Glutathione revisited: a vital function in iron metabolism and ancillary role in thiol‐redox control

Hgt1 is an exclusive GSH and GSSG transporter (Bourbouloux et al, 2000). Cells that overexpressed HGT1 were unable to grow on minimal medium plates that contained GSH at concentrations as low as 20 μM (Figure 1A). Similarly in liquid medium, HGT1 cells were growth inhibited at GSH concentrations of 50 μM and above (Figure 1B), but then resumed growth after a few hours. These cells died massively within 4 h after GSH addition (Figure 1C). Lack of a dose effect on cell killing at GSH concentration above 100 μM probably reflects saturation of HGT1, the Km of which for GSH is 50 μM (Bourbouloux et al, 2000). Saturation of GSH uptake also explains the lag period before growth resumption to be longer at 200 μM than at 100 μM GSH, as GSH remains longer in the medium under the former condition. In the absence of GSH, HGT1 cells had a wild‐type GSH content (7.5 mM); however, on adding GSH (100 μM), this concentration peaked at ∼70 mM after 30 min (Figure 1D; Table I). GSSG levels paralleled the increase in GSH levels peaking at 1.7 mM from a baseline value of 0.2 mM. The increases in GSH and GSSG resulted in a significant decrease of the GSH redox potential (E_GSH) from approximately −220 mV to approximately −250 mV at 30 min. GSH levels then decreased after an hour to normalize at 8 h (Figure 1D; Table I). Such decrease, which explains the growth resumption of HGT1 cells in liquid medium and only partial lethality (see Figure 1B and C), is not a consequence of GSH efflux, as the GSH concentration in the culture medium gradually decreased (Supplementary Figure S1A), but of degradation by the γ‐glutamyl transpeptidase (γ‐GT)‐independent Dug1–Dug2–Dug3 pathway (Ganguli et al, 2007; Kaur et al, 2009), as the corresponding mutants maintained elevated GSH levels for a longer period and were sensitive to lower GSH concentrations (Supplementary Figure S1B and C). The decreased GSH tolerance of the Dug pathway mutants indicates that GSH toxicity is caused by GSH itself and not by a GSH catabolite. Fast correction of GSH levels also indicates a fast turnover rate of GSH.

To search for causes of GSH toxicity, we established the mRNA profiles of HGT1 cells cultured with 50 μM GSH during 5, 30 and 240 min. The 5‐ and 30‐min transcript profiles had similarities, but were distinct from the 4‐h transcript profile that was not further considered as reflecting a non‐specific stress response (Supplementary Figure S2). The 5‐min response was modest (70 genes, Supplementary Table S1), and its significance was best appreciated in view of the 30‐min response that was concurrent to the GSH peak. The 30‐min response was large (375 genes, Supplementary Table S2) and included ∼50% of the genes regulated at 5 min (Supplementary Figure S3). Functional clustering revealed two major components in this response (Figure 2). The first one was highly similar to the response of iron‐starved cells (Shakoury‐Elizeh et al, 2004; Puig et al, 2005), and even more similar to that of the atm1Δ mutant (Hausmann et al, 2008). ATM1 encodes a mitochondrial inner‐membrane ABC transporter required for extra‐mitochondrial ISC assembly (Hausmann et al, 2008). The HGT1 cell iron response comprised 25 genes of the iron regulon (Supplementary Table S2; Yun et al, 2000; Foury and Talibi, 2001; Rutherford et al, 2003) controlled by the Aft1 and Aft2 transcription factors, which regulates the uptake, compartmentalization and use of iron. Many repressed genes identified as part of the iron starvation response were also present here. These included genes encoding Fe–S‐ and haeme‐containing proteins of mitochondrial respiration and the tricarboxylic acid cycle and of other pathways, and haeme, ergosterol, biotin and amino‐acid/nitrogen biosynthetic enzymes. The second major component of the response to high GSH levels (115 genes, ∼30%) was a typical unfolded protein response (UPR) that is regulated by the Ire1 kinase and Hac1 transcription factor (Figure 2; Supplementary Table S2). It encompassed all major UPR gene categories induced by the model ER stress agents DTT and tunicamycin (Travers et al, 2000; Supplementary Figure S4A and B). Induction of several cytoplasmic chaperone genes could also be considered as satellite of the UPR, as the encoded proteins are known to help cells cope with ER stress (Liu and Chang, 2008). Repression of membrane proteins and proteins involved in ribosome biogenesis and translation might also be considered as part of the ER stress response, as shown in C. albicans (Wimalasena et al, 2008).

The UPR and iron starvation‐like responses were independent of each other, as the latter persisted in HGT1 ire1Δ cells that had otherwise lost the UPR (Supplementary Figure S4C and D), but their concurrence probably caused conflicting gene regulation. For instance, haeme and ergosterol biosynthesis genes that are found repressed here are upregulated during the UPR.

We explored the link between toxic GSH levels and iron metabolism. Addition of 100 μM GSH to HGT1 cells induced the Aft1 target‐gene FET3, which peaked between 30 and 60 min (Figure 3A), confirming the microarray data. FET3 was maximally induced at concentrations of GSH as low as 10 μM (Figure 3B). Aft1p is cytoplasmic in iron‐replete cells, and accumulates in the nucleus in iron‐deprived cells (Yamaguchi‐Iwai et al, 2002). In accordance with FET3 expression, in HGT1 cells, GFP‐Aft1 became nuclear from a diffuse cellular pattern 15 min after addition of GSH (Figure 3C), and at concentrations of 10 μM and above (Supplementary Figure S5A). Although Aft1 was activated, iron was not limiting, as GSH‐grown HGT1 cells had a slightly higher iron content than wild‐type cells (Supplementary Figure S5B), and excess iron >100 μM worsened GSH toxicity (Figure 3D). Reciprocally, inactivating high‐affinity iron uptake by deletion of FET3, which encodes the iron oxidase of the high‐affinity iron uptake system, or AFT1, or AFT1 and AFT2 did not decrease GSH toxicity (Supplementary Figure S5C). Aft1 senses iron through the activity of mitochondrial ISC assembly (Chen et al, 2004) and mitochondrial ISC export systems, which are also required for extra‐mitochondrial Fe–S protein assembly (Rutherford et al, 2005). In HGT1 cells, the activity of the mitochondrial Fe–S enzyme aconitase (Aco1) was not altered by GSH addition, in contrast to that of the cytoplasmic Fe–S enzyme isopropylmalate isomerase (Leu1), which decreased with a kinetics that inversely paralleled the GSH level changes (Figure 3E; see Figure 1D). Aco1 and Leu1 protein levels remained unchanged under all conditions (Figure 3F).

GSH toxic levels thus promptly trigger Aft1 activation and alter extra‐mitochondrial but not mitochondrial Fe–S enzyme activity, both events occurring with time courses suggestive of a cause‐and‐effect relationship. Such alteration in iron metabolism may contribute to the lethality of high GSH levels, as several Fe–S proteins are essential.

We explored the link between GSH toxic levels and the UPR. In HGT1 cells, the Hac1 target‐gene Kar2 was induced 5 min after GSH addition (Figure 4A), confirming the microarray data. Induction initially paralleled the increase in GSH levels, but then remained high for up to 4 h while GSH levels had corrected (see Figure 1D). KAR2 induction occurred at GSH concentrations as low as 5 μM and reached a plateau at 50 μM (Figure 4B). As a thiol reductant, GSH might have induced the UPR by interfering with oxidative protein folding. We explored the impact of GSH on the Ero1 and Pdi1 redox states. In untreated HGT1 cells, Ero1 was fully oxidized and Pdi1 was almost fully oxidized. Addition of GSH caused a dose‐dependent reduction of these proteins with a kinetics that paralleled the UPR, occurring fast and lasting for up to 1–3 h after GSH levels had corrected, depending on the GSH amount added (Figure 4C and D). Such kinetics suggested that GSH rapidly diffused into the ER and was trapped there, perpetuating Ero1 cycling by concomitantly reducing its negative regulatory disulphide and providing an unlimited reduced substrate supply. The presence of an actively cycling Ero1 was reflected by the GSH‐dependent increase in the rate of GSH oxidation (Figure 4E), and by production of ROS (Figure 4F, upper) that are expected to be generated on reduction of oxygen by active Ero1 (Gross et al, 2006). An hour exposure of HGT1 cells to 100 μM GSH caused the ROS‐specific probe rhodamine to fluoresce with a signal comparable to that produced by a 15‐min exposure to 400 μM H₂O₂. The nature of this ROS signal was H₂O₂, as it was detected by the H₂O₂‐specific PF1 probe (Supplementary Figure S6A), and its cellular origin the ER, as suggested its disappearance upon UPR prevention by Ire1 kinase inactivation (Figure 4F, lower). A further indication of Ero1‐active cycling was the full reduction of Ero1, but not of Pdi1 (see Figure 4C and D), which indicated a balance between Ero1‐mediated Pdi1 oxidation and its reduction by GSH. An active Ero1 was needed to counter the GSH‐reductive load as HGT1 cells carrying the Ero1‐defective ero1‐1 allele was intolerant to very low GSH concentrations (Supplementary Figure S6B).

We tested the effect of GSH on ER secretory transport by following the maturation of carboxypeptidase Y (Cpy), a protein requiring five disulphides for proper folding and ER to vacuole transport (Figure 4G). In HGT1 cells, a subtoxic concentration of GSH (20 μM) delayed Cpy maturation and a toxic one (50 μM) totally blocked this process, with accumulation of the ER form (p) and lack of appearance of the matured vacuolar form (m). The impact of GSH at 20 μM was equivalent to that of DTT at 5 mM, a dose–response difference that further supported the notion that GSH but not DTT accumulates into the ER at very high levels. Finally, inactivation of the UPR by IRE1 deletion (ire1Δ) not only eliminated H₂O₂ production (see Figure 4F, lower), but also made HGT1 cells intolerant to very low and otherwise non‐toxic GSH concentrations (Figure 4H; Supplementary Figure S6C). The latter data indicate that the impact of toxic levels of GSH on ER secretion contributes to its lethality, and that this lethality is not caused by the H₂O₂ produced in this condition.

When grown in medium devoid of GSH, yeast cells lacking GSH1 (gsh1Δ) become growth arrested after 8–10 divisions by near‐complete GSH depletion (Spector et al, 2001); they activate Aft1 and are defective in extra‐mitochondrial but not mitochondrial ISC maturation (Sipos et al, 2002; Rutherford et al, 2005; Figure 5A and C; Supplementary Figure S7A). However, it is not clear whether it is this defect or a defect in thiol‐redox control that causes cell inviability by GSH depletion. To answer this question, we established the transcript profiles of gsh1Δ cells withdrawn of GSH for three or six divisions, the latter sample having extremely low GSH cellular levels (60 μM∼0.8% of WT) (Table I). The two GSH‐depleted samples had almost identical mRNA profiles (Supplementary Figure S2), and were thus analysed as replicates of the same experiment. The observed response was significant (178 genes), and highly similar to that of the atm1 mutant (Hausmann et al, 2008; Figure 2; Supplementary Table S3). In fact, with few exceptions, most genes responding to GSH depletion were categorized as part of an iron starvation‐like response. A few gene targets of the H₂O₂‐reponsive Yap1 transcription factor were also moderately induced, as noticed elsewhere (Wheeler et al, 2003), but their very limited number, several of which are in fact also controlled by Aft1 (see Supplementary Table S3), and lack of perceptible Yap1 oxidative activation (Figure 5B) indicate at best a very minor activation of Yap1. Interestingly, HGT1 was also induced, which fulfilled the demand for increased GSH uptake.

The iron starvation‐like response as the only significant genome‐wide consequence of GSH depletion did not befit ascribing a thiol‐redox control function at the origin of the GSH requirement for viability, which was previously established by the partial growth rescue of GSH‐depleted cells by DTT (Sipos et al, 2002). We enquired of the effects of iron on GSH‐depleted cells. Ferric iron (100 μM) partially decreased the high FET3 expression of GSH‐depleted cells within 15 min, while GSH totally corrected it (Figure 5C). As decrease in FET3 mRNA levels by iron might reflect mRNA degradation rather than ISC assembly remediation (Hassett et al, 1998), we also inspected Leu1 activity (Figure 5D). A 1‐h exposure of six‐division GSH‐depleted cells to FeCl₃ (300 μM) increased Leu1 activity from levels that were undetectable to more than half of those that were reached on a 1‐h exposure to 1 μM GSH. Ferric iron (100 μM) also enabled some growth of gsh1Δ cells on medium lacking GSH, under anaerobiosis slightly better than aerobiosis (Supplementary Figure S7B). Iron can boost the loss of mitochondrial DNA (mtDNA) occurring during GSH depletion (Kistler et al, 1986; Ayer et al, 2010). We thus also tested its effect on a fully homogeneous petite gsh1Δ cell population, assuming that any adverse influence of iron on growth rate by mtDNA loss would be avoided. Growth of GSH‐depleted petite gsh1Δ cells was now better rescued by ferric iron, irrespective of the presence of ambient oxygen, and with a dose‐dependent effect up to 500 μM (Figure 5E and F). In comparison, growth rescue by DTT (1 mM) was smaller and was observed with cells less depleted of GSH (Figure 5F). Similar to DTT‐rescued cells (Sharma et al, 2000; Spector et al, 2001), iron‐rescued ones did not grow upon re‐inoculation on medium containing iron but not GSH, indicating only transient, partial rescue.

Thus, the intense iron starvation‐like response constituting the only significant genome‐wide alteration of GSH‐depleted cells suggests that it is a defect in iron metabolism that causes the inviability of these cells. The partial correction by iron of the ISC assembly and growth defects of these cells supports this conclusion.

So far, by highlighting its role in iron metabolism while minimizing its impact on thiol‐redox control, except for the high GSH levels on the ER, all data that were presented questioned the notion of the primary role of GSH in reducing oxidized Cys residues. Previous studies indicated a quasi‐exclusive role of the thioredoxin pathway in peroxide metabolism (Le Moan et al, 2006), and a dominant role in RNR reduction (Camier et al, 2007). As another way to evaluate the actual GSH contribution to thiol‐redox control, we evaluated how GSH coped with the redox load increase resulting from inactivation of thioredoxin reductase (trr1Δ). This strain carries a 1.5‐ to 3‐fold increase in total GSH, with proportional increase of GSH and GSSG (Trotter and Grant, 2003; see Table I), presumably indicating a compensatory response to an increased redox load. In medium lacking GSH, trr1Δ cells grew very poorly. Adding high GSH amounts (100–200 μM) improved growth to near wild‐type levels (Figure 6A), indicating that the redox peptide was still very limiting despite its 1.5‐ to 3‐fold increase. To rigorously quantify the GSH demand caused by thioredoxin inactivation, we compared the gsh1Δ strain with a strain lacking both GSH1 and TRR1 (gsh1Δtrr1Δ). While the former strain only required 0.5 μM GSH to achieve a wild‐type growth, the latter required 400‐fold more GSH (200 μM) (Figure 6A). Anaerobiosis did not change the gsh1Δ GSH optimal growth requirement, but totally corrected the trr1Δ one (Figure 6B). Growth of the gsh1Δtrr1Δ strain was also improved, as now requiring 20‐fold less GSH (10 versus 200 μM) for optimal growth (compare Figure 6A and B). However, such GSH requirement was still 20‐fold higher than that of gsh1Δ (10 versus 0.5 μM), which indicated that an increased cellular redox load was still present in anaerobically grown trr1Δ. In trr1Δ, as in gsh1Δtrr1Δ, thioredoxin was largely oxidized (ox/red ∼70%) compared with its oxidation in wild‐type cells (∼10%), and GSH addition decreased this oxidation back to 50% (Figure 6C) indicating that the extra‐GSH improved the growth of these cells by balancing their redox load increase. Furthermore, note here (Figure 6C) that trr1Δ had more than a 10‐fold increase in thioredoxin abundance, further demonstrating the tremendous compensatory response to an increased redox load. Conversely, however, in the absence of GSH, no increased redox load was applied on the thioredoxin pathway under aerobiosis, as indicated by total preservation of a wild‐type redox state (ox/red ∼10%) and levels of thioredoxin in gsh1Δ cells grown for up to 10 divisions in the absence of GSH (Figure 6D).

In summary, GSH is not capable on its own of efficiently bearing the redox load of the cell, whereas thioredoxin can do so without any increase in levels, which places GSH at best as backup of thioredoxin. As corollary, the contrast between the very high and trace amounts of GSH needed for thiol‐redox maintenance and viability, respectively, should disqualify this function as explaining its viability requirement.

We enquired whether GSH linked thiol‐redox and iron metabolism by monitoring FET3 expression in thioredoxin reductase mutants. In trr1Δ, FET3 had a 2.5‐fold increased expression, but adding GSH returned this expression to wild‐type levels (Figure 7A), which indicated that in these cells GSH is also limiting for iron metabolism despite its 1.5‐ to 3‐fold compensatory increase in levels. In gsh1Δtrr1Δ, FET3 expression was further induced, in keeping with the very low uncompensated GSH levels of these cells. Adding GSH decreased FET3 expression, but only slightly and very transiently, as FET3 mRNA levels increased again after 30 min (Figure 7B). The latter observation indicates that thioredoxin deficiency not only led to an increased GSH demand (see Figure 6A and B), but also caused very fast consumption of the redox peptide.

In conclusion, under conditions increasing its thiol‐redox maintenance duties, GSH becomes limiting for its function in iron metabolism.

Links between GSH and iron metabolism were initially suggested by the presence of increased and decreased levels of the tripeptide in yeast ATM1 (Kispal et al, 1997) and frataxin mutants (Auchere et al, 2008), respectively. The observations that GSH depletion causes impairment of extra‐mitochondrial ISC maturation (Sipos et al, 2002) and activates Aft1 (Rutherford et al, 2005), allowed identifying GSH as part of mitochondrial ISC export, a system providing to the CIA machinery a compound of undefined nature emanating from the mitochondrial ISC machinery (Lill, 2009), also serving as a signal that negatively regulates Aft1/2 (Chen et al, 2004). We now show that the near‐unique genome‐wide consequence of GSH depletion is a major iron starvation‐like response remodelling most of iron‐dependent mitochondrial and extra‐mitochondrial pathways (see Figure 2). We also show that, unexpectedly, toxic GSH levels recapitulate the iron phenotype of GSH depletion. The molecular function of GSH in iron metabolism may not involve a conventional thiol‐disulphide reductase activity, as suggests the ability of iron (Figure 5D), but not DTT (Sipos et al, 2002), to partially correct the ISC defect of GSH‐depleted cells. GSH probably engages a partnership with Grx3 and Grx4, with which it forms in vitro unusual Fe–S‐bridged dimeric complexes that are coordinated by its Cys residue and the glutaredoxin active site ones (Li et al, 2009; Rouhier et al, 2009). A recent report has provided proofs of the in vivo presence of these unusual Fe–S clusters, and has proposed that they deliver iron to all iron‐containing proteins and to mitochondria (Muhlenhoff et al, 2010). The slightly higher GSH requirement of cells lacking GSH1, GRX3 and GRX4 in comparison with those only lacking GSH1 (Supplementary Figure S7C) is tantamount of synthetic lethality, and indicate that the corresponding gene products operate in the same essential pathway. Nevertheless, the ability of iron to improve the ISC assembly defect of GSH‐depleted cells could also indicate a specific function of GSH in cytosolic iron delivery. However, while Grx3/4 depletion inactivates both mitochondrial and extra‐mitochondrial ISC assembly (Ojeda et al, 2006; Muhlenhoff et al, 2010), GSH depletion only inactivates the latter (Sipos et al, 2002) (this study), which does not fit the notion of shared functions of GSH and Grx3/4, by Fe–S cluster coordination. These seeming discrepancies might be, at least in part, linked to the incomplete nature of GSH cellular depletion that causes cell death by affecting first extra‐mitochondrial ISC assembly before the full phenotypic scope of GSH deficiency reveals, which could also suggest wider functions of GSH in iron metabolism than Grx3/4.

Toxic GSH levels also impaired Leu1 activity, while sparing Aco1 activity. Impairment of the cytosolic Fe–S enzymes could be caused by disruption of the Grx3/4‐GSH Fe–S clusters by titration of Grxs by GSH; however, overexpressing Grx4 in HGT1 cells only corrected GSH‐induced high FET3 expression but not growth inhibition (Supplementary Figure S5D, E and F). High GSH levels could also prevent formation of the GSH‐Grx3/4 Fe–S clusters by iron sequestration, or could directly affect cytosolic Fe–S enzymes, as suggests the very fast inhibition of Leu1 (see Figure 3). The fact that toxic GSH levels spared mitochondrial iron metabolism was surprising, and might relate to the exclusion from this organelle of the high GSH levels reached in HGT1 cells (not shown). Further work will be needed to establish the molecular function and toxicity of GSH in iron metabolism and its partnership with Grx3/4.

The data presented here, all point to a limited cytosolic redox function of GSH. GSH depletion should have caused Cys residue oxidation, but the genome‐wide mRNA profiles of GSH‐depleted cells did not reveal any other hint of redox imbalance than induction of a very few Yap1 targets (see Figure 2; Supplementary Table S3). This result, and the observation that GSH depletion does not cause any proteome‐wide thiol oxidation (Le Moan et al, 2006), suggest that GSH is totally dispensable for cytosolic thiol‐redox maintenance. Reciprocally high GSH levels should have caused Cys residue reduction, but this effect was only observed in the ER, in which GSH led to total reduction of Ero1 and Pdi1, which blocked secretion and induced the UPR (see Figure 4). We also show, using cells with inactivation of the thioredoxin pathway, that GSH is not capable of supporting the redox load of the cell on its own. Despite the presence of a 1.5‐ to 3‐fold compensatory increase of total GSH in trr1Δ (Trotter and Grant, 2003; see Table I), the redox peptide was still very limiting for growth (Figure 6A) and was unable to sustain the redox load increase resulting from the lack of thioredoxin reductase. A large part of the redox load increase in trr1Δ was linked to aerobic metabolism, but was also present under anaerobiosis (Figure 6A and B). Reciprocally, cells that were severely depleted of GSH preserved thioredoxin in their wild‐type redox state, without any protein compensatory increase, indicating that the thioredoxin pathway could entirely assume the thiol‐redox maintenance duties of the cell. These observations and the quasi‐exclusive role of the thioredoxin pathway in peroxide metabolism (Le Moan et al, 2006) and a dominant role in RNR reduction (Camier et al, 2007), place GSH as an ancillary system of the thioredoxin pathway in cytosolic thiol‐redox control.

The GSH requirement for viability could theoretically be linked to either its thiol‐redox or iron‐metabolism functions, as both these processes are essential (Toledano et al, 2007; Lill, 2009). Notwithstanding, the notion that GSH thiol‐redox maintenance only serves as backup of thioredoxin, disqualifies this function as causing the GSH essential requirement. The contrast between the very high GSH levels needed for thiol‐redox maintenance in the absence of thioredoxin and the very low intracellular GSH levels required for viability (Figure 6A; Table I), at which GSH should behave as a thiol oxidant, further support this idea. Still, GSH could have one exclusive essential redox target, but this is unlikely, as DTT should have then totally rescued the growth of GSH‐depleted cells, as it does in E. coli that is made unviable by inactivation of thioredoxin reductase and glutathione reductase (Ortenberg et al, 2004; Gon et al, 2006a). Contrasting with the negligible importance of its thiol‐redox function, the function of GSH in iron metabolism appeared pre‐eminent. The intense iron starvation‐like response that constituted the near‐unique genome‐wide consequence of GSH depletion befit the idea that it is this function that makes GSH essential, which is supported by the partial rescue by iron of the ISC and growth defects of GSH‐depleted cells. DTT can also partially rescue GSH auxotrophy (Grant et al, 1996; see Figure 5F), presumably by substituting for some of its redox functions, thereby decreasing its consumption (Sharma et al, 2000; Spector et al, 2001; Ayer et al, 2010), or by recruiting GSH by reduction of GSSG, thus temporarily sparing its vital iron function.

The lethal effect of toxic levels of GSH is clearly linked to its impact on secretion, as indicated by the partial growth rescue of HGT1 cells by the thiol oxidant diamide that decreased ER reductive stress (not shown), and by the much lower GSH tolerance of HGT1 ire1Δ. The latter observation also indicates that cell death caused by prolonged ER stress is not linked to H₂O₂ production (Haynes et al, 2004), as HGT1 ire1Δ cells did not generate H₂O₂ when exposed to GSH, but were much more sensitive to its toxicity. The GSH impact on iron metabolism may also affect viability, but how it does so will only be deduced from knowledge of its molecular function in this pathway.

The data presented here provide a framework of the cellular organization, compartmentation and functions of thiol‐redox pathways in eukaryotic cells that has been lacking. They also establish an unsuspected crosstalk between thiol‐redox and iron homeostasis that probably applies to higher eukaryotes, in which GSH is also essential for viability (Shi et al, 2000) and partners with glutaredoxins to assemble Fe–S clusters (Rouhier et al, 2009). Although higher eukaryotes have necessarily more complex thiol‐redox systems, as here but not in yeast, GSH is also involved in peroxide scavenging by serving as exclusive proton donor for the peroxide scavengers glutathione peroxidases (Toledano et al, 2007), and by constituting a cellular nitric oxide storage in the form of nitrosoglutathione (GS‐NO) (Hess et al, 2005). The redox‐iron homeostasis crosstalk described here predicts a novel toxic mechanism for redox stresses caused by peroxides, electrophilic xenobiotics and metals that may endanger fitness not only by direct effect but by consumption of GSH and inhibition of cytoplasmic Fe–S protein maturation. Knowledge of such crosstalk should contribute to elucidating the intricate pathophysiological links between redox imbalance, GSH depletion, alterations of iron metabolism and oxidative stress underlying age‐related diseases, such as neurodegeneration and inflammation (Altamura and Muckenthaler, 2009; Bharath et al, 2002; Lee et al, 2009).

The S. cerevisiae strains ABC 1144 (WT) and ABC 1230 (HGT1 cells) are derived from YPH499 (MATα ura3‐52 leu2Δ‐1 lys2‐801 his3Δ‐200 trp1‐63 ade2‐101) and carry an integrated URA3 cassette or TEF‐HGT1, respectively (Srikanth et al, 2005). Y252, isogenic gsh1Δ (Spector et al, 2001) and trr1Δ (Le Moan et al, 2006) were described. The gsh1Δtrr1Δ strain was produced by integrating KanMX4 at the TRR1 locus in gsh1Δ. BY4742 and isogenic derivative ire1Δ are from EUROSCARF. The pTEF‐416‐HGT1 plasmid (CEN URA3 TEF‐HGT1) has been described (Bourbouloux et al, 2000); pRS314‐TEF‐HGT1 (CEN TRP1 TEF‐HGT1) was constructed by subcloning a partial SacI–KpnI digest fragment from pTEF416‐HGT1 into pRS314; pAFT1‐GFP is a gift from MA de la Torres‐Ruiz (Lleida, Spain) (Pujol‐Carrion et al, 2006); pTEF‐AFT1‐GFP was constructed by inserting a pTEF‐416 400‐bp SacI–XbaI fragment containing the TEF1 promoter in place of the MET25 promoter in pAFT1‐GFP; pAF85 (CEN LEU2 ERO1–Myc) is a gift from Chris Kaiser (Sevier et al, 2007). pRS316–Myc–Yap1 was described (Delaunay et al, 2002). Cells were grown in YPD (1% yeast extract, 2% peptone and 2% glucose), YPG (same as YPD, with 2% glycerol instead of glucose), or minimal media (SD) (0.67% yeast nitrogen base w/o amino acids, 2% glucose), or minimal media glycerol (SG) (same as SD with 2% glycerol instead of glucose) with amino‐acid supplements as appropriate, at 30°C. Anaerobic growth was performed as described (Spector et al, 2001). For generating respiratory incompetent (petite) gsh1Δ, cells were grown overnight in YPD, and then diluted in SD without GSH until they were growth arrested, re‐inoculated in SD+100 μM GSH, and tested for their ability to grow on YPG. GSH, GSSG and ferric chloride (FeCl₃) were obtained from Sigma.

The pellet of cold H₂O‐washed yeast cells was suspended in perchloric acid (0.1%, pH 2), boiled (95°C, 5 min) and centrifuged. An aliquot of the supernatant, which correspond to the metabolic fraction, was mixed with an aliquot of an ¹⁵N‐labelled reference metabolic fraction, and diluted with formic acid (0.1%) to reach the equivalent of 4 × 10⁷ cells/ml, and injected into the chromatographic system for analysis, as described (Lafaye et al, 2005). GSH and GSSG concentrations in mM/cell were calculated on the basis of an average yeast cell volume of 4.5 × 10⁻¹⁴l, which assumes that yeast cells are spherical, with an average diameter of 4.4 μM, as determined by microscopic measurement of 100 cells. We ensured that these size parameters did not change with strain backgrounds and growth conditions.

For each growth condition, microarray experiments were performed in two replicates with RNA isolated from independent cultures grown on different days. Each replicate comprised a dye swap between the experimental and reference samples. Totals RNAs were extracted using the Agilent Total RNA isolation mini kit. Labelled complementary RNAs (cRNAs) were generated using the Agilent Low RNA Input Linear Amplification Kit. The cRNA synthesis and Cy‐Dye incorporation efficacies were checked by spectrophotometry. Equal amounts of Cy3‐ (reference sample) and Cy5 (experimental sample)‐labelled probes, and the reverse for the dye swap reaction, were mixed together and hybridized onto the Agilent Yeast Oligo Microarray (V2) slides. Slides were scanned using the GenePix 4000B scanner (Molecular Devices). Raw data of spot intensities corrected for background signals were calculated using GenePix Pro 6.0 (Molecular Devices) and exported to GeneSpring Gx 7.3 (Silicon genetics Agilent) for normalization and statistical analyses (see Supplementary data). Genes with an expression change of ⩾2‐fold were considered induced or repressed. A lower cutoff of ⩾1.7‐fold was used for the 5‐min HGT1 condition.

Unique oligonucleotide primer pairs were designed with the QuantPrime software (http://www.quantprime.de) and assayed on efficiency (>80%) and specificity. Six micrograms of RNeasy Mini kit (Quiagen) purified total RNA, digested with RQ1 RNase‐Free DNase (Promega) was reverse‐transcribed with SuperScript II reverse transcriptase (Invitrogen) with random hexanucleotides (Roche Diagnostics). Real‐time quantitative PCR was performed in triplicate as described (Desaint et al, 2004).

Aft1‐GFP subcellular localization used cells carrying pTEF‐AFT1‐GFP stained with DAPI, washed, suspended into DABCO and visualized as described (Delaunay et al, 2002). For detection of H₂O₂, cells were collected, washed, incubated for 10 min in PBS containing Peroxyfluor‐1 (PF1; 10 μM; a gift from Chris Chang, California, USA) (Miller et al, 2005), collected, washed, suspended into PBS containing the indicated amount of GSH or H₂O₂, and analysed by fluorescence microscopy. Alternatively, cells were collected and suspended into PBS containing the indicated amount of GSH and dihydrorhodamine 123 (20 μM) (Invitrogen), incubated for 1 h at 30°C, and analysed by flow cytometry as described (Desaint et al, 2004).

The pellet of yeast cell cultures was suspended in lysis buffer (Tris–HCl (10 mM, pH 7.4), EDTA (2.5 mM), NaCl (150 mM), glycerol (10%, v/v) and Triton X‐100 (0.5%, v/v)), and lysed by agitation with glass beads. Cell extracts were centrifuged and supernatants were used for enzyme assays. Aco1 activity was measured by the disappearance of cis‐Aconitate at 240 nm, and Leu1 by the decrease in absorbance at 235 nm using citraconate as described (Kispal et al, 1999). Aco1 and Leu1 protein levels were quantified by western blot using polyclonal anti‐Aco1, a gift from Anne Bulteau (Paris, France), and anti‐Leu1, a gift from Rolland Lill (Marburg, Germany), antibodies. Total cellular iron was determined in mole/cell from a 250‐ml culture of exponentially growing HGT1 cells (OD₆₀₀=0.3–0.4) as described (Doeg and Ziegler, 1962).

Exponential phase cells were suspended in 10‐ml SD medium lacking methionine (∼6 × 10⁷ cells/ml), pulse‐labelled with 500 μCi of [³⁵S] methionine (NEN, Perkin‐Elmer) for 7 min and chased with excess cold methionine. A volume of 1.5 ml samples (1 × 10⁸ cells) were collected in 20 mM NaN₃, lysed in 100 μl of lysis buffer (Tris–HCl (50 mM, pH 7.5), EDTA (1 mM), sodium dodecyl sulphate (SDS; 1%), urea (6 M) protease inhibitor complete (Roche) (1 ×), PMSF (1 mM)) by agitation with glass beads, and by boiling. Cell extracts were suspended in 1 ml of IP buffer (Tris–HCl (50 mM, pH 7.5), NaCl (150 mM), EDTA (1 mM), Tween‐20 (0.5%), protease inhibitor complete (1 ×) and PMSF (1 mM)), mixed with Sheep anti‐Mouse IgG Dynabeads® (Invitrogen) coupled to mouse monoclonal anti‐Cpy1 antibodies. The mix was tumbled for 2 h at 4°C, washed three times in IP buffer, recovered in 20 μl 1 × Laemmli buffer, boiled and separated by 8% SDS–PAGE. Gels were analysed with a Storm PhosphorImager and ImageQuant TL Software (GE Healthcare).

Trichloroacetic acid‐precipitated cell extracts (Delaunay et al, 2002) were dissolved in SB buffer (Tris–Cl (100 mM, pH 8.8), EDTA (10 mM) and SDS (1%)) containing 30 mM 4‐acetamido‐4′‐maleimidylstilbene‐2,2′‐disulphonic acid (AMS, Invitrogen), or 50 mM N‐ethylmaleimide and incubated at 37°C for 2 h. Protein concentration was measured by bicinchoninic acid‐based colorimetric detection (micro BCA kit, Pierce). For Ero1 and Pdi1, 20 μg protein extracts were mixed with 3 × Laemmli buffer without β‐mercaptoethanol, boiled and deglycosylated by incubation with 250 units of endoglycosidase H (NEB) for 60 min at 37°C after addition of sodium citrate (67 mM, pH 5.5). Proteins were resolved by non‐reducing SDS–PAGE, and revealed by western blot with 9E10 anti‐Myc monoclonal (a gift from C Créminon, Saclay, France), anti‐Pdi1 rabbit polyclonal (a gift from J Winther, Denmark), rabbit polyclonal anti‐thioredoxin (a gift from C Grant, UK) or anti‐His antibodies, anti‐mouse IgG or anti‐rabbit IgG secondary antibodies labelled with fluorophores of different wavelengths, followed by Infrared Imaging technology (Odyssey, LI‐COR).

Supplementary data are available at The EMBO Journal Online (http://www.embojournal.org).