Abstract C120: Testing a gene expression based signature predicting Cetuximab sensitivity in wild type KRAS tumors expressing EGFR

Abstract

Abstract Cetuximab (C), an antibody targeting EGFR, is approved for the treatment of colon and head & neck cancer. Since the costs of such new targeted therapies are high, the FDA strongly recommends to develop companion diagnostic tests. Recently, KRAS mutation has been associated with resistance to EGFR-targeted therapies in both colorectal (CRC) and Non Small Cell Lung cancer (NSCLC) and its detection has been recommended by the National Comprehensive Cancer Network. However, a large proportion of tumors expressing EGFR and KRAS wild type (wt) remain unresponsive to anti-EGFR and efforts should continue for a better selection of cancer patients that will benefit from this therapy. Based on the micro-array profiling of a large collection of patient derived tumor growing sc in nude mice and treated with C at 30 mg/kg, qd7x3, Oncotest previously identified a 26 gene expression signature (26-GES) predicting response to C. In this study, we investigated the 26-GES to further discriminate tumors responding to C among models being KRAS wt and expressing but not over-expressing EGFR. C was tested in 59 established cancer xenografts originating from CRC (20), NSCLC (29), gastric (5) and head & neck (5) cancer patients, respectively. Using a T/C of <37% as a cut-off for activity, 20% (12/59) of all tumors were sensitive, 7/29 NSCLC, 2/20 CRC, 1/5 gastric and 2/5 head & neck. The tumor biomarker examinations were done by sequencing for KRAS mutational status and by using Affymetrix HG U133 Plus 2.0 arrays and qRT-PCR for expression analysis. The 26-GES was combined with the in vivo response classification data by a Support Vector Machine (SVM) as class prediction algorithm. The results were compared to C activity in the corresponding tumor models as determined by in vivo testing. KRAS mutations were found in 21 out of 59 (36%) tumors, (11/20) CRC, (8/29) NSCL, (2/5) gastric and (0/5) head & neck and all except one were unresponsive to C. In the KRASWT tumors, as expected EGFR mRNA levels were significantly associated to the response to C (p=0.03), 2 tumors with gene amplification and overexpression of EGFR responded and the 4 tumors poorly expressing the receptor not. In the remaining of 32 tumors (KRAS wt, expressing but not over-expressing EGFR), the 26-GES-based prediction using SVM discriminated 10 tumors as responsive and 22 tumors as unresponsive to C. In the group predicted as unresponsive, median T/C value was 76.5% and in the group predicted as responsive median T/C value was 11.5% (p<0.001). In this group of 32 tumors, 26-GES increased response rate to C from 9/32 (28%) to 8/10 (80%). The 26-GES-based predictor of C response was shown to be useful for stratification of tumors with KRAS wt and expressing but not over-expressing EGFR. The recent development and validation of a qRT-PCR-based test performable in Formalin Fixed Paraffin Embedded samples will allow the retrospective and prospective validation of the test. Citation Information: Mol Cancer Ther 2009;8(12 Suppl):C120.