Complex Formation between Azotobacter vinelandiiFerredoxin I and Its Physiological Electron Donor NADPH-Ferredoxin Reductase*

In Azotobacter vinelandii, deletion of the fdxA gene, which encodes ferredoxin I (FdI), leads to activation of the expression of the fpr gene, which encodes NADPH-ferredoxin reductase (FPR). In order to investigate the relationship of these two proteins further, the interactions of the two purified proteins have been examined. AvFdI forms a specific 1:1 cross-linked complex with AvFPR through ionic interactions formed between the Lys residues of FPR and Asp/Glu residues of FdI. The Lys in FPR has been identified as Lys258, a residue that forms a salt bridge with one of the phosphate oxygens of FAD in the absence of FdI. UV-Vis and circular dichroism data show that on binding FdI, the spectrum of the FPR flavin is hyperchromatic and red-shifted, confirming the interaction region close to the FAD. Cytochrome c reductase assays and electron paramagnetic resonance data show that electron transfer between the two proteins is pH-dependent and that the [3Fe-4S]+ cluster of FdI is specifically reduced by NADPH via FPR, suggesting that the [3Fe-4S] cluster is near FAD in the complex. To further investigate the FPR:FdI interaction, the electrostatic potentials for each protein were calculated. Strongly negative regions around the [3Fe-4S] cluster of FdI are electrostatically complementary with a strongly positive region overlaying the FAD of FPR, centered on Lys258. These proposed interactions of FdI with FPR are consistent with cross-linking, peptide mapping, spectroscopic, and electron transfer data and strongly support the suggestion that the two proteins are physiological redox partners.

chain folds around these two clusters in a manner similar to most eight-iron ferredoxins that contain two [4Fe-4S] 2ϩ/ϩ clusters. This NH 2 -terminal region therefore represents a core structural frame shared among different classes of ferredoxins (7). However, AvFdI contains an additional ϳ60-amino acid extension at the COOH terminus that is unique to this class of seven-iron ferredoxins. Proteins homologous to AvFdI have been reported from a number of organisms including Azotobacter chroococcum, Pseudomonas stutzeri, Pseudomonas ovalis, Pseudomonas aeruginosa, Caulobacter crescentus, Rhodobacter capsulatus, Streptomyces griseus, Thermus thermophilus, and Bacillus schlegelii with the pseudomonas protein sequences showing Ͼ90% identity to AvFdI (8 -13). Although it has been known for some time that FdI has a metabolic function unrelated to nitrogen fixation that is important for cell growth (14), the specific cellular function of FdI and related seven-iron ferredoxins in other organisms has yet to be determined.
In 1988, it was reported that disruption of the fdxA gene that encodes FdI in A. vinelandii led to a dramatic increase in the levels of another small acidic protein (13). In order to investigate the mechanism of regulation by FdI, the small acidic protein that is overexpressed in response to fdxA deletion was purified and characterized (15), and the gene encoding the protein was cloned and sequenced (16). The protein was shown to be a M r ϳ29,000 NADPH-ferredoxin reductase that was designated FPR because its physical properties and amino acid sequence showed striking similarity to the FPR from Escherichia coli (15,17,18). In E. coli, the protein appears to be part of a system that activates at least three different enzymes involved in anaerobic metabolism: anaerobic ribonucleotide reductase (19), pyruvate formate-lyase (20), and cobalamin-dependent methionine synthase (21). These activation reactions not only require NADPH and FPR but also ferredoxin or flavodoxin and S-adenosylmethionine. In those systems, FPR functions by mediating electron transfer between NADPH and the ferredoxin or flavodoxin (22). In E. coli, the fpr gene is activated in response to oxidative stress by the SoxRS regulon (23). In A. vinelandii, the fpr gene is similarly activated in response to oxidative stress; however, in that case the system is very specific for AvFdI (24,25), a ferredoxin type that is not found in E. coli.
Here, we have examined the possibility that in A. vinelandii both the metabolic and regulatory functions of FdI involve direct and highly specific interactions between FdI and FPR. As confirmed by both chemical cross-linking and spectroscopic characterization, AvFdI forms a complex with AvFPR at the stoichiometric ratio of 1:1 and the [3Fe-4S] ϩ cluster of AvFdI is specifically reduced by NADPH via FPR. Recently, the x-ray structures of both E. coli and A. vinelandii FPR have been determined (26,27). Taken together, these data are used to model the interaction sites between the two proteins in A. vinelandii. (28) and FPR (15) were purified to homogeneity, and FdI strain LM100 was constructed (14) as described elsewhere. FdI concentrations were estimated from the absorbance at 405 nm, using an extinction coefficient of 29.8 mM Ϫ1 cm Ϫ1 . FPR concentrations were determined by the method of Bradford (29), using bovine serum albumin as a standard. All chemicals used in the protein modifications were obtained from Sigma. The chemiluminescence kit used in Western blot analysis was purchased from DuPont or Amersham Pharmacia Biotech. Sequencing grade porcine trypsin, modified by reductive methylation to reduce autolysis of the enzyme and thereby increase its stability, was purchased from Promega.

Materials-FdI
Cross-linking Experiments-For these experiments, 60 M FdI was treated in the presence of 20 M FPR with 5 mM EDC in 25 mM MOPS, pH 7.2. The reaction mixture was incubated for 2 h at room temperature followed by quenching the reaction with the addition of the solubilization buffer used for SDS-PAGE. SDS-PAGE (12% acrylamide) was performed in 80 mM Tris/glycine, pH 8.8. The gel was then stained with 0.1% Coomassie Blue in a fixative solution (40% methanol and 10% acetic acid), or the proteins were electrotransferred to nitrocellulose membranes. Antigen-antibody interaction was visualized using horseradish peroxidase-conjugated secondary antibody.
Activity Assays-FdI dependent NADPH-cytochrome c reductase activity of FPR was performed at room temperature under strictly anaerobic conditions using both an anaerobic glove box and a glucose (25 mM)/glucose oxidase (0.5 unit) system. The reaction mixture consisted of 50 M cytochrome c, 2.5 M FdI, 0.2 mM NADPH, and 3 M FPR. The reaction was initiated by the addition of 0.2 mM NADPH, and the increase in absorbance at 550 nm was measured. Acetylene reduction assays for nitrogenase were performed as described previously (30) under argon in 100 mM TES, pH 7.4, or 100 mM MES, pH 6.5, with 20 mM sodium dithionite or NADPH-FPR as a reductant.
Chemical Modification of FdI and FPR-Pyridoxal 5Ј-phosphate followed by NaBH 4 reduction was used to modify lysine residues in both proteins. 5 M FPR or 5 M FdI was incubated in the dark in a reaction mixture containing 25 mM HEPES, pH 8.0, 5% glycerol, 0.1 mM EDTA, and 20 mM pyridoxal 5Ј-phosphate for 30 min at room temperature. The pyridoxal 5Ј-phosphate was then neutralized by addition of 40 mM fresh NaBH 4 . Unreacted chemicals were then removed using small Sephadex G-25 columns (0.9 ϫ 15 cm). Phenylglyoxal was used for modification of arginine residues. 5 M FPR or 5 M FdI was incubated in the dark with 20 mM phenylglyoxal for 2 h at room temperature and made free of unreacted chemicals using Sephadex G-25 columns (0.9 ϫ 15 cm). 10 mM EDC for modification of glutamate or aspartate residues was used in 50 mM HEPES, pH 7.5, containing 100 mM NaCl at a 5 M protein concentration. After a 20-min incubation, the reaction was terminated with 30 mM ethanolamine.
Spectroscopy-EPR spectra were obtained using a Bruker ESP300E spectrometer, interfaced with an Oxford liquid helium cryostat. Absorptional spectra were recorded in 0.5-ml quartz cuvettes on a Hewlett Packard 8452A diode array spectrophotometer. CD spectra were obtained using a Jasco J-500C spectropolarimeter. CD measurements were carried out using small volume cylindrical cells with fused quartz windows. For anaerobic EPR measurements, the samples were loaded into the EPR tubes in an O 2 -free (O 2 Ͻ 1ppm) glove box (Vacuum Atmosphere, Hawthorne, CA). The tubes were sealed with a cap, wrapped with parafilm, and then transferred to the liquid nitrogen Dewar.
Proteolytic Cleavage, Peptide Mapping, and Amino Acid Sequences-Apoprotein obtained by 10% trichloroacetic acid was dissolved in 7.6 M urea containing 0.4 M ammonium bicarbonate phosphate, pH 8.0, and 2 mM EDTA and reduced by addition of a 5-fold molar excess of dithiothreitol per mol of SH, under argon at 50°C for 15 min. The alkylation of the thiol groups was performed by incubation for 30 min with 10 mM iodoacetamide at room temperature. After adjusting the urea concentration of sample solution to Ͻ2 M with distilled water, tryptic digestion of carboxylmethylated apoprotein was carried out by addition of trypsin (10% w/w) at 37°C for 20 h. Peptides were separated by reverse phase high performance liquid chromatography. An aliquot containing 1-30 nmol of digested protein was resolved with a HP1090 system with a filter photometer equipped with a Vydac C-18 column (internal diameter, 0.45 mm). Peptides separations utilized a solvent system consisting of aqueous 0.1% trifluoroacetic acid (solvent A) and 90% acetonitrile/0.09% trifluoroacetic acid (solvent B), with the linear gradient changing 1% per min from 5 to 100% solvent B (31, 32). The NH 2terminal protein sequencing was performed by the Biotechnology Resource Facility of the University of California, Irvine, on a Perkin-Elmer Applied Biosystems Division model 477A automated protein sequencer with the on-line model 120A phenylthiohydantoin-amino acid analyzer equipped with strip chart recorder and 610A data analysis system.
Electrostatic Calculations-The electrostatic potentials for FdI and FPR were calculated with the program UHBD (33), which uses finite difference methods to solve the Poisson-Boltzmann equations for the electrostatic potential on a three-dimensional grid. Coordinates for FdI (Protein Data Bank entry 6FD1 (5)) include all hydrogen atoms except for His and Cys residues. All nonpolar hydrogen atoms were removed. Polar hydrogen atoms not present in the crystallographic coordinates were added using the molecular graphics program Insight (MSI Inc., San Diego, CA). Hydrogen atoms were attached to the sulfur atoms of the nonligating Cys residues 11 and 24 and to the N⑀ atoms of His residues 35 and 103. N␦ atoms for His 35 (34).
Polar hydrogens were also included in the crystallographic coordinates of FPR (Protein Data Bank entry 1A8P (27)). Except for His 109 , His side chains were modeled with only N⑀ being protonated, giving them an overall neutral charge. For His 109 , both imidazole side chain N atoms were protonated; His 109 is solvent exposed and forms a salt bridge to an Asp side chain through N␦. Polar protons were added to model a fully reduced FAD hydroquinone. Partial charges for FAD atoms were assigned to be consistent with the AMBER charges for DNA and RNA. All other partial charges for FdI and FPR were assigned using the library of AMBER partial charges provided with UHBD.
Electrostatic potentials were calculated on a grid with 1.0-Å spacing in a medium equivalent to 150 mM ionic strength. The electrostatic potential mapped onto the molecular surfaces (see Fig. 12) is the value of potential evaluated at a series of points 1.4 Å above and normal to the surfaces. Molecular surfaces were calculated with the program MSMS (35) using a 1.4-Å probe sphere, which is the van der Waals radius of a water molecule.

Interaction of FdI and FPR with the Chemical Cross-linking
Reagent EDC-While working out the purification scheme for FPR, we prepared an affinity column that had purified FdI attached to a resin via lysine residues (15). In 0.1 M NaCl, FPR appeared to bind specifically to that column, giving us the first indication that these two proteins might be redox partners in vivo. To characterize this interaction further, we undertook a study using the chemical cross-linking reagent EDC, which links lysine residues in one protein with glutamate or aspartate residues in the other (36). To begin this study, we examined the behavior of each of these proteins by itself with EDC.
As shown in Fig. 1A, FdI, which has an isoelectric point of 4.03, migrates abnormally on SDS-PAGE as if its molecular weight were 26,000 rather than 12,050 (37), a phenomenon that also occurs for other highly acidic proteins (38,39). Addition of EDC to FdI causes modification of the carboxylate groups, resulting in the protein migrating as expected based on its molecular weight. As also shown in Fig. 1A, the FPR monomer does not show any abnormality on SDS-PAGE either before or after treatment of the protein with EDC with respect to its molecular weight position. However, Coomassie-stained gels always also revealed a small amount of higher molecular weight complexes. The molecular weights of these complexes suggest that they corresponded to mostly FPR homodimer, possibly with different orientations, yielding complexes that run is somewhat different positions on the gel. Fig. 1B shows a corresponding Western blot, which both confirms that the high molecular weight complexes react with antibodies raised against FPR and shows that very small amounts of FdI dimer (not observed on the Coomassie-stained gel) are also formed by reaction of FdI alone with EDC. By varying conditions, we found that the amount of FdI dimer could be increased to the point where it could be easily observed on Coomassie-stained gels by lowering the pH (Fig. 2). The electrostatic calculations that are discussed below offer a clear explanation for this observation because the extreme negative charge of FdI would have to be neutralized by protonation of surface residues prior to dimer formation. It should also be noted that we never observed complexes greater than FdI dimers.
The formation of FdI dimers could be an artifact of nonspecific cross-linking or it could be physiologically relevant. Although in the crystalline state AvFdI is clearly a monomer (2)(3)(4)(5), previous studies have suggested that the homologous seven-iron ferredoxins from S. griseus (40), P. ovalis (41), and Sulfolobus acidocaldarius (42) might form dimers under certain conditions in solution. In general, the more physiological conditions of low salt (40) and, as shown in Fig. 2, lower pH tend to favor the formation of dimers. In addition, FdI does have a regulatory function, and it has been proposed that this function might involve dimer formation (43). None of these data are definitive, but they do leave open the possibility that either the metabolic or regulatory functions of FdI may involve dimer formation.
Formation of a 1:1 FdI:FPR Complex Using EDC-As shown in Fig. 3A, when FdI and FPR are incubated together with EDC, a new major band having an apparent M r ϳ43,000 appeared, concurrent with a reduction in the bands corresponding to FdI alone or FPR alone. The molecular weight of the band is consistent with a 1:1 complex of the two proteins, and Western analysis ( Fig. 3B) using antibodies raised against the individual FPR and FdI proteins confirmed that both are present in the complex. The efficiency of cross-linking was strongly dependent upon the ionic strength, pH, and the molar ratio of FdI to FPR. The amount of cross-linked product decreased as either [NaCl] or pH increased. Ionic strength less than 0.2 M, pH lower than 7.0, and a molar ratio of FdI to FPR greater than 3 are optimal conditions for complex formation (data not shown). The ionic strength and pH dependences of the interaction between the two proteins points to a strong contribution of electrostatic forces in the early stages of complex formation. The same efficiency in the cross-linking reaction was observed when the reaction was carried out anaerobically in the presence of 2 mM Na 2 S 2 O 4 . Under those conditions, the [3Fe-4S] ϩ/0 cluster of FdI is reduced, and the FAD of FPR is in the reduced hydroquinone state. This result shows that the formation of a 1:1 complex does not depend upon the oxidation state of the chromophores. The absence of any additional reaction products indicates that the cross-linking of FdI with FPR is highly specific. In addition, no cross-linked products were obtained when either FdI or FPR was heat-denatured prior to addition of EDC, showing that the reaction required both partners to be in their native conformations.
Identification of FPR as Lysine and FdI as Glutamate/Aspartate Donor-EDC is a water-soluble, zero-length carbodiimide cross-linker that covalently links salt bridges formed between carboxylate groups from glutamate or aspartate residues in one protein to amino groups from lysine residues in the other (36). The pK a values of arginine residues are too high to form a stable amino bond, so arginines are generally not present in the cross-linked complexes. The absence of a spacer in the zero-length cross-linker also increases the probability that the residues that are involved in the cross-linked product represent the actual sites of protein-protein interaction. In this case, FPR has 11 lysine and 33 aspartate and glutamate residues, whereas FdI has 6 lysine and 23 aspartate and glutamate residues, so our first experiments were directed toward determining which protein was the lysine donor and which was the glutamate or aspartate donor. To obtain this information, we did a series of experiments in which we chemically modified either the lysine or the glutamate/aspartate residues in FdI and/or FPR prior to attempting to form the cross-linked complex with EDC.
As shown in Fig. 4, when lysine residues from FPR and aspartate/glutamate residues from FdI were modified, the complex formation was greatly reduced. On the other hand, when aspartate/glutamate residues from FPR and lysine residues from FdI were modified, the complex formation is indistinguishable from the control (Fig. 4). Complex formation was also greatly impaired with only lysine modified FPR but was not affected with arginine modified FPR (data not shown). These results show that the lysine from FPR and glutamate/aspartate from FdI form a salt bridge at the interface between the two proteins. This result is also consistent with our previous observation that when FdI was cross-linked via lysines to a resin, a reaction that would leave its glutamate/aspartate residues available, FPR bound tightly to the affinity column (15). We have now tried the reciprocal experiment and find that if FPR is cross-linked to a resin via its lysine residues, FdI does not bind to that affinity column, which also supports our conclusion that in the cross-linked complex FPR is the lysine donor and FdI is the glutamate/aspartate donor.
Identification of Lys 258 as the Lysine Donated by FPR-In order to identify the cross-linking site, the FPR:FdI crosslinked complex was purified away from the FPR homodimer using FPLC. Then, EDC-treated FPR alone, EDC-treated FdI alone, and the cross-linked complex were individually S-cysteine-alkylated and digested with trypsin, and the resulting peptides were separated by high performance liquid chromatography. By comparing the high performance liquid chromatography profiles for the three samples, we isolated a peak that is observed in the cross-linked product and FdI alone, but not in FPR alone. The NH 2 -terminal sequence of that peak showed that two sets of residues were found for each cycle, one from FPR and the other from FdI. FVEX matched those of the known FPR peptide 254 AFVEK 258 . The Lys 258 was not observed in the sequencing reaction, which would be the expected result if that lysine were involved in the cross-linked complex. No peptide ran close to the same position in the profile of trypsin digested FPR alone and the FPR peptide with the same sequence eluted much earlier in the chromatogram when the tryptic peptides of FPR were run under the same conditions. This result is also consistent with Lys 258 being the cross-linked residue in FPR. The other set of residues matched those from a known very large tryptic peptide of FdI, 13 YTDCVEVCPV . . . K 85 . NH 2terminal sequencing of that fragment was successful through residue Glu 48 when all aspartates or glutamates were observed, consistent with the FdI site being contained within the peptide Cys 49 -Lys 85 . This site was not further defined.
FPR Cross-linked Residue Lys 258 Is Salt-bridged to FAD in the Absence of FdI-As shown in Fig. 5, in the crystal structure of AvFPR (refined at a 2.0-Å resolution) the COOH-terminal residue, Lys 258 interacts via a hydrogen bond (3.2 Å) with a phosphate oxygen of the AMP portion of the FAD (27). This interaction is unique among homologous FAD containing enzymes of known structure. In the most closely related protein, E. coli flavodoxin reductase (the E. coli fpr gene product), the COOH-terminal residue, Trp 248 , stacks on the adenine ring of FAD (26). In AvFPR, the side chain of Phe 255 stacks on the adenine; consequently, the conformation of the bound FAD is FIG. 4. Cross-linking between chemically modified FdI and chemically modified FPR. Pyridoxal 5Ј-phosphate was used for modification of lysine residues, phenylglyoxal for arginine residues, and EDC for glutamate or aspartate residues as described under "Experimental Procedures." Unreacted excess chemicals were then removed using small Sephadex G-25 columns (0.7 ϫ 15 cm). The cross-linking was carried out using the conditions described in Fig. 3, and the samples were analyzed by 12% SDS-PAGE with Coomassie staining. Lane a, unmodified FPR and unmodified FdI were incubated with 5 mM EDC; lane b, lysine-modified FPR and aspartate-or glutamate-modified FdI were incubated with 5 mM EDC; lane c, aspartate-or glutamate-modified FPR and lysine-modified FdI were incubated with 5 mM EDC. FdI Binding Perturbs the Spectrum of the Flavin-A model that is consistent with the data presented above is that FdI binding to FPR disrupts the interaction of cross-linking residue Lys 258 with the FAD. Independent evidence for FdI binding perturbing the environment of the flavin is shown in Fig. 6. In this experiment, UV-visible spectrophotometry was used to investigate possible changes in the FAD chromophore upon FdI binding. The absorption spectrum of FdI is featureless in the 350 -600 nm region (Fig. 6A), whereas FPR exhibits the strong characteristic flavin spectrum in this region (Fig. 6B). The spectrum in Fig. 6C is that of a physical 1:1 mixture of the two proteins, not of a cross-linked complex. The optical difference spectrum shown in Fig. 6D was obtained by computer subtraction of the sum of the spectra exhibited by the individual proteins (Fig. 6, A and B) from the physical mixture shown in Fig. 6C. This difference spectrum is very similar in shape to that exhibited by FPR, but it is red-shifted, which indicates that when FdI binds to FPR in solution, it appears to perturb the spectrum of the flavin. No difference spectrum was observed when FPR was incubated with bovine serum albumin for a control experiment. As shown in Fig. 7 this reaction was easily quantitated by varying the ratio of FdI to FPR and monitoring the absorbance change at 498 nm, the wavelength at which the total change was largest. The absorbance increased with increasing ratio of [FdI] to [FPR] up to a maximum when the two proteins were in equimolar concentration. This result again indicates that FdI forms a specific 1:1 com-plex with FPR in solution, strengthening the conclusion from the cross-linking experiments and extending the result to show that the flavin environment is altered when FdI binds. Assuming ⌬A 498 comes from the FAD signal of the complex and that its M r is ϳ43,000, the value of ⌬⑀ 498 was calculated to be 1.1 ϫ 10 3 M Ϫ1 cm Ϫ1 .
As shown in Fig. 8, in attempts to see whether the [Fe-S] clusters of FdI were also perturbed upon complex formation, we employed CD spectroscopy because the visible region CD spectrum of oxidized FdI is very unique to this class of ferredoxins. Fig. 8, A and B, shows the CD spectra for oxidized FdI and FPR respectively. The oxidized FdI has strong positive CD bands above 500 nm, but oxidized FPR does not. Furthermore, even at wavelengths shorter than 500 nm, the two CD spectra were easily distinguishable because FdI has two strong positive CD bands in this region, whereas FPR has a broad negative band. The same strategy was used to obtain the difference spectrum shown in Fig. 8D as was used for the UV-visible experiment. Again, the difference spectrum is similar to the FPR CD spectrum, indicating that the FAD environment is perturbed upon FdI binding but giving no indication of whether the FdI chromophores are perturbed. Finally, as shown in Fig. 9, the EPR spectrum of FdI that arises from the [3Fe-4S] ϩ cluster is also not perturbed by binding to FPR. This negative result may indicate that the FdI binding site is distant from the FdI chromophore or simply that the EPR spectral properties are dominated by the structure of this buried cluster and are not influenced by the formation of salt bridges between residues on the surface of the two proteins.
Electron Transfer from NADPH to FPR to FdI-In the last step in photosynthesis, proteins that are somewhat homologous to FPR receive reducing equivalents from low reduction potential [2Fe-2S] containing ferredoxins and transfer them to NADP ϩ to form NADPH (44). On the other hand, for E. coli FPR, which is much more similar to AvFPR, the reaction appears to go in the reverse direction, with electrons transferred from NADPH to FPR and onto ferredoxins or flavodoxins, al- though in that case, the specific ferredoxin or flavodoxin has not been identified (23). The reduction potentials of NADPH and of the FAD in AvFPR are reported to be similar, at approximately -320 mV versus the standard hydrogen electrode (8,15,45), making the reduction of FPR very sensitive to the ratio of NADPH/NADP ϩ .
In this study, we used two methods to measure electron transfer from NADPH to FPR to FdI. The reduction potential of the [4Fe-4S] 2ϩ/ϩ cluster of FdI is far too negative at -650 mV versus standard hydrogen electrode to be reduced by this system (8). The reduction potential of the [3Fe-4S] ϩ/0 cluster is pH-dependent (pK a ϭ 7.8; slope ϭ -55 mV per pH unit) due to direct protonation of the reduced cluster. This proton transfer reaction is known to be physiologically relevant (8,45). At high pH, the alkaline limit is ϳϪ430 mV versus standard hydrogen electrode, but at pH 6 it is ϳϪ320 mV, or very similar to the potential of the NADPH/FPR couple. These potentials were measured for free FdI in direct electrochemical experiments and could be modified somewhat upon FdI binding to FPR, as has been demonstrated for the photosynthetic ferredoxin/ferredoxin-NADP ϩ reductase complex from Anabaena (39).
As shown in Fig. 10, electron transfer from NADPH to FPR to FdI can be demonstrated using the nonphysiological NADPH-cytochrome c reductase assay. The background activity obtained by mixing NADPH and FPR with oxidized cytochrome c is specifically enhanced by addition of FdI. The specific activity of 0.8 mole cytochrome c reduced/mg of FPR/ min at pH 6.5 is much higher than that of the spinach ferredoxin-NADP ϩ reductase (FNR): ferredoxin system (published value ϭ 1.2 ϫ 10 Ϫ4 mol min Ϫ1 mg Ϫ1 ) but slightly lower than that of cytochrome P450 BM-3 (published value ϭ 19 mol min Ϫ1 mg Ϫ1 ). As shown in Fig. 10, the electron transfer is pH-dependent, as is the reduction potential of the [3Fe-4S] ϩ/0 cluster with the lower, more physiological values, bringing the FdI reduction potential closest to that of FPR and giving the highest activities. To confirm that electron transfer from NADPH to FPR to the [3Fe-4S] ϩ cluster occurs in the absence of cyto- chrome c, we directly monitored the reaction using EPR. As shown in Fig. 11, even in the absence of pulling the reaction with cytochrome c and with a 8-fold excess of FdI over FPR present, 60% of the FdI [3Fe-4S] ϩ clusters became reduced to the 0 oxidation level. The incomplete reduction of the [3Fe-4S] ϩ cluster by NADPH/FPR indicates that the system is in equilibrium. Within the A. vinelandii cell, the system may be driven toward complete reduction of the [3Fe-4S] ϩ cluster by the as yet unidentified FdI electron acceptor. No reduction of the cluster was observed upon addition of either FPR alone or NADPH alone, and the absence of any new signals demonstrates, as we expected based on reduction potential considerations, that the [4Fe-4S] 2ϩ cluster was not reduced under those conditions.
The Interaction of A. vinelandii FPR:FdI-To further investigate the interactions of these two proteins, the electrostatic potentials of FdI and FPR were calculated (Fig. 12). The calculations were done assuming a physiological ionic strength of 150 mM salt concentration; the protonation and oxidation states were taken into account, with partial charges assigned to all residues and cofactors (see under "Experimental Procedures"). The electrostatic potential around each protein at a distance of 1.4 Å (the radius of a water molecule) from the molecular surface is plotted onto the molecular surface. In Fig.  12A, FdI and FPR are oriented to suggest a specific mode of interaction for the two proteins. The electrostatic potential of FdI is strongly negative on the face of the protein where the [3Fe-4S] cluster is nearest the surface. The residues on the surface in this region include the cluster ligand Cys 49 , Pro 50 , and Ala 51 , and the adjacent loop containing Glu 57 , Asp 58 , Glu 59 , Glu 62 , and Asp 63 and are in the peptide Cys 49 -Lys 85 , identified by cross-linking. The acidic loop participates in defining the negative electrostatic potential around the cluster, and it is interesting to note that these residues are all contained within the 60-amino acid COOH-terminal extension that is unique to this class of ferredoxins. In contrast, the opposite face of FdI is largely neutral or only weakly negative, except for a region adjacent to the [4Fe-4S] cluster (Fig. 12B). The surface directly over the [4Fe-4S] cluster, a large, almost flat protuberance, is essentially neutral as well, although it is surrounded by a ring of negative potential (Fig. 12C).
The surface of FPR displays a striking concentration of pos-itive potential within and around a cavity bounded by the ribityl atoms of FAD, N10 of the flavin, N6 and N7 of adenine, the side chains of Arg 51 and Lys 258 , and the amides of Ala 52 and Val 256 (Fig. 12A). Also contributing to the positive potential is the dipole of the only helix in the NH 2 -terminal FAD binding domain (Fig. 12A, left of Arg 51 ). A unique interaction of Arg 51 with the phosphate of the FMN portion of FAD contributes significantly to this electropositive cavity. Arg 51 is a conserved residue in the motif RxYS/T that is involved in FAD binding in all homologous oxidoreductase structures; however, AvFPR is unique in that a second basic residue that was identified as the cross-linking site, Lys 258 , also interacts with the FAD phosphates (27). The presence of Lys 258 causes Arg 51 to interact with only the FMN phosphate deeper within the cavity. The methyl groups of the dimethylbenzene portion of the flavin isoalloxazine ring and the side chain of Val 256 are adjacent to the cavity on the protein surface. Surrounding this region, the surface is mildly positive over much of the NH 2 -terminal domain. The surface of the larger COOH-terminal domain is mildly negative, whereas the circumference of the entire protein is weakly negative (Fig. 12A). In contrast, the opposite face of FPR is virtually neutral, again with the circumference being weakly negative (Fig. 12B). Therefore, the only pronounced positive electrostatic potential on the surface of FPR is directly over the active site.
One way in which the electrostatic surfaces of FdI and FPR may interact, and in fact the only way with obvious electrostatic complementarity, is for the strongly negative region surrounding the [3Fe-4S] cluster to dock with the positive region overlaying of FAD. In Fig. 12A, this can visualized by imagining that the page is creased between the two surfaces, allowing them to be folded toward each other. This interaction would provide charge complementarity for much of the FdI surface, especially that to the right of the [3Fe-4S] cluster in the view in  The weak negative potential over the rest of the larger FPR molecule could further act to repel the very negatively charged, smaller FdI molecule, thereby focusing FdI toward the positive region of FPR. Therefore, both attractive and repulsive electrostatic forces appear to play a role in orienting FdI and FPR for interaction. This orientation is consistent with our cross-linking data and electron transfer between FAD and the [3Fe-4S] cluster.
The proposed interaction of FdI with FPR is consistent with several independent observations and results. The strong electrostatic complementarity of the two protein surfaces would account for the binding of FPR to FdI on an affinity column where 300 mM salt concentration is required to displace FPR (15). It is also consistent with the observation that the formation of a specific 1:1 complex is favored at low ionic strength. As discussed above, the reduction potential of the [3Fe-4S] ϩ/0 cluster is more compatible with that of the NADPH/FAD couple than is the [4Fe-4S] cluster, and electron transfer from flavin to the [3Fe-4S] ϩ cluster has been demonstrated directly. Furthermore, the essentially neutral electrostatic potential on the surface of FdI at the [4Fe-4S] cluster (Fig. 12C) argues that this cluster is not involved in the electron transfer reaction. Electron transfer via exposed methyl groups of the flavin ring has been proposed in other systems in which protein-protein interactions were known (46,47,26). Mutagenesis studies have implicated the complementary interaction of basic residues on Anabaena ferredoxin-NADP ϩ reductase with acidic residues on the Anabaena [2Fe-2S] Fd (48,49). Fig. 13 compares the sequences of AvFPR to that of the proteins in the data base that are most similar to the protein.
The sequence from P. aeroginosa is new and was obtained from the Pseudomonas Genome Project. 7Fe ferredoxins Ͼ90% identical to AvFdI have been identified in a number of Pseudomo-nas species, including P. aeruginosa, and as shown in Fig. 13, the P. aeruginosa FPR sequence is also highly homologous; it is 85% identical and 91% similar to AvFPR. The COOH-terminal Lys 258 residue that we have identified as the FPR cross-linking residue is also conserved in the P. aeruginosa sequence, strengthening our view that the functions and interactions of the proteins in the two organisms should be the same. The next most similar protein in the data base to AvFPR is the protein from E. coli (Fig. 13). Although this organism does not appear to synthesize a homologous ferredoxin, there are close parallels between the regulation of the fpr genes in the two organisms (25). The specific redox partner of FPR in E. coli has not been identified but is believed to be flavodoxin or ferredoxin. The x-ray crystal structure of E. coli FPR has recently been determined, and three localized arginine residues (Arg 236 -Arg 237 -Arg 238 ) near the COOH-terminal end were proposed as anchoring points for the redox partner flavodoxin or ferredoxin (26). As shown in Fig. 13, these localized basic residues are not conserved in the AvFPR or PaFPR. The ferredoxin-NADP ϩ reductases involved in photosynthesis (designated FNR) are less homologous to the proteins from A. vinelandii, P. aeruginosa, and E. coli (Fig. 13). In Anabaena PCC 7119 FNR, the residues Lys 53 , Arg 77 , and Lys 294 are suggested to be involved in ferredoxin binding, and Arg 77 directly interacts with FAD (50 -52). These three residues are located in the FAD and NADP ϩ binding domains, implying that ferredoxin binds the two domains. For spinach FNR, the ferredoxin binding site has been suggested to be a loop containing Lys 85 and Lys 88 , residues that are exclusively in the FAD domain (53)(54)(55)(56). A homologous loop is present in the FNR from Anabaena but is missing in the AvFPR and EcFPR structures (Fig. 13).
Physiological Considerations-Taken together, the above data show that FPR and FdI bind specifically to one another and clearly demonstrate that electron transfer can occur from NADPH to FPR to the [3Fe-4S] ϩ cluster of FdI, strongly suggesting that this is the physiological direction of electron transfer. The next step is to try to identify the protein that accepts electrons from the protonated reduced [3Fe-4S] 0 -H ϩ cluster of  27. The x-ray crystal structure of PaFPR is not known, but it was assumed to be the same as AvFPR for the structure-based sequence alignment.
FdI. One possibility is that reduced FdI could serve, at least under some conditions, as an electron donor to nitrogenase. Indeed, many years ago, electron transfer from NADPH to crude preparations of nitrogenase was reported in a reaction that required addition of FdI and spinach ferredoxin reductase (57,58). Another report suggested that FdI, which is not encoded by a nif gene, might substitute for the nifF gene product flavodoxin under certain conditions because nifF deletion strains (like FdI deletion strains) still fix dinitrogen at wildtype rates (59). This idea was tested by making a double mutant that did not synthesize either flavodoxin or FdI. That strain still fixed dinitrogen; however, FPR levels in that strain were greatly overexpressed, leading to the suggestion that FPR might serve as another substitute electron donor to nitrogenase (60). In this study, we have directly tested these ideas using a system composed of all purified components. In these experiments, we were unable to demonstrate any electron transfer from NADPH to FPR to nitrogenase in the presence or absence of FdI, leading to the conclusion that this is not the physiological function of either protein.
A second possibility for a metabolic function for the FPR/FdI system would be in the activation of enzymes involved in anaerobic metabolism, because this is known to be the function of the homologous FPR from E. coli (19 -21). Although A. vinelandii is an obligate aerobe, it is likely to have internal anaerobic metabolism to protect the extremely oxygen sensitive enzyme nitrogenase from oxygen, and future experiments will be directed toward examining this possibility. We note here that repeated attempts to delete the fpr gene from the A. vinelandii chromosome have been unsuccessful, making it likely that this protein is necessary for the survival of the cell.
In addition to its metabolic function, which is important for cell growth, FdI has a regulatory function in controlling the expression of its redox partner FPR, a system that is considered in detail elsewhere (24,25). As in E. coli, the A. vinelandii fpr gene is activated when the cells are exposed to oxidative stress via a transcriptional activator that binds to a SoxS-like consensus sequence just upstream of the fpr gene (24). The transcriptional activator in A. vinelandii has not yet been purified, but it is known that the role of FdI in the system is to shut down the activation of fpr transcription. The current working hypothesis, which is consistent with all of the data presented here and discussed in detail elsewhere (23)(24)(25), is that when oxidative stress is over, NADPH levels build up, FPR becomes reduced, and the reduced FPR specifically reduces the [3Fe-4S] ϩ cluster of FdI, which then goes on to reduce the next protein in the regulatory cascade, leading to inactivation of the system. Future experiments are being directed toward identifying the member of the cascade that is specifically reduced by FdI.