
quantitative variable-like personality [20, 26, 28]. For
example, antisocial personality disorder is more rare in
women than in men, and antisocial women carry a
greater genetic loading (measured by more antisocial
relatives) than do antisocial men.
Such findings about the quantitative nature of
variation in personality traits provides an important
justification for analysis of the number of symptoms
of personality disorders, rather than focusing on cat-
egorical classifications based on a more or less arbi-
trary number as a diagnostic threshold. As we will see,
studies using only categorical cut-offs have much less
information and also fail to consider the full range of
variation in the study population.
Inheritance of personality disorder
symptoms and categories
Some studies have been carried out using categorical
diagnoses despite the theoretical and practical disad-
vantages of categorical approaches. Svenn Torgersen
and his colleagues [29] carried out a twin study based
on a wide range of categorical diagnoses in Norway.
They ascertained 92 MZ and 129 DZ twin pairs in
which at least one proband had a diagnosis of person-
ality disorder, and divided the cases according to
DSM clusters and categories. The probandwise con-
cordances for any definite personality disorder were
40% for MZ pairs and 29% for DZ pairs, indicating
substantial genetic influences (p < 0.01), as summar-
ized in Table 26.1. Concordance for membership in
personality disorder clusters also could not be
318
DZ correlation 100 (r
SE) Heritability (%
SE)
36
10 44
20
9
11 37
25
31
12 59
23
23
11 59
20
explained without taking genetic variability into
account. Estimates of heritability for the clusters and
specific categories were moderate (i.e. between 40 and
60%), much as observed for quantitative measures of
normal [30, 31] and abnormal personality traits [32].
The correlations between DZ pairs were usually less
than half of those of MZ pairs, suggesting that gene–
gene and gene–environment interactions are import-
ant for categorical diagnoses as they are for personality
dimensions in twins reared apart [33–36]. The small
number of individuals with specific diagnoses made
any conclusions about the heritability of the categor-
ies imprecise, as shown by the large standard errors
in Table 26.1. Nevertheless, the consistency of the
overall results with those of studies of personality
dimensions support the moderate heritability of per-
sonality disorders.
Even stronger evidence for the moderate heritabil-
ity of personality disorders has been provided by
more recent studies of the inheritance of the number
of symptoms in each of 10 categories of personality
disorders. The sample included 1386 Norwegian twin
pairs aged 19–35 as assessed by SIDP-IV interviews.
The total heritability and the heritability unique to
symptoms of each disorder within the three DSM-IV
clusters are shown in Table 26.2. The data are com-
piled from separate reports for the number of symp-
toms of disorders in personality cluster A [37], cluster
B [38], and cluster C [39]. The total heritability of
liability to the number of symptoms in each category
was weak to moderate, ranging from 28% for para-
noid personality symptoms to 38% for antisocial per-
sonality symptoms.
The unique heritability in Table 26.2 indicates
variability for specific disorders that are not shared
by other disorders in the same cluster. For example,
obsessive–compulsive personality disorder has 85%
unique heritability, indicating that its genetic
 Chapter 26: Genetics of personality disorders

Table 26.2 Heritability of the number of symptoms in determinants are largely distinct from those of
personality categories in a sample of 1386 Norwegian twin pairs
aged 19–35 years as assessed by Structured Interview for DSM-IV
avoidant and dependent personality disorders. This
Personality (SIDP-IV) interviews. Total heritability and the supports the recommendation that obsessional traits
heritability unique to each disorder within the three clusters constitute a fourth cluster with high persistence or
are shown.
anankastic traits [7]. Likewise unique genetic deter-
Clusters of Total Unique minants predominate for dependent personality
personality disorder heritability heritability (52%) in cluster C and for schizoid personality
symptoms (%) (%) (74%) and paranoid personality (54%) in cluster A.
Cluster A (“odd”)
Cluster B is relatively more homogeneous but even
Schizotypal 26 0 there antisocial and borderline personality disorders
Paranoid 21 57 are more closely related to each other than to histri-
Schizoid 28 74 onic and narcissistic disorders.
The extent of sharing of common genetic factors
Cluster B (“dramatic”)
Antisocial 38 15
among the various personality disorders was further
Histrionic 31 0 evaluated by estimating the number of common
Narcissistic 24 10 factors and unique determinants of all 10 DSM-IV
Borderline 35 0 personality disorders simultaneously by path analysis
[40]. Three latent genetic factors were inferred as
Cluster C (“anxious”)
Avoidant 35 17
summarized in Table 26.3. The first factor is a
Dependent 31 52 common determinant of all personality disorders.
Obsessive–compulsive 27 85 This common personality factor corresponds to what
is measured as neuroticism or as low self-directedness
Compiled from Kendler et al. [37] for Cluster A, Torgersen et al. [38]
for Cluster B, and Reichborn-Kjennerud et al. [39] for Cluster C. (i.e. irresponsible, blaming) using the Temperament
and Character Inventory (TCI) [2, 7]. The second
genetic determinant contributes most strongly to
antisocial and borderline personality disorder, and
most likely corresponds to high TCI Novelty Seeking

Table 26.3 Path coefficients for contributions to heritability of numbers of personality disorder symptom groupings measured by the
Structured Interview for DSM Personality (SIDP) from 3 latent genetic factors (GF) and unique contributions for each grouping of
symptoms in 2794 Norwegian twins.

Clusters of Path from GF1 Path from GF2 (? High Path from GF3 (? High Path from
personality (? Low self- novelty seeking/low harm avoidance/low unique or
disorder directedness) cooperativeness) reward dependence) unexplained
symptoms GF
Cluster A
Schizotypal 11 22 22 31
Paranoid 30 13 22 28
Schizoid 2 7 37 34
Cluster B
Antisocial 6 63 9 0
Histrionic 49 23 14 0
Narcissistic 35 12 7 33
Borderline 32 43 16 24
Cluster C
Avoidant 14 7 59 0
Dependent 29 15 23 37
Obsessive 29 6 13 41
Adapted from Kendler et al. [40].

319
 Chapter 26: Genetics of personality disorders

(i.e. impulsive, quick-tempered) and/or low TCI Table 26.4 Heritability and concordances in 236 monozygotic
(MZ) and 247 dizygotic (DZ) twin pairs for 18 basic scales of the
Cooperativeness (i.e. hostile, revengeful). The third Differential Assessment of Personality Pathology in Canada.
genetic determinant contributed most strongly to
schizoid and avoidant personality disorders, and most Scale label MZ DZ Heritability
likely corresponds to high TCI Harm Avoidance (i.e. correlation correlation (%)
anxious, shy) and/or low TCI Reward Dependence (r  100) (r  100)
(i.e. aloof, detached). Affective 49 12 45
In addition, most of the disorders have substantial lability
genetic determinants that are not explained by these Anxiousness 42 25 44
three latent variables, so the description of the per-
sonality disorders by a few genetic factors is far from Callousness 56 32 56
complete. The specification of personality disorders is Cognitive 48 31 49
better explained by comprehensive multidimensional distortion
personality inventories that measure normal and/or Compulsivity 40 19 37
abnormal personality traits [2, 7, 12].
Conduct 53 36 56
problems
Inheritance of general personality Identity 51 28 53
dimensions problems
Multidimensional inventories for assessing compon- Insecure 45 27 48
ents of personality disorders have been developed to attachment
measure both abnormal traits and also traits that char- Intimacy 47 24 48
acterize the full range of variability in the general problems
population. John Livesley and his colleagues developed
the Differential Assessment of Personality Pathology Narcissism 51 22 53
(DAPP) to measure the self-report of a hierarchy of Oppositionality 41 29 46
dimensions of personality problems. The questionnaire Rejection 33 19 35
is composed of 560 items measuring 18 factorially
derived dimensions, each with at least three specific Restricted 48 26 41
expression
facet scales [10, 32]. The heritability and concordances
in 236 MZ twin pairs and 247 DZ twin pairs is shown Self-harm 39 26 41
for all 18 DAPP traits in Table 26.4. The heritability Social 52 27 53
ranged from 35% for rejection sensitivity to 56% for avoidance
conduct problems and callousness, once again showing
Stimulus- 38 21 40
that personality traits are moderately heritable. seeking
Factor analysis of the 18 basic scales yields
four higher-order factors labeled “emotional lability Submissiveness 41 29 45
or dysregulation”, “antagonism or dissocial behavior”, Suspiciousness 42 29 45
“interpersonal responsiveness or inhibition”, and Adapted from Jang et al. [32].
“compulsivity” [10]. These higher-order dimensions
of abnormal personality resemble the dimensions
of normal personality. For example, emotional dys- TCI self-directedness and cooperativeness, or low
regulation is similar to high neuroticism in the five- DAPP emotional dysregulation and dissocial behavior,
factor model (measured by the NEO) [41] and low or low NEO neuroticism and agreeability) and person-
self-directedness in the seven-factor model of tempera- ality disorder at the other extreme (e.g. low TCI self-
ment and character (measured by TCI) [2]. Antagon- directedness and cooperativeness, etc.) [7, 10, 41].
ism or dissocial behavior is similar to low agreeability Twin studies of personality dimensions in the
in the five-factor model and low cooperativeness in the general population also demonstrate moderate herit-
seven-factor model. Hence these dimensions define ability of normal personality traits. For example, the
healthy personality at one extreme (namely, high TCI was developed based on specific neurobiological

320

Table 26.5 Total heritability of each of the 7 temperament and
character inventory personality dimensions estimated in 2517
twins in Australia. Unique effects exclude genetic contributions
shared with other personality dimensions.
Personality Total Unique
dimension heritability (%) heritability (%)
Harm avoidance 42 29
Novelty seeking 39 32
Reward 35 20
dependence
Persistence 30 23
Self-directedness 34 25
Cooperativeness 27 16
Self- 45 26
transcendence
Adapted from Gillespie et al. [47].
and psychosocial data [1, 2, 42–44]. It measures four
dimensions of temperament (describing behavioral
biases in response to basic emotional stimuli) and
three dimensions of character (describing higher cog-
nitive processes influencing the maturity of a person’s
goals and values). Initial twin studies were carried out
using only measures of temperament [45, 46]. More
recent twin studies have used the TCI and show that
each of the seven TCI dimensions has a unique gen-
etic variance that is not explained by the other dimen-
sions [47]. The heritability of each of the seven TCI
dimensions in a sample of 2517 Australian twins is
summarized in Table 26.5. Total heritability varied
from 27 to 45% without correcting for the reduced
reliability of the short form of the TCI used in this
study. Both temperament and character traits are
roughly equally heritable, and each dimension had
genetic determinants unique to it (that is, not overlap-
ping with the genetics of other dimensions). Thus
personality is moderately heritable whether measured
as normal traits, abnormal traits, number of symp-
toms, or categorical diagnoses. Alternative measure-
ment methods are highly convergent with one another.
Epigenetics of human personality
The estimates of heritability in twin studies are
inflated by contributions from gene–gene and gene–
environment interactions. Adoption and linkage
Chapter 26: Genetics of personality disorders
studies indicate that the narrow heritability of person-
ality is about 20–30% [36, 48], rather than the 40–50%
estimated by twin studies. The prominence of
gene–gene and gene–environment interaction for
personality traits confirms the expectation of non-
linear dynamic interactions from evolutionary theory
[19, 49].
The prominence of genetic complexity has several
practical consequences. Meta-analyses of candidate
genes may grossly underestimate the importance of
individual genetic polymorphisms when interactions
with other genes and/or environmental variables are
not measured. There is now substantial direct evi-
dence that personality development depends on the
nonadditive effects of gene–gene interactions [50–52],
as discussed in detail elsewhere [15]. For example, TCI
novelty seeking depends on the three-way interaction
of DRD4 with COMT and the serotonin transporter
locus promoter’s regulatory region (5HTTLPR). In
the absence of the short 5HTTLPR allele (5-HTTLPR
L/L genotype) and in the presence of the high activity
COMT Val/Val genotype, novelty seeking scores
are higher in the presence of the DRD4 seven-repeat
allele than in its absence [53]. Furthermore, within
families, siblings who shared identical genotype
groups for all three polymorphisms (COMT, DRD4,
and 5HTTLPR) had significantly correlated TCI nov-
elty seeking scores (r ¼ 0.4 in 49 subjects, p < 0.01).
In contrast siblings with dissimilar genotypes in at
least one polymorphism showed no significant correl-
ation for novelty seeking. Similar interactions were
observed between these polymorphisms and novelty
seeking in an independent sample of unrelated sub-
jects [53] and have been replicated by independent
investigators [51].
Likewise there is also substantial evidence for spe-
cific gene–environment interactions on personality
development [54–63]. For example, TCI novelty seek-
ing scores in adulthood are associated with particular
DRD4 polymorphisms only if the children were
reared in a hostile childhood environment with meas-
ures during childhood of maternal reports of emo-
tional distance and punitive discipline [61].
Assessments of childhood environment were based
on maternal reports and were made prior to the
independent assessment of adult personality.
Such complex nonlinear interactions present
great challenges for identification and replication of
relationships. Although the heritability of personality
traits is substantial, genome-wide association studies
321
 Chapter 26: Genetics of personality disorders

(GWAS) of personality traits have failed to discover
specific genetic variants [64]. Nearly all genetic vari-
ants contributing to variation in personality traits
remain “hidden” for all measures that have been
tested by GWAS, including measures developed by
Verweij et al. [64]. Resolution of the hidden genetic
variation may be impossible if investigators lack an
understanding of the functional systems underlying
personality development and vulnerability to psycho-
pathology. It may therefore be useful to develop a
testable model of the functional structure of brain
systems that is relevant for the development of per-
sonality and its underlying brain functions.
A model of the evolution of brain functions may
be the only way to make sense of the psychobiology of
personality and psychopathology. Progress in
molecular phylogenetics now allows the specification
of the ancestral line from the earliest eukaryotes to
human beings. Likewise progress in comparative
neuroanatomy and anthropology allows the specifica-
tion of a testable model of the emergent brain struc-
tures and functional capacities along the line of
ancestors leading to human beings [3].
An evolutionary approach to understanding
human functioning has many potential advantages
over traditional descriptive approaches, but much
work will need to be done by many people in many
fields to delineate and explore each of the components
of human functioning. A renewed focus on adaptive
functioning in evolution and development is likely to
be more fruitful than biologically blind testing of
candidate genes or polymorphic markers of uncertain
function. It should not be surprising that such an
evolutionary model is needed for us to begin to
understand psychiatric genetics in a way that corres-
ponds to natural functional systems. As Dobzhansky
said, “Nothing in biology makes sense except in the
light of evolution” [65].

References 9. Krueger RF. Arch Gen Psychiatry

1. Cloninger CR. Arch Gen Psychiatry
1987;44(6):573–588.
2. Cloninger CR, et al. Arch Gen
Psychiatry 1993;50(12):975–990.
3. Cloninger CR. Aust N Z
J Psychiatry 2009;43(11):
994–1006.
4. Cloninger CR. In O’Donohue WT,
et al. (eds.). Personality Disorders:
Toward the DSM-V. Los Angeles:
Sage Publications; 2007.
5. American Psychiatric Association.
Diagnostic and Statistical Manual of
Mental Disorders, Fourth Edition
(DSM-IV). Washington, DC:
American Psychiatric Association;
1994.
6. Eysenck HJ. In Millon T, et al.
(eds.). Contemporary Directions in
Psychopathology. New York:
Guilford; 1986.
7. Svrakic DM, et al. Arch Gen
Psychiatry 1993;50(12):
991–999.
8. Widiger TA, et al. Dimensional
Models of Personality Disorders:
Refining the Research Agenda
for DSM-V. Washington, DC:
American Psychiatric Association;
2006.
1999;56(10):921–926.
10. Livesley WJ, et al. Arch Gen
Psychiatry 1998;55:941–948.
11. Markon KE, et al. J Pers 2002;
70(5):661–693.
12. Grucza RA et al. J Pers Assess
2007;89(2):167–187.
13. World Health Organization. The
Constitution of the World
Health Organization.
Official Records. Geneva: WHO;
1946.
14. World Health Organization.
Mental Health: New
Understanding, New Hope.
Geneva: WHO; 2001.
15. Cloninger CR. Feeling Good: The
Science of Well-Being. New York:
Oxford University Press; 2004.
16. Cloninger CR. World Psychiatry
2006;5(2):71–76.
17. Wilkinson RG. The Impact of
Inequality: How to Make Sick
Societies Healthier. New York:
New Press; 2005.
18. Roysamb E, et al. J Pers Soc
Psychol 2003;85(6):1136–1146.
19. Wright S. Evolution and the
Genetics of Populations: Genetics
and Biometric Foundations.
Chicago: University of Chicago
Press; 1984.
20. Cloninger CR, et al. Br J
Psychiatry 1975;127: 11–22.
21. Schulsinger F. Int J Mental Health
1972;1:190–206.
22. Crowe RR. Arch Gen Psychiatry
1972;27:600–603.
23. Crowe RR. Arch Gen Psychiatry
1974;31:785–791.
24. Cloninger CR, et al. Arch Gen
Psychiatry 1982;39(11):1242–1247.
25. Mednick SA, et al. Science
1984;22:891–894.
26. Cloninger CR, et al. In Mednick
SA, et al. (eds.). The Causes of
Crime: New Biological Approaches.
Cambridge: Cambridge
University Press; 1987.
27. Rhee SH, et al. Psychol Bull
2002;128(3):490–529.
28. Cloninger CR, et al. Arch Gen
Psychiatry 1978;35(8):941–951.
29. Torgersen S, et al. Comp
Psychiatry 2000;41:416–425.
30. Eaves LJ, et al. Genes, Culture and
Personality: An Empirical
Approach. London: Academic
Press; 1989.

322

31. Loehlin JC. Genes and
Environment in Personality
Development. Newbury Park: Sage
Publications; 1992.
32. Jang KL, et al. Acta Psychiatr
Scand 1996;94:438–444.
33. Pedersen NL, et al. J Pers Soc
Psychol 1999;55:950–957.
34. Tellegen A, et al. J Pers Soc Psychol
1988;54:1031–1039.
35. Bergeman CS, et al. J Personal
1993;61:159–179.
36. Plomin R, et al. J Pers Soc Psychol
1998;75:211–218.
37. Kendler KS, et al. Psychol Med
2006;36:1583–1591.
38. Torgersen S, et al. Psychol Med
2008;38:1617–1625.
39. Reichborn-Kjennerud T, et al.
Psychol Med 2007;37:645–653.
40. Kendler KS, et al. Arch Gen
Psychiatry 2008;65(12):
1438–1446.
41. Costa PTJ, et al. J Pers Disord
1990;4:362–371.
42. Cloninger CR. Psychiatr Dev
1986;4(3):167–226.
Chapter 26: Genetics of personality disorders
43. Bohman M, et al. J Psychiatr Res
1987;21(4):447–452.
44. Cloninger CR. Psychol Assess
2008;20(3):292–299; discussion
300–304.
45. Heath AC, et al. J Pers Soc Psychol
1994;66(4):762–775.
46. Stallings MC, et al. J Pers Soc
Psychol 1996;70(1):127–140.
47. Gillespie NA, et al. Pers Individ
Diff 2003;35:1931–1946.
48. Cloninger CR, et al. Am J Med
Genet 1998;81(4):313–317.
49. Wright S. Annu Rev Genet
1982;16:1–19.
50. Benjamin J, et al. Molecular Genetics
and the Human Personality.
Washington, DC: American
Psychiatric Publishing; 2002.
51. Strobel A, et al. Mol Psychiatry
2003;8(4):371–372.
52. Van Gestel S, et al. Mol Psychiatry
2002;7(5):448–450.
53. Benjamin J, et al. Mol Psychiatry
2000;5(1):96–100.
54. Caspi A, et al. Nat Rev Neurosci
2006;7(7):583–590.
55. Caspi A, et al. Science 2003;301
(5631):386–389.
56. Hintsanen M, et al. J Psychosom
Res 2009;67(1):77–84.
57. Jokela M, et al. J Affect Disord
2007;100(1–3):191–197.
58. Keltikangas-Jarvinen L, et al.
Am J Med Genet B
Neuropsychiatr Genet 2009;
150B(3):389–394.
59. Keltikangas-Jarvinen L, et al.
Genes Brain Behav 2007;
6(4):305–313.
60. Keltikangas-Jarvinen L, et al.
Genes Brain Behav 2006;
5(1):11–18.
61. Keltikangas-Jarvinen L, et al.
Mol Psychiatry 2004;9(3):308–311.
62. Keltikangas-Jarvinen L, et al.
Scand J Psychol 2009;
50(6):574–582.
63. Kim-Cohen J, et al. Mol Psychiatry
2006;11(10):903–913.
64. Verweij KJH, et al. Biol Psychology
2010;85:306–317.
65. Dobzhansky T. American Biology
Teacher 1973;35:125–129.
323
 Ethical issues in behavioral genetics
Chapter
27 Stephen H. Dinwiddie, Jinger Hoop, and Elliot Gershon
Introduction
The principles of Mendelian genetics became widely
known around 1900; over the next 20 years the
theoretical basis of quantitative genetics was estab-
lished. Molecular investigation followed statistical
genetics as Watson and Crick characterized the
structure of DNA in 1953. By 1990 the Human
Genome Project had began. Now, two decades later,
the Personal Genome Project has made genetic and
health data on its first volunteers available and hopes
to eventually recruit 100 000 subjects. The tempo of
discovery and the power of technology will continue
to increase synergistically, and advances in genetics
will pose in concrete terms ethical questions which a
handful of years ago were only speculative and
theoretical.
It should be acknowledged that advances in genet-
ics have not so much raised new ethical questions as
simply refined and rephrased established ones. More-
over, few if any of the major ethical questions now
being raised have relevance solely to behavioral gen-
etics, while some, such as those touching on issues
such as gene therapy or genetic engineering of
humans, are unlikely to be relevant to considerations
of behavior and mental illness in the short term. On
the other hand, the role of genetics goes far beyond
the determination of physical characteristics or risk of
medical disease – it touches on how we must now
think about such fundamental concepts as personal
responsibility.
In this chapter, we will concentrate mainly on
three areas: The use (and misuse) of genetic infor-
mation for social ends; issues that arise at the inter-
face of medicine and other social institutions such as
the law; and issues of privacy and control over infor-
mation about one’s genotype.
Principles of Psychiatric Genetics, eds John I. Nurnberger, Jr. and Wade H. Berrettini. Published by Cambridge University
Press. © Cambridge University Press 2012.
324
The use (and misuse) of genetic
information for social ends
Eugenics
Artificial selection by controlled breeding can be
traced back to the beginnings of pastoralism and
agriculture, and thus pre-dates by millennia any sci-
entific understanding of the mechanisms of biological
inheritance. Behavioral characteristics cannot have
been ignored in this endeavor; presumably domestic
animals were selected very early on for docility at
least, and hunting dogs (to take but one example)
have long been bred for very specific traits of tem-
perament and behavior.
Family resemblance for many behaviors and attri-
butes among humans, such as personality or intellect,
has also been widely acknowledged, and while select-
ive mating to influence these was proposed at least as
far back as Plato’s Republic, philosophers have more
commonly emphasized alterations in environmental
factors to enhance the well-being of society. Popular
acceptance of the proposition that the social good
might be substantially improved by differential repro-
duction of selected citizens appears to be of much
more recent origin. The invention of the term “eugen-
ics” is attributed to Sir Francis Galton, who in 1883
wrote:
Whenever a low race is preserved under conditions of
life that exact a high level of efficiency, it must be
subjected to rigorous selection. The few best specimens
of that race can alone be allowed to become parents, and
not many of their descendants can be allowed to live. On
the other hand, if a higher race be substituted for the low
one, all this terrible misery disappears. The most
merciful form of what I ventured to call “eugenics”
would consist in watching for the indications of superior

strains or races, and in so favoring them that their
progeny shall outnumber and gradually replace that of
the old one . . . ([1], p. 199).
The historical context of this passage must be kept in
mind: It was written a quarter-century after Darwin’s
On the Origin of Species was published, and Galton (a
cousin of Darwin’s) was a Victorian Englishman par
excellence: a geographer, meteorologist, mathemat-
ician, and biologist who seemed to have had in
adequate measure that comfortable certainty of his
own superiority and that of his society so characteris-
tic of the age – an intellectual and cultural chauvinism
that, in tandem with what was later called “social
Darwinism”, was to shape the eugenics movement.
Galton’s scientific achievements lent considerable
weight to his views and he can therefore be properly
considered a founder of “scientific racism”.
The propositions that the human race could be
improved by consciously selecting for (or against)
certain behavioral characteristics and that many such
characteristics were biologically heritable fit well also
with the intellectual climate of the US Progressive era:
In an 8–1 decision by the US Supreme Court, in Buck
v. Bell [2] (challenging a Virginia statute that allowed
compulsory sterilization of the mentally retarded)
Justice Oliver Wendell Holmes, Jr. described the case
as follows:
Carrie Buck is a feeble-minded white woman who was
committed to the State Colony . . . She is the daughter of
a feeble-minded mother in the same institution, and the
mother of an illegitimate feeble-minded child . . . The
Commonwealth [of Virginia] is supporting in various
institutions many defective persons who if now
discharged would become a menace but if incapable of
procreating might be discharged with safety and become
self-supporting with benefit to themselves and to society;
and that experience has shown that heredity plays an
important part in the transmission of insanity,
imbecility, etc.
Holmes went on to write (no doubt thinking of his
own military experiences in the Civil War):
We have seen more than once that the public welfare
may call upon the best citizens for their lives. It would be
strange if it could not call upon those who already sap
the strength of the State for these lesser sacrifices, often
not felt to be such by those concerned, in order to
prevent our being swamped with incompetence. It is
better for all the world, if instead of waiting to execute
degenerate offspring for crime, or to let them starve for
their imbecility, society can prevent those who are
Chapter 27: Ethical issues in behavioral genetics
manifestly unfit from continuing their kind . . . Three
generations of imbeciles are enough.1
Prior to Buck v. Bell, sterilization laws in the United
States had been legally problematic but the decision
cleared the way for their expansion such that, for
example, the state of Oklahoma could mandate steril-
ization of “habitual criminals”, defined as individuals
convicted of repeated felonies involving “moral turpi-
tude” [4, 5]. The focus of eugenics had shifted from
encouraging the “fit” to reproduce to sterilizing the
“unfit” as attention moved from disease or physical
weakness to constructs such as “feeblemindedness”
and habitual criminality, with the implicit assumption
that all of these constructs were equally biological and
heritable. But, as with kindred “sexual psychopath”
laws that proliferated at about the same time, by
uncritically accepting scientific-appearing assertions
and using them as justification for policy, these stat-
utes served social aims: surely no reasonable person
could oppose governmental efforts to lighten the
burden on upstanding, tax-paying citizens who other-
wise would continue to pay ever-increasing amounts
for the support of unproductive individuals who were
at best doomed to spend their lives in misery and at
worst would imperil the public by their acts of
immorality or violence. The punitive subtext was
not as readily apparent, either in the United States
or other countries (including Canada and a number
in Europe) that adopted eugenic measures.
But it was in Nazi Germany that the process
reached its end-point. As in the United States and
elsewhere, eugenic arguments were used in Germany
first to justify involuntary sterilization of individuals
with a variety of mental or neurological illnesses
including schizophrenia, manic-depressive illness,
alcoholism, epilepsy, and hereditary blindness or
deafness, among other conditions; over time, those
deemed to be unworthy of life were not merely steril-
ized but executed, supposedly to promote the welfare
of society as a whole [6]. Soon, entire ethnic groups
were decreed to be inferior, indeed not fully human
and a threat to the Reich – and so deserving of death.
It had taken little more than half a century to go from
1
As it happened, conclusions about Carrie Buck’s
“imbecility” and that of her daughter were at best
overstated. There is also convincing evidence that the
pregnancy was the result of rape rather than “promiscuity”
and that the commitment was motivated by an effort to
protect the family’s reputation [3].
325
 Chapter 27: Ethical issues in behavioral genetics
encouraging reproduction of the “fit” to discouraging
and preventing reproduction by the “unfit” to whole-
sale murder.
What can be learned from these atrocities? The
history of the eugenics movement shows how easily
social value judgments can be taken for incontrovert-
ible scientific fact. It also demonstrates how easily the
rights of the defenseless and marginalized can be
violated – even by individuals many of whom would
(we suspect) honestly assert they had nothing but
altruistic motives. It is easy in retrospect to trace this
evolution – or devolution – and condemn it; but it is
far too simplistic to ascribe intellectual dishonesty
and malevolent motives to all those who supported
the eugenics movement: History amply shows that
even the most meritorious of ideas and programs
can be hijacked and used to support depraved and
corrupt ends. Rather, the tragedy stemmed from the
fact that from early on, constructs of variable and
often minimal validity were taken out of proper con-
text and uncritically used to justify the victimization
of powerless individuals, by so doing serving social
agendas ranging from dubious to overtly evil.
Issues at the interface of medicine
and the law
Genetics and responsibility
Personality traits appear substantially biologically
transmissible. Though estimates vary, it is likely that
roughly half of observed variance in normal person-
ality traits can be explained by genetic factors [7].
For specific psychiatric illnesses, additive genetic
factors (assuming no radical environmental changes,
of course), may account for as much as 80% of the
observed phenotypic variance in schizophrenia and
bipolar illness [8–10] and perhaps around 65% for
alcohol dependence and conduct disorder [11, 12],
though the latter two appear to have some common
genetic determinants [13].
Progress in identifying specific alleles that have
differential influence on normal personality variance
has to date lagged behind identification of loci
involved in moderating risk for psychiatric illness,
an endeavor that itself can as yet boast few successes.
But such alleles may certainly be identified, using
presently available methods. Is it too early to ask what
effect characterizing such influences might have on
how we think of individual choice and responsibility?
326
By virtue of training and orientation, mental
health professionals may tend to concentrate on the
role of explanatory, putatively causal factors in think-
ing about behavior, particularly that which seems to
arise from mental illness; both psychoanalytic theory
and biological psychiatry, still the two dominant para-
digms in clinical practice, tend to be philosophically
deterministic, though of course the causal factors
proposed are radically different. It would seem, at
least at first glance, that a better characterization of
the underlying biology would provide a whole new
range of explanations, and hence excuses at least for
undesirable or self-defeating behavior. (At least in this
formulation the father as well as the mother would be
blamed!)
But of course to do so would be to commit a
fundamental error. A distinction must be made
between finding an explanation for a given behavior
and assigning that behavior a moral weight: It is not
clear that a better understanding of the underlying
biology must necessarily impact that value judgment.
Regardless of the biology behind it, we subjectively
have a sense of our own “agency”, and we assign
moral responsibility based on that perception. Any
mechanistic account (regardless of the extent to which
the factors involved are heritable) must include that
experience. After all, as Dennet has noted:
What we want when we want free will is the power to
decide our courses of action, and to decide them wisely,
in the light of our expectations and desires. We want to
be in control of ourselves, and not under the control of
others. We want to be agents, capable of initiating, and
taking responsibility for, projects and deeds. All of this is
ours, I have tried to show, as a natural product of our
biological endowment, extended and enhanced by our
initiation into society ([14], p. 169).
Dennet also points out that the role of development in
becoming a moral agent cannot be ignored. The pro-
cess begins, after all, with an infant who has a unique
set of innate (potential) qualities but minimal self-
awareness or volition; the end result is the contingent
product of reciprocal interactions with the environ-
ment (a point made, after all, by the biometric mod-
eling that indicates that roughly half of personality
variance can not be attributed to heritable factors).
Over time, the actor more and more actively shapes the
surrounding environment as well as being shaped by it
(there is compelling evidence that gene expression can
be powerfully affected by social environment, for
example [15]). In short, there is a profound distance

between genotype and the final product of that devel-
opmental process; indeed, given the powerful inter-
play between environment and gene expression it may
be misleading to talk of the “final product” of a
dynamic process. This is consistent with a position
sometimes called “compatibilism” or “soft determin-
ism”, or perhaps “conditional free will”:
[S]ocial human behavior is contingent on a countless
number of possible decisions from among which the
individual may choose. Not all of those decisions are
feasible, however, nor are the resources available that are
required to act on them. Choosing a course of action,
therefore, is limited by preset boundaries . . . [which]
include current circumstances and opportunities,
learning experiences, physiological abilities, and genetic
predispositions ([16], pp. 30–31).
To observe that individuals’ behaviors are influenced
and constrained by a host of factors (some more and
some less accessible to reflection or responsive to
environmental factors) is hardly a revolutionary state-
ment. More importantly, to assume that once one has
an understanding (from some perspective) of the
“cause” of an action that the actor is relieved of
responsibility is simply to commit a category error
(why should knowledge of a causal process necessarily
impact moral judgment?). There is, moreover, no
logical reason that biological characteristics which
might make some choices for a given actor harder
should be privileged over well-known environmental
factors that similarly influence behavior. Perhaps a
better way of thinking of how to assign moral respon-
sibility for an action, determined or not, would be to
consider the extent to which: (a) the act follows a
rational as opposed to irrational deliberative process;
and (b) the extent to which the actor him- or herself
accepts and endorses the desires motivating the act
[17]. In this formulation, genetic factors are of no
greater or lesser import than any other causative
influences.
Implications for criminal responsibility
and punishment
Individuals are held accountable for their actions,
despite knowledge that some behavioral choices are
difficult, every day. The preeminent social mechanism
for assigning responsibility for behavior is the legal
system. We will concentrate on criminal law here,
though of course the issue might arise in a variety of
other contexts, for example in personal injury litiga-
tion. As a matter of social control, in keeping with his
Chapter 27: Ethical issues in behavioral genetics
views regarding the developmental process of respon-
sibility, Dennet has argued:
Any finite control system (such as a human brain) will
always be prone to making mistakes or arriving at
decisions that a more leisurely analysis would condemn;
it is an inevitable feature of human character, even
perfected to its limit. Original sin, naturalized. It is wise,
however, to adopt policies that minimize the bad effects of
these inevitable defects of character . . . [B]y somewhat
arbitrarily holding people responsible for their actions, and
making sure they realize that they will be held responsible,
we constrain the risk-taking in the design (and redesign)
of their characters within tolerable bounds ([14], p. 165).
It should be kept in mind that the legal system makes
no pretense at being scientific; ultimately it is simply a
social means of settling disputes short of bloodshed,
and it can be argued that the criminal law is at base a
means of ensuring public safety and order, backed by
the armed power of the state. (In that view, fairness or
at least the public perception of fairness might be seen
as no more than the easiest way to have citizens
acquiesce to legal restrictions on their behavior.)
The usual justifications for criminal sanction
include reformation of the wrongdoer, restraint (or
incapacitation) of further wrongdoing (for example
by imprisonment), retribution (“an eye for an eye”),
and deterrence of the actor or others by the example
of punishment. Though the relative emphasis on
these factors shifts over time, all have in common
the idea that the criminal is a moral agent, responsible
for his actions, and as a result, for most serious crimes
both the commission of a wrongful act (actus reus)
and the concurrent presence of intent (mens rea)
must be demonstrated. It is generally (though not
universally) accepted that severe mental illness may,
under some circumstances, negate or at least minim-
ize punishment (i.e. by undermining mens rea entirely
or via the insanity defense in the former case, or by
diminished capacity or the equivalent in the latter). It
would seem obvious that, to the extent inborn char-
acteristics are beyond the control of the individual,
they should be taken into account when holding indi-
viduals to account for their behavior, and certainly it
would be better (contra Dennet) if this process were
scientifically based and not “somewhat arbitrary”.
But tempting as this approach might be, as dis-
cussed above, it is problematic, and, indeed, an analo-
gous approach has been tried – and failed. In 1954 the
US Court of Appeals for the District of Columbia
federal circuit found, in Durham v. US, that a criminal
327
 Chapter 27: Ethical issues in behavioral genetics
defendant should not be held accountable if it could
be shown that the “unlawful act was the product of
mental disease or defect” [18], the hope then being
that mental health professionals could identify to a
sufficient degree of accuracy and reliability, at least in
selected cases, the psychological factors motivating
specific criminal acts and could therefore assist the
court in assigning the appropriate degree of criminal
responsibility and hence the appropriate severity of
punishment. In the event, there was sufficient dis-
agreement among the experts regarding diagnosis,
causation, and even the boundaries of “mental disease
or defect” that, despite legal holdings which attempted
to bring some degree of order to the process, this
approach was abandoned less than 20 years later [19].
Substitute “genetic factors causing unlawful activ-
ity” for “mental disease or defect”, and the same confu-
sion would likely eventuate (though the tautological
nature of this approach is more explicit in this formu-
lation). No matter what phraseology is used, what
might properly be included (or excluded) is a matter
of perspective. The characterization of genetic factors
associated with, for example, antisocial personality
disorder or perhaps elevated risk for violent behavior
would undoubtedly lead to defense attorneys arguing
for clemency or exculpation on this basis – as has
indeed been done sporadically at least since Clarence
Darrow’s defense of Leopold and Loeb in 1924 [20].
But given the many processes operating in the space
between genotype and behavior, many “causes” of
behavior could be adduced, operative at many levels.
Note that this concern has nothing to do with the
scientific foundation upon which a diagnostic con-
struct should be based; the concern is over the misap-
plication of such a construct. Surely the identification
of specific genetic factors that promote the develop-
ment of a given behavioral syndrome would count
toward the validation of that syndrome, and it has
long been established that some heritable factors do,
indeed, appear to be associated with an elevated pro-
pensity for crime and violence [21], though few such
factors have so far been identified. However, given
that “criminality” is a meaningless concept except in
the context of the society that defines specific acts as
unlawful, and the observation that at the clinical level
a number of psychiatric disorders appear to be asso-
ciated with elevated risk for criminal and/or violent
behavior [22], it would seem unlikely that heritable
factors will prove to map very precisely onto legal
concepts such as criminality. It is intriguing to note
328
that genetic variants that appear to be associated with
violence seem either to be quite rare in the general
population [23] or to result in violence only in the
context of adverse social circumstances [24]. The
variability in expression even in the case of very
specific mutations again underscores the complexity
of the interplay between genetic factors and specific
triggering environmental factors. Thus, while it is
conceivable that (in some environments) rare genetic
variants might be strongly associated with specific
criminal acts, it seems more likely that heritable
factors that explain a substantial portion of observed
criminal behavior will moderate risk in a rather non-
specific way, perhaps through decreasing intelligence
or heightening impulsivity.
Any exculpatory power that such findings might
have is likely also to vary with the degree of associated
risk: if a given genotype is associated with a markedly
elevated risk of violence compared to the population
base rate, but many or most of the individuals with
that genotype still do not act violently, it is not likely
to persuade the justice system very strongly. On
the other hand, if the genotype were associated with
violence in a large majority [17], particularly in any
condition whose expression is strongly influenced by
gene–environment interaction, there is the concern
that individuals with a “high risk” genotype might be
inappropriately labeled as violence-prone, to their harm.
It is not clear that the identification of such gen-
etic factors would necessarily be exculpatory in any
event. While the presence of a genotype associated
with low monoamine oxidase A (MAOA) has been
associated with markedly increased (in the setting of
childhood maltreatment) risk of violent behavior
[24], such behavior was far from universal in the
cohort studied. Clearly other factors (heritable, envir-
onmental, or very likely both) further moderate risk,
and there is nothing in the nature of genetic factors that
makes them inherently inaccessible to amelioration
after birth. Moreover, the presence of a statistical
association between a given genotype and an elevated
likelihood of expressing a class of behaviors (violence
in this example) says nothing of the subjective experi-
ence of choice and congruence with values in the
context of a specific antisocial act [25]. Thus, it would
be a tremendous oversimplification to equate pres-
ence of a given genotype with lack of moral responsi-
bility. To parse out the relative contribution of
specific genetic factors to a given antisocial act would
be to go far beyond the limits of science.

Indeed, it could be argued that, to the extent that
such potentially violent individuals see their acts as
consistent with their values and as appropriate means
to achieve their ends, they might reasonably be held to
be more, rather than less, culpable [17]. Regardless of
any biological roots of antisocial attitudes and behav-
iors, there is a general sense that individuals are given
multiple opportunities in life to observe and model
their behaviors along more acceptable lines: their
failure to do so is precisely what society feels is worthy
of condemnation.
Sentencing might conceivably become harsher
rather than more lenient. Certain groups of offenders
(e.g. those diagnosed with antisocial personality dis-
order or who have elevated scores on measures of
psychopathy [26]) are known to be less likely to learn
from experience and have higher risk for reoffense.
To the extent that any genetic findings overlap with
such clinical constructs, genetic results could be used
by the State to increase the penalty or deny parole for
a given offense, under the theory that such individuals
are more unlikely to respond to less harsh punish-
ment or that they present a greater danger to the
public and so should be incarcerated longer. If the
condition is relatively rare, any such test, even with
relatively high specificity and sensitivity, will have
quite a high “false positive” rate [27], thus risking an
unjust result.
Moreover, even if it were accepted that such indi-
viduals might not be morally culpable, it seems likely
that such individuals would be treated much the same
as insanity acquitees – who are on average deprived of
their liberty longer than those simply found guilty, by
virtue of being perceived by society as dangerous.
Legislatures might well craft statues allowing indefin-
ite incarceration of such individuals, much as many
US states now allow “sexually violent predators” (and
in England and Wales, individuals found to have
“dangerous and severe personality disorders”) to be
indefinitely civilly committed for purposes of “treat-
ment” – even if no clearly effective treatment exists
and (if there were) even though there is no obvious
way to determine whether it might be effective until
after the release of such individuals back into the
community.
On balance, it seems unlikely that genetic infor-
mation (at least if properly weighed) will contribute
much to the legal determination of blameworthiness
or its absence. Genetic factors are likely to be too far
removed from causation of behavior, and are likely to
Chapter 27: Ethical issues in behavioral genetics
be quite nonspecific in their effects. But as happened
with the eugenics movement, there is the risk that
such information might well be taken out of context
and misapplied to support specific social agendas.
Privacy and control of genetic
information
Beyond the offender
The use of DNA testing on biological samples has
proven to be of great value in identifying criminals;
also of great importance has been its use in excul-
pating individuals wrongly convicted. There are many
other uses of genetic information in forensic settings,
including elimination of suspects during the investi-
gative process, identification of human remains, and
establishment of paternity. In the United States, the
Combined DNA Index System (CODIS), based on
States’ databases, has information at 13 loci on more
than 5.5 million people [28]; in the United Kingdom,
the National Criminal Intelligence DNA database
(NDNAD) has information at 10 loci on more than
4 million – about 5% of the population [29]. In neither
system is acquisition and/or nor storage of genetic
information restricted to those convicted of a crime.
In the United Kingdom, hair and samples from
mouth swabs can be taken without consent if the
individual is arrested or detained in a police station
even if the sample is not relevant to the crime being
investigated. In both the US and UK databases, many
samples are from juveniles or those who have not
been accused or convicted of crimes.
In theory, genetic information from a crime scene
could result in a partial match to a profile already in
the database, thus raising the possibility that the
offender might be a close relative of that individual.
This has been done on occasion [30], though since the
pool of suspects can be quite large (there might be
numerous partial matches), in practice the utility of
this approach is limited.
Despite the value of these databases to criminal
investigators, their establishment has raised concern.
One concern is privacy: in the case of partial matches,
there is the possibility of uncovering family secrets
such as a child fathered outside of the primary rela-
tionship or the discovery of nonpaternity. Because
samples may be stored there remains the possibility
of expanding the genotypic information beyond –
possibly far beyond – the loci currently assessed, with
329
 Chapter 27: Ethical issues in behavioral genetics
unclear ramifications. There is also the issue of the
proper use of this information in the case of juveniles
or those not convicted perhaps not even of a crime.
Finally, despite legal limitations on the use and dis-
closure of this information, requests to obtain identi-
fying information about profiles have been granted
(at least in the case of the UK database) for unclear
reasons:
The first of the two approved operational requests
was made by police to check for “named individuals”,
but it is not clear what this might mean. Further
clarification was provided which explained that such
requests related to “seeking named suspects in a specific
inquiry at the police’s request”. This is still far from
explaining such a use of the NDNAD. If the police seek
a DNA match on the NDNAD and one is found, then
that individual’s name will be readily known by the
police. If the police sought a named individual’s DNA
profile for purposes other than making a match to one
found at a crime scene, this might signal a departure
from the purposes for which the NDNAD may be
lawfully used ([30], p. 83).
This may relate to another area of concern, the possi-
bility of “function creep” –using the information for
purposes other than criminal investigation, for
example the establishment of paternity so that child
support payments can be made. Other uses with sig-
nificant potential social benefit could easily be identi-
fied; but no matter how laudable the goals, such use
would be to go beyond the initial rationale for the
databank’s establishment.
Issues of consent
Beneficence (the principle that the physician should
act for the good of the patient) and nonmaleficence
(the principle that harmful acts should be avoided and
if harm is unavoidable it should be balanced against
potential good) are two widely accepted foundations
for medical ethics. Other foundational principles
include respect for autonomy and fairness or justice
[31]. Subsumed under these are the additional prin-
ciples of honesty, dignity, and confidentiality.
Autonomy is generally privileged; that is, assum-
ing sufficient decisional ability, it is generally believed
that an individual should be free to make his or her
own choices without external constraint, even if the
decision leads to an undesirable consequence (the
paradigmatic case is the Jehovah’s Witness who
refuses a blood transfusion, risking death but adher-
ing to his or her religious principles).
330
But autonomy can be undermined by limitations
in rational thinking (for example the presence of an
active psychotic illness) or by limited understanding
of the nature and consequences of the decision to be
made. A related principle is therefore that of
informed consent, which is generally held to require
three elements: voluntariness, possession of sufficient
information upon which to base a decision, and com-
petence – the ability to make decisions for oneself in
both the legal sense (i.e. an adult not under guardian-
ship) and in the clinical sense of having sufficient
mental ability to process information and communi-
cate a choice. This “decisional capacity”, in turn,
might exist only at a very basic level, demonstrated
for example by the ability to make a choice, whether
rational or not. A somewhat higher threshold would
be demonstration of the ability to comprehend the
facts of the situation; higher yet would be evidence
that the individual was able to meaningfully appreci-
ate the risks and likely consequences. Finally (a very
high standard) would be demonstration of an ability
to rationally manipulate the information and inte-
grate it with his or her individual values.
The issues of autonomy and informed consent
may arise in the context of testing for genetic risk. It
appears generally accepted that if genetic testing is
offered, it should be voluntary. The individual should
be given relevant information regarding the nature of
the information to be obtained and the risks of the
procedure, should have sufficient time to consider the
options, and should be free to withdraw from testing,
even after the fact. But it is less clear what level of
decisional capacity would be appropriate. Presumably
a more conservative standard would be superior in
cases where risks and benefits are less clear-cut.
How does this translate into practice? A current
example is the recent offering of a test that screens
for several single-nucleotide polymorphisms (SNPs)
associated with elevated risk for breast cancer [32].
As opposed to BRCA1 and 2, which are relatively
less common but associated with substantial risk of
cancer, in this test the SNPs identified appear to be
associated with relatively small increases in risk of
disease individually, and it is likely that the variants
so far identified constitute only a small subset of the
total [32]. Thus, negative results cannot be equated
with lower overall risk, while positive results are open
to misinterpretation as to the real degree of risk.
It might be assumed that the physician, by virtue of
training and access to information, would have a

better understanding of the probabilistic nature of
any such tests and their implications. Unfortunately,
it is not clear that this is the case [33] and physicians
are not immune to common cognitive errors [34].
One can easily imagine situations in which patients
request testing but neither the patient nor the phys-
ician fully understands the implications. Given the
complexity of the information and the probabilistic
nature of any testing results, how likely is it that the
patient will be able to meaningfully understand the
implications of the test? What level of understanding
on the part of the patient is acceptable, and how
should one assess it? Who should pay for the time
spent educating the patient? How would sufficient
competence to order, interpret, and explain the results
on the part of the physician be assured?
Testing and addictive disorders
The same difficulties regarding the interpretation of
positive and negative test results could easily arise in
the context of psychiatric illness. The use of such testing
might lead to a false sense of inevitability and fear of a
“worst-case” scenario, even though many mental ill-
nesses are eminently treatable. The situation could be
further muddled if testing were applied to relapsing
and remitting rather than chronic conditions. In fact,
particularly if test sensitivity were low, one might have
a test that missed many severe cases and identified
numerous “cases” who through much of their lives
needed little if any intervention [35].
This appears to be the case with addiction to (at
least) nicotine and alcohol, both of which appear to
have substantial heritable components [36, 37]. Both
are relapsing and remitting conditions; for alcohol-
ism, only about half those diagnosed are likely to be
drinking and having problems at any one time [38].
Its course is quite variable and treatment quite effect-
ive. The clinical utility of any genetic test (even if
accurate) is unclear since screening for early diagno-
sis, which consists of asking a few questions, is quick
and cheap. Moreover, health risks (cancer, heart dis-
ease, and the like) are due to repeated use regardless
of whether the individual merits a clinical diagnosis of
dependence, and in the case of tobacco and illicit
drugs presumably all patients should be counseled to
avoid initiation or continuation of use regardless of
genetic liability to addiction.
More likely, screening would be seen as most
useful for purposes of identifying “at-risk” individual
Chapter 27: Ethical issues in behavioral genetics
for employment or health insurance reasons. Would
it make sense to identify higher risk individuals and
adjust premiums accordingly? Perhaps not, since ini-
tiation of substance use is frequently at a young age.
Thus the individual paying the premiums (or the
company paying the premiums for the employee)
would be financially penalized, not for his or her
own actions or even her own risk status but strictly
based on genotype.
It would appear that in the United States, with the
signing of the Genetic Information Nondiscrimina-
tion Act (GINA) on May 21, 2008 (when it became
Public Law 110–233) these issues may have been
resolved. GINA prohibits denial of insurance cover-
age to healthy individuals or requiring payment of
increased insurance premiums by healthy individuals
based on genetic screening; it also prohibits use of
genetic information by employers when making deci-
sions on hiring, promotion, or termination.
For-profit testing for genetic susceptibility
to psychiatric illness
Commercial testing for susceptibility to severe mental
illness using molecular genetic methods has been
available since at least 2007, when a test for variants
in the GRK3 gene was offered directly to consumers
by an internet laboratory [39]. According to the com-
pany, positive test results indicate a doubling to trip-
ling of the risk of bipolar disorder as compared to the
general population – though of course this still
amounts to only a 2 or 3% chance of developing the
disorder.
Other companies have recently announced plans
to market similar tests, (e.g. for susceptibility to
schizophrenia, autism, and antidepressant-induced
suicidal thinking) [40, 41]. As of the time of this
writing, one company offers testing (at a cost of
$399 plus shipping and handling) for some 91 traits
and conditions – though it notes that of those, only 23
have been found to be “supported by multiple, large,
peer-reviewed studies” while the remaining 68 are
“research reports” with less strong support. The “dis-
eases, traits, and conditions” assessed range from eye
color, bitter taste sensation, and earwax type to resist-
ance to HIV/AIDS, celiac disease, and diabetes melli-
tus, while assessments in the “research report” range
from odor detection and freckling to alcohol depend-
ence, schizophrenia, and progressive supranuclear
palsy [42].
331
 Chapter 27: Ethical issues in behavioral genetics
In the United States at this time, there is little
government regulation of genetic testing, and test
providers are not required to provide information
supporting clinical validity [43–45]. There is valid
concern that commercial interests may exploit fears
of psychiatric patients and their families by offering
tests with limited benefit and significant potential
for harm. In the case of some Mendelian disorders
and hereditary cancers, potential harm may include
complications of preventive treatment as well as
psychological stress, family strain, and social stigma-
tization [46–51]. Fortunately, the risk associated
with testing for these disorders appears to be low,
at least if is accompanied by safeguards such as
informed consent, confidentiality protection, genetic
counselling, and pretest screening. However, it is
possible that the impact of testing for common psy-
chiatric disorders may be greater. Mental illnesses
more directly affect emotions, cognition, and behav-
ior – key aspects of an individual’s “personhood”.
Learning that one has a genetic susceptibility to such
disorders may threaten one’s sense of self and future
prospects in ways different from the threats posed to
those at risk for cancer or other nonpsychiatric med-
ical conditions [46]. In addition, those who seek
testing for psychiatric disorders because they have
affected relatives may (partially because of know-
ledge of that family history) have higher baseline
levels of anxiety and depression [52, 53], and this
pre-test distress has been associated with post-test
distress [51]. On the other hand, a person identified
by testing as at elevated risk might be more inclined
to seek monitoring (and preventive treatment if
justified by future research).
The social risks of susceptibility testing may be
greater for psychiatric disorders than for other condi-
tions, because psychiatric disorders are more stigma-
tized. It is not clear currently whether progress in
psychiatric genetics will improve or worsen stigma.
It may be that further evidence of the biological
nature of psychiatric illness will reduce negative social
attitudes (though the massive amount of evidence to
date seems not to have sufficed). On the other hand,
some reports seem to indicate that elucidation of the
genetic bases for major mental illnesses may make
these disorders appear more permanent and less
amenable to treatment [54–56]. Nonetheless, it is to
be hoped that better understanding of the causation
(including genetics) of psychiatric illness will lead to
better treatment and social acceptance of those
332
afflicted – if that understanding is properly and con-
sistently communicated to the public.
A duty to warn?
Clinicians are generally expected to maintain the
confidentiality of health information. However, this
duty is not absolute, and must in some situations be
violated if other ethical obligations are to be met.
Reporting is mandated for certain communicable
diseases or injuries, where presumably the harm
due to breach of confidentiality is balanced by the
benefit accrued by society by tracking and interven-
ing in identified cases. In psychiatry, it is now gen-
erally accepted also that confidential information
may be released if there is reason to believe that
doing so would avert significant danger (for example
if a patient disclosed an intention to harm an identi-
fied [or identifiable] victim [57]), and various courts
have held that under some circumstances the
treater’s duty can extend to such a third party (some-
times referred to as a “duty to warn” or “duty to
protect”).
Knowledge of an individual’s genotype gives one
some knowledge of the genotype of family
members. Even if only probabilistic and uncertain,
there is an evolving body of malpractice case law
that indicates that health professionals must under
some circumstances disclose such information
under a “duty to warn” theory. In one such case
[58], after developing medullary thyroid carcinoma,
the plaintiff sued the physicians who had treated
her mother for the same condition, alleging that
given the heritable nature of risk for the disorder
she should have been warned so that she could have
taken preventive measures. The Florida Supreme
Court held that the physicians did indeed have a
duty to inform a third party (the child of the
patient) under such circumstances.
Similar reasoning was used in a New Jersey case
[59], which also involved heritable risk factors for
neoplastic disease. It has been suggested that impos-
ition of this “duty to warn” in the presence of a
known risk might eventually lead to the creation of
an affirmative obligation for physicians to routinely
test for certain genetic conditions, even though the
consequences of providing such information are
unknown [60]. Ideally, it might prompt the (poten-
tially) affected individual to take steps to increase
monitoring or prevention efforts. However, the

concern has been voiced that knowledge of this risk
might prompt parents to overestimate its degree, with
unknown but potentially adverse effects on the
parent–child relationship [61], though the basis for
this concern and the likely magnitude of harm has
been disputed [62].
The impact of such cases on behavioral genetics is
unclear. For traits or disorders with substantial gene–
environment interaction it is likely that imposition
of a duty to warn is at best premature [60]. However,
for conditions such as certain forms of mental retard-
ation a similar duty has been established [63]. In a
recent Minnesota case, the plaintiff had a child with
fragile X syndrome. Testing for the condition was
considered, but not done. A second child who also
had the syndrome was subsequently born to the
plaintiff. In this case there was a legal finding
that the “duty” extended past the patient (the first
child) to the mother (the carrier of the condition)
on the theory that it was foreseeable that she would
have other children. The court did not address
whether a similar duty might be established to other
relatives.
But what of the situation of an individual who, for
reasons of personal privacy or even malicious reasons,
refuses permission for such sharing? This issue
appears to have been addressed in 1982 by the US
President’s Commission for the Study of Ethical
Problems in Medicine and Biomedical and Behavioral
Research [64], which concluded that confidentiality
could be breached if: (a) reasonable attempts to obtain
voluntary consent to disclose the information failed;
(b) there were a high probability of harm if the infor-
mation were withheld and disclosure could prevent
the harm; (c) the potential harm were serious; and (d)
disclosure was limited to the genetic information
necessary for diagnosis and treatment of the disease.
However, this “duty to warn” might conflict with the
Health Insurance Portability and Accountability Act
(HIPAA), which places broad restrictions on release
of medical information. Under current regulations
(Section 164.512[j] of the final HIPAA privacy rules),
one circumstance under which protected health infor-
mation may be released without the individual’s
authorization would be to “prevent or lessen a serious
and imminent threat to the health or safety of a
person or the public” [65]. Fortunately, with the kind
of information at issue here, it is difficult to see how
the harm to be prevented could be considered
“imminent”.
Chapter 27: Ethical issues in behavioral genetics
Testing in the workplace
While the potential for misuse of genetic information
has been recognized and opposed at least since the
Declaration of Bilbao in 1993 and addressed in UNES-
CO’s Universal Declaration on the Human Genome
and Human Rights (1997) and its International Dec-
laration on Human Genetic Data (1993), this has not
generally resulted in passage of statutory protections
internationally; in the United States, while workplace
discrimination on the basis of genetic risk has been
alleged many times, it does not appear that this has led
to significant litigation. The best-known (and to date
perhaps the only) such case involved an employer that
surreptitiously tested employees for a rare genetic vari-
ant associated with the development of carpal tunnel
syndrome and threatened to fire a worker who declined
testing. In this case, the Equal Employment Opportun-
ity Commission (EEOC) filed suit against the employer
(Burlington Northern Santa Fe Railroad) in 2001; the
case was quickly settled and testing suspended. Since
passage of GINA, it would appear that in the United
States genetic information cannot be used for purposes
of hiring, promotion, or retention. Nonetheless,
markers indicating elevated liability for various illnesses
would likely be of interest to employers for planning
purposes, to the extent that these illnesses may be
associated with absenteeism, decreased productivity,
or increased health service utilization. Given the high
prevalence of mental illness and its impact on worker
productivity, should tests for liability to common psy-
chiatric conditions become available, they would
undoubtedly be of great interest to employers.
The availability of genetic screening would raise
issues not only of privacy but perhaps of benefit to the
employee. How might (or should) a test that indicated
heightened risk for mood or anxiety disorders be used if
an employee is considering a high-stress position?
Under GINA, the company would not be able to bar
the employee from taking the position, but the
employee might factor that information into the deci-
sion as to whether or not to accept the position. Perhaps
such information might encourage him/her to make
lifestyle changes or investigate preventive therapies.
Conversely, if he or she does accept the position, should
the company then be liable if the employee becomes ill?
Might the company be expected to provide additional
support or training for stress management?
Under the Americans with Disabilities Act (ADA),
a “disability” is a physical or mental impairment that
333
 Chapter 27: Ethical issues in behavioral genetics

substantially limits one or more major life activity;
individuals can qualify as disabled based on having a
record of such impairment or being regarded as
having such impairment. The ADA prohibits (among
other things) job discrimination against qualified indi-
viduals who have disabilities as so defined. However,
the US Supreme Court has held that companies may
refuse to employ an individual with a pre-existing
condition in a position that might worsen that condi-
tion. In that case, Chevron U.S.A., Inc. v. Echazabal
[66], a worker who was discovered to have hepatitis
C was fired from a maintenance job at an oil refinery
on the basis of the employer’s concern that the liver
disease could be worsened by exposure to toxins in
the workplace. It appears, therefore, that a company
would be permitted to bar a vulnerable employee from
a position that would worsen (or perhaps precipitate?)
a health condition. This holding potentially would
conflict with GINA should a genetic test become
available that would indicate, for example, heightened
risk of psychiatric illness in the presence of job stress.
Conclusion
Many behavioral traits and psychiatric illnesses are
known to have a substantial heritable component.
Investigators are currently localizing and characteriz-
ing the genetic factors involved. We are confident that
this endeavor will lead to a greater understanding of
the neurobiology of psychiatric illness as well as
normal personality variation, and to radically new
and better ways of relieving the suffering caused by
these disorders.
However knowledge can be put to the service of
bad ends, and the history of the eugenics movement
demonstrates the danger of uncritical acceptance
of constructs that contain social value judgments.
There is nothing inherent in the study of behavioral
genetics that should alter our concepts of personal
responsibility or our respect for individual autonomy.
But this knowledge, if not carefully employed, can be
misused – to marginalize unpopular individuals or
groups as, perhaps, deserving of harsher punishment;
to discriminate on the job; or to increase profits by
offering tests of unclear benefit. These concerns are
perhaps more salient given the complex, nuanced,
and probabilistic nature of genetic information,
which is susceptible to oversimplification and errone-
ous application – or, as H. L. Mencken is quoted as
saying, “For every complex problem there is a solu-
tion that is simple, neat – and wrong”.

References 10. Cardno AG, et al. Arch Gen

1. Galton F. Inquiries into Human
Faculty and Its Development.
London: Macmillan and Co.;
1883.
2. Buck v. Bell, 274 US 200 (1927).
3. Lombardo PA. New York Law
Review 1985;60(1):50–62.
4. Skinner V. Oklahoma, Ex Rel
Williamson 316 US 535 (1942).
5. Nourse VF. In Reckless Hands:
Skinner v. Oklahoma and the Near
Triumph of American Eugenics.
New York: WW Norton and Co.;
2008.
6. Sofair AN, et al. Ann Intern Med
2000;132:312–319.
7. Van Gestel S, et al. Mol Psychiatry
2003;8:840–852.
8. Sullivan PF, et al. Arch Gen
Psychiatry 2003:60(12):1187–92.
9. Kendler KS, et al. Behav Genetics
1995;25:127–232.
Psychiatry 1999;56:162–168.
11. Heath A, et al. Psych Medicine
1997;27(6):1381–1396.
12. Slutske WS, et al. J Abn Psychology
1997;106(2):266–279.
13. Slutske WS, et al. J Abn Psychology
1998;107(3):363–374.
14. Dennet DC. Elbow Room.
Cambridge, MA: MIT Press;
1984.
15. Robinson GE, et al. Science
322:896–900.
16. Fishbein DH. Criminology
1990;28;27–72.
17. Wasserman D. J Law Med Ethics
2004;32;252–256.
18. Durham v. US 214 F.2d 862
(1954).
19. US v. Brawner 471 F.2d 969 (1972).
20. Summer DA. In Botkin JR, et al.
(eds.). Genetics and Criminality:
The Potential Misuse of Scientific
Information in Court.
Washington, DC: American
Psychological Association; 1999.
21. Dinwiddie SH. AAPL Bulletin
1994;22(3):327–342.
22. Guze SB: In Botkin JR, et al. (eds.).
Genetics and Criminality: The
Potential Misuse of Scientific
Information in Court.
Washington, DC: American
Psychological Association; 1999.
23. Brunner HG, et al. Science
1993;262:578–580.
24. Caspi A, et al. Science
2002;297:851–854.
25. Dinwiddie SH. AAPL Bulletin
1996;24(1):95–108.
26. Salekin RT, et al. Clin Psychol – Sci
Pr 1996;3(3):203–215.
27. Dinwiddie SH. In Fishbein DH
(ed.). The Science, Treatment, and
Prevention of Antisocial Behaviors.
Kingston, NJ: Civic Research
Institute, Inc.; 2000.

334

28. Information from http://www.fbi.
gov/hq/lab/html/codis1.htm,
accessed 11/6/08.
29. Graham EAM. Forensic Science,
Medicine and Pathology
2007;3:285–288.
30. Nuffield Council on Bioethics.
The Forensic Use of
Bioinformation: Ethical
Issues. London: Nuffield
Council; 2007.
31. Beauchamp T: In Bloch S, et al.
(eds.). Psychiatric Ethics, third
edition. New York: Oxford
University Press; 1999.
32. Couzin J. Science 2008;322:357.
33. Greendale K, et al. Am J Med Gen
2001;106:223–232.
34. Graber ML, et al. Arch Intern Med
2005;165(13):1493–1499.
35. Quaid KA, et al. Alcohol Clin Exp
Res 1996;20(8):1430–1437.
36. Munafo M, et al. Addict Biol
2001;6:109–117.
37. Enoch M-A, et al. Curr Psychiatry
Rep 2001;3(2):144–151.
38. Helzer JE, et al. In Robis LN, et al.
(eds.). Psychiatric Disorders in
America. New York: Free Press;
1991.
39. Psynomics. https://psynomics.
com/. Accessed April 5,
2012.
40. Couzin J. Science 2008;319
(5861):274–277.
41. SureGene, LLC. Available at:
http://www.suregene.net.
Accessed April 5, 2012.
Chapter 27: Ethical issues in behavioral genetics
42. 23AndMe, Inc. Available at
https://www.23andme.com/.
Accessed April 5, 2012.
43. Genetics and Public Policy Center.
Who regulates genetic tests?
Available at http://www.
dnapolicy.org/policy.issue.php?
Accessed April 5, 2012.
44. Javitt GH, et al. Oklahoma Law
Rev 2004;57(2):251–302.
45. Secretary’s Advisory Committee
on Genetic Testing. Enhancing the
Oversight of Genetic Tests.
Bethesda, MD: National Institutes
of Health; 2000.
46. Hoop JG. Harva Rev Psychiatry
2008;16(6):332–338.
47. Clayton EW. N Engl J Med
2003;349(6):562–569.
48. Committee on Energy and
Commerce. The Potential for
Discrimination in Health
Insurance Based on Predictive
Genetic Tests: Hearing Before the
Subcommittee on Commerce,
Trade and Consumer Protection of
the Committee on Energy and
Commerce. Washington, DC: US
Government Printing Office;
2001.
49. Anderlik MR, et al. Annu Rev
Genomics Hum Genet 2001;
2:401–433.
50. Lapham EV, et al. Science
1996;274(5287):621–624.
51. Lerman C, et al. J Consult Clin
Psychol 2002;70(3):784–797.
52. Burke L. Int Rev Psychiatry
2003;15(3):243–255.
53. Poobalan AS, et al. Br J Psychiatry
2007;191:378–386.
54. Phelan JC. Trends Neurosci 2002;
25(8):430–431.
55. Link BG, et al. Ann Rev Sociol
2001;27:363–385
56. Corrigan PW, et al. Schizophr Bull
2004;30(3):477–479.
57. American Psychiatric Association.
The Principles of Medical Ethics
With Annotations Especially
Applicable to Psychiatry.
Washington, DC: American
Psychiatric Association; 2008.
58. Pate v. Threlkel (1995), 661 So. 2d
278 (Fla.).
59. Safer v. Pack (1996), 677 A.2d 188
(App. Div. NJ).
60. Petrila J. Behav Sci Law 2001;
19:405–421.
61. Wertz DC, et al. JAMA 1994;
272:875–880.
62. Malpas PJ. J Med Ethics 2008;
34:275–278.
63. Molloy v. Meier (2004), 679 N.W.
2d 711, 714 (Minn.).
64. President’s Commission for the
Study of Ethical Problems in
Medicine and Biomedical and
Behavioral Research. Making
Health Care Decisions.
Washington, DC: US Government
Printing Office; 1982.
65. Department of Health and
Human Services. Federal Register
2000;65(250):82813–82817.
66. Chevron USA, Inc. v. Echazabal
(2002), 536 US Supreme Court 73.
335
 Genetics of Tourette syndrome
Chapter
28 and related disorders
Maria G. Motlagh, Thomas V. Fernandez, and James F. Leckman
Introduction
Tourette syndrome (TS) is a developmental neurop-
sychiatric disorder of childhood onset. TS is a
chronic, potentially disabling condition characterized
by the presence of both motor and vocal tics. Tics
are characterized by brief, stereotypical but non-
rhythmic movements and vocalizations. Common
tics include eye blinking, grimacing, jaw, neck,
shoulder or limb movements, sniffing, grunting,
chirping, or throat clearing. In the natural history
of TS, motor tics often begin between the ages of 3
and 8, several years before the appearance of vocal
tics. Tics typically follow a waxing and waning
course [1, 2]. Tic severity usually peaks early during
the second decade of life with many patients showing
a marked reduction in severity by the end of adoles-
cence [3, 4]. Only 20% or less of individuals with TS
continue to experience a moderate level of impair-
ment of global functioning by the age of 20 years [3].
However, tic disorders that persist into adulthood
can be associated with the most severe symptoms
including episodes of self-injurious motor tics (sec-
ondary to hitting or biting) or socially stigmatizing
coprolalic utterances (e.g. shouting obscenities or
racial slurs).
The characterization of tics as intermittent trains
of involuntary motor discharge is incomplete. Many
tics are under partial voluntary control, evidenced by
patients’ capacity to suppress them for brief periods
of time. A key feature of tics is that they are frequently
associated with antecedent sensory phenomena,
including either a general sense of inner tension or
focal “premonitory urges” or both. These urges are
often experienced as nearly irresistible, can be a major
source of impairment, and likely reflect an underlying
deficit in sensorimotor gating [5].
Principles of Psychiatric Genetics, eds John I. Nurnberger, Jr. and Wade H. Berrettini. Published by Cambridge University
Press. © Cambridge University Press 2012.
336
Over the past 25 years, TS has emerged as a model
developmental disorder at the interface of neurology
and psychiatry. The identification of abnormalities
involving the basal ganglia in neuropathological
[6, 7] and neuroimaging studies [8, 9], the possibility
of a post-infectious form of the disorder [10, 11] and
the growing appreciation of the impact of environ-
mental, neurobiological, and developmental factors in
disease expression have all contributed to making TS
a model disorder for understanding developmental
psychopathology more broadly [1, 12].
Three decades of research has led to widespread
agreement that genes play a central role in the eti-
ology of this disorder. TS and related conditions not
only aggregate within families but cluster in patterns
that suggest a high degree of heritability. Moreover,
high rates of comorbid obsessive–compulsive dis-
order (OCD) and attention-deficit hyperactivity dis-
order (ADHD) appear to be accounted for, at least in
part, by a common genetic diathesis. Yet perhaps the
most important conclusion from nearly 30 years of
work in the field is that the pathogenesis of TS
involves the reciprocal interaction of genes and envir-
onment. There has been increasing excitement and
anticipation as an international consortium and other
major laboratories have begun to collect the pheno-
typic and genetic data needed to take full advantage of
the accelerating advances in the genomic sciences,
from gene microarray studies, to genome-wide asso-
ciation studies, to the study of noncoding RNA and
gene copy number variation.
This review summarizes the data suggesting that
the liability to TS and related disorders is heritable. It
will then review nearly three decades of efforts to
identify the specific genes involved and chronicle the
shifting strategies. First, we will examine the evidence
for the general nature and extent of genetic risk in TS

and comorbidities. Next, we will briefly examine the
results of traditional parametric linkage analyses, both
in extended families and within genetic isolates that
likely share the same subset of vulnerability genes
by descent. Here we focus on the apparent role of
L-histidine decarboxylase (HDC) in a two-generation
family with nine affected individuals. Third, we con-
sider the results and future promise of nonparametric
and other “model-free” methods of gene identifica-
tion. Fourth, we consider the merits of cytogenetic
methods to search for rare genetic variants based on
chromosomal rearrangements, with particular atten-
tion to Slit and Trk-like family member 1 (SLITRK1).
Fifth, we summarize the investigations of various
candidate genes. Sixth, we review some of the putative
environmental risk factors which may contribute
toward the development of TS. Finally, we consider
the future prospects for TS genetic research and close
with a discussion of the clinical and research implica-
tions of the increasing understanding of the genetic
underpinnings of this syndrome.
TS is a heritable condition
Multiple lines of evidence indicate that TS is an
inherited disorder. The first and most important piece
of evidence is that TS and related tic disorders tend to
aggregate across multiple generations within families.
The risk of a first-degree relative of someone with TS
also having TS or a lesser variant is substantially
greater than the risk to the general population. For
instance, the frequency with which siblings are
affected has consistently been found to be on the
order of 5% for sisters and more than 10% for broth-
ers [12]. Judged against the best available epidemi-
ological data [13, 14], this represents at least a 10-fold
increase in risk for first-degree relatives compared to
the overall prevalence of the disorder.
The evidence for familial aggregation of TS, while
suggestive, does not itself confirm a role for genes in
the etiology of the disorder. There are a variety of
reasons why the syndrome might run in families,
including an infectious etiology or a clustering of
other nongenetic (e.g. environmental) risk factors.
Twin studies complement heritability data, providing
direct evidence concerning the relative contributions
of inherited versus environmental factors. Several
studies have shown that monozygotic (MZ) twins
are far more likely to also suffer from TS than dizy-
gotic (DZ) twins (who share on average just 50% of
Chapter 28: Genetics of Tourette syndrome
their DNA) [15–17]. MZ concordance rates have been
shown to be approximately 53–56% versus less than
10% in DZ twins [16]. When study methodology has
allowed for direct patient examination and included
the diagnosis of chronic tics (CT) in addition to TS,
MZ concordance has been shown to approximate
100% [17].
In addition to family and twin studies, segregation
analyses provide circumstantial evidence for the role
of genes in TS. These studies examine pedigrees with
affected individuals and compare the actual patterns
of transmission to hypothetical patterns that would be
expected under varying modes of genetic transmis-
sion. Using this methodology, several studies have
suggested that TS is transmitted in an autosomal
dominant fashion with partial penetrance [18–20].
Alternatively, Walkup et al. [21] found that TS was
most likely the result of a gene of major effect confer-
ring more than half of the overall risk for the dis-
order, with the remainder accounted for by genetic
background and environmental factors. Hasstedt et al.
[22] completed a segregation analysis incorporating
assortative mating in a single large pedigree contain-
ing 182 members. The analysis provided evidence of a
major locus with an intermediate inheritance pattern
with a significant assortative mating correlation.
However, when assortative mating was not included
in the model, intermediate inheritance was not
inferred. In addition, other studies have been unable
to demonstrate a convincing Mendelian pattern of
inheritance, even taking into account TS spectrum
phenotypes [23]. In sum, although a consistent pic-
ture has emerged from nearly three decades of
research which strongly suggest that genes play a
major role in etiology of TS, the exact mode of inher-
itance remains elusive. As believed for other neurop-
sychiatric disorders, TS, in most instances, will likely
be found to be the result of multiple interacting gene
alleles, each with relatively small contributions
compared to the genetic effects observed in Mendel-
ian disorders.
Evidence for genetic relationships between
TS and other psychiatric disorders
When Georges Gilles de la Tourette published his
first descriptions of TS more than a century ago
(1885), he made several prescient observations: first,
that tics and TS aggregate in families; and second,
337
 Chapter 28: Genetics of Tourette syndrome
that patients with the disorder also suffer from
obsessions and compulsions. In fact, clinicians
familiar with TS have consistently noted a wide
range of psychiatric comorbidity in their patients.
Indeed, simple and transient tics in the absence of
comorbid conditions are common and occur in at
least 5% of children [13]. In clinical and population-
based samples, TS alone is the exception rather than
the rule [22].
In family studies, as many as 30–60% of TS
probands will meet diagnostic criteria for OCD.
These rates compare to population frequencies for
this disorder between 2–3% [20, 24]. In addition,
when relatives of TS patients are examined, there is
a significant increase in the risk for OCD in add-
ition to the increased risk for tic disorders. How-
ever, such an increase in the risk for a disorder
among relatives does not necessarily imply an alter-
native expression of a single genetic diathesis. The
same finding could be explained by comorbidity
due to common pathophysiological (but not gen-
etic) pathways, or the clustering of environmental
risk factors. However, multiple studies have found
that relatives of patients with TS (but without OC
symptoms) have higher rates of OCD (without tics)
than would be found in unaffected families [17, 20,
24–26]. Conversely, family studies of OCD pro-
bands have consistently observed elevated rates of
tic disorders [26–28]. In addition, several clinical
case series have documented that individuals with a
tic-related form of OCD are more likely to report
obsessions of symmetry and exactness and a need
to redo activities to achieve a sense of completion
or a sense of things looking, feeling or sounding
“just right” [29, 30]. In sum, the high rates of OCD
in TS patients and their families, the patterns with
which obsessive–compulsive symptoms are
expressed in TS pedigrees, and the presence of a
distinctive natural history in tic-related OCD all
suggest that in some families a shared genetic
diathesis may represent itself as either TS, OCD,
or both disorders.
ADHD is frequently diagnosed in children with
TS, with a prevalence as high as 60–70% [6, 31]. This
co-occurrence of TS and ADHD can be associated
with disruptive behaviors such as aggression, explo-
sive behavior, low frustration tolerance, and noncom-
pliance. When comorbid ADHD is present, it is
frequently associated with academic difficulties, peer
rejection, and family conflict [32–35].
338
The issue of whether ADHD and TS may be
variable expressions of a shared genotype has been
controversial [36–38]. Although there is a substantial
body of data indicating that these disorders are trans-
mitted independently within families, the large
comorbidity rate leaves open the possibility that in
some instances TS and ADHD may share a common
genetic vulnerability. Indeed, a recent latent class
analysis (LCA) of 952 individuals from 222 TS fam-
ilies indicated that individuals with TS, OCD, and
ADHD had the most heritable form of the disorder
[39]. In addition to LCAs, there have been several
factor analytic studies that have sought to refine her-
itable phenotypes [40–43].
The co-occurrence of depression and anxiety
symptoms with TS also may reflect either shared
genetic vulnerability or the cumulative psychosocial
burden of having tics or other shared biological diath-
eses [31]. A co-occurrence with autism has also been
reported. Indeed, among autistic subjects the preva-
lence of TS has been reported to be 6.2%, about 10
times the prevalence in the general population
[44, 45]. Furthermore, a recent study of genomic copy
number variation (CNV) in TS showed a significant
amount of overlap among rare CNVs in TS and
autism spectrum disorders (ASD), but not schizo-
phrenia or intellectual disability [46].
Parametric linkage analysis
Once family and twin studies have demonstrated that
genetic factors are likely to play a role in the patho-
genesis of a disorder, there are several different means
toward identification of specific genetic loci involved.
Studies suggesting an autosomal dominant mode of
transmission made TS a clear candidate for paramet-
ric linkage analysis. However, with one exception,
parametric linkage studies over three decades have
failed to identify and confirm a specific genetic locus
involved in the etiology of this disorder.
L-histidine decarboxylase (HDC). The one excep-
tion is the report by Ercan-Sencicek et al. [47] that
described a two-generation family with nine affected
members with TS. The logarithm of odds (LOD)
score of 2.05 for a region on chromosome 15
approximated the maximum theoretical LOD score
given a model of dominant transmission. They iden-
tified a heterozygous G-to-A transition at nucleotide
position 951 in exon 9 of the HDC gene, resulting in
a W317X substitution, predicted to result in a
 Chapter 28: Genetics of Tourette syndrome

truncated protein lacking key segments of the active

domain. Studies of mRNA from patient cells indi-
Nonparametric linkage analysis
With a few exceptions, there is a growing recognition
cated that the mutation escaped nonsense-mediated
that TS is unlikely to be inherited in a predominantly
decay. The mutation was not found in 3000 control
Mendelian fashion. As a result, researchers have
chromosomes from northern and western Europe.
favored the use of nonparametric approaches. Such
In vitro studies in Escherichia coli indicated that the
investigations do not require the specification of a
mutant protein acted in a dominant-negative
hypothesis regarding the mode or character of inherit-
manner, resulting in loss of enzyme activity. Of the
ance. An initial affected sibling pair analysis by The
nine affected individuals, four also had OCD and
Tourette Syndrome International Consortium for
one also had Asperger syndrome. HDC is the rate-
Genetics [59] identified two regions of the genome
limiting enzyme in histamine (HA) biosynthesis sug-
with evidence suggestive of linkage, one on chromo-
gesting that histaminergic neurotransmission is
some 4q and another on chromosome 8p, that
involved in the pathobiology of TS in this family
achieved LOD scores of greater than 2. The same
and perhaps more generally. HA signaling in the
consortium more recently reported the results of the
central nervous system is mediated by four
largest TS genetic linkage study yet undertaken. The
G protein-coupled receptors, located both presynap-
sample included 238 nuclear families yielding 304
tically (predominantly H3 as well as H4) and post-
independent sibling pairs and 18 separate multigenera-
synaptically (H1–H3). Presynaptic HA receptors
tional families, totaling 2040 individuals [60]. Suggest-
regulate not only the release of HA, but also a variety
ive evidence of linkage was observed for a region on
of other neurotransmitters, including dopamine. Sev-
chromosome 2p in the analyses that included individ-
eral lines of evidence suggest that HA acts in a
uals with TS or CT disorder as affected. Of note,
counter-regulatory fashion, with increased HA
neither of the two earlier reported sites on chromo-
resulting in decreased DA signaling and vice versa
some 4q or 8p was confirmed despite the fact that the
[48, 49]. H2 and H3 receptors are enriched in the
same families were a subset of the larger study.
striatum and cortex, regions of the brain implicated
Regions on chromosome 17q have also provided
in TS [50], and studies of rodents with decreased
suggestive evidence for linkage [61, 62]. For example,
brain HA show increased sensitivity to stereotypies
an initial scan of chromosome 17 in two large pedi-
when administered DA agonists [51]. The finding of
grees by Paschou et al. [61] provided a nonparametric
a loss of function of HDC has the potential to lead to
LOD score of 2.41. Fine mapping with 17 additional
the development of animal models and eventually
microsatellite markers increased the LOD score to
the development of novel therapeutics for TS.
2.61. Genotyping data from 25 single nucleotide poly-
Other than this important finding, assuming auto-
morphisms (SNPs) within 3 candidate genes in the
somal dominant transmission and genetic homogen-
region of interest on 17q was obtained from the
eity, at least 90% of the remaining genome has been
original families plus additional independent families
excluded [21, 52–58]. In light of the substantial body
with one or two affected children. These data yielded
of evidence supporting a major genetic contribution
several SNPs and three-marker haplotypes signifi-
to TS, these initial linkage results suggest that genes
cantly associated with TS.
conferring the phenotype are positioned in the
remaining 10% of the genome or that the methods
by which investigators have sought to identify these Population isolates and special
loci have been flawed or that the view of TS as a
unitary genetic entity is erroneous. populations
Historically, the failure to discover a significant TS Well-powered genome-wide association studies
genetic locus using a traditional parametric linkage have proven to be invaluable in identifying vulner-
strategy has lead to a number of alternative ability genes for a number of complex traits, but no
approaches including nonparametric linkage analyses, such study has yet been undertaken for TS. Efforts
the examination of population isolates, cytogenetic toward such studies are underway and investigators
methods to search for rare genetic variants, and asso- have so far focused genome-wide association efforts
ciation studies, including candidate gene studies and on genetically isolated populations. Examining
family-based tests of association. population isolates restricts the number of potential

339
 Chapter 28: Genetics of Tourette syndrome
vulnerability genes, as linkage disequilibrium
should extend for much greater distances than in
outbred populations. For example, Simonic et al.
[63, 64] completed an early genome-wide associ-
ation study for TS in a sample from the Afrikaner
population of South Africa. This population was
founded in the seventeenth century with some
admixture from English, south Asian, and indigen-
ous populations. The authors found evidence of an
association of TS to markers on chromosomes 2p,
8q, and 11q. Subsequently, Merette et al. [65] geno-
typed the 24 markers that achieved a nominally
significant association in the Afrikaners in a large
kindred from an isolate of French Canadians in
Quebec and found suggestive evidence of linkage
on chromosome 11 (11q23). Several additional
genome-wide association studies of TS are under-
way in samples that are now available from several
isolated populations in Costa Rica, Antioquia
[Colombia], Quebec, and Ashkenazi Jewish
diaspora.
The contribution of recessive loci as a mechanism
of genetic vulnerability in TS with reduced penetrance
is one possibility that has been difficult to explore.
Homozygosity mapping has been successfully used to
detect recessive loci within populations with high
rates of consanguinity. Using this technique, even
quite small inbred families can be informative due
to autozygosity in which the two alleles at an auto-
somal locus are identical by descent (i.e. copies of a
single ancestral gene). Motlagh et al. [66] identified 12
consanguineous Iranian families with TS. Remark-
ably, these families presented with an unusual natural
history characterized by the early onset of vocal tics
and coprolalia and frequent comorbidity with OCD.
Genotyping the affected and unaffected members of
these pedigrees has the potential to identify rare reces-
sive contributions to this disorder.
Cytogenetic abnormalities
Cytogenetic techniques such as karyotyping, fluores-
cence in situ hybridization (FISH), and array com-
parative genomic hybridization (aCGH), have been
used to identify patients with chromosomal abnor-
malities (i.e. translocations, deletions, duplications).
Identifying genes at or near the disrupted chromo-
somal regions may flag potential susceptibility genes
that warrant further investigation. A number of cyto-
genetic abnormalities have been reported in TS
340
families including sites at 2p12, 3p21.3, 7q35–36,
8q21.4, 9pter, 13q31, and 18q22.3 [69–72].
Boghosian-Sell et al. [68] described a family where
TS was segregating with a balanced 7;18 translocation.
The breakpoint on chromosome 7 was mapped to
within chromosomal bands 7q22 and 7q31. Subse-
quently, Kroisel et al. [73] described a de novo dupli-
cation on chromosome 7 (7q22.1–31.1) observed in a
13-year-old boy with TS, moderate mental retard-
ation, and minor physical anomalies. Both break-
points were within or close to the breakpoint region
described by Boghosian-Sell et al. [68], suggesting that
a gene located at or near this region may be involved
in the pathogenesis of TS in some cases. Further
molecular analysis of this same proband by Petek
et al. [70] revealed that his de novo abnormality was
an inverted duplication which resulted in disruption
of the IMMP2L gene, a human homologue of the
yeast mitochondrial inner membrane peptidase sub-
unit 2. More recently, this research group screened 39
TS patients and 95 multiplex autistic disorder (AD)
families (due to the localization of IMMP2L in the
critical region for an AD candidate locus on chromo-
some 7q, AUTS1), for sequence and CNV in IMMP2L
[74]. No coding mutations were found in either TS or
AD patients and expression studies provided no evi-
dence of parental imprinting at this gene locus, sug-
gesting that IMMP2L may not be a common
etiological factor in either disorder.
State et al. [71] reported on a young man with
CT and OCD who was found to carry a paracentric
inversion involving chromosome 18q22. This team
mapped the telomeric end of the inversion to a
genomic location that was within 1 Mb of a trans-
location breakpoint previously described by Bogho-
sian-Sell et al. [68]. Although no genes were
structurally disrupted by this inversion, functional
studies of two transcripts in the region showed rep-
lication timing dysregulation. This finding suggested
the possibility that epigenetic influences could affect
gene expression and influence phenotype, and left
open the possibility that genes in this region may
serve as candidates in TS.
Verkerk and colleagues [72] reported a shared
complex rearrangement in a father with TS and his
affected children, disrupting the contactin-associated
protein 2 gene (CNTNAP2) on chromosome 7q35.
This gene encodes a membrane protein located at
nodes of Ranvier of axons that may be important
for the distribution of potassium channels, which

would affect signal conduction along myelinated
neurons. However, Belloso et al. [75] recently
described a familial balanced reciprocal translocation
t(7;15)(q35;q26.1) in phenotypically normal individ-
uals. In this family the 7q35 breakpoint disrupted
CNTNAP2, indicating that disruption of this gene
does not necessarily lead to TS.
Cuker et al. [69] reported a 14-year-old girl with
severe OCD and a CT disorder with a t(2;18)(p12;
q22) translocation. The patient’s chromosome 18
breakpoint localized to the same chromosomal band
as two previously reported rearrangements associated
with TS, OCD, and CT disorder [68, 71], and mapped
to a genomic position approximately 5 Mb from these
rearrangements. The clustering of these three break-
points within a relatively small genetic interval sug-
gests that 18q22 may be a promising region for
containing a gene or genes of etiological importance
in the development of the TS/OCD phenotypic
spectrum.
Slit and Trk-like family member 1 (SLITRK1).
Abelson et al. [67] identified and mapped a de novo
chromosome 13 inversion in a patient with TS. The
gene designated as Slit and Trk-like family member 1
(SLITRK1) was identified as a brain expressed candi-
date gene mapping approximately 350 kb from the
13q31 breakpoint. Mutation screening of 174 unre-
lated TS patients of European ancestry identified one
SLITRK1 identification of a truncating frame-shift
mutation in a second family affected with TS. In
addition, two patients were identified with a rare
variant (var321) in a highly conserved region of the
30 untranslated region (3’UTR) of this gene, corres-
ponding to a brain expressed micro-RNA (miRNA)
binding domain. miRNAs are short, 20–22 bases,
noncoding RNAs that typically suppress translation
and destabilize messenger RNAs that bear comple-
mentary target sequences. Many miRNAs are
expressed in a tissue-specific manner and may con-
tribute to the maintenance of cellular identity. In vitro
studies showed that both the frame-shift and the
miRNA binding site variants had functional potential
and were consistent with a loss-of-function mechan-
ism. Studies of both SLITRK1 and the miRNA pre-
dicted to bind in the variant-containing 30 region
showed expression in basal ganglia and deep layers
of cortex in both mouse and human.
As a result of the Abelson et al. [67] study,
SLITRK1 has emerged as a strong candidate gene for
rare cases of TS and related conditions. Although
Chapter 28: Genetics of Tourette syndrome
quite uncommon, functional mutations in this gene
have been identified in patients with TS, trichotillo-
mania, and ADHD [67, 76, 77]. SLITRK1 contains
one exon and encodes 696 amino acids. There are
six known members in the SLIT and TRK-like gene
family, SLITRK1 is unique in that it lacks tyrosine
phosphorylation sites in its short intracellular domain
[78]. SLITRK family proteins are characterized as
integral membrane proteins that have two leucine-
rich repeat (LRR) domains and a carboxy-terminal
domain that is partially similar to trk neurotrophin
receptor proteins [79]. SLITRK1 demonstrates a
developmentally regulated pattern of expression
within cortical and striatal structures, particularly
within the striosomal compartment and the so-called
direct pathway of the CTSC loop, characterized at the
molecular level in part by dopamine D1 receptors.
A number of studies [80–85] have sought to repli-
cate the findings reported in Abelson et al. [67] and
have identified var321 of SLITRK1 in a total of nine
TS families, with the variant segregating along with
affected status in five of these families. Concerns that
the original SLITRK1 findings may have been errone-
ous as a result of occult ethnic differences in cases
versus controls (population stratification) have not
been supported by recent evidence [76].
Candidate gene studies
Current theories of the pathogenesis of TS have
guided the selection of several candidate genes for
association testing in individuals with TS. These
candidate genes have included various dopamine
receptor genes (DRD2, DRD3, and DRD4 [86–89]),
the dopamine transporter (DAT [90]), catecol-O-
methyltransferase (COMT [89, 91]), three noradre-
nergic receptor genes (ADRA1C, ADRA2A, and
ADRA2C [92, 93]), dopamine b-hydrolyase (DBH
[90]), monoamine oxidase A (MAOA [87]), and a
few serotonergic genes including tryptophan hydox-
ylase 2 (TPH2), the serotonin receptor 3 (5-HT3),
and the serotonin transporter (5-HTTLR) [91, 94–
96]. Several additional candidate genes involved in
neuronal development, neuroendocrine, and
immunological function have also been studied with
negative results [90, 97–100]. Genetic variation at
any one of these loci is unlikely to be a major source
of vulnerability to TS, but in concert, certain alleles
could have cumulative effects and contribute to
phenotypic variability.
341
 Chapter 28: Genetics of Tourette syndrome
Environmental factors: perinatal
events, psychosocial stress, infection,
and immune response
A number of environmental factors have been impli-
cated in the pathogenesis of TS including psycho-
social stress, gestational and perinatal insults,
exposure to androgens, heat and fatigue, and post-
infectious autoimmune mechanisms. For example,
perinatal hypoxic/ischemic events appear to increase
the risk of developing TS [14, 101, 102], and one
recent retrospective study added prenatal maternal
smoking as a risk factor for TS [103]. Maternal stress
during pregnancy may also influence the individual’s
tic severity in adolescence [104].
Male sex is a risk factor for TS. While this could be
understood by genetic mechanisms, frequent male-to-
male transmissions within families appear to rule out
the presence of a predominant X-linked vulnerability
gene. The increased prevalence of TS in males has led
to the hypothesis that the presence of androgenic
steroids during critical periods in fetal development
may play a role in the later development of the illness
[105]. This notion is supported by the observation
that gender-related behaviors in children and adults
with TS are similar to such behaviors in children with
known elevated prenatal androgens, and these behav-
iors correlate with tic severity [106]. While these
effects may be due to androgenic steroids expressed
early in development, it is likely that there are sex-
specific patterns of gene expression in male versus
female brains that influence their differentiation and
function [107].
Patients with TS report higher levels of psycho-
social stress, and latent class modeling of prospective
longitudinal data indicate that antecedent stresses can
increase future tic and obsessive–compulsive symp-
tom severity [108].
Temperature dysregulation involving some
change in hypothalamic function has also been pro-
posed as a factor in the pathobiology of some individ-
uals with TS [109]. In a case series [110], an increase
in ambient temperature as well as core body tempera-
ture was associated with a transient increase in tics in
some patients. This increase in tics was correlated
with patients’ local sweat rate – via a dopamine medi-
ated pathway in the hypothalamus.
Speculation concerning a post-infectious auto-
immune etiology for TS and OCD dates from the late
342
1800s [111] and has recently become an intense and
controversial area of research [112]. It is well estab-
lished that group A beta hemolytic streptococci
(GABHS) can trigger immune-mediated disease in
genetically predisposed individuals. Rheumatic fever
is characterized by inflammatory lesions involving the
joints, heart, and/or central nervous system. The cen-
tral nervous system manifestations are referred to as
Sydenham’s chorea (SC). In addition to chorea, some
SC patients display motor and phonic tics as well as
OCD and ADHD symptoms, suggesting the possibil-
ity that at least in some instances these disorders share
a common etiology [113]. Case reports have also
implicated other infectious processes in TS etiology
including Lyme disease [114] as well as mycoplasma
pneumonia [115].
In 1998, Swedo and colleagues proposed that Pedi-
atric Autoimmune Neuropsychiatric Disorder Asso-
ciated with Streptococcal infection (PANDAS)
represents a distinct clinical entity and includes some
cases of TS and OCD. In PANDAS, it is postulated
that although GABHS is the initial autoimmunity-
inciting event, viruses, other bacteria, or even nonin-
fectious immunological responses are capable of trig-
gering subsequent symptom exacerbations via
molecular mimicry, such that antibodies directed
against GABHS attack cells in the brain because of a
similar structure [116].
Future directions for TS genetics
In this final section, we briefly consider a few of the
emerging areas of great promise in TS genetics,
including the availability of increasingly high-reso-
lution DNA microarrays, expanded use of genome-
wide association studies, the importance of noncod-
ing RNAs, gene copy number variation, and epigen-
etic programming.
Microarray studies and the future of genome-
wide association studies. Since the complete sequen-
cing of the human genome, it is now possible to
genotype millions of SNPs and monitor expression
levels of thousands of genes simultaneously using
DNA microarrays. Gene expression profiling of per-
ipheral blood from a small number of TS patients has
led to the preliminary identification of a subset of
cases in which altered expression of an interrelated
set of immune genes discriminated between TS and
age- and gender-matched controls that included indi-
viduals with a variety of other neuropsychiatric

disorders [117–121]. Many of these genes are associ-
ated with natural killer (NK) cell function. NK cells
are a type of cytotoxic lymphocytes that constitute a
major component of the innate immune system. If
replicated, further work needs to be done to assess the
heritability of these NK gene variants and expression
profiles within TS families. Longitudinal studies will
also be important to determine whether the expres-
sion of any of these genes is associated with fluctu-
ations in symptom severity.
Work using DNA microarrays and mRNA expres-
sion in key brain regions has begun, and in the future
this method will permit investigators to use hundreds
or perhaps thousands of susceptibility genes to assess
an individual’s genetic vulnerability and to explore
developmental change and continuity, comorbidity
with other disorders, gene–gene, and gene–
environment interactions [122].
Noncoding RNAs. Another recent discovery with
far-reaching implications for future genetic research
in TS is the importance of noncoding RNAs. The
discovery that one of the rare variants of SLITRK1
alters the binding region for a miRNA [67] has
focused the field’s attention on the potentially
important role of noncoding RNAs in the genesis
of TS and related disorders. Remarkably, recent stud-
ies suggest that tissue-specific miRNAs may function
at multiple hierarchical stages within gene regulatory
networks, from targeting hundreds of effector genes
to controlling the levels of key transcription factors
[123]. This multilevel regulation may permit individ-
ual miRNAs to profoundly affect the gene expression
program of differentiated cells. It is possible that
specific miRNAs could alter cellular identities.
Although unlikely, it is conceivable that variant miR-
NAs could affect the number of fast spiking
GABAergic interneurons in the striatum of individ-
uals who go on to develop TS [6]. Specifically, the
number of fast spiking GABAergic interneurons in
the striatum in an animal model of idiopathic dysto-
nia appears to decrease and then return to normal
levels over the course of development [124]. The
changes in the number of these interneurons in the
striatum closely parallel the severity of their move-
ment disorder. The molecular mechanisms under-
lying this change in interneuron number are
unknown but might include cellular reprogramming
mediated in part by miRNAs. It is also of interest to
note that the one area in the adult brain where
SLITRK1 continues to be expressed in adulthood is
Chapter 28: Genetics of Tourette syndrome
in the tonically active cholinergic interneurons that
regulate the activity of the fast spiking GABAergic
interneurons in the striatum [1, 125].
Copy number variation. The emergence of
microarray technologies that can detect sub-micro-
scopic structural variation revealed extensive copy
number variation (CNV) across the human genome
[126, 127] and provided opportunities for genome-
wide assessment of rare variation. Studies in schizo-
phrenia [128–134] and ASD [135–138] demonstrated
an over-representation of rare CNVs, particularly
genic de novo variants [131, 135, 136, 139], and high-
lighted molecular mechanisms that likely play a role
in these conditions. Fernandez et al. [46] conducted a
genome-wide analysis of rare (< 1% frequency)
CNVs in 460 individuals with TS, including 148
parent–child trios and 1131 matched controls. The
results support recent findings implicating histami-
nergic neurotransmission in the etiology or modula-
tion of tics and highlight the potential involvement of
GABAergic mechanisms as well. In addition, the
results reinforce the notion of shared genetic risks
among ASD and TS, and identify three novel, large,
rare, genic, de novo CNVs that are likely carrying risk
in the individuals in which they were identified, based
on their de novo status and high gene content relative
to controls.
Epigenetic programming. Epigenetics is another
emerging field of potential promise for understanding
the origins of TS. Epigenetic programming is a fun-
damental part of eukaryotic biology, involving the
modification of DNA and the chromatin proteins that
associate with it during key periods of development.
Frequently these modifications have enduring effects
on gene expression in key brain regions. In addition,
some epigenetic alterations can be passed from one
generation to the next. For example, there is now
compelling data for the presence of developmental
windows during which the genetically determined
microcircuitry of key limbic–hypothalamic–midbrain
structures are susceptible to early environmental
influences and that these influences powerfully shape
an individual’s responsivity to psychosocial stressors
and their capacity to parent the next generation [140].
These early environmental influences have been
shown to alter the pattern of methylation of the pro-
moter region of the glucocortoid receptor gene in the
hippocampus, stably altering the level of glucocortoid
receptor gene expression and hypothalamic-pituitary-
adrenal (HPA) responses to stress. This finding could
343
 Chapter 28: Genetics of Tourette syndrome

be relevant to the observation that high levels of
maternal stress during pregnancy are associated with
a more severe form of TS [104]. Cross-sectional and
longitudinal studies of TS and early-onset OCD have
consistently suggested that these disorders are sensi-
tive to psychosocial stress [108], and that TS patients
show a heightened stress response via the HPA axis
[141, 142] and higher levels of cerebrospinal fluid
(CSF) corticotropin releasing hormone [143]. These
findings are consistent with an epigenetic effect of the
sort reported by Meaney and colleagues. In addition
to the evidence summarized above regarding gene
expression changes of the glucocorticoid receptor
gene in the hippocampus, it also appears that the
expression of the estrogen receptor alpha gene in the
hypothalamus is under epigenetic control. If such
changes occur in two genes, it seems possible that
the expression of multiple other central nervous
system and immune genes relevant to TS can be
altered through epigenetic programming.
Conclusion
The last three decades of research have demonstrated
that TS is more genetically heterogeneous than ini-
tially thought. Important progress toward the under-
standing of genetic influences in TS has been made by
the combination of family and twin studies, segrega-
tion analyses, parametric and nonparametric linkage
analyses, and association studies, as well as the study
of rare genetic variants. The currently available data
suggest that, with rare exceptions, TS is typically a
multidimensional, polygenic disorder in which the
effects of individual genes are much smaller than
previously believed. This implies that in a majority
of cases, TS vulnerability is determined by the com-
bined effect of multiple genes interacting with specific
environmental risk factors, such as maternal stress
and maternal smoking during pregnancy, as well as
inflammatory processes and psychosocial stressors
affecting the child. We believe that there is reason
for considerable optimism, as much of the conven-
tional wisdom about psychiatric genetics needs to be
re-examined in light of recent discoveries such as the
importance of CNV and the effect of environment on
DNA methylation and gene expression. One may
anticipate that the fast pace of genetic discoveries will
continue and will increasingly affect research in all of
medicine and psychiatry, especially developmental
neuropsychiatry.
Acknowledgements
Portions of the research described in this review
were supported by grants from the National Insti-
tutes of Health (NIH): MH49351, MH30929, and
K05MH076273, and by the Tourette Syndrome
Association.

References 9. Bloch MH, et al. Neurology

1. Leckman JF. Lancet 2002;
360(9345):1577–1586.
2. Lin H, et al. J Am Acad Child
Adolesc Psychiatry 2002;
41(9):1070–1077.
3. Bloch MH, et al. Arch Pediatr
Adolesc Med 2006;160(1):65–69.
4. Leckman JF, et al. Pediatrics
1998;102(1 Pt 1):14–19.
5. Swerdlow NR, et al.
Psychopharmacology (Berl)
2006;186(2):246–254.
6. Kalanithi PS, et al. Proc Natl Acad
Sci U S A 2005;102(37):
13307–13312.
7. Kataoka Y, et al. J Comp Neurol
2010;518(3):277–91.
8. Peterson BS, et al. Arch Gen
Psychiatry 2003;60(4):415–424.
2005;65(8):1253–1258.
10. Mell LK, et al. Pediatrics 2005;
116(1):56–60.
11. Swedo SE, et al. Am J Psychiatry
1998;155(2):264–271.
12. Leckman JF, et al. J Child Neurol
2006;21(8):642–649.
13. Khalifa N, et al. Dev Med Child
Neurol 2003;45(5):315–319.
14. Khalifa N, et al. Acta Paediatr
2005;94(11):1608–1614.
15. Hyde TM, et al. Neurology 1992;
42(3 Pt 1):652–658.
16. Price RA, et al. Arch Gen
Psychiatry 1985;42(8):815–820.
17. Walkup JT, et al. Psychopharmacol
Bull 1988;24(3):375–379.
18. Baron M, et al. Am J Hum Genet
1981;33(5):767–775.
19. Curtis D, et al. Br J Psychiatry
1992;160:845–849.
20. Pauls DL, et al. N Engl J Med
1986;315(16):993–997.
21. Walkup JT, et al. Am J Hum Genet
1996;59(3):684–693.
22. Hasstedt SJ, et al. Am J Hum Genet
1995;57(3):682–689.
23. Seuchter SA, et al. Genet
Epidemiol 2000;18(1):33–47.
24. Hebebrand J, et al. J Psychiatr Res
1997;31(5):519–530.
25. Pauls DL, et al. Am J Hum Genet
1991;48(1):154–1630.
26. Rosario-Campos MC, et al. Am
J Med Genet B Neuropsychiatr
Genet 2005;136B(1):92–97.
27. Nestadt G, et al. Arch Gen
Psychiatry 2000;57(4):
358–363.

344

28. Pauls DL, et al. Am J Psychiatry
1995;152(1):76–84.
29. Kwak C, et al. Mov Disord 2003;
18(12):1530–1533.
30. Woods DW, et al. J Dev Behav
Pediatr 2005;26(6):397–403.
31. Coffey BJ, et al. J Am Acad Child
Adolesc Psychiatry 2000;
39(5):562–568.
32. Carter AS, et al. J Child Psychol
Psychiatry 2000;41(2):215–223.
33. Hoekstra PJ, et al. J Clin Psychiatry
2004;65(3):426–431.
34. Peterson BS, et al. J Am Acad
Child Adolesc Psychiatry 2001;40
(6):685–695.
35. Sukhodolsky DG, et al. J Am Acad
Child Adolesc Psychiatry 2003;
42(1):98–105.
36. Comings DE, et al. J Am Acad
Child Psychiatry 1984;
23(2):138–146.
37. Pauls DL, et al. J Am Acad Child
Adolesc Psychiatry 1993;
32(5):1044–1050.
38. Stewart SE, et al. J Am Acad Child
Adolesc Psychiatry 2006;
45(11):1354–1362.
39. Grados MA, et al. Biol Psychiatry
2008;64(3):219–225.
40. Alsobrook JP, II, et al. Am
J Psychiatry 2002;159(2):291–296.
41. Mathews CA, et al. Biol Psychiatry
2007;61(3):292–300.
42. Robertson MM, et al. Psychiatr
Genet 2007;17(3):143–152.
43. Cavanna AE, et al. J Neurol
Neurosurg Psychiatry 2011;82
(12):1320–1323.
44. Baron-Cohen S, et al. Psychol Med
1999;29(5):1151–1159.
45. Kadesjo B, et al. J Am Acad Child
Adolesc Psychiatry 2000;39
(5):548–555.
46. Fernandez TV, et al. Biol
Psychiatry 2012;71(5):392–402.
47. Ercan-Sencicek AG, et al. N Engl
J Med 2010;362(20):1901–1908.
Chapter 28: Genetics of Tourette syndrome
48. Ferrada C, et al.
Neuropharmacology
2008;55:190–197.
49. Munzar P, et al.
Neuropsychopharmacology
2004;29:705–717.
50. Haas HL, et al. Physiol Rev
2008;88:1183–1241.
51. Kubota Y, et al. J Neurochem
2002;83:837–8451.
52. Barr CL, et al. Am J Med Genet
1999;20:437–445.
53. Curtis D, et al. Psychiatr Genet
2004;14(2):83–87.
54. Heutink P, et al. Adv Neurol
1992;58:167–172.
55. McMahon WM, et al. Adv Neurol
1992;58:159–165.
56. Pakstis AJ, et al. Am J Hum Genet
1991;48(2):281–294.
57. Pauls DL, et al. J Am Acad Child
Adolesc Psychiatry 1990;
29(2):195–203.
58. Verkerk AJ, et al. Mol Psychiatry
2006;11(10):954–964.
59. Tourette Syndrome Association
International Consortium
for Genetics. Am J Hum
Genet 1999;65(5):
1428–1436.
60. Tourette Syndrome Association
International Consortium for
Genetics. Am J Hum Genet
2007;80(2):265–272.
61. Paschou P, et al. Am J Hum Genet
2004;75(4):545–560.
62. Zhang H, et al. Am J Hum Genet
2002;70(4):896–904.
63. Simonic I, et al. Am J Hum Genet
1998;63(3):839–846.
64. Simonic I, et al. Am J Med Genet
2001;105(2):163–167.
65. Merette C, et al. Am J Hum Genet
2000;67(4):1008–1013.
66. Motlagh MG et al. Mov Disord
2008;23(14):2079–83.
67. Abelson JF, et al. Science 2005;
310(5746):317–320.
68. Boghosian-Sell L, et al. Am J Hum
Genet 1996;59(5):999–1005.
69. Cuker A, et al. Am J Med Genet A
2004;130A(1):37–39.
70. Petek E, et al. Am J Hum Genet
2001;68(4):848–858.
71. State MW, et al. Proc Natl Acad
Sci U S A 2003;100(8):4684–4689.
72. Verkerk AJ, et al. Genomics
2003;82(1):1–9.
73. Kroisel PM, et al. Am J Med Genet
2001;101(3):259–261.
74. Petek E, et al. Mol Genet Genomics
2007;277(1):71–81.
75. Belloso JM, et al. Eur J Hum Genet
2007;15(6):711–713.
76. O’Roak B, et al. Mol Psychiatry
2010;15(5):447–450.
77. Zuchner S, et al. Mol Psychiatry
2006;11(10):887–889.
78. Aruga J, et al. Gene 2003;
315:87–94.
79. Aruga J, et al. Mol Cell Neurosci
2003;24(1):117–129.
80. Keen-Kim D, et al. Hum Mol
Genet 2006;15(22):3324–3328.
81. Deng H, et al. Acta Neurol Scand
2006;114(6):400–402.
82. Chou IC, et al. Pediatr Neurol
2007;37(6):404–406.
83. Scharf JM, et al. Neurology
2008;70(16 Pt 2):1495–1496.
84. Orth M, et al. Mov Disord 2007;
22(14):2090–2096.
85. Fabbrini G, et al. Mov Disord
2007;22(15):2229–2234.
86. Cruz C, et al. Neurosci Lett
1997;231(1):1–4.
87. Diaz-Anzaldua A, et al.
Mol Psychiatry 2004;9(3):
272–277.
88. Grice DE, et al. Am J Hum Genet
1996;59(3):644–652.
89. Tarnok Z, et al. Am J Med
Genet B Neuropsychiatr
Genet 2007;144B(7):
900–905.
345
 Chapter 28: Genetics of Tourette syndrome
90. Yoon DY, et al. Am J Med Genet
B Neuropsychiatr Genet
2007;144B(5):605–610.
91. Cavallini MC, et al. Psychiatry Res
2000;97(2–3):93–100.
92. Chou IC, et al. Psychiatr Genet
2007;17 (6):359.
93. Xu C, et al. Am J Med Genet
B Neuropsychiatr Genet
2003;119B(1):54–59.
94. Brett PM, et al. Am J Psychiatry
1995;152(3):437–440.
95. Mössner R, et al. Mol Psychiatry
2007;12(7):617–619.
96. Niesler B, et al. Psychiatr Genet
2005;15(4):303–304.
97. Comings DE, et al. Neuroreport
1999;10(7):1589–1592.
98. Kindler J, et al. Psychiatry Res
2008;15:235–9.
99. Laurin N, et al. Am J Med Genet
B Neuropsychiatr Genet
2009;150B(1):95–103.
100. Miranda DM, et al. Am J Med
Genet B Neuropsychiatr Genet
2008;147B(1):68–72.
101. Burd L, et al. J Perinat Med
1999;27(4):295–302.
102. Whitaker AH, et al. Arch Gen
Psychiatry 1997;54(9):847–856.
103. Mathews CA, et al. Am
J Psychiatry 2006;
163(6):1066–1073.
104. Leckman JF, et al. J Am Acad
Child Adolesc Psychiatry 1990;29
(2):220–226.
105. Peterson BS, et al. J Clin
Psychopharmacol 1998;
18(4):324–331.
106. Alexander GM, et al. Dev
Psychopathol 2004;
16(2):407–420.
346
107. Dewing P, et al. Curr Biol 2006;16
(4):415–420.
108. Lin H, et al. J Child Psychol
Psychiatry 2007;48(2):157–166.
109. Kessler AR. J Child Neurol
2002;17(10):738–744.
110. Scahill L, et al. Percept Mot Skills
2001;92(2):419–432.
111. Kushner HI. A Cursing Brain? The
Histories of Tourette Syndrome.
Cambridge, MA: Harvard
University Press; 1999.
112. Hoekstra PJ, et al. Cell Mol Life Sci
2004;61(7–8):886–898.
113. Maia AS, et al. Rev Hosp Clin
Fac Med Sao Paulo 1999;54(6):
213–221.
114. Riedel M, et al. Lancet 1998;351
(9100):418–419.
115. Muller N, et al. Psychiatry Res
2004;129(2):119–125.
116. Snider LA, et al. Mol Psychiatry
2004;9(10):900–907.
117. Liao IH, et al. Pharmacogenomics
2010;11(12):1733–41.
118. Lit L, et al. Am J Med Genet
B Neuropsychiatr Genet
2007;144B(7):958–963.
119. Tang Y, et al. Arch Neurol 2005;62
(2):210–215.
120. Tian Y, et al. Brain Res
2011;1381:228–236.
121. Tian Y, et al. Am J Med Genet
B Neuropsychiatr Genet
2011;156B(1):72–78.
122. Johnson MB, et al. Neuron
2009;62(4):494–509.
123. Makeyev EV, et al. Science
2008;319(5871):1789–1790.
124. Hamann M, et al. Brain Res
2007;1150:190–199.
125. Stillman A, et al. J Comp Neurol
2009;513(1):21–37.
126. Sebat J, et al. Science
2004;305:525–528.
127. Eichler EE. Nature Genetics
2006;38:9–11.
128. Stefansson H, et al. Nature
2008;455:232–236.
129. Walsh T, et al. Science 2008;320
(5875):539–543.
130. Wilson G, et al. Hum
Mol Genet 2006;15:
743–749.
131. Xu B, et al. Nat Genet
2008;40:880–885.
132. Mulle JG, et al. Am J Hum Genet
2008;87:229–236.
133. Consortium IS. Nature
2008;455:237–241.
134. McCarthy S, et al. Nat Genet
2009;41:1223–1227.
135. Marshall C, et al. Am J Hum Genet
2008;82:477–488.
136. Sebat J, et al. Science 2007;316
(5823):445–449.
137. Szatmari P, et al. Nat Genet
2007;39:319–328.
138. Glessner J, et al. Nature
2009;459:569–573.
139. Sanders SJ, et al. Neuron
2011;70:863–885.
140. Kaffman A, et al. J Child
Psychol Psychiatry 2007;48
(3–4):224–244.
141. Chappell P, et al. Biol Psychiatry
1994;36(1):35–43.
142. Corbett BA, et al.
Psychoneuroendocrinology
2008;33(6):810–820.
143. Chappell P, et al. Biol Psychiatry
1996;39(9):776–783.
 Endophenotypes in psychiatric genetics
Chapter
29 Andrew C. Chen, Madhavi Rangaswamy, and Bernice Porjesz
It has been more than a half century since Nobel
laureates Drs. Watson, Crick, and Wilkin discovered
the structure of DNA. Since that time, rapid advance-
ment of molecular biological technology has facili-
tated progress in the field of genetics, with the
complete sequencing of the human genome by the
Human Genome Project accomplished in 2003. It is
now possible, not only to identify single gene Men-
delian disorders, but also to understand “complex”
disorders that are the result of multiple small gene
effects that predispose individuals to these disorders,
as they interact with environmental factors. As we try
to envision the future direction of human genetics
research, particularly the genetics of complex genetic
diseases such as the majority of psychiatric diseases, it
would be helpful to briefly review some important
milestones for the development of genetics in terms
of phenotypes before the era of molecular biology.
A few years after the term “genetics” was first
introduced by William Bateson in 1902, the Danish
botanist Wilhelm Johanssen attempted to provide the
clarifying distinction we now take for granted
between the concepts of “genotype” and “phenotype”.
He also introduced the word “gene” in 1909, a half
century before the discovery of the structure of the
genetic molecule, DNA. Through his research on self-
fertilized lines of beans, Johanssen found that the
phenotype is often an imperfect indicator of the geno-
type: the same genotype may give rise to a wide range
of phenotypes. Similarly, the same phenotype may
have arisen from different genotypes. This important
observation was contradictory to the genetic theory at
that time, which endorsed the idea that there was a
one-to-one correlation between genotype and its
“consequential” phenotype. About the same time,
Herman Nilsson-Ehle, a Swedish expert in plant
breeding and genetics, provided evidence for genetic
Principles of Psychiatric Genetics, eds John I. Nurnberger, Jr. and Wade H. Berrettini. Published by Cambridge University
Press. © Cambridge University Press 2012.
and nongenetic contributions to a continuous pheno-
type on the basis of observations of seed colors in
crosses of oats and wheat [1]. As we now know, a
genotype is the composition of DNA sequences; in
contrast, a phenotype represents observable charac-
teristics of an organism, and is the joint product of
both genotypic and environmental influences. In gen-
etic diseases that are transmitted through the classic
or Mendelian model, genotypes are usually indicative
of phenotypes in terms of the presence or absence of
the disease. However, this certain correlation between
genotypes and phenotypes does not exist in complex
genetic diseases.
Recently, there has been a surge of interest in the
use of endophenotypes in psychiatric research,
although the concept was first introduced to the field
of psychiatry by Gottesman and Shields more than
three decades ago [2]. This has been driven by con-
cerns about the limited success and relatively poor
reproducibility of the current psychiatric genetic
research approach, and the fact that current psychi-
atric diagnostic systems, DSM or ICD, are primarily
based on phenomenology and lacking of justification
by etiology. Psychiatry’s classification systems
describe heterogeneous disorders [3]. The brain is
the most complex of all organs. It is subject to com-
plex interactions not only among genes, proteins,
cells, and circuits of cells, but also across individuals
and their changing experiences [4]. As such, the
phenotypic output from the brain, i.e. behavior, is
not simply a sum of all its parts. Therefore, it would
be a more reasonable approach to apply more opti-
mally reduced measures of neuropsychiatric function-
ing than a behavioral complex, the diagnosis, in
studies of the biological and genetic components of
psychiatric disorders. It has been suggested that
ideally, molecular genetic studies should not be
347
 Chapter 29: Endophenotypes in psychiatric genetics
performed on psychiatric diagnoses alone, which
reflect distal and variable effects of genes, but on
quantitative neurobiological measures or markers
that reflect more proximal effects of genes involved
in the genetic predisposition for developing psychi-
atric disorders [5].
The endophenotype concept
An endophenotype is typically an unobserved pheno-
type, such as metabolite level, cell or organ activity,
biosignal, or other “biomarker”, that is thought to
contribute to the etiology of a visible phenotype or
disease susceptibility. It is associated with a disease
but can be measured independently of disease status.
Gottesman and Gould have defined endophenotype
as “measurable components unseen by the unaided
eye along the pathway between disease and proximal
genotype” [6]. In their writings summarizing genetic
theories in schizophrenia 30 years ago, Gottesman
and Shields described “endophenotypes” as internal
phenotypes that lie on the pathway between genes and
disease, and can be observed only by a “biochemical
test or microscopic examination” [2]. These quantita-
tive biological markers serve as covariates that correl-
ate with the main trait of interest (diagnosis) and
serve to better define that trait or its underlying
genetic mechanism. The term “endophenotypes” was
adapted from a 1966 paper explaining concepts in
evolution and insect biology by John and Lewis [7];
they wrote that the geographical distribution of grass-
hoppers was a function of some feature not apparent
in their “exophenotypes” that are obvious and exter-
nal. In contrast, these features were “microscopic and
internal”, hence the term, “endophenotype”. Thus, the
term endophenotype appears to fit the needs of psy-
chiatric genetics, and this concept bridges the gap
between the gene and the complexity of psychiatric
disease processes. Typically, endophenotypes are
quantitative traits that can be measured in a general
population. That is, the traits are present in individ-
uals who have manifested the symptoms of the disease
as well as in those who do not have the disease,
including those at risk prior to the onset of the
disease.
Generally speaking, the utility of the endopheno-
types is relevant for two types of studies. First, as
originally proposed [2, 6], they are used to aid in the
discovery of novel genes that are associated with the
disorders. Second, an increasingly popular method is
348
to apply endophenotypes to study the functional con-
sequences of risk alleles.
Characteristics useful for the
identification of endophenotypes
in psychiatric genetics
There have been a number of attempts to devise
criteria to define the optimal characteristics of an
endophenotype [6, 8–10]. As the diagnostic categor-
ies of psychiatric disorders in use today were initially
formulated in the late nineteenth and early twentieth
centuries by a small number of psychiatrists
who relied on perceived similarities in behavioral
syndromes and clinical outcomes, we should be
aware that such categories reflected only observable
behaviors rather than dysfunctions in distinct ana-
tomical substrates or patho-physiological processes.
Any given disorders with singular labels, such as
schizophrenia or attention-deficit hyperactivity dis-
order (ADHD), are probably better thought of as a
heterogeneous group of dysfunctions whose final
pathways eventually lead to similarity in symptoms,
course of illness, and other clinical features. For
endophenotypes to be useful in psychiatric genetic
analyses, they should have several properties [6, 11].
We summarize the six properties below. The first
two are most necessary for any endophenotype to
be useful in medical genetics. The last four, although
not strictly necessary, are properties that can aid
in the successful genetic dissection of complex
behaviors.
The endophenotype is heritable
Complex traits or phenotypes are the outcome of an
interaction between genetic and environmental
factors. One of the most important criteria defining
an endophenotype is that it be a heritable trait. For
endophenotypes to aid in the genetic dissection of
complex traits, there must be some evidence of herit-
ability of the phenotype. It is not very useful to link a
trait to genes if genetic factors do not contribute to
the individual differences in the trait, i.e. when it is
largely influenced by environmental factors. How-
ever, it should be kept in mind that using endophe-
notypes of relatively low heritability may still be
important, particularly when there is a potential for
gene
environment interaction. As environmental
effects may be crucial in altering the effects of even

highly heritable disorders, these endophenotypes can
also aid in the development of pharmacological or
behavioral treatments and interventions.
The endophenotype is associated
with the disorder
There should be reasonable evidence that the endo-
phenotype is associated with the pathophysiology of
the illness. Ideally, it should be a part of the causal
pathway from gene to disorder rather than effects
(sequelae) of disorders or their treatment. This (to
be associated with the causes of the disorders) is
important when the endophenotype is being used to
highlight the risk genes, and relatively less important
when the endophenotype is applied to study the func-
tional consequences of known risk alleles.
The endophenotype is present in both
affected and nonaffected individuals, and it
varies continuously in the general population
Although the diagnosis of mental disorders is dichot-
omous, i.e. whether a certain disorder is present or
not, it is unlikely that common mental disorders vary
in a discrete manner between individuals. Statistical
analyses of continuous traits that are artificially cat-
egorized into discrete groups have substantially less
power than analyses conducted on the original con-
tinuous scale. Therefore, it would be more powerful
to analyze endophenotypes that can be measured on
continuous scales. This also highlights another
important advantage of analyzing endophenotypes
that vary continuously in the nonaffected population
because it can simplify the sampling process and
reduce ascertainment biases in population-based
studies [12].
The endophenotype is primarily
state-independent
As mentioned, to be useful for genetic analysis, the
endophenotype should be present in both affected
and unaffected individuals. To aid in the discovery
of novel genes that are associated with the disorders,
the endophenotype should also manifest in an indi-
vidual at a similar quantitative level regardless of
whether or not the illness is active. It should
demonstrate adequate test–retest stability and
reliability.
Chapter 29: Endophenotypes in psychiatric genetics
Within families, endophenotype
and illness co-segregate
Affected members of families manifest the endophe-
notypic marker significantly more than unaffected
members of the families.
The endophenotype found in affected family
members is found in nonaffected family
members at a higher rate, or at a higher level,
than in the general population
This will allow small to moderate effect sizes between
biological relatives of affected patients and commu-
nity controls to be observed more easily [13].
Application of endophenotypes
in psychiatric genetics
A fundamental assumption for the application of
endophenotype in psychiatric genetic research is that
the number of genes required to produce variations in
these traits is fewer than those involved in producing
a psychiatric diagnostic entity, which may reflect a
more distal end-point of the combined contributions
from multiple genetic and environmental factors, as
well as the interactions among them. Quantitative
neurobiological measures of risk as endophenotypes
are closer to gene action involved in the predispos-
ition for the disorder. Therefore, in searching for
genetic variations that are involved in complex
(non-Mendelian) psychiatric diseases, endopheno-
types provide a more powerful strategy than the
dichotomous diagnosis of the disease [6, 14, 15]. In
psychiatry, a number of attempts have been made to
develop and determine the feasibility of candidate
endophenotypes. Although few candidates have
met all the criteria listed earlier, some linkage and
association studies using the endophenotype strategy
have had moderate success.
This chapter is organized into two main sections.
In the first part of this chapter, we summarize repre-
sentative published studies that have successfully used
various endophenotypes to find candidate genes
involved in several psychiatric disorders, including
schizophrenia, mood disorders, Alzheimer’s disease
(AD), ADHD, and suicidal behavior. The second half
of this chapter focuses on alcoholism, and more fully
illustrates the successful use of the endophenotype
349
 Chapter 29: Endophenotypes in psychiatric genetics
strategy in the Collaborative Study on the Genetics of
Alcoholism (COGA) project, where we have used
brain oscillations as endophenotypes in the identifi-
cation and understanding of genes involved in alco-
holism and related disinhibitory disorders.
Sensory motor gating as endophenotypes
for schizophrenia
Deficits in sensory motor gating are consistent neuro-
psychological findings in individuals suffering from
schizophrenia [16]. The hypothesized association has
face validity primarily on the basis of patients’ reports
that they may have difficulty filtering information
from multiple sources that occur in everyday life
[16–18]. Neurophysiological tests, including assess-
ments of P50 suppression and pre-pulse inhibition
of the startle response, have been developed to discern
efficiencies in these capabilities.
In tests of pre-pulse inhibition, startling sensory
stimuli (e.g. loud noise, bright light) are used to elicit
an unconditional reflexive startle response in individ-
uals. If a weaker pre-stimulus is provided before the
startling stimulus, the subsequent startle response is
generally diminished. A relatively reproducible find-
ing is that this diminution of the second response is
attenuated in patients with schizophrenia [16, 17]. Of
note, abnormal pre-pulse inhibition is not specific to
schizophrenia; studies have identified this abnormal-
ity in obsessive–compulsive disorder [19] and Hun-
tington’s disease [20].
The P50 suppression test uses two auditory stim-
uli presented at 500 ms intervals. Event-related poten-
tials (ERPs) for both stimuli are measured by
electroencephalogram (EEG). In normal individuals,
the ERP to the second stimulus is of lower amplitude
than the first. However, suppression of P50 amplitude
is compromised in patients with schizophrenia [16,
17, 21, 22] as well as in their unaffected first-degree
relatives [22–24]. Studies on the heritability of P50
suppression have strongly suggested that the variation
in P50 is under the influence of genetic factors and
that P50 suppression is a good candidate endopheno-
type [25]. Freedman and colleagues [26] used P50
suppression to identify a potential susceptibility locus
for schizophrenia on chromosome 15, a chromo-
somal region where the CHRNA7 gene encoding the
alpha-7 nicotinic acetylcholine receptor resides. Link-
age disequilibrium in this region [27] has subse-
quently been reported, and variations in the
350
promoter region of the CHRNA7 gene have been
found to be associated with schizophrenia and/or
P50 suppression abnormalities [27, 28].
Eye-tracking dysfunction as endophenotypes
for schizophrenia and bipolar disorder
Eye movements are generally of two forms: saccadic
(brief and extremely rapid movements) and smooth
(“smooth pursuit” eye movements occur only when
the subject is following an object moving at a constant
velocity, such as a pendulum or bright dot on a
computer monitor). Initiation and maintenance of
smooth pursuit eye movements (SPEMs) involve inte-
gration of functions of the prefrontal cortex frontal
eye fields, visual and vestibular circuitry, thalamus,
and cerebellum, as well as the muscles and neural
circuitry directly responsible for eye movement [29].
Studies have found that patients with schizophrenia
and their unaffected relatives have increased rate of
deficiencies in SPEMs [30, 31]. Similar results have
also been reported in bipolar disorder [31, 32]. Stud-
ies have reported evidence for linkage of SPEM
phenotypes to chromosome 6p23–21 [33, 34]; the
region harbors two genes associated with risk of
schizophrenia, ATXN1 (SCA1) and NOTCH4 [32].
Neurocognitive endophenotypes
for schizophrenia and ADHD
Family and twin studies have suggested moderate
heritability of working memory deficits in schizo-
phrenia [35–37]. Recently, researchers such as the
Consortium on the Genetics of Schizophrenia
(COGS), a seven-site collaboration that examines the
genetic architecture of quantitative endophenotypes
in families with schizophrenia, have attempted to
select neurocognitive tasks as endophenotypic meas-
ures in genetic studies [38, 39]. The COGS neurocog-
nitive assessment includes measures of attention,
verbal memory, working memory, and a computer-
ized neurocognitive battery that also includes facial
processing tasks [40].
A study of Finnish twins [41] using the sum of
performance scores on four neuropsychological tests
as an endophenotype suggested linkage and associ-
ation to a region of chromosome 1q, a region previ-
ously suggested in traditional linkage studies of
schizophrenia [42]. Studies [43] have also shown an
association of poorer performance on a working

memory task with the gene encoding the enzyme
catechol-O-methyltransferase (COMT) located at
chromosome 22q and increased risk for developing
schizophrenia. This chromosomal region has been
reported to be linked to both schizophrenia and bipo-
lar disorder and overlaps with a deletion that has been
associated with velo-cardio-facial syndrome
(DiGeorge syndrome) and schizophrenia [44].
A functional polymorphism, Val158Met, which
causes a valine to methionine mutation at peptide
158 in the gene encoding COMT, results in four-fold
decreased activity in the rarer Met allele, hence pref-
erentially increasing prefrontal extra-synaptic dopa-
mine, because COMT provides the major clearing
step for dopamine released from the synapse in this
region [45]. The more common Val allele is associ-
ated with less efficient cognitive processing by the
brain, and is transmitted at a higher rate than the
Met allele to patients with schizophrenia than to their
nonaffected siblings [46].
Similarly, many neurocognitive measures of
executive functions, for example, Attention Network
Task, Continuous Performance Test, Matching
Familiar Figures Test, Span of Apprehension Test,
Spontaneous Selective Attention Task, and Wisconsin
Card Sorting Test, have been proposed as endophe-
notypes for ADHD [47, 48]. Studies have shown that
deficits in neurocognition are a correlate of ADHD
and show preliminary evidence of heritability and
association with relevant candidate genes. Nonethe-
less, studies that have assessed the familial and genetic
overlap of neurocognitive impairments with ADHD
have yielded inconsistent results. In order for execu-
tive function deficits to be used as an endophenotype
for ADHD, more studies with greater attention to the
neurocognitive heterogeneity of this disorder and to
the precision of measurement of the neuropsycho-
logical tests are required [49].
Neuroimaging endophenotypes
Another powerful and rapidly evolving technique
used in psychiatric genetic studies is functional mag-
netic resonance imaging (fMRI), which combines the
neurocognitive tasks with the anatomical location of
brain activities. Studies have demonstrated that the
more common Val allele of the functional poly-
morphism Val158Met in the COMT gene is associated
with less efficient cognitive processing by the brain,
and that more blood flow in the prefrontal cortex
Chapter 29: Endophenotypes in psychiatric genetics
(PFC) was needed to accomplish the same task in
the fMRI assessment, suggesting that activation of
the dorsolateral prefrontal cortex is less efficient in
those subjects [50]. This inefficiency was observed in
schizophrenic subjects as well as unaffected siblings of
schizophrenic patients [51].
In agreement with this discovery, variation in
COMT also modulates other PFC-dependent neuro-
psychological performance [45] and the cortical
response to amphetamine, which increases synaptic
dopamine [52]. The latter finding suggests that the
COMT genotype links prefrontal dopamine stimula-
tion and neuronal activities, in which homozygotes
for the Val-encoding allele presumably possesses less
synaptic dopamine due to maximal COMT activity
than the Met allele carriers. Additional evidence for
this comes from a positron emission tomography
(PET) study [53] showing that the COMT genotype
has an impact on the prefrontal regulation of mid-
brain dopamine synthesis.
In addition to fMRI, phenotypes measured by
anatomical imaging such as structural MRI have also
been proposed as endophenotypes for some psychi-
atric disorders. For example, reduction in orbitofron-
tal volume in ADHD [54], decreased gray matter in
insular cortex, and temporal lobe gray matter abnor-
malities in schizophrenia [55]. Recent studies using
diffusion tensor imaging (DTI) and magnetization
transfer ratio (MTR) protocols have begun to shed
more light on the white matter deficits among schizo-
phrenia patients, and studies on their heritability
(including their potential as endophenotypes) are
under way [55, 56].
Neurochemical metabolites as
endophenotypes
In addition to soft neurological signs and neurocog-
nitive performance, many studies have focused on
neurochemicals, neurohormonal substances, and
their metabolites in search of candidate endopheno-
types for psychiatric disorders. Here we use a few
recent studies in AD and suicide as examples.
Aggregation and deposition of amyloid beta (Ab)
in the brain is thought to be central to the pathogen-
esis of Alzheimer’s disease (AD). Recent studies sug-
gest that cerebrospinal fluid (CSF) Ab levels are
strongly correlated with AD status and progression.
Using CSF Ab levels as a phenotype for AD, a recent
study has identified a DNA sequence variation in the
351
 Chapter 29: Endophenotypes in psychiatric genetics

(a) 12

Frequency (Hz)
Fz

Cz
6

Pz
3

Theta
–216 –31 153 338 522 706 891 1075 1259
(b) Time (ms)

Chromosome 7
3.5 θ
Fz, Max LOD = 3 .16 at 161 cM
3
Cz, Max LOD = 3 .6 at 164 cM

Pz, Max LOD = 2 .29 at 162 cM

2.5

2
LOD

1.5

0.5 CHRM2
GRM8

0
D7S1790

D7S1802

D7S1838

D7S2846

D7S1830

D7S3046
D7S1870

D7S1797

D7S1796

D7S1799

D7S1817

D7S2847
D7S1809

D7S1804

D7S1824

D7S1805
D7S513

D7S629
D7S673

D7S817

D7S521
D7S691
D7S478
D7S679
D7S665

D7S820

D7S821

D7S490

D7S509

D7S794
NPY2

0 20 40 60 80 100 120 140 160 180

Chromosome position (cM)
CHRM2
(c)
exon

exon

exon

exon

exon

1 2 3 4 5 6
Coding
5’-UTR 3’-UTR
Sequence

(d)

SNPs rs6948054 rs1424548

rs1378646 rs1378650
rs1455858
rs7782965
rs978437

Figure 29.1 Illustration of endophenotype strategy: from neurocognitive endophenotypes to genes. (a) Endophenotype (Theta oscillation),
Theta (y, 3–7 Hz) event-related oscillations (EROs) underlying the processing of target stimuli during P3 production (300–700 ms) in a visual
oddball task as seen in this time-frequency representation (right panel). Head plot (left) displays scalp topography of theta ERO power.

352
 Chapter 29: Endophenotypes in psychiatric genetics

presenilin 1 (PSEN1) gene that is a novel disease-
causing mutation in clinically characterized research
subjects with extreme values of CSF [57].
Another example is cortisol response to psycho-
social stress with regard to suicidal behavior. Disturb-
ances in hypothalamic-pituitary-adrenal (HPA) axis
function have been observed in suicide attempters
using various indices, including CSF corticotrophin
releasing hormone (CRH), cortisol levels following
dexamethasone challenge, and urinary cortisol [58,
59]. Twin studies of cortisol levels in blood, urine,
and saliva estimated heritability to be approximately
60% [60]. A twin study of cortisol response to psy-
chosocial stress (Trier Social Stress Test [TSST])
estimated heritability of cortisol response with repeti-
tion of the stressor to be between 56 and 97% [61].
Several polymorphisms that have been shown to be
associated with cortisol response to TSST include the
mineralocorticoid and glucocorticoid receptor genes
[62], the 5HTTLPR [63], and the g-aminobutyric
acid (GABA)-A a-6 receptor gene [64]. Studies are
necessary to establish if the same deficits in cortisol
response are more frequent in unaffected relatives
of suicidal individuals compared with the general
population.
Neurocognitive phenotypes in alcoholism
and related disorders: brain oscillations as
endophenotypes
In this section, we more fully illustrate the endo-
phenotype approach – from endophenotype to
genes, using brain oscillations as endophenotypes
in the search for genes predisposing to alcoholism
and related disinhibitory disorders in the COGA
project (Figure 29.1). Alcoholism is a common
complex (non-Mendelian) disorder with contribu-
tions from both genetic and environmental influ-
ences and their interactions. Recent evidence
suggests that alcohol dependence is part of a
spectrum of disinhibitory disorders, which include
externalizing and other substance use disorders.
Many of the same genetic risk factors are postulated
to underlie these disinhibitory co-occurring dis-
orders which are explained by a common under-
lying genetic liability involving impulse control
[65]. These findings are in keeping with electro-
physiological findings reflecting a similar electro-
physiological profile in these related disinhibitory
disorders [15]. Brain dysfunction is likely to be
involved in a genetic predisposition to develop alco-
holism and other psychiatric disorders, and neuro-
electric events may serve as excellent biological
markers or endophenotypes.
Brain oscillations as endophenotypes
Brain oscillations provide a rich source of potentially
useful endophenotypes for psychiatric genetics, as
they represent important neural correlates of infor-
mation processing and cognition in humans. These
oscillations are dynamic indices of the millisecond
by millisecond balance between excitation and inhib-
ition in brain neural networks and have exquisite
temporal resolution. They reflect ensembles of
neurons firing in synchrony and represent the basic
mechanism of neural communication. High frequency
(beta [12–28 Hz], gamma [28þ Hz]) synchronizations
are involved in short-range communication, while low
frequencies (delta [1–3 Hz], theta [4–7 Hz], alpha [8–
12 Hz]) are involved in longer range communication
between brain areas [66]. Brain oscillations represent
traits less complex and more proximal to gene function
than either diagnostic labels or traditional cognitive
measures. Therefore, these oscillations may be utilized
as phenotypes of cognition, and as valuable endophe-
notypes for the understanding of some complex gen-
etic brain disorders. Several brain oscillations meet
criteria for endophenotypes (see above); they are highly
heritable: delta 76%, theta 89%, alpha 89%, and beta
86% [67].

Figure 29.1 (cont.) (Theta EROs are correlates of impaired cognitive brain processes in alcoholism and risk.). (b) Genetic linkage analysis scans
the entire genome to assess chromosomal regions that contain polymorphic genetic markers that are linked to a quantitative trait (theta
ERO power) within families. This is a logarithm of odds (LOD) score plot of linkage for theta ERO power at frontal (Fz), central (Cz), and parietal
(Pz) electrodes on chromosome 7 with a significant linkage peak that contains 2 candidate genes under the linkage peak – CHRM2 and GRM8.
(c) Candidate gene studies focus on genes underlying significant linkage peaks and/or genes with relevant biological significance. This is a
schematic diagram of the CHRM2 gene indicating its coding region (gray) and exons. (d) Genetic association tests are performed for each
single nucleotide polymorphism (SNP) in the candidate gene and the trait variable. SNPs are identified within and flanking the candidate
CHRM2 gene (exons and introns). SNPs with significant associations (gray boxes) with the endophenotype indicate loci of interest. (The right-
most box with two SNPs lies in the region beyond the 3’ UTR of the gene.) See plate section for color version.

353
 Chapter 29: Endophenotypes in psychiatric genetics
In our strategy for finding susceptibility genes for
alcohol dependence and related disorders, we select
brain oscillations that not only differentiate between
alcoholics and controls, but also between high risk
offspring of alcoholics and controls, to be sure that we
are selecting “trait” rather than “state” measures that
meet criteria for good endophenotypes. COGA is a
genetic study of densely affected families ascertained
through affected probands in treatment, where there
were at least three affected (DSM III-R and Feighner
definite for alcohol dependence) first-degree relatives.
We focused on neuroelectric endophenotypes (such
as resting eyes-closed EEG [e.g. power and coher-
ence], ERPs [e.g. the P3 component], and event-
related oscillations [EROs] [e.g. theta and delta
event-related oscillations]) recorded during sensory
and cognitive tasks. As will be shown in this chapter,
we have successfully used this approach with brain
oscillations as endophenotypes first to target chromo-
somal regions using linkage analysis, and secondly
identifying potential candidate genes that may be
involved in both brain function and diagnosis of
alcohol dependence and related disorders contained
in the chromosomal regions under that linkage peak.
These genes are then more closely investigated using
association analyses (Figure 29.1).
Resting EEG beta power: GABRA2 gene
EEG signatures of the resting state of the brain have
revealed characteristic patterns in individuals with
alcoholism and also those with high risk for develop-
ing alcoholism and related spectrum of externalizing
conditions [15]. Across most studies published in this
field, increased beta power in the EEG has emerged as
one such important feature; it is noted in affected
[68–73] and high risk offspring of alcoholics [74,
75], including a large sample of offspring of alcoholics
at high risk from the COGA project [76]. Increased
resting beta power has been reported at frontal leads
in those who also have a diagnosis of antisocial per-
sonality disorder (ASP) [77]. Also, increased beta
power may be associated with increased vulnerability,
as female high risk subjects with a larger number of
affected first-degree relatives displayed significantly
elevated beta power compared to those with just one
affected parent [72]. Thus evidence of elevated beta
power provides strong support for the excitation–
inhibition imbalance model proposed to underlie the
predisposition to alcohol dependence [78].
354
As increased resting beta power is already observed
in offspring before the onset of alcohol dependence,
it is considered to be a trait rather than a state measure.
Significant linkage and linkage disequilibrium between
the EEG beta frequency and a g-aminobutyric acid
(GABAA) receptor gene on chromosome 4 has been
reported by COGA (Figure 29.2) [79]. With the use
of multiple single nucleotide polymorphisms (SNPs)
across this cluster of GABAA receptor genes on
chromosome 4, that includes GABRA2, GABRA4,
GABRB1, and GABRG1, we were able to specifically
identify that it was variations only in the GABRA2
receptor gene that accounts for the linkage /linkage
disequilibrium findings with the beta frequency. Thus,
variations in GABRA2 (the gene encoding the a-2
subunit of the GABAA receptor) affect brain oscilla-
tions and are directly involved in the level of neural
excitability (balance between excitation and inhib-
ition). There is a strong relationship between the most
significant SNP (rs279836) in the GABRA2 receptor
gene and beta EEG power. Of note individuals who
are homozygous for the rarer genotype (15%) of this
SNP have significantly increased EEG beta compared
to individuals with all other genotypes, indicating
underlying central nervous system (CNS) disinhibi-
tion. The same GABRA2 gene associated with the
EEG beta endophenotype was subsequently found
to be associated with DSM-IV diagnosis of alcohol
dependence [80], substance dependence [81], ASP
[82], and childhood conduct disorder [83]. The associ-
ation between GABRA2 and alcohol dependence has
been replicated by a number of groups throughout the
world [84–89].
Fast synaptic inhibition in the mammalian CNS is
mediated largely by activation of GABAA receptors
[90]. GABAA actions are a fundamental requirement
for both gamma and beta oscillations to occur, and
blockade of these receptors results in loss of syn-
chronization [91]. Although the recording of elec-
trical oscillations from a neural population reflects
the firing of multiple excitatory pyramidal cells, the
mechanism underlying beta and gamma oscillations
depends on the firing patterns of a network of inhibi-
tory interneurons [92, 93], that are gated by their
mutually induced GABAA action [94].
These genetic findings relating EEG beta and the
GABRA2 gene suggests that variations in the GABRA2
gene affect the inhibitory tone and thus affect the level
of neural excitability, which may underlie the predis-
position to develop alcohol dependence and related
 (a) Chromosome 4 (b) Chromosome 7

3.5

2.5

LOD
1.5

CHRM2
GRM8
0.5
GABA-A
0
D4S2639
D4S2382
D4S1627
D4S3025
D4S1536

D4S1630

D4S1645
D4S2432
D4S3253
D4S2367
D4S1558
D4S2393
D4S2630
D4S3088
D4S2361
D4S1544
D4S2404
D4S1559

D4S1570
D4S1651

D4S1625

D4S1629
D4S1626

D4S2374

D4S1652
GABRB1

D7S1790

D7S1802

D7S1838

D7S2846

D7S1830
D7S3046
D7S1870

D7S1797

D7S1796

D7S1799

D7S1817

D7S2847
D7S1809

D7S1804

D7S1824

D7S1805
D4S400

D4S171

D7S513

D7S629
D7S673

D7S817

D7S521
D7S691
D7S478
D7S679
D7S665

D7S820

D7S821

D7S490

D7S509

D7S794
FABP2
ADH3

NPY2
50 100 150 200 0 20 40 60 80 100 120 140 160 180
Chromosome position (cM) Chromosome position (cM)

Figure 29.2 (a) Linkage plots showing maximum logarithm of odds (LOD) scores with significant linkage peaks over the GABAA receptor gene cluster on chromosome 4 with two resting
EEG phenotypes. Red trace: Resting EEG beta band power – beta 2 (16.5–20.0 Hz). The dataset consists of 1553 individuals from 250 families. Green trace: Resting EEG high theta band (6–7
Hz) interhemispheric coherence at parieto-occipital bipolar pairs of electrodes (P4_O2-P3_O1). The dataset consists of 1312 individuals from 251 families.
(b) Linkage plot showing maximum LOD scores with significant linkage peaks on chromosome 7, over the region harboring two candidate genes: a cholinergic muscarinic receptor gene
(CHRM2) and a glutamate receptor gene (GRM8). Blue trace: the central midline theta (4–5 Hz) ERO band power (between 300–700 ms, P3 latency window for visual target case during visual
oddball task) on chromosome 7. The dataset consists of 1337 individuals from 253 families; Green trace: Resting EEG theta band (6–7 Hz) interhemispheric coherence at centro-parietal
bipolar pairs of electrodes (C4_P4-C3_P3) on chromosome 7. The dataset consists of 1312 individuals from 251 families. See plate section for color version.
 Chapter 29: Endophenotypes in psychiatric genetics
disorders. Alcoholics and offspring at high risk mani-
fest increased power in EEG beta oscillations, suggest-
ing an imbalance between excitation–inhibition (CNS
disinhibition). This provides a biological hypothesis
relating the underlying CNS disinhibition to the gen-
etic risk for alcohol dependence and related disorders
[78]. The involvement of the GABAergic system in
alcoholism is supported by neuroimaging studies,
which report specific deficits in GABA benzodia-
zepine receptor function in the brains of alcoholics
[95–97] and individuals at risk [98]. Taken together,
these findings suggest GABA deficits probably con-
tribute to a state of hyperexcitability in the brains
of alcoholics and individuals at risk, and this may
underlie the predisposition to develop alcoholism.
Interhemispheric theta coherence: GABRA2
and CHRM2 genes
Another EEG measure that provides a good endo-
phenotype for genetic study is EEG coherence – a
measure of cortical synchronization in neural net-
works, indexing communication in populations of
neurons. Recent studies have suggested a significant
role for theta frequency coherence in normal and
aberrant thalamocortical interactions [99, 100].
Impairments in neural synchrony have been reported
in several psychiatric conditions, including alcohol
dependence. There is evidence in the literature that
not only do alcoholics manifest differences in EEG
power in specific frequency bands, but they also
manifest increased interhemispheric coherence [101,
102]. It has been reported that bilateral intrahemi-
spheric coherences in alpha and beta frequency
bands were increased in both long-term abstinent
and nonabstinent alcoholics compared to controls
[103], particularly for alpha when depressiveness
was included as a covariate; there was no effect of
length of abstinence on these findings. In our labora-
tory we have observed significant increases in resting
EEG interhemispheric high theta (6–7 Hz) coherence
in alcohol-dependent subjects when compared to
normal controls, particularly posteriorly at parietal-
occipital and centroparietal regions [Rangaswamy
et al., unpublished]. Peak theta band coherence was
highest for posterior electrode pairs in alcohol
dependent subjects, displaying a shift from a more
frontocentral prominence as seen in normal control
subjects. Hence, this increased EEG coherence (cor-
tical synchronization) observed in alcoholics may be
356
indicative of altered functional thalamocortical and
cortico-cortical connectivity.
We conducted a whole genome linkage analysis in
COGA using the high theta coherence at parietal-
occipital leads as the phenotype. Highly significant
linkage was found on chromosome 4 in the same region
spotlighted by the EEG beta power phenotype [104;
Rangaswamy et al., unpublished]. Family-based associ-
ation analyses with the cluster of GABAA receptor
genes under this linkage peak revealed strong associ-
ation with a large number of SNPs (several at p < 0.001)
genotyped in GABRA2 for the Caucasian-only subset,
and not with other genes in the cluster (Figure 29.2).
Another earlier study [105] reports that three
exonic variants of the gene encoding the human
GABAB receptor on chromosome 6 modify scalp-
recorded EEG coherence (cortical synchronization).
Parietotemporal coherence showed statistical signifi-
cance associated with exon 7, suggesting that this
exon may be functionally meaningful and impact on
cortical EEG oscillations.
These results suggest that GABRA2 may indeed
influence susceptibility to alcohol and drug depend-
ence, not just by modulating level of neural excitation,
but also by influencing functional connectivity of
interhemispheric networks. In their model of thala-
mocortical dysrhythmia, Linas and co-workers have
proposed that the enhanced low-frequency (theta)
oscillations in the thalamocortical module can affect
the lateral inhibitory drive in the cortex and eventu-
ally result in high frequency coherent activation of
cortical module [106]. This is particularly significant
in light of our genetic findings where two resting state
electrophysiological signatures – beta power (high
frequency activity associated with arousal) and theta
coherence (low frequency synchrony) are both linked
to the same GABAA receptor gene.
The same interhemispheric high theta coherence
phenotype at centroparietal leads indicated significant
linkage on chromosome 7 and significant associations
with a muscarinic acetylcholine receptor (M2) gene
(CHRM2), underlying the linkage peak using family-
based association tests [104] (Figure 29.2). Significant
linkage and association for evoked theta band
responses at the same CHRM2 gene have been previ-
ously reported in COGA [107, 108] (see below).
Both GABAergic and cholinergic systems interact
significantly in the functions of local inhibitory cir-
cuits, thus affecting network functions and influen-
cing cortical synchronization. Increased GABAergic

inhibition is likely to be a mechanism underlying
impaired synaptic plasticity observed with M2 knock-
out mice, who demonstrate impaired behavioral flexi-
bility and memory deficits [109]. A recent study
suggested that M2 receptor activation produces a
presynaptic inhibition of GABA release by long-range
inhibitory neurons of the perirhinal cortex projecting
to the entorhinal cortex [110].
Theta and delta event-related oscillations underlying P3
during target detection: CHRM2 and GRM8 genes
One of the most consistent robust findings in the
literature is the reduced P3 amplitude in alcoholics
and in offspring at high risk prior to alcohol exposure
[111] (for review see [15]), providing a good endo-
phenotype for genetic studies. However, the P3 com-
ponent is not a unitary phenomenon, but emanates
from multiple sources in the brain with contributions
from frontal cortex (including anterior cingulate),
amygdala and hippocampus [112–115]. Low P3
amplitudes coupled with weaker and less well-organ-
ized sources in alcoholics and offspring at risk suggest
inefficient allocation of resources during neural pro-
cessing. This undifferentiated neurophysiological pat-
tern suggests a level of cortical disinhibition in
alcoholics and individuals at risk. The low P3 ampli-
tude is not only observed in abstinent alcoholics and
offspring of alcoholics, but is also present in various
disinhibitory conditions, such as substance abuse,
ASP, conduct disorder, and ADHD. Moreover, indi-
viduals with low P3 amplitudes manifest a signifi-
cantly higher incidence of externalizing disorders
and disinhibitory traits than those with high P3
amplitudes [15].
More recently, event-related activity recorded
during cognitive tasks, including paradigms
eliciting the P3 component, has been examined in
the frequency domain (i.e. EROs). Several studies
have demonstrated that P3 responses are primarily
the outcome of theta and delta EROs elicited during
cognitive processing of stimuli [116–120]. Topo-
graphically, delta ERO power peaks at the posterior
region, while the theta power peak is located in the
frontocentral region [119]; theta oscillations also
contribute strongly to N2 components of ERPs.
Theta and delta EROs underlying P3 are related to
different cognitive functions: Theta is associated
with memory processes and attention, and involves
fronto-limbic or cortico-hippocampal interactions,
and is taken as an index of frontal processing;
Chapter 29: Endophenotypes in psychiatric genetics
reciprocal synchronization has been observed in
the theta range between hippocampus and frontal
and parietal regions in the brain during attentional
tasks [66]. Delta is related to signal detection and
decision-making, and is generated by cortico-
cortical interactions and is prominent after the
target stimuli.
In a visual oddball paradigm, alcoholics mani-
fest significantly less evoked theta and delta ERO
amplitudes while processing the target stimuli
[121]; these findings are most significant anteriorly
for theta, and posteriorly for delta. In order to
determine whether these deficits in theta and delta
oscillations antecede the development of alcoholism
we examined adolescent high risk children of alco-
holics compared to normal children, using the same
paradigm [122]. The results showed that the adoles-
cent offspring of alcoholics have reduced delta and
theta band ERO amplitude (underlying P3) while
processing the target stimuli compared to controls;
differences were most prominent centroparietally
for theta, and parietally for delta. The ERO meas-
ures have provided robust group differentiation and
at times function as more sensitive measures than
P3 in differentiating between high risk and low risk
offspring. Thus, the results of these two studies
demonstrate that decreased theta and delta EROs
to target stimuli may antecede the development of
alcoholism and represents a strong trait marker
(Figure 29.3).)
We have reported significant linkage and associ-
ation between the CHRM2 gene on chromosome 7
and frontal theta oscillations to target stimuli in a
visual oddball task; association analyses using both
population-based tests (Measured Genotype) and
pedigree-based tests (Quantitative Pedigree Disequi-
librium Test, QPDT) indicate significant association
of the frontal theta band ERO phenotype with several
SNPs surrounding exon 4 of CHRM2. Further, an
examination of the slower frequency parietal-occipital
delta band EROs also revealed significant association
with several SNPs [107, 108] (Figure 29.3).
These findings implicate a role of CHRM2 in the
generation and modulation of evoked oscillations.
Theta and delta EROs depend on the level of acetyl-
choline (muscarinic activation). M2 receptors inhibit
presynaptic release of acetycholine, leading to inhib-
ition of irrelevant networks. Muscarinic receptors are
especially concentrated in the forebrain and possibly
serve to maintain the effective balance of relevant/
357
 Chapter 29: Endophenotypes in psychiatric genetics

(a)
Adult subjects
Control Alcohol dependent
Frequency (Hz)

Frequency (Hz)
10 10
θ θ θ
5 δ 5
δ
0 0
0 200 400 600 0 200 400 600
Time δ Time

1 2 3 4 1 2 3 4
Amplitude (μV) Amplitude (μV)

(b) Adolescent subjects

Low risk High risk
12

Frequency (Hz)
12
Frequency (Hz)

θ
7
7 θ
θ
δ 0
δ
0
0 200 400 600 0 200 400 600
Time
Time δ
2 4 6 2 4 6
Amplitude (μV) Amplitude (μV)

Figure 29.3 S-transform derived time-frequency representations of the average instantaneous amplitude that are z-scored for each
frequency. Instantaneous amplitudes were averaged across individual trials of subjects so that nonphase locked or imprecise phase locked
oscillatory energy is preserved. (a) Plots for the target condition at central (Cz) electrode in 120 alcoholic (right) and 120 control (left)
subjects. Center panel – Topographic headplots of the time-frequency regions of interest (TFROIs) indicate that the theta band (4–5 Hz)
power at 300–500 ms shows frontal maxima while the delta band (1–3 Hz) power at 350–700 ms has posterior maxima. Note that
alcoholics have weaker responses in both theta and delta bands. (b) Plots for the target condition at frontal (Fz) electrode in visual oddball task
in 87 high risk adolescent offspring of alcoholics from Collaborative Study on the Genetics of Alcoholism (COGA) families (right) and 57
matched low risk offspring of nonalcoholics from control families (left). Center panel – Topographic head plots for the TFROI that extends
from 300 to 700 ms for each band revealing frontal maxima for theta band power and parietal maxima for delta band power. Note that
high risk adolescents have weaker responses in both theta and delta bands to target stimuli, similar to the alcoholics. See plate section
for color version.

irrelevant networks, hence, having a direct influence
on P3 generation [123]. Genetic underpinnings of
evoked oscillations are likely to stem from regulatory
genes that control the neurochemical processes of the
brain and, therefore influence neural function. The
three major neurochemical substrates contributing to
theta and delta rhythms and P3 involve strong
GABAergic, cholinergic, and glutamatergic system
interactions [123].
Moreover, the cholinergic muscarinic genes have a
major role in memory and cognition [124, 125]. Sev-
eral studies, including the COGA study, have found
evidence that the CHRM2 gene may be involved in
intelligence [125–127]. In the COGA study, we found
evidence of association with multiple SNPs across
CHRM2 and Performance IQ, as measured by the
Wechsler Adult Intelligence Scale-Revised (WAIS-R).
These results remain significant after taking into
account alcohol dependence and depression diagnoses
in the sample [126].
Evidence from the COGA project indicates that
the CHRM2 gene is not only associated with brain
oscillations and cognition, but also clinical diagnosis.
Significant linkage and association were reported for
the CHRM2 gene and a diagnosis of alcohol depend-
ence and depression [128], comorbid alcohol and
drug dependence (a more severe addiction profile)
[129], as well as a spectrum of externalizing disorders
[130]. Other groups have replicated these findings,
reporting that the CHRM2 gene predisposes to alco-
hol dependence, drug dependence and affective dis-
orders [131], and major depression in women [132].

358

Thus genes important for the expression of the endo-
phenotype (brain oscillations) help in identification of
genes that increase the susceptibility for risk of alco-
hol dependence and related disorders [68, 133].
Another likely candidate gene located under the
same theta ERO linkage peak on chromosome 7 is a
metabotropic glutamate receptor (GRM8) gene. The
glutamatergic system is one of the major players
modulating CNS electrophysiological networks,
and in particular is also involved in theta oscilla-
tions and P3 [123]. Family-based association ana-
lyses of theta EROs revealed significant associations
with several SNPs in the GRM8 gene and theta
EROs to target stimuli at frontal, central, and par-
ietal regions [134]. An interesting finding is that
several SNPs were also significantly associated with
the diagnosis of alcohol dependence using ICD-10
diagnostic criteria.
A 3-T proton magnetic resonance spectroscopy
(1H-MRS) study has suggested the involvement of
glutamatergic neurotransmission in integrative
frontal-hippocampal processing [135] and the sensa-
tion-seeking personality dimension [136]. The study
demonstrated a robust relationship between glutam-
ate levels in the hippocampus and frontal theta
activity during auditory stimulus processing. Gluta-
matergic neurotransmission and its neuroadaptive
changes have been proposed as important molecular
determinants of craving and relapse [137, 138]. In
particular, it is suggested that a hyperglutamatergic
state mediates, at least in part, alcohol relapse behav-
ior and maintenance of alcoholism [139]. Several
studies have suggested the involvement of glutamate
receptors – NMDA and metabotropic, in alcohol
relapse [68, 140, 141]. Acamprosate, a drug used to
prevent relapse in alcoholic patients [142], has
been suggested to act through a suppression of a
hyperglutamatergic state created by alcohol addiction
[143, 144].
In light of the theta oscillations showing strong
association to both CHRM2 and GRM8 genes, one
could speculate a synergistic genetic mechanism
underlying this electrophysiological phenotype, thus
opening doors to future research in this direction.
Interestingly, in studies on the rat hippocampus,
authors have reported that a majority of interneurons
strongly immunopositive for the muscarinic M2 or
the mGlu1 receptors were the primary targets of
mGluR8-containing terminals [145]. Rare neurons
coexpressing calretinin and M2 were consistently
Chapter 29: Endophenotypes in psychiatric genetics
targeted by mGluR8-positive boutons. The postsy-
naptic interneuron type-specific expression predicts
a role in adjusting the activity of interneurons
depending on the level of network activity.
Conclusions
Psychiatric disorders result from a complex inter-
action of changing genetic and environmental liabil-
ities across development, possibly with greater genetic
loading in individuals who have early onset of the
disorders. The use of quantitative endophenotypes
provides the power to more easily localize and char-
acterize disease susceptibility genes than diagnostic
categories. The measures of these endophenotypes
reflect the genetic liability of the disorder among
nonaffected relatives of affected individuals. Many of
the same genes important for the expression of the
endophenotypes help in identification of genes that
increase the susceptibility for risk of the disease.
Hence, as this chapter has illustrated, the utility of
quantitative biological endophenotypes for the study
of genetic risk of psychiatric disorders continues to be
very promising.
Endophenotypes may have additional uses in
psychiatry, including uses in diagnosis, classification,
and the development of animal models. The lack of a
biological basis for the classification of psychiatric
disorders has led to limited success in research in
the neurobiology and genetics of psychiatric dis-
orders. Endophenotype-based analysis would be
useful for establishing a biological underpinning for
diagnosis and classification; a net outcome would be
improved understanding of the neurobiology and
genetics of psychopathology.
While the endophenotype approach is not a new
idea, it is an approach whose time has come. Because
of the rapidly evolving technology in the fields of
molecular and statistical genetics, new methods are
emerging that will facilitate the search for genes
underlying complex traits in the near future. With
the advent of SNP genotyping technology, genome-
wide association studies using dense sets of SNPs
across the genome, together with recent advances in
novel statistical genetic techniques and computational
power, have facilitated the identification of genes that
are associated with heritable, genetically influenced
quantitative traits that are risk factors involved in
the etiology of various psychiatric disorders. Once
genes are identified and understood, risk genotypes
359
 Chapter 29: Endophenotypes in psychiatric genetics
and haplotypes can be studied in prospective studies
of young individuals who have not yet developed the
disease, and can lead to an improved understanding
References
1. Akerberg E. Hereditas
1986;105:1–5.
2. Gottesman II, et al. Schizophrenia
and Genetics: A Twin Study
Vantage Point. New York:
Academic Press; 1972.
3. Lewis DA. Am J Psychiatry
2002;159:1467–1469.
4. Kandel ER. Am J Psychiatry
1998;155:457–469.
5. Tsuang MT. Am J Psychiatry
2000;157:489–491.
6. Gottesman, II, et al. Am
J Psychiatry 2003;160:636–645.
7. John B, et al. Science
1966;152:711–721.
8. Munafo MR. Genes Brain Behav
2006;5 Suppl 1:3–8.
9. Skuse DH. Br J Psychiatry
2001;178:395–396.
10. Waldman ID. Biol Psychiatry
2005;57:1347–1356.
11. Walters JT, et al. Mol Psychiatry
2007;12:886–890.
12. Almasy L, et al. Am J Med Genet
2001;105:42–44.
13. Leboyer M, et al. Trends Neurosci
1998;21:102–105.
14. Almasy L, et al. Am J Hum Genet
1998;62:1198–1211.
15. Porjesz B, et al. Clin Neurophysiol
2005;116:993–1018.
16. Braff DL, et al.
Psychopharmacology (Berl)
2001;156:234–258.
17. Braff DL, et al. Arch Gen
Psychiatry 1990;47:181–188.
18. Geyer MA, et al. Brain Res Bull
1990;25:485–498.
19. Swerdlow NR, et al. Biol
Psychiatry 1993;33:298–301.
20. Swerdlow NR, et al. J Neurol
Neurosurg Psychiatry
1995;58:192–200.
360
of how genes contribute to susceptibility, which in
turn can lead to the design of well-targeted prevention
initiatives.
21. Freedman R, et al. Biol Psychiatry
1983;18:537–551.
22. Siegel C, et al. Arch Gen Psychiatry
1984;41:607–612.
23. Clementz BA, et al. Am
J Psychiatry 1998;155:1691–1694.
24. Myles-Worsley M. Am
J Psychiatry 2002;159:2007–2012.
25. Myles-Worsley M, et al. Biol
Psychiatry 1996;39:289–295.
26. Freedman R, et al. Proc Natl Acad
Sci U S A 1997;94:587–592.
27. Freedman R, et al. Am J Med
Genet 2001;105:20–22.
28. Leonard S, et al. Arch Gen
Psychiatry 2002;59:1085–1096.
29. Munoz DP. Prog Brain Res
2002;140:89–96.
30. Hong LE, et al. Arch Gen
Psychiatry 2006;63:259–264.
31. Kathmann N, et al. Am
J Psychiatry 2003;160:696–702.
32. Lin PI, et al. Schizophr Bull
2008;34:791–797.
33. Arolt V, et al. Am J Med Genet
1996;67:564–579.
34. Matthysse S, et al. Am J Med Genet
B Neuropsychiatr Genet
2004;128B:30–36.
35. Cannon TD, et al. Am J Hum
Genet 2000;67:369–382.
36. Conklin HM, et al. Am
J Psychiatry 2000;157:275–277.
37. Park S, et al. Arch Gen Psychiatry
1995;52:821–828.
38. Calkins ME, et al. Schizophr Bull
2007;33:33–48.
39. Gur RE, et al. Schizophr Bull
2007;33:49–68.
40. Greenwood TA, et al. Arch
Gen Psychiatry
2007;64:1242–1250.
41. Gasperoni TL, et al. Am J Med
Genet B Neuropsychiatr Genet
2003;116B:8–16.
42. Hovatta I, et al. Am J Hum Genet
1999;65:1114–1124.
43. Egan MF, et al. Proc Natl Acad Sci
U S A 2001;98:6917–6922.
44. Sklar P. Annu Rev Genomics Hum
Genet 2002;3:371–413.
45. Tunbridge EM, et al. Biol
Psychiatry 2006;60:141–151.
46. Weinberger DR, et al. Biol
Psychiatry 2001;50:825–844.
47. Doyle AE, et al. J Child Psychol
Psychiatry 2005;46:774–803.
48. Doyle AE, et al. Biol Psychiatry
2005;57:1324–1335.
49. Rommelse NN. Expert Rev
Neurother 2008;8:1425–1429.
50. Meyer-Lindenberg A, et al. Mol
Psychiatry 2006;11:867–877, 797.
51. Callicott JH, et al. Am J Psychiatry
2003;160:709–719.
52. Mattay VS, et al. Proc Natl Acad
Sci U S A 2003;100:6186–6191.
53. Meyer-Lindenberg A, et al. Nat
Neurosci 2005;8:594–596.
54. van ‘t Ent D, et al. Neuroimage
2007;35:1004–1020.
55. Prasad KM, et al. Schizophr Bull
2008;34:774–790.
56. Burmeister M, et al. Nat Rev Genet
2008;9:527–540.
57. Kauwe JS, et al. Ann Neurol
2007;61:446–453.
58. Lindqvist D, et al.
Psychoneuroendocrinology
2008;33:1061–1068.
59. Mann JJ, et al. Biol Psychiatry
2009;65:556–563.
60. Bartels M, et al.
Psychoneuroendocrinology
2003;28:121–137.
61. Federenko IS, et al. J Clin
Endocrinol Metab
2004;89:6244–6250.
62. Derijk RH, et al. Eur J Pharmacol
2008;585:492–501.

63. Gotlib IH, et al. Biol Psychiatry
2008;63:847–851.
64. Uhart M, et al.
Neuropsychopharmacology
2006;31:2255–2263.
65. Kendler KS, et al. Arch Gen
Psychiatry 2003;60:929–937.
66. von Stein A, et al. Int
J Psychophysiol 2000;38:301–313.
67. van Beijsterveldt CE, et al. Am
J Hum Genet 1996;58:562–573.
68. Bachteler D, et al.
Neuropsychopharmacology
2005;30:1104–1110.
69. Bauer LO.
Neuropsychopharmacology
2001;25:332–340.
70. Costa L, et al. Drug Alcohol
Depend 1997;46:87–93.
71. Propping P, et al. Hum Genet
1981;59:51–59.
72. Rangaswamy M, et al. Biol
Psychiatry 2002;52:831–842.
73. Winterer G, et al. Psychiatry Res
1998;78:101–113.
74. Finn PR, et al. Alcohol Clin Exp
Res 1999;23:256–262.
75. Gabrielli WF, Jr, et al.
Psychophysiology
1982;19:404–407.
76. Rangaswamy M, et al. Neuroimage
2004;21:329–339.
77. Bauer LO, et al. J Stud Alcohol
1993;54:577–589.
78. Begleiter H, et al. Alcohol Clin Exp
Res 1999;23:1125–1135.
79. Porjesz B, et al. Proc Natl Acad Sci
U S A 2002;99:3729–3733.
80. Edenberg HJ, et al. Am J Hum
Genet 2004;74(4):705–714.
81. Agrawal A, et al. Behav Genet
2006;36(5):640–650.
82. Dick DM, et al. J Stud Alcohol
2006;67:185–194.
83. Dick DM, et al. Behav Genet
2006;36:577–590.
84. Covault J, et al. Am J Med Genet
B Neuropsychiatr Genet
2004;129:104–109.
Chapter 29: Endophenotypes in psychiatric genetics
85. Drgon T, et al. Am J Med Genet
B Neuropsychiatr Genet
2006;141:854–860.
86. Enoch MA, et al. Am J Med Genet
B Neuropsychiatr Genet
2006;141:599–607.
87. Fehr C, et al. Psych Genet
2006;16:9–17.
88. Lappalainen J, et al. Alcohol Clin
Exp Res 2005;29:493–498.
89. Soyka M, et al. J Psychiatr Res
2008;42:184–191.
90. Tobler I, et al. Proc Natl Acad Sci
U S A 2001;98:6464–6469.
91. Haenschel C, et al. Proc Natl Acad
Sci U S A 2000;97:7645–7650.
92. Faulkner HJ, et al. Br
J Pharmacology 1999;128:1813–
1825.
93. Kopell N, et al. Proc Natl Acad Sci
U S A 2000;97:1867–1872.
94. Whittington MA, et al. Int
J Psychophysiol 2000;38:315–336.
95. Abi-Dargham A, et al. Am
J Psychiatry 1998;155:1550–1555.
96. Krystal JH, et al. Arch Gen
Psychiatry 2006;63:957–968.
97. Lingford-Hughes AR, et al. Br
J Psychiatry 1998;173:116–122.
98. Volkow ND, et al. Alcohol Clin
Exp Res 1995;19:510–516.
99. Sarnthein J, et al. J Neurosci
2007;27:124–131.
100. Sarnthein J, et al. Int
J Psychophysiol 2005;57:87–96.
101. Kaplan RF, et al. J Stud Alcohol
1985;46:122–127.
102. Michael A, et al. Acta Psychiatr
Scand 1993;87:213–217.
103. Winterer G, et al. Acta Psychiatr
Scand 2003;108:51–60.
104. Porjesz B, et al. Scientific World J
2007;7:131–141.
105. Winterer G, et al. Am J Med Genet
B Neuropsychiatr Genet
2003;117:51–56.
106. Linas R. Rebound Excitation as the
Physiological Basis for Tremor:
A Biophysical Study of the
Oscillatory Properties of
Mammalian Central Neurons In
Vivo. London: Macmillan; 1984.
107. Jones KA, et al. Behav Genet
2006;36:627–639.
108. Jones KA, et al. Int J Psychophysiol
2004;53:75–90.
109. Seeger T, et al. J Neurosci
2004;24:10117–10127.
110. Apergis-Schoute J, et al. J Neurosci
2007;27:4061–4071.
111. Begleiter H, et al. Science
1984;225:1493–1496.
112. Ardekani BA, et al. Brain Res Cogn
Brain Res 2002;14:347–356.
113. Halgren E, et al. Science
1980;210:803–805.
114. Kiehl KA, et al. Schizophr Res
2001;48:159–171.
115. Menon V, et al. Neuroreport
1997;8:3029–3037.
116. Basar-Eroglu C, et al. Int
J Psychophysiol 1992;13:161–179.
117. Basar E, et al. IEEE Eng Med Biol
Mag 1999;18:56–66.
118. Karakas S, et al. Neuroscience Lett
2000;285:45–48.
119. Karakas S, et al. Clin Neurophysiol
2000;111:1719–1732.
120. Yordanova J, et al. Neuroreport
1996;8:277–280.
121. Jones KA, et al. Clin Neurophysiol
2006;117:2128–2143.
122. Rangaswamy M, et al. Int
J Psychophysiol 2007;63:3–15.
123. Frodl-Bauch T, et al.
Neuropsychobiology
1999;40:86–94.
124. Calabresi P, et al. Eur
J Neuroscience
1998;10:3020–3023.
125. Comings DE, et al. Mol Psychiatry
2003;8:10–11.
126. Dick DM, et al. Behav Genet
2007;37:265–272.
127. Gosso FM, et al. BMC Med Genet
2007;8:66.
128. Wang JC, et al. Hum Mol Genet
2004;13:1903–1911.
361
 Chapter 29: Endophenotypes in psychiatric genetics
129. Dick DM, et al. Addiction
2007;102:1131–1139.
130. Dick DM, et al. Arch Gen
Psychiatry 2008;
65:310–318.
131. Luo X, et al. Hum Mol Genet
2005;14:2421–2434.
132. Comings DE, et al. Am
J Med Genet 2002;
114:527–529.
133. Begleiter H, et al. Int
J Psychophysiol 2006;60:162–171.
134. Chen AC, et al. Am J Med
Genet B Neuropsychiatr
362
Genet 2009;150B:
359–368.
135. Gallinat J, et al.
Psychopharmacology
2006;187:103–111.
136. Gallinat J, et al. Neuroimage
2007;34:671–678.
137. Cornish JL, et al. J Neurosci
2000;20:RC89.
138. Tzschentke TM, et al. Crit
Rev Neurobiol 2000;
14:131–142.
139. Tsai G, et al. Annu Rev Med
1998;49:173–184.
140. Holter SM, et al. Pharmacol
Biochem Behav 2000;
66:143–151.
141. Krystal JH, et al. Pharmacol Ther
2003;99:79–94.
142. Mann K, et al. Alcohol, Clin
Experiment Res 2004;
28:51–63.
143. Dahchour A, et al. Prog Neurobiol
2000;60:343–362.
144. Spanagel R. Addiction
2005;100:1813–1822.
145. Ferraguti F, et al. J Neurosci
2005;25:10520–10536.
 Developmental disorders
Chapter
30 Craig A. Erickson, Khendra I. Peay, and Christopher J. McDougle
Introduction
This chapter will review recent evidence describing
the genetic underpinnings of several developmental
disorders. While a full discussion of the genetics of
autistic disorder occurs elsewhere in this text, the
phenotypic overlap between autism and other devel-
opmental disorders whose genetic underpinning are
more readily understood will be highlighted in this
chapter. Discussion of Down syndrome, Rett’s dis-
order, Prader–Willi syndrome, Angelman syndrome,
Smith–Magenis syndrome, and fragile X syndrome
will place the focus of this chapter on disorders with
well characterized genetic abnormalities combined
with phenotypic overlap with autism. The genetics
of these well characterized developmental disorders
provide a window of understanding to aide investi-
gation into specific genetic contributions to autism.
Additionally, given the relative genetic homogeneity
of these disorders, investigators can begin to better
understand the link from gene(s) to neurochemistry/
neuroanatomy to behavior.
Down syndrome: disorder of gene
dosage
Down syndrome represents the most common known
cause of developmental disability. Down syndrome
occurs in 1 in 1000 live births and results from tri-
somy of chromosome 21 [1]. Down syndrome rates
increase with advancing maternal age. Physical
markers of the disorder include hypotonia, flat facies,
slanted palpebral fissures, small ears, and single or
Simian palmar crease [2]. Individuals with Down
syndrome are also at increased risk for cardiac anom-
alies, thyroid dysfunction, acute leukemia, seizures,
celiac disease, diabetes mellitus, and early onset of
Principles of Psychiatric Genetics, eds John I. Nurnberger, Jr. and Wade H. Berrettini. Published by Cambridge University
Press. © Cambridge University Press 2012.
dementia among other disorders [3]. Despite being
associated with language impairment and intellectual
disability, many persons with Down syndrome show
good social communication skills [1]. However, there
is also a subgroup of Down patients with autism
spectrum disorders [1]. About 7–15% of persons with
Down syndrome additionally meet diagnostic criteria
for an autism spectrum disorder (ASD), a rate that is
more than 10 times the rate of ASD in the general
population [1, 4–7].
Down syndrome is caused by genomic-dosage
imbalance [8]. Since there are 3 copies of chromo-
some 21 rather than 2, it would be expected that in
affected persons, 1.5-fold gene expression would
occur for genes coded on that chromosome [8]. While
on average gene expression on chromosome 21 in
Down syndrome is upregulated near the expected
1.5-fold rate, great variation in gene expression has
been found in Down syndrome cell line analysis [8].
This gene expression variability is likely related to the
significant phenotypic variability that is characteristic
of the disorder [8]. In Down syndrome cell line analy-
sis, only between 39 and 62% of genes on chromo-
some 21 showed significant upregulation compared
to control samples [8].
Utilizing cell line gene expression analysis in
Down syndrome, researchers have been able to char-
acterize genes on chromosome 21 as: (1) potentially
characteristic of core features of Down syndrome; (2)
characteristic of phenotypic variability seen in Down
syndrome; and (3) not likely involved in the presen-
tation of Down syndrome [8]. Genes whose expres-
sion is consistently higher in Down syndrome cell
lines compared to controls are genes that likely
involve aspects of the Down syndrome phenotype
that exist in all affected persons [8]. Genes in this
group are potentially involved in nuclear signaling
363
 Chapter 30: Developmental disorders
and apoptosis, post-translational modification of pro-
teins, regulation of DNA replication, and interferon
signaling pathways [8]. Genes with partially overlap-
ping expression between Down syndrome and control
samples may represent genes that are associated with
the variable phenotypic expression of Down syn-
drome [8]. Many genes representing between one-half
and one-third of genes expressed from chromosome
21 fall into this category [8]. Genes with similar cell
line expression in Down syndrome cells and controls
are thought to likely not contribute to the disorder.
The gene encoding for amyloid precursor protein
(APP), which has been shown to be dosage sensitive
in the brain and is known to confer risk for early-
onset dementia, is included among genes whose
dosage mirrors control samples in Down syndrome
fibroblast and lymphoblast cell line analysis [8]. This
result is contrary to the expectation that the gene for
APP would be upregulated in Down syndrome cell
lines given the association between Down syndrome
and early dementia. This finding points to the likely
need for cell line analysis in Down syndrome beyond
utilization of fibroblast and lymphoblast cells [8].
Given the developmental nature of some aspects of
the Down syndrome phenotype, in particular the
onset of dementia, future cell line work in this field
may also need to utilize cell lines at different develop-
mental stages/ages [8].
Down syndrome is a phenotypically variable, rela-
tively common cause of developmental disability.
Much remains to be done in mapping the contribu-
tion of specific genes on chromosome 21 to the
pathopysiology of the disorder. Future work in this
area will likely aid understanding not only of the
cause of core features of Down syndrome, but also
help to define potential risk genes involved in the
phenotypic variability of the disorder. Identifying
those genes which may define variable presentations
of the disorder will likely aid in the understanding of
the genetics of autistic disorder given the over-repre-
sentation of autism in persons with Down syndrome.
Rett’s disorder: persistence of genetic
heterogeneity
Rett’s disorder is an X-linked neurodevelopmental
disorder that affects approximately 1 in 10 000
females [9]. Rett’s disorder rarely occurs in males
and it is thought that genetic mutations associated
with Rett’s are often not compatible with life in males.
364
In 1999, mutations at Xq28 within the gene encoding
methyl-CpG-binding protein 2 (MECP2) were identi-
fied as a major cause of Rett’s disorder [10]. In about
70–80% of clinical cases of Rett’s disorder, mutations
in this gene exist [10]. Rare familial cases of Rett’s
disorder do exist when carrier females with signifi-
cantly skewed X inactivation patterns remain asymp-
tomatic or mildly affected and then potentially pass
on their MECP2 mutation to their children [11].
Rett’s disorder is marked by apparently normal
prenatal and perinatal development, normal psycho-
motor development through the first 5 months of life,
normal head circumference at birth, and deceleration
of head growth between ages 5 and 48 months [12].
Rett’s disorder is also defined by losses in previously
gained skills including loss of purposeful hand move-
ments between ages 5 and 30 months with the subse-
quent development of stereotypic hand movements
(classic hand ringing or hand washing) and loss of
social engagement early in the disorder [12]. Severe
psychomotor retardation, severely impaired language
skills, and poor coordination also mark the clinical
presentation of the disorder [12]. Individuals with
Rett’s disorder frequently also have seizures, a dis-
ordered sleep pattern, and bruxism [13, 14].
With allelic heterogeneity marked by over 200
specific possible mutations identified in the MECP2
gene accounting for the majority of Rett’s disorder
cases, some work has been done to test for association
of specific mutations with clinical features [13]. In a
study of 135 females with Rett’s disorder in a cohort
from the United Kingdom, the overall mutation effect
on clinical presentation was found to be minimal [13].
Other reports have noted that mutations more prox-
imal in the MECP2 gene are associated with more
significant neurological impairment and gastrointest-
inal problems [14]
MECP2 functions in initiating and/or maintaining
neuronal maturation [15]. In animals, expression of
MECP2 has been shown to correlate with neuronal
maturation immediately prior to synaptogenesis [16,
17]. These animal findings correlate well with the
clinical course of Rett’s disorder with symptom onset
during a period of rapid human neuronal develop-
ment and growth marked by synaptogenesis [14].
While the genetic underpinnings of Rett’s disorder
most often relate to mutations in a single gene, the
disorder is still marked by some significant pheno-
typic variation, and a significant minority of cases,
20–30%, lack classic MECP2 mutations. While variant

X inactivation could play a part in this presentation,
other factors not yet quantified (variable gene expres-
sion, epistasis, gene–environment interactions, etc.)
likely contribute to this occurrence. It is likely that
MECP2 deficits represent only one means by which a
neurobiological effect leads to the classic clinical pre-
sentation of the disorder.
Prader–Willi and Angelman syndromes:
a study in genetic imprinting
Prader–Willi and Angelman syndromes represent dis-
orders of genomic imprinting. Although people
inherit one seemingly equivalent gene copy from each
parent, in some cases gene pairs are not functionally
equivalent (i.e. they are “imprinted”) as gene expres-
sion is based upon parent of origin. Imprinting is
termed epigenetic because the production of func-
tional protein is governed by factors that regulate
protein expression and not the actual nucleotide
sequence [18]. Genomic imprinting has been identi-
fied in about 30 genes [18], including those on
chromosome 15q11–13 that result in Prader–Willi
and Angelman syndromes.
Prader–Willi syndrome is a rare genetic disorder
with a prevalence of 1 in 10 000–15 000 births [19].
Newborns with Prader–Willi syndrome show hypo-
tonia and failure to thrive followed within the first few
years of life by hyperphagia, food preoccupations, and
frequent obesity [18]. Prader–Willi syndrome is also
marked by mild to moderate mental retardation and
frequently with temper tantrums, aggression, and
obsessive–compulsive behaviors including skin
picking [20, 21]. Prader–Willi syndrome significantly
overlaps with autism: approximately one in four per-
sons with Prader–Willi syndrome have a comorbid
autism spectrum disorder (ASD) [22]. Prader–Willi
syndrome is caused by the absence of paternally
derived imprinted genes within the 15q11–13 region
[23]. While 70% of Prader–Willi syndrome cases are
derived from paternally derived chromosomal dele-
tions, the rest of nondeletion cases are caused by
uniparental disomy in which both copies of chromo-
some 15 are derived from the mother [23]. Addition-
ally, a small minority of cases are derived from
mutations in a separate locus controlling gene
imprinting [23]. The final common effect of all
abnormalities leading to Prader–Willi syndrome is a
lack of paternally derived gene expression from the
15q11–13 region.
Chapter 30: Developmental disorders
Angelman syndrome generally results from
maternally derived chromosomal deletion also in the
15q11–13 region [18]. The prevalence of Angelman
syndrome is between 1 in 10 000–40 000 births [24].
The majority (68–75%) of Angelman syndrome cases
are due to maternally derived gene deletions that
impact expression of the gene UBE3A [25]. Approxi-
mately 2–7% of cases are due to paternal uniparental
disomy, 2–5% due to imprinting center mutations,
and 8–11% due to specific maternally inherited muta-
tions in the UBE3A gene [25]. UBE3A codes for a
HECT-domain ubiquitin ligase involved in synapto-
genesis whose substrates and function are not com-
pletely understood [22, 26]. Work utilizing knockout
drosophila and mouse models is underway to better
understand the implication of deficient UBE3A
expression [26, 27]. Angelman syndrome also repre-
sents an example of potential locus heterogeneity
because approximately 12% of people with a clinical
Angelman syndrome diagnosis do not have identifi-
able chromosome 15 abnormalities [28]. This points
to the possibility that other proteins involved with
UBE3A could manifest mutations that also result in
the Angelman syndrome phenotype.
Clinical characteristics found in all persons with
Angelman syndrome include developmental delay,
speech impairment, movement disorder, and “behav-
ioral uniqueness” defined by a combination of fre-
quent laughter/smiling, apparent happy demeanor,
easily excited personality, hypermotoric behavior,
and short attention span [29]. Frequent characteris-
tics found in about 80% of affected persons include
microcephaly, seizures, and abnormal electroenceph-
alogram (EEG) [29]. Sleep problems and feeding dif-
ficulty are also frequently seen [30]. Angelman
syndrome also has significant overlap with autism
with estimates utilizing standard diagnostic instru-
ments reporting an ASD comorbidity rate of 42–
81% [31, 32]. However, ASD rates in pooled case
series using clinical ASD diagnoses are as low as
1.9–3.6% [22]. The wide range of estimated ASD
comorbidity in Angelman syndrome may be due in
some cases to potential ASD over-diagnosis due to the
very low mental age of many persons with Angelman
syndrome [31].
Given the phenotypic overlap between ASD, Pra-
der–Willi, and Angelman syndrome, it is not surpris-
ing that in some otherwise idiopathic ASD cases,
chromosome 15q11–13 abnormalities have been
implicated. In particular, the reports of maternally
365
 Chapter 30: Developmental disorders
derived duplications at 15q11–13 in ASD has gener-
ated the hypothesis that, among cases of Prader–Willi
or Angelman syndrome, the rate of ASD would be
highest in persons with uniparental disomy Prader–
Willi syndrome due to increased expression of mater-
nal 15q11–13 in these cases [22]. In a pooled analysis
assessing reports on the comorbidity of ASD in these
disorders, Veltman et al. noted a significantly
increased prevalence of ASDs in uniparental disomy
Prader–Willi syndrome (38%) compared to deletion
Prader–Willi cases (18%) [22]. Additionally in this
report, uniparental disomy Prader–Willi syndrome
cases had a significantly increased rate of comorbid
ASD compared to a combined group of deletion
Prader–Willi cases and all cases of Angelman syn-
drome (group ASD rate of 20%) [22].
Overall Prader–Willi and Angelman syndromes
represent developmental disorders whose genetic
underpinnings are better understood than autism.
However, despite their relative genetic homogeneity,
the molecular basis of these disorders remains some-
what undefined. Molecular understanding has begun
in Angelman syndrome where a specific deficit
(UBE3A) provides the basis for animal modeling.
Smith–Magenis syndrome: a complex
congenital disorder
Smith–Magenis syndrome (SMS) is a complex devel-
opmental disorder encompassing over 30 clinical fea-
tures affecting neurological, behavioral, and
metabolic functioning [33]. About 90% of SMS cases
are due to a 17p11.2 deletion containing the retinoic
acid induced 1 (RAI1) gene [34]. The reported preva-
lence of SMS, 1 in 25 000 births, may be an under-
estimate since the diagnosis is frequently missed
because of the phenotypic similarity of the disorder
to Down syndrome, Prader–Willi syndrome, and
fragile X syndrome [33].
Physical markers of SMS include short stature,
small hands, obesity, and subtle facial dysmorphism
marked by a short, tented philtrum and relative prog-
nathism [6]. Medical problems common to persons
with SMS include hypercholesterolemia/hypertrigly-
ceridemia, renal abnormalities, thyroid dysfunction,
hypotonia, failure to thrive as an infant, vocal cord
nodules/polyps, hearing loss, and congenital cardiac
anomolies [35–37]. A severe sleep disturbance
marked by an altered circadian rhythm is commonly
seen in persons with SMS [33, 38, 39]. Self-hugging
366
marked by spasmodic tensing of the upper body and
clasping of the hands at the chest while tightly squeez-
ing the arms is thought to be a behavior potentially
unique to SMS [40]. Eighty percent of SMS patients
exhibit self-injury, including repetitive fingernail
picking, wrist-biting, head-banging, and insertion of
foreign objects into the ear [6]. Additional maladap-
tive behavior frequently seen in SMS includes affect-
ive lability, physical aggression, nervousness, and
intense attention-seeking behavior [41]. It is esti-
mated that 80–100% of persons with SMS meet cri-
teria for an ASD [6].
Analysis of the impact of RAI1 haploinsufficiency
has been conducted in a cell line study [33]. Testing
utilizing lymphoblastic cell lines from SMS patients
and HEK293T transfected cells with 50% knockdown
of RAI1 were utilized to assess for altered gene expres-
sion associated with insufficient RAI1 [33]. Many
candidate genes, both upregulated and downregu-
lated, were identified in this manner. These included
genes implicated in lipid biosynthesis, circadian
rhythm, gene expression, cell growth, and neuronal,
cardiovascular, renal, and skeletal function [33].
Overall, RAI1 insufficiency led to dysregulation of
genes impacting a wide variety of homeostatic func-
tions. Genome-wide expression findings from SMS
also showed significant overlap with genes dysregu-
lated in other disorders including fragile X syndrome,
Prader–Willi syndrome, and Down syndrome [33].
This suggests common mechanisms for the overlap-
ping phenotypes of these disorders. Given the broad
clinical presentation of SMS marked by numerous
metabolic, neurologic, medical, and behavioral mani-
festations, the finding of numerous dysregulated
genes in cells with RAI1 haploinsufficiency is not
surprising. Overlap between SMS and other develop-
mental disorders, both phenotypically and genetically,
will allow for a better future understanding of the link
between specific gene dysregulation and phenotypic
expression.
Fragile X syndrome: understanding the
molecular basis of a developmental
disorder
Fragile X syndrome (FXS) is an example of genetic
findings leading to increased molecular understand-
ing of disease pathogenesis and then potential dis-
order-specific treatment modalities. FXS represents

the most common inherited form of intellectual dis-
ability. It is the result of a cysteine-guanine-guanine
(CGG) trinucleotide repeat expansion (> 200 repeats)
located at the 5’ untranslated region of the fragile
X mental retardation gene (FMR1) located on the
long arm of the X chromosome at Xq27.3. The cyto-
sines on the CGG islands can be methylated (epigen-
etically impacting the pattern of gene expression)
[18]. The increase in methylation associated with the
expanded CGG repeats correlates directly with the
extent of the loss of functional fragile X mental
retardation protein (FMRP) [18]. FXS is inherited
from pre-mutation carrier parents (most often
mothers) who have between 55 and 200 CGG repeats.
The increase in repeat numbers with ensuing gener-
ations is an example of genetic anticipation (in which
subsequent generations potentially suffer from more
significant impairment). On rare occasion, FXS has
been associated with FMR1-point mutations instead
of CGG repeat expansion with subsequent methyla-
tion [18]. This is another example of allelic hetero-
geneity in which different mutations in a gene can
lead to similar phenotypes [18].
FXS occurs in approximately 1 in 4000–6000 live
births. Among all persons with intellectual disability,
between 1.9% [42] and 6.0% [43] are thought to
have FXS. As an X-linked disorder, FXS occurs more
frequently in males and the impact of the disorder
is more marked in this gender. Due to variable
X inactivation patterns, the full fragile X mutation
in females can be associated with variable phenotypic
expression ranging from the full FXS phenotype to
persons with normal cognition with little, if any, clear
clinical impact from the mutation. The clinical
phenotype of full mutation FXS is frequently marked
by physical features including a long, narrow face,
high arched palate, enlarged ears, joint laxity, and
macroorchidism [44]. Behaviorally, persons with FXS
frequently suffer from attention problems, hyperactiv-
ity, aggression, self-injury, and anxiety [45].
FXS has significant overlap with ASDs. While
only approximately 2% of persons with autism will
have FXS [46], ASDs impact the majority of males
with FXS with about 1 in 3 males meeting criteria
for autism and an additional 1 in 3 meeting dia-
gnosis of Pervasive Developmental Disorder Not
Otherwise Specified (PDD-NOS) criteria [47–50].
The rate of ASDs in females with full mutation
FXS is assumed to be much lower given the overall
phenotypic variability of this population. Given the
Chapter 30: Developmental disorders
significant overlap between FXS and ASDs, under-
standing the molecular basis of FXS may contribute
significantly to the understanding of the neurobiology
of ASDs.
Following the discovery of the FMR1 gene in 1991,
a large body of research has demonstrated that FMRP
is an RNA binding protein associated with actively
translating dendritic ribosomes [51]. This finding has
led to the hypothesis that FMRP may play a role in
activation-mediated protein synthesis, and thus may
be a regulator of synaptic plasticity [52]. Further work
in understanding the function of FMRP has focused
on the relationship between FMRP and metabotropic
glutamate (mGluR) receptor activity. Synthesis of
FMRP is enhanced after group 1 mGluR (includes
mGluR1 and mGluR5) receptor activation [53].
Excess activation of mGluR5 receptors has been asso-
ciated with a form of protein synthesis-dependent
long-term depression (LTD) [54], a phenomenon
enhanced in the hippocampus of FMR1 knockout
mice [55]. Activation of mGluR-mediated LTD
results in weakening of synaptic connections leading
to internalization of AMPA receptors and what
appear to be structurally immature, elongated den-
dritic processes [56]. Weak, elongated, and immature
synaptic connections which would be expected with
enhanced mGluR-mediated LTD have been docu-
mented in the FMR1 knockout mouse and in FXS
post-mortem brain tissue [52]. These findings have
led to the mGluR theory of FXS which postulates that
FMRP normally acts as a brake on mGluR neuro-
transmission and that excessive mGluR activation
(specifically mGluR5) leads to the FXS phenotype
including increased risk for seizures, hypersensitivity
to tactile stimuli, intellectual disability, and hyper-
activity among other characteristics [56].
The treatment implications of the mGluR theory
of FXS have been evaluated to date in animal models
and have recently entered human study. Administra-
tion of the mGluR5 antagonist MPEP has been
reported to reverse phenotypic characteristics associ-
ated with the drosophila and mouse models of FXS
[57, 58]. Fenobam, a nonbenzodiazepine anxiolytic
with demonstrated mGluR5 antagonist activity [59],
is the first mGluR5 antagonist to be studied in
humans with FXS [60]. Fenobam was recently studied
in six males and six females with full mutation FXS in
a single-dose design. Reportedly, 4 (67%) males and 2
(33%) females had a response to treatment and the
drug was generally well tolerated [60]. Currently, a
367
 Chapter 30: Developmental disorders

Table 30.1 Characteristics of selected developmental disorders.

Disorder Genetic Incidence Overlap Physical Medical Characteristic

findings with PDDs appearance comorbidities behavior
Down Trisomy 21 1 in 1000 7–15% meet Flat facies, Leukemia, thyroid Majority exhibit
syndrome live births PDD criteria slanted dysfunction, early strong social
palpebral dementia, cardiac skills for level of
fissures, small anomalies cognitive
ears, single functioning
palmar crease
Rett’s Mutation at 1 in A PDD by Small head Seizure disorder Stereotypic
disorder Xq28 in gene 10 000 definition hand
coding for females movements
MECP2 (classic hand
ringing or
hand washing)
Prader–Willi Paternally 1 in 1 in 4 Obese Hypotonia Obsessive–
syndrome derived gene 10 000– compulsive
deletion within 15 000 behavior (skin
15q11–13 births picking)
Angelman Maternally 1 in 42–81% Frequent Seizure disorder, Frequent
syndrome derived gene 10 000– meet PDD smiling, severe sleep laughter,
deletion within 40 000 criteria microcephaly disturbance hyperactive,
15q11–13 births short attention
span
Smith– Gene deletion 1 in 80–100% Short stature, Severe sleep Self-hugging,
Magenis at 17p11.2 25 000 meet PDD obesity, disturbance, severe self-
syndrome containing RAI1 births criteria prognathism, hyperlipidemia, injury
gene short/teneted renal anomolies,
philtrum hearing loss,
hypotonia, thyroid
dysfunction
Fragile CGG repeat 1 in 4000– 2 in 3 meet Long face, Otitis media, seizure Gaze aversion,
X syndrome expansion in 6000 live PDD criteria large ears, disorder inattention/
FMR1 gene on births large testicles, hyperactivity,
long arm of high arched anxiety,
X chromosome palate aggression/
self-injury
FMR1, fragile X mental retardation gene; MECP, methyl-CpG-binding protein 2; PDD, pervasive developmental disorder; RAI1, retinoic acid
induced 1 gene.

proprietary mGluR5 antagonist from Novartis is
under study in subjects with FXS in Europe [61].
FXS stands as an example of a genetic finding
leading to neurobiological understanding with further
progress linking neurobiology to phenotypic expres-
sion. This translational approach, resulting in a new
targeted treatment strategy is a major goal of research
in all neuropsychiatric disorders. The status of FXS as
a single gene disorder has made this progression
feasible. While other developmental disorders such
as autism have a much more complex and heteroge-
neous etiology, lessons can be learned from FXS when
investigating the genetic underpinnings and resulting
neurobiology of developmental disorders.
The phenotypic overlap between FXS and autism
gives investigators a solid foundation upon which to
build investigation into autism susceptibility. Future
work may start with thorough genetic analysis of

368
 Chapter 30: Developmental disorders

persons with FXS with and without autistic disorder.
All subjects with FXS appear to suffer from a lack of
FMRP; the research community has not yet identified
additional factors that may lead to some persons with
FXS expressing classic autism, while others maintain
a modicum of social ability. Increased understanding
of glutamatergic neurotransmission on the basis of
findings in FXS may also hold promise for future
study of the neurobiology and treatment of ASDs.
Conclusion
A review of the genetics of developmental disorders
highlights several concepts in psychiatric gene-
tics including gene dosage, genetic heterogeneity
(different mutations leading to the same/similar
phenotypes), imprinting, epigenetics, X inactivation
patterns, and anticipation. Interestingly, among the
disorders discussed (Table 30.1), while clear genetic
differences exist, there is phenotypic overlap between
many of the disorders. This overlap gives researchers
reason to use findings from the better characterized
genetic disorders highlighted in this chapter as a
potential key to unlock further understanding of
seemingly more complex idiopathic disorders
(ASDs).
Acknowledgements
This work was supported, in part, by research grants
from the National Institute of Mental Health (NIMH)
including U10 MH066766 (McDougle), R01
MH072964 (McDougle); by a FRAXA Research
Foundation Grant (McDougle & Erickson); by an
Indiana Clinical Research Center grant UL1
RR025761 from the National Institutes of Health
(NIH) to Indiana University; by a Indiana University
Clinical and Translational Sciences Institute (CTSI)
Career Development Award KL2 (Erickson); and by a
Indiana Bureau of Developmental Disabilities
Services Fellowship in Autism and Developmental
Disorders (Peay).

References 12. American Psychiatric Association.

1. Feinstein C, et al. Child Adolesc
Psychiatr Clin N Am 2007;16
(3):631–647.
2. Jones KL. Smith’s Recognizable
Patterns of Human Malformation.
Philadelphia: W. B. Saunders
Company; 1988.
3. Schieve LA, et al. Pediatrics
2009;123(2):e253–260.
4. Lowenthal R, et al. J Autism
Dev Disord 2007;37(7):
1394–1395.
5. Kent L, et al. Dev Med Child
Neurol 1999;41(3):153–158.
6. Cohen D, et al. J Autism Dev
Disord 2005;35(1):103–116.
7. Dykens EM. Ment Retard Dev
Disabil Res Rev 2007;
13(3):272–278.
8. Prandini P, et al. Am J Hum Genet
2007;81(2):252–263.
9. Leonard H, et al. Eur Child
Adolesc Psychiatry 1997;6 S1:8–10.
10. Amir RE, et al. Nat Genet 1999;
23(2):185–188.
11. Wan M, et al. Am J Hum Genet
1999;65(6):1520–1529.
Diagnostic and Statistical Manual
of Mental Disorders, Fourth
Edition, Text Revision (DSM-IV-
TR). Washington, DC: American
Psychiatric Association; 2000.
13. Robertson L, et al. Am J Med
Genet B Neuropsychiatr
Genet 2006;141B(2):
177–183.
14. Naidu S, et al. J Child Neurol
2003;18(10):662–668.
15. Shahbazian MD, et al. Hum Mol
Genet 2002;11(2):115–124.
16. Jung BP, et al. J Neurobiol 2003;
55(1):86–96.
17. Cohen DR, et al. Mol Cell Neurosci
2003;22(4):417–429.
18. Venkitaramani DV, et al. Child
Adolesc Psychiatr Clin N Am
2007;16(3):541–556.
19. Cassidy SB. J Med Genet 1997;
34(11):917–923.
20. Dykens EM, et al. J Child Psychol
Psychiatry 1996;37(8):995–1002.
21. Dykens EM, et al. Am J Med Genet
1995;60(6):546–549.
22. Veltman MW, et al. Psychiatr
Genet 2005;15(4):243–254.
23. Milner KM, et al. J Child
Psychol Psychiatry 2005;
46(10):1089–1096.
24. Buckley RH, et al. Am J Med Genet
1998;80(4):385–390.
25. Williams CA, et al. Am J Med
Genet 2001;101(1):59–64.
26. Wu Y, et al. Proc Natl Acad Sci
U S A 2008;105(34):
12399–12404.
27. Heck DH, et al. Hum Mol Genet
2008;17(14):2181–2189.
28. Clayton-Smith J, et al. J Med
Genet 2003;40(2):87–95.
29. Williams CA, et al. Am J Med
Genet 1995;56(2):237–238.
30. Horsler K, et al. J Intellect Disabil
Res 2006;50(Pt 1):33–53.
31. Trillingsgaard A, et al. Autism
2004;8(2):163–174.
32. Peters SU, et al. Clin Genet
2004;66(6):530–536.
33. Girirajan S, et al. Clin Genet
2009;75(4):364–374.
34. Girirajan S, et al. Genet Med
2006;8(7):417–427.
35. Greenberg F, et al. Am J Med
Genet 1996;62(3):247–254.

369
 Chapter 30: Developmental disorders
36. Di Cicco M, et al. Int J Pediatr
Otorhinolaryngol 2001;59(2):
147–150.
37. Sweeney E, et al. J Med Genet
1999;36(6):501–502.
38. Dykens EM, et al. J Autism Dev
Disord 1997;27(2):203–211.
39. Hicks M, et al. J Dev Behav Pediatr
2008;29(1):42–46.
40. Greenberg F, et al. Am J Hum
Genet 1991;49(6):1207–1218.
41. Dykens EM, et al. J Intellect
Disabil Res 1998;42 (Pt 6):
481–489.
42. Gerard B, et al. Ann Genet 1997;
40(3):139–144.
43. Florencia G, et al. J Biochem Mol
Biol 2006;39(6):766–773.
44. Erickson CA, et al. Curr
Psychiatry 2006;5(10):80–92.
370
45. Bailey DB, Jr., et al. Am J Med
Genet A 2008;146A(6):720–729.
46. Kielinen M, et al. Autism 2004;8
(1):49–60.
47. Clifford S, et al. J Autism Dev
Disord 2007;37(4):738–747.
48. Garcia-Nonell C, et al. Am
J Med Genet A 2008;
146A(15):1911–1916.
49. Bailey DB, Jr., et al. J Autism Dev
Disord 2001;31(2):165–174.
50. Rogers SJ, et al. J Dev Behav
Pediatr 2001;22(6):409–417.
51. Jin P, et al. Trends Biochem Sci
2003;28(3):152–158.
52. Beckel-Mitchener A, et al. Ment
Retard Dev Disabil Res Rev
2004;10(1):53–59.
53. Weiler IJ, et al. Proc Natl Acad Sci
U S A 1997;94(10):5395–5400.
54. Huber KM, et al. J Neurophysiol
2001;86(1):321–325.
55. Huber KM, et al. Proc Natl
Acad Sci U S A 2002;
99(11):7746–7750.
56. Bear MF, et al. Trends Neurosci
2004;27(7):370–377.
57. McBride SM, et al. Neuron
2005;45(5):753–764.
58. Yan QJ, et al. Neuropharmacology
2005;49(7):1053–1066.
59. Porter RH, et al. J Pharmacol Exp
Ther 2005;315(2):711–721.
60. FRAXA.org. First Fenobam Trial
Results are Positive. 2008; http://
www.fraxa.org/, accessed April 6,
2012.
61. ClinicalTrials.gov. http://www.
clinicaltrials.gov. Accessed April 6,
2012.
 Alzheimer’s disease
Chapter
31 Carlos Cruchaga, John S. K. Kauwe, and Alison M. Goate
In 2006 there were an estimated 26 million people
worldwide with Alzheimer’s disease (AD) [1]. The
most recent studies of AD predict a rapid increase
in prevalence, projecting that by 2050 there will be a
nearly four-fold increase in AD cases worldwide [1].
This complex neurodegenerative disorder is charac-
terized by gradual and progressive memory loss as
well as deficits in executive functioning, language,
visuo-spatial abilities, personality, behavior, and self-
care. Individuals with AD live from 8 to 20 plus years
after the onset of symptoms. In 2007, the US national
costs of caring for individuals with AD were calcu-
lated to be in excess of 148 billion dollars [2]. Just a
five-year delay in the onset of disease could result
in half as many AD cases in just one generation [3].
The identification of genetic risk factors for AD can
elucidate novel disease-related biological pathways
and drug targets, making it possible to develop new
approaches to prevention and treatment.
Standard diagnostic criteria have been developed
by The National Institute of Neurological and Com-
municative Disorders and Stroke-Alzheimer’s Disease
and Related Disorders Association (NINCDS-
ADRDA) [4]. AD can be divided into two categories
based on age of onset and familial aggregation, famil-
ial (FAD), and late onset (LOAD). Cases with evi-
dence of Mendelian inheritance (autosomal
dominant) and early onset (< 60 years) are categor-
ized as FAD. Less than 1% of all AD cases fall into this
category [5–7]. LOAD is characterized by onset after
approximately 60 years of age and complex patterns
of inheritance. FAD and LOAD share the same clin-
ical and pathological features. The pathological
changes include neuronal loss, beta-amyloid (Aβ)
plaques and neurofibrillary tangles (NFTs)
(Figure 31.1). Plaques are extra-cellular deposits of
insoluble proteins composed mainly of Aβ peptides,
Principles of Psychiatric Genetics, eds John I. Nurnberger, Jr. and Wade H. Berrettini. Published by Cambridge University
Press. © Cambridge University Press 2012.
which are derived from the β-amyloid precursor pro-
tein (APP) [8]. NFTs are intracellular deposits of
hyper-phosphorylated tau protein.
Genetic epidemiology
Familial aggregation of AD was clearly described in
1934 [9]. Unfortunately, the late onset of AD can
make family studies difficult. In most families direct
examination of parents is impossible. Older siblings
may already be dead, while younger siblings and
children may not have reached the risk period for
AD. Despite these difficulties, risk for first-degree
relatives has been reported to be 10–40% greater than
in unrelated individuals [10–13]. Sibling relative risk
ratios are estimated to be between 4 and 5 [11, 14, 15].
Twin studies indicate the presence of a genetic com-
ponent to disease risk, with monozygotic twins show-
ing greater concordance (0.49) than dizygotic twins
(0.18) [16, 17].
Genetic risk factors for familial Alzheimer’s
disease
Mutations in three genes, APP, presenilin 1 (PSEN1),
and presenilin 2 (PSEN2), cause FAD by affecting Aβ
levels [18–20] (Figure 31.1). APP is encoded by 18
exons spanning 290 kb on chromosome 21 [21]. Each
of the seven known transcripts encodes a multi-
domain protein with one transmembrane domain
[22]. APP is expressed in all tissues. The major tran-
script in most cell types is longer (APP770) than the
transcript in neurons (APP695). APP can be proteo-
lytically processed via one of two pathways. In most
cell types, APP is cleaved first by a-secretase, then by
g-secretase. In neurons APP is cleaved by β-secretase
and g-secretase resulting in a variety of Aβ species
varying in length from 37–43 amino acids. The most
371
 Chapter 31: Alzheimer’s disease

a-secretase
PSEN1
b-secretase g-secretase
-
PSEN2
APP

Ab
Ab degradation
APOE
SORL1

CALHM1
CALHM1 ACE
TF
Synapses impairment Ab Oligomers
Ab40–42 APOE APOE
Plaques

Neuronal death

TF
Oxidative stress Kinases/
& Inflammation Phosphatases
P
P
P
MAPT
Neurofibrillary P P P PP
P
tangles GAB2 P Tau P Tau
PPPP
Dementia
Fig 31.1 Role of Alzheimer’s disease (AD) -associated genes in likely pathways of neurodegeneration. Genes implicated in risk for AD are
marked with black-border rectangles. The amyloid precursor protein is encoded by APP. APP gives rise to Aβ through serial cleavage by
a-secretase, g-secretase and β-secretase. Mutations in APP, PSEN1 and PSEN2 (g-secretases) are found in early-onset familial AD. APOE, SORL1,
CALHM1 or TF could be implicated in amyloid precursor protein (APP) processing and risk for disease. APOE may also be involved in
amyloid-beta (Aβ) oligomerization and Aβ plaques formation. On the other hand, it is thought that the angiotensin-converting enzyme (ACE)
pathogenic mechanism is related to the Aβ oligomerization and plaques cleavage. According to the amyloid cascade hypothesis,
Aβ pathology would trigger other AD pathogenic events such as tau deposition. Genetics variants in the MAPT gene and GAB2,
favor the formation of NTF, decreasing age at onset or increasing risk for AD.

common fragment is 40 amino acids in length
(Aβ40); but the major component of amyloid plaques
observed in AD is the 42 amino acid peptide (Aβ42).
The 32 known mutations in APP account for about
10% of FAD cases (http://www.molgen.ua.ac.be/
ADMutations). Most of these mutations occur at
codons near the β and g-secretase cleavage sites in
APP. Mutations near the g-secretase cleavage site
result in an increase in the ratio of Aβ42 to Aβ40
[23, 24]. The “Swedish mutation” (KM670/671N)
occurs at the β-secretase cleavage site and results in
an increase in total Aβ species but does not affect the
Aβ42/Aβ40 ratio. Individuals with this mutation
show pathological evidence of both amyloid plaques
and cerebral amyloid angiopathy (CAA). Duplica-
tions of APP also lead to disease, causing a familial
disorder characterized by hemorrhagic strokes as well
as dementia and pathological evidence of neuritic
plaques and CAA [25]. It appears that elevated levels
of total Aβ lead to both AD and CAA pathology,
while elevation of Aβ42/Aβ40 ratios leads to AD
pathology. Mutations in APP may lead to a spectrum
of clinical phenotypes including both dementia and
hemorrhagic strokes.
After linkage studies provided evidence for a locus
on chromosome 14, Sherrington et al. identified five
missense mutations in PSEN1 that segregated with
FAD in their samples [20]. The PSEN1 contains 12
exons spanning 84 kb on chromosome 14q24.2. The
full-length protein includes nine transmembrane
domains [26–29]. With some exceptions [30], muta-
tions in PSEN1 cause a very early-onset form of the

372

disease, with a mean age at onset (AAO) of 26–60
years [31–33]. To date 182 mutations in PSEN1 have
been identified, accounting for a large proportion of
FAD cases (http://www.molgen.ua.ac.be/ADMuta-
tions). These mutations occur throughout the mol-
ecule but all result in an increased Aβ42/Aβ40 ratio
[31–33].
Levy-Lahad et al. identified a mutation in PSEN2
that segregated with AD in the Volga German kin-
dreds [19]. PSEN2 is homologous to PSEN1 and is
encoded by a gene on 1q42.13. It also has 12 exons but
spans just 25 kb. Presenilins may have other functions
but their role in g-secretase activity is of particular
relevance to AD. PSEN1 and PSEN2 have distinct but
overlapping g-secretase activities [34] and are thought
to be the catalytic core of the g-secretase complex,
which also includes APH1, NCT, and PEN2 [35, 36].
Fourteen FAD mutations in PSEN2 have been identi-
fied. These mutations result in a later AAO (40–75
years) than homologous mutations in PSEN1 and
may exhibit incomplete penetrance [37].
Risk factors for LOAD
While much is known about genetic risk factors for
FAD, the vast majority of AD cases are LOAD. The
single most important known risk factor for LOAD is
age [38–40]. APOE genotype is the strongest genetic
risk factor for LOAD. The APOE epsilon 4 allele
(APOE ε4) has been genotyped in samples of many
racial and ethnic origins, and consistently shows evi-
dence for association with LOAD [41, 42]. The APOE
gene is located in a region of chromosome 19 that has
been identified as a risk region in genetic linkage
studies in LOAD families [43]. It spans less than 4
kb and has 4 exons.
APOE exists in three isoforms (APOE ε2, -ε3, and
-ε4) which differ from each other at two amino acid
positions (codon 112 and codon 158). The APOE ε3
allele (Cys112, Arg158) is the most frequent isoform
in all populations [41]. In most populations the
second most common allele is APOE ε4 (Arg112,
Arg158) and the third allele is known as APOE ε2
(Cys112, Cys158). APOE mediates the binding,
internalization, and catabolism of lipoprotein par-
ticles, and has been implicated in cardiovascular dis-
ease as well as LOAD (for review see [44]).
APOE ε4 shows a dose-dependent increase in risk
for AD. European-Americans who are heterozygous
for the APOE ε4 allele exhibit a three-fold increase in
Chapter 31: Alzheimer’s disease
risk while homozygotes exhibit an eight-fold increase
[45]. Nearly all individuals homozygous for the APOE
ε4 allele will develop LOAD by 80 years of age [46].
It is also clear that the APOE ε2 allele has a protec-
tive effect [45]. APOE ε4 also explains some of the
variance in AAO in families with a known FAD
mutation [47].
Several hypotheses of the mechanism by which
APOE affects risk for AD have been proposed. As is
the case with the known FAD mutations, it appears
that APOE affects risk for LOAD via an Aβ-related
mechanism (Figure 31.1). Patients carrying at least
one APOE ε4 allele have more Aβ plaques than non-
carriers [48]. In vitro, APOE ε4 binds to Aβ with
higher affinity than APOE ε3 [49]. APOE and Aβ
may also compete for clearance through the same
receptor [50]. Mice over-expressing human APP con-
taining an FAD mutation show fibrillar Aβ deposition
only when the mouse apoE gene is expressed: Aβ does
not form amyloid in the absence of apoE [51]. In
humans APOE ε4 may also influence fibril formation
and clearance of Aβ, causing increased Aβ deposition
[52]. Indeed, mice expressing a human APOE ε4 allele
have earlier and more severe pathology than mice
expressing APOE ε3 [53]. The APOE ε2 allele is pro-
tective for AD and appears to have the opposite effect
on amyloid deposition; mouse studies suggest that
although Aβ is deposited it cannot form amyloid in
the presence of APOE ε2 [53]. It is also possible that
additional variation in APOE, such as promoter variants,
may alter risk for LOAD [54].
Disease mechanisms
Our knowledge of the mechanisms by which the
known genetic risk factors for AD alter risk clearly
indicates that amyloid metabolism and deposition is
central to the pathology of AD. There is some dispute
as to whether Aβ plaques are themselves neurotoxic
or whether the toxic species of Aβ may be soluble
oligomers [55]. This evidence is the basis for the
substantial body of research investigating mechan-
isms of Aβ production, clearance, degradation, and
regulation. Drugs targeting g-secretase and several
aspects of plaque formation and clearance are cur-
rently in various stages of clinical trials (for a sum-
mary of drugs in clinical trials see www.alzforum.org/
drg/drc/).
While it does appear that Aβ is central to the
pathology of AD, it is clear that tau is also important.
373
 Chapter 31: Alzheimer’s disease
The microtubule-associated protein tau, is encoded
by a single gene (MAPT, on 17q21) containing 15
exons. Under normal conditions, tau plays a role in
microtubule stabilization and neuronal integrity [56].
Six different isoforms are known to exist in the
human brain and result from alternative splicing of
the MAPT gene. Exon 10 of MAPT can undergo
alternative splicing resulting in a protein with 3 (iso-
form 3R) or 4 (isoform 4R) microtubule-binding
repeats. The ratio of these 2 isoforms is generally 1 :
1 but appears to be disrupted in specific frontotem-
poral dementia (FTD) subtypes involving tau path-
ology and is detectable upon brain autopsy. For
example, Pick’s disease is characterized by an increase
in 3R tau, whereas progressive supranuclear palsy
(PSP), corticobasal degeneration (CBD), and most
cases with MAPT mutations share an overabundance
of 4R tau. In contrast, there is no alteration in the 4R/
3R isoform ratio in AD brains [57]. To date, 44
mutations have been identified in the MAPT gene
(http://www.molgen.ua.ac.be/ADMutations/) causing
FTD. Mutations generally occur within exons 1 and
9–13. Missense and intronic splice site mutations are
most common and involve either disruption of the
normal 4R/3R isoform ratio, altered microtubule
assembly, or promotion of fibril aggregation [58].
However, no mutations in MAPT have been found
to cause AD.
In vitro experiments show that APOE ε3 binds to
tau with a greater affinity than APOE ε4, suggesting
that APOE may also have some effect on NFTs [59].
Recent studies of genetic variation in microtubule-
associated protein tau (MAPT) suggest that changes
in tau expression may alter risk and/or AAO of
LOAD [60–62] (Figure 31.1).
Transgenic mouse studies have also shown that
Aβ pathology is influenced by levels of tau expres-
sion [63]. Several tau-based therapeutics have been
studied extensively, including treatments that work
to inhibit tau aggregation or target tau phosphatases
or kinases. Results of Phase II clinical trials for the
drug RemberTM, reported at the International Con-
ference on Alzheimer’s Disease, were very promis-
ing. The active compound of RemberTM is methyl
thioninium chloride (MTC), commonly known as
methylene blue. The drug developers suggest that
MTC interferes with tau aggregation and even
works to clear existing aggregates. There are likely
to be genetic effects independent of APOE ε4, as this
shows only a modest effect on risk in Amish and
374
Hispanic patients [64, 65] and approximately 60% of
Caucasian AD patients do not carry an APOE
ε4 allele. The discovery of novel genetic risk factors
may provide us with unique insight into disease-
related biological pathways and thus, novel thera-
peutic targets.
The search for novel risk loci
Linkage and candidate gene studies
Linkage studies have led to the identification of many
potential disease genes that are found in regions of
linkage, or positional candidate genes. Studies of the
role of Aβ, tau, and even cholesterol processing in AD
have led to the identification of many biological can-
didate genes. Many of these genes have been studied
using genetic association studies. More than 1000
candidate gene studies examining hundreds of genes
and SNPs have been published. A detailed summary
of these studies and linkage regions can be found on
AlzGene (www.alzgene.org [43]). AlzGene is a “com-
prehensive, unbiased and regularly updated collection
of genetic association studies performed on Alzhei-
mer’s disease phenotypes”. This website provides a
meta-analysis of published data for each SNP. Linkage
and candidate gene studies have not successfully iden-
tified polymorphisms that provide consistent evi-
dence for association with LOAD across multiple
studies [66].
Genome-wide association studies
Genome-wide association studies (GWAS) have had a
profound effect on the search for genetic risk factors
for human disease. To date over 190 GWAS have
been published. Several studies using very large
samples (> 10 000), including exploratory and repli-
cation datasets, have successfully used GWAS to iden-
tify novel genetic risk factors for complex diseases like
diabetes and breast cancer [67–73].
Several groups have reported the results of a
GWAS for LOAD [74, 75]. Li et al. carried out a
GWAS by genotyping approximately 500 000 SNPs
in a series of 753 AD cases and 756 elderly controls
from Canada. An additional 418 AD cases and 249
elderly controls from the United Kingdom were used
for follow-up studies. Only SNPs in high linkage
disequilibrium (LD) with APOE ε4 showed associ-
ation with risk of AD after multiple test correction.
A SNP (rs10519262) located in an intergenic region

on chromosome 15 between ATP8B4 and SLC27A2, was
found to be associated with AAO of AD. Li et al. [76]
suggest that rs7176805, located in the ATP8B4 distal
promoter region and in strong LD with rs10519262,
could be the functional variant. Rs7176805 potentially
modifies a CCAAT box transcription factor binding site
affecting the expression of ATP8B4, which is involved in
phospholipid transport within the cell membrane, with
low levels of expression in hippocampus, caudate, sub-
stantia nigra, and cerebellum.
Coon et al. [74] also genotyped approximately 500
000 SNPs in histopathologically verified AD cases (n ¼
664) and elderly controls (n ¼ 442). Only rs4420638, in
LD with APOE ε4, was associated with AD after mul-
tiple test correction. However, in a subsequent study,
rs2373115, located within the gene encoding GRB-
associated binding protein 2 (GAB2), was associated
with AD (p ¼ 9  10–11; OR ¼ 4.1; CI ¼ 2.8–14.7) in
APOE ε4 carriers [77]. GAB2 is involved in multiple
signaling pathways, and may be related to tau, Aβ,
and other aspects of AD pathology and cell survival
(Figure 31.1). The association between GAB2 and
LOAD in APOE ε4 carriers was independently repli-
cated in a Belgian LOAD case-control series [78].
To obtain results similar to those in other complex
diseases, GWAS of LOAD will require the use of
much larger datasets.
For this reason, several groups, including the
AD Genetics Consortium (AGDC) have combined
existing GWAS data and/or performed large collab-
orative GWAS. Two studies published in 2009, 16
years after the identification of APOE as a risk factor
for LOAD, found 3 new candidate genes that showed
consistent and compelling evidence for association
with risk for LOAD [79, 80]. These three new genes
are Clusterin (CLU, also called APOEJ; chromosome 8),
the gene for the phosphatidylinositol binding clathrin
assembly protein (PICALM; chromosome 11) and the
gene for complement component (3b/4b) receptor 1
(CR1; chromosome 1).
Because pathogenic mutations in APP, PSEN1, and
PSEN2 directly affect the processing of β-amyloid and
isoforms of APOE could also increase risk for AD by
modifying Aβ aggregation and clearance, researchers
have tried to find evidence that links CLU, PICALM,
and CR1 with Aβ. It has been proposed that variants in
CLU could affect risk for AD by promoting the aggre-
gation of Aβ. PICALM, is involved in the intracellular
trafficking of proteins and lipids, and variants in this
gene could influence Aβ levels by affecting trafficking
Chapter 31: Alzheimer’s disease
of APP or the enzymes that cleave APP. Finally, several
observations suggest that pathways involving CR1
products are involved in Alzheimer’s disease pathogen-
esis, particularly in Aβ clearance.
It is important to note that CLU, PICALM, and
CR1 participate in other processes not related to Aβ
fibrillogenesis, processing or clearance, and therefore
studies of the role of these genes in the brain may
reveal evidence for additional disease mechanisms,
which go beyond Aβ accumulation. For example,
two of the identified AD susceptibility genes (CLU,
CR1) have known functions in the immune system,
suggesting a possible role for the immune system in
the risk for AD. It is hoped that the new association
evidence for these genes leads to a better understand-
ing of the pathological processes implicated in AD.
The proportion of AD cases associated with these
risk genes has been calculated to be approximately
25.5% for APOE, 8.9% for CLU, 5.8% for PICALM,
and 3.8% for CR1. In addition, it is clear that none of
these genes is pathogenic by itself. Although these are
only crude estimates it is very likely that other genetic
risk factors remain to be identified. The identification of
such factors may be possible by combining the results
from previous studies (meta-analyses) or by the collab-
oration of several research groups and consortiums in
order to carry out larger studies. It is expected that
additional genetic variants will be identified in future
studies and that these variants will explain only a small
percentage of the risk for AD. The importance of these
studies will not be the identification of specific variants,
with small OR, but the identification of new genes and
pathways implicated in AD that will enable the identifi-
cation of new therapeutic targets.
The most recent technologies used for genotyping
the complete genome also allow the quantification of
gene copy number. Recent data suggest that copy
number variants (CNVs) are surprisingly common;
it is estimated that between 5 and 12% of the genome
could have CNVs [81]. Association between CNVs
and several diseases have already been identified
[82–86]. Rapid progression of technology for the
detection, quantification and analysis of CNVs has
made it possible to begin to understand the extent to
which common CNVs affect risk for human disease.
LOAD endophenotypes
The idea of using endophenotypes in the study of
psychiatric disease was introduced by Gottesman
375
 Chapter 31: Alzheimer’s disease
and Shields in 1973 [87]. They defined endopheno-
types as “internal phenotypes discoverable by a bio-
chemical test or microscopic examination”. An ideal
endophenotype should be a risk “factor” (a factor that
increases risk of developing a disease and plays a role
in the development of the disease) and not solely a
risk “marker” (a marker of incidental changes associ-
ated with disease but not part of the causal pathway)
[88–90]. It should also be both inherited and state-
independent [91, 92]. The use of an endophenotype
for studying the genetics of complex disease may
provide greater power because it is less heterogeneous
than clinical diagnoses and more directly affected by
genetic variation. Endophenotypes may also provide a
biological model of disease and of the possible effects
of the associated genetic variation.
There is consistent evidence that levels of Aβ42 in
cerebrospinal fluid (CSF) correlate with LOAD status
and may be a useful endophenotype for LOAD [17,
93–103]. Furthermore, Fagan et al. showed that CSF
Aβ42 levels vary inversely with Aβ deposition as
measured by positron emission tomography (PET)
using the amyloid imaging tracer Pittsburgh com-
pound B (PIB) [104]. Recent animal studies have also
illustrated the potential importance of Aβ40 levels:
elevated Aβ40 levels in a transgenic model of Aβ
deposition substantially delayed Aβ42 deposition in
these animals [105]. These results and the observa-
tions from FAD mutations suggest that risk for AD
may be increased by either increasing Aβ42 levels or
decreasing Aβ40 levels. Given that the known genetic
risk factors for AD affect Aβ processing, CSF Aβ
levels appear to be useful endophenotypes for genetic
studies of AD risk. CSF tau and tau phosphorylated at
residue 181 (phospho-tau) levels are also useful endo-
phenotypes for AD. CSF tau and phospho-tau levels
are increased in AD patients compared with controls
and phospho-tau levels correlate with the number of
NFTs and the load of hyperphosphorylated tau pre-
sent in brain [106]. APOE ε4 shows replicable associ-
ation with decreased CSF Aβ42 [30, 62, 107–114] as
well as tau and phospho-tau levels [30, 62, 107–114].
Finally, sensitive and specific ELISA tests to measure
Aβ42 as well as tau and phospho-tau181 in the CSF
are readily available [115].
Several recent studies illustrate the potential of
this endophenotype-based approach to identify both
rare and common variants that influence disease. In
2007, Kauwe et al. used CSF Aβ levels to identify an
individual with an FAD mutation in PSEN1 [30].
376
Several studies have also evaluated common variation
for association with CSF Aβ levels. An assortment of
SNPs in SORL1 are associated with risk for AD in
several association studies [116–121]. Association
between these SNPs and CSF Aβ levels has been
observed in one small sample but has not been repli-
cated [113].
In 2003 Kehoe et al. published data suggesting that
SNPs in ACE were associated with CSF Aβ42 [122].
These findings were recently replicated in a large,
independent sample [112]. The study by Kauwe
et al. went on to identify SNPs in BDNF, DAPK1,
and TF that show significant association with LOAD
in AlzGene.org and CSF Aβ levels [112]. Rs4878104
(DAPK1) and rs6265 (BDNF) showed association
with total Aβ levels while rs1800764 (ACE) and
rs1049296 (TF) showed association with CSF Aβ42/
Aβ40 levels. Data showing that cell lines overexpres-
sing a TF cDNA containing the minor allele of
rs1049296 have a significantly higher Aβ42/Aβ40
ratio than those overexpressing wild-type TF cDNA
and are consistent with the CSF associations [112].
TF encodes transferrin, which is the major circulating
glycoprotein involved in iron metabolism and is
highly expressed in the brain. Iron levels are increased
in the brains of AD patients and several reports
suggest that iron misregulation may influence
neurodegeneration by increasing oxidative stress
[123] (Figure 31.1).
It has been proposed that polymorphisms that
modify CSF tau levels may modify the AAO or the
severity of disease [62]. Laws et al. found association
between rs242557 (MAPT) and CSF tau levels [124].
In 2008, Kauwe et al. found SNPs in MAPT that
showed association with both CSF tau and phospho-
tau181. The minor allele of rs3785883 was associated
with higher CSF tau and phospho-tau181 levels in AD
cases and with higher mRNA in AD patients. The
authors suggest that the increase in tau expression
accelerates the formation of tau deposits and as a
result of this, the minor allele carriers have an earlier
AAO than the noncarriers [62]. In a candidate gene
study analyzing SNPs in tau and in 33 other genes
related to tau phosphorylation, dephosphorylation,
and other tau post-translational modifications, Cru-
chaga et al. identified a SNP (rs1868402) associated
with higher CSF phospho-tau181 levels. People with
either one or two copies of the minor allele had higher
levels of CSF phospho-tau181 than those homozygous
for the common allele of this SNP. Cruchaga et al.

followed this up by looking for a connection between
this SNP and disease parameters such as risk, age of
onset, and rate of progression. They found that people
who carried the allele associated with higher CSF
phospho-tau181 levels had a six-fold faster cognitive
decline (as measured by change in the clinical demen-
tia rating per year) than those homozygous for the
common allele. In brains with Aβ pathology, but not
in normal brains, the harmful SNP was also associ-
ated with lower levels of protein phosphatase
B mRNA, and with more neurofibrillary tangles
[125]. The authors of this study suggest that genetic
variants associated with CSF Aβ42 levels may have
higher influence on risk and AAO (e.g. APOE) but
variants associated with CSF phospho-tau181 levels
have a greater impact on rate of progression.
Regular case-control studies are not designed to
identify genetic variants associated with the rate of
progression. For this reason endophenotypes may be
useful to identify such variants. Several studies have
reported that the ratio of CSF tau/Aβ42 is a useful
biomarker to predict progression to Alzheimer’s from
mild cognitive impairment [126–131] and also rate of
progression among AD patients [128]. Large-scale
studies to identify genetic variants associated with
disease progression have not been performed because
it requires longitudinal data collection on a large
number of subjects. The use of the ratio of CSF tau/
Aβ42 as a proxy for disease progression would allow
GWAS in a large sample without the need for longi-
tudinal data. A drawback of this approach is that
there are few large CSF series available for such stud-
ies. The genes and pathways implicated in disease
progression may be different from the genes impli-
cated in risk for disease; by using disease progression
as a phenotype additional genes and pathways may be
identified. Variants associated with disease progres-
sion may be useful in predicting more accurately the
time from diagnosis to functional impairment that
may require nursing home placement. Stratification
of samples by such biomarkers will enable cheaper
and more efficient clinical trials by selecting individ-
uals expected to have faster rates of progression. By
targeting different facets of AD biology, this approach
may identify a broader range of potential therapeutic
targets than a conventional case-control design.
While CSF, Aβ, and tau levels are the current
focus of endophenotype-based approaches, several
other molecules are being studied as potential bio-
markers/endophenotypes for AD. These candidate
Chapter 31: Alzheimer’s disease
molecules include single and multicomponent pro-
teins in the cerebrospinal fluid, blood, and urine as
well as biomarkers based on gene expression (for
review see [132]). Efforts to use proteomics and other
novel methods to identify novel biomarker proteins in
various fluids are also underway [133–136]. These
new endophenotypes may be useful in elucidating
and understanding novel pathways that contribute
to risk for AD.
PET imaging of the amyloid-binding agent, Pitts-
burgh Compound-B (PIB) has also emerged as a
possible endophenotype for AD studies. PIB retention
appears to be a good marker of amyloid deposition in
the brain and may be detected before the presence of
clinical symptoms. There is a strong inverse relation
between in vivo amyloid imaging load and CSF Aβ42
in humans [104, 109, 137, 138]. As larger samples of
individuals with PIB scans are procured these data
may be useful in genetic studies of AD.
To date, studies using this endophenotype-based
approach have been biased toward the study of
existing candidate genes. This can be easily addressed
by the performance of GWAS in samples with endo-
phenotype information. The success of these studies,
even with biased approaches and small sample sizes,
highlights the potential of this endophenotype-based
approach to help us understand the genetic etiology
and pathobiology of LOAD.
Gene expression levels
New high-throughput technologies for measuring
gene expression make it possible to interrogate up to
47 000 transcripts from more than 23 000 genes and
to quantify the different splicing isoforms. Genome-
wide expression profiles have been combined with
complete genome sequence data in order to detect
SNPs that modify the expression of genes (or eQTLs)
implicated in AD pathology. This approach has been
used to show significant association of the H1 haplo-
type in MAPT with increased tau expression [61],
which could increase the risk for AD [60] or modify
the AAO [62].
The genome-wide expression profiles have also
been used to detect genes that show differential
expression in AD patients and identify potential gen-
etic risk factors for AD. The APOE, BACE1, STUB1
(CHIP), FYN, GGA1, and SORL1 genes, have been
identified as differently expressed in AD [139].
A recent study used genome-wide expression data to
377
 Chapter 31: Alzheimer’s disease
identify AD associated variation in a novel candidate
gene. A variant in CALHM1, which codes for a trans-
membrane glycoprotein that controls cytosolic Ca2þ,
was identified by screening genes that showed expres-
sion in the hippocampus and are located in AD
linkage regions. The associated SNP, rs2986017, is a
P86L substitution and confers a partial loss of func-
tion which reduces permeability of CALHM1 for
Ca2þ. The partial loss of function results in lower
levels of cytosolic calcium, increasing Aβ levels by
an APP-dependent mechanism. In the same study
the minor allele frequency of rs2986017 was higher
in AD cases than in age matched controls in five
independent populations [140].
The ability to measure differences in total expres-
sion, splicing, or transcript ratios between AD cases
and controls will further inform our future studies,
making it possible to identify genetic variants which
have direct effects on expression.
Common themes in neurodegenerative
diseases
Loci that cause autosomal dominant forms of many
neurodegenerative diseases also carry common gen-
etic variants that are associated with increased risk for
the sporadic forms of the same disease [141]. For
example, mutations in PRNP have been shown to
segregate with Creutzfeldt–Jakob disease [142] and
polymorphisms upstream of PRNP exon 1 and also
a missense polymorphism (met/val129) are associated
with the sporadic form of Creutzfeldt–Jakob disease
[143]. In Parkinson’s disease (PD), mutations in a-
synuclein and duplications and triplications of a-
synuclein are associated with familial PD [144], and
polymorphisms that alter a-synuclein expression have
been shown to contribute to risk for sporadic forms of
PD [145]. Especially interesting is the case of MAPT.
Tau deposits are found in several neurodegenerative
diseases including AD, FTD, progressive supranuclear
palsy (PSP), and corticobasal degeneration (CDB).
MAPT missense and splicing mutations can cause
familial FTD, but the H1-MAPT haplotype is associ-
ated with sporadic FTD, PSP, CDB, and even AD [60,
146, 147]. Other proteins such as TARDBP (TDP-43)
are also part of pathogenic brain protein deposits
found in several neurodegenerative diseases. TDP-43
is found primarily, in frontotemporal lobar deg-
eneration with ubiquitin-positive, tau- and alpha-
synuclein negative inclusion bodies (FTLD-U), but it
378
can be also present in AD [148, 149] or in
amyotrophic lateral sclerosis [150]. In FTLD-U, the
presence of TDP-43 deposits are associated with
mutation in the GRN gene [151–153]. Most GRN
pathogenic mutations are nonsense, splice-site or
frameshift mutations that generate a loss of GRN
function. It has also been reported that GRN expres-
sion is lower in sporadic AD cases and that the minor
allele of rs5848, located in the 3’ UTR of GRN, is
associated with lower GRN gene expression [154]
and protein levels [155]. These data suggest that vari-
ation in GRN may modify risk for AD. Individuals
with low plasma or serum GRN levels can easily be
detected using an ELISA assay, allowing early
identification of at risk asymptomatic subjects [156].
Conclusion
Progress in understanding the genetic etiology of AD
has been slow. It appears clear that genetic risk for
LOAD is the result of numerous variants with small
effects as well as one variant with a moderate to large
effect (ApoE4). The task of unraveling this complex
genetic architecture is daunting. However, results from
recent GWAS and approaches incorporating the use of
expression data and quantitative endophenotypes are
promising. As scientists come together in the ADGC
and move forward with the application of new tech-
nologies to investigate SNPs, structural variation, and
expression at a genome-wide scale, the future is bright.
Web-site resources
AD and FTD mutation database:
http://www.molgen.ua.ac.be/ADMutations
Alzgene database: http://www.alzforum.org/
Database of genotype and phenotype:
http://www.ncbi.nlm.nih.gov/gap
National Center for Biotechnology information:
http://www.ncbi.nlm.nih.gov
Summary of drugs in clinical trials:
http://www.alzforum.org/drg/drc/
Abbreviations
AAO, age at onset
Aβ, β-amyloid
Aβ40, Aβ of 40 amino acids in length
Aβ42, Aβ of 42 amino acids in length
AD, Alzheimer’s disease
AD LOAD, late onset AD
 Chapter 31: Alzheimer’s disease

AGDC, AD Genetics Consortium MAPT, microtubule-associated protein tau

APOE ε4, APOE epsilon 4 allele MTC, methyl thioninium chloride
APP, β-amyloid precursor protein NFTs, neurofibrillary tangles
dbGaP, genotype and phenotype NINCDS-ADRDA, National Institute of Neurological
CAA, cerebral amyloid angiopathy and Communicative Disorders and Stroke–Alzheimer’s
CDB, corticobasal degeneration Disease and Related Disorders Association
CNVs, copy number variants PD, Parkinson’s disease
CSF, cerebrospinal fluid PET, positron emission tomography
FAD, familial AD Phosphor-tau, tau phosphorylated at residue 181
FTD, frontal temporal dementia PIB, Pittsburgh compound B
GAB2, GRB-associated binding protein 2 PSEN1, presenilin 1
GWAS, genome-wide association studies PSEN2, presenilin 2
LD, linkage disequilibrium PSP, progressive supranuclear palsy

References 16. Bergen A. Clinical Genetics

1. Brookmeyer R, et al. Alzheimers
Dement 2007;3(3):186–191.
2. Alzheimer’s Association.
Alzheimer’s Disease: Facts and
Figures. Chicago: Alzheimer’s
Association; 2007.
3. Brookmeyer R, et al. Am J Public
Health 1998;88(9):1337–1342.
4. McKhann G, et al. Neurology
1984;34(7):939–944.
5. Lopera F, et al. JAMA 1997;277
(10):793–799.
6. Crook R, et al. Nat Med 1998;4
(4):452–455.
7. Ezquerra M, et al. Neurology
1999;52(3):566–570.
8. Selkoe DJ. Annu Rev Neurosci
1994;17:489–517.
9. Lowenberg K. Arch Neurol
Psychiatr 1934;31:737.
10. Mohs RC, et al. Arch Gen
Psychiatry 1987;44(5):
405–408.
11. Breitner JC, et al. Am J Epidemiol
1998;128(3):536–548.
12. Huff FJ, et al. Neurology 1988;38
(5):786–790.
13. Mayeux R, et al. Arch Neurol
1991;48(3):269–273.
14. Sadovnick AD, et al. Genet
Epidemiol 1989;6(5):633–643.
15. Hirst C, et al. Genet Epidemiol
1994;11(4):365–374.
1994;46:144–149.
17. Bergen A, et al. Arch Gen
Psychiatry 1997;54:264–270.
18. Goate A, et al. Nature 1991;349
(6311):704–706.
19. Levy-Lahad E, et al. Science
1995;269(5226):973–977.
20. Sherrington R, et al. Nature
1995;375(6534):754–760.
21. Hattori M, et al. Nature 2000;405
(6784):311–319.
22. Yoshikai S, et al. Gene 1990;87
(2):257–263.
23. Citron M, et al. Nature 1992;360
(6405):672–674.
24. De Strooper B. et al. J Cell Sci
2000;113(Pt 11):1857–1870.
25. Rovelet-Lecrux A, et al. Nat Genet
2006;38(1):24–26.
26. Li X, et al. Neuron 1996;17
(5):1015–1021.
27. Li X, et al. Proc Natl Acad Sci
U S A 1997;94(22):12204–12209.
28. Li X, et al. Proc Natl Acad Sci
U S A 1998;95(12):7109–7114.
29. Laudon H, et al. J Biol Chem
2005;280(42):35352–35360.
30. Kauwe JS, et al. Ann Neurol
2007;61(5):446–453.
31. Xia W, et al. J Biol Chem 1997;272
(12):7977–7982.
32. Gustafson L, et al. Hum Genet
1998;102(3):253–257.
33. Dermaut B, et al. Am J Hum Genet
1999;64(1):290–292.
34. Lai MT, et al. J Biol
Chem 2003;278(25):
22475–22481.
35. Kimberly WT, et al. Proc Natl
Acad Sci U S A 2003;
100(11):6382–6387.
36. Schroeter EH, et al. Proc Natl
Acad Sci U S A 2003;100
(22):13075–13080.
37. Ezquerra M, et al. Arch Neurol
2003;60(8):1149–1151.
38. Jorm AF, et al. Acta Psychiatr
Scand 1987;76(5): 465–479.
39. Rocca WA, et al. Ann Neurol
1991;30(3):381–390.
40. Ritchie K, et al. Lancet 1995;
346(8980):931–934.
41. Finckh U. J Neural Transm
2003;110(3):253–266.
42. Bertram L, et al. The AlzGene
Database. Alzheimer Research
Forum. http://www.alzgene.org.
43. Strittmatter WJ, et al. Proc Natl
Acad Sci U S A 1993;
90(5):1977–1981.
44. Stampfer MJ. J Intern Med
2006;260(3):211–223.
45. Bertram L, et al. Nat Genet
2007;39(1):17–23.
46. Corder EH, et al. Science 1993;
261(5123):921–923.
47. Pastor P, et al. Ann Neurol
2003;54(2):163–169.

379
 Chapter 31: Alzheimer’s disease
48. Schmechel DE, et al. Proc Natl
Acad Sci U S A 1993;
90(20):9649–9653.
49. Strittmatter WJ, et al. Proc Natl
Acad Sci U S A 1993;90(17):
8098–8102.
50. Kounnas MZ, et al. Cell 1995;82
(2):331–340.
51. Bales KR, et al. Nat Genet 1997;
17(3):263–264.
52. Holtzman DM, et al. Ann Neurol
2000;47(6):739–747.
53. Fryer JD, et al. J Neurosci 2003;23
(21):7889–7896.
54. Wang JC, et al. Neurology 2000;55
(11):1644–1649.
55. Cerpa W, et al. Curr Alzheimer
Res 2008;5(3):233–243.
56. See TM, et al. J Geriatr Psychiatry
Neurol 2010;23(4):260–268.
57. Cairns NJ, et al. J Pathol 2004;204
(4):438–449.
58. Gasparini L, et al. Neurogener Dis
2007;4(2–3):236–253.
59. Strittmatter WJ, et al. Proc Natl
Acad Sci U S A 1994;
91(23):11183–11186.
60. Myers AJ, et al. Hum Mol Genet
2005;14(16):2399–2404.
61. Myers AJ, et al. Neurobiol Dis
2007;25(3):561–570.
62. Kauwe JS, et al. Proc Natl Acad Sci
U S A 2008;105:8050–8054.
63. Roberson ED, et al. Science
2007;316(5825):750–754.
64. Pericak-Vance MA, et al. Ann
Neurol 1996;39(6):700–704.
65. Tang MX, et al. JAMA 1998;279
(10):751–755.
66. Bertram L. et al. Hum Mol Genet
2004;13(Spec. No. 1):R135–R141.
67. Easton DF, et al. Nature 2007;
447(7148):1087–1093.
68. Hunter DJ, et al. Nat Genet
2007;39(7):870–874.
69. McPherson R, et al. Science
2007;316(5830):1488–1491.
70. Saxena R, et al. Science 2007;316
(5829):1331–1336.
380
71. Scott LJ, et al. Science 2007;316
(5829):1341–1345.
72. Skol AD, et al. Genet Epidemiol
2007;31(7):776–788.
73. Zeggini E, et al. Science 2007;316
(5829):1336–1341.
74. Coon KD, et al. J Clin Psychiatry
2007;68(4):613–618.
75. Li H, et al. Arch Neurol 2008;
65(1):45–53.
76. Nagase T, et al. DNA Res 2001;8
(6):319–327.
77. Reiman EM, et al. Neuron 2007;
54(5):713–720.
78. Sleegers K, et al. Hum Mutat
2009;30(2):E338–E344.
79. Harold D, et al. Nat Genet 2009;
41(10):1088–1093.
80. Lambert JC, et al. Nat Genet
2009;41(10):1094–1099.
81. Redon R, et al. Nature 2006;
444(7118):444–454.
82. Breunis WB, et al. Blood 2008;
111(3):1029–1038.
83. Cusco I, et al. BMC Med Genet
2008;9:27.
84. Marshall CR, et al. Am J Hum
Genet 2008;82(2):477–488.
85. McCarroll SA, et al. Nat Genet
2008;40(9):1107–1112.
86. Stefansson H, et al. Nature
2008;455(7210):232–236.
87. Gottesman II et al. Br J Psychiatry
1973;122(566):15–30.
88. Kannel WB, et al. Ann Intern Med
1961;55:33–50.
89. Nadella RK. Lancet 1979;1
(8130):1354.
90. Hanefeld M, et al. Przegl Lek
1989;46(7):588–594.
91. Gershon ES, et al. Acta Psychiatr
Scand 1986;74(2):113–118.
92. Gottesman II, et al. Am
J Psychiatry 2003;160(4):636–645.
93. Motter R, et al. Ann Neurol
1995;38(4):643–648.
94. Tamaoka A, et al. J Neurol Sci
1997;148(1):41–45.
95. Galasko D, et al. Arch Neurol
1998;55(7):937–945.
96. Shoji M, et al. J Neurol Sci
1998;158(2):134–140.
97. Mehta PD, et al. Arch Neurol
2000;57(1):100–105.
98. Kawarabayashi T, et al. J Neurosci
2001;21(2):372–381.
99. Strozyk D, et al. Neurology
2003;60(4):652–656.
100. Sunderland T, et al. JAMA
2003;289(16):2094–2103.
101. Hampel H, et al. Mol Psychiatry
2004;9(7):705–710.
102. Jia JP, et al. Neurosci Lett 2005;383
(1–2):12–16.
103. Schoonenboom NS, et al. Ann
Neurol 2005;58(1):139–142.
104. Fagan AM, et al. Ann Neurol
2006;59(3):512–519.
105. Kim J, et al. J Neurosci 2007;27
(3):627–633.
106. Buerger K, et al. Arch Neurol
2002;59(8):1267–1272.
107. Bouwman FH, et al. Neurobiol
Aging 2007;28(7):1070–1074.
108. Brys M, et al. Neurobiol Aging
2009;30(5):682–690.
109. Fagan AM, et al. Arch Neurol
2007;64(3):343–349.
110. Hansson O, et al. Dement Geriatr
Cogn Disord 2007;23(5):316–320.
111. Wiltfang J, et al. J Neurochem
2007;101(4):1053–1059.
112. Kauwe JS, et al. Neurogenetics
2009;10(1):13–17.
113. Kolsch H, et al. Neurosci Lett
2008;440(1):68–71.
114. Smach MA, et al. Neurosci Lett
2008;440(2):145–149.
115. Vanderstichele H, et al. Amyloid
2000;7(4):245–258.
116. Lee JH, et al. Arch Neurol 2007;
64(4):501–506.
117. Li Y, et al. Neurobiol Dis 2008;
29(2):293–296.
118. Liu F, et al. Am J Hum Genet
2007;81(1):17–31.

119. Rogaeva E, et al. Nat Genet
2007;39(2):168–177.
120. Bettens K, et al. Hum Mutat
2008;29(5):769–770.
121. Lee JH, et al. Neurology 2008;
70(11):887–889.
122. Kehoe PG, et al. Hum Mol Genet
2003;12(8):859–867.
123. Shcherbatykh I. et al. J Alzheimers
Dis 2007;11(2):191–205.
124. Laws SM, et al. Mol Psychiatry
2007;12(5):510–517.
125. Cruchaga C, et al. PLoS Genet
2010;6(9):e1001101.
126. Kester MI, et al. Neurology
2009;73(17):1353–1358.
127. Petrie EC, et al. Arch Neurology
2009;66(5):632–637.
128. Snider BJ, et al. Arch Neurol
2009;66(5):638–645.
129. Landau SM, et al. Neurology
2010;75(3):230–238.
130. Wallin AK, et al. Neurology
2010;74(19):1531–1537.
131. Okonkwo OC, et al. Arch Neurol
2011;68(1):113–119.
132. Lonneborg A. Mol Diagn Ther
2008;12(5):307–320.
133. Hu Y, et al. Proteomics Clin App
2007;1:1373–1384.
134. Ray S, et al. Nat Med 2007;13
(11):1359–1362.
135. Sonnen JA, et al. Expert Rev
Neurother 2007;7(8):1021–1028.
136. Craig-Schapiro R, et al. Biol
Psychiatry 2010;68(10):903–912.
137. Fagan AM, et al. Neurobiol Dis
2002;9(3):305–318.
138. Fagan AM, et al. Ann Neurol
2000;48(2):201–210.
139. Liang WS, et al. Physiol Genomics
2008;33(2):240–256.
140. Dreses-Werringloer U, et al. Cell
2008;133(7):1149–1161.
141. Hardy J. Biochem Soc Trans
2005;33(Pt 4):578–581.
142. Owen F, et al. Brain Res Mol Brain
Res 1990;7(3):273–276.
143. Mead S, et al. Am J Hum Genet
2001;69(6):1225–1235.
144. Zarranz JJ, et al. Ann Neurol
2004;55(2):164–173.
Chapter 31: Alzheimer’s disease
145. Chiba-Falek O, et al. Hum Genet
2003;113(5):426–431.
146. Hutton M, et al. Nature 1998;
393(6686):702–705.
147. Pittman AM, et al. J Med Genet
2005;42(11):837–846.
148. Hu WT, et al. Acta Neuropathol
2008;116(2):215–220.
149. Uryu K, et al. J Neuropathol
Exp Neurol 2008;67(6):
555–564.
150. Rutherford NJ, et al. PLoS Genet
2008;4(9):e1000193.
151. Baker M, et al. Nature 2006;
442(7105):916–919.
152. Cruts M, et al. Nature 2006;
442(7105):920–924.
153. Mukherjee O, et al. Ann Neurol
2006;60(3):314–322.
154. Fenoglio C, et al. J Neurology
2008;255:47.
155. Rademakers R, et al. Hum
Mol Genet 2008;17(23):
3631–4642.
156. Ghidoni R, et al. Neurology
2008;71(16):1235–1239.
381
 Index
1000 Genomes Project, 32
AAV (adeno-associated virus)
technology, 73
ABCA13, and schizophrenia, 243
ABCB1 (P-glycoprotein gene),
variants, and methadone
dose, 302
Ab (amyloid beta), and Alzheimer’s
disease, 351–353
absolute risk, 8
ACE, and Alzheimer’s disease, 376
acetylation, histones, 79–81, 88
ACN9, and alcoholism, 284
ACTH see adrenocorticotropic
hormone (ACTH)
actus reus, 327
AD see Alzheimer’s disease (AD)
addictions
testing issues, 331
vulnerability, 303
see also alcohol dependence;
alcoholism; drug addiction; nicotine
dependence
adeno-associated virus (AAV)
technology, 73
adenosine receptor 2A
(ADORA2A), 95
adenosine system, genes
involved in, 95
ADH
and alcoholism, 281, 283–285
arrangement, 281
ADH (alcohol dehydrogenases), 280
ADHD see attention-deficit
hyperactivity disorder
(ADHD)
admixture mapping, 19
adoption studies
antisocial behavior, 145–146, 156
gene–environment
interactions, 152
antisocial personality disorder,
148, 317
attention-deficit hyperactivity
disorder, 149–150, 168
and disease etiology, 2–3
and disorder transmission studies, 3
genetic epidemiology, 2–3
382
major depressive disorder, 215
mood disorders, 4–5
post-traumatic stress disorder, 136
ADORA2A (adenosine
receptor 2A), 95
ADRA2A, and obesity, 276
ADRB3, and obesity, 272–273
adrenocorticotropic hormone
(ACTH)
and opioid addiction, 300
roles, 300
affected sibling pair tests, 18
affective disorders, and schizophrenia,
233–234
Affymertrix GeneChip Mapping
Arrays, 25
aggression, 145
agoraphobia (AG)
classification, 112
diagnosis, 112
familial transmission, 113
genomic studies, 118
and panic disorder, 118
prevalence, 90, 112–113
AGP see Autism Genome
Project (AGP)
agranulocytosis, antipsychotic-
induced, 59
AGRE see Autism Genetic Resource
Exchange (AGRE)
AKT1, and schizophrenia, 247
alcohol, 280
alcohol dehydrogenases (ADH), 280
alcohol dependence
diagnostic criteria, 279
genetics, 295
linkage plots, 355
and nicotine dependence, 294
see also alcoholism
alcoholism
candidate genes, 280
and DNA methylation, 298–299
endophenotypes, 352
family studies, 282–283
follow-up linkage studies, 283–284
gene–environment interactions, 284
in treatment, 284
genetic linkage studies, 282–283
genetics, 279
environmental factors, 279–280
future research, 284–285
genome-wide association
studies, 284
heritability, 5–6
neurocognitive phenotypes,
353–359
and opioid system, 282
risk factors, 279–280
twin studies, 279–280
see also alcohol dependence
alcohol use disorders (AUDs), 279
aldehyde dehydrogenases
(ALDH), 280
ALDH2, variations, 280–281
ALDH2*2, and alcoholism, 280–281,
284
ALDH, and alcoholism, 281, 284–285
ALDH (aldehyde
dehydrogenases), 280
AliBaba2, 40
alignments, 41–43
categorization, 42
global, 42
local, 42–43
motif-based, 42
multiple, 42
pairwise, 42
sequence, 31, 41–43
structural, 43
alleles, 8–9
allelic association, concept of, 231
Allen Brain Atlas, 44
AlzGene, 37–38
Alzheimer’s disease (AD)
and amyloid beta, 351–353
candidate genes, 372, 374
categorization, 371
databases, 37–38
deterministic alleles, 8
diagnostic criteria, 371
disease mechanisms, 373–374
future trends, 378
gene expression, 377–378
genetic epidemiology, 371–373
genetics, 371–378
genome-wide association studies,
374–375
linkage studies, 374
 Index

murine studies, 374 diagnosis, 262 studies, 152–156

prevalence, 371
risk loci, 374–378
web site resources, 378
see also familial Alzheimer’s
disease (FAD); late-onset
Alzheimer’s disease (LOAD)
American Civil War, post-traumatic
stress disorder, 134
amiodarone, drug interactions,
53–54
amphetamines
attention-deficit hyperactivity
disorder studies, 176, 179
epigenetic mechanisms, 83–85
metabolism, 176
amphetamine-type stimulants (ATS),
abuse, 306
amyloid beta (Ab), and Alzheimer’s
disease, 351–353
AN see anorexia nervosa (AN)
Angelman syndrome, 365–366
animal models
advantages, 72–73
attention-deficit hyperactivity
disorder, 173, 176–179
pharmacogenetics, 179
behavioral assays, 75, 76
depression studies, 85
drug-addiction studies, 83–85
epigenetic mechanisms, 88
gene–environment interactions,
178–179
gene expression studies, 49
genetic susceptibility factor studies,
72–77
grooming disorders, 128
hemizygous deletions, 178
knockdown, 178
knockout, 176–177
locomotor activity changes, 177
noradrenergic pathway, 177
obsessive–compulsive
disorder, 128
post-traumatic stress
disorder, 141
psychiatric disorders, 73
schizophrenia, 73, 76, 251–253
selected inbred strains, 178
transgenic, 178
animal phobias, sex differences,
115
ANK3, and bipolar disorder,
205–206
ANKK1, and alcoholism, 281–282
anorexia nervosa (AN)
candidate gene association studies,
266–267
future research, 267–270
diagnostic criteria, 263
and exercise, 263
family studies, 264–267
genetics, 262–270
heritability, 264
linkage studies, 265–266
malnutrition and, 263–264
and mood disorders, 263
morbidity, 262
neurocognitive deficits, 263
overview, 262
prevalence, 262
recovery, 264
risk factors, 264
and setshifting, 263
state–trait characteristics, 263–264
symptoms, 262
twin studies, 264–265
anterior cingulate, roles, 102
anterior frontal lobes, functions,
192–193
anterior temporal lobes, functions,
192–193
anticipation
and bipolar disorder, 198
and schizophrenia, 230–231
anticonvulsants, 59–61
antidepressants
adverse reactions, pharmacogenetic
studies, 56
drug interactions, 53–54
genome-wide association
studies, 225
pharmacodynamics, genetic effects
on, 54
pharmacogenetics, 53–57, 225
pharmacokinetics, genetic effects
on, 53–54
see also tricyclic antidepressants
(TCAs)
antipsychotic drugs, 58–59
antisocial behavior
adoption studies, 145, 156
and attention-deficit hyperactivity
disorder, 145, 149
common pathways model, 147
comorbidities, 145, 157
DSM-IV disorders related to,
heritability, 148–150
endophenotypes, 154–156
environmental factors, 156
factors affecting, 156
family studies, 145, 153–154
future research, 157
gene–environment interactions,
145–157
adoption studies, 152
approaches, 152–153
twin studies, 152
gene identification, 153–154
genetics, 145–146
externalizing factors, 146–148
and heart rate, 155–156
heritability, 146
interventions, 156–157
measured risk factors, 154–156
molecular genetic studies, 152
and psychiatric disorders, 145
and substance use disorders, 150
symptoms, 148
treatment, 157
twin studies, 145–148, 156
types of, 145
use of term, 145
variance, 147
antisocial personality disorder, 145
adoption studies, 148, 317
family studies, 317
heritability, 148
prevalence, 148
sex differences, 317–318
symptoms, 148
twin studies, 317
anxiety disorders
comorbidities, 115–116
genetic epidemiology, 5
genetic variance, 116
and Tourette syndrome, 338
see also generalized anxiety
disorder (GAD);
obsessive–compulsive
disorder (OCD)
APOE E2 (apolipoprotein-E E2) allele,
139–140
APOE E4 (apolipoprotein-E E4) allele,
8
APOE
and Alzheimer’s disease, 374
and late onset Alzheimer’s
disease, 373
and major depressive disorder,
218–219
apolipoprotein-E E2 allele, 139–140
apolipoprotein-E E4 allele, 8
APP, and Alzheimer’s disease, 375
Applied Biosystems
SNPplex assays, 25
SOLiD, 28, 29
TaqMan assays, 25
arginine vasopressin (AVP), 95
Ariadne Genomics Pathway
Studio, 46
ArrayExpress, 44
artificial selection, 324
ASDs see autism spectrum disorders
(ASDs)

383
 Index
association studies
major depressive disorder, 217–218
psychostimulant dependence,
308–310
schizophrenia, 238
assortative mating, bipolar disorder
and, 198
atomoxetine, attention-deficit
hyperactivity disorder
studies, 179
ATP2C2, and language impairment,
164
ATP6V1B2, and major depression, 221
ATS (amphetamine-type stimulants),
abuse, 306
attention-deficit hyperactivity disorder
(ADHD), 148, 168–179
adoption studies, 149–150, 168
animal models, 173, 176–179
pharmacogenetics, 179
and antisocial behavior, 145, 149
and autism, 193
and bipolar disorder, 208
brain imaging, 176
candidate genes, 163, 169–173
catecholaminergic pathways, 172
cholinergic pathways, 172
comorbidities, 126–127, 175
as complex phenotype, 173–176
dopaminergic pathways, 169–171
endophenotypes, neurocognitive,
350–351
family studies, 168
gene–environment interactions, 173
genetic risk factors, 179
genome-wide association studies, 173
genome-wide linkage studies,
168–169, 171
glutaminergic neurotransmission,
172
heritability, 149–150, 168, 169
heterogeneity, 173–175
neurocognitive tests, 175
neuroimaging studies, 176
pharmacogenetics, 176
prevalence, 149, 168
serotonergic pathways, 172
structural variants, 173
symptoms, 149, 168
and Tourette syndrome, 338
twin studies, 149–150, 168, 176
attributable risk, 8
AUDs see alcohol use disorders
(AUDs)
autism, 183–193
and attention-deficit hyperactivity
disorder, 193
candidate genes, 189–190, 192
comorbidities, 192
384
copy number variations in, 191–192
diagnostic criteria, 183
etiology, 183
biochemical, 192–193
family studies, 186–187
genetics, 183
genetic studies, historical
background, 184
genetic variants, 191–192
genome-wide association studies,
190–191
genome-wide linkage studies,
186–187
heterogeneity, 187
Ingenuity network analyses, 192
linkage studies
historical background, 183–188
pathways, 192–193
and Prader–Willi syndrome
compared, 365
sex differences, 188, 191
twin studies, 183
Autism Genetic Resource Exchange
(AGRE), 186–187
endophenotype studies, 188
genome-wide association studies,
190–191
linkage studies, 187
Autism Genome Project (AGP)
genome-wide associations
studies, 191
linkage studies, 187
studies, 186
autism spectrum disorders (ASDs),
183–193
endophenotypes in, 188–189
and fragile X syndrome
compared, 367
genetics, 183
loci, 186
autonomic dysregulation
model, 101
avoidant personality disorder, and
social phobia compared, 117
AVP (arginine vasopressin), 95
AVPR1A, and autism, 190
AVPR1B, and panic disorder, 95
azacitidine, 299
balanced selection, 202
BAM (binary SAM), 31
BDNF, 85–86
and cocaine administration, 83–84
and histone acetylation, 87–88
and suicidal ideation, 56–57
BDNF see brain-derived neurotrophic
factor (BDNF)
behavior
and environmental factors, 327
suicidal, 353
see also antisocial behavior
behavioral assays, animal
models, 75
behavioral disorders, anticonvulsant
therapy, 59
behavioral endophenotypes, 188
behavioral genetics
and criminal responsibility,
327–329
developments, 324
ethical issues, 324–334
and law, ethical issues, 326–329
and medicine, ethical issues,
326–329
meta-analyses, 146
and punishment, 327–329
Bell, Buck v., 325
benzodiazepines, 59
panic disorder treatment,
95–96
binary SAM (BAM), 31
BIND (Biomolecular Interaction
Network Databank), 45
BIOBASE, 40
biochemical bases, for autism,
192–193
BioGPS, 44
bioinformatics, 44
BiologicalNetworks, 46
biomarkers, 49–50
see also genetic markers
Biomolecular Interaction Network
Databank (BIND), 45
Biomolecular Object Network
Databank (BOND), 45
BioNetSQL, 46
BioPAX, 45
bipolar disorder, 4
and anticipation, 198
and assortative mating, 198
and attention-deficit hyperactivity
disorder, 208
candidate genes, 202–204
comorbidities, 207–208
familial transmission, 207
family epidemiology, 196–199
family studies, 196
future research, 208
gene mapping methods, 199–200
genetic counseling, 208
genetics, 209
genetic susceptibility factors, 69
genome-wide association studies,
200, 205–206
linkage studies, 200–202
lithium treatment, 59–60
and migraine, 208
pharmacotherapy, 57

and schizophrenia, 230–231,
233–234, 208
copy number variant differences,
250–251
genetic association differences,
250–251
subphenotypes, 199
genetics, 198–199
mapping genes, 207
transmission mode, 196–198
twin studies, 196
blepharospasm, 127
BMI see body mass index (BMI)
BMP7, and major depressive
disorder, 225
BN see bulimia nervosa (BN)
body mass index (BMI)
heritability, 272
variance, 275
BOND (Biomolecular Object Network
Databank), 45
brain
adolescent, dynamic changes, 75
gene expression in, research
issues, 87
human postmortem, gene
expression profiling, 49
structural changes, and panic
disorder, 103
brain abnormalities, and panic
disorder, 102
brain alterations, gene-induced, 156
BrainArray, 36
brain-derived neurotrophic factor
(BDNF), 56–57
genes, 140, 201–202
roles, 85–86, 203–204, 219
brain dysfunction, and genetic
susceptibility factors, 77
brain function, regulatory
mechanisms, 82
brain imaging
attention-deficit hyperactivity
disorder, 176
humans, genetic susceptibility factor
studies, 77
schizophrenia studies, 234
brain maturation, postnatal, 75
brain oscillations
as endophenotypes, 353–359
brain responses, to fear, 103–104
BRD1, and schizophrenia, 247–248
BRITE, 45
Buck v. Bell, 325
bulimia nervosa (BN)
diagnosis, 262
diagnostic criteria, 263
family studies, 264, 265
genetics, 262–270
heritability, 264
linkage studies, 265–266
malnutrition and, 263–264
and mood disorders, 263
overview, 262
prevalence, 262
risk factors, 264
state–trait characteristics,
263–264
symptoms, 262, 263
twin studies, 264–265
C2ORF3, and reading disability, 163
CACNA1C, and bipolar disorder,
205–206
CALHM1, and Alzheimer’s disease,
377–378
candidate gene association
studies, 28
anorexia nervosa, 266–269
post-traumatic stress disorder, 137,
138
psychiatric disorders, 49
candidate genes
alcoholism, 280
Alzheimer’s disease, 372, 374
attention-deficit hyperactivity
disorder, 163, 169–173
autism, 189–190, 192
bipolar disorder, 202–204
cocaine dependence, 309–310
language impairment, 164–165
major depressive disorder, 218–219
obesity, 273, 277
obsessive–compulsive disorder, 121,
123–126
personality disorders, 321
reading disability, 161–163
schizophrenia, 204, 231, 238–248
animal and cell biology studies,
251–253
functional polymorphisms and
mutations, 251
stimulant dependence, 312
Tourette syndrome, 341
types of, 217–218
candidate gene studies, versus
genome-wide studies, 18
cannabinoid receptor 1 gene (CNR1)
see CNR1
cannabinoid-related genes, and
stimulant dependence, 312
carbamazepine (CBZ), 59
adverse reactions, 60–61
applications, 60–61
Stevens–Johnson syndrome
induction,
pharmacogenetics, 61
carbon dioxide hypersensitivity, 96
Index
catecholaminergic pathways,
attention-deficit hyperactivity
disorder, 172
catechol-O-methyltransferase
(COMT)
and antisocial behavior, 154
and attention-deficit hyperactivity
disorder, 172
and bipolar disorder, 204
localization, 300
and opioid addiction, 300–301
panic disorder studies, 94, 105
polymorphisms, 95
roles, 94–95, 300
CBZ see carbamazepine (CBZ)
CCK (cholecystokinin), 95
CDH17, and major depressive
disorder, 225
cdk5, and drug addiction, 83
cell biology
in genetic susceptibility factor
studies, 71–72
schizophrenia studies, 251–253
cell cultures, primary, 71
cellular assays, in vitro, 72
Center for Information Biology
Gene Expression Database
(CIBEX), 44
CFG see Convergent Functional
Genomics (CFG)
c-fos
and drug addiction, 83
and histone phosphorylation, 81
Chevron USA, Inc. v. Echazabal, 334
childhood apraxia of speech, 160
children
conduct disorder, 145
schizophrenia, 4
chlorpromazine, 57
CHNRA5, and stimulant dependence,
307–308
CHNRA7, and schizophrenia, 252
cholecystokinin (CCK), 95
cholinergic pathways, attention-deficit
hyperactivity disorder, 172
CHRH1, and panic disorder, 95
CHRM2, and alcoholism, 356–357
CHRNA3
and lung disease, 292
and nicotine dependence, 290
CHRNA4
in attention-deficit hyperactivity
disorder, 172
and lung disease, 292
CHRNA5
and alcohol dependence, 294
and nicotine dependence, 290–292,
295
CHRNA7, and schizophrenia, 350
385
 Index
CHRNB4, and lung disease, 292
chromatin
epigenetic mechanisms, 79–81, 88
future research, 88
remodeling, 80, 87
structure, 84, 87
cocaine-induced alterations, 84
chromosomal anomalies, in
obsessive–compulsive
disorder, 130, 130
chromosomal haplotypes, 16
chromosomal regions, in
obsessive–compulsive
disorder, 123
chromosome 1
1p36, and major depressive
disorder, 216
1p36.3–34.3, and eating
disorders, 265
and alcoholism, 282–283
and schizophrenia, 238–239,
350–351
chromosome 2
2p, and Tourette syndrome, 339
2q, and autism spectrum disorders,
186, 188
and alcoholism, 282–283
and schizophrenia, 239–240
chromosome 3
3p12.3, and reading disability, 163
3q27–28, fine mapping, 122
and alcoholism, 282–283
and schizophrenia, 240
chromosome 4
4p, and alcoholism, 283
4q
and alcoholism, 283
arrangements, 281
chromosome 5
5p15, and autism, 187
and schizophrenia, 240–241
chromosome 6
6p22, and reading disability,
162–163
6q27, 187
and schizophrenia, 241–243
chromosome 7
7p15.3, and major depressive
disorder, 221
7q31, and language
impairment, 164
7q32, and diabetes, 275
7q35
and language impairment, 165
and Tourette syndrome,
340–341
7q
and alcoholism, 284
and autism, 186, 187
386
and alcoholism, 282–283
and antisocial behavior, 154
and schizophrenia, 243
and Tourette syndrome, 340
chromosome 8, and schizophrenia,
243, 246
chromosome 9, 9p24, and obsessive–
compulsive disorder, 122
chromosome 10
10p12, and diabetes, 275
10q, and alcoholism, 282–283
chromosome 11
11p14, and autism, 187
11p15
bipolar disorder, 200–201
and diabetes, 275
and schizophrenia, 245–246
chromosome 12
12p13, and major depressive
disorder, 221
12q24, and diabetes, 275
chromosome 13
13q21, and language
impairment, 164
13q31, and schizophrenia, 206–207
and schizophrenia, 246–247
chromosome 14, and schizophrenia,
247
chromosome 15
15q11–13
and Angelman syndrome, 365
and autism spectrum disorder,
192, 365–366
and obesity, 275
15q21, and reading disability, 162
15q25 26, and major depressive
disorder, 216, 217
15q25, and nicotine dependence,
290, 292
and Prader–Willi syndrome, 365
chromosome 16
16p11, and autism spectrum
disorder, 184, 185, 192
16q23–24, and language
impairment, 164
chromosome 17
17p11.2, and Smith–Magenis
syndrome, 366
17q12, and major depressive
disorder, 216
17q
and autism, 186
and depression, 54–55
and Tourette syndrome, 339
chromosome 18
18p11, and bipolar disorder, 207
18q22, and Tourette syndrome,
340, 341
and bipolar disorder, 201–202
chromosome 19, 19q13, and reading
disability, 164
chromosome 20, 20p13, and
autism, 187
chromosome 21, and Down
syndrome, 363–364
chromosome 22
22q11–13, and autism spectrum
disorder, 192
22q12, and bipolar disorder,
206–207
and schizophrenia, 247–248
chromosome 22q deletion syndrome,
and obsessive–compulsive
disorder, 127–128
chromosomes, and meiosis, 14–15
chronic obstructive pulmonary disease
(COPD), genetics, 292–294
chronic social defeat stress, 85–86
CIBEX (Center for Information
Biology Gene Expression
Database), 44
cigarette consumption, and lung
cancer, 287
cigarette smoking
mortality, 287
see also nicotine dependence
cimetidine, drug interactions, 53–54
citalopram, studies, 55–56
CJD (Creutzfeldt–Jakob disease), 378
clinical depression see major
depressive disorder (MDD)
CLINT1, and schizophrenia, 240–241
clozapine, 57–59
CLSA (Cooperative Linkage Study in
Autism), 186–187
CLU, and Alzheimer’s disease, 375
ClustalW, 42
Cluster-Buster, 36, 40
CMIP, and language impairment, 164
CNPs see copy number
polymorphisms (CNPs)
CNR1
and post-traumatic stress
disorder, 139
and stimulant dependence, 312
CNTNAP2
and language impairment, 165, 189,
192–193
and Tourette syndrome, 340–341
CNVs see copy number variations
(CNVs)
cocaine
epigenetic mechanisms, 83–85
and histone phosphorylation, 82
cocaine dependence, 306, 314
candidate genes, 309–310
genome-wide association studies,
308, 313, 314

cognitive deficits, in
schizophrenia, 234
cohort effect, 198
color space, 29
common environment, use of
term, 114
common genetic variants, 23
autism, 191–192
common pathways model, antisocial
behavior, 147
common trait rare gene hypothesis
(CTRV), 273
comorbidities, psychiatric disorders, 7
Complete Genomics, next-generation
sequencing, 29
COMPOUND database, 45
compulsive hoarding, 122
genetics, 130–131
computer-based analysis and
resources, for psychiatric
genetics research, 34–47
COMT see catechol-O-
methyltransferase (COMT)
COMT
murine studies, 300–301
and personality disorders, 321
and schizophrenia, 247–248,
350–351
and stimulant dependence, 311
variants, 128, 300
conditioning, heritability, 118
conduct disorder, 148–149
children, 145
conscience, development
failure, 156
consent issues, genetic information,
330–331
controls, selection criteria, 7–8
Convergent Functional Genomics
(CFG), 50
applications, 49
approaches, 51
biomarkers, 50
development, 49
and genome-wide association
studies, 49
p-values and, 51
Cooperative Linkage Study in Autism
(CLSA), 186–187
COPD (chronic obstructive
pulmonary disease), genetics,
292–294
copy number analysis, 26
copy number polymorphisms (CNPs)
analysis, 26
and attention-deficit hyperactivity
disorder, 173
copy number variations (CNVs), 4,
30–31, 375
in autism, 191–192
identification, 26, 175
in obesity, 273
schizophrenia, 249–250
and bipolar disorder compared,
250–251
in Tourette syndrome, 343
use of term, 173
cortico-striato-thalamocortical
(CSTC) pathways,
alterations, 127
corticotropin releasing hormone
(CRH), 95
cortisol levels, and suicidal
behavior, 353
co-twin control studies, schizophrenia,
235–237
CR1, and Alzheimer’s disease, 375
CREB1, and major depressive
disorder, 215–216
Creutzfeldt–Jakob disease (CJD), 378
CRH (corticotropin releasing
hormone), 95
CRHR1, major depressive disorder
studies, 224
criminal responsibility, and behavioral
genetics, 327–329
criminals, sterilization, 325
CSTC (cortico-striato-
thalamocortical) pathways,
alterations, 127
CTNNA2, and bipolar disorder,
206–207
CTRV (common trait rare gene
hypothesis), 273
cultural risk factors, migration studies
and, 3
CYP1A2, 57–58
CYP2D6, polymorphisms, 57, 276
cytochrome p450 system, 53–54
amphetamine metabolism, 176
pharmacokinetic studies, 57
cytogenetic abnormalities
schizophrenia, 249–250
Tourette syndrome, 340–341
Cytoscape, 45–46
cytosine, methylation, 298
Danio rerio (zebrafish), as animal
model, 73
DAO, and bipolar disorder, 216
DAOA, and schizophrenia, 204,
246–247
Darrow, Clarence, 328
Darwin, Charles, On the Origin of
Species, 325
databases
in linkage studies, 35
mutations, 36–37, 39
Index
single nucleotide polymorphisms,
36–39
data mining, 40
DAT (dopamine active transporter)
gene, 139
DBH
and attention-deficit hyperactivity
disorder, 172
and post-traumatic stress disorder,
137–139
and stimulant dependence, 311–312
dbSNP (Single Nucleotide
Polymorphism Database), 39
DCDC2, and reading disability,
162–163
DCTN5, and bipolar disorder, 205
DCX, and reading disability, 162
debrisoquin, pharmacokinetics, 53–54
deletions, schizophrenia, 249–250
Dennet, D. C., 326–327
depression
animal models, 85
definition, 212
epigenetic mechanisms, 79–88
etiology, 212
neurotrophic model of, 219
and nicotine dependence, 294
pharmacotherapy, 53
symptoms, 85
and Tourette syndrome, 338
treatment, 85
see also major depressive
disorder (MDD)
deterministic alleles, 8
developmental disconnection,
hypothesis, 193
developmental disorders
characteristics, 368
future research, 369
genetics, 363–369
developmental dyspraxia, 160
DGKH, and bipolar disorder, 206
diabetes, 275
Diagnostic and Statistical Manual of
Mental Disorders (DSM)
DSM-I, traumatic neurosis,
134–135
DSM-III
generalized anxiety disorder, 119
post-traumatic stress disorder, 134
social phobia, 112–113
trauma exposure, 135
DSM-IV
alcohol abuse disorders, 279
alcohol dependence, 279
anorexia nervosa, 263, 266–267
antisocial behavior related
disorders, 148–150
autism, 183
387
 Index
Diagnostic and Statistical Manual of
Mental Disorders (DSM) (cont.)
bulimia nervosa, 263
nicotine dependence, 288
panic disorder, 90
phobias, 112–113
post-traumatic stress disorder, 134
trauma exposure, 135
DSM-V, 107
diathesis–stress model, and
post-traumatic stress
disorder, 134
dietary factors, drug metabolism, 53–54
DISC1, 73
and schizophrenia, 204, 238–239,
251
structure, 70–71
studies, 69, 72
animal models, 76–77
murine, 76
disease etiology, adoption studies
and, 2–3
disease mapping, single nucleotide
polymorphisms, 23–24
diseases
complex, 27
genetic variants, 35
genome-wide association studies,
34–35
disorder transmission studies,
adoption studies and, 3
disrupted in schizophrenia 1 (DISC1)
see DISC1
dizygotic (DZ) twins, 2, 14
antisocial behavior studies, 145
antisocial personality disorder
studies, 317
autism studies, 183
bipolar disorder studies, 196
carbon dioxide hypersensitivity
studies, 96
extraversion studies, 117
major depressive disorder,
214–215
neuroticism studies, 117
obsessive–compulsive disorder
studies, 121
panic disorder studies, 92
personality disorder studies, 318
post-traumatic stress disorder
studies, 136
schizophrenia studies, 235
co-twin controls, 235–237
stimulant dependence studies,
307–308
Tourette syndrome studies, 337
DMS (dystonia myoclonus syndrome),
and obsessive–compulsive
disorder, 127
388
DNA demethylases, existence of,
82–83
DNAm see DNA methylation
(DNAm)
DNA methods, 23
DNA methylation (DNAm), 79–82,
224
and alcoholism, 298–299
catalysis, 82–83
current research, 224
and gene expression repression, 82
and opioid addiction, 298–299
roles, 87
DNA samples, whole-blood, 23
DNA sequence variations, 21, 36
analysis tools, 37
future research, 46–47
DNA sequencing
analysis, 35
focused, 35
and genetic studies, 20–21
tools and resources, 36
dopamine active transporter (DAT)
gene, 139
dopamine beta-hydroxylase (DBH)
gene see DBH
dopamine receptor gene, 137–139
post-traumatic stress disorder
studies, 140
dopamine-related genes, and
stimulant dependence,
310–312
dopaminergic pathways, attention-
deficit hyperactivity
disorder, 169–171
dopaminergic system, 58
dopamine system
genes
in obsessive–compulsive
disorder, 126
in post-traumatic stress disorder,
137–139
genes involved in, 94–95
dopamine transporter gene (SLC6A3)
see SLC6A3
Down syndrome, 363–364
DPYSL2, and schizophrenia, 243
DRD1, polymorphisms, 171
DRD2
alleles, 154
and alcoholism, 281–282
animal models, 178
polymorphisms, 58, 171, 266–267
and schizophrenia, 245–246
and stimulant dependence, 310–311
DRD3
polymorphisms, 58
and schizophrenia, 240
and stimulant dependence, 311
DRD4, 154
animal models, 178
and personality disorders, 321
polymorphisms, 171
variants, 175
DRD5, polymorphisms, 171
Drosophila melanogaster (fruit fly),
as animal model, 73
drug addiction
epigenetic mechanisms, 79–88
treatment, 83
see also cocaine dependence, heroin
addiction; opioid addiction;
psychostimulant dependence;
stimulant dependence
drug-addiction studies, animal
models, 83–85
DRUG database, 45
drug interactions, antidepressants,
53–54
drug metabolism, dietary factors,
53–54
drug relapse, mechanisms, 83
drug responses, environmental factors,
61–65
DSM see Diagnostic and Statistical
Manual of Mental Disorders
(DSM)
DTNBP1, and schizophrenia, 241–243
duplications, schizophrenia, 249–250
Durham v. US, 327–328
dynamic programming algorithms, 42
dyslexia see specific reading disability
(SRD)
dystonia myoclonus syndrome (DMS),
and obsessive–compulsive
disorder, 127
DYX loci, 161
DYX1, 162
DYX1C1, and reading disability, 162
DYX2, 162–163, 165–166
DYX5, 163
DYX8, 165–166
EAAC1 see SLC1A1
eating disorders (EDs)
linkage studies, 265–266
morbidity, 262
prevalence, 262
symptoms, 263
see also anorexia nervosa (AN);
bulimia nervosa (BN); obesity
EBI see European Bioinformatics
Institute (EBI)
EBI Dali, 43
Echazabal, Chevron U.S.A.,
Inc. v., 334
egr-1, and cocaine withdrawal,
83–84

electroencephalography (EEG)
interhemispheric theta coherence,
356–357
resting, beta power, 354–356
Elston–Stewart algorithm, 26–27
EMBOSS, 42–43
E-MSD (Macromolecular Structure
Database), 43
EN2, and autism, 189–190
ENCODE Project, 20, 32, 47
regions, 40–41
endophenotypes
alcoholism, 352
for antisocial behavior, 154–156
attention-deficit hyperactivity
disorder, 350–351
in autism spectrum disorders,
188–189
behavioral, 188
brain oscillations as, 353–359
concept of, 348
conceptual development, 347–348
definition, 187–188
family studies, 349
heritability, 348–349
late onset Alzheimer’s disease,
375–377
neurochemical metabolites as,
351–353
neuroimaging, 351
obsessive–compulsive disorder, 188
for psychiatric disorder
classification, 9
psychiatric disorders and, 349
in psychiatric genetics, 347–360
applications, 349–359
future trends, 359–360
identification criteria, 348–349
schizophrenia, 350–351
state independence, 349
variations, 349
ENU (N-ethyl-N-nitrosourea)
mutants, 178
environmental factors
alcoholism, 279–280
antisocial behavior, 156
and behavior, 327
drug responses, 61–65
identification, 9
in obsessive–compulsive
disorder, 130
Tourette syndrome, 342
see also gene–
environment interactions
(G  E)
environmental risk factors, migration
studies and, 3
environmental variance, phobias, twin
studies, 114
EPHB1, and major depressive
disorder, 225
EPIC (Events Preceding Interstitial
Cystitis), 99–100
epidemiology
phobias, 112–113
see also genetic epidemiology
epigenetic mechanisms, 79–83
animal models, 88
in depression, 79–88
in drug addiction, 79–88
future research, 88
and gene expression, 79
epigenetic modifications, 51
obesity, 275–276
epigenetics
definition, 79
field of study, 79
future research, 88
major depressive disorder,
223–224
future research, 226
mu-opioid receptor, 298–299
personality, 322
post-traumatic stress disorder,
140–141
research issues, in psychiatry, 87
schizophrenia, 231–232
Tourette syndrome, 343–344
EPSIN4, and schizophrenia, 240–241
eQTL see expression quantitative trait
locus (eQTL)
EROs (even-related oscillations), theta
and delta, 357–359
escitalopram, studies, 55–56
ethanol, consumption, 280
ethical issues
behavioral genetics, 324–334
eugenics, 324–326
sterilization, 325
N-ethyl-N-nitrosourea (ENU)
mutants, 178
eugenics, 324–326
European Bioinformatics
Institute (EBI)
GeneWise, 35
InterProScan, 43
Macromolecular Structure
Database, 43
even-related oscillations (EROs), theta
and delta, 357–359
Events Preceding Interstitial Cystitis
(EPIC), 99–100
exercise, and anorexia nervosa, 263
expanded spectrum approach,
97–102
ExPasy Translate, 35
expression quantitative trait
locus (eQTL)
Index
autism studies, 188–189
resources, 44
externalizing disorders
heritability, 146–148
twin studies, 116–117
use of term, 145
extraversion, 117
eye-tracking dysfunction, and
schizophrenia, 350
FAAH, and stimulant dependence, 312
FACS (fluorescence activated cell
sorting), 87
factor analysis, personality
disorders, 320
Fagerström Test of Nicotine
Dependence (FTND), 288
familial aggregation, 3
indicators, 1–2
familial Alzheimer’s disease (FAD),
376
diagnostic criteria, 371
genetic susceptibility factors, 371
familial recurrence, schizophrenia,
232–233
familial specificity, phobias, 114
familial transmission
bipolar disorder, 207
phobias, 113
schizophrenia, 207
family epidemiology, bipolar disorder,
196–199
family history screen (FHS), 96–97
Family Interview for Genetic Studies
(FIGS), 96–97
family studies
advantages, 2
alcoholism, 282–283
anorexia nervosa, 264
antisocial behavior, 145, 153–154
antisocial personality disorder, 317
attention-deficit hyperactivity
disorder, 168
autism, 186–187
bipolar disorder, 196
bulimia nervosa, 264
early, 34
endophenotypes, 349
generalized anxiety disorder, 119
genetic epidemiology, 1–2
major depressive disorder,
213–214
obsessive–compulsive disorder,
121, 123
panic disorder, 90–92
post-traumatic stress disorder,
135–136
psychiatric disorders, 69
sampling, 7
389
 Index
family studies (cont.)
schizophrenia
cognitive deficits and imaging
changes, 234
segregation analysis, 237–238
subtypes, 232–233
variance component analysis,
237–238
single nucleotide polymorphism
genotyping, 26–27
social phobia traits, 117–118
stimulant dependence, 306–307
substance use disorders, 5, 6
Tourette syndrome, 337
FASTA, 42
fastq, 31
fear
brain responses to, 103–104
conditioning, susceptibility to, 118
irrational, 113
and post-traumatic stress disorder,
103–104
prevalence, 113
processing, in panic disorder, 107
stimuli, masked versus unmasked,
104
unconditioned, 102–103
fearful faces, amygdala
activation, 106
FHS (family history screen), 96–97
FIGS (Family Interview for Genetic
Studies), 96–97
FKBP5
and major depressive
disorder, 219
and post-traumatic stress disorder,
139–140
fluorescence activated cell sorting
(FACS), 87
fluoxetine, 86, 179, 224
pharmacokinetics, 53–54
FMR1, and fragile X syndrome, 366–367
fMRI (functional magnetic resonance
imaging), 351
focal dystonias, 127
FosB, and drug addiction, 83
FOXP2, and language
impairment, 164
fragile X mental retardation protein
(FMRP), 366–367
fragile X syndrome (FXS)
and autism spectrum disorders
compared, 367
etiology, 366–367
future research, 368–369
genetics, 366–369
and mGluR theory, 367–368
prevalence, 367
sex differences, 367
390
fruit flies, 73 animal models, 178–179
FTND (Fagerström Test of Nicotine antisocial behavior, 145–157
Dependence), 288 adoption studies, 152
functional analysis, computer- approaches, 152–153
based, 35 studies, 152–156
functional assessment, genes, 38 twin studies, 152
functional imaging studies attention-deficit hyperactivity
limitations, 103 disorder, 173
panic disorder, 102–104, 107 definition, 151–152
functional magnetic resonance major depressive disorder, 224
imaging (fMRI), 351 obesity, 272–273
functional polymorphisms, and post-traumatic stress
schizophrenia, 251 disorder, 142
functional validation, of genetic substance use disorders, 150
susceptibility factors, for gene expression, 44
psychiatric disorders, 69–77 Alzheimer’s disease, 377–378
FXS see fragile X syndrome (FXS) in brain, research issues, 87
FXYD6, and schizophrenia, 245–246 databases, 44
FZD3, and schizophrenia, 243 drug-induced, 83
epigenetic mechanisms and, 79
GABA-ergic system, genes involved in, major depressive disorder,
95–96 222–223
GABA system see gamma- future research, 226
aminobutyric acid (GABA) repression, 82
system Gene Expression Omnibus
GABHS (group A beta hemolytic (GEO), 44
streptococci), 342 gene expression profiling, 50
GABRA1, and schizophrenia, 240–241 gene expression studies
GABRA2, and alcoholism, 282–283, animal models, 49
354–357 developments, 51
GABRA4, and alcoholism, 354–356 future trends, 51
GABRA6, and schizophrenia, 240–241 human postmortem brain, 49
GABRB1, and alcoholism, 354–356 large-scale, viability, 49
GABRB2, and schizophrenia, 240–241 in psychiatric disorders, 49–51
GABRG1, and alcoholism, 283, p-values in, 50–51
354–356 GeneGo MetaCore, 46
GABRP, and schizophrenia, 240–241 GeneHunter program, 26–27
GAD2, and obesity, 275 gene identification
GAD see generalized anxiety disorder antisocial behavior, 153–154
(GAD) future trends, 49
GAD (glutamic acid decarboxylase), genetic epidemiology and, 7–10
95–96 gene localizations, in reading
Galton, Sir Francis, 324–325 disability, 161
gamma-aminobutyric acid (GABA) gene mapping, major depressive
system disorder, 215–222
panic disorder studies, 95–96 gene mapping methods
receptor genes, and alcoholism, 283 bipolar disorder, 199–200
GENDEP (Genome-Based subphenotypes, 207
Therapeutic Drugs for Gene Ontology (GO) database, 45
Depression), 55–56 generalized anxiety disorder (GAD),
gender differences see sex differences 119–120
gene association studies, hypothesis- comorbidities, 117
driven, opioid addiction, 302 genetic epidemiology, 5
GENE database, 45 genetics, 112–120
gene dosage, disorders, 363–364 generalized social phobia, 117
gene–environment correlations, 6 genes
gene–environment interactions additive
(G  E), 6, 151–152 generalized anxiety disorder, 120
alcoholism, 284 and phobias, 114–115, 117

in adenosine system, 95
in dopamine system, 94–95
functional assessment, 38
in GABA-ergic system, 95–96
in G-protein signaling regulators, 96
in neuropeptide systems, 95
in obsessive–compulsive disorder,
121, 123–126
rare, in obsessive–compulsive
disorder, 130, 130
related to carbon dioxide
hypersensitivity, 96
in serotonin system, 93–94
see also candidate genes
GENESEQ, 45
gene set expression analysis
(GSEA), 44
genes-to-brain-to-antisocial-behavior
model, 156
genetic analysis, issues, 19–20
genetic association
bipolar disorder, 200
schizophrenia and bipolar disorder
compared, 250–251
genetic bases, of psychiatric disorders,
methods, 23
genetic case-control association
studies, post-traumatic stress
disorder, 137–140
genetic counseling, bipolar disorder, 208
genetic engineering, murine, 73
genetic epidemiology, 1–3
adoption studies, 2–3
Alzheimer’s disease, 371–373
applications, to gene identification,
7–10
early studies, 1
family studies, 1–2
field of study, 1
migration studies, 3
panic disorder, 90
phobias, 113–118
future research, 119
of psychiatric disorders, 1–10
risk estimation, 8
schizophrenia, 3–4
twin studies, 2
genetic factors, interactions, 20
genetic heterogeneity, 6, 20
genetic imprinting, 365–366
genetic information
applications, social, 324–326
consent issues, 330–331
control of, 329
forensic, privacy issues, 329–330
health professionals and, 332–333
privacy of, 329
genetic linkage
alcoholism, 282–283
bipolar disorder, 200
schizophrenia, 238
genetic markers, 34–35
see also biomarkers
genetic risk factors see genetic
susceptibility factors
genetics
alcohol dependence, 295
alcoholism, 279
alcohol use disorders, 279
Alzheimer’s disease, 371–378
Angelman syndrome, 365–366
anorexia nervosa, 262–270
antisocial behavior, 145
autism, 183
autism spectrum disorders, 183
bipolar disorder, 209
subphenotypes, 198–199
bulimia nervosa, 262–270
chronic obstructive pulmonary
disease, 292–294
compulsive hoarding, 130–131
developmental disorders, 363–369
Down syndrome, 363–364
fragile X syndrome, 366–369
generalized anxiety disorder,
112–120
grooming disorders, 128–130
heroin addiction, 301, 302–303
historical background, 324
lung cancer, 292–295
lung disease, 292–295
major depressive disorder, 212–227
Mendelian, 324
nicotine dependence, 288
and psychiatric disorders,
294–295
obesity, 272–277
obsessive–compulsive disorder,
121–131
future research, 130–131
opioid addiction, 297–303
origin of term, 347
personality disorders, 316–322
phobias, 112–120
Prader–Willi syndrome, 365–366
psychostimulant dependence,
308–310
reading disability, 161
and responsibility, 326–329
Rett’s disorder, 364–365
schizophrenia, 230–253
Smith–Magenis syndrome, 366
speech sound disorder, 165–166
stimulant dependence, 306–314
Tourette syndrome, 126–127,
336–344
weight loss, 276
see also behavioral genetics;
Index
epigenetics; pharmacogenetics;
psychiatric genetics
genetic studies
autism, historical background, 184
DNA sequencing and, 20–21
limitations, 61–65
panic disorder, 90
issues, 107
limitations, 106
post-traumatic stress disorder,
135–140
see also molecular genetic studies
genetic susceptibility factors
animal models, 72–77
assays, 74–75
DISC1 studies, 76–77
readouts relevant to psychiatric
disorders, 74–75
biochemical studies, 69
bipolar disorder, 69
and brain dysfunction, 77
cell biology studies, 71–72
during neurodevelopment, 75
familial Alzheimer’s disease, 371
for-profit testing, 331–332
human studies, with brain
imaging, 77
late onset Alzheimer’s disease, 373
pathways, 70–71
protein chemistry studies, 69–71
for psychiatric disorders, functional
validation, 69–77
schizophrenia, 4, 69
genetic variance
anxiety disorders, 116
phobias, twin studies, 114
genetic variants
autism, issues, 191
diseases, 35
functional assessment, 38
germ-line, 23
identification, 61–65
linkage and linkage disequilibrium,
14–17
multiple comparisons, 19–20
types of, 23
see also common genetic variants;
copy number polymorphisms
(CNPs); copy number
variations (CNVs); rare genetic
variants
gene transfer, in utero, 73–74, 76–77
GeneWise, 35
GenMAPP, 46
Genome-Based Therapeutic Drugs for
Depression (GENDEP), 55–56
genome scan meta-analysis
(GSMA), 202
genome sequencing, complete, 35
391
 Index
genome-wide association studies
(GWAS), 19–20, 55–56
alcoholism, 284
Alzheimer’s disease, 374–375
antidepressants, 225
applications, 27
disease studies, 34–35
attention-deficit hyperactivity
disorder, 173
autism, 190–191
bipolar disorder, 200, 205–206
case-control, 27
cocaine dependence, 308, 314
and Convergent Functional
Genomics, 49
development, 24–25
errors, sources of, 28
late onset Alzheimer’s disease,
374–375
limitations, 35, 49
major depressive disorder, 219–222
nicotine dependence, 289–292,
295
obesity, 274
obsessive–compulsive disorder,
122–123, 130
opioid addiction, 302–303
panic disorder, 93, 106
permutational analysis, 28
personality disorders, 321–322
post-traumatic stress disorder,
141–142
single nucleotide polymorphism
genotyping, 27–28
stimulant dependence, 312–313
Tourette syndrome, 342–343
use of term, 23
genome-wide linkage studies
attention-deficit hyperactivity
disorder, 168–169, 171
autism, 186–187
genome-wide studies, versus candidate
gene studies, 18
genomic imprinting disorders, 365
genomics
future research, 47
impacts, on psychiatric science and
practice, 9–10
see also Convergent Functional
Genomics (CFG)
genomic studies, phobias, 118–119
genotype likelihood format (GLF), 31
genotyping
concept of, 347
future research, 46–47
GEO (Gene Expression Omnibus), 44
Gibbs sampling, 42
GLF (genotype likelihood format), 31
global alignments, 42
392
glucocorticoid receptor (GR),
polymorphisms, 139
glutamate system genes, 123
glutamic acid decarboxylase (GAD),
95–96
glutaminergic neurotransmission, in
attention-deficit hyperactivity
disorder, 172
P-glycoprotein gene (ABCB1),
variants, and methadone
dose, 302
GNB3, and major depressive disorder,
218–219
GO (Gene Ontology) database, 45
GPR158, and obesity, 275
GPRIN2, and obesity, 273
G-protein signaling regulators, genes
involved in, 96
GR (glucocorticoid receptor),
polymorphisms, 139
GRIN1, and weight loss, 276
GRK3, and bipolar disorder, 204
GRM3, and schizophrenia, 243, 253
GRM7
and attention-deficit hyperactivity
disorder, 176
and major depressive disorder,
221–222
GRM8, and alcoholism, 357–359
GRN, and Alzheimer’s disease, 378
grooming disorders, 128–130
group A beta hemolytic streptococci
(GABHS), 342
GSEA (gene set expression analysis), 44
GSMA (genome scan meta-
analysis), 202
GWAS see genome-wide association
studies (GWAS)
G  E see gene–environment
interactions (G  E)
haloperidol, 57, 224, 252
haplotype blocks, 39
haplotypes, chromosomal, 16
HapMap Project see International
HapMap Project
Haseman–Elston linkage algorithm,
118–119
HATs (histone acetyltransferases), 81
HD see Huntington’s disease (HD)
HDACs see histone deacetylases
(HDACs)
HDC, and Tourette syndrome,
336–339
HDMs (histone demethylases), 82
health professionals, and genetic
information, 332–333
heart rate, and antisocial behavior,
155–156
hemizygous deletions, animal
models, 178
HEP (Human Epigenome Project), 47
heritability
alcoholism, 5–6
anorexia nervosa, 264
antisocial behavior, 146
antisocial personality disorder,
148
attention-deficit hyperactivity
disorder, 149–150, 168–169
body mass index, 272
bulimia nervosa, 264
conditioning, 118
conduct disorder, 148–149
of DSM-IV disorders related to
antisocial behavior, 148–150
endophenotypes, 348–349
establishment, 13
estimates, 3
externalizing disorders, 146–148
migration studies, 14
missing, 274–275
mood disorders, 196
nicotine dependence, 289
obesity, 272
oppositional defiant disorder, 149
panic disorder, 92
personality disorders, 318–321
personality traits, 326
phobias, sex differences, 114
psychopathic traits, 150–151
schizophrenia, 230–231
smoking initiation, 288
stimulant dependence, 306–308
substance use disorders, 150
Tourette syndrome, 337
traits, 113
twin studies, 14
heritability estimates, psychiatric
disorders, 3
heroin addiction, genetics, 301–303
heterogeneity
autism, 187
phenotypic, 20
see also genetic heterogeneity
HGMD (Human Gene Mutation
Database), 37–38
HGVBase, 37–38
high-density single nucleotide
polymorphism (SNP)
genotyping
arrays, 23
technologies, 24–25
high-throughput assays, 44
L-histidine decarboxylase (HDC)
see HDC
histone acetyltransferases (HATs),
81

histone deacetylases (HDACs), 81
cocaine and, 84
inhibitors, 86
histone demethylases (HDMs), 82
histone methyltransferases (HMTs),
82
inhibition, 85
histones
acetylation, 79–81, 88
and bdnf, 87–88
genome-wide studies, 81
and epigenetic mechanisms, 79–81
methylation, 79–82, 88
localization, 82
regulatory mechanisms, 82
phosphorylation, 79–82, 85
localization, 81
regulatory mechanisms, 82
HLA, and schizophrenia, 241–243
HLA-B*1502 allele test, 65
HMTs see histone methyltransferases
(HMTs)
hoarding see compulsive hoarding
Holmes, Oliver Wendell, 325
HPA axis see hypothalamic-pituitary-
adrenal (HPA) axis
5-HT1A receptor, 93–94
5-HT1B (5-hydroxytriptamine-1B
receptor), and opioid
addiction, 301
5-HT2A receptors
and major depressive disorder, 55
and obesity, 276
polymorphisms, 58, 94
5-HT2C receptors, 58–59
5-HT (hydroxytryptamine), 154
HTR1B, 301
HTR1D, and eating disorders, 265
HTR2A
anorexia nervosa studies, 267–269
and attention-deficit hyperactivity
disorder, 172
and major depressive disorder,
55–56
htt (huntingtin), and Huntington’s
disease, 69
5-HTTLPR see serotonin-transporter-
linked polymorphic region
(5-HTTLPR)
Human Epigenome Project (HEP), 47
Human Gene Mutation Database
(HGMD), 37–38
human genetics research
analysis strategies, overview, 13–21
developments, 13
field of study, 13
future trends, 21
human genome
properties, 13
sequencing, 347
variant identification, 13
strategies, 34
Human Genome Project, 324
human studies, genetic susceptibility
factors, with brain imaging, 77
huntingtin (htt), and Huntington’s
disease, 69
Huntington’s disease (HD),
etiology, 69
hybridization-based target
enrichment, 32
6-hydroxydopamine (6-OHDA), 177
5-hydroxytryptamine-1B receptor
(5-HT1B), and opioid
addiction, 301
hydroxytryptamine (5-HT), 154
hypercapnia, 96
hypothalamic-pituitary-adrenal
(HPA) axis, 224
and major depressive disorder, 219
IC see interstitial cystitis (IC)
ICD-10 (International Classification of
Diseases), alcohol abuse
disorders, 279
idiopathic generalized epilepsy
(IGE), and schizophrenia,
249–250
IGE (idiopathic generalized epilepsy),
and schizophrenia, 249–250
IGF2, and obesity, 276
IL15, and obesity, 273
IL28RA, and suicidal ideation,
56–57
Illumina
BeadArray Mapping arrays, 25
Genome Analyzer, 28, 29
HiSeq, 29
procedures, 25
imaging changes, in
schizophrenia, 234
imaging genetics, 104–106
IMGSAC see International Molecular
Genetic Study of Autism
Consortium (IMGSAC)
imipramine, 224–225
pharmacokinetics, 53–54
IMMP2L, and Tourette
syndrome, 340
inbred strains, selected, 178
incomplete penetrance, 6
induced pluripotent stem (iPS)
cells, 71
infections
and immune response, Tourette
syndrome, 342
streptococcal, 130
viral vectors, 71
Index
Ingenuity network analyses,
autism, 192
Ingenuity Pathway Analysis, 46
inheritance
general personality dimensions,
320–321
Mendelian laws of, 1–2
modes of, 6
personality disorders, 318–320
interhemispheric theta coherence,
356–357
interleukin receptors, encoding,
56–57
internalizing disorders, twin studies,
116–117
International Classification of Diseases
(ICD-10), alcohol abuse
disorders, 279
International HapMap Project, 18,
24–25
databases, 39
International Molecular Genetic Study
of Autism Consortium
(IMGSAC), 183–186
two-stage genome scan, 186
International Schizophrenia
Consortium, 46–47
InterProScan, 43
interstitial cystitis (IC), 99
case-control studies, 98, 99–100
in utero gene transfer, 73–74, 76–77
inversions, 30–31
in vitro cellular assays, 72
iPS cells see induced pluripotent stem
(iPS) cells
irrational fears, and phobias, 113–115
JAligner (software), 42
kappa-opioid receptor (KOPr), 299
KE family, 164
KIAA0319, and reading disability,
162–163, 166
knockdown animal models, 178
knock-in mice, 73
knockout animal models, 176–177
knockout mice, 73, 176–177
KOPr see kappa-opioid receptor
(KOPr)
Kyoto Encyclopedia of Genes and
Genomes (KEGG), 45–46
lactate dehydrogenase (LDH), 96
Lander–Green algorithm, 26–27
language impairment (LI), 164–165
deficits, 166
future research, 166
see also specific language
impairment (SLI)
393
 Index
large deletions, 30–31
large insertions, 30–31
late onset Alzheimer’s disease
(LOAD)
diagnostic criteria, 371
endophenotypes, 375–377
genetic susceptibility factors, 373
genome-wide association studies,
374–375
law, and behavioral genetics, ethical
issues, 326–329
LD see linkage disequilibrium (LD)
LDH (lactate dehydrogenase), 96
learning disabilities, 160–166
definitions, 160–161
diagnostic issues, 160–161
prevalence, 160–161
leptin gene, 58–59
LI see language impairment (LI)
linkage analysis, 17
Tourette syndrome, 338–339
linkage disequilibrium (LD),
200, 309
biological significance, 20
concept of, 27, 231
functional assessments, 20
genetic variants, 14–17
mapping, 17–18
single nucleotide
polymorphisms, 24
and variant segregation in
populations, 15–17
linkages
biological information, leveraging
strategies, 35
biological significance, 20
functional assessments, 20
genetic variants, 14–17
mapping, 17–18
special designs, 18–19
and recombination, 14–15
linkage studies, 26–27
Alzheimer’s disease, 374
analysis, 26–27
anorexia nervosa, 265–266
autism
historical background, 183–188
overview, 187–188
bipolar disorder, 200–202
bulimia nervosa, 265–266
databases in, 35
early, 34
eating disorders, 265–266
follow-up, alcoholism, 283–284
major depressive disorder, 215–216
obesity, 274
obsessive–compulsive disorder, 122
panic disorder, 92, 99
psychostimulant dependence, 308
394
schizophrenia, 231
single nucleotide polymorphism
genotyping, 26–27
surrogate markers, 26
linkage variables, identification,
188–189
lithium, 59–60
LOAD see late onset Alzheimer’s
disease (LOAD)
local alignments, 42–43
loci, linkages, 14–15
locomotor activity changes, animal
models, 177
logarithm of odds (LOD) scores, 18,
24, 26–27, 122, 231
lung cancer
and cigarette consumption, 287
genetics, 292–295
prevalence, 287
lung disease, genetics, 292–295
lymphocyte protein studies, 50
lyonization, 79–81
Macromolecular Structure Database
(E-MSD), 43
major depressive disorder (MDD), 4
adoption studies, 215
association studies, 217–218
candidate genes, 218–219
citalopram studies, 55
diagnosis, 227
diagnostic criteria, 212
epigenetics, 223–224
future research, 226
etiology, 212
family studies, 213–214
future research, 225–227
gene–environment interactions, 224
gene expression, 222–223
future research, 226
gene function, 222–223
future research, 226
murine models, 222–223
gene mapping, 215–222
genetic epidemiology, 212–215
genetics, 212–227
genome-wide association studies,
219–222
linkage studies, 215–216
pathophysiology, 227
pharmacogenetics, 53, 225
future trends, 227
pharmacokinetics, 54
phenotypes
alternative, 216–217
definition, 212
future research, 227
sex differences, 215–216
twin studies, 214–215, 224
malnutrition, 263–264
MANEA, and stimulant
dependence, 313
manic-depression see bipolar
disorder
MAOA see monoamine oxidase A
(MAOA)
MAPT, and Alzheimer’s disease,
373–374, 376–378
MAST, 40
MATCH, 40
mate-pair sequencing, 29
mating, non-random, 7
Maudsley Twin Register, 121
MC2R see melanocortin receptor
type 2 (MC2R)
MDD see major depressive disorder
(MDD)
measured risk factors, 154–156
MECP2, and Rett’s disorder, 364–365
medicine, and behavioral genetics,
ethical issues, 326–329
MedScan, 46
meiosis, chromosomes and, 14–15
melanocortin receptor type 2 (MC2R)
and opioid addiction, 299–300
roles, 299–300
MEME, 40
Mendel, Gregor Johann, 112
Mendelian genetics, 324
Mendelian laws, of inheritance, 1–2
Mendelian ratio, 2
mens rea, 327
mental disorders see psychiatric
disorders
mental illnesses see psychiatric
disorders
MERLIN program, 26–27, 122
MET, and autism, 189–190, 192–193
methamphetamine dependence, 314
genome-wide association
studies, 313
methadone, opioid addiction
treatment, 302
methylation
cytosine, 298
histones, 79–82, 88
see also DNA methylation
(DNAm)
methylphenidate, attention-deficit
hyperactivity disorder studies,
176, 179
methylthioninium chloride
(MTC), 374
mGluR theory, and fragile
X syndrome, 367–368
mice
in animal models, 72–73
COMT studies, 300–301

DISC1 studies, 76, 251
knock-in, 73
knockout, 73, 176–177
major depressive disorder studies,
222–223
over-grooming behavior, 128–130
schizophrenia studies, 251–253
transgenic, 73
microarray platforms, 49
microarray studies, Tourette
syndrome, 342–343
migraine, and bipolar disorder, 208
migration studies
genetic epidemiology, 3
heritability, 14
missing heritability, 274–275
mitral valve prolapse (MVP), 101
MoD Tools, 40
molecular genetic studies
antisocial behavior, 152
panic disorder, 92–97
molecular manipulations, 71, 73–74
molecular profiling, 75
monoamine oxidase A (MAOA)
antisocial behavior studies,
152–154
attention-deficit hyperactivity
disorder studies, 172
panic disorder studies, 94
monozygotic (MZ) twins, 2, 14
antisocial behavior studies, 145
antisocial personality disorder
studies, 317
attention-deficit hyperactivity
disorder studies, 176
autism studies, 183
bipolar disorder studies, 196
carbon dioxide hypersensitivity
studies, 96
extraversion studies, 117
major depressive disorder, 214–215
neuroticism studies, 117
obesity studies, 272
obsessive–compulsive disorder
studies, 121
panic disorder studies, 92
personality disorder studies, 318
post-traumatic stress disorder
studies, 136
schizophrenia studies, 235–237
stimulant dependence studies,
307–308
Tourette syndrome studies, 337
mood disorders
adoption studies, 4–5
and anorexia nervosa, 263
and bulimia nervosa, 263
genetic epidemiology, 4
heritability, 196
mood stabilizers, pharmacogenetics,
59–61
MOPr see mu-opioid receptor (MOPr)
motif-based alignments, 42
MRPL19, and reading disability, 163
MSK1, 85
MSP (multiple scan probability),
202
MTC (methylthioninium
chloride), 374
MTHFR, and major depressive
disorder, 216, 218–219
multiple alignments, 42
multiple comparisons, genetic
variants, 19–20
multiple scan probability (MSP), 202
multi-threshold multifactorial model,
196–198
multivariate genetic models, measured
risk factors, 155–156
mu-opioid receptor (MOPr)
epigenetics, 298–299
and opioid addiction, 297–298
roles, 297
MUSCLE, 42
Mutation Discovery, 37–38
mutations
analysis tools, 37
databases, 36–37, 39
and non-uniform recombination
rates, 19
schizophrenia, 251
MVP (mitral valve prolapse), 101
MYOCD, and heroin addiction,
302–303
nAChRs (nicotinic acetylcholine
receptors), 290–291
NAP1L5, and obesity, 273, 275
Naples high-excitability (NHE) rats,
178–179
Naples low-excitability (NLE) rats, 178
National Center for Biotechnology
Information (NCBI),
39, 42, 44
National Institute of Diabetes and
Digestive and Kidney Diseases
(NIDDK), 100–101
Nazis, eugenics, 325–326
NCAM1, and alcoholism, 281–282
NCBI Blast, 42
NCBI (National Center for
Biotechnology Information),
39, 42, 44
NCBI VAST, 43
Needleman–Wunsch algorithm, 42
Neuregulin-1, 72–73
neurochemical metabolites, as
endophenotypes, 351–353
Index
neurocognitive deficits, anorexia
nervosa, 263
neurocognitive phenotypes,
alcoholism, 353–359
neurocognitive tests, attention-deficit
hyperactivity disorder studies,
175
neurodegenerative diseases, common
themes in, 378
neuroimaging, endophenotypes, 351
neuroimaging studies
attention-deficit hyperactivity
disorder, 176
panic disorder, 102–106
neurons, olfactory, 71
neuropeptide systems, genes involved
in, 95
neuropsychological deficits,
schizophrenia, 234–235
neuropsychopharmacology, field of
study, 53
neuroticism, 117
neurotrophic model, of depression, 219
next-generation sequencing, 23
analysis, 30–31
data analysis, 30–31
data standards, 31
emerging platforms, 29
sequencing technologies, 31–32
standards, 30–31
technologies, 28–29
NFKB1, and alcoholism, 283
NHE (Naples high-excitability) rats,
178–179
nicotine dependence, 287–295
and alcohol dependence, 294
allele frequency distribution, 293
comorbidities, 287–288, 294–295
shared etiologies, 294
definitions, 289
and depression, 294
diagnostic criteria, 288
etiology, 287–288
genetics, 288
future research, 295
genome-wide association studies,
289–290, 295
limitations, 290
heritability, 289
prevalence, 287–288, 295
and psychiatric disorders, genetics,
294–295
stages, 288
twin studies, 288
nicotinic acetylcholine receptors
(nAChRs), 290–291
NIDDK (National Institute of
Diabetes and Digestive and
Kidney Diseases), 100–101
395
 Index
NLE (Naples low-excitability)
rats, 178
noncoding RNA, Tourette syndrome
studies, 343
nonparametric linkage (NPL) method,
122
Tourette syndrome, 339
nonpsychiatric disorders, and
psychiatric disorders, 101–102
nonrandom mating, 7
nonsynonymous polymorphisms,
39–40
noradrenergic pathway, animal
models, 177
nortriptyline
pharmacokinetics, 53–54
studies, 55–56
NOS1AP, and schizophrenia, 238–239
NPAS3, and schizophrenia, 247
NPL (non-parametric linkage)
method, 122
NPY
and alcoholism, 282
and cocaine administration, 83–84
NRG1, and schizophrenia, 204, 243,
252
NTRK2
and bipolar disorder, 203–204, 206
and suicidal ideation, 56–57
NTRK3, and major depressive
disorder, 219
obesity
candidate genes, 273, 277
copy number variations in, 273
disparate approaches, 275
epigenetic modifications, 275–276
gene–environment interactions,
272–273
genetic expression differences, 276
genetics, 272–277
applications, 276
future research, 276
genome-wide association studies, 274
heritability, 272
linkage studies, 274
parent of origin effects, 275
polygenic model, 275
prevalence, 272
prevention, 276
protective genes, 276
rare genetic variants, 273
resistance to, 276
therapy, 276
twin studies, 272
obsessive–compulsive disorder (OCD)
age of onset, as secondary
phenotype in linkage
studies, 122
396
animal models, 128
biological bases, 121
candidate genes, 121, 123–126
chromosomal anomalies, 130, 130
chromosomal regions,
candidate, 123
and chromosome 22q deletion
syndrome, 127–128
and dystonia myoclonus
syndrome, 127
endophenotypes, 188
environmental factors, 130
family studies, 121, 123
genes
rare, 130
SAPAP3, 128–130
SLC1A1, 123–124, 130
genetic epidemiology, 5
genetics, 121–131
formal, 122
future research, 130–131
genome-wide association studies,
122–123, 130
heterogeneity, 130
linkage studies, 122
onset, 121
prevalence, 121
and Tourette syndrome, 338
twin studies, 121
OCD see obsessive–compulsive
disorder (OCD)
odds ratios (ORs), 18
in generalized anxiety disorder, 5
in genome-wide association
studies, 27
offending, age of onset, 154
6-OHDA (6-hydroxydopamine), 177
olanzapine, 57–58, 276
olfactory neurons, 71
oligogenic modes, of inheritance, 6
Online Mendelian Inheritance in Man
(OMIM), 36–37
opioid addiction
and catechol-O-methyltransferase,
300–301
current research, 303
gene association studies,
hypothesis-driven, 302
genetics, 297–303
genome-wide association studies,
302–303
and 5-hydroxytryptamine-1B
receptor, 301
and kappa-opioid receptor, 299
and melanocortin receptor type 2,
299–300
methadone treatment, 302
and mu-opioid receptor,
297–298
risk factors, 297
and tryptophan hydroxylase 1,
301–302
and tryptophan hydroxylase 2,
301–302
opioid system, and alcoholism, 282
oppositional defiant disorder, 145
heritability, 149
prevalence, 149
symptoms, 149
twin studies, 149
OPRD1
and alcoholism, 282
and eating disorders, 265
OPRK1
and alcoholism, 282
and opioid addiction, 299
OPRM1
and alcoholism, 282
and opioid addiction, 297–298
ordered subset analysis
(OSA), 188
ORegAnno, 40–41
ORs see odds ratios (ORs)
OSA (ordered subset analysis), 188
OSBPL1A, and bipolar disorder,
206–207
OXTR, and autism, 190
P2RX7, and bipolar disorder, 204
P50 suppression test, 350
PAG see periaqueductal gray region
(PAG)
pairwise alignments, 42
PALB2, and bipolar disorder, 205
PANDAS (pediatric autoimmune
neuropsychiatric disorders
associated with streptococcal)
infections, 130
panic attacks (PAs)
induction, 96
and panic disorder, 91
panic disorder (PD), 90–107
and agoraphobia, 118
and brain abnormalities, 102
case control studies, 98
comorbidities, 90–92
diagnosis, 90
diagnostic criteria, 90
variance, 92
epidemiology, 90
family studies, 90–92
fear processing, 107
functional imaging studies,
102–104, 107
future research, 107
genes
in adenosine system, 95
in dopamine system, 94–95

in GABA-ergic system, 95–96
in G-protein signaling regulators,
95–96
in neuropeptide systems, 95
related to carbon dioxide
hypersensitivity, 96
in serotonin system, 93–94
genetic epidemiology, 5, 90
genetic studies, 90
conclusions, 96–97
issues, 107
limitations, 106
molecular, 92–97
genome-wide association studies,
93, 106
heritability, 92
imaging genetics, 104–106
linkage studies, 92, 99
neuroimaging studies, 102–106
and panic attacks, 91
and phobias, 118–119
prevalence, 90
probands, 101
and structural brain changes, 103
structural imaging studies, 102
symptoms, 90, 97
twin studies, 90–92
panic disorder syndrome
expanded spectrum approach,
97–102
study characteristics, 99
tests, 100
panic phenotype, studies, 106–107
papilin, encoding, 56–57
PAPLN, and suicidal ideation, 56–57
parent of origin effects, in obesity, 275
Paris Autism Research International
Sibpair Study (PARIS),
183–186
Parkinson’s disease, databases, 37–38
paroxetine
panic disorder studies, 93
pharmacokinetics, 53–54
PAs see panic attacks (PAs)
Pathguide, 45
PathSys database, 46
PATHWAY, 45
pathway analysis, 44–46, 70
commercial tools, 46
future trends, 46
information sources, 44
pathway data repositories, 45
pathways
autism, 192–193
as networks, 45
pathway visualization, 45–46
PATIKAweb, 46
PCLO, and major depressive
disorder, 221
PCL-R (Psychopathy Checklist-
Revised), 150–151
PCM1, and schizophrenia, 243
PD see panic disorder (PD)
PDGene, 37–38
PDYN, and alcoholism, 282
pediatric autoimmune
neuropsychiatric disorders
associated with streptococcal
(PANDAS) infections, and
obsessive–compulsive
disorder, 130
PENK, and alcoholism, 282
periaqueductal gray region (PAG),
102–103
perinatal events, Tourette
syndrome, 342
Perlegen, 25
personality
abnormal vs. normal compared, 316
epigenetics, 322
general dimensions, inheritance,
320–321
well-functioning, requirements, 317
personality disorders
candidate genes, 321
common genetic factors, 319–320
definitional issues, 317
factor analysis, 320
genetics, 316–322
genome-wide association studies,
321–322
heritability, 319–321
inheritance, 318–320
multidimensional inventories, 320
quantitative multifactorial
variability, 317–318
twin studies, 318, 320–321
see also antisocial personality
disorder
personality traits, heritability, 326
PET (positron emission
tomography), 102
Pfam, 43
pharmacodynamics, 53–54
pharmacogenetics
anticonvulsants, 59–61
antidepressants, 53–57
adverse reaction studies, 56
antipsychotic drugs, 57
adverse reaction studies, 58–59
attention-deficit hyperactivity
disorder, 176
animal models, 179
of carbamazepine-induced
Stevens–Johnson syndrome, 61
in clinical practice, 62
commercial tests, 65
conceptual issues, 61–65
Index
definition, 53
lithium, 59–60
major depressive disorder, 53, 225
future trends, 227
mood stabilizers, 59–61
profiling, 65
in psychiatry, 53–65
schizophrenia, 251
see also psychiatric
pharmacogenetics
pharmacokinetics
antidepressants
genetic effects on, 53–54
antipsychotic drugs, genetic
factors, 57
definition, 53
pharmacotherapy
applications, 57
psychiatric disorders, 53
and single nucleotide
polymorphisms, 61–65
phenotypes
concept of, 347
see also endophenotypes
phenotypic heterogeneity, 20
phenytoin, 59
phobias
and additive genes, 114–115, 117
classification, 112
comorbidities, 113
diagnosis, 112
epidemiology, 112–113
familial specificity, 114
familial transmission, 113
genetic epidemiology, 5, 113–118
future research, 119
genetics, 112–120
genomic studies, 118–119
heritability, sex differences, 114
and irrational fears, 113–115
meta-analyses, 103–104
and panic disorder, 118–119
prevalence, 112–113
twin studies, 114–115
variance, 116
see also agoraphobia (AG), social
phobia, specific phobia
phobic disorders see phobias
phonologic disorders, 160
phosphoacetylation, 81
phosphorylation, histones, 79–81
PICALM, and Alzheimer’s disease,
375
PKHD1, and obesity, 276
plasmids, transfection, 71
Plato, Republic, 324–325
pleiotropic models, 101
pleiotropy, 6
PLINK (software), 27
397
 Index
P-MATCH, 40
PMCH, and obesity, 276
PMut, 39–40
polygenic model, obesity, 275
polygenic modes, of inheritance, 6
polymorphisms
analysis tools, 37
nonsynonymous, 39–40
regulatory, 40–41
see also copy number
polymorphisms (CNPs); single
nucleotide polymorphisms (SNPs)
Polyphen, 39–40
POMC, and alcoholism, 282
population attributable risk, 8
population isolates, Tourette
syndrome studies, 339–340
population prevalence, 1–2
population stratification, 308–309
positron emission tomography
(PET), 102
postnatal brain maturation, 75
post-traumatic stress disorder (PTSD),
134–143
adoption studies, 136
animal models, 141
comorbidities, genetic factors, 139,
140
conceptual development, 134–135
diagnosis, 134
epidemiology, 135
diagnostic criteria, issues, 142
and diathesis–stress model, 134
epigenetics, 140–141
family studies, 135–136
and fear, 103–104
future trends, 140–142
and gene–environment
interactions, 142
genetic studies, 135–140
candidate gene approach, 137,
138
case-control association, 137–140
future trends, 142–143
methodological issues, 142
molecular, 137–140
quantitative, 135–136
genome-wide association studies,
141–142
historical background, 134–135
prevalence, 135
sex differences, 135
trauma exposure, 135–137
twin studies, 136, 140
PPP3C, and schizophrenia, 252
PPP3CC, and schizophrenia, 243
PPYR1, and obesity, 273
Prader–Willi syndrome, 79–81, 275
and autism compared, 365
398
genetics, 365–366
and genomic imprinting
disorders, 365
prevalence, 365
predictive algorithms, 40
PRINTS, 43
privacy issues, genetic
information, 329
PRNP, mutations, 378
PRODH, and schizophrenia,
247–248
prodynorphin gene, and stimulant
dependence, 312
PROSITE, 43
protein chemistry, in genetic
susceptibility factor studies,
69–71
protein kinases, 82
protein phosphatases, 82
proteins, 43
overexpression, 71
protein sequencing, 35
protein structures, 43
PRPF31, and autism, 188–189
PSEN1
and Alzheimer’s disease, 375
and familial Alzheimer’s disease,
371, 376
PSEN2
and Alzheimer’s disease, 375
and familial Alzheimer’s
disease, 373
pseudo single-molecule sequencing,
concept of, 28
PSI-MI, 45
psychiatric disorders
animal models, 73
and antisocial behavior, 145
biomarkers, identification, 49–50
candidate gene association
studies, 49
classification
endophenotypes, 9
lack of validity, 6
comorbidities, 7
complexity in, sources of, 6
and endophenotypes, 349
family studies, 69
gene expression studies, 49–51
genetic bases, methods, 23
genetic epidemiology of, 1–10
genetic susceptibility factors for,
functional validation, 69–77
heritability estimates, 3
impacts of genomics on, 9–10
and nicotine dependence, genetics,
294–295
and non-psychiatric disorders,
101–102
pharmacotherapy, 53
risk factors, 3
risk ratios, 3
transmission, complex patterns of, 6
psychiatric genetics
endophenotypes in, 347–360
applications, 349–359
future trends, 359–360
identification criteria, 348–349
psychiatric genetics research,
computer-based analysis and
resources, 34–47
psychiatric pharmacogenetics
field of study, 53
future trends, 61–65
implications for clinical practice,
61–65
psychiatric phenomics, field of
study, 49
psychiatric phenotypes,
limitations, 49
psychiatry
epigenetic research, issues, 87
pharmacogenetics in, 53–65
psychic trauma, 134–135
see also post-traumatic stress
disorder (PTSD)
psychopathy, 150–151
and antisocial behavior, 145
Psychopathy Checklist-Revised
(PCL-R), 150–151
psychopharmacology, limitations,
61–65
psychosocial stress, Tourette
syndrome, 342
psychostimulant dependence, 308–310
psychostimulants
use of term, 306
see also stimulant dependence
psychotic depression,
pharmacotherapy, 57
PTSD see post-traumatic stress
disorder (PTSD)
PubMed, 46
punishment, and behavioral genetics,
327–329
PupaSuite, 36
p-values, 50–51
QTL (quantitative trait loci), 275
quantitative multifactorial variability,
personality disorders, 317–318
quantitative trait loci (QTL), 275
RAI1, and Smith–Magenis
syndrome, 366
rare genetic variants, 23
autism, 191–192
obesity, 273

rare variant hypothesis, 191–192
RAVEN (Regulatory Analysis of
Variation in Enhancers), 40
RCSB PDB, 43
RD see reading disability (RD)
Reactome, 45
read depth (RD), 30–31
reading disability (RD), 161–163
comorbidities, 165
deficits, 166
future research, 166
gene localizations, 161
genetics, 161
and language impairment
compared, 164
sex differences, 161
read pairs (RPs), 30–31
recombination, and linkages, 14–15
recombination rates, non-uniform,
and mutations, 19
RefSeq, 45
regulator of G-protein signaling
(RGS), 96
Regulatory Analysis of Variation in
Enhancers (RAVEN), 40
regulatory motifs, 41
regulatory polymorphisms, 40–41
regulatory RNAs, 51
relative risk, 8
RELN, and schizophrenia, 243, 252
responsibility, and genetics, 326–329
Rett’s disorder, 364–365
Rett syndrome, etiology, 82
RGS4, and schizophrenia, 238–239
RGS (regulator of G-protein
signaling), 96
risk alleles, 8
risk estimation, 8
risk factors
antisocial behavior, 154–156
see also measured risk factors
risk ratios, psychiatric disorders, 3
risperidone, 58
RNA interference (RNAi), 71,
73–74
RNA sequencing, analysis, 35
ROBO1, and reading disability, 163
Roche, 454 Sequencing System,
28–29
Roche Diagnostic AmpliChip CYP450
test, 65
rodents
depression studies, 85
drug-addiction studies, 83–85
molecular manipulations, 73, 74
see also mice
RORA, and major depressive
disorder, 225
RPs (read pairs), 30–31
SAD (social anxiety disorder),
100–101, 104
sampling, family studies, 7
SAM (sequence alignment/MAP), 31
Sanger sequencing technology, 28–29
SAPAP3, 128–130
SBML, 45
schizophrenia (SZ)
adoption studies, 237
and affective disorders, 233–234
age of onset, 231–232
anatomical and histological
abnormalities, 76
animal models, 73, 76
anticipation and, 230–231
association studies, 238
and bipolar disorder, 230–231,
233–234, 208
copy number variant differences,
250–251
genetic association differences,
250–251
candidate genes, 204, 231, 238–248
animal and cell biology studies,
251–253
functional polymorphisms and
mutations, 251
children, 4
copy number variations, 249–250
co-twin control studies, 235–237
cytogenetic abnormalities, 249–250
deletions, 249–250
diagnostic criteria, 233–234
diagnostic issues, 232
duplications, 249–250
endophenotypes
eye-tracking dysfunction, 350
neurocognitive, 350–351
sensory motor gating, 350
environmental risk factors, 4
epigenetics, 231–232
familial recurrence, 232–233
familial transmission, 207
family studies
cognitive deficits and imaging
changes, 234
segregation analysis, 237–238
variance component analysis,
237–238
functional polymorphisms, 251
future research, 46–47
genetic epidemiology, 3–4
genetic linkage, 238
genetics, 230–253
future research, 253
genetic susceptibility factors, 4, 69
heritability, 230–231
and idiopathic generalized epilepsy,
249–250
Index
Kraepelinian type, 232–233
linkage studies, 231
loci, 239
murine studies, 251–253
neuropsychological deficits, with
family genetic control, 234–235
pharmacogenetics, 251
pharmacotherapy, 57
subtypes, family studies, 232–233
twin studies, 235, 350–351
segregation analysis, 237–238
variance component analysis,
237–238
SCOPE, 40
second-generation sequencing
see next-generation sequencing
segregation analysis
schizophrenia, 237–238
Tourette syndrome, 337
selection, artificial, 324
selective serotonin reuptake inhibitors
(SSRIs)
adverse reaction studies, 56
depression treatment, 85
major depressive disorder
treatment, 225
panic disorder treatment, 93
pharmacokinetics, 53–54
SEMA5A, and autism, 187
sensory motor gating, and
schizophrenia, 350
sequence alignment/MAP (SAM), 31
sequence alignments, 41–43
sequence analysis, 35
see also DNA sequencing
sequence conservation, 41
Sequenced Treatment Alternatives to
Relieve Depression (STAR*D)
Study, 55–56, 226
treatment emergent suicidal
ideation studies, 56–57
sequence read format (SRF), 31
sequencing, 28
future research, 46–47
see also DNA sequencing, next-
generation sequencing
Sequenom, 25
serotonergic pathways, attention-deficit
hyperactivity disorder, 172
serotonergic system, 58, 105–106
serotonin, in carbon dioxide
hypersensitivity, 96
serotonin norepinephrine reuptake
inhibitors (SNRIs)
adverse reaction studies, 56
depression treatment, 85
pharmacokinetics, 53–54
serotonin receptor gene (HTR2A)
see HTR2A
399
 Index
serotonin receptors, 107
polymorphisms, 93–94
serotonin receptor subunit genes, and
antipsychotic drugs, 58
serotonin system, genes involved in,
93–94, 123–126
serotonin transporter gene (SLC6A4)
see SLC6A4
serotonin-transporter-linked
polymorphic region
(5-HTTLPR), 54–55
and adverse reactions, 56
anorexia nervosa studies, 267
bipolar disorder studies, 203
major depressive disorder
studies, 224
obsessive–compulsive disorder
studies, 123–126
and personality disorders, 321
polymorphisms, 93, 105–106
post-traumatic stress disorder
animal studies, 140
studies, 140
see also SLC6A4 SERT,
organization, 125
setshifting, and anorexia nervosa, 263
sex differences
animal phobias, 115
antisocial personality disorder,
317–318
autism, 188, 191
fragile X syndrome, 367
major depressive disorder,
215–216
phobias
heritability, 114
prevalence, 113, 115
post-traumatic stress disorder, 135
reading disability, 161
Tourette syndrome, 342
shank3 point mutation, 192
shell shock, 134–135
see also post-traumatic stress
disorder (PTSD)
SHRs (spontaneously hypertensive
rats), 178
Sibling-pair designs, autism studies,
183
SIFT, 39–40
simple phobia see specific phobia
Single Nucleotide Polymorphism
Database (dbSNP), 39
single nucleotide
polymorphism (SNP)
genotyping, 23–28
copy number analysis, 26
family studies, 26–27
genome-wide association studies,
27–28
400
linkage studies, 26–27
see also high-density single
nucleotide polymorphism (SNP)
genotyping
single nucleotide polymorphisms
(SNPs), 13
applications
disease mapping, 23–24
early, 24–25
classic, 24
classification, 24
databases, 36–39
definition, 23–24
functional, 24
linkage disequilibrium, 24
mapping, 24
nonfunctional, 24
and pharmacotherapy, 61–65
site-directed mutagenesis, 69–70
SJS (Stevens–Johnson syndrome),
carbamazepine-induced,
pharmacogenetics, 61
SLA6A2, animal models, 177
SLC1A1
in obsessive–compulsive disorder,
123, 130
studies, 124
SLC1A3, in attention-deficit
hyperactivity disorder,
172–173
and schizophrenia, 252
SLC6A2
and attention-deficit hyperactivity
disorder, 172
and major depressive disorder,
55–56, 218–219
and weight loss, 276
SLC6A3, 137–139
animal models, 176–178
in attention-deficit hyperactivity
disorder, 169, 173, 176
and stimulant dependence, 310
variants, 169–171, 175
SLC6A4
anorexia nervosa studies, 267–269
and autism, 190
bipolar disorder studies, 203
major depressive disorder studies,
218–219
obsessive–compulsive disorder
studies, 123–126
panic disorder studies, 93
pharmacogenetic studies, 54–55
post-traumatic stress disorder
studies, 140
see also serotonin-transporter-
linked polymorphic region
(5-HTTLPR)
SLI1 locus, 164, 166
SLI2 locus, 164, 166
SLI see specific language
impairment (SLI)
SLITRK1, and Tourette syndrome,
336–337, 341
Slit and Trk-like family member 1
(SLITRK1) see SLITRK1
Smith–Magenis syndrome
(SMS), 366
Smith–Waterman algorithm, 42
smoking initiation, heritability, 288
SMS see Smith–Magenis
syndrome (SMS)
SNAP-25, and attention-deficit
hyperactivity disorder,
172–173
SNPExpress, 44
SNP genotyping see single nucleotide
polymorphism (SNP)
genotyping
SNPs3D, 39–40
SNPs see single nucleotide
polymorphisms (SNPs)
SNPs (single nucleotide
polymorphisms), 13
SNRIs see serotonin norepinephrine
reuptake inhibitors (SNRIs)
social anxiety disorder (SAD),
100–101, 104
social phobia
and avoidant personality disorder
compared, 117
classification, 112
diagnosis, 112
familial transmission, 113
genomic studies, 118
prevalence, 112–113
traits, family studies, 117–118
see also generalized social
phobia
sodium butyrate, 224
SP4, and major depressive
disorder, 221
spasmodic torticollis, 127
SPCH1 locus, 164
specific environment, use of
term, 114
specific language impairment (SLI),
160
specific phobia
classification, 112
diagnosis, 112
familial transmission, 113
genomic studies, 118
prevalence, 112–113
specific reading disability (SRD), 160
speech sound disorder (SSD), 165–166
definition, 160
prevalence, 160
 Index

spontaneously hypertensive rats subPSEC, 39–40 tobacco use

(SHRs), 178
SRD see specific reading
disability (SRD)
SRF (sequence read format), 31
SSD see speech sound
disorder (SSD)
SSRIs see selective serotonin reuptake
inhibitors (SSRIs)
stable tubule-only polypeptide
(STOP) proteins, and
schizophrenia, 252
STAR*D Study see Sequenced
Treatment Alternatives to
Relieve Depression (STAR*D)
Study
state–trait characteristics
anorexia nervosa, 263–264
bulimia nervosa, 263–264
stem cells, 71
see also induced pluripotent
stem (iPS) cells
STEP-BD (Systematic Treatment
Enhancement Program for
Bipolar Disorder), 205–206
sterilization, 325
Stevens–Johnson syndrome (SJS),
carbamazepine-induced,
pharmacogenetics, 61
stimulant dependence
candidate genes, 312
and dopamine-related genes,
310–312
family studies, 306–307
future research, 314
genetics, 306–314
genome-wide association studies,
312–313
heritability, 306–308
markers, 309–310
prevalence, 306
twin studies, 306–307
see also psychostimulant
dependence
STOP (stable tubule-only polypeptide)
proteins, and
schizophrenia, 252
stratification, 20
streptococcal infections, and
obsessive–compulsive
disorder, 130
structural alignments, 43
structural imaging studies, panic
disorder, 102
structural variants
attention-deficit hyperactivity
disorder, 173
types of, 30–31
STY13, and cocaine dependence, 313
substance dependence, and antisocial
behavior, 145
substance use disorders, 148
and antisocial behavior, 150
family studies, 5, 6
gene–environment interactions,
150
genetic epidemiology, 5
heritability, 150
twin studies, 5–6
suicidal behavior, and cortisol
levels, 353
suicidal ideation, 56–57
Systematic Treatment Enhancement
Program for Bipolar Disorder
(STEP-BD), 205–206
SZ see schizophrenia (SZ)
TAAR6, and schizophrenia,
241–243
TACR3, and alcoholism, 283
TAMAL, 36
tamoxifen, 276
TaqI, 172
tardive diskinesia (TD), antipsychotic-
induced, 58–59
target enrichment, hybridization-
based, 32
TCAs see tricyclic antidepressants
(TCAs)
TCI (Temperament and Character
Inventory), 320–321
T-Coffee, 42
TD see tardive diskinesia (TD)
TDTs see transmission disequilibrium
tests (TDTs)
Technical University of Denmark,
Center for Biological Sequence
Analysis, 43
Temperament and Character
Inventory (TCI), 320–321
TEN (toxic epidermal necrolysis),
carbamazepine-induced,
60–61
TESIs (treatment emergent suicidal
ideations), antidepressant-
induced, 56–57
testing
for-profit, for genetic susceptibility,
331–332
issues, addictions, 331
in workplace, 333–334
TF, and Alzheimer’s disease, 376
TH (tyrosine hydroxylase), 201–202,
204
tics
characterization, 336
comorbidities, 337–338
prevalence, 287
see also nicotine dependence
Tourette syndrome (TS)
and anxiety disorders, 338
and attention-deficit hyperactivity
disorder, 338
candidate genes, 341
comorbidities, 337–338
copy number variations, 343
cytogenetic abnormalities, 340–341
and depression, 338
environmental factors, 342
epigenetics, 343–344
family studies, 337
genetics, 126–127, 336–344
future research, 344
future trends, 342–344
genome-wide association studies,
342–343
heritability, 337
infection and immune response, 342
linkage analysis, 338–339
microarray studies, 342–343
noncoding RNA studies, 343
non-parametric linkage
method, 339
and obsessive–compulsive
disorder, 338
perinatal events, 342
population isolates, 339–340
prevalence, 337
psychosocial stress, 342
research, conclusions, 336
segregation analysis, 337
sex differences, 342
studies, 336
symptoms, 336
twin studies, 337
toxic epidermal necrolysis (TEN),
carbamazepine-induced,
60–61
TPH1, 301–302
TPH2, 301–302
TPH (tryptophan hydroxylase), 94
TRa1, mutations, 178
training data, 40
traits, heritability, 113
TRANSFAC database, 40
transgenic animals, 178
transgenic mice, 73
Alzheimer’s disease studies, 374
translocations, 30–31
transmission disequilibrium tests
(TDTs), 18–19
advantages, 27–28
overview, 27–28
tranylcypromine, 86, 224
TRAR4, and schizophrenia, 241–243

401
 Index
trauma exposure
assaultive versus non-assaultive,
136–137
epidemiology, 135
genetic factors, 136–137
non-genetic factors, 137
traumatic neurosis, 134–135
see also post-traumatic stress disorder
(PTSD)
treatment emergent suicidal ideations
(TESIs), antidepressant-
induced, 56–57
trichotillomania, genetics, 128–130
tricyclic antidepressants (TCAs), 53
TRRD database, 40
tryptophan hydroxylase 1 (TPH1)
see TPH1
tryptophan hydroxylase 2 (TPH2)
see TPH2
tryptophan hydroxylase (TPH), 94
TS see Tourette syndrome (TS)
TStag, 44
TTC12, and alcoholism, 281–282
T variant, 15–17
twin studies
alcoholism, 279–280
anorexia nervosa, 264–265
antisocial behavior, 145–148,
156
gene–environment
interactions, 152
antisocial personality disorder, 317
attention-deficit hyperactivity
disorder, 149–150, 168, 176
autism, 183
bipolar disorder, 196, 197
bulimia nervosa, 264–265
carbon dioxide hypersensitivity, 96
conduct disorder, 149
externalizing disorders, 116–117
extraversion, 117
generalized anxiety disorder, 119–120
genetic epidemiology, 2
heritability, 14
internalizing disorders, 116–117
major depressive disorder, 214–215,
224
neuroticism, 117
nicotine dependence, 288
obesity, 272
obsessive–compulsive disorder, 121
oppositional defiant disorder, 149
402
panic disorder, 90–92
personality disorders, 318,
320–321
phobias, 114–115
environmental variance, 114
genetic variance, 114
prevalence, 113
post-traumatic stress disorder, 136,
140
psychopathy, 151
schizophrenia, 235, 350–351
segregation analysis, 237–238
variance component analysis,
237–238
stimulant dependence, 306–307
substance use disorders, 5–6
Tourette syndrome, 337
see also dizygotic (DZ) twins,
monozygotic (MZ) twins
tyrosine hydroxylase (TH), 201–202,
204
UBE3A, and Angelman syndrome, 365
UBE3C, and major depressive
disorder, 225
UCP1, and obesity, 273
UCSC Genome Browser, 40–41
UHMK1, and schizophrenia, 238–239
unconditioned fear, 102–103
unipolar depression see major
depressive disorder (MDD)
U.S., v. Durham, 327–328
Val66Met, polymorphisms, 140
val158met, polymorphisms, 105
valproate, 59, 224
valproic acid, 299
Val/Val genotype, 172
variable expressivity, 6
variable number tandem repeats
(VNTRs), 94, 203
variance component analysis,
schizophrenia, 237–238
variant consensus format (VCF), 31
variant identification
in human genome, 13
strategies, 34
variant segregation, and linkage
disequilibrium, 15–17
VBRs (ventricle-brain ratios), 234
VCFS see velo-cardio-facial syndrome
(VCFS)
VCF (variant consensus format),
31
velo-cardio-facial syndrome (VCFS),
4, 249–250
genetics, 127–128, 247–248
venlafaxine, pharmacokinetics,
53–54
ventricle-brain ratios (VBRs), 234
vesicular neurotransmission,
regulatory mechanisms,
172–173
violence, 145
genetic factors, 328–329
viral vectors, infection, 71
VisANT, 46
VISTA Enhancer Browser, 40–41
VMP, and reading disability, 163
VNTRs (variable number tandem
repeats), 94, 203
warfarin, pharmacogenetic
profiling, 65
war neurosis, 134–135
see also post-traumatic stress
disorder (PTSD)
WDR59, and bipolar disorder,
206–207
WebLogo, 41
weight gain
antipsychotic-induced, 58–59
drug-induced, 276
weight loss, genetics, 276
Weka, 40
WGAS see genome-wide association
(GWA) studies
WGA studies see genome-wide
association (GWA)
studies
Wistar Kyoto (WKY) strain, 178
word methods, 42
workplace, testing in, 333–334
WorldWide Protein Data
Bank, 43
X-inactivation, 79–81
yeast two-hybrid assays, 69–71
zebrafish, as animal models, 73
zinc finger proteins, 87–88
ZNF804A, and schizophrenia,
239–240