Introduction

The continuously increasing interest in renewable energy technologies has also stimulated research into hydrogen production using microorganisms [1–3]. Many sulfate-reducing bacteria and monocellular algae make use of H2 and hydrogenases as part of their energy metabolism.

Three types of hydrogenases can be identified according to the metal content of their active site [1, 4, 5]. For an overview of the field, the reader is referred to the 100th thematic issue of Chemical Reviews on hydrogen [6]. [NiFe] hydrogenases have one nickel and one iron atom in their active site [7–9], while [FeFe] hydrogenases accommodate six iron atoms and no other metals [6, 10, 11]. The third class of hydrogenases, the [Fe] hydrogenases, contain a single iron atom in their cofactors [12–14]. This type of enzyme is also called the iron–sulfur-cluster-free hydrogenase [15], or “Hmd” (H2-forming N 5,N 10-methylenetetrahydromethanopterin dehydrogenase).

The structures of two [FeFe] hydrogenases are known so far from X-ray crystallographic studies. One is a periplasimic [FeFe] hydrogenase from Desulfovibrio desulfuricans [16] (abbreviated as DdH). The other is a cytoplasmic [FeFe] hydrogenase I from Clostridium pasteurianum [17] (abbreviated as CpI). The structures of these hydrogenases exhibit some differences. DdH was found to be a heterodimer containing in total two [4Fe–4S] clusters in addition to the active site. CpI is a monomer with three additional [4Fe–4S] clusters and one [2Fe–2S] cluster. Nevertheless, the structures of the active site (the so-called H-cluster) in these two hydrogenases were found to be very similar [18].

The H-cluster contains six iron atoms arranged in two connected subclusters [16, 18] (Structure 1): a Cys-coordinated [4Fe–4S]H subcluster, connected to a [2Fe]H subcluster via a Cys-thiol ligand. Remarkably, each iron in the binuclear subcluster is coordinated by CO and CN− ligands [19, 20]. The distal iron (relative to the [4Fe–4S]H subcluster) has an open coordination site, which is believed to be the site for hydrogen binding [6, 21, 22]. Despite extensive investigations of the structure of the H-cluster by various methods, the identity of the central atom in the dithiol bridging ligand has not been resolved yet, so far CH2, NH, and O were proposed [17, 20, 23, 24]. On the basis of mechanistic considerations, most researchers are inclining to the dithiolmethylamine ligand. However, a recent X-ray crystallographic study of CpI in combination with a theoretical study favored a dithiomethyl ether ligand [24].

Structure 1
structure 1

 

Full size image

Several states of the H-cluster have been observed and characterized by various spectroscopic methods [6, 21]. In contrast to most other [FeFe] hydrogenases, DdH can be isolated aerobically. In this case, it is inactive (Hinact state) and needs to be activated under reducing conditions. During the activation, it passes through an intermediate state called Htrans, which is characterized by a reduction of the [4Fe–4S] 2+H subcluster to a [4Fe–4S]+ state, exhibiting an S = 1/2 electron paramagnetic resonance (EPR) signal [25, 26]. The binuclear subcluster remains in an Fe(II)–Fe(II) state [6]. Further lowering of the reduction potential leads to the active oxidized state (Hox), where the [4Fe–4S]H subcluster is diamagnetic, but the binuclear subcluster is an S = 1/2 system [27]. The Hox state of DdH is characterized by a vacant coordination position on the distal iron [6, 28]. In the case of CpI, an X-ray crystallographic study indicated that the open coordination site of the H-cluster is occupied by an oxygen species (H2O or OH–) [24]. It is assumed that the irons in the binuclear subcluster are formally in the mixed valence state Fe(I)–Fe(II) and adopt a low-spin configuration, resulting in an S = 1/2 EPR signal [26, 27, 29, 30]. According to Mössbauer spectroscopy data, the [4Fe–4S]H subcluster remains in the formal 2+ (S = 0 ground state) state in all active states of the H-cluster investigated [29].

The fully reduced state of the H-cluster (Hred) is EPR-silent. It is characterized by the Fe(I)–Fe(I) valence state of the binuclear subcluster [6, 30, 31]. According to Fourier transform (FT) IR and crystallographic studies, the bond between the bridging CO ligand and the proximal iron is broken upon reduction of Hox and so this CO ligand is then bound in a terminal position to the distal iron [19, 20, 31].

Inhibition of the active enzyme by CO has been investigated by various methods, including X-ray crystallography. It has been shown that extrinsic CO binds to the distal iron, leading to inactivation of the [FeFe] hydrogenase (the so-called Hox–CO state) [31–34]. The CO-inhibited state shows an S = 1/2 EPR signal [35]. Earlier investigations of this state showed that it is an Fe(I)–Fe(II) mixed valance state, similar to the Hox state [31, 33, 34]. Our recent 57Fe study has shown that in this state the spin density is mostly located around the proximal iron. On the basis of these results, together with recent FTIR studies, a formal Fep(I)–Fed(II) state has been suggested [30, 31, 34].

Upon illumination of a liquid solution of enzyme in the Hox–CO state at temperatures between 275 and 278 K, the extrinsic CO ligand can be reversibly dissociated [19, 31]. In the frozen state and at temperatures below 60 K this leads to the generation of the Hox state. In addition, another photodissociated state is formed and is characterized, most probably, by the loss of the bridging CO ligand. It has been shown that by increasing the temperature above 150 K, one can fully restore the Hox–CO state [31, 36].

Clostridium pasteurianum was believed to contain two [FeFe] hydrogenases. Both hydrogenases have been investigated in the Hox–13CO state by continuous-wave (CW) EPR and CW electron–nuclear double resonance (ENDOR) [37, 38] at X-band frequencies. CpI revealed a single quite isotropic 13C hyperfine coupling of 20–22 MHzFootnote 1 [37]. [FeFe] hydrogenase II from C. pasteurianum showed a rather different 13C hyperfine coupling, with principal values of A1 = 34.0 MHz, A2 = 36.0 MHz, and A3 = 29.0 MHz [38]. Since it is not clear whether this species represents a genuine second hydrogenase in the bacterium, or merely a breakdown product of CpI [6], we will not discuss these data in detail. DdH has also been studied in the 13CO-inhibited form. The CW EPR spectrum of the Hox–13CO state has been compared with that of nonenriched hydrogenase. Simulation of the line broadening indicated a single isotropic 13C hyperfine coupling of about 17 MHz [34]. The same authors investigated the effects of illumination at 275 K of the Hox–13CO state [31, 34]. It was found that upon prolonged illumination the IR bands assigned to the bridging CO and to the terminal CO of the distal iron shifted to longer wavelengths; therefore, it was proposed, that these two ligands exchanged with the extrinsic 13CO (so-called scrambling effect) [31, 34]. However, comparison of the EPR spectra before and after illumination did not reveal any additional broadening of the lines; therefore, it was concluded that only the 13C of the extrinsic CO ligand has a large hyperfine coupling, while the hyperfine coupling constants of 13C of the COd and the CObr ligands are too small to be observed at X-band EPR.

Structurally, the H-cluster is believed to be well preserved in all [FeFe] hydrogenases. However, variations in the spectroscopic properties of the H-clusters from different organisms (including the newly studied [FeFe] hydrogenases from green algae) may indicate distinct structural differences [6, 30, 40]. It is very important to determine the electronic structure of the H-cluster in great detail to understand what may cause these variations. Our recent study of the 57Fe hyperfine couplings in DdH showed delocalization of the unpaired spin over both irons in the binuclear subcluster, depending on the state of the H-cluster [30]. More details about the electronic structure can be obtained from investigations of magnetic ligand nuclei in the binuclear subcluster. Here we present a pulse EPR investigation of the 14N and 13C nuclear spin interactions of the CO and CN– ligands for the Hox–CO state of DdH. Also the effects of illumination of the H-cluster detected by this technique will be discussed.

It is shown that 13C hyperfine couplings can be extracted from hyperfine sublevel correlation spectroscopy (HYSCORE) and ENDOR spectra with high precision, allowing the assignment of the signals obtained to different CO ligands. The observed 13C signals were also used to understand the scrambling of the CO ligands under light. In addition, information about 14N interactions has been obtained, which is important to understand the distribution of the unpaired spin density over the CN– ligands of the irons in the binuclear subcluster.

Materials and methods

EPR setup

X-band measurements were performed using a Bruker Elexsys E-580 X-band spectrometer equipped with a SuperX-FT microwave bridge and an Oxford Instruments CF935 helium-flow cryostat. Pulse EPR, ENDOR, and HYSCORE spectra were obtained using a Bruker EN 4118X-MD4 dielectric ENDOR resonator with an Applied Systems Engineering 1-kW traveling wave tube amplifier (model 117x). CW EPR measurements were performed with the same spectrometer using a Bruker ER 4118X-MD5 dielectric resonator without ENDOR coils.

All Q-band HYSCORE and pulse ENDOR measurements were performed using a Bruker Elexsys E 580 Q-band spectrometer equipped with a SuperQ-FT microwave bridge and an Oxford Instruments CF935 flow cryostat at temperatures ranging from 10 to 20 K. For these measurements we used a slightly overcoupled cylindrical TE011 homebuilt resonator with a construction similar to that described by Sienkiewicz et al. [30, 41].

For pulse ENDOR experiments with the random-acquisition procedure we used a homebuilt data acquisition system, based on SpecMan software on a personal computer [42]. The Bruker spectrometer was used to generate microwave pulses and to trigger the SpecMan system, which in turn controls the generation of radiofrequency pulses and records the signal coming from the Bruker spectrometer. In these experiments the radiofrequency pulses were generated by an Agilent E4420B radiofrequency generator and amplified by a high-power AR 2500L radiofrequency amplifier from Amplifier Research™, running in CW mode (2,500-W output). To suppress the “harmonics” of the 1H ENDOR signals (around 51 MHz at 1.2 T), a Trilithic™ H4LE35-3-AA-R high-power low-pass filter (cut-off frequency about 35 MHz) was used.

EPR methods

Q-band EPR spectra were obtained using free induction decay detected EPR with a 1-μs microwave pulse. This length of the microwave pulse was found to be sufficient to suppress field-dependent distortions in the absorption-like spectrum due to the presence of magnetic nuclei. To facilitate the comparison with CW X-band spectra, free induction decay detected spectra were differentiated using the so-called pseudomodulation procedure described by Hyde et al. [43].

HYSCORE [44] was used to extract most of the parameters of 13C and 14N nuclear interactions. The pulse sequence for this method was

$$ [\pi /2] - \tau - \,[\pi /2\left] {\, - t_{1} - \,} \right[\pi \left] {\, - t_{2} - \,} \right[\pi /2] - \tau - \left( {\text{ESE}} \right). $$

This is a 2D method, in which the intensity of the stimulated electron spin echo (ESE) is detected as a function of delays t 1 and t 2. The 2D modulation pattern obtained was processed using third-order background polynomial subtraction and Hamming apodization followed by zero-filling and 2D fast Fourier transformation.

X-band HYSCORE spectra were obtained using an 8-ns microwave pulse for the “π/2” pulses and a 16-ns microwave pulse for the “π” pulse. The delay between the first two pulses (τ) was adjusted according to an ESE envelope modulation (ESEEM) experiment in which a three pulse ESEEM spectrum is recorded as a function of τ [45, 46]. To suppress the effects of unwanted echoes, a four-step phase cycling of the microwave pulses was used [47]. An additional set of experiments was performed using pulse ENDOR at Q-band frequency utilizing the Davies ENDOR sequence [48]:

$$ [\pi \left] { - t_{{{\text{d}}1}} - } \right[{\text{RF}}\left] {\, - t_{{{\text{d}}2}} - } \right[\pi /2] - \tau - [\pi ] - \tau - \left( {\text{ESE}} \right). $$

The excitation of nuclear spin transitions is detectable through a reduction of the ESE intensity. The length of the radiofrequency (RF) pulse was adjusted to maximize the ENDOR effect of the high-frequency feature in the 13C ENDOR spectrum. The delay before the radiofrequency pulse (t d1) was set to 1 μs. It has been found that the delay between the radiofrequency pulse and the detection sequence (t d2) needs to be longer then 3 μs to avoid distortions in the ENDOR spectra.

Simulations

Simulation of the spectra were based on the spin Hamiltonian approach using

$$ {\hat{\mathcal{H}}}_{0} = \beta \vec{B} \cdot {\mathbf{g}} \cdot \hat{S} + \sum {\mu_{n} g_{n} {}^{i}\vec{\user2{B}}} \cdot \hat{I}_{i} + \sum {\hat{S} \cdot {\mathbf{A}}_{i} \cdot \hat{I}_{i} + } \sum {\hat{I} \cdot \user2{P}_{i} \cdot \hat{I}_{i} }, $$
(1)

where β is the Bohr magneton, \( \vec{B} \) the magnetic field vector, \( \hat{S} \) the electron spin operator, g the g tensor, μ n the nuclear magneton, g n the nuclear g value, A i the hyperfine tensor, \( \hat{I}_{i} \) the nuclear spin operator, P i the nuclear quadrupole tensor, and the sums run over all nuclei (i), interacting with the unpaired electron spin. The first and the second terms represent the electron and the nuclear Zeeman effects, the third term is the hyperfine interaction of the unpaired electron and the nuclear spins, and the last term represents the quadrupole interaction of the nuclear spins for nuclei with I > 1/2.

The quadrupole tensor P is traceless. The representation in its principal axis system (in frequency units) can be written in the following way:

$$ \left[ {P_{x} ,\,P_{y} ,\,P_{z} } \right] = \frac{{e^{2} qQ}}{4I(2I - 1)h}\left[ { - (1 - \eta ), - (1 + \eta ),\,2} \right]. $$
(2)

Here we will use two variables to characterize the quadrupole coupling:

$$ K = \frac{{e^{2} qQ}}{4I(2I - 1)h} $$

and

$$ \eta = \frac{{P_{x} - P_{y} }}{{P_{z} }}. $$

Orientations of the hyperfine and quadrupole tensors are presented with respect to the g tensor using Euler angles. In the calculations we used the “y” convention for the Euler angles, in which the first angle (α) “rotates” along the z-axis, the second angle (β) along the y′-axis, and the third (γ) along the z″-axis [45].

ENDOR spectra were simulated using the “salt” routine of the EasySpin package [49] for MATLAB™ [50]. Frequency-domain calculations of HYSCORE spectra were facilitated using home-written routines for MATLAB™ [50] utilizing a general expression for the nuclear modulation signal for the case of ideal microwave pulses, presented by Shane et al. [51]. The case of several interacting nuclei was treated according to the product rule for HYSCORE [52].

Growth of D. desulfuricans ATCC 7757

Desulfovibrio desulfuricans cultures were grown for three days in a 10-L glass fermenter with an automatic pH control unit, which kept the pH within a range of 6.5–8. The medium contained 2.0 g MgSO4·7H2O, 5.0 g sodium citrate·2H2O, 1.0 g CaSO4·2H2O, 1.0 g NH4Cl, 0.5 g K2HPO4, 7.0 g sodium lactate, and 1.0 g yeast extract in 1 L H2O, pH 7.5. Lactate, K2HPO4, and yeast extract were autoclaved separately to avoid an undesired precipitation of the medium. These three media were afterwards mixed aseptically and 5 mL of a sterile filtered solution of 20% Fe(NH4)2(SO4)2·6H2O was added. During the growth, the culture was fed with 200 mL 50% lactic acid per day, divided into small portions.

The growth medium was kept anaerobic by constant bubbling with nitrogen. The dense culture was then harvested aerobically by centrifugation and the wet cell pellet (46 g) was stored at 193 K.

Isolation of DdH

The isolation followed the purification protocol of Hatchikian [53], originally established by van der Westen [54]. The wet cell pellet (46 g) of a harvested D. desulfuricans was incubated for 30 min at room temperature in a buffer with 50 mM Na2-EDTA and 50 mM tris(hydroxymethyl)aminomethane (Tris)/HCl (adjusted to pH 9 with HCl) under gentle stirring to obtain the periplasmic fraction, which contains the [FeFe] hydrogenase. This fraction was then separated from the cells by centrifugation at 35,000g for 30 min. The periplasm was precipitated with ammonium sulfate in 50 and 80% saturation steps at room temperature and centrifuged as mentioned above. The second precipitate contained the hydrogenase.

The precipitate was dialyzed overnight against 10 mM Tris/HCl buffer, concentrated, and then loaded on a DEAE 52 ion-exchange column. The resulting chromatogram showed three major peaks. From these peaks, selected fractions were tested for hydrogenase activity via a qualitative test with methyl viologen: 100-μL sample and 2 mL 25 mM Tris/HCl buffer comprising an excess of methyl viologen (approximately 0.8 M) were sealed in a glass vial and the glass vial was connected to a hydrogen-gas line. After 2–3 min of flushing with hydrogen, hydrogenase-containing fractions turn blue as the electrons from the oxidation of molecular hydrogen convert methyl viologen into its blue reduced radical form. Active fractions were pooled and further purified by a Sephacryl S 200 gel filtration column. A Biogel K-phosphate column finally separated the pooled concentrated fractions. The elution profile showed a single peak of active fractions, which were desalted by a PD-10 column and concentrated to a volume of 50 μL by Millipore Centriprep concentrators. The purity of the sample was monitored with sodium dodecyl sulfate polyacrylamide gel electrophoresis and showed two distinct bands representing the small and the large subunit, respectively (not shown). The concentration of the protein was calculated with the aid of UV/vis spectra of the sample. The purified enzyme (0.6 mg) was stored in a liquid nitrogen tank. The procedure was repeated twice and yielded samples of similar concentrations.

Treatment of the samples with gas

To activate the aerobically isolated [FeFe] hydrogenase, we followed the procedure described by Hagen et al. [55]. The sample in the sealed glass vial was flushed with hydrogen gas after evacuation of the vial. The procedure was repeated several times within 15 min. To obtain the Hox state, hydrogen was replaced by argon and the procedure was repeated several times for another 15 min. The CO-inhibited state (Hox–CO) was obtained by flushing the enzyme in the Hox state with CO gas for 15 min. To prepare the Hox–13CO state, 13C-enriched CO gas (99% enriched in 13C and containing less than 10% 18O) from Cambridge Isotope Laboratories was used. After a gas treatment, the sample was transferred to the EPR tubes in an anaerobic glove box (Coy Laboratory Products) under a nitrogen atmosphere with an admixture of 2% hydrogen. Subsequently, the EPR samples were frozen in liquid nitrogen.

Results

Figure 1 shows the X- and Q-band CW EPR spectra of Hox–CO from DdH. The spectrum is characteristic for all [FeFe] hydrogenases, from both bacterial and algal sources [6]. The slight rhombicity of the g tensor has already been discussed by Silakov et al. [30] and Albracht et al. [34] on the basis of X-band EPR but it is better resolved at Q-band.

Fig. 1
figure 1

X- and Q-band electron paramagnetic resonance (EPR) spectra of the Hox–CO state of the H-cluster measured at 40 K (blue) and simulations (red) using the g values shown for each experimental spectrum. Spectra are displayed on the same g value scale to facilitate a direct comparison. The experimental conditions were as follows: a continuous-wave (CW) EPR, P mw 20 μW (40-dB attenuation), ν mw 9.7134 GHz, time constant 40.96 ms, conversion time 81.92 ms, modulation amplitude 0.5 mT, modulation frequency 100 kHz; b free induction decay detected EPR, microwave pulse length 1 μs, shot repetition time 500 μs, ν mw 33.8485 GHz. To facilitate direct comparison with the X-band CW EPR spectrum (a), the first derivative of the spectrum is presented

Full size image

The spectra can be simulated using the following principal g values:

$$ g_{1} = 2.065,\;g_{2} = 2.007,\;g_{3} = 2.001. $$

Remarkably, the slight rhombic distortion observed for the D. desulfuricans spectra was not present in the EPR spectra of other [FeFe] hydrogenases from C. pasteurianum [37, 38] and from C. reinhardtii, C. submarinum, and C. moewusii [40]. The unpaired spin density distribution is very sensitive to the geometry of the molecule. Thus, differences in the observed g values are most probably related to slight geometrical variations, caused by the protein surroundings in the different [FeFe] hydrogenases. Accurate measurements of the 13C and 14N nuclear spin couplings in the H-cluster of the CO ligands and the CN– ligands, respectively, yield information on the electronic structure and its variation among the different [FeFe] hydrogenases. The findings of such experiments, performed on DdH, are presented in the following sections.

14N interactions of the CN– ligands

The 14N hyperfine couplings provide important information about the distribution of unpaired spin density over the CN− ligands in the binuclear subcluster (Structure 1). Since 14N has a nuclear spin of I = 1, additional information can be obtained from the quadrupole interaction, which probes the electric field gradient and thus the intimate surroundings of the 14N nucleus. Nuclear quadrupole couplings are characteristic for a nucleus in a specific bonding situation.

The investigation was performed at X-band frequency using HYSCORE. Several HYSCORE spectra were measured at various positions of the magnetic field. It should be noted that the rhombic distortion of the high-field component of the X-band EPR spectrum is about 0.8 mT, while the excitation bandwidth of the 16-ns microwave pulse is about 3.0 mT. Thus, the HYSCORE spectrum measured at g 2 is considered to contain all orientations related to both g 2 and g 3. The resulting X-band HYSCORE spectra measured for the H-cluster in the Hox–CO state are shown in Fig. 2.

Fig. 2
figure 2

X-band hyperfine sublevel correlation spectroscopy (HYSCORE) spectra of the nonlabeled Hox–CO state of the H-cluster in Desulfovibrio desulfuricans measured at 20 K: a complete spectrum, measured at 347.9 mT (g 2) [the (+–) quadrant shows no signals and thus is not shown]; b low-frequency part (14N signals) of a; c low-frequency part of the HYSCORE spectrum taken at 338.1 mT (close to g 1); d and e simulations of the 14N signals in the spectra shown in b and c, respectively, involving only one 14N nucleus and using the parameters given in Table 1. The experimental conditions were as follows: ν mw 9.7833 GHz; τ 120 ns; t 1 and t 2 step 16 ns; shot repetition time 2 ms

Full size image

The HYSCORE spectra revealed two separate sets of crosspeaks in the (++) quadrant, while no signals were observed in the (+−) quadrant (not shown). One set of peaks around 15 MHz is centered at the Larmor frequency of the 1H nucleus [ν n(1H) = 14.81 MHz at 347.9 mT]. It is assumed that these signals originate from the nonexchangeable β-protons of the Cys ligands of the [4Fe–4S]H subcluster [37]. As observed for various [4Fe–4S]-containing systems [56, 57], these 1H nuclear spins exhibit relatively large dipolar hyperfine couplings. Since all 1H signals are overlapping in the “powder” spectra,Footnote 2 the analysis cannot yield all hyperfine coupling parameters. We expect that EPR and ENDOR studies of single crystals of [FeFe] hydrogenase would be ideal to determine these interactions.

Apart from the 1H signals, the low-frequency part of the HYSCORE spectra (3–4 MHz) shows another set of crosspeaks. The field dependence of these crosspeaks can be fitted as 14N nuclear spin interactions. According to our experience, crosspeaks of ΔM I  = 2 transitions can be more pronounced in HYSCORE spectra than ΔM I  = 1 transitions, depending on the magnitude of the quadrupole and the hyperfine interactions. Therefore, the peaks around 3.5 MHz are assigned to double quantum transitions of a 14N nuclear spin. The shallow peak at about 1.7 MHz in Fig. 2c might represent a single quantum transition. The low-frequency region of the HYSCORE spectra, taken at the g 2 and g 1 positions, is shown in Fig. 2b and c.

The simulation of these spectra (Fig. 2d, e) yields the hyperfine and quadrupole parameters summarized in Table 1. Remarkably, the weak crosspeaks at 1.7 MHz could also be reproduced (Fig. 2e).

Table 1 Parameters of the hyperfine and quadrupole coupling of the 14N nucleus, from the analysis of hyperfine sublevel correlation spectroscopy spectra (HYSCORE)
Full size table

It was anticipated that the nitrogens from two CN− ligands would contribute to the HYSCORE spectrum; however, only one interacting 14N nucleus was sufficient to simulate the experimental data (Fig. 2d, e). This indicates a localization of the unpaired spin density on one of the iron atoms. Thus, strongly different hyperfine couplings for the two nitrogen nuclei of the CN− ligands are expected. If the observed 14N signals are from the nucleus with the largest hyperfine coupling, then the other 14N hyperfine coupling might be too small to be observed. On the other hand, since these crosspeaks are not well resolved, a contribution from two 14N nuclear spins cannot be completely excluded. Remarkably, the magnitude of the quadrupole coupling is somewhat smaller than is usually found for CN− ligands (0.9–1.5 MHz) [58, 59]. According to the analysis of the X-ray crystallographic structure, a hydrogen bond from the closest amino acid to the nitrogen of the CN− ligand may be formed [23, 24]. In the DdH X-ray crystallographic structure (Protein Data Bank ID 1HFE) they are Ile204, Lys237, and Ala109. Hydrogen bonding to CN− is expected to decrease the magnitude of the quadrupole coupling.

Here one should mention another possible source of the observed 14N signals. Reanalysis of the X-ray crystallographic structure together with some theoretical studies have proposed a –CH2–NH–CH2– moiety for the bridging dithiol ligand in the binuclear subcluster (Structure 1) [20, 60]. Therefore, the observed 14N signals could also originate from this ligand, though it seems unlikely, owing to its remote position with respect to the iron atoms. Moreover, the quadrupole coupling of this 14N nucleus (secondary amine) is expected to be larger (about 1.2 MHz) [58] than was observed experimentally. Since the observed hyperfine interaction points to a weakly coupled 14N nucleus, we conclude that there is only very little unpaired spin density on the CN− ligands of the binuclear subcluster.

13C interactions of the CO ligands

More information about the distribution of the unpaired spin density can be obtained from the investigation of the 13C hyperfine couplings of the CO ligands. The 13CO-inhibited state (Hox–13CO) can be easily obtained from the oxidized active state of the H-cluster by flushing it with 13C-enriched CO gas for 15 min. The Hox–13CO sample reveals a moderately broadened X-band CW EPR signal in comparison with the nonenriched sample (Fig. 3). The broadening of the EPR spectrum due to labeling with the 13C isotope can be simulated using one isotropic 13C hyperfine coupling of 0.60 mT (16.8 MHz), as described before [34].

Fig. 3
figure 3

Comparison of X-band CW EPR spectra of the Hox–CO (red) and the Hox–13CO (blue) states. The experimental conditions are the same as for the X-band CW EPR spectrum presented in Fig. 1a. Asterisks indicate a minor contribution of the Hox state due to incomplete inhibition of the H-cluster by CO

Full size image

However, since the hyperfine splitting is not resolved in the EPR, more precise methods need to be applied to obtain the parameters of the hyperfine tensor with high accuracy. Therefore, we performed pulse EPR and ENDOR spectroscopy on Hox–13CO.

HYSCORE spectra at X-band were measured at several positions of the external magnetic field. Apart from the 14N and 1H signals described above, an additional pair of crosspeaks was identified in these spectra. The spectrum measured at the position of highest EPR intensity (g 2) is shown in Fig. 4.

Fig. 4
figure 4

X-band HYSCORE spectrum [(+−; left) and (++; right) quadrants] of the Hox–13CO state. The experimental conditions were as follows: B 0 346.8 mT (g 2); ν mw 9.7607 GHz; τ 120 ns; t 1 and t 2 step 16 ns; shot repetition time 2 ms

Full size image

The pair of crosspeaks in the (+−) quadrant can be exclusively assigned to one 13C nucleus because it is not present in the spectra of the nonenriched sample. Moreover, the crosspeaks are separated by twice the Larmor frequency of a 13C nucleus (ν n (13C) = 3.75 MHz at 350 mT). The orientation-selective experiments reveal that this hyperfine interaction is very anisotropic. The field-dependent hyperfine coupling constant increases with decreasing magnetic field. Unfortunately, a complete analysis is not possible using this method since at the lower magnetic field (around g 1) these signals are not observable in the HYSCORE spectra (data not shown here). It is well known that signals from nuclei with hyperfine couplings close to exact cancellation (A ≈ 2ν n) are most prominent in ESEEM spectra, while signals from strongly coupled nuclei vanish with increasing hyperfine coupling constant [61]. Apparently, at this field position the NMR frequencies of the 13C signals are too large to be detected using this method; therefore, additional Q-band ENDOR measurements were performed to extract all the parameters of this 13C hyperfine coupling. At this microwave frequency (about 34 GHz) the 13C signals are well separated from the 1H signals and only partially overlap with the 14N peaks, which are, in fact, hardly detectable. Figure 5 shows Davies ENDOR spectra together with their simulation (see hyperfine coupling A 1C from Table 2).

Fig. 5
figure 5

Q-band Davies electron–nuclear double resonance (ENDOR) spectra of the Hox–13CO state, measured at several positions of the magnetic field (blue) and simulations using values of the A 1C hyperfine coupling from Table 2 (red). The experimental conditions were as follows: temperature 15 K; ν mw 33.8505 GHz; shot repetition time 2 ms; length of the radiofrequency pulse 25 μs; B 0 a 1,207.8 mT (g 3), b 1,204.7 mT (g 2), c 1,170.7 mT (g 1)

Full size image
Table 2 Principal values of the 13C hyperfine tensor and assignment of different couplings to the CO ligands of the binuclear subcluster
Full size table

From the X-ray crystal structure of CpI it is known that upon reaction of the H-cluster with CO gas, the extrinsic CO binds to the open coordination site of the distal iron [32] (Structure 1). Since the FTIR spectra of DdH and CpI are rather similar for the CO-treated samples (Hox–CO state), it can be assumed that in the case of D. desulfuricans the binding site for the exogenous CO is the distal iron as well. Hence, the extracted 13C hyperfine tensor was assigned to the extrinsic CO ligand of the distal iron (COext). Earlier ENDOR studies of the Hox–13CO state of the active site of CpI revealed a single 13C hyperfine coupling of about 21 MHz [37]. The extracted 13C hyperfine tensor was found to be rather isotropic (the dipolar contribution is less than 2 MHz). This differs from the A 1C coupling, which has a larger dipolar contribution and is somewhat smaller. This is in agreement with our previous study, in which the 57Fe hyperfine couplings of the Hox and Hox–CO states of DdH were also found to be different from those extracted for CpI [30, 37].

With use of the light-induced ligand exchange procedure described above [34], two other CO ligands of the distal iron can be labeled with 13C via exchange of the respective CO ligands during illumination of the Hox–13CO sample for 2–3 h at 273 K in the presence of 13CO gas. This effect thus provides an opportunity to undertake a more detailed investigation of the distribution of the unpaired spin over the bridging CO and terminal CO ligands of the distal iron. The “scrambled” sample after illumination will be called Hox–(13CO)3. The Hox–(13CO)3 sample was probed by X-band HYSCORE using settings similar to those for the Hox–13CO sample to provide a direct comparison of the spectra.

As shown in Fig. 6, two additional sets of crosspeaks can be clearly identified in the HYSCORE spectra of Hox–(13CO)3. Both of them are absent in the respective spectra of the Hox–13CO and the nonenriched samples.

Fig. 6
figure 6

X-band HYSCORE spectra of the Hox–(13CO)3 state measured at several field positions (left) and corresponding simulations (right) accounting for three interacting 13C nuclei using the parameters given in Table 2. The experimental conditions were as follows: t 1 and t 2 step 16 ns; shot repetition time 2 ms; B 0 a 346.2 mT (g 2), b 341.5 mT, c 337.0 mT (g 1); ν mw a 9.7328 GHz, b 9.7469 GHz, c 9.7576 GHz; τ a 136 ns, b, c 120 ns

Full size image

One pair of peaks remains in the (++) quadrant in all HYSCORE spectra, which is an indication of a weak (|2ν n| > |A|) and rather isotropic hyperfine coupling. Another set of signals shows up in the (+−) quadrant at higher magnetic field (g 2), while at low field it is in the (++) quadrant, which shows the large anisotropy of this 13C hyperfine coupling. The simulation of these HYSCORE spectra is also shown in Fig. 6; the parameters are listed in Table 2.

According to a recent FTIR study [31], the bridging CO ligand (CObr) and the terminal CO ligand of the distal iron (COd) are exchanged with 13COext and free 13CO upon illumination and thus become 13C-labeled with an average efficiency of about 60% after 4 h. It has also been shown that the “scrambling” does not affect the terminal CO ligand of the proximal iron (COp). On the basis of that observation, the additional 13C signals (A 2C and A 3C) were assigned to the CObr and the COd ligands, respectively (Structure 1).

There is no direct evidence for which of these two observed signals corresponds to which CO ligand; however, some assumptions based on the character of the hyperfine tensors can be made. Our recent investigation of the 57Fe hyperfine couplings [30] shows that the spin density is somewhat delocalized between the irons in the binuclear subcluster, which is also supported by the relatively large 13C hyperfine coupling of COext (A 1C). Therefore, the 13C nucleus of the CObr ligand is expected to have a large hyperfine coupling with a rhombic character of the hyperfine tensor rather than an axial one. Hence, we are inclined to assign the A 2C hyperfine tensor to the bridging CO ligand. Consequently, the other hyperfine coupling (A 3C) is assigned to the terminal CO ligand of the distal iron. According to density functional theory (DFT) calculations, no direct spin density is expected at this CO ligand [62]; therefore, a rather weak 13C hyperfine coupling is expected. This fits to the character of the observed A 3C hyperfine coupling.

Discussion

Spin density distribution

Recently we presented a study of the 57Fe-enriched [FeFe] hydrogenase in the Hox and Hox–CO state using pulse EPR spectroscopy [30]. It was found that both irons carry some spin density. On the basis of the observed 57Fe hyperfine couplings it was suggested that most of the spin density is located around the proximal iron in the case of the Hox–CO state, while for the Hox state a complete delocalization over the two irons in the binuclear subcluster was proposed.

The relatively large 13C hyperfine coupling found for the 13COext ligand of the distal iron may indicate that the spin delocalization is larger in the Hox–CO state than was estimated from the 57Fe hyperfine couplings.

According to recent theoretical calculations by Fiedler and Brunold [62], the unpaired spin in the Hox state is located mostly at the distal iron, while in the case of the Hox–CO state the spin density is distributed between both iron centers. When a complete H-cluster was taken into account, including all six iron atoms, the calculated Mulliken spin population ratio was found to be about 2:1 in favor of the proximal iron. For investigation of the electronic structure, Fiedler and Brunold used a truncated model [62] which consisted of only the binuclear subcluster. On the basis of the composition of the singly occupied molecular orbital it was observed that in this model spin density is more localized on the proximal iron (with a ratio of about 3.7:1 in favor of the proximal iron). This shows that the “cubane” plays an important role in the electronic structure of the H-cluster. Unfortunately no calculations of EPR parameters have been presented for the complete 6Fe model. Some of the calculated values for the 2Fe models with protonated (oxSH–CO) and not protonated (oxS–CO) sulfur of the bridging Cys residue are compared with experimental data in Table 3.

Table 3 Comparison of the experimental and calculated isotropic 13C hyperfine values and singly occupied molecular orbital (SOMO) composition ratio between Fep and Fed
Full size table

In both calculated 2Fe models of the Hox–CO state, the spin density around the distal iron is located in a \( d_{{z^{2} }} \)-shaped orbital and points towards the extrinsic CO ligand. An axial character of the extracted 13C hyperfine coupling (A 1C in Table 2) supports this. Therefore, it is expected to find a strong coupling of the 13C atom of COext even if the 57Fe hyperfine coupling on the distal iron is quite small. In general, this agrees with the experimental findings. Nevertheless, the closest calculated hyperfine coupling of 29 MHz (oxSH–CO model in Table 3) is much larger than the one observed experimentally [A iso (A 1C) = 17.1 MHz; Table 2]. Therefore, we conclude that the calculated spin population on the distal iron is overestimated.

In a first estimation we assume that the 13C hyperfine coupling of the extrinsic CO ligand is proportional to the amount of spin density on the distal iron. Thus, to lower the 13C hyperfine coupling to about 17 MHz, the spin population on the distal iron must be reduced about 2 times from what has been calculated. To achieve this, the spin density must be “shifted” towards the proximal iron. Consequently, the spin-population ratio between the proximal and the distal irons in the binuclear subcluster should be about 5:1 for the Hox–CO state, which better agrees with the picture of a rather localized spin density, concluded from the analysis of the 57Fe hyperfine interactions [30]. Therefore, despite the seemingly large 13C coupling of COext, we incline to the conclusion that the proximal iron carries most of the spin density in the Hox–CO state.

In the Hox state we found an equal distribution of the unpaired spin density over both irons, on the basis of the 57Fe hyperfine couplings [30]. This differs dramatically from the results of the DFT calculations [62] in which about 80% of the spin density was calculated to be localized at the distal iron. However, if we again assume that the spin population on the distal iron is overestimated 2–3 times, it would lead to the conclusion of an almost equal distribution of the unpaired spin density over both irons in the binuclear subcluster. This would fit the experimental data quite well.

One of the possible origins of this overestimation of the spin distribution on the distal iron by DFT may be the truncation of the H-cluster to a 2Fe model used for calculation of the spin distribution and EPR parameters. Nevertheless, only large-scale theoretical calculations of the structure and the magnetic resonance parameters could clarify this point. Moreover, we believe that the protein environment should also be taken appropriately into account to reproduce the experimental data.

Scrambling

Upon illumination of the Hox–13CO state, two other CO ligands of the distal iron can be exchanged with 13CO [so-called Hox–(13CO)3 state] [31, 34]. This result from FTIR spectroscopy has been corroborated by our EPR experiments. However, an exchange of the terminal CO ligand of the proximal iron was not completely excluded by the FTIR data.

According to our interpretation of the EPR data, the unpaired spin density is located at the proximal iron in Hox–CO; therefore, it is expected, that the 13CO ligand of this iron atom should produce sufficiently large 13C hyperfine coupling (although it could be mostly dipolar), which should be detectable by either HYSCORE or ENDOR spectroscopy. However, neither of these methods showed any trace of a fourth 13C hyperfine coupling, although a weakly coupled 14N nucleus has been observed. Therefore, our results are in agreement with the notion that under the experimental conditions used only the CO ligands of the distal iron exchange upon illumination with sufficient rates.

Comparison with previous results

Our investigation has shown a quite anisotropic 13C hyperfine coupling centered at A iso = 17.1 MHz for the COext ligand (at Fed). The ENDOR study of Telser et al. [37] on the active site of CpI yielded a somewhat larger and more isotropic 13C hyperfine coupling of A iso = 21.5 MHz for the extrinsic CO ligand. The difference cannot be explained by the uncertainty of the experiment. As discussed in our previous paper [30], the 57Fe hyperfine couplings are also different for these two species (DdH and CpI). We have observed that in the case of DdH both 57Fe and 13C hyperfine couplings are smaller than those found for the H-cluster in CpI. Therefore, it was concluded that the H-clusters of these two organisms are somewhat different in geometry, causing variations in the electronic structures. Unfortunately, no study on the Hox–(13CO)3 state has been performed so far for CpI.

One possible origin of the difference in the electronic structure between the two species can be a difference in the exchange coupling between the [2Fe]H and the [4Fe–4S]H subcluster [29]. The formally diamagnetic [4Fe–4S]2+ subcluster reveals strong 57Fe hyperfine couplings due to an exchange interaction between the electronic spin of the closest iron of the “cubane” and the unpaired electron spin of the binuclear subcluster. As discussed by Popescu et al. [29], the exchange interaction between the [4Fe–4S]H and the [2Fe]H subcluster has a major influence on the detected spin distribution in these subunits. Apparently, this effect is not only observed in the hyperfine couplings of the 57Fe nuclei, but also in the general distribution of the unpaired spin.

Several attempts have been made to elucidate the influence of the [4Fe–4S]H subcluster on the electronic structure of the binuclear subcluster by quantum chemical methods [62, 63]. Schwab et al. [63] discovered that in the 6Fe model of the H-cluster a strong delocalization of the frontier molecular orbitals is observed. The electron density difference plot between the 6Fe cluster and separate 4Fe and 2Fe subclusters shows large changes in the molecular orbitals. This indicates a strong influence of the connected [4Fe–4S] subcluster on the electronic structure of the binuclear subcluster. Therefore, differences in the geometry of the [4Fe–4S]H subcluster from species to species may affect the [4Fe–4S]H–[2Fe]H exchange interaction. In turn, this can alter the unpaired-spin distribution in the binuclear subcluster. As is known from X-ray crystallography, the [4Fe–4S]H subcluster is coordinated by Cys ligands, i.e., it is tightly bound to the protein environment. A geometrical difference in this part of the protein in DdH and CpI may be the reason for a difference in the geometry of the [4Fe–4S]H subcluster and thus in the electronic structure. Additionally, the protein surroundings may also affect the structure of the binuclear subcluster via electrostatic, dipole, and hydrogen-bonding interactions, as recently discussed [24].

Summary and conclusions

An extensive investigation of the distribution of the unpaired spin density over the ligands of the binuclear subcluster was performed for the Hox–CO state of the H-cluster of DdH.

The 13CO-inhibited state of the H-cluster (Hox–13CO) was studied by pulse EPR methods. A single 13C hyperfine coupling was observed and assigned to the external CO ligand. The extracted isotropic part of the hyperfine coupling agrees with the previously obtained hyperfine coupling for DdH obtained by Roseboom et al. [31, 34]. Use of advanced EPR methods allowed us to resolve all principal values of the 13C hyperfine coupling for the external CO ligand.

This hyperfine coupling, however, is substantially different with respect to both the isotropic and the anisotropic part from the one previously reported for CpI by Telser et al. [37, 39]; therefore, these results indicate a difference between the electronic structures of the H-cluster of these species. This has also been concluded from the observed 57Fe hyperfine couplings [30].

The reaction of the Hox–13CO state with light was investigated. As observed earlier by Roseboom [31, 34], illumination between 275 and 278 K resulted in scrambling of the CO ligands in the binuclear subcluster and thus in a 13C labeling of the terminal CO ligand of the distal iron and of the bridging CO ligand. For the first time 13C hyperfine couplings of these ligands have been extracted, which give insight into the extent of the spin density distribution over the [2Fe]H subcluster.

In addition, the unlabeled H-cluster in the Hox–CO state was investigated. A single weakly coupled 14N nucleus was observed. This shows that there is some spin density at one of the CN ligands, while the other one carries negligible spin density.

These data allow a refinement of the spin-distribution picture of the Hox–CO state. On the basis of the extracted hyperfine couplings we have concluded that although most of the spin density is indeed localized at the proximal iron, there is also a substantial distribution of the spin density over the distal iron and its CO ligands, but not over the CN ligands. On the basis of our results we conclude that the spin population at the proximal iron must be about 5 times larger than that at the distal iron in order for it to fit the experimental findings.

To summarize, we have mapped the distribution of the unpaired spin in the binuclear subcluster in the Hox–CO state for all CO and CN ligands except for the CO ligand of the proximal iron. We believe that the data obtained will be useful for further theoretical studies aiming at a correlation of the electronic structure of the H-cluster and the function of the enzyme.