Associations between Maternal Psychosocial Stress, DNA Methylation, and Newborn Birth Weight Identified by Investigating Methylation at Individual, Regional, and Genome Levels

Hum Biol. 2019 Apr;91(2):117-131. doi: 10.13110/humanbiology.91.2.04.


Stress is known to affect health throughout life and into future generations, but the underlying molecular mechanisms are unknown. We tested the hypothesis that maternal psychosocial stress influences DNA methylation (DNAm), which in turn impacts newborn health outcomes. Specifically, we analyzed DNAm at individual, regional, and genome-wide levels to test for associations with maternal stress and newborn birth weight. Maternal venous blood and newborn cord blood (n = 24 and 22, respectively) were assayed for methylation at ∼450,000 CpG sites. Methylation was analyzed by examining CpG sites individually in an epigenome-wide association study (EWAS), as regional groups using variably methylated region (VMR) analysis in maternal blood only, and through the epigenome-wide measures using genome-wide mean methylation (GMM), Horvath's epigenetic clock, and mitotic age. These methylation measures were tested for association with three measures of maternal stress (maternal war trauma, chronic stress, and experience of sexual violence) and one health outcome (newborn birth weight). We observed that maternal experiences of war trauma, chronic stress, and sexual assault were each associated with decreased newborn birth weight (p < 1.95 × 10-7 in all cases). Testing individual CpG sites using EWAS, we observed no associations between DNAm and any measure of maternal stress or newborn birth weight in either maternal or cord blood, after Bonferroni multiple testing correction. However, the top-ranked CpG site in maternal blood that associated with maternal chronic stress and sexual violence before multiple testing correction is located near the SPON1 gene. Testing at a regional level, we found increased methylation of a VMR in maternal blood near SPON1 that was associated with chronic stress and sexual violence after Bonferroni multiple testing correction (p = 1.95 × 10-7 and 8.3 × 10-6, respectively). At the epigenomic level, cord blood GMM was associated with significantly higher levels of war trauma (p = 0.025) and was suggestively associated with sexual violence (p = 0.053). The other two epigenome-wide measures were not associated with maternal stress or newborn birth weight in either tissue type. Despite our small sample size, we identified associations even after conservative multiple testing correction. Specifically, we found associations between DNAm and the three measures of maternal stress across both tissues; specifically, a VMR in maternal blood and GMM in cord blood were both associated with different measures of maternal stress. The association of cord blood GMM, but not maternal blood GMM, with maternal stress may suggest different responses to stress in mother and newborn. It is noteworthy that we found associations only when CpG sites were analyzed in aggregate, either as VMRs or as a broad summary measure of GMM.