Plasma Proteomics Identifies Molecular Signatures and Biomarkers Associated with High Structural Damage in Ankylosing Spondylitis

OBJECTIVE: Pathological bone formation in ankylosing spondylitis (AS) is not fully reflected by systemic inflammation, limiting the utility of inflammatory markers for structural damage assessment. This study aimed to identify plasma protein signatures associated with structural damage severity in AS. METHODS: Deep plasma proteomics was conducted in a prospective 88-participant discovery cohort comprising 25 and 31 patients with high and low structural damage, respectively, stratified by the SPARCC sacroiliac joint structural score, and 32 normal controls. Age- and sex-adjusted differential expression analysis, WGCNA, and integrated machine learning were used to characterize damage-associated molecular programs and prioritize biomarkers. Five candidate proteins were validated by ELISA in an independent 88-participant cohort (32 high, 32 low, and 24 normal controls). RESULTS: Proteomic profiling identified 536 differentially expressed proteins between the high structural damage group and normal controls. WGCNA revealed a structural-damage-associated module, enriched in platelet activation and cell-matrix adhesion, together with an AS-associated reduction in a module related to RNA processing and translation. Integrated analysis prioritized a five-protein signature comprising ITGA2B, HSPG2, ITGB1, EPHB2, and SAA2. The combined model achieved an AUC of 0.917 in the validation cohort and significantly outperformed individual biomarkers (AUCs 0.732-0.819; all Holm-adjusted p < 0.05). CONCLUSION: This study supports a dual-pathway model in which SAA2-marked systemic inflammation and ITGA2B-associated platelet-ECM remodeling appear partially dissociated yet remain biologically interconnected in AS. The validated five-protein panel provides a noninvasive tool for identifying patients with current high structural damage and complementing imaging-based assessment by capturing biological heterogeneity.