Exploring Structure and Function of Redox Intermediates in [NiFe]‐Hydrogenases by an Advanced Experimental Approach for Solvated, Lyophilized and Crystallized Metalloenzymes

Abstract To study metalloenzymes in detail, we developed a new experimental setup allowing the controlled preparation of catalytic intermediates for characterization by various spectroscopic techniques. The in situ monitoring of redox transitions by infrared spectroscopy in enzyme lyophilizate, crystals, and solution during gas exchange in a wide temperature range can be accomplished as well. Two O2‐tolerant [NiFe]‐hydrogenases were investigated as model systems. First, we utilized our platform to prepare highly concentrated hydrogenase lyophilizate in a paramagnetic state harboring a bridging hydride. This procedure proved beneficial for 57Fe nuclear resonance vibrational spectroscopy and revealed, in combination with density functional theory calculations, the vibrational fingerprint of this catalytic intermediate. The same in situ IR setup, combined with resonance Raman spectroscopy, provided detailed insights into the redox chemistry of enzyme crystals, underlining the general necessity to complement X‐ray crystallographic data with spectroscopic analyses.


Introduction
Tr ansition metals are often involved in chemical and enzymatic catalysis.I nn ature,m etal-containing enzymes catalyze av ariety of reactions,e specially the conversion of small gaseous molecules like CO 2 ,N 2 ,o rH 2 .T his type of chemistry is relevant for establishing alternative strategies for energy conversion and the production of carbon-neutral fuels. Many of these metalloenzymes,i ncluding carbon monoxide dehydrogenase,n itrogenase and hydrogenase,a re attractive targets for biotechnological application and can serve as blueprints for bioinspired chemistry. [1][2][3][4][5][6] However,t heir rational utilization requires at horough mechanistic understanding,t ypically requiring multiple spectroscopic techniques.
Infrared (IR) spectroscopy in various implementations has been successfully used to monitor catalytic processes and intermediates at and beyond biological metal sites.A vailable approaches comprise surface-sensitive techniques like surface-enhanced infrared absorption (SEIRA) and/or attenuated total reflection (ATR) spectroscopy,o ptionally combined with electrochemistry,i llumination, gas-atmosphere and/or temperature control. [7][8][9][10][11][12] Recently,t ime-resolved IR studies in the nanosecond range as well as ultrafast pumpprobe and two-dimensional IR techniques have also been introduced into metalloenzyme research. [13,14] Electron paramagnetic resonance (EPR) spectroscopy provides additional structural and electronic insight into paramagnetic states, [15][16][17][18] while resonance Raman (RR) spectroscopy has been successfully applied to monitor the characteristic metalÀligand modes of specific redox states and cofactors. [19][20][21][22][23][24] Additionally, 57 Fe nuclear resonance vibrational spectroscopy (NRVS), as ynchrotron-based technique that can selectively probe iron-specific normal modes,h as been lately established for characterizing iron-containing enzymes. [25][26][27][28][29][30][31][32] Here,w eh ave designed an ew experimental setup for spectroscopic analyses of gas-converting metalloenzymes in various sample forms.Gas composition and temperature can be adjusted, thereby enabling the preparation of specific redox states for subsequent characterization by complementary spectroscopic tools,s uch as RR, EPR, or NRVS.A dditionally,r edox transitions between catalytically relevant intermediates and resting states can be monitored by in situ IR spectroscopy.
To illustrate the versatility of our approach and its benefits for exploring metalloenzyme catalysis,w ei nvestigated two O 2 -tolerant model enzymes,t he regulatory and the membrane-bound [NiFe]-hydrogenase from Ralstonia eutropha (ReRH and ReMBH, respectively). Briefly,[ NiFe]-hydrogenases catalyze the reversible cleavage of molecular hydrogen into protons and electrons ( Figure 1). All of these enzymes feature aheterodimeric core structure,consisting of al arge subunit harboring the [NiFe] active site and as mall subunit comprising iron-sulfur clusters that form an electron transfer relay.T he proposed catalytic cycle and redox intermediates of [NiFe]-hydrogenases discussed in this study are displayed in Figure S1.

Results and Discussion
Thee xperimental setup introduced here is designed for the preparation of well-defined redox states of metalloenzymes and their concomitant multi-spectroscopic analysis under various experimental conditions and in different sample forms.T he sample compartment consists of ag astight chamber connected to an external vacuum pump and av ariety of dry or humidified gases (Figures 2a nd S2). The temperature can be set between 80 Ka nd 333 K. Our newly developed approach also allows the in situ analysis of protein solutions (Figures S3 and S11), lyophilizates ( Figure 3C)and single crystals (Figures 5a nd S10) by IR transmission spectroscopy.S ince the whole sample compartment is portable and gas-tight, it can also be disconnected from the remaining setup and moved ( Figure S2), for example,for RR spectroscopic analysis of the same sample or further sample treatment in an anaerobic chamber.
To illustrate the capabilities of our platform, we first describe the preparation and spectroscopic characterization of highly concentrated lyophilized samples of ReRH, suitable for conducting NRVS experiments.Upon incubation with H 2 , ReRH resides predominantly in the so-called Ni a -C catalytic intermediate.This species harbors abridging hydride between the Fe II and Ni III ions ( Figure S1), which is of general interest in biological and chemical catalysis. [33][34][35] So far, experimental evidence for the presence of the hydride in this state has only been provided by EPR spectroscopy. [35][36][37] Currently,the only vibrational spectroscopic method suitable to monitor metalÀ ligand bonding in this light-sensitive species ( Figure S4) is NRVS. [19,20] Observation of active-site metalÀligand vibrations with this technique,h owever,t ypically requires protein concentrations above 1mm, [27,28] which is the upper concentration limit for purified ReRH in solution. To overcome this limitation, we used our setup for ag entle lyophilization to prepare highly concentrated enzyme samples while preserving their catalytic activity.NRVSwas performed subsequently in combination with density functional theory (DFT) calculations to reveal structural details of the Ni a -C state.
In order to obtain highly concentrated samples in the Ni a -Cr edox state,0 .5 mm protein solution of ReRH was flashfrozen at 77 Ka nd transferred to the sample compartment (Figures 2A and S2). Subsequently,w ater was removed at controlled temperature and pressure by applying am ild vacuum of 0.1 mbar while slowly warming up the sample to 243 K. After completion of the lyophilization, the sample compartment was flushed with dry H 2 gas to reduce the enzyme.F inally,c ompression of the protein lyophilizate, using as patula and ap estle,y ielded highly dense samples (equivalent to 4-5 mm,s ee Supporting Information for details) suitable for demanding methods such as NRVS (Fig-ure 2C). Forf urther details,s ee the Materials and Methods section in the Supporting Information and Figure S2.
To verify structural integrity and functionality of the lyophilized oxidized and H 2 -reduced enzyme,w em easured the H 2 -oxidation activity and applied EPR, IR and RR spectroscopy ( Figure 3). This multi-spectroscopic approach was only applicable due to the design of the setup.T he IR signature of the amide Ia nd amide II vibrational modes is related to the absorption of the (polyamide) backbone of the protein and therefore reflects its secondary and tertiary structure. [38] Thes imilarity between the spectra of the lyophilized and solution-phase samples indicates that the freeze-drying process did not affect the integrity of the protein structure ( Figure 3A). Additionally,l yophilized and subsequently re-dissolved ReRH showed the same hydrogenase activity as freshly isolated enzyme,indicating that the metal cofactors and amino acid side chains responsible for substrate conversion and proton/electron transfer were not altered by lyophilization ( Figure 3A inset). Thei ntegrity of the iron-sulfur cluster relay was verified by almost identical FeÀSs ignals in the corresponding RR spectra ( Figure 3B). Likewise,metal À ligand vibrations,characteristic for the most oxidized, hydrogen-binding [NiFe] intermediate,termed Ni a -S( Figure S1), were observed, thereby confirming an ative active-site structure of oxidized ReRH lyophilizate (Figure 3B). [20] Intriguingly,t he lyophilized enzyme still reacts with molecular hydrogen, as monitored by in situ IR spectroscopy ( Figure 3C). Thebiologically unusual CO and CN À ligands at the active site of [NiFe]-hydrogenases give rise to valuable IR marker bands that allow the identification of individual redox states of the [NiFe] center. [39][40][41] TheH 2 -dependent formation of the Ni a -C state indicates ar edox-active ReRH in the compressed and H 2 -reduced lyophilizate.C omplementary EPR spectra display the typical rhombic signature of the Ni a -Cs tate, [35] confirming an ative active-site structure after lyophilization, reduction and compression ( Figure 3D).
Thel yophilization procedure described above yielded highly concentrated, 57 Fe-enriched ReRH samples,w hich ) ReRH compared to freshly prepared enzyme solution (sol.). A) IR spectroscopic signature reflecting the amide Iand II vibrational modes [38] of the oxidized enzyme. The inset shows the specific H 2 -oxidation activity of freshly prepared, freeze-thawed and re-dissolved lyophilized samples. B) RR spectra of the oxidized enzyme obtained with 458 nm excitationa nd normalizedt othe most intense active site signal at 550-552 cm À1 .The small shifts of 1-2 cm À1 likely reflect differences between protein purified from cells grown with either 57 Fe (lyo.) or iron with natural isotope distribution (sol.). The sharp signal marked with an asterisk derives from an optical artefact of the Raman spectrometer.C )IRs pectra of ReRH, oxidized (top), H 2 -reduced (middle) and subsequently reoxidized with air (bottom).F or details about the assigned active-siter edox states, see Figure S1 and Table S1. D) EPR spectra of the H 2 -reduced enzyme in solution and as compressed lyophilizate (both measured at 80 Ka nd with 1mWmicrowave power). IR and EPR spectra of the protein solution are normalized to the intensity of the lyophilized sample.
allowed NRVS-based detection of active-site vibrations of the Ni a -C state.T he NRVS data shown in Figure 4c omprise the spectral regions characteristic for Fe À CO/CN modes (400-650 cm À1 )a nd Ni À H À Fe wagging vibrations (650-800 cm À1 ). [25][26][27][28] Thecount rate of the elastic peak accumulated by the detector increased by af actor of four for the lyophilized sample in comparison to the solution-phase sample ( Figure 4A,i nset), providing as ignificant improvement of the signal-to-noise ratio.Consequently,the error bars of the two sample spectra vary,onaverage,byafactor of four ( Figure 4A,s ee Supporting Information for details).
In order to specifically probe the metal-bound hydride of the Ni a -C state,lyophilized ReRH samples were treated with both H 2 and D 2 ( Figure 4B). Ther esulting experimental spectra were compared to simulated data obtained by DFT calculations (Figure 4B), based on an ReRH active-site homology model [42,43] (Figures S5 and S6;f or details,s ee the DFT Methods section of the Supporting Information). Analysis of the Ni a -C spectrum obtained after incubation of ReRH with H 2 ( Figure 4B,t op blue trace) revealed two intense bands at 554 and 598 cm À1 and aw eaker one at 574 cm À1 .Additional spectral features appear at 446, 470, 500 and 508 cm À1 .P reviously,[ NiFe]-hydrogenase vibrational signals at these energies were attributed to normal modes dominated by Fe À CO and Fe À CN stretching and bending motions. [19][20][21][25][26][27][28] Overall, the agreement between the experimental and calculated spectra is good ( Figure S7). The isolated NiÀHÀFe wagging motions are predicted to produce only very weak 57 Fe-PVDOS intensities in the 670-740 cm À1 region. Them ost noticeable spectral feature of this type is expected at 694 cm À1 and formed by two modes at 692 and 696 cm À1 ,e ach characterized by approximately 30 % mH-PVDOS but only subtle 1% 57 Fe-PVDOS.Adiscrete hydride band could not be resolved by NRVS due to the low signal-tonoise ratio in this spectral region, far away from the elastic Mçssbauer peak. However,adistinct H/D-sensitive band observed at 574 cm À1 is predicted to arise from two modes calculated at 574 and 576 cm À1 (7 % 57 Fe-PVDOS each), both of which are combinations of mHw agging and C À NH 2 bending of the nearby Arg411 guanidium group ( Figures S7A  and S8). TheNRVSdata recorded from ReRH incubated with D 2 (Figure 4B,top red trace) reveal small shifts of 1-5 cm À1 to lower frequencies for several bands.This can be explained by deuteride motions contributing to the Fe À CO(CN)-dominated modes, [20,44,45] with a mD-PVDOS maximum (23 %) calculated at 501 cm À1 for Ni a -C(mD) (Figures S7B and S8, normal mode animations provided in Supporting Information). Most prominently,t he 574 cm À1 band detected for the H 2 -reduced sample is absent in spectra from both the experimental D 2treated sample and the calculated Ni a -C(mD) model. Moreover, anewly emerging signal (calculated Ni a -C(mD) mode at 581 cm À1 )can be observed as aweak shoulder at ca. 586 cm À1 , close to the nearby high-intensity feature centered at 596 cm À1 ,i ndicating the presence of the mDd euteride.I n summary,a pplication of the novel setup proved to be beneficial for the enrichment of ReRH, thereby enabling NRVS characterization of the reduced enzyme.Rationalized by DFT calculations,t hese data allowed insights into activesite metalÀligand vibrations including normal modes that reflect hydride coordination in the Ni a -C redox state.
Using ReMBH as am odel metalloenzyme,w en ext demonstrate the applicability of our setup for the characterization of protein crystals by multiple spectroscopic techniques ( Figures 2B and S10). To obtain ac omprehensive understanding of redox processes in [NiFe]-hydrogenase crystals,w efirst compared in situ IR spectra of aerobically grown protein crystals and solution ( Figure 5A,B). Unlike Ash et al.,w ho employed electrochemical control in combination with redox mediators to induce redox transitions within crystals of Hydrogenase 1f rom Escherichia coli, [46] our design allows reduction of the hydrogenase with its native substrate H 2 and (re)oxidation with the inhibitor O 2 .T hus, our setup allows monitoring physiologically relevant redox processes at the [NiFe] active site,a se xemplified for the Figure 4. Nuclear resonancevibrational spectra of ReRH in the Ni a -C state. A) NRVS data of lyophilized,H 2 -reduced and compressed ReRH (lyo.,black trace) compared to the corresponding protein solution at 0.5 mm (sol.,gray trace). Both spectra were recorded for 12 hours. The inset displayst he corresponding count rates at the elastic peak. B) Experimental NRVS data of lyophilized ReRH (top) incubated with H 2 (blue traces, 26 hours accumulation)orD 2 (red traces, 16 hours accumulation) in comparison to the corresponding DFT-calculated 57 Fe-PVDOS spectra, which were obtained using the model shown in Figure S6. NRVS data includingerror bars are displayed in Figure S9.

Angewandte Chemie
Research Articles reductive activation process of oxidized ReMBH in the following.
When exposed to oxidizing conditions,for example,upon incubation with air ( Figure 5A,B,g reen traces), O 2 -tolerant hydrogenases like ReMBH typically switch to anon-catalytic but easily reactivatable "resting" state,c alled Ni r -B,w hich features ab ridging hydroxide between Fe and Ni (Figure-s5A,B and S1). [47] In principle,N i r -B may be directly activated by reaction of the [NiFe] site with H 2 ,assuggested by Kurkin et al. [48] Alternatively,a ctivation could be accomplished indirectly by reverse,F e ÀSc luster-mediated transfer of H 2 -derived electrons delivered by catalytically active hydrogenase molecules. [49] Within crystals,h owever, each ReMBH heterodimer is locked at afixed position and unable to rotate,t hereby severely hampering electron transfer between different hydrogenase molecules.T his would affect indirect but not direct reactivation. Thus,acomparison of solute and crystalline ReMBH samples provides au nique possibility to analyze H 2 -mediated reactivation behavior of [NiFe]-hydrogenase residing in the Ni r -B state.
In fact, complete active-site reduction in an ReMBH crystal required 5h incubation with 100 %H 2 (Figures 5A, red trace and S11A), whereas complete reduction in solution was accomplished within 30 min (Figures 5B,r ed trace and S11B). Strikingly,O 2 (although larger than H 2 )r e-oxidized previously reduced ReMBH crystals ( Figure 5A,b lue trace) to full extent within less than 5min ( Figure S11A), demonstrating that the slow,H 2 -mediated reactivation of the Ni r -B state in crystalline samples is not caused by rate-limiting gas diffusion. We therefore propose that the fast activation of Ni r -B-as observed in solution-occurs via the indirect route, involving electron transfer from other hydrogenase molecules.S low activation-as observed in the crystal phasecould be ascribed to less efficient direct activation by H 2 or long-range intermolecular electron transfer [50] between metal centers of (translationally and rotationally locked) ReMBH Figure 5. IR and RR spectra of ReMBH in the crystal and solution phase. In situ IR spectra of the oxidized (green), reduced (red) and reoxidized (blue) state of A) an enzyme crystal and B) protein solution after consecutive exposure to various humidified gases (N 2 ,H 2 ,s ynthetic air) at 277 K. The insets show the reduced-minus-oxidized differencespectra after different times of H 2 exposure. Fordetails on the spectroscopic transformations, see Figure S11. Details regardingt he redox states assigned to the active site are described in Figure S1 and Table S2. C) IR spectrum (top trace) of ap rotein crystal grown under reducing atmosphere( 95 %N 2 and 5% H 2 )a nd the corresponding RR spectrum obtained with 568 nm laser excitation (bottom trace) of the same crystal, both measured at 80 K. D) RR spectra displaying the spectral region characteristic for iron-sulfur cluster vibrationalm odes of oxidized (blue trace) and partially reduced (red trace) ReMBH crystals (the latter was grown under an atmosphere of 5% H 2 and 95 %N 2 ). Spectra were recorded at 80 Kwith 514 nm laser excitation. RR spectra are normalizedtoanon-resonant band of the amino acid side chain of phenylalanineat1005 cm À1 .E )IRspectrum of ap rotein crystal grown under reducing atmosphere( 95 %N 2 and 5% H 2 )before (dark, cyan) and after (light, orange) 8hours illumination with the focused beam of acollimated 455 nm LED at 80 K. The corresponding light-minus-dark difference spectrum is shown in black. heterodimers ( Figure S12). These data demonstrate how our setup allows studying redox processes of gas-converting enzymes in the crystal phase.
Theredox state and behavior of ametalloenzyme may not only depend on the sample phase (crystal vs.solute), but also on the temperature and the sequence of steps during sample preparation that lead to the final state (which may not reflect thermodynamic equilibrium). These aspects are typically inaccessible by crystallography,t hus requiring additional insight from in situ spectroscopy (vide supra) performed under identical conditions and on the same set of crystals. [22,51] As most crystallographic data are acquired at cryogenic temperatures,i ti so fu tmost importance to perform the corresponding spectroscopic experiments both at cryogenic and ambient temperatures.T his way,i nformation from cryogenic crystal structures can be adequately related to ambient-temperature and/or solution-phase data, thereby revealing the functionality of the native protein. Thel owtemperature configuration for IR and RR spectroscopy of our new setup allows to conduct such experiments.
As an example,w einvestigated ReMBH crystals grown under moderately reducing conditions of 5%H 2 and 95 %N 2 . In case of hydrogenases,H 2 specifically acts as ar educing agent, thereby defining the redox status of the sample.A s revealed by IR spectra recorded from several crystals at 80 K (Figures 5C,top trace and S13A), the [NiFe] active site forms one dominant redox state,w hose IR signature has not been reported for ReMBH so far.T oobtain further insight into this previously unknown active-site state,w ev aried both the temperature (up to 283 K, Figure S13B) and the gas atmosphere (100 %H 2 or 20 %O 2 /80% N 2 , Figure S13C) over the crystal in situ. Surprisingly,n either had an otable impact on the IR pattern of the ReMBH crystals.T his is in sharp contrast to crystals grown under aerobic conditions,which are easily manipulatable by adding either of these gases (vide supra), indicating that the new redox state is inactive. Moreover,c rystals grown under 100 %H 2 revealed similar spectroscopic signatures as those grown under 5% H 2 and 95 %N 2 ( Figure S13B), indicating that, for so-far unknown reasons,the continuous presence of H 2 during crystal growth traps ReMBH in this state.T aking advantage of the transferable sample holder,wenext recorded complementary RR spectra of the same crystal grown under 5% H 2 /9 5% N 2 ( Figure 5C,b ottom trace). Surprisingly,t he corresponding spectra reveal the characteristic signature of the Ni a -S state [21] and partial reduction of the FeÀSc lusters [52] (Figure 5D). While the latter observation suggests successful reduction of the protein by H 2 ,t he former either indicates that the IRdetected redox state is indistinguishable from Ni a -S or that Ni a -S is formed in situ as aphotoproduct, due to illumination by the intense Raman probe laser. To test this hypothesis,we illuminated the crystal for 8hours at 80 Kw ith the focused beam of ac ollimated 455 nm LED ( Figure 5E). Tw on ew species emerged in the concomitantly recorded IR spectra, indicating that the IR-detected "dark" state is light-sensitive and not identical with the Raman-probed Ni a -S state.T his finding offers the opportunity to study the activation mechanism of inhibited redox states in detail. [53,54] On the one hand, these observations demonstrate that the redox status of crystallized metalloenzymes may not necessarily represent the expected equilibrium state,w hich is highly relevant for the interpretation of crystallographic data. On the other hand, the identification of anew redox state in the crystal phase offers an interesting avenue for combined Xray diffraction and vibrational spectroscopic studies,a s previously performed with the same setup on the F 420reducing hydrogenase from Methanosarcina barkeri [22] and the [FeFe]-hydrogenase from Desulfovibrio desulfuricans. [51] In total, the results presented here demonstrate the importance of in situ spectroscopic studies on protein crystals and the wide applicability of our new experimental platform.

Conclusion
We have developed an experimental setup for multispectroscopic in situ studies on gas-converting metalloenzymes.U sing this approach, enzymes can be studied in the form of single crystals,i ns olution or as lyophilizate-at various temperatures and gas atmospheres.
Freeze-drying and mechanical compression yields protein samples with concentrations that are unattainable for many sensitive metalloenzymes in solution. This strategy can be utilized for studying such targets with low-sensitivity spectroscopic techniques that provide otherwise inaccessible information on the enzyme structure and mechanism. The feasibility of this approach has been demonstrated by probing the hydride-containing Ni III intermediate of [NiFe]-hydrogenase (Ni a -C) by nuclear resonance vibrational spectroscopy.
Another valuable application is the in situ spectroscopic analysis of metalloenzyme samples in the crystalline and solution phase under variable experimental conditions.T his allows the preparation, identification and functional analysis of catalytic intermediates in protein crystals,thereby enabling amore reliable interpretation of crystallographic data. This is particularly important for enzymes that are sensitive towards crystallization conditions and structural species that cannot be readily identified by X-ray diffraction, as demonstrated for [NiFe]-hydrogenases.M oreover,s uch studies allow evaluating the impact of the sample form on the reactivity towards different gaseous substrates.I nt his respect, we have compared activation and inhibition kinetics of crystallized and solvated [NiFe]-hydrogenase,i ndicating that fast H 2 -mediated activation in solution is accomplished via an indirect electron transfer mechanism.
In summary,the obtained results highlight the potential of our multi-spectroscopic approach for studying metalloenzymes or other metal-containing chemical systems,e .g. heterogeneous catalysts,u nder tight experimental control, thereby allowing new insights into structure and function.