DIA Proteomics of FFPE Renal Biopsies Reveals Two Molecular Subtypes of Lupus Nephritis and Identifies APOL1 As Candidate Biomarker for Stratification

Introduction Lupus nephritis (LN) exhibits substantial clinical and pathological heterogeneity. We aimed to define proteomics-based molecular subtypes of LN and identify candidate biomarkers for subtype discrimination. Methods We analysed formalin-fixed paraffin-embedded (FFPE) renal biopsy specimens from 292 patients with biopsy-proven LN from four tertiary hospitals using data-independent acquisition (DIA)-liquid chromatography–tandem mass spectrometry (LC-MS/MS) proteomics. Molecular subtypes were identified by non-negative matrix factorisation. Differential proteins, functional enrichment, immune pathway activity, protein–protein interaction networks and subtype-associated clinical/pathological features were evaluated. Extreme Gradient Boosting (XGBoost) with SHapley Additive exPlanations (SHAP) and Least Absolute Shrinkage and Selection Operator (LASSO) logistic regression were used to identify key subtype-related features and derive a protein panel distinguishing proliferative (class III/IV) from membranous (class V) LN. Results Two stable molecular subtypes were identified, with 1002 differential proteins between them. Subtype_2 was enriched for interferon-related innate immunity, complement activation, phagocytosis-endocytosis-lysosome pathways and ribosome biogenesis/RNA metabolism, whereas Subtype_1 was characterised by keratinisation and epithelial structural remodelling. Subtype_2 was associated with higher serum creatinine, lower estimated glomerular filtration rate and higher chronicity index. APOL1 showed discriminatory value between subtypes, and serum ELISA demonstrated a consistent pattern with FFPE proteomic findings. A five-protein LASSO panel achieved an area under the curve of approximately 0.76 for distinguishing class III/IV from class V LN. Conclusion DIA-based proteomic profiling of FFPE renal biopsies identifies biologically and clinically relevant LN molecular subtypes and may support tissue-informed classification and risk stratification.