Pharmacological Myristoylation Inhibition by Zelenirstat Disrupts Growth Signalling, Energy Metabolism, and Migration in Triple-Negative Breast Cancer Cells

Abstract

Background

Metastatic breast cancer cells rely on oxidative phosphorylation (OXPHOS) for energy production, yet selectively targeting mitochondrial metabolism in cancer cells remains a major therapeutic challenge. Myristoylation, the modification of proteins with the 14-carbon fatty acid myristate by N-myristoyltransferases NMT1 and NMT2, dually impacts cancer cell signalling and metabolism. The pan-NMT inhibitor zelenirstat selectively inhibits these enzymes to prevent the myristoylation of target proteins.

Methods

We assessed the effects of zelenirstat on the protein levels and metabolic activity of multiple triple-negative breast cancer (TNBC) cell lines. NDUFAF4 knockout (KO) MDA-MB-231 cells were engineered using CRISPR/Cas9. Functional assays included multiple types of respirometry, cell migration measurements, and spheroid formation.

Results

Zelenirstat treatment depleted myristoylated complex I assembly factor NDUFAF4 levels in four different TNBC cell lines. This loss effectively abrogated mitochondrial OXPHOS through impairment of complex I, resulting in reduced cellular energy production. Zelenirstat treatment resulted in NADH accumulation, loss of AMPK, and decreased glycolysis. Given that zelenirstat impacted cell signalling and OXPHOS, we further demonstrated that it also significantly impaired migration and invasion. Importantly, both zelenirstat treatment and NDUFAF4 KO TNBC MDA-MB-231 cells dramatically reduced mammosphere formation, indicating potent activity against breast cancer stem cells (BCSCs).

Conclusion

In addition to the previously shown inhibitory effect of zelenirstat on TNBC cell growth in vitro and in vivo, our new data demonstrate the widespread effects of zelenirstat on both aggressive TNBC cells and BCSCs. These findings highlight the potential of zelenirstat as an anti-metastatic agent, warranting further clinical trial evaluation in patients with TNBC.

Graphical Abstract

Introduction

Worldwide, over 2.3 million individuals are diagnosed with breast cancer every year.¹ Triple-negative breast cancer (TNBC) is a highly aggressive subset of breast cancer, affecting nearly 200,000 patients worldwide annually, and currently lacks effective treatment options.² This is due to the absence of estrogen, progesterone, and human epidermal growth factor receptors (ER, PR, and HER2, respectively), which serve as therapeutic targets in other subtypes of breast cancer.³ Metastasis of solid tumor cancers, such as TNBC, causes around 90% of cancer-related mortality.⁴ New cancer therapeutic approaches are currently in development, several of which focus on interfering with cancer metabolism in these cells.⁵

The Warburg effect is a hallmark of cancer that describes how cancer cells preferentially generate energy through aerobic glycolysis, yielding pyruvate that is converted to lactate rather than fully oxidized using the tricarboxylic acid (TCA) cycle and oxidative phosphorylation (OXPHOS) despite the availability of oxygen.⁶ While this is common across most primary cancer tumors, recent evidence suggests that cancer cells switch to an oxidative-driven metabolism in order to metastasize, as shown in TNBC.⁷ Others have used murine models to show that the expression of peroxisome proliferator-activated receptor gamma coactivator 1-alpha (PGC-1α), a master mitochondrial biogenesis regulator, is increased in both circulating and metastatic TNBC cancer cells.⁸ Additionally, the activation of Src, an oncogenic Src-family kinase, is linked to increased fatty acid oxidation (FAO) and OXPHOS to drive TNBC aggressiveness.⁹

While attempts to target mitochondrial respiration by inhibiting complex I showed some promise in cultured cancer cells, compounds such as IACS-010759 have been shown to be overly toxic in clinical trials.¹⁰ Therefore, the search for compounds that can safely reduce OXPHOS activity in patients is ongoing. One potential way to achieve this goal is to use N-myristoylation inhibitors (NMTis) since these were recently shown to dually inhibit cell growth signalling and OXPHOS in numerous cancer cell lines.^11–13 N-myristoylation is the modification of the N-terminal glycine residue of >200 different mammalian proteins by the 14-carbon fatty acid myristate.¹⁴ Myristoylation is catalyzed by two enzymes in humans, N-myristoyltransferases 1 and 2 (NMT1 and NMT2), which have distinct yet overlapping specificities.¹⁵ Importantly, when myristoylation is inhibited, proteins that fail to be myristoylated are degraded through a glycine-specific N-degron pathway.¹⁶ Proteomic analysis of NMT1 knockout (KO) in haploid HAP1 leukemia cells unexpectedly revealed that 36 mitochondrial proteins exhibited the most significant reduction compared to WT HAP1 cells, and of these, 18 were from mitochondrial complex I with the myristoylated mitochondrial complex I assembly factor NDUFAF4 showing the greatest decrease.¹¹ Notably, several mitochondrial electron transport system components are myristoylated.¹⁷ Interestingly, NMT1 knockdown results in reduced in vitro cell migration as well as in vivo tumor growth in TNBC.¹⁸ A first-in-class pan-NMT inhibitor, zelenirstat (PCLX-001, DDD86481), preferentially killed hematologic cancer cell lines in vitro¹⁹ and reduced the growth of solid tumors in murine xenografts, including breast cancer.²⁰ Zelenirstat treatment inhibited OXPHOS by decreasing complex I activity¹¹ and also reduced glycolytic metabolism in acute myeloid leukemia (AML) cells.¹² It was safely administered to cancer patients and demonstrated early signs of efficacy in a Phase I clinical trial for advanced refractory lymphoma and solid tumors and is currently being investigated in a Phase 1/2 trial for relapsed/refractory AML.^21–23 Other NMTis have shown preclinical activity in blood and solid tumor models.^13,24

We previously generated a myristoylation inhibition sensitivity signature composed of 54 genes, MISS-54.¹¹ While the MISS-54 score predicted that hematologic cancer cells were the most sensitive to myristoylation inhibition, it also revealed a potential for use in breast cancer.²⁵ In this study, we reveal that TNBC is predicted to be the breast cancer subtype most sensitive to zelenirstat treatment. We further unveil that zelenirstat treatment causes degradation of mitochondrial complex I in numerous TNBC cell lines, as well as a significant reduction in its activity, leading to a loss of OXPHOS. Zelenirstat reduced the levels of numerous myristoylated and non-myristoylated proteins associated with cancer cell migration, in vitro migration/invasion, and mammosphere formation. These results warrant further investigation into the potential of zelenirstat as an anti-metastatic agent in TNBC.

Materials and Methods

Antibodies and Materials

Rabbit anti-GAPDH (1:20000) and rabbit anti-actin (1:5000) were obtained from the laboratory stock and are available through Eusera (https://www.eusera.com). Rabbit monoclonal anti-Lyn (clone C13F9, #2796), rabbit anti-Src (clone 36D10, cat#2109S), rabbit anti-phospho-Src family (clone D49G4, cat#6943S), rabbit anti-Flotillin-2 (clone C32A3, cat#3436S), rabbit anti-pyruvate dehydrogenase (clone C54G1, cat#3205T), rabbit anti-PDHK1 (clone C47H1, cat#3820T), rabbit anti-phospho-pyruvate dehydrogenase α1 (Ser293) (cat#31866S), rabbit anti-phospho-pyruvate dehydrogenase α1 (Ser232) (cat#15289S), rabbit anti-AMPKα (cat#5831S), and rabbit anti-AMPKβ1/2 (clone 57C12, cat#4150S) were purchased from Cell Signalling Technology (Danvers, MA, USA). Rabbit anti-NDUF3 (cat#ab122270), rabbit anti-NDUFS8 (cat#ab170936), rabbit anti-NDUFB7 (cat# ab188575), rabbit anti-Mic19 (cat#ab240598), rabbit anti-VDAC1 (cat#ab154856), rabbit anti-NDUFB11 (cat#ab183716), rabbit anti-pan PDK (cat#ab115321), and mouse Total OXPHOS rodent cocktail (cat#Ab110413) were purchased from Abcam (Cambridge, MA, USA). Rabbit anti-NDUFAF4 (#A14345) was purchased from Abclonal (Woburn, MA, USA). Rabbit anti-FAM49B (cat#PA5-52647 and #sc-390478) was purchased from Thermo Fisher Scientific (Waltham, MA, USA) and Santa Cruz Biotechnology (Dallas, TX, USA), respectively. Mouse monoclonal NMT1 (6F8D5) (1:500) and NMT2 (6C5E8) (1:1000) were developed and validated in our laboratory.²⁰ SuperSignal West Pico PLUS Chemiluminescent substrate was purchased from Thermo Scientific (Rockford, IL, USA). IACS-010759 (cat#25867) was purchased from Cayman Chemical (Ann-Arbor, MI, USA). Zelenirstat was provided by Pacylex Pharmaceuticals (Edmonton, AB, Canada).

All chemicals were of the highest purity available and were purchased from Sigma-Aldrich unless otherwise indicated.

NMT Expression and MISS-54 Bioinformatic Analysis

Gene expression data were downloaded from The Cancer Genome Atlas (TCGA) database and categorized into breast cancer tissue subtype or normal breast tissue. MISS-54 score and clustering analysis was applied to samples as previously described.¹¹ Values were transferred into GraphPad Prism to analyze the data and perform statistical analysis.

Cell Culture

MDA-MB-231, MCF-7, COS-7, and MCF 10A cells were purchased from the ATCC. BT-549, MDA-MB-468, and MDA-MB-453 cells were acquired from the ATCC by Dr. Ing Swie Goping and provided as generous gifts using cells frozen from original ATCC vials. MDA-MB-231, COS-7, and MCF-7 cells were maintained in DMEM (Gibco) supplemented with 5–10% fetal bovine serum (Avantor or Gibco), 100 U/mL penicillin, 0.1 mg/mL streptomycin. BT-549, MDA-MB-468, and MDA-MB-453 cells were maintained in RPMI supplemented with 10% fetal bovine serum (Avantor), 100 U/mL penicillin, 0.1 mg/mL streptomycin. Cell lines of breast origin were obtained from female donors. MCF 10A cells were cultured in Lonza MEGM complete growth medium and passaged using 0.05% trypsin and soybean trypsin inhibitor (Sigma, cat#10109886001). All other adherent cell lines were passaged with 0.25% trypsin. All cell lines were maintained at 37°C and 5% CO₂ in a humidified incubator and tested negative for mycoplasma contamination using MycoStrip tests (Cat# rep-mysnc-50, InvivoGen).

Cell Lysis

Cells were harvested, washed in cold PBS, and lysed in 0.1% SDS-RIPA buffer (50 mM Tris-HCl (pH 8.0), 150 mM NaCl, 1% Igepal CA-630, 0.5% sodium deoxycholate, 2 mM MgCl₂, and 2 mM EDTA with 1x complete protease inhibitor; Roche Diagnostics, Basel, Switzerland) by rocking for 20 min at 4°C. The lysates were centrifuged at 21,000 × g for 10 min at 4°C and the post-nuclear supernatant was collected.

Immunoblotting

Protein concentrations were determined using the BCA assay (Thermo Fisher Scientific), according to the manufacturer’s instructions. Samples were prepared for electrophoresis by adding 5× loading buffer and boiling for 7 min. Unless stated otherwise, 20 μg of total protein per lane was loaded onto a 12.5% acrylamide gel. After electrophoresis, gels were transferred onto a 0.2 μm nitrocellulose membrane (Bio-Rad), probed with antibodies, developed using SuperSignal™ West Pico PLUS Chemiluminescent Substrate (ThermoScientific cat#34577), and visualized using ImageQuant 800 (Amersham). Uncropped blots are available to view in a supplementary file.

For total OXPHOS rodent blots, aliquots of each lysate were separated from the rest and heated to 50°C. Lysates were run as normal and transferred onto a 0.2 μm PVDF membrane. Antibody blotting was performed as described previously.

Generation of NDUFAF4 Knock-Out Cell Line

Guide RNAs targeting human NDUFAF4 were prepared using Alt-R CRISPR-Cas9 tracrRNA, ATTO-550 (IDT), and Alt-R CRISPR-Cas9 crRNA (IDT) as previously described.²⁶

EnGen Cas9 NLS and S. pyogenes (NEB) proteins were incubated with prepared gRNAs to form ribonucleoprotein complexes, which were subsequently transfected into MDA-MB-231 cells using Lipofectamine CRISPRMAX according to the manufacturer’s instructions. The next day, cells positive for ATTO-550 were sorted individually into wells of a 96-well plate on a FACSAria-III instrument (Flow Cytometry Core, University of Alberta) to establish monoclonal populations. MDA-MB-231 cells were validated for NDUFAF4 gene knockout using DNA sequencing (ABI 3730, Molecular Biology Facility, University of Alberta) and western blotting.

Zelenirstat Treatment, Cell Viability and Cell Counts

Cells were grown in 96-well plates and incubated with increasing concentrations of zelenirstat for 48 h. The viability or count of cells treated with zelenirstat was measured using the CellTiter-Blue Cell Viability Assay (Promega) or the CyQUANT Cell Proliferation Assay (Thermo Fisher), respectively. Assays were performed using a Cytation 5 cell imaging multimode reader (BioTek Instruments) and Gen5 Data Analysis software using nonlinear regression.

Wound Healing Assay

Cells were plated in 6-well plates and allowed to grow to confluence. Cell monolayers were scratched using a 200 µL pipet tip, washed with PBS, and grown in media containing 0.5–1% FBS with increasing concentrations of zelenirstat. Scratches were visualized using a Cytation 5 at 0, 24, and 48 h of treatment using bright-field microscopy. At 48 h, the medium was replaced with medium containing calcein viability dye, incubated for 20 min, and visualized using fluorescence in a Cytation 5. For the measurements, Gen5 was used to autogenerate a mask around the edge of the scratch. Three equally spaced measurements were taken between the masks and then averaged for each time point.

Transwell Migration/Invasion

Cells were grown in 10 cm dishes and treated with zelenirstat for 24 h. Cells were scraped and counted, and 50,000 cells were plated onto the top chamber of 24-well Corning Transwell Fluoroblok HTS plates. For the invasion assay, the top chamber wells were incubated with 750 µg/mL Matrigel before starting the assay. The top chamber medium contained no serum, whereas the bottom chamber medium contained 10% FBS as a chemoattractant. Invasion was measured by adding matrigel to the top transwell chamber. Each group was allowed to migrate or invade in the presence of zelenirstat for an additional 24 h. The top chamber was removed and placed in a 24-well plate containing calcein medium. Fluorescence was read using bottom-reading microscopy in the Cytation 5. Migrated/invaded cells were counted using Gen5.

Mammosphere Formation Assay

A total of 6,000 cells per well were plated in ultra-low adherent plates in DMEM/F12 media supplemented with B27, hFGF, and EGF, as previously described.²⁷ Cells were cultured for 7 days ± zelenirstat treatment. Mammospheres were manually counted using Fiji software (ImageJ).

Confocal Microscopy

For comparison of NDUFAF4 localization, COS-7 cells were split into small glass-bottom dishes, transfected with either EGFP or pyruvate dehydrogenase (PDH) plasmids for 4–6 hours, and fresh media containing zelenirstat (0, 0.1, 1, and 10 μM). After 48 h of treatment, NucBlue (cat#R3760; Thermo Fisher) or Nuclear ID Blue (Enzo, cat# ENZ-CHM103-0200) was added directly to the cell culture medium and incubated with the cells for 30 min. They were then washed three times with PBS and 2 mL of pre-warmed indicator-free DMEM (Agilent Seahorse XF) supplemented with glucose, pyruvate, and L-glutamine was pipetted for visualization under the microscope using the Volocity program. The files were processed using Fiji software (ImageJ).

For mitochondrial network visualization, Parental or NDUFAF4 KO MDA-MB-231 cells were split into small glass-bottom dishes and treated with 1 μM zelenirstat or not. After 48 h of treatment, MitoTracker Green and Nuclear ID Blue stains were prepared in warmed Hank’s Balanced Salt Solution (HBSS) and incubated with the cells for 30 min. They were then washed twice with HBSS and 2 mL of HBSS was pipetted for visualization under the microscope using the Volocity program. Files were processed using the Fiji software (ImageJ).

SiRNA Transfection

Opti-MEM was warmed before transfection. A volume of 100 μL of each siRNA (Scrambled, NMT1, or NMT2) was pipetted into each microtube/Eppendorf tube. Lipofectamine + Opti-MEM solution (10 μL of Lipofectamine and 100 μL of Opti-MEM) was pipetted into respective tubes. One extra microtube was added for the double knockdown condition, in which double of every solution was added; thus, 200 μL of Opti-MEM, 10 μL of siNMT1, 10 μL of siNMT2, and 220 μL of Lipofectamine + Opti-MEM (20 μL of Lipofectamine and 200 μL of Opti-MEM).

After 10 min, treated petri dishes were prepared by aspirating media, washed with PBS, and then incubated with 4 mL Opti-MEM. SiRNA contents were slowly dropped/pipetted from microtubes into respective dishes, with the control dish receiving no siRNA. The Petri dishes were then incubated for 4 h. Opti-MEM was aspirated and 13 mL of complete medium was added to each Petri dish. The dishes were incubated for three days (72 h).

High-Resolution Respirometry

Cell respiration was measured at 37°C using an Oxygraph-2k (Oroboros Instruments Inc., Innsbruck, Austria). Chambers were calibrated with 2 mL of mitochondrial respiration medium MiR05 (0.5 mM EGTA, 3 mM MgCl2-6H2O, 60 mM K-lactobionate, 20 mM taurine, 10 mM KH2PO4, 20 mM HEPES, 110 mM sucrose, and 1 g/L BSA essentially fatty-acid free, pH 7.1) prior to experimentation. Cells were treated with zelenirstat for 48 h, after which they were collected, scraped, counted, and resuspended in MiR05. The cell suspension was then placed in oxygraph chambers, and the O₂ flow (pmol·min⁻¹·million⁻¹ cells) was measured under routine conditions (routine state). After adding the complex I-linked substrates, pyruvate (5 mM) and malate (2 mM), and glutamate (10 mM), the LEAK state was initiated by permeabilization of the plasma membrane with digitonin (3–5 μg·mL⁻¹; optimal concentration determined with experiments on intact and treated cells, as described by Mann et al 2019).²⁸ The various pathways and steps were then measured with the addition of ADP (2.5 mM; NADH pathway through complex I), cytochrome c (10 μM; outer mitochondrial membrane integrity test), succinate (10 mM; combined NADH and succinate pathways respiration), and rotenone (1 μM; succinate pathway alone), antimycin A (5 μM; non-mitochondrial residual oxygen consumption, ROX, subtracted from all previously described measurements), ascorbate, and TMPD (2 mM and 0.5 mM, respectively; complex IV activity), and sodium azide (100 mM; chemical background subtracted from complex IV activity measurements). Measurements were taken using DatLab 7.4 software (Oroboros Instruments Inc). Citrate synthase activity was analyzed in MDA-MB-231 and MCF 10A cells to normalize for mitochondrial content between samples and experiments.

Mitochondrial Isolation

Mitochondria were isolated from the cells as described previously.²⁹ The medium was aspirated, and the cells were scraped in PBS, pelleted, and washed a second time in PBS. Following pelleting, the cells were resuspended in 1.05 mL of mitochondrial isolation buffer (MIB), and 50μL was removed and stored at −80°C for whole cell analysis. Mitochondria were isolated by drawing into and expelling the cell solution from a 3 mL syringe using 18.5 gauge and 27 gauge needles respectively, for a total of six times. Cell debris was pelleted twice: first in a 50 mL conical tube, followed by transfer of the supernatant, and a second centrifugation step in a 15 mL conical tube. The MIB enriched in mitochondria was transferred to a 1.5 mL microfuge tube, and mitochondria were pelleted at 10,000 × g for 5 min, washed in 1 mL MIB, pelleted again as described above, and finally resuspended in 100 μL MIB. Isolated mitochondria were stored at −80°C until further processing.

Mitochondrial Membrane Solubilization

Mitochondrial membranes were solubilized as previously described³⁰ at 5 mg·mL⁻¹, and the final supernatant was stored at −80°C until blue native PAGE was performed.

Glucose Quantification (HPLC)

Samples were run on a Shimadzu 20 HPLC system equipped with an Alltech 3300 ELSD detector. The column used was Phenomenex Luna NH2 (100 × 2 mm, 3 μm). The column oven temperature was maintained at 40°C. The program was run isocratically using 15:85 water/acetonitrile with a run time of 6 min and a flow rate of 1.0 mL·min⁻¹. Lactose was used as an internal standard.

Blue Native PAGE of Mitochondrial Electron Transport System Complexes

Blue native PAGE was performed on solubilized mitochondria using the buffers and methods described by Fernandez-Vizarra et al,³⁰ with the following modifications: PAGE was performed using 10–20 µg protein/lane for western blotting and 20–40 µg/lane for in-gel complex I activity assays using 8% resolving and 4% stacking wide-pore gels prepared from stock acrylamide solution containing 50% (w/v) acrylamide and 0.5% (w/v) bisacrylamide (100:1 acrylamide: bisacrylamide).³¹ Electrophoresis was carried out at 10 mA/gel until the voltage reached 250 V, at which point the current was reduced to 5 mA/gel. Electrophoresis was performed for 40 min after the front left the gel. Cathode buffer A was replaced by cathode buffer B when the front reached the halfway point of the gel.

Transfer and Western Blotting of Blue Native PAGE

Following electrophoresis, the gels were transferred to PVDF membranes following the protocol described in reference³⁰, with the transfer time at 300 mA increased to 4 h to obtain the best results. Membranes were blocked for 1 h in 5% BSA in PBS with 0.1% Tween-20 (blocking buffer) and incubated overnight in blocking buffer containing either Rb α-NDUFB11 to detect intact complex I, total OXPHOS antibody cocktail (ab110413, Abcam) to detect the levels of the remaining electron transport system complexes, or VDAC1 to detect total mitochondrial loading. The secondary goat α-Rb and goat α-Ms HRP-conjugated antibodies were incubated in blocking buffer for 1 h at room temperature. After washing, the membranes were developed using SuperSignal™ West Pico PLUS Chemiluminescent Substrate (Thermo Scientific 34577) and visualized using ImageQuant 800 (Amersham).

In-Gel Complex I Activity Assays

In-gel complex I activity assays were performed following blue native PAGE, as described in reference³⁰, for 1 h. The gels were visualized and imaged using an ImageQuant 800 (Amersham, UK) before and after the activity assay. To accurately quantify the nitro blue tetrazolium precipitate, the blue background (Coomassie G-250) was removed by soaking the gels in 5% methanol for 2 h (to shrink the swollen gels), followed by overnight incubation in destaining solution containing 10% methanol and 10% acetic acid.

Complex I Enzymatic Assay

MDA-MB-231 cells were treated with increasing concentrations of zelenirstat and incubated for 96 h. Cells were scraped, centrifuged, and resuspended in PBS. Complex I activity was determined using the Complex I Enzyme Activity Microplate Assay Kit (Abcam #ab109721). Briefly, protein concentrations were determined using the Pierce BCA Protein Assay Kit. The cell concentration of the initial solution was calculated, and the extracts were prepared using a detergent kit followed by centrifugation. A final sample concentration of 750 μg·mL⁻¹ was used to incubate the extracts in the plate. After 3 h, each well was washed thrice with wash buffer. After addition of the assay buffer, the absorbance of each well was measured for 60 min at 450 nm using a Cytation 5 cell imaging multimode reader (BioTek Instruments) and Gen5 Data Analysis software.

NAD⁺/NADH Quantification

Following pre-treatment with zelenirstat for 48 h, cells were collected, counted, and lysed before being processed to detect NAD⁺ or NADH using the NAD/NADH-Glo™ Assay (Promega) following the manufacturer’s instructions. After processing, luminescence was detected using a Cytation 5 plate reader.

ATP Quantification

Following pre-treatment with zelenirstat for 48 h, cells were collected and ATP was released with a Somatic Cell ATP Releasing Agent (Sigma-Aldrich) following the manufacturer’s instructions. The collected ATP was quantified using an ATP Determination Kit (Invitrogen #A22066) and analyzed using a Cytation 5 plate reader.

Timewise Respirometry

Cells were trypsinized and counted using a Countess 3 automated cell counter (Thermo Fisher Scientific). Cells were then seeded at 5000–20,000 cells/well. Cells were allowed to adhere to the plate for 1 h at 37°C with 5% CO₂ before changing the lid to an Oxygen Consumption Rate (OCR) sensing lid and attaching the Resipher cell culture monitor (Lucid Scientific) to the lid after placing the plate back into the incubator. OCR was measured for 24 h before refreshing the media with normal media or media containing increasing concentrations of zelenirstat or 1 μM IACS-010759.³² OCR was measured for approximately 120 h after treatment. OCR data were downloaded from the Lucid Lab online portal, processed using Microsoft Excel, and subsequently graphed and analyzed using GraphPad Prism software.

Seahorse Metabolic Analysis

Cells were cultured with zelenirstat or IACS-010759 for 24 h, then collected and counted with trypan blue (Thermo Fisher). 40,000 live cells were plated per well in Agilent XF DMEM with the recommended supplementation. After further incubation for 24 h, metabolic activity was measured using a Seahorse XFe96 Analyzer (University of Alberta Advanced Cell Exploration Core) using a Glycolytic Rate Assay Kit according to the manufacturer’s instructions. Cells plated and treated in identical pairwise conditions with zelenirstat or not were stained with NucBlue and Calcein-AM, then imaged using the Cytation 5 imager and counted using the Gen5 program.

Glucose Quantification (Glucose-Glo Assay)

Cells were cultured and treated with siRNA targeting AMPKβ1 (PRKAB1, Invitrogen #4390824), AMPKβ2 (PRKAB2, Invitrogen #4392420), or both using the method described under siRNA transfection, modified to 2 days treatment instead of 3 days. Parental cells were treated with scrambled siRNA. After refreshing the media, the media was collected after 24, 48, and 72 h of incubation and glucose was measured using the Glucose-Glo^TM Assay (Promega) according to the manufacturer’s instructions. The Cytation 5 plate reader was used to acquire luminescence readings after processing the samples.

Quantification of Western Blots

Image analysis was performed using the software program GIMP (GNU Image Manipulation Program) (GIMP, 2019) to quantify the intensity of the protein bands. This involved selecting the regions of interest (ROIs) corresponding to the protein bands and measuring the intensity of the pixels within these ROIs. Once protein bands were selected, they were opened in Image Studio Lite and were quantified. Values were transferred into Excel and intensity of the protein bands were normalized to an internal control (GAPDH or Actin). Once values have been corrected, they were transferred into GraphPad Prism to analyze the data statistically to determine the significance of the differences between the groups being compared. Lastly, the data were presented as a graphical representation (bar graphs).

Statistical Methods

Data were analyzed using Prism 10.0.0 software (GraphPad) and generally expressed as mean ± standard error of the mean (SEM) unless stated otherwise. Data were statistically analyzed to assess the significance of differences between groups using unpaired t-tests for two-group comparisons, ANOVA for multiple-group comparisons, or other appropriate statistical tests. * = p<0.05, ** = p<0.01, *** = p<0.001, **** = p<0.0001, ns = non-significant.

Data Availability

Data were extracted from publicly available datasets, including tumor RNA expression data from TCGA dataset (https://portal.gdc.cancer.gov) and Tumormap (https://tumormap.ucsc.edu).

Manuscript Editing

The final version of the manuscript was proofread using Paperpal software (Cactus Communications) for language accuracy and consistency.

Results

TNBC is Predicted to Be Sensitive to Myristoylation Inhibition

Myristoylation inhibition has been demonstrated to be especially effective in cancer cells highly dependent on NMT1.^11,19 Thus, we investigated the expression of NMT1 and NMT2 in breast cancer subtypes using BRCA expression data from The Cancer Genome Atlas (TCGA).³³ All breast cancer subtypes expressed both NMT1 and NMT2 at lower levels than paired normal breast tissues (Figure 1A). In addition to HER2⁺ samples, basal (TNBC) cancers expressed the lowest levels of NMT1. While basal and TNBC terms are often used interchangeably, it is important to note that these classifications are not the same, where mRNA expression markers are used to classify basal breast cancers, while TNBC is commonly classified by the histological absence of three hormonal receptors.³⁴ We used MISS-54 to predict sensitivity of breast cancer subtypes to zelenirstat. Basal samples had the highest MISS-54 score, indicating the highest predicted sensitivity to myristoylation inhibition compared to other subtypes of breast cancer and normal breast cells (Figure 1B and C). Breast cancer subtypes are readily separated from other tumor types and among themselves (Figure S1). TNBC tumors had a significantly higher MISS-54 score than non-TNBC BRCA subtypes (Figure 1D). Examining MISS-54 score across patient characteristics of age, pathological stage, and pathology factors produced limited significant differences (Figure S1A). Basal tumors also had a higher MISS-54 score than both normal breast and all normal tissue samples (Figure 1E). We then compared the viability of MDA-MB-231 TNBC cells and MCF 10A normal breast epithelial cells treated with zelenirstat. After 48 h of treatment, MDA-MB-231 cells displayed an EC₅₀ of 0.3461 µM, whereas MCF 10A cells displayed an EC₅₀ of 1.601 µM (Figure 1F, p<0.001). As we have previously demonstrated that zelenirstat reduces cellular respiration,^11,12 and since OXPHOS drives metastasis in TNBC cells,^7,8 we next investigated if MISS-54 levels in relation to the expression of the 200 genes of the gene set enrichment analysis (GSEA) OXPHOS hallmark.³⁵ All BRCA tumor samples were split evenly into “Low” and “High” OXPHOS categories according to their expression of metabolism-related genes. In basal samples alone, tumors that express lower levels of OXPHOS genes are predicted to be more sensitive to myristoylation inhibition (Figure 1G and S1B). When examining glycolysis using the same methodology, while no significant differences were found, most basal tumors highly expressed glycolysis genes, unlike other BRCA subtypes, indicating that glycolytic metabolism is also important in TNBC (Figure S1C).

Figure 1.

MISS-54 score predicts TNBC to be sensitive to zelenirstat treatment. (A) Violin plot of NMT1 and NMT2 expression levels across subtypes of breast cancer and normal breast cells. (B) Cluster map of breast tissue indicated by subtype or MISS-54 score. MISS-54 gene set sensitivity score to NMT inhibition was generated from 4 cell line screens (1200 cell lines) using myristoylation inhibitors. (C) Boxplot of MISS-54 score for TCGA samples across BRCA subtypes. (D) Comparison of MISS-54 score in triple-negative breast cancer to all non-TNBC breast cancer subtypes. (E) MISS-54 score of basal breast cancer compared to either normal breast tissue or all other normal tissues. (F) Viability of TNBC MDA-MB-231 cells compared to normal immortalized MCF 10A cells across increasing levels of zelenirstat treatment over 48 hours. Statistical analyses were performed using Prism software. (G) MISS-54 score comparison across low or high OXPHOS status basal breast cancer cells. * = p<0.05, ** = p<0.01, *** = p<0.001, **** = p<0.0001.

Myristoylation Inhibition Leads to the Degradation of Complex I and Loss of Energy Regulation

Given the association between MISS-54 and OXPHOS expression, we further examined metabolism in TNBC. We previously demonstrated that NMT1 KO in haploid HAP1 cells significantly reduced the levels of NDUFAF4, a myristoylated mitochondrial complex I structural assembly factor.^11,17,36 Thus, we examined whether this loss occurs in TNBC MDA-MB-231 cells under pharmacological myristoylation inhibition both below and above the EC₅₀. Zelenirstat treatment for 48 h reduced NDUFAF4 levels in a dose-dependent manner, starting at 0.005 µM (Figure 2A and B). MDA-MB-231 cells treated with siRNA against NMT1, NMT2, or both revealed NDUFAF4 to be an NMT1-specific substrate, with only NMT1 KD leading to a reduction in NDUFAF4 levels (Figure S2A and S2B). 48 hours of treatment with zelenirstat also reduced multiple complex I proteins including NDUFAF3, which binds to NDUFAF4 to assemble complex I,³⁶ NDUFS8 and NDUFB8, two proteins which are present in separate modules of complex I,³⁷ and NDUFB7, which has recently been shown to require myristoylation for its localization and function (Figure 2A–D).³⁸ These effects were also consistently observed in TNBC MDA-MB-453, MDA-MB-468, and BT-549 cell lines, with varying sensitivity to zelenirstat after 48 h of treatment (Figure S3A and B). In MCF 10A normal breast epithelial cells, zelenirstat treatment for 48 h reduced NDUFAF4 and NDUFS8, but not NDUFAF3, NDUFB8, or NDUFB7 (Figure S4A and B). We then investigated whether zelenirstat reduced the levels of total assembled complex I. Blue native gel electrophoresis revealed that zelenirstat reduced the total complex I levels in MDA-MB-231 cells (Figure 2E). Zelenirstat also reduced complex I levels in MCF 10A cells to a lesser extent (Figure S4C). Zelenirstat treatment did not reduce the levels of other complexes in the mitochondrial respiratory transport system (Figure S5). We also investigated whether a shorter exposure time to zelenirstat for 24 h was sufficient to reduce respiratory proteins. NDUFAF4 and NDUFB8, but not NDUFAF3 or ATP5A, were significantly reduced after 24 h (Figure S6).

Figure 2.

Impact of zelenirstat treatment on MDA-MB-231 mitochondrial proteins. MDA-MB-231 cells treated for 48 hours with zelenirstat were examined for the myristoylated mitochondrial complex I assembly factor NDUFAF4 (A and B), its non-myristoylated binding partner NDUFAF3, the non-myristoylated mitochondrial complex I proteins NDUFS8 and NDUFB8 (A–C), and the myristoylated mitochondrial complex I protein NDUFB7 by western blot (A and D). All protein levels were normalized to GAPDH (n≥3). (E) Blue native PAGE analysis displaying the total assembled complex I protein level (apparent MW>670 kDa) in MDA-MB-231 cells treated with zelenirstat or not. (F) Mic19 was analyzed using the same procedures as in A-D (n≥4). Blots shown are representative of primary data. Quantification was performed using densitometry analysis. * = p<0.05, ** = p<0.01, *** = p<0.001, **** = p<0.0001, ns = non-significant. (G) Role of myristoylation in the mitochondrial electron transport system. Zelenirstat inhibits NMT1 activity, which is involved in mitochondrial complex I assembly/activity by myristoylating NDUFAF4. NDUFAF4 combines with NDUFAF3 to assemble mitochondrial complex I. MIB indicates the mitochondrial intermembrane space bridging complex. Created with BioRender.com.

Mic19 is a myristoylated mitochondrial protein involved in cristae formation through the mitochondrial contact site and cristae organizing system (MICOS) complex.³⁹ Both NMT siRNA treatment and higher concentrations of zelenirstat for 48 h were not able to reduce the levels of this protein (Figures 2F, S2A and B).⁴⁰ Finally, we propose a model summarizing the role of myristoylated NDUFAF4 in the assembly of mitochondrial complex I within the electron transport system and that zelenirstat treatment, through inhibition of NMT1, prevents this assembly from occurring (Figure 2G).

NDUFAF4 is Required for the Proper Assembly of Mitochondrial Complex I

As NDUFAF4 is the most downregulated protein in NMT1 KO cells,¹¹ we next investigated whether its loss after zelenirstat treatment drives complex I degradation. We generated a CRISPR/Cas9 constitutive NDUFAF4 KO in MDA-MB-231 cells and confirmed a 19 base pair deletion in exon 1 by DNA sequencing (Figure S7). NDUFAF4 KO cells were validated for the absence of NDUFAF4 (Figure 3A). Protein levels of NDUFAF3, NDUFS8, NDUFB8, and NDUFB7 were drastically reduced (Figure 3A–C) in NDUFAF4 KO cells. NDUFAF4 KO also significantly reduced the levels of fully assembled complex I, as measured by blue native immunoblot (Figure 3D). NDUFAF4 KO cells exhibited increased UQCRC2 (complex III) and ATP5A (complex V) protein levels while having mitochondrial protein content similar to that of parental cells, as indicated by VDAC levels (Figure S8).

Figure 3.

Impact of NDUFAF4 KO in MDA-MB-231 on various mitochondrial complex I proteins. (A–C) NDUFAF4 was knocked out in MDA-MB-231 cells using CRISPR/Cas9 and the resulting cells were examined for the myristoylated mitochondrial complex I assembly factor NDUFAF4 (A) as well as non-myristoylated proteins NDUFAF3, NDUFS8, and NDUFB8 (A and B), and the myristoylated complex I protein NDUFB7 (A and C). Protein levels were normalized to Actin (n≥3). (D) Blue native PAGE analysis displaying the total assembled complex I protein level in parental or NDUFAF4 KO MDA-MB-231. *** = p<0.001, **** = p<0.0001. (E) Confocal microscopy of COS-7 cells transfected with plasmids expressing either EGFP, WT-NDUFAF4-EGFP, or G2A-NDUFAF4-EGFP cDNAs. Localization of NDUFAF4 was compared with cells both transfected with WT NDUFAF4 and treated with zelenirstat for 48 hours. Transfection of a plasmid expressing pyruvate dehydrogenase (PDH-mCherry) was used to visualize mitochondria. Scale bars represent 5µm.

As myristoylation is important for the localization of many proteins, we next explored the impact of ablating the specific myristoylation of NDUFAF4 by transiently expressing WT or mutated non-myristoylated G2A-NDUFAF4-EGFP in COS-7 cells. The expression of both chimeric proteins was validated by western blotting (Figure S9A). We observed that while WT-NDUFAF4 colocalized with the mitochondria, as indicated by its merging with a pyruvate dehydrogenase (PDH) vector, G2A-NDUFAF4 showed cytoplasmic localization similar to that of the EGFP control alone (Figure 3E, S9B). This localization was replicated when using pharmacological myristoylation inhibition with zelenirstat. G2A-NDUFAF4 did not colocalize with cathepsin, indicating it was not undergoing lysosomal degradation (Figure S9C). Overall, we confirmed that zelenirstat reduced NDUFAF4 protein levels, leading to misassembly and degradation of complex I.

Zelenirstat Causes a Reduction in TNBC Mitochondrial Complex I Activity and Metabolic Disruption

As we observed a significant reduction in mitochondrial complex I protein levels after zelenirstat treatment, we next tested the physiological effects of this reduction. Zelenirstat treatment for 48 h reduced mitochondrial complex I in-gel and enzymatic activity in MDA-MB-231 cells (Figure 4A and B). Zelenirstat also reduced complex I activity in MCF 10A cells (Figure S4D). This effect was recapitulated in MDA-MB-231 NDUFAF4 KO cells, indicating that it was NDUFAF4-dependent (Figure 4C). We next explored whether this reduction in complex I activity in zelenirstat-treated or NDUFAF4 KO cells reduces cellular ATP levels. ATP levels were slightly reduced by zelenirstat treatment or NDUFAF4 KO alone, but this effect was potentiated in NDUFAF4 KO cells treated with zelenirstat (Figures 4D and S10). As we observed a disruption of ATP levels in zelenirstat-treated cells, we analyzed the cell culture media for glucose levels, a common substrate used to generate ATP. We also identified that zelenirstat treatment for 72 h decreased the amount of glucose consumed in the media (Figure S11).

Figure 4.

Impact of zelenirstat and NDUFAF4 KO on complex I activity and cellular energetics. (A and B) MDA-MB-231 treated with zelenirstat or not were analyzed for complex I in-gel activity using blue native PAGE (n=4) (A) or a complex I enzymatic colorimetric assay (n=3) (B). (C) Parental or NDUFAF4 KO complex I in-gel activity analyzed using blue native PAGE (n=3). (D) ATP levels of wildtype or NDUFAF4 KO MDA-MB-231 cells treated with varying concentrations of zelenirstat (n=4). (E) Individual NAD or NADH levels in response to zelenirstat treatment for 48 hours. (F) NAD/NADH ratio of wildtype or NDUFAF4 KO MDA-MB-231 cells treated with varying concentrations of zelenirstat (n=4). (G) Ratio of phosphorylated to non-phosphorylated PDH in MDA-MB-231 cells treated with zelenirstat examined through western blotting. (H–J) AMPKβ1/2 or AMPKα protein levels were analyzed for protein content in either wildtype, zelenirstat-treated, or NDUFAF4 KO cells using western blot. Blots shown are representative of primary data. Quantification was performed using densitometry analysis. * = p<0.05, ** = p<0.01, *** = p<0.001, **** = p<0.0001.

To further explore the effect of zelenirstat or NDUFAF4 KO on cellular metabolism, we examined the levels of NAD⁺ and NADH inside the cells. Since complex I activity is reduced, we hypothesized that this may affect the processing of NADH to NAD⁺. While NAD⁺ levels remained stable, NADH levels increased in both parental and NDUFAF4 KO cells upon zelenirstat treatment (Figure 4E and F). As NADH is created predominantly through the citric acid cycle, we investigated whether pyruvate was being processed to acetyl-CoA through pyruvate dehydrogenase (PDH) in this cycle, because accumulation of NADH can cause inhibition of PDH, leading to further metabolic disruption.⁴¹ We observed an increase in PDH phosphorylation at serine residues 293 and 232, indicating PDH inhibition (Figure 4G). The overall protein levels of non-phosphorylated pyruvate dehydrogenase and pyruvate dehydrogenase kinases (PDKs) did not change with zelenirstat treatment (Figure S12A).

With a significant decrease of complex I activity leading to NADH accumulation and subsequent PDH inhibition, we next explored how the cell might react to this metabolic disruption, especially via AMPK. AMPKβ is a myristoylated subunit of AMPK, a protein that acts as an ATP/ADP sensor to mediate a switch to glycolytic metabolism upon activation.^42,43 We have previously shown that AMPKβ and AMPKα levels decrease upon zelenirstat treatment in AML cells.¹² Zelenirstat treatment reduced both the levels of myristoylated AMPKβ1 and β2 subunits in MDA-MB-231 cells (Figure 4H). These effects were also consistently observed in MDA-MB-453, MDA-MB-468, and BT-549 cell lines (Figure S12B). We next demonstrated that this effect is NDUFAF4-independent, as NDUFAF4 KO cells did not display decreased AMPKβ1/2 levels (Figure 4I). Finally, we observed a significant and dose-dependent decrease in the non-myristoylated catalytic AMPKα subunit, confirming that destabilization of AMPK through the loss of myristoylation is sufficient to decrease total AMPK (Figure 4J).

Zelenirstat Reduces Both OXPHOS and Glycolysis in TNBC Cells

Considering the high impact of zelenirstat on mitochondrial complex I and metabolic regulators, we next examined its impact on cellular respiration. We first used Oroboros respirometry to distinguish the separate contributions of the NADH or Succinate respiration pathways, all normalized to CS activity and expressed as the % of the control cells. NADH pathway (through complex I) respiration decreased in MDA-MB-231 cells when the cells were treated with increasing zelenirstat concentrations for 48 h (Figure 5A). Succinate pathway respiration (through complex II) and total respiration (through complexes I and II) decreased with increasing drug treatment (Figure S13A and B). Importantly, the NADH/succinate pathway ratio decreased upon zelenirstat treatment, indicating a reduced contribution of the NADH-linked pathway to mitochondrial respiration (Figure 5B). While MCF 10A cells had reduced respiration, their NADH/succinate pathway ratio was stable, indicating no specific reduction in NADH pathway respiration (Figure S13C and D). Unlike MDA-MB-231 cells, MCF 10A cells showed increased citrate synthase activity when treated, indicating increased mitochondrial content in response to zelenirstat. This response indicated that the observed decrease in respiration was due to a loss of overall mitochondrial function rather than a specific complex I defect. To further examine the impact of zelenirstat on mitochondrial respiration, we utilized a Resipher cell culture monitor that monitors the routine oxygen consumption rate (OCR) over time. Zelenirstat significantly decreased OCR in as little as 24 h after treatment (Figures 5C and D). IACS-010759 treatment rapidly decreased respiration to near-absent levels. NDUFAF4 KO cells displayed decreased respiration compared to parental cells, and peaked at maximal respiration faster than that of parental cells. These differences indicate that NDUFAF4 KO cells progressively exhaust extracellular substrates that fuel alternative mitochondrial electron entry pathways, thereby limiting non-complex I-linked respiration. In these cells, zelenirstat also produced an additive decrease in OCR, thus highlighting sensitivity to zelenirstat when no NDUFAF4 is present, indicating an NDUFAF4-independent effect of zelenirstat on respiration. The observed reduction in OCR was not solely dependent on the impact of zelenirstat on paired viability and cell counts over time (Figure S14). These effects were also consistently observed in MDA-MB-453, MDA-MB-468, and BT-549 cell lines (Figure S15). Importantly, although there was not a significant decrease in viability of MDA-MB-453 cells, their respiration was drastically reduced upon treatment with 1 µM of zelenirstat.

Figure 5.

Zelenirstat reduces breast cancer cell oxidative phosphorylation. (A and B) Measurement of oxygen consumption rate (OCR) from individual electron transport system pathways in MDA-MB-231 cells after zelenirstat treatment using Oroboros respirometry. Respiration was standardized for mitochondrial content using citrate synthase enzyme activity (n≥6). (C) Average OCR of MDA-MB-231 cells after zelenirstat or IACS-010759 (complex I inhibitor) treatment over time using a Resipher respirometer (n=5). (D) OCR was quantified at different time points after drug treatment. (E) OCR or extracellular acidification rate (ECAR) of MDA-MB-231 cells treated with zelenirstat or IACS-010759 (IACS) measured using a Seahorse XFe96 Metabolic Analyzer. Rotenone/Antimycin A was added after 18 minutes of measurement, followed by addition of 2-DG 18 minutes later (n=4). (F) Basal and compensatory glycolysis of MDA-MB-231 cells derived from (E). * = p<0.05, ** = p<0.01, *** = p<0.001, **** = p<0.0001.

With a large effect on mitochondrial respiration, we investigated the morphology of the mitochondrial network following zelenirstat treatment or NDUFAF4 KO. We observed an intricate and well-defined mitochondrial network in parental cells, whereas the network in zelenirstat-treated cells showed less complexity, branching, and swollen mitochondria (Figure S16). NDUFAF4 KO alone did not show any significant qualitative differences compared to parental MDA-MB-231 cells.

Next, we explored the possibility that glycolysis is affected by zelenirstat treatment through AMPK and other myristoylated proteins. Parental, zelenirstat-treated, IACS-010759-treated, or NDUFAF4 KO MDA-MB-231 cells were examined using a Seahorse metabolic analyzer with a glycolytic rate assay kit. We observed a significant decrease in routine OCR following zelenirstat treatment without significant loss in cell viability (Figure 5E and S17A). The extracellular acidification rate (ECAR) was used to measure glycolytic rate in each group. Zelenirstat decreased the glycolytic rate of both parental and NDUFAF4 KO cells to the same level, indicating that zelenirstat decreased glycolysis in an NDUFAF4-independent manner (Figure 5E and F). Untreated NDUFAF4 KO cells displayed elevated glycolysis compared to parental cells, whereas IACS-010759 treatment significantly increased basal glycolysis. Compensatory glycolysis, which is the level of glycolysis following the addition of OXPHOS inhibitors Rotenone and Antimycin A, was significantly blunted in zelenirstat-treated cells. Dual siRNA-mediated knockdown of AMPKβ1 and AMPKβ2 also caused an increase in the amount of remaining glucose in cell media (Figure S17B). Taken together, zelenirstat reduces both OXPHOS and glycolysis in TNBC MDA-MB-231 cells.

Zelenirstat Reduces Cell Migration, Invasion, and Cancer Stem Cell Growth

As OXPHOS is heavily involved in the migration and metastasis of cancer cells,^7,8 we explored the specific effects of myristoylation inhibition on myristoylated proteins associated with migration. SFKs, such as Src and Lyn, are myristoylated oncoproteins that have been shown to influence migration and cancer aggressiveness.^9,44–46 Zelenirstat treatment dose-dependently reduced both Src and Lyn in MDA-MB-231 TNBC cells (Figure 6A). We also consistently observed these effects in MDA-MB-453, MDA-MB-468, and BT-549 cell lines (Figure S18A). NMT1 siRNA reduced Lyn, but not Src levels (Figure S2C and D). Zelenirstat significantly reduced Lyn, but not Src levels in normal breast epithelium MCF 10A cells (Figure S18B). Phosphorylation of SFKs is required for their activation, and zelenirstat treatment for 48 h reduced total SFK phosphorylation in MDA-MB-231 cells (Figure 6B).

Figure 6.

Zelenirstat inhibits TNBC migration and mammosphere formation. (A–C) MDA-MB-231 cells were treated for 48 hours with zelenirstat at varying concentrations and their Src and Lyn (A), P-SFK (B), or Flotillin-2 and FAM49B (C) protein level was analyzed by western blotting. Protein levels were normalized to GAPDH (n≥3). Blots shown are representative of primary data. Quantification was performed using densitometry analysis. (D) MDA-MB-231 confluent cell layers were scratched in low-serum media and imaged over 48 hours with or without zelenirstat treatment. Cells were stained with Calcein-AM viability dye after 48 hours. Yellow mask represents scratch from t=0 overlaid onto endpoint image. Scale bar represents 500µm. (E) Wound healing assay plot showing wound width remaining over time in MDA-MB-231 cells during treatment timepoints. Error bars represent standard error of the mean (n=5). (F) Corning FluoroBlokTM Transwell migration and Matrigel invasion assays in the presence of zelenirstat or not for 48 hrs. MDA-MB-231 cells stained with Calcein-AM viability dye to mark live invaded cells. Cell masks generated by Gen5 program on Cytation 5. Images representative of experiments conducted in triplicate (n≥3). Scale bars represent 2000µm. (G) Quantification of Transwell invasion and migration amounts as counted by Gen5. (H) Mammosphere formation assay in the presence of zelenirstat or not for 7 days. MDA-MB-231 cells were cultured in ultra-low attachment plate, visualized with Cytation 5, and counted. Images representative of 3 experiments completed in duplicate. Scale bar represents 100µm. (I) Quantified mammosphere forming efficiency from 6000 total cells initially plated in each condition. * = p<0.05, ** = p<0.01, *** = p<0.001, **** = p<0.0001, ns = non-significant.

We also investigated the impact of zelenirstat on Flotillin-2, another myristoylated protein involved in lipid rafts that has been previously shown to influence cancer cell motility.^47,48 Zelenirstat dose-dependently reduced Flotillin-2 in both MDA-MB-231 and MCF 10A cells (Figure 6C, S18B). NMT siRNA failed to lower Flotillin-2 levels (Figure S2D). FAM49B is another myristoylated protein that is heavily involved in breast cancer motility via its interaction with Rac1.⁴⁹ Zelenirstat drastically reduced FAM49B levels in MDA-MB-231 cells and additional TNBC cell lines (Figure 6C, S18A).

Next, we explored whether inhibition of myristoylation affected cell migration. We initially hypothesized that its effect on OXPHOS would decrease cell migration. Zelenirstat reduced wound healing migration in a dose-dependent manner for MDA-MB-231 cells (Figure 6D and E, S19). Calcein was added to the cells after 48 h of treatment to ensure that both the migrated and surrounding cells remained viable after drug treatment. No significant loss of viability was observed, indicating the anti-migratory effects are independent of cytotoxicity. These effects were also consistently observed in MDA-MB-468 and BT-549 cell lines (Figure S20A). NDUFAF4 KO did not affect wound or transwell migration (Figure S20B and C). ER⁺ MCF-7 cells also displayed a dose-dependent reduction in migration (Figure S20D and E). Zelenirstat did not affect MCF 10A cell migration, which could only be measured at the fluorescence endpoint due to cell morphology (Figure S20F). To strengthen our findings, we conducted transwell migration and invasion assays. Zelenirstat dose-dependently reduced both transwell migration and invasion in response to chemoattraction using fetal bovine serum (Figure 6F and G).

Cancer stem cells (CSCs) are a small subpopulation of cancer cells that have self-renewal capacity and are thought to be a source of cancer relapse and metastasis.⁵⁰ CSCs importantly exhibit metabolic plasticity and can use either OXPHOS or glycolysis, making them resistant to multiple types of therapies.^51–53 Since zelenirstat reduces both of these pathways, we hypothesized that zelenirstat could also reduce the growth of CSCs in vitro. Both zelenirstat treatment for 7 days and NDUFAF4 KO individually reduced mammosphere formation and produced dose-dependent additive effects on CSC growth when combined (Figure 6H and I). Altogether, we show that zelenirstat reduces migration, invasion, and the growth of cancer stem cells.

Discussion

Targeting cancer cell metabolism has emerged as a means to exploit new vulnerabilities, find novel therapies to kill cancer cells, and combat chemotherapeutic resistance. We explored the impact of genetic and/or pharmacological loss of myristoylation on signalling and mitochondrial protein levels, as well as on the metabolism and motility of TNBC cells. We revealed that MISS-54 predicts TNBC to not only be the most zelenirstat-sensitive breast cancer subtype but also that it is more sensitive to zelenirstat than all normal tissues. As there are few effective therapies for TNBC, the most aggressive subtype of breast cancer,⁵⁴ this warrants the preclinical evaluation of zelenirstat as a potential treatment for TNBC.

We demonstrate decreased expression of NMT1 and NMT2 in TNBC tissue compared to normal breast tissue. These results corroborate our previously reported data on NMT2 protein levels, which were significantly decreased in breast cancer tissue.²⁰ These results suggest that targeting the remaining NMT1 is a viable therapeutic approach. Compared to normal MCF 10A cells, zelenirstat preferentially killed MDA-MB-231 TNBC cells in vitro. This corroborates previously observed in vivo results using an MDA-MB-231 cell line-derived xenograft (CDX) model.²⁰ Furthermore, abolishing NMT1 in HAP1 cells abrogated the function of the myristoylated NDUFAF4, causing a significant reduction in complex I activity.¹¹ In MDA-MB-231 TNBC cells, zelenirstat not only reduces levels of NDUFAF4, but also reduces levels of other myristoylated and non-myristoylated complex I proteins, leading to the failure to properly assemble complex I and loss of its activity. As NDUFAF4 KO recapitulates these effects, the loss of NDUFAF4 after zelenirstat treatment appears to be the main driver of complex I degradation, thereby detrimentally impacting its activity and OXPHOS.

Interestingly, zelenirstat treatment did not lead to the degradation of all myristoylated mitochondrial proteins. We observed that myristoylated Mic19 was resistant to degradation, perhaps because of slow protein turnover or failure to expose its N-terminal glycine to the glycine-specific N-degron,¹⁶ despite its myristoylation being critical for its import into the mitochondria.⁵⁵ In NDUFAF4 KO cells, the specific degradation of complex I, which impaired the electron transport system and ATP production, also appeared to be compensated by increased levels of other respiratory complexes. It is also noteworthy that myristoylation of NDUFAF4-EGFP is required for proper mitochondrial targeting, since genetic (G2A mutation) or pharmacological (zelenirstat) abrogation of myristoylation prevented mitochondrial localization.

Physiologically, zelenirstat diminishes NADH pathway (through complex I) respiration. As NADH is the main source of electrons for OXPHOS, complex I failure results in a significant loss in the respiration of these cells and accumulation of this reduced dinucleotide within the cell. Interestingly, the reduction in total ATP level was enhanced in NDUFAF4 KO cells treated with zelenirstat compared to parental cells. This difference may indicate that zelenirstat impacts energy-producing pathways other than those relying on complex I. NDUFAF4 KO cells reached peak OCR faster than parental cells, albeit at a 50% reduced level compared to parental cells, which declined rapidly after the peak OCR was reached. This contrast suggests NDUFAF4 KO cells possibly rely upon an alternative energy source, which appeared to run out shortly after the peak OCR was reached. The identity of this energy source is unknown and is currently under investigation.

OCR, as well as the previously investigated mitochondrial complex I proteins, were also reduced in multiple other TNBC cell lines, including MDA-MB-453 cells. This cell line was resistant to cell killing even after 96 h of zelenirstat treatment, further highlighting physiological effects of zelenirstat on cancer cell metabolism independent of cell killing.

While the anaerobic use of glucose in the presence of oxygen (Warburg effect) is essential for tumor growth, mitochondrial OXPHOS is now also known to be crucial for tumor cell metastasis.^7,8 AMPK is a master cellular energy regulator whose homeostatic function is to balance energy levels by promoting the selective use of different energy-generating fuels under various physiological conditions.⁴² The loss of myristoylated AMPKβ and its degradation through the glycine-specific N-degron pathway¹⁶ as well as the subsequent decrease of non-myristoylated AMPKα subunit levels upon zelenirstat treatment can explain in part why basal and compensatory glycolysis were also reduced in MDA-MB-231 cells alongside blunted OCR due to improperly assembled complex I. The fact that NDUFAF4 KO cells could readily use basal and compensatory glycolysis, leading to increased ECAR, supports this critical role for myristoylated AMPK as a metabolic switch. Recently, multiple attempts have been made to target mitochondrial metabolism in the treatment of solid tumors.^56–58 However, many trials, such as those for IACS-010759, CPI-613, CB-839, and metformin, have failed due to lack of efficacy or adverse toxicity, presumably as a result of normal cell mitochondria being impacted by these therapies.^59–61 As zelenirstat already cleared phase I trials in patients with advanced refractory lymphoma or solid tumors and MISS-54 predicts that it has a plausible therapeutic window for the treatment of TNBC, we think it could selectively target both the oxidative metabolism of TNBC cells and their use of glucose, thereby potentially preventing their metastasis.

As a premise to study the potential impact of zelenirstat on metastasis, we monitored its effect on several myristoylated proteins, including SFKs, involved in cell motility.^9,44 Zelenirstat treatment reduced Src, Lyn, and total phosphorylated SFK levels, as well as that of flotillin-2. These observations, combined with the reduction in TNBC cell migration and invasion, reaffirmed its potential not only as a growth signalling and metabolic inhibitor but also as a potential anti-metastatic agent. Since neither NDUFAF4 KO nor IACS-010759 reduced migration, the effect of zelenirstat on migration is likely independent of its metabolic inhibition properties. When only OXPHOS is inhibited, various cancer cells are able to reprogram their metabolism to maintain ATP production, thus maintaining migratory capacity in these cells.^62,63 Zelenirstat also abrogated the growth of MDA-MB-231 mammospheres, indicating that it could also impact TNBC CSC survival in vivo. CSCs can use either glycolysis, OXPHOS, or both pathways to provide energy to fuel their proliferation.⁵¹ CSCs can also use FAO, a process mediated by Src in the mitochondria,⁹ to generate energy and inhibit apoptosis.⁵¹ Importantly, given that CSCs display metabolic plasticity and can show heterogeneity within the same tumor,⁵¹ targeting multiple metabolic pathways using zelenirstat could further justify its use as a potential anti-metastatic pharmacological treatment. We have already shown that zelenirstat selectively reduces leukemic cancer stem cell growth,¹² and, although we previously demonstrated that zelenirstat reduces TNBC tumor growth in vivo,²⁰ further investigation is needed to evaluate its potential anti-metastatic effects on TNBC and other aggressive, highly metastatic solid tumors. Importantly, the phase I clinical trial with zelenirstat revealed that zelenirstat treatment showed early signs of efficacy by achieving stable disease status in several patients with metastatic cancers.²³

In summary, in addition to inhibiting growth signalling, we show that the NMTi zelenirstat also causes the loss of NDUFAF4, which promotes complex I degradation and abrogates OXPHOS through complex I. Zelenirstat further prevents cells from using glucose as an alternative energy source by promoting AMPK degradation. Consequently, zelenirstat inhibited TNBC cell migration and invasion, and promoted the killing of TNBC CSCs. These multiple inhibitory properties are unique and highly desirable for zelenirstat as an orally available agent with minimal clinical toxicity. Our findings warrant further preclinical and potentially clinical studies of zelenirstat as a single agent or in combination therapies.

Funding Statement

RP was funded by scholarships from the Canadian Institutes of Health Research (CIHR) CGS-M, Alberta Innovates (AI) summer studentship, Cancer Research Institute of Northern Alberta (CRINA) travel award, and the Cancer Research Society (CRS) Doctoral Research Award. This work was supported by the Alberta Cancer Foundation (ACF) (grant numbers 26362 and 26927), the Award in Hematological Cancers’ Research in Memory of Dr. Rachel Mandel, an Alberta Paving Ltd. donation to LGB, a generous donation from Dr. Heleen and Rod McLeod through CRINA, Eusera (https://www.eusera.com), and Pacylex Pharmaceuticals Inc. (https://www.pacylex.com). All experiments were performed under the Cell Biology Research Program at the University of Alberta.

Abbreviations

AML, Acute myeloid leukemia; BCSC(s), Breast cancer stem cell(s); BRCA, Breast invasive carcinoma cohort from The Cancer Genome Atlas (TCGA); CDX, Cell line–derived xenograft; CSC(s), Cancer stem cell(s); EC₅₀, Half-maximal effective concentration; ECAR, Extracellular acidification rate; FAO, Fatty acid oxidation; GSEA, Gene set enrichment analysis; KO, Knockout; MISS-54, Myristoylation inhibition sensitivity signature (54-gene score); N-degron, N-terminal degradation signal; NMT, N-myristoyltransferase; NMT1, NMT2, N-myristoyltransferase 1, N-myristoyltransferase 2; NMTi(s), N-myristoylation inhibitor(s); OCR, Oxygen consumption rate; OXPHOS, Oxidative phosphorylation; PDH, Pyruvate dehydrogenase; PDHK1, Pyruvate dehydrogenase kinase 1; PDK(s), Pyruvate dehydrogenase kinase(s); PGC-1α, Peroxisome proliferator–activated receptor gamma coactivator 1-alpha; RIPA, Radioimmunoprecipitation assay buffer; ROX, Residual oxygen consumption; SFK(s), Src family kinase(s); TCA (cycle), Tricarboxylic acid cycle; TCGA, The Cancer Genome Atlas; TNBC, Triple-negative breast cancer; UQCRC2, Ubiquinol–cytochrome c reductase core protein 2; VDAC1, Voltage-dependent anion channel 1; WT, Wild-type.

Data Sharing Statement

The datasets generated and/or analyzed during the current study are available in TCGA database. https://www.cancer.gov/ccg/research/genome-sequencing/tcga.

Ethics Approval and Consent to Participate

All human data acquired from TCGA was public anonymous data. Secondary use of anonymous information or anonymous biological materials is exempt from University of Alberta ethics approval.

Author Contributions

All authors made a significant contribution to the work reported, whether that is in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas; took part in drafting, revising or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work.

Disclosure

As the co-founder(s) and/or shareholder(s) and/or convertible note-holder(s) of Pacylex Pharmaceuticals Inc. (https://www.pacylex.com), which owns the rights to patent applications PCT/2012/000696 and PCT/2013/050821, L.G.B., E.B., and R.P. declare potential financial competing interests. As employees of Pacylex Pharmaceuticals, L.G.B. and E.B. declare potential non-financial interests. The authors report no other conflicts of interest in this work.