Introduction
Exposure to stress is considered a risk factor for the emergence of psychiatric and physical disorders1,2. However, the mechanistic pathways through which these disorders occur are broad, involving hormonal, neural, immune and microbial pathways. The neuroendocrine response to stress, involving the hypothalamic-pituitary-adrenal (HPA) axis, functions, in part, to coordinate both the short- and long-term behavioral and biological consequences of stress exposure through increases in glucocorticoid release within the circulatory system1. In humans, acute stress exposure is associated with short-term elevations in cortisol, interleukin-6, C-reactive protein (CRP), and both increases and decreases in positive and negative affect have been observed within the acute stress literature3,4,5. Moreover, the impact of acute stress on mood and immune biomarkers may vary depending on the degree of hormonal reactivity3,5 which varies significantly between individuals. Both heightened and blunted HPA response to stress are predictive of negative health outcomes6,7 suggesting that understanding the biological factors that react to stressors may provide critical insights into the pathways leading to risk and resilience.
Epigenetics has emerged as a molecular mechanism that can shift in response to a broad range of environmental exposures, while also exhibiting stability across the lifespan8,9. Whereas initial investigations of epigenetic plasticity, particularly in DNA methylation, focused on early life environmental exposures, such as prenatal stress and childhood adversity10,11,12 both increases and decreases and DNA methylation have been observed to occur in adulthood. Studies in rodents suggest that dynamic changes in DNA methylation play a critical role in learning and memory13 and that global DNA methylation levels and expression of DNA methyltransferases in the brain are altered by acute stress exposure14,15. Despite this potential for plasticity in adulthood, analyses of DNA methylation in adults have typically focused on stressful or traumatic events occurring during development or have considered DNA methylation in adulthood as a predictor of HPA functioning that confers risk or resilience to subsequent stress exposure16,17. These studies generally highlight the impact of chronic stress on the epigenome and the relationship between HPA function and the epigenome18but do not explore the immediate effects following a stressor.
Though experimental studies of acute stress effects in humans are challenging, lab-based stressors such as the Trier Social Stress Test (TSST) have been used to demonstrate acute epigenetic changes. Using the TSST19 it has been demonstrated that altered plasma DNA methylation can be observed at CpG sites within the oxytocin receptor gene when contrasting pre-stress to post-stress DNA methylation levels20. However, these effects were found to be primarily driven by changes in cell types within the pre-stress compared to post-stress samples. In a genome-wide analysis of DNA extracted from plasma, CpG sites that typically show high degrees of stability were found to be altered by acute stress exposure including a site within the glutathione-disulfide reductase (GSR) gene and 10 additional CpG sites, which are reduced in DNA methylation following stress exposure21. These studies suggest that acute stress exposure in humans may alter DNA methylation levels in the periphery but also highlight the critical role that stress-induced changes in immune cell populations have in shifting the epigenome.
In the current study, we test whether genome-wide DNA methylation levels in humans can shift dynamically following exposure to an acute stressor using a within-subjects design in which we contrast the salivary epigenome using the Illumina EPIC array prior to and after stress exposure in a semi virtual version of the TSST (SV-TSST)22. Saliva is a non-invasive biological sample that has been used extensively in both hormonal and epigenetic analyses, and analyses of saliva has contributed to our understanding of the impact of stress and early life adversity on psychiatric risk23,24,25. Given that both psychological and hormonal indices of stress reactivity can be induced by an acute stressor, and evidence that this reactivity is an important consideration in predicting both acute and long-term outcomes, we also assessed whether stress reactivity was associated with altered DNA methylation. Finally, we explore the relationship between stress and epigenetic aging to ascertain whether this predictor of healthspan and mortality relates to the dynamics of stress reactivity. Though DNA methylation is considered a highly stable epigenetic modification to DNA, we predicted that dynamic changes in DNA methylation would occur in response to acute stress.
Results
Relationship between subjective stress and cortisol reactivity
We used measures of area under the curve with regards to increase (AUCi) to describe the stress reactivity of participants during the SV-TSST26 (see Fig. 1 for the SV-TSST timeline). AUCi was calculated for salivary cortisol levels and self-report psychological stress using a visual analogue scale (VAS)22. It should be noted that this operationalization of “reactivity” considers change from baseline salivary cortisol levels and self-report psychological stress during the SV-TSST and there may be other strategies for calculating AUC levels that are more restrictive in the use of pre- and post-stress sampling timepoints27. A summary of VASAUCi and CortisolAUCi, for SV-TSST participants and a small sample of participants not exposed to stress is provided in Table 1. Independent t-tests revealed a significant difference in VASAUCi (p = 0.0001) and CortisolAUCi, (p = 0.0002) between SV-TSST and no-stress subjects confirming that the SV-TSST induces both psychological and physiological stress in participants. We did not find significant correlations between VASAUCi and CortisolAUCi (p > 0.05) in males or females, nor did we find sex differences in these stress reactivity measures. The lack of sex differences in CortisolAUCi is consistent with our previous analyses of the cortisol levels in these participants, though this previous analysis indicated that females had higher self-report stress during the SV-TSST22 a finding that was not observed in the current study when utilizing VASAUCi.
Study timeline indicating biosampling and stress assessments. DNA methylation was assessed pre-stress (t1) and post-stress (t6) in all participants. Salivary cortisol was assessed at baseline (t1) and at 3 timepoints during the recovery phase (t4-t6). Psychological stress using the VAS was assessed throughout the study phases (t1-t6).
Cell type accounts for majority of variance in DNAm
Saliva samples were estimated to consist of 62.5–100% immune cells (mean: 98%, median: 100%) and 0-37.5% epithelial cells (mean: 2.0%, median: 0%), which is consistent with findings that suggest that saliva samples collected with Oragene kits may be enriched for immune cell DNA28. DNA methylation beta values were analyzed using principal component analysis before and after ComBat adjustment and revealed that principal component 1, associated with cell type, explained the majority of the variance, about 56%. Slide was associated with PC2 and explained about 6% of the variance in the data. We also applied EpiDISH deconvolution to confirm the cell type variation in the sample. As seen in Supplemental Fig. 1, immune cells (primarily neutrophils) compose the primary cell type estimate in the samples. The estimates of cell type composition were highly congruent across methods (epithelial estimates, r = 0.76, p < 0.001; leucocyte estimates, r = 0.73, p < 0.001) and cell type estimates did not shift from pre-stress to post-stress timepoints. There were significant individual differences in immune cell proportions that remained stable across time (r ≥ 0.72, p < 0.001), emphasizing the importance of using immune cell type proportions in the analytic models.
Change in DNAm from pre-stress to post-stress
Analyses of genome-wide DNAm levels as a function of time (before (t1) and after (t6) the acute stressor during the SV-TSST; see Fig. 1) revealed 8 significant CpG sites that survived FDR correction (FDR adjusted p < 0.05) and were robust to adjustment for cell type proportions (Fig. 2). Effect sizes ranged from − 0.002 to 0.023, with 3 CpGs showing hypomethylation and 5 CpGs showing hypermethylation at the post-stress timepoint (see Fig. 2). Overall, the effect sizes were small with cg13148544, located in a CpG island on chromosome 20, showing the strongest positive association with time (b = 0.023, FDR adjusted p = 0.033, Fig. 3D). cg19679989 located in a CpG island on chromosome 10 of the GATA3/FLJ45983 gene, showed the strongest negative association with time (b = -0.003, FDR adjusted p = 0.040, Fig. 3A). In addition, DNAm at the following CpG sites had a significant association with time: cg05050358 located in a CpG island of the NINL gene on chromosome 20 (b = -0.002, FDR adjusted p = 0.040; Fig. 3B), cg16897634 located in a CpG island of the BACH2 gene on chromosome 6 (b = -0.002, FDR adjusted p = 0.040; Fig. 3C), cg15064868 located in a gene body of the PHACTR1 gene on chromosome 6 (b = 0.006, FDR adjusted p = 0.041; Fig. 3E), cg00355314 on chromosome 12 (b = 0.009, FDR adjusted p = 0.040; Fig. 3F), cg18355410 located in a transcription start site (TSS) of the FAXDC2 gene on chromosome 5 (b = 0.010, FDR adjusted p = 0.033; Fig. 3G), and cg08969621 located in a TSS of the KRTAP19-3 gene on chromosome 21 (b = 0.018, FDR adjusted p = 0.040; Fig. 3H). No CpG sites related to a priori selected stress-relevant genes (NR3C1, FKBP5, HSP90, BDNF, SLC6A4, OXTR, DNMT1, and DNMT3A) survived FDR correction (FDR adjusted p > 0.05) when examining the effects of acute stress (pre-stress vs. post-stress comparison).
Stress-associated sex differences in DNAm
When the interaction between sex and time, and its association with DNAm was examined, only cg15416440 located in an open sea region of the COL4A1 gene on chromosome 13 (b = -0.041, FDR adjusted p = 0.010) survived multiple testing correction (Fig. 4). Females had an increase in DNAm while males exhibited a decrease in DNAm at this CpG site after the acute stressor.
Differential DNAm from pre-stress to post-stress during the SV-TSST. (A) Volcano plot of significant CpGs in the association between time and DNAm. Effect sizes of CpGs from linear mixed model testing association between time (t1 and t6) and DNAm. Black line represents significance threshold. Hypermethylated CpGs in orange. Hypomethylated CpGs in blue. (B) CpGs with significant difference in DNAm associated with time, between baseline (t1) and recovery period (t6) 1 h after SV-TSST challenge.
CpGs with significant difference in DNAm between pre-stress (t1) and post-stress (t6). CpGs significantly (FDR p < 0.05) associated with time: (A) cg19679989 related to the GATA3/FLJ45983 gene, (B) cg05050358 related to the NINL gene, (C) cg16897634 related to the BACH2 gene, (D) cg13148544 (UCSC RefGene unavailable), (E) cg15064868 related to the PHACTR1 gene, (F) cg00355314 (UCSC RefGene unavailable), (G) cg18355410 related to the FAXD2 gene, (H) cg08969621 related to the KRTAP19-3 gene.
CpG with a significant interaction between sex and time associated with DNAm. cg15416440 related the COL4A1 gene had a significant (FDR p < 0.05) interaction between sex and time in the relationship with DNAm. Females had an increase while males showed a decrease in DNAm after the acute stressor.
DNAm associated with subjective stress reactivity during the SV-TSST
When the association between DNAm (t6, post-stress) and VASAUCi was examined, 47 significant novel CpG sites survived FDR correction (FDR adjusted p < 0.05) and were robust to adjustment for cell type proportions (Table 2). 25 CpGs had a negative relationship with VASAUCi and 22 CpGs had a positive relationship with VASAUCi. We also examined the association between change in DNAm and VASAUCi within the 8 CpG sites (listed in Fig. 2) identified as exhibiting significant differences in post-stress compared to pre-stress DNAm levels. Change in DNAm (post-stress DNAm – pre-stress DNAm) at cg16897634 (BACH2 gene) was positively associated with VASAUCi (p < 0.05) and change in DNAm at cg19679989 (GATA3/FLJ45983 gene) was negatively associated with VASAUCi (p < 0.05). No significant CpG sites related to a priori stress-relevant genes survived multiple hypothesis testing correction (FDR adjusted p > 0.05).
DNAm associated with cortisol stress reactivity during the SV-TSST
When the association between DNAm and CortisolAUCi was examined, there were no significant CpG sites that survived multiple hypothesis testing correction (FDR adjusted p > 0.05). We also examined the association between change in DNAm and CortisolAUCi within the 8 CpG sites (listed in Fig. 2) identified as exhibiting significant differences in post-stress compared to pre-stress DNAm levels. No significant associations were found between change in DNAm (post-stress DNAm – pre-stress DNAm) and CortisolAUCi.
Association between acute stress and DNAm age
We assessed PhenoAge and PhenoAge acceleration (PhenoAgeAcc) as a measure of epigenetic age using genome-wide DNA methylation data29. PhenoAgeAcc (mean: -0.92 ± 5.8, median: -1.02) ranged from − 9.86 to 19.88. Females had a higher PhenoAge than males, which approached significance (p = 0.068) when controlling for age. There was no significant relationship between subjective stress reactivity and DNAm age when the relationship between VASAUCi and PhenoAge or PhenoAgeAcc was examined (p > 0.05). When the association between cortisol stress reactivity and DNAm age was examined, a significant relationship between CortisolAUCi and PhenoAgeAcc (b = 0.008, SE = 0.002, t = 3.165, p = 0.003) emerged when the analysis was adjusted for sex and ethnicity/race (Fig. 5). In addition, there was a significant relationship between CortisolAUCi and PhenoAge (b = 0.008, SE = 0.002, t = 3.167, p = 0.003) when the model was adjusted for sex, age and ethnicity/race but lost significance when two extreme outliers were removed. However, there was a significant relationship between CortisolAUCi and PhenoAge in males (b = 0.010, SE = 0.005, t = 2.137, p = 0.043) and females (b = 0.007, SE = 0.003, t = 2.613, p = 0.015) when the analysis included age and was stratified by sex.
The relationship between AUCi cortisol and PhenoAge acceleration stratified by sex. PhenoAge, a measure of biological aging based on genome-wide DNA methylation analyses, was positively associated with cortisol reactivity (AUCi cortisol) during acute stress exposure.
Discussion
Though plasticity in the epigenome in response to acute stress has been demonstrated experimentally in the brain using animal models14,15,30 there has been limited exploration of these effects in humans20,21. Using a within-subjects design, we were able to compare pre-stress and post-stress DNA methylation levels in individuals subjected to an experimentally induced stressor. Our findings suggest that within a 90-minute period, both hypermethylation and hypomethylation within salivary DNA can occur within individuals exposed to acute stress (Figs. 2 and 3). We also found some evidence for a differential epigenetic response to stress in males and females (Fig. 4). Whereas the SV-TSST protocol used in this study has a demonstrated impact on both psychological stress and salivary cortisol levels22 (Table 1), our analyses suggest that although psychological stress reactivity during the SV-TSST is associated with post-stress genome-wide DNA methylation levels (Table 2), cortisol reactivity during the stressor shows no robust genome-wide level associations with DNA methylation. However, cortisol reactivity during the SV-TSST is associated with an increased epigenetic aging metric that is predictive of health and longevity29. Taken together, these findings suggest that while dynamic epigenetic changes occur in response to exposure to an acute stressor, the mechanism through which these molecular changes occur may require careful consideration of the psychological reactivity of stress-exposed individuals and the interplay between HPA reactivity and the psychological experience of stress.
DNA hypomethylation and hypermethylation in response to acute stress
A primary goal of this study was to determine if DNA methylation could shift within the genome in the immediate aftermath of an acute stressor. Overall, we found 8 CpG sites that do significantly shift in methylation levels within the 90-minute timeframe of the stress exposure. Overall, the effect sizes were small with the largest positive change in DNAm between pre- and post-stress observed at a site (cg13148544) that was not annotated to a gene but was located in a CpG island on chromosome 20 and has previously been determined to overlap with histone mono-methylation (H3K4me1) which may be involved in enhancer-promotor interactions that influence transcriptional activation31,32. The largest negative change in DNAm observed between pre-stress compared to post-stress was at a site (cg19679989) in a CpG island of the GATA3/FLJ45983 gene on chromosome 10. This gene encodes for a transcription factor which modulates cytokines, T-cell proliferation, and mitochondrial biogenesis33,34,35. It has previously been found that exposure of peripheral blood mononuclear cells to dexamethasone for 24 h results in decreased expression of the GATA-3 gene36. In addition, blocking glucocorticoid signaling in the choroid plexus in mice exposed to severe stress has been found to reduce anxiety, increase T-cell trafficking, and increase Gata3 expression37. Overall, these findings suggest a functional role of GATA3 in stress-induced responses that impact immune and energy metabolism.
Several CpGs that were differentially methylated when comparing pre-stress and post-stress DNAm are within genes associated with stress and immune function. A CpG site (cg05050358), located in a CpG island of the NINL (Ninein Like) gene on chromosome 20, was hypomethylated following stress exposure. The NINL gene, which encodes the ninein-like protein, plays a role in microtubule anchoring activity38. Reduced expression of NINL has been implicated in brain atrophy in Alzheimer’s disease39 and neurodevelopmental disability associated with ciliopathies40. Our analyses also indicated that a CpG site within the BACH2 gene was hypomethylated in response to stress. The BACH2 gene encodes a transcription factor that plays a role in the adaptive immune response involving T- and B-cells and is involved in apoptosis consequent to oxidative stress41,42. Expression of BACH2 in lymphocytes has been found to decreases with age and may account for age-related immune dysfunction43.
Stress-induced hypermethylation was observed in CpG sites within the PHACTR1, FAXDC2, and KRTAP19-3 genes. The PHACTR1 gene encodes the phosphatase and actin regulator 1 protein, and its elevated expression has been found in carotid plaques compared to normal arteries possibly due to its involvement in inflammatory processes and response to oxidative stress44. In a recent study, increased Phactr4 expression was found within the hippocampus of rats induced to depression-like behavior through exposure to unpredictable chronic mild stress and these effects were reversed through hippocampal downregulation of Phactr4, likely mediated by reduced neuroinflammation and improved synaptic plasticity45. The FAXDC2 (fatty acid hydroxylase domain containing 2) gene is involved in positive regulation of protein phosphorylation and its expression in blood has been found to be reduced in heavy alcohol drinkers after acute psychological stress46. Lipid metabolism may be altered by FAXDC2 in response to stress47. The KRTAP19-3 (keratin-associated protein 19 − 3) gene has been found to be increased in expression in blood among individuals diagnosed with bipolar disorder compared to healthy controls48. KRTAP19-3 has also been found to be differentially expressed in studies of UV and smoking exposure which may be a consequence of increased oxidative stress49.
Gender differences in HPA functioning and differential effects of acute stress exposure in males and females, have been demonstrated previously, though a meta-analyses of these effects indicates a high degree of heterogeneity associated with methodological variations in the implementation of the TSST50,51. Our previous analyses of cortisol and self-report psychological stress in males and females in the SV-TSST suggest heighted psychological stress in females compared to males with no differences in stress-induced cortisol levels22. When considering interactions between sex and acute stress at a genome-wide level, we find only one CpG that shifts in DNA methylation in response to acute stress in a sex-specific manner. In a CpG site (cg15416440) within the COL4A1 (collagen, type IV, alpha 1) gene, females showed an increase while males showed a decrease in DNAm after the acute stress challenge (t6) compared to baseline (t1). This gene has previously been reported to be near a differentially hypomethylated region in children with high hair cortisol levels52. In a review of central and peripheral immune dysregulation in PTSD, ~ 50% of convergent genes were enriched for cell type markers in the brain, including COL4A1 which was enriched for endothelial cells53. The COL4A1 gene has also been identified within gene by sex analyses of sex differences in blood pressure, with female-specific epigenetic variation within this gene associated with female bias in vascular and cardiovascular dysfunction54. Our data highlight the potential for sex by stress interactions in predicting DNAm and the importance of considering both concordance and discordance in DNAm when comparing males and females.
Although we did observe significant shifts in DNA methylation in response to acute stress, we did not observe these effects within an a priori list of genes that have previously been associated with stress and HPA regulation. In addition, the CpG sites and genes we identified as exhibiting altered DNA methylation in response to acute stress did not overlap with those previously identified in studies examining DNA in blood samples from individuals who underwent the TSST protocol20,21. In the case of stress-associated candidate genes, it may be that these genes exhibit epigenetic plasticity in response to acute stress but are: (1) downstream of more immediate molecular changes and not identifiable within the timeframe of the current epigenetic analyses, (2) highly specific to certain cell types that are not abundant within the saliva samples analyzed in the current study, (3) display small changes that do not survive multiple hypothesis testing within the context of a genome-wide analyses, and (4) display altered DNA methylation within CpG sites that are not represented within the Illumina EPIC array. Although many CpG sites within each gene are included in genome-wide arrays, this methodology may not be well suited to examining critical gene regulatory regions within specific genes that may be sensitive to stress-induced epigenetic variation. There may also be critical methodological issues to consider when comparing stress-induced epigenetic changes in studies utilizing genome-wide DNAm arrays, including variables such as the method of inducing stress, sex and age of study participants, ethnicity/race, time of DNA sampling relative to stress exposure, tissue type, estimated cell subtype proportions, and batch effects. Consolidation of large genome-wide DNA methylation datasets, where these factors are annotated, may provide critical insights into the heterogeneity of epigenetic responses to stress.
Association between subjective stress vs. physiological stress reactivity and DNAm
The general model of adaptive stress responses suggests that exposure to or anticipation of stress or threat results in HPA activation with a consequence for physiological, emotional and behavioral outcomes that collectively constitute the self-perception of “being stressed”55. Maladaptive stress responses are typically characterized by a relatively heightened or blunted HPA reactivity to a stressor or a failure of negative feedback systems to regulate HPA activation, resulting in sustained elevation in HPA activation, particularly chronic elevations in glucocorticoids1,6,7. These maladaptive responses are associated with depression, anxiety, cardiovascular risk, and metabolic disorder. Individual differences in HPA function are well established as a consequence of genetic and early life developmental experiences56. Though the TSST is considered a robust paradigm for acute stress induction in humans, variation in both HPA and psychological stress reactivity in response to the TSST are well established, with both “responders” and “non-responders” typical of this task19,57. This variation illustrates the discordance that can occur between stress exposure and HPA activation. Using the SV-TSST, we similarly find individual differences in response to the stress challenge. In addition, we found that though there is an increase in both cortisol and self-report psychological stress that is apparent at a group-level, cortisol and psychological stress reactivity are not significantly correlated within participants in the study. This lack of concordance between cortisol reactivity and psychological stress reactivity may be the norm rather than the exception and reflect differences in neural processing57,58,59. It should be noted that though cortisol is a typical physiological measure of stress utilized in human studies, the adrenal hormones dehydroepiandrosterone (DHEA) and dehydroepiandrosterone sulfate (DHEA-S) have also been found to be responsive to psychological stress, may be particularly sensitive measures of acute vs. chronic stress, and be predictive of mood related outcomes60,61,62. Thus, thought we find no concordance between cortisol reactivity and psychological stress reactivity, future work could consider a broader survey of hormonal response to stress within the SV-TSST and the relationship to self-reported stress.
The lack of concordance in the association between acute stress-evoked cortisol reactivity and psychological stress reactivity is further illustrated within measures of DNA methylation. We find that post-stress genome-wide DNA methylation levels are associated with psychological stress reactivity but not cortisol reactivity in response to acute stress. Similarly, we find that significant changes in DNA methylation that occur during the acute stress protocol (post-stress compared to pre-stress) can be associated with psychological stress reactivity but not cortisol reactivity. Although increases in cortisol in the circulatory system provide a plausible mechanistic explanation for differential methylation in peripheral DNA samples (i.e. blood or saliva)63 the pathways linking psychological stress reactivity to acute DNA methylation levels in peripheral DNA are less clear. Further, the direction of this relationship would need careful consideration as variation in DNA methylation is also a predictor of psychological outcomes16,17. However, the prediction of psychological outcomes based on DNA methylation levels has typically focused on genes involved in stress regulation. These genes were included in our a priori analyses and found not to be associated with cortisol or psychological reactivity. The relationship between acute stress exposure, cortisol, psychological stress and DNA methylation is clearly complex. A compelling next step will be to understand the epigenetic pathways associated with psychological stress and how these relate to dynamic changes in DNA methylation.
Relationship between acute stress and epigenetic aging
Genome-wide DNA methylation profiling has enabled the development of “epigenetic clocks” which can be trained to predict a broad range of exposures and outcomes, particularly those associated with stress-related biological weathering29,64,65. In the current study, we utilized the PhenoAge clock, which is a DNAm clock that is predictive of all-cause mortality, healthspan, and physical and cognitive functioning29. We find that cortisol reactivity (but not psychological stress reactivity) during the SV-TSST is positively correlated with epigenetic age measured in post-stress saliva samples when accounting for chronological age. Though the causal direction of this relationship is unclear, these findings highlight the potential role of the degree of HPA activation in response to stress and aging related outcomes. Our findings suggest that the study of the HPA response to acute stress in healthy young adults may provide valuable insights into long-term aging-related outcomes in prospective study designs. Future studies will need to elucidate whether a correlation between acute stress responses and epigenetic age has biological significance for predicting health outcomes in later life. Another important consideration is how the experience of cumulative stressors interacts with acute stress responses to predict epigenetic aging18. However, these findings illustrate the discordance between the hormonal response to stress and the psychological response to stress in the prediction of epigenetic and health outcomes.
Limitations of the study
There are several limitations of the study that should be considered. The current study includes only one post-stress assessment of DNA methylation and the dynamics of epigenetic variation in response to stress are likely variable across time. The study design was focused on the within-subject analyses and only included a small sample of no-stress condition subjects. Thus, we were not powered to conduct between-group analyses beyond establishing that the SV-TSST AUCi values for cortisol and self-report stress were different between the stress and no-stress conditions. Our index of physiological stress is limited to salivary cortisol and a more expanded profiling of stress physiology may yield a more nuanced understanding of acute stress effects on DNA methylation. Our study did not find significant sex differences in cortisol or self-report stress AUCi values, though meta-analyses suggest sex-differences in both physiological and psychological responses to the TSST66. Theses characteristics of our protocol and participants may limit generalizability of the findings. Though we included the PhenoAge clock in our analyses of salivary DNA, there may be a weak relationship between within-person blood and saliva estimates of PhenoAge acceleration that limits the utility of this epigenetic clock67. There are a broad range of epigenetic clocks currently available that may be important to explore in studies of acute stress reactivity. There are also multiple alternative strategies that can be considered for conducting cell deconvolution and the primary approach used in our analyses28 though consistent with findings from alternative approaches, is calibrated to saliva samples in children. While we did not observe significant shifts in cell type proportions associated with stress, cell type proportion did account for a high proportion of variation in the DNA methylation data and varied between individuals. Careful consideration of the specific types of immune cells that create variability in within- and between-subjects DNA methylation levels remains an important issue within this work. Finally, the age range of participants in the current study is very narrow, which may impact the degree to which variation in epigenetic aging is evident and may also have implications for the generalizability of the findings.
Concluding remarks
Overall, our findings support the hypothesis that the epigenome responds dynamically to acute stress exposure and highlights epigenetic plasticity within genes related to stress and immune function. These novel findings also suggest that psychological and cortisol reactivity to stress may have distinct relationships to post-stress variation in DNA methylation with consequences for the prediction of long-term health outcomes. This work provides a foundation for exploration of the epigenetic pathways associated with psychological and physiological manifestations of stress that can contribute to the biological embedding of experiences across the lifespan. Future research exploring the epigenome at multiple post-stress timepoints may provide further insights into the plasticity and stability of stress-associated molecular changes.
Materials and methods
Participants
All participants were recruited from a large undergraduate class at the University of Texas at Austin and were required to be 18–25 years of age. Study participants (N = 32 female, N = 30 male) comprised the following race/ethnicity groups: Black/African American (2%), Hispanic White (24%), Non-Hispanic White (31%), Asian/Pacific Islander (40%) and Other/Mixed (3%). Average age was 19.63 ± 1.46 years. Respondents to a screening survey were excluded from the study if they reported taking endocrine-related medications (other than hormonal contraception), reported current substance abuse, indicated any current nicotine use, reported being pregnant, breastfeeding, or having an irregular menstrual cycle. Female participants were scheduled for the lab visit during the luteal phase of their menstrual cycle. Qualtrics was used to obtain informed consent, collect demographic data and administer study surveys. Participants who completed all components of the study received course credits, a Fitbit®, and were entered in a drawing for $20 USD. This study was approved by the Institutional Review Board at the University of Texas at Austin and all procedures were conducted in accordance with the guidelines and regulations of the University of Texas at Austin Institutional Review Board.
Acute stress exposure
The Semi-Virtual Trier Social Stress Test (SV-TSST)22an adapted version of the classic TSST, was used to induce acute stress in participants19. The SV-TSST consists of (a) subjects giving a speech with little preparation time and (b) an impromptu mental arithmetic exercise before a panel of judges with neutral/blank expressions during a live Zoom session. Phases of the SV-TSST (see Fig. 1) include a baseline period (participants watch a nature film), anticipatory period (participants are told to prepare for an impromptu speech), challenge period (participants deliver the speech and are additionally required to engage in mental arithmetic), and a recovery phase (participants watch a nature film). We have previously found this protocol effective at increasing subjective stress and salivary cortisol in the subjects included in the current study22.
Salivary cortisol
During the SV-TSST, participants provided four saliva samples for cortisol assay (see Fig. 1). Saliva was collected after the 30-minute baseline period and at three additional time points during the recovery period (at 15, 30, and 60 min after the SV-TSST). Participants were instructed to drool into a collection tube using a sterile straw until the volume reached the 2 mL mark on the cryogenic vial, while avoiding bubbles and phlegm. After collection, saliva samples were immediately stored at -80 °C until further processing. Cortisol concentration was measured using a high throughput liquid chromatography–tandem mass spectrometry assay by Dresden LabService GmbH68.
Saliva samples for DNA methylation analyses
Salivary DNA was used in the current study due to the ease of collection, minimal invasiveness, and appropriateness within the study of stress, DNA methylation, and psychiatric risk23. Two saliva samples for DNA extraction were collected from each participant; one at the beginning (t1) and end of the experiment (t6; see Fig. 1) using the Oragene-500 collection kit per the manufacturer’s instructions (DNA Genotek Inc.). Participants were instructed not to eat, drink, or use chewing gum for 30-minutes prior to providing their saliva sample. Participants were instructed to drool into the tube funnel until the saliva volume reached the black line on the tube while avoiding bubbles and phlegm. Saliva collection was monitored by a researcher to preserve the integrity of the sample. After collection, saliva samples were immediately transferred to Corning™ cryogenic vials and stored at -80 °C until DNA extraction.
Subjective stress assessment
Throughout the experimental session, participants self-reported their subjective stress using a visual analog scale (VAS) slider with end points at 0 and 100 and lower numbers indicating less stress and higher numbers indicating more stress69. Study participants were instructed to indicate on the slider the value that best represented their feelings of subjective stress. The VAS was administered after the 30-minute baseline period (t1), 7 min into the anticipatory period (t2), immediately after the acute stress induction challenge (t3), and at three sampling times during the recovery period (at 15 (t4), 30 (t5), and 60 (t6) minutes after the stress challenge), for a total of six assessments (see Fig. 1)22.
Study protocol
Details about SV-TSST study session sample collections, timeline and set-up are described elsewhere22 (see Fig. 1). At the time of scheduling, participants were assigned to the SV-TSST (N = 28 female, 27 male). As previously described, a small no-stress condition (N = 4 females, 3 males) was also included22. SV-TSST participants were exposed to the psychosocial stressor whereas no-stress participants did not undergo the stressor and instead watched a nature documentary in between saliva and VAS score collections. Because the study design is a within-subjects design, the small no-stress group was included only to confirm that presence in the lab environment during the study period did not induce psychological or physiological stress22. After placing their belongings in a separate storage room, all participants were asked to enter the participant room for the 30-minute baseline period. At this time, all communication between the participant and researcher was carried out via FaceTime (Apple Inc.) on an iPad, with the researcher keeping track of study progress using a timer. After the 30-minute baseline, participants rated their subjective stress level using the VAS and provided the two saliva samples (t1). Sixty minutes (t6) after the challenge, the sixth VAS and final two saliva samples were collected before the participant was debriefed on the study objective and released from the study.
Genome-wide DNA methylation assay
Prior to extraction, samples were thawed on ice then centrifuged at 1800 × g at 4 °C for 5 min. 1 mL of saliva was extracted using the MagMAX™ DNA Multi-Sample Ultra 2.0 Kit (Thermo Fisher Scientific Catalog Number: A36570) by following the large volume high-throughput automated DNA purification workflow on the KingFisher™ Flex System. Extracted DNA was eluted in 150 µL of elution solution and stored at − 20 °C. Salivary DNAm levels were determined using the Infinium MethylationEPIC v1.0 BeadChip kit (Illumina, Inc) at the Genomic Sequencing and Analysis Facility (GSAF) at the University of Texas at Austin. Samples with at least 500 ng of DNA first underwent sodium bisulphite conversion using the EZ-96 DNA Methylation™ Kit (Zymo Research, Cat# D5004) according to the manufacturer’s protocol. Bisulfite-converted DNA samples were then loaded onto each MethylationEPIC BeadChip slide with samples from the same participant loaded on the same slide to minimize batch effects. Samples were then processed per Illumina’s Infinium HD Methylation Assay protocol, which measures DNAm at over 850,000 CpG sites70,71. BeadChips were then scanned using the Illumina NextSeq 550 scanner which measures the fluorescence intensities and stores the data as IDAT files.
Data type and availability
These are DNA methylation data generated from a BeadChip microarray. The data are derived from DNA samples that have been treated with sodium bisulfite to induce mutagenesis and identify the location of methyl groups that are chemically bound to the DNA. These are not genome sequencing data (SNPs or whole genome). The IDAT files for this project are available at the following link: https://github.com/fchampagneUT/AcuteStressMethylation.
Pre-processing and analyses of DNA methylation levels
Processing of the EPIC BeadArray IDAT image files and data analysis was performed using RStudio, primarily with the minfi package. The data were first examined for low signal intensity. Then Illumina control metrics were evaluated using the ewastools package. Predicted sex from DNAm of the sex chromosomes was verified against survey demographic data. As a quality check, SNPs were compared between samples to verify genetic relatedness between samples. Background correction with dye-bias normalization was carried out using preprocessNoob. Beta (β) methylation values, which is the estimated proportion of cells in a sample that are methylated at a given CpG site, was then determined from the fluorescent signal intensity and used for subsequent analyses. Principal component analysis (PCA) was used to identify batch effects by linear regression analysis of the principal components and beta values as well as visual inspection of the PCA plots.
After initial quality checks, preprocessing and calculation of beta values and detection p-values, used to assess background fluorescence, were calculated and 0.689% of probes were found to be above a threshold of 0.01. In addition, the number of hybridizing beads measuring DNAm at each CpG site (nbeads) were calculated and 0.22% of probes were found to have low bead numbers (n < 4). 28,548 CpGs, for which 5% of samples were flagged for detection p-values or nbeads, were removed. All samples had fewer than 10% unreliable probes and thus no samples were filtered. 27,623 probes with a SNP in the CpG were removed. 40,050 cross-reactive probes were excluded and 4,389 gap probes were also removed. After initial quality control and filtering, there were 124 samples remaining with a total of 765,628 probes.
Saliva consists of epithelial cells and immune cells present in the oral cavity72 with each cell type contributing its own unique DNAm signature. Saliva cell proportions were estimated using the Bioconductor BeadSorted.Saliva.EPIC package developed from a reference panel of Illumina EPIC data from BeadSorted saliva cells28which was implemented with the ewastools function estimateLC. We additionally used EpiDISH, a reference based deconvolution approach that provides more detailed estimates of immune cell subtypes and has been validated for use in saliva73,74.
We targeted our analysis on autosomal chromosomes, thus excluding 12,275 probes on sex chromosomes from the analysis. The data were analyzed by PCA before and after batch adjustment to assess the variance. Sentrix ID and position-related batch effects were identified by regressing the first three principal components with possible batch covariates. Prior to batch correction, samples with missing data in key covariates were removed. ComBat, a supervised empirical Bayes batch-correction method as implemented in the Bioconductor sva package75 was used to adjust for slide batch effects using a model matrix which controlled for sex, age, time, ethnicity/race, condition and stress-related covariates.
Age-related DNA methylation calculation
The PhenoAge clock was developed from the analysis of DNAm to predict aging outcomes, including all-cause mortality, physical functioning and healthspan29. Whereas PhenoAge was developed using blood samples, it has been shown to be strongly associated with chronological age in other tissue types, including saliva29. The PhenoAge of samples in this study was estimated using the methyAge function from the ENmix R package version 1.36.08.
Data analysis
All statistical analyses were performed in RStudio. Figures were made using ggplot2 (package version 3.4.1). Linear regression models implemented the lm function, from the stats package version 4.0.3. Descriptive statistics were reported as means ± SD, unless otherwise specified, with p-values less than 0.05 reported as statistically significant. Pearson correlations were carried out using the cor.test function. Ethnicity and race (Non-Hispanic Asian/Pacific Islander, Non-Hispanic Black/African American, Hispanic White, Non-Hispanic Other/Mixed, and Non-Hispanic White) were combined into a single variable based on precedence76 which resulted in the following groups: Asian, Black, Hispanic, White and Other/Mixed. Given the small sample size of the Black and Other/Mixed ethnicity/race groups in our study, we combined these into a single group to improve statistical power.
Stress reactivity measures
Area under the curve with respect to increase (AUCi) was calculated using cortisol concentrations as an index of cortisol stress reactivity (CortisolAUCi)26. CortisolAUCi calculations included the 4 sampling timepoints at which cortisol was assessed and represents the total change in cortisol levels during the TSST with reference to the initial cortisol levels. In addition, AUCi was calculated using VAS scores (VASAUCi) as an index of subjective stress reactivity26. VASAUCi calculations included the 6 sampling timepoints at which self-report stress was assessed and represents the total change in self-report stress levels during the TSST with reference to the initial self-report stress levels. Correlations were carried out between VASAUCi and CortisolAUCi.
Stress and DNA methylation
The association between post-stress DNAm (see Fig. 1) and stress variables of interest was examined with linear regression models by applying an empirical Bayes method using the limma package version 3.46.0. For the analysis of repeat measures of DNAm with time (pre-stress and post-stress), non-independence was accounted for by fitting linear mixed-effects regression models implemented with the lme4 package version 1.1–31 and lmerTest version 3.1-3. Significance for the genome-wide DNAm analysis was evaluated at a false discovery rate (FDR) adjusted p-value < 0.05 using the Benjamini-Hochberg method. Version GRCh37/hg19 from the EPIC annotation manifest was used to annotate significant sites. Significance for DNAm was tested genome-wide, including at CpG sites within the following a priori stress-related genes of interest: NR3C1 (82 CpGs), FKBP5 (49 CpGs), HSP90 (38 CpGs), BDNF (103 CpGs), SLC6A4 (31 CpGs), OXTR (22 CpGs), DNMT1 (48 CpGs), and DNMT3A (112 CpGs). Change in DNAm associated with time was estimated as the change in average beta-values over time. The association between time and DNAm was tested using linear mixed effects models of the ComBat adjusted data that were fitted to beta values for each CpG site regressed on time. Samples with missing data in key covariates were removed prior to batch correction, leaving 124 samples.
Several analyses were conducted to determine the relationship between stress and DNA methylation. To test the association between DNAm and time (pre-stress and post-stress), DNAm was modeled as a function of time. The model controlled for the fixed effects of sex, age, ethnicity/race, time, array and cell proportions. Participant ID was included as a random intercept term to control for repeated measures. In addition, a sex interaction and main effect was tested a priori in the relationship between time and DNAm. To test the association between DNAm and cortisol reactivity during the SV-TSST, post-stress DNAm was modeled as a function of CortisolAUCi, controlling for age, sex, ethnicity/race, array and cell proportions. To test the association between DNAm and subjective stress, post-stress DNAm was modeled as a function of VASAUCi, controlling for age, sex, ethnicity/race, array and cell proportions. We used mixed effects models to examine the association between change in DNAm (post-stress DNAm – pre-stress DNAm) and both VASAUCi and CortisolAUCi. Epigenetic age acceleration (PhenoAgeAcc) was calculated by regressing PhenoAge on chronological age and calculating the residuals. Analyses with epigenetic age at the post-stress timepoint were carried using linear regression models, which tested the relationships between PhenoAge and sex, between PhenoAge and VASAUCi, and between PhenoAge and CortisolAUCi.
Data availability
The raw epigenomic data files and datasets generated and analyzed during the current study are available at the following link: https://github.com/fchampagneUT/AcuteStressMethylation.
References
McEwen, B. S. Central effects of stress hormones in health and disease: Understanding the protective and damaging effects of stress and stress mediators. Eur. J. Pharmacol. 583, 174–185 (2008).
Kivimäki, M. & Steptoe, A. Effects of stress on the development and progression of cardiovascular disease. Nat. Rev. Cardiol. 15, 215–229 (2018).
Man, I. S. C. et al. Multi-systemic evaluation of biological and emotional responses to the Trier social stress test: A meta-analysis and systematic review. Front. Neuroendocrinol. 68, 101050 (2023).
Izawa, S. et al. An increase in salivary interleukin-6 level following acute psychosocial stress and its biological correlates in healthy young adults. Biol. Psychol. 94, 249–254 (2013).
Kimura, K. et al. The biological effects of acute psychosocial stress on delay discounting. Psychoneuroendocrinology 38, 2300–2308 (2013).
Turner, A. I. et al. Psychological stress reactivity and future health and disease outcomes: A systematic review of prospective evidence. Psychoneuroendocrinology 114, 104599 (2020).
Carroll, D., Ginty, A. T., Whittaker, A. C., Lovallo, W. R. & de Rooij, S. R. The behavioural, cognitive, and neural corollaries of blunted cardiovascular and cortisol reactions to acute psychological stress. Neurosci. Biobehav Rev. 77, 74–86 (2017).
Kanherkar, R. R., Bhatia-Dey, N. & Csoka, A. B. Epigenetics across the human lifespan. Front Cell. Dev. Biol 2, (2014).
Champagne, F. A. Early adversity and developmental outcomes: interaction between genetics, epigenetics, and social experiences across the life span. Perspect. Psychol. Sci. 5, 564–574 (2010).
Weaver, I. C. G. et al. Epigenetic programming by maternal behavior. Nat. Neurosci. 7, 847–854 (2004).
Jensen Peña, C., Monk, C. & Champagne, F. A. Epigenetic effects of prenatal stress on 11β-hydroxysteroid dehydrogenase-2 in the placenta and fetal brain. PloS One. 7, e39791 (2012).
Roth, T. L., Lubin, F. D., Funk, A. J. & Sweatt, J. D. Lasting epigenetic influence of early-life adversity on the BDNF gene. Biol. Psychiatry. 65, 760–769 (2009).
Morris, M. J. & Monteggia, L. M. Role of DNA methylation and the DNA methyltransferases in learning and memory. Dialogues Clin. Neurosci. 16, 359–371 (2014).
Spinieli, R. L. et al. Acute restraint stress regulates brain DNMT3a and promotes defensive behaviors in male rats. Neurosci. Lett. 820, 137589 (2024).
Rodrigues, G. M. et al. Acute stress affects the global DNA methylation profile in rat brain: modulation by physical exercise. Behav. Brain Res. 279, 123–128 (2015).
Cecil, C. A. M., Zhang, Y. & Nolte, T. Childhood maltreatment and DNA methylation: A systematic review. Neurosci. Biobehav Rev. 112, 392–409 (2020).
Houtepen, L. C. et al. Genome-wide DNA methylation levels and altered cortisol stress reactivity following childhood trauma in humans. Nat. Commun. 7, 10967 (2016).
Harvanek, Z. M., Fogelman, N., Xu, K. & Sinha, R. Psychological and biological resilience modulates the effects of stress on epigenetic aging. Transl Psychiatry. 11, 601 (2021).
Kirschbaum, C., Pirke, K. M. & Hellhammer, D. H. The ’trier social stress Test’–a tool for investigating Psychobiological stress responses in a laboratory setting. Neuropsychobiology 28, 76–81 (1993).
Unternaehrer, E. et al. Dynamic changes in DNA methylation of stress-associated genes (OXTR, BDNF) after acute psychosocial stress. Transl Psychiatry. 2, e150–e150 (2012).
Apsley, A. T. et al. Biological stability of DNA methylation measurements over varying intervals of time and in the presence of acute stress. Epigenetics 18, 2230686 (2023).
Miller, M. et al. A semi-virtual trier social stress test (SV-TSST). Psychoneuroendocrinology (2025).
Smith, A. K. et al. DNA extracted from saliva for methylation studies of psychiatric traits: evidence tissue specificity and relatedness to brain. Am. J. Med. Genet. Part. B Neuropsychiatr Genet. Off Publ Int. Soc. Psychiatr Genet. 168B, 36–44 (2015).
Hellhammer, D. H., Wüst, S. & Kudielka, B. M. Salivary cortisol as a biomarker in stress research. Psychoneuroendocrinology 34, 163–171 (2009).
Papale, L. A., Seltzer, L. J., Madrid, A., Pollak, S. D. & Alisch, R. S. Differentially methylated genes in saliva are linked to childhood stress. Sci. Rep. 8, 10785 (2018).
Pruessner, J. C., Kirschbaum, C., Meinlschmid, G. & Hellhammer, D. H. Two formulas for computation of the area under the curve represent measures of total hormone concentration versus time-dependent change. Psychoneuroendocrinology 28, 916–931 (2003).
Hellhammer, J. & Schubert, M. The physiological response to Trier social stress test relates to subjective measures of stress during but not before or after the test. Psychoneuroendocrinology 37, 119–124 (2012).
Middleton, L. Y. M. et al. Saliva cell type DNA methylation reference panel for epidemiological studies in children. Epigenetics 17, 161–177 (2022).
Levine, M. E. et al. An epigenetic biomarker of aging for lifespan and healthspan. Aging 10, 573–591 (2018).
Witzmann, S. R., Turner, J. D., Mériaux, S. B., Meijer, O. C. & Muller, C. P. Epigenetic regulation of the glucocorticoid receptor promoter 17 in adult rats. Epigenetics 7, 1290–1301 (2012).
Kubo, N. et al. H3K4me1 facilitates promoter-enhancer interactions and gene activation during embryonic stem cell differentiation. Mol. Cell. 84, 1742–1752e5 (2024).
Grundberg, E. et al. Global analysis of DNA methylation variation in adipose tissue from twins reveals links to disease-associated variants in distal regulatory elements. Am. J. Hum. Genet. 93, 876–890 (2013).
Tai, T. S., Pai, S. Y. & Ho, I. C. GATA-3 regulates the homeostasis and activation of CD8 + T cells. J. Immunol. Baltim. Md. 1950. 190, 428–437 (2013).
Callender, L. A. et al. GATA3 induces mitochondrial biogenesis in primary human CD4 + T cells during DNA damage. Nat. Commun. 12, 3379 (2021).
Das, A. et al. Effector/memory CD4 T cells making either Th1 or Th2 cytokines commonly co-express T-bet and GATA-3. PLoS ONE. 12, e0185932 (2017).
Xiang, L. & Marshall, G. D. Glucocorticoid receptor BclI polymorphism associates with Immunomodulatory response to stress hormone in human peripheral blood mononuclear cells. Int. J. Immunogenet. 40, 222–229 (2013).
Kertser, A. et al. Corticosteroid signaling at the brain-immune interface impedes coping with severe psychological stress. Sci. Adv. 5, eaav4111 (2019).
Tillery, M. M. L., Zheng, C., Zheng, Y. & Megraw, T. L. Ninein domains required for its localization, association with partners dynein and ensconsin, and microtubule organization. Mol Biol. Cell Mbc E. https://doi.org/10.1091/mbc.E23-06-0245 (2024). 23-06-0245.
Jiang, S., Yang, W., Qiu, Y., Chen, H. Z. & Alzheimer’s Disease Neuroimaging Initiative (ADNI). Identification of novel quantitative traits-associated susceptibility loci for APOE ε 4 non-carriers of alzheimer’s disease. Curr. Alzheimer Res. 12, 218–227 (2015).
Parisi, M. A. The molecular genetics of Joubert syndrome and related ciliopathies: the challenges of genetic and phenotypic heterogeneity. Transl Sci. Rare Dis 4, 25–49 .
Liu, G. & Liu, F. Bach2: A key regulator in Th2-related immune cells and Th2 immune response. J. Immunol. Res. 2814510 (2022). (2022).
Muto, A. et al. Activation of Maf/AP-1 repressor Bach2 by oxidative stress promotes apoptosis and its interaction with promyelocytic leukemia nuclear Bodies *. J. Biol. Chem. 277, 20724–20733 (2002).
Chi, V. L. D. et al. Age-related changes in the BACH2 and PRDM1 genes in lymphocytes from healthy donors and chronic lymphocytic leukemia patients. BMC Cancer. 19, 81 (2019).
Zhang, Z., Jiang, F., Zeng, L., Wang, X. & Tu, S. PHACTR1 regulates oxidative stress and inflammation to coronary artery endothelial cells via interaction with NF-κB/p65. Atherosclerosis 278, 180–189 (2018).
Li, Y. et al. Targeting Phactr4 to rescue chronic stress-induced depression-like behavior in rats via regulating neuroinflammation and neuroplasticity. Int. J. Biol. Macromol. 273, 132854 (2024).
Beech, R. D. et al. Stress-related alcohol consumption in heavy drinkers correlates with expression of miR-10a, miR-21 and components of the TAR-RNA binding protein (TRBP)-associated complex. Alcohol Clin. Exp. Res. 38, 2743–2753 (2014).
Farhadi, A. et al. Identification of key immune and stress related genes and pathways by comparative analysis of the gene expression profile under multiple environmental stressors in Pacific white shrimp (Litopenaeus vannamei). Fish. Shellfish Immunol. 135, 108695 (2023).
Hess, J. L. et al. Transcriptomic abnormalities in peripheral blood in bipolar disorder, and discrimination of the major psychoses. Schizophr Res. 217, 124–135 (2020).
Pease, L. I., Wordsworth, J. & Shanley, D. Transcriptomic and epigenetic assessment of ageing female skin and fibroblasts identifies age related reduced oxidative phosphorylation is exacerbated by smoking. 08.16.504111 Preprint at (2022). https://doi.org/10.1101/2022.08.16.504111 (2022).
Liu, J. J. W. et al. Sex differences in salivary cortisol reactivity to the Trier social stress test (TSST): A meta-analysis. Psychoneuroendocrinology 82, 26–37 (2017).
Kudielka, B. M. & Kirschbaum, C. Sex differences in HPA axis responses to stress: a review. Biol. Psychol. 69, 113–132 (2005).
Nätt, D., Johansson, I., Faresjö, T., Ludvigsson, J. & Thorsell, A. High cortisol in 5-year-old children causes loss of DNA methylation in SINE retrotransposons: a possible role for ZNF263 in stress-related diseases. Clin. Epigenetics. 7, 91 (2015).
Núñez-Rios, D. L. et al. Central and peripheral immune dysregulation in posttraumatic stress disorder: convergent Multi-Omics evidence. Biomedicines 10, 1107 (2022).
Yang, M. L. et al. Sex-specific genetic architecture of blood pressure. Nat. Med. 30, 818–828 (2024).
Fink, G. Stress Science: Neuroendocrinology (Academic, 2010).
Ellis, B. J., Jackson, J. J. & Boyce, W. T. The stress response systems: universality and adaptive individual differences. Dev. Rev. 26, 175–212 (2006).
Dimitrov, A. et al. Differences in neural recovery from acute stress between cortisol responders and Non-responders. Front Psychiatry 9, (2018).
Campbell, J. & Ehlert, U. Acute psychosocial stress: does the emotional stress response correspond with physiological responses? Psychoneuroendocrinology 37, 1111–1134 (2012).
Dalile, B., La Torre, D., Verbeke, K., Van Oudenhove, L. & Vervliet, B. When the Mind says one thing, but the HPA axis says another: lack of coherence between subjective and neuroendocrine stress response trajectories in healthy men. Psychoneuroendocrinology 139, 105692 (2022).
Morgan, C. A. Relationships among plasma dehydroepiandrosterone sulfate and cortisollevels, symptoms of dissociation, and objective performance in humans exposedto acute stress. Arch. Gen. Psychiatry. 61, 819–825 (2004).
Lennartsson, A. K., Kushnir, M. M., Bergquist, J. & Jonsdottir, I. H. DHEA and DHEA-S response to acute psychosocial stress in healthy men and women. Biol. Psychol. 90, 143–149 (2012).
Jiang, X. et al. Attenuated DHEA and DHEA-S response to acute psychosocial stress in individuals with depressive disorders. J. Affect. Disord. 215, 118–124 (2017).
Braun, P. R. et al. Genome-wide DNA methylation investigation of glucocorticoid exposure within buccal samples. Psychiatry Clin. Neurosci. 73, 323–330 (2019).
Ryan, C. P. Epigenetic clocks: theory and applications in human biology. Am. J. Hum. Biol. 33, e23488 (2021).
Oblak, L., van der Zaag, J., Higgins-Chen, A. T., Levine, M. E. & Boks, M. P. A systematic review of biological, social and environmental factors associated with epigenetic clock acceleration. Ageing Res. Rev. 69, 101348 (2021).
Gu, H., Ma, X., Zhao, J. & Liu, C. A meta-analysis of salivary cortisol responses in the Trier social stress test to evaluate the effects of speech topics, sex, and sample size. Compr. Psychoneuroendocrinology. 10, 100125 (2022).
Zarandooz, S. & Raffington, L. Applying blood-derived epigenetic algorithms to saliva: cross-tissue similarity of DNA-methylation indices of aging, physiology, and cognition. Clin. Epigenetics. 17, 61 (2025).
Gao, W., Stalder, T. & Kirschbaum, C. Quantitative analysis of estradiol and six other steroid hormones in human saliva using a high throughput liquid chromatography–tandem mass spectrometry assay. Talanta 143, 353–358 (2015).
Barré, R., Brunel, G., Barthet, P. & Laurencin-Dalicieux, S. The Visual Analogue scale: An easy and reliable way of assessing perceived stress. 1, (2017).
Moran, S., Arribas, C. & Esteller, M. Validation of a DNA methylation microarray for 850,000 CpG sites of the human genome enriched in enhancer sequences. Epigenomics 8, 389–399 (2016).
Pidsley, R. et al. Critical evaluation of the illumina methylationepic BeadChip microarray for whole-genome DNA methylation profiling. Genome Biol. 17, 208 (2016).
Zheng, S. C. et al. A novel cell-type Deconvolution algorithm reveals substantial contamination by immune cells in saliva, buccal and cervix. Epigenomics 10, 925–940 (2018).
Wong, Y. T. et al. A comparison of epithelial cell content of oral samples estimated using cytology and DNA methylation. Epigenetics 17, 327–334 (2022).
Teschendorff, A. E., Breeze, C. E., Zheng, S. C. & Beck, S. A comparison of reference-based algorithms for correcting cell-type heterogeneity in Epigenome-Wide association studies. BMC Bioinform. 18, 105 (2017).
Leek, J. T., Johnson, W. E., Parker, H. S., Jaffe, A. E. & Storey, J. D. The Sva package for removing batch effects and other unwanted variation in high-throughput experiments. Bioinforma Oxf. Engl. 28, 882–883 (2012).
Goosby, B. J. & Walsemann, K. M. School racial composition and race/ethnic differences in early adulthood health. Health Place. 18, 296–304 (2012).
Acknowledgements
This work was supported by Whole Communities—Whole Health, a research grand challenge at the University of Texas at Austin, United States.
Ethics declarations
Competing interests
The authors declare no competing interests.
Additional information
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Electronic supplementary material
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.
About this article
Cite this article
Miller, M., McAfee, C., Brown, R.D. et al. Impact of acute stress exposure on genome-wide DNA methylation. Sci Rep 15, 23931 (2025). https://doi.org/10.1038/s41598-025-09299-y
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41598-025-09299-y