Site-specific mutagenesis of Rhodobacter capsulatus ferredoxin I, FdxN, that functions in nitrogen fixation. Role of extra residues.

One of the two [4Fe-4S]-type clusters of the Rhodobacter capsulatus ferredoxin I, FdxN, was modified through site-specific mutagenesis of the distinctive features of the second cluster-binding motif, Cys38-X2-Cys41-X8-Cys50-X3-Cys54-X4-Cys59. First, various mutagenized products were tested to learn whether they could rescue the decreased capacity of an fdxN-null strain MSA1 to fix nitrogen: the phenotype of MSA1 was reassessed to Nifs (low growth by trogen ixation) from our previous description of Nif− (Saeki, K., Suetsugu, Y., Tokuda, K., Miyatake, Y., Young, D. A., Marrs, B. L. and Matsubara, H. (1991) J. Biol. Chem. 266, 12889-12895). Substitution of Cys59 to Ser yielded an almost fully active product, while that of Cys54 did not. Gradual deletions and deletion-substitution of the 8 residues between Cys41 and Cys50 also yielded active products. Second, three of the modified FdxN proteins were subjected to purification. Only the GA protein, whose 8 residues between positions 42 and 49 were replaced by the Gly-Ala sequence, was purified. The GA protein and the authentic FdxN showed similar optical properties. The two clusters in the former had Em values of −490 and −430 mV, while those in the latter had an identical value of −490 mV, when determined by EPR analysis. It was concluded that: 1) Cys59 is not a ligand to [4Fe-4S] clusters but is important for structural integrity, 2) the residues between positions 42 and 49 may form a “loop-out” from a structure analogous to the Peptococcus aerogenes ferredoxin, and 3) the loop-out region does not have functional significance in nitrogen fixation but may be responsible for maintaining the highly negative redox potential of one of the two clusters.

Ferredoxins are generally small iron-sulfur proteins that function in diverse metabolic pathways. They can be divided into several classes depending on the nature and number of the  (1,2). Most 2[4Fe-4S]-type ferredoxins con-sist of less than 60 amino acids and have internal sequence similarities between the amino-terminal and carboxyl (COOH)terminal halves. They possess two cluster-binding motifs of a sequence consensus, Cys-X 2 -Cys-X 2 -Cys-X 3 -Cys. Each [4Fe-4S] cluster is chelated by the first three Cys residues of one motif and the last Cys residue of the other motif. These properties reflect on their tertiary structures that display an approximate internal 2-fold rotation axis to relate the two clusters and most of the polypeptide chain (3)(4)(5).
Among the 2[4Fe-4S]-type ferredoxins, there is a subclass in that the second cluster-binding motif conforms to a distinct sequence consensus, Cys-X 2 -Cys-X 7ϳ9 -Cys-X 3 -Cys-X 3ϳ5 -Cys, as depicted in Fig. 1 (1). This consensus can be characterized by two features: 1) possession of "extra" residues, where the number of residues between second and third Cys is larger than the common number two; and 2) possession of an "additional" fifth Cys, which can be a potential ligand to the Fe-S cluster. Members of this subclass have been purified from photosynthetic bacteria that are diazotrophs (6 -11). In addition, several members have also been identified as products of potential ferredoxin genes, which have often been designated as fdxNs, in the nif (or anf) gene region of photosynthetic (12) and non-photosynthetic diazotrophs (13)(14)(15)(16)(17). Here, we tentatively designate this subclass as photosynthetic bacterial and nif-associated ferredoxins. The distribution among such bacterial species and the results of gene disruption experiments performed with Bradyrhizobium japonicum (13), Rhizobium meliloti (15), and Rhodobacter capsulatus (18,19) may lead to the inference that these members have a function(s) in nitrogen fixation. Nevertheless, a gene to encode a ferredoxin of this subclass is also identified in a non-diazotrophic bacterium Haemophilus influenzae, whose whole nucleotide sequence of the genome has recently been determined (20).
It is of interest if the above two characteristics of the unique second cluster-binding motif have physiological and/or physicochemical significance, but there have been only a few studies on this subject. At the physiological level, the site-specific mutagenesis of R. meliloti FdxN has shown that the substitution of the additional Cys to Ser abolishes the capacity to support nitrogen fixation in planta, while the deletion of the extra residues has little effect (21). At the structural level, the nuclear magnetic resonance (NMR) analyses of the Chromatium vinosum ferredoxin, which has a unique COOH-terminal extension of 22 amino acids, have revealed that this extension sequence forms an ␣-helix and interacts with a turn formed by the extra residues between the Cys residues. It was also shown that the additional Cys residue is not a ligand to the [4Fe-4S] clusters (22). These studies designed under different disciplines remain to be interrelated.
The photosynthetic bacterium R. capsulatus contains as many as six ferredoxins of different classes (23-28); among them, the ferredoxin I encoded by the fdxN gene belongs to the photosynthetic bacterial and nif-associated subclass (29,30). Despite the possession of a variety of ferredoxins and a flavodoxin (31), the disruption of fdxN severely decreases the capacity to fix nitrogen (18,19), and this defective phenotype can be complemented by the reintroduction of the gene (18,32). In this paper, we have site-specifically mutagenized the fdxN gene to investigate the functional significance of the additional Cys residue and the extra residues. We further purified the engineered gene product that lacks the extra residues and compared its physicochemical properties with those of the authentic protein.

EXPERIMENTAL PROCEDURES
Bacterial Strains and Culture Conditions-Escherichia coli strains DH5␣ (33) for plasmid construction, TG1 (33) for preparation of single strand DNA (ssDNA), 1 CJ236 (34) for preparation of uracil-containing ssDNA, and Tec5 (35) for conjugative transfer of pBR322 derivative to R. capsulatus) were grown at 37°C in either 2 ϫ YT or LB medium (33). The R. capsulatus strains used were wild-type SB1003 (36) and fdxNnull MSA1 (18). Photosynthetic growth on either malate-(NH 4 ) 2 SO 4 minimal medium RCV (37), glutamate medium RCV-E (38), ammonium-free medium RCV-NF (38), or peptone yeast extracts medium PYS (37) was at 30°C with illumination from arrays of 60-W tungsten lamps, in either screw-cap tubes or bottles, or on agar plates placed in a GasPak anaerobic jar (BBL Microbiology System, Cockeysville, MD). The light intensity at the surface of the tube or plate was above 5 mW/cm 2 as measured with an optical power meter TQ8210 equipped with a TQ82017 silicon photodiode sensor (Advantest, Tokyo, Japan). Chloramphenicol was used for E. coli at 30 g/ml; the concentrations of other antibiotics have been described previously (18).
Molecular Genetic Techniques-Genetic procedures and recombinant DNA techniques were as reported previously (18) or according to Sambrook et al. (33).
To construct a vector with R. capsulatus puf promoter, an 806-base pair XhoI-EcoRI fragment from the plasmid pDAY1 (39) was cloned into pBR322 digested with SalI and EcoRI to make the plasmid pBRPuf.
Because the ampicillin-resistant (Amp r ) gene in pBR322 does not work in R. capsulatus, a 2-kilobase pair EcoRI-PstI fragment that contained spectinomycin-resistant gene was excised from plasmid pYS⍀2 (38,40) and was used to replace the Amp r gene. The resulting vector pTSV1 has the puf promoter followed by the pufQ gene, and the EcoRI/SacI/KpnI/ SmaI/BamHI multi-cloning site originated from pUC118 (41).
The plasmid pMFN3, which was the template plasmid in site-specific mutagenesis, was constructed by inserting a BamHI linker, pCG-GATCCG, into the pFA9d3 (18), which was digested with XhoI and HindIII and then blunted with Klenow polymerase.
Site-specific mutagenesis of fdxN was performed according to Kunkel et al. (34) using uracilated ssDNA produced by E. coli CJ236 harboring the plasmid pMFN3 (pMFN3/CJ236). The mutagenic oligonucleotide primers used are listed below.  The obtained plasmids were subjected to sequence determination using an Applied Biosystems DNA sequencer (model 373A) after the thermal cycling reaction with Taq DNA polymerase. Those confirmed for the expected mutagenesis events were named pMTM3-pMTM5 and pKTM3-pKTM7, respectively, according to the primers used. Plasmid pMTM4 was used to generate uracilated ssDNA, and a second mutagenesis was performed with the MT5 primer to obtain plasmid pMTM45.
To construct the plasmid to express fdxN under the control of the puf promoter in R. capsulatus, pMFN3 was digested with EcoRI and BamHI to excise the DNA fragment with authentic fdxN. The fragment  (29). Hyphens symbolize gaps introduced to maximize homology. (ϩ) indicates extending COOHterminal sequence, and total numbers of amino acids are shown in parentheses. Cysteine residues were highlighted by inverse contrast. The squares symbolize the ligands for a [4Fe-4S] cluster and the triangles for a [3Fe-4S] cluster. Clusters I and II are shown with open and filled symbols, respectively. aa, amino acids was then subcloned in pTSV1 to make plasmid pFNP. The other 10 expression plasmids, pPM3-pPM5, pKT3-pKT7, and pPM45 with engineered fdxN, were likewise constructed by ligating the fragments excised from pMTM3-pMTM5, pKTM3-pKTM7, and pMTM45 to the pTSV1 vector.
Construction of the plasmid, pYSA111, to express the fdxA gene under the control of the puf promoter will be described elsewhere. 2 Nitrogenase Activity-Nitrogenase activity with R. capsulatus cells under illumination at 5 mW/cm 2 was measured by the acetylene reduction method as described previously (38). R. capsulatus cells were grown in RCV-E medium for 26 h and washed twice with anaerobic RCV-NF medium before use.
Purification of Engineered Ferredoxin-R. capsulatus MSA1 cells harboring either plasmid pFNP, pPM3, pPM4, or pPM5 were photosynthetically grown for overnight in a PYS medium supplemented with 40 M iron sulfate and 1 g/ml spectinomycin, harvested by centrifugation, and stored at Ϫ70°C until use. All the procedures, except the cell disruption and centrifugation steps, were performed in a Coy anaerobic chamber (Coy Laboratory Products Inc., Ann Arbor, MI) using anaerobic buffers prepared by bubbling with N 2 gas and degassing. Cells, approximately 80 g (wet weight), were thawed and suspended with 100 ml of 0.1 M Tris-Cl buffer (pH 8.0) containing 5 mM dithiothreitol. The suspension was passed through a French pressure cell at approximately 20,000 lb/in 2 and was anaerobically transferred to a screw-cap tube under a flow of N 2 gas, then centrifuged at 160,000 ϫ g for 90 min. The supernatant was degassed and applied to a DEAE-Sepharose fast-flow column (2.5 ϫ 10 cm). The column was first washed with 100 ml of the 0.1 M Tris-Cl buffer (pH 8.0) containing 2 mM sodium dithionite (DT buffer) and then with 100 ml of DT buffer containing 0.1 M NaCl (0.1SDT buffer). The brown fraction containing ferredoxins was eluted by the 0.5SDT buffer. The eluate, usually 5-6 ml, was chromatographed on a Sephadex G-75 column (2.5 ϫ 50 cm) with 0.1SDT buffer. The ferredoxin fraction was diluted with an equal volume of DT buffer and applied to a second DEAE-Sepharose column (1.5 ϫ 10 cm). The column was developed by a linear gradient of NaCl to 0.5 M in DT buffer. At this step authentic or engineered FdxN was separated from other proteins that included ferredoxin II. The FdxN fraction was concentrated using a small DEAE-Sepharose column and chromatographed on tandemly connected two Toyopearl HW-50F columns (2.5 ϫ 50 cm ϫ 2) with 0.1 M Tris-Cl (pH 8.0) buffer.
When the exchange of the buffer system was needed, the sample was adsorbed on a small DEAE-Sepharose column, eluted with a new buffer containing 0.5 M NaCl and passed through a Sephadex G-25 column to remove NaCl.
Protein Chemical Analysis-Amino acid composition of the purified samples and their derivatives was determined by the method of Spackman et al. (42) using an Irica A-5500 analyzer (Irica Co., Kyoto, Japan). S-Carboxymethylation of reduced apo-ferredoxin was performed as described previously (43). Molecular mass of Cm-ferredoxins was determined by a Hitachi M-1200H LC/MS system equipped with an electrospray ionization interface (Hitachi Ltd., Tokyo, Japan).
Spectroscopic Analysis-Light absorption and circular dichroism (CD) spectra were measured by a Shimadzu UV-3101PC spectrophotometer and a Jasco J-500A circular dichrometer, respectively. For anaerobic measurement, the quartz cells with a rubber stopper were loaded in the Coy anaerobic chamber. The oxidized form of the purified samples was achieved by empirically adding a small amount of air through the rubber stopper using a gas tight syringe. Electron paramagnetic resonance (EPR) spectra were obtained using an X-band EPR spectrometer (model EPR-200, Bruker, Germany) with an Oxford Instruments ESR-900 liquid helium cryostat. Redox titration was conducted using the method of Dutton (44). Redox potential (E h ) was versus the normal hydrogen electrode.
Molecular Modeling of R. capsulatus FdxN and FdxN GA -The threedimensional structures of R. capsulatus FdxN and FdxN GA were modeled based on the structure of ferredoxin from P. aerogenes 3 determined by x-ray crystallography. Replacements and insertions of amino acid residues were performed using the Homology module of Insight II 95.0 (Biosym/MSI, San Diego, CA). The energy of the two modeled structures was minimized by the method of conjugate gradient until the root mean square of atomic forces reached less than 0.01 kcal/mol/Å. The molecular dynamics simulation was then performed on FdxN and FdxN GA for 100 ps. The energy minimization and the molecular dynamics simula-tion were executed on BIOGRAF 3.2.1 (45). The dielectric constant was set to ⑀ ϭ r, where r is the distance between two atoms. The coordinates of the [4Fe-4S] clusters and sulfur atoms of ligand cysteines were fixed during the calculation.

RESULTS
This study aims to analyze the physical and chemical properties of engineered FdxNs (FdxN* proteins). Before such analysis, the capacity of R. capsulatus fdxN-disrupted mutant MSA1 (18) to utilize molecular nitrogen was reinvestigated first at 5 mW/cm 2 intensity of the illumination for growth that was at least 3-fold stronger than that used in our previous study (18). The strain displayed no growth on the solid RCV-NF medium, which lacked a nitrogen source, in 3 days as reported previously (18), while the wild-type SB1003 had fully grown during the same period. However, the mutant displayed a trace of, but evident, growth after 7 days, unlike our previous results with lower light intensity. The nitrogenase activity of MSA1 cells grown anaerobically in RCV-E medium was about 5% of that of the wild-type cells (Fig. 2). Thus, the disruption of fdxN was less critical on the capacity to fix nitrogen under stronger light. We, therefore, revised the phenotype of MSA1 from Nif Ϫ to Nif s (slow growth by nitrogen fixation) and concluded that FdxN is not absolutely required for nitrogen fixation by R. capsulatus.
Restoration of Nitrogenase Activity of fdxN-null Mutant by puf Promoter-driven Expression of Authentic fdxN-We employed a system that enabled us to test the functional competency of an fdxN* and would also facilitate the purification of its product. Namely, we used a vector pTSV1 with R. capsulatus puf promoter to express an engineered ferredoxin gene in the strain MSA1. The promoter is strongly induced under anaerobic conditions regardless of the nitrogen sources, so that any fdxN* gene subcloned should be highly transcribed even when other nif-related genes are repressed. For the validation of this system, the authentic fdxN gene was subcloned in the vector to construct plasmid pFNP and conjugatively transferred into MSA1. The transconjugant strain pFNP/MSA1 showed full Nif ϩ phenotype in terms of both growth rate and nitrogenase activity as pCP-BH2/MSA1 did (Fig. 2). About 20 mg of FdxN could be purified from 80 g (wet weight) cells of pFNP/MSA1 grown in PYS medium under illuminated anaerobic conditions. The yield was comparable with that from the wild-type cells grown under the nif-derepressed conditions.
Effects of Overexpression of fdxA on Nif s Phenotype of fdxNnull Mutant-Although fdxN is disrupted in MSA1, the strain still contains ferredoxin II that is a [4Fe-4S][3Fe-4S]-type ferredoxin encoded by the fdxA gene and can function as electron donor to nitrogenase in vitro (24). The cellular concentration of FdxA is relatively low (29) and constant regardless of the nitrogen source (40). Therefore, the effect of the overexpression of fdxA in MSA1 was investigated. We transferred the plasmid pYSA111, which was a pTSV1 derivative. Immunoblot analysis using anti-FdxA antibodies revealed more than a 5-fold increase in FdxA concentration in cells of pYSA111/ MSA1 grown under anaerobic conditions (data not shown). The elevated expression of fdxA from pYSA111 enabled us to purify 10 times more FdxA protein from SB1003 harboring the plasmid than from the wild-type. However, the transconjugant strain pYSA111/MSA1 showed only a slight increase in nitrogenase activity, and its growth rate on nitrogen-free medium was comparable with those of MSA1 and pTSV1/MSA1 (Fig. 2). These results indicated that the level of increase in the [4Fe-4S][3Fe-4S]-type FdxA was still insufficient to compensate for the loss of the 2[4Fe-4S]-type FdxN, in vivo.
Site-specific Mutagenesis of Cys Residues of fdxN-We have made engineered genes to substitute Cys 54 and/or Cys 59 to Ser (Fig. 3). The three fdxN* genes, fdxN C54S , fdxN C59S , and fdxN C54S/C59S , were subcloned, respectively, into the vector pTSV1 to construct plasmids, pPM4, pPM5 and pPM45, and conjugatively transferred to R. capsulatus MSA1. The resulting transconjugants were surveyed for their capacity to fix nitrogen (Fig. 3). The strain pPM5/MSA1, which possessed fdxN C59S , grew on RCV-NF plates as fast as SB1003 and pFNP/ MSA1. This strain exhibited a nitrogenase activity approximately 90% of those of the strains with the authentic fdxN. In contrast, pPM4/MSA1 and pPM45/MSA1, which have fdxN C59S and fdxN C54S/C59S , respectively, showed the Nif s phenotype indistinguishable from MSA1 and pTSV1/MSA1, in terms of growth rates and nitrogenase activities. These results demonstrated that the substitution of Cys 54 eliminates the functional capacity, while that of additional Cys 59 does not.
Site-specific Mutagenesis of Extra Residues of fdxN-To survey the physiological significance of the extra residues, we made seven mutant genes, fdxN 42-49/7r -fdxN 42-49/3r and fdx-N GA , whose products have gradual deletions between the 42 and 49 positions and finally the replacement of the 8 residues by Gly-Ala sequence (Fig. 3). The fdxN* genes were subcloned in pTSV1 to make plasmids pPK7-pPK3 and pPM3 and introduced into the strain MSA1. All the resulting transconjugant strains grew on N 2 at rates comparable with SB1003 and pFNP/MSA1. Most of them displayed nitrogenase activities comparable with those of the wild-type and pFNP/MSA1, though three strains, pPK4p/MSA1, pPK4a/MSA1, and pPK3/ MSA1, which should have FdxN* proteins with the deletion of 4 -5 residues, displayed slightly decreased activities (Fig. 3). The decrease in activity apparently correlated to the extent of the deletion, but it must be noted that the strain pPM3/MSA1 which has FdxN GA protein lacking whole extra residues exhibited activity equal to those of pFNP/MSA1 and SB1003. These results clearly showed that the extra residues of FdxN are not FIG. 2. Complementation test of fdxN ؊ mutant MSA1 with puf promoter-driven plasmid-borne fdxN and fdxA. At left, the restriction map of the fdxN loci (white bar) of R. capsulatus strains SB1003 and MSA1 is aligned with the nucleotide fragments in the various related plasmids. The growth capacity on N 2 and acetylene reduction activity of each relevant strain are shown at right. The fdxN, fdxA, and known genes were shown as black, gray, and white arrows, respectively: Km r denotes kanamycin-resistant gene. The plasmids pCP-BH2, pKTS1, and pKTS2 have been described previously (18). A hatched bar represents the XhoI-EcoRI fragment containing the puf promoter. The nif and puf promoter sequences are shown by black and white triangles. The restriction enzyme sites are indicated as B, M, E, S, and H corresponding to BamHI, MluI, EcoRI, SalI, and HindIII, respectively. (ϩϩϩ) indicates full growth within 3 days under illuminated anaerobic condition on solid RCV-NF medium that lacked nitrogen source, while (Ϯ) indicates slight growth after 5-7 days under the same conditions. Nitrogenase activity was measured by acetylene reduction method using R. capsulatus cells that had been anaerobically grown in a medium containing glutamate as the sole nitrogen source (RCV-E) medium. indispensable to function in nitrogen fixation.
Purification of Products of Engineered fdxN Genes-Three engineered ferredoxins, FdxN C54S , FdxN C59S , and FdxN GA , were subjected to anaerobic purification from the cells of appropriate strains. We could usually purify approximately 25 mg of FdxN GA from 80 g (wet weight) cells of pPM3/MSA1; the yield was about 125% of that for authentic FdxN from the cells of pFNP/MSA1. On the other hand, we could not purify FdxN C54S and FdxN C59S protein by the same procedures, respectively, from cells of pPM4/MSA1 and pPM5/MSA1. Because the strain pPM5/MSA1 showed diazotrophic growth comparable with SB1003 (Fig. 3), the results indicate that the FdxN C59S is functional in vivo, but is unstable in vivo or becomes unstable during the purification.
The amino acid composition of the purified FdxN GA protein differed, as expected, from the authentic FdxN protein (data not shown). The molecular mass of the fully carboxymethylated Optical Spectroscopic Properties of the FdxN GA Protein-The ultraviolet (UV)-visible light absorption and CD spectra of Fdx-N GA protein were compared to those of the authentic FdxN. The absorption spectra of oxidized and dithionite-reduced Fdx-N GA protein were almost similar to those of the authentic FdxN protein (Fig. 4). The CD spectra of the two proteins were almost the same in near-UV and visible regions; the apparent difference in near-UV region might be due to the strong difference in far-UV region (Fig. 5). Both proteins exhibited peaks at 295, 320, 428, and 567 nm, suggesting little difference in the structures of the [4Fe-4S]-clusters. In contrast, far-UV CD spectra of FdxN GA and the authentic FdxN significantly differed, suggesting a change in the secondary structure by the removal of extra residues (Fig. 5).
EPR Analysis-Similar to clostridial 2[4Fe-4S]-type ferredoxins, oxidized forms of FdxN and FdxN GA were EPR-silent (data not shown; see Fig. 7). Their fully reduced forms with excess sodium dithionite exhibited EPR spectra significantly distinct from each other (Fig. 6). Fully reduced authentic FdxN showed a complex EPR spectrum with signals at g values 2.10, 2.02, 1.95, and 1.83, which suggested spin coupling between the two reduced [4Fe-4S] clusters (46). In contrast, fully reduced FdxN GA showed another form of complex EPR spectrum with signals at g values 2.13, 2.01, 1.97, 1.89, and 1.82. The spectrum was similar to that of reduced Clostridium pasteurianum ferredoxin (47) and appears also to be produced by the spin coupling between two reduced [4Fe-4S] clusters (46).
We performed redox titrations at pH 10.0 at 283 K (typical EPR spectra measured at 10 K are shown in Fig. 7); titrations conducted at pH 8.0 gave essentially the same results (data not shown). The authentic FdxN exhibited a very weak rhombic signal, which should correspond to one reduced [4Fe-4S] cluster with spin ϭ 1/2 at the E h range between Ϫ360 mV and Ϫ460 mV. Some fraction of FdxN should be singly reduced, i.e. at the state that one of the two clusters was reduced. This signal was overtaken by the complex signals of the interacting two reduced [4Fe-4S] clusters at the E h values more negative than Ϫ460 mV. The peak at g ϭ 1.89 detected above Ϫ469 mV in Fig.  7 seemed to represent the fraction of the protein of singly reduced form, although the signal intensity became a mixture of fractions of singly and doubly reduced forms below this potential. The signal intensity at g ϭ 1.92, on the other hand, seemed to represent primarily the fraction of doubly reduced form. Therefore, the signal intensities at g ϭ 1.89 and 1.92 were plotted against the redox potential after the normalization of the amplitudes to those at the fully reduced conditions (Fig. 8, upper). The two plots were almost similar but showed a difference above Ϫ500 mV. The apparent E m read as an E h value for 50% normalized amplitude was about Ϫ510 mV and matched the value electrochemically determined by Naud et al. assuming that the two [4Fe-4S] clusters have an identical redox midpoint potential (E m ) of Ϫ490 mV and are independent, i.e. behave as the case of twice as many molecules each bearing a single [4Fe-4S] cluster (47). The theoretical curve for the doubly reduced fraction fits the data derived from the peak intensities at g ϭ 1.92. The curve for the singly reduced fraction does not correlate to the two plots, but it resembles the deviation between them. The relatively weak signal of the singly reduced form, however, made further analysis to be rather difficult under the present experimental conditions.
The titration conducted with FdxN GA gave somewhat different results. The protein showed a rhombic signal (g x,y,z ϭ 1.85, 189 and 2.08) at the E h range between Ϫ360 mV and Ϫ460 mV. The signal can be assigned to one reduced [4Fe-4S] cluster with spin ϭ 1/2. A considerable fraction of FdxN GA should be at the singly reduced form in this E h range. As the potential was lowered below Ϫ460 mV, the signal was replaced by the complex signals that have features of spin coupling between two reduced [4Fe-4S] clusters and represent the fraction of FdxN GA of the doubly reduced form. The peaks at g ϭ 2.08 and 1.90 seemed to represent the fractions of singly and doubly reduced forms, respectively. Their intensities were plotted against E h (Fig. 8, lower); the heights of g ϭ 2.08 peak were plotted after the subtraction of the minor contribution of the doubly reduced form. Simulations of the titration curves were achieved by assuming the two [4Fe-4S] clusters with different E m values of Ϫ490 and Ϫ430 mV, respectively. Each theoretical curve fits the corresponding plot. This indicates that the E m of one Fe-S cluster shifted by approximately ϩ60 mV with a minor effect on the other cluster that retains an E m value of Ϫ490 mV.

DISCUSSION
The protein engineering in this study indicated the functional and physicochemical significance of the unique clusterbinding motif of the ferredoxin I of R. capsulatus. We used the fdxN-null mutant MSA1, which has a deletion-insertion of the kanamycin-resistant gene in fdxN, with an expression vector pTSV1 that has the puf promoter. The phenotype of the mutant MSA1 that was revised from Nif Ϫ to Nif s (18) agrees with that of the fdxN-null mutant with the insertion of gentamycinresistant gene reported by Schmehl et al. (19). The absence of FdxN was less critical under the high light intensity. The role of FdxN must be, therefore, more important under non-optimal conditions for nitrogen fixation. The [2Fe-2S]-type ferredoxin encoded by the fdxC gene that situates upstream of fdxN may also be required for optimization of diazotrophic growth, because we observed that the fdxC-null mutant MSB1 grew very slowly under increased illumination with the nitrogenase activity of nearly 10% of that of the wild-type strain (49).
The remaining capacity to fix N 2 of the fdxN-null mutants of R. capsulatus might depend on either direct electron transfer between the still unidentified reductase and oxidase of FdxN, possibly the Fe protein of nitrogenase (32,50), or other ferredoxin(s) (24 -28). We had previously pointed out the possibility that the cellular concentration of FdxA is simply too low to support the capacity (18). This [4Fe-4S][3Fe-4S]-type ferredoxin is constitutively expressed (40) and can donate electrons Conditions of measurements were essentially the same as in Fig. 6, except that the temperature was 10 K.
to nitrogenase in vitro with the apparent K m value higher than that for FdxN (24,51). However, the 10-fold overexpression of FdxA showed practically no effect on the Nif s phenotype in MSA1. Therefore, the functional difference between the 2[4Fe-4S]-type FdxN and the [4Fe-4S][3Fe-4S]-type FdxA in vivo must be based not only on the differential control of gene expressions but also on their structural difference.
The FdxN C59S protein was almost fully functional in vivo, whereas neither FdxN C54S nor FdxN C54S/C59S was functional in vivo. The results indicate that the additional Cys 59 is normally a free cysteine that does not participate in cluster binding. Thus the two [4Fe-4S] clusters should be chelated by two sets of Cys residues: Cys 9 , Cys 12 , Cys 15 , and Cys 54 for the first cluster (cluster I), and Cys 38 , Cys 41 , Cys 50 , and Cys 19 for the second cluster (cluster II), as depicted in Fig. 1. Consequently, the extra 8 residues between Cys 41 and Cys 50 of the R. capsu-latus FdxN should form a "loop-out" from a structure analogous to that of the Peptococcus aerogenes (i.e. Peptostreptococcus asaccharolyticus) ferredoxin (3, 52) ( Fig. 9 and description below), as estimated for the case of the C. vinosum ferredoxin that has a COOH-terminal extension of 22 amino acids, which interacts with the extra region (22). Although the FdxN C59S protein was functional in vivo, we failed to isolate this protein. This implies that the additional Cys 59 , which is not essential for function, is important for the structural integrity of the protein. Such an interpretation may explain the contradictory report that a similar substitution of the additional Cys to Ser in R. meliloti FdxN suppresses its capacity to support nitrogen fixation in planta (21) Their extra residues might be too short to form stable proteins that fulfill the physiological function. This may explain why we have not found extra residues with less than 6 residues among the natural ferredoxins of the subclass. Our results confirm the report by Masepohl et al. (21), who showed that the deletion and substitution of the extra residues of R. meliloti FdxN has almost no effect on the capacity to support nitrogen fixation in planta. On the other hand, Rhodospirillum rubrum is reported to produce FdxN protein even under nif-repressed growth conditions (11). A gene to encode a ferredoxin with similar extra residues has also been identified in the non-diazotrophic bacterium H. influenzae (20). These findings, together with the results of the protein engineering studies with R. meliloti (21) and R. capsulatus (this study), suggest that the possession of loop-out and extra residues in the second clus- ter-binding motif does not necessarily indicate a functional link to nitrogen fixation.
The FdxN GA protein that had the Gly-Ala sequence in replacement of the extra residues was essentially indistinguishable from the authentic FdxN in its capacity to support nitrogen fixation in vivo, as well as in the optical absorption spectra and the visible CD spectrum. However, the two proteins showed distinct paramagnetic and redox properties. The authentic FdxN showed a weak signal of a singly reduced [4Fe-4S] cluster in a narrow range of E h . The EPR spectrum in its fully reduced state suggests an interaction between two clusters that have essentially identical E m values at around Ϫ490 mV. The shape of the signal was somewhat different from those of typical 2[4Fe-4S]-type ferredoxins (46,47). In contrast, FdxN GA showed a clear signal of a singly reduced [4Fe-4S] cluster in a relatively wide range of E h and a spectrum similar to that of the typical clostridial 2[4Fe-4S]-type ferredoxins (46,47) in the fully reduced state. The deletion of the extra loop seems to have shifted the E m value of one of the two clusters from Ϫ490 to Ϫ430 mV and modified the spin interaction mode of the two clusters. The different signal shapes of fully reduced FdxN and FdxN GA may represent the modified spin interaction.
We simulated the tertiary structure of the FdxN on the basis of the known structure of P. aerogenes ferredoxin 3 as shown in Fig. 9. The putative structure of the authentic FdxN obtained by energy minimization and molecular dynamics simulation indicated that the extra loop shields cluster II from the outer medium. The deletion of the extra loop, thus, would expose cluster II to the medium with a slight effect on cluster I. Because the shielding of a [4Fe-4S] cluster from the medium is expected to destabilize the electrons on the cluster and to shift its E m to the more negative side according to theoretical investigations (55,56), mutagenesis might induce the positive shift of the E m of cluster II. It can be concluded that the extra loop together with its potential interaction with the COOH-terminal residues is responsible for keeping the negative E m of cluster II. This conclusion may be a clue in explaining the reason why ferredoxins with the extra residues are found in many diazotrophs and some non-diazotrophs, even though the feature is dispensable for the diazotrophic growth of R. meliloti (21) and R. capsulatus. Possession of cluster(s) with an extremely negative E m value, which might be favorable for interaction with ferredoxin oxidase(s) that has a very negative E m value, such as nitrogenase, could be an evolutional advantage. Namely, it can be speculated that the redox properties of ferredoxin become a limiting factor for the total efficiency of nitrogen fixation under some non-optimal diazotrophic conditions which we have not examined in this study. If so, the extra loops had been functional modules for the survival of those bacteria.