Deep Sequencing of Urinary RNAs for Bladder Cancer Molecular Diagnostics.

Abstract

<b>Purpose:</b> The majority of bladder cancer patients present with localized disease and are managed by transurethral resection. However, the high rate of recurrence necessitates lifetime cystoscopic surveillance. Developing a sensitive and specific urine-based test would significantly improve bladder cancer screening, detection, and surveillance.<b>Experimental Design:</b> RNA-seq was used for biomarker discovery to directly assess the gene expression profile of exfoliated urothelial cells in urine derived from bladder cancer patients (<i>n</i> = 13) and controls (<i>n</i> = 10). Eight bladder cancer specific and 3 reference genes identified by RNA-seq were quantitated by qPCR in a training cohort of 102 urine samples. A diagnostic model based on the training cohort was constructed using multiple logistic regression. The model was further validated in an independent cohort of 101 urines.<b>Results:</b> A total of 418 genes were found to be differentially expressed between bladder cancer and controls. Validation of a subset of these genes was used to construct an equation for computing a probability of bladder cancer score (P_BC) based on expression of three markers (<i>ROBO1, WNT5A</i>, and <i>CDC42BPB</i>). Setting P_BC = 0.45 as the cutoff for a positive test, urine testing using the three-marker panel had overall 88% sensitivity and 92% specificity in the training cohort. The accuracy of the three-marker panel in the independent validation cohort yielded an AUC of 0.87 and overall 83% sensitivity and 89% specificity.<b>Conclusions:</b> Urine-based molecular diagnostics using this three-marker signature could provide a valuable adjunct to cystoscopy and may lead to a reduction of unnecessary procedures for bladder cancer diagnosis. <i>Clin Cancer Res; 23(14); 3700-10. ©2017 AACR</i>.