Zelenirstat Inhibits N-Myristoyltransferases to Disrupt Src Family Kinase Signaling and Oxidative Phosphorylation, Killing Acute Myeloid Leukemia Cells

Acute myeloid leukemia (AML) is a hematologic malignancy characterized by heterogeneous molecular subtypes and difficulty achieving stable remission. Although AML represents 1% of new cancers, the 5-year survival rate in adults lags behind other cancer types at 31.7% (1). AML is typically a disease of the elderly with a median age at diagnosis of 69 years but is also the second most common pediatric leukemia. Drug treatments for AML are driven by molecular subtypes and clinical factors including age and comorbidities. Although induction therapies (2) frequently achieve complete remission, most patients relapse due to the persistence of leukemia stem cells (LSC; refs. 3, 4). Thus, there is an important unmet need for new druggable targets focusing on LSCs and AML subtypes with poor prognosis.

N-myristoylation is a posttranslational modification that links the 14-carbon fatty acid myristate to the N-terminal glycine residue of a protein and is catalyzed by two N-myristoyltransferases (NMT; NMT1 and NMT2). Numerous oncogenes encode myristoylated proteins, such as Src family kinases (SFK; refs. 5, 6) and c-ABL (7), suggesting that NMTs are druggable cancer targets. Myristoylation promotes membrane association (8), organelle targeting, and interactions with other proteins and complexes (9) and enhances protein stability (10). More than 200 human proteins are myristoylated (11), affecting diverse cellular pathways including proliferation signaling (12), survival, apoptosis (13), and mitochondrial function, including oxidative phosphorylation (OXPHOS; ref. 14). As the glycine-specific N-degron pathway rapidly degrades proteins with an exposed N-terminal glycine residue (10), NMT inhibitors (NMTi) promote the removal of unmyristoylated proto-oncogenic proteins. Zelenirstat (also known as PCLX-001 and DDD86481) is a potent small-molecule pan-NMTi that has recently completed phase I evaluation for the treatment of relapsed/refractory B-cell lymphomas and advanced solid tumors (15–17) and is now in phase IIa studies. Zelenirstat and one of its analogs have been independently validated as specific, on-target inhibitors of NMT1 and NMT2 (18, 19). Furthermore, asciminib (Scemblix), which binds to the myristoylation binding pocket of BCR-ABL tyrosine kinase to inhibit its catalytic function, has been approved for the treatment of chronic myeloid leukemia (20, 21). These studies have validated myristoylation as a therapeutic target in cancer.

We have previously reported variations in NMT expression levels in numerous cancer cell lines (19) and noted the prevalent loss of NMT2 in hematologic cancers. We also observed a correlation between NMT2 expression and dependence on NMT1, and recently developed a 54-gene myristoylation inhibitor sensitivity signature, MISS-54, to predict the sensitivity of untested tumors (14). Herein, we demonstrated the variability and prognostic value of NMT expression and MISS-54 scores in human AML. Zelenirstat effectively inhibited myristoylation in AML cells, promoting the degradation of SFKs, leading to apoptosis in vitro. In vivo, zelenirstat was well tolerated and reduced leukemic burden in multiple AML cell line–derived and patient-derived xenograft (PDX) models. Importantly, LSC-enriched populations were particularly vulnerable to zelenirstat. Our data suggest that this sensitivity could be the result of OXPHOS disruption by zelenirstat. Taken together, our findings support the clinical evaluation of zelenirstat as a novel therapeutic agent for AML treatment.

Zelenirstat was provided for this study by Pacylex Pharmaceuticals Inc. The structure of zelenirstat is described in Table 1 of US Patent#9828346 B2 (15), in which zelenirstat is listed as DDD86481. “EXAMPLE DDD86481” describes the chemical synthesis of zelenirstat.

CCLE(23Q2) and The Cancer Genome Atlas (TCGA) RNA sequencing data were obtained using DepMap and TCGA Biolinks, respectively. Clinical and mutational data for patients in TCGA-LAML were obtained using cBioPortal. GSE37642 data were obtained from the Gene Expression Omnibus website (https://www.ncbi.nlm.nih.gov/geo/). Data processing was performed using R v4.2.2 and RStudio (RRID: SCR_000432).

Authenticated MV-4-11 (RRID: CVCL_0064), KG-1 (RRID: CVCL_0374), U937 (RRID: CVCL_0007), KG-1a (RRID: CVCL_1824), HEL 92.1.7 (RRID:CVCL_2481), and HL 60 (RRID: CVCL_0002) cell lines were purchased from ATCC and DSMZ. The cells were maintained at 37°C and 5% CO₂ in a humidified incubator in RPMI-1640 medium (Gibco) supplemented with 10% FBS and 1 U penicillin/streptomycin (Gibco). Cell lines were cultured for a maximum of 2 months prior to experimental usage. Cells were validated as Mycoplasma free prior to usage.

Peripheral blood mononuclear cells (PBMC) were isolated from the blood of patients with AML with informed, written consent (Health Research Ethics Board of Alberta #26117) using density gradient centrifugation. CD34⁺ blast cells were isolated from samples using the EasySep CD34 selection kit (STEMCELL Technologies #18056), and lymphocytes were isolated from whole blood using the EasySep lymphocyte isolation kit (STEMCELL Technologies #19655).

Cells were stained with anti-CD45 (BD Biosciences Clone HI30), anti-NMT1 (Proteintech #11546-1-AP), and anti-NMT2 (BD Biosciences #611310 Clone 30) antibodies or an isotype control (BD Biosciences Clone MOPC-21) using a BD Fortessa Cell Analyzer (BD Biosciences). Samples were gated for live singlets, and blasts were identified via CD45/SSC gating on FlowJoV10.8.0. The levels of NMT1 or NMT2 were expressed as the mean fluorescence intensity ratio to the isotype control.

MV-4-11 and KG-1 cell lines, cells from patients with AML, providing informed, written consent, and healthy PBMCs and lymphocytes were incubated with ω-alkynyl myristate for 1 hour, as previously described (22). Cells were lysed, then “clicked” with azido-biotin, separated via SDS-PAGE, and then transferred to polyvinylidene difluoride (PVDF) membranes. Membranes were probed with a neutravidin–horseradish peroxidase conjugate and visualized using SuperSignal West Pico PLUS Chemiluminescent Substrate (Thermo Fisher Scientific #34580).

Cells were incubated with zelenirstat for 96 hours, and cell viability was determined using the CellTiter-Blue assay (Promega #G8080). Apoptosis induction was measured by annexin V and propidium iodide staining and analyzed by flow cytometry using a BD Fortessa Cell Analyzer (BD Biosciences) and FlowJo10.8.0 (RRID: SCR_008520).

OCI-AML22 cells were cultured in cytokine-rich media as previously described (23). Cells were treated with vehicle control or zelenirstat at the indicated concentrations for 48 or 72 hours, and the immunophenotype and viability were determined by flow cytometry using a BD Symphony A1 Flow Analyzer (BD Biosciences) and FlowJoV10.8.0 (RRID: SCR_008520).

Collection and analysis of intracellular proteins by immunoblotting were performed as previously described (19).

The experiments were performed with approval from the University of Alberta ACUC and Pharmaron ICUC. Flank xenografts were performed by Pharmaron. Six-to eight-week old female NOD/SCID mice (Pharmaron) received 10m MV-4-11 cells via flank SubQ injection. Once tumors reached 100 to 125 mm³, mice were randomized into treatment groups: vehicle, 10, 20, and 50 mg/kg zelenirstat daily. The treatment was administered once daily for 5 consecutive days. The tumor size was measured using a caliper.

Experiments were performed with the approval of the University of Alberta ACUC and Champions Oncology ICUC. Orthotopic xenografts were performed by Champions Oncology. Sublethally irradiated female NOG-EXL mice (RRID: IMSR_TAC:13395) received 2 × 10⁶ CTG-3439 AML cells via lateral tail vein injections. Once peripheral blood blasts reached 10 to 20 000 blasts/μL, surrogate animals were sacrificed and hCD45⁺ chimerism was analyzed to confirm engraftment. After randomization to treatment groups, mice received vehicle, 20, 35, or 50 mg/kg zelenirstat daily via oral gavage following a 5-day on and 3-day off regimen, with dosing holidays where necessary to mitigate weight loss. In-life bleeds, terminal blood, and bone marrow samples were collected, and immunophenotypes were determined via flow cytometry.

All samples were collected with informed, written consent according to procedures approved by the University Health Network Research Ethics Board. Primary AML samples were obtained from the Princess Margaret Hospital Leukaemia Bank and magnetically depleted of CD3⁺ cells prior to transplantation. Eight- to ten-week-old female NOD.Cg-Prkdc^scid Il2rg^tm1Wjl/SzJ mice (RRID: BCBC_4611) received 225cGy irradiation. Mice were intrafemorally injected with 4 million patient cells from patients with AML in 15 μL PBS. Three weeks later, the mice were randomized into groups and received zelenirstat via oral gavage at 50 mg/kg 3× weekly, supplemented with oral rehydration solution. After 4 weeks of treatment, the mice were sacrificed, and the bone marrow was flushed for flow cytometry analysis.

MV-4-11 and U937 cells were pretreated with zelenirstat for 48 hours. Three hours prior to analysis, cells underwent viability assessment using trypan blue (Thermo Fisher Scientific), and 50,000 live cells per well were plated. Metabolic function was analyzed with a Seahorse XFe96 Analyzer (RRID: SCR_019545) using a Glycolytic Rate Assay Kit (Agilent Technologies, #103344-100) and Seahorse XF RPMI media (Agilent Technologies, #103681-100) according to the manufacturer’s instructions. Each replicate of oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) values were individually normalized prior to statistical comparison by two-way ANOVA.

MV-4-11, U937, and KG-1 cells were pretreated with zelenirstat for 48 hours. ATP was released using Somatic Cell ATP Releasing Reagent (Sigma-Aldrich, #FLSAR) and quantified using an ATP Determination Kit (Invitrogen, #A22066) according to the manufacturer’s instructions. Luminescence values were normalized as percentages of untreated controls.

Oxygen consumption of MV-4-11 cells was measured using a Resipher cell culture monitor (Lucid Scientific). A total of 25, 000 cells/well were plated, and OCR was monitored in real time using Resipher sensors. After 24 hours of baseline data collection, 0, 0.1, 0.5, or 1 μmol/L of zelenirstat was added to the wells. “Peak OCR” represents the highest average OCR achieved by each group in 120 hours.

Following zelenirstat pretreatment, mitochondria were isolated from MV-4-11 and U937 cells as previously described (24). The mitochondria were then solubilized at a protein concentration of 5 mg/mL. Proteins were separated using blue native PAGE, transferred to nitrocellulose membranes, and probed. Membranes were developed and visualized as previously described.

In-gel activity assays were performed using blue-native PAGE gels. Gels were incubated with nitroblue tetrazolium buffer, as previously described (25), followed by fixation in 50% methanol and overnight de-staining. The bands were visualized using an ImageQuant 800 (GE Healthcare) and quantified using ImageStudioLite.

Complex I activity was measured in MV-4-11 cells pretreated with zelenirstat for 48 hours using an Abcam Complex I Activity Assay according to the manufacturer’s instructions. Absorbance was measured using a Cytation5 plate reader and Gen5 software.

Statistical analyses were performed using GraphPad Prism 10.1.2 (RRID: SCR_002798). Data represent the average of three or more biological replicates and are expressed as mean ± SEM unless stated otherwise.

The dataset used in the calculation of MISS-54 is publicly available through the GDC data portal (https://portal.gdc.cancer.gov/) and cBioPortal (https://www.cbioportal.org/). All other data generated in this study are available upon request to the corresponding author.

To evaluate the therapeutic potential of targeting myristoylation in AML, we investigated the expression levels of NMT1 and NMT2. AML cell lines universally expressed NMT1, whereas NMT2 expression was highly variable, with two of seven having nondetectable NMT2 (Fig. 1A and B), as assessed by Western blotting. Likewise, although all samples from patients with AML contained NMT1, NMT2 was low or undetectable in six of seven samples, and 18 of 39 had MFI ratios below 1 (Fig. 1C and D). Notably, monocytes had no detectable NMT2. Analysis of ∼1,400 CCLE cell lines (26) and ∼11,000 TCGA patient samples (27) also supported these observations (Supplementary Fig. S1). To test the effects of zelenirstat on NMTs in AML, AML cell lines (Fig. 1E; Supplementary Fig. S2A), patient cells (Fig. 1F), healthy PBMCs (Fig. 1G), and lymphocytes (Supplementary Fig. S2B) were incubated with zelenirstat for 1 hour prior to labeling with alkynyl-myristate and the detection of myristoylated proteins by click chemistry (22). Zelenirstat strongly inhibited myristoylation in both AML cell lines and patient cells in a dose-dependent manner but had less impact on healthy PBMCs and lymphocytes.

We next explored the potential prognostic significance of NMTs in AML using the TCGA-LAML project (28), looking at the correlations between NMT expression and overall survival (OS) of patients. Patients were separated into quartiles depending on the expression of NMT1 or NMT2, and survival data were analyzed using log-rank tests. Although NMT1 expression was not significantly associated with OS (Fig. 2A; P = 0.3559; HR = 1.346), NMT2 expression was highly associated with OS (Fig. 2B; P ≤ 0.0001; HR = 0.2945). Patients with low NMT2 levels had a median OS of 7.8 months, whereas patients with high NMT2 levels did not reach the median OS within the timeframe of the study. We validated this finding in the GSE37642 dataset, in which a significant correlation between NMT2 expression and survival was also found (Supplementary Fig. S3). To further elucidate the role of NMTs in AML, we compared NMT1 and NMT2 expression with commonly used prognostic AML mutations FLT3, NPM1, DNMT3A, IDH1, IDH2, RUNX1, and TP53. Patients were categorized into wild-type (WT), driver mutant, or other mutant groups based on genomic data (29). No significant difference was observed in NMT1 expression, regardless of the mutation status (Supplementary Fig. S4A). NMT2 expression, however, was significantly lower in patients with driver NPM1 mutations than in WT and was significantly higher in patients with driver IDH1 mutations than in WT (Fig. 2C and D). No other significant associations were observed (Supplementary Fig. S4B). Additionally, we did not find any significant correlation between NMT1 or NMT2 and ELN2022 risk groups or TCGA-LAML molecular risk (Supplementary Fig. S4C; ref. 28). Despite the lack of an association with other common prognostic factors, our analysis showed a clear and significant prognostic value of NMT2 expression.

MISS-54 is a gene signature developed to predict the sensitivity of cells to NMTis, including zelenirstat, and high MISS-54 scores indicate a higher potential sensitivity to NMTis (14). We explored the relationship between MISS-54 scores and the clinical characteristics of patient in TCGA-LAML. Mean MISS-54 scores for patients in the TCGA-LAML cohort were 6.62 compared with 5.51 in all other cancers, indicating higher NMTi sensitivity in AML (Fig. 2E; range: 2.5–7.5). MISS-54 was significantly associated with OS; a comparison of patients in the highest and lowest quartiles of MISS-54 scores showed that a high score was associated with poor survival (Fig. 2F; median survival: 21 vs. 56.3 months; P = 0.0226; HR = 1.949). Recently, a cellular hierarchy framework was developed to provide insight into disease heterogeneity and predict drug response in patients with AML (30). MISS-54 scores were significantly higher in patients with primitive or intermediate hierarchy subtypes (associated with worse outcomes) than in those with GMP-dominant hierarchies (associated with better outcomes; Fig. 2G). No association was found between the MISS-54 scores and ELN2022 risk classification or TCGA-LAML molecular risk groups (Supplementary Fig. S5). Overall, MISS-54 scores in AML are among the highest in any cancer (14) and are associated with poor OS and prognosticaly adverse hierarchies, suggesting that NMTi may be an effective therapeutic approach, particularly in high-risk patients.

Next, we performed viability assays using AML cell lines (Fig. 3A), patient cells, PBMCs, and lymphocytes obtained from healthy individuals (Fig. 3B). The cells were treated with zelenirstat for 96 hours, after which viability was measured. AML cell lines and primary patient cells underwent >70% viability loss, with EC₅₀ values between 0.1 and 1 μmol/L. In contrast, healthy lymphocytes and PBMCs exhibited only 20% and 40% viability losses respectively, even at 10 μmol/L.

To evaluate the efficacy of zelenirstat against AML cells in vivo, we used an ectopic MV-4-11 cell line–derived xenograft model (Fig. 3C). Control mice and those treated with 10 mg/kg zelenirstat exhibited rapid tumor growth. Treatment with 20 mg/kg zelenirstat exerted a tumorostatic effect in the treated mice. Treatment with 50 mg/kg zelenirstat resulted in complete tumor regression, with no discernible tumor reappearing by 28 days despite cessation of treatment. The observed toxicity was limited to reversible body weight loss in mice treated with 50 mg/kg zelenirstat (Supplementary Fig. S6A).

Next, we used an orthotopic AML model in which patient-derived AML cell CTG-3439 were xenografted via tail vein injection (Fig. 3D). Zelenirstat treatment reduced the hCD45⁺ leukemic burden in a concentration-dependent manner, which reached significance at day 42 compared with the vehicle control (P = 0.010, 0.012, and 0.008 for 20, 35, and 50 mg/kg, respectively). Pooled terminal blood and bone marrow analyses revealed near-complete eradication of hCD45⁺ cells at 50 mpk. Responses were observed as early as day 16 for all doses, and toxicity was again limited to reversible weight loss at 50 mg/kg (Supplementary Fig. S6B).

To evaluate the effects of zelenirstat on primary AML cells, we generated PDX models from four patient samples through intrafemoral injection. Three weeks posttransplantation, mice were randomized to receive either 50 mg/kg zelenirstat or vehicle control three times weekly for 4 weeks. Zelenirstat treatment significantly reduced the human (hCD45⁺) leukemic burden in three of the four AML PDX models compared with controls (Fig. 3E). Collectively, these data demonstrated that zelenirstat potently targeted AML cells in vitro and in vivo.

The persistence of LSCs after standard chemotherapy, leading to relapse, remains a major challenge in AML treatment. To assess the effect of zelenirstat on LSCs, we used OCI-AML22, a hierarchical AML patient-derived cell model that retains a well-characterized, functionally validated LSC population within the CD34⁺CD38⁻ fraction that recapitulates the transcriptional, epigenetic, and functional properties of LSCs from primary patient samples (23, 31). OCI-AML22 cells were treated with zelenirstat for 48 (Fig. 4A) or 72 hours (Fig. 4B), followed by flow cytometric analysis. Although loss of viability was observed in all cell fractions after 72 hours, with an IC₅₀ of 479 nmol/L, only the LSC-containing CD34⁺CD38⁻ fraction showed a significant loss of viability at 48 hours, with near-total viability loss by 72 hours with an IC₅₀ of 191 nmol/L. These data demonstrated the selective activity of zelenirstat against disease-driving LSC-enriched populations.

Next, we explored the mechanisms by which zelenirstat targeted AML cells. We previously demonstrated that zelenirstat strongly disrupted B-cell receptor signaling by promoting the degradation of numerous SFKs (19). Here, we explored whether zelenirstat causes loss of SFK signaling, which is critical to FLT3 and cKIT survival pathways in AML. We selected three cell lines: MV-4-11 with nondetectable NMT2 expression and FLT3-ITD mutation, KG1 with high NMT2 expression and WT FLT3, and U937 with moderate NMT2 expression and null-FLT3. We incubated these cell lines with zelenirstat for 48 hours and then co-stimulated them with FLT3 ligand and cKIT ligand stem cell factor. Zelenirstat promoted similar degradation of five SFKs (Supplementary Fig. S7). As HCK was the most readily detectable SFK subsequent analyses were carried out using HCK as a representative SFK. Western blot analysis showed that zelenirstat induced a significant dose-dependent loss of total HCK, phosphorylated SFKs, and phosphorylated Stat5 protein levels (Fig. 5A; Supplementary Fig. S8). Next, the cells were treated with zelenirstat for 48 to 96 hours to investigate the impacts on endoplasmic reticulum (ER) stress and apoptosis using BiP and cleaved caspase 3, respectively, as markers. We observed increases in ER stress after 48 hours of treatment with 0.1 μmol/L zelenirstat and induction of apoptosis after 48 hours of 1 μmol/L treatment (Fig. 5B and C; Supplementary Fig. S9). Interestingly, U937 cells showed a similar induction of ER stress and apoptosis despite lacking the FLT3 receptor, suggesting that additional mechanisms beyond FLT3 signaling are at play. To determine whether viability loss was the result of apoptosis induction, samples from patients with AML and healthy PBMCs were treated with zelenirstat for 96 hours, stained with propidium iodide and annexin V, and analyzed by flow cytometry (Fig. 5D). We observed a significant increase in annexin V staining, indicative of apoptosis with zelenirstat treatment, with increases in propidium iodide staining and cell death at higher concentrations. Again, we observed up to a 90% viability loss in AML cells, with only a 30% viability loss in healthy PBMC populations. Together, these findings demonstrate that zelenirstat treatment disrupts SFK signaling, resulting in cellular stress and a stronger induction of apoptosis in AML cells than in healthy cells.

Our recent proteomics analyses revealed that numerous mitochondrial proteins were downregulated in response to CRISPR/Cas9 NMT1 KO or zelenirstat treatment (14). The most downregulated protein was NDUFAF4, a myristoylated mitochondrial protein responsible for complex I assembly. Therefore, we explored the effects of zelenirstat on the oxidative metabolism of AML cells. For this purpose, we measured the OCR and ECAR in MV-4-11 and U937 cells treated with zelenirstat using the Agilent Seahorse glycolytic rate kit (Fig. 6A and B; Supplementary Figs. S10 and S11). We observed a significant, dose-dependent reduction in oxygen consumption upon treatment of MV-4-11 cells with zelenirstat, which was largely abolished by the addition of rotenone or antimycin A (Fig. 6B), suggesting significant impairment in OXPHOS. Similar results were obtained in the highly glycolytic U937 cell line (Supplementary Fig. S10). Interestingly, we also observed a dose-dependent reduction in the ECAR, suggesting a similar suppressive effect on the glycolytic rates of these cells (Supplementary Fig. S11). Zelenirstat treatment significantly reduced cellular ATP content in MV-4-11, KG-11, and U937 cell lines (Fig. 6C). We also used the Resipher platform to monitor oxygen consumption over time in MV-4-11 cells incubated with zelenirstat. We observed a dose-dependent reduction in peak oxygen consumption without a significant loss of viability in this timeframe (Fig. 6D; Supplementary Fig. S12). We isolated mitochondria from MV-4-11 and U937 cells treated with zelenirstat to evaluate the impact on complex I assembly and activity. Using in-gel activity assays, we observed a significant reduction in complex I activity in mitochondria isolated from cells treated with zelenirstat compared with that in control cells (Fig. 6E and F). This result was validated using a complex I activity assay, in which MV-4-11 cells treated with zelenirstat showed significant decrease in ΔA450, indicating impaired complex I activity (Fig. 6I). Further Western blotting analyses revealed the loss of NDUFB11, a non-myristoylated complex I protein, in blue-native gels (Fig. 6E and G), and loss of NDUFAF4 in SDS-PAGE (Fig. 6E and H) using mitochondria isolated from cells treated with zelenirstat. Together, these data indicate that zelenirstat treatment promotes the loss of myristoylated mitochondrial complex I assembly factor NDUFAF4, complex I assembly and activity, OXPHOS, and respiration in AML cells.

In this study, we explored the clinical relevance of NMTs in AML and the pharmacologic effects of the potent first-in-class NMTi zelenirstat on the survival of AML cells, including LSCs, in vitro and in vivo. Although the expression of NMTs was found to be highly variable in AML cell lines and patient samples, as seen in other cancers (14), only low NMT2 was significantly linked to unfavorable outcomes. The recently developed MISS-54 score (14), which predicts sensitivity to zelenirstat, was the highest in patients in TCGA-LAML with the shortest survival, suggesting that zelenirstat could address an unmet clinical need. Zelenirstat targeted AML cells via inhibition of FLT3 and c-KIT signaling, induction of ER stress, and apoptosis. Zelenirstat also disrupted complex I assembly and OXPHOS, which are processes essential for LSCs (32–34). Overall, the effects of zelenirstat were more potent in AML cells than those in healthy hematopoietic cells, suggesting a therapeutic index for AML similar to that seen in zelenirstat clinical trials in solid tumors and B-cell lymphoma (16). Indeed, in phase I studies, there has been no neutropenia or myeloid lineage suppression in patient receiving up to 280 mg daily, and thus no suggestion of hematopoietic stem cell toxicity with clinically achievable exposures of zelenirstat, even in a population of patients with extensive prior cytotoxic chemotherapy. Thus, zelenirstat could represent a novel therapeutic with the potential to provide significant benefits to patients with AML with the poorest outcomes with current therapies.

Zelenirstat effects were largely independent of the AML molecular subtype. FLT3 is among the most clinically relevant genes in AML, is the single most common mutation, and is significantly associated with poor survival. Similarly, cKIT is heavily mutated in AML and is associated with stem cell differentiation and proliferation (35, 36). As with the B-cell receptor (19), FLT3 and cKIT both require SFK myristoylation for their association with receptor complexes and signal amplification. Pretreatment of cells with zelenirstat, followed by stimulation with FLT3 and SCF ligands, resulted in a dramatic loss of SFKs and SFK phosphorylation, compromising the ability of cells to transduce these critical signals. The importance of FLT3 in AML has led to the development of targeted agents, such as quizartinib and gilteritinib, with efficacy largely restricted to AML subtypes dependent on FLT3 (2). Interestingly, zelenirstat also showed signaling effects in FLT3-deficient U937 cells, but this is not entirely unexpected as SFKs are critical to numerous growth and proliferation pathways, and their inhibition suppresses AML growth regardless of FLT3 status (37). We observed drastic increases in BiP and cleaved caspase 3 protein levels, indicating ER stress and the induction of apoptosis, irrespective of the FLT3 status of the cells.

LSCs must be successfully targeted to effectively treat AML and mitigate residual disease and relapse. One process that has been consistently identified as a vulnerability of LSCs is mitochondrial OXPHOS (32–34), with LSCs being functionally characterized by a low rate of energy metabolism and relative inability to compensate for the loss of OXPHOS with glycolysis. We previously reported that 36/108 downregulated proteins in HAP1 NMT1KO cells are mitochondrial, suggesting a key role for myristoylation in mitochondrial function (14). Of note, 20 of these 36 proteins belong to complex I, prompting the investigation of the impact of zelenirstat on cellular respiration. Herein, we demonstrate that zelenirstat causes the loss of the myristoylated complex I assembly factor NDUFAF4, leading to complex I misassembly and loss of oxygen consumption in AML cells. As OXPHOS is critical for LSC survival, this finding provides a potential mechanism by which zelenirstat selectively targets LSC-enriched populations, as demonstrated in OCI-AML22 cells. Typically, cells experiencing OXPHOS disruptions will upregulate glycolysis to compensate for deficient ATP production. We did not observe this; rather, we observed a decrease in glycolytic rate alongside OXPHOS. We believe this absence of a glycolytic switch to be due to the loss of AMPKβ (Supplementary Fig. S13), a myristoylated scaffolding subunit of the energy sensing AMPK (38, 39). The absence of AMPKβ could lead to the failure to upregulate glycolysis in response to ATP depletion, but the effects of zelenirstat on glycolysis require additional exploration.

Overall, our data provide a proof of concept for the clinical development of zelenirstat in AML, and MISS-54 scores suggest broad activity across various molecular subtypes and risk groups, with the greatest benefit expected in those currently experiencing the worst outcomes. Zelenirstat acts as a dual-action agent that inhibits cell signaling and OXPHOS, allowing it to effectively target both AML blasts (40) and LSCs. In addition, the novel mechanisms of action of zelenirstat present unexplored potential in combination with currently approved therapies. Initial investigations suggested striking synergy with venetoclax in vitro (Supplementary Fig. S14). This report demonstrates the combined computational and experimental validation of zelenirstat and its potential to meet an important unmet clinical need. Zelenirstat has already been evaluated in a phase I clinical trial and demonstrated favorable pharmacokinetics, safety, and efficacy in patients with relapsed/refractory lymphoma and advanced solid malignancies (16, 17). Our preclinical data strongly support the development and clinical evaluation of zelenirstat as a novel therapeutic agent for AML.

J.M. Gamma reports grants from Alberta Cancer Foundation, Leukemia & Lymphoma Society of Canada, Award in Hematological Cancers’ Research in Memory of Dr. Rachel Mandel, Alberta Paving Ltd., Dr. Heleen & Rod McLeod, Eusera, and Pacylex Pharmaceuticals during the conduct of the study and personal fees from Pacylex Pharmaceuticals outside the submitted work, as well as a pending US patent application #63/573,885 assigned to Pacylex Pharmaceuticals, and is a current shareholder of Pacylex Pharmaceuticals. Q. Liu reports personal fees from Revolution Medicines outside the submitted work. E. Beauchamp reports personal fees from Pacylex Pharmaceuticals during the conduct of the study and personal fees from Pacylex Pharmaceuticals outside the submitted work, as well as a pending US patent application #63/573,885 assigned to Pacylex Pharmaceuticals, and patents #11,135,218 and #11,788,145 issued to Pacylex Pharmaceuticals Inc. M.C. Yap reports grants and personal fees from University of Alberta during the conduct of the study, as well as patents #11,135,218 and #11,788,145 issued to Pacylex Pharmaceuticals Inc. C. Ekstrom reports grants from Alberta Cancer Foundation and Leukemia & Lymphoma Society of Canada during the conduct of the study. R. Pain reports grants from Leukemia & Lymphoma Society of Canada, Alberta Cancer Foundation, Award in Hematological Cancers’ Research in Memory of Dr. Rachel Mandel, Alberta Paving Ltd., Dr. Heleen and Rod McLeod, Eusera, and Pacylex Pharmaceuticals Inc. during the conduct of the study, as well as a convertible note for investment into Pacylex. M.A. Kostiuk reports grants from Dr. Heleen and Rod McLeod through the Cancer Research Institute of Northern Alberta during the conduct of the study, as well as other support from Pacylex Pharmaceuticals Inc. outside the submitted work. J.R. Mackey reports other support from Pacylex Pharmaceuticals during the conduct of the study, as well as other support from illumiSonics outside the submitted work. J. Brandwein reports personal fees from AbbVie, Astellas, Bristol Myers and Squibb, Pfizer, Servier, Jazz, Taiho, and Amgen outside the submitted work. J.C.Y. Wang reports grants from University of Alberta during the conduct of the study, as well as other support from Pfizer and grants from Leukemia & Lymphoma Society of Canada outside the submitted work. L.G. Berthiaume reports personal fees and other support from Pacylex Pharmaceuticals Inc. (Pacylex) outside the submitted work, as well as pending US patent applications #63/573,885 and #63/093,970 assigned to Pacylex Pharmaceuticals Inc., and patents #11,135,218 and #11,788,145 issued to Pacylex Pharmaceuticals Inc. No disclosures were reported by the other authors.

J.M. Gamma: Conceptualization, data curation, software, formal analysis, supervision, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. Q. Liu: Conceptualization, formal analysis, investigation, visualization, methodology, writing–review and editing. E. Beauchamp: Conceptualization, resources, data curation, formal analysis, supervision, validation, investigation, visualization, methodology, project administration, writing–review and editing. A. Iyer: Investigation, visualization. M.C. Yap: Conceptualization, resources, formal analysis, supervision, validation, investigation, visualization, methodology, project administration. Z. Zak: Investigation, visualization. C. Ekstrom: Investigation, visualization. R. Pain: Formal analysis, investigation, visualization. M.A. Kostiuk: Formal analysis, validation, investigation, visualization, methodology. J.R. Mackey: Conceptualization, funding acquisition, methodology, writing–review and editing. J. Brandwein: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, validation, methodology, project administration, writing–review and editing. J.C.Y. Wang: Conceptualization, resources, data curation, formal analysis, supervision, validation, investigation, visualization, methodology, project administration, writing–review and editing. L.G. Berthiaume: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, validation, methodology, project administration, writing–review and editing.

This work was supported by the Leukemia & Lymphoma Society of Canada, Alberta Cancer Foundation grants 26362 and 26927, the Award in Hematological Cancers’ Research in Memory of Dr. Rachel Mandel, an Alberta Paving Ltd. donation to L.G. Berthiaume, a generous donation from Dr. Heleen and Rod McLeod through the Cancer Research Institute of Northern Alberta, Eusera (www.eusera.com), and Pacylex Pharmaceuticals Inc. (www.pacylex.com).