Glutathione Degradation Is a Key Determinant of Glutathione Homeostasis*

Background: Intracellular concentration of glutathione, an essential sulfur compound, is tightly controlled. Results: In yeast, glutathione degradation is faster than previously published, and glutathione intracellular concentration does not affect its synthesis. Conclusion: Glutathione degradation is a key determinant of glutathione homeostasis. Significance: This work challenges notions on glutathione synthesis and degradation, which were considered as established. Glutathione (GSH) has several important functions in eukaryotic cells, and its intracellular concentration is tightly controlled. Combining mathematical models and 35S labeling, we analyzed Saccharomyces cerevisiae sulfur metabolism. This led us to the observation that GSH recycling is markedly faster than previously estimated. We set up additional in vivo assays and concluded that under standard conditions, GSH half-life is around 90 min. Sulfur starvation and growth with GSH as the sole sulfur source strongly increase GSH degradation, whereas cadmium (Cd2+) treatment inhibits GSH degradation. Whatever the condition tested, GSH is degraded by the cytosolic Dug complex (composed of the three subunits Dug1, Dug2, and Dug3) but not by the γ-glutamyl-transpeptidase, raising the question of the role of this enzyme. In vivo, both DUG2/3 mRNA levels and Dug activity are quickly induced by sulfur deprivation in a Met4-dependent manner. This suggests that Dug activity is mainly regulated at the transcriptional level. Finally, analysis of dug2Δ and dug3Δ mutant cells shows that GSH degradation activity strongly impacts on GSH intracellular concentration and that GSH intracellular concentration does not affect GSH synthesis rate. Altogether, our data led us to reconsider important aspects of GSH metabolism, challenging notions on GSH synthesis and GSH degradation that were considered as established.


Glutathione (GSH) has several important functions in eukaryotic cells, and its intracellular concentration is tightly controlled. Combining mathematical models and
S labeling, we analyzed Saccharomyces cerevisiae sulfur metabolism. This led us to the observation that GSH recycling is markedly faster than previously estimated. We set up additional in vivo assays and concluded that under standard conditions, GSH half-life is around 90 min. Sulfur starvation and growth with GSH as the sole sulfur source strongly increase GSH degradation, whereas cadmium (Cd 2؉ ) treatment inhibits GSH degradation. Whatever the condition tested, GSH is degraded by the cytosolic Dug complex (composed of the three subunits Dug1, Dug2, and Dug3) but not by the ␥-glutamyl-transpeptidase, raising the question of the role of this enzyme. In vivo, both DUG2/3 mRNA levels and Dug activity are quickly induced by sulfur deprivation in a Met4-dependent manner. This suggests that Dug activity is mainly regulated at the transcriptional level. Finally, analysis of dug2⌬ and dug3⌬ mutant cells shows that GSH degradation activity strongly impacts on GSH intracellular concentration and that GSH intracellular concentration does not affect GSH synthesis rate. Altogether, our data led us to reconsider important aspects of GSH metabolism, challenging notions on GSH synthesis and GSH degradation that were considered as established.
Glutathione (GSH) is a very important tripeptide present in quantity ranging from 1 to 10 mM in all cellular types. GSH is involved in a multitude of cellular processes, including detoxification of xenobiotics, antioxidant defense, proliferation, and apoptosis. Deficiency in GSH contributes to oxidative stress, which plays a key role in aging and in the development of many human diseases, including neurodegenerative diseases, liver diseases, diabetes, cystic fibrosis, AIDS, heart attack, and cancers. GSH is synthesized in the cytosol by two enzymes, the ␥-glutamylcysteine synthetase (␥-GCS), 3 considered as the rate-limiting enzyme of the pathway, and the glutathione synthetase (see Fig. 1).
In mammalian cells GSH degradation occurs exclusively in the extracellular space. ␥-Glutamyl-transpeptidase (␥-GT) cleaves GSH into cysteinylglycine (Cys-Gly) and a Glu residue that is transferred on another acceptor amino acid. Cys-Gly is further cleaved by a dipeptidase (for review see Ref. 1). GSH homeostasis is thus the result of its synthesis and utilization inside the cell, its export, and its degradation outside the cell, with a fast turnover (half-life of 2-3 h in the rat liver (2)). It is generally considered that the cellular pool of GSH is tightly controlled at the step of its synthesis through the regulation of ␥-GCS activity and Cys availability (for review, see Ref. 3), but little is known concerning regulations that could occur at the level of GSH export and degradation. In yeast, GSH is transported inside the cell by Hgt1 and localized in the plasma-membrane (4) and can be transported from the cytoplasm to the vacuole by Ycf1 (5) (see Fig. 1). Preliminary experiments have suggested that GSH turnover is slow (6,7). GSH was presumed to be degraded through the ␥-GT pathway, localized in the vacuole (see Fig. 1), at least under sulfur and nitrogen starvation (8,9), which are conditions strongly inducing ␥-GT activity. However, mutants deleted for the gene coding for ␥-GT, ECM38, are able to utilize GSH as an exogenous sulfur source (7). This observation led to the characterization of another degradation pathway, a cytosolic complex composed of three proteins, Dug1, Dug2 and Dug3 (see Fig. 1) (10). Dug2 and Dug3 are required to cleave the ␥-glutamyl linkage between Glu and Cys-Gly, and Dug1 is a dipeptidase able to cleave Cys-Gly (11). Mutants lacking any of these proteins (dug1⌬, dug2⌬, dug3⌬) cannot grow with GSH as sole sulfur source (10). However, the respective contributions of this pathway and of the ␥-GT path-* This work was supported by the Agence Nationale de la Recherche. way in GSH degradation under other growth conditions and under different stress conditions have not been studied.
GSH can be considered as an end-product of the sulfur metabolism. Yeast sulfur metabolism and its regulations have been extensively studied (for review see Refs. 12 and 13). One crucial actor of these regulations is the transcriptional activator Met4, which tightly controls the expression of most of the enzymes of this pathway (12,14). Met4 transcriptional activity is repressed by addition of repressive amounts of Met in the medium (12) and highly induced under conditions leading to sulfur deprivation (e.g. sulfur starvation (14) or chromate treatment (15)) or GSH demand (e.g. cadmium (Cd 2ϩ ) treatment Refs. 16 and 17).
In the course of our work on yeast sulfur metabolism, we developed quantitative techniques of 35 S labeling combined to mathematical models. The data inferred from these models suggested that GSH turnover would be higher than anticipated. We set up additional experiments and models to evaluate sulfur fluxes in this pathway and, more specifically, estimate GSH half-life. We confirmed the rapid turnover of intracellular GSH and found that whatever the condition used, the only pathway involved in GSH degradation is the Dug pathway. Moreover, we showed that GSH degradation activity strongly impacts on GSH intracellular concentration.
Growth Media-The medium used in this study is yeast nitrogen base minimal medium (20 g/liter glucose), modified as in Lafaye et al. (22) to control both sulfur (as sulfate, GSH or Met) and nitrogen (as ammonium, GSH or Glu) contents, and supplemented with Lys, Leu, Trp, His, adenine, and uracil, except when specified for strain S288c. Sulfur-free medium is minimal medium with 30 mM ammonium but without any added sulfur source. Nitrogen-free medium is minimal medium with 0.1 mM sulfate but without added ammonium. Standard medium is minimal medium with 0.1 mM sulfate and 30 mM ammonium. In this medium, W303-1A doubling-time is 120 min.
Determination of Intracellular Concentration of GSH-Metabolite extracts were prepared in perchloric acetic acid 0.1% as described (22). GSH concentration in the extracts was measured as described by Rahman et al. (23). GSH intracellular concentration was deduced from the average volume of a cell (4.5 ϫ 10 Ϫ14 liter) and the quantity of cells used to prepare the extract.
Metabolite Fraction (MF) and Protein Fraction (PF) Preparation-0.9 ml of a culture of 35 S-labeled cells in exponential phase were mixed to 0.9 ml of 86% cold ethanol (Ϫ60°C) and centrifuged. The cell pellet was washed in cold water, suspended in 100 l of cold water, heated for 5 min at 95°C, and centrifuged. The thus-obtained supernatant corresponds to the MF. The pellet, which corresponds to the PF, was suspended in 100 l of water.
Thin-layer Chromatography (TLC) Experiments-MF was prepared as above, except that the cells were suspended in 30 l instead of 100 l before heating at 95°C. 3 l of MF were mixed to 1 l of performic acid and incubated for 20 min at room temperature. 0.5 l of these samples as well as controls ([ 35 S]Cys, [ 35 S]GSH, [ 35 S]Met, PerkinElmer Life Sciences) were applied on cellulose plates TLC, where they migrated for ϳ35 h (5 migrations of ϳ7 h) using a butanol-1/acetic acid/water (90/ 25/15) solvent that allows the separation of Cys and GSH. The dried plates were quantified as previously described (24).
[ 35  Stoichiometric Models of Yeast Sulfur Metabolism-The model presented in Fig. 2 is detailed in supplemental information, part 1. It is based on the assumption of steady state: each concentration and metabolite flux is constant. Therefore, for each metabolite, the difference between synthesis fluxes and catabolism fluxes is the amount of metabolite required for cellular growth (supplemental Equation 3). From this, each "forward" flux (g, h, s, o, q, d, a, b, f U ) can be expressed as a combination of metabolite concentrations (considered as known; see the supplemental information), doubling-time (T, measured experimentally), and "reverse" fluxes (n, m, l, p, c, f URev ). Therefore, to find flux values compatible with our experimental data, we only had to explore six reverse fluxes and two additional parameters (percent Stable and percent Vac ; supplemental information, Part 1). The reverse fluxes were the main parameters adjusted to fit the experimental data, obtained using 35 S radiolabeling (see Fig. 3   imental data. The reverse fluxes were modified as described in the supplemental information to find the best fit of the experimental data. The selected fit is presented in Fig. 3 and discussed in the supplemental information, part 1.8, where the corresponding numerical values are also given. Determination of app T 1/2 GSH Using [ 35 S]GSH-Cells grown to early exponential phase in standard medium were labeled for 30 -40 min with [ 35 S]GSH (1 Ci/ml). At the end of the labeling, about 90% of the intracellular 35 S radioactivity was in the metabolites, the remaining 10% being associated to the proteins (supplemental Fig. S8). Culture medium was removed by centrifugation, and the cells were suspended in the same volume of a nonradioactive medium. At each time point, 0.9 ml of culture was collected, and MF and PF were prepared. Radioactivity of MF and PF was measured as previously described (24) and normalized to the total cellular radioactivity (MF ϩ PF), which was constant along the time course (no excretion of 35 S-labeled compounds). app T 1/2 GSH is the time for half-MF radioactivity (mostly GSH) to be transferred in PF.
Determination of Prot Cys /Meta Cys -Cells grown in standard medium to early exponential phase were transferred to the defined medium. After 15 min (H 2 O 2 ), 30 min (sulfur starvation), 60 min (nitrogen starvation, chromate, Glu, GSH, Met, H 2 O 2 ), 90 min (Cd 2ϩ ), or 120 min (GSH, Met, H 2 O 2 ) cells were labeled for 10 min with [ 35 S]Cys (10 -20 Ci/ml) and processed for MF and PF. In the case of exposure to H 2 O 2 , fresh H 2 O 2 was also added at times 40 and 80 min. Prot Cys /Meta Cys is the ratio of radioactivity measured in PF and MF.
Determination of GSH Synthesis Rate-Cells grown to early exponential phase in standard medium were labeled with [ 35 S]Cys for 10 min (50 Ci/ml) and processed for MF and PF. Analysis by TLC showed that MF radioactivity was mainly [ 35 S]GSH ([ 35 S]Cys stood for 6 -10% of the signal). For each strain, the radioactivity corresponding to [ 35 S]GSH, measured on TLC as described (24), was normalized to the radioactivity in PF.
Determination of T1 ⁄ 2 Using gsh1⌬ Mutant-gsh1⌬ cells were grown to early exponential phase in standard medium supplemented with 100 M GSH. Cells were pelleted and suspended in GSH-free medium. A constant volume of culture was collected at the indicated times for extract preparation and GSH determination. Values were plotted and fitted to an exponential function (supplemental Fig. S11), from which was determined app T 1/2 GSH ϭ cor T 1/2 GSH , the time for half-GSH to be converted into something else.
Quantitative Real-time-PCR Analysis-Around 5 ϫ 10 8 cells grown in the indicated condition were collected, washed in cold water, pelleted, and stored at Ϫ80°C. Total RNA was extracted as described (25). Expression of the target genes was determined using the ⌬⌬Ct method with ACT1 as the reference gene. The primers used are given in supplemental information, Part 11.

Stoichiometric Models of Yeast Sulfur Metabolism Suggest
Significant Recycling of GSH into Cys-In a precedent article, we built mathematical models of the yeast sulfur metabolism to optimize a new method of proteomic quantifications using radiolabeling (24). We pursued this work to develop a comprehensive representation of the yeast sulfur metabolism and evaluate the sulfur fluxes in its different parts. The model of the sulfur pathway that we used to simulate [ 35 Fig. 2. It corresponds to growth on sulfate as unique sulfur source (standard condition). Thanks to their high specific radioactivity, 35 S-labeled compounds can be used in very small amounts as tracers without the risk of affecting yeast metabolism. In this model GSH is divided into two stocks; one that can be recycled, with the flux p, and one (called "vacuolar" for simplicity) that is not recycled. Similarly, proteins are divided into two stocks, one that can be recycled and one that is stable (Fig. 2). Under the assumption of steady state, each forward or fixed flux can be expressed as a combination of reverse fluxes, concentrations, and doublingtime (see "Experimental Procedures" and supplemental information, Part 1.4). Therefore, to find flux values compatible with our experimental data, we only had to explore eight parameters ("Experimental Procedures" and supplemental information, Part 1). We fitted two kinds of experimental data: (i) kinetics of radioactivity distribution between the MF and the PF (for  Fig. 3. The best fit we found required the flux p, from GSH to Cys (Fig. 2), to be equal to the minimal flux required for synthesis of GSH from ␥-glutamylcysteine. We were surprised that the best fit was only obtained with a high recycling of GSH into Cys. As this observation was at odds with previous results (6, 7), we refined this work. We labeled the cells with [ 35 S]Cys and [ 35 S]GSH for, respectively, 30 and 45 min, put them at time t ϭ 0 in nonradioactive medium, and measured the radioactivity in MF and PF (see below for details). Again, the flux values found above were in good accordance with these experiments (supplemental Fig. S5). Thus, our model strongly suggests that the time required for converting half the concen- tration of total GSH into Cys ( cor T 1/2 GSH ) is around 90 min (supplemental information, Part1).

S]Met, [ 35 S]Cys, or [ 35 S]GSH labeling is presented in
Determination of GSH Half-lives-The half-life of a molecule is usually defined as the time required to degrade 50% of the initial amount of this molecule (e.g. cor T 1/2 GSH for GSH; see above). For practical reasons, we also use in this work the notion of apparent half-life of GSH ( app T 1/2 GSH ), which we define as the time required to incorporate into proteins 50% of the sulfur atoms of the initial pool of GSH. This conversion requires GSH to be first degraded into Cys, from which it can either be incorporated into proteins, into cystathionine, or converted back into GSH (Fig. 2). To measure app T 1/2 GSH , cells were labeled with [ 35 S]GSH, washed, and suspended in nonradioactive medium. A constant volume of culture, corresponding to a constant amount of radioactivity, was harvested at each time point, and the amounts of intracellular 35 S in MF and PF were quantified (Fig. 4). We found app T 1/2 GSH to be around 174 min (  (Fig. 1). We evaluated the importance of Cys back-conversion into GSH using [ 35 S]Cys labeling ( Table 1). The Prot Cys /Meta Cys ratio thus measured gives an estimate of the relative importance of the fluxes coming from Cys to proteins and to other sulfur-containing metabolites. Using these additional data in small mathematical models, we calculated f GSH , defined as the fraction of GSH converted into Cys in 1 min  (Table 1, supplemental information, Part 4). If we consider that the intracellular concentration of GSH is constant, which is the case under standard conditions, we can calculate the corrected half-life of GSH ( cor T 1/2 GSH ) as cor T 1/2 GSH ϭ 0.5/f GSH . As expected, cor T 1/2 GSH (63 min) is markedly shorter than app T 1/2 GSH (174 min). As previous studies concluded that GSH is highly stable (6, 7), we tried to confirm our results by a different method (7). It consists of using a gsh1⌬ mutant unable to synthesize GSH in monitoring the pool of intracellular GSH when cells are transferred to a GSH-free medium. In the case of stability, the amount of intracellular GSH for a constant volume of culture should be constant. As shown in supplemental Fig. S11, the amount of GSH decreases exponentially with a half-life (corresponding to both app T 1/2 GSH and cor T 1/2 GSH in this case) around 92 min. This result, obtained with a mutant strain, is in good accordance with the cor T 1/2 GSH calculated for the wild type (WT) strain using radioactivity and confirms that GSH degradation is higher than previously thought.
GSH Turnover Depends on Physiological State of Cell-We measured app T 1/2 GSH in S288c and BY4742, two strains with a genetic background different from W303-1A, and found that app T 1/2 GSH was slightly shorter in S288c and BY4742 compared with W303-1A (around 135 min, supplemental information, Part 8). Sulfate concentration of the culture medium seems to have no effect on app T 1/2 GSH (supplemental information, Part 8). We pursued the study with the W303-1A strain and tested different conditions that were expected to increase GSH degradation: sulfur starvation (8), nitrogen starvation (9,26), growth with GSH as sole sulfur source (7), growth with Glu as nitrogen source (6), and chromate treatment (15). We also tested Cd 2ϩ treatment (22), the effect of adding repressive amounts of Met to the culture, and exposure to H 2 O 2 . app T 1/2 GSH measured under these conditions are reported in Table 1. We also measured Prot Cys /Meta Cys to extrapolate f GSH and cor T 1/2 GSH (Table 1, details are in supplemental information, Part 4). A notable result is that GSH degradation is strongly stimulated by sulfur starvation ( app T 1/2 GSH around 32 min, Table 1 and Fig. 4). This is consistent with the notion that GSH is an endogenous sulfur supply during sulfur starvation (8). The app T 1/2 GSH of 57 min when GSH is used as the sole sulfur source reflects an important activity of GSH degradation. When cells are treated with chromate, app T 1/2 GSH is also short (around 71 min), in accordance with the observation that chromate causes a depletion of sulfur metabolites (15). Supplementation of repressive amounts of Met, exposure to Cd 2ϩ , or exposure to H 2 O 2 , each, block 35 S incorporation into proteins ( app T 1/2 GSH Ͼ 500 min, Table 1). This could be due to the important conversion of [ 35 S]Cys into [ 35 S]GSH, as under these conditions only 11% (H 2 O 2 ), 14% (Cd 2ϩ ), and 18% (Met) of the sulfur flux coming from Cys is converted into proteins or to a real inhibition of GSH degradation. To directly detect such an inhibition, we used a gsh1⌬ strain and confirmed that exposure to Cd 2ϩ or Met markedly reduces GSH degradation (supplemental information, Part 6),  whereas H 2 O 2 has no measurable effect on GSH recycling (supplemental information, Part 7). When cells are starved for nitrogen or when Glu is used as a unique source of nitrogen, app T 1/2 GSH is longer than under standard conditions ( Table 1). The high app T 1/2 GSH observed under nitrogen starvation contrasts with previous results, suggesting that under this condition, intracellular GSH could be used as a nitrogen source (9,26). We decided to carefully re-examine this question in the WT strains used in this study, W303-1A, BY4742, and S288c. We found that neither GSH (10 mM) nor glutathione disulfide (GSSG, 5 mM) was able to sustain growth of these strains, whereas Glu (5 to 10 mM) allowed it (supplemental information, Part 10). Therefore, in the three tested genetic backgrounds, GSH cannot be used as a unique nitrogen source. As Glu is the first amino acid to be released when GSH is degraded (Fig. 1), this result shows that GSH is not degraded at a sufficient rate to provide enough nitrogen for growth. This is not surprising as, in yeast, nitrogen requirement is about 30-fold higher than sulfur requirement. To complete this study, we measured GSH intracellular concentration in W303-1A cells grown in nitrogen-free medium and observed that this concentration is significantly increased ( Table 2). This result is consistent with a recent study in Schizosaccharomyces pombe (27) but contrasts with other data in Saccharomyces cerevisiae (9).
GSH Is Degraded via Dug Pathway-Two pathways of GSH degradation have been described in yeast, the ␥-GT pathway and the Dug pathway (Fig. 1). To evaluate the relative importance of these pathways, we built deletion mutants ecm38⌬, dug1⌬, dug2⌬, and dug3⌬. In standard medium, growth of the mutant strains was indistinguishable from growth of W303-1A, the isogenic WT strain. Consistent with previous results (7), dug1⌬, dug2⌬, and dug3⌬ (but not ecm38⌬) were unable to grow using GSH or GSSG as unique sulfur source. We compared their ability to incorporate the sulfur atoms of 35 S-labeled GSH in proteins. The test was performed under three conditions; (i) growth in standard medium, (ii) growth with GSH as unique sulfur source, and (iii) sulfur starvation (Fig. 5). Notably, synthesis of 35 S-labeled proteins was not affected in the ecm38⌬ strain, suggesting that the ␥-GT pathway is not involved in GSH turnover in any of the three tested conditions. In contrast, incorporation of GSH sulfur atoms into proteins was markedly reduced in the dug2⌬ and dug3⌬ mutants in all tested conditions. We checked that this decrease was not due to a reduced protein synthesis by measuring the Prot Cys /Meta Cys ratio under the same conditions (Fig. 5). The Dug pathway is thus of major importance for GSH degradation under these three conditions. The dug1⌬ mutant presented a less marked defect than dug2⌬ and dug3⌬. This intermediate pattern suggests that other peptidases can complement Dug1 absence.
Regulation of GSH Degradation-As above, transferring cells to sulfur-free medium leads to a strong and rapid activation of GSH degradation. This suggests a tight control of the activity and/or amounts of the Dug enzymes. As this may involve a very fast transcriptional induction, we measured the transcript levels of DUG1, DUG2, and DUG3 by quantitative real-time-PCR analyses. ECM38 was also analyzed, and ACT1 was used as an internal standard. Cells were collected 5, 10, and 30 min after transfer to the sulfur-free medium. Whereas DUG1 and ECM38 mRNA levels did not change significantly, DUG2 and DUG3 transcript levels increased rapidly after the onset of sulfur deprivation (Fig. 6A). On the contrary, exposure to repressive amounts of Met (0.5 mM) repressed DUG2 and DUG3 expression. Notably, induction and repression were more pronounced for DUG3 than for DUG2. To test whether these inductions were controlled by Met4, the transcriptional activator of the sulfur pathway, we analyzed DUG2 and DUG3 mRNA levels in a met4⌬ mutant. W303-1A and met4⌬ cells were grown using 50 M Met as a unique sulfur source (for growth of the mutant) and transferred to sulfur-free medium, and cells were collected 10 and 30 min after the transfer. The results presented in Fig. 6B show that DUG2 and DUG3 inductions are dependent of the presence of Met4.
We also observed marked inductions of DUG2 and very high inductions of DUG3 when cells were treated with chromate for 30 min or grown with GSH as unique sulfur source. Under these conditions, DUG1 and ECM38 were not significantly induced (Fig. 6C). When cells were treated with Cd 2ϩ for 1 h, DUG1, DUG2, DUG3, and ECM38 all appear to be induced. However, these apparent inductions may be a consequence of using ACT1 as the internal control. In fact, in the presence of Cd 2ϩ , ACT1 transcription is slightly repressed. Therefore, only DUG2 and DUG3 induction may be relevant under this condition. Finally,

Yeast Glutathione Homeostasis Revisited
FEBRUARY 10, 2012 • VOLUME 287 • NUMBER 7 we measured GSH turnover in a met4⌬ mutant; the mutant and its isogenic WT strain W303-1A were labeled with [ 35 S]GSH and transferred to sulfur-free medium to induce sulfur starva-tion, and MF and PF radioactivity was measured. As shown in Fig. 7, 35 S incorporation in proteins is almost totally blocked in the mutant strain, although the basal level of ECM38, DUG1, DUG2, and DUG3 mRNAs were similar in both strains before the shift (Fig. 6B, 0 min). Altogether, these data suggest that the increased degradation of GSH during sulfur starvation is mainly controlled by the transcriptional activation of DUG2 and DUG3. Dug Complex Contributes to GSH Homeostasis-The above results strongly suggest that GSH turnover is an important feature of GSH metabolism. In particular, GSH degradation could affect the intracellular content of GSH. We measured GSH concentration in WT, dug2⌬, and dug3⌬ strains and found that under standard conditions, the amount of intracellular GSH in the degradation mutants is approximately twice that in the WT strain (Table 2). Interestingly, this is the concentration we find in our mathematical model by simply suppressing GSH degradation (flux p ϭ 0, Fig. 2) and adapting GSH concentration to satisfy the steady state hypothesis. Indeed, the best fit of WT experimental data with our mathematical model was obtained with a flux from ␥-glutamylcysteine to GSH (Fig. 2) that was twice the minimal flux required for growth of the GSH pool (above and supplemental information, Part 1). Transposing the metabolic model to a mutant strain without changing any flux in the GSH cycle (except the one put to 0) postulates that changes in GSH concentration do not affect the GSH synthesis rate. This is not trivial because in vitro experiments have shown that GSH is an inhibitor of ␥-GCS in organisms as diverse as Escherichia coli (28), rats (28,29), Arabidopsis thaliana (30), and yeasts (31,32), and it is generally admitted that S. cerevisiae ␥-GCS is also subjected to such a feedback inhibition in vivo (33).
We evaluated GSH synthesis rate in WT, dug2⌬, and dug3⌬ strains under standard conditions. For this, cells were labeled with [ 35 S]Cys for 10 min, and the amount of radioactivity in GSH was measured on TLC plates and, to allow comparison between different strains, normalized to the 35 S incorporated into proteins during the same time. The values obtained were 0.67, 0.72, and 0.72 for the WT, dug2⌬, and dug3⌬ strains, respectively, with relative S.D. among two independent experiments of 12, 7, and 7% respectively. They show that GSH synthesis was not significantly changed among the three strains. In particular, doubling the cellular content of GSH had no significant effect on GSH synthesis rate in vivo.

DISCUSSION
The stoichiometric model of S. cerevisiae sulfur metabolism presented in this work led us to reappraise the values of yeast sulfur fluxes, especially GSH recycling into Cys. In this model, fed with in vivo 35 S labeling data, half of the synthesized GSH is used for cellular growth, and the other half is degraded. This result is corroborated by the observation that a mutant strain impaired for GSH recycling has twice as much GSH as a WT strain and also by the estimation of GSH half-life by two independent methods (giving, respectively, 63 and 92 min), which is in good accordance with the value inferred from our model (87 min). A major result of this work is thus that under standard conditions, GSH is degraded at a high rate. Considering that

Yeast Glutathione Homeostasis Revisited
two ATP molecules are consumed at each synthesis of one GSH molecule from Cys, this "waste" of cellular energy may be surprising. It is actually in accordance with studies showing that under optimal growth conditions, microbial cells produce energy in excess, leading to the concept of "energy spilling" (34). Our results are at odds with the general belief, based on [ 14 C]GSH labeling experiments, that GSH is very stable in yeast (6). Because detailed experimental data of this previous study were not provided, it is difficult to comment on this discrepancy.
A major interest of the [ 35 S]GSH-based assay developed in this work is that it allows the determination of GSH apparent half-life in any physiological condition and mutant as long as GSH can enter the cell. Without any additional measurements, this assay shows that GSH recycling is faster than previously published. Combined with [ 35 S]Cys-labeling experiments and small mathematical models, it allows the estimate of GSH degradation rate. However, in some cases (H 2 O 2 or Cd 2ϩ treatment, Met supplementation) our assay can be ineffective due to a strong recycling of Cys into GSH under the considered conditions. Despite this shortcoming, our [ 35 S]GSH assay represents an important improvement for the study of GSH turnover. An alternative method consists of monitoring GSH intracellular concentration in gsh1⌬ mutant cells transferred to a GSH-free medium (7). Although this method has the drawback of requiring a specific genetic background, we tested it and confirmed a good consistency with our 35 S tracer method. This shows that GSH degradation is not significantly altered in a strain unable to synthesize GSH and when GSH intracellular concentration drops. Therefore, the intracellular concentration of GSH does not seem to play a major role in GSH degradation.
We used our 35 S tracer assay to analyze degradation mutants under different growth conditions. We established that, whatever the physiological condition tested, GSH is degraded through the Dug pathway. Degradation of GSH is almost completely blocked in dug2⌬ and dug3⌬ mutants, indicating that both Dug2 and Dug3 are essential for this process. However, GSH turnover is only partly impaired in the dug1⌬ mutant. This result suggests that another peptidase(s) can partially complement the absence of Dug1. Consistent with this hypothesis, the defect observed in the dug1⌬ mutant is more marked under sulfur starvation and GSH growth conditions, which are conditions with higher rates of GSH degradation, than under standard conditions. Although the ability of the dug1⌬ strain to degrade GSH is not completely impaired, the mutant cannot grow using GSH as unique sulfur source (10). This apparent discrepancy could be due to the accumulation of Cys-Gly, a putative toxic compound (35,36), in the dug1⌬ mutant strain (11).
GSH intracellular concentration results from a balance between its synthesis, its degradation, and cellular growth. Previous studies have shown that increasing GSH synthesis through overexpression of GSH1 augments the GSH pool by 50 -66% (33,37). This work shows that blocking GSH degradation by deletion of DUG2 or DUG3 augments GSH cellular content by 100%. Both genetic data thus confirm that synthesis and degradation have a major impact on GSH intracellular con-centration. While studies generally focus on GSH synthesis, regulation of GSH homeostasis through degradation has never been explored in yeast. We show here that sulfur starvation and growth with GSH as a unique sulfur source highly activate GSH degradation. These regulations are physiologically relevant, as they provide the cell with sulfur. Conversely, Cd 2ϩ treatment or Met supplementation slow GSH turnover. In response to Cd 2ϩ , GSH intracellular content increases, probably to optimize Cd 2ϩ detoxification (22). This increase was considered to be the consequence of an enhanced synthesis of GSH (22,38). Our present study shows that it is also the consequence of a decreased degradation of GSH.
Sulfur starvation stimulates GSH degradation by the Dug2/3 enzyme. Because yeast sulfur metabolism is mainly regulated at the mRNA level (12,14), we examined the hypothesis of a transcriptional regulation of this activation. DUG2 and DUG3 mRNAs are highly and quickly induced by sulfur starvation, and both the induction of DUG2 and DUG3 transcription and the activation of GSH degradation by sulfur starvation are dependent on the presence of the transcriptional activator Met4. The data thus suggest that GSH-increased degradation in response to sulfur starvation is mainly controlled by the level of Dug2/3 complex. Similarly, we show that Met supplementation, which is known to repress Met4 activity (12), strongly decreases DUG2 and DUG3 mRNA levels and reduces GSH degradation. Therefore, the effect of Met supplementation on GSH degradation may be merely a consequence of the Met4-dependent regulation of DUG2 and DUG3 expression.
Of note, the transcriptional control observed under sulfur starvation or upon Met supplementation is not physiologically relevant in the case of Cd 2ϩ exposure. Indeed, Cd 2ϩ strongly induces DUG2 and DUG3 transcription, consistent with Met4 activation by exposure to Cd 2ϩ (16,17), but decreases GSH degradation. Different hypotheses may account for this discrepancy; (i) the Dug complex may be inhibited by the metal, (ii) when chelated to Cd 2ϩ , GSH (in the form of Cd(GS) 2 ) may not be a substrate of the Dug complex, and (iii) Cd(GS) 2 complexes may be rapidly translocated into the vacuole where they cannot be processed by the cytosolic enzyme Dug2/3. The development of an in vitro test of the Dug complex activity would help to identify mechanisms involved in the response to Cd 2ϩ exposure.
␥-GT (coded by ECM38) is highly expressed under different conditions, including nitrogen starvation (9), sulfur starvation (8), and growth with Glu as nitrogen source (6). Therefore, it was thought to be implicated in the utilization of GSH under these conditions. This work shows that whatever the condition (standard condition, sulfur starvation, and nitrogen starvation, growth with GSH as the sole sulfur source, or growth with Glu as nitrogen source), ␥-GT is never required for GSH degradation. In addition, we show that GSH cannot be used as a nitrogen source, and GSH degradation is low under nitrogen starvation or growth with Glu as unique nitrogen source despite the strong ␥-GT activity of protein extracts prepared under these conditions (6,9). We thus have no evidence that ␥-GT is involved in GSH degradation in vivo. Although GSH is a substrate of ␥-GT in vitro (39), our data suggest that free GSH is not Yeast Glutathione Homeostasis Revisited FEBRUARY 10, 2012 • VOLUME 287 • NUMBER 7 JOURNAL OF BIOLOGICAL CHEMISTRY 4559 a substrate of ␥-GT in vivo, raising the question of the real function of ␥-GT in yeast.
According to recent publications (40,41), ␥-GT plays a critical role in the detoxification of monochlorobimane. This electrophilic xenobiotic is first conjugated to GSH, giving glutathione S-bimane. Glutathione S-bimane is then catabolized to Cys-bimane through two pathways, both requiring ␥-GT for the cleavage of a ␥-bond (cleavage of glutathione S-bimane into Cys-Gly-bimane and cleavage of ␥-Glu-Cys-bimane) (41). This suggests that the physiological substrates of the ␥-GT are the GSH conjugates. The vacuolar localization of both ␥-GT and GSH conjugates (transported with high affinity by Ycf1) is consistent with such a function. Therefore, in yeast, GSH would be degraded in the cytosol mainly by the Dug pathway, and the GSH conjugates would be metabolized in the vacuole by ␥-GT. In mammalian cells, GSH and GSH conjugated are described as being exclusively catabolized outside the cell by ␥-GT (1). Mammals do not seem to possess true orthologs of Dug2 and Dug3 components (10). However, a Dug1 ortholog, CNDP2, can complement a dug1⌬ yeast mutant in regard to the Cys-Gly peptidase function (11). Such a function in mammals remains to be confirmed.
An important result of this work is that in vivo, a 2-fold increase of GSH does not seem to affect GSH synthesis, although based on in vitro experiments, GSH is considered as an inhibitor of ␥-GCS in many organisms (28 -30), including S. cerevisiae (32). In yeast, the K i value for GSH, 2.1 mM (32), is close to the intracellular concentration of GSH (Refs. 15 and 22) and this work). However, our data suggest that either this regulation does not occur in S. cerevisiae or this inhibition does not impact on GSH synthesis rate. Such a possibility is well explained by the metabolic control theory (42,43). In particular, in a metabolic pathway, decreasing the activity of one enzyme may lead to an increase of the precursor metabolites, thus restoring a close-to-normal metabolic flux (44). Therefore, inhibition of ␥-GCS may increase Cys concentration and restore the normal rate of GSH synthesis. Another explanation could simply be the fact that GSH is a competitive inhibitor of ␥-GCS with respect to Glu (32). Considering that the K m for Glu is 1.2 mM (32) and that in yeast Glu intracellular concentration is more than 20 mM (45,46), doubling GSH concentration from 3 to 6 mM may not significantly decrease ␥-GCS activity. We thus suggest that the notion that ␥-GCS would be the "rate-limiting enzyme" of the GSH synthesis pathway, a notion mainly based on in vitro data (3), should be reconsidered, at least in yeast, in the light of in vivo data. Although ␥-GCS activity is undeniably a factor influencing GSH synthesis, there are certainly other major determinants in this control, one of them being the Cys availability (47).