Most of the maternal stress studies in the past have been limited to targeted DNA methylation analyses in candidate genes (29, 30). It is only in the recent years that epigenome wide studies of DNA methylation have gained popularity allowing to evaluate locus-specific methylation across the entire genome(31–34). These different approaches have been recently summarized in two reviews and a meta-analysis (18, 35, 36). When analyzing these studies, a general conclusion on what type of epigenetic signature is observed in prenatally stressed infants is difficult to draw since many methodological differences are still observed in terms of the type and timing of the prenatal insult studied, the age of the child and the tissue employed to detect DNA methylation. This makes comparison among studies very difficult leading to inconclusive evidence on the association between PS and DNA methylation in the neonate. To bridge this gap, we have examined the association between psychological, molecular and biophysical maternal-fetal stress measures and the genome wide methylation profile in newborn saliva. Our findings validate the hypothesis that PS biomarkers are associated with epigenome–wide DNA methylation in newborn saliva across multiple CpG sites, in particular, those relevant to neuronal, immune and endocrine homeostasis.
Maternal stress measures and DNA methylation patterns
In our study, PDQ, but not PSS, and cortisol showed a significant association with five CpG sites. Out of these five CpGs, three were annotated to YAP1, TOMM20 and CSMD1; two CpGs were not annotated to any gene but lay within 50kb of SSBP4, SCAMP1 and LRRC25. We discuss the functional implications of these associations in the following paragraphs.
YAP1 and its related protein WWdomain- containing transcription regulator 1 (WWTR1; also known as TAZ) (YAP/TAZ) are the main effectors of the Hippo signaling pathway (37). This evolutionarily conserved signaling cascade regulates cell proliferation, stemness, organ size control, and regeneration. Its dysregulation has been associated to multiple forms of cancers, the immunity response and cardiovascular diseases (37, 38). Although widely expressed in several tissues, YAP is selectively expressed in astrocytes and neural stem cells in the mouse developing brain and its deletion causes reactive astrogliosis and astrocyte-driven microglial activation (39). Moreover, Passaro et al (40),demonstrated that the transient downregulation of YAP in mouse embryonic stem cells disrupts cellular homeostasis altering the ability to differentiate properly. In our study the hypermethylated CpG cg06542869 annotated to YAP1 is associated with specific pregnancy worries (PDQ score). The functional consequence of the hypermethylation of one single CpG site in the open sea of the YAP1 gene is highly speculative without evaluating the translated protein. However, it has been demonstrated that the methylation of one single CpG can impact on the methylation levels of neighboring CpG sites (41). Assuming that hypermethylation is generally associated with transcriptional silencing of genes, the modification of methylation status of the YAP1 gene might potentially lead to alterations in cell proliferation, cell differentiation and astrogliosis. In fact, the network analysis of the protein encoded by YAP1 using STRING-db showed an interaction with several proteins of the Hippo signaling pathway such as the TEAD family of transcription proteins. The phosphorylation and inhibition of YAP/TAZ activate the Hippo pathway limiting tissue growth and cell proliferation. Upon dephosphorylation, YAP/TAZ translocate to the nucleus, binding to TEAD and inducing transcriptional programs related to cell proliferation, survival, and migration (37).
TOMM20 (Translocase of outer mitochondrial membrane 20) is involved in glucose/ energy metabolism and deubiquitination. Together with TOMM22 functions as a transit peptide receptor at the surface of the mitochondrial outer membrane and facilitates the movement of preproteins (42, 43). Diseases associated with TOMM20 include Optic Atrophy 1 and 11. Our results show that the hypermethylated CpG site cg25252839 is associated with cortisol levels and annotates to TOMM20.
The CSMD1 gene has been proposed to have brain specificity since it encodes a cell adhesion molecule highly expressed in membrane associated proteins in the CNS, with almost no detection in other tissues (44). The CSMD1 protein is related to immune function playing a crucial role in regulating complement activation and inflammation in the developing brain (44, 45) and may also play a role in growth cone function (46). The CSMD1 protein is predominantly expressed in neurons mainly in the cerebral cortex and the hippocampus and has been involved in brain circuits development, neurotransmission, axon guidance, regeneration and plasticity (44). CSMD1 protein coding gene has been previously associated to Autism Spectrum Disorders (ASD) (47, 48). Corroborating the above statement, CSMD1 also appears on the SFARI database listing genes associated with ASD. It scored as level 3, meaning there is suggestive evidence from significant but non-replicated association studies. Moreover, CSMD1 has been associated to posttraumatic stress disorder (49, 50), schizophrenia (44, 45, 51, 52), and bipolar disorders (53).
In our study, we found that the hypo-methylated CpG cg05306225 annotates for the gene CSMD1 and is associated with high maternal cortisol levels. Although it is difficult to predict the functional consequences of this single site hypomethylation as mentioned above, it is interesting to observe that the probable destabilization of the methylation status of flanking CpGs mentioned before, is in a gene with high brain specificity and associated with several neuropsychiatric disorders. In particular, the association of this gene with ASD refers back to several reports showing that the risk for ASD is linked to PS (11, 54) .
The two other CpGs that were significantly associated with cortisol levels but are not annotated to any gene are cg11409463 and cg20905655, both hypermethylated. The nearest gene to the CpG site cg11409463 is SCAMP1 whose protein is involved in secretion and transportation. Diseases associated with this gene include Childhood Kidney Cell Carcinoma and Branchiootorenal Syndrome 1. This same CpG site overlaps with several transcription factor binding sites from the AP-1 family (the Jun, the Fos and ATF-2 subfamily) such as JUNB, FOS, SETDB1, ATF3, CBX3, TRIM28, ZNF143. The AP-1 family is responsible for cell growth, differentiation (55) and apoptosis (56). The nearest genes to CpG site cg20905655 are SSBP4 and LRRC25, the latter related to autophagic degradation. So far, not much is known about the functional role of SSB4 and its relation to stress yet.
Of interest, cortisol-associated methylation disbalances in several genes found in neonatal saliva suggest that the transplacental barrier might be impaired and abnormally permeable to steroid hormones. In fact, it has been described that the metabolizing enzymes that lay within trophoblasts and protect the fetus from overexposure to glucocorticoids, are sensitive to maternal stress (57, 58). For example, the glucocorticoid-inactivating enzyme, 11β-hydroxysteroid dehydrogenase type-2 (11βHSD2), showed a reduced placental expression in relation to maternal anxiety and depressed mood in humans (59, 60). The reduced placental expression of 11βHSD2 will potentially lead to a fetal glucocorticoid overexposure affecting developmental events such as fetal growth restriction, altered HPA axis development, impaired offspring brain function, permanent changes in the expression of specific transcription factors and early development of proliferative neural precursors (57, 61). Our observation that the newborn saliva shows cortisol associated epigenetic changes in genes related to energy metabolism, cell differentiation and function of the developing brain might be highlighting that one of the underlying mechanisms linking maternal stress with childhood outcomes is through transplacental mediated methylation disbalances in specific genes, among other mechanisms, such as transcriptional regulation of placental gene expression as suggested by Aushev et al (62).
To expand the search of epigenetic signatures associated with stress measures during pregnancy we considered DMRs. Two DMRs were detected, one associated to PDQ (DAXX) and the other to cortisol (ARL4D). DAXX gene encodes for a protein that resides in multiple locations in the nucleus and cytoplasm. Pathways related to Daxx are apoptosis and survival caspase cascade as well as TGF-β signalling pathways (63). Diseases associated with DAXX include Gastric neuroendocrine neoplasm, intellectual disability and Alpha-Thalassemia. Interestingly, ATRX gene which has been previously linked with ASD, interacts with DAXX in histone chaperone complex and influences DNA methylation (64–66). Moreover, DAXX is known to be an extended Class II, non-antigen binding HLA (human leukocyte antigen) gene associated with autoimmune diseases that interacts with death receptor Fas related to ASD (67). ARL4D belongs to ADP-ribosylation factors (ARFs), members of the Ras family of small GTPases, involved in membrane transport, membrane lipid modifications and maintenance of organelle integrity (68). Interestingly, the transcription of Arl4d was found to be consistently regulated by glucocorticoids such as cortisol (69). So far, not much is known about its function but it has been shown that Arl4D is involved in neurite growth (70), adipogenesis (71) and actin remodeling (72). In adult mice, Arl4d is expressed in neocortical layer 1 and hippocampus, mostly in cortical interneurons (CIN), whose loss or alteration have been related to neurological disorders such as autism, schizophrenia, and epilepsy (73). Interestingly, both DMRs are directly or indirectly related to neurological disorders such as ASD. To the best of our knowledge, this is the first report showing significant DMRs in the PS context in newborn saliva samples. Previously Drzymalla et al (74), have identified DMRs related to maternal stress but using cord blood.
Previous studies employing EPIC array on neonatal tissues in association with maternal stress and/or anxiety are limited to one study by Kallak et al,(75). These authors investigated DNA methylation in cord blood of newborns exposed to maternal depression and anxiety. They found two DMPs: one upstream of the ATP Binding Cassette Subfamily F Member 1 gene (ABCF1) and the other upstream of Homo sapiens integrator complex subunit 10 gene (INTS10). Although the maternal stress model is different from ours, it is interesting to note that ABCF1 was previously associated with ASD in a multi-omics data analysis(76). Other comparable studies employing Illumina Infinium 450 BeadChip found mismatching results when studying DNA methylation in infant tissues in relation to maternal stress. Rijlaarsdam et al.,(77) showed no associations between PS exposure and neonatal cord blood DNA methylation, whereas Wikenius et al.(78) studying maternal depressive symptoms, found no significant association with 6 weeks infant´s saliva DNA methylation. In contrast, Non et al.(33) reported the identification of CpGs located in a cluster of genes related to transcription, translation and cell division processes in cord blood of neonates exposed to non-medicated depression or anxiety.
To summarize these results, we conclude that these genes have been related to several regulatory processes of tissue and cellular homeostasis that, when disturbed, can elicit a stress response (79). Moreover, dysregulation of the expression of CSMD1 and YAP1 have been related to disorders of the immune system as well as of the central and autonomic nervous systems.
Biophysical signature of chronic stress in mother-fetus dyad and DNA methylation
No significant CpG sites were observed in association with FSI. This may be either due to insufficient study power or reflect an underlying mechanism. Namely, it is possible that regardless of the non-specific chronic stress perceived by the mother (PSS) and the ANS response of the fetus (FSI), what most impacts the fetal epigenetic profile is the stress generated by specific worries related to pregnancy (captured by PDQ) and the associated high circulating levels of cortisol that is crossing the maternal-fetal placental barrier and impacting the fetal physiology on the scale of epigenome. It is possible that FSI is not the appropriate biophysical correlate of epigenome-level alterations due to PS. In future studies, to investigate this relationship further we intend to analyze in more depth the relationships between the neonatal epigenome and the biophysical features of ANS derived from maternal and fetal HRV.
Strengths and Limitations
In the present study we report the findings of the largest prospectively followed cohort of its kind to date. Several strengths are to be highlighted. Firstly, saliva cells are easy and non-invasive to obtain in newborns. Even though epigenetic changes like DNA methylation are cell and tissue-specific, some CpG sites show cross tissue relevance. Changes in peripheral tissues such as saliva could serve as potential biomarkers for disease risk while also giving an advantage of being non-invasively obtainable. Since the primary organ affected by stress is not available in human studies and post-mortem brain tissue samples cannot capture the fluid state of the epigenome (80), more accessible samples such as saliva and blood are often used as substitutes. Binder and colleagues showed that saliva reflects better DNA methylation patterns of the brain than methylation in blood, highlighting that saliva is the sample medium of choice for epigenetic studies of psychiatric traits, especially in small children (81).
Secondly, we believe that our study’s findings can be generalized to the population of pregnant women in most clinics, as this study includes mothers experiencing typical daily stress situations rather than extreme stress exposures.
Limitations to our study are as follows. Our study has a relatively small sample size, which makes identifying subtle differences in methylation difficult. Since there is no other available study with cohorts of pregnant women and newborn saliva samples obtained, we have not yet been able to verify our findings in an independent cohort. Thus, the novel findings of DMPs and DMRs related to these stress measures should be considered as hypothesis-generating and requiring further validation in larger cohorts.
Assessing the DNA methylation levels as soon as the baby is born in association with four stress measures shows the impact of maternal stress on epigenetic marks during the fetal life. However, to serve as early neurodevelopmental biomarkers these marks have to be related to the corresponding neurodevelopmental appraisals. Since epigenetic marks are not fixed at birth and methylation patterns change with age, we are presently carrying out a longitudinal study in this cohort. The DNA methylation status at two years of age will allow us to detect the epigenetic drift defined as the difference in the DNA methylation status over time (78). Moreover, the 2-years’ time point will allow us to test for an association with the neurodevelopmental outcome showing the influence of the environment during the first two years of life on the epigenetic traits and whether the present early neonatal epigenetic differences can serve as biomarkers for early interventions to help restore optimal neurodevelopmental trajectories (23).