Cindy Boer

Genetics of Osteoarthritis Consortium GWAS Meta-Analyses | 171 4.1 Although the majority of SNVs confer risk or protection for disease consistently across phenotypes, two lead SNVs for hand osteoarthritis demonstrated opposite di-rection of effects with other osteoarthritis phenotypes ( Figure 2 ): rs3993110 in the TEA domain family member 1 ( TEAD1 ) gene with hip osteoarthritis and THR, and rs11071366 in the aldehyde dehydrogenase 1 family member A2 (ALDH1A2) gene with knee osteoarthritis and TKR. They are both common variants with MAF of 0.393 and 0.387, respectively ( Supplementary Information ). Genetic links between phenotypes We examined the extent to which subtypes of osteoarthritis phenotypes share a genet- ic component by cross trait LD Score regression using equally sized, non-overlapping, subgroups of the sample-sets from the meta-analysis[16, 17]. We found osteoarthritis subtypes to share substantial genetic components, albeit with a wide range. The lowest genetic correlation among anatomically very distinct joints (hip, knee, spine, finger and thumb) was observed between finger and hip osteoarthritis (rg=0.38), and the highest was observed between osteoarthritis of finger and thumb (rg=0.72)( Figure 2 b , sup- plementary Table 5 ). We also investigated if osteoarthritis genetic components are shared with other traits using cross trait LD score regression[16, 17]. Using LD-Hub, we found that 274 of the 763 traits in this database were significantly correlated with osteoarthritis (Bonfer- roni p-value<6.0x10 -06 ): anthropometric traits (BMI, obesity, weight and fat mass), type 2 diabetes, education, depressive symptoms, smoking behaviour, bone mineral density, reproductive phenotypes and intelligence as previously reported (PMID: 29559693, PMID: 30664745), and, for the first time, several pain phenotypes ( Supplementary Table 5 and supplementary Table 6, Supplementary Information ). We also investigated cross-trait LD score regression of these phenotypes and others not included in the LD-Hub database using the UK Biobank and the Icelandic deCODE genetic datasets ( Supplementary Table 5 ). In these analyses, BMI is also ge- netically correlated with all osteoarthritis phenotypes, with highest correlation with knee osteoarthritis, rg= 0.42. We note that although fat mass measured by DXA is high- ly correlated using LD-Hub, neither fat mass nor lean mass are very significant in our cross-trait analysis where we correct the lean mass and fat mass measures for sex, age, and importantly, BMI ( Supplementary Table 5 ), whereas weight is. This indicates that increased weight, rather than increased fat tissue that results in higher BMI, is causal in osteoarthritis pathogenesis.