Single-cell transcriptomics deciphers cancer-associated fibroblast heterogeneity and immune regulatory mechanisms in the bladder cancer tumor microenvironment
TL;DR
Single-cell transcriptomics of publicly available BLCA data identifies two transcriptionally distinct fibroblast states in adjacent tissue and supports an exploratory ligand–receptor interaction framework, warranting prospective validation with primary CAF populations and adequately powered multi-specimen cohorts.
Abstract
Background Bladder cancer (BLCA) is a common malignancy, with muscle-invasive bladder cancer (MIBC) associated with a 5-year survival rate below 50%. Cancer-associated fibroblasts (CAFs) are heterogeneous stromal components of the tumor microenvironment (TME) that contribute to tumor progression, therapy resistance, and immune evasion. However, the molecular basis of CAF diversity and immune regulation in BLCA remains incompletely understood. Methods We analyzed single-cell transcriptomic data from GEO dataset GSE135337 using dimensionality reduction, unsupervised clustering, differential expression analysis, GO/KEGG/GSEA, diffusion pseudotime inference, and LIANA ligand–receptor analysis. TCGA-BLCA bulk transcriptomic data (n = 408 total; n = 401 after stage filtering) were used for score validation. RT–qPCR was performed to assess selected myCAF- and iCAF-associated genes in RT4 and T24 bladder cancer cells. Results Analysis of 9,253 post-QC cells across two non-paired specimens (6,035 adjacent-tissue; 3,218 tumor) yielded 9 DE-marker-annotated cell populations. Among 1,923 retained adjacent-tissue fibroblasts, two transcriptionally distinct states were identified: Homeostatic fibroblasts (n = 1,197; enriched for COL1A1, DCN) and CCL2-high activated-like fibroblasts (n = 726; CCL2 log2FC = 0.806, FDR = 8.08 × 10−11). No fibroblast-like cells were retained in the tumor specimen. Diffusion pseudotime positioned CCL2-high cells at a later within-sample state (median DPT 0.813 vs. 0.672; FBLN1 rho = −0.808). Exploratory LIANA analysis identified 1,347 fibroblast-outgoing ligand–receptor interactions. A pre-specified six-gene myCAF score (ACTA2, COL1A1, MMP11, MYL9, TAGLN, TPM2) tracked pathological stage in TCGA-BLCA (P = 1.60 × 10−12) and showed an unadjusted overall survival association (HR = 1.21, 95% CI: 1.02–1.42, P = 0.025) that attenuated to non-significance after age and stage adjustment (HR = 1.03, P = 0.748); the association was not replicated in GSE31684 (n = 93; log-rank P = 0.443). RT–qPCR confirmed higher expression of ACTA2, POSTN, MMP11, FAP, IL6, and CXCL12 in T24 versus RT4 cells (4.33–16.56-fold; all BH-adjusted q < 0.01). Conclusion Single-cell transcriptomics of publicly available BLCA data identifies two transcriptionally distinct fibroblast states in adjacent tissue and supports an exploratory ligand–receptor interaction framework. A six-gene myCAF-associated score tracks pathological stage but does not independently predict overall survival after covariate adjustment and was not replicated in an independent cohort. These findings constitute an exploratory computational framework warranting prospective validation with primary CAF populations and adequately powered multi-specimen cohorts.