You are on page 1of 9

REVIEWS AND OVERVIEWS

Thinking About Schizophrenia in an Era of


Genomic Medicine
Daniel R. Weinberger, M.D.

Genetic discoveries about human brain development and epistemological questions about classification. In clinical re-
neuropsychiatric syndromes have changed the landscape of search, the application of polygenic risk scores—the cumu-
psychiatric research. The genotyping of hundreds of thou- lative sum of associated alleles in an individual genome—to
sands of individuals has identified many hundreds of genomic prediction models of environmental influences and outcome
regions that are associated with psychiatric diagnoses, and is gaining enthusiasm because these scores explain more liability,
progress is being made in uncovering the specific genes that but predictions are small and not yet actionable for individuals.
underlie these statistical associations, although most are still Psychiatry is intertwined with genomic medicine, and our
undetermined. While there are great expectations that such understanding of what we call schizophrenia and the possi-
genetic discoveries will lead to novel treatments based on bilities of improving the lives of affected individuals have never
fundamental mechanisms of illness, there are important ca- seemed more promising. Yet the research challenges are
veats. Individual risk-associated common variants explain daunting; psychiatric syndromes ultimately reflect how the
only a tiny fraction of individual liability. The degree to which brain mishandles environmental information, which at the
common risk variants represent “core” pathogenic insights is systems level is far “downstream” of the effect of genes in cells.
controversial. Individuals with rare and more penetrant risk
variants are often intellectually disabled, which raises Am J Psychiatry 2019; 176:12–20; doi: 10.1176/appi.ajp.2018.18111275

Why are genes important in psychiatry? The simple answer is Causation, however, is not discernible from these phenom-
the same reason they are important in other areas of biology enological approaches. In contrast, genes are not merely
and medicine—because most human traits and clinical dis- phenomenological observations. They are the proverbial
orders have a substantial genetic component. Genes influence “egg” rather than the “chicken.” The study of gene func-
and encode mechanisms of disease; they provide objective tion and genetic variation generates insights into etiolog-
insights into individual predisposition or risk; they may help ical mechanisms and into the root causes of psychiatric
contextualize the interacting environment; and they are a syndromes.
potential tool for the discovery of treatments based on cau- Genes are also important because they make it possible
sation and pathogenesis. to begin to unravel the critical contributory role of the in-
It has been clear for more than a century that psychiatric teracting environment in complex disease. Genes are finite
syndromes run in families. Archival studies of concordance and quantifiable biological entities; the environment, by con-
differences between identical and fraternal twins and oc- trast, is potentially far more complex, and without an under-
currence of schizophrenia in individuals who were adopted standing of an individual’s discrete biological background
away at birth and in their biological parents led to the con- (i.e., “genome”), the environment is a much more nebulous
clusion that the main reason psychiatric diagnoses run in and indefinable entity. If life is imagined as a game of Mo-
families is because of inheritance rather than culture or nopoly, genes will never determine how many railroads one
family dynamics. Psychiatry even before Freud has been owns or where hotels will be built. But genes are the only
essentially a phenomenological discipline, based on differ- square on the board that can be known with certainty—they
ences that can be observed between individuals with a are the “Go” square. Genes represent the biological tool box
particular diagnosis and those without one. Such phenom- with which one plays the game, the underlying basis of in-
enology has been rich and compelling, with persuasive dividual characteristics that interact with the environment
inferences about mechanisms from various schools of de- and lead to personal gain or resilience to loss. Daunting stress
scriptive psychology and psychopathology and, more re- to one individual may be an exciting opportunity to another.
cently, evidence for biological associations from molecular As Paul Simon sang it, “one man’s ceiling is another man’s
studies of patient tissues and from neuroimaging studies. floor.”

12 ajp.psychiatryonline.org Am J Psychiatry 176:1, January 2019


WEINBERGER

There is abundant excitement in the research and phar- presumed to contain genes that increase susceptibility to
maceutical industry communities about the prospect of ge- schizophrenia.
netic discoveries leading to novel treatments based on The finding of genome-wide statistically significant loci
elucidation of fundamental mechanisms of illness. Finding a for schizophrenia has transformed the landscape of research
disease-associated gene and translating that discovery into a in psychiatry, representing compelling evidence of the po-
therapeutic agent is a daunting challenge, and so far, in all of tential biological underpinnings of schizophrenia at a basic
medicine, very few discoveries of disease-associated genes molecular level. The PGC2 study has recently been expanded
have led to novel treatments. In psychiatry, the treatments (“PGC3”), and, as reported at the 2018 World Congress of
currently used, be it talk, electricity, magnetism, or phar- Psychiatric Genetics in Glasgow, with 30,000 additional
macology, were discovered through understanding of cause subjects, the number of GWAS-significant loci is now 246.
and pathogenesis; rather, nearly all were identified seren- As expected, most of the new loci were near statistical sig-
dipitously. Having insights into mechanisms of psychiatric nificance in the PGC2 version, confirming that increasing
illness based on genetic information will surely be a more sample size will tend to elevate near statistically significant
successful approach to therapeutic discoveries. loci into the rarefied range of GWAS significance.

DISCOVERY OF SCHIZOPHRENIA SOME CAVEATS ABOUT COMMON VARIANTS


RISK–ASSOCIATED GENETIC VARIANTS
While increasing the size of study samples has successfully
The landscape of genetic research in psychiatry has changed provided increased power to identify loci that achieve sta-
dramatically over the past decade. From family studies that tistical significance, larger samples do not increase the effect
were state of the art during the closing decades of the size, as reflected by the odds ratio of an individual risk-
20th century, the current fashion in medical genetics is the associated SNP. Even the SNPs with the greatest statistical
genome-wide association study (GWAS) comparing the DNA significance increase risk on an individual basis only to a
“genotypes” of clinical case subjects with control subjects. minuscule degree. Most of the SNPs increase risk less than
This approach is based on technological advances that allow 1.1-fold from general population risk (1, 2). Moreover, the
inexpensive and high-throughput analysis of sequence var- differences between patient and control samples in the fre-
iation across the human genome in DNA extracted from quency of alleles associated with risk is typically no more than
blood or saliva. Genotyping millions of common sequence 2%. Thus, by themselves, individual GWAS-significant SNPs
variants, called single-nucleotide polymorphisms (SNPs), have little or no role as predictors of a person’s clinical risk.
across the 3 billion nucleotides of a person’s genome can As sample sizes grow even further, the odds ratios of newly
now be accomplished for little more than $100. significant GWAS loci will predictably be even smaller, and
To facilitate discovery of genetic factors with rigorous it is conceivable that within a few years, we will have in our
statistical objectivity, based on no a priori expectation that gene quivers many hundreds of associations that increase risk
any particular sequence variant will be associated with a fractions of a fraction. This is an important caveat to overly
particular trait or clinical illness, large samples have been enthusiastic interpretations of existing data and promises of
acquired through collaborative consortia of investigators future GWAS analyses. It also is important to realize that the
around the world. Indeed, GWAS has been applied to hun- genetic analyses of this large case-control cohort charac-
dreds of traits and complex diseases to date, successfully terize primarily individuals of European ancestry. Compar-
yielding statistically significant risk-associated genomic re- isons with other populations will require similar large studies
gions for cancers, endocrinological disorders, and cardio- of those populations, and preliminary evidence from Asian
vascular diseases, as well as for psychiatric illnesses, including samples, although consistent to a moderate degree in terms of
schizophrenia, autism, mood disorders, posttraumatic stress loci and effect sizes, has uncovered substantial divergences
disorder, and eating disorders. (3). In African American samples, the attribution of pop-
The current state of the art in schizophrenia genetics is ulation risk based on common alleles identified in European
the product of the schizophrenia working group of the Psy- ancestry samples is likely to be much less consistent.
chiatric Genomics Consortium (PGC). In what is called It is also important to appreciate that the risk loci iden-
the PGC2 schizophrenia study (1, 2), which analyzed over tified are not themselves susceptibility genes. They are rel-
40,000 individuals with schizophrenia diagnoses and over atively low-resolution map points of regions of the genome
64,000 individuals identified as “controls” from over that enable more detailed analysis to identify the specific
50 centers in Europe and the United States, SNP allele fre- genetic components accounting for the statistical associa-
quencies that significantly differ between the patient and tion between DNA variation and clinical status. Most of the
control samples have been identified in 145 regions (called schizophrenia-associated loci contain more than one gene,
“loci”) across the genome. Because millions of SNPs were and often contain many genes. From the physical gene map
surveyed for allele frequency differences between the sam- and the clinical statistics alone, it is impossible to conclusively
ples, a statistical threshold for potential false positives was identify the causative gene in any locus. This task is further
set at a conservative p value of 531028. These regions are complicated by the fact that most of the risk-associated SNPs

Am J Psychiatry 176:1, January 2019 ajp.psychiatryonline.org 13


THINKING ABOUT SCHIZOPHRENIA IN AN ERA OF GENOMIC MEDICINE

are not in the coding portion of any known genes, meaning Most of the genes within these loci, however, encode proteins
that they do not directly change the amino acids in the involved in biological and molecular processes not previ-
protein. Rather, they may be located within regulatory do- ously emphasized as potentially related to schizophrenia
mains in the DNA that influence gene expression. While the or any psychiatric disorder. Many genes encode proteins in
exons of most genes encode proteins, the vast majority of the intermediate metabolism and the core processes of cell sig-
genome is noncoding. Recent work by the National Institute naling and cell cycle biology, while many others reflect
of Health ENCODE and National Institute of Mental Health cellular functions about which very little is known. In an
PsychENCODE projects (4, 5) has demonstrated that much effort to reduce the potential biological complexity, if not
of the noncoding genome is involved in regulating how genes the opacity, of the diverse functions involved and find
are expressed, and individual noncoding variants in “en- common biological pathways of the sets of genes contained in
hancers” can influence the processing of genes many thou- the GWAS loci, investigators have applied bioinformatic
sands of nucleotides from their physical location in the tools, hoping to find some functional convergence. The as-
genome. In other words, because a SNP of interest lies near or sumption here is that despite the apparent biological diver-
even inside the physical boundaries of a gene does not sity of the implicated genes, there may be common functions
necessarily mean that it is a factor in the biology of that gene. that underlie core phenotypes. The bioinformatic tools
employed in these “in silico” studies give only crude ap-
proximations of biological reality, as they are based on the
EARLY LESSONS FROM COMMON RISK
available research literature and databases and often on
VARIANT DISCOVERY
peripheral, but not CNS, biological data. Nevertheless, the
Although the current data only scratch the surface of the results have suggested that some of the gene sets within
underlying biology of risk, the schizophrenia GWAS results schizophrenia GWAS-significant loci converge on several
have arguably taught us a number of seminal lessons already. quite general aspects of cell biology, including calcium
First, there is no single schizophrenia gene, per se. Individual channel signaling, synaptic function, and basic neurodevel-
SNPs indicate minor components of risk, not fate. Second, opmental processes (11–13). These results are consistent
there is no royal road to risk. The many loci that are high- with the general idea that genes associated with a behavioral
lighted span the entire genome, involving every chromosome, disorder influence the development of the brain and the
and a schizophrenia-associated locus is likely to be found in plasticity of cells, particularly how they “talk to each other”
virtually any 5 million base pair section of the genome. In- and respond to environmental stimuli. However, as appealing
deed, there are many roads that potentially converge to as these results are, the validity and utility of conclusions
contribute to risk, a characteristic of most so-called polygenic about discrete pathogenesis or discrete pathophysiological
disorders. Third, the GWAS loci associated with schizo- mechanisms derived from bioinformatic gene set and path-
phrenia are not unique to schizophrenia. Many of the same way analyses should be interpreted with caution. The bi-
loci have been associated with other psychiatric diagnoses, ological roles that genes play in brain function are highly
including bipolar disorder, autism, and attention deficit hy- context dependent, spatially (cell and tissue dependent),
peractivity disorder (ADHD) (6, 7). It should come as no temporally (in terms of development and time of life), and
surprise that the human genome did not evolve to validate in terms of background (what other genes are doing in the
the DSM-5 or ICD-10 schema of psychiatric nosology. Fourth, same cell).
many SNPs associated with schizophrenia appear to influ- Another lesson learned from the large-scale schizophrenia
ence the epigenetic state of the genome, that is, the molec- GWAS is that the heritability of schizophrenia—that is, the
ular machinery that regulates gene expression by changing fraction of risk accounted for by common genetic variation as
chromatin conformation and methylation of DNA (8–10). analyzed under simple additive models—is not explained by
Epigenetic factors respond to environmental events and the cumulative associations reported to date, nor even by the
underlie gene-by-environment interactions that are critical GWAS results expected over the next decade. Indeed, the
for development and aging, for learning, and for shaping significant SNPs in the PGC2 study explain only a few percent
behavior patterns based on experience. These studies suggest of heritability, and predictions are that less than 30% will be
that many genetic variations related to risk for schizophre- explained by GWAS findings in the coming years (1, 2). The
nia (and likely other psychiatric diagnoses) influence the reasons for this apparent “missing heritability” are unclear
nature and sensitivity of the genomic response to environ- and the subject of considerable debate. One view holds that
mental events. the model for calculating heritability is based on a simple
Fifth, most genes located within GWAS-significant loci are linear addition of risk factors and does not accurately rep-
not genes that confirm earlier hypotheses about the biology resent how genes influence biology. In most nonhuman
of schizophrenia. Notwithstanding the aforementioned un- contexts, including in yeast, worms, flies, and rodents (14–17),
certainty of which, if any, genes within a locus are actually and also in human cancer cells (18), genes interact in non-
causative, there are a handful of genes in glutamate and GABA additive ways—biological processes in living organisms tend
signaling pathways in GWAS loci, and one GWAS-significant not to be linear. Whether this is a major factor in the ar-
locus contains the dopamine type 2 receptor gene (DRD2). chitecture of schizophrenia risk is unknown.

14 ajp.psychiatryonline.org Am J Psychiatry 176:1, January 2019


WEINBERGER

A recent and provocative hypothesis, the “omnigenic diagnosis have a 22q11 hemideletion (21). A dozen or so
hypothesis” (19), opines that heritability related to common other such recurrent CNVs have also been found to be
genetic variation will eventually be more fully explained more frequent in individuals with schizophrenia diagnoses
when sample sizes are in the millions, based on the as- than in control groups, but these CNVs have smaller effect
sumption that variation in any gene expressed in a relevant sizes on risk (21). Protein-altering SNVs that have been
cell type will contribute to risk. The omnigenic hypothesis observed in patients diagnosed with schizophrenia are
argues in effect that any gene expressed in a neuron matters to much rarer, and to date only one gene has been statistically
the biology of that neuron, and because cell function depends confirmed as being a schizophrenia risk factor based on rare
on many layers of interacting protein networks, any gene SNVs (24).
expressed in a relevant cell will exert at least a small influ- It is important to note that the CNVs and rare SNVs as-
ence on the function of multiple networks in that cell. While sociated with schizophrenia are not specific to schizophrenia
some networks are presumed to contain more pathogenically but rather are also associated with other psychiatric and
critical, or “core” genes, most of the genes in GWAS loci will medical illnesses, including autism, ADHD, seizure disor-
contribute to risk simply because they are expressed in the ders, and, in particular, intellectual disability. Indeed, in-
relevant cell, even though they are relatively peripheral to dividuals with schizophrenia-associated CNVs are more
core pathogenesis. In other words, according to this hy- intellectually disabled than patients without these genetic
pothesis, variations in most genes expressed in neurons will variations (25). Moreover, in the few reported cases of in-
contribute to the heritability of schizophrenia and eventually dividuals with schizophrenia diagnoses who have mutations
be in GWAS-significant loci (each with a trivial effect size) in SETD1A, which is the only gene so far statistically con-
but will be of minimal biological relevance to understand- firmed as a risk factor based on a protein-altering SNV, a
ing the pathogenesis of schizophrenia. This also is a con- history of developmental disability (e.g., epilepsy and in-
troversial concept, in part because it is so challenging to the tellectual disability) is typical (24). There is reason to
prevailing GWAS zeitgeist. question whether the psychosis in these cases is a primary
phenomenon based directly on the genetic variant or a sec-
ondary associated feature often observed in individuals with
RARE VARIANTS AND SCHIZOPHRENIA RISK
intellectual deficits. It is unclear whether loss-of-function
Genome-wide genotyping and, more recently, exome and SNVs and large CNVs may be found in patients with
whole genome DNA sequencing studies have discovered that schizophrenia diagnoses who do not have intellectual defi-
approximately 2%23% of patients with schizophrenia di- cits, and claims to the contrary must be viewed with some
agnoses have rare variations in their genomes that have skepticism (25). Cognitive function is considered “normal”
deleterious effects on protein expression or sequence and in across a broad range, but for any given individual, being
general have a greater influence on risk—that is, are more within that range does not mean that function is normal for
penetrant—than individual SNPs (20–22). These include that individual. As twin and family studies have shown, in-
relatively large DNA segments that are deleted or duplicated, dividuals with seemingly normal cognition and a schizo-
called copy number variants (CNVs), and loss-of-function phrenia diagnosis typically have poorer cognition than their
mutations of individual coding nucleotides. When DNA is siblings and other family members, even before clinical di-
copied, during meiosis or even during cell division, occa- agnosis (26). The 22q11 hemideletion syndrome, the most
sional errors occur, and sections of DNA may be duplicated penetrant CNV associated with “schizophrenia,” highlights
or deleted, or single nucleotides may be mutated (single- the nosological ambiguity that underlies diagnosis of the
nucleotide variants, or SNVs). If the mutation alters a protein schizophrenia syndrome. Recall that CNS syphilis was di-
or if the CNV is large enough, and typical risk-associated agnosed as “schizophrenia” at the turn of the 20th century
CNVs are on the order of 100,000 bases or more, genes within before the discovery of the spirochete. The velocardiofacial
the altered region will be either deleted or duplicated. Be- syndrome is associated not only with intellectual deficits
cause most of these more penetrant rare variants tend to arise but also with cardiac abnormalities, CNS malformations,
during gametogenesis, they are typically found on one pa- and facial dysmorphia. Studies of psychosis in individuals
rental chromosome, so affected individuals are heterozygotic with major intellectual disability, which used to be called
for the variation. It is thought that every person has at least “pfropfschizophrenie,” have shown that the phenomenology is
one large CNV in their genome and probably several hundred highly similar to that of patients with schizophrenia whose
protein-altering SNVs (23). Most of these do not “cause” a intellectual function is less impaired (27). We are once again
clinical disorder. The most common CNV that consistently confronted with an epistemological conundrum because
does so is the 22q11 hemideletion of approximately 3 million of the inaccuracies of our phenomenologically based syn-
bases associated with the velocardiofacial syndrome, a con- dromes. Furthermore, to the extent that many deleterious
dition in which approximately 30% of individuals will meet CNVs and mutations arise during gamete differentiation
criteria for a diagnosis of schizophrenia in early adulthood. and are thus “de novo” and not inherited (i.e., not in the pa-
Large-scale studies of patients with schizophrenia diagnoses rental genome), it is unlikely that they contribute to missing
suggest that approximately 0.5% of individuals with this heritability.

Am J Psychiatry 176:1, January 2019 ajp.psychiatryonline.org 15


THINKING ABOUT SCHIZOPHRENIA IN AN ERA OF GENOMIC MEDICINE

As if de novo mutations and segmental duplications and particular time of life (33) or perhaps even a specific cell type.
deletions in gamete formation did not add another non- Figure 1 illustrates this problem for a highly statistically sig-
Mendelian level of genetic complexity to human develop- nificant risk-associated locus, where many genes are mapped
ment, it has also become apparent that while every cell in a to the region. RNA sequencing and molecular characterization
human body is a daughter of the fertilized egg, not every cell of transcripts in the region identified two genes underlying
has the same genome. This is called somatic mosaicism. The the clinical association—BORCS7 and AS3MT—and for
same mistakes that can be made in DNA replication during AS3MT, the transcript was a previously unknown, human-
meiosis can be made during cell division and perhaps even unique isoform of the gene (30). To understand the molecular
during gene expression when DNA is replicated or repaired. mechanisms of clinical association, precise characterization of
If these mosaic events occur early in development, they will the transcript association is necessary. For most GWAS loci,
characterize many subsequent daughter cells, but if they this degree of specification is not yet available.
occur relatively late, for example, during the final stages of In the meantime, while the biological research commu-
neuronal migration in the brain, they will affect only a small nity engages the tools of functional genomics to elucidate
group of cells. The number of cells affected with a particular the biological import of common and rare schizophrenia-
mutation or CNV is in essence a lineage marker for the his- associated risk variants, the clinical psychiatry community
tory of that assembly of cells. Facial hemangiomas are a classic has begun to stratify patients by genetic risk and, in particular,
example of somatic mosaicism arising late in face maturation. by a cumulative sum of risk-associated variants across the
Single-cell DNA sequencing studies in brain neurons have genome, called the polygenic risk score. The clinical research
revealed that as many as 20%230% of such cells have vari- landscape is beginning to be transformed by the availability
ations in their genome, including SNVs and CNVs (28). While of genetic risk factors in prediction models of risk, of envi-
rare variations in individual neuronal genomes are found in ronmental influence, and of outcome.
the human brain, their role in CNS disorders, with rare ex-
ceptions (29), has yet to be determined.
POLYGENIC RISK SCORE
One of the principal goals of genetic counseling is to advise
BEYOND SCHIZOPHRENIA RISK VARIANT
about risk. As mentioned above for schizophrenia, rare ge-
DISCOVERY
nomic variants (CNVs) tend to be more penetrant compared
The landmark discovery of common and rare variants as- with common genomic variants. The velocardiofacial syn-
sociated with schizophrenia has initiated a tidal wave of drome deletion may increase risk of a diagnosis of schizo-
biological research to identify risk genes in the implicated phrenia up to 30-fold, and other identified rare variants
genomic regions and elucidate specific disease mechanisms associated with schizophrenia increase risk on average 5- to
and disease pathways. Ongoing biological research, includ- 10-fold. While this is potentially of value in clinical coun-
ing analyses of gene expression, epigenetics, proteomics, seling, the odds are still much greater that schizophrenia will
and single cellular analyses, as well as animal modeling, has not occur in the context of these rare variants than that it will.
intensified as a result of these recent genetic discoveries. Each common variant identified in GWAS, in contrast, in-
However, translating a GWAS positive locus into meaning- creases risk only a tiny amount and thus cannot be used to
ful or actionable biology or a clinical translation or even a predict risk. However, single variants are inherited on a
specific genetic entity is not straightforward and requires background of millions of other variants in the genome, and
extensive further research. It is not a simple task to discover genetic risk is presumably a cumulative manifestation of
how a SNP identifying a GWAS locus relates to the biologi- “genome type.”
cal activity of genes within the tissue thought to be of In an effort to achieve a more meaningful prediction of
pathogenic relevance—the brain in the case of schizophre- genomic risk, a polygenic risk score (PRS) has been applied in
nia. Genetic variants that regulate expression can engage many areas of medicine (34, 35). The PRS is an approach to
many molecular mechanisms accounting for their regulatory assigning to an individual a numerical score that quanti-
effects, such as transcription factor binding, small RNA in- tates his or her level of genomic risk for a particular trait. The
terference, epigenetic variation, transcript splicing, and so on, PRS is calculated by summing all the alleles (weighted by
but all these effects read out in the transcriptome. This is the their individual odds ratios) that have been associated with
rationale behind brain RNA sequencing studies being carried an illness in the latest GWAS data set for that illness. For
out in many centers around the world. As an illustration of the example, for schizophrenia, considering SNPs in 145 inde-
complexity of this problem, recent work has shown that even pendent loci contributing to risk in the current schizo-
when association of the clinical risk SNP is made in brain to a phrenia GWAS, a score can be assigned to an individual on
particular expressed gene, the transcript showing association the basis of how many of those specific risk-associated al-
may not be the common, known form of the gene but rather an leles for those SNPs are found in that individual’s genome.
isoform not previously characterized (30), or the classes of Scores can also be calculated from SNPs showing statisti-
putative risk isoforms may vary across different brain re- cal association at varying p value thresholds (e.g., 531028,
gions (31, 32), or the association in brain may depend on a 131026, 0.05). In principle, two individuals could have the

16 ajp.psychiatryonline.org Am J Psychiatry 176:1, January 2019


WEINBERGER

FIGURE 1. Diagram of SNPs in a GWAS-significant locus for schizophrenia risk on chromosome 10a

1.0
18 0.8
0.6 a
16 0.4
0.2 3 1 2
14 0.1 4 5
12
Observed (–logP)

10

2
filter: p<0.04
0
GBF1 MIR1468 SUFU CYP17A1 NT5C2 USMG5

NFKB2LOC100505781 TRIM8 C10orf32 LOC729020 MIR1307

PSD C10art95 ARL3 C10orf32–AS3MT INA PDCD11

FBXL15 TMEM180 SFXN2 AS3MT PCGF6

CUEDC2 ACTR1A WBP1L CNNM2 TAF5

104100 104300 104600 104900 105100

Chromosome 10 (kb)
a
Each dot is a single-nucleotide polymorphism (SNP), and the y-axis shows the log of the p value for association of SNP genotype with diagnosis. The
SNPs are heat-scale colored according to the magnitude of genotype correlation (i.e., R2, as shown in the box) with the most significant SNP
(the diamond). The horizontal green line in the graph is the statistical threshold for genome-wide association study (GWAS) significance. Many SNPs
spanning a large block are GWAS significant, and because many genes are mapped to this 1-Mb region of the genome, as depicted in the lower
part of the figure, it is impossible to determine from these data alone which, if any, or how many, are causative. The x-axis shows the standard
physical coordinates in nucleotide numbers (in kb) for this section of chromosome 10. (See the text for more about this particular region.)

same score but actually share none of the same alleles that performance (6, 38–40), as well as aspects of brain physi-
make up their score. The PRS is thus a gross aggregate ology associated with risk for schizophrenia (41, 42). It is
measure of genomic risk, based on a simple additive model. important to appreciate that associations and predictions in
Nevertheless, current PRS explains more risk than individual clinical samples based on PRS, although at times statistically
SNPs (on the order of 3%215%, depending on which p value strong, are still weakly predictive at the level of an individual
threshold is used [34]). Moreover, because it is a score that and are not clinically actionable. Furthermore, PRSs to date
can be calculated for any individual who has been genotyped, are derived principally from samples of European ances-
it has generated considerable interest as an independent try, and questions have been raised about the potential
measure for clinical research. confounding role of ancestral genetic components in medi-
The PRS for schizophrenia has become a popular and ating some of the clinical associations (43).
dependable variable in differentiating groups of patients The current incarnation of schizophrenia PRS has fo-
with schizophrenia diagnoses from control subjects, as pa- cused on index SNPs from the large GWAS study as an
tient samples have consistently higher scores than control omnibus measure of genomic risk for clinical illness, but the
samples. The schizophrenia PRS also tends to be increased in concept can also be applied to basic biological data to identify
individuals with several other psychiatric disorders, partic- genomic profiles that represent that biology in the clinical
ularly in the context of psychotic features (6, 7). Use of the setting. For example, a PRS derived from a set of SNPs that
PRS has become a popular strategy in many studies as a influence expression of a network of genes co-expressed in
predictor of associated clinical characteristics, such as cog- brain with the dopamine D2 receptor (DRD2) has been shown
nitive deficits and treatment response, and it has also been to predict response to antipsychotic drugs (44). A PRS de-
associated with an expanding catalog of diverse measures and rived from a set of SNPs indexing genes in schizophrenia
traits in unaffected people (e.g., adolescent cannabis use GWAS loci expressed in the human placenta and dynami-
[36] and body mass index [37]). Perhaps not surprisingly, as cally regulated in placenta from complicated pregnancies
a measure of overall genomic risk, the schizophrenia PRS predicts the interaction of genetic risk and obstetrical
is associated with aspects of early child development, in- complications on liability for schizophrenia (45). The devel-
cluding intellectual and emotional development and school opment of genomic profiles for specific biological processes

Am J Psychiatry 176:1, January 2019 ajp.psychiatryonline.org 17


THINKING ABOUT SCHIZOPHRENIA IN AN ERA OF GENOMIC MEDICINE

has considerable potential for parsing the role of genes in Lithium is not antipsychotic; clozapine does not benefit in-
biological processes that account for clinical variation. It is dividuals with ADHD; and stimulants are not antimanic
reasonable to hypothesize that schizophrenia PRSs decon- drugs. From a clinical perspective, treatment is a litmus test
volved into “gene sets” reflecting specific molecular pathways of biological relevance.
or networks will have differential and stronger effects on There is also concern that some of the genes within GWAS
specific clinical parameters than the whole-genome based PRS. loci may not substantially contribute to symptomatology in
adulthood. Studies of gene expression in adult postmortem
brain samples of patients with schizophrenia do not generally
COMMENT
find differential expression of genes within GWAS loci (47,
From a biological point of view, genetic research over the past 48). In contrast, in a consistent emergent literature, many of
decade has led to an unprecedented level of raw information the genes found in these loci have been found to be prefer-
about schizophrenia risk, as have genomic investigations of entially expressed and dynamically regulated in fetal life (49),
other traits and complex diseases across medicine, including consistent with the prevailing assumption that schizo-
in cancer, hypertension, diabetes mellitus, and cardiovascular phrenia, like most psychiatric syndromes, has early de-
disease. But how far will genetic insights take psychiatry? velopmental origins (50). A recent study (45) found that a
Will they change how patients are clinically diagnosed? substantial fraction of genes in the GWAS-significant loci are
Will they predict outcome and personalize medicine? Will abundantly expressed in the placenta and are dynamically
they lead to novel therapies based on pathogenesis? My view regulated in placenta from complicated pregnancies, possibly
is that the latter is a good bet, and the others are much more explaining the link between schizophrenia risk and preg-
speculative. Given the substantial role of environmental nancy complications. While this finding has potential im-
variables in the etiology and course of all complex disorders, it plications for primary prevention based on enhancing
is unlikely that even sequencing the DNA of everyone on the placental health, it does not suggest specific treatment op-
planet would lead to a full understanding of schizophrenia. tions for an affected adult.
It is important to continually emphasize that genes do not These caveats notwithstanding, genes are critical and
encode for psychopathology. Genes associated with schizo- valid entry points to the biology of cells, and genes related to
phrenia implicate subtle malfunctions at the molecular and risk for schizophrenia are building blocks for model systems
cellular level, most likely within neurons, altering micro- and that may lead to novel insights and novel therapeutic targets.
macrocircuits in the developing brain. Cells process molec- There is considerable enthusiasm in basic research labora-
ular information, and circuits process environmental in- tories for creating neuronal models based on human plu-
formation (e.g., sensory information, cognitive information, ripotent stem cells from patients with schizophrenia and
social information). Behavior is an emergent phenomenon of from cells with genomes that contain risk-associated variants
circuit physiology, which may be aberrant because the cells (51). In principle, such approaches may identify early de-
comprising the circuit are altered, for example, by not sig- velopmental phenotypes that can be rescued experimentally.
naling to each other in an appropriate manner for a given While this work also may not translate simply to the adult
context, based on the genetic and epigenetic programs that brain, it is worth remembering that the molecular programs
entrain their development and function. Psychiatric dis- that build synapses are at least in part the same ones that
orders ultimately reflect how the brain mishandles envi- sustain them and modify them during adult life. Furthermore,
ronmental information, which at the systems level is far animal research has shown that some early developmental
“downstream” of the effect of genes in cells. Thus, given the abnormalities, even in brain structure, may be reversed
complexity of the development and function of the brain during adulthood (52). This is a new frontier, but it is not
and its emergent properties, determining what it means for science fiction.
a brain to be at increased risk for schizophrenia is a far For good reason, 21st-century medicine has become highly
more complex and challenging problem than finding sus- intertwined with genomic-based medicine. Psychiatry is part
ceptibility loci. of this new age, and the opportunities to change our nosology
There is also reason to wonder to what degree GWAS and our understanding of what we call mental illness and
associations sharpen the focus on the core pathogenic to find new approaches to improve the lives of affected in-
processes of schizophrenia. The omnigenic hypothesis men- dividuals have never seemed more promising. That being
tioned above cautions that many GWAS associations are said, as usual, the more we learn, the more we need to know,
likely to be relatively peripheral factors in causation, detected and with such a complex task, it’s likely to be a bumpy ride.
because they play a role in the biological function of relevant
cells. This may account in part for the substantial overlap AUTHOR AND ARTICLE INFORMATION
between schizophrenia and not just other psychiatric di-
The Lieber Institute for Brain Development and Maltz Research Labora-
agnoses but even disorders as seemingly distant pathogeni- tories, Baltimore; and the Departments of Psychiatry, Neurology, and
cally as amyotrophic lateral sclerosis (46). Moreover, while Neuroscience and the McNusick Nathans Institute of Genomic Medi-
the loci that have been found seem not to respect our clinical cine, Johns Hopkins School of Medicine, Baltimore.
diagnoses, the drugs we use to treat patients generally do. Send correspondence to Dr. Weinberger (drweinberger@libd.org).

18 ajp.psychiatryonline.org Am J Psychiatry 176:1, January 2019


WEINBERGER

The author thanks Richard Straub, Rebecca Birnbaum, and Michael 22. Fromer M, Pocklington AJ, Kavanagh DH, et al: De novo mutations
O’Donovan for their helpful suggestions and review of the manuscript. in schizophrenia implicate synaptic networks. Nature 2014; 506:
The author reports no financial relationships with commercial interests. 179–184
23. 1000 Genomes Project Consortium, Abecasis GR, Altshuler D, et al:
Accepted November 15, 2018.
A map of human genome variation from population-scale se-
quencing. Nature 2010; 467:1061–1073
REFERENCES 24. Singh T, Kurki MI, Curtis D, et al: Rare loss-of-function variants in
1. Schizophrenia Working Group of the Psychiatric Genomics Con- SETD1A are associated with schizophrenia and developmental
sortium: Biological insights from 108 schizophrenia-associated ge- disorders. Nat Neurosci 2016; 19:571–577
netic loci. Nature 2014; 511:421–427 25. Singh T, Walters JTR, Johnstone M, et al: The contribution of rare
2. Pardiñas AF, Holmans P, Pocklington AJ, et al: Common schizo- variants to risk of schizophrenia in individuals with and without
phrenia alleles are enriched in mutation-intolerant genes and in re- intellectual disability. Nat Genet 2017; 49:1167–1173
gions under strong background selection. Nat Genet 2018; 50:381–389 26. Goldberg TE, Ragland JD, Torrey EF, et al: Neuropsychological
3. Li Z, Chen J, Yu H, et al: Genome-wide association analysis assessment of monozygotic twins discordant for schizophrenia.
identifies 30 new susceptibility loci for schizophrenia. Nat Genet Arch Gen Psychiatry 1990; 47:1066–1072
2017; 49:1576–1583 27. Meadows G, Turner T, Campbell L, et al: Assessing schizophrenia in
4. Birney E, Stamatoyannopoulos JA, Dutta A, et al: Identification and adults with mental retardation: a comparative study. Br J Psychiatry
analysis of functional elements in 1% of the human genome by the 1991; 158:103–105
ENCODE pilot project. Nature 2007; 447:799–816 28. McConnell MJ, Moran JV, Abyzov A, et al: Intersection of diverse
5. Akbarian S, Liu C, Knowles JA, et al: The PsychENCODE project. neuronal genomes and neuropsychiatric disease: the Brain Somatic
Nat Neurosci 2015; 18:1707–1712 Mosaicism Network. Science 2017; 356:eaal1641
6. Anttila V, Bulik-Sullivan B, Finucane HK, et al: Analysis of shared 29. Poduri A, Evrony GD, Cai X, et al: Somatic activation of AKT3
heritability in common disorders of the brain. Science 2018; 360: causes hemispheric developmental brain malformations. Neuron
eaap8757 2012; 74:41–48
7. Bipolar Disorder and Schizophrenia Working Group of the Psy- 30. Li M, Jaffe AE, Straub RE, et al: A human-specific AS3MT iso-
chiatric Genomics Consortium: Genomic dissection of bipolar dis- form and BORCS7 are molecular risk factors in the 10q24.32
order and schizophrenia, including 28 subphenotypes. Cell 2018; schizophrenia-associated locus. Nat Med 2016; 22:649–656
173:1705–1715.e16 31. Ma L, Semick SA, Chen Q, et al: Schizophrenia risk variants in-
8. Hannon E, Spiers H, Viana J, et al: Methylation QTLs in the de- fluence multiple classes of transcripts of sorting nexin 19 (SNX19).
veloping brain and their enrichment in schizophrenia risk loci. Nat Mol Psychiatry (in press)
Neurosci 2016; 19:48–54 32. Ramasamy A, Trabzuni D, Guelfi S, et al: Genetic variability in the
9. Jaffe AE, Gao Y, Deep-Soboslay A, et al: Mapping DNA methylation regulation of gene expression in ten regions of the human brain.
across development, genotype, and schizophrenia in the human Nat Neurosci 2014; 17:1418–1428
frontal cortex. Nat Neurosci 2016; 19:40–47 33. Tao R, Cousijn H, Jaffe AE, et al: Expression of ZNF804A in hu-
10. Won H, de la Torre-Ubieta L, Stein JL, et al: Chromosome con- man brain and alterations in schizophrenia, bipolar disorder, and
formation elucidates regulatory relationships in developing human major depressive disorder: a novel transcript fetally regulated by
brain. Nature 2016; 538:523–527 the psychosis risk variant rs1344706. JAMA Psychiatry 2014; 71:
11. Gaspar HA, Breen G: Drug enrichment and discovery from 1112–1120
schizophrenia genome-wide association results: an analysis and 34. Purcell SM, Wray NR, Stone JL, et al: Common polygenic variation
visualisation approach. Sci Rep 2017; 7:12460 contributes to risk of schizophrenia and bipolar disorder. Nature
12. Forstner AJ, Hecker J, Hofmann A, et al: Identification of shared 2009; 460:748–752
risk loci and pathways for bipolar disorder and schizophrenia. PLoS 35. Khera AV, Chaffin M, Aragam KG, et al: Genome-wide polygenic
One 2017; 12:e0171595 scores for common diseases identify individuals with risk equivalent
13. Schijven D, Kofink D, Tragante V, et al: Comprehensive pathway to monogenic mutations. Nat Genet 2018; 50:1219–1224
analyses of schizophrenia risk loci point to dysfunctional post- 36. Hiemstra M, Nelemans SA, Branje S, et al: Genetic vulnerability to
synaptic signaling. Schizophr Res 2018; 199:195–202 schizophrenia is associated with cannabis use patterns during ad-
14. Szappanos B, Kovács K, Szamecz B, et al: An integrated approach olescence. Drug Alcohol Depend 2018; 190:143–150
to characterize genetic interaction networks in yeast metabolism. 37. Peterson RE, Maes HH, Holmans P, et al: Genetic risk sum score
Nat Genet 2011; 43:656–662 comprised of common polygenic variation is associated with body
15. Lehner B, Crombie C, Tischler J, et al: Systematic mapping of mass index. Hum Genet 2011; 129:221–230
genetic interactions in Caenorhabditis elegans identifies common 38. Riglin L, Collishaw S, Richards A, et al: Schizophrenia risk alleles and
modifiers of diverse signaling pathways. Nat Genet 2006; 38:896–903 neurodevelopmental outcomes in childhood: a population-based
16. Mackay TF: Epistasis and quantitative traits: using model organisms cohort study. Lancet Psychiatry 2017; 4:57–62
to study gene-gene interactions. Nat Rev Genet 2014; 15:22–33 39. van Os J, van der Steen Y, Islam MA, et al: Evidence that polygenic
17. Shao H, Burrage LC, Sinasac DS, et al: Genetic architecture of risk for psychotic disorder is expressed in the domain of neuro-
complex traits: large phenotypic effects and pervasive epistasis. development, emotion regulation, and attribution of salience. Psy-
Proc Natl Acad Sci USA 2008; 105:19910–19914 chol Med 2017; 47:2421–2437
18. Horlbeck MA, Xu A, Wang M, et al: Mapping the genetic landscape 40. Mistry S, Harrison JR, Smith DJ, et al: The use of polygenic risk
of human cells. Cell 2018; 174:953–967.e22 scores to identify phenotypes associated with genetic risk of
19. Boyle EA, Li YI, Pritchard JK: An expanded view of complex traits: schizophrenia: systematic review. Schizophr Res 2017; 197:2–8
from polygenic to omnigenic. Cell 2017; 169:1177–1186 41. Kauppi K, Westlye LT, Tesli M, et al: Polygenic risk for schizo-
20. Marshall CR, Howrigan DP, Merico D, et al: Contribution of copy phrenia associated with working memory-related prefrontal brain
number variants to schizophrenia from a genome-wide study of activation in patients with schizophrenia and healthy controls.
41,321 subjects. Nat Genet 2017; 49:27–35 Schizophr Bull 2015; 41:736–743
21. Kirov G, Rees E, Walters JTR, et al: The penetrance of copy number 42. Chen Q, Ursini G, Romer AL, et al: Schizophrenia polygenic risk
variations for schizophrenia and developmental delay. Biol Psy- score predicts mnemonic hippocampal activity. Brain 2018; 141:
chiatry 2014; 75:378–385 1218–1228

Am J Psychiatry 176:1, January 2019 ajp.psychiatryonline.org 19


THINKING ABOUT SCHIZOPHRENIA IN AN ERA OF GENOMIC MEDICINE

43. Curtis D: Polygenic risk score for schizophrenia is more strongly 48. Fromer M, Roussos P, Sieberts SK, et al: Gene expression elucidates
associated with ancestry than with schizophrenia. Psychiatr Genet functional impact of polygenic risk for schizophrenia. Nat Neurosci
2018; 28:85–89 2016; 19:1442–1453
44. Pergola G, Di Carlo P, D’Ambrosio R, et al: DRD2 co-expression 49. Jaffe AE, Shin J, Collado-Torres L, et al: Developmental regulation
network and a related polygenic index predict imaging, behavioral, of human cortex transcription and its clinical relevance at single
and clinical phenotypes linked to schizophrenia. Transl Psychiatry base resolution. Nat Neurosci 2015; 18:154–161
2017; 7:e1006 50. Birnbaum R, Weinberger DR: Genetic insights into the neuro-
45. Ursini G, Punzi G, Chen Q, et al: Convergence of placenta biology developmental origins of schizophrenia. Nat Rev Neurosci 2017; 18:
and genetic risk for schizophrenia. Nat Med 2018; 24:792–801 727–740
46. McLaughlin RL, Schijven D, van Rheenen W, et al: Genetic corre- 51. Brennand KJ, Marchetto MC, Benvenisty N, et al: Creating patient-
lation between amyotrophic lateral sclerosis and schizophrenia. specific neural cells for the in vitro study of brain disorders. Stem
Nat Commun 2017; 8:14774–14778 Cell Reports 2015; 5:933–945
47. Jaffe AE, Straub RE, Shin JH, et al: Developmental and genetic 52. Berry-Kravis EM, Lindemann L, Jønch AE, et al: Drug development
regulation of the human cortex transcriptome illuminate schizo- for neurodevelopmental disorders: lessons learned from fragile X
phrenia pathogenesis. Nat Neurosci 2018; 21:1117–1125 syndrome. Nat Rev Drug Discov 2018; 17:280–299

20 ajp.psychiatryonline.org Am J Psychiatry 176:1, January 2019

You might also like