Yeast Mitochondrial Protein, Nfs1p, Coordinately Regulates Iron-Sulfur Cluster Proteins, Cellular Iron Uptake, and Iron Distribution*

Nfs1p is the yeast homolog of the bacterial proteins NifS and IscS, enzymes that release sulfur from cysteine for iron-sulfur cluster assembly. Here we show that the yeast mitochondrial protein Nfs1p regulates cellular and mitochondrial iron homeostasis. A strain of Saccharomyces cerevisiae, MA14, with a missenseNFS1 allele (I191S) was isolated in a screen for altered iron-dependent gene regulation. This mutant exhibited constitutive up-regulation of the genes of the cellular iron uptake system, mediated through effects on the Aft1p iron-regulatory protein. Iron accumulating in the mutant cells was retained in the mitochondrial matrix while, at the same time, iron-sulfur proteins were deficient. In this work, the yeast protein was localized to mitochondria, and the gene was shown to be essential for viability. Furthermore, Nfs1p in the MA14 mutant was found to be markedly decreased, suggesting that this low protein level produced the observed regulatory effects. This hypothesis was confirmed by experiments in which expression of wild-type Nfs1p from a regulated galactose-induced promoter was turned off, leading to recapitulation of the iron regulatory phenotypes characteristic of the MA14 mutant. These phenotypes include decreases in iron-sulfur protein activities coordinated with increases in cellular iron uptake and iron distribution to mitochondria.

Iron-sulfur (Fe-S) clusters are cofactors of proteins involved in oxidation-reduction, electron transport, metabolic conversions, and regulatory functions (1). The iron and sulfur are assembled in fixed stoichiometries (e.g. 2Fe-2S, 4Fe-4S) characteristic of the particular protein and coordinated to critical cysteines in the primary peptide backbone (2). Within cells, iron availability for synthesis of iron-sulfur proteins and other biological functions must be tightly regulated, because excess iron is toxic (3). Excess iron leads to free radical reactions that damage membranes, proteins, and DNA (4). Here we describe a regulatory control mechanism that coordinates iron uptake, iron distribution, and the levels of iron-sulfur cluster proteins in the eukaryote Saccharomyces cerevisiae. The regulator responsible for these effects is Nfs1p.
Examination of the S. cerevisiae genome data base reveals that Nfs1p is the single yeast homolog of bacterial IscS (5,6) and NifS (7). There is strong evidence, both biochemical and genetic, showing that the bacterial protein NifS mobilizes sulfur from cysteine and mediates Fe-S cluster assembly. Bacterial mutants of NifS were found to be deficient in the assembly of both Fe protein and MoFe protein subunits of nitrogenase (8,9). NifS through its enzymatic activity was found to reactivate the apo form of nitrogenase in which the Fe-S cluster was removed by chelation (10). Elegant biochemical work has elucidated this catalytic process: NifS was shown to be a pyridoxal phosphate-containing homodimer that catalyzes the formation of elemental sulfur from L-cysteine (7). A conserved lysine residue in the bacterial NifS protein (equivalent to Lys-299 in the yeast protein) is the covalent attachment site for pyridoxal phosphate. A conserved cysteine residue (equivalent to Cys-421 in the yeast protein) forms a catalytic persulfide intermediate involved in sulfur abstraction from the L-cysteine substrate (11). In this work, we demonstrate that the yeast homolog, Nfs1p, is an essential mitochondrial protein. We show that Nfs1p is required not only for activities of Fe-S cluster proteins, but also for regulatory effects on cellular iron metabolism. Regulatory effects resulting from decreased levels of Nfs1p include up-regulation of iron uptake to the cell and distribution of cellular iron to mitochondria.

EXPERIMENTAL PROCEDURES
Growth Media-Methods for yeast manipulations and growth media have been described (12). For experiments with different concentrations of metals, standard defined medium was modified by addition of iron and copper. Iron was added as ferric ammonium sulfate and copper was added as copper sulfate. For some experiments, in order to avoid catabolite repression, the yeast was grown in defined medium with 2% raffinose as the carbon source. For some experiments, the yeast was grown in rich medium (2% yeast extract, 2% peptone) with 2% raffinose (YPR) or 2% lactate (YPL) as the carbon source. The GAL1 promoter was induced by the addition of 0.5% galactose and repressed by the addition of 0.2% glucose to YPR medium.
Yeast Strains-Strains 81 and 61, containing the FRE1-HIS3 integrated fusion and the parental strains CM3260 and CM3262 have been described (13). The diploid CM3263 was obtained by mating CM3260 and CM3262. MA14 was derived from the strain 81 after selection on defined medium supplemented with 10 M copper sulfate and 20 M ferric ammonium sulfate. Strains x316 -1A-MA14 and x404-6D-MA14 were derived from MA14 by back-crossing with the wild-type strains CM3260 or YPH499, respectively. Heterozygote knock outs of NFS1 were constructed in strains CM3263 and YPH501 by transforming with the construct, nfs1KOUra. Sporulation yielded only uracil auxotrophs with 2 of 4 viable tetrad clones. Diploid wild-type strain D273-10B (ATCC 24657) was used for testing import into isolated mitochondria. AFT1 was interrupted in YPH499 and x404-6D-MA14 by transforming with the XhoI-KpnI fragment of pT20 (14) and selecting for tryptophan prototrophy, creating strains 499⌬aft1 and 6D⌬aft1.
Plasmids and DNA Constructions-pA95, pA96, and pB48 contained inserts of genomic DNA in the vector YCp50 and were selected by screening a library (15) for complementation of MA14. pB48⌬ SalI was constructed by digesting pB48 with SalI and religating, thus removing the genomic fragment between the SalI site within NFS1 and the SalI site in the vector. pB48⌬EcoRI was constructed by digesting pB48 with EcoRI and religating, thus removing the genomic EcoRI fragment. The genomic 3.4-kilobase EcoRI fragment contained the NFS1 ORF 1 and 1133 base pairs of 5Ј-flanking region, including a portion of the adjacent YCL016c ORF. This fragment was subcloned into pRS406, creating plasmid pRS406-R1(NFS1) and into pRS416 creating plasmid pRS416(NFS1). Plasmid pRS406-R1(nfs1-fs) contained a frameshift introduced into the NFS1 open reading frame by digesting with AflII, rendering the ends blunt with Klenow polymerase, and religating. Plasmid pRS416(NFS1) linearized with SphI and PflMI was used to recover the mutant allele (nfs1-14) from the strain x404-6D-MA14, as described in the gap repair method (16). The mutated base pair was identified by dideoxy DNA sequencing of the NFS1 ORF. The NFS1 interruption/deletion constructs nfsKOUra and nfsKOHis were made by subcloning the NFS1 EcoRI fragment into pBluescript SKϩ and inserting URA3 or HIS3 cassettes into the BglII sites within the coding region. For partial deletion and disruption of the gene, the modified genomic fragment was released by EcoRI digestion and used for transformation of yeast. For plasmid shuffle experiments, the NFS1 EcoRI fragment was subcloned in pRS318, which carries LEU2 and CYH2 genes for selection and counterselection, respectively. For regulated expression of Nfs1p in yeast, the ORF was amplified using polymerase chain reaction and Pfu polymerase, adding XbaI and XhoI sites at the 5Ј and 3Ј ends, respectively. The polymerase chain reaction product was subcloned into the XbaI and XhoI sites of pBluescript. The sequence was confirmed, and the XbaI-XhoI fragment was moved to the same sites of pEMBLyex4-i. In this integrating vector the NFS1 ORF was placed under control of the GAL1 promoter, creating Gal-Nfs1. For integration into the genome, this plasmid was linearized at a unique StuI site and integrated into ura3-52. Site-directed mutations were introduced into the NFS1 ORF in plasmid pRS406-R1(NFS1) using the QuickChange kit (Stratagene). The mutations involved changing codon 299 from AAG (Lys) to GCG (Ala), and changing codon 421 from TGT (Cys) to GCT (Ala).
Plasmid Shuffling-For plasmid shuffling, a diploid strain resistant to cycloheximide, YPH501(cyh2/cyh2), was generated as described (17). This strain was transformed with nfs1KOHis and the heterozygous knock out (NFS1/nfs1::HIS3, cyh2/cyh2) was transformed with plasmid pR318-R1(NFS1, LEU2) and sporulated. A tetrad clone with Leuϩ and Hisϩ growth, designated x478 -6C was identified and used in the experiments requiring plasmid shuffling. In this strain the chromosomal deletion of NFS1,⌬nfs1::HIS3, was covered by the NFS1 carried on the pRS318 plasmid with selectable (LEU2) and counterselectable (CYH2) markers. For ejection of the covering plasmid the strain was grown on media containing 10 g/ml cycloheximide. Additional constructs containing NFS1 or mutant alleles in plasmid pRS406 were integrated at the ura3-52 locus.
Bacterial Expression and Antibody Generation-For expression in bacteria, the NFS1 ORF was amplified with 5Ј BamHI and 3Ј EcoRI sites. The product was cloned into the corresponding sites of pGEX2T (Amersham Pharmacia Biotech) making an in-frame fusion with glutathione S-transferase (GST), and the sequence was confirmed. The Escherichia coli strain BL21(DE3) was transformed with the pGEX2T-NFS1 plasmid. Expression of GST-NFS1 was induced by 2 mM isopropyl-␤-D-1-thiogalactopyranoside for 4 -5 h at 30°C. The GST-Nfs1p fusion protein was purified on a glutathione-Sepharose column and eluted with 50 mM glutathione. The eluate was analyzed by polyacrylamide gel electrophoresis and found to contain a single major band by staining with Coomassie Blue. The band was cut from the gel and used to inject rabbits for generation of polyclonal antibodies. Other antibodies used as markers in the localization experiments were as follows: anti-Put2p was a rabbit polyclonal directed against Put2p, a mitochondrial matrix protein; anti-Por1p was a rabbit polyclonal against porin, the outer mitochondrial membrane protein; anti-Pgk1p (Molecular Probes) was a mouse monoclonal antibody directed against 3-phosphoglycerate kinase, a soluble cytosolic protein.
RNA Blotting and Protein Immunoblotting-Cultures grown in YPAD were harvested and RNA was isolated using hot acidic phenol (65°C, pH 5.5). The RNA was separated by formaldehyde-agarose gel, blotted to nitrocellulose, and hybridized with 32 P-labeled probes. The probes were labeled by random prime synthesis using fragments derived from the coding regions of the genes to be studied. The blot was washed at high stringency (65°C, 0.1 ϫ SSC, 0.1% SDS) and exposed to a PhosphorImager screen. For immunoblotting, proteins were separated by electrophoresis on polyacrylamide gels and electrophoretically transferred to nitrocellulose. The primary antibodies were rabbit polyclonal antibodies and the signal was developed using a goat anti-rabbit IgG peroxidase conjugate and the ECL kit (Amersham). Fractionation of Cells-Mitochondria were isolated from yeast (18). The cytosolic cell fraction was derived from a concentrated translationcompetent yeast lysate. This lysate was centrifuged at 386,000 ϫ g for 20 min and the supernatant was designated as cytosol.
Fractionation of Mitochondria-Intact mitochondria were suspended in 20 mM Hepes-KOH, pH 7.5, 0.6 M sorbitol. For releasing the soluble contents of the intermembrane space, the mitochondria were subjected to hypotonic shock by diluting the sorbitol to 0.1 M and incubating at 0°C for 10 min (19). The mitoplasts were separated from the outer membrane and intermembrane space components by centrifuging at 12,000 ϫ g. The mitoplasts were solubilized with 0.5% Triton X-100 and the lysate was centrifuged for 10 min at 15,000 ϫ g. For tracking the 55 Fe label in mitochondria, aliquots of the fractions (mitochondria, intermembrane space, mitoplasts, Triton X-100 mitoplast supernatant, and Triton X-100 insoluble mitoplast pellet) were suspended in scintillation mixture and counted in a Beckman scintillation counter.
Mitochondrial Import-The NFS1 ORF was amplified between NdeI (5Ј) and XhoI (3Ј) sites and cloned into pSP64T (20). Radiolabeled Nfs1p preprotein was generated by transcription followed by translation in the presence of a mixture of [ 35 S]methionine and [ 35 S]cysteine, and the radiolabeled product was used in a mitochondrial import study (21). Briefly, reactions containing 100 g of mitochondria were initiated by adding the radiolabeled preprotein. Import reaction mixtures contained 4 mM ATP and 1 mM GTP. Following import at 20°C for 15 min, reaction mixtures were treated with trypsin (0.1 mg/ml) for 30 min at 0°C. The protease was inactivated and the samples were analyzed by SDS-polyacrylamide gel electrophoresis and exposed to film.
Assays-The assay for ferric reductase was a filter lift assay (22), modified by the addition of 50 M copper sulfate and 10 M ferric ammonium sulfate to YPD agar plates for growth of the colonies to be assayed. Under these conditions the wild-type exhibited repressed activity (RedϪ) and the MA14 strain was strongly positive (Redϩ). Assays for ferric reductase and high affinity ferrous iron uptake have been described (13). Aconitase was assayed by measuring the formation of cis-aconitate at 240 nm as described (23). The reaction mixture of 1 ml consisted of 20 mM iso-citrate, 90 mM Tris-HCl, pH 8.0. Purified mitochondria (150 g) were solubilized with 0.5% Triton X-100 and added to the assay mixture. The ⌬A 240 nm was measured for 2 min at room temperature. The extinction coefficient for cis-aconitate is 3.6 mM Ϫ1 and the activity was expressed as nanomoles of cis-aconitate formed per mg of mitochondria/min. Succinate dehydrogenase was assayed by following the reduction of p-iodonitrotetrazolium violet (INT) to the INTformazan (24). For the assay reactions, 300 l of 0.01 M succinate was mixed with 100 l of 2.5 mg/ml INT solution, and the solubilized mitochondria (150 g) in 0.5% Triton X-100 was added. The reaction was incubated at 37°C for 5-8 min and stopped with ethyl acetate: ethanol:trichloroacetic acid in 5:5:1 ratio (v/v). The reaction tube was centrifuged for 1 min at 15,000 ϫ g and the OD 490 of the supernatant was measured. For control reactions, the succinate substrate was substituted with 300 l of 0.05 M sodium phosphate, pH 7.5. The extinction coefficient for INT-formazan is 19,310 M Ϫ1 at 500 nm and the activity of succinate dehydrogenase was expressed as nanomoles of INT-formazan formed per mg of mitochondria/min (24).
Chemical Assay for Iron-Mitochondria were lysed with 0.6% SDS in 0.1 M sodium phosphate buffer, pH 7.2. Iron was reduced by dithionite and the indicator bathophenanthroline disulfonic acid was added at 10 mM final concentration. The colored ferrous iron-bathophenanthroline disulfonic acid complex was measured by determining the absorption at 520 nm, and mitochondrial iron content was calculated from a standard curve.

New Complementation Group Identified Using FRE1-HIS3
Selection for Mutants of Iron Metabolism-A selection for mutants involved in cellular iron homeostasis was modified from our previous method (25). The FRE1 gene encodes the major cell surface reductase and is required for iron acquisition from ferric iron chelates. FRE1 is active when cells are starved for iron, and inactive when cells are replete. To isolate mutants unable to respond to environmental iron, the FRE1 promoter was fused to the HIS3 coding region, and this construct was integrated into the chromosome of a ⌬his3 strain. The strain carrying the integrated FRE1-HIS3 marker on the chromosome was cultured in rich (YPAD) medium and a dilute inoculum was spread on agar plates of the same composition, allowing colonies to arise from single cells. The colonies were replicated to selective plates consisting of defined medium lacking histidine and supplemented with iron (20 M ferric ammonium sulfate). Under these conditions, uptake of the metal could proceed by genetically distinct high and low affinity uptake systems (26). Thus, mutations abrogating only one system were not selected, and instead, regulatory mutants and mutants affecting metal distribution were identified. Most colonies did not grow at all on the selection plates, but after 4 days, a small number of colonies emerged as papillae on a background of non-growing cells. The growing cells were transferred to rich (YPAD) plates. One of the mutants selected in this manner, MA14, was for further genetic analysis. The mutant exhibited non-repressed surface reductase, consistent with the selection scheme (Redϩ for non-repressed reductase). Crossing with the parental strain of the opposite mating type revealed that the mutation was recessive, i.e. the diploid was RedϪ. Sporulation of this diploid revealed that a single locus was involved, because the reductase phenotype segregated as 2ϩ:2Ϫ in 20 tetrads.
Several other mutants with similar phenotypes were selected, including numerous isolates of ssq1 mutants (27). Crossing of MA14 with 191-33C which carries a mutant allele of SSQ1, yielded diploid strains with completely normal iron metabolism. Iron uptake, surface reductase, and regulation of the above diploid strain were normal, indicating that the MA14 and SSQ1 mutant loci were not allelic. Similar analyses with mutants at other loci implicated in iron homeostasis were performed. Crosses with ⌬atm1 (28) and ⌬yfh1 (29) strains ruled out that MA14 was mutated at those loci.
Altered Cellular Iron Homeostasis in MA14 Mutant: Dependence on Aft1p-We wondered if the dysregulation of the FRE1 promoter fusion in the MA14 mutant reflected a more general alteration in cellular iron homeostasis. In wild-type yeast, uptake of iron from ferric chelates in the medium requires the concerted action of several proteins (30). The surface reductases encoded by FRE1 and FRE2 act on external ferric iron chelates, releasing ferrous iron which can then be transported by an iron transport complex. The latter consists of the FTR1 permease and FET3 multicopper oxidase (31,32). The genes for the proteins involved in iron uptake are coordinately regulated at the level of transcription by AFT1, an iron sensor-regulator (33). Iron deprivation induces AFT1-dependent transcription of the target genes. The mutant grown in standard rich (YPAD) medium was evaluated for up-regulation of the genes controlled by AFT1. As shown in Fig. 1A, the mRNA levels for surface reductases, FRE1 and FRE2, were increased in the MA14 mutant compared with the wild-type. FRE1 expression was present in the wild-type and increased roughly 2-fold in the mutant. FRE2 expression was undetectable in the wildtype and highly induced in the mutant. The different effects on FRE1 and FRE2 are probably due to the dual regulatory control of FRE1, which is dependent on both iron (14) and copper (34) levels, whereas FRE2 is dependent only on iron levels (35). The FET3 and FTR1 transcripts were also increased in the mutant. The extent of up-regulation of these components was similar to that observed in an AFT1-1 up strain, M2, in which the iron sensor-regulator remains constitutively induced due to a point mutation (Fig. 1A).
To investigate the role of AFT1 in the induction of the cellular iron uptake system in the MA14 mutant, the AFT1 gene was interrupted in the MA14 strain. Surface ferric reductase activity was evaluated in the double mutant. The increment in ferric reductase in MA14 was found to be partially dependent on the presence of an intact copy of AFT1 (Fig. 1B). However, FIG. 1. Dysregulated cellular iron uptake in MA14 mutant. Panel A, RNA blot showing increased transcripts for genes of the high affinity cellular iron uptake system in the MA14 mutant. Yeast strains, YPH499 (WT), x404-6D-ma14 (MA14), M2 (AFT1-1 up ), were grown to logarithmic phase in rich medium (YPAD) and total RNA was isolated. RNAs (30 g/lane) from the three strains were separated by formaldehyde-agarose gel and blotted to nitrocellulose. The blots were probed with random primed 32 P-labeled DNA from the reductase genes FRE1 and FRE2, the permease gene FTR1, the multicopper oxidase FET3, and ACT1 as a control. Panel B, reductase and high affinity ferrous iron uptake are increased in MA14 mutant dependence on AFT1. Strains YPH499 (WT, AFT1), x404-6D-MA14 (MA14, AFT1), 499⌬aft1 (WT, ⌬aft1), and 6D⌬aft1 (MA14, ⌬aft1) were grown in YPAD prior to measuring ferric reductase and ferrous iron uptake. Results are averages of triplicate determinations with standard deviation less than 15%. Panel C, dysregulation of high affinity ferrous iron uptake in response to media iron levels in MA14 mutant. Yeast strains YPH499 (WT) and x404-6D-MA14 (MA14) were grown in defined medium with different concentrations of iron added as ferric ammonium sulfate. Copper was present at 1 M in the medium. The cells were grown to stationary phase, diluted 20-fold, and allowed to enter logarithmic growth prior to assay of ferrous iron uptake. Results are mean values of triplicate determinations with standard deviations less than 15%.
increased activity was still observed in the double mutant compared with the wild-type (compare WT AFT1 with MA14 ⌬aft1 in Fig. 1B). This effect could be due to regulators other than AFT1 that are capable of inducing FRE1 or FRE2 expression. Regulation of FRE1 by the MAC1 regulator has been described (34,36), and an AFT1 homologous gene present in the yeast genome (YPL202C) might also mediate the increased ferric reductase expression. By contrast, the increment in high affinity ferrous iron uptake observed in the MA14 mutant was completely dependent on AFT1 (Fig. 1B). The AFT1 sensor regulator is involved in homeostatic responses of the cell to environmental iron. Therefore, the response of the MA14 mutant to a more graded and controlled exposure to iron was tested. The wild-type and the mutant were grown in defined media with varying concentrations of iron, and high affinity ferrous iron uptake was measured. The results show that at every medium iron concentration tested, the iron uptake activity was higher in the MA14 than the wild-type (Fig. 1C). The wild-type exhibited a progressively decreasing iron uptake rate with increasing iron exposure, as would be predicted for a homeostatically regulated process. The magnitude of this regulation was about 5-fold between lowest and highest iron exposures. By contrast the magnitude of the iron uptake regulation by the MA14 mutant was only 2-fold. Together these data suggest that in the mutant, Aft1p was not responding appropriately to iron in the environment and in the cell. Possible explanations for this are that intracellular iron was sequestered away from Aft1p or that a factor required for the Aft1p iron response was lacking.
Altered Mitochondrial Iron Homeostasis in MA14 Mutant: Mitochondrial Iron Sequestration-The observation that the rate of cellular iron uptake was increased in the MA14 mutant led us to wonder how the excess iron was distributed within the mutant cells. To address this question, the wild-type and MA14 strains were cultured in media containing different concentrations of iron to which radioactive 55 Fe was added as a tracer. After 16 h of growth during which the labeling of intracellular iron pools reached a steady state, mitochondria were isolated from the radiolabeled cells. The iron levels in mitochondria isolated from wild-type and MA14 were indistinguishable at the lowest iron medium concentration of 0.1 M (.512 versus .541 pmol/g, Table I). This value is comparable to previously published values for yeast (37) and mammalian (38) mitochondrial iron levels. However, when the medium iron concentration was increased, MA14 mutant mitochondria accumulated dramatically more iron. At 1 M the iron in the MA14 mitochondria was 18.6 (pmol/g), at 5 M the level was 61.5, and at 50 M the level was 75.3, whereas in the wild-type the maximum level achieved was 4.9 (pmol/g) ( Table I). The distribution of the iron within the mitochondria was also examined (Fig. 2). The iron content of the intermembrane space was evaluated after hypotonic shock that disrupts the outer mem-brane. No differences between wild-type and mutant were observed. The mitoplasts, consisting of intact inner membrane and matrix, were evaluated. Most of the iron in the MA14 mutant mitochondria was found in the mitoplasts. The mitoplasts were then lysed with 0.5% Triton X-100 and separated into a soluble supernatant fraction and an insoluble pellet fraction. Remarkably, for the mutant grown in iron concentrations of 1 M or greater, most of the mitochondrial iron (53-60%) remained in the Triton X-100-insoluble pellet fraction. By contrast, only a small proportion of the iron (5-12%) was found in this fraction isolated from wild-type cells (Table I, Fig. 2). The mitochondrial fraction resistant to detergent solubilization would be expected to include membranes, large protein complexes and aggregates. The very marked increase of iron distributed to this fraction in the MA14 mutant suggests that the mutant phenotype alters iron trafficking to the mitochondria and iron solubility within that compartment.
Decreased Iron-Sulfur Protein Activities in the MA14 Mutant-Iron entering the mitochondria might have different final destinations: insertion into heme for use in heme proteins, insertion into Fe-S clusters for use in Fe-S cluster proteins, or retention in a matrix pool of unutilized iron. Mitochondria were isolated from the wild-type and the MA14 mutant after exposure to increasing media iron and were compared for the status of these iron pools. The results showed that as iron levels in the medium increased, the iron in the mitochondria of the MA14 mutant increased (see Fig. 2, iron ranging from 0.5 to 75 pmol of Fe/g of mitochondrial protein), but heme proteins were largely unaffected. We evaluated cytochrome c levels as an indicator of the status of heme proteins in mitochondria. No changes of cytochrome c, evaluated by specific antibody, or heme, evaluated by pyridine hemochrome spectra (39), were noted in the MA14 mutant (not shown). By contrast, Fe-S protein activities were compromised at all media iron concentrations (Fig. 3). Aconitase is a soluble single subunit enzyme of the mitochondrial matrix and catalyzes the conversion of citrate to isocitrate. The enzymatic activity is mediated by substrate interaction with a 4Fe-4S complex on the active surface of the protein (40). In the MA14 mutant, aconitase activity was decreased to 17-33% of the wild-type level (Fig. 3). Succinate dehydrogenase activity was also measured. This enzyme consists of four nuclear encoded subunits assembled as a complex in the inner mitochondrial membrane with FAD and Fe-S cofactors (41). Succinate dehydrogenase activity was markedly decreased in the MA14 mutant, to 18 -53% of the wild-type level (Fig. 3). The increased mitochondrial iron levels in the MA14 mutant (Fig. 2) had very little, if any, effect on restoring Fe-S protein enzymatic activities. On the other hand, the increased mitochondrial iron did not exacerbate the deficiencies (Fig. 3).
Cloning the Wild-type Allele for the Mutant Gene in MA14: Identification as NFS1-A wild-type genomic library (15) was screened for the ability to complement the non-repressed ferric reductase phenotype of the MA14 mutant. Three independently isolated plasmids with complementing activity were isolated and found to contain identical 9470-base pair genomic DNA inserts (B48 in Fig. 4). Complementation activity for both the reductase (Redϩ) and ferrous uptake phenotypes was localized to the NFS1 open reading frame, and a frameshift introduced at the AflII site in this reading frame abrogated activity (Fig. 4). Theoretically, complementation by the NFS1containing genomic DNA fragment could be indirect. To rule this out, the EcoRI genomic fragment containing NFS1 was marked with URA3 and integrated into the unique MscI site of the chromosomal NFS1 in a wild-type strain. This strain was evaluated in a cross with MA14. Tetrad clones derived from the cross were RedϪ and Uraϩ (wild-type phenotype) or Redϩ and UraϪ (mutant phenotype) (Fig. 4). The absence of recombination between the marked wild-type NFS1 and the MA14 mutant phenotype suggests that they are allelic. Thus, complementation data and meiotic mapping both supported that the mutation in MA14 was in NFS1. We next proceeded to rescue the mutant allele, called nfs1-14, by gap repair and to determine its DNA sequence. A single nucleotide change was found within the coding region compared with the wild-type se-quence. Nucleotide 572 was changed from T to G, altering codon 191 from ATC (isoleucine) to AGC (serine). The mutation resulted in changing a conserved residue (Fig. 5). The altered amino acid was conserved in the bacterial IscS sequence (Fig.  5) and predicted to be present in the mature protein rather than in the mitochondrial leader. The location of the mutated amino acid in the primary sequence was distant from the pyridoxal phosphate-binding domain and the active site cysteine involved in sulfur transfer. DNA sequencing of the mutant and wild-type NFS1 alleles revealed another deviation from the data base sequence. Nucleotide 448 was a T in the data base and C in the sequence of both the wild-type and mutant alleles rescued from strain 81 and the congenic MA14. The sequence of the wild-type allele cloned from the library also had a T at this position. Thus the sequence difference with the data base may represent a polymorphism due to strain variation or a sequence error. The effect of the nucleotide change would be to replace tyrosine at position 150 of the ORF in the data base sequence with histidine. Neither amino acid resembles the bacterial IscS sequence, which has a glutamine in this position (Fig. 5).
Identification of NFS1 as an Essential Gene and Plasmid Shuffle to Evaluate NFS1 Mutant Alleles-The effects of deleting NFS1 from the haploid genome were evaluated in three different ways, and in each case, the gene was found to be required for viability. An interruption-deletion construct was designed that removes the central portion of the NFS1 ORF including the putative pyridoxal phosphate-binding site and replaces it with a HIS3 marker (Fig. 5). Transformation of two different haploid strains (CM3260, YPH499) and selection for histidine prototrophy yielded no viable transformants. Diploid transformants in both backgrounds (CM3263, YPH501) were sporulated, and the tetrads showed a 2ϩ:2Ϫ pattern for viability with all the viable tetrads being auxotrophic for histidine. The tetrad clones carrying the interruption/deletion were presumed to be non-viable or unable to germinate. Finally, a plasmid shuffle strategy (see below) confirmed the essentiality of the NFS1 gene.
Methodology for studying essential genes using plasmid shuffling has been described (17). Briefly, a shuffle strain was constructed. The NFS1 gene was interrupted with HIS3 in a cycloheximide-resistant diploid strain YPH501 (cyh2/cyh2). Plasmid pRS318, carrying the NFS1 wild-type allele, a LEU2 marker gene, and the CYH2 gene for counterselection, was transformed into the nfs1::HIS3/NFS1 heterozygous diploid,  55 Fe radionuclide was added as a tracer. After 16 h of growth, the mitochondria were isolated and an aliquot equivalent to 100 g of protein was fractionated according to standard protocols (19). The iron concentration was calculated for each fraction. The fractions were mitochondria, intermembrane space (IMS), mitoplasts, Triton X-100 supernatant, and Triton X-100 insoluble pellet. The intermembrane space contents were released by hypotonic shock in 0.1 M sorbitol for 10 min. The mitoplasts were solubilized in 0.5% Triton X-100 and separated into supernatant and pellet fractions by centrifugation at 15,000 ϫ g for 10 min. and the diploid transformant was sporulated. A spore clone x478-6C with the nfs1::HIS3 allele covered by the plasmid borne NFS1 was identified by its Hisϩ and Leuϩ growth. Into this shuffle strain, we integrated various NFS1 constructs targeted to the ura3-52 locus on pRS406. These transformants thus carried an inactivated chromosomal copy of NFS1, a sec-ond modified copy of NFS1 integrated at the ura3-52 locus, and a wild-type NFS1 plasmid with counterselectable marker. When these strains were exposed to cycloheximide, the covering plasmid was ejected and the phenotype resulting from the modified copy of NFS1 at the ura3-52 locus could be evaluated.
The results confirmed that NFS1 was essential. When the covering plasmid was ejected leaving the empty vector pRS406 at the ura3-52 locus and no functional copy of NFS1, the cells were rendered non-viable (Fig. 6, row 1). When NFS1 or nfs1-14 was inserted at ura3-52 and then uncovered, the cells were viable (Fig. 6, rows 2 and 5). The nfs1-14 mutant was slightly impaired in its growth rate compared with the wildtype (compare rows 2 and 5). We expected that since the only copy of NFS1 in the shuffle strain after cycloheximide treatment was the nfs1-14 mutant allele, this strain would recapitulate the phenotype of the MA14 strain. Indeed, high affinity ferrous iron uptake in this strain was significantly increased (24 pmol/10 6 cells/h for nfs1-14 versus 2.5 for NFS1). The increment was not as marked as in the original MA14 mutant, but this could be due to genetic background differences and different culture conditions. Point mutations were constructed in the ORF for NFS1 (depicted in Fig. 5). A frameshift was introduced at a unique AflII site, creating nfs1-fs. The pyridoxal phosphate-conjugating lysine was changed to alanine (K299A) by site-directed mutagenesis. In a separate construct, the critical active site cysteine was mutated to alanine (C421A). Each of these mutant forms of NFS1 was unable to rescue the knock out in the shuffle assay (Fig. 6, rows 3, 4, and 6). The regulated GAL1 promoter is repressed by glucose and induced by galactose as the carbon source. The NFS1 coding region placed under GAL1 control was able to rescue the shuffle strain when induced by galactose (Fig. 6, row 7Ј, YPAG ϩ CHX) but not when repressed by glucose (Fig. 6, row 7, YPAD ϩ CHX).
Nfs1p Localized to Mitochondria-Alignment of the predicted amino acid sequence of the NFS1 ORF with the bacterial IscS homolog revealed a large (98 amino acid) amino-terminal extension of the Nfs1p sequence (Fig. 5). Features of the initial portion of this sequence resemble a mitochondrial leader sequence. To directly test whether the Nfs1p preprotein contained a leader sequence that could mediate import into mitochondria, the protein was synthesized and radiolabeled in reticulocyte lysate. The Nfs1 precursor was incubated with purified mitochondria, and the mitochondria were reisolated and analyzed by SDS-gel electrophoresis. In the presence of mitochondria, the precursor (p) was imported and proteolytically processed to a mature species (m) (Fig. 7A, compare lanes  1 and 2). The generation of the mature product was dependent Ϫ, in the complemented mutant) and by ferrous iron uptake (high in the mutant and normal in the complemented mutant). Complementing activity was present in the EcoRI fragment as shown (pRS406-RI(NFS1)), but not in the same fragment into which a frameshift was introduced in the NFS1 ORF (pRS406-RI(nfs1-fs)). For meiotic mapping, the pRS406-RI(NFS1) plasmid was integrated at MscI in the genome of a wild-type strain, YPH500, placing the URA3 marker close to this site. This marked strain was crossed with the MA14 mutant and sporulated. Sixteen tetrads or 64 spore clones were analyzed for ferric reductase activity (Redϩ) and uracil prototrophy (Uraϩ). Shown are the numbers of spore clones in each category.
FIG. 5. Sequence features of Nfs1p. The predicted Nfs1p was aligned with the bacterial protein IscS using ClustalW. Features of the alignment are 55% identity and 71% similarity with the bacterial protein, excluding the amino terminus. There is a large amino-terminal overhang present in Nfs1p, which is absent in the bacterial protein. A predicted mitochondrial leader sequence for Nfs1p is shown with underlining of the predicted cleavage site between residues 33 and 34. A clear downward arrow indicates a sequence difference with the data base. Codon 450 is TAC (codes for Y) in the data base but is CAC (codes for H) in our sequence. A black downward arrow indicates the location of the frameshift introduced at the AflII site. The change resulting from the point mutation in the nfs1-14 allele is in bold. Codon 191 was changed from ATC (codes for I) to AGC (codes for S). The ⌬nfs1 deletion construct which removed amino acids between two BglII sites is depicted by overlining. The conserved pyridoxal phosphate-binding lysine (Lys-299) is in bold. This was mutated to alanine for some experiments. The surrounding domain includes residues (*) that fit a consensus for pyridoxal phosphate (62). The conserved Cys-421 is bold and highlighted by a black circle above. This was mutated to alanine for some experiments.
on the presence of a membrane potential and was completely inhibited by valinomycin (Fig. 7A, compare lanes 2 and 3). Judging from the approximate size of the precursor and mature forms, the leader sequence removed by the processing cleavage was ϳ3 kDa. This would be consistent with removal of a 33amino acid leader sequence with arginine residues in Ϫ2 and Ϫ3 positions with respect to the processing cleavage site (Fig. 5, putative cleavage site is underlined). The mature processed protein (m) formed after import was protected from externally added protease (Fig. 7A, lane 4).
The predicted mitochondrial localization of the mature Nfs1p was evaluated directly using a monospecific polyclonal antibody generated against the bacterially expressed protein. Enriched cytoplasmic and mitochondrial fractions were separated on polyacrylamide gel, blotted to nitrocellulose, and examined by immunoblotting with marker antibodies for subcellular compartments or with the anti-Nfs1p antibody (Fig. 7B). The marker antibody studies confirmed the effectiveness of the fractionation procedure. Anti-Put2p, directed against a mitochondrial matrix marker protein, reacted only with the mitochondrial fraction. Anti-Pgk1p (anti-phosphoglycerokinase), directed against a soluble cytoplasmic enzyme, reacted only with the cytoplasmic fraction. The results show that Nfs1p was present in mitochondria. No signal for Nfs1p was detected in cytoplasmic fractions, even with an overloaded gel (Fig. 7B). The size of the mature Nfs1p by immunoblotting was consistent with the predicted size for the processed mitochondrial form (51 kDa) and migrated more rapidly than the purified full-length preprotein expressed in bacteria (not shown).
Nfs1p Deficient in MA14 Mutant Strain-Nfs1p was evaluated in mitochondria isolated from two strains bearing the nfs1-14 mutation. Initial examination by immunoblotting revealed no signal at all from the mutant mitochondria (Fig. 7C,  anti-Nfs1p (1Ј)). Only upon overexposure of the film could we discern a weak signal, approximately 1/50 of the intensity of the wild-type (Fig. 7C, anti-Nfs1p (30Ј)). This weak signal in the mutant migrated at the same position as the wild-type protein at roughly 51 kDa, suggesting that the minimal residual Nfs1p in the mutant was processed to the mature form (Fig.   FIG. 6. Shuffle of NFS1 alleles. Strain x478-6C contains a chromosomal deletion of NFS1 covered by the wild-type NFS1 gene carried on plasmid pRS318. The latter contains the counterselectable CYH2 marker. This strain was transformed with various constructs targeted to the ura3-52 locus. These constructs were: 1, pRS406, empty vector; 2, pRS406-RI(NFS1), carrying the wild-type NFS1 allele; 3, pRS406-RI(K299A), carrying a mutant allele in which the lysine at position 299 was mutated to alanine; 4, pRS406-RI(C421A), carrying a mutant allele in which the cysteine at position 421 was mutated to alanine; 5, pRS406(nfs1-14), carrying the mutant allele rescued from MA14; 6, pRS406(nfs1-fs), carrying the NFS1 gene with a frameshift in the ORF; 7 and 7Ј, Gal-Nfs1, carrying the NFS1 ORF under the control of the GAL1 promoter. The transformants were plated on rich medium (YPAD) or the same medium with 10 g/ml cycloheximide added to eject the covering plasmid (YPAD ϩ CHX). For the Gal-Nfs1 transformant, the covering plasmid was also ejected on medium containing 0.5% galactose and 2% raffinose (YPGal ϩ CHX). Serial 10-fold dilutions of 10 6 cells are shown. FIG. 7. Nfs1p, mitochondrial localization and decrease in MA14. Panel A, import and processing of Nfs1p preprotein into isolated mitochondria. The Nfs1p preprotein was synthesized and radiolabeled with [ 35 S]methionine in reticulocyte lysate. The precursor was incubated with 100 g of isolated mitochondria from strain D273-10B (Mito, lanes [2][3][4][5] in the import reactions. In some cases, 5 g/ml valinomycin (Val, lanes 3 and 5) was added to the mitochondria prior to the addition of the precursor. After 15 min, samples were exposed to 0.1 mg/ml trypsin (Trypsin, lanes 4 and 5), and the protease was inactivated. The mitochondria were retrieved by pelleting at 12,000 ϫ g and analyzed by SDS-polyacrylamide electrophoresis. The Nfs1p precursor (p) and mature forms (m) are indicated. Panel B, mitochondrial localization of Nfs1p. Cytoplasmic (Cyto) and mitochondrial (Mito) fractions were isolated from strain D273-10B grown to late logarithmic phase in YPL medium. Three pairs of identical cytoplasmic and mitochondrial samples (100 g/lane) were separated by polyacrylamide gel electrophoresis and transferred to nitrocellulose. Each pair of lanes was cut from the nitrocellulose and exposed to a single antibody: anti-Pgk1p (3-phosphglycerate kinase), a mouse monoclonal antibody used at 1:5000 dilution; anti-Put2p, a rabbit polyclonal antibody used at 1:1000; anti-Nfs1p, a rabbit polyclonal antibody used at 1:1000. The secondary antibodies used were goat anti-mouse or goat anti-rabbit peroxidase, and the signal was developed with ECL (Amersham Pharmacia Biotech). Molecular mass markers in kDa are shown to the left and right of the panel. The mobilities of the major bands were 45 kDa for PGK, 61 kDa for Put2p, and 51 kDa for Nfs1p. Panel C, decreased Nfs1p in MA14. Mitochondria were isolated from strains YPH499 (WT), x316-1A (M1), and x404-6D (M2) grown to late logarithmic phase in YPR medium. Each of the latter two strains carried the nfs1-14 mutation. The mitochondria were separated on a polyacrylamide gel, blotted to nitrocellulose, and probed with rabbit polyclonal antibodies to Nfs1p or Por1p (porin). The Nfs1p blot was exposed for 30 min (30Ј) or 1 min (1Ј). Molecular mass markers in kDa are shown for the long exposure of the Nfs1 blot. The mobilities of the major bands were 51 kDa for Nfs1p and 30 kDa for porin. 7C). A second more weakly reactive band migrating just above the 45-kDa marker likely represents a Nfs1p degradation product. Porin, a mitochondrial outer membrane protein, was not affected in the mutant (Fig. 7C).
Repressing Expression of Nfs1p from the GAL1 Promoter Recapitulates the MA14 Phenotypes-The NFS1 ORF protein was placed under control of the inducible GAL1 promoter, and the wild-type allele was replaced with this regulated construct using the shuffle strategy. This strain grew when maintained under inducing conditions for Nfs1p expression but stopped growing when the inducer was removed and Nfs1p was no longer expressed (see Fig. 6). We examined in detail the events occurring during the shift from inducing to noninducing conditions. Nfs1p disappeared with a half-life of approximately 1 h under these conditions (data not shown). Time points included immediately before shifting to galactose and at 2.5, 5, 7.5, and 10 h after the shift. At each time point, Nfs1p level, cellular iron uptake rate, and mitochondrial iron level were measured. Finally the activities of iron-sulfur proteins, succinate dehydrogenase, and aconitase, were assayed on purified mitochondria. The results showed (Fig. 8) that as Nfs1p declined, iron uptake to the cell was perturbed in the opposite direction, increasing rapidly and reaching a plateau at a level 5 times the wild-type level. The absolute ferrous iron uptake rate was higher in the MA14 mutant, however, and this may relate to genetic background differences. Iron accumulating in the cell as a consequence of Nfs1p deficiency appeared in the mitochondria (Fig.  8, see the 10-h time point assayed for mitochondrial iron). The evidence of iron in the mitochondria only at this time point implies that iron accumulation is a late event, occurring only after iron-sulfur protein activities have been compromised.
The activities of iron-sulfur proteins decreased rapidly as Nfs1p levels declined. The kinetics of this decrease was remarkably similar for aconitase, a mitochondrial matrix enzyme with one subunit, and for succinate dehydrogenase, a multisubunit complex of the mitochondrial inner membrane (Fig. 8). A control experiment was performed in which the wild-type strain, YPH499, was subjected to identical growth conditions. Negligible changes (less than 10%) in Nfs1p levels, iron-sulfur protein activities, cellular iron uptake, or mitochondrial iron levels were observed (not shown). Thus, the decrease in the quantity of wild-type Nfs1p produces iron regulatory effects similar to those observed for the MA14 mutant. The iron regulatory phenotypes of the MA14 mutant are very likely due to the decreased amount of Nfs1p, although other allele specific effects are not completely excluded by these experiments.

DISCUSSION
Nfs1p is the yeast homolog of the bacterial proteins NifS and IscS. The bacterial proteins possess cysteine desulfurase activities that provide elemental sulfur from L-cysteine for Fe-S cluster synthesis (5,7). We have shown here in two ways that yeast Nfs1p is required for activity of Fe-S proteins. First, the mutant strain carrying the nfs1-14 allele was found to have a reduced amount of Nfs1p, and both aconitase and succinate dehydrogenase activities were deficient in this strain. These enzymes are Fe-S proteins of the mitochondrial matrix and inner membrane, respectively. Second, repression of Nfs1p expression using the regulated GAL1 promoter led to rapid decline of both aconitase and succinate dehydrogenase activities.
In this work, the NFS1-encoded protein was localized to mitochondria, and the gene was found to be essential for viability. Point mutations in the conserved pyridoxal phosphate lysine (Lys-299) or in the conserved active site cysteine (Cys-421), both predicted to disrupt cysteine desulfurase activity (5), resulted in cell inviability. Therefore, a critical Fe-S protein is likely to be required for viability. The specific target of this effect is not easily discerned, because most of the Fe-S proteins of yeast cells are non-essential. These include Fe-S proteins within mitochondria (Rieske iron-sulfur protein (42), succinate dehydrogenase (42), aconitase (40), homoaconitase (43,44), or outside mitochondria (3-isopropylmalate isomerase (45)). Other proteins that reside outside mitochondria and are likely to contain Fe-S clusters include an endonuclease III-like protein (46) and RNase L inhibitor ortholog (47). The localization of Nfs1p to mitochondria and its apparent absence from the cytoplasm raises an intriguing question. Is mitochondrial Nfs1p required for synthesis or maintenance of cytoplasmic iron-sulfur proteins? Leu1p is a cytoplasmic protein (45) with 3-isopropylmalate isomerase activity (48) that is likely to contain an iron-sulfur cluster. Although MA14 is not auxotrophic for leucine, further work will be required to evaluate if the activities of cytoplasmic iron-sulfur proteins are diminished in the MA14 mutant. The human homolog of Nfs1p has been localized to both cytoplasmic as well as mitochondrial cellular compartments (49), but we find no evidence of the yeast Nfs1p in the cytoplasm.
A new finding of this work is that Nfs1p regulates iron homeostasis. Thus, Nfs1p regulates both the activities of Fe-S cluster proteins and cellular iron uptake and distribution. The The NFS1 ORF was placed under control of the GAL1 promoter and the wild-type allele was replaced with this regulated construct. The strain was grown to mid-logarithmic phase under inducing conditions in 1.5 liter of YPR 0.5% galactose. The cells were pelleted by centrifugation and resuspended in YPR 0.2% glucose, which does not induce the GAL1 promoter. At 0 time (before the shift) and at 2.5, 5, 7.5, and 10 h, 750 ml of culture volume was removed and replaced with fresh media. High affinity ferrous iron uptake rate (picomole/10 6 cells/h) was measured using 55 Fe radionuclide. The standard assay was modified in that raffinose was substituted for glucose in the assay buffer, which consisted of 50 mM sodium citrate pH 6.5 and 2% raffinose. Mitochondria were purified and analyzed for Nfs1p and Por1p by immunoblotting. Mitochondria were analyzed for iron content using bathophenanthroline disulfonate as a color indicator. Aconitase activity (nanomole/mg/ min) and succinate dehydrogenase activity (nanomole/mg/min) were measured on purified mitochondria. decreased level of Nfs1p, due to the nfs1-14 point mutation (changing Ile to Ser at amino acid 191) or due to repressed expression from a galactose-regulated promoter, led to up-regulation of cellular iron uptake. The surface reductases, FRE1 and FRE2, were expressed at higher levels. The FET3 and FTR1 transcripts were present at higher levels, consistent with induction of the high affinity ferrous transport system. The cellular iron uptake system remained active under iron conditions that normally repress expression in a wild-type strain. These effects on cellular iron uptake were dependent upon Aft1p, the iron sensor-regulator. Nfs1p levels also affected iron distribution within the cell. The iron assimilated by the nfs1-14 mutant cells was retained in mitochondria and sequestered in a fraction that was easily sedimented and resistant to solubilization with Triton X-100. The biological function, if any, of this iron pool is not clear. By contrast, the cytoplasm was relatively depleted of iron as compared with the wild-type (data not shown). The time course following repressed expression of Nfs1p was remarkable. Reduction in Fe-S enzyme activities occurred prior to mitochondrial iron overload, implying that the primary changes affect Fe-S enzymes. Mitochondrial iron accumulation is thus likely to be a secondary effect. This interpretation is in conflict with the hypothesis proposed to explain Fe-S protein deficiencies in clinical iron overload states such as Friedreich ataxia. According to this hypothesis, primary iron overload causes oxidative stress by iron-catalyzed Fenton chemistry. Fe-S proteins represent critical targets for oxygenfree radicals (50) and lose activity as a consequence of the oxidative stress (51).
The mitochondrial location of Nfs1p suggests that a signaling pathway exists that couples mitochondrial events with cellular iron uptake. According to this speculative pathway, Nfs1p enhances the bioavailability of iron and its delivery to Fe-S proteins within mitochondria. The soluble iron, perhaps "sensed" through a specific regulatory Fe-S protein, then represses iron accumulation in the mitochondria. Signals must also emanate from the mitochondria to the plasma membrane iron uptake system, likely mediated via effects on Aft1p.
The involvement of Fe-S proteins in regulatory functions has been described recently (52). Fe-S clusters are well suited for functions involving uptake and donation of electrons, and in some cases changes in oxidation state may transduce regulatory signals. Oxidation of the E. coli SoxR protein by O 2 . induces reversible oxidation of the [2Fe-2S] ϩ clusters, and that in turn renders the DNA-bound protein competent to activate transcription. In this way, an environmental input alters gene expression (53). In another type of regulation, Fe-S clusters may be alternately destroyed and rebuilt as part of regulatory switching in response to environmental signals. The FNR protein in bacteria contains a [4Fe-4S] 2ϩ cluster that is required for transcriptional activity. Upon exposure to oxygen, the cluster is intially converted to a more oxygen stable [2Fe-2S] 2ϩ form and then becomes completely disassembled into the apo form. The latter forms of FNR are inactive as transcription factors (54). Perhaps the most relevant example of a regulatory iron-sulfur protein is the cytosolic aconitase/iron-regulatory protein iron sensor. In iron-deprived mammalian cells, the apo form of this protein (IRP1), binds to the iron response element in mRNA, inducing expression of the target genes. These include genes of the cellular iron uptake system such as the transferrin receptor. In iron replete cells, the Fe-S cluster of the regulator is rebuilt and inserted into the apo protein, inactivating the mRNA-binding function of the protein and activating the aconitase function. Thus, this iron-dependent regulatory switch depends on rebuilding the iron-sulfur cluster (55).
A protein homolog of IRP1 other than the mitochondrial acon-itase (Aco1p) is lacking from the yeast genome, nor has cytosolic aconitase activity been detected in yeast. However, the role for Nfs1p in yeast described here suggests that rebuilding iron-sulfur clusters in an iron regulatory protein (other than IRP1) may be involved in the control of cellular and mitochondrial iron homeostasis.
A key to further defining how this regulation works will be the identification of the proteins involved. An operon encoding proteins dedicated to the in vivo synthesis of Fe-S clusters has been characterized in prokaryotes. In addition to the IscS protein, homologous to Nfs1p, other proteins of this operon include HscA and HscB, IscU, and ferredoxin (5). Each of these bacterial proteins has a yeast counterpart containing an NH 2 -terminal extension with features of a mitochondrial leader sequence (6). Mutants of the yeast homologs of IscS (NFS1), HscA (SSQ1), and HscB (JAK1) were identified in a screen for suppressors of the lysine auxotrophy of ⌬sod1 mutants (6). The mechanism by which mutations in these putative mitochondrial proteins suppress defects in the cytoplasmic copper-zinc superoxide dismutase is not entirely clear. Furthermore, an iron regulatory function that recalls the NFS1 effects has been ascribed to SSQ1. Yeast mutants lacking Ssq1p, like mutants lacking Nfs1p, were selected using the FRE1-HIS3 selection scheme. These loss-of-function mutants exhibited up-regulation of cellular iron uptake activity and diversion of iron to mitochondria reminiscent of the nfs1-14 mutant (27). Ssq1p belongs to the Hsp70 class of chaperones with functions in protein translocation, protein folding, and proteolysis. The specific substrates of Ssq1p that mediate control of iron metabolism and Fe-S cluster protein synthesis are unknown.
One possible substrate for Ssq1p with a role in iron metabolism is Yfh1p, the yeast frataxin homolog (29,56). Yfh1p is synthesized on cytoplasmic ribosomes and imported into mitochondria, where the preprotein is processed in two steps. The second processing step is impaired in ssq1 mutants, suggesting a functional association between these two proteins (27). Yfh1p is thought to mediate iron export from the mitochondria (57). The human homolog of Yfh1p, called FRDA or frataxin, has been implicated in the neurodegenerative disease Friedreich ataxia (58). Another protein with a role in mitochondrial iron homeostasis is Atm1p. Atm1p is a transporter of the inner mitochondrial membrane oriented with a probable ATP-binding site in the matrix, so that it could function to export substrates from the matrix to the cytoplasm (59). The human gene has been implicated in a rare disorder involving ataxia and sideroblastic anemia (60). Loss-of-function mutations in yeast lead to iron accumulation in the mitochondria (61) and deficiency of heme proteins (59,61). The substrates for this putative mitochondrial export pump are unknown.
The specific roles of Nfs1p and interacting proteins in mediating iron homeostasis remain to be defined. There must be mechanisms for transport of iron into and out of mitochondria, for delivery of iron in functional form within the mitochondrial matrix, for assembly of the Fe-S clusters in mitochondria and in the cytoplasm, for unfolding of proteins for insertion of Fe-S clusters and for refolding these proteins. Finally, all these processes must be coordinated so that iron is present in sufficient amounts but not in toxic excess. Our work suggests that Nfs1p plays a role in this coordination. The conservation between yeast and humans of the proteins involved in normal iron homeostasis (Nfs1p, Yfh1p, and Atm1p) makes it likely that mechanisms observed in yeast will be relevant to human biology and disease.
Note Added in Proof-After this manuscript was submitted, an article by Lill and coworkers (Kispal, G., Csere, P., Prohl, C., and Lill, R. (1999) EMBO J. 18, 3981-3989) was published indicating that Nfs1 is an essential protein of mitochondria with a role in maintaining or synthesizing cytosolic Fe-S proteins.