Unifying activity, structure and spectroscopy of [NiFe] hydrogenases: combining techniques to clarify mechanistic understanding

cyclic voltammogram of E. coli Hyd1 recorded under a H 2 atmosphere at a rotating disc electrode using PFE; (C) cyclic voltammogram of E. coli Hyd1 recorded under a H 2 atmosphere in the PFIRE spectroelectrochemical cell; (D) IR spectrum recorded at +356 mV showing active site ν CO and ν CN bands of the electrochemically-generated Ni-B redox state. Other conditions: pH 6.0 buffered electrolyte.

states are color-coded to match IR spectra shown later, ν CO band positions refer to Hyd1, pH 6.0. 21 The catalytic cycle of [NiFe] hydrogenases comprises a series of linked electron and proton transfer steps ( Figure 1B). The active site Fe atom remains formally Fe(II) throughout, although there is substantial delocalization as indicated by changes in CO and CNwavenumber positions (ν CO and ν CN ) in IR spectra, highly sensitive to electron density on Fe. 22 The ν CO band positions of the active site of E. coli hydrogenase I (Hyd1) are given in Figure 1B. 21,22 Starting at the Ni(II) level in the Ni a -SI state, and considered in the direction of H 2 oxidation, H 2 activation leads to Ni a -R in which nickel remains formally Ni(II), with a bridging hydride ligand and nearby proton. The data we present here are recorded on E. coli Hyd1, an O 2 -tolerant membrane-associated [NiFe] hydrogenase which natively exchanges electrons with the quinone pool via a membrane anchor cytochrome. 37 Electron-relay chains, a common feature of many redox metalloenzymes, make hydrogenases well-equipped to exchange electrons with an electrode; the technique of protein film electrochemistry (PFE), 3,4,38 in which the protein is adsorbed onto a carbon electrode and probed under direct electronic control has been applied extensively to Hyd1 and its variants. [34][35][36]39,11 Hyd1 also crystallizes readily; high resolution structures are available for the native enzyme and several variants. [34][35][36][37] Therefore Hyd1 is an ideal model system to demonstrate how activity-based, spectroscopic and structural approaches can be combined to build a more comprehensive picture of enzyme function.

Linking PFE and IR Spectroscopy: Protein Film IR Electrochemistry
We have developed a method called protein film IR electrochemistry (PFIRE) that combines direct electrochemical control at a carbon electrode with simultaneous in situ IR spectroscopy. 21,40 Enzyme is adsorbed onto high surface area (>1000 m 2 g -1 ) 41  within a (spectro)electrochemical cell ( Figure 3A). The high electrode surface area is necessary to provide sufficient material in the IR sampling area for spectroscopic detection; for some hydrogenase variants, covalent coupling to the surface has proved beneficial in increasing coverage and therefore IR signal. 35 Rapid solution flow sustains fast electrocatalysis through supply of fresh substrate, and cyclic voltammograms analogous to those recorded using PFE are obtained in the presence of H 2 -saturated buffer ( Figure 3B and C). (All potentials in this manuscript are quoted vs the standard hydrogen electrode, SHE.) We can also obtain information under non-turnover conditions equilibrated with Ar; Figure 3D shows an IR spectrum of Hyd1, recorded at +356 mV, showing the ν CO and ν CN bands of the active site in the 'pure' oxidized, inactive Ni-B state.

Non-turnover studies
Although no significant Faradaic current is observed for a Hyd1-modified electrode under an Ar atmosphere ( Figure 4A), a redox titration of electrode-immobilized Hyd1 under these nonturnover conditions reveals the equilibrium ('static') behavior of the [NiFe] active site as the enzyme responds to electrode potential ( Figure 4B). At pH 6.0 Hyd1 is strongly biased towards H 2 oxidation and is very poor at proton reduction. 42 Nevertheless, the fact that we are unable to obtain pure Ni a -R states even at very negative potentials is indicative of slow catalytic proton reduction under these conditions. Transformations between all redox states are fully reversible, 21,40 allowing assignment of the various states of Hyd1 on the basis of their potential dependence, and not only by comparison to the substantial literature on [NiFe] hydrogenase redox chemistry in solution. 19,22,43,44 The proportion of different states at each potential can be estimated from the ν CO intensities, and Figure 4C shows a graphic representation of a 'slice' of the redox titration at −199 mV, where each bar represents the population of each redox state.   Figure 3A). 21,40 A positive catalytic current is now observed at the electrode at −74 mV and +356 mV ( Figure 5A), as enzymatic H 2 oxidation channels electrons to the electrode. The current is very stable over the measurement time at −74 mV, but decays slowly at +356 mV due to well-established oxidative inactivation. 11,19,45 IR spectra recorded at these potentials are shown in Figure 5B: in contrast to the +356 mV spectrum under argon which shows only the inactive state, Ni-B ( Figure 3D), the corresponding spectrum under H 2 shows a distribution of states which reflect steady-state catalysis. Importantly, Ni a -L states are populated suggesting that restoration of Ni a -SI from Ni a -C does go via a detectable Ni(I) intermediate.
Together with evidence from time-resolved photolysis and light-induced potential jump experiments, this strongly implies involvement of the Ni(I) states in catalysis. 15,[23][24][25]46,47 The ν CO bands of Ni a -L are very low in wavenumber for a terminal CO ligand, consistent with the presence of a metal-metal bond and extensive delocalization of electron density onto Fe. 48 Two different forms of Ni a -L and of Ni a -R are observed in Hyd1 spectra, and these have been assigned to differentially protonated states. 15,19,22,46 Spectroscopic data are also presented in Figure 5C as turnover minus non-turnover difference spectra to reflect changes in speciation in response to catalysis. Species that are populated in response to turnover (as opposed to simply a response to electrode potential) give rise to positive peaks. The speciation is highlighted graphically in Figure 5D, where the area of each bar represents the population of that state during steady-state turnover. At +356 mV, some Ni-B is present even during turnover under H 2 , consistent with the high potential inactivation evidenced by decay in catalytic current ( Figure   5A). Inactivation is fully reversible when the potential is stepped to less positive values. 21

The effect of pH
A major advantage of PFE is that measurements can be made on the same enzyme sample whilst perturbing solution conditions, such as substrate and inhibitor concentration, or pH. 2-4,11,39 1980 1960 1940 1920 1900   These advantages are retained by PFIRE, with the additional benefit of observing, spectroscopically, chemical changes brought about by the perturbation. Solution pH has a significant effect on [NiFe] hydrogenase catalysis: E. coli Hyd1 becomes good at H 2 production at low pH. 42 By varying solution pH we found clear evidence for an acid-base equilibrium between the Ni a -C and Ni a -L states of Hyd1. 49 The two Ni a -L states observed for Hyd1 do not share the same pH dependence ( Figure 6), strongly implying the existence of distinct proton acceptor sites. 22 Similar pH dependence of Ni a -L states has been reported for the [NiFe] hydrogenase from D. vulgaris MF, where an additional Ni a -L state was formed at high pH. 46 Taken together with transient absorption spectra following a laser-induced potential jump 15,24 these data imply multiple Ni a -L states can be involved in catalysis sequentially.
It has been suggested that concerted proton and electron transfer takes place during the Ni a -LNi a -SI transition. 24 It is also known that the proximal iron-sulfur cluster ( Figure 2) must be in an oxidized state in order to accept an electron from the [NiFe] active site; Ni a -SI formation following photolysis of Ni a -C only occurs when the proximal cluster is oxidized. 23,24 Therefore, the Ni a -CNi a -LNi a -SI steps during the catalytic cycle are 'gated' by both the proximal cluster and the proton acceptor site(s). Unlike the highly-conserved nature of amino acid residues surrounding the [NiFe] active site, distinct proximal cluster structures have been identified in hydrogenases with different catalytic characteristics. 19 hydrogenases, allowing more time for acid-base chemistry around the active site, and making a wider range of Ni a -L states accessible during turnover: the proton released from Ni a -C can travel further before an electron transfers to the proximal cluster. This provides a tentative mechanistic rationale for the Ni a -L states observed in O 2 -tolerant hydrogenases, which tend to have very low ν CO . In these Ni a -L states the proton is likely to be further removed from the primary coordination sphere of the [NiFe] active site. 22 Figure 6. The relative populations of Ni a -C and Ni a -L are pH dependent. Additionally, the Ni a -L states observed in E coli Hyd1 do not share the same pH dependence, implying multiple proton acceptor sites.

PFIRE applied to a variant of Hyd1 with impaired proton transfer: E28Q
Site-directed mutagenesis is widely used by molecular biologists to probe the importance and role of specific amino acids on enzyme activity in order to gain mechanistic insight into the native enzymes. Assumptions are often made that mutations only have a single well-defined effect on catalysis, commonly on the basis that active sites 'look' the same as judged from As a case study, we consider a highly-conserved glutamate (residue E28 in E. coli Hyd1 numbering), located within 4 Å of the terminal cysteine residues of the [NiFe] active site ( Figure   2) and widely reported to be critical for catalytic proton transfer during the catalytic cycle. 29,31,33,35,54,55 Mutation of glutamate 28 to glutamine (Q, to form the E28Q variant in Hyd1), greatly suppressing the proton-transfer ability of the site by raising the pK a , has been shown to attenuate substantially the activity of EQ variants in a range of hydrogenases. These observations, coupled with apparently unaltered active site redox chemistry or structure, have led to consideration of glutamate E28 as the essential 'gating' residue for proton transfer during catalysis. 29 This interpretation may be over-simplistic, however, and belies the fact that: (a) [NiFe] hydrogenase active site 'cavities' contain not only highly conserved amino acid residues but also highly conserved, crystallographically ordered water molecules that can also act as H + transfer sites (at least within the active site cavity); 34  Using a combination of non-turnover and steady-state turnover PFIRE measurements, we can compare active site chemistry of E28Q Hyd1 with the native enzyme. 35 Figure 7A shows a difference spectrum (steady-state H 2 oxidation minus non-turnover under Ar) of E28Q Hyd1 at −100 mV. In line with previous reports we find that the non-turnover redox behavior of the [NiFe] active site remains largely unaltered (negative bands, Figure 7A), and further that the ν CO and ν CN peak positions of all redox states remains unchanged, although in contrast to native Hyd1 we see no conclusive evidence for Ni a -L formation under any conditions in E28Q Hyd1. Figure 7B shows    Figure 8B). Similar high potential inactivation was observed with E28Q Hyd1 ( Figure 8C), but in this instance PFIRE studies did not reveal Ni-B formation during high potential inactivation ( Figure 8D, red line): the fairly flat difference spectrum shows minimal effects at the active site accompanying the significant loss of catalytic current. In contrast, removal of H 2 by flushing with Ar (difference spectrum in Figure 8D, black line) does result in rapid active site oxidation and Ni-B formation, proving that Ni-B is still an accessible state. This observation implies that there must be alternative high-potential inactivation processes that can take place during turnover (perhaps linked to super-oxidation of the proximal iron sulfur cluster), 35  to be slow to re-reduce due to a significant structural change accompanying the superoxidized/oxidized redox couple. 35,37,56 In situ spectroscopic data are lacking on other hydrogenases under turnover conditions, and so the generality of Ni-B formation during inactivation is largely unexplored. Electrochemistry alone can only identify relative activities or the existence of inactivation processes, but cannot definitively diagnose the underlying active site chemistry.

Single crystal microspectroscopy coupled to electrochemical control
In the same way that in situ spectroscopy is increasingly important in mechanistic studies of redox enzymes, spectroscopic methods are required to complement structural studies of metalloproteins. 57 Single crystal microspectroscopic methods to confirm the integrity, physiological relevance, and redox state of crystallized proteins, and to assess radiation damage are becoming more common. [57][58][59][60] Such techniques have been applied to on-and off-line monitoring, but there is a need to develop methods which offer control over the oxidation state of crystallized proteins, or go beyond 'static' study of proteins in the crystalline state. For complex enzymes with multiple redox states, it can be difficult to structurally characterize all states due to challenges in crystal sample preparation. 59 The state of crystallized proteins can be manipulated by soaking in reductants, oxidants, substrates or inhibitors, or the protein can be chemically treated prior to crystallization. We combined these ideas with electrochemical control over a pool of redox mediators spanning the potential range relevant to redox states of [NiFe] hydrogenases, using a custom-designed IR microspectroscopic-electrochemical cell ( Figure 9). 14   Figure 10A). Importantly the peak positions and equilibrium potential dependencies of redox states observed in crystallo, in solution, and in PFIRE measurements are the same. Redox transitions involving chemical steps (for example proton transfer) appear to be retarded in the crystalline state. This is shown by the time-dependent formation of two Ni a -R states in Hyd1 crystals ( Figure 10B), in contrast to the appearance of these states together under turnover in PFIRE measurements (see Figure 5B). The existence of multiple Ni a -R states is well documented, and although the relative population of different Ni a -R states is known to be pH dependent there is no general mechanistic rationalization of these observations. The pH dependence implies structural differences in protonation state, and the observation that two Ni a -R substates form at different rates in crystalline Hyd1 suggests that individual Ni a -R states may represent sequential proton transfer steps away from the [NiFe] active site following initial H 2 activation. There is also the tantalizing prospect that one of the Ni a -R states could be closely related to the Michaelis-Menten complex of H 2 at the active site. The examples we have presented here from studying E. coli Hyd1 using IR spectroscopy applied in different ways demonstrate the value of coupling activity-based measurements with spectroscopy. The observation that high potential inactivation of the E28Q variant of Hyd1 does not arise from formation of the well-known Ni-B state but seems to be associated with an oxidation of the proximal iron sulfur clusters challenges common assumptions about hydrogenase inactivation and highlights possible pitfalls in over-interpreting current-activity data from PFE alone. We also show the power in carrying out in situ spectroscopy during catalytic turnover to report on steady-state intermediates during catalysis. Using this approach, we have been able to strengthen the evidence for involvement of the Ni(I) intermediate Ni a -L in the [NiFe] hydrogenase catalytic cycle, for example, and to diagnose catalytic 'bottlenecks' introduced by the targeted E28Q mutation in Hyd1, allowing us to propose a critical mechanistic role for E28Q in the second proton transfer during H 2 oxidation.
The PFIRE method is particularly valuable for studying hydrogenases, as intrinsic CO and CNligands are strong IR absorbers and report directly on changes in redox state. Since the protein is immobilized on the electrode, variable temperatures or pH conditions can readily be applied in PFIRE measurements by controlling the temperature or pH of the buffered electrolyte.
We further show that IR spectroscopy combined with electrochemical control can be extended to the crystalline state, where the enzyme can still be regarded as a dynamic system able to cycle through redox states relevant to catalysis. This is a valuable tool for mechanistic studies, as chemical steps seem to be slowed in crystallo, so that, for example, formation of different Ni a -R states could be kinetically resolved. It will also be extremely useful to link this control to structure determination. It is possible to generate several redox states of Hyd1 in almost pure form at specific potentials (see Figure 4B), which could enable structural study of states that have thus far been inaccessible for X-ray diffraction. Preliminary results (unpublished) show no noticeable drop in X-ray diffraction quality for crystals of Hyd1 after electrochemical manipulation. It will also be important to use spectroscopic methods to characterize crystals that have been generated for structural study in order to correlate known catalytic and solution behavior of redox states to atomic resolution structures. We therefore see possibilities for a unified approach to bioinorganic chemistry, with coherent crossover of methods from solution, electrochemical, and solid state studies helping to bridge gaps in mechanistic understanding and reconcile confusion between different types of data set.
Where else will this be useful? There have been several recent fascinating X-ray crystallographic studies of metalloproteins that are yet to be reconciled with solution behavior of these enzymes. Rees and coworkers reported a CO bound structure of molybdenum nitrogenase where the molybdenum iron catalytic cluster has opened up with loss of sulfide to coordinate the inhibitor. 63 IR spectroscopy on crystals of the CO-inhibited nitrogenase would be valuable in relating this structure to the wealth of CO-inhibited states observed spectroscopically during stopped-flow studies of CO inhibition during turnover. 64 Other valuable systems to target with a suite of turnover spectroelectrochemical studies and in crystallo spectroscopic studies will be the molybdopterin enzymes nitrate reductase and formate dehydrogenase where disparate mechanisms have been suggested for similar enzyme active sites. 65 Carbon monoxide dehydrogenase is another fast turnover enzyme for which it is difficult to prepare samples for spectroscopy in catalytically relevant states, and the in situ and in crystallo substrate-turnover methods may provide new mechanistic understanding.
In conclusion, we hope that solution, electrode and crystal methods will be utilized in a much more unified way in future studies in bio-inorganic chemistry so that information can be reconciled more readily across different sample types. It will also be important to apply the same range of spectroscopic and activity-based methods to enzymes from different organisms in order to understand how general some of these mechanistic steps are to metalloenzymes from different environments or with different cellular roles. We hope that in the future, the versatile sampling possibilities offered by IR spectroscopy will bridge key divides between the chemical, biochemical, and structural biology communities in mechanistic enzymology.