Superoxide Inhibits 4Fe-4S Cluster Enzymes Involved in Amino Acid Biosynthesis CROSS-COMPARTMENT PROTECTION BY CuZn-SUPEROXIDE DISMUTASE*

Among the phenotypes of Saccharomyces cerevisiae mutants lacking CuZn-superoxide dismutase (Sod1p) is an aerobic lysine auxotrophy; in the current work we show an additional leaky auxotrophy for leucine. The lysine and leucine biosynthetic pathways each contain a 4Fe-4S cluster enzyme homologous to aconitase and likely to be superoxide-sensitive, homoaconitase (Lys4p) and isopropylmalate dehydratase (Leu1p), respectively. We present evidence that direct aerobic inactivation of these enzymes in sod1 (cid:1) yeast results in the auxotrophies. Located in the cytosol and intermembrane space of the mitochondria, Sod1p likely provides direct protection of the cytosolic enzyme Leu1p. Surprisingly, Lys4p does not share a compartment with Sod1p but is located in the mitochondrial matrix. The activity of a second matrix protein, the tricarboxylic acid cycle enzyme aconitase, was sim-ilarly lowered in sod1 (cid:1) mutants. We measured only slight changes in total mitochondrial iron and found no detectable difference in mitochondrial “free” (EPR-detectable) iron making it unlikely that a gross defect in mitochondrial iron metabolism is the cause of the decreased enzyme activities. Thus, we conclude that

The antioxidant enzyme copper/zinc-superoxide dismutase (CuZn-SOD or Sod1p) 1 plays an integral role in the protection of many organisms from the oxidative aerobic environment.
CuZn-SOD is localized to the cytosol, nucleus, and the intermembrane space (IMS) of the mitochondria, suggesting that it exerts its protective effect in multiple compartments (1). Another SOD that contains manganese (Mn-SOD or Sod2p) is located in the matrix of mitochondria. Saccharomyces cerevisiae lacking CuZn-SOD (sod1⌬) have distinct and well established aerobic phenotypes including diminished growth, auxotrophies for lysine and either methionine or cysteine, decreased ability to grow on nonfermentable carbon sources (2,3), increased "free" (EPR-detectable) iron (4), hypersensitivity to millimolar concentrations of zinc (5), and exquisite sensitivity to redox-cycling drugs such as the herbicide paraquat (6,7). All of these phenotypes (except zinc sensitivity) are also observed in mutants lacking the copper chaperone for CuZn-SOD 2 (lys7⌬ or ccs1⌬) and thus contain a form of the Sod1p polypeptide that is inactive because of a lack of copper in the active site (5,8). Mutants lacking Sod2p have a less dramatic phenotype; they are sensitive to redox cycling drugs and grow poorly on nonfermentable carbon sources but show no aerobic auxotrophies (2).
Many cellular components are susceptible to oxidative damage, including proteins, lipids, and DNA (9,10). However, targets of superoxide-specific damage are much more limited. A particular type of protein prosthetic group, solvent-exposed 4Fe-4S clusters occurring in nonelectron transfer proteins, have been shown to be specifically damaged by superoxide at low concentrations (11). These clusters differ from the ironsulfur clusters found in electron transfer proteins in that one iron atom, rather than being covalently bound to the enzyme, is labile and able to interact with its substrate to promote dehydration. The most extensively studied enzyme of this class is aconitase, a dehydratase of the tricarboxylic acid cycle located in the mitochondrial matrix. Aconitase is subject to reversible inactivation by superoxide resulting in loss of the labile iron atom, leaving a 3Fe-4S cluster (12,13). We have observed previously reversible inactivation of aconitase in yeast lacking Mn-SOD (14) supporting a model of superoxide-mediated inactivation of 4Fe-4S cluster enzymes in the mitochondrial matrix. Studies in bacterial systems have shown that two 4Fe-4S cluster containing dehydratases, dihydroxy-acid dehydratase and 6-phosphogluconate dehydratase (15), are specifically sensitive to superoxide (16,17). Inactivation of these enzymes leads to a deficiency in branched chain amino acid biosynthesis and a decreased capacity to utilize gluconate as a carbon source, respectively.
In the eukaryotic organism S. cerevisiae, two enzymes share distinct homology with aconitase, homoaconitase (Lys4p) and isopropylmalate dehydratase (Leu1p). These enzymes are involved in the biosynthesis of lysine and leucine, respectively. Lys4p catalyzes the conversion of cis-homoaconitate to homoisocitrate as part of the ␣-aminoadipate pathway of lysine biosynthesis (18). Although it has not been directly proven to contain a 4Fe-4S cluster, Lys4p is homologous to aconitase, particularly at active site residues involved in the 4Fe-4S cluster coordination (18,19), and its activity can be inhibited by iron chelators (20). Lys4p is therefore a potential target for superoxide-mediated inactivation. Leu1p catalyzes a two-step reaction that converts ␣-isopropylmalate to ␤-isopropylmalate during leucine biosynthesis and is located in the cytosol (18). Yeast defective in formation of cytosolic Fe-S clusters exhibit a leucine auxotrophy (21) indicating the necessity of the Leu1p Fe-S cluster.
Fe-S cluster metabolism is an important process as indicated by (a) the abundance of enzymes requiring some form of cluster, and (b) the fact that mutations affecting cluster synthesis are often lethal. Fe-S synthesis occurs in the mitochondria, and defects in the biosynthetic pathway generally result in dramatic accumulation of iron within the mitochondria (22)(23)(24). Iron entry into the mitochondria for Fe-S cluster production, as well as heme synthesis, requires a mitochondrial membrane potential, a reducing environment, and the availability of ATP and NADH (25,26).
A tight lysine auxotrophy is a classic phenotype of sod1⌬ yeast. We hypothesized that Lys4p is susceptible to superoxide inactivation and that this sensitivity leads to the lysine auxotrophy. A leucine auxotrophy has not been reported for sod1⌬ yeast, but based on the above information we decided to look for and found a previously undescribed leaky leucine auxotrophy in sod1⌬ yeast. In this study we demonstrate inactivation of the iron-sulfur cluster enzymes Lys4p and Leu1p in the sod1⌬ mutant under aerobic conditions. In addition, we localize Lys4p in the mitochondrial matrix and show that levels of mitochondrial total and free iron are similar in wild type and sod1⌬ strains. Overall, we conclude that both auxotrophies are caused by direct inactivation of the respective protein by superoxide and that Sod1p protects them by lowering superoxide levels. In the case of Lys4p, the protection by Sod1p is exerted from one compartment (the IMS) across a membrane barrier to the matrix and is required despite the presence of Mn-SOD in the matrix.

EXPERIMENTAL PROCEDURES
Yeast Strains, Media, and Growth Conditions-Strains are listed in Table I (27-29). Plasmid pADCl (30) was obtained from Catherine F. Clarke. Plasmid pADCL-LYS4-HA, a high copy plasmid carrying an epitope-tagged homoaconitase gene, was prepared as follows. A 2-kb fragment containing the LYS4 gene was obtained by PCR of genomic DNA isolated from yeast strain S288C using PCR primers 5Ј-TCGCG-TCGACATGCTACGATCAACCACATTTACTCG and 5Ј-AGATATTTC-GGGCCGCCCTAGTTGGGATTTGACCCAACCTTCC. This fragment was digested with SalI and NotI and cloned into the polylinker site of plasmid pADCL, which contains a hemagglutinin (HA) epitope tag, creating plasmid pADCL-LYS4-HA. This plasmid was transformed into the wild type yeast strains EG103, MO-59 -13c, and X3356-1B by using standard yeast transformation procedures. To reintroduce the LEU2 gene, strains EG103 and EG118 were transformed with plasmid YIp351 (31) to create EG103L (wild type LEU2) and EG118L (sod1⌬ LEU2). In strain JW101 (lys1⌬) the LEU2 gene was introduced as part of the deletion of CCS1(LYS7). These strains were used to examine the activity of Leu1p.
Unless otherwise stated, all experiments in liquid media were performed in synthetic medium with 2% glucose (SD) supplemented with amino acids, adenine, and uracil as described (32). In addition, a 4-fold excess of the supplements tryptophan, leucine, uracil, adenine, methionine, and histidine was added (SDC). Specific amino acids were dropped out as indicated (e.g. SD-Leu is SDC with leucine omitted). Growth was followed by monitoring the turbidity at 600 nm (A 600 ; A 600 of 1 ϭ 1 ϫ 10 7 cells/ml). For paraquat studies, yeast pre-cultures were grown in standard YPD media and inoculated at an initial A 600 of 0.05, although all other experiments used pre-cultures grown in SDC. For the Lys4p experiments, cultures were inoculated in SDC at an A 600 of 0.025 for EG103/EG110 and an A 600 of 0.05 for EG118. These cultures were allowed to grow at 30°C, 220 rpm until mid-log phase (A 600 of 1-2), spun down, washed, and resuspended in SD-Leu-Lys-Trp. This medium was used so that the effect of the absence of lysine could be measured under conditions where neither strain could grow, because removing only lysine would have allowed the wild type but not the sod1⌬ mutant to grow. These cultures were then incubated at 30°C, 220 rpm and collected at the indicated times. For Leu1p experiments, cultures were inoculated in SD-Leu at an A 600 of 0.05 for EG103L and EG118L. Yeast for aconitase experiments were inoculated in SDC medium at an A 600 of 0.05. The above cultures were allowed to grow at 30°C, 220 rpm until mid-log phase (assays) or with timed absorbance readings as indicated.
Protein Extracts-Glass bead lysis was used to prepare crude soluble extracts. Samples were processed and stored under argon or nitrogen to avoid exposure of the lysates to oxygen, preventing inactivation of the enzymes during and after lysis. For Lys4p experiments, equal volumes of concentrated cell suspension (A 600 10 -30) and 0.5-mm glass beads were mixed and subjected to 6 -10 rounds of 1 min of vortexing followed by 1 min on ice, under argon in a septum-sealed glass test tube. Particulate matter was spun out by centrifugation at 3080 ϫ g in a Jouan CR422 centrifuge. For Leu1p and aconitase assays, lysis was performed under nitrogen in a septum-sealed test tube using 0.5-mm glass beads with a ratio of 30 s vortexing followed by 30 s on ice, six times.
Enzymatic Assays-Lysine levels were determined by measuring the oxidation of NADH to NAD ϩ coupled to the conversion of lysine to saccharopine by saccharopine dehydrogenase as described by Nakatani et al. (33). Lys4p activity was determined as described by Broquist (34) and Strassman and Ceci (35) using homoisocitrate as a substrate and measuring spectrophotometrically the production of homoaconitate, which absorbs at 240 nm. Briefly, the absorbance at 240 nm of 1 nmol of homoisocitrate in 0.33 M potassium phosphate, pH 8.5, was followed for 10 min at 30°C. The activity of the extracts was taken as the difference between the slopes obtained with and without added substrate. Leu1p activity was determined spectrophotometrically in a sim- ilar manner as described by Kohlhaw (36), by watching the disappearance of citraconate (Aldrich) at 235 nm. Briefly, a 0.5-ml assay mixture containing 4 mM potassium phosphate, pH 7.0, 0.4 mM citraconate, and 100 -300 g of protein was assayed for 3 min in a 2-mm path length cuvette. Aconitase activity was determined spectrophotometrically as described by Gardner et al. (37) by monitoring the formation of NADPH at 340 nm. Briefly, the assay mixture contained 50 mM Tris-HCl, pH 7.5, 5 mM sodium citrate, 0.6 mM MnCl 2 , 0.2 mM NADP ϩ , 2 units of NADP ϩ isocitrate dehydrogenase, and 50 -100 g of protein.
The absorbance change was measured for 5 min, and the slope was calculated from the linear portion.
Isolation of Mitochondria and Localization of Homoaconitase-Mitochondria were isolated from EG103, LL101, and EG118 and purified on linear Nycodenz density gradients as described by Glick and Pon (38). The partial lysis experiment described in Fig. 5 was performed as described (39,40) except that trypsin was used, because CuZn-SOD is resistant to proteinase K under the conditions of the experiment. Briefly, samples of purified mitochondria were diluted 10-fold into 20 mM K-HEPES, pH 7.4, containing 0.1 mg/ml trypsin and sorbitol such that the final sorbitol concentrations ranged from 0.6 M (no change in sorbitol concentration to maintain whole mitochondria) to 0.06 M (10fold decrease in sorbitol concentration to rupture the outer membrane, forming mitoplasts and releasing the intermembrane space proteins into solution where they are digested by the trypsin). Mitochondria/ mitoplasts were then pelleted, and samples were analyzed by SDS-PAGE followed by Western blotting to determine subcellular localization of the proteins. An ECL kit (Amersham Biosciences) was used for detection. Polyclonal antibodies raised against yeast Sod1p were affinity-purified using cyanogen bromide-activated Sepharose covalently linked to the appropriate protein (41). Polyclonal antibodies against cytochrome b 2 (cyt b 2 ), Hsp60, and ␣-ketoglutarate dehydrogenase (KGD1) were kindly provided by C. Koehler (UCLA). Affinity-purified anti-HA antibody was purchased from Sigma.
EPR Analysis-Mitochondrial samples for EPR were prepared by placing 150 l of 25 mg/ml purified mitochondria (see above) into a 4-mm EPR tube (Wilmad), flash-frozen in dry ice/ethanol, and stored at Ϫ80°C. "Free iron" (high spin rhombic Fe(III), g ϭ 4.3) spectra were collected on a Bruker X-band spectrophotometer at 96 K using a variable temperature nitrogen gas setup with a Wilmad quartz Dewar insert. Parameters were as follows: center field, 1560 G; sweep width, 500 G; frequency 9.33 GHz; microwave power, 31.8 milliwatts; modulation amplitude, 20 G; modulation frequency, 100 kHz; receiver gain, 1.1 ϫ 10 5 ; time constant, 81.92 ms. Each spectrum is representative of an average of eight scans. Free iron quantification was done as described (42). All EPR data processing was performed with Bruker WINEPR software.
Metal Analysis-Total mitochondrial iron levels were determined on a Thermo-Jarrel Ash Iris 1000 inductively coupled plasma atomic emission spectrometer (ICP-AE). 1 mg of purified mitochondria was placed in an Eppendorf tube, allowed to dry at 95°C for 4 h, dissolved in 1 ml of 20% nitric acid Optima (Fisher), and incubated at 95°C for 18 h in a sealed tube. Digestion volume was confirmed to be constant, and the full 1 ml was added to 3.5 ml of metal-free water. A blank (1 ml of nitric acid to 3.5 ml metal-free water) was also included, and the value was subtracted from each sample.

RESULTS
The Leaky Leucine Auxotrophy of CuZn-SOD Delete Yeast-A classic phenotype of yeast lacking CuZn-SOD (sod1⌬) is their inability to grow aerobically in the absence of the amino acids lysine and methionine (or cysteine) in their growth medium (43). These auxotrophies are very tight; there is little to no growth in air without the addition of lysine and methionine to the sod1⌬ strain ( Fig. 1B) (3) or to the ccs1⌬ strain (Fig. 1C). Note that the sod1⌬ and ccs1⌬ strains grow well in complete medium or in medium lacking threonine.
Although there is a 4Fe-4S cluster protein in the leucine biosynthetic pathway, Leu1p, that is potentially superoxidesensitive, no defects in leucine synthesis have been reported previously for sod1⌬ yeast. Such an auxotrophy may have been overlooked in the earlier studies because many of the parental strains utilized carried a leu2 mutation and were therefore unable to synthesize leucine, regardless of the status of Leu1p (43). Therefore, we constructed LEU2 ϩ strains EG103L and EG118L by transforming EG103 and EG118 strains with an integrating plasmid carrying the LEU2 gene. We then examined the ability of EG103L and EG118L to grow aerobically in medium lacking leucine. These experiments showed that there is a leaky auxotrophy for leucine in strains lacking CuZn-SOD activity, and strains with sod1⌬ or ccs1⌬ mutations grew much more slowly without leucine (Fig. 1). Similar results were obtained with sod1⌬ and ccs1⌬ mutants derived from the S. cerevisiae strain BY4741 (data not shown). As with the methionine and lysine auxotrophies, the leucine auxotrophy was not observed in sod1⌬ cells grown anaerobically or in strains lacking the matrix enzyme Mn-SOD (SOD1 ϩ sod2⌬) (data not shown).
Induced Amino Acid Auxotrophies in Wild Type Yeast Exposed to Paraquat-The bipyridyl herbicide paraquat (methyl viologen) is known to give rise to intracellular superoxide production, and sod1⌬ yeast are exquisitely sensitive to it. Elevated concentrations of paraquat in wild type yeast are expected to replicate the phenotype of sod1⌬ strains assuming enough superoxide is generated to overtax the ability of the cells to cope with it. Wild type cells can grow in complete medium with concentrations of paraquat as high as 5 or 10 mM (28). However, 0.25 mM paraquat induced amino acid auxotrophies for lysine, methionine, and leucine (Fig. 2), indicating that the pathways producing these amino acids are especially sensitive to superoxide. Indeed, the observed pattern is quite similar to that seen in sod1⌬ yeast, and paraquat causes al- most complete growth inhibition of the wild type strain in media lacking lysine or methionine and partial growth inhibition in medium lacking leucine. These data support the conclusion that the three amino acid auxotrophies of sod1⌬ yeast are specifically due to the presence of superoxide and prompted us to look for inactivation of specific 4Fe-4S cluster proteins involved in these amino acid biosynthetic pathways.
The Activity of 4Fe-4S Cluster Enzymes-We analyzed the activity of the three aconitase-like 4Fe-4S cluster enzymes in our wild type and sod1⌬ strains in order to determine whether their activities are affected by the absence of Sod1p.
Aconitase is the classic example of an enzyme that is inactivated by superoxide-mediated oxidation of its 4Fe-4S cluster (13). Aconitase has been reported to be inactivated in yeast lacking either Mn-SOD (sod2⌬) (14) or CuZn-SOD (44), and we found similar results under the conditions used in the present study (Fig. 3A). The addition of 1 mM paraquat to the wild type strain also caused a decrease in aconitase activity. Thus aconitase is susceptible to inactivation by superoxide produced by both natural and artificial means. Because aconitase is not a rate-limiting enzyme, decreases in its activity do not necessarily affect growth or respiration rate. Mn-SOD and aconitase are both located in the mitochondrial matrix, so it is perhaps not surprising that aconitase activity is decreased in sod2⌬ strains. The fact that aconitase activity was also affected in sod1⌬ yeast is harder to explain, because the two proteins are not present in the same cellular compartment.
The activity of Lys4p, the 4Fe-4S cluster containing enzyme in the lysine biosynthetic pathway, was measured in the sod1⌬, the sod2⌬, and wild type yeast that had been grown for 12 h in medium lacking lysine, tryptophan, and leucine. We observed a dramatic difference between wild type and the mutants, although the specific activity of Lys4p was low even in wild type cells (Fig. 3B). This result is somewhat surprising, as the sod2⌬ mutant does not exhibit a lysine auxotrophy. To address this point, we measured free lysine levels in the three strains, and we found that there was 12.5 Ϯ 2 nmol per mg of protein in wild type, 5.0 Ϯ 0.8 nmol per mg of protein in sod1⌬, and 16.0 Ϯ 2 nmol per mg of protein in sod2⌬. The level of free lysine in the sod1⌬ mutant was less than half that of wild type, indicating that the inactivation of Lys4p in this strain affected the amount of available lysine and providing a good explanation for the observed lysine auxotrophy. The sod2⌬ strain actually had a slightly higher level of lysine than the wild type strain, consistent with the fact that sod2⌬ yeasts do not exhibit a lysine auxotrophy. It remains unclear why the sod2⌬ strain can still make wild type levels of lysine with the drastically lowered enzyme activity that we observed.
To test whether the inactivation of the cytosolic 4Fe-4S cluster protein Leu1p could be the cause of the leaky leucine auxotrophy, we measured its activity in sod1⌬, sod2⌬, and wild type yeast, as well as wild type yeast exposed to paraquat. In sod1⌬ and wild type plus paraquat, the Leu1p activity was approximately half that of the wild type control (Fig. 3C). The moderate level of inactivation correlates well with the observed leakiness of the leucine auxotrophy; enough leucine is produced to support some growth but at a lower rate. Because Leu1p and Sod1p are both cytosolic proteins, it seems likely that Sod1p exerts a direct protective effect. The sod2⌬ strain had the same Leu1p activity as wild type, indicating that the matrix-localized Mn-SOD plays little or no role in protecting this cytosolic protein.
Restoration of 4Fe-4S Cluster Enzyme Activity by Low Oxygen-When cells are grown anaerobically, the lysine (and leucine) auxotrophies of the sod1⌬ mutant are not observed (43). We therefore sought to determine whether anaerobiosis affected the Lys4p and Leu1p activities. After 12 h of aerobic growth in SDC-Leu-Lys-Trp (Lys4p) or in SDC-Leu (Leu1p), cultures were divided in two. One of these cultures was placed , and aconitase were measured on extracts from EG103 (WT), EG103 ϩ 1 mM paraquat (PQ), EG118 (sod1⌬), and EG110 (sod2⌬) yeast strains. Extracts were prepared under argon or nitrogen to minimize enzyme inactivation, as Fe-S clusters are very sensitive to oxidation. A, aconitase activity was assayed on extracts of cells grown in SDC medium to an A 600 of 1.5. B, homoaconitase (Lys4p) activity was obtained after growth in SDC to an A 600 of 1.0 and subsequent incubation in SD-Lys-Leu-Trp medium for 12 h to induce Lys4p activity and to prevent further growth. C, isopropylmalate isomerase (Leu1p) activity was measured on extracts from cells grown in SD-Leu medium to an A 600 of 1.5. EG103L and EG118L were used to restore leucine biosynthetic competency. Asterisks indicate a significant difference from WT (p Ͻ 0.05 by the Student's t test). Values are averages of 5-8 separate cultures.
in low aeration conditions (either under 3% oxygen with no shaking or under 100% nitrogen), and the other was returned to normal culture conditions, shaking with ambient oxygen conditions (high aeration). As before, under high aeration the activities of these enzymes were reduced, but 3 h in low oxygen conditions fully restored the Lys4p activity in both sod1⌬ and sod2⌬ cells (Fig. 4A) and the Leu1p activity in sod1⌬ (Fig. 4C). Restoration of activity under low oxygen conditions was independent of protein synthesis, and the increase occurred even if cycloheximide was added to the culture medium at the time of the split (Fig. 4, B and D). These data are consistent with a reactivation event that involves repair or resynthesis of the 4Fe-4S cluster prosthetic group in existing protein molecules. In other words, Lys4p and Leu1p are inactivated but not degraded under superoxide stress.
Localization of Homoaconitase-Lys4p has been reported to be a mitochondrial enzyme (19), and it contains a potential mitochondrial localization signal similar to aconitase as suggested by hydrophobicity plots (data not shown). Because it is inactivated in sod1⌬ mutants, we thought it possible that Lys4p shared a location with Sod1p, i.e. that it was located in the IMS. To test this possibility, we constructed a hemagglutinin (HA)-tagged version of the enzyme. The HA-tagged protein complemented the lysine growth defect of two different lys4 mutant strains (M0-59-13c and X3356-1B), indicating that it was properly processed and active in vivo (data not shown).
Mitochondria purified from wild type yeast (LL101) expressing HA-tagged Lys4p were used in a partial lysis experiment to distinguish matrix proteins from IMS proteins. Samples of mitochondria were incubated in buffer containing trypsin and various concentrations of sorbitol, ranging from 0.6 M, where intact mitochondria are preserved, to 0.06 M, where complete lysis of the outer membrane occurs, leaving mitoplasts but releasing intermembrane space proteins into solution where they are degraded by the trypsin. As can be seen on the Western blot in Fig. 5, the bands for the matrix marker proteins Hsp60 and Kgd1p remain constant across the entire range of sorbitol concentrations, indicating that they are located in the mitoplasts (i.e. in the matrix), whereas the intensity of the IMS marker cyt b 2 band decreases dramatically as the sorbitol concentration decreases, indicating a location in the IMS. The HA-tagged Lys4p band remains constant at all sorbitol concentrations indicating that it is located in the mitochondrial matrix. In support of previous reports (1), we found that CuZn-SOD is present in the IMS. Thus, Lys4p, like aconitase, is present in the matrix and yet is dependent on Sod1p, a protein located in a different cellular compartment, for its full protection.
Total and Free Mitochondrial Iron-Previous results from this lab indicate that sod1⌬ mutant yeast exhibit alterations in iron metabolism, including a slight increase in total iron (45) and a large (3-4-fold) increase in free iron, or iron detectable by EPR at g ϭ 4.3 (4). Large increases (10-fold) in total mitochondrial iron (22-24) not bound to proteins or heme (23) have been reported for several mutations that affect Fe-S cluster biosynthesis. Therefore, we measured total and free iron in purified mitochondria from wild type and sod1⌬ yeast to see if our earlier whole cell results were indicative of a mitochondrial iron processing defect that could be affecting Fe-S cluster biosynthesis. Total iron, as measured by ICP-AE, was slightly increased in mitochondria from sod1⌬ cells (Fig. 6A), similar to what we previously observed in whole cells but unlike the large increases observed in mutants defective in Fe-S cluster biosynthesis. Free iron as measured by EPR at g ϭ 4.3 was not significantly increased in sod1⌬ mitochondria (Fig. 6B), in contrast to what is observed in whole cells or in mitochondria defective in Fe-S biosynthesis. Both the ICP-AE and EPR data indicate that sod1⌬ does not have a severe defect in mitochondrial iron metabolism, suggesting that Fe-S cluster production mechanisms are grossly intact. DISCUSSION Superoxide is produced by a number of cellular processes, the predominant source being leakage from the electron transport chain. Superoxide and its downstream products, including hydrogen peroxide and hydroxyl radical, are collectively known as reactive oxygen species. The toxic effects of reactive oxygen species are numerous and have been implicated in a number of damage paradigms and human diseases (9,10,46). However, damage due specifically to superoxide itself is much more limited. The best established target for superoxide is the exposed 4Fe-4S cluster in the active sites of many dehydratase enzymes, the best known of which is aconitase. Attack by super- (sod1⌬), and EG110 (sod2⌬) yeast were grown as described in Fig. 3 and switched to SD-Lys-Leu-Trp medium to induce amino acid biosynthetic pathways and to prevent further growth. Lys4p activity (A and B) and Leu1p activity (C and D) were determined. A and C, cultures were divided into 2 aliquots and incubated at 30°C for 3 h, with 1 aliquot shaken at 220 rpm under ambient oxygen (high aeration, H) and the other aliquot with no shaking under 3% oxygen (A) or nitrogen (C) (low aeration, L). B and D, cells were switched to SD-Lys-Leu-Trp under low aeration in the absence or presence of cycloheximide (Cyclohex) to inhibit protein synthesis. An asterisk indicates a significant difference between the high and low aeration condition (p Ͻ 0.05 by the Student's t test). One sample has no detectable activity (N.D.).
FIG. 5. Localization of Lys4p and CuZn-SOD. Mitochondria from LL101 cells expressing HA-tagged Lys4p were isolated and fractionated as described under "Experimental Procedures" to incrementally release IMS proteins as sorbitol concentration decreased. Immunoblotting was performed using antibodies that recognize Sod1p (IMS), cyt b 2 (IMS), HA (Lys4 tag), Kgd1p (matrix), and Hsp60 (matrix). Trypsin was used to degrade proteins not protected by a membrane, and Triton was used to completely solubilize all membranes and expose all proteins to trypsin. Lys4p was determined to be localized in the matrix. oxide oxidizes the cluster and leads to loss of the labile iron resulting in a 3Fe-4S cluster and enzyme inactivation (12)(13)(14)(15)(16)(17).
In the present study, we explored the molecular basis for the aerobic lysine auxotrophy of yeast lacking CuZn-SOD, and we found a previously unreported air-dependent leaky auxotrophy for leucine (Fig. 1). These same auxotrophies are induced in wild type yeast by treatment with the superoxide-generating drug paraquat (Fig. 2), indicating that oxidative stress specifically leads to the auxotrophies. This stress acts on 4Fe-4S cluster enzymes in the metabolic pathway of each amino acid, as evidenced by their inactivation in sod1⌬ yeast grown aerobically (Fig. 3). Reactivation of these 4Fe-4S cluster enzymes occurred in low oxygen even in the presence of the protein synthesis inhibitor cycloheximide (Fig. 4), implying that such reactivation is due to repair of the cluster and not through synthesis of new protein. Thus, we have found two more targets of superoxide-mediated inactivation, Lys4p (homoaconitase) and Leu1p (isopropylmalate dehydratase).
Leu1p and CuZn-SOD are both located in the cytosol, so it is logical that Sod1p provides protection to Leu1p by decreasing the local concentration of superoxide. We were more puzzled by the ability of CuZn-SOD to provide protection to the mitochondrial enzyme Lys4p. We considered three possible explanations.
We first considered the possibility that Lys4p was located in the IMS and thus shared a common location with CuZn-SOD.
Upon exploring the sub-mitochondrial localization of Lys4p, however, we found Lys4p to be in the matrix (Fig. 5), whereas Sod1p is in the IMS. Although this result was somewhat surprising, it is not completely unprecedented, as the activity of aconitase, also a matrix enzyme, is decreased when Sod1p is absent (Fig. 3) and (44). The activity of the matrix enzyme succinate dehydrogenase, which contains several nonexposed Fe-S clusters and which is inactivated in yeast and mice lacking Sod2p, was also observed to decrease in sod1⌬ yeast (data not shown). These results fit the general emerging picture, CuZn-SOD protects some Fe-S clusters in the mitochondrial matrix from its location in (an)other compartment(s).
Second, we considered the possibility that disruptions in iron metabolism could be the cause of the lowered activity of these Fe-S cluster proteins. Such disruptions could conceivably lead to two different manifestations, low or high mitochondrial iron. Because our previous work indicated that sod1⌬ yeast feel iron-starved (45), we considered it possible that, due to oxidizing conditions outside of the matrix in mutants without active CuZn-SOD, iron was failing to reach the mitochondria and that their mitochondrial iron was consequently low. Alternatively, many disruptions in Fe-S cluster synthesis pathways cause large increases (up to 10-fold or more) in mitochondrial iron (22)(23)(24), in a non-heme, non-iron sulfur state (23). To test these hypotheses, we measured total and free iron in mitochondria purified from wild type and sod1⌬ mutant yeast (Fig. 6). We observed only a very small increase in total mitochondrial iron in sod1⌬ cells, an increase similar in magnitude to what we have observed in whole cells (45) but different from that observed in yfh1 mutants (22) that are defective in Fe-S synthesis, for example. We found no difference in free iron (iron detectable by EPR at g ϭ 4.3) between wild type and sod1⌬ yeast in purified mitochondria. We thus conclude that the sod1⌬ mutation does not severely reduce iron levels in the mitochondria, nor does it have a global effect on the synthesis of Fe-S clusters. This conclusion is supported by the fact that respiration, which requires the action of Fe-S centers and cytochromes, is not decreased in sod1⌬ mutants. 3   During aerobic growth, superoxide is generated on the inner and outer surfaces of the inner mitochondrial membrane, as well as in the cytoplasm. In wild type cells, Sod1p in the IMS and cytoplasm intercepts superoxide generated in those locations, and the 4Fe-4S cluster enzymes Aco1p (aconitase), Lys4p, and Leu1p remain active. When Sod1p is absent (sod1⌬) the superoxide concentration increases in the IMS and cytosol. Superoxide anion (pK a ϭ 4.8) in the IMS near the inner membrane becomes protonated due to the local low pH created by the respiratory proton gradient. This uncharged form of superoxide is then able to diffuse across the inner membrane into the matrix where it deprotonates and inactivates Aco1p and Lys4p. Leu1p is inactivated by superoxide in the cytosol. In either case, Sod2p continues to intercept superoxide made by the respiratory chain at the inside (matrix) surface of the inner membrane, but it does not react with superoxide that enters the matrix by direct diffusion, either because it is not correctly positioned, because it is not present in sufficient quantity, or because it is product-inhibited at higher superoxide concentration. OM, outer mitochondrial membrane; IM, inner membrane. Inverted text is used to indicate Fe-S enzymes in an inactive state. Single-sided arrows indicate movement of a single species; regular arrows indicate a chemical reaction, and bars indicate inhibition. The pores that make the outer mitochondrial membrane nonspecifically permeable to small ions are indicated by gaps; there are no such pores in the inner membrane, and only uncharged hydrophobic molecules can penetrate into the matrix in the absence of a specific transporter protein. At high superoxide concentration and relatively low pH, spontaneous disproportionation can occur, as indicated by the dashed arrow. growth conditions used in these experiments, whole sod1⌬ cells show a 4 -5-fold increase in free iron, so the absence of such iron in mitochondria indicates that excess free iron is located elsewhere in the cell.
Thus we are left with the third possible explanation, the specific sensitivity of exposed 4Fe-4S clusters to superoxide, as the most likely explanation for the lysine and leucine auxotrophies. As mentioned above, understanding the protection of Leu1p by Sod1p is straightforward because both enzymes are present in the cytosol. It is more difficult to account for the cross-compartmental protection of Lys4p and aconitase by Sod1p.
We believe movement of superoxide into the matrix is likely in yeast based on our data and on the following considerations. The pK a of superoxide is 4.8, so that at physiological pH (around 7 in the cytosol) it will exist mainly as the anion. However, at the outer surface of the inner mitochondrial membrane, the local pH may be substantially lower due to the export of protons by respiration to form the pH gradient. In the absence of Sod1p, excess superoxide may become protonated in this local environment, diffuse across the inner membrane, and become trapped in the matrix where the pH is typically higher (above 7.5) (47,48). Such behavior has been documented in the case of acetate/acetic acid with isolated mitochondria (49,50).

In both pairs, O 2
Ϫ /HO 2 and CH 3 CO 2 Ϫ /CH 3 CO 2 H, the conjugate acid is lipid-soluble and can diffuse freely across membranes, whereas the conjugate base is anionic and relatively membrane-impermeant. In both cases, the pK a of the acid is ϳ4.8. This model is diagrammed in Fig. 7.
At this point a question arises as to why Sod2p, which is present in the matrix, does not protect Lys4p and aconitase in sod1⌬ yeast. It is possible that Sod2p is not present in the correct location or in sufficient quantity to successfully combat the influx of superoxide across the inner membrane of sod1⌬ yeast. In this case, CuZn-SOD activity outside of the matrix would be required to prevent an overwhelming inward flow of superoxide. In addition, it is known that Mn-SOD is subject to reversible product inhibition by H 2 O 2 , rendering it less active at high superoxide concentrations (51). This inhibition is quite severe in the human enzyme, less so in bacterial ones (52), and is apparently an evolutionarily conserved trait because a Mn-SOD mutant engineered to have reduced product inhibition (and thus be a more efficient dismutase) was strongly growth inhibitory when expressed in cultured cells (53). Such product inhibition in the yeast Mn-SOD could help explain its inability to compensate for loss of CuZn-SOD.
Apparently, superoxide originating from the matrix does not cause lysine auxotrophy, because it is not observed in sod2⌬ mutants. This fact remains especially puzzling because Lys4p activity is decreased in sod2⌬ mutants (Fig. 3B), although the level of lysine itself remains normal in these cells. The lysine level is greatly decreased in sod1⌬ cells. These data fit the observed phenotypes but do little to explain them. A number of observations support the idea that Mn-SOD is less important than CuZn-SOD in the overall protection of yeast growing in glucose media (2,3). The total amount of CuZn-SOD is 10 (2) to 50 times (54) greater than that of Mn-SOD in yeast. Furthermore, overexpression of Mn-SOD in the matrix of various cell types enhances their ability to combat oxidative stress (55)(56)(57) indicating that Mn-SOD is expressed at a level that is just sufficient to protect against normal superoxide levels. The production of lysine in sod2⌬ yeast cells despite the inactivation of Lys4p could be explained if cluster repair is more efficient in sod2⌬ than in sod1⌬ cells.
The methionine auxotrophy of sod1⌬ has been proposed to result from a depletion of NADPH in the cell by oxidative stress leading to a lowered synthesis of methionine because its biosynthetic pathway requires a great deal of NADPH (58,59). This model is supported by the observation that stimulation of NADPH production by overexpression of transketolase, an enzyme of the pentose phosphate pathway, rescues the methionine auxotrophy of the sod1⌬ mutant but not the lysine auxotrophy (59). The methionine auxotrophy is inducible in a wild type strain exposed to the redox-cycling drug paraquat (Fig. 2), confirming that this auxotrophy is in fact due to an increase in oxidative stress. However, it is interesting to note that sulfite reductase (YSiR), the enzyme that carries out the NADPH-dependent six-electron reduction of sulfite (SO 3 2Ϫ ) to sulfide (S 2Ϫ ) in the methionine biosynthetic pathway, contains a 4Fe-4S cluster (60,61), and it catalyzes the step at which methionine production is interrupted. Inactivation of this enzyme could result in a methionine auxotrophy directly and could lead to further toxicity by preventing the removal of sulfite, a toxin to which sod1⌬ mutants are particularly susceptible (58).
We conclude that the role of CuZn-SOD in S. cerevisiae lysine and leucine metabolism is to protect the 4Fe-4S cluster enzymes Lys4p and Leu1p from direct attack by superoxide on their Fe-S clusters. We find no indication that the primary defect is an alteration in mitochondrial iron metabolism. Our data also indicate that superoxide is able to pass through the mitochondrial inner membrane of yeast and damage Fe-S proteins in the matrix despite the presence of Mn-SOD, so that Sod1p plays an important role in protecting mitochondrial matrix proteins, as well as cytoplasmic proteins, from superoxidemediated damage.