Abstract 5063: Preclinical evaluation of an EMT multi-gene biomarker index for predicting response to inhibitors of EGFR and IGF1R

Abstract

Abstract In oncology, the goal of personalized medicine is to improve patient outcome by tailoring therapy to the biochemical signaling within the individual's tumor. This requires identifying effective predictors of response that can be measured in biopsy material. High expression of the epithelial marker E-cadherin is associated with increased sensitivity in cultured cells and an improved survival benefit in response to erlotinib in patients. Downregulation of E-cadherin is a critical event in epithelial to mesenchymal transition (EMT) and is associated with decreased sensitivity to erlotinib. In order to determine if a more complete characterization of the EMT state would better predict sensitivity to erlotinib as well as other epithelial-targeting drugs, we developed an 88-gene signature that can be used to calculate a numerical index value which represents the EMT state of cells or tumors. In a panel of human tumor cell lines, the EMT index predicted erlotinib sensitivity correctly in 21 out of 24 lines. The index also predicted sensitivity to the IGF1R inhibitor OSI-906. While the EMT index marginally improved sensitivity prediction in vitro compared to E-cadherin, the numerical index allowed for a comparison of relative EMT states that was not possible with E-cadherin alone. Using a microarray database of human tumor sections, we computed EMT index values across 8 solid tumor types in order to compare their relative EMT states. Breast and lung tumors had a fairly even distribution between epithelial and mesenchymal tumors, while colon tumors were more epithelial, and kidney tumors were more mesenchymal. When comparing EMT index values to E-cadherin mRNA levels, some tumor types showed good agreement between E-cadherin and the EMT index (colon, kidney, lung) while others showed less agreement (breast, pancreas). We next compared our EMT signature to other published signatures of EMT and EGFR inhibitor response. Interestingly, most signatures were equally predictive of erlotinib sensitivity in vitro, even though less than 20% of the genes overlaped between any two signatures. We also analyzed signatures developed for individual characteristics of EMT such as invasion or stem cells. These did not correlate with erlotinib sensitivity, suggesting the protection from current therapeutics is not controlled solely by these specific aspects of EMT. Finally, we characterized sub-types of lung cancer for EMT state to determine if the EMT classification might correlate with established response rates to erlotinib. Adenocarcinomas, which tend to respond well, had more epithelial index values while squamous cell carcinomas, which tend to be less sensitive, had more mesenchymal values. This suggests that more epithelial index values might predict for erlotinib response in patients, potentially providing clinicians with a tool to more effectively treat patients based on the EMT status of their tumors. Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 102nd Annual Meeting of the American Association for Cancer Research; 2011 Apr 2-6; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2011;71(8 Suppl):Abstract nr 5063. doi:10.1158/1538-7445.AM2011-5063