Cell-of-origin and genetic drivers define advanced bladder cancer subtypes and potential therapeutic response in mouse models
Bladder cancer (BC) remains a major clinical challenge owing to its high recurrence, limited treatment options, and molecular heterogeneity. Despite recent therapeutic advances, prognosis remains poor, and resistance is frequent, underscoring the need for improved experimental models to study tumorigenesis and therapeutic response. A key limitation of advanced BC research is the scarcity of in vivo models that accurately reflect invasive disease, with even fewer capturing the complexity of metastasis. To investigate how the cell-of-origin and specific combinations of driver mutations influence in bladder tumorigenesis, we developed and characterized four genetically engineered mouse models of advanced BC by targeting two combinations of tumor suppressor genes (Pten and Trp53, or Pten, Trp53, Rb1, and Rbl1) in basal or suprabasal urothelial cells through intravesical of Cre-adenovirus delivery. Loss of the retinoblastoma family reduced cancer-specific survival and was associated with more differentiated carcinomas. In both genetic backgrounds, luminal-derived tumors developed earlier but showed fewer metastatic events. Histopathological and transcriptomic analyses revealed that these tumors resemble human basal-squamous and stroma-rich subtypes, sharing regulatory networks and activated signaling pathways with human invasive tumors. Notably, tumors lacking retinoblastoma family genes exhibited increased immune infiltration, reinforcing the value of these models for diverse preclinical applications. To overcome detection and latency limitations, we established tumor-derived cell lines and generated syngeneic graft models. These were validated as preclinical platforms, exhibiting therapeutic responses to CDK4/6 inhibition and anti-PD-L1 immunotherapy. Our findings highlight the value of these novel models for studying BC progression and evaluating emerging therapeutic strategies in immunocompetent settings.