Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2022 Jun 1.
Published in final edited form as: J Trauma Stress. 2021 Feb 2;34(3):607–615. doi: 10.1002/jts.22655

Epigenome-Wide Study of Posttraumatic Stress Disorder Symptom Severity in a Treatment-Seeking Adolescent Sample

Christina M Sheerin 1, Eva E Lancaster 1,3, Timothy P York 1,3, Jesse Walker 2,4, Carla Kmett Danielson 2, Ananda B Amstadter 1
PMCID: PMC8217087  NIHMSID: NIHMS1665855  PMID: 33529416

Abstract

Emerging research has demonstrated that psychosocial trauma exposure may elicit epigenetic changes, with downstream effects on the transcriptional regulation of genes. Epigenome-wide association studies (EWAS) offer an agnostic approach to examine DNA methylation (DNAm) associations and are a valuable tool to aid in the identification of biological pathways involved in posttraumatic stress disorder (PTSD). This study represents the first EWAS of PTSD in an adolescent sample, an important group given the significance of this developmental period regarding both DNAm changes and PTSD risk. The sample (n = 39, M age = 15.41 years, SD = 1.27, 84.6% female) comprised adolescents who experienced interpersonal trauma and were enrolled in a treatment study. Participants were assessed using the UCLA PTSD Reaction Index for DSM-IV–Adolescent Version and provided a blood sample at baseline. Genomic DNA was isolated from whole blood and assayed using the Illumina Infinium MethylationEPIC BeadChip. The primary analysis estimated the associations among individual CpG sites and PTSD symptom scores. Of the 793,575 screened probes tested, two were significant at a false discovery rate (FDR) < 10%. Hypomethylation of both sites was associated with increased PTSD symptom scores. Analysis of differentially methylated regions (DMR) identified a DMR associated with PTSD symptom scores at an FDR < 10%. Results from follow-up models are also discussed. Findings from this preliminary investigation suggest the importance of further research conducted in adolescent samples. The analytic pipeline and results are documented for use in future meta-analytic work as more such samples become available.

Childhood exposure to traumatic events is common, with large, prospective, population-based studies demonstrating that between 30% (Copeland et al., 2018) and over 66% (Copeland et al., 2007) of children report exposure to at least one traumatic event by the age of 16 years. Although many youths are resilient following trauma exposure, it is well documented that the conditional risk of posttraumatic stress disorder (PTSD) varies by trauma type (i.e., from 1.3% to 8.8%; Atwoli et al., 2015), with a higher likelihood of PTSD often associated with interpersonal trauma, such as sexual assault (Liu et al., 2017), as well as with trauma exposure that occurs during childhood compared to adulthood (McCutcheon et al., 2010). Evidence indicates that childhood and adolescence represent important developmental periods for PTSD risk (Dunn, Nishimi, et al., 2017), especially regarding exposure to interpersonal trauma during this period (Dunn, Wang, et al., 2017), with multiple neurobiological systems particularly susceptible to its deleterious impact (see Cross et al., 2017, for a review).

It is important to examine differences in responses to trauma exposure as a function of factors such as trauma type and timing, as well as with regard to individual differences, that may account, in part, for this variable risk. Genetically informed research can help inform biological models for understanding this differential risk. The present study specifically sought to identify the biological processes underlying symptom presentation during an important developmental period for PTSD. A growing body of molecular genetic research (i.e., genome-wide association studies [GWAS]) has been successful in identifying novel risk alleles associated with PTSD (for a review, see Daskalakis et al., 2018). Another important aspect related to genetic sequence is the epigenome, through which environmental experience, such as trauma exposure, are posited to affect change regarding chromatin features, transcription factors, enhancers, and other functional proteins that can influence transcriptional activity without changing the underlying genetic code (Feil & Fraga, 2012). To date, DNA methylation (DNAm) has been the most commonly studied type of epigenetic feature (Toyokawa et al., 2012). Once thought to be stable, it is now known that the methylation of DNA is a dynamic process that can be influenced by cellular and environmental factors across the life course (Teperino et al., 2013). Specifically, changes in methyl-(CH3) group status at 5’-C-phosphate-G-3’ (CpG) dinucleotides underlie this dynamic modification through their effects on chromatin, gene expression, and transcriptional regulation (Teperino et al., 2013).

Broadly speaking, human methylation studies have typically examined the relation between methylation status at CpG sites and a phenotype of interest. Epigenome-wide methylation studies (EWAS), like GWAS, offer an agnostic approach to examining epigenetic associations and are a valuable tool to aid in the identification of biological pathways involved in PTSD. To date, several EWAS have been conducted in relation to PTSD and have provided support for genes involved in the stress response system, such as hypothalamic–pituitary–adrenal (i.e., HPA) axis genes and inflammatory genes (see recent review by Mehta et al., 2020). Although this promising literature has been growing, it is important to note that to our knowledge, all DNAm studies of PTSD to date have been conducted in adult samples.

This preliminary investigation aimed to add to the existing EWAS literature by performing a genome-wide examination of DNAm in a treatment-seeking adolescent sample; thus, it was the first study of which we are aware to examine the differences in DNAm methylation associated with PTSD symptom severity in adolescents. We focused on the associations among PTSD symptoms, as the whole sample consisted of individuals with clinically relevant symptomatology.

Method

Participants and Procedures

The present sample (N = 39, M age = 15.41 years, SD = 1.27, 84.6% female) represents a subsample of adolescents participating in a larger U.S. National Institute on Drug Abuse (NIDA)–funded clinical trial (R01 DA08686; Danielson et al., 2020; examining the efficacy of the Risk Reduction Through Family Therapy (RRFT) intervention; N = 140). Shortly after data collection began, funding was received for an additional genetic component of the study, after which all participants were approached and given the option to participate. The present study sample includes all participants who consented or assented to have blood drawn for genomic analysis. Participants were recruited into the parent study through a community clinic specializing in trauma treatment. The RRFT intervention was designed for adolescents with a history of interpersonal trauma exposure and substance use and aimed to reduce both PTSD symptoms and substance use problems (Danielson, Adams, & Hanson, 2019). The inclusion criteria for the parent study were: (a) age of 13–18 years; (b) exposure to interpersonal violence (i.e., at least one experience of sexual abuse, physical abuse, physical assault, threat with a weapon, and/or witnessing violence); (c) the presence of at least five PTSD symptoms, as measured using the UCLA PTSD Reaction Index (UCLA-PTSD-RI); and (d) at least one nontobacco substance-using day in the past 90 days. Youth with active suicidal or homicidal ideation, active psychosis, intellectual disability, or pervasive developmental disability, as well as those already engaged in treatment, were excluded.

All participants (i.e., parents/caregivers of minors, adolescents) provided written informed consent or assent, respectively, for the larger treatment study and the blood draw as part of the add-on study. This study was approved by the institutional review board of the Medical University of South Carolina. Following consent, participants completed a battery of self-report measures before beginning the parent study intervention. Blood samples (< 30 ml) were collected during the baseline pretreatment assessment session by a trained nurse or phlebotomist. Only data from the baseline measurement were analyzed in the present study.

Measures

Trauma History and PTSD Symptoms

Trauma history and PTSD symptoms were assessed using the adolescent and caregiver versions of the UCLA-PTSD-RI (UCLA-PTSD-RI; Steinberg et al., 2004) for the fourth edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV); data collected from the adolescent self-report version was used for the present study. The UCLA-PTSD-RI has demonstrated good-to-excellent test–retest reliability (Goenjian et al., 2001). In the present sample, Cronbach’s alpha was .90.

The UCLA-PTSD-RI includes a list of 12 potentially traumatic events (PTEs), with an additional item (i.e., “other”) meant to reflect PTEs not included in the list. We summed the number of items endorsed to create a total baseline trauma event count, which was used as a covariate in the analyses. If a participant endorsed more than one traumatic event, they were asked to identify the item that bothered them the most and respond to subsequent PTSD symptom items associated with that event. The UCLA-PTSD-RI also includes a list of 20 symptoms, the past-month frequency of which respondents are asked to rate on a Likert-type scale ranging from 0 (none) to 4 (most days). A cutoff score of 38 indicates probable PTSD (Steinberg et al., 2004). In the EWAS, the total symptom score was the variable of interest. The mean total score in the present sample was 33.80; based on the established cutoff criteria for the measure, 43.6% of participants were considered to have scored above the threshold for probable PTSD. Thus, although all participants reported trauma exposure and at least some PTSD symptoms, variability existed, allowing us to examine DNAm associated with symptom severity. The timeline follow-back method was used to determine any nontobacco substance use, including alcohol, and the number of substance-use days during the previous 90 days (Sobell & Sobell, 1992).

DNA Preparation

Genomic DNA was isolated from whole blood according to standard methods using the Puregene DNA Isolation Kit (Qiagen; Hilden, Germany). Fresh blood samples were isolated via standard procedures and stored. An aliquot of 500 ng of DNA per participant was sent to the HudsonAlpha Institute for Biotechnology (Huntsville, AL, USA) in one shipment for bisulfite conversion and genome-wide methylation measurement. HudsonAlpha was blind to the study purpose and received only the unique identifier; thus, technicians were blind to PTSD symptom status. Samples were randomized on the plate, and standardized quality control measures were employed (e.g., negative controls used). An analysis of variance (ANOVA) was conducted to confirm that PTSD symptom count was randomly distributed across arrays (i.e., groups of eight samples). Genome-wide methylation was assayed using the Illumina (San Diego, CA, USA) Infinium MethylationEPIC BeadChip, which interrogates over 850,000 methylation sites per sample; this array retained more than 90% of the original CpGs from the Illumina HumanMethylation 450k BeadChip array plus an additional 350,000 CpGs largely in enhancer regions. Both procedures were completed according to standard protocols.

Data Analysis

Genome-Wide DNAm Analysis

Quality control and analytic scripts are available on the Open Science Framework (https://osf.io/cza45/?view_only=c3dc042f086d4723b4aa2100a96fcb22). The minfi Bioconductor package (Aryee et al., 2014) in R was used to process intensity values from the scanned arrays (Wright et al., 2016). Single nucleotide polymorphism (SNP) probes were used to verify sample identity and the independence of samples. Quality control for major steps in sample processing was assessed quantitatively and visually to identify the presence of deviant samples (Aryee et al., 2014). Beta values were determined based on the ratio of the methylated probe intensity to the sum of the methylated and unmethylated probe intensities (Du et al., 2010). Density plots for beta values from each array were inspected to identify poor performing arrays based on a large deviation from the rest of the samples. Probes were filtered if they had a detection p value greater than .01 in at least 10% of samples or have been previously identified as cross-hybridizing (Chen et al., 2013). Probes with an SNP present at the CpG site or the single-base extension were filtered. Quantile normalization adapted to DNAm arrays (Touleimat & Tost, 2012) was applied to adjust the distribution of Type I and Type II probes to the final set of screened sample arrays and probes. See Supplementary Figures S1–S9 for figures that resulted from the quality-control analyses.

Ancestry determination was made by sample clustering derived from principal components (PC) estimated from ancestry informative probes (Barfield et al., 2014; see Supplementary Figures S11–S20); in the present analyses, we controlled for the three ancestry PCs that correlated with race (i.e., PC3, PC4, PC5). Cellular heterogeneity was accounted for by inferring blood cell proportions for each sample (Houseman et al., 2015). Neutrophil cell proportions were included as a covariate as they represent the largest cell type proportion and were highly correlated with all other cell types, rs = .59 to −.79 (see Table 1 and Supplementary Figure S5). In an effort to decrease the number of covariates in the model and prevent overfitting, particularly given the relatively smaller sample size, this was the only cell type included as a covariate in the model (see Supplementary Table S2 for a study variable correlation table).

Table 1.

Demographic and Clinical Characteristics of the Epigenome-Wide Study Sample

Variable n % M SD
Sex (female) 33 84.6
Race (White) 25 64.1a
Age (years) 15.41 1.27
Number of PTEs 3.08 1.53
PTSD symptom scoreb 33.80 14.58
Met diagnostic cutoff for probable PTSD 17 43.6
Past 90-day alcohol use 27 69.2
Number of days of alcohol use in the past 90 days 11.85 16.85
Any drug use in the past 90 days 17 43.6
Open in a new tab

Note. N = 39.

a

20.5% Black, 12.8% biracial, 2.6% Hispanic/Latino.

b

Evaluated using the UCLA PTSD Reaction Index for DSM-IV–Adolescent Version.

For the primary analysis, beta values were transformed using the M value procedure (i.e., calculated as a logit transformation of the methylated and unmethylated intensity ratio along with an added constant to offset potentially small values) to promote normality (Du et al., 2010). To identify an extraneous structure that may account for batch effects, correlations between experimental variables, and the top 10 principal components derived from the entire set of filtered M values across all arrays were inspected (Leek et al., 2010; Supplementary Figure S21). The ComBat function in R was used to remove average differences across arrays (Leek et al., 2012). We note that neither PTSD symptoms nor trauma count was correlated with array and, thus, did not expect batch correction to remove variability on our outcome of interest (Supplementary Figure S22). In more conservative follow-up models, we also included row number (e.g., 1–8) as a numeric covariate. Finally, as smoking status was not assessed in the present sample, we examined the potential impact of smoking by assessing the variation in beta values of a specific CpG that has been shown to be highly sensitive to smoking and pollution (Tantoh et al., 2019); no variation was identified (Supplementary Figure S23).

Differentially Methylated Positions Associated With PTSD Symptom Count

Single-probe analysis (i.e., differentially methylated positions [DMPs]) was initially conducted. The primary analysis consisted of M values for each individual probe, i, regressed on PTSD symptom score. This analysis was conducted using the limma package in R (Ritchie et al., 2015), which promotes stable estimates in analyses with small sample sizes (Smyth, 2005). The initial model was specified as

DNAmi~PTSD symptom count+trauma history count+ancestry PC3+ancestry PC4+ancestry PC5+Neu+sex+error

The predictor of interest was PTSD symptom count. Given the small sample size, covariates were limited and included (a) trauma history, to separate the effects of PTSD symptomatology from trauma exposure itself; (b) the ancestry PCs that correlated with self-reported race, (c) cell-type proportions for neutrophils, as they represent the largest proportion and correlated with all other cell types; and (d) sex, given the associations among sex and both the study outcome and DNAm in general. In follow-up sensitivity analyses, we controlled for the first five ancestral PCs along with the effect of row. The p value distribution was used to estimate the false discovery rate (FDR; qvalue package in R; Storey et al., 2019). To capture the incremental validity of significant CpGs, unadjusted and adjusted (i.e., with and without the inclusion of the predictor of interest) R2 values were calculated and compared to determine the variance in PTSD symptoms explained by CpGs in the model. As none of the items related to substance use (i.e., alcohol frequency, any drug use, cannabis use specifically) correlated with PTSD symptoms or technical PCs, they were not included as covariates in the primary model; however, they were included in follow-up analyses.

To further characterize any PTSD symptom severity–associated CpGs, gene boundaries were annotated using the University of California Santa Cruz (UCSC) hg19 known gene track (Kent et al., 2002). The region 2 kb upstream of the transcriptional start site was included in the genic definition to capture proximal regulatory elements (i.e., promoter regions; Carlson et al., 2016).

Differentially Methylated Regions Associated with PTSD Symptom Count

Regional analysis (i.e., differentially methylated regions [DMRs]) was also conducted using the limma package in R and the DMRcate package in Bioconductor (Peters et al., 2015). The DMR analysis was run on the same model as specified earlier. Regions were built from the initial DMP analysis. A candidate DMR required a minimum of two CpGs within 1 kb for each region of interest. The DMRcate regional association method retained the same FDR threshold as was set for the DMP analysis.

Results

The demographic and clinical characteristics of the sample are presented in Table 1. The current sample (i.e., participants in the add-on EWAS) did not significantly differ from participants in the overall study sample with regard to age, race, or PTSD symptom count. There was a higher percentage of male participants in the EWAS compared to the full sample, p = .034, as well as a lower average traumatic event count, p = .031. In the present sample, the average number of traumatic events was 3.08 (range: 1–6), and the average UCLA-PTSD-RI score was 33.80, representing moderate symptoms scores.

Differentially Methylated Probes

All samples passed array-based quality control and were retained in the analyses. Following quality control and processing, 793,575 screened CpG probes remained for association tests. The histogram of p values and Q–Q plot from the primary model, which does not indicate bias in our set of results, are shown in Figure 1. Of the included probes, two were significantly associated with PTSD symptom count at an FDR of less than 10%. We also considered DMPs at a p value threshold set at 10−6 based on existing simulation studies shown to control the family-wise error rate (Tsai & Bell, 2015), although we note that a p value of 10−8 is now considered more appropriate (Mansell et al., 2019). This examination (i.e., p value threshold of 10−6) only added one additional DMP. These individual probes were hypomethylated CpG sites associated with PTSD symptom severity. One of the three DMPs mapped onto the MAML3 gene (i.e., mastermind-like transcriptional coactivator 3). Table 2 lists the nominally significant DMPs, estimates, their respective gene symbols, and their and Entrez ID numbers. The mean R2 change value for the three nominally significant DMPs at a p value of 10−6 (i.e., the change in R2 with and without the inclusion of PTSD symptom scores) was .48 (SD = .04, range: .45–.53), suggesting a large amount variance explained between these DMPs and PTSD symptom score. Due to the modest size of the present sample, only DMPs of relatively large effects would be expected.

Figure 1.

Figure 1

Open in a new tab

Histogram and Q–Q Plot of p Values From the Primary Study Model Assessing DNA Methylation (DNAm) Associated with Posttraumatic Stress Disorder (PTSD) Symptoms

Note. The figure presents the results of the primary model examining the association between DNAm and PTSD symptom severity, covarying for trauma count, sex, ancestry principal components, and neutrophils. Panel A presents a histogram of the p-value distribution from the model. Panel B presents the Q–Q plot of p values from this model. The Q–Q plot is consistent with the degree of significant results that was indicated by false discovery rate estimation.

Table 2.

5’-C-phosphate-G-3’ (CpG) Sites Associated With Posttraumatic Stress Disorder Symptom Severity

CpG probe Chromosome Coefficient t(31) p q Gene Entrez ID Gene name
cg01170198 1 −.02 −6.96 1.27×10−7 .08 NA Intergenic
cg20012601 4 −.01 −6.83 2.19×10−7 .08 55534 MAML3
cg16562342 1 −.001 −6.07 7.67×10−7 .19 NA Intergenic
Open in a new tab

Note. This table presents the nominally significant CpG sites, along with the genes they fall within, that are associated with PTSD symptom severity. Bold font denotes CpG sites at FDR < 10% and p < 10−6, which have been found to be significantly associated with at least one of four brain tissues (i.e., prefrontal cortex, entorhinal cortex, superior temporal gyrus, and cerebellum), using the Blood Brain DNA Methylation Comparison Tool (Hannon et al., 2015).

Differentially Methylated Regions

The DMR analysis identified one region associated with PTSD symptom score at an FDR less than 10%. This region included two CpGs and overlaps with the MOBP gene (i.e., myelin-associated oligodendrocyte basic protein).

Follow-up Models

Given the exploratory nature of epigenome-wide studies, several follow-up models were conducted examining DMPs. Sensitivity analyses began by first focusing on self-reported race; given that the sample was quite diverse and admixed, we reran a model that removed the ancestry PCs and instead used self-reported race, including only participants who endorsed being White or Black. This model largely aligned with the original model, with slightly attenuated results (i.e., same significant DMPs, but at a higher FDR or p value threshold). Next, further analyses were conducted with additional covariates. A more conservative model covarying for all five ancestry PCs as well as row was conducted. The DMPs with an FDR of less than 10% did not remain. A second sensitivity model, which included alcohol consumption as a covariate, as it was the most commonly endorsed substance, resulted in a similar outcome with slightly attenuated results: There were no DMPs at an FDR of less than 10% but 28 DMPs at an FDR less than 30%.

Discussion

To our knowledge, the present study was the first examination of DNAm and PTSD symptoms in an adolescent, treatment-seeking sample. This study identified sites differentially methylated as a function of PTSD severity while controlling for trauma history. The analyses identified two DMPs associated with PTSD symptom severity at an FDR threshold of less than 10%. One was in an intergenic region, and the other overlapped with the MAML gene (i.e., mastermind-like transcriptional coactivator). This gene is a key component of the Notch signaling pathway, has been implicated in central nervous system plasticity across the lifespan (Ables et al., 2011), and was identified in a GWAS of multisite chronic pain (Johnston et al., 2019). Although the present findings need replication, this association was of interest given the well-established association between PTSD and chronic pain from a phenotypic perspective (Siqveland et al, 2017). The analyses also identified a DMR at an FDR threshold of less than 10% in the MOBP gene (i.e., myelin-associated oligodendrocyte basic protein). The MOBP gene was one of several identified as significantly associated with PTSD status in a meta-analysis of two gene expression datasets (Chitrala et al., 2016), suggesting some consistency across DNAm and gene expression studies.

Although most EWAS to date have focused on PTSD case status, three such EWAS have identified significant CpG sites associated with symptom severity. Mehta and colleagues (2017) identified five CpGs sites at an FDR of 10%; Rutten and colleagues (2018) identified 17 CpG sites at an FDR of 5%, and Uddin and colleagues (2013) identified 80 such sites (p < .01). The present study findings identified a much smaller number of CpG sites and only at a slightly higher FDR. The differences in these results may reflect the importance of considering developmental timing in comparing findings across the lifespan, as a range of reactions following early life stressors change over the course of development (Trickett et al., 2010), and the majority of work to date examines adult samples.

The present results must be viewed in the context of a number of study limitations. First, the sample size is modest, and replication with independent samples is needed. However, EWAS sample sizes in the PTSD literature have ranged widely from, from 14 to 513 participants. The mean estimated R2 change value of .48 was consistent with the propensity of the study design to detect large effect sizes, although CpGs with smaller effects would be expected to exist, and larger sample sizes would increase the ability to find these effects. Although unlikely, in the context of the small sample size, the number of covariates in the model may have resulted in overfitting. Alternatively, it is also likely that power was impacted due to the increased number of covariates in the follow-up models. Temporally informative studies that include the integration of multiomic measurements (i.e., transcriptomics, proteomics) are also needed to provide more evidence for potential mechanistic processes underlying the associations reported herein. We note that the present data were from a treatment outcome study; thus, when data become available, pre–post analyses can be conducted. Although it was outside the scope of the present study, an examination of methylation age in the present sample in particular and trauma-exposed adolescents more broadly will add to our understanding of the impact of traumatic events during this developmental period. However, we note that to date, estimates of methylation age for younger samples have not been shown to be robust compared to adult samples.

The lack of an independent replication sample for analysis is a limitation that requires interpretation of the present study findings as preliminary in nature. It is also important to note that the present sample represents adolescents with additional comorbid conditions (e.g., substance use problems and depression symptomatology). Although psychiatric comorbidity is common among individuals with PTSD, it could also mean that these findings are not specific to PTSD. However, substance use was not associated with PTSD symptom severity in this sample, and its inclusion in models did not notably alter the results. Further, information on nicotine use was not assessed, and its potential impact on methylation in this sample cannot be specifically determined. However, an examination of variation at a CpG site that is highly sensitive to smoking and pollution demonstrated no variation, suggesting the impact of smoking was not a strong concern in this sample. The use of an adolescent sample, while important for identifying relevant factors in the risk for PTSD during an important developmental window, also means that the findings may not generalize across the broader lifespan. As datasets that include adolescent participants grow, comparisons of findings between children, adolescents, and adults will add important information on the developmental impact of trauma exposure and the impact on methylation signatures. Finally, this study measured DNAm in peripheral blood, whereas PTSD is considered a disorder of the central nervous system. The rationale for utilizing peripheral blood in epigenetic research is supported by both technical and biological considerations. Blood is often chosen as a surrogate tissue because of its unique ability to reflect the state of other regions in the body. White blood cells are able to interact with other tissues, and there is evidence that these cells pick up epigenetic markers from those areas (Bakulski et al., 2016). Further, in the most recent EWAS of PTSD to date, Logue and colleagues (2020) found cross-replication in an independent sample of DNAm generated from prefrontal cortex tissue; the authors concluded that this suggests DNAm in peripheral tissue can yield replicable PTSD associations that may mirror associations in the brain. The nominally significant CpGs in the present study were examined using the Blood Brain DNA Methylation Comparison Tool (Hannon et al., 2015), and one CpG was associated with at least one brain tissue.

The aim of the present preliminary investigation was to extend the DNAm literature on PTSD to include an examination in adolescents who had experienced childhood interpersonal trauma given the importance of developmental timing and this type of trauma exposure with regard to PTSD risk. Although the study was limited due to its modest sample size, and the results should be interpreted with caution given the lack of replication and nominally significant findings (e.g., two DMPs and one DMR at an FDR less than 10%), the present findings on methylation in genes of potential interest do align with those from other samples related to trauma exposure and PTSD. This work also highlights the need for an examination of epigenetic profiles associated with trauma exposure and PTSD during this developmental stage that may be meaningful for longer-term mental health outcomes across the lifespan. It is our hope that the present study results and analyses can be used in future work as a replication sample as well as in meta-analyses.

Supplementary Material

Supplementary Materials

Acknowledgments

The authors have no conflicts of interest to disclose. The overarching study was funded by the National Institute on Drug Abuse (R01 DA08686 to Carla Kmett Danielson), and the genetic component was funded by the NARSAD Research Institute (independent investigator grant to Ananda B. Amstadter). Christina M. Sheerin’s time was funded by the National Institute on Alcohol Abuse and Alcoholism (NIAAA; K01 AA025692). Ananda B. Amstadter’s time was partially funded by the NIAAA (K02 AA023239).

Footnotes

Open Practices Statement

The study reported here was not formally preregistered. The raw data have not been made available on a permanent third-party archive due to institutional review board restrictions. Data analysis scripts and results files are posted at https://osf.io/brdsu/?view_only=5251dd1bbb914d5f9d79fdc20e2cc2ad

References

  1. Ables JL, Breunig JJ, Eisch AJ, & Rakic P (2011). Not(ch) just development: Notch signalling in the adult brain. Nature Reviews Neuroscience, 12(5), 269–283. 10.1038/nrn3024 [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Aryee MJ, Jaffe AE, Corrada-Bravo H, Ladd-Acosta C, Feinberg AP, Hansen KD, & Irizarry RA (2014). Minfi: A flexible and comprehensive Bioconductor package for the analysis of Infinium DNA methylation microarrays. Bioinformatics, 30(10), 1363–1369. 10.1093/bioinformatics/btu049 [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Atwoli L, Stein DJ, Koenen KC, & McLaughlin KA (2015). Epidemiology of posttraumatic stress disorder: Prevalence, correlates, and consequences. Current Opinions in Psychiatry, 28(4), 307–311. 10.1097/yco.0000000000000167 [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Bakulski KM, Halladay A, Hu VW, Mill J, & Fallin MD (2016). Epigenetic research in neuropsychiatric disorders: The “tissue issue”. Current Behavioral Neuroscience Reports, 3(3), 264–274. 10.1007/s40473-016-0083-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Barfield RT, Almli LM, Kilaru V, Smith AK, Mercer KB, Duncan R, Klengel T, Mehta D, Binder EB, Epstein MP, Ressler KJ, & Conneely KN (2014). Accounting for population stratification in DNA methylation studies. Genetic Epidemiology, 38(3), 231–241. 10.1002/gepi.21789 [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Bock C, Walter J, Paulsen M, & Lengauer T (2007). CpG island mapping by epigenome prediction. PLoS Computational Biology, 3(6), e110. 10.1371/journal.pcbi.0030110 [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Carlson MR, Pages H, Arora S, Obenchain V, & Morgan M (2016). Genomic annotation resources in R/bioconductor. Methods in Molecular Biology, 1418, 67–90. 10.1007/978-1-4939-3578-9_4 [DOI] [PubMed] [Google Scholar]
  8. Chen YA, Lemire M, Choufani S, Butcher DT, Grafodatskaya D, Zanke BW, Gallinger S, Hudson TJ, & Weksberg R (2013). Discovery of cross-reactive probes and polymorphic CpGs in the Illumina Infinium HumanMethylation450 microarray. Epigenetics, 8(2), 203–209. 10.4161/epi.23470 [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Chitrala KN, Nagarkatti P, & Nagarkatti M (2016). Prediction of possible biomarkers and novel pathways conferring risk to post-traumatic stress disorder. PloS one, 11(12), e0168404. 10.1371/journal.pone.0168404 [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Copeland WE, Keeler G, Angold A, & Costello EJ (2007). Traumatic events and posttraumatic stress in childhood. Archives of General Psychiatry, 64(5), 577–584. 10.1001/archpsyc.64.5.577 [DOI] [PubMed] [Google Scholar]
  11. Copeland WE, Shanahan L, Hinesley J, Chan RF, Aberg KA, Fairbank JA, van den Oord EJCG, & Costello EJ (2018). Association of childhood trauma exposure with adult psychiatric disorders and functional outcomes. JAMA Network Open, 1(7), e184493. 10.1001/jamanetworkopen.2018.4493 [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Cross D, Fani N, Powers A, & Bradley B (2017). Neurobiological development in the context of childhood trauma. Clinical Psychology: Science and Practice, 24(2), 111–124. 10.1111/cpsp.12198 [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Danielson CK, Adams ZA, & Hanson RF (2019). Risk reduction through family therapy (RRFT): Exposure-based treatment for co-occurring PTSD and substance use problems among adolescents In Vujanovic AA & Back SE (Eds.), Posttraumatic stress and substance use disorders: A comprehensive clinical handbook (pp. 165–182). Routledge. [Google Scholar]
  14. Danielson CK, Adams ZW, McCart MM, Chapman J, Sheidow A, Walker J, Smalling A, & de Arellano M (2020). The safety and efficacy of exposure-based risk reduction through family therapy (RRFT) for co-occurring substance use problems and PTSD among adolescents: A randomized controlled trial. JAMA Psychiatry, 77(6), 574. 10.1001/jamapsychiatry.2019.4803 [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Daskalakis NP, Rijal CM, King C, Huckins LM, & Ressler KJ (2018). Recent genetics and epigenetics approaches to PTSD. Current Psychiatry Reports, 20(5), 30. 10.1007/s11920-018-0898-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Du P, Zhang X, Huang CC, Jafari N, Kibbe WA, Hou L, & Lin SM (2010). Comparison of beta-value and m-value methods for quantifying methylation levels by microarray analysis. BMC Bioinformatics, 11, 587. 10.1186/1471-2105-11-587 [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Dunn EC, Nishimi K, Powers A, & Bradley B (2017). Is developmental timing of trauma exposure associated with depressive and post-traumatic stress disorder symptoms in adulthood? Journal of Psychiatric Research, 84, 119–127. 10.1016/j.jpsychires.2016.09.004 [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Dunn EC, Wang Y, Tse J, McLaughlin KA, Fitzmaurice G, Gilman SE, & Susser ES (2017). Sensitive periods for the effect of childhood interpersonal violence on psychiatric disorder onset among adolescents. The British Journal of Psychiatry, 211(6), 365–372. 10.1192/bjp.bp.117.208397 [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Feil R, & Fraga MF (2012). Epigenetics and the environment: Emerging patterns and implications. Nature Reviews Genetics, 13(2), 97–109. 10.1038/nrg3142 [DOI] [PubMed] [Google Scholar]
  20. Goenjian AK, Molina L, Steinberg AM, Fairbanks LA, Alvarez ML, Goenjian HA, & Pynoos RS (2001). Posttraumatic stress and depressive reactions among Nicaraguan adolescents after Hurricane Mitch. American Journal of Psychiatry, 158(5), 788–794. 10.1176/appi.ajp.158.5.788 [DOI] [PubMed] [Google Scholar]
  21. Hannon E, Lunnon K, Schalkwyk L, & Mill J (2015). Interindividual methylomic variation across blood, cortex, and cerebellum: Implications for epigenetic studies of neurological and neuropsychiatric phenotypes. Epigenetics, 10(11), 1024–1032. 10.1080/15592294.2015.1100786 [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Houseman EA, Kelsey KT, Wiencke JK, & Marsit CJ (2015). Cell-composition effects in the analysis of DNA methylation array data: A mathematical perspective. BMC Bioinformatics, 16(1), 95. 10.1186/s12859-015-0527-y [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Johnston KJ, Adams MJ, Nicholl BI, Ward J, Strawbridge RJ, Ferguson A, McIntosh A, Bailey MES, & Smith DJ (2019). Genome-wide association study of multisite chronic pain in UK Biobank. PLoS genetics, 15(6), e1008164. 10.1101/502807 [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Kent WJ, Sugnet CW, Furey TS, Roskin KM, Pringle TH, Zahler AM, & Haussler D (2002). The human genome browser at UCSC. Genome Research, 12(6), 996–1006. 10.1101/gr.229102 [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Leek JT, Johnson WE, Parker HS, Jaffe AE, & Storey JD (2012). The sva package for removing batch effects and other unwanted variation in high-throughput experiments. Bioinformatics, 28(6), 882–883. 10.1093/bioinformatics/bts034 [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Leek JT, Scharpf RB, Bravo HC, Simcha D, Langmead B, Johnson WE, Geman D, Baggerly K, & Irizarry RA (2010). Tackling the widespread and critical impact of batch effects in high-throughput data. Nature Reviews Genetics, 11(10), 733–739. 10.1038/nrg2825 [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Liu H, Petukhova MV, Sampson NA, Aguilar-Gaxiola S, Alonso J, Andrade LH, Bromet EJ, Girolamo G, Haro JM, Hinkov H, Kawakami N, Koenen KC, Kovess-Masfety V, Lee S, Medina-Mora E, Navarro-Mateu F, O’Neill S, Piazza M, Posada-Villa J, … Kessler RC (2017). Association of DSM-IV posttraumatic stress disorder with traumatic experience type and history in the World Health Organization World Mental Health Surveys. JAMA Psychiatry, 74(3), 270–281. 10.1001/jamapsychiatry.2016.3783 [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Logue MW, Miller MW, Wolf EJ, Huber BR, Morrison FG, Zhou Z, Zheng Y, Smith AK, Daskalakis NP, Ratanatharathorn A, Uddin M, Nieverwgelt CM, Ashley-Koch AE, Baker DG, Beckham JC, Garrett ME, Boks MP Geuze E, Grant GA, … Traumatic Stress Brain Study Group (2020). An epigenome-wide association study of posttraumatic stress disorder in U.S. veterans implicates several new DNA methylation loci. Clinical Epigenetics, 12(1), 1–14. 10.1186/s13148-020-0820-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Mansell G, Gorrie-Stone TJ, Bao Y, Kumari M, Schalkwyk LS, Mill J, & Hannon E (2019). Guidance for DNA methylation studies: Statistical insights from the Illumina EPIC array. BMC Genomics, 20(1), 1–15. 10.1186/s12864-019-5761-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. McCutcheon VV, Sartor CE, Pommer NE, Bucholz KK, Nelson EC, Madden PA, & Heath AC (2010). Age at trauma exposure and PTSD risk in young adult women. Journal of Traumatic Stress, 23(6), 811–814. 10.1002/jts.20577 [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Mehta D, Bruenig D, Carrillo-Roa T, Lawford B, Harvey W, Morris CP, Smith AK, Binder EB, Young RM, & Voisey J (2017). Genomewide DNA methylation analysis in combat veterans reveals a novel locus for PTSD. Acta Psychiatrica Scandinavica, 136(5), 493–505. 10.1111/acps.12778 [DOI] [PubMed] [Google Scholar]
  32. Mehta D, Miller O, David G, Bruenig D, & Shakespeare-Finch J (2020). A systematic review of DNA methylation and gene expression studies in posttraumatic stress disorder, posttraumatic growth, and resilience. Journal of Traumatic Stress, 33(2), 171–180. 10.1002/jts.22472 [DOI] [PubMed] [Google Scholar]
  33. Peters TJ, Buckley MJ, Statham AL, Pidsley R, Samaras K, Lord RV, Clark SJ, & Molloy PL (2015). De novo identification of differentially methylated regions in the human genome. Epigenetics & Chromatin, 8(6). 10.1186/1756-8935-8-6 [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Ritchie ME, Phipson B, Wu D, Hu Y, Law CW, Shi W, & Smyth GK (2015). limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Research, 43(7), e47. 10.1093/nar/gkv007 [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Rutten BPF, Vermetten E, Vinkers CH, Ursini G, Daskalakis NP, Pishva E, de Nijs L, Houtepen LC, Eijssen L, Jaffe AE, Kenis G, Viechtbauer W, van den Hove D Schraut KG, Lesch K−P, Kleinman JE, Hyde TM, Weinberger DR, Schalkwyk L, … Boks MPM (2018). Longitudinal analyses of the DNA methylome in deployed military servicemen identify susceptibility loci for post-traumatic stress disorder. Molecular Psychiatry, 23(5), 1145–1156. 10.1038/mp.2017.120 [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Siqveland J, Hussain A, Lindstrøm JC, Ruud T, & Hauff E (2017). Prevalence of posttraumatic stress disorder in persons with chronic pain: A meta-analysis. Frontiers in Psychiatry, 8, 164. 10.3389/fpsyt.2017.00164 [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Smyth G (2005). limma: linear models for microarray data. Springer. [Google Scholar]
  38. Sobell LC, & Sobell MB (1992). Timeline follow-back: A technique for assessing self-reported ethanol consumption. In Litten RZ & Allen JP (Eds.), Measuring alcohol consumption: Psychological and biochemical methods (pp. 41–72). Humana Press. 10.1007/978-1-4612-0357-5_3 [DOI] [Google Scholar]
  39. Steinberg AM, Brymer MJ, Decker KB, & Pynoos RS (2004). The University of California at Los Angeles Post-traumatic Stress Disorder Reaction Index. Current Psychiatry Reports, 6(2), 96–100. 10.1007/s11920-004-0048-2 [DOI] [PubMed] [Google Scholar]
  40. Storey JD, Base AJ, Dabney A, & Robinson D (2019). qvalue: Q-value estimation for false discovery rate control. R package (Version 2.10. 0. 2015) [Software program]. R Corporation. [Google Scholar]
  41. Tantoh DM, Lee KJ, Nfor ON, Liaw YC, Lin C, Chu HW, Chen PH, Hsu SY, Liu WH, Ho CC, Lung CC, Wu MF, Liaw YC, Debnath T, & Liaw YP (2019). Methylation at cg05575921 of a smoking-related gene (AHRR) in non-smoking Taiwanese adults residing in areas with different PM 2.5 concentrations. Clinical Epigenetics, 11(1), 69. 10.1186/s13148-019-0662-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Teperino R, Lempradl A, & Pospisilik JA (2013). Bridging epigenomics and complex disease: The basics. Cellular and Molecular Life Sciences, 70(9), 1609–1621. 10.1007/s00018-013-1299-z [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. Touleimat N, & Tost J (2012). Complete pipeline for Infinium Human Methylation 450K BeadChip data processing using subset quantile normalization for accurate DNA methylation estimation. Epigenomics, 4(3), 325–341. 10.2217/epi.12.21 [DOI] [PubMed] [Google Scholar]
  44. Toyokawa S, Uddin M, Koenen KC, & Galea S (2012). How does the social environment “get into the mind’? Epigenetics at the intersection of social and psychiatric epidemiology. Social Science & Medicine, 74(1), 67–74. 10.1016/j.socscimed.2011.09.036 [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Trickett PK, Noll JG, Susman EJ, Shenk CE, & Putnam FW (2010). Attenuation of cortisol across development for victims of sexual abuse. Development and Psychopathology, 22(1), 165–175. 10.1017/s0954579409990332 [DOI] [PMC free article] [PubMed] [Google Scholar]
  46. Tsai PC, & Bell JT (2015). Power and sample size estimation for epigenome-wide association scans to detect differential DNA methylation. International Journal of Epidemiology, 44(4), 1429–1441. 10.1093/ije/dyv041 [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. Uddin M, Galea S, Chang SC, Koenen KC, Goldmann E, Wildman DE, & Aiello AE (2013). Epigenetic signatures may explain the relationship between socioeconomic position and risk of mental illness: Preliminary findings from an urban community-based sample. Biodemography and Social Biology, 59(1), 68–84. 10.1080/19485565.2013.774627 [DOI] [PMC free article] [PubMed] [Google Scholar]
  48. Wright ML, Dozmorov MG, Wolen AR, Jackson-Cook C, Starkweather AR, Lyon DE, & York TP (2016). Establishing an analytic pipeline for genome-wide DNA methylation. Clinical Epigenetics, 8(1). 10.1186/s13148-016-0212-7 [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Materials
NIHMS1665855-supplement-Supplementary_Materials.pdf (1.3MB, pdf)

RESOURCES