Mechanism of Nitrogen Fixation by Nitrogenase: The Next Stage

1. Introduction

Nitrogen is an essential element contained in many biomolecules necessary to sustain life.^1,2 This element is abundantly available in Earth’s atmosphere in the form of dinitrogen (N₂) gas, yet most organisms are unable to metabolize N₂ because it is relatively inert.^3,4 Instead most organisms must obtain their N from “fixed” forms such as ammonia (NH₃) or nitrate (NO₃^–).⁵⁻⁷ Because fixed forms of N are continuously sequestered into sediments, rendering them unavailable for metabolism, and because they are also continuously converted to N₂ through the combined processes of nitrification and denitrification, life can only be sustained by conversion of N₂ to NH₃.^6,7 This latter process is known as N₂ fixation⁸ and is a critical step in the biogeochemical N cycle.^5,7,9 N₂ fixation occurs in three different ways: (i) through geochemical processes such as lightning,⁹ (ii) biologically through the action of the enzyme, nitrogenase,^10,11 found only in a select group of microorganisms,^12,13 and (iii) industrially through the Haber–Bosch process.^2,14,15 From the evolution of nitrogenase, approximately two billion years ago¹⁶ until the widespread use of the Haber–Bosch process in the 1950s, all life derived N from biological nitrogen fixation, with geochemical processes representing a minor contributor to the supply of fixed nitrogen.^2,7 Since the increase in use of the Haber–Bosch process, the biological and industrial processes contribute comparably to N₂ fixation.^5,7,9

Nitrogen fixation has a profound agronomic, economic, and ecological impact owing to the fact that the availability of fixed nitrogen represents the factor that most frequently limits agricultural production throughout the world.² Indeed, nearly half of the existing human population could not exist without application of the Haber–Bosch process for production of nitrogen fertilizers.^2,5 Given that over half of the fixed nitrogen input that sustains Earth’s population is supplied biologically, there has been intense interest in understanding how the nitrogenase enzyme accomplishes the difficult task of N₂ fixation at ambient temperature and pressure.^17,18 An understanding of biological N₂ fixation may further serve as the foundation for achieving two highly desirable, although so far unmet, goals: genetically endowing higher plants with the capacity to fix their own nitrogen,¹⁹⁻²¹ and developing improved synthetic catalysts based on the biological mechanism.^3,4,22−25

It has been over 150 years since Jodin first suggested that microbes could “fix” N₂,²⁶ and more than a century since the first isolation of N₂-fixing bacteria around 1900. In 1934, Burk coined the term “nitrogenase”^10,11 for the enzyme that catalyzes the conversion of N₂ to a bioaccessible form of nitrogen, and initiated the first meaningful studies of nitrogenase in living cells. Methods for extracting nitrogenase in an active form were developed in the early 1960s,²⁷⁻²⁹ opening the way for serious mechanistic investigations. The next 35 years witnessed intensive efforts by numerous investigators to reveal the structure and catalytic function of nitrogenase.³⁰⁻³⁴ These developments were summarized in the magisterial review by Burgess and Lowe in 1996.¹⁷ Key advances in understanding nitrogenase structure and function during those intervening years included the following: (i) It was determined that nitrogenase is a two-component system³⁵⁻³⁷ composed of the MoFe protein (also called dinitrogenase or component I) and the electron-transfer Fe protein (also called dinitrogenase reductase or component II).^34,38−41 (ii) A reducing source and MgATP are required for catalysis.⁴²⁻⁴⁵ (iii) Fe protein and MoFe protein associate and dissociate in a catalytic cycle involving single electron transfer and MgATP hydrolysis.³⁸ (iv) It was discovered that the MoFe protein contains two metal clusters: the iron–molybdenum cofactor (FeMo-co),^30,46 which provides the active site for substrate binding and reduction, and P-cluster, involved in electron transfer from the Fe protein to FeMo-co.^39,47−50 (v) Crystallographic structures were solved for both Fe⁵¹ and MoFe^{32,48,52−54} proteins. (vi) Also, the alternative V- and Fe-type nitrogenases, in which the Mo of FeMo-co is replaced by V or Fe, were discovered.¹⁸ Despite this accumulation of functional and structural information, the catalytic mechanism remained elusive.

The years since the Burgess and Lowe review¹⁷ have seen profound advances in understanding many aspects of nitrogenase structure and function. For example, the solutions of a number of high-resolution X-ray structures of the nitrogenase component proteins⁵⁵⁻⁶⁹ have provided insights into the nature of the active site FeMo-cofactor, most recently identifying the presence of an interstitial C atom,⁷⁰⁻⁷⁷ while structures of the two proteins in the complex⁷⁸⁻⁸¹ have identified their binding interface (Figures 1 and 2) and its alterations with the state of the bound nucleotide.⁶⁷ Likewise, great strides have been made in understanding the biosynthesis and insertion of the metal clusters of nitrogenase to form the mature proteins,^21,82−89 and the properties of the V-type nitrogenase.⁹⁰⁻⁹⁸ Recent studies have begun to shed light on the order of events during the catalytic cycle,⁹⁹⁻¹⁰³ including the nature of electron transfer between the metal clusters^62,104−111 and the roles of ATP binding and hydrolysis in these processes.^{55,68,99,112−121} Considerable progress has been made in the application of theoretical methods to various aspects of the nitrogenase mechanism.^122−140 Finally, progress has been made in expanding in the substrates of nitrogenases^93,141−149 to include CO^{95,96,98,150,151} and CO₂.^148,152

Figure 1.

Molybdenum nitrogenase. (A) One catalytic half of the Fe protein:MoFe protein complex with the Fe protein homodimer shown in tan, the MoFe protein α subunit in green, and the β subunit in cyan. (B) Space filling and stick models for the 4Fe–4S cluster (F), P-cluster (P), and FeMo-co (M). Made with Pymol and ChemDraw using PDB:2AFK.

Figure 2.

FeMo-cofactor and the side chains of selected amino acid residues of the MoFe protein. Numbering of iron atoms is according to the structure PDB coordinate 2AFK. Iron is shown in rust, molybdenum in magenta, nitrogen in blue, sulfur in yellow, carbon in gray, and oxygen in red.

The present narrative focuses on recent progress in understanding the mechanism of N₂ activation and reduction to ammonia by Mo-nitrogenase. The discussion begins with a short reminder of the kinetic scheme that describes nitrogenase catalysis.^33,103 It then turns to the successes in trapping catalytic intermediates of the MoFe protein by rapid freezing of turnover mixtures of Fe protein and of MoFe proteins, both wild-type and variants containing selected amino acid substitutions as a means to modulating reactivity.^{146,148,153,154} The use of EPR/ENDOR/ESEEM spectroscopic techniques applied to isotopically substituted trapped intermediates has allowed the identification and characterization of key intermediates along the N₂ reduction pathway.^154−156 This led to the formulation of a reaction mechanism based on the properties of catalytic intermediates and grounded in the reaction of hydrides associated with FeMo-co.¹⁵⁶ The mechanism not only satisfies all constraints on the mechanism provided by earlier studies, but has suggested and passed a stringent test.¹⁵⁷ This report recounts these advances, and expands on them.

2. Background

Two issues require consideration as a basis for discussion of recent advances in nitrogenase mechanism.^155,156 The first is the kinetic model that has been developed to describe the multistep reduction of N₂ to two NH₃, and its implications for the stoichiometry of this reaction,^33,103 implications that were mutually supported by experiment.¹⁵⁸ The second is the strategies and procedure that at last enabled the trapping of catalytic intermediates whose characterization by advanced paramagnetic resonance techniques underlies the progress in mechanism described here.

2.1. Kinetics and Stoichiometry

A “kinetic” foundation for a nitrogenase mechanism was developed by extensive studies in the 1970s and 1980s by many groups, especially Lowe and Thorneley and their co-workers.^17,33,103 The culmination of these extensive kinetic studies, which involved steady-state, stopped-flow, and freeze–quench kinetics measurements, was the Lowe–Thorneley (LT) kinetic model for nitrogenase function,^17,33,103 which describes the kinetics of transformations among catalytic intermediates (denoted E_n) where n is the number of steps of electrons/protons delivery to MoFe protein, Figure 3. Electron transfer from Fe protein to MoFe protein is driven by the binding and hydrolysis of two MgATP species within the Fe protein;⁹⁹ the release of the Fe protein after delivery of its electron is the rate-limiting step of catalysis.³³

Figure 3.

Simplified LT kinetic scheme that highlights correlated electron/proton delivery in eight steps. Although in the full LT scheme N₂ binds at either the E₃ or E₄ levels, the pathway through E₃ is de-emphasized here. LT also denotes the protons bound to FeMo-co (e.g., E₁H₁); for clarity we have omitted these protons in this scheme.

A central consequence of the kinetic measurements and defining feature of this scheme, Figure 3, is that the limiting enzymatic stoichiometry for enzyme-catalyzed nitrogen fixation is not what would be given by the simple balanced equation for reduction of N₂ to two NH₃ by six electrons/protons, but is given by eq 1

1

This is a conclusion that is in agreement with stoichiometric experiments by Simpson and Burris.¹⁵⁸ This equation highlights several key aspects of the nitrogenase mechanism, including the involvement of ATP hydrolysis in substrate reduction and the obligatory formation of 1 mol of H₂ per mole of N₂ reduced, an apparent “waste” of two reducing equivalents and four ATP per N₂ reduced.^17,33

Although the close of the previous millennium saw the accumulation of a vast breadth and depth of information about the reduction of N₂, H⁺, and a variety of other nonphysiological substrates,¹⁷ it was not until recently that studies have succeeded in characterizing E_n intermediate states beyond the resting-state E₀.^154−156 Thus, the early studies provided little direct experimental evidence regarding a reaction pathway, and hence, there was no possibility of integrating a reaction pathway and kinetic scheme, as is central to development of a mechanism based on the properties of catalytic intermediates.¹⁵⁶

2.2. Trapping and Characterization of Substrates

The first 40 years of study of purified nitrogenase did not see the definitive characterization of any intermediates associated with the binding and reduction of N₂,¹⁵⁹ leaving the identity of the reaction pathway unresolved. The way forward was provided by studies of nitrogenases with individual amino acid substitutions, which revealed that the residue at position α-70 within the MoFe protein, a valine, acts as a “gatekeeper” that sterically controls the access of substrate to the active site FeMo-co (Figure 2).^146,154 The side chain of this amino acid residue is located over one FeS face of FeMo-cofactor (that includes Fe atoms 2, 3, 6, and 7) thereby also implicating Fe as the site of substrate binding, while the α-195^His was inferred to be involved in proton delivery (Figure 2).^160−165

Use of MoFe protein substituted at one or both of these residues enabled freeze–quench trapping of a number of nitrogenase turnover intermediates, almost all of which show an EPR signal arising from an S = ¹/₂ state of FeMo-co, rather than the S = ³/₂ state of resting-state FeMo-co.^153,154 The procedures developed with these variants even enabled N₂−intermediate trapping with enzyme.¹⁶⁶ The first fruit of this approach was the trapping of a state during reduction of the alkyne propargyl alcohol to the corresponding alkene.^145,167 An intermediate trapped using MoFe protein variants was shown by ENDOR studies to be a wholly novel bio-organometallic structure in which the alkene product of alkyne reduction by nitrogenase binds as a π-complex/ferracycle to a single Fe ion of FeMo-cofactor, presumed to be Fe6.¹⁶⁸ This was followed by characterization of intermediates formed during the reduction of H⁺ under Ar,¹⁶⁹ and, finally, identification of four associated with N₂ reduction itself.^{147,153,160,166,170,171}

Paramagnetic resonance methods have proven to be uniquely advantageous for characterization of trapped nitrogenase intermediates.¹⁵⁵ At the most basic level, FeMo-co in the E₀ resting-state of MoFe protein is an odd-electron (“Kramers”; half-integer spin, S = ³/₂¹⁷²), EPR-active cluster, and therefore, intermediate states that have accumulated an even number of electrons also will be EPR-active. Focusing on nitrogen fixation, FeMo-co then will be EPR-active in the E_n states, n = 2, 4, 6, 8, formed along the pathway for accumulation of the stoichiometrically required eight [e^–/H⁺], eq 1. In contrast, E_n states n = 1, 3, 5, 7 will be even-electron, and FeMo-co will either be diamagnetic or in an integer-spin (“non-Kramers” spin-state)¹⁷³ cluster, which also can be EPR-active under appropriate conditions.^173,174

As will be illustrated below, electron–nuclear double-resonance (ENDOR) spectroscopy,^175,176 supported by related techniques ESEEM and HYSCORE,¹⁷⁷ is uniquely suited for the study of freeze–quench trapped intermediates. These techniques give NMR-like spectra of nuclei that are hyperfine-coupled to the electron spin of an EPR-active cluster. The importance of the techniques rests on several aspects. ENDOR is broad-banded: with isotopic enrichment it can monitor every atom in a metalloenzyme active site. Thus, when interpreted in the context of the X-ray structure of the resting-state, it can reveal the electronic and metrical structure of a catalytic intermediate. It is selective: it interrogates only EPR-active states. It is high-resolution: it can resolve and interrogate the signals from multiple distinct EPR-active centers. It is sensitive: we have successfully analyzed the properties of intermediates present in ∼20% abundance in a sample containing ∼100 μM MoFe protein. Viewed another way, ENDOR is capable of selecting and characterizing a small fraction of the MoFe protein in a sample. In contrast, for example, Mössbauer and X-ray absorption techniques, which have made enormous contributions to the study of resting-state nitrogenase, interrogate all FeMo-co in a sample, and if the state of interest is a small minority, its signal is buried and lost. Recently, however, an X-ray spectroscopic study has given information about a freeze–quenched nitrogenase intermediate.¹⁷⁸

3. Intermediates of Nitrogenase Activation

According to the simplified LT kinetic scheme of Figure 3, the first four of the eight [e^–/H⁺] of nitrogen fixation accumulate prior to N₂ binding, which occurs at the E₄ stage. The complete scheme^17,33,103 allows for N₂ binding at E₃ as well, but E₄ uniquely places the enzyme on the pathway to N₂ hydrogenation.

3.1. E₁–E₃

The E₁ state contains one-electron reduced cofactor, and has been assigned as an integer-spin species on the basis of Mössbauer studies of MoFe protein trapped during turnover under N₂.^179,180

High-spin EPR signals (S = ³/₂), denoted as 1b and 1c, thought to be associated with E_n states, n ≤ 4, were first observed 35 years ago for samples of wild-type nitrogenase trapped during turnover using a variety of conditions,¹⁸¹ and more recently were studied by rapid freeze–quench EPR.¹⁸² The kinetics of appearance of 1b and 1c demonstrated that they must be assigned to reduced states of cofactor, n > 1, rather than just as conformers of the FeMo-co resting-state. However, the kinetics of appearance of the stronger 1b signal was a puzzle: they were best described by assigning 1b to the E₃ state, which would seem to require that FeMo-co be in an integer-spin (non-Kramers) state, contrary to observation. Most likely, this apparent contradiction reflects uncertainties in the rate constants used in the kinetics analysis, and 1b represents an E₂ state. During cryoannealing experiments¹⁸³ discussed below, we definitively observed that FeMo-co of E₂ is in a high-spin (S = ³/₂) state, but at least in the α-70^Ile variant its g-values were distinct from those of 1b. The spectrum of the 1c species is weaker in intensity. It may represent a conformer of the resting or 1b states, or may correspond to even more reduced states, as its effective formation requires a high molar ratio of Fe protein to MoFe protein, corresponding to higher electron flux.

3.2. E₄: The “Janus Intermediate”

In this subsection we describe the trapping and EPR/ENDOR characterization of the E₄ intermediate as activated by the accumulation of four [e^–/H⁺] for binding and reduction of N₂. The structure of E₄ as determined by ENDOR spectroscopy, and integrated into the LT kinetic scheme, has been the key to recognizing the central role of hydrides in the mechanism for nitrogen fixation.¹⁵⁶ We then discuss the E₁–E₄ states associated with electron accumulation by MoFe protein; subsequent sections discuss the trapped states associated with the N₂ reduction pathway following N₂ binding.

Early in the search for intermediates,¹⁴⁶ the α-70^Val→Ile substitution in the MoFe protein was shown to deny access of all substrates to the active site, except protons.^169,184 Samples of this substituted MoFe protein freeze–quenched during turnover under Ar exhibited a new S = ¹/₂ EPR signal,¹⁶⁹ which also can be observed at lower concentrations during turnover of wild-type MoFe protein under Ar.^181,1851,2H ENDOR spectroscopic analysis of this trapped state¹⁶⁹ revealed the presence of two strongly hyperfine-coupled, metal-bridging hydrides [M–H–M′]: (i) The finding that the bound hydrides have a large isotropic hyperfine coupling, a_iso ≈ 24 MHz, led to their assignment as hydrides bound to metal ion(s) of the core. (ii) The anisotropic hyperfine contribution, T = [−13.3, 0.7, 12.7] MHz, exhibits almost complete rhombicity, as defined by the form T_rh ≈ [t, 0, −t]. This form rules out terminal hydrides, which would have a roughly axial T,¹⁸⁶ and is precisely the form first predicted¹⁸⁷ and then confirmed¹⁸⁸ to be associated with a hydride bridging two paramagnetic metal ions, namely as [Fe–H–Fe] and/or [Mo–H–Fe] fragments.

⁹⁵Mo ENDOR measurements subsequently established that both hydrides bridge two Fe ions, forming two [Fe–H–Fe] fragments (Figure 4), as follows.¹⁸⁹ Equations for the anisotropic hyperfine interaction matrix, T, of a nucleus that undergoes through-space dipolar interactions to two spin-coupled metal ions¹⁸⁷ were generalized to describe an arbitrary [M₁–H–M₂] fragment of a spin-coupled cluster. The components of T are a function of the [M₁–H–M₂] geometry and of the coefficients [K₁, K₂] that describe the projection of the total cluster spin on the two local M-ion spins. The ⁹⁵Mo ENDOR measurements of the intermediate showed a very small isotropic hyperfine coupling, a_iso(⁹⁵Mo) ∼ 4 MHz, which indicated that K_Mo is too small to yield the rhombic dipolar coupling, T_rh, observed in this intermediate.¹⁸⁹ The model for E₄ displayed in Figure 4 is completed by placement on sulfurs of the two protons^190,191 that form part of the delivery of 4[e^–/H⁺] (Figure 3). The protons are so placed because they must be near to the negative charge density associated with the hydrides in order to obtain the electrostatic stabilization implicit in the required accumulation of one proton for each electron delivered to MoFe protein;¹⁷ other arrangements are possible, such as putting both protons on doubly bridging sulfur, but see below.

Figure 4.

Depiction of E₄ as containing two [Fe–H–Fe] moieties, emphasizing the essential role of this key “Janus intermediate”, which comes at the halfway point in the LT scheme, having accumulated four [e^–/H⁺], and whose properties have implications for the first and second halves of the scheme. Janus image adapted from http://www.plotinus.com/janus_copy2.htm. Figure adapted with permission from ref (156). Copyright 2013 American Chemical Society.

Cryoannealing this “dihydride” intermediate in the frozen state at −20 °C, which prevents further delivery of electrons from the Fe protein, showed that it relaxes to the resting FeMo-co state by the successive loss of two H₂ molecules.¹⁸³ According to the LT scheme, only E₄ would undergo this two-step relaxation process (Scheme 1), with the first relaxation step of E₄ yielding H₂ and the E₂ state, the second step returning FeMo-co to the E₀ stage with loss of a second H₂, and the production of H₂ being revealed by solvent kinetic isotope effects in both stages. This relaxation protocol thus revealed that the trapped intermediate is the E₄ state, which has accumulated n = 4 electrons and protons.¹⁸³ As the relaxation measurements involved tracking the kinetically linked conversion of E₄ into E₂, and the conversion of E₂ into resting-state E₀, the measurements further allowed an unambiguous identification of the EPR signal associated with E₂ (see above).

Scheme 1.

Examination of the simplified version of the LT scheme of Figure 3 reveals that E₄ is a key stage in the process of N₂ reduction.^33,103 Indeed, we have denoted it as the “Janus” intermediate, referring to the Roman God of transitions who is represented with two faces, one looking to the past and one looking to the future (Figure 4).¹⁵⁶ Looking “back” from E₄ to the steps by which it is formed, E₄ is the culmination of one-half of the electron/proton deliveries during N₂ fixation: four of the eight reducing equivalents are accumulated in E₄, before N₂ even becomes involved. Looking “forward”, toward NH₃ formation, E₄ is the state at which N₂ hydrogenation begins, and it is involved in one of the biggest puzzles in N₂ fixation: “why” and “how” H₂ is lost upon N₂ binding.

To date, we have visualized E₄ by placing its two hydrides on the Fe2, 3, 6, 7 face of resting-state FeMo-co and sharing a common vertex at Fe6, Figure 4. Although the hydrides may well exhibit fluxionality at ambient temperature, their ability to adopt a configuration with a common vertex is required by the reductive elimination (re) mechanism of reversible H₂ release upon N₂ binding (section 7), and Fe6 is favored from earlier kinetic studies on MoFe protein variants.^{69,144,145,154,166,168} However, this model is only one of four possible configurations based on the resting structure that have two hydrides sharing an Fe6 vertex. To visualize these structures we have built the bound hydrides onto the crystal structure of resting-state FeMo-co using Fe–H distances from model complexes,^188,192 Figure 5.

Figure 5.

Mockups of the “Janus” E₄ intermediate in which the two bridging hydrides [Fe–H–Fe] revealed by ENDOR spectroscopy are built onto the resting-state crystal structure. These models of FeMo-co have Fe6 as a “vertex” for the two bridging hydrides to facilitate reductive elimination. The figure was generated using the coordinate file PDB:2AFK. Iron is shown in rust, molybdenum in magenta, sulfur in yellow, carbon in dark gray, and hydrogen in light gray.

Quantum chemical computations will test these alternatives. However, the experimentally determined relative orientation of the hyperfine tensors of the two hydrides provides a significant constraint on their placement within E₄. Given the stability of the FeMo-co structure that is likely imparted by the interstitial carbide, it seemed plausible to us that consideration of the constructed models of Figure 5 would allow us to test these alternative hydride distributions, even though it is beyond doubt that the structure of FeMo-co will distort upon substrate binding. This exercise (see Supporting Information) provides support for the topology of hydride binding pictured for the Janus E₄ intermediate in Figure 4, with hydrides bridging Fe2/Fe6 and Fe6/Fe7 (Figure 5A,B), as opposed to Figure 5C,D, but does not discriminate between the structures of Figure 5A,B. In discussions below, we retain the placement of the E₄ hydrides shown in Figure 4 (Figure 5A) as being more readily visualized in discussions of mechanism.

The characterization of the hyperfine interactions of the metal-ion core of E₄ that began with the ⁹⁵Mo ENDOR measurements¹⁸⁹ was completed by an ENDOR study of the ⁵⁷Fe atoms of the E₄ FeMo-co through use of a suite of advanced ENDOR methods.¹⁹³ The determination of hyperfine interactions for two ligand hydrides and all eight metal ions of FeMo-cofactor in this state will provide the experimental test that guides future computational studies that seek to characterize the geometric and electronic structure of E₄.

Storage of the reducing equivalents accumulated in the E₄ state as bridging hydrides has major consequences. A bridging hydride is less susceptible to protonation than a terminal hydride, and thus bridging hydride(s) diminish the tendency to lose reducing equivalents through the formation of H₂ (Scheme 1), thereby facilitating the accumulation of reducing equivalents by FeMo-co. This mode also lowers the ability of the hydrides to undergo exchange with protons in the environment, a characteristic that is shown to be of central importance below. However, the bridging mode also lowers hydride reactivity toward substrate hydrogenation, relative to that of terminal hydrides.^194,195 As a result, substrate hydrogenation most probably incorporates the conversion of hydrides from bridging to terminal binding modes.¹⁹⁶ We next discuss how the structure found for E₄ guides assignment of structures for the E₁–E₃ states. Subsequently, we show how the E₄ structure defined possible mechanisms for coupling H₂ loss to N₂ binding.

3.3. Redox Behavior and Hydride Chemistry of E₁–E₃: Why Such a Big Catalytic Cluster?

Given that the four accumulated electrons of E₄ reside not on the metal ions but, instead, are formally assigned to the hydrides of the two Fe-bridging hydrides, what then are the proper descriptions of E₁–E₃? The addition of one electron/proton to the MoFe protein results in the E₁ state, and a Mössbauer study of nitrogenase trapped during turnover under N₂¹⁸⁰ suggested that this state contains the reduced metal-ion core of FeMo-co, denoted M^– in Figure 6A. The presence in E₄ of two bridging hydrides/two protons led us to propose that upon delivery of the second electron/proton to form E₂ the metal–sulfur core of the FeMo-cofactor “shuttles” both electrons onto one proton to form an [Fe–H–Fe] hydride, leaving the second proton bound to sulfur for electrostatic stabilization and the core formally at the resting-state, M⁰, redox level (also commonly referred to as, M^N),¹⁹⁷ Figure 6A. A subsequent, analogous, two-stage process would then yield the E₄ state, with its two [Fe–H–Fe] hydrides, two sulfur-bound protons, and the core at the resting-state, M⁰, redox level.¹⁹³

Figure 6.

Formulations of E₁–E₄ derived from consideration of E₄ as containing two bound hydrides and two protons. (A) Assuming reduction of the core in n = 1, 3 states. (B) Alternative formulation of E₁–E₄ under the assumption of hydride formation at every stage, in which case the core is formally oxidized for E_n, n = 1, 3. Symbols: M represents FeMo-co core; superscripts are charge difference between core and that of resting-state (commonly denoted M^N); the number of bound protons/hydrides are indicated. Adapted with permission from ref (156). Copyright 2013 American Chemical Society.

Such a process of acquiring the four reducing equivalents of E₄ involves only a single redox couple connecting two formal redox levels of the FeMo-co core of eight metal ions; M⁰ the resting-state, and M^– the one-electron reduced state of the core, Figure 6A.¹⁹³ Indeed, comparisons of the ⁵⁷Fe ENDOR results for the E₄ intermediate with earlier ⁵⁷Fe ENDOR studies and “electron inventory analyses”^155,198 of nitrogenase intermediates led us to the remarkable suggestion that, throughout the nitrogenase catalytic cycle, the FeMo-cofactor would cycle through only two formal redox levels of the metal-ion core. On reflection, it seems obvious that only by “storing” the equivalents as hydrides is it possible to accumulate so much reducing power at the constant potential of the Fe protein. We further proposed that such “simple” redox behavior of a complex metal center might apply to other FeS enzymes carrying out multielectron substrate reductions.¹⁹³

Considering the critical role of hydrides in storing reducing equivalents, we also suggested that the E₁ and E₃ states, respectively, might well contain one and two bridging hydrides bound to a formally oxidized metal-ion core (Figure 6B),¹⁰⁰ in which case the single redox couple accessed would formally be that between M⁰ and M⁺. In section 9, below we adopt this “oxidative” formulation of the E₁–E₃ structures. We emphasize that a third formulation of E₁(E₃), with hydride(s) bound to M⁰ and the presence of oxidized P-cluster, is ruled out by the absence of EPR signals from P⁺ in samples trapped under turnover conditions.

If the FeMo-cofactor does not utilize more than one redox couple during catalysis, then why is it constructed from so many metal ions? As discussed above, the hydrides of E₄ bind to at least two, and plausibly three Fe atoms of a 4-Fe face of FeMo-co, as shown in Figures 4 and 5. It is further possible that catalysis is modulated by the linkage of Fe ion(s) to the anionic atom C that is centrally located within the metal–sulfur core of the FeMo-cofactor.^70,71 Formation of such a 4Fe face and the incorporation of C is not likely with less than a trigonal prism of six Fe ions linked by sulfides to generate these structural features. In this view, the trigonal prismatic FeMo-cofactor core of six Fe ions plus C generates the catalytically active 4Fe face. This prism is capped, and its properties are likely “tuned”, by two “anchor” ions, one Fe plus a Mo, or a V or Fe in the alternative nitrogenases.

Finally, and far from least, as we have consistently noted (see section 7), there is good reason to imagine N₂ and/or the N₂H_x reduction intermediates may interact with multiple Fe ions on a FeMo-co face.

3.4. Why Does Nitrogenase Not React with H₂/D₂/T₂ in the Absence of N₂?

The following question is commonly raised: If electrons accumulated in E_n intermediates, n = 2–4, can relax to E_n-2 through formation and release of H₂ during turnover, as captured in the partial LT scheme, Scheme 1, why does the enzyme not exhibit the reverse of this reaction, and react with H₂/D₂ in what might appear to be the “microscopic reverse” of H₂ release? We have proposed that H₂ formation involves protonation of an [Fe–H–Fe], and at a basic level, all three relaxation processes of Scheme 1 should have much the same characteristics. For simplicity in addressing this issue, we focus on the “first” of these, the E₂ → E₀ relaxation, and ask why E₀ is not reduced by H₂ to form E₂, eq 2

2

A logical answer to this question begins with the recognition that the LT kinetic scheme for N₂ fixation, Figure 3 (also denoted the “MoFe protein cycle”), and the segment presented in Scheme 1, omit the reactions of the Fe protein for clarity; these are treated as a separate “Fe-protein cycle”.^{17,33,103,154} A stoichiometrically correct scheme that merges the Fe protein and MoFe protein cycles is given in Figure 7. It reminds us that E₂ is formed by two steps of Fe → MoFe protein ET, with each step involving hydrolysis of two ATP molecules to drive a reaction that is highly “uphill” energetically.

Figure 7.

Formation and relaxation of E₂. In-line: The “on-path” two-step, ATP-dependent addition of two H⁺/e^– to MoFe protein to form E₂. Off-line: Representation of the exergonic (free energy, +|ΔG_h|) “off-path” relaxation of E₂, liberating H₂ and directly regenerating E₀ without intervention of Fe protein, and of the energetically (free energy, +|ΔG_h|) and kinetically forbidden reverse of this process; E₀′ is a putative intermediate state that causes the reaction of E₀ not to be the microscopic reverse of the release of H₂ from E₂ (see text).

Clearly the E₂ → E₀ relaxation with accompanying loss of H₂ is not the “reverse” of the turnover formation of E₂ from E₀; neither Fe protein reduction nor ATP formation is involved. Instead, it is a side-reaction of E₂. Indeed, it is even quite unlikely that a direct reaction of H₂ with E₀ to form E₂ (eq 2) would be the microscopic reverse of the E₂ → E₀ relaxation with accompanying loss of H₂. Moreover, the steric congestion caused by the sulfurs at the six tetrahedral [FeS₃C] sites of the FeMo-co “waist” requires that the core must relax for the Fe to bind any ligand; in particular it is probable that the structure of the [Fe₇S₉MoC] core of FeMo-cofactor of E₀ (denoted M^N) is altered during the reduction of E₀ by two [e^–/H⁺] to form E₂ (also see section 9.2). In this case, as illustrated in Figure 7, the relaxation of E₂ with loss of H₂ to form the resting E₀ state would be a 2-step process. The loss of H₂ by E₂ would be expected to form a state (denoted here E₀′) that contains FeMo-cofactor in a conformation approximating that of E₂, corresponding to a metastable conformation of its resting redox level (denoted M^N′); this conformer would in turn undergo a M^N′ → M^N structural relaxation associated with E₀′ → E₀ relaxation. The reduction of E₀ by H₂ (eq 2) by the microscopic reverse of this two-step relaxation would correspondingly take place in two steps, Figure 7, with the initial E₀ → E₀′ thermal activation associated with the conformational change, M^N → M^N′, adding an activation free energy (denoted |ΔG^†|) and kinetic barrier to the endergonic reduction of E₀′ → E₂ by H₂ (free energy denoted |ΔG_h′|).

What would be the free energy for reduction of nitrogenase by H₂, as in eq 2? An upper bound for the free energy change for this reaction, ΔG_h2, would be 4 times the negative of the free energy change for the hydrolysis of ATP to form ADP and P_i (−ΔG_Hyd ∼ +7 kcal/mol; total, endergonic by ∼ +28 kcal/mol), that is required for the formation of E₂ through the delivery of reducing equivalents by Fe protein; roughly compatible with that, oxidative addition of H₂ to an Fe–S center (hydrogenase), corresponding to |ΔG_h′|, is uphill by at least +20 kcal/mol,^199,200 to which must be added the conformational free energy |ΔG^†|. Given the strongly endergonic nature of eq 2, coupled with the kinetic penalty associated with the activation of E₀ to E₀′, it becomes clear why H₂ is not observed to reduce FeMo-cofactor.

4. “Dueling” N₂ Reduction Pathways

Researchers have long considered two competing proposals for the second half of the LT kinetic scheme, the reaction pathway for N₂ reduction that begins with the Janus E₄ state.^17,139,155 These invoke distinctly different intermediates, Figure 8, and computations suggest they likely involve different metal-ion sites on FeMo-co.¹³⁹ The “distal” (D) pathway is associated with the Chatt^4,201 or Chatt–Schrock cycle³ because it is utilized by inorganic Mo complexes discovered by these investigators to cleave N₂ (Chatt and co-workers^202,203) and, most dramatically, to catalytically fix N₂ (Schrock and co-workers^24,204−206). In this cycle, which has been suggested to apply to nitrogen fixation by nitrogenase with Mo as the active site,¹³⁹ a single N of N₂ is hydrogenated in three steps until the first NH₃ is liberated, and then the remaining nitrido-N is hydrogenated three more times to yield the second NH₃. In the “alternating” (A) pathway that has been suggested to apply to catalysis at Fe of FeMo-co,^25,131 the first two hydrogenations generate a diazene-level intermediate, the next two form hydrazine, and the first NH₃ is liberated only by the fifth hydrogenation (Figure 8). As one can imagine alternative structures for the intermediates, the figure focuses on the defining difference between D and A pathways as being the release of the first NH₃ in the D as occurring after three hydrogenations of substrate, the addition of three [e^–/H⁺] to substrate, but only after five hydrogenations in A.

Figure 8.

Comparison of distal (D) and alternating (A) pathways for N₂ hydrogenation, highlighting the stages that best distinguish them, most especially noting the different stages at which NH₃(1) is released.

Simple arguments can be made for both pathways and for either Fe and Mo as the active site.^{17,154,155,207} For example, the A route is suggested by the fact that hydrazine is both a substrate of nitrogenase and is released upon acid or base hydrolysis of the enzyme under turnover,^17,208−211 and is favored in computations with reaction at Fe,¹³¹ while the D route was suggested by the fact that until recently the only inorganic complexes that catalytically fix N₂ employ Mo and function via the D route,²⁴ which is computationally favored for reaction at Mo.¹³⁹ Interestingly, this argument is somewhat weakened by a recent study that reported small W clusters fix N₂ by the A pathway.²¹² More significantly, the argument based on N₂ cleavage and catalytic N₂ fixation by Mo model complexes has lost ground by the quite recent discovery of Fe model complexes that cleave N₂ (Holland and covworkers^22,213) and indeed that also catalytically fix N₂ (Peters and co-workers²¹⁴).

Further support for the A pathway is provided by considerations of the alternative nitrogenases. It is most economical to suggest that both the Mo-dependent nitrogenase studied here and the V-type nitrogenase reduce N₂ by the same pathway. As V-nitrogenase produces traces of N₂H₄ while reducing N₂ to NH₃,²¹⁵ then according to Figure 8 this enzyme can be concluded to function via the A pathway, implying the same is true for Mo-nitrogenase.

5. Intermediates of N₂ Reduction: E_n, n ≥ 4

As can be seen in Figure 8, characterization of catalytic intermediates formed during the reduction of N₂ could distinguish between the D and A pathways. However, such intermediates had long eluded capture until four intermediates associated with N₂ fixation were freeze-trapped and characterized by ENDOR spectroscopic studies.^154,155 These four states were generated under the hypothesis that intermediates associated with different reduction stages could be trapped using N₂ or semireduced forms of N₂ or their analogues: N₂; NH=NH; NH=N–CH₃; H₂N–NH₂.^17,153,154 These included a proposed early (e) stage of the reduction of N₂, e(N₂), obtained from wild-type (WT) MoFe protein with N₂ as substrate;^166,170 two putative “midstage” intermediates, m(NH=N–CH₃), obtained from α-195^Gln MoFe protein with CH₃–N=NH as substrate^170,171 and m(NH=NH), obtained from the doubly substituted, α-70^Ala/α-195^Gln MoFe protein during turnover with in-situ-generated NH=NH;¹⁴⁷ and a “late” stage, l(N₂H₄), from the α-70^Ala/α-195^Gln MoFe protein during turnover with H₂N-NH₂^160,170 as substrate. Both hydrazine and diazene are substrates of wild-type nitrogenase that, like N₂, are reduced to ammonia.^{17,147,160,211,216}

5.1. Intermediate I

A combination of X/Q-band EPR and ¹⁵N,^1,2H ENDOR measurements on the intermediates formed with the three semireduced substrates during turnover of the α-70^Val→Ala/α-195^His→Gln MoFe protein subsequently showed that in fact they all correspond to a common intermediate (here denoted I) in which FeMo-co binds a substrate-derived [N_xH_y] moiety (Figure 9).^{154−156,207} Thus, both the diazenes and hydrazine enter and “flow through” the normal N₂-reduction pathway (Figure 8), and the diazene reduction must have “caught up” with the “later” hydrazine reaction.

Figure 9.

Comparison of 35 GHz ReMims pulsed ¹⁵N ENDOR spectra of intermediates trapped during turnover of the α-70^Ala/α-195^Gln MoFe protein with ¹⁵N₂H₄, ¹⁵N₂H₂, and ¹⁵NH=N—CH₃ (denoted ¹⁵MD). Adapted with permission from ref (207). Copyright 2011 American Chemical Society.

^1,2H and ¹⁵N 35 GHz CW and pulsed ENDOR measurements next showed that I exists in two conformers, each with metal ion(s) in FeMo-co having bound a single nitrogen from a substrate-derived [N_xH_y] fragment.^154,155 Subsequent high-resolution 35 GHz pulsed ENDOR spectra and X-band HYSCORE measurements showed no response from a second nitrogen atom, and when I was trapped during turnover with the selectively labeled CH₃—¹⁵N=NH, ¹³CH₃—N=NH, or C²H₃—N=NH, no signal was seen from the isotopic labels.²⁰⁷ From these results we concluded the N–N bond had been cleaved in forming I, which thus represents a late stage of nitrogen fixation, after the first ammonia molecule already has been released and only a [NH_x] (x = 2 or 3) fragment of substrate is bound to FeMo-co.²⁰⁷

5.2. Nitrogenase Reaction Pathway: D versus A

Given that states that could correspond to I are reached by both A and D pathways (Figure 8), the identity of this [NH_x] moiety need not in itself distinguish between pathways. However, the spectroscopic findings about I, in conjunction with a variety of additional considerations, led us to propose that nitrogenase functions via the A reaction pathway of Figure 8 for reduction of N₂.²⁰⁷ As one example, to explain how nitrogenase could reduce each of the substrates, N₂, N₂H₂, and N₂H₄, to two NH₃ molecules via a common A reaction pathway, one need only postulate that each substrate “joins” the pathway at the appropriate stage of reduction, binding to FeMo-co that has been “activated” by accumulation of a sufficient number of electrons (possibly with FeMo-co reorganization), and then proceeds along that pathway. Energetic considerations,¹³⁹ in combination with the strong influence of α-70^Val substitutions of MoFe protein without modification of FeMo-co reactivity, then implicate Fe, rather than Mo, as the site of binding and reactivity.^146,154,217

5.3. Intermediate H

When nitrogenase is freeze–quenched during turnover, the EPR signals from trapped intermediates in odd-electron FeMo-co states (Kramers states; S = ¹/₂, ³/₂,...; E_n, n = even),^154,155 plus the signals from residual resting-state FeMo-co, never quantitate to the total FeMo-co present, indicating that EPR-silent states of FeMo-co must also exist. These silent MoFe protein states must contain FeMo-co with an even number of electrons, and thus correspond to E_n, n = odd (n = 2m +1, m = 0–3) intermediates in the LT scheme. As noted above, such states may contain diamagnetic FeMo-co, or FeMo-co in integer-spin (S = 1, 2, ...), “non-Kramers (NK)” states,^179,180,218 but no EPR signal from an integer-spin form of FeMo-co had been detected until careful examination of samples that contain intermediate I(154−156) revealed an additional broad EPR signal at low field in Q-band spectra that arises from an integer-spin system with a ground-state non-Kramers doublet with spin S ≥ 2 (Figure 10).²¹⁹

Figure 10.

2K Q-band CW EPR spectrum of α-70^Val→Ala, α-195^His→Gln MoFe protein in resting-state (S = ³/₂) and trapped during turnover with ¹⁴N₂H₄. Kramers intermediate I and non-Kramers intermediate, H, are noted in the turnover spectrum. Adapted with permission from ref (219). Copyright 2012 National Academy of Sciences.

Earlier work showed how to characterize a non-Kramers doublet with ESEEM spectroscopy (NK-ESEEM),^173,174 so NK-ESEEM time-waves were collected for the NK intermediates trapped during turnover with: ¹⁴N and ¹⁵N isotopologs of N₂H₂ and N₂H₄ substrates; ⁹⁵Mo-enriched α-70^Val→Ala/α-195^His→Gln MoFe protein; H—¹⁴N=¹⁴N—CH₃, H—¹⁵N=¹⁴N—CH₃, and H—¹⁴N=¹⁴N—CD₃. Figure 11 presents representative 35 GHz (2 K) three-pulse NK-ESEEM time-waves collected at several relatively low fields from the nitrogenase NK intermediates generated with isotopologs of the three substrates. The NK-ESEEM time-waves for the intermediates trapped during turnover with the corresponding ¹⁴N and ¹⁵N isotopologues of N₂H₂, N₂H₄, and HN₂CH₃ substrates are identical at all fields, indicating that they are associated with a common intermediate, denoted H, trapped during turnover with all three substrates. ⁹⁵Mo enrichment of α-70^Val→Ala, α-195^His→Gln MoFe protein produces a significant change of the NK-ESEEM time-wave. This analysis established that the NK-EPR signal of H arises from the Mo-containing FeMo-co in an integer-spin-state with S ≥ 2, and not the all-iron electron-transfer P cluster, also present in the MoFe protein, or even the [4Fe–4S] cluster of the Fe protein.²¹⁹

Figure 11.

Three-pulse ESEEM traces after decay-baseline subtraction for NK intermediate H of α-70^Val→Ala, α-195^His→Gln MoFe protein trapped during turnover with ¹⁴NH=¹⁴NH, ¹⁴NH=¹⁴NCD₃, ¹⁴NH₂—¹⁴NH₂, ¹⁵NH=¹⁵NH, ¹⁵NH=¹⁴NCH₃. Adapted with permission from ref (219). Copyright 2012 National Academy of Sciences.

Comparison of the ¹⁴N/¹⁵N NK-ESEEM of H in Figure 11 indicates that a nitrogenous ligand derived from substrate is directly bound to FeMo-co of α-70^Val→Ala/α-195^His→Gln MoFe protein. Modulation is absent from the second ¹⁴N that would be present if the N–N bond of substrate remained intact, as shown by comparison of the time-waves for the H prepared with H—¹⁵N=¹⁵N—H versus H—¹⁵N=¹⁴N—CH₃, as is modulation from ²H of H–¹⁴N=¹⁴N-CD₃. This indicates that H contains an NH_x fragment that remains bound to FeMo-co after cleavage of the N–N bond and loss of NH₃. Quadrupole coupling parameters for the NH_x fragment indicated it is not NH₃, and that H has bound [−NH₂].²¹⁹

6. Unification of the Nitrogenase Reaction Pathway with the LT Kinetic Scheme

The H and I intermediates provide “anchor-points” that allow assignment of the complete set of E_n intermediates that follow E₄, 5 ≤ n ≤ 8. As illustrated in Figure 12, the loss of two reducing equivalents and two protons as H₂ (eq 1) upon N₂ binding to the FeMo-co of E₄ leaves FeMo-co activated by two reducing equivalents and two protons. We argued that when N₂ binds to FeMo-co it is “nailed down” by prompt hydrogenation, Figure 12, with N₂ binding, H₂ loss, and reduction to the diazene level, all occurring at the E₄ kinetic stage of the LT scheme.²¹⁹ The identification of H with its E_n stage is achieved as follows. (i) As the same intermediate H is formed during turnover with the two diazenes and with hydrazine, the diazenes must have catalytically “caught up” to hydrazine, and H must occur at or after the appearance of a hydrazine-bound intermediate. (ii) As noted above, H contains FeMo-co in an integer-spin (NK) state, and thus corresponds to an E_n state with n = odd. As H is a common intermediate that contains a bound fragment of substrate, it must, therefore, correspond to E₅ or E₇, and analysis of the pathway alternatives in the light of the EPR/ESEEM measurements indicated that H corresponds to the [NH₂]⁻-bound intermediate formed subsequent to N–N bond cleavage and NH₃ release at the E₇ stage of the P–A pathway.

Figure 12.

Integration of LT kinetic scheme with “prompt” (P) alternating (A) pathway for N₂ reduction. The ? represents the product of N₂ binding with H₂ release, whose identity is discussed below. Also shown is how diazene and hydrazine join the N₂ reduction pathway. Note: M denotes FeMo-co in its entirety, and substrate-derived species are drawn to indicate stoichiometry only, not mode of substrate binding. E_n states, n = even, are Kramers states; n = odd are non-Kramers. M^N denotes resting-state FeMo-co. Individual charges on M and a substrate fragment, not shown, sum to the charge on resting FeMo-co. Adapted with permission from ref (156) with corrections based on the re mechanism for H₂ loss upon N₂ binding discussed below. Copyright 2013 American Chemical Society.

By parallel arguments, the only possible assignment for the S = ¹/₂ state I, which we showed earlier to occur after N–N bond cleavage,²⁰⁷ is as E₈: I must correspond to the final state in the catalytic process (Figure 12), in which the NH₃ product is bound to FeMo-co at its resting redox state, prior to release and regeneration of the resting-state form of the cofactor. The trapping of a product-bound intermediate I is analogous to the trapping of a bio-organometallic intermediate during turnover of the α-70^Val→Ala MoFe protein with the alkyne, propargyl alcohol; this intermediate was shown to be the allyl alcohol alkene product of reduction.¹⁶⁸ With assignments of E₄, E₇, and E₈, then filling in the LT “boxes” for E₅, E₆ of Figure 3 is straightforward, thus unifying the reaction pathway for N₂ reduction with the LT scheme.

Figure 12 adopts a “prompt” (P)–alternating pathway for the stages following N₂ binding and H₂ loss, which offers explanations for how the hydrogenated reaction intermediates, diazene and hydrazine, join the N₂ reduction pathway. Key to this issue was the finding that H₂ inhibits the reduction of diazene,¹⁴⁷ but not hydrazine.²¹¹ We took the simplest view, that under turnover, diazene and hydrazine each joins the N₂ reduction pathway at its own characteristic entry point, and each then proceeds to generate both H and I. As shown in Figure 12, diazene binds to E₂ with the release of H₂, thereby entering the N₂ pathway as the “final” interconverting form of the E₄ state. N₂H₄ instead binds to E₁ (as proposed for another two-electron substrate, C₂H₂^{17,33,103,220}), joining the N₂ pathway with the release of NH₃ to form a stage corresponding to E₇ in the N₂ reduction scheme.¹⁵⁶

7. Obligatory Evolution of H₂ in Nitrogen Fixation: Reductive Elimination of H₂

The E_n assignments of Figure 6 plus those of Figure 12 give proposed structures to all E_n states of the LT kinetic scheme (Figure 3), but the assignments have been developed through independent analyses of the two four-electron halves of the eight-electron catalytic cycle (eq 1). In the first half (part I) of the pathway, accumulation of four electrons/protons activates FeMo-co, generating E₄; in the second half (part II), bound N₂ is hydrogenated by two of those electrons/protons plus an additional four electrons/protons. However, the assignments are silent about the mechanism by which the E₄ Janus intermediate, Figure 4, connects these two halves: the obligatory production of an H₂ molecule upon N₂ binding, as shown in Figure 12.^33,156,158 Why nitrogenase should “waste” fully 25% of the ATP required for nitrogen fixation through H₂ generation (eq 1) has remained a mystery, and indeed is not even accepted uniformly.^23,221

Consideration of the finding that E₄ stores its four reducing equivalents as two bridging hydrides (Figure 4) within the context of the well-known organometallic chemistries of hydrides^194,222 and dihydrogen²²³ led us to examine the two alternative mechanisms by which this state might bind and activate N₂ with release of H₂, and proceed to the prompt formation of FeMo-co with a bound diazene-level species (N₂H₂) without additional accumulation of [e^–/H⁺], as featured in the P–A reaction pathway, Figure 12. In one, H₂ is formed by hydride protonation (hp mechanism), Figure 13, upper; the other forms H₂ through reductive elimination (re),¹⁹⁵ Figure 13, lower. We first describe these two mechanisms, and then show that the re mechanism is operative.

Figure 13.

Visualization of hp and re mechanisms for H₂ release upon N₂ (blue) binding to E₄. The following is shown: the Fe-2,3,6,7 face of resting FeMo-co; the structure of FeMo-co must distort in different stages of catalysis. The Fe that binds N₂ is presumed to be Fe6, as indicated by studies of α-70^Val variants; when bold, red, Fe6 is formally reduced by two equivalents (see text). The bridging hydrides of E₄ (green) are positioned to share an Fe “vertex”, as suggested by re mechanism of H₂ release upon N₂ binding. Alternative binding modes for N₂-derived species can be envisaged.

7.1. Hydride Protonation (hp) Mechanism

In the hp scheme (Figure 13, upper), N₂ binding is accompanied by the activation of one bridging hydride to the terminal form and protonation of this hydride by a sulfide-bound proton to form and release H₂. Such a mechanism for H₂ formation is invoked in discussions of hydrogenases,^223,224 and there is strong precedence for replacement of a metal-bound H₂ with N₂. In this context, by analogy to the mechanism for the (much less demanding) reduction of alkynes/alkenes one might propose that transient terminalization of the “second” hydride would then lead to hydrogenation and protonation of the bound N₂, to form FeMo-co bound N₂H₂ (see Figure 13, upper, below). For reasons that will become clear below, the hydrogenation of N₂ to form metal bound-N₂H₂ must be reversible.

7.2. Reductive Elimination (re) Mechanism

The second mechanism for H₂ loss upon N₂ binding begins with transient terminalization of both E₄ hydrides, Figure 13, lower. This is followed by reductive elimination of H₂ as N₂ binds, steps with considerable precedence.^{4,194,222,223,225} Of key importance, the departing H₂ carries away only two of the four reducing equivalents stored in E₄, while the Fe that binds N₂ becomes highly activated through formal reduction by two equivalents; for example a formal redox state of Fe(II) would be reduced to Fe(0). This delivery of two reducing equivalents to the FeMo-co core which otherwise is reduced by at most one equivalent during electron/proton activation (Figures 6 and 7) would poise the cofactor to deliver the two activating electrons to N₂, whose π acidity could be further enhanced by electrostatic interactions with the two sulfur-bound protons: combined delivery of the two electrons and protons would directly yield cofactor-bound N₂H₂. This amounts to a “push–pull” mechanism for the hydrogenation of N₂, in which the “push” of electrons from the doubly reduced cofactor onto N₂ is enhanced by the electrostatic “pull” of the protons bound to sulfur. As discussed in section 9, below, models of E₄(N₂) constructed by placing N₂ and two protons on the doubly reduced FeMo-co core modeled with its resting-state structure provide a convincing illustration of this mechanism (and other insights, as well). The diminished electron donation to Fe by protonated sulfides would not only facilitate reductive elimination, but also would act to localize the added electrons on the Fe involved, limiting charge delocalization over the rest of the cofactor. This mechanism provides a compelling rationale for obligatory H₂ formation during N₂ reduction: the transient formation of a state in which an electrostatically activated N₂ is bound to a highly activated, doubly reduced site, thereby generating a state optimally activated to carry out the initial hydrogenations of N₂, the most difficult process in N₂ fixation.

7.3. Mechanistic Constraints Reveal That Nitrogenase Follows the re Mechanisms

A clear choice between hp and re mechanisms is achieved by testing them against the numerous constraints that are associated with the reaction of D₂ with the diazene-level E₄(N₂/N₂H₂) state formed when N₂ binds to the cofactor and is reduced. The three principal constraints are listed in Chart 1. The first test that they provide for a mechanism is that it must accommodate the finding that when nitrogenase turns over in the presence of both N₂ and D₂, then two HD are formed through D₂ cleavage and solvent-proton reduction, with the stoichiometry summarized as constraint i of Chart 1.^17,226−228 Such HD formation only occurs in the presence of N₂, and not during reduction of H⁺ or any other substrate.^226,229,230

Chart 1.

The second key constraint and mechanistic test was revealed by Burgess and co-workers 30 years ago; the absence of exchange into solvent of D⁺/T⁺ derived from D₂/T₂ gas, Chart 1, constraint ii.²²⁶ When nitrogenase turns over under a mixture of N₂ and T₂, HT is formed with stoichiometry corresponding to Chart 1, constraint i, but during this process only a negligible amount of T⁺ is released to solvent (∼2%). The third constraint is provided by a later study of α-195^His- and α-191^Gln-substituted MoFe proteins.¹⁶¹ It provided persuasive evidence that HD formation under N₂/D₂ requires that the enzyme be at least at the E₄ redox level, with a FeMo-co-bound N–N species at the reduction level of N₂H₂ or beyond, corresponding to the third constraint, Chart 1, iii.¹⁶¹ Constraint iii, plus the stoichiometry of HD formation according to constraint i implies a process described as

3

Thus, N₂H₂ formation is reversible, as shown in Figure 13.

Figure 14, upper, shows that the characteristics of HD formation during turnover under N₂/D₂cannot be reconciled with the hp mechanism. In the reverse of this mechanism, D₂ binding and N₂ release would generate an E₄ state that has one deuteride bridge, which is deactivated for exchange with solvent. However, it carries the other deuteron in the form of D⁺ bound to sulfur (or a protein residue), which likely would be solvent exchangeable. Exchange of that D⁺ would violate the stoichiometric constraint, of eq 3 (line i, Chart 1), as relaxation of E₄ to E₂ within the reverse-hp mechanism would generate only one HD per D₂, not two as required. Correspondingly, replacement of D₂ by T₂ in Figure 14, upper, with exchange of T⁺ bound to sulfide would lose roughly one T⁺ per T₂ to solvent, contrary to the few percent loss observed by Burgess et al. (constraint ii).²²⁶ The possibility that the proton-bearing site is “shielded” from exchange seems implausible for a catalytic cluster that depends on proton delivery for its catalytic function, and in any case solvent exchange need not be fast; the rate-limiting step in nitrogenase turnover is the off-rate for Fe protein after it has delivered its electron to MoFe,^33,99 and this process is quite slow, with a rate constant of ∼6 s^–1.

Figure 14.

Reversal of hp and re mechanisms upon D₂ binding. Details as in Figure 13. Bold arrows replace equilibrium arrows to emphasize the relaxation process.

If this proton were nonetheless shielded from exchange, relaxation to E₂ would occur with regeneration of D₂, without the generation of HD, in disagreement with the stoichiometric constraint of Chart 1. This objection would be overcome if at ambient temperatures the hydrides/protons can “migrate” over the FeMo-co face, but this instead would require multiple sites to be “shielded” for slow exchange, while FeMo-co is accessible to rapid proton delivery. Overall, we conclude that the hp process fails to satisfy the constraints of Chart 1, as the reverse hp process satisfies neither the stoichiometry of eq 3 nor the constraint that T⁺ is not released to solvent (Chart 1).

In contrast, in the reverse of the re mechanism, shown in Figure 14, lower, D₂ binding and N₂ release would generate E₄(2D), the E₄ isotopomer in which both atoms of D₂ exist as deuteride bridges. This state would relax with loss of HD to E₂(D), and then to E₀ with loss of the second HD, thus satisfying the stoichiometry of eq 3. If the reaction were carried out under T₂, essentially no T⁺ would be lost to solvent because the bridging deuterides are deactivated for exchange with the protein environment and solvent, thus satisfying the “T⁺ exchange” constraint, Chart 1.

One alternative fate of the E₄(2D) formed by D₂ replacement of N₂ would be to rebind an N₂, but this would merely release the D₂ that had started the reverse process, creating a cycle invisible to detection. As a second alternative, E₂(D) could acquire two additional electrons/protons to achieve the monodeutero E₄ state. However, as shown in Figure 14, lower, if this state then bound N₂ it would release the second HD, again without solvent exchange, whereas if it ultimately relaxed to E₀ it would release the second HD along with an H₂. Thus, the re mechanism for N₂ binding and H₂ release not only has the compelling chemical rationale discussed above, but also satisfies the three critical HD constraints for the various alternatives that arise when it is run in reverse, Chart 1.

In short, the re mechanism, Figure 13, lower, satisfies the constraints summarized in Chart 1, as visualized in Figure 14, lower; to the best of our knowledge, it likewise satisfies all other constraints on the mechanism provided by earlier studies, most of which are not directly tied to D₂ binding.

8. Test of the re Mechanism

Subsequent to formulation of the re mechanism for the activation of FeMo-cofactor to reduce N₂ (Figure 13, lower),¹⁵⁶ we noted that addition of C₂H₂ to a N₂/D₂ reaction mixture should offer a rigorous test of the mechanism. The test is founded on a defining characteristic of nitrogenase catalysis, an exact distinction between hydrons (H/D/T) associated with the gaseous diatomics, H₂/D₂/T₂, and those derived from solvent water. Thus, when nitrogenase in protic buffer is turned over under N₂/D₂, gaseous D₂ can displace N₂ from the E₄(N₂/N₂H₂) state (Figure 14, lower), stoichiometrically yielding two HD.^226−228 This and other observations clearly show that diatomic H₂/D₂ is not used to reduce N₂ during turnover under N₂//H₂/D₂ (in particular, T incorporated into the ammonia product of N₂ fixation would exchange with solvent).¹⁷ Likewise, as demonstrated below, when C₂H₂ is reduced in the presence of D₂, no deuterated ethylenes are generated.

8.1. Predictions

With this foundation, we recognized that the re mechanism predicts that turnover under C₂H₂/D₂/N₂ should not only incorporate H from solvent to generate C₂H₄ by the normal reduction process, but through the agency of the added N₂ also should breach the separation of gaseous D₂ from solvent protons by generating both C₂H₃D and C₂H₂D₂ (Figure 15). According to the re mechanism, when turnover is carried out under N₂/D₂, D₂ can react with E₄(N₂/N₂H₂), replacing the N₂ and undergoing oxidative addition to generate E₄(2D). We recognized that this state in fact might be expected to react with C₂H₂ to form C₂H₂D₂ through the idealized mechanism (Figure 16A) involving terminalization of an [Fe–D–Fe] bridge of E₄, and migratory insertion of bound C₂H₂ into the Fe–D bond to form an Fe-alkenyl intermediate, followed by reductive elimination of C₂H₂D₂.^195,231 Previous studies^17,103 could not distinguish reaction at the E₄ state from reaction at the E₂ state when C₂H₂ is reduced in the absence of N₂, as N₂ is required to enable gaseous D₂ to enter the nitrogenase catalytic process. The possibility that acetylene can access different nitrogenase redox states, however, had been suggested on the basis of experiments using a nitrogenase variant that exhibits N₂ reduction that is resistant to inhibition by acetylene.^232,233

Figure 15.

Formation of deuterated acetylenes during turnover under N₂/D₂/C₂H₂ as predicted according to re mechanism. Cartoons again depict the Fe2,3,6,7 face of resting-state FeMo-co, with no attempt to incorporate likely structural modifications. Figure shows that the “reverse” of re mechanism through displacement of N₂ by D₂ produces, successively, E₄(2D) and E₂(D), further showing potential reaction channels for capture of E₄(2D) and E₂(D) intermediates with C₂H₂.

Figure 16.

Schematic mechanism for reaction of C₂H₂ with E₄(2D) and E₂(D). (A) Formation of C₂H₂D₂, which follows Scheme 15.20 of Hartwig:²³¹ mi = migratory insertion; re = reductive elimination. In braces: Possible alternative reaction channel that leads to formation of C₂H₃D, ap = alkenyl protonolysis. (B) Schematic mechanism for formation of C₂H₃D from reaction of C₂H₂ with E₂(D). (C) Illustration of possibility that C₂H₂ displaces D₂ formed by reductive elimination of the E₄(2D) deuterides, leading to direct formation of C₂H₄ without D incorporation.

The E₄(2D) state also would relax through the loss of HD to form E₂(D), an E₂ state whose unique isotopic composition can be generated in no other way. Interception of the E₂(D) state by C₂H₂ would then generate C₂H₃D, with Figure 16B presenting a plausible mechanism: deuteride terminalization and insertion, followed by alkenyl protonolysis.^195,231 This reaction also might occur through an alternative reaction channel of E₄(2D), as noted in Figure 15.

8.2. Testing the Predictions

We tested the predictions based on the re mechanism of an unprecedented involvement of gaseous D₂ in substrate reduction by use of C₂H₂ reduction under N₂/D₂/C₂H₂ gas mixtures to intercept the E₄(2D) and E₂(D) states. As expected, the control reaction of turnover under D₂/C₂H₂ generates only C₂H₄, without incorporation of D from gaseous D₂ to generate either C₂H₃D or C₂H₂D₂ (Figure 17). In dramatic contrast, C₂H₂ reduction by nitrogenase under a N₂/D₂/C₂H₂ gas mixture in fact produces readily measured amounts of C₂H₂D₂ and even greater amounts of C₂H₃D (Figure 17).¹⁵⁷

Figure 17.

Time-dependent formation of ¹³C₂H₃D and ¹³C₂H₂D₂, catalyzed by nitrogenase reduction of ¹³C₂H₂. ¹³C₂H₃D determined by GC/MS monitoring of m/z = 31 for a reaction mixture containing ¹³C₂H₂ and including D₂ and N₂ (■), just D₂ (x inside □), or H₂ and N₂ (□). Inset: ¹³C₂H₂D₂, m/z = 32, formation starting with ¹³C₂H₂/D₂/N₂ (●), just D₂ (◊), or H₂/N₂ (○). Partial pressures of 0.02 atm ¹³C₂H₂, 0.25 atm N₂, and 0.7 atm H₂/D₂, where present. The molar ratio of Fe protein to MoFe protein was 2:1. All assays incubated at 30 °C. Adapted with permission from ref (157). Copyright 2013 National Academy of Sciences.

On reflection, the success of this test for the formation of E₄(2D) is a consequence of the greater reactivity of C₂H₂ compared to that of N₂ and/or of the difference in the likely ways that these two substrates bind to FeMo-co: side-on for C₂H₂, end-on for N₂. Otherwise, in a process analogous to that for N₂H₂ formation in the re mechanism (Figure 13, lower), C₂H₂ might in principle displace D₂ formed by reductive elimination of the E₄(2D) deuterides, leading to direct formation of C₂H₄ without D incorporation, Figure 16C. The yield of C₂H₂D₂ may be less than that of C₂H₃D, because the contribution from this reaction channel diminishes the yield of the former, but it is perhaps more likely that the binding and reduction of C₂H₂ by E₄(2D) is substantially less likely than the relaxation of E₄(2D) to E₂(D) through loss of HD, and the reduction of C₂H₂ by E₂(D) (Figure 15).

These observations are enriched by consideration of the dependences of the yields of C₂H₃D and C₂H₂D₂ on the partial pressures of C₂H₂, D₂, N₂, and electron flux, all of which are understandable in terms of the production of the E₄(2D) and E₂(D) states under these turnover conditions, as predicted by the re mechanism for FeMo-cofactor activation for N₂ binding and reduction.¹⁵⁷ For example, reduction of acetylene and N₂ are mutually exclusive, with complicated inhibition kinetics between these two substrates.^217,234 Therefore, it was of interest to determine the effect of varying the N₂ partial pressure on the formation of C₂H₃D and C₂H₂D₂ at fixed C₂H₂ and D₂ pressures. The yields of C₂H₃D and C₂H₂D₂ increase in parallel with increasing partial pressure of N₂ (Figure 18). This can be explained by enhanced formation of E₄[N₂/N₂H₂] by reaction of N₂ with E₄. Increased formation of E₄[N₂/N₂H₂] in turn would enhance reaction with D₂ to form E₄(2D), which can be intercepted by acetylene to form deuterated ethylenes (Figure 15).¹⁵⁷

Figure 18.

Deuterated ethylene formation as a function of N₂ partial pressure. The partial pressure of C₂H₂ was 0.02 atm and D₂ was 0.6 atm. The molar ratio of Fe protein to MoFe protein was 4:1. Assay conditions as in Figure 17. Adapted with permission from ref (157). Copyright 2013 National Academy of Sciences.

It is of interest to note that the reduction of C₂H₂ to C₂H₃D by reaction with E₂(D) formally corresponds to the reduction of C₂H₂ by the HD that otherwise would form during relaxation of E₂(D) to E₀, a perspective that highlights the contrast between this result, achieved in the presence of N₂, with the failure of nitrogenase to use H₂/D₂ to reduce any substrate in the absence of N₂. As an eleboration on this perspective, the formation of HD during turnover under N₂/D₂, with stoichiometry (eq 3, above),¹⁷ can be seen to correspond to the nitrogenase-catalyzed reduction of protons by D₂ and electrons with N₂ as cocatalyst, eq 3′

3′

as the reaction neither proceeds without N₂ nor consumes N₂. Likewise, although C₂H₂D₂ is well-known to form during nitrogenase reduction of C₂D₂ in H₂O buffer (or C₂H₂ in D₂O buffer),¹⁴⁸ formation of this species during turnover under C₂H₂/D₂/N₂ corresponds to the previously unobserved reduction of C₂H₂ by gaseous D₂ with N₂ as cocatalyst (eq 4).

4

Correspondingly, the formation of C₂H₃D involves incorporation of D^– derived from D₂ along with H⁺ from solvent with N₂ as cocatalyst.

9. Completing the Mechanism of Nitrogen Fixation

Figure 12, above, presents a formal integration of the reaction pathway for nitrogen fixation (intermediates E₄–E₈) with the LT kinetic scheme, the key to the resulting mechanism being N₂ binding and H₂ release through the re mechanism, Figure 13, lower. This mechanism is built on the structure of the E₄ intermediate and its implication that hydride chemistry is central to nitrogen fixation by nitrogenase (section 3). As a corresponding implication, we further offered the two alternative sets of proposed structures for the “early”, E₁–E₃, intermediates (Figure 6). We now discuss in greater depth the E₅–E₈ intermediates of nitrogen fixation, proposing in Figure 19, II not only more detailed structures for the stages following the formation of N₂H₂-bound FeMo-co, the E₄(N₂H₂) state, written as binding diazene itself, but also the nature of the chemical transformations that link these stages during the delivery of the ‘second half’ of the eight [e^–/H⁺] that comprise the stoichiometry of nitrogen fixation, eq 1. The analysis further leads us to provisionally assign the early, “first half” intermediates to the alternative described in Figure 6B, now visualized in Figure 19, I. When combined with the reductive elimination (re) mechanism for the binding N₂ and release of H₂, Figure 13, lower, the result, Figure 19, is a self-consistent proposal for the structures of all intermediates in the nitrogen fixation mechanism and a formal description of the transformations that convert each stage to the subsequent one: a complete, though of course still simplified, mechanism for nitrogen fixation by nitrogenase.

Figure 19.

Proposed mechanism displaying structures of all intermediates in nitrogen fixation, inspired by the assumption of primacy of hydride chemistry associated with the Fe2,3,6,7 face of FeMo-co, and containing a formal description of the transformations that convert each stage to the subsequent one. In I the mechanism tentatively adopts and visualizes the view of E_n states n = 1–4 presented in Figure 6B; in II it visualizes bridging hydrides by analogy, without evidence for or against terminal hydrides for n = 5–7. Likewise, the structure of the N₂H₂ species as end-on bound diazene is suggestive, not definitive, etc. I and II are connected by the re mechanism, Figure 13, lower. Formal charges are included as useful to help guide the reader.

Figure 19, II, is constructed on two assumptions that (i) the formation and reactions of hydrides is key; (ii) beginning with N₂H₂, the hydrogenation of reduced forms of N₂ involves migratory insertion into Fe–H bonds. These assumptions lead to the conclusion that [e^–/H⁺] transfer to FeMo-co of the E₄(N₂H₂) and E₆ states creates E₅ and E₇ that each contain an [Fe–H–Fe] bridging hydride moiety bound to an oxidized FeMo-co, Figure 19, II, in correspondence with the analogous [e^–/H⁺] transfer to E₀ and E₂, shown in the cartoon of Figure 6B, and now visualized in Figure 19, I. In the case of E₅, an accompanying migratory insertion of the N₂H₂ into an Fe–H bond (presumably formed by terminalization of the bridge) forms the [N₂H₃]⁻ moiety bound to the oxidized cluster; in the case of E₇, migratory insertion leads to N–N bond cleavage and formation of [NH₂]⁻ bound to the formally oxidized cofactor (Figure 19, II). The follow-up [e^–/H⁺] transfer to E₅, E₇ generates the E₆ and E₈ states, respectively. This mechanistic picture is anchored by the final stages, E₇ and E₈, whose structures match those proposed in the EPR/ENDOR/ESEEM studies of intermediates H, assigned to E₇, and I, assigned to E₈, Figure 12.

The proposal completes a mechanism in which the stoichiometrically required delivery of all 8 [e^–/H⁺] to FeMo-co is controlled by the hydride chemistry of the cofactor. The clearly understandable differences between the first “half” of the catalytic cycle, visualized in Figure 19, I, and the “second half”, Figure 19, II, arise because the former involves accumulation of reducing equivalents while the latter involves delivery of reducing equivalents to substrate.

The two halves are similar in that addition of [e^–/H⁺] to form an n = odd intermediate (n = 1, 3, first half; n = 5, 7, second half) generates an [Fe–H–Fe] bridging hydride attached to a formally oxidized FeMo-co core. They differ in that the hydride is “stored” in n = 1, 3, but is promptly transferred to substrate in n = 5, 7 to form a (formally) anionic reduced substrate. Upon addition of [e^–/H⁺] to any one of these four n = odd intermediates, to form the subsequent n = even intermediate, the electron formally reduces the core to the resting-state redox level. In the first half (n = 2, 4), the H⁺ is delivered to a sulfur and its charge balances that on a hydride; in the second half (n = 6, 8), the proton neutralizes the anionic nitrogenous ligand, to form the neutral, N₂H₄ of E₆, NH₃ of E₈.

The two halves of the nitrogen fixation mechanism are joined at the E₄ stage, as described above and displayed as Figure 19, re: the E₄(2H) intermediate formed by accumulation of four [e^–/H⁺] and containing two bridging hydrides undergoes reductive elimination as it binds N₂ and releases the two “sacrificial” reducing equivalents as H₂. Figure 19 thus represents a complete mechanism for nitrogen fixation by nitrogenase that invokes the primacy of the hydride chemistry of FeMo-co.

9.1. Uniqueness of N₂ and Nitrogenase

The mechanistic proposal of Figure 19 invokes the primacy of hydride chemistry associated with a 4Fe face of FeMo-co, a structural feature made possible only with a cluster of at least six metal ions. The hydrogenations of reduced forms of N₂, starting with N₂H₂, involve migratory insertion of substrate into Fe–H bonds, one at a time. This is the same mechanism visualized for the “normal” reduction of C₂H₂ at the E₂ stage, and even for the rare trapping of E₄ by C₂H₂, Figure 16; we suggest migratory insertions are likely to be involved in the hydrogenation of all other substrates.

But N₂ is not reactive to hydride insertion. So nitrogenase adopts a different “strategy” for attacking its physiological substrate. It is forced to accumulate four reducing equivalents as two Fe hydrides, which requires a 4-Fe face, and thus the large cluster is “held together” by the carbide at its core. We have concluded that this cluster can only become activated for N₂ hydrogenation through reductive elimination of two of those equivalents in the form of H₂.^156,157 The “push” of the doubly reduced metal-ion core of the cluster, compounded by the electrostatic “pull” of sulfur-bound protons, is required to overcome the high barrier to the initial hydrogenation of N₂, directly to N₂H₂, Figure 19.

9.2. Structure of the E₄(N₂) Intermediate: Some Implications

As an exercise to illustrate four points worth noting, we have modeled alternative structures of the E₄(N₂) intermediate by building the bound substrate onto the crystal structure of resting-state FeMo-co using structural information from model complexes.^235−240 It seems most likely, on the basis of the structures of model complexes, that N₂ binds end-on, rather than bridging. As illustrated in Figure 20, and emphasized over the years,^137,241,242 end-on bound N₂ can bind to FeMo-co in two basic, alternative modes: endo, with the N₂ “nestled” in the pocket above the Fe2,3,6,7 face; exo, with N₂ pointed away from that face. The first point is as follows. According to our mechanism, E₄(N₂) contains doubly reduced metal-ion core with two protons bound to sulfur. There are multiple potential dispositions of the H⁺ on different sulfurs, but distance measurements with the mockups show that the atoms of N₂ and protons can indeed be in close enough proximity to support the electrostatic “pull” postulated above.

Figure 20.

Models for the two alternative modes for N₂ binding at Fe6 of FeMo-cofactor in the E₄(N₂) state, with two protons bound to two adjacent sulfides as in Figure 4: (A) endo mode; (B) exo mode. The side chains of selected amino acid residues are shown as sticks. The figure was generated in Pymol by building N₂ onto the resting-state of FeMo-co using the coordinate file PDB:2AFK. Iron is shown in rust, molybdenum in magenta, sulfur in yellow, carbon in dark gray, hydrogen in light gray, nitrogen in blue, and oxygen in red.

Second, this mockup demonstrates the commonly understood need for the FeMo-co core to “relax” upon substrate binding. In the resting-state the Fe ions are roughly tetrahedral, and without such relaxation, the N₂–S distances would be far too short. The normal assumption would be that Fe6 roughly forms a plane with three S atoms, with a major contribution to the relaxation being an elongation of the bond trans to N₂.

The third issue is the resulting structural/electronic-structure consequences of the identity of the trans ligand in exo versus endo N₂ binding, and it does not appear to have been widely discussed. The modulation of metal-ion reactivity by variations in the trans ligand (the “trans effect”) is well-known,²³¹ and recently, a series of trigonal Fe complexes that are biomimetic of nitrogenase have shown that the trans ligands to a terminal Fe–N₂ can regulate the ability of the complex to catalytically reduce N₂.²¹⁴ In the exo binding mode, the interstitial carbide is trans to N₂. This mode would favor the idea that carbide modulates the properties of Fe6 through the trans effect, and may well act as a hemilabile ligand. However, in the endo mode, which has been favored by some computations, the trans ligand is now a S that bridges to Mo. As C is (roughly) an “in-plane” ligand, not trans, its influence on reactivity would be different than for endo binding, in which case modulation of Fe6 reactivity by the trans effect would involve [S–Mo] being “axial” ligand.

There is a corollary to considerations of the endo binding mode. It is widely assumed that the catalytic centers of the alternative nitrogenases have the same structure as FeMo-co, with the heterometal atom Mo being replaced by V or Fe.¹⁸ Thus, if N₂ does bind endo to Fe6 of a FeMo-co-like structure in all three systems, its reactivity would be modulated by differences in the axial −S–M “ligand” caused by differences in the properties of Mo, V, and Fe.

The fourth point is the possible importance of interactions of substrate with adjacent amino acid residues. In the endo binding mode the N₂ is nestled within a binding pocket capped by the side-chain of α-70^Val; in the exo mode, the N₂ is pointed to a pocket surrounded by α-191^Gln and the homocitrate ligand of Mo. The present mockups suggest that the protein environment in either binding mode could readily accommodate, and even stabilize, N₂ with no more than minor conformational rearrangements.

10. Summary of Mechanistic Insights

10.1. Catalytic Intermediates of N₂ Fixation

Two major points can be made regarding intermediates trapped: (1) Characterization of the E₄ “Janus” intermediate as bearing four reducing equivalents in the form of two [Fe–H–Fe] bridging hydrides has provided the foundation for proposals that the FeMo-co core is never oxidized or reduced by more than one equivalent relative to the resting-state, and that the oxidative couple in fact is operative, Figure 19, I. (2) The characterization of the common intermediates H and I, trapped during turnover with nitrogenous substrates, led to the proposed unification of kinetic scheme and A reaction pathway, Figure 12.

10.2. re Mechanism

Reductive elimination of two hydrides upon N₂ binding (re mechanism) provides an explanation for the nitrogenase stoichiometry (eq 1) and for the obligatory formation of H₂ upon N₂ binding. This mechanism for H₂ production upon N₂ binding to E₄, Figure 13, lower, satisfies both the stoichiometric constraint of HD formation (Chart 1, line i) and the “T⁺” constraint against exchange of gas-derived hydrons with solvent (Chart 1, line ii), whereas the hp mechanism (Figure 13, upper) satisfies neither. The re mechanism further involves D₂ binding to a state at the “diazene level” of reduction, as required by the constraint of eq 3 and Chart 1, line iii. Finally, to the best of our knowledge, all other constraints on the mechanism, most of which are not directly tied to D₂ binding, are satisfied, as well.

This mechanism answers the following long-standing and oft-repeated question: Why does nature “waste” four ATP/two reducing equivalents through an obligatory loss of H₂ when N₂ binds? The answer follows: reductive elimination of H₂ upon binding of N₂ to FeMo-co of the E₄ state generates a state in which highly reduced FeMo-co binds N₂, which likely is activated for reduction through electrostatic interactions with the remaining two sulfur-bound protons. Transfer of the two reducing equivalents generated by the reductive elimination, combined with transfer of the two activating protons, then forms N₂H₂, Figure 13, lower, in keeping with the P–A scheme of Figure 12. It appears that only through this activation is the enzyme able to hydrogenate N₂.

10.3. Turnover under N₂/D₂/C₂H₂ as a Test of the re Mechanism

This mechanism has been supported by a rigorous test which provided experiments in which C₂H₂ is added to an N₂/D₂ reaction mixture. Although diatomic D₂ does not reduce nitrogenase C₂H₂ in the absence of N₂, the re mechanism successfully predicted that turnover under C₂H₂/D₂/N₂ would breach the separation of gaseous D₂ from solvent protons by generating both C₂H₃D and C₂H₂D₂.

The conclusions regarding H₂ formation upon N₂ binding reached from this study are as follows. (i) The unprecedented incorporation of D from D₂ into the nitrogenase reduction products C₂H₂D₂ and C₂H₃D during turnover under C₂H₂/D₂/N₂ in H₂O demonstrates the presence of the E₄(2D) and E₂(D) states under these conditions. In our view any model that fails to incorporate obligatory H₂ loss as a fundamental aspect of N₂ activation is unlikely to provide a robust description of the chemistry associated with the biological process.²⁴² (ii) This incorporation provides a very clear demonstration of the essential mechanistic role for obligatory, reversible loss of H₂ upon N₂ binding and thus of the eight-electron stoichiometry for nitrogen fixation by nitrogenase embodied in eq 1. Until now, the data indicating that some H₂ must be evolved during N₂ reduction has been viewed as being much more compelling than the data indicating an obligatory evolution of one H₂ for every N₂ reduced, leading to the stoichiometry of eq 1.¹⁷ (iii) The formation of E₄(2D) and E₂(D) during turnover under D₂/N₂ in H₂O is predicted by the re mechanism for the activation of FeMo-cofactor for reduction of N₂, and the interception of these intermediates by C₂H₂ thus provides direct experimental evidence in support of this mechanism (Figure 17). (iv) The well-known reduction of protons by D₂ to form 2HD during turnover under D₂/N₂ in H₂O and the newly discovered reductions of C₂H₂ by D₂/N₂ should be viewed as being catalyzed by nitrogenase with N₂ as cocatalyst. (v) This review has proposed an explanation of the inability of H₂/D₂ to reduce nitrogenase and/or catalyze substrate reduction in the absence of N₂.

11. Conclusions

The trapping and characterization of five nitrogenase catalytic intermediates, which correspond to three of the five stages involved in binding and reduction of nitrogen (Figure 3), most especially the “Janus intermediate”, E₄, and including the nitrogenous intermediate states H (E₇) and I (E₈), have identified the “prompt–alternating (P–A)” pathway of Figure 12, carried out on a four-Fe face of FeMo-co, as most likely operative for nitrogenase and led to the unification of the nitrogenase reaction pathway and the LT kinetic scheme.

The recognition of the central role played by conversion of accumulated [e^–/H⁺] into metal hydrides has led to the proposal that this most complex of biological catalytic clusters, and by extension perhaps all biological clusters involved in multielectron substrate hydrogenation, function through a limited set of redox couples, and indeed most likely through a single couple, with multiple reducing equivalents being stored as hydrides rather than as reduced metal ions, Figures 6, 19. Only in this way can a cluster accumulate equivalents delivered at a constant potential set by its biological partners. These considerations provide part of the reason why such a large cluster is required for nitrogenase catalysis.

Simple energetic considerations have further illuminated the heretofore puzzling observation that states of nitrogenase activated by the accumulation of multiple [e^–/H⁺] can relax through release of H₂ (Scheme 1), but H₂ cannot reduce nitrogenase in what appears to be the reverse process: the answer is that the processes are not microscopic reverses.

Perhaps the central question of nitrogen fixation by nitrogenase has been that of stoichiometry: “Why does (or even, does) nature ‘waste’ four ATP/two reducing equivalents through an obligatory loss of H₂ when N₂ binds?” An answer has been proposed on the basis of further consideration of hydride chemistry exhibited by E₄: the enzyme exhibits the stoichiometry of eq 1 because reductive elimination (re) of two [e^–/H⁺] in the form of H₂ activates FeMo-co for hydrogenation of N₂ to N₂H₂ via a “push–pull” mechanism, Figure 13, lower. A test of the mechanism involving turnover under N₂/D₂/C₂H₂, as in Figure 15, validated the re mechanism, and in so doing confirmed the stoichiometry of nitrogen fixation, eq 1, as requiring eight [e^–/H⁺].

The test reaction further highlighted the role of N₂ as cocatalyst in reductions catalyzed by nitrogenase that would not occur in the absence of N₂. We have further noted some issues regarding the uniqueness of N₂ and nitrogenase as the catalyst for its hydrogenation, and of the implications of alternative structures of the N₂ complex.

The result of these efforts is the mechanism for nitrogen fixation presented in Figure 19 for further tests, both experimental and theoretical.

Biographies

Brian M. Hoffman was an undergraduate at the University of Chicago, received his Ph.D. from Caltech, and spent a postdoctoral year at MIT. From there, he went to Northwestern University, where he is the Charles E. and Emma H. Morrison Professor in the Departments of Chemistry and of Molecular Biosciences. He is a member of the National Academy of Sciences.

Dmitriy Lukoyanov received a M.S. degree and a Ph.D. from Kazan State University. He is a postdoctoral fellow at Northwestern University.

Zhi-Yong Yang received a B.S. degree in chemistry and a Ph.D. degree in organic chemistry from Nankai University in Tianjin, China. Following a position at Shanghai ChemPartner as an organic scientist, he joined the Ph.D. program in biochemistry at Utah State University, where he recently completed his degree. During the past 12 years, his research has covered biomimetic chemistry of hydrogenases, and the mechanism of nitrogenase. His research interest is in the activation and formation of chemical bonds catalyzed by biological and synthetic transition-metal catalysts.

Dennis R. Dean received a B.A. from Wabash College and a Ph.D. from Purdue University. He is currently a University Distinguished Professor at Virginia Tech where he also serves as the Director of the Fralin Life Science Institute and the Virginia Bioinformatics Institute.

Lance C. Seefeldt received a B.S. degree in chemistry from the University of Redlands in California and a Ph.D. in biochemistry from the University of California at Riverside. He was a postdoctoral fellow at the Center for Metalloenzyme Studies at the University of Georgia before joining the faculty of the Chemistry and Biochemistry Department at Utah State University, where he is now Professor. He was recently awarded the D. Wynne Thorne Career Research Award and elected a Fellow of the American Association for the Advancement of Science. Over the past 25 years, his research has focused on elucidating the mechanism of nitrogenase.

Supporting Information Available

Analysis of the relative orientations of bridging hydrides of E₄ intermediate. This material is available free of charge via the Internet at http://pubs.acs.org.

The authors declare no competing financial interest.

Funding Statement

National Institutes of Health, United States