Relation between anaerobic inactivation and oxygen tolerance in a large series of NiFe hydrogenase mutants

Abstract

Nickel-containing hydrogenases, the biological catalysts of oxidation and production, reversibly inactivate under anaerobic, oxidizing conditions. We aim at understanding the mechanism of (in)activation and what determines its kinetics, because there is a correlation between fast reductive reactivation and oxygen tolerance, a property of some hydrogenases that is very desirable from the point of view of biotechnology. Direct electrochemistry is potentially very useful for learning about the redox-dependent conversions between active and inactive forms of hydrogenase, but the voltammetric signals are complex and often misread. Here we describe simple analytical models that we used to characterize and compare 16 mutants, obtained by substituting the position-74 valine of the -sensitive NiFe hydrogenase from Desulfovibrio fructosovorans. We observed that this substitution can accelerate reactivation up to 1,000-fold, depending on the polarity of the position 74 amino acid side chain. In terms of kinetics of anaerobic (in)activation and oxygen tolerance, the valine-to-histidine mutation has the most spectacular effect: The V74H mutant compares favorably with the -tolerant hydrogenase from Aquifex aeolicus, which we use here as a benchmark.

The nickel–iron hydrogenases that have been crystallized and/or thoroughly studied so far are very similar from a structural point of view: They all are either soluble heterodimers or heterodimers isolated from a membrane-associated complex. The amino acids that surround the NiFe active site are conserved (1) and yet the kinetic properties of these enzymes are diverse. For example, some NiFe hydrogenases can oxidize and produce , whereas others preferentially catalyze one direction of the reaction (2 –4). Another property of some hydrogenases that has attracted considerable interest is their sensitivity (and sometimes their resistance) to . This interest stems from the fact that hydrogenases could be used for oxidation in fuel cells or production in photo-electrochemical cells if they were functional under aerobic conditions (5).

The NiFe hydrogenases that have been studied first, referred to as “standard,” were purified from Allochromatium vinosum or Desulfovibrio species. Upon exposure to , they convert into two inactive forms called NiA and NiB, where an oxygenated ligand bridges the Ni and the Fe. The NiB and NiA states can be reactivated by reduction, the former more quickly than the latter (6). The membrane-bound NiFe hydrogenases from, e.g., Ralstonia eutropha or Aquifex aeolicus are reversibly inhibited by and termed “ resistant.” Apparently, this resistance results from (i) the enzyme reacting with to form only the NiB state and (ii) this NiB state reactivating much more quickly than in standard hydrogenases (4, 7). The most patent differences between -resistant and -sensitive hydrogenases are the structures and redox properties of the FeS cluster that is proximal to the active site. Recent structural and spectroscopic investigations have shown that in -resistant enzymes, it is an unprecedented 4Fe3S cluster (8 –10) that has a high reduction potential and the ability to cycle through three redox states in a narrow range of potential (11).

The rate of reactivation of NiB partly determines tolerance, but the reason this reaction is orders of magnitude faster in oxygen-resistant hydrogenases than in standard hydrogenases has not been investigated. The hypothesis that fast reactivation may result from a thermodynamic advantage (that is, NiB having a higher reduction potential) (7) has found no experimental support: The redox potential of the NiB state in the NiFe hydrogenase from A. aeolicus is not significantly greater than in standard hydrogenases (4), and the V74C mutation in Desulfovibrio fructosovorans has no effect on the spectral and redox properties of the NiB state, whereas it does increase significantly its rate of reduction (12).

The anaerobic conversion between active enzyme and the NiB state is often studied in protein film voltammetry experiments, where the enzyme is adsorbed or attached to a rotating-disk electrode and electron transfer is direct. The value of the electrode potential (E) is set to drive the catalytic reaction while the turnover rate (the activity) is simultaneously measured as a current (13). Changing the electrode potential also triggers the conversion between active and inactive states. In chronoamperometry (CA) experiments, the (in)activation is seen as current transients following potential steps (14, 15). In cyclic voltammetry (CV) experiments, the oxidative inactivation gives the electrochemical response the complex shape that is shown in Fig. 1A (14 –16). Starting from the low potential limit, the current initially increases as the electrode potential increases (and so does the driving force for oxidation) and then bends down or even decreases at high potential, as a result of the oxidative formation of the inactive state. As the potential is lowered back on the reverse scan, the reactivation produces an increase in current, which is eventually outcompeted by the decrease in steady-state activity that results from the potential being less oxidizing. Hysteresis occurs because the conversion between (in)active states is slow on the voltammetric timescale (16), irrespective of scan rate in the experimentally accessible range of scan rate ( mV/s). The wave shape is defined by the steady-state electrochemical response of the active enzyme (17) and the potential-dependent kinetics of conversion between active and inactive states, but the latter, crucial information is not easily disclosed. In particular, the position of the first inflection point on the reverse scan (the so-called “switch potential,” , shown as vertical segments in Fig. 1A ) is very dependent on scan rate ^† and should therefore not be interpreted as a thermodynamic quantity. Considering that voltammetry is now used in several research groups to characterize hydrogenases isolated from various organisms (e.g., refs. 3, 18 –21), there is a crucial need for simple theoretical models that will ease quantitative investigations.

Fig. 1.

Electrochemical characterization of the D. fructosovorans V74N mutant. (A) CVs recorded with the enzyme attached to a rotating-disk graphite electrode. Potentials are quoted versus the standard hydrogen electrode (SHE). One atmosphere , pH 5.5, 40 °C, electrode rotation rate ω = 3,000 rpm. The vertical lines mark the “switch” potentials. (B and C) Solid lines show portions of the voltammograms in A. The dashed lines are the best fits to Eq. 5 (C) and Eq. 7 (B). (D) Current response (after subtraction of the capacitive current recorded in a control experiment with no adsorbed enzyme) when the potential was stepped according to the sequence , 190, , , 240, 40, mV, as shown using the dashed line and the Y-right axis. (E) The relation between and scan rate. The straight dashed red line is a fit to Eq. 10 . F shows as open squares the time constants determined by fitting exponential functions to the current transients. The blue horizontal lines indicate the values of (solid line) (dashed line), and (dotted line) determined by fitting the portions of CVs shown in C. The green line shows the rate of reactivation at low potential, determined by fitting the data in B. The red line shows the rate of reactivation at low potential, determined by fitting the change in switch potential against scan rate.

Valine 74 shapes a bottleneck in the channel that guides from solvent to the active site of D. fructosovorans NiFe hydrogenase (22). Substitutions of this amino acid have a strong effect on the rates of intramolecular diffusion (23, 24) and, consequently, on the Michaelis constant for (24), on the catalytic bias of the enzyme (2), and on the rates of inhibition by CO and (23, 24). For example, the V74Q, V74M, and V74E mutants are inhibited by more slowly than the wild-type (WT) enzyme, but this effect on the rate of inhibition is not strong enough to make the enzyme oxygen resistant. Indeed, in contrast to standard hydrogenases, those that are naturally resistant reactivate after they have been transiently exposed to even under oxidizing conditions. A high rate of reactivation is therefore more beneficial than a slow rate of reaction with , and oxygen resistance cannot be achieved only by obstructing the gas channel (24). The V74M (25) and V74C (12) mutations significantly increase the resistance to because they modify the way reacts at the active site, independently of the effect of the rate of diffusion toward the active site. The V74M variant can oxidize over extended periods of time in the presence of (25) and the V74C mutant reacts with to form preferentially a NiB state that has the same thermodynamic and spectral properties as in the WT enzyme, but reactivates much more quickly (12): These are precisely the properties one seeks if the goal is to mimic -tolerant hydrogenases.

Suspecting that V74 mutants other than V74M and V74C may have interesting properties, and aiming at understanding the reason these substitutions affect tolerance, we developed electrochemical methods and used them for screening in a quantitative manner 16 position-74 mutants. We discuss the interpretation of the electrochemical data and the relation between kinetics of anaerobic inactivation and oxygen tolerance, and we identify a mutation (V to H) that results in a spectacular improvement.

Results and Discussion

We interpret the electrochemical data by considering the anaerobic, redox-dependent transformation between active and inactive enzyme:

and are the first-order rate constants of inactivation and reactivation, respectively. The change in concentration of active enzyme, , is obtained by solving

The enzyme activates or reactivates when the electrode potential is changed because and/or depend on E.

Fitting Chronoamperograms Returns k _i(E)+k _a(E) at Every Potential.

The (in)activation is clearly detected in CA experiments where the hydrogenase is attached to a rotating-disk graphite electrode and the potential is changed stepwise. Fig. 1D shows a sequence of potential steps and the corresponding current response of the V74N mutant of D. fructosovorans NiFe hydrogenase. The -oxidation current instantly increases each time the potential is stepped up as a result of the increase in driving force and then decreases slowly as the enzyme inactivates. After each step down, the catalytic current immediately drops and then slowly increases as activity is recovered.

A straightforward analysis of these experiments consists of fitting each current transient to an exponential function to measure the time constant at any potential (4, 15); the results are shown as squares in Fig. 1F .

Whether the enzyme will activate or inactivate after a certain potential step can be predicted only if one knows the variations of both and . For example, if the ratio increases as the electrode potential increases, the enzyme inactivates after a step up. We have shown that, to independently determine the two rate constants from CA experiments, one must simultaneously interpret the time constant and the magnitude of the transients (15). The result of such detailed investigation of the oxygen-tolerant hydrogenase from A. aeolicus in ref. 15 is summarized by the following empirical equations (plotted in Fig. S1A), which can be used to simulate the whole voltammogram (15):

The prefactor k is a rate constant whose meaning is discussed below. We note .

Eq. 3 appears to be general: It is consistent with the results of CA experiments carried out with A. vinosum (14) and A. aeolicus (4, 15) WT NiFe hydrogenases and also with the V74N D. fructosovorans mutant (Fig. 1F ): The rate of interconversion between active and inactive states () levels off at high potential and increases exponentially with at low potential. The voltammetric traces are also consistent with Eq. 3 . Indeed, the oxidative inactivation proceeds gradually because both and are independent of E at high potential (Eq. 3 and Fig. S1A). On the reverse scan, as the electrode potential is lowered, increases whereas does not, and the enzyme therefore reactivates; the increase in current is very sudden around the so-called switch potential because the exponential increase in rate of reactivation (Eq. 3a ) produces a sharp voltammetric feature.

We now assume that Eq. 3 also applies in the case of all other V74 mutants of D. fructosovorans NiFe hydrogenase and we derive models that can be used to determine the parameters in Eq. 3 (k, n, , and ) from voltammetric experiments. Fitting the entire voltammogram recorded at slow scan rate is not straightforward because Eq. 2 must be numerically solved to obtain the time-dependent fraction of enzyme that is in the active form, (15). However, we show hereafter that there are potential ranges where the equations for and are simple enough that closed-form expressions can be derived and used to fit fragments of the cyclic voltammograms (CVs) that embed the relevant information. The dependence of the switch potential on scan rate can also be rigorously interpreted.

Wave Shape at High Potential Depends on and .

If E is high enough that and are constant, then the fraction of active enzyme decreases exponentially with time, with a time constant that depends mainly on , toward a limiting value ,

where is the fraction of active enzyme at (herein, subscript “1” is used to denote an initial condition that will be treated as an adjustable parameter in a fitting procedure). The current is obtained by multiplying the time-dependent fraction of active enzyme by the steady-state current response of the active enzyme (15), which, in the potential range of interest, is a linear function of E (cf. equation 9 in ref. 17),

where is the vertex potential and the scan rate v is counted as positive in both directions. The linear steady-state contribution results from the fact that enzyme molecules that are weakly coupled to the electrode contribute to the current only at high driving force. Eq. 5 is defined by five independent parameters: a and b (which can be measured from the slope of the voltammogram in the potential range where there is no inactivation), and (both in the range [0–1]), and τ. The initial potential (), vertex potential (), and scan rate (v) are known.

The solid lines in Fig. 1C are the high-potential parts of the CVs shown in Fig. 1A . The dashed lines are the fits to Eq. 5 obtained with a single pair of values for the parameters and τ. The blue, solid, horizontal line in Fig. 1F indicates the values of : It agrees with the results of the CA measurements. The dashed and dotted horizontal blue lines mark the values of and , respectively.

Wave Shape on the Reverse Scan Gives k and n, Which Define (E) at Low Potential.

On the reverse scan, when the electrode potential becomes low, starts to increase and so does . Here we assume that (i) the potential is low enough (lower than a certain potential ) that and (ii) the reactivation is irreversible (that is, the second term in the right-hand side of Eq. 2 dominates). If when , obeys

This equation is solved using the initial condition that at , and the current equation for the reverse sweep is obtained by multiplying by the linear current response of the active enzyme,

Five parameters must be adjusted to fit the data: a and b, (the initial fraction of active enzyme at ), n, and k.

The solid lines in Fig. 1B show the fragments of the CVs shown in Fig. 1A that can be described by Eq. 7 . The best fit (dashed lines) has been obtained with a single pair of parameters for k and n. Using these values, Eq. 3a was used to recalculate the low-potential part of , shown as a green line in Fig. 1F : Again the agreement with the results of the CA measurements is very good.

Switch Potential Is a Kinetic, Rather than Thermodynamic, Parameter.

We demonstrate in SI Text that if the reactivation is irreversible [condition (ii) in the previous section], then passes through an inflection point when

The physical meaning of Eq. 8 is that reactivation depends on two different timescales: the rate of reactivation, , and the rate at which increases, . On the high-potential part of the return sweep, , and the fast increase in makes the current increase faster and faster. On the low-potential part of the return sweep, the process slows down before reaching completion because is large enough that the decrease in the amount of inactive species counteracts the increase in . The switch potential marks the transition between the two regimes.

Using the Change in Against v to Measure k and n.

Substituting in Eq. 8 gives

and after rearrangement

Eq. 10 predicts the value of and explains the linear relation between and that has been observed every time the effect of scan rate has been examined.^† Because Eq. 10 predicts a mV change in per decade of scan rate, a switch potential should always be quoted along with the value of the scan rate.

We analyzed the data in Fig. 1A by determining the values of from the position of the minimum on the first derivative of each reverse sweep (Fig. S2). The relation between and is shown as squares in Fig. 1E . Fitting a straight line to these data (Eq. 10 ) returned the values of k and n that we used to calculate the low-potential part of shown in red in Fig. 1F .

At fast scan rates, the switch is not detected on the reverse scan because the enzyme has had no time to inactivate on the forward scan (e.g., purple line in Fig. 1A ). We carried out experiments (Fig. S3) where the electrode potential is poised at mV for about 15 min (to let the enzyme inactivate) before quickly sweeping the potential down. We used these data to determine the switch potentials whose values are shown as open circles in Fig. 1E . There is no deviation from the linear relationship between and that is observed at a slower scan rate. The switch potential decreases by about 200 mV when the scan rate is increased from 0.5 mV/s to 0.4 V/s.

Relation Between the Switch Potential and the Result of Equilibrium Titrations.

As activation proceeds, the fully oxidized inactive state NiB is first one-electron reduced (with a reduction potential ), into an inactive “NiSIr” state, before it gives the active form NiSIa that binds as part of the catalytic cycle (scheme 1 in ref. 4):

The above mechanism is consistent with the kinetics of reactivation. Indeed, in the case of all mutants analyzed here, the changes in against and the fits of the CVs to Eq. 7 returned values of n close to unity (in the range 0.8–1.1) ^‡ , suggesting that the low potential increase in , which determines , is the onset of the one-electron sigmoid plotted in Fig. S1B: . Hence k in Eq. 3a equates to . The rate of reactivation is independent of , consistent with the earlier observation that and are independent of concentration (14, 15). The low-potential plateau at was not approached in our experiments, suggesting that is, in each case, lower than the lowest potential probed in the CA experiments and lower than the switch potential measured even at the fastest scan rates.

When the scan rate is decreased, the value of increases and drifts away from . At an infinitely low scan rate, tends to the potential where and intersect; ^§ this value depends on concentration because binding prevents inactivation. We have indeed observed that decreases when the concentration of increases (figure 4 in ref. 15).

In redox titrations (in the absence of ), NiSIr and NiSIa are in equilibrium and the measured potential is more positive than by an amount .

Summary About the Switch Potential.

The value of this parameter is determined by the rate of reactivation. It is not the reduction potential of the inactive state. It is independent of [], dependent on scan rate (Fig. 1E ), and always greater than (defined by Eq. 11 ). Its pH dependence should be the same as that of (Eq. 10 ). Redox titrations return an equilibrium potential that is also greater than and may, accidently, match .

Histidine Variant Exhibits the Fastest (In)activation Rates and a Slightly Biphasic Behavior.

Fig. S4 and Fig. 2 summarize the results obtained with the V74H mutant, which exhibits the fastest rates of equilibration between active and inactive states. Fig. S4 shows that the hypothesis that a single inactive state is reversibly formed at high potential in a redox-dependent process that is defined by Eq. 3 leads to a satisfactory description of the CV data.

Fig. 2.

Characterization of the V74H mutant and comparison with the enzyme from A. aeolicus. (A) The result of a CA experiment performed with the V74H mutant. The electrode potential was stepped according to the sequence , , , , , , mV vs. SHE. (B) Change in against scan rate, for V74H, fitted to Eq. 10 . (C) Open squares show the time constants for V74H, determined by fitting the sum of two exponential functions to CA traces like those in A; the symbols that mark the slower time constant have a surface that is proportional to the relative amplitude of the slow phase. The blue line in C indicates the value of determined by fitting portions of CVs recorded in the high-potential range (Fig. S4C and Eq. 5 ). The green line in C shows the rate of reactivation determined by fitting portions of CVs recorded in the intermediate range of potential (Fig. S4B and Eq. 7 ). The red line shows the rate of reactivation at low potential, determined by fitting the change in switch potential against scan rate (B and Eq. 10 ). The open red circles in C show the time constants measured by fitting CA traces recorded with the -resistant enzyme from A. aeolicus under the same conditions, 40 °C, pH 5.5, 1 atm of , 3,000 rpm.

However, the CA data obtained with the V74H mutant attest to the formation of more than one inactive state. Indeed, the current transients are better described by a sum of two exponentials than by a single exponential function. The deviation from a clean exponential decay is clear, for example, on the current traces shown in Fig. 2A . Fitting a sum of two exponential functions to these data returned the time constants shown as open black squares in Fig. 2C ; the symbols that mark the slower time constant have a surface that is proportional to the relative amplitude of the slow phase, ranging from 1/2.5 at high potential to about 1/45 at low potential. The correspondence between the rates of the largest-amplitude transients and the lines in Fig. 2C (deduced from the interpretation of the voltammetry) shows that analyzing the voltammetric data with a model that assumes the formation of a single inactive state returns the rate of (in)activation that corresponds to the major species.

Fig. 2C compares the time constants of (in)activation determined in CA experiments carried out with the V74H mutant (open black squares) and the -resistant enzyme from A. aeolicus (open red circles) under the same conditions. The fast reactivation phase that dominates the V74H transients is faster than in A. aeolicus (the latter exhibits monophasic transients) (15). The slow phase of V74H has about the same rate as the unique phase of A. aeolicus NiFe hydrogenase.

We repeated the above electrochemical characterization of the V74N and V74H mutants with 14 other V74 mutants, and we found that the above conclusions are always true: The model that assumes the formation of a single inactive state describes well the CV data, showing (i) that this reversible (in)activation is responsible for the major voltammetric features in all cases and (ii) that the mutation of the position-74 valine affects the rate but not the mechanism of this process. However, many mutants, like V74H, exhibit nonideal (biphasic) behavior in CA experiments. When the CA traces reveal that two inactive states are formed and/or reactivated on the timescale of the electrochemical experiments, the CVs inform on the inactive state whose formation or activation dominates the chronoamperometric transients.

Hydrophilic Position-74 Side Chains Accelerate Anaerobic (In)activation.

We analyzed the data obtained with all mutants and collected the results in Fig. 3, which shows the rate of reactivation at −90 mV (an arbitrary value) against the limiting value of at high potential (compare Eqs. 3 ).

Fig. 3.

The rate of reactivation at mV plotted against the rate of inactivation at high potential for all D. fructosovorans NiFe hydrogenase mutants (squares; the red letter indicates the position-74 residue) and the -resistant enzyme from A. aeolicus (open circle).

Fig. 3 and Table S1 show that substituting V74 may change the rate of reactivation by more than three orders of magnitude, from s for V74M to 30 s⁻¹ for V74H (at −90 mV, pH 5.5, 40 °C). The rate of reactivation of the histidine mutant is 20 times that of the V74C mutant, and close to that observed with the O₂-tolerant enzyme from A. aeolicus (4, 15).

The mutants that show the fastest reactivation rates are those where valine is replaced by a polar amino acid (threonine is an exception). The pK_a of the position-74 side chain does not seem be determining, because, for example, the V74K and V74E mutants reactivate at about the same rate. Therefore, the rates k _i and k _a are not dependent on a proton transfer mediated by the position-74 side chain.

Fig. 3 shows that there is a strong positive correlation between the rate of reactivation at a low potential, k _a(E = −90 mV), and the rate of inactivation at high potential, . The CVs shown in this paper illustrate this correlation: The V74N mutant inactivates more slowly at high potential and reactivates more slowly (hence at a lower potential) than the V74H mutant (compare Fig. 1A and Fig. S4A). The relation between and shows that, unexpectedly, the mutations affect the rate of a step that is common to both processes.

Mutants That (In)activate More Quickly Under Anaerobic Conditions Resist Better.

We tested the O₂ sensitivity of all mutants, and we found that the mutants that exhibit the fastest rates of (in)activation are also those that exhibit the greatest resistance to O₂.

Fig. 4 shows how transient expositions to affect the activity of the most resistant mutants (additional results in Fig. S5). The experiments were carried out by poising the electrode potential at mV vs. SHE, under an atmosphere of , and repeatedly injecting in the cell solution small amounts of an -saturated buffer, reaching concentrations of 4, 8 and 20 μM. After each injection, is flushed away by the stream of , and its concentration decreases exponentially with time, with a time constant of about 20 s (26). The red and blue traces in Fig. 4 can be used to benchmark the tolerance of the mutants. WT D. fructosovorans NiFe hydrogenase quickly loses nearly all activity after the first exposure to (red trace), whereas the enzyme from A. aeolicus (blue trace) recovers activity after each injection. The time constant of the reactivation after the enzyme has been exposed to ( s) is equal to the time constant of anaerobic equilibration between the active and the inactive state [, open red circles in Fig. 2C ], which implies that anaerobic oxidation and reaction with form the same inactive state (see discussion of figure 7 in ref. 4).

Fig. 4.

Effect of transient exposure to on the turnover rate of WT D. fructosovorans NiFe hydrogenase (red); the V74H (green), V74P (gray), V74N (black), and V74C (purple) mutants; and the WT enzyme from A. aeolicus (blue). All experiments were performed at 40 °C, mV vs. SHE, pH 5.5, under 1 atm of , 3,000 rpm.

Regarding the hydrogenases from D. fructosovorans, we observed that the V74 mutations have little effect on the bimolecular rate of reaction with (the parameter we called in ref. 24) unless diffusion along the channel is so slow that the diffusion step limits the rate of inactivation, as is the case for V74Q, V74E, and V74M (24). For all other mutants, the rate-limiting step for the reaction with is the reaction at the active site (24), and substituting V74 results in a smaller than twofold effect on the rate of reaction with : This rate appears to be independent of the nature of the position-74 side chain.

The purple, gray, and green traces in Fig. 4 show the response of the V74H, V74P, and V74C mutants to transient exposures to . The histidine variant reactivates as quickly as the enzymes from A. aeolicus, consistent with the observation in Fig. 2C that the sum takes the same value for the two enzymes at the potential used in the experiment in Fig. 4 ( mV vs. SHE). The experiment in Fig. 4 clearly shows that the reactivation after exposure to is not complete, but full activity was recovered after a 800-s poise at mV (Fig. S6), suggesting that at least two inactive states have been formed and that they reactivate at different rates. This conclusion was confirmed by FTIR experiments carried out in the absence of potential control (Fig. S7). After aerobic purification, the V74H mutant is mostly in the NiA state. After reduction and reoxidation with air, the FTIR signal shows that two major species have been formed: One has a typical NiB FTIR signature, and the other (with a CO band at 1,943 cm) is unprecedented.

Kinetic Determinants of Tolerance Are and , Not or the Low-Potential Part of (E).

Certain NiFe hydrogenases are resistant because they react with to form a single inactive state, NiB, that reactivates more quickly than the NiB state of standard hydrogenases. So far, this kinetic property has been related only to the rate of reactivation at low potential or, equivalently, to the value of the switch potential (and indeed, the two are related to each other by Eq. 10 ).

The experiments reported here illustrate this relation: The mutants that exhibit the fastest reactivation rates (at low potential under anaerobic conditions) are also those that reactivate quickly, at high potential, after exposure to . However, it is important to realize that it is the large value of the inactivation rate constant ( in Eq. 3 , Fig. S1A) that is beneficial. Indeed, if the same inactive state is formed under aerobic and anaerobic conditions, then the recovery of activity after the enzyme has been inhibited by occurs with a time constant , which, at high potential, depends mainly on ; a fast rate of anaerobic inactivation at high potential is therefore synonymous with fast recovery of activity after the enzyme has been exposed to . The fact that at low potential and are correlated (Fig. 3) is the only reason tolerance seems to relate to .

Future Directions.

It has recently been suggested that certain enzymes are tolerant because a 4Fe3S cluster that is proximal from the active site prevents the formation of NiA. However, the fact that this cluster has a very high reduction potential (11) should not favor fast electron transfer toward the active site and fast reduction of NiB. Recent results show that modifying the proximal cluster has no effect on the switch potential (27) and, therefore, on the rate of reactivation. The reason the inactive state reactivates quickly in -tolerant NiFe hydrogenases has yet to be understood.

The observation herein that hydrophilic position-74 residues accelerate the reactivation of the enzyme from D. fructosovorans is reminiscent of the recent finding that water-filled cavities are present in the environment of the NiFe active site of the -tolerant membrane-bound hydrogenase from R. eutropha (8). It has been suggested that they are part of a hydrophilic channel that guides away from the active site the water molecules that are constantly produced upon complete reduction of the attacking (8). This has not been supported yet by the results of mutagenesis experiments. In R. eutropha, mutating amino acids that shape this hydrophilic cavity (G80 and C81) have no effect on the inhibition constant relative to (1), and in D. fructosovorans, the reverse mutations (Y77G and V78C) have no effect on the bimolecular rate of inhibition by (28). Whether these amino acids affect the rates of anaerobic (in)activation now deserves investigation.

Materials and Methods

The D. fructosovorans NiFe hydrogenase mutants were purified as described previously (12). Table S1 lists their purification yields, specific activities, and Michaelis constants for . The purification of the enzyme from A. aeolicus is described in ref. 15. All voltammetric experiments were carried out with the enzyme molecules covalently attached (29) to a rotating-disk pyrolytic graphite edge electrode, in a glove box filled with nitrogen, using the equipment described previously (12). We analyzed the electrochemical data using the in-house programs SOAS and Qsoas. The former is available free and free of charge on our Web site at http://bip.cnrs-mrs.fr/bip06/software.html (30). Qsoas can be used to simultaneously fit a series of datasets by forcing certain parameters to have the same values for all datasets: This was useful for finding the best values of and τ defined in Eq. 5 or k and n in Eq. 7 , by fitting a series of CVs recorded over a range of scan rates.