Identification and validation of gene expression subtypes in a large set of colorectal cancer samples

• Published in: Journal of clinical oncology Vol. 30; no. 15_suppl; p. 3511
• Main Authors: Budinska, Eva, Popovici, Vlad Calin, Sikora, Katarzyna Otylia, Lapique, Nicolas, Tejpar, Sabine, Hodgson, John Graeme, Weinrich, Scott, Roth, Arnaud, Bosman, Fred, Delorenzi, Mauro
• Format: Journal Article
• Language: English
• Published: 20.05.2012
• ISSN: 0732-183X; 1527-7755
• DOI: 10.1200/jco.2012.30.15_suppl.3511

Abstract only 3511 Background: From a clinical perspective colorectal cancer (CRC) is a heterogeneous disease whose biological background is insufficiently understood. In order to adapt targeted treatment to biologically different categories of CRC patients, in depth understanding of the molecular mechanisms involved in CRC heterogeneity is urgently needed. Methods: Consensus cluster analysis of 1113 stage II-III CRC gene expression profiles from PETACC3 and 4 public datasets defined a set of subtypes which were confirmed in an independent set of 720 samples. The similarity between tumors was based on a set of 54 meta-genes. This approach improves the robustness to measurement noise and gives equal chances to each biological process to be accounted for in the subtype definition. We tested the association of the subtypes with clinical variables, molecular markers and patient survival. Results: We identified and validated 5 major subtypes A-E, with different levels of expression in 6 main molecular processes: epithelial-mesenchymal transition (EMT), immune response, colon crypt differentiation, proliferation, Wnt signaling and chromosome 20q. Significant differences in survival and an enrichment for markers such as MSI, BRAF mutation, site or p53 status were found between the subtypes. In addition, the new subtype classification uncovered heterogeneity within groups defined by these commonly used markers. Survival analysis showed significant prognostic value for meta-genes connected to EMT, proliferation and differentiation. Identical data processing and clustering applied to the validation set on the same meta-genes resulted in subtypes with almost identical expression patterns. Conclusions: We identified and validated robust molecular subtypes in the largest set of stage II-III CRC samples as a combination of multiple molecular processes, that complement current disease stratification based on clinico-pathological variables and molecular markers. The biological relevance of these subtypes was reflected in significant differences in survival. These insights open new perspectives for improving prognostic models and rationalize the prediction of tumor-specific drug sensitivity.