Structure of C42D Azotobacter vinelandii FdI

All naturally occurring ferredoxins that have Cys-X-X-Asp-X-X-Cys motifs contain [4Fe-4S]2+/+ clusters that can be easily and reversibly converted to [3Fe-4S]+/0 clusters. In contrast, ferredoxins with unmodified Cys-X-X-Cys-X-X-Cys motifs assemble [4Fe-4S]2+/+ clusters that cannot be easily interconverted with [3Fe-4S]+/0 clusters. In this study we changed the central cysteine of the Cys39-X-X-Cys42-X-X-Cys45of Azotobacter vinelandii FdI, which coordinates its [4Fe-4S]2+/+ cluster, into an aspartate. UV-visible, EPR, and CD spectroscopies, metal analysis, and x-ray crystallography show that, like native FdI, aerobically purified C42D FdI is a seven-iron protein retaining its [4Fe-4S]2+/+ cluster with monodentate aspartate ligation to one iron. Unlike known clusters of this type the reduced [4Fe-4S]+ cluster of C42D FdI exhibits only an S = 1/2 EPR with no higher spin signals detected. The cluster shows only a minor change in reduction potential relative to the native protein. All attempts to convert the cluster to a 3Fe cluster using conventional methods of oxygen or ferricyanide oxidation or thiol exchange were not successful. The cluster conversion was ultimately accomplished using a new electrochemical method. Hydrophobic and electrostatic interaction and the lack of Gly residues adjacent to the Asp ligand explain the remarkable stability of this cluster.

A fundamental question in [Fe-S] protein biochemistry concerns how cysteine ligand and neighboring residue organization determines [Fe-S] cluster type, and whether or not one type can be converted to another (for reviews, see Refs. [1][2][3][4][5][6][7][8][9]. These issues are important, in part, because there are physiological situations where 3Fe to 4Fe or 4Fe to 2Fe cluster interconversion reactions modulate the activity of an enzyme or a regulatory protein (1, 8 -13). In addition, understanding these reactions is of importance in attempts to create clusters with new reactivities (14), to the de novo design of [Fe-S]-containing proteins (14,15), and to the study of [Fe-S] cluster assembly (16).
This study focuses on the simplest of these reactions, the interconversion of [3Fe-4S] ϩ/0 and [4Fe-4S] 2ϩ/ϩ clusters. As shown in Fig. 1, these two cluster types differ only by the presence or absence of a single Fe atom at one corner of the cube. The well characterized seven-iron ferredoxin from Azotobacter vinelandii, which contains one [3Fe-4S] ϩ/0 and one [4Fe-4S] 2ϩ/ϩ cluster, is proving to be an excellent model system for elucidating the properties of these redox centers. As shown in Fig. 2, the amino acid sequence Cys-X-X-Cys-X-X-Cys, which provides three of the four cysteine ligands to a [4Fe-4S] 2ϩ/ϩ cluster, is a very common motif. Interestingly, with one exception, all naturally occurring ferredoxins or ferredoxin variants that contain 4Fe clusters, which can be easily interconverted with 3Fe clusters have instead a Cys-X-X-Asp-X-X-Cys motif. The Asp serves as the ligand that is lost during the cluster interconversion process (23,(25)(26)(27)(28)(29). The exception is DgFdII where the central Cys is covalently modified in the [3Fe-4S] cluster-containing state and where the interconversion reaction is complicated by a change in subunit composition (10,30). Aconitase, which also easily interconverts, also has a non-Cys (water or OH Ϫ ) ligand (31,32). In contrast, for ferredoxins or variants that contain [4Fe-4S] 2ϩ/ϩ clusters with four Cys ligands, attempts to convert the 4Fe center to a 3Fe center more often produce only poor yields or result in complete degradation accompanied by denaturation (33)(34)(35)(36)(37)(38). This is certainly the case for the protein of interest in this study, AvFdI, where the oxidative destruction of its [4Fe-4S] 2ϩ/ϩ cluster has been studied in some detail (38 -40). In this study, we changed the central cysteine of the Cys 39 -X-X-Cys 42 -X-X-Cys 45 motif, which coordinates the [4Fe-4S] 2ϩ/ϩ cluster, into an aspartate. This paper describes the redox and spectroscopic properties of the new [4Fe-4S] 2ϩ/ϩ cluster, and our various attempts to convert it into [3Fe-4S] ϩ/0 . It also is the first report of the x-ray structure of an aspartate-liganded [4Fe-4S] 2ϩ/ϩ cluster.

EXPERIMENTAL PROCEDURES
Mutagenesis and Expression of fdx-The oligonucleotide used for the mutagenesis has the sequence 5Ј-TGCATCGACGACGCGCTCTG-3Ј, which differs from the wild-type sequence by the substitution of TGC (encoding Cys) for GAC (encoding Asp). The oligonucleotide-directed mutagenesis procedure was similar to that previously described (41,42) except that a pBluescript SK Ϫ phagemid was used as the mutagenesis vector instead of M13mp18. The mutation was confirmed at the DNA level by DNA sequencing. A 1.9-kilobase pair ClaI-XhoI fragment of A. vinelandii carrying wild-type fdxA gene was inserted into pBluescript SK Ϫ . The fragment containing the mutation was subcloned into pKT230 as described elsewhere (43), and electroporation was used to introduce the plasmid into the A. vinelandii strain LM100, which contains a disruption of the fdxA gene by a kanamycin resistance cartridge.
Protein Purification-All A. vinelandii cells were grown under N 2fixing conditions in a 200-liter New Brunswick fermentor (42). The protein purification was carried out at room temperature, and fractions were monitored by their absorbance at 405 nm and by cross-reactivity with antibodies raised against purified FdI, FdIII, or FdIV. The purification was carried out aerobically. Cell-free extracts were prepared from 1 kg of cells, and the first DEAE-cellulose column was run as described previously for the purification of other FdI variants (43). From the first DEAE-cellulose column, the FdI fraction eluted at 70 -80% of the linear 0.1-0.5 M NaCl gradient (total volume ϭ 3 liters) as a very shallow, yet well resolved brown peak. The protein fraction was immediately diluted with three volumes of 0.1 M potassium phosphate buffer, pH 7.4, and loaded onto a 2.5 ϫ 20-cm DEAE-cellulose column that was preequilibrated with the same buffer. Following loading, the column was washed at the elution rate of 3.0 ml/min for 24 h with 0.12 M KCl in the same buffer until the greenish-brown FdIII (34) eluted. The brownish FdI with the red-colored FdIV (44), which was visible on the column immediately following FdIII, was then eluted with 0.3 M KCl. Ammonium sulfate was added slowly to 75% saturation with stirring. After a 30-min incubation, the sample was centrifuged at 20,000 ϫ g for 20 min. The pellet was then collected and resuspended in 0.025 M Tris-HCl, pH 7.4, before loading onto a 1.5 ϫ 110-cm Superdex G-75 (Amersham Pharmacia Biotech) column preequilibrated with 0.025 M Tris-HCl, pH 7.4, 0.1 M NaCl. The brown-colored C42D FdI protein and red-colored FdIV eluted as a single broad peak as observed by monitoring at 405 nm; however, the FdIV actually moved slightly faster than C42D FdI. For that reason, the last third of the peak was collected and concentrated using a Centriplus-10 concentrator (Amicon) prior to further purification on a 1-ml Mono-Q FPLC system (Amersham Pharmacia Biotech) with a flow rate of 1 ml/min and a linear gradient of 0.3-0.8 M NaCl in 0.05 M Tris-HCl, pH 8.0, over 20 ml. Two fractions were well resolved: the red-colored fraction (FdIV), which eluted between 0.46 and 0.48 M NaCl, and a brown-colored fraction (FdI), which eluted between 0.54 and 0.58 M NaCl. Each fraction was collected, buffer-exchanged with 0.025 M Tris-HCl, pH 7.4, and concentrated to 0.5 ml. The C42D protein fraction was further purified with triclinic crystallization. The yield for C42D FdI protein is ϳ 0.5-1 mg/kg of cell paste.
Protein Characterization-To determine iron content, samples were digested, and the analysis was carried out as described previously (45) using FeCl 3 ⅐6H 2 O to generate a standard curve with FdI and FdIII (34) as controls. EPR 1 spectra were obtained using a Bruker ESP300E spectrometer, interfaced with an Oxford liquid helium cryostat. For spin quantitation of [3Fe-4S] ϩ , the spectrum was recorded at 10 K using 100 M purified C42D FdI at microwave power of 1 mW. Under those conditions, the EPR signal of FdI is linearly proportional to (microwave power) 1/2 . 100 M Cu 2ϩ -EDTA, which was used as a standard, was recorded at the same EPR setting and temperature, where the signal is also linear with respect to (power) 1/2 . The spin concentration was determined by double integration of the signal over the entire field sweep (46). For anaerobic EPR measurement, the samples were loaded into the EPR tubes in an O 2 -free (O 2 Ͻ 1 ppm) glove box (Vacuum Atmosphere, Hawthorne, CA). The tubes were sealed with a cap, wrapped with parafilm, and then transferred to the liquid nitrogen Dewar flask. Reduction of FdI was carried out by mixing degassed FdI, 5-deazariboflavin and EDTA (final concentration, 0.2, 0.15, and 10 mM, respectively) in 0.05 M Tris-HCl, pH 8.0, and then illuminating the mixture for 1 min using white light from a fiber optics illuminator, model 20 (150 W, Edmond Scientific Co., Barringston, NJ) (34). Using this system, the extent of reduction of the [4Fe-4S] 2ϩ cluster of FdI depends upon the light intensity, light illumination time, media anaerobicity, and absence of other electron acceptors. Oxidation with K 3 Fe(CN) 6 was carried out anaerobically in 0.05 M potassium phosphate buffer, pH 7.5; FdI was incubated at room temperature with K 3 Fe(CN) 6 at a molar ratio of 1:3 for 2 h prior to freezing (39,40). For NMR, samples were exchanged into a 0.05 M Tris-buffered D 2 O solution, pH 7.8, by passage through a Sephadex G-25 column (1.5 ϫ 15 cm). The FdI fraction was then concentrated to ϳ0.6 -1.0 mM by microfiltration with a Centriprep-10 (Amicon, Beverly, MA). 1 H 500-MHz NMR spectra were recorded at room temperature using 90°pulses of 8 s with 0.5-s delay times, using a Bruker GN-500 Fourier transform NMR spectrometer at the NMR facility, University of California, Irvine, CA. About 6000 scans were accumulated for each spectrum over a 2.5-h period. UV-visible absorption spectra were each recorded in a 0.5-ml quartz cuvette on a Hewlett Packard 8452A diode array spectrophotometer. CD spectra were obtained using a Jasco J-500C spectropolarimeter, with the sample contained in a small volume cylindrical cell with fused quartz window.
Electrochemistry-Purified water of resistivity ϳ 18 megohm⅐cm Ϫ1 (Millipore, Bedford, MA) was used in all experiments. The buffers MES, HEPES, and TAPS and the co-adsorbates neomycin or polymyxin (sulfate salts) were purchased from Sigma. An AutoLab electrochemical analyzer (Eco Chemie, Utrecht, The Netherlands) was used to record DC cyclic voltammograms. All experiments were carried out under anaerobic conditions in a Vacuum Atmospheres glove box with an inert atmosphere of N 2 (Ͻ1 ppm). The three-electrode configuration featuring all-glass cells has been described previously (47); this featured a jacketed main cell compartment (typically holding 500 l), which could be maintained at 0°C to optimize protein film stability. The saturated calomel reference electrode was held in a Luggin side arm. Prior to each experiment, the pyrolytic graphite "edge" electrode (surface area typically 0.18 cm 2 ) was polished with an aqueous alumina slurry (Buehler Micropolish; 1.0 m) and sonicated extensively to remove traces of Al 2 O 3 . Reduction potentials (E 0Ј ) from cyclic voltammetry were determined from the average of the anodic and the cathodic peak potentials, E 0Ј ϭ 1 ⁄2(E pa ϩ E pc ) and adjusted to the standard hydrogen electrode scale using E SCE Ϫ E SHE ϩ 243 mV at 22°C (where SCE is saturated calomel reference electrode and SHE is standard hydrogen electrode).
Bulk electrochemistry solutions contained between 0.05 and 0. ground current by subtracting a polynomial base line (48). Pulsing experiments were carried out as follows. Films were prepared and then scanned over the normal potential range until the film had stabilized (usually three scans). This voltammogram was then recorded. The protein film was then subjected to an oxidative potential pulse of 0.643 V for 1-4 s. A scan was then taken immediately afterward in the normal potential region, in order to observe the changes that had occurred.
Crystallization-Crystals of C42DFdI were grown by the standard methods as described previously (49,50). However, the protein appeared to denature quickly in drops that contained higher concentrations of ammonium sulfate. Touchseeding and attempts to macroseed did not work. Macroseeds quickly denatured when placed in preequilibrated drops. A few crystals grew spontaneously after several months' incubation at 4°C and appeared to be small but regular. These crystals grew in only two of all the drops seeded.
Attempts to cryoprotect these crystals did not work. The crystals quickly melted when placed in ammonium sulfate saturated drop buffer supplemented with 20% or 25% glycerol. As a result, after several failed attempts to mount a crystal, the crystal used for data collection was scooped directly out of the drop conditions and frozen in the N 2 cryostream without cryoprotectant.
Structure Determination-The structure of native FdI was modeled with a glycine at position 42, and refined by rigid body and positional refinement against all of the data to 2.3-Å resolution (Table I) using Xplor version 3.8 (51). The unbiased 2͉F o  Ϫ ͉F c  and ͉F o  Ϫ ͉F c  electron density maps clearly revealed the position of the Asp 42 side chain with very little other perturbation in the structure. The Asp 42 side chain was modeled into the density with Xfit/Xtalview (52), and the structure was refined with isotropic B-factors and a bulk solvent correction using Xplor version 3.8. The Asp 42 side chain was restrained to form a 2.1-Å bond to Fe of the [4Fe-4S] cluster, as indicated by the density. The refined model was checked and adjusted against a 2.

RESULTS AND DISCUSSION
Cell Growth and Protein Purification-A. vinelandii is a nitrogen-fixing soil bacterium that appears to synthesize at least 12 different ferredoxins (44). FdI is the only seven-iron ferredoxin that has been identified in that organism. Previous studies have established that FdI has a regulatory function as part of an oxidative stress response system in A. vinelandii (53)(54)(55)(56)(57)(58) and a metabolic function unrelated to nitrogen fixation that is important for cell growth (59). Mutagenesis experiments have shown that even subtle alterations in the environment of the [3Fe-4S] ϩ/0 cluster have profound negative effects on the growth rate of the organism (42). In contrast, much more drastic alterations in the environment of the [4Fe-4S] 2ϩ/ϩ cluster have no effect on cell growth (60). In keeping with these obser-vations, the A. vinelandii strain expressing C42D FdI and the strain expressing native FdI exhibit identical lag phases and growth rates.
Previous studies have also shown that mutations involving FdI [Fe-S] cluster ligands always lead to proteins that accumulate to much lower levels in vivo than does the native protein (41,61). This is also true for C42D FdI. Despite the low levels of accumulation of the C42D FdI in vivo, however, once the protein is purified it appears to be as stable as native FdI. What is particularly important for this study is the fact that the protein is completely stable toward oxygen.
Like Native FdI, C42D FdI Is a Seven-iron Protein-Native AvFdI contains one [4Fe-4S] 2ϩ/ϩ cluster and one [3Fe-4S] ϩ/0 cluster (49,62,63). The C42D mutation in AvFdI converted the central cysteine of the [4Fe-4S] 2ϩ/ϩ binding motif, Cys 39 -X-X-Cys 42 -X-X-Cys 45 , to an aspartate. Naturally occurring ferredoxins that contain Cys-X-X-Asp-X-X-Cys motifs assemble [4Fe-4S] 2ϩ/ϩ clusters that are easily converted by air oxidation to [3Fe-4S] ϩ/0 clusters (10, 19 -24); this was the expected outcome in this case. We were therefore surprised to observe that the UV-visible absorption spectrum of the air-oxidized C42D FdI was nearly identical to that of the native FdI (Fig. 3), suggesting that the protein retained its original cluster composition. Oxidized [3Fe-4S] ϩ clusters exhibit characteristic g ϭ 2.01 EPR signals that integrate to one spin per molecule. If air-oxidized C42D FdI contained two [3Fe-4S] ϩ clusters, then we would expect the size of its EPR signal to double, integrating to two spins per molecule. In contrast, as shown in Fig. 4, the EPR spectrum of air oxidized C42D FdI exhibits a g ϭ 2.01 signal, similar in shape and size to that of native FdI, which is again consistent with its retaining the original cluster composition. Iron analysis was used to confirm that C42D FdI had not lost an iron to become a 6Fe-containing protein, and a value of  7.5 Ϯ 0.4 iron atoms/molecule for C42D FdI was obtained. That the cluster composition of air oxidized C42D FdI is identical to that of native FdI was subsequently confirmed by x-ray crystallography.
The [4Fe-4S] 2ϩ/ϩ Cluster of C42D FdI Has Only Three Cysteine Ligands-For reference, Fig. 5 shows the [4Fe-4S] 2ϩ/ϩ region of native FdI. In general, [4Fe-4S] 2ϩ/ϩ clusters like the one in native FdI obtain three of their four cysteine ligands from a Cys-X-X-Cys-X-X-Cys sequence, whereas the fourth cysteine ligand comes from a remote part of the protein, in this case Cys 20 . As shown in Fig. 5 there is also a fifth free cysteine (Cys 24 ) in van der Waals contact with the [4Fe-4S] 2ϩ/ϩ cluster of native FdI. Previous studies have shown that both C20A (64) and C20S (50) variants of FdI assemble [4Fe-4S] clusters with four cysteine ligands by recruiting Cys 24 . In order to use that cysteine as a new ligand, polypeptide rearrangements were necessary in both cases (50,64). For the C42D mutant, we wanted to address the possibility that the [4Fe-4S] 2ϩ/ϩ cluster might ignore the introduced aspartate and assemble instead using the four cysteine residues Cys 20 , Cys 24 , Cys 39 , and Cys 45 .
In general, visible region CD spectroscopy is very sensitive to the type of [Fe-S] cluster a protein contains. As shown in Fig. 6, oxidized native FdI exhibits a distinctive CD spectrum with contributions from both clusters. Previous studies have shown that both the wavelength dependence and form of this spectrum is dramatically changed for both the C20A and C20S variants, each of which undergo ligand exchange and structural rearrangement (50,64). In contrast, as shown in Fig. 6, the CD spectrum of oxidized C42D FdI is similar in wavelength dependence and form to that of native FdI. These data argue against a ligand exchange with an accompanied structural rearrangement.
Further support for the conclusion that [4Fe-4S] 2ϩ/ϩ cluster of C42D FdI does not recruit the free cysteine at position 24 (Fig. 5) as a new ligand comes from studies of the oxidation of the protein by ferricyanide. Previous studies have established that native FdI reacts with Fe(CN) 6 3Ϫ in a three-step degradative process in which the first step is a three-electron oxidation, involving the [4Fe-4S] 2ϩ/ϩ cluster and giving rise to a paramagnetic species (39,40). As shown in Fig. 7, this exhibits EPR up to quite high temperature and is most easily monitored in the range 40 -60 K (c and d), where the [3Fe-4S] ϩ cluster EPR is not detectable (a and b) (39,40). It has been established that this reaction absolutely requires a free cysteine at position 24 and the paramagnetic species is not observed for a C24A variant or for the C20A variant that has rearranged to utilize C24 as a new ligand (61). Fig. 7 shows that the behavior of C42D FdI is indistinguishable from that of native FdI in this reaction, confirming that C24 remains free in this variant. Taken together, the CD and the ferricyanide oxidation experiments lead to the conclusion that in solution the [4Fe-4S] 2ϩ/ϩ cluster of C42D FdI is ligated by only three cysteine residues.
Aspartate Ligation Is Confirmed by X-ray Crystallography-Prior to this report, there were no x-ray structures available for any [4Fe-4S] clusters with aspartate ligation. After 2 years of trying, and only after the original version of this paper had been submitted, we were finally successful in obtaining diffrac- tion quality crystals of C42D FdI. As expected from the spectroscopy, the structure of C42D FdI is very similar to native FdI with Asp 42 superposing closely on Cys 42 . The root-meansquare deviation of all main chain atoms in the two structures is 0.27 Å following least squares superposition. The main chain is displaced outward by 0.27 Å at the C␣ of residue 42 due to the mutation. There is a similar small displacement of the Pro 21 and Leu 44 side chains, and at Cys 45 as well. Overall the electron density for residues 39 -45 is well defined in C42D FdI, and these residues have comparable B-factors to the native protein, except for Leu 44 , which has significantly weaker density for its side chain.
In C42D FdI the O␦2 atom of Asp 42 is located within ϳ0.2 Å of the position of S␥ in the native protein, and so it is able to form a covalent bond, 2.11 Å in length, with Fe of the [4Fe-4S] cluster (Fig. 8A). In this orientation of the side chain, the O␦1 atom is farther from Fe, at 3.14 Å; therefore, Asp 42 acts as a monodentate ligand. It is important to note that the orientation of Asp 42 in Fig. 8A, with O␦1, O␦2 to Fe distances of 3.14 and 2.11 Å, respectively, is not the same orientation as the bidentate orientation proposed for the [4Fe-4S](Cys 3 Asp) cluster of P. furiosus Fd based on an nuclear magnetic resonance (NMR) structure (28) or for a [4Fe-4S] model complex with three thiolate ligands and one carboxylate ligand (65). Insertion of the aspartate side chain into the cluster-binding loop results in three short contacts to adjacent hydrophobic residues (Fig. 8A). These involve the O␦1 atom with C␥1 of Ile 40 and C␥ of Pro 21 , and the O␦2 atom with C␤ of Leu 44 . The tight packing of these three hydrophobic residues around the Asp 42 side chain is illustrated in Fig. 8B and will be discussed below. Fig. 3 compares the UV-visible absorption spectra of air oxidized (A) and dithionitereduced (B) C42D and native FdI. Because of its very low reduction potential, the [4Fe-4S] 2ϩ/ϩ cluster in FdI remains oxidized in the presence of dithionite, while the [3Fe-4S] ϩ/0 cluster is reduced. The similarity of spectra between native and C42D in both the oxidized and reduced states confirms that there is no major structural rearrangement due to the C42D mutation.

Direct Comparison of the Properties of the Oxidized [4Fe-4S] 2ϩ Clusters of C42D and Native FdI-
The wavelength dependence and form of the visible region CD spectra exhibited by different [Fe-S] proteins is not only sensitive to cluster type and protein environment but also changes substantially upon oxidation and reduction. Fig. 6B compares the CD spectra exhibited by dithionite-reduced C42D and native FdI. Again, the [4Fe-4S] 2ϩ/ϩ cluster remains oxidized under these conditions. The similarity of the wavelength dependence and overall shape of the spectrum confirm that there is no major structural change occurring upon reduction of the [3Fe-4S] ϩ cluster. The small changes in form of the reduced spectrum likely arise from the Cys to Asp change in ligation for the [4Fe-4S] 2ϩ cluster.
The 1 H NMR spectrum of oxidized native AvFdI shows several well resolved, paramagnetically shifted resonances (66). The five most downfield resonances are shown in Fig. 9. Three of these, designated A-C, have previously been assigned to the methylene protons of cysteine ligands to the [3Fe-4S] ϩ cluster, while the other two are believed to arise from methylene pro- FIG. 7. X-band EPR spectra of native (a, c, and e) and C42D (b,  d, and f) FdI either not treated (a and b) or treated (c-f) with 3 equivalents of K 3 Fe(CN) 6 (67). A sequence-specific assignment of resonances D and E has not clearly been made for A. vinelandii FdI; however, by analogy to a homologous ferredoxin from Bacillus schlegelii (68,69), it is probable that resonance D arises from ␤-CH 2 protons from Cys 42 , the residue of interest in this study, and that resonance E arises from ␤-CH 2 protons from Cys 45 . This assignment is supported by the observation of weak dipole contacts between the ␤-CH 2 protons of Cys 16 in the [3Fe-4S] ϩ cluster and the ␤-CH 2 protons of Cys 45 that ligates [4Fe-4S] 2ϩ , together with the observation in the 1 H NMR structure of a nuclear Overhauser effect between signals C and E (66,67). As shown in Fig. 9, the spectrum of C42D indicates that replacement of ligand Cys 42 by Asp perturbs the structures of both the Cys 42 site and the Cys 45 site, which results in one new signal appearing at 14.4 ppm. This perturbation at Cys 45 is observed in the x-ray structure described above (Fig. 8, A and B). This perturbation also gives rise to a splitting of the Cys 16 signal into two equal components with half-intensities. The substitution, however, does not change the other ␤-CH 2 ligand resonances for the [3Fe-4S] ϩ cluster, as seen in the same chemical shifts for signals A and B.
Reduction of the [4Fe-4S] 2ϩ Cluster of C42D Versus Native FdI-The [4Fe-4S] 2ϩ/ϩ cluster of native FdI has an unusually low reduction potential and cannot be reduced by dithionite at neutral pH. Previous studies have established, however, that it can be reduced using direct electrochemical methods (47). Fig.  10 shows bulk solution voltammetry of C42D AvFdI at pH 7.0. Couple A arises from [3Fe-4S] ϩ/0 , whereas couple B arises from [4Fe-4S] 2ϩ/ϩ , confirming again that the cluster composition of native and C42D FdI are the same. Examination of the data in Fig. 10 shows a shift of ϩ30 mV in the reduction potential of the [4Fe-4S] 2ϩ/ϩ couple; from Ϫ0.620 V (for native FdI) to Ϫ0.590 V (for C42D FdI) versus standard hydrogen electrode. As expected, there was no change in the reduction potential of the [3Fe-4S] ϩ/0 cluster, which is remote from the site of the muta-tion (60). The observation that the reduction potential of the [4Fe-4S] 2ϩ/ϩ cluster of C42D FdI is similar to that in native FdI is not unexpected. Thus, as shown in Table II, mutagenesis of other ferredoxins where a naturally occurring Cys-X-X-Asp-X-X-Cys motif is changed to Cys-X-X-Cys-X-X-Cys also resulted in only minor changes in reduction potential (36,71).
In most cases, reduced [4Fe-4S] ϩ clusters that have four cysteine ligands exhibit g av ϳ1.94 signals that arise from an S ϭ 1/2 ground state. This is true not only for naturally occurring ferredoxins but also for variants that have been constructed by mutagenesis where naturally occurring Cys-X-X-Asp-X-X-Cys motifs have been converted to Cys-X-X-Cys-X-X-Cys motifs (36,71). In contrast, where information is available, naturally occurring reduced ferredoxins that contain [4Fe-4S] ϩ cluster ligated by three cysteine and one aspartate ligands exhibit a g ϳ 5.3 signal that arises from the ground state m s ϭ Ϯ3/2. In some cases, these clusters exhibit only the S ϭ 3/2 spin state (for example, DaFdIII (Refs. 25, 72, and 73) and an A33Y mutant of PfFd (Ref. 74)), whereas, in others, they exhibit a mixture of S ϭ 1/2 and S ϭ 3/2 states (23,71). In order to compare the spectra of the reduced [4Fe-4S] ϩ clusters in C42D and native FdI, a low reduction potential system was employed, utilizing the photoreduction of 5Ј-deazariboflavin with EDTA as sacrificial electron donor. As shown in Fig. 11A, the reduced [4Fe-4S] ϩ cluster of C42D FdI exhibits an S ϭ 1/2 EPR signal with resonances, g ϭ 2.14, 1.86, and 1.79. The signal is much broader than that observed for the native protein. The same extent of signal broadness has been reported previously for other oxygen-liganded proteins, substrate-complexed aconitase 2 (75), aspartate-liganded PfFd (76), and probably OH Ϫ - Protein samples were suspended in 60 mM mixed buffer, 0.1 M NaCl, and 4 mM neomycin at 0°C, pH 7.0, scan rate 5 mV s Ϫ1 . Voltammograms measured on the third cycle. The concentration of C42D was lower than that of native FdI; hence, the currents are smaller. liganded dihydroxy-acid dehydratase (77). As shown in Fig. 11 (B and C), the C42D non-cysteine-liganded S ϭ 1/2 [4Fe-4S] ϩ resonance undergoes more rapid relaxation than the cysteineonly native [4Fe-4S] ϩ resonance since it is not observable at temperatures above 18 K and it is not easily saturated. One striking aspect is that no S ϭ 3/2 or other high spin signals were detected for 200 M C42D FdI under the conditions that were adjusted in temperature from 4.2 K to 30 K and/or in microwave power from 0.1 mW to 100 mW.
The [4Fe-4S] 2ϩ/ϩ cluster of C42D FdI is the only known example of an O 2 -stable 4Fe cluster with one aspartate and three cysteine ligands. The above data show that it is also the only example so far established of a cluster of that type that is exclusively S ϭ 1/2 in the reduced state. This result suggests that the prior observation of S ϭ 3/2 EPR for aspartate-liganded [4Fe-4S] ϩ clusters is not due to the aspartate ligand per se but rather to some other property of the cluster. One property that may unite the easily convertible, O 2 -sensitive, and S ϭ 3/2 clusters is their accessibility to solvent. In this regard, we note that the [4Fe-4S] ϩ cluster of the Fe protein of nitrogenase is solvent-exposed, extremely oxygen-sensitive, and exhibits a mixed S ϭ 1/2 and S ϭ 3/2 spin state despite the fact that it has four cysteine ligands rather than three cysteine and one aspartate (78). There are several reports in the literature showing that spin crossover is solvent-induced (23, 79 -85). For the nitrogenase Fe protein, the addition of a denaturing agent such as 0.4 M urea stabilizes the S ϭ 3/2 state, whereas both 50% (v/v) ethylene glycol and glycerol favor the S ϭ 1/2 spin state (79). In PfFd, 50% (v/v) glycerol, but not 0.4 M urea affects the relative populations of the S ϭ 1/2 and S ϭ 3/2 states and results in a shift in favor of the S ϭ 1/2 species (23). Solvents such as acetonitrile or dimethyl sulfoxide also induce spin mixtures in the synthetic analog [Fe 4 S 4 (SR) 4 ] 3Ϫ (R ϭ alkyl or aryl) (80). Finally, the 2[4Fe-4S] 2ϩ/ϩ PsaC protein mutants, C14D and C51D, are also known to be easily converted from S ϭ 3/2 to S ϭ 1/2 at the modified cluster upon binding to the photosystem I core complex, a situation that is likely to affect solvent accessibility (81)(82)(83)(84)(85).
Interestingly, there are several examples of other proteins showing S Ͼ 1/2 spin states with cysteine-only ligated [4Fe-4S] 2ϩ/ϩ clusters. For example, the Bacillus subtilis glutamine phosphoribosylpyrophosphate aminotransferase has the S ϭ 3/2 [4Fe-4S] ϩ as the dominant species and S ϭ 1/2 and 5/2 as minor components (86). In Thauera aromatica ferredoxin, which contains cysteine only-ligated 2[4Fe-4S] 2ϩ/ϩ clusters, a spin mixture of S ϭ 3/2 and 5/2 is observed for one cluster in the reduced state (87). Structures are not yet available for these clusters, and their solvent accessibility has not been reported.

The [4Fe-4S] 2ϩ/ϩ Cluster of C42D FdI Is Extremely Stable and Is Not Converted to a [3Fe-4S] ϩ/0 Cluster by Conventional
Methods-As shown above, like native FdI, C42D FdI is a completely air-stable protein that contains one [3Fe-4S] ϩ/0 cluster and one [4Fe-4S] 2ϩ/ϩ cluster. This is the only known example of a [4Fe-4S] 2ϩ/ϩ cluster with three cysteines and one water or aspartate ligand that is not converted to a [3Fe-4S] ϩ/0 cluster upon exposure to oxygen. We therefore attempted to do the conversion using alternative methods that had been successful with other proteins. One of these methods involves addition of potassium ferricyanide (10, 88 -91). However, in the case of C42D FdI, addition of Fe(CN) 6 3Ϫ did not result in cluster conversion, as monitored by quantitative examination of the g ϭ 2.01 signal that arises from the [3Fe-4S] ϩ clusters. A successful conversion should have doubled the size of that signal, but, as shown in Fig. 7, the [3Fe-4S] ϩ signal exhibited by C42D FdI was both qualitatively and quantitatively unchanged following the addition of ferricyanide.
In studies of DaFdIII, it was demonstrated that externally added thiolates could compete with the easily displaced aspartate ligand to form [4Fe-4S] 2ϩ/ϩ clusters with three endogenous cysteine ligands and one exogenous thiolate ligand (92). The products could be easily identified using direct electrochemical methods because their reduction potentials were significantly different from that of the original cluster. For example, the transformed (O 2 -sensitive) [4Fe-4S] 2ϩ/ϩ cluster in DaFdIII coordinates exogenous 2-mercaptoethanol as a ligand, whereupon its reduction potential becomes approximately 190 mV more negative. By contrast, no such shift in reduction potential was observed for C42D FdI when an electrode coated with a protein film was placed in a buffered solution of 715 mM 2-mercaptoethanol at pH 8.0. Thus, all of our attempts to use conventional methods to displace the aspartate ligand of C42D FdI were unsuccessful, thus demonstrating the extreme stability of this cluster.
Conversion of the C42D FdI [4Fe-4S] Cluster to a [3Fe-4S] Cluster Was Ultimately Accomplished Using a New Electrochemical Method-The mechanism of the 4Fe to 3Fe cluster conversion reaction in any protein has yet to be established but,

TABLE II
The comparison of reduction potentials for Cys-X-X-Cys-X-X-Cys versus Cys-X-X-Asp-X-X-Cys in the same protein Both DaFdIII and PfFd mutants have Cys instead of Asp in the Cys-X-X-Asp-X-X-Cys motif, whereas C42D FdI has Asp instead of Cys in the Cys-X-X-Cys-X-X-Cys motif. as shown in Reaction I, it has been suggested to involve transient oxidation of the [4Fe-4S] 2ϩ cluster to the superoxidized HiPIP [4Fe-4S] 3ϩ state prior to removal of iron (75,(93)(94)(95).

REACTION I
This possibility has recently been studied using direct electrochemical methods with Clostridium pasteurianum Fd, which contains two [4Fe-4S] 2ϩ/ϩ clusters with complete cysteine ligation. The experiment, which is discussed in detail elsewhere (96), involves subjecting a protein film at an electrode surface to short pulses (e.g. 0.1 to 4 s) at various high potentials to create the superoxidized [4Fe-4S] 3ϩ state. This species is usually unstable and breaks up, but it is possible to optimize conditions (pulse potential and duration) to control this oxidative fragmentation, so that just one iron atom is ejected, thus forming a [3Fe-4S] cluster. The success of the experiment is easily monitored by immediate cyclic voltammetry of the same film because the new [3Fe-4S] cluster exhibits two signals. One, designated AЈ, arises from the [3Fe-4S] ϩ/0 transition; the other, designated CЈ, which is more prominent, arises from a further two-electron [3Fe-4S] 0/2Ϫ transition.
The above method was successful in converting both of the two [4Fe-4S] 2ϩ/ϩ clusters of C. pasteurianum Fd to a [3Fe-4S] cluster, even though the original clusters each have four cysteine ligands. This method, however, has not so far been successful in converting the [4Fe-4S] 2ϩ/ϩ cluster of native FdI. Although the [4Fe-4S] 2ϩ/ϩ cluster of C42D FdI is extremely stable when compared with aspartate-ligated clusters in other proteins, we thought that it might be less stable than the cluster in native FdI. To test this, a variety of pulse times, pulse potentials, and other conditions were used. Optimum formation of new signals attributable to a second [3Fe-4S] cluster was observed when a film of C42D FdI was poised, at pH 8.0, at a potential of ϩ0.643 V for 4 s, 3 s, and finally 2 s with voltammograms recorded between each poise. The result is shown in Fig. 12. The 7Fe protein that is present at the start of the experiment (Fig. 12A, top panel) clearly shows three signals, AЈ, BЈ, and CЈ, corresponding to the [3Fe-4S] ϩ/0 , [4Fe-4S] 2ϩ/ϩ , and [3Fe-4S] 0/Ϫ2 redox couples, respectively (97). As is commonly found in the film voltammetry of [3Fe-4S]-clustercontaining proteins, the CЈ signal is kinetically complex and displays a large peak separation; thus, even at a relatively slow scan rate of 50 mV s Ϫ1 , the CЈ peak overlaps the BЈ peak in the oxidative scan whereas they are clearly separated in the reducing direction. Fig. 12A (bottom panel) shows the voltammetry of the product; the BЈ couple that arises from the indigenous [4Fe-4S] 2ϩ/ϩ cluster has disappeared and has been replaced by two new signals labeled A*Ј and C*Ј. Their appearance is fully consistent with the formation of a second [3Fe-4S] cluster.
The change in ratios AЈ/CЈ after the pulse probably arises because application of high potentials also causes extensive desorption of the protein from the electrode surface (96). The desorbed ferredoxin molecules, many of which will escape without undergoing the cluster transformation, are initially still able to diffuse back to the electrode and undergo electron transfer, but without becoming adsorbed. The AЈ and CЈ signals in a diffusional voltammogram do not reflect the expected 1-to-2 stoichiometry, i.e. the CЈ signal is rarely observed unless the protein is tightly bound to the electrode. (This is most likely due to the complexity of the cooperative two-electron redox reaction, which involves sequential electron-proton transfers that require a sufficiently long residence time at the electrode (Ref. 97).) Fig. 12B compares the pH dependence of the reduction potentials of the new and indigenous [3Fe-4S] cluster signals for C42D FdI. The slopes for both redox couples and their pK a values are similar for both clusters, with the new AЈ couple having pK a ϭ 7.7 and E alk ϭ Ϫ0.341V. The pK a arises from coupled proton-electron transfer reactions that are well characterized for the native protein (97). The reduction potential of the new [3Fe-4S] ϩ/0 cluster is about 100 mV more positive than that of the indigenous cluster. This is not surprising since [3Fe-4S] ϩ/0 clusters in different proteins vary greatly in their reduction potentials.
Even though we were successful in converting the 4Fe cluster of C42D FdI to a 3Fe cluster by electrochemical pulsing, the product could not be readily converted back to a 4Fe cluster. Thus, unlike the situation with easily converted clusters (70,98), when films of the 6Fe form of C42D FdI were placed in 5 mM solutions of Fe 2ϩ , Zn 2ϩ , and Tl ϩ , no signs of metal uptake were observed.
Why Can't the C42D FdI [4Fe-4S] 2ϩ/ϩ Cluster Be Easily Converted to a [3Fe-4S] ϩ/0 Cluster?-Taken together, the above data demonstrate that it is not the presence of an aspartate ligand in the central position per se that allows the facile conversion of a [4Fe-4S] 2ϩ/ϩ cluster to a [3Fe-4S] ϩ/0 cluster upon oxygen and ferricyanide addition. Fig. 8 (A and B) shows the structure for the [4Fe-4S] 2ϩ/ϩ cluster region of C42D FdI. The aspartate fits tightly and is held in place and restrained by van der Waals contacts with the three hydrophobic residues Ile 40 , Leu 44 , and Pro 21 . These interactions would be expected to limit access of solvent to the [4Fe-4S] 2ϩ/ϩ cluster at the site of Asp 42 ligation, and at the same time prevent dissociation of the aspartate from the cluster. Although the carboxylate carries the same negative charge as a cysteine ligand, its oxygen atoms are less polarizable than sulfur. The negative charge, within the hydrophobic cavity formed by Pro 21 , Ile 40 , and Leu 44 , is compensated not only by the [4Fe-4S] 2ϩ cluster, but also by two electrostatic interactions, one involving O␦2 with the amide of Leu 44 , and the other O␦1 with Pro 21 (Fig. 8A). The former interaction has good geometry for a hydrogen bond, and the latter, justified by the positions of the residues in the electron density, may represent a CH . . . O hydrogen bond. Alternatively, O␦1 of Asp 42 may be protonated. Together, these hydrophobic contacts and electrostatic interactions help account for the stability of Asp 42 as a ligand.
Examination of the sequences of the easily convertible proteins (Fig. 2) with a central aspartate shows that in most cases the residue at position 44 is much smaller and/or not hydrophobic, which may facilitate the movement of the aspartate. An exception is PfFd, which does easily convert, albeit at a slower rate than for DaFdIII (70), and has a hydrophobic Ile at the position homologous to the FdI residue number 44. Thus, although the residue 44 may be important in facilitating the cluster conversion, it cannot be the controlling factor. In order for the cluster conversion to occur, the backbone must be able to move to allow the aspartate to move away from the cluster. Examination of the sequences of the easily convertible proteins (Fig. 2) shows that, without exception, those proteins that have a central aspartate have a glycine on one or both sides of the aspartate to facilitate this backbone movement. In contrast, the Asp 42 residue of C42D FdI is flanked by aspartate and alanine, which restrict the corresponding backbone movement. Future experiments could therefore test the hypothesis that construction of a D41G/C42D double mutant in FdI ought therefore to lead to an easily convertible cluster or that conversion of the homologous Gly 41 position in PfFd to an aspartate would lead to a [4Fe-4S] 2ϩ/ϩ cluster that could no longer be converted to [3Fe-4S] ϩ/0 .