Crystal clear


Editor's Introduction

Ligand binding to the FeMo-cofactor: Structures of CO-bound and reactivated nitrogenase

annotated by

The reduction of dinitrogen (N2) by the enzyme nitrogenase is one of the most widely applicable transformations in the world, converting the stubbornly inert gas that makes up 78% of the atmosphere into the reactive atomic building blocks of proteins, vitamins, explosives, and fertilizers. Despite its importance, however,  the reaction itself (known as nitrogen fixation) is still very poorly understood. This article attempts to elucidate some of the nitrogenase fixation mechanism by replacing the substrate N2 with the comparable inhibitor CO, temporarily halting nitrogenase's activity in order to obtain an atomic structure of the enzyme mid-reaction. X-ray crystallography is used to produce the first example of an inhibitor-bound crystal structure of nitrogenase (during and after CO inhibition) and the results are analyzed for mechanistic implication such as the relocation of a sulfur atom during the reaction. Ultimately, this groundbreaking work provides a glimpse into the mechanism of nitrogenase's enigmatic reduction of N2 and paves the way for future studies of other substrates and inhibitors which could eventually allow scientists to mimic the enzyme's valuable and unique activity.

Paper Details

Original title
Ligand binding to the FeMo-cofactor: Structures of CO-bound and reactivated nitrogenase
Original publication date
Vol. 345, Issue 6204, pp. 1620-1623
Issue name


The mechanism of nitrogenase remains enigmatic, with a major unresolved issue concerning how inhibitors and substrates bind to the active site. We report a crystal structure of carbon monoxide (CO)–inhibited nitrogenase molybdenum-iron (MoFe)–protein at 1.50 angstrom resolution, which reveals a CO molecule bridging Fe2 and Fe6 of the FeMo-cofactor. The μ2 binding geometry is achieved by replacing a belt-sulfur atom (S2B) and highlights the generation of a reactive iron species uncovered by the displacement of sulfur. The CO inhibition is fully reversible as established by regain of enzyme activity and reappearance of S2B in the 1.43 angstrom resolution structure of the reactivated enzyme. The substantial and reversible reorganization of the FeMo-cofactor accompanying CO binding was unanticipated and provides insights into a catalytically competent state of nitrogenase.


Biological nitrogen fixation is nature’s pathway to convert atmospheric dinitrogen (N2) into a bioavailable form, ammonia (NH3). Nitrogenase, the only known enzyme capable of performing this multielectron reduction, consists of two component metalloproteins, the iron (Fe)–protein (Av2) and the molybdenum-iron (MoFe)–protein (Av1) (13). The Fe-protein, containing a [4Fe:4S]-cluster, mediates the adenosine triphosphate (ATP)–dependent electron transfer to the MoFe-protein to support dinitrogen reduction (4). The MoFe-protein is an α2β2 heterotetramer with one catalytic unit per αβ heterodimer (5). To achieve the elaborate redox properties required for reducing the N–N triple bond, two metal centers are present in the MoFe-protein: the P-cluster and the FeMo-cofactor. The P-cluster, an [8Fe:7S] entity, is the initial acceptor for electrons, donated from the Fe-protein during complex formation between the two proteins (68). Electrons are subsequently transferred to the FeMo-cofactor, a [7Fe:9S:C:Mo]-R-homocitrate cluster that constitutes the active site for substrate reduction and is the most complex metal center known in biological systems (5912).

Substrates and inhibitors bind only to forms of the MoFe-protein reduced by two to four electrons relative to the resting, “as-isolated” state, which can only be generated in the presence of reduced Fe-protein and ATP (1). Mechanistic studies must take into account the dynamic nature of the nitrogenase system, which requires multiple association and dissociation events between the two component proteins, as well as the ubiquitous presence of protons that are reduced to dihydrogen even in competition with other substrates (1,1315). The resulting distribution of intermediates under turnover conditions complicates the structural and spectroscopic investigation of substrate interactions. Hence, even the fundamental question of whether molybdenum or iron represents the site for substrate binding at the FeMo-cofactor is still under debate, and as a consequence, a variety of mechanistic pathways have been proposed based on either molybdenum or iron as the catalytic center, mainly following Chatt-type chemistry (16).

Inhibitors are potentially powerful tools for the preparation of stably trapped transient states that could provide insight into the multielectron reduction mechanism. In this regard, carbon monoxide (CO), a noncompetitive inhibitor for all substrates except protons (17,18), has a number of attractive properties; CO is isoelectronic to the physiological substrate, is a reversible inhibitor, and only binds to partially reduced MoFe-protein generated under turnover conditions. Although noncompetitive inhibitors are traditionally considered to bind at distinct sites from the substrate, for complex enzymes—such as nitrogenase—with multiple oxidation states and potential substrate-binding modes, this distinction is not required (19). More recently, it has also been shown that CO is a substrate, albeit a very poor one, whose reduction includes concomitant C–C bond formation to generate C2 and longer-chain hydrocarbons, in a reaction reminiscent of the Fischer-Tropsch synthesis (20, 21). Therefore, CO binding as inhibitor or substrate must involve important active-site properties common to the reduction of the natural substrate dinitrogen. For this reason, CO binding has been investigated by a variety of spectroscopic methods, most notably electron paramagnetic resonance and infrared spectroscopy, and depending on the partial pressure, multiple CO-bound species have been observed; yet, a structurally explicit description of any CO-binding site has been elusive (182227).

Building on these observations, we have determined a high-resolution crystal structure of a CO-bound state of the MoFe-protein from Azotobacter vinelandii. This necessitated overcoming several obstacles. First, the experimental setup for all protein-handling steps, including crystallization, was deemed to require the continuous presence of CO. Second, because inhibition requires enzyme turnover, a prerequisite was the ability to obtain crystals of the MoFe-protein from activity assay mixtures, rather than from isolated protein (see supplementary material for assay details), conditions that are typically contradictory to crystallization requirements. Finally, rapid MoFe-protein crystallization (≤5 hours) was crucial and was achieved on the basis of previously developed protocols in combination with seeding strategies and in the presence of CO (10).

Crystals of the inhibited MoFe-protein (Av1-CO) yielded structural data at 1.50 Å resolution, which allowed clear identification of a CO ligand (Fig. 1, A and C, and fig. S1, A to D). The Av1-CO structure directly demonstrates the binding of one molecule of CO per active site in a μ2-bridging mode between Fe2 and Fe6 that forms one edge of the trigonal six-iron prism (Fe2-3-4-5-6-7) of the FeMo-cofactor (Fig. 1, A, C, and D). CO binding is accompanied by a displacement of one of the belt sulfur atoms (S2B), although it retains the essentially tetrahedral coordination spheres for Fe2 and Fe6. As a result, the two Fe are coordinated by two sulfur and two carbon atoms—a geometry, to our knowledge, not previously observed in metalloclusters [although higher–coordination number geometries have been observed in FeFe-hydrogenases (28)]. Confirmation of the S2B displacement was provided by anomalous difference Fourier maps calculated with diffraction data measured at 7100 eV (Fig. 1B); this energy is just below the Fe K-edge so that the anomalous scattering from S is substantially enhanced relative to Fe. The carbon atom of CO is located at a distance of 1.86 Å from each of the irons (Fe2 and Fe6), compared with a previous distance of 2.2 Å for S2B (Fig. 1C). The altered ligand environment results in a small adjustment of the FeMo-cofactor geometry, with the Fe2-Fe6 distance (2.5 Å) slightly shortened relative to the unchanged Fe4-Fe5 and Fe3-Fe7 distances (2.6 Å) (Fig. 1C).

Fig. 1. CO-inhibited MoFe-protein (Av1-CO).Refined structure of the CO-bound FeMo-cofactor at a resolution of 1.50 Å. (A) View along the Fe1-C-Mo direction. The electron density (2Fobs – Fcalc) map is contoured at 4.0 σ and represented as blue mesh. The density at the former S2B site is substantially decreased and in excellent agreement with bound CO [see also (C)]. (B) Same orientation as (A) superimposed with the anomalous density map calculated at 7100 eV (green) at a resolution of 2.1 Å contoured at 4.0 σ, showing the absence of anomalous electron density at the CO site. (C) Side view of FeMo-cofactor highlighting the μ2 binding geometry of CO. The electron density (2Fobs Fcalc) map (blue mesh) surrounding CO-Fe2-Fe6-C is contoured at 1.5 σ. (D) Same orientation as (C) highlighting the ligand environment of the metal center. The catalytically important side-chain residues α-Val70 and α-His195 are in close proximity to the CO-binding site. Iron atoms are shown in orange, sulfur in yellow, molybdenum in turquoise, carbon in gray, nitrogen in blue, and oxygen in red.

What evidence indicates that the CO inhibitor binds to the FeMo-cofactor?

Where and how does the CO ligand bind?

How does this affect the atomic structure of nitrogenase as compared with the unbound cofactor?

What other structural characteristics of the protein might have an affect on the binding of CO?


The structure of the CO-bound cofactor was obtained through crystallography and is represented with numbered atoms in four different ways:

(A) This structure is an illustration of the FeMo-cofactor centered on the titular Fe1 and Mo atoms extending away from the viewer where electron density is represented by blue mesh. Electron density at the position bridging the iron atoms Fe2 and Fe6, which in the uninhibited structure is occupied by the belt sulfur atom S2B, closely matches that of a CO ligand.

(B) The cofactor is displayed in the same orientation as in (A); however, the illustration now accounts for anomalous electron density (a result of the presence of heavier atoms such as sulfur) represented by green mesh. It is absent at the bridging, indicating that there is therefore no belt sulfur atom at the CO binding site.

(C) A side view of (A), this provides a more contoured illustration of the manner in which the CO bridges the two Fe atoms and allows for the determination of the length of the relevant atomic bonds.

(D) In the same orientation as (C), this structure also displays the close proximity of the ligand side chains to the CO binding site, implying that their position and identity likely affect the properties of the metal centers.

Given the complete displacement of S2B, we assessed whether the CO-inhibited protein could be reactivated or if it was irreversibly modified. Crystals of CO-inhibited MoFe-protein were active when dissolved in an assay mixture in the absence of CO. Furthermore, when we newly assayed CO-inhibited MoFe-protein from the original inhibition preparations after removal of CO (see supplementary material for assay details), a quantitative recovery (94 ± 4%) of the initial activity was obtained (Table 1). The reactivated MoFe-protein was subsequently reisolated from activity assay mixtures and crystallized, which yielded a structure at 1.43 Å resolution. The structural data of the protein (Av1 reactivated) revealed that S2B is regained by replacing the previously bound CO ligand, which results in the recovery of the resting state FeMo-cofactor. The full occupancy of the sulfur at the S2B site in the reactivated enzyme is evident by inspecting the 2Fobs – Fcalc electron density map, as well as the anomalous electron density map verifying the anomalous scattering contribution expected for S2B (Fig. 2, A and B).

Table 1. Nitrogenase activity. The acetylene reduction activity of “as-isolated” Av1, Av1-CO, and Av1 reactivated was measured by the quantification of ethylene production. Nitrogenase activity is quantitatively recovered upon reactivation. Errors represent standard deviations of three measurements.
Fig. 2. Reactivated MoFe-protein (Av1 reactivated). Refined structure of the FeMo-cofactor at a resolution of 1.43 Å. (A) View along the Fe1-C-Mo direction. The electron density (2Fobs – Fcalc) map is contoured at 4.0 σ and represented as blue mesh. Electron density at the S2B site is in excellent agreement with a regained sulfur. (B) Same orientation as (A) superimposed with the anomalous density map (green) at a resolution of 2.15 Å contoured at 4.0 σ, showing the presence of anomalous density at the S2B site. Color scheme is according to Fig. 1.

How does the structure of the CO-bound FeMo-cofactor change on the removal of the inhibitor?

Does the displaced S2B sulfur atom return to its original, uninhibited belt position, reactivating the protein? p>


The structure of the reactivated, CO-unbound cofactor was obtained through crystallography and is represented with numbered atoms in two different ways.

(A) This structure is an illustration of the FeMo-cofactor centered on the titular Fe1 and Mo atoms extending away from the viewer. Blue mesh represents the electron density; in the position bridging the iron atoms Fe2 and Fe6, it in agreement with sulfur atom occupation as opposed to the CO inhibitor.

(B) Now accounting for anomalous electron density represented by green mesh, the cofactor is displayed in the same orientation as in (A) where the presence of anomalous electron density at the bridging site is also indicative of the return of the S2B atom.

The finding that S2B can be reversibly replaced by CO raises the question of where this atom is located in the CO-inhibited state. If the S2B binding site is ordered, candidate locations should be evident by an inspection of the 7100 eV anomalous difference Fourier map. In this manner, one site per catalytic unit was identified with anomalous density compatible with sulfur. This potential sulfur binding site (SBS) is positioned ~22 Å away from the S2B position in the FeMo-cofactor and consists of a small protein pocket at the interface of the α and β subunits, formed by the side chains of residues α-Arg93, α-Thr104, α-Thr111, α-Met112, β-Asn65, β-Trp428, β-Phe450, and β-Arg453 (Fig. 3, A and B). The positive surface charge of the cavity is suited to accommodate an anionic species such as HS– or S2– (Fig. 3B). In previous structures of the resting state enzyme [Protein Data Bank (PDB) IDs: 1M1N and 3U7Q], this site has been assigned as water; note that the density at this site is also decreased in the structure of reactivated Av1 (fig. S2). The potential SBS is connected to the FeMo-cofactor–binding pocket by a noncontinuous water channel, and conformational rearrangements would be needed to accommodate the reversible migration of S2B from and to the active site. Although we have observed a correlation between density at the potential SBS and the CO-inhibited state of the MoFe-protein, the identity of this site as the displaced S2B cannot be achieved solely based on crystallographic data. In assessing the relevance of this site, it should be noted that the residues forming the pocket are poorly conserved, the site seems rather remote from the FeMo-cofactor, and because sulfur and chloride have similar anomalous scattering properties at 7100 eV, it is possible that this is a general anion-binding site in Av1.

Fig. 3. Overview of the potential sulfur binding site (SBS) in the CO-inhibited MoFe-protein (Av1-CO). (A) Location of the potentially bound sulfur in a protein cavity on the interface between the α and β subunits of the α2β2 MoFe-protein. The potential SBS is located 22 Å away from its former position in the FeMo-cofactor (S2B-site). (B) Close-up view of the binding cavity. Positive surface charge is represented in blue, negative surface charge in red. The anomalous density map at a resolution of 2.1 Å is represented as green mesh and contoured at 4.0 σ, showing the presence of anomalous density at the potential SBS. The side-chain sulfur of α-Met112 provides an internal standard for full occupancy. Alternate conformations for β-Arg453 are indicated. The color scheme is according to Fig. 1.

If the belt sulfur atom S2B is displaced by CO inhibition, to where does it relocate?

What evidence supports this identification of the potential sulfur binding site (SBS)?

Why are the side chains included in the analysis of the SBS?


The potential binding site (SBS) of the displaced belt sulfur atom S2B is determined in the context of the greater CO inhibited MoFe-protein.

(A) This illustrates the interface between the and α and β subunits of the protein, showing the placement of the proposed SBS relative to the FeMo-cofactor and highlighting the significance of the 21.8-Å distance between the two.

(B) A representation of the binding cavity wherein the potential SBS is located; positive and negative surface charges on the cavity are represented by blue and red, respectively, while anomalous electron density at the SBS, indicative of the presence of a sulfur atom S2B, is represented by green mesh. Sides chains are included as references of electron density and for to provide a full illustration of possible alternative protein conformations.

The crystallographic characterization of CO-bound FeMo-cofactor of the MoFe-protein has important implications for the mechanism of substrate reduction by nitrogenase: The CO-binding site is close to the side chains of residues α-His195 (2.8 Å, NE2–OC distance) and α-Val70 (3.4 Å, closest methyl–OC distance). Modifications to both side chains were reported to substantially alter the catalytic properties of the enzyme. An α-His195 to α-Gln195 mutation resulted in the loss of N2 reduction activity, while an α-Val70 to α-Ala/Gly70 alteration was reported to enable the reduction of longer carbon-chain substrates and highlighted the potential involvement of Fe6 in substrate reduction (29–33). In the structure presented here, α-His195 is in hydrogen bonding distance to the oxygen of CO, and α-Val70 directly flanks the binding site (Fig. 1D).

The displacement of S2B could be facilitated by a proton donation from α-His195 to yield either HS- or H2S, which would generate a better leaving group than S2-. Although the dissociation of a sulfur may seem surprising, it opens up the ligand binding site, because the large radius of S2- effectively shields the cofactor Fe atoms in the resting state from substrate and/or inhibitor attack (2). The more general implication that binding of exogenous ligands can be accompanied by the reversible dissociation of at least one belt sulfur from the metal sites of the FeMo-cofactor changes the present view of the structural inertness of the [7Fe:9S:C:Mo]-R-homocitrate cluster toward ligand exchange. The relative lack of reactivity of the resting state is a striking property of the FeMo-cofactor, and the requirement for more highly reduced forms to bind substrate and inhibitors may reflect the need to dissociate sulfur ligands from Fe sites.

The displacement of the belt sulfur S2B by carbon monoxide causes the FeMo-cofactor scaffold to lose its intrinsic three-fold symmetry. Additionally, Fe1, the interstitial carbon, and molybdenum are no longer aligned, creating an asymmetry in the resulting [7Fe:8S:C:Mo] cluster (Fig. 1C). The modest adjustments of the remaining scaffold upon CO binding are suggestive of an important role for the interstitial carbon in stabilizing the cofactor during rearrangements and substitutions to the coordination environment of the irons (34,35).

The experimental manipulations used to generate the CO-inhibited structure are distinct from those reported in previous spectroscopic studies; hence, it is not possible to unambiguously assign the structure to one of the many annotated spectroscopic states. Note that many of the previously identified states undergo dynamic interchanges, including photoinduced transitions between states (25). Like the structure presented here, the spectroscopically identified “lo-CO” state has been proposed to involve one molecule of CO bound to the active site in a bridging mode (2226). A state with two CO bound to Fe2 and Fe6 could correspond to the “high-CO” form (2223) and might represent an intermediate relevant to the C–C coupling reaction.

The generation and successful stabilization of CO-bound MoFe protein under turnover conditions has culminated in a crystal structure that provides a detailed view of a ligand bound to the nitrogenase active site. The observations that CO is isoelectric to N2, is a potent yet reversible inhibitor of substrate reduction without impeding proton reduction to dihydrogen, and is bound in close proximity to previously determined catalytically important residues emphasize the relevance of the CO-bound structure toward understanding dinitrogen binding and reduction. This sheds light on N2 activation based on a di-iron edge of the FeMo-cofactor and in this respect resembles the Haber-Bosch catalyst that also uses an iron surface to break the N–N triple bond. The demonstrated structural accessibility of CO-bound MoFe-protein opens the door for comparable studies on a variety of inhibitors and substrates, with the goal of understanding the detailed molecular mechanism of dinitrogen reduction by nitrogenase.

Supplementary Materials

Materials and Methods
Figs. S1 and S2
Tables S1 and S2
References (3643)

References and Notes

  1. B. K. Burgess, D. J. Lowe, Chem. Rev. 96, 2983–3012 (1996).

  2. J. B. Howard, D. C. Rees, Proc. Natl. Acad. Sci. U.S.A. 103, 17088–17093 (2006).

  3. L. C. Seefeldt, B. M. Hoffman, D. R. Dean, Annu. Rev. Biochem. 78, 701–722 (2009).

  4. M. M. Georgiadis et al., Science 257, 1653–1659 (1992).

  5. J. Kim, D. C. Rees, Nature 360, 553–560 (1992).

  6. H. Schindelin, C. Kisker, J. L. Schlessman, J. B. Howard, D. C. Rees, Nature 387, 370–376 (1997).

  7. J. W. Peters, K. Fisher, W. E. Newton, D. R. Dean, J. Biol. Chem. 270, 27007–27013 (1995).

  8. D. J. Lowe, K. Fisher, R. N. F. Thorneley, Biochem. J. 292, 93–98 (1993).

  9. O. Einsle et al., Science 297, 1696–1700 (2002).

  10. T. Spatzal et al., Science 334, 940 (2011).

  11. K. M. Lancaster et al., Science 334, 974–977 (2011).

  12. Y. Hu, M. W. Ribbe, Coord. Chem. Rev. 255, 1218–1224 (2011).

  13. D. C. Rees et al., Philos. Trans. A Math. Phys. Eng. Sci. 363, 971–984 (2005).

  14. D. J. Lowe, R. N. F. Thorneley, Biochem. J. 224, 895–901 (1984).

  15. R. N. F. Thorneley, D. J. Lowe, Biochem. J. 215, 393–403 (1983).

  16. Chatt, J. R. Dilworth, R. L. Richards, Chem. Rev. 78, 589–625 (1978).

  17. J. C. Hwang, C. H. Chen, R. H. Burris, Biochim. Biophys. Acta 292, 256–270 (1973).

  18. L. M. Cameron, B. J. Hales, Biochemistry 37, 9449–9456 (1998).

  19. L. C. Davis, M. T. Henzl, R. H. Burris, W. H. Orme-Johnson, Biochemistry 18, 4860–4869 (1979).

  20. C. C. Lee, Y. Hu, M. W. Ribbe, Science 329, 642–642 (2010).

  21. Y. Hu, C. C. Lee, M. W. Ribbe, Science 333, 753–755 (2011).

  22. H.-I. Lee, L. M. Cameron, B. J. Hales, B. J. Hoffman, J. Am. Chem. Soc. 119, 10121–10126 (1997).

  23. L. Yan et al., Eur. J. Inorg. Chem. 2011, 2064–2074 (2011).

  24. S. J. George, G. A. Ashby, C. W. Wharton, R. N. F. Thorneley, J. Am. Chem. Soc. 119, 6450–6451 (1997).

  25. Z. Maskos, B. J. Hales, J. Inorg. Biochem. 93, 11–17 (2003).

  26. R. C. Pollock et al., J. Am. Chem. Soc. 117, 8686–8687 (1995).

  27. L. C. Davis, M. T. Henzl, R. H. Burris, W. H. Orme-Johnson, Biochemistry 18, 4860–4869 (1979).

  28. D. W. Mulder et al., Structure 19, 1038–1052 (2011).

  29. P. M. C. Benton et al., Biochemistry 42, 9102–9109 (2003).

  30. C. H. Kim, W. E. Newton, D. R. Dean, Biochemistry 34, 2798–2808 (1995).

  31. P. C. Dos Santos, S. M. Mayer, B. M. Barney, L. C. Seefeldt, D. R. Dean, J. Inorg. Biochem. 101, 1642–1648 (2007).

  32. D. J. Scott, H. D. May, W. E. Newton, K. E. Brigle, D. R. Dean, Nature 343, 188–190 (1990).

  33. J. Christiansen, V. L. Cash, L. C. Seefeldt, D. R. Dean, J. Biol. Chem. 275, 11459–11464 (2000).

  34. J. Rittle, J. C. Peters, Proc. Natl. Acad. Sci. U.S.A. 110, 15898–15903 (2013).

  35. I. Dance, Chem. Asian J. 2, 936–946 (2007).