Mapping of Multilineage Tumor Cell Populations in Mouse Bladder Cancer

Abstract

In bladder cancer (BLCA), the cellular expression of lineage markers with predictive biomarker function can have important implications for tumor progression, treatment response, and survival. However, tumor heterogeneity and the coexistence of distinct tumor subpopulations can complicate the utility of using a primary, clinically assigned tumor signature. In this report, we have applied the reference atlas, Tabula Sapiens, and deep learning model, UniCell, to conduct in-depth transcriptional analysis of unique lineage marker-defined cell clusters in a carcinogen (BBN) induced bladder cancer model. UniCell deconvolution has identified tumor populations, including urothelial, adenocarcinoma, squamous cell carcinoma, and mesenchymal tumor populations, each with cell-intrinsic gene expression signatures relevant to human BLCA progression. The identified tumor clusters contain uniquely basal, luminal, stromal, or hybrid cells in cell lineage marker expression. To understand the significance of these populations during progression, we used trajectory and pseudo-time analysis to show that cells uniquely basal are plastic in lineage identity and can evolve to form tumor populations composed of other lineage marker-defined signatures. Finally, pathway and drug enrichment analysis of tumor cell clusters were used to identify therapeutics that may preferentially target the identified tumor cell populations. These data collectively define a molecular template that may uniquely profile molecular plasticity occurring during progression and response to therapy with important implications for human disease. SignificanceTumor heterogeneity is a mechanism for treatment resistance. Our study defines unique tumor subpopulations having differential therapeutic sensitivities and potential for lineage plasticity. Our modeling may impact the treatment of BLCA patients.