Molecular Profiling of Small Cell Bladder Cancer Reveals Gene Expression Determinants of an Aggressive Phenotype

Abstract

SCBC is a rare variant of bladder cancer with relatively poor outcomes. Treatment options are often extrapolated from other anatomic sites. We performed molecular profiling to better characterize unique features in SCBC with the goal of identifying novel therapeutic targets and biomarkers for risk stratification. A retrospective analysis of 63 patients with biopsy-confirmed SCBC treated with radiotherapy, chemotherapy, and surgery at a single high volume academic center between 1994-2015. Small cell histology (SC%) was quantitatively determined by independent pathologic review. DLL3 and PD-L1 protein expression were measured by IHC in 53 patients. Formalin fixed and paraffin embedded tumor tissue was available for gene expression analysis in 39/63 primary tumors (median SC%=100%), 1 metastatic sample, and 6 adjacent normal samples (n=46 total) using the HTG EdgeSeq OBP assay with probes for 2568 genes. An average of 1.1x107mRNA probes were sequenced per sample and analyzed using the RNAseq workflow in the Partek Genomics Suite. Median age was 71 (n=63, range 39-90), 83% were men, SC% range was 5-100%, estimated median OS was 22.8 months. Unsupervised hierarchical clustering of gene expression patterns from 46 samples produced 4 distinct clusters; independent consensus clustering confirmed 4 stable and well-demarcated gene clusters. Patients with tumors in cluster 1 (where 5 of 6 normal samples were located) did not have metastases at diagnosis and did not develop distant recurrence, both of which were over-represented in the other 3 clusters. Kaplan-Meier analysis revealed a trend towards longer overall survival in cluster 1 patients (log rank p = 0.065). Higher gene expression of PRC1, NCAM1 (CD56) and DLL3 correlated with higher SC%, as did lower gene expression of ERBB2, PD-L1 and HPGD (p < 0.01). PD-L1 protein expression determined by IHC (≥1% cells) was noted in 30% of patients but did not correlate with outcome, SC%, DLL3 protein expression (IHC), or PD-L1 gene expression. DLL3 protein expression (≥1% cells) was noted in 68% of patients and DLL3 > 10% correlated with decreased OS (p = 0.03). Higher DLL3 protein expression correlated with DLL3 gene expression (Spearman r = 0.70, p < .01) and with SC% (r = 0.33, p = 0.01). This is the first study to reveal distinct gene expression patterns that define aggressive behavior, metastatic potential and outcomes in SCBC. The prognostic value of differential gene expression networks and the presence of underlying genetic and epigenetic alterations is the subject of ongoing prospective validation in a larger cohort.