Unification of Reaction Pathway and Kinetic Scheme for N2 Reduction Catalyzed by Nitrogenase Dmitriy Lukoyanova, Zhi-Yong Yangb, Brett M

Home , Nitrogenase

Unification of reaction pathway and kinetic scheme for N2 reduction catalyzed by nitrogenase Dmitriy Lukoyanova, Zhi-Yong Yangb, Brett M. Barneyb,1, Dennis R. Deanc, Lance C. Seefeldtb, and Brian M. Hoffmana,2

aDepartment of Chemistry, Northwestern University, 2145 Sheridan Road, Evanston, IL 60208; bDepartment of Chemistry and Biochemistry, Utah State University, 0300 Old Main Hill, Logan, UT 84322; and cDepartment of Biochemistry, Virginia Tech, 110 Fralin Hall, Blacksburg, VA 24061

Contributed by Brian M. Hoffman, February 7, 2012 (sent for review January 10, 2012)

Nitrogenase catalyzes the reduction of N2 and protons to yield two at the fifth hydrogenation step (8). However, this conclusion left NH3 and one H2. Substrate binding occurs at a complex organo- many questions unresolved, in particular the correspondence be- metallocluster called FeMo-cofactor (FeMo-co). Each catalytic cycle tween intermediates formed along the reaction pathway and the involves the sequential delivery of eight electrons/protons to this LT kinetic scheme for the accumulation of eight reducing equiva- cluster, and this process has been framed within a kinetic scheme lents during catalysis. developed by Lowe and Thorneley. Rapid freezing of a modified Intermediates formed during turnover have been trapped by nitrogenase under turnover conditions using diazene, methyldia- freeze-quench methods and studied by paramagnetic resonance ¼ zene (HN N-CH3), or hydrazine as substrate recently was shown techniques. The resting-state FeMo-co of E0 exhibits an EPR- ¼ 1 I ¼ 3 to trap a common S 2 intermediate, designated . It was further active, S 2 spin state. When nitrogenase is freeze-quenched concluded that the two N-atoms of N2 are hydrogenated alter- during turnover, the resting-state EPR spectrum partially or fully nately (“Alternating” (A) pathway). In the present work, Q-band disappears as FeMo-cofactor accumulates electrons. Accumula- 95 ¼ CW EPR and Mo ESEEM spectroscopy reveal such samples also tion of an even number of electrons generates En, n even, in- contain a common intermediate with FeMo-co in an integer-spin termediates with EPR-active, odd-electron (half-integer spin; state having a ground-state “non-Kramers” doublet. This species, 1 3 Kramers) FeMo-co states (S ¼ 2 ; 2) (3, 4). Such states have been designated H, has been characterized by ESEEM spectroscopy using , , observed upon freeze-quench of wild-type and amino-acid substi- 14 15 1 2 CHEMISTRY a combination of N isotopologs plus H isotopologs of tuted MoFe proteins during turnover with a variety of different H methyldiazene. It is concluded that: has NH2 bound to FeMo-co substrates (3, 4). However, the signals from Kramers forms of and corresponds to the penultimate intermediate of N2 hydroge- FeMo-co in the quenched samples never quantitate to the total nation, the state formed after the accumulation of seven electrons/ FeMo-co present, indicating that apparently EPR-silent states of I protons and the release of the first NH3; corresponds to the final FeMo-co must also exist. These silent MoFe protein states con- intermediate in N2 reduction, the state formed after accumulation tain FeMo-co with an even number of electrons, and correspond of eight electrons/protons, with NH still bound to FeMo-co prior to ¼ ¼ 2 þ 1 ¼ 0–3 3 to En, n odd (n m , m ) intermediates in the LT release and regeneration of resting-state FeMo-co. A proposed uni- scheme. These quenched samples may include states with dia- fication of the Lowe-Thorneley kinetic model with the “prompt” magnetic FeMo-co, but Mössbauer studies indicated the presence BIOCHEMISTRY alternating reaction pathway represents a draft mechanism for of reduced FeMo-co in integer-spin (S ¼ 1; 2, .....), “non-Kra- N2 reduction by nitrogenase. mers (NK)” states (9). Although NK-EPR signals at conventional microwave frequencies are well known for other enzymes (10, EPR ∣ ESEEM ∣ non-Kramers 11), until now, no EPR signal from an integer-spin form of FeMo-co has been detected. itrogen fixation—the reduction of N2 to two NH3 molecules Recently, rapid freezing during turnover of a doubly-substi- N—is catalyzed by the enzyme nitrogenase according to the tuted nitrogenase MoFe protein (α-70Val→Ala, α-195His→Gln), limiting stoichiometry (1, 2): which favors reduction of large nitrogenous substrates (3, 4), with diazene, methyldiazene (HN¼N-CH3), or hydrazine as substrate − þ N2 þ 8e þ 8H þ 16MgATP → 2NH3 þ H2 þ 16MgADP 1 was shown to trap a common S ¼ 2 intermediate (denoted I) (8). 95 þ 16Pi [1] Here, Q-band CW EPR and Mo ESEEM (electron spin-echo envelope modulation) spectroscopy reveal that these samples H The Mo-dependent enzyme studied here (2–4) consists of two also contain a second common intermediate, denoted , one component proteins, denoted the Fe protein and the MoFe pro- that is observed in which FeMo-co is in an EPR-active integer- spin state with a ground-state NK doublet. The NK-EPR signal tein. The former delivers electrons one-at-a-time to the MoFe 95 protein, where they are utilized at the active site iron-molybde- in samples prepared with Mo-enriched FeMo-co and with 14 15 num cofactor ([7Fe-9S-Mo-C-R-homocitrate]; FeMo-co) to re- ; N-enriched substrates has been studied by pulsed-EPR mea- duce substrate (2, 3). In the Lowe-Thorneley (LT) kinetic surements of the nuclear modulation of the electron spin-echo scheme for N2 reduction by nitrogenase (2, 5, 6), the eight steps (ESE) amplitude, denoted NK-ESEEM (12). These measure- 1 – H of electron/proton delivery implied by Eq. are denoted En, ments allow us to infer the state of the N N bond in and where n ¼ 0 to 8, with E0 representing the resting-state enzyme. the relationship of H to intermediate, I. Most importantly, the It is known that N2 binds only after the MoFe protein accumu- analysis allows us to assign the correspondence of both intermedi- lates three or four electrons (E3 or E4 states), with N2 reduction ates with specific En states, and to infer how the hydrogenated then proceeding along a reaction pathway that comprises the sequence of intermediate states generated as N2 bound to FeMo-co −∕ þ Author contributions: D.L., Z.-Y.Y., B.M.B., D.R.D., L.C.S., and B.M.H. designed research; undergoes six steps of hydrogenation (e H delivery to sub- D.L. performed research; Z.-Y.Y. and B.M.B. contributed new reagents/analytic tools; strate) (1–4, 7). Recently, we concluded that nitrogenase follows D.L. analyzed data; and D.L., Z.-Y.Y., B.M.B., D.R.D., L.C.S., and B.M.H. wrote the paper. an “alternating” (A) reaction pathway, in which the two N’s are The authors declare no conflict of interest. hydrogenated alternately, with a hydrazine-bound intermediate 1Present address: Department of Bioproducts and Biosystems Engineering, University of formed after four steps of hydrogenation, and with cleavage of Minnesota, St. Paul, MN 55108. the hydrazine N-N bond and liberation of the first NH3 only 2To whom correspondence should be addressed. E-mail:[email protected].

www.pnas.org/cgi/doi/10.1073/pnas.1202197109 PNAS ∣ April 10, 2012 ∣ vol. 109 ∣ no. 15 ∣ 5583–5587 Downloaded by guest on October 2, 2021 reaction intermediates, diazene and hydrazine, join the N2 reduc- of a field establishes the coupling and introduces modulation, tion pathway. These conclusions integrate the sequence of N2 with a depth that increases quadratically with B (15). reduction intermediates (the reaction “pathway”) with the LT ki- As illustrated in Fig. 2A, the NK-ESEEM time-waves for the netic scheme for the accumulation of the eight reducing equiva- NK intermediates trapped during turnover with the corresponding 14 15 lents (and protons), thereby generating a draft mechanism for Nand N isotopologs of N2H2,N2H4,andHN2CH3 substrates nitrogen fixation by nitrogenase. are identical at all fields, indicating that a common intermediate, denoted H, is trapped during turnover with all three substrates. Results Fig. 2A, Left and Fig. 2C, Left further show that 95Mo enrichment → The EPR spectra of the doubly substituted α-70Val Ala, of α-70Val→Ala, α-195His→Gln MoFe protein produces significant → α-195His Gln MoFe protein trapped during turnover in the pre- change of the NK-ESEEM time-wave. This establishes that the sence of hydrazine, diazene, and methyldiazene show a significant NK-EPR signal of H arises from the Mo-containing FeMo-co ¼ 3 loss of the S 2 resting state of FeMo-co and the appearance of a in an integer-spin state, and not the all-iron electron-transfer- 1 S ¼ 2 signal from intermediate I in the vicinity of g ¼ 2, along active P cluster also present in the MoFe protein, or even the 1þ with the spectrum from the ½4Fe4S cluster of the Fe protein [4Fe-4S] cluster of the Fe protein. The absence of any signal in 1 14 15 (Fig. 1). Recent H, ; N ENDOR and HYSCORE spectro- Fig. 1 from the P cluster further indicates that all electrons deliv- scopic experiments demonstrated that I represents a late stage ered to MoFe protein of H during turnover reside on FeMo-co. of nitrogen fixation, when the first ammonia molecule already The “turn-on” of nuclear modulation for H with applied field is ¼ 2 has been released (8), and only a [NHx](x or 3) fragment quite gradual in comparison with that observed for the carboxy- of substrate is bound to FeMo-co (8). late-bridged diiron centers (16). In addition to the hyperfine cou- We here note that samples freeze-trapped during turnover pling and g-value along the unique magnetic axis of the center, with all three substrates also show an additional broad EPR signal Ajj and gjj, there are three factors that influence the field depen- at low field in Q-band spectra collected at a temperature of 2 K dence of the modulation depth (15). First, the earlier diiron cen- ∼ 9 (Fig. 1). This signal begins near zero-field and extends beyond ter measurements were performed at X band (ve GHz), 3 5,000 G, where the signals from S ¼ 2 FeMo-co in both resting whereas the present ones employ a fourfold higher frequency ∼ 35 state and trapped high-spin intermediates (13) appear. The (ve GHz). According to our analysis (15), this causes a four- low-field signal arises from an integer-spin system (S ≥ 2) (10) fold decrease in the rate at which the effective hyperfine coupling ~ that exhibits a ground-state non-Kramers doublet that is split in (Aeff ) and thus the modulations increases with field. Second, the zero applied field by an energy, hΔ (h, Planck’s constant) (14), in earlier measurements were performed with the microwave field υ ¼ 35 the range of the microwave quantum, e GHz; the breadth parallel to the external field, whereas the current ones are per- of the signal indicates that Δ exhibits a considerable distribution formed in perpendicular mode, which requires a roughly three- ~ in values. fold higher field to achieve the same Aeff (17). Finally, the diiron ESEEM can be used to characterize a non-Kramers doublet centers exhibit a total spin state S ¼ 4 (16); assigning intermedi- (NK-ESEEM) (15); we begin by discussing the ESEEM time- ate H a total spin of S ¼ 2 would further lower the rate at which ~ waves. Fig. 2A presents representative 35 GHz (2 K) three-pulse Aeff increases with field by an additional factor of four (15). This NK-ESEEM time-waves collected at several relatively low fields value of S for H is supported by the finding from Q-band EPR from the nitrogenase NK intermediates generated with isotopo- spectroscopy (Fig. 1) that Δ ∼ 35 GHz for H, not approximately logs of the three substrates. The electron spin-echo (ESE) ampli- 9 GHz for the diiron centers, for it is generally expected that Δ tude from the NK intermediates is maximum at zero applied increases as S decreases* (11). magnetic field (B) and decreases very slowly with increasing field, Figs. 2A, Right and 2B, Left show that intermediates prepared being observable at all fields up to approximately 5;000 G, where with 14N and 15N substrates yield quite different NK-ESEEM 3 signals from S ¼ 2 FeMo-co begin. This slow decrease of echo time-waves, demonstrating that the modulation arises from a sub- intensity with field contrasts sharply with the rapid decrease strate fragment bound to FeMo-co. The state of the N-N bond H for carboxylate-bridged di-ferrous centers (S ¼ 4), where the in is revealed by the behavior of the common intermediate H echo (at X band) vanished by approximately 100 G (16). As illu- upon isotopic substitution of the three nitrogenous substrates, in particular selective labeling of methyldiazene. The time-waves strated in Fig. 2A, Left (Upper), at the lowest applied magnetic 14 field (B ¼ 18 G) none of the nitrogenase intermediates show nu- of Figs. 2A, Right show that the N modulation is abolished and 15 clear modulation of the ESE envelope. As the field is increased, replaced with N modulation not only when both nitrogens of 14 14 15 nuclear modulation from protons first appears (B ¼ 296 G), then N2H2 and N2H4 are substituted with N, but also when 14 15 14 modulation associated with hyperfine-coupled N(B¼ 362 G). H- N¼ N-CH3 is used as substrate; furthermore, there is no 2 14 14 The appearance of modulation with increasing field reflects the added H modulation from H- N¼ N-CD3. The absence of 14 14 14 fact that at zero applied field the electron and nuclear spins are modulation from the second NofH- N¼ N-CH3 or from 2 14 14 uncoupled and so there can be no ESE modulation. Application HofH- N¼ N-CD3 indicates that the N–N bond has been cleaved prior to the formation of H, which thus contains an NHx fragment bound to FeMo-co. This conclusion parallels 1 the conclusion that the common S ¼ 2 intermediate, I, formed during turnover with the same substrates (see Fig. 1) likewise is formed after the first ammonia is released. H To characterize the NHx substrate fragment of , frequency- domain NK-ESEEM spectra were obtained by Fourier transform of the time-waves. At low applied fields the electron-nuclear coupling for 14N(I¼ 1) is effective in introducing ESEEM, but the effects of hyperfine couplings on the frequency-domain spectrum are small, and a 14N frequency spectrum exhibits transitions at the three pure nuclear–quadrupole frequencies, α 70Val→Ala α 195His→GIn Fig. 1. Q-band CW EPR spectrum of - , - MoFe protein in denoted (¯vþ > ¯v− > ¯v0; ¯vþ ¼ ¯v− þ ¯v0), with relative intensities 14 resting state (S ¼ 3∕2) and trapped during turnover with N2H4. Conditions: microwave frequency, 35.015 GHz (resting), 35.022 GHz (turnover); modulation amplitude, 2 G; time constant, 128 ms; field sweep, 67 G∕s; T ¼ 2 K. *As a rule, S ¼ 1 states do not exhibit a NK doublet.

5584 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1202197109 Lukoyanov et al. Downloaded by guest on October 2, 2021 A determined by the relative orientations of the quadrupole and zero-field splitting tensors (15). The frequencies yield the quadrupole coupling constant, e2qQ∕h and asymmetry parameter, η (usual symbols). In contrast, the low-field measurements of nuclei 1 1 15 with I ¼ 2 ( H, N) exhibit only a single line at the nuclear Lar- mor frequency. Examination of spectra taken at multiple fields for H prepared 14 14 by turnover with N2H2 discloses two N peaks whose frequencies 0.71 and 1.92 MHz change little with field from approximately 250 G up to at least 1,200 G, and that can be assigned as part of the 14N quadrupole triplet. As expected from the change in the time-waves upon generating H with 15N-labeled substrate (Fig. 2A, Right,2B), the 14N quadrupole peaks are abol- 15 ished in the NK-ESEEM spectra of H prepared with N2H2, and replaced by a single peak at the 15N Larmor frequency (Fig. 2B). As the frequencies of the three peaks obey the relationship, ¯vþ ¼ ¯v− þ ¯v0, there are two options for the frequency of the un- B observed peak: 1.21 and 2.63 MHz. This peak could go unob- served either because it lies at 1.21 MHz, where features presumed to arise from the molybdenum isotopes with nonzero nuclear spin could obscure the 14N signal, or because the relative orientation of quadrupole and zero-field splitting tensors causes it to have negligible intensity (15). Assignment of the quadrupole frequencies as ¯v0 ¼ 0.71 MHz, ¯v− ¼ 1.21 MHz, and ¯vþ ¼ 1.92 MHz gives the quadrupole coupling parameters: e2qQ∕h ¼ 2.08 MHz, η ¼ 0.68; the alternative assignment of the missing 2 peak as ¯vþ ¼ 2.63 MHz gives: e qQ∕h ¼ 3.04 MHz, η ¼ 0.47.

η ∼ 0 CHEMISTRY A cofactor-bound NH3 would have , so the NHx fragment of H cannot be NH3; as bound −NH does not appear on the A pathway, this fragment therefore must be NH2. The frequency-domain NK-ESEEM spectra obtained from H prepared with 95Mo-enriched α-70Val→Ala, α-195His→Gln MoFe protein give information about the environment of the molybdenum of FeMo-co in that intermediate. Fig. 2C shows that 95Mo enrichment introduces a doublet centered at approximately

0.3 MHz in the low-field spectrum, as well as an additional, un- BIOCHEMISTRY resolved broad signal between one and two MHz at higher fields. C As the doublet is barely observable in the natural-abundance samples and the broad signal is not seen, the new signals arise from 95Mo of FeMo-co. Although the 14N signal is not detectable below 250 G, the 95Mo signal clearly is present in spectrum at 150 G, and likely could be detected even at lower fields but for overlap with the strong 1H signal. This difference can be ex- plained by a larger hyperfine coupling to 95Mo in comparison with 14N (15), combined with differences in the dependence of 5 95 the spectrum on magnetic field for nuclei with I ¼ 2 ( Mo) and I ¼ 1 (14N) that will be discussed elsewhere. The low-frequency 95Mo doublet observed is provisionally assigned to two pure-quadrupole transitions. The quadrupole in- 95 5 teraction of a Mo (I ¼ 2) in zero applied field splits the 2 þ 1 ¼ 6 I mI sublevels into three doublets for each electron- 14 → ¼ 1 Fig. 2. (A): Three-pulse ESEEM traces for NK intermediate H of α-70Val Ala, spin manifold. As with N(I ), this yields three pure-nuclear His→Gln 14 ¯ ¯ ¯ α-195 MoFe protein under turnover. (A, Left) Time domain for N2H2 quadrupole transitions, which we denote as v3 > v2 > v1. Matrix 14 95 14 (black), N2H4 (red) substrates, and for Mo-enriched MoFe with N2H4 diagonalization of the full electron-nuclear Hamiltonian and ex- (blue). Most spectra taken with τ selected to suppress the 1H response; none- amination of the calculated modulation depth parameters (12) 1 theless, residual H signals typically are observed. Conditions: microwave fre- suggests that the highest frequency peak, ¯v3 will generally be 14 14 95 quency, 34.756 GHz ( N2H2), 34.767 GHz ( N2H4), 34.756 GHz ( Mo-FeMo- of low intensity, so we further assign the observed doublet as 14 π∕2 ¼ 50 co, N2H4); ns, 30 ns time steps; repetition time 2 ms, 50 shots/point ¯ ∼ 0 2 ¯ ∼ 0 35 2 ∕ 14 – ¼ 2 v1 . MHz, v2 . MHz. This assignment yields, e qQ (10 ms, 10 shots/point for N2H2), 50 150 scans; T K. (A, Right) Time- ∼ 1 2 η ∼ 0 2 H 14 ¼14 14 ¼14 h . MHz, . , values comparable to those found for waves for trapped with: NH NH (black), NH NCD3 (green), þ 14 14 15 15 15 14 ðM Þ2ðMoO4Þ salts (18). The higher-frequency feature seen at NH2- NH2 (red), NH¼ NH (blue), NH ¼ NCH3 (magenta). Condi- tions:asfor(A, Left), except for microwave frequency, 34.772 GHz higher field presumably arises through the influence of the larger 14 14 15 15 15 14 95 ( NH¼ NCD3), 34.764 GHz ( NH¼ NH), 34.725 GHz ( NH¼ NCH3); Mo hyperfine interaction and nuclear spin. 140–500 scans. Time-waves are shown after decay baseline subtraction. (B): Field dependence of three-pulse ESEEM (B, Left) time and (B, Right) Discussion 14 15 frequency traces of H prepared with N2H2 (black) and N2H2 (red). Con- The key experimental findings of this report can be summarized ditions:asfor(A). Triangles represent suppressed frequencies n∕τ, n ¼ 1; 2, ... as follows. 14 (C): Time and frequency domain traces for H formed with N H4 substrate: 2 95 natural abundance (black) and 95Mo-enriched (red) α-70Val→Ala, α-195His→Gln i. CW EPR and Mo NK-ESEEM spectroscopy reveal that → → MoFe protein. Conditions:asfor(A). samples of the α-70Val Ala, α-195His Gln substituted nitrogen-

Lukoyanov et al. PNAS ∣ April 10, 2012 ∣ vol. 109 ∣ no. 15 ∣ 5585 Downloaded by guest on October 2, 2021 ase MoFe protein trapped during turnover in the presence of activated by two reducing equivalents and two protons, presum- hydrazine, diazene, and methyldiazene each contain a com- ably in the form of one hydride and one proton. mon intermediate, H, with FeMo-co in an even-electron, in- Recently, we concluded that the subsequent catalytic stages teger-spin state (plausibly S ¼ 2) characterized by a low-lying follow an “Alternating” (A) pathway, in which the two N atoms NK doublet. The NK-ESEEM measurements yield quadru- of N2 are hydrogenated alternately and the first NH3 is released 95 pole coupling parameters for the Mo of FeMo-co in H. after N2 has been hydrogenated five times (Fig. 3), rather than a 14 15 ii. N∕ N NK-ESEEM of H formed with substrate isotopologs “Distal” pathway in which a specific nitrogen of N2 is hydroge- indicates that a nitrogen of the substrate is directly bound to nated three times, then released as NH3. We also noted (4) that FeMo-co of α-70Val→Ala, α-195His→Gln MoFe protein. the A pathway offers multiple ways to deal with the two reducing 14 iii. The absence of modulation from either a second N of sub- equivalents that remain on FeMo-co after N2 binding. In the 15 14 2 strate, in particular from H- N¼ N-CH3, or from Hof “Prompt” (P) pathway, when N2 binds to FeMo-co it is “nailed 14 14 H- N¼ N-CD3 indicates that H is formed after cleavage down” by prompt hydrogenation, Fig. 3, with N2 binding, H2 loss, of the N–N bond of N2H4 bound to FeMo-co and loss of and reduction to diazene all occurring at the E4 kinetic stage of NH3. Quadrupole coupling parameters for this cofactor- the LTscheme. In the “Late” (L) model, the two reducing equiva- bound NHx fragment indicate it is not NH3, whose threefold lents associated with FeMo-co after N2 binding are delivered to symmetry would yield η ∼ 0, and thus must be NH2. substrate after one or more steps of hydrogenation. In the limiting version of Fig. 3, this comes after hydrazine is formed. This With these findings, it is possible to propose a complete uni- only occurs at the final, E8 kinetic stage of the LT scheme, and in fication of reaction pathway and LTscheme. In the LTscheme the this case N–N bond cleavage and release of the first NH3, fol- MoFe protein is optimally activated for N2 binding at the E4 lowed by conversion of the remaining bound NH2 to NH3 and stage, in which MoFe protein has accumulated four reducing its release all occur at E8. Thus, as illustrated, a given state of equivalents and four protons. Our studies (19) further show N2 hydrogenation can correspond to a different En state along that the four reducing equivalents of E4 exist in the form of two the two A pathway branches. (Fe-bridging) hydrides, presumably with an additional two pro- The present results establish the identities of H and I, as well P tons bound to sulfide (20). As illustrated, in Fig. 3, N2 binding as their correspondence with En states along the pathway of to FeMo-co of E4 is accompanied by the loss of two reducing Fig. 3, while unambiguously ruling out the limiting version of the L pathway. (i) As the same intermediate H is formed during equivalents and two protons as H2 (Eq. 1), leaving FeMo-co turnover with the two diazenes and with hydrazine, the diazenes must have catalytically “caught up” to hydrazine, and H must oc- P A Pathway L cur at or after the appearance of a hydrazine-bound intermediate. N N (ii) As noted above, H contains FeMo-co in an integer-spin (NK) E4 N state, and thus corresponds to an E state with n ¼ odd. As H is a H2 N2 N2 H2 N n E4 M common intermediate that contains a bound fragment of sub- M M H - + - + - + strate, it must therefore correspond to E5 or E7. But cannot H H (H )2 (H )2 H H e-/H+ be either E5 or E7 on the L path or E5 on the P pathway, for they NH all come before hydrazine appears (Fig. 3). Indeed, in the L path- E NH 4 N way of Fig. 3, the MoFe protein is in the E8 state at the binding of N E5 M hydrazine and after, and thus FeMo-co is in a Kramers (half- M - + H H+ integer-spin; odd-electron) spin state. As a result, no state on H e-/H+ NH this pathway can correspond with the non-Kramers intermedi- NH NH ate, H.(iii) We are left to conclude that H corresponds to the NH E − M 6 ½NH2 -bound intermediate formed by N–N bond cleavage at M P H- + the E7 stage on the pathway. By parallel arguments, the only - + H - + 1 e /H e /H ¼ I NH2 possible assignment for the S 2 state , which we showed earlier NH2 – I NH to occur after N N bond cleavage (8), is as E8: must correspond E5 NH E7 M to the final state in the catalytic process (Fig. 3), in which the NH3 M - + product is bound to FeMo-co at its resting oxidation state, prior - + H H e /H e-/H+ to release and regeneration of the resting-state form of the NH2 NH2 cofactor. NH2 NH2 E6 Although the above discussion focuses on the limiting version M M L - + of the pathway, the stabilization of bound substrate by prompt - + H H e /H hydrogenation along the P pathway for N2 reduction and the NH NH3 3 E8 respective correspondences of H and I with E7 and E8 of this NH2 NH2 L E7 H pathway greatly favor it over a version of the pathway in which M M + the reducing equivalents from E4 are transferred to substrate at I H e-/H+ an intermediate stage of substrate reduction, E5–7 (Fig. 3). The NH3 I E8 trapping of a product-bound intermediate is analogous to the M trapping of a bioorganometallic intermediate during turnover of → NH3 α 70Val Ala MN the - MoFe protein with the alkyne, propargyl alcohol; this intermediate was shown to bind the allyl alcohol alkene pro- Fig. 3. Integration of LT kinetic scheme with Alternating (A) pathways for duct of reduction (21). N2 reduction. Note: M denotes FeMo-co in its entirety and substrate-derived A final mechanistic question to be addressed is: how do the species are drawn to indicate stoichiometry only; nothing is implied as to hydrogenated reaction intermediates, diazene and hydrazine, join mode of substrate binding. Bold arrows indicate transfer to substrate of hy- the N2 reduction pathway? Key to this issue is the finding that dride remaining after N2 binding in E4; P represents “Prompt” and L “Late” H2 inhibits the reduction of diazene (22), but not hydrazine transfer; as L pathway is ruled out by experiment, it is in gray. En states, n ¼ even, are Kramers states; n ¼ odd are non-Kramers. MN denotes rest- (23). We take that to mean that H2 and diazene bind competi- ing-state FeMo-co. Individual charges are not assigned to M and any sub- tively, as do H2 and N2 (2, 5, 6), but that this is not the case strate fragment, but they sum to the charge on resting FeMo-co. for hydrazine. Provisionally, we further take the simplest view,

5586 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1202197109 Lukoyanov et al. Downloaded by guest on October 2, 2021 P A Pathway is proposed for another two-electron substrate, C2H2 (2, 5, 6), and joins the N2 pathway at a stage corresponding to E7 in the N2 reduction scheme. NH H2 E E4 NH 2 The proposed identification of nitrogenous intermediate states NH NH M H and I and their correlation with LTstages E7 and E8, plus their M - H H+ incorporation in the “Prompt-Alternating” pathway of Fig. 3, re- e-/H+ present a unification of the nitrogenase reaction pathway and the NH2 LT kinetic scheme into a draft mechanism for N2 reduction by E5 NH nitrogenase. We view this scheme as a “draft” because of the nu- M e-/H+ merous additional questions it poses. Among the key issues are “ ” NH2 the electron/proton inventory that relates the redox levels of

E6 NH2 the substrate and metal-ion core, along with the site of protona- M tion, in individual En states subsequent to N2 binding. For exam- e-/H+ E1 NH2 ple, consider reduction of N2 to diazene at the E4 stage. This is M NH3 NH2 shown as occurring by hydride transfer followed by protonation, + H but there are other alternatives. Likewise, it is postulated in Fig. 3 E NH2 7 that E5 contains a diazenido ligand, yet there are alternative ways M H to form this species, as well as other formulations of that inter- e-/H+ NH3 mediate. Framed differently, the LT scheme focuses on stable in- E8 I M termediates, and a full mechanism would include any transient

NH3 states formed on the way to such intermediates. MN Materials and Methods Fig. 4. Proposed pathways for reduction of N2H2 and N2H4. Sample preparation followed freeze-quench procedures previously described (22, 24, 25); the 35 GHz CW (26) and pulsed (27) EPR spectrometers also have that under turnover, diazene and hydrazine each join the N2 re- been described. Fourier transforms were performed with Bruker software,

duction pathway at their own characteristic entry point, and then Win-EPR. CHEMISTRY proceed to generate both H and I. These two assumptions alone lead to the scheme of Fig. 4. In this scheme, diazene binds to E2 ACKNOWLEDGMENTS. This work has been supported by the National Institutes with the release of H2, and enters the N2 pathway as the final of Health (HL 13531, BMH; GM 59087, DRD and LCS) and National Science form of the E4 state. In contrast, N2H4 instead binds to E1,as Foundation (MCB 0723330, BMH).

1. Christiansen J, Dean DR, Seefeldt LC (2001) Mechanistic features of the Mo-containing 14. Griffith JS (1963) Spin Hamiltonian for even-electron systems having even multiplicity. nitrogenase. Annu Rev Plant Physiol Plant Mol Biol 52:269–295. Phys Rev 132:316–319. 2. Burgess BK, Lowe DJ (1996) Mechanism of molybdenum nitrogenase. Chem Rev 15. Hoffman BM (1994) ENDOR and ESEEM of a non-Kramers doublet in an integer-spin

96:2983–3011. system. J Phys Chem 98:11657–11665. BIOCHEMISTRY 3. Seefeldt LC, Hoffman BM, Dean DR (2009) Mechanism of Mo-dependent nitrogenase. 16. Sturgeon BE, et al. (1997) Non-Kramers ESEEM of integer-spin diferrous carboxylate- Annu Rev Biochem 78:701–722. bridged clusters in proteins. J Am Chem Soc 119:375–386. 4. Hoffman BM, Dean DR, Seefeldt LC (2009) Climbing nitrogenase: Towards the mechan- 17. Doan PE, Hoffman BM (2000) How important is parallel-mode EPR in electron spin ism of enzymatic nitrogen fixation. Acc Chem Res 42:609–619. echo envelope modulation studies of non-Kramers systems? Inorg Chim Acta – 5. Thorneley RNF, Lowe DJ (1985) Kinetics and mechanism of the nitrogenase enzyme 297:400 403. system. Molybdenum Enzymes, ed TG Spiro (Wiley-Interscience, New York), Metal Ions 18. Forgeron MAM, Wasylishen RE (2008) Molybdenum magnetic shielding and quadru- 95 in Biology, 7, pp 89–116. polar tensors for a series of molybdate salts: a solid-state Mo NMR study. Phys Chem – 6. Wilson PE, Nyborg AC, Watt GD (2001) Duplication and extension of the Thorneley and Chem Phys 10:574 581. 19. Lukoyanov D, Barney BM, Dean DR, Seefeldt LC, Hoffman BM (2007) Connecting ni- Lowe kinetic model for Klebsiella pneumoniae nitrogenase catalysis using a MATHE- trogenase intermediates with the kinetic scheme for N2 reduction by a relaxation pro- MATICA software platform. Biophys Chem 91:281–304. tocol and identification of the N2 binding state. Proc Natl Acad Sci USA 104:1451–1455. 7. Thorneley RNF, Eady RR, Lowe DJ (1978) Biological nitrogen fixation by way of an 20. Lukoyanov D, Yang Z-Y, Dean DR, Seefeldt LC, Hoffman BM (2010) Is Mo involved in enzyme-bound dinitrogen-hydride intermediate. Nature 272:557–558. hydride binding by the four-electron reduced (E4) intermediate of the nitrogenase 8. Lukoyanov D, et al. (2011) ENDOR/HYSCORE studies of the common intermediate MoFe protein? J Am Chem Soc 132:2526–2527. trapped during nitrogenase reduction of N2H2,CH3N2H, and N2H4 support an alter- 21. Lee H-I, et al. (2004) An organometallic intermediate during alkyne reduction by nating reaction pathway for N2 reduction. J Am Chem Soc 133:11655–11664. nitrogenase. J Am Chem Soc 126:9563–9569. 9. Yoo SJ, Angove HC, Papaefthymiou V, Burgess BK, Muenck E (2000) Moessbauer study 22. Barney BM, et al. (2007) Diazene (HN ¼ NH) is a substrate for nitrogenase: Insights into of the MoFe protein of nitrogenase from Azotobacter vinelandii using selective 57Fe the pathway of N2 reduction. Biochemistry 46:6784–6794. – enrichment of the M-centers. J Am Chem Soc 122:4926 4936. 23. Davis LC (1980) Hydrazine as a substrate and inhibitor of Azotobacter vinelandii 10. Hendrich MP, Debrunner PG (1989) Integer-spin electron paramagnetic resonance of nitrogenase. Arch Biochem Biophys 204:270–276. – iron proteins. Biophys J 56:489 506. 24. Barney BM, et al. (2005) Trapping a hydrazine reduction intermediate on the nitro- 11. Münck E, Surerus K, Hendrich MP (1993) Combining Mössbauer spectroscopy with in- genase active site. Biochemistry 44:8030–8037. – teger spin electron paramagnetic resonance. Methods Enzymol Part C 227:463 479. 25. Barney BM, et al. (2006) A methyldiazene (HN ¼ N-CH3)-derived species bound to the 12. Hoffman BM, et al. (1994) ESEEM and ENDOR magnetic resonance studies of the non- nitrogenase active-site FeMo cofactor: Implications for mechanism. Proc Natl Acad Sci Kramers doublet in the integer-spin diiron(II) forms of two methane monooxygenase USA 103:17113–17118. hydroxylases and hemerythrin azide. J Am Chem Soc 116:6023–6024. 26. Werst MM, Davoust CE, Hoffman BM (1991) Ligand spin densities in blue copper pro- 13. Fisher K, Newton WE, Lowe DJ (2001) Electron paramagnetic resonance analysis of teins by Q-band 1H and 14N ENDOR spectroscopy. J Am Chem Soc 113:1533–1538. different Azotobacter vinelandii nitrogenase MoFe-protein conformations generated 27. Zipse H, et al. (2009) Structure of the nucleotide radical formed during reaction of during enzyme turnover: Evidence for S ¼ 3∕2 spin states from reduced MoFe-protein CDP/TTP with the E441Q-a2b2 of E. coli ribonucleotide reductase. J Am Chem Soc intermediates. Biochemistry 40:3333–3339. 131:200–211.

Lukoyanov et al. PNAS ∣ April 10, 2012 ∣ vol. 109 ∣ no. 15 ∣ 5587 Downloaded by guest on October 2, 2021