Abstract 1199: Systems-based approach identifies altered carbohydrate metabolism as a predictor of a malignant phenotype in ovarian cancer

Abstract

Abstract Objective: Metabolic changes occur initially at a molecular level, while genetic alterations further contribute to this shift and promote cancer cell survival and proliferation. We sought to identify genomic correlates of metabolic dysregulation in high grade serous ovarian cancer (HGSOC). Methods: We profiled 101 HGSOC samples and 15 normal ovarian tissues samples by LC/MS and GC/MS metabolic profiling; 172 significantly altered metabolites were identified. We classified these metabolites into altered pathways and carried out full-scale gene expression analyses. Results: We compared expression of measured metabolites for normal ovarian tissues and HGSOC and classified them into super pathways. We created a random forest classifier to generate a prediction model using metabolic profiles from normal tissue versus tumor within 3% error. From the random forest classification, the top 10% of significantly altered metabolites included gluconate, ADMA, and NAA. Carbohydrate, amino acid, and lipid super pathways were identified as the most important, with carbohydrate enrichment as significant (p = 0.03). Metabolites from the pentose phosphate pathway (PPP) and glycolysis were identified with this prediction model and found to be globally downregulated. Gene expression for enzymes in the PPP and glycolysis were compared between HGSOC and normal ovary and not found to be different. Gene expression ratios from the rate limiting steps in these pathways were evaluated. No significant difference was identified between gene ratios from normal and tumor tissues (p = 0.22) within our data set, but relative expression was significantly different within The Cancer Genome Atlas (TCGA) data set (p = 0.009). We subsequently generated a network that merged metabolic and gene level changes for enzymes coding for the synthesis and degradation of these metabolites, while accounting for time to recurrence in ovarian cancer patients for each gene within glycolysis and PPP. We then merged the genes identified by our network analysis with data from a whole-genome siRNA synthetic lethality screen (3 HGSOC chemoresistant cell lines). When GPI, the gene that encodes glucose-6-phosphate isomerase (PGI), was silenced, cellular lethality was observed across all cell lines tested. In our network analysis, GPI was among the most upregulated within the carbohydrate pathway within our cohort of ovarian cancer samples. In the TCGA data, ovarian cancer patients with tumoral GPI levels higher than the median had worse overall survival (p = 0.0002). Conclusions: Here, we present a novel systems-based approach using altered metabolites and genes to predict a malignant phenotype specific to HGSOC patients. Altered metabolism, coupled with genomic analyses, identified the most interconnected gene-biochemical networks that will lead to novel biomarkers and therapeutic targets. Citation Format: Rebecca A. Previs, Tyler J. Moss, Behrouz Zand, Rajesha Rupaimoole, Heather J. Dalton, Jean M. Hansen, Guillermo Armaiz-Pena, Susan Lutgendorf, Robert L. Coleman, Pratip Bhattacharya, Prahlad Ram, Anil K. Sood. Systems-based approach identifies altered carbohydrate metabolism as a predictor of a malignant phenotype in ovarian cancer. [abstract]. In: Proceedings of the 106th Annual Meeting of the American Association for Cancer Research; 2015 Apr 18-22; Philadelphia, PA. Philadelphia (PA): AACR; Cancer Res 2015;75(15 Suppl):Abstract nr 1199. doi:10.1158/1538-7445.AM2015-1199