Modeling Human Cancer Therapy Response in Patient Derived Xenografts

Patient‐derived xenografts (PDXs) generated by implanting human tumor tissue into a transplant compliant mouse host have been of increasingly importance to preclinical development and have been demonstrated to have advantages compared to cancer cell lines and cell‐line xenografts (CLX) for modeling therapeutic responses in cancer. Nevertheless, many open questions remain regarding the relationship between study design factors and classification of treatment response and the molecular fidelity of tumors passaged in PDXs relative to the original patient tumor(s). The research described in this dissertation addresses both of these significant issues related to the use of PDXs as a tool for modeling human cancer therapy response. This work investigated how study design factors (enrollment criteria, group size, study duration) and data analysis method influenced treatment responses in PDXs. Each of the methods evaluated consistently classified responsiveness which suggests it is feasible to ii combine response data across studies that apply different methods assuming consistent response value thresholds have been used. The results demonstrated that a cohort size of three is sufficient for identifying the four highly responsive and nine highly non‐responsive cisplatin treated tumors, suggesting that the use of a low cohort size to screen for chemotherapies that have a high degree of activity or models that are highly responsive is possible. Mutational and expression profiles of engrafted tumors were compared with the original patient tumors and demonstrated that the molecular fidelity of PT and passaged tumors is high at a global level; the majority of variants with known clinical significance are preserved between PT and PDX tumors. However, individual markers are not invariably concordant and in some cases the molecular discrepancies observed between PDX and PT occurred in variants that are clinically relevant to either disease prognosis or to selection of therapy. Confirming the mutation status of donor tumors used to establish testing cohorts should be routine practice when PDXs are used as models for modeling treatment responses of individual patients