Cu/Zn superoxide dismutase (Sod1) is a key antioxidant enzyme, and its importance is underscored by the fact that its ablation in cell and animal models results in oxidative stress; metabolic defects; and reductions in cell proliferation, viability, and lifespan. Curiously, Sod1 detoxifies superoxide radicals (O2•−) in a manner that produces an oxidant as byproduct, hydrogen peroxide (H2O2). While much is known about the necessity of scavenging O2•−, it is less clear what the physiological roles of Sod1-derived H2O2 are. We discovered that Sod1-derived H2O2 plays an important role in antioxidant defense by stimulating the production of NADPH, a vital cellular reductant required for reactive oxygen species scavenging enzymes, as well as redox regulating a large network of enzymes.
Cu/Zn superoxide dismutase (Sod1) is a highly conserved and abundant antioxidant enzyme that detoxifies superoxide (O2•−) by catalyzing its conversion to dioxygen (O2) and hydrogen peroxide (H2O2). Using Saccharomyces cerevisiae and mammalian cells, we discovered that a major aspect of the antioxidant function of Sod1 is to integrate O2 availability to promote NADPH production. The mechanism involves Sod1-derived H2O2 oxidatively inactivating the glycolytic enzyme, GAPDH, which in turn reroutes carbohydrate flux to the oxidative phase of the pentose phosphate pathway (oxPPP) to generate NADPH. The aerobic oxidation of GAPDH is dependent on and rate-limited by Sod1. Thus, Sod1 senses O2 via O2•− to balance glycolytic and oxPPP flux, through control of GAPDH activity, for adaptation to life in air. Importantly, this mechanism for Sod1 antioxidant activity requires the bulk of cellular Sod1, unlike for its role in protection against O2•− toxicity, which only requires <1% of total Sod1. Using mass spectrometry, we identified proteome-wide targets of Sod1-dependent redox signaling, including numerous metabolic enzymes. Altogether, Sod1-derived H2O2 is important for antioxidant defense and a master regulator of metabolism and the thiol redoxome.
Superoxide dismutases (SODs) serve on the frontline of defense against reactive oxygen species (ROS). SODs, which detoxify O2•− by catalyzing its disproportionation into O2 and hydrogen peroxide (H2O2), are unique among antioxidant enzymes in that they also produce a ROS byproduct. While much is known about the necessity of scavenging O2•−, it is less clear what the physiological consequences of SOD-derived H2O2 are. Paradoxically, increased expression of Cu/Zn SOD (Sod1), which accounts for the majority of SOD activity in cells (1), is associated with reduced cellular H2O2 levels (2), suggesting there may be additional unknown mechanisms underlying Sod1 antioxidant activity.
The cytotoxicity of O2•− stems from its ability to oxidize and inactivate [4Fe-4S] cluster-containing enzymes, which results in defects in metabolic pathways that utilize [4Fe-4S] proteins and Fe toxicity due to its release from damaged Fe/S clusters (3⇓⇓–6). The released Fe can catalyze deleterious redox reactions and, in particular, production of hydroxyl radicals (•OH) via Haber-Weiss and Fenton reactions, which indiscriminately oxidizes lipids, proteins, and nucleic acids (4, 7). The importance of Sod1 in oxidative stress protection is underscored by reduced proliferation, decreased lifespan, and numerous metabolic defects, including cancer, when SOD1 is deleted in various cell lines and organisms (7⇓⇓⇓–11). It was previously proposed that Sod1 limits steady-state H2O2 levels because of its ability to prevent the O2•−-mediated oxidation of Fe/S clusters, which results in the concomitant formation of H2O2 (2, 12, 13). However, since vanishingly small amounts of Sod1 (<1% of total cellular Sod1) is sufficient to protect cells against O2•− toxicity, including oxidative inactivation of Fe/S enzymes (14⇓–16), any changes in Sod1 expression would not be expected to alter H2O2 arising from O2•− oxidation of Fe/S clusters. How then can Sod1, an enzyme that catalyzes H2O2 formation, act to reduce cellular [H2O2]?
Two previously reported but unexplained metabolic defects in sod1Δ strains of Saccharomyces cerevisiae point to a potential role for Sod1 in regulating the production of NADPH, a key cellular reductant required for reductive biosynthesis and the reduction and regeneration of H2O2 scavenging thiol peroxidases (17) and catalases (18, 19). Yeast strains lacking SOD1 exhibit increased glucose consumption (20) and defects in the oxidative phase of the pentose phosphate pathway (oxPPP) (21), the primary source of NADPH. Inhibition of key rate-limiting enzymes in glycolysis—including phosphofructose kinase (22), GAPDH (23, 24), and pyruvate kinase (25, 26)—reduces glucose uptake (27⇓–29) and increases the concentration of glucose-6-phosphate (G6P), a glycolytic intermediate that is also the substrate for the first enzyme in the oxPPP, G6P dehydrogenase (G6PDH), which in turn increases oxPPP flux and NADPH production (30⇓⇓⇓⇓–35). Taking these data together, we surmised that Sod1 negatively regulates a rate-determining enzyme in glycolysis, thereby accounting for the observed metabolic defects in glucose utilization and the oxPPP in sod1Δ cells (20, 21).
GAPDH, which catalyzes a rate-determining step in glycolysis (36, 37), is very abundant (38), and contains a H2O2-reactive catalytic Cys (k ∼ 102 to 103 M−1s−1), represents a critical redox regulated node that can toggle flux between glycolysis and the oxPPP (32). As such, we hypothesized that a novel aspect of the antioxidant activity of Sod1 is to oxidatively inactivate GAPDH using Sod1-catalyzed H2O2, which would in turn stimulate NADPH production via the oxPPP and enhance cellular peroxide scavenging by thiol peroxidases. This mechanism for Sod1-mediated antioxidant activity would explain a number of prior observations, including the findings that elevated Sod1 expression decreases peroxide levels and loss of SOD1 increases glucose consumption and attenuates oxPPP activity. In addition, more generally, since Sod1-derived H2O2 has previously been implicated in the redox regulation of other enzymes, including protein tyrosine phosphatases (39) and casein kinases (15, 16, 40), we also sought to identify proteome-wide redox targets of Sod1.
In the present report we provide evidence highlighting an antioxidant function for Sod1-derived H2O2 in integrating O2 availability to control NADPH production to support aerobic growth and metabolism. The mechanism involves the conversion of O2 to O2•− by mitochondrial respiration and an NADPH oxidase, followed by the Sod1-catalyzed conversion of O2•− to H2O2, which in turn oxidatively inactivates GAPDH. The inhibition of GAPDH serves to reroute metabolism from glycolysis to the oxPPP in order to maintain sufficient NADPH for metabolism in air. The aerobic oxidation of GAPDH is dependent on and rate-limited by Sod1, suggesting that it provides a privileged pool of peroxides to inactivate GAPDH under physiological conditions. Finally, we revealed a larger network of cysteine-containing proteins that are oxidized in a Sod1-dependent manner using mass spectrometry-based redox proteomics approaches. Altogether, these results highlight a mechanism for O2 sensing and adaptation, reveal an important but previously unknown antioxidant role of Sod1 that goes beyond O2•− scavenging to include the stimulation of aerobic NADPH production, and places Sod1 as a master regulator of proteome-wide thiol oxidation and multiple facets of metabolism.
Sod1 Regulates Glycolysis.
In many eukaryotes, including S. cerevisiae, glucose uptake negatively correlates with pO2 (41⇓⇓–44). Indeed, we find that batch cultures of WT yeast cells grown anaerobically consume more glucose per cell than aerobically grown cultures (Fig. 1A). Media glucose concentration is plotted versus cell density, rather than time, to correct for differences in growth rate (SI Appendix, Fig. S1 A and B). Consistent with previous studies (20), aerobic cultures of sod1Δ strains consume more glucose than WT cells (Fig. 1A). However, in the absence of oxygen, both WT and sod1Δ cells consume similar amounts of glucose (Fig. 1A).
Since defects in glycolytic enzymes, including hexokinase 2 or GAPDH, decrease glucose uptake (27, 28), we sought to determine if the increased glucose consumption of sod1Δ cells was associated with altered glycolytic flux. Cytosolic pH is a reporter of glycolytic activity and can be monitored using the GFP-based ratiometric pH sensor, pHluorin. Supplementing glucose-starved WT yeast with glucose results in a rapid decrease in pH due to proton release from glycolytic phosphorylation reactions (Fig. 1 B and C, phase 1) (45), followed by a slower realkalization phase due to activation of Pma1, a cell surface H+-ATPase that pumps H+ into the extracellular space (Fig. 1 B and C, phase 2) (45). In response to glucose, sod1Δ cells exhibit more rapid rates of intracellular acidification, indicating that glycolysis is more active compared to WT cells. In contrast, sod1Δ cells exhibit a diminished rate of realkalization, indicating that Pma1 is less active in response to glucose, which is consistent with prior work that found that sod1Δ cells exhibit a defect in Pma1 activity (16, 46). To rule out that the intracellular acidification phase is affected by Pma1-dependent realkalization, glucose-dependent changes in intracellular pH were monitored in a strain expressing a hypomorphic allele of PMA1, pma1-tap, containing a C-terminal tandem affinity purification (TAP)-tag (SI Appendix, Fig. S1C). In pma1-tap cells, the rate of glucose-induced acidification is similar to WT cells (Fig. 1C, phase 1), but there is a significant decrease in the rate of realkalization (Fig. 1C, phase 2). Taken together, the data demonstrate that Sod1 negatively regulates glucose uptake and glycolytic activity.
Sod1 Interacts with and Regulates GAPDH Oxidation and Activity.
We hypothesized that Sod1-derived H2O2 may negatively regulate glucose uptake and glycolytic activity through the oxidative inactivation of GAPDH, which catalyzes a rate-determining step in glycolysis and contains a peroxide-sensitive active site Cys (23). S. cerevisiae encodes three GAPDH isoforms—TDH1, TDH2, and TDH3—with TDH3 being the most highly expressed in log-phase cultures, accounting for >50% of total cellular GAPDH (47). Yeast GAPDH has only two cysteines, catalytic C150 and C154, which sensitizes C150 to oxidation by H2O2 (SI Appendix, Fig. S1D). In order to probe the Sod1-dependence of GAPDH oxidation, we employed a thiol alkylation assay that exploits the reactivity of methoxypolyethylene glycol maleimide (mPEG-mal) with reduced but not oxidized thiols (48). The extent of GAPDH labeling with mPEG-mal, which is 5 kDa, is assessed by determining the changes in electrophoretic mobility of PEGylated GAPDH, corresponding to single- and double-labeled GAPDH at reduced C150 and/or C154 (Fig. 1D and SI Appendix, S1 E and F). Thus, the fraction of GAPDH oxidized in vivo can be determined by quantifying the ratio of the intensity of unlabeled GAPDH (oxidized GAPDH) to total GAPDH (SI Appendix, Fig. S1 F–O). The mPEG-mal approach was validated by treating cells with H2O2 and observing a twofold increase in GAPDH oxidation (SI Appendix, Fig. S1 F and G). Moreover, the identity of the specific sites of mPEG-mal labeling was confirmed by observing that a yeast strain expressing a single allele of Tdh3C154S was found to have only two GAPDH proteoforms corresponding to single and unlabeled GAPDH (SI Appendix, Fig. S1P).
To determine if Sod1 oxidizes GAPDH in vivo, we analyzed GAPDH PEGylation in sod1Δ cells expressing empty vector (EV) (sod1Δ + EV) or WT SOD1 (sod1Δ + SOD1). sod1Δ + EV cells exhibit an increase in mPEG-mal labeling compared to sod1Δ + SOD1 cells (Fig. 1 D and E and SI Appendix, Fig. S2 A and B), indicating that GAPDH is more oxidized in cells expressing SOD1. Analysis of 17 replicates over 6 independent trials revealed that expression of SOD1 results in an approximately twofold increase in GAPDH oxidation, with the median and average GAPDH oxidation in cells expressing SOD1 being 26% and 35%, respectively, and 13% and 16%, respectively, in cells lacking SOD1 (Fig. 1E). The values of GAPDH oxidation in SOD1 expressing cells are on par with prior studies of glucose-grown WT yeast that found GAPDH is 26% oxidized (49). Due to variations in the absolute values of GAPDH oxidation across multiple trials (Fig. 1E), subsequent comparisons of GAPDH oxidation involved normalizing GAPDH oxidation to that of SOD1-expressing cells within each trial, though absolute values are indicated in SI Appendix.
In order to assess the extent to which Sod1 mediates the reversible oxidation of GAPDH, we employed an mPEG-mal labeling scheme that involved first capping free GAPDH thiols with N-ethylmaleimide, followed by DTT reduction of reversibly oxidized thiols and labeling them with mPEG-mal (SI Appendix, Fig. S2 C–G). Using this strategy, it was similarly found that Sod1 accounts for an approximately twofold increase in reversible GAPDH oxidation, with three independent trials giving an average GAPDH oxidation value of ∼50% in SOD1 expressing cells and 25% in cells lacking SOD1 (SI Appendix, Fig. S2 C–G). In addition, >95% of GAPDH is reducible and alkylatable in both WT and sod1Δ cells, indicating that differences in apparent GAPDH oxidation are not due to variations in the ability to reduce and alkylate GAPDH in the two strains (SI Appendix, Fig. S2H).
We next found that Sod1 interacts with and regulates GAPDH activity. GAPDH catalyzes the oxidative phosphorylation of GAP to 1,3 BPG and requires Cys150 for activity. Cells lacking SOD1 (sod1Δ + EV) exhibit threefold greater GAPDH activity (Fig. 1F and SI Appendix, Fig. S3 A and B), which correlates with the approximately twofold decrease in GAPDH thiol oxidation relative to SOD1-expressing cells (Fig. 1 D and E). Moreover, GAPDH and Sod1 interact in vivo as assessed by coimmunoprecipitation (co-IP) experiments (Fig. 1G and SI Appendix, Fig. S2I). IP of HA-Tdh3 using anti-HA antibody resulted in the detection of Sod1 by immunoblotting. In contrast, a control strain expressing untagged Tdh3 did not result in co-IP of Sod1, indicating that the apparent interaction between Sod1 and GAPDH is not due to Sod1 nonspecifically binding to the anti-HA resin.
Sod1-mediated GAPDH oxidation is not dependent on the carbon source. Galactose is a fermentable carbon source that alleviates glucose-mediated respiration repression, resulting in more mitochondrial respiratory activity (16). The absolute and relative amounts of GAPDH oxidation in sod1Δ + EV and sod1Δ + SOD1 cells cultured in 2% galactose is similar to cells cultured in 2% glucose (SI Appendix, Fig. S3 C–E).
The effect of Sod1 localization on GAPDH oxidation was also evaluated. Sod1 is primarily cytosolic but is also present in the mitochondrial intermembrane space (IMS). sod1Δ cells expressing an IMS-targeted allele of SOD1, Sco2-SOD1 (15, 16, 50), exhibit comparable GAPDH oxidation to sod1Δ + EV cells, both of which are significantly lower than cells expressing WT SOD1 (Fig. 1H and SI Appendix, Fig. S3 F and G). Altogether, these results indicate that extramitochondrial Sod1 interacts with and oxidatively inactivates GAPDH, thereby explaining the previous observations that sod1Δ cells exhibit increased glucose uptake (Fig. 1A) and glycolytic activity (Fig. 1 B and C).
Yno1 and Mitochondrial Respiration Are Sources of Superoxide for GAPDH Oxidation.
Sod1 requires a superoxide source to catalyze peroxide production for the control of GAPDH oxidation and activity. As with higher eukaryotes, two primary sources of superoxide in yeast include mitochondrial respiration and the yeast NADPH oxidase, Yno1 (51). Both sources contribute toward GAPDH oxidation as respiration-deficient rho0 and yno1Δ cells exhibit an approximately threefold lower degree of GAPDH oxidation relative to WT cells and phenocopy the sod1Δ mutant (Fig. 2A and SI Appendix, Fig. S4 A and B). Furthermore, overexpression of Yno1 on a galactose-inducible promoter (pYES-YNO1, 3%GAL), which resulted in a ∼30% increase in dihydroethidium (DHE)-detectable superoxide (SI Appendix, Fig. S4C) promoted GAPDH oxidation by twofold as compared to cells expressing EV (pYES2-EV, 0 and 3%GAL) or that were cultured in noninducing media (pYES2-YNO1, 0%GAL) (SI Appendix, Fig. S4 D–F). In total, these results indicate that both Yno1 and mitochondrial respiration are sources of the superoxide substrate required by Sod1 to drive the H2O2-dependent oxidation of GAPDH.
O2-Dependent GAPDH Oxidation Is Dependent on and Rate-Limited by Sod1.
All metabolic sources of superoxide and peroxide are ultimately derived from O2. We therefore sought to determine if Sod1 is the sole enzymatic adapter that links oxygen availability to the control of GAPDH oxidation and if GAPDH oxidation is rate-limited by Sod1. Toward this end, we first asked if Sod1 mediates O2-dependent GAPDH oxidation. Since WT Sod1 is transcriptionally and posttranslationally down-regulated in response to hypoxia and anoxia (52⇓⇓–55), we utilized sod1Δ cells expressing ADH1-driven Sod1P144S, which is a mutant previously engineered to constitutively express mature enzymatically active Sod1 even in the absence of O2 (16, 54). Indeed, the Sod1P144S mutant is enzymatically active in lysates derived from both aerobic and anaerobic cultures, whereas WT Sod1 is only fully active in lysates derived from aerobically cultured cells (Fig. 2B and SI Appendix, Fig. S4G). sod1Δ cells expressing WT or Sod1P144S exhibit a nearly twofold decrease in GAPDH oxidation when cultured anaerobically, consistent with the requirement for O2 as the metabolic origin of superoxide and peroxide (Fig. 2C and SI Appendix, Fig. S4 H and I). However, remarkably, the O2-dependence of GAPDH oxidation is completely lost in cells lacking SOD1. Furthermore, we determined that the oxidation of GAPDH is rate limited by Sod1, finding that GAL-regulated titration of Sod1 levels results in a dose-dependent increase in GAPDH oxidation (Fig. 2 D and E and SI Appendix, Fig. S4J). Altogether, these results indicate that the aerobic oxidation of GAPDH is dependent on and rate-limited by Sod1. It is also worth noting that there is a certain level of O2-independent GAPDH oxidation that is not dependent on Sod1, albeit the nature of the oxidant in this case is not known.
Sod1-Mediated Oxidative Inactivation of GAPDH Results in Increased NADPH Production and Oxidative Stress Resistance.
We next determined if the Sod1-dependent oxidative inactivation of GAPDH results in rerouting of glycolytic metabolism toward oxPPP to produce NADPH and increase resistance to oxidative stress. Titration of Sod1 expression using a GAL-driven SOD1 allele results in both a dose-dependent increase in GAPDH oxidation as well as NADPH levels (Fig. 3 A and B and SI Appendix, Fig. S5 A–F). Control experiments in which GAL is titrated into cells expressing a non-GAL–driven SOD1 allele indicate that GAL alone does not alter NADPH levels (SI Appendix, Fig. S5G). Moreover, tdh3Δ cells, which express ∼60% less GAPDH than WT cells (Fig. 3C and SI Appendix, Fig. S5H) (47), exhibit increased NADPH levels (Fig. 3D), consistent with the finding that oxidative inactivation of GAPDH increases NADPH.
NADPH is required for the reduction and regeneration of numerous cellular antioxidants. Therefore, we determined if the alterations in Sod1-mediated NADPH production correlated with the oxidation state of Tsa1, a yeast peroxiredoxin that scavenges peroxide and that is maintained in a reduced state using NADPH. Peroxide oxidizes Cys47 to a sulfenic acid, which in turn forms a disulfide oxidized dimer with Cys170 on another Tsa1 monomer. At high peroxide concentrations, (>250 μM exogenous H2O2), Cys47 is overoxidized to sulfinylated or sulfonylated states rather than becoming disulfide oxidized. To assess Tsa1 oxidation state, we expressed an N-terminal HA tagged allele of TSA1 in yeast and ascertained its oxidation state by immunoblotting for monomeric or disulfide oxidized dimeric states of Tsa1 using anti-HA antibodies or sulfonylated states using an anti–Prx-SO3 antibody. HA-Tsa1 complements the H2O2 sensitivity of tsa1Δ cells (SI Appendix, Fig. S6A), and we can assign various proteoforms corresponding to reduced monomeric and reversibly oxidized monomeric and dimeric Tsa1 based on the electrophoretic mobility of HA-Tsa1 and HA-Tsa1C47S expressed in WT and tsa1Δ cells on SDS/PAGE gels (SI Appendix, Fig. S6 B–D). zwf1Δ cells lacking G6PD, which catalyzes the first committed step of the oxPPP, exhibited decreased NADPH levels (Fig. 3E) and elevated disulfide oxidized Tsa1 dimers (Fig. 3 F and G), thereby confirming that NADPH limitation results in elevated Tsa1 oxidation. Likewise, sod1Δ cells also exhibited increased disulfide oxidized Tsa1 dimer formation (Fig. 3 F and G).
Since steady-state intracellular peroxide levels are similar between WT, sod1Δ, and zwf1Δ cells (SI Appendix, Fig. S6 E and F), we ascribe the increased Tsa1 disulfide oxidation as arising from diminished NADPH levels. Surprisingly, for reasons that are unclear at this time, tsa1Δ and sod1Δ cells also exhibited increased Tsa1 sulfonylation, despite these cells having similar steady-state H2O2 as WT cells (SI Appendix, Fig. S7 A–E). Titration of a copper chelator, bathocuproinedisulfonic acid (BCS), resulted in a dose-dependent decrease in Sod1 activity, without affecting expression, and consequent increase in the formation of disulfide-dimerized Tsa1 (Fig. 3H and SI Appendix, Fig. S7 F and G). Overall, these results are consistent with Sod1 mediated oxidative inactivation of GAPDH promoting NADPH production via the oxPPP, thereby allowing for the regeneration of peroxide scavengers like Tsa1.
In order to determine if the oxidative inactivation of GAPDH by Sod1 provides a physiological benefit for cells, we measured the aerobic growth and oxidative stress resistance of cells expressing low and high levels of GAPDH. Although WT and tdh3Δ cells have similar aerobic and anaerobic growth rates in 2% GLU (Fig. 3I), when cultured in 2% GAL, a fermentable carbon source that promotes respiration in yeast, tdh3Δ cells have a marked enhancement in growth rate compared to WT cells (Fig. 3J). Moreover, tdh3Δ cells exhibit greater resistance to peroxide (SI Appendix, Fig. S7H) and diamide induced oxidative stress (Fig. 3K and SI Appendix, Fig. S7I). Altogether, the enhanced aerobic fitness and oxidative stress resistance of cells depleted of GAPDH is consistent with a beneficial role for the Sod1-mediated inactivation of GAPDH.
Sod1 Regulates GAPDH Oxidation in Human Cells.
Since both Sod1 and GAPDH are highly conserved, from yeast to humans, we sought to establish if Sod1 regulates GAPDH oxidation in human embryonic kidney HEK293 cells. Human GAPDH has an additional peroxide reactive Cys, which would result in three PEGylated proteoforms, corresponding to triple-, double-, and single-labeled GAPDH in immunoblots (SI Appendix, Fig. S8A). As expected, cells treated with H2O2 exhibited less mPEG-mal labeling, indicating a larger fraction of oxidized GAPDH compared to nontreated cells (SI Appendix, Fig. S8 B and C). To determine if Sod1 promoted GAPDH oxidation, we depleted Sod1 in HEK293 cells using small-interfering RNA against Sod1 (siSOD1) or scrambled control RNAi (siCTRL). Across various trials, we consistently observed a depletion of ∼60 to 80% of Sod1 (SI Appendix, Fig. S8 D and E) and a corresponding decrease in GAPDH oxidation (SI Appendix, Fig. S8 F–H). Sod1 increases GAPDH by ∼7 to 16%, which is comparable to the contribution observed from exogenous peroxide treatment (SI Appendix, Fig. S8B). Moreover, as with yeast, we find that Sod1 expression levels across multiple trials positively correlate with GAPDH oxidation (SI Appendix, Fig. S8I). Notably, the y-intercept of the linear regression (r2 = 0.76; P = 0.002) is close to 0, indicating that in the complete absence of Sod1, GAPDH oxidation is expected to be ∼0% (SI Appendix, Fig. S8I). Furthermore, cell lines that overexpress Sod1, such as the breast cancer cell line MCF7 (56), exhibit a nearly threefold increase in GAPDH oxidation compared to HEK293 cells (SI Appendix, Fig. S8 J and K). Altogether, these results indicate that Sod1-mediated oxidation of GAPDH is conserved in humans.
Redox Proteomics Identifies Additional Putative Targets of Sod1 Redox Regulation.
The high abundance and broad cellular distribution of Sod1 suggests that it may function as a redox regulator of a broad variety of substrates in addition to GAPDH. To test this, we conducted a high-powered quantitative redox proteomics screen of WT and sod1Δ yeast to identify Sod1-dependent redox substrates. We used a combined stable isotope labeling with amino acids in cell culture (SILAC)-tandem mass tags (TMT) approach, whereby sod1Δ-dependent changes in protein abundance are quantified through SILAC and changes in reversible cysteine oxidation are quantified through cysteine-reactive iodoacetyl TMT (iodo-TMT) before and after reduction with DTT (Fig. 4A) (57). In this case, the same peptide from WT and sod1Δ cells is distinguished by a mass shift, allowing quantitative comparison of the peptide’s abundance in each strain [protein expression difference = log2(sod1Δlight/WTheavy)]. Subsequent fragmentation/sequencing of each peptide reveals the peptide/protein identity and, if cysteine is present, releases TMT reporter ions enabling quantification of the cysteine oxidation percentage in each strain (oxidation difference = sod1Δ%Ox − WT%Ox) (Materials and Methods). Consequently, the study reveals a broad range of proteins that undergo significant changes in abundance, cysteine oxidation, or both (Fig. 4B and SI Appendix, Fig. S9A).
Focusing first on the effects of SOD1 deletion on proteome-wide protein abundance, we independently analyzed the SILAC mass spectrometry (MS) data. A total of 4,409 proteins were confidently detected and quantified, and ∼9% of these (373) exhibited significant differences in abundance between the two strains (Fig. 4C and Dataset S1). Of these, 114 (30.6%) exhibit a significant decrease in protein abundance in sod1Δ cells compared to 259 (69.4%) proteins that exhibit a significant increase in abundance. The changes in protein expression reflect known metabolic defects associated with loss of SOD1, including alleviation of glucose repression, increased mitochondrial mass, induction of the iron starvation and antioxidant responses, and diminished plasma membrane casein kinase expression (Fig. 4D and SI Appendix, Fig. S9B). Gene ontology (GO) enrichment analysis of proteins undergoing a statistically significant change in abundance revealed a set of outlier ontologies (i.e., GO terms falling outside the 90th percentile in the distribution of fold-enrichment) that could be clustered into discrete bins. Notably, we observed the greatest enrichment for ontologies linked to glycolysis and pentose phosphate pathways. This included chitin synthesis (median fold-enrichment [FE]: 29.6), which feeds directly from D-fructose-6-phosphate, a precursor to the GAPDH substrate D-glyceraldehyde-3-phosphate and a key metabolite in the PPP; cellular polysaccharide catabolism (starch and sucrose catabolism) (FE: 21.1); amino acid biosynthesis (FE: 20.6), which relies on several glycolytic intermediates; as well as purine biosynthesis/metabolism that feeds on D-ribose-5-phosphate from the PPP (FE: 14.1) (Fig. 4E and Dataset S1). We observed lower but still significant GO enrichment of proteins involved in redox processes and functions (FE: 14.1), including ontologies encompassing oxidoreductase activity, NADH oxidation, and ROS metabolic processes; and enrichment of proteins associated with ATP synthase/proton pumping (FE: 17.6).
Next, we focused on the effects of SOD1 deletion on proteome-wide cysteine oxidation as revealed by iodo-TMT MS data. A total of 2,077 cysteine residues were confidently detected ranging in oxidation level from 1 to 100% and a median level of 28.9% across all sites measured. Cross-referencing all sites against protein functional residues revealed 54 of these sites associated with metal binding (38 sites, 74%), active sites (11 sites, 20%), and protein or DNA binding (5 sites, 9%) and cysteine residues involved in these functions also exhibited statistically higher overall oxidation levels in both WT and sod1Δ (44.6%, P < 0.0001) (SI Appendix, Fig. S9C and Dataset S1). Mean global oxidation levels were nearly identical between both strains (32.8% vs. 32.9%) (SI Appendix, Fig. S9A).
Approximately 5% of the quantified cysteine sites (99 sites, 96 proteins) undergo significant changes in oxidation between WT and sod1Δ cells, and of these, we found that 65% exhibit reduced oxidation levels in the absence of SOD1 (Fig. 4F and Dataset S1). GAPDH peptides, and C150/C154 specifically, were readily detected but failed to undergo iodo-TMT reporter release under any of the three collision energies used for peptide fragmentation, preventing quantitation of SOD1-dependent oxidation in these experiments. However, we were able to detect DTT-irreversible GAPDH thiol oxidation products, including sulfinic acid (SO2) and dehydroalanine, and found they were elevated up to 1.6-fold in WT cells relative to sod1Δ cells (Dataset S1). These results support the conclusion that Sod1 contributes to both reversible and irreversible GAPDH oxidation as detected by targeted mPEG-mal/GAPDH immunoblotting experiments and MS, respectively.
On average, cysteine residues in sod1Δ cells appear to be less oxidized compared to WT cells, consistent with a role for Sod1 in providing a source of peroxides for thiol oxidation. On the other hand, cysteine residues that are more oxidized in sod1Δ cells compared to WT may be due to oxidative stress, in part mediated by the decrease in NADPH production due to diminished Sod1-mediated GAPDH oxidation and the subsequent reduction in flux through the oxPPP.
Over half of the proteins harboring the 99 significantly changing oxidation sites are classified as enzymes, with the next most populous class represented at less than 10% in comparison (Fig. 4G and Dataset S1). Most of these proteins are also associated with metabolism, synthesis of secondary metabolites, and biosynthesis of amino acids. Clustering of ontologies that were statistical outliers for FE supported these findings and produced three distinct ontology bins. Ontologies associated with amino acid biosynthesis were predominantly enriched (FE: 133.9) and to a far greater extent than what we observed for Sod1-dependent protein abundance changes (Fig. 4H and Dataset S1). Moreover, many of the proteins within this bin exhibit reduced oxidation in the absence of SOD1, and map to multiple different amino acid biosynthesis/metabolism pathways including those for lysine (Lys20/21), arginine (Car1), threonine (Hom2), and s-adenosyl methionine (Sam1/2), to name a few (SI Appendix, Fig. S9 D and E). Ontologies encompassing redox processes and functions were also highly enriched (FE: 90.4), and include proteins involved in catalysis of redox reactions in which a CH-NH2 group acts as a hydrogen or electron donor and reduces NAD+ or NADP+. Finally, we also found significant enrichment of proteins involved in Ras and Rho signal transduction (median FE: 36.5), which hint at a unique and unrealized connection between SOD1, redox homeostasis, actin mobilization, and cell motility.
Considering the role of Sod1 as a direct regulator of GAPDH through C150 oxidation, we asked whether our proteomic data might contain evidence of a similar relationship between Sod1 and other proteins, wherein: 1) the protein is specifically affected at one out of multiple possible cysteine residues; and 2) oxidation of this cysteine is significantly reduced in the absence of SOD1. To create this filter, we binned each of the 99 sites based on whether the change in oxidation was a statistical outlier (>90th percentile) from other cysteine residues in the protein, revealing 18 proteins (Fig. 4 I and J). In seven of these cases, the change in oxidation was also found to be an outlier with respect to other oxidized cysteine residues found in other proteins within the associated bioprocess for the given protein (Fig. 4I, blue halos, SI Appendix, Fig. S10A, and Dataset S1). Deconvolution of their site-specific changes in oxidation revealed a range of responses reflecting the sensitivity of specific cysteine residues to the presence or absence of SOD1 (Fig. 4J). Some of these coincided with proteins enriched previously through ontology, most notably Sfa1 and Gfa1 involved in pyruvate metabolism and conversion of D-fructose-6-phosphate, respectively. Sfa1-C12 and Gfa1-C562, but not other detected cysteine residues in either protein, undergo a significant decrease in oxidation in the absence of SOD1 (Dataset S1). We also observed that in one case, Nbp35 (required for maturation of extramitochondrial Fe-S proteins), the site of dynamic oxidation is a functional metal binding site in the protein (Dataset S1). Finally, other proteins not revealed by previous analyses were also evident, most notably Uba1, which is the sole enzyme necessary for activating ubiquitin in yeast. Uba1-C447, which is immediately adjacent to the ATP binding site for the protein, undergoes a large drop in oxidation in the absence of SOD1, going from a maximum ∼89% in WT cells down to ∼28% in sod1Δ cells (SI Appendix, Fig. S10B and Dataset S1). In comparison, C600, which is the site of ubiquitin attachment, undergoes very little change in oxidation upon deletion of SOD1. Collectively, these data provide an unparalleled look into the dynamic oxidation of specific cysteine thiols as well as protein abundance changes in response to the loss of SOD1.
SODs are unusual “antioxidants” in that they catalyze the production of one ROS, H2O2, as a byproduct of detoxifying another ROS, O2•−. Herein, we sought to understand if there were physiological roles for Sod1-derived H2O2 in redox regulating antioxidant defenses and the thiol proteome. We identified a redox circuit in which Sod1 senses the availability of O2 via metabolically produced O2•− radicals and catalyzes the production of H2O2 that oxidatively inactivates GAPDH. GAPDH inactivation in turn promotes flux through the oxPPP to generate the NADPH required for aerobic metabolism and antioxidant defenses (30, 31). Moreover, using MS-based redox proteomics approaches, we identified a larger network of proteins whose redox state, like GAPDH, is sensitive to Sod1 levels. Altogether, our results highlight a mechanism for the antioxidant activity of Sod1—namely that Sod1-derived H2O2 can stimulate NADPH production—and place Sod1 as a master regulator of the cellular redox landscape through two potential mechanisms: either by providing a direct source of thiol oxidizing H2O2 or by altering the NADPH/NADP+ redox balance through the Sod1/GAPDH signaling axis (Fig. 5).
We propose a proximity-based model in which Sod1 provides a highly localized and privileged pool of peroxide capable of physiological GAPDH oxidation and control of glycolytic flux. Since GAPDH and Sod1 are among the most abundant soluble proteins in cells, present at levels >100 μM (58), rapid peroxide diffusion due to short distances between Sod1 and GAPDH, or peroxide channeling due to transient interactions between these proteins may allow for the Sod1-mediated aerobic oxidation of GAPDH. This proximity-model is supported by the observations that Sod1 and GAPDH interact (Fig. 1G and SI Appendix, Fig. S2I), WT and sod1Δ cells have similar levels of intracellular peroxide (SI Appendix, Fig. S11A), and exposure to concentrations of exogenous H2O2 up to 0.25 mM, which is sufficient to substantially increase intracellular peroxides (SI Appendix, Fig. S11A), does not result in GAPDH oxidation (SI Appendix, Fig. S11 B–D). Altogether, these findings are consistent with a model in which only localized production of H2O2 by Sod1 is capable of physiological GAPDH oxidation. Sod1-mediated redox signaling complements other paradigms of peroxide signaling, such as the transfer of oxidizing equivalents through thiol-disulfide exchange (59) or the “flood-gate” effect, which posits that a burst of H2O2 inactivates antioxidants like peroxiredoxins so that sufficient peroxide may diffuse far enough to transmit redox signals (60).
How physiologically significant is the Sod1-mediated oxidative inactivation of GAPDH? Prior metabolic flux analysis in yeast found that the inactivation of up to ∼20% of the total GAPDH pool does not result in oxPPP rerouting or increase in NADPH (31). In contrast, inactivation of greater than 20% of the GAPDH pool results in substantially increased carbohydrate flux through the oxPPP to elevate NADPH in a manner that has a power dependence on GAPDH oxidation (31). We find herein that in the absence of SOD1, GAPDH oxidation is at or below this ∼20% threshold. When SOD1 is expressed, the average and median GAPDH oxidation exceeds 20%, leading to elevated NADPH. Thus, Sod1-derived H2O2 ensures that there is sufficient steady-state oxidized GAPDH to enable oxPPP-dependent NADPH production for optimal growth in air. However, it is also possible that there are other redox sensitive nodes that contribute to Sod1-dependent changes in [NADPH] independent of GAPDH.
The aerobic oxidation of GAPDH is rate-limited by Sod1 (Fig. 2 D and E), suggesting that the oxidative inactivation of GAPDH and glycolytic flux may be dynamically regulated by Sod1 expression, maturation, which itself is O2-dependent, its interactions with GAPDH, or posttranslational modifications that regulate Sod1 localization and activity (61⇓⇓–64). With respect to Sod1 posttranslational modifications, it is tempting to speculate that the O2-sensing redox circuit we identified herein may be integrally linked to the nutrient sensing target of rapamycin (TOR) pathway. TOR-dependent phosphorylation at Ser38 in yeast (Thr39 in humans) suppresses Sod1 activity in response to nutrient abundance to promote cell growth, whereas starvation promotes Sod1 activity (64). TOR/Sod1 signaling may be coupled to the redox regulation of GAPDH and glycolytic flux, thereby providing a means to integrate oxygen and nutrient availability to control metabolism. With respect to Sod1 expression, since many cancer cells overexpress Sod1, it is conceivable that the Sod1/GAPDH signaling axis promotes cancer cell survival and drug resistance due to its potentiation of NAPDH production via the PPP (65). Thus, our work also sheds new light on the benefits of anti-Sod1 therapeutic interventions (64, 66⇓⇓–69). Indeed, we find that the breast cancer MCF-7 cell line exhibits greater levels of GAPDH oxidation relative to HEK293 cells, correlating with its higher levels of Sod1 expression. Like Sod1, MnSOD (Sod2), which is highly inducible, is a biomarker for cancer progression and is involved in the metastasis of certain cancers (70⇓⇓–73). As Sod2 may have its own distinct set of redox targets that control cell metabolism and physiology, therapeutic interventions that target Sod1, Sod2, or both potentially offers a means to titrate mitochondrial and extramitochondrial peroxide-based signaling to combat disease.
The integration of O2 availability through the Sod1/GAPDH redox signaling axis is dependent on key sources of O2•−, including the yeast NADPH oxidase Yno1 and mitochondrial respiration (Fig. 2). Given that Sod1 and GAPDH are primarily cytosolic enzymes, it is not surprising that we found Yno1, which is localized at the endoplasmic reticulum (ER) membrane and produces O2•− on the cytosolic side, is required for transducing O2 availability to the O2•− signal required for Sod1/GAPDH signaling (51). Indeed, Yno1 was previously found to regulate Sod1-mediated redox regulation of yeast casein kinase Yck1/2 signaling, which also occurs in the cytosol in proximity to the plasma membrane. However, it was rather surprising to find that mitochondrial respiration, as assessed in a respiratory incompetent yeast mutant lacking mitochondrial DNA, could also contribute O2•− for Sod1-mediated GAPDH oxidation in the cytosol. Complex I (missing in S. cerevisiae) and III are the primary sources of electron leakage from the electron transport chain and produce O2•− in the mitochondrial IMS (74). Yet, IMS-targeted Sod1 does not mediate GAPDH oxidation (Fig. 1H). Thus, IMS O2•−, which is charged and membrane impermeable, must exit the mitochondria for Sod1/GAPDH signaling via voltage-dependent anion channels (75) or as membrane-permeable neutral hydroperoxyl radicals (HO2•). Our results highlight how mitochondrial respiration may be an important sensor for O2, providing a source of O2•− that can act as a retrograde signal to control extramitochondrial metabolism for adaptation to increasing pO2. The present study complements prior work that found mitochondrial respiration is a key source of ROS required for adaption to hypoxia via hypoxia inducible factor signaling (76).
The finding that Sod1 can redox regulate GAPDH prompted us to determine if other redox targets of Sod1 exist. Indeed, MS-based redox proteomics identified a number of proteins that, like GAPDH, are more oxidized in the presence of Sod1. Interestingly, enzymes associated with pathways orthogonal to glycolysis and the PPP, namely involved in amino acid biosynthesis, were prominently enriched over other groups (Fig. 4H and SI Appendix, Fig. S9 D and E), and suggest that Sod1-dependent growth control may also be modulated through additional targets within these pathways that may have evolved Sod1-sensitive cysteine oxidation sites. On the other hand, all but one of these proteins (Nbp35) are oxidized at Cys residues not associated with a direct role in catalysis or function, raising the possibility that oxidation of such sites may function allosterically. For example, the E1 ubiquitin activating enzyme Uba1 is differentially oxidized at C447 (SI Appendix, Fig. S10B and Dataset S1), which is immediately adjacent to the ATP binding site, and the cysteine desulfurase Nfs1 is differentially oxidized at C199, which is adjacent to the pyridoxal phosphate cofactor. Future work will probe the functional consequences of Sod1-dependent protein oxidation. We also identified proteins that are more oxidized in sod1Δ cells, which we interpret as arising due to oxidative stress associated with the loss of GAPDH/Sod1 signaling and the concomitant decrease in NADPH levels and/or O2•− toxicity. It is worth noting that although past redox proteomics studies revealed much about potential targets of redox regulation, they typically involved treatment with exogenous oxidants mimicking pathological stress levels outside the physiological range (24, 77⇓⇓–80). Our study is notable in that we have identified redox targets of a physiological endogenous source of H2O2.
To what degree does Sod1 regulate thiol oxidation in animal models? A recent elegant redox proteomics study from Couchani and coworkers (81), termed “Oximouse,” describes tissue specific protein thiol oxidation in a mouse model. Since active Sod1 requires the formation of an intramolecular disulfide bond between Cys-147 and Cys-56, we plotted percent Sod1-Cys147 oxidation as a proxy for mature active Sod1 against overall average Cys oxidation in each tissue and found a statistically significant linear correlation (r2 = 53; P = 0.04) (SI Appendix, Fig. S12A). To determine if this correlation between active Sod1 and overall thiol oxidation was unique to Sod1, we also plotted tissue Cys oxidation against the oxidation of active site Cys residues in other peroxide metabolizing enzymes, including peroxiredoxins 1 and 5 (PRDX1, PRDX5) or glutaredoxin 3 (GLRX3). Unlike Sod1, the activity of these other enzymes, as assessed by their active site oxidation, do not significantly correlate with overall Cys oxidation (SI Appendix, Fig. S12 B–F). Moreover, we also found that the oxidation of other enzymes known to be redox regulated by Sod1, including tyrosine phosphatases PTP1N (PTP1B in humans) and PTPN11 (82, 83) and a GTPase involved in vesicular protein transport from the ER to the Golgi, Rab1α (84) (SI Appendix, Fig. S12 G–I), exhibit a linear correlation between active Sod1 and their active site oxidation. Notably, there is no correlation with PRDX1 Cys oxidation, which is typically associated with the transmission of oxidizing equivalents to regulate redox signaling (59) (SI Appendix, Fig. S12 J–L). Together, the data we extracted from the Oximouse study supports our principal findings from yeast that Sod1-derived peroxide regulates proteome-wide thiol oxidation.
Our studies have also finally explained a mystery regarding the role of Sod1 in antioxidant defense. Prior to the present work, the primary role of Sod1 in antioxidant defense was thought to be O2•− scavenging. However, we previously found that the vast majority of Sod1 (>99%) was dispensable for protection against cell wide markers of O2•− toxicity (15). We surmised that exceedingly low amounts of Sod1 were sufficient to protect against O2•− because the targets of O2•− toxicity are limited in scope, primarily Fe-S cofactor-containing proteins. Indeed, all O2•−-related toxicity phenotypes arise from diminished activity of Fe-S enzymes and iron toxicity due to iron release from damaged clusters. Our work indicates that a broader role for Sod1 in oxidant defense is to promote the production of NADPH via the GAPDH/Sod1 signaling axis. As a major cellular reductant for numerous antioxidant systems, NADPH offers more expansive protection against redox stress than just defending against O2•−. Since GAPDH and Sod1 are among the most ancient and highly conserved enzymes in aerobic life, we propose that this reported function of Sod1 to produce NADPH via GAPDH oxidation was a key requirement for O2 sensing and integration for adaptation to life in air.
Materials and Methods
All cell lines, culture conditions, reagents, methods, including MS-based redox proteomics, and statistical analyses are described in SI Appendix, SI Materials and Methods. Yeast strains used in this study were in the BY4741 genetic background and HEK293 and MCF-7 cells were obtained from American Type Culture Collection (ATCC). Plasmids and primers (SI Appendix, Table S1) are described in SI Appendix, SI Materials and Methods. The synthetic gene for HyPer7 (85) was purchased from Synbio Technologies and was codon-optimized for expression in S. cerevisiae (SI Appendix, Table S2). Published methods for immunoblotting and IP (15, 16), mPEG-mal labeling (48, 86), GAPDH activity (87), SOD activity (15, 16), Pma1 activity (16), DHE (15), and HyPer7-mediated (85) superoxide and peroxide detection, respectively, pH measurements using pHluorin (88, 89), NADPH detection (90), Sod1 silencing (40), MS-based redox proteomics (57), and cell growth were employed and are described in detail in SI Appendix, SI Materials and Methods.
All study data are included in the manuscript and/or supporting information. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium (91) via the PRIDE partner repository (92) with the dataset identifier PXD030372.
We thank Ms. Kathrin Ulrich for assistance with mPEG-mal labeling protocols, Prof. Valeria Culotta for the Sod1 antibody, Prof. Tobias Dick for the tdh1Δ tdh2Δ tdh3Δ strains, Prof. Dennis Thiele for the mitochondrial targeted Sod1 allele, Prof. Michael Breitenbach for the Yno1 plasmids, and Prof. David Eide for the HA-Tsa1 constructs. We acknowledge support from the NIH grant GM118744 (to A.R.R. and M.P.T.) and the Blanchard Professorship (A.R.R.).
- Accepted November 14, 2021.
Author contributions: C.M.-A., H.K., A.E.T., A.P.J., M.P.T., and A.R.R. designed research; C.M.-A., H.K., A.E.T., and A.P.J. performed research; C.M.-A. contributed new reagents/analytic tools; C.M.-A., H.K., A.E.T., M.P.T., and A.R.R. analyzed data; and C.M.-A., M.P.T., and A.R.R. wrote the paper.
The authors declare no competing interest.
This article is a PNAS Direct Submission.
This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.2023328119/-/DCSupplemental.
- Copyright © 2021 the Author(s). Published by PNAS.
Read more here: Source link