Feng, Felix Y.

337 Background: Current risk stratification schema have limited prognostic performance in predicting outcome within National Comprehensive Cancer Network (NCCN) high-risk to very high-risk prostate cancer. Methods: Two multicenter high-risk cohorts were used for training (n = 214) and validation (n = 151) of a novel RNA microarray-based integrated clinical-genomic Classifier Optimized for Outcome in High-risk Prostate cancer (COOHP) to classify patients as COOHP favorable high-risk, standard high-risk, or very high-risk. Cox analysis was used to model metastasis-free survival (MFS), prostate cancer-specific survival (PCSS), and overall survival (OS). Model performance was compared to prior sub-classification systems using time-dependent c-indices. Results: Among NCCN high/very high-risk patients in the training cohort, 11% were classified as COOHP favorable high-risk, 70% as COOHP standard high-risk, and 18% as COOHP very high-risk. Patients with COOHP favorable high-risk disease had better rates of 5-year MFS compared to those with COOHP standard high-risk disease (94% vs 76%, hazard ratio [HR] 0.10, p = 0.02), and patients with COOHP very high-risk disease had worse 5-year MFS compared to those with COOHP standard high-risk disease (34% vs 76%, HR 3.5, p < 0.0001). Similarly, patients with COOHP very high-risk disease had worse 10-year PCSS compared to those with COOHP standard high-risk disease (36% vs 82%, HR 4.4, p < 0.0001). The c-indices for 5-year MFS and 10-year PCSS in the training cohort were 0.80 and 0.74, significantly improved compared to prior clinical and clinical-genomic risk stratification systems (0.62-0.69 for 5-year MFS and 0.56-0.63 for 10-year PCSS). These results were consistent in the validation cohort, where 5-year MFS significantly varied among the three COOHP subgroups (100% vs 89% vs 79%, p = 0.020), as did 10-year OS (100% vs 71% vs 53%, p = .040). Conclusions: A clinical-genomic risk stratification system specifically designed to discriminate prognosis in high-risk prostate cancer better identified favorable high-risk and very high-risk subsets of disease compared to prior clinical and clinical-genomic stratification systems.

327 Background: Though Black men with prostate cancer are more likely to have aggressive disease features than White men, race-specific differences in initial treatment responses in localized disease remains unknown. Methods: Individual patient data were obtained for 9259 patients (including 1674 [18.1%] Black men and 7585 [81.9%] White men) enrolled on eight randomized controlled trials evaluating definitive radiotherapy (RT) ± short-term or long-term androgen deprivation therapy (STADT and LTADT). The primary endpoints were biochemical recurrence (BCR), distant metastasis (DM), and prostate cancer-specific mortality (PCSM). Fine-Gray subdistribution HR (sHR) models were developed to evaluate the cumulative incidences of all endpoints after stratification by National Comprehensive Cancer Network risk grouping. A meta-analysis was done to estimate pair-wise comparisons of treatments within and between Black and White men, after adjusting for age, Gleason score, clinical T stage, and initial PSA. Results: Black men were more likely to have NCCN high-risk disease at enrollment (656/1674 [39.2%] vs 2506/7585 [33%], p<0.001). However, within the high-risk stratum Black men had lower 10-year rates of BCR (46.1% vs. 50.4%, p=0.02), DM (14% vs. 21.6%, p<0.001), and PCSM (4.9% vs. 9.8%, p<0.001). After adjusting for age and disease characteristics, Black men with high-risk prostate receiving RT+STADT had lower rates of BCR (sHR 0.73, 95% CI 0.62-0.86, p<0.001), DM (sHR 0.64, 95% CI 0.49-0.84, p=0.001) and PCSM (sHR 0.49, 95% CI 0.25-0.95, p=0.04). There were no differences in BCR, DM, or PCSM among men receiving RT+LTADT. The interaction between race and the impact of adding STADT to RT alone on BCR was statistically significant (p=0.003). Conclusions: Black men enrolled on randomized trials with long-term follow-up have higher risk disease at enrollment, but have better BCR, DM, and PCSM outcomes with RT-based therapy compared with White men, particularly with the addition of STADT.

295 Background: While 18F-fluciclovine PET/CT is approved in the US and recommended by the NCCN, prostate-specific membrane antigen (PSMA) PET/CT is more common in Europe/Australia and recommended by the EAU. Less is known about the biology of lesions detected by either modality. 18F-fluciclovine PET relies on radiotracer uptake by amino acid transporters LAT1-4 and ASCT1-2. PSMA PET is dependent on surface expression of PSMA. We compared relative expression of PSMA and fluciclovine transporter genes in radical prostatectomy (RP) samples to determine their distribution across subtypes and correlation with outcomes. Methods: Gene expression data of 19,102 RP samples were analyzed using the Affymetrix Human Exon 1.0 ST microarray. 1,135 patients had long term follow up. Associations between expression of PSMA and fluciclovine transporter genes (LAT1-4 and ASCT1-2) and pathologic variables, molecular subtypes, and clinical outcomes were conducted. Results: All fluciclovine transporter genes (LAT 1-4, ASCT1-2) were expressed at lower levels than PSMA (p <0.0001). PSMA expression was positively correlated with genomic risk score and pathologic Gleason score (p<0.0001), but LAT2-3 and ASCT2 were inversely correlated with genomic risk in primary tumors (p<0.0001) and less expressed in GS 9-10 tumors (p<0.0001). PSMA expression was associated with worse metastasis-free survival (MFS) (HR 1.45, p=0.001) and lymph node involvement (HR 2.14, p<0.0001). Expression of LAT2, LAT3, ASCT2 expression was associated with better MFS (HR 0.85, 0.63, 0.74, p<0.0001-0.04). After multivariable adjustment, PSMA expression remained independently prognostic of poorer MFS (HR 1.3, p=0.028). Luminal B subtype was notable for PSMA overexpression; Luminal A was enriched in ASCT2 and LAT2 (p<0.0001). PSMA expression did not correlate with ERG fusion prostate cancers, but LAT2, ASCT1, and ASCT2 were overexpressed in ERG fusion negative tumors (p<0.0001). Conclusions: PSMA expression is associated with more aggressive disease and poorer clinical outcomes than fluciclovine transporter genes in localized prostate cancer. Molecular subtypes of prostate cancer vary in PSMA and fluciclovine transporter gene expression.

276 Background: Decipher is a genomic classifier (GC) that estimates risk of prostate cancer (PCa) distant metastases (DM) post-radical prostatectomy (RP). Herein, we validate the GC within the context of a randomized phase 3 trial. Methods: RP specimens from patients on the NRG/RTOG 9601 phase 3 placebo-controlled randomized trial of salvage radiotherapy (sRT) +/- 2 years of bicalutamide (NCT00002874) were centrally reviewed and the highest-grade tumor underwent RNA extraction. Samples passing quality control (QC) were run on a clinical-grade whole-transcriptome assay to assign the GC score (scale 0-1). Our NCTN-CCSC approved pre-specified statistical plan (NRG-GU-TS002 CSC0075) included the primary objective of validating the ability of the GC to independently prognosticate the cumulative incidence of DM, with secondary endpoints of prostate cancer-specific mortality (PCSM) and overall survival (OS). Results: Of patients with tissue available, 352 passed QC and were included for analysis. The final GC cohort was a representative sample of the overall cohort, with a median follow-up of 13 years. On multivariable analysis, the GC (continuous variable) was independently associated with DM (HR 1.19 [95%CI 1.06-1.35], p=0.003), PCSM (HR 1.37 [95%CI 1.18-1.61], p<0.001), and OS (HR 1.16 [95%CI 1.06-1.28], p=0.002) after adjusting for age, race, Gleason score, T-stage, margin status, entry PSA, and treatment arm. The estimated absolute impact of bicalutamide on 12-year OS was less in patients with lower vs higher GC scores (2.4% vs 8.9%), further demonstrated in men receiving early sRT at PSA <0.7 ng/mL (-2.0% vs 5.0%). Conclusions: This is the first validation of this GC in a prospective randomized trial cohort and demonstrates association of the GC with DM and PCSM independent of standard clinicopathologic variables. The GC may help personalize shared decision-making to weigh the absolute benefit from the addition of bicalutamide to sRT.

BackgroundThe progression of prostate cancer is a complex, multistep process that involves molecular alterations in cells of the tumor and the microenvironment, with associated interactions between the stroma and epithelium. Genomic expression analyses of stromal infiltration markers were performed to determine the significance thereof in prostate cancer.MethodsGenome‐wide expression profiles of formalin‐fixed, paraffin‐embedded radical prostatectomy samples were evaluated from a prospective registry cohort (n = 5239) and 3 retrospective institutional cohorts (n = 1135). Two independent stromal gene expression signatures implied stromal infiltration. Cox proportional hazards regression defined the association between stromal infiltration expression and metastasis‐free survival (MFS).ResultsStromal expression scores were correlated with stromal signature genes and with other key stromal markers (CAV1, VIM, and TAGLN), basal activity, and CD3 and CD4 immune biomarkers (r > 0.5 for all). The top decile of stromal expression was associated with high genomic risk scores (Decipher ≥ 0.6) , high Cancer of the Prostate Risk Assessment–Postsurgical scores, Gleason 9 to 10 disease, and a higher risk for metastasis (hazard ratio, 2.35; 95% CI, 1.37‐4.02; P = .001). A higher stromal infiltration score was also associated with decreased expression of DNA repair genes and higher radiation sensitivity genomic scores. Postoperative radiation therapy (RT) was associated with an MFS benefit for patients with high stromal scores, but not for patients with low stromal scores (Pinteraction = .02).ConclusionsExpression of stromal infiltration markers is correlated with prostate cancer aggressiveness/progression and may be predictive of a response to RT. Stromal infiltration markers should be studied and considered for incorporation into clinical prognostication and decision making.

Abstract Purpose: The heterogeneity of androgen receptor (AR)-activity (AR-A) is well-characterized in heavily treated metastatic castration-resistant prostate cancer (mCRPC). However, the diversity and clinical implications of AR-A in treatment-naïve primary prostate cancer is largely unknown. We sought to characterize AR-A in localized prostate cancer and understand its molecular and clinical implications. Experimental Design: Genome-wide expression profiles from prostatectomy or biopsy samples from 19,470 patients were used, all with independent pathology review. This was comprised of prospective discovery (n = 5,239) and validation (n = 12,728) cohorts, six retrospective institutional cohorts with long-term clinical outcomes data (n = 1,170), and The Cancer Genome Atlas (n = 333). Results: A low AR-active subclass was identified, which comprised 9%–11% of each cohort, and was characterized by increased immune signaling, neuroendocrine expression, and decreased DNA repair. These tumors were predominantly ERG and basal subtype. Low AR-active tumors had significantly more rapid development of recurrence or metastatic disease across cohorts, which was maintained on multivariable analysis [HR, 2.61; 95% confidence interval (CI), 1.22–5.60; P = 0.014]. Low AR-active tumors were predicted to be more sensitive to PARP inhibition, platinum chemotherapy, and radiotherapy, and less sensitive to docetaxel and androgen-deprivation therapy. This was validated clinically, in that low AR-active tumors were less sensitive to androgen-deprivation therapy (OR, 0.41; 95% CI, 0.21–0.80; P = 0.008). Conclusions: Leveraging large-scale transcriptomic data allowed the identification of an aggressive subtype of treatment-naïve primary prostate cancer that harbors molecular features more analogous to mCRPC. This suggests that a preexisting subgroup of patients may have tumors that are predisposed to fail multiple current standard-of-care therapies and warrant dedicated therapeutic investigation.

Abstract Purpose: Protein kinases are known to play a prominent role in oncogenic progression across multiple cancer subtypes, yet their role in prostate cancer progression remains underexplored. The purpose of this study was to identify kinases that drive prostate cancer progression. Experimental Design: To discover kinases that drive prostate cancer progression, we investigated the association between gene expression of all known kinases and long-term clinical outcomes in tumor samples from 545 patients with high-risk disease. We evaluated the impact of genetic and pharmacologic inhibition of the most significant kinase associated with metastatic progression in vitro and in vivo. Results: DNA-dependent protein kinase (DNAPK) was identified as the most significant kinase associated with metastatic progression in high-risk prostate cancer. Inhibition of DNAPK suppressed the growth of both AR-dependent and AR-independent prostate cancer cells. Gene set enrichment analysis nominated Wnt as the top pathway associated with DNAPK. We found that DNAPK interacts with the Wnt transcription factor LEF1 and is critical for LEF1-mediated transcription. Conclusions: Our data show that DNAPK drives prostate cancer progression through transcriptional regulation of Wnt signaling and is an attractive therapeutic target in aggressive prostate cancer.

Abstract Background Predictive molecular biomarkers to select optimal treatment for patients with glioblastoma and other cancers are lacking. New strategies are needed when large randomized trials with correlative molecular data are not feasible. Methods Gene signatures (GS) were developed from 31 orthotopic glioblastoma patient-derived xenografts (PDXs), treated with standard therapies, to predict benefit from radiotherapy (RT-GS), temozolomide (Chemo-GS), or the combination (ChemoRT-GS). Independent validation was performed in a heterogeneously treated clinical cohort of 502 glioblastoma patients with overall survival as the primary endpoint. Multivariate Cox analysis was used to adjust for confounding variables and evaluate interactions between signatures and treatment. Results PDX models recapitulated the clinical heterogeneity of glioblastoma patients. RT-GS, Chemo-GS, and ChemoRT-GS were correlated with benefit from treatment in the PDX models. In independent clinical validation, higher RT-GS scores were associated with increased survival only in patients receiving RT (P = 0.0031, hazard ratio [HR] = 0.78 [0.66–0.92]), higher Chemo-GS scores were associated with increased survival only in patients receiving chemotherapy (P < 0.0001, HR = 0.66 [0.55–0.8]), and higher ChemoRT-GS scores were associated with increased survival only in patients receiving ChemoRT (P = 0.0001, HR = 0.54 [0.4–0.74]). RT-GS and ChemoRT-GS had significant interactions with treatment on multivariate analysis (P = 0.0009 and 0.02, respectively), indicating that they are bona fide predictive biomarkers. Conclusions Using a novel PDX-driven methodology, we developed and validated 3 platform-independent molecular signatures that predict benefit from standard of care therapies for glioblastoma. These signatures may be useful to personalize glioblastoma treatment in the clinic and this approach may be a generalizable method to identify predictive biomarkers without resource-intensive randomized trials.

5000 Background: It is clinically challenging to integrate genomic classifier results that report a continuous numerical risk of recurrence into treatment decisions for prostate cancer (PCa). We aimed to develop a novel clinical-genomic risk system that can readily be incorporated into treatment guidelines for localized PCa. Methods: Four multi-center cohorts (n = 6928 men; 5937 prospective samples and 991 retrospective samples with long-term follow-up) were utilized to identify and validate our clinical-genomic risk system in radical prostatectomy (RP) samples and subsequently in pre-treatment biopsy samples. All patients’ FFPE tissue underwent microarray analysis, and the expression values for 22 prespecified biomarkers that constitute Decipher were extracted. Cumulative incidence curves were constructed to estimate metastasis risk. C-indices were calculated to compare NCCN and CAPRA score to our clinical-genomic system. Results: With a median follow-up of 8 years for men in our RP cohort, the 10-year distant metastasis rates for NCCN low, favorable-intermediate, unfavorable-intermediate, and high-risk were 7.8%, 9.4%, 40.1%, and 41.4%, respectively. Our 3-tier clinical-genomic risk groups had 10-year distant metastasis rates of 3.7%, 30.7%, and 57.7%, for low, intermediate, and high-risk, which were validated in our pre-treatment biopsy cohort with 10-year rate of distant metastasis of 0%, 30.3%, and 63.2%, respectively. C-indices for the clinical-genomic system (0.84, 95%CI 0.62-0.92) were significantly improved over NCCN (0.71, 95%CI 0.59-0.84) and CAPRA (0.71, 95%CI 0.60-0.81) score. A total of 33.4% of men would be reclassified by the clinical-genomic system, and specifically 17.1%, 41.3%, and 19.4% of men in NCCN low, intermediate and high risk groups would be reclassified by our new system. Conclusions: The use of a readily available genomic classifier in combination with clinicopathologic variables can generate a simple to use 3-tier clinical-genomic risk system that is highly prognostic for distant metastasis, is more accurate than clinical risk, and can be easily incorporated into NCCN guidelines to inform treatment decisions.