Very long chain fatty acid metabolism is required in acute myeloid leukemia

Acute myeloid leukemia (AML) is a malignant hematological disease characterized by poor patient outcomes and limited chemotherapy options. Mitochondrial phenotypes of leukemia and leukemia stem cells (LSCs) have distinguishing features that separate them from normal hematopoietic stem cells (HSCs).^1-3  This altered phenotype is characterized by increased mitochondrial mass¹  as well as modified metabolic features such as a greater reliance on oxidative phosphorylation (OXPHOS)^1,2  and fatty acid oxidation (FAO)³  for survival. Unlike normal HSCs, LSCs are less able to compensate for added metabolic^4,5  or oxidative stress,⁶  which further defines the mitochondria as a vulnerable target that is essential for AML and LSC survival. As such, drug-induced perturbations of mitochondrial activity through the inhibition of mitochondrial protein synthesis,¹  OXPHOS,^4,6  or FAO^2,3,7  can induce or sensitize leukemia and LSCs to death. Given that compounds targeting these pathways and other metabolic perturbations in AML are being evaluated in human clinical trials,⁸  targeting mitochondria and altering leukemia cell metabolism is an attractive approach to AML therapy.

Mitochondrial FAO is the primary catabolic pathway for lipids <18 carbons in length.⁹  FAO supports leukemia cell survival by generating reductive equivalents (ie, NADH and FADH₂) for the electron transport chain (ETC), providing acetyl-CoA to the tricarboxylic acid (TCA) cycle,³  preventing accumulation of cytosolic palmitate, and indirectly contributing to the cytosolic antioxidant pool (through NADPH).³  Pharmacological inhibition of FAO results in antileukemia activity; however, seminal studies using etomoxir (ETO; 50% effective concentration >100 µM), which blocks fatty acid entry into mitochondria by inhibiting carnitine palmitoyltransferase I (CPT1), only sensitized leukemia cells to cytarabine-induced apoptosis.³  ETO’s hepatotoxicity prevented further clinical use,¹⁰  and only a few FAO inhibitors are clinically available; they are limited by their unique tissue specificity and unknown mechanism of activity.^11-16  Together, this highlights the role of FAO in AML and the need to better understand how FAO impacts AML cell metabolism.

To better understand the role of FAO in AML cell survival, we first compared gene expression profiles between normal and leukemia patient populations and found that ACADVL, the gene encoding very long chain acyl-CoA dehydrogenase (VLCAD), was elevated. We next showed that VLCAD is an enzyme that is critical to leukemia cell FAO, survival, and engraftment in vitro and in vivo. Genetic or pharmacological inhibition of VLCAD inhibited long chain FAO, the contribution of fat-derived carbons into the TCA cycle, and downstream OXPHOS. This triggered an insufficient compensatory shift in energy metabolism, ultimately resulting in selective AML cell death while sparing normal cells. Thus, we highlight a novel role for VLCAD in AML and validated a metabolic vulnerability in AML.

TLS-ERG immortalized (TEX) leukemia cells were maintained in Iscove's Modified Dulbecco medium (IMDM), 15% fetal bovine serum, 2 mM l-glutamine, 2% penicillin-streptomycin, 20 ng/mL stem cell factor, and 2 ng/mL interleukin-3.^17,18  OCI-AML2 (AML2) leukemia cells were maintained in IMDM, 10% fetal bovine serum, and 2% penicillin-streptomycin. Cells were cultured in 5% CO₂ at 37°C and passaged up to 25 times. The sources for human AML and normal samples are detailed in supplemental Methods (available on the Blood Web site); AML patient cytogenetic profiles are detailed in supplemental Figure 10.

Cell growth and viability and clonogenic growth were carried out as previously described; 7-aminoactinomycin D was used on a Guava EasyCyte 8HT flow cytometer.¹⁹  For rescue experiments, sodium heptanoate (TCI America) was solubilized, according to the manufacturer’s instructions.

Data sets (GSE9476, GSE13204, and GSE1159) were downloaded from the National Center for Biotechnology Information’s Gene Expression Omnibus (https://www.ncbi.nlm.nih.gov/geo/) and separated into 2 groups: normal healthy bone marrow (BM) and AML patient groups. ACADVL expression between normal healthy BM and AML patient groups were compared with a 2-tailed, unpaired Student t test in GraphPad Prism 7.0. All non-AML patient samples were excluded from the analysis.

Coimmunoprecipitation (co-IP) of VLCAD was conducted using the Crosslink Magnetic Co-IP Kit (Pierce) and an anti-VLCAD antibody for co-IP (NBP2-15238; Novus Biologicals); subsequent quantification of avocadyne (AYNE) was carried out using our published method.²⁰ 

Total cell lysates were prepared as described previously¹⁹  and detailed in supplemental Methods.

Oxygen flux was measured using a high-resolution respirometer (Oxygraph-O2k; Oroboros Instruments), with all protocols detailed in supplemental Methods. Intact cell respiration was measured as previously described²¹ ; fat-supported electron transfer flavoprotein (ETF) respiration with FAO substrates was assessed as previously described²² ; Complex I and II–supported respiration was assessed as previously described.²³ 

FAO activity in leukemia cell lines was determined as previously described,²⁴  with slight modification, for suspension cell lines and as detailed in supplemental Methods.

Acyl-CoA dehydrogenase (ACAD) activity was assessed as previously described²⁵  and detailed in supplemental Methods.

Lentiviral transductions were performed as described¹⁹  and detailed in supplemental Methods.

Acylcarnitine (AC) profiling was carried out as previously described²⁶  and detailed in supplemental Methods.

Adenosine triphosphate (ATP; Sigma), pyruvate dehydrogenase (PDH) activity (Sigma), and mitochondrial mass (MitoTracker; Thermo Fisher) were quantified using kits and following the manufacturer’s instructions, with no deviations. Cells were supplied with 200 µM palmitic acid: bovine serum albumin (PA:BSA) and 400 µM l-carnitine for 3 hours prior to ATP and PDH quantification. Primary AML and mononuclear cells (MNCs) were treated with AYNE for 12 hours prior to ATP quantification.

TEX cells, AML2 cells, primary AML cells, or normal MNCs (2.5 × 10⁵ cells per milliliter) were treated with a solvent vehicle, AYNE, or ETO and incubated with stable ¹³C isotope substrates, including uniformly labeled ¹³C₁₆ palmitate or uniformly labeled ¹³C₆ glucose (both from Cambridge Isotopes), for 12 hours. Cells were then collected, counted, washed 3 times in phosphate-buffered saline, and frozen at −82°C until ultra high performance liquid chromatography (UHPLC) analysis. Additional details can be found in supplemental Methods. Cells were extracted in methanol/acetonitrile/water (5:3:2 volume-to-volume-to-volume ratio [%]) prior to UHPLC–mass spectrometry (MS) analyses (Vanquish-Q Exactive; Thermo Fisher), as described previously.²⁷  Peak areas and isotopologue deconvolution were determined as described.²⁷ 

Nonobese diabetic severe combined immunodeficiency γ (NSG) mice (The Jackson Laboratory, Bar Harbor, ME or Western University, London, ON) were used for xenograft and engraftment assays as previously described^7,28  and detailed in supplemental Methods.

Assessment of VLCAD/drug binding was performed using a previously published protocol²⁹  as detailed in supplemental Methods.

Unless otherwise stated, in vitro results are presented as mean ± standard deviation, whereas in vivo results are presented as mean ± standard error of the mean. Data were analyzed with GraphPad Prism 7.0 (GraphPad Software) using 1-way analysis of variance (ANOVA) with Tukey’s post hoc analysis for between-group comparisons or standard Student t tests where appropriate; Mann-Whitney U tests were used for animal experiments. P values <.05 were considered statistically significant.

Pharmacological inhibition of CD36 or CPT1 targeted AML mitochondrial metabolism and resensitized previously resistant cells to current induction chemotherapy.^2,3,5  Given this importance of FAO in leukemia survival and relapse, we profiled levels of intramitochondrial enzymes⁹  associated with the carnitine shuttle; long, medium, and short chain FAO; and ACAD electron transfer in normal healthy BM and AML populations from publicly available databases (Figure 1A; supplemental Figure 1A). Across GSE9476, GSE13204, and GSE1159, the gene most consistently overexpressed among FAO enzymes in the AML patient vs normal healthy BM population was ACADVL, the gene encoding VLCAD, the enzyme catalyzing the first of the 4 steps in the intramitochondrial long chain FAO cycle (Figure 1B; supplemental Figure 1B). When expression was stratified, ACADVL was overexpressed in 8 of 12 quartiles (Figure 1C; supplemental Figure 1C). We next assessed VLCAD protein levels in patient-derived AML cells and normal mononuclear (MNC) samples and found that VLCAD levels were only elevated in AML cells (Figure 1D). Taken together, VLCAD is a previously unexplored intramitochondrial FAO enzyme that is overexpressed in the AML patient population.

Lentiviral-mediated short hairpin RNA knockdown of VLCAD protein was achieved in the AML cell lines TEX [F(2,8) = 75.16; P < .001; VLCAD knockdown lines are denoted TEX D6 and TEX D8; Figure 2A-2B] and AML2 (supplemental Figure 2A-B). A near-complete VLCAD knockdown significantly reduced leukemia cell proliferation [day 3: F(2,5) = 121.1; P = .0014; day 5: F(2,5) = 480.2; P = .0002; Figure 2C; AML2 cells, supplemental Figure 2C] and clonogenic growth [F(2,11) = 89.68; P < .001; Figure 2D; AML2 cells, supplemental Figure 2D].

Because VLCAD plays a critical role in FAO, we investigated the impact of protein knockdown on mitochondrial respiration of intact cells and, specifically, on palmitate (C16)-supported respiration through the ETF (Figure 1A). Basal respiration of intact cells is supported by substrates supplied in media and is tightly coupled to ATP production; maximal respiration is the highest respiration rate following ATP synthase uncoupling (Figure 2E).²¹  Knockdown of VLCAD significantly decreased intact cell respiration [Basal F(2,8) = 9.881; P = .01; maximal F(2,8) = 24.76; P = .001; Figure 2E; AML2 cells, supplemental Figure 2E]. In permeabilized cells, ETF receives electrons from multiple dehydrogenases and passes them directly to complex III via ubiquinone, bypassing complex I and II. Inhibiting these latter complexes and supplying palmitate results in the exclusive contribution of VLCAD to ETF-supported respiration (ie, indirect measure of VLCAD activity; Figure 2F).²²  Loss of VLCAD decreased C16-supported, but not C8-supported, ETF respiration [C16: F(2,8) = 141; P < .001; Figure 2G; AML2 cells: supplemental Figure 2F; C8: F(2,8) = 0.1179; P = .89; Figure 2G; AML2 cells: supplemental Figure 2F]. These results suggest that decreased C16-supported ETF respiration is specific to the knockdown of VLCAD and not the general family of ACADs.

In response to the loss of VLCAD, PDH activity [F(2,8) = 28.7; P = .0008; Figure 3A; AML2 cells: supplemental Figure 2G] was significantly increased, and ATP levels were decreased [F(2,8) = 50.15, P < .001; Figure 3B; AML2 cells, supplemental Figure 2H]. However, mitochondrial mass of the knockdown cells was unchanged, confirming that the observed mitochondrial functional changes (eg, ETF respiration, PDH, and ATP) were not due to a loss of mitochondria numbers [F(2,8) = 0.1281; P = .8821; Figure 3C; AML2 cells, supplemental Figure 2I]. The failure of PDH as a compensatory mechanism to maintain ATP levels further highlights the importance of VLCAD to AML mitochondrial metabolism.

To confirm the role of VLCAD in AML engraftment, the ability of knockdown cells to repopulate NSG mouse BM was examined (Figure 3D). Injection of an equal number of TEX cells with endogenous or reduced (TEX D6) VLCAD protein levels showed that loss of VLCAD significantly hindered leukemia cell engraftment and increased survival time more than twofold (P = .03, Mann-Whitney U test = 0; Figure 3E; P = .0134, log-rank test; Figure 3F). Taken together, VLCAD is critical to leukemia mitochondrial metabolism, proliferation, and engraftment; loss of VLCAD in leukemia cells prolongs mouse survival.

To identify a pharmacological inhibitor that could recapitulate the genetic loss of VLCAD (ie, no experimental or clinical-grade compound currently exists), a respirometry-based screen of an in-house chemical library was executed. A polyhydroxylated fatty alcohol with a terminal alkyne, avocadyne (AYNE; 16-heptadecyne-1,2,4-triol), was identified as the most potent small molecule capable of inhibiting C16-supported ETF respiration (Figure 4A).

To confirm a physical interaction, AML2 cells were treated with AYNE, and VLCAD was isolated by co-IP (Figure 4B). Magnetic co-IP produced a flow-through fraction, containing all protein with the exception of VLCAD, and a VLCAD-enriched fraction containing only VLCAD. Fraction purity was confirmed by immunoblotting (Figure 4C); VLCAD was isolated in lanes #3 and #4 from the flow-through fraction in lanes #1 and #2. AYNE was then quantified in the VLCAD-enriched fraction using a validated liquid chromatography–MS method and compared with a commercially available analytical standard (Figure 4D).²⁰  AYNE was detected in VLCAD-enriched fractions in a dose-dependent manner (Figure 4E), suggesting a physical interaction between VLCAD and AYNE.

Death of AML cells, as determined using a 7-aminoactinomycin D exclusion assay and flow cytometry, was observed following pharmacological inhibition of VLCAD using AYNE (IC₅₀ for TEX cells: 3.10 ± 0.12 µM; IC₅₀ for AML2 cells: 11.39 ± 0.90 µM; Figure 5A). Similarly, clonogenic growth of patient-derived AML cells was suppressed, with no effect on the normal hematopoietic population [AML cells: t(4) = 8.871; P < .001; normal cells: t(4) = 1.26; P = .25; Figure 5B]. This demonstrates that inhibition of VLCAD imparts selective death to AML cells. Using respirometry and a standard substrate-uncoupler-inhibitor-titration (SUIT) protocol (as outlined in "Materials and methods"), VLCAD inhibition suppressed basal and maximal mitochondrial respiration of intact leukemia cells [TEX cells, basal: F(2,8) = 21.53; P = .002; TEX cells, maximal: F(2,8) = 25.86; P = .001; AML2 cells, basal: F(2,8) = 6.57; P = .004; AML2 cells, maximal: F(2,8) = 36.75; P = .001; Figure 5C]. Similarly, VLCAD inhibition resulted in complete FAO inhibition, as determined by liquid scintillation counting using radiolabeled palmitate [TEX cells: F(4,14) = 27.28; P < .001; AML2 cells: F(4,14) = 48.59; P < .001; Figure 5D]. Therefore, inhibition of VLCAD decreased FAO in leukemia cells resulting in selective AML cell death.

Direct inhibition of VLCAD was confirmed [AML2 cells: F(2,8) = 14.79; P = .005; Figure 5E; TEX cells: supplemental Figure 3A], with no effect on medium chain acyl-CoA dehydrogenase (MCAD) [AML2 cells: F(2,8) = 10.8; P = .095; Figure 5E; TEX cells: supplemental Figure 3A], complex I (AML2 cells: supplemental Figure 4), or complex II (AML2 cells: supplemental Figure 4). Importantly, C16-supported ETF respiration (ie, very long chain FAO) was hindered by direct VLCAD inhibition [AML2 cells: F(2,8) = 14.97; P = .005; Figure 5F; TEX cells: supplemental Figure 3B], with no effect on C8-supported ETF respiration [ie, medium chain FAO; AML2 cells: F(2,8) = 0.158; P = .86; Figure 5F; TEX cells: supplemental Figure 3B]. This confirms that inhibition of VLCAD with AYNE recapitulates the effects of VLCAD knockdown on impairing mitochondrial respiration and decreasing FAO in leukemia cells.

Profiling of ACs, which are FAO metabolites, is a clinical screening tool that is used to identify and diagnose patients with FAO deficiencies.^26,30  Deficient VLCAD activity results in a characteristic accumulation of long chain AC, one of which is fully saturated C16, the substrate of VLCAD. As a result of the upstream blockage of long chain FAO, there is a subsequent decrease in fully saturated medium and short chain ACs.^26,30  Thus, to further confirm VLCAD inhibition, we compared AC profiles from AYNE-treated nontransformed fibroblasts with functional VLCAD and nontransformed fibroblasts from a patient with VLCAD deficiency. In these studies, cells were placed in minimal essential medium supplemented with palmitate–bovine serum albumin and carnitine (ie, to promote FAO); the spent cell media was subjected to AC profiling by tandem MS.²⁶  Treatment with AYNE resulted in a similar AC profile to fibroblasts lacking VLCAD with noted accumulations in C16 carnitine and depletions in C10, C8, C6, and C4 carnitines [C16 carnitine: F(3,16) = 52.78; P < .001; C10 carnitine: F(3,16) = 20.48; P < .001; C8 carnitine: F(3,16) = 12.10; P < .001; C6 carnitine: F(3,16) = 17.35; P < .001; C4 carnitine: F(3,16) = 31.83; P < .001) (Figure 5G).

The specificity of AYNE for VLCAD was further confirmed using cell lines with partial VLCAD knockdown (denoted as D5, D7, and D9; supplemental Figure 5A). D5 and D9 showed moderate, but significant, VLCAD knockdown, whereas D7 did not have any change in VLCAD compared to the parental line. Interestingly, partial VLCAD knockdown coincided with reduced C16-supported ETF respiration in D5 and D9 (supplemental Figure 5C), which was further suppressed with the addition of AYNE (supplemental Figure 5D). Consistent with VLCAD specificity, FAO inhibitors that do not target VLCAD (eg, ETO and ranolazine [RAN]) did not have any effect (supplemental Figure 5D). Taken together, we confirm that AYNE is an FAO inhibitor that specifically targets VLCAD.

Sodium heptanoate is the salt of the active metabolite of triheptanoin, a glycerol esterified to 3 seven-carbon (C7) fatty acids used to treat long chain FAO disorders.³¹  To confirm that AYNE causes a blockade of long chain FAO, we coincubated cells with C7, a medium chain fatty acid. Medium chain fatty acids bypass long chain mitochondrial FAO enzymes (ie, VLCAD) to be broken down by medium and short chain FAO enzymes, forming TCA cycle–bound acetyl-CoA and propionyl-CoA.³¹  As such, we reasoned that treatment with C7 would bypass VLCAD inhibition and rescue AML cell viability. Indeed, cotreatment with C7 increased the 50% inhibitory concentration twofold, reducing AYNE-imparted cell toxicity in TEX cells [t(4) = 14.59; P < .001; Figure 5H] and AML2 cells [t(4) = 12.25; P < .001; Figure 5I].

To characterize the effects of inhibiting VLCAD, we assessed viability and ATP levels in primary AML and normal populations following CD34⁺ cell isolation (supplemental Figure 6). VLCAD inhibition did not have any effect on CD34⁺ normal cell viability [t(4) = 0.2992; P = .78], but it was cytotoxic to bulk primary AML cells [t(4) = 6.793; P = .002] and CD34⁺ primary AML cells [t(6) = 6.047; P < .001] (Figure 6A). Preceding AML cell death, ATP levels were depleted; ATP levels were not depleted in normal CD34⁺ cells coinciding with cell survival [AML #18 cells: t(6) = 5.609; P = .001; AML #20 cells: t(4) = 52.23; P < .001; AML #21 cells: t(4) = 35.41, P < .001; normal CD34⁺ #1: t(4) = 0.239, P = .82; #2: t(4) = 0.8484, P = .44; #3; t(4) = 0.2689; P = .8) (Figure 6B).

FAO supplies fat-derived acetyl-CoA to the TCA cycle.⁹  To assess the contribution of FAO to TCA metabolism, we conducted stable isotope tracing experiments with a uniformly labeled ¹³C₁₆ palmitate tracer in leukemia cell lines (TEX and AML2 cells; supplemental Figure 7A), patient-derived AML cells (Figure 6C-D), and normal MNC samples (Figure 6D). In TEX and AML2, treatment with AYNE and ETO, a positive control for FAO inhibition, as well as the loss of VLCAD induced an accumulation of ¹³C₁₆ palmitate and a depletion of ¹³C₂ acetyl-carnitine (below the detection limit), confirming FAO inhibition (supplemental Figure 7A). In patient-derived AML cells, AYNE also inhibited FAO, with a similar accumulation in ¹³C₁₆ palmitate and a decrease in ¹³C₂ acetyl-carnitine (Figure 6C). Similar to our AC profiling with the nontransformed fibroblasts (Figure 5G), we noted significant decreases in medium and short chain ACs derived from the ¹³C₁₆ palmitate tracer, further confirming long chain FAO inhibition (Figure 6C). Following VLCAD knockdown or pharmacological inhibition of VLCAD in AML lines (supplemental Figure 7A) or primary AML (Figure 6D), we observed significantly decreased enrichment of fat-derived carbon into TCA metabolites. At the concentration (50 µM) and time point (12 hours) that suppressed AML mitochondrial metabolism, normal MNC samples did not alter the enrichment of fat-derived ¹³C atoms into FAO or TCA metabolites.

To assess potential compensatory glycolysis in response to FAO inhibition, the enrichment of glucose-derived ¹³C atoms from a ¹³C₆ glucose tracer into pyruvate, lactate, and citrate pools were quantified in AML2 cells (supplemental Figure 7B), primary AML, and normal MNC samples (Figure 6E). Pyruvate, the terminal metabolite of glycolysis, stands at the junction between lactate production and mitochondrial oxidation.³²  Across AML2 and primary AML cells, FAO inhibition did not alter levels of ¹³C₃ pyruvate, confirming no alteration of glycolysis. Lactate production is also viewed as a measure of glycolytic activity.^6,33-35  VLCAD knockdown in AML2 cells resulted in a slight increase in ¹³C₃ lactate (supplemental Figure 7B). In contrast, pharmacological inhibition of FAO did not alter levels of ¹³C₃ lactate in AML2 (supplemental Figure 7B) or primary AML cells (Figure 6E). Compared to the vehicle control, there was no significant change in ¹³C₃ pyruvate or ¹³C₃ lactate, which suggests no compensatory upregulation of glycolysis or lactate production in response to pharmacological VLCAD inhibition.

Enrichment of glucose-derived carbons into the ¹³C₂ citrate isotopomer is used as an analog to measure PDH activity.^36,37  Knockdown or pharmacological inhibition of VLCAD in AML2 cells (supplemental Figure 7B) and primary AML cells (Figure 6E) resulted in significantly increased ¹³C₂ citrate, in agreement with increased PDH activity as confirmed earlier with a commercial kit (Figure 3A; supplemental Figure 2G). In normal MNCs receiving ¹³C₁₆ palmitate tracer, there was no significant change in ¹³C₂ TCA metabolites, coinciding with no FAO inhibition at this concentration (50 µM) and time point (12 hours) (Figure 6D). Normal MNCs also did not alter the levels of ¹³C₃ pyruvate, ¹³C₃ lactate, or ¹³C₂ citrate following incubation with the ¹³C₆ glucose tracer (Figure 6E). Statistical values for all tracer experiments (Figure 6C-E; supplemental Figure 7) are presented in supplemental Figure 11. Together, this confirms that basal levels of glycolysis are not upregulated in response to VLCAD inhibition; however, compensatory PDH activation occurs but is ultimately insufficient to prevent death of leukemia cell lines or patient-derived AML cells. Thus, the selective nature of VLCAD inhibition, in exclusively impairing altered AML cell metabolism, results in the death of AML but not normal cells.

Pharmacological inhibition of VLCAD was tested in vivo in a functionally defined subset of primitive human AML and normal hematopoietic cell populations using a mouse engraftment assay (Figure 7A). Normal HSCs were injected into the tail vein of NSG mice. After 1 week, intraperitoneal injections of AYNE (300 mg/kg per week) were administered for 6 weeks (U = 1; P = .06; Figure 7B). There was no significant effect on the repopulation potential of normal human myeloid cells (CD33⁺/CD45⁺) in mouse BM compared with vehicle control, as measured by flow cytometry (Figure 7B).

Next, patient-derived AML cells were injected into the tail vein of NSG mice, and AYNE or a vehicle control was administered after 7 or 28 days. Inhibition of VLCAD significantly reduced the repopulation potential of leukemia cells compared with vehicle control–treated mice (U = 0; P = .03; Figure 7C; U = 0; P = .02; Figure 7D; U = 0; P = .005; Figure 7E). Analysis of complete blood counts, bilirubin, serological markers (ie, alkaline phosphatase and creatine kinase, which are markers of kidney function and liver function, respectively), and mouse body weights did not reveal any differences between the control and AYNE treatment groups (supplemental Figure 8). Together, these demonstrate the selective activity of VLCAD inhibition against primitive leukemia cells in established preclinical models.

To confirm VLCAD inhibition in vivo, patient-derived cells from AML #28 were injected into NSG mice and allowed to engraft for 8 weeks (Figure 7F). Mice were then given a bolus dose of AYNE (150 or 200 mg/kg) or a vehicle control. After 24 hours, femoral BM was harvested (Figure 7G), and human myeloid cells (CD33⁺/CD45⁺) were isolated and purified (Figure 7H). C16-supported ETF respiration, as an indirect measure of VLCAD activity, and a cellular thermal shift assay, as a measure of drug-target binding, were performed. Prior to the cellular thermal shift assay with primary leukemia cells, the melting point of VLCAD was determined to be 60°C with AML2 cells (supplemental Figure 9). Isolated engrafted human cells (>95% pure) from mice treated with AYNE showed inhibition of ETF respiration [F(2,7) = 13.29; P = .01; Figure 7I] and a shift in VLCAD melting temperature, which confirmed inhibition of VLCAD activity and an AYNE-VLCAD interaction, respectively [F(2,8) = 14.95; P = .005; Figure 7J]. Taken together, these studies confirm in vivo efficacy against leukemia populations, while sparing hematopoietic cells, as well as pharmacological inhibition of VLCAD.

Compared with the hematopoietic population, leukemia cells have higher rates of FAO that contribute to dysregulated metabolism, resulting in altered rates of proliferation and increased survival during chemotherapy.^2,3  Although FAO is potentially a novel pharmacological target in AML, previous reports have focused on inhibiting the entry of fats into the cytoplasm via CD36⁵  or into the mitochondria via CPT1.^2,3  In this study, we investigated intramitochondrial FAO enzymes to determine a novel anti-AML target. Database profiling confirmed that ACADVL, the gene encoding the enzyme catalyzing the first intramitochondrial step of long chain FAO (ie, VLCAD),⁹  is overexpressed in AML. Although VLCAD in AML has not been explored previously, we show here that it plays a critical role in FAO, downstream oxidative metabolism, and overall AML cell survival. Indeed, protein knockdown or pharmacological inhibition of VLCAD resulted in selective toxicity of AML cells in vitro and in vivo. Thus, this study highlights the importance of VLCAD in AML and validates a novel metabolic target of this disease.

VLCAD is a homodimer that is anchored to the inner mitochondrial membrane and interacts with the ETF, an electron acceptor protein equipped with a FAD cofactor.³⁸  Long chain acyl substrates are oxidized by VLCAD, which introduces a double bond across carbons 2 to 3 while reducing the FAD cofactor.³⁸  Through a series of redox reactions, electrons are transferred by ETF dehydrogenase from ETF to ubiquinone and, finally, to complex III.⁹  In this study, we identified a crucial role for VLCAD in altered leukemia cell metabolism. In leukemia cell lines, loss of VLCAD activity inhibited C16-supported ETF respiration to hinder mitochondrial FAO and decrease fat-derived ¹³C atom enrichment into the TCA cycle, while triggering insufficient compensatory PDH activity. This resulted in reduced proliferation, clonogenic potential, and engraftment, confirming that VLCAD is critical to leukemia cell growth and repopulation potential.

VLCAD deficiency is a rare human disorder that is defined by a mutation in the ACADVL gene causing decreased VLCAD enzymatic activity.³⁹  A range of clinical symptoms with varied severity are due to the depletion of ATP and TCA metabolites.^31,39  Adult VLCAD-deficient patients manage the severity of their symptoms by avoiding a diet high in long chain fats, stress, cold exposure, fasting, and strenuous exercise.³⁹  Although the severe form of VLCAD deficiency can present with cardiomyopathy and rhabdomyolysis, hypoketotic hypoglycemia is also a typical symptom.³⁹  Xiong et al demonstrated cardiomyopathy in murine cardiomyocytes upon conditional ACADVL deletion.⁴⁰  In a previously published pharmacokinetic study by our group, we demonstrated that AYNE was detected in murine cardiac tissue.⁴¹  Although studies reported here showed that 6 weeks of repeated AYNE therapy was well tolerated, direct cardiotoxicity needs be confirmed in future murine studies.

Pharmacological inhibition of FAO has been applied with varying success in preclinical AML studies.^2,3,5,7  Clinical FAO inhibitors exist, most of which target components of the mitochondrial trifunctional protein^12,16 ; however, no experimental or clinical VLCAD-specific inhibitors are available. In this study, AYNE recapitulated the effects of VLCAD knockdown, resulting in selective toxicity of primary AML cells in vitro and in vivo. In primary AML cells, chemical inhibition of VLCAD blocked FAO to decrease enrichment of ¹³C atoms from the ¹³C₁₆ palmitate tracer into TCA metabolites. FAO inhibition did not significantly upregulate glycolysis with no changes in glucose-derived carbon enrichment into pyruvate and lactate. Despite no change in glycolysis, primary leukemia cells increased PDH activity that was ultimately insufficient to prevent ATP depletion or cell death. At the dose and time point that eliminated primary AML cells, donor-derived normal MNCs were spared from FAO inhibition, resulting in no change in ATP levels or viability. The divergent metabolic fates observed with primary AML and normal MNCs demonstrate a therapeutic window for VLCAD inhibition in AML. Together, this study highlights VLCAD as a novel drug target that, when inhibited, results in selective leukemia cell death due to AML metabolic inflexibility.

Glucose,⁴²  fatty acid,^3,43  and amino acid³⁵  oxidation are rewired to support the energy demands of rapidly proliferating leukemia cells. Specific leukemia subpopulations harboring these metabolic dependencies are unable to upregulate alternative catabolic processes under metabolic stress. In contrast, normal cells switch to compensatory metabolism for survival, further demonstrating the therapeutic potential of targeting altered AML metabolism.⁶  In agreement with numerous studies that used pharmacological inhibitors of mitochondrial proteins (ie, BCL2, complex I) or substrate withdrawal,^6,33-35  inhibition or genetic knockdown of VLCAD results in no compensatory glycolysis, leading to leukemia cells with greatly reduced proliferation and engraftment potential. Downstream from glycolysis, AYNE and VLCAD knockdown increased ¹³C₂ citrate levels from the ¹³C₆ glucose tracer, suggesting increased PDH activity. Previously, increased PDH activity supported leukemia mitochondrial respiration and ATP production, resulting in decreased proliferation and engraftment potential.^44,45  In agreement with these previous observations, the increase in PDH activity could explain the decreased proliferation and engraftment observed upon VLCAD knockdown. Despite demonstrating significantly increased PDH activity, primary AML samples still showed significant cell death, confirming that this compensatory increase was insufficient to rescue the leukemia population.

Interestingly, although unable to use other fuel substrates, depletion of ATP and TCA metabolites can be averted by supplementing starved leukemia cells with a downstream product of the inhibited pathway.^27,46  For example, inhibition of glutaminolysis by CB-839 or glutamine starvation could be rescued with exogenous α-ketoglutarate²⁷  or oxaloacetate,⁴⁶  averting ATP depletion and apoptosis. Similarly, medium chain triglyceride oils are provided to patients with VLCAD deficiency with varied clinical success. In our study, supplying leukemia cells with a medium chain fat partially attenuated AYNE-induced cell death by bypassing the blockage at long chain FAO.³¹  Taken together, alterations in leukemia cell metabolism caused by substrate inhibition (ie, fatty acids^3,43  or amino acids³⁵ ), reductions in TCA cycle activity, or perturbations of OXPHOS^6,8  resulted in selective cell death.

AML is a highly heterogeneous cancer with diverse patient cytogenetics. Although this study confirmed FAO inhibition and reduced contributions of fat-derived carbon to the TCA cycle in primary AML samples, additional samples are required to make a conclusive statement about whether VLCAD inhibition, or, more generally, FAO inhibition will be particularly efficacious in a specific AML patient subpopulation. Therefore, future studies will correlate primary AML sample sensitivity with VLCAD activity.

In summary, VLCAD plays a central role in AML metabolism. Decreased VLCAD activity reduced AML cell viability, which resulted in an insufficient metabolic response that was mechanistically defined by no change in glycolysis, insufficient compensatory PDH activity, a decreased contribution of fat-derived carbons to the TCA cycle, and ultimately, ATP depletion. These results highlight, for the first time, VLCAD as a novel metabolic regulator of AML cell viability.

The authors thank John Dick for the generous gift of TEX cells.

This work was supported by grants from the Leukemia Research Foundation (P.A.S.), Cancer Research Society (P.A.S.), National Sciences and Engineering Research Council (P.A.S.), the Ontario Institute for Cancer Research (P.A.S.), and by National Institutes of Health, National Institute of Diabetes and Digestive and Kidney Diseases grant R01-DK78755 (J.V.).

Contribution: P.A.S. conceived the project and designed experiments; M.T. designed and performed experiments; A.R. and P.J. conducted in vivo experiments; N.A., R.W.S., A.D., and E.B. conducted UHPLC and/or MS analyses; M.D.M., D.A.H., and K.H. provided primary human samples; K.A.R. and T.A.A. conducted liquid scintillation counting; A.-W.M. and J.V. conducted AC profiling experiments; and A.D.S. and J.V. provided technical advice.

Conflict-of-interest disclosure: A.D.S. has received honoraria or consulting fees from Novartis, Jazz Pharmaceuticals, Otsuka, and Takeda Pharmaceuticals and research support from Medivir AB and Takeda. A.D.S. owns stock in AbbVie Pharmaceuticals. A.D.S. and P.A.S. are named on a patent application for the use of avocado compounds for the treatment of leukemia; however, this patent was not pursued passed the jurisdictional phase. P.A.S. has received honoraria from the Hass Avocado Board. The remaining authors declare no competing financial interests.

Correspondence: Paul A. Spagnuolo, Department of Food Science, University of Guelph, 50 Stone Rd, Guelph, ON N1G 2W1, Canada; e-mail: paul.spagnuolo@uoguelph.ca.