Interspecies Trait Genetics Reveals Association of

Adcy8 with Mouse Avoidance Behavior and a Human

Mood Disorder
Annetrude (J.G.) de Mooij-van Malsen, Hein A. van Lith, Hugo Oppelaar, Judith Hendriks, Marina de Wit,
ь
Elzbieta Kostrzewa, Gerome Breen, David A. Collier, Berend Olivier, and Martien J. Kas
Background: Identifying susceptibility genes for endophenotypes by studying analogous behaviors across species is an important
strategy for understanding the pathophysiology underlying psychiatric disorders. This approach provides novel biological pathways plus
validated animal models critical for selective drug development. One such endophenotype is avoidance behavior.

Methods: In the present study, novel automated registration methods for longitudinal behavioral assessment in home cages are used to
screen a panel of recently generated mouse chromosome substitution strains that are very powerful in quantitative trait loci (QTL) detection
of complex traits. In this way, we identified chromosomes regulating avoidance behavior (increased sheltering preference) independent of
motor activity levels (horizontal distance moved). Genetic information from the mouse QTL-interval was integrated with that from the
homologous human linkage region for a mood disorder.

Results: We genetically mapped a QTL for avoidance behavior on mouse chromosome 15, homologous with a human genome region
(8q24) linked to bipolar disorder. Integrating the syntenic mouse QTL-interval with genotypes of 1868 BPD cases versus 14,311 control
subjects revealed two associated genes (ADCY8 and KCNQ3). Adenylyl cyclase 8 (Adcy8) was differentially expressed in specific brain regions
of mouse strains that differ in avoidance behavior levels. Finally, we showed that chronic infusion of the human mood stabilizer carbamaz-
epine (that acts via adenylyl cyclase activity) significantly reduced mouse avoidance behavior, providing a further link between human
mood disorders and this mouse home cage behavior.

Conclusions: Our data suggest that Adcy8 might encode a translational behavioral endophenotype of bipolar disorder.

Key Words: Animal model, chromosome substitution strains, en-
dophenotype, home cage environment, mood disorder, psychiatric
disorders, quantitative trait loci
across species, to obtain food or mediate social interactions
while avoiding threatening situations. Such behavior is influ-
enced by genetic variation, as shown by behavioral differences
between inbred strains of mice and subsequent quantitative trait
loci (QTL) analysis (5– 8). Avoidance and motor activity levels are

M
ood disorders have a major impact on the quality of life
of many people, with a prevalence of 10%–20% world-
wide (1). Finding the mechanisms underlying these
heterogeneous psychiatric disorders and obtaining valid animal
models is essential for the development of selective pharmaco-
logical treatments (2). Interspecies genetic analysis of mood
disorder endophenotypes is an important approach to the dis-
covery of novel insights in causality and to identify translational
preclinical models (3,4). Here, we focus on the genetic dissection
of avoidance behavior in mice with the aim of finding more
selective and effective pharmacological targets for behavioral
disorders in humans.
The balance between approach and avoidance behavior is
part of a behavioral strategy, which has been highly conserved
From the Rudolf Magnus Institute of Neuroscience (JGdM-vM, HO, JH, MdW,
EK, BO, MJK), Department of Neuroscience and Pharmacology, Univer-
sity Medical Centre Utrecht; Faculty of Veterinary Medicine (HAvL), De-
partment of Animals, Science, Society, Division of Laboratory Animal
Science, Utrecht University; Utrecht Institute for Pharmaceutical Sci-
ences (BO), Department of Psychopharmacology, Faculty of Science,
Utrecht University, Utrecht, The Netherlands; Social, Genetic and Devel-
opmental Psychiatry Research Centre (GB, DAC), Institute of Psychiatry,
King’s College London, United Kingdom; and the Department of Psychi-
atry (BO), Yale University School of Medicine, New Haven, Connecticut.
Address correspondence to Martien Kas, Ph.D., Rudolf Magnus Institute of
Neuroscience, Department of Neuroscience and Pharmacology, Univer-
sity Medical Centre Utrecht, Universiteitsweg 100, 3584 CG, Utrecht, The
Netherlands; E-mail: m.j.h.kas@umcutrecht.nl.
Received Apr 2, 2009; revised Jun 18, 2009; accepted Jun 21, 2009.
commonly studied in rodent species with traditional anxiety
tests, such as the elevated plus maze and open field. However,
the nature of these tests makes it difficult to differentiate between
these two behavioral components, and experimenter-effects can
have a great impact on the behavioral outcome (9). In the present
study, novel automated registration methods for longitudinal
behavioral assessment in home cages are used to screen a panel
of recently generated mouse chromosome substitution strains
(CSSs) that are very powerful in QTL-detection of complex traits
(10). The automated home cage environment (Figure 1A–1C) is
designed to increase behavioral resolution by dissociating be-
havioral endophenotypes in mice (11). It assesses levels of
avoidance behavior (sheltering) independent of motor activity
levels (horizontal distance moved) and with minimal human
interference. Longitudinal automated assessment of anxiety-
related behaviors might also overcome inconsistent results, de-
pending on subtle, short-term variations in the laboratory or test
environment (12). Furthermore, the avoidance behavior in the
home cage is sensitive to benzodiazepines (11), providing pre-
dictive validity for this anxiety-related endophenotype that might
relate, for example, to mood disorders with anxious symptoms.
In contrast to the concepts of face and predictive validity,
construct validity (similarity to the underlying causes and mech-
anisms of the disease) is the most difficult to provide for animal
models of psychiatric disorders, simply because of the lack of
knowledge about the underlying etiological mechanisms of such
complex disorders. Furthermore, for animal models of psychiat-
ric disorders it has been proven difficult to provide a 1:1

0006-3223/09/$36.00 BIOL PSYCHIATRY 2009;66:1123–1130

doi:10.1016/j.biopsych.2009.06.016 © 2009 Society of Biological Psychiatry
1124 BIOL PSYCHIATRY 2009;66:1123–1130 J.G. de Mooij-van Malsen et al.

Figure 1. Behavioral screening of a chromosome substitution strain (CSS) panel revealed a locus for avoidance behavior in an automated home cage
environment on mouse chromosome 15. (A,B) The automated home cage environment is equipped with a home base shelter (in which mice mainly sleep
during the light phase), a drinking spout, and two feeding platforms (C); on one feeding platform the mouse is exposed to the environment while the other
platform allows sheltered feeding. The PhenoTyper top unit contains an infrared camera and infrared light-emitting diode lights allowing continuous
recording independent of lighting conditions in the test room. (D) The C57BL/6J and CSS15 females; feeding duration on the two platforms on the 3 days of
testing. The CSS15 females show an increasing preference for the sheltered platform over the consecutive days. (E,F) Representative track samples during
automated baseline behavioral registration, showing a reduction in visitation of the exposed feeding platform for CSS15 (F) compared with C57BL/6J (E).
*Exposed versus sheltered feeding platform/day, p ⬍ .05.

translation with respect to the face validity criteria. Currently,
large-scale genome wide association studies (GWAS) are being
performed and have in some cases revealed (unexpected) can-
didate genes that have low odds ratios and explain a small
percentage of the variance. In light of this, genetic validity might
provide a new entrance to the biology of psychiatric diseases
with animal models (13). By integrating mouse and homologous
human genetic mapping data, we identified a gene, adenylyl
cyclase 8 (ADCY8), connecting mouse avoidance behavior ob-
tained in automated home cage environments to human bipolar
affective disorder. These findings point to novel mechanisms
underlying bipolar affective disorders and open new roads for
treatment and translational research of its psychiatric endophe-
notypes.
Methods and Materials
Animals
CSSs. The CSS panel (C57BL/6J-Chr; 1A/NaJ to C57BL/6J-Chr;
19A/NaJ, C57BL/6J-Chr; XA/NaJ, C57BL/6J-Chr; YA/NaJ; hence-
forth referred to as CSS1, CSS2, and so forth) was studied (except
for CSS7 due to low availability). See Supplement 1 for further
information on amount of mice tested, the generation of the
F2-population, and housing conditions.
Mouse Behavioral Testing
All experimental procedures were approved by the ethical
committee for animal experimentation of the University Medical
Center Utrecht, The Netherlands.
Home Cage Environment. For a description of the home
cage environment dimensions, please see Kas et al. (11). One
hour before the start of the dark phase, mice (10 –14 weeks old)
were weighed and placed individually in a home cage environ-
ment for 73 hours. No experimenter interference took place
during these 3 days. At the end of the experiment, body weight
change and food and water intake were measured. Smears were
taken from the female mice to account for possible effects of the
estrous cycle on behavior. Distance moved, enter frequency, and
duration spent in the shelter and on the two feeding platforms
were obtained with the video-tracking system (PhenoTyper PT10S/
P/N Version 1.01, Noldus Information Technology, Wageningen,
The Netherlands).
For information on the open field, elevated plus maze, and
light-dark box tests that were used in this study, please see
Supplement 1.
Animal Surgery and Drug Infusion. Female C57BL/6J mice
were anesthetized (2% isoflurane in medial grade oxygen) and
received surgical implantation with 2 minipumps (Model 1007 D,
Alzet, Cupertino, California) placed subcutaneously on their

www.sobp.org/journal
J.G. de Mooij-van Malsen et al.
back (incision between the shoulder blades, 1 pump at each
shoulder blade). The minipumps were filled with either vehicle
(42.5% dimethyl sulfoxide; 42% propylene glycol 400; 15% etha-
nol) or carbamazepine, dissolved in the same vehicle (Sigma-
Aldrich Chemie BV, Zwijndrecht, The Netherlands) such that
there was a daily release of 40 mg/kg carbamazepine for a total
of 7 days (until the end of experiment). Animals were allowed to
recover for 3 days after surgery. Subsequently, all animals were
tested in the home cage for 4 consecutive days.
QTL Analysis
A total of 13 markers were selected across chromosome 15
(average spacing of consecutive markers was 4.2 centimor-
gans). Genotyping methods, Map Construction, and QTL
analysis have been described previously (14). For details, also
see Supplement 1.
In Situ Hybridization
In situ hybridization analysis on Adcy8 and Kcnq3 was
performed on brain sections from an additional set of female
C57BL/6J (n ⫽ 9) and CSS15 (n ⫽ 9) mice. In situ hybridization
analysis was performed as previously described (15). For further
details, see Supplement 1.
Statistics
Mouse Data. Values were expressed as mean ⫾ SEM. For the
statistical analyses SPSS 11.5 for Windows was used (SPSS,
Chicago, Illinois). Data for mice that lost more than 1.5 g during
the 3 days of testing (e.g., due to failing water bottle) were
discarded (in total 7 female mice and 2 male mice; no specific
strain effects were observed in this respect). Male and female
mice were analyzed separately due to observed gender effects.
Parameters were considered for normal distribution and trans-
formed where appropriate. Data were analyzed with a linear
mixed model with day as the repeated factor and an unstructured
repeated covariance type (best fit), with the least significant
difference test to compare the main effects/day/strain with
C57BL/6J. For the nonrepeated measures, a one-way analysis of
variance (ANOVA) by strain was used, with Dunnett’s post hoc
comparing all CSS with the C57BL/6J control strain. Significance
levels (␣ ⫽ .05) were corrected with the Dunnett method to
account for the multiple strain comparison (p ⫽ .003) (16).
Preference for either feeding platform was analyzed/strain with a
one-sample t test on the difference between the open and the
sheltered feeding platform (“delta”) (␣ ⫽ .05). In situ hybridiza-
tion results were analyzed with a one-way ANOVA (␣ ⫽ .05).
Treatment effect of carbamazepine (measured by the difference
between treatment groups in fraction of the total feeding dura-
tion spent on the exposed feeding platform) was analyzed with
a one-sample t test (␣ ⫽ .05).
Human Data. This study makes use of data generated by the
Wellcome Trust Case Control Consortium (WTCCC). For materi-
als and methods for genome-wide association studies, see refer-
ence (17). Summary statistics for the Affymetrix 500K gene chip,
subset “bipolar disorder” (http://www.wtccc.org.uk) were ob-
tained (1868 cases vs. the 14,311 combined control subjects
[control group consisting of all individuals within the WTCCC
study without psychiatric diagnosis] or 2938 basic control sub-
jects [shared control group for all diseases within the WTCCC
dataset]).
The p values of all single nucleotide polymorphisms (SNPs)
within the Affymetrix 500K gene chip located within these
regions were examined (total of 391 SNPs within the 9 selected
BIOL PSYCHIATRY 2009;66:1123–1130 1125
genes [Mlze; Ddef1; Adcy8; Kcnq3; Lrrc6; Phf20l1; Wisp1; Ndrg1;
Zfat1]) with additive/genetic models with or without sex-strati-
fication. To identify genes of interest, a significant corrected p
value of p ⬍ .0056 was chosen (p ⫽ .05/9 genes tested). The
Ensemble Genome Browser (http://www.ensembl.org/index.
html) was used to detect whether SNPs were located in intronic or
coding regions.
Results
Genetic Dissection of Mouse Avoidance Behavior
To efficiently screen for genetic regions influencing motor
activity levels or avoidance behavior, the CSS mouse panel, their
genetic background strains (C57BL/6J host, A/J donor strain),
and a F2-progeny for QTL mapping were tested in an automated
home cage environment for 3 continuous days. In the automated
home cage environment (Figure 1A–1C) the animals can choose
to eat at two different feeding platforms, where they had ad
libitum access to regular chow. At one feeding platform the
animals could eat while exposed to the environment and at the
other while sheltered. Because total feeding duration is very low
during the light phase but no differences in preference were
observed between the dark and light phase, data were analyzed/
24-hour day.
A/J and C57BL/6J. The genetic background strains of the CSS
panel showed marked behavioral differences in the automated
home cage environment. Under novelty conditions (the first hour
of testing), both male and female A/J mice exhibited significantly
lower motor activity levels (horizontal distance moved) than
C57BL/6J males and females, respectively (p ⬍ .001; data not
shown). In line with these suppressed motor activity levels, A/J
mice did not explore the feeding platforms during this first hour
(data not shown), illustrating the confounding effects of novelty-
induced motor activity levels on the assessment of avoidance
behavior during short-lasting behavioral tasks in new environments.
During the subsequent longitudinal measurements, C57BL/6J
males showed significant avoidance of the exposed feeding
platform (Figure 2A) [t (37) ⫽ 3.013, p ⫽ .005; data for day 3 are
shown]. Both A/J females and C57BL/6J females exhibited no
preference for either feeding platform (Figure 2B) (data for day 3
are shown). Thus, male C57BL/6J mice were found to have a
preference in visit and feeding duration for the sheltered feeding
platform on all 3 days, whereas female C57BL/6J mice showed
no preference on any day of the experiment.
With respect to the longitudinal assessment of (horizontal)
distance moved in the home cage, A/J males showed higher
motor activity levels during both the light and the dark phase
compared with C57BL/6J males (Figure 2C) [total distance moved
day 3: F (2,71) ⫽ 41.255, p ⬍ .001; data for day 3 are shown]. In
contrast, A/J females showed a reduction in motor activity levels
compared with C57BL/6J females (Figure 2D) [total distance
moved day 3: F (2,55) ⫽ 8.438, p ⫽ .005; data for day 3 are
shown].
The CSS Panel. As with the genetic background strains,
significant gender differences were also observed in the CSS
panel (Figure 2, Table 1). In males, overall preference for the
sheltered feeding platform was significant for C57BL6/J (t ⫽ 3.0,
p ⫽ .005), CSS1 (t ⫽ 4.4, p ⫽ .005), CSS3 (t ⫽ 4.8, p ⫽ .002), CSS4
(t ⫽ 3.5, p ⫽ .004), CSS6 (t ⫽ 3.2, p ⫽ .012), CSS8 (t ⫽ 7.1, p ⬍
.001), CSS14 (t ⫽ 2.6, p ⫽ .020), and CSSY (t ⫽ 8.4, p ⬍ .001). In
females, only CSS15 (t ⫽ 3.4, p ⫽ .027) and CSS19 (t ⫽ 3.7, p ⫽
.004) had a significant overall preference for the sheltered
feeding platform. None of the strains exhibited a preference for
www.sobp.org/journal
1126 BIOL PSYCHIATRY 2009;66:1123–1130 J.G. de Mooij-van Malsen et al.

Figure 2. Genetic dissection of mouse avoidance behavior and motor activity levels (horizontal distance moved) in a chromosome substitution strain (CSS)
panel. (A,B) Preference for sheltered feeding on the third day of home cage environment testing. Sheltering preference is calculated as difference between
duration of feeding on the sheltered platform and the exposed platform (“delta”). A positive number indicates preference for the sheltered feeding platform
(thus, avoidance of the exposed feeding platform). (C,D) Distance moved on the third day of home cage environment testing. *Delta versus 0 (“no
preference”), p ⬍ .05. ✧A/J versus C57BL/6J, p ⬍ .05. #CSS versus C57BL/6J, p ⬍ .003.

the exposed feeding platform. Different chromosomes were
found to influence basal motor activity levels (horizontal distance
moved), confirming that the home cage environment measures
avoidance levels independent of motor activity levels (Figure 2).
CSS15-F2. A gender-specific effect of A/J chromosome 15 on
avoidance behavior was selected for further genetic fine map-
ping on the basis of the following criteria: 1) in contrast to female
C57BL/6J control subjects, female CSS 15 showed significant
lower absolute amounts of time on the exposed platform than on
the sheltered platform on day 2 (t ⫽ 2.829, p ⫽ .047) and day 3
(t ⫽ 4.371, p ⫽ .012) (Figures 1D–1F); 2) the relative amount of
time spent on the sheltered platform (as calculated by the
difference between duration of feeding on the sheltered platform
and the exposed platform [“delta”]) was significantly different
from 0 in CSS 15 females and not in C57BL/6J females (Figure
2B); and 3) mouse chromosome 15 has repeatedly been found to
contain a region involved in avoidance behavior (6,18,19). A
CSS15-F2 population was bred (n ⫽ 105) for QTL analysis and
tested in the open field, automated home cage environment,
elevated plus maze, and light-dark box. Behavioral correlations
revealed no significant relationship between the avoidance be-
havior assessed in the home cage environment and the standard
behavioral tests for exploration and avoidance (except for motor
activity levels during novelty in the home cage environment and

Table 1. Location of QTLs for Several Parameters with Consomic Mouse Strains on the 3 Days of Measurement in the Home Cage Environment

The black boxes show quantitative trait loci (QTLs) indicated by a significant difference (p ⬍ .003) between the corresponding chromosome substitution
(CS) strain and C57BL/6J.
Grey boxes indicate suggestive QTLs (.05 ⬎ p ⬎ .003). The lower two rows of the table indicate whether a preference in feeding or visit duration for either
platform is present; a black box indicates a significant preference (one-sample t test on difference between sheltered and exposed platform, p ⬍ .05).

www.sobp.org/journal
J.G. de Mooij-van Malsen et al.
Figure 3. Genetic fine mapping of behavioral quantitative trait loci (QTLs)
on mouse chromosome 15. Genetic fine mapping of QTL analysis revealed 3
significant peaks on chromosome 15 showing specific genetic control for
distinct behaviors. Black bar on x-axis indicate ⫺1 logarithm of odds (LOD)
support interval of the QTL for baseline avoidance behavior.
open field; females p ⫽ .001, r ⫽ .41). Because the anxiolytic
benzodiazepine chlordiazepoxide reduces sheltered feeding
preference in the automated home cage environment (11), these
findings emphasize that the home cage avoidance phenotype
reveals additional behavioral characteristics related to anxiety when
compared with the avoidance parameters in the traditional behav-
ioral tests (open field, elevated plus maze, and light-dark box).
Genetic Mapping of Avoidance Behavior on Mouse
Chromosome 15
The QTL analysis on the CSS15-F2 population revealed several
significant logarithm of odds (LOD) scores for the females in the
home cage environment and open field but not for the elevated
plus maze and light-dark box (for LOD scores and peak markers,
see Table S2 in Supplement 1). Three distinct loci were identified
for different home cage behaviors (Figure 3): one for baseline
avoidance behavior (visit duration exposed feeding platform day
3), one for motor activity levels (horizontal distance moved)
during novelty, and one for novelty avoidance behavior. It was
determined that the QTL for baseline avoidance behavior ac-
counted for 22.7% of the variance in the F2-population, the QTL
for novelty-induced motor activity accounted for 25.2%, and the
QTL for novelty avoidance behavior accounted for 25.7% of the
variance.
Homology with 8q24 Linkage Region for Human Bipolar
Disorder
The QTL for baseline avoidance behavior in the home cage is
located around markers rs13482612 and D15MIT234. The signif-
icant 13-mega base pair QTL-interval (with the ⫺1.0 LOD support
interval) (20) comprises 42 known genes (http://www.informatics.
jax.org/) and is located between 56,114,878 bp and 69,568,581 bp
(with interpolation calculations). Interestingly, this QTL-interval
is homologous with human genetic region 8q24, repeatedly
found in human linkage studies of bipolar affective disorder (21)
and suicidal behavior (22). In a genome scan meta-analysis
(GSMA) (23) of bipolar affective disorder, 8q24 (bin 8.6) was
ranked as the fifth most significant region of linkage, of 120
genome bins under a narrow diagnostic model of bipolar
disorder, and the third most significant under a broad diagnostic
model, which included unipolar depression (21).
BIOL PSYCHIATRY 2009;66:1123–1130 1127
For the application of comparative genomics on the basis of
the homology with the 8q24 human linkage region, all known
brain-expressed genes within the mouse QTL for baseline avoid-
ance containing known SNPs (with http://www.jax.org/phenome/
snp.html and http://www.brain-map.org) between the progenitor
strains of the CSS panel (A/J and C57BL/6J) were selected as
candidate genes to be tested in a human sample of bipolar
affective disorder patients and control subjects. These genes
were Mlze; Ddef1; Adcy8; Kcnq3; Lrrc6; Phf20l1; Wisp1; Ndrg1;
and Zfat1. To obtain human data, GWA data for bipolar disorder
generated by the WTCCC was examined and the summary
statistics for the Affymetrix 500K gene chip, subset “bipolar
disorder”, were obtained (1868 cases vs. the 14,311 combined
control subjects [control group consisting of all individuals within
the WTCCC study without psychiatric diagnosis] or 2938 basic
control subjects [shared control group for all diseases within the
WTCCC dataset]). For the selected mouse genes, the human
location was determined with the University of California Santa
Cruz genome browser (http://genome.ucsc.edu/cgi-bin/hgGateway,
human set March 2006), and p values of the GWA study for all
SNPs located within these regions were examined (total of 391
SNPs within the 9 selected genes). Two genes were found to be
associated with BPD (Table S3 in Supplement 1). Within the
combined control dataset, a significant association was found for
ADCY8 (rs3914071). We analyzed this SNP for association with
other disorders in the WTCCC study (e.g., type 1 and type 2
diabetes) and found evidence for association of rs3914071 with
type 2 diabetes (p ⫽ .001). Because this suggests that there might
be cross-association of ADCY8 with type 2 diabetes and BPD, the
type 2 diabetes cases were excluded from the case-control
analysis. After this correction the association between the ADCY8
SNP and BPD remained significant (p ⫽ .0055; genetic model,
without sex-stratification). In addition, within the basic control
dataset, a significant association was found for KCNQ3 (rs1437812;
p ⫽ .0029; additive model, without sex-stratification). These findings
are consistent with recent studies on human candidate genes
identified for bipolar affective disorder within the human 8q24
region with an independent sample (24,25).
Differential Expression of Brain Adcy8
To examine whether differential expression of these two
genes is associated with mouse avoidance behavior, brain gene
expression patterns were studied. The Adcy8 and Kcnq3 mes-
senger RNA (mRNA) expression levels in brain regions known to
be involved in mood were examined in C57BL/6J and CSS15
females with radioactive in situ hybridization. For Kcnq3, no
changes were found in expression levels in the examined brain
areas. However, CSS15 females showed an increased expression
of Adcy8 in the ventromedial hypothalamus (p ⫽ .01) and the
piriform cortex when compared with C57BL/6J females (p ⫽ .01)
(Figure 4), indicating that the levels of Adcy8 in these brain
regions are associated with avoidance behavior.
Treatment with Carbamazepine
Because the genetic data revealed an association between
mouse avoidance behavior and human bipolar disorder, we
hypothesized that a mood stabilizer should alter the avoidance
behavior in mice. For that reason, mouse behavior in the
automated home cage was compared in mice chronically infused
with vehicle or with the mood stabilizer carbamazepine (40
mg/kg/day) that is known to attenuate the adenylyl cyclase
(Adcy-) signaling pathway (for review see Gould et al.) (26).
Interestingly, Adcy8 is an adenylyl cyclase involved in the cyclic
www.sobp.org/journal
1128 BIOL PSYCHIATRY 2009;66:1123–1130 J.G. de Mooij-van Malsen et al.

Figure 4. Adcy8 is differentially expressed in specific brain regions as a function of avoidance behavior. The messenger RNA expression levels of (A) Kcnq3 and
(B) Adcy8 in the basal amygdala (Amyg B), lateral amygdala (Amyg L), CA1 region of the hippocampus, dentate gyrus (DG), dorsomedial hypothalamus
(Hypothal D), ventromedial hypothalamus (Hypothal V), piriform cortex, and thalamus. The CSS15 mice showed increased expression of Adcy8 in the
ventromedial hypothalamus (C: C57BL/6J, D: CSS15) and in the piriform cortex (E: C57BL/6J, F: CSS15). Please note that the ventromedial hypothalamic region
is circled in Figures C and D. The arrow (E,F) points to the piriform cortex (pir) *CSS15 versus C57BL/6J, p ⬍ .05.

adenosine monophosphate signaling pathway. These studies
revealed that carbamazepine significantly reduced mouse avoid-
ance behavior when compared with vehicle-treated mice (p ⫽
.03) (Figure 5).
Discussion
Understanding the pathophysiology of affective disorders,
thus identifying underlying disease pathways and hence identify-
ing novel drug targets, is crucial for the development of selective
treatments. One approach to this is the mapping of susceptibility
genes that influence behavioral endophenotypes of the disorder,
because they might identify treatable components of the disease.
Thus, endophenotypes that can be examined across species not
only provide tools for mapping genes and therefore biological
insights into disease but also validate animal models necessary
for drug development (3,4). So far, the identification of animal
models for affective disorders has been hampered by their poor
resemblance to the human symptoms and the lack of knowledge
about the etiology of these heterogeneous disorders. However,
recent genome-wide association studies in humans have pro-
vided new insights in the etiology of some psychiatric disorders
such as bipolar disorder (27). Although these findings generally
revealed gene mutations with small effect sizes, they offer new
leads to the cause of these complex disorders. On the basis of the
notion that in different species the same genes might indepen-
dently give rise to alleles with similar functional and phenotypic
effects, genetic validity could be used as a tool for identifying
analogous pathology between animals and human affective
disorders. In the present study an association was found for
Adcy8 with mouse avoidance behavior and bipolar affective
disorder.
With chromosome substitution strains of mice, we found that
different chromosomes contributed to avoidance behavior or
motor activity levels in the automated home cage environment.
As commonly seen in human psychiatric disorders, the behav-
ioral parameters in the automated home cage environment were
generally differentially affected in female and male mice. The
relationship between increased avoidance behavior in CSS15

www.sobp.org/journal
J.G. de Mooij-van Malsen et al.
Figure 5. Chronic infusion of the mood stabilizer carbamazepine reduced
avoidance behavior in mice. Feeding fraction on the exposed platform as a
percentage of vehicle on the fourth day of testing. Female C57BL/6J mice
treated chronically with 40 mg/kg carbamazepine for 7 days show a 3⫻
larger feeding fraction on the exposed feeding platform (feeding duration
open platform/total feeding duration both platforms) than vehicle control
animals. *Significant difference in avoidance behavior between vehicle- and
carbamazepine-treated mice (p ⫽ .03).
females and bipolar affective disorders is not known, although
some population-based epidemiological studies have shown that
the prevalence of bipolar disorder is greater in women (28).
Further genetic mapping showed that mouse avoidance be-
havior in the home cage is not directly related to anxiety-related
parameters obtained from traditional behavioral tests for rodent
species, such as the open field test and the elevated plus maze.
However, the home cage QTL on mouse chromosome 15
obtained for motor activity levels during novelty and open field
activity mapped to the same location as “emo2”, a QTL for motor
activity levels and defecation in the open field test (8). Further-
more, the additional locus for visiting the exposed feeding
platform in the first hour of home cage testing, a measure for
avoidance and exploration during novelty, is located approxi-
mately 2 centimorgans from marker rs13482748. The ⫺1 LOD
support interval for this QTL encompasses a large genetic region
and contains genes previously found to be involved in neopho-
bia (Vdr, vitamin D receptor) (29) and spatial learning (Nell2)
(30). These findings further underscore the need for novel
behavioral phenotyping methods that allow dissociation of be-
havioral components and assessment of both novelty-induced
and baseline behaviors (9).
By means of comparative genomics, we have explored the
underlying genetic origin of a potential behavioral endopheno-
type of bipolar disorder, which occurs across mice and humans.
The QTL we identified for mouse avoidance behavior on chro-
mosome 15 is homologous with human chromosome 8q24, a
locus linked to bipolar affective disorder and possibly also
unipolar depression (21). This finding provides a unique oppor-
tunity to combine human and mouse data to identify homolo-
gous mechanisms relevant to mood disorders, especially because
the validity of animal models for this class of psychiatric disor-
ders has been controversial (31). To further investigate the
relationship between mouse avoidance behavior and human
bipolar disorder, we studied the effect of a mood stabilizer on the
avoidance behavior in mice and found that chronic infusion of
carbamazepine significantly reduced avoidance behavior.
We found two genes from this locus showing association with
bipolar disorder in humans, which are expressed in the brain and
are plausible candidates on the basis of their known function.
KCNQ3, a voltage-gated potassium channel, is known to be
BIOL PSYCHIATRY 2009;66:1123–1130 1129
involved in epilepsy (32). Because anticonvulsants are successful
in treating bipolar disorder, subtle changes in this gene might
contribute to both disorders. ADCY8 is a member of the adenylyl
cyclase family of genes. Adenylyl cyclase activity is known to be
altered in bipolar affective disorders (33). For example, the levels
of cyclic adenosine monophosphate (that is synthesized from
adenosine triphosphate by adenylyl cyclase) are increasing in
bipolar patients moving toward a less depressed mood, whereas
the levels are decreasing in patients moving away from mania
(34). These findings indicate that adenylyl cyclase activity is
directly related to the different mood states that are characteristic
for bipolar affective disorder. Furthermore, mood stabilizers,
such as lithium and carbamazepine, are known to modulate the
adenylyl cyclase pathway (26,35). Interestingly, Adcy1/Adcy8
double knockout mice show altered sucrose preference and
alternations in neurotrophic signaling (36). Also, Adcy8-deficient
mice showed changes in anxiety-related behavior after exposure
to stressful events (37). Because the stress-hormone axis has
been associated with bipolar affective disorder, the relationship
between environmental stressors and the expression of anxiety-
related behavior in Adcy8 knockout mice is highly interesting.
To further analyze the association between these two candi-
date genes and avoidance behavior in mice, gene expression
analysis was performed as a function of mouse avoidance
behavior. Analysis of mRNA expression levels of Kcnq3 in a
variety of brain regions showed no differential regulation. In
contrast, Adcy8 was differentially expressed in a hypothalamic
region and in the piriform cortex. Other brain regions such as the
amygdala and hippocampus did not show differentially regu-
lated Adcy8, suggesting a brain region specific function of Adcy8
in mouse avoidance behavior. The high avoidance behavior
strain (CSS15) showed upregulation of Adcy8 in these specific
brain regions, consistent with data from Adcy8 knockout mice
that exhibit decreased anxiety-related behavior (37).
In conclusion, by combining mouse QTL mapping data with
human GWA data, two candidate genes (KCNQ3 and ADCY8)
were identified potentially linking mouse avoidance behavior to
a human mood disorder. Interestingly, we showed that a human
mood stabilizer significantly reduced mouse avoidance behavior,
providing a further link between human mood disorders and this
innate mouse behavior. By analyzing the expression of these two
genes in relation to mouse avoidance behavior, we found a
relationship between Adcy8 signaling in the piriform cortex and
hypothalamus and this behavioral trait. The differential regula-
tion of Adcy8 in the ventromedial hypothalamus and piriform
cortex might offer novel insights in the neurocircuitry underlying
affective disorders and could provide a potential mechanism
underlying an endophenotype in bipolar affective disorder (e.g.,
linked to the depressive phase) to aid the identification of more
selective, individual treatments for this psychiatric disorder.
This research was supported by the Academic Biomedical
Centre Neurogenomics program of the Utrecht University and
University Medical Centre Utrecht. This study makes use of data
generated by the Wellcome Trust Case Control Consortium. A full
list of the investigators who contributed to the generation of the
data is available from http://www.wtccc.org.uk. Funding for the
project was provided by the Wellcome Trust under award
076,113 (17).
The authors report no biomedical financial interests or po-
tential conflicts of interest.
www.sobp.org/journal
1130 BIOL PSYCHIATRY 2009;66:1123–1130
Supplementary material cited in this article is available
online.
1. Ormel J, Petukhova M, Chatterji S, Aguilar-Gaxiola S, Alonso J, Angerm-
eyer MC, et al. (2008): Disability and treatment of specific mental and
physical disorders across the world. Br J Psychiatry 192:368 –375.
2. de Mooij-van Malsen AJ, Olivier B, Kas MJ (2008): Behavioural genetics in
mood and anxiety: A next step in finding novel pharmacological targets.
Eur J Pharmacol 585:436 – 440.
3. Gould TD, Gottesman II (2006): Psychiatric endophenotypes and the
development of valid animal models. Genes Brain Behav 5:113–119.
4. Kas MJ, Fernandes C, Schalkwyk LC, Collier DA (2007): Genetics of be-
havioural domains across the neuropsychiatric spectrum; of mice and
men. Mol Psychiatry 12:324 –330.
5. Flint J, Valdar W, Shifman S, Mott R (2005): Strategies for mapping and
cloning quantitative trait genes in rodents. Nat Rev Genet 6:271–286.
6. Henderson ND, Turri MG, DeFries JC, Flint J (2004): QTL analysis of
multiple behavioral measures of anxiety in mice. Behav Genet 34:267–
293.
7. Mackay TF (2009): Q&A: Genetic analysis of quantitative traits. J Biol 8:23.
8. Turri MG, Talbot CJ, Radcliffe RA, Wehner JM, Flint J (1999): High-resolu-
tion mapping of quantitative trait loci for emotionality in selected
strains of mice. Mamm Genome 10:1098 –1101.
9. Kas MJ, Van Ree JM (2004): Dissecting complex behaviours in the post-
genomic era. Trends Neurosci 27:366 –369.
10. Singer JB, Hill AE, Burrage LC, Olszens KR, Song J, Justice M, et al. (2004):
Genetic dissection of complex traits with chromosome substitution
strains of mice. Science 304:445– 448.
11. Kas MJ, de Mooij-van Malsen AJ, Olivier B, Spruijt BM, Van Ree JM (2008):
Differential genetic regulation of motor activity and anxiety-related
behaviors in mice using an automated home cage task. Behav Neurosci
122:769 –776.
12. Izidio GS, Spricigo L Jr, Ramos A (2005): Genetic differences in the ele-
vated plus-maze persist after first exposure of inbred rats to the test
apparatus. Behav Proc 68:129 –134.
13. Kas MJ, Gelegen C, Schalkwyk LC, Collier DA (2009): Interspecies com-
parisons of functional genetic variations and their implications in neu-
ropsychiatry. Am J Med Genet B Neuropsychiatr Genet 150B:309 –317.
14. Kas MJ, Mooij-van Malsen A, de Krom M, van Gassen KL, van Lith HA,
Olivier B, et al. (2009): High resolution genetic mapping of mammalian
motor activity levels in mice. Genes Brain Behav 8:13–22.
15. Hillebrand JJ, Kas MJ, Adan RA (2005): Alpha-MSH enhances activity-
based anorexia. Peptides 26:1690 –1696.
16. Belknap JK (2003): Chromosome substitution strains: some quantitative
considerations for genome scans and fine mapping. Mamm Genome
14:723–732.
17. Wellcome Trust Case Control Consortium (2007): Genome-wide associ-
ation study of 14,000 cases of seven common diseases and 3000 shared
controls. Nature 447:661– 678.
18. Laarakker MC, Ohl F, van Lith HA (2008): Chromosomal assignment of
quantitative trait loci influencing modified hole board behavior in lab-
oratory mice using consomic strains, with special reference to anxiety-
related behavior and mouse chromosome 19. Behav Genet 38:159 –184.
19. Turri MG, Datta SR, DeFries J, Henderson ND, Flint J (2001): QTL analysis
identifies multiple behavioral dimensions in ethological tests of anxiety
in laboratory mice. Curr Biol 11:725–734.
20. Lynch M, Welsh B (1998): Genetics and Analysis of Quantitative Traits.
Sunderland, Massachusetts: Sinauer.
www.sobp.org/journal
J.G. de Mooij-van Malsen et al.
21. Segurado R, Detera-Wadleigh SD, Levinson DF, Lewis CM, Gill M, Nurn-
berger JI Jr, et al. (2003): Genome scan meta-analysis of schizophrenia
and bipolar disorder. Part III: Bipolar disorder. Am J Hum Genet 73:49 – 62.
22. Cheng R, Juo SH, Loth JE, Nee J, Iossifov I, Blumenthal R, et al. (2006):
Genome-wide linkage scan in a large bipolar disorder sample from the
National Institute of Mental Health genetics initiative suggests putative
loci for bipolar disorder, psychosis, suicide, and panic disorder. Mol
Psychiatry 11:252–260.
23. Levinson DF, Levinson MD, Segurado R, Lewis CM (2003): Genome scan
meta-analysis of schizophrenia and bipolar disorder. Part I: Methods
and power analysis. Am J Hum Genet 73:17–33.
24. Zandi PP, Avramopoulos D, Willour VL, Huo Y, Miao K, MacKinnon DF, et
al. (2007): SNP fine mapping of chromosome 8q24 in bipolar disorder.
Am J Med Genet B Neuropsychiatr Genet 144:625– 630.
25. Zandi PP, Zollner S, Avramopoulos D, Willour VL, Chen Y, Qin ZS, et al.
(2008): Family-based SNP association study on 8q24 in bipolar disorder.
Am J Med Genet B Neuropsychiatr Genet 147B:612– 618.
26. Gould TD, Quiroz JA, Singh J, Zarate CA, Manji HK (2004): Emerging
experimental therapeutics for bipolar disorder: insights from the molec-
ular and cellular actions of current mood stabilizers. Mol Psychiatry
9:734 –755.
27. Ferreira MA, O’Donovan MC, Meng YA, Jones IR, Ruderfer DM, Jones L, et
al. (2008): Collaborative genome-wide association analysis supports a
role for ANK3 and CACNA1C in bipolar disorder. Nat Genet 40:1056 –
1058.
28. Laursen TM, Munk-Olsen T, Nordentoft M, Bo MP (2007): A comparison
of selected risk factors for unipolar depressive disorder, bipolar affective
disorder, schizoaffective disorder, and schizophrenia from a Danish
population-based cohort. J Clin Psychiatry 68:1673–1681.
29. Minasyan A, Keisala T, Lou YR, Kalueff AV, Tuohimaa P (2007): Neopho-
bia, sensory and cognitive functions, and hedonic responses in vitamin
D receptor mutant mice. J Steroid Biochem Mol Biol 104:274 –280.
30. Matsuyama S, Doe N, Kurihara N, Tanizawa K, Kuroda S, Iso H, et al.
(2005): Spatial learning of mice lacking a neuron-specific epidermal
growth factor family protein, NELL2. J Pharmacol Sci 98:239 –243.
31. Kato T, Kubota M, Kasahara T (2007): Animal models of bipolar disorder.
Neurosci Biobehav Rev 31:832– 842.
32. Singh NA, Westenskow P, Charlier C, Pappas C, Leslie J, Dillon J, et al.
(2003): KCNQ2 and KCNQ3 potassium channel genes in benign familial
neonatal convulsions: expansion of the functional and mutation spec-
trum. Brain 126:2726 –2737.
33. Young LT, Li PP, Kish SJ, Siu KP, Kamble A, Hornykiewicz O, et al. (1993):
Cerebral cortex Gs alpha protein levels and forskolin-stimulated cyclic
AMP formation are increased in bipolar affective disorder. J Neurochem
61:890 – 898.
34. Abdulla YH, Hamadah K (1970): 3=,5= cyclic adenosine monophosphate
in depression and mania. Lancet 1:378 –381.
35. Mork A, Geisler A (1989): The effects of lithium in vitro and ex vivo on
adenylate cyclase in brain are exerted by distinct mechanisms. Neuro-
pharmacology 28:307–311.
36. Krishnan V, Graham A, Mazei-Robison MS, Lagace DC, Kim KS, Birnbaum
S, et al. (2008): Calcium-sensitive adenylyl cyclases in depression and
anxiety: behavioral and biochemical consequences of isoform target-
ing. Biol Psychiatry 64:336 –343.
37. Schaefer ML, Wong ST, Wozniak DF, Muglia LM, Liauw JA, Zhuo M, et al.
(2000): Altered stress-induced anxiety in adenylyl cyclase type VIII-defi-
cient mice. J Neurosci 20:4809 – 4820.