3529 Background: Epithelial-mesenchymal transition (EMT) has been identified as a dominant molecular subtype of colorectal cancer (CRC). This EMT phenotype as recognized by complex gene signatures is prognostic and associated with chemoresistance, but a biomarker for EMT suitable for clinical utilization has not yet been validated. The purpose of this study was to compare MET protein expression with protein/gene expression of EMT markers and to evaluate its impact on overall survival (OS). Methods: We performed an exploratory analysis of 139 untreated primary CRC samples using data from The Cancer Genome Atlas. Protein and gene expressions were measured using reverse-phase protein array (RPPA) and RNA-sequencing, respectively. MET high/overexpressed group was defined by protein level in the highest quartile. Mann-Whitney U-test and Spearman rank correlation was used to determine association between MET protein expression and protein/gene expression of EMT markers and EMT gene signature scores. Regression tree method and Kaplan-Meier estimates were used to assess overall survival (OS). Results: The MET protein distribution is right skewed, demonstrating a unique population of MET high expressing tumors (P < 0.01). Colon tumors had higher MET protein levels compared to rectal tumors (P < 0.01). MET overexpression was associated with decreased OS (HR 2.92; 95% CI: 1.45 - 5.92). MET protein expression correlated strongly with protein expressions of SLUG (transcription factor for EMT) (r = 0.6) and ERCC1 (a marker for oxaliplatin chemo-resistance) (r = 0.6) (P < 0.01). Higher MET protein levels were associated with higher gene expression of 28 EMT markers including AXL, VIM, ZEB1, ZEB2, FGF1, TGFB1I1 and MMP11 (P < 0.05). Higher MET protein levels were also associated with higher gene scores derived from three published EMT gene signatures (P < 0.05). MET protein expression did not correlate with MET gene expression (r = 0.16). Conclusions: Increased MET protein expression strongly correlates with a molecular EMT phenotype and poor survival in patients with CRC. MET protein expression may be used as a surrogate biomarker to represent and select for this unique molecular subset of CRC driven by EMT biology.
Gene signatures associated with drug resistance to irinotecan and oxaliplatin predict a poor prognosis in patients with colorectal cancer