DNA methylation partially mediates the relationship between childhood adversity and depressive symptoms in adolescence

Background: Exposure to adversity during childhood is estimated to at least double the risk of depression later in life. Some evidence suggests childhood adversity may have a greater impact on depression risk, if experienced during specific windows of development called sensitive periods. During these sensitive periods, there is evidence that adversity may leave behind biological memories, including changes in DNA methylation (DNAm). Here we ask if those changes play a role in the link between adversity and later adolescent depressive symptoms. Methods: We applied a method for high-dimensional mediation analysis using data from a subsample (n=627-675) of the Avon Longitudinal Study of Parents and Children. We first assessed the possibility of time-dependent relationships between seven types of childhood adversity (caregiver abuse, physical/sexual abuse, maternal psychopathology, one-adult household, family instability, financial stress, neighborhood disadvantage), measured on at least four occasions between ages 0-7 years, and adolescent depression at mean age 10.6. Specifically, we considered three types of life course hypotheses (sensitive periods, accumulation, and recency), and then evaluated which of these hypotheses had the strongest association in each adversity-adolescent depression relationship using the structured life course modeling approach (SLCMA; pronounced slick-mah). To conduct the mediation analyses, we used a combination of pruning and sure independence screening (a dimension reduction method) to reduce the number of methylated CpG sites under consideration to a viable subset for our sample size. We then applied a sparse group lasso penalized model to identify the top mediating loci from that subset using the combined strength of the coefficient measuring the relationship between the childhood adversity and a CpG site () and of the coefficient measuring the relationship between the CpG site and depressive symptoms ({beta}) as a metric. Using a Monte Carlo method for assessing mediation (MCMAM), we assigned a significance level and confidence interval to each identified mediator. Results: Across all seven adversities, we identified a total of 70 CpG sites that showed evidence of mediating the relationship between adversity and adolescent depression symptoms. Of these 70 mediators, 37 were significant at the p < 0.05 level when applying the MCMAM, a method tailored to estimating the significance of SEM-derived mediation effects. These sites exhibited four different mediating patterns, differentiated by the direction of and {beta}. These patterns had signals that were: (1) both positive (19 loci), (2) both negative (18 loci), (3) positive and negative {beta} (23 loci) or (4) negative and positive {beta} (10 loci). Conclusion: Our results suggest that DNAm partially mediates the relationship between different types of childhood adversity and depressive symptoms in adolescence. These findings provide insight into the biological mechanisms that link childhood adversity to depression, which will ultimately help develop treatments to prevent depression in more vulnerable populations.