Cofactor composition and function of a H2-sensing regulatory hydrogenase as revealed by Mössbauer and EPR spectroscopy† †Electronic supplementary information (ESI) available: Tables with the simulation parameters and details of the Mössbauer, and EPR spectra (Tables S1–S4). additional EPR and Mössba

A regulatory hydrogenase is characterised by Mössbauer, EPR and FTIR yielding insight into structure and function of this dihydrogen sensor.


Introduction
Hydrogenases, the biocatalysts carrying out the reversible oxidation of H 2 in nature, are complex metalloenzymes, which harbor metal cofactors with a fascinating coordination chemistry. [1][2][3][4] According to their active site metal content, they are grouped into [NiFe]-, [FeFe]-and [Fe]-hydrogenases. 4 The enzyme investigated in this study belongs to the class of [NiFe]hydrogenases, whose basic module consists of two subunits.
The large subunit harbors the heterobimetallic [NiFe] center, whereas the small subunit carries up to three iron-sulfur clusters, which serve as an electron relay during catalysis. [4][5][6][7][8] [NiFe]-hydrogenases are classied into ve phylogenetically distinct groups on the basis of their structure and cellular function. 9,10 Although most [NiFe]-hydrogenases are strongly inhibited by molecular oxygen, at least four of the subfamilies contain members that retain catalytic activity in the presence of O 2 , 10,11 which is in particular a most interesting property with respect to any potential application for photocatalytic hydrogen production. 4 Both the degree and the biophysical/biochemical origin of this O 2 -tolerance differ between the different subgroups. In selected representatives of group 1, O 2 -tolerance is crucially linked to the presence of a novel [4Fe-3S] cluster with unprecedented redox properties. 7,8,[12][13][14][15][16] H 2 sensors belong to the group 2 of hydrogenases and are virtually resistant towards O 2 and CO. This property has been primarily associated with a narrow intramolecular hydrophobic gas channel that impedes O 2 diffusion to the [NiFe] site. [17][18][19] The facultatively lithoautotrophic b-proteobacterium Ralstonia eutropha (Re) H16 harbors four different O 2 -tolerant [NiFe] hydrogenases: a membrane-bound hydrogenase (MBH), a NAD + -reducing soluble hydrogenase (SH), an actinobacterialtype hydrogenase (AH) and a regulatory hydrogenase (RH). 10,11,[19][20][21][22][23][24][25][26] The RH is a group 2b protein and acts as H 2 sensor in a signal transduction pathway, which ensures that the energy-generating hydrogenases, MBH and SH, are only synthesized if their substrate H 2 , is available. 9,24 The H 2 -sensing unit is composed of two RH heterodimers that are tightly connected to a histidine protein kinase homotetramer (Fig. 1). 22 A single RH heterodimer consists of the large, active sitecontaining subunit, HoxC, and the small subunit, HoxB, that harbors three iron-sulfur clusters (FeS) of so far unknown structure. Amino acid sequence analysis predicts the presence of three [4Fe-4S] clusters. 24 This is in contrast to prototypical group 1 hydrogenases, which usually possess two [4Fe-4S] clusters and one [3Fe-4S] cluster, 2,4,17,27 but is similar to the group 1 subclass of [NiFeSe]-hydrogenases. 28,29 Though the molecular structure of regulatory hydrogenases is unknown, a number of biochemical and biophysical investigations by EPR, 20,21 FTIR, 30 Resonance Raman 31 and X-ray Absorption Spectroscopy (XAS) 22,26 have provided valuable information on the cofactor structure, composition and function of the RH from R. eutropha. Due to the high complexity of the system, in vitro studies were mainly carried out on a truncated version of the protein (designated as RH stop ), which allows for both isolation by affinity chromatography and formation of a single heterodimer (HoxBC), the tertiary structure of which is similar to that of classical hydrogenases. 5,22 The as-isolated (oxidized) RH stop exhibits no EPR signals related to paramagnetic Ni or FeS clusters. In the most oxidized state, the [NiFe] active site resides in the EPR-silent, catalytically competent Ni-SI state, 20 which contrasts the case of "standard" [NiFe]-hydrogenases that exhibit a superposition of paramagnetic states, known as Ni-A and Ni-B, reecting oxidative modications at the active site. [32][33][34] The absence of the Ni-A and Ni-B states in RH, both of which require reductive reactivation, is fully consistent with the sensory character of the protein that needs to react immediately in the presence of H 2 . 22,24,35,36 Incubation of RH with H 2 in absence of articial electron acceptors affords the paramagnetic Ni-C state characterized by a hydride ligand in the bridging position between the Ni III and Fe II ions. 20 This state appears to be a thermodynamic "bottleneck". Even prolonged incubation with H 2 does not detectably yield the otherwise occurring most reduced Ni-R state(s), typically observed in standard hydrogenases. Ni-SI and Ni-C are therefore the only intermediates to accumulate in signicant amounts in RH. 20 The hydride ligand in Ni-C is photolabile at cryogenic temperatures due to its presumed translocation as a proton. This yields two well-dened paramagnetic forms, termed Ni-L 1 and Ni-L 2 , where the Ni ion resides formally in the Ni 1+ state. 28,37,38 Paramagnetic states of the FeS clusters in RH so far have not been amenable to characterization by EPR spectroscopy. Therefore, the identity, electronic properties, and their possible role in the O 2 -tolerance of this enzyme have remained elusive. This prompted us to employ Mössbauer spectroscopy, which has proven to be a powerful tool for the characterization of the electronic structure of iron-containing cofactors in general and in [Fe]-, 1 [FeFe]-, 39,40 [NiFe]-13,41-44 and [NiFeSe]-hydrogenase, in particular. 45,46 The study was complemented by EPR and FTIR spectroscopy. The Mössbauer spectra of RH can be explained by the presence of three [4Fe-4S] clusters, at least two of them with distinct redox properties. Strong reducing agents were employed to ascertain complete reduction of all three FeS clusters, which for the rst time also afforded detectable EPR signals of the reduced FeS centers. Moreover, the low-spin Fe II site in the [NiFe] center could be unambiguously identied, based on its unique properties which differ from those of other [NiFe] enzymes studied so far. The information obtained by Mössbauer spectroscopy was complemented by a 57 Fe HYS-CORE characterization of the [NiFe] center in the Ni-C state. Overall, the present study contains a comprehensive characterization of the metallocofactors in RH and sheds light onto their structural and electronic properties, which were so far enigmatic. On the basis of the present results the mechanism of the regulatory hydrogenase is revisited and discussed with regard to its physiological functionsensing of H 2 .

Reagents
All chemicals were of analytical grade and used without further purication. 57 Fe was purchased as metallic iron from Chemotrade Handelsgesellscha, Düsseldorf (Germany) and subsequently dissolved in HCl to yield FeCl 3 for preparing the fermentation media.

Protein preparation
For the investigation of the RH by Mössbauer, EPR and FTIR spectroscopy, we have employed the so-called RH stop protein, which represents a simplied version of the dimeric wild-type RH, comprising only one functional moiety of the dimer (Fig. 1). 22,26 Because of the replacement of a 50 amino acid extension of the RH small subunit HoxB with a Strep-tagII affinity peptide, formation of the double-dimeric RH is prevented and instead the heterodimeric RH stop protein can be obtained and further puried by affinity chromatography. The RH stop derivative was puried from the transconjugant strain R. eutropha HF574 (pGE567), 19 which was cultivated under hydrogenase-depressing conditions in the presence of 57 FeCl 3 as the sole source of iron. The RH was then puried by Strep-Tactin affinity chromatography as described previously. 19 The elution fractions were pooled and the RH protein was subsequently concentrated to a nal concentration of ca. 400 mM, suitable for spectroscopic analysis.
Reduction with H 2 was performed by ushing the sample with hydrated H 2 gas (99.99%, Air Liquide) for 60 minutes in an anaerobic glovebox (Coy). From these preparations, samples were obtained for parallel EPR, IR and Mössbauer experiments, respectively.
Reduction with reducing agents was carried out anaerobically through addition of either Ti 3+ citrate or sodium dithionite (Sigma-Aldrich, Germany) to a nal concentration of 10 mM. Sodium dithionite was freshly prepared in 100 mM Na 2 CO 3 / NaHCO 3 buffer (pH 10). Ti 3+ citrate was prepared by dissolving the necessary amount of TiCl 3 (anhydrous, Sigma-Aldrich, Germany) in a solution containing ten-fold excess of trisodium citrate. The color of the solution turned immediately from violet to dark orange (nal pH was 7.5). The Ti 3+ stock concentration was 15 mM. In a rst set of experiments, 1.5 equivalents of Ti 3+ were added anaerobically to an RH sample and were allowed to react for 15 min, before dividing the sample in two aliquots and separate freezing for parallel examination by Mössbauer and EPR spectroscopy. In a second set of experiments, four more equivalents of Ti 3+ were added to the previous Mössbauer sample to ensure complete reduction. No signicant changes were observed in the spectra recorded with this sample, and it was therefore further used for eldand temperature-dependent Mössbauer spectroscopic studies.

Mössbauer spectroscopy
Mössbauer spectra were recorded on an alternating constant acceleration spectrometer. The minimum experimental line width was 0.24 mm s À1 (full-width at half maximum). The sample temperature was maintained constant either in an Oxford Variox or an Oxford Mössbauer-Spectromag cryostat. The latter is a split-pair superconducting magnet system for applying elds of up to 8 Tesla to the samples that can be kept at temperatures in the range 1.5-250 K. The eld at the sample is perpendicular to the g beam. The 57 Co/Rh source was positioned at room temperature inside the gap of the magnet in a reentrant bore tube at a distance of about 85 mm from the sample. The eld is zero at this position. All isomer shis are quoted relative to the centroid of the spectrum of metallic airon at 300 K. Mössbauer spectra were simulated employing the usual spin Hamiltonian formalism. 47 Theoretical details are described in the ESI. †

Electron paramagnetic resonance (EPR) spectroscopy
Continuous wave (cw) X-Band measurements were carried out using an X-band Bruker ESP 300E instrument (9.4 GHz, TE 012 resonator) equipped with a helium ow cryostat (Oxford Instruments, ESR910) and an ITC 503 temperature controller. CW S-band (2-4 GHz, loop gap resonator) and CW Q-Band (34 GHz, TE 011 resonator) measurements were carried out using a Bruker ESP 300 spectrometer equipped with an Oxford Instruments helium ow cryostat.
Pulse Q-Band experiments were carried out on a Bruker ELEXSYS E-580 FT EPR Q-band spectrometer equipped with a SuperQ-FT microwave bridge, a home-built slightly overcoupled cylindrical TE 011 resonator, with a construction described by Reijerse et al., 48 and an Oxford CF935 Helium ow cryostat. The solid-state microwave amplier in this bridge produces a power of 3 W at the resonator. Pulsed X-band EPR measurements were performed on a Bruker ELEXSYS E-580 Xband spectrometer with a SuperX-FT microwave bridge and a CF935 Oxford ow cryostat in the temperature range 10-20 K. An over-coupled Bruker ER 4118X-MD4-W1 dielectric ring ENDOR resonator was used for these experiments. The MW pulses were amplied by using an Applied Systems Engineering Traveling Wave Tube (TWT) amplier (1 kW). Q-band and Xband HYSCORE spectra were recorded using the standard Bruker data acquisition soware. W-band measurements were carried out on a Bruker ELEXSYS E-680 spectrometer, using the commercial W-band ENDOR probehead (Bruker).
Spin quantitation was carried out by using as a standard 1.0 mM Cu 2+ chloride in water (1 mM Cu 2+ , 2 M NaClO 4 , 10 mM HCl), measured under non-saturating conditions. Simulations of the cw and pulse pseudo-modulated spectra, and HYSCORE spectra were done using home-written simulation programs implemented in the Kazan soware (Dr Alexey Silakov, The Pennsylvania State University, and Prof. Boris Epel, University of Chicago) that employs MATLAB (Mathworks) as an interface. Details are described in the ESI † or in previous publications of our group. 49

Fourier transform infrared (FTIR) spectroscopy
Measurements were performed on a Bruker IFS 66v/s FTIR spectrometer with 2 cm À1 resolution. The detector was a photovoltaic mercury cadmium telluride (MCT) element. Room temperature measurements were done with a liquid cell that consists of two CaF 2 windows (3.5 cm diameter) separated by a 0.1 mm Teon spacer. The temperature was regulated by a thermostatic bath. The soware for data recording consisted of the OPUS package (Bruker Optics). Analysis and further processing were performed with home-built routines written in MATLAB version 6.5 (Mathworks).

Results
The as-isolated, oxidized state of RH stop The FTIR spectrum of the as-isolated RH stop (Fig. 2) 26,50 exhibits three characteristic signals from the diatomic ligands at the Fe in the [NiFe] center that can be assigned to the Ni-SI state, with a strong band at 1943 cm À1 (corresponding to the stretching vibration of CO) and two weaker bands at 2081 cm À1 , 2072 cm À1 (corresponding to the vibrationally coupled CN À stretches). 2,51 A small band at 1961 cm À1 coincides with the CO stretch of the Ni-C state, the concentration of which however, is presumably too low for the detection of the corresponding Ni-related EPR signals. Also the concentration of another minor component, causing a weak band at 1932 cm À1 appears to be undetectably low by EPR. The EPR spectrum of the as-isolated enzyme shows only faint signals at g ¼ 4.23 and g av $ 2.01, which we assign to adventitiously bound high-spin Fe III (S ¼ 5/2) and [3Fe-4S] 1+ clusters (S ¼ 1/2), respectively. Since in particular the [3Fe-4S] 1+ clusters are present to only substoichiometric amounts (their spin quantication is in agreement with the abundance of only 0.3 clusters per protein dimer, as estimated below from the Mössbauer spectra), both EPR-active species appear to be oxidation-derived degradation products. This is a common observation for hydrogenases, because of the partial accessibility of O 2 to the FeS clusters (the distal FeS cluster is particularly exposed to the protein surface). 52 The Mössbauer spectra of the as-isolated 57 Fe labeled RH stop recorded at various magnetic elds are shown in Fig. 3. In these spectra all Fe containing species are detected irrespective of their redox and spin states. The dominant contribution (corresponding to $79% of the total iron content, blue lines) originates from a diamagnetic species with intermediately strong isomer shi and quadrupole coupling constants ( Table 1). The values are typical of iron in the high spin state with quasitetrahedral sulfur coordination and mixed oxidation state (+2.5). Together with the diamagnetic behavior the features are characteristic for valence-delocalized [4Fe-4S] 2+ clusters with four essentially indistinguishable Fe 2.5+ sites and total spin S ¼ 0. 53 Other FeS clusters with Fe 2.5+ sites would be either paramagnetic or have additional iron sites with different spectroscopically detectable oxidation states. Similar signals have been found in all [NiFe] hydrogenases studied by Mössbauer spectroscopy to date (Table S1 †). [41][42][43][44][45][46] A well-resolved shoulder observed in the zero-eld spectra at ca. À0.8 mm s À1 can be ascribed to another, rather particular subspectrum with unusually low isomer shi and large quadrupole splitting (d ¼ 0.10 mm s À1 , DE Q ¼ 1.60 mm s À1 , green lines), accounting for $7% of the total Fe content. Because the isomer shi is clearly below the range known for physiological FeS sites (and their degradation products in proteins) 54-56 this subspectrum can be associated with the low-spin Fe II ion of the [NiFe] active site. As expected for the Ni-SI state, the Mössbauer spectrum reveals diamagnetic behavior, as can be seen, e.g., from the eld dependence of the low-energy line, yielding a shoulder throughout the series of spectra shown in Fig. 3. Fig. 2 FTIR spectra of the RH from R. eutropha in its as-isolated (green trace), H 2 -reduced (red trace) and Ti 3+ citrate-reduced (blue trace) states, which demonstrate the presence of 82% Ni-SI, 100% Ni-C and 85% Ni-C state, respectively. In the Ti 3+ -citrate treated sample the band at 1948 cm À1 , may be correlated with a substoichiometric accumulation of Ni-R, an assignment that is however precluded, because the expected bands of the corresponding CN ligands of Ni-R are absent. The absence of any additional Ni signals in EPR, suggest that the 1948 cm À1 band corresponds either to a different conformation of Ni-C or to an EPR-silent state. Moreover, two weak paramagnetic components are present in the spectra, but they are of only minor intensity ($7% of the total iron content each, Tables 1 and S1 †). These can be attributed to the adventitiously bound high-spin Fe III and a contamination with [3Fe-4S] + clusters, respectively, which combined with their correspondingly weak EPR signals at g ¼ 4.3 and g av $2.01, suggest them to be oxidative degradation products (vide supra).
In summary, the relative intensities of the Mössbauer components correspond to a model with 13 genuine iron sites in RH stop , arising from 12 Fe in three [4Fe-4S] clusters and one iron in the [NiFe] active site (one Fe of 13 would correspond to 7.7% relative intensity, we observed $7% for the [NiFe] species). In practice minor decomposition of some cubanes can be inferred from the presence of nuisance Fe III and [3Fe-4S] clusters, which changes the percentages slightly, revealing ca. 0.3 [3Fe-4S] + clusters per heterodimer (see ESI †).
The H 2 -reduced state of the RH stop FTIR. The FTIR spectrum of the H 2 -reduced RH stop exhibits three bands that are characteristic of the Ni-C state, 19 with a CO stretching vibration at 1961 cm À1 and two CN-related stretches at 2083 and 2071 cm À1 , respectively. This spectrum demonstrates that exposure to H 2 leads to a well-dened and homogeneous reduction of the [NiFe] center to the Ni-C state (Fig. 2).
Mössbauer. The Mössbauer spectra of the H 2 -reduced RH stop recorded at different magnetic elds and temperatures are shown in Fig. 4 The results provide profound evidence for the reduction of tetranuclear iron-sulfur cofactors in the regulatory hydrogenase, a conclusion that has been drawn previously on the basis of UV/VIS and XAS spectroscopy, albeit invoking a different FeS cofactor composition. 19 The intensity ratio of the Mössbauer subspectra demonstrates that maximally two of the three clusters in HoxB can be reduced by molecular H 2 .
Unique, model-free simulations of the complex magnetic Mössbauer spectra shown in Fig. 4 are virtually impossible in all details because of the large number of overlapping subspectra, which, even with the variety of eld conditions and temperatures, cannot be fully disentangled. We were therefore aiming for 'generic' solutions, which allowed us to characterize the general types and numbers of different FeS clusters, and iron in the catalytic site. To this end, a set of Mössbauer parameters could be obtained (  43,45,46,53 These yielded the nice global simulation of the spectra shown in Fig. 4, but we cannot readily exclude ambiguities for each and every value of magnetic hyperne coupling constants or individual isomer shis and quadrupole splittings. However, the general assignments to oxidized and reduced [4Fe-4S] clusters are clear and the result excludes major deviations from the typical properties of cubane clusters. The subspectrum of low-spin Fe II in the catalytic center in the Ni-C state shows a similar low isomer shi (0.07 mm s À1 ) as found above for the oxidized Ni-SI state (0.10 mm s À1 ). This observation is in agreement with the active site iron persisting in its low-spin Fe II state upon reduction of the RH with H 2 , which is supported by the diamagnetic behavior of the component in the magnetic Mössbauer spectra, revealing very small spin density on the Fe of the active site in the Ni-C state (vide infra). However, the quadrupole splitting (0.69 mm s À1 ) of the Fe II ion in Ni-C is much lower than that in Ni-SI (1.60 mm s À1 ), revealing a signicant change in the ligand environment. a Obtained from the simulations of the spectra at different applied elds perpendicular to the g beam at 4.2 K. Isomer shis (d) and quadrupole splittings (DE Q ) that are shown in parenthesis correspond to 160 K (these parameters were also used to simulate the spectrum at 80 K). h is the asymmetry parameter, A i are the hyperne tensor components and G is the line width. g ¼ 2.0 was used for all the species. For the Fe III high spin species S ¼ 5/2, D ¼ 1.0 cm À1 and E/D ¼ 0.33 were assumed. The slow relaxation limit was assumed at 4.2 K and the fast relaxation limit was used at 160 and 80 K. Errors are estimated as follows: A minor contribution from [3Fe-4S] clusters, as seen above by EPR for the oxidized protein, could not be unambiguously derived from the Mössbauer spectra of the reduced RH. Typical subspectra for the corresponding 1+ and 0 oxidation states of cubane trinuclear FeS centers with spins S ¼ 1/2 and S ¼ 2, respectively, were introduced in the simulations, but these did not signicantly affect the quality of the ts or the distribution of the native FeS clusters (Fig. S2 †), and therefore have not been further considered in our analysis. Presumably, magnetic broadening or heterogeneity in structure and environment of the non-physiological 3Fe clusters hampered their detection in the complex spectra of H 2 -reduced RH stop . Eventually, also reconstitution of [4Fe-4S] clusters from [3Fe-4S] species under reductive conditions in the presence of adventitious Fe II ions cannot be completely ruled out as a rationale for the apparent absence of such signals in the spectra. 58 Reduction of RH stop with strong reducing agents FTIR. The FTIR spectrum of the Ti 3+ citrate-reduced sample (Fig. 2) shows the typical Ni-C state signals. 26,50 Even under these strongly reducing conditions (E m ¼À510 mV at pH 7.5), the oneelectron more reduced Ni-R state was not generated in detectable amounts, which is in agreement with previous observations. 23 The Ni-R state, which is believed to be the direct outcome of the reaction of the Ni-SI state with H 2 , 59-62 was accumulated neither with H 2 (physiological) nor with more potent chemically reducing agents (non-physiological). Thus, Ni-C represents the only tractable reduced state of RH stop , which might be rationalized by the explicit redox equilibria of the RH cofactors (discussed below). A small band at 1948 cm À1 , may indicate a substoichiometric generation of the Ni-R state.
Mössbauer. The Mössbauer spectrum of the Ti 3+ citratereduced RH stop protein recorded at 4.2 K with 1 T applied magnetic eld is shown in Fig. 5 (top), and a series of eldand temperature-dependent Mössbauer spectra is displayed in Fig. S7. † Employing the same approach and starting parameters as for the simulation of the spectra of the H 2 -reduced RH stop (Tables 2 and S2 †), analysis of the spectra of the Ti 3+ citratereduced samples reveals almost full reduction of all three [4Fe-4S] clusters to the 1+ state ($82% of the Fe in the sample, 2.67 clusters per protein). Similar results were obtained with samples reduced with sodium dithionite at pH 10 (data not shown). It should be noted that super-reduction of the cubanes to the [4Fe-4S] 0 state, which would result in high isomer shi values and large quadrupole splittings exhibited by typical Fe II S 4 sites, was not observed. [63][64][65][66] The diamagnetic Mössbauer spectrum of the low-spin Fe II ion in the [NiFe] center was identied reasonably well again, in particular in the spectra recorded at low temperature and small elds ( Fig. 5 and S7 †). In detail, the parameters could be taken the same as those determined for the H 2 -reduced samples (d ¼ 0.10 mm s À1 , DE Q ¼ 0.69 mm s À1 ), which is supported by the FTIR spectra, demonstrating that the majority of the [NiFe] sites remains in the Ni-C state. Non-cluster-related high-spin Fe II was also present as an impurity and amounted to $8%.
EPR spectra of reduced RH stop 'Unsplit' Ni-C signal of the H 2 -reduced RH stop . The X-band cw spectrum of the H 2 -reduced RH stop shows at 5 K a single S ¼ 1/2 signal with principal g-values (2.199, 2.140, 2.015) that are characteristic of the [NiFe] center in the Ni-C state (Fig. S2 †), and which have been explained formally by adopting a conguration of a low-spin Fe II (S ¼ 0) and a Ni III (S ¼ 1/2) center, with the singly occupied orbital exhibiting mainly d z 2 character. 20,21 Remarkably, no indications from FeS resonances were found in the cw EPR spectra, although the majority of clusters were in the reduced state with spin S ¼ 1/2 (see below). In addition, line broadening in the Ni-C spectra due to 57 Fe hyperne coupling   Fig. S4 † discloses a 'splitting' in the g y component of Ni-C. Because this splitting originates neither from 57 Fe hyperne interactions (spectrum of the 56 Fe sample is identical) nor from magnetic interactions with the proximal [4Fe-4S] 1+ cluster (splitting is eld-independent), it is rather consistent with a conformational heterogeneity of Ni-C, which appeared to slightly vary between different sample preparations. This scenario is supported also by spectra obtained at higher microwave frequencies (e.g., W-Band, Fig. S4 †). 57 Fe hyperne coupling in the Ni-C related EPR signal of H 2reduced RH stop . The weak magnetic hyperne coupling of the low-spin Fe II site (S ¼ 0) in the Ni-C state, which is caused by (covalent) spin density delocalization from the Ni III ion, could not be resolved by Mössbauer spectroscopy. Such weak interactions can be probed and studied by advanced pulsed EPR techniques. 32,67 Indeed, 2-and 3-pulse ESEEM experiments at Qand X-band frequencies showed intense modulations, originating from the electron spin interaction with the 57 Fe nuclear spin (I ¼ 1/2), which were absent in the 56 Fe-labeled samples.    Analyses of the orientation-selective HYSCORE spectra recorded at Q-band ( Fig. 6 and S5 †) and X-band (Fig. S6 †) frequencies yielded principal values A Fe ¼ (5.0, 1.1, À0.5) (AE0.5) MHz for the 57 Fe hyperne coupling tensor. Additional signals due to 1 H and weakly coupled 14 N nuclei from the protein backbone were also detected, in particular in the X-band HYSCORE spectra ( Fig. S6 †), but not further analyzed. 'Split' Ni-C EPR signal in the chemically reduced RH stop . Incubation of the RH stop with either Ti 3+ citrate or dithionite resulted in the same Ni-C EPR spectrum without any discernable differences in the overall signal intensity. This is in contrast to canonical hydrogenases, in which the Ni-C signal diminishes upon prolonged activation with H 2 as well as electrochemically at reduction potentials similar to those of the two strong reductants. 4,51 The spectra of the Ti 3+ citrate-and dithionite-treated RH stop recorded at 40 K are shown in Fig. S8. † Lowering the temperature to #10 K led to the appearance of a 'split signal' (see also Fig. 7). This 'splitting' is indicative of magnetic interactions between the [NiFe] site and another paramagnetic species, which on the basis of its spatial localization (distance # 14Å) has to be the reduced proximal [4Fe-4S] cluster in the 1+ state. Similar 'interaction' spectra have been reported for all [NiFe]hydrogenases studied so far, with the exception of the RH-type proteins. 20 Hence, the 'magnetic ngerprint' found here is the rst direct evidence of such a spin-spin interaction between adjacent metallocenters in a regulatory hydrogenase.
The Mössbauer results suggest a certain heterogeneity with respect to the degree of reduction of the FeS clusters of the chemically reduced samples of RH stop , a fact that is reected also in a mixture of 'split' and 'unsplit' Ni-C signals contributing to the EPR spectra. It is difficult to ascertain the precise amount of 'unsplit' Ni-C species in the spectra of the chemically reduced protein. However, by using the Ni-C signal of H 2reduced RH stop as a reference for an almost pure 'unsplit' Ni-C spectrum, an upper threshold of less than 30% can be estimated for the fraction of magnetically uncoupled Ni-C in the dithionite-reduced sample ( Fig. 7 and S8 †). This estimation is in full agreement with the corresponding Mössbauer results on samples prepared under identical conditions, which demonstrated that 2.7 clusters (of three) are in the 1+ reduced state. With this presumption, the X-and Q-band Ni-C-related EPR spectra in the dithionite-reduced RH stop could be well simulated to obtain a reasonable solution of the magnitude and orientation dependence of the electron spin-spin interaction between the participating paramagnetic centers. The spectral features can be reproduced by a simple three-spin model accounting for magnetic interaction between the [NiFe] site in the Ni-C state and the proximal [4Fe-4S] 1+ cluster as well as between the proximal and the medial [4Fe-4S] 1+ cluster. The interaction between the [NiFe] site and the medial cluster was assumed to be negligible and therefore not considered. The g-values of the [4Fe-4S] 1+ clusters were chosen similar to those of other typical tetranuclear clusters, Fig. 6 Q-band 57 Fe HYSCORE spectra of the H 2 -reduced forms of 57 Fe-enriched RH stop at magnetic fields corresponding to two principal g-components of the Ni-C spectrum. Experimental conditions: mw frequency ¼ 33.90 GHz, T ¼ 20 K, p/2 ¼ 40 ns, s ¼ 300 ns, shot repetition time ¼ 1 ms. Simulation parameters are given in Table 3. Fig. 7 X-band CW EPR spectra of the dithionite-reduced RH stop at two different temperatures, 5 K and 80 K, respectively. The experimental signals (black lines) and the simulations (red lines) correspond to the spin-coupled Ni-C sites since the contribution of the uncoupled Ni-C signal ('unsplit'), estimated to be <30% in the spectra, has been subtracted from the spectra. The part of the simulated spectrum that corresponds to the [4Fe-4S] 1+ cluster signals is depicted with a dashed line (this was experimentally not detectable presumably due to g-strain effects and relaxation enhancement caused by spin-spin interactions). Because of the fast electronic relaxation of the FeS cluster signals at 80 K, their signals are not detectable and thus the 'fingerprint' of their magnetic interaction on the [NiFe] is absent from the NiC-related EPR signals. Experimental conditions: mw frequency 9.47 GHz, modulation amplitude 0.7 mT, mw power 0.2 mW.
constrained, however, to reproduce the respective Q-band EPR spectra (Fig. 8, Table S2 †). The interaction between the [NiFe] center and the proximal cluster was described by an anisotropic J tensor (H electron spinÀspin coupling ¼ ÀS 1 $J$S 2 ) with an isotropic component J iso ¼ 31.7 Â 10 À4 cm À1 (95 MHz) and an anisotropic component J dip ¼ [À60, 110, À50] MHz. The rst is the trace of J and accounts for the through-bonds interaction of the paramagnetic centers, whereas the second part is traceless and accounts for through-space dipole interaction. Using the point-dipole approximation, J dip is consistent with the [NiFe] site residing at a distance of r ¼ 9.8Å from the proximal [4Fe-4S] 1+ cluster with spherical polar coordinates (q, 4) ¼ (80 , 80 ) relative to the magnetic axes given by the two g-matrices, which were taken collinear. 68 The interaction between the two [4Fe-4S] 1+ centers was in good approximation considered to be isotropic with J iso ¼ 43.4 Â 10 À4 cm À1 (130 MHz). The signals of the [4Fe-4S] 1+ clusters are not clearly detectable (presumably due to g-strain effects), which is similar to the case of standard O 2 -sensitive group 1 hydrogenases, but unlike [NiFeSe] hydrogenases. [68][69][70] EPR spectra of the FeS clusters. Whereas the 'split' Ni-C EPR signal in chemically reduced RH stop samples is readily observable by using conventional cw spectroscopy with eld modulation, the signals of the reduced [4Fe-4S] 1+ clusters are hardly detectable. Even measurements with dispersion-detection at temperatures as low as 2 K and high microwave power (providing rapid-passage conditions) were unsuccessful. Probably spin-spin interactions between the clusters, g-tensor anisotropy and g-strain effects broaden the signals beyond recognition. However, eld-swept 2-pulse echo-detected Q-band spectra of H 2 -and dithionite-reduced RH stop showed for the rst time two distinct signals that can be assigned to Ni-C and the [4Fe-4S] 1+ clusters (Fig. 8), respectively. The assignment is based on the typical low g-values of FeS clusters in conjunction with fast spin relaxation (the broad quasi-absorption signals are detectable only at lower temperatures, Fig. S3 †). The intensity of the FeS subspectrum is signicantly higher for the dithionitereduced samples than that for the H 2 -reduced samples, given relative to the resolved Ni-C contribution, appearing at higher g values. This is in agreement with the Mössbauer results that demonstrated reduction of almost all three [4Fe-4S] clusters with dithionite. The region of the spectra in which the FeS clusters are expected could not be explored in the Ti 3+ -reduced samples, because the strong signal from residual reductant overlaps with the [4Fe-4S] 1+ signals (Fig. S8 †), hindering their detection and analysis. 21,24

Discussion
The RH active site The low-spin Fe II site in the catalytic center of [NiFe] hydrogenases has not been discerned in most Mössbauer spectroscopic studies to date, because its spectrum is usually masked by the iron-sulfur cluster signals that impede site-resolution and characterization. In this study, however, we obtained wellresolved signals attributed to the corresponding low-spin Fe II center in RH stop . This site exhibits a low isomer shi characteristic of low-spin Fe II centers, because shielding of the 57 Fe nucleus is less effective compared to the case of high-spin Fe II centers. 71,72 Furthermore, the quadrupole splitting parameter varies between the two different redox states. Upon transition from the oxidized Ni-SI to the one-electron more reduced Ni-C state, the quadrupole splitting DE Q decreases from 1.60 mm s À1 to 0.69 mm s À1 , reecting signicant changes in the ligand environment of the Fe II ion, whereas the isomer shi remains almost invariant. Only minor differences in the isomer shi of the Fe II ion were observed for the Ni-SI and Ni-C states, which suggest that there is no marked effect on the average bond lengths or the strength of the p-/s-bond interaction with the diatomic Fe ligands. The quadrupole splitting, however, was signicantly different for the two redox states, which might result from a change in the direct coordination environment of the Fe II . In fact, a vacant bridging position between the two heterometals has been proposed for the Ni-SI state, whereas in the Ni-C state, the presence of a bridging hydride between the Fe and Ni ions has been demonstrated. 20,37 The observed decrease in the quadrupole splitting for the RH can therefore be correlated with the change in the Fe coordination number from Ni-SI (5-coordinate) to Ni-C (6-coordinate), the latter of which reects a more symmetric environment. This signicant difference of the quadrupole splitting between the Ni-SI and Ni-C states might not be a unique feature of H 2 -sensors and is expected to be observed also in other [NiFe]-hydrogenases. The unprecedented resolution of the low-spin Fe II site (at least) in the as-isolated RH is presumably due to the almost stoichiometric presence of the Ni-SI state and the concomitant absence X-and Q-band 57 Fe HYSCORE experiments were carried out to characterize the electron spin density at the Fe center, from which detailed information about the electronic structure of the active site can be obtained. The anisotropic hyperne tensor has an isotropic component of 1.9 MHz, which is two-fold larger than the 57 Fe isotropic hyperne coupling constants determined for the Ni-B state of the D. vulgaris hydrogenase (A iso ¼ 0.8 MHz) 33 and the Ni-A state of the D. gigas enzyme (A iso ¼ 1.0 MHz). 73 Because the isotropic hyperne constant is proportional to the electron spin density at the nucleus, all data indicate a higher (although very small) electron spin density at the Fe center in the Ni-C state compared to the Ni-A and Ni-B states. Furthermore, the data are consistent with a low-spin Fe II center in all states, with most of the spin density located at the Ni ion, as has been previously shown by 61 Ni enrichment and subsequent determination of the respective 61 Ni hyperne tensor in the standard [NiFe] hydrogenase from D. vulgaris. 4,74,75 These results are in full agreement with density functional theory calculations of the 61 Ni hyperne coupling constants. 76 In the present as well as in previous studies of RH, the fully reduced Ni-R state could not be generated in signicant amounts, neither by incubating the enzyme with H 2 nor with strong reducing agents. In the FTIR spectra of the chemically reduced RH, a small intensity band at 1948 cm À1 is present and could represent the CO stretching vibration characteristic of the reduced Ni-R state. However, there is no evidence for the corresponding bands of the CN ligands, which for all reported Ni-R forms are typically shied at least by 10 cm À1 to lower wavenumbers with respect to those of Ni-C. 51 Nevertheless, we cannot exclude that Ni-R accumulated in small amounts, but the CN bands have been too low in intensity to be observed.

The FeS clusters
Amino acid sequence analysis of the RH small subunit HoxB revealed 12 conserved cysteine residues to be the likely ligands of three predicted tetranuclear clusters. 24 In previous studies, however, the iron-sulfur cofactors could not be successfully detected by EPR under any conditions. 20,21,23,50 The unprecedented O 2 -tolerance (or rather insensitivity) of the RH protein and the limited spectroscopic accessibility of the iron-sulfur clusters invoked a series of proposals for the chemical nature and role of these metallocofactors in the function and O 2resistance of H 2 -sensing hydrogenases. In addition, previous XAS results suggested the presence of [2Fe-2S] clusters and an unusual [4Fe-3S-3O] cluster in the RH wild-type protein. 23,26 However, these assumptions are not in line with our Mössbauer and EPR results, which neither showed signals attributable to [2Fe-2S] clusters 55,56,77 nor to a [4Fe-4S] cluster with unusual oxygen ligation at one or more of the Fe sites. In the latter case, the more ionic oxygen ligands are expected to cause a greater electric eld gradient asymmetry around that Fe site and result in a large quadrupole splitting, which, however, was not observed in our study. The Mössbauer and EPR results obtained for the RH stop protein in combination with the presence of twelve conserved cysteine residues arranged in characteristic FeS cluster binding motifs are consistent with the presence of three [4Fe-4S] clusters in H 2 -sensing hydrogenases and afford for the rst time the characterization of their electronic properties, which were found to be similar to those of classical tetranuclear cubanes. Without a crystal structure of the RH, however, unusual structures of any of the tetranuclear cubanes cannot be readily excluded.
In the as-isolated RH stop , all three clusters reside in their oxidized, diamagnetic [4Fe-4S] 2+ form. Upon reduction with H 2 , approximately two (1.8) of the clusters are reduced to the 1+ state, whereas stronger (non-physiological) reducing agents, i.e. Ti 3+ citrate and sodium dithionite, afford almost complete reduction of all three clusters (2.7). The expected EPR signals of the reduced [4Fe-4S] 1+ clusters in both the H 2 and dithionite/ Ti 3+ citrate-treated samples are hardly detectable, which may be a result of g-strain effects and the inter-cluster spin-spin interactions, as has been previously observed for group 1 hydrogenases. 43,68 On the basis of the amino acid sequence, two of the three [4Fe-4S] clusters in the RH are coordinated by four cysteinederived thiolate ligands, while the 'distal' cluster is coordinated by one His and three Cys residues. Thus, at least with respect to the rst coordination sphere, the proximal and the distal clusters are expected to have similar electronic properties as their counterparts in the mostly O 2 -sensitive [NiFe]-hydrogenases. However, compared to other [NiFe]-hydrogenases, including the O 2tolerant membrane-bound enzymes, it is rather unusual that H 2 incubation does not lead to detectable accumulation of the reduced state of the proximal [4Fe-4S] cluster. On the basis of our observations, the cluster remains largely oxidized under steady state conditions, which is also supported by the absence of a magnetic splitting in the Ni-C-derived EPR signal generated upon H 2 incubation. The geometric arrangement of the clusters relative to the active site dictates that the proximal cluster is the rst FeS cofactor to become reduced upon H 2 oxidation. Thus, the present results may support a fast discharging of the proximal

Interaction between the [NiFe] active site and the FeS clusters
Although the FeS clusters of reduced RH stop are essentially 'silent' in the cw EPR spectra, a spin coupling model of three spins S i ¼ 1/2 including the Ni center, the proximal and the medial clusters (both in their reduced 1+ forms) could be successfully employed to simulate the spectra of Ni-C in chemically (dithionite) reduced RH stop . In order to limit the number of parameters, the fourth paramagnetic species, the distal [4Fe-4S] cluster, was not included into the simulations.
In most hydrogenases studied so far, the long-range spinspin interaction between the reduced proximal cluster and the [NiFe] center can be considered, in good approximation, to be isotropic. In the case of RH stop , however, the spin coupling tensor, J, is quite anisotropic as reected in the Ni-C-related EPR spectra. The largest component of the anisotropic magnetic interaction tensor is 110 MHz and directed along the y-axis of the g-tensor of the Ni-C center, which is consistent with the absence of resolved splittings in the g x and g z components in the rhombic Ni-C spectrum. The isotropic component caused by the through-bond exchange interaction is found to be 95 MHz (assuming a traceless tensor for the anisotropic magnetic dipolar interaction). 78 This value is comparable to the exchange coupling constant of J iso ¼ 120 MHz (40 Â 10 À4 cm À1 ) used for modeling the magnetic interaction between the paramagnetic active site in the Ni-C state and the proximal [4Fe-4S] 1+ cluster in the standard hydrogenase of D. gigas. 68 Recently, the exchange coupling constant between the active site in the Ni-B state and the superoxidized [4Fe-3S] 3+ proximal cluster of the O 2 -tolerant hydrogenase I from Aquifex (A.) aeolicus was also found to lie in this range (J iso ¼ 107 MHz, 36 Â 10 À4 cm À1 ). 13 The isotropic exchange coupling constant between the medial and the proximal clusters in RH stop was determined to be at 130 MHz, which is noticeably larger compared to the corresponding value found in the case of A. aeolicus Hydrogenase I (21 MHz, 7 Â 10 À4 cm À1 ). 13 This difference might provide an additional rationale for the difficulty to detect signals from the reduced [4Fe-4S] clusters in RH preparations due to a correspondingly increased sensitivity of the spectra for strain (distributions of different local electronic changes/interactions) and relaxation enhancement effects.

Structure and function of H 2 -sensor hydrogenases
As described above, the RH harbors three [4Fe-4S] clusters and is in this respect structurally similar to the group 1 [NiFeSe] enzymes, which contain the same iron-sulfur cluster complement. 28,29 The presence of a tetranuclear FeS cofactor in the medial position, which is usually occupied by a "high-potential" [3Fe-4S] center in most other hydrogenases, is consistent with the cytoplasmic localization of RH. In functional terms, the RH protein deviates from most other [NiFe(Se)] hydrogenases in some or all three of the following features: (i) it does not form the oxidized, inactive Ni-A and Ni-B states (ii) it hardly accumulates the most reduced Ni-R form upon reduction with H 2 , and (iii) H 2 does not lead to detectable population of the proximal [4Fe-4S] cluster in the 1+ state. The rst property is most likely related to the RH-specic narrow hydrophobic gas channel that restricts access of molecular oxygen to the active site, 18,19 which apparently renders H 2 sensors O 2 -insensitive rather than O 2 -tolerant. The second and third features are presumably relevant for the catalytic function underlying the H 2 -sensing mechanism of regulatory hydrogenases. Accumulation of the Ni-C state upon incubation of the RH with H 2 and the apparent inability of RH to generate detectable amounts of the Ni-R state under steady-state conditions can be explained by one (or both) of the following thermodynamic considerations. First, the redox potential of the proton-coupled Ni-C to Ni-R transition might be more negative compared to that of standard hydrogenases. Second, the proximal [4Fe-4S] cluster may have an unusual low potential, which is supported by its difficult reducibility through H 2 . Assuming a redox potential that is appreciably more negative than that of the proton-coupled Ni-SI to Ni-C transition, 38 the RH would be trapped in the Ni-C state under steady state conditions. The low-potential Ni-C to Ni-R transition may be promoted by alterations in the rst and second coordination sphere of the RH active site, which might also result in lowering of the pK a of the cysteine thiol-based proton present in the Ni-R state (e.g., due to a more hydrophobic environment). 61 One characteristic of the second coordination sphere of RH proteins is the presence of a glutamine residue instead of a histidine that is conserved in all other [NiFe]-hydrogenases. This glutamine residue might play a role in the redox properties of the active site but seems not to be involved in H 2 sensing. 22 The low redox potential of the proximal [4Fe-4S] cluster might also be caused by the immediate protein environment. In fact, a number of glycine residues in the vicinity of the proximal cluster, which are conserved only in sensory hydrogenases (Fig. S9 †), are situated at the interface of the small and large subunits. It is difficult to predict how the presence of these conserved glycine residues may change the reduction potential, but presumably these may alter the local electrostatic environment.
Remarkably, a low redox potential of the proximal cluster would impede particularly the Ni-C to Ni-SI transition and thereby slowing down the catalytic activity of the enzyme. This is in-line with the low H 2 turnover rate observed for H 2 sensors, which, in turn, prevents energy loss and saves H 2 for the energyconverting hydrogenases. 36

Summary and conclusions
The present study provides new insight into the nature and role of the metal cofactors in H 2 -sensing regulatory hydrogenases. Mössbauer spectroscopy allowed the identication of the lowspin Fe in the [NiFe] active site. As the Mössbauer parameters are sensitive to changes in the ligand eld coordination and symmetry around the active site Fe ion, they can be used to probe structural and electronic changes otherwise inaccessible by other techniques. Only, the quadrupole splitting, but not the isomer shi, is markedly dependent on the redox state (Ni-SI and Ni-C) of the catalytic site. It is considerably smaller in the Ni-C state compared to that of the Ni-SI state, which correlates well with the iron being 6-and 5-fold coordinated respectively.
Mössbauer and EPR spectroscopy allowed for the rst time the identication and characterization of three [4Fe-4S] clusters as constituents of the electron transfer relay in the RH. Their electronic properties are similar to those of conventional tetranuclear FeS centers in low-potential ferredoxins. The extent of their reduction is dependent on the chemical reductant employed, i.e. non-physiological, strong reducing agents led to reduction of essentially all three [4Fe-4S] clusters, while H 2 afforded the reduction of only the medial and distal clusters. Upon H 2 treatment, a magnetic 'splitting' in the EPR signal of the active site-related Ni-C state, which is typical for most [NiFe]hydrogenases, was not observed for the RH. The splitting was detectable, however, when the enzyme was treated with dithionite and identied the proximal cluster as the last cofactor being reduced. This observation supports a low redox potential of the proximal cluster that promotes rapid transfer of electrons to the adjacent [4Fe-4S] 2+ centers, which, on the basis of our data, are supposed to have higher redox potentials.
Concomitantly with the absence of the 'split' Ni-C EPR signal, also the Ni-R state as the most reduced catalytic intermediate is not accumulated upon "physiological" treatment of the RH with H 2 . Even in the presence of strong reducing agents, solely the Ni-C state is observed. Stabilization of the Ni-C state is presumably accomplished through appropriate adjustment of the redox potential of both the Ni-C-to-Ni-R transition (and the concomitant proton transfer) and the proximal [4Fe-4S] cluster. The potential of the latter seems to be too low to support an efficient redox transition from Ni-C to Ni-SI, which coincides with the low H 2 turnover rates of H 2 -sensing regulatory hydrogenases. In summary, the present spectroscopic study unravels the structural basics that have been evolved to convert an energy-generating into a cytoplasmic, H 2 -sensing hydrogenase.