1. Introduction
The complex interaction between genetic risk alleles and environmental stimuli such as adverse events in early life (early-life adversity; ELA) can modulate the response to stressors later in life. Therefore, those interactions play a major role in the resilience to stress and, thus, in the development of psychopathologies such as Borderline Personality Disorder (BPD), Major Depressive Disorder (MDD) and Social Anxiety Disorder (SAD). The underlying molecular mechanisms responsible for these effects are still not fully understood. Recent evidence indicates that the interplay between environmental and genetic factors in the development of psychiatric disorders is partially mediated by epigenetic regulation [
1,
2]. Environmental stimuli and stress factors shape the epigenome, which in turn is a regulator of gene expression (reviewed in [
3]). Epigenetic alterations following ELA might, therefore, potentially lead to changes in gene expression, resulting in the dysregulation of important biological pathways, which is the basis for structural changes as a response to environmental stress.
An association between ELA as a stressor, long-term epigenetic alterations and lasting gene expression changes in peripheral tissue and the brain has been identified in several studies. For example, in baboons, early-life adversity, i.e., resource limitation in their environment, was associated with aberrant DNA methylation patterns in adults [
4]. In the human brain, Lutz et al. (2021) reported that ELA was associated with epigenetic, as well as transcriptomic, changes, indicating that immune-related and small GTPase signaling pathways are consistently impaired in the amygdala of ELA individuals [
5].
Mental disorders, for which ELA is particularly described as strong risk factor, include MDD, BPD and anxiety disorders, e.g., SAD [
6]. SAD is a psychiatric disorder characterized by severe fear of social situations and avoidance of these, and it is especially prevalent in women [
7]. MDD is one of the most commonly occurring mental diseases, with a lifetime prevalence of 15.7% in Germany, and it has a major impact on quality of life, autonomy, social integration and life expectancy [
8]. BPD is a severe psychiatric disorder associated with substantial impairment in daily life, a high risk of self-mutilation and suicidal tendencies [
9]. As mentioned before, ELA may contribute to the risk of those mental disorders by modulating the response to additional stressors on an epigenetic and/or transcriptomic level. For example, Kos et al. (2023) identified cell type-specific changes in the transcriptional states of glutamatergic and GABAergic neurons in the ventral hippocampi of ELA mice after exposure to acute social stress in adulthood [
10]. Furthermore, SAD—as a psychiatric disorder corresponding to high levels of social stress—was shown to be associated with ELA in an indirect manner by the aberrant expression of genes involved in neuronal immune signaling pathways [
11]. On a gene regulatory level, CpG methylation in
BDNF and
NR3C1 mediated ~20% of the association between childhood trauma and depressive symptoms [
12]. Additionally, several epigenetic modifications such as brain
SLC6A4 methylation or the expression of the miRNA miR-450a-2-3 were associated with the link between ELA and adolescent depression (for a review, see [
13]). Furthermore, the differential DNA methylation of, e.g.,
FKBP5 or oligodendrocyte genes (e.g.,
LINGO3 and
POU3F1) was identified as a potential molecular link between ELA and adult MDD (for a review, see [
14]). In a cross-species and cross-tissue approach, Nieratschker et al. identified the differential methylation of
MORC1 in response to ELA, for which an association with MDD was suggested via gene-set analysis from a genome-wide association study [
15]. In the blood of a female BPD cohort, Teschler et al. (2013, 2016) identified increased methylation at the CpG sites of several genes, including
APBA2 and
PRIMA1 [
16,
17]. In an epigenome-wide association study (EWAS) conducted by Arranz et al. (2021), the authors identified several epigenetic alterations potentially modulating the development of BPD following the exposure of ELA [
18]. Interestingly, comparisons between BPD patients with and without ELA identified changes in cg10888111 methylation within the gene
PXDN that were, although only nominally significant in the EWAS performed in the discovery cohort, significant when validated with a targeted EpiTYPER assay in the same cohort [
18]. Additionally, the significantly lower methylation level of
PXDN, a gene that was associated with intellectual disability and obesity before [
19], was also found in a replication cohort when comparing BPD patients with high levels of ELA to patients with low levels of ELA and healthy control individuals [
18]. Therefore, this CpG site can be assumed to be relevant for mental disorders such as BPD in the context of ELA.
The aim of this study was, therefore, to contribute to a better understanding of the biological mechanisms involved in the association between ELA and adult mental disorders by investigating whether epigenetic alterations within the gene PXDN could be a molecular link between ELA and mental disorders strongly negatively influenced by ELA (especially BPD, MDD and SAD). Therefore, we analyzed the methylation levels of the CpG site cg10888111 identified by Arranz et al. in a large cohort comprising patients suffering from either BPD, MDD or SAD with high or low levels of ELA, as well as healthy control individuals with and without the experience of ELA, respectively.
5. Discussion
In this study, we aimed to replicate the differential DNA methylation of
PXDN in BPD patients and control individuals in the context of ELA, as described in a previous study by Arranz et al. (2021) [
18]. In addition, the goal of our study was to analyze the generalizability of the findings and investigate the role of epigenetic dysregulation of
PXDN as a potential molecular link between ELA and adult mental illness not only in BPD but also other adult mental disorders associated with traumatic childhood experiences, especially MDD and SAD.
We indeed observed the differential DNA methylation of
PXDN—more specifically cg10888111, which is located in proximity to several other CpG sites in the last exon of the gene—in the context of ELA when analyzing our overall cohort, consisting of BDP, MDD and SAD patients and their respective healthy control individuals. Interestingly, the methylation alteration of cg10888111 between individuals with high and low levels of ELA was especially pronounced when considering the status of mental disorder, with significantly higher
PXDN methylation detected in mentally ill patients who experienced high levels of ELA compared to mentally ill patients with low levels of ELA, supporting the finding by Arranz et al. (2021) [
18], although in our study the effect was reversed. The
PXDN DNA methylation of healthy control individuals alone was not affected by ELA. Notably, participants suffering from a mental disorder scored significantly higher in the CTQ compared to healthy control individuals, supporting the view that ELA is a risk factor for mental disorders. In healthy control individuals where ELA was less pronounced, it can potentially be compensated for by protective factors, or there might be a dose-dependent relationship between ELA and the risk of mental disorders [
40].
When comparing the
PXDN methylation levels of the group with a diagnosed mental disorder with healthy control individuals disregarding ELA status, no difference in cg10888111 methylation levels was observed, suggesting that differential
PXDN DNA methylation is not influenced by mental disorders alone but rather by an interaction between ELA and adult psychopathology. Our finding of no significant differences in
PXDN DNA methylation between healthy control individuals with high and low levels of ELA is also in line with this hypothesis. However, fitting a quantile regression model to account for covariates (i.e., age, sex and substudy) did not reveal a significant effect of ELA, but subcohort status (SAD), on the cg10888111 methylation levels. On one hand, this could be the result of confounding effects of age and sex, although they did not correlate with the DNA methylation values. As we observed significant differences together in age between the subcohorts and between aMD and HC, as well as the subgroups with high and low levels of ELA, we cannot exclude an effect of age on the DNA methylation of
PXDN, especially as age was generally shown to be affected by DNA methylation (for a review, see [
41]) On the other hand, together with the rather small effect observed in this study, the small sample size might have led to a lack of power to conduct this analysis.
Furthermore, we observed differential overall
PXDN methylation levels between the substudies. As cg10888111 methylation was measured via pyrosequencing in the BPD and the MDD cohorts (displaying similar overall DNA methylation levels) and via the Infinium MethylationEPIC BeadChip (Illumina, San Diego, CA, USA) in the SAD cohort, using two different techniques most likely contributes to imbalances between the cohorts. Moreover, accounting for the individual blood cell type proportions was only possible in the SAD cohort using the Houseman reference data for epigenome-wide approaches [
42]. Unfortunately, the blood cell counts of the participants in the MDD and BPD substudies were not available.
However, the observed effect was mainly driven by the cohort of MDD patients and their matched healthy control individuals. Therefore, we fitted a quantile regression model with an interaction term of ELA and disorder on
PXDN methylation for this substudy, as well as for the overall cohort, which revealed no significant effect. Additionally, we fitted the model for the effect of ELA and disorder on cg10888111 methylation separately, which revealed a significant effect of ELA on the DNA methylation of
PXDN. Therefore, we cannot mathematically detect an effect of interaction between MDD and ELA on
PXDN methylation in our subcohort. Nevertheless, an association between ELA and MDD was shown previously (e.g., [
43]; for a review, see [
44]), and an indirect link between ELA and MDD that is partially regulated on an epigenetic level is still possible, as demonstrated in SAD in a previous work [
11].
A possible explanation for the effect of ELA on cg10888111 methylation, especially pronounced in the MDD cohort, may be an age effect, as participants with high levels of ELA were generally older, and the participants in our MDD cohort were older than those in the BPD and SAD cohorts. However, age was not significantly associated with
PXDN DNA methylation. Nevertheless, we hypothesize that age is a relevant factor for experiencing of ELA, which could have two possible reasons: 1. European Baby Boomers (born 1946–1964) and Generation X (born 1965–1980) grew up during times of significant social upheaval, economic instability and less stringent child protection laws. These factors contributed to higher instances of various forms of childhood trauma, including physical and emotional abuse and neglect. The increasing awareness and better implementation of child protection laws and mental health services in later years have helped to reduce such adversities for Millennials (born 1981–1996) and Generation Z (born 1997–2012) individuals. Additionally, the types of ELA changed between those generations and experiences that are more stressful for younger generations, such as digital bullying, are not well covered by the CTQ. 2. Increasing awareness of individuals and society over time, as well as a greater possibility of patients being treated, may enhance the understanding and awareness of ELA, and it could also be a contributing factor. Therefore, ELA and its perception may be more pronounced in older individuals. The usage of the CTQ itself reveals other limitations, although various studies confirm its consistency and reliability [
45,
46,
47,
48], e.g., a certain risk of retrospective bias, especially within patient groups [
49,
50]; the limited scope of other forms of adversity, such as childhood interpersonal trauma [
51], are not captured; and the lack of context in terms of details of the frequency, duration or severity of adverse experiences [
52].
Moreover, Arranz et al. (2021) [
18] observed hypomethylation upon ELA in their BPD patient cohort. We were not able to replicate these results in our study. First of all, in our overall cohort,
PXDN methylation levels were increased in individuals who experienced early-life trauma, which is in contrast to their results. As mentioned before, the MDD cohort was found to drive this reversed effect. MDD and BPD are two mental disorders that share factors such as affective instability [
53]. MDD often presents as comorbidity in BPD patients [
54]. However, these mental disorders also reveal distinct sets of symptoms that may involve distinct epigenetic patterns. Perroud et al. (2011) reported higher overall
NR3C1 exon 1F methylation levels in BPD than in MDD subjects in peripheral blood leucocytes that also correlated with the forms of ELA mentioned in our study [
55]. Moreover, whereas Perroud et al. (2013) identified significantly higher methylation levels of
BDNF in BPD subjects [
56], MDD was associated with lower [
57] and higher
BDNF methylation status [
58,
59]. These results imply that although MDD and BPD share a lot of neurobiological aberrations from healthy control individuals, at the single-gene level, epigenetic differences may be possible.
Additionally, our BPD patients did not reveal PXDN methylation upon ELA. The BPD cohort was not recruited with ELA in mind; therefore, there is no balance of individuals with and without ELA between the groups. In particular, in the BPD cohort, this led to a very small number of patients with low levels of ELA (n = 2), which prevented us from accurately replicating this study. However, these numbers support the fact that ELA is associated with the development of mental disorders in adult life.
PXDN is a heme-containing peroxidase best known for its role in external matrix formation, especially the formation of a sulfilimine bond, which cross-links collagen IV in basement membranes via the catalyzed oxidation of bromide to hypobromous acid, providing structure and mechanical stability throughout tissue development, homeostasis and wound healing [
60]. It is associated with a variety of human diseases such as obesity and intellectual disability [
19], several forms of cancer ([
61,
62,
63]; for a review, see [
60]), Autism Spectrum Disorder [
64] and, interestingly, Post-Traumatic Stress Disorder (PTSD, [
65]). Recent studies have investigated its roles in innate immunity, cardiovascular physiology and diseases and extracellular matrix formation and found that PXDN-generated reactive oxidants are important components for host defense, collagen IV synthesis in basement membrane development and tissue genesis and signaling pathways and homeostasis under physiological conditions (for a review, see [
66]).
Hence, its immune-related function is especially interesting in the context of ELA as it is known that stress and trauma impact the immune response [
67]. We identified increased CpG methylation in the last exon of the gene. DNA methylation in gene bodies is thought to be positively correlated with gene expression [
68], transcriptional elongation and alternative splicing [
69]. Gene body methylation could assist in the silencing of potentially detrimental repetitive DNA elements such as LINE1 and
Alu [
70]. Exonic DNA methylation was shown to cause C → T transition mutations, leading to disease-causing mutations in the germline and cancer-causing mutations in somatic cells [
71]. However, most evidence points towards the relevance of exon DNA methylation in alternative splicing. Exons were found to be more highly methylated than introns, and transitions in the degree of methylation occur at exon–intron boundaries, possibly suggesting a role for differential DNA methylation in transcript splicing [
72]. Manipulating DNA methylation in vivo in a site-specific manner using the deactivated endonuclease Cas9 fused to DNA methylation associated enzymes, Shayevitch et al. (2018) demonstrated that changes in the DNA methylation pattern of alternatively spliced exons, but not constitutively spliced exons or introns, altered inclusion levels [
73]. Moreover, induced inhibition of DNA methylation led to alternative splicing events in human cell cultures [
74,
75]. Furthermore, a positive correlation between methylation density and the exon expression level of intragenic exons was observed [
76]. It could, therefore, be assumed that the methylation of the analyzed CpG site could either be involved in alternative splicing of
PXDN or lead to higher
PXDN gene expression. Therefore, epigenetic dysregulation, leading to altered expression regulation, in turn leading to an increased expression as an immune-related response to early adversity, may be possible. This hypothesis is supported by the protein PXDN interaction network (
Figure S9, [
77,
78]): Interactions with proteins such as MAPK3 and STAT3, whose gene expression levels were previously associated with the molecular link of ELA and adult mental disorders in [
11], suggest the involvement of
PXDN in neuronal inflammatory signaling as a response to ELA, which may ultimately result in altered neuronal plasticity and brain activity [
79,
80].
Some mental disorders are correlated with certain subtypes of ELA such as BPD and SAD, in particular significantly correlating with emotional ELA [
11,
81,
82,
83]. We investigated the link between cg10888111 methylation levels and identified a significant correlation with emotional neglect and emotional abuse. We also observed a weak but significant correlation of cg10888111 methylation levels and sexual abuse, which was mainly seen in the SAD cohort. A link between
PXDN and emotional ELA has not been explicitly analyzed previously, but we assume that
PXDN is involved in a regulatory network that responds to the experience of emotional ELA as part of the inflammatory system, as is known for ELA in general (e.g., [
84,
85,
86]). However, an increase in inflammatory activity was associated with the occurrence of early-life sexual abuse [
87]. Several inflammatory markers such as IL-6 and TNF-α—both interaction partners of PXDN—showed up-regulation upon childhood sexual abuse [
87], which supports the aberrant regulation of
PXDN through these forms of childhood adversity.