Presentation Authors: Simpa Salami*, Jeremy Kaplan, Srinivas Nallandhingal, Ann Arbor, MI, Mandeep Takhar, San Diego, CA, Jeffrey Tosoian, Matthew Lee, Ann Arbor, MI, Junhee Yoon, Los Angeles, CA, Daniel Hovelson, Komal Plouffe, Samuel Kaffenberger, Arvin George, Jeffrey Montgomery, Matthew Davenport, Ann Arbor, MI, Sungyong You, Los Angeles, CA, Scott Tomlins, Nicole Curci, Ann Arbor, MI, Hyung Kim, Los Angeles, CA, Daniel Spratt, Aaron Udager, Ganesh Palapattu, Ann Arbor, MI
Introduction: Up to 20% of patients with negative multiparametric magnetic resonance imaging (MRI) harbor Gleason score â‰¥7 prostate cancer (PCa). In this study, we sought to elucidate the molecular basis of and determine the prognostic significance of PCa visibility on MRI.
Methods: We identified a retrospective cohort of patients who underwent MRI prior to radical prostatectomy with both MRI visible (PIRADS 3 - 5) and invisible PCa. MRI for each patient was re-reviewed and co-registered with whole-mount histopathology. DNA and RNA were co-isolated from all tumor foci pre-identified on formalin-fixed paraffin-embedded specimens. High depth, targeted DNA and RNA next generation sequencing was performed to characterize the molecular profile of each tumor focus using the Oncomine Comprehensive Panel (DNA) and a custom targeted RNAseq panel assessing PCa relevant alterations. A multigene RNAseq model was developed and validated in two independent cohorts to predict MRI visible PCa and to determine the prognostic significance of MRI visibility.
Results: A total of 26 primary tumor foci from 10 patients were analyzed. The median number of PCa foci was 3. Of the 14 (54%) invisible lesions on MRI, 5 (36%) were Gleason 3+4=7 and the remainder were Gleason 6. We detected high-confidence prioritized mutations in 54% (14/26) of tumor foci, 43% (6/14) of which were in MRI invisible lesions. Notable point mutations were in APC, AR, ARID1B, ATM, ATRX, BRCA2, FAT1, MAP3K1, NF1, SPEN, SPOP, TP53, and a frameshift mutation was detected in SOX2. A 9-gene RNA signature, the majority of which were under-expressed cellular organization and structure genes, was developed to predict MRI visible tumors with an AUC of 0.89. Validation of this signature in an independent data set (n = 16) yielded an AUC of 0.88 with a specificity of 100% for predicting MRI visible tumors (Figure 1). Using this RNAseq signature in a cohort of 375 patients with clinical follow up, we found that predicted MRI visibility was not an independent predictor of biochemical recurrence, metastasis-free survival, or PCa specific mortality (all p > 0.05).
Conclusions: We observed under-expression of cellular organization and structural genes in MRI visible tumors compared to MRI invisible lesions. Using our validated signature to predict MRI visibility status, we found that MRI visibility is not a significant predictor of oncological outcomes.
Source of Funding: Prostate Cancer Foundation