Identification of transcriptomic determinants to predict the efficacy of anti-EGFR treatment of colorectal cancer patients.

Abstract

130 Background: It is now well known that in colorectal cancer (CRC) activating mutations in KRAS, NRAS, and BRAF confer resistance to anti-EGFR therapy. However, only ~ 40% of patients with RASWT (KRASWT/NRASWT/BRAFWT) tumors respond to therapy. We have used transcriptomic data in addition to somatic mutations to decipher intrinsic resistance mechanisms and identify novel biomarkers of anti-EGFR response in CRC. Methods: The transcriptomic profiles of metastatic CRC patients treated with the anti-EGFR drug cetuximab plus chemotherapy (n = 193) or as a single agent (n = 80) were retrospectively obtained from two independent clinical cohorts (Okita et al., 2018; & Khambata-Ford et al., 2007). Tumor samples were further divided into RASWT and RASMut groups and subsequently in four consensus molecular subtypes (CMS) based on their transcriptomic profile. The transcriptomic profiles of cetuximab treated CRC cell lines (n = 148, Medico et al., 2015) and PDX models (n = 230, Bertotti et al., 2015) were used for validation. Results: In RASWT population of cetuximab plus chemotherapy cohort, the disease control rates (DCR) observed for cetuximab in the second line were 92% (33/36) for CMS2 and 83% (15/18) for CMS4 tumors relative to 50% (2/4) and 59% (13/22) for CMS1 and CMS3 respectively (Chi-square test p-value = 0.037). Similarly, in cetuximab single agent cohort, the DCR were 0% (0/4), 68% (15/22), and 29% (5/17) for CMS1, CMS2, and CMS4, respectively (Chi-square test p-value = 0.029). In the preclinical study of 148 CRC cell lines, 60% (12/20) of CMS2 cell lines with RASWT genes were sensitive to cetuximab, versus no sensitive cell lines from other CMS (CMS1 = 0/4; CMS3 = 0/5; CMS4 = 0/12). In the PDX study, 84% (11/13) of CMS2 tumor showed clinical benefit, followed by CMS4 with 75% (6/8) in RASWT group (CMS1 = 2/5; CMS3 = 2/4). Gene Set Enrichment Analysis (GSEA) of tumor tissue samples identified that Myc (NES = 1.32, FDR = 0.3), E2F (NES = 2.24, FDR = 0.02), and mTOR (NES = 1.54, FDR = 0.1) pathways were active in the cetuximab refractory RASWT CMS2 patients relative to cetuximab sensitive tumors. Similarly, these pathways were also activated (FDR ≤ 0.05) in cetuximab resistant RASWT cell lines and PDX models, suggesting that preclinical models are appropriate for testing of potential therapeutic combinations to overcome resistance to anti-EGFR therapy. Conclusions: Resistance to anti-EGFR therapy in extended RASWT CRC tumors is mediated in part by transcriptional activation and can be predicted by consensus molecular subtype (CMS) and quantifying transcriptional states of Myc, E2F, and mTOR pathway gene sets.