Large-scale plasma proteomics identifies prediagnostic patterns and implicates NECTIN2 in age-related macular degeneration

Background Age-related macular degeneration (AMD) develops years before clinical diagnosis; however, the prediagnostic plasma proteomic landscape remains poorly characterized. This study aimed to characterize plasma proteomic patterns in relation to time to AMD diagnosis and to identify candidates relevant to AMD pathogenesis. Methods This prospective cohort study analyzed plasma proteomic data from 45,379 AMD-free UK Biobank participants (median follow-up 15.2 years) using Cox proportional hazards models, LOESS-based trajectory clustering, segmented regression, and repeated-subsampling LASSO stability selection. The lead candidate was further evaluated through OCT-derived retinal structure analysis, human macular single-nucleus transcriptomics, and a laser-induced choroidal neovascularization (CNV) mouse model with siRNA- and AAV-mediated silencing. Results Ninety-two proteins were associated with incident AMD in the fully adjusted Cox analysis. A separate effect-size-based selection retained 98 proteins for trajectory analysis; k-means clustering identified three distinct population-level patterns. Segmented regression identified a breakpoint in the mean Cluster 3 trajectory 8.81 years before diagnosis. In internal cross-validation, the combined protein-demographic model achieved an AUC of 0.81 for AMD diagnosed within 3 years of baseline. NECTIN2 demonstrated convergent support from a significant Cox association, Cluster 3 membership, and the highest cross-window LASSO stability score among the panel proteins. Higher plasma NECTIN2 levels were correlated with reduced thickness of photoreceptor-associated and RPE-adjacent retinal layers. Human macular snRNA-seq analysis showed prominent AMD-associated upregulation of NECTIN2 in the retinal pigment epithelium. In a laser-induced CNV mouse model, NECTIN2 silencing before or immediately after laser injury reduced lesion size, vascular leakage, and the expression of Vegfa, Il6, and Mmp9. Conclusions AMD-associated plasma proteomic patterns varied with time to diagnosis, with a transition in the mean Cluster 3 trajectory approximately 9 years before diagnosis. NECTIN2 emerged as a candidate supported by observational and experimental evidence and warrants further validation.