Purpose: Previous investigations successfully uncovered molecular characteristics of bladder tumors (BT), dealing with non-muscle-invasive (NMIBC) and muscle-invasive (MIBC) bladder cancers, separately. At the molecular level, however, there is a great need to aggregate these subtypes, which may share biological characteristics. With recent publications of large-scale genomic data, we aim to identify distinct molecular subtypes of BT and their prognostic values.
Experimental Design: A total of 1636 BT samples from six patient cohorts including The Cancer Genome Atlas (TCGA) and the multicenter European UROMOL consortiums were used. We constructed a subtype classifier by applying Significance Analysis of Microarrays (SAM) and Prediction Analysis of Microarrays (PAM) algorithms. Various statistical methods, including Chi-square and log-rank tests, were applied to estimate an association between subtypes and mutation/prognosis.
Results: Hierarchical clustering revealed four molecular subtypes of BT with different clinical outcomes: Class 1 with low-grade NMIBC and the best prognosis; Class 2 characterized by active FGFR3 and inhibited immune pathways; Class 3 with high-grade NMIBC and the worst progression-free survival; and Class 4 mainly comprised of MIBC along with EMT activation. By applying the classifier based on these characteristics, we stratified all BT samples into newly identified molecular subtypes. When comparing previously reported subtypes, our subtypes well agreed with their molecular characteristics regardless of breast cancer-based biology, and showed a strong prognostic relevance in Class 3. Integrative analysis of mutation and gene expression suggested that Class 3 may have the potential benefit from anti-PD-L1 immunotherapy.
Conclusions: Our classifier, constructed by NMIBC and MIBC integration, successfully stratified BT patients into distinct subtypes with different clinical outcomes and a possible treatment option.