A T14C Variant of Azotobacter vinelandii Ferredoxin I Undergoes Facile [3Fe-4S] 0 to [4Fe-4S] 2 1 Conversion in Vitro but Not in Vivo *

[4Fe-4S] 2 / 1 clusters that are ligated by Cys- X-X- Cys-X-X- Cys sequence motifs share the general feature of being hard to convert to [3Fe-4S] 1 /0 clusters, whereas those that contain a Cys- X-X- Asp- X-X- Cys motif undergo facile and reversible cluster interconversion. Little is known about the factors that control the in vivo assem-bly and conversion of these clusters. In this study we have designed and constructed a 3Fe to 4Fe cluster conversion variant of Azotobacter vinelandii ferredoxin I (FdI) in which the sequence that ligates the [3Fe-4S] cluster in native FdI was altered by converting a nearby residue, Thr-14, to Cys. Spectroscopic and electrochemical characterization shows that when purified in the presence of dithionite, T14C FdI is an O 2 -sensitive 8Fe protein. Both the new and the indigenous clusters have reduction potentials that are significantly shifted compared with those in native FdI, strongly suggesting a significantly altered environment around the clusters. Interestingly, whole cell EPR have revealed that T14C FdI exists as a 7Fe protein in vivo . This

One of the most interesting and important features of protein-bound [Fe-S] clusters is their ability to convert from one form to another (for reviews see Refs. [1][2][3][4][5][6][7]. The simplest of these reactions involves the interconversion of [3Fe-4S] ϩ/0 and [4Fe-4S] 2ϩ/ϩ clusters. These two cluster types are structurally very closely related and differ only by the presence or absence of a single iron atom at one corner of a cube. The physiological relevance of an iron loss or uptake mechanism is best illustrated by the example of aconitase and related dehydratases. In the best studied system, aconitase, the [3Fe-4S] cluster is in-active, and the spontaneous conversion to the [4Fe-4S] form of the enzyme represents a self-activation of the enzyme (1,8). For the related iron-responsive element mRNA-binding protein, which is involved in iron homeostasis, the reverse reaction may be the first step on the route to apoprotein production (9). In a quite different protein, Desulfovibrio gigas ferredoxin II (DgFdII), 1 the 3Fe to 4Fe cluster conversion reaction, which is proposed to be controlled by the physiological effector pyruvate, is accompanied by a change in subunit composition of the protein, and the two forms participate in completely different electron transfer pathways (10,11). For Bacillus subtilis amidotransferase the stability of the enzyme depends on the presence of a [4Fe-4S] 2ϩ/ϩ cluster, and exposure to dioxygen inactivates the enzyme by destroying its cluster (12). Indeed numerous [4Fe-4S] 2ϩ/ϩ -containing proteins are extremely sensitive to dioxygen, and the destruction of the cluster often initiates with the formation of a [3Fe-4S] cluster.
Most proteins that contain [3Fe-4S] ϩ/0 clusters do not undergo this type of cluster interconversion, and those that do vary in the ease and reversibility of the reaction (1)(2)(3)(4)(5)(6)(7). At present the structural features that are responsible for the variable reactivity of 3Fe and 4Fe clusters are not understood. In the majority of proteins that are known to undergo facile 3Fe to 4Fe cluster interconversion, including aconitase, the [4Fe-4S] 2ϩ/ϩ cluster has one noncysteine ligand, and it is the iron atom that is coordinated by that ligand that is labile (13)(14)(15)(16)(17)(18)(19)(20)(21). In Desulfovibrio africanus FdIII, Desulfovibrio vulgaris FdI, and Pyrococcus furiosus Fd, the [4Fe-4S] 2ϩ/ϩ forms of the proteins appear to have three cysteine and one aspartate as ligands, and again the iron atom associated with the aspartate is the labile iron (13)(14)(15)(16)(17)(18)(19). It is interesting in those cases that the aspartate replaces the cysteine at the central position of a [4Fe-4S] 2ϩ/ϩ Cys-X-X-Cys-X-X-Cys-binding motif (Fig. 1). If the aspartate is replaced by a cysteine using site-directed mutagenesis, a stable 4Fe cluster results, but it can no longer be easily converted to a 3Fe cluster (22)(23)(24). Like these mutants, most native proteinbound [4Fe-4S] 2ϩ/ϩ clusters that have four cysteine ligands can be converted to the [3Fe-4S] ϩ/0 state only during destructive reactions en route to apoprotein, and stable 3Fe cluster forms are not available to study the reverse reaction (25)(26)(27). An exception is DgFdII where the central Cys of the Cys-X-X-Cys-X-X-Cys binding motif is covalently modified by a thiomethane group and rotated away from the cluster and is therefore not used as a ligand in the [3Fe-4S] ϩ/0 form of the protein (28).
Azotobacter vinelandii FdI is an extremely well characterized 7Fe protein, which contains a [3Fe-4S] cluster that cannot be easily converted to a [4Fe-4S] cluster (29 -33). As shown in Fig. 1 the AvFdI [3Fe-4S] cluster is ligated by cysteine residues at positions 8, 16, and 49. The 3Fe region of the protein therefore does not have either a Cys-X-X-Cys-X-X-Cys or a Cys-X-X-Asp-X-X-Cys motif. There is an additional free Cys at position 11 but that Cys is far removed from the cluster and is not used as a ligand. Our interest in converting the 3Fe cluster of AvFdI to a 4Fe cluster by mutagenesis is motivated both by a desire to understand the structural features that control the interconversion reactions and by the suggestion that the 3Fe to 4Fe conversion reaction, which is so difficult to accomplish with native AvFdI in vitro, might be physiologically relevant (34). That suggestion is based in part on the surprising observation that when apo-FdI is reconstituted in vitro with iron and sulfide the product is an 8Fe protein containing two [4Fe-4S] 2ϩ/ϩ clusters rather than the extremely stable and well characterized 7Fe form of the protein that contains one [3Fe-4S] ϩ/0 and one [4Fe-4S] 2ϩ/ϩ cluster (32).
In a previous study we constructed a Y13C FdI variant that could easily be modeled with a 4Fe cluster in the 3Fe cluster position. That design was unsuccessful, resulting in a 7Fe protein possibly because A. vinelandii covalently modified the cysteine to form a persulfide that could not be used as a ligand (35). That interpretation is consistent with a recent report that the same mutation in a related ferredoxin from Bacillus schlegelii, which was heterologously expressed in Escherichia coli, appears to contain a 4Fe cluster in the 3Fe position (24). 2 In another study we were successful in constructing an 8Fe version of FdI by deleting residues Thr-14 and Asp-15 to create a Cys-X-X-Cys-X-X-Cys motif in the region of the 3Fe cluster (36). However, like the four-Cys coordinated clusters discussed above, the conversion of the new 4Fe cluster back to the 3Fe cluster was a destructive reaction, and a pure sample of the 7Fe form of ⌬T14/⌬D15 FdI could not be isolated (36). Here we report the construction, characterization, and interconversion of a T14C variant of FdI that can be isolated both in 7Fe and 8Fe forms and the observation that it is the 7Fe form that accumulates in vivo.

EXPERIMENTAL PROCEDURES
Mutagenesis of fdxA, Expression, and Purification of T14C FdI-The oligonucleotide used for the mutagenesis had the sequence 5Ј-AAGTG-CAAGTACTGCGATTGTGTTGAA-3Ј which differs from the wild-type sequence by the substitution of TGC (encoding Cys) for ACC (encoding Thr). The oligo-directed mutagenesis, FdI overexpression in A. vinelandii, and cell growth were performed as described previously (37). The parent A. vinelandii strain used for the expression of T14C FdI, designated DJ138, is lacking the genes encoding AvFdI and flavodoxin. The constructed T14C FdI variant strain is designated DJ138/pBS3A1.
The purification of native FdI and T14C FdI follows a modification of the purification method described previously (38). All steps were carried out anaerobically with or without (as specified below) 2 mM Na 2 S 2 O 4 present in all buffers. A. vinelandii T14C FdI variant was grown under the N 2 -fixing condition, and cell-free extracts were prepared from about 1 kg of cell paste. The cell-free extracts were loaded onto a 5 ϫ 20-cm DE52 cellulose column equilibrated with 0.025 M Tris⅐HCl, pH 7.4. A 3-liter linear 0.1-0.5 M NaCl gradient was applied, and T14C FdI was eluted completely by a NaCl concentration of 0.45 M. The ferredoxin fraction, while being diluted with 2 volumes of 0.1 M potassium phosphate buffer, pH 7.4, was immediately loaded onto a 2.5 ϫ 12 cm DE52 cellulose column equilibrated with the same buffer. The column was then washed with 2 liters of 0.1 M potassium phosphate buffer, pH 7.4, 0.12 M in KCl, at 240 ml/min. T14C FdI was eluted with 0.3 M KCl. Saturated ammonium sulfate in 0.05 M Tris⅐HCl, pH 7.4, was added very slowly with stirring to the T14C FdI fraction to 75% saturation in ammonium sulfate. The precipitated protein was recovered by centrifugation and dissolved in the minimum volume of 0.025 M Tris⅐HCl, pH 7.4, 0.1 M NaCl. The T14C FdI solution was then loaded onto a 2.5 ϫ 100-cm Sephadex G-50 superfine (Amersham Pharmacia Biotech) gel filtration column equilibrated with 0.025 M Tris⅐HCl, pH 7.4, 0.1 M NaCl. The eluted T14C FdI fraction was concentrated using an Amicon ultrafiltration unit with a YM-10 membrane.
Spectroscopy-For spectroscopic studies all protein samples were prepared anaerobically under argon in a Vacuum Atmospheres glove box (O 2 Ͻ1 ppm) using degassed buffers. Where concentrated protein was required, the samples were concentrated and buffer exchanged using Centricon-10 microconcentrators. The photoreduction reaction was carried out by mixing T14C FdI, 5Ј-deazariboflavin and EDTA at final concentrations of 100, 200, and 20 M, respectively, in 0.1 M Tris⅐HCl, pH 7.4, and then illuminating for 1 min using white light from a slide projector. UV-visible spectra were obtained with a Hewlett-Packard 8452 Diode Array UV-visible spectrophotometer. CD spectra were recorded using a JASCO J720 spectropolarimeter. EPR spectra were obtained using a Bruker 300 Ez spectrometer, interfaced with an Oxford Instruments ESR-9002 liquid helium continuous flow cryostat. Samples for whole cell EPR were prepared as described elsewhere (36).
Iron Content Analysis-To determine iron content, samples were digested, and the analysis was carried out as described elsewhere (39) using FeCl 3 ⅐6H 2 O to generate a standard curve with the native FdI as control. A parallel Lowry assay (40) was carried out to confirm the quantities of proteins used in the iron analysis.
tions of 165 M and 1.32 mM, respectively. Polymerization occurred after 5 min of gel casting. To the stacking gel solution, 10% ammonium persulfate and TEMED were added to give final concentrations of 300 M and 5 mM, respectively. Polymerization occurred within 30 min. The gels were pre-run for 30 min at 70 V with the running buffer containing 2 mM Na 2 S 2 O 4 . After the protein samples were loaded, the gels were run for 2.5 h at 70 V. The cathode buffer was refreshed with fresh running buffer at 30-min intervals to ensure the continuous presence of sodium dithionite in the gel. The completed gels were treated with Coomassie Blue staining.
Electrochemistry-Purified water of resistivity ϳ18 M⍀⅐cm (Millipore) was used in all experiments. The buffers MES, HEPES, and TAPS, and co-adsorbates polymyxin B sulfate and neomycin sulfate were purchased from Sigma. Other reagents were purchased from Aldrich and were of at least analytical grade. An AutoLab electrochemical analyzer (Eco Chemie, Utrecht, The Netherlands) was used to record DC cyclic voltammograms (in the digital mode) and square-wave voltammetry. The three-electrode configuration featuring all-glass cells (typically holding 600 l and either a single pot or multipot for pH dependence and metal-uptake experiments) has been described previously (42). The sample compartment was maintained at 0°C. All potential values are given with reference to the standard hydrogen electrode. The saturated calomel reference electrode was held at 22°C, which we have adopted as E(saturated calomel reference electrode) ϭ ϩ243 mV versus standard hydrogen electrode. Reduction potentials from cyclic voltammograms were calculated as the average of the anodic and cathodic peak potentials, E 0 Ј ϭ 1/2(E pa ϩ E pc ). The pyrolytic graphite "edge" (PGE) electrode (surface area typically 0.18 cm 2 ) was polished prior to each experiment with an aqueous alumina slurry (Buehler Micropolish For film experiments, the solutions used to coat the electrodes were as follows. For the 7Fe form, a 300 M protein solution was prepared using 45 l of the mixed buffer cell solution at pH 7.0, mixed with 10 l of a stock solution of 1.6 mM protein in 25 mM Tris⅐HCl at pH 7.4. For films of the 8Fe ferredoxin, the solution contained 120 M protein in 0.1 M Tris⅐HCl, 0.1 M NaCl, 200 g/ml polymyxin, with 2.5 mM sodium dithionite, and 1 mM Fe(II) at pH 7.4. Films were produced by painting a freshly polished electrode surface with about 1 l of chilled protein solution using a Pasteur pipette drawn to a fine capillary tip, following which the electrode was placed promptly into the cell solution.
For bulk solution voltammetry of the 7Fe form, the solution contained 160 M protein, 25 mM Tris⅐HCl, and 0.1 M NaCl at pH 7.4. For the 8Fe form, the solution contained 120 M protein, in 0.1 M Tris⅐HCl and 0.1 M NaCl with 1 mM Fe(II), and 2.5 mM sodium dithionite, at pH 7.4. Small aliquots of neomycin stock solution were added (final concentration 4.0 mM) to promote a strong and persistent electrochemical response. This solution was used for both cyclic and square-wave voltammetry.
X-ray Crystallography-Crystals of oxidized T14C FdI in the 7Fe form were grown by seeding using small native FdI crystals as described previously (48). A large, single crystal, 0.5 ϫ 0.7 ϫ 0.9 mm in size, was mounted in a thin-walled glass capillary using a synthetic mother liquor of 4.8 M (NH 4 ) 2 SO 4 , 0.45 M Tris⅐HCl, pH 8.1. Data were collected at room temperature using CuK␣ radiation from a RU200 x-ray generator operated at 40 kV, 80 mA, and equipped with a graphite monochromator and Xentronics X-1000 area detector mounted on a Siemens P4 goniostat. Data were collected with a single phi scan of 180°w ith 0.25°oscillations and 6-min exposures per frame. The crystal did not exhibit significant deterioration in the 3 days required for data collection. The data were indexed, integrated, reduced, merged, and scaled using the Xengen suite of programs (Table I) (49).
The structure was solved by molecular replacement using Xplor version 3.8 (50). Starting coordinates of native FdI (31) were modified to contain a Gly at both residues 11 and 14, and all H 2 O molecules were omitted from the model. This model was refined as a rigid body using data in the resolution range 8.0 to 4.0 Å and then by positional refinement using data to 4.0-, 3.5-, 3.0-, 2.7-, and 2.5-Å resolution. The resulting phases, unbiased for the side chains of residues 11 and 14, were used to calculate ͉F o ͉ Ϫ ͉F c ͉ and 2͉F o ͉ Ϫ ͉F c ͉. Fourier maps with all data in the resolution range 20.0 to 2.5 Å. The difference Fourier map revealed large positive 5 peaks at the positions of S␥ of Cys-11 and C-␥ 2 of Thr-14 of native FdI, providing an internal check on the correctness of the phases and revealing the position of the S␥ atom in the mutated Cys-14 side chain. The 2.5-Å resolution 2͉F o ͉ Ϫ ͉F c ͉ map was consistent with Cys-11 and Cys-14 side chains and was used to model the Cys-14 side chain and remodel the Cys-11 side chain. Density fitting and model building were carried out with the Xtalview suite of programs (51). The electron density and difference electron density did not indicate any significant changes in the [3Fe-4S] cluster or other portions of the structure. The structure was refined by Powell minimization and isotropic B factor refinement at 2.3 Å and then 2.1 Å resolution using Xplor version 3.8 (50). Side chains of three residues in contact with Thr-14 in native FdI, Asp-15, Lys-84, and Lys-85, were adjusted slightly to fit the 2͉F o ͉ Ϫ ͉F c ͉ electron density. Difference Fourier maps were used to locate peaks Ͼ3 for H 2 O molecules; 48 H 2 O molecules refined with B factors Ͻ50.0 A 2 were retained in the model. Statistics for the final refined structure are given in Table I. There are no non-glycine outliers in the Ramachandran plot. Coordinates for oxidized 7Fe T14C FdI have been deposited with the Protein Data Band code 1A6L.
In an attempt to convert the [3Fe-4S] cluster into a [4Fe-4S] cluster in the crystal lattice, as observed in solution, two additional data sets were collected. A large single crystal of 7Fe T14C FdI was wedged on a capillary and bathed in a synthetic mother liquor of 4.8 M (NH 4 ) 2 SO 4 , 0.45 M Tris⅐HCl, pH 8.1, and the capillary was sealed with mineral oil. The crystal was reduced as previously reported in the determination of the structure of reduced native FdI (52), i.e. a few crystals of methyl viologen and subsequently a few crystals of sodium dithionite were introduced through the oil. After about 8 h the blue color of the reduced methyl viologen diffused to the crystal indicating that the mother liquor surrounding the crystal was anaerobic. At pH 8.1 the [3Fe-4S] cluster is readily reduced by sodium dithionite. A complete data set to 2.5-Å resolution was collected to test the effect of reduction on T14C FdI in the crystal. The blue color of reduced methyl viologen remained in the synthetic mother liquor surrounding the crystal throughout data collection. The structure of oxidized 7Fe T14C FdI was refined against this data set as described above. The R factor for the reduced T14C FdI model was 0.237 for all data in the range 8.0 to 2.5 Å and without H 2 O molecules included (data not shown). The difference Fourier map using all data in the range 20.0 to 2.5 Å showed no positive or negative 3 peaks except at the positions of H 2 O molecules with low B factors in the oxidized 7Fe T14C FdI structure, which appeared as ϩ4 and ϩ5 peaks. Therefore, no large scale conformational change occurs upon reduction of the [3Fe-4S] cluster in crystals of T14C FdI.
Immediately after collecting the 2.5-Å data set for reduced 7Fe T14C FdI, a few additional crystals of sodium dithionite were introduced through the mineral oil followed by a few crystals of ferrous ammonium sulfate. After 8 h the darker blue color of freshly reduced methyl viologen had diffused to the crystal. Complete 2.9-Å resolution data were then collected. Again the blue color of the synthetic mother liquor around the crystal remained throughout data collection, whereas slow oxidation at the top of the capillary (nearest the oil) revealed a pale yellow color of the synthetic mother liquor, indicating that the concentration of ferrous ammonium sulfate was significant, probably in the millimolar range. The structure of reduced 7Fe T14C FdI already refined at 2.5-Å resolution was refined against the 2.9-Å data set resulting in an R factor of 0.198 for all data in the range 8.0 to 2.9 Å without H 2 O molecules included (data not shown). Again the difference Fourier map was completely featureless except for ϩ4 and ϩ5 peaks at the positions of tightly bound H 2 O molecules. Therefore, the [3Fe-4S] to [4Fe-4S] cluster conversion reaction observed in solution does not occur in crystals of T14C FdI. Attempts to crystallize the 8Fe T14C FdI directly from solution were unsuccessful.

RESULTS AND DISCUSSION
As revealed by spectroscopy and x-ray crystallography, when isolated aerobically or anaerobically in the presence of excess dithionite, A. vinelandii ferredoxin I (AvFdI) contains one [4Fe-4S] 2ϩ/ϩ and one [3Fe-4S] ϩ/0 cluster (29 -31, 38), and the [3Fe-4S] ϩ cluster of the protein can be easily observed in vivo by whole cell EPR (see Ref. 36 and see below). Unlike for many 3Fe-containing proteins, the [3Fe-4S] ϩ/0 cluster of FdI cannot be easily converted to a 4Fe cluster. However, an 8Fe version of the protein can be produced in low yield by reconstitution of apoprotein (32,33). An 8Fe version can also be produced directly from the 7Fe form by a method that involves use of the denaturing agent guanidine⅐HCl (33). Although the ligands for the new 4Fe cluster in native FdI have not been identified, the ligands may be cysteines 8, 11, 16, and 49 with the free Cys at position 11 being recruited as the fourth ligand (Fig. 1). In this study we have attempted to create a 4Fe cluster in the 3Fe cluster position by introducing an additional Cys at position 14 in the region of the cluster (Fig. 1).
Spectroscopic Characterization of an 8Fe Form of T14C FdI-Native FdI is an air-stable protein. The eight-iron form of native FdI (32, 33) and a number of FdI mutant variants (35,36), however, are highly oxygen-sensitive. Therefore, as discussed under "Experimental Procedures," our first attempts to purify the T14C variant were carried out under strictly anaerobic conditions in the presence of excess dithionite. The UVvisible absorption spectrum of anaerobically oxidized T14C FdI (data not shown) in the 400-nm region is quite different from the spectrum exhibited by native FdI and is similar to the spectrum exhibited by the A. vinelandii 2[4Fe-4S] 2ϩ/ϩ FdIII (53), which is very characteristic of 8Fe ferredoxins (54). The magnitude of the absorbance, however, is very similar for both T14C and native FdI indicating that there has been no loss of iron in the T14C variant. That two [Fe-S] clusters must be present in T14C FdI is also supported by direct iron analysis, which gives 6.8 Ϯ 0.2 iron atoms per molecule for native FdI versus 8.0 Ϯ 0.2 iron atoms per molecule for T14C FdI.
Further support for anaerobically isolated T14C containing two [4Fe-4S] 2ϩ/ϩ clusters comes from the comparison of the CD spectra of native FdI, T14C FdI, and FdIII from A. vinelandii that are shown in Fig. 2. Again the wavelength dependence and form of the mutant spectrum are very characteristic of spectra obtained for 8Fe ferredoxins and quite different from that of native FdI (55,56).
In general the presence of oxidized [3Fe-4S] ϩ clusters or reduced S ϭ 1/2 [4Fe-4S] ϩ clusters can be easily confirmed by EPR spectroscopy where the former cluster type gives a very characteristic g ϭ 2.01 signal (38) and the latter gives a characteristic anisotropic g av ϳ1.94 spectrum (56). Fig. 3A shows that in sharp contrast to native FdI, anaerobically oxidized T14C FdI is EPR silent strongly supporting the conclusion that it no longer contains a [3Fe-4S] ϩ/0 cluster. For native FdI, because of the very low reduction potential of its [4Fe-4S] 2ϩ/ϩ cluster, the addition of excess dithionite only reduces the [3Fe-4S] ϩ cluster to the zero oxidation level without reducing the [4Fe-4S] ϩ cluster. Thus the reduced spectrum of native FdI is EPR silent in the perpendicular mode. For the 8Fe AvFdIII the reduction potentials of both [4Fe-4S] 2ϩ/ϩ clusters are sufficiently low that neither can be reduced by excess dithionite (53). Thus the addition of excess dithionite to oxidized FdIII does not change the UV-visible or the CD spectrum, and FdIII remains EPR silent in the presence of excess dithionite (53). The same results are obtained for T14C FdI (data not shown) strongly suggesting that its two [4Fe-4S] 2ϩ/ϩ clusters have reduction potentials too low to be reduced by excess dithionite. For FdIII, reduction was accomplished using 5Ј-deazariboflavin (53), and the two [4Fe-4S] 2ϩ clusters of T14C FdI can also be reduced by this system. Thus, as shown in Fig. 3B, following reduction with the 5Ј-deazariboflavin system, a g av ϭ 1.94 signal is observed that has power and temperature dependence characteristic of [4Fe-4S] ϩ clusters. This signal integrates to 1.4 spins per molecule and thus represents partial reduction of the two [4Fe-4S] 2ϩ clusters contained within T14C FdI. 3 Electrochemical Characterization of the 8Fe Form of FdI-Taken together, the iron analysis, UV-visible, CD, and EPR spectroscopic data lead to the conclusion that when purified in the presence of excess dithionite T14C FdI is an 8Fe protein containing two very low potential [4Fe-4S] 2ϩ/ϩ clusters. To characterize further the redox properties of T14C FdI and to define the reduction potentials for the two [4Fe-4S] 2ϩ clusters, direct electrochemical experiments were employed. Voltammetry of the 8Fe T14C FdI showed the presence of two centers with unequal reduction potentials, although sufficiently similar in value to produce overlapping signals. Protein film cyclic voltammetry at pH 7.0 revealed one broad, asymmetric, reductive peak at Ϫ527 mV and a similar oxidative peak at Ϫ562 mV. Both peaks gave the appearance of being comprised of two one-electron peaks at slightly different potentials. Between pH 6.0 and 9.0, the peak potentials and shapes showed little pH dependence. Films appeared unstable below pH 6.0 and above pH 9.0. Similar observation of two closely spaced reduction potentials was made using bulk solution cyclic voltammetry, so consequently, square-wave voltammetry was used in order to obtain improved resolution. Fig. 4 shows a square-wave voltammogram obtained for a solution of 8Fe T14C FdI at pH 7.4, using an amplitude of 10 mV, frequency 25 Hz, and a potential step increment of 1 mV. The two redox couples are now sufficiently resolved to enable estimation of their reduction potentials, the values being Ϫ519 and Ϫ565 mV. We did not attempt to analyze this voltammetry further, but it should be noted that these data together confirm the absence of a 3Fe cluster which, as discussed below, always exhibits a third wave arising from the [3Fe-4S] 0/2Ϫ couple.
Until very recently it was believed that the two [4Fe-4S] 2ϩ/ϩ clusters in naturally occurring 8Fe ferredoxins must have very similar or identical reduction potentials. Currently, the only known naturally occurring exceptions to this general rule are the homologous proteins, A. vinelandii FdIII (53) and Chromatium vinosum ferredoxin (57). Our results now show that the T14C FdI variant can accommodate two [4Fe-4S] 2ϩ/ϩ clusters which, like the exceptions just discussed and our previous 8Fe ⌬T14/⌬D15 FdI variant, have different reduction potentials. Significantly, although it is uncertain which reduction potential corresponds to each cluster, it is clear that the value for the indigenous center in T14C FdI has become more positive by at least 60 mV relative to the value in the 7Fe protein. In previous studies we have shown that the reduction potential of the 4Fe cluster could be varied over a 150-mV range (58 -60) without affecting the reduction potential of the 3Fe cluster and that the 3Fe cluster could be varied over a 60-mV range (35-37) without affecting the 4Fe cluster potential. One explanation for the large shift in reduction potential for the indigenous [4Fe-4S] 2ϩ/ϩ cluster in T14C FdI is that the structural change required to produce the new 4Fe cluster has caused structural changes near the indigenous cluster as well. Another possibility is that the indigenous [4Fe-4S] 2ϩ/ϩ cluster senses the positive change in core charge that has occurred at the neighboring center, something that did not occur for the 8Fe ⌬T14/⌬D15 FdI variant whose indigenous [4Fe-4S] 2ϩ/ϩ reduction potential did not change (36).
Physiological Considerations-A. vinelandii FdI has an electron transfer function in a metabolic process that is unrelated to N 2 fixation but is important for cell growth (61). In previous studies we have shown that even subtle mutations near the [3Fe-4S] ϩ/0 cluster have serious negative effects on the growth rate of the cell (37) and that converting that cluster to a [4Fe-4S] 2ϩ/ϩ cluster in a ⌬T14/⌬D15 variant led to a strain with a very slow growth rate compared with the strain expressing native FdI (36). In addition to its electron transfer function, FdI has a regulatory function in controlling the expression of the fpr gene product, NADPH:ferredoxin reductase (62)(63)(64). While growing strains that express the 8Fe-containing ⌬T14/⌬D15 variant, we observed that the strains overproduced the siderophore azotobactin D such that the cells appeared to respond as if they have less iron than is really available (36). Although this result was counterintuitive, we expected that it was somehow a manifestation of the regulatory function of FdI which resulted from the conversion of the 3Fe cluster to a 4Fe cluster. The strain expressing T14C FdI grew at a similar slow rate as the ⌬T14/⌬D15 strain (and other strains carrying more subtle mutations near the 3Fe cluster), but it did not overproduce azotobactin D. This led us to suspect that unlike the situation for ⌬T14/ ⌬D15 FdI, which accumulates as an 8Fe protein in vivo, the form of T14C FdI that accumulates in vivo might be a 7Fe rather than an 8Fe form of the protein. This suspicion was confirmed by the whole cell EPR data in Fig. 5 which clearly show that strains overexpressing T14C FdI and native FdI accumulate similar amounts of [3Fe-4S] ϩ -containing proteins.
Spectroscopic and Electrochemical Characterization of a 7Fe Form of T14C FdI-To confirm that cells accumulate a stable 7Fe form of T14C FdI, we purified the protein anaerobically but in the absence of dithionite. Because oxidized [3Fe-4S] ϩ clusters contain three ferric iron atoms and oxidized [4Fe-4S] 2ϩ clusters formally contain only two ferric iron atoms, the conversion of a 3Fe cluster to a 4Fe cluster is well known to require prior reduction of one of the iron atoms in the 3Fe cluster (1). Eliminating dithionite during cell rupture and protein purification should thus prevent the cluster conversion reaction. Fig. 6 compares the UV-visible spectra of T14C FdI isolated in the absence of dithionite to that of native FdI. In contrast to the situation for 8Fe T14C FdI the two spectra are now indistinguishable (Fig. 6A). T14C FdI can be reduced by dithionite as shown by the UV-visible spectrum, which is again extremely similar to that of the dithionite-reduced native FdI (Fig. 6B). The same result is obtained for the CD spectra (Fig. 2B). The EPR spectrum of the air-oxidized T14C FdI exhibits a g ϭ 2.01 signal which integrates to 1.0 spin per molecule (Fig. 3C). This signal shows a slight difference in the shape from that of native FdI (Fig. 3C) which is also evident in the whole cell EPR data of Fig. 5. The EPR spec-trum of dithionite-reduced T14C FdI, like that of the native FdI, is EPR silent (data not shown).
Direct electrochemical experiments were then performed to characterize further the 7Fe form of T14C. Fig. 7 shows cyclic voltammograms obtained for the 7Fe form. Bulk solution voltammetry measured at pH 7.4 and 0°C reveals two well defined reversible couples A and B with reduction potentials Ϫ405 and Ϫ611 mV, respectively. Corresponding peak separations at 5 mV s Ϫ1 are 47 and 51 mV, and plots of peak currents against (scan rate) 1/2 were linear up to 20 mV s Ϫ1 , as expected for a diffusion-controlled electrode reaction, although the peak separations are smaller than the expected theoretical value (54 mV at 0°C) indicating that the response arises at least in part from some more tightly adsorbed protein. Since it is reducible by dithionite, and based upon assignments established for native FdI, couple A is assigned to the [3Fe-4S] ϩ/0 transition (37). Likewise, by analogy with native FdI, couple B is assigned to the [4Fe-4S] 2ϩ/ϩ cluster. A third redox couple, E 0 Ј (estimated) ϭ Ϫ719 mV, can be seen at the negative limit, beyond which background proton reduction interferes. The film voltammograms shown in Fig. 7 for pH 8.0 have a form that is very similar to that observed for native AvFdI, with three signals AЈ, BЈ, and CЈ, which are assigned to [3Fe-4S] ϩ/0 (Ϫ426 mV), [4Fe-4S] 2ϩ/ϩ (Ϫ633 mV), and [3Fe-4S] 0/2Ϫ (Ϫ756 mV), respectively. The characteristically sharp signal CЈ has peak areas approximately twice that of AЈ or BЈ (65) and confirms the presence of the [3Fe-4S] cluster. We note that unlike the situation discussed above for the 8Fe form of T14C, the indigenous [4Fe-4S] 2ϩ/ϩ cluster of the 7Fe form has the same reduction potential as in the native protein.
Protonation of the [3Fe-4S] 0 cluster is now a well established property of AvFdI and other proteins (37, 65-67) and for that reason the film voltammetry was measured over a wide pH range, with the results shown in Fig. 8. The curved dependence displayed by signal AЈ gave an excellent fit to Equation 1 (68). This describes the binding of a single proton to the "0" but not the "ϩ" oxidation level of the [3Fe-4S] cluster, and for which E alk is the limiting reduction potential at pH values sufficiently above pK red .
The fit gave pK red ϭ 8.4 with a limiting alkaline region reduction potential (E alk ) of Ϫ464 Ϯ 10 mV. In the acid region, the slope is Ϫ54 mV. Signal CЈ shows linearity throughout the entire range, and the gradient of Ϫ54 mV/pH unit is again consistent with the uptake of one proton per electron (i.e. two protons for the 0/2Ϫ transition). Signal BЈ (the [4Fe-4S] 2ϩ/ϩ cluster) is partly hidden by the more prominent CЈ couple below pH 8.0; however, little variation of reduction potential is evident over the limited range of pH able to be studied. As is generally seen, the reduction potentials measured for the bulk solution sample lie within 20 mV of those measured for films under comparable conditions of pH. The protonation of the [3Fe-4S] 0 cluster can easily be observed because the protonated and deprotonated forms of reduced FdI exhibit different CD spectra (37,38). To confirm that the pK red for the cluster in T14C FdI had increased from 7.8 (37) to 8.4, we therefore monitored the pH dependence of the CD spectra of dithionite-reduced 7Fe T14C FdI (Fig. 9). The spectra of the T14C FdI samples reduced at pH 6.0 and pH 9.0 are indistinguishable from those of the native FdI obtained at pH 6.0 and pH 8.3, respectively (38), which represent the proteins containing the protonated (acidic) and the deprotonated (alkaline) [3Fe-4S] 0 cluster, respectively. The CD spectrum of T14C FdI at pH 8.3 has mixed features of those spectra obtained at pH 6.0 and 9.0 (Fig. 9), consistent with the [3Fe-4S] cluster being about half protonated at around its pK a .
Taken together, the above data show that the form of T14C FdI observed in vivo (Fig. 5) can be purified to give a 7Fe protein that is spectroscopically and electrochemically similar to native FdI. Both proteins are also completely air-stable, and the purification yields for the 7Fe form of T14C FdI are the same as for the native protein and much greater than for the 8Fe form of T14C FdI.
Crystal Structure of 7Fe T14C FdI-To date we have been unable to crystallize the 8Fe form of T14C FdI, whereas crystals of the 7Fe form were readily obtained. The crystal structure of T14C FdI in the oxidized 7Fe form has been refined at 2.10-Å resolution ( Table I). The structure is isomorphous with native FdI (31) with a root mean square deviation of only 0.18 Å for 526 pairs on common nitrogen, C␣, carbon, oxygen, and C␤ atoms. The portion of the structure including the loop of residues 8 -16 in the native and T14C FdI structures is shown in Fig. 10. The conformation of the two loops is virtually identical, and the mutated Cys-14 side chain is oriented such that the S␥ atom occupies the position of the methyl group of Thr-14. Consequently, a 4.1-Å contact of the methyl group to a divalent inorganic sulfur atom of the [3Fe-4S] cluster is preserved by the Cys-14 S␥ atom in T14C FdI. At the same time, hydrophobic contacts of the Thr-14 methyl group and Cys-14 S␥ atom to alipathic atoms of Lys-12, Leu-32, Ile-81, and Lys-85 are very similar. These side chains and neighboring Lys-84 and Pro-87 residues also have very similar conformations in the two structures. A tightly bound H 2 O molecule conserved in both structures is 3.8 Å from S␥ of Cys-14. Therefore, the effect of the T14C replacement is to substitute the S␥ atom for a methyl group with loss of the Thr hydroxyl group but with very little other effect on the structure (Fig. 10) consistent with the observed very small changes in spectroscopic and electrochemical properties of the protein.
3Fe to 4Fe Cluster Conversion-When wild-type A. vinelandii or strains overexpressing native FdI are ruptured anaerobically in the presence of dithionite, the [3Fe-4S] ϩ cluster is reduced, but it is not converted to a [4Fe-4S] 2ϩ cluster. In contrast to that result the data shown above demonstrate that when cells overexpressing a 7Fe form of T14C FdI (Fig. 5) are ruptured in the presence of dithionite, the [3Fe-4S] ϩ cluster is reduced and converted to a [4Fe-4S] 2ϩ cluster. This suggested that we should be able to take the stable purified 7Fe form of T14C FdI and easily convert it to an 8Fe form in vitro by anaerobic addition of dithionite and iron, a reaction that does not occur under the same conditions for native FdI (33). As determined by UV-visible absorption and CD spectra (data not shown), T14C FdI can be easily converted (within 30 min) to give an 8Fe form of the protein that is indistinguishable from the form purified anaerobically in the presence of dithionite whose characterization is described above. Fig. 11 shows an additional assay we have developed to monitor the cluster conversion reaction, which should now be applicable to other systems. The data show that the 7Fe form of T14C FdI runs in the same place as the native protein, thus the mutation of Thr to Cys does not by itself influence the behavior of the protein during anaerobic native gel electrophoresis. The 8Fe form of T14C FdI, however, runs in a very different place (Fig. 11) from the 7Fe form of the protein. Furthermore, the 8Fe forms produced by purification in the presence of dithionite or by addition of dithionite and iron to the purified 7Fe protein are indistinguishable (Fig. 11).
Previous studies have shown that the reactivity of [3Fe-4S] 0 clusters with regard to incorporation of a fourth iron to produce a [4Fe-4S] cluster can be studied using direct electrochemical methods. Incorporation of iron or other metals into the fourth site occurs readily at the electrode surface for some [3Fe-4S]containing proteins, particularly the labile ferredoxin III from D. africanus. While uptake of divalent metal ions is restricted to the 0 oxidation level state, Tl(I) binds very rapidly and to both [3Fe-4S] ϩ and [3Fe-4S] 0 forms (69 -71). An electrode coated with a film of 7Fe T14C FdI was therefore transferred to a pot containing buffer/electrolyte solutions and Fe(II), Zn(II), or Tl(I). To hold the [3Fe-4S] cluster at the 0 level, the electrode potential was poised at Ϫ507 mV for iron and zinc or at Ϫ482 mV for Tl, for various times depending on film stability. The most extreme conditions used were for iron, with a poising time of 1 h and an iron concentration of 5 mM. In all cases no new signal was observed attributable to a new [M-3Fe-4S] cluster. Thus, the conversion reaction that so readily occurs in solution with the 7Fe form of T14C FdI is somehow prevented from occurring when the protein is bound to the electrode.
Probable Ligands to the New [4Fe-4S] 2ϩ/ϩ Cluster-At present there is no structure available for the 8Fe form of T14C FdI, and we are therefore left with three possibilities for the ligand arrangement for the new 4Fe cluster in the 8Fe T14C FdI. The first would require that the cluster ignore what is normally a 3Fe cluster ligand, Cys-16, and use instead Cys-11 (a free Cys in native FdI) and the introduced Cys-14 along with Cys-8 and Cys-49 (Fig. 1). This possibility seems unlikely to us for a number of reasons. For example, the very large structural rearrangement required would seem unnecessary based on the available structure of the 7Fe form of T14C FdI (Fig. 10) and would appear to be at least as difficult as the rearrangement required for native FdI that does not undergo facile cluster conversion. This possibility has the advantage of introducing a Cys-X-X-Cys-X-X-Cys ligand motif that should have produced a [4Fe-4S] 2ϩ/ϩ cluster in vivo (36). In contrast, our whole cell EPR studies (Fig. 5) show that in vivo the cluster conversion does not occur.
A second possibility would involve the use of a solvent molecule as the new ligand. The [3Fe-4S] to [4Fe-4S] cluster conversion that occurs in both crystals and in solution in mitochondrial aconitase (72) does not involve a fourth cysteine but rather a solvent molecule that acts as the ligand to the introduced iron atom. We cannot rule out the possibility that a solvent molecule acts as the fourth ligand to the converted [4Fe-4S] cluster in 8Fe T14C FdI, although we consider that possibility to be unlikely in part because earlier studies have already shown that, when given a choice, the cluster has a strong preference for sulfur over oxygen ligation (60). It also does not appear very logical that a mutation that introduces a  10. The structure of a Cys-X-X-Cys-X-X-Thr-X-Cys loop near the [3Fe-4S] cluster in native AvFdI (FdI) and the corresponding Cys-X-X-Cys-X-X-Cys-X-Cys loop in the T14C FdI variant (T14C FdI). The [3Fe-4S] clusters in both proteins are ligated by Cys-8, -16, and -49 (not shown). The distances (Å) between the Thr-14-C␥ and the nearest cluster sulfide, and between the Cys-14-S␥ and the nearest cluster sulfide are indicated by the dotted line. Carbon, nitrogen, oxygen, sulfur, and iron atoms are shaded white, light gray, gray, dark gray, and black, respectively. new Cys would lead to the formation of a 4Fe cluster that does not use the introduced residue. Based on the 7Fe structure of T14C FdI (Fig. 10), a water or hydroxyl oxygen coordinated to the fourth iron in a tetrahedral manner with an Fe-O bond of 2.0 Å would have a severe steric clash with Tyr-13 nitrogen and C␣ atoms (contacts of 1.0 -1.1 Å). Therefore, a non-Cys-14 ligand to the new [4Fe-4S] cluster also would require significant conformational change in the T14C FdI protein. Furthermore, [4Fe-4S] 2ϩ/ϩ clusters with one noncysteine ligand can be easily converted back to [3Fe-4S] 2ϩ/ϩ clusters (13)(14)(15)(16)(17)(18)(19)(20)(21). In contrast, with the exception of DgFdII (10,11), [4Fe-4S] 2ϩ/ϩ clusters with four cysteine ligands have not been reported to convert easily back to [3Fe-4S] ϩ/0 clusters (18,(22)(23)(24)(25)(26)(27). As examined by EPR, the 8Fe form of T14C FdI behaves like these proteins, exposure to oxygen or anaerobic oxidation with ferricyanide leads to small amounts of [3Fe-4S] ϩ/0 , but the reaction is destructive and rapidly leads to apoprotein.
The third possibility that we currently feel is most probable is that the 4Fe cluster uses Cys-8, -14, 16, and -49 as ligands. The isomorphism of the T14C (Fig. 10) and native 7Fe FdI structures requires that in order for the T14C FdI [3Fe-4S] cluster to be converted to a [4Fe-4S] cluster using these ligands, a conformational change must still occur. If Cys-14 were to ligate a newly incorporated fourth iron atom as in other [4Fe-4S] clusters, the main chain torsion angles would have to change significantly (ϳ90°), and the C␣ atom would have to be displaced approximately 2 Å. This change in the main chain T14C could be translated to the region of Cys-20, a ligand to the indigenous [4Fe-4S] 2ϩ/ϩ cluster, thus explaining its change in reduction potential. The distance between Cys-14 S␥ and the new fourth iron atom position is about 5.2 Å. The failure of the conversion to occur in crystals of T14C FdI soaked in sodium dithionite and ferrous ammonium sulfate (see "Experimental Procedures") must be due to restraints of the crystal lattice and/or the fact that residues 11-14 are packed against residues 84 -88 on the surface of the protein. In contrast, the rearrangement that does occur in the C20A and C20S mutants of FdI (58,60), allowing Cys-24 to be a new ligand to the [4Fe-4S] cluster, involves residues in an exposed surface loop.
The only naturally occurring example of a stable [3Fe-4S] ϩ/0 cluster containing protein that can be converted to a [4Fe-4S] 2ϩ/ϩ cluster with four cysteine ligands in vitro is DgFdII, a reaction that appears to occur in vivo and is complicated by accompanying changes in subunit composition (10,11). All other examples of 3Fe to 4Fe cluster conversions involve formation of a 4Fe cluster by recruiting one noncysteine ligand (13)(14)(15)(16)(17)(18)(19)(20)(21). Data shown above strongly suggest the resemblance of the new [4Fe-4S] cluster in T14C FdI with the all-cysteine ligand [4Fe-4S] clusters in other ferredoxins. This would make T14C FdI the first example of a stable [3Fe-4S] ϩ/0 cluster containing protein that can be easily converted in vitro to a [4Fe-4S] 2ϩ/ϩ cluster by using an extra cysteine residue. This reaction is not accompanied by a change in subunit composition as both the 7Fe and 8Fe forms are monomers. The isolation of large quantities of an air-stable 7Fe form of T14C FdI should therefore enable future investigation of the factors that control both the 4Fe to 3Fe cluster conversion reaction and the extreme oxygen sensitivity of the new 4Fe cluster.
It is finally interesting to note that little is known about how [Fe-S] clusters are assembled in vivo. T14C FdI provides a very clear example of a protein that can easily be converted to an 8Fe form in vitro and yet remains as an 7Fe form in vivo. The cellular mechanism for assembling a [4Fe-4S] 2ϩ/ϩ cluster must therefore be much more specific than the in vitro self-assembly reaction. As discussed above and considered in Figs. 1 and 10, the new 4Fe cluster in T14C FdI is likely to receive three of its ligands from a Cys 8 -X-X-X-X-X-Cys 14 -X-Cys 16 sequence which is quite unlike the Cys-X-X-Cys-X-X-Cys motif that is often used to ligate [4Fe-4S] 2ϩ/ϩ clusters in native ferredoxins. It is therefore possible that the in vivo specificity relies both on recognition of that sequence and the context of that sequence.