Glucocorticoid-Modulated Molecular Signatures and Prediction of Treatment Efficacy in IgA Nephropathy
Lan Wang, Yu He, Wenmin Tian, Jinwei Wang, Sufang Shi, Lijun Liu, Helen Monaghan, Muh Geot Wong, Jicheng Lv, Yang Chen, Yuemiao Zhang, Hong ZhangBackground:
Glucocorticoids remain an important anti-inflammatory therapy for IgA nephropathy but are associated with substantial treatment-related adverse effects, highlighting the urgent need for early prediction of treatment efficacy. We hypothesized that randomized controlled trial-based serum proteomic profiling could delineate glucocorticoid-modulated proteins and more accurately stratify patients likely to benefit from glucocorticoid therapy.
Methods:
We performed proteomic analyses of samples from the TESTING trial, in which patients with IgA nephropathy were randomized to receive methylprednisolone or placebo. 479 longitudinal serum samples from 241 Chinese participants collected at baseline and at 6 and 12 months were analyzed. Linear mixed effect models were used to identify glucocorticoid-modulated proteins and pathways in the methylprednisolone group relative to placebo. Cox proportional hazards and penalized ridge regression models were used to prioritize proteins and develop a prediction model for therapeutic efficacy. The candidate proteins were further validated by ELISA.
Results:
Among more than 1,500 detected proteins, 302 were glucocorticoid-modulated and enriched in pathways of cytoskeletal stabilization and immunometabolism. Notably, pathways associated with long-term eGFR decline, such as complement activation and endothelial injury, were not modulated by glucocorticoids. Among the glucocorticoid-modulated proteins, Cox proportional hazards and penalized ridge regression models identified those whose baseline levels improved prediction of treatment efficacy compared to clinical variables alone. ELISA validated selected candidate proteins LEP and C1QTNF5. The prediction model incorporating these two proteins significantly outperformed the traditional clinical model in predicting glucocorticoid efficacy (5-year AUC, 0.85 [95% confidence interval (CI), 0.72 to 0.88] versus 0.80 [95% CI, 0.78 to 0.91]).
Conclusions:
Data from this randomized trial-based proteomic study helped delineate glucocorticoid-modulated proteins and pathways, and identified robust protein biomarkers that enabled stratification of patients likely to benefit from glucocorticoid therapy.