Methylome-wide association study of antidepressant exposure

Antidepressants are widely prescribed for Major Depressive Disorder (MDD), although individual response is unpredictable, and their mechanism of action is incompletely understood. Many medications are associated with changes in DNA methylation that have links to their mode of action. Therefore, studying the effects of antidepressant exposure on DNA methylation may provide mechanistic insights into their efficacy and varied side effects. This study presents a comprehensive examination of the association of antidepressant exposure and DNA methylation using both self-report and prescription derived antidepressant phenotypes from Generation Scotland (n ∼ 20,000). We derived two phenotypes of antidepressant exposure, derived from electronic prescription record linkage (Ncases = 863, Ncontrols= 12,213) and self-report data (Ncases = 1509, Ncontrols = 15,051) respectively. Methylome-wide association analyses (MWAS) were conducted on each phenotype using the MOA model in OSCA software which included age, sex, AHRR probe methylation, monocyte cell counts, lymphocyte cell counts as fixed effects and an omics-relatedness-matrix as a random effect. Significance was determined as p < 9.42 × 10-8. Differentially methylated regions (DMR) were assessed using the R package dmrff. The genes annotated to the top 100 CpG hits were tested for enrichment of synapse-related pathways (obtained from SynGO database) by the missMethyl R package. We found differential methylation at 7 CpGs associated with self-reported antidepressant exposure, with the top CpG (cg262777237) annotated to a gene involved with cytoskeleton formation, KANK1 (b= 0.084, p= 3.87 × 10-13). There was no statistically significant DMRs identified, although a region on Chromosome 12 close to SHMT, an enzyme related to NMDA-receptor neurotransmission, was the most enriched region. Downstream SynGO pathway analysis found annotation of 6 genes (SHISA6, DGKI, DOCK10, AKT1, ALK and STAU2) relating to post-synaptic density, however there was no significant enrichment for any of the pathways. In the MWAS using the prescription-derived antidepressant phenotype, we found a strong concordance of effect sizes to those found the self-report analysis (R = 0.56, p < 2.2 × 10-16). We found differential methylation of 4 CpGs were significantly associated with prescription-derived antidepressant exposure, with the top hit (cg04173586) annotated to DOT1L, a histone methyltransferase linked to stress-sensitivity responses (b = 0.046, p = 3.98 × 10-12). The remaining hits were annotated to genes seen in the self-reported MWAS results (DGUOK-AS1, DOT1L, and KANK1). Similar analyses are currently being performed in the Netherlands Twin Registry (NTR) and Avon-Longitudinal Study of Parents and Children (ALSPAC) cohorts for the self-reported antidepressant MWAS. Findings to date do not strongly point to involvement of the serotonin transporter, the TrkB receptor or other mechanisms postulated to explain the action of antidepressants.