Introduction

Hydrogen gas (H2) is a promising energy vector for the coming energy transition, due to its high energy/mass ratio and clean combustion [1, 2]. As of today, the lack of a cheap and efficient catalyst for H2 production prevents any possible large-scale application, since currently available technologies all rely on rare or precious elements [3,4,5,6]. [FeFe]-hydrogenases (Fig. 1A) are central to hydrogen metabolism in many microorganisms, as they catalyze both H2 oxidation and evolution [7]. Since their H2 evolution rates are as high as 10,000 s−1, these enzymes are also highly relevant in the context of developing sustainable biological and bio-inspired synthetic systems for H2/H+ interconversion [1, 2, 8,9,10,11]. The active site of [FeFe]-hydrogenases (the H-cluster, Fig. 1B) contains a unique diiron subcluster ([2Fe]H) where the low valent Fe ions are bridged by a bidentate azadithiolate ligand (−SCH2NHCH2S−, adt) and further coordinated by three CO and two CN− ligands [2, 12,13,14,15]. The [2Fe]H subsite is, in turn, attached to a canonical [4Fe-4S] cluster ([4Fe-4S]H) through a cysteine residue. The ligand geometry of the H-cluster leaves an open-coordination site on the [2Fe]H subsite positioned in close proximity to the nitrogen of the adt ligand, and the latter is believed to act as a proton relay during catalysis. [FeFe]-hydrogenases are extremely O2 sensitive enzymes, and this sensitivity represents a key challenge for technological development. The precise inactivation mechanism is still under debate, but O2 is, with few exceptions, an irreversible inhibitor of the enzyme [16,17,18,19,20,21]. However, it is possible to protect the H-cluster by forming a reversibly inhibited state, as first observed already in the 1980s in the [FeFe]-hydrogenases from sulfate reducing bacteria including Desulfovibrio vulgaris and Desulfovibrio desulfuricans (DvHydAB and DdHydAB, respectively, whose protein sequences are identical), following isolation of the enzymes under aerobic conditions [22,23,24,25]. The chemical nature of this O2 stable form has since then been characterized in-depth under in vitro conditions [20, 26,27,28,29,30]. The catalytically active, but O2 sensitive, Hox state has been observed in all [FeFe]-hydrogenases studied so far, and it exhibits an [Fe(II)Fe(I)]H subsite while the [4Fe-4S]H-cluster resides in the oxidized (2 +) state [2, 9, 13, 27] (Fig. 1B). The Hox state is also well-known for its affinity for CO, generating the reversibly inhibited Hox-CO state [31, 32]. In DdHydAB the coordination of a sulfide ligand (SH−) to the [2Fe]H subsite converts Hox into the isoelectronic Htrans state ([4Fe-4S]H+-[Fe(II)Fe(II)-SH]H), which instead of degrading upon O2 exposure is oxidized to the so-called Hinact state ([4Fe-4S]H2+-[Fe(II)Fe(II)-SH]H), also referred to as Hoxair [33, 34] (Fig. 1B). Albeit an inhibited state, the Hinact state is readily reactivated following release of the sulfide ligand under reducing conditions [19, 27, 29, 35].

Fig. 1
figure 1

Schematic representation of [FeFe]-hydrogenase and the H-cluster. A The dimeric Desulfovibrio desulfuricans [FeFe]-hydrogenase (DdHydAB) in the Hinact state, based on the crystal structure reported in ref [33] (PDB ascension number 6SG2). Subunit A is colored in gray, subunit B is colored in red, the color scheme for the H-cluster and for the accessory [FeS] clusters is based on the elements (Fe: cyan; S: yellow; N: blue; O: red; C: gray). B The H-cluster in the catalytically active state Hox and in the inhibited state Hinact

Full size image

Despite having been widely studied in purified form, our insight into the catalytic cycle and stability of [FeFe]-hydrogenases in a cellular environment remains limited. Improved understanding of [FeFe]-hydrogenase reactivity under whole-cell conditions is critical not only to verify the physiological relevance of states observed in vitro, but also to improve the performance of hydrogenase-based biological and biohybrid energy applications. Herein, we take advantage of our capacity to generate semi-synthetic [FeFe]-hydrogenases in E. coli at concentrations suitable for whole-cell spectroscopy [9, 36,37,38], to investigate the formation of the Htrans and Hinact states under intracellular conditions. More specifically, we explore two different model enzymes: HydA1 from the green algae Chlamydomonas reinhardtii (CrHydA1) and the aforementioned bacterial DdHydAB. Moreover, a recent mechanistic proposal underscores the need for efficient proton-transfer during the formation of these sulfide-inhibited states [34]. Thus, we probe the influence of sterics and the proton-transfer network on the inhibition chemistry of the CrHydA1 enzyme, through modifications of the organometallic cofactor as well as single-point mutations in the active-site pocket.

Results and discussion

Formation of an Htrans-like state under whole-cell conditions

To generate enzyme concentrations enabling H-cluster detection by electron paramagnetic resonance (EPR) spectroscopy CrHydA1 was heterologously expressed in standard E.coli BL21, as previously described [36,37,38,39]. As E. coli does not natively express an [FeFe]-hydrogenase, the organism lacks the [FeFe]-hydrogenase-specific accessory proteins (HydEFG) and produces the enzyme in an inactive form, complete with the [4Fe-4S]H cluster but lacking the diiron [2Fe]H cofactor (apo-CrHydA1) [40,41,42]. After protein overproduction, cell cultures were concentrated and the active semi-synthetic enzyme (holo-CrHydA1, herein denoted [2Fe]adt-CrHydA1) was subsequently generated through the addition of the synthetic [2Fe]H cofactor mimic [Fe2(adt)(CO)4(CN)2]2− ([2Fe]adt) under strictly anaerobic conditions. The final concentration of [2Fe]adt in the cell suspension was 80 μM, as this has previously been shown to afford close to quantitative maturation of apo-CrHydA1 under these conditions [36,37,38,39, 43]. The cell suspensions produced H2 for 1–2 h following the addition of [2Fe]adt, before slowing down significantly and halting production after 3–4 h (Fig. S1). In parallel to quantifying H2 production, whole-cell samples for EPR analysis were collected and flash frozen at different time points to monitor the formation and stability of the [2Fe]adt-CrHydA1 holoenzyme over the course of 23 h. EPR spectra recorded on samples obtained after a 20 min incubation with [2Fe]adt under an argon atmosphere displayed a mixture of one rhombic (gz ≠ gy ≠ gx) and one pseudo-axial (gx = gy ≠ gz) feature, respectively attributable to the Hox (gzyx = 2.100, 2.040, 1.998) and to the CO-inhibited Hox-CO states (g∥⊥ = 2.054, 2.007) [31] (Fig. 2). The Hox and Hox-CO species continued to be the main constituents of the signal over the first 3 h, with the latter species gradually releasing CO and converting into the Hox state. (Fig. 2B).

Fig. 2
figure 2

The effect of time on the population of H-cluster states in [2Fe]adt-CrHydA1 monitored by EPR spectroscopy. A Whole-cell EPR spectra of [2Fe]adt-CrHydA1 recorded on samples of E. coli cells expressing apo-CrHydA1 and flash frozen following incubation with the [2Fe]adt cofactor for 1 h, 2 h, 3 h and 23 h at 37 ˚C under anaerobic conditions. Experimental spectra from a representative dataset are overlaid with simulations (dashed lines), including Hox, Hox-CO and Htrans contributions. All spectra were corrected by subtracting the apo-CrHydA1 control sample, to eliminate any signal coming from the cells and from unmatured CrHydA1 (Fig. S4). The poor fitting around 330 mT on the 23 h spectrum is due to the imperfect subtraction of the variable g = 2.02 isotropic signal arising from the cell background, which is a minor contribution in the difference spectrum (see Fig. S4). EPR experimental conditions: T = 20 K, P = 1 mW, ν = 9.28 GHz. B Contribution of the different catalytic states to the signal intensity at the different time points on the same dataset. The relative contributions are calculated based on their respective weight in the simulations

Full size image

However, after 23 h of incubation, a significant decrease in the features specific to these two states is observed, concomitantly with the rise of a relatively broad rhombic signal with gzyx = 2.064, 1.972, 1.910. Simulations show that the latter species begins to form already after 2 h of incubation, at which point it represents ~ 10% of the total spin count. This rhombic signal accumulates with time to then become the dominant species in the sample, contributing for ~ 80% of the total signal intensity after 23 h. At this point, the cells are able to resume H2 production when re-suspended in fresh media containing glucose (Fig. S1); however, the media exchange has no effect on the relative proportion of the different states (Fig. S2). Similarly, the broad rhombic signal displayed negligible changes when parallel samples incubated for 23 h were exposed to strongly reducing conditions via the addition of sodium dithionite and 1 atm H2 (data not shown). In combination, these observations suggest that, once formed, the species giving rise to the rhombic signal is highly stable. Thus, the H2 produced by aged cells is attributed to the small residual population still residing in the catalytically competent Hox state.

The broad rhombic EPR signal (gzyx = 2.064, 1.972, 1.910) was not observed when cells expressing apo-CrHydA1 were incubated for up to 23 h in the absence of [2Fe]adt, nor in standard BL21 cells incubated with the same complex (Fig. S3), supporting the assignment of the signal to an H-cluster species. The g values of this new broad EPR signal are in good agreement with a signal previously attributed to the sulfide-inhibited Htrans state in DdHydAB (gzyx = 2.060; 1.960; 1.890) [25, 29]. To assess whether the origin of this new signal could be attributed to the binding of a sulfide or hydrosulfide ligand to the diiron subcluster, 375 μM Na2S were added to the incubation mixture concomitantly with [2Fe]adt. The addition of sulfide increased the overall intensity of the rhombic Htrans-like signal, as reported for the in vitro formation of the Htrans state in DdHydAB [34]. The same effect could be observed using L-cysteine as the source for the sulfide, albeit with a lower magnitude (Fig. S5).

Intracellular Htrans formation in DdHydAB

The heterodimeric DdHydAB hydrogenase has been the primary model system for elucidating the chemistry of the Htrans and Hinact states [29, 33, 34]. Thus, this enzyme was also studied under intracellular conditions for comparative purposes, with DdHydAB overproduced in E. coli using the same methods as for CrHydA1. In addition to the [4Fe-4S]H component of the H-cluster, DdHydAB features two additional [4Fe-4S] “F-clusters” serving as electron relays from the protein surface to the active site (Fig. 1A). The apo-DdHydAB protein, containing only the three [4Fe-4S] clusters, shows an interaction spectrum of the [4Fe-4S] clusters’ signals under intracellular conditions (Fig. 3, apo-DdHAB), as observed earlier for the purified enzyme [44]. In vitro activation of apo-DdHydAB using [2Fe]adt to generate [2Fe]adt-DdHydAB (i.e. holo-DdHydAB) is a relatively slow process. The mechanistic rationale for this remains to be elucidated, but is arguably attributable to the relative stabilities of the so-called “open” and “closed” conformations of the enzyme [44]. Nevertheless, treating apo-DdHydAB expressing E. coli cells with [2Fe]adt for just 1 h yielded a complex spectrum, with an axial signature attributable to an Hox-CO species clearly discernible under the dominant [4Fe-4S] interaction spectrum. The reason for the apparent higher rate of H-cluster assembly in DdHydAB under whole-cell conditions is unknown. Extending the incubation time to 24 h resulted in a decrease of the [4Fe-4S] signals, possibly due to gradual oxidation, allowing also for the detection of a contribution arising from a small population of [2Fe]adt-DdHydAB featuring the H-cluster in the Hox state and a one-electron-reduced F-cluster (denoted (Fred)Hox) [45]. However, no features attributable to Htrans or an Htrans-like state were observed, even when extending the incubation time to 72 h, at which point the only remaining H-cluster signal is Hox-CO (Fig. 3). These experiments verify that [2Fe]adt-DdHydAB can be generated through artificial maturation under whole-cell conditions, at a seemingly faster rate than during normal in vitro studies. Still, the lack of any Htrans-like species in the case of DdHydAB shows that the stability of this state under intracellular conditions can vary significantly between different [FeFe]-hydrogenases, as expected from variations in structure as well as reported affinities for other inhibitors such as carbon monoxide [46].

Fig. 3
figure 3

The effect of time on the population of H-cluster states in [2Fe]adt-DdHydAB monitored by EPR spectroscopy. Whole-cell EPR spectra of apo- and [2Fe]adt-DdHydAB recorded on samples of E. coli cells expressing apo-DdHydAB and flash frozen prior to or following incubation with the [2Fe]adt cofactor for 1 h, 24 h and 72 h under anaerobic conditions at 37 °C. The figure reports the g values for the Hox-CO state. (Inset): zoom in on the 320–330 mT area. Vertical dotted lines indicate the g = 2.065 feature belonging to Hox-CO and the g = 2.054 from the [4Fe-4S] clusters interaction spectrum. Asterisks mark a feature attributable to the (Fred)Hox interaction spectrum. EPR experimental conditions: T = 20 K, P = 1 mW, ν = 9.28 GHz. Note that the spectrum of apo-DdHydAB has been normalized by a factor 0.3

Full size image

Reactivity of the Htrans-like state towards molecular oxygen.

As mentioned in the introduction, in DdHydAB the Hinact state forms via Htrans under oxidative conditions (e.g. during aerobic purification) and has been demonstrated to grant protection from O2-induced H-cluster degradation. Conversely, CrHydA1 is known to be an extremely oxygen-sensitive enzyme and aerobic purification of an active enzyme has not been reported [20]. Still, CrHydA1 has also been shown to be able to enter the O2-protected Hinact state under controlled in vitro conditions [34]. To examine if exposing whole-cell [2Fe]adt-CrHydA1 samples to molecular oxygen could trigger the formation of Hinact or Htrans, and whether these states were able to grant resistance towards O2 damage, cell suspensions were exposed to air and the effects on the H-cluster were monitored via EPR spectroscopy and enzyme activity assays. As described above, incubating cells expressing apo-CrHydA1 with [2Fe]adt for 1 h under anaerobic conditions, generates an enzyme population consisting of Hox and Hox-CO. Subsequent exposure of such samples to O2, by placing them in air for 1 h, resulted in a noticeable decrease of the overall EPR signal intensity (total residual signal intensity ≈ 45%) (Fig. 4, panel A). Visibly, all of the residual signal is attributable to the Hox-CO state; this is expected as it is known to be more resistant to oxygen and to be formed by CO released upon H-cluster degradation (a process known as “cannibalization”) [27, 29]. The same air exposure treatment on samples pre-incubated under argon for 23 h, in which the H-cluster resides primarily in the Htrans-like state, yielded a similar result, with a decrease in the intensity of the spectral features of both the Htrans-like species and residual Hox (total residual signal intensity ≈ 35%) (Fig. 4, panel B). This loss of EPR signal intensity can be attributed either to the formation of an EPR silent Hinact-like state or to O2-induced degradation of the H-cluster. To separate these two possibilities, the EPR experiments were complemented with in vitro activity assays on lysed cells. The use of a strong reductant (dithionite), in combination with methyl viologen as electron mediator, is expected to re-activate any H-cluster fraction residing in Hinact or Htrans, and thus enable quantification of the entire intact H-cluster population.

Fig. 4
figure 4

Reactivity towards molecular oxygen of whole-cell [2Fe]adt-CrHydA1 samples. Panel A: The EPR spectrum of a whole-cell [2Fe]adt-CrHydA1 sample initially incubated for 1 h with [2Fe]adt under anaerobic conditions, and subsequently exposed to air for 1 h (red spectrum) is overlaid with an equivalent sample kept strictly anaerobic (gray spectrum). (Inset): Enzymatic activities for the same samples measured through in vitro assays. Panel B: The EPR spectrum of a whole-cell [2Fe]adt-CrHydA1 sample initially incubated for 23 h with [2Fe]adt under anaerobic conditions, and subsequently exposed to air for 1 h at ambient temperature (blue spectrum) is overlaid with an equivalent sample kept anaerobic (gray spectrum). (Inset): Enzymatic activities for the same samples measured through in vitro assays. EPR data shown come from representative spectra. Activity data represent the average values of two biological replicates with standard deviations reported as error bars. EPR experimental conditions: T = 20 K, P = 1 mW, ν = 9.28 GHz

Full size image

It should be noted that already prior to O2 exposure the 23 h pre-incubated samples did display slightly diminished activities, as compared to the 1 h pre-incubated samples (compare Fig. 4A and B, insets). We attribute this to H-cluster degradation, as in vitro assays show no indication of degradation processes involving the protein scaffold on a 24 h time scale (Fig. S4). The readily observable H2 production from the lysed cells enable a relative quantification of intact enzyme to determine the effect of O2 on the different preparations. Regardless of pre-incubation time (1 or 23 h), the signal loss observed by EPR spectroscopy following oxygen exposure was larger than the activity loss. The 1 h-pre-incubated sample retained ~ 70% of the H2-producing activity and the 23 h-pre-incubated sample retained ~ 50% of its activity, when compared to parallel samples kept under strictly anaerobic conditions (Fig. 4, insets).

Consequently, the observed EPR signal loss is attributable to a combination of H-cluster degradation and formation of an EPR silent Hinact-like state. Considering the well-established reactivation of Hinact under reducing conditions [29, 34], we expect the latter species to re-enter the catalytic cycle under the conditions employed in the in vitro activity assays, while the former causes a definitive loss in activity. Moreover, pre-established accumulation of the Htrans-like state did not provide any apparent increase in O2-tolerance, as samples pre-incubated for 23 h showed larger relative activity losses as compared to samples pre-incubated only for 1 h. In contrast, the cellular environment showed remarkable capabilities in terms of O2-protection, as a significant fraction of the H-cluster population was not degraded nor oxidized to an Hinact-like state during oxygen exposure. This is attributed to the low intracellular O2 concentrations ensured by cellular respiration, a process which is indeed expected to be more efficient in “young” and metabolically active cells [47].

Structural factors influencing formation of the Htrans and Hinact-like states

The presence of an efficient proton-transfer chain has been proposed to be important for the formation of the Htrans and Hinact states, as the sulfide ligand is thought to enter the active-site pocket as H2S and to undergo a protein-assisted deprotonation event to yield the SH− ligand [34]. Thus, modifications were made to both the [2Fe]H subsite as well as to the active-site pocket to assess the importance of the proton-transfer pathway on formation of the Htrans and Hinact-like states. More specifically, holo-CrHydA1 samples were prepared in which the [2Fe]adt mimic was replaced with [Fe2(pdt)(CO)4(CN)2]2− ([2Fe]pdt, pdt = −SCH2CH2CH2S−, propanedithiolate), generating [2Fe]pdt–CrHydA1. In this modified H-cluster, the amine bridgehead present in the natural [2Fe]adt cofactor is substituted with a non-protonatable methylene group. Earlier whole-cell and in vitro spectroscopic studies, as well as crystallographic studies, have shown that [2Fe]pdt–CrHydA1 is an organometallic variant where the [2Fe] subsite is locked in an oxidized Fe(II)Fe(I) state, strikingly similar to the EPR active Hox state of the native enzyme [14, 37,38,39, 43, 48, 49]. As expected, EPR samples consisting of apo-CrHydA1-expressing cells incubated with the [2Fe]pdt complex revealed the exclusive formation an Hox-like species. The H-cluster population remained locked in the Hox state for up to 23 h of incubation under anaerobic conditions; and no features attributable to an Htrans-like species were discernable at any point during the experiment (Fig. S6).

These results seem to confirm the importance of a proton relay in the second coordination sphere of the H-cluster for the formation of the Htrans and Hinact-like states. To proceed one step further down the proton-transfer pathway, the C169 residue was mutated to a serine (CrHydA1-C169S). This cysteine residue is strictly conserved in “prototypical” [FeFe]-hydrogenases, and it is the first amino acid involved in the proton-transfer chain towards the adt-nitrogen of the H-cluster [9]. The C169S mutant has been studied extensively under in vitro conditions, and demonstrated to be a variant with a severely compromised proton-transfer pathway and significantly decreased catalytic activity [50, 51]. Incubating E. coli cells expressing CrHydA1-C169S with [2Fe]adt resulted in the rapid formation of a broad rhombic EPR signal with gzyx = 2.064; 1.970; 1.910, very closely matching the Htrans-like signal in the wild-type enzyme (gzyx = 2.064; 1.972; 1.910). A signal with comparable EPR properties has been reported from in vitro studies of the C169S mutant (gzyx = 2.065; 1.969; 1.906) [51, 52]. It was originally assigned to an Htrans-like species, potentially featuring a hydride ligand. However, despite the similarity in g values the signal observed under our in vivo conditions is highly unlikely to represent a metal hydride species. No other H-cluster derived features were discernable in the EPR spectrum (Fig. 5).

Fig. 5
figure 5

Monitoring the formation of the Htrans-like state in [2Fe]adt-CrHydA1-C169S under whole-cell conditions by EPR spectroscopy. EPR spectra recorded for E. coli cells expressing apo-CrHydA1-C169S flash frozen after incubation with [2Fe]adt for 1, 3, 24 or 48 h at 37 °C. A characteristic Htrans-like signature became visible already after 1 h of incubation; a spectral simulation for this species is shown as a stacked dashed line and g values obtained from the fittings are reported. Unassigned, sample-specific weak signals potentially arising from the imperfect subtraction of the cell background are indicated with asterisks. EPR experimental conditions: T = 20 K, P = 1 mW, ν = 9.28 GHz

Full size image

The evolution of this Htrans-like signal over time was examined with long-term (1 h, 3 h, 23 h, 48 h) incubation of the cells with the [2Fe]adt cofactor, and no significant change was detected neither in the shape, nor intensity of the signal (Fig. 5). The stability of this state was further confirmed by its isolation. In contrast to the wild-type protein, [2Fe]adt-CrHydA1-C169S could be purified in its Htrans-like state via affinity chromatography following anaerobic cell lysis. However, this required supplementing the buffers with 100 mM Na2S, as omitting sulfide resulted in a complete loss of the Htrans-like rhombic EPR signal (Fig. S7). Albeit EPR spectroscopy revealed rapid H-cluster formation, no in vivo or in vitro hydrogen production could be detected from [2Fe]adt-CrHydA1-C169S samples, as expected from this reportedly inactive variant [50].

To gain more insight into potential intermediates formed on route to the Htrans-like state in the C169S mutant, the process was studied at lower temperature. More specifically, apo-CrHydA1-C169S containing cells were treated with [2Fe]adt at 12 ˚C. The first discernable state generated under these conditions was Hox (20–60 min), while the Htrans-like state only started to be detectable after 2 h of incubation with [2Fe]adt (Fig. S8). The Htrans-like signal reached full intensity after overnight incubation with the cofactor.

As modifications of the bridgehead atom showed clearly diverging effects as compared to the C169S mutant, with regards to formation of the Htrans-like state, the combined effect of the latter mutation and the methylene-bridgehead mimic was investigated. Whole-cell samples of the C169S mutant matured with [2Fe]pdt ([2Fe]pdt-CrHydA1-C169S) initially displayed a behavior consistent with the [2Fe]pdt wild-type, with an EPR spectrum showing an H-cluster apparently locked in an Hox- like state (Fig. 6). However, a signal corresponding to an Htrans-like state started to accumulate after 3 h of incubation and became predominant after 24 h. No other signals arose thereafter, on a time scale of days (96 h).

Fig. 6
figure 6

Monitoring the formation of the Htrans-like state in CrHydA1-C169S with the [2Fe]pdt cofactor. EPR spectra of E. coli cells expressing apo-CrHydA1-C169S after incubation with [2Fe]pdt for variable amounts of time at 37 °C. The Htrans-like signature only became visible after 23 h of incubation. Unassigned, sample-specific weak signals potentially arising from the imperfect subtraction of the cell background are indicated with asterisks. EPR experimental conditions: T = 20 K, P = 1 mW, ν = 9.28 GHz

Full size image

In summary, the [2Fe]pdt-variant of the wild-type CrHydA1 does not form detectable amounts of any Htrans or Hinact-like states, in agreement with previous studies performed in vitro [34]. Conversely, the [2Fe]pdt-CrHydA1-C169S ‘double’ variant showed accumulation of the Htrans-like state. This shows that the presence of a nitrogen-bridgehead facilitates, but is not critical, for the formation of these inhibited states. Rather, it underscores that additional factors in the outer coordination sphere also affect their formation. These include a decrease of steric bulk resulting from the replacement of a thiol group with an alcohol that provides easier access for exogenous ligands and a change in the hydrogen-bonding dynamics involving the first and second coordination spheres.

Conclusions

Semi-synthetic [FeFe]-hydrogenases have been shown to remain catalytically active in cyanobacteria on a time scale of days [43, 53]. The observation that E. coli cells expressing C. reinhardtii [FeFe]-hydrogenase can produce hydrogen gas only transiently following enzyme activation has raised the question on the stability of these enzymes in the bacterial cytoplasm.

In the case of CrHydA1, EPR spectra recorded on long time-scale samples show the gradual accumulation of an Htrans-like state under the slow-turnover conditions assayed here. However, the rate of formation of this Htrans-like state clearly varies between [FeFe]-hydrogenases, as DdHydAB did not form a similar state under identical experimental conditions. The latter observation is tentatively attributed to DdHydAB’s high affinity for CO, resulting in a relatively stable Hox-CO population unable to convert to the Htrans-like state under the conditions employed here.

In light of its stability, we propose that the observed Htrans-like state represents a thermodynamic sink for the enzyme accessed under the whole-cell conditions employed here. Moreover, its formation is evidently accelerated by mutating amino-acid residues close to the open-coordination site of the cofactor, and by the presence of an amino-group in the bridging dithiolate ligand. Models have been proposed that focus on the role of the hydrogen-bonding network in stabilizing apical ligands, therefore, it can be hypothesized that the nature of the H-bond partners in the vicinity of the apical site can significantly influence the formation of an inhibited state. [54] In addition, a cysteine-to-serine mutation has the effect of reducing the steric hindrance close to the open-coordination site of the cofactor, which could further facilitate the insertion of an inhibitor ligand. The positive effect of adding sulfide for generating this species and stabilizing it, as during isolation of [2Fe]adt-CrHydA1-C169S, suggests that the observed rhombic EPR signal can be attributed to the in vitro characterized Htrans state. The observation of Htrans under whole-cell conditions would support a physiological relevance of this state. However, we note that the in vivo data does not unequivocally prove the structure of this species, and further investigations on the nature of this Htrans-like state are underway.

Albeit the Htrans-like state appears stable, it does not seem to offer substantial protection against oxygen damage, in contrast to in vitro observations [33, 34]. Conversely, E. coli cells were found to provide a surprisingly high level of protection versus air, highlighting the potential of E. coli as a suitable host organism for oxygen-sensitive enzymes in bio (hybrid) technological applications.

In closing, the spontaneous formation of inactive species could pose a problem in the development of systems that take advantage of bacterial hosts to develop whole-cell hydrogen production systems, as it potentially shortens their service life. Having discovered the critical role of the steric occupancy and the hydrogen-bonding network of the active-site pocket in regulating the kinetics of this phenomenon, a possible path for research could be to fine-tune these parameters and thus develop potentially more robust catalysts.

Experimental procedures

General. All chemicals were purchased from Sigma-Aldrich or VWR and used as received unless otherwise stated. All anaerobic work was performed in an MBRAUN glovebox ([O2] < 10 ppm). The expression vector encoding the hydA1 gene (pETDuet-CrHydA1-His) was kindly provided by Prof. Marc Fontecave (College de France, Paris/CEA, Grenoble). (Et4N)2[Fe2(adt)(CO)4(CN)2] ([2Fe]adt]) and (Et4N)2[Fe2(pdt)(CO)4(CN)2] ([2Fe]pdt]) were synthesized in accordance to literature protocols with minor modifications, and verified by FTIR spectroscopy [55,56,57,58]. The complexes were dissolved in anaerobic potassium phosphate buffer (100 mM, pH 6.8) at 10 µg/µL concentration and used directly. Protein content was analyzed by 10% SDS-PAGE minigels in a BioRad Mini-PROTEAN Tetra Cell system. Protein bands were stained with Page Blue protein staining solution (Thermo Fisher Scientific) according to the supplier`s instructions.

Generation of CrHydA1-C169S mutant. To generate the C169S mutant in HydA1 protein, the triplet coding for the cysteine amino acid in position 169 (TGC) was replaced with one coding for serine (TCC). The pETDuet-CrHydA1-His was mutated with the Quick Change Site-Directed Mutagenesis Kit (Agilent) according to the manufacturer instructions using 5` GTTTACCAGCTCCTGCCCGGGCTGGATTGC` 3` and 5` GCAATCCAGCCCGGGCAGGAGCTGGTAAAC` 3` primers. The amino-acid change was verified by sequencing.

Overexpression of the apo-CrHydA1 hydrogenase. Escherichia coli BL21(DE3) cells containing the CrHydA1 plasmid were grown in 50 mL M9 medium [22 mM Na2HPO4, 22 mM KH2PO4, 85 mM NaCl, 18 mM NH4Cl, 0.2 mM MgSO4, 0.1 mM CaCl2, 0.4% (v/v) glucose] under aerobic conditions until O.D.600 = 0.6 – 0.8 in the presence of ampicillin. The protein overproduction was induced with 1 mM IPTG and persisted at 20 °C for 16 – 18 h with continuous aeration. The media was supplemented with 100 µM FeSO4 at the time of the induction. Final O.D.600 of the cultures were 1.4 ± 0.2.

Overexpression of the apo-DdHydAB hydrogenase. Escherichia coli BL21(DE3) cells containing the pACYCDuet-DdHydAB plasmid[44] were grown in 50 mL M9 medium [22 mM Na2HPO4, 22 mM KH2PO4, 85 mM NaCl, 18 mM NH4Cl, 0.2 mM MgSO4, 0.1 mM CaCl2], supplemented with 0.4% (w/v) glucose, under aerobic conditions until O.D.600 = 0.6 – 0.8 in the presence of chloramphenicol. The protein overproduction was induced with 1 mM IPTG and persisted at 20 °C for 16–18 h with continuous aeration. The media was supplemented with 400 µM FeSO4 at the time of the induction. Final O.D.600 of the cultures were 1.4 ± 0.2.

In vivo formation of [2Fe]adt-CrHydA1, [2Fe]pdt-CrHydA1, [2Fe]adt–DdHydAB and [2Fe]pdt-DdHydAB. The preparation of the semi-synthetic hydrogenase was performed following literature protocols with minor modifications [38, 39]. The apo-protein was expressed in 50 mL E. coli cultures as described in the “Overexpression of the apo-CrHydA1 hydrogenase” and “Overexpression of the apo-DdHydAB hydrogenase” sections. After the 16–18 h expression period the cells were harvested, deaerated and transferred to the glovebox. The cells were re-suspended in fresh M9 medium (2 mL final volume), and formation of [2Fe]adt–CrHydA1, [2Fe]pdt–CrHydA1, [2Fe]adt–DdHydAB and [2Fe]pdt-DdHydAB was achieved by incubating the cell suspensions with 100 µg (156 nmol) [2Fe]adt or 100 µg (156 nmol) [2Fe]pdt complex (80 μM final conc.), for 1 – 96 h at 37 ˚C under strictly anaerobic conditions. When indicated, the medium was supplemented with 375 µM L-cysteine or 375 µM Na2S.

For the medium exchange experiments, after 23 h incubation the cell suspensions were transferred in the glovebox, centrifuged and the resulting pellet was re-suspended in fresh M9 media, supplemented with 0.4% (w/v) glucose when necessary. The cell suspensions were then incubated for 2 h at 37 ˚C under strictly anaerobic conditions.

Whole-cell EPR sample preparation. The 2 mL concentrated cell suspensions generated via the “In vivo formation of [2Fe]adt-HydA1, [2Fe]pdt-HydA1, [2Fe]adt–DdHydAB and [2Fe]pdt-DdHydAB” protocol were centrifuged and the cell pellet washed with 1 mL TRIS–HCl buffer (100 mM TRIS, 150 mM NaCl pH 7.5) three times under anaerobic conditions. After washing, the cells were diluted to a final volume of 400 µL using the same buffer and transferred into EPR tubes. The tubes were capped and directly frozen in liquid nitrogen.

Isolation of [2Fe]adt-CrHydA1-C169S. 100 mL of C169S-CrHydA1 cells were activated with 100 µg [2Fe]adt for 1 h. After activation the cells were lysed in lysis buffer (50 mM sodium phosphate buffer pH 7.8, 100 mM NaCl, 20 µg/mL DNaseI, 40 µg/mL RNaseI, 1 mM MgCl2, 0.1 mg/mL lysozyme) using three freeze–thaw cycles. The soluble fraction of the cell lysate was separated with 13,000 rpm centrifugation for 20 min under anaerobic conditions. The protein was purified using a 1 mL Ni-sepharose High Performance (GE Heathcare) gravity column equilibrated with 10 column volumes of equilibration buffer (50 mM sodium phosphate buffer pH 7.8, 100 mM NaCl) in the glovebox. The non-specifically bound proteins were removed with wash buffer (50 mM sodium phosphate buffer pH 7.8, 100 mM NaCl, 20 mM imidazole). The [2Fe]-C169S-CrHydA1 protein was eluted with 1.5 mL elution buffer (50 mM sodium phosphate buffer pH 7.8, 100 mM NaCl, 300 mM imidazole), the elution fraction was directly used for EPR samples. The purity of the purified protein was confirmed with SDS-PAGE.

Hydrogenase activity measurements. Activity measurements were performed on whole-cell samples as well as under in vitro conditions using published protocols [39]. Selected technical details are re-iterated for clarity. Whole-cell activities were determined from cell suspensions (2 mL final volume), incubated in gastight vials with a total volume of 9 mL. In vitro assays were performed on cell lysates. The reaction mix contained potassium phospate buffer (pH 6.8, 100 mM), Triton-X 1.5% v/v and methyl viologen (14 mM);the reaction was initiated with the addition of dithionite (200 mM) and the sample vials incubated for 15 min at 37 °C prior to headspace sampling. Hydrogen production was determined by analyzing the headspace gas, using a gas chromatograph (GC; PerkinElmer LLC, MA, USA) equipped with a thermal conductivity detector (TCD) and a stainless-steel column packed with Molecular Sieve (60/80 mesh). A calibration curve was established by injecting known amounts of hydrogen. The operational temperatures of the injection port, the oven and the detector were 100 °C, 80 °C and 100 °C, respectively. Argon was used as the carrier gas at a flow rate of 35 mL min−1.

EPR measurements. The EPR spectra shown are representative signals from at least two individual experiments. The individual experiments show some preparation dependent differences, but the amplitude of these background signals are negligible compared to the signal intensity of the [2Fe]adt activated CrHydA1. Measurements were performed on a Bruker ELEXYS E500 spectrometer using an ER049X SuperX microwave bridge in a Bruker SHQ0601 cavity equipped with an Oxford Instruments continuous flow cryostat and using an ITC 503 temperature controller (Oxford Instruments). Measurement temperatures ranged from 10 to 20 K, using liquid helium as coolant, with the following EPR settings unless otherwise stated: microwave power 1 mW, modulation amplitude 1 mT, modulation frequency 100 kHz. The spectrometer was controlled by the Xepr software package (Bruker).

EPR spectra processing and simulations. The EPR spectra were processed using the softwares Matlab (Mathworks, Inc) and QSoas [59]. Matlab served for converting the EPR files to ascii format, while QSoas was used to display the spectra as a function of g values, for visual inspection and subtraction of background signals emerging from the cells. The processed signals were used for Figs. 2, 3, 4, 5, 6, S2 and S3, S5 to S8. The simulations were performed using the Easyspin toolbox (5.2.23) within Matlab [60]. Other details of the procedure can be found in [38].