Longitudinal analysis of blood-based dna methylation profiles in major depressive disorder

Major depressive disorder (MDD) is a complex multifactorial condition with an estimated heritability of 40%, characterized by episodes of depressed mood, loss of interest and additional symptoms such as weight changes, sleep disturbances and fatigue. Both a high relapse and recurrence risk of MDD are known from epidemiological studies, however, biomarkers that could facilitate early detection and prediction of prognosis are lacking. DNA methylation (DNAm) has been suggested as a potential molecular marker for disorder states and traits due to its capacity to integrate both genetic and environmental signals into an epigenetic memory. In a sample of individuals with a lifetime diagnosis of MDD and healthy controls, we here performed a longitudinal analysis of blood-based DNAm profiles with regard to case-control status and longitudinal readouts, including the number of depressive episodes, number of hospitalizations and phase of illness at measurement. For 185 MDD cases and 188 age- and sex-matched healthy controls drawn from the deeply phenotyped German FOR2107 cohort, DNAm at baseline and at a two-year follow-up study visit was profiled with the Infinium MethylationEPIC BeadChip in peripheral blood. Data quality control and stratified quantile normalization of methylation signals were conducted in R using the minfi and ewastools packages, resulting in a longitudinal analysis sample of 178 MDD cases and 178 healthy controls with two measurements each. Differential methylation was examined across more than 760k CpG sites with linear mixed models using the limma package, including proband identity as a random effect. For each independent variable of interest, a separate model was fitted. As fixed effect covariates, sex, age, cell type composition estimated with the Houseman approach, ten genotyping-based principal components (PC) as well as ten methylation control-probe PCs were included in each model. Associations with p < 9e-8 were considered methylome-wide significant. In the analysis of case-control status, significant differential methylation was detected at four CpG sites, including hypomethylation of cg04847273, which is located within the GABRD gene encoding a GABA receptor subunit. Regarding the number of depressive episodes and hospitalizations, 13 and 40 CpG sites, respectively, showed significant differential methylation. When comparing different categories reflecting phase of illness (acute episode, partial remission, full remission, healthy), significant differential methylation was visible between samples of MDD cases in full remission and healthy controls at two CpG sites. To the best of our knowledge, our study represents one of the largest methylome-wide longitudinal investigations of MDD so far. The detected differences in blood-based DNAm profiles provide a promising starting point for the identification of trait- and state-dependent biomarkers in MDD. The hypomethylation within GABRD appears to be particularly interesting, as it might be related to GABAergic dysregulation in mood disorders implicated by previous studies. Larger sample sizes and a higher number of longitudinal measurements per individual are required to increase statistical power, and any identified associations should be replicated in independent cohorts.