Chronic social stress induces DNA methylation changes at an evolutionary conserved intergenic region in chromosome X

ABSTRACT

Chronic stress resulting from prolonged exposure to negative life events increases the risk of mood and anxiety disorders. Although chronic stress can change gene expression relevant for behavior, molecular regulators of this change have not been fully determined. One process that could play a role is DNA methylation, an epigenetic process whereby a methyl group is added onto nucleotides, predominantly cytosine in the CpG context, and which can be induced by chronic stress. It is unknown to what extent chronic social defeat, a model of human social stress, influences DNA methylation patterns across the genome. Our study addressed this question by using a targeted-capture approach called Methyl-Seq to investigate DNA methylation patterns of the dentate gyrus at putative regulatory regions across the mouse genome from mice exposed to 14 days of social defeat. Findings were replicated in independent cohorts by bisulfite-pyrosequencing. Two differentially methylated regions (DMRs) were identified. One DMR was located at intron 9 of Drosha, and it showed reduced methylation in stressed mice. This observation replicated in one of two independent cohorts. A second DMR was identified at an intergenic region of chromosome X, and methylation in this region was increased in stressed mice. This methylation difference replicated in two independent cohorts and in Major Depressive Disorder (MDD) postmortem brains. These results highlight a region not previously known to be differentially methylated by chronic social defeat stress and which may be involved in MDD.

Introduction

Chronic stress results from prolonged exposure to negative life events and increases the risk of psychiatric illnesses such as mood and anxiety disorders. This has been highlighted by previous studies showing that prior exposure to stress [1,2], including chronic stress, increases risk for depression [3]. Studies have also shown that the chronicity of stress increases the likelihood of post-traumatic stress syndrome [4,5]. In rodent models of chronic stress, anxiety-like behavior has been observed, including reduced time in the open arms of an elevated plus maze [6], and depression-like behavior has similarly been observed, such as reduced time to immobility in the forced swim test [7]. Due to its pernicious impact on mental health, much effort has been directed towards dissecting the cellular and molecular impact of negative stressors on the brain.

Stressors are known to alter the expression of genes and cellular processes that influence behavior. For example, the hypothalamus-pituitary-adrenal (HPA) axis constitutes a set of neuroendocrine structures that regulate adaptive response to stress [8]. Rodents exposed to stressors have altered expression of genes along the HPA axis, such as those coding for corticotrophin releasing factor (CRF), vasopressin, and adrenocorticotropic hormone, and these alterations impair its negative feedback response to stress [9–11]. Similarly, the expression of Bdnf, a gene that regulates neuroplasticity and neurogenesis, is reduced by chronic stress, resulting in depression-like behavior in rodents [12,13]. Although the predominant mechanisms that mediate stress-induced changes to gene expression are still unclear, one molecular process that could contribute to this change is DNA methylation.

DNA methylation is an epigenetic process whereby a methyl group is added onto nucleotides without changes to the DNA sequence. In mammals, DNA methylation occurs on cytosines, predominantly in the CpG context, and has been shown to regulate cellular phenotypes through differential methylation patterns at transcriptional regulatory elements resulting in tissue-specific changes to gene expression [14,15]. Studies have also observed methylation outside of the CpG context, but to a lesser degree than CpG methylation [16]. Importantly, DNA methylation patterns can be altered by stress leading to changes in the expression of genes that regulate behavior. For example, studies using early life stress and chronic social defeat stress models in mice have observed decreased methylation at the promoter and regulatory regions of the vasopressin and CRF genes in the paraventricular nucleus of the hypothalamus, resulting in increased expression of those genes [9,11]. These changes were associated with anxiety and depressive-like behavior in the animals. Since the chronic social defeat paradigm closely models the human experience of social stress, and it is currently unknown to what extent chronic social stress influences genome-wide DNA methylation, in this study we sought to investigate this.

For this investigation, we had previously optimized a targeted-capture approach whereby promoters and other putative regulatory elements such as enhancers were isolated using a custom hybridization array followed by bisulfite conversion and next-generation sequencing [14]. Using this approach, we assayed DNA methylation to determine the contribution of chronic social defeat to changes in methylation patterns across these regions of the mouse genome.

Results

Anxiety and depression-like phenotypes in chronic social defeat stress mice

Mice were tested in behavioral tests following 14-days social defeat. Social defeat stress resulted in greater anxiety-like behavior. In the open field test, Stress group spent more time immobile (P < 0.05) (Figure 1(a)) and less time exploring the inner zone of the open field (P < 0.05) (Figure 1(b)) compared to Control group. In the elevated plus maze, the Stress group spent more time in the closed arm compared to the Control group (Figure 1(c)) and less time in the center zone (P < 0.05) (Figure 1(d)). Time spent in the open arm was not statistically significant (P = 0.36) (Figure 1(e)). Social stress did not alter the total volume of fluid consumed (data not shown), but resulted in lower preference for 1% sucrose solution in the sucrose preference test (Figure 1(f)) suggestive of anhedonia- and depression-like behavior. Stress mice showed adrenal (Figure 1(g)) and splenic hypertrophy (Figure 1(h)) and thymic involution following stress (Figure 1(i)). Basal plasma corticosterone levels were significantly increased in Stress compared to Control mice (P < 0.05) (Figure 1(j)). Together, our experiments confirm that chronic social defeat had significant effects on behavioral and physiological measures in our mice.

Figure 1.

Behavioral phenotypes, organ weight and corticosterone level of mice following 14-day social defeat stress (N = 16) compared to control mice (n = 16). (a–b) Open field test. (c–e) Elevated plus maze. (f) Sucrose preference test. (g–i) Organ weights. (a) Time spent immobile. (b) Time spent in inner zone. (c) Time spent in closed arm. (d) Time spent in center. (e) Time spent in open arm. (f) Consumption of sucrose solution. (g) Adrenal gland hypertrophy. (h) Splenic hypertrophy. (i) Thymic atrophy. (j) Basal corticosterone level. N.S.: not significant; *: P < 0.05; **: P < 0.0001. Figure shows mean ± S.E.M.

Single base resolution maps of 5-methylcytosine generated by Methyl-Seq from the dentate gyrus of chronic social defeat and non-stress control mice

Single base-pair resolution maps of 5-methylcytosine were generated for seven chronic social defeat and eight non-stress control mice. DNA extracted from the dentate gyrus of these animals was bisulfite-sequenced at targeted regions containing RefSeq genes and other putative regulatory elements, as previously described [14]. This generated an average of 68 million reads that were filtered for duplicates leaving an average of ~50 million reads for downstream analysis. From this, 91% of bases (~4 Gb) were within targeted reads. About 74% of targeted bases had at least 10X coverage. The average depth of coverage across all samples was ~40×. An overview of these results is described in Supplementary Table S1.

Cytosines in the CpG context that had a minimum of 10X coverage and showed good base quality (≥Q30) were extracted for differential methylation analysis. About 1.2 million CpGs satisfied these criteria. Although numerous differences in DNA methylation levels between stressed and unstressed mice were observed, only one CpG was observed to be differentially methylated at a genome-wide level of statistical significance (Supplementary Fig. S1). The magnitude of change between stressed and unstressed mice for this CpG was ~8%. To compensate for technical variations which may arise from bisulfite-sequencing experiments, we chose to employ a conservative threshold of at least a 20% methylation difference between stress and control groups, along with a nominal P value cut-off of ≤0.05 in order to establish a list of CpGs we considered to be of interest. This is a similar threshold that has been used in other studies [14,17]. Using these criteria, 4,896 DMCs were identified. As previous studies have observed robust validation of DMCs clustered in DMRs and have shown functional changes corresponding with DMRs [18,19], we ranked the DMRs by the number of DMCs clustered in the region and created a list of top hits (i.e., DMRs with the greatest number of DMCs) for replication in two separate cohorts (Table 1).

Table 1.

List of differentially methylated regions (DMRs) and corresponding differentially methylated cytosines (DMCs) from Methyl-Seq analysis. Table is sorted by descending number of DMCs in DMRs.

Chromosome	Start	End	Number of DMCs	Number of CpGs	DMC density	P-value (DMR)	FDR (DMR)	Gene Symbol	Dist. from gene (bps)	Base	% Meth. Diff [Stress] vs [Control]	P-value (DMC)	FDR (DMC)
chrX	11,577,185	11,577,543	5	15	0.33	5.38 E-04	1.06E-02		> 10,000	11,577,253	22	4.30E-02	9.58E-01
										11,577,264	29	9.84E-03	9.23E-01
										11,577,298	29	1.22E-02	9.32E-01
										11,577,300	29	1.71E-02	9.46E-01
										11,577,326	48	2.79E-03	8.67E-01
chr4	129,642,504	129,642,622	4	7	0.57	3.82E-04	1.06E-02	1700003M07Rik	1,745	129,642,603	29	4.21E-02	9.58E-01
										129,642,607	30	4.85E-03	9.03E-01
										129,642,611	32	3.00E-02	9.58E-01
										129,642,613	30	3.70E-03	8.83E-01
chr12	87,039,073	87,039,638	4	38	0.11	3.94E-05	3.62E-03	Batf	0	87,039,130	−29	5.82E-04	7.67E-01
										87,039,142	−24	4.19E-02	9.58E-01
										87,039,146	−25	4.46E-03	8.98E-01
										87,039,187	−32	1.15E-03	8.29E-01
chr5	124,329,590	124,330,091	3	19	0.16	5.43E-04	1.06E-02	Gpr81	0	124,329,900	21	2.51E-02	9.57E-01
										124,329,908	22	2.04E-03	8.50E-01
										124,329,931	25	2.00E-02	9.51E-01
chr7	17,471,014	17,471,612	3	96	0.03	6.95E-04	1.06E-02	Dact3	0	174,715,57	−27	3.30E-02	9.58E-01
										17,471,563	−28	3.33E-03	8.77E-01
										17,471,566	−25	2.29E-02	9.54E-01
chr11	119,873,218	119,873,658	3	33	0.09	1.69E-03	1.64E-02	Aatk	0	119,873,507	23	3.09E-02	9.58E-01
										119,873,516	23	3.66E-02	9.58E-01
										119,873,542	21	2.49E-03	8.65E-01
chr14	69,867,139	69,867,617	3	16	0.19	5.35E-03	2.56E-02	Slc25a37	0	69,867,321	−26	2.62E-02	9.58E-01
										69,867,323	−26	5.67E-03	9.07E-01
										69,867,325	−22	9.18E-03	9.23E-01
chr1	34,489,113	34,489,351	2	6	0.33	2.08E-02	4.03E-02	Ccdc115	4,164	34,489,288	23	7.70E-03	9.18E-01
										34,489,290	28	1.28E-02	9.34E-01
chr4	140,545,241	140,545,479	2	7	0.29	4.48E-03	2.45E-02	Atp13a2	0	140,545,365	−23	2.16E-03	8.51E-01
										140,545,395	28	7.88E-03	9.18E-01
chr5	116,096,119	116,096,477	2	9	0.22	2.59E-02	4.33E-02	Rab35	0	116,096,155	−32	3.91E-03	8.85E-01
										116,096,191	−23	9.54E-03	9.23E-01
chr10	79,512,931	79,513,169	2	3	0.67	2.54E-02	4.33E-02	Gpx4	3,086	79,513,000	−44	1.14E-02	9.25E-01
										79,513,147	−33	3.85E-02	9.58E-01
chr11	106,530,978	106,531,456	2	10	0.2	1.88E-02	3.85E-02	Pecam1	0	106,531,439	30	4.49E-02	9.58E-01
										106,531,445	34	7.11E-03	9.14E-01
chr15	12,778,006	12,778,364	2	7	0.29	1.04E-03	1.37E-02	Drosha	0	12,778,019	−23	1.77E-02	9.46E-01
										12,778,022	−21	3.89E-03	8.85E-01

Bisulfite-pyrosequencing in replication cohorts of social defeat mice

Two DMRs were successfully replicated in separate cohorts of social defeat mice. One of the DMRs is located in an intergenic region on chromosome X, and the original experiment showed increased methylation of this DMR in stressed mice (Table 1). Bisulfite-pyrosequencing was performed on two separate replication cohorts at DMCs observed by Methyl-Seq and three additional CpGs proximal to those DMCs, i.e., CpG5, CpG7, and CpG8 (Figure 2(a–b)). In cohort 2, CpG2, CpG5, CpG7, and CpG8 were observed to be significantly more highly methylated in stressed mice compared to unstressed mice (Figure 2(a)). Although the other DMCs were not statistically significant, they were more highly methylated in stressed animals (Figure 2(a–b)) as similarly observed by Methyl-Seq (Table 1). In cohort 3, while CpG2 and CpG5 remained significantly highly methylated in stressed mice compared to control mice, methylation of CpG1, CpG2, and CpG4-6 were also observed to be significantly elevated in stressed mice (Figure 2(b)). Consistent with the original experiment, this DMR was elevated in DNAm in stressed mice compared to control mice in cohort 2 and cohort 3 (Figure 3(a–b)).

Figure 2.

Bisulfite-pyrosequencing of DMCs in the intergenic region on Chromosome X and intron 9 Drosha. Bisulfite-pyrosequencing of intergenic region on chromosome X in (a) cohort 2 (n = 7 stressed and 12 control mice) and (b) cohort 3 (n = 15 stressed and 15 control mice). Bisulfite-pyrosequencing in intron 9 of Drosha for (c) cohort 2 and (d) cohort 3. Figures shown are mean ± S.E.M. Data was analyzed using one-way ANOVA and corrected for multiple testing using Holm-Šídák test. ***: P < 0.001; *: P < 0.05; N.S.: not significant.

Figure 3.

Methylation differences of DMRs at intergenic region of chromosome X and intron 9 of Drosha. Methylation of CpGs detected by bisulfite-pyrosequencing were averaged across the region and compared between stressed and control mice. DMR at the intergenic region of chromosome X in replication (a) cohort 2 and (b) cohort 3. DMR at intron 9 of Drosha in replication (c) cohort 2 and (d) cohort 3. Figures shown are mean ± S.E.M of the DMR for each group. #: P = 0.052; *: P < 0.05; **: P < 0.01; N.S; not significant. P values were calculated using student’s t-test.

The second region that showed some evidence of replication was a DMR in intron 9 of Drosha. We examined DMCs there, along with the addition of two neighboring CpGs at that site, i.e., CpG3 and CpG4, in the same two replication cohorts as mentioned above (Figure 2(c–d)). In cohort 2, CpG2 and CpG4 were observed to have significantly lower methylation levels in stressed mice than in unstressed ones (Figure 2(c)), as was seen in the Methyl-Seq data (Table 1). Although CpG1 and CpG3 did not show statistically significant differences between stressed and non-stressed mice in cohort 2, they did show the same trend towards lower methylation in the stressed animals. In cohort 3, however, none of the CpGs replicated with statistically significant lower methylation in stressed animals (Figure 2(d)). Overall, the DMR in intron 9 of Drosha was reduced in cohort 2 as observed in Methyl-Seq experiment, but this was not replicated in cohort 3 (Figure 3(c–d)).

The 11 other DMRs listed in Table 1 were tested for replication in cohort 2. None of them showed statistically significant levels of concordance with the Methyl-Seq results. As such, they were not further investigated in cohort 3.

Determining if expression of neighboring genes are altered by differential methylation at chromosome X

Since CpGs in the DMR of chromosome X were consistently differentially methylated across three independent cohorts of chronic social defeated mice, we explored whether this DMR influenced the expression of nearby genes. To accomplish this, we examined a fourth cohort of chronic social defeated mice. This DMR is distal from neighboring genes such as Bcor, ATP6ap2, and Med14 and is located ~83 kb, ~591 kb and ~765 kb, respectively, from their transcriptional start sites. It also contains hallmarks of an active enhancer, as enrichment has been observed for histone 3 lysine 4 monomethylation (H3K4me1) and histone 3 lysine 27 acetylation (H3K27ac) in mouse brain, marks associated with enhancer functions (Supplementary Fig. S2). Since distal enhancers have been observed to regulate the expression of genes as far away as 1–1.5 Mb [20,21], the expression of Bcor, ATP6ap2, and Med14 genes were investigated to determine if their expression changes with the methylation status of the DMR. However, quantitative PCR did not show significant change between chronic social defeated and unstressed mice for those genes (Supplementary Fig. S3A-C). Similarly, no significant difference in Drosha expression was observed between the two groups of mice (Supplementary Fig. S3D).

Evolutionary conserved region in chromosome X had higher methylation in the brain of major depressed patients compared to healthy controls

Since the DMR at chromosome X and in intron 9 of Drosha are evolutionary conserved in humans, we compared DNA methylation patterns of these regions in postmortem brain samples of major depressive disorder (MDD) patients and healthy controls (Supplementary Table S2). Analysis using linear mixed model showed that MDD samples (n = 20) had significantly higher methylation (β = ~5.7%, SE = 2.3, P < 0.01) than control samples (n = 10) after controlling for age, race, sex, postmortem interval, and pH (Figure 4(a)). When MDD cases were stratified into suicide (n = 10) and non-suicide groups (n = 10), methylation remained significantly higher in MDD suicide postmortem brains than control postmortem brains methylation (β = ~5.8%, SE = 3.5, P < 0.05) (Figure 4(b)). Methylation was also significantly higher in MDD non-suicide postmortem brains than control postmortem brains by ~5% (β = ~5.1%, SE = 2.2, P < 0.01) (Figure 4(c)). In contrast, no significant difference was observed between MDD and controls postmortem brains at the DMR of intron 9 of DROSHA (Figure 4(d)). When cases were stratified into MDD suicide and MDD non-suicide groups, no significant difference was observed between MDD suicide vs. control (Figure 4(e)) and MDD non-suicide vs. control for intron 9 of DROSHA (Figure 4(f)). Similar to gene expression level in the mouse brain, neither BCOR nor DROSHA was differentially expressed between MDD and control post-mortem brains (Supplementary Fig. S4A-F).

Figure 4.

Comparison of DNA methylation between MDD and controls at the evolutionary conserved intergenic region at chromosome X and intron 9 of DROSHA in Broadman Area 46. (a–c) Intergenic region of chromosome X. (d–f) Intron 9 of DROSHA. Linear mixed modeling was used to compare DNA methylation at the DMRs between (a, d) all MDD cases vs. controls, (b, e) MDD suicide cases vs. controls and (c, f) MDD non-suicide cases vs. controls. Figures shown are mean ± S.E.M of the DMR for each group. P values were calculated using likelihood ratio test as described in material and methods controlling for age, race, sex, postmortem interval, and pH. **: P < 0.01 [χ² (1) = ~7.3]; *: P < 0.05 [χ² (1) = ~4]; N.S.: not significant.

Discussion

Although chronic social defeat stress has been known to affect behavior, little is known as to how this negative stressor might alter DNAm changes in the brain at a genome-wide level. A recent study by Blaze et al. showed that bone marrow transferred from mice exposed to repeated social defeat stress resulted in recipient mice being susceptible to sub-threshold social defeat [22]. This study further showed that these recipient mice had reduced global DNA methylation level in the hippocampus compared to recipient mice that obtained bone marrow from naïve mice [22]. The study therefore suggests that social defeat stress could affect behavior by altering global DNAm pattern in the brain. Similar to this study, the present study assessed the effects of chronic social stress on DNA methylation patterns in the dentate gyrus of mice. The dentate gyrus is known to play an important role in neurogenesis, which can affect behavior. Reduced neurogenesis has been shown to contribute to anxiety and depression whereas there is evidence that enhancing neurogenesis can ameliorate these phenotypes [23,24]. Our study provides the first genome-wide methylome assessment of the dentate gyrus of mice exposed to chronic social stress. This was accomplished by using a targeted genome-wide methylation approach that focused on methylation patterns at promoters of genes, and at their distal regulatory elements. To test the robustness of our initial findings, we sought replication in additional cohorts of chronic social defeated mice. Evidence for at least some replication emerged from two distinct regions: one in intron 9 of Drosha, and one in an intergenic region on chromosome X.

Drosha is a RNase type III protein that plays a pivotal role in the biogenesis of microRNAs. It cleaves the lower stem loop of pre-microRNAs prior to their export from the nucleus to the cytoplasm for further processing to become mature microRNAs [25]. Loss of Drosha has been shown to abolish majority of microRNA expression in a cell line [26]. In addition to its involvement in micro-RNA biogenesis, recent studies have also revealed its role outside of micro-RNA processing. These include regulating mRNA stability, transcriptional activation, RNA splicing, transcriptional termination of long noncoding RNAs that serve as primary micro-RNAs, maintenance of genome integrity and antiviral defense [27]. Importantly, Drosha has been observed to be involved in adult neurogenesis. In a recent study by Rolando et al., the authors used a cre-lox system to selectively delete Drosha in transgenic mice to investigate its role in neurogenesis. The loss of Drosha not only significantly reduced the number of adult neural stem cells generated by neurogenesis, but also changed the cell fate of their progeny [28]. Whereas adult neural stem cells in the dentate gyrus normally generate glutamatergic granule neurons and astrocytes, knockout of Drosha switched the cell fate to that of oligodendrocytes [28]. Consistent with this finding, change to Drosha expression in the hippocampus has also been correlated with depression in mice. A study by Mulligan et al. observed Drosha expression in the hippocampus to correlate with measures of depression in the BXD gene expression and trait datasets [29]. Although changes to Drosha expression can affect behavior, chronic social defeat stress was not observed to alter its expression in our study despite an observable change to its methylation pattern at intron 9. However, it should be noted that although this study identified a DMR in intron 9 of Drosha, this finding was only replicated in the first mouse replication cohort, but not in the second one. In addition, it was not observed in postmortem MDD brains. Thus, although this study provides some evidence that chronic social defeat stress alters DNAm in the intron of Drosha, further replication is required to validate this finding.

The region whose methylation pattern was most consistently altered by chronic social stress in our study was an intergenic region at chromosome X. Increased methylation was observed across all three independent cohorts tested. Interestingly, this region which is conserved in humans is similarly highly methylated in MDD compared to control brain. This DMR contains histone modification marks characteristic of an enhancer region. Despite changes to its methylation pattern, changes in the expression of the genes neighboring this DMR were not observed. There are a number of potential explanations for this.

First, enhancers may not influence the closest neighboring genes. For example, a polymorphic variant in intron 9 of CLEC16A was observed to influence the expression of DEXI, ~150 kb away, but not CLEC16A itself [30]. A chromosome conformational capture assay, together with histone modification marks of enhancers, revealed that this intron is a distal enhancer of DEXI [30]. Similarly, a previous study that identified an enhancer present in intron 5 of Lmbr1 showed that it regulates the expression of the sonic hedgehog gene, which is 1Mb downstream from Lmbr1 [21]. Given these prior observations, it is possible that the enhancer region we identified on chromosome X may be a regulator of other genes not investigated in this study.

Second, the effect of DNA methylation changes may be offset by other epigenetic mechanisms. Although a strong connection between changes to DNA methylation pattern and gene expression have been previously described [18,31], some studies have demonstrated that other epigenetic mechanisms may play a greater role at regulating gene expression in certain stress paradigms. For example, a previous study that also exposed mice to chronic social defeat stress showed a change in BDNF expression in the absence of DNA methylation changes in that locus [32]. Changes to BDNF gene expression were, however, linked with changes to histone modification [32]. Another study showed that microRNAs that regulate the expression of Nr3c1, Nr3c2, and Fkbp4 were increased by repeated social defeat [33]. These studies highlight how other epigenetic mechanisms might alter or override the effect of DNA methylation patterns on gene expression levels.

Third, dynamic changes to RNA temporal profiles may preclude detectable changes to transcriptional levels. RNA expression has a dynamic temporal profile observed as peaks and troughs of RNA levels over time and is determined by their transcriptional and degradation rate. In a study by Rabani et al., RNAs labelled with 4-thiouridine were observed to vary in temporal profiles for different genes [34]. Transcriptional rate and degradation rates were both observed to determine the temporal profiles of gene expression. Genes that showed a constant degradation rate had temporal profiles that were shaped by their transcriptional rate [34]. In contrast, genes that have a more variable degradation rate over time have temporal profiles shaped by their degradation rate [34]. Due to this dynamic change of RNA expression, a single time point may not provide an optimal window to detect changes in transcript levels. As such, a time course may be required to fully determine the impact of chronic defeat stress on the transcriptional profiles in the dentate gyrus. The number of animals required for such an investigation is beyond the scope of the present study.

The present study has several limitations. Although some consistency in methylation changes were observed between cohort 1 and cohort 2 at intron 9 of Drosha, and across three cohorts for the intergenic region in chromosome X, the differences seen were not entirely uniform. The magnitude of difference varied, with the average fluctuation being 18%, which is consistent with the variation across Methyl-Seq and pyrosequencing platforms reported in our previous study [14]. Reasons for this difference have been described [14]. Second, inherent variation in the effect of social defeat stress on methylation change was observed in cohort 3. The effect size was distinctly reduced for the intergenic region at chromosome X and not observed at intron 9 of Drosha. We cannot exclude the possibility that variations between cohorts could have prevented detectable change of Drosha expression and that of genes surrounding the intergenic region of chromosome X, since RNA expression change was queried in a separate cohort. Third, although DNAm at the intergenic region of chromosome X was higher in MDD than in control brain, which is consistent with the methylation pattern observed in all three social defeat stress mice, this observation was performed in a small MDD cohort. A larger cohort is required to replicate this finding to strengthen its validity. Fourth, as we did not have any information regarding the negative stress experience of the MDD cohort, we were not able to determine temporal or causal relationships between this DNAm pattern, negative stressors, and MDD.

In summary, we have identified an evolutionary conserved region in chromosome X whose methylation pattern can be altered by stress and depression. This region is consistently highly methylated in the brain of mice exposed to chronic social defeat. Importantly, this region is similarly more highly methylated in postmortem MDD brains compared to control brains. Further studies with a larger sample of postmortem MDD brains are required to replicate this finding.

Material and methods

Animals

Subjects

A total of 121 adult male C57BL/6J mice (Jackson Laboratories, approximately 8–10 weeks old on arrival) were used. Resident aggressors were 48 male CD-1 retired breeder mice (Charles River). The resident aggressive mice were 16–32 weeks old and weighed between 33 and 42 grams. During the Social Defeat protocol, all mice were housed in pairs in standard mouse cages; all mice were separated by a clear, ventilated divider. All mice had ad libitum access to standard rodent diet (2018 Teklad, Envigo) and water. All procedures were approved by the Animal Care and Use Committee at Johns Hopkins University School of Medicine and were performed in accordance with guidelines established in the National Research Council’s Guide for the Care and Use of Laboratory Animals.

Animals were allowed to habituate for one week before starting social defeat experiments. Residents were screened for aggression by allowing them to interact with an unfamiliar C57BL/6 mouse for 10 minutes a day for 5 days prior to starting. Animals were divided into 2 experimental groups: Unstressed control (CON) and Social Defeat (SD). CON groups remained group housed in pairs separated by a divider and were undisturbed except for daily body weight measurement. SD mice were pair housed with a CD-1 resident and received social stress as described in detail below.

Social stress procedure

Social Defeat was conducted according to Golden et al., 2011 [35]. Briefly, SD mice were paired with a different CD-1 aggressor each day. The SD mouse was placed into the aggressor’s home cage and mice were allowed to interact for 10 minutes. Aggression scores were assigned to each dyad to ensure that each intruder received similar levels of aggression. After 10 minutes, mice were separated by a clear, ventilated barrier wall within the same cage and they remained co-housed for the next 24 hours. SD mice were moved to a different resident aggressor’s cage each day. Before each stress period, SD mice were weighed and fur was inspected and scored on a subjective scale of 1–4 according to coat condition. A score of 4 indicated a shiny, smooth coat and a score of 1 was assigned for coats that appeared ruffled, dull, and uneven over the whole body.

Behavioral tests

Open field test

Locomotor activity of all mice was measured during a 10 minutes Open Field Test. The open field apparatus consists of an opaque plastic box (40 cm square chamber, 30 cm high walls) with a clearly marked central zone (circle with a 35 cm diameter). We recorded each mouse’s behavior using a digital camera. A computerized detection system (Digiscan) recorded activity level (distance traveled). Videos were scored using Hindsight software to record the time spent in the central zone, exploring, immobile, rearing, and grooming. The open field apparatus was cleaned with ethanol solution between each subject.

Elevated plus maze (EPM) test

All mice were tested for anxiety-like behavior during a 5-minute EPM Test. The EPM apparatus is constructed of opaque black plastic and consists of four arms (6 × 35 cm) adjoined by a square intersection (6 × 6 cm) to form a ‘+’ shape. Two opposing arms have walls that are 15 cm high. The remaining two arms are open and have no walls. The base of the maze is constructed such that the arms are elevated 50 cm above the ground. Each mouse was started at the center of the maze facing an open arm. Activity was recorded using a digital camera mounted above the EPM. Video recordings were later scored for the time spent in the open arm, closed arm, or the center of the EPM. The apparatus was cleaned with ethanol solution between each subject.

Sucrose preference test

All mice were tested for anhedonia-like behavior using a Sucrose Preference Test. Mice were habituated to a 1% w/v sucrose solution 3 days before the start of the test. Mice were given ad libitum access to two bottles: one containing water, and one containing 1% sucrose solution at the start of the dark cycle. Intake of water and sucrose solutions were recorded after 12 and 24 hours of access to calculate a preference ratio during the dark and light cycles and over the entire testing period.

Corticosterone measurement

A blood sample (~20 μl whole blood) was collected from all mice on Day 14 of stress 3 hours after the onset of the light cycle to measure basal plasma corticosterone levels. Blood was centrifuged at 4°C and plasma was collected for corticosterone measurement by radioimmunoassay (MP Biomedicals) according to manufacturer’s instructions.

Tissue collection

Animals were sacrificed by rapid decapitation. Blood was collected in EDTA coated collection tubes and centrifuged at 4°C at 3,000 RPM for 15 minutes to collect plasma which was stored at -80°C. Brains were removed, frozen immediately on powdered dry ice, and stored at -80°C. Spleen, adrenal glands, and thymus were dissected and weighed.

Dentate gyrus dissection

Brains from non-behaviorally tested cohorts of Stress and Control mice were used for DNA methylation assays to avoid potential stress effects of behavioral testing on DNA methylation and gene expression. Frozen mouse brains were sectioned using a cryostat and 200 µm sections were mounted on plain glass slides. A 19G needle (ID 0.686 mm × OD 1.086 mm) was used to punch out the dentate gyrus from sections containing the hippocampus (bregma −1.22 mm through −2.30 mm). Samples were stored at −80°C until processed for genomic DNA.

DNA and RNA extraction

DNA and RNA were extracted using MasterPure™ Complete DNA and RNA Purification Kit (Cat. No. MC85200, Epicenter) as described in the manufacturer’s protocol.

Methyl-Seq library

Methyl-Seq was performed using the Agilent SureSelect Methyl-Seq kit (Cat. No. G9651A, Agilent) as described in the manufacturer’s protocol (Methyl-Seq protocol, version A.1, April 2012). A detailed description of the protocol was previously described [14]. In short, 1 µg of DNA, quantified by qubit double-stranded DNA broad range assay (Cat. No. Q32850, Thermo Fisher Scientific), was used for Methyl-Seq library prep. DNA was fragmented by Covaris sonication and ligated to Methyl-Seq adapters. Targeted regions were then isolated by using RNA baits that were complementary to the strand of the targeted regions. Isolated targets were then bisulfite converted using EZ DNA Methylation-Gold^TM (Cat. No. D5005, Zymo Research) followed by PCR amplification. Concentration and sizes of DNA fragments were analyzed by Agilent Bioanalyzer 2100 after shearing, adaptation ligation and indexing, and were observed to be within the range advised by the manufacturer’s protocol.

Next-generation sequencing

Sequencing was performed using 100 bp paired-end sequencing by Illumina HiSeq2000 as previously described [14]. Briefly, quality and concentration of library preparation were determined by bioanalyzer using a high sensitivity DNA kit (Cat. No. 50674626, Agilent) and a KAPA real-time PCR assay (Cat. No. #KK4835, KAPA Biosystem, USA). Three samples were multiplexed in a HiSeq lane to a final concentration of 12 pmole for sequencing. This provided a cluster density between 600–700 K/mm². 1% of phiX genome was spiked into the reaction to overcome color imbalance inherent to low complexity in a bisulfite-converted genome. A control lane containing regular genome that was not bisulfite converted was also processed in the same flow cell as bisulfite-converted libraries. Q30 scores of bases from HiSeq reactions were within the threshold recommended by the manufacturer.

Bisulfite conversion

DNA used for bisulfite-pyrosequencing was bisulfite-converted using EZ DNA Methylation-GoldTM (Cat. No. D5005, Zymo Research) as described in the manufacturer’s instructions. DNA was eluted in 12 µl of elution buffer.

Nested PCR for validation experiment

Validation of differentially methylated cytosines as detected by Methyl-Seq was performed by bisulfite pyrosequencing as previously described [14]. Briefly, nested PCR primer sets were designed using Primer 3 and Pyromark assay design SW 2.0 (Cat. No. 9019077, Qiagen) and primers were purchased from Integrated DNA Technology, USA. For inner PCR primer sets, one of the inner primers was biotinylated and HPLC purified. Outer PCR amplification was performed with 1 µl of bisulfite-converted DNA using Taq polymerase with Thermopol® buffer (Cat. No. M0267S, New England Biolabs). Inner PCR amplification was performed using 2 µl of outer PCR product. Thermal cycling conditions were performed as described in the manufacturer’s instructions. Annealing temperatures for outer and inner primer sets were 57°C and 53°C, respectively. The extension times were 45 and 30 seconds for inner and outer primer sets, respectively. The number of thermal cycles was 30 and 40 for inner and outer primer sets, respectively. The PCR products were electrophoresed on a 1% agarose gel to check for product specificity. Primers used are described in Supplementary Table S3.

Pyrosequencing

Pyrosequencing was performed using capillary dispensing tips on the PyroMark Q96 MD (Qiagen) as described previously [14]. PyroMark CpG software (Cat. No. 9019067, Qiagen) was used in this assay. Bisulfite conversion was shown to be efficient for all samples as the fluorescence signal by cytosine in a non-CpG context was ≤1% of the signal produced by thymine. Sequence specificity was also ascertained by introducing a base that is not in the target sequence.

Quantification of mRNA transcripts

RNA was DNase treated (Thermo Fisher Scientific, USA) and reverse transcribed using the SuperScript III First-Strand Synthesis System (Thermo Fisher Scientific) as described in the manufacturer’s protocol. The following Taqman probes (Thermo Fisher Scientific) were used to detect transcript levels in mice: Atp6 (Mm00510398_g1), BCor (Mm00551516_m1), Drosha (Mm01310009_m1), Med14 (Mm00488809_m1), Hprt (Mm00446968_m1), and Hprt (Mm00446968_m1) in mice. For human transcripts, BCOR (Hs00372378_m1), DROSHA (Hs00203008_m1), and GAPDH (Hs03929097_g1) were used. Thermal cycling was performed using the ViiA™ 7 Real-Time PCR System (Thermo Fisher Scientific) as follows. Hold stage: 50°C for 2 mins followed by 95°C for 10 minutes. PCR stage: 40 cycles of 95°C for 15 seconds followed by 60°C for 1 minute. Primer efficiency was calculated for each reaction using LinregPCR version 2016.1 [36] to ensure consistent amplification across all samples. Data for the target genes were normalized to the housekeeping gene Hprt for mice transcripts and GAPDH for human transcripts using the relative fold change 2^−ΔΔCt method, as previously described [37].

Data analysis

Bisulfite sequencing data from HiSeq were trimmed using Trim Galore! (http://www.bioinformatics.babraham.ac.uk/projects/trim_galore/) to remove standard Illumina paired-end adaptors using the settings – paired -t -q 30 44. Data were subsequently aligned to the mouse genome mm9 using Bismark and Bowtie2 with default settings. Mapping efficiency was ~84%. Bisulfite error rate was estimated from the non-CpG context giving an average of ~1.0% in the dentate gyrus. This is consistent with our previous finding and reports from other investigations [14,38]. A higher bisulfite conversion error rate from brain tissue is expected as methylation in the non-CpG context has been observed to be elevated in the brain [39,40]. Sequence Alignment/Map (SAM) files from Bismark output were subsequently imported to GeneSpring NGS version 12.6.1 (Agilent Technologies) for further analysis. Data were filtered for CpGs within targeted regions with a minimum base quality of Q30 followed by deduplication. For methylation detection, only CpGs with a minimum coverage of 10X were used for the analysis. Percentage of methylated cytosine at a CpG site was calculated by counting the number of cytosines as a proportion of the total number of cytosines and thymines. Differentially methylated cytosines (DMCs) were considered as cytosines whose difference in methylation level between stress and control was >20% with a t-test P value ≤0.05. For a targeted region to be considered a differentially methylated region (DMR), at least 10% of cytosines in the CpG context had to be differentially methylated with a Fisher’s combined P value of ≤0.05. Quantile-quantile and Manhattan plots were generated using R. Genome-wide significance thresholds were calculated by using Bonferroni corrections. In short, the default P value of 0.05 was divided by the total number of CpGs being tested.

DNAm identified by bisulfite pyrosequencing for replication cohorts of social defeat stressed mice were analyzed by either Student’s t-test or one-way ANOVA with Sidak’s post-hoc multiple comparison test using Graphpad prism. MDD postmortem brains were obtained from Maryland Brain Collection at the Maryland Psychiatric Research Center, Baltimore, Maryland. DNAm data obtained by bisulfite pyrosequencing from postmortem brains were analyzed using R with a linear mixed-effects model using the ‘lme4’ package. For the null model, DNAm was the response variable while phenotype (i.e., MDD or control) was the independent variable. Other independent variables included in the analysis were age, pH, post-mortem interval and sex as provided by Maryland Brain Collection. This was done to control for the effect of these variables. Subjects were specified as a random factor. The alternative model was the null model without phenotype, but included the aforementioned variables and subjects as a random factor. To determine if phenotype affects DNAm, a likelihood ratio test was performed using ANOVA incorporating the null model against the alternative model.

For behavioral experiments and organ weights, data was analyzed using Statistica 7 software (Systat, Tulsa, OK) to determine if there was an effect of social defeat stress using a factorial or repeated measures ANOVA test. P < 0.05 was considered to be significantly different.

Funding Statement

This work was supported by the National Institute of Mental Health [R01MH090595];National Institutes of Health [T32GM008629];

Acknowledgments

This work was supported by the National Institute of Mental Health [R01MH090595].

P.R.B. received training funding from the National Institutes of Health Predoctoral Training Grant T32GM008629, PI Daniel Eberl.

Disclosure statement

No potential conflict of interest was reported by the authors.

Supplementary material

Supplemental data for this article can be accessed here.