Intercluster Redox Coupling Influences Protonation at the H-cluster in

Article Cite This: J. Am. Chem. Soc. 2017, 139, 15122-15134

pubs.acs.org/JACS

Intercluster Redox Coupling Influences Protonation at the H‑cluster in [FeFe] Hydrogenases Patricia Rodríguez-Maciá, Krzysztof Pawlak, Olaf Rüdiger, Edward J. Reijerse, Wolfgang Lubitz,* and James A. Birrell* Max Planck Institute for Chemical Energy Conversion, Stiftstraße 34-36, D-45470 Mülheim an der Ruhr, Germany

Downloaded via UNIV OF CAMBRIDGE on July 8, 2018 at 21:02:34 (UTC). See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

S Supporting Information *

ABSTRACT: [FeFe] hydrogenases catalyze proton reduction and hydrogen oxidation displaying high rates at low overpotential. Their active site is a complex cofactor consisting of a unique [2Fe] subcluster ([2Fe]H) covalently bound to a canonical [4Fe−4S] cluster ([4Fe−4S]H). The [FeFe] hydrogenase from Desulfovibrio desulf uricans is exceptionally active and bidirectional. This enzyme features two accessory [4Fe−4S] F clusters for exchanging electrons with the protein surface. A thorough understanding of the mechanism of this efficient enzyme will facilitate the development of synthetic molecular catalysts for hydrogen conversion. Here, it is demonstrated that the accessory clusters influence the catalytic properties of the enzyme through a strong redox interaction between the proximal [4Fe−4S]F cluster and the [4Fe−4S]H subcluster of the H-cluster. This interaction enhances proton-coupled electronic rearrangement within the H-cluster increasing the apparent pKa of its one electron reduced state. This may help to sustain H2 production at high pH values. These results may apply to all [FeFe] hydrogenases containing accessory clusters.

■

has an open coordination site where the substrates H2 or H+ bind, as well as inhibitors including CO and O2. The nitrogen atom in the adt bridge serves as a base facilitating heterolytic splitting of H2 at Fed. For the [FeFe] hydrogenase from Chlamydomonas reinhardtii (CrHydA1) several redox states of the H-cluster have been identified, providing an almost complete picture of a potential catalytic cycle (Figure 1B).15−22 The oxidized state (Hox) has an oxidized [4Fe−4S]H subcluster, and the [2Fe]H subcluster is in a mixed valence Fe(I)Fe(II) redox state. One-electron reduction occurs at [4Fe−4S]H generating the Hred state. Protonation of the adt moiety is coupled to electron transfer from [4Fe−4S]H to [2Fe]H affording the HredH+ state with a homovalent Fe(I)Fe(I) [2Fe]H subcluster.19 This enables a second reduction at [4Fe−4S]H yielding the HsredH+ state,15,18 which may rearrange to form a hydride-bound state with Fe(II)Fe(II) at [2Fe]H and a reduced [4Fe−4S]H.20−22 A second protonation may again be coupled to electron transfer to [2Fe]H, followed by H2 formation. While the H-cluster in CrHydA1 interacts directly with its redox partner (PetF),23 bacterial [FeFe] hydrogenases contain additional iron−sulfur clusters, the so-called F-clusters, for efficient electron transfer between the buried H-cluster and the protein surface.24 However, the F-clusters may also serve additional functions. For instance, the redox midpoint potential

INTRODUCTION

Hydrogen is considered to be an ideal primary fuel for a future society based on renewable energy due to its high energy density and clean combustion, which produces only water as a waste product.1 At present H2 is produced mostly from fossil fuels or by water electrolysis using expensive noble metal catalysts. Hydrogenases are biocatalysts that interconvert protons and electrons and molecular hydrogen in a highly efficient and reversible manner, with rates comparable to that at platinum electrodes,2,3 but using the abundant metals nickel and/or iron.4−6 While [NiFe] hydrogenases are the most common in nature, it is the [FeFe] hydrogenases that are the most active (up to 10 000 s−1 in H+ reduction),7 but at the same time they are highly oxygen sensitive.8 A remarkable feature of these enzymes is their reversibility, requiring very small overpotentials for catalysis in both directions.9 In recent years, great efforts have been made to incorporate similar reversibility in the design of molecular catalysts.10 The active site of [FeFe] hydrogenase, the H-cluster, consists of a unique [2Fe] subcluster ([2Fe]H) bridged by a cysteine sulfur to a canonical [4Fe−4S] cluster ([4Fe−4S]H) (Figure 1A).11,12 The [2Fe]H subcluster has two irons bridged by the thiol groups of an aza-propane-1,3-dithiolate (adt) ligand,13,14 a bridging carbon monoxide ligand, and two terminally bound cyanide and carbon monoxide ligands (one on each iron). The iron directly bound to the [4Fe−4S]H subcluster (the proximal iron (Fep)) is coordinately saturated, while the distal iron (Fed) © 2017 American Chemical Society

Received: August 2, 2017 Published: September 14, 2017 15122

DOI: 10.1021/jacs.7b08193 J. Am. Chem. Soc. 2017, 139, 15122−15134

Article

Journal of the American Chemical Society

for the development of synthetic catalysts with comparable activity.

■

RESULTS AND DISCUSSION A Redox Interaction is Observed Between the Proximal F-cluster and the H-cluster in DdHydAB(pdt). A substantial difficulty in thermodynamic studies of hydrogenases is that one of their substrates, the proton, is always present. This prevents the establishment of a thermodynamic equilibrium during redox titrations at very negative potentials, unless high pressures of hydrogen gas can be maintained. Furthermore, the redox-coupled protonation events at the H-cluster are complex. Therefore, in order to allow titrations to very negative potentials, and to focus on the redox processes at the [4Fe−4S]H subcluster, we started our investigations with FTIR monitored redox titrations of the catalytically inactive H-cluster variants in which the proton-coupled electronic rearrangement process was blocked. This is the case for variants that were maturated with a [2Fe]H analogue in which the adt bridge is replaced by a propane dithiolate (pdt) moiety.14,33 Since the methylene headgroup of the dithiol bridge cannot be protonated, no reduction of the [2Fe]H subcluster is possible under physiological conditions. The same behavior is displayed by the CO inhibited form of the native enzyme, which will also be studied here.31 The H-cluster in these preparations only exists in two forms corresponding to the redox state of the [4Fe−4S]H subcluster (Figure 1B). FTIR spectroscopy can record the stretch vibrations of the CO and CN− ligands, unique to the [2Fe]H subcluster. Reduction of the H-cluster, either at [4Fe−4S]H or at [2Fe]H, increases the electron density on [2Fe]H, causing characteristic shifts in the FTIR bands. FTIR spectra can be measured over a range of applied electrochemical potentials in an FTIR cell equipped with a standard three-electrode system. The midpoint potentials of the redox events occurring at the H-cluster can then be extracted by fitting the amplitude of FTIR bands associated with specific redox states to the Nernst equation. Figure 2A shows FTIR spectra of DdHydAB maturated with [2Fe]pdt (DdHydAB(pdt)) at three different potentials. At less negative potential (−200 mV vs SHE) the spectrum resembles that of the Hoxpdt state reported for CrHydA1(pdt) and at more negative potential (−680 mV) the spectrum resembles that of the Hredpdt state.31 By following the absorbance of the most intense peak of each state (1941 cm−1 for Hoxpdt and 1934 cm−1 for Hredpdt) it can be seen that the transition between them occurs at about −500 mV, 100 mV more negative than the same transition in CrHydA1. In the redox titration experiment the peak amplitudes clearly show non-Nernstian behavior (Figure 2B) that is dependent on the measurement position within the FTIR line shape (Figure S1). This behavior is reproducible in both the reductive and oxidative titrations and was observed in multiple repeats of the same experiment (Figure S2A). This effect is associated with small shifts (≈1 cm−1) in the FTIR peak positions of the Hoxpdt and Hredpdt states as a function of the potential (Figure 2C). The nonNernstian behavior is also observed if the area under the main peaks for Hoxpdt and Hredpdt (estimated by integration) are plotted against the potential (Figure S2B and C). Similar peak shifts are found for the other peaks in the FTIR spectrum (Figure S3). Increasing the resolution of the FTIR experiment to 0.2 cm−1 (instead of 2 cm−1) only slightly improved the resolution of these minute FTIR band shifts (Figure S4). The time required for the 0.2 cm−1 resolution measurements is ten

Figure 1. Structure and proposed mechanism of [FeFe] hydrogenases. A close-up view of the F-clusters and the H-cluster is shown with the shortest interatomic distances between the clusters (A). The figure was created in Pymol using PDB file 1HFE.12 The accessible states for the F-cluster free [FeFe] hydrogenase from Chlamydomonas reinhardtii CrHydA1 (B) maturated with [2Fe]adt (green and red parts of the cycle) and [2Fe]pdt (red part of the cycle only).

of the most distal cluster from the active center is predicted to be an important determinant of the enzyme’s catalytic bias (the relative rates of H2 production and oxidation at comparable overpotentials).25,26 Additionally, the F-clusters may influence the oxygen sensitivity since it was demonstrated that oxygentolerant [NiFe] hydrogenases containing a modified proximal F-cluster are able to quickly supply enough electrons to reduce oxygen to water.27,28 The [FeFe] hydrogenase from Desulfovibrio desulf uricans, DdHydAB, is an exceptionally active bidirectional enzyme that operates reversibly over a wide pH range.7,29,30 It contains two accessory clusters bound in a bacterial ferredoxin fold. FTIR spectro-electrochemistry of the [FeFe] hydrogenase from both C. reinhardtii and D. desulf uricans has been previously reported.18,19,31,32 Yet a number of conflicting observations made for the two enzymes remain to be understood. First, the super-reduced state, Hsred, originally identified in DdHydAB at very negative potentials, occurs at much more positive potentials in CrHydA1. Second, the Hred−CO state identified in CrHydA1 has so far not been observed for DdHydAB, despite measurements under similarly reductive conditions.32 Third, only one of the two Hred states recently identified in CrHydA1 have been observed in DdHydAB. We speculated that these differences may be related to the presence of the F-clusters in DdHydAB, and therefore, we investigated this further using FTIR specto-electrochemistry and EPR spectroscopy. A detailed understanding of the mechanism of this highly active hydrogen converting enzyme may provide crucial insight 15123


Article


Figure 2. FTIR spectro-electrochemistry of DdHydAB(pdt). FTIR spectra recorded from the spectro-electrochemical redox titration of 1 mM DdHydAB(pdt) in 50 mM MES, 50 mM HEPES at pH 8 and 100 mM KCl (as supporting electrolyte) containing redox mediators (see experimental section) at various applied potentials in a spectro-electrochemical cell. Selected spectra at −200, −460, and −680 mV are shown (A) highlighting the Hoxpdt (dark/light blue) and Hredpdt (red/orange) states. (B) Change in FTIR absorbance with redox potential at the peak positions 1941 cm−1 (blue circles) and 1935 cm−1 (red circles), where non-Nernstian behavior is clearly observed. The blue and red lines correspond to n = 1 Nernst fits with Em = −500 mV for the oxidized and reduced species, respectively. (C) Close-up view of the 1925 to 1950 cm−1 region at the same three potentials as in (A), highlighting the peak shifts. The peaks have been fit with Gaussian functions in order to extract the relative contributions from the FoxHoxpdt, FredHoxpdt, FoxHredpdt, and FredHredpdt states. The relative contributions from each state are plotted against the redox potential (D), and the data are fitted with Nernst equations based on the model shown in (E) (see also Figure S7). The fitted redox potentials are shown in (E). All potentials are quoted versus the standard hydrogen electrode (SHE). 15124


Article

Journal of the American Chemical Society times longer than that for the 2 cm−1 resolution measurements. The protein showed some instability over such long time periods. Therefore, subsequent measurements were performed at 2 cm−1 resolution. Such small shifts cannot be attributed to changes in the redox state of the H-cluster itself, and so they must arise from redox changes at a nearby site, most likely the proximal F-cluster. This results in two substates of each of the H-cluster redox states depending on the redox state of the proximal F-cluster, which we call FoxHoxpdt, FredHoxpdt, FoxHredpdt, and FredHredpdt. Therefore, the data analysis is complex and requires meticulous care in fitting the peaks as is described below. The relative contributions of the Fox/Fred associated states were extracted by fitting the shifting FTIR peaks using Gaussian functions with fixed peak positions and line widths but with variable amplitudes (Figure 2C and Figure S5). The fixed parameters (peak position, line width, and maximum amplitude) were determined at potentials where the pure states could be observed. For the FoxHoxpdt and FredHredpdt this was done at −200 and −680 mV, respectively, and for the FredHoxpdt and FoxHredpdt states −460 and −380 mV were used, respectively. The maximum amplitudes of the FredHoxpdt and FoxHredpdt states were assumed to be the same as those of the FoxHoxpdt and FredHredpdt states, respectively. The peaks could be fitted very well with Gaussian functions, as has been demonstrated previously for IR peaks from other metalbound CO ligands in proteins.34−36 Negligible improvement in the fitting was achieved by using Voigtian functions (Figure S6), and the peak positions were unaffected. IR peaks from molecules in solution normally display Voigtian line shapes compared with Gaussian line shapes for solids due to free rotation in the former. However, in the concentrated (≈1 mM), viscous protein solutions used in these experiments, the CO ligands of the H-cluster show mostly Gaussian behavior due to restricted rotation. Therefore, to simplify the data fitting, we assumed Gaussian line shapes. At the potential limits, small residual absorbance could be observed at the peak positions of the Hoxpdt and Hredpdt states. This was included in the fitting and subtracted as background for the data in Figure 2D. The intensities of each of the four peaks corresponding to the FoxHoxpdt, FredHoxpdt, FoxHredpdt, and FredHredpdt states are then plotted against the applied potential (Figure 2D) and fitted using a model of the Nernst equation based on the scheme shown in Figure 2E (see also Figure S7). This scheme allows electrons to distribute between the proximal F-cluster and the H-cluster and ignores the reduction of the distal F-cluster, as this is unlikely to be detectable by FTIR. Importantly, the data in Figure 2D can only be fitted with a scheme in which the redox midpoint potential of [4Fe−4S]H is dependent on the redox state of the proximal F-cluster. This is clear from the fact that the redox midpoint potentials of the FoxHoxpdt/FoxHredpdt and FredHoxpdt/FredHredpdt transitions are not identical. Likewise, the potential of the proximal F-cluster depends on the redox state of the [4Fe−4S]H (the FoxHoxpdt/ FredHoxpdt and FoxHredpdt/FredHredpdt transitions are not identical). Instead the pairs of midpoint potentials differ by 115 mV. Simulation of the peaks with Voigtian curves gave essentially identical results (Figure S8). Attempts to simulate the data in Figure 2D with the same model but assuming independent midpoint potentials (i.e., forcing the FoxHoxpdt/FoxHredpdt and FredHoxpdt/FredHredpdt potentials to be equal) were not successful (Figure S9). The non-Nernstian behavior observed for the peak areas plotted against the potential in Figure S2C can also be

reproduced using the model in Figure 2E (Figure S10). This demonstrates that a redox interaction is observed even when the minute peak shifts are not considered. The redox interaction between the F- and H-cluster apparently causes the H-cluster cubane to be reduced at a lower potential when the proximal F-cluster is already reduced and vice versa. This mutual redox interaction is known as redox anticooperativity and has been extensively described in the literature for a number of redox proteins.37−42 In particular, MauG, which contains two hemes, was shown to undergo two redox events at potentials separated by ≈100 mV, where the two spectroscopically distinct hemes could be observed as being partially reduced in the intervening potential region.38 Redox anticooperativity can be mediated by electrostatic repulsion between the clusters or redox coupled conformational changes of the protein. A simple way to explain this is that the F-cluster/H-cluster pair accepts an electron that equilibrates between the proximal F-cluster and [4Fe−4S]H. The ratio between the two is determined by the difference in the intrinsic midpoint potentials of the clusters. A second electron enters at lower potential to reduce the remaining clusters. If, indeed, the FoxHoxpdt and FredHoxpdt states can be distinguished in FTIR, the same could be true in EPR. The Hoxpdt state of the H-cluster has a mixed valence Fe(I)Fe(II) [2Fe]H subcluster and an oxidized [4Fe−4S]H subcluster. Overall this state is paramagnetic (S = 1/2) giving a characteristic rhombic EPR spectrum. The Hredpdt state has a reduced [4Fe−4S]H subcluster (S = 1/2), which interacts antiferromagnetically with the [2Fe]H subcluster to give a spin state S = 0 and has, therefore, an EPR silent ground state.6 The F-clusters are EPR silent when oxidized and paramagnetic (S = 1/2) when reduced. As such, in the FredHoxpdt state we expect to observe EPR contributions from both the H-cluster and the proximal F-cluster. If the two paramagnetic species are sufficiently close, as would be expected based on the crystal structure12 and their redox anticooperativity, the two sites should be magnetically coupled. First, we compared the EPR spectra and FTIR spectra of oxidized and partially reduced DdHydAB(pdt) samples (Figure 3A and B). In the oxidized sample a mixture of the Hoxpdt state and a reduced [4Fe−4S] cluster was observed. The latter most likely originates from the distal F-cluster since no spin-coupling is observed to the H-cluster, as would be expected for the reduced proximal F-cluster. In the partially reduced sample, the FTIR spectrum clearly shows a mixture of Hoxpdt and Hredpdt, yet the Hoxpdt state in the EPR spectrum is completely absent. This is because it is spin coupled to the reduced proximal F-cluster giving rise to a completely altered EPR spectrum, with features around g = 2.01. To study this in more detail and compare the redox potential dependence with the FTIR data, we generated 22 EPR samples of DdHydAB(pdt) at different applied potentials in an EPR redox titration (Figure S11). At less negative potentials (−200 to −300 mV) the Hoxpdt spectrum dominates. Between −300 and −450 mV, the Hoxpdt spectrum is lost and both the FredHoxpdt spectrum and the distal [4Fe−4S] cluster spectrum appear. From −450 mV until −560 mV (most negative achieved in this experiment) the FredHoxpdt signal disappears and the [4Fe−4S] cluster signal further increases, with contributions from the coupled spectrum between the distal and proximal F-clusters. These results agree with our interpretation that at intermediate potentials a state exists where the proximal F-cluster is reduced and the H-cluster is oxidized. 15125


Article


The potential dependent changes in the EPR signal intensities of field positions attributable to the different species are shown in Figure S11B−D. The data could be fitted with the same scheme used for the FTIR data presented in Figure 2 using identical midpoint potential values. However, without simulation of the FredHoxpdt spectrum, the relative concentration of this state at each potential cannot be calculated. Hence, the fitting of the EPR redox titration is only qualitative. Therefore, the EPR redox titration data presented here support the observation of redox anticooperativity between the proximal F-cluster and the H-cluster. Recently, the redox potentials of the F-clusters in the [FeFe] hydrogenase from Clostridium pasteurianum (CpI) maturated with [2Fe]pdt were measured using EPR redox titration.43 The authors did not observe similar effects to those we describe here. However, the major feature that we attribute to the spin-coupled H-cluster/proximal F-cluster pair has a g-value around 2.01. Artz and co-workers observed a similar feature in their work, but they attributed it to the Hox−CO state, despite the fact that this Hox−CO was not observed in their FTIR spectra. It would be interesting if the same phenomenon can also be observed in more complex enzymes such as CpI. To our knowledge, this is the first observation of such an effect in the [FeFe] hydrogenases. It is logical to expect similar behavior in the native form of DdHydAB, maturated with [2Fe]adt. It was previously demonstrated that the CO-inhibited form of CrHydA1(adt) shows similar redox behavior to CrHydA1(pdt) (i.e., reduction only occurs on [4Fe−4S]H) but with a lower midpoint potential. This is due to the CO bound to the active site that shifts the electron density from the [2Fe]H subcluster to the [4Fe−4S]H subcluster.31,44 Therefore, we next investigated the CO-inhibited form of DdHydAB(adt). Redox Anticooperativity Observed in CO-inhibited DdHydAB(adt). An analogous FTIR spectro-electrochemical study was carried out on DdHydAB(adt) inhibited with CO which showed very similar behavior to that described above for DdHydAB(pdt). At less negative potential (−280 mV) the sample was in the Hox−CO state (Figure 4A). The Hox−CO/ Hred−CO transition was found to occur at −570 mV (Figure 4B), about 100 mV more negative than in CrHydA1. Again non-Nernstian behavior was observed in the titration of the Hox−CO state with the potential. The behavior changes depending on the wavenumber value followed (Figure S12), and it is caused by small peak shifts (Figure 4C). The peak shifts are observed for all the CO and CN− bands in the FTIR spectrum (Figure S13). Peak fitting with Gaussian functions allowed the separated titrations of the two different Hox−CO states (FoxHox−CO and FredHox−CO) to be extracted (Figure 4C and Figure S14). The data could be fitted with the same scheme used before for DdHydAB(pdt) but using different midpoint potentials (Figure 4D and E). The FoxHred−CO state could not be observed, and was, therefore, not included in the fit. The best fits were achieved using the same level of redox anticooperativity as for the DdHydAB(pdt) data (≈115 mV) but with a slightly more negative midpoint potential for the proximal F-cluster (−405 mV) and a much more negative potential for the H-cluster (−455 mV). The latter is similar to that measured for the CO-inhibited state of CrHydA1(adt) (−470 mV).31 Simulation of the data with independent midpoint potentials gave poorer fits to the data (Figure S15). Our data suggest that the “intrinsic” midpoint potential of the H-cluster in DdHydAB is essentially the same as that in CrHydA1. This agrees well with the sequence and structure

Figure 3. Comparison of FTIR and CW X-band EPR spectra of DdHydAB(pdt). (A) FTIR spectra of DdHydAB(pdt) (100 μM in 50 mM MES, 50 mM HEPES, 150 mM NaCl, pH 8) under N2 without any added reductant and with addition of 50 μM sodium dithionite. (B) CW X-band EPR spectra of the same samples. The upper spectrum is overlaid with a spectral simulation (red dotted line), the components of which are shown underneath. Component 1 is the Hoxpdt spectrum, and component 2 is the distal F-cluster spectrum. Red arrows in (B) indicate the loss of the Hoxpdt EPR spectrum, and red asterisks indicate features attributable to a coupled spectrum in the FredHox state. EPR spectra were measured at 20 K and 0.1 mW microwave power. FTIR spectra were measured at room temperature. 15126


Article


Figure 4. FTIR spectro-electrochemistry of CO-inhibited DdHydAB(adt). FTIR spectra were measured on 2 mM DdHydAB(adt) in 50 mM MES, 50 mM HEPES at pH 8 and 100 mM KCl (as supporting electrolyte) containing redox mediators (see Experimental Section) at various applied potentials in a spectro-electrochemical cell. Selected spectra at −280 mV, −470 mV and −600 mV are shown (A) highlighting the FoxHox−CO (dark blue), FredHox−CO (light blue), and FredHred−CO (red) states. (B) Change in FTIR absorbance with redox potential at the peak positions 2017 cm−1 (blue circles) and 2001 cm−1 (red circles), where non-Nernstian behavior is clearly observed. The blue and red lines correspond to n = 1 Nernst fits with Em = −570 mV for the oxidized and reduced species, respectively. (C) Close-up view of the 2000 to 2030 cm−1 region at −280 and −470 mV, highlighting the peak shifts. The peaks have been fit with Gaussian functions in order to extract the relative contributions from the FoxHox−CO, FredHox−CO, and FredHred−CO states. The relative contributions from each state are plotted against the redox potential (D), and the data are fitted with Nernst equations based on the model shown in (E) (see also Figure S7). The fitted redox potentials are shown in (E). All potentials are quoted versus the standard hydrogen electrode (SHE). 15127


Article

Journal of the American Chemical Society conservation of these two enzymes. The apparent midpoint potential of the H-cluster in DdHydAB is shifted due to redox anticooperativity, as a consequence of the proximal F-cluster being reduced first. This explains why the FoxHred−CO state is not observed in our titrations. The proximal F-cluster has a more positive potential, and so this is reduced preferentially. It is likely that the same phenomenon is responsible for the low midpoint potential of the HredH+/HsredH+ transition in native DdHydAB.32 As was observed for DdHydAB(pdt), the redox states of the various cofactors in CO-inhibited DdHydAB(adt) can be observed with EPR spectroscopy (Figure 5). For CO-inhibited DdHydAB(adt) without added reductant the EPR spectrum displays the Hox−CO spectrum and a spectrum from the reduced distal F-cluster. Reduction with sodium dithionite does not generate any of the Hred−CO state, as observed by FTIR (Figure 5A), yet the EPR spectrum is completely altered, due to spin-coupling between the H-cluster and the reduced proximal F-cluster. Since under these conditions the sample was purely in the FredHox−CO state, we investigated the spin-coupling further by measuring the samples using higher frequency (Q-band and W-band) EPR spectroscopy (Figure S16). At Q-band, the spectrum is radically altered as the effect of the spin−spin coupling is diminished. This is because the Zeeman interaction starts to dominate the zero-field interactions at higher magnetic fields. The spin-coupling interaction is independent of the microwave frequency, and thus appears smaller on the g-value scale. The spectrum at W-band is very similar to that at Q-band suggesting the spin-coupling interactions are already smaller than the line broadening at Q-band. This confirms that the FredHox−CO spectrum derives from spin-coupling of the H-cluster and the proximal F-cluster. The redox anticooperative effect described so far for DdHydAB(pdt) and the CO-inhibited form of DdHydAB(adt) is predicted to have some influence on the redox properties of the active enzyme. However, due to the ability to protonate the H-cluster at the bridging nitrogen of the ADT ligand, the effects are likely to be complex. It was recently demonstrated that protonation of the H-cluster in CrHydA1(adt) is coupled to electronic rearrangement, with transfer of an electron from the [4Fe−4S]H subcluster to the [2Fe]H subcluster.19 We speculate that redox anticooperativity may influence this process. Therefore, we next performed spectro-electrochemistry with DdHydAB(adt) at several different pH values. Effect of Redox Anticooperativity in Active DdHydAB(adt). Spectro-electrochemical measurements at various pH values (pH 6, 8, 9, and 10) were performed in which the active redox states of the enzyme were titrated with the applied potential. Selected data are presented in Figure 6A, and the titrations of the main peaks of each state at pH 6, 8, 9, and 10 are shown in Figure 6B−E. At pH 8, DdHydAB(adt) gave data similar to those reported previously for this enzyme.32,45 At less negative potential (−200 mV) the Hox state (main CO stretch at 1940 cm−1) dominates. Between −300 and −500 mV, the Hox state is replaced by the HredH+ state (main CO stretch at 1894 cm−1). At a very negative potential the HredH+ state population decreases and is partially replaced by the HsredH+ state (main CO stretch at 1883 cm−1). It should be noted that at very negative potentials H+ reduction becomes significant and leads to deviations from thermodynamic equilibrium. In addition, at −400 mV a peak at 1934 cm−1 can be identified. We assign this to the (unprotonated) Hred state (main CO stretch at 1934 cm−1) that was recently reported for

Figure 5. Comparison of FTIR and CW X-band EPR spectra of COinhibited DdHydAB(adt). (A) FTIR spectra of CO-inhibited DdHydAB(adt) (100 μM in 50 mM MES, 50 mM HEPES, 150 mM NaCl, pH 8) under N2 without any added reductant and with addition of 10 mM sodium dithionite. (B) CW X-band EPR spectra of the same samples. The upper spectrum is overlaid with a spectral simulation (red dotted line), the components of which are shown underneath. Component 1 is the Hox−CO spectrum and component 2 is the distal F-cluster spectrum. Red arrows on (B) indicate the loss of the Hox−CO EPR spectrum, and red asterisks indicate features attributable to a coupled spectrum in the FredHox−CO state. EPR spectra were measured at 20 K and 0.1 mW microwave power. FTIR spectra were measured at room temperature.

CrHydA1(adt).19,46 The two singly reduced states Hred and HredH+ in CrHydA1(adt) are related to each other by protoncoupled electronic rearrangement.19 While the Hred state is characterized by a reduced [4Fe−4S]H and a mixed valence [2Fe]H subcluster, protonation of the adt bridge is coupled to electron transfer from [4Fe−4S]H to [2Fe]H affording the HredH+ state with an oxidized [4Fe−4S]H and a reduced [2Fe]H in the Fe(I)Fe(I) configuration. 15128


Article


Figure 6. FTIR spectro-electrochemistry of DdHydAB(adt) at pH 6, 8, 9, and 10. FTIR spectra were measured on 1−2 mM DdHydAB(adt) samples in a mixed buffer system containing redox mediators (see Experimental Section) at pH 6, 8, 9, and 10 over a range of applied potentials in a spectroelectrochemical cell. Selected spectra are shown (A) highlighting the four states (Hox, Hred, HredH+, and HsredH+) observed under the various conditions. Two spectra are shown from the pH 8 titration to highlight how the Hox, Hred, and HredH+ features are changing during this titration. The main peaks from each state are colored blue (Hox), red (Hred), green (HredH+), and purple (HsredH+), and all other peaks assigned to these states are indicated with dashed vertical lines. Unassigned peaks at 2016, 1972, and 1812 cm−1 belong to small amounts of the CO-inhibited state and at 1988, 1921, and 1843 cm−1 are due to incorporation of small amounts of degraded [2Fe]adt complex during maturation. These features do not significantly overlap with the main peaks from the active states and do not significantly change during the potential titration; hence, they do not interfere with the interpretation of these data. The potential titration of the main peaks of the four states at pH 6 (B), 8 (C), 9 (D), and 10 (E) are shown. All potentials are quoted versus the standard hydrogen electrode (SHE). It should be noted that, at very negative potential, catalytic H+ reduction becomes significant and leads to deviations from thermodynamic equilibrium. 15129


Article


Figure 7. Fitting of the FTIR spectro-electrochemistry of DdHydAB(adt) at pH 6, 8, 9, and 10. FTIR spectra were measured on 0.5−2 mM DdHydAB(adt) samples at various applied potentials in a spectro-electrochemical cell at pH 6, 8, 9, and 10. Two spectra are shown from the pH 8 titration to highlight how the Hox, Hred, and HredH+ features are changing during this titration. The relative contributions at the main peak positions from the FoxHox (dark blue), FredHox (light blue), FoxHred (red), FredHred (orange), FoxHredH+ (dark green), and FredHredH+ (light green) states were estimated by fitting with Gaussian curves (A) and are plotted against the applied potential at pH 6, 8, and 10 (B). The data are fitted using equations derived from the scheme shown in Figure S18, and selected fitting parameters are shown in the box in this figure (for an extended list of the parameters see Table S1). In (A), the purple area represents the sum of the line shapes. The Hsred states were not taken into account due to their low contribution to the FTIR spectra. The dotted lines in (B) indicate the equilibrium H2/2H+ potential at each pH. The unassigned peak at ≈1888 cm−1 derives from small amounts of degraded [2Fe]adt cofactor incorporation during artificial maturation and does not contribute significantly to the spectra at any potential.

In CrHydA1(adt), which lacks the F-clusters, the Hred and HredH+ states have the same potential dependence and their relative populations are governed solely by a protonation equilibrium, defined by a pKa of ≈7.2.19 In DdHydAB(adt) the two singly reduced states initially appear together, but the Hred state is then converted to HredH+ around the point at which we expect the proximal F-cluster to be reduced. It is clear, therefore, that the process of proton-coupled electronic rearrangement in the H-cluster is affected by the redox behavior of the proximal F-cluster in DdHydAB(adt). At pH 10, a greater amount of Hred is observed during the redox titration. This dominates over HredH+ at less negative potentials. At more negative potentials both have similar contributions. At pH 6 the Hred state is not observed at all. These data suggest that the pKa of the one-electron reduced Hred state is shifting during the redox titration. The pKa appears to be close to 8 at less negative potentials and closer to 10 at more negative potentials. This can be explained considering the redox anticooperative effect.

A closer inspection of the potential dependence of the main peaks in each of the H-cluster redox states at each pH value reveals that the peaks also show non-Nernstian behavior (Figure S17). This is again caused by peak shifts, as was observed for DdHydAB(pdt) and CO-inhibited DdHydAB(adt). Again, we attribute this behavior to reduction of the proximal F-cluster. Therefore, for every H-cluster state (Hox, Hred, HredH+, and HsredH+) a counterpart with a reduced proximal F-cluster might be observable through minute band shifts in its FTIR spectrum. In Figure 7A, the FTIR marker bands of Hox, Hred, and HredH+ are analyzed at pH 6, 8, 9, and 10 for the redox event at the proximal F-cluster. Subsequently, the line shapes were simulated using Gaussian curves at fixed positions corresponding to the redox state of the proximal F-cluster. The amplitudes of these components are plotted against the applied potentials for pH 6, 8, and 10 (Figure 7B). As can be seen in Figure 7A, the HredH+ state at pH 8 is already associated with a reduced proximal F-cluster: FredHredH+. Only at pH 6 is the FoxHredH+ state populated. The unprotonated singly reduced state 15130


Article


Figure 8. Scheme describing how redox anticooperativity influences the catalytic cycle in DdHydAB(adt). Each state in the catalytic cycle is depicted with cubes representing the proximal cluster and [4Fe−4S]H and a long rectangle representing [2Fe]H. Green color indicates an oxidized site (proximal F-cluster, [4Fe−4S]H or [2Fe]H), yellow indicates a reduced site, and blue indicates a superoxidized site (only for FoxHhyd). Hydrogen production follows the cycle in the clockwise direction while hydrogen oxidation follows the cycle in the anticlockwise direction. The conversion of FoxHred to FredHredH+ can follow two pathways depending on conditions: either reduction of the proximal F-cluster precedes protonation of the bridging amine, forcing electron transfer to [2Fe]H and driving protonation, or protonation occurs directly without requiring F-cluster reduction. During hydrogen oxidation deprotonation forces electron transfer from [2Fe]H to [4Fe−4S]H leading to electron transfer from the proximal Fcluster to the distal F-cluster. The colored states were observed in the experiments described in this study, while the gray states are hypothetical. The FoxHsredH+ state (shown in the square brackets) is bypassed here to indicate that it may not be essential for the catalytic cycle in the F-cluster containing hydrogenases.

pH should be similar to the inability to protonate the H-cluster when the bridgehead atom is replaced with carbon in [2Fe]pdt. The intrinsic pKa of the FoxHred to FoxHredH+ transition (pK2 ≈ 7.7) is similar to that determined in CrHydA1(adt) (≈ 7.2),19 which agrees with the fact that the active sites of the two enzymes have highly conserved structures, and that the apparently high pKa (pK4 = 9.3) in DdHydAB(adt) is due to destabilization of FredHred by redox anticooperativity between the proximal F-cluster and [4Fe−4S]H. This is the same effect that decreases the redox potential at which the HsredH+ state can be observed in comparison to the situation for CrHydA1.18 We predict that the observed redox anticooperativity may influence catalysis: during hydrogen production, a fully reduced set of F-clusters favors a ground state where the electron is localized on [2Fe]H, which drives protonation of the H-cluster, even when the concentration of protons is low (Figure 8). Similarly, deprotonation of the H-cluster during H2 oxidation is coupled to electron transfer from [2Fe]H to [4Fe−4S]H, forcing electron transfer from the proximal F-cluster to the distal F-cluster. This complex interplay between the deprotonation and electron transfer steps should allow rapid communication between the two ends of the electron transfer chain, an effect suggested previously based on theoretical calculations.47 A further consideration is that in CrHydA1, which lacks F-clusters, H+ reduction must proceed from a two-electron reduced H-cluster in the HsredH+ state. The second electron enters at moderately lower potential than the first, potentially requiring a modest overpotential before catalysis proceeds. In DdHydAB, formation of the HsredH+ state is not required, because the second electron can remain in the F-clusters and

(FoxHred) is mainly populated at high pH (>8) and the FredHred state is best observed at pH 10 and a very negative potential (−560 mV). Since, each oxidation state of the H-cluster can be protonated and/or associated with a reduced proximal F-cluster, the fitting of the data is challenging. In order to extract the intrinsic potentials and pKa values for all these different processes the relative contributions from each state at different potentials and pH values (Figure 7B) were fitted using the scheme shown in Figure S18, similar to those used for the DdHydAB(pdt) and CO-inhibited DdHydAB(adt), but including the possibility for protonation of each state. The Hsred states (FoxHsred, FredHsred, FoxHsredH+, and FredHsredH+) are excluded in this scheme because only very small contributions from them are observed, and only at high pH. The fits to this scheme are shown as solid lines in Figure 7B, and the associated parameters are presented in the accompanying table (extended in Table S1). The same trends are observed at each pH: the FoxHox state decays initially as the potential is lowered and is replaced by a mixture of one electron reduced states. At pH 6 this is a mixture of the FoxHredH+ and FredHox states, while at pH 8 the FredHox state dominates, and at pH 10 there is a mixture of the FredHox and FoxHred states. As the potential is further decreased the enzyme is reduced by a second electron. At pH 6 and 8 the FredHredH+ state is exclusively observed, while at pH 10 almost equal contributions from the FredHredH+ and FredHred states exist. At pH 10 the behavior is quite similar to that observed for DdHydAB(pdt) with similar midpoint potentials for the two reduction events. This makes sense since the difficulty to protonate the bridgehead nitrogen atom at the H-cluster at high 15131


Article


to ensure the potential stability during the course of the experiment. The potential was controlled by an Autolab PGSTAT101 potentiostat using Nova software with an equilibration time of 30 min at each potential. Spectra were collected in the double-sided, forward− backward mode with a resolution of 2 or 0.2 cm−1, an aperture setting of 1.5 mm, and a scan velocity of 20 Hz. Data were processed using home-written routines in the MATLAB environment. The temperature of the cell was maintained using a water circulator system (Huber, Offenburg). All other parameters were varied and are detailed in the figure legends. All potentials referred to in the text are quoted versus the standard hydrogen electrode (SHE). X-band (≈9.6 GHz) EPR samples (200 μL) were prepared in an anaerobic chamber and transferred to 4 mm (o.d.) quartz EPR tubes, frozen in liquid nitrogen, and measured on a Bruker ELEXSYS E500 CW X-band EPR spectrometer with an Oxford Instruments ESR900 helium flow cryostat connected to an ITC503 temperature controller. The parameters used were as follows: microwave frequency 9.64 GHz, time constant 81.92 ms, conversion time 81.92 ms, modulation frequency 100 kHz. All other parameters varied and are given in the figure legends. EPR spectral simulations were performed using Easyspin.51 Q-band (≈33.6 GHz) EPR spectra were measured on a Bruker ELEXYS E580 Q-band pulsed EPR spectrometer with a SuperQ-FT microwave bridge and a home-built resonator described earlier.52 The magnetic field was calibrated using a Bruker ER035D NMR gaussmeter. Cryogenic temperatures were maintained using a custom-made closed cycle Helium Cryostat (Cryogenic Ltd.). W-band (≈94.0 GHz) EPR experiments were performed using a Bruker ELEXSYS E680 pulsed EPR spectrometer. Cryogenic temperatures were obtained using an Oxford Instruments Helium flow cryostat. Spectra were collected at Q- and W-band using free induction decay (FID) detected EPR at 20 K with a microwave pulse length of 1 μs. For EPR redox titrations, 0.2 mM DdHydAB was mixed with 0.03 mM of the same redox mediators used for FTIR spectro-electrochemistry. The potential was adjusted using concentrated stock solutions of sodium dithionite and potassium ferricyanide, and the potential was measured in a two-electrode electrochemical cell. A pyrolytic graphite working electrode, an Ag/AgCl (sat. KCl) electrode was used as a reference, and the potential was recorded using a potentiometer. At ≈10 mV intervals, 0.25 mL samples were transferred to X-band EPR tubes and frozen in liquid nitrogen, before being measured at X-band as described above. Data fitting was performed by using the equations derived in the Supporting Information together with home-written routines in the MATLAB environment. A grid search was used to attain the global minimum based on the least squared deviation (χ2 value) from the experimental data. Error ranges were estimated by comparing fits obtained with a range of parameter sets. The error ranges presented in Table S1 (ΔE = ± 10 mV and ΔpK = ± 0.1) caused minimal deviations (

Intercluster Redox Coupling Influences Protonation at the H-cluster in

Recommend Documents