Layered genetic control of DNA methylation and gene expression: a locus of multiple sclerosis in healthy individuals

Abstract

DNA methylation may contribute to the etiology of complex genetic disorders through its impact on genome integrity and gene expression; it is modulated by DNA-sequence variants, named methylation quantitative trait loci (meQTLs). Most meQTLs influence methylation of a few CpG dinucleotides within short genomic regions (<3 kb). Here, we identified a layered genetic control of DNA methylation at numerous CpGs across a long 300 kb genomic region. This control involved a single long-range meQTL and multiple local meQTLs. The long-range meQTL explained up to 75% of variance in methylation of CpGs located over extended areas of the 300 kb region. The meQTL was identified in four samples (P = 2.8 × 10(-17), 3.1 × 10(-31), 4.0 × 10(-71) and 5.2 × 10(-199)), comprising a total of 2796 individuals. The long-range meQTL was strongly associated not only with DNA methylation but also with mRNA expression of several genes within the 300 kb region (P = 7.1 × 10(-18)-1.0 × 10(-123)). The associations of the meQTL with gene expression became attenuated when adjusted for DNA methylation (causal inference test: P = 2.4 × 10(-13)-7.1 × 10(-20)), indicating coordinated regulation of DNA methylation and gene expression. Further, the long-range meQTL was found to be in linkage disequilibrium with the most replicated locus of multiple sclerosis, a disease affecting primarily the brain white matter. In middle-aged adults free of the disease, we observed that the risk allele was associated with subtle structural properties of the brain white matter found in multiple sclerosis (P = 0.02). In summary, we identified a long-range meQTL that controls methylation and expression of several genes and may be involved in increasing brain vulnerability to multiple sclerosis.

Figure 1.

Identification of the long-range meQTL with GWAS of PC1 in the SYS adolescents (top) and parents (bottom). Separate analyses were conducted in the samples of SYS adolescents (n = 132) and SYS parents (n = 278). Pink shaded area indicates the region of CpGs loading into PC1. In both the top and bottom panels, the purple dot indicates the top SNP associated with DNAm (PC1) in the discovery sample of SYS adolescents.

Figure 2.

The relative contributions of the long-range and local meQTLs to methylation of CpGs within the studied genomic region. The contributions are shown as proportions of variance explained (r²) by the long-range meQTL (purple lines) and respective local meQTLs (yellow lines with stars). Gray triangles indicate the combined contribution of the long-range and respective local meQTLs at each assessed CpG. The top panel includes the 25 highly variable CpGs loading into PC1 (including 19 polymorphic CpGs). The middle panel includes a total of 456 CpGs (including 85 polymorphic CpGs). The bottom panel includes a total of 5663 CpGs (including 273 polymorphic CpGs) located within the region of high linkage disequilibrium with the long-range meQTL (as shown in Fig. 1).

Figure 3.

Genomic landscape of regions containing the long-range meQTL (center), HLA-DRB1 (left) and HLA-DQB1 (right). DNase-I hypersensitivity signals; histone modification signals for H3K27ac, H3K4Me1 and H3K4Me3; RNA-seq transcription signals; CTCF-binding positions by Chip-seq, CpG islands and CpGs assessed with the 450K BeadChip are those observed in the lymphoblastoid GM12878 cells (ENCODE/Duke). Binding of transcription factors is that compiled from 72 cell types in Chip-seq ENCODE V2 (capital G denotes binding in GM12878 cells).

Figure 4.

Associations of the long-range meQTL with mRNA expression (top) and hypothesized models (bottom) of how minor and major alleles of the long-range meQTL affect regional chromatin structure and mRNA expression. The data are plotted by the meQTL genotype. The long-range meQTL is located within a super-enhancer. Depending on its allele, the super-enhancer may interact with either the HLA-DRB1 promoter (minor allele) or the HLA-DQB1 promoter (major allele). When interacting with the HLA-DRB1 promoter (minor allele), it could activate mRNA expression of HLA-DRB1 and two neighboring genes transcribed in the same direction (HLA-DRB6 and HLA-DRB5); note that HLA-DRB6 is a transcribed pseudogene (30). When interacting with the HLA-DQB1 promoter (major allele), it could create a chromatin loop and activates mRNA expression of HLA-DQB1 but not that of the neighboring HLA-DQA1 because that gene is transcribed in the opposite direction.

Figure 5.

Pairwise linkage disequilibrium between the long-range meQTL and previously identified disease loci (the NHGRI GWAS Catalog, only loci with r² > 0.2 are shown). Linkage disequilibrium (LD) was calculated with the 1000 Genomes European reference sample (March 2012).