Glutaredoxin 2 is essential for AML survival through mitochondrial permeability transition pore regulation

Key Points

• •The deglutathionylating enzyme, GLRX2, is essential for the survival of primary human AML.
• •GLRX2 regulates the activity of the mPTP through the deglutathionylation of ATP5PO.

Visual Abstract

Abstract

Patients with acute myeloid leukemia (AML) have a poor 5-year survival rate, highlighting the need for the identification of new approaches to target this disease. AML is highly dependent on glutathione (GSH) metabolism for survival. Although the metabolic role of GSH is well characterized in AML, the contribution of protein glutathionylation, a reversible modification that protects protein thiols from oxidative damage, remains largely unexplored. Therefore, we sought to elucidate the role of protein glutathionylation in AML pathogenesis. Here, we demonstrate that protein glutathionylation is essential for AML cell survival. Specifically, the loss of glutaredoxin 2 (GLRX2), an enzyme that removes GSH modifications, resulted in selective primary AML cell death while sparing normal human hematopoietic stem and progenitor cells. Unbiased proteomic analysis revealed increased mitochondrial protein glutathionylation upon GLRX2 depletion, accompanied by mitochondrial dysfunction, including impaired oxidative phosphorylation, reduced mitochondrial membrane potential, and increased opening of the mitochondrial permeability transition pore (mPTP). Further investigation identified adenosine triphosphate synthase subunit O (ATP5PO), a key regulator of mPTP opening and a component of the ATP synthase complex, as a critical GLRX2 target. Disruption of ATP5PO glutathionylation partially restored mPTP function and rescued AML cell viability after GLRX2 depletion. Moreover, both genetic and pharmacological inhibition of mPTP opening restored the leukemic potential of primary AML specimens in the absence of GLRX2. By disrupting glutathionylation-dependent mitochondrial homeostasis, this study reveals a novel vulnerability in AML that could inform future therapeutic strategies.

Their heightened dependence on mitochondrial energy production makes acute myeloid leukemia (AML) cells particularly vulnerable to loss of mitochondrial function. Opening of the mitochondrial permeability transition pore (mPTP) leads to loss of mitochondrial membrane potential and apoptosis. Ling and colleagues reveal that the glutathione-dependent oxidoreductase glutaredoxin 2 (GLRX2) is upregulated in AML cells and that, through its deglutathionylating action on a component of electron transport chain complex V, this increases closure of the mPTP, linked to AML resistance to chemotherapy. Loss of GLRX2 function results in sustained opening of the mPTP, identifying it as a potential future target for AML therapy.

Introduction

Acute myeloid leukemia (AML) is a hematological malignancy characterized by the clonal proliferation of undifferentiated myeloid precursors within the bone marrow.¹ The 5-year survival rate in AML is only 30%.¹ Although advancements have been made in the treatment of AML including the development of novel therapies,² most patients will relapse and succumb to their disease. This underscores the urgent need for innovative therapeutic approaches to more effectively eliminate AML cells and improve patient outcomes.

Glutathione (GSH) is an antioxidant ubiquitously found in mammalian tissues and plays a critical role in scavenging reactive oxygen species (ROS).^3,4 Rewiring of antioxidant pathways has been reported in many different cancer types, enabling them to withstand oxidative stress.^{5, 6, 7} Although the role of GSH homeostasis in cancer is complex, studies have consistently demonstrated that AML cells are dependent on GSH for survival.^{8, 9, 10, 11} Despite success inhibiting GSH biosynthesis in preclinical models, the translation of these strategies into clinical practice has been impeded by toxicity and pharmacokinetic limitations.¹² To address these challenges, we sought to interrogate alternative mechanisms regulated by GSH as potential targets in AML.

Glutathionylation is a posttranslational modification that involves the conjugation of GSH to reactive protein cysteine thiol derivates to prevent irreversible damage from further oxidant exposure.^{13, 14, 15} This oxidative event can be induced by normal ROS signaling or by oxidative stress,¹⁶ commonly observed in cancer, including AML.¹⁷ When proteins are glutathionylated, the activity of most enzymes is reduced.^{13, 14, 15} Therefore, once the oxidative event is resolved, it is critical for cells to reverse protein glutathionylation and restore proteins to their deglutathionylated state. The process of deglutathionylation requires enzyme catalysis, which is, in part, mediated by the glutaredoxins (GLRXs), primarily GLRX and GLRX2.¹⁸ Here, we examine the consequences of disrupting deglutathionylation regulatory proteins in AML.

GLRX2 is a dithiol oxidoreductase¹⁸ that has been implicated in cell survival and resistance to therapy in cell lines and animal models.^{19, 20, 21, 22, 23} Mechanistically, GLRX2 has been shown to be important for transcriptional regulation in glioma cells,¹⁹ MAPK signaling in osteoclastogenesis,²⁰ and migration in HeLa cells.²¹ GLRX2 has also been shown to protect HeLa cells from chemotherapy-induced oxidation.^22,23 In cardiac muscle, liver, and macrophages, GLRX2 has been shown to regulate mitochondrial function,^{24, 25, 26, 27} but the role of GLRX2 in mitochondrial biology has, to our knowledge, never been described in cancer. Furthermore, GLRX2 has yet to be functionally examined in a primary human tumor, limiting our ability to determine the potential of GLRX2 as a novel therapeutic target in cancer.

Here, we show that GLRX2 is essential for the survival of primary human AML cells. Furthermore, we demonstrate that an essential GLRX2 deglutathionylation target is adenosine triphosphate synthase subunit O (ATP5PO), which is involved in the regulation of mitochondrial permeability transition pore (mPTP) formation. We show that depletion of GLRX2 induces mPTP opening and results in loss of mitochondrial integrity in AML cell lines and primary human AML cells. Overall, these data demonstrate that GLRX2 represents an essential protein in AML through the regulation of mitochondrial redox biology.

Methods

Primary specimens

Patient-derived AML cells were obtained from donors who gave informed consent under the Princess Margaret Leukemia Tissue Bank, the University of Colorado tissue procurement protocol, or under the Cincinnati Children’s Hospital Tissue protocol. Human bone marrow mononuclear cells and human bone marrow CD34⁺ progenitor cells were purchased from Lonza. See supplemental Table 1, available on the Blood website, for additional details on the human AML specimens.

Other methods

Cell culture methods, colony assays, xenograft studies, flow analysis, lentivirus production, gene expression analysis, quantitative reverse transcription polymerase chain reaction (qRT-PCR), proliferation assays, biotinylated GSH ethyl ester (BioGEE) switch assay, immunoprecipitation and immunoblotting, small interfering RNA (siRNA) transduction, Seahorse analysis, translocase of outer mitochondrial membrane 20 staining, and proteomics and metabolomics analysis were performed as previously described^{9,28, 29, 30, 31, 32, 33} and detailed methods are available in the supplemental Methods.

Studies were performed in accordance with animal use protocols at University Health Network and institutional animal care and use committee approval at Cincinnati Children's Hospital Medical Center. Patient specimens were obtained from biobanks at Princess Margaret Hospital, Cincinnati Children's Hospital, and the University of Colorado.

Results

GLRX2 is essential for AML cell survival

The human GLRX family consists of 4 members encoded by the GLRX, GLRX2, GLRX3, and GLRX5 genes. We first assessed the relative expression of these genes encoding these proteins in leukemia stem cells and normal hematopoietic stem and progenitor cells (HSPCs). Single-cell RNA sequencing data from van Galen et al³⁴ show that the expression of GLRX2 was significantly higher in the leukemia stem cell population than in HSPCs (Figure 1A), whereas GLRX, GLRX3, and GLRX5 showed no significant difference (supplemental Figure 1A-C). We further assessed the expression of GLRXs in bulk AML using the Leucegene (GSE481173), Beat-AML, TCGA, and He et al^{35, 36, 37, 38} RNA-sequencing data sets. In 3 of 4 transcriptomic data sets, GLRX2 expression was upregulated in bulk AML compared with normal specimens (Figure 1B), whereas no consistent pattern of expression for GLRX, GLRX3, and GLRX5 was observed (supplemental Figure 1D-F). Additionally, GLRX2 protein was elevated in primary AML cells compared with normal cells (Figure 1C). We then determined whether GLRX2 expression was enriched in specific AML subtypes. Although GLRX2 expression is elevated in some mutation or morphological subtypes, the difference did not follow a uniform trend, indicating that GLRX2 upregulation is not specifically associated with any single genetic abnormality or morphological feature (supplemental Figure 1G-I). Altogether, data show that GLRX2 gene expression is elevated in AML compared with normal specimens across a wide range of patients with AML.

Figure 1.

GLRX2 KD reduces the growth of AML cells in vitro. (A) GLRX2 expression in LSCs and normal HSPCs. Single-cell RNA sequencing data from van Galen et al.³⁴ Horizontal lines indicate the interquartile range; unpaired t test, ∗∗∗P < .001. (B) GLRX2 expression in bulk AML cells compared with normal hematopoietic cells. Data from Leucegene (GSE48173), TCGA, Beat-AML, and He et al^{35, 36, 37, 38} data sets. Horizontal lines indicate the interquartile range; unpaired t test, ∗P < .05; ∗∗P < .01; ∗∗∗∗P < .0001. (C) GLRX2 protein expression in normal peripheral mobilized blood cells and primary AML samples (AML1-7). (D) Experimental design of GLRX2 KD experiments. MOLM13, PL21, and MV411 were assessed with viability assay or plated for proliferation assay, and colony-forming unit assay 96 hours after transduction. Created with BioRender. Ling C. (2025) BioRender.com/k96q695. (E) Representative immunoblot measuring protein expression of GLRX2 in MOLM13, PL21, and MV411 96 hours after lentivirus transduction delivering control short hairpin RNA (shRNA) or GLRX2 targeting shRNAs (shRNA1 and shRNA2; n = 3-4 biological replicates). (F) Representative experiment of viable cell counts of MOLM13, PL21, and MV411 after GLRX2 KD. Statistical significance was determined from 3 biological replicates calculating area under the curve. Mean ± standard deviation (SD); ordinary 1-way analysis of variance (ANOVA), ∗P < .05; ∗∗P < .01; ∗∗∗∗P < .0001. (G) Colony-forming ability of MOLM13, PL21, and MV411 upon GLRX2 KD. Mean ± SD (n = 3 biological replicates); ordinary 1-way ANOVA, ∗P < .05; ∗∗P < .01; ∗∗∗∗P < .0001. (H) Representative staining of annexin-V and DAPI in control or GLRX2-KD MOLM13 cells. (I) Bar graphs show early apoptotic (annexin-V⁺/DAPI⁻) and late apoptotic (annexin-V⁺/DAPI⁺) MOLM13, PL21, and MV411 cells. Mean ± SD (n = 3 biological replicates); total apoptotic (annexin-V⁺) population compared with ordinary 1-way ANOVA, ∗∗P < .01; ∗∗∗P < .001. GAPDH, glyceraldehyde-3-phosphate dehydrogenase; LSCs, leukemia stem cells; ns, not significant; shControl, short hairpin control.

To determine whether GLRX2 loss targets AML cells, we transduced 3 AML cell lines (MOLM13, PL21, and MV411) with lentiviral vectors expressing a control sequence or 1 of 2 short hairpin RNA sequences targeting GLRX2 (Figure 1D). GLRX2 knockdown (KD) (Figure 1E) significantly impaired the proliferation and clonogenic growth of the AML cells (Figure 1F-G). Consistently, CRISPR/CRISPR–associated protein 9–mediated GLRX2 knockout in MOLM13 and MV411 cells reduced cell proliferation and colony formation (supplemental Figure 1J-L). Next, we measured cell death using the apoptotic markers annexin-V and DAPI (4′,6-diamidino-2-phenylindole) after GLRX2 depletion (Figure 1H). The apoptotic populations were significantly increased in all cell lines upon GLRX2 KD (Figure 1I). These data demonstrate that GLRX2 depletion induces cell death in AML cells.

Next, we sought to investigate whether GLRXs are required for the growth of primary AML specimens. First, we achieved efficient siRNA knockdown of each GLRX gene (supplemental Figure 2A) in 3 primary human AML specimens and measured colony-forming potential. Consistent with the gene expression data (Figure 1A-B; supplemental Figure 1A-F), we observed that transient loss of GLRX2, but not other GLRXs, reduced colony-forming capacity of AML specimens (Figure 2A). Next, we knocked down GLRX2 in 10 primary AML specimens. Because we did not observe an enrichment of GLRX2 expression in any specific AML subtype, we used primary patient specimens that covered a wide range of genetic abnormalities (eg, FLT3, NPM1, and IDH1; supplemental Table 1). Twenty-four hours after transfection with siRNA, cells were seeded in methylcellulose medium and assessed for colony-forming ability (Figure 2B). GLRX2 mRNA expression and, when possible, protein expression was measured 48 hours after transfection to confirm knockdown in the primary AML specimens (Figure 2C; supplemental Figure 2B). GLRX2 KD reduced the colony-forming potential of all AML specimens evaluated (Figure 2D). Importantly, cell viability before colony-forming unit assay was not significantly affected by GLRX2 targeting, indicating that the observed differences in colony-forming capacity are unlikely to be caused by the differences in cell death at the time of plating (supplemental Figure 2C). To determine whether GLRX2 is essential for normal HSPCs, we performed the same experiment using normal bone marrow (NBM) cells from healthy donors. Upon GLRX2 KD (supplemental Figure 2D), colony-forming potential was not altered in NBM cells (Figure 2E).

Figure 2.

GLRX2 KD selectively impairs primary AML specimens in vitro and in vivo. (A) CFU assay of AML8, AML15, and AML16 with individual GLRX KD. Mean± SD (n = 3 biological replicates); repeated measures (RM) 1-way ANOVA, ∗∗P < .01. (B) Experimental design of the clonogenic assay. Ten primary AML and 1 NBM specimens were transfected with scramble or GLRX2 targeting siRNA and plated in MethoCult medium 24 hours after transfection. Created with BioRender. Ling C. (2025) BioRender.com/k96q695. (C) Expression of GLRX2 determined by western blot in 3 bulk primary AML specimens (AML13-15) 48 hours after transfection. (D-E) Colony-forming potential of 10 bulk AML samples (AML8-17) and 1 NBM sample. Mean ± SD; unpaired t test, ∗∗P < .01; ∗∗∗P < .001; ∗∗∗∗P < .0001. (F) Human engraftment evaluation in the femur of humanized mice. Three primary AML and 3 NBM specimens were injected into NSG-SGM3 mice through tail vein injection 24 hours after transfection with scramble or GLRX2 targeting siRNA. Eight to 10 weeks after injection, bone marrow cells were collected from femurs to assess disease burden and differentiation. One representative mouse from AML and NBM specimens is shown. Created with BioRender. Ling C. (2025) BioRender.com/k96q695. (G) Engraftment of 3 primary AML specimens (AML18-20) in NSG-SGM3 mice after GLRX2 KD. Each point represents a single mouse. Mean ± SD; unpaired t test, ∗P < .05; ∗∗∗∗P < .0001. (H) Engraftment of 3 primary NBM specimens in NSG-SGM3 mice after GLRX2 KD. Each point represents a single mouse. Mean ± SD; unpaired t test. CFU, colony-forming unit; FSC-A, forward scatter area; GAPDH, glyceraldehyde-3-phosphate dehydrogenase; ns, not significant; PE, phycoerythrin; PECY7, phycoerythrin-cyanine 7; Scr, scrambled siRNA.

To examine the role of GLRX2 in the leukemia stem and progenitor function, we performed GLRX2 KD in 3 primary AML cells and HSPCs enriched from 3 healthy donor specimens and transplanted them into NOD-scid IL2Rg^null-3/GM/SF (NSG-SGM3) mice (Figure 2F; supplemental Figure 2E-G). GLRX2 KD resulted in a significant decrease in leukemic burden, suggesting that GLRX2 is required for leukemia initiation and leukemia stem and progenitor function in AML (Figure 2G; supplemental Figure 2H). In contrast, NBM cells engrafted similarly to controls upon GLRX2 KD (Figure 2H; supplemental Figure 2I). Moreover, lineage distribution within the NBM grafts was not affected, overall suggesting that transient depletion of GLRX2 is better tolerated by NBM than leukemia cells (supplemental Figure 2J-L).

GLRX2 regulates the deglutathionylation of mitochondrial proteins

To determine the mechanism(s) by which GLRX2 is essential for AML cell survival, we sought to identify GLRX2 deglutathionlyation targets. To identify proteins with increased glutathionylation upon GLRX2 loss, we used the BioGEE assay established by Sullivan et al.³⁹ In this assay, endogenous glutathionylation is removed and replaced with a BioGEE moiety.³⁹ Biotin-tagged proteins can then be enriched and visualized by measuring streptavidin binding on a western blot or identified by mass spectrometry (Figure 3A). GEE, a nonbiotin conjugation GSH analog, was used as a control for nonspecific binding to the streptavidin bead in these experiments. In AML cells, BioGEE-incorporated samples (Figure 3B, lane 1) displayed a strong enrichment of BioGEE-bound proteins when probed with a streptavidin IRDye. This signal was reduced in BioGEE-incorporated samples after treatment with disulfide reducing agent, dithiothreitol (Figure 3B, lane 2). Furthermore, almost no signal was detected when AML cells were treated with the control GEE (Figure 3B, lane 3). Proteins were only partially labeled with BioGEE when the reaction lacked the glutathionylation catalyst (ie, prooxidant) diamide (Figure 3B, lane 4). Overall, this strategy allowed us to enrich endogenous putatively glutathionylated proteins from AML cells.

Figure 3.

Identification of glutathionylated proteins. (A) Schematic of methods for BioGEE switch assay. Created with BioRender. Ling C. (2025) BioRender.com/a9h05yv. (B) Western blot probed with streptavidin IRDye to identify glutathionylated proteins enriched from BioGEE switch assay in MOLM13 cell line (n = 1). (C) Summary of 49 hits grouped by function identified from control and GLRX2 KD MOLM13 cells detected by liquid chromatography–tandem mass spectrometry–based BioGEE switch assay. Significant hits were defined as proteins with <5% false discovery rate (FDR) and a greater than fourfold change upon GLRX2 KD compared with nontargeting control shRNA–expressing cells. (D) Selected gene ontology terms with <5% FDR and −log₁₀-adjusted P value >3 in the GLRX2-regulated proteins. DTT, dithiothreitol; HS, sulfhydryl group; LC-MS/MS, liquid chromatography-tandem mass spectrometry; NEM, N-ethylmaleimide; NEM-S, N-ethylmaleimide bound to sulfur atom; SH, free thiol; SSG-Bio, biotinylated GSH ethyl ester; SSG, glutathionylation; TCA, tricarboxylic acid cycle.

Next, we performed the BioGEE switch assay coupled with liquid chromatography–tandem mass spectrometry to identify glutathionylated proteins. Proteins enriched in the BioGEE condition were compared with the GEE-treated controls in MOLM13, PL21, and MV411 AML cells. Putative glutathionylated proteins were defined as those with a BioGEE/GEE spectral count ratio of ≥2 (log₂ > 1) in at least 2 cell lines. Glutathionylated hits (n = 676) identified in each of the 3 cell lines were consistent between biological replicates (R² = 0.92, 0.93, and 0.94 for MOLM13, PL21, and MV411, respectively; supplemental Figure 3A-C). Furthermore, we detected several previously characterized glutathionylated proteins, such as CASP3,⁴⁰ PRDX6,⁴¹ and NPM1.⁴² Of our putatively glutathionylated proteins, ∼53% (358/676 proteins) overlap with those in previous studies (supplemental Figure 3D).^{43, 44, 45} Nonoverlapping proteins may result from different cysteine oxidation sensitivity due to differences in cell type or stressors. Together, this proteomic approach allowed us to unbiasedly identify putative glutathionylated proteins in AML cells.

To define the GLRX2-regulated glutathionylome, we conducted the BioGEE switch assay on MOLM13 cells with inducible GLRX2 KD. We used an inducible GLRX2 KD system to ensure that any changes detected in protein glutathionylation would not be caused by the cell death observed upon GLRX2 depletion (supplemental Figure 3E-F). To minimize false positives and highlight only the most robustly regulated targets, we applied a fourfold change threshold in spectral enrichment with a false discovery rate of <5% and identified 49 proteins with significant increase in glutathionylation after GLRX2 KD (Figure 3C). Importantly, we identified NDUFS1, a bona fide GLRX2 substrate.²⁴ Annotated enrichment analysis revealed targets to be localized in the mitochondria, nucleus, cytoplasm, and endoplasmic reticulum (supplemental Figure 3G). Furthermore, these proteins are responsible for the regulation of cellular metabolism, RNA binding, transcription and translation, protein homeostasis, protein import, and signaling (Figure 3C). Although most proteins observed had a significant increase in glutathionylation, 2 proteins had a significant decrease (Figure 3C), indicating that GLRX2 in some instances may be involved in conjugating GSH to a protein thiol, consistent with previous findings in bovine heart mitochondria.⁴⁶ Pathway analysis revealed that many of the top enriched pathways upon GLRX2 KD are involved in mitochondrial metabolism including ribonucleotide biosynthesis, ATP synthesis, and oxidative phosphorylation (OXPHOS; Figure 3D). AML cells are known to be highly dependent on mitochondrial metabolism for their survival.^47,48 Therefore, these data suggest that GLRX2 may be critical for AML survival by mediating mitochondrial activity.

GLRX2 regulates mitochondrial function

Mitochondrial metabolism is essential for AML survival and represents a promising therapeutic target.⁴⁹ To determine the consequences of GLRX2 depletion on mitochondrial biology we measured parameters of mitochondrial function in AML cell lines and primary AML specimens. GLRX2 depletion resulted in a decrease in OXPHOS in both AML cell lines and primary AML specimens (Figure 4A-C; supplemental Figure 4A-B). Consistent with a decrease in OXPHOS, GLRX2 KD resulted in a reduction of mitochondrial ATP production (Figure 4D), and total cellular ATP levels, as detected by mass spectrometry–based metabolomics (Figure 4E). In addition to reduced ATP levels, we observed a decrease in metabolites involved in purine and pyrimidine biosynthesis (supplemental Figure 4C), including nucleotides (supplemental Figure 4D). Nucleotide biosynthesis is essential in AML cells and is dependent on mitochondrial function.^50,51 To determine whether the reduction in nucleotide levels is part of the mechanism by which GLRX2 KD leads to AML cell death, we supplemented control and GLRX2 KD AML cells with individual and combinations of nucleosides. Notably, cell viability was not rescued through nucleoside supplementation, indicating that the reduction in nucleotides upon GLRX2 KD is not the cause of AML cell death (supplemental Figure 4E).

Figure 4.

GLRX2 regulates ATP5PO to mediates mPTP opening. (A) Representative oxygen consumption curve from viable MOLM13 5 days after GLRX2 KD. For all Seahorse experiments, viable cells were enriched using a dead cell removal bead kit before analysis. (B) Basal respiration in viable MOLM13, PL21, and MV411 5 days after GLRX2 KD. Mean ± SD (n = 3-4 biological replicates); ordinary 1-way ANOVA, ∗P < .05; ∗∗P < .01. (C) Oxygen consumption rate measured in primary AML specimens 8, 10, 11, and 12, 2 days after GLRX2 KD using the Seahorse assay. Mean ± SD; unpaired t test, ∗∗P < .01. (D) Effects of GLRX2 KD 5 days after transduction on ATP production in viable MOLM13, PL21, and MV411 measured using the Seahorse Mito Stress Test. Mean ± SD (n = 3-4 biological replicates); ordinary 1-way ANOVA, ∗P < .05; ∗∗P < .01. (E) Total cellular ATP level determined by mass spectrometry in MOLM13, PL21, and MV411, 5 days after GLRX2 KD. Mean ± SD (n = 4 technical replicates); ordinary 1-way ANOVA, ∗∗P < .01; ∗∗∗∗P < .0001. (F) Representative data of MOLM13 calcein blue signaling treated with CoCl₂, and ionomycin in DAPI⁻ population (n = 3 biological replicates). (G) mPTP opening in DAPI⁻ MOLM13, PL21, and MV411 4 days after expressing shControl or shRNAs against GLRX2. Mean ± SD (n = 3 biological replicates); ordinary 1-way ANOVA, ∗∗P < .01; ∗∗∗P < .001, ∗∗∗∗P < .0001. (H) mPTP opening measured in 4 DAPI⁻ primary AML specimens (AML14-17) 2 days after GLRX2 KD. Mean ± SD; unpaired t test, ∗∗P < .01. (I) Mitochondrial membrane potential in DAPI⁻ MOLM13, PL21, and MV411 4 days after GLRX2 KD. Mean ± SD (n = 3 biological replicates); ordinary 1-way ANOVA, ∗∗∗P < .001; ∗∗∗∗P < .0001. (J) Mitochondrial membrane potential measured in DAPI⁻ primary AML specimens 14, 15, and 17, 2 days after GLRX2 KD. Mean ± SD; unpaired t test, ∗∗P < .01. AU, arbitrary units; FCCP, carbonyl cyanide-p-trifluoromethoxyphenylhydrazone; MFI, mean fluorescence intensity; OCR, oxygen consumption rate; Scr, scrambled siRNA; TMRE, tetramethylrhodamine, ethyl ester.

Mitochondrial activity can also be modulated by membrane permeability. mPTP is a nonselective channel within the mitochondrial membrane, which allows diffusion of molecules up to 1.5 kDa in size.⁵² Transient opening of the mPTP is critical for regulating OXPHOS, mitochondrial Ca²⁺ levels, and mitochondrial ROS,^53,54 whereas prolonged opening of mPTP results in loss of mitochondrial membrane potential and cell death.⁵⁵ To determine whether GLRX2 regulates mPTP opening, we used calcein staining to measure mPTP opening in MOLM13, PL21, and MV411 after GLRX2 KD. Calcein is a cell-permeable dye that can passively diffuse into the mitochondria. In the presence of CoCl₂, a cytosolic calcein quencher, mitochondrial calcein retention can be used as measurement of mPTP closure. As a positive control, we used ionomycin treatment, which stimulates mPTP opening, leading to loss of calcein signal (Figure 4F). Compared with control AML cells, GLRX2 KD and knockout cells exhibited a significant reduction in mitochondrial calcein retention, indicating that GLRX2 depletion results in mPTP opening (Figure 4G; supplemental Figure 4F). Importantly, GLRX2 depletion also resulted in a significant increase in the mPTP opening in primary AML specimens (Figure 4H; supplemental Figure 5A).

Prolonged mPTP opening can be damaging to mitochondrial function because it causes reduced mitochondrial membrane potential.⁵⁵ Importantly, GLRX2 depletion also compromised the membrane potential in AML cell lines, and to a modest extent in primary AML specimens (Figure 4I-J). Furthermore, additional mitochondrial dysfunction was observed in AML cell lines such as reduction in total mitochondrial number (supplemental Figure 5B), increase in mitochondrial ROS (supplemental Figure 5C), and decrease in mitochondrial mass (supplemental Figure 5D). However, primary AML specimens demonstrated inconsistent changes in mitochondrial number upon GLRX2 KD (supplemental Figure 5E). We also did not see any changes in mitochondrial ROS or mitochondrial mass in primary AML samples (supplemental Figure 5F-G). This discrepancy could be due to many factors including proliferation differences between primary AML cells. It is also important to note that primary AML cells do not survive in culture long term as in cell lines; therefore, we measured mitochondrial function at earlier time points in primary AML specimens compared with the cell lines to reduce artifacts that could be a result of cell death occurring from in vitro culture. Altogether, our data show that GLRX2 is essential in the regulation of OXPHOS, ATP production, mPTP opening, and mitochondrial membrane potential in both cell line models and primary AML specimens.

GLRX2 mediates mPTP opening through ATP5PO

Next, we sought to identify how GLRX2-mediated deglutathionylation regulates mitochondrial function. To prioritize the evaluation of GLRX2 targets, we first measured their essentiality in AML using the DEPMAP database (supplemental Figure 6A). One of the genes essential in AML (supplemental Figure 6A), ATP5PO, is known to regulate mitochondrial biology including OXPHOS, mPTP opening, and membrane potential, each of which is dysregulated upon GLRX2 depletion.⁵⁶ ATP5PO is a component of the F1FO-ATP synthase peripheral stalk and coordinates the structural and functional stability of the enzyme.⁵⁷ ATP5PO also modulates the formation of the mPTP in a cyclophilin D (CypD) interaction–dependent manner. Cysteine-141 (C141) of ATP5PO was identified as a redox-sensitive residue regulating mPTP formation.⁵⁸ However, ATP5PO has, to our knowledge, not previously been identified as a glutathionylated protein or GLRX2 substrate. To validate that ATP5PO was indeed deglutathionylated by GLRX2 in AML cells, we performed proteomic analysis on flag tagged ATP5PO immunoprecipitated from MOLM13 cells expressing inducible GLRX2 or control short hairpin RNA. Using mass spectrometry, we identified a +305 Da mass shift consistent with the incorporation of a GSH moiety on the cysteine residue within the ATP5PO peptide sequence (GEVP¹⁴¹CTVTSASPLEEATLSELK; Figure 5A; supplemental Figure 6B). Importantly, GLRX2 KD led to an increased ratio of glutathionylated to total ATP5PO protein, supporting our finding from the BioGEE switch assay (Figure 5B; supplemental Figure 6C).

Figure 5.

GLRX2 mediates mPTP opening through reversible ATP5PO glutathionylation. (A) Tandem mass spectrum of the glutathionylated peptide GEVPC[305]TVTSASPLEEATLSELK from immunoprecipitated ATP5PO. The precursor ion with an m/z ratio of 856.07 is highlighted by the gray arrow, corresponding a GSH conjugated (+305 Da) triply charged peptide. (B) Quantitative measurement of ATP5PO glutathionylation, calculated by the ratio of modified to total ATP5PO MS1 peptide signal intensity 48 hours after doxycycline induction. Mean ± SD (n = 3 biological replicates); paired t test, ∗∗∗P < .005. (C) mPTP opening 4 days after GLRX2 KD in DAPI⁻ ATP5PO C141C wild-type (WT) and C141S (mutation) MOLM13 cells. Mean ± SD (n = 6 biological replicates); RM 1-way ANOVA, ∗P < .05; ∗∗P < .01. (D) Oxygen consumption curve of live ATP5PO C141C (WT) or C141S (mutation) MOLM13 with or without GLRX2 KD 5 days after transduction evaluated by Seahorse assay. (E) Mitochondrial membrane potential in DAPI⁻ ATP5PO C141C (WT) and C141S (mutation) MOLM13 cells 4 days after control or GLRX2 KD. Mean ± SD (n = 8 biological replicates); RM 1-way ANOVA. (F) Percentage of DAPI⁻ ATP5PO C141C (WT) and C141S (mutation) MOLM13 cells 4 days after transduction with control or GLRX2 shRNA. Mean ± SD (n = 3 biological replicates); ordinary 1-way ANOVA, ∗P < .05. (G) Opening of mPTP in GLRX2-KD MOLM13 cells is inhibited by 72-hour 250nM CsA treatment. Mean ± SD (n = 4 biological replicates); ordinary 1-way ANOVA, ∗∗∗∗P < .0001. (H) Colony-forming ability of MOLM13, PL21, and MV411 cells transduced with control or GLRX2-targeting shRNA in 250nM CsA-supplemented methylcellulose media. Mean ± SD (n = 4 biological replicates); ordinary 1-way ANOVA, ∗P < .05. FCCP, carbonyl cyanide-p-trifluoromethoxyphenylhydrazone; MFI, mean fluorescence intensity; MT, mutation; m/z, mass-to-charge; ns, not significant; NT, nontargeting shRNA; TMRE, tetramethylrhodamine, ethyl ester.

To determine whether ATP5PO glutathionylation is essential for AML mitochondrial biology and survival, we generated a stable MOLM13 AML line bearing an ATP5PO C141S point mutation using CRISPR/CRISPR–associated protein 9, which resulted in undetectable levels of glutathionylation at position 141 (supplemental Figure 6D-E). To control for any potential nonspecific effects from the modification, we produced MOLM13 cells carrying a C141C silent mutation (wild-type). Expression of the C141S mutation partially reduced mPTP opening after GLRX2 KD (Figure 5C). Interestingly, expression of the C141S mutation did not rescue OXPHOS (Figure 5D; supplemental Figure 6F), mitochondrial membrane potential (Figure 5E), mitochondrial ATP production (supplemental Figure 6G), mitochondrial ROS levels (supplemental Figure 6H), or mitochondrial mass (supplemental Figure 6I). However, expression of the C141S mutation was sufficient to partially rescue cell viability upon GLRX2 depletion, demonstrating that ATP5PO glutathionylation is an essential part of the mechanism by which GLRX2 loss targets AML cells (Figure 5F).

To examine whether reducing mPTP opening could rescue AML cell survival upon GLRX2 depletion we inhibited CypD, a positive regulator of mPTP, which promotes the conformational change of the ATP synthase through ATP5PO interaction. This interaction can be impaired by cyclosporin A (CsA), a well characterized mPTP inhibitor. Low-dose CsA treatment reduced mPTP opening caused by GLRX2 KD (Figure 5G; supplemental Figure 6J) and partially rescued the GLRX2-mediated reduction in MOLM13, PL21, and MV411 colony-forming potential (Figure 5H) without changing ATP5PO glutathionylation (supplemental Figure 6K). Our data show that GLRX2-mediated ATP5PO deglutathionylation is crucial to repress mPTP opening and for AML survival.

GLRX2-mediated mPTP activity is essential in primary AML cells

To determine whether repression of mPTP opening by GLRX2 was essential in primary AML, we measured colony-forming potential of 5 patient-derived primary AML samples treated with the mPTP inhibitor, CsA, upon GLRX2 depletion. CsA partially rescued the colony-forming potential and mPTP opening of all primary AML specimens upon GLRX2 KD (Figure 6A-B; supplemental Figure 7A). To complement these studies, we knocked down CypD in 3 primary AML specimens and examined whether reduced CypD expression could rescue mPTP opening and colony-forming potential upon GLRX2 depletion (supplemental Figure 7B-C). Although knocking down GLRX2 or CypD separately significantly affected mPTP opening and decreased the colony-forming potential, double knockdown restored mPTP and colony-forming potential similar to the control cells (Figure 6C-D). Using a patient-derived xenograft model we knocked down GLRX2, CypD, or the combination; transplanted an equal number of live cells into NSG-SGM3 mice; and measured engraftment potential and overall survival. GLRX2 or CypD knockdown resulted in reduced disease burden within the bone marrow and increased overall survival compared with control mice. However, compared with controls, engraftment potential and survival were not significantly different in the mice that received a transplant with the combination of GLRX2 and CypD knockdown cells (Figure 6E-F; supplemental Figure 7D). This suggests that mPTP opening is an important part of the mechanism by which GLRX2 depletion target AML cells and that mPTP regulation may be an essential process in AML.

Figure 6.

GLRX2 regulates mPTP opening and OXPHOS in primary AML specimens. (A) Colony-forming ability of 5 bulk primary AML specimens (AML13-17) supplemented with 250nM CsA in methylcellulose media. Mean ± SD (n = 3 technical replicates); ordinary 1-way ANOVA, ∗P < .05; ∗∗P < .01; ∗∗∗P < .001; ∗∗∗∗P < .0001. Number of colonies formed by AML17 with Scr and GLRX2 KD duplicated in Figure 2D. (B) mPTP opening in 3 DAPI⁻ primary AML specimens (AML13-15) transfected with scramble or GLRX2-targeting siRNAs, treated with or without CsA. Mean ± SD; ordinary 1-way ANOVA, ∗P < .05; ∗∗∗P < .001. (C) mPTP opening measured in 3 DAPI⁻ primary AML specimens (AML13-15) upon GLRX2 and CypD KD. Mean ± SD; ordinary 1-way ANOVA, ∗P < .01; ∗∗∗∗P < .0001. (D) Clonogenic assays of AML13-15 transfected with scramble, GLRX2, CypD, or GLRX2 + CypD siRNA. Mean ± SD (n = 3 technical replicates); ordinary 1-way ANOVA, ∗∗P < .01; ∗∗∗P < .001; ∗∗∗∗P < .0001. (E) Engraftment of AML21 in NSG-SGM3 mice after GLRX2 and CypD KD. Each point represents a single mouse. Mean ± SD; ordinary 1-way ANOVA, ∗P < .05. (F) Kaplan-Meier survival curves for NSG-SGM3 mice engrafted with AML21 (patient-derived xenograft model) after GLRX2 and CypD KD. Curve comparison done between NT and each KD group with log-rank (Mantel-Cox) test. (G) mPTP opening measured in 4 viable primary AML specimen populations (AML13, 14, 17, and 19) treated with or without 2.5μM cytarabine upon siRNA-mediated GLRX2 KD. Mean ± SD; ordinary 1-way ANOVA, ∗P < .05; ∗∗P < .01. (H) Colony-forming potential of AML11-15, 17, and 19 transfected with scramble or siRNA targeting GLRX2, treated with or without 2.5μM cytarabine. Mean ± SD (n = 3 technical replicates); ordinary 1-way ANOVA, ∗P < .05; ∗∗P < .01; ∗∗∗P < .001; ∗∗∗∗P < .0001. Number of colonies formed by AML11 and AML12 with Scr and GLRX2 KD conditions duplicated in Figure 2D. Comb, combination knockdown of CyPD and GLRX2; Cyt, cytarabine; MFI, mean fluorescence intensity; ns, not significant; NT, nontargeting shRNA; Scr, scrambled siRNA; Veh, vehicle.

mPTP opening is induced upon cytarabine treatment in AML, with reduced mPTP opening being implicated in the development of chemotherapy resistance.⁵⁹ Therefore, we sought to determine whether GLRX2 depletion could resensitize primary AML cells to the commonly used AML chemotherapy, cytarabine. GLRX2 KD in primary specimens resulted in enhanced mPTP opening and resensitization to cytarabine treatment (Figure 6G-H; supplemental Figure 7E). These data suggest that GLRX2 targeting may be a promising approach to target therapy resistant AML cells.

Overall, our data revealed that GLRX2 is essential for AML survival. GLRX2 mediates reversible ATP5PO glutathionylation to modulate mPTP opening. Independent of ATP5PO glutathionylation, GLRX2 regulates other mitochondrial properties including OXPHOS, membrane potential, and ROS, which also likely contributes to AML cell death upon GLRX2 depletion.

Discussion

GLRX2 is an oxidoreductase with a regulatory role on redox equilibrium and mitochondrial dynamics in cardiovascular, hepatic, immune, and central nervous systems.^24,26,46,60 Although these biological processes have been highlighted to be important in leukemic cells, the function of GLRX2 in AML was previously unknown. Herein, we demonstrated that GLRX2 is crucial for the survival of AML but not healthy HSPCs, indicating that a therapeutic window may exist to target GLRX2 in AML. Although transient GLRX2 KD using siRNA improved the survival outcomes, future studies using long-term genetic or pharmacological perturbations are critical. Importantly, GLRX2 knockout mice are viable, suggesting that it is not essential in mammals and could thus represent a therapeutic target.²⁴ Although GLRX2 knockout mice do develop heart hypertrophy and fibrosis as they age, heterozygous mice do not have cardiac-related problems, suggesting partial loss of GLRX2 is more tolerable than complete and prolonged knockout.²⁴

GLRX2 is upregulated in AML cells compared with normal hematopoietic cells irrespective of genetics, cytogenetics, French-American-British classification, or relapse status, suggesting that targeting GLRX2 may be a promising approach in a broad range of patients with AML. One of the greatest challenges in the treatment of AML is disease heterogeneity, but targeting GLRX2 has the potential to overcome this obstacle. To identify specific downstream targets of GLRX2, we used an unbiased proteomic approach that enriched putatively glutathionylated proteins. The BioGEE switch assay has the advantage of enriching for modified proteins allowing for large-scale screening of putative glutathionylated proteins. Caveats to this assay include the potential to identify proteins with a disulfide bond or other modifications such as sulfenylation as false positives. However, focusing on the differences between GLRX2-depleted and control cells helps to mitigate the inclusion of these false positives. Furthermore, it will be essential to use orthogonal approaches to validate that the other putative GLRX2 targets identified in our BioGEE assay are indeed GLRX2 substrates, as we did for ATP5PO.

Our proteomics data revealed that a large proportion of potential GLRX2 substrates are involved in the regulation of mitochondrial function. Mechanistically, we identified that GLRX2 depletion enhances mPTP opening through increased ATP5PO glutathionylation. ATP5PO is a major mPTP component and plays a regulatory role in the onset and progression of cardiovascular and neurodegenerative disease. In hypertrophic cardiac cells, and neurons from individuals with Alzheimer disease, ATP5PO expression is negatively correlated with mPTP opening, ROS production, mitochondrial membrane depolarization, and a defect in energy production.⁶¹ Our findings also highlighted that ATP5PO glutathionylation is crucial for cancer cell survival through the maintenance of mitochondrial homeostasis.

Loss of ATP5PO glutathionylation only partially rescues cell death upon GLRX2 KD, therefore suggesting that GLRX2 is crucial for AML survival through several mechanisms. We identified 11 additional mitochondrial proteins that are differentially glutathionylated upon GLRX2 depletion each of which may play a role in AML targeting upon loss of GLRX2. For example, previous studies on cardiac mitochondria showed an impairment on mitochondrial respiration due to enhanced glutathionylation of electron transport chain component NDUFS1 in GLRX2-deficient mice.²⁴ Interestingly, we also identified elevated glutathionylation levels of complex II subunits, succinate dehydrogenases A and B upon GLRX2 depletion. Complex I and complex II are both essential electron transport chain components for AML survival, so it is possible that the glutathionylation of these proteins contributes to GLRX2 KD–mediated AML cell death.^{9,62, 63, 64}

In summary, our data show that GLRX2 is essential for the survival of AML. We identified 49 putative GLRX2 targets, many of which were enriched for mitochondrial proteins, including ATP5PO. Glutathionylation of ATP5PO regulates the formation of mPTP, which mediates AML survival. Given the importance of GSH homeostasis in cancer, it is highly likely that the mechanisms described here in AML are relevant in other cancer types. There is no commercially available inhibitor for GLRX2. Future work is required to fully elucidate the role of protein glutathionylation and GLRX2 in cancer as well as explore the feasibility of therapeutically targeting GLRX2.

Conflict-of-interest disclosure: J.E.D. receives royalties from Trillium Therapeutics Inc, a commercial research grant from Celgene/Bristol Myers Squibb, and institutional licensing fees for acute myeloid leukemia models. The remaining authors declare no competing financial interests.