Abstract
BACKGROUND: Bladder cancer (BLCA) is a malignant tumor originating from the urothelial lining, characterized by a complex tumor microenvironment (TME) and heterogeneous tumor mutation burden (TMB). Cancer-associated epithelial cells (EpiCs) exhibit substantial heterogeneity during BLCA initiation and progression. Therefore, elucidating the diversity and functional states of EpiCs is essential for improving future diagnostic and therapeutic strategies. METHODS: We integrated multi-omics datasets, including 13 single-cell RNA-seq samples, 514 bulk transcriptome profiles, and 30 whole-exome sequencing (WES) samples, to comprehensively characterize EpiC subtypes. Nonlinear dimensional reduction (UMAP) and clustering analyses were performed to identify major epithelial subsets, followed by secondary clustering. TMB values calculated from self-generated WES data were incorporated into scAB and Ro/e algorithms to determine the TMB-associated epithelial subset, ultimately identifying the key cluster Epi14. Differentially expressed genes (DEGs) of Epi14 were analyzed, and CellChat was used to infer intercellular communication networks. CytoTRACE and Monocle2 were applied to assess stemness potential and differentiation trajectories. Random survival forest (RSF) combined with DEG analysis was used to identify hub genes. Immune infiltration, drug sensitivity, and functional pathway analyses were subsequently conducted. Spatial transcriptomics were deconvoluted using spacexr, CellChat, and PROGENy to map cellular composition, signaling activity, and pathway nodes. Finally, qPCR and Western blot assays were performed to validate hub gene expression in tumor versus adjacent tissues. RESULTS: A total of 77,263 cells and 3,000 highly variable genes were included, yielding 32 annotated cell clusters. Secondary clustering combined with WES-derived TMB identified 14 epithelial subpopulations, among which Epi14 was confirmed as the key TMB-associated subset using the Ro/e algorithm. Integration of DEGs, RSF, and multi-cohort datasets revealed ABRACL and ARPC3 as the pivotal hub genes, from which a risk-score model was constructed. Notably, ABRACL expression showed a strong positive association with tumor TMB and exhibited pronounced enrichment in spatial transcriptomic tumor regions. CONCLUSION: By integrating multi-omics and spatial datasets, this study reveals the epithelial heterogeneity of BLCA and identifies ABRACL and ARPC3 as key TMB-associated hub genes within EpiCs. The established risk-score model and validated functional markers provide valuable insights for future mechanistic studies and potential clinical translation in BLCA.