Establishing a molecular taxonomy for epithelial ovarian cancer (EOC) from 363 formalin-fixed paraffin embedded (FFPE) specimens

C Gourley, C O Michie, K E Keating, S Deharo, E J O'Brien, A Winter, F A McDyer, J M Mulligan, L A Hill, T S Davison, T Halsey, L McCoy, C Wilson, A Williams, D J Harrison, D P Harkin, R D Kennedy

Research output: Contribution to journalArticlepeer-review


5000 Background: Despite histological heterogeneity and evidence of some molecular subtypes (e.g. BRCA1/2-deficient) EOC continues to be treated as a single disease. Using unsupervised hierarchical clustering of DNA microarray expression data we aimed to identify molecular subgroups with a view to personalized therapy.

METHODS: Microarray expression analysis was performed on 363 FFPE EOC samples (by far the largest dataset to date) with linked prospectively collected clinical data. Unsupervised clustering was applied to the most variable genes. This was followed by computational classification in the serous samples only.

RESULTS: Unsupervised analysis of all samples identified 6 subgroups which were most significantly related to histology (p=3.65 x 10(-27)) and had different overall survival (OS; p<0.0001). Clear cell tumors were present in one group, mucinous in another, serous and endometrioid were spread across 4 groups each. When a similar approach was applied to pure serous samples (no mixed tumors; n=199) three subgroups were identified with different OS (p<0.0001). Intriguingly, the dominant discriminatory biology within the serous cohort was related to vasculature development (FDR x 10(-7)) and more specifically to angiogenic processes (FDR p.Angiogenesis= 8.59 x 10(-5)). Up-regulation of these angiogenesis-related genes predominated in only one of the 3 serous subgroups (58 out of 199 tumours). Class labels were assigned to these samples based upon expression of these angiogenesis-related genes and a gene expression signature was developed which could identify the novel angiogenesis molecular subgroup.

CONCLUSIONS: 1) It is possible to consistently perform transcriptomic analysis from small FFPE ovarian tumor sections and cluster similar tumors. 2) We have identified 6 main clusters of EOC, strongly associated with histology and survival. 3) In the pure serous subgroup, 3 clusters were identified with significantly differing survival; one of these subgroups was defined by upregulation of multiple angiogenesis genes. In light of the recently reported activity of bevacizumab in EOC (GOG 218 and ICON7) this subgroup may explain the observed response.

Original languageEnglish
Pages (from-to)5000
JournalJournal of Clinical Oncology
Issue number15_suppl
Publication statusPublished - 20 May 2011

Cite this