Role of Glutaredoxin-3 and Glutaredoxin-4 in the Iron Regulation of the Aft1 Transcriptional Activator in Saccharomyces cerevisiae*

The transcription factors Aft1 and Aft2 from Saccharomyces cerevisiae regulate the expression of genes involved in iron homeostasis. These factors induce the expression of iron regulon genes in iron-deficient yeast but are inactivated in iron-replete cells. Iron inhibition of Aft1/Aft2 was previously shown to be dependent on mitochondrial components required for cytosolic iron sulfur protein biogenesis. We presently show that the nuclear monothiol glutaredoxins Grx3 and Grx4 are critical for iron inhibition of Aft1 in yeast cells. Cells lacking both glutaredoxins show constitutive expression of iron regulon genes. Overexpression of Grx4 attenuates wild type Aft1 activity. The thioredoxin-like domain in Grx3 and Grx4 is dispensable in mediating iron inhibition of Aft1 activity, whereas the conserved cysteine that is part of the conserved CGFS motif in monothiol glutaredoxins is essential for this function. Grx3 and Grx4 interact with Aft1 as shown by two-hybrid interactions and co-immunoprecipitation assays. The interaction between glutaredoxins and Aft1 is not modulated by the iron status of cells but is dependent on the conserved glutaredoxin domain Cys residue. Thus, Grx3 and Grx4 are novel components required for Aft1 iron regulation that most likely occurs in the nucleus.

Iron, an indispensable nutrient in cell physiology, is used in ironsulfur clusters, hemes, and diiron-oxo metal centers in enzymes. Saccharomyces cerevisiae, a model organism in metal metabolism, maintains iron homeostasis largely through the regulation of iron uptake and storage. In this yeast, survival under low iron conditions is ensured through the utilization of the iron-responsive transcriptional activators Aft1 and Aft2 (1)(2)(3). These factors are activated in iron-deficient cells and induce the expression of more than 20 genes that are referred to as the iron regulon (4 -7). This regulon includes genes whose products function in ionic iron acquisition, iron siderophore uptake, and vacuolar iron utilization. Activated Aft1 also induces the expression of CTH2 that encodes an RNA-binding protein. Cth2 mediates the degradation of transcripts of some iron-requiring enzymes to conserve iron in the cell (8).
Aft1 is localized to the nucleus under low iron conditions and to cytoplasm under iron-sufficient conditions (9). Under iron-sufficient conditions, Aft1 remains inactive due to its cytoplasmic localization where it is unable to drive transcription (9). Aft1 has two nuclear local-ization sequences and a nuclear export sequence (NES), 2 which map to its N-terminal DNA binding domain (9,10). Mutations of two leucines within the NES result in retention of Aft1 within the nucleus and constitutive transcriptional activity regardless of iron levels. In addition, Aft1 contains a functionally important conserved 291 CXC 293 sequence motif adjacent to the DNA binding domain and 190 residues downstream of the NES. Cys to Phe substitutions at either Cys within this motif in Aft1 result in constitutive transcriptional activation in ironreplete cells (9,11). As expected, the constitutively active C291F Aft1 variant (Aft1-1 up ) is retained within the nucleus. Thus, iron-regulation of Aft1 is dependent on its cycling between the nuclear and the cytoplasmic compartments.
The mechanism by which Aft1 and Aft2 sense cellular iron levels has been a topic of interest and intense research. Clues on the mechanism of iron sensing came from the observation that cells defective for Fe-S cluster biogenesis within the mitochondrial matrix exhibited constitutive expression of the iron regulon (12). Since disruption of Fe-S cluster biogenesis results in mitochondrial iron accumulation, it was initially thought that Aft1 was constitutive by virtue of depletion of cytosolic iron (12). It was later shown that disruption of Fe-S cluster biogenesis by diminution in the levels of the cysteine desulfurase (Nfs1) or the frataxin homologue (Yfh1) did not decrease cytosolic iron (13). This is an indication that Aft1 becomes constitutive due to impairment of a signal created by the mitochondrial Fe-S biosynthetic machinery and not to an indirect effect of alteration in iron compartmentalization.
In S. cerevisiae mitochondria are required for maturation of Fe-S proteins both inside and outside of the organelle (14). For synthesis of cytosolic and nuclear Fe-S proteins, mitochondria export a still unknown compound via the mitochondrial inner membrane transporter Atm1 (15). Other components of this export machinery are the mitochondrial intermembrane space sulfhydryl oxidase Erv1 as well as glutathione (15,16). Depletion of glutathione activates Aft1 (11,17). After export to the cytosol, the cytosolic Fe-S protein assembly machinery (CIA) matures Fe-S clusters and inserts them into target proteins (14). The CIA machinery includes the proteins Nar1, Cfd1, Nbp35, and Cia1 (18 -20).
Iron sensing by Aft1 and Aft2 requires proper mitochondrial Fe-S cluster biosynthesis as well as a functional export to the cytoplasm. However, it does not require the CIA machinery (11), demonstrating that iron sensing by Aft1/Aft2 is not linked to the maturation of cytosolic 4Fe-4S clusters.
Since the CIA complex is not required to mediate iron inhibition of Aft1 function, we predicted that other proteins may be involved in sensing the iron inhibitory signal extruded by Atm1. One attractive candi-date protein was the nuclear glutaredoxin-3 (Grx3), which was reported to interact with Aft1 in a global yeast two-hybrid interaction study (21). Glutaredoxins are glutathione-dependent thiol-disulfide oxidoreductases that function in maintaining the cellular redox homeostasis. S. cerevisiae has two dithiol glutaredoxins (Grx1 and Grx2) and three monothiol glutaredoxins (Grx3, Grx4, and Grx5) (22)(23)(24). The monothiol glutaredoxins are believed to reduce mixed disulfides formed between a protein and glutathione in a process known as deglutathionylation. In contrast, dithiol glutaredoxins can participate in deglutathionylation as well as in the direct reduction of disulfides (25). Grx5, the most studied monothiol glutaredoxin, is localized to the mitochondrial matrix, where it participates in the maturation of Fe-S clusters (24). Grx3 and Grx4 are predominantly localized to the nucleus (26). These proteins can substitute for Grx5 when overexpressed and targeted to the mitochondrial matrix (23); no information on their natural function has been reported. In addition to the reported interaction between Grx3 and Aft1, iron inhibition of Aft1 requires glutathione (11). Based on these clues, we evaluated the role of Grx3 and Grx4 in the iron inhibition of Aft1 and show presently that iron sensing is dependent on the presence of the redundant Grx3 and Grx4 proteins.

MATERIALS AND METHODS
Yeast Strains and Culture Conditions-The yeast strains used in this study are listed in Table 1. BY4742, ⌬grx3, ⌬grx4, and ⌬glr1 strains were obtained from Research Genetics. A PCR-created LEU2 cassette was integrated by homologous recombination at the GRX3 locus in a ⌬grx4 cell to create the ⌬grx3⌬grx4 strain. The ⌬trr1 and YM4271 strains were previously described (27). A PCR-created HIS3 cassette was integrated by homologous recombination at the AFT1 locus in a YM4271 strain to create the ⌬aft1 strain used in yeast two-hybrid experiments. Cells were grown at 30°C either in YPD medium, containing yeast extract, tryptone, and dextrose, or in complete synthetic medium (CM) or incomplete synthetic medium lacking, for example, uracil (CMϪUra) or leucine (CMϪLeu). For several experiments, the growth medium was supplemented with 0.1 mM bathophenanthroline sulfonate (BPS) as a ferrous iron chelator to lower the availability of iron or supplemented with 0.1 mM FeCl 2 . Doxycycline was added to the medium at a 5 g/ml final concentration for the indicated periods of time to modulate expression from the tetO 7 promoter (28). All cells were harvested during log phase.
Plasmids-All plasmid constructs were confirmed by DNA sequencing. Full-length wild type GRX3 and GRX4 coding sequences as well as their mutant forms (GRX3 C211S and GRX4 C171S) were tagged at the 3Ј-end with one Myc epitope. In addition, full-length wild type GRX4 as well as its mutant forms (GRX4 C171S and GRX4 GPm) were also tagged at the 3Ј-end with His 6 epitopes. All of these constructs were cloned in the YCp pCM189 and YEp pCM190 plasmids under the control of the doxycycline-regulated tetO 7 promoter (28). In the previous cloning procedures, GRX3 constructs were cloned between NotI/PstI sites, and GRX4 constructs were cloned at BamHI/PstI sites.
For the two-hybrid experiments, the plasmid pBG4D-1, which contains the ADH1 promoter and the GAL4 (codons 1-147) DNA binding domain, was used. The GRX3 and GRX4 coding sequences were amplified and ligated into pBG4D-1, resulting in GRX3 and GRX4 3Ј-ends being fused in frame with the GAL4 DNA binding domain. The VP16 activation domain fused to the CYC1 terminator (5Ј to 3Ј orientation) was amplified and ligated into pRS416. The AFT1 promoter and open reading frame was PCR-amplified with SpeI/BglII sites and ligated into cut pRS416 VP16-CYC1 plasmid, resulting in AFT1 being fused in frame with VP16. The resulting AFT1 VP16 was used as a template for PCR mutagenesis of C291F and C293F. The previous AFT1 constructs are under the control of the AFT1 promoter.
AFT1 was TAP-tagged at its C terminus by homologous recombination in its chromosomal locus (29). The genomic AFT1-TAP was later used as a template in a PCR where AFT1-TAP was amplified, cut, and ligated into pCM190, where it is under the control of the tetO 7 promoter. Wild type AFT1, AFT1-1 up , and AFT1 L99A were subcloned in plasmid pRS416 under the control of its own promoter.
The C-terminal 375 bp of GRX4 (including the glutaredoxin domain but excluding the thioredoxin domain) as well as the C-terminal 381 bp of GRX5 (excluding the mitochondrial target sequence) was PCR-amplified and ligated into pCM190, where they were under the control of the tetO 7 promoter. S1 Nuclease Assays-RNA was extracted from cells grown to midlog phase using the hot acid phenol method, and S1 analysis was performed as previously described (30). For each reaction, 12 g of total RNA were hybridized to a 32 P end-labeled DNA oligonucleotide probe before digestion with S1 nuclease and separation on an 8% polyacrylamide, 8 M urea polyacrylamide gel. Dried gels were imaged using a Bio-Rad FX phosphor imager and quantified using Quantity One software prior to autoradiography.
DNA Microarray Analysis-RNA was extracted from wild type BY4742 and ⌬grx3⌬grx4 cells grown in YPD medium supplemented with 200 M FeCl 2 . Total RNA was isolated using the hot acid phenol method. mRNA was isolated from total RNA by using the Poly(A) Tract mRNA isolation system IV kit from Promega following the manufacturer's instructions. Fabrication of DNA microarray, synthesis of fluorescence-labeled cDNA, hybridization of the microarrays, and subsequent scanning were performed in the Huntsman Cancer Institute Microarray Core Facility at the University of Utah.
␤-Galactosidase Assays-Cells were grown to midexponential phase (A 600 0.5) in CMϪUraϪLeuϪHisϪTrp, 2% glucose either with supplemented iron or in the presence of BPS. ␤-Galactosidase activity was measured in permeabilized cells as previously described (31) and is expressed in Miller units that are calculated as follows (⌬420 ϫ 1000)/ (min ϫ ml of culture used ϫ absorbance of the culture at 600 nm). Immunoprecipitation and Immunodetection-Cellular lysates for immunoprecipitation analysis were prepared by glass beading in 50 mM Tris-Cl, pH 7.5, 150 mM sodium chloride, 0.1% Nonidet P-40, 0.05% sodium deoxycholate, and a protease inhibitor mixture. The supernatant was incubated with a rabbit polyclonal anti-Myc antibody for 1 h at 4°C. Protein A-agarose was added and incubated overnight at 4°C. The protein A-agarose was collected by centrifugation, washed three times, and boiled in SDS sample buffer. The immunoprecipitated protein was resolved by SDS-10% polyacrylamide gel electrophoresis and transferred to nitrocellulose. Membranes were blocked and probed with either PAP peroxidase anti-peroxidase (for TAP detection) or rabbit polyclonal anti-Myc. Detection was performed by enhanced chemiluminescence after incubation with a horseradish peroxidase-conjugated secondary antibody.
In addition, an aliquot of the supernatant was used for immunodetection analysis by immunoblotting, using PAP (Sigma), rabbit polyclonal anti-Myc (Santa Cruz Biotechnology, Inc., Santa Cruz, CA), mouse monoclonal anti-Pgk1 (Molecular Probes), and mouse monoclonal anti-His (Novagen).
Cellular lysates were prepared for immunoblotting by glass beading using 10% trichloroacetic acid in Tris acetate buffer, pH 8. Proteins were resolved by SDS-polyacrylamide gel electrophoresis and transferred to nitrocellulose. The membranes were probed with antibodies previously described and detected using chemiluminescence (ECL; Pierce).
Labeling of yeast cells with radioactive iron ( 55 Fe) and the determination of iron incorporation into Fe-S proteins by immunoprecipitation and liquid scintillation counting were carried out as previously described (15).
Miscellaneous Procedures-The following published methods were used. The sulfite reductase assay was performed as previously described (11). For aconitase activity assays, cells were lysed by bead beating, and aconitase activity was determined by coupled reaction of aconitase (EC 4.2.1.3) and isocitrate dehydrogenase (EC 1.1.1.42) (32). Mutagenesis was performed by either PCR mutagenesis or by introducing the mutation in the primer followed by homologous recombination (33). Yeast transformation was performed using standard lithium acetate protocol (34).

RESULTS
To evaluate the role of Grx3 and Grx4 in the iron inhibition of Aft1, we quantified the expression of two iron regulon genes, FET3 and FIT3, in cells lacking either Grx3 or Grx4 or in cells lacking both molecules (Fig. 1, A and B). Gene expression was assessed by quantifying mRNA levels using the S1 nuclease protection assay. Whereas expression of FET3 and FIT3 was inhibited in iron-supplemented wild type cells, expression of FET3, but not FIT3, was elevated 3.5-and 2.5-fold in iron-supplemented ⌬grx3 or ⌬grx4 cells, respectively, relative to wild type cells. The absence of Grx3 or Grx4 did not affect the full induction of FET3 observed when the iron bioavailability is limited in cells treated with the iron chelator bathophenanthroline sulfonate (BPS).
Cells lacking both Grx3 and Grx4 exhibited constitutive expression of both FET3 and FIT3. To verify that the major iron regulon genes were expressed in ⌬grx3⌬grx4 cells, DNA microarray analysis was performed comparing wild type and ⌬grx3⌬grx4 cells cultured in YPD medium supplemented with iron ( Table 2). The same genes induced by the constitutively active Aft1-1 up were highly expressed in ⌬grx3 ⌬grx4 cells, although the observed induction ratios varied. Such variation in the induction ratios of iron regulon genes is observed under other conditions that activate Aft1 (4 -7). The high expression of the Aft2 target gene MRS4 (7, 35) suggested that both Aft1 and Aft2 are constitutively active in the ⌬grx3⌬grx4 strain. These data confirm that the iron regulon is induced in the absence of Grx3 and Grx4. The induced expression of iron regulon genes observed in ⌬grx3⌬grx4 cells cultured in ironreplete medium was due to the absence of Grx3 and Grx4, since transformation of the double null cells with GRX4 under the TET promoter restored inhibition of FET3 expression (Fig. 2). As expected, the addition of doxycycline to repress GRX4 expression yielded elevated FET3 expression.
Grx3 and Grx4 appear to be redundant molecules. ⌬grx4 cells transformed with a low copy plasmid containing either GRX3 or GRX4 restored full iron inhibition of FET3 transcription (Fig. 3A). Grx3 and Grx4 are members of the monothiol glutaredoxin family and as such possess a single functional cysteinyl residue within a CGFS sequence motif (36). We tested whether the single conserved cysteine residue within the CGFS motif in each protein is essential for iron inhibition of Aft1 activity. Transformation of ⌬grx4 cells with a YCp plasmid-borne GRX4 restored iron inhibition of FET3 expression, whereas transformants containing a mutant GRX4 allele encoding a C171S substitution failed to inhibit FET3 expression (Fig. 3A). In the same way, mutant Grx3 containing a C211S substitution failed to mediate iron inhibition

TABLE 2 Genes induced in ⌬grx3⌬grx4 cells compared with Aft1-1 up -containing cells
Microarray analysis was conducted on ⌬grx3⌬grx4 cells compared with wild type (WT) cells cultured in YPD containing 0.2 mM FeCl 3 . RNA was extracted from these cells, and poly(A) RNA was recovered. The mean from two duplicate experiments is shown. The -fold induction data are compared with transcript profile data of AFT1-1 up cells published previously. Only a subset of the iron regulon genes are shown to document that the iron regulon genes are induced in ⌬grx3⌬grx4 cells. In the duplicate experiments with ⌬grx3⌬grx4 cells, the variation in the -fold induction was within 10%. of FET3 expression in ⌬grx3 cells (Fig. 3B). Although the mutant Grx3 and Grx4 proteins were inactive, they were stably expressed (data not shown). Thus, the putative functional cysteinyl residue in each glutaredoxin is important to mediate iron inhibition of Aft1. Since depletion of Grx3 and Grx4 resulted in constitutive Aft1 activity, the effect of overexpression of GRX4 on Aft1 function in wild type cells was evaluated. Wild type cells cultured in iron-limited SC medium showed partial FET3 expression that could be completely inhibited by the addition of iron salts to the culture medium. The overexpression of GRX4 in cells cultured in this iron-limited medium markedly attenuated FET3 expression (Fig. 4, A and B, two lanes on the left). The inhibitory effect of GRX4 overexpression was also seen when Aft1 was fully activated in BPS-supplemented, iron-deficient cells (data not shown). In addition, the C171S substitution in Grx4 partially abrogated the ability of overexpressed Grx4 to attenuate FET3 expression in wild type cells (Fig. 4A).
Aft1 becomes constitutively active when the 291 CXC 293 motif or the NES motif is mutated (1,9). To address whether overexpression of Grx4 can attenuate the function of constitutively active variants of Aft1, the TET-GRX4-containing high copy vector was transformed into ⌬aft1 cells containing either an AFT1 allele encoding the C291F variant (Aft1-1 up ) or the L99A NES variant. Overexpression of Grx4 inhibited wild type Aft1 activity and resulted in a partial, reproducible attenuation of Aft1 (L99A) (Fig. 4, C and D) but no significant attenuation of the activity of the Aft1-1 up constitutive mutant (Fig. 4E).
Monothiol glutaredoxins are believed to function in the deglutathionylation of target proteins (37). In the deglutathionylation reaction, monothiol glutaredoxins are predicted to be transiently glutathionylated themselves (36). We evaluated whether the conserved residues that form the glutathione-binding pocket in other monothiol glutaredoxins are important for Grx4-mediated iron inhibition of Aft1 activity. We tested a double mutant of Grx4 in which the conserved 209 WP 210 was converted to 209 DA 210 . The 209 DA 210 mutant Grx4, designated Grx4 GPm for "glutathione pocket mutant," was unable to mediate iron inhibition of FET3 expression in ⌬grx4 cells (Fig. 3C). The Grx4 GPm mutant did not affect the full induction of FET3 by iron deprivation and was shown to be equally stable to the wild type protein by immunoblotting (data not shown). Thus, glutathione binding may be important for Grx4 activity.
Grx3 and Grx4 differ from the mitochondrial monothiol glutaredoxin Grx5 in that they contain an N-terminal thioredoxin domain in addition to the C-terminal glutaredoxin domain (26). The thioredoxin domain of Grx3 is believed to be responsible for the predominant nuclear localization of this glutaredoxin (26). To test whether the thioredoxin domain is important for iron inhibition of Aft1, we engineered a Grx4 truncate lacking the N-terminal thioredoxin domain. FET3 expression was iron-inhibited in ⌬grx3⌬grx4 cells harboring a high copy plasmid containing the Grx4 truncate, designated Grx4T (Fig. 5A). However, when the Grx4T was expressed in a low copy plasmid, it had no effect (data not shown). A green fluorescent protein fusion of the Grx4 truncate, expressed in a low copy plasmid, was found to have a diffused localization throughout the cell (data not shown), suggesting that insufficient protein existed within the nucleus to mediate iron inhibition of Aft1 activity.
Cells lacking Grx3 and Grx4 exhibit a growth defect in synthetic culture medium (Fig. 5B) but are less impaired in rich YPD medium (data not shown). The thioredoxin domain is nonessential for normal cell growth, because expression of Grx4 or Grx4T restored wild type growth (Fig. 5B).
Grx5 functions in the mitochondrial Fe-S biogenesis pathway (24). Grx3 and Grx4 were shown to substitute for Grx5 in mitochondrial function when overexpressed and targeted to the mitochondrial matrix (23). Suppression of the ⌬grx5 phenotypes requires both the thioredoxin and glutaredoxin domains and the essential glutaredoxin domain Cys residue (23). To address whether the mitochondrial Grx5 can complement ⌬grx3⌬grx4 cells, a truncated GRX5 construct was engineered that lacked the N-terminal mitochondrial target sequence (designated Grx5 C). Expression of the Grx5 truncate restored wild type growth of ⌬grx3⌬grx4 cells (Fig. 5B), yet FET3 expression was constitutive (Fig.  5A), suggesting that the growth defect is unrelated to constitutive Aft1 FIGURE 2. The iron regulon induction phenotype seen in ⌬grx3⌬grx4 cells is reversed by introducing a wild type GRX4 gene. ⌬grx3⌬grx4 cells transformed with a low copy plasmid containing GRX4 were grown in CMϪUra, 2% glucose for 19 h in the presence or absence of doxycycline (5 g/ml). S1 nuclease analysis was used to assess the expression of FET3 mRNA. CMD1 encoding calmodulin was used as the loading control.  Grx4 C171S). B, ⌬grx3 cells transformed with a low copy plasmid alone or with the same plasmid containing either the wild type GRX3 gene or the GRX3 C211S mutant gene (Grx3 C211S). Cells from A and B were grown in CMϪLeu, 2% glucose plus 50 M FeCl 2 prior to S1 nuclease analyses of FET3 and CMD1 mRNA levels. C, ⌬grx4 cells were transformed with a low copy plasmid alone or with the same plasmid containing either the wild type GRX4 gene or the GRX4 glutathione pocket mutant (Grx4 GPm). These cells were grown in CMϪLeu, 2% glucose supplemented with either 100 M BPS or 50 M FeCl 2 prior to S1 nuclease analyses of FET3 and CMD1 mRNA levels.
activity. Since the Grx5 truncate was not stably expressed in the yeast cytoplasm (data not shown), we cannot be certain whether the lack of iron inhibition of FET3 expression was a result of low protein levels or inactivity in that function.
The previous results are consistent with Grx3 and Grx4 having a direct role in the iron inhibition of Aft1 activity. The original motivation to consider Grx3 was its reported interaction with Aft1 in a global twohybrid study. A two-hybrid assay system for detecting protein-protein interactions was set up to confirm the binding interaction between Grx3 (or Grx4) and Aft1. The Gal4 DNA binding domain was fused to Grx3 and Grx4, generating Grx3/Gal4 and Grx4/Gal4 fusion proteins. The transactivation domain from the herpes simplex VP16 was fused to the C terminus of Aft1. Cells harboring the two fusion proteins and a GAL1/ lacZ reporter fusion were assayed for ␤-galactosidase activity. The combination of either glutaredoxin fusion with Aft1/VP16 resulted in elevated ␤-galactosidase activity (Fig. 6A). Mutation of the critical Cys in either glutaredoxin abrogated lacZ expression consistent with a loss of interaction. The mutant Grx3/Gal4 and Grx4/Gal4 fusion proteins were shown to be equally abundant as the wild type fusion proteins by immunoblotting (data not shown). In addition, the interaction of Grx3 and Grx4 with Aft1 was markedly diminished when the constitutively active Aft1 C291F,C293F mutant variant was used (data not shown). The observed interaction between the two glutaredoxins and Aft1 was not altered by changes in the cellular iron status (Fig. 6B).
To confirm the observed interaction between the glutaredoxins and Aft1, constructs were engineered in which AFT1 was TAP-tagged and GRX4 was either Myc-tagged or poly-His-tagged. Cells harboring vec- A, wild type cells transformed with a high copy plasmid alone (V) or with the same plasmid containing either the wild type GRX4 gene or the GRX4 C171S mutant gene were grown in CMϪUra, 2% glucose containing the BIO101 low iron nitrogen base. S1 nuclease analysis was used to assess the expression of FET3 mRNA. CMD1 encoding calmodulin was used as the loading control. B, the FET3 mRNA levels in A were quantified and normalized to CMD1 mRNA levels. C, ⌬aft1 cells were transformed with two plasmids, a low copy plasmid containing either a wild type AFT1 (Aft1) or a NES mutant AFT1 (Aft1 L99A) gene and a high copy plasmid alone or containing the wild type GRX4 gene. These cells were grown in CMϪUraϪLeu, 2% glucose containing the BIO101 low iron nitrogen base. S1 nuclease analysis was used to assess the expression of FET3 mRNA. D, FET3 mRNA levels in C were quantified and normalized to CMD1 mRNA levels. E, cells containing Aft1-1 up either in the presence or absence of GRX4 were grown in CMϪUraϪLeu, 2% glucose, and FET3 expression was assessed by S1 nuclease analysis. In three independent experiments done in duplicate, the mean FET3 expression in Grx4-overexpressing Aft1-1 up cells was 85% of that in control Aft1-1 up cells. FIGURE 5. The thioredoxin domain of Grx4 is not necessary for complementation of either growth or the control of Aft1 activity in a ⌬grx3⌬grx4 strain. ⌬grx3⌬grx4 cells were transformed with a low copy plasmid containing the wild type GRX4 gene (Grx4) or with a high copy plasmid either alone (V) or containing the GRX4 gene lacking the thioredoxin domain (Grx4 T) or the GXR5 gene lacking the mitochondrial target sequence (Grx5 C). A, cells were grown in CMϪUra, 2% glucose prior to S1 nuclease analyses of FET3 and CMD1 mRNA levels. B, cells from A were plated on CMϪUra, 2% glucose plates. The double null cells have a long lag phase in growth and this manifests in the low growth seen in B.
tors containing the AFT1-TAP under the control of the TET promoter and tagged GRX4 also under the control of the TET promoter were cultured in iron-replete SC medium. Immunoprecipitation of Grx4-Myc in lysates with anti-Myc antisera resulted in the co-immunoprecipitation of Aft1-TAP. In contrast, no co-immunoprecipitation of Aft1-TAP was observed if cell lysates containing the poly-His-tagged Grx4 were used instead of Grx4 Myc lysates in the same experiment (Fig. 7).
In addition, we noticed that when Aft1 was overexpressed in a high copy plasmid under the TET promoter, the overexpressed protein was not stable unless Grx4 was also overexpressed (Fig. 8A). Quantitation of blots revealed a striking 15-fold Grx4-dependent accumulation of Aft1-TAP protein. The Grx4-mediated stabilization of Aft1-TAP occurred in both iron-limited and iron-replete cells (Fig. 8B). However, the Grx4induced accumulation of Aft1 protein was attenuated when the C171S mutant Grx4 was overexpressed instead of the wild type. Despite the dramatic accumulation of Aft1 protein in Grx4-overexpressing cells, FET3 transcription was increased only 2-fold in iron-replete cells (Fig.  8C) and not increased in iron-deficient cells. Thus, the bulk of the Aft1 protein remains inactive in these cells.
When AFT1-TAP is present chromosomally under the control of its own promoter, overexpression of Grx4 does not increase Aft1-TAP protein levels significantly (data not shown). The endogenous Aft1-Tap levels are probably already stabilized by the endogenous Grx4 protein.
Glutaredoxins function to maintain reduced thiols in cellular proteins through glutathione disulfide redox reactions (23). The constitutive Aft1 activity in ⌬grx3⌬grx4 cells may arise from secondary effects due to changes in the cellular redox status. However, cells lacking either Grx3 or Grx4 individually have no change in the GSH/ GSSG ratio, and deglutathionylation of Tdh3 is unaffected (38). In addition, ⌬grx3⌬grx4 cells display only a modest increase in protein carbonylation, are not altered in glutathione reductase activity, and are only slightly affected in total glutaredoxin activity (22). Consistent with these data, we observed no sensitivity of the ⌬grx3⌬grx4 strain to growth in 100% oxygen (data not shown). Thus, ⌬grx3⌬grx4 cells do not appear to be grossly altered in their redox status. Nevertheless, we tested whether oxidative stress caused by alterations in the glutathione system induced constitutive activity in Aft1. To artificially induce oxidative stress, we used cells lacking either the thioredoxin reductase (Trr1) or the glutathione reductase (Glr1) enzymes. Cells lacking Trr1 exhibit elevated levels of glutathionylation (27). In contrast, cells lacking Glr1 or stressed with superoxide FIGURE 6. Yeast two-hybrid analyses shows that Aft1 interaction with Grx3 or Grx4 requires the conserved glutaredoxin domain cysteine and is independent of iron levels. ⌬aft1 cells were transformed with a high copy plasmid containing the GAL1/lacZ reporter, a low copy plasmid either empty (V) or containing the AFT1 VP16 construct, and a low copy plasmid with either GRX3 or GRX4 in frame with the GAL4 DNA binding domain. Interaction between Aft1 with the glutaredoxins results in increased lacZ expression. A, cells were grown in CMϪUraϪLeuϪTrpϪHis, 2% glucose supplemented with 100 M FeCl 2 prior to lacZ activity assays. Aft1 interaction with Grx3 or Grx4 depends on the conserved glutaredoxin domain cysteine. B, cells were grown in CMϪUraϪLeuϪTrpϪHis, 2% glucose supplemented with either 100 M BPS or 100 M FeCl 2 prior to lacZ activity assays. Aft1 interaction with Grx3 or Grx4 is independent of the iron levels in the medium. FIGURE 7. Aft1 co-immunoprecipitates with Grx4. Wild type cells were transformed with a high copy plasmid containing AFT1 (TAP-tagged) and a high copy plasmid containing GRX4 either Myc-or His-tagged. The lysate from these cells was immunoprecipitated with an antibody that recognizes the Myc-tagged Grx4, and both a sample of total cellular lysate and the immunoprecipitated fraction underwent immunoblotting and were probed with the antibodies shown in the figure. TAP detection is by peroxidase anti-peroxidase (PAP). Phosphoglycerate kinase Pgk1 is used as a loading control.
anions have an abnormally low GSH/GSSG ratio (27). To determine whether Aft1 activity is responsive to either excessive glutathionylation or to a decreased GSH/GSSG ratio, trr1⌬ and glr1⌬ cells were cultured in medium with high or low iron levels (Fig. 9). FET3 expression was iron-inhibited in both mutants, suggesting that Aft1 is not activated due to changes in the glutathione redox status. FET3 expression was low in ⌬trr1 cells cultured in iron-deficient medium for an unknown reason, but iron inhibition is still observed.
Aft1 is known to be constitutively active in cells impaired in Fe-S cluster biogenesis within the mitochondrion, including cells lacking Grx5 (13,24). The role of Grx5 in mitochondrial Fe-S cluster biogenesis led us to explore whether Fe-S maturation may also be impaired in ⌬grx3⌬grx4 cells. The enzymatic activity of the Fe-S enzyme aconitase (Aco1) was quantified in cells lacking either Grx3 or Grx4 or lacking both glutaredoxins. Aconitase activity was wild type in either single null mutant but was decreased in the double null strain (Fig. 10A). The diminution in aconitase activity was not due to oxidative disruption of the Fe-S cluster, since de novo biosynthesis of the Fe-S center in aconitase was also compromised, as shown by 55 Fe incorporation into immunoprecipitated Aco1 (Fig. 10B). 55 Fe incorporation into Aco1 was similar in wild type cells and ⌬grx3 or ⌬grx4 cells but reduced in ⌬grx3⌬grx4 cells. As with aconitase, enzymatic activities of the cytosolic Fe-S enzymes sulfite reductase and Leu1 are normal in single ⌬grx3 or ⌬grx4 cells and are only depressed in the double null strain (data not shown). Likewise, 55 Fe incorporation into Leu1 was similar in wild type cells and ⌬grx3 or ⌬grx4 single null cells but impaired in ⌬grx3⌬grx4 cells. The partial activity of Aft1 in ⌬grx3 or ⌬grx4 cells does not appear to arise from a defect in Fe-S cluster biogenesis or maintenance, although we cannot definitively rule out an indirect effect of Fe-S cluster disruption leading to Aft1 activation.

DISCUSSION
Iron inhibition of the Aft1 transcriptional activators is shown to require the monothiol glutaredoxins Grx3 and Grx4. Under iron-re-   . A, wild type (WT), ⌬grx3, ⌬grx4, and the double ⌬grx3⌬grx4 cells were grown in YPD medium, and their total cellular extracts were used for aconitase activity measurements as described under "Materials and Methods." B, wild type, ⌬grx3, ⌬grx4, and the double ⌬grx3⌬grx4 mutant were grown in iron-poor medium (SC) supplemented with glucose for 16 h. Cells were radiolabeled with 55 Fe for 2 h, a cell lysate was prepared, and Aco1 was immunoprecipitated with Aco1-specific antibodies. The amount of 55 Fe that co-immunoprecipitated with the antibodies was quantified by liquid scintillation counting. plete conditions, the iron regulon genes are partially induced in the absence of either Grx3 or Grx4 and fully induced in the absence of both proteins.
Grx3 and Grx4 contain a thioredoxin-like domain in addition to the glutaredoxin domain. In a previous study, the thioredoxin domain was shown to be required for the nuclear localization of these glutaredoxins (39). Expression of a mutant form of Grx4 that lacks the thioredoxin domain behaved as a wild type form only when overexpressed. We hypothesized that when overexpressed, a fraction of the truncated Grx4 leaks into the nucleus and exerts its effect on Aft1. Thus, the thioredoxin domain would be required to target Grx4 to the nucleus but not necessary for the control of Aft1 activity.
In contrast, the cysteinyl residue that is part of the conserved CGFS motif in monothiol glutaredoxins is essential for each glutaredoxin to modulate Aft1 activity. In addition, disruption of a putative glutathione pocket in Grx4 abolishes its capacity to regulate Aft1. These observations emphasize the importance of the glutaredoxin domain in the control of Aft1 and suggest a role for glutathione in this process.
An important clue to the function of Grx3 and Grx4 is their ability to bind Aft1. We reproduced the initial observation of an interaction by two-hybrid analysis and confirmed this interaction by co-immunoprecipitation. Furthermore, we observed a striking Grx4-mediated stabilization of Aft1 in cells overexpressing AFT1 and GRX4 under the strong TET promoter. The observed interaction and modulation of Aft1 function are dependent on the conserved Cys residue in the glutaredoxin domain and the 291 CXC 293 motif in Aft1. Likewise, the stabilization of Aft1 by Grx4 is impaired if Grx4 is mutated in the conserved glutaredoxin cysteine residue. Consistent with the Grx3/4-mediated stabilization of overexpressed Aft1, preliminary studies show that Aft1 is destabilized in ⌬grx3⌬grx4 cells (data not shown).
High levels of Grx4 inhibit Aft1 function, and the glutaredoxin domain Cys is important for inhibition of Aft1. Together, these observations underscore the importance of interaction between the monothiol glutaredoxins and Aft1 for proper control of Aft1 function.
Both Grx3 and Grx4 reside within the nucleus in S. cerevisiae cells. Green fluorescent protein fusions of both Grx3 and Grx4 were found predominantly within the nucleus, and this localization is dependent on the thioredoxin domain (39). Further proof that thioredoxin-containing glutaredoxins are nuclear is the observed nuclear staining of Schizosaccharomyces pombe Grx4, the only glutaredoxin with a thioredoxin-like domain in this species (40).
As mentioned, Aft1 cycles between the nucleus and cytoplasm, depending on the iron status of the cells (9). Aft1 is enriched within the nucleus in iron-deficient cells. Mutations that abrogate nuclear export result in constitutive Aft1 activity (9). Overexpression of Grx4 attenuates the activity of both wild type Aft1 and the NES Aft1 mutant. This suggests that Grx4 inhibits Aft1 activity in the nucleus, where this glutaredoxin normally resides. The decreased Grx4 inhibition of the Aft1-1 up mutant protein is consistent with the attenuation in the interaction between Grx4 and the Aft1-C291F,C293F mutant.
Iron inhibition of Aft1 requires a mitochondrial iron signal, which is exported to the cytoplasm through the Atm1 transporter localized in the mitochondrial inner membrane (11). It is unclear how Aft1 senses this signal and where the signal is sensed. We postulate that the interaction of Aft1 and Grx3/Grx4 within the nucleus preconditions Aft1 to sense the inhibitory signal. However, the iron-inhibitory signal is not the Grx/Aft1 interaction, since this interaction is not modulated by the iron status of cells. The inhibitory signal is postulated to cause a conformational change in Aft1 that destabilizes DNA binding and induces nuclear export. According to this model, Aft1 is unable to be exported from the nucleus in ⌬grx3⌬grx4 cells. The constitutive activity of the Aft1-C291F,C293F allele may arise from an attenuated Grx3/4 interaction and subsequent export.
In the published studies on localization of Grx3 and Grx4, the nuclear localization did not appear exclusive (38), suggesting that Grx3 and Grx4 may be in both the cytoplasm and nucleus. If these glutaredoxins cycle between these compartments, it is conceivable that Grx3 and Grx4 transport the putative iron-inhibitory signal from the mitochondrion to the nucleus and mediate iron inhibition of Aft1 within the nucleus. Alternatively, limited Grx3 or Grx4 within the cytoplasm may trap Aft1 in the cytoplasm in iron-replete cells. Current studies are focused on addressing these issues.
Little information exists on other functional roles of Grx3 and Grx4 in S. cerevisiae. The mitochondrial monothiol glutaredoxin Grx5 has an important role in Fe-S protein biogenesis that is an essential process. Cytosolic Fe-S protein maturation is essential for cell viability due to the Fe-S-containing Rli1 that participates in ribosome biogenesis and translation initiation (41). In this study, we addressed whether Grx3 and Grx4 together with the CIA machinery contributed to Fe-S maturation. We observed that in the absence of either Grx3 or Grx4, the activities of mitochondrial and cytosolic Fe-S clusters enzymes were unaffected. Since Aft1 is partially activated, we suggest that the activation of Aft1 is independent of Fe-S cluster biogenesis, as is the case in ⌬grx5 cells (42). However, since Fe-S cluster formation is attenuated in ⌬grx3⌬grx4 cells, we cannot definitely rule out that Aft1 is sensing a defect in Fe-S cluster biogenesis or maintenance.
The activity and 55 Fe incorporation into mitochondrial and cytosolic Fe-S enzymes are affected in the ⌬grx3⌬grx4 cells. Since Grx3 and Grx4 are largely nuclear proteins, it is difficult to assign a direct role of these glutaredoxins in Fe-S maturation within the cytoplasm. We speculate that this effect is indirect and will be addressed in future studies.
The cytosolic dithiol glutaredoxins Grx1 and Grx2 have functional roles in oxidative stress responsiveness (23). Grx1 appears important for protection against superoxide radicals, whereas Grx2 appears important for protection against peroxide stress (43). In contrast, Grx3 and Grx4 do not seem to be playing a critical role in the oxidative stress response, since the ⌬grx3⌬grx4 strain is not sensitive to hyperoxia (data not shown); however, it displays a modest decrease in the levels of glutaredoxin activity and a modest increase in the level of protein carbonylation (22). In this study, we addressed whether broad changes in the ratio of GSH/GSSG can influence the iron regulon and showed that in a ⌬glr1 strain, where this ratio is disrupted, Aft1 is normally regulated. Thus, Aft1 does not appear to be activated in the ⌬grx3⌬grx4 strain by a change in the redox status.
Dithiol glutaredoxins are able to reduce mixed disulfides formed between a protein and glutathione in a process known as deglutathionylation. Grx3 and Grx4 do not appear to influence deglutathionylation (38).
Furthermore, we did not detect any glutathione adducts in either Grx4 or Aft1 using an anti-glutathionyl antibody. We cannot, however, rule out that transient glutathionylation occurs in these molecules. It is unlikely that broad cellular glutathionylation increases Aft1 activity, since the deletion of the thioredoxin reductase gene TRR1 does not render Aft1 constitutively active. In addition, Aft1 is constitutively active when glutathione levels are depleted in a ⌬gsh1 strain (11), an indication that glutathione is not required to activate Aft1.
Monothiol glutaredoxins are present in organisms from bacteria to humans (26). The human homologue of Grx3 and Grx4, PICOT, is a negative regulator of protein kinase C- (44). Transient overexpression of PICOT in T cells inhibits the activation of the c-Jun N-terminal kinase (a target of protein kinase ) and the transcription factors AP-1 and NF-B (44). As in humans, yeast monothiol glutaredoxins Grx3 and Grx4 seem to play an important role in the control of specific transcription factors. Grx3 and Grx4 will probably have a function in addition to iron inhibition of Aft1, since ⌬grx3⌬grx4 cells exhibit growth defects. Grx4 is known to interact and be phosphorylated by the Bud32 kinase in vitro, although the functional significance of these processes is unclear (39). Finally, the importance of monothiol glutaredoxins is underscored by the fact that S. pombe cells lacking Grx4 are nonviable (40).