Antileukemic efficacy of a potent artemisinin combined with sorafenib and venetoclax

Most patients with acute leukemias are still treated principally with the same drug regimens used for the past 4 decades, even though these therapies result in substantial mortality and serious morbidities. Although current treatments cure most children with acute lymphoblastic leukemias (ALLs), ∼20% of pediatric and ∼50% of adult ALL cases still relapse, particularly in poor prognostic categories including cases harboring mixed lineage leukemia chromosomal rearrangements (MLLr).^1,2  Relapsed ALL generally requires intensive treatment, such as chimeric antigen receptor–T-cell and/or allogeneic stem cell transplantation.^3,4  Acute myeloid leukemia (AML) is cured in only ∼30% of patients treated in the United States,^5,6  and standard AML chemotherapy causes severe morbidity with treatment-related deaths in ∼10% of younger patients, precluding use in frail senior patients; this is a substantial limitation because the median AML patient is 67 years old.^7,8  AML is now recognized to be an extremely heterogenous disease characterized by parallel evolution and emergence in an individual patient of multiple codominant cellular clones, each harboring a different set of mutations as well as chromatin modifications that result in gene-expression abnormalities.^9,10  This molecular understanding not only rationalizes disappointing clinical observations that even mutation-targeted inhibitors have predominantly induced only partial responses as single agents in patients,^11,12  but also motivates efforts to develop novel effective, low-toxicity multiagent combinations of antileukemic agents.^13-15 

Artemisinins, a large class of derivatives of the natural compound artemisinin, are used worldwide to treat malaria with essentially no human toxicity.^16,17  Artemisinins are also active against multiple human cancer cell types,^18-29  including leukemia cells resistant to standard clinical antineoplastic drugs and harboring poor-prognosis genetic aberrations, for example, BCR-ABL, FMS-like receptor tyrosine kinase-3 internal tandem duplication (FLT3-ITD), and MLLr.^18-20,30,31  We reported that artemisinin-derived trioxane diphenylphosphate dimer 838 (ART838)^32-34  was ∼88-fold more potent than artesunate (AS), the major clinical artemisinin derivative, against 23 human AML and ALL cell lines tested, with minimal in vitro toxicity to normal human hematopoietic stem-progenitor cells (HSPCs).¹⁸  Our prior observation that artemisinins synergized in vitro with kinase inhibitors, including sorafenib (SOR), against leukemia cell lines¹⁸  stimulated us to seek additional targeted drugs that might synergize with ART838 and SOR as a tolerable 3-drug combination. As we chose here to emphasize AMLs containing the prevalent and particularly poor-prognosis MLLr and FLT3-ITD mutations, all experiments depicted in our main figures evaluated MOLM14 and/or MV4;11 AML cell lines, each harboring both MLLr and FLT3-ITD, and additional acute leukemia cell lines and primary cases were investigated less extensively.

Leukemia cells cultured with ART838 reduced myeloid leukemia cell differentiation protein (MCL1) and upregulated integrated stress response messenger RNA (mRNA)/protein DNA damage–inducible transcript 3 (DDIT3)/CCATT/enhancer-binding protein (C/EBP) homologous protein (CHOP), potential biomarkers for artemisinin action. SOR reduced MCL1 with minimal CHOP elevation, suggesting that SOR plus ART838 plus venetoclax (VEN; SAV) combine to target MCL1 and B-cell lymphoma 2 (BCL2) via discrete mechanisms. The SAV combination demonstrated strong synergy in vitro and clear efficacy/tolerability in AML xenograft and primagraft models. Thus, SAV warrants further development for leukemia treatment.

Supplemental Table 1 details the reagents used.

Leukemia cell lines, primary cells, and luciferase (luc)-yellow fluorescent protein (YFP) labeling of AML cells are described in supplemental Table 2 and supplemental Methods. Western blots of several of the cell lines and primary cases for BCL2, MCL1, and BCLxL are in supplemental Figure 1, along with the 50% inhibitory concentrations (IC₅₀s) of competitive inhibitors of these BCL2-family proteins.

Cell metabolism was assessed by alamarBlue assays (Life Technologies, Grand Island, NY) following the manufacturer’s guidelines. Cell death was verified by Annexin V/7-aminoactinomycin D (7AAD) staining (BioLegend, San Diego, CA) per the manufacturer’s guidelines.¹⁸ 

Immunodeficient NOD.Cg-Rag1^tm1MomIl2rg^tm1Wjl/SzJ (NRG) mice were bred at the University of Maryland Baltimore (UMB) from breeders purchased from The Jackson Laboratory (JAX; Bar Harbor, ME). Seven to 14 days after IV (tail vein) transplant of 0.5 × 10⁶ to 1 × 10⁶ luc/YFP-labeled leukemia cells, luminescence (AML burden) was quantitated on treatment day 0 in each NRG mouse by bioluminescence imaging (BLI) (Xenogen IVIS Spectrum; PerkinElmer, Waltham, MA). Mice were allocated to treatment groups (usually ≥5-10 mice per group) so that each group had similar average day 0 luminescence, then groups were administered drug orally (by mouth; gavage). Luminescence of each mouse was assessed over time and compared with that mouse’s day 0 AML burden (fold-change). Clinical behavior, appearance, weight, and survival were also monitored.

Cellular protein was extracted, quantitated, electrophoresed, and western blotted with monoclonal antibodies. Antibody-specific band intensities were quantitated by densitometry (ChemiDOC XRS⁺ and Image Laboratory system; Bio-Rad, Hercules, CA) and normalized to β-actin.³⁵ 

Total RNA was isolated, complementary DNA was synthesized, and SYBR Green quantitative polymerase chain reaction (PCR) was performed in triplicate (Applied Biosystems QuantStudio 6 Flex and Real-Time PCR software; ThermoFisher Scientific); primer sequences are shown in supplemental Methods. Cycle threshold values were normalized to glyceraldehyde-3-phosphate dehydrogenase.^35,36 

The different drug treatments of cells and mice are color-coded consistently across all figures. Statistical analyses were performed using Prism 8.4 GraphPad software (San Diego, CA). Data are presented as arithmetic mean plus or minus standard error of the mean (SEM) from ≥3 independent experiments unless otherwise indicated. AML burden (luminescence) fold-change values are geometric means. To compare experimental groups, P values were calculated by analysis of variance, followed by the Dunnett multiple-comparisons test unless otherwise indicated (*P < .05; **P < .01; ***P < .001; no asterisk, P > .05). For time-to-event end points of in vivo xenograft experiments, Kaplan-Meier survival curves were compared by log-rank (Mantel-Cox).

We screened a library of 111 targeted antineoplastics^37-43  to identify drugs that synergized with ART838 and/or AS.¹⁸  To identify broad synergism against AMLs and ALLs, screening was performed in 3 human acute leukemia cell lines: MOLM14 AML (harboring MLLr KMT2A-MLLT3, FLT3-ITD), KOPN8 ALL (harboring MLLr KMT2A-MLLT1), and RCH-ACV ALL (non-MLLr) (supplemental Table 2). Each drug was ranked by “synergy score,” the observed growth inhibition of the drug pair compared with the additive effect predicted by the Bliss independence model, averaged across all 3 cell lines^44-46  (supplemental Methods; Figure 1A; supplemental Figure 2A). Two of the 3 most synergistic drugs were the only BCL2-competitive inhibitors in the library: ABT737 (chemically similar to clinically tested navitoclax [NAV]; inhibits BCL2, BCLxL, and BCLW)^47,48  and ABT199 (renamed VEN; selectively inhibits BCL2).⁴⁹  Cooperativity of ABT737, VEN, or NAV with ART838 and AS was confirmed in MOLM14 AML (supplemental Figure 2B-D). Because NAV causes clinical thrombocytopenia,^50,51  we prioritized VEN for further studies focused on MOLM14 and MV4;11 AMLs (both containing MLLr and FLT3-ITD, along with other mutations).

In Chou-Talalay drug combination index (CI) synergy analyses of 48-hour alamarBlue assays,^52,53  VEN and ART838 exhibited significant synergy at their IC₅₀s (shown in bold in Figure 1B) against KOPN8 ALL (MLLr) and 2 MLLr AML cell lines: ML2 (harboring KMT2A-MLLT4) and MV4;11, neither tested in the synergy screen (Figure 1B). Significant synergy between VEN and AS was observed against MOLM14 AML and 5 other leukemia cell lines (supplemental Figure 3).

After 18-hour culture of KOPN8 ALL, ML2, or MV4;11 AML with VEN or ART838 alone, percentages of viable (Annexin⁻/7AAD⁻) cells were only slightly reduced; these drug concentrations and relatively short culture durations were used to avoid extensive cytolysis observed at later time points and with higher concentrations of single drugs. In contrast, parallel culture with the VEN plus ART838 combination resulted in significantly lower viable cell percentages with high levels of apoptotic/dead (Annexin⁺7AAD⁻ and Annexin⁺7AAD⁺) cells (Figure 1C; supplemental Figure 4). Apoptotic cell death was blocked by pretreatment with pan-caspase inhibitor Q-VD-OPh (QVD) (Figure 1C).⁵⁴  VEN similarly enhanced AS-induced apoptosis (supplemental Figure 5).

Because we found previously that SOR synergized with artemisinins,¹⁸  we next tested the antileukemic activity of the SAV combination. In 48-hour alamarBlue assays with each drug at ∼25%, ∼50%, and ∼100% (shown in bold in Figure 2A) of its IC₅₀ against MOLM14 and MV4;11 AML cells, SAV demonstrated synergistic CI values (Figure 2A).^52,53  Similar results were observed testing KOPN8 ALL and ML2 AML cells (supplemental Figure 6). SAV induced ≥95% apoptosis in flow cytometric assays of the tested 3 leukemia cell lines and 6 primary AML/ALL cases. SAV was statistically superior to VEN plus SOR against MV4;11, MOLM14, AML45, and AML1526. Although SAV was not substantially superior to VEN plus SOR in all tested cell lines/primary cases, it is possible that the high efficacy of VEN plus SOR obscured the additional contribution of ART838 and/or the high spontaneous apoptosis of primary cells cultured ex vivo reduced the dynamic range of these assays (Figure 2B; supplemental Figure 7).

After culture of primary human CD34⁺ HSPCs with SAV, normal hematopoietic colonies were reduced by only ∼20% (Figure 2C) relative to vehicle, similar to the effects of the single drugs. In contrast, parallel culture of MV4;11 cells with SAV reduced total AML colonies by >95% (Figure 2C).

Because artemisinins, including dihydroartemisinin (DHA), the active metabolite of several artemisinins, reduce MCL1 protein levels via the integrated stress response pathway⁵⁵  in ALLs harboring BCR-ABL,²⁰  we assessed whether ART838 (not metabolized to DHA^18,34 ) reduced MCL1 in AML/ALL cell lines lacking BCR-ABL. To avoid nonspecific cell death–associated reduction of MCL1, we used drug concentrations and a short (18-hour) artemisinin exposure where >85% of cells remained viable (Figure 2B). AS, ART838, or SOR reduced MCL1 protein levels in MOLM14, MV4;11, ML2 AML, and KOPN8 ALL cell lines (Figure 3A; supplemental Figure 8). Specificity was observed, in that none of these drugs affected BCL2 or BCLxL protein levels, VEN increased MCL1 levels, and cytosine arabinoside (AraC; the most effective clinical drug against AML^56,57 ) did not affect MCL1 levels. MCL1 mRNA levels were not changed by ART838 or AS (Figure 3B; supplemental Figure 9), and ART838 did not significantly alter MCL1 protein half-life in KOPN8 ALL or ML2 AML cells (supplemental Figure 10). Importantly, all of these leukemia cell lines are sensitive to the selective competitive MCL1 inhibitor S63845⁵⁸  (IC₅₀ range, 7-177 nM; supplemental Figure 1B), suggesting that reduction of MCL1 protein levels in these cells would similarly result in cell death.

Culture of MOLM14 and MV4;11 AML cells with ART838 or AS, but not AraC, induced more than fivefold to 10-fold elevations in levels of the integrated stress-response protein CHOP (Figure 3A; supplemental Figure 8) and DDIT3 mRNA (GADD153 [encodes CHOP protein] Figure 3B). SOR caused only 1.3-fold to 3.6-fold increases in DDIT3/CHOP levels. In time-course experiments, DDIT3 mRNA levels were elevated in MV4;11 AML cells cultured for ≥4 hours with ART838 (supplemental Figure 11A). These in vitro observations translated to in vivo, with a twofold to fourfold increase in DDIT3 mRNA in cells harvested 6 to 8 hours after ART838 administration from heavily leukemia-infiltrated bone marrow and spleens of luc/YFP-labeled MV4;11-bearing mice (supplemental Figure 11B). In contrast, deoxyART838, which lacks endoperoxide pharmacophores,^59,60  neither reduced MCL1 levels nor elevated CHOP levels (supplemental Figure 12). Additionally, levels of several other known MCL1-regulatory proteins, including translation initiation factor eIF4B,⁶¹  ubiquitin-associated protein FBXW7,⁶²  and deubiquitinase USP13,⁶³  were not altered by culture with any of these drugs (supplemental Figure 13A). GSK3β, another known regulator of MCL1,⁶⁴  was not altered by ART838 or AS, whereas SOR treatment reduced the phosphorylated (inactive) form of GSK3β in MOLM14 AML cells (supplemental Figure 13B), consistent with previous reports.⁶⁵  These results suggest that ART838 and SOR downregulate MCL1 protein levels mainly by distinct mechanisms.

For in vivo preclinical and potential clinical use, we measured the stability of ART838 in human and mouse plasma and liver microsomes. Under all tested conditions, ART838 was degraded by ≤20% (supplemental Table 3).

In pilot experiments in NRG mice, we determined the single-dose maximum tolerable dose (MTD) for ART838 as 250 mg/kg by mouth and the repeated dose MTD for ART838 as 50 mg/kg daily ×5 by mouth (supplemental Methods). Plasma concentrations of ART838 were >30 nM (the in vitro IC₅₀ for ART838 against MOLM14) for ∼8 hours, suggesting that daily dosing is reasonable. Based on these results, we developed a tolerable 5-day on/9-day off cyclic SAV treatment dose schedule (Figure 4A).

Novel drugs and combinations initially developed via in vitro experiments require in vivo evaluation to better model patient treatment. In several experiments, we treated NRG mice bearing luc/YFP-labeled MOLM14 or MV4;11 AML xenografts with ≥5 cycles of ART838, VEN, and/or SOR. As compared with treatment with vehicle, monotherapy with either SOR, ART838, or VEN slowed the increase in AML burden at day 7 (taken as time point of near-maximal response to the first treatment cycle). However, none of the monotherapies reduced day 7 MOLM14 or MV4;11 AML burden to levels substantially below day 0 luminescence. These results were confirmed at day 14, the earliest time when NRG mice can tolerate a second treatment cycle (supplemental Figures 15A-C and 16B-D, visualized best in the waterfall plots and line graphs). Consistent with these observations, Kaplan-Meier survival curves of MOLM14 or MV4;11 AML xenograft-bearing mice administered the 3 monotherapies were not markedly longer than for vehicle-treated mice in these experiments involving mice with substantial initial AML burdens (supplemental Figures 15 and 16).

Supplemental Figures 15A and 16A provide direct comparisons of results of the SAV regimen vs each of the 2-drug combinations against MOLM14 and MV4;11 AML xenografts, respectively; for convenience, the most relevant data are also shown in Figure 4, panels B-D and E-G. SAV treatment of MOLM14 xenografts resulted in the greatest fold reduction of day 7 AML burden, followed by VEN plus SOR, ART838 plus SOR, and ART838 plus VEN; day 7 AML burden was 83- and 250-fold reduced from day 0 by VEN plus SOR and SAV, respectively (Figure 4B-C). Assessments at subsequent time points confirmed the day 7–based rank order of AML burden reductions following SAV or the 2-drug combinations (supplemental Figure 15A). Survival was fourfold increased for both SAV and VEN plus SOR compared with vehicle (Figure 4D). All 5 mice from the VEN plus SOR group and 3 of 5 (1 mouse censored due to death with low AML burden) from the SAV group had undetectable AML burdens on day 103 (treatment ended on day 60; supplemental Figure 15A).

Similar results were observed in the MV4;11 AML xenograft model (Figure 4E-G; supplemental Figure 16A). Treatment with SAV or each of the 2-drug combinations substantially reduced day 7 AML burdens (Figure 4E-F). VEN plus SOR and SAV extended survival by threefold over vehicle (Figure 4G). All 5 mice from the VEN plus SOR group and 4 of 5 from the SAV group had undetectable AML burdens at experiment termination on day 91 (treatment completed on day 60; supplemental Figure 16A). These interpretations, summarized in Figure 4H, were confirmed by statistical analysis of all experiments in supplemental Figures 15 and 16, (not dependent on censoring of mice that died with low AML burdens). There was a highly significant inverse association between day 7/day 0 fold-change in AML burden and fold-change in survival relative to vehicle for MOLM14 (Spearman rank correlation coefficient, Rs = −0.98; P < .0001) and MV4;11 AML xenografts (Rs = −0.97; P < .0001). Because VEN plus SOR was highly effective against both these xenograft models, these models could not definitively ascertain the contribution of ART838 in SAV.

AML-bearing mice tolerated all treatment regimens without adverse clinical signs or treatment-limiting weight loss (<20%), and there were few unexpected deaths of mice with AML burdens undetectable or lower than on day 0 (censored in Figure 4). Blood cell counts were not significantly reduced in non–AML-bearing mice treated in parallel (supplemental Figure 17).

Each of the single drugs and combinations slowed the increase in AML burden over time vs vehicle using the luc/YFP-labeled human AML45 primagraft model,⁶⁶  but day 7 AML burden was reduced to below day 0 in only the SAV group and 1 of the ART838 plus VEN mice (Figure 5A-B; supplemental Figure 18A). All mice had AML burdens exceeding day 0 at all subsequent time points (supplemental Figure 18A). The SAV group had the longest survival (Figure 5C; twofold over vehicle). Figure 5G summarizes this and an additional experiment (supplemental Figure 18B) evaluating AML45 primagrafts. SAV was the most active regimen against AML45 primagrafts, and ART838 plus VEN was next best combination.

Similar results were obtained in our luc/YFP-AML1526 primagraft model, with SAV the most effective regimen (survival extended 1.6-fold over vehicle; supplemental Figure 19; Figure 5D-G). VEN plus SOR and ART838 plus VEN had modest effects on AML1526 burden and survival, and ART838 plus SOR was the least active combination.

As with the xenograft models, there was a significant inverse association between the day 7/day 0 fold-change in AML burden and survival for AML45 (Spearman rank correlation coefficient, Rs = −0.71; P = .048) and AML1526 primagrafts (Rs = −0.71; P = .050).

Artemisinins are cytotoxic to many acute leukemia cell lines,^18-20,23,25  suggesting that the potent ART838 analog, which we found to have favorable stability and pharmacokinetics, might serve as a component of leukemia treatment regimens. From our previous report that 3 tested kinase inhibitors (midostaurin, lestaurtinib, and SOR) each synergized with artemisinins,¹⁸  we chose to combine ART838 with SOR because SOR is a potent inhibitor of the FLT3-ITD mutation common in AML and also inhibits several other kinases (eg, KIT, PDGFR, VEGFR) that may be mutated or overexpressed to drive acute leukemias.^67,68  Moreover, SOR is approved by the US Food and Drug Administration (FDA), has modest toxicity,⁶⁹  and is an effective addition to standard clinical AML treatment regimens.^70,71  In this project, our goals were to identify an additional antileukemic drug that might be combined synergistically with SOR and ART838, and then to begin to evaluate this empirically derived 3-drug regimen for efficacy, mechanism, and tolerability, focusing on poor-prognosis AMLs harboring FLT3-ITD and/or MLLr mutations.^72,73 

Our screen of a library of targeted antineoplastics^37-43  revealed that both BCL2 inhibitors in the drug library, ABT737 and VEN, synergized strongly with ART838 (and AS). In vitro culture with ART838 plus VEN robustly reduced growth and enhanced apoptotic cell death of several human acute leukemia cell lines (5 AMLs, 2 ALLs) and primary acute leukemia samples (5 AMLs, 1 ALL), validating the results of the screen. Previous studies indicated that artemisinins induced high cellular levels of reactive oxygen species (ROS), triggered proapoptotic signaling,^18,30,74,76  and altered expression of BCL2 family members in several cancer types.^77-79  Notably, a recent study aimed at identifying drugs that synergize with the dual BCL2/BCLxL inhibitor, NAV, against poor-prognosis B-precursor ALLs harboring the BCR-ABL translocation, found that DHA synergized strongly.²⁰  Thus, both this and our own validated screen identified that artemisinins and inhibitors of BCL2 family members synergize against at least some acute leukemias, and our results extended this synergy beyond BCR-ABL⁺ ALLs. Because NAV induced thrombocytopenia in clinical trials,^51,80  we focused on the selective BCL2 inhibitor, VEN. Clinical trials of VEN as a single agent document an encouraging response frequency in AML, but responses were typically only partial and transient.¹¹  VEN combined with azacytidine or decitabine showed efficacy and low toxicity, which led to FDA approval of VEN-based combination regimens for frail senior patients, poor candidates for standard intensive AraC/anthracycline induction chemotherapy.⁸¹  Finally, SOR and VEN synergize against AML cell lines and patient samples in vitro.⁸² 

Further understanding of ART838’s mechanism(s) of action might explain its empiric synergies with SOR and VEN and suggest additional drugs that might cooperate. All artemisinins contain an endoperoxide pharmacophore(s), which upon activation by iron, induces high levels of intracellular ROS, necessary for artemisinins’ antimalarial and antileukemic activities.^30,74-76,83  High-level ROS evokes integrated stress response,^20,84,85  cell-cycle arrest,^18,86  and lysosomal disruption.⁸⁷  We assessed several proteins previously shown to regulate MCL1 protein; none were altered by ART838 exposure. Additionally, we inhibited mRNA translation, which did not significantly alter MCL1 protein abundance in the presence of ART838. This experiment takes into account only protein half-life; it does not rule out the possible modulation of MCL1 protein translation via reduced phosphorylated eIF4B.⁸⁸  We confirmed and extended the observation that artemisinins induce the integrated stress response, which increases DDIT3/CHOP mRNA/protein and reduces MCL1 protein levels in multiple subtypes of acute leukemia cells in vitro and in vivo,²⁰  suggesting that MCL1 protein and DDIT3/CHOP levels may serve as useful biomarkers of the antileukemic action of ART838 and SAV. Indirect reduction of MCL1 by artemisinins might explain their broad antineoplastic activity and their synergy with inhibitors of BCL2 and BCLxL. Furthermore, SOR has been reported to indirectly reduce MCL1 protein stability by increasing its ubiquitination via GSK3β activation,⁸⁹  consistent with our finding that SOR reduced the phosphorylated (inactive) form of GSK3β. Because SOR caused only minor elevations of DDIT3/CHOP in acute leukemia cells, ART838 and SOR appear to reduce MCL1 protein predominantly by distinct mechanisms; their combined targeting of MCL1 by discrete pathways may explain their antileukemic synergy.

The SAV combination reduced growth and enhanced apoptotic cell death of all tested leukemia cell lines (3 AMLs, 1 ALL) and primary patient samples (5 AMLs, 1 ALL) in vitro, although the extent of drug synergy may have been obscured in these assays by the high efficacy of 1 to 2 of the 3 drugs or by the high spontaneous ex vivo death rates of some of the primary patient AML cases. For these and other reasons, in vivo evaluations may better model clinical trials. For example, despite their in vitro activity as single drugs in this and previous studies, none of these monotherapies substantially reduced AML burdens in our experiments using 4 AML xenograft/primagraft models. Perhaps we would have detected greater in vivo responses had we used an AML progression model in which the xenograft/primagraft-bearing mice were treated before they had large AML burdens; thereby, we and others have documented the single-agent activity of artemisinin analogs.^18,25  In this project, we chose instead to use more challenging in vivo models in an attempt to reveal the potential clinical impact of the drug combinations. In contrast to single drugs, SAV was significantly active against all 4 AML xenograft/primagraft models, causing deep responses in the MOLM14 and MV4;11 AML xenograft models at day 7, which persisted for months during and after treatment completion, without substantial reductions in mouse complete blood counts or body weights. Several mice had undetectable leukemia-based luminescence at the final imaging done in these experiments. Against MOLM14 and MV4;11 AML xenografts, VEN plus SOR was approximately as active as SAV, and much more active than the other 2-drug combinations. The potency of VEN plus SOR may have obscured recognition of the contribution of ART838 in SAV using these 2 xenograft models.

Despite their in vitro sensitivity, the 2 AML primagraft models, both derived from older adults with relapsed, refractory AMLs, were much less sensitive to all tested drug combinations than the MOLM14 and MV4;11 AML xenografts (both derived from young patients harboring FLT3-ITD and MLLr). Notably, SAV was the most active combination against the AML45 primagraft, reducing day 7 AML burdens below day 0 levels, and the ART838 plus VEN combination was the next most effective combination. SAV was also the most active combination against the AML1526 primagraft model, although it only slowed leukemia growth. Thus, these in vivo results in the 2 AML primagraft models confirmed the observed in vitro synergism among the SAV drugs and the contribution of ART838 to the SAV combination.

In the future, newer kinase inhibitors (eg, gilteritinib) might be investigated for combination with ART838 and VEN, especially for cases harboring FLT3 kinase domain mutations (not inhibited by SOR).^90,91  Combination of ART838 and VEN with hypomethylating agents might also be investigated because combinations of VEN with azacytidine or decitabine are clinically active and tolerable. Although cardiotoxicity has been observed in clinical trials of MCL1-competitive inhibitors (eg, NCT03465540),⁹²  CDK9 inhibitors (eg, voruciclib, alvocidib, or AZD4573) might synergize with SAV to further reduce MCL1 because CDK9 inhibition reduces MCL1 transcription, whereas ART838 and SOR each reduce MCL1 via posttranscriptional mechanisms.^93-95  Finally, AMLs might resist SAV based on high levels of BCLxL,⁹⁶  and therefore selective BCLxL inhibitors or dual BCL2/BCLxL inhibitors (eg, NAV) might be evaluated for use in combination with SAV, although their clinical use would require close attention to platelet counts.^97,98 

In summary, our findings, especially the impressive outcomes of SAV in the MLLr⁺FLT3⁺ MOLM14 and MV4;11 AML xenograft models and the cooperativity of the SAV drugs in the AML primagraft models, demonstrate that ART838 and the SAV combination warrant consideration for clinical development for acute leukemias. In parallel, we suggest preclinical evaluation of additional acute leukemia cell line xenografts and primagrafts, especially those containing FLT3-ITD and/or MLLr driver mutations, as well as exploration of the herein-described modifications of the SAV regimen.

The authors thank the following students who participated in this research during their medical education at the University of Maryland School of Medicine: Kalyani Kumar, Taylor Rosenbaum, Allison Durham, Sang Tran, and Adam Ketchum. The luc/YFP-expressing MV4,11 AML cell line⁷²  was the kind gift of Sharyn Baker and the St. Jude Children’s Research Hospital Viral Vector Core (Memphis, TN). Unlabeled primary AML45 and AML1526 cells were kindly provided by Martin Carroll and Alexander Perl (University of Pennsylvania, Philadelphia, PA).

This work was supported in part by awards from the Emmert Hobbs Foundation Endowment (B.S.M.); the National Foundation for Cancer Research (C.I.C.); the Maryland Industrial Partnership (#MIPS-6502 and #MIPS-1801) (C.I.C.); the Maryland Innovation Initiative (#2018-MII-4916) (C.I.C.), the Leukemia & Lymphoma Society (TRP-10335016) (C.I.C.); the National Institutes of Health: T32CA154274 (B.S.M.), P30CA134274 (C.I.C., A.E., S.M.B., R.G.L., T.J.K., and M.K.), and P30CA006973 (M.A.R.) from the National Cancer Institute, UL1TR001079 (M.A.R.) from the National Center For Advancing Translational Sciences, S10RR026824 (M.A.R.) from the National Center for Research Resources, and R01CA183947 (J.W.T.), U01CA217862 (J.W.T.), and U54CA224019 (J.W.T.) from the National Cancer Institute; the V Foundation for Cancer Research (J.W.T.); Gabrielle’s Angel Foundation for Cancer Research (J.W.T.); the Silver Family Foundation (J.W.T.); and the Mark Foundation for Cancer Research (J.W.T.).

The contents are solely the responsibility of the authors and do not necessarily represent the official view of the funding sources.

Contribution: C.I.C. supervised all areas of the study; C.I.C., B.S.M., and M.A.R. designed experiments and developed methodology; C.I.C., B.S.M., T.J.K., and M.A.R. analyzed and interpreted data; B.S.M., S.M., and J.M.F. performed in vitro studies with cell lines and primary cells; B.S.M. carried out in vivo experiments under an approved institutional animal care and use committee protocol; M.A.R. supervised and analyzed stability and pharmacokinetics experiments; B.S.M., C.I.C., M.K., and T.J.K. wrote, reviewed, and revised the first and subsequent drafts of the manuscript; A.E. and R.G.L. provided the luciferase-labeled primary leukemia cells; J.W.T. provided the library of targeted antineoplastics; B.T.M. guided ART838 synthesis; S.M.B. and B.S.M. performed statistical analyses; and all authors reviewed and critically edited the manuscript.

Conflict-of-interest disclosure: J.W.T. has received research support from Agios, Aptose, Array, AstraZeneca, Constellation, Genentech, Gilead, Incyte, Janssen, Petra, Seattle Genetics, Syros, and Takeda. J.M.F. is a current employee of AstraZeneca and owns stock in the company. A.E. has received research grants from Jazz Pharmaceuticals, NewLink Genetics and Servier; is a global oncology advisory board member for Amgen; has served as an advisory board member for Genentech and Servier; and is a Co-Founder and Scientific Advisor for KinaRx, Inc. M.A.R. has received research grants to her institution from Celgene Corporation, Cullinan Apollo, and RenovoRx and institutional support for clinical trials conduct from Taiho; M.A.R.’s spouse is employed by GlaxoSmithKline. C.I.C. is chairman of the scientific advisory board for GPB Scientific, Inc. and owns stock in the company. The remaining authors declare no competing financial interests.

The current affiliation for B.S.M. is Jacobio (US) Pharmaceuticals Inc., Lexington, MA.

The current affiliation for S.M. is Arcellx Inc., Gaithersburg, MD.

The current affiliation for J.M.F. is AstraZeneca, Gaithersburg, MD.

Correspondence: Curt I. Civin, University of Maryland School of Medicine, BRB 14-021, 655 W Baltimore St, Baltimore, MD 21201; e-mail: ccivin@som.umaryland.edu.