Evidence for the Saccharomyces cerevisiae ferrireductase system being a multicomponent electron transport chain.

We have studied the relationships between in vivo (whole cells) and in vitro (plasma membranes) ferrireductase activity in Saccharomyces cerevisiae. Isolated plasma membranes were enriched in the product of the FRE1 gene and had NADPH dehydrogenase activity that was increased when the cells were grown in iron/copper-deprived medium. The diaphorase activity was, however, independent of Fre1p, and Fre1p itself had no ferrireductase activity in vitro. There were striking similarities between the yeast ferrireductase system and the neutrophil NADPH oxidase: oxygen could act as an electron acceptor in the ferrireductase system, and Fre1p, like gp91, is a glycosylated hemoprotein with a b-type cytochrome spectrum. The ferrireductase system was sensitive to the NADPH oxidase inhibitor diphenylene iodonium (DPI). DPI inhibition proceeded with two apparent Ki values (high and low affinity binding) in whole wild-type and Δfre2 cells and with one apparent Ki in Δfre1 cells (high affinity binding) and in plasma membranes (low affinity binding). These results suggest that the Fre1-dependent ferrireductase system involves at least two components (Fre1p and an NADPH dehydrogenase component) differing in their sensitivities to DPI, as in the neutrophil NADPH oxidase. A third component, the product of the UTR1 gene, was shown to act synergistically with Fre1p to increase the cell ferrireductase activity.

We have studied the relationships between in vivo (whole cells) and in vitro (plasma membranes) ferrireductase activity in Saccharomyces cerevisiae. Isolated plasma membranes were enriched in the product of the FRE1 gene and had NADPH dehydrogenase activity that was increased when the cells were grown in iron/copper-deprived medium. The diaphorase activity was, however, independent of Fre1p, and Fre1p itself had no ferrireductase activity in vitro. There were striking similarities between the yeast ferrireductase system and the neutrophil NADPH oxidase: oxygen could act as an electron acceptor in the ferrireductase system, and Fre1p, like gp91, is a glycosylated hemoprotein with a b-type cytochrome spectrum. The ferrireductase system was sensitive to the NADPH oxidase inhibitor diphenylene iodonium (DPI). DPI inhibition proceeded with two apparent K i values (high and low affinity binding) in whole wild-type and ⌬fre2 cells and with one apparent K i in ⌬fre1 cells (high affinity binding) and in plasma membranes (low affinity binding). These results suggest that the Fre1-dependent ferrireductase system involves at least two components (Fre1p and an NADPH dehydrogenase component) differing in their sensitivities to DPI, as in the neutrophil NADPH oxidase. A third component, the product of the UTR1 gene, was shown to act synergistically with Fre1p to increase the cell ferrireductase activity.
The ferrireductase system of Saccharomyces cerevisiae is a plasma membrane electron transfer system which reduces extracellular ferric chelates to their ferrous counterparts (1)(2)(3). This reduction step thermodynamically favors the dissociation of iron from its ligands, which seems to be a prerequisite for iron uptake. Free ferrous ions, which are far more soluble than ferric ions, are then taken up by an unidentified high affinity transport system (4). Translocation of iron into the cells could be accompanied by a second oxidoreduction step catalyzed by Fet3p (5).
The ferrireductase activity of cells is greatly increased when the cells are grown in iron-deficient conditions (1)(2)(3). Reciprocally, the reductase system is repressed by excess iron in the growth medium, but also by copper (6) and by cadmium ions (7). Copper could be taken up via the same reductase system (8).
Dancis et al. (9) identified the gene FRE1 as essential for the cell ferrireductase activity. The product of this gene is expected to be a plasma membrane protein with several transmembrane domains. Interestingly, Fre1p has significant homologies with gp91 phox , a component of the neutrophil cytochrome b 558 (9). Cytochrome b 558 , together with other plasma membrane and cytosolic proteins, constitute the NADPH oxidase complex which, in response to various stimuli, uses intracellular NADPH to reduce extracellular oxygen to superoxide anions (reviewed in Ref. 10). Strains lacking the FRE1 gene still have a residual activity (about 10% of the wild-type) which is completely lost when a second gene, FRE2, is deleted (11). The product of the FRE2 gene presents significant homologies with Fre1p (24.5% identity) and with gp91 phox (11). Transcription of FRE1 and FRE2 are regulated differently, and their products could be components of separate reductase systems with different physiological roles.
We partially purified a NADPH flavodehydrogenase whose activity in plasma membranes from iron-deprived cells was higher than that of iron-rich cells (12). However, we also showed that the NAD(P)H-dependent ferrireductase activity associated with isolated plasma membranes did not reflect the ferrireductase activity of whole cells (13). First, the increase in ferrireductase activity upon iron/copper starvation was considerably less in purified plasma membranes than in intact cells. Second, the reductase activities in plasma membranes isolated from various strains (wild-type, fre1, hem1, or ras1 mutants), which had very different ferrireductase activities in vivo, were similar. It is possible that Fre1p and/or Fre2p are lost during plasma membrane purification, or the measured NAD(P)H-dependent reductase activities are due to mitochondrial and/or endoplasmic reticulum contamination. We therefore re-examined the question of the link between in vivo and in vitro ferrireductase activity and provide evidence against the model of Fre1p/Fre2p as a transmembrane NAD(P)H dehydrogenase.

EXPERIMENTAL PROCEDURES
Yeast Strains and Growth Conditions-The reference strain used was S150-2B (a, his3, leu2, trp1, ura3). The following strains, provided by D. Alexandraki (Institute of Molecular Biology and Biotechnology, University of Crete), were used in some experiments: S288C (wildtype), S288C⌬fre1, S288C⌬fre2, S288C⌬fre1⌬fre2. Cells were grown as described previously (6) except when cultures were used for plasma membrane isolation. For the latter, cells were precultured in 50 ml of YNB-glucose medium for 24 h and then inoculated to a final A 600 of 0.005 in 2.5 liters of the same medium. When the A 600 reached 1.2, the medium was diluted 2-fold with medium containing 4% glucose, 2% yeast extract, 2% peptone, and either 200 M BPS 1 (iron/copper-deprived medium) or 25 M CuSO 4 (iron/copper-rich medium). Cells were harvested by centrifugation when the A 600 reached 1.5.
Isolation of Plasma Membranes-Cell fractionation and plasma membrane purification were done as described by Dufour et al. (14), using disruption of cells with glass beads followed by differential cen-* This work was supported by grants from CNRS and Université Paris 7. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ To whom correspondence should be addressed. Tel: 33-1-43-54-04-79; Fax: 33-1-44-27-57-16; E-mail: elu@ccr.jussieu.fr. trifugation and acid precipitation (acetic acid) of mitochondrial membranes. Aliquots of plasma membranes were suspended at about 5 mg/ml in 10 mM Tris acetate buffer (pH 7.5) and frozen. Low temperature spectra (Ϫ191°C) of plasma membrane suspensions were obtained as described previously (15), with an optical path length of 1 mm with one sheet of wet filter paper in the reference path. Spectra were corrected for the baseline shift.
The cell suspension was about 1 mg (wet weight) per ml. When resazurin was used as the electron acceptor, the reductase activity was recorded by the appearance of resorufin at 30°C with a Jobin Yvon JY3D spectrofluorimeter ( ex 560 nm, em 585 nm, slit widths of 2 nm for both excitation and emission). The incubation mixture in 50 mM sodium citrate buffer (pH 6.5) containing 10 M resazurin was magnetically stirred. Activity was quantified from a resorufin calibration curve prepared under the conditions of the enzymatic assay. The reductase activity of plasma membranes (10 -50 g/ml) was measured in the same way, using 500 M Fe(III)-EDTA, or 10 M resazurin and 100 M NAD(P)H as substrates. When DPI was added, the cells or plasma membranes were preincubated for 10 min with the inhibitor (whole cells) and NAD(P)H (membranes) before adding the electron acceptor. Cytochrome c reduction was followed spectrophotometrically at 550 nm. Protoporphyrinogen oxidase, used as a mitochondrial marker, was assayed as described by Camadro et al. (16).
Plasmids and DNA Constructions-Disruption by TRP1 of the FRE1 gene was performed as described previously (7). For overexpression, FRE1, FRE2, and UTR1 were cloned in YEp351 and YEp352. The FRE1-lacZ gene fusion was constructed as described previously (7). For the UTR1-lacZ construction, a DNA fragment containing 600 base pairs of the UTR1 promoter and the first 135 base pairs of the coding sequence was cloned in the multicopy vector YEp353 (17). For epitope tagging, FRE1 was cloned in YEp352. New BamHI restriction sites were created at different points in the FRE1 gene by site-directed mutagenesis (18). In each case, it was verified that the mutations did not affect the activity of the resulting Fre1p. Double-stranded synthetic DNA encoding the c-myc epitope and flanking pentaglycine linkers (GGGGGMEQKLISEEDLNGGGGG) were then cloned in-frame into the new BamHI sites.
Electrophoresis and Immunoblotting-Native PAGE of plasma membrane proteins and in situ detection of reductase activity were performed using 3-20% polyacrylamide gradient gels containing 0.1% Triton X-100 (1.5 mm thick) (19). Gels were electrophoresed for 16 h at 4°C (30 mA per gel, 50 mM Tris-glycine, pH 9) and then incubated at 30°C in 50 mM citrate buffer (pH 6.5) containing 100 M NADPH and either 0.5 mg/ml NBT or 200 M Fe(III)-EDTA plus 1 mM BPS. Similar results were obtained with either NBT or Fe(III)-EDTA as electron acceptors, but bands of NADPH dehydrogenase activity were enhanced with NBT. SDS-polyacrylamide gels and transfer to nitrocellulose were done using standard procedures. Immunodetection was with anti-c-Myc monoclonal antibody (immunodetection of Fre1(c-Myc)p) or with antiserum specific to Sss1p (20) and employed horseradish peroxidase-conjugated anti-mouse IgG second antibody followed by ECL chemiluminescence (Amersham). For quantitative analysis, Sss1p was immunodetected on dot blots by successive incubation of the nitrocellulose with anti-Sss1p antiserum and radioiodinated protein A, followed by counting of the radioactive spots in a liquid scintillation counter.

Isolated Plasma Membranes Are Enriched in Fre1p and
Show Specific NADPH-dependent Ferrireductase Activity-The FRE1 gene was engineered with an epitope tag from c-myc at the carboxyl-terminal domain of the protein to allow Fre1p to be monitored during cell fractionation, but this led to complete loss of Fre1p activity (data not shown). A DNA fragment encoding the c-myc epitope and flanking pentaglycine linkers was then introduced at different positions in the FRE1 gene in a multicopy plasmid. Insertion sites (positions 137-138, 246 -247, and 314 -315 of the protein) were in regions encoding putative hydrophilic loops of the protein. A ⌬fre1 strain then was transformed with the recombinant genes. Ferrireductase activity was retained when the c-myc epitope was inserted at positions 137-138 and 246 -247, but not at position 314 -315. Subcellular fractions were then prepared from transformants grown under iron/copper starvation, and the Fre1(c-Myc) pro-tein was immunodetected in the plasma membrane fraction ( Fig. 1). In some cases, a doublet was apparent for Fre1p ( Fig.  1) which could correspond to a partial degradation of the protein.
Mitochondrial contamination of the plasma membrane fraction was monitored by assaying protoporphyrinogen oxidase, an enzyme of the inner mitochondrial membrane. The extent of mitochondrial contamination of the plasma membrane preparations was estimated to be 3-8%. Contamination of the plasma membrane fraction with endoplasmic reticulum was estimated by immunodetection of Sss1p, an endoplasmic reticulum protein involved in protein secretion (20). Sss1p (75%) was mostly associated with the mitochondrial fraction (Fig. 2).
NADH-dependent ferrireductase activity (99 nmol/min/mg of protein) was higher than NADPH-dependent ferrireductase activity (14 nmol/min/mg of protein) in the mitochondrial fraction. The opposite was true in the plasma membrane fraction (69 nmol/min/mg of protein for NADH-dependent activity and 186 nmol/min/mg of protein for NADPH-dependent activity in one representative experiment). Intact mitochondria prepared by osmotic lysis of protoplasts showed no NADPH-dependent ferrireductase (12). Therefore, the activity in the mitochondrial fraction was probably due to contamination by endoplasmic reticulum and plasma membrane. Reciprocally, the NADPH ferrireductase activity of the plasma membrane fraction could not result from contamination with either mitochondria or  (1), mitochondria (2), and plasma membranes (3) of S. cerevisiae (S150-2B⌬fre1 transformed by FRE1(c-myc) on YEp352). Ferrireductase activity was induced by growth on iron/copper-deprived medium. endoplasmic reticulum since endoplasmic reticulum mostly contaminates the mitochondrial fraction which showed little NADPH-dependent activity. Finally, the NADH-dependent ferrireductase activity in the plasma membrane fraction could be due to a lack of specificity of the plasma membrane-associated NADPH dehydrogenase or to mitochondrial contamination. In addition, we confirmed (see Ref. 12) that the NADPH-(and not the NADH-) dependent ferrireductase activity was increased about 2-fold in plasma membranes isolated from cells with induced ferrireductase activity (growth with 200 M BPS) compared to plasma membranes from control cells (growth with 25 M Cu 2ϩ ) (data not shown).
Relationship between Plasma Membrane-bound NADPH-dependent Ferrireductase Activity and Fre1p-Plasma membranes were purified from a wild-type strain and from ⌬fre1, ⌬fre2, and ⌬fre1⌬fre2 mutants. In vivo (whole cells) and in vitro (plasma membranes) ferrireductase activity was measured. There was no correlation between the in vivo and the in vitro activities (Table I). While the double disruptant strain showed almost no activity in vivo, its plasma membranes had the highest NADPH-dependent ferrireductase activity. Therefore Fre1p (Fre2p) does not contribute to ferrireductase activity in vitro. This hypothesis was tested further by isolating plasma membranes from cells transformed by FRE1(c-myc) in a multicopy plasmid. After treatment with various detergents, the resulting soluble and insoluble fractions were separated by centrifugation. Each fraction was assayed for ferrireductase activity and tested for the presence of Fre1(c-Myc)p. The NADPH-dependent ferrireductase was solubilized more easily than Fre1(c-Myc)p, and there was no correlation between the presence/absence of Fre1p and the ferrireductase activity in a given fraction (Fig. 3). Similarly, the Fre1(c-Myc)p and the plasma membrane-bound ferrireductase could be separated by electrophoresis (Fig. 4). Native PAGE of total plasma membrane proteins followed by in situ detection of the NADPH-dependent reductase activity identified two main bands with reductase activity. The pattern from the plasma membrane fraction was clearly different from that of the mitochondrial fraction and the whole homogenate (Fig. 4). Gels were treated with SDS and blotted onto nitrocellulose for immunodetection of Fre1(c-Myc)p. A single band was detected as a high molecular weight entity where there was little NADPH dehydrogenase activity. In vitro, Fre1(c-Myc)p therefore is devoid of NADPH dehydrogenase activity.
Similarities between the Yeast Ferrireductase System and the Neutrophil NADPH Oxidase-Fre1p is a glycosylated hemoprotein. Dancis et al. (9) previously showed that there were significant homologies between the sequences of Fre1p and gp91, the main component of cytochrome b 558 . Treatment of plasma membrane proteins with endoglucosidase H showed that Fre1(c-Myc)p is glycosylated, like gp91. The apparent molecular mass of Fre1(c-Myc)p was decreased by this treatment to a value close to the calculated molecular mass of the protein (81 kDa) (data not shown). Plasma membranes isolated from cells overexpressing FRE1 had a b-type cytochrome absorbance spectrum, with peaks at 556 -557 nm (␣), 526 -527 nm (␤), and 424 nm (␥) (Fig. 5). This spectrum is identical with that of a typical neutrophil cytochrome b (21). It was observed in plasma membranes isolated from a strain overexpressing FRE1 and grown on iron/copper-deprived medium, but not when FRE1 transcription was repressed by adding copper to the growth medium (Fig. 5). Nor was it observed in plasma membranes isolated from a ⌬fre1⌬fre2 strain or a ⌬fre1⌬fre2 strain bearing FRE2 in a multicopy plasmid, regardless of the growth conditions (Fig. 5). These data clearly indicate that Fre1p is a hemoprotein, which explains why heme-deficient mutants also lack ferrireductase activity (2). No cytochrome spectrum was detected when FRE2 was overexpressed in a double disruptant strain, indicating that either Fre2p is not a hemoprotein or that it was not abundant enough, even under conditions of induction, to allow detection of its putative cytochrome spectrum.  1-8, respectively, 82%, 94%, 64%, 98%, 87%, 96%, 99%, and 99% of the total ferrireductase activity was associated with the soluble fractions.

FIG. 4. Detection of the NADPH dehydrogenase activity associated with different subcellular fractions on native PAGE and immunodetection of Fre1(c-Myc)p.
Cells (S150-2B⌬fre1 transformed by FRE1(c-myc) in a multicopy plasmid) were grown in iron/ copper-deficient medium. 100 g of protein of the plasma membranes (1,4), mitochondria (2, 5), or the whole homogenate (3, 6) were separated by native PAGE, and the NADPH dehydrogenase activity was revealed in situ with NBT as the electron acceptor (lanes 1-3). The proteins on the gel were then denatured (16-h incubation with 1% SDS in 50 mM Tris-glycine buffer, pH 8.3) and blotted onto nitrocellulose for immunodetection of Fre1(c-Myc)p (lanes 4 -6).

TABLE I In vivo (whole cells) and in vitro (plasma membranes) ferrireductase
activities of different yeast strains Cells (S288C and isogenic mutants) were grown in iron/copper-deprived medium to induce the cell ferrireductase activity. Whole cell ferrireductase activities (mean Ϯ S.D. from 3 experiments) are expressed in nanomoles/min/mg of total protein to allow direct comparison with the in vitro values. It was assumed that 100 mg wet weight of cells/ml (corresponding to an A 600 of 75) ϭ 10 mg of total protein/ml. In vitro ferrireductase activities were measured with NADPH (100 M) as electron donor. Results are for two experiments (wild-type, ⌬fre1) or one experiment (⌬fre2, ⌬fre1⌬fre2).

Strain
Wild type ⌬fre1 ⌬fre2 ⌬fre1⌬fre2 In vivo 9 Ϯ 0.8 0.75 Ϯ 0. However, the ferrireductase activity of double disruptant cells that overexpressed FRE1 (8 nmol min Ϫ1 A 600 Ϫ1 ) was only 2-fold higher than when FRE2 was overexpressed in the same conditions (3.5 nmol min Ϫ1 A 600 Ϫ1 ). Cellular Production of Superoxide-Cytochrome c was slowly reduced by resting cell suspensions. This was fully inhibited by adding superoxide dismutase to the reaction medium (data not shown), indicating that superoxide production was involved. The rate of superoxide generation was low (20 -100 pmol min Ϫ1 (mg wet weight) Ϫ1 ), which explains why there was no observable effect of superoxide dismutase on the reduction of ferric citrate by the cells (which, under our experimental conditions, proceeded at a 100 -500-fold higher rate). The rate of superoxide production was 2-3-fold higher with iron/copper-deprived cells than with control cells, and it was observed with both a wild-type or a ⌬fre2 strain, but not with a ⌬fre1 strain (data not shown). These results suggest that oxygen is an alternative, although low efficient, electron acceptor of the Fre1-dependent ferrireductase system.

Effects of DPI and Detergents on the Reduction of Resazurin and Ferric Citrate by Whole Cells and Isolated Plasma Membrane-Whole cells of S. cerevisiae
can reduce a variety of physiological and nonphysiological electron acceptor via the ferrireductase system (22). The nonpermeant dye resazurin is reduced to its highly fluorescent derivative resorufin, providing a very sensitive assay for the cell reductase activity. The reduction of resazurin was increased when cells were grown under iron/copper limitation and was strongly decreased in a ⌬fre1 mutant (data not shown). However, the residual reductase activity of a ⌬fre1 mutant (compared to the wild-type) varied, depending on the substrate used: the rate of ferric citrate reduction by ⌬fre1 cells was 10 -13% (iron/copper-de-prived cells) or 3-4% (iron/copper-rich cells) of the reduction rate by the wild-type grown in the same conditions. When resazurin was used as electron acceptor, these values were 25-27% and 20 -21%, respectively (data not shown). In addition, a ⌬fre1⌬fre2 strain showed no residual activity with ferric citrate as substrate and 10 -15% residual activity with resazurin as electron acceptor. This activity was still inducible since it was increased 2-3-fold when cells were grown under iron/ copper deprivation (data not shown). These observations suggest that a component other than Fre1p and Fre2p is able to transfer electrons to resazurin but not to ferric citrate. Reduction of both substrates by resting cells was strongly inhibited by DPI, a powerful inhibitor of the neutrophil NADPH oxidase (21). However, while ferric citrate reduction was completely inhibited at DPI concentrations of 5-10 M, there was always a residual activity (15-20%) when resazurin was the electron acceptor (Fig. 6A). This difference was even more pronounced when a ⌬fre1 strain was used instead of the wild-type (Fig. 6A). In this case, 35-40% of the resazurin reductase activity was insensitive to 5-10 M DPI. Fig. 6 (B, C, and D) shows the effect of the DPI concentration on the reduction of Fe 3ϩ and resazurin by wild-type cells, ⌬fre1 cells, and isolated plasma membranes (NADPH as electron donor). The results obtained with wildtype cells were very similar to those obtained with a ⌬fre2 strain, and the results obtained with isolated plasma membranes were identical, whatever the strain used (data not shown). Only the NADPH-dependent ferrireductase activity associated with plasma membranes was sensitive to DPI. The inhibitor had no effect on the NADH-dependent activity (data not shown). The effect of DPI on ferric reduction by wild-type (or ⌬fre2) cells was biphasic (Fig. 6B), and two apparent K i values (0.2-0.3 M and 1-2 M) were calculated. A single K i of about 0.3 M (high affinity binding) was calculated when ⌬fre1 cells were used (Fig. 6C), and a K i of about 1 M (low affinity binding) was calculated with isolated plasma membranes (Fig.  6D). These results suggest that there are at least two DPIsensitive components in the Fre1-dependent ferrireductase system, differing in their sensitivity to the inhibitor. This is directly comparable to the neutrophil NADPH oxidase (21). In this case, it was suggested that the component sensitive to low DPI concentrations was the heme-binding component of cytochrome b 558 , while higher concentrations of DPI might inhibit the NADPH dehydrogenase component of the oxidase complex (21). We suggest that a similar two-component system is involved in ferric reduction by wild-type and ⌬fre2 cells, with the heme-binding component, Fre1p, being sensitive to low concentrations of DPI, and the NADPH dehydrogenase component being sensitive to higher DPI concentrations. Only this latter component, acting as a diaphorase, would contribute to iron reduction by isolated plasma membranes (Fig. 6D). This is consistent with (i) our observation that Fre1p does not contribute to the ferrireductase activity in isolated plasma membranes (see above) and (ii) our previous attempts to isolate the plasma membrane-bound ferrireductase that led to partial purification of a flavoprotein (FMN) with NADPH-dependent diaphorase activity (12). The NADPH dehydrogenase component of the neutrophil NADPH oxidase can also act as a diaphorase (23). In contrast, the Fre2-dependent reductase system seemed to depend on a single DPI-sensitive component (Fig. 6C). The situation was more complex when resazurin was used as the electron acceptor (Fig. 6, A-D) because part of the reductase activity was not inhibited by 5-10 M DPI. This part of the reductase activity probably corresponded to the reductase activity that was measured in a ⌬fre1⌬fre2 strain with resazurin, but not iron, as substrate (see above) and which was not sensitive to DPI (data not shown). Resazurin may be able to bypass FIG. 5. Low temperature absorbance spectra of purified yeast plasma membranes. S150-2B⌬fre1 was transformed by FRE1(c-myc) in a multicopy plasmid (YEp352) and grown under iron/copper deprivation. The plasma membranes were purified, reduced with a few crystals of dithionite, and the main spectrum was obtained with 18 mg of protein/ml. Inset A, absorbance spectra of plasma membranes isolated from cells grown in either iron/copper-rich medium (dotted line) or iron/copper-deficient medium (continuous line) (about 15 mg of protein/ml in both cases). Inset B, absorbance spectra of plasma membranes isolated from S288C⌬fre1⌬fre2 (dotted line, 20 mg of protein/ml), S288C⌬fre1⌬fre2 overexpressing FRE2 (dashed line, 8 mg of protein/ ml), or S288C⌬fre1⌬fre2 overexpressing FRE1 (continuous line, 13 mg of protein/ml); cells were grown under iron/copper deprivation. The spectra in dashed/dotted lines have a shoulder at 551 nm and a small peak at 558 nm, which probably correspond to the ␣ bands of contaminant mitochondrial b and c 1 cytochromes. the DPI-sensitive electron flow of the reductase system, similar to 2,6-dichloroindophenol with the neutrophil NADPH oxidase (21). Resazurin would then be able to accept electrons from Fre1p (Fre2p) in a DPI-sensitive way and from another component, possibly the NADPH dehydrogenase component, upstream of the DPI-sensitive sites. This would also account for the complex effects of detergents on the cell reductase activity. We previously reported the strong inhibitory effect of some ionic detergents on the ferrireductase activity of whole cells (22). Deoxycholate has a similar effect on the neutrophil NADPH oxidase, and it was suggested that the detergent could abrogate interactions between two components of the oxidase system (21). Working with whole cells (wild-type or ⌬fre2 cells) we found that the detergent Zwittergent 3-16 (6 -8 M) inhibited the reduction of ferric citrate by 85-100% and the reduction of resazurin by 50 -65%. The same detergent had no effect on the NADPH-dependent reductase activity associated with isolated plasma membranes (data not shown). These observations suggest that, as in the neutrophil oxidase, the detergent could act by weakening interactions between two components of the reductase system. Resazurin reduction would be less sensitive to this effect because of its capacity to partly bypass the electron flow from NADPH to Fre1p. Fig. 7 is a model of the system.
Utr1p, a Cytosolic Factor Involved in Ferrireductase Activity of the Cells-In their search for ferrireductase-deficient mutants, Anderson et al. (24) identified a strain which was mutated in the UTR1 gene encoding a protein of unknown function (25). The product of the UTR1 gene was predicted to be cytosolic, was not involved in the regulation of FRE1 expression, and transcription of this gene was not regulated by iron. When our wild-type strain S150-2B was transformed by a multicopy plasmid bearing either the UTR1 gene or the FRE1 gene, the ferrireductase activity of the cells was only slightly increased (by less than 20%). However, when cells were transformed by both genes on multicopy plasmids, the resulting transformants showed ferrireductase activity that is 5-7-fold higher than in wild-type grown under iron/copper-deficient conditions and up to 20-fold higher than in wild-type grown under iron/copperrich conditions (data not shown). No significant increase in ferrireductase activity was observed when FRE2, and not FRE1, was overexpressed together with UTR1. In contrast to FRE1, transcription of the UTR1 gene was not affected by iron/copper deprivation (data not shown). However, our results clearly show that the FRE1 and UTR1 gene products act synergistically on the cell ferrireductase activity. Utr1p should therefore be considered to be a true component of the ferrireductase system, and we propose to rename the UTR1 gene FRE5 (it was originally named FRE2 by Anderson et al. (24) but this name is now devoted to the gene described by Georgatsou and Alexandraki (11); we do not suggest "FRE3" since the sequences of two new genes homologous to FRE1/FRE2 are available in data banks). The effect of manipulating the FRE1 and UTR1 copy number on ferrireductase activity was strongly dependent on the yeast strain used. For example, wild-type cells of W303-1B showed a 2-3-fold lower reductase activity than S150-2B, but this activity was increased (about 3-fold and 2-fold, respectively) when either FRE1 alone or UTR1 alone was overexpressed. The effect of overexpressing both genes was less pronounced in W303-1B (4-fold increase in reductase activity) than in S150-2B (data not shown).
The predicted product of the UTR1 gene is a 59.4-kDa protein of 530 amino acids which has significant homologies with the product of POS5, a gene recently described whose disruption causes cells to become more sensitive to hydrogen peroxide (26). It is tempting to speculate that the UTR1/FRE5 gene product could play a role in the yeast ferrireductase system similar to that of the cytosolic factors (p47 and p67) in the neutrophil NADPH oxidase. Sequence alignments using Clustal w (version 1.5) multiple sequence analysis software did not allow to show any overall homology between the product of UTR1/FRE5 and the neutrophil cytosolic factors p47 and p67. However, a short segment of Utr1p (amino acids 324 -354) showed 30% identity with amino acids 161-189 of p47, and 19% identity with amino acids 245-273 of p67 (data not shown). In p47 and p67, these segments are part of larger motifs (50 amino acids) similar to a motif found in the noncatalytic domain of src-related tyrosine kinases (27). DISCUSSION Genetic studies have indicated that the ferrireductase activity of S. cerevisiae depends on two transmembrane reductases, the products of the FRE1 and FRE2 genes, which are believed to transfer electrons directly from intracellular NAD(P)H to extracellular ferric chelates (28). However, enzymological and biochemical data do not fit with this simple model. The increase in ferrireductase activity of whole cells subjected to iron/copper deprivation does not correlate with the increase in FRE1 transcription, 2 and cells bearing FRE1 on a multicopy plasmid do not always show, depending on the strain used, a significant increase in ferrireductase activity over that of cells with a single copy of the gene, suggesting that other component(s) could be limiting. A study on reduction of different substrates by whole cells and isolated plasma membranes led us to suggest that the yeast plasma membrane redox system is actually a complex chain of electron carriers (22). The present study demonstrates that, at least in vitro, Fre1p as such has no reductase activity and that a plasma membrane NADPH dehydrogenase is responsible for the ferrireductase activity in vitro and is probably also involved in transmembrane electron transfer in vivo. In isolated plasma membranes, electrons are directly transferred from this component, acting as a diaphorase, to the electron acceptor(s). Thus, the electron acceptors have direct access to this component in isolated membranes, but this is probably not the case in vivo, except for hydrophobic substrates like resazurin. It also seems probable that the electron flow from NADPH to Fe 3ϩ in vivo needs the transplasma membrane potential (⌬) to be preserved (22).
Our results show that there are striking similarities between the ferrireductase system of yeast and the neutrophil NADPH oxidase. The latter is composed of two plasma membranebound redox centers, a NADPH flavodehydrogenase and a lowpotential cytochrome b 558 , both of which are sensitive to DPI, plus additional proteins which are located in the cytosol of dormant cells and are translocated to the plasma membrane during activation (reviewed in Ref. 10). One of these proteins (p47) is phosphorylated during neutrophil activation. The yeast ferrireductase system can use oxygen as electron acceptor to produce superoxide radicals, and the Fre1p has the spectral characteristics like the neutrophil cytochrome b 558 . In addition, there is a NADPH flavodehydrogenase (12) distinct from Fre1p in the yeast plasma membranes. Its activity is increased under the same conditions that cause whole cell ferrireductase activity to be induced. A protein kinase A-dependent phosphorylation step is also likely to occur during "activation" of the ferrireductase system, since full derepression under iron/copper deprivation needs the integrity of the Ras/cAMP pathway (29). Such a phosphorylation step may affect the putative cytosolic factor Utr1/Fre5p. The ferrireductase activity is probably regulated at several levels, both transcriptional and post-transcriptional. For example, a peak of ferrireductase activity occurs in the late exponential growth phase even in iron-rich medium (12), i.e. when FRE1 transcription is repressed while UTR1/FRE5 transcription remains unaffected. Finally, it was shown recently that gp91, the main component of cytochrome b 558 , can act as a proton pump (30), which also parallels our previous findings that ferricyanide reduction by the cells is accompanied by a measurable decrease in the extracellular pH (22) and that ferrireductase-deficient mutants are also affected in their capacity to acidify the surrounding medium (31). Fre1p involvement in proton translocation is currently being examined, and we are also working on the nature of Utr1/Fre5p and its role in regulating ferrireductase activity.