Using machine learning analysis of RNA sequencing data, the researchers identified four molecular subtypes with distinct tumor microenvironments: luminal desert (33%), stromal (23%), immune (20%), and basal (24%). Two subtypes showed a significant overall survival benefit with atezolizumab compared to standard of care - the immune subtype (HR 0.67, p=0.001, median OS 17.5 vs. 11.3 months) and basal subtype (HR 0.72, p=0.002, median OS 10.3 vs. 8.3 months). The immune subtype was characterized by high lymphocyte infiltration, including T cells, NK cells, and B/plasma cells, while the basal subtype showed increased neutrophil frequency and PD-L1 expression. Importantly, digital pathology analysis of H&E slides using deep learning achieved 86% accuracy in predicting molecular subtypes. Somatic mutation analysis revealed subtype-specific alterations, with FGFR3 mutations enriched in luminal desert tumors (p=5.23e-26), and TP53/RB1 mutations more common in immune and basal subtypes (p=4.84e-17 and p=8.27e-09 respectively).
This study highlights that both tumor microenvironment and cancer cell-intrinsic properties are important for response to immunotherapy and that molecular subtyping incorporating these elements could guide the selection of patients most likely to benefit from immune checkpoint inhibition. The immune and basal subtypes were most likely to benefit from PD-L1 blockade. Additional therapeutic targets were identified for each subtype, including FGFR3 inhibition for luminal desert tumors. Future investigation of methods to increase immune cell trafficking and infiltration into immune-excluded tumors is warranted to improve responses.
Written by: Bishoy M. Faltas, MD, Director of Bladder Cancer Research, Englander Institute for Precision Medicine, Weill Cornell Medicine
Reference:
- Hamidi H et al. (2024). Molecular heterogeneity in urothelial carcinoma and determinants of clinical benefit to PD-L1 blockade. Cancer Cell 42:1-15.
Read the Abstract