Electronic structure of two catalytic states of the [FeFe] hydrogenase H-cluster as probed by pulse electron paramagnetic resonance spectroscopy

Abstract

The active site of the [FeFe] hydrogenase (HydA1), the H-cluster, is a 6-Fe cofactor that contains CO and CN⁻ ligands. It undergoes several different oxidation and protonation state changes in its catalytic cycle to metabolize H₂. Among them, the well-known H_ox state and the recently identified H_hyd state are thought to be directly involved in H₂ activation and evolution, and they are both EPR active with net spin S = 1/2. Herein, we report the pulse electronic paramagnetic spectroscopic investigation of these two catalytic states in Chlamydomonas reinhardtii HydA1 (CrHydA1). Using an in vitro biosynthetic maturation approach, we site-specifically installed ¹³C into the CO or CN⁻ ligands and ⁵⁷Fe into the [2Fe]_H subcluster of the H-cluster in order to measure the hyperfine couplings to these magnetic nuclei. For H_ox, we measured ¹³C hyperfine couplings (¹³CO a_iso of 25.5, 5.8, and 4.5 MHz) corresponding to all three CO ligands in the H-cluster. We also observed two ⁵⁷Fe hyperfine couplings (⁵⁷Fe a_iso of ~17 MHz and 5.7 MHz) arising from the two Fe atoms in the [2Fe]_H subcluster. For H_hyd, we only observed two distinct ¹³CO hyperfine interaction (¹³CO a_iso of 0.16 and 0.08 MHz), and only one for ¹³CN⁻ (¹³CN a_iso of 0.16 MHz)—the couplings to the ¹³CO/¹³CN⁻ on the distal Fe of [2Fe]_H may be too small to detect. We also observed a very small (< 0.3 MHz) ⁵⁷Fe HFI from the labeled [2Fe]_H subcluster, and four ⁵⁷Fe HFI from the labeled [4Fe-4S]_H subcluster (⁵⁷Fe a_iso of 7.2, 16.6, 28.2, and 35.3 MHz). These hyperfine coupling constants are consistent with the previously proposed electronic structure of the H-cluster at both H_ox and H_hyd states, and provide a basis for more detailed analysis.

Graphical Abstract

Introduction

Hydrogenases are used by nature to catalyze the reversible conversion between H₂ and H⁺/e⁻.^1–2 In [FeFe] hydrogenases, the catalytic center “H-cluster” consists of a tetracysteine (Cys) coordinated [4Fe-4S]_H subcluster linked through one Cys residue thiol to an organometallic [2Fe]_H subcluster that has CO and CN⁻ ligands along with an azadithiolate (adt) ligand that bridges the two iron ions (Scheme 1).^3–6 The H-cluster has been shown to adopt different oxidation and protonation states that have unique spectroscopic properties: the most oxidized state, H_ox, the one-electron reduced state (H_red, H_redH⁺, H_red*), and the two-electron reduced state (H_sred, H_sredH⁺, H_sred*).^7–12 Most recently, another state of the H-cluster with a terminal iron hydride, H_hyd, has been identified in Chlamydomonas reinhardtii HydA1 (CrHydA1), in the wild-type enzyme and several variants.^13–17 Importantly, H_hyd is considered to be one of the key state in the catalytic cycle of [FeFe] hydrogenase,^{13, 15, 18} as the terminal hydride is thought to coordinate with the pendant amine of the adt ligand to form a frustrated Lewis pair that is necessary for H₂ activation and evolution.^19–20 The full characterization of H_hyd is therefore crucial to further understand the mechanism of H-cluster-based catalysis.

Scheme 1.

Proposed catalytic cycle of the H-cluster highlighting the two S = 1/2 states, H_ox and H_hyd.

Both S = 1/2 EPR-active states, H_ox and H_hyd, represent different forms of the H-cluster in the proposed catalytic cycle of [FeFe] hydrogenase (Scheme 1).^15–16 Results from Mössbauer spectroscopy, electron nuclear double resonance (ENDOR) spectroscopy, and DFT calculations support a [4Fe-4S]_H²⁺ -[Fe^IIFe^I]_H configuration for H_ox, with a net spin of S = 1/2 that is primarily localized on the mixed-valance [2Fe]_H, and less spin density on the [4Fe-4S]_H²⁺ (which would be a diamagnetic cluster if isolated from the paramagnetic [Fe^IIFe^I]_H).^{12, 21–22} Hyperfine interactions (HFI) for specifically ¹³CN⁻ and ⁵⁷Fe -labeled [2Fe]_H have been reported for H_ox and its CO-inhibited state, H_ox-CO.^23–27 These ⁵⁷Fe and ¹³C hyperfine coupling constants have been used to estimate the spin density distribution in the H-cluster. More recently, the ¹H NMR spectrum was also reported for H_ox, providing further insights into its electronic structure.²⁸ Despite these efforts to characterize H_ox, questions still remain. One in particular is why only one ⁵⁷Fe HFI has been detected for the [2Fe]_H of Clostridium pasteurianum CpI and Chlamydomonas reinhardtii HydA1,^{27, 29} yet two distinct HFI for the coordinating ¹³CN⁻ have been found, one assigned as a ligand to the proximal, nominally Fe^II ion (Fe_p) (¹³C a_iso = 4.9 MHz), and one assigned as a ligand to the spin-bearing distal Fe^I (Fe_d) (¹³C a_iso = 27 MHz).²⁵ Simulations of Mössbauer results also implied that only one large ⁵⁷Fe HFI should be present in H_ox.¹² Similarly in Desulfovibrio desulfuricans DdH, while identical ⁵⁷Fe hyperfine couplings were experimentally measured for the two Fe sites in [2Fe]_H,³⁰ DFT calculations using a variety of models suggested that their couplings should be significantly different.³¹ Clarifying this inconsistency is not only of fundamental interest, but would also shed light in designing molecular models for hydrogenase.

As for H_hyd, while the proposed formula of H_hyd may be viewed as binding of a terminal hydride on the Fe_d of H_ox, previous studies suggested that H_hyd adopts a [4Fe-4S]_H⁺-[Fe^IIFe^II]_H-H⁻ configuration, with a net spin of S = 1/2 that is now primarily localized on the [4Fe-4S]_H.¹⁶ This assignment is supported by the g-tensor of H_hyd (g = [2.08, 1.94, 1.88]) that is similar to that found for typical [4Fe-4S]⁺ clusters, the presence of two low-spin diamagnetic Fe²⁺ in the [2Fe]_H with small isomer shifts (δ = 0.03 and 0.23 mm/s) as revealed by Mössbauer spectroscopy, as well as by DFT calculations.^15–16 Interestingly, the proposed electronic structure of H_hyd is analogous to that of a reaction intermediate we recently trapped in HydG,³² the enzyme responsible for the early steps of H-cluster bioassembly.^{29, 33} Specifically, we identified a [4Fe-4S]⁺-[(Cys)Fe^II(CO)(CN)] species and observed small ¹³C (a_iso = 0.22 and 0.24 MHz) and ⁵⁷Fe (a_iso = 0.42 MHz) HFI from the specifically labeled ¹³CO, ¹³CN⁻ ligands and the low-spin dangler ⁵⁷Fe^II respectively.³² Therefore, we reason that similarly small HFI could be observed for H_hyd and these HFI could be used to clearly define the electronic configuration of H_hyd.

The Swartz lab has developed an in vitro approach to assemble the H-cluster by using apo-hydrogenase (that harbors only the [4Fe-4S]_H), cell lysate of maturation enzymes—HydE, HydF and HydG—responsible for the biosynthesis of the [2Fe]_H,^34–36 and necessary substrates/cofactors.^37–39 This “cell free synthesis” method enables us to facilely and precisely install various isotopes into the [2Fe]_H^{25, 29} In this work, we have taken advantage of this strategy and selectively labeled CrHydA1 with ¹³CO, ¹³CN⁻, and ⁵⁷Fe, allowing us to determine the electronic structures of the H-cluster at both H_ox and H_hyd states via the HFI of these various magnetic nuclei. These results provide further insights into the structure and function of this unique cofactor.

Methods

General Consideration.

⁵⁷Fe and, ¹³C- and ¹⁵N- labeled Tyr were purchased from Cambridge Isotope Inc. All other chemicals were from Sigma-Aldrich unless specified. All handling of Fe-S cluster proteins, which includes purification of CrHydA1, preparation of HydE, HydF and HydG lysate, maturation of CrHydA1, and preparation of EPR samples, were carried out in an anaerobic chamber with O₂ level < 1 ppm.

In vitro maturation of CrHydA1.

Maturation experiments were performed according to previous procedure with slight modifications.^{29, 37} These experiments required apo-CrHydA1, lysate of cells overexpressing Shewanella oneidensis HydE, HydF and HydG, and necessary small molecules. Briefly, apo-CrHydA1 was expressed in a recombinant E. coli BL21(DE3) ΔiscR strain that harbors the plasmid containing codon-optimized Chlamydomonas reinhardtii hydA1 gene with an N-terminal Strep-II purification tag. Four liters of E. coli cells were grown in 25 g/L LB medium with 40 mg/L kanamycin, 100 mg/L ampicillin, 2 mM ammonium ferric citrate, 0.5% (w/v) glucose and 100 mM 3-(N-morpholino)propanesulfonic acid (MOPS, pH = 7.8), at 30 °C to a OD₆₀₀ of ~0.5, at which point the cultures were pooled, transferred into the anaerobic chamber, and supplemented with 5 mM Cys and 10 mM fumarate. The pooled culture was stirred in the chamber for 30 min to deplete the remaining oxygen in the solution and protein expression was then induced by 0.25 mM isopropyl β-D-1-thiogalactopyranoside (IPTG). After 20 h, cells were harvest by centrifugation and frozen. For protein purification, cells were lysed in a HEPES buffer (buffer A, 50 mM HEPES, 150 mM KCl, pH = 8.0, with 1 mM freshly made sodium dithionite added) containing 1xBugbuster detergent solution (EMD Millipore), 25 U/mL benzonanse (EMD Millipore), 1 kU/mL rLysozyme (EMD Millipore) and one EDTA-free protease inhibitor cocktail tablet (Roche). After the removal of cell debris by centrifugation, the clear supernatant was loaded to Strep-Tacin resin (~ 50 mL resin) in a gravity column. The column was washed with buffer A and then CrHydA1 was eluted by buffer A containing 3 mM desthiobiotin. The dark fractions were collected, concentrated as necessary and stored at −80 °C.

To make cell lysate containing untagged HydE, HydF or HydG, recombinant E. coli cells containing the corresponding plasmid were grown and lysed (~3 mL buffer A per gram of cell) same as above. Clear cell lysate was aliquoted and stored at −80 °C. Expression levels of the maturation enzymes in the cell lysate were examined by SDS-PAGE. Desthiobiotin in the apo-CrHydA1, introduced in protein purification, was removed by using a PD-10 desalting column (GE Healthcare) prior to the maturation reaction. For the maturation of CrHydA1, each 25 mL reaction contained 12.5 mL HydG lysate, 4 mL HydF lysate, 1.5 mL HydE lysate, 4 mM DTT, 1 mM Fe²⁺ (⁵⁷Fe²⁺ if necessary), 0.5 mM Na₂S, 2 mM S-adenosylmethionine (SAM), 2 mM Cys, 1 mM pyridoxal phosphate (PLP), 20 mM guanosine triphosphate (GTP), 2 mM tyrosine (Tyr, labeled as desired), 2 mM sodium dithionite and ~8 μM apo-CrHydA1 (added in that order). The pH of the reaction mixture was adjusted to ~7.5 before dithionite was added. The reaction mixture was incubated at room temperature in an anaerobic chamber containing 2% H₂ for ~16 h, and then clarified by centrifugation. Maturated CrHydA1 was re-purified from the supernatant as abovementioned (~5 mL Strep-Tactin resin). Fractions containing CrHydA1 were pooled and concentrated as necessary.

Purification of uniformly ⁵⁷Fe-enriched CrHydA1.

This procedure is same as reported in the previous study.²⁹ Briefly, [6⁵⁷Fe]-CrHydA1 (that harbors the maturated H-cluster with all six iron labeled) was expressed in a recombinant E. coli BL21(DE3) ΔiscR strain that harbors two plasmids, encoding the Chlamydomonas reinhardtii hydA1 gene and Shewanella oneidensis hydEFGX genes, respectively. Cell growth and protein purification were performed same as above, except that the LB media was supplemented with ⁵⁷Fe instead of natural abundance Fe.

EPR sample preparation.

EPR samples were made from freshly purified or maturated holo-CrHydA1. To make H_ox samples, 2 mM thionine was added to ~ 500 μM CrHydA1. The solution was immediately transferred into the EPR tube and frozen for storage. To make H_hyd samples, 300 mM dithionite made in buffer A was added to ~ 500 μM CrHydA1.¹⁶ The mixture was immediately transferred into the EPR tube and frozen for storage.

EPR spectroscopy.

EPR spectroscopy was performed in the CalEPR center in Department of Chemistry, University of California at Davis. X-band (9.4 GHz) Continuous Wave (CW) EPR spectra were recorded on a Bruker Biospin EleXsys E500 spectrometer equipped with a super high Q resonator (ER4122SHQE). Cryogenic temperatures were achieved and controlled by using an ESR900 liquid helium cryostat, a temperature controller (Oxford Instrument ITC503) and a gas flow controller. All CW EPR spectra were recorded at slow-passage, non-saturating conditions. Spectrometer settings were: conversion time, 40 ms; modulation amplitude, 0.5 mT; modulation frequency, 100 kHz; and other settings given in figure captions. Pulse Q-band (34 GHz) hyperfine sublevel correlation (HYSCORE) and electron nuclear double resonance (ENDOR) experiments were performed on the Bruker Biospin EleXsys 580 spectrometer using a R.A. Isaacson cylindrical TE₀₁₁ resonator.⁴⁰ Cryogenic temperatures were achieved and controlled with an Oxford Instrument CF935 cryostat. The pulse sequences employed were as follows: free induction decay (FID)-detected field swept EPR (π/2-FID), electron spin-echo detected field swept EPR (π/2-τ-π-τ-echo), HYSCORE (π/2-τ-π/2-t1-π-t2-π/2-τ-echo), Davies-ENDOR (π-RF-π/2-τ-π-τ-echo), and Mims-ENDOR (π/2-τ-π/2-RF-π/2-τ-echo). HYSCORE time domain data were baseline corrected with 3^rd polynomial, hamming-window applied, zero-filled to eight-fold of data points and fast Fourier-transformed to obtain the frequency domain results. Simulations of CW and pulse EPR spectra were performed in Matlab 2014a (MathWorks, Inc,) with EasySpin 5.2.13 toolbox.⁴¹ Euler angles are from A tensors to g tensors and follow z-y-z convention.

Orientationally disordered radical systems usually exhibit anisotropic hyperfine coupling interaction (HFI) that can be decomposed into an isotropic component, a_iso and an anisotropic component, T. At specific orientations and for nuclei with nuclear spin I = 1/2 (¹³C and ⁵⁷Fe in this study), the ENDOR transitions for the m_s = ± 1/2 electron manifolds are observed, to first order approximation, at frequencies: v± = |v_N ± A/2|, where v_N is the nuclear Larmor frequency and A is the orientation-dependent hyperfine coupling.⁴² In the weak coupling case where v_N > A/2, the two ENDOR peaks are centered at v_N, and split by A. In the strong coupling case where v_N > A/2, the two ENDOR peaks are centered at A/2, and split by 2v_N. M ims-ENDOR⁴³ can be used to detect very small A with superb sensitivity, but its response intensity is modulated by a factor related to A and the time interval between the first and the second microwave pulse (τ): R ~ [1-cos(2πAτ)]. Therefore, the intensity of Mims-ENDOR vanishes at A = n/τ (n = 0, 1, 2 …), known as the “Mims-holes”. As such, larger hyperfine couplings are typically probed by Davies-ENDOR.⁴³

HYSCORE spectroscopy detects the modulation of an electron spin-echo by the spin flipping of nearby magnetic nuclei.⁴³ The two-dimensional spectra reveal crosspeaks, the positions of which rely on A and v_N. For an S = 1/2 and I = 1/2 system and in the weak coupling case, the crosspeaks appear in the first quadrant at (v_N + A/2, v_N - A/2) and (v_N - A/2, v_N + A/2). In the strong coupling case, the crosspeaks appear in the second quadrant at (v_N - A/2, v_N + A/2) and (-v_N - A/2, - v_N + A/2). For HFI with large anisotropy, empirical analysis using these simple formula is not straightforward, and spectral simulation is necessary to extract the HFI parameters.

Results and discussion

[¹³CO] in H_ox.

We previously reported the ¹³C HFI of the two ¹³CN⁻ ligands in the Clostridium pasteurianum hydrogenase (CpI) in the H_ox and H_ox-CO states,²⁵ and here we turn our focus to the three CO ligands. The CO and CN⁻ ligands in the H-cluster are sourced from Tyr which is cleaved by the enzyme HydG.^{33, 44} In order to make ¹³CO-specifically labeled CrHydA1, we used 1-¹³C-Tyr in the in vitro maturation reaction that generates the ¹³CO ligands to the H-cluster. The maturated CrHydA1 poised in the H_ox state was examined by CW EPR (Figure 1A), which revealed a rhombic g-tensor of [2.103, 2.042, 1.998], consistent with published results for CrHydA1.⁴⁵ A small amount of H_ox-CO contamination (17%, g = [2.056, 2.008, 2.008]) is also present, as is often observed in [FeFe] hydrogenase samples.⁴⁵ The observed ~1 mT splitting at the three principal g values indicates a large HFI from ¹³CO, which is similar to the observation in CpI [¹³CN]-H_ox.²⁵ The ¹³C HFI were further analyzed by the field-dependent Q-band Davies-ENDOR spectra recorded at the field positions without the contribution from H_ox-CO. We observed a total of three sets of ¹³C ENDOR signals corresponding to the three ¹³CO ligands in H_ox (Figure 1B). In line with the previous study on the HFI of the ¹³CN⁻ ligands and the established electronic structure of H_ox,^11–12 we assigned the ¹³C ENDOR signal with the largest HFI, a_iso = 25.5 MHz (Figure 1B, green shade), to the distal ¹³CO (¹³CO_d), the ¹³C ENDOR signal with the smallest HFI, a_iso = 4.5 MHz (Figure 1B, blue shade), to the proximal ¹³CO (¹³CO_p), leaving the remaining ¹³C ENDOR signal with an HFI in-between, a_iso = 5.8 MHz (Figure 1B, red shade), that must arise from the bridging ¹³CO (¹³CO_bridge). Notably, the ¹³C HFI of ¹³CO_d is close to the cancellation limit, A~2v_c, and only the higher frequency ENDOR transition is detected as the lower frequency one is close to zero and is best observed with Q-band HYSCORE spectra (Figure S1). The complex ENDOR lineshape of the higher frequency transition in the center field positions are due to narrow orientation selection of this relatively anisotropic HFI. The ¹³C HFI of the other two ¹³CO ligands are also confirmed by X-band HYSCORE spectroscopy (Figure S1).

Figure 1.

(A) X-band CW EPR spectrum of thionine-oxidized [¹³CO]-CrHydA1 (black trace) and the total simulation (magenta trace). The two spectral components are: H_ox, g = [2.103, 2.042, 1.998], 83% (red trace); H_ox-CO, g = [2.056, 2.008, 2.008], 17% (blue trace). Conditions: frequency, 9.4 GHz; temperature, 15 K; microwave power, 4 μW. (B) Q-band field-dependent Davies-ENDOR spectra (black trace) of [¹³CO]-CrHydA1 H_ox simulated with three sets of ¹³C HFI. Simulation: g = [2.103, 2.042, 1.998]; ¹³CO_d (green shade): A = [20.5, 29.9, 26.0] MHz, Euler angle = [37, 26, 0]°; ¹³CO_p (blue shade): A = [5.3, 4.0, 4.3] MHz, Euler angle = [25, 25, 0]°; ¹³CO_bridge (red shade): A = [9.0, 3.8, 4.5] MHz, Euler angle = [0, 20, 0]°. Conditions: frequency, 34.1 GHz; temperature, 15K; inversion pulse, 80 ns; RF pulse, 30 μs; π/2, 12ns; τ = 300 ns.

The ¹³C HFI of ¹³CO_p and ¹³CO_d both have magnitudes and degrees of anisotropy similar to those of the ¹³CN_p and ¹³CN_d ligand in CpI H_ox, which is consistent with the proposed local z-axis assignment of both Fe as pointing approximately along each Fe-CO_bridge bond.^{21, 25} The a_iso (¹³CO_d) : a_iso (¹³CO_p) ratio of 5.7 is essentially the same with the value determined for the ¹³CN⁻ ligands in CpI H_ox (5.8).²⁵ These findings support our previous electronic structure descriptions of H_ox; namely, that most spin density is carried by Fe_d and localized in the d_z² orbital of this low spin 3d⁷ Fe^I.²⁵ The ¹³C HFI of ¹³CO_bridge is more dipolar, with a degree of anisotropy similar to that of ¹³CO_bridge in DdH H_ox-CO, although the latter has a slightly larger a_iso (7.4 MHz).²⁴ The relatively small and anisotropic ¹³C HFI of the ¹³CO_bridge reflects the longer Fe-CO_bridge distance compared to Fe-CO_d (PDB ID: 3C8Y³), which diminishes the Fermi contact of the lobe of the spin-carrying Fe_d 3d_z² with the ¹³C nucleus.

In another hydrogenase sample that has higher H_ox-CO percentage (80%), we were able to observe four ¹³C ENDOR signals on the Davies-ENDOR spectra recorded at g = 2.008 (Figure S2), g2 of H_ox-CO, with HFI values of 4.8, 7.3, 10.9 and 21.1 MHz, corresponding to the distal, proximal, bridging, and external ¹³CO, respectively. These values are assigned by comparing to the ¹³CO HFI values found in DdH and CpI H_ox-¹³CO and the ¹³CN⁻ HFI values found in CpI H_ox-CO.^24–26 The HFI values we found are in good agreement with previous results. The ¹³C HFI values and tensors for diatomic ligands in different states of the H-cluster are summarized in Table 1.

Table 1.

Summary of ¹³C HFI of the diatomic ligands in different states of the H-cluster

H-cluster state	A 13C (MHz)	[α, β, γ] (°)e	assignment	reference
Hoxa	[5.3, 4.0, 4.3]	[25, 25, 0]	COp	this work
	[20.5, 29.9, 26.0]	[37, 26, 0]	COd	this work
	[9.0, 3.8, 4.5]	[0, 20, 0]	CObridge	this work
	[5.22, 5.24, 4.16]	[30, 90, 0]f	CNp	25
	[30.9, 23.3, 30.2]	[60, 120, 170]f	CNd	25
Hox-COb	7.3		COp	this work
	[3.2, 3.7, 4.4] 4.8		COd	24 this work
	[8.5, 9.8, 3.9] 10.9		CObridge	24 this work
	[15.6, 16.6, 19.2]; 21d, 21.1;		COexternal	24, 26 this work
	[7.2, 7.0, 7.0]	[0, 0, 0]	CNp	25
	[3.90, 3.75, 3.75]	[0, 0, 0]	CNd	25
Hhydc	[0.30, −0.18, −0.59]	[25, 0, 0]	COp	this work
	[−0.51, 0.16, 0.10]	[35, 0, 0]	CObridge	this work
	[0.07, 0.17, 0.22]	[0, 10, 0]	CNp	this work

^a.13CO HFI are from CrHydA1 H_ox; ¹³CN HFI are from CpI H_ox.

^b.13CO HFI are from CrHydA1 and DdH H_ox-CO; ¹³CN HFI are from CpI H_ox-CO.

^c.From CrHydA1 H_hyd.

^d.From CpI H_ox-CO.

^e.Euler angle are relative to g-frame as g1>g2>g3.

^f.Euler angles are relative to g-frame as g1<g2<g3.

[¹³CO] and [¹³CN] in H_hyd.

Following the published procedures to reduce CrHydA1 H_ox with high concentration of dithionite,^14–16 we were able to readily generate CrHydA1 H_hyd. The Q-band echo-detected EPR spectrum of the resulting sample reveals an S = 1/2 species with g = [2.076, 1.936, 1.881] (Figure 2A), identical to the reported g values of H_hyd generated from wild-type CrHydA1.¹⁶ A small amount of H_ox-CO is also present.¹⁵ Since both Fe_p and Fe_d in H_hyd are proposed to be low-spin diamagnetic Fe²⁺,¹⁶ the ¹³C HFI of their diatomic ligands are expected to be small. Indeed, we observed small ¹³C HFI on the Mims-ENDOR spectra collected across the EPR absorption envelope of the [¹³CO]-H_hyd sample (Figure 2B). Two sets of splittings, centered at the Larmor frequency of ¹³C and separated by 0.2–0.5 MHz, were observed at most field positions. To clarify whether they arise from one or two ¹³C HFI tensors, we also performed variable mixing time (VMT) Mims-ENDOR experiments which measure the absolute signs of the HFI tensors.⁴⁶ The VMT Mims-ENDOR collected at g =1.936 (Figure S3), g2 of H_hyd, indicates that the two sets of peaks do not change in the same manner as t_mix is increased from 1 μs to 200 μs, and they also have oppositely signed HFI. Therefore, they must correspond to two different HFI tensors. Accordingly, we simulated the field-dependent ENDOR spectra with two sets of ¹³C HFI, with a_iso = −0.16 MHz (Figure 2B, blue shade) and −0.08 MHz (Figure 2B, red shade), respectively. Since in H_hyd, the spin density is proposed to localize primarily on the [4Fe-4S]_H that is at +1 state, the magnitude of the small ¹³C HFI, arising from spin polarization, would be largely governed by their distances to the [4Fe-4S]_H cluster. Recent DFT calculation has also suggested that Fe_p has higher spin density than Fe_d.¹⁸ As such, we assign the two observed ¹³C HFI tensors to the ¹³CO_p (−0.16 MHz) and ¹³CO_bridge (−0.08 MHz) respectively (Table 1), and we reason that the ENDOR features corresponding to ¹³CO_d are not observed likely because it is too distant from the unpaired electron spin leading to too small of a HFI.

Figure 2.

(A) Pseudo-modulated Q-band echo-detected EPR spectrum of dithionite-reduced CrHydA1 (black trace) and the total simulation (magenta trace). The two spectral components are: H_hyd, g = [2.076, 1.936, 1.881], 97% (green trace); H_ox-CO, g = [2.056, 2.008, 2.008], 3% (blue trace). Conditions: frequency, 34.1 GHz; temperature, 10 K; τ = 300 ns; modulation amplitude, 0.5 mT. (B, C) Q-band field-dependent Mims-ENDOR spectra (black trace) of [¹³CO]-CrHydA1 H_hyd simulated with two sets of ¹³C HFI (B), and that of [¹³CN]-CrHydA1 H_hyd simulated with one ¹³C HFI (C). Simulation: g = [2.076, 1.936, 1.881]; ¹³CO_p (blue shade): A = [0.30, −0.18, −0.59] MHz, Euler angle = [25, 0, 0]°; ¹³CO_bridge (red shade): A = [−0.51, 0.16, 0.10] MHz, Euler angle = [35, 0, 0]°; ¹³CN_p (red trace): A = [0.07, 0.17, 0.22] MHz, Euler angle = [0, 10, 0]°. Conditions for Mims-ENDOR: frequency, 34.1 GHz; temperature, 10K; RF pulse, 30 μs; π/2, 12 ns; τ = 600 ns.

To bolster this assignment, we prepared the ¹³CN⁻-labeled CrHydA1 H_hyd by using 2-¹³C-Tyr as the source for the ¹³CN⁻ ligand in the in vitro maturation, followed by reduction of the resulting enzyme with dithionite. In this case, the ¹³C ENDOR signals will be less complicated as only two ¹³CN⁻ ligands, bound to Fe_p and Fe_d respectively, are present. Indeed, field-dependent ¹³C Mims-ENDOR spectra of [¹³CN]-H_hyd disclose only one set of ¹³C ENDOR signals, with |a_iso| = 0.16 MHz (Figure 2C, red trace; Table 1), the same coupling as we assigned to ¹³CO_p, and therefore this ¹³C HFI can be assigned to ¹³CN_p. Again the ENDOR features corresponding to ¹³CN_d are not observed probably due to its very small HFI.

The fact that we detected two sets of ¹³CO but only one set of ¹³CN⁻ indicates that the second ¹³CO HFI must arise from the bridging, not distal, ¹³CO. W ere this ¹³CO to bind terminal on Fe_d, it would also have too small a HFI to be observed. The presence of a bridging CO is consistent with the FT-IR results for H_hyd where the observed 1857 cm⁻¹ peak is within the range of observed bridging CO stretch frequency.^{16 13}CO/¹³CN HFI results also indicate that while Fe_d is almost completely diamagnetic, Fe_p still carries a small but non-negligible amount of spin density, which may account for the isomer shift differences (~0.2 mm/s) between these two Fe centers in Mössbauer spectroscopy.¹⁶

Interestingly, the ¹³C of ¹³CN_p presents almost the same a_iso as the ¹³C in ¹³CO_p of H_hyd, but with a much smaller dipolar part, T, 0.05 MHz vs. 0.30 MHz. It is even smaller than that of the ¹³CO_bridge, 0.22 MHz. Since the spin density in H_hyd is primarily on the [4Fe-4S]_H cluster, the magnitude T can be estimated accounting for the distances between each ¹³C and the nearest Fe in the [4Fe-4S]_H cluster, using the point dipole approximation and projection factors typical for [4Fe-4S]⁺ clusters.^{42, 47–49} In the X-ray crystal structure of CpI H_ox (PDB ID: 3C8Y³), such distances are 3.8 Å, 5.0Å, and 5.6 Å, for ¹³CO_p, ¹³CN_p, and ¹³CO_bridge, corresponding to T of 0.4, 0.2, and 0.15 MHz, respectively. Neglecting in this approximation possible structural rearrangements in forming H_hyd, it appears that the T of ¹³CO_p and ¹³CO_bridge are in agreement with the observed dipolar interactions, while the T of ¹³CN_p is much smaller. The small anisotropy of ¹³CO/¹³CN⁻ HFI on Fe centers has been attributed to the strong covalent bonds between Fe and these π-acid ligands.⁵⁰ Upon formation of H_hyd, the [2Fe]_H electronic configuration shifts from Fe^IIFe^I to Fe^IIFe^II, leading to weakening of the Fe-CO bonds, as demonstrated by the increased FT-IR stretch frequencies of CO in H_hyd compared to H_ox indicating stronger C-O bonds, and with that, weaker Fe-CO bonds.^{10, 16} It is therefore possible that the weaker Fe-CO bonds in H_hyd result in the increased rhombicity of ¹³C HFI whereas the effects of Fe-CN bonds are much smaller.

[⁵⁷Fe] in H_ox.

The magnitudes of ¹³CO and ¹³CN⁻ HFI serves as reporters for the spin density of the Fe centers to which these diatomic ligands bind. In H_ox, however, while the ¹³CO/¹³CN⁻ HFI on Fe_p and Fe_d are very different (vide supra), previous studies on [FeFe] hydrogenases from several organisms found only one set of nearly isotropic ⁵⁷Fe HFI corresponding to the [2Fe]_H (A ⁵⁷Fe = 1 2.4 or 16–18 MHz).^{23, 27, 29} In an attempt to resolve this discrepancy, we sought to investigate the ⁵⁷Fe HFI of Fe_p and Fe_d using a combination of ENDOR and HYSCORE spectroscopies. While the molecular mechanism of the H-cluster bioassembly remains only partly understood, we found that selective installation of ⁵⁷Fe into the [2Fe]_H subcluster can be achieved by simply supplementing 1 mM of ⁵⁷Fe²⁺ in the maturation reaction instead of natural abundance (n.a.) Fe²⁺, using cell lysate and apo-HydA1 previously expressed with n.a. Fe-containing media. It seems likely that the ⁵⁷Fe²⁺ in the solution may reconstitute into or exchange with the somewhat labile dangler Fe site in the auxiliary cluster of HydG,⁵¹ leading to the formation of an ⁵⁷Fe-labeled Fe(CO)₂(CN)(Cys) product of HydG which is then incorporated in the subsequent steps of H-cluster biosynthesis.²⁹ We also used ¹⁵N-Tyr in the reaction simply to label the CN⁻ ligands as C¹⁵N⁻ in order to minimize any interference from the strong ¹⁴N features in HYSCORE spectra which could overlap with potential ⁵⁷Fe crosspeaks.

We first studied the H_ox state of this [2⁵⁷Fe]-CrHydA1 compared to a uniformly ([6⁵⁷Fe]) ⁵⁷Fe-labeled sample. Q-band Davies ENDOR spectra (Figure 3A, red trace) recorded at g1 (2.103) and g₃ (1.998) of the [2⁵⁷Fe] sample exhibit one major set of ⁵⁷Fe HFI with A = 16 MHz (Fe_A), which is consistent with the previously reported ENDOR spectra of [2⁵⁷Fe]_H-labeled CrHydA H_ox.²⁹ In contrast, the spectra lack the isotropic feature with A = 10.5 MHz (Fe_B) that has been previously assigned to the [4Fe-4S]_H as here seen in the ENDOR spectra (Figure 3A, blue trace) of the [6⁵⁷Fe]-CrHydA1 H_ox,²⁹ validating our selective labeling strategy. In parallel, we recorded the Q-band HYSCORE spectra of the [6⁵⁷Fe]- and [2⁵⁷Fe]- CrHydA1 H_ox samples, which disclosed two sets of ⁵⁷Fe peaks in the [6⁵⁷Fe] sample and one set of ⁵⁷Fe peaks in the [2⁵⁷Fe] sample, both in the second quadrant and absent in a parallel sample without ⁵⁷Fe labeling (Figure 3B, C; Figure S4). Spectral simulation indicates that the extra ⁵⁷Fe peak in the [6⁵⁷Fe] sample, which is quite isotopic, corresponds to the Fe_B feature ([4Fe-4S]_H) observed in the ENDOR spectra (Figure S4). However, the ⁵⁷Fe peak in the [2⁵⁷Fe] sample, Fe_C, does not correspond to the Fe_A feature observed via ENDOR, since the HFI of the latter is much larger. The relative intensity between these Fe_B and Fe_C feature in the [6⁵⁷Fe] vs. [2⁵⁷Fe] sample also rules out the remote possibility that Fe_C arises from a second ⁵⁷Fe coupling in the small amount of the labeled [4Fe-4S]_H in the [2⁵⁷Fe] sample. Based on these considerations, we assert that the Fe_C feature we observed in the Q-band HYSCORE spectra of the [2⁵⁷Fe]-CrHydA1 H_ox sample arises from the second Fe in the [2Fe]_H.

Figure 3.

(A) Comparison of the Q-band Davies ENDOR spectra of [2⁵⁷Fe]-CrHydA1 H_ox (red trace) and [6⁵⁷Fe]-CrHydA1 H_ox (blue trace) recorded at g = 2.103 and 1.998. (B) Comparison of the Q-band HYSCORE spectra of [6⁵⁷Fe]- (left panels), [2⁵⁷Fe]-(middle panels), and n.a. Fe (right panels) CrHydA1 H_ox samples recorded at g = 2.103 and 1.998 showing the ⁵⁷Fe peaks in the second quadrant. (C) Field-dependent Q-band ⁵⁷Fe HYSCORE spectra of the [2⁵⁷Fe]-CrHydA1 H_ox sample (red contours) and simulations of Fe_C (blue contours). Simulation: g = [2.103, 2.042, 1.998], A ⁵⁷Fe_C = [6.5, −1.4, 12.0] MHz, Euler angle = [17, 50, 25]° Conditions for Davies-ENDOR: frequency, 34.1 GHz; temperature, 15 K; RF pulse, 50 μs; inversion pulse, 80 ns; π/2,12 ns; τ = 300 ns. Condition for Q-band HYSCORE: frequency, 34.1 GHz; temperature, 15 K; π/2, 12 ns; τ = 368, 392, 388, 368 ns at g = 2.103, 2.075, 2.042, 1.998, respectively; time increment along both dimensions was 24 ns with 270 steps.

The field-dependent HYSCORE spectra can be well-simulated with a rhombic ⁵⁷Fe HFI tensor of A = [6.5, −1.4, 12.0] MHz (a_iso = 5.7 MHz and T = [0.8, −7.1, 6.3] MHz, Figure 3C, Table 2). The β of 50° used in the simulation is in good agreement with the Fe_p-Fe_d-CO_bridge angle (44°) from the X-ray structure of the H-cluster (PDB: 3C8Y³), which is consistent with A₃ being along the Fe_d-Fe_p vector if we assign g₃ as pointing along the Fe_d-CO_bridge bonding vector (vide infra) No corresponding ENDOR signal could be identified for this HFI tensor (Figure S5), probably due to its large rhombicity. Since selective labeling of a single specific Fe in the [2Fe]_H subcluster is not yet feasible, we cannot assign this ⁵⁷Fe HFI to a specific Fe. However, the relatively smaller a_iso suggests that this ⁵⁷Fe HFI may be assigned to Fe_p, with the stronger 16–18 MHz (a_iso ~ 17 MHz) coupling arising from Fe_d. In this case, the a_iso (Fe_d)/a_iso (Fe_p) ratio of 3.0 is more consistent with Fe_d carrying the majority spin density as inferred from ¹³C HFI results and theoretical predictions,^{21, 25, 31} although it is still approximately two-fold smaller than the a_iso (¹³CO_d)/a_iso (¹³CO_p) ratio (or that of the ¹³CN⁻ ligands). This difference may be attributed to the rhombicity of the A-tensor of Fe_p, which may affect the ¹³C HFI in a manner we have not fully accounted.

Table 2.

Summary of ⁵⁷Fe HFI of CrHydA1 H-cluster at different states

H-cluster state	A 57Fe (MHz)	aiso (MHz)	[α, β, γ] (°)	assignment	reference
Hox	FeA: 16–18	~17		Fed	29
	FeB: [10.4, 9.1, 10.2]	9.9	[0, 0, 0]	[4Fe-4S]H	29, this work
	Fec: [6.5, −1.5, 12.0]	5.7	[17, 50, 25]	Fep	this work
	[2.2, 5.5, 5.5]	4.4	[0, 0, 0]	Fep	53
Hox-CO	[−1.7, 2.8, 2.81	1.3	[0, 15, 0]	Fed	53
	see ref 51	27.6; 28.0; 30.0; 33.2		[4Fe-4S]H	53
Hhyd	<0.3			Fep and Fed	this work
	Fe1: [20.0, 29.5, 35.0]	28.2	[140, 35, 40]	[4Fe-4S]H	this work
	Fe2: [35.0, 31.5, 39.5]	35.3	[120, 35, 0]
	Fe3: [20.2, 21.0, 8.5]	16.6	[125, 35, 0]
	Fe4: [16.5, −0.5, 5.5]	7.2	[130, 35, 50]

More importantly, the fact that the two Fe sites of the [2Fe]_H are indeed electronically different is crucial for the catalytic activity of the [FeFe] hydrogenase. This Fe sites heterogeneity, together with the magnitudes of the hyperfine couplings, can be used as a guide for designing molecular models for [FeFe] hydrogenases. For instance, one such synthetic complex has been characterized, showing that the ⁵⁷Fe HFI on the two Fe sites are different by a factor of ~2 in H_ox models, similar to that found in the hydrogenase enzyme.⁵²

[⁵⁷Fe] in H_hyd

When the [2⁵⁷Fe]-CrHydA1 sample is poised in the H_hyd state, we expect to see small ⁵⁷Fe HFI as the ¹³CO and ¹³CN⁻ HFI on these Fe are very small (vide supra). Indeed, in the field-dependent Q-band HYSCORE spectra, we were only able to see intensity on the diagonal line at the Larmor frequency of ⁵⁷Fe, likely from the two overlapping ⁵⁷Fe HFI. The largest principal value of these ⁵⁷Fe HFI was estimated by the feature at g2 which showed a width of ~0.3 MHz. Notably, since there was a small amount of H_ox-CO in the H_hyd sample, HYSCORE spectra recorded at g = 2.008, g2 of H_ox-CO, showed two sets of crosspeaks in the first and the second quadrant (Figure S6), corresponding to the Fe_d and Fe_p in this [2⁵⁷Fe]-H_ox-CO, respectively. This HYSCORE patterns are identical to those of the previously reported CrHydA1 H_ox-CO sample maturated using the synthetic [⁵⁷Fe₂ (CO)₄(CN)₂(adt)]²⁻, and can be well-simulated with the reported ⁵⁷Fe HFI tensors.⁵³ This further proves that both Fe_d and Fe_p are labeled in the sample.

The very small ⁵⁷Fe HFI arising from the [2Fe]_H supports the diamagnetic low-spin Fe²⁺ assignment of these two Fe sites. The ⁵⁷Fe HFI of the remaining paramagnetic [4Fe-4S]_H⁺ was then measured using the uniformly ⁵⁷Fe-labeled H_hyd sample generated by reducing the [6⁵⁷Fe]-CrHydA1 with dithionite. The X-band CW EPR of the [6⁵⁷Fe]-CrHydA1 H_hyd, but not the [2⁵⁷Fe]-H_hyd, shows line broadening at g₃ (Figure S7), indicating the presence of large ⁵⁷Fe HFI from the [4Fe-4S]_H of H_hyd, as expected. The Q-band field-dependent Davies ENDOR spectra further reveal these large ⁵⁷Fe couplings (Fe₁ and Fe₂, Figure 5A), and are simulated with two sets of ⁵⁷Fe HFI, with a_iso of 28.2 MHz and 35.3 MHz, respectively (Table 2). These ⁵⁷Fe hyperfine values are similar to that of the mixed-valance Fe^2.5+ pair found in typical [4Fe-4S]⁺ clusters.⁴⁸ More interestingly, additional ⁵⁷Fe HFI are disclosed by Q-band HYSCORE spectroscopy, as indicated from the crosspeaks in the second quadrant of the HYSCORE spectra collected at g = 1.881, g₃ of H_hyd (Fe₃ and Fe₄, Figure 5B). Simulation of the field-dependent HYSCORE spectra (Figure 5B and S8) gives two ⁵⁷Fe HFI tensors, a rhombic one with a_iso of 7.5 MHz (Figure 5B and S8, blue contours) and an axial one with a_iso of 16.6 MHz (Figure 5B and S8, green contours), corresponding to the two remaining Fe²⁺ in the [4Fe-4S]_H (Table 2). The a_iso of 16.6 MHz is comparable to typical ⁵⁷Fe HFI of ferrous pair (Fe²⁺) found in the [4Fe-4S]⁺ clusters of various ferredoxins and synthetic systems, whereas the a_iso of 7.2 MHz is much smaller than these typical values.⁴⁸ It has been noted in aconitase that the unique substrate-binding Fe has the smallest ⁵⁷Fe HFI among the four Fe.^{48, 54} It is likely that in H_hyd, this small ⁵⁷Fe HFI also arises from the unique Fe²⁺ that is coordinated by the bridging Cys residue by which the [4Fe-4S]_H is linked to the [2Fe]_H. Together, these ⁵⁷Fe HFI results depict the electronic structure of [4Fe-4S]_H in H_hyd as having antiferromagnetically coupled Fe^{2 5+} pair and Fe²⁺ pair, leading to an S = 1/2 subcluster. However, the presence of the [2Fe]_H in close proximity has considerably affected the electronic structure of the [4Fe-4S]_H, most significantly so on the unique Fe, despite the former being formally diamagnetic.

Figure 5.

(A) Q-band field-dependent Davies ENDOR spectra of uniformly ⁵⁷Fe-labeled CrHydA1 (black traces) simulated using two ⁵⁷Fe A-tensors: g = [2.076, 1.936, 1.881]; A ⁵⁷Fe₁ = [20.0, 29.5, 35.0] MHz, Euler angle = [140, 35, 40]° (purple traces); A ⁵⁷Fe₂ = [35.0, 31.5, 39.5] MHz, Euler angle = [120, 35, 0]° (red traces). (B) Q-band HYSCORE spectrum of uniformly ⁵⁷Fe-labeled CrHydA1 recorded at g₃ = 1.881 (left) and simulation with two ⁵⁷Fe A-tensors (right): g = [2.076, 1.936, 1.881]; A ⁵⁷Fe₃ = [20.2, 21.0, 8.5] MHz, Euler angle =[120, 35, 0]° (green contours); A ⁵⁷Fe₄ = [16.5, −0.5, 5.5] MHz, Euler angle = [130, 35, 50]° (blue contours). See Figure S8 for HYSCORE spectra and simulations recorded at other g values. Conditions for Davies-ENDOR: frequency, 34.0 GHz; temperature, 10 K; RF pulse, 50 μs; inversion pulse, 80 ns; π/2, 12 ns; τ = 300 ns. Condition for Q-band HYSCORE: frequency, 34.0 GHz; temperature, 10 K; π/2, 12 ns; τ = 300 ns; time increment along both dimensions was 24 ns with 270 steps.

The ¹³C and ⁵⁷Fe HFI results of H_hyd validate the electronic structural description proposed from Mössbauer spectroscopy that the [2Fe]_H contains two diamagnetic low-spin Fe²⁺.¹⁶ The small ¹³C and ⁵⁷Fe HFI in the [2Fe]_H most likely arise from spin delocalization via covalency and spin polarization, and the exchange interaction to the [4Fe-4S]_H is almost nil because of the much higher energy gap between the S = 0 and S = 1 states for low-spin Fe²⁺. It is therefore safe to write H_hyd as [4Fe-4S]_H⁺-[Fe^IIFe^II]-H⁻. The diamagnetic nature of the [2Fe]_H is also consistent with the recently reported ¹H NMR spectra of H_hyd, in which the terminal hydride has a very small, if any, paramagnetic shift.¹³ As noted above, the small ¹³C and ⁵⁷Fe HFI in H_hyd are reminiscent of that in the HydG reaction intermediate we recently characterized—the [4Fe-4S]⁺ -[(Cys)Fe^II(CO)(CN)] intermediate termed “Complex A” also harbors a CO and CN⁻ bound low-spin Fe²⁺.³² But compared to Complex A, the ¹³C and ⁵⁷Fe HFI are even smaller in H_hyd, indicating a smaller spin density on Fe_p and Fe_d in H_hyd. This difference could be due to a variety of factors. The nearest iron of the [4Fe-4S] could be closer to the dangler Fe in Complex A; it could have a larger spin projection value; or the Fe-S_Cys-Fe bond angle could be more amenable for spin delocalization via covalency.

If we view H_hyd as resulting from the binding of an H⁻ onto the Fe_d of H_ox, this is accompanied by the movement of one electron from the [2Fe]_H to the [4Fe-4S]_H. This is exactly opposite to the cases of H_red/H_redH⁺—recently identified to be [4Fe-4S]_H⁺-[Fe^IFe^II]_H and [4Fe-4S]_H²⁺-[Fe^IFe^I]_H-(NH₂⁺) respectively—where the protonation on the adt amine is accompanied by one electron moving from the [4Fe-4S]_H to the [2Fe]_H.⁸ It appears that the net charge on the [2Fe]_H tends to remain unchanged, as implied in previous work as the “neutralization effect”.⁸ The [4Fe-4S]_H—in addition to its roles in mediating electron transfer in catalysis—could serve as an electron sink or source to facilitate electronic rearrangement during the catalytic cycle in order to maintain the overall structural integrity of the H-cluster. In this regard, further protonation of H_hyd on the adt amine would lead to a H_hydH⁺ state that is proposed to adopt a [4Fe-4S]_H²⁺-[Fe^IIFe^I]_H-(NH₂⁺)-H⁻ configuration and to release H₂ rapidly before returning to H_ox.¹⁸ Spectroscopic characterization on this transient species would be of great interest if conditions favoring its accumulation can be found since in H_hydH⁺, the spin density is purported to be localized on the [2Fe]_H again, and one would expect to see the much larger ¹H HFI of the terminal hydride (or H₂ molecule) bound to Fe_d directly by EPR/ENDOR spectroscopy.

Conclusion

The in vitro biosynthetic route to the H-cluster has demonstrated its convenience and versatility to site-specifically label the H-cluster with desired magnetic nuclei. This biochemical approach, combined with advanced EPR spectroscopic techniques, allow us to study the electronic structure of the H-cluster at two catalytic states, H_ox and H_hyd. For H_ox, we measured the ¹³C HFI of the three ¹³CO ligands, which is a natural outgrowth of our previous work on the two ¹³CN⁻ ligands. We re-investigated the ⁵⁷Fe HFI of the [2Fe]_H, and identified for the first time a second set of ⁵⁷Fe HFI. This work shows that the two Fe in the [2Fe]_H have different HFI, which agrees with the results from the spectroscopic study of the ¹³CO/¹³CN ligands on these Fe sites, and largely solves a controversy from previous spectroscopic and theoretical studies. For H_hyd, we detected very small ¹³C and ⁵⁷Fe HFI in the [2Fe]_H subcluster, and four much larger ⁵⁷Fe HFI in the [4Fe-4S]_H subcluster, consistent with the proposed electronic structural description of H_hyd as [4Fe-4S]_H⁺[Fe^IIFe^II]-H⁻. Overall, our work represents a comprehensive spectroscopic investigation on the H-cluster. These results, together with the numerous recent advances with the [FeFe] hydrogenases, helps enhance our understanding on this unique cofactor, and guide de novo design of molecular catalysts for hydrogen evolution.

Figure 4.

Q-band HYSCORE spectra of [2⁵⁷Fe]-CrHydA1 H_hyd recorded at three principal g values. Conditions: temperature, 10 K; τ = 300 ns; and other settings as indicated in Figure 3.

Synopsis:

The electronic structures of the two catalytic states of the [FeFe] hydrogenase H-cluster, H_ox and H_hyd, are studied by probing the hyperfine couplings of the ¹³C and ⁵⁷Fe nuclei in site-specific isomiddlee-labeled H-cluster. For H_ox, two different ⁵⁷Fe hyperfine interactions are observed for the two Fe sites of the spin-carrying [2Fe]_H subcluster. For H_hyd, the weakly coupled ¹³C and ⁵⁷Fe are detected, consistent with previous description of this newly identified form of H-cluster.