Cindy Boer

176 | Chapter 4.1 methylation and molecular QTLs from eQTLGen and GTEx ( Methods, Supplementary Information ), three osteoarthritis phenotype associations for 2 proteins were consid- ered robust ( Supplementary Table 9 ). Matrix gla protein (MGP) levels showed strong MR and colocalization evidence for hand and finger osteoarthritis. The lead SNV at this signal (rs7294636) is highly cor- related with UTR and missense variants in MGP  (rs1800801 and rs4236; r 2 >0.8), which have previously been found to be associated with hand osteoarthritis severity and with allele-specific expression of MGP [38]. Whilst it is possible that the pQTL proxies the effect of the missense variant on probe binding, previous functional studies have identi- fied correlation between the expression of MGP and osteoarthritis in cartilage and other joint tissues[38, 39]. MGP is a vitamin K-dependent protein secreted by chondrocytes and plays a role in the inhibition of ectopic tissue calcification. Spondin 2 (SPON2) protein levels demonstrated MR and colocalization evidence for knee osteoarthritis. The pQTL of SPON2 (rs878323) is in strong LD with a missense variant (rs11247975) therefore epitope-binding artefacts cannot be ruled out. SPON2 is expressed in human osteoarthritis synovial fluid[40], and is a secreted extracellu- lar matrix protein. It is essential for the initiation of immune responses and a unique pattern recognition molecule for microbial lipopolysaccharide[41, 42], which has been linked to knee osteoarthritis severity[43]. Amassing evidence to identify disease pathways By combining the results from the complementary functional genomics and compu- tational approaches outlined above, we identified 376 genes with at least one line of evidence pointing to a putative effector gene ( Table 3 , Supplementary Table 8 ) and assessed pathway and gene set signal enrichment in osteoarthritis pathology ( Supple- mentary Table 10 ). We found significant enrichment of signal in biological process- es known to be involved in osteoarthritis, specifically in pathways relevant to skele- tal, bone and cartilage development. Bone and cartilage development pathways were also found to be enriched in signals traversing weight-bearing and non-weight-bearing joints ( Supplementary Table 10 ), indicating joint development as a common mecha- nisms for any form of osteoarthritis[13]. Interestingly, when we only examined genes from SNVs associated with weight-bearing joints only, the most significant pathways were related to cellular response to endogenous stimuli, neurogenesis and Alzheimer’s ( Supplementary Table 10 ). Skeletal developmental pathways remained in the top 20 most significantly enriched pathways. We did not have sufficient power to detect signif- icant enrichment for the non-weight-bearing joint signals.