Pooled Clustering of High-Grade Serous Ovarian Cancer Gene Expression Leads to Novel Consensus Subtypes Associated with Survival and Surgical Outcomes

Abstract

Purpose: Here we assess whether molecular subtyping identifies biological features of tumors that correlate with survival and surgical outcomes of high-grade serous ovarian cancer (HGSOC).Experimental Design: Consensus clustering of pooled mRNA expression data from over 2,000 HGSOC cases was used to define molecular subtypes of HGSOCs. This de novo classification scheme was then applied to 381 Mayo Clinic HGSOC patients with detailed survival and surgical outcome information.Results: Five molecular subtypes of HGSOC were identified. In the pooled dataset, three subtypes were largely concordant with prior studies describing proliferative, mesenchymal, and immunoreactive tumors (concordance > 70%), and the group of tumors previously described as differentiated type was segregated into two new types, one of which (anti-mesenchymal) had downregulation of genes that were typically upregulated in the mesenchymal subtype. Molecular subtypes were significantly associated with overall survival (P < 0.001) and with rate of optimal surgical debulking (≤1 cm, P = 1.9E-4) in the pooled dataset. Among stage III-C or IV Mayo Clinic patients, molecular subtypes were also significantly associated with overall survival (P = 0.001), as well as rate of complete surgical debulking (no residual disease; 16% in mesenchymal tumors compared with >28% in other subtypes; P = 0.02).Conclusions: HGSOC tumors may be categorized into five molecular subtypes that associate with overall survival and the extent of residual disease following debulking surgery. Because mesenchymal tumors may have features that were associated with less favorable surgical outcome, molecular subtyping may have future utility in guiding neoadjuvant treatment decisions for women with HGSOC. Clin Cancer Res; 23(15); 4077-85. ©2017 AACR.

Figure 1

Five de novo molecular subtypes and signatures comprised of 100 genes in the public HGSOC expression set (n=2,103) and Mayo Clinic HGSOC expression set (n = 381).

Figure 2

Overall survival plots of five de novo molecular subtypes in the Mayo clinic cohort, for (A) all late-stage HGSOC patients, (B) late-stage patients with no macroscopic residual diseases (RD0), (C) late-stage patients with optimal-debulking but visible residual disease (RD1: ≤1cm), and (D) late stage patients with suboptimal diseases (RD2: >1cm).

Figure 3

Surgical outcomes and complexity with versus tumor molecular subtypes in Mayo Clinic cohort. (A) and (B) are barplots of number and proportion of cases with trinary scale of surgical outcomes in Mayo Clinic cohort and TCGA cohort, respectively. (C–E) Boxplots of surgical complexity scores in optimally debulked cases of the mesenchymal subtype vs. other molecular subtypes according to the de novo molecular subtypes defined in this work (C), TCGA molecular subtypes (D) and Tothill molecular subtypes (E), respectively.

Figure 4

Schematic summaries of de novo molecular subtypes with respect to previous subtype systems and associated changes.