The H2 sensor of Ralstonia eutropha. Biochemical characteristics, spectroscopic properties, and its interaction with a histidine protein kinase.

Previous genetic studies have revealed a multicomponent signal transduction chain, consisting of an H(2) sensor, a histidine protein kinase, and a response regulator, which controls hydrogenase gene transcription in the proteobacterium Ralstonia eutropha. In this study, we isolated the H(2) sensor and demonstrated that the purified protein forms a complex with the histidine protein kinase. Biochemical and spectroscopic analysis revealed that the H(2) sensor is a cytoplasmic [NiFe]-hydrogenase with unique features. The H(2)-oxidizing activity was 2 orders of magnitude lower than that of standard hydrogenases and insensitive to oxygen, carbon monoxide, and acetylene. Interestingly, only H(2) production but no HD formation was detected in the D(2)/H(+) exchange assay. Fourier transform infrared data showed an active site similar to that of standard [NiFe]-hydrogenases. It is suggested that the protein environment accounts for a restricted gas diffusion and for the typical kinetic parameters of the H(2) sensor. EPR analysis demonstrated that the [4Fe-4S] clusters within the small subunit were not reduced under hydrogen even in the presence of dithionite. Optical spectra revealed the presence of a novel, redox-active, n = 2 chromophore that is reduced by H(2). The possible involvement of this chromophore in signal transduction is discussed.

The detection of physiologically important gases by organisms is mediated by biological sensors that convert the molecular signal into a cellular response. Sensors for O 2 , CO, and NO have been described, and the signaling mechanism is the subject of current research (1)(2)(3). One of the best studied examples is the two-component FixL-FixJ system of Rhizobium meliloti and Bradyrhizobium japonicum. In this case the presence of O 2 is detected by a heme-containing histidine protein kinase (4). The heme group in FixL binds the oxygen molecule that induces a transition of the ferrous iron from high spin to low spin. This triggers the inactivation of the kinase domain of FixL. The release of O 2 , at low O 2 tensions, restores the S ϭ 2 state of the heme iron, which in turn leads to activation of the kinase by autophosphorylation. Subsequent phosphoryl transfer to the response activator FixJ finally stimulates gene transcription (5).
In the facultative chemolithotrophic proteobacterium Ralstonia eutropha (formerly Alcaligenes eutrophus), the oxidation of molecular hydrogen is catalyzed by two [NiFe]-hydrogenases, a membrane-bound (MBH) 1 and a cytoplasmic NAD-reducing hydrogenase (SH) (6,7). The structural genes of both [NiFe]hydrogenases together with sets of accessory genes are grouped in the MBH and SH operons, which are induced in the presence of molecular hydrogen (8,9). Hydrogenase gene transcription is controlled by a multicomponent regulatory system consisting of the proteins HoxB, HoxC, HoxJ, and HoxA, which are encoded in the MBH operon (8 -10). HoxJ and HoxA share typical features of a bacterial two-component regulatory system that recognizes and responds to various environmental stimuli (9,11). Our studies showed that HoxJ displays autokinase activity (9) and communicates with the activator HoxA, 2 a member of the NtrC family of response regulators (12). HoxA, the final target of the H 2 -sensing signal transduction chain, binds to the upstream region of the hydrogenase promoters and activates open complex formation by 54 RNA polymerase (8,13).
Genetic studies revealed that recognition of H 2 requires in addition to HoxA and HoxJ the protein HoxBC (9). Proteins similar to HoxBC, designated HupUV, have been identified in Rhodobacter capsulatus and B. japonicum (14,15). HoxBC-like proteins show typical features of a [NiFe]-hydrogenase (16). Although HoxBC is essential for lithoautotrophic growth of R. eutropha (9), it cannot compensate for the loss of the MBH and the SH. This observation points to a regulatory rather than an energy-yielding function of the HoxBC protein (16). The low level of expression combined with an extremely low activity allowed only preliminary biochemical analysis of HoxBC in crude extracts (17). Attempts to express a functional HoxBC protein in Escherichia coli were unsuccessful. This prompted us to develop a homologous overexpression system in R. eutropha (16) and to use it successfully for the purification of HoxBC, later named regulatory hydrogenase (RH).
Biochemical and spectroscopic analysis of the homogenous RH uncovered unique enzymatic features that are clearly distinct from the properties of standard hydrogenases. The data suggest that the RH shows a common [NiFe] active site but displays significant changes in the protein environment. In order to study the mechanism of H 2 signal transduction in more depth, we began to establish an in vitro system, using purified components. First data show that the RH forms a specific complex with the sensor kinase HoxJ, supporting the notion that the RH is a direct component of the signal transduction chain.

EXPERIMENTAL PROCEDURES
Cell Growth-R. eutropha strain HF371, a derivative of R. eutropha H16, harboring plasmid pGE378, was used for protein purification (16). Cells were heterotrophically grown in a mineral medium in a 10-liter Braun Biostat fermentor (Braun, Melsungen, Germany) at 30°C under hydrogenase derepressing conditions. At an OD 436 of 11 the cells were harvested, washed in 50 mM potassium phosphate buffer, pH 7.0 (K-PO 4 buffer), and stored frozen in liquid nitrogen.
RH Purification-Cells (83 g, wet weight) were resuspended in 30 ml of K-PO 4 buffer containing 0.1 mM phenylmethylsulfonyl fluoride. Cells were disrupted by two passages through a chilled Amicon French press cell at 1100 pounds/square inch (75.8 bar). Soluble extracts were prepared by high speed centrifugation (100 000 ϫ g, 60 min, 4°C). The resulting supernatant was degassed and saturated with hydrogen. The extract was kept under an atmosphere of 100% H 2 and subsequently incubated for 10 min at 65°C. After the heat treatment the sample was chilled on ice. All further purification steps were carried out under air. The denatured proteins were removed by centrifugation (13 000 ϫ g, 20 min, 4°C), and the supernatant was fractionated by addition of (NH 4 ) 2 SO 4 to give a final concentration of 1 M. The precipitated proteins were removed by centrifugation (13 000 ϫ g, 20 min, 4°C), and the clear supernatant was directly applied to a POROS 20ET column (Applied Biosystems; ethyl ether; 10 ϫ 100 mm), preequilibrated with K-PO 4 buffer containing 1 M (NH 4 ) 2 SO 4 at a flow rate of 40 ml/min (BioCAD Perfusion Chromatography Workstation). The column was washed with 2 bed volumes of K-PO 4 buffer containing 1 M (NH 4 ) 2 SO 4 . The protein was eluted with K-PO 4 buffer containing 0.4 M (NH 4 ) 2 SO 4 , and fractions of 4 ml were collected. The active fractions of several column runs were combined, concentrated, and dialyzed against K-PO 4 buffer. The RH was further purified on a POROS 20HQ column (Applied Biosystems; quarternized polyethyleneimine; 4.6 ϫ 100 mm) preequilibrated with K-PO 4 buffer. The eluent was pooled, concentrated (Centriprep-10; Amicon), and directly frozen in liquid N 2 . Protein concentrations were determined according to the methods of Lowry et al. (18).
Complex Formation Assay-The histidine protein kinase HoxJ was overproduced in E. coli and purified as a polyhistidine-tagged protein, His 6 -HoxJ, by metal chelate affinity chromatography (9). Purified His 6 -HoxJ and RH were mixed and subsequently applied to a native 4 -15% polyacrylamide gel. Native gel electrophoresis was carried out as described previously (19). Complex formation was either monitored by in-gel hydrogenase activity staining (19) or by protein staining using Coomassie Blue.
Metal Analysis-Nickel and iron were determined with a Hitachi 180-80 polarized Zeeman atomic absorption spectrophotometer against a standard series. Samples were made devoid of extraneous metal ions by passage over a Chelex-100 column (Bio-Rad).
Activity Measurements-Hydrogen uptake activity was measured amperometrically at 30°C in a cell (2.15 ml) with 50 mM Tris-HCl, pH 8.0, using a Clark-type electrode (YSI 5331) according to Coremans et al. (20). As O 2 did not affect the activity, no efforts were made to remove air. Hydrogen, in the form of H 2 -saturated water, was added to final concentrations varying from 36 to 100 M. As electron acceptor either benzyl viologen (4.2 mM, E m ϭ -359 mV) or methylene blue (4 mM, E m, pH 7 ϭ ϩ11 mV) were used. The measured specific activities were plotted against the H 2 concentration. The dependence was simulated using the program Leonora by Cornish-Bowden, assuming Michaelis-Menten kinetics (21). Protein concentrations in the assay were typically 2.5-5 nM RH (␣ 2 ␤ 2 ). Benzyl viologen-dependent H 2 evolution was determined amperometrically at 30°C. The reaction mixture contained 50 mM acetate buffer, 1 mM benzyl viologen, and 3 mM sodium dithionite. D ϩ /H ϩ exchange activity was measured in a stirred membrane leak chamber fitted to a mass spectrometer (Masstorr 200 DX quadrupole, VG Quadrupoles Ltd.). Two different assays were used. In the first assay 10 ml of Mes/Mops/Tris buffer solution (ionic strength 90 mM; pH 6.5) was saturated with 20% D 2 and 80% argon, and 1 mol of sodium dithionite was added to eliminate residual oxygen. The reaction was started by the addition of RH (␣ 2 ␤ 2 ) to a final concentration of 0.12 M. Masses 1-6 were scanned at 1 atomic mass unit/s. In the second assay the buffer solution was in 99.9% D 2 O (Aldrich) and saturated with H 2 . A control experiment was done in D 2 /D 2 O in order to evaluate the HD production catalyzed by the protein due to contaminant H ϩ . This effect was subtracted to the H 2 /D 2 O assay. The pD of the assay mixtures was measured with a glass electrode calibrated with pH standards in H 2 O. It was taken into account that pD ϭ pH ϩ 0.41 (22,23). All experiments were done at 30°C.
EPR Spectroscopy-X-band (9.4 GHz) spectra with a 100 kHz field modulation frequency were recorded on a Bruker ECS106 EPR spectrometer equipped with an Oxford Instruments ESR900 helium flow cryostat with an ITC4 temperature controller. The magnetic field was calibrated with an AEG Magnetic Field Meter. The frequency was measured with a Hewlett-Packard 5350B Microwave Frequency Counter. Illumination of the samples was performed by shining white light (Osram Halogen Bellaphot, 150 watts) via a light guide through the irradiation grid of the Bruker ER 4102 ST cavity. Spectra were simulated according to the formulas published by Beinert and Albracht (24).
FTIR Spectroscopy-Fourier transform infrared (FTIR) spectra were taken on a Bio-Rad FTS 60A spectrometer equipped with an MCT detector. Spectra were recorded at room temperature with a resolution of 2 cm Ϫ1 . Typically, averages of 1524 spectra were taken against proper blanks. Enzyme samples (10 l) were loaded into a gas-tight transmission cell (CaF 2 , 56 m path length). The spectra were corrected for the base line using a spline function provided by the Bio-Rad software.
Ultraviolet/Visible Spectroscopy-Optical spectra were taken on an Aminco DW-2000 spectrophotometer interfaced with an IBM computer.

RESULTS
Purification of the RH Protein-To avoid interferences with the dominant activities of the MBH and the SH, we started the purification of the RH protein with mutant R. eutropha HF371, in which the MBH and SH genes had been deleted by mutation. After cell disruption and high speed centrifugation, the soluble extract was incubated at 65°C for 10 min under an atmosphere of H 2 (100%). The heat treatment was necessary prior to high resolution hydrophobic interaction chromatography to provide an effective and rapid isolation of the RH. Purification to apparent homogeneity was achieved by subsequent anion exchange chromatography. The total procedure is summarized in Table I. Beginning with 83 g of cells (wet weight), 10.9 mg of RH was obtained. The specific activity of the preparation was 0.94 units/mg of protein with methylene blue as the electron acceptor. The H 2 concentration in the assay was 57.8 M. The protein was purified 26-fold with a yield of 6%. Biochemical Properties-Two protein bands occurred after denaturing the RH by SDS-PAGE corresponding to molecular masses of 37 and 55 kDa, respectively (Fig. 1A). These values are in good agreement with those predicted from the nucleotide sequence of hoxB (36.5 kDa) and hoxC (52.4 kDa). The identity of the purified protein was confirmed by immunoblot analysis, using an antibody raised against the HoxC subunit of the RH (Fig. 1B). Analysis of the enzyme on a Superdex G-200 (Amersham Pharmacia Biotech) revealed a single peak corresponding to a molecular mass of ϳ165 kDa (data not shown) indicating that the RH was purified as a tetramer with an ␣ 2 ␤ 2 structure. Atomic absorption spectroscopy showed an average metal content of 11.2 iron/nickel. After Chelex treatment this ratio decreased to 7.6 iron/nickel. The activity after the Chelex-100 column was 75% of the initial activity.
The oxidation of H 2 by the purified RH turned out to be O 2 -insensitive. The level of activity was the same in aerobic and anaerobic buffers. Moreover, the rate of H 2 oxidation determined with methylene blue as the electron acceptor did not show the typical lag phase that is found with most as isolated [NiFe]-hydrogenases. This observation is consistent with the result obtained with soluble extract (17). The K m for H 2 was 25 Ϯ 5 M, and the calculated specific activity at V max conditions was 1.2 Ϯ 0.2 units/mg of protein. The activity of the RH remained constant over a broad pH range between 5 and 10 irrespective of the used buffers (potassium acetate, K-PO 4 , and Tris-HCl; 50 mM each), whereas most hydrogenases show a distinct pH optimum. In contrast to the H 2 uptake activity, the production of H 2 by the RH was pH-dependent. Highest H 2 evolution rates (0.8 units/mg of protein) were obtained at pH 4.0 with benzyl viologen as electron donor. Acetylene has been shown to be a competitive inhibitor for several hydrogenases (25,26). Incubation of the RH with C 2 H 2 did not affect the RH activity (data not shown).
Storage of the purified RH at 4°C under air or an atmosphere of 100% O 2 resulted in a loss of 50% of the H 2 -dependent methylene blue-reducing activity within 48 h. Replacement of the air atmosphere by 100% argon or N 2 caused a decrease of 20% of the activity within the same period. Addition of metal ions (Fe 3ϩ , Ni 2ϩ , Mn 2ϩ , Mg 2ϩ , and Zn 2ϩ ) or glycerol (20%) or addition of KCl up to 0.5 M did not affect the stability of the RH. The supply of dithionite or ferricyanide under anoxic conditions also did not contribute to the stability of the RH. Moreover, storage of the isolated RH under an atmosphere of 100% H 2 inactivated the RH rapidly; 50% of its activity disappeared within 12 h. The H 2 sensitivity contrasts with data obtained with the soluble extract, which showed constant RH activity over a period of 24 h under comparable conditions. In all cases inactivation of the RH was irreversible.
D ϩ /H ϩ Exchange Activity-The D ϩ /H ϩ exchange assay with the RH yielded only H 2 production but no HD formation ( Fig.  2A). The initial rate of H 2 production at pH 6.5 was 2.1 Ϯ 0.1 units/mg of protein. This behavior is distinct from that of other [NiFe]-hydrogenases, which show higher initial rates of HD production than of H 2 production. When the exchange activity assay was measured in deuterated water saturated with H 2 some HD production was detected with the RH, although the rate of D 2 evolution was definitively higher (Fig. 2B). The initial rate of D 2 production at pD 6.5 was 1.3 Ϯ 0.2 units/mg of protein, whereas the initial rate of HD production was 0.5 Ϯ 0.1 units/mg of protein. The pH optimum of the D ϩ /H ϩ exchange activity of the RH was at pH 5.5 (data not shown).
EPR Spectroscopy-Preliminary studies of the RH in crude cell extracts prohibited a study of the EPR properties of its Fe-S clusters (17). The purified enzyme now allowed this approach. The as isolated RH showed no EPR signals at temperatures between 4.2 and 100 K. Also after addition of the oxidizing agent DCIP (dichlorophenolindophenol, E m ϭ ϩ230 mV), no signal occurred. Upon reduction of the RH (15 min under 100% H 2 at room temperature in 50 mM Tris-HCl, pH 8.0), a rhombic EPR signal with g values at 2.19, 2.13, and 2.01 appeared (Fig.  3, trace A). The double-integrated intensity of the signal amounted to a spin concentration equal to 69% of the nickel concentration.
The EPR signal is very similar to the well studied Ni a -C* signal observed in standard hydrogenases (e.g. from Desulfovibrio gigas and Allochromatium vinosum), and it is due to a paramagnetic state of the active site nickel in the 3ϩ state (27). A typical feature of enzyme in the Ni a -C* state is its light sensitivity at cryogenic temperatures, yielding the so-called Ni a -L* signal as a result of the photodissociation of a hydrogen (28). A model for this photodissociation has been described by Happe et al. (27).
In the case of the RH the Ni a -C* signal also showed this light-sensitive behavior. Upon illumination at 30 K a spectrum (Ni a -L*: g xyz ϭ 2.045, 2.09, and 2.24), only slightly different from the Ni a -L* signal of standard hydrogenases, appeared (Fig. 3, trace B). The small difference concerns the position of the g z (2.24 in the RH as compared with 2.28/2.30 in standard [NiFe] hydrogenases). This points to a small structural difference around the active site nickel. Upon warming of the sample to 200 K for 15 min in the dark a third, transient, spectrum came up with g values at 2.047, 2.069, and 2.30 (Fig. 3, trace C). Only after several hours at 200 K the sample returned to the Ni a -C* state.
Contrary to observations in standard [NiFe]-hydrogenases no signal of a [3Fe-4S] ϩ cluster could be observed in the oxidized protein, not even after treatment with excess DCIP. This is in agreement with the presence of three [4Fe-4S] clusters as predicted from the amino acid sequence data. When the protein was treated with 100% H 2 , however, no signals due to reduced cubanes were detectable, not even if 20 mM dithionite was added. None of the nickel signals (Ni a -C*, Ni a -L*, or the transient signal) showed any spin coupling due to a reduced proximal [4Fe-4S] cluster (Fig. 3). This indicates that this cluster was in the oxidized, diamagnetic state in the RH under H 2 . In standard [NiFe]-hydrogenases the proximal cluster is usually reduced under 100% H 2 . The interaction of the nickel with the reduced proximal cluster is observed as a clear 2-fold splitting of the Ni a -C* signal at 4.5 K. At low temperatures it was also possible to completely saturate the Ni a -C* signal at high microwave power (260 milliwatts), which is again indicative of an oxidized proximal cluster (28,29). Reduction of the RH with dithionite in the presence or absence of low potential electron acceptors (methyl viologen and benzyl viologen) under 100% H 2 did not evoke any signal of a reduced Fe-S cluster. Also inspection of the integrated EPR signals did not uncover any broad signal due to reduced Fe-S clusters as can be seen in the right-hand panel in Fig. 3 for the Ni a -L* signal.
FTIR Spectroscopy-The FTIR measurements on purified RH confirmed the presence of only two redox states described earlier to be present in the RH from soluble extracts (17). Untreated protein showed a spectrum (Fig. 4A) with two small bands (2082 and 2071 cm Ϫ1 ) and one large band (1943 cm Ϫ1 ) in the 2150 -1850 cm Ϫ1 spectral region. This EPR-silent state of the active RH resembles the Ni a -S state of standard [NiFe]hydrogenases. Maximal reduction, already obtained after a few minutes under 100% H 2 at room temperature, yielded the Ni a -C* state (Fig. 4C) as identified previously in other [NiFe]hydrogenases (30,31). This state showed a CO stretch vibration at 1960 cm Ϫ1 . The two bands at 2082 and 2071 cm Ϫ1 , which did not shift, are ascribed to the symmetrical and antisymmetrical coupled vibrations of two cyanides bound to iron in the active site (17). It was not possible to reduce further this state by adding excess dithionite (20 mM, spectrum not shown).
When the gas phase was changed from 100% H 2 to 100% CO (equilibration time 60 min) a mixture of the Ni a -C* and Ni a -S state was observed (Fig. 4B). The spectrum clearly showed that it was not possible for exogenous CO to bind to the active site of the RH since no extra peak around 2060 cm Ϫ1 could be seen. Such a band from added CO is observed in the A. vinosum and D. gigas enzyme 3 (32). A similar change was observed by replacing H 2 with argon (results not shown). Upon complete oxidation with excess DCIP (2 mM) the sample returned to the Ni a -S state.
UV-Visible Absorption Spectroscopy-UV-visible spectra of oxidized and reduced RH showed differences in absorption between the two species. Incubation of the RH under 100% H 2 resulted in an increase in absorption in the 250 -280 and 300 -400 nm spectral regions (Fig. 5). The difference spectrum of reduced minus oxidized RH showed a large peak at 251 nm and a smaller one at 342 nm with an apparent shoulder at 305 nm. The calculated ⑀ 251 was 11.96 mM Ϫ1 cm Ϫ1 based on protein concentration. Similarly the ⑀ 342 was calculated to be 5.36 mM Ϫ1 cm Ϫ1 . The protein concentration used (0.64 mg/ml) was such that the absorption at 280 nm was about 1.0. At this intensity the detector is still sensitive enough to pick up reliable differences in the UV, meaning that these are not due to 3   Complex Formation-To elucidate the nature of the interaction between the RH and the signal transduction chain, purified kinase HoxJ and the RH were mixed, and the sample was subjected to native PAGE (Fig. 6). In one experiment the gel was resolved by protein staining (Fig. 6A), and in the parallel experiment a hydrogenase in-gel activity staining with phenazine methosulfate as electron acceptor was performed (Fig. 6B). Incubation of the RH with increasing amounts of HoxJ led to a bandshift indicating the formation of a high molecular weight complex (Fig. 6A, lanes 2-4). A bandshift was not observed in the control containing the RH and an excess of bovine serum albumin (Fig. 6, A and B, lane 5). The in-gel assay (Fig. 6B) demonstrated that the hydrogenase activity of the RH was maintained at high level upon complex formation (Fig. 6B,  lanes 3 and 4). Exposure to H 2 resulted in considerable loss of hydrogenase activity (Fig. 6B, lane 6), which is consistent with the observed instability of the RH in the presence of H 2 .

DISCUSSION
Genetic and biochemical studies uncovered a signal transduction chain, which directs H 2 -dependent gene activation in R. eutropha. This signal transduction chain consists of the transcription activator HoxA, the histidine protein kinase HoxJ, and the H 2 sensor RH. The RH is absolutely necessary for the recognition of dihydrogen suggesting its primary role in signal reception (9). Sequence alignment revealed that the RH contains typical signatures of [NiFe]-hydrogenases (16), and a preliminary EPR and FTIR study showed an active site similar to that of prototypic [NiFe]-hydrogenases (17). Characterization of the purified RH achieved in this study confirmed some biochemical features that are compatible to those of standard [NiFe]-hydrogenases. On the other hand, some characteristics were uncovered that are obviously uniquely assigned to the subgroup of H 2 -sensing proteins (16). Unlike standard [NiFe]hydrogenases, which usually have H 2 uptake activities of about 200 -300 units/mg of protein, the RH displayed a specific activity at V max of only 1.2 Ϯ 0.2 units/mg protein. Moreover, in the air-oxidized state the RH showed no lag phase, suggesting that it does not require a reductive activation step before the protein is enzymatically active.
Interestingly, the activity of the RH was not inhibited by O 2 , CO, or C 2 H 2 . Most hydrogenases are sensitive to these gases with the exception of the SH of R. eutropha (33). In this case a modified catalytic center probably excludes the binding of CO and O 2 (34). Although the EPR and FTIR spectra of the [NiFe] site of the RH resemble those of standard [NiFe]-hydrogenases, the active site of the RH exhibits some important redox differences. Only the Ni a -S and Ni a -C* states are attainable, and CO cannot bind to the active enzyme. This indicates that the nickel site (where in standard [NiFe]-hydrogenases CO binds and where H 2 is proposed to react under turnover conditions (27)) is altered such that it cannot react with CO or H 2 . This would restrict the reaction with H 2 to the iron site resulting in the Ni a -C* state only. The very low activity of the RH is in line with this idea. The D 2 /H ϩ exchange data suggest that D 2 diffusion to and from the active site is severely restricted resulting in a molecular cage effect (35). The formed HD then reacts again to form H 2 , before diffusion of HD from the enzyme to the bulk occurs. In the H 2 /D ϩ exchange, the formed D 2 escapes slower FIG. 4. FTIR spectra of the RH. Trace A shows the RH in the oxidized state. After reduction under H 2 the RH ends up in the reduced state (trace C). If the gas phase was then exchanged for CO, a mixture of oxidized and reduced RH was observed (trace B). A similar spectrum was obtained by flushing with argon.

FIG. 5. Difference spectrum of reduced minus oxidized RH.
Aerobic RH was diluted to a concentration of 0.64 mg/ml and divided over two cuvettes. One cuvette was put under 100% H 2 and measured against the aerobic sample in the reference cuvette at 2 nm resolution. than HD allowing some HD detection. The gas channel detected in the x-ray structures of [NiFe]-hydrogenases (36,37) points right to the nickel site. Changes in the amino acid composition of this channel close to the nickel site, e.g. the presence of more bulky residues, could explain both the redox and the exchange properties. The kinetic behavior of the RH in the D 2 /H ϩ activity assay is in agreement with the low activity of the RH in the other assays.
The described EPR and FTIR data on the purified RH do not differ from those presented earlier for the protein in crude extracts, so purification does not change these properties. A previously unobserved state occurred when the Ni a -L* state was warmed up to 200 K. A transient state was then observed with g values at 2.047, 2.069, and 2.30. This points to changes induced in the vicinity of the nickel site. As yet, we do not understand the nature of these changes.
Another typical feature of the light sensitivity in the RH is that all conversions are much slower than in several other hydrogenases tested in this laboratory using the same experimental set up (e.g. Allochromatium vinosum, Methanococcus thermoautotrophicum, and Wollinella succinogenes). The Ni a -C* to Ni a -L* conversion in membrane-bound hydrogenase (MBH) of A. vinosum is completed within 5 min, whereas in the RH it took about 15 min. The difference in the reverse reaction was even more pronounced. After 2 h at 200 K the RH was still in the transient dark state, whereas the MBH of A. vinosum requires only 10 -15 min at 200 K to return completely to the Ni a -C* state. This slow photolysis and the extremely slow annealing might be due to a less spacious, obstructed active site.
It was shown that it was impossible to reduce the three [4Fe-4S] clusters (predicted to be present from sequence data), although highly reductive conditions were applied (100% H 2 with benzyl viologen, methyl viologen and/or 20 mM dithionite). Also no splitting of the Ni a -C* or Ni a -L* signals at 4.5 K by a reduced proximal cluster was observed.
Our current model of signal transduction in the RH is as follows; H 2 binds to the active site (presumably at the iron site (27)) and causes a formal oxidation of the nickel ion from the 2ϩ to the 3ϩ state. The released electron is transferred to the Fe-S clusters. However, no spectroscopic evidence for a reduced Fe-S cluster was found, and no other S ϭ 1/2 EPR signal was detected. Since the RH is apparently functional as an ␣ 2 ␤ 2 tetramer, the possibility exists that two unpaired spins released by the two [Ni-Fe] sites in the tetramer are united in a yet undetected, diamagnetic prosthetic group. Hence UV-visible spectroscopy was applied. Much to our surprise reduction of the RH by H 2 resulted in an increase in absorption with clear maxima at 251 and 342 nm. We tentatively conclude that this increase is caused by the reduction of a two electron accepting cofactor, shared by the two dimer (␣␤) molecules in the RH (␣ 2 ␤ 2 ). The exact identity of this cofactor is currently under investigation. The position of the 342 nm band and its approximate molecular absorption coefficient (5.4 mM Ϫ1 cm Ϫ1 ) resemble those of NADH.
Transmission of the H 2 -induced changes in the RH to the histidine protein kinase HoxJ proceeds via direct protein-protein interaction as shown by complex formation. The N-terminal part of HoxJ, the so-called input domain, is the most likely region for the signal-accepting site. Sequence comparison revealed that this domain is a member of the PAS domain superfamily, which is found in a wide variety of regulatory systems involved in the sensing of light, oxygen, or redox potential (38)(39)(40). Several PAS domain proteins mediate signal transmission by the way of an associated cofactor (40), like the FAD in the aerotaxis signal transducer Aer of E. coli (41). Such a twoelectron cofactor in HoxJ might be a good candidate to be reduced by the yet unidentified cofactor in the RH. In this scenario, electron flow from the RH to the histidine kinase should induce a conformational switch to modulate the activity of the HoxJ transmitter domain and thereby affect the autophosphorylation activity of HoxJ. To resolve such a mechanism we intend to block electron transport within the RH by sitedirected mutagenesis. Attractive targets will be the ligands of the three FeS clusters and the nonmetal cofactor of the RH.