Introduction

Maternal depression and anxiety, both during and after pregnancy, are common and a major public health problem in the United States, affecting over 1 in 10 mothers [1]. Epidemiologic studies have suggested an association between prenatal maternal depression and anxiety with adverse child outcomes such as low birthweight and preterm delivery [2, 3], as well as developmental delays and emotional and behavioral problems [4]. Low birthweight, preterm delivery, and developmental delays have been associated with changes in methylation profiles [5,6,7]. Epigenetic modifications have been proposed as possible mechanisms that may help explain the association between prenatal stress and adverse child developmental outcomes [8]. The motivation for this approach is that DNA methylation may alter gene expression in ways that influence early-infancy and later-life developmental outcomes.

Epigenetic markers, including DNA methylation, may be responsive to environmental factors throughout life, especially during in utero development [9]. With the exception of imprinted regions, the genome is demethylated prior to implantation with totipotency restored and the appropriate sex and tissue type specificity patterns reestablished throughout development [10,11,12,13,14,15,16]. Prenatal smoking [17], body mass index [18], and exposure to certain chemicals [19, 20] have been associated with changes to the infant’s epigenetic profile and the long-term health outcomes of offspring. It has thus been hypothesized that in utero exposure of the fetus to maternal depression may influence infant epigenetic profiles that alter fetal and child health and development. However, exposure to maternal depression is highly related and confounded with many potential variables such as smoking [21], maternal age [22], and maternal socioeconomic status [23]. Alterations in DNA methylation patterns in response to maternal depression may be adaptive changes that help an infant to anticipate a stressful or scarce environment; alternatively, they could be induced by pathological changes associated with medications or increased oxidative stress, or perhaps simply reflect different underlying sequence variation.

Epigenome-wide associations studies (EWAS) investigate associations between a phenotype (e.g. maternal mental health) and epigenetic variants in various tissues across the genome spanning 27,000 to a million or more CpG methylation sites [24]. While most studies of prenatal stress and infant epigenetic outcomes have focused on candidate gene methylation sites, epigenome-wide studies take an agnostic approach to identifying novel associations. Determining the influence of prenatal exposures on offspring epigenomic patterns requires complex study design and careful consideration of potential confounding and effect modification because the genetic, epigenetic, and environmental exposures over time of two linked but independent individuals must be considered. Although several reviews have attempted to synthesize related research [25,26,27], to our knowledge, ours is the first comprehensive, systematic review of EWAS of prenatal mental health and infant epigenetic profiles.

We performed a systematic review and critical assessment of EWAS studies on maternal mental health during pregnancy and the epigenetic profile of the offspring to assess whether depression, depression treatment, or anxiety in pregnancy women may influence offspring epigenetic profiles, compared with the epigenetic profiles in offspring born to mothers without depression or anxiety during pregnancy. We examined each study’s design, statistical analyses, and reporting, as well as its methodological quality and risk of bias. We assessed replication of CpG findings among studies and conducted gene enrichment analysis. Our findings underscore the importance of replication of research and study design in this area; more studies are needed to clarify the associations between maternal mental state and offspring epigenetic changes, which may influence offspring early-infancy and later-life health and developmental outcomes.

Methods

Literature search

We conducted our systematic review in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [28] and registered it prospectively in PROSPERO (registration ID number: CRD42022335595). We conducted a review of EWAS to assess the association between mothers with vs mothers without either maternal depression, maternal anxiety, or depression treatment during pregnancy and their offspring’s epigenetic profile to explore the quality of these studies and compare their findings. The exposures were maternal depression, maternal anxiety, or depression treatment during pregnancy. The comparison groups were mothers who did not experience maternal depression, maternal anxiety, or depression treatment during pregnancy. The outcomes were epigenetic profiles of the offspring. Epigenome wide association studies were eligible for inclusion if they 1) specified the epigenetic profile of the offspring as the outcome, 2) measured exposures occurring during pregnancy, and 3) were published in the English language. Studies were excluded if they 1) analyzed only candidate genes, or 2) were published only as conference papers or abstracts. The search was conducted on May 22, 2023, on PubMed, Scopus, and Embase. The following search terms were applied: “epigenome wide association study”, “DNA methylation”, “pregnancy”, “prenatal”, “depression”, “anxiety”, and “psychiatric” (detailed search strategy is available in the Supplement).

Data extraction and quality assessment

The results of the search were exported to Excel and two reviewers (ED, KSC) conducted title abstract screening full text of articles that passed screening. Risk of bias was assessed by two reviewers (ED and DR) through the Risk of Bias in Non-randomized Studies of Interventions (ROBINS-I) tool [29] and the quality of reporting was assessed using the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines [30]. From each study, we extracted data pertaining to sample size, recruitment time period, time of enrollment, country, recruitment location, ethnicity/race, age, study design, analytic method, methylation chip, tissue sample type, cell type correction, covariates, covariate data collection methods, ancestral markers, genetic interactions, genome-wide associations, CpG sites excluded from analysis, P-values for significance, replication/validation analyses, exposures, exposure measurements, timing of exposure measurement, main CpG findings, main DMR findings, and relevant genes. The significant or top CpG sites and DMRs were compared within each exposure category to find any replicated CpG sites and DMRs or overlapping DMRs. Gene enrichment of the gene ontology (GO) categories of the significant and top ranked CpG sites and DMRs from the studies was conducted using the gometh function of the missMethyl R package [31] (version 1.31.0). Gometh tests GO enrichments for inputted CpG sites by empirically calculating the probability of differential methylation as a function of the number of CpGs, which accounts for biases in the number of probes per gene on the array and for CpGs that are annotated to multiple genes [31].

We conducted separate analyses for three specific associations: prenatal maternal depression and offspring DNA methylation profile (directed acyclic graph Fig. 1A), prenatal maternal depression treatment and offspring DNA methylation profile (directed acyclic graph Fig. 1B), and prenatal maternal anxiety and offspring DNA methylation profile (directed acyclic graph Fig. 1C). Due to the direct relationship between depression and depression treatment shown in Fig. 1B and the effect medications may have on DNA methylation, depression treatment was examined as its own category. Anxiety treatment was not examined due to the lack of articles on anxiety treatment in the search results.

Fig. 1: Directed acyclic graph for prenatal maternal health exposures and offspring DNA methylation profile.
figure 1

A Directed acyclic graph for prenatal maternal depression and offspring DNA methylation profile. Many factors have been associated with prenatal maternal depression, offspring DNA methylation, or both resulting in a complex network between prenatal maternal depression and offspring DNA methylation. The arrows represent associations between two factors in which one factor influences the other. The hyphened arrows refer to a theoretical association. The purple arrows refer to associations where prenatal maternal depression influences another factor, and the black arrows refer to associations between the covariates and other factors. These covariates came from common covariates used in the studies for prenatal maternal depression and offspring DNA methylation in this review. Prenatal maternal depression can be connected to or influence offspring DNA methylation through multiple pathways. A main pathway through which prenatal maternal depression influences offspring methylation in this DAG is through preterm birth/gestational age [79, 80]. Multiple factors including maternal place of birth [21, 81], maternal age [22, 82], maternal smoking [21, 83, 84], maternal BMI [18, 85], maternal education [21, 86, 87], maternal SES [23, 88], parity [87, 89], and race/ethnicity [88, 90] influence both prenatal maternal depression and offspring DNA methylation. Though not “directly” connected in this DAG, offspring DNA methylation is influenced by maternal education through maternal smoking [87], by maternal socioeconomic status through preterm birth/gestational age [88], and by parity through maternal smoking [87] and maternal BMI [89]. Infant sex [91] and child age [91] also influence the offspring’s DNA methylation profile. B Directed acyclic graph for prenatal maternal depression treatment and offspring DNA methylation profile. Many factors have been associated with prenatal maternal depression treatment, offspring DNA methylation, or both resulting in a complex network between prenatal maternal depression and offspring DNA methylation. The arrows represent associations between two factors in which one factor influences the other. The hyphened arrows refer to a theoretical association. The arrows represent associations between two factors in which one factor influences the other. The green arrows refer to associations where prenatal maternal depression treatment influences another factor, the purple arrows refer to association where prenatal maternal depression influences another factor, and the black arrows refer to associations between the covariates and another factors. These covariates came from common covariates used in the studies for prenatal maternal depression treatment and offspring DNA methylation in this review. Like prenatal maternal depression, prenatal maternal depression treatment influences offspring DNA methylation through preterm birth/gestational age [92]. As with the previous DAG in A, maternal age [22, 82], maternal smoking [21, 83], maternal BMI [18, 85], and parity [87, 89, 93] maternal education [21, 81] influence both prenatal maternal depression and offspring DNA methylation. As a result, these factors also influence prenatal maternal depression treatment through prenatal maternal depression. Infant sex [91] and child age [91] also influence the offspring’s DNA methylation profile. C Directed acyclic graph for prenatal maternal anxiety and offspring DNA methylation profile. Many factors have been associated with prenatal maternal anxiety, offspring DNA methylation, or both. The arrows represent associations between two factors in which one factor influences the other. The blue arrows refer to associations where prenatal maternal anxiety influences another factors and the black arrows refer to associations between the covariates and another factors. The hyphened arrow refers to a theoretical association. These covariates came from common covariates used in the studies for prenatal maternal anxiety and offspring DNA methylation in this review. Prenatal maternal anxiety is connected to or influences offspring DNA methylation through multiple pathways. An important route in which prenatal maternal anxiety influences offspring DNA methylation is through preterm birth/gestational age [94]. Similarly, to A and B, except with prenatal maternal anxiety instead of prenatal maternal depression, maternal education [86, 95], maternal socioeconomic status [88, 95], maternal smoking [83, 96], maternal age [82, 87], maternal BMI [18, 85], and parity [87, 89] influence both prenatal maternal anxiety and offspring DNA methylation. Infant sex [91] also influences the offspring’s DNA methylation profile.

Full size image

Results

Included studies

We identified 1321 articles from our PubMed, Scopus, Embase search. Of these, 676 were excluded as duplicates leaving 645 for abstract screening. After removing 598 during abstract screening, we performed full text screening on the remaining 47 articles and removed an additional 27 for one of the following reasons: non-human study population, candidate genes only, postnatal exposure, not being a full article, or exposure other than maternal depression, maternal depression treatment, or maternal anxiety, The remaining 20 articles were included our analysis (Fig. 2). Several articles described analyses of more than one exposure, resulting in 16 analyses [32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47] of maternal depression, 8 analyses [32, 35, 38, 39, 44, 45, 47, 48] of maternal depression treatment, and 9 analyses [27, 32, 33, 35, 38, 41, 47, 49, 50] of maternal anxiety (Table S1). Based on STROBE and ROBINS-I criteria, we found that 13 analyses had a moderate risk of bias and 7 had a severe risk of bias (Tables S2 and S3).

Fig. 2: Preferred Reporting Items for Systematic Reviews and Meta-Analyses flow diagram.
figure 2

Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flow diagram for epigenome wide association studies about maternal mental health and infant epigenetic profile curation.

Full size image

Depression during pregnancy

Of the 16 studies that analyzed associations between maternal depression during pregnancy and the infant’s epigenetic profile, we found the risk of bias to be moderate in 12 studies and severe in 4 studies. Main factors increasing the risk of bias were not accounting for all important confounders and not accounting for selection bias. Covariates and confounders adjusted for in more than one of the included studies are shown in Fig. 1A. No statistically significant CpG site was reported from more than one study. However, one CpG site, cg25157095, was reported as a top ranked CpG site in two studies. Nine pairs of overlapping DMRs were found in at least two maternal depression studies (Table 1). The gene enrichment analysis had 62 significant pathways for study significant DMRs and 65 signficant pathways for top ranked DMRs (Table 2). The gene enrichment analysis did not produce any significant results for the CpG sites. The top ten pathways for the CpG sites are presented in Table S4.

Table 1 Overlapping differential methylated regions and repeated CpG site for maternal depression.
Full size table
Table 2 Significant pathways for DMRs.
Full size table

Depression treatment during pregnancy

Eight studies analyzed associations of maternal depression treatment during pregnancy with the infant’s epigenetic profile. We found the risk of bias to be moderate in 7 of these studies and severe in 1 study. Main factors increasing the risk of bias in these studies were not accounting for all important confounders, not accounting for selection bias, and missing data. Covariates and confounders adjusted for in more than one of the included studies are shown in Fig. 1B. No statistically significant or top-ranked CpG sites were reported from more than one study. Many studies either did not perform a DMR analysis or did not provide ranked results. Only one DMR, chr12: 56325797–56325867, was reported as a top ranked DMR. The gene enrichment analysis did not produce any significant results for the CpG sites. The top ten pathways are presented in Table S5.

Anxiety during pregnancy

Of the 9 studies that analyzed the association between maternal anxiety during pregnancy and the infant’s epigenetic profile, 7 had a moderate risk of bias while 2 had a severe risk of bias. Main factors increasing the risk of bias in these studies were not accounting for all important confounders, not accounting for selection bias, and missing data. Covariates and confounders adjusted for in more than one of the included studies are shown in Fig. 1C. No statistically significant or top-ranked CpG sites were reported from more than one study. The studies shared no overlapping DMRs. The gene enrichment analysis did not produce any significant results for the CpG sites or the DMRs. The top ten pathways are presented in Table S6.

Discussion

To our knowledge, the present work is the first systematic review of epigenome-wide association studies of prenatal maternal mental health and infant epigenetic profiles. We identified 20 EWAS studies which together reported 803 CpG sites and 19,440 DMRs in infants associated with maternal anxiety, depression, or depression treatment. Among the studies within any maternal mental health category, there was only one top ranked CpG site, cg25157095, reported in more than one study. Only nine overlapping DMRs were reported in at least two studies on maternal depression. We identified significant pathways only for maternal depression DMRs in the gene enrichment analysis. The main limitations of the studies we identified were small sample sizes, concerns about exposure misclassification and suboptimal statistical analyses that increased the risk of bias.

We found that no replicated single significant CpG sites were reported from more than one study in any of the three categories (maternal depression, maternal depression treatment, and maternal anxiety). However, one CpG site, cg25157095 was found among the top non-significant CpG sites for maternal depression in two studies [40, 47]. This CpG site is in an intron for the RIPK4 gene which has been implicated in keratinocyte differentiation, modulation of the actin cytoskeleton, and restricting intercellular adhesion [51]. Three pairs of overlapping DMRs were reported in two of 15 studies of maternal depression during pregnancy. One pair of overlapping DMRs, chr8:70378380-70378994 and chr8:70378380-70378995, is in the SULF1 gene. The SULF1 gene is involved in the regulation of multiple cellular pathways for editing heparan sulfate chains [52]. This gene has been associated with nervous system development in studies of SULF1 deficient mice, providing evidence that a non-functioning SULF1 gene is associated with impaired neurite growth and impaired long-term potentiation [53, 54], a form of synaptic plasticity [55]. Single nucleotide polymorphisms (SNPs) in this gene have been associated with cancer risk [56]. Another pair of overlapping DMRs, chr8: 143859669-143859991 and chr8: 143859369-143860092, is closest to the LYNX1 gene. Evidence has been provided through mouse models for the possible role of this gene in synaptic plasticity [57].

The chr11: 85195094-85195206 and chr11: 85195119-85195288 pair of overlapping DMRs are in the DLG2 gene. This gene encodes for the postsynaptic density 93 protein which has been thought to have roles in synaptic stability and regulation [58, 59]. Genetic variations in DLG2 have been associated with schizophrenia [60], attention deficit hyperactivity disorder [61], and bipolar disorder [62]. Another pair of overlapping DMRs were chr7:27183643-27184853 and chr7:27183133-27184522, which were significantly associated in opposite directions in the two studies. This DMR is located in HOXA-5 in the HOXA gene cluster, which is important in human development [63]. Hypermethylation in this gene has also been associated with various cancers [64] and hypermethylation in this region of the gene has also been associated with Alzheimer’s disease [65]. Another pair of overlapping DMRs, chr1:62660188-62660861 and chr1:62660038-62661010, span the L1TD1 gene. The L1TD1 gene has been connected to pluripotency maintence [66]. Higher expression of L1TD1 has also been associated with longer disease free colon cancer survival [67] and L1TD1 has been found to have higher levels of methylation in non-small cell lung cancer tissue [68]. The chr6: 151125848-151125886 and chr6: 151125729-151125904 pair of DMRs are within the PLEKHG1 gene. A SNP within this gene has been associate with white matter hyperintensities and ischemic stroke [69].

Continued research in this area is needed to determine if changes in these DMR regions result in functional changes that may correspond to the adverse outcomes seen in the epidemiological literature [2,3,4]. The overall lack of replication across studies highlights the difficulty of interpreting these types of analyses.EWAS studies are limited by the types of CpG sites included and may systematically overlook critical regions [70, 71]. These types of studies also limited by available tissue type, typically blood or saliva. It is well established that DNA methylation patterns are tissue specific and the lack of data on critical regions such as the brain is problematic [11, 70, 72]. Poor correlation of DNA methylation results measured using different Illumina platforms (e.g., the EPIC, 450k, and 27k arrays) presents problems for replication [73]. Underlying sequence variation is also of concern in EWAS studies as methylation can be a direct result of the underlying genetic sequence. Correction for population stratification and genetic variation/interaction will be critical in future studies.

Methods for determining and classifying maternal depression and maternal anxiety during pregnancy varied among our included studies, as did measurements of medication intake and dosage. The studies we considered to have a severe risk of bias all lacked appropriate adjustment for confounding and consideration of selection bias.

Maternal depression is a critical public health problem that is both under-treated and under-diagnosed, especially in minority and underserved populations. Untreated prenatal depression has been associated with detrimental health outcomes for both the pregnant woman and the baby, including an elevated risk for postpartum depression in the mother and increased infant risk for preterm birth and low birth weight [74]. Antenatal maternal anxiety and stress can impact the psychological and intellectual development of the infant [75], with some studies suggesting increased risk for emotional and cognitive problems, attentional deficit, and language delay. Prenatal anxiety and depression are also associated with increased risk for suicidality in mothers [76], with the greatest increases seen among Non-Hispanic Black, low-income, and younger individuals. Maternal mental health is tied to racial and ethnic disparities, with a higher overall prevalence of maternal depression among non-Hispanic Blacks and Hispanics compared to non-Hispanic whites [77]. Emerging evidence indicates that the COVID-19 pandemic increased the prevalence of mental health issues during pregnancy, with a meta-analysis of 37 studies suggesting that more than one in four pregnant women experienced prenatal depression and one in three experienced clinically significant anxiety [78]. Early screening and treatment for women at risk for maternal depression and anxiety may help prevent long-term adverse outcomes on maternal and infant well-being.

The EWAS studies included in this systematic review explored associations of prenatal anxiety or depression with DNA methylation patterns in offspring. Among the included studies, there was a lack of replication for a majority of the studies’ findings. However, a limitation of this study includes the potential to miss relevant articles in the initial search. Further studies of larger sample sizes are needed to identify and replicate findings and further investigate the role of maternal mental health on infant epigenetic profiles as well as take greater steps to control for confounding and address selection bias.