Abstract
<title>Abstract</title> <p>Background Osteoarthritis (OA) is a major cause of chronic pain and disability, yet no approved therapy modifies disease progression. Human genetic evidence can improve therapeutic target selection, but OA genome-wide association studies often identify loci without resolving the causal gene or direction of effect. We therefore evaluated genes within two OA-associated regions on chromosomes 16 and 20, selected from an earlier exploratory screen, to identify phenotype-specific candidate effector genes. Methods We analyzed 14 candidate genes using whole-blood cis-eQTL data from eQTLGen (N = 31,684) and four Genetics of Osteoarthritis consortium phenotypes: AllOA (N = 826,690), KneeHipOA (N = 490,345), KneeOA (N = 396,054), and HipOA (N = 353,388). Evidence was integrated using cis-eQTL Mendelian randomization (MR), Bayesian colocalization (coloc.abf and coloc.susie), and summary-data-based MR with the HEIDI heterogeneity test. Each phenotype was analyzed separately, and gene–phenotype associations were classified using a predefined evidence framework. Results MAPK3 was the leading candidate. Higher genetically predicted blood expression was associated with lower OA risk, with negative MR and SMR estimates across all four phenotypes. Concordant evidence was observed for KneeHipOA, KneeOA, and AllOA, with PP.H4 values of 0.991, 0.979, and 0.945, respectively; MR-IVW p values were 7.1×10⁻⁴, 1.0×10⁻⁴, and 7.5×10⁻⁴, and HEIDI tests were non-significant. The HipOA MR-IVW association was not significant (p = 0.091). UQCC1 showed a risk-increasing direction and phenotype-dependent support, strongest for AllOA (PP.H4 = 0.919; SMR p = 7.6×10⁻¹²). Evidence was weaker for KneeHipOA and HipOA, while KneeOA findings suggested linkage rather than a shared causal variant. Conclusions Integrated genetic evidence prioritizes MAPK3, and secondarily UQCC1, as phenotype-dependent candidate OA targets. Because the exposure was whole-blood eQTL and the UQCC1 signal could not be fully distinguished from the adjacent GDF5 locus, these findings support further experimental validation rather than definitive causal-gene assignment.</p>