Characterization of [FeFe] Hydrogenase O2 Sensitivity Using a New, Physiological Approach*

[FeFe] hydrogenases catalyze rapid H2 production but are highly O2-sensitive. Developing O2-tolerant enzymes is needed for sustainable H2 production technologies, but the lack of a quantitative and predictive assay for O2 tolerance has impeded progress. We describe a new approach to provide quantitative assessment of O2 sensitivity by using an assay employing ferredoxin NADP+ reductase (FNR) to transfer electrons from NADPH to hydrogenase via ferredoxins (Fd). Hydrogenase inactivation is measured during H2 production in an O2-containing environment. An alternative assay uses dithionite (DTH) to provide reduced Fd. This second assay measures the remaining hydrogenase activity in periodic samples taken from the NADPH-driven reaction solutions. The second assay validates the more convenient NADPH-driven assay, which better mimics physiological conditions. During development of the NADPH-driven assay and while characterizing the Clostridium pasteurianum (Cp) [FeFe] hydrogenase, CpI, we detected significant rates of direct electron loss from reduced Fd to O2. However, this loss does not interfere with measurement of first order hydrogenase inactivation, providing rate constants insensitive to initial hydrogenase concentration. We show increased activity and O2 tolerance for a protein fusion between Cp ferredoxin (CpFd) and CpI mediated by a 15-amino acid linker but not for a longer linker. We suggest that this precise, solution phase assay for [FeFe] hydrogenase O2 sensitivity and the insights we provide constitute an important advance toward the discovery of the O2-tolerant [FeFe] hydrogenases required for photosynthetic, biological H2 production.

Hydrogenases catalyze the reversible formation of H 2 from two protons and two electrons. They are classified as [FeFe], [NiFe], or [Fe]-only with respect to the metal atoms included in the active site. Although [NiFe] hydrogenases are generally biased toward oxidation of H 2 , [FeFe] hydrogenases have a catalytic bias toward H 2 formation. The most prolific enzymes require overpotentials lower than that for platinum, one of the best metal catalysts reported to date (1). Consequently, [FeFe] hydrogenases offer significant potential for biological production of H 2 (2)(3)(4)(5)(6).
We have approached such applications with a particular focus on the [FeFe] hydrogenase from Clostridium pasteurianum. It is a 63.8-kDa protein with one of the highest reported H 2 production-specific activities (2,7). However, it is also a complex enzyme with three accessory [4Fe-4S] clusters and one [2Fe-2S] cluster that deliver electrons to or from the active site consisting of an [FeFe] sub-cluster bridged to a [4Fe-4S] cluster by a cysteinyl thiol. Much work has been done during the past decade to enable the heterologous expression, maturation, and purification of the active form of CpI 2 in Escherichia coli (8,9). Based on these advances, we are now working on two distinct routes of biological H 2 production using this enzyme: fermentative (2, 10) and photosynthetic (11).
Unfortunately, [FeFe] hydrogenases are highly sensitive to O 2 (12)(13)(14), imposing a significant technical barrier for oxygenic photosynthetic H 2 production. In fact, the half-life of CpI in air-saturated buffer is estimated to be 2-3 min (13,15,16). Although the exact mechanism of inactivation remains to be elucidated, previous research has provided insights and hypotheses. Strip et al. (17) proposed that O 2 first binds to the distal Fe atom of the [FeFe] sub-cluster of the active site. Reactive oxygen species are then formed and cause dissociation of the neighboring [4Fe-4S] sub-cluster. Cohen et al. (18) suggested that O 2 molecules migrate to the active site through channels within the enzyme. Goldet et al. (13) then developed a kinetic model that consists of two rate-limiting steps: O 2 first diffuses to the active site and then permanently destroys activity. More recently, Swanson et al. (16) reported that the extent of inactivation depends on the initial redox state of the hydrogenase and that the inactivation process results in the loss of the [FeFe] sub-cluster from the active site. Despite these insights, an inability to quantitatively and conveniently assess O 2 sensitivity in a biological context has limited progress in the discov- Most studies on the oxygenic inactivation of [FeFe] hydrogenases to date have relied on the experimental technique known as protein film voltammetry (PFV). In this technique, the generated by applying either negative or positive potential bias (12,19,20). Although this technique allows precise measurements, it also has drawbacks (21). Most importantly, the technique employs an unnatural electron supply mechanism, and O 2 reduction at the electrode complicates interpretation. Concern has also been expressed that individual hydrogenase molecules may adsorb to the electrode in different orientations, thereby providing different electron transfer distances between the electrode and conducting Fe-S clusters in the enzyme (22). As a result, the findings from PFV experiments may not accurately predict in vivo hydrogenase O 2 sensitivity.
In this study, we describe an alternative approach to assessing O 2 tolerance of hydrogenases that more closely mimics physiological H 2 production as well as our envisioned processes for biological H 2 production. This approach allows us to assess two critical parameters connected to O 2 tolerance: the inactivation rate constant for hydrogenase activity in the presence of O 2 , and the residual hydrogenase activity after O 2 exposure. In assessing both, we use the natural electron source for hydrogenases, a solution phase reduced Fd; the difference lies in how the oxidized Fd is reduced (Fig. 1). The first assay (Fig. 1a) uses a reaction sequence developed as an integral part of our fermentative H 2 production process (2). It delivers electrons from NADPH to CpI via FNR and Fd. This assay assesses the inactivation of CpI while it produces H 2 in the presence of O 2 . The second assay (Fig. 1b) uses a sequence in which electrons are delivered from DTH to hydrogenase via Fd. In this assay, the strong reducing power of DTH (E o Ј Ϫ0.66 V versus normal hydrogen electrode at pH 7) consumes any remaining O 2 , and apparently maintains most of the Fd in the reduced state (Fd red ), allowing an assessment of the full residual activity of hydrogenase after O 2 exposure.
Using these assays, we studied the kinetics of inactivation with varying concentrations of O 2 and CpI. Next, we evaluated the influence of other reagents. Noting that the O 2 concentration decreases over the course of the NADPH-driven assay, we confirmed a significant rate of electron loss directly from the reduced Fd to O 2 . Nonetheless, our approach provides convenient assessment of the O 2 tolerance of [FeFe] hydrogenases and engineered mutants. As an example, we show that a CpI-Fd fusion protein with a higher hydrogenase activity is also less sensitive to O 2 inactivation.

Influence of CpI and O 2 Concentration on the Kinetics of
Inactivation-Due to our interest in photosynthetic H 2 production, we conducted our initial experiments with the Fd from Synechocystis sp. PCC 6803 (SynFd). Oryza sativa (rice root) FNR (RrFNR) was chosen over the FNR native to SynFd or CpI because it results in significantly faster H 2 production within the NADPH-driven assay (11). With these choices, we first evaluated the kinetics of CpI inactivation caused by different concentrations of O 2 . The H 2 accumulation rate appeared to decrease immediately upon the introduction of O 2 and then further diminished over the course of an hour (Fig. 2). As expected, higher concentrations of O 2 were more deleterious but, surprisingly, no additional H 2 production was detectable after introducing 10 vol % O 2 or higher. These reaction mixtures were well mixed to encourage O 2 transfer into the liquid phase, and 10 vol % O 2 in the headspace is expected to produce a dissolved O 2 concentration of roughly 120 M.
Goldet et al. (13) Fig. 2b indicates that an attempt to fit the full data set to this model was unsuccessful. The decline in hydrogenase activity measured immediately after O 2 addition was consistently lower than predicted. We show later that the immediate drop in H 2 production rate is caused by direct electron loss from SynFd red to O 2 . However, this loss remains relatively constant, and the subsequent decreases in H 2 production rate primarily reflect CpI inactivation. The same pattern, immediate increase in current upon O 2 introduction and gradual decrease afterward, was observed with PFV as well (13). In our assay, using only the rates after O 2 addition provides data sets that indicate the expected exponential decay curves ( Fig. 2c) with k inact,O2 ranging from 0.025 to 0.099 min Ϫ1 for 1.0 to 5.0 vol % O 2 (Table 1).
To further test the assay, we next varied the initial concentration of CpI from 1 to 100 nM while adding 5.0 vol % O 2 after 10 min (Fig. 3). First order inactivation will only be valid if the inactivation rate constant does not vary as a function of active hydrogenase concentration. In all cases, significant H 2 production activity was retained for only 30 min after the introduction of 5.0 vol % O 2 (Fig. 3a), limiting analysis to that time period. The magnitude of electron loss to O 2 also precluded analysis with 1 nM CpI. Nonetheless, the available H 2 production rate measurements (Fig. 3b) indicate first order inactivation with consistent inactivation rate constants (0.097 Ϯ 0.005 min Ϫ1 ) within the accuracy of our assay over a 20-fold range of CpI concentrations. For convenience, the first slope after O 2 addition is assumed to approximate the rate after 5 min of exposure, whereas the rates at other time points are the average of the preceding and subsequent slopes.
Influence of Other Reagents on the Kinetics-We next examined the influence of the concentrations of the two potentially rate-limiting electron transfer proteins, FNR and Fd, used for NADPH-driven H 2 production (Fig. 4). Increasing

[FeFe] Hydrogenase O 2 Sensitivity Assay
[FNR] slowed inactivation by 5.0 vol % O 2 (Fig. 4a). An increase from 5 to 50 M increased the anaerobic H 2 production rate by about 10% and decreased the inactivation rate constant, k inact,O2 , from 0.102 Ϯ 0.004 to 0.089 Ϯ 0.008 min Ϫ1 . This observation suggests that increasing the flux of electrons to the hydrogenase may also increase O 2 tolerance. Note that the anaerobic CpI turnover numbers (TON) in these experiments were about 4.5 s Ϫ1 relative to a k cat of about 400 s Ϫ1 with DTH as the electron source. This factor will be discussed in more detail later.
Increasing [Fd] from 5 to 50 M had a similar impact on the inactivation kinetics (Fig. 4b). k inact,O2 decreased from 0.097 Ϯ 0.006 to 0.082 Ϯ 0.008 min Ϫ1 , whereas the CpI TON stayed the same. More significant changes were observed when [Fd] was decreased from 5 to 1 M. The anaerobic CpI TON decreased to 78% of the level with 5 M Fd, and after O 2 addition, the H 2 production rate immediately dropped by 87% for 1 M versus 68% for the higher concentrations. k inact,O2 also increased to 0.118 Ϯ 0.010 min Ϫ1 . The larger k inact,O2 is in agreement with our aforementioned conjecture about the influence of electron flux on the hydrogenase inactivation rate.
Electron Loss from Reduced Fd to O 2 -As mentioned previously, several observations suggested that O 2 addition to the reactor headspace lowered the rate of electron supply to the hydrogenase. As shown in Fig. 5a, O 2 addition to the NADPHdriven assay initiates significant O 2 consumption. Although the O 2 tolerance mechanism for some [NiFe] hydrogenases (reduction of O 2 to H 2 O) (23) suggests that the hydrogenase could deplete O 2 , this was not apparent in our experiments. Incremental omission of reagents starting at the end of the reaction sequence clearly indicated that the reduced Fd was responsible. Omitting the electron source avoided O 2 depletion as expected. A 1974 study indicates that Clostridial ferredoxins are capable of reducing O 2 to form reactive oxygen species (24).
The study described here was conducted with CpFd because we reasoned that it would interact more favorably with CpI, its native partner. CpFd contains two [4Fe-4S] clusters, whereas SynFd used in our initial experiments has only one [2Fe-2S] cluster. SynFd transferred electrons to O 2 at about 50% of the rate observed for CpFd over an hour within the NADPH-driven assay in the absence of CpI (Fig. 5b). Adding catalase and superoxide dismutase to the assay reduced the rate of O 2 depletion consistent with the hypothesis that anionic O 2 radicals are being formed (data not shown), but a full investigation is beyond the scope of this study.
Next, we monitored the time course of simultaneous H 2 production and O 2 depletion with and without the presence of CpI

TABLE 1 Inactivation rate constants for CpI exposed to varying O 2 concentrations
The values were obtained from the aerobic H 2 production rates shown in Fig. 2c.  Fig. 3a because the slope decreases consistently by ϳ3 nmol/min, although the hydrogenase concentration varies by 100-fold. Fig. 2a indicates that the rate of electron loss from Fd red increases with increasing [O 2 ]. The immediate reduction in H 2 production rate after O 2 addition increases from 1.5 to 5.0 nmol/min as [O 2 ] increases from 1.0 to 5.0 vol %. After the addition of 10% vol % O 2 and higher, it appears that the rate of electron loss from Fd red to O 2 is so high that no additional H 2 production can be produced. Fig. 4b also indicates an immediate and consistent H 2 production rate decrease of 3 nmol/min, although [Fd] was increased 10-fold relative to the concentration used in the typical NADPH-driven assay (5 M). This observation suggests that the FNR activity is rate-limiting such that [SynFd red ] remains relatively constant and is primarily controlled by oxygen reactivity.
From the results shown in Fig. 4, the anaerobic TON with 50 nM CpI is about 4.5 s Ϫ1 . This can be compared with a k cat of about 400 s Ϫ1 when SynFd is reduced by DTH, implying that in the NADPH-driven reactions, [SynFd red ] is far below the K m of about 18 M. [SynFd red ] is estimated at about 0.2 M or only 4% of the 5.0 M Fd in a typical assay. This is consistent with the interpretation that FNR is rate-limiting in the NADPHdriven assay because G6P and G6PD will maintain a stable NADPH supply. The diversion of ϳ60% of the electron flux to O 2 will tend to stabilize [SynFd red ] at this low level, and this will also stabilize the rate of electron loss to O 2 . Consequently, the continuous decreases in H 2 production rate in the presence of O 2 will primarily reflect decreases in active  [CpI] times k cat , most likely the former quantity. Thus, exponential decay in H 2 production rate after O 2 addition appears to be an appropriate indicator of [FeFe] hydrogenase O 2 sensitivity.
Validating the NADPH-driven Assay with the DTH-driven Assay-To provide additional evidence that the previously described assay is correctly assessing O 2 tolerance, we implemented an alternative assay that assesses residual hydrogenase activity at multiple time points. Because reductive reactivation of O 2 -exposed [FeFe] hydrogenase has been reported (14, 16), we evaluated hydrogenase activity using an assay in which CpI received electrons from DTH via CpFd to maximize the available redox potential. Periodic samples were taken from the NADPH-driven assay, O 2 was removed, and CpI activity was measured using the DTH-driven reaction sequence. Unlike in the NADPH-driven reaction sequence, DTH maintains CpFd primarily in the reduced state so that the higher effective reducing potential might reactivate partially modified CpI. In addition, we show that the anaerobic rate of H 2 production from the DTH-driven assay is linearly proportional to active [CpI], at least between 0.1 and 20 nM (Fig. 6a). The changes in H 2 production rate from this assay, therefore, indicate changes in residual CpI activity.
We also used CpFd in this alternate assay because it supports a higher CpI TON (1700 versus 400 s Ϫ1 from SynFd) (10) and might be more likely to reactivate the hydrogenase. As shown in Fig. 6b, we obtained the same inactivation rate constant (k inact,O2 ϭ 0.099 min Ϫ1 ) as we did from the much more convenient NADPH-driven assay. In our opinion, this confirms that the time course of changes in aerobic H 2 production rates observed in the NADPH-driven assay (Figs. 2 and 3) represents oxygenic inactivation of CpI. All O 2 -exposed CpI samples produced H 2 at constant rates, suggesting that enzyme reactivation did not occur. Interestingly, we also obtained the same inactivation rate constant after incubating CpI with O 2 (without an electron supply) and assessing residual activity with the anaerobic DTH-driven assay (data not shown).

O 2 Tolerance of CpI-Fd Fusion Proteins Versus CpI-Previous studies have described [FeFe]
hydrogenase-ferredoxin fusions that appear to increase H 2 production rates and reduce electron loss to competing pathways (25,26). We hypothesized that such fusion proteins might also have different O 2 sensitivities. We therefore evaluated two fusion proteins that differ in the number and type of amino acids that link CpI to CpFd. In the first fusion protein, CpI-15aa-CpFd, the linker extended the CpI C terminus and consisted of 15 amino acids with three repeats of four glycines followed by a serine. In the second construction, CpI-55aa-CpFd, the linker was 55 amino acids long (details on the linker design can be found in Table 2) and was designed to be long enough so that the C-terminal extension could allow CpFd to approach its putative binding pocket on CpI.
Only the fusion protein with the shorter linker exhibited a significant enhancement in O 2 tolerance (Fig. 7) with a k inact,O2 of 0.079 Ϯ 0.009 min Ϫ1 as compared with about 0.092 Ϯ 0.010 min Ϫ1 for CpI and the longer fusion protein, a decrease in sensitivity of about 13%. These studies were conducted with 50 M RrFNR, 10 nM CpI or fusion protein, and 5.0 M free CpFd in [FeFe] Hydrogenase O 2 Sensitivity Assay OCTOBER 7, 2016 • VOLUME 291 • NUMBER 41 the NADPH-driven assay to allow a valid comparison between the wild-type (WT) CpI and the fusion proteins. Interestingly, CpI-15aa-CpFd also catalyzed H 2 production significantly faster (1.7-fold), presumably by supplying electrons to the active site at a faster rate (Fig. 7a). Previously, we speculated that higher electron flux to CpI might improve O 2 tolerance based on observations with higher [FNR] and lower [SynFd].
Here we observe the same correlation.

Discussion
In this study, we report the development of a convenient assay for assessing [FeFe] hydrogenase O 2 sensitivity. Reducing equivalents for H 2 production are supplied from NADPH via FNR and Fd as a more physiological solution phase alternative to PFV.
During assay development, we discovered a significant rate of electron loss directly from reduced Fd to O 2 . This occurred   with both SynFd and CpFd, although they differ significantly in molecular weight and the type of [Fe-S] clusters. Although the gradual decrease in [O 2 ] has the potential to lower the k inact,O2 over time, the excellent agreement of the rate data with the first order model suggests that this effect is not significant in the assay we offer. The assay was also validated using the DTH-driven activity assay that assesses periodic samples from the NADPHdriven assay. We obtained the same inactivation rate constant using the anaerobic DTH-driven assay (Fig. 6), which is performed with an electron source of higher redox potential. This validation would not have been possible if the hydrogen accumulation rates shown in Figs. 2 and 3 had been dominated by Fd red oxidation by O 2 rather than CpI inactivation. By using only the H 2 production rates after O 2 addition, the data from the NADPH-driven assay consistently indicate first order inactivation with rate constants insensitive to changes in initial [FeFe] hydrogenase concentrations. To further demonstrate the utility of the new assay, we show that a CpI-Fd fusion protein has a decreased inactivation rate constant relative to CpI possibly by improving the rate of electron supply to the enzyme.

CpI-15aa-CpFd
Biological photosynthetic H 2 production holds great promises for sustainable production of H 2 as an important fuel and industrial chemical. However, a major barrier has been the lack of a production enzyme that tolerates the O 2 produced as the unavoidable side product of photosynthesis. We believe that this convenient and potentially predictive assay for O 2 sensitivity will provide an important tool for the discovery of O 2 -tolerant hydrogenases. We also suggest that similar assay formats may be useful in assessing the functional properties of other redox-sensitive metallo-enzymes.

Experimental Procedures
Protein Preparation-RrFNR, SynFd, CpFd, and CpI were expressed in vivo in E. coli and purified as described previously (2,10,15). Two different fusion proteins in which CpI is linked to CpFd were designed and tested for effects on O 2 tolerance. The gene for the CpI-15aa-CpFd fusion (with a 15-amino acid linker composed of three repeats of Gly-Gly-Gly-Gly-Ser) was synthesized via assembly PCR and cloned into the pET21b vector with Gibson assembly. The gene for the other fusion protein, CpI-55aa-CpFd (with a 55-amino acid linker), was synthesized in the following two steps. First, the variation of the pET21b vector that contains the genomic sequence of the WT CpI was opened using NdeI and SaCI. Then, Gibson assembly was used to extend the gene to encode the linker and CpFd ( Table 2). The same expression and purification system used for obtaining WT CpI was employed in harvesting both fusion proteins (8). SDS-PAGE was used to verify the sizes of fusion proteins. The Bradford protein assay and methyl viologen assay were used to determine the total and active protein concentrations (8,9).
Investigation of the Oxygenic Inactivation of CpI Using Gas Chromatography (GC)-NADPH was purchased from Sigma-Aldrich, and DTH, G6P, and G6PD were purchased from Santa Cruz Biotechnology Inc. The solutions for both assays were prepared inside a N 2 -only glovebox (Vacuum Atmosphere Co.).
Each reagent (unless otherwise stated) was added in the order indicated to the following final concentrations: 50 mM Tris-HCl buffer, pH 7.0, 10 mM G6P, 2.0 units of G6PD, 5.0 mM NADPH, 5.0 M RrFNR, 5.0 M SynFd, and the CpI concentration indicated. 200-or 840-l reaction volumes were prepared in 2.0-ml target screw thread vials (National Scientific) or 8.4-ml crimp vials (Fisher Scientific), respectively. The 8.4-ml vials were used only in investigating the kinetics of O 2 consumption. Before sealing the vials with rubber septa, magnetic stir bars were added for mixing. 500 mM DTH was prepared separately in a new vial and added outside the glovebox by using a gas-tight syringe with a 25-gauge needle (Hamilton Co.).
After removal from the glovebox, the sealed vials were placed on a stir plate to initiate mixing at 235 rpm. Because all reagents for the NADPH-driven assay had been added inside the glovebox, some H 2 production took place before making the first time point measurement (denoted by t ϭ 0 min); the typical amounts ranged from 5 to 15 nmol, which we subtracted when presenting data. A gas-tight syringe with a valved 23-gauge needle (Hamilton Co.) was used to sample 100 or 200 l of the headspace every 10 min. H 2 and O 2 concentrations were determined using a Hewlett-Packard 6890 gas chromatograph (Hewlett-Packard) with a ShinCarbon ST 100/120 mesh column (Restek). All experiments were done in triplicate, and error bars in the figures represent S.D. O 2 was introduced to the headspace of the vials by injecting varying volumes of air and removing the same volume afterward (maintaining a constant headspace pressure before and after the introduction of O 2 ). In the experiments with 5.0 vol % O 2 , for example, the samples were prepared inside the N 2 -only glovebox, and 0.58 ml of air was typically injected after 10 min of anaerobic catalysis.
Measuring the Full Residual Activity after O 2 Exposure-Glass vials containing the NADPH-driven assay solution with 10 nM CpI (and 5.0 M CpFd replacing 5.0 M SynFd) were incubated with 5.0 vol % O 2 for 0, 10, 20, and 30 min. Next, the vials were purged with 100% N 2 for 2.5 min at a flow rate of 630 ml/min. This removed the O 2 previously introduced as well as accumulated H 2 . H 2 production was stimulated by adding 6.0 l of 500 mM DTH to a final concentration of 15 mM. H 2 concentration in the headspace was measured at three time points to determine the rate of DTHdriven H 2 production as an indication of residual CpI activity. All experiments were done in duplicate on two separate occasions, and error bars in the figures represent S.D.