Conserved epigenetic sensitivity to early life
experience in the rat and human hippocampus
Matthew Sudermana,b,c,1, Patrick O. McGowand,1, Aya Sasakid,1, Tony C. T. Huangb, Michael T. Hallettc,
Michael J. Meaneya,e,f,g, Gustavo Tureckie, and Moshe Szyfa,b,g,2
aSackler Program for Epigenetics and Developmental Psychobiology at McGill University andbDepartment of Pharmacology and Therapeutics, McGill
University, Montreal, QC, Canada H3G 1Y6;cMcGill Centre for Bioinformatics, Montreal, QC, Canada H3G 1Y6;dCentre for the Neurobiology of Stress,
Department of Biological Sciences, University of Toronto Scarborough, Toronto, ON, Canada M1C 1A4;eDouglas Mental Health University Institute, Montreal,
QC, Canada H4H 1R3;fSingapore Institute for Clinical Sciences, Agency for Science, Technology and Research, Singapore 117609; andgExperience-Based Brain
and Biological Development Program, Canadian Institute for Advanced Research, Toronto, ON, Canada M5G 1Z8
Edited by Gene E. Robinson, University of Illinois at Urbana–Champaign, Urbana, IL, and approved May 10, 2012 (received for review December 23, 2011)
Early life experience is associated with long-term effects on
behavior and epigenetic programming of the NR3C1 (GLUCOCOR-
TICOID RECEPTOR) gene in the hippocampus of both rats and
humans. However, it is unlikely that such effects completely cap-
ture the evolutionarily conserved epigenetic mechanisms of early
adaptation to environment. Here we present DNA methylation pro-
files spanning 6.5 million base pairs centered at the NR3C1 gene in
the hippocampus of humans who experienced abuse as children
and nonabused controls. We compare these profiles to correspond-
ing DNA methylation profiles in rats that received differential lev-
els of maternal care. The profiles of both species reveal hundreds of
DNA methylation differences associated with early life experience
distributed across the entire region in nonrandom patterns. For
instance, methylation differences tend to cluster by genomic loca-
tion, forming clusters covering as many as 1 million bases. Even
more surprisingly, these differences seem to specifically target reg-
ulatory regions such as gene promoters, particularly those of the
protocadherin α, β, and γ gene families. Beyond these high-level
similarities, more detailed analyses reveal methylation differences
likely stemming from the significant biological and environmental
differences between species. These results provide support for an
analogous cross-species epigenetic regulatory response at the level
of the genomic region to early life experience.
well as humans. For example, differences in maternal care in
rats during the first week of life are associated with long-term
effects on behavior and brain function that persist into adult-
hood, including alterations in the stress response (1). In humans,
similar effects are observed. For instance, childhood maltreat-
ment associates with development of both externalizing and in-
ternalizing personality traits and psychopathology in adulthood
(2). The association in both rats and humans of stable devel-
opmental phenotypes with early life experience suggests that
molecular mechanisms may serve as a memory of these early life
experiences in both species. In fact, there is evidence that these
long-term effects are, at least in part, mediated by epigenetic
alterations in the brain. In particular, recent studies have found
aberrant DNA methylation in the NR3C1 (GLUCOCORTICOID
RECEPTOR) gene promoter of the hippocampi of both rats and
humans associated with differential early life experience (3, 4).
Exposure of infant rats to stressed caretakers displaying abusive
behavior produced persisting changes in methylation of the
BDNF gene promoter in the adult prefrontal cortex (5). Early
life stress in mice caused sustained DNA hypomethylation of an
important regulatory region of the AVP gene (6).
Although explanations involving a single site are appealing,
it is unlikely that the broad systemic response to early life ex-
perience would be associated with a few site-specific epigenetic
changes. Indeed, we have previously shown that several hundred
ariation in early life experience is associated with differences
in life-long health and behavioral trajectories in animals as
genes are differentially expressed in the hippocampi of adult rat
offspring that received low compared with high maternal licking
and grooming (LG) (7). Moreover, in the hippocampi of humans
with documented childhood abuse, we have recently discovered
methylation differences in the rRNA gene promoters that are
scattered across the genome (8). Furthermore, recent evidence
suggests that epigenetic regulation is not restricted to the few
thousand bases around the transcription start sites of genes.
Epigenetic changes associated with transcriptional changes can
appear within the body of a gene (9) or even at high frequency
across megabase-sized domains simultaneously deactivating
dozens of neighboring genes (10, 11). These results led us to hy-
pothesize that the epigenetic response to early life experience is
not limited to a single gene promoter but that NR3C1, along with
neighboring genes, might belong to a domain under coordinated
control. To test this hypothesis, we recently investigated DNA
methylation, H3K9 acetylation, and transcriptional profiles in a
region encompassing 6.5 million base pairs centered at NR3C1 in
the hippocampus of adult rat offspring of high and low LG (12).
We confirmed our hypothesis by identifying hundreds of robust
DNA methylation differences between the offspring of high and
low LG that were scattered across this large region.
Considering the parallel behavioral and epigenetic responsesin
humans and rats to early life environments described above, it is
reasonable to assume that at least part of the broad epigenetic
responses observed in rats to early life experiences may be evo-
lutionarily conserved in humans. Therefore, in this study we in-
vestigated the extent of this conservation in humans by generating
epigenetic profiles of the analogous region in humans, the 6.5
million base pair region centered at NR3C1 (heretofore referred
as the NR3C1 locus). Such a cross-species investigation is further
supported by the fact that there is an ∼35% sequence homology
between the rat and human NR3C1 loci, and that 80% of the
genes in the human region have orthologs in the rat region.
This paper results from the Arthur M. Sackler Colloquium of the National Academy of
Sciences, “Biological Embedding of Early Social Adversity: From Fruit Flies to Kindergart-
ners,” held December 9–10, 2011, at the Arnold and Mabel Beckman Center of the
National Academies of Sciences and Engineering in Irvine, CA. The complete program and
audio files of most presentations are available on the NAS Web site at www.nasonline.org/
Author contributions: M.J.M., G.T., and M. Szyf designed research; P.O.M., A.S., and T.C.T.H.
performed research; M. Suderman analyzed data; P.O.M. and A.S. analyzed the clinical di-
agnosis and the qPCR validation data; and M. Suderman, P.O.M., M.T.H., G.T., and M. Szyf
wrote the paper.
The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
Data deposition: The DNA microarray data reported in this paper have been deposited
into the Gene Expression Omnibus, www.ncbi.nlm.nih.gov/geo (accession no. GSE38352).
1M. Suderman, P.O.M., and A.S. contributed equally to this work.
2To whom correspondence should be addressed. E-mail: email@example.com.
This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.
| October 16, 2012
| vol. 109
| suppl. 2www.pnas.org/cgi/doi/10.1073/pnas.1121260109
Methylation Profiles in the NR3C1 Locus of Rat and Human
Hippocampi. We generated DNA methylation profiles in hippo-
campal samples obtained from the Quebec Suicide Brain Bank
of 12 suicide completers with a history of severe childhood abuse
and 12 nonabused controls. Profiles covered the genomic region
from 3.25 Mb upstream to 3.25 Mb downstream of the NR3C1
gene at 100-bp spacing and were created by using the method of
methylated DNA immunoprecipitation (meDIP) followed by
hybridization to a custom-designed Agilent 44K tiling micro-
array. Fig. 1 depicts the locus tiled with probes including the
locations of genes along with estimated methylation levels and
differences between the abuse group and controls. Previously
published rat methylation profiles were generated using identical
methods from the hippocampi of adult rat offspring of high and
low LG and covering the synteneic region from 3.25 Mb up-
stream to 3.25 Mb downstream of the NR3C1 gene at 100-bp
Conservation of the NR3C1 Locus Gene Architecture. Overall orga-
nization of the NR3C1 locus is significantly conserved, as shown by
the almost identical order of orthologous genes across the locus
(Fig.2).Fig.2,Top shows the positions ofgenes inhuman,and Fig.
2, Bottom shows their positions in rat. Fig. 2, Middle shows the rat–
human “hybrid” created by assigning orthologous genes to posi-
tions similar to their relative locations in the human and rat
genomes. Gray vertical lines in each panel coincide with tran-
scription start sites. Black lines between adjacent panels link
transcriptionstart sites of orthologousgenes inneighboring panels.
Expected DNA Methylation Patterns Confirmed. Methylation levels
were estimated from microarray meDIP profiles by deconvolut-
ing individual CpG methylation levels from the intensities of
nearby probes (13). Estimates of these levels across the locus in
human and rat are shown in Fig. 2 and are compared directly
in the middle panel showing the human–rat hybrid. Overall,
methylation levels seem to rise and fall in unison. Indeed, they
do have a small but statistically significant correlation (P <
0.0013; R = 0.048). Given the regulatory role of DNA methyl-
ation, these patterns are unlikely to be random. For example,
dips should correspond to active transcription start sites, CpG
islands, and the 3′ ends of genes (14, 15). As shown for the rat
profiles, we observed lower methylation levels around tran-
scription start sites (P ≤ 1.68 × 10−272, Wilcoxon rank sum test),
at the 3′ ends of genes (P ≤ 9.9 × 10−28), and inside CpG islands
(P ≤ 10−300). In contrast to rat methylation levels, human
methylation levels were lower near methylation-sensitive tran-
scription factor binding sinks (P ≤ 0.06). As the name suggests,
methylation-sensitive transcription factor binding sinks are
regions enriched for methylation-sensitive transcription factor
binding sites as predicted by binding motifs. The lack of meth-
ylation decrease in these regions in rats is likely due to the fact
that most transcription factor binding motifs have been derived
from human studies rather than rat studies.
Conservation of a Widespread Methylation Response. Both the rat
and human profiles revealed hundreds of differentially methyl-
ated regions (DMRs) associated with early life experience scat-
tered unevenly across the NR3C1 locus (Fig. 1). In total, there
were 281 human DMRs, of which 126 had increased methylation
in controls (cDMRs), and 155 had increased methylation in the
individuals with histories of childhood abuse (aDMRs). Real-
time PCR of meDIP samples was used to validate selected
DMRs. We investigated 11 of these differences inside gene
promoters located across the locus (Fig. 3). The rat profiles
revealed more than twice as many DMRs (723), of which 373
were more methylated in high-LG offspring (hDMRs) and 350
were more methylated in low-LG offspring (lDMRs). This larger
number in rat is possibly due to the greater genetic similarity and
less environmental variability leading to increased power to de-
tect differences within rat groups compared with human groups.
As observed in the rat profiles, the placement of the DMRs
across the locus is nonuniform, resulting in large regions
enriched with DMRs and others almost completely depleted of
any DMRs (Fig. 1). In the sections below we explore these
patterns in more detail.
Conservation of Long-Range Methylation Dependencies. In both
species, DMRs showing the same direction of change according
to environmental experience seem to form clusters covering large
genomic regions, supporting a high-level organization linking
distant sites. In general, there seem to be consistent dependencies
between methylation differences as far apart from NR3C1 as 1
million base pairs in both species (Fig. 4; figure 3a in ref. 12). As
Santa Cruz genome browser (human genome assembly hg18) show % 5meC: average methylation levels across all samples estimated from microarray probe
intensities; Δ 5meC: mean log2fold differences between abused and control sample probe intensities, where positive values are shown in black and indicate
higher methylation in abused samples, and gray values indicate higher methylation in control samples; cDMR: locations of cDMRs (significantly higher
methylation in control samples); and aDMR: locations of aDMRs (significantly higher methylation in abuse samples). The locations of the protocadherin
families of genes and NR3C1 are identified by shading.
Associations of human DNA methylation with early life abuse in the 6.5-Mb NR3C1 locus. Track images obtained from the University of California,
Suderman et al.PNAS
| October 16, 2012
| vol. 109
| suppl. 2
an example of this long-range clustering, observe in Fig. 1 that if
thehumanNR3C1locusis partitionedintotwo parts,onepartleft
and the other part right of NR3C1, aDMRs are enriched in the
left part (P ≤ 2.2 × 10−32; hypergeometric), and cDMRs are
enriched in the right part (P ≤ 2.2 × 10−32). Partitioning the rat
NR3C1 locus in the same way, hDMRs are enriched in the left
part (P ≤ 0.031), and lDMRs are enriched in the right part (P ≤
0.013). To determine whether any of the rat DMRs were con-
served in human, hDMR and lDMR sequences were mapped to
the human genome using BLAT. In total, 111 of these sequences
mapped successfully to the human locus; however, none of them
overlapped with human DMRs. Interestingly though, just as
hDMRs are enriched to the left and lDMRs are enriched to the
rightof NR3C1 inrat, themapped hDMRsare also enriched to the
5MeC diff: abused − control
5meC diff: high−low
6.5-Mb NR3C1 locus. Top: Human locus. Bottom: Rat locus. Middle: Human–rat “hybrid” panel created by assigning orthologous genes to positions similar to
their relative locations in the human and rat genomes. Each panel is divided into five labeled parts: genes: black horizontal arrows denote genes and the
direction of mRNA synthesis; variation: graph indicates regions of high and low methylation variation across all human subjects; % homology: graph shows
the percentage of bases in the human genome that were mapped by the lastz alignment tool to the rat genome; 5meC diff: graph shows mean log2fold
differences between sample groups (i.e., between abused and control humans and between high- and low-LG rats); and % 5meC: graph shows methylation
levels estimated from microarray probe intensities. Across each panel, gray vertical lines demark transcription start sites. Black lines between panels link the
positions of transcription start sites of orthologous genes. The lack of crossings between these lines illustrates conservation of gene architecture around
NR3C1 between rats and humans.
Associations of human and rat DNA methylation with early life abuse in the 6.5-Mb NR3C1 locus. Top, Middle, and Bottom: Each panel shows the
| www.pnas.org/cgi/doi/10.1073/pnas.1121260109Suderman et al.
left and the mapped lDMRs are enriched to the right of NR3C1 in
the human locus (Fig. S1). This suggests that change according to
environment is conserved across species at a high level, although
details about those changes differ between species.
Conservation of Enriched Methylation Response in Suspected
Regulatory Sites. Given the regulatory role that DNA methyla-
tion plays, one might expect to see DMRs near known or sus-
pected regulatory sites such as near transcription start sites,
particularly coinciding with CpG islands, and transcription factor
binding sites. Indeed, these regions tend to be enriched with
DMRs in both rats and humans, although some of the details
differ. Approximately 8% of DMRs in both rat and human in-
tersect promoter regions (−2,000...+200 bp of the transcription
start site; Table S1); however, whereas this intersection is sta-
tistically significant in humans (P < 0.001), it does not reach
significance in rats (P > 0.14). When analogous regions at the 3′
ends of genes are included, the overlap of both human and rat
DMRs is significant (P < 0.001 and P < 0.0032, respectively). In
humans, this enrichment extends to 1,000 bp past transcription
start sites (P < 0.001).
Interestingly, much of this enrichment in humans is explained
by aDMR enrichment (P < 2 × 10−4, promoters; P < 0.034, 3′
ends of genes; P < 2 × 10−4, 1,000 bp after transcription start
sites; P < 0.019, first exons) because cDMRs are depleted in
nearly all of these regions (P < 0.039, promoters; P < 0.031,
1,000 bp after transcription start sites; P < 0.02, first exons).
Supporting the regulatory nature of the sites targeted by aDMRs
is the observation that they are enriched for methylation-sensitive
transcription factor binding sinks (P < 0.08; Methods) and highly
enriched with CpG sites (P < 4.4 × 10−14). Not surprisingly,
cDMRS are depleted in these regions (P < 0.02) and depleted of
CpG sites compared with the rest of the locus (P < 1.4 × 10−8).
In rat, such a simple characterization of lDMRs and hDMRs is
not possible. lDMRs are enriched in some of these regions (P <
0.02, promoters; P < 0.003, 3′ ends of genes) but depleted in
others (P < 0.019, 1,000 bp after transcription start sites; P <
0.0011, first exons). In contrast, hDMRs are enriched primarily
in first exons (P < 0.0008) and, interestingly, also in last exons
(P < 0.0038). On the other hand, depletion of lDMRs (rat) and
cDMRs (human) are observed in last exons (P < 0.0063 and P <
Differential Methylation Across NR3C1. We have previously shown
that NR3C1 gene expression is lower in the abuse group and that
this decrease in expression associates with increased methylation
levels in the promoter of a splice variant (1F) of the NR3C1 gene
(4). The comprehensive mapping of the NR3C1 locus presented
here identified a total of seven DMRs in and around NR3C1: two
upstream cDMRs, four aDMRs within the first and second
introns, and one aDMR downstream of the gene (Fig. S2). The
increased number of aDMRs compared with cDMRs is consis-
tent with the repression of NR3C1 in the abuse group. In rats
there are similar DMRs throughout the gene, with the majority
being lDMRs (figure 4a in ref. 12), also consistent with the re-
pression of NR3C1 in the low-LG group.
Conserved Methylation Sensitivity in the Protocadherin Families of
Genes. Notable methylation differences in the NR3C1 locus of
both rats and humans are located downstream of NR3C1 within
the α-, β-, and γ-protocadherin (PCDH) gene clusters. All three
clusters together are highly enriched for aDMRs (P < 2 × 10−4)
entially methylated by microarray (Table S1) were subjected to real-time PCR quantification of enrichment. The y axis represents concentration values
generated by methylation-enriched and input DNA. Left: Concentration levels in the abuse group. Right: Concentration levels in the control group. Each real-
time PCR was performed in triplicate. Error bars indicate SEM.
Validation of microarray calls. Real-time PCR validation of microarray meDIP data is shown. Eleven of the 28 promoters identified as being differ-
Suderman et al. PNAS
| October 16, 2012
| vol. 109
| suppl. 2
and depleted of cDMRs (P < 0.0014). Of the three clusters,
α-PCDH is most enriched for DMRs (P < 0.054) and, particularly,
for aDMRs (P < 2 × 10−4). Fig. S3 depicts the methylation dif-
ferences within the PCDH gene clusters. Similarly to aDMRs,
lDMRs are highly enriched in the PCDH gene clusters (P < 0.01).
These methylation differences observed in human hippocam-
pus are of interest because the protocadherin families of genes
are known to be regulated by promoter methylation (16–18) and
have been implicated in synaptic function and neuronal con-
Regions That Lack Differential Methylation. The existence of DMR
clusters implies the existence of regions lacking methylation
differences. For example, despite the fact that gene promoters
are enriched with DMRs compared with other genomic regions
(P < 0.001), only 28 of the 171 gene promoters (−2000...+200
bp around the transcription start site of a gene) contain a DMR.
That leaves a lot of gene promoters unaffected by differential
methylation, despite the fact that DMRs are widely distributed
across the locus. For some reason, these promoters were “avoi-
ded.” In fact, there are eight regions of more than 100 Kb within
the NR3C1 locus that contain no differentially methylated sites
(Table S2). Permutation tests show that the expected number of
such regions is only 2.2, with a maximum of six found in 1,000
such tests (DMR positions were randomly permuted within the
locus). Three of the eight regions contained no genes, and one of
the eight regions contained at least 10 genes, more than three
times the number genes expected. Hence, these methylation
profiles identify large gene-rich and gene-poor regions without
any DMRs, evidence for a widespread but selective effect on
DNA methylation levels within the NR3C1 locus.
There is growing evidence for association between variation in
early life experience and differential methylation in several
genes. Past studies focused on documented or highly predicted
regulatory regions around the transcription start sites of these
genes. Such an approach might miss important DMRs and
ignore the larger scope of the DNA methylation response to
environmental cues. Our recent study of the epigenetic response
to maternal care in rats (12) determined that the epigenetic re-
sponse is in fact not limited to a few sites but affects broader
genomic regions. We asked here to what extent this broad re-
sponse might be conserved across species. Hence, we performed
a human study similar to our rat study covering the 6.5-Mb re-
gion centered at the NR3C1 gene, wherein we examined differ-
ences in DNA methylation in the hippocampi of subjects who
committed suicide and experienced severe abuse during child-
hood vs. control individuals with negative histories of abuse.
Similarly to our rat study, we showed that differential methyla-
tion is not restricted solely to NR3C1 promoters but instead
appears at many sites throughout the NR3C1 locus, both within
as well distant from promoters (Fig. 1).
Although there are many methylation differences, they are not
uniformly distributed across the locus, and our analysis describes
several levels of structural organization of this association with
early life experience showing surprising agreement with our
parallel rat analysis. The differences in DNA methylation tend to
concentrate in specific regions relative to transcription start sites
and in specific regions containing dozens of genes within the
NR3C1 locus (Fig. 1), suggesting high-level organization. Espe-
cially remarkable was the discovery of a division of the entire 6.5-
Mb locus into two major domains characterized by genomic sites
with reduced DNA methylation in the abuse group upstream to
the NR3C1 locus and genomic sites with increased in DNA
methylation downstream to the NR3C1 locus (Fig. 1).
A strikingly significant number of DMRs can be found in the
promoters of the protocadherin families of genes (Fig. 1) in both
humans and rats, supporting the hypothesis that protocadherins
play a key role in the response to early life experience. This
hypothesis is consistent with previous findings that the complex
expression patterns of the protocadherins are regulated by DNA
methylation leading to differential promoter activation and al-
ternative pre-mRNA splicing (16–18). That our methylation
differences were observed in the hippocampus and that proto-
cadherins have been implicated in synaptic function and neuro-
nal connectivity (19–22) suggests that regional DNA methylation
in response to early life experience.
The potential regulatory roles of the methylation differences
thatdo notmap toregulatory sites, suchastranscription startsites
and methylation-sensitive transcription factor binding sites, are
more difficult to characterize. However, the conserved enrich-
ment of methylation differences around these regulatory sites in
humans and rats supports the existence of a regulatory role for
other methylation differences yet to be elucidated.
Not surprisingly, given the important differences between rats
and humans and the nature of their early life environments,
comparison between the rat and human methylation changes
associated with early life experience was neither simple nor
straightforward. For example, the methylation profiles do not
support a direct analogy at the individual base level between low
maternal care in rats and childhood abuse in humans. However,
such an analogy was not the purpose of our study. Instead, we
reasoned that a cross-species comparison of DNA methylation
associated with variation in early life environment would identify
genomic regions beyond the promoters regions of NR3C1 that
are epigenetically labile in response to a range of early life
experiences. Our results support this hypothesis. In both rats and
humans, we identified a broad but selective response to early life
experience that is enriched in suspected regulatory regions,
exhibits evidence of a long-range coordination between distant
sites, and seems to particularly target the regulation of the
protocadherin families of genes, suggesting that these genes
may also be involved in the response to early life experience.
Such a conserved response motivates the development of novel
abuse in the 6.5-Mb NR3C1 locus. Pearson correlations of DNA methylation
differences between the subject groups at various genomic distances. Error
bars show 95% confidence intervals for the correlation values. The gray
highlight shows the expected 95% confidence interval if there is no corre-
lation between methylation differences at different genomic sites. This
confidence interval does not overlap with the error bars associated with
distances lessthan 1Mb, suggesting the existence ofsystematic dependencies
between methylation differences at distances up to 1 Mb.
Correlation of human DNA methylation associations with early life
| www.pnas.org/cgi/doi/10.1073/pnas.1121260109Suderman et al.
experimental approaches to understand how these DNA meth-
ylation modulations affect genome function.
Methods related to the human samples only are provided here because
methods and analyses related to the rat samples have already been pub-
Ethics Statement. Studies with human subjects were approved by the McGill
University institutional review board, and signed informed consent was
obtained from next of kin. All procedures involving rodents were performed
the protocol was approved by the McGill University Animal Care Committee.
Subjects and Tissue Preparation. Hippocampal samples obtained from the
Quebec Suicide Brain Bank included 12 suicide subjects with histories of
severe childhood abuse and 12 controls with validated negative histories of
childhood abuse who did not differ in postmortem interval, sex, age at death,
and brain pH (all P > 0.05). Psychiatric diagnoses were obtained by means of
the Structured Clinical Interview for DSM-III-R (23) interview adapted for
psychological autopsies, which is a validated method to reconstruct psychi-
atric and developmental history by means of extensive proxy-based inter-
views, as outlined elsewhere (24). To be considered in this study, all suicide
subjects had to have a positive history of severe childhood sexual and/or
physical abuse or severe neglect, as determined by most severe scores in the
respective scales of the structured Childhood Experience of Care and Abuse
(25) questionnaire adapted for psychological autopsies (26). Conversely,
controls had to have validated evidence of negative lifetime histories of
abuse and/or neglect.
All samples were from male suicide and control subjects of French-Ca-
nadian origin. Samples were dissected at 4 °C and stored in plastic vials at
−80 °C until analysis. All samples were processed and analyzed blind to
demographic and diagnostic variables. To be included in this study, all
subjects had to die suddenly, with no medical or paramedic intervention and
no prolonged agonal period. Suicide as the cause of death was determined
by the Quebec Coroner’s Office.
DNA Immunoprecipitation and Microarray Hybridization. The procedure for
methylated DNA immunoprecipitation was adapted from previously pub-
lished work (27–29). The amplification (Whole Genome Amplification kit;
Sigma) and labeling reaction (CGH labeling kit; Invitrogen), and all of the
steps of hybridization including washing and scanning were performed
according to the Agilent protocol for chip-on-chip analysis. Microarrays were
hybridized in triplicate for each sample.
Quantitative Real-Time PCR of Immunoprecipitated Samples. Gene-specific
real-time PCR validation of microarray was performed for DNA methylation
enrichment (30) for the same samples used for microarray experiments.
Triplicate reactions were performed, and relative concentration was de-
termined as a ratio of the crossing point threshold (Ct). The average con-
centration for each set of replicates was plotted along with its SEM. Primers
for each amplicon are given in Table S3.
Microarray Design and Analysis. Custom 44K tiling arrays were designed using
eArray (Agilent). Probes of ∼55 bp were selected to tile all unique regions
within ∼3.25 MB upstream and downstream of the NR3C1 gene described in
Ensembl (version 44) at 100-bp spacing. Probe intensities were extracted
from microarray scan images using Agilent’s Feature Extraction 9.5.3 Image
Analysis Software and analyzed using the R software environment for sta-
tistical computing (31). Background corrected log-ratios of the bound (Cy5)
and input (Cy3) microarray channel intensities were computed for each
microarray. Microarrays were normalized to one another using quantile
All genomic coordinates are given with respect to the hg18 human
In some cases, DNA methylation levels at genomic locations were estimated
from microarray probe intensities. In these cases, a Bayesian convolution al-
gorithm was used to incorporate probe values from nearby probes (13).
Differential methylation between groups was determined in two stages to
ensure both statistical significance and biological relevance. In the first stage,
linear models implemented in the “limma” package (33) of Bioconductor
(34) were used to compute a modified t statistic at the individual probe level.
An individual probe was called differentially methylated if the significance
of its t statistic was at most 0.05 (uncorrected for multiple testing) and the
associated difference of log-normalized means between the groups was at
least 0.5. Given that the DNA samples were sonicated into 200- to 700-bp
fragments before hybridization, we assumed that probes within 500 bp
should have approximately similar probe scores. Therefore, in the second
stage, we computed differential statistics for 1,000-bp intervals from the
differential statistics of the probes that they contained. The intervals tiled
the entire 6.5-Mb region under investigation at 500-bp spacing. Differential
significance of these intervals was determined using the Wilcoxon rank-sum
test comparing t statistics of the probes within the interval against those of
all of the probes on the microarray. Significance levels were then adjusted to
obtain false discovery rates. An interval was called differentially methylated
if it satisfied each of the following: (i) its false discovery rate was at most 0.2,
and (ii) the 1,000-bp interval contained at least one probe called differen-
tially methylated. The first requirement ensured that several probes in the
interval had similar group differences, and the second requirement ensured
that the difference was not simply weakly distributed across the entire in-
terval and consequently difficult to validate. Intervals satisfying these tests
were called differentially methylated regions (DMRs). Consecutive DMRs for
which the difference of means showed greater methylation in the abused
group were called aDMRs and the converse cDMRs. Consecutive a/cDMRs
were coalesced into single a/cDMRs.
Statistically significant enrichment or depletion of DMRs in specific regions
such as CpG islands orgene promoters was computed using permutation tests
on the locations of DMRs. More specifically, the statistic used the number of
base pairs overlapping between DMRs and the regions in question, for ex-
ample CpG islands. A distribution for this statistic was computed by re-
peatedly (1,000 times) randomly assigning theoretically possible coordinates
(based on the locations of probes) to the DMRs and then calculating the
overlap between the regions and the newly located DMRs.
Methylation-sensitive transcription factor sinks were computed by posi-
tion weight matrices for specific transcription factors from the Transfac (35)
and Jaspar (36) databases, including AP2, CBF, CREB, ETS, FOXP3, GABP,
GATA1, NF-kappaB, NGFIA/EGR1, NR3C1, P53, RUNX, SP1, SP3, TCF, and
USF1. There is evidence that the activity of each of these transcription fac-
tors is affected by the presence or absence of DNA methylation (3, 37–48).
For some of these transcription factors, the databases contained multiple
identical position weight matrices. To avoid having these matrices bias the
identification of transcription factor sinks in favor of a single transcription
factor, we removed one position weight matrix for any pair whose targets
overlapped 75% of the time. Transcription factor targets were identified by
scanning the sequence with a second-order position weight matrix adjusting
log-likelihoods with the 500-bp sequence background context (49). Sites
with a log-likelihood greater than 14 were called binding sites. To make the
remaining computation to identify transcription factor sinks more efficient,
the genome was then partitioned into 100-bp segments, and the binding
score for each transcription factor in each segment was set to the maximum
log-likelihood in that segment. Each segment was called a transcription
factor sink if its transcription factor scores were significantly higher than
average as determine by the Wilcoxon rank sum test (P ≤ 1 × 10−9or 6 × 10−5
after Bonferroni correction).
The variability of a probe was quantified as the number of sample pairs for
which all normalized replicate log-ratios for one sample were at least 0.5
greater than all normalized replicate log-ratios for the other sample.
Overall homology between human and rat were computed using the lastz
program (50). Specifically, the lastz program was used with default settings
to align the human and rat NR3C1 locus sequences. The percentage of ho-
mology was given as the percentage of human sequence that was success-
fully mapped to the rat sequence.
Rat DMR sequences, all 1,000 bp long, were mapped onto the human NR3C1
using BLAT (51) with the following settings: tile_size = 10, step_size = 10,
min_match = 2, min_score = 400, min_identity = 75, and max_intron = 250.
Fig. 4 illustrates the correlation of methylation differences across various
genomic distances as Pearson correlations of modified t statistics computed
by limma for all pairs of probes at specified distances (with a 10% tolerance).
Error bars denote 95% confidence intervals obtained from 1,000 bootstraps
composed of randomly selected probe pairs with replacement. The gray
rectangle denotes the 95% confidence interval for correlations of probe
pairs independent of their distance. Independence was simulated by with
500 random permutations of the probe coordinates.
All microarray data are MIAME compliant, and the raw data have been
deposited in the Gene Expression Omnibus.
ACKNOWLEDGMENTS. This study was supported by grants from the
Canadian Institutes of Mental Health and the Sackler Foundation (to M.J.M.
Suderman et al. PNAS
| October 16, 2012
| vol. 109
| suppl. 2
1. Francis D, Diorio J, Liu D, Meaney MJ (1999) Nongenomic transmission across gen- Download full-text
erations of maternal behavior and stress responses in the rat. Science 286:1155–1158.
2. Gilbert R, et al. (2009) Burden and consequences of child maltreatment in high-in-
come countries. Lancet 373:68–81.
3. Weaver IC, et al. (2004) Epigenetic programming by maternal behavior. Nat Neurosci
4. McGowan PO, et al. (2009) Epigenetic regulation of the glucocorticoid receptor in
human brain associates with childhood abuse. Nat Neurosci 12:342–348.
5. Roth TL, Lubin FD, Funk AJ, Sweatt JD (2009) Lasting epigenetic influence of early-life
adversity on the BDNF gene. Biol Psychiatry 65:760–769.
6. Murgatroyd C, et al. (2009) Dynamic DNA methylation programs persistent adverse
effects of early-life stress. Nat Neurosci 12:1559–1566.
7. Weaver IC, Meaney MJ, Szyf M (2006) Maternal care effects on the hippocampal
transcriptome and anxiety-mediated behaviors in the offspring that are reversible in
adulthood. Proc Natl Acad Sci USA 103:3480–3485.
8. McGowan PO, et al. (2008) Promoter-wide hypermethylation of the ribosomal RNA
gene promoter in the suicide brain. PLoS ONE 3:e2085.
9. Hellman A, Chess A (2007) Gene body-specific methylation on the active X chromo-
some. Science 315:1141–1143.
10. Coolen MW, et al. (2010) Consolidation of the cancer genome into domains of re-
pressive chromatin by long-range epigenetic silencing (LRES) reduces transcriptional
plasticity. Nat Cell Biol 12:235–246.
11. Frigola J, et al. (2006) Epigenetic remodeling in colorectal cancer results in coordinate
gene suppression across an entire chromosome band. Nat Genet 38:540–549.
12. McGowan PO, et al. (2011) Broad epigenetic signature of maternal care in the brain
of adult rats. PLoS ONE 6:e14739.
13. Down TA, et al. (2008) A Bayesian deconvolution strategy for immunoprecipitation-
based DNA methylome analysis. Nat Biotechnol 26:779–785.
14. Lister R, et al. (2009) Human DNA methylomes at base resolution show widespread
epigenomic differences. Nature 462:315–322.
15. Ball MP, et al. (2009) Targeted and genome-scale strategies reveal gene-body
methylation signatures in human cells. Nat Biotechnol 27:361–368.
16. Tasic B, et al. (2002) Promoter choice determines splice site selection in protocadherin
alpha and gamma pre-mRNA splicing. Mol Cell 10:21–33.
17. Kawaguchi M, et al. (2008) Relationship between DNA methylation states and tran-
scription of individual isoforms encoded by the protocadherin-alpha gene cluster.
J Biol Chem 283:12064–12075.
18. Yagi T (2008) Clustered protocadherin family. Dev Growth Differ 50(Suppl 1):
19. Bass T, Ebert M, Hammerschmidt M, Frank M (2007) Differential expression of four
protocadherin alpha and gamma clusters in the developing and adult zebrafish:
DrPcdh2gamma but not DrPcdh1gamma is expressed in neuronal precursor cells,
ependymal cells and non-neural epithelia. Dev Genes Evol 217:337–351.
20. Junghans D, Haas IG, Kemler R (2005) Mammalian cadherins and protocadherins:
About cell death, synapses and processing. Curr Opin Cell Biol 17:446–452.
21. Suzuki ST (2000) Recent progress in protocadherin research. Exp Cell Res 261:13–18.
22. Weiner JA, Wang X, Tapia JC, Sanes JR (2005) Gamma protocadherins are required for
synaptic development in the spinal cord. Proc Natl Acad Sci USA 102:8–14.
23. Spitzer RL, Williams JB, Gibbon M, First MB (1992) The Structured Clinical Interview
for DSM-III-R (SCID). I: History, rationale, and description. Arch Gen Psychiatry 49:
24. Dumais A, et al. (2005) Risk factors for suicide completion in major depression: A case-
control study of impulsive and aggressive behaviors in men. Am J Psychiatry 162:
25. Bifulco A, Brown GW, Harris TO (1994) Childhood Experience of Care and Abuse
(CECA): A retrospective interview measure. J Child Psychol Psychiatry 35:1419–1435.
26. Zouk H, Tousignant M, Seguin M, Lesage A, Turecki G (2006) Characterization of
impulsivity in suicide completers: Clinical, behavioral and psychosocial dimensions.
J Affect Disord 92:195–204.
27. Weber M, et al. (2005) Chromosome-wide and promoter-specific analyses identify
sites of differential DNA methylation in normal and transformed human cells. Nat
28. Keshet I, et al. (2006) Evidence for an instructive mechanism of de novo methylation
in cancer cells. Nat Genet 38:149–153.
29. Brown SE, Szyf M (2008) Dynamic epigenetic states of ribosomal RNA promoters
during the cell cycle. Cell Cycle 7:382–390.
30. Sadikovic B, et al. (2008) In vitro analysis of integrated global high-resolution DNA
methylation profiling with genomic imbalance and gene expression in osteosarcoma.
PLoS ONE 3:e2834.
31. R Development Core Team (2007) R: A Language and Environment for Statistical
Computing (R Foundation for Statistical Computing, Vienna).
32. Bolstad BM, Irizarry RA, Astrand M, Speed TP (2003) A comparison of normalization
methods for high density oligonucleotide array data based on variance and bias.
33. Smyth GK (2005) Limma: Linear models for microarray data. Bioinformatics and
Computational Biology Solutions Using R and Bioconductor, eds Gentleman R,
Carey V, Dudoit S, Irizarry R, Huber W (Springer, New York), Vol 1, pp 397–420.
34. Gentleman RC, et al. (2004) Bioconductor: Open software development for compu-
tational biology and bioinformatics. Genome Biol 5:R80.
35. Matys V, et al. (2006) TRANSFAC and its module TRANSCompel: Transcriptional gene
regulation in eukaryotes. Nucleic Acids Res 34(Database issue):D108–D110.
36. Bryne JC, et al. (2008) JASPAR, the open access database of transcription factor-
binding profiles: New content and tools in the 2008 update. Nucleic Acids Res 36
37. McPherson LA, Baichwal VR, Weigel RJ (1997) Identification of ERF-1 as a member of
the AP2 transcription factor family. Proc Natl Acad Sci USA 94:4342–4347.
38. McPherson LA, Weigel RJ (1999) AP2alpha and AP2gamma: A comparison of binding
site specificity and trans-activation of the estrogen receptor promoter and single site
promoter constructs. Nucleic Acids Res 27:4040–4049.
39. Zheng Y, et al. (2010) Role of conserved non-coding DNA elements in the Foxp3 gene
in regulatory T-cell fate. Nature 463:808–812.
40. Zhang X, et al. (2005) Genome-wide analysis of cAMP-response element binding
protein occupancy, phosphorylation, and target gene activation in human tissues.
Proc Natl Acad Sci USA 102:4459–4464.
41. Polansky JK, et al. (2010) Methylation matters: Binding of Ets-1 to the demethylated
Foxp3 gene contributes to the stabilization of Foxp3 expression in regulatory T cells.
J Mol Med (Berl) 88:1029–1040.
42. Lucas ME, Crider KS, Powell DR, Kapoor-Vazirani P, Vertino PM (2009) Methylation-
sensitive regulation of TMS1/ASC by the Ets factor, GA-binding protein-alpha. J Biol
43. Hutchins AS, et al. (2002) Gene silencing quantitatively controls the function of
a developmental trans-activator. Mol Cell 10:81–91.
44. Siegfried Z, Cedar H (1997) DNA methylation: A molecular lock. Curr Biol 7:
45. Nabilsi NH, Broaddus RR, Loose DS (2009) DNA methylation inhibits p53-mediated
survivin repression. Oncogene 28:2046–2050.
46. Zhu WG, et al. (2003) Methylation of adjacent CpG sites affects Sp1/Sp3 binding and
activity in the p21(Cip1) promoter. Mol Cell Biol 23:4056–4065.
47. Aoki M, et al. (2008) Kidney-specific expression of human organic cation transporter
2 (OCT2/SLC22A2) is regulated by DNA methylation. Am J Physiol Renal Physiol 295:
48. Choy MK, et al. (2010) Genome-wide conserved consensus transcription factor bind-
ing motifs are hyper-methylated. BMC Genomics 11:519.
49. Stormo GD (2000) DNA binding sites: Representation and discovery. Bioinformatics
50. Harris RS (2007) Improved pairwise aligment of genomic DNA. PhD dissertation
(Pennsylvania State Univ, University Park, PA) (Pennsylvania State Univ, University
51. Kent WJ (2002) BLAT—the BLAST-like alignment tool. Genome Res 12:656–664.
| www.pnas.org/cgi/doi/10.1073/pnas.1121260109 Suderman et al.