Molecular classes in CRC: Characterization of MSI by expression profiling in the translational study of the PETACC 3-EORTC 40993- SAKK 60-00 trial.

Abstract

3597 Background: Microsatellite instability (MSI) is the hallmark of a deficient mismatch repair system (MMR) in about 15% of colorectal cancers (CRC). Other studies and our own confirm MSI as an independent prognostic marker associated with a better outcome in stage II and III CRC (ASCO 2009, abstr. #4002). As MSI can be caused by different mechanisms, and dMMR leads to secondary oncogenic alterations, heterogeneity in the clinical and molecular features of MSI CRC is likely but not well understood. In this transcriptome expression study, we explore which genes differentiates MSI from Microsatellite stable (MSS) tumors and look for evidence of additional subclasses. Methods: RNA extracted from formalin-fixed paraffin-embedded (FFPE) tissue blocks was used for expression profiling on the ALMAC platform Colorectal Cancer DSA. Classifiers were constructed using AdaBoost and DLDA algorithms and assessed with area under the curve (AUC) by cross-validation. Survival analysis was based on Cox regression. Results: 240 chips passed data quality control (42 MSI, 17.5%). Unsupervised methods allowed only weak separation of MSI and MSS specimen, despite 494 genes with significantly different expression (≤ 1% FDR), due to high variation in both groups. Gene expression differences were in agreement with results from reanalysis of three public datasets. Classifiers discriminated MSI and MSS with AUC of 0.96 [95% CI: 0.95-0.97], using 40-80 selected genes. Prominent discriminatory genes include various pathways: Wnt (f.ex. Axin2), MAPK (DUSP4); inflammation-immunity (REGs, STAT1), differentiation (TNNC2, mucins), metallothioneins. Association of these genes with RFS is heterogeneous. Conclusions: Efficient discrimination of MSS and MSI in gene expression profiles can be obtained, with good quality FFPE material, using a multigene classifier. Inside both classes there is high residual heterogeneity. More samples will be profiled in order to further define molecular subgroups and to search for prognostic genes. The ability to obtain reliable profiles form FFPE material implies that relevant information can be obtained from archival material stored in many biobanks. Author Disclosure Employment or Leadership Position Consultant or Advisory Role Stock Ownership Honoraria Research Funding Expert Testimony Other Remuneration Pfizer Pfizer Pfizer Pfizer Pfizer