Heterologous Hydrogenase Overproduction Systems for Biotechnology—An Overview

Hydrogenases are complex metalloenzymes, showing tremendous potential as H2-converting redox catalysts for application in light-driven H2 production, enzymatic fuel cells and H2-driven cofactor regeneration. They catalyze the reversible oxidation of hydrogen into protons and electrons. The apo-enzymes are not active unless they are modified by a complicated post-translational maturation process that is responsible for the assembly and incorporation of the complex metal center. The catalytic center is usually easily inactivated by oxidation, and the separation and purification of the active protein is challenging. The understanding of the catalytic mechanisms progresses slowly, since the purification of the enzymes from their native hosts is often difficult, and in some case impossible. Over the past decades, only a limited number of studies report the homologous or heterologous production of high yields of hydrogenase. In this review, we emphasize recent discoveries that have greatly improved our understanding of microbial hydrogenases. We compare various heterologous hydrogenase production systems as well as in vitro hydrogenase maturation systems and discuss their perspectives for enhanced biohydrogen production. Additionally, activities of hydrogenases isolated from either recombinant organisms or in vivo/in vitro maturation approaches were systematically compared, and future perspectives for this research area are discussed.


Introduction
The world's still increasing hunger for energy and the resulting excessive use of fossil fuels is a major reason for environmental pollution and global warming. These problems are one of the most urgent and actual questions of humankind. Therefore, the search for and utilization of so-called alternative energy sources is an important task. A promising solution is the utilization of "green" molecular hydrogen (H 2 ) as it contains a high energy potential and its combustion releases only water [1][2][3][4]. Unfortunately, the already well-established non-biological systems for the production, storage and utilization of H 2 rely on noble metal catalysts, which are resource-limited, expensive, and can easily be destroyed by, e.g., CO or H 2 S impurities in commercially available H 2 gas [5][6][7]. H 2 -converting biocatalysts, the so-called hydrogenases, however, utilize abundant metals, such as iron and nickel, for catalysis and are often resistant towards CO or H 2 S. During evolution, the harsh environments and low substrate availability lead to the development of highly specific, but robust hydrogenases with sometimes extremely high affinity towards hydrogen [8]. Based on the metal ion composition of their active center, hydrogenases can be divided into the three classes of [Fe]-, [FeFe]-and [NiFe]-hydrogenases [9]. In contrast to the FeS cluster-free [Fe]-hydrogenases, Recent studies showed that the reconstitution of active [Fe]-hydrogenase in vitro is feasible ( Figure 1). However, since chemical synthesis of the FeGP cofactor was not possible so far, the cofactor had to be extracted from native [Fe]-hydrogenases. The inactive [Fe]-hydrogenase apo-enzymes from some methanogens have been successfully produced in E. coli either as soluble proteins or in the form of inclusion bodies [36]. To obtain an active cofactor, Shima et al. [36] optimized the extraction process, yielding purified FeGP cofactor from M. marburgensis, which had the ability to restore a specific activity of approx. 700 U from 1 mg of enzyme. This activity is actually higher than that of purified [Fe]-hydrogenase (400-600 U mg −1 ) [34]. The activity of the reconstituted Hmd from M. jannaschii (300-400 U mg −1 ) is equivalent to the full activity of the Hmd purified from M. marburgensis (400-600 U mg −1 ) [34]. Remarkably, the inactive apo-Hmd from M. jannaschii was reactivated up to a specific activity of 1100 U mg −1 by the addition of the FeGP cofactor purified via ultrafiltration from active Hmd from M. marburgensis, which had been denatured in 8 M urea instead of MeOH, 2-mercaptoethanol and ammonia solution to release the cofactor [31]. However, the difference in MjaHmd activity during activation with purified FeGP is mainly attributed to the different reaction temperatures used for the activity measurements, while 1100 U mg −1 were obtained at 65 • C [31], the measurement at 40 • C resulted in 300-400 U mg −1 Hmd activity [34]. Similarly, heterologously produced Hmd from M. kandleri was activated by the purified FeGP cofactors from M. jannaschii or M. kandleri [31]. These studies clearly showed that the cofactor is released in an intact form upon enzyme denaturation in the presence of urea and thiol reagents. At least seven hmd co-occurring genes (hcgA-G) are required for FeGP cofactor synthesis [30,38,39]. So far, the functions of five of them, HcgB, HcgC, HcgD, HcgE and HcgF, in FeGP cofactor biosynthesis have been elucidated [40][41][42][43][44]. The functions of HcgA and HcgG are, however, still not known, and only a putative role in CO ligand synthesis or pyridinol assembly is proposed [42]. The seven Hcg proteins are, in contrast to the Hmd-paralogue HmdII, essential for Fe hydrogenase maturation [39]. Nevertheless, Goldman et al. suggested a function of HmdII as a scaffold for the synthesis of the FeGP cofactor [45], while Fujishiro et al. excluded a scaffold function and proposed that HmdII could act as a sensor of intracellular methylene H4MPT concentration and thus be involved in the regulation of hmd gene expression [46]. However, further experimental evidence is required in both cases.
Native hydrogenases exclusively utilize Fe and/or Ni for H 2 activation. However, other transition metals are also known to activate or catalyze the production of hydrogen in synthetic chemical systems. Consequently, a non-native metal hydrogenase has been constructed by incorporation of a Mn-complex into the apo-[Fe]-hydrogenase from M. jannaschii heterologously produced in E. coli, resulting in a [Mn]-hydrogenase with a comparable activity to semi-synthetic [Fe]-hydrogenase [37,47]. However, compared to the hydrogenase reconstituted with the native FeGP cofactor the semi-synthetic enzymes showed significant less activity. This might be attributed to the lack of the guanosine monophosphate (GMP) moiety and two methyl groups in the pyridinol group of the synthetic cofactors used for reconstitution hydrogenase by denaturation under the indicated conditions. The unfolded protein is removed by salt precipitation to obtain highly purified FeGP cofactor. The active reconstituted holoenzyme with a fully assembled active site is formed by mixing of the isolated apoenzyme heterologously produced in E. coli and the extracted FeGP cofactor. In addition to this, many Fe-complexes have been chemically synthesized to mimic the active site of [Fe]-hydrogenase. These complexes become active only when incorporated into the apo-enzyme of Fe-hydrogenases forming so-called semi-synthetic [Fe]-hydrogenases. X represents a solvent molecule that is bound at the "open" site of the crystal structure of the native enzyme. GP: guanylylpyridinol.
Native hydrogenases exclusively utilize Fe and/or Ni for H2 activation. However, other transition metals are also known to activate or catalyze the production of hydrogen in synthetic chemical systems. Consequently, a non-native metal hydrogenase has been constructed by incorporation of a Mn-complex into the apo-[Fe]-hydrogenase from M. jannaschii heterologously produced in E. coli, resulting in a [Mn]-hydrogenase with a comparable activity to semi-synthetic [Fe]-hydrogenase [37,47]. However, compared to the hydrogenase reconstituted with the native FeGP cofactor the semi-synthetic enzymes showed significant less activity. This might be attributed to the lack of the guanosine monophosphate (GMP) moiety and two methyl groups in the pyridinol group of the synthetic cofactors used for reconstitution of both semi-synthetic [Fe]-and [Mn]-hydrogenase. Nevertheless, the apo-enzyme reconstituted with the Mn(I)-complex exhibits the highest activity and broadest scope in catalytic hydrogenation among all engineered variants [37]. Compared to other synthetic Mn catalysts, the semisynthetic [Mn]-hydrogenase displays a unique activity in Mncatalyzed hydrogenation of compounds analogous to the natural substrate of [Fe]-hydrogenase, methenyl-H4MPT + [37]. These results demonstrate the potential use of biomimetic catalysts in hydrogenation reactions.
Since their discovery about 30 years ago, [Fe]-hydrogenases are still the least studied hydrogenases, but they have uncovered new structural and mechanistic surprises that often required the re-assessment of the proposed catalytic mechanism [48][49][50]. Apart from that, research on [Fe]-hydrogenases is still conducted in only a few research groups, which is in marked contrast to the situation for [FeFe]-and [NiFe]-hydrogenases. [Fe]-hydrogenase by denaturation under the indicated conditions. The unfolded protein is removed by salt precipitation to obtain highly purified FeGP cofactor. The active reconstituted holoenzyme with a fully assembled active site is formed by mixing of the isolated apoenzyme heterologously produced in E. coli and the extracted FeGP cofactor. In addition to this, many Fe-complexes have been chemically synthesized to mimic the active site of [Fe]-hydrogenase. These complexes become active only when incorporated into the apo-enzyme of Fe-hydrogenases forming so-called semi-synthetic [Fe]-hydrogenases. X represents a solvent molecule that is bound at the "open" site of the crystal structure of the native enzyme. GP: guanylylpyridinol.
Since their discovery about 30 years ago, [Fe]-hydrogenases are still the least studied hydrogenases, but they have uncovered new structural and mechanistic surprises that often required the re-assessment of the proposed catalytic mechanism [48][49][50]. Apart from that, research on [Fe]-hydrogenases is still conducted in only a few research groups, which is in marked contrast to the situation for [FeFe]-and [NiFe]-hydrogenases.
well as in some unicellular green algae [54,55]. Consequently, they are synthesized under strictly anoxic conditions [9]. [FeFe]-hydrogenases occur as monomeric species, but hetero-dimeric, -trimeric and -tetrameric protein are also known [56][57][58][59]. They harbor a unique catalytic site, the so-called Hcluster. Many of them also contain additional iron-sulfur clusters [56]. The H-cluster comprises a diiron [FeFe] sub-cluster, which is equipped with carbonyl and cyanide ligands and connected to a [4Fe-4S] cluster via a bridging cysteine residue ( Figure 2A) [56,60]. -hydrogenases requires the co-expression of specific maturases. First, the iron-sulfur cluster is inserted into the apo-enzyme by the host's universal ISC (iron-sulphur cluster assembly machinery) pathway producing an inactive pro-enzyme. In the second step, the [FeFe]H subcluster precursor is produced by the co-synthesized maturases HydE and HydG, while the third maturase, HydF, serves as scaffold protein, which also inserts the cofactor into Pro-HydA. This results in active [FeFe] hydrogenase. The question mark denotes unknown function(s). (B) Artificial in vitro maturation is achieved by the heterologous production of the apo-hydrogenase and the three maturases and subsequent mixing of the purified proteins. (C) Alternatively, artificial in vitro maturation can be achieved by the heterologous production of the apo-hydrogenase and subsequent mixing of the purified protein with a chemically synthesized [2Fe] subcluster. The bridging ligand X might be NH/NH2 + in the native maturation or additionally CH2, or O when using artificial maturation.
As early as in the 1980s, Voordouw and coworkers tried to heterologously express the periplasmic [FeFe]-hydrogenase from Desulfovibrio vulgaris in E. coli, resulting in an inactive -hydrogenases requires the co-expression of specific maturases. First, the iron-sulfur cluster is inserted into the apo-enzyme by the host's universal ISC (iron-sulphur cluster assembly machinery) pathway producing an inactive pro-enzyme. In the second step, the [FeFe] H sub-cluster precursor is produced by the co-synthesized maturases HydE and HydG, while the third maturase, HydF, serves as scaffold protein, which also inserts the cofactor into Pro-HydA. This results in active [FeFe] hydrogenase. The question mark denotes unknown function(s). (B) Artificial in vitro maturation is achieved by the heterologous production of the apo-hydrogenase and the three maturases and subsequent mixing of the purified proteins. (C) Alternatively, artificial in vitro maturation can be achieved by the heterologous production of the apo-hydrogenase and subsequent mixing of the purified protein with a chemically synthesized [2Fe] subcluster. The bridging ligand X might be NH/NH 2 + in the native maturation or additionally CH 2 , or O when using artificial maturation.
As early as in the 1980s, Voordouw and coworkers tried to heterologously express the periplasmic [FeFe]-hydrogenase from Desulfovibrio vulgaris in E. coli, resulting in an inactive hydrogenase lacking the catalytic H-cluster with a yield of about 5% w/w of total protein [61,62]. However, to date, the successful (over)production of several [FeFe]-hydrogenases has been achieved by using homologous as well as heterologous hosts (summarized in Table 2).

Recombinant [FeFe] Hydrogenase Production in the Presence of the Maturases HydE, F and G
Initial attempts to heterologously produce functional [FeFe]-hydrogenases had only limited success, due to the limited understanding of their maturation. The heterologous hosts were not capable of synthesizing the complete H-cluster, resulting in either catalytically inactive apo-enzymes [61,90] or in protein with only very low activity [74,91]. In 2004, Posewitz and coworkers produced Chlamydomonas reinhardtii HydA1 in E. coli and co-expressed the hydEFG genes encoding the maturases from C. reinhardtii [78]. The purified recombinant CreHydA1 displayed catalytic activity, which was, however, ca. 3-fold lower than the activity of the enzyme isolated from its native host. Despite the low expression levels of the C. reinhardtii genes, this study demonstrated impressively that the three maturases, HydF, HydE and HydG, are essential for the maturation and integration of the H-cluster to produce active [FeFe]-hydrogenase ( Figure 2A). Later, the production of CreHydA1 was improved by using Clostridium acetobutylicum as a heterologous expression host. With this strategy, functional CreHydA with a final activity of about 760 µmol H 2 /(min·mg) was obtained [76]. Moreover, King et al. co-expressed the C. acetobutylicum maturase genes hydEFG with the genes encoding the monomeric [FeFe]-hydrogenases from C. acetobutylicum, C. reinhardtii, Clostridium pasteurianum and Clostridium saccharobutylicum in E. coli, yielding stable and active HydA enzymes with 10-fold higher productivity compared to the native hosts [64,92]. The purification process has also been improved over the years.
[FeFe]-hydrogenase activity can markedly decrease during storage of the cells, storage of the purified proteins and during purification. Hence, Girbal et al. [76] introduced modifications into the hydrogenase purification protocol. In this, optimized protocol cells are flushed with pure H 2 at the end of the cultivation to prevent oxidative damage of the protein. Furthermore, cell disruption was directly performed after the cultivation with freshly harvested cells. With this optimized purification protocol, Demuez et al. [63] increased the specific activities of C. acetobutylicum [FeFe]-hydrogenase by 16-fold for H 2 production and by 130-fold for H 2 oxidation.
Based on the beneficial codon usage and its hydrogenase maturation ability, Sybirna et al. [82] found that Shewanella oneidensis is a suitable host for heterologous production of functional [FeFe]-hydrogenase from C. reinhardtii. S. oneidensis is a facultatively anaerobic, Gram-negative γ-proteobacterium that possesses a [FeFe]-hydrogenase operon [93]. Compared to the production of CreHydA1 in C. acetobutylicum [76], heterologous production in S. oneidensis resulted in a similar specific hydrogenase activity but a 5-fold higher enzyme yield [82]. By co-expression of S. oneidensis hydEFG during CreHydA1 production in E. coli, an even 75-fold higher enzyme yield was obtained without any loss of enzyme activity [70]. Co-expression of S. oneidensis hydEFG maturation proteins was also successfully demonstrated for heterologous production of functional [FeFe]-hydrogenases from C. pasteurianum (Hyd1) and Caldanaerobacter subterranus (HydABCD) in E. coli [70,85]. Additionally, expression of the S. oneidensis [FeFe]-hydrogenase operon, including the structural and maturation genes, in Anabaena sp. PCC7120 resulted in heterologous production of active SonHydA [89]. In contrast, SonHydA was only produced in an inactive form in both E. coli and Anabaena sp. PCC 7120 without the co-expression of its own maturation proteins [87], thus further demonstrating the importance of the co-production of specific maturation proteins for successful heterologous production of functional [FeFe]-hydrogenases.

In Vitro Maturation Systems for [FeFe]-Hydrogenases
An alternative to the in vivo production of active [FeFe]-hydrogenases is the in vitro maturation and activation of in vivo produced inactive apo-enzymes using independently produced maturases ( Figure 2B) or chemically synthesized di-iron cluster ( Figure 2C) [69,73,81,84,94]. An in vitro, cell-free maturation has been developed to generate fully active [FeFe]-hydrogenase. To this end, both C. saccharobutylicum pro-HydA and the maturases HydEFG from C. acetobutylicum were individually produced in E. coli, and subsequently all E. coli cell extracts were mixed in a reaction tube, yielding active CsaHydA [84]. Furthermore, HydEFG from S. oneidensis were used in vitro to activate hydrogenases HydA and HydI from C pasteurianum but with varying success. Whereas CpaHydA was produced with only 20% of its original activity, CpaHydI was obtained with a 2-fold higher specific activity compared to the production in E. coli in the presence of the S. oneidensis maturation proteins [69,72]. In contrast, apo-HydA from C. reinhardtii was in vitro activated by the addition of a chemically synthesized di-iron cluster either bound to Thermotoga maritima HydF [81] or even in the absence of any auxiliary protein [73]. In both cases, the addition of the chemically synthesized co-factor was sufficient to restore hydrogenase activity comparable to the native hydrogenases [73,81].
E. coli can be used for the recombinant production of [FeFe]-hydrogenase apoenzymes from essentially any organism in high yield and purity. The subsequent use of in vitro maturation systems allows to produce active enzymes on demand by mixing the apo-enzyme with any desired set of maturation enzymes or chemically synthesized cofactors. This approach can also be applied for rapid randomized mutant library screening with regard to desired properties, e.g., improved catalytic activity or decreased O 2 -sensitivity [71,80]. Taken together, these studies demonstrate that the in vitro production of active [FeFe]-hydrogenases is possible even in the absence of the otherwise essential specific maturases.

[FeFe]-Hydrogenase Production in Cyanobacteria and Microalgae
In most cases, [FeFe]-hydrogenases are irreversibly inactivated by trace amounts of O 2 [95][96][97]. This O 2 -sensitivity is one of the most critical drawbacks that hamper the biotechnological use of recombinant hydrogenases. Consequently, several recent studies focus on exploiting particularly more O 2 -tolerant enzymes to improve the understanding of the hydrogenase biogenesis and their catalytic mechanisms in vivo and in vitro.
Recently, cyanobacteria were successfully applied for the heterologous production of [FeFe]hydrogenases with improved O 2 -tolerance [98]. Asada et al. expressed the hydrogenase I gene from C. pasteurianum under control of a strong promoter in the cyanobacterium Synechococcus sp. PCC7924, which naturally encodes a [NiFe]-hydrogenase [74]. This resulted in a significant increase in hydrogen production compared to wild-type Synechococcus strain even without co-expression of maturation genes [74]. Nevertheless, compared to the heterologous production of the enzyme in E. coli with co-expression of the maturase genes, the specific activity of the clostridial hydrogenase produced in Synechococcus is very low [45,48]. Moreover, a C. acetobutylicum [FeFe]-hydrogenase was produced in the cyanobacterium S. elongatus sp. 7942 upon co-expression of the C. reinhardtii hydEFG genes and C. acetobutylicum ferredoxin with, however, again very low specific activity [66]. Berto et al. [83] produced the active [FeFe]-hydrogenase from C. reinhardtii in the cyanobacterium Synechocystis sp. PCC6803 without the co-expression of maturase genes and also without significant activity. Taken together, these results show that cyanobacteria are able to synthesize and correctly maturate a functional [FeFe]-hydrogenase in the presence [66] and maybe even in the absence of exogenous auxiliary maturases [83]. Their ability to enable the correct maturation of [FeFe] hydrogenases in the absence of co-expressed auxiliary maturase genes has to be considered as an interesting exception in heterologous hydrogenase production and an alternative for industrial applications.
The photobiological H 2 production by unicellular green algae has recently become particularly attractive as the raw materials required for this process, solar energy and water, are available in sufficient quantities and at low cost [99][100][101]. Even though, the expression of recombinant hydrogenases in green algae is a promising alternative to bacterial expression, this process is, however, exposed to greater challenges due to O 2 -presence and difficulties in genetic manipulation [102,103]. Recent studies have focused on the use of photosynthetic model organisms, such as C. reinhardtii for the production of [FeFe]-hydrogenases thereby coupling the photosynthetic transport chain via ferredoxin with H 2 production. Among all hydrogenases, the algal HydA shows an excellent H 2 production activity of up to 8 mmol H 2 /(min·mg) [99]. Additionally, some progress in the recombinant production of natural and engineered [FeFe]-hydrogenases has been achieved through optimizing methods of nuclear and chloroplast transformation of C. reinhardtii, e.g., by bead milling, particle bombardment or physical disruption of the cell wall [104][105][106][107].
As discussed above, to date the maturation process allowing the H-cluster assembly is still incompletely characterized. Nevertheless, although further work is needed, the controllable expression of hydrogenase genes in green algae and the coordinated development of light-driven hydrogen production and post-translational processing of hydrogenases is interesting for potential biotechnological and economical applications.
[NiFe]-hydrogenases are comprised of at least one large subunit (about 60 kDa) containing the [NiFe] active center and one small subunit (about 30 kDa) harboring three almost linearly arranged FeS clusters. The [NiFe] center is deeply buried in the large subunit and linked to the enzyme through four cysteine-derived thiolates. The nickel is coordinated by all four cysteines. Two of them serve as bridging ligands and also coordinated the Fe atom, which is further ligated by a CO and two CN − ligands [110,111]. The two metal ions as well as the diatomic ligands are inserted post-translationally into the apo-hydrogenase through a complex maturation process ( Figure 3A) [112]. At least six auxiliary proteins, HypABCDEF, are necessary for biosynthesis of the [NiFe(CN) 2 (CO)] cofactor and its integration into the apo form of the large subunit [11,112]. The maturation systems can be highly specific for their corresponding hydrogenases [9,10]. Thus, efficient heterologous production of functional [NiFe]-hydrogenases is quite challenging.
Nevertheless, several heterologous production systems for functional [NiFe]-hydrogenases have been reported (Table 3), which are discussed in the following section. and two CN − ligands [110,111]. The two metal ions as well as the diatomic ligands are inserted posttranslationally into the apo-hydrogenase through a complex maturation process ( Figure 3A) [112]. At least six auxiliary proteins, HypABCDEF, are necessary for biosynthesis of the [NiFe(CN)2(CO)] cofactor and its integration into the apo form of the large subunit [11,112]. The maturation systems can be highly specific for their corresponding hydrogenases [9,10]. Thus, efficient heterologous production of functional [NiFe]-hydrogenases is quite challenging. Nevertheless, several heterologous production systems for functional [NiFe]-hydrogenases have been reported (Table 3), which are discussed in the following section.   For comparison, homologous productions of hydrogenases are highlighted in gray. NR: not reported; NA: no activity; a = µmol H 2 -evolved min −1 mg −1 -protein by sodium dithionite-reduced methyl viologen; b = µmol H 2 -consumed min −1 mg −1 -protein for reduction of NAD + ; c = µmol H 2 -consumed min −1 mg −1 -protein for reduction of methylene blue; d = s −1 D 2 production rate; e = mU mL −1 reaction mixture; f = mg L −1 -culture; g = mg g −1 -wet mass; * indicates activity measured from crude cell extracts.

Heterologous Production of Hydrogenases in Hosts Encoding Closely Related Native Enzymes
Numerous examples of non-functional recombinant expression of hydrogenase genes have been reported. As early as 1987, the heterodimseric [NiFe]-hydrogenases from Desulfovibrio vulgaris Hildenborough has been produced heterologously in E. coli, resulting in inactive protein lacking the active site in the large subunit [61]. Later, heterologous expression of [NiFe]-hydrogenase genes from Acetomicrobium flavidum [117], Rhodococcus opacus [134] and Synechocystis sp. PCC6803 [135] was attempted, which also resulted in the production of non-functional hydrogenases.
To this end, recombinant hydrogenase gene expression either in the native hosts or closely related species was considered resulting in the (over-)production of functional hydrogenases. For example, the [NiFe]-hydrogenase from Desulfovibrio gigas was successfully produced in an active form in Desulfovibrio fructosovorans MR400. The hydrogenase subunits of both species show 64% identity and 80% similarity. However, the specific H 2 uptake activity of the recombinant D. gigas hydrogenase was only 16% of that observed in the native host [119,136]. Transfer of the entire Rhizobium leguminosarum hup/hyp gene cluster (hupSLCDEFGHIJK and hypABFCDEX) into hydrogenase mutant strains of the closely related α-proteobacteria Mesorhizobium loti, Rhizobium etli and Bradyrhizobium japonicum led to functional HupSL protein with, however, low H 2 uptake activity [137].
The soluble NAD + -reducing [NiFe]-hydrogenase (SH) from the actinomycete Rhodococcus opacus was functionally overproduced in the β-proteobacterium Ralstonia eutropha, which also hosts an NAD + -reducing hydrogenase. Despite the large phylogenetic difference of the two species, their SH proteins are closely related with an overall amino acid identity of these four subunits (HoxFUYH), ranging between 71% and 86%. The recombinantly produced R. opacus SH displayed almost 30% of the activity observed for the R. eutropha SH [132]. A similar approach was used for successful production of the cytoplasmic NADP + -dependent soluble [NiFe]-hydrogenase I (SHI) from Pyrococcus furiosus in E. coli. Here, thirteen P. furiosus genes (four structural and nine maturation genes) were co-expressed in the heterologous host, resulting in a recombinant hydrogenase with half of the specific activity of native PfuSHI [125].
After the [NiFe]-hydrogenase HynSL from the marine bacterium, Alteromonas macleodii (AmaHynSL, previously named HyaAB [116]) was successfully produced in aerobically grown, recombinant A. macleodii cells [113], Weyman et al. [114] surveyed the genetic elements required for the synthesis of active AmaHynSL and HynSL from Thiocapsa roseopersicina in aerobically grown E. coli. They showed that the accessory hyp genes from A. macleodii are sufficient to mature HynSL from T. roseopersicina, indicating that the assembly machines of the two species must be very similar. Later double-substitution of two residues in the small subunit of AmaHynS afforded a 4-fold improvement of its specific H 2 evolution activity without affecting its hydrogen uptake activity [115]. Following this, comprehensive studies confirmed that the modification of the proximal FeS cluster in HupS improved the H 2 production by reversing the electron flow within the H 2 -oxidizing hydrogenase HupSL from nitrogen-fixing cyanobacterium Nostoc punctiforme [124,138].
The examples described above suggest that a high degree of similarity improves the likelihood of successful production of [NiFe]-hydrogenases in the homologous or closely related heterologous hosts. In most of these cases, co-production of the structural hydrogenase proteins along with their specific maturases is required to achieve correct metal center assembly in the recombinant hydrogenases. As the specific hydrogenase activities of the recombinant enzymes are lower (16-45%) than those of native hydrogenases, one can speculate that an ineffective maturation of the recombinant enzymes is responsible.

Recombinant Hydrogenase Production in the Presence of Specific Accessory Proteins
The composition of hydrogenase operons is mostly conserved and exhibits a high degree of similarity; however, the high specificity of the cis-acting maturation system to the corresponding hydrogenase is a barrier that makes heterologous production more challenging [139]. Lenz et al. constructed a broad-host-range plasmid carrying the entire membrane-bound hydrogenase (MBH) operon of R. eutropha H16 encompassing 21 genes. In fact, hydrogenase activity was observed when this plasmid was transferred to the hydrogenase-free host, Pseudomonas stutzeri, indicating the presence of fully assembled and functional MBH [128]. This successful heterologous MBH production implied that an entire operon including structural genes (hoxKGZ), seven hyp genes (hypABFCDEX), MBH-specific accessory genes (hoxMLOQRTV) and a two-component regulatory system (hoxABCJ) are essential for producing MBH activity in heterologous systems [128].
Similarly, the regulatory hydrogenase (RH) of R. eutropha was produced in the cytoplasm of E. coli by coproduction of eight R. eutropha proteins, including HoxBC forming the RH heterodimer and the maturation proteins HypABFCDE [140]. However, the truncated R. eutropha HypF protein turned out to be nonfunctional in E. coli, and the E. coli HypF took over cyanide ligand synthesis for recombinant ReuRH, indicating substantial differences in HypF-mediated CN ligand biosynthesis [128,141,142]. As mentioned above, E. coli has been used for the functional production of PfuSHI (four structural and nine maturation genes hypABFCDE, hycI, slyD, frxA) [125]. A remarkable novel finding of this study was that only the four structural genes and one gene encoding ferredoxin oxidoreductase A (frxA) from P. furiosus are required for production of functional PfuSHI in E. coli. Their expression was induced under anaerobic conditions similar to the native hydrogenase-related genes of E. coli, suggesting that the E coli maturases can mediate post-translational maturation of an archaeal [NiFe]-hydrogenase.
In a metagenomics approach, Maroti et al. successfully produced a novel [NiFe]-hydrogenase from the Sargasso Sea in Thiocapsa roseopersicina by co-expressing only two accessory genes, hyaD and hupH, from A. macleodii [116]. The structural proteins showed 99% identity to the A. macleodii hydrogenase subunits HyaA and HyaB. This result emphasizes that the level of similarity of the hydrogenases appears to be more important than the similarity of the maturation apparatuses for the successful production of an active hydrogenase in a foreign host. However, the activity of the recombinant environmental hydrogenase was only 15%-20% of that of the native T. roseopersicina hydrogenase. The low homology between accessory maturation proteins may be the reason for the low activity of the recombinantly produced hydrogenases without co-expression of specific maturases in the expression system.
We anticipate that synthetic biology approaches will allow for the high-level production of recombinant [NiFe]-hydrogenase more easily. In fact, Schiffels et al. [130] have used an innovative cloning platform to achieve high-yield production and maturation of the SH from R. eutropha H16 in E. coli. Using this cloning platform, each gene received its own T7 promoter and terminator. The genes were placed on two different vectors, one comprising the structural genes hoxFUYHI as well as the specific endopeptidase hoxW and the second harboring hypC1D1E1A2B2F2X, encoding the maturases. Further addition of R. eutropha hoxN1, encoding a high-affinity nickel permease, increased maturation efficiency in E. coli. Compared to previously reported SH production in the native host R. eutropha, the authors obtained a 3-fold increase in both protein yield and specific activity in whole cells. The recombinant SH was isolated from E. coli cells grown under both aerobic and anaerobic conditions with the same purity and stability. Remarkably, protein purified from anaerobically grown cells showed a 1.8-fold higher specific hydrogenase activity when compared to native SH isolated from the native host R. eutropha [130,143]. Thus, this platform based on synthetic biology provides a novel direction to produce recombinant [NiFe]-hydrogenases and facilitate directed evolution approaches to optimize the enzymes and their reaction conditions. However, current methods still have low throughput for measuring hydrogenase activity, which is a major limit for research studies and biotechnological applications. Lacasse et al. [144] developed a whole-cell colorimetric hydrogenase activity assay suitable for high-throughput applications based on the reduction of benzyl viologen with a very high specificity. This assay is a promising method for future screenings of growth conditions and factors facilitating the heterologous production of active [NiFe] hydrogenases.

In Vitro Reconstitution Systems for [NiFe]-Hydrogenases
The in vitro reconstitution of [FeFe]-hydrogenases by outfitting the apo-enzyme with a chemically synthesized metal cofactor enhanced yield and activity and also greatly facilitated the investigation of unknown [FeFe]-hydrogenases from various origins (see above). Development of a similar in vitro reconstitution strategy for [NiFe] hydrogenases is much more complicated because of the higher complexity of the active site architecture and the bipartite nature of the hydrogenase module. So far, different studies attempted to in vitro reconstitute catalytic active [NiFe]-hydrogenase by mixing only purified independently produced proteins ( Figure 3B). Although the mechanism of small subunit maturation is not well understood, the FeS clusters are often stably incorporated into the apo-protein by the host's Isc/Suf machineries during heterologous small subunit production in E. coli [11,24,112,139]. The catalytic activity of the NAD + -reducing SH (HoxFUYH) from R. eutropha could be reconstituted by mixing cell extracts containing the sub-modules of the enzyme. This strategy led to the recovery of 72% of the enzymatic activity of the native SH [131], indicating that the HoxHY and HoxFU are synthesized as stable and functional sub-modules.
A little earlier, an in vitro maturation system for the large subunit HycE of E. coli hydrogenase 3 (Hyd-3) was described that was based on mixing extracts of nickel-free HycE precursor, HypBCDEF and HycI in the presence of nickel to generate an active enzyme (12% of the wild-type) under anaerobic conditions [123]. Recently, Soboh and coworkers mixed extracts containing the precursors of Hyd-2 from E. coli in vitro with the purified HybG-HybDE that serves as assembly site and carrier of the [Fe(CN) 2 (CO)] unit of the catalytic center [122,145]. They obtained catalytically active [NiFe]-hydrogenases, which gained further activity upon addition of purified HypF and HypE [122]. Furthermore, the addition of an "activation mixture" including reductant, ATP, nickel and carbamoyl phosphate was necessary, suggesting that at least a part of the reconstitution process was mimicking the catalytic in vivo biosynthesis process.
Taken together, these findings strongly suggest that it is possible to isolate incompletely processed intermediates during the maturation process and to utilize these as a basis for the in vitro maturation of apo-[NiFe]-hydrogenase. The studies described above advanced our understanding of how these accessory proteins work together in active site assembly. Moreover, this innovative in vitro maturation system for [NiFe]-hydrogenases provides an appealing strategy to engineer hydrogenases with desired properties, e.g., improved O 2 tolerance or incorporation of alternative catalytic metal ion centers, which, in turn, offers the possibility for creating new enzymatic reactions.
At present, the investigation of the isolated large subunits of [NiFe]-hydrogenases stands in the focus of research to understand the role of the individual subunits in the H 2 activation process. In fact, it was demonstrated that the isolated large subunit of the membrane-bound hydrogenase of R. eutropha was capable in H 2 activation as it showed catalytic hydrogen/deuterium exchange activity even in the presence of O 2 [127,[146][147][148]. Such a size reduction, that is the removal of the FeS cluster containing a small subunit, greatly facilitates spectroscopic investigation of the catalytic center, as demonstrated for the recombinantly overproduced large subunit of the regulatory [NiFe]-hydrogenase of R. eutropha [149]. Notably, the cofactor-free apo-large subunits protein provide a suitable target for semiartificial reconstitution trials with chemically synthesized Fe-cofactor compounds in conjunction with separate insertion of nickel irons.
Despite of the fact that considerable progress has been made in the heterologous production of [NiFe]-hydrogenases, most recombinant enzymes have significantly lower specific activities than their native counterparts, indicating that the hydrogenase maturation efficiency within the heterologous host cells is still limiting. Therefore, the complex process of post-translational maturation of [NiFe]hydrogenases requires further exploration.

Biohydrogen Production through Heterologous Gene Expression
Hydrogen is considered a promising alternative to classical fossil fuels owing to its various merits. Despite the fact that current H 2 production mainly depends on fossil fuels [150][151][152][153], biological approaches based on photosynthetic and fermentative processes have been greatly developed to generate biohydrogen in a sustainable way. These biological approaches mainly aim at the heterologous production of hydrogenases but also the production of other hydrogen-generating enzymes, such as nitrogenases [154][155][156], modified photosystem I [157][158][159], or semisynthetic catalysts composed of a chemically synthesized metal catalysts and a recombinantly produced protein [160][161][162]. Here, the fermentative hydrogen production is generally more efficient than the photosynthetic one because of its numerous benefits: (i) independence from the availability of light in dark fermentation [163,164]; (ii) higher H 2 production rates [165,166]; (iii) use of a wide range of carbon sources (more attractively from wastes); (iv) requirement of less energy; and (v) technical much simpler and more stable process [167][168][169].
In biological H 2 production processes, photosynthetic microorganisms collect the energy of sunlight and use it to activate special hydrogen-producing enzymes such as hydrogenases or nitrogenases in biophotolysis or photofermentation. So far, a number of strategies have been afforded to use these organisms in H 2 production [170,171]. Sulfur or magnesium deprivation results in the inactivation of photosystem II and subsequently reduced O 2 evolution in order to protect the O 2 -sensitive cyanobacterial [FeFe] hydrogenases and improves hydrogen production [172,173] More recently, co-production of cyanoglobin GlbN from Nostoc commune was used to protect the heterologously produced C. acetobutylicum [FeFe]-hydrogenase HydA from oxidation when produced in Nostoc PCC7120. This resulted in an increased H 2 yield of about 20-fold under aerobic conditions [174]. Additionally, the reduction of antennas was successfully used to enlarge the utilized spectrum and the quantity of the captured light at high-light intensities, thereby improving the light efficiency [175,176]. Similarly, metabolic and genetic engineering were applied for improvement of the H 2 yield in dark fermentation [177][178][179]. Nevertheless, the sensitivity of hydrogenases to O 2 , their poor catalytic efficiency for H 2 production as well as the low H 2 yield on substrates in dark fermentation are the major obstacles to develop H 2 production systems using photosynthetic microorganisms [180][181][182][183].
Up to date, several improvements in biohydrogen production have been made to overcome the current major obstacles of slow H 2 production and low H 2 yield in dark fermentation by heterologous hydrogenase production with the aid of the native E. coli maturation machinery. At the earliest, H 2 production was enhanced by 3-times through recombinant production of C. butyricum [FeFe]-hydrogenase in an E. coli mutant lacking a native hydrogenase activity [184]. A similar improvement was accomplished by overproducing recombinant HydA from Ethanoligenes harbinenese in non-hydrogen producing E. coli BL21 [86]. Furthermore, recombinant Rhodobacter sphaeroides HupSL hydrogenase in E. coli produced 200-fold more H 2 than wild-type R. sphaeroides cells under dark anaerobic conditions [185]. Akhtar and collegues constructed a synthetic YdbK-dependent pyruvate:H 2 pathway in E. coli BL21(DE3) by co-producing six proteins, including E. coli YdbK, C. pasteurianum [4Fe4S]-ferredoxin and C. acetobutylicum HydEFGA [186,187]. The deletion of iscR and/or the addition of thiamine pyrophosphate to the medium enhanced both total YdbK activity and H 2 yield per glucose (19 µmol H 2 /(h·mg)). Additional co-production of B. subtilis α-amylase, AmyE, enabled starch-dependent H 2 production in E. coli BL21 [65,187]. In another approach, the [FeFe]-hydrogenase HydA from Enterobacter cloacae was employed in E. coli BL21. Here, recombinant HydA production increased the H 2 yield 1.4-fold compared to the wild type E. cloacae strain [188,189]. Maeda et al. produced the cyanobacterial [NiFe]-hydrogenase HoxEFUYH from Synechocystis sp. PCC 6803 together with its maturation factors HypABFCDE and HoxW in E. coli, resulting in a 41-fold higher H 2 production (10 µmol H 2 /(h·mg)) compared to E. coli producing only its native hydrogenase 3 [190]. Interestingly, single deletion of any of the seven cyanobacterial hydrogenase maturation factors HypABFCDE and HoxW affected both hydrogenase activity and H 2 production in a hydrogenase negative E. coli strain to varying degrees [133]. All seven factors show a high degree of specificity towards their particular hydrogenase target and are required for optimal hydrogenase maturation. Nevertheless, only two of them, HypA and HoxW, need to be co-expressed, since their function in activating the cyanobacterial hydrogenase cannot be taken over by their E. coli homologues [109].
In contrast to cyanobacteria, E. coli cannot utilize light energy directly. Nevertheless, several attempts have been made to make light energy available for E. coli to be used for the production of hydrogen. On the one hand, it has been shown that protons generated by rhodopsin can migrate along the cytoplasmic membrane [191]. These protons could serve as a substrate for H 2 production. In fact, Kim et al. introduced genes for the synthesis of proteorhodopsin and retinal in addition to a [NiFe]-hydrogenase encoding gene from H. marinus into E. coli BL21(DE3) for light-driven biohydrogen production. The presence of proteorhodopsin and retinal increased the hydrogen production~1.3-fold (4.25 µmol H 2 /(h·mg)) compared to the hydrogenase only strain [192]. On the other hand, the use of bioinorganic hybrid systems, comprised of a semiconductor and hydrogenase-producing bacterial cells, can be used [193]. This strategy was successfully applied to engineer E. coli cells that synthesizes a metal ion complex-binding protein on their surface that collects the light energy in addition to a hydrogenase that uses the solar energy for hydrogen production [15]. Moreover, the additional encapsulation of hydrogen-producing bacteria within a biomimetic silica matrix allowed us to use O 2 -sensitive hydrogenases, even under aerobic conditions [15]. However, compared to the activity of purified hydrogenases, the hydrogen yield that can be obtained from these bioinorganic hybrid systems is very low (0.5 µmol H 2 /10 8 cells within 36 h) [15], and it requires the use of hazardous heavy metal ion complexes.
Another example is the membrane-bound [NiFe]-hydrogenase, HupSL, from the photoautotrophic bacterium Rhodopseudomonas palustris that plays a key role in the oxidation of H 2 produced as a side product in the nitrogenase reaction [194] in R. palustris. Zhou et al. [195] used this hydrogenase for the construction of an engineered E. coli BL21(DE3) strain with remarkably enhanced H 2 production activity (2.23 µmol H 2 /(h·mg)).
However, the research on H 2 production based on microbial biotechnology is currently still in its infancy. Both the maturity of the technology and the production scale are far away from meeting the requirements of commercial production. Therefore, significant improvements in H 2 production rates as well as yields in engineered E. coli strains are required.

Conclusions
Over the past few years, enormous progress has been made in our understanding of hydrogenase structure, function and their applications in biohydrogen production in vivo or in vitro mediated by recombinantly produced enzymes. However, despite these progresses, still the state of the production of recombinant hydrogenases is not satisfactory with regard to potential applications. Most recombinant hydrogenases have not yet reached the catalytic properties of natural enzymes and exhibit O 2 -sensitivity of their active site which needs to be protected. For the potential industrial applications of these enzymes, more efficient production systems need to be developed that produce recombinant hydrogenases with a larger quantity, higher specific activity and improved O 2 tolerance. Nevertheless, work on [NiFe]-hydrogenases in particular has shown that many microbes have developed a natural solution to confer these enzymes' O 2 tolerance and catalytic bias. For example, a variety of production systems for R. eutropha H16 [NiFe]-hydrogenases have been succeeded in different organisms [196]. However, our ability to produce these O 2 -tolerant hydrogenases in heterologous hosts is still challenging, and this limitation extremely hinders the viable potentials in hydrogen production and fuel cells. An interesting attempt to circumvent the O 2 sensitivity is to use specifically designed low-potential viologen-modified redox polymers, thus providing a protection matrix against oxidative damage and high-potential deactivation of O 2 -sensitive hydrogenases. Recent developments in the viologen-based polymer matrix offered the possibility of using O 2 -sensitive enzymes for H 2 oxidation in an oxidative environment [197][198][199][200]. Hence, highly O 2 -sensitive catalysts can now be considered for applications under very harsh oxidative conditions in H 2 /O 2 mixed feeding biofuel cells and other energy-converting devices.
In addition, the development of artificial in vitro maturation systems for hydrogenases in a great variety of different scaffolds could help to improve our fundamental understanding of hydrogenases, including the potential limiting factors, spectroscopic studies of the active center and the effects of O 2 on the enzyme activity. The successful in vitro maturation system can provide catalytically active hydrogenase subunits used for protein-protein docking studies with enzymes demanding low-potential electrons such as formate dehydrogenase or CO dehydrogenase [127,201]. Indeed, this strategy facilitates their potential application in industrially useful H 2 -conversion catalysts. Moreover, another important direction of recombinant research is to promote the production of minimized artificial hydrogenase proteins, thus providing the original enzyme model for the development of hydrogenase chemical mimicry catalysts.
In order to facilitate their biotechnological application, the recombinant hydrogenase should ideally not only have high oxygen tolerance and long-term catalytic stability, but also possess enzymatic characteristics that are absent in natural hydrogenases, such as the ability to use cheap electron donors. Moreover, the biotechnological production process of these enzymes should be flexible with regard to the heterologous expression strain and the cultivation conditions without affecting product quality and quantity, which was not yet achieved due to the complicated maturation mechanisms.
We have discussed the efforts contributed to the in vivo and in vitro maturation in this short perspective. Undoubtedly, we hope the combined efforts of molecular biology, bioinformatics and synthetic biology may practically help to improve our understanding of the different catalytic mechanisms of hydrogenases and the molecular details of oxygen tolerance and maturation, as well as to find solutions to the imminent energy exhaustion by the development of a hydrogen economy in the future.