Cindy Boer

76 | Chapter 2.1 Table 9 ), as well as expression in bone and cartilage tissue in mice using data from the Jackson lab database ( Supplementary Table 7 ). To further explore which genes are possibly underlying the genetic associations identified in this study, we analyzed gene expression in a paired set of non-lesioned and OA-lesioned cartilage samples of the RAAK study acquired from 33 donors at the time of joint replacement surgery for primary OA[19]. We first examined which genes are expressed in a set of seven human healthy cartilage samples ( Supplementary Table 7 ). Additionally, we tested which of the genes located in 1MB region surrounding the lead SNP were differentially expressed in OA-lesioned cartilage versus non-lesioned carti- lage of the same hip. Of the 152 genes that were selected, 129 genes were represented on the array. Of those, 64 genes were significantly expressed in the cartilage samples. For eight of the twelve loci, we found genes that were differentially expressed in OA lesioned cartilage versus non-lesioned cartilage ( Table 4 , Supplementary Table 10 ). Differential expression in cartilage healthy vs OA affected cartilage was performed like- wise ( Supplementary Table 10 ), while additionally adjusting for sex and age. Given the relatively small number of healthy samples (n=7) with large age range these data are less robust and we did not use these data in gene prioritization. For each gene a prioritization score was computed, based on equally weighting of the ten lines of evidence ( Table 4 ). Following this approach, RUNX2 is highly likely to be the causal gene associated with rs12206662 and rs10948155. Similar strong evidence is found for rs788748 ( IGFBP3 ) and rs10492367 ( PTHLH ). In addition, suggestive ev- idence for a causal gene is found for the following: rs10471753 ( PIK3R1 ), rs835487 ( CHST11 ), rs2862851 ( TGFA ), rs6094710 ( SULF2 ), rs9350591 ( COL12A1 ) and rs11177 ( GNL3 ). However for some loci the current evidence is ambiguous, suggesting more than one gene as the potentially causal one; rs2236995 ( FGFR3 or SLBP ), rs11880992 ( GADD45B or DOT1L ) and rs496547( KMT2A or UPK2 ) ( Table 4 ) Exome sequencing of prioritised genes In 2,628 individuals from the Rotterdam Study, exome sequencing was performed at a mean depth of 55x. Of those, 2,050 individuals also had mJSW and hip OA phenotype data. Baseline characteristics of those individuals were similar to the source popula- tion, mean age was 67.3 years, 57% of the individuals were female and mean of mJSW was 3.81 mm (SD 0.82). Details of the experimental procedure and variant calling are given in the supplementary material ( Supplementary Text 2 ). Only the variants with a minimal allele count of three in the total population were selected for analysis. With- in the sixteen prioritized genes, a total number of 158 variants were identified in the protein-coding region, of which 85 were non-synonymous and one was a stop-gain mu- tation ( Table 5 , Supplementary Table 11 ). We first performed a single variant test,