RNAseq Analysis of Formalin Fixed Paraffin Embedded Radical Prostatectomy Specimens Identifies Predictors of Biochemical Recurrence in Prostate Cancer
Carlos S Moreno, Qi Long, Soma Sannigrahi, Jianpeng Xu, Brent A Johnson, Wei Zhou, Arun K Seth, Robert K Nam, John A Petros, Theresa W Gillespie, Adeboye O Osunkoya. Emory University School of Medicine, Atlanta, GA; Sunnybrook Health Sciences Center, University of Toronto, Toronto, ON, Canada
Background: One of the challenges in prostate cancer research is the development of effective predictors of tumor recurrence following radical prostatectomy, to determine whether immediate adjuvant therapy is warranted. This is even more important in patients who have negative margins and may be lost to follow up. To date, very few biomarkers have been identified that accurately predict the possibility of tumor recurrence or biochemical failure in this setting.
Design: RNAseq libraries were prepared from RNA samples derived from a set of 51 Formalin Fixed Paraffin Embedded (FFPE) prostatectomy samples from two academic institutions. Complete Time to Recurrence, PSA, Stage, Age, and Gleason Score data was available for 47 of the 51 samples. High-throughput sequencing was performed. RNAseq data was mapped using TopHat open source software and fragments per kilobase per million reads (FPKM) values were computed using Cufflinks in the Galaxy analysis environment. We mapped sequencing reads to 24,179 mRNA and microRNA genes and performed a pre-selection step to identify the top 100 predictive genes using univariate Proportion Hazard Models. We then applied Lasso PH models to identify a final set of 9 selected genes and their estimated coefficients. Subjects were divided into high and low risk groups based on the median predictive score, and we performed log rank test to compare biochemical recurrence (BCR) between the two risk groups.
Results: Of the 47 samples analyzed, 16 (34%) had no BCR and 31 (66%) had BCR. Median follow-up for patients without BCR was 72 months (range: 35-171 months) and median time to BCR was 15 months (range: 0-49 months). Gleason scores included 3 (6%) cases of 3+3=6, 24 (51%) cases of 3+4=7, 11 (24%) cases of 4+3=7, 5 (11%) cases of 4+4=8, 2 (4%) cases of 4+5=9, and 2 (4%) cases of 5+4=9. Clinical variables alone were highly predictive of recurrence (log-rank test p=4.06e-08). The final set of 9 genes from the RNAseq analysis outperformed the clinical variables (log-rank test p=6.27e-12) with or without inclusion of the clinical variables in the model.
Conclusions: We have identified a set of 9 biomarkers in FFPE tissue that may be useful in the prediction of BCR in patients following radical prostatectomy, irrespective of pathologic stage, Gleason score, patient race, or pre-operative PSA levels. Patients with tumors that demonstrate increased expression of these markers may benefit from close follow-up after radical prostatectomy.
Category: Genitourinary (including renal tumors)
Monday, March 4, 2013 1:00 PM
Poster Session II # 141, Monday Afternoon