The majority of drugs tested in clinical trials for hematologic malignancies ultimately fail to confer benefit. In many cases, this results from the inability to accurately identify patients most likely to benefit and mechanisms that mediate resistance. At the same time, clinical trials commonly fail to capture essential specimens at early and late timepoints that could be used to identify predictors of response, pharmacodynamic markers of on-target activity and targetable resistance pathways. To overcome these challenges, we have developed a repository of >200 patient-derived xenografts (PDXs) of leukemias and lymphomas now available through the DFCI Public Repository of Xenografts (PRoXe; http://www.proxe.org). Because PDX models can be developed and propagated from a large number of patients, randomized in vivo studies using adequate numbers of PDXs offer the first opportunity to capture the diversity of disease biology in pre-clinical drug testing. We tested the novel MDM2 inhibitor CGM097, which is currently undergoing clinical testing in solid tumors, in 24 B-cell acute lymphoblastic leukemia (B-ALL) PDXs (including hypodiploid, near haploid, MLL- rearranged, CRLF2-rearranged, and BCR-ABL models) in a randomized, phase II-like trial. Only a small subset of B-ALLs harbor de novo TP53 mutations, suggesting that MDM2 antagonists may have broad activity in this disease. Each PDX was injected into 4 NOD.SCID.IL2Rɣ-/- (NSG) mice. Upon engraftment (>2% hCD45+/hCD19+ cells in the peripheral blood), mice were randomized to vehicle or CGM097 treatment arms. One animal from each treatment arm was sacrificed 26 hours after beginning treatment to examine pharmacodynamic endpoints. The remaining two mice continued on daily therapy until moribund. CGM097 markedly improved overall survival (median 73 vs 28 days for vehicle; p=0.0008). All 19 models with survival benefit from CGM were TP53 wild-type. Among 6 models (all TP53 wild-type) derived from patients with relapsed disease, the median survival improvement compared to vehicle was 53 days (p=0.0059), consistent with robust single-agent activity in relapsed disease. Specimens at the 26 hour timepoint and upon progression to moribund were captured from the majority of mice in the trial, allowing for comprehensive characterization of the trial population. Dynamic BH3 profiling (Montero et al. Cell 2015), in which CGM097 or vehicle is added to leukemia cells harvested from mice and the effect of CGM097 on "priming" for apoptosis was performed on 10 models and demonstrated 100% accuracy in predicting response to CGM097. To characterize the effects of MDM2 inhibition on p53-dependent gene expression, we measured expression of 120 p53-related genes using a custom Nanostring gene expression panel. Differential expression analysis identified 11 genes that were significantly upregulated (p≤0.05) by CGM097 treatment at the 26 hour timepoint, including the canonical p53 targets BBC3, CDKN1A and MDM2. All mice treated with CGM097 ultimately became moribund from progressive leukemia. Targeted deep sequencing identified acquired TP53 mutations in only 2 leukemias after progression on CGM097. This indicates that p53 mutation is not the primary genetic driver of resistance to MDM2 inhibition in B-ALL PDXs. Despite this, CGM097-dependent transcriptional changes were largely abolished in the majority of leukemias collected from mice upon progression on CGM097. In summary, we established a paradigm for "Phase II-like" trials in panels of human leukemia PDX models. With this approach, we defined CGM097 as a highly active agent across the diverse spectrum of TP53-wildtype B-ALL, and established 19 independent models of acquired resistance that are the ideal reagents for defining mechanisms and then testing combinations in vivo that overcome those mechanisms. The same paradigm could be applied as a new standard for pre-clinical testing of drugs to minimize the empiric nature of current drug development strategies. DisclosuresBarzaghi-Rinaudo:Novartis Institutes for Biomedical Research: Employment. Letai:AbbVie: Consultancy, Research Funding; Tetralogic: Consultancy, Research Funding; AstraZeneca: Consultancy, Research Funding. Jeay:Novartis: Employment, Equity Ownership. Wuerthner:Novartis: Employment, Equity Ownership. Halilovic:Novartis: Employment, Equity Ownership.