Metabolic circuitry of NADH buffering reveals a vulnerability in redox-adaptive cancer cells

Abstract

Mitochondrial dysfunction elevates cellular NADH/NAD⁺ ratios, inducing reductive stress that impairs biosynthesis and cell viability. However, the mechanisms by which cells buffer excess NADH to maintain redox homeostasis remain unclear. In this study, we identify a mitochondrial-cytosolic metabolic circuitry involving pyruvate carboxylase (PC), malate dehydrogenases (MDH1/2), and malic enzyme 1 (ME1) that mitigates NADH overload by coupling anaplerotic flux with NADH oxidation. Under reductive stress induced by electron transport chain (ETC) dysfunction, oxaloacetate derived from PC is converted to malate by MDH1/2, which is then oxidized by ME1, transforming both cytosolic and mitochondrial NADH into cytosolic NADPH. Although cellular NADPH is typically associated with antioxidant defense and biosynthesis, our experiments show that TPNOX, which can oxidase NADPH to NADP⁺, restores proliferation under ETC inhibition, and this rescue is entirely dependent on the mitochondrial-cytosolic metabolic circuitry. This finding highlights that the primary function of the circuit is to maintain NADH homeostasis rather than to generate NADPH. Notably, glucose-6-phosphate dehydrogenase (G6PD) enhances ME1 activity independently of its catalytic function by acting as a scaffold, thereby preventing net NADPH production while facilitating the conversion of NADH to NADPH. Disruption of the PC-MDH1/2-ME1 pathway through PC inhibition sensitizes cells to complex I inhibitors and/or glutaminase blockade, synergistically suppressing tumor growth both in vitro and in vivo. These findings uncover a redox-buffering strategy that redirects NADH into NADPH production, revealing a metabolic vulnerability in redox-adapted cancer cells.

Graphical abstract

Public summary

• •Glucose-6-phosphate dehydrogenase (G6PD) promotes malic enzyme 1 (ME1) activity independent of catalysis.
• •The PC-MDH1/2-ME1 circuitry supports cytosolic and mitochondrial NADH redox homeostasis.
• •Blocking PC sensitizes cancer cells to reductive stress and enhances antitumor synergy.

Introduction

Redox balance is a fundamental determinant of metabolic homeostasis, with NADH/NAD⁺ ratios critically regulating glycolysis, the tricarboxylic acid (TCA) cycle, and biosynthesis.^1,2,3,4,5 Under normal physiological conditions, mitochondrial respiration efficiently oxidizes NADH to regenerate NAD⁺, sustaining metabolic reactions and ATP production.^1,6 However, when electron transport chain (ETC) activity is compromised by hypoxia, mitochondrial dysfunction, oncogenic transformation, or pharmacological inhibition, NADH accumulates in both the cytosol and mitochondria, leading to reductive stress that threatens cell viability by disrupting central carbon metabolism, redox signaling, and anabolic capacity.^1,2,3,7

To mitigate this threat, cells engage multiple adaptive responses to restrain NADH production and facilitate its clearance.^1,3 Metabolic reprogramming under reductive stress involves suppressing NADH-generating pathways, including downregulation of the TCA cycle, and rerouting glucose metabolism toward lactate fermentation, which not only regenerates NAD⁺ independently of mitochondria but also sustains ATP production without exacerbating redox imbalance.^1,8 More recently, reductive stress has been shown to trigger glycerol biosynthesis and secretion, a cytosolic pathway that consumes NADH, albeit at the cost of carbon substrates and ATP.⁹ Additionally, cells activate glutamine-dependent lipid biosynthesis as a compensatory electron sink under reductive stress, but this pathway primarily consumes NADPH rather than NADH,^3,10,11,12 highlighting the compartmentalized redox challenge imposed by ETC dysfunction.

The inability to effectively dissipate excess NADH contributes to pathologies characterized by mitochondrial dysfunction and redox imbalance. In cancer, metabolic rewiring and hypoxic microenvironments often result in elevated NADH levels, promoting reductive stress that fuels genomic instability and treatment resistance.^3,13 Similarly, ischemic injury, acute alcoholic liver injury, fatty liver disease, mitochondrial myopathy, and neurodegenerative diseases involve impaired mitochondrial respiration and NADH accumulation, underscoring the clinical relevance of redox homeostasis in diverse disease contexts.^{7,14,15,16,17,18}

Efficient conversion of NADH into NADPH would simultaneously relieve reductive pressure and bolster antioxidant and biosynthetic capacity. While the mitochondrial nicotinamide nucleotide transhydrogenase (NNT) catalyzes NADH-to-NADPH conversion, its activity is membrane-potential dependent and restricted to the mitochondrial matrix, limiting its utility under ETC inhibition and offering no cytosolic redox relief.¹⁹ Consequently, a gap remains in our understanding of how cells coordinate NADH-to-NADPH conversion across subcellular compartments, especially within the cytosol where biosynthesis and reductive stress intersect.

Here, we identify a mitochondrial-cytosolic metabolic circuitry involving pyruvate carboxylase (PC), malate dehydrogenases (MDH1/2), and malic enzyme 1 (ME1) that enables cells to dissipate excess NADH by generating NADPH. This pathway is activated under ETC inhibition, linking anaplerotic carbon flux to cytosolic redox detoxification. We further show that glucose-6-phosphate dehydrogenase (G6PD) enhances ME1 activity through a non-enzymatic mechanism. Disruption of this circuit compromises redox balance and sensitizes cancer cells to ETC and glutamine metabolism inhibitors, highlighting a therapeutically targetable vulnerability in redox-adapted tumors.

Materials and methods

Cell culture and treatments

HeLa, SK-Hep1, 143B, and HEK293T cells were obtained from the American Type Culture Collection (ATCC) and cultured in Dulbecco’s modified Eagle’s medium (DMEM; Gibco) supplemented with 10% fetal bovine serum (FBS; Gibco) and 1% penicillin-streptomycin (Gibco) at 37°C in a humidified 5% CO₂ atmosphere. To minimize the interference of exogenous pyruvate on cellular NADH levels, NADH-related experiments were performed using pyruvate-free DMEM (Gibco). All cell lines were routinely tested and confirmed to be free of mycoplasma contamination, and no commonly misidentified cell lines were used in this study.

Measurement of NADH/NAD⁺ and NADPH/NADP⁺ ratios

Intracellular NADH/NAD⁺ and NADPH/NADP⁺ ratios were measured using NAD/NADH-Glo and NADP/NADPH-Glo assays (Promega) following the manufacturer’s protocols. Briefly, 5 × 10⁴ cells were seeded in 24-well plates overnight and incubated with the treatment medium. Cells were washed with PBS and lysed with 200 μL of ice-cold PBS/bicarbonate buffer (1:1) with 1% Dodecyl Trimethyl Ammonium Bromide (DTAB). For the acid treatment, 50 μL of the cell lysate was mixed with 25 μL of 0.4 N HCl and incubated at 60°C for 15 min. For the base treatment, the same volume of lysate was incubated at 60°C for 15 min. Neutralization was achieved by adding 25 μL of 0.5 M Trizma base (acid treated) or 50 μL of HCl/Trizma solution (base treated). Neutralized samples (10 μL) were transferred to 384-well white plates and mixed with 10 μL of freshly prepared detection reagent. After 30 min of gentle shaking at room temperature, luminescence was measured using a Synergy H1 Hybrid Multi-Mode reader (BioTek).

Cell viability assays

Cells were seeded in triplicate into 24-well plates (5 × 10⁴ or 1 × 10⁴ cells per well). The next day, cells were treated with indicated compounds for the specified duration. Then cells were trypsinized and the viable cells were counted using Muse Cell Analyzer (Luminex Corporation).

Tumor xenografts in nude mice

All of the xenograft experiments were approved by the Institutional Animal Care and Use Committee (IACUC) at Capital Medical University in accordance with the principles and procedures outlined in the National Institutes of Health Guide for the Care and Use of Laboratory Animals. Female nude mice (4–5 weeks old) were purchased from Beijing Vital River Laboratory Animal Technology. The mice were housed in a pathogen-free animal barrier facility. The housing conditions were maintained at a temperature of 19°C–23°C, humidity of 40%–70% and a 12/12 h light/dark cycle. The maximal tumor burden permitted by the ethics committee was less than 10% of the mouse’s body weight, and this limit was not exceeded. A total of 5 × 10⁶ HeLa cells or 8 × 10⁶ SK-Hep1 cells were injected subcutaneously into mice. Once the tumor reached 100–200 mm³, the mice were assigned randomly into different groups. Metformin was dissolved in PBS and administered by oral gavage at 100 mg kg⁻¹ daily or every other day. ZY444 and CB839 were dissolved in 10% dimethyl sulfoxide and 90% corn oil; ZY444 was administered by intraperitoneal injection at 2.5 mg kg⁻¹ daily or every other day, while CB839 was administered orally at 150 mg kg⁻¹ on the same schedule. Tumor volume was calculated using the formula: volume (mm³) = length (mm) × [width (mm)]² × ½.

Quantification and statistical analysis

The results of all the experiments were generated from at least three independent replicates unless noted otherwise in the figure legend. All results are presented as mean ± standard deviation (SD). Plots and statistical analysis were performed using Prism 10 software (GraphPad). Statistical significance was calculated as appropriate using a two-tailed Student's t test or two-way ANOVA as indicated in the figure legends.

Results

G6PD promotes ME1 activity independently of its enzymatic activity

Cytosolic NADPH is primarily generated by G6PD, ME1, and IDH1.²⁰ To assess their roles in redox regulation, we knocked down each enzyme in HeLa cells using small interfering RNAs (siRNAs). Silencing G6PD or ME1 significantly impaired cell proliferation, whereas IDH1 knockdown had no effect (Figures S1A–S1C). Notably, G6PD-depleted cells exhibited marked sensitivity to mitochondrial ETC inhibition by antimycin A and phenformin (Figure S1D), a phenotype not observed with ME1 or IDH1 knockdown (Figures S1E and S1F).

To further validate these findings, we generated CRISPR knockout (KO) lines of each enzyme. G6PD KO phenocopied the siRNA results, with impaired growth and enhanced susceptibility to ETC inhibition, both of which were rescued by re-expression of wild-type G6PD (Figures 1A–1C). In contrast, ME1 KO suppressed proliferation but did not affect ETC inhibition sensitivity (Figures S1G and S1H), and IDH1 KO had no detectable effects on either parameter (Figures S1I and S1J).

Figure 1.

G6PD promotes ME1 activity independently of its enzymatic activity

(A) Immunoblot analysis of G6PD expression in HeLa/wild-type (WT) cells and G6PD-KO cells re-expressing empty vector (EV), G6PD WT, or R198P mutant.

(B) Cell viability of HeLa/G6PD-KO and re-expression cells over 4 days.

(C) Cell viability of HeLa/G6PD-KO and G6PD re-expression cells treated with electron transport chain (ETC) inhibitors antimycin A (AntA, 1 μM) or phenformin (Phen, 100 μM) for 36 h.

(D) NADH/NAD⁺ and NADPH/NADP⁺ ratios in HeLa/WT, HeLa/G6PD-KO, and G6PD re-expression cells.

(E) ME1 enzymatic activity assay in HeLa/WT cells and G6PD-KO cells re-expressing EV, G6PD WT, or R198P with immunoblotting of ME1 and G6PD.

(F) ME1 activity in HEK293T cells transfected with FLAG-vector, G6PD-WT, or R198P, with immunoblotting of ME1 and FLAG-tagged proteins.

(G and H) In vitro ME1 activity with purified FLAG-ME1 incubated with increasing amounts of purified His-G6PD-WT (G) or R198P (H). Protein loading was verified by Coomassie-stained SDS-PAGE.

(I) Co-immunoprecipitation (coIP) of FLAG-ME1 in HEK293T cells followed by immunoblotting. WCL, whole-cell lysate.

(J) CoIP of endogenous G6PD with anti-G6PD antibody in HeLa cells. Rabbit immunoglobulin (Ig) G served as negative control.

(K) CoIP of endogenous ME1 with anti-ME1 antibody in HeLa cells. Mouse IgG served as negative control.

(L) FLAG pull-down assay of purified FLAG-ME1 incubated with purified His-G6PD-WT or R198P, followed by immunoblotting.

Data are presented as mean ± SD (n = 3 biologically independent samples). ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001 by two-tailed Student's t test.

Given the metabolic roles of these enzymes, we examined their effects on redox ratios. G6PD loss reduced the NADPH/NADP⁺ ratio and concurrently elevated the NADH/NAD⁺ ratio (Figure 1D), consistent with our previous reports.^21,22 Interestingly, ME1 KO did not affect NADPH/NADP⁺ ratio but increased NADH/NAD⁺ ratio (Figure S1K), suggesting a distinct role in redox homeostasis. IDH1 loss had no effect on either redox pair (Figure S1L).

We previously demonstrated that both wild-type G6PD and naturally occurring G6PD mutants found in patients with G6PD deficiency are capable of rescuing cell proliferation and redox balance in G6PD-deficient cells.²¹ In this study, we employed a catalytically inactive G6PD-R198P mutant²¹ and found that both the wild-type and R198P variants were able to restore the NADPH/NADP⁺ and NADH/NAD⁺ ratios (Figures 1A–1D), further supporting a non-enzymatic role for G6PD in maintaining redox homeostasis.

Considering that inactive G6PD maintained NADPH/NADP⁺ homeostasis, we next tested whether G6PD regulates ME1 or IDH1 activity. G6PD KO reduced ME1 enzymatic activity without affecting IDH1 (Figures 1E and S1M). Strikingly, both wild-type G6PD and R198P restored ME1 activity, and overexpression of either protein in 293T cells enhanced ME1 activity (Figures 1E and 1F), with no effect on IDH1 (Figure S1N).

In vitro assays using purified proteins confirmed that both G6PD and R198P directly enhances ME1 activity (Figures 1G and 1H). Co-immunoprecipitation experiments further demonstrated that G6PD interacts with ME1 at both the endogenous and exogenously expressed levels (Figures 1I–1K). Moreover, pull-down assays using purified His-tagged G6PD or G6PD-R198P and FLAG-tagged ME1 proteins revealed a direct physical interaction between G6PD and ME1 (Figure 1L). To further define the interaction interface between G6PD and ME1, we performed truncation mutant analyses based on the known structural domains of ME1 and G6PD (Figure S1O). Our results show that ME1 interacts only with full-length G6PD, as none of the G6PD truncation mutants retained binding capacity (Figure S1P), indicating that the structural integrity of G6PD is essential for its interaction with ME1.Truncation analysis of ME1 revealed that G6PD binds to the N-terminal region of ME1. Specifically, interaction with G6PD was observed in the ME1 truncation containing amino acids 1–271, whereas deletion of the first 121 amino acids completely abolished the interaction (Figure S1P). These findings collectively demonstrate that G6PD promotes ME1 activity through direct interaction, independently of its dehydrogenase activity.

ME1 maintains NADH/NAD⁺ balance via NADPH-generating metabolism

Our previous findings revealed that G6PD exerts multiple functions independent of its enzymatic activity, broadly regulating metabolic homeostasis.²¹ Consequently, G6PD-KO cells are highly susceptible to NADH accumulation-induced reductive stress under various stress conditions.²¹ Here, we challenged G6PD-KO HeLa cells with ETC inhibitors, antimycin A, or phenformin and found that treatment with α-ketobutyrate,²³ a substrate that can be reduced by lactate dehydrogenase (LDH) and thereby lowers the NADH/NAD⁺ ratio, significantly improved cell viability (Figures S2A–S2C). To further modulate the NADH/NAD⁺ ratio, we employed a doxycycline-inducible system to express LbNOX,² an engineered enzyme that transfers electrons from NADH to oxygen, generating NAD⁺ and water (Figures 2A and 2B). LbNOX expression also markedly enhanced the survival of G6PD-deficient cells under ETC inhibition (Figure 2C). This makes the G6PD-deficient cell line a suitable model for investigating the regulation of reductive stress.

Figure 2.

ME1 attenuates NADH reductive stress by generating NADPH

(A) Schematic of LbNOX catalysis. Immunoblot of inducible Tet-on FLAG-LbNOX in HeLa/G6PD-KO cells treated with Dox (0.1 μg/mL, 24 h).

(B) NADH/NAD⁺ and NADPH/NADP⁺ ratios in HeLa/G6PD-KO-Tet-on-LbNOX cells.

(C) Cell viability of HeLa/G6PD-KO-Tet-on-LbNOX cells under ETC inhibition.

(D) Immunoblot analysis of ME1, IDH1, and G6PD expression in HeLa/G6PD-KO cells reconstituted with EV, G6PD WT, or overexpression of ME1 or IDH1.

(E) Cell viability of HeLa/G6PD-KO cells expressing EV, G6PD WT, ME1, or IDH1 under AntA (1 μM) or Phen (100 μM) treatment for 36 h.

(F) NADH/NAD⁺ and NADPH/NADP⁺ ratios in HeLa/G6PD-KO cells expressing EV, G6PD WT, ME1, or IDH1 under AntA (1 μM) or Phen (100 μM) treatment for 8 h.

(G) Schematic of the enzymatic reaction catalyzed by TPNOX. Immunoblot analysis of inducible Tet-on FLAG-tagged TPNOX expression in HeLa/G6PD-KO cells treated with Dox (0.1 μg/mL) for 24 h.

(H) NADH/NAD⁺ ratios in HeLa/G6PD-KO-Tet-on-TPNOX cells.

(I) Relative SoNar fluorescence ratios (F_420nm/F_485nm) in HeLa/G6PD-KO-Tet-on-TPNOX cells.

(J) Cell viability of HeLa/G6PD-KO-Tet-on-TPNOX cells under ETC inhibition.

(K) Immunoblot analysis of ME1 and FLAG-TPNOX expression in HeLa/G6PD-KO-Tet-on-TPNOX cells with or without ME1 knockdown.

(L) Cell viability of HeLa/G6PD-KO-Tet-on-TPNOX cells with or without ME1 knockdown under ETC inhibition.

(M) Relative SoNar fluorescence ratios (F_420nm/F_485nm) in HeLa/G6PD-KO-Tet-on-TPNOX cells with or without ME1 knockdown.

(N and O) Cell viability of G6PD-KO cells expressing EV or ME1 under AntA (1 μM) or Phen (100 μM) treatment for 36 h with or without ND646 (10 μM) in HeLa (N) or SK-Hep1 cells (O).

Data are presented as mean ± SD (n = 3 biologically independent samples). ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001 by two-tailed Student's t test.

Given that G6PD regulates ME1 activity and ME1 deletion leads to elevated NADH/NAD⁺ ratios (Figures 1E–1H and S1K), we next investigated the role of ME1 in the cellular response to reductive stress. We ectopically expressed ME1 or IDH1 in G6PD-KO HeLa cells, with G6PD re-expression serving as a control (Figure 2D). Similar to G6PD, ME1 overexpression significantly improved cell viability under ETC inhibition (Figure 2E) and mitigated the increase in the NADH/NAD⁺ ratio and the decline in the NADPH/NADP⁺ ratio (Figure 2F). In contrast, IDH1 overexpression had minimal effects on these parameters (Figures 2D–2F). Consistent results were observed in another cell line, SK-Hep1, where ME1 overexpression rescued cell viability, reduced the NADH/NAD⁺ ratio, and restored the NADPH/NADP⁺ ratio in G6PD-deficient cells under ETC inhibition (Figures S2D–S2F). We also examined whether other malic enzymes or NADPH-generating enzymes, including ME2, ME3, and IDH2, could affect the proliferation of G6PD-KO cells under ETC inhibition. Our results showed that overexpression of none of these enzymes restored proliferative capacity (Figures S2G–S2H), indicating that they are unable to functionally substitute for ME1 in this context.

To dissect whether ME1-mediated rescue depends on restoring NADH/NAD⁺ or NADPH/NADP⁺ balance, we expressed doxycycline (Dox)-inducible TPNOX (Figure 2G), a synthetic oxidase that transfers electrons from NADPH to oxygen.²⁴ Interestingly, TPNOX often led to a reduction in the NADH/NAD⁺ ratio without significantly affecting the NADPH/NADP⁺ ratio, likely due to an as-yet-undefined metabolic pathway that converts NADH into NADPH.²⁴ Consistently, TPNOX expression in HeLa/G6PD-KO cells did not significantly reduce NADPH/NADP⁺ ratio (Figure S2I) but attenuated ETC-inhibition-induced NADH/NAD⁺ elevation (Figure 2H), as confirmed by the genetically encoded cytosolic NADH/NAD⁺ biosensor SoNar²⁵ (Figure 2I). Importantly, TPNOX significantly enhanced the survival of G6PD-KO cells under ETC inhibition (Figure 2J). A similar outcome was achieved with EcSTH-LbNOX, another NADPH oxidase,¹⁵ which not only lowered the NADPH/NADP⁺ ratio but also decreased the NADH/NAD⁺ ratio under ETC inhibition (Figures S2J–S2L).

Strikingly, siRNA-mediated depletion of ME1 abrogated the ability of TPNOX to rescue both the elevated NADH/NAD⁺ ratio and the impaired viability of G6PD-deficient HeLa or SK-Hep1 cells under ETC inhibition (Figures 2K–2M and S3A–S3C). De novo fatty acid synthesis requires a substantial amount of NADPH. By employing the inhibitor ND646 to restrict NADPH consumption in cells, we observed a significant reduction in the rescue effect of ME1 on the impaired viability of G6PD-deficient HeLa and SK-Hep1 cells under ETC inhibition (Figures 2N and 2O). These findings suggest that ME1 contributes to redox homeostasis not merely by simply providing cellular NADPH but through a metabolic mechanism that couples NADPH production to NADH oxidation. Thus, ME1 plays a key role in alleviating NADH reductive stress by functionally linking NADPH generation to the maintenance of NADH/NAD⁺ balance.

ME1 cooperates with MDH1/2 to mitigate NADH reductive stress

Malate, the substrate for cytosolic ME1, is generated by mitochondrial MDH2 and cytosolic MDH1 (Figure 3A). To determine whether ME1 activity relies on upstream malate supply, we silenced MDH1 in G6PD-KO HeLa cells. Knockdown of MDH1 markedly attenuated the TPNOX-mediated decrease in the NADH/NAD⁺ ratio under ETC inhibition, as detected by the SoNar biosensor (Figures 3B and 3C). MDH1 depletion also impaired TPNOX-mediated rescue of cell proliferation under these conditions (Figure 3D). Interestingly, similar effects were observed upon knockdown of MDH2, suggesting that both cytosolic and mitochondrial malate dehydrogenases are required for the redox-buffering function of ME1 (Figures 3E–3G). These findings were corroborated in SK-Hep1 G6PD-KO cells, where knockdown of either MDH1 or MDH2 similarly impaired TPNOX function (Figures S3D–S3I).

Figure 3.

ME1-MDH1/2 alleviates NADH reductive stress

(A) Schematic of the mitochondrial malate-aspartate shuttle, which transfers reducing equivalents from mitochondrial NADH to the cytosol.

(B) Immunoblot analysis of MDH1 and FLAG-TPNOX expression in HeLa/G6PD-KO-Tet-on-TPNOX cells with or without MDH1 knockdown.

(C) Relative SoNar fluorescence ratios (F_420nm/F_485nm) in HeLa/G6PD-KO-Tet-on-TPNOX cells with or without MDH1 knockdown.

(D) Cell viability of HeLa/G6PD-KO-Tet-on-TPNOX cells with or without MDH1 knockdown under ETC inhibition.

(E) Immunoblot analysis of MDH2 and FLAG-TPNOX expression in HeLa/G6PD-KO-Tet-on-TPNOX cells with or without MDH2 knockdown.

(F) Relative SoNar fluorescence ratios (F_420nm/F_485nm) in HeLa/G6PD-KO-Tet-on-TPNOX cells with or without MDH2 knockdown.

(G) Cell viability of HeLa/G6PD-KO-Tet-on-TPNOX cells with or without MDH2 knockdown under ETC inhibition.

(H) Relative SoNar fluorescence ratios (F_420nm/F_485nm) in HeLa cells with individual knockdown of ME1, MDH1, or MDH2. Corresponding immunoblot analyses of ME1, MDH1, or MDH2 expression in respective knockdown cells.

(I) Cell viability of HeLa-Tet-on-TPNOX cells with or without ME1 knockdown upon AntA (0.1 μM) or Phen (50 μM) treatment for 4 days. Corresponding immunoblot analysis of ME1 and FLAG-TPNOX expression.

(J) Cell viability of SK-Hep1-Tet-on-TPNOX cells with or without ME1 knockdown upon AntA (0.1 μM) or Phen (50 μM) treatment for 4 days. Corresponding immunoblot analysis of ME1 and FLAG-TPNOX expression.

Data are presented as mean ± SD (n = 3 biologically independent samples). ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001 by two-tailed Student's t test.

To evaluate the role of this pathway under basal conditions, we knocked down ME1, MDH1, or MDH2 in wild-type HeLa, SK-Hep1, and 143B cells. All three perturbations resulted in elevated SoNar signals, indicating increased NADH/NAD⁺ ratios and underlying a conserved role in redox homeostasis (Figures 3H and S4A–S4F). We further found that TPNOX conferred protection against ETC inhibition in wild-type HeLa and SK-Hep1 cells. However, this protective effect was significantly blunted by knockdown of ME1, MDH1, or MDH2 (Figures 3I, 3J, and S5A–S5D), indicating that TPNOX function relies on an intact MDH1/2-ME1 metabolic axis.

Together, these results demonstrate that the oxidation of NADPH by TPNOX creates a metabolic “pull” that increases ME1-mediated malate consumption. This elevated ME1 flux stimulates MDH1/2 activity to regenerate malate, leading indirectly to NADH oxidation. ME1 and its upstream malate-generating enzymes, MDH1 and MDH2, form a functional metabolic pathway that is essential for buffering cellular NADH, maintaining NADH/NAD⁺ homeostasis, and mitigating reductive stress.

Reductive stress activates PC-dependent rerouting of glucose carbon toward malate and aspartate synthesis

To investigate metabolic changes associated with the ME1-MDH1/2 pathway under conditions of NADH reductive stress, we performed [U-¹³C]-glucose tracing in HeLa cells. ETC inhibition using antimycin A markedly suppressed ¹³C incorporation into TCA cycle intermediates, as evidenced by decreased isotopic labeling proportions of multiple citrate isotopologues (Figures 4A–4C). In contrast, we observed a pronounced increase in the ¹³C-labeling proportions of malate and aspartate, particularly the m+3 isotopologues (Figures 4D–4G). Given that ¹³C-labeling of citrate was decreased (Figure 4C), this enrichment was not due to enhanced TCA cycling but rather suggests that m+3-labeled malate and aspartate were derived from ¹³C₃-pyruvate converted to oxaloacetate (OAA) via PC, bypassing the TCA cycle (Figure 4A). CRISPR-mediated deletion of PC significantly reduced the fractional ¹³C-enrichment of malate m+3 and aspartate m+3 under both basal and antimycin A-treated conditions, and these effects were reversed upon PC re-expression (Figures 4B–4G), confirming a PC-dependent pathway for glucose-derived OAA generation under ETC inhibition.

Figure 4.

NADH reductive stress promotes PC-mediated glucose-derived malate and aspartate

(A) Schematic diagram illustrating the metabolic flux of ¹³C₆-glucose through glycolysis, the TCA cycle, and TCA anaplerosis via PC.

(B) Immunoblot analysis of PC expression in HeLa/WT cells and PC-KO cells re-expressing EV or PC WT.

(C–E) ¹³C-labeled fractions of citrate (C), malate (D), and aspartate (E) in the indicated HeLa cells cultured with 10 mM ¹³C₆-glucose for 8 h under AntA (1 μM) treatment.

(F) Quantification of malate m+3 isotopologues (¹³C-labeled fraction), absolute m+3 levels, and total malate abundance under the same conditions as in (C–E).

(G) Quantification of aspartate m+3 isotopologues, absolute m+3 levels, and total aspartate abundance under the same conditions as in (C–E).

(H) The ratios of 2-hydroxyglutarate (2HG) to α-ketoglutarate (2KG) in the indicated HeLa cells after 8 h of AntA(1 μM) treatment.

Data are presented as mean ± SD (n = 3 biologically independent samples). ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001 by two-tailed Student's t test.

A hallmark of NADH reductive stress is the activation of reductive glutamine metabolism, which can be monitored via increased ¹³C-labeling proportions of citrate m+5 from [U-¹³C]-glutamine^3,10 (Figure S6A). In HeLa cells, ETC inhibition led to elevated ¹³C-enrichment in citrate m+5 and its derivatives, malate m+3 and aspartate m+3 (Figures S6B–S6G). Interestingly, PC KO cells exhibited higher baseline ¹³C-labeling proportions of these metabolites, and this effect was further amplified upon ETC inhibition. Reintroduction of PC reversed these increases (Figures S6E–S6G), suggesting that PC constrains reductive glutamine flux. Additionally, PC deletion led to an increase in the ratio of the reductive metabolite 2-hydroxyglutarate and its oxidized form 2-oxoglutarate (Figure 4H), further supporting a role for PC in mitigating reductive stress.

PC supports cytosolic and mitochondrial NADH redox homeostasis

To evaluate the role of PC in redox regulation, we overexpressed PC in G6PD-KO HeLa cells. Similar to G6PD expression, PC expression robustly rescued cell viability under various stress conditions, including ETC inhibition (Figures 5A, 5B, and S7A). PC overexpression also attenuated the increase in NADH/NAD⁺ ratio caused by ETC inhibition, but its effect on NADPH/NADP⁺ levels was minimal (Figure 5C). Conversely, silencing PC impaired the ability of TPNOX to rescue G6PD-KO HeLa and SK-Hep1 cells under ETC inhibition (Figures 5D, 5E, S7B, and S7C). This requirement for PC was similarly observed in wild-type HeLa and SK-Hep1 cells (Figures 5F, 5G, and S7D), underscoring the importance of PC in countering reductive stress.

Figure 5.

PC plays a critical role in maintaining NADH/NAD+ homeostasis

(A) Immunoblot verification of the overexpression of PC in HeLa/G6PD-KO cells.

(B) Cell viability of HeLa/G6PD-KO cells overexpressing PC under AntA (1 μM) treatment for 36 h.

(C) NADH/NAD⁺ and NADPH/NADP⁺ ratios in HeLa/WT, HeLa/G6PD-KO cells expressing EV or PC under AntA (1 μM) treatment for 8 h.

(D) Immunoblot analysis of PC and FLAG-TPNOX expression in HeLa/G6PD-KO-Tet-on-TPNOX cells with or without PC knockdown.

(E) Cell viability of HeLa/G6PD-KO-Tet-on-TPNOX cells with or without PC knockdown under ETC inhibition.

(F) Immunoblot analysis of PC and FLAG-TPNOX expression in HeLa-Tet-on-TPNOX cells with or without PC knockdown.

(G) Cell viability of HeLa-Tet-on-TPNOX cells with or without PC knockdown upon AntA (0.1 μM) or Phen (50 μM) treatment for 4 days.

(H) NADH/NAD⁺ ratios in HeLa cells with or without PC knockdown.

(I) Relative SoNar fluorescence ratios (F_420nm/F_485nm) in HeLa cells with or without PC knockdown.

(J) NADH/NAD⁺ and NADPH/NADP⁺ ratios in HeLa cells treated with varying concentrations of PC inhibitor ZY444 for 12 h.

(K) Effects of respiratory-chain inhibition on cell viability in HeLa/G6PD-KO cells overexpressing different PC mutants. Corresponding immunoblot analysis of PC expression.

(L) Mitochondrial NADH assessed by Frex-Mit fluorescence signal in HeLa cells treated with ZY444 for 12 h. Fluorescence values were corrected for pH using cpYFP-Mit signal.

(M) Effects of PC knockdown on Frex-Mit fluorescence in HeLa cells. Corresponding immunoblot analysis of PC expression.

Data are presented as mean ± SD (n = 3 biologically independent samples). ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001 by two-tailed Student's t test.

PC depletion by siRNA in wild-type HeLa cells led to a significant increase in the NADH/NAD⁺ ratio and elevated SoNar signal (Figures 5H, 5I, and S7E), consistent with an induction of reductive stress. Similar effects were observed in SK-Hep1 and 143B cells, where PC knockdown increased NADH/NAD⁺ ratio and SoNar signal but did not alter the NADPH/NADP⁺ ratio (Figures S7F–S7H). Pharmacological inhibition of PC using ZY444²⁶ also dose-dependently increased NADH/NAD⁺ ratios without affecting NADPH/NADP⁺ in HeLa and SK-Hep1 cells (Figures 5J and S7I), further confirming PC’s role in buffering excess NADH.

A recent study reported that, under stress, PC may translocate to the cytosol and interact with MDH1 and ME1 to form a complex facilitating NADH-to-NADPH conversion.²⁷ To determine the subcellular site of PC action, we expressed wild-type or N-terminally FLAG-tagged PC constructs in G6PD-KO HeLa cells. Only wild-type PC, but not N-terminally FLAG-tagged PC, rescued growth under ETC inhibition, correlating with proper mitochondrial localization (Figures 5K and S8A–S8C). Similarly, PC constructs lacking the first 20 amino acids of the mitochondrial targeting sequence (PCΔ1-20Aa) or its N-terminally FLAG-tagged counterpart were unable to rescue viability (Figure 5K). However, when a mitochondrial targeting signal was reintroduced into these truncated or FLAG-tagged forms, mitochondrial localization was restored (Figures S8C–S8F), and the rescue of cell survival was regained (Figure 5K), indicating that PC functions within mitochondria to combat reductive stress.

Cytosolic NADH/NAD⁺ is metabolically associated with the mitochondrial NADH/NAD⁺ pool. While the SoNar biosensor selectively detects cytosolic NADH/NAD⁺, we employed the Frex-Mit biosensor to directly monitor mitochondrial NADH levels²⁸ in order to determine whether PC, as well as ME1, contributes to the regulation of mitochondrial reductive stress. Our results showed that pharmacological inhibition using ZY444, as well as siRNA-mediated knockdown of PC, ME1, MDH1, or MDH2, markedly increased Frex-Mit signal in HeLa, SK-Hep1, and 143B cells (Figures 5L, 5M, and S9A–S9M), indicating mitochondrial NADH accumulation. Collectively, these results demonstrate that PC, ME1, MDH1, and MDH2 play a critical role in maintaining NADH redox homeostasis across cellular compartments, thereby protecting both the cytosol and mitochondria from reductive stress.

Blocking PC sensitizes cancer cells to reductive stress and enhances antitumor synergy

We further observed that PC KO sensitized HeLa and SK-Hep1 cells to ETC inhibition, whereas re-expression of PC reversed this sensitivity (Figures 6A and 6B). As a mitochondrial enzyme, PC produces OAA, which can be converted to malate by MDH2 or to aspartate via GOT2; both metabolites are exported to the cytosol and contribute to NADH detoxification pathways (Figure 3A). To dissect their relative contributions downstream of PC, we silenced MDH2 and GOT2 individually or in combination in HeLa cells. While single knockdowns only mildly increased sensitivity to ETC inhibition, combined knockdown of MDH2 and GOT2 mimicked the effect of PC loss, indicating that both pathways downstream of PC are required for resistance to reductive stress (Figures 6C and 6D).

Figure 6.

Blocking PC sensitizes cancer cells to NADH reductive stress

(A) Cell viability of HeLa/WT cells and HeLa/PC-KO cells reconstituted with EV or PC under AntA (0.1 μM), Phen (50 μM), or metformin (2 mM) treatment for 5 days. Corresponding immunoblot analysis of PC expression.

(B) Cell viability of SK-Hep1/WT cells and SK-Hep1/PC-KO cells reconstituted with EV or PC under the same treatment conditions as in (A). Corresponding immunoblot analysis of PC expression.

(C) Immunoblot analysis of MDH2 and GOT2 expression in HeLa cells with single or combined knockdown of MDH2 and GOT2.

(D) Cell viability of HeLa cells with MDH2 or GOT2 single or combined knockdown under Phen (50 μM) or metformin (2 mM) treatment for 5 days.

(E) Cell viability of HeLa, 143B, and SK-Hep1 cells treated with ZY444 (2 μM), metformin (0.5 mM), or their combination for 5 days.

(F) NADH/NAD⁺ ratios in HeLa, 143B, and SK-Hep1 cells treated with ZY444 (2 μM), metformin (0.5 mM), or their combination for 12 h.

(G) Tumor growth of HeLa xenografts in nude mice treated with metformin (100 mg kg⁻¹ per day, i.g.), ZY444 (2.5 mg kg⁻¹ per day, intraperitoneal [i.p.]), CB839 (150 mg kg⁻¹ per day, i.g.), or their combinations. The control group was treated with the same volume of vehicle (10% DMSO and 90% corn oil).

(H) Tumor growth of SK-Hep1 xenografts in nude mice under the same treatment regimens as in (G), administered every other day.

Data are presented as mean ± SD (n = 3 biologically independent samples or n = 6 biologically independent mice). ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001 by two-tailed Student's t test.

Next, we evaluated the therapeutic potential of targeting PC in combination with ETC inhibition. The PC inhibitor ZY444 synergized with complex I inhibitors, including metformin, antimycin A, and phenformin, to impair cell viability in HeLa, SK-Hep1, and 143B cells (Figures 6E, S10A, and S10B). These combinatorial treatments also caused a more pronounced increase in the intracellular NADH/NAD⁺ ratio (Figure 6F), indicating enhanced reductive stress.

A previous study has reported that PC inhibition reduces cell viability under glutamine deprivation,²⁹ consistent with the notion that reductive stress increases dependence of cancer cells on glutamine metabolism. Supporting this, we found that PC-deficient HeLa cells were more sensitive to both glutamine withdrawal and glutaminase (GLS) inhibition by CB839³⁰ (Figures S10C and S10D). Furthermore, the combination of ZY444 with both metformin and CB839 resulted in greater suppression of cell viability than dual combinations across all three cell lines tested (Figure S10E). These results reinforce the role of PC in buffering reductive stress and maintaining metabolic flexibility under nutrient limitation.

In xenograft models using HeLa and SK-Hep1 cells, individual treatment with ZY444, metformin, or CB839 showed no or only modest tumor inhibition. However, dual combinations (ZY444 + metformin or ZY444 + CB839) significantly impaired tumor growth, and the triple combination nearly abolished tumor progression, indicating a strong synergistic effect in vivo (Figures 6G, 6H, S10F, and S10G).

These findings demonstrate that PC is a critical node in protecting cancer cells from NADH reductive stress and that its inhibition synergizes with ETC blockade and glutamine utilization suppression, offering a potential combinatorial therapeutic strategy.

Discussion

Impairment of the ETC leads to NADH accumulation in both cytosol and mitochondria, causing reductive stress that disrupts glycolysis, biosynthesis, and antioxidant defense.^2,3,31 We identify a dual-branched, PC-driven mechanism that channels excess NADH into NADPH production via malate metabolism, coupling mitochondrial anaplerosis to cytosolic redox detoxification (Figure 7). PC converts pyruvate to OAA in mitochondria, after which OAA exits through two routes. In one branch, MDH2 reduces OAA to malate using mitochondrial NADH; malate is exported and oxidized by ME1 to generate NADPH, converting mitochondrial NADH equivalents into cytosolic NADPH. In the other, GOT2 transaminates OAA to aspartate for export; cytosolic GOT1 regenerates OAA, MDH1 reduces it to malate using cytosolic NADH, and ME1 oxidizes malate to produce NADPH. Together, these pathways allow both mitochondrial and cytosolic NADH pools to fuel ME1, increasing NADPH when reductive stress is most severe. The resulting NADPH supports antioxidant and anabolic reactions, thereby relieving NADH burden.

Figure 7.

Schematic model of the PC-MDH1/2-ME1 pathway that linking NADH detoxification to NADPH generation

Under conditions of ETC inhibition, a mitochondrial-cytosolic metabolic circuitry centered on PC, MDH1/2, and ME1 is activated to buffer excess NADH. PC-mediated anaplerotic flux channels glucose-derived carbon into OAA, which is transformed into malate via either the mitochondrial MDH2 branch (consuming mitochondrial NADH) or the GOT2-GOT1-MDH1 cytosolic shuttle (consuming cytosolic NADH). The resulting malate is oxidized by ME1 to produce cytosolic NADPH, thereby coupling mitochondrial carbon metabolism to redox detoxification. G6PD enhances ME1 activity through a non-enzymatic scaffold function, enabling efficient NADH-to-NADPH flux. Disruption of this pathway compromises redox homeostasis and sensitizes cells to ETC and glutamine metabolism inhibition, revealing a therapeutically targetable redox vulnerability.

This dual architecture provides two advantages. The MDH2-malate-ME1 axis directly consumes mitochondrial NADH, preserving NAD⁺ to sustain residual oxidative phosphorylation and dehydrogenase activity.³² The GOT2-GOT1/MDH1-malate-ME1 shuttle removes excess cytosolic NADH generated by ETC blockade and converts it into NADPH to maintain antioxidant and biosynthetic capacity. In both cases, NADH is redirected into adaptive reducing power that supports survival.

We further show that G6PD enhances ME1 activity through a non-catalytic scaffolding function (Figure 1). By promoting efficient malate oxidation, G6PD sustains continuous NADH-to-NADPH flux. Loss of this interaction impairs activation of the PC-MDH1/2-ME1 pathway and renders cells susceptible to reductive collapse.

Although PC has been implicated in ME1-dependent NADPH production during oxidative stress,^27,32 our findings redefine the PC-ME1 axis as an NADH disposal system functioning as a metabolic transhydrogenation-like cycle that converts mitochondrial and cytosolic NADH into NADPH. This circuit is driven by the need to maintain NADH/NAD⁺ homeostasis under reductive stress, with NADPH consumption—physiological (biosynthesis and antioxidant defense) or experimental (eg, TPNOX-mediated oxidation)—providing the pull that sustains flux (Figure 7). Thus, its dominant role is redox balancing rather than net NADPH accumulation. G6PD optimizes this cycle by scaffolding ME1 into a high-activity complex while suppressing its own NADPH-producing activity, ensuring efficient malate oxidation and uninterrupted NADH clearance. Disruption of this regulation markedly compromises the PC-MDH1/2-ME1 axis.

Therapeutically, this dual PC pathway represents a metabolic vulnerability. PC inhibition blocks both malate export routes and sensitizes cancer cells to complex I and glutamine metabolism inhibitors in vitro and in vivo (Figure 6). Because many cancers exhibit elevated NADH/NAD⁺ ratios, they may depend on the PC-ME1 network for redox homeostasis. Targeting this axis may therefore impair NADH clearance and antioxidant capacity, undermining tumor viability.

In summary, we define a two-branched, PC-dependent malate shuttle that converts NADH into NADPH, linking mitochondrial anaplerosis with cytosolic redox homeostasis. This mechanism supports survival under respiratory stress and represents a tractable target for combination cancer therapies. Future studies should determine how this circuit is regulated across tissues and stress contexts and how it integrates with other redox-buffering systems.

Resource availability

Materials availability

All unique/stable reagents generated in this study are available from the lead contact with a completed materials transfer agreement.

Data and code availability

Data are available from the corresponding author upon reasonable request.

Funding and acknowledgments

We thank Dr. Xiaohui Liu (Metabolomics Facility at Tsinghua University Branch of China National Center for Protein Sciences, China) for technical help. This work was supported by grants from the National Natural Science Foundation of China (82325038, 82530094, and 82303227), the Chinese Institutes for Medical Research, Beijing (CX23YZ18), and the Beijing High-Level Innovation and Entrepreneurship Talent Support Program-Young Backbone Talent Projects (G202532306).

Author contributions

Conceptualization, B.L.; investigation, X.L., R.Y., P.H., X.Z., L.L., and Z.G.; funding acquisition, B.L. and R.Y.; advice and discussion, P.C., L.Q., Y.F., and J.N.; supervision, B.L.; writing – original draft, B.L.; writing – review & editing, B.L., R.Y., and X.L. All authors contributed to the manuscript and approved the final version.

Declaration of interests

The authors declare no competing interests.

Published Online: April 1, 2026