The Iron Transporter Fth1p Forms a Complex with the Fet5 Iron Oxidase and Resides on the Vacuolar Membrane*

Iron transport across the plasma membrane appears to be a unidirectional process whereby iron uptake is essentially irreversible. One of the major sequestration sites for iron is the vacuole that stores a variety of metals, either as a mechanism to detoxify the cell or as a reservoir of metal to enable the cell to grow when challenged by a low iron environment. Exactly how the vacuole contributes to the overall iron metabolism of the cell is unclear because mutations that affect vacuolar function also perturb the assembly of the plasma membrane high affinity transport system composed of a copper-containing iron oxidase, Fet3p, and an Fe3+-specific iron transporter, Ftr1p. Here, we characterize the iron transporter homologue Fth1p, which is similar to the high affinity plasma membrane iron transporter Ftr1p. We found that Fth1p was localized to the vacuolar surface and, like other proteins that function on the vacuole, did not undergo Pep4-dependent degradation. Co-immunoprecipitation experiments showed that Fth1p also associates with the Fet3p oxidase homologue, Fet5p; and disruption of the FET5 gene results in the accumulation of Fth1p in the endoplasmic reticulum. We also found that loss of this protein complex leads to elevated transcriptional activity of the FET3 gene and compromises the ability of the cell to switch from fermentative metabolism to respiratory metabolism. Because the Fet5 protein is oriented such that the oxidase domain of Fet5p is lumenal, this complex may be responsible for mobilizing intravacuolar stores of iron.

Iron transport across the plasma membrane appears to be a unidirectional process whereby iron uptake is essentially irreversible. One of the major sequestration sites for iron is the vacuole that stores a variety of metals, either as a mechanism to detoxify the cell or as a reservoir of metal to enable the cell to grow when challenged by a low iron environment. Exactly how the vacuole contributes to the overall iron metabolism of the cell is unclear because mutations that affect vacuolar function also perturb the assembly of the plasma membrane high affinity transport system composed of a copper-containing iron oxidase, Fet3p, and an Fe 3؉ -specific iron transporter, Ftr1p. Here, we characterize the iron transporter homologue Fth1p, which is similar to the high affinity plasma membrane iron transporter Ftr1p. We found that Fth1p was localized to the vacuolar surface and, like other proteins that function on the vacuole, did not undergo Pep4-dependent degradation. Coimmunoprecipitation experiments showed that Fth1p also associates with the Fet3p oxidase homologue, Fet5p; and disruption of the FET5 gene results in the accumulation of Fth1p in the endoplasmic reticulum. We also found that loss of this protein complex leads to elevated transcriptional activity of the FET3 gene and compromises the ability of the cell to switch from fermentative metabolism to respiratory metabolism. Because the Fet5 protein is oriented such that the oxidase domain of Fet5p is lumenal, this complex may be responsible for mobilizing intravacuolar stores of iron.
The mechanism by which iron is transported across the plasma membrane in yeast has been intensively studied (1,2). High affinity transport is catalyzed by a complex of proteins that comprise the gene products of FET3 and FTR1 (3,4). FET3 encodes a multicopper iron oxidase, whereas FTR1 encodes a polytopic membrane protein that is thought to be a Fe 3ϩ -specific transporter. Together, Fet3p and Ftr1p are responsible for the transport of Fe 2ϩ across the plasma membrane; the observation that a functional FET3 is required for Ftr1p to exit the endoplasmic reticulum (ER) 1 strongly indicates that this complex assembles in the ER prior to moving to the plasma membrane (4). The coupling between the iron oxidase and the Fe 3ϩ -specific transporter may ensure the specificity of this transport system (1,2). The assembly of a functional Ftr1p-Fet3p complex is dependent on copper, which is supplied to the Fet3 protein by the copper-specific P-type ATPase Ccc2p and the copper-specific chaperone Atx1p. Loading of Fet3p with copper also requires a functional V-ATPase as well as the chloride channel Gef1p (5,6) and may take place in the Golgi apparatus or in some yet uncharacterized post-Golgi compartment. This model is based on the observations that Ccc2p and Gef1p are localized to such compartments and that perturbations in the function of the post-Golgi endocytic system lead to a defect in copper loading of Fet3p (5)(6)(7)(8)(9)(10)(11). In the absence of the Ftr1p-Fet3p complex, other transporters such as the Fet4 protein are up-regulated (12,13). Fet4p is a broad specificity transporter that has been shown to transport iron across the plasma membrane and is the major transporter responsible for low affinity iron uptake (12,13), whereas other transporters such as Smf1p/Smf2p may also contribute to nonselective uptake of iron (14 -16).
Iron transport across the plasma membrane in living yeast appears to be a unidirectional process whereby iron uptake and sequestration is essentially irreversible even when iron chelators are added to the exogenous medium (17). One of the major sequestration sites for iron is the vacuole that stores a variety of metals either as a mechanism to detoxify the cell or as a reservoir of metal to enable the cell to grow when challenged by a low iron environment (21). Exactly how the vacuole may contribute to the overall iron metabolism of the cell is unclear. Mutations in the vacuolar ATPase cause a defect for growth under iron-limiting conditions (5,6). Also, mutations that disrupt vacuolar biogenesis result in similar defects (14). The explanation for this effect is that these mutations alter the function or localization of the P-type ATPase Ccc2p, which in turn compromise the copper-dependent activation of Fet3p (7,8,14). However, how much these mutations would affect the capacity of the vacuole to store iron has not been examined directly, thus providing the possibility that these mutations may also perturb vacuolar storage function.
Because the vacuole stores iron, we surmised that the vacuole may have a specific iron transport system. Here, we characterize the iron (Fe) transporter homologue, Fth1p, which is similar to the high affinity plasma membrane iron transporter Ftr1p (4). We found that Fth1p is localized to the vacuolar surface, and like other proteins that function on the vacuole, it does not undergo PEP4-dependent degradation. Fth1p also appears to associate with the Fet3p oxidase homologue, Fet5p, as disruption of the FET5 gene results in the accumulation of Fth1p in the ER. Furthermore, immunoprecipitation experiments show that Fet5p associates with Fth1p in detergent lysates. Finally, we find that disruption of the Fth1p-Fet5p complex alters iron homeostasis and can inhibit the transition from fermentative to respiratory metabolism.

EXPERIMENTAL PROCEDURES
Materials-Glutathione-agarose beads were purchased from Amersham Pharmacia Biotech. Yeast nitrogen base was purchased from DIFCO. Amino acid supplements were purchased from Bio101 (La Jolla, CA). Chemiluminescent developer was purchased from Pierce. Bathophenathroline disulfonic acid (BPS) was purchased from Sigma. Bathocuproinedisulfonic salt was purchased from Aldrich. All other chemicals were of high grade and were purchased from commercial sources. Oxyliticase was purchased from Ensogenetics (Corvallis, OR). Restriction enzymes were purchased from New England Biolabs (Beverly, MA). Protease inhibitor mixture (Complete TM ) was purchased from Roche Molecular Biochemicals.
Yeast Strains and Culture Conditions-The yeast strains used in this study were SEY6210 (MAT␣ ura3-52 leu2-3-112 his3-⌬200 trp1-901 lys2-801 suc2-⌬9) (19 A fet5⌬ disruption strain was made by amplifying the HIS3 gene from the plasmid pRS303 using the oligos: ATGTTGTTCTACTCGTT-CGTGTGGTCTGTACTGGCCGCTAGTGTTGCTTTTTGTACTGAGAG-TGCACCAT; GCCGCGAGAAACCTCTATTTCATTTTCAGCCAAGAT-TTCCCTTAACGTGTGGGTATTTCACACCGCATA. The resulting PCR fragment was transformed into SEY6210. Genomic DNA was prepared from Hisϩ transformants by pelleting 1-2 optical density units of cells and incubating them in 300 l of 50 mM Tris, pH 8.0, 5 mM EDTA, 50 mM ␤-mercaptoethanol, containing 10 g/ml oxyliticase overnight to produce a cell lysate in which vacuolar proteases were allowed to digest cellular proteins. The cell lysate was then extracted with phenolchloroform, and DNA was precipitated with the addition of ethanol and NaCl. Genomic DNA from putative fet5⌬::HIS3 mutants was then analyzed by PCR using oligos flanking the HIS3 gene insertion. The fet5⌬::HIS3 deletion removes amino acids 17-620 from the FET5 open reading frame. A fet5⌬ fth1⌬ strain was made by amplifying the loci of the FTH1 gene from RGY3347 using oligos that hybridized 400 bp outside of either end of the FTH1 open reading frame. This PCR product (containing the KanMX4 neomycin resistance gene surrounded by a 400-bp sequence homologous to the 5Ј and 3Ј untranslated regions of FTH1) was transformed into PLY1313. Neo ϩ colonies were confirmed for the fth1⌬::NEO disruption by PCR amplification of the genomic DNA.
For metabolic shift experiments, cells were cultured in SD medium until they had reached mid-log phase (A 600 ϭ 1). Cells were centrifuged and resuspended in SD medium containing 200 M BPS and cultured for an additional 12 h. Cells were then washed with water, and 5 l of a 1:10 serial dilution of each cell type was then plated onto minimal plates containing 10 M BPS and 500 M FeSO 4 with either glucose or ethanol/glycerol (2:4%).
␤-Galactosidase Activity-Cells transformed with promoter-lacz fusion plasmids were grown in SD medium lacking uracil overnight to A 600 ϭ 2.5. Cells were then diluted to an A 600 of 0.05 in SD-Ura medium containing supplements and grown for an additional 10 h. Cells were collected by centrifugation, and cell pellets were frozen at Ϫ20°C. Cells were then resuspended in 1 ml of Z buffer (60 mM Na 2 PO 4 , 40 mM NaH 2 PO 4 , 10 mM KCl, 1 mM MgSO 4 , and 0.27% (v/v) 2-mercaptoethanol). 30 l of CHCl 3 and 20 l of 10% SDS were added, and cells were vortexed and adjusted to 0.8 mg/ml o-nitrophenyl-␤-D-galactosidase. Cell suspensions were incubated for 10 min at 22°C. Reactions were stopped by the addition of 500 l of 1 M Na 2 CO 3 , pH 11.0. Cells were pelleted, and supernatants were measured for absorbance at 405 nm. Each condition was measured in triplicate using three separate cultures derived from separate transformants for each plasmid.
Fluorescence Microscopy-Immunofluorescence microscopy using anti-HA, anti-Vma2p, and anti-Kar2p was performed essentially as previously described (19). Briefly, cells were grown in selective SD medium and adjusted to 3.7% formaldehyde. After 30 min, cells were then resuspended and incubated in 4% fresh paraformaldehyde containing 50 mM KPO 4 for 12 h. Cells were washed in 200 mM Tris, pH 8.0, ϩ 20 mM EDTA ϩ 1% ␤-mercaptoethanol, and then spheroplasted in 1.2 M sorbitol, 50 mM Tris, pH 8.0, containing 10 g/ml oxyliticase for 30 min. Cells were then washed in 1.2 M sorbitol, permeabilized for 60 s in 1% SDS, 1.2 M sorbitol, and washed three times in 1.2 M sorbitol. Cells were then adhered to polylysine-coated glass slides prior to immunolabeling.
For labeling with FM4-64 (Molecular Probes, Eugene, OR), cells were labeled for 30 min at 30°C in 500 nM FM4-64 in SD medium containing 50 mM KPO 4 , pH 7.5. Cells were then washed in water and cultured in unbuffered SD medium or yeast extract-peptone-dextrose for 30 min. Yeast cells harboring GFP-expressing plasmids were grown in SD medium to a density of 0.8 -1.2. An 0.5 volume of GFP-KILL buffer (1 M Tris, pH 8.0, 5.0% sodium azide) was added to ensure that membrane traffic had ceased and that the GFP protein was kept at the optimal pH level for fluorescence detection. For some experiments, 10 M DAPI was included in the GFP-KILL buffer. GFP fusion proteins were imaged using a 470-nm excitation filter, 495-nm dichroic mirror, and 525-nm emission filter. Images were captured with a Hammamatsu ORCA CCD camera mounted on an Olympus BX-60 microscope equipped with a 100ϫ oil objective. Image sets were processed and overlaid using Adobe Photoshop TM .
Antibodies-Plasmid pJLU54 was transformed into MC1061 Escherichia coli, which were then induced with isopropyl-1-thio-␤-D-galactopyranoside as described previously (22,23). The resulting recombinant GST-GFP fusion protein was purified over glutathione-agarose, dialyzed against phosphate-buffered saline, and used to immunize rabbits. Anti-GFP antibodies were affinity purified over an Affi-Gel 10/15 column to which was attached the GST-GFP fusion protein according to manufacturer's directions (Bio-Rad). Polyclonal anti-GST antibodies were purified from serum from rabbits immunized with irrelevant GST fusion proteins over an Affi-Gel 10/15 column to which was attached the GST protein alone.
The mouse anti-Vma2p (13D11) and anti-ALP (Pho8p; 1D3) monoclonal antibodies were purchased from Molecular Probes. The polyclonal anti-HA antibody was a kind gift of Kathryn Hill and Tom Stevens (University of Oregon, Eugene, OR). The monoclonal anti-HA antibody 16B12 and the monoclonal anti-6xHis antibody, which recognizes an epitope composed of six tandem histidine residues, was purchased from Babco (Berkeley, CA). Rabbit polyclonal anti-Kar2p was a kind gift of Mark Rose (Princeton University). Texas Red-labeled secondary antibodies were purchased from Molecular Probes. Horseradish peroxidase-conjugated secondary antibodies were purchased from Amersham Pharmacia Biotech. Digoxin labeling of antibodies was performed according to the manufacturer's instructions using a digoxin labeling kit (Roche Molecular Biochemicals).
Immunoblotting-Whole cell lysates from growing cultures of yeast cells were made by pelleting yeast, washing the pellet once with water, and resuspending it in 50 mM Tris, pH 6.5, 8 M urea, 2% SDS. Glass beads (150 -200 m) were added, and the sample was vortexed for 2 min. Samples were then adjusted to 1ϫ Laemmli sample buffer (20). For screening for the presence of the HA epitope, colonies were transferred into wells of a 96-well microtiter plate containing 50 mM Tris, pH 6.5, 5% SDS, 8 M urea, and 10% ␤-mercaptoethanol. The dish was then incubated at 65°C for 5 min prior to loading the resulting samples. To make whole membrane fractions, yeast cells were washed in water and resuspended in 200 mM Tris, pH 9.0, 0.05% sodium azide, containing 10 mM dithiothreitol, and incubated for 5 min at 25°C. Cells were then pelleted and resuspended in 10 ml of 1.2 M sorbitol, 50 mM KPO 4 , containing 10 g/ml oxyliticase, and incubated for 30 min at 30°C. The spheroplasts were then layered onto a sucrose cushion (1.2 M sucrose, 20 mM HEPEPS, pH 7.2, 2 mM EDTA) and collected as a pellet after centrifugation at 7,000 ϫ g for 20 min. Cells were lysed in 700 mM sorbitol, 20 mM HEPES, pH 7.2, and 2 mM EDTA. Postnuclear supernatants were generated by centrifugation at 1,000 ϫ g for 5 min, and the membranes were isolated by centrifuging the postnuclear supernatants at 40,000 ϫ g for 30 min in a Beckman TLA-45 rotor.
Immunoprecipitation-Yeast spheroplasts were made from 100 ml of yeast culture grown in minimal medium to A 600 ϭ 1.0 as described above. The spheroplast pellet was resuspended in phosphate-buffered saline, 1 mM EDTA, 1% Triton X-100, and protease inhibitor mixture (Complete TM , Roche Molecular Biochemicals). Cell lysates were then centrifuged at 15,000 ϫ g for 15 min, and the supernatant was removed and added to tubes containing affinity-purified anti-GFP antibody, affinity-purified anti-GST antibody, anti-HA monoclonal antibody 16B12, or the monoclonal anti-6xHis antibody. Extracts were incubated on ice for 2 h after which 50 l of protein G-coupled Sepharose was added (Santa Cruz Biotechnology, Santa Cruz, CA). After 60 min, beads were washed three times in phosphate-buffered saline, 1 mM EDTA, 0.5% Triton X-100. Samples were then solubilized in Laemmli sample buffer and analyzed by immunoblotting using a mouse monoclonal anti-HA antibody or digoxin-labeled rabbit anti-GFP antibody to assess the amount of HA-and GFP-tagged protein immunoprecipitated by the rabbit polyclonal antibodies. A polyclonal anti-HA was used to assess the amount of HA-tagged protein precipitated with the monoclonal anti-HA antibody.
Plasmids and DNA Manipulations-All PCR amplifications performed to obtain gene fragments were done using genomic DNA prepared from SEY6210. All PCR fragments were generated using the High Fidelity PCR kit (Roche Molecular Biochemicals). The plasmid pALP-GFP3-3ЈUT1 (pPL650) consisting of a NotI/BamHI fragment composed of bp-(440 -1698) of PHO8, a BamHI/EcoRI fragment of mutant 3 GFP (21) bp-(1-714), and an EcoRI/KpnI fragment of the 3ЈUT of the PHO8 gene (bp-(1701-3530)) in the polylinker of pRS316 was used as the vector for tagging various open reading frames with GFP. The plasmid was made by serially subcloning PCR fragments generated with High Fidelity polymerase using the oligos: CACGCGGCCGCAC-AAGGGAAGCAGGCCTCTTGC; CACGGATCCTGGTCAACTCATGG-TAGTATTC; CACGGATCCTGTCTAAAGGTGAAGGAATTATTC; GAAGAATTCTTATTTGTACAATTCATCCAT; CACGAATTCACGAAT-TCCCTATCTAAGTGTCCCTCTTTTTTC; and TCCAGGGATATCGAA-AAAATC.
pJLU73, which carries the FTH1-GFP fusion gene in a LEU2-containing, centromere-based vector, was made by cotransforming pJLU50, which had been cut with PvuII along with pRS315, which had been linearized. Yeast colonies that contained the correct recombinant plasmid were identified by screening Leuϩ Uraϩ transformants for the presence of GFP by fluorescence microscopy. The plasmid was then rescued by selecting for ampicillin resistant E. coli transformed with genomic DNA prepared from the respective Leuϩ Uraϩ GFPϩ yeast. pJLU61 was made by generating a PCR fragment containing a 6X-HA epitope that also had ends that overlapped the insertion site of GFP within pJLU38. This fragment was generated using the template pRCP48 (22) with the oligos: CCACACGTTAAGGGAAATCTTGGCT-GAAAATGAAATAGAGGTTTCTCGCGGCATGGATTCTATTAGATCT and AAATTATAACGTATTAAATAATATGTGAAAAAAGAGGGACAC-TTAGATAGGGCATGCACATAGAAGGATC. The fragment was then cotransformed with pJLU38 in which the GFP had been excised by cutting with BamHI and EcoRI. Uraϩ GFPϪ transformants were then screened for the presence of the HA epitope by immunoblot. This procedure resulted in a plasmid, pJLU61, that encoded a Fet5 fusion protein containing the following additional amino acids prior to the stop codon: DSIRSADLGRIF (YPYDVPDYAG) 3 AAQCGPDLGRIF-(YPYDVPDYAG) 3 AAQCGPD.
The predicted amino acid sequences of the Fth1-GFP, Pho8-GFP, and Fet5-HA fusion proteins were confirmed by DNA sequencing of the respective plasmids. pJLU54 was made by subcloning the BamHI/ EcoRI fragment from pPL650 (encoding mutant 3 GFP) into the BamHI/EcoRI site of pGEX-3X (22).
To ensure the sequence of FTH1 was correct, we fully sequenced the FTH1 plasmid (pJLU47) and a duplicate clone (pJLU48), as well as a full-length PCR product amplified and analyzed independently. Several changes from the original sequence were noted and submitted to Gen-Bank TM (accession number AF177330).
We validated the authenticity of the various GFP constructs by complementation: Vph1-GFP could restore vacuolar acidification to a vph1⌬ mutant; Ste3-GFP could restore mating efficiency to a ste3⌬ mutant; and GFP-PHO8 was found to traffic appropriately to the vacuole in an APM3-dependent manner. 2

Fth1p and Fet5p
Localize to the Vacuole Surface-We tagged Fth1p with the green fluorescent protein as part of a separate study to find membrane protein markers of different subcellular compartments. Surprisingly, we found Fth1p localized to the vacuole and went on to characterize this protein further. Using a set of GFP-tagged constructs, we found that Fth1p tagged with GFP at the C terminus was clearly localized to the limiting membrane of the vacuole, where it colocalized with the endocytic tracer FM4-64 that was allowed to transit to the vacuole for 1 h (Fig. 1). Despite localization to the vacuole, these data did not necessarily indicate that Fth1p functions at the vacuole. Indeed, overexpression of many membrane proteins that function elsewhere, such as the Golgi apparatus or plasma membrane, can accumulate within the vacuole upon overexpression or upon deletion of Golgi-targeting motifs. This phenomenon is most apparent in cells lacking the vacuole proteases that allow these proteins to accumulate without degradation. Recently, it has been found that membrane proteins destined for degradation in the vacuole actually localize within the lumen of the vacuole, whereas other proteins that function at the vacuole are localized to the vacuole surface (24). Other examples of this type of intravacuolar localization of proteins destined for degradation have been found previously (25)(26)(27)(28)(29)(30). Thus, to further examine the subcellular distribution of Fth1p, we compared the labeling pattern of Fth1-GFP to a panel of GFP fusion proteins made with either proteins known to function at the vacuole or proteins known to undergo PEP4-dependent degradation in the vacuole. This analysis was performed in a pep4 mutant strain that lacked vacuole proteases to ensure that proteins localized to the interior of the vacuole could be visualized. In the case of Fth1p, Fth1-GFP could clearly be seen on the vacuole surface in Pep ϩ and pep4 mutant cells, indicating that Fth1p does not localize within the interior of the vacuole. This same labeling pattern was also observed for Vph1p and Pho8p GFP fusion proteins. Vph1p is a component of the vacuolar-ATPase known to function at the vacuole, whereas PHO8 encodes the type II membrane protein alkaline phosphatase, which has been localized to the vacuole surface in previous studies (24,28). In contrast, proteins such as Ste3p, which function at the plasma membrane and then undergo PEP4-dependent degradation, were clearly seen within the vacuole and did not overlap with the FM4-64 label that was restricted to the limiting membrane of the vacuole characteristic of FM-64 labeling (31).
As a further analysis, we examined Fth1-GFP from wild-type and pep4 mutant cells by immunoblot analysis (Fig. 2). We found that with anti-GFP antibodies we could label a single ϳ75-kDa band from both wild-type and pep4 cells. This is the predicted molecular mass of the fusion protein; 51502 kDa for the FTH1 open reading frame and 26.8 kDa for the GFP open reading frame. When equal amounts of whole cell extract were loaded, the levels of Fth1p were the same. In contrast, this same analysis performed with Ste3p showed that the pep4 mutation dramatically increased the level of Ste3p, consistent with previous studies showing that Ste3p is rapidly degraded by vacuolar proteases after endocytosis from the plasma membrane (27). Taken together, these data indicate that Fth1p is a resident protein of the vacuole and thus likely functions at this compartment.
The Fth1p homologue, Ftr1p, is a Fe 3ϩ -specific transporter that forms a complex with the iron oxidase Fet3p. Fet3p is a type I membrane protein that is a multi-copper-dependent oxidase and bears homology to the iron oxidase ceruloplasmin. The Ftr1p-Fet3p complex has been localized to the plasma membrane and is responsible for high affinity uptake of iron into the cell (4). Another protein that bears close homology with the Fet3 iron oxidase is Fet5p (32). FET5 was originally identified as a high copy suppresser of iron-limited growth of a fet3 fet4 double mutant in which both the high affinity iron trans-port system (FET3-dependent) and the low affinity transport system (FET4-dependent) were disrupted. FET5 itself is responsible for ϳ85% of the total cellular iron oxidase activity, although disruption of FET5 does not lead to a decrease in the apparent transport of iron into the cell (32). Although the suppresser activity of high copy FET5 was dependent on the plasma membrane transporter Ftr1p, the observation that disruption of FET5 did not result in a decrease in iron uptake indicates that Fet5p might normally function elsewhere. Thus, given the functional interaction between Ftr1p and Fet3p, we were interested in the potential interaction between Fth1p and the oxidase Fet5p. Examining this possibility, we localized Fet5p using indirect immunofluorescence in pep4 mutant cells to determine whether Fet5p was also localized to the vacuole surface. To facilitate localization, we inserted an epitope tag encoding 6 tandem epitopes derived from the hemagglutinin spike glycoprotein (HA epitope). Immunoblot analysis showed that the Fet5-HA protein migrates as an ϳ100-kDa band con- sistent with what has been observed previously as glycosylation of the predicted 71-kDa core protein (32) . Fig. 3 shows double immunofluorescence localization of Fet5-HA with the vacuolar protein Vac8p as well as with the ER protein Kar2p in pep4 mutant cells. Fet5p could clearly be seen on the vacuolar membrane in a pattern very similar to Vac8p, a cytosolic protein associated with the vacuole (34). In contrast, Fet5p did not colocalize with the ER marker Kar2p. Importantly, we did not detect any fluorescence labeling within the vacuole but rather only on the limiting membrane of the vacuole, indicating that, like Fth1p, Fet5p is a resident protein of the vacuole.
Interaction between Fth1p and Fet5p-Given the observation that Fet5p colocalizes with Fth1p on the vacuole, we sought to determine whether these two proteins indeed formed a complex as do Ftr1p and Fet3p. The model that Fet3p forms a complex with Ftr1p is based on the observation that disruption of FET3 resulted in the accumulation of Ftr1p in the endoplasmic reticulum (4). Thus, we reasoned that if Fet5p and Fth1p also formed a complex before exiting the ER, we might observe a similar mislocalization of one putative subunit upon deletion of the other partner subunit. Fig. 4 shows the localization of Fth1p to the ER in a strain lacking the FET5 gene. In fet5⌬ cells, Fth1-GFP was not localized to the vacuole as denoted by FM-64 labeling. Rather, Fth1p was localized to structures similar to the ER, located underneath the plasma membrane and around the nucleus identified by DAPI labeling (Fig. 4A). To confirm this localization, we also performed double labeling experiments with Kar2p and Vma2p (Fig. 4, B and C). Here we observed good colocalization of Fth1p with the ER marker Kar2p and very little, if any, colocalization with the vacuolar ATPase subunit, Vma2p. In this experiment we chose to increase the levels of Fth1p as much as possible to enable better detection by growing cells in an iron-limited medium. Growth of cells in SD containing the Fe 2ϩ chelator BPS at 100 M was shown to significantly increase the levels of Fth1 protein (see below). To control for any effects of higher levels of Fth1p expression on localization, we also localized Fth1p in wild-type cells grown in 100 M BPS (Fig. 4D). Under these conditions, the higher levels of Fth1p were localized to the vacuole with no Fth1p apparent in the ER in wild-type cells. Also, the localization of Fth1-GFP to the ER was observed in fet5⌬ cells grown in the absence of BPS (data not shown).
The observation that Fth1p was dependent on the presence FIG. 2. Fth1p does not undergo PEP4-dependent degradation.  SF838-9Da pep4⌬ mutant cells (lanes 1 and 3) or congenic RPY10 wild-type Pep ϩ cells (lanes 2 and 4) were transformed with the Fth1-GFP-expressing plasmid pJLU47 (lanes 1 and 2) or the Ste3-GFPexpressing plasmid pJLU34 (lanes 3 and 4). Cell lysates were immunoblotted with anti-GFP antibodies or anti-ALP antibodies (loading control). of Fet5p for exiting the ER strongly indicated that Fth1p forms a complex with Fet5p. As a more direct test, we performed coimmunoprecipitation experiments from cells harboring the Fth1-GFP-producing plasmid (pJLU73) and the Fet5p-HA-producing plasmid (pJLU61). Detergent-solubilized spheroplasts were subjected to immunoprecipitation with anti-HA or anti-GFP antibodies, and the immunoprecipitates were analyzed by Western blot for GFP-and HA-tagged proteins. As a control for nonspecific precipitation, we also performed reactions with a polyclonal anti-GST antibody and a monoclonal antibody that recognizes a 6-histidine tag. Fig. 5 shows the anti-GFP antibody immunoprecipitated the Fth1-GFP protein and the Fet5-HA protein. In contrast, neither the anti-GST nor the anti-6xHis antibody immunoprecipitated the Fth1-GFP or the Fet5-HA protein. Comparable results were obtained when detergent extracts were made with CHAPS (data not shown).
Orientation of the Fth1p-Fet5p Complex-Given the similarities in the primary structure of Fet5p and Fet3p, one would expect the orientation of the Fet5p-Fth1p complex to be similar to the orientation of the Fet3p-Ftr1p complex. The fact that the Fet5p glycoprotein has a predicted short N-terminal hydrophobic signal sequence as well as having all of the predicted Nlinked glycosylation sites on the N-terminal side is strong evidence that Fet5p shares the same topology as Fet3p in that the copper-binding oxidase domain is lumenal (32). To confirm this orientation as well as probe the orientation of Fth1p, we subjected membrane fractions prepared from cells producing Fet5p-HA and Fth1-GFP to limited proteolysis and then immunoblotted with anti-GFP antibodies. As a control, we also  treated membrane fractions prepared from cells producing the Ste3-GFP fusion protein as well as Vph1-GFP (Fig. 6). In the case of Ste3-GFP, all of the Ste3p was protected from protease treatment because this protein is localized within the vacuole (refer to Fig. 1). In contrast, all of the GFP moiety from the Vph1-GFP was susceptible to protease treatment consistent with previous studies showing that the C-terminal portion of Vph1p (where GFP was attached) is oriented to the cytosol (33). As an internal control, we immunoblotted Pho8p; note that upon protease treatment, Pho8p migrates slightly faster indicating that the short N-terminal portion of Pho8p was cleaved by proteases. In the case of Fet5-HA, the HA epitope was susceptible to limited protease treatment, confirming that the orientation of the Fet5 protein is similar to that of a type I transmembrane protein such that the short C-terminal portion is oriented toward the cytosol. Likewise, the C-terminal GFP domain within the Fth1p-GFP protein was also completely accessible to protease treatment. These data not only confirm the localization of Fth1p to the vacuole surface, but they also indicate that the C terminus Fth1p is oriented toward the cytosol.
Role of the Fth1p-Fet5p Complex in Iron Metabolism-One of the key responses to growth in low iron conditions is the transcriptional up-regulation of the high affinity plasma membrane iron transport system (1). Consistent with a role in iron metabolism, FET5 message levels have been shown to increase during growth in low iron conditions (32). Although not directly tested, the most likely explanation is that transcriptional activation of the FET5 gene is under the control of the ironresponsive Aft1p transcription factor. The observation that FET5 is within the transcriptional program that cells use to respond to low iron indicates that this protein is indeed somehow involved in iron metabolism and is required more under low iron conditions. Previous studies have also shown a similar response for FTH1 as the levels of FTH1 mRNA are augmented in strains carrying an activated allele of AFT1 (35). As confirmation of the iron responsiveness of the FTH1 gene, we found that the levels of Fth1 protein are significantly elevated in cells grown in low iron medium (SD medium containing the iron chelator BPS at 100 M ) (Fig. 7A). Likewise, we found that activation of the FTH1 promoter, as assessed by ␤-galactosidase activity produced from the appropriate FTH1 promoterlacz fusion plasmid, was greatly elevated in response to 100 M BPS (Fig. 7B), whereas no effect was observed upon the addition of the copper chelator bathocuproinedisulfonic acid. Thus, as shown in previous studies, the transcription of FTH1 and FET5 is activated by low iron conditions (1, 35). Importantly, we found that the addition of 500 M FeSO 4 further decreased the level of FET3 and FTH1 promoter-dependent lacz production, indicating that our SD medium had intermediate levels of iron such that the level of intracellular iron was low enough to induce some activation of the FET3 gene.
Given the ability of the cell to rapidly and markedly upregulate transcription of FET3 to compensate for iron-poor conditions, we decided to use the transcriptional activity of the FET3 gene as a functional indicator of the overall level of intracellular iron. We reasoned that if the Fet5p-Fth1p complex was important for providing iron to the cell from intracellular stores, the absence of this complex might lead to activation of FET3 under conditions of intermediate levels of iron. Thus, we assessed the transcriptional activity of the FET3 promoter by measuring the ␤-galactosidase activity produced by the FET3-lacz fusion plasmid in wild-type and fet5⌬ cells grown in SD medium containing 10 M BPS, 100 M BPS, or 500 M FeSO 4 . At 100 M BPS, the level of ␤-galactosidase activity was the same for both the wild-type strain and the fet5⌬ mutant, showing that the maximal response of the FET3 promoter to low levels of iron was similar between wild-type and fet5⌬ mutants (data not shown). Likewise, the FET3 promoter activity in high levels of iron was the same between both FIG. 6. Protease susceptibility of the Fth1p C terminus. Spheroplasts were made from SF838-9Da (pep4⌬) cells carrying the Fth1-GFPexpressing plasmid pJLU47, the Fet5-HA expressing-plasmid pJLU61, the Ste3-GFP-expressing plasmid pJLU34, or the Vph1-GFP-expressing plasmid pJLU40. Postnuclear supernatants were prepared from lysates obtained after resuspension in hypotonic medium. Membranes were pelleted, resuspended, and treated with trypsin in the absence or presence of 0.5% Triton X-100, as indicated, for 15 min at 22°C. Reactions were stopped with the addition of SDS-polyacrylamide gel electrophoresis sample buffer and analyzed by immunoblot analysis using anti-GFP, anti-HA, or anti-ALP antibodies.

FIG. 7. Iron homeostasis and regulation of Fth1p levels.
A, SEY6210 cells harboring the Fth1-GFP-expressing plasmid pJLU47 were cultured in SD medium or SD containing 100 M BPS for 12 h as indicated. Equivalent extracts were prepared and immunoblotted with anti-GFP. B, SEY6210 cells were transformed with the FTH1-lacz construct pJLU65, the FET3-lacz construct pJLU62, or the FET4-lacz construct pJLU68 and grown in SD medium. Cells were then shifted to SD medium containing 500 M FeSO 4 , 100 M BPS, 100 M BCS or no additions for 6 h prior to assaying for ␤-galactosidase activity. Enzyme activity was normalized to cell number and expressed relative to the specific activity measured in cultures grown in SD for each respective construct. C, wild-type (SEY6210) and fet5⌬ (PLY1313) cells carrying the FET3-lacz construct pJLU62 were grown overnight in SD medium and then cultured for 6 h in SD containing 500 M FeSO 4 , no additions, or 10 M BPS prior to measuring ␤-galactosidase activity. Enzyme activity was normalized to cell number and expressed as the mean Ϯ S.D. All measurements in B and C were made in triplicate using three different transformants for each condition.
wild-type and fet5⌬ cells (Fig. 7C). In contrast, fet5⌬ cells grown in SD medium or SD medium containing only 10 M BPS showed a 2-fold increase in the level of ␤-galactosidase activity, indicating that iron levels were compromised enough in fet5⌬ mutants to result in activation of the FET3 gene.
As a more substantial test of Fth1p-Fet5p function, we sought to develop a regimen by which vacuolar iron stores might be necessary for growth. Previous studies have shown that during the switch from growth on glucose to growth on glycerol that vacuolar iron stores are depleted (18). This depletion is observed even when the cells are cultured in iron-replete medium and most likely represents the large amount of iron that is required for the proper synthesis of the heme-containing mitochondrial cytochrome oxidases (18). Thus, we tested whether fet5 fth1⌬ mutant cells were compromised in making a transition from fermentative to respiratory metabolism. In the course of these experiments we found that prior treatment of cells with the iron chelator BPS accentuated the defects we observed, possibly by limiting the amount of iron stored in vacuolar or mitochondrial compartments. Fig. 8 shows that wild-type cells (SEY6210) or fet5⌬ fth1⌬ (PLY1546) cells harboring the FET5-HA and FTH1-GFP plasmids were both capable of quickly converting from fermentative growth to respiratory growth when switched from glucose to ethanol/glycerol. In contrast, fth1⌬ fet5⌬ mutant cells alone lagged well behind their wild-type counterparts in converting to respiratory metabolism, provided that they were subjected to treatment with BPS prior to metabolic challenge. Interestingly, this growth restriction was observed even when the cells were plated onto iron-rich medium, suggesting that utilization of intracellular iron is key to this transition. After prolonged incubation, we did observe the growth of fth1⌬ fet5⌬ cells, suggesting that the particular defect in these cells is a greater lag period for metabolic conversion. If cells were not treated with BPS prior to plating on ethanol/glycerol plates, no defects were seen, demonstrating that each strain was inherently capable of respiratory growth. Furthermore, prior treatment of cells with BPS did not alter their viability when plated onto glucose containing plates. DISCUSSION We have found that the iron transporter homologue, Fth1p, is localized to the vacuolar surface and does not undergo PEP4dependent degradation, which indicates that it functions at the vacuole. Fth1p appears to function with the iron oxidase Fet5p because Fet5p specifically coimmunoprecipiates with Fth1p from detergent lysates and is required for the exit of Fth1p out of the ER. We propose the following model for the Fth1p-Fet5p complex, shown in Fig. 9, based on the orientation of Fet5p that has been solved in previous studies (32), the sequence changes we found upon fully sequencing FTH1 from SEY6210 yeast, and the observation that the C terminus is likely to be oriented in the cytosol. The FTH1 open reading frame shows homology with Ftr1p throughout the length of the protein. Fth1p possesses two REXXE motifs in a similar position to those within Ftr1p and an extra REXXE motif within the loop between transmembrane domain 1 and 2. These motifs have been shown to coordinate the binding of iron within ferritin and the iron transporter SFT and to be functionally important within Ftr1p (4,36,37). In the original model for Ftr1p, it was proposed that the N-terminal hydrophobic region serves as a cleavable signal sequence leaving Ftr1p with six transmembrane domains such that the C terminus is lumenal. However, other topology prediction algorithms also score this region, adopting the conformation proposed in Fig. 9. Given the significance of the REXXE motif and the observation that it is within the very N-terminal hydrophobic region of both Ftr1p and Fth1p, we believe that this N-terminal region may form another transmembrane domain. Although this conclusion remains to be tested directly, we also note that in this model both of the predicted N-linked glycosylation sites (N 89 KS and N 218 KT) are predicted to be cytosolic. Consistent with this understanding, we observed no effect on the migration of Fth1p by SDS-polyacrylamide gel electrophoresis after endoglycosidase H treatment, indicating that these sites are not accessible to lumenal glycosyltransferases (data not shown). Also, there was no difference in the size of Fth1p in wild-type cells versus pep4 mutant cells, indicating that the large predicted loops are either resistant to vacuolar proteases or cytosolically disposed.
As an iron transport complex on the vacuole, Fth1p-Fet5p could serve one of two functions. One possibility is that this complex could be used to detoxify the cell of high intracellular levels of iron. Perhaps the iron oxidase activity of the complex could modify iron to be bio-unavailable as Fe 3ϩ , which could be sequestered by binding to polyphosphates or other factors within the vacuole. However, this possibility is counterintuitive given that Fth1p levels increase when cells are challenged by low iron conditions (Fig. 7) and that message levels from both FTH1 and FET5 promoters are greatly elevated in iron-poor media (32,35). Instead, the model we favor is that the vacuole can act as a store for iron and that the Fth1p-Fet5p complex serves as a means to mobilize stored iron under conditions of iron starvation. One of the difficulties in assessing the role of these proteins in overall iron-dependent growth of the cell is that there are many mechanisms in place to compensate for the lack of iron. When iron levels are low in the cell, the production of the iron transport systems are elevated by activation of the transcription factor Aft1p to restore intracellular iron levels (35). In the absence of the high affinity Ftr1p-Fet3p system or in low iron medium, Fet4p levels increase and iron uptake across the plasma membrane is maintained by this low affinity/ low specificity transport system (12,13). Although these mechanisms may not be viable strategies in the wild because they may incur increased risk to other environmental factors such as toxic heavy metals, they do obscure functional analysis of these iron transport proteins (14). We find that when cells are grown in medium containing intermediate levels of iron, disruption of the vacuolar iron transport complex caused the activation of FET3 transcription, indicating that intracellular iron levels are perturbed enough to activate the FET3 promoter. Iron can be accumulated to high levels within vacuoles and then mobilized from these stores and incorporated into other cellular proteins, especially during the transition from fermentative to respiratory growth (18). Our analysis indicates that the Fth1p-Fet5p complex plays an important role in the mobilization of vacuolar iron stores during this transition and that absence of the complex results in a significant delay in the synthesis of many iron-containing enzymes required for respiration.
One question that arises from these studies is how iron storage might be regulated. Iron might be stored in the vacuole as Fe 3ϩ and thus not as a substrate for the Fth1p-Fet5p complex. This model would predict a role for a vacuolar localized iron reductase. Currently, there are seven recognized iron reductase-related genes in yeast (38). All but one, FRE7, are activated by growth in low iron conditions, advancing the pos-sibility that one of these proteins is localized exclusively to the vacuole. Fet5p has been shown to be a potent iron oxidase, and disruption of FET5 led to an 85% decrease in total cellular oxidase activity (32). Fet5p was originally characterized as a multicopy suppresser of the iron-limited growth defect of a fet3 fet4 mutant. Interestingly, this suppression was dependent on Ftr1p, suggesting that the overexpressed Fet5p was helping to assemble a functional Ftr1p complex capable of transporting iron across the plasma membrane. Importantly, in a fet3 fet4 mutant grown under iron-poor conditions, the expectation would be that Fet5p levels would already be induced because FET5 transcript levels are increased in iron limited medium, yet this is not enough to suppress the fet3 mutation. As expressed from a 2plasmid, Fet5p levels under these conditions would be expected to be exceptionally high and may allow some association with Ftr1p and exit from the ER. In our experiments, Fth1p-Fet5p was associated only with the vacuole even when levels of this complex were elevated by growing the cells in limited iron medium. Given the similarities between Fet3p and Fet5p and Fth1p and Ftr1p, it will be interesting to determine not only what domains of the iron oxidase are required for recognition of the respective transporter and vice versa but also to determine the mechanisms by which the Fth1p-Fet5p complex is sorted to the vacuole while the Fet3p-Ftr1p complex is sorted to the plasma membrane.