Molecular Insights into Nitrogenase FeMoco Insertion

Biosynthesis of the FeMo cofactor (FeMoco) of nitrogenase MoFe protein is arguably one of the most complex processes in metalloprotein biochemistry. Here we investigate the role of a MoFe protein residue (Trp-α444) in the final step of FeMoco assembly, which involves the insertion of FeMoco into its binding site. Mutations of this aromatic residue to small uncharged ones result in significantly decreased levels of FeMoco insertion/retention and drastically reduced activities of MoFe proteins, suggesting that Trp-α444 may lock the FeMoco tightly in its binding site through the sterically restricting effect of its bulky, aromatic side chain. Additionally, these mutations cause partial conversion of the P-cluster to a more open conformation, indicating a potential connection between FeMoco insertion and P-cluster assembly. Our results provide some of the initial molecular insights into the FeMoco insertion process and, moreover, have useful implications for the overall scheme of nitrogenase assembly.

Nitrogenase is a multicomponent metallo-enzyme that catalyzes the reduction of atmospheric dinitrogen to bioavailable ammonia (for recent reviews see Refs. [1][2][3][4][5][6][7][8]. The best studied, Mocontaining nitrogenase of Azotobacter vinelandii is composed of two proteins, the Fe protein and the MoFe protein. The homodimeric Fe protein (Av2) 2 has one ATP binding site per subunit and a single [4Fe-4S] cluster bridged between the subunits; whereas the ␣ 2 ␤ 2 tetrameric MoFe protein (Av1) contains, in each ␣␤ subunit pair, two complex metal clusters. One, designated the P-cluster, is an [8Fe-7S] cluster (9) that is located at the ␣␤-interface and coordinated by six Cys ligands. The other, designated FeMoco, is a [Mo-7Fe-9S-X] 3 cluster that is situated within the ␣-subunit and bound to His-␣442, Cys-␣275 and an endogenous homocitrate ligand (10). These metal clusters are essential for nitrogenase reactivity, mediating the ATP-driven, interprotein electron transfer from the [4Fe-4S] cluster of the Fe protein to the P-cluster of the MoFe protein and finally to the FeMoco where substrate reduction takes place.
Biosynthesis of P-cluster and FeMoco has attracted considerable attention, because these clusters are biologically unique and chemically unprecedented. Both P-cluster and FeMoco are composed of smaller substructures: the P-cluster consists of two [4Fe-4S] subclusters sharing a 6 -sulfide (9); and the FeMoco comprises [Mo-3Fe-3S] and [4Fe-3S] subcubanes sharing three 2 -sulfides and a central 6 -atom (10). Nonetheless, while the P-cluster is likely assembled through fusion of its substructural [4Fe-4S] units in its target location (11)(12)(13), FeMoco is first assembled on a scaffold NifEN protein and then inserted into its binding site in the MoFe protein (14 -17). Biosynthesis of FeMoco presumably starts with the production of an Fe/S core by NifB (encoded by nifB), which is then transferred to, and further processed on the NifEN complex (14 -19). Previously, we identified a molybdenum-free, NifEN-bound FeMoco precursor that bears a striking resemblance to the Fe/S core of the mature FeMoco (15). Most recently, we showed that, upon MgATP hydrolysis, Fe protein inserts molybdenum and homocitrate into the FeMoco precursor while it is still bound to NifEN, resulting in the formation of a fully complemented cluster that can be subsequently inserted into its final location in the MoFe protein (16,17). We also established that the transfer of the cluster from NifEN to MoFe protein likely occurs through direct protein-protein interactions (16,17). These findings not only provide important insights into the biosynthesis and biomimetic chemical synthesis of FeMoco, they may also bear useful implications for the biosynthetic mechanism of other complex metal-containing clusters (20 -22). Though progress has been made toward understanding the assembly of FeMoco prior to its insertion into the FeMoco binding site, little is known about the final step of FeMoco incorporation into its target location within the MoFe protein.
Our current knowledge in this regard is largely based upon the crystal structure of a ⌬nifB MoFe protein (Av1 ⌬nifB ) from a nifB-deletion strain of A. vinelandii (23). Consistent with the hypothesis that nifB encodes for a protein that is involved in the biosynthesis of FeMoco, Av1 ⌬nifB is FeMoco-deficient, yet it contains normal P-clusters (23). In vitro, Av1 ⌬nifB can be reconstituted, without additional factors, into an active holoprotein by isolated FeMoco (23,24). Given that the FeMoco binding site is fully buried at ϳ10 Å below the surface of the wild-type MoFe protein (Av1 wild-type ), this observation indicates that Av1 ⌬nifB undergoes significant conformational rearrangements relative to its holo counterpart, a process that renders its FeMoco site open and allows the subsequent insertion of isolated FeMoco (23). Indeed, when compared with Av1 wild-type , the ␣III domain of Av1 ⌬nifB has some major structural changes in that nearly all ␤-strands are shorter toward their C termini while ␣-helices A, C, and D are shorter toward their N termini. These conformational alterations create a positively charged insertion funnel in Av1 ⌬nifB , notably missing in its holo counterpart, that presumably steers the negatively charged FeMoco all the way down the funnel into its final location (23). A closer look at the FeMoco binding sites of Av1 wild-type and Av1 ⌬nifB (Fig. 1) reveals that, in the case of Av1 ⌬nifB , the C ␣ of His-␣442 shifts ϳ5 Å during the rearrangement of ␣III domain and joins two other residues, His-␣274 and His-␣451, in the formation of a striking His triad (23). Coupled to this rearrangement, residues His-␣442 and Trp-␣444 switch their relative positions (23). Additionally, substantial changes take place in non-liganding residues as well; particularly, the stretch from ␣355 to ␣359, with Gly-␣356, Gly-␣357, and Arg-␣359 that normally form hydrogen bonds to FeMoco sulfurs in holo Av1 (23). These residues are likely the key players in the process of FeMoco insertion and, therefore, serve as ideal targets for us to tackle the mechanism of the final step of FeMoco assembly on a molecular basis.
In the current study we examine the role of Trp-␣444 of Av1 in FeMoco insertion by combined mutational, biochemical, and spectroscopic approaches. Our data show that substitution of this large aromatic residue for small uncharged ones results in a drastically decreased level of FeMoco insertion/retention, thus suggesting that the sterically prohibiting Trp-␣444 may function in the FeMoco insertion process by tightly locking the FeMoco into its binding site.

Construction of Variant A. vinelandii Strains-
Cell Growth and Protein Purification-All A. vinelandii strains were grown in 180-liter batches in a 200-liter New Brunswick fermentor (New Brunswick Scientific, Edison, NJ) in Burke's minimal medium supplemented with 2 mM ammonium acetate. The growth rate was measured by cell density at 436 nm using a Spectronic 20 Genesys (Spectronic Instruments, Westbury, NY). After the consumption of ammonia, the cells were de-repressed for 3 h followed by harvesting using a flow-through centrifugal harvester (Cepa, Lahr/Schwarzwald, Germany). The cell paste was washed with 50 mM Tris-HCl (pH 8.0). Published methods were used for the purification of all Av2 proteins (29) and His-tagged Av1 proteins (12,28).
EPR Spectroscopy-All EPR samples were prepared in a Vacuum Atmospheres dry box (Hawthorne, CA) with an oxygen level of less than 4 ppm. Unless noted otherwise, all samples were in 25 mM Tris-HCl, pH 8.0, 10% glycerol and 2 mM Na 2 S 2 O 4 . Av1 protein samples were oxidized by incubation with excess indigo disulfonate (IDS) for 30 min. Subsequently, IDS was removed by a single passage over an anion exchange column as described elsewhere (30). All perpendicular and parallel mode EPR spectra were recorded using a Bruker ESP 300 E z spectrophotometer (Bruker, Billerica, MA), interfaced with an Oxford Instruments ESR-9002 liquid helium continuous flow cryostat (Oxford Instruments, Oxon, UK). All spectra were recorded at 10 K using a microwave power of 50 milliwatt, a gain of 5 ϫ 10 4 , a modulation frequency of 100 kHz, and a modulation amplitude of 5 Gauss. The microwave frequencies of 9.62 and 9.39 GHz were used for the perpendicular (10 scans) and parallel (20 scans) mode EPR spectra, respectively. Spin quantitation of EPR signals was carried out as described in detail earlier (28).
Activity Assays and Metal Analysis-All nitrogenase activity assays were carried out as described previously (31). The products H 2 and C 2 H 4 were analyzed as published elsewhere (32). Ammonium was determined by a high performance liquid chromatography fluorescence method (33). FeMoco maturation assays were carried out as published earlier (14). Molybdenum (34) and iron (35) were determined as published elsewhere.
Protein Stability Experiments-Two approaches were used to determine the stability of the purified Av1 proteins: (i) heat treatment and (ii) prolonged storage at room temperature. A total amount of 100 mg of purified Av1 (in 25 mM Tris-HCl (pH 8.0), 10% glycerol, 250 mM NaCl, and 2 mM Na 2 S 2 O 4 ) was incubated in a crimped anaerobic vial (volume, 8.7 ml; gas atmosphere, 100% Ar) at 56°C for 30 s (i) or stored at room temperature for 8 h (ii). Precipitated protein was subsequently removed in both cases by centrifugation at 10,000 rpm for 10 min (Biofuge fresco, Heraeus, Germany) in a Vacuum Atmospheres dry box. Concentrations of Av1 proteins were subsequently determined using Bio-Rad protein assay, and enzymatic activities of these proteins were determined as described above.

RESULTS AND DISCUSSION
Trp-␣444 is a highly conserved (ϳ80%) residue among the currently known MoFe proteins from various organisms; the rest of them contain Tyr (ϳ20%) in place of Trp. 4 The 100% natural  AvYM18A W444A-nifD His-tagged MoFe protein with site-directed W444A-nifD mutation Av1 W444A-nifD Av2 W444A-nifD YM19A AvYM19A W444F-nifD His-tagged MoFe protein with site-directed W444F-nifD mutation Av1 W444F-nifD Av2 W444F-nifD YM20A AvYM20A W444G-nifD His-tagged MoFe protein with site-directed W444G-nifD mutation His-tagged MoFe protein, deletion of nifB gene Av1 ⌬nifB -d a The component proteins of the Mo-nitrogenase are designated by the initials of the organism from which they are isolated and the arabic numeral of the component. For example, the MoFe protein and Fe protein of the Mo-nitrogenase of A. vinelandii are designated Av1 and Av2, respectively. b Note that, although Av1 and Av2 from the same variant strain have the same superscript, the site-directed mutations are located in Av1, not Av2. c Note that the Av2-encoding nifH gene is deleted in AvDJ1165, hence no expression of Av2 in this case. d Av2 of AvDJ1143 is not used in this study.
Under N 2 -fixing conditions, 5 the doubling time of AvYM19A W444F-nifD or AvYM21A W444Y-nifD , like that of AvYM13A wild-type , is ϳ5 h (Fig. 2). However, the doubling time of AvYM18A W444A-nifD is ϳ12 h, which is significantly longer compared with that of AvYM13A wild-type (Fig. 2). In the case of AvYM20A W444G-nifD , only a negligible amount of cell growth is detected (Fig. 2). Given that cell growth under N 2 -fixing conditions is proportionately correlated with the amount of active nitrogenase, these observations indicate that a mutation of the aromatic Trp residue at ␣444 to a small uncharged residue (in the case of AvYM18A W444A-nifD or AvYM20A W444G-nifD ) results in a significantly diminished enzymatic activity of nitrogenase; whereas nitrogenase reactivity is largely unaffected when the Trp-␣444 is replaced by another aromatic residue (in the case of AvYM19A W444F-nifD or AvYM21A W444Y-nifD ). Apparently, the replacement of Trp-␣444 with Gly has a more dramatic effect than that with Ala, which is not surprising, considering that Gly is a smaller residue than Ala. Meanwhile, although the doubling time of AvYM19A W444F-nifD or AvYM21A W444Y-nifD is roughly the same as that of AvYM13A wild-type , compared with AvYM13A wild-type , AvYM21A W444Y-nifD reaches nearly the same cell density while AvYM19A W444F-nifD achieves a slightly reduced cell mass, suggesting that the change of Trp-␣444 to Phe (no natural occurrence) has a trifle more significant impact on the enzyme activity than that to Tyr (20% natural occurrence). This could be explained by the fact that the side chain of Phe is slightly smaller than that of the Tyr or Trp.
The decreased enzymatic activities do not originate from the Av2 proteins of these ␣444 variant strains. 6 An approximate amount of 400 mg of non-tagged Av2 was purified from 200 g of cells of AvYM21A W444Y-nifD , AvYM20A W444F-nifD , AvYM18A W444A-nifD , or AvYM19A W444G-nifD , as was from AvYM13A wild-type (data not shown), 7 suggesting that Av2 expression is unaffected in these variant strains. The monomer of Av2 W444Y-nifD , Av2 W444F-nifD , Av2 W444A-nifD or Av2 W444G-nifD , like that of Av2 wild-type , is ϳ30 kDa (Fig. 3A). The molecular masses of all these Av2 proteins are ϳ60 kDa based on their elution profiles on gel filtration Sephacryl S-200 high resolution columns (data not shown), indicating that they are all homodimers. Compared with their wild-type counterpart, all Av2 proteins of the ␣444 variant strains have approximately the same metal content of 4 mol Fe/mol protein (Table 2) and exhibit, in the dithionite-reduced state, the same characteristic 5 Only organisms capable of expressing active nitrogenase enzymes are able to grow under the N 2 -fixing conditions, where no fixed or organic nitrogen (such as ammonia and urea) is present in the culture media. 6 Although point mutations of Av1 have no apparent effect on Av2, it has been well established that the latter is involved in FeMoco biosynthesis and insertion (11). Because this work deals with FeMoco insertion into Av1 variants, the activities of which are tested by assays that involve Av2, it is crucial to show that the Av2 of the variant strains have normal cluster compositions and function normally in both catalytic and biosynthetic capacities. 7 Note that for protein purification, all strains were grown in the presence of a limited amount of ammonia, so that a certain amount of cell mass could be accumulated regardless of the ability of the strain to fix nitrogen. After the consumption of ammonia, cells were derepressed for 3 h, during which process the nitrogenase protein (active or inactive) was expressed. Under these conditions, a consistent yield of ϳ200 g of cells per 180 liters of cell growth batch could be obtained for the wild-type and variant strains of A. vinelandii.   OCTOBER 13, 2006 • VOLUME 281 • NUMBER 41 S ϭ 1/2 EPR signal of rhombic line shape in the g ϭ 2 region (Fig. 4) Table 3), suggesting that they are all fully proficient in their catalytic capacities. Additionally, these Av2 proteins all function as normally as the wild-type protein in the FeMoco maturation assay (Table 3), indicating that they are perfectly competent in FeMoco assembly, a second function that has been definitively assigned to this multitask component of nitrogenase recently (14,16,17). Like the Av2 proteins, the Av1 proteins of the Trp-␣444 variant strains are expressed at practically the same level as Av1 wild-type , as evidenced by Western blot analysis of Av1specific antibody against the crude extracts of these variant strains under non-saturated conditions (supplemental Fig. S1). In addition, ϳ600 mg of His-tagged Av1 was purified from 200 g of cells of AvYM21A W444Y-nifD , AvYM20A W444F-nifD , AvYM18A W444A-nifD or AvYM19A W444G-nifD , as was from AvYM13A wild-type (data not shown), 7 further supporting the notion that Av1 expression is unperturbed with point mutations at ␣444 of Av1. Moreover, like Av1 wild-type , all these ␣444 Av1 variants are composed of ␣ (ϳ56 kDa)-and ␤ (ϳ59 kDa)subunits (Fig. 3B) and have the same, ␣ 2 ␤ 2 -tetrameric molecular mass of ϳ230 kDa based on their elution profiles on gel filtration Sephacryl S-200 columns (data not shown). Given the presence of nearly the same amount of Av1 in the variant strains, the diminished nitrogenase activity, particularly in the case of AvYM18A W444A-nifD or AvYM20A W444G-nifD , must arise from a decrease in the activity of the variant Av1 protein.
EPR analyses of the Av1 variant proteins provide further support to our argument that a mutation of Trp-␣444 to a small uncharged residue results in a decreased amount of FeMoco accumulation in Av1. In the dithionite-reduced state, Av1 wild-type (Fig. 5, trace 1) exhibits a distinct, FeMocospecific, S ϭ 3/2 signal (1). This S ϭ 3/2 signal is also observed in the case of Av1 W444Y-nifD (Fig. 5, trace 2), Av1 W444F-nifD (Fig. 5, trace 3), or Av1 W444A-nifD (Fig. 5, trace  4), although spin integration data indicate that the intensities of the signals are 98, 80, and 21% of that of Av1 wild-type for Av1 W444Y-nifD , Av1 W444F-nifD , and Av1 W444A-nifD , respectively (Table 4). In the case of Av1 W444G-nifD , this characteristic S ϭ 3/2 signal is virtually non-existent, integrating to 1% of that of Av1 wild-type (Fig. 5, trace 5). The spin integration data from these EPR experiments align well with those from the growth curves (Fig. 2), metal analyses ( Table 2) and activity assays (Table 4), once again confirming the presence of less FeMoco in Av1 variant proteins with mutations from aromatic Trp-␣444 to small uncharged residues like Ala and Gly. Thus, all results (summarized in supplemental Fig. S2) come together and corroborate with our theory that Trp-␣444 is specifically involved in FeMoco insertion, that the large size and steric rigidity of Trp-␣444 is crucial for locking FeMoco into its binding site, and that removal of this restraint compromises the ability of the Av1 to insert and/or retain FeMoco.
Interestingly, in the dithionite-reduced state, Av1 W444A-nifD (Fig. 5, trace 4) and Av1 W444G-nifD (Fig. 5, trace 5) show additional S ϭ 1/2 signals that integrate to 0.08 and 0.1 spin per protein, respectively. This particular S ϭ 1/2 signal has been previously assigned to a P-cluster analog comprising two [4Fe-4S]-like fragments, which may represent a physiologi- EPR samples (20 mg/ml) were prepared and measured as described under "Experimental Procedures." The FeMoco-specific, S ϭ 3/2 EPR signals of Av1 W444Y-nifD , Av1 W444F-nifD , Av1 W444A-nifD , and Av1 W444G-nifD integrate to 98, 80, 21, and 1% of that of Av1 wild-type , respectively. The S ϭ 1/2 EPR signals in the g Ϸ 2 region of the spectra of Av1 W444A-nifD and Av1 W444G-nifD integrate to 0.08 and 0.1 spin per Av1 wild-type , respectively.  (12,13,37). Consistent with this observation, in the IDS-oxidized state, Av1 variant proteins exhibit P-cluster specific (P 2ϩ state), g ϭ 11.8, parallel mode EPR signals (38,39), which integrate to 106, 99, 92, and 93% of that of Av1 wild-type (Fig. 6, trace 1) for Av1 W444Y-nifD (Fig. 6, trace 2), Av1 W444F-nifD (Fig.  6, trace 3), Av1 W444A-nifD (Fig. 6, trace 4) and Av1 W444G-nifD (Fig. 6, trace 5), respectively. The S ϭ 1/2 signal-associated analog of P-cluster, therefore, could account for the missing portion of P-cluster in Av1 W444A-nifD or Av1 W444G-nifD . Apparently, when Trp-␣444, one of the protein residues in a stretch of polypeptide that reorients with FeMoco insertion, is mutated to small uncharged residues, the [8Fe-7S] P-cluster is converted, partially, to a more open conformation consisting of a pair of [4Fe-4S]-like clusters. It is likely that the mutation of Trp-␣444 to Ala or Gly leads to a conformational rearrangement that in turn affects the P-cluster site and forces the substructural units of P-cluster apart. This observation, therefore, suggests a potential association between the insertion of FeMoco and the assembly of P-clus-ter and, perhaps more significantly, places these seemingly independent events in an overall, interactive scheme of nitrogenase biosynthesis. It is important to note that, the mutations at Trp-␣444 do not affect the overall stability of the Av1 variant proteins in this study. Like Av1 wild-type , variant proteins Av1 W444Y-nifD , Av1 W444F-nifD , Av1 W444A-nifD , and Av1 W444G-nifD all exhibit considerable stability upon heat treatment or prolonged storage at room temperature (Table 5). In addition, in the presence of MgATP, each of these ␣444 Av1 variants can form a complex with Av2 wild-type that is similarly stable compared with that formed between Av1 wild-type and Av2 wild-type (supplemental Fig. S3). Furthermore, structural predictions based on program MUMBO (40) 8 show that the Trp-␣444 mutations in this study present no steric problems that could lead to the instability of the Av1 protein (data not shown).
In summary, using a combined mutational/bio-chemical/ spectroscopic approach, we show that the Trp-␣444 of Av1 is specifically involved in the FeMoco insertion/retention. Mutations of this aromatic residue to small uncharged ones result in dramatically decreased levels of FeMoco insertion/retention and drastically reduced activities of the Av1 proteins, suggesting that Trp-␣444 may lock the FeMoco tightly into its binding site through the sterically restricting effect of its bulky, aro-FIGURE 6. Parallel mode EPR spectra of IDS-oxidized Av1 wild-type (trace 1), Av1 W444Y-nifD (trace 2), Av1 W444F-nifD (trace 3), Av1 W444A-nifD (trace 4), and Av1 W444G-nifD (trace 5). EPR samples (20 mg/ml) were prepared and measured as described under "Experimental Procedures." The P-cluster specific (P 2ϩ state), g ϭ 11.8, parallel mode EPR signals of Av1 W444Y-nifD , Av1 W444F-nifD , Av1 W444A-nifD , and Av1 W444G-nifD integrate to 106, 99, 92, and 93% of that of Av1 wild-type , respectively. a Protein in mg, determined before and after heat treatment or prolonged storage as described under "Experimental Procedures." b Specific activity in nmol of C 2 H 4 evolution/min/mg protein. c The FeMoco-deficient, P-cluster precursor-containing Av1 ⌬nifH was used as a negative control in these experiments, which showed 90% and 64% of degradation, respectively, upon heat treatment and prolonged storage. d Av1 protein samples were incubated at 56°C for 30 seconds as described under "Experimental Procedures." e Av1 protein samples were stored at room temperature for 8 hours as described under "Experimental Procedures."

Trp-␣444 of MoFe Protein Is Essential for FeMoco Insertion
matic side chain. Additionally, these mutations cause partial conversion of the P-cluster to a more open conformation, indicating a potential connection between FeMoco insertion and P-cluster assembly. Our results provide some of the initial insights into the process of FeMoco insertion on the molecular basis. Future studies will focus on addressing the effects of other key residues involved in FeMoco insertion and investigating the link between the assembly processes of FeMoco and P-cluster, in hope of elucidating a more detailed mechanism of nitrogenase biosynthesis.