Cindy Boer

Cartilage Thickness and Hip Osteoarthritis | 79 2.1 where we tested each of the 86 variants changing the amino-acid sequence for asso- ciation with the mJSW trait ( Supplementary Table 11 ). We observed four nominal significant associations, with rare variants in SULF2 , TGFA , RUNX2 and FGFR3 . None of these rare exonic variants explained the original association between the GWAS hit and mJSW or hipOA when tested in a multivariate model ( Supplementary Table 12 ). Next, we performed a burden test (SKAT)[23], to investigate whether the cumulative effects of the variants present in the sixteen selected genes were associated to mJSW, while adjusting for age and gender ( Table 5 ). We observed a nominal significance burden test (p-value<0.05) for TGFA , SULF2 , CHST11 and RUNX2 for mJSW. However, none of these findings reached significance after correction for multiple testing. mJSW associated SNPs in regulatory regions For most of the loci, no obvious protein-coding variants were found that could explain the associations. In previous studies it was shown that disease-associated variants are enriched in regulatory DNA regions[24,25]. We therefore examined whether the iden- tified DNA variants (or SNPs in high LD) resided in chondrocyte and/or osteoblast spe- cific enhancer regions, using data from ENCODE and ROADMAP[26–28]. To this end, we compared CHIP-seq signals from five different chromatin state markers (H3K4me3, H3K4me1, H3K36me3, H3K27me3, H3K9me3) in chondroblasts and osteoblasts to four cell lines from another origin. Together, these chromatin state markers identify promot- er and enhancer activity in each of the cell lines. With the exception of rs2862851, we observed that for all mJSW genetic loci, SNPs in high LD were located in cell regulatory regions in chondroblast and/or osteoblast cells (see Figure 3 for an overview and Sup- plementary Tables 13-18 . for each locus). Discussion Only a modest number of genetic variants has been successfully identified through ge- nome-wide association studies for OA This can in part be explained by the phenotypic heterogeneity of OA. Therefore, we used mJSW, a proxy for cartilage thickness in the hip joint, as one of the structural components of joint health. An additional advantage of this phenotype is its continuous nature, which increases power compared to a di- chotomous trait, such as OA-status. We identified six independent loci associated with cartilage thickness in the hip joint, of which four surpassed genome-wide significance ( TGFA , PIK3R1, SUPT3H-RUNX2, DOT1L ) and two were suggestive for association with mJSW ( SLBP/FGFR3, TREH-DDX6 ). Four of these loci ( TGFA, SUPT3H - RUNX2, DOT1L and FGFR3 ) were also associated with hip OA.