Methylation quantitative trait locus analysis of osteoarthritis links epigenetics with genetic risk

Human Molecular Genetics, 2015, Vol. 24, No. 25 7432–7444 doi: 10.1093/hmg/ddv433 Advance Access Publication Date: 13 October 2015 Association Studies Article A S S O C I AT I O N S T U D I E S A R T I C L E Methylation quantitative trait locus analysis of osteoarthritis links epigenetics with genetic risk 1 Musculoskeletal Research Group, Institute of Cellular Medicine and, 2Institute of Genetic Medicine, International Centre for Life, Newcastle University, Newcastle upon Tyne NE1 3BZ, UK and 3Freeman Hospital, High Heaton, Newcastle upon Tyne NE7 7DN, UK *To whom correspondence should be addressed at: Institute of Cellular Medicine, Newcastle University, 4th Floor, Catherine Cookson Building, The Medical School, Framlington Place, Newcastle upon Tyne NE2 4HH, UK. Tel: +44 1912087178; Fax: +44 1912085455; Email: Abstract Osteoarthritis (OA) is a common, painful and debilitating disease of articulating joints resulting from the age-associated loss of cartilage. Well-powered genetic studies have identified a number of DNA polymorphisms that are associated with OA susceptibility. Like most complex trait loci, these OA loci are thought to influence disease susceptibility through the regulation of gene expression, so-called expression quantitative loci, or eQTLs. One mechanism through which eQTLs act is epigenetic, by modulating DNA methylation. In such cases, there are quantitative differences in DNA methylation between the two alleles of the causal polymorphism, with the association signal referred to as a methylation quantitative trait locus, or meQTL. In this study, we aimed to investigate whether the OA susceptibility loci identified to date are functioning as meQTLs by integrating genotype data with whole genome methylation data of cartilage DNA. We investigated potential genotype–methylation correlations within a 1.0–1.5 Mb region surrounding each of 16 OA-associated single-nucleotide polymorphisms (SNPs) in 99 cartilage samples and identified four that function as meQTLs. Three of these replicated in an additional cohort of up to 62 OA patients. These observations suggest that OA susceptibility loci regulate the level of DNA methylation in cis and provide a mechanistic explanation as to how these loci impact upon OA susceptibility, further increasing our understanding of the role of genetics and epigenetics in this common disease. Introduction Osteoarthritis (OA) is characterized by the age-related gradual thinning and eventual focal loss of articular cartilage and as such is a painful disease that severely impacts on normal joint function (1,2). There are no disease-modifying pharmacological therapies for OA, with pain management and joint replacement the principal clinical treatments. As a result, the health economic burden of the disease is large and will increase with a progressively ageing population. OA is polygenic and a number of candidate gene and genomewide association scan (GWAS) studies have reported OA riskconferring loci for knee, hip or hand OA that exceed or are close to the genome-wide significance threshold (3–14). While some of the signals encompass genes with prior known roles in joint † These authors contributed equally to this work. These authors jointly supervised this work. Received: June 19, 2015. Revised: September 9, 2015. Accepted: October 9, 2015 ‡ © The Author 2015. Published by Oxford University Press. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited. 7432 Michael D. Rushton1,†, Louise N. Reynard1,†,‡, David A. Young1, Colin Shepherd1, Guillaume Aubourg1, Fiona Gee1, Rebecca Darlay2, David Deehan1,3, Heather J. Cordell2 and John Loughlin1, *,‡ Human Molecular Genetics, 2015, Vol. 24, No. 25 (NOF) fracture and who are free of OA. In total, we assessed 16 SNPs representing 16 loci that have been reported to be associated with OA. Results Identification of OA meQTLs Genome-wide methylation was measured using our previously published data (24) using the Illumina Infinium HumanMethylation450 BeadChip. This array, which captures over 450 000 CpGs, will henceforth be referred to as the 450k array. For each of the 16 loci, we covered at least the whole of the LD region; therefore, for 15 loci we spanned 1 Mb while for the rs10948172 locus we spanned 1.5 Mb. Table 1 lists the loci and the number of CpG probes from the 450k array that were assessed at each locus for genotype–methylation correlations. Initially, we assessed correlations in all samples combined—OA knee, OA hip and NOF. This analysis identified four SNPs that correlated with methylation (Benjamin–Hochberg P < 0.05): chromosome 3p21.1 (GLT8D1), SNP rs6976 and three CpGs; chromosome 6p21.1 (SUPT3H), SNP rs10948172 and four CpGs; chromosome 15q21.3 (ALDH1A2), SNP rs3204689 and one CpG; and chromosome 20q11 (GDF5), SNP rs143383 and one CpG (Table 2). Patient information and genotypes at the four SNPs are shown in Supplementary Material, Table S1. For each of these nine CpGs, we next compared the methylation levels without genotype stratification but stratified by OA knee, OA hip and NOF. This revealed that the degree of methylation was on average significantly lower in the NOF patients for the 3p21.1 (GLT8D1, rs6976) and 6p21.1 (SUPT3H, rs10948172) CpGs, but that there was no difference between OA knee and OA hip (data not shown). We therefore repeated the analysis for all 16 loci separating OA from NOF. This analysis of only OA samples identified the same four SNPs and no OA-specific or NOF-specific genotype–methylation correlations (Table 2). As mentioned previously, many OA genetic association signals were only discovered following stratification by joint and/or sex. When we repeated our analysis using such stratification, we did not identify any further genotype–methylation correlations. In summary, we identified four potential meQTLs, and these are summarized in Table 2 and discussed in more detail below. Genotype at GLT8D1 SNP rs6976 correlated with the methylation of three CpGs located ∼31.5 kb upstream of the LD block and ∼170 kb from the SNP: cg18099408, cg15147215 and cg18591801. All three CpGs are within the gene body of STAB1 (Fig. 1A), and for each CpG, the OA-associated T allele of rs6976 correlated with lower methylation (Fig. 1B; data for the OA patients only shown). Significant correlations were seen when all of the patient samples were combined (OA knee, OA hip and NOF) as well as in OA samples alone (Table 2). Genotype at the SUPT3H SNP rs10948172 correlated with the methylation of four adjacent CpGs located ∼82 kb upstream from the SNP: cg13979708, cg19254793, cg20913747 and cg18551225. All four are located within the LD region marked by rs10948172, which encompasses a ∼667 kb region from hg19 chromosome 6:44 683 049–45 349 877 (Fig. 2A). (...truncated)