Oxygen-tolerant H2 Oxidation by Membrane-bound [NiFe] Hydrogenases of Ralstonia Species

Knallgas bacteria such as certain Ralstonia spp. are able to obtain metabolic energy by oxidizing trace levels of H2 using O2 as the terminal electron acceptor. The [NiFe] hydrogenases produced by these organisms are unusual in their ability to oxidize H2 in the presence of O2, which is a potent inactivator of most hydrogenases through attack at the active site. To probe the origin of this unusual O2 tolerance, we conducted a study on the membrane-bound hydrogenase from Ralstonia eutropha H16 and that of the closely related organism Ralstonia metallidurans CH34, which was purified using a new heterologous overproduction system. Direct electrochemical methods were used to determine apparent inhibition constants for O2 inhibition of H2 oxidation (\batchmode \documentclass[fleqn,10pt,legalpaper]{article} \usepackage{amssymb} \usepackage{amsfonts} \usepackage{amsmath} \pagestyle{empty} \begin{document} \(K_{I(\mathrm{app})}^{\mathrm{O}_{2}}\) \end{document}) for each enzyme. These values were at least 2 orders of magnitude higher than those of “standard” [NiFe] hydrogenases. Amino acids close to the active site were exchanged in the membrane-bound hydrogenase of R. eutropha H16 for those from standard hydrogenases to probe the role of individual residues in conferring O2 sensitivity. Michaelis constants for H2 (\batchmode \documentclass[fleqn,10pt,legalpaper]{article} \usepackage{amssymb} \usepackage{amsfonts} \usepackage{amsmath} \pagestyle{empty} \begin{document} \(K_{M}^{\mathrm{H}_{2}}\) \end{document}) were determined, and for some mutants these were increased more than 20-fold relative to the wild type. Mutations resulting in membrane-bound hydrogenase enzymes with increased \batchmode \documentclass[fleqn,10pt,legalpaper]{article} \usepackage{amssymb} \usepackage{amsfonts} \usepackage{amsmath} \pagestyle{empty} \begin{document} \(K_{M}^{\mathrm{H}_{2}}\) \end{document} or decreased \batchmode \documentclass[fleqn,10pt,legalpaper]{article} \usepackage{amssymb} \usepackage{amsfonts} \usepackage{amsmath} \pagestyle{empty} \begin{document} \(K_{I(\mathrm{app})}^{\mathrm{O}_{2}}\) \end{document}) values were associated with impaired lithoautotrophic growth in the presence of high O2 concentrations.

dues, two of which are bridging ligands to the Fe, which additionally coordinates two CN Ϫ and one CO (7,8). The structure of the active site of the "standard" O 2 -sensitive [NiFe] hydrogenase from Desulfovibrio gigas is shown in Fig. 1A.
A crucial feature of hydrogenases is their sensitivity to O 2 . Although the [NiFe] hydrogenases are generally more robust toward O 2 damage than the [FeFe] enzymes (9,10), the vast majority of [NiFe] hydrogenases act under anaerobic conditions in vivo, and their activity is normally subject to reversible inactivation by O 2 . Dioxygen is a -acceptor ligand like H 2 and CO and is expected to enter the active site easily. However, it subsequently behaves as an oxidizing agent, leading to inactive "resting" states. EPR studies on a set of standard [NiFe] hydrogenases, such as those isolated from Allochromatium vinosum, D. gigas, Desulfovibrio vulgaris Miyazaki F, and Desulfovibrio fructosovorans, have identified distinct Ni(III) features associated with two inactive states, known as Ni-A ("Unready") and Ni-B ("Ready"), and these are also associated with subtly distinct (CO) and (CN) frequencies in IR spectra (11,12). In conjunction with crystallographic studies on the D. gigas enzyme, for example, it has been established that Ni-B incorporates a hydroxide ligand into the bridging position between Ni and Fe, whereas further electron density in the active site of Ni-A suggests a bridging peroxide ligand or an O atom bonded to cysteinyl S; both ligands are released through reductive activation under H 2 (13)(14)(15). The ready state of the enzyme is also formed under anaerobic oxidative conditions (10 -12, 16).
In contrast to standard hydrogenases, a few [NiFe] hydrogenases, especially those synthesized by Knallgas bacteria, catalyze aerobic H 2 oxidation using O 2 as the terminal oxidant (9). In the light of the above comments, this situation is seemingly paradoxical. Furthermore, there are intense efforts to produce organisms and synthetic catalysts that cycle H 2 without interference from O 2 . We define hydrogenases that exhibit catalytic activity in the presence of O 2 as being "O 2 tolerant" (10). The [NiFe] hydrogenases expressed by chemolithoautotrophic Ralstonia species fall into this special category (17). For example, electrochemical experiments have shown that although the membrane-bound hydrogenase (MBH) 5 enzymes from Ralstonia eutropha H16 (Re H16) and Ralstonia metallidurans (Rm CH34) react rapidly and reversibly with O 2 , substantial H 2 oxidation activity remains even in air (9,18). A practical demonstration of this O 2 tolerance was provided by the construction of an H 2 fuel cell employing Rm CH34 MBH as the anode catalyst and laccase (a multicopper oxidase catalyzing the clean four-electron reduction of O 2 to water) at the cathode. With this device it was possible to produce sufficient electricity from a quiescent atmosphere of just 3% H 2 in air to power a wristwatch for over 24 h (19).
No crystal structure has yet been obtained for any O 2 -tolerant hydrogenase. IR spectroscopic experiments suggest that Re H16 MBH has the same coordination arrangement of CO and CN Ϫ ligands at the active site as the standard O 2 -sensitive [NiFe] hydrogenases (20). 6 An amino acid sequence comparison reveals that Re H16 MBH shares only ϳ40% identity with standard [NiFe] hydrogenases. However, the amino acids closest to the Re H16 MBH Ni-Fe active site are identical to those in the O 2 -sensitive periplasmic hydrogenase from D. gigas, for which a structure has been determined (4). The nearest nonconserved residues to the active site are (according to D. gigas hydrogenase numbering) tyrosine 70 and valine 71, which correspond to glycine 80 and cysteine 81 in Re H16 MBH (Fig. 1B). To probe whether these specific residues confer O 2 tolerance, a series of exchanges via site-directed mutagenesis was introduced into the Re H16 MBH.
Experiments on the H 2 -sensing hydrogenases (RH) from Re H16 and Rhodobacter capsulatus (21,22) have provided evidence that O 2 tolerance arises from the bulky residues isoleucine and phenylalanine restricting access of O 2 through a narrow gas channel to the active site. In Re H16 MBH, the corresponding residues are valine 77 and leucine 125. To probe whether introducing bulky residues into the MBH further enhances the O 2 tolerance of the enzyme, the residues found in the sensory hydrogenases were introduced into Re H16 MBH by genetic engineering.
The effects of these mutations were investigated in terms of cell growth characteristics and the catalytic properties of the isolated enzymes. The isolated enzymes were studied electrochemically using the suite of techniques known as protein film voltammetry (PFV), which has proved to be very useful in studying hydrogenases from various organisms (9,10,20,23,24). The enzyme is adsorbed onto an electrode as a sub-monolayer film such that the electrode replaces physiological electron donors and acceptors. Catalytic currents report directly on enzyme activity under conditions of controlled potential (driving force). Using PFV, activity can be measured even under aerobic conditions (18), in which soluble electron donors would be oxidized by O 2 .

EXPERIMENTAL PROCEDURES
Strains and Plasmids-The bacterial strains and plasmids and the primers used in this study are listed in Table 1 and Table  2, respectively. Escherichia coli JM109 (25) was used as host in standard cloning procedures, and E. coli S17-1 (26) served as a donor in conjugative transfers. Both Re H16 and Rm CH34 are wild-type strains, and strains carrying the letters HF are derivatives of Re H16.
For purification of the Re MBH variants, we used the plasmid pGE636 carrying the complete MBH operon with a hoxK gene fused with a StrepTag II sequence at its 3Ј end (27). An equiva-  (4). A shows the structure of the active site in its active state, and B highlights amino acid residues within a 12-Å distance of the Ni-Fe active site, which are shown in gray. Nickel, iron, and acid-labile sulfur atoms are shown as green, brown, and yellow spheres, respectively. The diatomic CN Ϫ and CO ligands to the iron and the variable amino acids of interest are shown as ball-and-stick models. Labels indicate nature and position of the respective amino acids residues of D. gigas enzyme and those of the Re H16 MBH. The organisms stated in parentheses contain hydrogenases harboring the specific amino acid residues that have been chosen for alterations of Re H16 MBH. lent plasmid for overproduction of the MBH from Rm CH34 was constructed as follows. A HindIII restriction site within the coding region for the leader sequence of HoxK and a BsrGI site downstream of hoxG were introduced by inverse PCR with primers 638 and 639 using pCH1229 as a template. The 4.39-kbp amplificate was digested with XbaI and re-ligated, resulting in pCH1237. A 2.94-kbp PCR fragment, amplified with primers 640 and 641 on Rm CH34 genomic DNA, was digested with HindIII-BsrGI and inserted into pCH1237 resulting in pCH1266. Plasmid pCH1266 was cut with SpeI and AsiSI, and a  A StrepTag II sequence was fused to the 3Ј end of hoxK from Rm CH34 by inverse PCR using primers 653 and 654 with pCH1269 as the template. The resulting 3.48-kbp PCR fragment was digested with RsrII and re-ligated, resulting in pCH1294. A 0.41-kbp PspOMI-SalI fragment from pCH1294 was cloned into pCH1266 resulting in pCH1295, which carries the StrepTag II sequence at the 3Ј end of hoxK. A 4.44-kbp SpeI-AsiSI fragment was transferred from pCH1295 into pCH1351 yielding pCH1296. From there a 9.06-kbp SpeI-Ecl136II fragment was cloned into pCH785 resulting in pCH1297. Finally, a 21.6-kbp SpeI-XbaI fragment from pCH1297, carrying the MBH operon with CH34 hoxK Strep G, was cloned into pEDY309 resulting in pGE621. The plasmids pGE615, pGE621, and pGE636 were transferred via conjugation into Re HF631, which is devoid of megaplasmid pHG1 (29), yielding Re strains HF688 (Rm MBH), HF717 (Rm MBH Strep ), and HF641 (Re MBH Strep ), respectively.
Amino acid exchanges in HoxG of Re H16 were constructed via PCR using the forward primer 649 and the mutagenic primers 646 (for C81A), 648 (for C81S), 647 (for C81T), and 661 (for C81V), respectively. Plasmid pCH1234, carrying a 69-kbp Acc65I-BmgBI hoxG fragment, served as the template. The resulting PCR products were inserted as 143-bp AgeI-SphI fragments into pCH1234. Also, the amplificates resulting from PCRs with primers 645, 649, and 662 (for G80Y) or 663 (for G80Y/C81V) and pCH1351 as template were introduced as a 179-bp Acc65I-MscI fragment into pCH1234. For the V77I exchange, an 88-bp Acc65I-AgeI-digested PCR fragment, generated with primers 644 and 645 and pCH1351 as the template, was transferred into pCH1234. The L125F exchange was obtained by cloning a 234-bp MscI-BamHI-digested PCR product, generated with primers 650, 651, and 652 and pCH1351 as the template, into pCH1234. A 1.80-kbp AgeI-DraIII fragment from the L125F derivative was transferred into the previously constructed V77I derivative of pCH1234 yielding a hoxG mutant fragment encoding a V77I/L125F exchange. From all pCH1234 derivatives carrying the hoxG mutations, a 0.68-kbp Acc65I-BstZ17I fragment was transferred into pCH1351 and pCH1353. From the pCH1351 derivatives 4.82-kbp SalI-PstI fragments were cloned into pLO2, yielding plasmids that were used for introduction of the mutations into Re HF388 by double homologous recombination as described previously (30). The resulting strains HF681-HF740 are listed in Table  1. From the pCH1353 derivatives, 9.06-kbp SpeI-Ecl136II fragments were transferred into pCH785 and, from there, as 21.60kbp SpeI-XbaI fragments into pEDY309. The resulting plasmids pGE610-614, pGE622, and pGE639-641 were transferred into Re HF631 by conjugation. This allowed the overproduction and purification of the MBH variants via a StrepTag II fused to the C terminus of the small subunit HoxK. All PCR amplificates were verified by sequencing.
MBH Purification-Cell pellets were resuspended in resuspension buffer (100 mM Tris/HCl, pH 8.0, 150 mM NaCl), 4 ml per 1 g of cells (wet weight), containing protease inhibitor mixture (Complete EDTA-free protease inhibitor mixture, Roche Applied Science) and DNase I. The cell suspension was subsequently disrupted in a French pressure cell (Constant Cell Disruption Systems or SLM Aminco). The resulting crude extract was treated by sonication (2 min, level 2.5, 75%, Branson Sonifier), and the membrane and soluble fractions were separated by ultracentrifugation (100,000 ϫ g at 4°C for 60 min). The brownish membranes were removed, homogenized in an appropriate volume of washing buffer (resuspension buffer ϩ 1 mM EDTA), and ultracentrifuged again (100,000 ϫ g at 4°C for 30 min). For MBH purification, the membrane proteins were solubilized by adding washing buffer containing Triton X-114 at a final concentration of 2% w/v and subsequently stirring on

Oxygen-tolerant H 2 Oxidation
ice for 1.5 h. After ultracentrifugation (100,000 ϫ g, 4°C, 20 min), the supernatant, containing the solubilized membranes, was loaded on Strep-Tactin Superflow columns (IBA, Göttingen, Germany; 1-ml bed volume for up to 25 ml of solubilized membrane extract). The columns were washed with 12 column volumes of washing buffer, and proteins were eluted with 6ϫ 0.5 ml of elution buffer (washing buffer ϩ 5 mM desthiobiotin). MBH-containing fractions were pooled and concentrated using a centrifugal filter device (Amicon Ultra-15 (PL-30), Millipore). The buffer was changed twice by adding 10 ml of storage buffer (50 mM Tris/HCl, pH 8.0, 50 mM NaCl, 1 mM EDTA, 20% glycerol) and subsequent recentrifugation. Aliquots of the resulting concentrate were frozen in liquid N 2 and stored at Ϫ80°C. Protein concentrations were determined by the Bradford method (33) using bovine serum albumin as standard. The purity of the samples was estimated by visual inspection of SDS-polyacrylamide gels stained by Coomassie Brilliant Blue G-250 (34). Western immunoblot analysis was performed according to a standard protocol (35) with anti-HoxK serum (1:5000) and anti-HoxG serum (1:10,000). H 2 -dependent Methylene Blue Reduction Assays-A spectrophotometric assay was used to measure MBH (hydrogen:acceptor oxidoreductase; EC 1.18.99.1) activity using methylene blue as the electron acceptor in an H 2 -flushed cuvette sealed with a rubber septum (36). Activity measurements on isolated membrane fractions were carried out in potassium phosphate buffer (50 mM) at pH 7.0, whereas measurements on MBH solubilized from the membrane were conducted at pH 5.5.
For initial measurements of the Michaelis constants for H 2 oxidation (K M H 2 ), MBH activities were quantified with a Clarktype oxygen electrode (Oxygraph, Hansatech Instruments) adapted for H 2 measurements (37). A sample of H 2 -saturated phosphate buffer (1 ml, pH 5.5) was mixed with an equal volume of N 2 -saturated buffer, resulting in a starting H 2 concentration of 0.4 mM (at 30°C). After addition of methylene blue (0.6 mM final concentration) and dithionite (80 M final concentration), the reaction was initiated by injecting aliquots (5-20 l) of MBH sample.
Electrochemical Measurements of H 2 Oxidation Activity-Protein film voltammetry experiments were carried out in an anaerobic glovebox (M. Braun) comprising a N 2 atmosphere (O 2 Ͻ2 ppm). A pyrolytic graphite "edge" rotating disk electrode (rotating disk electrode, area 0.03 cm 2 ) was used in conjunction with an electrode rotator (EcoChemie Autolab rotating disk electrode) in a specially designed, gas-tight, glass electrochemical cell incorporating a Pt wire counter electrode. A saturated calomel reference electrode (SCE) was contained in a side arm containing 100 mM NaCl, separated from the main cell compartment by a Luggin capillary. Potentials (E) are quoted with respect to the standard hydrogen electrode (SHE) using the correction E SHE ϭ E SCE ϩ 242 mV at 298 K (38). Electrochemical experiments were carried out using an electrochemical analyzer (Autolab PGSTAT10 or 20) controlled by a PC operating GPES software (EcoChemie).
To prepare an enzyme film, the pyrolytic graphite edge electrode was first polished for 30 s with an aqueous slurry of ␣alumina (1 m, Buehler) and sonicated for 5 s in purified water, before enzyme solution (1.5 l, 0.2-1 g of enzyme, pH 5.5) was repeatedly applied and withdrawn from the electrode surface over a period of 10 s. The electrode was then placed in enzymefree buffered electrolyte so that all enzyme molecules addressed in the experiments were subjected to the same regime of strict potential control. In all experiments the electrode was rotated at a constant rate (2500 -6000 rpm) to provide efficient supply of substrate and removal of product. For experiments carried out at low levels of H 2 , the electrode tip was gently polished with damp cotton wool to lower the enzyme coverage to minimize the effect of limiting transport of H 2 to the electrode. The method is sensitive to extremely low levels of H 2 (18) and has a very small sample requirement; determination of an average value for K M H 2 from Ͼ12 repeated experiments can be achieved using just 2 g of pure protein.
Experiments were performed in phosphate buffer (50 mM; Sigma) containing NaCl (100 mM; Fisher) and titrated to the desired pH at the experimental temperature. All solutions were prepared using purified water (Millipore; 18.2 megaohms⅐cm). Experiments were performed under gas atmospheres of H 2 (Premier Grade, Air Products), 1% H 2 in N 2 (Air Products), or N 2 (BOC gases). To measure inhibition by O 2 , known volume fractions of O 2 (Air Products) were injected into the headspace of the electrochemical cell. In experiments to determine K M H 2 , aliquots of H 2 -saturated temperature-equilibrated buffer solution (of the same composition to that already in the cell) were injected into the cell solution. Concentrations of O 2 or H 2 in solution were calculated using the corresponding Henry's Law constants. In analyzing results from O 2 injection experiments, currents were corrected for film loss and normalized by fitting to the exponential curve given in Equation 1, in which the time (t) dependence of the current, i t , is described in terms of the current at the start (t 0 ) of the data to be fitted (i 0 ), the time constant for film loss (), and the limiting current to which the exponential tends (i ∞ ).

RESULTS
New Protocol for Purification of MBH Proteins from Re H16 and Rm CH34-In Re H16 the genes coding for the MBH subunits, specific accessory proteins, a set of Hyp proteins, and the H 2 -sensing apparatus are clustered in a single operon (Fig. 2), which maps on megaplasmid pHG1 (29,39,40). The gene organization of the Rm CH34 MBH gene cluster, which is located on chromosome 1 (DOE Joint Genome Institute), resembles that of Re H16 with few exceptions (Fig. 2). The hoxZ gene, encoding a b-type cytochrome, is embedded between hupE encoding a putative nickel transporter and a second open reading frame, with unknown function, annotated as vapI. Unlike the situation in Re H16, the Rm CH34 MBH cluster does not harbor an hypX gene or the response regulator gene hoxA. However, both genes are constituents of a gene cluster encoding the soluble, NADreducing [NiFe] hydrogenase, which is located at a different site on chromosome 1 in Rm CH34 (DOE Joint Genome Institute).
To obtain pure hydrogenase for biochemical and electrochemical studies, we exploited an MBH overexpression system that was initially designed for the MBH from Re H16 (29). For production of Rm CH34 MBH, only hoxK and hoxG of the Re H16 MBH operon were replaced by the corresponding genes from Rm CH34 (Fig. 2) using genetic techniques resulting in plasmid pGE615 (see "Experimental Procedures"). This strategy was chosen because the precursors of the MBH small (pre-HoxK) and large (preHoxG) subunits, respectively, of the Re H16 and Rm CH34 MBH proteins are highly similar with an identity of 87 and 81%.
The recombinant plasmid pGE615 carrying the hoxK and hoxG genes of Rm CH34 restored H 2 -dependent autotrophic growth of the Re H16-derived megaplasmid-free strain HF631 (data not shown). This observation indicates that the gene products HoxK and HoxG of Rm CH34 form a catalytically active hydrogenase and, hence, are compatible with the MBH maturation machinery of the host Re H16 whose single components show identities between 37 and 81% to the corresponding proteins of Rm CH34 (Fig. 2).
For straightforward protein purification, a StrepTag II peptide was genetically fused to the C termini of the small subunits of the MBH proteins from Re H16 and Rm CH34 (Fig. 2). Representative results of the purification procedure using Strep-Tactin affinity chromatography are given in Fig. 3 and Table 3 for both MBH proteins. Typically, 60 g (wet weight) of cells harboring the MBH overproduction plasmids yielded ϳ10 mg of nearly homogeneous enzyme with a specific activity of ϳ50 units mg Ϫ1 (Table 3).
Specific activities were routinely obtained in the range 20 -120 units mg Ϫ1 protein. Using the new purification protocol, we confirmed an earlier observation that the pH optimum for MBH H 2 oxidation activity, assayed with methylene blue as electron acceptor, shifts from pH 7.0 to pH 5.5 upon solubilization from the membrane (36) (we attribute the shift in pH optimum to the disconnection of the MBH from its physiological membrane anchor cytochrome b). Although the pH profiles for H 2 oxidation activity differed slightly between MBH preparations of Re H16 and Rm CH34, maximum activity was obtained in both cases at pH 5.5 (supplemental Fig. S1).
Growth Characteristics of Re H16 MBH and Variants Carrying Amino Acid Mutations Close to the Ni-Fe Active Site-To elucidate the origin of the O 2 tolerance of the Ralstonia MBH proteins, amino acid residues in the vicinity of the active site that are unique to this class of hydrogenases were exchanged. In Re H16 MBH, Gly-80 and Cys-81 were exchanged for amino acid residues found at equivalent positions in O 2 -sensitive hydrogenases, as indicated in Fig. 1B. This resulted in the series of exchanges G80Y, C81A, C81V, C81S, C81T, and G80Y/ C81V, which were introduced into the large subunit (HoxG) of Re H16 MBH via site-directed mutagenesis. In addition, Val-77 and Leu-125, which are constituents of the postulated gas channel, were exchanged for bulkier residues, resulting in the exchanges V77I, L125F, and V77I/L125F (as in the RH enzymes from Re H16 and Rhodobacter capsulatus (21,22)), to determine whether the MBH became more tolerant to O 2 .
To test the effect of amino acid alterations on lithoautotrophic growth of Re H16 in the presence of high O 2 concentrations, the strains were cultivated in liquid medium under an H 2 atmosphere containing either 5 or 20% O 2 . The respective growth rates, derived from the logarithmic growth phase, were determined and are summarized in Table 4, as well as MBH activity and protein stability in the membrane.
Wild-type Re H16 responds to the higher O 2 concentration with an ϳ2-fold lower growth rate (0.045 h Ϫ1 compared with 0.079 h Ϫ1 at 5% O 2 ) as determined by optical density measure-  H16 and Rm CH34. The designation of Rm CH34 genes is according to the Re H16 nomenclature. Genes belonging to the same class/function are displayed in the same pattern. Horizontal arrows in the Re H16 gene cluster indicate the region that has been transferred to a broad-host-range vector, yielding the Re H16 MBH overexpression plasmid (28). The DNA fragment indicated by horizontal arrows in the Rm CH34 cluster was introduced into the Re H16 MBH overexpression plasmid, yielding the Rm CH34 MBH overexpression system. Triangles indicate the position at which the StrepTag II coding sequence was fused to the 3Ј end of the respective hoxK gene. The numbers represent the percentage of amino acid residues that are identical in the Re H16 and Rm CH34 gene products. The genes are as follows: hoxKGZ, genes encoding hydrogenase small and large subunits and the cytochrome b; hoxMLOQRTVV, MBHspecific accessory genes; hyp genes are responsible for metallocluster assembly; hoxAJ, genes encoding the response regulator and the histidine protein kinase; hoxBC, genes for the small and large subunit of the regulatory hydrogenase; hoxN and hupE encode nickel permeases. ments. The G80Y mutant strain showed no lithoautotrophic growth on H 2 . This was consistent with Western immunoblot assays revealing only traces of MBH protein in the membrane fraction of mixotrophically grown cells that also lack any MBH activity. Introduction of a second alteration to mimic the D. gigas structure (G80Y/C81V) did not reverse the detrimental effect of the G80Y mutation (Table 4).
In the presence of 5% O 2 , the growth rates of the Cys-81 mutant strains were comparable with that of the wild type (Table 4). At 20% O 2 , however, both C81T and C81A mutant strains displayed significantly lower growth rates than the wild type (0.039 and 0.030 h Ϫ1 ), whereas the C81V mutant was only slightly affected (growth rate 0.051 h Ϫ1 ). Growth of the C81S variant was indistinguishable from wild type.
At 5% O 2 the V77I strain grew similarly to the wild type, whereas the L125F mutant was significantly retarded under these conditions; very slow growth was observed for both mutants at 20% O 2 . The double exchange V77I/L125F abolished the ability to grow lithoautotrophically even at 5% O 2 . These effects are not merely caused by lowering the activity of the MBH in the membrane; anaerobic assays (Table 4) show that although the MBH activity was significantly lower in the V77I mutant than in the L125F strain, the L125F mutant had an even more pronounced growth defect under high O 2 .
Except for the Gly-80 mutations, none of the alterations had a significant effect on the amount of MBH protein in the mem-brane as determined by Western blot, and most of the MBH mutants showed similar H 2 oxidation activity to that of wild type when measured photometrically in an anaerobic, H 2 -saturated assay ( Table 4).
The hydrogenase activity in membranes of the strain expressing Rm CH34 MBH was only about 70% of that determined for Re H16 MBH, which correlated with a lower level of protein as determined by immunological analyses (data not shown). Therefore, the Re H16 strain carrying Rm CH34 MBH was excluded from the H 2 -dependent growth experiments.
Affinity for H 2 in Air, Defining an O 2 Tolerance Factor-We next used PFV to measure the ability of the purified wild-type and mutant hydrogenases to oxidize H 2 in the presence of O 2 . When the electrode, modified with a film of hydrogenase, is immersed in electrolyte solution containing H 2 , the catalytic current reports directly on the activity of the enzyme.
An electrode with adsorbed MBH as described (see "Experimental Procedures") was placed in a sealed cell with gas flowing through the headspace and was rotated at 4500 rpm. Electrocatalytic H 2 oxidation was observed for films of wild-type and all mutant MBH enzymes with the exception of the G80Y and G80Y/C81V variants, for which no current was observed (supplemental Fig. S2). 7 Maximum catalytic activity was attained at pH 5.5, consistent with photometric measurements (supplemental Fig. S1). For measurements of K I(app) O 2 , the electrode was poised at Ϫ8 mV versus SHE (as indicated in supplemental Fig.  S2), a potential sufficiently positive to give a detectable H 2 oxidation current and minimize reduction of O 2 at unmodified regions of the graphite, yet sufficiently negative to avoid anaerobic inactivation (18). A typical experiment for the Re H16 MBH (wild type) is shown in Fig. 4. First, H 2 (100%, corresponding to 0.8 mM in the electrolyte solution) was flushed through the headspace of the cell to obtain the background current. Aliquots of O 2 (1 ml) were then injected into the headspace of the cell (prior to this, the carrier gas flow was halted to prevent the O 2 being immediately flushed out), giving 20% O 2 in the headspace corresponding to a solution concentration of 0.3 mM at 30°C. The current was allowed to stabilize after each injection. We note that relatively high activity is retained even at 0.3 mM O 2 . After the last injection, the cell was again flushed with H 2 to remove O 2 from the cell. The data were corrected for film loss by extrapolating an exponential fit to the anaerobic data points, as described under "Experimental Procedures." A plot of 7 Because noncatalytic redox signals could not be detected for any of the MBH variants, it is not clear whether the variations in current magnitude result from differences in inherent turnover frequency (k cat ) or differing electrode coverages.  We therefore adopted a different approach and used a modification of the direct electrochemical method described by Léger et al. (24), for determining affinity constants for enzymes reacting with gaseous substrates. This concept involves introducing a gas and monitoring the change in catalytic current as this gas is flushed out of the electrochemical cell by a stream of carrier gas. Values of K M H 2 are extracted by analysis of the current time profiles described by Equation 2, which incorporates the Michaelis-Menten equation and the exponential profile for loss of H 2 from solution. The H 2 concentration at zero time (t ϭ 0) is denoted C H 2 ; i max is the maximum catalytic current corresponding to substrate saturation, and is the time constant for the removal of substrate by the carrier gas. In contrast to solution assays in which a soluble electron acceptor is employed, the potential experienced by the enzyme is strictly defined in the electrochemical experiment. Using a slight adaptation of this method, in which gas concentrations were determined by the composition of the gas flowing through the headspace of the sealed electrochemical cell, we determined values for K M H 2 for Rm CH34 and Re H16 wild-type MBH enzymes as well as the genetic variants of Re H16 MBH. Within the window of electrochemical potential over which H 2 oxidation is observed, we chose a sufficiently negative potential, Ϫ0.108 mV versus SHE, to avoid anaerobic inactivation of the Ni-Fe center that occurs at high potential (9). Fig. 5A shows a typical result for Rm CH34 MBH. At a given time, defined as t ϭ 0, a known volume of H 2 -saturated buffer (of identical composition and temperature to that already in the electrochemical cell) was injected to give an initial H 2 concentration of 0.4 mM in solution, causing the electrocatalytic current to rise rapidly. The H 2 concentration immediately begins to drop because of the N 2 flow through the cell headspace, and independent experiments confirmed that the drop in H 2 concentration follows an exponential decay, as indicated in Fig. 5A (right axis). The initial current (i max ) remains fairly stable for more than 200 s indicating the range of H 2 concentrations over which the enzyme is substrate-satu-  rated. The current at time t (i t ) subsequently drops back to its background level as the H 2 concentration falls to zero.
To obtain values of K M H 2 that could be compared across the different MBH enzymes and mutants, a nonlinear least squares regression analysis was used to fit Equation 2 to the raw current versus time data (Fig. 5A, dashed line). From the average of 12 experiments, a value for K M H 2 of 0.57 M was obtained for wildtype Rm CH34 MBH.
A number of factors may contribute to the deviation of the current versus time trace from being a good fit to Equation 2. 8 First, voltammograms of all Ralstonia MBH enzymes (see supplemental Fig. S2) suggest heterogeneity; this is observable as an additional minor local current maximum in the region between Ϫ0.1 and Ϫ0.2 V. The origin of this feature is not known and is currently being investigated. Second, Ralstonia MBH enzymes undergo anaerobic inactivation at high potentials, and there is no region of potential available in which the catalytic current (activity) is independent of potential. This is more pronounced for Re H16 MBH (and most of the mutants) than for Rm CH34 MBH (supplemental Fig. S2). Third, there is the possibility of back diffusion of low levels of H 2 into the cell from the glovebox; one of the assumptions underlying Equation 2 is that the H 2 concentration in the cell tends to 0 over time. Finally, mass transport limitation is unavoidable at extremely low levels of H 2 , even at high rotation rates, and this is not accounted for in Equation 2.
Analogous experiments were performed for wild-type Re H16 MBH (Fig. 5B). In this case there was always a short delay following the injection of H 2 -saturated buffer before the maximum current, i max , was attained. This slow rise in activity is not simply because of gas mixing (as evidenced by the long time constant for this phase; the current continues to rise for ϳ100 s). This rise is also noticeable for Rm CH34 MBH, although it is far less pronounced. Experiments to probe this feature are underway. The initial plateau in the current, reflecting the range of H 2 concentrations over which the enzyme remains substrate-saturated, is shorter for Re H16 than Rm CH34 MBH, indicating that the former has a lower affinity for H 2 . This is reflected in the value of K M H 2 obtained for Re H16 MBH, 6.1 M.

Values of K M
H 2 were also determined at 20°C (Table 5). This procedure was applied to the series of mutants of Re H16 MBH (Fig. 5 and supplemental Fig. S3). All values of K M H 2 are given in Table 5. The H 2 oxidation current versus time profile as H 2 is flushed out of solution gives an immediate qualitative indication of relative affinity for H 2 , provided experimental conditions, including gas flow rate and electrode rotation rate, are held constant. In experiments on some mutants, in particular C81A, L125F, and V77I/L125F, a current plateau was not attained following injection of H 2 -saturated solution, indicating that these variants are not substrate-saturated at 0.4 mM H 2 . 8 It was expected that a logarithmic transform of the data, log((i max /i t ) Ϫ 1) versus t, would yield a straight line, with the y intercept defining K M H2 (24). However, in numerous repetitions of these experiments, it was commonly found that the logarithmic plots deviated from linearity, with distinct kinks in the transformed data; in some plots this tended to two linear phases joined by a short transition phase. This prevented the reliable fitting of a straight line through the transformed data, meaning that K M H2 could not be determined accurately from the y intercept. , indicated by the asterisk, is reached when the current is equal to i max /2. C, the protocol was modified to measure K M H2 for Re V77I/L125F MBH. At t ϭ 0 s, after establishing the headspace atmosphere with a flow of 100% H 2 , the gas flow was changed to 100% N 2 . The lack of any initial current plateau shows that this variant is unsaturated even at 100% H 2 , preventing precise determination of K M H2 . Conditions: E ϭ Ϫ108 mV versus SHE, pH 5.5, temperature 30°C, electrode rotation rate 4500 rpm.

Oxygen-tolerant H 2 Oxidation
To achieve a higher initial H 2 concentration (C H 2 (0)), H 2 was flushed through the headspace of the cell to give an initial concentration of 0.8 mM, and at the time designated t ϭ 0, the gas flow was changed to N 2 to flush out H 2 . Experiments conducted in this way on the wild-type enzymes confirmed that values of K M H 2 are consistent with those obtained using the injection method.
Qualitative analysis of the current versus time profiles for the C81A and V77I/L125F MBH enzymes (Fig. 5C and

DISCUSSION
We have carried out an integrated physiological, biochemical, and electrochemical analysis of the catalytic properties of the membrane-bound [NiFe] hydrogenases derived from the two prominent aerobic H 2 oxidizers R. eutropha H16 and R. metallidurans CH34. For hydrogenase purification, overproduction plasmids were used that contain the entire genetic information for synthesis and maturation of catalytically active enzyme (29). The genes encoding the small subunit were equipped with a StrepTag II sequence allowing a mild and easy one-step purification of the fully matured, heterodimeric MBH protein. Purification of active CH34 MBH was achieved simply by exchanging the structural genes of Re H16 with the respective counterparts from Rm CH34. The heterologous production of active [NiFe] hydrogenase has been reported for several examples (43,44). We have demonstrated here the cross-compatibility of a protease that is normally highly substrate-specific; i.e. the endopeptidase of Re H16 MBH was able to process, efficiently, the premature large subunit of Rm CH34 MBH. Furthermore, the Rm CH34 MBH is entirely compatible with the Re H16 maturation apparatus and functionally interacts with the MBH-specific cytochrome b of Re H16. These observations are probably because of the fact that the H16 and CH34 MBH subunits share more than 80% identity. However, immunological analysis revealed that the content of the CH34 MBH was lower in the membrane of Re H16 than that of the indigenous H16 MBH indicating a less strong connection to the primary electron acceptor, the cytochrome b HoxZ of Re. For this reason we did not include the Re strain carrying the CH34 MBH in our growth experiments.
In vivo, the [NiFe] hydrogenases of aerobic, H 2 -oxidizing Knallgas bacteria permanently face challenging conditions. The H 2 concentrations in their habitats are usually extremely low (possibly 1-10 nM (45), equivalent to 1-10 ppm gas fraction), whereas the O 2 concentration is close to the 21% prevailing in the atmosphere. Consequently, the H 2 -converting enzymes of Knallgas bacteria must be highly selective for H 2 with a concomitant high tolerance toward O 2 , although the origin of this selectivity is unclear. The enzymes clearly do react rapidly with O 2 , even though they are not completely inhibited in its presence. (Note that in Fig. 4 the time taken for the current to stabilize at each stage is determined by the gas mixing time following injections of O 2 into the headspace of the cell and not directly into the cell solution.) Therefore, the origin of their O 2 tolerance cannot simply be a physical exclusion of O 2 from the active site. The conserved nature of the amino acid residues around the active site between O 2 -sensitive and O 2 -tolerant hydrogenases, and the consistent ligation of the iron and nickel atoms, raises interesting questions about the origin of the differences in affinities for H 2 and sensitivity to O 2 of hydrogenases from different organisms.
The term that we refer to as K I(app) should be regarded as giving an empirical measurement of O 2 tolerance (a high K I(app) O 2 implies a high O 2 tolerance). In addition to temperature and pH, K I also depends not only on partial pressure of H 2 but also on electrode potential. In principle, as the partial pressure of H 2 tends toward 0, K I(app) O 2 becomes a realistic measurement of the O 2 tolerance that is likely to prevail in vivo.
In an early investigation, Schink and Probst (46) presented evidence that O 2 inhibits the Re H16 MBH activity in a competitive manner. However, it is most unlikely that inhibition by O 2 is this straightforward as it is now established that O 2 reacts with the Ni-Fe active site in a range of hydrogenases to produce chemically modified states that require reductive re-activation at well defined potentials (9,47). Reaction of O 2 with standard [NiFe] hydrogenases gives a mixture of Ni-A and Ni-B inactive states. However, in recent spectroscopic investigations on Re H16 MBH, reaction with O 2 produced only Ni-B; formation of the O 2 -inhibited Ni-A state was not observed under any conditions. 6 This evidence, combined with our electrochemical data, leads to the model shown in Fig. 6. Although O 2 cannot be viewed as a simple competitive inhibitor in the accepted sense, H 2 and O 2 may thus be regarded as "competitive substrates": the O 2 -inactivated forms are recovered by reduction, with the net result that O 2 is reduced (and is therefore, technically, a substrate). The turnover frequency for this "oxidase" activity is likely to be limited by the rate-determining reactivation of Ni-B that completes the competing cycle. 9 A high affinity for H 2 is likely to keep the MBH enzyme in the productive H 2 oxidation cycle. Experiments have shown that the Ralstonia MBH enzymes are reductively reactivated much more rapidly, and at a higher potential, than standard O 2 -sensitive [NiFe] hydrogenases such as the enzymes from A. vinosum and D. gigas (9,48 tivity (21,22). In wild-type Re H16 MBH, Leu and Val are present at the equivalent positions, and mutation to Phe and Ile, respectively, was therefore expected to confer increased O 2 tolerance. Interestingly, the double exchange brought about a 100fold increase of K M H 2 relative to the wild-type enzyme, whereas the catalytic activity was only reduced 3-fold, consistent with a restricted gas flow to the catalytic center. However, the mutant proteins were actually more O 2 -sensitive than the wild-type protein, as expressed by the 4-fold decrease in K I(app) . In a very recent study the same exchange strategy was applied to proposed gas channel of the standard [NiFe] hydrogenase from D. fructosovorans (57). The corresponding V74I/L122F mutant also showed a significant increase in the K M H 2 but concomitantly exhibited a 10-fold higher tolerance toward CO (which is a potent inhibitor for the D. fructosovorans wild-type enzyme (57) but not for Re MBH (20)).
The experiments presented in this study have explored several aspects of the issue of "O 2 tolerance" in [NiFe] hydrogenases. Although it is still not possible to directly identify molecular factors that render the MBH enzymes from Ralstonia significantly less O 2 -sensitive than corresponding enzymes from other species, a number of conclusions can be drawn. is sensitive to H 2 concentration, but this is not always the case (the correlation does not hold for the C81A and C81S mutants).
The O 2 tolerance of Ralstonia MBH enzymes cannot be simply linked to single point mutations in the vicinity of the active site; mutants of Re H16 MBH in which the closest varying residues were exchanged for those found in O 2 -sensitive hydrogenases did not show significantly enhanced O 2 sensitivity compared with the wild type. Tolerance to O 2 is clearly a complex factor and is determined by a well adapted spatial and electronic structure of the active site rather than a simple restriction of diffusion of inhibitory gases such as O 2 . Further studies to investigate kinetic and thermodynamic details of the reactions of the active site with H 2 and O 2 are required.