Reductive Elimination of H₂ Activates Nitrogenase to Reduce the N≡N Triple Bond: Characterization of the E₄(4H) Janus Intermediate in Wild-Type Enzyme

Abstract

We have proposed a reductive elimination/oxidative addition (re/oa) mechanism for reduction of N₂ to 2NH₃ by nitrogenase, based on identification of a freeze-trapped intermediate of the α-70^Val→Ile substituted MoFe protein as the Janus intermediate that stores four reducing equivalents on FeMo-co as two [Fe-H-Fe] bridging hydrides (denoted E₄(4H)). The mechanism postulates that obligatory re of the hydrides as H₂ drives reduction of N₂ to a state (denoted E₄(2N2H)) with a moiety at the diazene (HN=NH) reduction level bound to the catalytic FeMo-cofactor. In the present work, EPR/ENDOR and photophysical measurements show that a state freeze-trapped during N₂ reduction by wild type (WT) MoFe protein is the same Janus intermediate, thereby establishing the α-70^Val→Ile intermediate as a reliable guide to mechanism, and enabling new experimental tests of the re/oa mechanism with WT enzyme. These allow us to show that the re/oa mechanism accounts for the longstanding Key Constraints on mechanism. Monitoring the S = ½ FeMo-co EPR signal of Janus in WT MoFe during N₂ reduction under mixed-isotope condition, H₂O buffer/D_2, and the converse, establishes that the bridging hydrides/deuterides do not exchange with solvent during enzymatic turnover, thereby explaining earlier observations and verifying the re/oa mechanism. Relaxation of E₄(2N2H) to the WT resting-state is shown to occur via oa of H₂ and release of N₂ to form Janus, followed by sequential release of two H₂, demonstrating the kinetic reversibility of the re/oa equilibrium. The relative populations of E₄(2N2H) and E₄(4H) freeze-trapped during WT turnover furthermore show that the rapidly reversible re/oa equilibrium between [E₄(4H) + N₂] and [E₄(2N2H) + H₂] is roughly thermoneutral (Δ_reG⁰ ~ −2 kcal/mol), whereas hydrogenation of gas-phase N₂ would be highly endergonic. These findings establish (i) that re/oa satisfies all key constraints on mechanism, (ii) that Janus is the key to N₂ reduction by WT enzyme, which (iii) indeed occurs via the re/oa mechanism. Thus emerges a picture of the central mechanistic steps by which the nitrogenase MoFe protein carries out one of the most challenging chemical transformation in biology, the reduction of the N≡N triple bond.

Graphical abstract

Introduction

By catalyzing biological nitrogen fixation — the reduction of N₂ to two NH₃ molecules — nitrogenase generates the nitrogen-containing nutrients that support most of the biosphere, including over half the human population.¹ But an understanding of the mechanism of N₂ reduction by nitrogenase has been elusive. A ‘kinetic’ foundation for the mechanism of the Mo-dependent nitrogenase was developed through extensive studies in the 1970’s and 1980’s.^2-4 The culmination of these measurements was the Lowe-Thorneley (LT) kinetic scheme for nitrogenase function, Fig 1,^2-4 which describes the transformations among catalytic intermediates, denoted E_n where n is the number of steps of electrons/protons delivered from the nitrogenase Fe protein to the MoFe protein, which contains the active site iron-molybdenum cofactor ([7Fe-9S-Mo-C-R-homocitrate, denoted M]; FeMo-co) Fig 2. A defining feature of this scheme is the obligatory formation of one mole of H₂ per mole of N₂ reduced, which leads to a limiting stoichiometry for enzyme-catalyzed nitrogen fixation given by eq 1,

$${\text{N}}_2+8{\text{e}}^{-}+16\text{ATP}+8{\text{H}}^{+}\to 2{\text{NH}}_3+{\text{H}}_2+16\text{ADP}+16{\text{P}}_{\text{i}}$$ (1)

in agreement with stoichiometric measurements by Simpson and Burris.⁵ However, the obligatory requirement for H₂ formation has not been universally accepted, and was even questioned in the culminating review of Burgess and Lowe: “Thus the data that support the obligatory evolution of one H₂ for every N₂ reduced are much less compelling than the data that require us to believe that some H₂ will always be evolved during N₂ reduction.”²

Figure 1.

Simplified Lowe-Thorneley (LT) kinetic scheme for nitrogen reduction^2-4 that focus on the electron-accumulation and FeMo-co activation (boxed) stages. In the E_n notation, n = number of [e⁻/H⁺] added to FeMo-co; in parentheses, the stoichiometry of H/N bound to FeMo-co, except n = 7, 8, where structures are shown.

Figure 2.

Crystal structure of FeMo-co. Fe is show in rust, Mo in magenta, S in yellow, carbide in dark-grey, carbon in gray, N in blue and O in red. The Fe atoms of catalytic 4Fe-4S face are labelled as 2, 3, 6, and 7. Two amino acids, α-70^Val and α-195^His, that approach the FeMo-co are also shown. The image was created using PDB coordinate 2AFI.

We recently proposed^6,7 that obligatory H₂ formation during nitrogenase N₂ reduction is in fact required to explain the catalytic function of nitrogenase. This proposal originates with the characterization of an intermediate (FeMo-co spin S = ½) trapped during turnover of MoFe protein having the α-70^Val→Ile substitution, which apparently inhibits access by all substrates to the active site, other than protons.⁸ A combination of ^1,2H/⁹⁵Mo ENDOR spectroscopy^8,9 and cryoannealing ‘electron counting’¹⁰ showed this state to be the key E₄(4H) ‘Janus’ intermediate, which has accumulated four of the eight reducing equivalents required by eq 1, storing them as two [Fe-H-Fe] bridging hydrides.^8-10 E₄(4H) sits at a transition in the N₂ reduction pathway, Fig 1, poised to ‘fall back’ to the E₀ resting state by successive release of two H₂,¹⁰ but equally poised to eliminate H₂ and proceed to the reduction of N₂ to two NH₃ through the accumulation of four more equivalents, hence the appellation, ‘Janus’.⁷

The discovery that the four reducing equivalents accumulated by E₄(4H) are stored as bridging hydrides forged a connection between nitrogenase catalysis and the organometallic chemistry of metal hydrides^11-13 that offered explanations of a multitude of features of nitrogenase mechanism that had defied explanation for decades.² At the most basic level, this hydride formation helped to explain how a constant-potential electron donor, reduced Fe protein (Fe^red), could reduce FeMo-co by four equivalents. Moreover, bridging hydrides are less susceptible to protonation to form H₂, or to solvent exchange, than terminal hydrides. As a consequence of the latter, bridging hydrides diminish the tendency of FeMo-co to ‘fall back’ to resting state through the formation of two H₂ (Fig 1) However, the bridging mode also lowers hydride reactivity, relative to that of terminal hydrides. How the release of H₂ contributes to the activation of this ‘deactivated’ intermediate for the hydrogenation of N₂ to a moiety at the N₂H₂-reduction level is thus a central ‘mystery’ of N₂ reduction by nitrogenase.

Importantly, reference to the inorganic chemistry of metal-dihydrides offered an explanation to this mystery, as well. Once it is recognized that E₄(4H) contains two bridging hydrides, then the chemistry of metal-dihydride complexes^11-13 identifies the LT E₄(4H)↔E₄(2N2H) mechanistically coupled equilibrium (Fig 1) as the reductive elimination (re) of H₂ and its reverse, the oxidative addition (oa) of H₂, Fig 3. We proposed that for nitrogenase, the re of H₂ carries off two of the four reducing equivalents formally stored as the H⁻, driving the first, and most difficult, step of N₂ cleavage; reduction of the N₂ triple bond to a diazene-level (2N2H) moiety bound to FeMo-cofactor, Fig 3, by the remaining two reducing equivalents. This process fulfills one of the long-standing key constraints on mechanism: Chart 1, (i).²

Figure 3. Schematic of re/oa Equilibrium.

The cartoon represents the Fe 2,3,6,7 face of FeMo-co, and the ‘2N2H’ implies a species at the diazene reduction level of unknown structure and coordination geometry. In the indicated equilibrium the binding and activation of N₂ is mechanistically coupled to the re of H₂, as described in the text. Some of the potential complexities associated with this enzymatic process that underlie this cartoon are discussed in reference 18..

Chart 1.

Key Constraints on Nitrogenase Mechanism²

The proposal that N₂ fixation requires the obligatory re of H₂ was first supported by its ability to provide explanations of key constraints on nitrogenase mechanism summarized in Chart 1.² These include not only the E_n state at which N₂ is first reduced (Key Constraint (i)), but most especially previously baffling results from turnover in the presence of D₂/T₂. It was found that: Key Constraints (ii)), D₂ only reacts with nitrogenase during turnover under an N₂/D₂ atmosphere; (iia) in this reaction D₂ is stoichiometrically reduced to two HD (with H₂O buffer); (iib), corresponding turnover with T₂ does not lead to exchange of T⁺ into H₂O solvent.² These constraints are explained as arising through the reverse of the re process, Fig 3, namely the oxidative addition (oa) of H₂ with loss of N₂.^6,7 According to the mechanism, during turnover under D₂/N₂, reaction of the E₄(2N2H) intermediate with D₂ generates dideutero-E₄ with two [Fe-D-Fe] bridging deuterides which do not exchange with solvent.¹⁴ This E₄(4H) isotopologue, which we denote E₄(2D⁻;2H⁺) (Chart 2), would relax through E₂(D⁻;H⁺) to E₀ with successive loss of two HD ^7,10.

Chart 2.

The proposed formation of the E₄(2D⁻;2H⁺) and E₂(D⁻;H⁺) states through the thermodynamically allowed reverse of re, the oa of D₂ with accompanying release of N₂, led to a successful test of their formation. During turnover under N₂/D₂, the non-physiological substrate acetylene (C₂H₂) intercepted these states to generate deuterated ethylenes (C₂H₃D and C₂H₂D₂).¹⁴ More recently, we identified an S = ½ EPR signal that appears during N₂ reduction by wild type (WT) MoFe protein as arising from FeMo-co of the E₄(2N2H) state formed by re of H₂ with N₂ reduction, Fig 3, and in so doing confirmed the prediction that the (re/oa) activation equilibrium is not only thermodynamically, but also kinetically reversible.¹⁵ Characterization of the E₄(4H) ‘Janus’ intermediate as carrying four reducing equivalents in the form of two [Fe-H-Fe] bridging hydrides thus laid the foundation for the re/oa mechanism, Fig 3, with its obligatory formation of one H₂ per N₂ reduced and resultant limiting stoichiometry of eq 1,¹⁵ while the success of predictions based on the mechanism provides powerful support for the mechanism.

We here report two major types of advance in our understanding of nitrogenase catalysis. The first was motivated by our recognition that, although the predictions of re/oa were tested in WT MoFe, the mechanism was founded on the presence and properties of the two [Fe-H-Fe] bridging hydrides in the Janus intermediate trapped and characterized in the α-70^Val→Ile MoFe protein, which shows a much decreased ability to bind and reduce N₂. This raised the question: Do the properties of the Janus intermediate in the MoFe variant accurately reflect those of the WT intermediate, and are mechanistic conclusions based on studies of the variant applicable to WT enzyme? This report answers these coupled questions: yes. We here establish that the same E₄(4H) Janus intermediate, with its two [Fe-H-Fe] bridging hydrides, in fact does participate in catalysis by the WT enzyme through the re/oa mechanism for N₂ reduction, thereby showing the α-70^Val→Ile Janus intermediate to be a reliable guide to mechanism.

The second type involves observations founded on the ability to monitor both the E₄(4H) and E₄(2N2H) partners of the re/oa equilibrium in WT enzyme (Fig 3). This ability enables us to experimentally demonstrate that the re/oa mechanism indeed satisfies the Key Constraints of Chart 1, and beyond that to measure the energetics and kinetics of the equilibrium interconversion. Overall, this report yields a picture of the key mechanistic steps by which nitrogenase carries out one of the most challenging chemical transformation in biology, the reduction of the N≡N triple bond.¹⁶

Materials and Methods

Materials and Protein purifications

All the reagents were obtained from SigmaAldrich (St. Louis, MO) or Fisher Scientific (Fair Lawn, NJ) and were used without further purification. Argon and N₂ gases were purchased from Air Liquide America Specialty Gases LLC (Plumsteadville, PA). Experiments were carried out with WT Azotobacter vinelandii MoFe protein expressed and purified as described elsewhere.¹⁷ The handling of all buffers and proteins were done anaerobically unless stated otherwise.

EPR and ENDOR samples

Samples were prepared as described under turnover conditions specified in figure legends; X-band samples contain 50 μM MoFe protein, 75 μM Fe protein, and the concentration of intermediates trapped during turnover is typically ~ 20 μM.^15,18 For Q band samples, 100 μM MoFe and 150 μM Fe was used. All the samples were allowed to turnover for 20-25s, transferred to a X or Q band tube, and frozen with liquid nitrogen..^15,18 It was found that during turnover of WT enzyme under low partial pressures of N₂ a g₁ = 2.15 species, shown below to be the E₄(4H) intermediate in WT enzyme could be trapped in populations adequate for study, typically along with its ‘equilibrium partner’ the E₄(2N2H) intermediate. The properties of g₁ = 2.15, E₄(4H) are compared with those of E₄(4H) trapped in the α-70^Val→Ile and α-70^Val→Ile/α-195^His→Gln MoFe proteins as described.^17,18

EPR and ENDOR Measurements

X-band CW EPR spectra and Q-band CW EPR and ^1,2H ENDOR spectra, including those during in situ photolysis by constant 450nm diode laser illumination, were collected as described.¹⁸ Data points for kinetics of photolysis and annealing were obtained as amplitudes of EPR signals at the well-resolved g₁ features of the E₄ states linked by re/oa. Tests of this protocol by EPR simulations show that the error of this quantitation approach should not exceed 15% when both E₄(4H) and E₄(4D) intermediates are present. We recall a time constant for photolysis is inversely proportional to the intensity of illumination and photolysis quantum yield.¹⁸ As a result the kinetic isotope effect on photolysis corresponds to the effect on quantum yield.

Results and Discussion

Enhancing the population of the WT g₁ = 2.15 intermediate

Figure 4 presents EPR spectra of WT MoFe protein freeze-quenched under turnover at high and low partial pressure of N₂, along with a spectrum of the E₄(4H) intermediate trapped during turnover of α-70^Val→Ile MoFe protein. The spectrum for the α-70^Val→Ile enzyme shows the signal from the E₄(4H) Janus intermediate, g = [2.15, 2.007, 1.965]. The WT enzyme turned over under high partial pressure of N₂ (P(N₂)) shows the EPR signal from the E₄(2N2H) intermediate, in which FeMo-co binds a diazene-level product of N₂ binding/reduction, g = [2.09, 1.99, 1.97]. However, WT enzyme freeze-trapped under low P(N₂) turnover with high solution concentration of H₂ (high effective P(H₂), see figure legend and ref ¹⁵) shows only traces of the signal from E₄(2N2H), and instead shows an EPR signal of an intermediate with g₁ = 2.15, which is indistinguishable from that of E₄(4H) in the MoFe α-70^Val→Ile variant.^8,10 Closer inspection of the high-P(N₂) signal then shows a trace of the g₁ = 2.15 feature, associated with a very low population of this intermediate, confirming its presence under all turnover conditions.

Figure 4.

X-band EPR spectra of WT nitrogenase turnover samples trapped under 1 atm of N₂ (with stirring to facilitate transfer of H₂ formed during turnover into the headspace¹⁵), and under low P(N₂) in H₂O buffer (without stirring) shown in comparison with spectrum of E₄(4H) state trapped during turnover of α-70^Val→Ile MoFe protein of the same concentration (see M&M). EPR conditions: temperature, 12 K; microwave frequency, 9.36 GHz; microwave power, 10 mW; modulation amplitude, 9 G; time constant, 160 ms; field sweep speed, 20 G/s.

Such a correlation of relative intensities with P(N₂) and effective P(H₂) is as expected if this g₁ = 2.15 signal is indeed from WT E₄(4H). In this case its concentration relative to that of E₄(2N2H) is governed by the equilibrium of Fig 3, when, as expected, the forward and reverse steps of the re/oa equilibrium are rapid compared to the slow delivery of the next electron from Fe^red (Fig 1),

$$\frac{E_4\left(2N2H\right)}{E_4\left(4H\right)}=K_{re}\frac{P\left(N_2\right)}{P\left(H_2\right)}\propto P\left(N_2\right)$$ (2)

Here K_re is the equilibrium constant for the re/oa equilibrium, and a simple proportionality to P(N₂) follows from our observation that at these high enzyme concentrations, turnover produces a roughly constant (saturating) concentration of H₂ regardless of P(N₂).¹⁵ Not only does this P(N₂) dependence supports the idea that the g₁ = 2.15 state is indeed the WT E₄(4H), but of central importance to this study, the ability to prepare samples whose dominant EPR signal is that of the g₁ = 2.15 state enables its characterization and identification, as now described.

^1,2H Hyperfine Coupling to Bridging [Fe-H/D-Fe] in WT MoFe

ENDOR spectroscopy revealed the presence in E₄(4H) of α-70^Val→Ile and α-70^Val→Ile/α-195^His→Gln MoFe protein of two [Fe-H-Fe] bridging hydrides with anisotropic hyperfine tensors whose principal values are virtually identical, A1 = [11, 25, 37] MHz, A2 = [10, 24, 33] MHz, but whose orientations differ.⁸ To begin this investigation, we examined how the hyperfine coupling to the two hydrides influence the EPR spectrum of E₄(4H) of α-70^Val→Ile MoFe protein. Figure 5A presents CW X-band EPR spectra of this intermediate prepared in H₂O and D₂O buffers, along with expansions of the g₁ and g₃ regions of the spectra plus simulations that incorporate the ENDOR-derived hyperfine couplings to the hydrides/deuterides (the accompanying protons have only small couplings and do not significantly contribute). The difference in breadth of the g₁ feature in H₂O and D₂O buffers is well-captured by the simulated ¹H hyperfine broadening from the two hydrides, as calculated using the ENDOR-derived couplings. In particular, at g₃, the EPR spectrum of the intermediate in H₂O buffer even shows the 1-2-1 pattern arising from comparable couplings to the two ¹H hydrides, and this too is reproduced by the simulation. The loss of this pattern in D₂O buffer definitively confirms its identification with coupling to the two hydrides.

Figure 5.

(A) EPR spectra of WT and α-70^Val→Ile turnover samples trapped in H₂O (red) and D₂O (blue) buffers. Spectra are shown after normalization to the same g₁ feature amplitude and their extended g₁ and g₃ fragments are compared with EPR simulations obtained with hydrides/deuterides hyperfine interaction parameters known from previous ENDOR study.⁸ Red arrows indicate resolved features associated with hydrides hyperfine interaction. EPR conditions: temperature, 12 K; microwave frequency, 9.36 GHz; microwave power, 10 mW; modulation amplitude, 2.3 G; time constant, 160 ms; field sweep speed,10 G/s; 4 scans. (B) Mims ²H ENDOR spectra of WT low P(N₂) turnover in D₂O (middle) shown in comparison with previously obtained stochastic CW ENDOR of hydrides of E₄(4H) trapped in α-70^Val→Ile/α-195^His→Gln protein during turnover in H₂O (top).¹⁸ Spectra of ²H and ¹H are scaled to the same Larmor frequency. In D₂O spectra, down triangles indicate 'blind spots' of Mims ENDOR spectra with suppressed hyperfine couplings of A = n/τ , n = 1, 2, …; *-labeled signal in the H₂O background Mims τ = 400 ns spectrum is associated with 5^th harmonics of the matrix ¹H response. Mims ENDOR conditions: microwave frequency, 34.743 GHz; π/2 = 50 ns; RF 20 μs; repetition time 50 ms; ~ 400-800 scans; temperature, 2 K.

The EPR spectrum of the WT g₁ = 2.15 species in H₂O buffer has the same breadth at g₁ = 2.15 as that of E₄(4H) of α-70^Val→Ile, and shows the same narrowing in D₂O. At g₃ the signals from residual Fe^red and E₄(2N2H) largely obscure the g₁ = 2.15 intermediate signal for the WT enzyme, but the spectrum in H₂O buffer nonetheless shows clear evidence for the same 1-2-1 hyperfine-coupling pattern from two [Fe-H-Fe] hydrides as seen for the α-70^Val→Ile variant, a feature that is lost in D₂O buffer as with the α-70^Val→Ile intermediate. Thus, the WT and α-70^Val→Ile enzymes have equivalent hyperfine-coupled hydrides.

Comparison of ENDOR responses from E₄(4H) in α-70^Val→Ile MoFe protein and the g₁ = 2.15 intermediate of WT enzyme confirms the presence of the two bridging hydrides in the WT intermediate. Figure 5B shows a 2 K Q-band stochastic CW ¹H ENDOR spectrum collected at g₂ = 2.01 for E₄(4H) α-70^Val→Ile/α-195^His→Gln.¹⁸ This spectrum is one component of the 2D field-frequency pattern of spectra collected across the EPR envelope used to carry out the ENDOR analysis. In this spectrum, taken at the field of maximum EPR intensity,⁸ the strongly-coupled ¹H signals from the two hydrides are completely overlapped, yielding a single composite (structured) doublet whose feature of maximum intensity corresponds to a hyperfine splitting, A(¹H) ~ 24 MHz.

The population of the g₁ = 2.15 intermediate trapped during turnover of WT enzyme in H₂O buffer is low, even when enhanced by low P(N₂), which prevented collection of a satisfactory ENDOR response for strongly-coupled [Fe-¹H-Fe] protons. However, during turnover of the WT enzyme in D₂O buffer the intermediate is trapped with a more than two-fold higher population. As a result, we could obtain ²H ENDOR signals from strongly-coupled [Fe-²H-Fe] deuterons. The Q-band ²H Mims pulsed-ENDOR spectrum at g = 2.01 shows a narrow ‘distant deuteron’ signal centered at the ²H Larmor frequency (Fig 5B), but in addition shows a pair of peaks offset from the ²H Larmor frequency by ~ ±1.85 MHz. The identification of this pair as a hyperfine-split doublet with A(²H) ~ 3.7 MHz is confirmed by its suppression when the interval, τ, in the Mims microwave pulse sequence is chosen appropriately for the hyperfine coupling, in this case A(²H)τ ~ 2, and by the absence of the signal in a sample prepared in H₂O buffer, Fig 5B.¹⁹ When scaled by the nuclear g-factors, the coupling for this feature corresponds to a proton coupling of A(¹H) ~ 24 MHz, the value seen in the ¹H spectrum from E₄(4H) α-70^Val→Ile/α-195^His→Gln.

Together, the EPR and ENDOR measurements demonstrate that the WT g₁ = 2.15 intermediate not only has the same g-values as α-70^Val→Ile E₄(4H) (vide supra), but also exhibits hyperfine couplings that correspond to the pair of [Fe-H-Fe] bridging hydrides of α-70^Val→Ile E₄(4H), thereby establishing that the intermediate in WT MoFe protein corresponds to that in the α-70^Val→Ile variant: it is the E₄(4H) Janus intermediate of N₂ reduction by the WT enzyme.

Photoinduced re of WT and α-70^Val→Ile E₄(4H) Hydrides; oa of H₂ by cryoannealing E₄(2H)*

The assignment of the WT g₁ = 2.15 intermediate to E₄(4H), with its two [Fe-H-Fe] bridging hydrides, is confirmed by the behavior of this intermediate under photolysis. We were inspired to take this approach by considering that the photolysis of transition metal dihydride complexes (with mutually cis hydride ligands) commonly results in the release of H₂,^20-28 which represents “a typical example of reductive elimination”, while the thermal reverse reaction “is the prototype example of an oxidative addition reaction.”²⁰ Thus, we tested whether the two bridging hydrides of E₄(4H) would behave in this fashion. ENDOR and EPR measurements showed that photolysis of E₄(4H) α-70^Val→Ile at 4 K and above generates a new FeMo-co state, denoted E₄(2H)*, through the photoinduced re of the two bridging hydrides as H₂. The E₄(2H)* thus trapped relaxes to the initial, equilibrium, E₄(4H) form during cryoannealing of the frozen solid at temperatures above 175 K, where the oxidative addition (oa) of the eliminated H₂ by E₄(2H)* becomes kinetically allowed. The photolysis quantum yield is temperature invariant at liquid helium temperatures and shows a large kinetic isotope effect, KIE ≈ 10. These observations imply the photoinduced release of H₂ involves a barrier to the combination of the two nascent H atoms and further suggest that H₂ formation involves nuclear tunneling through that barrier.

Figure 6 compares EPR signals collected during the in situ irradiation by 450 nm light of E₄(4H) α-70^Val→Ile, WT E₄(4H), and E₄(2N2H) held at 12 K in an EPR cavity. Figure 7A plots the time course for the intensity of the EPR signals of these intermediates during the in situ photolysis. Irradiation converts E₄(4H) α-70^Val→Ile (g = [2.15, 2.007, 1.965) to E₄ (2H)* (g = [2.098, 2.0, 1.956]), Fig 6. The figure shows that irradiation of WT E₄(4H) likewise converts this state to an E₄(2H)* state with g-values identical to those in the α-70^Val→Ile MoFe variant. In contrast, the EPR spectra of Fig 6 and progress curves of Fig 7A, show that the E₄(2N2H) signal is unaffected by photolysis. That a bound nitrogenous moiety, the N₂-derived 2N2H moiety of E₄(2N2H), is not photodissociable supports the idea that the photosensitivity is associated with the presence of bound hydrides. The photolyis results thus confirm the presence of the two bridging hydrides in the WT g = 2.15 intermediate, and its identification as E₄(4H).

Figure 6.

Photoinduced changes in EPR spectra of WT and α-70^Val→Ile freeze-trapped during enzymatic turnover in H₂O during 24.5 minutes of 450 nm diode laser irradiation at 12 K. EPR conditions: the same as in Figure 4.

Figure 7.

(A) Time courses of E₄(4H) and E₄(4D) states in α-70^Val→Ile and WT and E₄(2N2H) in WT 1 atm N₂ turnover in H₂O during irradiation with 450 nm diode laser at 12 K. Data points in this and following kinetics plots were obtained as described in Materials and Methods; decays of E₄(4H) and E₄(4D) are fitted as stretched exponential with following parameters: τ = 9.2 min, m = 0.51 for H₂O and τ = 83.3 min, m = 0.45 for D₂O. EPR conditions: for this and the following kinetics figures are the same as in Figure 4. (B) Time courses of E₄(4H)/ E₄(4D) recovery during 193 K annealing of irradiated WT and α-70^Val→Ile turnovers in H₂O/D₂O. Time constants obtained from exponential fit: 8.4 min (H₂O) and 41.5 min (D₂O), KIE ~ 5.

The progress curves for in situ irradiation of E₄(4H) in WT and α-70^Val→Ile MoFe protein (Fig 7A) are the same, within error, showing that the quantum yield for re of H₂ is independent of environment - WT enzyme vs α-70^Val→Ile variant. As a result, the re progress curves for WT and α-70^Val→Ile in H₂O buffer have been jointly fit to the stretched exponential behavior (exp(−(t/τ)^m))^29,30 that is a consequence of photolysis in a non-glassy sample,¹⁸ and likewise for those in D₂O buffer. As reported for photolysis of E₄(4H) α-70^Val→Ile, the joint progress curves show a large KIE, defined as the ratio of the median decay times for D₂O and H₂O buffers, KIE ~ 9, which implies that photoinduced re of the two hydrides and release of H₂ involves a barrier to the combination of the two nascent H atoms. Likewise, as reported for the α-70^Val→Ile intermediate, the decay time for the WT intermediate is temperature invariant, within error, for T = 4-12 K, which suggests that the formation of H₂ involves nuclear tunneling through that barrier.³¹

Figure 7B plots the timecourse for the oa of H₂ by the E₄(2H)* and E₄(2D)* to regenerate the corresponding E₄(4H)/E₄(4D) states during 193 K cryoannealing of WT and α-70^Val→Ile MoFe samples.¹⁸ In these experiments the sample is annealed at 193 K multiple times, with cooling for collection of EPR spectra at 12 K between annealing periods. For both H₂O and D₂O buffers the relaxation of E₄(2H)* oa of H₂ regenerates the E₄(4H) state formed during turnover, and the progress curves are well-fitted as a single-exponential processes. As with photoinduced re, the progress curves during cryoannealing oa are the same for the two MoFe variants, and their joint fit yields KIE ~ 5, rather larger than seen for closed-shell monometallic complexes.^32,33 Combined with a strong temperature dependence of the time-constant (not shown), this KIE implies that oa of H₂ involves traversal of an energy barrier.

Overall, the measurements of photoinduced, re loss of H₂ from E₄(4H) and thermal regeneration of the E₄(4H) state via oa of H₂ by the resulting E₄(2H)* confirm the identification of the WT and α-70^Val→Ile intermediates as the same Janus, in showing that both the excited and ground state energy surfaces associated with these processes are essentially the same in the WT and variant MoFe proteins.

Accumulation of E₄(4H) Isotopologues During Mixed-Isotope Turnover; Bridging Hydrides are Exchange-Inert

The kinetic reversibility of the re/oa mechanism in WT enzyme, as well as the stability of the bridging hydrides to solvent exchange, as required by Key Constraint (ii, b), Chart 1, are here established by the use of EPR and photolysis to measure the isotopic composition of the bridging hydrides in WT Janus trapped during N₂ turnover under isotopically mixed conditions: H₂O buffer under an atmosphere that include D₂ and conversely, Fig 8. These measurements show that the re/oa mechanism satisfies Key Constraints, Chart 1, on catalysis by WT nitrogenase

Figure 8.

(A) The g₁ features of Janus state EPR recorded for WT protein turnovers trapped under mixtures of 0.1 atm N₂ with 0.9 atm of H₂ or D₂ in H₂O and D₂O buffers with stirring. EPR conditions: temperature, 12 K; microwave frequency, ~9.36 GHz; microwave power, 10 mW; modulation amplitude, 4.5 G; time constant, 160 ms; field sweep speed, 5 G/s; 4-8 scans. (B) Photolysis of WT Janus intermediate formed through isotopically-mixed turnover; irradiation with 450 nm light at 12 K. Progress curves for N₂/H₂, H₂O and N₂/D₂, D₂O are fitted as stretched exponential decays with parameters shown in the figure. Kinetics of other two samples can be well fitted with following parameters: τ = 52.6 min, m = 0.47 for N₂/D₂ turnover in H₂O and τ = 38.3 min, m = 0.46 for N₂/H₂ turnover in D₂O. Dotted lines present alternative fits as sums of two decays corresponding to photolysis of E₄(4H) in H₂O and E₄(4D) in D₂O turnover samples with ratios shown in the figure.

According to the re/oa mechanism, N₂ turnover in H₂O buffer under D₂ would generate E₄(4H) through turnover accumulation of reducing equivalents and protons derived from solvent (Fig 1), but in addition, oa of D₂ from the gas phase by E₄(2N2H) would generate E₄(2D⁻;2H⁺), with two [Fe-D-Fe] bridges and two bound protons, and conversely for turnover in D₂O buffer under H₂ (Fig 3). As a result, during N₂ turnover under both mixed-isotope conditions, two isotopologues of Janus are expected to accumulate, one in which the two bridges have the solvent H/D isotope, the other with the two bridges generated through oa of the diatomic D₂/H₂ in the gas phase.¹⁴

Figure 8A shows that the g₁ = 2.15 feature from the EPR spectrum of the WT E₄(4H) intermediate, as trapped in H₂O buffer during turnover under N₂/D₂, is distinctly narrower than for turnover under N₂/H₂, thus demonstrating the accumulation of an intermediate in which the reverse of the re/oa equilibrium has generated [Fe-D-Fe] bridging deuterides with loss of the ¹H hyperfine broadening (Fig 5A, above). Conversely the g₁ = 2.15 feature for the WT Janus intermediate in D₂O buffer is correspondingly broader for turnover under N₂/H₂ than for turnover under N₂/D₂, demonstrating the accumulation of an intermediate with bridging [Fe-H-Fe], which contribute ¹H hyperfine broadening. Simulations that sum roughly equal contributions of the limiting spectra for E₄(4H) and E₄(4D) in fact reproduce the mixed-isotope spectrum quite well (not shown). Note especially, that the demonstration that deuterides/hydrides acquired by oa of the gas-phase diatomic accumulate in the catalytic intermediate, rather than exchanging with a solvent of opposite isotopic composition, confirms our proposal^6,7 that these bridging hydrides are exchange-inert, and that their formation during oa of a gas-phase diatomic explains why turnover under N₂/T₂ does not lead to the exchange of T⁺ into the solvent.²

The accumulation of mixtures of E₄(4H) isotopologues during turnover by oa of H₂/D₂ under isotopically-mixed conditions is actually seen most dramatically when comparing measurements of the KIE for the photolysis of WT Janus trapped during turnover under N₂ in isotopically homogeneous conditions – H₂O buffer with addition of 0.9 atm of H₂, or D₂O buffer with added D₂ – with those for isotopically-mixed turnover conditions – H₂O buffer with added D₂ or D₂O buffer with added H₂. As expected, the photolysis traces from isotopically homogeneous intermediates, formed in H₂O buffer with N₂/H₂ and in D₂O buffer with N₂/D₂, Fig 8B, show a large KIE ~ 11, within error the same as seen for re of the two hydrides/deuterides formed in H₂O/D₂O buffers without the diatomics in the atmosphere (Fig 7A).

However, according to the re/oa mechanism, oa of D₂ by E₄(2N2H) during turnover in H₂O buffer under D₂ generates E₄(2D⁻;2H⁺), with two [Fe-D-Fe] bridges and two bound protons, and conversely for turnover in D₂O buffer under H₂. To a good approximation each of these E₄(4H) mixed isotopologues should undergo photoinduced re with quantum yield associated with the isotopic composition of the bridges, as determined by the complementary diatomic (see above), independent of the isotopic character of the protons/deuterons on sulfur, as determined by the solvent. Thus, if the H/D bridges introduced from the diatomic do not exchange with solvent, the intermediates formed in mixed isotope turnover should have apparent rates of photolysis roughly midway between those of E₄(4H) and E₄(4D). Indeed, Fig 8B shows that the Janus intermediate formed during turnover under N₂/D₂ in H₂O buffer photolyzes more slowly than with N₂/H₂ in H₂O, while the intermediate formed under N₂/H₂ in D₂O buffer photolyzes more rapidly than with N₂/D₂ in D₂O. The photolysis of these mixed-isotope samples each can be described by time constant roughly midway between those of the two isotopically homogeneous samples (see figure legend). Alternatively, as shown in Fig 8B, each isotopically mixed trace can be fit as the sum of a roughly f ~ 50% contribution from the progress curve for photolysis of E₄(4H), with two bridging hydrides, plus a contribution of (1-f) from the curve for E₄(4D) with two bridging deuterides.³⁴

These EPR and photochemical observations thus confirm that turnover of WT MoFe protein under N₂ involves a rapidly reversible re/oa equilibrium between the E₄ Janus intermediate and the E₄ state with diazene-level dinitrogen reduction product, Fig 3. Furthermore, the buildup of E₄(2D⁻;2H⁺) with bridging deuterides during N₂/D₂ turnover in H₂O buffer, and the converse, confirm that the bridging hydrides/deuterides of the Janus intermediate do not exchange with solvent during turnover. Thus, these measurements show that the re/oa mechanism indeed satisfies Key Constraints (i), (ii)b,c, Chart 1, as proposed.^6,7

Kinetic reversibility of the re/oa equilibrium during N₂ reduction by WT MoFe

The E₄(4H) α-70^Val→Ile intermediate was shown to have accumulated four reducing equivalents by a quench-cryoannealing relaxation protocol corresponding to that described above for the regeneration of E₄(4H) by oa of H₂ to the photogenerated E₄(2H)*. Keeping the sample frozen prevents any additional accumulation of reducing equivalents because binding of reduced Fe protein to and release of oxidized protein from the MoFe protein both are abolished in a frozen solid. As recently confirmed,³⁵ the frozen intermediate can relax towards the resting state only through steps that release a stable species from FeMo-co, with the E_n states formed prior to N₂ binding losing two equivalents per relaxation step in the release of H₂. By this approach, E₄(4H) was identified by its relaxation to the resting state E₀ through the release of a total of four reducing equivalents in a two-step process, each step involving hydride protonation with release of H₂ (two equivalents per step), with formation of E₂(2H) in the first step (Fig 1).¹⁰

As an extension of this procedure, the FeMo-co S = ½ E₄(2N2H) intermediate trapped during catalytic turnover of WT enzyme was identified by the finding that the decay of this state in a frozen reaction mixture is accelerated by increasing [H₂] and slowed by increasing [N₂], which directly demonstrated that the intermediate is the product of the kinetically, as well as thermodynamically reversible (re/oa) activation equilibrium, Fig 3. However, in that study the E₄(4H) precursor to re was accumulated in such low levels that its kinetic progress curve during cryoannealing could not be monitored directly, and we were forced to analyze the decay of E₄(2N2H) to resting state, Chart 3, through a steady-state approximation for the concentration of E₄(4H).

Chart 3.

The preparation of freeze-quenched WT samples that contain significant populations of both E₄(4H) and E₄(2N2H), obtained through control of P(N₂) during turnover, has now enabled us to directly monitor the progress curves in WT enzyme of all LT species on the cryoannealing relaxation pathway of E₄(2N2H) to resting state (E_n states with n ≤ 4, even), Chart 3. In so doing we directly track the kinetics of the species linked through the oa reverse, of the re/oa equilibrium, Figs 1, 3. These progress curves for the relaxation of E₄(2N2H) and kinetically linked intermediates (Chart 3) during cryoannealing of WT enzyme at −50 °C are presented in Fig 9. These are well described by the curves calculated by fitting them to the sequential kinetic scheme of Fig 9, which corresponds to the reverse of the LT scheme starting at E₄(2N2H), Fig 1 and Chart 3. The fitting procedure allowed for each step to exhibit distributed kinetics, and the time-constants and ‘stretch’ parameters are collected in the legend to Fig 9. The amplitudes of the kinetic phases correspond to the steady-state turnover populations of the E_n states, and of course vary widely depending on P(N₂) and the electron flux (see Fig 4). In the sample represented by Fig 9 these correspond to: E₀ ~ 8%; E₂ ~ 16%; E₄(4H) ~8%; E₄(2N2H) ~ 12 %. Thus, in this sample the two states linked by the re/oa equilibrium correspond to fully ~ 20% of the MoFe protein. The remainder of the MoFe protein is in EPR-silent E_n, n = odd states, ~ 56%; these cannot relax to E₀ during cryoannealing.³⁵

Figure 9.

Time courses of four EPR detected states during −50 °C cryoannealing of WT low P(N₂) ~ 0.05 atm turnover in H₂O. The data colors correspond to those in the kinetic scheme (Top) and the lines correspond to fits to that scheme as previously described.¹⁰ Stretched exponential parameters of the first two fast steps are τ₁ = 43 min, m₁ = 0.79 and τ₂ = 6 min, m₂ = 0.8, the third slow step fitted as exponential with τ₃ = 330 min. Intensities of E₂ and E₀ states obtained and quantitated with previously described procedures;¹⁵ intensities of E₄ states converted to concentration units by scaling three-step kinetic scheme for corresponding decays. EPR conditions: for E₄(4H) spectra the same as in Fig 4; for E₄(2N2H), E₂ and E₀ the same as used in previous studies.¹⁵

The kinetic coupling of the loss of E₄(2N2H) through oa of H₂, with the formation of E₄(4H) directly establishes the operation of the re/oa mechanism and its kinetic reversibility. The above demonstration that oa of D₂ by E₄(2N2H) in H₂O yields E₄(2D⁻;2H⁺) with exchange-inert deuteride bridges then shows that in this case each of the two steps by which this WT mixed-isotope intermediate relaxes to E₀ would involve protonation of the deuteride bridges by protons from solvent, generating 2HD, confirming that the re/oa mechanism explains the final Key Constraint (ii a), Chart 1, as proposed.^6,7

Thermodynamic reversibility of the re/oa equilibrium during N₂ reduction by WT MoFe

The mere observation of both E₄(4H) and E₄(2N2H) in samples freeze-quenched during N₂ fixation (Fig 4) establishes that the equilibrium constant for re/oa, K_re = k_r/k_b (Fig 3; Chart 3), is small. The cryoannealing experiments only give k_b, not k_r, but as a rough quantitative estimate of K_re, we assign the relative concentrations of E₄(2N2H) and E₄(4H) during turnover under low N₂ partial pressure to the zero-time values determined by the fit to the annealing kinetics (Fig 9), E₄(2N2H)/E₄(4H) ~3/2, in keeping with the conclusion (vide supra) that the forward and reverse steps of the re/oa equilibrium are rapid compared to other steps in the catalytic cycle (Fig 1). In this case we can rewrite eq 2 to approximate K_re as,

$$K_{re}=\frac{E_4\left(2N2H\right)}{E_4\left(4H\right)}\cdot \frac{P\left(H_2\right)}{P\left(N_2\right)}$$ (3)

The N₂ partial pressure is fixed by the experimental conditions; earlier observations suggest that saturating concentrations of H₂ are formed during turnover under all P(N₂), which suggests an effective P(H₂) ~ 1atm.¹⁵ As a result, one obtains:

$$K_{re}~\left(~\frac{3}{2}\right)\ast \left(\frac{~1}{0.05}\right)~30$$ (4)

$${\Delta }_{re}{G}^0=-RT\ln \left(K_{ro}\right)~-2kcal∕mol$$ (5)

The LT kinetic measurements likewise yielded values for the re process: K_re ~ 0.7 and Δ_reG⁰ ~ +0.2 kcal/mol.^2,4 Given the difference in methodologies – direct observation of species in equilibrium in the present study, analysis of turnover kinetics in the former – and the differences in origin of the MoFe proteins – Azotobacter vinelandii in the present study and Klebsiella pneumoniae in the former - we consider the measurements to be in excellent agreement: nitrogenase catalysis, driven by the re of H₂, turns the highly endothermic first step in the reduction of the N₂ triple bond, (to the diazene level) into the essentially thermoneutral re/oa equilibrium conversion of Fig 3.

Conclusions/Summary

The re/oa mechanism for N₂ reduction by nitrogenase postulates that the reduction of the N₂ triple bond to a 2N2H (diazene) level is driven by the obligatory formation of one H₂ for each N₂ reduced. This proposal was based on identification of a FeMo-co S = ½ state trapped during turnover of the α-70^Val→Ile MoFe protein as the Janus intermediate: the E₄(4H) state, which stores four reducing equivalents as two [Fe-H-Fe] bridging hydrides. Once this identification is made, then the LT E₄(4H)↔E₄(2N2H) equilibrium (Fig 1) ceases to be a ‘mystery’: the connection with the organometallic chemistry of metal-dihydride complexes identifies this process as the mechanistically coupled reductive elimination (re) of H₂ (Fig 3). The H₂ formed during re carries away two of the four reducing equivalents stored in E₄(4H) and drives the reaction, while the metal-ion core of FeMo-co becomes activated to reduce N₂ through the simultaneous acquisition of two reducing equivalents. The mechanism was first supported by its proposed explanation of the Key Constraints of Chart 1, and then two new predictions regarding turnover by WT MoFe were promptly verified,^14,15 adding direct experimental support.

As we now summarize, the present report establishes the presence of Janus in WT enzyme, its participation in N₂ reduction, and the operation of the re/oa equilibrium during N₂ reduction. Experiments on WT enzyme further show that this mechanism satisfies all the Key Constraints imposed on N₂ reduction by decades of careful study by others summarized in Chart 1² and that the re/oa equilibrium in WT MoFe is indeed kinetically and thermodynamically reversible, thereby establishing the role of the re/oa equilibrium in N₂ reduction (Fig 3).^6,7

(i) EPR/ENDOR, and photophysical measurements establish that the intermediate with g₁ = 2.15 trapped during nitrogen fixation by WT MoFe in fact is the E₄(4H) Janus intermediate of N₂ reduction, which has accumulated four reducing equivalents stored as [Fe-H-Fe] bridging hydrides, whose properties are identical to those of the Janus intermediate first trapped in the α-70^Val→Ile MoFe variant. This observation thereby establishes the freeze-trapped α-70^Val→Ile Janus intermediate as a reliable guide to mechanism, but most importantly, enables direct observation of the participation of the Janus intermediate in N₂ reduction and the re/oa process during catalysis by the WT enzyme. The new findings and conclusions that build on this foundation are summarized next.

(ii) Examination of the isotopic composition of WT Janus during turnover in H₂O buffer under D₂ (or D₂O buffer under H₂) establishes that oa of D₂ from the gas phase by E₄(2N2H) accumulates E₄(2D⁻;2H⁺), with two [Fe-D-Fe] bridges and two bound protons from solvent (or E₄(2H⁻;2D⁺)) (Fig 3), and that the bridging hydrides/deuterides do not exchange with solvent during turnover. This demonstrates experimentally that the re/oa mechanism accounts for the longstanding Key Constraints on mechanism, Chart 1, (i),(ii)b,c.

(iii) The observation and successful modeling of the entire relaxation pathway of WT enzyme by which E₄(2N2H) relaxes to the resting-state, E₀, including the oa of H₂, to form Janus with release of N₂, and two subsequent steps of hydride protonation each with release of H₂, Fig 1, eq 2, demonstrates the kinetic reversibility of the re/oa equilibrium, Fig 3, Chart 3, and shows that re/oa satisfies the last of the Key Constraints, (ii)a, in addition to confirming that it satisfies (ii)c.

(iv) An estimate of the free energy for the re of H₂ by FeMo-co that has accumulated four reducing equivalents, with reduction of N₂ to generate a 2N2H-level species (Fig 3) in WT enzyme, quantifies the thermodynamic reversibility of the first step in the reduction of N₂ by nitrogenase: this reaction is essentially thermoneutral, Δ_reG⁰ ~ −2 kCal/M, whereas direct hydrogenation of gas-phase N₂ is highly endergonic.¹⁶

In summary, in this report we have described the central mechanistic steps by which the WT nitrogenase carries out one of the most challenging chemical transformation in biology, the reduction of the N≡N triple bond.