Background <p>Adenosine-related RNA modifications, including N6-methyladenosine (m⁶A), N1-methyladenosine (m¹A), alternative polyadenylation (APA), and adenosine-to-inosine (A-to-I) editing, critically regulate transcriptional programs in cancer. However, their coordinated contribution to bladder cancer biology and response to immunotherapy remains poorly understood.</p> Methods <p>We systematically characterized four classes of RNA modification “writer” genes in bladder cancer, assessing somatic mutations, copy-number variations, and mRNA expression patterns across seven independent cohorts, including TCGA-BLCA, E-MTAB-4321, and GEO datasets (GSE32894, GSE13507, GSE48276, GSE70691, and GSE69795). Unsupervised consensus clustering identified RNA modification-based molecular subtypes, which were used to construct an RNA modification-based prognostic and immunotherapy response score (RMP_Score). This score was validated across multiple external cohorts, including the IMvigor210 anti-PD-L1 dataset. Differentially expressed genes (DEGs) between subtypes were assessed and integrated through Kaplan-Meier survival analysis and Cox proportional hazards regression to construct the RMP_Score. Bulk and single-cell transcriptomic analyses, complemented by in vitro functional assays, were performed to investigate the role of carboxylesterase 1 (CES1), a key component of the RMP_Score. Tumor microenvironment (TME) composition was quantified using single-sample gene set enrichment analysis (ssGSEA).</p> Results <p>RNA modification “writer” genes exhibited frequent somatic mutations and copy-number amplifications, which were strongly associated with transcriptional dysregulation and inferior overall survival. Unsupervised clustering revealed two distinct subtypes: a metabolism-enriched cluster associated with favorable prognosis, and an immune- and stroma-enriched cluster characterized by abundant regulatory T-cell infiltration and adverse clinical outcomes. The RMP_Score robustly stratified patient survival across six independent cohorts, correlated with differential sensitivity to inhibitors targeting the JNK/p38, PI3K/mTOR, and WNT pathways, and accurately predicted clinical responses to anti-PD-L1 immunotherapy, with low RMP_Scores demonstrating superior responses. CES1 emerged as a robust predictor of poor survival and immunotherapy resistance, closely associated with tumor progression, extracellular matrix remodeling, and an immunosuppressive, fibroblast-rich TME. Functionally, CES1 silencing significantly attenuated bladder cancer cell viability, invasive capacity, and migratory potential in vitro.</p> Conclusions <p>These findings established an RNA modification-based framework for molecular classification and risk stratification in bladder cancer, and identified CES1 as a metabolically driven regulator of the TME with potential utility as a prognostic and predictive biomarker for immune checkpoint blockade.</p>