Identification of a novel predictive genomic biomarker for response to combination bevacizumab in metastatic colorectal cancer (mCRC).

Abstract

3580 Background: Somatic copy number alterations (SCNA) are genomic alterations evident in cancers including mCRC. These alterations support biomarker discovery, allowing identification of variants that can be used to predict response to therapy. Herein, we studied the impact of SCNAs on mCRC patient response to bevacizumab (BVZ). Methods: SCNA data was assembled from mCRC tumors in the TCGA cohort (n = 676), from the CAIRO 2 study [n = 143] and from the ANGIOPREDICT cohort (n = 258) [Betge et al. Digestion 2016 94(3):129-137]. GISTIC v2.0 was used to identify the most frequent and overrepresented chromosomal aberrations. A region was considered deleted if the logR value was < 0.1 and amplified when the logR was > 0.1. A cutoff q-value of 0.25 was used to select significantly overrepresented SCNAs. To further explore the impact of new ANGIOPREDICT clusters and consensus molecular subtypes (CMS), [Guinney J et al Nat Med. 2015 21(11):1350-6], a panel of seven xenografts representing each CMS subtype was treated with FOLFOX (40mg/kg 5-FU, 13.4mg/kg folinic acid & 2.4mg/kg oxaliplatin) + B20 antibody (mouse avastin, 10mg/kg) for 4 weeks. Results: Unsupervised hierarchical clustering classified all 1077 tumors into 3 consensus SCNA subgroups termed ‘ANGIOPREDICT’ clusters 1-3. Concordance between CMS and ANGIOPREDICT clusters was evident: CMS1 - Cluster 1 78%, CMS3 - Cluster 1 50%. CMS2 - Cluster 2/3 92% and CMS4 - Cluster 2/3 84%. Tumors with intermediate or high copy number instability (ANGIOPREDICT cluster 2 & 3) showed improved progression free survival (PFS): 369d vs. 227d p = 1.36e-7) whereas tumors with low copy number alterations displayed a poor response to BVZ therapy (PFS: 147d vs 152d p = 0.906). Xenografts corresponding to CMS 2&4/ ANGIOPREDICT cluster 2-3 showed improved response to FOLFOX + B20 therapy and increased PFS (p < 0.001) vs. CMS1&3/ ANGIOPREDICT cluster 1. Conclusions: For the first time, in a large test and validation cohort including retrospectively collected and randomized controlled trial patient samples, we have identified a predictive genomic biomarker for BVZ therapy in mCRC. We have further demonstrated the utility of CMS subtyping to stratify patients to BVZ .