Electron transport chain inhibition increases cellular dependence on purine transport and salvage

SUMMARY Mitochondria house many metabolic pathways required for homeostasis and growth. To explore how human cells respond to mitochondrial dysfunction, we performed metabolomics in fibroblasts from patients with various mitochondrial disorders and cancer cells with electron transport chain (ETC) blocakde. These analyses revealed extensive perturbations in purine metabolism, and stable isotope tracing demonstrated that ETC defects suppress de novo purine synthesis while enhancing purine salvage. In human lung cancer, tumors with markers of low oxidative mitochondrial metabolism exhibit enhanced expression of the salvage enzyme hypoxanthine phosphoribosyl transferase 1 (HPRT1) and high levels of the HPRT1 product inosine monophosphate. Mechanistically, ETC blockade activates the pentose phosphate pathway, providing phosphoribosyl diphosphate to drive purine salvage supplied by uptake of extracellular bases. Blocking HPRT1 sensitizes cancer cells to ETC inhibition. These findings demonstrate how cells remodel purine metabolism upon ETC blockade and uncover a new metabolic vulnerability in tumors with low respiration.


INTRODUCTION
Cancer and inborn errors of metabolism (IEMs) are characterized by mutations that perturb cellular metabolism.Although cancer and IEMs are clinically very different, they share some pathogenic mechanisms.Oncogenic signaling regulates many of the same metabolic pathways that become dysfunctional in IEMs, including glycolysis, amino acid oxidation, the urea cycle and others [1][2][3][4] .Dozens of human IEMs are caused by mutations of mitochondrial enzymes, particularly in the TCA cycle and subunits of the electron transport chain (ETC) necessary for oxidative phosphorylation (OXPHOS) 5 .Some of these same enzymes, including succinate dehydrogenase and fumarate hydratase, are tumor suppressors in adult-onset cancers [6][7][8] .Therefore, studying metabolic reprogramming in cancer can provide insights relevant to the pathophysiology of IEMs, and vice versa 9 .
Mitochondria house pathways that contribute to cell growth by producing energy, biosynthetic precursors, and signaling molecules.Human tumors appear to vary in their need for OXPHOS, with some showing genetic evidence for suppression of ETC function 6,8,10- 12 .However, although many tumors contain nonsynonymous mitochondrial DNA point mutations, including ones predicted to impair the ETC, tumors often select against these mutations as they progress 13,14 .Efforts to assess human lung tumor metabolism in vivo using intra-operative 13 C-glucose infusions demonstrate variable 13 C labeling of TCA cycle intermediates among tumors from different patients and in different regions of the same tumor 15,16 .This variation could be related to intrinsic properties of tumor cells 17,18 , or to environmental factors such as hypoxia, which may limit oxidation of glucose and other nutrients.In mice, tumors display low tricarboxylic acid (TCA) cycle turnover at the site of origin, but higher activity at some metastatic sites 19 .Genetically eliminating ETC components severely suppresses xenograft growth, even at the site of implantation, indicating the requirement for at least some OXPHOS in those models 20,21 .
A marker of ETC dysfunction relevant to cancer cell growth is NADH reductive stress (i.e., low NAD + :NADH ratio), which occurs when NADH production outpaces its oxidation to NAD + , particularly by ETC Complex I.This form of metabolic stress has broad implications for intermediary metabolism because many NAD(H)-dependent oxidoreductases are sensitive to the NAD + :NADH ratio.High demands for NAD + can impose a bottleneck on growth because some pathways that produce precursors for macromolecular synthesis contain NAD + -dependent oxidoreductases 22,23 .The Warburg effect, which describes the conversion of glucose to lactate in the presence of oxygen observed in most cancer cells, is thought to reflect the need to regenerate NAD + by lactate dehydrogenase when mitochondrial redox shuttles become oversaturated 24 .Identifying consistent responses to low redox ratios may provide new insights about metabolic diseases, including cancer.
Here we sought to identify consistent metabolic alterations downstream of mitochondrial dysfunction in humans.We report that ETC dysfunction changes how cells produce purines, with the salvage enzyme HPRT1 becoming essential for growth in the setting of mitochondrial impairment.

Altered purine metabolism is a common feature of human mitochondrial disease
To identify the effects of mitochondrial dysfunction in humans, we first analyzed metabolomics in fibroblasts from patients with genetic diseases affecting mitochondrial RNA processing and translation, lipoylation of mitochondrial enzymes, the ETC or other processes (Figure 1A).Metabolite set enrichment analysis (MSEA) revealed that the TCA cycle is commonly perturbed in fibroblasts from these patients (Figure 1B).Unexpectedly, even more cells from this panel display alterations in purine metabolism (Figure 1C).We also analyzed plasma from a previously-reported patient with Lipoyltransferase-1 (LIPT1) deficiency, which results in dysfunction of pyruvate dehydrogenase and oxoglutarate dehydrogenase (PDH and OGDH) 25 .We compared plasma collected from this patient during several hospitalizations to plasma from healthy controls.As expected, plasma from the LIPT1-deficient patient had elevated lactate, alanine, and α-ketoglutarate (Figure 1D).Several purine metabolites were also increased in plasma from this patient (Figure 1E).

ETC blockade increases purine metabolites
To assess mitochondrial dysfunction in a simpler system, we treated human H460 non-small cell lung cancer (NSCLC) cells with IACS-010759, a potent mitochondrial ETC complex I inhibitor (Figure 2A) 26 .The drug reduced the cellular NAD + :NADH ratio (Figure 2B) and induced a compensatory increase in glucose uptake and lactate secretion (Figures S1A and S1B).Metabolomics revealed widespread effects of IACS-010759, including reduced levels of several metabolites related to TCA cycle function (Figure S1C and S1D) [27][28][29] .Many purine metabolites were elevated upon IACS-010759 treatment, despite the depletion of aspartate, a precursor for de novo purine synthesis (Figures 2C and S1D).IACS-010759 also increased purine monophosphates in other cell lines (Figures S1E and  S1F).To test whether IACS-010759 impacts purine metabolism in NSCLC xenografts in vivo, H460 cells were subcutaneously implanted into immunocompromised mice, and metabolomics was performed after treating the mice with IACS-010759 or vehicle.Among many alterations, purine metabolism stood out as by far the most affected pathway, with marked accumulations in GMP and guanosine (Figure S1G-S1I).
To verify that IACS-010759 induced these changes through complex I inhibition, we expressed Saccharomyces cerevisiae alternative NADH dehydrogenase (NDI1) to restore OXPHOS (Figure 2D).NDI1 boosted basal mitochondrial respiration in H460 cells, rendered them resistant to IACS-010759 and rotenone, and normalized the NAD + :NADH ratio, glucose uptake, and lactate secretion (Figure 2E, F, S2A and S2B).Control H460 cells expressing an empty vector displayed growth suppression upon IACS-010759 treatment, and extensive cell death in medium containing galactose instead of glucose.These effects were reversed by NDI1 (Figure 2G).Metabolomics revealed that nearly all IACS-010759-induced metabolic alterations, including those involving purines, were corrected by NDI1 (Figures 2H, 2I and S2C).MSEA identified purine metabolism as the top-scoring pathway from metabolites that were altered by IACS-010759 in control but not NDI1-expressing cells (Figure 2J), identifying this pathway as the most responsive to complex I blockade.
To evaluate whether altered purine metabolism is a common phenotype associated with defects in other ETC complexes, we generated H460 cell lines depleted for UQCRC2, a component of ETC complex III (i.e., UQCRC2 −/− cells, Figure 2K).UQCRC2 ablation impaired mitochondrial respiration (Figure 2L), suppressed cell growth (Figure S2D), and induced a distinct metabolomic profile (Figures S2E and S2F), including marked purine accumulation (Figures 2M and 2N).
Reduced cell proliferation in cells with ETC defects could result in suppressed purine consumption for nucleic acid synthesis, thereby leading to purine accumulation.To examine the relationship between proliferation and purine accumulation, we treated H460 cells with nocodazole, an inhibitor of microtubule polymerization, at a dose that arrests cell proliferation (Figure S2G).This had no effect on purine monophosphate levels (Figure S2H).We then used SK-N-DZ neuroblastoma cells, which maintain proliferation despite reduced oxygen consumption during IACS-010759 treatment 30 (Figures S2I and S2J).These cells still accumulated purine monophosphates when complex I was inhibited (Figure S2K).Therefore, although ETC blockade can suppress growth, this is neither sufficient nor necessary to induce purine nucleotide accumulation.

Cytosolic NAD + :NADH ratio impacts purine metabolism upon ETC blockade
Mitochondrial respiration is coupled to oxidation of reducing equivalents (i.e., NADH and FADH 2 ), ATP synthesis, and mitochondrial membrane potential.It was unclear which of these modulates purine abundance.Given that NAD + is required for many oxidoreductase reactions, we reasoned that the decreased NAD + :NADH ratio impacts purine metabolism when the ETC is compromised.To dissociate NADH oxidation from OXPHOS, we expressed the water-forming NADH oxidase from Lactobacillus brevis (LbNOX) that utilizes oxygen to convert NADH to NAD + (Figure 3A) 31 .We localized LbNOX to cytosol or mitochondria in UQCRC2 −/− cells to ameliorate NADH accumulation in either compartment (Figures 3B and 3C) 31 .In whole-cell lysates, Mito-LbNOX but not Cyto-LbNOX increased the NAD + :NADH ratio (Figure 3D).Nevertheless, consistent with previous studies 31 , both versions improved UQCRC2 −/− cell growth in the absence of pyruvate and uridine, and this was more pronounced with Cyto-LbNOX (Figure 3E).Cyto-LbNOX also enhanced aspartate abundance to a greater extent (Figure 3F).Importantly, LbNOX did not restore oxygen consumption in UQCRC2 −/− cells, indicating that it increases the NAD + :NADH ratio independently of OXPHOS (Figure 3G).We assessed metabolomic profiles of WT and UQCRC2 −/− cells expressing either an empty vector (EV) or Mito/Cyto-LbNOX, all grown without pyruvate and uridine supplementation.Under these conditions, Cyto-LbNOX had a greater overall metabolomic impact than Mito-LbNOX (Figure 3H), although Mito-LbNOX had a more pronounced effect on glutamine reductive carboxylation (Figure S3A).Neither Cyto-LbNOX nor Mito-LbNOX completely alleviated the metabolomic effects of UQCRC2 loss (Figure S3B).MSEA identified numerous pathways modulated by Mito-LbNOX and Cyto-LbNOX in ETC-deficient cells (Figure 3I and 3J).However, in terms of nucleotide metabolism, Cyto-LbNOX primarily affected purines while Mito-LbNOX primarily affected pyrimidines (Figure 3I and 3J).Accordingly, Cyto-LbNOX, but not Mito-LbNOX reduced IMP and hypoxanthine levels in UQCRC2 −/ − cells (Figure 3K).AMP and GMP were not normalized by Cyto-LbNOX but instead increased further.The increased GMP level is likely attributed to enhanced activity of inosine monophosphate dehydrogenases 1/2 (IMPDH1/2), which require cytosolic NAD + in the GMP synthesis pathway 28 .The elevation in AMP may result from increased availability of aspartate 28 .We also supplemented the medium with α-ketobutyrate (AKB), a compound utilized to mitigate cytosolic NADH accumulation (Figure S3C) 28 .Similar to Cyto-LbNOX, AKB reduced hypoxanthine and IMP levels and enhanced cell growth (Figures S3D and  S3E).

ETC blockade suppresses de novo purine synthesis
To further examine purine metabolism in cells with ETC dysfunction, we cultured cells with uniformly 13 C-labeled glucose ([U- 13 C]glucose).While de novo purine synthesis yields various purine nucleotide isotopologues reflecting labeling in both the ribose backbone and purine base, purines produced from the salvage pathway are dominated by m+5 labeling in the ribose backbone (Figure 4A).Vehicle-treated cells displayed the expected heterogeneity in IMP isotopologues, but labeling was almost entirely m+5 in IACS-010759-treated cells (Figure 4B).NDI1 eliminated the effect of IACS-010759 on purine labeling (Figure 4B).Similar effects occurred in GTP and ATP, although overall labeling was lower than for IMP (Figure S4A).IACS-010759 also increased time-dependent m+5 labeling of both IMP and GMP (Figure S4B).
We next conducted kinetic [amide- 15 N]glutamine tracing in H460 cells with or without IACS-010759 to assess the de novo purine synthesis pathway through which two and three 15 N nuclei are incorporated into IMP and GMP, respectively (Figure 4C).Compared to untreated cells, control cells treated with IACS-010759 exhibited lower fractional enrichment of m+2 IMP and m+3 GMP throughout the time course, and this was reversed by NDI1 (Figure 4D).Adding hypoxanthine to the medium did not reverse the suppressed labeling caused by IACS-010759 (Figure S4C).IACS-010759 nearly eliminated 5-aminoimidazole-4-carboxamide ribonucleotide (AICAR) and phosphoribosylaminoimidazolesuccinocarboxamide (SAICAR), two intermediates in the de novo purine synthesis pathway (Figure S4D), as previously observed in ETC Complex III-deficient cells 28 .It also suppressed m+2 labeling of AICAR in control but not NDI1-expressing cells (Figure S4E).These findings indicate a suppression of de novo purine synthesis upon IACS-010759 treatment.
As an orthogonal analysis of the relationship between purine metabolism and the ETC, we used the Cancer Dependency Map (DepMap) to identify co-essential genes with PPAT32, which encodes Phosphoribosyl Pyrophosphate Amidotransferase, the rate limiting enzyme of the de novo purine synthesis pathway.We observed a strong correlation of PPAT with genes involved in purine metabolism and other related pathways, including the pentose phosphate pathway and one carbon metabolism (Figure S4F).Notably, the highest-scoring pathways were the TCA cycle and oxidative phosphorylation, indicating a critical role of mitochondrial oxidative metabolism in supporting de novo purine synthesis (Figure S4F).IACS-010759 had little effect on mRNA or protein abundance of enzymes in the de novo purine synthesis pathway (Figure S4G and S4H), as expected if pathway suppression occurred as a metabolic effect of ETC inhibition.
To test whether ETC blockade impacts purine metabolism in tumors in vivo, H460 xenograft-bearing mice were dosed with IACS-010759 and infused with [amide- 15 N]glutamine (Figure 4E).AICAR abundance declined in tumors treated with IACS-010759 (Figure 4F).IACS-010759 had no effect on tumor enrichment of m+1 glutamine (Figure 4G) but it suppressed m+2 labeling in IMP and AMP (Figure 4H), indicating suppressed de novo purine synthesis in vivo.
Defective mitochondrial respiration limits synthesis of aspartate, which is required for de novo purine synthesis 27,28,33,34 .This raises the possibility that aspartate depletion explains reduced de novo purine synthesis in IACS-010759-treated cells.Previous studies demonstrated that ETC inhibition leads to increased incorporation of exogenous aspartate into nucleotides 34 , but it is unknown whether aspartate supplementation restores de novo purine synthesis in ETC-deficient cells.To test this, we generated isogenic cell lines that overexpress the aspartate transporter SLC1A3 (SLC1A3 OE ) (Figure 4I).These cells display enhanced aspartate uptake from the medium (Figure S4I).However, despite restoration of intracellular aspartate abundance in SLC1A3 OE cells exposed to IACS-010759 (Figure S4J), purine nucleotide labeling from [amide- 15 N]glutamine remained low (Figure 4J).Therefore, the ETC supports de novo purine nucleotide synthesis through mechanisms beyond supplying cellular aspartate.

Mitochondrial ETC deficiency enhances purine salvage
Despite reduced de novo purine synthesis, ETC blockade increases purine monophosphate abundance.In IACS-010759-treated cells, essentially 100% of the IMP pool is labeled as m+5 after 6 hours of culture with [U- 13 C]glucose (Figure S4B); this indicates that the entire IMP pool has turned over in 6 hours.However, under identical conditions, 6 hours of culture with [amide- 15 N]glutamine results in approximately 50% m+2 IMP fractional enrichment (Figure 4D), indicating that only 50% of the IMP that has turned over arises from de novo purine synthesis, with the rest presumably arising from purine salvage.To assess purine salvage, we first challenged cells with lometrexol (LTX) or methotrexate (MTX) to inhibit de novo purine synthesis, and traced with either [amide- 15 N]glutamine or [ 15 N 4 ]hypoxanthine (Figure 5A).As expected, both drugs suppressed de novo purine synthesis but stimulated purine salvage (Figure S5A and S5B).After 6 hours of culture with [ 15 N 4 ]hypoxanthine, IACS-010759-treated control cells displayed higher m+4 IMP enrichment than untreated cells, with NDI1 reversing this effect (Figure 5B).Kinetic tracing revealed increased labeling of both AMP and GMP from [ 15 N 4 ]hypoxanthine in IACS-010759-treated cells (Figure S5C).H460 cells deficient in LIPT1 also showed higher contribution of hypoxanthine to purine pools (Figure S4D and S4E).To ascertain whether purine salvage was necessary for purine nucleotide accumulation by IACS-010759, we generated H460 cells defective in the purine salvage enzyme hypoxanthine phosphoribosyl transferase 1 (HPRT1, Fig. S4F).Ablation of HPRT1 led to an elevation in intracellular hypoxanthine and nearly eliminated enrichment of m+4 purine nucleotides induced by IACS-010759 (Fig 5C and 5D).The accumulation of IMP, AMP, and GMP induced by IACS-010759 was also blunted in the absence of HPRT1 (Figure 5E), indicating that HPRT1-mediated purine salvage contributes to purine monophosphate accumulation upon ETC blockade.Neither HPRT1 expression nor its enzymatic activity in cell lysates was different between control and ETC-deficient cells (Figures S4G-H and S5G).Therefore we next tested whether enhanced purine salvage in ETC-deficient cells results from an increase in HPRT1 substrates.

ETC blockade enhances the PPP
Purine salvage requires purine nucleobases and phosphoribosyl diphosphate (PRPP), an activated form of ribose-5-phosphate (R5P) produced in the pentose phosphate pathway (PPP).Therefore, we examined the effects of complex I inhibition on the PPP.R5P pools were depleted after 24 hours of IACS-010759 (Figure S5H).However, after addition of fresh medium containing [U- 13 C]glucose, both the abundance and m+5 labeling of R5P rose faster in IACS-010759-treated than DMSO-treated cells, indicating rapid synthesis of R5P from glucose during complex I blockade (Figures S5I to S5K).IACS-010759-treated cells also displayed higher enrichment of m+6 6-phosphogluconate (6-PG) and m+7 sedoheptulose 7-phosphate (S7P), two other PPP intermediates (Figure S5L), and more rapid labeling and higher abundance of PRPP (Figure 5F and 5G).Because the concentration of PRPP in human cells (~10 μM) is below its reported K m for HPRT1 (approximately 200 μM) 35,36 , increasing PRPP abundance may facilitate purine salvage during ETC blockade.
To assess whether the oxidative or non-oxidative branch of the PPP predominated in these cells, we used [1,2-13 C]glucose as a tracer.In this tracing scheme, both m+1 and m+2 R5P are produced, with m+1 arising predominantly from the oxidative branch and m+2 arising from the non-oxidative branch (Figure 5H).IACS-010759 treatment increased both m+1 and m+2 R5P abundance, but the fractional enrichment of m+1 increased while m+2 decreased in response to IACS-010759, and both were normalized by NDI1.(Figures 5I and 5J).NDI1 also eliminated the increase in PRPP m+5 induced by IACS-010759 (Figure 5K).Taken together, the data indicate an activation of PRPP synthesis, primarily through the oxidative branch of the PPP, in response to ETC blockade.

HPRT1 is important for NSCLC growth especially when ETC is impaired
Since defective mitochondrial respiration inhibits de novo purine nucleotide synthesis, we hypothesized that purine salvage is essential for ETC-deficient cells.Indeed, compared to control cells, HPRT1-deficient cells are more sensitive to IACS-010759 (Figure 6A).Inhibiting de novo purine synthesis does not exacerbate the growth defect caused by ETC blockade (Figure S6A), indicating that ETC-deficient cells are less dependent on de novo purine synthesis for growth.To assess HPRT1's role in purine metabolism in vivo, we subcutaneously injected control and HPRT1-deficient H460 cells into immunocompromised mice and dosed the mice with vehicle or IACS-010759.HPRT1 deficiency reduced tumor growth even without IACS-010759 treatment, but compared to the control tumors, HPRT1deficient tumors were more sensitive to IACS-010759 (Figure 6B).These data indicate that although H460 cells tolerate HPRT1 loss in culture, this enzyme is required for maximal tumor growth in vivo and complex I blockade increases dependence on HPRT1-mediated purine salvage.While de novo purine synthesis is the target of multiple chemotherapeutic drugs, the role of purine salvage in tumor growth remains underappreciated.Analysis of The Cancer Genome Atlas (TCGA) showed higher expression of HPRT1 in human lung adenocarcinomas and squamous cell carcinomas compared to nonmalignant lungs (Figure 6C).We also observed enhanced HPRT1 expression in NSCLCs relative to patient-matched lung tissue from our own cohort (Figure 6D).Moreover, high expression of HPRT1 correlates with poor overall survival of patients with NSCLC (Figure 6E).
We next examined how mitochondrial function affects purine metabolism in human NSCLC in vivo.Intra-operative infusion of [U- 13 C]glucose during surgical NSCLC resection leads to variable labeling in TCA cycle intermediates extracted from the tumors (Figure 6F) 15,16 .In xenografts, ETC activity within NSCLC cells contributes to TCA cycle intermediate labeling from glucose 30 , so for this analysis we asked how labeling of these intermediates correlates with markers of purine metabolism.In human NSCLCs, there was a strong correlation between m+2 glutamate and m+2 malate, indicating label propagation around the TCA cycle (Figure S6B).Analysis of 13 C labeling and metabolite abundance revealed that tumors with low malate m+2 enrichment contained more IMP (Figure 6G).RNA-sequencing revealed that tumors with low malate m+2 enrichment exhibited higher HPRT1 expression (Figure 6H).Similar results were also obtained if we used m+2 glutamate for the analyses (Figure S6C and S6D).These data may indicate an enhanced propensity for purine salvage in human NSCLCs when glucose-dependent labeling of TCA cycle intermediates is low, as would be the case if OXPHOS is relatively impaired.

Purine nucleotide accumulation induced by complex I inhibition is independent of macroautophagy
We next explored how cells acquire purine nucleobases for the salvage reaction.Cancer cells can use autophagy to generate purine nucleotides 37 .Some 80% of cellular RNA is ribosomal RNA, which accounts for most of the ribosomal mass 38,39 .The selective degradation of ribosomes by autophagy (ribophagy) contributes to nucleotide pools during nutrient starvation, and this process is negatively regulated by mTORC1 40,41 .Consistent with a previous study 29 , we observed repressed mTORC1 signaling upon IACS-010759 treatment, and this was restored by NDI1 (Figures S7A and S7B).Since mTORC1 inhibits ribophagy, we tested whether mTORC1 suppression induces ribosomal degradation and supplies nucleobases for purine salvage during IACS-010759 treatment.We generated an H460 ribophagy reporter cell line that expresses ribosomal protein 3 (RPS3) fused with a Keima-Red protein (Figure S7C).During ribophagy, RPS3-Keima-Red is cleaved to release Keima protein (Figure S7D, lower band) 40,42 .Treatment with the mTORC1 inhibitor Torin1 led to the expected cleavage of RPS3-Keima, and this was prevented by the autophagy inhibitor Bafilomycin A (Figure S7D).In contrast, IACS-010759 did not induce ribophagy (Figure S7D).IACS-010759 did decrease p62/SQSTM1, indicating augmented macroautophagy (Figure S7D).To examine whether macroautophagy contributes to purine accumulation, we generated H460 cells deficient in ATG5 or ATG7, two essential autophagy factors (Figure S7E).Deletion of ATG5 or ATG7 had no impact on the accumulation of IMP, GMP or AMP by IACS-010759 (Figure S7F), indicating that ETC blockade induces purine nucleotide accumulation independently of macroautophagy.

ETC-deficient cells depend on nucleobase uptake to provide purine nucleotides for growth
Metabolic stress induces nutrient scavenging from the microenvironment to sustain cell survival and growth.ETC-deficient cells rely on environmental lipids for cell growth, and pancreatic cancer cells use macropinocytosis when aspartate synthesis is disrupted 43,44 .Of note, an unbiased CRISPR screen in pancreatic cancer cells identified both HPRT1 and SLC29A1, a nucleoside/base transporter, as conditionally essential during ETC blockade 44 .To investigate whether ETC-deficient cells rely on extracellular purine nucleobases for purine salvage and proliferation, we cultured cells in medium supplemented with either FBS or dialyzed FBS (dFBS).These two sera are metabolically different including much lower purine levels in dFBS (Figure S7G and S7H).Cells growing in dFBS-supplemented medium were more sensitive to ETC blockade, and growth was partially rescued by supplementing with purine nucleosides (Figure S7I).This led us to hypothesize that ETC-deficient cells take up nucleosides or nucleobases from the medium to supply the salvage pathway.
Two main nucleoside transporter groups, the SLC28 and SLC29 families, transport most purine nucleosides and nucleobases 45 .From RNA-seq data, we determined that H460 cells only express appreciable levels of SLC29A1 and SLC29A2 (Figure S7J).We treated H460 cells with DMSO, IACS-010759, or a combination of IACS-010759 and the SLC29A1/SLC29A2 inhibitor nitrobenzylthioinosine (NBMPR) and monitored consumption of unlabeled purine metabolites and [ 15 N 4 ]hypoxanthine from the medium.Hypoxanthine was rapidly depleted, but none of the other bases were taken up (Figures 7A  and S7K).IACS-010759 did not potentiate hypoxanthine uptake, and NBMPR suppressed it (Figures 7A and S7K).Hypoxanthine uptake was dependent on HPRT1, because cells lacking HPRT1 displayed no net hypoxanthine consumption over time (Figure S7L and S7M).NBMPR reduced labeling of cellular purine nucleotides from [ 15 N 4 ]hypoxanthine (Figure 7B), and diminished purine nucleotide accumulation induced by IACS-010759 (Figure 7C).NBMPR did not alter proliferation of H460 cells under control conditions, but enhanced the effect of IACS-010759 (Figure 7D) and reduced proliferation in H460 cells lacking UQCRC2 (Figure S7N).These data indicate that ETC-deficient cells depend on extracellular hypoxanthine for purine salvage to sustain growth.
It is worth emphasizing that while total purine salvage contributes to approximately 50% of the IMP pool in IACS-010759-treated cells, the fractional enrichment of m+4 IMP was only around 20% after 6 hours of [ 15 N 4 ]hypoxanthine tracing (Figure 5B).These data indicate that approximately 30% of the IMP pool arises from HPRT1-dependent salvage reactions involving unlabeled bases.These unlabeled bases likley arise from purine recycling inside the cell, because unlabeled extracellular hypoxanthine is essentially absent in medium supplemented with dFBS (Figure S7H), including medium used in [ 15 N 4 ]hypoxanthine experiments.These results, along with the glucose and glutamine tracing data above indicate that the bases feeding purine salvage arise both inside and outside the cell, with rapid purine turnover contributing substantially to HPRT1-dependent purine salvage.SLC29A1 overexpression promoted hypoxanthine uptake and blunted the effect of IACS-010759 on cell proliferation (Figures 7E-7G).SLC29A1 overexpression was also sufficient to enhance H460 xenograft growth, suggesting that purine nucleoside uptake is limiting in vivo for growth of these tumors (Figure 7H).In line with this, both SLC29A1 and SLC29A2 are more highly expressed in human lung adenocarcinoma relative to adjacent lungs, suggesting a role in human lung cancer (Figure 7I).

DISCUSSION
We find that mitochondrial metabolism -specifically, the ability to engage in OXPHOSdictates the pathway by which cells maintain pools of purines.ETC-deficient cells exhibit suppressed de novo purine synthesis and require purine uptake and salvage to maximize growth (Figure 7J).We observed alterations in purine metabolites in fibroblasts from patients with mitochondrial dysfunction, and in human NSCLCs where the contribution of glucose to the TCA cycle correlates inversely with markers of purine salvage.We also note that patients with cancer receiving IACS-010759 in a Phase I clinical trial exhibited elevated purine nucIeotides in the blood 46 , and that defects in mitochondrial DNA replication perturb purine-related metabolites in patients and mice 47 .These findings provide support for the disease relevance of our study.
Aspartate is required for the synthesis of SAICAR, a pivotal step in de novo IMP synthesis, and becomes limiting when the cell's ability to recycle NADH to NAD + is impaired 27,28 .Therefore, the suppression of de novo purine synthesis by ETC blockade is intuitively understandable.However, merely restoring intracellular aspartate is insufficient to restore de novo purine nucleotide synthesis in ETC-deficient cells, indicating the involvement of other factors in suppressing de novo purine synthesis under these conditions.For example, ETC dysfunction also disrupts one-carbon metabolism which provides N 10formyl-tetrahydrofolate for de novo purine synthesis pathway 48 .
An intriguing aspect of the data is that ETC inhibition not only results in a switch from de novo purine synthesis to purine salvage to maintain purine monophosphate pools, but also expands these pools.We speculate that accumulation of purine monophosphates is an adaptive response to help cells cope with compromised ETC function.De novo purine nucleotide synthesis from PRPP to IMP is energetically demanding, requiring contributions from glutamine, glycine, aspartate, and N 10 -formyl-tetrahydrofolate, and is subject to feedback inhibition by purine monophosphates 49,50 .Constitutive de novo purine synthesis would be counterproductive and perhaps toxic when the cell's ability to produce and maintain a favorable energy state and pools of required intermediates is insufficient to complete the pathway 51 .The accumulated purine monophosphates may facilitate inhibition of PPAT 52 , which catalyzes the committed step of the de novo pathway, therefore preventing unnecessary energy expenditure.
The low NAD + :NADH ratio induced by ETC dysfunction is a key factor in elevated IMP and hypoxanthine levels.While previous studies have touched on the association between redox imbalance and IMP accumulation in cells with defective OXPHOS 28,[53][54][55] , our study further delineates the distinct roles of compartmentalized NAD + :NADH ratios in modulating metabolic responses to ETC blockade.Additional evidence linking purine accumulation to excess NADH includes the observation that expressing the Escherichia coli pyridine nucleotide EcSTH in HeLa cells reduces the NAD + :NADH ratio while increasing the abundance of several purines, and that ethanol ingestion decreases the NAD + :NADH ratio and induces purine monophosphate accumulation in the mouse liver 56 .Collectively, these observations indicate an important role for the NAD + :NADH ratio in regulating the mode of purine metabolism.
The balance between de novo purine synthesis and purine salvage is particularly relevant in cancer, where mitochondrial function is variable and drugs can be used to inhibit either pathway.In some tumors, mutations in mitochondrial enzymes may render cells permanently reliant on purine salvage.Recent data demonstrate that fumarate hydratase (FH)-deficient renal carcinoma cells have suppressed de novo purine synthesis and require purine salvage 57 .We note that H460 cells, which respire well and do not require HPRT1 for growth in culture, nevertheless require this enzyme for maximal growth of subcutaneous xenografts.These findings indicate that dependence on purine salvage can be imposed on respiration-competent cells by environmental factors.Rapidly growing tumors also experience hypoxia, which may further enhance salvage dependence 34,58,59 .This may also explain why NSCLCs tend to over-express HPRT1, SLC29A1, and SLC29A2.The fact that over-expressing SLC29A1 is sufficient to drive xenograft growth argues that access to purine nucleosides or nucleobases is a limiting factor for H460 cell growth in vivo.

Limitations of the study
Low metabolite abudnace and the limitations of LC/MS resolution prevented us from detecting all purine intermediates and pinpointing the exact mechanism by which ETC blockade suppresses de novo purine synthesis.We also have not fully defined the signals that regulate the switch from the de novo purine synthesis to purine salvage when mitochondrial respiration is compromised.Although the NAD + :NADH ratio is involved, neither Cyto-nor Mito-LbNOX expression in UQCRC2 −/− cells increased this ratio to the level observed in parental cells, and we lack tools to precisely measure the NAD + :NADH ratio in a compartment-specific manner.This leaves the possibility that other aspects of ETC function beyond redox maintenance regulate purine salvage.Finally, it is interesting that ETC inhibition does not enhance hypoxanthine uptake in the assays we used, given that ETC inhibition stimulates HPRT1-dependent salvage and HPRT1 stimulates hypoxanthine import.Given that SLC29A transporters are equilibrative, it may be that ETC blockade also increases the pool of intracellular hypoxanthine such that no increase in net import occurs.Further quantitative analysis of hypoxanthine uptake and metabolism may help resolve this issue.

Lead Contact
Further information and requests for resources and reagents should be directed to the lead contact, Ralph DeBerardinis, MD, PhD. Ralph.DeBerardinis@utsouthwestern.edu

Material Availability
Isogenenic cell lines and DNA constructs generated in this paper are available upon request.

Data and Code Availability:
The raw RNA-seq data reported in this study have been deposited in Gene Expression Omnibus (GEO) (GSE265923).The raw proteomics data in this study have been deposited in MassIVE (MSV000094553).Accession numbers are also listed in the key resource table.An Excel file containing the values to create graphs in the paper and a PDF file containing uncropped scans of all western blots are provided as Data S1.Raw metabolomics data are provided as Data S2.R script used to analyze the data can be found on the GitHub repository (https://github.com/wencgu/nac).

Clinical samples
All patients provided informed consent.Both the inborn errors of metabolism study (NCT02650622) and the lung cancer study (NCT02095808) were approved by the Institutional Review Board (IRB) at University of Texas Southwestern Medical Center (UTSW).For the inborn errors of metabolism study, plasma samples were obtained from fresh blood collected in heparinized tubes at the Children's Medical Center at Dallas.Punch biopsies of the skin for fibroblast culture were obtained from the patient, following standard culture procedures for clinical diagnostics.For the lung cancer study, patients were infused with [U- 13 C]glucose and samples were obtained as described 15 .

NAD + and NADH quantitation
Qualitative analysis of NAD + and NADH was performed as previously described 66 on a QExactive HF-X mass spectrometer (Thermo Scientific, Bremen, Germany).We perfomed quantitative analysis of NAD + and NADH according to our previous protocol 67 on a 6500+ mass spectrometer (AB Sciex, Framingham, MA).To prepare quantitative samples, cells were washed with saline and extracted with 40:20:20 acetonitrile:methanol:water (v/v) and 0.1 M formic acid and then neutralized with 15% ammonium bicarbonate (w/v).A 15 N 5 -AMP internal standard was added to each extract at the final concentration of 100 nM.
Samples were run the same day to minimize oxidation of the analytes of interest.
All cellular extracts were analyzed against an 8-point standard curve ranging from 5 nM to 1000 nM.All standard curves had R 2 values greater than or equal to 0.98 with greater than 6 calibrators having accuracies within 20% of their known concentration.

Targeted metabolomics
To extract metabolites, cells were rinsed with ice-cold saline twice and quenched by 80% cold methanol.Cells were incubated at −80°C for at least 20 minutes and then scraped.For tumor samples, the tissues were thawed and homogenized in cold 80% methanol using plastic pestles (Thermo Fisher Scientific, 12141364).Samples were subjected to three freeze-thaw cycles in liquid nitrogen and a 37°C water bath.Afterwards, the samples were vortexed for 1 minute and spun down at 4°C at 20,160 x g for 15 minutes.The supernatants were transferred into fresh Eppendorf tubes and dried in a SpeedVac concentrator overnight.
To measure metabolites from conditioned media, 10 μL of medium was collected and added into 100 μL of 80% methanol followed by vortexing.The samples were then dried in a SpeedVac concentrator overnight.
Metabolite abundance was analyzed using multiple mass spectrometers.For analysis on a Q-TOF mass spectrometer, dried metabolites were reconstituted in 0.1% formic acid in analytical-grade water and vortexed for 1 minute before spinning at 4°C at Raw data files (.d) were processed using Profinder B.08.00 SP3 software (Agilent Technologies, CA) with an in-house database containing retention time and accurate mass information on 600 standards from Mass Spectrometry Metabolite Library (IROA Technologies, MA).The in-house database matching parameters were: mass tolerance 10 ppm; retention time tolerance 0.5 min.Peak integration results were manually curated in Profinder for improved consistency and exported as a spreadsheet (.csv).
Samples prepared for analysis on a Q-Exactive were reconstituted in 80% acetonitrile and centrifuged at 4°C at 20,160 x g to remove insoluable material.Chromatographic separation of metabolites was carried out on a Vanquish UHPLC system equipped with a ZIC-pHILIC column (Millipore-Sigma, Burlington, MA) as previously described 66,68,69 .Extracted ion chromatograms (XICs) were generated with a mass tolerance of 5 ppm and integrated for relative quantitation.Identities of analytes were confirmed with purified standards and product ion spectra.
Principal component analyses and metabolite set enrichment analyses were conducted using MetaboAnalyst 5.0 70 .

Stable isotope tracing
For tracing with 13  The metabolites were extracted as described in the targeted metabolomics method.For analysis by Q-TOF, data acquisition was performed and analyzed according to the methods described above.
For analysis on the Q-Exactive, tSIM methods were used to increase the signal of isotopically-labeled intermediates of the purine biosynthetic pathway and pentose phosphate pathway.Both 15 N and 13 C nuclei were analysed using this approach.Quadrupole isolation windows for individual analytes were set to capture all relevant nuclei to calculate fractional enrichment values.We performed analysis of isotopologues according to our previously reported method 71 .Natural isotope abundances were corrected using a customized R script, which can be found at the GitHub repository (https://github.com/wencgu/nac).The script was written by adapting the AccuCor algorithm 72 .
For targeted analysis of purines and hypoxanthine tracing using AB SCIEX QTRAP 5500 LC/triple quadrupole MS (Apllied Biosystems SCIEX), metabolites were reconstituted in 0.1% formic acid in analytical water, vortexed, and spun down to remove insoluble material before being loaded onto the instrument as previously described 73

Gas chromatography/mass spectrometry (GC/MS)
Metabolites were extracted as described in the targeted metabolomics method and 1 μL D 27 -myristic acid was added as an internal control.The dried metabolites were re-suspended in 40 μL anhydrous pyridine and transferred to GC/MS autoinjector vials.The samples were incubated at 70°C for 15 min, followed by addition of 80 μL N-(tert-butyldimethylsilyl)-N-methyltrifluoroacetamide (MTBSTFA) derivatization reagent, as previously described 15 .
The samples were incubated at 70°C for 1 hour before being subjected to GC-MS analysis.1 μL of the sample was injected for analysis on Agilent 6890 or 7890 gas chromatographs coupled to an Agilent 5973N or 5975C Mass Spectrometer.The data were analyzed using EL-MAVEN, and observed distributions of mass isotopologues were corrected for natural abundance using a customized R script on the GitHub repository (https://github.com/wencgu/nac).

HPRT1 enzymatic activity analysis
HPRT1 enzyme activity was measured using the Precise HPRT1 assay kit (Novo CIB, K0709-01-2) according to manufacturer's instructions.In brief, H460 cells were seeded in 10 cm plates and treated with DMSO or 25 nM IACS-010759 for 24 hours.Cells were rinsed once with PBS, scraped, and lysed in ice-cold lysis buffer containing 10 mM Tris-HCl pH 7.4, 150 mM NaCl, 1% NP-40, 1 mM EDTA followed by centrifugation at 18,000 x g for 10 minutes at 4°C.Each enzymatic reaction contained 5 μL sample or positive control (human recombinant HPRT enzyme) and 200 μL reaction mixture containing DTT (cofactor 1), NAD (cofactor 2) and bacterial IMPDH in the absence (blank) or presence of 2 mM PRPP (enzyme reaction).The reaction was performed at 37°C.The absorbance at 340 nm was recorded at 2-minutes intervals for 120 minutes.HPRT1 protein abundance in each sample was assessed by immunoblotting using antibody against HPRT1.Image J was used to quantify the HPRT1 band intensity.HPRT1 catalytic rate was normalized to the protein abundance in each group with the DMSO control samples given a value of 1.

Glucose uptake and lactate secretion assay
Cells were seeded in 6 cm plates.The glucose uptake and lactate secretion analysis was started when cells reached 90% confluence.Cells were washed with PBS once. 2 mL medium containing 25 nM IACS-010759 or equal volume of DMSO was added into the plates for 6 hours.Medium from each plate was collected and spun down at 20,160 x g at 4°C. 1 mL supernatant of each sample was transferred to a fresh eppendorf tube and loaded in a NOVA instrument to measure glucose and lactate levels.Three or four tubes containing medium but no cells were used as blanks to calculate the amount of glucose and lactate taken up or secreted by cells.The cell number was counted from each plate to calculate the rate of glucose uptake and lactate secretion per cell.

Xenograft studies in mice
All mouse experiments complied with relevant ethical regulations and were performed according to protocols approved by the Institutional Animal Care and Use Committee at the University of Texas Southwestern Medical Center (protocols 2016-101360 and 2016-101694).H460 cells were suspended in serum-free RPMI medium and mixed with Matrigel (Thermo Fisher Scientific, CB-40234) at 1:1 volume ratio.One million cells were subcutaneously injected into the right flank of NOD.Cg-Psrkdcscid Il2rgtm1Wjl/SzJ (NSG) mice.Mice were randomized for IACS-010759 treatments and then administered vehicle or IACS-010759 daily through oral gavage (5 or 10mg/kg body mass in 100 μL of 0.5% methylcellulose and 4% DMSO) 26,30 .For tumor metabolomics and glutamine infusion experiments, mice were dosed with (10mg/kg) IACS-010759 for 5 days prior to the infusion as previously described 30 .After the infusion, tumors were collected and snap frozen for later metabolite extraction and LC-MS analyses.For tumor growth analyses, mice were administered 5 mg/kg IACS-010759 until the day before sacrifice.Two orthogonal measurements of tumor diameter were collected every other day and tumor volume was calculated using the formula V = (L 1 x(L 2 2 ))/2.

[amide-15 N]glutamine infusion
Mice were anesthetized using (30 mg/mL) ketamine/xylazine mix (30 μL/g).25-gauge catheters were placed in the lateral tail vein under anesthesia.The total dose of glutamine was 1.725 g/kg dissolved in 1.5 mL saline.Isotope infusions started with 150 μL/minute bolus for 1 minute followed by continuous infusion at rate of 150 μL/hour for 4 hours.Upon termination of the infusions, animals were euthanized immediately and tumors were collected and snap frozen in liquid nitrogen.

Seahorse XFe96 Respirometry
An XFe96 Extracellular Flux Analyzer (Agilent Technologies) was used to measure oxygen consumption rate.In brief, 20,000 cells per well were seeded and simultaneously treated with 25 nM IACS-010759.After 16 to 20 hours, cells were washed three times with Seahorse medium (Agilent Technologies, 102353) containing 2 mM glutamine, 1 mM pyruvate, 10 mM glucose and pen/strep (pH 7.4) and incubated in a CO 2 -free incubator at 37°C for at least 30 minutes prior to loading into the instrument.Final concentrations for oligomycin A, carbonyl cyanide m-chlorophenylhydrazone (CCCP), and rotenone were 2 μM, 1 μM, and 2 μM respectively.After the assay, cells were counted using Celigo Image Cytometer (see method: cell growth analysis) to normalize oxygen consumption rate.

Immunofluorescence and confocal microscopy
Coverslips were coated with 10 μg/mL fibronectin (Sigma-Aldrich, F1141-5MG) for 1 hour at 37°C and rinsed once with PBS.Cells were immediately seeded on the coverslips.Cells were fixed the next day with fresh warm 4% paraformaldehyde (PFA) solution in PBS for 15 minutes followed by permeabilization using 0.1% (v/v) Triton X-100 in PBS at room temperature for 10 minutes.Cells were then blocked in filtered PBS containing 1% BSA for at least 30 minutes at room temperature before incubation with primary antibodies against FLAG (1:200, F1804, Sigma-Aldrich) and HSP60 (1:500, 12165S, CST) for 1 hour at room temperature.Cells were washed 3 times for 5 minutes with PBS and then incubated with secondary antibodies (Alexa fluorophores 488 and 555, Invitrogen) for 1 hour in dark at room temperature.Coverslips were washed with PBS 3 times for 5 minutes and Mili-Q water once before being mounted on slides by Profade-Antifade (P36935, Invitrogen) overnight in dark.Cells were imaged using Zeiss LSM 880 Confocal Laser Scanning Microscope with Z-stacks acquired.The images labeled "Merge" are composites.All representative images were processed using Image J.

RNA-seq
RNA was extracted using Trizol (Thermo Fisher Scientific, 15596018) and an RNeasy Mini Kit (Qiagen, 74106).A Qubit fluorometer and Invitrogen Qubit RNA High Sensitivity kit (Invitrogen, Q32852) were used to measure total RNA levels.RNA-seq libraries were prepared using the NEBNext Ultra II directional RNA library prep kit with the NEBNext Poly(A) mRNA magnetic isolation module (New England Biolabs, E7490L, E7760L) according to manufacturer's instructions.Libraries were stranded using standard N.E.B indices according to manufacturer's instructions (New England Biolabs, E7730L, E7335L, E7500L).Sequencing reads from all RNA-seq experiments were aligned to hg19 reference genome by STAR v. 2.5.2b 74with the following parameters: --runThreadN 28 --outSAMtype BAM SortedByCoordinate --outFilterMultimapNmax 1 --outWigStrand Unstranded --quantMode TranscriptomeSAM.Output BAM files were converted to BED format using the "bamtobed" command from BEDtools v.2.29.2 [https:// bedtools.readthedocs.io/en/latest/].BED files were then converted to a normalized wiggle file using a custom python script.Normalized wiggle files were then converted to bigwig format using wigToBigWig with "-clip" parameter.Read counts were derived using HTSeq 75 with parameter "-s no" and 1 additional read count was added to each gene for each independent sample prior to downstream analyses.Differentially expressed genes were identified by DESeq2 (fold change ≥ 1.5, FDR-adjusted P value ≤ 0.05) 76 .Fragments Per Kilobase Of Exon Per Million Fragments Mapped (FPKM) value of genes were calculated by normalizing the gene length and sequencing depth.

Quantitative proteomics
H460 cells were treated with 25 nM IACS-010759 or DMSO for 24 hours.To isolate protein, cells were washed twice with ice-cold PBS followed by addition of freshly prepared lysis solution consisting of 5% SDS in 50 mM TEAB with protease and phosphatase inhibitors.Cells were scraped in lysis solution and transferred to 1.5 mL Eppendorf tubes.Samples were allowed to sit at room temperature for 15 minutes to complete lysis.Protein concentration was calculated with a BCA assay and all samples were normalized to the same protein concentration.Following disulfide bond reduction and alkylation, samples were digested overnight with trypsin using an S-Trap (Protifi).The peptide eluate from the S-Trap was dried and reconstituted in 100 mM TEAB buffer.A TMT10plex Isobaric Mass Tagging Kit (Thermo) was used to label the samples as per the manufacturer's instructions.The combined sample then underwent solid-phase extraction cleanup with an Oasis HLB plate (Waters) and was dried in a SpeedVac.The sample was then reconstituted in a 2% acetonitrile, 0.1% TFA buffer and diluted such that ~1 ug of peptides were injected.
Peptides were analyzed on a Thermo Orbitrap Eclipse MS system coupled to an Ultimate 3000 RSLC-Nano liquid chromatography system.Samples were injected onto a 75 um i.d., 75-cm long EasySpray column (Thermo) and eluted with a gradient from 0-28% buffer B over 180 minutes at a flow rate of 250 nL/minute.Buffer A contained 2% (v/v) ACN and 0.1% formic acid in water, and buffer B contained 80% (v/v) ACN, 10% (v/v) trifluoroethanol, and 0.1% formic acid in water.at a flow rate of 250 nL/minute.Spectra were continuously acquired in a data-dependent manner throughout the gradient, acquiring a full scan in the Orbitrap (at 120,000 resolution with a standard AGC target) followed by MS/MS scans on the most abundant ions in 2.5 s in the ion trap (turbo scan type with an intensity threshold of 5,000, CID collision energy of 35%, standard AGC target, maximum injection time of 35 ms and isolation width of 0.7 m/z).Charge states from 2-6 were included.Dynamic exclusion was enabled with a repeat count of 1, an exclusion duration of 25 s and an exclusion mass width of ± 10 ppm.Real-time search was used for selection of peaks for SPS-MS3 analysis, with searched performed against the human reviewed protein database from UniProt.Up to 1 missed tryptic cleavage was allowed, with carbamidomethylation (+57.0215) of cysteine and TMT reagent (+229.1629) of lysine and peptide N-termini used as static modifications and oxidation (+15.9949) of methionine used as a variable modification.MS3 data were collected for up to 10 MS2 peaks which matched to fragments from the real-time peptide search identification, in the orbitrap at a resolution of 50,000, HCD collision energy of 65% and a scan range of 100-500.
Protein identification and quantification used Proteome Discoverer v.         p(HR)=0.00097.The plot and statistics were generated using GEPIA 2 62 .F. Schematic illustrating intra-operative [U- 13 C]glucose infusion in patients with NSCLC followed by tumor resection and multi-omics analyses.G. Fractional enrichment of m+2 malate and relative IMP abundance in tumors displaying low or high malate labeling.The analysis was performed on the top and bottom 25% of tumors for malate m+2 labeling (n=7 tumors each with both isotope tracing and metabolomics analysis).
H. Fractional enrichment of m+2 malate and HPRT1 mRNA levels in tumors displaying low or high malate labeling.The analysis was performed on the top and bottom 25% of tumors for malate m+2 labeling (n=6 tumors each with both isotope tracing and RNA-Seq analysis).Unpaired, two-sided t tests (A, B, G, and H), and a paired t test (D) were used for the statistical analyses.****: P < 0.0001; ***: P < 0.001; **: P < 0.01, n.s.: P > 0.05.Error bars denote SEM.BioRender was used to generate the illustration.

•
Cytosolic NAD(H) imbalance induces purine accumulation in ETC-deficient cells•HPRT1-mediated purine salvage supports NSCLC growth during ETC inhibition•Purine uptake is required to supply salvage upon ETC inhibition

Figure 1 .
Figure 1.Metabolomic profiling of patients with mitochondrial disorders.A. Illustration of the human mitochondrial defects analyzed in panels B-E.B. and C. Altered metabolite abundances in the TCA cycle (B) and purine metabolism (C) in fibroblasts from patients with the indicated mitochondrial defects.Each dot represents a fibroblast line from a patient with a disorder affecting the mitochondria.The mutated gene from each disorder is indicated.D. and E. Plasma metabolite levels from a patient with LIPT1 deficiency and healthy controls.n = 60 (healthy); n = 28 (LIPT1 deficiency; samples collected on different days).

Figure 3 .
Figure 3. Cytosolic NAD(H) imbalance impacts purine accumulation in ETC-deficient cells.A. Schematic of LbNOX-catalyzed reaction.B. Western blot validating expression of Flag-tagged LbNOX in UQCRC2 −/− H460 cells.Vinculin is the loading control.C. Immunofluorescence showing the subcellular localization of the indicated Flag-tagged LbNOX proteins in UQCRC2 −/− H460 cells.HSP60 is a mitochondrial matrix marker.Scale bar represents 10 μm.

Figure 4 .
Figure 4. ETC blockade suppresses de novo purine nucleotide synthesis.A. Schematic illustrating13 C labeling of purines from [U-13 C]glucose.B.13 C labeling in IMP after 6 hours of culture with [U-13 C]glucose in control and NDI1expressing H460 cells pre-treated with DMSO or 25 nM IACS-010759 for 24 hours (n=3).C. Schematic illustrating15 N labeling from [amide-15 N]glutamine during de novo purine nucleotide synthesis.
15glucose, cells were cultured in base RPMI medium (Sigma, R1383-L) supplemented with 11 mM [U-13C]glucose (Cambridge Isotope Laboratories, CLM481-0.25) or [1,2-13 C]glucose (Cambridge Isotope Laboratories, CLM-504-0.5) and 10% dialyzed FBS (Gemini Bio-Products, 100108).For tracing with [amide-15N]glutamine, 3.0 SP1 (Thermo).Raw MS data files were analyzed against the human reviewed protein database from UniProt.Both Comet and SequestHT with INFERYS Rescoring were used, with carbamidomethylation (+57.0215) of cysteine and TMT reagent (+229.1629) of lysine and peptide N-termini used as static modifications and oxidation (+15.9949) of methionine used as a variable modification.Reporter ion intensities were reported, with further normalization performed by using the total intensity in each channel to correct discrepancies in sample amount in each channel.The false-discovery rate (FDR) cutoff was 1% for all peptides.Extracted reporter ions were further normalized by using the total intensity in each channel to correct for differences in sample amounts.
Error bars denote SEM.BioRender was used to generate the illustration.
Cell Metab.Author manuscript; available in PMC 2024 July 12.