Genetically engineered murine tumor models for drug response prediction: population based pathway profiles predict response to rapamycin in a HER2 driven breast model

2230

Traditional preclinical models of human cancer have been of limited utility in guiding clinical development of anti-cancer agents, such that successful inhibition of human cell line based xenograft tumor provides little indication of response in the respective human tumor type. To further confound this situation, clinical experience with both chemotherapeutic and targeted agents reveals significant variation in response even among subpopulations of human tumors, in some cases revealing significant variation in the genetic context of tumors within the subpopulation (e.g. EGFR inhibitors in EGFR mutant subsets of NSCLC, or the 30-50% response to Traztusumab in advanced HER2 amplified breast cancer).

To better understand this variation in response, and to more accurately model populations of human cancers, we have established a platform of in vivo genetically engineered tumors in which defined genetic alterations are combined with spontaneous and naturally arising variation to create populations of in vivo tumors suitable for preclinical drug response testing.

To create a population based model of human HER2 amplified breast cancer, we used novel ES cell based chimeric approach to generate over 100 murine breast adenocarcinomas engineered to combine HER2 overexpression and INK4A/ARF-/- knockout. Each of these primary tumors was then propagated in vivo to generate an archive of frozen tumor material suitable for inoculation into thousands of recipient immunocompromised host mice. Tumors from these archives were characterized for phenotypic and molecular variation, and found to show significant variation in growth rate, stromal involvement, angiogenesis, response to taxotere, and microarray profile. To provide a molecular characterization of this variation, we then created a novel computational approach to assess the coordinated variation in gene expression among genes involved in various biological functions. Using this approach we generated expression based indices that revealed variation among the population in proliferation, and hypoxia as well as variation in several signal transduction pathways, including mTOR pathway signaling.

The mTOR signaling index enabled us to make a prospective prediction that a subset of HER2 driven breast tumors should be sensitive to in vivo rapamycin treatment. Pilot rapamycin tumor response studies have now revealed variation in rapamycin sensitivity that correlates with the prospective prediction. Variation in the same mTOR signaling index is observed in human breast cancers providing a hypothesis for predicting rapamycin response in subsets of human breast cancer.