Impact of AR-V7 and other androgen receptor splice variant expression on outcomes of post-prostatectomy salvage therapy.

Abstract

274 Background: Radiotherapy (RT) with or without androgen deprivation therapy (ADT) plays a key role in salvage therapy of prostate cancer recurrent after prostatectomy. However, not all patients benefit from salvage therapy, and there is an unmet need for biomarkers to distinguish responders from non-responders. Prostate cancer depends on androgen receptor (AR) signaling, and expression of AR splice variants (ARVs) that enable androgen-independent AR signaling is associated with resistance to ADT in the metastatic setting. Recent in vitro data suggest that ARVs also mediate DNA repair after irradiation, suggesting that ARV expression may also be a biomarker of RT resistance. However, the landscape of ARVs in primary prostate cancer and its effect on treatment outcomes remain unexplored. Here we hypothesize that ARVs are detectable in primary prostate cancer and may modulate response to salvage RT + ADT. Methods: We retrospectively identified 46 prostate cancer patients treated with prostatectomy followed by salvage RT+ ADT at a single institution from 1995 to 2012. The indication for salvage therapy was biochemical failure after undetectable post-operative PSA in 72%, gross local recurrence in 17%, and persistently elevated PSA after surgery in 11%. Median RT dose was 64.8 Gy, and all patients received concurrent ADT. We comprehensively interrogated the landscape of ARVs by performing ultra-deep targeted RNA-seq of archival formalin-fixed paraffin-embedded prostatectomy samples. Using a custom library of > 3000 primers spanning all AR exons and introns, we evaluated 21 native splice junction sites and 20 splice variants with a mean depth of coverage of > 5000x. Decipher score was also evaluated. We tested for association between splice variant expression and clinical outcomes using the log-rank test and Cox proportional hazards model. Results: In total, 76% of patients had one or more detectable AR splice variants. The most commonly detected variants were AR-45 in 41%, AR-V9 in 20%, and AR-V7 in 13%. At a median follow-up of 33.8 months, biochemical progression-free survival (BPFS) at 3 years was 60%, distant metastasis-free survival was 90%, and overall survival was 100%. Among detected splice variants, only AR-V7 was associated with differential clinical outcomes, with a median BPFS of 10.9 months in AR-V7 positive vs 73.4 months in AR-V7 negative patients ( p = 0.0020, HR 5.23, 95% CI 1.62-16.87). Conclusions: Using ultra-deep targeted RNA-Seq, we provide among the first comprehensive descriptions of the ARV landscape in primary prostate cancer. Moreover, we show that detectable AR-V7 in prostatectomy specimens was associated with inferior outcomes following salvage RT+ADT, suggesting for the first time that AR-V7 may modulate outcomes for localized as well as metastatic disease. Ongoing work includes comparison with Decipher score and validation in independent cohorts.