Integrative epigenome-wide analysis demonstrates that DNA methylation may mediate genetic risk in inflammatory bowel disease

Abstract

Epigenetic alterations may provide important insights into gene-environment interaction in inflammatory bowel disease (IBD). Here we observe epigenome-wide DNA methylation differences in 240 newly-diagnosed IBD cases and 190 controls. These include 439 differentially methylated positions (DMPs) and 5 differentially methylated regions (DMRs), which we study in detail using whole genome bisulphite sequencing. We replicate the top DMP (RPS6KA2) and DMRs (VMP1, ITGB2 and TXK) in an independent cohort. Using paired genetic and epigenetic data, we delineate methylation quantitative trait loci; VMP1/microRNA-21 methylation associates with two polymorphisms in linkage disequilibrium with a known IBD susceptibility variant. Separated cell data shows that IBD-associated hypermethylation within the TXK promoter region negatively correlates with gene expression in whole-blood and CD8⁺ T cells, but not other cell types. Thus, site-specific DNA methylation changes in IBD relate to underlying genotype and associate with cell-specific alteration in gene expression.

Figure 1. Differentially methylated positions (DMP) analysis in Inflammatory bowel disease (IBD) cases and controls in whole blood.

(a) Manhattan plot of top differentially methylated positions (DMPs) in Inflammatory bowel disease (IBD) versus control. (b) Volcano plot of top DMPs and position of methylation probes in relation to the gene (IGR, intergenic region; TSS, transcription start site; UTR, untranslated region).

Figure 2. Detailed characterization of the VMP1/microRNA-21 region.

(a) VMP1/microRNA-21 region 450 K microarray probes (triangle) in inflammatory bowel disease cases (IBD, Red) and controls (Blue). (b) the same VMP1/microRNA-21 region mapped using whole genome bisulphite sequencing (WGBS) data (Red line=IBD cases, Blue line=Controls). (c) VMP1/microRNA-21 gene schematic diagram. Note only the first two exons of VMP1 shown. (d) Correlation between 450k microarray probes and WGBS data at same site. Correlation using Pearson's test. X axis denotes Chr 17 (h19) coordinates. DMR, differentially methylated region in IBD versus control case control 450k analysis.

Figure 3. Cell-type specificity of DNA methylation signals.

(a) Principal component analysis demonstrating the first two components of the entire (i) DNA methylation data set (ii) Gene expression data set. Both demonstrate tight clustering according to the cell type of origin of the sample. (b) Volcano plots for IBD versus controls differential methylation position (DMP) analysis for separated cells (CD4⁺, CD8⁺ T cells and CD14⁺ monocytes). (c,d) Boxplots show the median, 25th and 75th percentiles, and 1.5 * interquartile range (error bars) of methylation (beta) values. Cell-specific DNA methylation profiles. (c) The top differentially methylated position (RPS6KA2) was hypomethylated in whole blood and also monocytes. There was a larger difference between cases and controls in the separated cells compared with whole tissue (blood). (d) demonstrates monocyte specific DNA methylation at the histone deacetylase 4 (HDAC4) locus. Beta differences and uncorrected P values derived from linear models (IBD cases versus controls, including age and sex as covariates).

Figure 4. VMP1 genotype associates with DNA methylation.

(a) The genotype of rs8078424 strongly associates with VMP1 DNA methylation (cg16936953) (FDR corrected P=8.8 × 10⁻⁵, linear model). (b) VMP1 DNA methylation strongly associates with Inflammatory bowel disease (IBD) case status compared with controls (Holm corrected P=2.2 × 10⁻¹³, linear model). (c) The rs8078424 genotype associates with IBD status (Cochran-Armitage test 1df χ²=4.7 uncorrected P=0.03)(rs10853015 also associates Cochran-Armitage test 1df χ²=6.6 uncorrected P=0.01, both in Hardy-Weinberg equilibrium). Error bars denote s.e. (d) The two SNPs (rs10853015 and rs8078424, blue diamonds) that associate with methylation of the VMP1/microRNA-21 locus (red inverted triangle) are in linkage disequilibrium with the known IBD-susceptibility polymorphism (rs1292053, green diamond)(rs1292053-rs8078424, distance=13072, bp, D′=1, r²=0.43 and rs1292053–rs10853015, distance=185198, D′=0.93, r²=0.43). Plot generated using SNAP.

Figure 5. DNA methylation relates with cell-specific TXK gene expression.

TXK (Tyrosine Kinase) DNA methylation (a) and gene expression (c) in whole blood (Globin mRNA depleted), CD8⁺ T Cells, CD4⁺ T Cells and, CD14⁺ monocytes. Differences in methylation (Diff Beta), and gene expression (LogFC, Log fold change) and uncorrected P values (p) derived from linear models including age and sex as covariates. (b) demonstrates correlation (P Values derive from Pearson's correlation) between TXK gene expression and DNA methylation in matched samples. Boxplots show the median, 25th and 75th percentiles, and 1.5 * interquartile range (error bars).