Quantitative ex vivo synergy profiling uncovers heterogeneous combination responses in AML primary samples

Abstract Combination chemotherapy is a cornerstone of acute myeloid leukemia (AML) treatment, yet there remains a lack of patient-specific data to guide rational selection of drug combinations. Ex vivo drug sensitivity testing offers a functional approach to personalize treatment, but most efforts have focused on single-agent efficacy. Here, we apply a high-resolution ex vivo assay to quantify synergy across clinically relevant drug pairs in primary AML samples and demonstrate its potential to inform individualized combination strategies. To assess the potential of ex vivo synergy profiling for optimizing AML combination therapies, we obtained bone marrow (BM) aspirates from 15 patients with newly diagnosed (27%) or relapsed/refractory (R/R, 73%) AML. Mononuclear cells were isolated via density gradient centrifugation, red blood cells were removed, and the remaining viable leukemic cells were suspended in a custom culture medium for functional testing. Cells were seeded into 384-well plates and exposed to various concentrations of five drug pairs—two venetoclax (VEN)-based (VEN+decitabine, or VEN+azacitidine) and three cytarabine (CYT)-based (CYT+daunorubicin, CYT+mitoxantrone, or CYT+idarubicin). Drug concentrations for synergy testing were selected based on IC10 and IC20 levels derived from single-agent assays. Cell viability was measured after 72 hours using a viability assay, and synergy was quantitatively assessed using three complementary models: Combination Index (CI), Combination Sensitivity Score (CSS), and the cross-term interaction coefficient from the Quadratic Phenotypic Optimization Platform (QPOP). To validate our synergy analysis pipeline and optimize assay conditions, we first conducted preliminary experiments using HL-60 cells, a well-characterized CD20-negative leukemia cell line. Drug synergy, as measured by the CI, was highly sensitive to drug concentration, with a shift from synergy (CI <1) to antagonism (CI >1) observed upon modest reduction of cytarabine levels in a fixed CYT+idarubicin combination. QPOP-based 3D response surfaces mirrored these findings, showing steep concave curvature at synergistic concentrations and flattened profiles at antagonistic ones. Using optimized IC10 and IC20 drug concentrations, cytarabine-based pairs—particularly CYT+mitoxantrone—demonstrated the strongest synergy (CSS = 74, QPOP cross-term = –2092), whereas venetoclax-based combinations yielded uniformly lower synergy (CSS ≤ 22) compared to the former. These results confirmed the sensitivity and discriminatory power of the assay and revealed model-dependent differences in synergy ranking across drug pairs. We next applied the optimized ex vivo protocol to primary mononuclear cells from 15 AML patients to evaluate synergy across five clinically relevant drug pairs. Results revealed substantial inter-patient heterogeneity in response to a given combination, as well as intra-patient variability in synergy across different pairs. Synergy scores were strongly correlated between combinations sharing a common agent (e.g., VEN+DEC vs. VEN+AZA), but only weakly correlated across mechanistically distinct combinations (e.g., VEN+DEC vs. CYT+IDA). Cytarabine-based combinations yielded the highest synergy in 11 of 15 patients (73%), with CYT+mitoxantrone dominating this group (n = 8). Venetoclax-based pairs were top-ranked in 4 cases (27%). In three patients, synergy was comparably high for both VEN- and CYT-based pairs, whereas another three exhibited uniformly low synergy across all combinations. These findings underscore the potential of ex vivo synergy profiling to stratify AML patients for individualized combination therapy selection.