Evidence for the Selective Population of FeMo Cofactor Sites in MoFe Protein and Its Molecular Recognition by the Fe Protein in Transition State Complex Analogues of Nitrogenase*

We have collected synchrotron x-ray solution scattering data for the MoFe protein of Klebsiella pneumoniae nitrogenase and show that the molecular conformation of the protein that contains only one molybdenum per a 2 b 2 tetramer is different from that of the protein that has full occupancy i.e. two molybdenums per molecule. This structural finding is consistent with the existence of MoFe protein molecules that contain only one FeMo cofactor site occupied and provides a rationale for the 50% loss of the specific activity of such preparations. A stable inactive transition state complex has been shown to form in the presence of MgADP and AlF 4 2 . Gel filtration chromatography data show that the MoFe protein lacking a full complement of the cofactor forms initially a 1:1 complex before forming a low affinity 1:2 complex. A similar behavior is found for the MoFe protein with both cofactors occupied, but the high affinity 1:2 complex is formed at a lower ratio of Fe protein/MoFe protein. The 1:1 complex, MoFe protein-Fe protein z (ADP z AlF 4 2 ) 2 , formed with MoFe protein that lacks one of the cofactors, is stable. X-ray scattering studies of this complex have enabled us to obtain its low resolution structure at ; 20-Å resolution, which confirms the gel filtration finding that only one molecule of the Fe protein

We have collected synchrotron x-ray solution scattering data for the MoFe protein of Klebsiella pneumoniae nitrogenase and show that the molecular conformation of the protein that contains only one molybdenum per ␣ 2 ␤ 2 tetramer is different from that of the protein that has full occupancy i.e. two molybdenums per molecule. This structural finding is consistent with the existence of MoFe protein molecules that contain only one FeMo cofactor site occupied and provides a rationale for the 50% loss of the specific activity of such preparations. A stable inactive transition state complex has been shown to form in the presence of MgADP and AlF 4 ؊ . Gel filtration chromatography data show that the MoFe protein lacking a full complement of the cofactor forms initially a 1:1 complex before forming a low affinity 1:2 complex. A similar behavior is found for the MoFe protein with both cofactors occupied, but the high affinity 1:2 complex is formed at a lower ratio of Fe protein/MoFe protein. The 1:1 complex, MoFe protein-Fe protein ⅐(ADP⅐AlF 4 ؊ ) 2 , formed with MoFe protein that lacks one of the cofactors, is stable. X-ray scattering studies of this complex have enabled us to obtain its low resolution structure at ϳ20-Å resolution, which confirms the gel filtration finding that only one molecule of the Fe protein binds the MoFe protein. By comparison with the low resolution structure of purified MoFe protein that contains only one molybdenum per tetramer, we deduce that the Fe protein interacts with the FeMo cofactorbinding ␣-subunit of the MoFe protein. This observation demonstrates that the conformation of the ␣-subunit or the ␣␤ subunit pair that lacks the FeMo cofactor is altered and that the change is recognized by the Fe protein. The structure of the 1:1 complex reveals a similar change in the conformation of the Fe protein as has been observed in the low resolution scattering mask and the high resolution crystallographic study of the 1:2 complex where both cofactors are occupied and with the Fe protein bound to both subunits. This extensive conformational change observed for the Fe protein in the complexes is, however, not observed when MgATP or MgADP binds to the isolated Fe protein. Thus, the large scale conformational change of the Fe protein is associated with the complex formation of the two proteins.
Biological nitrogen fixation is catalyzed by nitrogenase, a two-component metalloenzyme system that couples the hydrolysis of MgATP to the reduction of dinitrogen in the reaction, N 2 ϩ 8 H ϩ ϩ 8 e Ϫ ϩ 16 MgATP ¡ 2 NH 3 ϩ H 2 ϩ 16 MgADP ϩ 16 P i REACTION 1 Molybdenum-containing nitrogenases are made up of a molybdenum-containing (MoFe protein or component 1; ϳ230 kDa) and an iron-containing protein (Fe protein or component 2; ϳ60 kDa). 1 During enzyme turnover the Fe protein functions as a specific MgATP-dependent electron donor to the MoFe protein (1)(2)(3)(4). The x-ray crystal structures of both individual proteins isolated from Azotobacter vinelandii (Av proteins) and Clostridium pasteurianum (Cp proteins) have been determined (5)(6)(7)(8)(9). The x-ray structure of MoFe protein from K. pneumoniae (Kp protein) has also been determined (10). The MoFe proteins have an ␣ 2 ␤ 2 subunit structure in which each subunit pair binds a unique Fe 8 S 7 cluster (P cluster) positioned at the subunit interface and the active site of the enzyme, a Fe 7 S 9 molybdenum homocitrate cluster (FeMo cofactor), within the ␣-subunit (11). The Fe protein is a ␥ 2 dimer that has a single [Fe 4 S 4 ] cluster at the subunit interface and two nucleotide binding sites, one on each subunit (6,9). The binding of MgADP or MgATP to the isolated Fe protein results in an altered reactivity and spectroscopic properties of the Fe-S cluster, which have been well documented (see Ref. 4).
The crystal structures of Av2 and Cp2 (6, 9) display a peptide folding pattern similar to other nucleotide-binding proteins, including the ras and G-protein family, and myosin, where transient protein complexes couple nucleotide hydrolysis to signal and energy transduction processes (1). MgATP hydrolysis by nitrogenase requires the presence of both the Fe protein and the MoFe protein, and recently several groups have exploited these similarities to form stable but inactive nitrogenase complexes of A. vinelandii (12,13) and K. pneumoniae (14,15) using AlF 4 Ϫ and MgATP or MgADP. Aluminum fluoride has been extensively used as a tool to examine MgATP binding by gated proteins (16 -18). It has been proposed that the ADP-AlF 4 Ϫ complex can be considered to be an analogue of E-ADP⅐P i species in which AlF 4 Ϫ mimics the trigonal bipyrimidal geometry of the terminal phosphate undergoing nucleophilic attack by a water molecule. More recently, Kp1 and Kp2 have been shown to form ADP-BeF 3 Ϫ stabilized complexes, being putative analogues of the MgATP-bound conformation (19).
Extensive kinetic and modeling work has shown that follow-* This work was supported by the Biotechnology and Biological Research Council as part of the Competitive Strategic Grant to the John Innes Center. Facilities were provided by the Daresbury Synchrotron Radiation Source. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
§ To whom correspondence should be addressed. Ϫ ) 2 ] 2 , was recently determined at ϳ20-Å resolution for K. pneumoniae nitrogenase using x-ray solution scattering (14) and for A. vinelandii nitrogenase at 3-Å resolution by x-ray crystallography (20) and solution scattering (21). For both species, a 13°rotation of subunits of the Fe protein compared with that of the isolated protein was observed (21), indicating that the Fe protein undergoes a substantial conformational change either on complex formation or nucleotide binding. In the complex, the conformational change undergone by the Fe protein brings the Fe 4 S 4 cluster of the Fe protein 4 Å closer to the P cluster of the MoFe protein, to a typical electron transfer distance of ϳ14 Å. It is unclear whether this change in conformation takes place in the complex or is a result of MgATP binding to the Fe protein. In the case of the Fe protein from Av, it has been suggested from a small angle x-ray scattering study that a substantial conformational change has already taken place upon MgATP binding (22). The hypothesis that this "priming" of the Fe protein conformation is required to allow effective docking with the MoFe protein to permit successful electron transfer (see Ref. 23 for discussion) lacks firm experimental evidence and has been questioned recently. An alternative, in which MgATP binds rapidly to a complex of Fe and MoFe proteins, followed by subsequent conformational changes in the Fe protein, has been proposed (24). Moreover, in the heterologous nitrogenase formed between the Fe protein of C. pasteurianum and the MoFe protein from A. vinelandii, MgATP is not required for complex formation (25,26).
We have recently reported that the kinetics of the formation of the K. pneumoniae transition state complex are consistent with MoFe protein lacking one FeMoco 2 center having an altered molecular conformation, which the Fe protein can recognize (15). To test this proposal experimentally, we have obtained solution x-ray scattering data of Kp1 containing one or two FeMoco centers per molecule and generated their molecular shapes at ϳ20-Å resolution for both forms. In addition, the 1:1 complex MoFe protein-(Fe protein⅐ADP⅐AlF 4 Ϫ ) was isolated using Kp1 with only one cofactor site occupied, and the molecular structure was determined from solution x-ray scattering data. The molecular structure of the isolated Fe protein in the absence and presence of nucleotide was also examined to establish the extent of conformational change in Fe protein on its own.

Sample Preparation
Nitrogenase Component Proteins-All manipulation of the air-sensitive nitrogenase components was done under an atmosphere of nitrogen. The nitrogenase proteins Kp1 and Kp2 were purified from K. pneumoniae to homogeneity as described previously (27,28). In the case of Kp1, this method (27) resolved species containing 1.1 and 1.9 molybdenum atoms per molecule and were shown to be free of any contaminating proteins by SDS gel electrophoresis. The specific activities expressed as nmol of hydrogen evolved/min/mg of protein were as follows: Kp1(1.9 Mo), 2100; Kp1(1.1 Mo), 1100; Kp2, 1399.
For data collection, Kp1 was in 25 mM HEPES buffer, pH 7.5, containing 250 mM NaCl and 2 mM Na 2 S 2 O 4 . Kp2 was in the same buffer except at 50 mM, and in some experiments 10 mM MgCl 2 and 5 mM ATP were present. After exposure to the x-ray beam and after the scattering data had been collected, samples of Kp2 were removed from the cell, and the specific activity was remeasured; typically, the recovery was 80% or higher.
Preparation of the 1:1 ADP⅐AlF 4 Ϫ Transition State Complex Analogue-The 1:1 Kp1-(Kp2⅐ADP⅐AlF 4 Ϫ ) 2 complex was prepared as described previously for the 2:1 complex (15), except that the Kp1 used contained 1.1 molybdenum atoms per ␣ 2 ␤ 2 tetramer. The complex was separated from the excess Kp2 and reaction mixture components by anaerobic gel filtration using an Amersham Pharmacia Biotech FPLC system and a 26/60 Superdex™ 200 column equilibrated with 25 mM HEPES buffer, pH 7.5, 2 mM Na 2 S 2 O 4 , 5 mM AlF 3 , and 50 mM KF. Further purification was carried out using an anion exchange Mono Q 5/5 column developed with a linear gradient of NaCl from 0.1 to 1 M in 50 mM Tris buffer, pH 7.5, containing 5 mM AlF 3 and 2 mM Na 2 S 2 O 4 . This procedure resulted in the separation of three species, which subsequent analytical gel filtration showed to be homogeneous species with retention volumes corresponding to Kp1 and the 1:1 and 1:2 complexes. The work described in this paper utilized the purified 1:1 complex.

X-ray Scattering Experiments and Data Analysis
Due to the extreme sensitivity to oxygen, all manipulations of nitrogenase proteins were conducted in an anaerobic chamber (glove box) in an atmosphere of nitrogen containing a very low concentration of oxygen (Ͻ5 ppm of O 2 ). All samples were filtered (0.2-m pore size) and loaded in the glove box into a brass cell (containing a Teflon ring sandwiched by two mica windows that defines the sample volume of 120 l and a thickness of 2.5 mm). The cell was sealed with plastocene and then transferred immediately to the x-ray station. Scattering data were collected on beamline 8.2 at the Synchrotron Radiation Source (Daresbury, UK) (29) at an electron energy of 2 GeV and with beam currents between 150 and 250 mA. At the sample-to-detector distance of 3.3 m (2.5 m) and the x-ray wavelength of ϭ1.54 Å, a momentum transfer interval of 0.002 (0.004) Å Ϫ1 Յ s Յ 0.030 (0.035) Å Ϫ1 was covered on a position-sensitive quadrant multiwire proportional counter (30). Values in parentheses refer to the measurements for the Fe protein only. The modulus of the momentum transfer is defined as s ϭ (2sin⌰)/, where 2⌰ is the scattering angle. The scattering pattern from an oriented specimen of wet rat tail collagen was used to calibrate the detector. Samples were measured at room temperature (ϳ20°C) at concentrations between 0.5 and 5 mg/ml. To minimize systematic errors, each data set consisted of buffer followed by protein data collection. The experimental data were recorded in frames of 100 s allowing on-line checks for changes in the scattering profiles and corrected for background scattering (subtraction of the scattering from the camera and a cell filled with buffer), sample transmission and concentration, and positional nonlinearities of the detector. Off-line data reduction was done with the OTOKO software package (31). Maximum particle dimensions D max , the radius of gyration R g , the distance distribution function p(r), and the extrapolated forward scattering value I(0) were evaluated with the program GNOM (32). The latter allows the estimation of molecular mass when calibrated against the scattering from proteins with known molecular mass (apart from fully loaded Kp1, 225 kDa, as standard for an anaerobic protein sample, nitrous oxide reductase, 134 kDa, was used). The volume V of the particle can be calculated from the Porod invariant (33), including the outer part of the scattering profile. A correction factor is applied to alleviate the difficulties of the limited range of scattering data (described in Ref. 34). More details concerning data collection and reduction are given elsewhere (35).
The computation of the molecular envelopes was based on the ab initio shape determination procedure of Svergun and Stuhrmann (36). If we assume that the scattering is caused by a globular, homogeneous molecule, one can define its molecular shape by the angular envelope function F(,) such that the particle density (r) is unity inside the molecular boundary and vanishes elsewhere. F(,) can be expanded into a series of spherical harmonics Y lm (,) according to Refs. 37 and 38, where f lm represents complex multipole coefficients. The determination of molecular shapes directly from the scattering profile alone in a model-independent manner exploits the information inherent in the wider angle scattering data. The resolution is determined by the maximum order of included harmonics, i.e. the highest value of L. A nonlinear equations system interrelates the f lm coefficients with coefficients of the power series describing the experimental scattering curve. A reliable computational procedure was used to evaluate the multipole coefficients by minimizing the least squares deviation (R 2 ) between experimental and calculated curves according to methods described by Svergun and Stuhrmann (36) and Svergun et al. (39). The shape calculation for the nitrogenase proteins has been performed with smoothed experimental curves (where the low angle part was adjusted to match the corresponding extrapolation to zero concentration). No molecular symmetry was assumed (except for the shape restoration of the isolated Fe protein as well as of the fully loaded MoFe protein; in both cases, a 2-fold symmetry is expected). The available experimental data range for the nitrogenase enzymes justifies the use of harmonics up to L ϭ 4 (using no symmetry; i.e. 19 free parameters) and L ϭ 5 (in the case of symmetry; i.e. 14 free parameters). To set the scene, atomic models have been fitted manually into the molecular shapes. The shape pictures were rendered and manipulated using a Silicon Graphics Indigo work station and the AVS graphics software (Advanced Visual Systems, Inc., Waltham, MA). The ribbon diagrams were produced using the molecular graphics program Insight II (Biosym/MSI). Crystallographic information was utilized to define the nature of the observed scattering features in structural terms. Scattering curves from atomic models of the fully loaded MoFe protein (10), a modeled MoFe protein with missing FeMo cofactor, a 1:1 complex based on the A. vinelandii nitrogenase complex (20), and the Fe protein in its free state (9) were evaluated using the program CRYSOL (40). This method takes the solvent effect into account by surrounding the protein with a hydration shell that has a thickness of 3 Å and uniform density (as fit parameter) different from that of bulk solvent.

RESULTS AND DISCUSSION
Formation of a 1:1 Complex-In a previous study, the rate of formation of the transition state complex Kp1-(Kp2⅐ ADP⅐AlF 4 Ϫ ) 2 was monitored by the loss of nitrogenase activity as the proteins were incubated in the presence of ADP and AlF 4 Ϫ (15). It was proposed that Kp1 that did not have all FeMoco and P cluster binding sites in the protein occupied had an altered conformation compared with fully loaded protein and that Kp2 reacted with these protein species at different rates to form the inhibited complex.
To detect complexes formed by Kp1 preparations that are partially active due to incomplete occupancy of the FeMoco binding sites, a method was developed to monitor complex formation that was independent of activity measurements. The extent of formation of the transition state complexes was determined using gel permeation chromatography of reaction mixtures containing a range of molar ratios of Kp1 and Kp2 from 0.25 to 5 Kp2/Kp1 in the presence of MgADP and AlF 4 Ϫ . Fig. 1 shows the elution profiles from a gel filtration column of the protein species present in incubation mixtures leading to the formation of the transition state complex for a preparation of Kp1 containing only 1.1 molybdenum atoms per ␣␤ dimer. At the lowest Kp2/Kp1 ratio tested, the profile is dominated by free Kp1, which has a retention volume of 10.3 ml (bottom trace in Fig. 1). At this ratio, no peak corresponding to free Kp2 is evident, but a shoulder on the elution profile of Kp1 arising from a higher molecular weight species with a retention volume of 9.8 ml has been formed. As the Kp2/Kp1 ratio was increased to 0.5 Kp2/Kp1, this species became the dominant feature, but at higher ratios it was replaced by a peak with a retention volume of 8.9 ml. A peak corresponding to free Kp2 with a retention volume of 12.35 ml was also detectable under these conditions. The Kp2 band was first evident at a ratio of 1:1 Kp2/Kp1 and continued to grow as the ratio was increased (Fig.  1). These data are consistent with the formation of two types of stable complexes by Kp1 lacking a full complement of cofactor centers, as the Kp2/Kp1 ratio is varied. We propose that initially a 1:1 complex is formed as an intermediate on its way to the 1:2 complex, which predominates at high Kp2/Kp1 ratios. The kinetic data of Yousafzai and Eady (15) are consistent with Kp2 in this complex binding to the ␣ 2 ␤ 2 subunit pair, which contain the metal redox centers. When similar experiments were carried out with Kp1 containing 1.9 molybdenum atoms per ␣ 2 ␤ 2 tetramer, a similar behavior was observed, but the formation of the 1:2 complex occurred at a lower ratio of Kp2/ Kp1 (data not presented), consistent with a difference in the stability of the 2:1 complexes formed by the two species of Kp1.
To isolate the 1:1 complex in sufficient quantity to allow structural studies, it was purified from the components formed in an incubation mixture containing Kp1 lacking a full complement of cofactor centers as described under "Experimental Procedures." This procedure resulted in the separation of three species, which subsequent analytical gel filtration showed to have retention volumes corresponding to Kp1 and the 1:1 and 1:2 complexes.
Solution Structure of Fully Loaded and Half-loaded MoFe Protein- Fig. 2 compares the x-ray scattering patterns (with error bars) for the fully loaded Kp1 (i.e. with the full complement of the cofactor) and Kp1 with only half the cofactor centers present (from now on denoted as Kp11 ⁄2 ). The two scattering profiles are distinct, crossing each other at an intermediate s, suggesting that a significant structural difference exists. As shown in Table I, the geometrical parameters increase in the absence of a full metal cofactor complement, indicating an expansion of the overall conformation. The main differences in the p(r) curves (Fig. 2, inset) occur for longer distances and in the maximum of p(r), which is shifted for Kp11 ⁄2 to larger distances (by approximately 5 Å) to 47 Å. To evaluate possible protein aggregation, scattering patterns have been obtained in the concentration range from 0.5 to 5 mg/ml. The concentrationdependent values for radii of gyration are revealed in Fig. 3a, highlighting the difference between the two protein samples. r itinf;g values (extrapolated to infinite dilution) differ by as much as 1.5 Å, and the R g versus concentration curve shows very different slopes for Kp1 and Kp11 ⁄2 . This may reflect a change in the electrostatic properties of the Kp1 and Kp11 ⁄2 surface. The difference in scattering behavior is further illustrated in Fig. 3b, where the ratio of the scattering curves for Kp1 and Kp11 ⁄2 are plotted (upper trace). It is clear that scattering data for the two protein samples differ over much of the range, and only beyond s Ϸ 0.02 Å Ϫ1 , the ratio hovers around unity. As a control, the lower trace in Fig. 3b shows the intensity ratio of scattering profiles from half-loaded Kp1 recorded at two different concentrations. It is clear that in this case the ratio is unity over almost the whole data range. Moreover, a careful analysis of I(0) did not reveal changes in molecular mass between Kp1 and Kp11 ⁄2 (the mass of the metal cofactor is small compared with that of the protein molecule).
Previously, we have reported the molecular shape of the fully loaded Kp1 (41) and shown this to be in good agreement with the overlaid crystal structure of the MoFe protein; i.e. essentially flexible polypeptide segments appear outside the molecular envelope (see also Fig. 4a). In these calculations, a 2-fold symmetry was assumed for the ␣ 2 ␤ 2 tetramer; thus, shape restoration up to harmonics L ϭ 6 was justified. To assess if there are differences in the molecular structure due to the absence of one of the cofactors, we have undertaken shape restoration without assuming a 2-fold symmetry (Fig. 4b). As a control, a shape restoration of Kp1 containing both cofactors was attempted, where no symmetry is assumed. In this case, spherical harmonics of only up to L ϭ 4 are permissible (19 free parameters). Fig. 4b (left panel, yellow envelope) shows two views of the molecular shape of Kp1 restored with L ϭ 4. A comparison with Fig. 4a shows that the two shapes resemble each other closely, demonstrating that shape restoration with L ϭ 4 is sufficient to recognize the characteristic features of the molecule such as the presence of a 2-fold symmetry. Fig. 4b (right panel, pink envelope) provides two views of the Kp11 ⁄2 shape at L ϭ 4, where again no symmetry was assumed (fits to the experimental data with final residual R ϭ 2.1 and 2.4% for Kp1 and Kp11 ⁄2 , respectively, are shown in Fig. 2). The shape for Kp11 ⁄2 differs significantly from Kp1; an extension or bulge appears that breaks the familiar view of 2-fold symmetry. An assessment of the differences of both molecular envelopes (Kp1 versus Kp11 ⁄2 at L ϭ 4) demonstrates that the left half of the molecule remains essentially the same in the two cases (see Fig. 4b), but significant expansion is observed on the right half of the molecule in the absence of the cofactor. These shapes indicate that the missing FeMo cofactor in Kp1 results in a significantly less compact structure, which is underlined by the geometrical parameters given in Table I. This structural expansion may also explain the slightly larger volume of Kp11 ⁄2 as a result of water filling the created cavities and clefts. Although the molecular shape for both Kp1 and Kp11 ⁄2 (represented by an average envelope deduced from several shape reconstruction runs using different starting conditions) offers a qualitative insight into the structural change as a result of FeMoco ab-

FIG. 2. Solution scattering profiles and distance distribution functions p(r) (inset) of half-loaded (E) and fully loaded (q) MoFe protein.
Smooth curves represent the scattering profiles from the restored shapes (thin lines) in Fig. 4b as well as from simulations (thick lines) using models of Kp1 and Kp11 ⁄2 based on crystal structure coordinates of Kp1 (10). See second section of "Results and Discussion" for further details. Apart from a few contacts between the N-terminal residues of the ␣-subunit and domain IЈ of the ␤-subunit (see below), the hinge movement was considered to be a rigid body rotation of the ␣-subunit alone with minor effects on the ␤-subunit (Fig.   4c), the latter playing a major role in tetramerization. However, since some of the helices in domain III of the ␣-subunit help to stabilize the tetramer interface (5), an influence on the arrangement of the two ␣␤ dimer pairs cannot be excluded. Furthermore, due to the likely nature of a flexible hinge, the possibility of multiple conformers cannot be ruled out. The result from the scattering pattern simulation for the model of Kp11 ⁄2 (together with the result for Kp1) is given in Fig. 2. The simulated profile based on the Kp1 crystal structure (10) agrees very well with the scattering curve of fully loaded Kp1 in solution. The profile from the theoretical model of Kp11 ⁄2 effectively reproduces the characteristic features of the experimental curve (this is also reflected in the goodness of fit ( 2 value); see Table I). Deviations for 0.015 Å Ϫ1 Յ s Յ 0.022 Å Ϫ1 may be rationalized, given that this only represents one (i.e. a model that has been obtained by rigid body movement of domain III of one of the ␣-subunits only) of several possible conformations. Besides, parts of the surrounding environment and even the tetrameric conformation are probably affected. Ribbon drawings representing the structure of Kp1 and the modeled structure of Kp11 ⁄2 have been superimposed on the shapes displayed in Fig. 4b. Solution Structure of 1:1 Complex and Changes in the Fe Protein-The x-ray solution scattering curve and pair distribution function for the isolated 1:1 complex, purified as described under "Experimental Procedures," is compared with the 1:2 complex data (21) in Fig. 5a. The scattering results for both complexes are considerably different (see also Table I). Analysis of the I(0) intensities of 1:2 and 1:1 complex revealed a ratio of 1.3, confirming the proposed stoichiometry for both complexes, since an intensity ratio of 1.21 would be expected based on the difference in molecular masses of the two complexes. The distance distribution function (Fig. 5a, inset) of 1:1 complex computed for infinite dilution shows a decrease of approximately 30 Å in long distances compared with the 1:2 complex being consistent with only one Kp2 bound to Kp1. For shape calculations, again no symmetry was assumed, and an envelope with harmonics up to L ϭ 4 could be restored. The fits to the experimental data are superimposed in Fig. 5a and yielded R factors of 2.2% (1:1 complex) and 1.8% (1:2 complex). Two views of the molecular shape thus obtained for the 1:1 complex are shown in Fig. 5b. This is superimposed with the model built for Kp11 ⁄2 (see above) and one Fe protein. Although this comparison clearly demonstrates that only one Fe protein can be included in the restored shape for the 1:1 complex, a scattering pattern simulation (distinguishing between the two known structural states of the Fe protein) confirms that the Fe protein undergoes a very similar conformational change (see simulated curves in Fig. 5a and 2 values given in Table I) as that documented for the 1:2 complex (20,21). Interestingly, in all simulations the goodness of fit improves when the Fe protein from the Av1-Av2 complex (20) is considered (Table I). This is also emphasized visually by a better agreement with the restored molecular envelope (see Fig. 5b) overlaying the compact conformation for the Fe protein when in complex with the MoFe protein (20) rather than the less tight conformation of the free Fe protein as suggested in the original docking model (5). Deviations between experimental and simulated scattering results for the 1:1 complex (in particular concerning the scattering range 0.018 Å Ϫ1 Յ s Յ 0.023 Å Ϫ1 (see Fig. 5a) as well as the R g value (Table I) 4. a, two orientations of the shape of fully loaded Kp1 restored with symmetry (L ϭ 5). Superimposed is a ribbon model of the Kp1 crystal structure (10) highlighting ␣and ␤-subunits in red and blue, respectively. Metal cofactors are depicted as ball and stick models (P clusters in black and FeMo cofactors in magenta). b, The left panel (yellow envelope) represents the shape of fully loaded Kp1 restored without symmetry (L ϭ 4); the panel on the right gives the molecular envelope for the half-loaded Kp1 (in pink, restored with harmonics up to L ϭ 4). The extension of the shape of Kp11 ⁄2 with regard to Kp1 (i.e. on the right of the pink envelope) is also highlighted by superimposed ribbon drawings of the Kp1 crystal structure (onto the yellow envelope) and a modelled structure of Kp11 ⁄2 (onto the pink envelope). In the latter model, domain III of the right-hand ␣-subunit (green colored ribbon segments) has been moved by a 22°hinge rotation. As indicated, the corresponding views on the top and bottom are related by a 90°rotation around the horizontal axis. c, ribbon model of an ␣␤ subunit pair from the structure of Kp1 with FeMo cofactor buried in the ␣-subunit (left) and from the modeled structure of Kp11 ⁄2 in which the missing FeMo cofactor causes an opening of the ␣-subunit (right). The same color code as in b is used. The view shows the approximate pseudo-2-fold symmetry. In addition, the two helices involved in docking the Fe protein are indicated.
view of the recent evidence for long range conformational changes in the MoFe protein upon Fe protein binding (26).
The absence of the second Fe protein from the opposite side of the molecule that lacks the cofactor provides support to the idea that the Fe protein is able to recognize the altered conformation due to the missing cofactor from the MoFe protein (Fig.  4b). At this point, it has to be mentioned that the "open" conformation of Kp11 ⁄2 (here modeled simply as rigid body ro-tation of domain III in one of the ␣-subunits) does not directly affect the surface area implicated in Fe protein docking. However, as a result of certain contacts between the N-terminal residues (␣1-␣56) in the ␣-subunit (forming part of domain III) and residues ␤111-␤140 in domain IЈ of the ␤-subunit, this particular region of the ␣␤-interface is likely to be modified. Most importantly, the latter polypeptide segment of the ␤-subunit contains one of the two helices for Fe protein docking (see Ref. 20; see also Fig. 4c). It is therefore most plausible to assume that destabilization or reorientation of this docking helix in the ␤-subunit (as a consequence of a hinge movement in the ␣-subunit due to the missing FeMo cofactor) leads to rearrangements of a considerable section in the interaction surface with the Fe protein. It is assumed that the other docking helix (located in domain I) in the ␣-subunit is unaffected due to stabilizing effects of the bound P cluster. This is an appealing structural scenario, considering that the gel permeation data presented above show the binding of a second Fe protein to Kp11 ⁄2 , albeit with lower affinity. Interestingly, besides binding the Fe protein, additional functional roles of the ␤ docking helix have been inferred from the structure of Kp1 (10). Our examination may even suggest that this helix is able to sense the absence of the FeMo cofactor.
Conformational Change of the Isolated Fe Protein upon Nucleotide Binding-The conformational changes observed for both Av2 and Kp2 in the transition state complexes are of significant functional importance, since they enable the Fe 4 S 4 cluster of Fe protein and MoFe protein to approach significantly closer together to typical electron transfer distances. This is likely to result in an efficient electron transfer between the Av2/Kp2 Fe 4 S 4 cluster to the Av1/Kp1 P cluster, which in the complex is same distance away from the FeMo cofactor, the site of nitrogen reduction.
There is a body of experimental data indicating that the binding of nucleotides to the Fe protein results in changes in the spectroscopic properties and reactivity of the Fe 4 S 4 center (see Ref. 4) and in the sensing of the redox level of the cluster by bound nucleotide (42). Both MgATP and MgADP, competitive inhibitors of electron transfer, are expected to bind at the same site, which is located some ϳ20 Å away from the Fe 4 S 4 cluster, thus their effect on the properties of the cluster has been rationalized to result from a conformational change in the Fe protein. It is of interest to see if this conformational change in the isolated protein is similar to that observed for Kp2 or Av2 in the transition state complex analogues. A preliminary x-ray scattering study has been reported on the effects of nucleotide binding on Av2, where the radius of gyration (R g ) deduced from the Guinier region alone has been determined (22). Recent advances in the x-ray scattering technique, particularly the ability to use a wider data range, have proved very powerful in studying conformational changes in proteins (14,35,43). Fig. 6a shows the calculated scattering profiles for Av2 obtained using the crystallographic structures of Av2 on its own (6,9) and that observed in the Av1-(ADP⅐AlF 4 Ϫ ⅐Av2) 2 complex (20). The compact nature of Av2 in the complex is reflected in reduction of both R g (⌬R g ϭ 1.9 Å) and D max (⌬D max ϭ 6 Å). The changes in the scattering profile over the extended scattering range are similar in nature to those observed in the halfmolecule of transferrin upon binding of iron (35) and thus should be easily accessible by x-ray scattering. Fig. 6b illustrates the experimental x-ray scattering data for Kp2 on its own, with MgADP and with MgATP. The profiles are practically indistinguishable. The lack of change compared with the one seen in Fig. 6a is apparent. This experiment has been repeated for the Fe protein of the vanadium nitrogenase system (see Ref. 44), and identical results were obtained (data not FIG. 5. a, scattering profile and p(r) function (inset) of the 1:1 complex (E) compared with the 1:2 complex (q). Solid curves obtained from the restored envelopes of 1:2 complex (20) and 1:1 complex are overlaid as thin lines, whereas curves based on scattering pattern simulations of the complexes are indicated by thick lines. In both cases, the simulation where the Fe protein is bound as compact molecule (20) rather than in the "relaxed" conformation according to the original docking model (5) results in a better goodness of fit value (see Table I). For the purpose of clarity, I(s) profiles have been shifted, and p(r) curves have been scaled according to molecular weight. b, two views of the molecular envelope of the 1:1 complex (shaded in green). Superimposed is an adapted ribbon model for this complex based on the Fe protein (in yellow) taken from the crystal structure of the Av1-Av2 complex (20) and docked onto the modeled Kp11 ⁄2 structure (color coding as in Fig. 4, a and b). The displayed views are equivalent in scale and orientation to those shown in Fig. 4. shown). In addition, data collection for the set of Kp2 (Kp2 plus MgATP and Kp2 plus MgADP) has been made on three different occasions, and very similar results have been obtained (data not shown). It is clear from these data that free Kp2 does not undergo a significant structural change upon nucleotide binding. Thus, the extensive changes observed in Kp2 and Av2 in the transition state complexes must primarily arise from the interaction with the MoFe protein; the role of the nucleotide may thus be to prime the interaction region of the Fe protein to respond to the MoFe protein. Consistent with this, Duyvis et al. (24) recently proposed from pre-steady state kinetic data that MgATP interacts with the A. vinelandii nitrogenase complex to trigger the conformational changes that are essential for effective electron transfer. We note that the surface incompatibility of Av1 and Av2 necessitates the large scale conformational changes in Av2 to accommodate specific interactions between the two proteins (20).
The excellent agreement between structures in solution and crystalline state is illustrated when the scattering data for Kp2 are compared (Fig. 6c, solid line) with the scattering profile calculated from the crystallographic structure of free Av2 (9) surrounded by a water layer (40). The R g from the crystal structure increases from 23.9 to 25.5 Å when hydration effects are taken into account and agrees neatly with the experimental R g value. The inclusion of a hydration shell improves the fit to the experimental x-ray scattering pattern considerably (45,46). In fact, the value for the hydrated structure of free Av2 is equivalent to what was observed in an earlier study by Chen et al. (22) for Av2 with MgATP but is substantially different from their values for native Av2 and Av2 plus MgADP. Their R g values (Ͼ27 Å) for the latter two states would suggest an Fe protein structure even less compact compared with the structures reported for the free Fe protein (6,9). CONCLUSION Our earlier kinetic work (15) showed that preparations of Kp1 containing nonintegral amounts of molybdenum did not contain apoprotein but were a mixture of species with different reactivity containing one or two molybdenum atoms. The gel filtration data presented here show that, depending on the molybdenum content, these species react with Fe protein to form both 1:1 and 2:1 complexes with different stabilities. These differences are consistent with our demonstration that the missing FeMoco in an ␣␤ subunit pair of Kp1 affects its molecular shape; the ␣␤ subunit pair with FeMoco bound has a more compact structure compared with that of the empty ␣␤ subunit pair (Kp11 ⁄2 ). The open structure of Kp11 ⁄2 provides a potential route for insertion of FeMoco from its precursor carrying proteins in MoFe protein maturation (4). The difference in conformation influences the interaction with the Fe protein significantly. Kp11 ⁄2 forms an inactive 1:1 ADP⅐AlF 4 Ϫ transition state complex analogue as an intermediate species before binding a second Fe protein with a lower affinity, compared with Kp1. The scattering profile of the Fe protein shows that in the 1:1 complex it undergoes a substantial conformational change, similar to that observed for the 1:2 complex, bringing the Fe4S 4 cluster of Fe protein 4 Å closer to the P cluster of the MoFe protein. However, this large scale structural change in the Fe protein does not take place in the isolated Fe protein upon nucleotide binding, thus providing strong evidence that the changes in conformation of the Fe protein are coupled with the complex formation.