How Escherichia coli Is Equipped to Oxidize Hydrogen under Different Redox Conditions*

The enterobacterium Escherichia coli synthesizes two H2 uptake enzymes, Hyd-1 and Hyd-2. We show using precise electrochemical kinetic measurements that the properties of Hyd-1 and Hyd-2 contrast strikingly, and may be individually optimized to function under distinct environmental conditions. Hyd-2 is well suited for fast and efficient catalysis in more reducing environments, to the extent that in vitro it behaves as a bidirectional hydrogenase. In contrast, Hyd-1 is active for H2 oxidation under more oxidizing conditions and cannot function in reverse. Importantly, Hyd-1 is O2 tolerant and can oxidize H2 in the presence of air, whereas Hyd-2 is ineffective for H2 oxidation under aerobic conditions. The results have direct relevance for physiological roles of Hyd-1 and Hyd-2, which are expressed in different phases of growth. The properties that we report suggest distinct technological applications of these contrasting enzymes.

Hydrogenases catalyze the reversible cleavage of H 2 into protons and electrons, and play an important role in the energy metabolism of a broad range of microorganisms (1). Hydrogenases are classified according to their active site metal ion content, and three phylogenetically distinct classes have so far been identified: di-iron [FeFe]-, nickel-iron [NiFe]-, and mono-iron [Fe]-hydrogenases (1). Nickel-iron hydrogenases are the most abundant of the three types (1), and many members of this class are membrane bound, with the membrane-extrinsic domain consisting of a large subunit containing the active site, and a small subunit accommodating one to three electron-transferring iron-sulfur clusters. The active sites of [NiFe]-hydrogenases contain a nickel atom coordinated by four cysteine-S ligands, two of which bridge to an iron atom that is further coordinated by three unusual diatomic ligands, two cyanides and one carbonyl (2).
Hydrogenases are inactivated by O 2 : reaction of the [NiFe] active site with O 2 results in at least two inactive Ni(III) forms, known as Ni-A ("unready") and Ni-B ("ready"). The ratio of these forms depends upon the conditions (especially the availability of electrons) under which O 2 attacks, and varies from one enzyme to another (3,4). The unready state appears to contain a product of partial reduction of O 2 (a coordinated peroxide has been assigned (2,5)) and it is re-activated only slowly upon reduction in the presence of H 2 (3). In contrast, the ready state contains a bridging hydroxo ligand (2, 6, 7) (the product of complete reduction of O 2 ) and is rapidly reactivated upon reduction (8).
Hydrogen plays a critical metabolic role in bacteria ranging from the strictly anaerobic, such as Desulfovibrio species (9), to the facultative aerobic, such as Ralstonia species (10). Furthermore, microaerophilic and facultative aerobic human pathogens, such as Helicobacter and Salmonella, respectively, are now known to require H 2 oxidation activity for virulence (11,12). Thus, whereas the majority of well characterized [NiFe]hydrogenases are inactivated by very low levels of molecular O 2 (4), in certain species there is a requirement for H 2 oxidation to occur in microaerobic and even fully aerobic environments. Hydrogenases capable of operating under aerobic conditions are defined as "O 2 tolerant" (13). To be O 2 tolerant a hydrogenase requires certain mechanistic features for dealing with O 2 attack, such as suppressing O 2 access to the active site (14), avoiding the formation of the unready state (Ni-A) and ensuring that any ready state (Ni-B) formed is reactivated rapidly (13).
The enterobacterium Escherichia coli synthesizes at least three [NiFe]-hydrogenases (15). Hydrogenase-3 (Hyd-3) is a cytoplasmic enzyme and forms part of the membrane-anchored formate hydrogen lyase system, which is responsible for H 2 evolution under fermentative conditions (16). In contrast, both Hyd-1 and Hyd-2 are membrane-bound periplasmic proteins, and are believed to function as "H 2 -uptake" enzymes (17,18). Hyd-1 is encoded by the first two genes of the hyaABCDEF operon, with hyaA coding for the [FeS] cluster-containing ␤-subunit and hyaB coding for the ␣-subunit, which binds the [NiFe] active site (19). Similarly, Hyd-2 is encoded by genes hybO and hybC of the hybOABCDEFG operon, with hybO coding for the ␤-subunit and hybC coding for the ␣-subunit (20). A membrane-anchoring hydrophobic helix is located at the extreme C terminus of the ␤-subunit of each enzyme. Hyd-1 couples H 2 oxidation to quinone reduction via the HyaC transmembrane cytochrome b (19), whereas Hyd-2 utilizes a periplasmic ferredoxin (HybA) to reduce quinone via the HybB integral membrane protein (21).
Regulation of the expression of Hyd-1 and Hyd-2 is distinct and complex. According to the available data, syntheses of both enzymes are repressed under aerobic conditions (22), but significant differences exist in the finer control of production. Repression by nitrate (23) or phosphate (24), stationary phase induction (25), and induction by acidic conditions (26) are all reported for the hya operon. In contrast, the hyb operon is expressed most strongly during the exponential phase of growth, and its expression is possibly catabolite repressed (23).
It has been proposed that the physiological role of Hyd-2 is to catalyze the oxidation of extraneous H 2 , with the resulting electrons eventually being used by fumarate reductase to produce succinate (15). Thus Hyd-2 allows E. coli to use H 2 as an energy source during growth on non-fermentable carbon compounds. The physiological role of Hyd-1 is yet to be firmly established, but is believed to be quite distinct from that of Hyd-2. A suggested function of Hyd-1 is the recycling of H 2 produced by Hyd-3 under fermentative conditions (16). Additionally, evidence has emerged to suggest that Hyd-1 might be involved in linking H 2 oxidation to O 2 reduction under certain conditions (27). Indeed, a fragment of Hyd-1 has been purified from aerobically grown E. coli, implying that conditions might exist under which this enzyme is expressed in the presence of high concentrations of O 2 (28).
In protein film electrochemistry (PFE), 4 an extremely small quantity of redox protein is adsorbed directly onto an electrode to give a mono-or submonolayer film. A wealth of information can be obtained from enzymes, provided that full catalytic and electron-transfer activities are maintained in the adsorbed state (29). The electrocatalytic activity of adsorbed enzymes responds directly to the electrode potential, and is monitored through the electrical current recorded in the presence of substrate. The current relates directly to the turnover frequency of the enzyme at the particular potential applied. Rates of change of catalytic state induced by stepping the electrode potential or introducing an inhibitor are measured from a current-time course. The environment of the adsorbed enzyme typically consists of 2 ml of enzyme-free buffer/electrolyte solution in a sealed electrochemical cell, through which precise mixtures of gases can be passed. Rotation of the electrode at controlled high speeds ensures that the observed reactions are not diffusion controlled, and also allows product inhibition to be quantified and suppressed (30).
In this article we describe the characterization of E. coli hydrogenases 1 and 2 and report striking differences in the catalytic properties of the two enzymes, most notably their contrasting activities under the different redox conditions and oxidizing environments likely to be encountered by the organism.

EXPERIMENTAL PROCEDURES
Bacterial Strains and Growth Conditions-The E. coli K-12 strains used in this study were FTH004 and FTH013, which are based on MC4100 (31). FTH004 (21) carries an engineered hyaABCDEF operon encoding a modified HyaA protein bearing a His 6 affinity tag at its extreme C terminus, whereas FTH013 carries an engineered hybOABCDEFG operon encoding a similarly modified HybO protein. FTH013 was constructed by PCR-amplifying a 500-bp fragment covering the 3Ј end of the hybO gene (minus stop codon), digesting with XbaI and BglII, and cloning into pFAT210 (21). Next, a 500-bp fragment covering the 5Ј region of the hybA gene (including the hybA ribosome binding site) was amplified by PCR, digested with ClaI and KpnI, and cloned into the pFAT210 construct. The hybO his allele now present on pFAT210 was then moved as an XbaI-KpnI fragment onto pMAK705 (32) before being transferred to the chromosome of MC4100 (32). The chromosomal modifications were carefully constructed so as to preserve identifiable regulatory elements, coding sequences, stop codons, and ribosome binding sites of genes flanking the hyaA and hybO loci.
FTH004 and FTH013, respectively, produce His-tagged Hyd-1 and His-tagged Hyd-2 at wild-type levels under native regulatory and biosynthetic control. The bacteria were cultured anaerobically at 37°C in LB medium supplemented with 0.5% (w/v) glycerol and 0.4% (w/v) sodium fumarate. The inoculum (1%) was grown aerobically at 37°C in LB medium. To isolate Hyd-1, cells were harvested during the stationary phase (centrifugation at 3,500 ϫ g for 15 min at 4°C). To isolate Hyd-2, the cells were harvested similarly, but during late exponential phase.
Isolation of the Membrane-bound Hydrogenases-The procedure for isolating His-tagged hydrogenase was essentially the same for both Hyd-1 and Hyd-2. All steps were carried out at 4°C or on ice. The pelleted cells (FTH004 or FTH013) were resuspended in 100 mM Tris, pH 7.5, 1 mM EDTA, 50 mM NaCl, supplemented with 10 g ml Ϫ1 of DNase I and 50 g ml Ϫ1 of lysozyme (both from Sigma), and Complete EDTA-free protease inhibitor mixture tablets (Roche Molecular Biochemicals) at the manufacturer's recommended concentration.
The cells were disrupted by three passages through a French pressure cell, and crude membrane fractions were prepared by differential centrifugation as described (33). The pelleted membrane fraction was resuspended in 10 ml of 100 mM Tris, pH 7.5, 50 mM NaCl, and the suspension adjusted to a protein concentration of 10 mg ml Ϫ1 in the same buffer. Protease inhibitor mixture tablets were again added at the appropriate concentration.
The detergent n-dodecyl-␤-D-maltoside was added to the suspension at a concentration of 1 mg/mg of protein, and the membranes were solubilized by gentle agitation at 4°C overnight. Insoluble material was removed by ultracentrifugation at 150,000 ϫ g for 1 h. Dithiothreitol was added to the cleared supernatant (giving 1 mM), and the solution was applied to a Ni 2ϩ -loaded 5-ml HisTrap Chelating HP column (Amersham Biosciences) equilibrated in 20 mM Tris, pH 7.5, 150 mM NaCl, 50 mM imidazole, 1 mM dithiothreitol, 0.02% (w/v) n-dodecyl-␤-D-maltoside (buffer A). The column was washed with 50 ml of buffer A and then developed in 40 ml of the same buffer using a linear gradient of imidazole to a final concentration of 500 mM. Fractions containing purified Hyd-1 or Hyd-2 as judged by SDS-PAGE followed by Coomassie Blue staining were pooled and subjected to two rounds of dialysis against 500 ml of buffer A containing no imidazole. The purified protein was then con-centrated by ultrafiltration, snap-frozen, and stored in liquid nitrogen.
Spectrophotometric Assays of Hydrogenase Activity-To measure H 2 uptake, the reduction of benzyl viologen (1 mM) (⑀ ϭ 7.4 mM Ϫ1 cm Ϫ1 ) was followed at 604 nm in anaerobic cuvettes containing 1 ml of 50 mM Tris/HCl, pH 7.0, 100 mM NaCl, saturated with H 2 . Appropriate amounts of purified enzyme (5-50 g) were injected to initiate the reaction. To measure H 2 production, the oxidation of methyl viologen (1 mM, pre-reduced with an equivalent of sodium dithionite) (⑀ ϭ 13.7 mM Ϫ1 cm Ϫ1 ) was followed at 604 nm in anaerobic cuvettes containing the same buffer, saturated with N 2 .
Protein Film Electrochemistry Experiments-Pyrolytic graphite "edge" electrodes were constructed in-house (30). The pyrolytic graphite edge electrode surface is rich in hydrophilic C-O groups that bind many enzymes strongly with full retention of activity (34,35). The electrode (geometric surface area 0.03 cm 2 ) was prepared by sanding with P800 Tufbak Durite sandpaper. Enzyme solution (2 l at ϳ20 M) was then pipetted onto the surface and, after 15 s, the electrode was held under a stream of water to remove non-adsorbed enzyme. All solutions used in the electrochemical cell are enzyme free, and therefore only enzymes adsorbed on the electrode and under the direct control of the electrode potential was examined. A mixed buffer system was used (8).
Electrochemical experiments were conducted in an allglass cell located in an anaerobic glove box (Vacuum Atmospheres, O 2 Ͻ2 ppm). The saturated calomel reference electrode (SCE) was housed in a side-arm linked to the main compartment by a Luggin capillary. All potential values were corrected to the standard hydrogen electrode (SHE) scale using the relationship E SHE ϭ E SCE ϩ 0.241 V at 25°C (36). The cell was sealed to a glass head fixed around an electrode rotator (EG&G). Gas inlet and outlet connectors allowed gas flow through the headspace. The electrode was rotated at variable high speed to control mass transport and ensure rapid equilibration of the cell solution with the head gas atmosphere. Precise gas mixtures were made using flow meters (Sierra Instruments). The total gas flow rate was typically 1000 standard cubic centimeters min Ϫ1 . The cell was water-jacketed and thermostated by a water circulator (Neslab). Kinetic analyses of hydrogenase inhibition by product and CO are described under supplemental data.
EPR Measurements and Data Analysis-Continuous wave EPR experiments were performed either using an X-band (9 -10 GHz) Bruker EMX spectrometer (Bruker BioSpin GmbH, Germany) equipped with an X-band Super High Sensitivity Probehead (Bruker), or a W-/X-band Bruker Elexsys 680 spectrometer using an X-band EMX High Sensitivity Probehead (Bruker), both equipped with a low temperature helium flow cryostat (Oxford Instruments CF935). For determining g values, the magnetic field was calibrated at room temperature with an external 2,2-diphenyl-1-picrylhydrazyl standard (g value 2.0036). Data analysis and simulations of the EPR spectra were performed using the program EasySpin (37).

Isolation of Hydrogenase-1 and Hydrogenase-2 from E. coli-
His 6 -tagged Hyd-1 and Hyd-2 were expressed from the chromosome of E. coli strains FTH004 and FTH013, respectively, at wild-type levels. Hyd-1 was purified from the solubilized membrane fraction to near homogeneity in a single affinity chromatography step, with a yield of ϳ0.15 mg of protein (g of cells) Ϫ1 . Hyd-2 was purified in a complex with its physiological redox partner HybA, again with low levels of contamination, with a yield of ϳ0.015 mg of protein (g of cells) Ϫ1 . SDS-PAGE analyses of both enzymes are shown in supplemental Fig. S1. Spectrophotometric activity assays of purified Hyd-1 and Hyd-2 yielded specific activities for H 2 uptake of 1 mol of H 2 min Ϫ1 mg Ϫ1 and 34 mol of H 2 min Ϫ1 mg Ϫ1 , respectively. No H 2 production by Hyd-1 was detected, but a specific activity of 2.5 mol of H 2 min Ϫ1 mg Ϫ1 was measured for Hyd-2.
Cyclic Voltammetry Experiments on Hyd-1 and Hyd-2-Both Hyd-1 and Hyd-2 are highly active in catalyzing H 2 oxidation. Fig. 1 shows electrochemical experiments on Hyd-1 and Hyd-2, each adsorbed as a film on a pyrolytic graphite edge electrode, at pH 6.0, 30°C under three different H 2 partial pressures (10, 3, and 0.3% H 2 in argon). The cyclic voltammograms were initiated at Ϫ0.560 V and the electrode potential was steadily increased to ϩ0.244 V at a rate of 1 mV s Ϫ1 . The scan direction was then reversed, and the potential lowered back to Ϫ0.560 V. Positive current is a direct measure of the net rate of steadystate H 2 oxidation (H 2 3 2H ϩ ϩ 2e Ϫ ), and negative current, similarly, provides a direct, net measurement of the reverse reaction, H 2 production (30). The underlying wave shapes for electrocatalysis by enzymes have been explained previously (38,39).
Important differences between the catalytic activities of the two enzymes are immediately apparent. The first of these differences relates to catalytic "bias" (the inherent ability to catalyze in one direction versus the other): whereas no H 2 production by Hyd-1 can be detected, Hyd-2 shows significant catalysis of this process.
The second difference relates to the ease of re-activation of the oxidized, inactive ready (Ni-B) state, which is formed under anaerobic oxidizing conditions (8,40). The Ni-B state contains Ni(III) with a OH Ϫ ligand in a bridging position (2). When the potential is scanned from a high (positive) value back to a low (negative) value, Ni-B reductively re-activates (a one-electron reduction of Ni(III) to Ni(II)); this reactivation is observed as an increase in current on the reverse scan, marked with asterisks in Fig. 1, top panel. The midpoint potential of reactivation is defined by the value E switch determined from the derivative di/dE (8). At pH 6.0, E switch is ϩ0.150 V for Hyd-1 and Ϫ0.085 V for Hyd-2, showing that the inactive Ni-B state is much less favored thermodynamically in Hyd-1 than in Hyd-2. Therefore, at equilibrium at a potential of, for instance, 0 V, Hyd-1 would exist entirely in its active state, whereas Hyd-2 would be entirely in its inactive state.
The current trace for Hyd-2 crosses the zero line sharply and, under 10% H 2 , this occurs at a potential of Ϫ0.33 V, the value predicted by the Nernst equation for the 2H ϩ /H 2 redox couple (the reversible cell potential) under these conditions. However, trace for Hyd-1 shows that H 2 oxidation under 10% H 2 becomes apparent only above a potential of Ϫ0.28 V, significantly more positive than the 2H ϩ /H 2 cell potential. Hyd-1 therefore requires an overpotential of ϳ0.05 V to begin catalysis of H 2 oxidation under these conditions. Using the Hyd-2 voltammogram as a reference for the reversible cell potential, a close examination of Fig. 1 shows that the overpotential required for H 2 oxidation by Hyd-1 decreases as the partial pressure of H 2 is lowered.
A comparison of all three panels of Fig. 1 also reveals the sensitivity of Hyd-1 and Hyd-2 to changes in the partial pressure of H 2 . Oxidation of H 2 by each enzyme is affected to a similar degree by changes in substrate concentration, and this implies that Hyd-1 and Hyd-2 have H 2 affinities of the same order of magnitude. Indeed, by carrying out Hanes analyses on data recorded at seven different H 2 concentrations, K m values of 9 Ϯ 1 and 17 Ϯ 4 M were determined for Hyd-1 and Hyd-2, respectively, at Ϫ0.175 V (see supplemental Fig. S2). From separate experiments to examine product inhibition of the H 2 production activity of Hyd-2, the inhibition constant K i H 2 was measured to be 210 Ϯ 19 M at pH 6.0, 30°C, and Ϫ0.600 V (see supplemental Fig. S3). H 2 is therefore a weak product inhibitor of Hyd-2. Hyd-1 shows negligible H 2 production activity, even under a 100% argon atmosphere. Hyd-1 is thus exclusively a H 2 oxidizer, whereas Hyd-2 is, at least in vitro, a bidirectional hydrogenase.
Electron Paramagnetic Resonance Spectroscopy and PFE Analysis of As-isolated Hyd-1 and Hyd-2- Fig. 2 shows the X-band continuous wave EPR spectra of as-isolated samples of Hyd-1 and Hyd-2, each recorded at 10 K and 80 K. At 10 K, both Hyd-1 and Hyd-2 spectra are dominated by the strong signal of the [3Fe-4S] ϩ cluster with g values (Hyd-1/Hyd-2: g x ϭ 2.02/ 2.03, g y ϭ 2.01/2.02, g z ϭ 2.00/2.01) consistent with those reported for [NiFe]-hydrogenases (e.g. Refs. [41][42][43][44][45][46]. The spectrum of Hyd-1 is more complicated than that of Hyd-2, and we discuss this below. At 80 K, only the EPR signals arising from nickel are observable, and can be examined without interference from the [3Fe-4S] ϩ cluster, which relaxes rapidly at this temperature. The spectra show that both Hyd-1 and Hyd-2 form Ni-A and Ni-B states upon aerobic isolation. The rhombic signals characteristic of both Ni-A and Ni-B show identical g values (to two decimal places) for Hyd-1 and Hyd-2 (Ni-A, g x ϭ 2.31, g y ϭ 2.24, and g z ϭ 2.01; Ni-B, g x ϭ 2.31, g y ϭ 2.16, and g z ϭ 2.01). These values are in good agreement with those previously reported for [NiFe]-hydrogenases (6), although we consistently observed relatively more Ni-B than was evident in the data presented by DerVartanian et al. (47). The exact ratio of Ni-A to Ni-B varies between preparations: in the spectra shown, the ratio is ϳ1:3 for Hyd-1 and 1:4 for Hyd-2.
The spectra of Hyd-2 contain no signals apart from those arising from Ni-A, Ni-B, and [3Fe-4S] ϩ , and are similar to those of the [NiFe]-hydrogenases from Desulfovibrio sp. (46, 48 -50). In contrast, the spectra of Hyd-1 contain several other peaks around the [3Fe-4S] ϩ cluster region at g ϭ 2.06, 1.96, 1.93, 1.91, and 1.87 (denoted with an asterisk in Fig. 2) as well as around the peaks arising from Ni-A and Ni-B at g ϭ 2.36, 2.27, and 2.21 (indicated with an open circle). These features have been observed previously for [NiFe]-hydrogenases from Allochromatium vinosum (41,(51)(52)(53), Thiocapsa roseopersicina (42), and most recently Ralstonia eutropha (43). Interestingly, the peaks denoted with the open circle in Fig. 2 at the low-field end of the 10 K spectrum (which do not overlap with the [3Fe-4S] ϩ cluster signals) are still present at 80 K, and thus there is no As indicated in the top panel, the negative current is a direct measurement of enzyme-catalyzed H 2 production, and the positive current is a direct measure of enzyme-catalyzed H 2 oxidation. Maximum currents were normalized to those of the same sample under 100% H 2 , which was measured prior to each voltammogram. Blank voltammograms, not shown, recorded using a bare pyrolytic graphite edge electrode with no enzyme adsorbed on the surface, showed that no H 2 oxidation or H ϩ reduction occurs directly on the electrode surface. SHE, standard hydrogen electrode. FEBRUARY 5, 2010 • VOLUME 285 • NUMBER 6 evidence of a spin-spin interaction between the [NiFe] center and the [3Fe-4S] ϩ cluster. Fig. 3 shows plots of current against time (chronoamperometry), which reveal reductive activation of as-isolated samples of Hyd-1 and Hyd-2 during their first exposure to H 2 after aerobic purification. The measurements were made at pH 6.0, 30°C, by holding the electrode potential at a value more nega-tive than E switch (Ϫ0.060 V was chosen for Hyd-1, and Ϫ0.200 V for Hyd-2), to drive the reductive reactivation process. In electrochemical experiments Ni-A is identified by slow activation kinetics (hence the term unready), whereas Ni-B always activates much more rapidly (3). Such experiments are therefore complementary to the EPR described above for comparing the relative amounts of Ni-A and Ni-B. Activation of both Hyd-1 and Hyd-2 is biphasic: the first phase is fast and includes the activation of Ni-B, whereas the second phase is much slower and represents activation of Ni-A. Therefore, both EPR data and PFE experiments show that as-isolated Hyd-1 and Hyd-2 both contain Ni-A and Ni-B, and again, the ratio of these two species varies between preparations. Such chronoamperometry experiments provide only a qualitative comparison of the as-isolated enzymes, because the current-time relationship is complicated by the rate at which the enzyme becomes favorably oriented on the electrode following its application, because of relatively rapid film loss immediately after adsorption, and also because of the need to commence measurements as soon as the electrochemical cell was sealed, and before gas concentrations had equilibrated.

How E. coli Is Equipped to Oxidize Hydrogen
A Comparison of the Effects of O 2 at High Potential on Active Hyd-1 and Hyd-2-The cyclic voltammograms in Fig. 4A show that fully activated Hyd-1 responds to high-potential O 2 exposure in a dramatically different manner to Hyd-2. The left-hand voltammogram shows the response of Hyd-1 to O 2 under an atmosphere of 100% H 2 . During the forward scan, O 2 -saturated buffer was injected into the cell solution when the electrode potential reached 0.03 V, to give a total O 2 concentration of 144 M. A sudden decrease in the current is observed, but the H 2 oxidation activity does not fall to zero. The decrease is attributable partly to enzyme inhibition, and partly to temporary dilution of H 2 in the cell solution. Importantly, there is a rapid recovery of activity, even though the sweep is continuing in the direction of more positive potential (see Fig. 4, asterisk), before the activity again begins to decrease. The continuing flow of H 2 through the cell headspace ensures that by the time the electrode potential has reached ϩ0.244 V, almost all of the injected O 2 has been flushed from the cell solution. On the reverse sweep (to negative potential) the activity increases and reaches a maximum at ϳ0 V. The E switch value of ϩ0.150 V for recovery of Hyd-1 from aerobic inactivation is the same as that recorded  during recovery from anaerobic inactivation (Fig. 1); in both cases, the scan rate is very low, thus approaching as closely as possible reversible conditions. The cyclic voltammogram on the right-hand side of Fig. 4A shows an experiment carried out on Hyd-2, in which all conditions were identical to those for Hyd-1 except for the amount of O 2 added to the electrochemical cell; this time, O 2 -saturated buffer was injected to give an O 2 concentration of 16 M, only one-ninth of the O 2 concentration to which Hyd-1 was exposed. The response of Hyd-2 shows that this enzyme is much more sensitive than Hyd-1 to O 2 . Immediately upon addition of 16 M O 2 at 0.03 V, the H 2 oxidation activity of Hyd-2 falls to less than 15% of the pre-O 2 activity. As the potential is scanned to more positive values and O 2 is flushed from the electrochemical cell, no recovery of Hyd-2 activity occurs. Furthermore, upon reversal of the scan direction, Hyd-2 does not reactivate until a much lower potential has been reached (Ϫ0.085 V) than that required for recovery of Hyd-1. Again, this E switch value is the same as that measured for recovery of Hyd-2 from the anaerobically inactivated state (Fig. 1). Similar experiments, the results of which are shown in supplemental Fig. S4, reveal that exposure of each enzyme sample to 42 M O 2 at high potential results in about 40% inhibition of Hyd-1 and 100% inhibition of Hyd-2. Fig. 4 also shows potential step kinetic experiments that distinguish between differences in the composition of the O 2 -inhibited states of Hyd-1 and Hyd-2, generated when active enzyme is exposed to O 2 at high potential, either under argon or H 2 . Unlike the experiments in Fig. 3, the enzyme films have had time to become established on the electrode. In Fig. 4B, samples of Hyd-1 or Hyd-2 were first held under a headgas atmosphere of 100% H 2 at a potential significantly below E switch but at which considerable H 2 oxidation activity can be observed (Ϫ0.060 V was chosen for Hyd-1 and Ϫ0.175 V for Hyd-2). This provided an initial measure of the catalytic current recorded from the particular enzyme film being studied. The headgas in the electrochemical cell was then changed from H 2 to argon, and the system was left for 800 s to equilibrate, still under control of the electrode potential; the current decreased as the substrate was purged from the solution, and reached a steady value by 1000 s. Because the K m H 2 of the enzymes is so low and glove box pressure was raised to permit injection of inhibitor, unavoidable leakage of trace amounts of H 2 into the electrochemical cell results in this steady-state current being higher than 0 A. The potential was next stepped to ϩ0.344 V, and O 2 -saturated buffer was immediately injected to give a final O 2 concentration of 325 M in the cell solution. It was important to ensure that O 2 injection occurred exactly simultaneously with the step to high potential, as the aim of the experiment was to analyze the effect of O 2 on active enzyme; this makes it essential that no anaerobic inactivation occurs prior to O 2 exposure. The system was next held at ϩ0.344 V for 1200 s, during which time the electrochemical cell re-equilibrated with H 2 and all O 2 was flushed from the cell solution. Finally, the potential was stepped back below E switch (to Ϫ0.060 V for Hyd-1, and Ϫ0.175 V for Hyd-2), and the kinetics of reactivation of each enzyme were monitored.
Reactivation of both enzymes occurs in two phases, the first being very fast and corresponding to reactivation of Ni-B. In the case of Hyd-1, this accounts for ϳ85% of the increase in current. The second phase (ϳ15%) represents the much slower reactivation of Ni-A, with a rate of 2 ϫ 10 Ϫ3 s Ϫ1 (t1 ⁄ 2 340 s at 30°C). In the case of Hyd-2, only 55% of the observed reactivation is characteristic of Ni-B recovery, and the remaining reactivation occurs over a period of several hours and is characteristic of Ni-A recovery. Thus, under these identical O 2 exposure conditions (in the absence of H 2 ), Hyd-2 is more than twice as prone as Hyd-1 to forming the Ni-A state.
Similar experiments in which O 2 was injected under an atmosphere of H 2 instead of argon are shown in Fig. 4C. The presence of H 2 , and the consequential greater electron availability, is known to enhance formation of Ni-B over Ni-A (3). The amount of Ni-A formed by Hyd-1 under these conditions is negligible, and although Hyd-2 still shows some Ni-A reactivation when the potential is lowered, the ratio Ni-A:Ni-B is less than one-third of that following O 2 exposure under a H 2 -free atmosphere.
The Activity of Hyd-1 under Aerobic Conditions-The ability of Hyd-1 to reactivate from rapid O 2 inhibition at high potential, in conjunction with its low K m H 2 value, suggests that there should be a window of potential within which the enzyme can maintain H 2 oxidation under a low H 2 partial pressure and 20% O 2 . Fig. 5 shows an experiment that proves that Hyd-1 is an O 2 -tolerant H 2 oxidizer. The cyclic voltammograms were recorded at 1 mV s Ϫ1 , pH 6.0, 37°C, under three different gas mixtures: 10% H 2 , 90% argon; 10% H 2 , 20% O 2 , 70% argon; and 20% O 2 , 80% argon. Before measurements were made, the cell was allowed to equilibrate for 15 min with the appropriate gas mixture at Ϫ0.560 V. Under 10% H 2 at Ϫ0.1 V, ϳ40% of the H 2 oxidation current measured in O 2 -free conditions was sustained in the presence of 20% O 2 . This value is actually a slight underestimate, because at potentials more negative than 0 V there is a negative contribution to the current due to direct reduction of O 2 at the graphite electrode surface, as shown in the scan recorded under 20% O 2 , 80% argon.
The Effect of CO on the Activity of Hyd-1 and Hyd-2-The classic inhibitor of H 2 -cycling catalysts is CO (54). Fig. 6 shows experiments in which the inhibition constant K i (CO/H 2 ) was measured for Hyd-1 (panel A) and Hyd-2 (panel B) during H 2 oxidation at a fixed potential below E switch , pH 6.0, 20°C. Under a constant headgas atmosphere of 20% H 2 in argon, the concentration of CO was increased stepwise from 0 to 50%. To demonstrate reversibility of inhibition, CO was then flushed from solution; in the case of both Hyd-1 and Hyd-2, H 2 oxidation activity returned to the level recorded prior to CO exposure. A plot of i max /i t versus [CO], where i max is the current recorded under 100% H 2 (measured prior to the experiment), gives a straight line of intercept (K m /[S]) ϩ 1 and gradient K m /(K i ⅐[S]) (see supplemental data). Such analyses are shown in the insets in Fig. 6. Under 20% H 2 , K i (CO/H 2 ) values were measured to be 51 Ϯ 6 M for Hyd-1 at Ϫ0.060 V and 3 Ϯ 1 M for Hyd-2 at Ϫ0.175 V (both potentials are below E switch of the respective enzymes). Thus Hyd-2, in addition to showing far greater sensitivity than Hyd-1 to O 2 , is almost 20-fold more sensitive than Hyd-1 to inhibition by CO during H 2 oxidation. Fig. 6C shows a similar experiment to measure K i app (CO/H ϩ ) for Hyd-2 during H 2 production at Ϫ0.560 V, pH 6.0, 20°C. The experiment was carried out under an argon atmosphere with increasing partial pressures of CO, and again inhibition was shown to be reversible. The analysis shown in the inset gives a gradient of (1/K i app (CO)), which allows the K i app (CO/H ϩ ) value to be calculated as 11 Ϯ 3 M at Ϫ560 mV.
The Effect of O 2 on the H 2 Production Activity of Hyd-2-For reasons explained in detail by Goldet et al. (55,56), experiments to investigate the effect of O 2 on enzymes under very reducing conditions are subject to particular difficulties, most notably the reaction of soluble electron donors with O 2 . These problems can be overcome using PFE, provided that hydrogenasecatalyzed H ϩ reduction is distinguishable from O 2 reduction, which occurs directly on the electrode surface. By introducing a gaseous inhibitor such as CO, which is specific for the hydrogenase and has no effect on O 2 reduction at the electrode, the FIGURE 5. Aerobic H 2 oxidation by Hyd-1. Cyclic voltammograms recorded at 1 mV s Ϫ1 , pH 6.0, 37°C, under three different gas mixtures: 10% H 2 , 90% argon; 10% H 2 , 20% O 2 , 70% argon; and 20% O 2 , 80% argon, a total flow rate 500 standard cubic centimeters min Ϫ1 in each case. Prior to each measurement, the cell was allowed to equilibrate for 15 min with the appropriate gas mixture. Note that an O 2 reduction current arises from direct reaction of O 2 with the electrode. H 2 production current resulting from enzyme catalysis can be extracted from the overall current. Fig. 7 shows such an experiment. The measurement was initiated under a 100% argon atmosphere, and the negative current recorded corresponds only to H 2 production catalyzed by Hyd-2 on the electrode. Changing the headgas mixture to 10% CO in argon resulted in a rapid loss of current, due to inhibition of the enzyme. After recording the effect of this CO concentration, the headgas was changed back to 100% argon, and the H 2 production current recovered as CO was purged from solution.
In the next part of the experiment, the headgas was changed to 1.25% O 2 in argon (from Henry's law, this gives an O 2 concentration of 16 M in the cell solution at 30°C). A correction is required to allow for removal of O 2 by its direct reduction on the electrode surface. Using the method of Goldet et al. (55,56), it was estimated that the O 2 concentration experienced by Hyd-2 under these conditions is 10 Ͻ [O 2 ] Ͻ 16 M. Non-enzymatic O 2 reduction produces an increase in negative current. Once a steady value was established, 10% CO was again added to the headgas. A rapid decrease in negative current occurred, due solely to specific CO inhibition of H 2 production by active enzyme; importantly, the magnitude of the current loss was very similar to that resulting from CO addition in the absence of O 2 in the previous stage. From this, it is concluded that the H 2 production activity of Hyd-2 is largely unaffected by the presence of the low level of O 2 (10 Ͻ [O 2 ] Ͻ16 M).
Upon switching the headgas back to 1.25% O 2 in argon, the negative current increased in magnitude as the inhibitory CO was flushed from solution. In the final stage, anaerobic conditions (i.e. 100% argon) were restored, and a further CO gas exchange was carried out to compare the activity of the enzyme after O 2 exposure to that recorded before. After 1750 s of catalysis in FIGURE 6. Sensitivity of Hyd-1 and Hyd-2 to CO. Chronoamperometry traces of Hyd-1 (A) and Hyd-2 (B and C) during exposure to varying CO concentrations, as indicated. Experimental conditions were: pH 6.0, 20°C, electrode rotation rate 3500 rpm, and total gas flow rate 1000 standard cubic centimeters min Ϫ1 . Panel A (Hyd-1), electrode potential Ϫ0.060 V, carrier gas 20% H 2 in argon with increasing partial pressures of CO (the resulting concentration of CO in the electrochemical cell solution is indicated). Panel B (Hyd-2), electrode potential Ϫ0.175 V, carrier gas 20% H 2 in argon with increasing partial pressures of CO. Panel C (Hyd-2), electrode potential Ϫ0.560 V, carrier gas argon with increasing partial pressures of CO. In each case, CO is flushed from solution at the end of the experiment, to demonstrate reversibility of inhibition. Insets show plots used to determine K i (CO) in each case (see "Results"). K i (CO/H 2 ) for H 2 oxidation by Hyd-1 (A) was measured to be 51 Ϯ 6 M; K i (CO/H 2 ) for H 2 oxidation by Hyd-2 (B) was 3 Ϯ 1 M; and K i app (CO/H ϩ ) for H 2 production by Hyd-2 (C) was 11 Ϯ 3 M. FEBRUARY 5, 2010 • VOLUME 285 • NUMBER 6 the presence of O 2 , ϳ70% of the initial H 2 production activity remained.

DISCUSSION
The strikingly different catalytic profiles of Hyd-1 and Hyd-2 are summarized in Fig. 8. The "activity window" for H 2 oxidation by hydrogenases is defined by the potential at which H 2 oxidation activity commences and the potential at which the enzyme anaerobically inactivates (4). Under anaerobic conditions, Hyd-2 shows high activity in more reducing environments, to the extent that it can produce as well as oxidize H 2 in electrochemical experiments. However, the active form of Hyd-2 is unstable at potentials more positive than Ϫ0.085 V because the inhibited Ni-B state is formed. In contrast, the H 2 oxidation activity window of Hyd-1 is shifted to more oxidizing potentials, and Hyd-1 shows no activity at potentials low enough for H 2 production. PFE is well suited for measuring relative catalytic bias (oxidation versus reduction) and the precise dependence of rates on potential, but it is often not possible to measure electroactive coverage (it is too low) and therefore absolute activities are difficult to determine. Conversely, solution assays can underestimate activity because of limitations due to the driving force and bimolecular encounters. The results obtained with benzyl viologen suggest that Hyd-1 is much less active than Hyd-2. This differential is at least partly explained by the voltammograms (Fig. 1) that show that Hyd-1 but not Hyd-2 requires an overpotential of ϳ50 mV before oxidation of H 2 commences: benzyl viologen (EЈ 0 ϭ Ϫ350 mV) is only a very weak oxidant for H 2 at pH 6.0. The absence of detectable H 2 production by Hyd-1 using methyl viologen (EЈ 0 ϭ Ϫ460 mV) also corresponds perfectly with our PFE results.
When O 2 reacts with the active site of [NiFe] hydrogenases, a more reducing environment favors the formation of Ni-B over Ni-A (3). A proficient supply of electrons is required to ensure that O 2 is rapidly and fully reduced to H 2 O or equivalent species (OH Ϫ ), and that intermediates such as peroxide do not become trapped. When conditions prevail in which Ni-B is the dominant or sole species produced on reaction with O 2 , the activity window of the enzyme in the presence of O 2 will be the same as that under anaerobic conditions. The upper limit is thus defined by the potential at which Ni-B is re-activated (E switch ). Furthermore, because the rate of reductive activation of Ni-B is much higher than that of Ni-A, a hydrogenase that forms negligible amounts of Ni-A upon exposure to O 2 will be O 2 tolerant (13). Under extended exposure to O 2 with no control of potential, as occurs during isolation, both Hyd-1 and Hyd-2 form Ni-A and Ni-B active site states in poorly reproducible ratios. However, during H 2 oxidation at a controlled electrode potential, the ratio Ni-B:Ni-A formed on exposure to O 2 is far greater in Hyd-1 than in Hyd-2. This, along with our measurement of a much more positive E switch for Hyd-1 than for Hyd-2, correlates with our finding that Hyd-1 is a truly O 2 -tolerant enzyme.
Hyd-1 and Hyd-2 are both located in the periplasm of E. coli, and thus are exposed to conditions that depend largely on the external environment. In particular, periplasmic O 2 levels will fluctuate considerably during the infection cycle of E. coli. The high potential of the activity window for Hyd-1 would correlate with the enzyme being expressed and utilized under more oxidizing conditions, and indeed Hyd-1 but not Hyd-2 has been detected in aerobically grown cultures of E. coli (28).
The contrast between the H 2 oxidation activities of Hyd-1 and Hyd-2 that is only evident from voltammetric experiments, namely that there is a region of potential in which Hyd-2 is operational and Hyd-1 is not, depends on H 2 concentration. This region is larger at higher H 2 levels, and diminishes so that at 0.3% H 2 the difference is smaller than 20 mV. The effect arises as a result of the overpotential requirement of Hyd-1 at higher H 2 levels, the origin of which is under investigation. Measurements at Ϫ0.175 V showed Hyd-1 to have a slightly lower K m than Hyd-2. Physiological evidence implies that Hyd-1 may permit the coupling of H 2 oxidation to O 2 reduction by E. coli (27), and it is highly likely that whenever E. coli is exposed to O 2 , H 2 levels will be very low. Therefore if Hyd-1 does oxidize H 2 in the presence of O 2 in vivo, the low K m that we  measure would be a pre-requisite for significant energy extraction.
Hyd-2 behaves like a "standard" [NiFe]-hydrogenase on exposure to O 2 at high potential; like the hydrogenases from the Desulfovibrio species it cannot catalyze oxidation of H 2 in the presence of O 2 , although the result shown in Fig. 6 reveals Hyd-2 to be capable of O 2 -tolerant H 2 production. This is possible because H 2 production catalyzed by Hyd-2 occurs at potentials well below E switch (in the experiment a potential of Ϫ0.560 V was used). Under such conditions, the [NiFe] active site of the enzyme has ready access to a continuous supply of electrons, facilitating recovery from the Ni-B product of attack by O 2 . Also, at such negative potentials, only a small amount of Ni-A is ever likely to accumulate. Consequently, after almost 2000 s of exposure of active Hyd-2 to O 2 under H 2 production conditions, the majority of the enzymes pre-O 2 exposure activity remains.
In addition to their different catalytic biases and the distinct rates and potentials associated with their reactions with O 2 , Hyd-1 and Hyd-2 are further distinguished by their EPR spectra. The presence of a split signal at g ϭ 2.01 for Hyd-1 is similar to that observed for the [NiFe]-membrane-bound hydrogenase from R. eutropha, the best established O 2 -tolerant hydrogenase to date (43). In contrast, the relatively simple spectrum of Hyd-2 resembles hydrogenases from Desulfovibrio (46, 48 -50), which are strict anaerobes.
The data presented here not only demonstrate how E. coli can oxidize H 2 under different redox environments, but also suggests why this bacterium needs to express two separate uptake hydrogenases. Our physical characterization, taken together with genetic analyses (e.g. Refs. 19 and 57), establish that Hyd-1 is an O 2 -tolerant, unidirectional H 2 oxidizer, the role of which seems to be H 2 scavenging under conditions of slow growth and fluctuating O 2 levels. The H 2 oxidation activity of Hyd-1 is probably tightly coupled to the generation of a proton motive force via a loosely associated transmembrane cytochrome b subunit (HyaC) (19). In this respect, Hyd-1 shares a predicted similar overall structure to the R. eutropha membrane-bound hydrogenase (58). The function of Hyd-1 as an energy-conserving enzyme perhaps explains why this isoenzyme demonstrates an inherent inability to catalyze H 2 production at low redox potentials: reverse electron transport from the quinol pool through the Hyd-1 complex in vivo would collapse the transmembrane electrochemical gradient. The unidirectional nature of Hyd-1 is therefore an important physiological feature.
The characterization here of Hyd-2 as a hydrogenase capable of bidirectional behavior also correlates with what is known about the genetics of this system. Hyd-2 is not thought to associate with a HyaC-type quinone reductase, but instead passes electrons to the quinone pool by a different route, without generating a proton electrochemical gradient (21,59). It is possible that Hyd-2 is fine-tuned for conditions of rapid growth under nutrient-rich, but strictly anaerobic, conditions. Indeed, because Hyd-2 is uncoupled, this enzyme can potentially act as an electron release valve and run "backwards" if the quinone/ quinol pool becomes over-reduced. Thus, Hyd-2 is not a suitable system for conserving energy by H 2 scavenging at high redox potentials, and vice versa, Hyd-1 is not suited to bi-directional activity at low redox potentials.
Finally we summarize some possible technological implications. If H 2 is to be harnessed as a fuel, H 2 O-splitting catalysts must be designed. An enzyme such as Hyd-2, which can tolerate exposure to O 2 during H 2 production, proves the possibility of Pt-free catalysts for H 2 production under ambient conditions. The contrasting properties of Hyd-1 and Hyd-2 that we demonstrate are already being exploited in two completely different applications. The ability of Hyd-2 to produce H 2 in the presence of CO (although CO does act as an inhibitor, considerable enzyme activity remains under low CO partial pressures) has allowed Hyd-2 to be coupled to a CO dehydrogenase from Carboxydothermus hydrogenoformans via a conducting graphite particle. This system catalyzes the "water-gas shift" reaction with a much higher efficiency than the technologies currently used in industry (60). In contrast, the ability of Hyd-1 to oxidize H 2 in the presence of O 2 , along with its great stability on a graphite surface, makes it an appropriate anode catalyst for small enzyme-based fuel cells.