Molecular characterization of Gleason patterns 3 and 4 prostate cancer using reverse Warburg effect-associated genes

Abstract

Background: Gleason scores (GS) 3 + 3 and 3 + 4 prostate cancers (PCa) differ greatly in their clinical courses, with Gleason pattern (GP) 4 representing a major independent risk factor for cancer progression. However, Gleason grade is not reliably ascertained by diagnostic biopsy, largely due to sampling inadequacies, subjectivity in the Gleason grading procedure, and a lack of more objective biomarker assays to stratify prostate cancer aggressiveness. In most aggressive cancer types, the tumor microenvironment exhibits a reciprocal pro-tumorigenic metabolic phenotype consistent with the reverse Warburg effect (RWE). The RWE can be viewed as a physiologic response to the epithelial phenotype that is independent of both the epithelial genotype and of direct tumor sampling. We hypothesize that differential expression of RWE-associated genes can be used to classify Gleason pattern, distinguishing GP3 from GP4 PCa foci. Methods: Gene expression profiling was conducted on RNA extracted from laser-capture microdissected stromal tissue surrounding 20 GP3 and 21 GP4 cancer foci from PCa patients with GS 3 + 3 and GS ≥ 4 + 3, respectively. Genes were probed using a 102-gene NanoString probe set targeted towards biological processes associated with the RWE. Differentially expressed genes were identified from normalized data by univariate analysis. A top-scoring pair (TSP) analysis was completed on raw gene expression values. Genes were analyzed for enriched Gene Ontology (GO) biological processes and protein-protein interactions using STRING and GeneMANIA. Results: Univariate analysis identified nine genes (FOXO1 (AUC: 0.884), GPD2, SPARC, HK2, COL1A2, ALDOA, MCT4, NRF2, and ATG5) that were differentially expressed between GP3 and GP4 stroma (p < 0.05). However, following correction for false discovery, only FOXO1 retained statistical significance at q < 0.05. The TSP analysis identified a significant gene pair, namely ATG5/GLUT1. Greater expression of ATG5 relative to GLUT1 correctly classified 77.4 % of GP3/GP4 samples. Enrichment for GO-biological processes revealed that catabolic glucose processes and oxidative stress response pathways were strongly associated with GP3 foci but not GP4. FOXO1 was identified as being a primary nodal protein. Conclusions: We report that RWE-associated genes can be used to distinguish between GP3 and GP4 prostate cancers. Moreover, we find that the RWE response is downregulated in the stroma surrounding GP4, possibly via modulation of FOXO1.