Stepwise isotope editing of [FeFe]-hydrogenases exposes cofactor dynamics

Significance

[FeFe]-hydrogenases are H₂-forming enzymes with potential in renewable energy applications. Their molecular mechanism of catalysis needs to be understood. A protocol for specific ¹³CO isotope editing of all carbon monoxide ligands at the six-iron cofactor (H-cluster) was established. Analysis of vibrational modes via quantum chemical calculations implies structural dynamics at the H-cluster in the active-ready state. Site-selective introduction of isotopic reporter groups opens new perspectives to identify intermediates in the catalytic cycle.

Abstract

The six-iron cofactor of [FeFe]-hydrogenases (H-cluster) is the most efficient H₂-forming catalyst in nature. It comprises a diiron active site with three carbon monoxide (CO) and two cyanide (CN⁻) ligands in the active oxidized state (H_ox) and one additional CO ligand in the inhibited state (H_ox-CO). The diatomic ligands are sensitive reporter groups for structural changes of the cofactor. Their vibrational dynamics were monitored by real-time attenuated total reflection Fourier-transform infrared spectroscopy. Combination of ¹³CO gas exposure, blue or red light irradiation, and controlled hydration of three different [FeFe]-hydrogenase proteins produced 8 H_ox and 16 H_ox-CO species with all possible isotopic exchange patterns. Extensive density functional theory calculations revealed the vibrational mode couplings of the carbonyl ligands and uniquely assigned each infrared spectrum to a specific labeling pattern. For H_ox-CO, agreement between experimental and calculated infrared frequencies improved by up to one order of magnitude for an apical CN⁻ at the distal iron ion of the cofactor as opposed to an apical CO. For H_ox, two equally probable isomers with partially rotated ligands were suggested. Interconversion between these structures implies dynamic ligand reorientation at the H-cluster. Our experimental protocol for site-selective ¹³CO isotope editing combined with computational species assignment opens new perspectives for characterization of functional intermediates in the catalytic cycle.

The reduction of protons to form molecular hydrogen (H₂) is catalyzed by [FeFe]-hydrogenases (1, 2). With a turnover rate of up to 10,000 H₂ molecules per second in a thermodynamically reversible reaction (3–5), [FeFe]-hydrogenases inspired synthetic hydrogen catalysts (6–8) and renewable fuel technology applications (9, 10). The mechanism of catalysis at the active site cofactor (H-cluster) needs to be elucidated. Further information on functional intermediates is required (11–16) and expected to emerge from spectroscopic studies on H-cluster constructs carrying site-selective isotopic reporter groups (17–20).

Protein crystallography has identified the H-cluster as a six-iron complex (21–23), in which a canonical cubane cluster ([4Fe4S]_H) is linked to a unique diiron moiety ([2Fe]_H) (Fig. 1). The two iron ions of [2Fe]_H are located in proximal (p) or distal (d) position relative to [4Fe4S]_H and carry a bridging amine-dithiolate group [adt; (SCH₂)₂NH] (19). Both iron ions bind a terminal carbonyl (CO) and a cyanide (CN⁻) ligand. In crystal structures, the “active-ready,” oxidized state (H_ox) of the H-cluster shows a third carbonyl in Fe-Fe bridging position (µCO) and an apical vacancy at Fe_d (23). On exposure to CO gas, a fourth carbonyl binds at [2Fe]_H (24–26) and was modeled in apical position at Fe_d in H_ox-CO (27). Formation of H_ox-CO does not affect the formal redox state of the H-cluster, but leads to increased spin delocalization over the diiron site (28). CO binding inhibits H₂ turnover and protects the enzyme against O₂ and light-induced degradation (24, 29, 30).

Fig. 1.

Crystal structure of [FeFe]-hydrogenase from C. pasteurianum (23). The H-cluster (ball-and-stick) with its cubane ([4Fe4S]_H) and diiron subcomplexes ([2Fe]_H with an amine-dithiolate = adt bridge) is protein-bound by four cysteine residues. An apical vacant site (*) at Fe_d was modeled in structures of oxidized enzymes (21–23, 46). The shown CO/CN⁻ ligand orientation herein is annotated standard.

The vibrational modes of the CO and CN⁻ ligands at the diiron site are well accessible by infrared (IR) spectroscopy because they are separated from protein backbone and liquid water bands. Infrared spectroscopy therefore has pioneered elucidation of the molecular structure of the H-cluster and identification of several redox states (24, 31). In particular, the CO stretching frequencies are highly sensitive to structural isomerism, redox transitions, ligand binding, and isotope exchange (11, 12, 15, 18, 24, 31, 32). ¹³CO editing of the H-cluster has been achieved using ¹³C-precursors during H-cluster assembly or exposure of [FeFe]-hydrogenases to ¹³CO gas (18, 24–26, 33). These experiments have yielded either a completely labeled H-cluster, mixtures of labeled species, and mostly the inhibited state. Selective ¹³CO editing of H_ox was hampered by tight binding of exogenous CO, which impaired quantitative regeneration of active enzyme (29, 34). H_ox is believed to be the starting state in the H₂ conversion cycle of [FeFe]-hydrogenases (1). Selective ¹³CO editing of H_ox thus may provide access to key catalytic H-cluster intermediates (14). Introduction of ¹³CO groups also facilitates analysis of structure–function relationships using quantum chemical calculations. However, relatively few computational studies to calculate vibrational modes of the diatomic ligands have been carried out (35–39).

We compared three different [FeFe]-hydrogenase proteins, HYDA1, from the green alga Chlamydomonas reinhardtii, and the bacterial enzymes CPI from Clostridium pasteurianum and DDH from Desulfovibrio desulfuricans. HYDA1 represents the “minimal unit” of biological hydrogen turnover as it exclusively binds the H-cluster, whereas CPI and DDH hold accessory iron-sulfur clusters (3, 40). Purified HYDA1 and CPI were reconstituted in vitro with a synthetic diiron site analog to yield the active H-cluster (35, 41, 42), whereas DDH was isolated with a complete cofactor (43). We report the generation of H_ox-CO and H_ox isotopic species with all possible labeling patterns upon exposure of [FeFe]-hydrogenase protein films to ¹³CO gas, visible light, and different levels of humidity as monitored by real-time attenuated total reflection Fourier-transform infrared spectroscopy (ATR-FTIR). Density functional theory (DFT) assigned the carbonyl vibrational modes. This approach has established a reaction scheme with 16 options to convert selectively labeled H_ox-CO into 8 H_ox isotopic species as entry points to the catalytic cycle.

Results

[FeFe]-hydrogenase protein films deposited on an ATR cell were exposed to ¹²CO, ¹³CO, or N₂ gas with controlled humidity either in darkness or under red or blue light irradiation, exploring the differential wavelength sensitivity of the iron-carbonyl bonds (44). Real-time detection of spectral changes of the stretching vibrations of the diatomic ligands (SI Appendix, Figs. S1–S4) yielded high-quality IR spectra of the thereby derived pure H_ox and H_ox-CO states (Fig. 2). Frequencies and intensities of IR bands were determined using least-squares fitting. Density functional theory calculations generated geometry-optimized models of the whole H-cluster for H_ox-CO and H_ox (SI Appendix, Fig. S5). Calculated IR spectra were used for assignment of experimental vibrational bands to individual CO ligands, specific isotopic labeling patterns, and molecular structures.

Fig. 2.

ATR-FTIR spectra of HYDA1 [FeFe]-hydrogenase films. (A) Isotopic species with a p¹²CO ligand. (B) Isotopic species with a p¹³CO ligand. IR bands due to stretching vibrations of CO ligands at the H-cluster were normalized to unity area sums. Spectra are attributed to H_ox (orange) or H_ox-CO (black); CO bands are denoted α, β, γ, and δ. For real-time ATR-FTIR experiments, see SI Appendix, Fig. S3. Straight arrows denote gas exposures (¹²CO, green; ¹³CO, magenta; and N₂, black), wiggled arrows denote red or blue light irradiation. Numerals i–xii annotate identified spectral species (Table 1).

IR Band Assignment for Unlabeled H_ox and H_ox-CO.

Under an N₂ atmosphere HYDA1 showed the typical three CO bands of the H_ox state (Fig. 2 A, i). Carbonyl bands shifted by ∼40 cm⁻¹ to lower frequencies due to ¹³CO isotope editing (see below), whereas the CN⁻ bands shifted less than 1 cm⁻¹ (SI Appendix, Fig. S6) and hence were not decisive for H-cluster species assignment. DFT consistently attributed the CO bands to the largely uncoupled vibrations of the Fe-Fe bridging carbonyl (µCO, band α at 1802 cm⁻¹) and the terminal CO ligands at Fe_d (dCO, band β at 1940 cm⁻¹) and Fe_p (pCO, band γ at 1964 cm⁻¹) (SI Appendix, Fig. S9). As a measure for correlation of calculated and experimental CO frequencies, the RMSD (Eq. S1) was calculated (Table 1 and SI Appendix, Table S1 and Fig. S11). A mean RMSD of ∼10 cm⁻¹ was obtained for the four possible H_ox rotamers with equatorial CO/CN⁻ ligands at Fe_p and Fe_d (Fig. 3A), which indicated good agreement between experimental and calculated CO frequencies. A similar small RMSD was obtained for a H_ox rotamer with dCN⁻ rotated toward a more apical position (Fig. 4), whereas a rotated apical dCO was disfavored. In the following, H-cluster rotamer structures are discussed relative to the “standard” model (24, 25, 45) with trans orientation of equatorial CO ligands and apical vacancy at Fe_d in H_ox (Fig. 1).

Table 1.

Assignment of IR spectra to isotopic labeling patterns

Spectrum	p12CO	RMSD	p13CO	RMSD
i	12 12 12	6 (6)	13 13 13	5 (5)
iv	12 12 13	12 (12)	13 13 12	12 (10)
vii	12 13 13	12 (12)	13 12 12	11 (10)
x	12 13 12	6 (6)	13 12 13	6 (6)
ii	12121213	11 (31)	13131312	6 (25)
iii	12121313	7 (26)	13131212	7 (21)
v	12121312	7 (17)	13131213	5 (20)
vi	12131313	7 (25)	13121212	8 (20)
viii	12131312	7 (17)	13121213	4 (20)
ix	12131212	5 (23)	13121313	5 (23)
xi	12131213	11 (31)	13121312	6 (25)
xii	12121212	5 (24)	13131313	5 (23)

Experimental IR spectra i–xii are shown in Fig. 2A (p¹²CO) and Fig. 2B (p¹³CO). All ¹²CO/¹³CO (12/13) labeling patterns are given in the order δ, γ, β, and α (band δ is missing in i, iv, vii, and x). Deviation values (RMSD; Eq. S1) were derived from comparison of experimental and DFT-calculated CO stretch frequencies of rotamers with apical CN⁻ at Fe_d in H_ox-CO and with the distal CN⁻ rotated toward apical position in H_ox, compared with standard ligand arrangements (values in parentheses). Correlations of experimental and calculated CO (and CN⁻) band frequencies and intensities for all 178 studied model structures are shown in SI Appendix, Tables S1–S4 and Figs. S11 and S12. Bold numbers indicate ¹³CO isotope labeling.

Fig. 3.

Correlation of experimental and calculated CO band frequencies. (A) H_ox: standard model (blue) and model with proximal CN⁻ rotated toward apical position (red). (B) H_ox-CO: standard model (blue) and model with distal CN⁻ in apical position (red). Diagonals show ideal correlation. Calculated CO frequencies were offset-corrected (31 ± 1 cm⁻¹, H_ox; 38 ± 2 cm⁻¹, H_ox-CO) for alignment with experimental data (SI Appendix, Tables S1 and S2). (Insets) Approximate rotamer probabilities from IR data analysis (SI Appendix, Table S4).

Fig. 4.

H-cluster rotamer structures of H_ox and H_ox-CO. A transition from H_ox structure (A) to H_ox-CO structure (C) is suggested in the standard model where exogenous CO binds at Fe_d in apical position (magenta arrow). Equilibrium between H_ox rotamers A and B facilitates CO binding at Fe_d in equatorial position (green arrow) and thereby transition to the H_ox-CO rotamer with apical CN⁻ at Fe_d (D). Octahedral coordination of Fe_d in H_ox-CO renders ligand rotation unlikely and prevents a transition between rotamers C and D.

Exchange of N₂ by ¹²CO gas in the headspace above the protein film resulted in the appearance of a forth CO band (δ) at higher IR frequencies due to an additional carbonyl ligand (d₂CO) in H_ox-CO (Fig. 2 A, xii). We calculated the IR bands of the six possible CO/CN⁻ rotamers. Similar large RMSD values (∼30 cm⁻¹) were observed for the four structures with apical d₂CO (SI Appendix, Table S2). An about sixfold improved RMSD (∼5 cm⁻¹) was observed for the H_ox-CO structure with apical dCN⁻ and d₂CO in the equatorial plane (Fig. 3B). DFT assigned band α to the µCO stretch mode (1808 cm⁻¹) and band β to an anti-symmetric coupled mode with smaller contributions from equatorial d₁CO and larger contributions from apical d₂CO (1962 cm⁻¹). Band γ was assigned to a coupled mode with similar contributions from the symmetric vibrations of d₁CO and d₂CO and the antisymmetric stretch mode of pCO (1968 cm⁻¹) and band δ to a coupled symmetric mode with contributions from all four carbonyls (2012 cm⁻¹) in the standard model (SI Appendix, Fig. S9). Except for the energetically separated band α due to the µCO ligand (SI Appendix, Fig. S7), pronounced vibrational coupling of d₁CO, d₂CO, and pCO precludes a priori assignment of IR bands to specific CO ligands in H_ox-CO.

Stepwise ¹³CO Editing of the H-Cluster.

For HYDA1 protein films, exposure of unlabeled H_ox-CO (xii) to ¹³CO gas caused a >20 cm⁻¹ shift to lower frequencies of bands β and δ, whereas band γ was less affected and α remained unchanged, suggesting a single ¹³CO ligand at Fe_d (Fig. 2A, ii). Red light irradiation under ¹³CO gas resulted in a further >20 cm⁻¹ down-shift of bands β and δ, indicative of a second ¹³CO ligand at Fe_d (iii). In the dark, species iii was converted under ¹²CO gas to a state differing from unlabeled H_ox-CO in band β, suggesting a d₁¹³CO exchange (v). Blue light irradiation of iii under ¹³CO caused an exclusive ∼40 cm⁻¹ down-shift of band α, whereas β, γ, and δ remained unchanged. Thus, a state with three shifted CO bands with respect to unlabeled H_ox-CO was populated, suggesting two distal ¹³CO ligands and µ¹³CO (vi). Exchange to a ¹²CO atmosphere resulted in a ∼30 cm⁻¹ up-shift of band δ, small shifts to higher frequencies of β and γ, and no change of band α. This pattern agrees with d₁¹³CO and μ¹³CO labeling (viii). Red light irradiation of viii under ¹²CO yielded a state showing similar β, γ, and δ frequencies as xii, but α remained at its low frequency so that only μ¹³CO was present (ix). ¹³CO exposure converted ix to a state reminiscent of spectrum ii, including d₂¹³CO and μ¹³CO labeling (xi). Finally, blue light irradiation of xi under ¹²CO regained unlabeled H_ox-CO (xii). These results suggested that µCO and the distal carbonyls were exchangeable in HYDA1, but not the proximal CO ligand.

At increased humidity of the ¹³CO aerosol and blue light irradiation, HYDA1 with three ¹³CO ligands (vi) produced down-shifts of all four CO bands compared with the unlabeled species. This state was assigned to completely ¹³CO-labeled H_ox-CO (33), including the proximal CO ligand (Fig. 2B, xii). Exposure to ¹²CO caused a ∼40 cm⁻¹ up-shift of δ with only minor changes for γ and β and no difference for α (ii). Further red light irradiation mainly up-shifted band γ by ∼40 cm⁻¹ (iii), which suggested stepwise replacement of the two ¹³CO ligands at Fe_d by ¹²CO in the presence of p¹³CO. Further ¹²CO exposure under blue light induced the exchange of µCO as indicated by a ∼40 cm⁻¹ up-shift of band α (vi). Rebinding of ¹³CO to iii or vi yielded species v or viii, their δ band positions suggesting a single distal ¹³CO ligand. Red light irradiation under ¹³CO of viii restored the frequency pattern of xii except for the down-shifted band α (ix). The latter was exchanged only under blue light (xii). ¹²CO exposure of viii finally regained species ii. Selective ¹³CO editing of pCO was facilitated only in sufficiently hydrated HYDA1 protein films.

Complementary ¹³CO editing experiments were performed for CPI and DDH (SI Appendix, Fig. S8). ¹³CO exchange of the two distal carbonyls was achieved already under red light in these enzymes, possibly related to increased light absorption in the presence of the accessory iron-sulfur clusters, whereas HYDA1 allowed sequential editing with red and blue light. Four of the eight possible H_ox-CO isotopic species excluding p¹³CO were populated in the bacterial enzymes. The CO frequencies, however, were similar in the three enzymes.

H_ox-CO Isotopic Species Assignment from DFT.

The IR experiments showed 16 distinct H_ox-CO isotopic species with all possible labeling patterns. We calculated IR spectra for 96 H_ox-CO models, including 16 possible ¹³CO-labeling patterns with six CO/CN⁻ rotamers each (SI Appendix, Fig. S12 and Table S2). Similarly large RMSD values (∼30 cm⁻¹) were observed for all isotopic species with an apical dCO, which precluded assignment of the experimental IR spectra for the “standard” H_ox-CO geometry. Species with an apical dCN⁻ showed significantly diminished RMSD values for all isotopic labeling patterns. These results facilitated the unambiguous attribution of each IR spectrum to a specific H_ox-CO species (Table 1). Both medium and large models showed diminished preference for the dCN⁻ rotamer compared with the small H-cluster model (SI Appendix, Table S2), but still a twofold smaller RMSD was observed for the structure with an apical dCN⁻ ligand. Comprehensive analysis of experimental and calculated IR band frequencies and intensities suggested that H_ox-CO structures with proximal CO/CN⁻ inversion were disfavored and further supported an apical dCN⁻ (SI Appendix, Table S4). These results indicated the cyclic isotope editing sequence shown in Fig. 5. The exogenous CO ligand (d₂CO) is exchangeable in darkness, red light sensitivity is attributed to the equatorial d₁CO, and blue light induces exchange of µCO and pCO, the latter being feasible only in sufficiently hydrated HYDA1 protein films.

Fig. 5.

Stepwise isotope editing of the H-cluster. Gray shadings highlight eight differently labeled H_ox species providing access to the catalytic cycle of hydrogen turnover. Carbonyl ligand patterns are shown in the order p µ d₁ d₂ (d₂ is present only in H_ox-CO). Exposure to ¹³CO (magenta) or ¹²CO (green) gas is indicated only for the dark steps (solid black arrows) and persisted during the following red or blue light irradiation steps (colored arrows) in the experimental cycle; dashed arrows denote N₂ exposure in darkness. The proximal CO ligand is prone to ¹³CO exchange only in sufficiently hydrated (H₂O) protein films.

Site-Selective ¹³CO Editing and Rotamers of H_ox.

Quantitative population of four H_ox isotopic species with zero to two ¹³CO ligands excluding pCO was achieved by N₂ gas exposure of HYDA1 protein films at low humidity (Fig. 2A). H_ox-CO species xii and ii were converted into unlabeled H_ox (i). In comparison with i, H_ox-CO species iii and v were converted into a state showing a ∼30 cm⁻¹ down-shift of band β and a smaller shift of γ, implying a single ¹³CO ligand at Fe_d (d¹³CO) (iv). Species vi and viii yielded a H_ox state similar to iv, but showing an additional ∼40 cm⁻¹ down-shift of α due μ¹³CO labeling (vii). Finally, species ix and xi were converted into a state with an exclusive ∼40 cm⁻¹ down-shift of α compared with unlabeled H_ox, indicative of μ¹³CO (x). Starting with completely ¹³CO-labeled and hydrated HYDA1 in the H_ox-CO state, four H_ox species with one to three ¹³CO ligands including pCO were populated by N₂ exposure (Fig. 2B). H_ox-CO species ii and xii were converted to H_ox species i with bands α, γ, and β shifted ∼40 cm⁻¹ to lower frequencies (complete ¹³CO exchange). H_ox-CO species with ¹³CO at Fe_p (p¹³CO) and ¹²CO at Fe_d (iv and vii) were converted to H_ox species iv and vii showing a γ band intensity (1955 cm⁻¹) exceeding the one of band β (1905 cm⁻¹), which was reversed for H_ox species with unlabeled pCO. These are the only H_ox isomers with pronounced vibrational coupling of dCO and pCO (SI Appendix, Fig. S9). H_ox-CO species ix and xi finally were converted to H_ox species x, which resembled species i except for presence of µ¹²CO.

IR band patterns for the 56 possible H_ox structures (7 CO/CN⁻ rotamers with 8 ¹³CO-labeling patterns each) were calculated (SI Appendix, Table S1). Comparison of experimental and calculated CO frequencies revealed by far lowest RMSD values only for isotopic patterns in agreement with the above experimental assignments (Table 1). Analysis of IR band frequencies and intensities of H_ox (SI Appendix, Table S1 and Fig. 11) and mutual comparison with the results for H_ox-CO (SI Appendix, Table S4) excluded dCO in more apical position. On the other hand, the calculated IR data of a structure with dCN⁻ rotated toward a more apical position were as well in agreement with the experimental data as the standard ligand configuration, for all isotopic species of H_ox (Table 1). Both these structures accounted for vibrational coupling of pCO and dCO in the presence of a proximal ¹³CO (SI Appendix, Fig. S9), which explained the inverted intensity ratio of the β and γ bands in H_ox species iv and vii.

Discussion

Our protocol for controlled gas exposure, irradiation, and hydration of [FeFe]-hydrogenase protein films facilitates quantitative population of 8 H_ox and 16 H_ox-CO species selectively labeled with zero to four ¹³CO ligands. Fourier-transform IR spectroscopy in ATR configuration facilitates rapid gas exchange for controlled and quantitative state population in [FeFe]-hydrogenase protein films. These experiments have provided an unprecedentedly large IR data set for comparison with quantum chemical calculations. The CO vibrational modes underlying the IR spectra were assigned unambiguously. In H_ox, experimentally observed CO stretching frequencies are well separated and differ by at least 24 cm⁻¹ (pCO/dCO). This feature facilitates direct band assignment via ¹³CO isotope editing. In contrast to H_ox, the three terminal carbonyls in H_ox-CO show pronounced vibrational coupling that results from changes in ligand geometry and [2Fe]_H spin distribution (24–28, 31, 36). Disentangling of spectral shifts as induced by stepwise isotope editing of H_ox-CO was achieved via DFT analysis. Our results imply a consistent reaction cycle for isotopic editing of the H-cluster (Fig. 5).

H_ox-CO in standard configuration (27, 46) is not in good agreement with the experimental carbonyl vibrations. Models comprising an apical CN⁻ ligand at Fe_d yielded a vibrationally uncoupled pCO, which is a characteristic feature of the H-cluster (24, 26). Only these models reproduced the altered origin of the pCO vibrational frequency and inverted band intensities for species including d₁¹³CO and d₂¹³CO. Improved correlation of experimental and calculated IR data for H_ox-CO with apical dCN⁻ has been discussed before, but evaluated against insufficiently small experimental IR datasets (39, 47, 48). We prove the effect for 16 H_ox-CO species, three phylogenetically distinct [FeFe]-hydrogenases, and varying computational approaches. However, our analysis clearly supports the ligand arrangement at the proximal iron ion in the crystallographic data (21–23).

Available H-cluster structures were modeled with trans equatorial carbonyls and square-pyramidal (H_ox) or octahedral geometries (H_ox-CO) at the distal iron ion (21–23, 27, 46, 49). At a resolution of ∼1.5 Å or less, however, CO/CN⁻ discrimination remains speculative. These ligands originally were assigned using potential hydrogen bonding of CN⁻ ligands to protein residues (21, 40, 49–51) (SI Appendix, Fig. S10) and before the identity of the adt ligand was unraveled (19). A computational study on the DDH crystal structure preferred the “standard” H_ox-CO geometry by ∼6 kJ/mol due to interaction of dCN⁻ with a backbone amine and the conserved Lys237 (39, 48). An interaction between Lys237 and dCN⁻ has also been inferred from EPR but was not supported later (20, 52). Our analysis for all model structures suggests slight distortion of octahedral Fe_d symmetry in the standard model, whereas for an apical CN⁻ weak H-bonding to the adt nitrogen base occurs (Fig. 4). This geometry was earlier calculated to be stabilized by ∼8 kJ/mol (48). It has been suggested that H₂ may form a similar H-bond to adt during the catalytic reaction (37, 49, 51, 53). Substrate (H₂) or inhibitor (CO) binding at the active site thus may be governed by intramolecular rather than protein–cofactor interactions. The detailed influence of the protein environment on the fine structure of the H-cluster is difficult to quantify both from experimental and theoretical viewpoints. Our general isotope editing scheme (Fig. 5), however, remains valid irrespective of the precise angular arrangement of the distal ligands.

The H_ox standard configuration and a rotamer with more apical CN⁻ and equatorial vacancy at Fe_d showed similar and superior agreement between experimental and calculated IR data. Accordingly, such structures appear equally probable. Our analysis further favors trans orientation of equatorial carbonyls and a proximal CO/CN⁻ arrangement as in crystallographic assignments (21–23, 49). The HYDA1 and CPI proteins used in this study were activated in vitro with a synthetic diiron site analog (39, 40). We observed no significant differences between our HYDA1 and CPI preparations and the natively maturated DDH so that rotamer formation during in vitro maturation can be excluded (12, 17, 20, 32, 54). The observation that only sufficient hydration of HYDA1 protein films facilitates isotope editing at the proximal iron ion rather indicates that structural flexibility of gas channels (55) is involved in ligand exchange. Under cryogenic conditions (i.e., for diffraction data collection), the standard H_ox structure thus dominates. Biologically relevant conditions (i.e., dissolved protein at room temperature as used here) could promote equilibrium between the two ligand geometries at Fe_d or even dominance of the rotamer with more apical dCN⁻. Such equilibria exist for diiron compounds in solution (56–58). This view is further reinforced by molecular dynamics simulations on DDH showing that distal ligand rotation is related to motions by up to 2 Å of a nearby phenylalanine side chain (36). Only in the rotated H_ox structure, CO can bind in equatorial position at Fe_d (Fig. 4). Ligand dynamics also impacts on possible motifs of substrate (H₂) interactions with the active site.

H_ox is the entry point to the hydrogen conversion cycle of [FeFe]-hydrogenases (1). At least two increasingly reduced H-cluster species were derived from H_ox; their molecular structures and involvement in catalysis yet remain to be defined (11–15, 36, 59). The fate of the Fe-Fe bridging carbonyl is of particular mechanistic interest. Binding of hydrogen species in apical position at Fe_d of the H-cluster is believed to be essential for catalysis (1). However, configurations with (semi) bridging or equatorial H-species were considered as well (12, 14, 36) and may result from structural flexibility of the H-cluster (36). Such structural dynamics may facilitate apical or equatorial ligand binding at the distal iron ion and may also be relevant for O₂ inactivation of the enzymes via reactive oxygen species formation (24, 29, 30, 60, 61). Our protocol for selective preparation of H_ox with eight distinct isotopic labeling patterns introduces spectroscopic probes at individual positions at the cofactor. This approach opens the road for investigations on novel isotopically labeled intermediates in the catalytic cycle to probe structural dynamics during the H₂ conversion chemistry of [FeFe]-hydrogenases.

Materials and Methods

HYDA1 Protein Preparation.

[FeFe]-hydrogenase HYDA1 and CPI apo-proteins were overexpressed in Escherichia coli, purified, and quantitatively reconstituted in vitro with a synthetic diiron complex [Fe₂(µ-adt)(CO)₄(CN)₂, adt = (SCH₂)₂NH] (23, 41). All protein preparation and handling procedures were carried out under strictly anoxic conditions and dim light. DDH was purified from D. desulfuricans with a complete H-cluster (43).

Infrared Spectroscopy.

ATR-FTIR spectroscopy (62) was performed with a Tensor27 spectrometer (Bruker) placed in an anaerobic glovebox and equipped with a mid-IR globar, a liquid-nitrogen–cooled MCT detector, and a silicon prism with two active reflections, which was capped by a sealed PCTFE head-space gas compartment. Infrared spectra were recorded with 1-cm^–1 spectral resolution using varying numbers of interferometer scans on thin protein films, corrected for background contributions, and evaluated using a least-squares-fit algorithm. Hydrogenase films were exposed to ¹³CO, ¹²CO, or N₂ gas by fast exchange of the head-space atmosphere using a multichannel mass flow controller (Sierra Instruments) at room temperature. All gases were sent pro rata through a water-filled wash bottle to create an aerosol that prevents dehydration of protein films. This allowed controlling the water/ protein ratio in the film (hydration) and influenced the velocity of any gas-processing reaction. Humidity refers to the water/gas ratio in the aerosol. A Schott white light source with band pass filters (center wavelengths 640 or 460 nm) was used for irradiation of protein samples. Details on real-time ATR-FTIR experiments and data evaluation are given in SI Appendix, Figs. S1–S4.

Quantum Chemical Calculations.

DFT calculations on H-cluster model structures with ¹²CO/¹³CO ligands were carried out using Gaussian09 (63) on the Soroban computer cluster of the Freie Universität Berlin. Starting structures of increasing complexity (SI Appendix, Fig. S5) were constructed using the crystal structure of CO-inhibited CPI [FeFe]-hydrogenase (27) as a template and geometry-optimized using the BP86/TZVP or TPSSh/TZVP functional/basis-set combinations (64–66), and IR spectra were calculated thereafter (67). Details of the computational methods are given in SI Appendix. The calculated structures can be accessed via Dataset S1.