The final steps of [FeFe]-hydrogenase maturation

Significance

The maturation of each hydrogenase class describes the process that converts an apo-hydrogenase into an active holoenzyme involving a machinery of maturases. For [FeFe]-hydrogenases, the synthesis and insertion of the catalytic dinuclear [2Fe_H]-cofactor is executed by the 3 maturases (HydE, HydF, and HydG) that have already been studied extensively. However, the final steps of maturation occurring within apo-hydrogenase (cofactor recognition and integration) have not been elucidated yet. Herein, we postulate a molecular mechanism, exemplifying how a synthetic cofactor mimic is directed to its binding cavity, which precisely locks the cofactor to unlock its catalytic potential. Given the striking similarities to the maturation processes of other redox enzymes, these findings illuminate common mechanistic principles that govern cofactor insertion into metalloproteins.

Abstract

The active site (H-cluster) of [FeFe]-hydrogenases is a blueprint for the design of a biologically inspired H₂-producing catalyst. The maturation process describes the preassembly and uptake of the unique [2Fe_H] cluster into apo-hydrogenase, which is to date not fully understood. In this study, we targeted individual amino acids by site-directed mutagenesis in the [FeFe]-hydrogenase CpI of Clostridium pasteurianum to reveal the final steps of H-cluster maturation occurring within apo-hydrogenase. We identified putative key positions for cofactor uptake and the subsequent structural reorganization that stabilizes the [2Fe_H] cofactor in its functional coordination sphere. Our results suggest that functional integration of the negatively charged [2Fe_H] precursor requires the positive charges and individual structural features of the 2 basic residues of arginine 449 and lysine 358, which mark the entrance and terminus of the maturation channel, respectively. The results obtained for 5 glycine-to-histidine exchange variants within a flexible loop region provide compelling evidence that the glycine residues function as hinge positions in the refolding process, which closes the secondary ligand sphere of the [2Fe_H] cofactor and the maturation channel. The conserved structural motifs investigated here shed light on the interplay between the secondary ligand sphere and catalytic cofactor.

Hydrogenases are the most interesting natural blueprints for designing sustainable H₂-producing catalysts (1–3). The highly active [FeFe]-hydrogenases contain a unique catalytic cofactor (H-cluster) that is housed in a strictly conserved protein domain. The H-cluster comprises a “standard” cubane [4Fe-4S] cluster ([4Fe_H] site) being directly coupled via the thiol group of a bridging cysteine to a unique binuclear iron moiety (the [2Fe_H] cluster) (4–6). The 2 Fe atoms of the [2Fe_H] cluster (proximal p and distal d, differentiated according to their relative distance to the [4Fe_H] cluster) are coordinated by 2 nonproteinogenic CN⁻ and 3 CO ligands that are responsible for stabilizing the [2Fe_H] cluster as a low-spin, low-oxidation-state complex (Fig. 1B) (7–11). The open coordination site at Fe_d in the oxidized ready state (H_ox) is regarded as the location of catalytic H⁺/H₂-turnover, which can reach significant turnover rates of up to 10,000 s⁻¹ (12–15). While the knowledge about the catalytic cycle has increased significantly during the last 5 y, data on the process of H-cluster maturation and cofactor integration are lagging behind. It is known that the unmaturated preform of the [FeFe]-hydrogenase (apo-hydrogenase), which contains the active site [4Fe_H] cluster as well as any accessory [FeS] centers for transferring electrons, is synthesized first (16, 17). The maturation process for the [2Fe_H] cluster describes the molecular pathway of the preassembly of a symmetric [2Fe_H] precursor, carrying 4 CO and 2 CN⁻ ligands, and its transfer and integration into the H-cluster binding site of the apo-hydrogenase. The biosynthesis of the [2Fe_H] precursor is accomplished by the synergy of 3 maturases (HydG, HydE, and HydF) and is to date not fully characterized (13, 16, 18). HydG and HydE are both monomeric radical S-adenosyl-l-methionine (SAM) proteins and contain the CX₃CX₂C motif, which is characteristic of this enzyme superfamily. This motif contains an incompletely coordinated [4Fe4S] cluster that cleaves SAM into methionine and a 5′-deoxyadenosyl radical, which is then used in reactions that require radical chemistry (19, 20). HydG synthesizes the 2 low-spin, mononuclear iron synthons ([Fe(CO)₂CN]) as subunits of the [2Fe_H] precursor (21). At the end of the HydG (αβ)₈ triosephosphate isomerase (TIM) barrel, a cubane [4Fe-4S] cluster binds SAM (22, 23). At the opposite end of the active site cavity, another auxiliary FeS-site can be observed, being either a [4Fe-5S] or a [5Fe-5S] cluster, the latter one corresponds to a cubane cluster that is coupled via a µ₂-sulfide to a mononuclear Fe^II-center. This “labile” dangler iron is assumed to be the site of synthon formation (24). Synthon assembly is a multilevel synthetic process which includes the cleavage of a tyrosine into p-cresol and dehydroglycine, providing one CO and one CN⁻ ligand (25). HydE, which has been suggested to provide the proton-shuttling azadithiolate (adt) ligand, is the least characterized of the 3 maturases as the chemical nature of its substrate(s) and product are still under debate (26). The homodimeric GTPase, HydF, is in the center of this process, fulfilling the dual role 1) of acting as a scaffold for the [2Fe_H] precursor assembly in which this precursor is docked to the [4Fe-4S] cluster of HydF via one inverted CN⁻ ligand and 2) of being the carrier which inserts the finished precursor into the apo-hydrogenase protein (27, 28). For the latter process, a cluster of 7 amino acids was suggested for HydA1 to partake in the electrostatic interaction, which facilitates HydF/apo-hydrogenase complex formation during [2Fe_H] cofactor transfer (29–31).

Fig. 1.

(A) Schematic representation of the [2Fe]^MIM-uptake channel, which guides the cofactor into the open H-cluster binding site of apo-CpI. This transfer process is initiated by a transient electrostatic contact to R449 and leads to anchoring beside the [4Fe_H] site via H-bond contact to K358. (Surface charge distribution has been calculated for apo-CrHydA1 [PDB ID code 3LX4] using the Adaptive Poisson–Boltzmann Solver plug in.) (B) Stick model of the H-cluster with nearby amino acids from the plug region as well as K358 (black) and its essential H-bond contact to CN_d. Color codes for H-cluster are as follows: gray, carbon; blue, nitrogen; red, oxygen; yellow, sulfur; and orange, iron. (C, Left) A major structural rearrangement of the 3 flexible loop elements plug (cyanine), lid (magenta), and lock (red) upon [2Fe]^MIM transfer through the cationic maturation channel (blue) is proposed to be essential for the stable and functional integration of the [2Fe_H] cofactor in its binding niche (gray). (C, Right) Cartoon model of the 3 distinct regions in CpI (PDB ID code 4XDC) in the closed conformation, which are shown in cyanine (plug: 405 to 423), magenta (lid: 437 to 453), and red (lock: 529 to 540), respectively. The positively charged arginine residue (blue), which is assumed to initiate the [2Fe]^MIM import, and 5 highly conserved glycine positions (black), which might serve as hinge sites in the structural rearrangement, are shown as sticks.

As recently shown, a synthetic mimic of the [2Fe_H] cofactor ([2Fe]^MIM) can be employed to activate apo-hydrogenase, yielding a semisynthetic holo-hydrogenase that is indistinguishable from its native form (5, 32). For the process of cofactor integration into apo-hydrogenase, either in vivo or in vitro, a self-assembly mechanism has been postulated, involving a positively charged [2Fe_H]-precursor uptake channel to guide the cofactor from the protein surface to the H-cluster binding site (16). This channel is visible in the crystal structure of unmaturated hydrogenase I from Chlamydomonas reinhardtii (CrHydA1) and is only accessible in apo-hydrogenase (Protein Data Bank [PDB] ID code 3LX4); in all available holo-[FeFe]-hydrogenase structures, it is blocked by 3 flexible loop regions (13, 16, 29). In apo-CrHydA1, these 3 regions have previously been referred to as “plug,” “lock,” and “lid” due to their potential functional contributions in the final stages of the H-cluster maturation process (Fig. 1C). In the last steps of cofactor integration, they presumably rearrange to the “closed” state, which is visible in the structure of holo-hydrogenase I from Clostridium pasteurianum (CpI) (Fig. 1C) (13). A multiple alignment of 828 [FeFe]-hydrogenase sequences revealed several highly conserved amino acids within the maturation channel (33), including a strictly conserved arginine located at its entrance whose basic residue might serve as a first interaction point for the incoming negatively charged [2Fe]^MIM complex (Fig. 1 A and C). In this study, we experimentally explored the 3 self-assembly steps of H-cluster biosynthesis in apo-hydrogenase after preassembly of the [4Fe_H] site: 1) the funneling of the [2Fe_H] complex into the maturation channel leading from the surface to the H-cluster binding site through intermittent contacts to basic residues; 2) the coupling between the [4Fe_H] and [2Fe_H] site, accompanied by the release of one CO ligand; and 3) the completion of the secondary ligand sphere of the [2Fe_H] site by closing the maturation channel via the structural rearrangement of plug, lid, and lock. Selected amino acids within the respective regions in CpI were targeted by site directed mutagenesis and compared with the wild-type (WT) enzyme with respect to efficiency and persistence of the maturation process by applying in-solution kinetic assays, attenuated total reflection–Fourier transform infrared (ATR-FTIR) spectroscopy and protein film electrochemistry (PFE).

Results

The process of cofactor integration was investigated in terms of 2 consecutive steps: (1) [2Fe_H]^MIM uptake into the H-cluster binding site and (2) the structural rearrangement of plug, lock, and lid, which completes the secondary coordination sphere of the incorporated [2Fe_H] cofactor and thus stabilizes it in its catalytically functional configuration. For step 1, the highly conserved, surface exposed arginine R449 was selected to investigate its potential “gatekeeper” function of capturing the doubly negatively charged [2Fe]^MIM complex and directing it into the maturation channel leading to the [2Fe_H]-binding site. To test this hypothesis, the positively charged arginine was replaced by an uncharged alanine or a histidine for which the charge can be altered (neutral/positive) by changing the pH. Being part of the H-cluster binding site, K358 represents one of the major anchor points of the [2Fe_H] cluster as in vivo maturated variants entirely lack the [2Fe_H] subsite when it is substituted by asparagine (34). Moreover, residue K358 was considered not only to be a final recognition site for the [2Fe]^MIM complex but also to be crucial for stabilizing the rotated and catalytically active configuration of the [2Fe_H] cluster after subcluster coupling. To investigate step 2, 5 highly conserved glycine residues within the plug region (405–423^CpI) were targeted individually by site-directed mutagenesis to examine their roles in encapsulating the [2Fe]^MIM complex after in vitro maturation. To limit the spatial freedom conveyed by these presumptive hinge positions, we substituted these glycines individually by histidine, being a more bulky residue which conveys less rotational freedom during structural reconfigurations. To test our hypothesis, 2 of these selected glycine residues were additionally substituted for alanine, as this residue retains a larger degree of flexibility in this loop region that might still enable the proposed structural rearrangement.

Characterization of Holoproteins.

Before all kinetic investigations, we determined the iron content of each exchange variant by the method of Fish (35). A fully assembled [4Fe_H] site is an essential precondition for a successful integration and coupling to the [2Fe]^MIM complex in the first place. Apo-CpI WT and all exchange variants featured ∼18 iron atoms per enzyme (SI Appendix, Table S1), confirming not only the presence of all accessory clusters but also a fully assembled [4Fe_H] site, which is a precondition for [2Fe_H]-subcluster coupling.

To obtain catalytic activity, CpI WT and all respective protein variants were maturated in vitro at pH 6.8 with a 10-fold molar excess of [2Fe]^MIM vs. apo-hydrogenase. Subsequent size-exclusion chromatography was carried out to remove any excess [2Fe]^MIM, and H₂-production rates were determined for each protein variant at pH 6.8. For the 2 exchange variants targeting position R449, which marks the entrance to the maturation channel, a drastic decrease in H₂-production activity was observed. The Arg → Ala substitution limited the remaining activity to 3%, while this residual activity was tripled if arginine was replaced by histidine (Fig. 2A).

Fig. 2.

(A) In vitro H₂-evolution assays for CpI WT and exchange variants. Rates were determined at pH 6.8 by analyzing the head space composition of the following reaction mixture: 0.1 M KpI (pH 6.8), 10 mM methyl viologen, 0.1 M NaDT, and 400 ng of protein. Shaded bars denote H₂-evolution activities after maturation with an E. coli lysate containing HydE, HydF, and HydG from S. oneidensis. Blue bars denote exchange variants involved in cofactor uptake; green bars denote exchange variants within the plug region. Error bars represent SDs (n = 3). (B) FTIR spectra of holo-CpI WT and the protein variants (0.5 mM). Note the different scale for the exchange variants as indicated by the red lines.

On the other hand, the exchange variant K358N showed no remaining H₂-evolution activity at all. To verify this role of R449 and K358 for the in vivo maturation process, we maturated CpI WT and both exchange variants with an Escherichia coli lysate containing coexpressed maturases HydE, HydF and HydG from Shewanella oneidensis instead of using the [2Fe]^MIM complex (36). The results indicate that the activation level of R449H is comparatively small (no activation for K358), regardless of whether maturases are used for cofactor transfer to apo-CpI or a surplus of synthetic [2Fe]^MIM complex is added.

All 5 plug region variants exhibited drastically decreased H₂-evolution activities ranging from 0 to 3% activity when glycine was substituted by the more sterically hindering histidine (Fig. 2A). The more conservative exchange to alanine at positions G414 and G418 instead led to residual activities of 98 and 14%, respectively. FTIR spectroscopy conveys deeper insights into the vibrational modes of the Fe-bound CO and CN⁻ ligands (2,100 to 1,780 cm⁻¹), which can be monitored without the interference of the strong vibrational signals from water or the polypeptide backbone (amide bands). Noncatalytic and catalytic states can be accumulated under certain conditions, reflected by the appearance of state-specific patterns of CO and CN⁻ stretching frequency signals of the [2Fe_H] site (37–41). We investigated the level of cofactor occupancy for all holoproteins via ATR-FTIR spectroscopy (Fig. 2B). This technique allowed us to correlate the H-cluster occupancy for each variant with respective in vitro H₂-production activities. For fully maturated hydrogenase, the ratio of the amplitudes of the ligand vibrational signal from CO_d in H_ox at 1,946 cm⁻¹ and amide band I at 1,650 cm⁻¹ typically amounts to ∼0.1. This ratio was selected as an internal standard to determine the fraction of maturated sample (SI Appendix, Fig. S1). While CpI WT and exchange variant G414A feature the characteristic H-cluster spectrum for the H_ox state in terms of signal intensity, all other protein variants exhibit a significantly lower H-cluster occupancy which corresponds to their greatly impaired H₂-evolution activities. Except for R449H, G418A, and G421H, none of the other exchange variants showed a typical H-cluster spectrum.

Monitoring the Maturation Process of Apo-Hydrogenase with [2Fe]^MIM.

To attain a deeper insight into the impact of each individual amino acid substitution on the maturation process, we performed CO-release tests. Monitoring the release of CO allows us to follow the rate of [4Fe_H]- and [2Fe_H]-subcluster coupling after successful uptake of [2Fe]^MIM. While WT enzyme shows a rapid release of CO within the first 50 s after addition of [2Fe]^MIM, an almost stationary phase is reached after 150 s, indicating that the reaction is complete (SI Appendix, Fig. S2). A diminished CO-release activity is observed for variant R449A, while its conservative amino acid substitution to histidine, which still retains the positive charge at this position, reaches a WT-like level of CO release after a lag phase within the first 150 s. For variant K358N, only a slight increase in Hb-CO concentration is observed within the time course of 500 s, which can be ascribed to decomposition of the [2Fe]^MIM complex in solution (32). This result correlates well with the lack of any H₂ production or visible H-cluster bands in FTIR spectroscopy for this variant. All CpI protein variants with G → H substitutions within the plug region showed strongly diminished CO-release activities, indicating correspondingly low rates of subcluster coupling.

The FTIR spectrometer was equipped with an accessible ATR-crystal interface to enable targeted and gradual manipulations of system parameters during repetitive spectroscopic measurements of the protein sample. These parameters included purging gas composition, gas flow, pH, and constituent concentrations, as well as sample drying, which increases the local protein concentration. We exploited the accessibility of this setup to follow the in vitro maturation process for CpI on the ATR crystal. A splitting of the vibrational bands of the CO and CN⁻ ligand signals of [2Fe]^MIM indicates the adoption of a unique secondary coordination sphere for each ligand. For CpI WT, this occurs within the range of 20 to 30 s (SI Appendix, Fig. S3). After only 10 min of drying in a continuous N₂ gas flow, we obtained the characteristic H-cluster spectrum of WT-CpI, confirming previous data suggesting that maturation is largely complete after several minutes (5, 32, 42). We supplemented apo-CpI WT and each of the variants with [2Fe]^MIM at 2 different ratios, 1:1 and 1:3, respectively. Using a 1:10 excess of [2Fe]^MIM in this experiment would entirely mask the spectrum of the integrated H-cluster. After loading both components on the ATR crystal, we recorded the development of H-cluster spectrum during a 10 min period of sample drying under N₂. This method allowed us to evaluate the relative kinetics of cofactor integration and of the attainment of the catalytically competent ligand configuration, without the requirement to take into account any long-term cofactor loss, which can be assumed to be a critical factor in case of the purified holoenzymes (Fig. 3). For CpI WT, the development of a characteristic H-cluster spectrum showed similar kinetics, regardless of whether maturation was monitored for a mixture with 3-fold molar excess of [2Fe]^MIM or for an equimolar ratio of complex and apo-CpI, thereby confirming that a 1:1 ratio is sufficient to achieve full activation (SI Appendix, Fig. S4).

Fig. 3.

FTIR spectra observed for in vitro activation on the ATR crystal for apo-CpI (0.5 mM) and variants at pH 7 with [2Fe]^MIM. (A) Amino acids putatively involved in [2Fe_H]-cofactor transfer highlight the limited tolerance toward structural changes for cofactor recognition and subsequent integration. (B) Amino acid substitutions within the plug region cause a severe impairment of [2Fe]^MIM integration for all G → H substitutions compared with WT. All apo-proteins were maturated with 3-fold molar excess of [2Fe]^MIM. The samples were applied on the ATR crystal (15 µL) and immediately dried under a constant N₂ stream for 10 min at 30 °C. The protein samples were kept in Tris buffer, pH 7, and 2 mM NaDT.

Monitoring the maturation process on the ATR crystal for CpI variants with impaired [2Fe_H]-cofactor integration allows for greater insight into the determinants of the last steps of H-cluster maturation. The conserved surface-exposed arginine (R449) at the entrance to the maturation channel most likely plays an essential role in the uptake of [2Fe]^MIM since infrared (IR) spectroscopy of exchange variant R449A shows no indication of H-cluster formation, an observation that correlates well with its very low residual H₂-production activity of 3%. In contrast, the development of the FTIR vibrational spectrum of R449H demonstrates the formation of an intact H-cluster when H-cluster activation is executed at pH 7, reaching ∼50% of the H-cluster occupancy of CpI WT (Fig. 3A). When activation was carried out at pH 9, however, only slight fractions of integrated [2Fe]^MIM were observed, which may be explained by the deprotonation and loss of positive charge at the histidine sidechain (pK_a: 6). The capability of K358N (located at the opposite end of the uptake passage) to generate an active H-cluster is significantly more affected as it neither adopts a characteristic H-cluster spectrum nor exhibits any catalytic activity, even after prolonged exposure to a 10-fold concentration of [2Fe]^MIM.

For the exchange variants targeting the plug region, this method provided deeper mechanistic insights into H-cluster integration and configuration. The His variant at position G412, which is located at the beginning of the plug region, had approximately 40% [2Fe_H] occupancy when using a 1:1 ratio of apo-hydrogenase and [2Fe]^MIM; however, this variant adopted a higher fraction of H_ox-CO compared with WT (SI Appendix, Fig. S5). An almost WT-like integration was achieved for variant G412H when using a 3-fold molar excess of [2Fe]^MIM, although the relative amount of H_ox-CO was unaffected (Fig. 3B). Similar results were obtained for the 2 exchange variants G421H and G422H, both of which are located at the opposite end of the plug region. They reach ∼50% of complex integration when maturated with a 3-fold molar excess of [2Fe]^MIM. Such a high level of occupancy seems rather surprising, given the holoproteins’ poor H₂-evolution activities of less than 3% compared with WT.

More pronounced impairments in terms of [2Fe_H]-cofactor integration were observed for exchange variants G414H and G418H. Both positions are located in the center of the plug region and thus directly shield the [2Fe_H] subsite from solvent exposure (Fig. 1C). For G414H, no integration was observed with an equimolar ratio of apo-hydrogenase and [2Fe]^MIM (SI Appendix, Fig. S5), while a 3-fold molar excess resulted in small shifts in the spectrum (Fig. 3B). Compared with free [2Fe]^MIM complex, the vibrational signals for the CO-ligands are clearly sharpened and red-shifted by −6 to −11 cm⁻¹, but they cannot be assigned to any specific H-cluster state. Also, the vibrational signals of the CN⁻ ligands are clearly separated (2,062 and 2,050 cm⁻¹), suggesting a less symmetric coordination. A weak signal at 1,807 cm⁻¹ suggests the formation of a bridging CO (µCO) ligand in a small fraction of the sample. This would presuppose a similarly rotated ligand configuration, as known for the mature [2Fe_H] subcluster, which was already presaged by small molecule models from synthetic chemists (43–45). Additionally, the vibrational signal for the H_ox-CO marker band at 2,015 cm⁻¹ is significantly red-shifted by −6 cm⁻¹, suggesting for the [2Fe_H] subsite in H_ox-CO a slightly deviating ligand coordination. The amino acid substitution of G414 to alanine on the other hand yields an almost WT-like FTIR spectrum upon maturation with a 3-fold molar excess of [2Fe]^MIM (Fig. 3B), highlighting that a conservative substitution at this position still allows for the formation of a stable H-cluster, while the sterically more demanding histidine severely hampers or destabilizes proper [2Fe_H]-cofactor placement. A maturation with an equimolar ratio of [2Fe]^MIM and apo-G414A, however, yielded only ∼30% of sample activation, whereas for WT, this ratio was sufficient to reach a maturation level of 100% (SI Appendix, Fig. S5). At this position, even a small side chain like CH₃ limits the success of stable cofactor integration, rendering the necessity to use an excess of [2Fe]^MIM for complete sample maturation.

When maturated with a 3-fold molar excess of [2Fe]^MIM, exchange variant G418H features only a very broad CO/CN⁻ ligand signal pattern, which corresponds to the vibrational spectrum of the free [2Fe]^MIM complex. While a slight splitting of the CN⁻ ligand signals is observed, there is no indication for a bridging CO species, demonstrating that the substitution in G418H prevents functional cofactor integration. However, the more conservative substitution of G418 to alanine permits a distinct incorporation of [2Fe]^MIM, as indicated by the formation of a µCO ligand signal at 1,808 cm⁻¹ and typical H-cluster–associated vibrational signals, although to a significantly lower extent compared with WT (∼20% H-cluster occupancy).

In a final FTIR-spectroscopic experiment, we attempted to reactivate the purified holoenzymes that largely lacked the [2Fe_H] site on the ATR crystal. Here, identical H-cluster FTIR spectra were obtained compared with the experiment in which apo-hydrogenase was maturated with [2Fe]^MIM on the ATR crystal, demonstrating that rematuration with a second [2Fe]^MIM complex is possible (SI Appendix, Fig. S6).

PFE.

In a previous study, we reported the successful application of PFE to investigate the final stages of H-cluster assembly in [FeFe]-hydrogenases (42). To investigate in more detail the role of the conserved arginine residue in [2Fe_H]-cofactor transfer, CpI WT and variant R449H were activated at pH 6 and 8 in a chronoamperometry experiment. With this approach, we aimed to probe the effect of switching, on and off, the positive charge of the substitute histidyl residue in R449H, similar to the corresponding ATR-FTIR experiment but now focusing on the kinetics of the development of catalytic activity. For measurements at pH 6, we monitored the proton reduction current, while at pH 8, the H₂ oxidation current was monitored. Whereas the apo-protein of CpI WT can be activated at both pH 6 (−0.434 V vs. standard hydrogen electrode [SHE]) and pH 8 (+0.075 V vs. SHE), activation of apo-R449H can only be observed at pH 6 and to a significantly lower extent compared with WT. At pH 8, little to no activation is detectable, thus verifying the results obtained from FTIR spectroscopy (Fig. 4 A and B). Even when employing a much higher concentration of [2Fe]^MIM, the activation level at pH 8 could not be enhanced. Surprisingly, the R → H substitution only influenced the total amount of integrated [2Fe]^MIM but not the rate. The degree of activation declines with increasing pH until at pH 8, no activation is measurable anymore, while the half-lives for activation do not change significantly (SI Appendix, Fig. S7). This result suggests that the proportion of deprotonated and protonated histidine residues influences the extent of activation but that the 2 states are poorly interchangeable since we do not observe a trend in half-life, whereby similar levels of activation are attained more slowly at higher pH values.

Fig. 4.

In vitro activation of CpI WT (black) and protein variants (red/orange) observed by chronoamperometry. (A) Activation of CpI WT and variant R499H at a constant potential of −0.434 V vs. SHE at pH 6. (B) As for A at 0.075 V vs. SHE and pH 8. For variants G412H (C), G421H (E), G414H (G), and G418H (H), the potential was initially held at −0.75 V vs. SHE for 1,000 s, during which the [2Fe]^MIM cofactor was injected at 500 s (final concentration, 0.6 µM). The potential was then stepped to −0.05 V for 5 separate intervals and returned to the starting potential at −0.75 V to monitor H₂-evolution activity. (D) An additional experiment was carried out for G412H, in which the potential was held at −0.75 V vs. SHE for 1,000 s, during which the [2Fe]^MIM cofactor was injected at 500 s. The potential was then stepped to −0.1 V to monitor H₂-oxidation activity (100% H₂). (F) As for C and E with a 10-fold buffer exchange after 1,000 s and potential jumps after 1,500 s. All experiments were under 100% Ar atmosphere (except for B and D), 10 °C, and 1,000-rpm electrode rotation.

As the Gly → His exchanges were expected to influence the final structural rearrangement of plug, lid, and lock, chronoamperometry experiments were performed to investigate the stability of [2Fe_H]-subcluster integration. Previous results demonstrated that under reducing conditions (less than −0.54 V vs. SHE), [2Fe]^MIM forms a stable precursor complex with the immobilized apo-hydrogenase on a pyrolytic graphite “edge” (PGE) electrode (42). Stepping to a more positive potential is required to release the [4Fe-4S] cluster from its reduced (and therefore electron-donating) state, which prevents formation of the [2Fe-μSCys-4Fe] bond. Final formation and reorientation of the H-cluster, presumably leading to rearrangement of the plug, lock, and lid and closure of the secondary coordination shell can now occur. The ability to stall the final stages of maturation simply by applying a very negative potential is a useful diagnostic feature. Before and after the injection of [2Fe]^MIM (final concentration, 0.6 µM), apo-hydrogenase was kept for 500 s at −0.75 V vs. SHE. After 1,000 s, the potential was periodically switched for 170 s to −0.05 V (activation potential) and then back to the lower starting potential (to assess the catalytic current) for 5 consecutive intervals, thus monitoring the development of H₂-evolution activity in the course of H-cluster maturation. We selected the “potential stepping” method in favor of a continuous measurement at a fixed, more positive potential to obtain a catalytic current that is only attributable to a fixed number of enzyme molecules preoccupied with [2Fe]^MIM. While monitoring the catalytic current at −0.75 V vs. SHE, no ongoing incorporation of [2Fe]^MIM will interfere with the current signal, which consequently can exclusively be attributed to the already present preformed complex of apo-hydrogenase and [2Fe]^MIM. This approach allowed us selectively to probe the ability of each variant to retain the [2Fe_H] site. For CpI WT, H₂-production current is detected after the first potential jump and increases up to the third interval (Fig. 4). A linear decline in catalytic current (∼10%) is observed during each measurement interval due to protein loss, which reflects the expected level of stability of CpI films on PGE. Catalytic current was also detected for variant G412H and was significantly lower than that measured for WT (Fig. 4C); it attained maximum hydrogen-production current earlier than the WT enzyme, at the second interval. Most strikingly, the more severe loss of ∼40% of catalytic current (compared with only ∼10% for the WT enzyme) during each measurement step does not correlate with the expected film loss alone but instead can rather be referred to an instability in [2Fe_H]-cofactor binding or retention. To verify this idea, we looked more closely at the activation kinetics of G412H by monitoring the H₂-oxidation current at pH 8 (Fig. 4D) which does not require periodic switches of the potential (as activation and assessment of catalytic current occur at the same voltage). While the activation of G412H initially proceeds even faster than for WT, a maximum catalytic current is attained after 330 s before the activity steadily drops by ∼50% over the next 3,000 s. In the case of WT CpI, a continuous activation process is observed without any loss in H₂ oxidation current. In contrast, the exchange variant G421H exhibits a WT-like activation in terms of H-cluster stability but at significantly lower current levels (about 20%) (Fig. 4E). To assess the stability of the precursor complex that is formed between apo-hydrogenase and [2Fe]^MIM, a 10-fold buffer exchange was performed for these 2 variants as described earlier (13, 42). The [2Fe]^MIM complex was injected at 500 s, while the buffer solution was exchanged after 1,000 s. At 1,500 s, the potential shifts were executed according to the potential-jump experiment. CpI WT still activates well without any additional [2Fe]^MIM in solution. Likewise, variants G412H and G421H (Fig. 4F) still feature catalytic currents (albeit lower), confirming that both exchanges to a certain extent still allow the entry, orientation, and retention of the [2Fe_H] cofactor.

In agreement with the maturation experiments of apo-hydrogenase and [2Fe]^MIM performed on the ATR crystal, the exchange variants G414H and G418H exhibit the strongest impairment of H₂-production activity and/or [2Fe_H]-cofactor stability in the potential-jump chronoamperometry experiment (Fig. 4 G and H). The very low catalytic H₂-production current that is attained with variant G414H returns to zero current level at the end of each activity-assessment period (170 s), indicating a very low H-cluster stability. On the other hand, exchange variant G418H showed no catalytic current at all.

To provide a better overview, a table summarizing the results of all of the experiments of each variant probed in this study is displayed in SI Appendix, Table S2.

Discussion

The 2 types of experiments must be distinguished from each other to allow for a sophisticated analysis of each amino acid substitution: In experiment 1, the purified holo forms of the R449H and all G → H variants feature significantly decreased H₂-evolution activities, which correlate with the low H-cluster occupancies as demonstrated by FTIR spectroscopy (Fig. 2). However, in this experiment, maturation was followed by size-exclusion chromatography, which might promote [2Fe_H]-cofactor loss. In experiment 2, we performed the maturation directly on the ATR crystal with [2Fe]^MIM and apo-hydrogenase dried as a stable protein film. In contrast to the low occupancies observed in experiment 1, H-cluster integration can be observed for most variants in experiment 2. The different outcome of both experiments suggests varying degrees of postintegration instability for the respective variants. The final FTIR-spectroscopic experiment, which demonstrated successful reactivation of purified holoenzymes on the ATR crystal (SI Appendix, Fig. S6), carried out analogous to experiment 2, proves that subcluster coupling is reversible and loss of the [2Fe_H] cluster does not result in a permanently inactive species, therefore indicating that the [4Fe_H] site remains intact and accessible.

A Conserved Arginine Residue Controls the Introduction of the [2Fe]^MIM Complex into the Maturation Channel.

Essentially, position R449 can be considered as a first recognition site, responsible for initiating and maintaining control over the passage of the [2Fe]^MIM complex along the cationic maturation channel. Our results obtained with the R → A and R → H exchange variants at position 449 clearly demonstrate the requirement for a positive charge located at the channel entrance for efficient cofactor uptake. However, the H₂-evolution activity of R449H is drastically decreased compared with CpI WT, demonstrating that the unique properties of the “gatekeeper” arginine go beyond the mere electrostatic recognition of the [2Fe_H] precursor. Importantly, the R → H substitution only influenced the total amount of integrated [2Fe]^MIM, not the rate at which maximum current was attained, emphasizing the important characteristics of arginine that might be essential for [2Fe_H]-cofactor recognition. In maturated CpI, R449 is translocated into the center of a cluster of 4 basic (R429, K446, R449, R540) and 4 acidic (E425, E444, D534, E535) residues, which form salt-bridge contacts between the lid and lock structural elements, thus presumably stabilizing the position of the plug in the closed conformation (SI Appendix, Fig. S8) (29). The smaller and less basic histidine residue may not be able to achieve this second function of R449 as a structure-stabilizing element. For nitrogenases, an accumulation of basic residues (involving 2 arginine and 4 histidine residues) was shown to be essential for the recognition and funneling of the negatively charged FeMo cofactor to its binding site (46). Additionally, as for the [FeFe]-hydrogenases, maturation was shown to induce subsequent conformational changes within the protein, thus stabilizing the integrated cofactor (47). The strong similarities between these 2 proteins are a hint of how metalloenzymes may have evolved to recognize and funnel a negatively charged cofactor into a tightly arranged cationic pocket; this feature is further addressed in SI Appendix, SI Discussion and Fig. S9. We could further show that this amino acid has the same function in vivo, because enzyme activation of R449H by an E. coli lysate containing the maturases HydE, HydF, and HydG was hampered to the same degree as in the maturation experiments with [2Fe]^MIM. We can therefore assume that R449 plays a similar decisive role in recognizing and introducing the [2Fe_H] precursor into the maturation channel after its release during the HydF/apo-hydrogenase interaction (Fig. 1A and SI Appendix, Fig. S7). Additionally, electrostatic contact to the surface exposed R449 may even be a precondition for the detachment of the [2Fe_H] precursor from the [4Fe4S] cluster of HydF in the HydF/apo-hydrogenase complex.

Residue K358 is located at the very end of the maturation channel close to the [4Fe_H] cluster (Fig. 1 and SI Appendix, Fig. S7), where it is directly involved in stabilizing the final ligand configuration of the [2Fe_H] cofactor in the catalytically functional H-cluster by forming an H-bond contact to CN_d. Variant K358N shows no H₂ production, which is well in line with earlier examinations on this variant that showed a complete lack of [2Fe_H]-cofactor content upon in vivo maturation (34). CO-release tests and ATR-FTIR spectroscopy confirm that no subcluster bond is formed and that apo-K358N does not interact with the [2Fe]^MIM complex at all. Even more essential than R449, the positive charge at this position is absolutely required as a final recognition point to incorporate the [2Fe_H] precursor into the H-cluster binding pocket. The ligand configuration is likely to readjust in a way that is vital for the success of the following steps (closure of the binding site and subcluster coupling under CO release).

H-Cluster Stability Is Ensured by Conserved Glycine Residues within the Plug Region That Are Required to Close a Secondary Coordination Sphere.

The current hypothesis is that successful [2Fe_H]-cofactor integration and its subsequent stabilization are ensured by closing the maturation channel via the structural rearrangement of plug, lid, and lock. Within the plug region, 5 highly conserved glycine residues were identified that, due to the absence of bulky sidechains, convey a high degree of rotational freedom within this peptide stretch and may thus function as hinge positions in its structural reconfiguration. The “closed” conformation completes the strongly conserved secondary coordination sphere of the [2Fe_H] moiety by a precise and stable placement of the plug over the H-cluster binding site, thus preventing an exposure of the active site to bulk solvent (33).

We individually substituted these 5 glycines of the plug region by the larger amino acid histidine to investigate if the larger side chain would indeed impede the closure of the channel. All G → H exchange variants showed severe impairments in [2Fe_H]-cofactor integration, in postintegration stability, and, as a consequence, in catalytic activity (<3% remaining H₂-evolution activity). However, the degree of impairment strongly depends on the proximity of the targeted amino acid to the [2Fe_H]-binding pocket. Despite the low activities of holoenzyme, for most variants, CO-release tests and the activation of apo-hydrogenase with the [2Fe]^MIM complex on the ATR crystal showed significant levels of [2Fe_H]-cofactor integration (∼5 to 80%). In this context, it should be mentioned that for ATR-FTIR measurements, the mixture of [2Fe]^MIM and apo-hydrogenase was immediately dried under N₂ until a stable protein film of holo-hydrogenase was obtained: a factor that might lead to a more rigid state that is generally less prone to cofactor loss compared with protein in solution. Previous work on the effects of lyophilizing [FeFe]-hydrogenases demonstrated that the freeze-drying process conserves the characteristic H-cluster features and furthermore significantly increases its stability, even in the presence of oxygen (48). Thus, the difference between residual activity and extent of cofactor integration determined via IR spectroscopy is likely due to a secondary loss of already-integrated cofactor, promoted by the substitution-dependent instability of the closed configuration.

The surface-exposed residues G412 (located at the beginning of the plug region) along with G421 and G422 (both located at the terminus of this region) feature similar effects in terms of postintegration H-cluster instability, which appears to contribute significantly to the diminished enzyme activity. Postintegration stability is conferred by the relocation of the lock and lid peptide stretches over the successfully positioned plug element (SI Appendix, Fig. S10 and SI Discussion). Exchange variants G414H and G418H retain an ATR-FTIR spectrum largely resembling that of free [2Fe]^MIM with slightly sharper CO and CN⁻ ligand signals, indicating merely some undefined interaction with the protein. The plug region between these 2 positions directly covers the [2Fe_H] site against bulk solvent; thus, the corresponding exchange variants show the strongest impairment. Both G414 and G418 are located in close proximity to the large aromatic sidechain of residue F417, which closes a large gap in the immediate ligand environment of the [2Fe_H] site (Fig. 1B) and, moreover, has been speculated to provide a tight environment which confers just enough steric flexibility to tolerate ligand reconfigurations that transiently protect the H-cluster from oxidative destruction (49, 50). The absence of side chains at positions G414 and G418 seem to provide the required space for the precise integration of F417 into the secondary ligand sphere, thereby creating a tight binding pocket that forces the [2Fe_H] subcluster to adopt the rotated ligand configuration of the catalytically competent H-cluster (SI Appendix, Fig. S10). This is in agreement with previous studies on [FeFe]-hydrogenase active-site models, in which the presence of such a secondary coordination sphere has been postulated to be a prerequisite for the rotated ligand configuration in native enzymes (51).

The fact that a decreased stabilization in the binding pocket of the [2Fe_H] cluster rapidly leads to a detachment of the already coupled [2Fe_H] complex further suggests that the bridging cysteine thiolate bond between the [4Fe_H] and [2Fe_H] subsite must be fairly weak (14). This is in line with previous results by Rüdiger and coworkers who could demonstrate that the stability of the intercluster bond is potential-dependent and that the H-cluster can disassemble reversibly (52).

Conclusions

Our results demonstrate that the bond between the [4Fe_H] and [2Fe_H] subsite is prone to disassembly if not tightly encapsulated by a closed secondary coordination sphere that predefines its catalytically competent structural state. The conserved structural motifs investigated in this work provide deep mechanistic insights into the mechanism underlying [2Fe_H]-cofactor recognition and transfer, yielding a blueprint of how metalloenzymes might have evolved to synchronize such complex molecular processes. Based on our results, the maturation process within apo-hydrogenase may occur by the following sequence of events: 1) introduction of the [2Fe]^MIM complex into the maturation channel via a first electrostatic contact to R449; 2) transfer through the channel and subsequent anchoring and precoordination of the [2Fe_H] cofactor in its binding pocket via the electrostatic contact to K358 and subsequent H-bond formation to CN_d; 3) plug placement promoted by flexible glycine residues and closure of the secondary coordination sphere by integrating F417. 4) The [2Fe_H] cofactor, now held tightly in place, adopts its rotated ligand coordination state, ultimately leading to subcluster coupling under CO-release. 5) Lock and lid relocate to close and stabilize the plug element in which R449 is introduced into a cluster of multiple salt bridges to finish the maturation process.

Experimental Section

Plasmid Preparation, Enzyme Expression, and Purification.

Site-directed mutagenesis was carried out as described previously, using the QuikChange protocol (Stratagene) (10). Recombinantly expressed apo-proteins of CpI were isolated as described previously, following the method reported by Kuchenreuther (36). All experiments were carried out in an anaerobic tent (Coy Labs) under a N₂/H₂ atmosphere (99:1). Solutions and buffers used for anaerobic cultivation were supplemented with 2 mM sodium dithionite (NaDT). The recombinant strep-tagged proteins were affinity-purified with Strep-Tactin Superflow high-capacity cartridges (IBA Lifesciences) according to the manufacturer’s instructions. Sodium dodecyl sulfate polyacrylamide gel electrophoresis was used to monitor the purity of the hydrogenase isolates, and the protein concentration was determined via Bradford assay (Bio-Rad) using bovine serum albumin as a standard (Biolabs). Protein samples were concentrated to 0.5 to 1 mM in 0.1 M tris(hydroxymethyl)aminomethane HCl (Tris/HCl), pH 8.0, and stored at −80 °C until further use.

Iron Quantification.

The iron content of each apo-hydrogenase sample was quantified in triplicate by applying the method of Fish (35).

Preparation of [2Fe_H] Mimics and In Vitro Maturation of [FeFe]-Hydrogenases.

The [2Fe_H]^MIM complex was synthesized as described previously (53). Before usage, the crystalline complex was dissolved in 100 mM potassium phosphate buffer (KpI), pH 6.8. Recombinant apo-proteins of CpI were activated with the [2Fe]^MIM complex as described earlier with some minor modifications (10, 32). Briefly, apo-protein and a 10-fold molar excess of [2Fe]^MIM were incubated for 1 h at 4 °C. Excess [2Fe]^MIM was afterward removed by size-exclusion chromatography (GE Healthcare) and the protein subsequently concentrated to 0.5 to 1 mM in 100 mM Tris/HCl, pH 8.0, and 2 mM NaDT.

Enzyme Activity Assays.

Hydrogen-production activity assays were performed as described previously (54). A 400-ng sample of holo-hydrogenase was incubated in 2 mL of reaction mixture, containing KpI buffer (0.1 M potassium phosphate, pH 6.8), NaDT (0.1 M), and methyl viologen (10 mM) using sealed 8-mL vessels (Suba). For apo-hydrogenase, the same reaction mixture was used, which was further supplemented with 1% Triton X-100 and 300 µL of E. coli lysate containing heterologously coexpressed HydE, HydF, and HydG from S. oneidensis. The samples were purged with argon for 5 min and incubated for 20 min at 37 °C (100 rpm) in a shaking water bath. Hydrogen production was monitored by analyzing the composition of a 400-µL sample of head space via gas chromatography (Shimadzu).

CO-Release Assays.

CO-release assays were used as an indirect means to determine the kinetics of the in vitro maturation process (32). Conversion of deoxyhemoglobin to hemoglobin-CO (Hb-CO) was monitored at 419 nm, using a UV-visible (UV-VIS) spectrometer (UV-2450; Shimadzu). Initially, 20 mg of hemoglobin (Merck) was completely reduced to deoxyhemoglobin in a 1 M NaDT solution. Reduced deoxyhemoglobin was separated from surplus NaDT, using a gel-filtration column (NAP-10)—the red fraction containing deoxyhemoglobin was eluted with KpI buffer. To monitor CO release during in vitro maturation, KpI buffer (0.1 M potassium phosphate, pH 6.8, 20 mM NaDT) was supplemented under anaerobic conditions with deoxyhemoglobin in a gas-tight UV-VIS cuvette, before adding apo-hydrogenase and [2Fe]^MIM. Release of CO, corresponding to the rate of cofactor integration, was documented by following the change in absorbance at 419 nm.

ATR-FTIR Spectroscopy.

ATR-FTIR spectroscopy was performed, using a Bruker Tensor27 spectrometer equipped with a BioATR cell II (Harrick) harboring a double-reflection ZnSe/Si crystal. The spectrometer was kept in an anaerobic N₂/H₂ atmosphere (99:1) (anaerobic chamber; Coy Labs). All measurements were performed at 30 °C, using a varying number of scans (depending on the experiment) and a resolution of 2 cm⁻¹. For the measurements of holo-hydrogenase, a sample of 15 µL was applied on the ATR crystal. The ATR cell was then closed with a custom-designed gas-tight lid and purged with N₂ (20 L/min) to obtain a semidried protein film after 10 min. To follow the in vitro maturation process of apo-hydrogenase with [2Fe]^MIM via FTIR spectroscopy, both components were mixed in a 200-µL reaction tube (7.5 µL each) and immediately transferred to the ATR crystal; a semidried protein film was obtained by N₂ gassing. For this experiment, each protein variant was measured as a biological replicate.

PFE.

All PFE experiments were carried out in an N₂ atmosphere (O₂ < 2 ppm) within an anaerobic glovebox (MBraun 150B-G) as described previously (42). An Autolab potentiostat (PGSTAT128N) was used, employing Nova software (Eco Chemie-Metrohm Autolab). The gastight electrochemical cell was water-jacketed to control the temperature. A PGE rotating-disk electrode was used as a working electrode and controlled by a rotator (Autolab). A nonisothermal side arm housing the reference electrode (saturated calomel electrode [SCE]) and containing a 0.1 M NaCl solution was connected to the main cell compartment by a luggin capillary. Platinum wire was used as a counter electrode. The reference potential was converted to the SHE scale, using the correction E_SHE = E_SCE + 0.241 V at 25 °C. For each experiment, defined gas mixtures [British Oxygen Company (BOC) gases] were adjusted, using mass flow controllers (Sierra Instruments).