A Higher Dysregulation Burden of Brain DNA Methylation in Female Patients Implicated in the Sex Bias of Schizophrenia

Abstract

Sex differences are pervasive in schizophrenia (SCZ), but the extent and magnitude of DNA methylation (DNAm) changes underlying these differences remain uncharacterized. In this study, sex-stratified differential DNAm analysis was performed in postmortem brain samples from 117 SCZ and 137 controls, partitioned into discovery and replication datasets. Three differentially methylated positions (DMPs) were identified (adj. p < 0.05) in females and 29 DMPs in males without overlap between them. Over 81% of these sex-stratified DMPs were directionally consistent between sexes but with different effect sizes. Down-sampling analysis revealed more DMPs in females than in males when the sample sizes matched. Females had higher DNAm levels in healthy individuals and larger magnitude of DNAm changes in patients than males. Despite similar proportions of female-related DMPs (fDMPs, 8%) being under genetic control compared with males (10%), significant enrichment of DMP-related SNPs in signals of genome-wide association studies was identified only in fDMPs. One DMP in each sex connected the SNPs and gene expression of CALHM1 in females and CCDC149 in males. PPI subnetworks revealed that both female- and male-related differential DNAm interacted with synapse-related dysregulation. Immune-related pathways were unique for females and neuron-related pathways were associated with males. This study reveals remarkable quantitative differences in DNAm-related sexual dimorphism in SCZ and that females have a higher dysregulation burden of SCZ-associated DNAm than males.

Figure 1. Sex-stratified differential DNAm in SCZ.

Volcano plot of SCZ-associated differential DNAm (A) for all tested CpGs and (C) for chr X CpGs in females (right) and males (left). Red points indicate the significantly hypermethylated DMPs in SCZ, and blue points indicate the hypomethylated DMPs. All other points are gray. The dash lines represent the threshold of adj.p < 0.05 (top), p < 1e-05 (middle) and p < 1e-04 (bottom), respectively. (B) Example of the methylation state in female and male subgroups for fDMP at cg20584841 and mDMP at cg17165266 with adj.p < 0.05. Red arrow shows the direction and magnitude of DNAm changes in SCZ. (D) Effect sizes correlation between sex-by-disease interactions and sex-stratified differential methylations in those interactive DMPs.DNAm, DNA methylation; SCZ, schizophrenia; DMP differentially methylated positions; chrX, chromosome X; fDMP female-related DMP; mDMP male-related DMP.

Figure 2. Comparing the magnitude of DNAm changes between females and males with SCZ.

(A) RRHO maps highlight the concordant DNAm changes between females and males with SCZ. The CpGs were ranked by the −log10 of DMP p-value multiplied by the effect size direction. The log10-transformed hypergeometric p-values were plotted in the heatmap as indicated by an accompanying color scale. Signals in the bottom left quadrant represent the overlap for CpGs upregulated in both sexes, while signals in the top right quadrant represent the overlap for CpGs downregulated in both sexes. (B) Effect sizes comparison for the RRHO defined concordant CpGs between females and males with SCZ. The right panel shows violin plot of the up-regulated CpGs, and the left panel for the down-regulated CpGs. (C) Effect sizes correlation between females and males with SCZ in RRHO defined concordant CpGs (with p < 0.05 in both sexes), chrX CpGs and all tested CpGs, respectively. (D) Circular-Manhattan plot of p-value by chromosome positions for fDMPs, mDMPs and sDMPs. The red dash lines represent the threshold of p < 1e-04 (for fDMPs and mDMPs) and adj.p < 0.05 (for sDMPs). DNAm, DNA methylation; SCZ, schizophrenia; RRHO, rank-rank hypergeometric overlap; DMP differentially methylated positions; fDMPs, female-related DMPs; mDMPs, male-related DMPs; sDMPs, sex-related DMPs.

Figure 3. Sex-stratified DMPs mediate genetic effects on gene expression.

(A) Four groups of significant SNPs-cg02167201-CALHM2 relationships in female SCZ. Three of those four SNPs were in LD with the reported susceptibility SNP (rs1163238) in SCZ (ref. 42). (B) Eight groups of significant SNPs-cg24278948-CCDC149 relationships in male SCZ. SNP single nucleotide polymorphism; SCZ, schizophrenia; LD, linkage disequilibrium; SDP SNP-DMP pairs.

Figure 4. Sex-specific PPI subnetworks and biological processes.

PPI subnetworks for (A) mPN5 (centered around TJP1gene), (B) fPN3 (centered around GRIA2 gene), (C) fPN1 (centered around PSMD14 gene) and fPN2 (centered around PSMB4gene). Every node represents a gene. The color of nodes represents differential methylation levels in corresponding promoters (yellow represents hypomethylation; blue means hypermethylation). The edges were built based on the protein-protein interaction in the Pathway Common. Red circles represent the up-regulated gene expression in SCZ-related differential expression analysis according to PsychENCODE results (p < 0.05) (ref.41), while the green circle indicates a down-regulation (p < 0.05). (D)Gene Ontology annotations for mPN5 and fPN3. (E) Top 10 Gene Ontology annotations for fPN1. PPI, protein-protein interaction; fPN, female-related PPI subnetwork; mPN, male-related PPI subnetwork; SCZ, schizophrenia.

Figure 5. The model for sex-biased DNAm burden hypothesis of SCZ.

In this model, a higher dysregulation burden of DNAm is required for females to manifest the SCZ phenotype than males. SCZ, schizophrenia; DNAm, DNA methylation