Cindy Boer

Cartilage Thickness and Hip Osteoarthritis | 71 2.1 from these eighteen loci were selected for replication in an additional 8,227 individuals from seven different cohorts. We observed that six of the eighteen SNPs significantly replicated (p-value<0.05) with the same direction of effect ( T able 1 ). When we com- bined discovery and replication results in a meta-analysis, the five SNPs that met ge- nome-wide significance in the discovery analysis became more significant and another two SNPs that replicated in independent cohorts reached suggestive evidence (p-val- ue≤1*10 −06) for association in the combined meta-analysis. The top signal in the combined meta-analysis, rs1180992 ( Table 1 , p- value combined = 3.2x10 -16 ), is located in the intronic region of the previously OA associated DOT1L gene. This variant is very close to and in linkage disequilibrium with rs12982744 (D’=1, r 2 =1), whichwas previously found in associationwith mJSW and hip OA [1,8]. The DOT1L signal was followed in strength of association by rs2862851 (p- value combined = 5.2x10 −11 ), which is annotated to the intronic region of TGFA ( Figure 2a ). Two variants near RUNX2 , rs10948155 and rs12206662, also reached genome-wide significance for association with mJSW ( Figure 2b ). The two variants in the RUNX2 locus were weakly correlated (r 2 <0.2). Conditional analysis, using GCTA, showed that both SNPs represented different signals ( Supplementary T able 4 ). Finally, the last signal that reached genome-wide significance was rs10471753, an intergenic variant closer to PIK3R1 (~450 Kb) than to SLC30A5 (~750Kb) ( Table 1 , p-value combined = 3.8x10 −9 ). Other suggestive signals for association with mJSW at a p-value≤ 1x10 −6 including signals with significant replica-tion were rs496547 (p-value=1.5*10 −07 ), a downstream gene variant located in the 3’ of TREH and, an intron variant annotated near SLBP (rs2236995; p-value=9x10 −07 ). All other additional signals selected in the discovery stage did not replicate. Association of identified loci with hip OA and other musculoskeletal phenotypes We examined whether the five GWS and two suggestive mJSW loci were also associated with hip OA in a total of 8,649 cases and >57,000 controls. Detailed description of the cohorts and OA definitions is given in supplementary Table 1. Table 2 shows the asso-ciations found with hip OA. We observed that five of the seven identified mJSW loci were also associated with hip OA (p-value<0.05). Apart from the known DOT1L locus, the vari-ant near TGFA was significantly associated with hip OA ( Table 2 , p-value=4.3x10 −05 ). In addition, the SNP near SLBP and the two SNPs near RUNX2 were associated with hip OA. One of the latter SNPs, rs10948155, is in high LD with a variant (rs10948172, D’=0.95 and r 2 =0.90) previously found in association with hip OA in males at borderline GWS level[2]. However, in our study, rs10948155 was just marginally associated with hip OA in the overall analysis ( Table 2 , p-value=0.021).