Dosage-dependent phenotypes in models of 16p11.2 lesions found in autism
Recurrent copy number variations (CNVs) of human 16p11.2 have been associated with a variety of developmental/neurocognitive syndromes. In particular, deletion of 16p11.2 is found in patients with autism, developmental delay, and obesity. Patients with deletions or duplications have a wide range of clinical features, and siblings carrying the same deletion often have diverse symptoms. To study the consequence of 16p11.2 CNVs in a systematic manner, we used chromosome engineering to generate mice harboring deletion of the chromosomal region corresponding to 16p11.2, as well as mice harboring the reciprocal duplication. These 16p11.2 CNV models have dosage-dependent changes in gene expression, viability, brain architecture, and behavior. For each phenotype, the consequence of the deletion is more severe than that of the duplication. Of particular note is that half of the 16p11.2 deletion mice die postnatally; those that survive to adulthood are healthy and fertile, but have alterations in the hypothalamus and exhibit a "behavior trap" phenotype-a specific behavior characteristic of rodents with lateral hypothalamic and nigrostriatal lesions. These findings indicate that 16p11.2 CNVs cause brain and behavioral anomalies, providing insight into human neurodevelopmental disorders.
Dosage-dependent phenotypes in models of 16p11.2
lesions found in autism
, Jacob Ellegood
, Jason P. Lerch
, Young-Eun E. Son
, Lakshmi Muthuswamy
, Hannes Vogel
Abba M. Krieger
, Andreas Buja
, R. Mark Henkelman
, Michael Wigler
, and Alea A. Mills
Cold Spring Harbor Laboratory, Cold Spring Harbor, NY 11724;
Mouse Imaging Centre, Hospital for Sick Children, Toronto, ON, Canada M5T 3H7;
Department of Pathology, Stanford University, Stanford, CA 94305; and
Wharton School, University of Pennsylvania, Philadelphia, PA 19104
Contributed by Michael Wigler, August 31, 2011 (sent for review July 18, 2011)
Recurrent copy number variations (CNVs) of human 16p11.2 have
been associated with a variety of developmental/neurocognitive
syndromes. In particular, deletion of 16p11.2 is found in patients
with autism, developmental delay, and obesity. Patients with de-
letions or duplications have a wide range of clinical features, and
siblings carrying the same deletion often have diverse symptoms.
To study the consequence of 16p11.2 CNVs in a systematic manner,
we used chromosome engineering to generate mice harboring
deletion of the chromosomal region corresponding to 16p11.2, as
well as mice harboring the reciprocal duplication. These 16p11.2 CNV
models have dosage-dependent changes in gene expression, viabil-
ity, brain architecture, and behavior. For each phenotype, the con-
sequence of the deletion is more severe than that of the duplication.
Of particular note is that half of the 16p11.2 deletion mice die
postnatally; those that survive to adulthood are healthy and fertile,
but have alterations in the hypothalamus and exhibit a “behavior
trap” phenotype—a speciﬁc behavior characteristic of rodents with
lateral hypothalamic and nigrostriatal lesions. These ﬁndings indi-
cate that 16p11.2 CNVs cause brain and behavioral anomalies, pro-
viding insight into human neurodevelopmental disorders.
ccumulating evidence suggests the importance of copy
number variations (CNVs) in the etiology of neuropsychi-
atric disorders, including autism (1), schizophrenia (2–4), de-
velopmental delay (5), and other complex traits (6). The 16p11.2
region is particularly intriguing. Whereas deletion of 16p11.2 has
been associated with autism (7–9), duplication of 16p11.2 has
been associated with autism (9, 10) as well as schizophrenia (11).
16p11.2 CNVs have also been reported in patients with de-
velopmental delay, mental retardation, repetitive behaviors (12–
16), and a highly penetrant form of obesity (17). A reciprocal
effect of 16p11.2 dosage on head size has been noted, as deletions
are associated with large head size or macrocephaly, whereas
duplications are associated with microcephaly (16). These studies
reveal the variability of symptoms in patients carrying the same
16p11.2 CNV, an extreme example being a family with three af-
fected members with symptoms so heterogeneous that they were
barely overlapping (18).
Mouse models allow direct assessment of CNVs while re-
ducing variability caused by genetic and environmental factors.
We and others have previously used chromosome engineering
(19) to model genetic alterations found in complex human dis-
eases including cancer (20) and genomic disorders (21–24),
allowing identiﬁcation of the causative gene and elucidation of
the mechanism involved (20, 25–27). Here we used a similar
approach to generate mouse models with deletion and duplica-
tion corresponding to those found in patients with 16p11.2
CNVs. Because of the evidence for clinical heterogeneity, we
screened these models for multiple changes in brain anatomy
and behavior by using a combination of high-resolution MRI
(28) and a monitoring system that assesses multiple behaviors
(29). We found that the deletion and the duplication affect be-
havior and brain anatomy in opposing ways, with deletion mice
exhibiting behaviors that resemble sensorimotor deﬁcits in rats
with lateral hypothalamic and nigrostriatal lesions (30, 31).
These ﬁndings provide evidence that brain anatomy and behav-
ior depend on dosage of the region corresponding to 16p11.2.
Generation of Mouse Models for Human 16p11.2 CNVs. We asked
whether altered dosage of the region corresponding to 16p11.2
causes abnormalities in mice. Genes mapping to the 0.52-Mb
16p11.2 CNV in humans cluster within a 0.44-Mb region of mouse
chromosome 7 (Fig. 1A). Using chromosome engineering (19) as
we have previously (20, 27, 32), we generated mice with one copy
[heterozygous for a deletion or deﬁciency (df) allele], as well as
mice with three copies [heterozygous for a duplication (dp) allele]
of the region corresponding to 16p11.2 (Fig. 1B and Fig. S1).
Endpoints for the rearrangement were selected based on human
data (1), with each gene in the interval being conserved in mouse
(Dataset S1). Gene targeting constructs were generated using
MICER (33), and sequential targeting in mouse ES cells resulted
in integration of loxP sites and selection cassettes at each endpoint
(Fig. 1B and Fig. S1). Cre-mediated recombination and drug se-
lection within eight independent doubly targeted clones revealed
that three clones had been targeted in cis and ﬁve clones had been
targeted in trans, which generated df/+ and df/dp ES cells, re-
spectively (Fig. 1B and Figs. S1 and S2). Five independent df/dp
clones were used for blastocyst injection, producing 40 different
male chimeras that were crossed to
females. Ten of these
chimeras (representing two independent ES cell clones) produced
df/+ and dp/+ mice that were identiﬁed by PCR (Fig. 1C). This
approach provides mouse models for directly assessing the con-
sequences of both the 16p11.2 CNV losses (i.e., deletion) and
gains (i.e., duplication) found in humans.
We established both df/+ and dp/+ mice, but at weaning we
noticed that df/+ mice were underrepresented and litter sizes
were smaller than expected (Table S1). Before weaning, df/+
mice were sometimes small (Fig. 1D), but as adults, they were
essentially the same size as their siblings and appeared healthy
(SI Experimental Procedures). To determine whether df/+ mice
were dying during embryogenesis, we crossed df/+ males to
females, and harvested embryos at day 13.5 of development [i.e.,
embryonic day (E) 13.5] as well as just before birth (E17.5–
E18.5); progeny from similar crosses using the same studs as well
as their male siblings were also genotyped at weaning (Table S1).
Whereas litter sizes during embryogenesis averaged 9.4 embryos
and the ratio of df/+ embryos was Mendelian, litter sizes at
weaning averaged only 5.0 mice and the ratio of df/+ mice was
half that expected. In addition, litter sizes were normal and df/+
Author contributions: G.H., M.W., and A.A.M. designed research; G.H., J.E., Y.-E.E.S., H.V.,
and A.A.M. performed research; G.H., R.M.H., and A.A.M. contributed new reagents/
analytic tools; G.H., J.E., J.P.L., L.M., A.M.K., A.B., and A.A.M. analyzed data; and G.H.,
J.E., M.W., and A.A.M. wrote the paper.
The authors declare no conﬂict of interest.
Data deposition: The data reported in this paper have been deposited in the Gene Ex-
pression Omnibus (GEO) database, www.ncbi.nlm.nih.gov/geo (accession no. GSE32012).
Present address: Ontario Institute for Cancer Research, Toronto, ON, Canada M5G 2M9.
To whom correspondence may be addressed. E-mail: firstname.lastname@example.org or email@example.com.
This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.
October 11, 2011
no. 41 www.pnas.org/cgi/doi/10.1073/pnas.1114042108
mice were present in expected ratios immediately after birth,
whereas dead pups lacking a milk pouch were sometimes found
later on. Therefore, some df/+ mice die after birth, indicating
that 16p11.2 loss can compromise survival.
Gene Expression in Multiple Brain Regions Corresponds to 16p11.2
To validate the models, we analyzed gene expression
proﬁles in the brain and determined whether expression corre-
sponded with dosage. We measured mRNA intensities in 37
microarray hybridizations representing four brain regions (ol-
factory bulbs, cortex, cerebellum, and brainstem; ﬁve samples
were hybridized twice for estimation of technical errors) in two
, and three dp/+ male mice. All mice were F1
C57BL/6N:129Sv hybrids; therefore, other than the engineered
CNV, their genomes were identical. A scatter plot of the gene
expression intensity difference between dp/+ and
difference between df/+ and
indicated that genes within
16p11.2 displayed a large difference between df/+ and
brain, and a much smaller difference between dp/+ and
brain (Fig. S3A). Two-way ANOVA with brain region and dos-
age as main factors indicated that, of 33 genes in the engineered
region, expression of 26 was affected directly by dosage (Dataset
S2). Gdpd3, mapping within the engineered region, showed ex-
treme up- and down-regulation; this reﬂected differences in
Gdpd3 expression in C57BL/6N vs. 129Sv strains. Further anal-
ysis indicated that expression of genes in the region was signiﬁ-
cantly altered in each of the four brain regions analyzed, and that
expression was affected more by deletion than by duplication
(Fig. S3B). These ﬁndings indicate that copy number dictates
gene expression levels in multiple brain regions, and that loss has
the largest effect.
df/+ and dp/+ Mice Have Distinctive Behavioral Phenotypes. General
survey of behavior.
The clinical evidence that patients with 16p11.2
CNVs have highly heterogeneous symptoms suggested that if the
corresponding genomic alteration did in fact cause behavioral
alterations in mice, the phenotypes might also be highly variable.
Therefore, we believed it imperative to monitor the 16p11.2
CNV models for multiple behaviors by using as quantiﬁable and
unbiased approaches as possible. We used HomeCageScan, a
system previously used to assess behavioral alterations caused by
neurodegenerative disease, neurotoxic agents, and pain (29, 34–
36). We investigated behavior of a cohort of 50 male and female
mice. The mice were progeny from df/+ × dp/+ crosses, and
and df/dp diploid controls (n = 15 and
n = 9, respectively), df/+ (n = 13), and dp/+ (n = 13). The parents
in these crosses came from two chimeras. Thirty-nine of these
were later used for MRI (as detailed later). Recording was done
in cages that were signiﬁcantly larger, with a ceiling that was
much higher, than a standard mouse cage. The reason for using
large cages is that, by minimizing the physical constraints on the
animals, a rich spectrum of behaviors evolves (37) and the dy-
namics of the change in behavior varies between genotypes (38).
In this experimental paradigm, the recording cages posed a new
environment to the mice being analyzed. In particular, mice had
to adapt their climbing abilities to this new environment. In each
session, the behavior of four individual mice was recorded
simultaneously. Multiple sessions were performed so that the
behavior of each of the 50 mice was analyzed. Mice were
transferred into the recording cages before the last 2 h of the
light period. The recording started immediately and continued
for 60 h after the onset of the ﬁrst dark period (i.e., over three 12-
h dark periods and two 12-h light periods). We also tested social
behavior and grip strength (Fig. S4). These analyses did not show
signiﬁcant differences and therefore we do not discuss those data.
However, six of eight distinct behavioral measures revealed
signiﬁcant genotype differences. The changes were evident im-
mediately after the mice were introduced into the new cages, as
well as throughout the entire period of the test (Fig. 2A and
Dataset S3). Five of these six differences were particularly in-
teresting, as these behaviors were affected in opposite directions
in df/+ and dp/+ mice relative to diploid (
and df/dp) con-
trols. As was the case for gene expression proﬁles, the effect of
the deletion on behavior was more severe than that of the du-
plication. As 16p11.2 CNV-associated syndromes sometimes
have a gender bias (16), we asked whether the changes in be-
havior were sex-speciﬁc. Two-way ANOVA did not reveal sig-
niﬁcant interactions between 16p11.2 dosage and sex for any of
the behavioral measures (Dataset S3). Later we describe in detail
the behaviors affected by 16p11.2 dosage.
Response to change in environment. The 16p11.2 CNV models
responded uniquely to environmental change: within the 2-h
period after being transferred to the test cage, the distance
traveled, as well as the time spent walking, lingering, and resting,
depended on genotype (P
= 0.0016, P
= 0.021, P
= 0.025; Dataset S3). Tukey’s conﬁ-
Fig. 1. Generation of 16p11.2 models. (A) Genes map-
ping to human 16p11.2 CNVs are conserved in mouse. (B)
Schematic of the chromosome engi neering strategy used
to generate mouse models of 16p11.2 CNVs. Step 1 is
gene targeting at the 135k15 locus; step 2 is gene tar-
geting at the 216k12 locus in the 135k15-targeted ES
cells; and step 3 is Cre-mediated recombination. Cis and
trans indicate that loxP sites (yellow triangles) had in-
tegrated on the same or different chromosome homo-
logues, respectively. (C) Molecular validation. PCR pro-
ducts using primers speciﬁc for the positive control
(β-actin), targeting at the ﬁrst and second endpoints
(135k15 and 216k12 loci, respectively), the df allele, and
the dp allele are shown. (D) Before weaning, df/+ mice
(Right) tend to be smaller than their
and 15.4 g, respectively, for the females shown). Note the
light-colored tail and ears of the df/+, which is a result of
the presence of the Agouti transgene. In A and B, chro-
mosome positions are shown in megabases. Further in-
formation is provided in Figs. S1 and S2, Dataset S1,and
SI Experimental Procedures.
Horev et al. PNAS
October 11, 2011
dence intervals were calculated to determine pair-wise differ-
ences between genotypes while controlling for false discovery
rate (FDR) (39). Distance traveled and time spent walking were
signiﬁcantly higher in df/+ relative to other groups, and the time
spent lingering was signiﬁcantly higher in df/+ relative to dp/+
mice. Resting was lower in df/+ and higher in dp/+ mice relative
to controls. We further analyzed the time course of these
changes during the ﬁrst 2 h of the test (Fig. 2 B and C). Mice of
all four genotypes were most active immediately after being
transferred to the test cage, with activity gradually declining
during the ﬁrst hour as the mice habituated to their new envi-
ronment (Fig. 2B). However, the df/+ cohort had a second burst
of activity that did not occur in other groups. The response to
being transferred to the test cage occurred in three sequential
stages deﬁned by the actions of the mice (Fig. 2C). For each
group, the ﬁrst stage was characterized by elevated walking and
rearing, and the second stage was characterized by elevated
grooming. The third stage was characterized by resting, which
was signiﬁcantly decreased and increased in df/+ and dp/+, re-
spectively, relative to
; indeed, df/+ mice had a burst of
walking and rearing, whereas resting was absent during this
stage. Thus, the rate of certain behaviors is affected reciprocally
by loss and gain of 16p11.2 dosage in response to environmental
challenge. In addition, the sequence of these behaviors is dis-
rupted in df/+ mice. Thus, 16p11.2 CNVs affect both the rate
and the timing of speciﬁc behaviors.
Diurnal deﬁcits. Sleeping disorders are frequently reported in many
psychiatric disorders, including autism (40). Because we recor-
ded behavior over sequential dark/light intervals, we could assess
the effect of 16p11.2 CNVs on light and dark cycling by using
previously established methods (29). Mice are nocturnal, and
indeed, each genotype was most active during the dark periods
(Fig. 3A). The
, df/+, and dp/+ mice were most active during
the initial dark period, with activity decreasing in successive dark
periods. Although activity of df/+ mice was highest in the ﬁrst
dark period, during subsequent dark cycles, the mice adapted
and had activity levels similar to controls. The activity of df/dp
mice was indistinguishable in light and dark periods; dp/+
mice were notably less active in the dark (but not light) period
(Fig. 3B). In striking contrast, df/+ mice were signiﬁcantly more
active than mice of other genotypes in both light and dark
periods. Furthermore, df/+ mice were unique, as they had
a higher ratio of light to dark activity compared with the other
Fig. 2. Multiple behaviors depend on 16p11.2 dosage. (A) The relationship
between 16p11.2 dosage and distance traveled (Top), hanging (Middle), and
resting (Bottom) is shown during the 2-h period immediately after cage
transfer (Left), the light periods ( Middle), and the dark periods (Right).
Averages and standard error of the mean (SEM) are displayed. Statistically
signiﬁcant pair-wise differences relative to the df/+ group (determined by
ANOVA followed by Tukey’sconﬁdence intervals) are depicted by the fol-
lowing: df/+ is cyan when it differs signiﬁcantly from at least one other
cohort, but is otherwise black; cohorts that differ signiﬁcantly from df/+ are
magenta, but are otherwise black. Dataset S3 shows pair-wise comparisons
between genotypes. (B) Median distance traveled vs. time during the ﬁrst 2
h after cage transfer. (C) Detailed behaviors during the ﬁrst 2 h after cage
transfer. Medians of the cumulative time of four distinct behaviors (rearing,
walking, grooming, resting) vs. time elapsed from the beginning of the trial
are shown for each genotype. During this period, df/+ mice do not rest;
instead, they have a second peak of activity (B and C). Further information is
provided in Dataset S3 and SI Experimental Procedures.
Fig. 3. Behavior/diurnal rhythms are affected in 16p11.2 CNV mice. Com-
parison of the distance traveled in cohorts of distinct genotypes (detailed in
the text) during 60 h of alternating 12-h dark and 12-h light cycles (i.e.,
spanning three dark periods plus two light periods). (A) The distance trav-
eled over successive dark (black bars) and light (gray bars) periods is shown in
bins of 12 h. Five successive dark or light periods are shown. (B) The average
distance traveled during light vs. the average distance traveled during dark.
(C) The ratio of light-to-dark activity indicates that df/+ mice are unusual in
that their activity is not as restricted to the dark periods as the controls are.
Average and SEM are presented in all panels.
www.pnas.org/cgi/doi/10.1073/pnas.1114042108 Horev et al.
genotypes (Fig. 3C). These ﬁndings indicate that 16p11.2 CNVs
affect diurnal behaviors.
Climbing deﬁcits. The most signiﬁcant genotype effect reported by
HomeCageScan was that df/+ mice remained on the ceiling of
the cage for extended periods (P
= 0.000021; Fig. 2A).
Therefore, we further investigated the climbing patterns of the
mice. The ceiling-climbing behavior of controls was dynamic and
changed over the course of the session. Shortly after being in-
troduced into the test cage, diploid controls climbed up to the
lower part of the V-shaped ceiling, remained there brieﬂy, and
then returned to the ﬂoor. During this early phase of testing,
control mice returned to the ﬂoor with the rear part of their
bodies leading, i.e., they hung on the ceiling with their forelimbs,
touched the ﬂoor with their hindlimbs, and then left the ceiling
(Movie S1). In subsequent climbing episodes, control mice
traveled to higher and more distant locations on the ceiling,
gradually progressing to the highest point of the cage. The
climbing behavior of controls developed in two dimensions: ﬁrst,
they left the ceiling from different locations and returned to
different places on the ﬂoor; second, they could climb down from
the ceiling with their head and forepaws leading, i.e., they hung
on the ceiling by their hindlimbs and then touched down on the
ﬂoor with their forelimbs.
In contrast to the adaptability of controls, the ceiling-climbing
behavior of df/+ mice was extremely stereotypic throughout the
test period. Like control mice during the early phase of being
introduced into the test cage, df/+ mice returned to the ﬂoor
with their hindlimbs leading. However, in contrast to control
mice, df/+ mice did not progress to the stage at which they were
able to climb down from the ceiling with their head and fore-
limbs leading. In further contrast to controls, df/+ mice did not
climb off the ceiling from different spots; they continued to go up
to and down from the ceiling at the same location (Table 1).
Some df/+ mice became “trapped” on the ceiling for extensive
periods, apparently lacking the ability to return to the ﬂoor of the
cage (Movie S2). Other df/+ mice developed stereotypic ways of
coming down from the ceiling (Movie S3) that they repeated
hundreds of times during the course of the session. This re-
petitive behavior continued throughout the recording period,
even after the mice had performed hundreds of climbing epi-
sodes. This analysis revealed that 16p11.2 deletion mice show
nonprogressive, stereotypic motor behavior that is similar to
stereotypic behavior caused by lateral hypothalamic and nigros-
triatal lesions (30, 31).
16p11.2 CNV Models Have Distinct Changes in Brain Architecture. To
identify brain regions altered in 16p11.2 CNV mice, we used
MRI to analyze the brains of 39 mice from the cohort that had
already been analyzed for behavioral phenotypes (Fig. 4) (28).
We included both male and female mice in the cohort, which
and df/dp diploid controls (n = 9 and n =8,
respectively), df/+ (n = 11), and dp/+ (n = 11). Anesthetized
mice were perfused and euthanized, and the brain (which
remained within the skull) was subjected to MRI. Sixty-two dif-
ferent brain regions (41) were examined, and their volumes were
assessed as the percentage of total brain volume averaged for
each of the four models (Dataset S4).
Signiﬁcant changes between brains of df/+ and
, as well as
between brains of df/+ and dp/+ mice, were noted (Fig. 4 and
Fig. S5). Although brains of
and dp/+ mice were not sig-
niﬁcantly different, a clear trend was found for some regions.
Brain structures signiﬁcantly affected after stringent correction
for multiplicity (with the Holm procedure) included the basal
forebrain, superior colliculus, fornix, hypothalamus, mammillo-
thalamic tract, medial septum, midbrain, and periaqueductal
gray (Fig. 4 A and B). For each structure, the volumetric changes
were more extensive between df/+ and dp/+ than between df/+
, indicating that loss and gain of 16p11.2 dosage affects
these regions in opposite ways (Fig. 4 and Fig. S5).
Because the “behavior trap” resembles a phenotype described
in rats with lesions in the lateral hypothalamus, we performed
detailed MRI analysis of the hypothalamus. Most changes be-
tween df/+ and dp/+ were located in the posterior region of the
hypothalamus, with pronounced changes in the lateral zone (Fig.
5). These ﬁndings support the hypothesis that the lateral hypo-
thalamus is affected in df/+ mice. In addition, we found that
Mapk3—which maps within the region corresponding to human
16p11.2—is expressed robustly in speciﬁc brain regions including
the lateral hypothalamus and the nigrostriatal tract (Fig. S6).
These ﬁndings demonstrate that altered dosage of 16p11.2 cau-
ses changes in the size of several brain structures, and that de-
letion and duplication have opposing effects.
16p11.2 CNV Models Provide Insight into Human Syndromes. CNVs
affecting 16p11.2 have been associated with autism and other
neurodevelopmental/neuropsychiatric syndromes (1, 7, 9, 12–
16), yet several issues remain unresolved. Are these conditions
unique to humans? Do loss and gain cause the same syndrome?
Does dosage of 16p11.2 affect brain architecture? Why are
symptoms of patients with the same CNV so diverse? To begin to
address these issues, we engineered mice heterozygous for de-
letion and duplication of the interval corresponding to 16p11.2
CNVs found in humans. The striking changes we discovered in
gene expression proﬁles, viability, brain architecture, and most
importantly behavior, provide functional evidence that 16p11.2
CNVs cause phenotypes in mice, that loss and gain have opposing
effects, and that multiple brain regions and behaviors are affected.
Our ﬁnding that brain volume size is affected reciprocally in de-
letion vs. duplication mice is concordant with the macrocephaly
and microcephaly observed in human subjects with 16p11.2 de-
letion and duplication, respectively, indicating that our animal
models recapitulate the human genomic disorders.
The ﬁnding that mice with the same CNVs present in humans
have neuroanatomical and behavioral phenotypes indicates that
16p11.2 genes are important for brain function in mammals other
than humans. For some human CNV-associated syndromes such as
7q11.23 deletion (i.e., Williams–Beuren syndrome) and the re-
ciprocal duplication (42), loss and gain are associated with opposing
clinical features. Indeed, this is the case for head size alterations
associated with 16p11.2 CNVs (16), but certainly not for behavioral
symptoms of these patients (10, 13, 16). In mice, we see that loss and
gain of 16p11.2 cause distinct and opposing behavioral phenotypes.
Similarly, mouse chromosome engineered models of human
17p11.2 deletion/duplication-associated syndromes had opposing
phenotypes for some, but not all, clinical phenotypes studied (22).
The side-by-side comparison of mice with deletion and dupli-
cation of the region corresponding to human 16p11.2 reveals that
expression of most genes within the engineered interval correlates
directly with dosage, and that a number of neuroanatomical and
behavioral phenotypes are affected in opposing directions by loss
and gain. Deletion has a more severe effect than duplication on
each phenotype—viability, gene expression, brain structure, and
behavior—in keeping with the severity of deletion vs. duplication
of 16p11.2 in humans. For examples, duplications are occasionally
seen in asymptomatic carriers, but carriers of the deletion are
Table 1. Number of mice from each cohort that performed
different climbing behaviors
Group Travel on ceiling
df/+ 13/13 2/13 1/13 4/13
15/15 15/15 15/15 0/15
df/dp 6/9 6/9 6/9 1/9
dp/+ 1/13 1/13 1/13 0/13
Data shown are from the second night of the test, as shown in Movies S1,
S2, and S3. Note that most of the dp/+ mice did not climb on the ceiling.
Horev et al. PNAS
October 11, 2011
extremely rare; in addition, patients with 16p11.2 deletions tend
to be diagnosed earlier than those with duplications (16).
16p11.2 CNVs Affect Many Brain Regions. Changes in head cir-
cumference and abnormal brain structure have been reported in
patients with 16p11.2 CNVs (14, 16). By using MRI, we ﬁnd
signiﬁcant volumetric changes in eight different brain regions.
Brains of df/+ (but not dp/+) mice, have signiﬁcant volumetric
changes relative to controls, but the most extensive difference is
between df/+ and dp/+ mice, emphasizing the opposing effects
that 16p11.2 dosage has on brain architecture. Importantly,
brains of df/dp diploid controls are not signiﬁcantly different
controls, providing genetic evidence that the structural
changes in df/+ and dp/+ models are dosage-dependent.
16p11.2 CNVs Affect Multiple Behaviors. Several human studies
compared the behavioral symptoms of patients with 16p11.2
deletions and duplications (10, 13, 16); however, to our knowl-
edge, there is no evidence that loss and gain of 16p11.2 affect
behavior in opposing ways. Even with patients harboring the same
16p11.2 lesion, there is a broad spectrum of clinical symptoms,
some patients being severely affected and others highly func-
tional. By simultaneously analyzing multiple behaviors in the
context of a new environment, we identify a number of behaviors
that are altered in 16p11.2 CNV mice, revealing that deletion and
duplication have opposite consequences. These highly signiﬁcant
changes survive strict statistical analyses (37, 43). Each genotype
responds to the new cage with heightened activity, but only df/+
mice have a second burst of activity at a time when controls are
already resting. When control mice have become accustomed to
their new environment, they have a gradual increase in freedom of
movement on the ceiling over the course of the trial, i.e., a mo-
bility gradient that recapitulates the ontogeny of movement (44).
In contrast, df/+ mice do not show the mobility gradient: their
ceiling-climbing behavior is restricted to speciﬁc locations and
their movements are stereotypic. Interestingly, this ceiling-
climbing behavior is similar to the behavior trap described in rats
with lateral hypothalamic lesions and 6-hydroxydopamine–in-
duced lesions (30, 31), a well characterized model of Parkinson
disease. Other phenotypes of these rats are feeding problems (45,
46), sensory neglect, and abnormal gait (30, 31, 47–49). Indeed,
abnormal gait and motor delay (13, 16, 18, 50), attention deﬁcits
(13), and feeding defects (16) are common in patients with
16p11.2 deletion. Moreover, motor development problems are
common in autism spectrum disorders and may serve as an in-
dicator for early intervention, as these features appear before the
core symptoms that deﬁne autism (51).
Fig. 4. MRI identiﬁes structural changes in brains of 16p11.2
CNV mice. The relative volume (percentage of total brain
volume) of eight brain regions is increased in df/+ mice. (A)
Three-dimensional representation of the mouse brain high-
lights eight regions (colored as in legend) affected by 16.p11.2
dosage. ( B) Relative volumes (shown as percentage of total
brain volume) are dependent on dosage. Mean and SEM are
shown. Statistically signiﬁcant pairwise differences to the df/+
group (determined by t test followed by Bonferroni–Holm
procedure) are depicted as follows: cyan indicates that df/+
differs from at least one other cohort, magenta indicates
cohorts that differ signiﬁcantly from df/+, and black indicates
groups that do not differ signiﬁcantly from df/+. Full pair-wise
comparisons are shown in Dataset S4.
Fig. 5. Details of alterations in the hypothalamus detected by MRI. Three-
dimensional models of the surface of the hypothalamus (Bottom), coronal
images showing the regions affected (Middle), and magniﬁcation focusing on
the hypothalamus (Top). Red indicates voxels that differ signiﬁcantly between
df/+ and dp/+ cohorts with an FDR of 0.05. The sections performed along four
locations marked A–D (A, most posterior; D, most anterior). Colors indicate
voxels that differ signiﬁcantly between df/+ and dp/+ cohorts, with brightness
indicating the signiﬁcance of the difference, as speciﬁed by the FDR. AHN,
anterior hypothalamic nucleus; DMH, dorsomedial hypothalamus; FX, col-
umns of the fornix; LHA, lateral hypothalamic area; MTT, mammillothalamic
tract; OPT, optic tract; PH, posterior hypothalamic nucleus; ZI, zona incerta.
www.pnas.org/cgi/doi/10.1073/pnas.1114042108 Horev et al.
Deletion of 16p11.2 Causes Lethality in Neonates. A major ﬁnding of
this work is that approximately half of df/+ neonates die after
birth, a ﬁnding that may have relevance to autism incidence. The
precise cause of death in df/+ mice could be related to feeding
deﬁcits, but this remains to be investigated. Based on our ﬁndings,
we suggest that efforts be made to determine whether 16p11.2
deletion is associated with unexplained cases of infant death. If
these ﬁndings generalize to other genotypes associated with au-
tism, they may explain puzzling aspects of the human condition.
The recent increase in autism incidence (52) might be partially
attributable to factors that improve pre- and/or postnatal survival.
Human studies are consistent with this idea, as it is much more
common for inherited rare copy number polymorphisms that af-
fect coding regions to be duplications than deletions (53).
Closing. This work demonstrates the value of using mice to model
CNVs found in human disorders. This approach provides func-
tional evidence that 16p11.2 CNVs affect brain anatomy and
behavior in mice, with loss and gain having opposing effects.
Multiple brain regions are affected, with deletion of 16p11.2
causing profound behavioral changes such as hyperactivity, dif-
ﬁculty adapting to change, sleeping abnormalities, and repetitive
or restricted behaviors. In addition, our ﬁndings suggest a po-
tential link between 16p11.2 copy number alterations and infant
mortality. Finally, we note a similarity in phenotype between
16p11.2 deletions and rats with lateral hypothalamic lesions.
These 16p11.2 CNV models should prove valuable for eluci-
dating the physiological basis of neurodevelopmental syndromes
and for evaluating their treatments.
Mice carrying rearrangements corresponding to the human CNVs (Dataset S1)
were established by using chromosome engineering as described previously (19,
20, 27, 32). HomeCageScan system (CleverSys) was used to analyze behavior in
, df/dp,anddp/+ mice. Thirty-nine of these mice
were also analyzed by MRI. Hypothesis testing was followed by correction for
multiplicity (SI Experimental Procedures provides additional details).
ACKNOWLEDGMENTS. We thank members of the A.A.M. laboratory,
D. Yekutieli, and I. Golani for input; L. Bianco, G. Munoz, and N. Alston for
animal husbandry; and C. Johns for microarray hybridizations. This project
was funded by Simons Foundation Autism Research Initiative.
1. Sebat J, et al. (2007) Strong association of de novo copy number mutations with
autism. Science 316:445–449.
2. Liu H, et al. (2002) Genetic variation at the 22q11 PRODH2/DGCR6 locus presents an unusual
pattern and increases susceptibility to schizophrenia. Proc Natl Acad Sci USA 99:3717–3722.
3. Stefansson H , et al.; GROUP (2008) Large recurrent microdeletions associated with
schizophrenia. Nature 455:232–236.
4. Stone JL, et al.; International Schizophrenia Consortium (2008) Rare chromosomal
deletions and duplications increase risk of schizophrenia. Nature 455:237–241.
5. Moeschler JB (2008) Medical genetics diagnostic evaluation of the child with global
developmental delay or intellectual disability. Curr Opin Neurol 21:117–122.
6. Stankiewicz P, Lupski JR (2010) Structural variation in the human genome and its role
in disease. Annu Rev Med 61:437–455.
7. Kumar RA, et al. (2008) Recurrent 16p11.2 microdeletions in autism. Hum Mol Genet
8. Marshall CR, et al. (2008) Structural variation of chromosomes in autism spectrum
disorder. Am J Hum Genet 82:477–488.
9. Weiss LA, et al.; Autism Consortium (2008) Association between microdeletion and
microduplication at 16p11.2 and autism. N Engl J Med 358:667–675.
10. Fernandez BA, et al. (2010) Phenotypic spectrum associated with de novo and in-
herited deletions and duplications at 16p11.2 in individuals ascertained for diagnosis
of autism spectrum disorder. J Med Genet 47:195–203.
11. McCarthy SE, et al.; Wellcome Trust Case Control Consortium (2009) Microduplications
of 16p11.2 are associated with schizophrenia. Nat Genet 41:1223–1227.
12. Bijlsma EK, et al. (2009) Extending the phenotype of recurrent rearrangements of
16p11.2: Deletions in mentally retarded patients without autism and in normal in-
dividuals. Eur J Med Genet 52:77–87.
13. Rosenfeld JA, et al. (2010) Speech delays and behavioral problems are the pre-
dominant features in individuals with developmental delays and 16p11.2 micro-
deletions and microduplications. J Neurodev Disord 2:26–38.
14. Schaaf CP, et al. (2011) Expanding the clinical spectrum of the 16p11.2 chromosomal
rearrangements: three patients with syringomyelia. Eur J Hum Genet 19:152–156.
15. Shimojima K, Inoue T, Fujii Y, Ohno K, Yamamoto T (2009) A familial 593-kb microdeletion of
16p11.2 associa ted withmental retardation andhemivertebrae. Eur J Med Genet 52:433–435.
16. Shinawi M, et al. (2010) Recurrent reciprocal 16p11.2 rearrangements associated with
global developmental delay, behavioural problems, dysmorphism, epilepsy, and ab-
normal head size. J Med Genet 47:332–341.
17. Walters RG, et al. (2010) A new highly penetrant form of obesity due to deletions on
chromosome 16p11.2. Nature 463:671–675.
18. Shen Y, et al. (2011) Intra-family phenotypic heterogeneity of 16p11.2 deletion carriers in
a three-generation Chinesefamily. Am J Med GenetB NeuropsychiatrGenet 156:225–232.
19. Mills AA, Bradley A (2001) From mouse to man: Generating megabase chromosome
rearrangements. Trends Genet 17:331–339.
20. Bagchi A, et al. (2007) CHD5 is a tumor suppressor at human 1p36. Cell 128:459–475.
21. Lindsay EA, et al. (1999) Congenital heart disease in mice deﬁcient for the DiGeorge
syndrome region. Nature 401:379–383.
22. Ricard G, et al. (2010) Phenotypic consequences of copy number variation: insights from
Smith-Magenis and Potocki-Lupski syndrome mouse models. PLoS Biol 8:e1000543.
23. Walz K, et al. (2003) Modeling del(17)(p11.2p11.2) and dup(17)(p11.2p11.2) contig-
uous gene syndromes by chromosome engineering in mice: Phenotypic consequences
of gene dosage imbalance. Mol Cell Biol 23:3646–3655.
24. Li HH, et al. (2009) Induced chromosome deletions cause hypersociability and other
features of Williams-Beuren syndrome in mice. EMBO Mol Med 1:50–65.
25. Nakatani J, et al. (2009) Abnormal behavior in a chromosome-engineered mouse
model for human 15q11-13 duplication seen in autism. Cell 137:1235–1246.
26. Anney R, et al. (2010) A genome-wide scan for common alleles affecting risk for
autism. Hum Mol Genet 19:4072–4082.
27. Stark KL, et al. (2008) Altered brain microRNA biogenesis contributes to phenotypic
deﬁcits in a 22q11-deletion mouse model. Nat Genet 40:751–760.
28. Bock NA, et al. (2003) High-resolution longitudinal screening with magnetic reso-
nance imaging in a murine brain cancer model. Neoplasia 5:546–554.
29. Steele AD, Jackson WS, King OD, Lindquist S (2007) The power of automated high-
resolution behavior analysis revealed by its application to mouse models of Hun-
tington’s and prion diseases. Proc Natl Acad Sci USA 104:1983–1988.
30. Schallert T, Whishaw IQ, Ramirez VD, Teitelbaum P (1978) Compulsive, abnormal
walking caused by anticholinergics in akinetic, 6-hydroxydopamine-treated rats. Sci-
31. Golani I, Wolgin DL, Teitelbaum P (1979) A proposed natural geometry of recovery
from akinesia in the lateral hypothalamic rat. Brain Res 164:237–267.
32. Guo X, et al. (2009) TAp63 induces senescence and suppresses tumorigenesis in vivo.
Nat Cell Biol 11:1451–1457.
33. Adams DJ, et al. (2004) Mutagenic insertion and chromosome engineering resource
(MICER). Nat Genet 36:867–871.
34. Bordone L, et al. (2007) SIRT1 transgenic mice show phenotypes resembling calorie
restriction. Aging Cell 6:759–767.
35. Roughan JV, Wright-Williams SL, Flecknell PA (2009) Automated analysis of post-
operative behaviour: Assessment of HomeCageScan as a novel method to rapidly
identify pain and analgesic effects in mice. Lab Anim 43:17–26.
36. Chen L, Zhang X, Chen-Roetling J, Regan RF (2011) Increased striatal injury and be-
havioral deﬁcits after intracerebral hemorrhage in hemopexin knockout mice. J
37. Benjamini Y, et al. (2010) Ten ways to improve the quality of descriptions of whole-
animal movement. Neurosci Biobehav Rev 34:1351–1365.
38. Horev G, Benjamini Y, Sakov A, Golani I (2007) Estimating wall guidance and at-
traction in mouse free locomotor behavior. Genes Brain Behav 6:30–41.
39. Benjamini Y, Krieger AM, Yekutieli D (2006) Adaptive linear step-up procedures that
control the false discovery rate. Biometrika 93:491–507.
40. Bourgeron T (2007) The possible interplay of synaptic and clock genes in autism
spectrum disorders. Cold Spring Harb Symp Quant Biol 72:645–654.
41. Dorr AE, Lerch JP, Spring S, Kabani N, Henkelman RM (2008) High resolution three-
dimensional brain atlas using an average magnetic resonance image of 40 adult
C57Bl/6J mice. Neuroimage 42:60–69.
42. Merla G, Brunetti-Pierri N, Micale L, Fusco C (2010) Copy number variants at Williams-
Beuren syndrome 7q11.23 region. Hum Genet 128:3–26.
43. Benjamini Y, Drai D, Elmer G, Kafkaﬁ N, Golani I (2001) Controlling the false discovery
rate in behavior genetics research. Behav Brain Res 125:279–284.
44. Golani I (1992) A mobility gradient in the organization of vertebrate movement: The
perception of movement through symbolic language. Behav Brain Sci 15:249–266.
45. Teitelbaum P, Epstein AN (1962) The lateral hypothalamic syndrome: Recovery of
feeding and drinking after lateral hypothalamic lesions. Psychol Rev 69:74–90.
46. Ungerstedt U (1971) Stereotaxic mapping of monoamine pathways in rat brain. Acta
Physiol Scand Suppl 367:1–48.
47. Marshall JF, Teitelbaum P (1974) Further analysis of sensory inattention following
lateral hypothalamic damage in rats. J Comp Physiol Psychol 86:375–395.
48. Marshall JF, Richardson JS, Teitelbaum P (1974) Nigrostriatal bundle damage and the
lateral hypothalamic syndrome. J Comp Physiol Psychol 87:808–830.
49. Marshall JF, Turner BH, Teitelbaum P (1971) Sensory neglect produced by lateral
hypothalamic damage. Science 174:523–525.
50. Hanson E, et al.; 16p11.2 Study Group Clinicians (2010) Cognitive and behavioral
characterization of 16p11.2 deletion syndrome. J Dev Behav Pediatr 31:649–657.
51. Lloyd M, Macdonald M, Lord C (2011) Motor skills of toddlers with autism spectrum
disorders. Autism 10.1177/1362361311402230.
52. Newschaffer CJ, et al. (2007) The epidemiology of autism spectrum disorders. Annu
Rev Public Health 28:235–258.
53. Levy D, et al. (2011) Rare de novo and transmitted copy-number variation in autistic
spectrum disorders. Neuron 70:886–897.
Horev et al. PNAS
October 11, 2011