Nitrogenase MoFe-Protein at 1.16 Å Resolution: A Central Ligand in the FeMo-Cofactor

See allHide authors and affiliations

Science  06 Sep 2002:
Vol. 297, Issue 5587, pp. 1696-1700
DOI: 10.1126/science.1073877


A high-resolution crystallographic analysis of the nitrogenase MoFe-protein reveals a previously unrecognized ligand coordinated to six iron atoms in the center of the catalytically essential FeMo-cofactor. The electron density for this ligand is masked in structures with resolutions lower than 1.55 angstroms, owing to Fourier series termination ripples from the surrounding iron and sulfur atoms in the cofactor. The central atom completes an approximate tetrahedral coordination for the six iron atoms, instead of the trigonal coordination proposed on the basis of lower resolution structures. The crystallographic refinement at 1.16 angstrom resolution is consistent with this newly detected component being a light element, most plausibly nitrogen. The presence of a nitrogen atom in the cofactor would have important implications for the mechanism of dinitrogen reduction by nitrogenase.

Biological nitrogen fixation provides the dominant route for the transformation of atmospheric dinitrogen into a bioavailable form, ammonia (1–4). This process is catalyzed by the enzyme nitrogenase, which consists of two component metalloproteins, the Fe-protein and the MoFe-protein. The homodimeric Fe-protein couples adenosine 5′-triphosphate hydrolysis to interprotein electron transfer and is the only known mechanistically competent source of electrons for the catalytically active component, the MoFe-protein. The latter is organized as an α2β2 tetramer that contains two copies each of two unique polynuclear metal clusters designated the P-cluster and the FeMo-cofactor. Whereas the P-cluster likely participates in interprotein electron transfer, the FeMo-cofactor is the active site of substrate binding and reduction. Despite detailed structural information and a multitude of kinetic, spectroscopic, and theoretical studies, little is known about the mechanistic details of dinitrogen reduction by nitrogenase (5), particularly the site and mode of substrate binding.

The structures of the P-cluster and FeMo-cofactor in the MoFe-protein have been determined crystallographically at resolutions between 2.8 and 1.6 Å (6–12). The FeMo-cofactor, with composition [Mo:7Fe:9S]:homocitrate, is coordinated to the protein through the side chains of only two residues bound to Fe and Mo sites located at opposite ends of the cluster. Perhaps the most unusual feature of the cofactor in these structures is the trigonal prismatic arrangement of the six central iron atoms. These iron atoms lie on the surface of a sphere with a radius of 2.0 Å from the cofactor center and are each coordinated to three inorganic sulfur atoms. Furthermore, all nine sulfur atoms of the FeMo-cofactor are themselves equidistant from the center on a second sphere with a radius of 3.3 Å. The structure of the P-cluster, with composition [8Fe:7S], can be considered composed of two [4Fe:3S] subclusters that are bridged by a hexacoordinate S, with the overall assembly coordinated to the protein through six cysteine ligands.

Analysis of crystallographic structures of the MoFe-protein at resolutions up to 1.7 Å (13) indicated a significant (>6σ), positive F oF c difference density peak in the central cavity of the FeMo-cofactor. However, the corresponding 2F oF c electron density maps did not show this feature. This contrasting behavior for the two maps suggested that the scattering properties of the whole cofactor might perturb the calculated electron density in its center through the influence of series termination effects. These are a well-known phenomenon in Fourier analyses, and in crystallography lead to resolution-dependent ripples around atomic positions (14–16); the effect is particularly pronounced around regions of high electron density such as metal sites. To illustrate the effect, the electron-density distribution, ρ(r), adjacent to an iron atom can be calculated as a function of the high-resolution limit from the expression (16).Embedded Image(1)where f Fe is the atomic form factor for iron; s = 1/d, whered is the resolution; and d max is the high-resolution limit for integration. If the Fourier transform is truncated by choosing a finite integration limit instead of 1/d max = ∞, the calculated ρ(r) will show resolution-dependent series termination errors (Fig. 1). At a distance of r = 2.0 Å from an iron atom, reminiscent of the situation in the central cavity of the FeMo-cofactor, an artificial minimum with negative electron density is created for d max(resolution) between 1.6 and 2.5 Å.

Figure 1

The effect of series termination errors on the resolution-dependent electron density profile around an iron atom. A plot of electron density ρ(r) versus distancer (Eq. 1) shows varying effects for high-resolution limits d max of 1.0 Å (black), 1.3 Å (brown), 2.0 Å (red), and 2.5 Å (orange). When ρ(r) is plotted versus d max for the distance r = 2.0 Å (as found in the FeMo-cofactor), a characteristic profile is obtained with resolution-dependent maxima and minima (inset).

To model the diffraction behavior inside the FeMo-cofactor, the influence of the entire [Mo:7Fe:9S] unit must be considered. When Eq. 1 is used to calculate the scattering contributions from the various individual components, it is apparent that the density in the central cavity is influenced mainly by the six iron atoms at 2.0 Å and the nine sulfur atoms at 3.3 Å (Fig. 2). At lower resolutions, the negative ripples surrounding these multiple iron and sulfur atoms combine to produce sufficiently negative electron density in the cofactor center to completely obscure the electron density of a light atom at this site. Consequently, this implies that the “hole” at the center of the cofactor in the 2F oF c electron density map is an artifact, rather than the peak in the F oF c difference density map (16).

Figure 2

Contributions of individual atom types to the resolution-dependent electron density profile in the central cavity of the FeMo-cofactor. Six iron atoms and all nine of the cluster's sulfur atoms are located on two concentric spheres. Having identical distances from the center (3.3 Å for sulfur, 2.0 Å for iron), they are the main contributors to the electron density profile there. The apical iron and molybdenum atoms exert only a minor influence. Plots of ρ(r) versus d max, calculated analogously to Fig. 1 (inset), illustrate this effect. The curves for six iron atoms at 2.0 Å (blue), nine sulfur atoms at 3.3 Å (dark yellow), and one apical iron (Fe1, gray) and the molybdenum (orange) at 3.5 Å are shown. The sum of of all these contributions is shown in black.

As indicated in Fig. 2, series termination effects become less pronounced with increasing resolution. Hence we initiated the structure determination of the Azotobacter vinelandii MoFe-protein at a sufficiently high resolution to overcome their influence (17). To achieve this, we improved the quality of dithionite-reduced MoFe-protein crystals to permit data collection and refinement of the structure (Table 1) at 1.16 Å resolution toR = 0.123, with a diffraction component precision index [DPI (18)], a measure of the coordinate error, of 0.027 Å. These crystals contain two α2β2tetramers per asymmetric unit, and consequently four crystallographically independent copies of the FeMo-cofactor are present. Each copy clearly shows electron density at the center of the FeMo-cofactor in maps calculated with both F o F c and 2F oF c Fourier coefficients (Fig. 3).

Figure 3

Stereo representation of the FeMo-cofactor with the central ligand modeled as a nitrogen atom. The electron density map shown is a weighted 2F oF c map of the 1.16 Å resolution structure of dithionite-reduced A. vinelandii MoFe-protein contoured at 3σ.

Table 1

Data collection and refinement statistics. For the calculation of R free, 1% of the observed reflections were removed at random before refinement. a.u., asymmetric unit; ref., refinement; FOM, figure of merit; rmsd, root mean square deviation.

View this table:

To characterize the atomic identity of the ligand at the center of the cofactor, we calculated a resolution-dependent electron density profile, analogous to that of Fig. 2, from the experimental diffraction data, which confirms the substantial impact of series termination errors (Fig. 4). The absolute values of electron density differ from those of the theoretical model (Fig. 2), owing to the absence of some low-resolution reflections, including theF(000) term, in the experimental data. Consequently, the experimental electron-density minimum is even more pronounced than in the theoretical model. The electron density at the cofactor center, as determined from the observed structure factor amplitudes and phases calculated from a model lacking a central atom (black dots), is negative for all resolutions <1.55 Å. If, however, a purely theoretical curve is generated with both calculated structure factor amplitudes and phases from this same model (black line), the density at the cofactor center is substantially lower. Clearly, a central atom needs to be added to the model to reproduce the experimentally observed data. If the same model calculations are performed with an atom in the center of the cavity, the results are qualitatively similar and demonstrate that the destructive interference of the surrounding atoms is sufficient to obliterate the densities of a carbon, a nitrogen, or an oxygen atom at this position, although not that of a sulfur atom.

Figure 4

Resolution-dependent electron density profiles as derived from the complete 1.16 Å resolution structure of dithionite-reduced A. vinelandii MoFe-protein. Solid lines are pure F calc electron density values, whereas dotted curves are read from actualF obs·ϕcalc maps computed with the experimental structure factor amplitudes. The calculated density without a central ligand in the cofactor (black line) is well below the one observed in the experimental map (black dots). The latter further shows that positive density at the central position is obtained only at resolutions beyond 1.55 Å. A sulfur atom in the central cavity, however, gives a profile (dark yellow line) that at no resolution leads to negative electron density. Thus, a sulfur atom in the structure would not have been missed. If carbon (green line), nitrogen (blue line), or oxygen (red line) is placed in the center, their density still disappears, but experimental densities from structures refined with any of these atoms result in nearly identical curves (green, blue, and red dots), which agree most closely to the theoretical curve with nitrogen.

Unambiguous identification of an atom type solely from its electron density is problematic, even at atomic resolution. We considered carbon, nitrogen, oxygen, and sulfur as chemically plausible candidates for the central atom and tested each with the available diffraction data. Because the surrounding atoms of the FeMo-cofactor are well defined and show very low temperature factor anisotropy, we assumed that the central ligand is fully occupied; partial occupancy should create at least a slight positional displacement in its environment. Although the possibility of a central sulfur atom was considered in initial models of the cofactor, of these four elements, we consider sulfur as the least likely candidate for the central atom because the observed density would allow only partial occupancy, the distances to the surrounding iron atoms are too short, and the destructive interference of the surrounding atoms is not sufficient to entirely cancel the density of the sulfur in the lower resolution structures, contrary to what is observed (Fig. 4). Refinement of a carbon, nitrogen, or oxygen atom at the central position in all cases yielded temperature factors for this site (C: 12.6 ± 2.2 Å2; N: 14.0 ± 2.0 Å2; O: 14.8 ± 2.5 Å2) that correspond well to the values observed for the surrounding irons (12.2 ± 1.1 Å2). Although the density for the ligand tends to be more anisotropic than for the inorganic components of the cofactor, it is not compatible with an ordered diatomic or larger species. Consequently, although the identity of the central ligand in the FeMo-cofactor cannot be unambiguously established from the crystallographic analysis, from the properties of the resolution-dependent electron-density profile (Fig. 4), and from the interaction of nitrogenase with dinitrogen and ammonia, we have tentatively assigned this central ligand as a fully occupied N.

Interatomic distances in the FeMo-cofactor between metals and the central nitrogen are summarized in Fig. 5. The central nitrogen ligand is hexacoordinate, with average iron-nitrogen distances of 2.00 ± 0.05 Å. This overall arrangement resembles that of a previously characterized cobalt carbonyl cluster containing interstitial nitrogen surrounded by a trigonal prismatic arrangement of metal (19), with an average Co-N distance of 1.94 Å. With average N–Fe–S bond angles of 102° ± 2°, the central ligand completes an approximately tetrahedral coordination environment for the six irons surrounding this group; consequently, it is no longer true that these iron sites are “three-coordinate” (6), at least in the dithionite-reduced form of the MoFe-protein.

Figure 5

Stereo representation of interatomic distances in the FeMo-cofactor. Mean values for the individual distances are the averages of the four crystallographically independent copies of the cofactor in the structure of the dithionite-reduced A. vinelandii MoFe-protein. Standard deviations are 0.01 Å for metal-metal and 0.03 Å for metal-nitrogen distances.

The presence of a ligand in the center of the FeMo-cofactor, particularly a nitrogen, has important implications for understanding the properties of nitrogenase. One potentially relevant observation is the evidence from electron spin echo envelope modulation (ESEEM) studies for one or more nitrogen nuclei interacting with the FeMo-cofactor (20–22). Whereas the ESEEM signals have been assigned to nitrogen atoms of surrounding protein residues, the presence of a nitrogen atom in the cofactor suggests an alternate, nonprotein, source for at least some of the signal. Although a central nitrogen atom could be a structural component of the cofactor, it is difficult to conceive of a process whereby it is inserted without some relation to dinitrogen reduction. Indeed, a monoatomic nitrogen is consistent with the Thorneley and Lowe (23) kinetic model that requires the resting state of the MoFe-protein to be reduced by three electrons before dinitrogen can bind. This may reflect the need to replenish the electrons used in reducing the nitrogen to the level of nitride (N3−) before it can be liberated as ammonia.

Theoretical studies of substrate binding to the FeMo-cofactor have indicated that the center of the trigonal prismatic arrangement of irons provides favorable interaction sites for dinitrogen and its reduction products (24–26). Furthermore, the spacing of iron atoms around this central site in the FeMo-cofactor closely parallels that of the iron surfaces used as catalysts for dinitrogen reduction in the industrial Haber-Bosch process (27). However, the distances between irons in the cofactor are longer (2.63 Å) than in regular metallic iron (2.47 Å), and such strained metal surfaces have been modeled to be particularly reactive as catalysts for dinitrogen dissociation (28). Notwithstanding the enormous disparity of reaction conditions, the parallels between the arrangement of metals in the nitrogenase FeMo-cofactor and the catalyst for the Haber-Bosch process suggest the possibility of common mechanistic elements in the reduction of dinitrogen to ammonia.

Supporting Online Material

Materials and Methods


  • * To whom correspondence should be addressed. E-mail: dcrees{at}


View Abstract

Navigate This Article