Identification of an Fe protein Residue (Glu146) ofAzotobacter vinelandii Nitrogenase That Is Specifically Involved in FeMo Cofactor Insertion*

The Fe protein of nitrogenase has three separate functions. Much is known about the regions of the protein that are critical to its function as an electron donor to the MoFe protein, but almost nothing is known about the regions of the protein that are critical to its functions in either FeMo cofactor biosynthesis or FeMo cofactor insertion. Using computer modeling and information obtained from Fe protein mutants that were made decades ago by chemical mutagenesis, we targeted a surface residue Glu146 as potentially being involved in FeMo cofactor biosynthesis and/or insertion. The Azotobacter vinelandii strain expressing an E146D Fe protein variant grows at ∼50% of the wild type rate. The purified E146D Fe protein is fully functional as an electron donor to the MoFe protein, but the MoFe protein synthesized by that strain is partially (∼50%) FeMo cofactor-deficient. The E146D Fe protein is fully functional in an in vitro FeMo cofactor biosynthesis assay, and the strain expressing this protein accumulates “free” FeMo cofactor. Assays that compared the ability of wild type and E146D Fe proteins to participate in FeMo cofactor insertion demonstrate, however, that the mutant is severely altered in this last reaction. This is the first known mutation that only influences the insertion reaction.

The Fe protein of nitrogenase has three separate functions. Much is known about the regions of the protein that are critical to its function as an electron donor to the MoFe protein, but almost nothing is known about the regions of the protein that are critical to its functions in either FeMo cofactor biosynthesis or FeMo cofactor insertion. Using computer modeling and information obtained from Fe protein mutants that were made decades ago by chemical mutagenesis, we targeted a surface residue Glu 146 as potentially being involved in FeMo cofactor biosynthesis and/or insertion. The Azotobacter vinelandii strain expressing an E146D Fe protein variant grows at ϳ50% of the wild type rate. The purified E146D Fe protein is fully functional as an electron donor to the MoFe protein, but the MoFe protein synthesized by that strain is partially (ϳ50%) FeMo cofactor-deficient. The E146D Fe protein is fully functional in an in vitro FeMo cofactor biosynthesis assay, and the strain expressing this protein accumulates "free" FeMo cofactor. Assays that compared the ability of wild type and E146D Fe proteins to participate in FeMo cofactor insertion demonstrate, however, that the mutant is severely altered in this last reaction. This is the first known mutation that only influences the insertion reaction.
Nitrogenase is composed of two separately purified proteins (for recent reviews see Refs. [1][2][3][4][5][6]. The iron protein (Fe protein) is a 60,000 M r dimer of two identical subunits encoded by the nifH gene. The two subunits are bridged by a single [4Fe-4S] cluster, and each of the subunits has a binding site for MgATP. The molybdenum-iron protein (MoFe protein) is a 230,000 M r ␣ 2 ␤ 2 tetramer with the ␣ and ␤ subunits encoded by the nifD and nifK genes, respectively. The MoFe protein contains two different types of metal clusters, the [8Fe-7S] P-clusters and the [Mo-7Fe-9S-homocitrate] clusters that are designated FeMo cofactor. Substrate reduction by the enzyme requires both component proteins with the Fe protein serving as a specific electron donor to the MoFe protein. To carry out this function in nitrogenase catalysis, the reduced Fe protein first binds two molecules of MgATP and then undergoes a global conformational change before forming a very specific complex with the MoFe protein. Electrons can then be transferred from the Fe protein to the P-clusters of the MoFe protein in a reaction that is somehow coupled to MgATP hydrolysis. Subsequently, the oxidized Fe protein dissociates from the MoFe protein, the Fe protein is rereduced, MgADP dissociates, and the cycle is repeated.
In addition to its role in dinitrogen reduction, the Fe protein has at least two other functions in the cell. One of them is in the initial biosynthesis of FeMo cofactor (for reviews see . Although the complete pathway for FeMo cofactor biosynthesis is not yet established, it has been known for some time that the FeMo cofactor is synthesized separately from the MoFe protein polypeptides (14 -16) and that its synthesis requires the combined action of the nif Q, B, N, E, X, and V genes (5,(7)(8)(9)(10)(11)(12)(13)17). In 1986, surprising reports appeared that mutants of Klebsiella pneumoniae and Azotobacter vinelandii that did not synthesize the nifH polypeptide also did not synthesize FeMo cofactor (15,16). Based on these and other experiments (18 -21), it is now accepted that the nifH gene product is required for the initial biosynthesis of FeMo cofactor. Its role in this process, however, has not been established, nor is it known what features of the Fe protein are required for its participation in FeMo cofactor biosynthesis.
A third function of the Fe protein involves the insertion of preformed FeMo cofactor into a FeMo cofactor-deficient form of the MoFe protein (for reviews see . This final maturation of the MoFe protein also appears to occur in a series of steps. First, a P-cluster containing but FeMo cofactordeficient MoFe protein is synthesized that has the FeMo cofactor site somehow inaccessible to FeMo cofactor insertion (18 -21). This form of the protein accumulates in strains that have deletions in the nifH gene. Next, in a reaction that requires both the Fe protein and MgATP, the FeMo cofactor-deficient MoFe protein is converted to another form with the FeMo cofactor site accessible for FeMo cofactor insertion (20,22). This reaction appears to require at least one additional protein, nifY in K. pneumoniae or gamma in A. vinelandii, and the role of the Fe protein/MgATP may be to facilitate the association of these proteins with the FeMo cofactor-deficient MoFe protein (23)(24)(25)(26)(27). If the Fe protein is available in vivo, as it is in nif B, N, or E minus strains, then this step has already occurred, and the FeMo cofactor-deficient MoFe protein that accumulates can be directly activated with isolated FeMo cofactor (20). Again, the exact role that the Fe protein plays in the conversion of the FeMo cofactor-deficient protein, from a form with the FeMo cofactor site inaccessible to one that can be directly activated with FeMo cofactor, is not known, nor is it known what features of the Fe protein are required for this reaction.
In recent years a large number of site-directed mutant variants of the Fe protein have been constructed to examine which features of the Fe protein are required for its first function, electron transport to nitrogenase (1)(2)(3)(4)(5)(6). In some cases, the strains carrying those mutations have also been examined with respect to whether or not the Fe proteins could carry out the other two Fe protein functions (28 -33). These and other studies (34) have led to the conclusions that to participate in FeMo cofactor biosynthesis and insertion, the Fe protein may have to bind MgATP, but it does not need to undergo the MgATPinduced conformational change, it does not need to form the normal complex with the MoFe protein, it does not need to transfer electrons to any protein, and it does not need to hydrolyze MgATP, at least using the mechanism that is used during nitrogenase catalysis. 1 Because these mutations were designed to study the mechanism of nitrogenase catalysis, all were defective in normal nitrogenase turnover (28 -33). As a result they could not be used to identify regions of the Fe protein that might be uniquely required for the other two functions of the protein.
Prior to the availability of site-directed mutagenesis, a large number of mutants of A. vinelandii and K. pneumoniae were created by chemical mutagenesis followed by selection for their inability to grow under dinitrogen-fixing conditions. Subsequently, several of these mutations were mapped to the nifH gene, and some were sequenced to identify their nifH genotype (28,(35)(36)(37)(38). Very recently, the structure of an Fe protein/MoFe protein complex from A. vinelandii became available (39), and we used that structure to search for those nifH mutations that were not located in any known region of importance for normal nitrogenase catalysis. Here we report that this information was used to construct a site-directed variant of the Fe protein, E146D, that is defective in its ability to participate in FeMo cofactor insertion but that participates normally in nitrogenase turnover and FeMo cofactor biosynthesis.

EXPERIMENTAL PROCEDURES
Unless otherwise noted, all chemicals and reagents were obtained from Fisher, Baxter Scientific, or Sigma.
Construction and Expression of Variant A. vinelandii Strains-A fragment of the A. vinelandii chromosome containing the entire nifH and a portion of the nifD genes was cloned into the bacteriophage M13mp18 (28). Site-directed mutagenesis was performed using the Mutagene mutagenesis kit, version 2, from Bio-Rad. The oligonucleotides used for mutagenesis were purchased from the Integrated DNA Technologies (Coralville, IA). They were both 32 bases long and were complementary to the region surrounding the Glu 146 codon. One oligonucleotide was degenerate at the Glu 146 codon, allowing for the production of several Fe protein mutants at this position, including the glutamate to lysine mutation. The other oligonucleotide was specific for the glutamate to aspartate mutation. After mutagenesis, bacteriophage containing mutated nifH genes were selected through DNA sequencing using the Sequenase II sequencing kit (U. S. Biochemical Corp.). Two mutated genes were selected, corresponding to E146K and E146D mutations. In each case double-stranded DNA was isolated from these phage using a Qiagen kit (Qiagen, Chatsworth, CA). This DNA was then transformed back into two A. vinelandii strains, the wild type Trans strain and the ⌬nifH strain DJ54, using a published method (40).
Here the chromosomal copy of the gene is replaced with the mutated gene through homologous recombination, allowing production of variant Fe protein in its native background, under the control of its native promoter. The recombination of the mutated nifH gene containing the Glu to Lys mutation into the bacterial chromosome of the Trans strain resulted in a strain that was not able to grow under dinitrogen-fixing conditions. The recombination of the gene containing the Glu to Asp mutation into the bacterial chromosome of the DJ54 strain resulted in a strain that grew slowly under dinitrogen-fixing conditions. All subsequent work was done on this DJ54-derived strain.
Cell Growth and Protein Purification-A. vinelandii strains expressing wild type and E146D Fe proteins were grown in 180-liter batches in a 200-liter New Bruswick fermentor under N 2 -fixing conditions on Burke's minimal media. The growth rate was measured by cell density at 436 nm using a Spectronic 20 Genesys (Spectronic Instruments, Rochester, NY). The cells were harvested at the end of the exponential phase using a flow through centrifugal harvester (Cepa, Germany). Published methods were used for the purification of wild type and E146D Fe proteins (41), wild type MoFe protein (42), and the MoFe protein synthesized by the strain expressing the E146D Fe protein (21). A. vinelandii strains E146K nifH and DJ54 (⌬nifH) were grown in 10-liter batches in a 12-liter New Brunswick fermentor on Burke's minimal medium supplemented with 2 mM ammonium acetate. After the consumption of the ammonia, the cells were derepressed for 3 h followed by harvesting using a Sorvall RC5C centrifuge (10,000 rpm, 30 min, 4°C). The cell paste was washed with 50 mM Tris-HCl, pH 8.0, and kept on dry ice until needed. Cell-free extracts were prepared from these strains by passing bacterial suspensions (10 g of wet weight in 20 ml of 50 mM Tris-HCl, pH 8.0) three times through a French pressure cell and centrifuging them (10,000 rpm, 30 min, 4°C) under anaerobic conditions.
Protein Characterization and Spectroscopy-All spectroscopic samples were prepared in a Vacuum Atmospheres dry box with an oxygen level of less than 1 ppm. With the exception of the all ferrous Fe protein, all reduced Fe protein, and MoFe protein samples that were characterized spectroscopically were 2 mM in Na 2 S 2 O 4 . Ti(III)citrate and the Ti(III)citrate reduced all ferrous Fe protein samples were prepared following published methods (43,44). For visible range absorption experiments reduced samples were prepared in anaerobic cuvettes that were previously blanked. Spectra were recorded on an HP 8452A Diode Array spectrophotometer. Visible region CD samples were prepared in the same manner, but in that case the Fe protein samples were first oxidized to the [4Fe-4S] 2ϩ state by passage over a specially prepared column as described elsewhere (29). Spectra were recorded on a Jasco spectrometer with a scan speed of 50 nm/min. Traces shown are the accumulation of five scans. All perpendicular and parallel mode EPR spectra were recorded on dithionite or Ti(III)citrate reduced samples using a Bruker ESP 300 E z spectrophotometer, interfaced with an Oxford Instruments ESR-9002 liquid helium continuous flow cryostat.
All nitrogenase activity assays (8.7-ml vials) were carried out as described previously (42). The products H 2 and C 2 H 4 were analyzed as published elsewhere (28). Ammonium was determined by a high performance liquid chromatography fluorescence method (45). Molybdenum (46) and iron (47) were determined as published elsewhere. The iron chelation assays were performed as described elsewhere (44) using a Fe protein concentration of 0.8 mg/ml.
FeMo Cofactor Biosynthesis and Insertion Assays-Experiments that monitored the insertion of FeMo cofactor into the FeMo cofactor-deficient MoFe protein synthesized by A. vinelandii ⌬nifH strain DJ54 were all done in the presence of excess isolated FeMo cofactor in Nmethyl formamide (NMF), 2 which was isolated as described elsewhere (42). The anaerobic mixture (argon gas) contained, in a 0.35-ml total volume, 25 mM Tris-HCl, pH 7.4, 1.7 mM ATP, 3.4 mM MgCl 2 , 20 mM creatine phosphate, 17 units of creatine phosphokinase, 20 mM Na 2 S 2 O 4 , 4 mg of DJ54 cell-free extract protein, and 45 g of purified wild type or E146D Fe protein. The reactions were started by injecting 2 l of concentrated FeMo cofactor in NMF. Following incubation at 30°C for the times indicated in the text, the insertion was stopped by addition of 40 nmol of (NH 4 ) 2 MoS 4 . Subsequently, the enzyme activity was measured by C 2 H 4 evolution. Those enzyme activity assays, containing 0.1 ml of the insertion mixture, were carried out as described previously (42). The assays designed to test for the presence of "free" or uninserted FeMo cofactor in E146D nifH cell-free extracts contained, in a 0.35-ml total volume, 25 mM Tris-HCl, pH 7.4, 20 mM Na 2 S 2 O 4 , and 0.15 mg of purified cofactor-deficient MoFe protein from ⌬nifB strain DJ1143 (48). The insertion was started by the addition of isolated FeMo cofactor in NMF or E146D nifH cell-free extracts. The samples were incubated at 30°C for 30 min, and subsequently, the enzyme activity of 0.1 ml of the insertion mixture was determined as described previously (42).
Assays designed to test the ability of the E146D Fe protein to participate in FeMo cofactor biosynthesis in vitro contained, in a 0. 35 (34,49), and the enzyme activity was determined as described previously (42). Homocitrate lactone was converted to the free acid as described elsewhere (50). 1 FeMo cofactor biosynthesis and insertion both appear to require MgATP hydrolysis and the binding of MgATP to the Fe protein. Many Fe protein variants that do not hydrolyze MgATP when bound to the MoFe protein still carry out FeMo cofactor biosynthesis and insertion. At present the possibility that these same mutants will hydrolyze MgATP when bound to another protein cannot be eliminated. 2 The abbreviation used is: NMF, N-methyl formamide.

RESULTS AND DISCUSSION
The Glu 146 Region of the Fe Protein-As discussed in the Introduction, the Fe protein of nitrogenase has at least three separate functions in the cell: electron donation to the MoFe protein to support N 2 reduction; participation in the initial biosynthesis of FeMo cofactor; and insertion of preformed FeMo cofactor in a FeMo cofactor-deficient MoFe protein. How the Fe protein participates in the last two reactions and what features of the Fe protein are required for those two reactions is not known. In this study we used computer modeling with the recently published structure of an Fe protein/MoFe protein complex (39) to try to locate conserved Fe protein residues that might be specifically involved only in one or both of the last two reactions. To do this we relied on information concerning the nifH genotypes of strains of K. pneumoniae or A. vinelandii that had been produced decades ago using chemical mutagenesis and that had Nif Ϫ phenotypes (35)(36)(37)(38). The idea was to look for residues that were not close to the [4Fe-4S] cluster or any of the regions of the Fe protein that were known to be important for MgATP binding, the MgATP-induced conformational change or for complex formation with the MoFe protein.
In this way Glu 146 was selected as a target for mutagenesis. Fig. 1 shows the region around Glu 146 in the Fe protein sequence. The residue is completely conserved in 57 known Fe proteins sequences, it is located in a highly conserved region of the protein, it is on the surface of the protein and part of ␤ sheet number 6, and it is not close to any regions known to be required for Fe protein participation in nitrogenase turnover (Fig. 1). Fig. 2 shows the position of the residue in the A. vinelandii complex structure relative to the [4Fe-4S] cluster, the nucleotide-binding region, the dimer interface, and the MoFe protein interaction surface. It is difficult to see from this figure why a mutation at this position would affect the ability of the protein to serve as an electron donor for the MoFe protein. Nonetheless one of the chemically generated nifH mutations that led to a Nif Ϫ phenotype was later shown to be the substitution of Glu by Lys at the homologous position, 147 in K. pneumoniae (36). Fig. 3A is a close up of Glu 146 showing again that this hydrophilic residue is on the surface of the protein and is salt-bridged to another surface residue, Arg 3 . Fig. 3B is a model of the nonconservative replacement by Lys, a mutation that introduces a positive charge at that position and that would by necessity break the salt bridge and cause a rearrangement of Arg 3 . 3 The chemically generated E147K K. pneumoniae NifH that led to a Nif Ϫ phenotype was never purified, but it was reported that it accumulated to lower levels in the cell than the wild type protein. This might be due to destabilization of the NH 2 terminus caused by breaking the salt bridge to Arg 3 . We constructed the homologous mutation E146K nifH in A. vinelandii and confirmed both the Nif Ϫ phenotype and the fact that it accumulated to lower levels in the cell making purification difficult (Fig. 4). In addition, because all three functions of the Fe protein are dependent on Fe protein concentration, it is 3 We have previously demonstrated by x-ray crystallography that breaking a salt bridge between surface Asp 15 and Lys 84 of another [Fe-S] protein, by construction of a D15N mutation, caused Lys 84 to rearrange substantially and find a new partner (56). difficult to determine to what extent the Nif Ϫ phenotype is due to a specific critical role played by Glu 146 rather than being due to simply having less Fe protein present in the cell.
To test the importance of Glu 146 per se we constructed a conservative mutation in which Glu was substituted by Asp leaving the negative charge at this position but decreasing the extent to which the residue protrudes into the solvent (Fig. 3C). In the absence of an x-ray structure, it is not known whether or not Arg 3 rearranges to retain the salt bridge. Fig. 4 shows however, that unlike the situation with E146K, the E146D and wild type Fe proteins do accumulate to the same levels in the cell, a fact that was also confirmed by purification yields (see below). Fig. 5 is a growth curve showing that nonetheless this conservative mutation leads to a slow growth phenotype under dinitrogen-fixing conditions with doubling time of 4 h versus 2 h for the strains expressing E146D and wild type Fe proteins, respectively.
The Slow Growth Phenotype Is Not Due to Defects in the Ability of the E146D Fe Protein to Serve as an Electron Donor to the MoFe Protein-To examine the ability of the E146D Fe protein to serve as an electron donor, the protein was purified and characterized. The protein was easily purified using the wild type procedure, and the final yields of protein were the same as those normally obtained from a wild type preparation (42). Fig. 6 compares the EPR spectra in the g ϭ 2 region exhibited by the wild type and E146D Fe proteins. These spectra are both qualitatively and quantitatively indistinguishable from each other, leading to the conclusion that the E146D Fe protein has a normal complement of [4Fe-4S] cluster and that the cluster can be reduced by dithionite to the [4Fe-4S] 1ϩ level.
Recent studies have shown that wild type Fe protein can be further reduced to an unprecedented all ferrous [4Fe-4S] 0 oxidation state in vitro (43,51) and have suggested that this state might be utilized in vivo (44,52). If that were the case, then a mutant Fe protein that is unable to achieve the 0 oxidation level should exhibit a slow growth phenotype. The all ferrous Fe protein exhibits a very distinctive UV-visible spectrum (43) The left side of A shows a ball and stick representation of Glu 146 (wild type) and some surrounding amino acids. The interaction between Glu 146 and Arg 3 is indicated by a dashed line between the terminal oxygen and nitrogen atoms. The right side of each panel shows a space filling model of the Fe protein surface with residues 146 and 3 shown as ball and stick representations. B and C show the hypothetical structures of the E146K and E146D mutants, respectively, in the region of the mutation. The structures used in these last two panels were generated by homology modeling the mutant sequences to existing Fe protein structures and subsequently energy minimizing the mutant structure using GROMOS96. This process is completely automated and is available at the SWISS-MODEL web site (60 -62). vinelandii. Both strains were grown in 10-liter batches in a 12-liter New Brunswick fermentor in Burke's minimal medium. In both cases the inocula was 200 ml, A 436 ϭ 1.6. Growth for the wild type (q) and E146D nifH A. vinelandii strain (E) was monitored by measuring the absorbance at 436 nm. that is observed for both the wild type and E146D Fe proteins following reduction by Ti(III) citrate to the all ferrous level (data not shown). Another distinctive feature of the all ferrous Fe protein is that it is S ϭ 4 and exhibits a g ϭ 16 EPR signal in the parallel mode (43). Fig. 7 compares the parallel mode EPR spectra exhibited by the wild type and E146D Fe proteins following reduction by Ti(III) citrate to the all ferrous level. Again these spectra are indistinguishable from each other, leading to the conclusion that the E146D Fe protein can be reduced by two electrons to the all ferrous level.
Following reduction of the [4Fe-4S] 2ϩ cluster, the next step in the nitrogenase reaction involves the Fe protein binding MgATP and undergoing a global conformational change (1,2,4). This reaction can be easily monitored using a chelation assay because in the absence of MgATP (or the presence of MgADP) the [4Fe-4S] 1ϩ or [4Fe-4S] 0 clusters are resistant to chelation by bathophenanthroline disulfonate, whereas in the presence of MgATP iron is rapidly removed from the protein (53,54). Many mutant Fe proteins have been described that are either blocked in this reaction or show altered chelation kinetics (e.g. 1,4,29,41). Fig. 8 compares the behavior of the wild type and E146D Fe proteins in a standard chelation assay. Again the results are indistinguishable, leading to the conclusion that the E146D Fe protein is not defective in its ability to bind MgATP and undergo the conformational change. The altered MgATP conformation has also been examined by EPR spectroscopy (Fig. 6), and again there are no differences observed between the wild type and E146D Fe proteins in these reactions.
Once the Fe protein is in the MgATP conformation, it is positioned to form a complex with the MoFe protein. When that occurs, the MoFe protein covers up the [4Fe-4S] cluster of the Fe protein, protecting it from chelation (39,53). Fig. 8 shows again that there are no observed differences in the behavior of the wild type and E146D Fe proteins in this chelation protection assay. Thus the slow growth phenotype cannot be attributed to an inability of E146D Fe protein to bind normally to the MoFe protein, an observation that is consistent with the location of the residue (Fig. 2). Following the formation of a productive complex, an electron is transferred from the Fe protein to the MoFe protein in a reaction coupled to MgATP hydrolysis. This is followed by complex dissociation, and the Fe protein is left in a conforma-

FIG. 6. EPR spectra of purified wild type (A, C, and E) and E146D Fe protein (B, D, and F) without nucleotide (A and B) and in the presence of MgATP (C and D) or MgADP (E and F).
A final concentration of 6.25 mM ADP or ATP and 13 mM MgCl 2 was used in the measurements containing nucleotide. All protein concentrations were 15 mg/ml, and the solutions contained 2 mM sodium dithionite. The spectra were recorded at 13 K using a microwave power of 20 mW, a microwave frequency of 9.43 GHz, and a gain of 5 ϫ tion with MgADP bound (1). Figs. 6 and 9 compare the MgADP conformations of the wild type and E146D Fe proteins using EPR and CD spectroscopies, respectively. Again the spectra are indistinguishable. Following repeated cycles of electron transfer, the MoFe protein accumulates enough electrons to reduce substrates. Table I shows that the specific activities of the E146D and wild type Fe proteins are within experimental error of each other not only in the H 2 evolution and C 2 H 2 reduction assays but also in the N 2 fixation assay. Thus the slow growth phenotype of the strain expressing E146D Fe protein cannot be explained by its inability to serve normally as an electron donor to the MoFe protein.
The Slow Growth Phenotype Can Be Explained by Defects in the MoFe Protein-Having established that the slow growth phenotype was not due to a defect in the first function of the Fe protein, we went on to examine whether or not the E146D Fe protein might be defective in FeMo cofactor biosynthesis and/or insertion. The MoFe protein synthesized by that strain was therefore purified and characterized. Once the pure protein was available, it was immediately obvious that the MoFe protein synthesized by the strain expressing E146D Fe protein was lighter in color than wild type MoFe protein, indicating that it did not have a full complement of metal clusters. Fig. 10 compares the EPR signals exhibited by wild type MoFe protein and MoFe protein from E146D nifH when the proteins are in the presence of excess dithionite. This well characterized S ϭ 3/2 EPR signal arises from the FeMo cofactor center of the MoFe protein (1, 7). Spin integration shows that the MoFe protein synthesized by the E146D nifH strain accumulates only 54% as much FeMo cofactor as the wild type.
This lower amount of FeMo cofactor is also consistent with the observed metal content of 0.83 Ϯ 0. cannot distinguish whether all molecules of MoFe protein are missing one FeMo cofactor or whether a statistical distribution is present with some holo protein and some totally FeMo cofactor-deficient protein. We think the latter is unlikely, however, because the FeMo cofactor-deficient MoFe protein from a NifH Ϫ strain is very difficult to purify and can only be obtained in small yield (21), whereas good yields of MoFe protein were obtained from the strain expressing E146D nifH. The independent assembly of the two halves of the protein is also not unprecedented and has been demonstrated for the vanadium nitrogenase system (55).
It is known that FeMo cofactor-deficient MoFe proteins synthesized by NifB Ϫ and NifH Ϫ strains are not the same (20). In this case we expect that the MoFe protein molecules that are missing one FeMo cofactor should be of the NifH Ϫ and not the NifB Ϫ type. This is confirmed by Fig. 11, which shows that the MoFe protein synthesized by the E146D nifH strain is an ␣ 2 ␤ 2 tetramer, like the FeMo cofactor-deficient MoFe protein from a NifH Ϫ strain (21) and not an ␣ 2 ␤ 2 ␥ 2 hexamer like the FeMo cofactor-deficient MoFe protein from NifB Ϫ strains (27).
As expected, this lower FeMo cofactor content correlates with lower MoFe protein specific activity such that the MoFe protein synthesized by the E146D nifH strain is ϳ55% as active as the wild type MoFe protein not only for H 2 evolution and C 2 H 2 reduction but also for N 2 fixation (Table II). This reduction in activity further correlates with the growth rate of the strain (Fig. 5), explaining the slow growth phenotype.
Evidence That the E146D Fe Protein Can Function Normally in FeMo Cofactor Biosynthesis-As discussed above, the MoFe protein synthesized by the E146D nifH strain accumulates only about half as much FeMo cofactor as the protein from wild type cells leading to a slow growth phenotype. One possibility that would explain that result would be that the E146D Fe protein was defective in its ability to function in the initial biosynthesis of FeMo cofactor such that FeMo cofactor biosynthesis could not keep up with protein synthesis/assembly. Table III shows, however, that we were unable to distinguish E146D Fe protein and wild type Fe protein in an in vitro biosynthesis assay, making this possibility seem unlikely.
In addition, it has been known for many years that in the   10. EPR spectra of dithionite reduced wild type (A) and E146D nifH MoFe protein (B). All protein concentrations were 5 mg/ml. The spectra were measured at 10 K using a microwave power of 50 mW, a microwave frequency of 9.43 GHz, and a gain of 5 ϫ 10 4 .
absence of the MoFe protein polypeptides in vivo small amounts of FeMo cofactor accumulate on some other protein (14 -16). We reasoned that if the E146D Fe protein were not defective in the initial biosynthesis of FeMo cofactor, then the cell-free extracts might contain small amounts of FeMo cofactor that had not been inserted into the MoFe protein. To test this we added purified FeMo cofactor-deficient MoFe protein from a NifB Ϫ strain to try to capture this free cofactor. It is known that the insertion of FeMo cofactor into that protein does not require the Fe protein (27,48), and therefore the presence of the E146D Fe protein in the cell-free extracts should not interfere. As shown in Table IV, the inactive NifB Ϫ FeMo cofactordeficient MoFe protein could be activated either by addition of isolated FeMo cofactor in NMF or by free FeMo cofactor contained within the E146D nifH, cell-free extracts. Thus, free FeMo cofactor does accumulate in the E146D strain leading to the conclusion that the E146D Fe protein is not defective in the initial biosynthesis of FeMo cofactor.

The E146D Fe Protein Is Defective in FeMo Cofactor
Insertion-Strains that do not synthesize the Fe protein accumulate a FeMo cofactor-deficient form of the MoFe protein that can be activated in vitro, in cell-free extracts, by addition of isolated FeMo cofactor in NMF (15, 18 -21). Previous studies have established that this "insertion" reaction also requires the Fe protein and MgATP and at least one additional protein, which has been designated gamma in A. vinelandii (18 -27). One role of the Fe protein and MgATP appears to be to promote the association of gamma with the FeMo cofactor-deficient MoFe protein in such a way that the protein is converted from a form with the FeMo cofactor site buried to one with the FeMo cofactor site accessible for FeMo cofactor insertion (26). Fig. 12 is a time course for the insertion reaction where either wild type Fe protein or E146D Fe protein were added to cell-free extracts from ⌬nifH strain DJ54 along with MgATP and isolated FeMo cofactor. The results show that E146D Fe protein is obviously defective in the insertion of isolated FeMo cofactor into the FeMo cofactor-deficient MoFe protein. The shape of the insertion curve indicates that the E146D Fe protein is specifically FIG. 12. Time course of FeMo cofactor insertion by purified wild type (q) and E146D (E) Fe proteins into FeMo cofactordeficient MoFe protein present in the extracts of ⌬nifH strain DJ54. As previously published (18,19), the insertion required the addition of Fe protein and MgATP. Excess isolated FeMo cofactor in NMF was inserted into the FeMo cofactor-deficient MoFe protein of ⌬nifH strain DJ54 using extracts. The insertion assays were performed as described under "Experimental Procedures" and stopped at the indicated time points by adding (NH 4 ) 2 MoS 4 (34,50). After the insertion, the enzyme activity was determined by measuring C 2 H 4 evolution using the standard activity assay.
FIG. 11. SDS-polyacrylamide gel electrophoresis of purified wild type and E146D nifH MoFe protein. Lane 1, protein standard (20 g); lane 2, purified wild type MoFe protein (4 g); lane 3, purified E146D nifH MoFe protein (4 g). A molecular weight of 23,000 -26,000 was reported for the monomeric gamma protein from SDS-polyacrylamide gel electrophoresis (26). The molecular weights of the native wild type and of the E146D nifH MoFe protein appeared to be identical based on the criteria of the elution profile of a gel filtration Ultrogel AcA34 (ICF, France) column.  defective in the initial phase of the reaction. After this barrier is overcome, the insertion can carry on at a higher rate. Previous studies would suggest that this early phase of the reaction may involve the Fe protein mediated association of the FeMo cofactor-deficient MoFe protein with gamma (26).
Conclusion-The Fe protein of nitrogenase has three functions in the cell: 1) electron donation to the MoFe protein; 2) the initial biosynthesis of FeMo cofactor; and 3) the insertion of preformed FeMo cofactor into the FeMo cofactor-deficient MoFe proteins synthesized by ⌬nifH strains. In this study we have used sequence and structural information to identify a residue that might be involved in only the second or third functions. The above data establish that the E146D Fe protein is competent in electron transfer and the initial biosynthesis of FeMo cofactor but is specifically blocked for the first phase of the insertion reaction. Thus, the cells expressing the E146D Fe protein accumulate uninserted FeMo cofactor and partially FeMo cofactor-deficient MoFe protein. This is the first mutant that has been described that is only defective in this reaction. Glu 146 is a negatively charged surface residue. When it is converted to the smaller aspartate, the protein is partially active in the insertion reaction, whereas when it is converted to a positively charged lysine, a Nif Ϫ phenotype is obtained. When combined with the information that the initial phase of the insertion reaction involves Fe protein/MgATP/gamma-mediated modification of the FeMo cofactor-deficient MoFe protein (26), the data suggest that Glu 146 is likely to be critical for specific protein-protein interactions. Future studies will be directed toward elucidating the details of those reactions.