Ex Vivo Drug Response Profiling Defines Novel Drug Sensitivity Patterns for Predicting Clinical Therapeutic Responses in Myeloid Neoplasms

Introduction: Myelodysplastic syndromes (MDS), chronic myelomonocytic leukemia (CMML) and acute myeloid leukemia (AML) are clonal myeloid neoplasms for which limited conventional treatment options exist in the relapsed / refractory setting, especially for older patients.

Methods: We provide proof of concept data by drug sensitivity profiling of 64 samples from 52 adult patients with both newly diagnosed and treatment refractory myeloid neoplasms (MDS = 38, CMML = 4, AML = 10). Fresh mononuclear cells from bone marrow aspirates and/or peripheral blood specimens were red blood cell lysed upon arrival and re-suspended in serum free media with cytokines. The samples were then plated in 384 well microtiter plates and were screened against a collection of investigational and FDA approved compounds (up to 85) in triplicate using a Labcyte Echo acoustic liquid handler. These specimens were treated for 72 hours and analyzed for drug sensitivity on an Intellicyt iQue Screener PLUS flow cytometer for both cytotoxicity and differentiation.

Results: Principal component analysis was performed to explore differential ex vivo sensitivity and resistance patterns between the MDS and AML samples. Individual MDS samples clustered according to their ex vivo responses, with distinct subgroups enriched for sensitivity to hypomethylating agents (HMAs), HDAC inhibitors, differentiation agents, and PARP inhibitors. Strong clustering of drug classes was also observed with distinct drug classes showing high correlation within samples. Clinical parameters contributed but did not explain all of the variability of ex vivo sensitivity patterns, including prior HMA therapy, cytogenetics, prognostic risk category and mutational profile. In patients resistant to or relapsed from HMAs where next line of treatment matched profiled compounds (n = 19), ex vivo drug sensitivity data demonstrated positive (PPV) and negative predictive values (NPV) of 85% and 83%, respectively, for clinical response prediction (Fisher's exact test p-value = 0.0095). In patients with serial samples, ex vivo sensitivity data corresponded to emergence of clinical resistance.

Conclusions: This novel platform, applied to predict ex vivo therapeutic responses of patient samples to various classes of drugs, recapitulates known clinical and molecular predictors of therapeutic efficacy and provides possible new biologically focused therapeutic options. The accuracy and reproducibility of this method coupled with short turnaround time demonstrate the potential of such an approach as a decision support system for therapeutic selection in the management of various myeloid neoplasms. A prospective study is ongoing to assess the feasibility of this technique for treatment decision support purposes in HMA-refractory MDS patients.