Superoxide Triggers an Acid Burst in Saccharomyces cerevisiae to Condition the Environment of Glucose-starved Cells*

Background: During prolonged periods without glucose yeast cells acidify the extracellular environment. Results: This acid burst is derived from a mitochondrial aldehyde dehydrogenase and a cascade involving superoxide damage to the TCA cycle. Conclusion: The acid burst can condition the medium to improve cell growth. Significance: Damage from mitochondrial superoxide can be advantageous to cells during carbon starvation stress. Although yeast cells grown in abundant glucose tend to acidify their extracellular environment, they raise the pH of the environment when starved for glucose or when grown strictly with non-fermentable carbon sources. Following prolonged periods in this alkaline phase, Saccharomyces cerevisiae cells will switch to producing acid. The mechanisms and rationale for this “acid burst” were unknown. Herein we provide strong evidence for the role of mitochondrial superoxide in initiating the acid burst. Yeast mutants lacking the mitochondrial matrix superoxide dismutase (SOD2) enzyme, but not the cytosolic Cu,Zn-SOD1 enzyme, exhibited marked acceleration in production of acid on non-fermentable carbon sources. Acid production is also dramatically enhanced by the superoxide-producing agent, paraquat. Conversely, the acid burst is eliminated by boosting cellular levels of Mn-antioxidant mimics of SOD. We demonstrate that the acid burst is dependent on the mitochondrial aldehyde dehydrogenase Ald4p. Our data are consistent with a model in which mitochondrial superoxide damage to Fe-S enzymes in the tricarboxylic acid (TCA) cycle leads to acetate buildup by Ald4p. The resultant expulsion of acetate into the extracellular environment can provide a new carbon source to glucose-starved cells and enhance growth of yeast. By triggering production of organic acids, mitochondrial superoxide has the potential to promote cell population growth under nutrient depravation stress.

Cells exist and thrive in environments that are both supportive of life and potentially hostile. The environment is in a constant state of change and the onset of stress conditions, such as nutrient depravation and exposure to toxins, require appropriate adaptation strategies. Cells have evolved a variety of mechanisms to appropriately sense and respond to variations in the environment through induction of stress pathways and adjustments in metabolism. Yet, in some cases, cells respond by modifying the microenvironment itself to enhance its suitability for their fitness and survival. The pathogen Plasmodium falciparum, for example, is known to secrete a variety of proteins that enhance the permeability of the red blood cell membrane to increase the availability of nutrients (1). Cancer cells are particularly adept at altering microenvironments. During the process of invasion, these cells secrete a variety of metalloproteases (2) and also acidify their surrounding environment to inhibit the toxic activity of immune cells and chemotherapies, and promote migration to more nutrient-rich environments (3,4). To make nutrients more accessible, the bakers' yeast Saccharomyces cerevisiae secretes phospholipases (5), a variety of enzymes that break down sugars (6), and phosphatases (7).
To S. cerevisiae, the pH of the environment is critical for regulating growth and survival (8 -10). Cells growing on the preferred carbon source glucose acidify the surrounding media by activating the proton ATPase, Pma1p (11), and secreting organic acids (12,13). This acidic environment (ϷpH 4 -5) helps to maintain the electrochemical gradient across the plasma membrane and drive uptake of many essential nutrients (14). On the other hand, when yeast cells grow on the less preferred non-fermentable carbon sources or are starved for glucose, they alkalize their surrounding environment, which slows growth but enhances survival under stress (8,15,16). This period of alkalization correlates with increases in volatile ammonia that serves as a signaling molecule to synchronize colony growth (15)(16)(17)(18). Interestingly, following prolonged growth on the non-fermentable carbon source glycerol, yeast cells can switch from this alkalization phase to producing acid (17,19,20). The mechanism or rationale for this sudden "acid burst" has been unknown. In previous studies completed by Palkova's group (17,19,20), the production of ammonia during the alkalization phase was reportedly defective in S. cerevisiae cells lacking the mitochondrial matrix manganese-containing superoxide dismutase, SOD2 (or yeast Sod2p). Sod2p was suggested to affect ammonia signaling, but through an unknown pathway (19,21). Superoxide dismutases (SODs) 3 represent a family of metalloenzymes responsible for detoxifying superoxide radicals generated as a by-product of aerobic metabolism. Most eukaryotes, including S. cerevisiae, contain both a Mn-SOD2 in the mitochondrial matrix as well as, a Cu,Zn-containing SOD1 that is largely cytosolic but also found in the mitochondrial intermembrane space (22,23). The presence of SODs on both sides of the mitochondrial inner membrane collectively protect against superoxide generated by the electron transport chain. The importance of SOD2 in guarding against mitochondrial oxidative damage has been underscored in various model organisms for SOD2 deficiency. SOD2Ϫ/SOD2Ϫ mice are neonatal lethal and early death in this case is thought to arise from severe oxidative damage to the mitochondrial respiratory chain (24,25). Loss of SOD2 also induces early mortality in Drosophila adults, associated with damage to the mitochondrial respiratory chain and TCA (tricarboxylic acid) cycle enzymes (26). In the bakers' yeast S. cerevisiae, cells lacking Sod2p are viable but are highly sensitive to hyperoxia and redox cycling agents (27)(28)(29), defects ascribed to increased superoxide damage to mitochondrial components. However, the means by which loss of yeast mitochondrial matrix Sod2p can control extracellular pH as described earlier (19) is not clear, particularly because the superoxide substrate for Sod2p should not cross the mitochondrial membranes (30).
Here, we describe a new connection between mitochondrial superoxide and the control of environmental pH. Specifically we find that the acid burst produced by cells during long-term respiratory conditions can be induced by mitochondrial superoxide. With either mutation in yeast SOD2 or by treatment with the redox cycler paraquat, the acid burst is accelerated in yeast cells starved for glucose. Moreover, the acid burst is eliminated through Mn-antioxidants that act as SOD mimics and remove intracellular superoxide. We provide evidence that superoxide damage to Fe-S enzymes in the TCA cycle results in massive production of acetate involving the mitochondrial aldehyde dehydrogenase Ald4p. The concomitant acetate burst during nutrient starvation provides a new carbon source to enhance cell growth during long-term nutrient depravation.
Culture Conditions and pH Measurements-To examine the effects of yeast colony growth on extracellular pH, solid growth medium containing 3% glycerol (or where indicated, 2% glucose), 1% yeast extract, 0.01% bromocresol purple (BCP; Sigma, B5880), and 2% bacto-agar was prepared precisely as described by Palkova and co-workers (15), except supplemental CaCl 2 was generally omitted. pH was typically adjusted to 5.75 with HCl prior to autoclaving but could range from 4.35 (enhance detection of media alkalization) to 6.5 (enhance detection of media acidification) without altering the cell alkaline and acid phases. When needed, the specified concentrations of paraquat (MP Biomedicals) or MnCl 2 were added immediately before pouring. In all cases, extracellular pH was monitored using "giant colony" growth as prescribed by Palkova et al. (15) where 2 ϫ 10 5 cells in 10 l were spotted onto plates and incubated at 30°C. Images were taken with a Sony Cybershot DSC-F828 on the days indicated. Cell viability measurements were obtained by removing cells at the designated times and measuring colony-forming units on YPD (1% yeast extract, 2% peptone, 2% glucose) averaged over 3 giant colonies as described (20). Results were normalized to WT at day 4. WT and sod2⌬ strains retain full viability through day 8 and at day 11 exhibited 92 (S.D. Ϯ 8) and 89% (S.D. Ϯ 11) viability, respectively.
For monitoring effects of yeast growth under low glucose conditions, cells were inoculated at an initial A 600 ϭ 0.05 and grown shaking at 220 rpm, at 30°C for 24 h in a liquid media consisting of 1% yeast extract and 0.2% glucose, adjusted to pH 5.5 with HCl. To determine the extracellular pH of these cultures, a colorimetric assay based on the pH-sensitive dyes BCP and bromocresol green (BCG; Sigma, 114359) was developed. Briefly, cells were removed by centrifugation at 17,000 ϫ g for 1 min and the media was collected (90 l each) in triplicate and applied to a 96-well plate. Water (blank), 0.1% BCP, or 0.1% BCG (10 l) were added and absorbance was measured at A 430 , A 589 , A 514 , and A 616 in a Synergy HT Multi-Mode Microplate Reader (BioTek). pH was calculated by subtracting the absorbance of the blank at each wavelength and using the following empirically derived equations.
For Plates were imaged on an HP Scanjet 8200 scanner. For conditioned media tests, cultures from low glucose conditions as described earlier were filter-sterilized (Millipore, Millex-GV, 0.22 m) to remove cells and the conditioned medium collected was then used for inoculation of WT cells at A 600 ϭ 0.1 followed by aerobic growth for the indicated time points.
Metabolite Measurements-To measure glucose, acetate, and succinate concentrations in media, conditioned medium was collected from cells grown in low glucose as described earlier and cells were removed by centrifugation. Measurements were performed with a Synergy HT Multi-Mode Microplate Reader using the QuantiChrom TM Glucose Assay Kit (Bio-Assay Systems), the Acetic Acid kit (R-Biopharm), and the Succinic Acid kit (R-Biopharm), respectively, according to the manufacturers' specifications. To adapt the Acetic Acid and Succinic Acid kits to a plate reader format, one-tenth the recommended volumes were used and each reading was path length corrected using the standard correction of the plate reader: (A 977 Ϫ A 900 )/0.18. Intracellular acetate measurements were performed in the same manner as those for media, except cells were first lysed by glass bead homogenization in 20 mM potassium phosphate, 1.2 M sorbitol, 1.0 mM PMSF, pH 7.4, and clarified by centrifugation at 17,000 ϫ g for 10 min. After determining the total protein by Bradford assay, lysates were heated 15 min at 85°C and the supernatant was collected for analysis after a second centrifugation. Intracellular acetate concentrations were normalized to total protein.
To measure intracellular metabolites of the TCA cycle, cells grown as described earlier in low glucose media were harvested at A 600 ϭ 1-1.2 and washed in PBS. Cells were resuspended in PBS at a ratio of 5 l of PBS/A 600 cells and lysed by glass bead homogenization as described (36). Lysates were clarified by centrifugation at 17,000 ϫ g for 10 min and total protein was determined by Bradford assay. Duplicate samples were analyzed by stable-isotope dilution GC-MS by the Kennedy Krieger Institute Biochemical Genetics Laboratory.
Enzymatic Assays-For measurements of aconitase and succinate dehydrogenase (SDH) activity, cells were collected from low glucose liquid media (10 ml) after 24 h growth by centrifugation or scraped from glycerol plates containing BCP (2-9 ϫ 10 8 cells collected from strains plated as in Fig. 1A) on the days indicated and stored at Ϫ20°C. Cells were then resuspended in lysis buffer consisting of 20 mM potassium phosphate, 1.2 M sorbitol, 1.0 mM PMSF, 0.1% Triton X-100, pH 7.4, at a ratio of 10 l of buffer/2 ϫ 10 7 cells and lysed by glass bead homogenization. Protein concentrations were determined by Bradford assay and enzymatic assay measurements were performed with 10 -50 g of lysate protein per reaction. Aconitase activity was measured as described (36) in 250 l of total reaction volume and calculated as the nanomoles of cis-aconitate consumed per mg of protein per min using the Beer-Lambert law and an extinction coefficient of 4.88 mM Ϫ1 cm Ϫ1 . SDH activity was measured with the procedure of Ackrell et al. (37), adapted to a 200-l 96-well plate format. Briefly, SDH in whole cell lysates was activated at 30°C for 10 min in the reaction mixture without electron acceptors/dyes (178 l volume; final concentration after addition of dyes ϭ 20 mM potassium phosphate, 20 mM succinate or malonate, 1 mM KCN, pH 7.4). After 10 min, prewarmed dyes (22 l volume; final concentration in complete reaction mix ϭ 1 mM phenazine methosulfate (Sigma), 0.05 mM 2,6-dichlorindophenol (Sigma)) were added to initiate the reaction. SDH activity was calculated as the malonate-sensitive nanomoles of 2,6-dichlorindophenol reduced per mg of protein per minute using an extinction coefficient of 21 mM Ϫ1 cm Ϫ1 .
For Pma1p ATPase activity, cells were grown in 1% yeast extract with 2 or 0.2% glucose, pH 5.5 to an A 600 ϭ 0.75-0.9. Plasma membranes were isolated (38) and Pma1p ATPase activity was measured by assaying orthovanadate-sensitive release of inorganic phosphate precisely as described by Monk et al. (39) and normalized to the activity of WT cells grown in 0.2% glucose.

RESULTS
A Role for Superoxide and SOD2 in Controlling the Acid Burst of Respiring Cells-Extracellular pH in yeast growth medium is readily monitored by pH indicators such as BCP and BCG. When grown in high glucose, fermenting yeast cells rapidly acidify the medium (Fig. 1A top), but when grown on nonfermentable carbon sources, the respiring yeast cells initially elevate the extracellular pH (Fig. 1A, middle and bottom). As prescribed by Č áp et al. (19), this alkalization is observed on glycerol medium supplemented with 30 mM calcium (Fig. 1A, middle), yet is also seen without added calcium (Fig. 1A, bottom); hence all subsequent studies were conducted in the absence of calcium supplements. Alkalization of the growth medium is not unique to non-fermentable carbon sources, but is also observed in liquid cultures with low glucose, conditions where cells switch from fermentation to respiration (8) (Fig.  1B). Cells grown in 0.2% glucose rapidly deplete their glucose (Fig. 1C) and the pH of the extracellular medium rises from 5.5 to Ն7.0 within 24 h (Fig. 1B). Following long-term growth on glycerol containing medium, S. cerevisiae cells can enter an acid phase as previously described, which we refer to herein as the acid burst ( Fig. 2) (17,19,20). We observe that the initiation of the acid burst is substantially accelerated in strains lacking Mn-SOD2, the mitochondrial matrix manganese-containing SOD. The sod2⌬ null cells show no defect in the initial alkalization of the medium (Fig. 2, Fig. 3A); however, their onset of entry into the acid phase is accelerated compared with WT strains (Fig. 2). This sod2⌬ impact on extracellular pH appears specific to respiratory conditions, as sod2⌬ cells are indistinguishable from WT cells in acid production under fermentative conditions with high glucose (Fig. 3A, right). Moreover, the early acid phase of sod2⌬ cells under respiratory conditions is not due to accelerated death of these cells, because our viability tests (see "Experimental Procedures") estimate that WT and sod2⌬ both retain Ϸ90% viability after 11 days on glycerol.
The premature acid burst appears specific to sod2⌬ strains and was not observed in other yeast mutants we tested that control oxidative stress resistance. For example, deletion of genes required for glutathione and thioredoxin reduction in the mitochondria (GLR1 (27) and TRR2 (40)) did not show the same accelerated acid burst of sod2⌬ cells, nor did deletion of the thioredoxin and glutathione peroxidases AHP1 (41), TSA1 (42), and GPX1 (43) (data not shown). We also observed that loss of SOD1, encoding the cytosolic and mitochondrial intermembrane space Cu,Zn-SOD, does not induce a premature acid burst (Fig. 3A). The effects on extracellular pH appear specific to loss of SOD activity on the matrix side of the mitochondrial inner membrane.
In addition to SOD enzymes, yeast cells have the capacity to remove superoxide through the action of small non-proteinaceous complexes of manganese such as Mn-phosphate. These redox active compounds of manganese, so called Mn-antioxidants, are effective SOD mimics in vitro and in yeast cells in vivo (44 -50). Mn-antioxidants are elevated when yeast cells accumulate high manganese through either supplementation of manganese salts to the growth medium or by mutations in the pmr1⌬ Golgi pump for manganese (46 -48). We observed that in both cases, the Mn-antioxidants prevented the premature acid burst of sod2⌬ strains (Fig. 3, B and C left). Moreover, boosting levels of Mn-antioxidants in WT cells completely prevented cells from exiting the alkaline phase and the acid burst was eliminated (Fig. 3C, right). These studies together with results from sod2⌬ cells strongly indicate that the transition to acid phase following long-term respiratory conditions is triggered by superoxide.
Redox cycling agents such as paraquat can induce production of cellular superoxide, including mitochondrial matrix superoxide (51)(52)(53). As seen in Fig. 4A, paraquat treatment dramatically accelerated the transition to the acid burst in glycerol and within 24 h, both WT and sod2⌬ cells treated with paraquat were producing copious levels of acid. Much less paraquat was required in the case of sod2⌬ cells (Fig. 4A), consistent with the anticipated high superoxide in these mutants that cannot catalytically remove mitochondrial matrix superoxide. In addition to effects on glycerol plates, paraquat treatment also caused cells grown in low glucose to produce acid rather than base (Fig.  4B). The paraquat-induced burst of acid was not due to cell death; in fact total cell growth over 24 h was slowed by only 25-30% with doses of paraquat that induced maximal acidification (Fig. 4C), which translates to an overall reduction in growth rate of only Ϸ15%. These studies support the notion that acid production during respiratory conditions can be induced by superoxide, specifically the mitochondrial superoxide relevant to Sod2p.
Fe-S Cluster Enzymes of the TCA Cycle as Targets of Superoxide during the Acid Burst-Superoxide can disrupt Fe-S clusters (54,55), and we tested whether the mitochondrial Fe-S proteins SDH and aconitase were inhibited in cells producing acid. In WT cells, both aconitase and SDH are affected by paraquat. Aconitase is inhibited even by low (50 M) levels of paraquat when cells are not producing acid, but maximal levels of SDH inhibition requires very high (800 M) paraquat, conditions where WT cells produce acid (Fig. 5A). In sod2⌬ cells, aconitase FIGURE 2. Acid production following the alkaline phase of yeast cell growth is accelerated by sod2⌬ mutations and eliminated by ald4⌬ mutations. The indicated yeast strains were analyzed for acid and base production following the designated times of incubation on glycerol containing plates (no calcium) as described in the legend to Fig. 1A. is constitutively low without paraquat consistent with previous studies (55,56), and the main effect of 50 M paraquat is inhibition of SDH activity (Fig. 5A). Hence with WT and sod2⌬ cells, acid production is associated with losses in both aconitase and SDH. The same trends were apparent with cells grown for extended periods on glycerol without paraquat: WT cells exhibited reductions in both aconitase and SDH, whereas sod2⌬ cells are constitutively low in aconitase and lose SDH activity over time (Fig. 5B).
We tested whether the degree of SDH and aconitase inhibition observed in vitro reflected disruptions in the TCA cycle, in vivo. TCA cycle metabolites were measured from cells grown in low glucose and treated with paraquat. As controls, mutants of SDH and aconitase were used. Yeast sdh4⌬ mutants have high succinate but low citrate, whereas aco1⌬ mutants lacking aconitase have high citrate and low succinate (Fig. 5C). WT cells treated with 800 M paraquat exhibited the elevation in succinate typical of sdh mutants, but did not lose citrate (Fig. 5C), consistent with the double inhibition of SDH and aconitase (as in Fig. 5A). In sod2⌬ null cells treated with 50 M paraquat, citrate was elevated consistent with losses in aconitase, but there was no loss of succinate (Fig. 5C), again consistent with TCA blockages in both SDH and aconitase.
The correlation between the acid burst and TCA cycle inhibition prompted us to test whether mutations in TCA cycle genes were by themselves able to induce an acid burst. As seen in Fig. 4D, the aco1⌬ and sdh2⌬ mutants lacking aconitase and SDH showed acid production in low glucose, analogous to what was seen with paraquat treatment. Moreover, there was no additive effect of paraquat, indicating that paraquat and loss of the TCA cycle operate in the same pathway to induce acid. Acid production is not specific to disruptions at aconitase and SDH, as mutations in citrate synthase (cit1⌬), ␣-ketoglutarate dehydrogenase (kgd1⌬), and fumarase (fum1⌬) all resulted in acid production (Table 1). Disruptions in 5 of 8 steps in the TCA cycle mimicked the effects of paraquat and induced acid production in low glucose. Only deletions in isocitrate dehydrogenase (idh1⌬), succinyl-CoA ligase (lsc1⌬), and malate dehy-drogenase (mdh1⌬) resulted in little to no acid production ( Table 1).
Because the TCA cycle is required for electron transport chain functioning, we addressed whether blockages in the electron transport chain also lead to the acid burst. As seen in Table  1, mutations that inhibit the electron transport chain, other than SDH loss, including deletions affecting ubiquinone, complex III, or complex IV, and rho Ϫ mutations lacking mitochondrial DNA, did not rapidly acidify the medium. Thus, it is not a general defect in mitochondrial function that leads to acid production in low glucose, but specific disruptions in the TCA cycle.
The Acid Burst Results from Acetate Production by the Aldehyde Dehydrogenase Ald4p-What is the source of the acid burst with low glucose? In the case of high glucose, yeast acidification of the environment involves activation of a cell surface proton ATPase Pma1p, as well as cellular export of organic acids (11)(12)(13). As seen in Fig. 6A, cells grown in low glucose have extremely low levels of Pma1p activity compared with cells grown in high glucose. However, Pma1p activity remained low in the TCA cycle mutants that produce acid (Fig. 6, B and C). We therefore examined effects of organic acids. Previous studies have shown that Fe-S cluster assembly or delivery defects result in extracellular acidification, marked by secretion of various organic acid mixtures (57)(58)(59) and, in yeast, sdh3 and sdh4 mutants secrete succinate (60). However, sdh mutants grown under glucose starvation conditions exhibit only small increases in extracellular succinate (Fig. 6D). The same is true with WT cells treated with paraquat (Fig. 6D). These small increases in extracellular succinate are not likely to account for the large changes in extracellular pH, and other acid producing cells (e. g. cit1⌬, aco1⌬, kdg1⌬, and fum1⌬) do not elevate extracellular succinate (Fig. 6D). Another organic acid more highly elevated in sdh mutants is acetate (61), which feeds into the TCA cycle via acetyl-CoA. WT yeast cells grown in high glucose produce abundant acetate (13) but with glucose-starved WT cells, extracellular acetate is virtually undetectable (Fig. 7A). Remarkably, when glucose-starved WT cells are treated with paraquat, extracellular acetate levels rise dramatically to levels comparable with that seen with high glucose (Fig. 7, A and B). The high expulsion of acetate into the growth medium was also seen in sod2⌬ cells treated with low paraquat (Fig. 7B) and with all the acid producing TCA cycle mutants (sdh2⌬, cit1⌬, aco1⌬, kgd1⌬, and fum1⌬) (Fig. 7D). In every case where cells make acid with low glucose, there is Ն100-fold higher levels of extracellular acetate in the growth medium than is seen with control cells not producing acid (Fig. 7, A, B, and D, and Table 2). Intracellular acetate is also elevated, albeit not to the same degree as extracellular acetate (Fig. 7, C and E). The high level of extracellular acetate can easily account for the acid burst observed, as spiking the growth medium with levels of acetic acid produced in paraquat-treated cells (0.276 g/l) was sufficient to lower the pH to roughly the same degree ( Table 2).
The acetate that normally enters the TCA cycle can be derived from the cytosolic and/or mitochondrial pyruvate dehydrogenase bypass pathways or from the hydrolysis of acetyl-CoA. In the pyruvate dehydrogenase bypass pathways, pyruvate is converted to acetaldehyde, which in turn, is converted to acetate via the action of aldehyde dehydrogenase (ALD) enzymes (62). S. cerevisiae expresses 5 aldehyde dehydrogenase isoforms (63) and we tested if mutants in any of these would prevent the acid burst. Of these, deletion of the major mitochondrial isoform Ald4p most significantly inhibited paraquat-induced acidification of the growth medium with low glucose (Fig. 8A). Loss of Ald4p also attenuated the constitutive acidification of low glucose medium by sdh4⌬ mutants (Fig.  8B). Consistent with these reversals of acidification, we observed corresponding decreases in acetate production when ald4⌬ is disrupted in paraquat-treated cells and in sdh4⌬ mutants grown with low glucose (Fig. 7F). The effect of ald4⌬ mutations was also examined in cells grown under long-term glycerol conditions. As seen in Fig. 8C, ald4⌬ mutations prevented the early acid burst of sod2⌬ cells on glycerol. Moreover, loss of ALD4 in the background of SOD2ϩ cells completely prevented the acid burst with long-term respiratory conditions, and the extracellular pH continued to rise over time with ald4⌬ mutants (Fig. 2). Under all cases studied, the acid burst produced by cells under glucose starvation is dependent on mitochondrial pyruvate dehydrogenase bypass enzyme, Ald4p.
Conditioning the Growth Medium by the Acid Burst-What is the significance of the Acid Burst? As one possibility, cells FIGURE 5. Activities of aconitase and SDH under respiratory conditions. A and C, the designated yeast strains were grown as described under "Experimental Procedures" in 0.2% glucose medium that was supplemented with the indicated concentrations of paraquat. A, extracellular pH shown at the bottom was measured in triplicate samples as described in the legend to Fig.  1B, and whole cell lysates were prepared and assayed for aconitase and SDH activity as described under "Experimental Procedures." Units are represented as either nanomoles of cis-aconitate converted/mg of protein/min (aconitase), or nanomoles of 2,6-dichlorindophenol reduced/mg of protein/min (SDH); values represent the average of triplicate samples; error bars ϭ S.D. The data are representative of 6 -9 independent cultures analyzed over 2-3 experimental trials, with the possible exception of paraquat effects on sod2⌬ aconitase. Paraquat either had little impact on the low aconitase of sod2⌬ cells (as shown) or, in some studies, would increase this activity, perhaps reflecting the induction of the corresponding gene by severe oxidative stress (70). B, the indicated cells were spotted on glycerol containing medium as described in the legend to Fig. 1A, and after the indicated incubation times in days, cells were harvested and assayed for aconitase and SDH activity. The data are representative of 2-3 independent experimental trials, each run in duplicate; error bars ϭ S.D. C, the indicated intracellular metabolites were measured by gas chromatography mass spectrometry as described under "Experimental Procedures." Values represent averages of duplicate samples; error bars ϭ S.D.

TABLE 1
Effects of TCA and ETC mutants on extracellular pH pH was measured as described under ЉExperimental ProceduresЉ after the indicated strains were grown 24 h in 0.2% glucose growth medium. The average pH of triplicate samples is denoted with S.D. in parentheses. For WT ϩ PQ, WT cells were grown in media containing 800 M paraquat (ETC, electron transport chain; mito, mitochondrial). might produce copious amounts of acetate to provide a carbon source for neighboring cells under conditions of long-term carbon source starvation. To determine whether acetate production and/or media acidification could enhance growth of yeast cells, we collected "conditioned" growth media that was conditioned by cells grown 24 h in low glucose. Low acetate/high pH medium was obtained from cultures of WT and ald4⌬ cells, and high acetate/low pH medium was obtained from cultures of sdh4⌬ cells and from sod2⌬ cells treated with 50 M paraquat, a low dose that is non-toxic to WT cells (Fig. 4C). After removal of cells by filtration, these various conditioned media were used to inoculate WT cells and growth was monitored over 10 h. As seen in Fig. 9A, the high acetate/low pH media of sdh4⌬ and paraquat-treated sod2⌬ cells supported strong growth of WT cells, whereas the low acetate/high pH media from WT cells or from ald4⌬ mutants was poorly effective at supporting growth.

TCA mutants ETC mutants
To determine whether acetate alone accounts for these differences, we supplemented the low acetate medium from WT and ald4⌬ cells with levels of glacial acetic acid (0.20 -0.22 g/l) equivalent to that in cultures of sdh4⌬ and paraquat-treated sod2⌬ cells. This supplementation of acetic acid was sufficient to eliminate differences in the various conditioned media, and all four media types supported growth of WT cells (Fig. 9B). Thus, the production of acetate can enhance growth of yeast cells under glucose starvation. This event triggered by superoxide disruptions in the TCA cycle may enable the organism to withstand long-term periods of starvation.

DISCUSSION
Herein, we describe a mechanism by which the S. cerevisiae yeast modifies its extracellular environment during prolonged nutrient depravation. Under long-term colonization on respiratory carbon sources, yeast cells shift from alkalization to producing acid and we demonstrate here that this acid burst is dependent on the acetate-producing Ald4p aldehyde dehydrogenase. Media conditioned by Ald4-acetate is indeed able to promote yeast cell growth under glucose starvation conditions. Our findings are consistent with a model in which the acid burst is produced in response to disruption of the TCA cycle by superoxide (Fig. 9C).
Perhaps the most intriguing finding of this work is the apparent role of mitochondrial matrix superoxide in conditioning the extracellular environment. Mitochondrial superoxide is not likely to cross mitochondrial membranes (30) and yet it can  influence extracellular pH by modulating organic acid production. In support of this effect of mitochondrial superoxide, we observed that the acid burst under long-term glycerol conditions is accelerated by sod2⌬ mutations lacking mitochondrial matrix Mn-SOD, but not by sod1⌬ mutations affecting the largely cytosolic Cu,Zn-SOD1. Previously, the loss of Sod2p was shown to reduce volatile ammonia production and prevent transition to the alkali phase under long-term glycerol conditions but the mechanism was unknown (19). This may be explained by the early acid burst of sod2⌬ cells which by lowering pH would trap a significant portion of ammonia (pK a Ϸ 9.25) in the non-volatile ammonium form. Alternatively, Sod2p may have separable roles in regulating extracellular pH through ammonia and acetate production. In further support of the role of superoxide in triggering the acid burst, we observe that acid production under both glycerol and low glucose conditions is dramatically accelerated when WT cells are treated with the redox cycler paraquat; conversely, the acid burst with long-term glycerol conditions is eliminated by Mn-antioxidants, chemical mimics of SOD enzymes that can remove superoxide in yeast (44 -50). Superoxide, more specifically mitochondrial superoxide, represents an ideal candidate to initiate the acid burst. Key Fe-S cluster enzymes in the TCA cycle (i.e. SDH and aconitase) are sensitive to superoxide inactivation (24 -26, 54, 55), and we show here that disruption in the TCA cycle can cause build-up of acetate derived from mitochondrial Ald4p. Although multiple reports have established a role for ROS in cell signaling and regulation (64 -67), the inactivation of TCA cycle enzymes by mitochondrial superoxide was up until now considered a detrimental outcome of oxidative damage. This is, perhaps, the first example where superoxide inactivation of mitochondrial enzymes can be beneficial, particularly under the stress conditions of nutrient depravation.
Superoxide is a natural by-product of respiration, yet in WT cells the acid burst only occurs after prolonged incubations under respiratory conditions (as in Fig. 2). The products of superoxide damage appear to accumulate slowly in these cells, as we observe a gradual loss in SDH activity over prolonged respiratory growth (Fig. 5B). In sod2⌬ cells the combined effects of this slow loss in SDH together with the already low aconitase of these cells can explain the acceleration in the acid burst observed. Once SDH and/or aconitase activity have been sufficiently inhibited to disrupt TCA cycle functioning, the Ald4p acid burst occurs. In the simplest model, the blockage in FIGURE 8. The mitochondrial Ald4p aldehyde dehydrogenase is required for the acid burst. The indicated strains were grown either in low glucose liquid medium (A and B) or on glycerol containing plates (C) and pH was monitored as described in the legend to Fig. 1. A and B, pH values represent the results of duplicate (A) or triplicate (B) samples where S.D. Յ0.12. ϩPQ ϭ cultures supplemented with 800 M paraquat. FIGURE 9. Medium conditioned during the acid burst can promote yeast growth. Growth medium was collected from the indicated cultures following 24 h incubation in 0.2% glucose medium. This cell-free "conditioned medium" was used to inoculate WT cultures of yeast as described under "Experimental Procedures" and growth was monitored turbidimetrically at A 600 . Results represent the averages of triplicate cultures; error bars ϭ S.D. In some cases, the error was too small to be seen by error bars. A, conditioned medium was used as isolated from cells. B, the levels of acetate in all four types of conditioned medium were normalized by supplementing 0.20 -0.22 g/l of glacial acetic acid to the medium conditioned by WT and ald4⌬ cells. C, a model for how mitochondrial superoxide produced during long-term respiratory conditions can lead to the acid burst and conditioning of the extracellular environment. Superoxide can lead to disruption of the TCA cycle through damage to the Fe-S enzymes aconitase and/or SDH. The blockage in the TCA cycle causes a build up in acetate derived from the mitochondrial Ald4p aldehyde dehydrogenase. The high levels of acetate expelled into the growth medium can condition the extracellular environment to promote growth under glucose starvation conditions. the TCA cycle itself prevents mitochondrial consumption of Ald4-acetate, and the accumulated metabolite is expelled from the cell. Ald4 production of acetate might also be enhanced in a futile attempt to correct the TCA cycle defect. It is noteworthy that blockages in 5 of 8 steps in the TCA cycle resulted in an acetate burst, including losses in SDH and aconitase, as well as citrate dehydrogenase, ␣-ketoglutarate dehydrogenase, and fumarase. It is possible that mutants for the remaining three steps (idh1⌬, mdh1⌬, and lsc1⌬ (68, 69)) do not yield the same disruption in TCA cycle functioning.
Although high acetate has previously been associated with increased chronological aging in yeast (8,9), these previous studies were conducted with fermenting yeast cells with abundant glucose. In contrast, our studies with respiring yeast (as in Fig. 9) indicate that high acetate can be beneficial in environments with scarce nutrients. Overall, these studies provide the first line of evidence for mitochondrial superoxide and the TCA cycle in conditioning the extracellular environment under glucose starvation conditions.