Abstract B41: Methylation profiling of ovarian cancer to study etiologic and prognostic heterogeneity and to develop a molecular classifier.

Abstract

Abstract Background: Ovarian cancer is a heterogeneous disease that is divisible into multiple subtypes with variable pathogenesis, etiology and biological behavior. We analyzed DNA methylation profiling data to identify biologic subgroups of ovarian cancer and study their relationship with histologic subtypes and prognosis. Additionally, we developed a molecular classifier in relation to standard histologic subtype for classification of ovarian tumors in epidemiologic studies. Methods: A total of 180 paraffin embedded ovarian epithelial tumor tissues, including the four major epithelial ovarian tumor subtypes (serous, endometrioid, mucinous and clear cell) and tumors of low malignant potential (LMP) were selected from two different sources: The Polish Ovarian Cancer study, an incident population-based case-control study conducted from 2001-2004, and the Surveillance, Epidemiology, and End Results Residual Tissue Repository (SEER RTR), which included ovarian tumors blocks collected between 1994 and 2004 from the Iowa and Hawaii SEER registries. The distributions of tumor histologic subtypes and grades from the studies were similar. All analyses were restricted to Caucasian women. Methylation profiling was conducted using the Illumina 450K methylation array. Analyses were restricted to the 22 autosomal chromosomes and non-SNP probes. Fourteen samples did not pass quality control and were excluded from the analysis. Of the 166 evaluable samples, 29 cases (17.5%) had their histologic subtype recoded after expert review. In order to compute and validate our histological signatures, the samples were divided into a training set (N=110) and a validation set (N=56). In addition, 10 high grade serous cases with 450K methylation data from the ovarian TCGA effort were included for replication. Signatures were computed using a LASSO logistic regression. Results: Among 166 samples, 32 (19%) were from the Polish study and 134 (81%) were from the SEER RTR. Unsupervised hierarchical clustering of the 5,000 most variable CpG sites showed 4 major clusters: Cluster 1 with 79% invasive serous and 14% endometrioid carcinomas, including the majority of high grade carcinomas; cluster 2 with 77% either endometrioid or clear cell carcinomas; cluster 3 with 71% serous LMP; and cluster 4 with 73% mucinous carcinomas. We observed significant survival differences across these clusters (long-rank test P= 4.64×10-6), similar to differences observed for histologic subtypes. We used the training set to determine a parsimonious classifier based on methylation markers for histologic subtypes. We applied these signatures to an independent validation set from the Polish Study, SEER RTR, and the ovarian TCGA and found that 77% of the samples were correctly classified. Among the cases for which the histology was recoded after expert review, the methylation signatures correctly classified the histology subtype in 76% of the cases. Conclusions: Unsupervised analysis of DNA methylation profiling identified 4 distinct clusters of ovarian carcinomas, consistent with data indicating that ovarian cancer is heterogeneous with respect to cells of origin, carcinogenic pathways and histology. High grade serous carcinomas were grouped with high grade endometrioid cancers, while the remaining endometrioid carcinomas clustered with clear cell carcinomas, consistent with a common origin for a subset of these tumors from orthotopic or ectopic endometrial tissue. Our results suggest that methylation signatures provide a classification of ovarian tumors that overlaps with histologic subtypes and probable mechanisms of origin. Ongoing analyses will compare performance of the methylation classifier with histologic subtype in relation to risk factors and prognosis. Citation Format: Clara Bodelon, Keith Killian, Joshua Sampson, Holly Stevenson, William Anderson, Rayna Matsuno, Louise Brinton, Jolanta Lissowska, Mark Sherman, Nicolas Wentzensen. Methylation profiling of ovarian cancer to study etiologic and prognostic heterogeneity and to develop a molecular classifier. [abstract]. In: Proceedings of the AACR Special Conference on Advances in Ovarian Cancer Research: Exploiting Vulnerabilities; Oct 17-20, 2015; Orlando, FL. Philadelphia (PA): AACR; Clin Cancer Res 2016;22(2 Suppl):Abstract nr B41.