DNA methylome signature in rheumatoid arthritis

Abstract

Objectives: Epigenetics can influence disease susceptibility and severity. While DNA methylation of individual genes has been explored in autoimmunity, no unbiased systematic analyses have been reported. Therefore, a genome-wide evaluation of DNA methylation loci in fibroblast-like synoviocytes (FLS) isolated from the site of disease in rheumatoid arthritis (RA) was performed.

Methods: Genomic DNA was isolated from six RA and five osteoarthritis (OA) FLS lines and evaluated using the Illumina HumanMethylation450 chip. Cluster analysis of data was performed and corrected using Benjamini-Hochberg adjustment for multiple comparisons. Methylation was confirmed by pyrosequencing and gene expression was determined by qPCR. Pathway analysis was performed using the Kyoto Encyclopedia of Genes and Genomes.

Results: RA and control FLS segregated based on DNA methylation, with 1859 differentially methylated loci. Hypomethylated loci were identified in key genes relevant to RA, such as CHI3L1, CASP1, STAT3, MAP3K5, MEFV and WISP3. Hypermethylation was also observed, including TGFBR2 and FOXO1. Hypomethylation of individual genes was associated with increased gene expression. Grouped analysis identified 207 hypermethylated or hypomethylated genes with multiple differentially methylated loci, including COL1A1, MEFV and TNF. Hypomethylation was increased in multiple pathways related to cell migration, including focal adhesion, cell adhesion, transendothelial migration and extracellular matrix interactions. Confirmatory studies with OA and normal FLS also demonstrated segregation of RA from control FLS based on methylation pattern.

Conclusions: Differentially methylated genes could alter FLS gene expression and contribute to the pathogenesis of RA. DNA methylation of critical genes suggests that RA FLS are imprinted and implicate epigenetic contributions to inflammatory arthritis.

Figure 1

Hierarchical clustering and heatmap of differentially methylated (DM) loci. The methylation levels at the 1859 significantly DM loci were used for hierarchical clustering. (A) The clustering of the samples is shown by the dendrogram at the top and the clustering of the loci is shown by the dendrogram on the left. The methylation levels at the loci are shown in the heatmap. (B) Representative correlation in DM loci between the HumanMethylation450 BeadChip and pyrosequencing. DM loci of five genes (TNF; 2, FERMT3; 3, CHI3L1; 1, ITGA4; 2, and MYEF2; 2 loci) of the HumanMethylation450 BeadChip were validated by pyrosequencing (all 11 samples; six rheumatoid arthritis (RA), five osteoarthritis (OA)). The correlation is shown for a hypermethylated locus (TNF) and hypomethylated loci (IGTA4 and CHI3L1) (TNF, chr6:31543300; ITGA4, chr2:182321354; CHI3L1, chr1:203156625, GRCh37/hg19). The y-axis shows the percent methylated cytosines. The x-axis shows the mean methylation score from the Illumina chip. p Value <0.01 for all loci. (C) Gene expression of hypomethylated genes in RA fibroblast-like synoviocytes (FLS). Gene expression was determined by PCR in 6–13 separate OA and RA FLS lines for seven genes that were significantly hypomethylated (CHI3L1, COL1A1, MYEF2, ITG4A, SYNJ2, STK24, MAP3K5). Representative examples are shown for hypomethylated genes. Taken as a group, expression of hypomethylated genes in RA was significantly greater than OA (p<0.01). As a group, expression of genes that were normally methylated in RA was similar to OA (representative examples shown) (AXIN, IKKE, TBK1, NANOG, POU5F1, MAP2K6, IRF3).

Figure 2

Hierarchical clustering and heatmap of differentially methylated (DM) loci. Hierarchical clustering was performed using the top 20 DM loci as described in the text. Note that the osteoarthritis (OA) fibroblast-like synoviocytes (FLS) lines were completely segregated from rheumatoid arthritis (RA) using these loci, including three new RA and six new OA FLS lines that were independent of the algorithm. In addition, six new normal FLS lines clustered with OA. See supplementary online table S5 for identification of the loci used in the model.