Unique gene expression profile based upon pathologic response in epithelial ovarian cancer. Homo sapiens

PURPOSE: We investigated whether tumor tissue obtained at diagnosis expresses a specific gene profile that is predictive of findings at second-look surgery in patients with epithelial ovarian cancer (EOC). PATIENTS AND METHODS: Tumor tissue obtained at the time of diagnosis was profiled with oligonucleotide microarrays. Class prediction analysis was performed in a training set of 24 patients who had undergone a second-look procedure. The resultant predictive signature was then tested on an independent validation set comprised of 36 patients. RESULTS: A 93-gene signature referred to as the Chemotherapy Response Profile (CRP) was identified through its association with pathologic complete response. When applied to a separate validation set, the CRP distinguished between patients with unfavorable versus favorable overall survival (median 41 months v not yet reached, respectively, log-rank P = .007), with a median follow-up of 52 months. The signature maintained independent prognostic value in multivariate analysis, controlling for other known prognostic factors such as age, stage, grade, and debulking status. There was no genetic overlap between the CRP and our previously described Ovarian Cancer Prognostic Profile (OCPP), which demonstrated similar prognostic value. The combination of the CRP and OCPP yielded better prognostic discrimination then either profile alone. Genes present in the CRP include BAX, a proapoptotic protein previously associated with chemotherapy response in ovarian cancer. CONCLUSION: Identification of a gene expression profile based on pathologic response in EOC provides independent prognostic information and offers potential insights into the mechanism of drug resistance. Efforts to identify a more tailored profile using selected genes from both the CRP and OCPP are underway. keyword(s): Correlation with outcome Overall design: All microarray experiments were performed at the BIDMC Genomics Center. Ovarian cancer samples were collected at the time of primary debulking surgery and frozen at 80°C. Tumor samples were pulverized in liquid nitrogen and homogenized in Trizol solution, followed by RNA isolation, probe labeling, and Affymetrix GeneChip hybridization using standard manufacturer protocols. The dChip algorithm was used for data normalization and for generating gene expression signal values. Two outlier arrays as identified by the dChip software were excluded from further analysis.