Integrated post-GWAS, single-cell, and functional analyses prioritize TCTN2 at an osteoarthritis locus associated...

Post-genome-wide association study (GWAS) interpretation of osteoarthritis (OA) increasingly requires integration with single-cell biology because many risk loci likely act through sterile inflammatory remodeling, regulatory and inflammatory chondrocyte states, and tissue-stress biology rather than through large disease-stage expression shifts. We therefore asked whether a published Transformer-based post-GWAS prioritization strategy could refine a multi-gene chr12 susceptibility locus into a tractable TCTN2-centered effector-gene hypothesis.

Using combined hip and knee osteoarthritis summary statistics (213,839 cases and 1,080,481 controls), we built a transfer-learning variant-prioritization layer that scored 20,651,210 variants with 21 engineered features, combined these with LDSC, positional MAGMA, fibroblast and muscle eMAGMA, six-tissue SMR, and fine-mapping, and then integrated the prioritized locus with a 56,000-cell chondrocyte atlas, independent OA cartilage expression resources, OA primary-tissue eQTL maps, and ATDC5 perturbation experiments.

The prioritization model, used solely as a ranking layer, assigned a high prioritization score (≥ 0.99) to 17,150 variants, including the chr12 lead rs11611450 (prioritization score 0.999853), and highlighted 6,275 additional non-genome-wide-significant high-priority variants. Positional MAGMA nominated eight Bonferroni-significant genes, and eMAGMA nominated four genes in skeletal muscle and seven in fibroblasts. Within the rs11611450-linked five-gene cluster, TCTN2 showed the strongest cross-layer convergence, with SMR support in five of six tissues, fibroblast eMAGMA support (P = 8.34 × 10^-10), hypertrophic-chondrocyte enrichment (z = 2.20), and localization to regulatory and inflammatory chondrocyte states. Independent validation showed modestly higher TCTN2 abundance in OA than in normal cartilage (delta = 0.45, P = 2.8 × 10^-2), no high-grade-versus-low-grade cartilage shift (logFC = 0.053, FDR = 1.0), and strong OA-cartilage eQTL support in both low-grade and high-grade cartilage (q = 4.86 × 10-⁴ and 1.13 × 10^-2). In ATDC5 cells, Tctn2 knockdown reduced expression to 0.21-fold of Mock and increased Col10a1, Mmp13, and Runx2 to 2.71-, 2.98-, and 3.53-fold, respectively, whereas overexpression and rescue reversed the same markers.

This staged analysis nominates TCTN2 as a plausible candidate within an unresolved chr12 osteoarthritis locus; ATDC5 perturbation indicates that TCTN2 can modulate hypertrophic markers. Because the human-tissue association is directionally discordant, we present a locus-to-cell-state hypothesis rather than a complete genetics-to-mechanism chain, with cross-sectional shifts remaining modest and context-dependent.