The crystalline state as a dynamic system: IR microspectroscopy under electrochemical control for a [NiFe] hydrogenase

Controlled formation of catalytically-relevant states within crystals of complex metalloenzymes represents a significant challenge to structure–function studies. Here we show how electrochemical control over single crystals of [NiFe] hydrogenase 1 (Hyd1) from Escherichia coli makes it possible to navigate through the full array of active site states previously observed in solution. Electrochemical control is combined with synchrotron infrared microspectroscopy, which enables us to measure high signal-to-noise IR spectra in situ from a small area of crystal. The output reports on active site speciation via the vibrational stretching band positions of the endogenous CO and CN− ligands at the hydrogenase active site. Variation of pH further demonstrates how equilibria between catalytically-relevant protonation states can be deliberately perturbed in the crystals, generating a map of electrochemical potential and pH conditions which lead to enrichment of specific states. Comparison of in crystallo redox titrations with measurements in solution or of electrode-immobilised Hyd1 confirms the integrity of the proton transfer and redox environment around the active site of the enzyme in crystals. Slowed proton-transfer equilibria in the hydrogenase in crystallo reveals transitions which are only usually observable by ultrafast methods in solution. This study therefore demonstrates the possibilities of electrochemical control over single metalloenzyme crystals in stabilising specific states for further study, and extends mechanistic understanding of proton transfer during the [NiFe] hydrogenase catalytic cycle.


Single-crystal IR microspectroscopic-electrochemical experiments
Electrochemical cell design Figure S1 shows a schematic representation of the electrochemical cell used for microspectroscopicelectrochemical measurements, based around previously-reported electrochemical cells. [1][2][3][4] A glassy carbon rod working electrode (4 mm diameter, Alfa Aesar) was sealed into a brass connector using silver-loaded epoxy resin (RS Components). This working electrode assembly was then sealed, using epoxy resin (Araldite), into a custom-built cell 'puck' constructed from either PEEK or Delrin. A graphite ring counter electrode, depth approximately 5 mm, was cut from a graphite tube (Goodfellow, 1.6 mm wall thickness) and was sealed into a groove in the cell puck using epoxy resin. Connection to the counter electrode was achieved using a Pt wire (Surepure Chemetals, 99.99 %, 26 gauge) fixed to the rear of the counter electrode (before sealing into the cell) using silver-loaded epoxy resin. The cell surface, glassy carbon and graphite ring electrodes were polished to a mirror finish using successive grades of silicon carbide paper (to 4000 grit, Kemet). The polished cell was then washed by ultrasonication in ultra-high purity water. A miniature Ag/AgCl reference electrode (3 M KCl, eDAQ, 2 mm diameter) was positioned close to the electrodes to minimize uncompensated resistance. The reference electrode was removed prior to polishing, and sealed into the cell during measurements using silicone sealant (Dowsil, SE 9187L Silicone RTV). Figure S1. Schematic representation of the IR microspectroscopic-electrochemical cell used in this work. The arrangement of the working, reference and counter electrodes (WE, RE, CE, respectively) are shown. Not to scale.

IR microspectroscopic data collection
Single crystal spectra were recorded using 15 × 15 μm 2 knife-edge apertures in the detection beampath to define the volume of crystal illuminated during a measurement. Spectra were recorded relative to a background spectrum of the crystal stabilisation buffer and redox mediator mixture, recorded in situ prior to each electrochemical redox titration using a reflection from the bare electrode surface approximately 30-45 μm from the crystal under investigation. We found that this provides a reasonable method to minimise interference artefacts due to roughness of the electrode surface and high brightness of the synchrotron IR source. We have previously shown that crystals of Hyd1 are stable over several days in the electrochemical cell, with no detectable dissolution of protein. All raw data presented in this work therefore represent absolute spectra of the crystal under investigation, and baseline-corrected data are free from significant artefacts in the background spectrum.

Solution infrared spectroscopic-electrochemical measurements
Solution spectroscopic-electrochemical measurements were carried out in a similar manner to those previously reprted. 5,6 A phosphate-buffered Nafion solution was prepared by titrating a 1:1 v/v mixture of Nafion 117 (Sigma Aldrich, 10% aqueous dispersion) and K 2 HPO 4 (BDH, 100 mM) to pH 6.0 using small quantities of concentrated NaOH (BDH). A 2.5 µL aliquot of phosphate-buffered Nafion was then mixed with 2.5 µL of carbon black particle dispersion (XC72R, DuPont, 20 mg mL -1 ) and 20 µL Hyd1 solution (ca 10 mg mL -1 ), and allowed to concentrate by evaporation to a volume of ca 15 µL. The Nafion-carbon-Hyd1 mixture was deposited onto the surface of a silicon prism (Crystal GmBH) mounted on an ATR-IR accessory (GladiatIR, Pike Technologies) in an anaerobic, dry N 2 -filled glovebox (Glove Box Technology, <1 ppm O 2 , <−75 °C dew point). The mixture was allowed to concentrate under N 2 to a final volume of ca 3-5 μL, at which point the carbon black particle dispersion forms a 3D particle network electrode encasing concentrated Hyd1 in the aqueous Nafion. Lateral contact was achieved using carbon paper (AvCarb P50, Ballard Power Systems), and electrical contact to the composite working electrode was made using a carbon rod (WH Smith, 0.9 mm HB) sealed into the body of an electrochemical cell. 6,7 The electrochemical cell contained additional feedthroughs for a fritted counter electrode (Pt wire) and a homemade miniature saturated calomel reference electrode (SCE) as previously described. 7 The ATR-IR solution spectroscopic-electrochemical cell was filled with N 2saturated, phosphate-buffered electrolyte solution (50 mM, pH 6, 100 mM KCl). A closed loop of N 2purged electrolyte was pumped through the cell during data collection to prevent build-up of any trace H 2 produced by Hyd1. IR spectra were recorded at a range of applied potentials, allowing a minimum of 15 minutes equilibration time after each potential step. Equilibration was confirmed by the absence of changes to the ν CO and ν CN bands of the active site over five successive spectra. Spectra were recorded using an Agilent 680-IR spectrometer controlled by ResPro 4 software, as an average of 1024 interferograms (ca 360 s measurement time). Electrochemical control was provided by an Autolab 128N potentiostat (Metrohm), and potentials (E) are reported relative to the standard hydrogen electrode (SHE) using the conversion E(mV vs SHE) = E(mV vs SCE) + 241 mV at 25 °C. 8 Baseline correction, and all subsequent data analysis was carried out using OriginPro software (OriginLab Corp., version 9.1).

Infrared spectroscopic-electrochemical measurements of electrode-adsorbed Hyd1
The data collected from electrode-adsorbed Hyd1 reported in this manuscript were reproduced from data previously reported by us in Hidalgo et al.. 7 Simultaneous IR spectroscopic-electrochemical measurements were carried out using the ATR-IR cell and protein film IR electrochemistry (PFIRE) method. Briefly, a 50 µL aliquot of as-isolated Hyd1 (6.3 mg mL -1 ) was exchanged into potassium phosphate buffer (15 mM, pH 5.8) by dilution and re-concentration using a 50 kDa molecular weight cut-off microcentrifugal concentrator (Amicon Ultra 0.5 mL, Merck). The enzyme sample was then mixed with 5 µL of a carbon black particle suspension (Black Pearls 2000, Cabot Corporation, dispersed in water by sonication to a loading of 20 mg mL -1 ) and left at 0 °C for 2 h to allow for enzyme adsorption. The particles modified with Hyd1 were then washed with buffer to remove un-adsorbed enzyme before re-concentration to a carbon black loading of 20 mg mL -1 . Enzyme adsorption was carried out in a N 2 -filled glovebox (< 1 ppm O 2 , Glove Box Technology Ltd.).
An aliquot (ca 1-2 μL) of Hyd1-modified particles was deposited onto the surface of a Si internal reflection element (Crystal GmBH) mounted onto an ATR-IR accessory (GladiatIR, PIKE technologies). Lateral electrical connection was provided by carbon paper (Toray, TGP-H-030) placed on top of the deposited particle film. A spectroscopic-electrochemical cell body housing a miniature saturated calomel reference electrode, Pt wire counter electrode, carbon rod working electrode connection, and an inlet and outlet to allow solution flow, was sealed onto the ATR-IR accessory and filled with Arsaturated buffer. Ar-saturated buffer was flowed through the cell throughout measurements to prevent build-up of any trace H 2 produced. A mixed buffer solution (pH 6.0, MES, HEPES, TAPS, CHES, NaOAc, 15 mM each) containing NaCl (100 mM) as supporting electrolyte was used as the flow solution. The aerobically-purified Hyd1 was reductively activated under 1 bar H 2 at −594 mV for at least 1 hour before collecting redox titration data.
IR spectroscopic data were collected using an Agilent 680-IR spectrometer controlled using ResPro 4.0 software. Baseline correction and all subsequent data handling was carried out using OriginPro software (OriginLab Corp., version 9.1).

Calculation for number of active sites illuminated with IR microspectroscopy per unit of crystal volume
Approximate Unit cell volume = 93x97×183 Å 3 = 1.642×10 6 Å 3

Unit cells / µm 3 = 609,013
In Spacegroup P2 1 2 1 2 1 the unit cell contains 4 asymmetric units Thus Hyd1 molecules / µm 3 = 2,436,000 Protein is a dimer, therefore 4.87×10 6 active sites / µm 3 Typically, a sampling area of 15×15 µm 2 is used in the IR microspectroscopy experiments: Active sites sampled during experiment = 1.1×10 9 for each µm of crystal depth. Scheme S1. The catalytic cycle of Hyd1, including oxidised inactive state Ni-B. States are coloured as in Scheme 1 and denoted in spectra throughout. Viewed in the direction of H 2 oxidation, H 2 is thought to bind at the level of Ni a -SI to form a transient Michaelis complex. The H 2 molecule is polarised between an active site metal (Lewis acid) and a nearby base, the identity of which is still under debate. Candidates include a terminal cysteine thiol (C576 in Hyd1 numbering) in the primary coordination sphere of the active site Ni, or the pendant arginine of R509 located ca 4 Å in a "canopy" above the bridging position between Ni and Fe. [9][10][11][12][13] Abstraction of the proton from H 2 leaves a hydride in a bridging position between the two metals, forming the most reduced state of the active site generally termed Ni a -R. Next, the Ni is oxidised from Ni II to Ni III , the electron being transferred to the proximal FeS cluster, and the proton abstracted from H 2 is transported away from the active site forming Ni a -C. From here a proton and an electron must be lost from the active site to re-form Ni a -SI. In the case of Hyd1 the oxidation of Ni a -C to Ni a -SI has been shown to occur via Ni a -L: the bridging hydride moves as a proton, via C576 to E28, 14,15 leaving its two electrons on the Ni, forming Ni I . Oxidation of Ni I to Ni II occurs via transfer of the second electron to the proximal cluster. However, this electron transfer step is likely retarded by the high potential of the FeS clusters of Hyd1, 16 ultimately leading to the ability to be able to identify Ni a -L as a true intermediate in the catalytic cycle. 7,17,18 The Ni-B inactive state can be formed by the oxidation of the active site, either by application of high potentials, or by the binding of O 2 (which requires 3 protons and 3 electrons to form Ni-B, 1 electron reduction of which re-forms Ni a -SI). Table S1. Mid-point potentials (E m ) and peak-to-peak separations (ΔEp) of redox mediators used, measured in the infrared microspectroscopic-electrochemical cell described in Figure S1 at Figure S2. Cyclic voltammogram demonstrating electrochemical control within the infrared microspectroscopicelectrochemical cell. Recorded at a scan rate of 10 mV s -1 , with a background electrolyte of hydrogenase crystal stabilisation buffer (pH 5.9) additionally containing the redox mediators methyl viologen, anthraquinone-2-sulfonate, indigo carmine, phenazine methosulfate, and 2,6-dichloroindophenol (each at a concentration of 1 mM ). See Table S1 for additional analysis.      Figure S8. Chronoamperometry traces recorded during Hyd1 pH 5.9 microspectroscopic-electrochemical oxidative titration shown in Figure 2. A potential of −597 mV was applied for 3720 s prior to IR spectra being recorded. For the purposes of constructing the redox titration in Figure 3A of the main text, equilibrium was judged spectroscopically. In general 'spectroscopic equilibrium' coincided with net zero current at most applied potentials. However at potentials more positive than +103 mV, or more negative than −347 mV, a small residual current (<|8 nA|) was observed. This small residual current is less than 0.5% of the peak current observed during voltammograms of the mediator cocktail in the microspectroscopic-electrochemical cell ( Figure S2), and small residual currents are expected in thin-layer spectroscopic-electrochemical cells. Critically, however, we find that crystal position on the working electrode surface (i.e. whether it is near the centre or near the edge) has no discernible effect on the recorded redox titrations.

+103 mV
Ni a -L II Ni a -L III Figure S9. Representative potential-dependent spectra at pH 5.9. Infrared spectra of a single crystal of Hyd1, pH 5.9, poised at selected potentials as noted. The potentials selected reflect those at which maximum intensity for each active site redox species is observed during the redox titration shown in Figure 3A. Fitted peaks are shown in grey. Table S2. Comparing active site species at pH 5.9. The relative v CO band positions of each active site species, to the nearest cm -1 , at pH 5.9 observed for Hyd1 in crystallo, in solution, or adsorbed on an electrode (PFIRE). We have reassigned the Ni a -L absorbances relative to the original literature, as discussed in the main text, such that both the Ni a -L I, II, III and Ni a -R I, II, III sub-states are labelled in order of decreasing wavenumber of the active site CO stretch, ν CO .   Ni a -L III Ni a -L II Figure S12. Representative potential-dependent spectra at pH 8.0. Infrared spectra of a single crystal of Hyd1, pH 8.0, poised at selected potentials as noted. The potentials selected reflect those at which maximum intensity for each active site redox species is observed during the redox titration shown in Figure 4. Fitted peaks are shown in grey.   Band positions, reported the nearest 0.1 cm -1 , are derived from Gaussian profiles fitted to baselinecorrected IR spectra with the aid of 2 nd derivative spectra.

Position of ν CO band centre / cm -1 Method
We have reassigned the Nia-L absorbances relative to the original literature, as discussed in the main text, such that both the Ni a -L I, II, III and Ni a -R I, II, III sub-states are labelled in order of decreasing wavenumber of the active site CO stretch, νCO. pH Figure S14. The relative populations of Ni a -C, Ni a -L II , and Ni a -L III are pH dependent. Peak absorbance data reproduced from PFIRE spectra, recorded as a function of pH at potentials at which the total [Ni a -C + Ni a -L] is optimised. The pH-dependence observed in PFIRE data is in good agreement with the pH dependence observed in single crystals shown in Figure 5 of the main text. The Ni a -L sub-states observed in E. coli Hyd1 do not share the same pH dependence, implying multiple proton acceptor sites.
We have reassigned the Ni a -L absorbances relative to the original literature, as discussed in the main text, such that both the Ni a -L I, II, III and Ni a -R I, II, III sub-states are labelled in order of decreasing wavenumber of the active site CO stretch, νCO.  Figure S15. Single crystal difference spectra show no evidence of cysteine-S protonation in either Ni a -R II/III or Ni a -L II/III . Difference spectra were calculated from raw single crystal microspectroscopy data recorded at pH 5.9 and reported in Figure S7: a shows the Ni a -C/Ni a -L redox level minus the Ni a -R redox level; b shows the Ni a -SI redox level minus the Ni a -C/Ni a -L redox level; c shows the Ni a -SI redox level minus the Ni a -R redox level. The approximate S-H stretching region (νSH) is shaded in grey. There is no evidence for a change in cysteine-S protonation at any redox level, which implies either the existence of S-H in all redox states or the absence of S-H in all redox states. Note that in Hyd1 crystals we do not observe the Ni a -L I state, for which Tai 15 Tai et al. additionally noted that cysteine-S protonation was not observed in either the Ni a -SI or Ni a -C states of DvMF. Our data therefore suggests that there are no changes in cysteine-S protonation between any of the Ni a -SI, Ni a -C, Ni a -L II/III , or Ni a -R II/III states in Hyd1. In combination with evidence from collected IR data which suggests that a terminal cysteine-S ligand becomes deprotonated during the transition from Ni a -L I to Ni a -L II/III , and from Ni a -R I Ni a -R II/III , 26 these data provide further evidence that terminal cysteine-S is not protonated in the Ni a -L II/III or Ni a -R II/III states.
Ni a -SI Ni a -L Figure S16. The transition of Ni a -L to Ni a -SI in single crystals of Hyd1 at pH 5.9 during an oxidative potential step from −197 mV to −172 mV. Raw overlaid IR spectra are shown, with baseline corrected spectra in Figure S17. 27 Ni a -L II Ni a -L III Figure S17. The transition of Ni a -L to Ni a -SI in a single crystal of Hyd1 at pH 5.9 during an oxidative potential step from −197 mV to −172 mV. Baseline corrected IR spectra are shown, for raw IR spectra see Figure S16. Figure S18. Time dependence of the active site redox species in a single crystal of Hyd1 at pH 5.9, following an oxidative potential step from −197 mV to −172 mV. These are the same data as reported in Figure  6B, but here also include the Ni a -R and Ni a -C species, whose intensities remain approximately constant after collection of the first spectrum following the potential step (ca 32 s). The Ni a -C data was subject to 10 % fitting error; the Ni a -R data were subject to 10 % fitting error. Qualitatively similar results were obtained from multiple crystals. Figure S19. X-ray diffraction data for mediator soaked Hyd1 crystals. X-ray diffraction image collected from Hyd1 crystal after extensive (6hr) soaking in redox mediator cocktail, demonstrating crystals still diffract X-rays to high resolution after prolonged exposure to redox mediators. Values are parenthesis refer to data in the highest resolution shell.