Reduction of unusual iron-sulfur clusters in the H2-sensing regulatory Ni-Fe hydrogenase from Ralstonia eutropha H16.

The regulatory Ni-Fe hydrogenase (RH) from Ralstonia eutropha functions as a hydrogen sensor. The RH consists of the large subunit HoxC housing the Ni-Fe active site and the small subunit HoxB containing Fe-S clusters. The heterolytic cleavage of H(2) at the Ni-Fe active site leads to the EPR-detectable Ni-C state of the protein. For the first time, the simultaneous but EPR-invisible reduction of Fe-S clusters during Ni-C state formation was demonstrated by changes in the UV-visible absorption spectrum as well as by shifts of the iron K-edge from x-ray absorption spectroscopy in the wild-type double dimeric RH(WT) [HoxBC](2) and in a monodimeric derivative designated RH(stop) lacking the C-terminal 55 amino acids of HoxB. According to the analysis of iron EXAFS spectra, the Fe-S clusters of HoxB pronouncedly differ from the three Fe-S clusters in the small subunits of crystallized standard Ni-Fe hydrogenases. Each HoxBC unit of RH(WT) seems to harbor two [2Fe-2S] clusters in addition to a 4Fe species, which may be a [4Fe-3S-3O] cluster. The additional 4Fe-cluster was absent in RH(stop). Reduction of Fe-S clusters in the hydrogen sensor RH may be a first step in the signal transduction chain, which involves complex formation between [HoxBC](2) and tetrameric HoxJ protein, leading to the expression of the energy converting Ni-Fe hydrogenases in R. eutropha.

Ni-Fe hydrogenases represent an important class of metalloenzymes that catalyze the reversible cleavage of molecular hydrogen into electrons and protons (reaction H 2 N 2H ϩ ϩ 2e Ϫ ) (1). The chemolithoautotrophic ␤-proteobacterium bacterium Ralstonia eutropha H16 houses three different Ni-Fe hydrogenases that are physiologically active in the presence of O 2 (2,3). The membrane-bound and soluble NAD-reducing enzymes are involved in energy conversion (4,5).
The regulatory Ni-Fe hydrogenase (RH) 1 belongs to a partic-ularly interesting type of Ni-Fe hydrogenase functioning as hydrogen sensors (6). Hydrogen sensors have also been described in Bradyrhizobium japonicum (7) and Rhodobacter capsulatus (8). Upon the interaction of the RH with molecular hydrogen, a complex signal transduction cascade is initiated that leads to the expression of the energy-converting hydrogenases (9). The RH consists of the large subunit HoxC that harbors the hydrogen-activating Ni-Fe site and the small subunit HoxB that contains iron-sulfur clusters (6). Several unusual properties of the RH (3,10,11) are remarkably different from those of the so-called standard Ni-Fe hydrogenases from inter alia, Desulfovibrio gigas and Allochromatium vinosum (1,12,13). In contrast to the dimeric standard Ni-Fe hydrogenases, the RH forms a double dimer [HoxBC] 2 (Fig. 1A) that is connected to a tetramer of the HoxJ protein (14). The N-terminal input module of HoxJ containing a PAS domain is required for the formation of the RH-HoxJ complex, whereas the C-terminal domain of HoxJ has histidine protein kinase activity (14). The RH cleaves H 2 only at extremely low rates (3,10). In contrast to standard hydrogenases, which can exist in up to nine different redox states (1,13), in the RH only two states of functional relevance have been detected (10). After aerobic isolation the enzyme is in its oxidized state containing Ni(II). This state does not need to be activated but is always ready to bind hydrogen, a prerequisite for the sensor function (10,11). In the presence of H 2 it is rapidly converted to a state revealing a typical EPR-signal, termed Ni-C, due to a Ni(III)-H Ϫ species formed during heterolytic H 2 cleavage (15). In standard Ni-Fe hydrogenases the nickel is coordinated by four conserved cysteine residues. X-ray absorption spectroscopy (XAS) investigations on the RH, however, revealed that nickel may be coordinated by less than four cysteines (11). The iron atom of the RH active site, on the other hand, carries two cyanides and one CO molecule, similar to standard Ni-Fe hydrogenases (3,10).
Although information has become available about the sequence of events that occur at the Ni-Fe active site upon interaction of the RH with H 2 (10,11,15), it is unclear whether electron transfer out of the Ni-Fe site takes place during H 2 cleavage and to where these electrons are transferred. Information on these points is expected to contribute to the understanding of the H 2 -sensing mechanism of the RH-HoxJ complex (14).
In standard Ni-Fe hydrogenases of the D. gigas type, the small subunit contains three Fe-S clusters, two [4Fe-4S] and one [3Fe-4S] (16), which are bound via conserved cysteines and one histidine residue found in all Ni-Fe hydrogenase sequences. During hydrogen turnover these clusters become reduced as detected by EPR spectroscopy (1,17). The RH small subunit HoxB also contains these conserved cysteines. Therefore, it was postulated that it might also harbor three Fe-S clusters (6). EPR investigation of the RH, however, did not show reduced Fe-S clusters when the Ni-C EPR signal was formed under H 2 (3,10,15). It has been proposed that a non-Fe-S cofactor may be involved in electron transfer instead (10).
In this work the double dimeric wild-type RH (RH WT , Fig.  1A) and a derivative denoted as RH stop (Fig. 1B), which forms a monodimer due to mutational truncation of the C terminus of HoxB (14), were compared. It has been suggested that the RH stop may lack the putative non-Fe-S cofactor (14). In both preparations, for the first time reduction of Fe-S clusters in the presence of H 2 was clearly detected in UV-visible spectra and by XAS at the iron K-edge. Support for reduction of a non-Fe-S cofactor was not obtained. Seemingly, the Fe-S clusters of the RH differ from those of standard hydrogenases.

MATERIALS AND METHODS
Bacteria Growth and Enzyme Purification-Strains with the initials HF were derived from R. eutropha H16 (DSM428, ATCC 17699). Large scale cultivation of R. eutropha strains, cell harvesting, cell disruption, and preparation of soluble protein extracts were published before (10,14). RH WT (Fig. 1A) was purified from the RH-overproducing strain R. eutropha HF371(pGE378) as described in Bernhard et al. (10). Starting with 50 g of cells (wet weight) yielded 3.7 mg of RH with a specific activity of 1.6 units/mg of protein. The RH stop protein ( Fig. 1B) was purified from R. eutropha HF574(pGE567) as a Strep-tag II fusion protein. 2 Starting with 18 g of cells yielded 1.5 mg of RH stop with a specific activity of 1.6 units/mg of protein.
The homogeneity of the respective protein preparations was investigated by SDS-PAGE analysis and subsequent Coomassie staining. The amount of impurities was quantified by using the Gelscan Professional V5.1 software (BioSciTec, Frankfurt, Germany). After background subtraction, the sum of the percent differential integrated density of the HoxC-and HoxB-specific bands was correlated to the sum of the percent differential integrated density from the contaminating proteins.
Assays of Hydrogenase Activity-H 2 -oxidizing activity was quantified by an amperometric H 2 uptake assay as in Pierik et al. (3) using a H 2 electrode with methylene blue as an electron acceptor. One unit of H 2 methylene blue oxidoreductase activity was the amount of enzyme that catalyzed the consumption of 1 mol of H 2 /min. Protein concentrations were determined according to the protocol of Bradford (18).
Analysis of Metal Contents-Atomic absorption spectroscopy (AAS) and total reflection x-ray fluorescence analysis (TRXFA) (19) were used for quantification of nickel and iron. For AAS, three aliquots of each RH preparation were solubilized overnight in concentrated HNO 3 (65%, Suprapur, Merck), then diluted with ultrapure water (Millipore) to 1-5% HNO 3 and further diluted for measurements using the transversely heated graphite furnace technique with longitudinal Zeemaneffect background correction on a PerkinElmer Life Sciences Aanalyst 800 spectrometer equipped with an autosampler and WinLab32 soft-ware in the laboratory of Dr. Klaus Irrgang (TU-Berlin). TRXFA for simultaneous nickel and iron quantification was performed on a Picotax spectrometer at Röntec (Berlin, Germany) using 1 l of concentrated and dried protein solutions. Nickel and iron contents were determined against commercial nickel and iron standards (Fluka) in AAS and relative to a gallium standard in TRXFA.
Fourier Transform Infrared Spectroscopy (FTIR)-FTIR measurements were carried out with a Bruker IFS66V/S spectrometer equipped with a photovoltaic MCT detector using a resolution of 2 cm Ϫ1 . The sample compartment was purged with nitrogen. Samples were held in a temperature-controlled (23°C) gas-tight liquid cell (volume ϳ7 l) with CaF 2 windows. FTIR spectra were baseline-corrected using the software available with the spectrometer.
EPR Spectroscopy-EPR measurements were performed in the laboratory of Prof. Robert Bittl (FU-Berlin) on an X-band Bruker Elexsys E580 spectrometer equipped with a SHQE resonator and a helium cryostat (Oxford) (microwave frequency of 9.6 GHz). For additional conditions see the figure legends. From each enzyme spectrum in Fig. 2 the background from sample holder and cavity was subtracted. EPR signals in enzyme samples were quantified by comparison with the integrated intensities of signals from CuSO 4 solutions used as spin standards (20).
UV-Visible Spectroscopy-Purified samples of wild-type RH and RH stop were diluted with 20 mM Tris-HCl (pH 8.0) to a protein concentration of 0.64 mg/ml. UV-visible spectra were recorded on a Cary 1E spectrophotometer (Varian) with a spectral resolution of 0.3 nm. To reduce the RH samples, protein solutions were flushed with hydrogen gas for 1 min. Subsequently, the sample was centrifuged (1 min, 12,000 ϫ g) to remove small amounts of precipitated protein. The clear supernatant was immediately re-transferred to the cuvette, and the UV-visible spectrum of the reduced sample was recorded.
XAS-X-ray absorption spectra at the iron K-edge were collected at beamline D2 of the EMBL at HASYLAB (DESY, Hamburg, Germany). XAS samples contained 10 -20 l of RH solution (protein concentration 0.4 -1 mM). Fluorescence-detected XAS spectra were measured with a 13-element solid-state germanium detector (Canberra) at 20 K as described elsewhere (11). An absolute energy calibration (accuracy, Ϯ0.1 eV) was performed by monitoring the Bragg reflections of a crystal positioned at the end of the beamline (21). 3 scans of ϳ60-min duration were taken on the same spot (4.5 ϫ 1 mm) of the sample. Comparison of the first and third scan revealed that radiation damage to the samples was absent because the iron K-edge energy remained unchanged. Six scans, obtained on two separate spots of the samples, were averaged for each EXAFS spectrum. Spectra were normalized, and EXAFS oscillations were extracted as in Dau et al. (22). The energy scale of iron EXAFS spectra was converted to a k-scale using an E 0 of 7112 eV; E 0 refined to 7120 eV during the EXAFS simulations. Unfiltered k 3weighted spectra were used for least-squares curve-fitting (22) and for calculation of Fourier transforms representing k-values ranging from 1.85 to 13 Å Ϫ1 . The data were multiplied by a fractional cosine window (10% at low and high k-side). For EXAFS simulation, complex backscattering amplitudes were calculated using FEFF 7 (23); the amplitude reduction factor S 0 2 was 0.9. For XAS sample preparation, RH solutions were degassed 3 times under vacuum and subsequently flushed for 10 min with H 2 gas. H 2 -reduced samples were filled under argon or H 2 atmosphere into Kapton-covered acrylic glass sample holders using syringes previously filled with Ar or H 2 . The same samples were used for EPR (before and after XAS) and XAS measurements.

Comparison of the Ni-Fe Active Site in RH WT and RH stop -To
test whether truncation of the C-terminal 55 amino acids of HoxB affected the Ni-Fe site, preparations of RH WT and RH stop were examined by FTIR and EPR measurements. The FTIR spectrum of oxidized RH stop was identical to that of RH WT ( Fig.  2A) in showing an intense peak at 1942 cm Ϫ1 originating from the Fe-CO and two peaks at 2072 cm Ϫ1 and 2081 cm Ϫ1 from the two Fe-CN stretching vibrations as previously reported (3,10). Upon reduction of the samples by H 2 , the (CO) vibration in both spectra shifted to higher frequencies by 18.5 cm Ϫ1 (not shown), whereas the (CN) vibrations remained unaffected as previously observed for RH WT (10).
The redox properties of the Ni-Fe active site of RH stop also turned out to be similar to those of RH WT as detected by EPR (Fig. 2B). Flushing of the RH with H 2 generates the Ni-C state 2 T. Buhrke, O. Lenz, and B. Friedrich, manuscript in preparation. (Ni(III)-H Ϫ ) of the Ni-Fe active site with the hydride in a bridging position between nickel and iron (15). The Ni-C EPR signals ( Fig. 2B) were similar in RH WT and RH stop . Signal quantification revealed that greater than 90% of the nickel was converted to Ni(III) in both preparations after incubation with H 2 for 10 min, implying similarly effective heterolytic hydrogen cleavage. However, EPR signals from singly reduced Fe-S clusters were completely absent in the Ni-C state. Reduced Fe-S clusters were also not detected when temperatures between 6 and 50 K and microwave power variations between 0.01 and 10 milliwatt were applied (data not shown).
In summary, RH WT and RH stop appeared to be similar with respect to the coordination of the iron of the Ni-Fe site both in the oxidized state and in the Ni-C state, which was nearly quantitatively formed in both cases in the presence of hydrogen. Moreover, both preparations showed identical hydrogenase activities (see "Materials and Methods"). The catalytic properties of RH stop were, thus, not affected by the truncation of the C terminus of HoxB.
Determination of Metal Contents-Analysis of the nickel contents by AAS in combination with protein determination yielded on the average about 0.6 mol of nickel in the RH WT and close to 1 mol of nickel in the RH stop (Table I). The relatively low nickel content in RH WT can be explained with impurities in the preparations. SDS-PAGE analysis indicated that preparations of RH WT contained sizable amounts of copurified proteins, whereas preparations of RH stop were homogenous (Fig. 3). The amount of impurities in the RH WT sample was quantified by calculating the intensities of HoxC-and HoxB-specific protein bands and the sum of the band intensities derived from the contaminating proteins. According to this analysis the RH WT sample contained about 16% of impurities, whereas the RH stop sample was pure. Therefore, assuming that each HoxBC unit of RH WT also contained one nickel similar to RH stop , it was estimated from the Fe/Ni ratios (Table I), determined by two independent techniques (AAS and TRXFA), that each HoxBC unit of RH WT contained about 8 -9 iron atoms, in accordance with previous estimates (10). The iron content of RH stop was about 4 -6 and, thus, distinctly lower than that of RH WT .
Based on sequence homologies (see "Discussion") one would expect HoxB to contain three [4Fe-4S] clusters (6). Taking also the iron atom of the Ni-Fe site into account, one HoxBC monodimer might contain 13 Fe/Ni. Thus, the experimentally determined iron content of RH WT was significantly lower than expected. Furthermore, HoxB WT seemed to contain more iron than HoxB stop . The double dimeric RH WT may, thus, comprise iron species that are absent in monodimeric RH stop .
Detection of Fe-S Cluster Reduction by UV-Visible Spectroscopy-The optical absorption spectra of the oxidized RH WT and RH stop showed, in addition to a peak at 280 nm due to the absorption of the aromatic amino acid residues of the protein, a broad shoulder around 410 nm, presumably due to the presence of Fe-S clusters (Fig. 4, solid lines). That both preparations were fully oxidized was apparent from the absence of any Ni-C EPR signal (data not shown). The extinction coefficients were calculated at 390 and at 420 nm and compared with data from the literature (Table II). The ⑀(420 nm) value determined for RH stop is in good agreement with the value that would be expected in the presence of two cysteinyl-coordinated [2Fe-2S] clusters (49). The ⑀(420 nm) of RH WT , on the other hand, is significantly higher, confirming the data from the metal analysis which indicate that RH WT contain additional iron species (see "Determination of Metal Contents"). The ⑀(390 nm) values estimated for RH WT as well as for RH stop were significantly larger than the value of ⑀(390 nm) that was expected if only one [4Fe-4S] cluster was present (Table II).
When the RH was reduced by flushing of samples with H 2 , increased absorption was observed in the whole spectral range (not shown); the curvature of the background (proportional to Ϫ4 ) suggested its origin from a scattering contribution. Because this background could be removed by centrifugation of FIG. 2. A, FTIR spectra of oxidized RH WT and RH stop (normalized on the CO bands). B, normalized Ni-C EPR spectra due to Ni(III)-H Ϫ (15) in the H 2 -reduced samples later used for XAS. EPR conditions: temperature 30 K, microwave power 250 microwatts, modulation amplitude/ frequency 2 millitesla/100 KHz. Quantification of measured Ni-C spectra in combination with nickel determination by AAS revealed that Ͼ90% of the nickel was present as Ni(III) in RH WT and RH stop . a.u., arbitrary units. H 2 -flushed samples, apparently a small portion of the RH protein precipitated, thereby light-scattering aggregates were formed. Aggregation was also observed after flushing with argon, which indicated that it was not specifically caused by H 2 .
Reduction of RH WT by flushing with H 2 and subsequent removal of aggregates by centrifugation yielded a spectrum with clearly decreased absorption between 350 and 600 nm compared with the oxidized state (Fig. 4A, dotted line); the absorption at ϳ280 nm was almost unchanged. Reduction of RH stop with H 2 yielded similar results (Fig. 4B, dotted line). The difference spectra (reduced Ϫ oxidized RH, insets in Fig. 4,  A and B) showed two main minima (bleachings) around 410 and 550 nm, respectively. Whereas bleaching solely around 410 nm has been observed upon reduction of [4Fe-4S] clusters (24,25), two minima as found in the RH spectrum may be attributed to the reduction of [2Fe-2S] clusters (26,27).
Characterization of Fe-S Clusters by XAS-By XAS (22, 28) at the iron K-edge, the iron sites in the RH preparations were selectively studied. Because the coordination of the iron of the Ni-Fe site was similar in RH WT and RH stop both in the oxidized and H 2 -reduced states and this iron remained in its divalent oxidation state, any changes in the iron XAS spectra were expected to be attributable to the putative Fe-S clusters.
The XANES spectra from both RH preparations (Fig. 5) were similar to typical spectra of Fe-S clusters (29 -31). A pronounced shift of the iron K-edge to lower energies by 1.0 -1.2 eV after reduction by H 2 was observed (dotted lines). The shift was by 0.2 eV larger in the RH WT (Table III). A shift by ϳ0.2 eV was also evident in the pre-edge peak of the XANES spectra (Fig. 5,  insets). Such spectral shifts are typical for Fe-S cluster reduction (29,31), implying that such clusters became reduced in RH WT and RH stop upon formation of the Ni-C state.
The maxima of the XANES spectra of the RH WT were larger, and the pre-edge peak magnitudes and areas were smaller than in the RH stop , pointing to an average coordination of iron in the RH WT by less sulfur and more O/N ligands (32). For one-electron reduction of single-Fe(III) compounds with mixed O/N/S ligation of iron, a shift of the K-edge by about Ϫ2.5 eV may be expected (29,31,33). Hence, it was estimated that 3-4 and ϳ2 iron atoms (on basis of 8 and 5 iron atoms in total) became reduced in the RH WT ϩH2 and RH stop ϩH2 , respectively (Table III). In both preparations obviously more than one electron was transferred to iron in the Ni-C state. Moreover, the overall iron coordination was different, and additional iron atoms seemed to become reduced in the RH WT .
The Fourier transforms (FTs) of iron EXAFS oscillations (Fig. 6) of oxidized RH WT and RH stop showed two prominent peaks immediately revealing the presence of at least two backscatterer shells, likely due to Fe-O/N/S and Fe-Fe/Ni interactions around 2-2.2 and 2.7 Å (the true iron-backscatterer distances are by about 0.4 Å larger than the reduced distances indicated in Fig. 6). The FTs of RH WT ox and RH stop ox were not identical; FT peak I was larger in RH stop ox , and an additional shoulder was present on FT peak II in RH WT ox (Fig. 6, arrow). Precise structural information was obtained from simulations of the EXAFS oscillations (Fig. 6, inset). In both preparations the EXAFS was expected to be dominated by contributions from Fe-C/N/O/S and Fe-Fe/Ni vectors in the first and second ligand spheres. The broad shoulder at lower distances on FT peak I in RH WT (Fig. 6) suggested the presence of C/O/N atoms with significantly shorter distance from iron than the S-atoms at about 2.3 Å. A first simulation where one Fe-C/O/N,

FIG. 4. UV-visible spectra of RH WT (A) and RH stop (B) (solid lines, oxidized; dotted lines, H 2 -reduced).
The insets show the difference of spectra (H 2 -reduced minus oxidized).  one Fe-S, and one Fe-Fe/Ni vector were included yielded an error factor R F of about 10% for both RH WT and RH stop (  (29, 31, 34 -36). These results seemed to imply the presence of an additional more unusual iron cofactor in the RH WT .
In a third simulation a C-shell was included to account for the three CN/CO ligands of the iron of the Ni-Fe site. The fit for RH WT ox was not improved, supporting its lower (CN,CO)/Fe ratio. A significantly improved fit was obtained for RH stop ox (Table IV,  To gain more support for the unexpected result that the Fe-Fe coordination number (N Fe-Fe ) was close to one, a particularly stringent analysis (37) of the Fe-Fe vector in RH stop was performed. The values of N Fe-Fe and 2 2 Fe-Fe were varied, an EXAFS simulation was carried out for each parameter couple, and the resulting fit errors (R F ) were depicted in Fig. 7. Clearly, the absolute minimum (R F ϭ 12%) in the fit function was obtained at a value of N Fe-Fe close to 1. The presence of only one [4Fe-4S] cluster was strongly disfavored because at the then expected N Fe-Fe of 2.6 (taking into account the iron from the Ni-Fe active site) R F was 3-fold increased. The presence of only 1 [3Fe-4S] or of 1 [2Fe-4S] plus 1 [3Fe-4S] in RH stop was also disfavored because at the then predicted N Fe-Fe values of 1.75 and 1.5 (including the single Fe-Ni vector) R F was already about doubled (Fig. 7). Thus, the presence of 2 [2Fe-2S] clusters appeared to be the most likely option in RH stop .
The Fe-Fe distances in [4/3Fe-4S] clusters in crystallized Ni-Fe hydrogenases range between 2.62 and 2.77 Å (Refs. 34 -36; unusually long (38) and short (39) Fe-Fe distances were neglected); the range is similar in [2Fe-2S] clusters (29,31). The Debye-Waller parameters of the ϳ2.7-Å Fe-Fe vectors corresponded to a smaller distance spread of only ϳ0.07 Å (Table  IV, (48) over a reduced distance range of 1-3 Å. A, three fit approaches to spectra from oxidized enzymes. B, fit of the spectra from H 2 -reduced enzymes. C, fit of the spectrum of the tentatively isolated contribution from the extra iron ions in RH WT (see Fig. 8). The following fit restraints were applied: #, the sum of N 1-3 was restricted to a value of 5.0; §, 2 2 was coupled to yield equal values for the respective backscatterer shells; *, parameters were not varied in the simulations. tors differing by less than 0.1 Å does not cause significant destructive interference (22,37), rendering underestimation of N Fe-Fe unlikely.
In RH WT the fit minimum for the ϳ2.7-Å Fe-Fe vector was also observed for N Fe-Fe ϳ1 irrespective of the specific fit approach, e.g. inclusion of 1-3 O/N/C ligand shells and/or of an additional ϳ3 Å Fe-Fe vector (not documented). Presumably, RH WT contained the same [2Fe-2S] clusters as RH stop and, furthermore, an additional Fe-S cluster.
In the H 2 -reduced state, the EXAFS spectra of both RH preparations were overall similar to the ones in Fig. 6; in particular, the shoulder on FT peak II due to long Fe-Fe distances was still present in RH WT ϩH2 (data not shown). Simulations revealed significantly elongated Fe-O/N and Fe-S distances; the Fe-Fe/Ni vectors of ϳ2.7-Å length were slightly shortened (Table IV, B). Very similar effects have been observed upon reduction of the [2Fe-2S] Rieske cluster in Pseudo-monas cepacia phthalate dioxygenase (29). Fe-S cluster reduction was clearly detectable both in the XANES and EXAFS of the RH in the Ni-C state.
Tentative Identification of the Additional Iron Cofactor in RH WT -The EXAFS of RH WT revealed Fe-Fe distances of ϳ3 Å that were absent in RH stop . Similar Fe-Fe distances were found in the [4Fe-3S-3O] cluster in the proximal position of the small subunit of the Ni-Fe hydrogenase from Desulfovibrio desulfuricans (36). To test the hypothesis of whether the RH WT may contain a [4Fe-3S-3O] cluster, a tentative isolation of the contribution of the extra iron to the RH WT EXAFS was performed (Fig. 8, see legend). The resulting EXAFS difference spectrum (from oxidized enzymes) revealed pronounced splitting of both FT peaks (the difference from the H 2 -reduced enzymes was similar (not shown)). The best fit was achieved using a (O/N) 1.25-1.75 S 2.25-2.75 iron coordination and two Fe-Fe vectors with N Fe-Fe ϭ 1 and ϳ2.7and ϳ3-Å length (Table IV, C). The [4Fe-3S-3O] cluster in D. desulfuricans reveals similar structural motifs (Fig. 8, inset) and also 4-coordinated iron ions (Table IV, C, in parentheses). The FT (Fig. 8, triangles) of an EXAFS spectrum calculated on basis of the crystal structure (36) was similar to the one tentatively attributed to the extra Fe-S species in RH WT . The binding of one extra [4Fe-3S-3O] to each HoxB subunit was, thus, a conceivable option in RH WT .

DISCUSSION
The RH of R. eutropha belongs to a subclass of Ni-Fe hydrogenases acting as hydrogen sensors (3). H 2 -sensing proteins have also been examined in B. japonicum (7) and R. capsulatus (8). Moreover, genome sequencing projects of a number of microorganisms have uncovered sequences of additional potentially H 2 -sensing Ni-Fe hydrogenases (Fig. 9). These proteins may all be similarly organized, the large subunit carrying a Ni-Fe site and the small subunit harboring Fe-S clusters. However, due to the lack of crystal structures, only little information about the chemical nature of the putative Fe-S clusters in the H 2 -sensing hydrogenases was available. In this study, for the first time the structure and function of Fe-S clusters in the oxygen-insensitive hydrogen sensor of R. eutropha was investigated.
Alignment of the HoxB amino acid sequence with those of the small subunits of crystallized Ni-Fe hydrogenases and of potential H 2 sensors (Fig. 9) shows that the conserved residues coordinating the iron ions of Fe-S clusters in standard hydrogenases are always present. Thus, on the level of amino acid primary sequences one might argue that the small subunits of H 2 -sensing hydrogenases possibly contain Fe-S clusters such as found in the standard hydrogenases. Crystal structure analyses (34 -36) revealed that the standard hydrogenases harbor a proximal [4Fe-4S] cluster coordinated by four cysteine residues (P 1 to P 4 in Fig. 9). The medial cluster usually is a [3Fe-4S] cluster coordinated by three cysteines except for the Ni-Fe-Se enzyme from Desulfomicrobium baculatum (38), where a [4Fe-4S] cluster is found due to the replacement of one proline residue by another cysteine (39). All four cysteines are present in the sequences of the small subunits of the H 2 sensors (M 1 to M 4 in Fig. 9). The distal [4Fe-4S] cluster is coordinated by three cysteines and one histidine, which are conserved in all Ni-Fe hydrogenases including the H 2 sensors (D 1 to D 4 in Fig. 9). One cysteine is shifted by one amino acid in the binding motif for the distal Fe-S cluster (D 3 in Fig. 9).
Despite these striking sequence similarities, our experimental data provided no evidence for [4Fe-4S] or [3Fe-4S] clusters in the RH but instead favored the presence of [2Fe-2S] clusters possibly of the Rieske type. These may be coordinated by two cysteines and possibly two histidines. Conserved histidine residues present only in the sequences of the H 2 sensors that could  Table I) to allow for comparison with the theoretical EXAFS spectrum (triangles) of the [4Fe-3S-3O] cluster (inset) in the D. desulfuricans crystal structure (39) calculated by using parameters given in Table IV, C (in parenthesis). a.u., arbitrary units. be involved in Fe-S cluster binding are highlighted in Fig. 9 (H 1 to H 4 ). The coordination of iron ions by non-sulfur ligands could also be explained by the oxidative conversion of cysteine thiols to sulfenates (Cys-SOH) or sulfinates (Cys-SO 2 H) observed in a variety of proteins (40,41), which would impair direct coordination of iron ions by Cys-S. The presence of [2Fe-2S] clusters instead of [4Fe-4S] clusters in the RH may, thus, be related to thiol group oxidation under aerobic conditions. Preliminary resonance Raman data suggested the presence of Fe-O bonds in RH stop . 3 Thus, in RH stop perhaps iron binding oxygens from oxidized thiol groups may be present. Future studies will employ Mössbauer and resonance Raman spectroscopy to elucidate further details of the Fe-S clusters of the RH. Interestingly, in the Ni-Fe hydrogenase of the aerobic Thiobacillus ferrooxidans only 8 irons per nickel were found instead of the 12-13 irons per nickel in standard Ni-Fe hydrogenases (42). Possibly, [2Fe-2S] clusters may be present in Ni-Fe hydrogenases from a variety of aerobic organisms.
Metal quantification and XAS analysis seemed to reveal an additional Fe-S cluster in RH WT that is absent in RH stop . In D. desulfuricans, a [4Fe-3S-3O] cluster is found in the proximal position (36) with Fe-Fe distances of ϳ3 Å as also observed in RH WT . The authors postulated its formation in a reaction of a normal [4Fe-4S] cluster with O 2 and H 2 O (39), causing release of H 2 S and an iron coordination change from Cys-30 to Glu-87, a strictly conserved residue in the sequences of all Ni-Fe hydrogenases (E in Fig. 9). Because the RH operates under aerobic conditions, similar [4Fe-3S-3O] cluster formation may readily occur. A [4Fe-3S-3O] cluster has also been found in the so-called "prismane" protein (43). The aerobic conversion of a [4Fe-4S] into a [4Fe-3S-3O] cluster may be a more general feature of Fe-S proteins.
The range of the iron content in RH WT was also compatible with the binding of only one extra Fe-S cluster per [HoxBC] 2 by residues from the two HoxB proteins. Such a binding mode was observed in Rhodospirillum rubrum dimeric carbon monoxide dehydrogenase (CODH) where a [4Fe-4S] cluster is ligated by two cysteines from each of the two CODH molecules (44). The binding of a [4Fe-3S-3O] in between two HoxB units may be less likely for symmetry reasons since this cluster should be coordinated by three cysteines and by one glutamate residue.
In RH stop where the C-terminal extension of HoxB was truncated, XAS-detectable ϳ3-Å Fe-Fe distances were absent. The RH stop protein does not form the RH double dimer [HoxBC] 2 and is also unable to bind the PAS domain of the HoxJ protein (14). The HoxB C terminus may, thus, mediate these protein interactions and stabilize the Fe-S species appearing to be lost in RH stop . Loss of a Fe-S cluster in this position in RH stop could affect the Ni-Fe site in HoxC of the RH and the efficiency of electron transfer. Indeed, the nickel coordination may differ 3 I. Zebger and P. Hildebrandt, unpublished results. between RH WT and RH stop . 4 However, the hydrogen cleavage activity at the Ni-Fe site was unchanged. Assuming a maximal distance between the Fe-S clusters of about 25 Å as in standard hydrogenases, electron transfer between these clusters may in any event proceed at least within seconds (45).
Compared with energy-generating hydrogenases, the RH displays a very low but clearly defined hydrogen cleavage turnover rate (3,10). Consequently, electrons have to be released from the Ni-Fe active site. In standard Ni-Fe hydrogenases these electrons are transferred to the Fe-S clusters, which when reduced show typical EPR signals (1,13). In the RH, the splitting of H 2 at the Ni-Fe site yields the EPR-detectable Ni-C state (10,11), but no EPR signals from reduced Fe-S clusters in the HoxB subunit were observed. Previous UV-visible data were interpreted to suggest reduction of a two-electron accepting organic cofactor in the RH (10). The more detailed UVvisible analysis performed in the present study clearly indicated the reduction of Fe-S clusters upon incubation of the RH with H 2 . No evidence for an additional organic redox cofactor was obtained.
In the presence of hydrogen the reduction of iron ions both in RH WT and RH stop was unambiguously detected by iron XAS spectroscopy. In case of the RH stop protein, two iron ions possibly from two [2Fe-2S] ϩ clusters in the medial and distal positions of HoxB seemed to become reduced. Their presumably small distance of about 12 Å may cause strong magnetic coupling between the Fe(II) ions, rendering them EPR-invisible by relaxation enhancement. In RH WT there was evidence for the reduction of up to four iron atoms. Thus, besides of formation of two [2Fe-2S] ϩ clusters, the putative [4Fe-3S-3O] cluster may become doubly reduced. The [4Fe-3S-3O] cluster in prismane proteins was shown to exist in four oxidation states (43); the doubly reduced state was EPR silent. This cluster, however, showed a bleaching around 400 nm in the UV-visible spectrum (46) upon reduction similar to that of RH WT . Thus, a doubly reduced [4Fe-3S-3O] cluster may also escape detection by EPR in the RH.
The sensing of H 2 involves complex formation between double dimeric RH and a tetramer of the histidine protein kinase HoxJ (14). A similar arrangement has been described in the R. capsulatus regulatory hydrogenase (47). In the RH, reduction of Fe-S clusters of the HoxB subunit may cause a structural change of the RH-HoxJ complex, thereby modifying its phosphorylation activity. These events may represent the first step in the signal transduction chain leading to the expression of the energy converting Ni-Fe hydrogenases in R. eutropha.