Spectroscopic Investigations of [FeFe] Hydrogenase Maturated with [⁵⁷Fe₂(adt)(CN)₂(CO)₄]^2–

Abstract

The preparation and spectroscopic characterization of a CO-inhibited [FeFe] hydrogenase with a selectively ⁵⁷Fe-labeled binuclear subsite is described. The precursor [⁵⁷Fe₂(adt)(CN)₂(CO)₄]^2– was synthesized from the ⁵⁷Fe metal, S₈, CO, (NEt₄)CN, NH₄Cl, and CH₂O. (Et₄N)₂[⁵⁷Fe₂(adt)(CN)₂(CO)₄] was then used for the maturation of the [FeFe] hydrogenase HydA1 from Chlamydomonas reinhardtii, to yield the enzyme selectively labeled at the [2Fe]_H subcluster. Complementary ⁵⁷Fe enrichment of the [4Fe-4S]_H cluster was realized by reconstitution with ⁵⁷FeCl₃ and Na₂S. The H_ox-CO state of [2⁵⁷Fe]_H and [4⁵⁷Fe-4S]_H HydA1 was characterized by Mössbauer, HYSCORE, ENDOR, and nuclear resonance vibrational spectroscopy.

Graphical abstract

INTRODUCTION

Since their crystallographic identification, the hydrogenases (H₂ases) represent perfect paradigms for bioinspired energy transformations. Utilizing only base metals and operating at near thermodynamic potentials, these enzymes promote reactions at extraordinary rates, in the case of [FeFe] H₂ases thousands of turnover per second.¹ Further motivating this theme are the practical implications of the H₂/H⁺ redox reaction, which is the basis for fuel cells.

Many [FeFe] H₂ases have been studied, but the algal H₂ase from Chlamydomonas reinhardtii, HydA1,² has received particular attention.³ The C. reinhardtii protein is ideal for studies on biosynthesis⁴ and spectroscopy because it represents a “minimal architecture” enzyme containing only the active site.⁵ In contrast, most [FeFe] H₂ases feature multiple accessory [Fe–S] clusters. The active site of all [FeFe] H₂ases is the “H-cluster”, consisting of a [4Fe-4S]_H cluster appended via a bridging cysteinyl thiolate to a diiron subunit called [2Fe]_H.⁶ In [2Fe]_H, a diiron center is bound to CO, CN⁻, and azadithiolate (adt^2– = [(SCH₂)₂NH]^2–) ligands (Figure 1).⁷

Figure 1.

H-cluster of HydA1 in the H_ox-CO state.

Recent work has shown that functional HydA1 enzyme can be produced through incubation of unmaturated HydA1 containing only the [4Fe-4S]_H subcluster, with a synthetic active site mimic (Et₄N)₂[⁵⁷Fe₂(adt)(CN)₂(CO)₄].⁸ This innovation exploits the fact that the active site is attached to the protein through few covalent bonds. Via this artificial maturation, the enzyme is now available with wide variety of chemically and isotopically labeled versions^9,10 of the [2Fe]_H subunit on a scale and at a pace that would not be readily achieved by in vitro maturation routes.^11,12 This “artificial maturation” allows a detailed characterization of individual active site states and the catalytic mechanism through a variety of spectroscopic techniques.⁹

The artificial maturation route in principle should allow the selective labeling of the [2Fe]_H subunit with ⁵⁷Fe, a nucleus highly responsive to Mössbauer and nuclear resonance vibrational spectroscopies (NRVS). With a nuclear spin I = 1/2, ⁵⁷Fe is also ideal for the suite of EPR techniques that provide exquisite insights into Fe-based enzymes.^{13 57}Fe-labeling of the [2Fe]_H subunit however poses a significant synthetic challenge because salts of [Fe₂(adt)(CN)₂(CO)₄]^2– are prepared by multistep sequences starting from reagents that would only be awkwardly and inefficiently labeled with ⁵⁷Fe. In this report these challenges are surmounted, as established by the preparation of HydA1 with a selectively ⁵⁷Fe-labeled [2Fe]_H site. The H_ox-CO state of HydA1 (Figure 1) can be obtained as a pure state and was therefore chosen to demonstrate the selective labeling. Using (Et₄N)₂[⁵⁷Fe₂(adt)(CN)₂(CO)₄] as precursor the [2Fe]_H subsite in H_ox-CO was labeled using artificial maturation. In a complementary experiment the [4Fe-4S] subcluster in H_ox-CO was labeled using FeS reconstitution. The two ⁵⁷Fe labeled versions of H_ox-CO were studied using Mössbauer, electron nuclear double resonance (ENDOR), hyperfine sublevel correlation (HYSCORE) as well as nuclear resonance vibrational (NRVS) spectroscopy.

RESULTS AND DISCUSSION

Synthesis and Characterization of [⁵⁷Fe₂(adt)-(CN)₂(CO)₄]^2–

The precursor to the target [⁵⁷Fe₂(adt)-(CN)₂(CO)₄]^2– is ⁵⁷Fe₂(adt)(CO)⁶, which undergoes dicyanation nearly quantitatively.¹⁴ Synthesis of the diiron hexacarbonyl, however, poses challenges because it is derived via a series of inefficient reactions from precursors that are not readily labeled with ⁵⁷Fe. Low yielding routes to unlabeled Fe₂(adt)(CO)₆ are tolerated¹⁴ because the relevant reagents, e.g., Fe(CO)₅, are inexpensive and the early steps in the preparation can be conducted on a multigram scale. The industrial method for production of Fe(CO)₅ involves the direct carbonylation of Fe metal at high temperatures and pressures, e.g., 175 atm at 150 °C.¹⁵ Such reactions require specialized autoclaves,¹⁶ which are not suited for producing small amounts of ⁵⁷Fe(CO)₅. A variety of laboratory syntheses of ⁵⁷Fe(CO)₅ have been described, but they suffer from low yields and difficult separations even when using specialized equipment.¹⁷

The above considerations led to a focus on routes that avoid the intermediacy of Fe(CO)₅. Retrosynthetic analysis reminds one that Fe₂S₂(CO)₆, the immediate precursor to Fe₂(adt)-(CO)₆, is formed from the [HFe(CO)₄]⁻ anion, not the pentacarbonyl. Thus, syntheses of [H⁵⁷Fe(CO)₄]⁻ from ⁵⁷FeX₂ are of interest. Literature methods¹⁸ for generating [HFe-(CO)₄]⁻ from iron halides proved low-yielding in our hands. Relevant to possible routes to [H₅₇Fe(CO)₄]⁻ is the fact that it is easily derived from [Fe(CO)₄]^2– by protonation. The anion [Fe(anthracene)₂]⁻, prepared by Ellis and co-workers in 61% yield from FeBr₂, carbonylates at ambient pressures.¹⁹ The product, obtained in 81% isolated yield, is [K(18-crown-6)]₂[Fe₂(CO)₈]. Unfortunately, attempts to convert this salt into Fe₂S₂(CO)₆ were unfruitful. Treatment of [K(18-crown-6)]₂[Fe₂(CO)₈] with S₂Cl₂ or S₈ gave complex mixtures including [Fe₂S₂(S₅)₂]^2– and intractable solids but no Fe₂S₂(CO)⁶. A successful method for the direct synthesis of Fe₂S₂(CO)₆ was inspired by details in the PhD thesis of W. W. Brennessel of the Ellis group, who describes the synthesis of K₂Fe(CO)₄ from FeBr₂ in ~50% yield.²⁰ His method involves treatment of FeBr₂ in THF with four equivalents of potassium anthracene at low temperature, followed by carbonylation at 1 atm. The Fe(-II) derivative is proposed to form via reduction of K₂Fe₂(CO)₈ by the fourth equivalent of K(anthracene). This reaction was reproduced. Treatment of the resulting K₂Fe-(CO)₄ with methanol efficiently afforded KHFe(CO)₄,²¹ which reacted with elemental sulfur to give Fe₂S₂(CO)₆ after standard workup.²¹ In this way, starting from 500 mg of ⁵⁷Fe, we prepared 180 mg of ⁵⁷Fe₂S₂(CO)₆ (11.8% from ⁵⁷Fe metal) together with 60.8 mg (4.5% yield) of ⁵⁷Fe₃S₂(CO)₉ (Scheme 1). Repeat synthesis using the same procedure resulted in the yield of ⁵⁷Fe₂S₂(CO)₆ improving to 371 mg (24.4% from ⁵⁷Fe metal). The ¹³C NMR spectra of ⁵⁷Fe₂S₂(CO)₆ and ⁵⁷Fe₃S₂(CO)₉ gives ¹J(⁵⁷Fe,¹³C) = 28.3 and 26.3 Hz, respectively. The most problematic step in the synthesis is the conversion of KHFe(CO)₄ into Fe₂S₂(CO)₆, which we estimate proceeded in 30% yield.

Scheme 1. Synthesis of [⁵⁷Fe₂(adt)(CN)₂(CO)₄]₂⁻a.

^a(i) 4 KC₁₄H₁₀, (ii) CO, (iii) MeOH, (iv) S₈, (v) H⁺, (vi) CH₂O, NH₄⁺, (vii) (NEt₄)CN.

Conversion of ⁵⁷Fe₂S₂(CO)₆ to [⁵⁷Fe₂(adt)(CN)₂(CO)₄]^2– requires two steps, which have been described previously for the unlabeled versions of these complexes.²² The first of these steps is the conversion of ⁵⁷Fe₂S₂(CO)₆ to ⁵⁷Fe₂(adt)(CO)₆. This conversion was achieved by reduction of ⁵⁷Fe₂S₂(CO)₆ with LiBEt₃H to give ⁵⁷Fe₂(SLi)₂(CO)₆, followed by protonation to ⁵⁷Fe₂(SH)₂(CO)₆ and subsequent adt bridge formation through reaction with a mixture of ammonia and formaldehyde to yield ⁵⁷Fe₂(adt)(CO)₆. Conversion of ⁵⁷Fe₂(adt)(CO)₆ to [⁵⁷Fe₂(adt)(CN)₂(CO)₄]^2– was achieved by reaction with 2 equiv of (NEt₄)CN. The isotopic purity of the anion [⁵⁷Fe₂(adt)(CN)₂(CO)₄]^2– was established by negative ion ESI mass spectrometry (see Supporting Information (SI), Figure S1).

Selective Labeling of HydA1

The insertion of [⁵⁷Fe₂(adt)(CN)₂(CO)₄]^2– into unmaturated HydA1 was monitored by Fourier transform infrared spectroscopy (FT-IR). The precursor undergoes substantial rearrangement upon incorporation into unmaturated HydA1 (Figure S2A and S2B, SI). One CO ligand migrates to the bridging position and the CN⁻ groups shift from apical to basal sites. Furthermore, in a major desymmetrizing event, one of the two previously equivalent Fe sites binds to cysteinyl sulfur, which links the [2Fe]_H and [4Fe-4S]_H clusters. Under the usual reductive maturation conditions one terminal CO ligand of the distal Fe dissociates affording an open coordination site where substrate and inhibitors can bind. After oxidation with thionin (Figure S2C, SI) and flushing with CO, a pure H_ox-CO state is created (Figure S2D, SI). In the H_ox-CO state, the [2Fe]_H subunit is in a delocalized mixed valence Fe(II)Fe(I) configuration.

In order to realize complementary ⁵⁷Fe enrichment, in a separate experiment the [4Fe-4S]_H cluster was ⁵⁷Fe-labeled as described in the Experimental Section. In a first step, unfolding and thereby loss of the unlabeled [4Fe-4S]_H cluster was observed by subsequent loss of the brownish color of the solution. The [4Fe-4S]_H cluster was reconstituted with ⁵⁷FeCl₃ and Na₂S, accompanied by recovery of the brownish color. Successful reconstitution was confirmed by activity measurements, UV–vis and EPR spectroscopy (see SI for detailed information).

EPR Characterization

The selective labeling of the two components of the H-cluster with ⁵⁷Fe greatly simplifies the assignment of the observed pulse EPR signals (see Figure 2). For the hybrid in which the [2Fe]_H subcluster is labeled, two spectral features in the HYSCORE spectra can be observed (Figure 2B,C). The correlation ridges in the right-hand section of the spectra (positive frequencies) are assigned to a weakly coupled ⁵⁷Fe nucleus while the ones in the left-hand section (negative frequencies) originate from a strongly coupled ⁵⁷Fe nucleus. These signals can both be simulated with axial hyperfine tensors (see Table S1 in the SI). Their A_iso values are 4.4 and 1.3 MHz, respectively, suggesting that one iron in the [2Fe]_H subsite has ~4 times more spin density than the other. According to the spin-exchange model describing the electronic structure of H_ox-CO (see below), the proximal iron in [2Fe]_H has the largest spin density. The obtained hyperfine parameters are in agreement with previously reported EPR data on the [FeFe] H₂ases from Desulfovibrio desulfuricans (DdH).¹³ Earlier Mössbauer studies on the [FeFe] H₂ases from Desulfovibriovulgaris Hildenborough (DvH) and Clostridium pasteurianum (CpI) provided the first estimates for the hyperfine ⁵⁷Fe interactions in the [2Fe]_H subsite.^23,25 These measurements, however, suffered from overlapping signals from the accessory [4Fe-4S] clusters and are therefore less accurate (see Table S1). The complete hyperfine tensor information on Fe1 and Fe2 as obtained from HYSCORE (Table S1) accurately defines the electronic structure of the binuclear subsite and can be used to validate quantum chemical calculations on the H-cluster properties.

Figure 2.

Q-band ⁵⁷Fe ENDOR and HYSCORE spectra (20 K) of selectively labeled HydA1 H_ox-CO recorded at field positions indicated in the EPR spectrum (inset); for experimental parameters see SI. (A) Davies ENDOR of H_ox-CO ⁵⁷Fe labeled at [4Fe-4S]_H (black) or [2Fe]_H (blue); simulations are indicated in red. (B) and (C) HYSCORE spectra recorded for H_ox-CO ⁵⁷Fe labeled at [2Fe]_H at g = 2.008 and g = 2.053 respectively; for all simulation parameters see Table S1.

As described above, it is also possible to selectively label the [4Fe-4S]_H subcluster with ⁵⁷Fe. For this sample in the H_ox-CO state, no signal in the HYSCORE spectra was observed. In the ENDOR spectra between 10–20 MHz signals from strongly coupled Fe nuclei are observed. As it is shown in Figure 2A, these signals are not observed when only the [2Fe]_H subcluster is labeled. This unequivocally confirms that these signals originate from the [4Fe-4S]_H subcluster. As presented in Figure 3, these signals can be simulated with four distinct anisotropic hyperfine interactions corresponding to the four Fe centers in the [4Fe-4S]_H subcluster (hyperfine interaction parameters are listed in Table S1). In earlier Mössbauer studies on CpI, these relatively large hyperfine interactions of the formally diamagnetic [4Fe-4S]_H subcluster were interpreted as originating from spin density induced by a substantial exchange coupling between the two subclusters.²³ The four hyperfine interactions were predicted to be divided in two groups representing the two S = 9/2 Fe(II)Fe(III) pairs which are mutually antiferromagnetically coupled to effective spin S = 0. Later ENDOR studies on DdH confirmed this analysis and showed that the hyperfine interaction of the two groups have opposite signs.¹³ The current data on HydA1 are fully in line with these studies. The two groups are defined by pair 1 with A_iso = 33.2 MHz (brown) and 30.0 MHz (orange), pair 2 A_iso = 28.0 MHz (blue) and 27.6 MHz (green).

Figure 3.

Q-band Davies ENDOR spectra and simulations of H_ox-CO HydA1 selectively labeled with ⁵⁷Fe at the [4Fe-4S]_H cluster. Spectra were recorded with an RF pulse of 45 µs, shot repetition time 800 µs, microwave frequency 33.93396 GHz, temperature 20 K at field positions: (A) 1182.9 mT ≈ g₁, (B) 1189.65 mT, (C) 1196.4 mT, (D) 1203.15 mT, (E) 1209.9 mT ≈ g₂. The black line represents experimental data and the red line the sum of the simulations. The colored lines below each experimental spectrum are the components of the simulation corresponding to the four hyperfine couplings: brown Fe³, orange Fe⁴, blue Fe⁵, green Fe⁶ (see Table S1).

Mössbauer Characterization

The selective ⁵⁷Fe-labeling of the [2Fe]_H site enables the investigation of this specific subsite by Mössbauer spectroscopy. Previous Mössbauer studies on the active [FeFe] H₂ases CpI^23,24 and DvH²⁵ are complicated by the presence of accessory [4Fe-4S] clusters in addition to the H-cluster. Using our approach of selective labeling, the Mössbauer spectra are simplified and Mössbauer parameters can be obtained without the cosimulation of other ⁵⁷Fe-components. Additionally, this method enables high concentrations of the sample and therefore high-quality spectra. The Mössbauer spectrum of the [4⁵⁷Fe-4S]_H cluster from HydA1 maturated with unlabeled [Fe₂(adt)(CN)₂(CO)₄]^2– in the H_ox-CO state is shown in Figure 4A. The counterpart spectrum, the ⁵⁷Fe-labeled [2Fe]_H site in the H_ox-CO state, is presented in Figure 4B. Both spectra are quite simple and can be simulated with relatively few components. Figure 4C displays a superimposition of the simulated spectra 4A and 4B. It clearly shows the overlap between the signals from the [4Fe-4S]_H cluster and the [2Fe]_H site, which would require 9 parameters for its simulation (apart from the background contributions).

Figure 4.

(A) Mössbauer spectrum and simulations of H_ox-CO HydA1 selectively labeled with ⁵⁷Fe at the [4Fe-4S]_H subcluster. (B) Mössbauer spectrum and simulations of H_ox-CO HydA1 selectively labeled with ⁵⁷Fe at the [2Fe]_H subcluster. (C) Superimposition of the simulations of the ⁵⁷Fe-labeled [2Fe]_H subunit and the [4Fe-4S]_H cluster, ratio 0.5:1. All spectra were measured at 160 K in zero magnetic field.

The isomer shift (δ = 0.42 mm/s) and the quadrupole splitting (ΔE_Q = 1.04 mm/s) found for the [4⁵⁷Fe-4S]_H cluster are typical for oxidized [4Fe-4S] clusters.²⁶ The spectrum of the ⁵⁷Fe-labeled [2Fe]_H site can be analyzed assuming two nonequivalent Fe sites. A small amount of an Fe(II) impurity is apparent as well. The data can be fitted assuming a slight nonequivalence either more pronounced in the quadrupole splitting (simulation 1, Figure 4B) or in the isomer shift (simulation 2, Figure S5). Simulation 1 (Figure 4B) is slightly favored. Taking into account the different ligands coordinating the two irons one would expect clear differences in quadrupole splitting whereas the isomer shift is relatively insensitive to electron density differences for low valence iron centers.²⁷ Simulation 1 is characterized by the isomer shifts δ1(1) = 0.16 mm/s, δ1(2) = 0.08 mm/s and quadrupole splittings of ΔE_Q1(1) = 0.89 mm/s, ΔE_Q1(2) = 0.55 mm/s. The parameters of all simulations can be found in Table S2 in the SI. The fitted Mössbauer parameters are not identical to those obtained previously by Pereira et al. for DvH (δ1(1) = 0.17 mm/s, δ1(2) = 0.13 mm/s and ΔE_Q1(1) = 0.70 mm/s, ΔE_Q1(2) = 0.65 mm/s, see also Table S2), but lie in the same range that is consistent with low-spin iron in a low oxidation state. The slight differences between the parameters obtained in this work and those by Pereira et al. could originate from the lower accuracy in the latter study caused by the overlapping F-cluster signals but it is also possible that the H-cluster of [FeFe] H₂ases in different organisms (DvH vs HydA1) slightly different electronic structures. To verify this, DvH could be maturated using the method described here to obtain a Mössbauer spectrum from the selectively labeled [2Fe]_H site. This would also pave the way to study in detail the elusive “inactive” oxidation states in DvH H_ox-air and H_trans as well as the “regular” active oxidation states H_ox and H_red.

NRVS Characterization

The [2⁵⁷Fe]_H-HydA1 H_ox-CO sample was also analyzed by nuclear resonance vibrational spectroscopy (NRVS). This technique is exclusively responsive to vibrations affecting the ⁵⁷Fe site. Previous attempts to record NRVS spectra of [FeFe] H₂ase (CpI)¹¹ were hampered by signals from other ⁵⁷Fe labeled components, and not purely originating from the [2Fe]_H subcluster (see Figure 6). In this study, NRVS spectra of (Et₄N)₂[⁵⁷Fe₂(adt)(CN)₂(CO)₄] and [2⁵⁷Fe]_H HydA1 H_ox-CO were obtained (Figure 5). The NRVS spectrum of the precursor displays intense bands in the regions assigned to Fe-CO stretching and Fe–C–O bending (six bands between 490 and 650 cm⁻¹) and the Fe-CN region (two bands between 400 and 500 cm⁻¹).^28,29 Comparing the spectrum of the precursor to that for the HydA1 H_ox-CO spectrum, one can clearly see that the Fe–C bands are strongly affected, with Fe- CO modes being (on average) red-shifted and Fe-CN modes being blue-shifted. The Fe-CO/-CN shift is consistent with major changes in the primary coordination sphere of the Fe centers upon insertion into the enzyme and conversion to H_ox-CO.

Figure 6.

NRVS spectrum of HydA1 H_ox-CO of this work (red) and CpI from previous NRVS on hydrogenase (mixed states, blue).¹¹ The assignments are not pure vibrational modes.

Figure 5.

NRVS spectrum of [2⁵⁷Fe]_H HydA1 H_ox-CO (red) and the precursor [⁵⁷Fe₂(adt)(CN)₂(CO)₄]^2– (blue). The peaks labeled in black are the same energy ±2 cm⁻¹ in both spectra.

The notable band at 653 cm⁻¹ is assigned as having significant contributions from a diequatorial Fe(CO)₂ stretch–bend by comparison with assignments of previous model complexes.^29–31 This band is absent in the [2⁵⁷Fe]_H HydA1 H_ox-CO spectrum as there are no longer two CO ligands in the basal positions of either iron center.³¹ The high intensity of the Fe-CO band at 576 cm⁻¹ in the [⁵⁷Fe₂(adt)(CN)₂(CO)₄]²⁻ spectrum is due to the high symmetry of this complex.

When comparing the low energy portion of the two spectra, it is clear that the bands at 170, 198, and 326 cm⁻¹ are largely unmoved and there are only minor shifts for other bands in the Fe–S region. This is an indication that there is little change in the Fe–S geometry upon incorporation of the precursor into the enzyme. A common feature of the NRVS of [4Fe-4S] clusters are bands due to Fe–S bending and breathing at 150 cm⁻¹.³⁰ The absence of any strong band at 150 cm⁻¹ demonstrates that there is no significant ⁵⁷Fe incorporation into the [4Fe-4S]_H cluster, verifying the results from the Mössbauer spectroscopy. Further discussion of bands and their assignments can be found in the Supporting Information.

The benefits of the selective labeling by the artificial maturation process are illustrated also by the NRVS measurements. The previously reported NRVS spectrum was recorded on CpI, which was only partially labeled by incubation/activation of the ⁵⁶Fe-labeled inactive apo protein with ⁵⁷Fe-enriched maturation proteins (Figure 6).¹¹ The CpI spectrum has characteristic features at 353 and 382 cm⁻¹ in the Fe–S region (300–400 cm⁻¹) assigned to Fe–S bands of oxidized [4Fe-4S]²⁺ clusters. These bands are indicative of unselective labeling of a [4Fe-4S] cluster. For the [2⁵⁷Fe]_H HydA1 H_ox-CO spectrum, only low intensity features are present in this region. Selective labeling is also evident in a notable increase in the intensity of the Fe-CN and Fe-CO signals in the 400–600 cm⁻¹ region.

CONCLUSIONS

In summary, the work demonstrates the selective ⁵⁷Fe-labeling of the HydA1 enzyme at both the [2Fe]_H subunit and the subcluster [4Fe-4S]_H. The route to ⁵⁷Fe-labeled iron carbonyls is described using only conventional Schlenk techniques. The method is clearly adaptable to labeling of other derivatives, perhaps other hydrogenases, and their synthetic analogues.³² Focusing on the protein prepared in the H_ox-CO state, all spectroscopic measurements confirm the selectivity of the labeling.

The ⁵⁷Fe hyperfine interactions from the [2Fe]_H and [4Fe-4S]_H subclusters in HydA1 obtained from HYSCORE and ENDOR spectroscopies are consistent with those obtained earlier from the H-cluster of the H₂ase DdH although slight species dependent variations were observed. The overall electronic structures of H_ox-CO of DdH and HydA1 seem therefore virtually identical. The Mössbauer parameters from the [2Fe]_H subsite in HydA1 revealed an intrinsic ambiguity (i.e., two equivalent combinations of isomer shift and quadrupole splitting) that was not detected in earlier studies on DvH.

NRVS analysis demonstrated that through this selective ⁵⁷Fe labeling the intensity of signals from the key [2Fe]_H subunit can be improved significantly. This improvement opens up the possibility of new modes involving the bridging CO, hydrides, or Fe–Fe interactions being identified. This future effort would require suitable isotopic labeling experiments of other atoms of the [2Fe]_H subunit and complementary simulations. These labeling experiments are facilitated by the frugal use of common reagents in this synthetic method, which allows easy substitution for their isotopically labeled counterparts ³⁴S₈, ¹³CO, and ¹³CN–.³³ Such isotope labeling of atoms bound directly to ⁵⁷Fe will facilitate assignment of the NRVS spectrum of catalytically significant states.

Further work is underway to spectroscopically characterize the “active” states of the enzyme, specifically H_ox, H_red, and H_sred. Through selective labeling the overlapping ⁵⁷Fe ENDOR spectra of the two subclusters in H_ox^13,34 can now be disentangled. The effects of, e.g., the bridging CO ligand and the exchange coupling to the [4Fe-4S]_H subcluster can now for the first time be studied on selectively labeled H-clusters. In addition, also “inactive” precursors [⁵⁷Fe₂(pdt)(CN)₂(CO)₄]^2– (pdt_2– = propanedithiolate = [(SCH₂)₂CH₂]^2–) and [⁵⁷Fe₂(odt)(CN)₂(CO)₄]^2– (odt^2– = oxadithiolate = [(SCH₂)₂O]^2–) will be studied in ⁵⁷Fe labeled form as a product of this work. Although largely inactive, these variants show “trapped” intermediates that may be related to the catalytic cycle.³⁵

EXPERIMENTAL SECTION

General Considerations

Unless otherwise indicated, reactions were conducted using standard Schlenk techniques or in a glovebox under an N₂ atmosphere at room temperature with stirring. Synthesis of ⁵⁷FeBr₂ from ⁵⁷Fe,^{36 57}Fe₂(adt)(CO)₆ from ⁵⁷Fe₂S₂(CO)₆,³⁷ and (Et₄N)₂[⁵⁷Fe₂(adt)(CN)₂(CO)₄] from ⁵⁷Fe₂(adt)(CO)₆,¹⁴ were achieved by modification of literature procedures for the equivalent unlabeled complexes. Details can be found in the Supporting Information.

Synthesis of ⁵⁷Fe₂S₂(CO)₆

A potassium anthracene solution was prepared by stirring thinly sliced potassium metal (1.41 g, 36.2 mmol) and anthracene (6.44 g, 36.1 mmol) in THF (120 mL) overnight using a glass covered stir bar to give a deep blue solution of potassium anthracene. In a separate flask ⁵⁷FeBr₂ (1.82 g, 8.40 mmol) was also stirred overnight in THF (100 mL) giving a suspension of yellow solid in a pale orange solution. Both solutions were chilled to −77 °C, and the ⁵⁷FeBr₂ solution was cannula transferred into the potassium anthracene solution over a period of 10 min. The solution was allowed to warm to room temperature over a period of 5 h, resulting in a color change from blue to very dark red/brown indicating the formation of K[Fe(anthracene)₂]. The solution was then filtered through Celite (the solution contains very fine particles that easily block glass frits and even block up Celite if too much pressure is applied). The volume of the filtrate was reduced to ~150 mL under a vacuum. The solution was cooled to −77 °C, and the headspace was replaced by an atmosphere of CO. The solution was then allowed to warm to room temperature over a period of 10 h under a constant 1.05 atm of CO. Once warmed to room temperature the solution is a dark color with a significant amount of pale precipitate. All solvent was removed to give a brown solid containing the desired K₂Fe(CO)₄. This solid was cooled to −77 °C, and treated with MeOH (50 mL) added over the course of 20 min. The initially brown/green solution changed to dark red as the mixture was warmed to room temperature over the course of an hour, indicating the formation of KHFe(CO)₄. The solution was filtered through Celite, and any remaining solid was washed with MeOH. The volume of the solution was reduced to ~10 mL, with the aim of keeping as much of the red color in solution, not coprecipitating with the colorless solids, which precipitate upon this volume reduction. The solution was cooled to 0 °C, and solid potassium hydroxide (1.30 g, 23.2 mmol) was added against a positive pressure of argon, followed by the addition of degassed water (15 mL). After 5 min of stirring at 0 °C, elemental sulfur (1.29 g, 40.2 mmol) was added directly against a flow of argon. The mixture was stirred at 0 °C for 1.5 h before degassed water (20 mL) and hexane (50 mL) were added. Solid NH₄Cl (3.22 g, 60.2 mmol) was added against a flow of argon, and a needle with attached bubbler was attached to the flask and the solution stirred for 16 h. Subsequent steps were performed in air. Stirring was ceased and the red hexane layer was decanted and reserved. The remaining aqueous solution was extract with hexane (4 × 20 mL) until no red color remained in the hexane layer, with the extracts then added to the previously reserved solution. The solution was passed through a short silica plug (2 cm) and the filtrate then evaporated on a rotary evaporator (Fe₂S₂(CO)₆ readily sublimes under oil pump vacuum at room temperature). An extract of the resulting red oil in hexane (3 mL) was chromatographed on a silica gel column (2 cm × 30 cm), eluting with hexane. The first orange band was identified as ⁵⁷Fe₂S₂(CO)₆ (180 mg, 0.520 mmol, 12.4% from ⁵⁷FeBr₂) and the second dark red band as ⁵⁷Fe₃S₂(CO)₉ (60.8 mg, 4.5% from ⁵⁷FeBr₂). The preparation was repeated with improved handling giving an improved yield of ⁵⁷Fe₂S₂(CO)₆ (371 mg, 1.07 mmol, 25.4% from ⁵⁷FeBr₂) and a small amount of ⁵⁷Fe₃S₂(CO)₉ (15 mg, 0.031 mmol, 1.1% from ⁵⁷FeBr₂) ⁵⁷Fe₂S₂(CO)₆: IR (pentane): ν_C≡O = 2085 (m), 2045 (s), 2008 (s), 1993 (w); ¹³C NMR (600 MHz, d₈-Toluene): δ 208.66 (d, J_C–Fe = 28.3 Hz, CO); ⁵⁷Fe₃S₂(CO)₉: IR (pentane): ν_C≡O = 2063 (s), 2045 (s), 2024(s), 2008 (m), 1988 (w); ¹³C NMR (600 MHz, d₈-Toluene): δ 209.23 (d, J_C–Fe = 26.3 Hz, CO). MS ESI- (m/z) 515.6 ((Et₄N)[⁵⁷Fe₂(adt)(CN)₂(CO)₄]⁻) IR (acetonitrile): ν_C≡N = 2075 (m) ν_C≡O = 1968 (s), 1924 (s), 1891 (s), 1873 (sh).

Preparation of ⁵⁷Fe-Labeled HydA1

All samples were handled strictly anaerobically. All buffers were carefully degassed. Unmaturated HydA1 was prepared as described by Kuchenreuther et al.³⁸ In order to selectively label the [4Fe-4S]_H cluster with ⁵⁷Fe, unmaturated HydA1 was treated with 6 M guanidinium chloride in 100 mM Tris/ HCl pH, 8.0, 20 mM EDTA to have a final concentration of 100–200 µM protein. After 30 min of incubation, samples were buffer exchanged into 100 mM Tris/HCl, pH 8.0, 150 mM NaCl using PD10 desalting columns (GE Healthcare). For the reconstitution reaction, a protocol similar to literature procedures was used.^{39 57}FeCl₃ was prepared from ⁵⁷Fe powder by dissolving in concentrated hydrochloric acid, evaporation of the solvent and resolubilizing in distilled water. The protein solution (50–70 µM) was incubated with 5 mM DDT for 5 min and slightly stirred. 10-fold excess of ⁵⁷FeCl₃ solution was added stepwise to the reaction mixture within 20 min. Thereafter, 10-fold excess of an aqueous solution of Na₂S was added stepwise. The reaction mixture turned from reddish (after addition of ⁵⁷FeCl₃) to a brownish color typical for unmaturated HydA1. After removing excess of ions using PD10 desalting columns, samples were concentrated and kept frozen until use.

In order to selectively label unmaturated HydA1 on the [2Fe]_H subcluster, it was incubated with a 2-fold excess of (Et₄N)₂[⁵⁷Fe₂(adt)-(CN)₂(CO)₄] following recent protocols.^8,10 In order to obtain the H_ox-CO inhibited state, samples were first oxidized with thionine and then flushed with CO for 20 min (Figure S2, SI). The redox states of all samples were verified using FT-IR spectroscopy.

EPR Analysis

Field swept Q-band EPR spectra were recorded in the pulsed mode using FID detection after a 1 µs π/2 excitation pulse. After a pseudomodulation transformation, the spectra obtained in this way are comparable to those using CW EPR.⁴⁰

Electron nuclear double resonance (ENDOR) was used to study the ⁵⁷Fe hyperfine interactions. In this investigation the Davies ENDOR sequence was used: [π]-t_d1-[RF]-t_d2-[π/2]-τ-[π]-τ-(ESE).^41,42 The excitation of nuclear spin transitions is detectable through an increase in the inverted ESE intensity. The microwave preparation pulse was set to 140 ns, whereas the length of the radiofrequency (RF) pulse was 45 µs and the shot repetition time 0.8 ms.

Q-band hyperfine sublevel correlation spectroscopy (HYSCORE) experiments were performed using the standard HYSCORE pulse sequence: [π/2]-τ-[π/2]-t₁-[π]-t₂-[π/2]-τ-(ESE).^42,43 The length of the microwave [π/2] and [π] pulses was adjusted to the maximum available microwave power (3 W). The delay between the first two pulses (τ) was 352 ns. The starting t₁ and t₂ delays in all measurements were 100 ns and were changed by a step of 16 ns. To suppress the effect of unwanted echoes, a four step phase cycling of the microwave pulses was used.

Q-band experiments were performed on a Bruker ELEXYS E580 spectrometer with a SuperQ-FT microwave bridge and a home-built resonator described earlier.⁴⁴ Cryogenic temperatures (20 K) were obtained by an Oxford CF935 flow cryostat. ENDOR experiments were performed using the random (stochastic) acquisition technique and making use of a 300W ENI 300 L RF amplifier. A Trilithic H4LE35-3-AA-R high-power low pass filter (cut off frequency around 35 MHz) was used to suppress the “harmonics” of the ¹H ENDOR signals.

All the simulations were performed in an EasySpin based program written in the MATLAB environment.⁴⁵ ENDOR spectra were simulated using the “salt” routine and the frequency domain calculations of HYSCORE spectra were simulated using the “saffron” routine. Signals corresponding to the different nuclei were simulated separately in order to reduce computing time.

Mössbauer Analysis

Mössbauer spectra were recorded on a conventional spectrometer with alternating constant acceleration of the γ-source. The sample temperature was maintained constant in an Oxford Instruments Variox cryostat. Isomer shifts are quoted relative to iron metal at 300 K. Mössbauer spectra were collected for frozen aqueous solution samples (1–2 mM, 650 µL) at 160 K and fitted using the program MFIT (written by Eckhard Bill, Max Planck Institute for Chemical Energy Conversion) with Lorentzian doublets.

NRVS

Nuclear resonance vibrational spectroscopy (NRVS) of [2⁵⁷Fe]_H HydA1 H_ox-CO and the precursor (Et₄N)₂[⁵⁷Fe₂(adt)-(CN)₂(CO)₄] were measured with a coldfinger helium cryostat maintained at 10 K. The real sample temperatures were obtained by the imbalance of the spectra and ranged from 40–60 K. NRVS measurement was performed at SPring-8, BL09XU and BL19LXU using C beam mode. A high heat load LN2-cooled Si(1,1,1) double crystal monochromators followed by a high resolution monochromators of Ge(4,2,2) and two Si(9,7,5) were used to achieve a 0.8 meV energy resolution. The delayed nuclear and K_α fluorescence were measured with a 2 × 2 APD array, and processed with the associated electronics.