Abstract
BACKGROUND: Cancer-associated fibroblasts (CAFs) are key drivers of tumor progression in bladder cancer (BLCA), yet their molecular heterogeneity and prognostic utility remain incompletely characterized. Single-cell studies have revealed distinct CAF subpopulations with divergent clinical impacts, necessitating refined prognostic frameworks that capture CAF-mediated progression. METHODS: We analyzed single-cell RNA sequencing data (GSE267718) from 8 BLCA patients to identify CAF populations and define progression-associated gene signatures. Using 359 TCGA-BLCA samples as the training cohort, we performed non-negative matrix factorization (NMF) consensus clustering on 85 prognostically significant CAF genes, identifying two molecular clusters with distinct survival outcomes. Through LASSO-Cox regression and stepwise selection, we constructed a four-gene Tumor-Progressing Fibroblast Riskscore model comprising FOXA1, TBX3, LRIG1, and RNF11. Model performance was validated in the E-MTAB-4321 cohort (n = 476). Functional validation of RNF11 was performed using shRNA-mediated knockdown in T24 and 5637 bladder cancer cell lines, followed by proliferation, migration, invasion assays, and transcriptomic profiling. RESULTS: Single-cell analysis identified 557 differentially expressed genes between non-muscle-invasive bladder cancer and muscle-invasive bladder cancer CAFs. NMF clustering stratified TCGA patients into 2 clusters with significantly different overall survival. The TPFR model showed consistent prognostic performance in both training and validation cohorts, with high-risk patients showing significantly worse survival. Functional enrichment analysis revealed that TPFR scores correlated with ECM-receptor interaction, focal adhesion, and cytoskeletal regulation pathways. Stratified analysis revealed superior model performance in elderly (>60 years), male, and early-stage patients. In particular, RNF11 knockdown significantly reduced proliferation, migration, and invasion in 5637 and T24 cells, while transcriptomic analysis revealed alterations in tumors after RNF11 knockdown including TNF and MAPK signaling pathway, indicating a potential mechanism by which RNF11 regulates bladder cancer progression. CONCLUSION: We established a CAF-based prognostic model that integrates single-cell insights with bulk transcriptomics for robust risk stratification in BLCA. The TPFR model shows clinical utility particularly in elderly and early-stage patients. Functional characterization showed that RNF11 regulates proliferation and migration of bladder cancer. These findings highlight the prognostic value of CAF signatures and provide a framework for precision medicine approaches in bladder cancer management.