1. Biochemistry and Chemical Biology
  2. Cancer Biology

Mitochondrial redox adaptations enable alternative aspartate synthesis in SDH-deficient cells

  1. Madeleine L Hart
  2. Evan Quon
  3. Anna-Lena BG Vigil
  4. Ian A Engstrom
  5. Oliver J Newsom
  6. Kristian Davidsen
  7. Pia Hoellerbauer
  8. Samantha M Carlisle
  9. Lucas B Sullivan  Is a corresponding author
  1. Human Biology Division, Fred Hutchinson Cancer Center, United States
  2. Molecular Medicine & Mechanisms of Disease Program, University of Washington, United States
  3. Department of Chemistry and Biochemistry, New Mexico State University, United States
https://doi.org/10.7554/eLife.78654
Share this article
Cite this article
  1. Madeleine L Hart
  2. Evan Quon
  3. Anna-Lena BG Vigil
  4. Ian A Engstrom
  5. Oliver J Newsom
  6. Kristian Davidsen
  7. Pia Hoellerbauer
  8. Samantha M Carlisle
  9. Lucas B Sullivan
(2023)
Mitochondrial redox adaptations enable alternative aspartate synthesis in SDH-deficient cells
eLife 12:e78654.
https://doi.org/10.7554/eLife.78654
  2. Download BibTeX
  3. Download .RIS

Abstract

The oxidative tricarboxylic acid (TCA) cycle is a central mitochondrial pathway integrating catabolic conversions of NAD +to NADH and anabolic production of aspartate, a key amino acid for cell proliferation. Several TCA cycle components are implicated in tumorigenesis, including loss-of-function mutations in subunits of succinate dehydrogenase (SDH), also known as complex II of the electron transport chain (ETC), but mechanistic understanding of how proliferating cells tolerate the metabolic defects of SDH loss is still lacking. Here, we identify that SDH supports human cell proliferation through aspartate synthesis but, unlike other ETC impairments, the effects of SDH inhibition are not ameliorated by electron acceptor supplementation. Interestingly, we find aspartate production and cell proliferation are restored to SDH-impaired cells by concomitant inhibition of ETC complex I (CI). We determine that the benefits of CI inhibition in this context depend on decreasing mitochondrial NAD+/NADH, which drives SDH-independent aspartate production through pyruvate carboxylation and reductive carboxylation of glutamine. We also find that genetic loss or restoration of SDH selects for cells with concordant CI activity, establishing distinct modalities of mitochondrial metabolism for maintaining aspartate synthesis. These data therefore identify a metabolically beneficial mechanism for CI loss in proliferating cells and reveal how compartmentalized redox changes can impact cellular fitness.

Editor's evaluation

Mutations in subunits of succinate dehydrogenase (SDH) are implicated in tumorigenesis, but it is not fully understood how cell proliferation occurs in cells that lack SDH. Hart et al. elegantly demonstrate that lack of mitochondrial complex I increases proliferation in cells lacking SDH. The authors have expertly dissected the mechanism behind the rescue and show that inhibition of complex I restores aspartate production by modulating NAD+/NADH. This study thus demonstrates that compartmentalized redox changes control distinct aspartate biosynthetic pathways that promote the proliferation of tumor cells.

https://doi.org/10.7554/eLife.78654.sa0

Introduction

Cell proliferation requires metabolic alterations to coordinate catabolism, the maintenance of cellular bioenergetics, with anabolism, the synthesis of macromolecules for cellular replication. Several metabolic pathways serve amphibolic roles, generating both bioenergetic molecules to support metabolic homeostasis, like ATP, NADH, or NADPH, while also producing metabolic precursors used for biosynthesis of molecules such as nucleotides, amino acids, or lipids. Since these metabolic outputs are interconnected, coordinating metabolic pathway activities is critical for efficient cell proliferation. However, specific mechanisms for how metabolic alterations support cell proliferation remain poorly understood, especially in disease contexts where cell lineage, genotype, and environmental factors can all influence metabolic requirements (Vander Heiden and DeBerardinis, 2017).

Central to cell metabolism is the mitochondrion, a double membrane-bound organelle that plays critical roles in cell function. While mitochondria are often viewed through the catabolic lens of efficient ATP synthesis, many proliferating cells can meet their ATP demands from aerobic glycolysis alone yet require intact mitochondrial metabolism for cell proliferation (Weinhouse, 1956; Tan et al., 2015; King and Attardi, 1989). Two fundamental processes of mitochondrial metabolism are the tricarboxylic acid cycle (TCA) and the electron transport chain (ETC). The TCA cycle is an amphibolic pathway, serving as both the major catabolic source of mitochondrial NADH and as a primary anabolic route supporting aspartate biosynthesis. Metabolite progression through the TCA cycle is linked to ETC activity through two mechanisms: (1) the shared succinate dehydrogenase (SDH, also known as complex II) reaction, and (2) redox coupling, where complex I (CI) accepts electrons from mitochondrial NADH to yield NAD+, in turn driving oxidative TCA cycle reactions that convert NAD +to NADH. Disruptions to ETC complexes I, III, and IV diminish mitochondrial NAD +regeneration, reducing the cellular NAD+/NADH ratio, impairing the TCA cycle, and slowing cell proliferation. Supplementation with molecules that regenerate NAD+, such as pyruvate, alpha-ketobutyrate, or duroquinone, can restore the NAD+/NADH ratio and the proliferation of ETC impaired cells (Birsoy et al., 2015; Luengo et al., 2021; Sullivan et al., 2015). Similarly, heterologous expression of the bacterial NADH oxidase LbNOX can also reestablish NADH oxidation independent of the ETC and restore the proliferation of ETC impaired cells (Titov et al., 2016). These findings highlight that redox regulation is an essential function of ETC activity in proliferating cells.

A major metabolic consequence of decreased NAD+/NADH upon ETC impairment is depletion of the amino acid aspartate (Birsoy et al., 2015; Sullivan et al., 2015). Most cells are dependent on de novo aspartate synthesis since aspartate is poorly permeable at physiological concentrations (Garcia-Bermudez et al., 2018; Sullivan et al., 2018). Aspartate production occurs through transamination of the TCA cycle metabolite oxaloacetate, which is typically produced by NAD+-dependent reactions in the TCA cycle, linking aspartate production to NAD +regeneration by the ETC. Aspartate is a central metabolic node in proliferative metabolism, serving as a direct substrate for protein synthesis and as a precursor for the synthesis of other essential metabolites, including asparagine, arginine, and both purine and pyrimidine nucleobases. Thus, aspartate depletion likely impairs core metabolic processes necessary for cell proliferation. Indeed, providing cells with exogenous sources of aspartate circumvents the proliferative defects of impairments to complexes I, III, and IV without correcting the NAD+/NADH imbalance, indicating that ETC impairments block cell proliferation primarily by suppressing aspartate synthesis (Birsoy et al., 2015; Sullivan et al., 2015). Nonetheless, it remains unclear how alterations to other components of mitochondrial metabolism affect redox homeostasis and aspartate production, as well as how cells with mitochondrial dysfunction may adapt to the disruption of canonical aspartate production pathways.

The dual ETC/TCA cycle enzyme SDH serves important functions in mitochondrial metabolism yet is also a target of biological disruption. SDH is a heterotetrameric nuclear-encoded protein complex located in the inner mitochondrial membrane (IMM) that catalyzes the oxidation of succinate to fumarate and shuttles electrons to ubiquinone in the ETC. Biallelic loss-of-function mutations to the subunits of SDH (A-D) or SDH assembly factors (SDHAF1-2), can lead to neoplasms in neuroendocrine tissues and the kidney (Astuti et al., 2001; Bardella et al., 2011; Baysal et al., 2000; Burnichon et al., 2010; Hao et al., 2009; Niemann and Müller, 2000). Cancer cells arising from mutations in SDH components typically display a complete loss of oxidative TCA cycle metabolism at the SDH step and have a substantial accumulation of the upstream metabolite succinate, which can impair enzymes involved in oxygen sensing, epigenetic regulation, and metabolism (Letouzé et al., 2013; Sullivan et al., 2016). SDH loss also blocks the canonical route for aspartate biosynthesis, making it unclear how SDH-deficient cells adapt to fulfill their anabolic demands for cell proliferation. Understanding how SDH-deficient cancer cells reconfigure their metabolism to support cell proliferation will help reveal the normal metabolic roles of SDH and could support the development of novel therapies targeting the unique metabolic liabilities of SDH-mutant cancers.

Here, we investigate the consequences of acute and chronic SDH deficiency and observe that SDH impairment causes aspartate-dependent proliferation defects. However, unlike impairments to other ETC complexes, exogenous electron acceptors are insufficient to restore aspartate levels and cell proliferation to SDH-deficient cells. Surprisingly, we find that additional impairment to CI is required to enable aspartate synthesis and cell proliferation in SDH-deficient cells. Specifically, we find that CI inhibition decreases mitochondrial NAD+/NADH, which drives pyruvate carboxylation and reductive carboxylation of glutamine-derived alpha-ketoglutarate (AKG) to generate precursors for aspartate synthesis in the cytosol, thereby supporting proliferation in SDH-impaired cells. Finally, we observe that disrupting or restoring SDH prompts selective pressure for progressive changes in CI activity to directly correspond to SDH activity, identifying distinct metabolic states that enable aspartate production. Altogether, our data reveal a metabolic mechanism where compartmentalized redox changes are required to enact alternative aspartate biosynthetic pathways that support cell proliferation during TCA cycle dysfunction.

Results

SDH inhibition impairs proliferation, which is partially restored by electron acceptors but robustly restored by aspartate

To understand the metabolic contributions of SDH/complex II (hereafter, SDH) to cell proliferation, we compared the proliferative consequences of the inhibitors atpenin A5 (AA5), a specific inhibitor of SDH, and the classic CI inhibitor rotenone in the respiration-intact osteosarcoma cell line 143B. Both inhibitors effectively blocked cell proliferation when cultured in media without electron acceptors (DMEM without pyruvate), consistent with an essential role of mitochondrial metabolism in support of cell proliferation (Figure 1A). Previous work from us and others identified that supplementation with the exogenous electron acceptors pyruvate (PYR) or alpha-ketobutyrate (AKB) can restore proliferation to cells with CI impairments by regenerating NAD +to support aspartate biosynthesis (Birsoy et al., 2015; Sullivan et al., 2015). While we confirm that PYR and AKB can robustly restore proliferation upon rotenone treatment, proliferation of AA5 treated cells was only modestly improved by electron acceptor supplementation (Figure 1A). In contrast, aspartate supplementation equivalently, albeit incompletely, restored cell proliferation to both rotenone and AA5-treated cells (Figure 1A). Aspartate is poorly permeable to most cells even at supraphysiological concentrations, so we generated 143B cells expressing the glial-specific aspartate transporter SLC1A3, which we confirm supports aspartate uptake at micromolar extracellular concentrations (Figure 1—figure supplement 1A–C). Notably, improved aspartate uptake with SLC1A3 expression allowed aspartate to fully restore proliferation upon AA5 treatment, regardless of PYR co-treatment (Figure 1—figure supplement 1D). Similarly, orthogonal aspartate acquisition by intracellular expression of the guinea pig asparaginase (gpASNase1) combined with asparagine treatment was also able to restore proliferation to AA5-treated cells (Figure 1—figure supplement 1E–G; Sullivan et al., 2018). These data indicate that aspartate is a primary metabolic limitation of SDH-impaired cells, consistent with previous findings (Cardaci et al., 2015; Lussey-Lepoutre et al., 2015). However, these data also indicate that the metabolic determinants of aspartate limitation upon SDH inhibition are distinct from those of CI loss.

Figure 1 with 1 supplement see all
Download asset Open asset
SDH inhibition blocks proliferation, which is incompletely rescued by electron acceptors but robustly restored by aspartate.

(A) Proliferation rates of 143B cells treated with vehicle (DMSO), 50 nM rotenone, or 5 µM atpenin A5 (AA5) cultured in pyruvate free DMEM with no addition, 1 mM pyruvate (PYR), 1 mM alpha-ketobutyrate (AKB), or 20 mM aspartate (ASP). (B) Aspartate levels measured by liquid chromatography-mass spectrometry (LCMS) metabolomics in 143B cells, comparing normalized aspartate ion counts (a.u.; arbitrary units) from cells treated with vehicle (DMSO), 50 nM rotenone, or 5 µM AA5 in pyruvate free DMEM with no addition or 1 mM PYR for 6 hr (n=3). (C) Proliferation rate (y-axis) versus normalized aspartate levels (x-axis) of 143B cells after 6 hours of the indicated treatments (n=3). Data points without borders correspond to a dose titration of rotenone (25 12.5, 6.25, 3.125, 1.565 and 0 nM) in pyruvate free DMEM. Data points with a black border correspond to cells treated with standard dosing of 50 nM rotenone and 5 µM AA5 in DMEM, with and without pyruvate as indicated. See Figure 1—figure supplement 1K-L for additional information. (D) NAD+/NADH measured by LCMS metabolomics of 143B cells treated with vehicle (DMSO), 50 nM rotenone, or 5 µM AA5 in pyruvate free DMEM with no addition or 1 mM PYR for 6 hours (n=3). (E) Schematic showing that rotenone inhibits the ETC, blocking NADH oxidation causing an indirect TCA cycle impairment that can be overcome by exogenous electron acceptors (PYR/AKB), while AA5 directly blocks the TCA cycle and aspartate synthesis. (F) Western blot for SDHB and α-tubulin (tubulin) in WT UOK269 cells or UOK269 cells with restoration of SDHB (SDHB UOK269) as indicated. Tubulin is used as a loading control. (G) Proliferation of WT UOK269 and SDHB UOK269 cells cultured in pyruvate-free DMEM supplemented with vehicle (H2O), 1 mM PYR, or 20 mM ASP (n=3). Data are plotted as means ± standard deviation (SD) and compared with an unpaired two-tailed student’s t-test. p<0.05*, p<0.01**, p<0.001***, p<0.0001****.

Figure 1—source data 1

Proliferation rates and relative metabolite levels by LCMS in Figure 1.

https://cdn.elifesciences.org/articles/78654/elife-78654-fig1-data1-v3.zip
Figure 1—source data 2

Uncropped western blot for Figure 1F.

https://cdn.elifesciences.org/articles/78654/elife-78654-fig1-data2-v3.zip

To evaluate the metabolic differences between impairments to CI or SDH, we conducted liquid chromatography-mass spectrometry (LCMS) metabolomics on cells treated with either inhibitor, with or without pyruvate supplementation. As expected for SDH impairment, AA5 treatment caused succinate accumulation and depleted fumarate levels (Figure 1—figure supplement 1H and I). Consistent with aspartate being the mediator of proliferation defects from either inhibitor, aspartate levels were depleted by both drugs, fully restored by pyruvate in rotenone treated cells, and only partially increased in cells treated with AA5 (Figure 1B). Thus, although either inhibitor is associated with proliferation defects corresponding with aspartate levels, CI impairments have distinct metabolic phenotypes from SDH inhibition.

While intracellular aspartate levels correlate with proliferation rate in our datasets of mitochondrial impaired cells, the quantitative relationship between the two is non-linear making it unclear how to predict the degree to which changes in aspartate levels will constitute a functionally meaningful change for proliferation rate (Figure 1A and B). To contextualize these findings, we generated a dose response curve measuring the effects of a range of rotenone concentrations on proliferation rate and intracellular aspartate levels, comparing them to SDH inhibition by AA5, with or without PYR treatment (Figure 1—figure supplement 1K and L). Notably, combining these datasets revealed that the correlation between proliferation rate and aspartate levels was consistent across inhibitors and environmental conditions, manifesting as a hyperbolic relationship (Figure 1C). These data are consistent with a model where the phenotypic consequences for changes in aspartate levels depend on the degree to which aspartate limits proliferative cell metabolism. Basal conditions are not aspartate-limited and so decreases in aspartate will only have a small effect on cell proliferation; however, as aspartate becomes metabolically limited (around 0.4, compared to vehicle treated cells in media containing PYR), changes in aspartate have more severe consequences on cell proliferation, until aspartate levels are insufficient to support proliferation (around 0.05). These data therefore provide critical context to understand how changes in aspartate levels correspond to proliferation upon mitochondrial inhibition.

We next investigated how each inhibitor affects intracellular NAD+/NADH, a critical mediator of the effects of CI inhibition on aspartate levels and cell proliferation. Unlike with rotenone treatment, AA5 treatment was not associated with suppression of NAD+/NADH, breaking the direct correlation between NAD+/NADH, aspartate levels, and proliferation rate seen with rotenone and other CI inhibitors (Figure 1D; Gui et al., 2016). Pyruvate supplementation increased NAD+/NADH in all cases; restoring NAD+/NADH in rotenone treated cells to near vehicle-treated levels and increasing NAD+/NADH beyond vehicle treated cells upon AA5 treatment (Figure 1D). These data are consistent with a model where CI impairments suppress aspartate production by lowering NAD+/NADH and thereby slowing the oxidative TCA cycle, whereas SDH inhibition directly blocks the oxidative TCA cycle, impairing aspartate synthesis and NADH production (Figure 1E). Indeed, we measured M+4 aspartate in cells cultured in U-13C glutamine, reflective of aspartate produced via canonical oxidative TCA cycle activity and found that pyruvate treatment restores metabolic progression through the TCA cycle in rotenone-treated cells but not in cells treated with AA5 (Figure 1—figure supplement 1J).

We also evaluated the metabolic effects of alterations to SDH activity in UOK269 cells, a patient derived renal cell carcinoma cell line that arose from a heterozygous point mutation (R46Q) in SDHB and deletion of the other allele (Saxena et al., 2016). Re-expression of wild type SDHB in UOK269 cells had the expected metabolic effects by alleviating the accumulation of succinate and increasing aspartate levels (Figure 1F, Figure 1—figure supplement 1M and N). Similar to 143B cells, an intact TCA cycle ameliorated the dependence of UOK269 cells on pyruvate or aspartate for cell proliferation (Figure 1G). However, in contrast to AA5 treated 143B cells, pyruvate and aspartate treatment were equally effective at restoring cell proliferation in UOK269 cells (Figure 1G). These data suggest that long-term SDH loss may be associated with metabolic adaptations that allow cells to fully take advantage of exogenous electron acceptors to restore aspartate and cell proliferation.

CI inhibition promotes aspartate synthesis and cell proliferation in SDH-deficient cancer cells

We next questioned what metabolic adaptations may occur in SDH-deficient cells to promote cell proliferation. Intriguingly, several cell lines generated with genetic defects in SDH components have diminished respiration rates (Cardaci et al., 2015; Lorendeau et al., 2017; Saxena et al., 2016). Since CI and SDH are functionally independent contributors to the ETC, and CI-derived electrons account for the majority of respiration, it is not inherently obvious that SDH-deficient cells would be unable to maintain robust respiration activity. In support of this, we found that treatment with AA5 only partially impairs respiration after several hours of treatment, whereas rotenone severely abrogates respiration within minutes (Figure 2—figure supplement 1A). One potential explanation for the respiration loss in SDH-mutant cells is the observation that they manifest with decreased CI expression and activity (Cardaci et al., 2015; Lorendeau et al., 2017; Saxena et al., 2016). Indeed, it was reported that dual inhibition of SDH and CI is required to mimic the metabolic phenotype of SDH-mutant cancer cells (Lorendeau et al., 2017). Nonetheless, while these data suggest that CI loss may serve a metabolic role in SDH-deficient cells, no functional benefit for CI loss in this context has yet been described.

We tested if, in the presence of electron acceptors, rotenone could alter the proliferation of cells treated with AA5 and, surprisingly, found it caused a robust restoration of cell proliferation (Figure 2A). Since aspartate levels define the proliferation rate of SDH-impaired cells, we then measured aspartate levels and found that CI co-inhibition partially restored aspartate to SDH-impaired cells to a level consistent with their improved proliferation rate (Figure 1C, Figure 2B). CI co-inhibition also restored proliferation and increased aspartate levels across a panel of diverse human cell lines treated with AA5, including both transformed (A549, HCT116) and non-transformed (293T, TF-1) cells, indicating that this metabolic relationship was generalizable (Figure 2—figure supplement 1B and C). To test these effects genetically, we used CRISPR/Cas9 to generate gene knockouts (KO) of SDHB, the most frequently mutated subunit in SDH-null human cancers (Amar et al., 2007; Badenhop et al., 2004; Klein et al., 2008; Linehan and Ricketts, 2013). While we were able to achieve efficient disruption of SDHB, with ~96% loss in 143B cells, we were initially unable to generate single-cell SDHB KO clones in standard media (Figure 2C). Based on our previous observations, we repeated colony selection with or without rotenone and found that CI inhibition was sufficient for successful generation of clones from Cas9/sgSDHB nucleofected 143B cells (Figure 2D). Notably, every tested clone generated in this manner was devoid of SDHB expression (Figure 2E and Figure 2—figure supplement 1D). We also found that AA5 treatment of an SDHB KO clone had no significant effects on proliferation, aspartate levels or the NAD+/NADH ratio, indicating on-target effects of AA5 (Figure 2—figure supplement 1G). In each SDHB KO cell line generated, proliferation was slowed relative to their parental cells but was improved by treatment with the CI inhibitors rotenone or metformin, or by supplementation with aspartate (Figure 2F, Figure 2—figure supplement 1H). We also restored SDHB expression in a KO clone and found that it rescued proliferation to parental rates and reinstated wild type relationships with rotenone and AA5 treatment (Figure 2—figure supplement 1I and J). We then measured aspartate levels in WT and SDHB KO cells with or without rotenone treatment and found that rotenone decreases aspartate levels in WT cells but increases aspartate in SDHB KO cells, consistent with proliferation restoration (Figure 2G). Finally, to test if this metabolic relationship was an artifact of our cell culture media choice, which contains high levels of pyruvate, we also tested the effects of SDH inhibition in 143B cells cultured in Human Plasma-Like Media (HPLM), which closely matches the nutrient composition of human plasma (Cantor et al., 2017). We found that AA5 still impaired cell proliferation and that rotenone or aspartate co-treatment restored cell proliferation in HPLM (Figure 2—figure supplement 1K). Collectively, our data demonstrate that CI inhibition is sufficient to induce aspartate levels and cell proliferation in diverse cell lines and media conditions upon SDH disruption.

Figure 2 with 1 supplement see all
Download asset Open asset
CI inhibition is sufficient to induce aspartate synthesis and cell proliferation in SDH-deficient cancer cells.

(A) Proliferation rates of 143B cells cultured in DMEM with 1 mM PYR treated with vehicle (DMSO), 50 nM rotenone, 5 µM AA5, or 5 µM AA5 and 50 nM rotenone (n=3). (B) Aspartate levels of 143B cells cultured in DMEM with 1 mM PYR treated with vehicle (DMSO), 50 nM rotenone, 5 µM AA5, or 5 µM AA5 and 50 nM rotenone for 6 hours (n=3). (C) Western blot for SDHB and vinculin from 143B cells 3 days after nucleofection with plasmid GFP (Mock) or sgRNAs and sNLS-SpCas9 for SDHB as indicated. Vinculin is used as a loading control. (D) Number of single cell clones that formed colonies from sgSDHB pool in 143B cells treated with vehicle (DMSO) or 30 nM rotenone during a 30 day period. (E) Western blot for SDHB and vinculin from WT 143B cells and SDHB KO 143B clone 2. Vinculin is used as a loading control. (F) Proliferation rates of SDHB KO 143B cells (clone 2) cultured in DMEM with 1 mM PYR and treated with vehicle (DMSO), 50 nM rotenone, 1 mM metformin, or 20 mM ASP (n=3). (G) Aspartate levels measured by LCMS metabolomics of WT 143B cells and SDHB KO 2 cells cultured in DMEM with 1 mM PYR and treated with vehicle (DMSO) or 50 nM rotenone for 6 hr (n=3). (H) Schematic showing that CI inhibition promotes alternative aspartate synthesis pathways upon SDH disruption. Data are plotted as means ± standard deviation (SD) and compared with an unpaired two-tailed student’s t-test. p<0.05*, p<0.01**, p<0.001***, p<0.0001****.

Figure 2—source data 1

Proliferation rates, relative metabolite levels by LCMS, and other data in Figure 2.

https://cdn.elifesciences.org/articles/78654/elife-78654-fig2-data1-v3.zip
Figure 2—source data 2

Uncropped western blots for Figure 2C and E.

https://cdn.elifesciences.org/articles/78654/elife-78654-fig2-data2-v3.zip

Mitochondrial redox alterations are required for CI inhibition to rescue SDH-deficient cells

We next investigated the metabolic mechanism for how CI impairment could be providing a benefit to SDH-deficient cells. CI activity contributes to several metabolic processes, including NADH oxidation, mitochondrial membrane potential generation, and the reduction of ubiquinone (Figure 3—figure supplement 1A). To ascertain which of these processes is deleterious to SDH-deficient cells, we co-treated cells with AA5 and conditions that target other aspects of the ETC and oxidative phosphorylation. While rotenone directly inhibits CI, the complex III inhibitor Antimycin A similarly impairs NADH oxidation and mitochondrial membrane potential generation but differs from CI inhibition by preventing ubiquinol oxidation (Figure 3—figure supplement 1A). Oligomycin blocks the IMM complex ATP synthase, inhibiting mitochondrial membrane potential utilization, inducing hyperpolarization of the IMM, and thereby causing collateral inhibition of ETC activity. Oligomycin treatment therefore similarly impairs NADH oxidation but differs from CI inhibitors by increasing the mitochondrial membrane potential and slowing electron flux through all three proton pumping complexes (Figure 3—figure supplement 1A). Lastly, hypoxia limits the terminal electron acceptor activity of complex IV, slowing upstream ETC function (Figure 3—figure supplement 1A). Importantly, we found that treatment with antimycin or oligomycin can substitute for CI inhibitors, albeit less effectively, to restore the proliferation and aspartate levels of AA5 treated cells (Figure 3—figure supplement 1B and C). Gratifyingly, since complex III inhibition collaterally impairs both complex I and II, antimycin treatment alone had near identical effects on aspartate levels and proliferation as AA5 and rotenone cotreatment (Figure 3—figure supplement 1B and C). Finally, culturing cells in severe hypoxia (1%) was sufficient to improve the proliferation of SDH-impaired cells, albeit to a smaller degree than CI inhibition, while physiological O2 tensions (3–11%) had little effect on the proliferation of AA5 treated cells (Figure 3—figure supplement 1D). Collectively, these results suggest that the shared effect of each treatment on mitochondrial NAD+/NADH, rather than other downstream effects on ETC function, is likely a primary mechanism by which ETC inhibition supports aspartate production and proliferation in SDH-deficient cells.

To investigate how CI inhibition changes the redox state of SDH-impaired cells, we measured the effects of co-inhibition on NAD+/NADH by LCMS. While AA5 treatment alone increased NAD+/NADH in the presence of pyruvate, as in Figure 1D, the effects of rotenone co-treatment were epistatic to AA5 treatment and reduced cellular NAD+/NADH to that of cells treated with rotenone alone (Figure 3A). The increase in NAD+/NADH in cells upon SDH impairment alone is thus likely a consequence of decreased NADH production in the TCA cycle coinciding with persisting NADH oxidation by active mitochondrial CI. We next adapted a selective membrane permeabilization protocol to rapidly isolate and extract metabolites from the cytosol and mitochondria and used an enzymatic assay that enables low sample inputs to measure NAD+/NADH in each compartment (Figure 3—figure supplement 1E; Lee et al., 2019; Sullivan et al., 2015). As expected, in vehicle-treated conditions, mitochondrial NAD+/NADH was lower than cytosolic NAD+/NADH (Figure 3B; Williamson et al., 1967). In addition, the effects on NAD+/NADH from rotenone and/or AA5 treatment were predominantly observed at their site of action in the mitochondria (Figure 3B). Notably, the absence of changes to cytosolic NAD+/NADH upon mitochondrial inhibitor treatments is expected in this context due to exogenous pyruvate pinning cytosolic NAD+/NADH through lactate dehydrogenase activity.

Figure 3 with 2 supplements see all
Download asset Open asset
CI inhibition decreases mitochondrial NAD+/NADH, which is required for aspartate synthesis and proliferation in SDH-impaired cells.

(A) Whole cell NAD+/NADH measured by LCMS metabolomics of 143B cells cultured in DMEM with 1 mM PYR and treated with vehicle (DMSO), 50 nM rotenone, 5 µM AA5, or 5 µM AA5 and 50 nM rotenone for 6 hr (n=3). (B) Cytosolic and mitochondrial NAD+/NADH measured by enzymatic assay of 143B cells cultured in DMEM with 1 mM PYR and treated with vehicle (DMSO), 50 nM rotenone 5 µM AA5, or 5 µM AA5 and 50 nM rotenone for 6 hr (n=3). (C) Schematic depicting the functions of FLAG-tagged cytoLbNOX and mitoLbNOX in each compartment as indicated. (D) Western blot for FLAG, Vinculin, and SDHB from 143B cells expressing FLAG-tagged cytoLbNOX or mitoLbNOX, in cytosolic or mitochondrial fractions isolated by differential centrifugation. Vinculin is a loading control for cytosol and SDHB is a loading control for mitochondria. (E) Proliferation rates of eGFP, cytoLbNOX, and mitoLbNOX expressing 143B cells cultured in DMEM with 1 mM PYR treated with vehicle (DMSO), 50 nM rotenone, 5 µM AA5, or 5 µM AA5 and 50 nM rotenone (n=3). (F) Aspartate levels measured by LCMS metabolomics of eGFP, cytoLbNOX, and mitoLbNOX expressing 143B cells cultured in DMEM with 1 mM PYR and treated with vehicle (DMSO), 50 nM rotenone, 5 µM AA5, or 5 µM AA5 and 50 nM rotenone for 6 hr (n=3). (G) Schematic showing how compartment-specific mitochondrial redox alterations promote alternative aspartate synthesis pathways in SDH-impaired cells. Data are plotted as means ± standard deviation (SD) with the exception of A which is plotted as means ± standard error of the mean (SEM) and compared with an unpaired two-tailed student’s t-test. p<0.05*, p<0.01**, p<0.001***, p<0.0001****.

Figure 3—source data 1

Proliferation rates, relative metabolite levels by LCMS, and enzymatic assay values in Figure 3.

https://cdn.elifesciences.org/articles/78654/elife-78654-fig3-data1-v3.zip
Figure 3—source data 2

Uncropped western blot for Figure 3D.

https://cdn.elifesciences.org/articles/78654/elife-78654-fig3-data2-v3.zip

We next asked if mitochondrial NAD+/NADH alterations are required for the proliferation benefits of CI inhibition in SDH-impaired cells. To do so, we expressed cytosolic or mitochondrial-targeted FLAG-tagged Lactobacillus brevis NADH oxidase (cytoLbNOX/mitoLbNOX) in 143B cells, an enzyme that directly utilizes molecular oxygen to oxidize NADH to NAD +without requiring ETC function or other carbon substrates (Figure 3C; Titov et al., 2016). Subcellular fractionation and immunoblotting verified successful subcellular targeting, as each construct was enriched in its expected compartment (Figure 3D). Notably, since these cells are cultured in the presence of the cytosolic electron acceptor pyruvate, we thus expected cytoLbNOX to be redundant and ineffective to change cell metabolism, whereas mitoLbNOX would be positioned to disrupt the mitochondrial redox changes from rotenone treatment. Indeed, whereas cells expressing cytoLbNOX behaved similarly to parental cells expressing eGFP, mitoLbNOX expression impeded the restorative effect of rotenone treatment on SDH-impaired cell proliferation (Figure 3E). Moreover, the increase in aspartate levels upon rotenone co-treatment was similarly diminished by mitoLbNOX expression, but not by cytoLbNOX expression (Figure 3F). These data indicate that decreased mitochondrial NAD+/NADH is required for the benefits of CI inhibition when SDH is impaired.

We next asked if decreased mitochondrial NAD+/NADH was sufficient to improve SDH-deficient cell function without CI inhibition. Another determinant of mitochondrial NAD+/NADH is the pyruvate dehydrogenase (PDH) reaction, which converts NAD +to NADH (Figure 3—figure supplement 2A). The drug AZD7545 can disinhibit PDH to increase its activity by blocking regulation by pyruvate dehydrogenase kinases (PDKs) (Morrell et al., 2003). In the context of electron acceptor limitation, AZD7545 can further drive down NAD+/NADH and impair cell function (Luengo et al., 2021) however, our data indicate that lowering mitochondrial NAD+/NADH is beneficial in SDH-deficient cells, so we hypothesized that AZD7545 may instead provide a benefit in this context. To ensure that the AZD7545 mechanism of action functions independently of CI inhibition, we confirmed that AZD7545 treatment does not impair mitochondrial oxygen consumption in AA5 treated cells (Figure 3—figure supplement 2B). Additionally, AA5 treated cells co-treated with AZD7545 maintained and marginally increased 13C glucose-derived cis-aconitate levels, whereas rotenone co-treatment depleted them, highlighting the distinct mechanisms by which these inhibitors drive mitochondrial metabolic changes (Figure 3—figure supplement 2C). Importantly, treatment with AZD7545 also slightly improved cell proliferation and increased aspartate levels in AA5-treated cells (Figure 3—figure supplement 2D and E). Notably, these effects are likely limited by the fact that they are occurring in the presence of intact CI activity, which would likely buffer major changes in mitochondrial NAD+/NADH. Nonetheless, these data provide additional support that compartment-specific mitochondrial redox alterations are both sufficient and required to reprogram aspartate synthesis in SDH-impaired cells (Figure 3G).

Mitochondrial redox alterations promote alternative aspartate synthesis pathways in SDH-impaired cells

While our results demonstrate that decreased mitochondrial NAD+/NADH is a critical driver of aspartate synthesis in SDH-impaired cells, it remains unclear how these redox changes induce aspartate production. One likely contributor is reductive carboxylation of glutamine-derived AKG (RCQ), whose relative contribution to aspartate production is increased in cells with defects in oxidative mitochondrial metabolism (Metallo et al., 2011; Mullen et al., 2011, Mullen et al., 2014; Wise et al., 2011). Another is pyruvate carboxylase (PC), which can have increased expression and activity in SDH-null cancer cells (Cardaci et al., 2015; Lussey-Lepoutre et al., 2015). To ascertain the source of increased aspartate in SDH-deficient cells upon CI inhibition we conducted stable isotope labeling of WT and low passage SDHB KO cells with or without rotenone from U-13C glutamine, which can delineate between aspartate generated from the oxidative TCA cycle or RCQ (Figure 4A). SDHB KO decreased total aspartate levels and abolished oxidative M+4 aspartate production, as expected, and RCQ-derived M+3 aspartate was slightly increased in SDHB KO cells upon rotenone treatment (Figure 4A and B). While RCQ can occur in both the cytoplasmic and mitochondrial compartments through IDH1 and IDH2, respectively, both RCQ routes rely on ATP-citrate lyase (ACLY) to cleave cytosolic citrate to oxaloacetate (OAA) and acetyl-CoA (Figure 4—figure supplement 1A). Therefore, we tested if preventing RCQ with the ACLY inhibitor BMS-303141 (BMS) would alter the proliferation rescue of CI inhibition in SDH-impaired 143B cells. Indeed, BMS had an anti-proliferative effect on AA5 and rotenone co-treated cells, which could be prevented by aspartate supplementation (Figure 4C). Correspondingly, aspartate levels were decreased in SDH/CI inhibited cells when treated with BMS, tracking with the small contribution of RCQ to aspartate production in this context (Figure 4B and D). These data indicate that RCQ is partially responsible for increased aspartate synthesis and cell proliferation upon CI inhibition in SDH-impaired cells.

Figure 4 with 1 supplement see all
Download asset Open asset
Reductive carboxylation and pyruvate carboxylation drive aspartate synthesis in SDH impaired cells upon CI inhibition.

(A) Schematic depicting metabolic pathway usage for isotopologue patterns of aspartate derived from U-13C glutamine. (B) Relative ion counts for all aspartate isotopologues derived from U-13C glutamine measured by LCMS metabolomics in WT and SDHB KO 143B cells treated with either vehicle (DMSO) or 50 nM rotenone in pyruvate-free DMEM with 1 mM AKB for 6 hr (n=3). (C) Proliferation rates of WT 143B cells treated with either vehicle (DMSO); 25 µM BMS-303141 (BMS); 50 nM rotenone; 5 µM AA5; 5 µM AA5 and 50 nM rotenone; 5 µM AA5, 50 nM rotenone and 25 µM BMS; or 5 µM AA5, 50 nM rotenone, 25 µM BMS, and 20 mM aspartate in DMEM with 1 mM pyruvate (n=3). (D) Relative aspartate levels measured by LCMS metabolomics of WT 143B cells treated with either vehicle (DMSO); 25 µM BMS; 50 nM rotenone; 5 µM AA5; 5 µM AA5 and 50 nM rotenone; 5 µM AA5, 50 nM rotenone and 25 µM BMS in pyruvate free DMEM with 1 mM AKB for 6 hr (n=3). (E) Schematic depicting metabolic pathway usage for isotopologue patterns of aspartate derived from U-13C glucose. (F) Relative ion counts for aspartate isotopologues derived from U-13C glucose measured by LCMS metabolomics in WT and SDHB KO 143B cells treated with either vehicle (DMSO) or 50 nM rotenone in pyruvate free DMEM with 1 mM AKB for 6 hr (n=3). (G) Western blot for PC in WT 143B cells, PC KO clone 2, and PC KO clone 2 with PC cDNA added back. Tubulin is used as a loading control. (H) Proliferation rates of PC KO and PC AB 143B cells treated with vehicle (DMSO); 25 µM BMS; 50 nM rotenone; 5 µM AA5; 5 µM AA5 and 50 nM rotenone; 5 µM AA5, 50 nM rotenone and 25 µM BMS; or 5 µM AA5, 50 nM rotenone, 25 µM BMS, and 20 mM aspartate in DMEM with 1 mM PYR (n=3). (I) Relative aspartate levels measured by LCMS metabolomics of PC KO and PC AB 143B cells treated with vehicle (DMSO); 25 µM BMS; 50 nM rotenone; 5 µM AA5; 5 µM AA5 and 50 nM rotenone; 5 µM AA5, 50 nM rotenone and 25 µM BMS; or 5 µM AA5, 50 nM rotenone, 25 µM BMS in pyruvate free DMEM with 1 mM AKB for 6 hr (n=3). (J) Schematic showing two alternative aspartate synthesis pathways induced by CI inhibition in SDH-impaired cells. Data are plotted as means ± standard deviation (SD) and compared with an unpaired two-tailed student’s t-test. p<0.05*, p<0.01**, p<0.001***, p<0.0001****.

Figure 4—source data 1

Proliferation rates and relative metabolite levels by LCMS in Figure 4.

https://cdn.elifesciences.org/articles/78654/elife-78654-fig4-data1-v3.zip
Figure 4—source data 2

Uncropped western blot for Figure 4G.

https://cdn.elifesciences.org/articles/78654/elife-78654-fig4-data2-v3.zip

Next, we measured the contributions of PC activity by culturing WT and SDHB KO cells with U-13C glucose, with or without rotenone treatment (Figure 4E). To avoid potential caveats that may confound labeling patterns from unlabeled media pyruvate, we conducted U-13C glucose isotope tracing experiments in pyruvate-free DMEM with AKB, which serves a similar electron acceptor function as pyruvate but cannot fulfill its carbon fates (Figure 1A; Altea-Manzano et al., 2022; Sullivan et al., 2015). We also confirmed that rotenone is still effective to restore proliferation and aspartate levels of SDH-impaired cells in pyruvate-free DMEM with AKB (Figure 4—figure supplement 1B and C). While glucose-derived carbon was a minor contributor to aspartate in WT cells, the majority of aspartate in vehicle-treated SDHB KO cells was labeled M+3 from glucose, as was most of the aspartate induced by rotenone treatment (Figure 4F). We next investigated the functional contributions of PC to SDH-impaired cells, with or without CI co-inhibition. First, we generated monoclonal PC KO 143B cells, and to control for variation from single cell cloning, we re-expressed PC cDNA in a KO clone to generate PC ‘addback’ (AB) cells to serve as a metabolically wild type control (Figure 4G). Consistent with previously published results, we find that PC KO cells are sensitive to SDH impairment (Figure 4H; Cardaci et al., 2015; Lussey-Lepoutre et al., 2015). Interestingly, rotenone treatment still provided a proliferative benefit to AA5-treated PC KO cells, although the resultant proliferation rate was still decreased compared to PC AB cells (Figure 4H). We hypothesized that this residual proliferation benefit was driven by existing or compensatory RCQ activity, so we added BMS to SDH/CI inhibited PC KO cells, which was sufficient to completely block cell proliferation and was ameliorated by aspartate treatment (Figure 4H). Consistent with the proliferation defects, BMS treatment also blocked the increase in aspartate levels upon rotenone treatment in SDH-impaired PC KO cells (Figure 4I). Collectively, these data demonstrate that both alternative aspartate synthesis pathways, RCQ and PC, contribute to aspartate levels and cell proliferation during SDH deficiency and are activated by CI impairment (Figure 4J).

Mitochondrial pyruvate entry is essential for alternative aspartate synthesis pathways

Both PC and RCQ have reactions that can occur in the cytosol or mitochondria, obfuscating the mechanism by which mitochondrial redox changes from CI inhibition might drive alternative aspartate synthesis pathways in SDH-impaired cells. To gain insight into these compartmentalized factors, we tested the dependence on mitochondrial pyruvate availability by generating knockout and addback clones for MPC1, a required subunit of the mitochondrial pyruvate carrier (Figure 5A). Proliferation of MPC1 KO cells was moderately decreased in vehicle-treated conditions, but completely ablated by SDH impairment and no longer improved by CI co-inhibition, while MPC1 AB control cells mirrored WT cells (Figure 5B). Consistent with MPC activity supporting the alternative aspartate synthesis pathways induced by CI inhibition in SDH-impaired cells, we found that aspartate levels were decreased by AA5 treatment in MPC1 KO cells and were not restored during rotenone cotreatment (Figure 5C). To confirm that the proliferation defects in SDH-impaired MPC1 KO cells were mediated by aspartate depletion, we supplemented MPC1 KO cells with aspartate, and observed robust proliferation rescues during both SDH inhibition and SDH/CI co-inhibition (Figure 5D). Thus, mitochondrial pyruvate import is a critical component of alternative aspartate synthesis pathways in SDH-impaired cells.

Figure 5 with 1 supplement see all
Download asset Open asset
Mitochondrial pyruvate import supports alternative aspartate synthesis in SDH-impaired cells.

(A) Western blot for MPC1 and SDHB in WT 143B cells, two MPC1 KO 143B clones, and MPC1 KO 143B clone 1 with MPC1-V5 cDNA added back (AB). SDHB is used as a loading control. (B) Proliferation rates of MPC1 KO 143B cells (clone 1) compared to MPC1 AB 143B cells treated with vehicle (DMSO), 5 µM AA5, or 5 µM AA5 and 50 nM rotenone in DMEM with 1 mM PYR (n=3). (C) Relative aspartate levels measured by LCMS metabolomics of MPC1 KO 143B cells (clone 1) compared to MPC1 AB 143B cells treated with vehicle (DMSO), 5 µM AA5, or 5 µM AA5 and 50 nM rotenone in pyruvate free DMEM with 1 mM AKB for 6 hr (n=3). (D) Proliferation rates of MPC1 KO 143B cells (clone 2) treated with vehicle (DMSO), 5 µM AA5, 5 µM AA5 and 50 nM rotenone, with and without 20 mM aspartate in DMEM with 1 mM PYR (n=3). (E) Relative AKG levels measured by LCMS metabolomics of MPC1 KO 143B cells (clone 1) compared to MPC1 AB 143B cells treated with vehicle (DMSO), 5 µM AA5, or 5 µM AA5 and 50 nM rotenone in pyruvate free DMEM with 1 mM AKB for 6 hours (n=3). (F) Schematic depicting the metabolic fates of mitochondrial pyruvate in SDH/CI impaired cells: PC activity to synthesize oxaloacetate and GPT2 activity to generate AKG and alanine as a byproduct. Data are plotted as means ± standard deviation (SD) and compared with an unpaired two-tailed student’s t-test. p<0.05*, p<0.01**, p<0.001***, p<0.0001****.

Figure 5—source data 1

Proliferation rates and relative metabolite levels by LCMS in Figure 5.

https://cdn.elifesciences.org/articles/78654/elife-78654-fig5-data1-v3.zip
Figure 5—source data 2

Uncropped western blot for Figure 5A.

https://cdn.elifesciences.org/articles/78654/elife-78654-fig5-data2-v3.zip

Our data suggest that MPC1 KO phenocopies the proliferative and aspartate production consequences of impairing PC and RCQ in SDH-deficient cells, raising the question of how mitochondrial pyruvate entry supports these processes. While it is evident that mitochondrial pyruvate supports PC at least in part by serving as an enzymatic substrate, the requirement of mitochondrial pyruvate for RCQ is less clear. RCQ proceeds by converting glutamine derived glutamate to AKG, which is reductively carboxylated by IDH1 or IDH2, steps toward generating cytosolic citrate that is cleaved to OAA for aspartate synthesis (Figure 4—figure supplement 1A). We noted that MPC1 KO was associated with a depletion in AKG, suggesting that mitochondrial pyruvate entry may support RCQ by driving AKG generation (Figure 5E). Glutamate conversion to AKG can conventionally occur via three mitochondrial enzymes: GDH, which consumes glutamate and NAD +to generate AKG, NADH, and ammonia; GOT2, which converts glutamate and OAA to aspartate and AKG; and GPT2, which converts glutamate and pyruvate to AKG and alanine. The requirement for MPC1 to maintain AKG levels suggests that pyruvate-dependent GPT2 activity is a dominant route for AKG production in these cells, consistent with recent studies (Figure 5—figure supplement 1A; Rossiter et al., 2021; Wei et al., 2022; Weinberg et al., 2010). Indeed, MPC1 KO was also associated with a decrease in alanine levels, corroborating studies that identify GPT2 as a dominant source of alanine production (Figure 5—figure supplement 1B; Rossiter et al., 2021). While AA5 treatment did not alter AKG levels in MPC1 KO cells, the addition of rotenone co-treatment was associated with further depletion of AKG, but not alanine, suggesting that rotenone blocked the secondary path of AKG generation through redox changes that impair the NAD +consuming GDH reaction (Figure 5E, Figure 5—figure supplement 1B). Together, these data support the hypothesis that a reduced mitochondrial environment promotes alternative aspartate synthesis in SDH-deficient cells through activating RCQ and PC activity, both of which require mitochondrial pyruvate import (Figure 5F).

Alternative aspartate synthesis upon CI inhibition in SDH-impaired cells is dependent on GOT1

Since alternative aspartate synthesis pathways occur across intracellular compartments, it is unclear whether aspartate transamination occurs in SDH-impaired cells via the cytosolic glutamic-oxaloacetic transaminase (GOT) isoform GOT1 or its mitochondrial counterpart GOT2. On one hand, the mitochondrial enzymes MPC and PC are required for aspartate synthesis in this context, indicating that aspartate production by GOT2 is a plausible path. Indeed, GOT2-mediated aspartate production is observed in wild-type cells using the oxidative TCA cycle for aspartate synthesis and in hypoxic pancreatic cancer cells (Garcia-Bermudez et al., 2022). However, RCQ derived OAA is produced in the cytosol, suggesting GOT1 may also drive aspartate synthesis, as has been described in cells with other mitochondrial disruptions (Birsoy et al., 2015). Therefore, we next asked whether CI inhibition drives aspartate production in SDH-impaired cells by GOT2 or GOT1. We first generated monoclonal 143B cells with GOT2 KO and re-expressed GOT2 to generate a paired GOT2 AB cell line (Figure 6A). Consistent with published phenotypes, GOT2 KO cells were unable to proliferate in pyruvate-free media (Figure 6—figure supplement 1A; Garcia-Bermudez et al., 2022; Kerk et al., 2022). In our standard media containing pyruvate, we found that GOT2 KO cells have increased sensitivity to SDH inhibition alone but maintained the benefits of CI co-inhibition on proliferation rate and aspartate production, similar to GOT2 AB and WT cells (Figure 6B and C). The enhanced depletion of aspartate upon AA5 treatment in GOT2 KO cells was also functionally related to its stronger proliferation impairment phenotype, as aspartate supplementation was able to phenocopy the proliferation benefits of rotenone co-treatment (Figure 6D). These data indicate that residual aspartate production in SDH-impaired cells is primarily driven by GOT2, but GOT2 is dispensable for the induction of aspartate production in SDH-impaired cells upon CI inhibition.

Figure 6 with 1 supplement see all
Download asset Open asset
GOT1 is required for increased aspartate synthesis in SDH impaired cells upon complex I co-inhibition.

(A) Western blot for GOT2 and Vinculin from WT 143B cells, GOT2 KO 143B cells, and the same GOT2 KO clone expressing GOT2 cDNA (AB). Vinculin is the loading control. (B) Proliferation rates of GOT2 KO 143B cells compared to GOT2 AB 143B cells treated with vehicle (DMSO), 5 µM AA5, or 5 µM AA5 and 50 nM rotenone in DMEM with 1 mM PYR (n=3). (C) Relative aspartate levels measured by LCMS metabolomics of GOT2 KO 143B cells compared to GOT2 AB 143B cells treated with vehicle (DMSO), 5 µM AA5, or 5 µM AA5 and 50 nM rotenone in pyruvate-free DMEM with 1 mM AKB for 6 hours (n=3). (D) Proliferation rates of GOT2 KO 143B cells treated with vehicle (DMSO), 5 µM AA5, 5 µM AA5 and 50 nM rotenone, or 5 µM AA5 and 20 mM aspartate in DMEM with 1 mM PYR (n=3). (E) Western blot for GOT1 and Vinculin from WT 143B cells, GOT1 KO 143B cells, and the same GOT1 KO clone expressing GOT1 cDNA (AB). Vinculin is the loading control. (F) Proliferation rates of GOT1 KO 143B cells compared to GOT1 AB 143B cells treated with vehicle (DMSO), 5 µM AA5, or 5 µM AA5 and 50 nM rotenone in DMEM with 1 mM pyruvate (n=3). (G) Relative aspartate levels measured by LCMS metabolomics of GOT1 KO 143B cells compared to GOT1 AB 143B cells treated with vehicle (DMSO), 5 µM AA5, or 5 µM AA5 and 50 nM rotenone in pyruvate-free DMEM with 1 mM AKB for 6 hr (n=3). (H) Proliferation rates of GOT1 KO 143B cells treated with vehicle (DMSO), 5 µM AA5, 5 µM AA5 and 50 nM rotenone, or 5 µM AA5, 50 nM rotenone and 20 mM aspartate in DMEM with 1 mM PYR (n=3). (I) Schematic showing how aspartate is synthesized via GOT1 from RCQ and PC in SDH-deficient cells when complex I is also impaired. Data are plotted as means ± standard deviation (SD) and compared with an unpaired two-tailed student’s t-test. p<0.05*, p<0.01**, p<0.001***, p<0.0001****.

Figure 6—source data 1

Proliferation rates and relative metabolite levels by LCMS in Figure 6.

https://cdn.elifesciences.org/articles/78654/elife-78654-fig6-data1-v3.zip
Figure 6—source data 2

Uncropped western blots for Figure 6A and E.

https://cdn.elifesciences.org/articles/78654/elife-78654-fig6-data2-v3.zip

We next established monoclonal 143B cells with GOT1 KO and re-expressed GOT1 to generate a paired GOT1 AB cell line (Figure 6E). GOT1 KO cells had a mild proliferation defect compared to GOT1 AB or WT cells, but unlike GOT2 KO cells they were not obviously sensitized to AA5 treatment (Figure 6F). Importantly, SDH-impaired GOT1 KO cells differed from GOT1 AB and WT cells in that CI inhibition no longer provided a proliferative benefit (Figure 6F). Correspondingly, aspartate levels were suppressed by SDH impairment in both GOT1 KO and GOT1 AB cell lines, but CI co-inhibition only increased aspartate levels in GOT1 replete cells (Figure 6G). Notably, CI inhibition still increased malate levels in SDH-impaired GOT1 KO cells, suggesting that the mitochondrial redox changes from CI inhibition still drive the biochemical pathways required for alternative aspartate synthesis in these cells (i.e. PC and RCQ), but that GOT1 is required for terminal aspartate generation (Figure 6—figure supplement 1B). These data also suggest that cytosolic malate dehydrogenase (MDH1) may play a role in alternative aspartate synthesis upstream of GOT1, so we also generated monoclonal MDH1 KO and paired MDH1 AB cells (Figure 6—figure supplement 1C and D). Similar to GOT1 KO cells, we found that neither proliferation rate nor aspartate levels were restored by CI inhibition in SDH-impaired MDH1 KO cells (Figure 6—figure supplement 1E and F). Interestingly, baseline aspartate levels were elevated in both GOT1 KO and MDH1 KO cells, consistent with their conventional aspartate-consuming role in the malate-aspartate shuttle (Figure 6G, Figure 6—figure supplement 1F). However, the observation that GOT1 KO or MDH1 KO prevents CI inhibition from restoring aspartate levels in SDH-impaired cells indicates a reversal of the malate-aspartate shuttle to synthesize aspartate in SDH/CI impaired cells (Figure 6—figure supplement 1C). Finally, we found that the inability of CI inhibition to restore the proliferation of SDH-impaired GOT1 KO cells was due to its lack of aspartate restoration, as aspartate rescues proliferation of SDH/CI impaired GOT1 KO cells (Figure 6H). The effects of each metabolic gene disruption on cell proliferation during SDH or SDH/CI inhibition were also replicated across additional KO clones (Figure 6—figure supplement 1G). Collectively, these results indicate that in SDH-impaired cells, CI inhibition increases mitochondrial NADH to drive alternative aspartate synthesis pathways via PC and RCQ, both of which yield cytosolic oxaloacetate that is converted to aspartate by GOT1 (Figure 6I). Notably, in the case of PC-derived aspartate, the divergent NAD+/NADH ratios in each compartment would support reversal of mitochondrial malate dehydrogenase (MDH2) to reduce OAA to malate, which can be exported to the cytosol and oxidized to OAA by MDH1 prior to transamination to aspartate by GOT1 (Figure 6I, Figure 6—figure supplement 1C).

SDHB-null cells adapt through progressive complex I loss

While conducting these studies, we also generated a monoclonal SDHB KO HEK293T cell line, which, as expected, had a proliferation defect that was ameliorated by CI inhibition (Figure 7—figure supplement 1A and B). Interestingly, after several months of passaging, we observed that the basal proliferation rate of SDHB KO HEK293T cells progressively increased, and that the proliferation benefit of CI inhibition was lost (Figure 7—figure supplement 1B). To understand this phenomenon, we created three categories of cell lines derived from one 143B SDHB KO clone: Early Passage (EP) SDHB KO,<5 passages; Late Passage (LP) SDHB KO,>15 passages; or Addback (AB), where SDHB was re-expressed immediately after SDHB KO clone generation (Figure 7A). We measured proliferation rates of parental cells (WT), AB, EP SDHB KO and LP SDHB KO cells and found that SDHB KO decreased their proliferation rate, but that culturing SDHB KO cells to LP allowed them to adapt to an improved proliferation rate (Figure 7B). Culturing EP SDHB KO cells in the presence of rotenone or aspartate was able to improve cell proliferation, consistent with the effects of acute SDH inhibition with AA5; however, those benefits were largely lost in LP SDHB KO cells (Figure 7B). Consistent with their proliferation rates, aspartate levels were lowest in EP SDHB KO cells and were increased by either rotenone treatment or by culturing to LP (Figure 7C). Also similar to the acute effects of AA5 treatment, EP SDHB KO cells had an increased NAD+/NADH ratio that was diminished upon rotenone treatment; however, the NAD+/NADH ratio was already lowered in LP SDHB KO cells and unaffected by rotenone, suggesting redox adaptations occurred during passaging (Figure 7D). To determine if these adaptive phenotypes were due to the selective proliferation advantages they provide or simply inherent to passaged SDH-null cells, we separately passaged EP SDHB KO cells in rotenone for 15 passages (Figure 7—figure supplement 1C). We found that cells cultured with continuous rotenone treatment remained phenotypically similar to EP cells when CI inhibitors were removed, suggesting the selective pressure is stalled by exogenous CI inhibition (Figure 7—figure supplement 1D).

Figure 7 with 1 supplement see all
Download asset Open asset
Adaptive CI activity loss supports proliferation in SDHB-null cells.

(A) Schematic showing how SDHB KO 143B clones were used to generate SDHB addback cells (AB), early passage SDHB KO cells (EP), or late passage (LP) SDHB KO cells. (B) Proliferation rates of wild-type (WT), SDHB addback (AB), early passage (EP) and late passage (LP) SDHB KO 143B cells cultured in DMEM with 1 mM PYR, treated with vehicle (DMSO), 50 nM rotenone, or 20 mM ASP (n=3). (C) Relative aspartate levels measured by LCMS metabolomics of wild-type (WT), SDHB addback (AB), early passage (EP) and late passage (LP) SDHB KO 143B cells cultured in DMEM with 1 mM PYR and treated with vehicle (DMSO) or 50 nM rotenone for six hours (n=3). (D) Whole cell NAD+/NADH measured by LCMS metabolomics of wild-type (WT), SDHB addback (AB), early passage (EP) and late passage (LP) SDHB KO 143B cells cultured in DMEM with 1 mM PYR and treated with vehicle (DMSO) or 50 nM rotenone for 6 hr (n=3). (E) Gene set enrichment analysis (GSEA) of mitochondrial protein expression in LP SDHB KO 143B cells compared to WT 143B cells using MitoCarta 3.0 pathways. (F) Volcano plot of compiled LP SDHB KO 143B cells compared to WT 143B cells showing all mitochondrial proteins (black), complex II subunits (blue), and complex I subunits (yellow) (n=4). (G) Box plots of a subset of CI subunits showing relative expression in WT 143B, SDHB AB 143B, EP SDHB KO 143B, and LP SDHB KO 143B cells (n=4). (H) Complex I activity assay of WT 143B, EP SDHB KO, and LP SDHB KO cells (n=2). (I) Western blot for NDUFA8 and tubulin from WT 143B, EP SDHB KO, LP SDHB KO, and SDHB AB cells. Tubulin is used as a loading control. (J) Schematic detailing how CI activity is suppressed to generate an ideal mitochondrial redox state in SDH-deficient cells to enable aspartate synthesis. Data are plotted as means ± standard deviation (SD) and compared with an unpaired two-tailed student’s t-test. p<0.05*, p<0.01**, p<0.001***, p<0.0001****.

Figure 7—source data 1

Proliferation rates, relative metabolite levels by LCMS, volcano plot data, and complex I activity values in Figure 7.

https://cdn.elifesciences.org/articles/78654/elife-78654-fig7-data1-v3.zip
Figure 7—source data 2

Uncropped western blot for Figure 7I.

https://cdn.elifesciences.org/articles/78654/elife-78654-fig7-data2-v3.zip

To learn more about the mitochondrial alterations occurring during the passaging of SDHB KO cells, we conducted tandem mass tag (TMT) quantitative proteomics on isolated mitochondria from WT cells, AB cells, EP SDHB KO cells, and LP SDHB KO cells (Figure 7—figure supplement 1E). Unsupervised clustering of protein levels found that each group clustered together and that pairs of SDHB KO and replete cells clustered together (Figure 7—figure supplement 1F). Comparing WT and LP samples, we conducted MitoPathways 3.0 pathway enrichment analysis (Pagliarini et al., 2008; Rath et al., 2021) of mitochondrial protein abundance changes and found that, following complex II proteins, CI proteins were the most statistically significantly different (Figure 7E). Indeed, the abundances of most CI subunits were decreased in LP SDHB KO cells compared to WT cells (Figure 7F). In addition, CI subunit levels progressively decreased from WT/AB to EP SDHB KO to LP SDHB KO and resulted in decreased CI activity, suggesting that SDHB KO cells adapt over time by selecting against CI activity (Figure 7G and H). We confirmed these findings by western blot, where LP SDHB KO 143B cells were associated with an ~80% reduction in the CI subunit NDUFA8 compared to WT cells (Figure 7I). Together, these data suggest that SDH-deficient cells suppress CI activity to generate an ideal mitochondrial redox state that enables aspartate synthesis (Figure 7J).

Complex I activity is deleterious to SDH-mutant cancer cell proliferation and tumor growth

A potential caveat of generalizing these effects to SDH-mutant cancers is that SDH wild type parental cell lines may not recapitulate the intrinsic metabolic features of naturally arising SDH-mutant cancer cells. We thus measured the CI activity of the SDH-null UOK269 cells and found that similar to LP SDHB KO 143B cells, UOK269 cells had substantially decreased CI activity compared to WT 143B cells (Figure 8—figure supplement 1A). We next investigated the metabolic state of late passage (LP) SDHB restored UOK269 cells. We first conducted stable isotope labeling metabolomics from U-13C glucose and U-13C glutamine to measure carbon contributions to aspartate in WT and LP SDHB UOK269 cells. Consistent with the metabolic phenotypes observed in 143B cells, aspartate was predominantly labeled M+3 from U-13C glucose and U-13C glutamine in WT UOK269 cells. SDHB restoration was associated with a loss of glucose-derived aspartate and a restoration of oxidative TCA cycle derived M+4 aspartate from glutamine (Figures Figure 8—figure supplement 1B and C). Consistent with their low CI activity, WT UOK269 cells had minimal NDUFA8 protein expression compared to 143B cells, but interestingly LP SDHB UOK269 cells had restored their NDUFA8 expression (Figure 8A). Correspondingly, LP SDHB UOK269 cells also had substantially increased mitochondrial oxygen consumption and CI activity compared to their SDH-deficient WT counterparts (Figure 8B and C). These results therefore indicate that the adaptations in CI status are reversible to achieve concordant activity with that of SDH.

Figure 8 with 1 supplement see all
Download asset Open asset
Complex I activity is deleterious in SDHB-mutant renal cell carcinoma cells.

(A) Western blot for NDUFA8 and tubulin from WT 143B, WT UOK269, and late-passage (LP) SDHB UOK269 cells. Tubulin is used as a loading control. (B) Mitochondrial oxygen consumption rate of WT UOK269 cells compared to LP SDHB UOK269 cells with indicated injections of 100 nM rotenone and 10 µM antimycin (n=3). (C) Complex I activity assay of WT UOK269 and LP SDHB UOK269 cells (n=3). (D) Mitochondrial oxygen consumption rates of pInducer20-NDI1 expressing UOK269 cells (UOK269 iNDI1) that were pretreated with vehicle (H2O) or 1 µg/mL doxycycline for 40 hr as indicated, followed by injections of 100 nM rotenone and 10 µM antimycin (n=3). (E) Proliferation rates of WT and SDHB UOK269 cells with expression of pInducer20-NDI1 (iNDI1) cultured in DMEM with 1 mM PYR and treated with vehicle (H2O) or 1 µg/mL doxycycline (n=3). (F) Aspartate levels measured by LCMS metabolomics of WT and SDHB UOK269 cells with expression of pInducer20-NDI1 (iNDI1) cultured in DMEM with 1 mM PYR and treated with vehicle (H2O) or 1 µg/mL doxycycline for 24 hr (n=3). (G) Tumor volumes of iNDI1 UOK269 cells implanted into the flanks of immunocompromised SCID mice. After tumors became palpable, mice were fed either standard or doxycycline-containing chow (n=10 tumors per condition). (H) Model demonstrating adaptations to SDH loss and SDH restoration, correlating to aspartate levels and proliferation rate. Data are plotted as means ± standard deviation (SD) except in B and D which are means ± standard error of the mean (SEM) and compared with an unpaired two-tailed student’s t-test. p<0.05*, p<0.01**, p<0.001***, p<0.0001****.

Figure 8—source data 1

Proliferation rates, mitochondrial oxygen consumption values, complex I activity values, relative metabolite levels by LCMS, and tumor volumes in Figure 8.

https://cdn.elifesciences.org/articles/78654/elife-78654-fig8-data1-v3.zip
Figure 8—source data 2

Uncropped western blot for Figure 8A.

https://cdn.elifesciences.org/articles/78654/elife-78654-fig8-data2-v3.zip

We next hypothesized that CI loss was critical for UOK269 cells by enabling aspartate production and cell proliferation, and that restoring CI activity would therefore be deleterious. Mammalian CI activity is dependent on the expression of >50 genes encoding core subunits and assembly factors, so increasing CI activity by traditional gene overexpression is not feasible. Instead, we expressed a doxycycline (DOX) inducible construct of the Saccharomyces cerevisiae NDI1, a rotenone-insensitive CI analog that can restore NADH oxidation and ETC activity in mammalian cells when CI is impaired (Seo et al., 1998; Wheaton et al., 2014). DOX treatment of WT UOK269 cells expressing the inducible (iNDI1) construct resulted in a substantial induction of mitochondrial oxygen consumption that was resistant to rotenone treatment, verifying NDI1 expression and confirming that decreased CI activity is a bottleneck for ETC function in UOK269 cells (Figure 8D). NDI1 expression decreased the proliferation of WT UOK269 cells but had no effect on the proliferation of their SDHB-restored counterparts (Figure 8E). Metabolomics measurements also found that NDI1 expression lowered aspartate levels in WT UOK269 cells, but not in the SDHB-restored UOK269 cells (Figure 8F). Since DOX can act as a mitochondrial inhibitor at some doses, we also confirmed that DOX had no effect on proliferation rates or aspartate levels in the absence of the iNDI1 construct (Figure 8—figure supplement 1D and E). Finally, we tested whether the deleterious effects of CI activity restoration on UOK269 cell proliferation were maintained in a physiological metabolic context by generating mouse xenograft tumors. Notably, DOX-mediated NDI expression also impaired the growth of tumors arising from iNDI1 UOK269 cells (Figure 8G). Together, these findings highlight that cells adapt to SDH loss by selecting for concordant CI loss, which is required to maximize aspartate production, cell proliferation rate, and tumorigenesis in SDH-null cancer cells (Figure 8H).

Discussion

Mitochondria are established contributors to catabolic cell metabolism through their canonical oxidative TCA cycle and ETC activities. In proliferating cells, the anabolic contributions of these processes in NAD+/NADH homeostasis and TCA cycle-derived aspartate production have recently gained appreciation, defining important amphibolic roles for oxidative mitochondrial metabolism (Birsoy et al., 2015; Sullivan et al., 2015; Titov et al., 2016). However, the fact that loss-of-function mutations in genes encoding the TCA cycle enzymes SDH and Fumarate Hydratase (FH) can promote tumorigenesis in some contexts challenges the notion that mitochondrial oxidative anabolism is strictly essential for aspartate production. Here, we investigate the metabolic roles of SDH and find that, in cells that employ oxidative TCA cycle function, SDH activity is primarily required for cell proliferation by supporting aspartate production. Interestingly, we found that cells can tolerate SDH loss if they also have concomitant CI impairment, which supports cell proliferation by decreasing mitochondrial NAD+/NADH to enable alternative aspartate synthesis in the cytosol, driven by PC and RCQ. The linked nature of SDH and CI was also revealed by reciprocal experiments where either SDHB was knocked out in wild type cells or restored in cancer cells arising from SDHB mutations. In both cases, functional SDH selects for cells with high CI activity and impaired SDH selects for cells with low CI activity, suggesting two modalities for mitochondrial anabolic metabolism: (1) CI regenerates NAD +to drive oxidative anabolism via the TCA cycle, or (2) CI loss prevents NADH consumption, supporting NADH-driven reductive anabolism in the mitochondria of cells without canonical TCA cycle function. Overall, these findings support a growing understanding that mitochondrial function has critical roles in supporting cell function beyond its classical role in oxidative ATP synthesis, which can be revealed by divergent phenotypic effects of impairments to discrete components of mitochondrial metabolism (Arnold et al., 2022; Martínez-Reyes et al., 2020; Mullen et al., 2011; Ryan et al., 2021; To et al., 2019).

Our results highlight the interconnectedness of cellular metabolism by demonstrating how intercompartmental metabolic variables can collaborate to promote metabolic adaptations and support cell function. The observations that CI inhibition primarily decreases mitochondrial NAD+/NADH and that only mitoLbNOX blocks the benefits of CI inhibition in SDH-impaired cells localize the site of redox effect to the mitochondria. Interestingly, we find that CI inhibition promotes aspartate synthesis in SDH-deficient cells by driving PC and RCQ activity; however, neither path directly consumes mitochondrial NADH for its function. Therefore, understanding the mechanistic connection between mitochondrial NAD+/NADH and the activation of those pathways is an important goal for future study. Additionally, we note that SDH/CI impaired cells also require a supply of cytosolic electron acceptors to proliferate, indicating that this metabolic state depends on intercompartmental metabolic differences including a reduced mitochondrial compartment and an oxidized cytosolic environment. These results therefore reveal that compartmentally localized redox state changes can functionally alter cell phenotypes, highlighting the need for tools to measure subcellular metabolic states to gain a comprehensive understanding of cell metabolism.

Metabolic changes are linked with tumorigenesis, which is epitomized by cancers deriving from mutations in TCA cycle genes. We found that cancer cells devoid of SDH activity enact further metabolic changes by suppressing CI, which is critical for optimal tumor growth and drives alternative aspartate synthesis pathways through MPC, PC, RCQ, and GOT1, suggesting that these pathways could be potential targets for SDH-mutant cancers. Interestingly, the phenomenon of CI loss has also been observed in several other kidney neoplasias, including FH-mutant type 2 papillary renal cell carcinomas (pRCC), renal oncocytomas driven by mitochondrial DNA mutations, and VHL-mutant clear cell renal cell carcinomas (ccRCC), which are characterized by suppression of oxidative mitochondrial metabolism (Courtney et al., 2018; Crooks et al., 2021; Mayr et al., 2008; Simonnet et al., 2003; Tomlinson et al., 2002). Decreases in CI expression and/or heteroplasmic mutations in mitochondrial DNA CI genes have also been reported across diverse cancers, suggesting that activation of reductive mitochondrial anabolism may be a metabolic feature of those cancers and correspond to similar metabolic vulnerabilities (Gorelick et al., 2021; Kumari et al., 2020; Reznik et al., 2017). On the contrary, CI inhibitors have also been identified as a therapeutic avenue for cancer treatment, which can decrease tumor growth by affecting redox homeostasis and suppressing intratumoral aspartate (Baccelli et al., 2019; Gui et al., 2016; Martínez-Reyes et al., 2020; Molina et al., 2018; Sullivan et al., 2018; Wheaton et al., 2014). Notably, the responses to these mitochondrial inhibitors have been variable across clinical trials and cancer models suggesting that heterogeneity in mitochondrial metabolism may govern the responses to CI inhibitors, and potentially to future inhibitors of reductive mitochondrial anabolism (Janku et al., 2021; Momcilovic et al., 2019; Yap et al., 2019; Yap et al., 2023). Thus, understanding mitochondrial phenotypes will be critical for leveraging cancer cell metabolic changes to improve cancer therapy.

Materials and methods

Cell culture

Cell lines were acquired from ATCC (143B, HEK293T, A549, HCT116), as a gift from Dr. W. Marston Linehan, NCI (UOK269), or as a gift from Dr. Stanley Lee, Fred Hutch (TF-1). 143B cells are deficient in thymidine kinase (TK). Cell identities were confirmed by satellite tandem repeat profiling and cells were tested to be free from mycoplasma (MycoProbe, R&D Systems). Cells were maintained in Dulbecco’s Modified Eagle’s Medium (DMEM) (Gibco, 50–003-PB) supplemented with 3.7 g/L sodium bicarbonate (Sigma-Aldrich, S6297), 10% fetal bovine serum (FBS) (Gibco, 26140079) and 1% penicillin-streptomycin solution (Sigma-Aldrich, P4333). TF-1 cells were also supplemented with the 2 ng/ml Human Granulocyte Macrophage-Colony Stimulating Factor (GM-CSF) (Shenandoah Biotechnology, 100–08). Cells were incubated in a humidified incubator at 37 °C with 5% CO2.

Proliferation assays

Cells were trypsinized (Corning, 25,051 CI), resuspended in media, counted (Beckman Coulter Counter Multisizer 4 or Nexcelom Auto T4 Cellometer), and seeded overnight onto six-well dishes (Corning, 3516) with an initial seeding density of 20,000 cells/well (143B, HEK293T, HCT116, 786-O, A549) or 30,000 cells/well (UOK269, SDHB KO 143B). After overnight incubation, 3–6 wells were counted for a starting cell count at the time of treatment. Cells were washed twice in phosphate-buffered saline (PBS) and 4 mL of treatment media was added. Experiments were conducted in DMEM without pyruvate (Corning 50–013-PB) supplemented with 3.7 g/L sodium bicarbonate 10% dialyzed fetal bovine serum (FBS) (Sigma-Aldrich, F0392) and 1% penicillin-streptomycin solution, with or without 1 mM sodium pyruvate (PYR) (Sigma-Aldrich, P8574), 1 mM 2-ketobutyric acid (AKB) (Sigma-Aldrich, K401), 1–20 mM Aspartic Acid (ASP) (Sigma-Aldrich, A7219), or 1 mM Asparagine (Sigma-Aldrich, A7094). Proliferation assays contain 1 mM PYR unless otherwise noted. Proliferation assays using Human Plasma-Like Medium (HPLM) (ThermoFisher, A4899101) were supplemented with 10% dialyzed FBS, 1% penicillin-streptomycin solution, with or without 20 mM ASP. Drug treatments included rotenone (Sigma-Aldrich, R8875), metformin (Sigma-Aldrich, D150959), piericidin A (Cayman Chemical, 15379), atpenin A5 (Cayman Chemical, 11898; AdipoGen, AG-CN2-0110; Abcam, ab144194; or Enzo Life Sciences, ALX-380–313), doxycycline hydrochloride (Sigma-Aldrich, D3447), AZD7545 (Cayman Chemical, 19282), antimycin A (Sigma-Aldrich, A8674), oligomycin (Sigma-Aldrich, 495455), BMS-303141 (Cayman, 16239; MedChemExpress, HY-16107) and DMSO vehicle (D2650). Cells were incubated in a humidified incubator at 37 °C with 5% CO2 then counted after 4–6 days. Cells for the hypoxia experiments were incubated in a 37 °C humidified hypoxia chamber (Coy Lab Products, 8375065) at various O2 concentrations after drug treatment. Proliferation rate was determined by the following equation: Proliferation rate (doublings per day, 1/d) = (log2(final cell count / initial cell count))/total days.

Lentiviral production and infection

The following plasmids were obtained from DNASU Plasmid Repository: pLenti6.3-V5 DEST_SLC1A3, pLenti6.3-V5-DEST_SDHB, pLX304_PC, pLX304_GOT1, pLenti6.3-V5-DEST_MPC1, pLX304_MDH1, and pDONR201-NDI1. pLX304-gpASNase1 and pDONR-cytoLbNOX were previously described. CytoLbNOX was cloned into pLX304 (Addgene, 25890) and NDI1 was cloned into pInducer20 (Addgene, 44012, gift from Stephen Elledge) using LR Clonase II (Fisher, 11791100). pLentiCMV-GOT2 was constructed by purchasing GOT2 cDNA (Integrated DNA Technologies), which was amplified by PCR and assembled with the pENTR1A vector using NEBuilder HiFI DNA Assembly Cloning Kit (New England BioLabs, E2621). pENTR1A-GOT2 was then used to transfer GOT2 to the pLenti-CMV-Hygro Destination vector (w117-1) (Addgene, 17,454 a gift from Eric Campeau & Paul Kaufman) using Gateway LR Clonase II (Fisher, 11791020). Lentivirus was generated by transfection of HEK293T cells with expression construct plasmid DNA with pMDLg/pRRE (Addgene, 12251), pRSV-Rev, (Addgene, 12253), and pMD2.G (Addgene, 12259) packaging plasmids. and FuGENE transfection reagent (Fisher, PRE2693) in DMEM (Fisher, MT10017CV) without FBS or penicillin-streptomycin. The supernatant containing lentiviral particles was filtered through 0.45 µM membrane (Fisher, 9720514) and was supplemented with 8 µg/µL polybrene (Sigma, TR-1003-G) prior to infection. 143B and UOK269 cells were seeded at 50% confluency in six-well dishes and centrifuged with lentivirus (900 xg, 90 min, 30 °C). After 24 hr media was replaced with fresh media and 48 hr after infection cells were treated with either 1 µg/mL blasticidin (Fisher, R21001), 150 µg/mL hygromycin (Sigma, H7772), or 1 mg/mL G418 (Sigma, A1720) and maintained in selection media until all uninfected control cells died.

Generation of mitoLbNOX expressing cells

LbNOX-FLAG DNA was amplified from pDONR-cytoLbNOX by PCR, an oligonucleotide containing the zmLOC100282174 mitochondrial translocation sequence (MTS), which causes potent mitochondrial localization, (Chin et al., 2018) was purchased (Integrated DNA Technologies) and the pENTR1A vector (Fisher, A10462) was amplified by PCR to create a linear fragment. The three fragments were assembled using the NEBuilder HiFi DNA Assembly Cloning Kit (New England BioLabs, E2621) to generate pENTR1A-mitoLbNOX, which was then used as a donor to transfer mitoLbNOX into pLenti-CMV-Hygro-DEST (w117-1) (Addgene, 17,454 a gift from Eric Campeau & Paul Kaufman) using Gateway LR Clonase II (Fisher, 11791020). Lentivirus was generated with pLentiCMV-mitoLbNOX as described above and 143B cells were infected and then selected in 150 µg/mL Hygromycin B (Sigma, H7772) for 4 days.

LbNOX-FLAG primers:

  • 5’-GCCGCGAGACCGTATGCTCATAAGGTCACCGTG-3’ (fw primer).

  • 5’-CAAGAAAGCTGGGTCTAGTTACTTGTCATCGTCATCC-3’ (rv primer).

pENTR primers:

  • 5’-CTAGACCCAGCTTTCTTGTAC-3’ (fw)

  • 5’-CTGGCTTTTAGTAAGCGAATTC-3’ (rv).

MTS sequence:

5’-CGCTTACTAAAAGCCAGGCCACCATGGCACTGCTTCGCGCCGCCGTTTCAGAACTCAGACGGAGAGGACGGGGTGCGCTTACTCCCCTCCCGGCGCTGTCTAGCTTGCTTTCCTCACTTAGCCCCCGAAGTCCCGCCTCAACGCGCCCAGAGCCAAACAATCCACACGCAGATCGACGCCATGTCATCGCTTTGAGGCGATGCCCCCCACTTCCTGCCTCTGCCGTTCTGGCACCTGAACTCCTGCATGCACGAGGATTGCTCCCGAGACATTGGTCTCATGCCTCTCCCTTGTCCACGTCCTCTTCATCCAGTAGACCAGCAGATAAGGCGCAGTTGACCTGGGTCGATAAATGGATCCCAGAAGCCGCGAGACCGTAT-3’.

Generation of knockout cell lines

Protocol was adapted from Hoellerbauer et al., 2020. Three chemically synthesized 2'-O-methyl 3’phosphorothioate-modified single guide RNA (sgRNA) sequences targeting the gene of interest were purchased (Synthego) and are listed in the table below. Each sgRNA was resuspended in nuclease-free water, combined with SF buffer (Lonza, V4XC-2032), and sNLS-spCas9 (Aldevron, 9212). 2x105 143 B or HEK293T cells were resuspended in the resulting solution containing ribonucleoprotein complexes (RNPs) and electroporated using a 4D-Nucleofector (Amaxa, Lonza) programs FP-133 (143B) and DS-150 (HEK293T). Nucleofected cells were then moved to a 12-well plate (Corning, 3513) and, after achieving confluence, were single-cell cloned by limiting dilution by plating 0.5 cells/well in a 96-well plate. Gene knockout was confirmed using western blots on the nucleofected pool and each single-cell clone used in this study.

GenesgRNA sequence (5’–3’)
SDHBUCGCCCUCUCCUUGAGGCG
AGAAAUUUGCCAUCUAUCGA
CUUUGUUAGAUGUGGCCCCA
PCGCAGGCCCGGAACACACGGA
GCUGGAGGAGAAUUACACCC
ACACCGGCCGCAUUGAGGU
MPC1UUGCCUACAAGGUACAGCCU
GGGCUACUUCAUUUGUUGCG
AUGUCAAAGAAUAGCAACAG
GOT1CAGUCAUCCGUGCGAUAUGC
GCACGGAUGACUGCCAUCCC
CGAUCUUCUCCAUCUGGGAA
GOT2UUUCUCAUUUCAGCUCCUGG
CGGACGCUAGGCAGAACGUA
UCCUUCCACUGUUCCGGACG
MDH1CCAAUCAGAGUCCUUGUGAC
CCACAAGAAUGGCCACAUCC
GUGAAAAUCUUCAAAUCCCA

Oxygen consumption

Oxygen consumption measurements were conducted using an Agilent Seahorse Xfp Analyzer. 143B or UOK269 cell lines were trypsinized and seeded at 2.5x105 cells/well of a Seahorse XFp cell culture miniplate (Agilent, 103025–1000) overnight in 80 µL media. The following day, 100 µL cell culture media was added with or without pretreatment, as indicated. Before the assay Seahorse XFp sensor cartridges (Agilent, 103022–100) were incubated with calibrant solution, following manufacturer’s instructions and loaded with injection solutions yielding the following final concentrations: AA5; 5 µM, rotenone; 100 nM. AZD7545; 5 µM, antimycin A; 10 µM. Following each assay, cells of each well were counted by Coulter Counter and mitochondrial oxygen consumption rate was determined as the measured oxygen consumption rate minus the average of post-antimycin oxygen consumption rates, per 100,000 cells.

Mitochondrial fractionation

A total of 2x107 cells were trypsinized, washed with PBS, and centrifuged (300 xg, 5 min, 4 °C). Pellets were washed with ice-cold PBS and resuspended in Homogenization buffer (10 mM Tris-HCl pH 6.7, 10 mM KCl, 0.15 mM MgCl2, 1 mM PMSF Sigma, 10837091001). Cells were then vortexed, incubated on ice for 2 min, then transferred to a Dounce Homogenizer (Sigma, D9063) and homogenized with 40–60 strokes. Cells were transferred to mitochondrial suspension buffer (10 mM Tris-HCl pH 6.7, 0.15 mM MgCl2, 0.25 mM sucrose, 1 mM PSMF) and centrifuged (700 xg, 10 min, 4 °C). The supernatant was transferred and centrifuged again (10,500 xg, 15 min, 4 °C). The resulting supernatant (cytosolic fraction) was transferred and centrifuged once more, (17,000 xg, 15 min, 4 °C) while the pellet containing the mitochondrial fraction was washed in suspension buffer and centrifuged again (12,000 xg, 15 min, 4 °C). Mitochondrial proteins were then harvested with RIPA buffer (see below).

Western blotting

Protein lysates were harvested in RIPA buffer (Sigma, R0278) supplemented with protease inhibitors (Fisher, A32953). Protein concentration was determined using a Bicinchoninic Acid Assay (Fisher, 23225) using bovine serum albumin (BSA) as a protein standard. Equal amounts of protein were denatured with Bolt 4 x Loading Dye (ThermoFisher, B0007) and Bolt 10 x reducing agent (ThermoFisher, B0004), heated at 95 °C for 5 min, and loaded onto 4–12% by SDS-polyacrylamide gels (Invitrogen, NW04127), apart from MPC1, for which we used 16% Novex Tricine gels for separation of low molecular weight proteins (Invitrogen, EC6695C). After electrophoretic separation, proteins were transferred onto a 0.22 mm nitrocellulose using iBlot2 transfer stacks (Fisher, IB23001) and transferred with the P0 system setting. Membranes were blocked with 5% milk in Tris-buffered saline with 0.1% Tween-20 (TBS-T) and incubated at 4 °C overnight with the following antibodies: anti-FLAG (Sigma, F1804; 1:1000), anti-SLC1A3 (Genetex, GTX20262; 1:500), anti-SDHB (Atlas, HPA002868; 1:1000), anti-PC (Proteintech, 66615–1-IG, 1:1000), anti-MPC1, (Cell Signaling, 14462 S, 1:1000), anti-GOT2 (Proteintech, 14800–1-AP, 1:750), anti-GOT1 (Cell Signaling, 34423 S, 1:1000), anti-MDH1 (Proteintech, 15904–1-AP, 1:5000), anti-NDUFA8 (Atlas, HPA041510; 1:1000), anti-Vinculin (Sigma, SAB4200729; 1:10,000), and anti-Tubulin (Sigma, T6199; 1:10,000). The next morning, membranes were washed three times with TBS-T and the following secondary antibodies were added: 800CW Goat anti-Mouse IgG (LiCOR, 926–32210; 1:15,000), 680RD Goat anti-Rabbit IgG (LiCOR, 926–68071; 1:15,000). Membranes were washed three more times with TBS-T and imaged on a LiCOR Odyssey Near-Infrared imaging system.

TMT-quantitative mitochondrial proteomics

Sample preparation

Mitochondrial fractions of WT 143B, SDHB AB 143B, EP SDHB KO 143B, and LP SDHB KO 143B cells in quadruplicate were fractionated using the protocol described above. Mitochondrial pellets were kept at –80 °C until analysis.

Disulfide bond reduction/alkylation

Protein solutions (100 μg) were diluted to 2 μg/μL in 100 mM ammonium bicarbonate. Protein disulfide bonds were reduced by adding tris (2-carboxyethyl) phosphine to a final concentration of 5 mM and mixing at room temperature for 15 min. The reduced proteins were alkylated by adding 2-chloroacetamide to a final concentration of 10 mM and mixing in the dark at room temperature for 30 min. Excess 2-chloroacetamide was quenched by the addition of dithiothreitol to 10 mM and mixing at room temperature for 15 min.

Methanol-chloroform precipitation and protease digestion

Samples (100 μg) were diluted to 1 μg/μL with 100 mM ammonium bicarbonate in a 1.5 mL Eppendorf low bind tube. Protein precipitation was done as follows: 400 μL of methanol was added to each sample and vortexed for 5 s. A total of 100 μL of chloroform was added to each sample and vortexed for 5 s. A total of 300 μL of water was added to each sample and vortexed for 5 s. The samples were centrifuged for 1 min at 14,000 xg. The aqueous and organic phases were removed, leaving a protein wafer in the tube. The protein wafers were washed with 400 μL of methanol and centrifuged at 21,000 g at room temperature for 2 min. The supernatants were removed, and the pellets were allowed to air dry, but not to complete dryness. The samples were resuspended in 70 μL 100 mM HEPES (pH 8.5) and digested with rLys-C protease (100:1, protein to protease) with mixing at 37 °C for 4 hr. Trypsin protease (100:1, protein to protease) was added and the reaction was mixed overnight at 37 °C.

TMT-labeling

TMTpro16plex labeling reagent (Pierce 500 μg) was brought up in 30 μL acetonitrile and added to the digested peptide solution (100 μg) yielding a final organic concentration of 30% (v/v) and mixed at room temperature for 1 hr. A 2 μg aliquot from each sample was combined, dried to remove the acetonitrile, concentrated on a C18 ZipTip (Millipore) and analyzed via LC/MS as a ‘label check’. The label check was used for two purposes. First, it ensured the TMT labeling efficiency was greater than 95%. Second, sample equalization volumes were determined after summing the total intensity from each sample TMT ‘channel’ and calculating the volume of each sample that should be mixed to provide equal amounts from each sample in the final mixture. After labeling efficiency was determined, the reactions were quenched with hydroxylamine to a final concentration of 0.3% (v/v) for 15 min with mixing. The TMTpro16plex-labeled samples were pooled at 1:1 ratio based on calculated equalization volumes and concentrated by vacuum centrifugation to remove acetonitrile. Half of the material was desalted over an Oasis HLB 3 cc cartridge (Waters) and taken to dryness.

bRP fractionation

Pierce’s High pH Reversed-Phased Peptide Fractionation kit (part# 84868) was used to fractionate the sample (100 μg) into eight fractions with cuts at 5, 7.5, 10.0, 12.5, 15, 17.5, 20.0, 22.5, 25.0, and 50.0% acetonitrile in 0.1% triethylamine following the manufacturer’s protocol. The fractions were dried in a speedvac.

Mass spectrometry analysis

The generated basic reverse phase fractions were brought up in 2% acetonitrile in 0.1% formic acid (20 μL) and analyzed (2 μL) by LC/ESI MS/MS with a Thermo Scientific Easy1200 nLC (Thermo Scientific, Waltham, MA) coupled to a tribrid Orbitrap Eclipse with FAIMS pro (Thermo Scientific, Waltham, MA) mass spectrometer. In-line de-salting was accomplished using a reversed-phase trap column (100 μm×20 mm) packed with Magic C18AQ (5 μm, 200 Å resin; Michrom Bioresources, Auburn, CA) followed by peptide separations on a reversed-phase column (75 μm×270 mm) packed with ReproSil-Pur C18AQ (3 μm, 120 Å resin; Dr. Maisch, Baden-Würtemburg, Germany) directly mounted on the electrospray ion source. A 180-min gradient from 4% to 44% B (80% acetonitrile in 0.1% formic acid/water) at a flow rate of 300 nL/min was used for chromatographic separations. A spray voltage of 2300 V was applied to the electrospray tip in-line with a FAIMS pro source using varied compensation voltage −40,–60, –80 while the Orbitrap Eclipse instrument was operated in the data-dependent mode. MS survey scans were in the Orbitrap (Normalized AGC target value 300%, resolution 120,000, and max injection time 50ms) with a 3 s cycle time and MS/MS spectra acquisition were detected in the linear ion trap (Normalized AGC target value of 100% and injection time 50ms) using CID activation with a normalized collision energy (NCE) of 32% using turbo speed scan. Selected ions were dynamically excluded for 60 s after a repeat count of 1. Following MS2 acquisition, real time searching (RTS) was employed, and spectra were searched against a Human database (UP00005640 Human 120119) using COMET. Searches were performed with settings for the proteolytic enzyme trypsin. Maximum missed cleavages were set to 1 and maximum variable modifications on peptides was set to 3. Variable modifications included oxidation (+15.995 Da on M) with static modifications TMTpro (+304.207 DA on K) and carbamidomethyl (+57.021 on C). Maximum search time was 35ms. Scoring thresholds were set to the following: Xcorr 1.4, dCn 0.1, precursor PPM 10 and charge state 2. TMT Synchronous precursor selection (SPS) MS3 was collected on the top 10 most intense ions detected in the MS2 spectrum. SPS-MS3 precursors were subjected to higher energy collision-induced dissociation (HCD) for fragmentation with an NCE of 45% and analyzed using the Orbitrap (Normalized AGC target value of 400%, resolution 60,000 and max injection time 118ms).

Data analysis

Data analysis was performed using Proteome Discoverer 2.5 (Thermo Scientific, San Jose, CA). The data were searched against a Human database (UP00005640 Human 120119) that included common contaminants (cRAPome). Searches were performed with settings for the proteolytic enzyme trypsin. Maximum missed cleavages were set to 2. The precursor ion tolerance was set to 10 ppm and the fragment ion tolerance was set to 0.6 Da. Dynamic peptide modifications included oxidation (+15.995 Da on M). Dynamic modifications on the protein terminus included acetyl (+42.–11 Da on N-terminus), Met-loss (–131.040 Da on M) and Met-loss +Acetyl (–89.030 Da on M) and static modifications TMTpro (+304.207 Da on any N-termius), TMTpro (+304.207 DA on K), and carbamidomethyl (+57.021 on C). Sequest HT was used for database searching. All search results were run through Percolator for scoring. Unsupervised clustering of the 16 proteomic samples was performed using the z-score based R (version 4.0.3) package GMD (Zhao and Sandelin, 2012) and visualized with the heatmap.3 function.

Polar metabolite extractions

Adherent cells

For standard metabolic analysis, cells were seeded overnight at 2x105 cells per well of a six-well dish. The next morning, cells were washed twice with PBS and changed to the indicated media supplemented with 10% dialyzed FBS, 1% penicillin-streptomycin, and treatments as indicated, and returned to the tissue culture incubator. After 6 hr, polar metabolites were extracted from cells by three rapid washes with ice-cold blood bank saline, (Fisher, 23293184) then 250 µL of 80% HPLC grade methanol in HPLC grade water containing 2.5 µM D8 Valine (Cambridge Isotope Laboratories, DLM-488) or 13C amino acid mix (Cambridgre Isotope Laboratories, MSK-CAA-1) was added to each well and cells were scraped with the back of a P1000 pipet tip and transferred to Eppendorf tubes. Tubes were centrifuged (17,000 xg, 15 min, 4 °C) and the supernatant containing polar metabolites was transferred to a new centrifuge tube and placed in a centrivap until lyophilized.

Suspension cells

TF-1 cells were seeded at 2.5x105 cells per well of a ix-well dish the night before treatment and extraction in standard media containing 2 ng/mL human GM-CSF. The next day, cells were washed with PBS and treated with media containing 2 ng/mL human GM-CSF, 10% dialyzed FBS, and the treatments as indicated. After 6 hr, cells were collected and washed in 1 mL of ice-cold saline containing 2% dialyzed FBS. Metabolism was quenched and metabolites were extracted with 600 µL of 80% HPLC grade methanol in HPLC grade water. Samples were centrifuged, (17,000 xg, 15 min, 4 °C) moved to new Eppendorf tubes, lyophilized, and stored at –20 °C until analysis.

Isotope Tracing

Aspartate uptake

SLC1A3 143B cells were plated at 2x105 cells per well of a six-well dish. The next morning, cells were washed twice with PBS and changed to DMEM without pyruvate supplemented with 10% dialyzed FBS, 1% penicillin-streptomycin, and treated with the indicated concentrations of 1,4-13C2 aspartate (Cambridge Isotopes, CLM-4455) for 1 hr.

Glutamine tracing

WT and SDHB KO 143B cells were seeded at 2x105 cells and UOK269 cell lines were plated at 3x105 cells per well of a six-well dish. The next morning, cells were washed twice with PBS and changed to DMEM without glucose, glutamine, pyruvate, or phenol red (Sigma, D5030) supplemented with 10% dialyzed FBS, 1% penicillin-streptomycin, 1 mM pyruvate, 25 mM 12C glucose (Sigma, G7528), and 4 mM U-13C glutamine (Cambridge Isotopes, CLM-1822). 143B cells were treated as indicated for 6 hr.

Glucose tracing

WT and SDHB KO 143B cells were seeded at 2x105 cells per well of a six-well dish. The next morning, cells were washed twice with PBS and swapped to DMEM without glucose, glutamine, pyruvate, or phenol red supplemented with 10% dialyzed FBS, 1% penicillin-streptomycin, 1 mM AKB, 25 mM U-13C glucose (Cambridge Isotopes, CLM-1396), and 4 mM 12C glutamine (Sigma, G5792). Polar metabolites were extracted with the above technique.

Rapid subcellular NAD+/NADH detection

Rapid subcellular fractionation

143B cells were plated at 1x105 cells per well of a six-well dish the night before rapid subcellular fractionation, a protocol adapted from Lee et al., 2019. Cells were washed twice with PBS and treated with DMEM containing 10% dialyzed FBS, 1% penicillin-streptomycin, 1 mM pyruvate, and treatments as indicated. After 6 hr of treatment, cells were washed with ice cold PBS and scraped into a 1.5 mL microcentrifuge tube. After a brief centrifuge (13,500 xg, 10 s, 4 °C), the supernatant was discarded, and the cells were resuspended in 250 µL of digitonin (Sigma, D141) dissolved at 1 mg/mL in ice cold PBS and triturated five times with a P1000 micropipette tip. The sample was then centrifuged (13,500xg, 10 s, 4 °C) and the supernatant was collected for the cytosolic fraction, while the remaining pellet contained the mitochondria. The cytosolic fraction was mixed 1:1 with 250 µL of 0.2 N NaOH containing 1% dodecyl trimethylammonium bromide (DTAB; Sigma, D5047). Metabolites for the mitochondrial fraction were extracted immediately using 500 µL of cold lysis buffer (1% DTAB in 0.2 N NaOH at 1:1 (v/v) with PBS). Whole cell lysates were obtained by the same manner as the mitochondrial fraction.

Compartmentalized NAD+/NADH measurements

NAD +and NADH were then detected using the NAD/NADH-Glo Assay (Promega, G9072) according to the manufacturer’s directions and previous reports (Gui et al., 2016; Sullivan et al., 2015). To detect NAD+, 50 µL of sample was transferred to a PCR tube and treated with 25 µL of 0.4 N HCl and heated to 60 °C for 15 min where acidic conditions selectively degrade NADH. For NADH detection, 50 µL of sample was transferred to a PCR tube and heated to 75 °C for 30 min for selective basic degradation of NAD+. Following incubation, samples were equilibrated at room temperature, then quenched by neutralizing the NAD +acid-treated samples with 25 µL of 0.5 M Trizma base (Sigma, T6066) and the NADH base-treated samples with 50 µL of 1:1 (v/v) 0.4 N HCl:0.5 M Trizma base. Once neutralized, 50 µL of each sample was then mixed with 50 µL of NAD/NADH-Glo Detection Reagent in an opaque white 96-well flat bottom plate. Selective degradation protocols and detection in the linear range were confirmed using chemical standards for NAD+ (Sigma, N1511) and NADH (Sigma N8129). Enzyme-linked luminescence measurements were then recorded on a Tecan Infinite M Plex microplate reader according to the manufacturer instructions.

Liquid chromatography-mass spectrometry

Lyophilized samples were resuspended in 80% HPLC-grade methanol in HPLC-grade water and transferred to liquid chromatography-mass spectrometry (LCMS) vials for measurement by LCMS. Metabolite quantitation was performed using a Q Exactive HF-X Hybrid Quadrupole-Orbitrap Mass Spectrometer equipped with an Ion Max API source and H-ESI II probe, coupled to a Vanquish Flex Binary UHPLC system (Thermo Scientific). Mass calibrations were completed at a minimum of every 5 days in both the positive and negative polarity modes using LTQ Velos ESI Calibration Solution (Pierce). Polar Samples were chromatographically separated by injecting a sample volume of 1 μL into a SeQuant ZIC-pHILIC Polymeric column (2.1x150 mm 5 mM, EMD Millipore). The flow rate was set to 150 mL/min, autosampler temperature set to 10 °C, and column temperature set to 30 °C. Mobile Phase A consisted of 20 mM ammonium carbonate and 0.1% (v/v) ammonium hydroxide, and Mobile Phase B consisted of 100% acetonitrile. The sample was gradient eluted (%B) from the column as follows: 0–20 min.: linear gradient from 85 % to 20 % B; 20–24 min.: hold at 20% B; 24–24.5 min.: linear gradient from 20 % to 85 % B; 24.5 min.-end: hold at 85% B until equilibrated with ten column volumes. Mobile Phase was directed into the ion source with the following parameters: sheath gas = 45, auxiliary gas = 15, sweep gas = 2, spray voltage = 2.9 kV in the negative mode or 3.5 kV in the positive mode, capillary temperature = 300 °C, RF level = 40 %, auxiliary gas heater temperature = 325 °C. Mass detection was conducted with a resolution of 240,000 in full scan mode, with an AGC target of 3,000,000 and maximum injection time of 250 msec. Metabolites were detected over a mass range of 70–1050 m/z. Quantitation of all metabolites was performed using Tracefinder 4.1 (Thermo Scientific) referencing an in-house metabolite standards library using ≤5 ppm mass error. Metabolite abundances were quantified by comparing ion counts relative to cells reflecting the most unmodified metabolic state for each experiment (WT cells, vehicle treatment, AB cells, as indicated) from cells cultured in DMEM containing an electron acceptor (PYR or AKB, as indicated). In situations where the LCMS dataset contained one or more samples had high sample loading variance, all samples in that dataset were normalized to isotopically labeled loading controls (either valine D8 or U-13C aspartate from 13C amino acid mix) as annotated in Supplementary file 1. Isotopically labeled standards for NAD +and NADH are not widely available, so LCMS experiments measuring NAD+/NADH ratios do not incorporate differences in ionization efficiencies or matrix effects for each analyte. Thus, while changes in the ratio of measured ion counts for NAD +and NADH reflect biological changes, they should not be interpreted as an absolute intracellular ratio, which is more accurately quantified by enzymatic methods. Individual ion counts for NAD +and NADH are available in Supplementary file 1. Data from stable isotope labeling experiments includes correction for natural isotope abundance using IsoCor software v.2.2.

Complex I activity assay

Complex I activity was measured using an assay kit (Abcam, ab109721) according to the manufacturer’s instructions. Absorbances at 450 nm were read every 30 s for 40 min on a Tecan Infinite M Plex microplate reader.

Gene set enrichment analysis

Gene Set Enrichment Analysis (GSEA) (Subramanian et al., 2005) was conducted on the proteomics dataset utilizing 149 hierarchical, mitochondrial specific pathways curated by MitoCarta3.0 (Rath et al., 2021) as the a priori defined gene/protein sets. Proteins were ranked by the signed (negative or positive by the direction of the fold-change), -log of the adjusted p-value, both of which were calculated during differential expression analysis between LP and WT cells and used as input for GSEA. Individual MitoCarta3.0 pathways were evaluated for enrichment of upregulated proteins (positive normalized enrichment score) and downregulated proteins (negative normalized enrichment score). GSEA analysis was implemented in R (R Core Team, 2020; version 4.0.3) using the Bioconductor package fgsea (Korotkevich et al., 2021). Adjusted p-values and normalized enrichment scores have been reported.

Mouse xenografts

All mouse work was performed in accordance with FHCC-approved IACUC protocol 51069 and AAALAS guidelines and ethical regulations. iNDI1 UOK269 cells were suspended in 100  μl PBS and injected subcutaneously into both flanks of ten 6–8 week-old, NSG (NOD.Cg-PrkdcscidIl2rgtm1Wjl/SzJ) male mice at 5 × 106 cells per injection. When tumors became palpable, one cage (n=5 mice per cage) was randomly assigned DOX chow with 625 mg/kg doxycycline hyclate (Envigo, TD.01306), while the other continued with standard chow. Tumor volumes were measured by calipers in two dimensions and estimated volumes were calculated using the equation V = (π/6)(L x W2).

Statistical analysis

All graphs and statistical analyses were made in GraphPad Prism 9.0. Technical replicates, defined as parallel biological samples independently treated, collected, and analyzed during the same experiment, are shown. Experiments were verified with independent repetitions showing qualitatively similar results. Details pertaining to all statistical tests can be found in the figure legends.

Data availability

All data generated or analyzed during this study are included in the manuscript and supporting files. Source data files have been provided for Figures 1–8. Ion counts from metabolomics experiments are available in Supplementary file 1, separated by tabs for each graph. The mitochondrial proteomics dataset is provided in Supplementary file 2.

References

Decision letter

  1. Mark S Sharpley
    Reviewing Editor; Cedars-Sinai Medical Center, United States
  2. Utpal Banerjee
    Senior Editor; University of California, Los Angeles, United States
  3. Heather Christofk
    Reviewer; University of California, Los Angeles, United States

Our editorial process produces two outputs: (i) public reviews designed to be posted alongside the preprint for the benefit of readers; (ii) feedback on the manuscript for the authors, including requests for revisions, shown below. We also include an acceptance summary that explains what the editors found interesting or important about the work.

Decision letter after peer review:

Thank you for submitting your article "Mitochondrial Redox Adaptations Enable Aspartate Synthesis in SDH-deficient Cells" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation has been overseen by Utpal Banerjee as the Senior Editor. The following individual involved in review of your submission has agreed to reveal their identity: Heather Christofk (Reviewer #2).

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

The manuscript has been reviewed by three reviewers who were all in agreement that the work has potential. However, all of the reviewers identified a number of issues with the manuscript that will need to be addressed for it to be accepted by eLife. The authors will see from the essential revisions and from the individual recommendations that a lot of work is expected and it will not be straightforward for the authors to sufficiently improve the manuscript within a reasonable period of time. However, the reviewers all agree that the authors should be given the opportunity to revise the manuscript if they wish to do so. Yet, given the extensive nature of the requested revisions, the reviewers ask that the authors carefully consider whether trying to revise the manuscript for publication in eLife is the best option for them and it might be that submission to another journal is preferable.

Essential revisions:

1) The reviewers feel that the data linking Asp to rescue of proliferation are mostly correlative. Much more evidence is needed to conclude that the small changes in aspartate abundance are sufficient to restore proliferation and the presented correlations between aspartate levels and proliferation are not sufficient evidence that aspartate biosynthesis is the primary mode of supporting cell proliferation in SDH-deficient cells.

Similarly, comparison of Asp rescue between different conditions is also challenging and the authors are inconsistent in their interpretation of meaningful changes in aspartate. For example, the recovery of Asp caused by "AA5 + pyruvate" (Figure 1D) appears to be greater than that caused by "AA5 and rotenone" (Figure 2B), yet the recovery of proliferation by "AA5 + pyruvate" is less. As another example, Asp levels in "AA5+mitoLbNOX" (Figure 3E) appear greater than Asp levels in "AA5+ rotenone" (Figure 2B). However, only "AA5 + rotenone" rescues proliferation. In other words, the relationship between Asp recovery and proliferation is not clear.

As one further example, in Figure 1E, the authors investigated the M4 aspartate isotopomer derived from U-13C-glutamine and conclude that pyruvate is insufficient to restore progression through the TCA cycle in AA5 treated cells because it resulted in only a 2.8X increase; yet in Figure 2B the authors observed a 2.3X increase in aspartate levels in cells treated with AA5 + rotenone, which leads the authors to conclude that CI and SDH co-inhibition partially restored aspartate levels.

Can the authors more clearly describe how they are determining their thresholds and when (and how much) an increase in Asp is considered noteworthy?

As a result of this uncertainty in the extent of Asp recovery, the proposed rescue of Asp as a factor in facilitating proliferation in SDH and CI inhibition is not convincing. Right now, a key question can't be assessed by the reader: how much Asp is needed for proliferation? The quantitative relationship between Asp levels and proliferation should be made clearer.

2) A full mechanistic model should be clearly presented by the authors and the manuscript would be strengthened if the authors would take the next steps and provide experimental support for potential mechanisms. For example, currently, it is not clear which pathways of Asp synthesis are controlled by mitochondrial NADH (e.g. is reductive carboxylation or PC activity at play here). Moreover, it is not clear in this system, whether Asp regulates proliferation via "reverse" operation of GOT to support biosynthesis or forward "operation" to generate NAD+ for aerobic glycolysis.

More generally, there is a lack of a coherent and consistent explanation for the phenotype and the authors would help the reader if they would present a clear model throughout the manuscript.

3) The data relating NAD+/NADH to the rescue of Asp and proliferation are hard to interpret. On a practical level, the data presentation is not clear and would be easier to interpret if the ion counts for NAD+ and NADH were presented rather than relative ratios.

There are also several conceptual issues with the NAD+/NADH mechanism that need to be cleared up. For example, the data shows that the NAD+/NADH pools are compartmentalized, but their connection to proliferation and aspartate synthesis is not clear from the presented data. Providing additional data would help interpret the role of NAD/NADH in the LbNOX experiments. Are cyto- and mito-LbNox causing the same level of reversal of the NAD+/NADH balance? Is the LbNOX construct just less functional when in the mitochondrial versus the cytosolic form?

Furthermore, the authors should address the variable correlation between the NAD+/NADH ratio and proliferation. For example, in Figure 3G, H, the change in proliferation seems to be relatively minor compared to other proliferation changes presented in this work.

To reiterate the comment above, clarifying and quantifying the mechanistic links between mitochondrial NAD+/NADH and Asp and proliferation is critical if this manuscript is to be published by eLife.

4) The data presentation needs to be clearer and to be more consistent. For example, the Asp recovery experiments are sometimes presented using log scales which make very small changes in Asp levels appear much more substantial than they are; linear scales may be more appropriate and easier to interpret. The metabolite measurement data would also be clearer if presented as normalized ion counts rather than relative amounts; the ion counts should at the very least be provided somewhere in the manuscript.

5) In the current version of the manuscript, key data are not always consistent between panels. For example, the recovery of proliferation in SDH inhibited cells by rotenone appears somewhat inconsistent between experiments (e.g. Figure 2A and Figure S3B). As another example, in Figure 1, SDHB expression in UOK269 cells led to a proliferation rate that was nearly 5X higher than the control cells. Yet in Figure 2l, the proliferation rate of UOK269 cells expressing SDHB is the same as those not expressing SDHB. Further still, in 1F, addition of AA5 without PYR leads to minimal change in the NAD+/NADH ratio, while in 3A, sole addition of AA5 roughly doubles the NAD+/NADH ratio. The lack of consistency between panels is a recurrent problem throughout the manuscript and it is critical that the authors address these inconsistencies.

6) As detailed in the individual responses, several panels lack important controls and the appropriate statistical comparisons are not always included.

7) Some of the data included in the manuscript does not add support to the model. For example, data regarding the combinatorial treatment of cells with AA5+AZD7545 (Figures 3G-H) shows only slight changes in proliferation and aspartate levels. This adds little to the argument that mitochondria-specific alterations in NAD+/NADH and aspartate biosynthesis are responsible for proliferation in the context of SDH inhibition.

8) Could the authors address/discuss the physiological relevance of their findings. Do CI subunits exhibit decreased expression in SDH-deficient tumors in patients? Is this a phenomenon of cell culture or is this also observed in vivo? Could more physiological oxygen tension (or even hypoxia) affect SDH-deficient cell proliferation?

Reviewer #1 (Recommendations for the authors):

1. The Asp recovery experiments are hard to interpret.

The panels showing Asp levels (e.g. in Figure 1D, 2B, 1G, 3E, 1H) would be easier for the reader to interpret if they were shown with a linear scale on the y-axis. On a linear scale the recovery of Asp by rotenone treatment in SDH inhibited cells (Figure 2B) would look quite minor compared to the loss of Asp that is caused by SDH inhibition. The authors also use different non-linear scales to show Asp levels in different panels of the same figure (e.g. Figure 2G and J), which is especially confusing.

Comparison of Asp rescue between different conditions is also challenging. For example, the recovery of Asp caused by "AA5 + pyruvate" (Figure 1D) appears to be greater than that caused by "AA5 and rotenone" (Figure 2B), yet the recovery of proliferation by "AA5 + pyruvate" is less. As another example, Asp levels in "AA5+mitoLbNOX" (Figure 3E) appear greater than Asp levels in "AA5+ rotenone" (Figure 2B). However, only "AA5 + rotenone" rescues proliferation. In other words, the relationship between Asp recovery and proliferation is not clear.

Part of the problem is that the authors use relative ion counts as well as non-linear scales. They also show the data for closely related conditions in different figures and panels. The best way to resolve the problem would be to present the data as absolute values rather than relative ion counts (e.g. fmol of Asp per cell). If this is not available, the authors should present the (normalized) ion counts (i.e. not the relative ion counts) on a linear scale, with error bars representing the variance of the ion counts. The authors do this in Figure 4B and it is much clearer.

Do the authors have the data from a control sample (143B wild-type cells) that could be included in Figure 4B, so that the reader could see how the rescue of Asp by SDHB KO compares to wild-type levels? Incidentally, Figure 4A would also be easier to interpret if the proliferation rate of wild-type 143B cells were also included.

More generally, if the experiments were performed at the same time, it would be helpful if the data for all the conditions were to be collated in a Sup. Table so that the recovery of Asp between different conditions can easily be compared.

As a result of this uncertainty in the extent of Asp recovery, the proposed rescue of Asp as a factor in facilitating proliferation in SDH and CI inhibition is not convincing. Right now, a key question can't be assessed by the reader: how much Asp is needed for proliferation? The quantitative relationship between Asp levels and proliferation should be made clearer.

2. The data relating NAD+/NADH to the rescue of Asp and proliferation is not entirely convincing and is quite hard to interpret (Figure 1F, G).

i) It is not quite clear what is being measured in Figure 1F. The NAD+/NADH ratio reported in the Figure (i.e. ratio is approx. "1" for controls) is far lower than values that are reported in the literature for this cell line using a different approach (e.g. in the corresponding authors previous work with this cell line) or using carefully validated methods to extract NAD from other cell lines (e.g. by the Rabinowitz group, 2018, Antioxid Redox Signal.). It may be too much to ask the authors to redo the experiments using standards (to provide absolute amounts or concentrations), but it would easier to interpret claims about the role of redox state in these processes easier if the ratio reported here more closely reflects the intracellular NAD+/NADH ratio.

Perhaps the authors are reporting a relative ratio (a ratio of ratios?) that is normalized to the control group? If so, the figure legend isn't clear. Furthermore, having a ratio of ratios is very hard to interpret and can't be compared to previous studies or between experiments in this study. Could the authors please report either the NAD+/NADH ratio that they have measured for each group (on a linear scale), or (better) report the measured values for NAD+ and NADH (e.g. fmol per cell).

ii) Could the authors comment on the lack of an increase in the NAD+/NAD ratio in control cells when they are treated with pyruvate. Isn't this unexpected and inconsistent with previous studies? Several previous studies in very similar systems have all shown an increase in the NAD+/NADH following pyruvate treatment (e.g. Gui et al. 2016 Cell Metabolism), confirming many older studies.

iii) The authors use global NAD+/NADH measurements to infer compartment specific changes. The follow-up experiments using LbNOX are quite an indirect way to check whether this is the case. FLIM would provide some indication if the cytoplasmic and mitochondria pools of NAD(P)H are acting as proposed by the authors.

3. A full mechanistic model is not clearly presented by the authors.

The scheme in Figure 1G presumably doesn't reflect the author's full model, and it would be informative for the authors to present their proposed mechanism somewhere in the manuscript. Right now, it is not clear (for example) how the increase of mitochondrial NADH by dual inhibition of SDH and CI leads to an increase in Asp, or how the rescued Asp levels restore proliferation in this system.

The authors address how mitochondrial NADH increase may increase Asp in the Discussion. However, the manuscript would be strengthened if these questions were addressed experimentally. For example, it is straightforward to test whether reductive carboxylation of glutamine or anaplerosis by pyruvate carboxylase is responsible for the rescue of Asp when both SDH and CI are inhibited.

Furthermore, the link between Asp levels and proliferation has been extensively described in other conditions, but not in this specific condition. Given that the role and "direction" of Asp metabolism varies markedly between different conditions (e.g., Birsoy et al. 2015 vs Gaude et al. 2018) the authors should test which model is operating here. E.g. are MDH1 and GOT1 operating in reverse to generate Asp that is used for biosynthetic reactions, or are GOT1 and MDH1 is operating in the forward direction to consume Asp in order to generate NAD+ for aerobic glycolysis (perhaps more plausible here?).

As a whole the manuscript covers a lot of previously covered ground (e.g. loss of SDH causes loss of CI in tumors (e.g. Fendt group)), loss of SDH and CI creates a requirement for Asp synthesis (Gottlieb group). A clear mechanism of how mitochondrial NADH rescues Asp levels and proliferation would contribute much to its novelty.

4. Key data are not always consistent between panels

For example, the recovery of proliferation in SDH inhibited cells by rotenone appears somewhat inconsistent between experiments (e.g. Figure 2A and Figure S3B). Some variability is to be expected, but, in this version of the manuscript, the experimental variability isn't reflected in the error bars.

5. Are the 143B cells used in these experiments capable of pyrimidine salvage?

It would be useful to know whether the 143B line that is used for most of the experiments is thymidine kinase negative (e.g, the 143B cells from ATCC are TK-). If so, this should be clearly stated in the methods, as lack of thymidine salvage will impact how these results are interpreted.

6. The rescue by AZD7545 does not add support to the model

The rescue of proliferation and especially Asp levels by AZD7545 is so small that if is hard to assess its importance. Is a difference between ~27% and ~31% of wild-type Asp levels really meaningful? When placed on a linear scale it would be hard to discern a difference between Asp levels in these groups.

Reviewer #2 (Recommendations for the authors):

1. While quite interesting, the physiological relevance of SDH and ETC complex I inhibition is unclear. Do CI subunits exhibit decreased expression in SDH-deficient tumors in patients? Is this a phenomenon of cell culture or is this also observed in vivo? Could more physiological oxygen tension (or even hypoxia) affect SDH-deficient cell proliferation?

2. The data regarding the importance of aspartate biosynthesis in the context of SDH inhibition are mostly correlative and lack important controls.

a. Figure 1C – How was the dose selected for Asp treatment in the SLC1A3-expressing cells? While the tracing data confirms the increased uptake of Asp relative to non-SLC1A3 expressing cells (S1B-C), the authors do not report Asp levels or indicate how the dosage was selected.

b. Figures 1I and S1K – SDHB expression in UOK269 cells permitted glutamine-dependent aspartate biosynthesis and increased cell proliferation, yet the authors did not report how SDHB expression affected aspartate levels.

c. The authors are inconsistent in their interpretation of meaningful changes in aspartate. For example, regarding Figure 1E, the authors investigated the M4 aspartate isotopomer derived from U-13C-glutamine and conclude that pyruvate is insufficient to restore progression through the TCA cycle in AA5 treated cells because it resulted in only a 2.8X increase; yet in Figure 2B the authors observed a 2.3X increase in aspartate levels in cells treated with AA5 + rotenone, which leads the authors to conclude that CI and SDH co-inhibition partially restored aspartate levels.

d. Is the increase in proliferation upon SDH and CI inhibition dependent on GOT?

3. Overstatements of results reduce clarity of the findings. For example:

a. The authors argument that "SDH inhibition blocks proliferation, which is restored by aspartate but not electron acceptors" (Figure 1 title) seems to be an overstatement. Though aspartate seems to increase cell proliferation to a greater degree than electron acceptors, data presented in 1A and 1C indicate that pyruvate, AKB, and LbNOX are all sufficient to increase cell proliferation.

b. The presented correlations between aspartate levels and proliferation are not sufficient evidence that aspartate biosynthesis is the primary mode of supporting cell proliferation in SDH-deficient cells. For example, in Supp. Figure 2B-C aspartate levels are not rescued to the same extent as proliferation across the panel of cell lines.

c. Data regarding the combinatorial treatment of cells with AA5+AZD7545 (Figures 3G-H) shows only slight changes to proliferation and aspartate levels. This adds little to the argument that mitochondria-specific alterations in NAD+/NADH and aspartate biosynthesis are responsible for proliferation in the context of SDH inhibition.

4. Statistical analysis was frequently missing or improperly conducted. For example, regarding panel 1A the authors noted that AA5 and rotenone treatment decreased cell proliferation and that co-treatment with pyruvate, AKB, LbNox, or Asp restored proliferation. Statistical comparisons were made between AA5 and rotenone for all groups; however, based on the text, the statistical analysis should be restricted to comparisons between groups for each AA5 and rotenone.

5. The data presentation throughout the manuscript is confusing. For example, in some panels the y-axis is log-scale while in complimentary panels the y-axis is linear, and for some graphs the y-axis does not start at 0.

6. In Figure 1, SDHB expression in UOK269 cells led to a proliferation rate that was nearly 5X higher than the control cells. Yet in Figure 2l, the proliferation rate of UOK269 cells expressing SDHB is the same as those not expressing SDHB. Why are the proliferation rates different?

Reviewer #3 (Recommendations for the authors):

During our reading of the manuscript, we felt the model the authors proposed within this work was not consistent or clear throughout the text. We feel that some attention could be made to try to emphasize and clarify this model throughout and would greatly assist other readers of the work. Currently, we feel like we as readers had to do a lot of work to try to make sense of the model and how the data are supporting that model.

The authors state that "this work provides the first mechanistic evidence that CI loss can directly support core metabolic functions":

1) We do not feel the mechanistic data provided in the current form of the manuscript are strong enough to make this claim.

2) Perhaps this is the first instance of loss of CI & II together are beneficial, but there seem to be several other works that seem to make a case that loss of CI is beneficial to core metabolism processes, such as:

a. PMID: 29915083

b. PMID: 18784283

While we appreciate that many of the scripts used to generate data within this work are using pre-existing toolkits, it is still important to present the code used (how the toolkits were used exactly) to aid in the reproducibility of the work. These scripts could be included in the supplement, or hosted on a public repository, such as GitHub. Raw and processed metabolomics data should also be included as supplements, or hosted on a public repository, such as Metabolomics Workbench.

In Figure S4F, the authors state what "unsupervised clustering" was used. This is very generic – what was the exact algorithm used in the hierarchal clustering of the proteomics data and why? For example, scipy.cluster.hierarchy.linkage provide 7 different clustering methods (https://docs.scipy.org/doc/scipy/reference/generated/scipy.cluster.hierarchy.linkage.html)

There are also a couple of changes to the presentation of the figures that would improve the manuscript:

– Figure 3B/F: There is a line showing up at the top of each of these subpanels in our version of the manuscript.

–

This is a very minor point, but perhaps bar plot colors could be more standardized between sub-panels for the different conditions?

In summary, while we do not feel like this manuscript in its current form meets the criteria for publication in eLife, we are certainly positive about its potential and suggest filling in some of the remaining gaps outlined above and some mechanisms to explain the observations would make the work more complete and suitable for publication.

https://doi.org/10.7554/eLife.78654.sa1

Author response

Essential revisions:

1) The reviewers feel that the data linking Asp to rescue of proliferation are mostly correlative. Much more evidence is needed to conclude that the small changes in aspartate abundance are sufficient to restore proliferation and the presented correlations between aspartate levels and proliferation are not sufficient evidence that aspartate biosynthesis is the primary mode of supporting cell proliferation in SDH-deficient cells.

Similarly, comparison of Asp rescue between different conditions is also challenging and the authors are inconsistent in their interpretation of meaningful changes in aspartate. For example, the recovery of Asp caused by "AA5 + pyruvate" (Figure 1D) appears to be greater than that caused by "AA5 and rotenone" (Figure 2B), yet the recovery of proliferation by "AA5 + pyruvate" is less. As another example, Asp levels in "AA5+mitoLbNOX" (Figure 3E) appear greater than Asp levels in "AA5+ rotenone" (Figure 2B). However, only "AA5 + rotenone" rescues proliferation. In other words, the relationship between Asp recovery and proliferation is not clear.

As one further example, in Figure 1E, the authors investigated the M4 aspartate isotopomer derived from U-13C-glutamine and conclude that pyruvate is insufficient to restore progression through the TCA cycle in AA5 treated cells because it resulted in only a 2.8X increase; yet in Figure 2B the authors observed a 2.3X increase in aspartate levels in cells treated with AA5 + rotenone, which leads the authors to conclude that CI and SDH co-inhibition partially restored aspartate levels.

Can the authors more clearly describe how they are determining their thresholds and when (and how much) an increase in Asp is considered noteworthy?

As a result of this uncertainty in the extent of Asp recovery, the proposed rescue of Asp as a factor in facilitating proliferation in SDH and CI inhibition is not convincing. Right now, a key question can't be assessed by the reader: how much Asp is needed for proliferation? The quantitative relationship between Asp levels and proliferation should be made clearer.

We thank the reviewers for bringing up this potential source of misunderstanding related to the relationship between aspartate levels and cell proliferation. We agree that, as previously written, the relationship between the relative levels of aspartate and the corresponding expectation for proliferation effects was not clearly described, leading to confusion about what fold change thresholds are considered meaningful. Compounding this issue was some variability in the LCMS methods used in the early stages of this project that occasionally distorted relative levels, leading to some inconsistencies in the actual numbers generated across experiments (e.g. when samples were too concentrated, aspartate detection could fall outside the linear range for our mass spectrometer, affecting relative levels). While the qualitative conclusions of the manuscript were not affected by those issues, we agree that describing a consistent quantitative relationship between aspartate levels and proliferation would improve the manuscript. To address these important issues, we have (1) improved our LCMS protocols and repeated many experiments and (2) conducted experiments to directly describe the quantitative relationship between ASP levels and proliferation during mitochondrial inhibition.

1) Standardizing aspartate quantification:

To ensure that we are accurately measuring changes in aspartate levels, we first standardized our protocol to ensure uniform cell material inputs by normalizing metabolite extract loading to total cell volume. Cell volumes were measured from parallel wells at the time of extraction using a Coulter Counter, which we have found to be a preferred method for calculating the cell volume from adherent cells (DOI: 10.21769/BioProtoc.4216). For each condition, the volume of metabolite extract containing 1 uL of total cell volume was calculated, aliquoted into new tubes, dried, and resuspended in 150 uL of solvent prior to running. Next, we incorporated an isotopically labeled amino acid mix containing aspartate at 1 μm into the resuspension solvent to serve as an internal standard. Finally, we also measured the linearity of aspartate detection across a standard curve of isotopically labeled amino acid mix. Importantly, we found that aspartate ion counts had a linear relationship with the loaded concentration of standard between 1 x 106 and 5 x 109 ions, and possibly higher (Author response image 1A). Using the standardized protocol described here, we also found that the ion counts of aspartate measured from cells cultured in representative conditions from this study was between 1 x 107 and 1 x109 ions, well within the linear range for linear aspartate quantification. Collectively these data support the conclusion that this method allows for accurate measurements of relative aspartate levels across conditions (Author response image 1B). The methods section was edited to describe these methodological differences (see: “liquid chromatography-mass spectrometry” section).

Author response image 1
Download asset Open asset
Quantitative analysis of aspartate detection by LCMS.

(A) Detection of isotopically labeled aspartate by LCMS across a standard curve of aspartate levels. (B) Measurement of aspartate from 143B cells treated with vehicle, 50 nM rotenone, or 5 µM AA5, with or without 1 mM pyruvate in DMEM for 6 hours. Aspartate ion counts are within the linear range for aspartate detection.

The presentation of aspartate quantification figures was also edited to improve clarity. Including changing to a linear Y axis (with break) instead of a log axis. The linear Y axis also “fold change” values unnecessary and so were removed, since they were an additional source of confusion.

2) Quantifying the relationship between aspartate levels and proliferation:

A critical conclusion from this study is that aspartate levels correlate with, and determine, the proliferation rate across conditions of SDH inhibition. This conclusion is reached by two fundamental observations: (1) The proliferation defects after SDH inhibition are ameliorated by aspartate restoration (Revised Manuscript Figures 1A, 1G, 2F, 4C, 4H, 5D, 6D, 6H, 7B, and Figure 1—figure supplement 1D, 1G, Figure 2—figure supplement 2H, 2K, and Figure 7—figure supplement 1D). (2) The proliferation rate upon SDH inhibition (and related cotreatments) directly correlates with aspartate levels. However, as noted by the reviewers, the relative levels of aspartate do not necessarily correlate linearly with proliferation levels, since a small increase in aspartate can have a large effect on proliferation when cells are proliferating slowly, leading to confusion about what is a meaningful change. To address the fundamental relationship between aspartate levels and proliferation, we generated a standard curve comparing the levels of aspartate and the corresponding proliferation rate in conditions well characterized to cause aspartate limitation – treatment with an ETC complex I inhibitor (PMIDs: 27746050, 26232225, 29941931, 26232224, 29941933, 36411320, 29892070). Indeed, when cultured in DMEM without pyruvate across eight doses of rotenone (0, 0.1, 1.25, 3.125, 6.25, 12.5, 25, 50 nM) cells showed a dose dependent decrease in cell proliferation that correlated with a dose dependent decrease in aspartate levels (Revised Manuscript Figures 1C, Figure 1—figure supplement 1K-L see below). Interestingly, the relationship between aspartate levels and proliferation rate followed a hyperbolic curve, where relatively small increases in aspartate levels cause a large change in proliferation rate when aspartate is very low, and only a minor change in proliferation rate when aspartate is near untreated levels. This curve is consistent with a model where cells have sufficient aspartate when untreated and decreasing aspartate does not immediately cause robust cell proliferation defects until it becomes limiting for a downstream metabolic process (or processes). Our data indicate that, relative to the aspartate levels present in untreated 143B cells (grown in DMEM with 1 mM pyruvate), aspartate only becomes a significant limitation for cell proliferation once depleted to around 0.4, and is unable to support proliferation around 0.05, with a strong correlation between proliferation and aspartate levels between 0.05-0.4. Culturing cells treated with high dose rotenone (50 nM) in media containing the electron acceptor pyruvate restored both aspartate levels and proliferation rate to values consistent with the measured proliferation rate to aspartate levels curve. Finally, we also included measurements of proliferation rate and aspartate levels upon AA5 treatment, with or without pyruvate in the media, and found that they too matched the values observed in rotenone treated cells. Together, these data provide detailed quantitative information about the relationship between aspartate levels and cell proliferation in the context of complex I and SDH inhibition, providing considerable support for our claim that small changes in aspartate levels can impact cell proliferation rate in SDH impaired cells and contextualizing aspartate level changes throughout the study

2) A full mechanistic model should be clearly presented by the authors and the manuscript would be strengthened if the authors would take the next steps and provide experimental support for potential mechanisms. For example, currently, it is not clear which pathways of Asp synthesis are controlled by mitochondrial NADH (e.g. is reductive carboxylation or PC activity at play here). Moreover, it is not clear in this system, whether Asp regulates proliferation via "reverse" operation of GOT to support biosynthesis or forward "operation" to generate NAD+ for aerobic glycolysis.

More generally, there is a lack of a coherent and consistent explanation for the phenotype and the authors would help the reader if they would present a clear model throughout the manuscript.

We concur with the reviewers’ suggestion that identifying the metabolic enzymes and pathways required for aspartate synthesis is a reasonable and important next step for understanding of this metabolic phenotype and for generating a clear mechanistic model. To address these goals we (1) Conducted stable isotope resolved metabolomics from U-13C glucose and U-13C glutamine to ascertain the source of aspartate in each condition; (2) Used genetic knockouts and metabolic inhibitors to identify key pathway components of aspartate synthesis in SDH impaired cells; (3) Generated several model figures to clarify our metabolic model(s) of our findings. These experiments generated a significant amount of new data, necessitating the addition of three new figures, each of which has a linked supplemental figure (Revised Manuscript Figures 4, 5, and 6 and Revised Manuscript Supplemental Figures 4, 5, and 6), as well as seven additional model panels.

1) Stable isotope resolved metabolomics.

Three primary paths of aspartate synthesis have been described in mammalian cells – the oxidative TCA cycle (primarily deriving from glutamine), reductive carboxylation of α-ketoglutarate (primarily deriving from glutamine, “RCQ”), and pyruvate carboxylation (primarily deriving from glucose, “PC”). Since the oxidative TCA cycle is blocked in SDH impaired cells, it raises the question of which metabolic source contributes to increased aspartate levels when complex I is co-inhibited. We first conducted isotope tracing from wild type (WT) or SDHB KO cells cultured in U-13C glutamine for 6 hours, with or without rotenone treatment (Revised Manuscript Figure 4A and B). As expected, glutamine was the predominant carbon source for aspartate in wild type cells, and rotenone predictably decreased aspartate levels and processive TCA cycling, based on M+2 label. Additionally, rotenone treatment increased aspartate levels in SDHB KO cells, as expected, with the only major aspartate isotopotologue species from glutamine being M+3, consistent with being derived from RCQ. While the abundance of M+3 aspartate was increased upon rotenone co-treatment in SDHB KO cells, the total amount of labeled carbon from glutamine remained a relatively small contributor to total aspartate levels in both cases. These data confirm that oxidative TCA cycling is impaired in SDHB KO cells and suggest that RCQ is a modest contributor to increased aspartate upon complex I inhibition.

We also conducted isotope tracing in identical conditions using U-13C glucose. Since unlabeled pyruvate in the media could contaminate the pyruvate labeling from U-13C glucose. we cultured cells in media containing 1 mM AKB, which had similar effects on proliferation as 1 mM PYR (Figure 1A). We also confirmed that rotenone co-treatment still restored cell proliferation and aspartate levels to cells cultured in AKB media instead of PYR media (Revised Manuscript Figure 4—figure supplement 1A and B, below). Consistent with the aforementioned glutamine labeling results, WT cells had minimal labeling from glucose and the small contribution oxidative acetyl-CoA production and processive TCA cycling (M+2) that was diminished by rotenone cotreatment (Revised Manuscript Figure 4E and F, above). However, also consistent with glutamine labeling results, glucose was a dominant contributor to aspartate carbon in SDHB KO cells, with the increase in aspartate levels upon rotenone treatment primarily having M+3 label from U-13C glucose. These data suggest that PC is a major contributor to aspartate levels in 143B SDHB KO cells, consistent with results from prior work (Cardaci et al., 2015), and that complex I inhibition can enhance this aspartate synthesis route in SDH impaired cells.

To corroborate these findings in the patient-derived SDHB mutant UOK269 cells without (WT UOK269) or with SDHB re-expression (SDHB UOK269), we also added U-13C glucose tracing to the U-13C glutamine isotope tracing conducted in the prior version of the manuscript. As in 143B cells, we found that SDH loss was associated with an increased contribution of glucose to aspartate by PC (M+3) and a switch of glutamine contribution to aspartate from the oxidative TCA cycle (M+4) to RCQ (M+3) (Revised Manuscript Figure 8B and C).

1) Genetic knockouts and metabolic inhibitors

a. Dependence on RCQ and PC

As suggested by the reviewers, we investigated which components of alternative aspartate synthesis pathways were required for CI inhibition to drive aspartate synthesis in SDH impaired cells – evaluating contributions from RCQ, PC, GOT enzymes, and related metabolic enzymes. We first evaluated the relative contributions of RCQ and PC to cell proliferation and aspartate levels in SDH impaired cells, with and without CI co-inhibition. To test the dependence of RCQ we leveraged BMS-303141 (BMS) an inhibitor of ACLY, a shared reaction for the RCQ pathway to aspartate regardless of whether RCQ occurs in the mitochondria by IDH2 or the cytosol by IDH1 (See Revised Manuscript Figure 4—figure supplement 1A, below, for metabolic model). We found that BMS treatment caused minor cell proliferation defects and lowered aspartate levels in SDH/CI dual-inhibited cells, and that this minor proliferation defect could be remedied by aspartate co-treatment (Revised Manuscript Figure 4C and D, above). Notably, these effects also track with the minor contributions of RCQ to aspartate synthesis, measured by isotope tracing. We next evaluated the dependence on PC, using CRISPR/cas9 to generate isogenic knockout clones for PC, comparing to paired “Addback” (AB) clones Revised Manuscript Figure 4G, above. We prefer to compare KO to AB cell lines for each gene knock out rather than to parental WT cells, as it better controls for any unrelated phenotypic changes that can result from the process of single cell cloning. As expected, PC KO sensitized cells to the proliferative and aspartate depleting effects of AA5 treatment and somewhat decreased the corresponding proliferative benefits of rotenone co-treatment (Revised Manuscript Figure 4H and I, above). To test whether residual or compensatory RCQ activity was contributing to the enduring aspartate and proliferation rescue in PC KO cells, we also co-treated with BMS and ASP. Importantly, the combination of PC KO and BMS treatment abolished the proliferation of SDH/CI impaired cells, which could be restored by aspartate co-treatment, indicating that the loss of those pathways prevents cell proliferation in SDH/CI impaired cells by decreasing aspartate availability. Correspondingly, while not quite statistically significant, BMS treatment mitigated the increase of aspartate upon rotenone co-treatment in PC KO cells, with the partial effect in PC AB cells mirroring that of wild type cells in Figure 4D. Collectively, these experiments reveal that RCQ and PC are essential contributors to the induction of aspartate synthesis in SDH impaired cells upon CI co-inhibition (Revised Manuscript Figure 4J, above).

b. Dependence on mitochondrial pyruvate import

While not explicitly requested by the reviewers, we sought to dive deeper to mechanistically understand how compartmentalized redox changes could drive alternative aspartate synthesis via RCQ and PC. While PC is conventionally reported to be active in the mitochondria, recent reports have concluded that it may have some cytoplasmic activity (PMID: 34547241) and so we investigated whether mitochondrial pyruvate entry was required for in the induction of aspartate and proliferation upon CI inhibition in SDH impaired cells. To do so, we generated knockout and AB cells for the essential mitochondrial pyruvate carrier subunit, MPC1. MPC1 KO and AB were verified by western blot (Revised Manuscript Figure 5A). We then tested the effects of AA5 with or without rotenone and found that MPC1 KO cells had severe proliferation defects upon AA5 treatment that were not rescued by rotenone co-treatment (Revised Manuscript Figure 5B). Importantly aspartate levels corroborated this result since rotenone co-treatment in AA5 treated MPC1 KO cells had minimal ability to increase aspartate levels (Revised Manuscript Figure 5C). Once again, this inability to increase aspartate upon rotenone treatment was functionally limiting in AA5 and AA5/rotenone treated MPC1 KO cells since aspartate treatment could still restore their cell proliferation (Revised Manuscript Figure 5D). Notably, the loss of mitochondrial pyruvate uptake not only decreases pyruvate for potential PC reactions, but also pyruvate for transamination by GPT2, which generates alanine from pyruvate while converting glutamate to AKG – a necessary substrate for RCQ driven aspartate synthesis. In support of this hypothesis, MPC1 KO cells were associated with depletion of both alanine and AKG in both vehicle and AA5 treated conditions (Revised Manuscript Figures 5E and Figure 5—figure supplement 1B). Notably, AKG levels (but not alanine levels) were further depleted in MPC1 KO cells when co-treated with rotenone, suggesting that residual GDH activity provides some AKG production in MPC1KO cells, which is blocked upon NAD+ depletion by rotenone treatment. Together, these data identify a novel metabolic dependency of SDH deficient cancer cells on MPC that supports both pathways of alternative aspartate synthesis.

c. Dependence on compartmentalized GOT enzymes

Aspartate is generated from oxaloacetate, using glutamate as a nitrogen donor, through either the cytosolic transaminase GOT1 or the mitochondrial transaminase GOT2. Recent studies have found that GOT2 generates aspartate during canonical oxidative TCA cycle metabolism and in PDAC cells cultured in hypoxia (PMID: 35726024), whereas GOT1 can generate aspartate during other forms of mitochondrial dysfunction (PMID: 26232224). We again turned to CRISPR/Cas9 to generate knockout clones for GOT1 or GOT2 and to compare the effects of AA5 with or without rotenone co-treatment, comparing KO cell lines to their restored counterparts (AB). We first tested the effects of GOT2 KO our phenotypes. GOT2 KO and AB were confirmed by western blot and functionally, since GOT2 KO cells cannot grow without pyruvate, as has been recently published (Revised Manuscript Figures 6A and Figure 6—figure supplement 1A) (PMID: 35815941). While GOT2 loss caused an increase in sensitivity to AA5, it had essentially no effect on the proliferation restoration from rotenone co-treatment when compared AB cell lines (Revised Manuscript Figure 6B). Measurement of aspartate levels corroborated these results, where aspartate levels were restored by rotenone in both GOT2 KO and AB cell lines (Revised Manuscript Figure 6C). Finally, the proliferative defects from AA5 treatment in GOT2 KO cells was found to result from aspartate deficiency, as treatment with aspartate could restore proliferation to AA5 treated cells to a similar degree as rotenone co-treatment (Revised Manuscript Figure 6D). We then generated KO and AB cell lines for GOT1, which were confirmed by western blot, and then conducted similar experiments (Revised Manuscript Figure 6E). Notably, GOT1 KO cells did not have notable sensitivity to AA5 treatment, consistent with their dependency on GOT2 in this condition, but the proliferation benefit from rotenone co-treatment was abolished in the absence of GOT1 (Revised Manuscript Figure 6F). Measuring aspartate levels similarly showed that the increase in aspartate levels upon rotenone co-treatment was absent in AA5 treated cells without GOT1, commensurate with their proliferation phenotype (Revised Manuscript Figure 6G). Interestingly, we noted that this defect in aspartate production was downstream of the increased anaplerosis following CI inhibition in SDH impaired cells, as malate levels still increase upon CI/SDH co-inhibition in GOT1 KO cells (Revised Manuscript Figure 6—figure supplement 1B). Nonetheless, the impaired ability to synthesize aspartate was functionally limiting for proliferation since aspartate treatment could still restore the proliferation to AA5/rotenone treated GOT1 KO cells (Revised Manuscript Figure 6H). Finally, we corroborated the finding that cells dependent on GOT1 for aspartate synthesis also depend on MDH1 as well, presumably to generate cytosolic OAA for GOT1 dependent aspartate synthesis (Revised Manuscript Figure 6—figure supplement 1D-F and Figure 6I).

Collectively, the above experiments indicate that the increased aspartate levels and proliferation upon complex I inhibition in SDH impaired cells arise primarily from mitochondrial redox changes driving pyruvate carboxylation and, to a lesser degree, reductive carboxylation of glutamine, both of which depend on mitochondrial pyruvate, to generate cytosolic OAA to ultimately being converted to aspartate in the cytosol via GOT1.

1) Metabolic models

As recommended by the reviewers, we have incorporated our data into seven additional model figure panels to help communicate our understanding of the metabolic pathways and phenotypes relevant to SDH deficient cells, see Revised Manuscript Figures 2H, 3G, Figure 4—figure supplement 1A, Figure 5—figure supplement 1A, Figure 6—figure supplement 1C, 7J, and 8H, (new models for Main Figures 4, 5, and 6 are shown above).

3) The data relating NAD+/NADH to the rescue of Asp and proliferation are hard to interpret. On a practical level, the data presentation is not clear and would be easier to interpret if the ion counts for NAD+ and NADH were presented rather than relative ratios.

There are also several conceptual issues with the NAD+/NADH mechanism that need to be cleared up. For example, the data shows that the NAD+/NADH pools are compartmentalized, but their connection to proliferation and aspartate synthesis is not clear from the presented data. Providing additional data would help interpret the role of NAD/NADH in the LbNOX experiments. Are cyto- and mito-LbNox causing the same level of reversal of the NAD+/NADH balance? Is the LbNOX construct just less functional when in the mitochondrial versus the cytosolic form?

Furthermore, the authors should address the variable correlation between the NAD+/NADH ratio and proliferation. For example, in Figure 3G, H, the change in proliferation seems to be relatively minor compared to other proliferation changes presented in this work.

To reiterate the comment above, clarifying and quantifying the mechanistic links between mitochondrial NAD+/NADH and Asp and proliferation is critical if this manuscript is to be published by eLife.

We are glad to have the opportunity to further clarify the connection between (compartmentalized) NAD+/NADH, aspartate levels, and cell proliferation. The comments raise two general issues, each of which will be addressed in turn: (1) Ion counts vs relative ratios, (2) Evidence for compartmentalized redox changes in the comments about Original Submission Figures3G and H will be incorporated into the discussion about those panels below.

1) NAD+/NADH ion counts, ratios, and relative ratios.

NAD+/NADH is a compound measurement of synthesis pathways of NAD(H) and the redox reactions that govern the NAD+/NADH ratio, which in turn affects other reactions that use those coenzymes. We have previously found that, in the context of mitochondrial inhibition, treatments that alter NAD+ and NADH pool sizes can alter NAD+/NADH ratios and dependent metabolic reactions, but that the ratio change from this approach is temporary as the ratio adjusts to the larger total pool size and that the resulting ratio ultimately governs the phenotypic effects on proliferation (PMID: 27746050). Thus, we believe that displaying the NAD+/NADH ratio is still the most relevant output for understanding changes in redox status and metabolism, rather than NAD+ or NADH ion counts individually. Nonetheless, in the interest of full transparency we have added all ion counts for NAD+ and NADH to Supplementary File 1 for easy reference.

We do however sympathize with the issue reviewers raise regarding to “relative ratios,” where we normalized the ratio of ion counts for NAD+ and NADH ratio from each sample to that of the vehicle treated cells. The reason we used this approach in the prior version is that, because standards for isotopically labeled NAD+ and NADH are not widely available, LCMS ion counts for NAD+ and NADH do not incorporate differences in matrix effects or ionization efficiencies for each analyte (and so are not comparable from an absolute quantitation perspective) and thus do not constitute a “true molar” NAD+/NADH ratio. We thought that, by normalizing samples it might decrease the likelihood that these data would be misperceived to a quantitative representation of the true cellular NAD+/NADH (which we have found to be a common misunderstanding). Nonetheless, we agree that is confusing to present a “relative ratio” and so we have changed all LCMS derived NAD+/NADH measurements to a simple ratio of ion counts, changing the Y axis legend to include “a.u./a.u.” to hopefully make this caveat apparent, as below:

In addition, we added the following text to the “Liquid Chromatography-Mass Spectrometry (LCMS)” section of the Materials and methods, to describe these caveats:

“Isotopically labeled standards for NAD+ and NADH are not widely available, so LCMS experiments measuring NAD+/NADH ratios do not incorporate differences in ionization efficiencies or matrix effects for each analyte. Thus, while changes in the ratio of measured ion counts for NAD+ and NADH reflect biological changes, they should not be interpreted as an absolute intracellular ratio, which is more accurately quantified by enzymatic methods. Individual ion counts for NAD+ and NADH are available in Supplementary File 1.”

2) Compartmentalized measurements of redox changes

Our data indicate that decreases in mitochondrial NAD+/NADH are required for the aspartate synthesis and proliferative benefits of complex I inhibition in SDH impaired cells, as evidenced by the result that these benefits can be blunted by expression of mitoLbNOX, but not cytoLbNOX. The lack of benefit for cytoLbNOX was expected since these cells are cultured in the presence of media pyruvate, which makes cytoLbNOX redundant for maintaining cytosolic NAD+/NADH. To underscore this experimental detail, we edited the text to make this point more clearly:

“Notably, since these cells are cultured in the presence of the cytosolic electron acceptor pyruvate, we thus expected cytoLbNOX to be redundant and ineffective to change cell metabolism, whereas mitoLbNOX would be positioned to disrupt the mitochondrial redox changes from rotenone treatment. Indeed, whereas cells expressing cytoLbNOX behaved similarly to parental cells expressing eGFP, mitoLbNOX expression impeded the restorative effect of rotenone treatment on SDH-impaired cell proliferation (Figure 3E).”

Nonetheless the reviewers raise a good point that since changes in mitochondrial compartmentalized NAD+/NADH are concluded to be the driver of our phenotype, it would be reasonable to measure them directly. To do so, we combined a recently published rapid subcellular enrichment protocol for metabolomics (Lee et al., 2019) with an enzymatic NAD+/NADH assay that allows measurements on NAD+/NADH from low sample input (protocol optimized in Sullivan et al., 2015), allowing us to separately measure NAD+/NADH from the cytosolic and mitochondrial fractions. We found that in basal conditions, as expected, cytosolic NAD+/NADH was higher than mitochondrial NAD+/NADH, indicating that the assay was likely reflecting endogenous redox differences between compartments (Revised Manuscript Figure 3B). Notably, mitochondrial inhibitors only had small effects on cytosolic NAD+/NADH, which was expected in this experimental context since exogenous pyruvate in the media can fix the cytosolic NAD+/NADH ratio through lactate dehydrogenase. We did however observe large changes in mitochondrial NAD+/NADH, where rotenone lowered the ratio, AA5 increased the ratio, and rotenone was epistatic to AA5 with regards to the effects on mitochondrial NAD+/NADH in co-treated cells. The disproportionate amount of NADH in the mitochondrial compartment thus likely drives the ”whole cell” NAD+/NADH changes observed by LCMS in the Revised Manuscript Figures 1D, 3A, and 7D, We believe these data provide important corroborative evidence for the conclusion of this figure that mitochondrial redox changes are required for the benefits of CI inhibition in SDH impaired cells.

4) The data presentation needs to be clearer and to be more consistent. For example, the Asp recovery experiments are sometimes presented using log scales which make very small changes in Asp levels appear much more substantial than they are; linear scales may be more appropriate and easier to interpret. The metabolite measurement data would also be clearer if presented as normalized ion counts rather than relative amounts; the ion counts should at the very least be provided somewhere in the manuscript.

We thank the reviewers for this suggestion and opportunity to improve the interpretability of the manuscript. Since SDH loss causes a severe depletion in aspartate by ~90% to a regime where small changes in aspartate levels can have large changes in proliferation rate (Revised Manuscript Figure 1C), we initially thought that a continuous log axis would be a fairer depiction of the data. It was certainly not our intention to mislead the readers and so we are happy to change the presentation to avoid that impression. We also recognize that log scales can be unintuitive and so, in accordance with the reviewers’ suggestion, we have changed all the metabolite graphs to linear scales, with appropriate line breaks to allow the reader to easily see relevant changes in aspartate levels when it is decreased. We still believe that relative ion counts are easier for reader interpretation compared to raw ion counts, particularly in light of the addition of Revised Manuscript Figure 1C that should help contextualize the relationship between relative aspartate levels and proliferation rates throughout the manuscript. Nonetheless, we have further clarified the quantification details by making the Y axis labels more descriptive (to make it easier to understand the condition for which all other metabolite levels are made relative) and by describing the metabolite quantification approach in detail in the methods section. Finally, to assuage any further concerns, we have included all ion counts used to generate the relative metabolite level graphs in Revised Manuscript Supplementary File 1.

5) In the current version of the manuscript, key data are not always consistent between panels. For example, the recovery of proliferation in SDH inhibited cells by rotenone appears somewhat inconsistent between experiments (e.g. Figure 2A and Figure S3B). As another example, in Figure 1, SDHB expression in UOK269 cells led to a proliferation rate that was nearly 5X higher than the control cells. Yet in Figure 2l, the proliferation rate of UOK269 cells expressing SDHB is the same as those not expressing SDHB. Further still, in 1F, addition of AA5 without PYR leads to minimal change in the NAD+/NADH ratio, while in 3A, sole addition of AA5 roughly doubles the NAD+/NADH ratio. The lack of consistency between panels is a recurrent problem throughout the manuscript and it is critical that the authors address these inconsistencies.

We apologize for this misunderstanding. While some variability in proliferation rates is a natural feature of conducting these experiments, the reviewers astutely noted several situations where the variability could be misinterpreted as potentially suggesting non-reproducible results, which is not the case. We have repeated each of the indicated experiments (and many others) and replaced them as necessary to ensure that the most representative experimental results are displayed.

However, we believe the discrepancies noted by the reviewers for UOK269 cells and NAD+/NADH relate to confusion about the media conditions used for each experiment, rather than inconsistent data. For example, the “5X higher” proliferation rate is true if comparing the proliferation rates of WT UOK269 cells in pyruvate-free DMEM (Original Submission Figure 1I/Revised Manuscript Figure 1G) to those cultured in DMEM with 1 mM pyruvate (Original Submission Figure 2I/Revised Manuscript Figure 8E). However, if the reviewers compare the proliferation rates of each cell line in the +PYR condition, they will see they are similar across figures. Likewise, the NAD+/NADH measurements of Original Submission 1F/Revised Manuscript Figure 1D are actually well in line with the data in Original Submission 3A/Revised Submission 3A, if comparing the +PYR conditions in the Figure 1 panel to those in Figure 3, which are from cells cultured in pyruvate containing media. Nonetheless, this perception of inconsistent results was our fault because we did not delineate experimental details clearly enough to guide the reader to the correct comparisons across figure panels. To address this issue, we have made several edits throughout the manuscript to remind the reader that all experiments after Figure 1 are conducted in media containing an electron acceptor (Pyruvate or AKB, as indicated).

Examples:

1) “We tested if, in the presence of electron acceptors, rotenone could alter the proliferation of cells treated with AA5 and, surprisingly, found it caused a robust restoration of cell proliferation (Figure 2A).”

2) Manuscript text edits highlighted in regards to “Compartmentalized measurements of redox changes,” (above):

“To avoid potential caveats that may confound labeling patterns from unlabeled media pyruvate, we conducted U-13C glucose isotope tracing experiments in pyruvate-free DMEM with AKB, which serves a similar electron acceptor function as pyruvate but cannot fulfill its carbon fates (Figure 1A) (Altea-Manzano et al., 2022; Sullivan et al., 2015). We also confirmed that rotenone is still effective to restore proliferation and aspartate levels of SDH-impaired cells in pyruvate-free DMEM with AKB (Figure 4—figure supplement 1B, 1C).”

3) In the discussion:

“Additionally, we note that SDH/CI impaired cells also require a supply of cytosolic electron acceptors to proliferate, indicating that this metabolic state depends on intercompartmental metabolic differences including a reduced mitochondrial compartment and an oxidized cytosolic environment.”

In addition, we have revised every figure legend for which the presence or absence of an electron acceptor (i.e. pyruvate or AKB) is a relevant variable for the experimental output to include the media condition explicitly. We hope these edits will help dispel any potential interpretation of variable results and aid in interpretation of our findings.

6) As detailed in the individual responses, several panels lack important controls and the appropriate statistical comparisons are not always included.

As suggested, we have repeated experiments to include important controls and have added the appropriate statistics corresponding to the individual reviewer comments.

7) Some of the data included in the manuscript does not add support to the model. For example, data regarding the combinatorial treatment of cells with AA5+AZD7545 (Figures 3G-H) shows only slight changes in proliferation and aspartate levels. This adds little to the argument that mitochondria-specific alterations in NAD+/NADH and aspartate biosynthesis are responsible for proliferation in the context of SDH inhibition.

We appreciate the reviewers concern for unnecessary data, but we respectfully disagree that these data do not add support to the model. While our mitoLbNOX data indicate that decreases in mitochondrial NAD+/NADH are required for the aspartate production and proliferative benefits of CI inhibition in SDH impaired cells, they do not necessarily confirm that decreased mitochondrial NAD+/NADH is sufficient to increase aspartate and proliferation in the absence of a CI inhibitor. We believe that the fact that AZD7545 increases aspartate and cell proliferation without blocking oxygen consumption provides evidence that decreased mitochondrial NAD+/NADH is both required and sufficient to induce aspartate and proliferation, without inherently requiring an ETC inhibitor. Nonetheless we agree that the very small effect sizes may confuse readers. So, we have de-emphasized these data in the revised manuscript by moving them to the supplemental figures (Revised Manuscript Figure 3—figure supplement A-E) and edited the manuscript to more accurately describe the effect sizes and clarify our interpretation:

“Importantly, treatment with AZD7545 also slightly improved cell proliferation and increased aspartate levels in AA5 treated cells (Figure 3—figure supplement 2D, 2E). Notably, these effects are likely limited by the fact that they are occurring in the presence of intact CI activity, which would likely buffer major changes in mitochondrial NAD+/NADH. Nonetheless, these data provide additional support that compartment-specific mitochondrial redox alterations are both sufficient and required to reprogram aspartate synthesis in SDH-impaired cells (Figure 3G).”

8) Could the authors address/discuss the physiological relevance of their findings. Do CI subunits exhibit decreased expression in SDH-deficient tumors in patients? Is this a phenomenon of cell culture or is this also observed in vivo? Could more physiological oxygen tension (or even hypoxia) affect SDH-deficient cell proliferation?

We thank the reviewers for bringing up this important point. Due to the rarity of these tumors, acquiring sufficient patient samples to analyze expression patterns of CI subunits on SDH-deficient tumors is difficult at the current moment. We did however seek to increase the physiological relevance of our findings, as suggested by the reviewers, by measuring these phenotypes in physiological oxygen tensions and in a tumor model. First, we asked whether decreased oxygen was sufficient to promote the proliferation of SDH impaired cells in the absence of a CI inhibitor. We conducted paired experiments culturing 143B cells in either 21% O2 (in a standard tissue culture incubator) or at 11%, 5%, 3%, or 1% O2 in a hypoxia chamber (See “Proliferation Assays” in the Materials and methods section for details). Interestingly, we found that physiological oxygen tensions were unable to significantly increase cell proliferation of AA5 treated cells from 3% or above, but that severe hypoxia at 1% O2 was able to partially improve the proliferation of AA5 treated cells, albeit to a lesser degree than rotenone (Revised Manuscript Figure 3—figure supplement 1D). These data suggest that severe hypoxia may benefit the proliferation of SDH deficient tumor cells and also provide plausibility that CI loss would be beneficial to SDH deficient cells at physiological oxygen concentrations in vivo.

Reviewer #1 (Recommendations for the authors):

1. The Asp recovery experiments are hard to interpret.

The panels showing Asp levels (e.g. in Figure 1D, 2B, 1G, 3E, 1H) would be easier for the reader to interpret if they were shown with a linear scale on the y-axis. On a linear scale the recovery of Asp by rotenone treatment in SDH inhibited cells (Figure 2B) would look quite minor compared to the loss of Asp that is caused by SDH inhibition. The authors also use different non-linear scales to show Asp levels in different panels of the same figure (e.g. Figure 2G and J), which is especially confusing.

Comparison of Asp rescue between different conditions is also challenging. For example, the recovery of Asp caused by "AA5 + pyruvate" (Figure 1D) appears to be greater than that caused by "AA5 and rotenone" (Figure 2B), yet the recovery of proliferation by "AA5 + pyruvate" is less. As another example, Asp levels in "AA5+mitoLbNOX" (Figure 3E) appear greater than Asp levels in "AA5+ rotenone" (Figure 2B). However, only "AA5 + rotenone" rescues proliferation. In other words, the relationship between Asp recovery and proliferation is not clear.

Part of the problem is that the authors use relative ion counts as well as non-linear scales. They also show the data for closely related conditions in different figures and panels. The best way to resolve the problem would be to present the data as absolute values rather than relative ion counts (e.g. fmol of Asp per cell). If this is not available, the authors should present the (normalized) ion counts (i.e. not the relative ion counts) on a linear scale, with error bars representing the variance of the ion counts. The authors do this in Figure 4B and it is much clearer.

Do the authors have the data from a control sample (143B wild-type cells) that could be included in Figure 4B, so that the reader could see how the rescue of Asp by SDHB KO compares to wild-type levels? Incidentally, Figure 4A would also be easier to interpret if the proliferation rate of wild-type 143B cells were also included.

More generally, if the experiments were performed at the same time, it would be helpful if the data for all the conditions were to be collated in a Sup. Table so that the recovery of Asp between different conditions can easily be compared.

As a result of this uncertainty in the extent of Asp recovery, the proposed rescue of Asp as a factor in facilitating proliferation in SDH and CI inhibition is not convincing. Right now, a key question can't be assessed by the reader: how much Asp is needed for proliferation? The quantitative relationship between Asp levels and proliferation should be made clearer.

These concerns are addressed in detail in response to Essential Revisions Comment 1. In brief, we have changed the presentation of aspartate levels to linear scales, standardized the presentation of aspartate levels amongst conditions, included ion counts for all LCMS data in Supplementary File 1, repeated experiments including additional control samples, and generated dose response curves that compare the effects of mitochondrial inhibitors on aspartate and proliferation to contextualize changes in aspartate throughout the manuscript (see revised submission Figure 1C; Figure 1—figure supplement 1K-L). We believe this latter point is a substantial improvement to the manuscript and so thank the reviewer again for the suggestion.

2. The data relating NAD+/NADH to the rescue of Asp and proliferation is not entirely convincing and is quite hard to interpret (Figure 1F, G).

i) It is not quite clear what is being measured in Figure 1F. The NAD+/NADH ratio reported in the Figure (i.e. ratio is approx. "1" for controls) is far lower than values that are reported in the literature for this cell line using a different approach (e.g. in the corresponding authors previous work with this cell line) or using carefully validated methods to extract NAD from other cell lines (e.g. by the Rabinowitz group, 2018, Antioxid Redox Signal.). It may be too much to ask the authors to redo the experiments using standards (to provide absolute amounts or concentrations), but it would easier to interpret claims about the role of redox state in these processes easier if the ratio reported here more closely reflects the intracellular NAD+/NADH ratio.

Perhaps the authors are reporting a relative ratio (a ratio of ratios?) that is normalized to the control group? If so, the figure legend isn't clear. Furthermore, having a ratio of ratios is very hard to interpret and can't be compared to previous studies or between experiments in this study. Could the authors please report either the NAD+/NADH ratio that they have measured for each group (on a linear scale), or (better) report the measured values for NAD+ and NADH (e.g. fmol per cell).

ii) Could the authors comment on the lack of an increase in the NAD+/NAD ratio in control cells when they are treated with pyruvate. Isn't this unexpected and inconsistent with previous studies? Several previous studies in very similar systems have all shown an increase in the NAD+/NADH following pyruvate treatment (e.g. Gui et al. 2016 Cell Metabolism), confirming many older studies.

iii) The authors use global NAD+/NADH measurements to infer compartment specific changes. The follow-up experiments using LbNOX are quite an indirect way to check whether this is the case. FLIM would provide some indication if the cytoplasmic and mitochondria pools of NAD(P)H are acting as proposed by the authors.

These concerns are addressed in detail in response to Essential Revisions Comment 3. Overall, we agree with the reviewer’s concerns – normalizing the ratios was intended to avoid the misperception that ion count ratios of NAD+ and NADH represent the true molar ratio, which cannot be done without isotopically labeled standards to account for matrix effects and differences in ionization efficiencies between analytes. Nonetheless, as detailed above, we changed all LCMS derived NAD+/NADH ratios to untransformed ratios of ion counts (with corresponding Y axis label changes to make this more obvious), explicitly describe these caveats in the Materials and methods section, and included all ion counts in Supplementary File 1. As suggested by the reviewer we also measured compartment specific NAD+/NADH changes. While we did not have access to FLIM for this purpose, we were able to leverage a rapid subcellular enrichment protocol combined with an enzymatic assay to measure compartmental changes to NAD+/NADH (see revised submission Figure 3B).

Related to comment ii, above – We agree that culturing cells in the presence of pyruvate is expected to increase NAD+/NADH, as was found in the paper mentioned, and many others. However, we respectfully disagree that our data are inconsistent with those studies. As the last author on this study was also a lead author on the referenced Gui et al. 2016 Cell Metabolism manuscript, we are able to directly compare the data from each. One issue with comparing across studies is that basal NAD+/NADH ratios were higher in the Gui paper, likely a result of using an enzyme-based assay specific for NAD+/NADH rather than LCMS (which has the caveats mentioned above). So, for the sake of comparing changes in NAD+/NADH in relevant treatment conditions across studies, we normalized each to the NAD+/NADH ratio measured in vehicle treated cells in DMEM-PYR. Importantly, regardless of differences in the cell line or measurement method, we see qualitatively similar changes upon the addition of pyruvate, rotenone, and both (Author response image 2). We believe these data highlight that, while the ratio of ion counts detected by LCMS may not reflect a true molar ratio in NAD+/NADH, it is still a valid method to measure changes in NAD+/NADH across treatment conditions.

Author response image 2
Download asset Open asset
Comparing relative changes in NAD+/NADH when adding pyruvate and/or rotenone, across studies.

NAD+/NADH measurements from the indicated figure panels in Gui et al. 2016 or this revised manuscript, with each made relative to vehicle treated cells cultured in pyruvate-free DMEM.

3. A full mechanistic model is not clearly presented by the authors.

The scheme in Figure 1G presumably doesn't reflect the author's full model, and it would be informative for the authors to present their proposed mechanism somewhere in the manuscript. Right now, it is not clear (for example) how the increase of mitochondrial NADH by dual inhibition of SDH and CI leads to an increase in Asp, or how the rescued Asp levels restore proliferation in this system.

The authors address how mitochondrial NADH increase may increase Asp in the Discussion. However, the manuscript would be strengthened if these questions were addressed experimentally. For example, it is straightforward to test whether reductive carboxylation of glutamine or anaplerosis by pyruvate carboxylase is responsible for the rescue of Asp when both SDH and CI are inhibited.

Furthermore, the link between Asp levels and proliferation has been extensively described in other conditions, but not in this specific condition. Given that the role and "direction" of Asp metabolism varies markedly between different conditions (e.g., Birsoy et al. 2015 vs Gaude et al. 2018) the authors should test which model is operating here. E.g. are MDH1 and GOT1 operating in reverse to generate Asp that is used for biosynthetic reactions, or are GOT1 and MDH1 is operating in the forward direction to consume Asp in order to generate NAD+ for aerobic glycolysis (perhaps more plausible here?).

As a whole the manuscript covers a lot of previously covered ground (e.g. loss of SDH causes loss of CI in tumors (e.g. Fendt group)), loss of SDH and CI creates a requirement for Asp synthesis (Gottlieb group). A clear mechanism of how mitochondrial NADH rescues Asp levels and proliferation would contribute much to its novelty.

We thank the reviewer for raising these issues, which are addressed in detail in Essential Revisions Comment 2. Briefly, we investigated these variables by conducting stable isotope tracing, using inhibitors of reductive carboxylation of glutamine, and making genetic knockouts for PC, MPC1, GOT2, GOT1, and MDH1 to reveal mechanisms for how CI inhibition promotes aspartate synthesis and cell proliferation in SDH deficient cells. These experiments generated a substantial amount of new data, with three new Figures added to the manuscript (Revised Manuscript Figures 3-5).

4. Key data are not always consistent between panels

For example, the recovery of proliferation in SDH inhibited cells by rotenone appears somewhat inconsistent between experiments (e.g. Figure 2A and Figure S3B). Some variability is to be expected, but, in this version of the manuscript, the experimental variability isn't reflected in the error bars.

Addressed in detail in Essential Revisions Comment 2.

5. Are the 143B cells used in these experiments capable of pyrimidine salvage?

It would be useful to know whether the 143B line that is used for most of the experiments is thymidine kinase negative (e.g, the 143B cells from ATCC are TK-). If so, this should be clearly stated in the methods, as lack of thymidine salvage will impact how these results are interpreted.

The 143B cells used in this study are from ATCC and are indeed TK-. However, this variable is unlikely to be a critical effector of this phenotype since DMEM does not contain thymidine (or any other nucleosides) in its formulation and experiments were done in media containing dialyzed FBS. In addition, our results were recapitulated in several other cell lines that are capable of thymidine salvage, without any discernable differences. Nonetheless, we aim to be fully transparent in our methods and so this information has been added to the Materials and methods (Cell Culture).

6. The rescue by AZD7545 does not add support to the model

The rescue of proliferation and especially Asp levels by AZD7545 is so small that if is hard to assess its importance. Is a difference between ~27% and ~31% of wild-type Asp levels really meaningful? When placed on a linear scale it would be hard to discern a difference between Asp levels in these groups.

Addressed in detail in Essential Revisions Comment 7.

Reviewer #2 (Recommendations for the authors):

1. While quite interesting, the physiological relevance of SDH and ETC complex I inhibition is unclear. Do CI subunits exhibit decreased expression in SDH-deficient tumors in patients? Is this a phenomenon of cell culture or is this also observed in vivo? Could more physiological oxygen tension (or even hypoxia) affect SDH-deficient cell proliferation?

Addressed in detail in Essential Revisions Comment 8. We believe the addition of the added hypoxia proliferation assays and inducible NDI1 xenograft tumor experiment add important physiological context to the study.

2. The data regarding the importance of aspartate biosynthesis in the context of SDH inhibition are mostly correlative and lack important controls.

a. Figure 1C – How was the dose selected for Asp treatment in the SLC1A3-expressing cells? While the tracing data confirms the increased uptake of Asp relative to non-SLC1A3 expressing cells (S1B-C), the authors do not report Asp levels or indicate how the dosage was selected.

b. Figures 1I and S1K – SDHB expression in UOK269 cells permitted glutamine-dependent aspartate biosynthesis and increased cell proliferation, yet the authors did not report how SDHB expression affected aspartate levels.

c. The authors are inconsistent in their interpretation of meaningful changes in aspartate. For example, regarding Figure 1E, the authors investigated the M4 aspartate isotopomer derived from U-13C-glutamine and conclude that pyruvate is insufficient to restore progression through the TCA cycle in AA5 treated cells because it resulted in only a 2.8X increase; yet in Figure 2B the authors observed a 2.3X increase in aspartate levels in cells treated with AA5 + rotenone, which leads the authors to conclude that CI and SDH co-inhibition partially restored aspartate levels.

d. Is the increase in proliferation upon SDH and CI inhibition dependent on GOT?

We thank the reviewers for the opportunity to clarify these issues:

a. We selected 1 mM aspartate for the indicated experiment (now Revised Manuscript Figure 1—figure supplement 1D) by identifying an aspartate dose that was sufficiently high to allow robust incorporation of media aspartate into SLC1A3 expressing cells (as measured in Revised Manuscript Figure 1—figure supplement 1C) and would be unlikely to be depleted over the course of a four-day proliferation assay. The goal of this experiment was to determine the magnitude cell proliferation changes upon SDH inhibition when ASP was not limiting, so an excess was desired. Nonetheless we estimated that 1 mM media aspartate would be reasonable based on net glutamine consumption (glutamine consumption – glutamate excretion) that we previously measured (Sullivan et al., 2018 Nature Cell Biology) to be around 170 fmol/cell*hour in 143B cells – resulting in a consumption of around 2.5 mM glutamine from the media during a standard proliferation assay (based on ~6E7 total cell-hours, 4 mL media with 4 mM glutamine starting concentration). Since glutamine is used for intracellular demands for glutamine, glutamate, aspartate, and aspartate fates (asparagine, pyrimidines, purines) we estimate that roughly 50% (+/- 25%) of glutamine consumption would be used towards aspartate and its fates, and so 1 mM aspartate would approximately meet those demands in the absence of functional aspartate synthesis.

b. Our apologies for this oversight – this data is now available in Revised Manuscript Figure 1—figure supplement 1N and Figure 8—figure supplement 1E.

c. We agree that changes in aspartate levels could be better presented and have overhauled our presentation of them (see reply to Essential Revisions Comment 1). We hope the reviewer will agree that the interpretation of meaningful aspartate changes is now consistent throughout the manuscript, especially when contextualized by Revised Manuscript Figure 1C.

d. This question is addressed in detail in response to Essential Revisions Comment 2 and by Revised Manuscript Figure 6.

3. Overstatements of results reduce clarity of the findings. For example:

a. The authors argument that "SDH inhibition blocks proliferation, which is restored by aspartate but not electron acceptors" (Figure 1 title) seems to be an overstatement. Though aspartate seems to increase cell proliferation to a greater degree than electron acceptors, data presented in 1A and 1C indicate that pyruvate, AKB, and LbNOX are all sufficient to increase cell proliferation.

b. The presented correlations between aspartate levels and proliferation are not sufficient evidence that aspartate biosynthesis is the primary mode of supporting cell proliferation in SDH-deficient cells. For example, in Supp. Figure 2B-C aspartate levels are not rescued to the same extent as proliferation across the panel of cell lines.

c. Data regarding the combinatorial treatment of cells with AA5+AZD7545 (Figures 3G-H) shows only slight changes to proliferation and aspartate levels. This adds little to the argument that mitochondria-specific alterations in NAD+/NADH and aspartate biosynthesis are responsible for proliferation in the context of SDH inhibition.

We are happy to have the ability to correct the text to avoid overstating our results:

a. This is a good point, we have changed the Figure 1 title to: “SDH inhibition blocks proliferation, which is incompletely rescued by electron acceptors but robustly restored by aspartate”

b. We have provided a detailed response with corresponding new data in the response to Essential Revisions Comment 2, which we hope the reviewer will agree provides data to support our conclusion that aspartate synthesis is a metabolic limitation for cell proliferation in SDH impaired cells. While we agree that different cell lines have different relationships between the degree of aspartate and proliferation rescue upon CI co-inhibition, these findings are in line with the non-linear relationship we found between aspartate levels and proliferation rate during mitochondrial inhibition (Revised Manuscript Figure 1C). The variability in aspartate levels and proliferation rate is likely a consequence of cell intrinsic differences in starting aspartate concentration, efficacy of the inhibitors to impair their target, and sensitivity differences to changes in aspartate levels. Nonetheless, we believe that the similar qualitative changes in aspartate and proliferation upon SDH and CI inhibition across cell lines supports the generalizability of our findings.

c. This comment is addressed in detail in response to Essential Revisions Comment 7.

4. Statistical analysis was frequently missing or improperly conducted. For example, regarding panel 1A the authors noted that AA5 and rotenone treatment decreased cell proliferation and that co-treatment with pyruvate, AKB, LbNox, or Asp restored proliferation. Statistical comparisons were made between AA5 and rotenone for all groups; however, based on the text, the statistical analysis should be restricted to comparisons between groups for each AA5 and rotenone.

This point is well taken, and we have reorganized Figure 1A to make the appropriate comparisons clear and added appropriate statistical tests throughout the manuscript.

5. The data presentation throughout the manuscript is confusing. For example, in some panels the y-axis is log-scale while in complimentary panels the y-axis is linear, and for some graphs the y-axis does not start at 0.

We have overhauled our data presentation in accordance with the reviewer critiques, see detailed response to Essential Revisions Comment 1.

6. In Figure 1, SDHB expression in UOK269 cells led to a proliferation rate that was nearly 5X higher than the control cells. Yet in Figure 2l, the proliferation rate of UOK269 cells expressing SDHB is the same as those not expressing SDHB. Why are the proliferation rates different?

We believe that apparent inconsistency is a misunderstanding based on comparing cells grown in different media conditions, a result of insufficient labeling clarity on our previous draft. Please see the detailed discussion of this point in response to Essential Revisions Comment 5.

Reviewer #3 (Recommendations for the authors):

During our reading of the manuscript, we felt the model the authors proposed within this work was not consistent or clear throughout the text. We feel that some attention could be made to try to emphasize and clarify this model throughout and would greatly assist other readers of the work. Currently, we feel like we as readers had to do a lot of work to try to make sense of the model and how the data are supporting that model.

We thank the reviewer bringing this to our attention and providing us the opportunity to clarify our model(s) and improve the manuscript. As detailed in response to Essential Revisions Comment 2, we have added a substantial amount of new data to better describe the metabolic pathways governing aspartate synthesis in this system. In addition, we have included models in each main figure to help underscore the conclusions of each, and the manuscript as a whole. We believe, and hope the reviewer will agree, that the inclusion of these models will help readers understand the conclusions of our work.

The authors state that "this work provides the first mechanistic evidence that CI loss can directly support core metabolic functions":

1) We do not feel the mechanistic data provided in the current form of the manuscript are strong enough to make this claim.

2) Perhaps this is the first instance of loss of CI & II together are beneficial, but there seem to be several other works that seem to make a case that loss of CI is beneficial to core metabolism processes, such as:

a. PMID: 29915083

b. PMID: 18784283

We are glad to have the opportunity to avoid overstating our data – That statement has been removed from the discussion of the revised submission and the section has been edited accordingly.

We do however feel that the novelty of our phenotype is distinct from the listed papers – The first (Gopal et al.) is an observational study comparing the exome, transcriptome, and metabolome of kidney cancers with and without mutations in mitochondrial CI. We agree that the association of mtDNA CI mutations with the disease raises the possibility that a loss of CI activity is beneficial, however no functional experiments are done in this study to test that hypothesis. It remains to be seen if these cancers (or others associated with mtDNA CI mutations, as mentioned in our Discussion) are truly driven to grow by the loss of CI activity per se, or as a result of another consequence of those mutations. We speculate that the finding that these oncocytomas are exclusively driven by mutations in the mtDNA, and not by mutations in CI subunits encoded by nuclear DNA, would suggest a more complicated mechanism than simply CI loss driving transformation and proliferation. The second paper (Garmier et al.) does investigate the functional effects of CI inhibition by rotenone treatment in Arabidopsis cells, measuring effects on cell growth and metabolic state. Importantly though, the authors note defects in cell growth upon rotenone treatment, in contrast to the effects observed in SDH impaired cells in our study. We believe a major novelty distinction between the listed studies and ours is the finding of experimental conditions where CI inhibition benefits cell proliferation rate and the characterization of the metabolic variables that govern that effect.

While we appreciate that many of the scripts used to generate data within this work are using pre-existing toolkits, it is still important to present the code used (how the toolkits were used exactly) to aid in the reproducibility of the work. These scripts could be included in the supplement, or hosted on a public repository, such as GitHub. Raw and processed metabolomics data should also be included as supplements, or hosted on a public repository, such as Metabolomics Workbench.

In Figure S4F, the authors state what "unsupervised clustering" was used. This is very generic – what was the exact algorithm used in the hierarchal clustering of the proteomics data and why? For example, scipy.cluster.hierarchy.linkage provide 7 different clustering methods (https://docs.scipy.org/doc/scipy/reference/generated/scipy.cluster.hierarchy.linkage.html)

We would also like to ensure we are aiding other groups in their ability to reproduce this work. As the reviewer mentioned, much of the data analysis was done using standard approaches from pre-existing software (e.g., LCMS data analysis by Tracefinder). Most other data (proliferation assays, cloning, western blots, Seahorse assays, Western blots, CI activity assays, etc.) did not require the generation of code beyond simple analyses in Excel or Prism and so does not seem applicable, although we are happy to add any missing information that the reviewer deems pertinent. We will however include all information that is available so we can to support the goal of transparency and reproducibility. Processed ion counts for all LCMS experiments are now included in Supplementary File 1. In addition, raw metabolomics files for key figures were uploaded onto Metabolomics Workbench with appropriate documentation. For the mentioned unsupervised clustering we have added the following statement to the methods section under TMT-Quantitative Mitochondrial Proteomics.

“Unsupervised clustering of the 16 proteomic samples was performed using the z-score based R (version 4.0.3) package GMD (Zhao & Sandelin, 2012) and visualized with the heatmap.3 function.”

There are also a couple of changes to the presentation of the figures that would improve the manuscript:

– Figure 3B/F: There is a line showing up at the top of each of these subpanels in our version of the manuscript.

Perhaps bar plot colors could be more standardized between sub-panels for the different conditions?

We thank the reviewer for identifying these issues and have worked to improve the aesthetics of our presentation, which we hope will aid in easier reader comprehension.

In summary, while we do not feel like this manuscript in its current form meets the criteria for publication in eLife, we are certainly positive about its potential and suggest filling in some of the remaining gaps outlined above and some mechanisms to explain the observations would make the work more complete and suitable for publication.

We again thank this reviewer, and the others, for their positive comments about the potential of our work. We believe the incorporating the data from the suggested experiments into the revised manuscript has addressed many of the mechanistic gaps and as a result increased the scope and impact of the work.

https://doi.org/10.7554/eLife.78654.sa2

Article and author information

Author details

  1. Madeleine L Hart

    1. Human Biology Division, Fred Hutchinson Cancer Center, Seattle, United States
    2. Molecular Medicine & Mechanisms of Disease Program, University of Washington, Seattle, United States
    Contribution
    Conceptualization, Data curation, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing – original draft, Writing – review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-9125-0627

  2. Evan Quon

    Human Biology Division, Fred Hutchinson Cancer Center, Seattle, United States
    Contribution
    Validation, Investigation, Writing – review and editing
    Competing interests
    No competing interests declared

  3. Anna-Lena BG Vigil

    Human Biology Division, Fred Hutchinson Cancer Center, Seattle, United States
    Contribution
    Validation, Investigation, Writing – review and editing
    Competing interests
    No competing interests declared

  4. Ian A Engstrom

    Human Biology Division, Fred Hutchinson Cancer Center, Seattle, United States
    Contribution
    Validation, Investigation, Writing – review and editing
    Competing interests
    No competing interests declared

  5. Oliver J Newsom

    Human Biology Division, Fred Hutchinson Cancer Center, Seattle, United States
    Contribution
    Validation, Investigation, Writing – review and editing
    Competing interests
    No competing interests declared

  6. Kristian Davidsen

    Human Biology Division, Fred Hutchinson Cancer Center, Seattle, United States
    Contribution
    Methodology, Writing – review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-3821-6902

  7. Pia Hoellerbauer

    Human Biology Division, Fred Hutchinson Cancer Center, Seattle, United States
    Contribution
    Methodology
    Competing interests
    No competing interests declared

  8. Samantha M Carlisle

    Department of Chemistry and Biochemistry, New Mexico State University, Las Cruces, United States
    Contribution
    Data curation, Formal analysis
    Competing interests
    No competing interests declared

  9. Lucas B Sullivan

    Human Biology Division, Fred Hutchinson Cancer Center, Seattle, United States
    Contribution
    Conceptualization, Resources, Data curation, Formal analysis, Supervision, Funding acquisition, Validation, Investigation, Visualization, Methodology, Writing – original draft, Project administration, Writing – review and editing
    For correspondence
    lucas@fredhutch.org
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-6745-8222

Funding

National Cancer Institute (P30CA015704)

  • Lucas B Sullivan

National Institute of General Medical Sciences (T32GM095421)

  • Madeleine L Hart

National Cancer Institute (R00CA218679-03S1)

  • Madeleine L Hart

National Cancer Institute (R00CA218679)

  • Lucas B Sullivan

National Institute of General Medical Sciences (R35GM147118)

  • Lucas B Sullivan

Andy Hill Cancer Research Endowment (CARE Award)

  • Lucas B Sullivan

National Cancer Institute (U54CA132381)

  • Samantha M Carlisle

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We thank Brian Milless, Dr. Philip Gafken and the staff of the Fred Hutch Proteomics and Metabolomics Shared Resource for LCMS metabolomics assay development, technical support, and for their assistance with conducting TMT-mitochondrial proteomics. We also thank members of the Sullivan lab, Dr. Lydia Finley, and members of the Finley lab for manuscript discussion and feedback.

Ethics

All mouse work was performed in accordance with FHCC-approved IACUC protocol 51069 and AAALAS guidelines and ethical regulations.

Senior Editor

  1. Utpal Banerjee, University of California, Los Angeles, United States

Reviewing Editor

  1. Mark S Sharpley, Cedars-Sinai Medical Center, United States

Reviewer

  1. Heather Christofk, University of California, Los Angeles, United States

Publication history

  1. Preprint posted: March 14, 2022 (view preprint)
  2. Received: March 15, 2022
  3. Accepted: March 6, 2023
  4. Accepted Manuscript published: March 8, 2023 (version 1)
  5. Accepted Manuscript updated: March 9, 2023 (version 2)
  6. Version of Record published: March 20, 2023 (version 3)

Copyright

© 2023, Hart et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 1,227
    Page views
  • 229
    Downloads
  • 0
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Madeleine L Hart
  2. Evan Quon
  3. Anna-Lena BG Vigil
  4. Ian A Engstrom
  5. Oliver J Newsom
  6. Kristian Davidsen
  7. Pia Hoellerbauer
  8. Samantha M Carlisle
  9. Lucas B Sullivan
(2023)
Mitochondrial redox adaptations enable alternative aspartate synthesis in SDH-deficient cells
eLife 12:e78654.
https://doi.org/10.7554/eLife.78654

Categories and tags

Research organisms

Further reading

    1. Biochemistry and Chemical Biology

    Targeting oncogenic KRasG13C with nucleotide-based covalent inhibitors

    Lisa Goebel, Tonia Kirschner ... Daniel Rauh
    Short Report

    Mutations within Ras proteins represent major drivers in human cancer. In this study, we report the structure-based design, synthesis, as well as biochemical and cellular evaluation of nucleotide-based covalent inhibitors for KRasG13C, an important oncogenic mutant of Ras that has not been successfully addressed in the past. Mass spectrometry experiments and kinetic studies reveal promising molecular properties of these covalent inhibitors, and X-ray crystallographic analysis has yielded the first reported crystal structures of KRasG13C covalently locked with these GDP analogues. Importantly, KRasG13C covalently modified with these inhibitors can no longer undergo SOS-catalysed nucleotide exchange. As a final proof-of-concept, we show that in contrast to KRasG13C, the covalently locked protein is unable to induce oncogenic signalling in cells, further highlighting the possibility of using nucleotide-based inhibitors with covalent warheads in KRasG13C-driven cancer.

    1. Biochemistry and Chemical Biology
    2. Cancer Biology

    Multi-targeted therapy resistance via drug-induced secretome fucosylation

    Mark Borris D Aldonza, Junghwa Cha ... Yoosik Kim
    Research Article

    Cancer secretome is a reservoir for aberrant glycosylation. How therapies alter this post- translational cancer hallmark and the consequences thereof remain elusive. Here we show that an elevated secretome fucosylation is a pan-cancer signature of both response and resistance to multiple targeted therapies. Large-scale pharmacogenomics revealed that fucosylation genes display widespread association with resistance to these therapies. In cancer cell cultures, xenograft mouse models, and patients, targeted kinase inhibitors distinctively induced core fucosylation of secreted proteins less than 60 kDa. Label-free proteomics of N-glycoproteomes identified fucosylation of the antioxidant PON1 as a critical component of the therapy-induced secretome (TIS). N-glycosylation of TIS and target core fucosylation of PON1 are mediated by the fucose salvage-FUT8-SLC35C1 axis with PON3 directly modulating GDP-Fuc transfer on PON1 scaffolds. Core fucosylation in the Golgi impacts PON1 stability and folding prior to secretion, promoting a more degradation-resistant PON1. Global and PON1-specific secretome de-N-glycosylation both limited the expansion of resistant clones in a tumor regression model. We defined the resistance-associated transcription factors (TFs) and genes modulated by the N-glycosylated TIS via a focused and transcriptome-wide analyses. These genes characterize the oxidative stress, inflammatory niche, and unfolded protein response as important factors for this modulation. Our findings demonstrate that core fucosylation is a common modification indirectly induced by targeted therapies that paradoxically promotes resistance.

    1. Biochemistry and Chemical Biology
    2. Developmental Biology

    O-GlcNAc glycosylation orchestrates fate decision and niche function of bone marrow stromal progenitors

    Zengdi Zhang, Zan Huang ... Hai-Bin Ruan
    Research Article Updated

    In mammals, interactions between the bone marrow (BM) stroma and hematopoietic progenitors contribute to bone-BM homeostasis. Perinatal bone growth and ossification provide a microenvironment for the transition to definitive hematopoiesis; however, mechanisms and interactions orchestrating the development of skeletal and hematopoietic systems remain largely unknown. Here, we establish intracellular O-linked β-N-acetylglucosamine (O-GlcNAc) modification as a posttranslational switch that dictates the differentiation fate and niche function of early BM stromal cells (BMSCs). By modifying and activating RUNX2, O-GlcNAcylation promotes osteogenic differentiation of BMSCs and stromal IL-7 expression to support lymphopoiesis. In contrast, C/EBPβ-dependent marrow adipogenesis and expression of myelopoietic stem cell factor (SCF) is inhibited by O-GlcNAcylation. Ablating O-GlcNAc transferase (OGT) in BMSCs leads to impaired bone formation, increased marrow adiposity, as well as defective B-cell lymphopoiesis and myeloid overproduction in mice. Thus, the balance of osteogenic and adipogenic differentiation of BMSCs is determined by reciprocal O-GlcNAc regulation of transcription factors, which simultaneously shapes the hematopoietic niche.