The Structure of the Complex between Yeast Frataxin and Ferrochelatase

Frataxin is a mitochondrial iron-binding protein involved in iron storage, detoxification, and delivery for iron sulfur-cluster assembly and heme biosynthesis. The ability of frataxin from different organisms to populate multiple oligomeric states in the presence of metal ions, e.g. Fe2+ and Co2+, led to the suggestion that different oligomers contribute to the functions of frataxin. Here we report on the complex between yeast frataxin and ferrochelatase, the terminal enzyme of heme biosynthesis. Protein-protein docking and cross-linking in combination with mass spectroscopic analysis and single-particle reconstruction from negatively stained electron microscopic images were used to verify the Yfh1-ferrochelatase interactions. The model of the complex indicates that at the 2:1 Fe2+-to-protein ratio, when Yfh1 populates a trimeric state, there are two interaction interfaces between frataxin and the ferrochelatase dimer. Each interaction site involves one ferrochelatase monomer and one frataxin trimer, with conserved polar and charged amino acids of the two proteins positioned at hydrogen-bonding distances from each other. One of the subunits of the Yfh1 trimer interacts extensively with one subunit of the ferrochelatase dimer, contributing to the stability of the complex, whereas another trimer subunit is positioned for Fe2+ delivery. Single-turnover stopped-flow kinetics experiments demonstrate that increased rates of heme production result from monomers, dimers, and trimers, indicating that these forms are most efficient in delivering Fe2+ to ferrochelatase and sustaining porphyrin metalation. Furthermore, they support the proposal that frataxin-mediated delivery of this potentially toxic substrate overcomes formation of reactive oxygen species.

In an oxidative environment, like that of the mitochondrial matrix (1), free Fe 2ϩ is rapidly oxidized to Fe 3ϩ with the subsequent formation of insoluble Fe(OH) 3 . This type of Fe 2ϩ oxidation generally produces oxygen radicals. In addition, through the Fenton reaction in which Fe 2ϩ reacts with hydrogen peroxide, highly toxic hydroxyl radicals are produced. Organisms have evolved mechanisms for the control of iron uptake, delivery, storage, and detoxification, including those for Fe 2ϩ handling within mitochondria. Frataxin, a major mitochondrial protein player in this process, has a central role in iron detoxification and iron delivery in heme and iron-sulfur cluster (ISC) 5 synthesis (2)(3)(4)(5) as well as in aconitase repair (6). Low levels of frataxin in humans are responsible for the progressive neurodegenerative disease Friedreich's ataxia, caused by trinucleotide repeat expansions in the first intron of the frataxin gene and the consequent gene silencing. Frataxin deficiency results in aberrations in cellular iron homeostasis, progressive accumulation of iron in mitochondria, high levels of oxidative stress, and deficiency in heme and ISC biosynthesis (4,(7)(8)(9).
Numerous studies focusing on human and Saccharomyces cerevisiae (Yfh1) frataxin (FXN) and their Escherichia coli ortholog, CyaY, showed the ability of this protein to bind different metal ions, among which are Fe 2ϩ and Co 2ϩ (10 -15). Metal ion binding has been linked to the oligomerization propensity of yeast and bacterial frataxin, which can form oligomeric complexes with 3-24 or even 48 subunits (10, 11, 16 -19). These oligomers and the oligomerization process have been studied using x-ray crystallography, electron microscopy (EM), and small angle x-ray scattering (SAXS) (18 -21). Furthermore, the different oligomeric forms have been suggested to be associated with the different functions of frataxin (22).
The iron-dependent oligomerization is directly linked to iron detoxification through the frataxin-catalyzed ferroxidation reaction, in which two Fe 2ϩ atoms are oxidized, whereas O 2 is reduced to O 2Ϫ (10,11,(23)(24)(25)(26). These events are followed by the formation of an insoluble ferrihydride iron core, similar in structure to the iron core of ferritin (21,27). Frataxin's functional role as metal ion chaperone and direct Fe 2ϩ donor to proteins have been substantiated in diverse experimental models (2, 6, 13, 28 -31). Both human and yeast frataxin have been shown to deliver iron to the ISC scaffold protein (yeast Isu1/ human ISCU) (29,32), interacting with the sulfur donor, a cysteine desulfurase (yeast Nfs1/human NFS1, stabilized by Isd11/ ISD11) during the synthesis of ISC cofactors. Frataxin has also been shown to interact with ferrochelatase and donate iron for heme synthesis (2,11,28). Ferrochelatase, the terminal enzyme of the heme biosynthesis pathway, catalyzes the insertion of Fe 2ϩ into protoporphyrin IX (33). Frataxin-mediated iron delivery to ferrochelatase was supported by the initial observations that Yfh1-bound Fe 2ϩ was not oxidized as readily as free Fe 2ϩ in solution and that the transfer of Fe 2ϩ from Yfh1 to ferrochelatase occurred even in presence of an excess of citrate, a physiological Fe 2ϩ chelator, which suggested that direct protein-protein contacts and metal ligand exchange should take place (11,24). In fact, Fe 2ϩ could remain bound to Yfh1 for periods long enough to allow its delivery to the appropriate Fe 2ϩ protein acceptors for either heme or ISC biosynthesis (11,28,29,32).
Eukaryotic ferrochelatase is a homodimeric protein peripherally associated with the matrix side of the inner mitochondrial membrane. Although the physiological substrate of ferrochelatase is Fe 2ϩ , the enzyme can also bind and catalyze the insertion of other divalent metal ions, such as Zn 2ϩ , Ni 2ϩ , Cu 2ϩ , and Co 2ϩ , into the porphyrin macrocycle (33). In 2006 it was proposed that due to its low metal ion specificity, ferrochelatase must rely on a metallochaperone (34). In early studies with Yfh1-deficient (⌬yfh1) S. cerevisiae cells, ferrochelatase was shown to catalyze the formation of zinc-protoporphyrin but not heme (28). These results demonstrated that, although catalytically competent, in the absence of frataxin ferrochelatase did not catalyze the insertion of Fe 2ϩ into protoporphyrin. Nanomolar-range values for the binding constant (K d ) of frataxin and ferrochelatase, as measured using various biophysical approaches and proteins from different organisms (13,28,35), indicated a high affinity interaction between these two proteins. Specifically, surface plasmon resonance (Biacore) yielded a K d of between 17 nM and 40 nM for the in vitro interaction between Yfh1 and yeast ferrochelatase (28), and similarly, isothermal titration calorimetry (ITC) returned a K d of 1.7 ϫ 10 Ϫ8 M for the binding affinity of iron-loaded, mature truncated form of FXN (covering residues 81-210 (FXN 81-210 )) to human ferrochelatase (35). Moreover, Bacillus subtilis frataxin homolog, Fra, has been recently shown to interact with B. subtilis ferrochelatase HemH and to supply intracellularly Fe 2ϩ to the enzyme for heme synthesis (36). Ferrochelatase-frataxin interaction was also corroborated with in vitro ferrochelatase activity assays in which oligomeric Yfh1 supported heme formation by providing iron to ferrochelatase at neutral pH in the presence of atmospheric oxygen and absence of reducing agents (11).
With the determination of the crystal structure of the 88 -208 variant of human frataxin (FXN 88 -208 ), Dhe-Paganon et al. (37) predicted that a conserved and predominantly hydrophobic region on the surface of the protein would mediate protein-protein interactions. Subsequently, Stemmler and coworkers (2, 13) by using nuclear magnetic resonance spectroscopy (NMR) identified frataxin helical plane as the possible ferrochelatase binding surface in the monomeric yeast Yfh1 and human FXN structures. However, the precise details of how Fe 2ϩ is delivered to, or made available to ferrochelatase without being oxidized to Fe 3ϩ remain to be defined. Zhang et al. (38), in teasing out the roles of S. cerevisiae mitochondrial carrier proteins Mrs3/Mrs4 and Yfh1, concluded that these proteins cooperate in supplying iron for heme synthesis; mrs3/4 carriers deliver iron into mitochondria, whereas Yfh1 makes iron bioavailable within mitochondria. A similar picture appears to be emerging from studies on mitochondrial iron metabolism in Drosophila (39,40) and mammalian cells (41,42).
Here we used presteady state kinetics to determine the rate of interaction between Yfh1 and ferrochelatase and the rate of ferrochelatase-catalyzed porphyrin metalation with either free Fe 2ϩ or Fe 2ϩ -bound Yfh1. The results support direct interaction between the proteins and are consistent with delivery of Fe 2ϩ from Yfh1 to ferrochelatase, suggesting that the enhanced ferrochelatase activity stems from the Yfh1-promoted Fe 2ϩ availability and direct supply to the enzyme. Single-particle reconstruction from negatively stained EM images was used to study the structure of the complex between ferrochelatase and frataxin. Protein-protein docking guided by cross-linking combined with mass-spectrometric analyses was used independently to create a quasi-atomic model of the complex between yeast ferrochelatase dimer and two Yfh1 trimers. This model was subsequently docked into the EM reconstruction and showed a good fit.
Study of Protoporphyrin IX Binding to Ferrochelatase-The dissociation constant for the ferrochelatase-protoporphyrin IX complex was determined by direct spectrophotometric titration based on the loss of porphyrin fluorescence upon binding to the enzyme. A RF-5301 PC Shimadzu spectrofluorimeter was utilized, and the excitation wavelength was set to 407 nm, whereas the emission was recorded at 635 nm. A buffer solution containing 20 mM MOPS, pH 7.0, 0.4 M NaCl, 0.2% Tween 80, and 0.5 M protoporphyrin IX was titrated incrementally with 93.2 M yeast ferrochelatase, which was in a similar buffer but without protoporphyrin IX. The total volume of ferrochelatase added was Ͻ5% of the total volume and was not accounted for in the computations. The titration was run three times, and for each set of data the change in fluorescence was plotted as a function of ferrochelatase concentration. The points did not suggest cooperative binding and were fit to a quadratic equation (Equation 1) to determine the dissociation constant (Ref.
where [ES] represents the concentration of the enzyme-substrate complex, which is plotted at the y axis as the change in fluorescence, max refers to the fitted maximal percentage of quenching, [E] 0 is the initial enzyme concentration plotted at the x axis, [S] 0 represents the initial protoporphyrin IX concentration, which was 0.5 M, K d is the fitted dissociation constant.
Determination of the Rates of Frataxin-Ferrochelatase Association and Ferrochelatase-catalyzed Porphyrin Metalation by Stopped-flow UV-visible Spectroscopy-The presteady state kinetics of the association of Yfh1 and yeast ferrochelatase was examined by measuring changes in the intrinsic protein fluorescence intensity. Yfh1 and ferrochelatase, in 20 mM MOPS, pH 7.0, containing 0.4 M NaCl and 0.2% (v/v) Tween 80, were maintained at 20°C in separate syringes before their mixing in the reaction chamber. The protein concentrations in the two syringes were 2-fold greater than the final concentrations in the reaction chamber (final concentrations are reported in Fig. 1  legend). The intrinsic protein fluorescence quenching was examined upon excitation at a wavelength of 280 nm. The emitted light was filtered using a 320-nm longpass filter placed over the photomultiplier detector. The time course data were fit to a single-exponential process (Equation 2), where F obs (t) is the observed fluorescence change (in arbitrary units) at time t, k obs is the observed first-order rate constant, A 1 is the pre-exponential factor, and A 0 is the offset.
The rate for the ferrochelatase-catalyzed insertion of Fe 2ϩ into protoporphyrin IX was monitored by following the decrease in protoporphyrin fluorescence intensity. The excitation wavelength was set to 407 nm, and a 520-nm longpass filter was used in the detection of the emitted light. The time course data were fit to Equation 2 as described above (the data were fit to a first-order exponential rather than a burst equation because the steady state rate was negligible or insignificant on this time scale, and the simpler equation facilitates a more precise determination of the burst rate, which is what the experiment was intended to measure). In addition, similar presteady state kinetic studies were performed in the presence and absence of EDTA, a strong metal ion chelator. All buffers were made with milliQ water, treated with the DIAION TM CR11 (Sigma) metal ion-chelating resin, and filtered (0.2 m) to remove metal ions.
Single-turnover Burst Kinetics of Ferrochelatase in the Absence and Presence of Frataxin-In all cases one of the two syringes contained 40 M purified recombinant yeast frataxin (Yfh1) and 30 M protoporphyrin IX. The dissociation constant for the yeast ferrochelatase-protoporphyrin complex was separately determined to be 0.29 Ϯ 0.03 nM, indicating that under these conditions all of the porphyrin is bound to the enzyme. The other syringe contained 400 M ferrous chloride and 0 -300 M Yfh1, as specifically indicated in the Fig. 1 legend. The buffer contained 20 mM MOPS, pH 7.0, 0.4 M NaCl, and 0.2% Tween 80. Before running the experiments, yeast ferrochelatase and Yfh1 were dialyzed into this buffer. The ferrochelatase-protoporphyrin syringe was prepared and loaded first, and Yfh1, when included, was added immediately after the ferrous chloride had been mixed into the buffer of the second syringe. Approximately 30 s elapsed between the time the ferrous chloride was added and the first trace was recorded.
Single-turnover Stopped-flow Experiments-All of the fluorescence stopped-flow experiments described in the following sections were conducted using a SF-2001 from KinTek Corp. The syringe and observation chambers were maintained at 20°C with an external water bath, which approximately corresponded to room temperature at the time of the experiments. The photomultiplier detector was mounted at a right angle to the path of excitatory light entering the 0.2-cm 2 observation cell. For experiments monitoring the depletion of protoporphyrin IX, the excitation wavelength was set to 407 nm, and the photomultiplier was equipped with a 520-nm longpass filter.
Transmission Electron Microscopy-For preparation of the complex of ferrochelatase with frataxin trimers, Yfh1 at a concentration of 3 mg/ml (ϳ22 M) was incubated with ferrous ammonium sulfate at 2-to-1 iron-to-Yfh1 molar ratio for 60 min at 30°C (19). This was followed by incubation of 1 part ferrochelatase (6 mg/ml) with 2 parts Yfh1 trimers (3 mg/ml) for 30 min at 30°C, yielding a final protein concentration of 4 mg/ml (2 mg/ml Yfh1 and 2 mg/ml ferrochelatase). Reported protein molar concentrations are based on monomeric molecular masses of 13,783 and 41,209 daltons for Yfh1 and S. cerevisiae ferrochelatase, respectively, as calculated from the primary amino acid sequences encoded by the cDNAs for the two recombinant proteins.
The Yfh-ferrochelatase complex was applied to a 10/300 GL Superdex 200 gel filtration chromatography column (GE Healthcare) at a flow-rate of 0.5 ml/min and eluted at room temperature with HN100 buffer (20 mM HEPES, pH 7.3, and 100 mM NaCl) in 0.1-ml fractions. The recording was done at 280 nm (A 280 ).
Samples from fractions containing both Yfh1 and ferrochelatase (10 l) were applied directly to a 400-mesh, carboncoated, copper grid (Electron Microscopy Sciences) that had been preincubated for 1 min in HN100 buffer. After 1 min, excess protein sample was blotted with filter paper, and the grid was washed with sterile water for 3 s. Excess water was again removed by blotting. For staining, 1% (w/v) uranyl acetate was applied to the grid, and after 30 s excess stain was blotted. The grid was left to dry for ϳ30 min, after which it was inserted into the sample holder of a Philips CM120 transmission electron microscope equipped with a GATAN GIF 100 energy filter and a GATAN 791 CCD camera (1024 ϫ 1024 pixels). Images were taken using 55,000ϫ magnification, and image processing was performed with the EMAN2 software package (51). Classification of particles was made using six averaging iterations. 1184 particles were used for the final three-dimensional reconstruction, and a C2 point group symmetry was applied to the model. Validation of the final reconstructions was accomplished by comparing class averages with projections of the three-dimensional reconstructions in the same angular orientation. The resolution of the reconstruction was estimated to 38 Å. Chimera software was used to visualize the reconstruction (52).
Chemical Cross-linking, Proteolysis, and MALDI Mass Spectrometry-The Yfh1-ferrochelatase complex was crosslinked with a 1:1 mixture of H12:D12 isotopically labeled bis-(sulfosuccinimidylsuberate) (BS 3 ), a primary amine-specific chemical and cleavable cross-linker. The Yfh1-ferrochelatase complex sample (20 l) was incubated with the BS 3 cross-linker in a 50:1 molar ratio of cross-linker:protein at 25°C for 15 min. The reaction was quenched by adding Tris to a final concentration of 20 mM. Subsequently, the protein was precipitated with freeze-cold acetone, centrifuged, and resuspended in 50 mM ammonium bicarbonate buffer, pH 7.8. Endoproteinase GluC was used to digest the cross-linked Yfh1-ferrochelatase complex for 17 h at 37°C. Because the Gluc endoproteinase, which cleaves the polypeptide C-terminal to glutamate residues, can also cleave C-terminal to aspartate residues, albeit at a 100 -300 times slower rate, the GluC reaction time was the same for all experiments to ensure reproducibility in the results.
The GluC-cleaved peptide mixture was separated by reversed phase liquid chromatography and collected into 192 fractions on a MALDI target plate. Subsequently, the fractions were analyzed on a 4700 Proteomics Analyzer (Applied Biosystems/MDS SCIEX) as previously described (53). All mass spectrometry (MS) spectra were internally calibrated using internal calibration peptides added to the matrix, except in fractions where the ion suppression decreased the intensity of the calibration peptides. Data from MS and tandem MS (MS/MS) were analyzed with the program FINDX (53). FINDX is an in-housedeveloped software designed for detection of cross-linked peptides using isotope-labeled cross-linker and MALDI mass spectrometry. In contrast to other programs designed for use with high resolution electrospray instruments, where cross-link detection is based on acquisition of a large amount of high quality MS/MS data, FINDX is designed to select cross-linked pep-tides only based on MS data. Thus, in the primary data analysis, which is based on well calibrated MS data, cross-linked peptides are identified at a tolerance setting of 7 ppm as doublet peaks with the mass difference of 12.07573 Da (due to the incorporation of deuterium from the H12:D12 isotopically labeled BS 3 ). In the subsequent step the data are validated by acquiring MS/MS-data for the peaks that correspond to the suggested cross-links. The MS/MS-spectra are inspected manually and analyzed by the second module of FINDX, which considers the following matches: 1) the mass of the fragments cross-linked to the peptide (referred to as x-fragment) and 2) masses that match stepwise fragmentation peptides in the putative crosslinked peptide (referred to as sequence fragments). At this stage potential cross-links are confirmed with more than three x-fragments in the list of matched masses, with both parts of the cross-linker represented and with more than three sequencefragments from each of the two cross-linked peptides. For this reason no statistics of MS/MS-data is provided by the FINDX report.
Protein-Protein Docking and Yfh1-Ferrochelatase Model Evaluation-Initial models (a total of 12,000) with randomized orientation of both docking partners in the complex between yeast frataxin Yfh1 trimer (PDB ID 3OEQ; Ref. 18) and yeast ferrochelatase (PDB ID 1LBQ; Ref. 20) were generated using the Rosetta 3.4 software package (54). The "low_res_protocol_ only" docking flag in Rosetta was used during this stage. The initial models were filtered using distance constraints established from the results of the cross-linking experiments and sorted according to the Rosetta energy score. The C␣ distance between two cross-linked lysines was set as Ͻ31.6 Å. This value was obtained after taking into account the length of the crosslinker spacer arm (11.4 Å) and the length of two lysine side chains (2 ϫ 6.5 Å). In addition, estimates of the resolution-dependent error in the positions of the protein atoms of Yfh1 and ferrochelatase (5.1 Å and 2.1 Å, respectively) needed to be added. These were estimated at 3 using Cruickshank's formula (55) rearranged by Blow (56). The single-particle reconstruction was converted from mrc format to a bead model in pdb format using the program EM2DAM from the ATSAS suite of programs (57). The program SUPCOMB (58) was then used to automatically fit the filtered, initial models from Rosetta to the bead model resulting from the EM reconstruction. All models that appeared to fit well to the reconstruction, as assessed by the scoring function of SUPCOMB and by manual inspection but in which the trimeric structures of Yfh1 would collide with the predicted position of the inner mitochondrial membrane, were removed. The remaining models (ϳ2000) were taken for further refinement in Rosetta using the "dock-ing_local_refine" as well as the standard "dock_pert 3 8" docking flags. The refined models were evaluated and sorted according to their interface score (the total score of the complex minus the total score of each partner in isolation, according to the Rosetta 3.4 manual).

Results
Transient Kinetic Analysis of Yfh1-mediated Fe 2ϩ Transfer to Ferrochelatase-As shown previously with small angle x-ray scattering and x-ray crystallography, at a Fe 2ϩ :Yfh1 ratio of 1:1 Yfh1 populates a mixture of trimeric and monomeric states (38%/61% monomer/trimer), whereas at a ratio of and 2:1 the oligomeric state is predominantly trimeric (19). To determine if ferrochelatase-catalyzed heme synthesis depends on Fe 2ϩ made available by Yfh1 and on the oligomeric state of Yfh1, single-turnover reactions were performed with Fe 2ϩ :Yfh1 ratios of 0, 1.3, and 2.0. The rate of the reaction was monitored by following the consumption of the protoporphyrin IX substrate (or porphyrin metalation). The results show that in the absence of Yfh1 (Fig. 1A), the rate of the reaction was clearly slower than the reaction in the presence of Yhf1 at Fe 2ϩ :Yfh1 ratios of 1.3 or 2 (Fig. 1, B and C) regardless of the "preincubation time" of Yfh1 with Fe 2ϩ (different colored kinetic traces in Fig. 1, B and C). Furthermore, in the absence of Yhf1 (Fig. 1A) the rate of heme formation decreased with increasing the preincubation time (yellow and blue circles compared with red and green on Fig. 1A), presumably due to the oxidation of the "free" Fe 2ϩ to Fe 3ϩ , which is not a substrate of ferrochelatase. Higher Fe 2ϩ :Yfh1 ratios (4, 7, and 10; data not shown), which result in higher order oligomers of frataxin, were also tested and showed dependence on incubation time similar to that of the Yfh1-free solution. This suggests that higher order frataxin oligomers, such as hexamers and dodecamers, etc., are less efficient in delivering iron to the ferrochelatase reaction. These results agree with our previous experiments performed under steady state conditions, which showed that at low Fe 2ϩ :Yfh1 ratios, Fe 2ϩ can be readily mobilized by chelators or made available to ferrochelatase to synthesize heme (11).
Rate of Yfh1-Ferrochelatase Association in Relation to the Rate of Heme Synthesis-To ascertain if the rate of complex formation between ferrochelatase and frataxin is consistent with direct transfer of Fe 2ϩ from Yfh1 to ferrochelatase, we determined and compared the rates of association of Yfh1 with ferrochelatase and ferrochelatase-catalyzed Fe 2ϩ insertion into protoporphyrin IX. The decrease in intrinsic protein fluorescence was used to follow the association of Yfh1 with ferrochelatase (green circles in Fig. 2A). The rate of complex formation was determined using transient (fluorescence stopped-flow spectroscopy) kinetics, which gave the value of 18 Ϯ 1 s Ϫ1 . The rate of the ferrochelatase-catalyzed metalation reaction was monitored by following the consumption of the protoporphyrin IX substrate (red circles, Fig. 2A), was 14.9 Ϯ 0.1 s Ϫ1 . The results show that the two proteins bind at a rate slightly faster than the rate of protoporphyrin IX metalation, in agreement with direct delivery of Fe 2ϩ from Yfh1 to ferrochelatase.
We also performed transient kinetic studies of the ferrochelatase-catalyzed heme synthesis reaction using either free Fe 2ϩ or Fe 2ϩ -loaded Yfh1 in the presence and absence of EDTA. The addition of EDTA completely abolished heme production in the absence as well as in the presence of Yfh1, which was preincubated with Fe 2ϩ at a Fe 2ϩ -to-Yfh1 molar ratio of 2.5-to-1 (Fig. 2B). Although the reaction mechanism of Fe 2ϩ dissociation from Yfh1 and Fe 2ϩ association with and chelation by EDTA (e.g. formation of a ternary Fe(II)-Yfh1-ferrochelatase complex, scavenging of released Fe(II) from Yfh1 by EDTA yielding the EDTA-Fe(III) complex, or Fe(II) transfer between Yfh and ferrochelatase by a dissociative or an associative mechanism) is unknown, our findings indicates that EDTA, a strong chelator and catalyst of Fe(II) autooxidation, presumably "removes" Fe(II) from Yfh1 (59), rendering it in a form incompatible with its ferrochelatase-catalyzed insertion into protoporphyrin IX. These results are consistent with the observations showing a relatively low metal binding affinity by frataxin with dissociation constants in the micromolar range (2-3 M; Refs. 13 and 14). Furthermore, the rate of heme synthesis was slightly accelerated when Fe 2ϩ was provided from "Fe 2ϩloaded Yfh1," in agreement with the chaperone role for Yfh1. It is important to note that protoporphyrin IX binds to ferrochelatase very tightly with a K d value of 0.34 Ϯ 0.04 M under the conditions of these stopped-flow experiments (data not shown). This means that there is virtually no free protoporphy-  Yfh1 (B and C). The Fe 2ϩ :Yfh1 molar ratios were 2.0 (B) and 1.3 (C). In all cases, the red, green, yellow, and blue circles represent the kinetic traces at 0, 2, 5, and 10 min, respectively, after the initial shot (i.e. filling the stopped-flow syringes with the reactants). MAY 27, 2016 • VOLUME 291 • NUMBER 22

JOURNAL OF BIOLOGICAL CHEMISTRY 11891
rin IX present in solution, and it is, therefore, truly a singleturnover reaction (Fig. 2B).
EM Single-particle Reconstruction-Based on earlier gel-filtration studies, 6 which showed the formation of a complex between Yfh1 trimer and ferrochelatase, on the kinetic data obtained here on frataxin-ferrochelatase association and ferrochelatase-catalyzed porphyrin metalation and to ensure a homogenous complex, for the EM reconstruction it was decided to focus on the Fe 2ϩ :Yfh1 ratio of 2:1. As mentioned above, at this ratio frataxin is expected to be primarily in the trimeric state. After formation (see "Experimental Procedures") the complex was applied to a gel filtration chromatography column. Sample analysis revealed three visible peaks corresponding to fractions containing proteins of approximate molecular mass of 240, 160, and 50 kDa (Fig. 3A). A small amount of sample from the fractions corresponding to the 160-and 240-kDa peaks was analyzed by SDS-PAGE (Fig. 3B) and protein staining. The analysis showed that the peaks contained both frataxin and ferrochelatase (Fig. 3B). Examination of negatively stained EM images of the fractions corresponding to the two peaks showed that the 240-kDa peak images contained mostly aggregated and irregular material, whereas the 160-kDa peak images together with other material contained repeating particles with recognizable shape (Fig. 4A). Furthermore, the apparent molecular mass of 160 kDa is very close to the molecular mass of a complex of 2 frataxin trimers with a ferrochelatase dimer (172 kDa). Thus, it was decided to use the images of the 160-kDa fractions for further analysis by single-particle reconstruction, as described under "Experimental Procedures." The complex was reconstructed using 1184 particles. Class averages of the collected particles and the final reconstructed volume are shown on Fig. 4B and Fig. 4, C and D, respectively. It should be noted that the complex was unstable under the conditions used to set up the grids, and presumably a large fraction of the pro-   tein dissociated and aggregated after the addition of uranyl acetate and during grid preparation. Nevertheless, the number of collected particles allowed for the overall shape of the complex to be reconstructed. Using the EM volume and a combination of cross-linking, modeling, and docking of the x-ray structures of the frataxin trimer and yeast ferrochelatase dimer (see below), the reconstruction of the complex could be interpreted with one ferrochelatase dimer positioned in the center of the density flanked by two Yfh1 trimers bound on the sides, one trimer to each ferrochelatase subunit (Fig. 5, A and B, details presented below). The total buried surface area between the molecules was estimated to ϳ700 Å 2 , which is within the range expected for transient molecular complexes.
Primary Amine-specific Cross-linking of the Complex between Yfh1 and Ferrochelatase-When the resolution of single-particle reconstructions from negatively stained EM images is not sufficient to unambiguously position high resolution structural models (from e.g. x-ray crystallography) into the reconstructed volume, distance constraints can be used to aid interpretation of the EM model. Here, we used primary amine-specific crosslinking of the Yfh1-ferrochelatase complex and, upon proteolysis with endoproteinase GluC, analyzed the cross-linked peptides by LC-MS and MS/MS. The established distance constraints were subsequently applied to filter the protein-protein docking simulations (see below).
Of several hundred masses, the FINDX software (53) Fig. 2, B and C. All structure figures were prepared using PyMOL (70). peaks consistent with cross-linking between Yfh1 and ferrochelatase (supplemental Table S1). These four identified crosslinked Yfh1-ferrochelatase peptides were further analyzed by MS/MS, and two of them (with masses at 1565.77 Da and 1621.90 Da in MS) were verified by peptide sequencing supplemental Table S2) as described under "Experimental Procedures." From the determined cross-linked peptide sequences 72 KYHEE 76 and 168 KAISKSQ 174 of Yfh1 cross-linked to 93 KQYRE 97 of ferrochelatase and with the theoretical masses of 1565.76 Da and 1621.89 Da, respectively, we found that either Lys-72 (from the flexible N terminus) or Lys-168/-172 (from the C-terminal helix) of Yfh1 were cross-linked to Lys-93 of ferrochelatase (Fig. 4A). Thus, and given the lengths of the BS 3 spacer arm and two lysine side chains (see "Experimental Procedures"), the C␣ atom distances between Yfh1 Lys-72 and ferrochelatase Lys-93 and between Yfh1 Lys-168 (or Yfh1 K172) and ferrochelatase Lys-93 should be Ͻ31.6 Å.
Protein-Protein Docking Using Rosetta, EM, and Crosslinking-The initial model for the complex between ferrochelatase and frataxin was built using protein-protein docking and distance constraints acquired from the cross-linking studies without taking into account the EM reconstruction volume. Using the protein-protein docking program within the Rosetta package, 12,000 random complex models were created with the low resolution docking flag. Of these, 1269 fulfilled the distance constraints acquired from the cross-linking studies. These models were sorted according to the Rosetta energy score, and the top models were compared against the reconstructed EM volume. The two best-fit final models (model 1 and model 2) were chosen such that steric clashes between the frataxin subunits and the expected position of the mitochondrial membrane would be avoided. Subsequent fitting of the models to the EM reconstruction showed that model 1 could fit the EM reconstruction slightly better than model 2 (scores 3.738 and 3.820, respectively), whereas model 2 had a slightly lower Rosetta energy score (the lower the energy the better the model) compared with model 1 (Ϫ95.361 and Ϫ103.335 for models 1 and 2, respectively). Toward the final refinement we decided to produce 2000 refined models for each of models 1 and 2. The refined protein-protein docking resulted in complexes with similar fits to the reconstructed EM volume (scores 1.439 and 1.435 for models 1 and 2, respectively). However, the interface score (the total score of the complex minus the total score of each partner in isolation, according to the Rosetta 3.4 manual) was significantly better for refined model 1 compared with model 2 (4.27 and 2.37, respectively). Thus, model 1 was selected for the final evaluation of the complex.
Evaluation of the Model of Yfh1-Ferrochelatase Complex- Fig. 5A shows the refined model 1 of the complex between Yfh1 trimers and the ferrochelatase dimer superimposed on the EM reconstruction. In this model the relative orientation of the ferrochelatase dimer and the two frataxin trimers is such that the hydrophobic surface of ferrochelatase (facing the reader in Fig. 5B) faces the presumed membrane surface, where it is expected to receive the substrate, the highly hydrophobic protoporphyrin IX. This mode of interaction supports the physiological relevance of the model. Two monomers of the frataxin trimer appear to be positioned for interactions with a ferrochelatase monomer, whereas the third frataxin monomer does not contribute directly to the interactions (Fig. 6A). Several conserved ferrochelatase and FIGURE 6. Structural details of the complex between yeast ferrochelatase and frataxin Yfh1. A, the cross-links between ferrochelatase (green ribbon model) and frataxin (orange ribbon model), which were analyzed in the MS experiments, are shown as dotted lines, and the corresponding amino acid side chains are shown in sticks representation. Only one ferrochelatase monomer is shown looking down into the porphyrin binding pocket. The cross-link distances between the amino acid residues involved in the cross-linking are shown in the insert. Only distance constrains obtained in the cross-linking experiments where used in the choice of the best model for docking into the EM density and for further analysis. Regions of frataxin helices ␣1 and ␣2, shown in NMR experiments (in which monomeric frataxin was used) to be involved in interactions with ferrochelatase, are colored in magenta. For clarity only two monomers of the frataxin trimer are shown on this figure. The third monomer points away from ferrochelatase and is not involved in complex formation. B, the side chains of frataxin and ferrochelatase amino acid residues involved in potential interactions within the complex are shown in sticks representation. A dotted line is drawn from the ferroxidation site of frataxin to Asp-320 of the -helix of ferrochelatase, and the distance between the two sites is shown. Both regions are colored in violet for clarity. From the model it appears that although the upper frataxin monomer is involved in the stabilization of the complex, the lower monomer would deliver the metal to ferrochelatase. frataxin residues are located within the interaction surface area between Yfh1 and ferrochelatase (Fig. 6B). These include ferrochelatase residues from a loop between helices 2 and 3 (residues 98 -103) and a conserved tryptophan (Trp-107). Residue Glu-97 together with His-317, which belongs to the -helix of ferrochelatase (residues 313-325), have been shown earlier to bind Cd 2ϩ in the yeast ferrochelatase structure (20). Conserved frataxin regions within the buried interface include residues from the end of the acidic ␣-helix (residues Glu-90 and Leu-91) as well as from the end of the second ␣-helix (Asn-154, Asp-160, and Glu-167). In an NMR spectroscopy study, which was limited to the use of frataxin monomers, most of the abovementioned residues where shown to be affected by ferrochelatase-frataxin interactions (13). The present model also suggests potential hydrogen bonding interactions and salt bridges between Yfh1 and ferrochelatase. These include interactions between frataxin-ferrochelatase residues Asn-146 -Tyr-110, Glu-164 -Lys-106, and Lys-174 -Glu-53 (Fig. 6A).
In our docked reconstruction of the complex between the frataxin trimers and the ferrochelatase dimer, the second frataxin monomer is positioned at a longer distance from the ferrochelatase monomer, too far for hydrogen bonding interactions to be formed. The residue His-83 from the ferroxidation site of Yfh1 is positioned at ϳ15 Å from the conserved Asp-320 of ferrochelatase (Fig. 4B); however, taking into account the low resolution of the model, this distance might be shorter or longer. Asp-320, like the above-mentioned His-317, is located in the -helix, which connects the surface of the protein to the porphyrin binding cleft and the substrate metal ion binding site of ferrochelatase. The conserved acidic residues, aligned along the axis of the -helix, have been proposed to constitute a route for Fe 2ϩ transfer from the surface to the active site of the enzyme, where the metal ion is inserted into protoporphyrin IX (60). Thus, it is plausible that although one of the frataxin monomers is involved in extensive complex-stabilizing interactions, the second monomer is positioned for Fe 2ϩ delivery to ferrochelatase.
It should be noted that the focus on the complex of frataxin trimer with ferrochelatase does not exclude the presence of other types of complexes in the images, for example complexes of ferrochelatase with monomers and dimers of frataxin. However, these would be difficult to clearly separate on the images for the sake of a separate single-particle reconstruction.

Discussion
In this work we present a structural model of the complex between yeast frataxin trimers and yeast ferrochelatase and show that the rate of heme production by ferrochelatase is dependent on the type of Yfh1 oligomers present in solution. In particular, when compared with free Fe 2ϩ , the rate of heme production is higher in the presence of monomers and trimer and is reduced in the presence of larger Yfh1 oligomers, like hexamers and 24-mers. Earlier, Park et al. (11) showed that Yfh1 could bind large quantities of iron, keeping it available for ferrochelatase during long time periods. The authors also showed that iron sequestered inside larger Yfh1 oligomers, like 24-mers, was less accessible to the ferrochelatase reaction. It should be noted that although iron:frataxin ratios higher than 2:1 trigger the formation of oligomers of order higher than trimers, they also reduce the percentage of available monomers and trimers (19), which as shown here are the most favorable forms of frataxin for iron delivery to ferrochelatase.
It is of interest to compare the mechanisms of the chaperone function of frataxin with those of the better-studied copper chaperone system. Similar to iron, copper supports redox chemistry in cells and, if unrestrained, can be highly toxic for organisms (61). The best-studied copper chaperone Atx1 appears to protect Cu(I) from nonspecific reactions, allowing rapid metal transfer to its partners (62). This process is believed to proceed through substitution of metal ligands on the donor chaperone by new ligands from the acceptor target protein after the formation of a complex between the proteins (61,63). Weak, transient protein-protein interactions have been shown to form the basis for complex formation (64). Moreover, it has been shown that copper ion is necessary for the formation of the complex between Atx1 and Ccc2 ATPase and that no interaction occurs between the apo-forms of the proteins (63).
The structure of the complex of ferrochelatase and frataxin presented here suggests that both proteins are engaged in close interactions with each other, creating a route for Fe 2ϩ transfer from the metal binding site of frataxin to the site of the ferrochelatase reaction. This route appears to involve the -helix of ferrochelatase, a conserved structural element that contains several acidic residues aligned along the helix axis. Only a -helix may provide such alignment of amino acid side chains along the helix axis (65).
The results also support earlier work in which different experimental methods used to study the interactions between ferrochelatase and frataxin clearly pointed to the existence of direct contacts between the two proteins (11,13,28,35,38,47), although the precise oligomeric state of frataxin required for the most favorable interaction was not discussed. In the presented model of the complex between yeast ferrochelatase and Yfh1, His-83 of the ferroxidation site of frataxin, believed to be one of the iron binding residues, is around 15 Å away from Asp-320 of the -helix of ferrochelatase (Fig. 4B). This distance could be too long for direct ligand exchange to take place, but taken into account the low resolution of the structure and the transient character of the complex, closer interactions between the two sites, which would include frataxin His-83, and ferrochelatase Asp-320 and possibly His-317 (previously shown to bind cadmium) are feasible in solution. Although in solution, Fe 2ϩ is normally coordinated by six water molecules, Fe 2ϩ at the di-iron ferroxidation site of frataxin, similarly to ferritin, may be coordinated to a histidine and acidic amino acid residues as well as to water (66,67). The water molecule could take part in ligand exchange by being initially replaced by one of the acidic residues of the -helix of ferrochelatase. Partial dehydration of the iron before arriving at the ferrochelatase active center histidine and glutamate residues would facilitate the reaction of iron insertion into protoporphyrin IX (68).
Under experimental conditions favoring the formation of Yfh1 trimers, the rate of complex formation between Yfh1 and ferrochelatase is slightly faster than that of heme formation (18 Ϯ 1 s Ϫ1 versus 14.9 Ϯ 0.1 s Ϫ1 ). This is consistent with the idea that a step involving complex formation takes place before porphyrin metalation. Furthermore, Fe 2ϩ needs to be loosely bound to Yfh1 for a delivery process to be efficient. The fact that EDTA inhibits porphyrin metalation in the presence of Yfh1 indicates that EDTA binds the metal ion much more tightly than the protein, which has instead evolved to chaperone the metal ion. In prior steady state kinetic experiments we postulated that if Yfh1-bound Fe 2ϩ is not transferred to a ligand (e.g. ferrochelatase), its oxidation and mineralization proceed to completion, with Fe 3ϩ becoming progressively less accessible, being sequestered within the iron core of frataxin (11). Although other Fe 2ϩ chaperones cannot be ruled out as donors of the metal substrate to ferrochelatase, frataxin appears to meet the structural and functional requirements for such a role. Moreover, recently, Mielcarek et al. (36) reported that the physical interaction between the B. subtilis homologs of frataxin and ferrochelatase (hemH) mediated Fe 2ϩ supply and sustained heme biosynthesis in vivo. A similar scenario is very likely in yeast and other eukaryotes. Conceivably, in the erythroblast, the inner mitochondrial transporter Mfrn1 (69), frataxin, and ferrochelatase may form a dynamic and transient complex, and at different times the complex could also include other protein partners. According to this model, Mfrn1 may import Fe 2ϩ across the inner mitochondrial membrane and frataxin chaperones and delivers Fe 2ϩ to ferrochelatase. The model corroborates the cooperative roles observed for the yeast Mfrn ortholog mrs3/4 and yeast frataxin Yfh1 (38) and the need for the proposed bridging proteins stabilizing the immunoprecipitable Mfrn1-ferrochelatase complex (47). In sum, a stable multicomponent complex with at least the ferrochelatase dimer and two frataxin trimers as protein partners would serve two important purposes: 1) to enable the protected delivery of potentially toxic Fe 2ϩ to ferrochelatase and 2) to enable the specificity of ferrochelatase for Fe 2ϩ over other divalent metal ions and thus heme synthesis under physiological conditions. Author Contributions-S. A.-K., G. C. F., and G. I. conceived and coordinated the study, contributed to the analysis of the data, and wrote the paper. C. S. designed, performed, and analyzed the EM study, docking, cross-linking, and mass spectrometric analysis. M. E. G. designed, performed, and analyzed the kinetic experiments shown in Figs. 1 and 2. E.-C. A. provided technical assistance in sample preparation for EM, EM data collection, and processing. G. A. H. and O. G. provided technical assistance and contributed to the preparation of the manuscript. All authors reviewed the results and approved the final version of the manuscript.