Cindy Boer

128 | Chapter 3.1 Visualisation of the associated loci and the regulatory landscape For the top GWAS-associated single nucleotide polymorphism (SNP), the linkage dis- equilibrium (LD) region (r 2  >0.8) was determined using the 1000G Phase-1 population using the HaploReg V3 tool.[21] Using the ROADMAP-generated reference epigenomes, we determined if any of the variants in high LD were located in potential gene regula- tory regions in primary osteoblasts (generated by ENCODE) and bone marrow-derived chondrocytes (ROADMAP).[22,23] The 18-state chromatin reference epigenomes were downloaded from the ROADMAP epigenomes data portal.[23] SNPs and regulatory an- notations were visualised using the UCSC Genome Browser on GRCh37/hg19.[24] For each variant, it was also determined if the alternative allele would disrupt a protein binding motif; this was done using the HaploReg V3 tool.[21] RNA sequencing data Post-RNA isolation (Qiagen RNeasy Mini Kit, RIN >7) of 40 knee (15 paired preserved (P) and OA lesioned (OAL), 7 P only and 3 OAL only) and 28 hip (six paired P and OAL, 14 P only and 2 OAL only) cartilage samples ( online supplementary table S2 ), paired-end 2×100bp RNA library sequencing (Illumina TruSeq RNA-Library Prep Kit, Illumina HiSeq2000) resulted in an average of 10million fragments per sample. Reads were aligned using GSNAP against GRCh37/hg19, in which SNPs from the Genome of the Netherlands consortium with a minor allele frequency (MAF) >1% were masked to prevent alignment bias. Number of fragments per gene were used to assess quan- tile-adjusted conditional maximum likelihood (edgeR, R-package). Subsequently, differ- ential gene expression analysis was performed pairwise between P and OAL samples for which we had RNA of both (n=21). Allele-specific expression (ASE) was assessed using SNVMix2 [25] with default settings(min coverage=25, 10 reads per allele). The extent of ASE was defined as the fraction of risk allele among all counts at the respective location. Meta-analysis was done only across P samples or OAL when no P counterpart sample was present. p Values were calculated using canonical binominal test (metagen R-package). TaqMan assay Conventional TaqMan genotyping was performed on both genomic DNA (gDNA), ar- ticular cartilage and subchondral bone cDNA. An allele-specific custom TaqMan assay for rs1800801 (Thermo Fisher Scientific) was used to quantify the allele ratio in cDNA samples and were normalised against the gDNA ratio, which was used as an 1:1 allele ratio reference. Each sample has been measured in four (cartilage) or eight (subchon-