Genome-wide DNA methylation profiling of blood samples from patients with major depressive disorder: correlation with symptom heterogeneity

Abstract

Background: Alterations in DNA, such as DNA methylation, may be key molecular events involved in the development of major depressive disorder (MDD). We sought to clarify correlations between DNA methylation profiles and symptom heterogeneity among patients with MDD.

Methods: We conducted a genome-wide DNA methylation analysis of blood samples from patients with MDD and controls, using the Infinium MethylationEPIC BeadChip.

Results: We analyzed 283 blood samples, including 141 from an initial cohort (69 patients with MDD, 72 controls) and 142 from a second validation cohort (67 patients with MDD, 75 controls). After adjustment for age, sex, and blood cell heterogeneity, DNA methylation status at 2699 CpG sites tended to differ between patients with MDD and controls in both the initial and second cohorts. Hierarchical clustering of patients based on DNA methylation status at these 2699 CpG sites revealed a significant correlation with scores for GRID-Hamilton Depression Rating Scale (GRID-HAMD) items (depressed mood, guilt, early insomnia, middle insomnia, work and activities, psychic anxiety, loss of appetite, general somatic symptoms, and total score), suggesting the feasibility of severity diagnostics based on blood DNA methylation testing. Pathway over-representation analysis revealed that genes whose DNA methylation status was correlated with epigenetic clustering were accumulated in molecular pathways involved in various cellular functions, especially nerve development. For PLEKHD1, STK10, and FOXK1, DNA methylation levels were inversely correlated with expression levels in the Clinical Proteomic Tumor Analysis Consortium database. DNA hypomethylation of PLEKHD1, STK10, and FOXK1 was correlated with higher GRID-HAMD scores in both cohorts.

Limitations: Although we performed marker exploration using 2 cohorts including 283 participants, the heterogeneity of the molecular mechanisms operating in MDD might necessitate a larger cohort for establishment of criteria with sufficient diagnostic impact.

Conclusion: These findings indicate that the DNA methylation status of specific genes may correlate with the severity of MDD symptoms, and that genome-wide DNA methylation analysis of blood samples would be useful for clarifying the DNA methylation profiles related to symptom heterogeneity.

Figure 1

Epigenetic clustering of patients with major depressive disorder (MDD). We performed hierarchical clustering (Manhattan distances, Ward method) for 69 patients with MDD from the initial cohort using the residuals of 2699 CpG sites for which DNA methylation status tended to differ between the patients and controls, using multivariate analysis of covariance, adjusted for the participant group (MDD and control), age, sex, and estimated proportions of blood cells. The 69 patients with MDD were clustered into 2 subclasses, cluster A (n = 34) and cluster B (n = 35). Heatmaps for the total score of the GRID-Hamilton Depression Rating Scale (GRID-HAMD-17) and the residuals are shown at the top left of the panel. Supporting data are presented in Table 2.

Figure 2

Scattergrams of the residuals of the representative CpG sites by multivariate analysis of covariance, adjusted for the participant group (major depressive disorder [MDD] and control), age, sex, and the estimated proportions of blood cells. Probe identifications for the Infinium assay are shown at the top of each panel. (A) Representative CpG sites (cg12290615, cg18346402, cg08128444, cg20889774, cg15313859, cg15216812) showing DNA hypermethylation among patients with MDD in cluster A (n = 34) with more severe symptoms than those in cluster B (n = 35) (Welch t test p < 0.05 after Bonferroni correction). (B) Representative CpG sites (cg10651583, cg16207408, cg14657417, cg01008854, cg05348421, cg10959848) showing DNA hypomethylation in cluster A compared with cluster B (Welch t test p < 0.05 after Bonferroni correction). The mean for control participants is shown as residual 0.0 in each panel. Therefore, at all of the CpG sites shown here, patients in cluster B had DNA methylation levels similar to control participants, confirming that DNA hypermethylation and hypomethylation in cluster A are relative to both cluster B and controls.

Figure 3

Pathway over-representation analysis of biological processes from the Gene Ontology database (version 2024–01–17, doi:10.5281/zenodo.10536401) using 395 genes whose DNA methylation status (the residuals by the multivariate analysis of covariance adjusted for the participant group [major depressive disorder (MDD) and control], age, sex, and the estimated proportions of blood cells) was associated with epigenetic clustering and for which CpG probes were designed within CpG islands, island shores, or island shelves around the transcription start sites. Over-representation of the 395 genes for which DNA methylation status was potentially associated with MDD symptom severity and potentially participated in transcriptional regulation was observed in 24 pathways (p < 0.05 and false discovery rate [FDR] < 0.3). A heatmap for FDR is shown on the right side. Among the 395 genes, the number involved in the 24 pathways is represented by the size of the circles in this figure, and these genes are summarized in Appendix 3, Table S2. GTP = guanosine triphosphate; JNK = c-Jun N-terminal kinase.