Identification of Effective Combinations of Targeted Drugs for DNMT3A and NPM1 Mutated Acute Myeloid Leukemia

Introduction: As the molecular heterogeneity underlying cancer becomes better understood, the capacity to define actionable targeted therapeutic strategies for patients appears essential to improving outcomes. These challenges are exemplified in patients with acute myeloid leukemia (AML), where a broad spectrum of driver mutations, emergence of genetically heterogeneous subclones, rescue signals in the microenvironment, and tumor-intrinsic feedback pathways frequently result in a lack of durable responses and rapid development of resistance to many single-agent targeted therapies. To identify biomarker-driven effective combination strategies for patients with AML, we built upon our previous ex vivo profiling of targeted drug combinations in primary patient specimens through integrated analysis of an expanded cohort of 133 AML patients.

Methods: Freshly isolated primary mononuclear cells from patients were cultured in the presence of a panel of 48 drug combinations in fixed molar dose series encompassing different classes of targeted small-molecule inhibitors already approved by the FDA or in clinical development. For comparison, cells were also tested against graded concentrations of each single-agent, and sensitivity was assessed by MTS-based viability assay. IC₅₀ and AUC values were derived using a probit regression model. Efficacy of each combination relative to its respective single-agents was calculated as a Combination Ratio (CR) value, defined as the combination IC₅₀ or AUC divided by the lowest single-agent IC₅₀ or AUC value, where a CR value < 1 indicates enhanced efficacy of the combination. Associated clinical, prognostic, mutational, cytogenetic, and surface antigen data were compiled for comparisons with combination efficacy.

Results: Analysis of our discovery cohort of 58 AML patients revealed several associations between molecular features and combination sensitivity. An intriguing finding was the significantly enhanced efficacy of combinations involving the bromodomain inhibitor, JQ1, in patients with mutations in NPM1 or DNMT3A. Specifically, the combination of JQ1 with the multi-kinase inhibitor sorafenib demonstrated superior efficacy compared to each inhibitor alone in patients with NPM1 mutations (median IC₅₀ CR: 0.437; p=0.010). JQ1 paired with the CDK4/6 inhibitor, palbociclib, demonstrated superior efficacy compared with single-agents for patients with DNMT3A mutations (median IC₅₀ CR: 0.119; p=0.017). These associations between the presence of NPM1 or DNTM3A mutations and increased JQ1-inclusive combination efficacy were also validated using our expanded AML cohort of 133 patients. For patients with NPM1 mutations, one-way ANOVA based comparison of AUC values for JQ1, sorafenib, and the combination showed JQ1 as the more effective of the two single agents, with the combination significantly reducing AUC beyond either drug alone (p<0.001 versus sorafenib; p<0.05 versus JQ1). Likewise, among patients positive for DNMT3A mutations, JQ1 showed superior efficacy compared to palbociclib alone, and the combination of JQ1 and palbociclib demonstrated significantly further reduced AUC relative to each drug alone (p<0.001 versus palbociclib; p<0.05 versus JQ1). Similar results were obtained for both combinations and their respective mutations in both discovery and validation patient sample cohorts. Additional synergy studies and analysis of the association of RNA-seq-based gene expression and JQ1-inclusive combination efficacy in the context of these mutations is ongoing.

Conclusions: Among the observed patterns of ex vivo drug combination efficacy associated with select, actionable diagnostic and genetic subsets in our study, our data suggest that AML patients harboring NPM1 or DNMT3A mutations demonstrate moderate sensitivity to the bromodomain inhibitor JQ1 and their viability is further inhibited by combinations of JQ1 with sorafenib or palbociclib, respectively. More broadly, these findings warrant clinical evaluation and highlight the utility of integrating drug sensitivity data with genomic information to identifying novel treatment strategies for improving disease control and patient outcomes.