Infrared spectroscopy of the nitrogenase MoFe protein under electrochemical control: potential-triggered CO binding

Electrochemical control over nitrogenase allows us to examine electrocatalytic proton reduction and potential-triggered CO inhibition using infrared spectroscopy.


Introduction
Biological dinitrogen xation occurs in diazotrophic bacteria that contain the enzyme nitrogenase. In the Mo-dependent nitrogenase, two protein components work together to achieve the reduction of dinitrogen (N 2 ) to ammonia (NH 3 ). The molybdenum-iron (MoFe) protein (a 2 b 2 ) houses the active site FeMo-cofactor [7Fe-9S-1Mo-C-homocitrate], FeMo-co, where substrates are reduced, and a [8Fe-7S] P cluster, that appears to function as an electron carrier. The Fe protein component (a 2 ) contains a single [4Fe-4S] cluster and functions to deliver one electron to the MoFe protein during a transient association of the two proteins, in a process that is coupled to the hydrolysis of two equivalents of ATP for each electron transferred. 1 The requirement for ATP hydrolysis by this system is energy intensive, comparable to the energy input required in the industrial Haber-Bosch process for N 2 reduction where the energy comes primarily from fossil fuels. The complexity of the fully functioning nitrogenase system makes studies to dene the mechanism of electron transfer and coupling with ATP challenging. The redox level of the MoFe protein is typically controlled indirectly by the ratio of the two component proteins (Scheme 1A), which complicates access to well dened, catalytically relevant, redox states of the MoFe protein. 2,3 As a step toward controlled delivery of electrons to nitrogenase, and as a way to establish potential dependence of catalytically relevant events, it is desirable to gain electrochemical control of the MoFe protein without the Fe protein. In this direction, earlier reports have demonstrated Fe protein-independent electron transfer to variants of the nitrogenase MoFe protein with amino acid substitutions. Introduction of a cysteine-bound Ru photosensitizer at position 158 of the a chain was shown to allow light-driven reduction of protons, acetylene and hydrogen cyanide at very low rates. 4,5 Recently, it was shown that CdS nanorods could be coupled to MoFe protein to enable photocatalytic N 2 reduction to ammonia. 6 Variants of the MoFe protein with amino acid substitutions in the region between the P cluster and FeMo-co have been shown to reduce protons to H 2 , azide to ammonia, and hydrazine to ammonia with low potential Eu II polyaminocarboxylate complexes as reductants. 7 A crystal structure of one of these variants bearing a Tyr to His substitution in the b chain of the MoFe protein (b-98 Tyr/His ) showed only minor structural changes that affected solvent ordering between the P cluster and FeMo-co, but no global structural changes. 8 Here we show that it is possible to use the Eu III/II -ligand redox couples to mediate control by an electrode of wild type and amino acid substituted Azotobacter vinelandii nitrogenase MoFe protein, as evidenced by electrocatalytic proton reduction (Scheme 1B), and potential-dependent interaction of CO with the MoFe protein detected with infrared spectroelectrochemistry.

Results and discussion
Electrochemical and IR spectroelectrochemical cell for mediated electrochemistry of nitrogenase MoFe protein Electrochemical experiments were conducted on samples of nitrogenase MoFe protein (ca. 2 nmol) contained within a layer of the polymer electrolyte, Naon, titrated to pH 7.4 in Tris-HCl buffer (Naon-Tris, see ESI Experimental methods †), in contact with a carbon paper working electrode, as shown in Fig. 1. Naon allows movement of small molecules and ions within hydrated channels, 9 and traps the protein within its structure. We have demonstrated previously that enzymes retain their native activity within an aqueous Naon environment titrated to near-neutral pH. 10 Assays for proton reduction activity and N 2 reduction activity were conducted for both wild type and b-98 Tyr/His variant MoFe proteins in the presence and absence of Naon as shown in Fig. S2 of the ESI. † These showed that more than 90% proton reduction activity and more than 80% N 2 reduction activity is retained aer exposure to Naon. EPR spectra showed no change in signal intensity or linewidth for either the wild type or b-98 Tyr/His MoFe protein aer exposure to Naon (see discussion under ESI Fig. S2 †).
For electrochemical and IR spectroelectrochemical experiments, europium(III) together with three ligands was added into the Naon-Tris and the electrolyte to give Eu-BAPTA, Eu-EGTA and Eu-DTPA each at 0.25 mM (collectively dened as Eu-L, see ESI for further details †). These complexes provide a convenient series of small-molecule electron transfer mediators with reduction potentials in water spanning the very negative potential range. Reduction potentials were measured at pH 8.0 as À634 mV (Eu-BAPTA), À868 mV (Eu-EGTA) and À1090 mV (Eu-DTPA) (see ESI, Fig. S3 †) in agreement with literature values, 11,12 and the Eu-DTPA potential is insensitive to pH above pH 4.0. 13 We chose to utilise carbon as the working electrode because this material exhibits minimal background proton reduction current down to potentials as low as À1 V vs. the standard hydrogen electrode (SHE). 10 The same cell design was used for both electrochemical and attenuated total reectance (ATR)-IR spectroelectrochemical experiments (Fig. 1). Further detail is provided in the ESI Experimental methods. † Flow of solution through the cell during experiments aided mass transport and allowed for exchange of gas saturated buffers. All experiments were conducted within an anaerobic glove box operating at <1 ppm O 2 .
Electrocatalytic proton reduction by nitrogenase MoFe protein  (Fig. 2b) the current is very close to the background level in the rst potential steps, but a clear negative current is observed aer stepping to À900 mV. The steady increase in current magnitude over time presumably reects the protein equilibrating with the Eu-L pool as it becomes slowly more reduced. An analogous experiment on the wild type MoFe showing that the current at À900 mV does not arise from electrocatalytic reduction of N 2 itself. We therefore attribute the current observed at À900 mV to enzyme-catalysed proton reduction mediated by the reduced Eu-L complexes. Fig. 2 also shows electrochemical data for two variants of the MoFe protein with substitution of amino acids that are located between the P-cluster and the FeMo-cofactor. The response for the b-98 Tyr/His variant (Fig. 2c) is similar to that of the wild type, but with higher electrocatalytic current at À900 mV. Substitution of the nearby 99 Phe residue by His gives a variant, b-99 Phe/His , that also shows higher electrocatalytic current than the wild type enzyme at À900 mV, and shows evidence of some electrocatalytic activity at À800 mV (Fig. 2d).
Assignment of the electrochemical response at À900 mV as electrocatalytic proton reduction by the MoFe protein was conrmed in a separate experiment in a larger volume, stirred electrochemical cell (see ESI Experimental methods, Fig. S1 †) in which H 2 gas produced during a 30 minute poise at À900 mV was detected by gas chromatography (see ESI Experimental methods †). As shown in Fig. 3, there is good agreement between the calculated H 2 produced on the basis of charge passed during the experiment (assuming 0.5 equivalents of H 2 per electron) and the level of H 2 detected in the headspace of the electrochemical cell, as measured by gas chromatography.
These results demonstrate electrochemically-controlled substrate reduction by nitrogenase. The rate of electrocatalytic H 2 production is 30 AE 3 nmol min À1 (mg protein) À1 for wild type MoFe protein at À900 mV, representing approximately 1.4% of the rate observed in an assay with Fe protein and MgATP (2221 AE 66 nmol min À1 (mg protein) À1 , see ESI Table S1 †). The low electrocatalytic rate for the wild type protein is consistent with the observation that it has negligible activity in solution assays with Eu II -EGTA or Eu II -DTPA as electron donors. 7 The fact that the low level of proton reduction by the apo-protein matches that of the control sample ('no protein') indicates that the FeMo-co centre within the wild type nitrogenase is responsible for its electrocatalytic proton reduction activity. Interestingly, the b-98 Tyr/His and b-99 Phe/His variants, which show signicantly lower proton reduction activity than wild type nitrogenase in assays with the Fe protein as electron donor (see ESI, Table S1 †), show higher electrocatalytic activity than the wild type, suggesting that the structural changes induced by these single-site substitutions equip the variants for more effective electron transfer with the reduced Eu-L mediators. The electrocatalytic rates observed for the b-98 Tyr/His and b-99 Phe/His variants represent 4-5% of the rates observed in proton reduction assays with Fe protein as electron donor (see ESI Table S1 †). These results are in agreement with a previous report showing the ability of Eu(II)-DTPA to support the two-electron reduction of hydrazine by the b-98 Tyr/His and b-99 Phe/His variants, but not wild type MoFe protein. 8 At À300 mV, the most positive potential shown for the electrochemical experiments in Fig. 2, the FeMo-co centre should be in the so-called 'resting state', termed M N . Oxidation of this state to the 'oxidised state' termed M OX has a midpoint potential of À42 mV. The midpoint potentials for reduction of FeMo-co below the resting state are not well understood. Assuming that the Eu-L mediator system does not introduce any signicant overpotential to catalysis by the MoFe protein, the onset of electrocatalytic proton reduction in the region of À800 to À900 mV suggests an approximate reductive driving force required for catalysis by nitrogenase. This is consistent  with the estimated potential likely to be experienced by the MoFe protein during photocatalysis on CdS nanorods, 6 although it represents a signicant overpotential relative to the H + /H 2 couple potential (E 0 ¼ À437 mV at pH 7.4).

Effect of CO on proton reduction by the MoFe protein
We next investigated electrochemical control of binding of the well-established inhibitor carbon monoxide (CO) to the MoFe protein. Under assay conditions using Fe protein/ATP, CO is known to inhibit reduction of all substrates catalysed by nitrogenase, 3 although inhibition of proton reduction by CO is weak and has only been detected at pH values above approximately pH 7. 15 Proton reduction assays of wild type and the b-98 Tyr/His and b-99 Phe/His variants of the MoFe protein conducted at pH 7.4 with Fe protein as electron donor (ESI Table S1 †) do not show signicant inhibition by CO.
In electrochemical experiments on the MoFe protein with the Eu-L mediators at pH 7.4, CO is seen to inhibit proton reduction irreversibly at À900 mV, causing a substantial drop in the electrocatalytic current. A representative trace for the b-98 Tyr/His variant is shown in Fig. 4; similar responses are observed for the wild type and b-99 Phe/His proteins. In electrochemical experiments conducted under a CO atmosphere, no hydrocarbon product of CO reduction was detected from either the wild type MoFe or variant MoFe proteins, showing that the proteins are not electrocatalytically reducing CO under these conditions.
Although proton reduction activity in the MoFe protein is generally reported to be fairly insensitive to CO in assays, CO-bound states of the MoFe protein prepared during turnover studies with Fe protein and ATP under a CO atmosphere have been detected by both stopped-ow IR spectroscopy [16][17][18] and freeze-quench EPR and ENDOR. 19 Several CO-bound species were observed in these studies, depending on the CO concentration. A crystal structure of the MoFe protein prepared under turnover conditions under a CO atmosphere showed a single enzyme form, in which the FeMo-cofactor had opened up with a bridging CO between two Fe sites on a face of the cluster. 20 We have applied ATR-IR spectroelectrochemistry to examine the effect of CO as a function of potential. Free CO in aqueous solution gives a weak IR stretching band (n CO ) at about 2143 cm À1 , which typically shis to lower wavenumber (ca. 2100-1800 cm À1 ) and becomes much more intense when CO coordinates to a metal centre, making it easy to determine features arising from CO ligands at metal sites. 21 Potentialdependent difference spectra for the MoFe proteins, recorded under an atmosphere of 100% CO, are shown in Fig. 5, relative to a background spectrum recorded at À100 mV prior to introduction of CO to the cell. For the apo-protein (Fig. 5a), no n CO bands are observed at any of the applied potentials. The wild type MoFe protein and the b-98 Tyr/His and the b-99 Phe/His variants exhibit no n CO bands at the more positive potentials, but all show evidence of potential-induced coordination of CO at more negative potentials (Fig. 5, panels b-d). The general features of the CO-bound spectra are similar in all three cases: a broad band centred at approximately 1940 cm À1 dominates in intensity, with shoulders at both higher and lower wavenumber (approximately 1965-1970 and 1910-1914 cm À1 , respectively), in addition to several features at higher wavenumber (>2000 cm À1 ). The absence of any detectable n CO bands for the apo protein provides strong evidence that all the n CO bands observed for the wild type and b-98 Tyr/His and b-99 Phe/His variants of the MoFe protein arise from CO bound at FeMo-co rather than the P-cluster.  The absence of n CO bands at potentials more positive than À700 mV shows that CO does not bind to resting state FeMo-co, but requires more reduced levels of the cofactor. This is consistent with studies on isolated FeMo-co in N-methyl formamide in which no perturbation of the oxidised/semi-reduced FeMo-co couple was observed under CO, and the onset of CO binding was only detected at potentials around À600 mV more negative than the oxidised/semi-reduced couple, coinciding with potentials at which electrocatalytic proton reduction by isolated FeMo-co is observed. 22,23 In the spectra shown in Fig. 5, the onset for CO binding also coincides fairly closely with the onset potential for electrocatalytic proton reduction for each of the proteins (see Fig. 2), suggesting that states generated during the MoFe protein catalytic H + reduction cycle may be targeted by CO. This is consistent with spectroscopic and crystallographic studies during turnover under CO, which all show CO binding to be dependent on catalysis. [16][17][18][19][20] In this context, the weak inhibitory effect of CO observed in biochemical assays is curious and will be explored further in future work.
The spectra in Fig. 5 resemble quite closely the 'hi-CO' spectra reported in stopped ow IR studies on the MoFe protein reduced by Fe protein in the presence of ATP and CO, which are dominated by a band at 1936 cm À1 , with smaller bands at 1906 and 1960 cm À1 , 16,18 as well as weak higher wavenumber features around 2020-2030 cm À1 which were not assigned. 18 This provides compelling evidence that the electrochemically generated states of nitrogenase MoFe protein studied here are relevant to states generated using the native Fe protein as electron donor. In the spectra shown in Fig. 5, the band centred around 1940 cm À1 appears to have contributions from two components. These are most clearly resolved for the b-99 Phe/His protein, at 1948 and 1934 cm À1 . A similar split n CO band and pattern of relative intensities was observed for isolated FeMo-co in the presence of imidazole which is presumed to mimic the native histidine coordination at molybdenum. 22,23 Both the b-98 Tyr/His and b-99 Phe/His variants show more intense features than the wild type protein. This could be related to their improved ability to take up electrons from the Eu-L system, but could also arise from slight structural alterations that modify CO access to FeMo-co. Altered CO access has been observed in a variant of the MoFe protein with a mutated residue close to the terminal Fe site of FeMo-co. 24 Previous spectroscopic studies of CO binding to FeMo-co within the MoFe protein have been interpreted in terms of an initial bridging CO ligand between two Fe sites that opens up to two terminal CO ligands at higher CO concentrations. 17,19,[25][26][27] Broad lower wavenumber features below 1900 cm À1 in the spectra in Fig. 5 may be associated with bridging CO ligands. Development of a peak at 1897 cm À1 in spectra of the b-99 Phe/His variant becomes clear in difference spectra following the onset of CO binding (see ESI, Fig. S6 †).
Additionally, a higher wavenumber band at 2050 cm À1 is particularly sharp in the spectra of the b-99 Phe/His variant, and indicates CO bound to a less electron-rich site, possibly arising from CO on Mo. Molybdenum hexacarbonyl exhibits a single n CO at 2004 cm À1 . 28 For the relatively few characterised examples of carbonyl complexes of Mo in higher oxidation states, the n CO bands are very sensitive to the co-ligands. The Mo(III) complex [CpMoCl(dppe)(CO)] + has its n CO band at 2002 cm À1 , while the Cp* derivative has n CO at 1971 cm À1 . 29 ‡ The reduced Mo(II) forms of these complexes have n CO bands at 1853 cm À1 and 1835 cm À1 respectively. 29 The Mo(III) complexes ((Me 3 SiNCH 2 CH 2 ) 3 N) Mo(CO) and ((C 6 F 5 NCH 2 CH 2 ) 3 N)Mo(CO) exhibit lower wavenumber n CO bands, at 1859 cm À1 and 1889 cm À1 respectively, however. 30 Flushing CO out of the cell, whilst poising the potential at À900 mV, does not lead to any noticeable decrease in intensity of the n CO bands, consistent with the fact that electrocatalytic H + reduction current does not recover when CO is ushed out with Ar (see Fig. 4). Switching to a N 2 atmosphere and stepping the potential to À100 mV aer the experiment on wild type MoFe protein shown in Fig. 5b resulted in partial depletion of all n CO features (see ESI, Fig. S7 †), showing that CO binding to the FeMo-co centre is somewhat reversible. The Eu-L mediator system used in this work will not effectively mediate the solution potential above approximately À400 mV (see ESI, Fig. S3 †) which may explain the incomplete reversibility.

Conclusions
The results presented here demonstrate that the nitrogenase MoFe protein can be placed under electrochemical control to drive substrate reduction or ligand binding using a Eu(III/II)ligand mediated system, without the need for the Fe protein/ ATP. The electrocatalytic rate of H 2 evolution we measure is modest compared to the rate for H 2 evolution with Fe protein/ ATP as electron donor, or for photocatalytic H 2 production by wild type MoFe protein on CdS (3000 nmol of H 2 min À1 (mg protein) À1 ), 6 but is greater than the rate observed in an assay with Eu-EGTA as electron donor without electrochemical regeneration (6.5 nmol of H 2 min À1 (mg protein) À1 ). 8 The Eu-L mediated electrocatalytic method was then used to examine the inhibitory effect of CO on electrocatalytic proton reduction by wild type and b-98 Tyr/His and b-99 Phe/His MoFe proteins using a combination of electrochemistry and infrared spectroelectrochemistry, showing the onset potential for H + reduction and CO binding to be closely related. Several CO ligands appear to bind to the FeMo-co cluster, possibly at different redox levels, and give rise to IR n CO bands that closely resemble the potential-dependent features observed in earlier studies on isolated FeMo-co, as well as those observed for the MoFe protein under turnover conditions with Fe protein/ATP as reductant. This approach opens up opportunities for future work to examine, systematically, electron transfer and ligand binding reactions to nitrogenase MoFe protein without the complexities of the Fe protein/ATP as a reductant system.
We are grateful to Henry Waite for carrying out initial Eu-L experiments and to Ricardo Hidalgo for the electrochemical ow cell design. This material is based upon work supported by the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences (DE-SC0010687 and DE-SC0010834 to LCS and DRD).