Oligomeric Yeast Frataxin Drives Assembly of Core Machinery for Mitochondrial Iron-Sulfur Cluster Synthesis*

Mitochondrial biosynthesis of iron-sulfur clusters (ISCs) is a vital process involving the delivery of elemental iron and sulfur to a scaffold protein via molecular interactions that are still poorly defined. Analysis of highly conserved components of the yeast ISC assembly machinery shows that the iron-chaperone, Yfh1, and the sulfur-donor complex, Nfs1-Isd11, directly bind to each other. This interaction is mediated by direct Yfh1-Isd11 contacts. Moreover, both Yfh1 and Nfs1-Isd11 can directly bind to the scaffold, Isu1. Binding of Yfh1 to Nfs1-Isd11 or Isu1 requires oligomerization of Yfh1 and can occur in an iron-independent manner. However, more stable contacts are formed when Yfh1 oligomerization is normally coupled with the binding and oxidation of Fe2+. Our observations challenge the view that iron delivery for ISC synthesis is mediated by Fe2+-loaded monomeric Yfh1. Rather, we find that the iron oxidation-driven oligomerization of Yfh1 promotes the assembly of stable multicomponent complexes in which the iron donor and the sulfur donor simultaneously interact with each other as well as with the scaffold. Moreover, the ability to store ferric iron enables oligomeric Yfh1 to adjust iron release depending on the presence of Isu1 and the availability of elemental sulfur and reducing equivalents. In contrast, the use of anaerobic conditions that prevent Yfh1 oligomerization results in inhibition of ISC assembly on Isu1. These findings suggest that iron-dependent oligomerization is a mechanism by which the iron donor promotes assembly of the core machinery for mitochondrial ISC synthesis.

ISC 3 biosynthesis is an essential function that eukaryotic cells initiate in mitochondria and probably other cellular compartments using three core components: a sulfur donor, an iron donor, and an ISC assembly scaffold (1,2). In yeast mitochondria, the cysteine-desulfurase, Nfs1, and the iron-chaperone, Yfh1, are believed to provide sulfur and iron, respectively, for ISC assembly on the Isu1 scaffold (1), whereas the Nfs1-binding protein, Isd11, has been shown to stabilize Nfs1 (3). These components are highly conserved and the human orthologues of Yfh1 (frataxin), Isu1 (ISCU), and Isd11 (ISD11) are implicated in the etiology of severe disorders including Friedreich ataxia and mitochondrial myopathy (4).
Previous studies have underscored the complexity of the interactions among eukaryotic ISC assembly components as well as their metal dependence. Supplementation of mitochondrial lysates with Fe 2ϩ under aerobic conditions led to co-isolation of Yfh1 and Isu1 along with Nfs1 and Isd11 by pulldown or immunoprecipitation assays (5)(6)(7). Furthermore, aerobic preincubation of histidine-tagged Yfh1 monomer with Fe 2ϩ enabled Isu1 to be pulled down by Yfh1 in the absence of other proteins (5). These studies have led to the current view that iron delivery for yeast ISC synthesis involves direct contacts between iron-loaded monomeric Yfh1 and Isu1 (5)(6)(7). Although Yfh1 oligomerization is normally coupled with iron binding, oxidation, and storage (5,8), the possibility that Isu1 might also interact with oligomeric Yfh1 has remained largely unexplored.
Similar to Yfh1, human frataxin was found to interact with multiple ISC assembly components in human cells; however, in this case immunoprecipitation data suggested that frataxin binds to ISCU indirectly, via nickel-dependent contacts with ISD11 (9). Whether direct interactions occur between Yfh1 and Isd11 has not yet been examined.
While previous studies focused primarily on Yfh1-Isu1 and frataxin-ISD11 interactions, it is likely that the coordinate delivery of potentially toxic sulfur and iron to Isu1/ISCU involves multiple close interactions whereby the sulfur donor and the iron donor simultaneously interact with each other and with the ISC scaffold, as proposed for prokaryotic ISC assembly (10). However, it is currently unknown whether monomeric Yfh1/frataxin may form direct contacts with more than one partner, and the structure of the eukaryotic ISC assembly machinery is completely undefined. We show that iron oxidation-dependent oligomerization enables Yfh1 to have simultaneous direct interactions with Nfs1-Isd11 and Isu1. Our data provide insights about the sequence of events and the molecular architecture required for the initial step in mitochondrial ISC assembly.

EXPERIMENTAL PROCEDURES
Protein Expression and Purification-Yfh1 monomer, trimer, and 24-mer were expressed in Escherichia coli and purified as described (11). For expression of Isu1 His , a DNA fragment corresponding to the predicted mature form of Isu1 (residues 28 -166) (12) was cloned in vector pET28b (Novagen) in two steps: first, the fragment was cloned into the EcoRI and HindIII sites of pET28b, and second, an NcoI-EcoRI fragment was removed. This resulted in the addition of seven amino acids plus a six-histidine tag to the Isu1 C terminus. Isu1 His was expressed in E. coli strain BL21 and purified by nickel affinity chromatography (5-ml HiTrap Ni-column, Amersham Biosciences) followed by preparative gel-filtration chromatography (16-mm ϫ 50-cm column packed with Superdex 75 support from Amersham Biosciences) essentially as described for Haemophilus influenzae IscU (13). Briefly, cells were grown at 25°C in 1 liter of Luria broth containing 30 g/ml kanamycin. Protein expression was induced at A 600 ϭ 0.6 with 0.5 mM isopropyl-␤-D-thiogalactopyranoside, and growth continued at 15°C for ϳ18 h. Cells were harvested and disrupted by sonication in 50 mM NaH 2 PO 4 , pH 8.0, 300 mM NaCl, 10 mM imidazole, 5 mM ␤-mercaptoethanol, containing 1 tablet/50 ml of buffer of complete protease inhibitor mixture EDTA-free (Roche Diagnostics). Isu1 His was eluted from the HiTrap Nicolumn in 50 mM NaH 2 PO 4 , pH 8.0, 300 mM NaCl, 5 mM ␤-mercaptoethanol, between 160 and 240 mM imidazole. Gelfiltration chromatography was carried out in 10 mM HEPES-KOH, pH 7.8, 100 mM NaCl, 5 mM ␤-mercaptoethanol. Isu1 His was stable in 10 mM HEPES-KOH, pH 7.8, 100 mM NaCl, 5 mM ␤-mercaptoethanol at a protein concentration Յ2.0 mg/ml.
For production of untagged Isu1, a DNA fragment corresponding to residues 28 -166 was cloned in the NdeI and XhoI sites of pET28b, which resulted in the addition of an N-terminal 6-histidine tag followed by a thrombin cleavage site. Protein was expressed as described above, and after nickel affinity chromatography, the histidine tag was removed by incubation with thrombin (Sigma, 0.1 mg of thrombin per 100 mg of His-Isu1) for ϳ12 h at 4°C. Uncleaved His-Isu1 was removed by nickel affinity chromatography, and fractions corresponding to untagged Isu1 were purified by Superdex 75 gel-filtration chromatography as described above.
Two DNA fragments corresponding to residues 36 -498 of Nfs1 and 12-95 of Isd11-the predicted mature forms of these proteins as determined with MitoProtII (14)-were cloned into the NdeI and XhoI, and NcoI and HindIII sites, respectively, of vector pCDFDuet (Novagen), which resulted in the addition of a C-terminal S-tag to Nfs1 and an N-terminal 6-histidine tag to Isd11. A thrombin cleavage site was also introduced immediately after the 6-histidine tag on Isd11. Nfs1 and Isd11 were co-expressed in E. coli essentially as described for the human orthologs (15). Briefly, BL21 cells were grown at 25°C in 1 liter of Luria broth containing 30 g/ml streptomycin. Protein expression was induced at A 600 ϭ 0.6 with 0.5 mM isopropyl-␤-D-thiogalactopyranoside. Growth was continued for ϳ18 h at 15°C, after which cells were harvested, resuspended in lysis buffer (50 mM NaH 2 PO 4 , pH 8.0, 300 mM NaCl, 20 mM imidazole, containing 50 M pyridoxal phosphate, 10% glycerol, and 1 tablet/50 ml of buffer of complete protease inhibitor mixture EDTA-free), and disrupted by sonication. The lysate was centrifuged at 25,000 ϫ g for 20 min at 4°C, and the soluble fraction was loaded onto a 5-ml HiTrap nickel column. The Nfs1-Isd11 His complex was eluted in lysis buffer without protease inhibitors between 150 and 250 mM imidazole and was further purified by preparative gel-filtration chromatography (16-mm ϫ 50-cm column packed with Superdex 200 support from Amersham Biosciences) in 10 mM HEPES-KOH, pH 7.8, 150 mM NaCl, 50 M pyridoxal phosphate. The Nfs1-Isd11 His complex was stable in this buffer at a protein concentration of Յ5 mg/ml.
A DNA fragment corresponding to the predicted mature form of Isd11 was cloned in the NdeI and HindIII sites of pET28b resulting in the addition of an N-terminal 6-histidine tag followed by a thrombin cleavage site as described above. For expression of Isd11 His we used the same procedure employed for co-expression of Nfs1-Isd11 His , except that induction with isopropyl-␤-D-thiogalactopyranoside was for 2 h at 37°C, because this resulted in a higher protein yield relative to other conditions tested. Isd11 His was expressed at very low levels and exhibited a strong tendency to precipitate out of solution. Therefore, upon HiTrap nickel-affinity chromatography, fractions containing mostly Isd11 His monomer were pooled and concentrated to ϳ0.4 mg of protein/ml in 10 mM HEPES-KOH, pH 7.8, 150 mM NaCl.
As determined by SDS-PAGE, Yfh1, Isu1 His , untagged Isu1, and the Nfs1-Isd11 His complex exhibited Ͼ90% purity, whereas isolated Isd11 His exhibited Ͼ80% purity (supplemental Fig. S1). Protein concentration was determined by use of the BCA kit (Pierce) and is expressed in all cases per subunit. Iron content of purified proteins was measured as described (8).
In Vitro Pulldown Assays-Nickel-nitrilotriacetic acid (Ni-NTA)-agarose (Qiagen) affinity pulldown experiments were started by incubating increasing concentrations of the appropriate His-tagged protein (Isu1 His or Nfs1-Isd11 His ) with a fixed concentration (typically 4 M) of each partner protein (Yfh1 and/or Isu1). Unless otherwise stated, binding was allowed to occur at 30°C for 30 min in 100 l of binding buffer (10 mM HEPES-KOH, pH 7.3, 100 mM NaCl, 20 mM imidazole, 5 mM ␤-mercaptoethanol). Stock solutions of ferrous ammonium sulfate (4 -10 mM, pH ϳ 4) were freshly prepared in water previously deaerated by purging with argon gas (Ͻ0.2 ppm O 2 ). To generate holo-Yfh1 proteins, each apoform (typically 40 M) was aerobically incubated with 10 equivalents of Fe(NH 4 ) 2 (SO 4 ) 2 for ϳ15 h in 10 mM HEPES-KOH, pH 7.3, at room temperature on a nutator mixer, and added to the binding reaction to achieve the appropriate final concentration. The Fe 2ϩ /subunit molar ratio used to generate holo-Yfh1 proteins was lower than the maximum loading capacities of both Yfh1 and Yfh1-Y73A (Ͼ50 and ϳ50 atoms of iron/subunit) (16,17). In addition, iron loading was carried out under established aerobic conditions that lead to formation of a stable ferric iron mineral inside Yfh1 or Yfh1-Y73A oligomers (16,18). Accordingly, ultrafiltration analysis of holo-Yfh1 forms in Ultrafree-0.5 devices (Millipore) (8,19) showed that Ն96.5% of the total iron was protein-bound (Ն38 M), whereas soluble free iron was Յ0.35 M and insoluble iron was Յ2 M. With holoproteins subjected to ultrafiltration, the absolute amounts of all three Yfh1 forms pulled down by Isu1 His were not significantly different as compared with untreated holoforms (see Fig. 1D), and therefore ultrafiltration was omitted from subsequent experiments using holo-Yfh1 forms. Beads were washed and equilibrated with one volume of binding buffer containing 0.01% bovine serum albumin. Equilibrated beads (20 l of a slurry containing equal volumes of beads and buffer) were added to each binding reaction and washed four times with 1 ml of binding buffer containing 0.1% Triton X-100. Bound proteins were eluted by boiling the beads for 5 min in Laemmli buffer and resolved by SDS-PAGE on Criterion precast Tris-HCl gel (Bio-Rad). The flow-through of each binding reaction was also analyzed on a separate gel as a control. This demonstrated that essentially 100% of the Isu1 His or Nfs1-Isd11 His added to the reactions was bound to the Ni-NTA beads. Gels were stained with SYPRO Orange (Invitrogen), and bound protein was quantified relative to the internal standard using ImageQuant 5.0 on a Storm 860 (Amersham Biosciences).
In Vitro Gel-filtration Chromatography Assays-Unless otherwise stated, samples (typically 250 l) containing one or more of the proteins under investigation (typically 40 M each protein) were incubated in binding buffer for 30 min at 30°C. The binding reaction was centrifuged for 5 min at 20,000 ϫ g and loaded onto the appropriate column equilibrated with the same buffer. Fractions were collected, and aliquots were analyzed by SDS-PAGE and SYPRO Orange staining. To generate holo-Yfh1 proteins, each apoform (typically 80 M) was aerobically incubated with 10 equivalents of Fe(NH 4 ) 2 (SO 4 ) 2 as described above and added to the binding reaction to achieve the required final concentration.
Iron-Sulfur Cluster Assembly Assay-Apo-␣ or apo-␣ 24 Y73A was aerobically incubated with Fe(NH 4 ) 2 (SO 4 ) 2 at a Fe 2ϩ /subunit ratio of 1:2 for 75 min in 20 mM HEPES-KOH, pH 7.3, 150 mM NaCl (19), in parallel with samples containing the same iron concentration but no Yfh1. Ultrafiltration analysis (8,19) showed that, in both the holo-␣ and holo-␣ 24 Y73A preparations, protein-bound iron was Ն46.5 M (Ն93% of the total 50 M iron added to each protein), whereas free and insoluble iron were Յ2 M and Յ2.5 M, respectively. To verify the extent of iron oxidation, each holo-Yfh1 form was centrifuged at 20,000 ϫ g for 10 min to eliminate insoluble iron, and the supernatant incubated with the Fe 2ϩ chelator, ␣-␣Ј-bipyridine (8,19), which showed that Ն95.5% of the protein-bound iron was present in ferric form as expected. All buffers and solutions were purged with argon gas (Ͻ0.2 ppm O 2 ) in vials tightly sealed with rubber septa. Each reaction was assembled in a quartz cuvette that was purged with nitrogen gas before and after addition of each component, and finally sealed with a tight fitting rubber septum. [2Fe-2S] synthesis was started with addition of 50 M iron (80 l), provided as holo-␣ or holo-␣ 24 Y73A or directly, to anaerobic reactions containing 6.25 M Isu1 His , 2.5 mM Na 2 S, 5 mM dithiothreitol, and 2 mM L-cysteine in 20 mM HEPES-KOH, pH 7.3, 150 mM NaCl (800-l final reaction volume). Reactions containing 0.5-1.0 M Nfs1-Isd11 His instead of Na 2 S were set up in a similar manner. Reactions containing holo-␣(Fe 2ϩ ) were similarly set up except that apo-␣ was incubated with Fe(NH 4 ) 2 (SO 4 ) 2 anaerobically (8). ISC synthesis was monitored spectrophotometrically at A 426 .
Mitochondria (800 l at ϳ12 mg of total protein/ml) were suspended in 10 mM HEPES-KOH, pH 7.3, 100 mM NaCl, and disrupted by sonication, followed by treatment with 0.5% Triton X-100 for 30 min on ice. The mitochondrial lysate was cleared by centrifugation at 20,000 ϫ g for 10 min at 4°C, and the supernatant (ϳ4.5 mg of total protein) was used for the experiment. Pulldown assays were performed with Ni-NTAagarose beads (40 l slurry) with or without bound Isu1 His or Nfs1-Isd11 His , which were added to the mitochondrial lysate described above in a final 1-ml volume of 10 mM HEPES-KOH, pH 7.3, 100 mM NaCl, 20 mM imidazole, 0.05% Triton X-100. The final concentration of Isu1 His or Nfs1-Isd11 His was 0.8 M.
Binding reactions were incubated for 2 h at 4°C on a nutator mixer. The agarose beads were washed three times in 800 l of buffer and resuspended in 40 l of Laemmli buffer after the last wash. Polyclonal antibodies were raised in rabbits using purified recombinant proteins as antigens.

Stable Yfh1-Isu1 Interactions Require Yfh1 Oligomerization-
We postulated that the iron dependence of Yfh1-Isu1 interactions (5-7) might reflect stepwise assembly of Yfh1 monomer into trimer, 24-mer, and higher order oligomers, which occurs in an iron-dependent manner both in vitro (5,21) and in yeast (17). We produced recombinant forms of Yfh1, Isu1, Nfs1-Isd11, and Isd11 in E. coli (supplemental Fig. S1). To distinguish whether Yfh1 interacts with its partners in an iron-or oligomerization-dependent manner, we used iron-free wildtype Yfh1 monomer (␣) as well as stable iron-free trimer (␣ 3 Y73A ) and 24-mer (␣ 24 Y73A ) produced during expression of the Yfh1-Y73A variant in E. coli (11). The Y73A mutation stabilizes an assembly-competent conformation of Yfh1 and thereby accelerates its assembly kinetics both in E. coli and in yeast without changing its iron-chaperone function (17).
We initially tested binding of iron-free (apo) Yfh1 proteins to Isu1 His by use of a pulldown assay, a method used in previous studies of Yfh1-Isu1 interactions (5,7). Binding to Isu1 His was much stronger with apo-␣ 24 Y73A protein than with apo-␣ or apo-␣ 3 Y73A (Fig. 1, A and B). The amounts of Yfh1 pulled down by Isu1 His were not simply proportional to the number of Yfh1 subunits in monomer, trimer, and 24-mer because we obtained essentially identical results with apo-␣ and apo-␣ 3 Y73A . Moreover, each of the three protein preparations contained Յ0.02 atoms of iron per Yfh1 subunit, and the binding of apo-␣ 24 Y73A to Isu1 His was only modestly decreased by EDTA (not shown), indicating that the interaction of Isu1 His with apo-␣ 24 Y73A was mostly metal-independent. Next, ␣ and ␣ 3 Y73A were preincubated with Fe 2ϩ aerobically under established conditions that induce stepwise assembly of monomer and trimer into 24-mer, and promote conversion of Fe 2ϩ to Fe 3ϩ inside Yfh1 oligomers via ferroxidase activity and autoxidation (8,16). Binding of holo-␣ or holo-␣ 3 Y73A to Isu1 His was stronger relative to the respective apo form and approached that of holo-␣ 24 Y73A (Fig.  1, C and D). We hypothesized that these results might reflect the presence of different levels of 24-mer in the ␣, ␣ 3 Y73A and ␣ 24 Y73A preparations before and after iron loading. To assess this, we analyzed Yfh1-Isu1 His interactions in solution by gelfiltration chromatography.
Isu1 His was eluted in fractions with apparent molecular mass ϳ44 -17 kDa (Fig. 2A). An untagged version of Isu1 exhibited a very similar elution profile (supplemental Fig.  S1d). The elution profiles of the apo-␣, -␣ 3 Y73A , and -␣ 24 Y73A proteins showed that they contained primarily Yfh1 monomer, trimer, and 24-mer, respectively, although additional species were present in each preparation (Fig. 2 (B and F) and supplemental Fig. S2b). The distribution of Isu1 His remained mostly unchanged upon incubation with apo-␣ or apo-␣ 3 Y73A (Fig. 2C and supple-mental Fig. S2c); in contrast, a significant proportion of Isu1 His was co-eluted with 24-mer upon incubation with apo-␣ 24 Y73A (Fig. 2G). When apo-␣ and apo-␣ 3 Y73A were preincubated with iron, they underwent iron-dependent oligomerization ( Fig. 2D and supplemental Fig. S2d), and upon incubation with these proteins, most Isu1 His was co-eluted with the newly assembled 24-mer ( Fig. 2E and supplemental Fig. S2e). The co-elution profiles were almost indistinguishable from that obtained upon incubation of Isu1 His with holo-␣ 24 Y73A (Fig. 2I). We verified that the distribution of Isu1 His remained unchanged when the protein was directly incubated with iron (not provided as holo-Yfh1; supplemental Fig. S3). We also verified that untagged Isu1 exhibited very similar interactions as Isu1 His upon incubation with the ␣ preparation or ␣ 24 Y73A preparation before and after iron loading (supplemental Fig. S4,  a-g).
Samples containing equivalent concentrations of Yfh1 (provided as holo-␣ 24 Y73A ), and untagged Isu1 were analyzed by Sephacryl 300 gel-filtration chromatography (fractionation range 1.5 MDa to 10 kDa). We observed a symmetrical peak with a molecular mass of ϳ400 kDa, and once again SDS-PAGE showed co-elution of 24-mer and Isu1 (Fig. 3). Together, these results indicate that Isu1 interactions with Yfh1 are initiated by Yfh1 oligomerization, which in the case of wild-type Yfh1 occurs concurrently with iron binding and oxidation. NaCl, 20 mM imidazole, 5 mM ␤-mercaptoethanol, as described under "Experimental Procedures." Ni-NTA-agarose was added, and bound protein (100%) was separated on 15% SDS-PAGE and detected by staining with SYPRO Orange. S, 2.5 g (180 pmol) Yfh1 used as internal standard. In B the amounts of bound Yfh1 pulled down in three independent experiments were measured by densitometry with ImageQuant 5.0, and data were fitted with a one site binding equation using GraphPad Prism 4.03. In C holo-Yfh1 proteins were obtained by aerobically incubating the corresponding apoforms with 10 iron equivalents as described under "Experimental Procedures." In D, the amounts of bound holo-Yfh1 pulled down in four independent experiments were measured as in B. Data were obtained with holoproteins that were (n ϭ 2) or were not (n ϭ 2) subjected to ultrafiltration to remove residual levels of unbound iron. Upon ultrafiltration, the absolute amounts of all three Yfh1 forms pulled down by the scaffold were somewhat lower, suggesting that this manipulation may cause some damage to the surface of Yfh1 24-mer. However, all four experiments consistently showed higher binding affinity of holo-␣ and holo-␣ 3 Y73A relative to their apoforms (D versus B).

FIGURE 2. Oligomerization-dependent interactions of Yfh1 with Isu1 His analyzed by gel-filtration chromatography. A-I, Isu1
His was incubated in the absence or presence of an equivalent concentration of the indicated Yfh1 preparations (40 M Yfh1 subunit), each sample was directly applied to a Superdex 75 column under the same solution conditions used in Fig. 1, and fractions were analyzed by 15% SDS-PAGE and protein staining. The distribution of Isu1 His is shown at the top of each set of gels for reference.

Nfs1-Isd11 Complex Directly
Interacts with Oligomeric Yfh1-Binding of Yfh1 proteins to Nfs1-Isd11 His complex was also examined by pulldown assay. Low amounts of apo-␣ and larger amounts of apo-␣ 24 Y73A were bound to the Nfs1-Isd11 His complex (Fig. 4, A and B). Preloading the Yfh1 proteins with iron resulted in an increased affinity of holo-␣ for Nfs1-Isd11 His to a level comparable to that of holo-␣ 24 Y73A (Fig. 4, A and B). Similar results were obtained with isolated Isd11 His (Fig. 4C), suggesting that the interactions observed above were mediated at least in part by Yfh1-Isd11 contacts. Isolated Nfs1 was not available to test for direct Yfh1-Nfs1 interactions.
Yfh1 (provided as holo-␣) and untagged Isu1 could bind to Nfs1-Isd11 His both separately and at the same time; notably, the amount of Yfh1 that was bound to Nfs1-Isd11 His did not change significantly in the absence or presence of Isu1 (Fig.  4D), suggesting that different surfaces of oligomeric Yfh1 are involved in interactions with Nfs1-Isd11 His and Isu1, respectively.

Iron-dependent Oligomerization of Yfh1 Promotes Formation of Stable Multicomponent Complexes-Equivalent concentrations of Isu1 His and
Nfs1-Isd11 His were incubated in the absence or presence of equivalent amounts of Yfh1 (as apo or holo ␣ or ␣ 24 Y73A ), and interactions analyzed in solution by Sephacryl 300 gelfiltration chromatography. Isu1 His was used in these experiments because it was less prone to degradation than untagged Isu1 (supplemental Fig. S1, a-d).
Low levels of Isu1 His were coeluted with Nfs1-Isd11 His , whereas the bulk of Isu1 His was eluted in fractions between 44 and 17 kDa (Fig.  5A, fractions ϳ72-84), as observed earlier by Superdex 75 gel-filtration chromatography ( Fig. 2A). Nfs1-Isd11 His and the associated Isu1 His were eluted around ϳ300 kDa (Fig.  5A, fraction 58). The elution profiles of Nfs1-Isd11 and Isu1 His remained essentially unchanged when these proteins were incubated with apo-␣ (Fig. 5B), consistent with the inability of monomeric Yfh1 to form stable interactions with either component, as described above. Upon incubation with apo-␣ 24 Y73A , low levels of Nfs1-Isd11 His and Isu1 His were otherwise shifted to higher molecular weight fractions and were co-eluted with 24-mer (Fig.  5D, fractions ϳ44 -50). However, smaller proportions of Isu1 His and Nfs1-Isd11 His were associated with apo-␣ 24 Y73A under these conditions as compared with earlier analyses in which apo-␣ 24 Y73A had been separately incubated with either Isu1 His or Nfs1-Isd11 His .
In striking contrast, upon incubation with holo-␣, significant levels of Nfs1-Isd11 His and the bulk of Isu1 His were shifted to higher molecular weight fractions and were co-eluted with newly assembled oligomeric Yfh1 (Fig. 5C, fractions ϳ38 -54). A more prominent shift was observed with holo-␣ 24 Y73A (Fig.  5E, fractions ϳ38 -54) owing to the larger proportion of 24-mer present in this protein preparation. Direct addition of Ni-NTAagarose beads to fractions eluted from the column enabled the simultaneous pulldown of Yfh1 along with Isu1 His and Nfs1 as well as tagged and untagged Isd11 (supplemental Fig. S5, a and  b). Together, these results indicate that simultaneous binding of Yfh1 to Nfs1-Isd11 and Isu1 requires oligomerization of Yfh1 and is stabilized by Fe 3ϩ loading of Yfh1 oligomers.
Sulfur Delivery to Isu1 Regulates Iron Release from Oligomeric Yfh1-Binding of Isu1 to Fe 3ϩ -loaded oligomeric Yfh1 was not sufficient to induce iron release from Yfh1 ( Fig. 3 and not shown). We tested whether the ferric iron stored inside oligomeric Yfh1 might be released in the presence of Isu1, elemental sulfur provided as Na 2 S, and L-cysteine, which is thought to serve as a physiological source of reducing equivalents for reduction of Fe 3ϩ (10,22). Apo-␣ and apo-␣ 24 Y73A were once again loaded with Fe 2ϩ aerobically to induce oligomerization of wild-type Yfh1 and promote conversion of Fe 2ϩ to Fe 3ϩ inside Yfh1 oligomers. We then used an anaerobic assay monitoring [2Fe-2S] cluster formation by UV-visible absorption analysis over time (10) (Fig. 6A). Reactions containing Isu1 His and ferric iron without Yfh1 showed a small increase in A 426 relative to control reactions, consistent with some spontaneous assembly of [2Fe-2S] clusters on Isu1 His . There was a much greater and faster increase in A 426 when holo-␣ or holo-␣ 24 Y73A was incu-bated together with Isu1 His , consistent with Yfh1-catalyzed synthesis of [2Fe-2S] clusters (Fig. 6A). When L-cysteine was not included, the reaction proceeded very slowly, and the yield of [2Fe-2S] clusters after 120 min was only slightly higher compared with a reaction containing Yfh1-bound Fe 3ϩ but no Isu1 (A 426 ϭ 0.15 versus 0.11) (not shown). A lower yield was obtained with holo-␣ 24 Y73A . This can be explained by the low iron/subunit ratio (2:1) used in these experiments, which was largely below the maximal iron loading capacity of ϳ50 iron/subunit. In addition, under these conditions wild-type Yfh1 underwent limited oligomerization, as expected (21), such that the holo-␣ preparation used in these assays contained ϳ5 times less 24-mer than the holo-␣ 24 preparation (not shown). Because both preparations contained equal concentrations of protein-bound iron (ϳ47 M) and Isu1 (6.25 M), the Fe/24-mer and the Isu1/24-mer ratios were ϳ5 times lower in the holo-␣ 24 preparation, causing a lower rate of [2Fe-2S] synthesis.
To limit sulfur availability, we replaced 2.5 mM Na 2 S with low concentrations of Nfs1-Isd11 His (0.5-1.0 M), to provide elemental sulfur via cysteine desulfuration (15). Once again we observed a significant increase in A 426 in reactions containing both Yfh1 (provided as holo-␣) and Isu1 His , which was not observed in control reactions lacking either protein (Fig. 6B). A low yield was obtained with holo-␣ 24 Y73A (not shown), as expected given the low Fe/24-mer, Isu1/24-mer, and Nfs1-Isd11/24-mer ratios, as discussed above.
In the presence of holo-␣ and Isu1 His , the A 426 versus time curve changed from a fast hyperbola in the presence of 2.5 mM Na 2 S to a slow sigmoid in the presence of low concentrations of Nfs1-Isd11 complex, although the maximal A 426 was similar between the two conditions (Fig. 6, A and B). The [2Fe-2S] yield estimated with an extinction coefficient ⑀ ϭ 8 -11 mM Ϫ1 cm Ϫ1 (23) indicated full utilization of the iron initially stored inside oligomeric Yfh1. These data suggest that the ability to store ferric iron enables oligomeric Yfh1 to adjust the rate of iron delivery to the availability of sulfur. Similarly, Fe 3ϩ -loaded oligomers of CyaY (the E. coli orthologue of Yfh1) were shown to serve as the iron donor for ISC assembly on the IscU scaffold in the presence of the sulfur donor, IscS (10).  Fig. 1 except that 50 M pyridoxal phosphate was added to the binding buffer. Ni-NTA-agarose-bound protein (100%) was separated on 18% SDS-PAGE and detected by protein staining. S, 2.5 g (180 pmol) of Yfh1. B, the amounts of bound Yfh1 pulled down in three independent experiments were measured as in Fig. 1B. C, pulldown assays were performed under the experimental conditions described above except that isolated Isd11 His was used instead of Nfs1-Isd11 His complex. S, 5 g (360 pmol) Yfh1. D, pulldown assays were performed with increasing concentrations of Nfs1-Isd11 His incubated with holo-␣ or untagged Isu1 or both proteins together (4 M each). Solution conditions were as in A except that 50 mM imidazole and 200 mM NaCl were used to minimize nonspecific binding of untagged Isu1 to Ni-NTA beads. S, 5 g of Yfh1; SЈ, 5 g of Isu1.
Conversely, under anaerobic conditions in which CyaY was stable as Fe 2ϩ -loaded monomer, the protein inhibited ISC assembly (24). We therefore measured [2Fe-2S] synthesis under anaerobic conditions in which Yfh1 is also known to be stable as Fe 2ϩ -loaded monomer (25). The rate of [2Fe-2S] synthesis was only slightly increased in the presence of Yfh1 as compared with a reaction in which Fe 2ϩ was provided directly (Fig. 6C) and was significantly lower than the rate observed in the presence of Fe 3ϩ -loaded holo-␣ (Fig. 6C,  inset). These data suggest that anaerobic conditions inhibit the iron-chaperone function of both Yfh1 and CyaY, likely because they prevent Fe 2ϩ oxidation and protein oligomerization.
Limited Amounts of Yfh1/Isu1/Nfs1 Complexes are Detected in Vivo-Mitochondrial lysates were prepared from yeast cultures that had been grown without or with iron supplementation under established conditions that cause a small increment in mitochondrial iron content, from ϳ3 to ϳ15 nmol/mg of protein, which is associated with limited assembly of endogenous Yfh1 (17). Recombinant Isu1 His interacted with significantly larger levels of endogenous Yfh1 or Nfs1 when pulldown assays were performed with mitochondrial lysate prepared from cells supplemented with iron (Fig. 7A). This result is consistent with the idea that Isu1 interacts preferentially with oligomeric Yfh1. The iron-dependent increment in the amount of Nfs1 pulled down by Isu1 His could be explained by the presence of a fraction of Nfs1 bound to oligomeric Yfh1 that was pulled down through Isu1 His -Yfh1 contacts. Accordingly, Nfs1-Isd11 His interacted with significantly larger levels of endogenous Yfh1 in mitochondrial lysate from cells supplemented with iron, although ϳ5-fold higher levels of Yfh1 were required to detect this interaction (Fig. 7, B  versus A, top).
In yeast, Yfh1-Y73A exists as stable trimer that oligomerizes into 24-mer in an iron-dependent manner (17). Accordingly, recombinant Isu1 His pulled down endogenous Yfh1-Y73A in mitochondrial lysate from cells supplemented with iron, with an efficiency comparable to that observed with endogenous wild-type Yfh1 (Fig. 7C). In contrast, limited interactions were observed with two Yfh1 proteins lacking ferroxidase activity (Yfh1-D79A;D82A) or iron mineralization activity (Yfh1-E93A; D97A;E103A), which are required for efficient iron-dependent oligomerization of Yfh1 (20). Low rates of iron-dependent oligomerization correlated with limited binding to Isu1 His in vitro (supplemental Fig. S6, a-e), whereas there was no detectable binding to Isu1 His in mitochondrial lysate (Fig. 7C), probably because the levels of endogenous Yfh1 oligomer interacting with Isu1 His were below the limit of detection of our assay.

DISCUSSION
In this study we set out to assess how known properties of Yfh1, i.e. the ability to bind and oxidize Fe 2ϩ and the ability to oligomerize, might influence the interactions of Yfh1 with its ISC assembly partners, Isu1 and Nfs1-Isd11. We studied wildtype Yfh1 and the Yfh1-Y73A variant, which oligomerize in an iron-dependent and an iron-independent manner, respectively (17). The Y73A mutation is adjacent to a patch of acidic residues involved in iron binding and oxidation (20,25), and most likely stabilizes a conformation required to initiate oligomerization (11). In the case of wild-type Yfh1, this conformation is stabilized upon the binding and oxidation of Fe 2ϩ (8). Thus, the Y73A mutation enabled us to uncouple oligomerization from iron binding/oxidation, which cannot be independently evaluated in wild-type Yfh1.
Our data show that Yfh1 oligomerization plays an important role in Yfh1-Isu1 interactions as well as Yfh1-Nfs1/Isd11 interactions. The large protein surface of Yfh1 oligomers (18) may facilitate simultaneous contacts with Nfs1-Isd11 and Isu1. Moreover, concurrent binding of oligomeric Yfh1 to both Isu1 and Nfs1-Isd11, and of Isu1 to both Yfh1 and Nfs1-Isd11, may help in stabilizing these multicomponent assemblies. The additional stabilizing effect of the iron stored inside oligomeric Yfh1 may involve iron-mediated contacts between Yfh1 and Isu1 and/or protein conformational changes. Finally, the capacity of Yfh1 oligomers to store iron in ferric form (16,18) may further provide a mechanism to adjust the rate of iron delivery to that of sulfur delivery, as proposed for prokaryotic ISC assembly (10,23).
Yfh1 oligomerization is normally tightly coupled with the Yfh1-catalyzed oxidation of Fe 2ϩ that leads to formation of a stable ferric mineral within Yfh1 oligomers (5,16,21). Previously, this led us to propose that the main function of Yfh1 oligomerization is to detoxify redox-active iron thereby providing mitochondria with antioxidant protection and iron storage capacity (18,20). The data presented here indicate that Fe 3ϩloaded oligomeric Yfh1 may also function as the iron donor for yeast ISC assembly. Similarly, Fe 3ϩloaded oligomeric CyaY was previously shown to serve as the iron donor for ISC assembly on the IscU scaffold in the presence of the sulfur donor, IscS (10). Other studies have analyzed iron delivery by Yfh1 or CyaY under strictly anaerobic conditions that inhibit Fe 2ϩ oxidation and Yfh1/CyaY oligomerization (24,26). The physiological relevance of these conditions is uncertain, however, given that both Saccharomyces cerevisiae and E. coli are facultative, not obligate, anaerobes. During aerobic growth these unicellular organisms are directly exposed to atmospheric air, which is expected to result in relatively high intracellular or intramitochondrial oxygen concentrations (27). Notably, under strictly anaerobic conditions, Fe 2ϩ -loaded CyaY monomer inhibited FIGURE 6. Iron release from oligomeric Yfh1 depends on Isu1 and sulfur availability. A, Yfh1 proteins were aerobically loaded with Fe 3ϩ , and synthesis of [2Fe-2S] cluster was subsequently measured anaerobically in the presence of the indicated components and 2.5 mM Na 2 S as described under "Experimental Procedures." Plots show the mean Ϯ S.D. of three independent assays. In the absence of protein, the A 426 in buffer containing iron was higher compared with buffer without iron. The A 426 in buffer containing iron decreased by ϳ50% after table-top centrifugation (not shown), suggesting the signal was due to scattering possibly originating from the formation of insoluble iron sulfides. B, holo-␣ or iron was incubated with or without Isu1 His as in A except that Nfs1-Isd11 His was used instead of Na 2 S. Each plot represents one measurement except plot Isu1 ϩ 1.0 M Nfs1-Isd11 ϩ Holo-␣, which shows the mean Ϯ S.D. of three independent assays. C, reactions were carried out as in B except that apo-␣ was incubated with Fe(NH 4 ) 2 (SO 4 ) 2 anaerobically (8). Each plot shows the average of two reproducible assays. The inset shows holo-␣(Fe 2ϩ )-catalyzed ISC assembly on Isu1 His after subtraction of spontaneous ISC assembly activity in the absence of Yfh1 (i.e. the activity present in Isu1 ϩ 1.0 M Nfs1-Isd11 ϩ Fe 2ϩ ).
ISC assembly (24), which led to the proposal that CyaY and other frataxin orthologues participate in ISC assembly not as iron chaperones but as iron-dependent inhibitors of ISC formation (24). However, we have shown here that when Fe 2ϩ binding to Yfh1 is allowed to occur physiologically under aerobic conditions, the resulting Fe 3ϩ -loaded oligomers promote ISC assembly on Isu1 (Fig. 6). In contrast, this process is inhibited under strictly anaerobic conditions in which Yfh1 is stable as an Fe 2ϩ -loaded monomer (Fig. 6).
Although it is possible that changes in oligomeric state enable CyaY and Yfh1 to either promote or inhibit ISC biosynthesis, the available evidence indicates that both CyaY and Yfh1 promote ISC assembly under physiological conditions in vivo (1,28). Likewise, lack of human or mouse frataxin causes early defects in Fe-S enzymes, prior to, or in the absence of, detectable mitochondrial iron accumulation (29,30). These data do not support the proposal that frataxins are inhibitors of ISC synthesis and that the primary consequence of frataxin deficiency is an ISC surplus (24). The latter effect would promote iron accumulation and delay appearance of Fe-S enzyme defi-synthesis and limit iron toxicity. Oligomerization of Yfh1 responds to increments in mitochondrial iron levels (17), and simultaneous interactions of iron-loaded Yfh1 oligomer with Isu1 and Nfs1-Isd11 may enable productive iron release for ISC assembly and prevent potentially toxic increases in the mitochondrial concentration of free iron. Disassembly of wild-type Yfh1 oligomers upon iron release may subsequently facilitate dissociation of holo-Isu1 from Yfh1 and/or enable new associations between holo-Isu1 and molecular chaperones required for the transfer of ISC from holo-Isu1 to appropriate apoenzymes. Additional studies will be required to elucidate this mechanism biochemically and structurally in vitro and in vivo. These studies will be technically challenging but will be critical to achieve a better understanding of how the ISC assembly process functions in physiological conditions and disease states.