PD53-01: Development of a Whole-Urine, Multiplexed, Next Generation RNA-Sequencing Assay for Aggressive Prostate Cancer Early Detection
Friday, May 15, 2020
7:00 AM – 9:00 AM
Andi K. Cani, Kevin Hu, Javed Siddiqui, Yingye Zheng, Sumin Han, Chia-Jen Liu, Daniel H. Hovelson, Srinivas Nallandhighal, Trinh Pham, Lanbo Xiao, Heng Zheng, Jeffrey J. Tosoian, Ganesh S. Palapattu, Todd M. Morgan, Aaron Udager, John T. Wei, Arul M. Chinnaiyan, Scott A. Tomlins, Simpa S. Salami
Introduction: Despite advances in biomarker development, early detection of aggressive prostate cancer (PCa) remains challenging. We previously developed the Michigan Prostate Score (MiPS) for individualized risk prediction of aggressive prostate cancer. MiPS uses transcription-mediated amplification to quantify expression of TMPRSS2:ERG (T2:ERG) and PCA3 from whole urine obtained after a digital rectal exam (DRE), combined with serum PSA. To improve upon MiPS, herein we describe the pre-clinical development and validation of a targeted next generation RNA sequencing assay (NGS-MiPS).
Methods: We selected patients with available MiPS scores as well as representing a spectrum of disease grade on biopsy (Benign to Grade Group 5). We used 2.5 mL of post-DRE whole urine to asses ~90 PCa transcriptomic biomarkers: including T2:ERG, PCA3, and additional isoforms of common PCa gene fusions, mRNAs, lncRNAs, and expressed mutations.
Results: NGS-MiPS showed a 98% informative sample rate, high technical reproducibility, robustness and concordance with orthogonal methods (TMA and RT-qPCR), and was able to detect an expressed HOXB13 p.G84E variant and SPOP mutations. NGS-MiPS accurately recapitulated clinical MiPS-measured risk scores for presence of PCa or high-grade PCa (Gleason Score >6) on biopsy as determined by clinical MiPS vs. the same model but with NGS-MiPS data. In an extreme design cohort (Benign or Gleason 6 vs. Gleason = 4+3=7 cancer) NGS-MiPS showed expected differences in the levels of T2:ERG T1E4 (p<0.00001) and PCA3 (p=0.02), with additional T2:ERG splice isoforms and other biomarkers also showing significantly different expression between low vs. high grade cancer. We used a machine learning approach trained on a subset of the extreme design cohort (n=73) to generate a 29-transcript model that outperformed MiPS and serum PSA in two validation cohorts: 1. A held-out set from the extreme design cohort n=36, (AUC 0.82 vs. 0.73 and 0.69, respectively); and 2. A separate PCa active surveillance cohort n=45, (AUC 0.66 vs. 0.58 and 0.53, respectively).
Conclusions: These results support the potential utility and continued development of our urine based targeted NGS assay to supplement serum PSA for improved early detection of aggressive prostate cancer. Source of
Funding: University of Michigan Prostate Cancer S.P.O.R.E. Grant P50 CA186786-05.
AKC was supported by the NIH Training Program in Translational Research T32-GM113900 and by the University of Michigan Precision Health, 2018 Scholars Awards.
TMM and SAT have been supported by the A. Alfred Taubman Biomedical Research Institute.