Site-selective Protonation of the One-electron Reduced Cofactor in [FeFe]-Hydrogenase

Hydrogenases are microbial redox enzymes that catalyze H2 oxidation and proton reduction (H2 evolution). While all hydrogenases show high oxidation activities, the majority of [FeFe]-hydrogenases are excellent H2 evolution catalysts as well. Their active site cofactor comprises a [4Fe-4S] cluster covalently linked to a diiron site equipped with carbon monoxide and cyanide ligands that facilitate catalysis at low overpotential. Distinct proton transfer pathways connect the active site niche with the solvent, resulting in a non-trivial dependence of hydrogen turnover and bulk pH. To analyze the catalytic mechanism of [FeFe]-hydrogenase, we employ in situ infrared spectroscopy and infrared spectro-electrochemistry. Titrating the pH under H2 oxidation or H2 evolution conditions reveals the influence of site-selective protonation on the equilibrium of reduced cofactor states. Governed by pKa differences across the active site niche and proton transfer pathways, we find that individual electrons are stabilized either at the [4Fe-4S] cluster (alkaline pH values) or at the diiron site (acidic pH values). This observation is discussed in the context of the natural pH ABSTRACT Hydrogenases are microbial redox enzymes that catalyze H 2 oxidation and proton reduction (H 2 evolution). While all hydrogenases show high oxidation activities, the majority of [FeFe]-hydrogenases are excellent H 2 evolution catalysts as well. Their active site cofactor comprises a [4Fe-4S] cluster covalently linked to a diiron site equipped with carbon monoxide and cyanide ligands that facilitate catalysis at low overpotential. Distinct proton transfer pathways connect the active site niche with the solvent, resulting in a non-trivial dependence of hydrogen turnover and bulk pH. To analyze the catalytic mechanism of [FeFe]-hydrogenase, we employ in situ infrared spectroscopy and infrared spectro-electrochemistry. Titrating the pH under H 2 oxidation or H 2 evolution conditions reveals the influence of site-selective protonation on the equilibrium of reduced cofactor states. Governed by pKa differences across the active site niche and proton transfer pathways, we find that individual electrons are stabilized either at the [4Fe-4S] cluster (alkaline pH values) or at the diiron site (acidic pH values). This observation is discussed in the context of the natural pH dependence of hydrogen turnover as catalyzed by [FeFe]-hydrogenase.


INTRODUCTION
[FeFe]-hydrogenases are gas-processing metalloenzymes that have been found in bacteria and green algae. [1][2][3] They serve various roles in the hydrogen metabolism of prokaryotes, including oxidation of H2 as an energy source and proton reduction (H2 evolution) to maintain the cellular redox equilibrium. 4 In the chloroplast of green algae, they are part of the photosynthetic electron transport chain, coupling H2O oxidation and H2 evolution at the reducing end of photosystem I. 5 The first crystal structures of [FeFe]hydrogenase helped identifying accessory and catalytic iron-sulfur clusters as well as gas channels and potential proton transfer (PT) pathways. [6][7][8][9][10][11][12] Additionally, various biophysical techniques were employed to characterize the electronic structure of the active site cofactor, the so-called 'H-cluster' ( Figure 1AB). 1 This iron-sulfur compound is formed by a [4Fe-4S] cluster connected to a bimetallic iron site via a bridging cysteine residue. Carbon monoxide (CO) and cyanide ligands (CN -) tune the redox potential of the H-cluster and anchor the diiron site within the protein. [13][14][15] The secondary amine of the H-cluster's aminodithiolate ligand (ADT) has been suggested to serve as proton relay between the diiron site and the amino acid residues of the catalytic PT pathway. [16][17][18][19][20][21] Moreover, it acts as an inner-sphere hydrogenbonding donor to a number of apical ligands at the distal iron ion (Fed). 22 Figure 1C depicts a schematic representation of the H-cluster with potential binding sites for hydrogen species. cluster) and Hred (reduced diiron site) as well as the 'super-reduced' state, Hsred. [23][24][25][26][27][28][29][30] An additional 2e --reduced state, Hhyd, is comprised of a reduced [4Fe-4S] cluster and a formally over-oxidized diiron site with a terminal hydride ligand (H -). [31][32][33] Additional states include the CO-inhibited states Hox-CO and Hred´-CO and the oxidized protonated state, HoxH. 27,30 While most authors agree on the importance of Hred´ and Hhyd in hydrogen turnover, the protonation state, cofactor geometry, and involvement in catalysis of Hred and Hsred are under discussion. 34 Both Hred´ and Hred are enriched upon proton-coupled electron transfer (PCET). Previously, we suggested that a cysteine residue coordinating the [4Fe-4S] cluster may bind a proton in Hred´ ( Figure   1B) 29 ; however, there is no consensus regarding the nature of protonation and cofactor geometry in Hred.
Sommer et al. presumed protonation of the ADT ligand ( + NH2) and a shift of the µCO ligand into a 'semibridging' position at Fed as seen in a crystal structure grown under H2. 35 38,39 In contrast, our infrared evaluation of Hred and Hsred at ambient temperatures implied the formation of a bridging hydride species (µH) and an apical CO ligand at Fed. 28 Such changes would exclude Hred and Hsred from the catalytic cycle as a µH geometry was calculated to be rather unreactive. [40][41][42] Following the time-dependent evolution of individual H-cluster bands by transient infrared spectroscopy at ambient temperature, Sanchez et al. proposed the kinetic competence of both Hred´ and Hred. 43 Unfortunately, the spectroscopic marker bands that were used to follow Hred in this study are nearly identical for the µCO and the µH geometry, impeding any kinetic discrimination. 38,39 We presume that the cryogenic states may represent kinetically trapped intermediates. 44 In this study, we investigate the pH-dependent accumulation of 1e --reduced H-cluster states in the [FeFe]hydrogenase from Chlamydomonas reinhardtii, CrHydA1, to understand the equilibrium of Hred´ and Hred under turnover conditions. Making use of in situ attenuated total reflection Fourier-transform infrared (ATR FTIR) spectroscopy and spectro-electrochemistry under H2 oxidation or H2 evolution conditions, we found consistent trends for an accumulation of Hred´ towards alkaline pH values whereas the accumulation of Hred increases towards acidic pH values. This observation is explained by siteselective PCET to either the diiron site or the [4Fe-4S] cluster, guided by differences in the proton affinity.
Our findings are employed to distinguish catalytic from regulatory H-cluster states and inspire a molecular understanding of the pH-dependent hydrogen turnover of [FeFe]-hydrogenase.

EXPERIMENTAL
Protein purification and activation. All experiments involving CrHydA1 were performed under strictly anaerobic conditions. Native apo-protein and amino acid variants C417SDH were expressed in Escherichia coli host strain BL21 (DE3) ΔiscR and purified by strep-tactin affinity chromatography as described previously. 45,46 The synthetic mimics of the native ADT complex and artificial propanedithiolate complex (PDT) were synthesized following literature procedures. 17 Apo-protein was activated in the presence of ADT or PDT at a 10-fold molar excess. After an incubation period of at least 1 h at 25 °C, size-exclusion chromatography was employed to remove redundant complex. 18 Activated CrHydA1 was eluted in 10 mM Tris/HCl (pH 8). For native CrHydA1, sodium dithionite was avoided to prevent accumulation of HoxH and Hhyd at low pH values. 30,32 Enzyme was concentrated to ~1 mM and stored anaerobically at −80 °C.
ATR FTIR spectroscopy. The FTIR spectrometer (Tensor27, Bruker) was equipped with a triple-reflection ZnSe/Si crystal ATR cell (Smith Detection) and placed in an anaerobic chamber. Infrared spectra were recorded with 80 kHz scanning velocity at a spectral resolution of 2 cm -1 (MCT detection). Under these conditions, the time-resolution of data acquisition is in the range of seconds (i.e., five interferometer scans in forward/ backward direction). ATR FTIR measurements were performed at 25°C and on hydrogenase films derived by controlled dehydration and rehydration of 1 µl protein sample as reported earlier. 27  was added to the N2 stream via separate flow controllers and passed through a wash bottle containing 150 mL mixed buffer (0.1 -100% at ambient pressure). The aerosol was fed into a gas-tight polychlorotrifluoroethylene (PCTFE) compartment, attached on top of the ATR crystal plate and equipped with six optional gas inlets, a manometer for pressure control, and a transparent glass window for UV/Vis irradiaton. 27 For each H2 concentration step, the film was equilibrated for 2.5 min ( Figure S1).
ATR FTIR spectro-electrochemistry. The pH-dependent reduction of CrHydA1 in the absence of H2 was analyzed by ATR FTIR spectro-electrochemistry. 28,29 For this, 1 µL protein sample (diluted with 50 mM mixed buffer pH 9 -5) was injected into a 9 µm thin gold mesh on top of an ATR silicon crystal. The mesh was covered with 8 kDa dialysis membrane to protect the film from dilution. A custom-made PCTFE electrochemical cell was attached to the ATR crystal plate and filled with 3 mL electrolyte buffer (50 mM mixed buffer pH 9 -5 including 500 mM KCl as electrolyte), that was purged with N2 throughout the whole experiment. After 60 -90 minutes, the film was fully hydrated. The gold mesh was connected with the working electrode, platinum wire was used as counter electrode, and an Ag/AgCl electrode served as reference (+230 mV vs SHE, as determined with 1 mM methyl viologen at pH 7). 29 After complete oxidation at -100 mV vs SHE, the potential was lowered incrementally from -150 mV to -850 mV in steps of 50 mV with a fixed duration of 20 minutes for each step ( Figure S2) until no further spectral changes were observed. Midpoint potentials were estimated from bi-sigmoidal fits. Although the lack spectral changes after 20 min suggested steady-state conditions, smaller changes in current hinted at imperfect equilibria, in particular at strongly reducing potential ( Figure S2). Thus, we refrained from Nernst-fitting and a quantitative, comparative analysis of midpoint potentials.
Data treatment. All absorbance spectra were derived from single channel spectra of reference (ZnSe/Si) and sample (ZnSe/Si + protein) in OPUS software. Then, data was exported to a home-written routine, as described previously. 30 In the frequency regime of the H-cluster (2150 -1750 cm -1 ), absorbance spectra were subtracted with a polynomial function simulating the broad combination band of liquid water underneath the sharp CO/CNbands of the H-cluster ( Figures S1A and S2A). This gave rise to baselinecorrected spectra as shown in Figures S1B and S2B. Reference spectra (Figure 2 and Figure S3) allowed determining fit parameters for all observed redox states (frequency, intensity, band width, and peak ratio, see Table S1), as described earlier. 30 The sum of peak area (2 CN -+ 3 CO) for a given redox state was obtained by simulation of spectral data with the fixed fit parameters that represent the population in relation to the population of the other redox states. This value (%) and was plotted against time illustrating how the system converges into new redox equilibria upon disturbance (i.e., changes in H2 concentration in Figure S1C or electrochemical potential in Figure S2C). For the last step of data evaluation, the population of redox states was plotted as a function of H2 concentration or electrochemical potential.

RESULTS
All experiments were performed with the [FeFe]-hydrogenase CrHydA1 under ambient conditions. In the first step, ATR FTIR spectroscopy and spectro-electrochemistry [27][28][29] were employed to extract the IR signatures of all relevant redox states (Table 1 and Figure S3). Figure 2A shows how ATR FTIR spectroelectrochemistry at alkaline or acidic conditions facilitated recording difference spectra for the transition from Hox into Hred´ (top, pH 9, -650 minus -450 mV vs SHE) and Hred into Hsred (bottom, pH 5, -750 minus -550 mV vs SHE). The overall downshift of the cofactor bands from Hox → Hred´ and Hred → Hsred has been attributed to a reduction of the [4Fe-4S] cluster ( Figure 2B). 25,26 Opposed to the Hox → Hred´ difference spectrum, the Hred → Hsred difference spectrum shows no signal around 1800 cm -1 (yellow mark-up, dashed line). This highlights the lack of a µCO ligand at the reduced diiron site ( Figure   2B). Moreover, note the absence of other H-cluster species, which confirms the assignment of bands at 1961 cm -1 and 1953 cm -1 to Hred and Hsred, respectively (blue mark-up). The small band at 1972 cm -1 is unrelated to any known redox state. It has been assigned to Hhyd:red under cryogenic conditions 39 ; however, the spectrum lacks the respective µCO band at 1851 cm -1 . Therefore, Hred and Hsred are depicted with a terminal CO ligand and a µH ligand instead of a µCO ligand in Figure 2B. 28 Dabs.   32 Moreover, H2 oxidation induces an accumulation of reduced Hcluster states so that we were able to follow the increase and decrease of redox state populations as a function of atmospheric H2 and time. As an example, Figure 3A depicts a series of baseline-corrected ATR FTIR absorbance spectra in the CO regime of the H-cluster recorded after 2.5 min under 0 -100% H2 (steady-state conditions, compare Figure S1). We note that residual, unidentified H-cluster states were 10 present in the spectra. As these contribution did not to interfere with the global fit analysis (i.e., χ 2 < 10 -4 ) we did consider them any further. Figure 3B illustrates how increasing the H2 concentration from 0 -0.1% resulted in an accumulation of 1e --reduced states Hred´ and Hred, which remained fairly stable between 0.1 -3% H2. At higher concentrations of H2, the 2e --reduced Hsred state dominated the spectrum. Only minor traces of the 2e --reduced Hhyd state were observed, most likely due to the lack of sodium dithionite in the sample (see Experimental section). 32 Analogous to the experiment shown in Figure 3, six individual CrHydA1 protein films between pH 10 -5 were analyzed (Figure 4). Largely independent of pH, the oxidized state Hox was the most prominent species in the absence of H2 while Hsred dominated the spectrum for H2 > 10% ( Figure S4). The steadystate population of Hred´ and Hred is plotted as a function of H2 and at different pH values in Figure   4AB. Here, we observed diverging trends for the accumulation of the 1e --reduced states as highlighted  Figure 4C depicts the accumulation of Hred´ and Hred at 3% H2 as a function of pH, which clearly illustrates this trend. At low pH and H2 > 10%, an increasing accumulation of Hhyd was observed, which may explain the mild suppression of Hsred that was otherwise expected to follow the same pH dependence as Hred ( Figure S4). Upon removal of H2 from the gas stream, the 2e --reduced states Hhyd and Hsred converted transiently into the 1e --reduced states Hred´ and Hred, indicating intermolecular electron transfer in the dense films 30 , before the equilibrium shifted back towards Hox upon autooxidation. 22 The diverging pH dependence for the steady-state accumulation of Hred´ and Hred in the presence of H2 was found to be well conserved in this transient increase, emphasizing the robustness of all trends observed for of the 1e --reduced states ( Figure S5). The simultaneous presence of Hred´ and Hred complicates unique conclusions regarding the mechanism of H-cluster protonation. Therefore, additional experiments were performed. First, we probed the pH dependence of H2 oxidation with CrHydA1 PDT . This cofactor variant lacks the secondary amine of the natural ADT ligand ( Figure 1B) and allows analyzing the reduction of the [4Fe-4S] cluster (i.e., the Hox/Hred´ transition) independent of redox chemistry at the diiron site. 26,29 Figure S6 shows that the H2 oxidation activity of CrHydA1 PDT increases between pH 10 -8. While this cannot be explained by the 5  The influence of cysteinyl ligand C417 on the catalytic properties of CrHydA1 has been addressed by site-directed mutagenesis earlier. 46,47 Here, we analyzed three cysteine variants to compare the composition of H-cluster states under H2 oxidation conditions ( Figure S7). (i) C417S behaved much like wild-type CrHydA1 but showed a reduced percentage of Hred´ under H2. (ii) Due to electron withdrawal from the [4Fe-4S] cluster by the imidazole ligand, C417H was reported with a less negative redox potential than wild-type CrHydA1. 47 In agreement with earlier observations, C417H adopted Hred´ as a resting state and even after 12 -18 h under N2 we did not observe accumulation of Hox, reflecting the lack of H2 evolution activity of CrHydA1 C417H. 47 In the presence of H2, the variant converted into In contrast to conventional, Moss-type 48 transmission cells, the ATR FTIR spectro-electrochemistry approach allowed H2 to be released from the protein film which precluded product oxidation. Figure 5 13 depicts how the oxidized states Hox and Hox-CO were lost at reductive potentials, followed by accumulation of the 1e --reduced states Hred´ and Hred. Upon further reduction, accumulation of Hsred was observed; however, in contrast to the experiments performed under H2 oxidation conditions ( Figure   3), Hhyd was not observed at all. The latter observation may hint at different catalytic pathways for H2 oxidation and H2 evolution. 21 14 4). The oxidized state Hox was the most prominent species at potentials more positive than -350 mV vs SHE, largely independent of pH. The accumulation of Hsred was found to be affected by pH more drastically, with a higher population at acidic pH values that reflects the lack of Hhyd in the experiment ( Figure S8). Overall, we observed a mean midpoint potential around -650 mV vs SHE for Hsred, which leaves a potential window of ~300 mV to analyze the accumulation of the 1e --reduced states. In Figure   6AB the   (including Hhyd and Hsred), unraveling the PCET chemistry of Hred´ and Hred was found to be challenging. 35 To this end, ATR FTIR spectro-electrochemistry facilitated analyzing the population of 1e --and 2e --reduced H-cluster states individually. The accumulation of Hred at acidic pH values and mildly reducing conditions (e.g., -350 mV vs SHE) suggests an apparent pKa < 6 , in agreement with the involvement of glutamic acid residues as 'bottle neck' in the catalytic PT pathway. 19 In contrast, our data for the population of Hred´ at more reducing potentials (e.g., -550 mV vs SHE) hints at an apparent pKa > 8, which reflects the greater ease of proton transfer to the [4Fe-4S] cluster via bulk solvent 30,34 and is in 16 excellent agreement with the formal pKa of 8. Both in vivo and in vitro, the pH dependence of H2 evolution of [FeFe]-hydrogenase shows a bell-shaped distribution with a maximal activity around neutral or mildly alkaline pH values. [50][51][52][53] This has recently been confirmed by bulk electrochemistry. 54 While the increase in H2 evolution activity between pH 9 -7 can be attributed to rising proton concentration, our data now allows correlating the activity decrease between pH 7 -5 to the formation of Hred, which clearly dominated over Hred´ at acidic pH values.
This behavior is in agreement with Hred and Hsred as 'H2-inhibited' states 55 that bind a bridging hydride at the diiron site 28 and have been shown to play a key role in sensory [FeFe]-hydrogenases. [56][57][58] The ligand flip required to form a reactive terminal hydride geometry disfavors fast catalysis. [40][41][42] The present data support the theory that reduction and site-selective protonation at the [4Fe-4S] cluster adjusts the redox potential of the H-cluster to stabilizes a reactive geometry necessary for efficient hydrogen catalysis. 34 We suggest that similar concepts may give rise to a novel generation of biomimetic hydrogen catalysts.