Abstract B18: Variation in response to VEGFR inhibitor tivozanib in a unique population‐based tumor model enables the development of a multigene response biomarker

Variation is a hallmark of human tumors. Within any human tumor population, significant inter‐tumor variation in genetic context underlies variation in phenotype, prognosis, and response to a given therapeutic agent. The ability to identify associations between genetic context and drug response lies at the core of much of translational research in cancer. However, current preclinical models fail to adequately represent this variation in a manner suitable for correlating genetic context with drug response. To address this challenge, we used complex genetically engineered murine tumors to develop a population based model comprising over 100 HER2 INK4A‐/‐ driven breast tumors, each with associated expression microarray, array CGH, histology, as well as protein and biochemical characteristics. Similar to that observed in human tumor populations, this breast tumor archive exhibited significant inter‐tumor variation in RNA and DNA profiles, and variation in many measurable tumor phenotypes, including tumor vasculature. The VEGF/VEGFR axis is postulated to be the dominant human tumor angiogenesis mediator. However, in contrast to robust activities observed in traditional xenograft models, anti‐VEGF/VEGFR agents have thus far elicited relatively modest activity in human clinical studies (e.g. single agent RECIST response rates vary between 0–11% in several breast cancer studies). To better model the observed human variation in response to VEGF pathway antagonism, and to explore the development of a predictive biomarker for patient selection, we determined the responsiveness of tumors across the population based model to a potent, selective VEGFR inhibitor, tivozanib (AV‐951). Tivozanib exhibits picomolar inhibitory activity against all three VEGF receptors, is active in a broad array of traditional xenografts, exhibits a multiday T1/2 in humans, and demonstrates robust clinical activity in renal cell carcinoma, with RECIST response rates of 25–40%. Interestingly, a significant variation in response to tivozanib was observed among 25 tumors in the population model, with the majority exhibiting intrinsic resistance to tivozanib9s anti‐angiogenic activity. Combining the efficacy data, extensive IHC analyses, the comprehensive expression profiles and a novel bioinformatics approach, we identified a multigene biomarker of on‐target drug resistance. Using the same multigene biomarker identified in the murine model to examine human tumor microarray datasets, subsets of breast, lung, colon and kidney cancer human populations were predicted to exhibit tivozanib resistance. These data demonstrate the promise of population based preclinical models for translational research and provide a readily testable multigene biomarker for the potent VEGF pathway antagonist tivozanib. Citation Information: Mol Cancer Ther 2009;8(12 Suppl):B18.