Abstract B1: Gene expression and TP53 mutation analysis predict sensitivity of leukemia cells to MDM2 inhibition by DS-3032b

Abstract

Abstract Background: MDM2 overexpression, by preventing p53 activation, contributes to the growth and development of a variety of solid tumors and hematologic malignancies; hence, MDM2 inhibition could be a promising novel therapeutic strategy. Several MDM2 inhibitors have shown promise in early clinical trials. While preclinical studies generally reveal a requirement of wild-type (wt) TP53 for activity, tumor response to MDM2 inhibitors varies widely in the clinic and may not be strictly linked to TP53 mutational status. Identification of predictive biomarkers is therefore needed to enrich for patients with high likelihood of response. We here propose two gene signature-based models to predict the sensitivity of AML cells to MDM2 inhibition using two different methods. Methods: Leukemia samples isolated from peripheral blood or bone marrow of patients with newly diagnosed or relapsed/refractory AML were treated using DS-3032b (Daiichi-Sankyo), a dispiropyrrolidine-based, highly potent MDM2 inhibitor currently undergoing clinical trials in solid and hematological malignancies. Forty-one primary AML samples were treated ex vivo for 48 hours with DS-3032b (0, 25, 50, 100, 250, 500, and 1000 nM), and live cell numbers were determined. To define drug sensitivity/resistance, area under the curve (AUC) values, based on%live cell number measured at each concentration, were calculated. Baseline whole-genome RNA expression profile (Affymetrix Human Genome U133 Plus 2.0 Array) and TP53 mutation status (next generation sequencing) were determined. In the first model, we validated a predictive 175-gene signature that was established in a wide range of cancer cells by Daiichi Sankyo. In the second model, we used the random forest method with cross validation to establish a new predictive gene signature. Results: Eight samples (20%) had TP53 mutations. 6/8 (75%) p53 mutant and 8/33 (24%) of p53 wt samples were resistant (p = 0.01). In the first model, 11 each p53 wt samples were selected as sensitive or resistant to DS-3032 based on AUC values, and the 175-gene signature was applied. The prediction accuracy was 72%. In the genotype mixed samples, 14 each sensitive and resistant samples were selected, and the prediction accuracy was 79%. In the second model, we focused on 33 p53 wt samples and trichotomize the samples in the same way as in the first model, and investigated the accuracy of gene expression-derived prediction model with (A) 1500 gene set with the highest variance in mRNA expression (unbiased approach), (B) 32 gene set derived from previous studies (referenced approach), (C) combined (A+B) gene set. The sensitivities to predict cases with high drug sensitivity were 72%, 73% and 82% in scenarios (A), (B) and (C), respectively. The analysis was then extended to all 41 samples and the sensitivity to predict cases with high drug sensitivity remained high (64%, 64% and 72%). The results indicate that an unbiased approach can create a prediction model as accurate as the referenced approach, and moreover, that the combining approach can provide the highest prediction of sensitivity to the MDM2 inhibitor. Conclusion: The two models reported here could provide a novel strategy to identify the optimal gene signatures for predicting the cases most sensitive to MDM2 inhibitors prior to therapy. These models will be tested in an ongoing AML phase 1 clinical study of DS-3032b. Citation Format: Jo Ishizawa, Kenji Nakamaru, Takahiko Seki, Koichi Tazaki, Kensuke Kojima, Dhruv Chachad, Archie Tse, Arvind Rao, Michael Andreeff. Gene expression and TP53 mutation analysis predict sensitivity of leukemia cells to MDM2 inhibition by DS-3032b. [abstract]. In: Proceedings of the AACR-NCI-EORTC International Conference: Molecular Targets and Cancer Therapeutics; 2015 Nov 5-9; Boston, MA. Philadelphia (PA): AACR; Mol Cancer Ther 2015;14(12 Suppl 2):Abstract nr B1.