Downer, Mary K.

Abstract Introduction: Epidemiologic data support an inverse association between statin use and risk of lethal prostate cancer, but contributing mechanisms have not been elucidated. Herein, using data from the Health Professionals Follow-up Study (HPFS), we updated our previous analysis of statins and lethal prostate cancer with a decade of additional follow-up, and incorporated immunohistochemistry (IHC)-based biomarkers and gene expression profiling to identify putative molecular mechanisms contributing to the association. Methods: We included data from 44,076 HPFS participants who were cancer free in 1990, with follow-up through 2012. Participants reported prediagnostic statin use on biennial questionnaires. Statin use was categorized as current vs. never/past use, with current users further categorized as long-term vs. short-term users (≥6 vs. <6 years). To classify prostate cancer molecular subtypes, genetically validated IHC assays for TMPRSS2:ERG fusion and PTEN loss were conducted on tissue microarrays for 891 and 660 cases, respectively. Whole-genome mRNA profiling of tumor and adjacent normal tissue was completed for 237 and 114 cases, respectively, using Affymetrix GeneChip Human Gene 1.0 ST arrays. We used Cox proportional hazards analysis to generate hazard ratios (HR) and 95% confidence intervals (CI) for associations between statin use and risk of overall, advanced (stage T3b or higher at diagnosis, metastatic or fatal), and lethal (metastatic or fatal) prostate cancer, as well as molecular subtypes, adjusted for PSA testing behavior and potential confounders including obesity and smoking. Moreover, we explored gene expression pathways in tumor and adjacent normal prostate tissue of statin users vs. nonusers using gene set enrichment analysis (GSEA). Results: During 22 years of follow-up, 6,144 men were diagnosed with prostate cancer; 1,051 (17%) had advanced disease and 827 (13%) developed lethal prostate cancer. Of cases with molecular subtype available, 48% were ERG-positive and 14% PTEN-null. In multivariable analysis, relative to never/past use, current statin use was inversely associated with risk of advanced (HR 0.84; 95% CI 0.68-1.04) and lethal (HR 0.78; 95% CI 0.68-1.01) prostate cancer. These associations were more pronounced in long-term users, where ≥6 years of prediagnosis statin use was significantly associated with a lower risk of advanced (HR 0.72; 95% CI 0.53-0.97) and lethal prostate cancer (HR 0.64; 95% CI 0.45-0.92), relative to nonuse. Long-term statin use was not associated with overall prostate cancer risk, whether ERG-positive (HR 0.89; 95% CI 0.59-1.35) or ERG-negative (HR 1.06; 95% CI 0.74-1.50). However, relative to nonuse, a longer duration of prediagnosis statin use was associated with reduced risk of PTEN-null cancers (HR 0.42; 95% CI 0.20-0.90), but not PTEN-intact disease (HR 1.18; 95% CI 0.95-1.46; p-heterogeneity=0.01). In GSEA, we identified T cell, B cell, and phosphatidylinositol (PI3K) signaling among the top pathways enriched in tumor-adjacent normal prostate tissue of current statin users, relative to never users. Intriguingly, no gene sets were differentially expressed by statin use in the tumor. Discussion: Our updated findings from the HPFS are consistent with our prior findings and other epidemiologic data supporting a role for statins in lethal prostate cancer prevention. Molecular classification of tumors identified PTEN/PI3K signaling and inflammation/immune activation as two potential mechanisms contributing to this association. Elucidating molecular mechanisms mediating the association between statin use and lower risk of lethal prostate cancer will provide support for a causal effect of statins on lethal prostate cancer risk and could inform statin clinical trials with appropriate intermediate endpoints. Citation Format: Emma H. Allott, Ericka M. Ebot, Amparo G. Gonzalez-Feliciano, Sarah C. Markt, Kathryn M. Wilson, Thomas U. Ahearn, Travis A. Gerke, Mary K. Downer, Konrad H. Stopsack, Jennifer R. Rider, Stephen J. Freedland, Elizabeth A. Platz, Meir J. Stampfer, Edward L. Giovannucci, Christopher J. Sweeney, Stephen P. Finn, Lorelei A. Mucci. Molecular tumor profiling to identify mechanisms linking statin use with lower risk of lethal prostate cancer: Results from the Health Professionals Follow-up Study [abstract]. In: Proceedings of the AACR Special Conference: Prostate Cancer: Advances in Basic, Translational, and Clinical Research; 2017 Dec 2-5; Orlando, Florida. Philadelphia (PA): AACR; Cancer Res 2018;78(16 Suppl):Abstract nr A015.

106 Background: Pharmacoepidemiologic studies suggest prognostic benefits of aspirin in prostate cancer. We hypothesized that prostate cancer tissue from regular aspirin users at cancer diagnosis compared to nonusers is characterized by specific transcriptional changes. Methods: We analyzed tumor whole-transcriptome data from men diagnosed with prostate cancer during prospective follow up of the Physicians Health Study 1 (PHS1, n1 = 150), initially a randomized-controlled trial of aspirin for prevention of cardiovascular disease and cancer. We compared the expression of putative aspirin target genes ( PTGS2, CTNNB1, MYC, AXIN2, PTGER3) between regular aspirin users ( ≥ 3 x 324 mg/week) and non-users at cancer diagnosis. We used Gene Set Enrichment Analysis to identify Gene Ontology biological process gene sets (total, 3609) associated with aspirin use. Leading edge genes of ≥ 3 functionally related gene sets were summarized in a meta-gene score. Results were validated among prostate cancer patients from the Health-Professionals Follow-up Study (HPFS, n2 = 254; aspirin use, ≥ 2 x/week). Results: Regular aspirin use was common (65% in PHS1; 45% in HPFS) and not associated with age, stage, and Gleason grade. In PHS1, none of the predefined genes were associated with aspirin use (all p > 0.05). 26 gene sets were downregulated in aspirin users at a false discovery rate (FDR) < 0.25 (9 at FDR < 0.05). 11 gene sets were clustered functionally around ribosome function and protein translation. A “ribosome” score of 96 genes was higher in aspirin users compared to nonusers (difference, 35.5 standard deviations [SD]; p = 0.0001). There was no dose-dependency for cumulative duration of aspirin use before cancer diagnosis ( ptrend = 0.17) or time since stopping use among non-users at diagnosis ( p = 0.32). In HPFS, neither the pre-defined target genes (all p > 0.18) nor the score were associated with aspirin use (difference, –9.8 SD; p = 0.18). Conclusions: Although RNA acetylation and resulting intratumoral gene expression changes due to aspirin use may be biologically plausible, we were unable to corroborate this association in two-long term prospective studies. Short-term trials might assess if aspirin has direct effects on tumor gene expression.