Abstract A045: All happy families are alike: Transcriptomic homogeneity in indolent prostate tumors is a useful prognostic biomarker

Abstract

Abstract Purpose: To derive a gene expression score of transcriptomic homogeneity in localized prostate tumor specimens, and assess its prognostic utility for metastatic or lethal disease. Background: At the time of prostate cancer diagnosis, improved biomarkers are needed to distinguish tumors with metastatic potential versus those likely to follow an indolent course. Previous evidence suggests that gene expression aberrations are homogeneous in nonaggressive tumors, while transcriptomic profiles are disregulated in widely varying patterns across metastasis-prone tumors. Here, we investigate whether a measure of transcriptomic homogeneity has prognostic utility. Study Population: Gene expression profiles of noncancerous prostate tissue from the Genotype-Tissue Expression (GTEx) project (n = 106) were used to develop the homogeneity measure. Prognostic utility of the score was assessed in a study nested in the Health Professionals Follow-up Study (HPFS) and Physicians’ Health Study (PHS), in which whole-transcriptome gene expression was quantified from archival surgical tumor tissue. Cases (n = 113) were men who died of prostate cancer or developed metastatic disease, and controls (n = 291) had at least 8 years of metastasis-free survival. Men were diagnosed with prostate cancer between 1982 and 2005, and median follow-up time was 14.0 years. TCGA data (n = 333) were further used to investigate associations of the expression score with Gleason. Methods: RNA-seq profiles from GTEx were filtered to retain genes persistently expressed across noncancerous prostate tissue specimens. Remaining genes were placed into decile bins on the basis of median expression across samples. The 25 lowest variance genes within each bin were selected to generate a 250-gene panel. Transcriptome homogeneity scores for patients in HPFS and PHS were calculated by summing the standard deviations over the decile bins, where standard deviations were calculated using patient-specific data across 25 genes in each bin. Associations between the score and lethal disease were assessed through logistic regression and AUC analyses. Associations between the homogeneity score and Gleason score in TCGA samples were evaluated via Pearson correlations. Results: High transcriptomic heterogeneity was strongly associated with lethal or metastatic outcomes. A crude odds ratio of 3.36 (95% CI 2.12-5.41; p<0.001) was estimated for patients with above median heterogeneity scores compared to those with homogenous expression (those with less than or equal to median heterogeneity). This association persisted with adjustment for Gleason (<7, 3+4, 4+3, 8, 9-10) with an odds ratio of 2.04 (95% CI 1.19-3.51; p=0.01). Addition of heterogeneity scores to Gleason improved the AUC of the Gleason-only model from 0.82 (95% CI 0.77-0.86) to 0.84 (95% CI 0.80-0.88; p=0.01 for improvement). Higher heterogeneity also tracked with higher Gleason score in TCGA with a correlation of 0.33 (95% CI 0.23-0.42; p<0.001). Conclusions: Transcriptomic homogeneity, as measured by variation across gene sets that are typically stable in normal prostate tissue, is able to distinguish indolent from lethal prostate cancers. Further investigation and discovery of gene sets using this novel approach is likely to deliver new biologic insights and clinically actionable biomarker sets. Citation Format: Travis Gerke, Svitlana Tyekucheva, Jordan H. Creed, Kathryn L. Penney, Jennifer A. Sinnott, Ericka Ebot, Anders E. Berglund, Massimo Loda, Meir J. Stampfer, Peter Kraft, Giovanni Parmigiani, Lorelei A. Mucci. All happy families are alike: Transcriptomic homogeneity in indolent prostate tumors is a useful prognostic biomarker [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 A045.