Bladder cancer variants share aggressive features including a CA125+ cell state and targetable TM4SF1 expression

Abstract

Histologic variant (HV) subtypes of bladder cancer are clinically aggressive tumors that are more resistant to standard therapy compared to conventional urothelial carcinoma (UC). Little is known about the transcriptional programs that account for their biological differences. Here we show using single cell analysis that HVs harbor a tumor cell state characterized by expression of MUC16 (CA125), MUC4, and KRT24. This cell state is enriched in metastases, predicted to be highly resistant to chemotherapy, and linked with poor survival. We also find enriched expression of TM4SF1, a transmembrane protein, in HV tumor cells. Chimeric antigen receptor (CAR) T cells engineered against TM4SF1 protein demonstrated in vitro and in vivo activity against bladder cancer cell lines in a TM4SF1 expression-dependent manner, highlighting its potential as a therapeutic target.

Fig. 1. Top level clustering analysis of tumor epithelial cells and characterization of a common tumor cluster.

A Clustering UMAPs of tumor epithelial cells (N = 8553) extracted from the main dataset color-coded by cluster and annotated according to tumor ID. Cluster 13 (ellipse) is annotated separately due to contributions from multiple tumors. B Cluster composition by patient/tumor. C Table displaying primary and secondary histologic patterns observed in each tumor and percentage of cells within Cluster 13. D Curated dot plot of top differentially expressed genes (DEGs) by tumor cluster. E Serial section immunohistochemistry of MUC4, CA125, and KRT24 in variant tumors (N = 2 of 2 cases). F Representative immunohistochemistry of CA125 in multiple histologic variants. All scale bars = 50 µm. G Preoperative serum CA125 values in bladder cancer patients stratified by tumor grade and histology (Two-way Mann–Whitney U-test, p = 0.007). Quartiles, centers and outliers are shown in the box plot. Source data are provided as a Source data file.

Fig. 2. Transcriptional relationship between Cluster 13 and parent tumor cells.

A Partition-based graphical abstraction of tumor cell clusters. B Dot plot of tumor signature scores relative to Cluster 13 tumors of origin. C UMAP of individual tumors color-coded by Cluster 13 cells (red) and pseudotime using Cluster 13 cells as the starting point. D Expression along the pseudotime of Cluster 13 and parent tumor DEGs. Data are also presented as fitted curves with confidence intervals. Source data are provided as a Source data file.

Fig. 3. Cluster 13 is associated with metastasis and chemotherapeutic resistance.

A Gene ontology analysis of Cluster 13 gene signature. B Representative CA125 immunohistochemistry in a primary HV bladder tumor and the corresponding lymph node metastasis (N = 4 of 5 cases examined). C Intratumoral comparison of Cluster 13 and parent tumor gene signature susceptibility to common bladder cancer chemotherapeutic agents in 5 HV tumors. D Kaplan–Meier curves of overall and disease-specific survival according to Cluster 13 signature enrichment in TCGA-BLCA (log-rank test, p = 0.01 for overall survival; p = 0.03 for disease-specific survival). Data are presented as mean values ± standard error. Source data are provided as a Source data file.

Fig. 4. Identification of TM4SF1 as a gene enriched in HVs.

A Volcano plot comparison of all UC and HV cells after downsampling (N = 150 cells per patient) (p-values are computed from the two-sided Wilcoxon rank-sum test). B Violin plots of TM4SF1 expression by tumor cluster. Source data are provided as a Source data file. C, D Immunohistochemistry of TM4SF1 in a validation cohort of HV and UC C primary tumors and D lymph node metastases. E Semiquantitative comparison of TM4SF1 staining in HV and UCs (Fisher’s exact test, p = 0.02). Scale bars = 50 µm.

Fig. 5. Efficacy of TM4SF1 CAR T cells in vivo and in vitro.

A Quantification of in vitro TM4SF1-CAR1 and CAR2 activity against bladder cancer cell lines using IncuCyte co-culture assay with 1:1 effector:target cell ratio. B In vitro co-culture assays comparing CAR activity (2:1 effector:target cell ratio) in UMUC3 cell lines with and without CRISPR-mediated TM4SF1 knockout (N = 1 replicate per condition, 5 conditions per experiment). C Evaluation of CAR1 tumor-killing efficacy in an in vivo xenograft model using UMUC3 (N = 9 mice) and 253JBV (N = 10 mice) cell lines. Data are presented as mean values ± standard error bars. D Kaplan–Meier survival analysis of CAR1 treated and untreated mice with UMUC3 (N = 9 mice) and 253JBV (N = 10 mice) tumors. E:T Effector:Target, UTD untransduced, CAR chimeric antigen receptor.