Specific micro-RNA expression patterns distinguish the basal and luminal subtypes of muscle-invasive bladder cancer.

Andrea E Ochoa,Colin Dinney,Xiaoping Su,Bogdan Czerniak,Woonyoung Choi,David J Mcconkey,Arlene Siefker-Radtke

doi:10.18632/oncotarget.13284

Abstract

The roles of non-coding RNAs in controlling clinical and biological heterogeneity in bladder cancer remain unclear. We used TCGA's published dataset (n = 405 tumors) as a discovery cohort and created a new validation cohort to define the miRNA expression patterns in the basal and luminal molecular subtypes of muscle-invasive bladder cancer (MIBC). We identified 63 miRNAs by PAM, which optimally identified basal and luminal tumors. The targets of the top luminal miRNAs were activators of EMT (ZEB1, ZEB2) and basal subtype transcription (IL-6, EGFR, STAT3), whereas the targets of the top basal miRNAs were involved in adipogenesis pathways and luminal breast cancer (ERBB2, ERBB3). We also identified a 15-miRNA signature that identified stromally infiltrated basal and luminal MIBCs corresponding to the “cluster IV/immune undifferentiated/claudin-low” and “cluster II/luminal immune” subtypes identified previously, which likely contain samples with higher infiltration rates. Using the 63-miRNA signature, we accurately assigned MIBCs to the basal and luminal subtypes and confirmed that patients with basal tumors had shorter overall survival. The results strongly suggest that miRNAs contribute to the control of the gene expression patterns observed in basal and luminal MIBCs and that they can be used as biomarkers and candidate therapeutic targets.

Highlights

Muscle-invasive bladder cancer (MIBC) is the fourth most common cancer type in men in the United States, occurring less frequently in women
Consistent with our previous conclusions [14], the results revealed that a three cluster (k = 3) solution best fit the data mathematically (Figure S1A), and the biomarkers that were enriched in each of the 3 consensus clustering (CC) subtypes overlapped significantly with those associated with the basal, p53like, and luminal subtypes we had identified previously (Figure S1)
Consistent with this idea, using a k = 2 solution (Figure 1A), we observed excellent overlap (93%) with subtype assignments made using an independent basal/luminal (k = 2) Prediction analysis of microarrays (PAM) classifier (BASE47) [6], and when the one-nearest neighbor (oneNN) p53-like tumors were omitted, we observed an overlap of 93% with the oneNN basal and luminal subtype assignments (Figure 1B)

Summary

Introduction

Muscle-invasive bladder cancer (MIBC) is the fourth most common cancer type in men in the United States, occurring less frequently in women. Basal cancers were enriched with squamous histopathological features [4, 5, 7], whereas luminal cancers were enriched with papillary histopathological features and activating DNA mutations and fusions involving fibroblast growth factor receptor-3 (FGFR3) [4, 5, 9]. It appears that basal and luminal MIBCs have very distinct clinical and biological properties and should be managed as distinct disease entities. There is strong interest in developing reliable clinical tools that can be used to accurately assign MIBCs to the basal and luminal subtypes

Objectives

Methods

Results

Conclusion