Engineering Synechocystis PCC6803 for Hydrogen Production: Influence on the Tolerance to Oxidative and Sugar Stresses

In the prospect of engineering cyanobacteria for the biological photoproduction of hydrogen, we have studied the hydrogen production machine in the model unicellular strain Synechocystis PCC6803 through gene deletion, and overexpression (constitutive or controlled by the growth temperature). We demonstrate that the hydrogenase-encoding hoxEFUYH operon is dispensable to standard photoautotrophic growth in absence of stress, and it operates in cell defense against oxidative (H2O2) and sugar (glucose and glycerol) stresses. Furthermore, we showed that the simultaneous over-production of the proteins HoxEFUYH and HypABCDE (assembly of hydrogenase), combined to an increase in nickel availability, led to an approximately 20-fold increase in the level of active hydrogenase. These novel results and mutants have major implications for those interested in hydrogenase, hydrogen production and redox metabolism, and their connections with environmental conditions.


Introduction
Cyanobacteria are the only known prokaryotes capable of oxygenic photosynthesis, which uses nature's most abundant resources, solar energy, water, CO 2 and mineral nutrients, to produce a large part of the oxygen and organic assimilates for the aerobic food chain. Furthermore, cyanobacteria are regarded as promising ''low-cost'' microbial cell factories for carbon capture and storage, and the sustainable production of biofuels, thanks to their simple nutritional requirements; their physiological robustness (in colonizing a wealth of biotopes they will enable future industrial productions to be performed near the sites of use, to reduce transportation costs); and the powerful genetics of some model strains. One of the fuels of special interest is hydrogen (H 2 ) because it is a high-energy fuel [1] that burns cleanly in producing only water as its by-product. Thus, it is important to investigate the cyanobacterial machine for the production of hydrogen. The complex cyanobacterial machine for hydrogen production ( Figure  S1) is best studied in the widely used unicellular strain Synechocystis PCC6803 [2] (hereafter designated as Synechocystis), which harbours a small genome (less than 4 Mb; See CyanoBase: http://genome.kazusa.or.jp/cyanobase/) easily manipulable [3][4][5][6]. The pentameric hydrogenase enzyme (HoxEFUYH; Hox for hydrogen oxidation), which is reversibly inactivated by oxygen [7], is a bidirectional enzyme with a bias to H 2 production [8]. This reaction, 2H + + 2e 2 « H 2 , uses NAD(P)H as the source of electrons originating from photosynthesis and/or sugar catabolism, and a nickel-iron center and several iron-sulfur clusters as redox cofactors [9]. The five Synechocystis genes hoxEFUYH are clustered in an octacistronic operon that comprises the following genes hoxE, hoxF, sll1222, hoxU, hoxY, ssl2420, sll1225 and hoxH in that order, which also encode the three proteins of unknown function Sll1222, Ssl2420 and Sll1225 [2,9]. The hoxEFUYH operon is expressed by a weak promoter [10], which generates a polycistronic transcript that initiates 168 bp upstream of the start codon of the proximal hoxE gene [11,12]. The HoxEFU subunits make up the diaphorase sub-complex that transfers the electrons provided by NAD(P)H [7] to the hydrogenase unit HoxHY that reduces the protons to generate H 2 ( Figure S1). After HoxYH assembly, the HoxH subunit is processed by the HoxW protease [2]. The Hox complex is assembled by the six-subunits HypABCDEF complex enzyme [2,9] encoded by the hypABCDEF genes, which are scattered onto the chromosome ( Figure S1). Physiological studies indicated that the hydrogenase enzyme acts as an emergency electron valve to release excess of photosynthetic electrons, for instance during the transition from (anaerobic) dark to light conditions, leading to weak and transient H 2 production [7,9].
To attempt increasing hydrogenase activity in Synechocystis Germer and co-workers have used the light-inducible promoter of the photosynthetic gene psbAII to increase the expression of the endogenous hoxEFUYH operon and the heterologous Nos hypABC-DEF operon from Nostoc PCC7120 [13]. The gain in activity was modest (3.2 fold; i.e. from 2.9 nmol H 2 .min -1 .mg chlorophyll 21 in wild-type cells up to 9.4 nmol H 2 .min 21 .mg chlorophyll 21 in mutant cells) for the following possible reasons. First, the lightinducible psbAII promoter used to increase the expression of the hoxEFUYH and the Nos hypABCDEF genes is more active under high light, which increases the photosynthetic production of O 2 , which inhibits hydrogenase activity. Second, the C-terminus of the HoxF subunit was fused to the strep tag, which might have interfererred with HoxF activity. Third, the Km r marker gene that was introduced downstream of the strep-tagged hoxF gene might have decreased the expression of the downstream genes hoxUYH, at least in increasing the spacing distance between them and the psbAII promoter. Fourth, it is also possible that the Nostoc HypABCDEF proteins are not fully active on the Synechocystis PCC6803 HoxEFUYH proteins, or could somehow interfere with hydrogen production (for instance interfere with the function of the endogenous Synechocystis HypABCDEF proteins). Fifth, the nos hy-pABCDEF operon contains a gene, asr0697 (located between hypD and hypE [14]) encoding a protein (a putative 4-oxalocrotonate tautomerase), which is not normally present in Synechocystis and could interfere with hydrogen metabolism. For the same objective of increasing the abundance of active hydrogenase we have characterized and deleted the Synechocystis AbrB2 repressor of the hoxEFUYH operon [10,15] with a moderate success (two fold increase in the level of active hydrogenase as compared to the wild-type strain).
In spite of these studies, the role of the hydrogen metabolism remains puzzling. A better understanding of this role is required to identify suitable environmental conditions and powerful genetic manipulations to improve the level of hydrogen production. In this prospect, we have studied the hydrogen production machine in the model unicellular strain Synechocystis PCC6803 through gene deletion, and overexpression (constitutive or controlled by the growth temperature). We report that the hydrogenase-encoding hoxEFUYH operon is dispensable to standard photoautotrophic growth in absence of stress, and it operates in cell defense against oxidative (H 2 O 2 ) and sugar (glucose and glycerol) stresses. Furthermore, we show that the simultaneous over-production of the endogenous proteins HoxEFUYH and HypABCDEF, combined to an increase in nickel availability led to an approximately 20-fold increase in active hydrogenase level. These novel results and mutants have major implications for the engineering of effective cyanobacterial factories for the biological production of hydrogen from solar energy.

Bacterial strains and culture conditions
Synechocystis PCC6803 was grown at 30uC or 39uC (depending on the experiments) under continuous white light (standard intensity is 2,500 lux; 31.25 mE m 22 s 21 ) on mineral medium (MM), i.e. BG11 enriched with 3.78 mM Na 2 CO 3 [16]. Some cultures were grown in MM supplemented with 17 mM Fe (provided as green ferric ammonium citrate) and 2.5 mM NiSO 4 , for hydrogenase assay, or with 10 mM glucose or 300 mM glycerol, as indicated. For survival analyses, exponentially growing cells were harvested at the density of 2.5610 7 cells.ml 21 , washed and resuspended in MM at the same concentration. Then, 1 ml aliquotes were incubated under anaerobic conditions for 30 min under dim light and for 1 h in darkness with H 2 O 2 at the indicated concentrations. Cells were washed and resuspended in an equal volume of MM, diluted, spread on MM solidified with 1% agar (difco) and subsequently incubated at 30uC under standard light. Surviving colonies were counted after 5-7 days of growth.
Construction of pFC1K, the Kmr replicative plasmid vector for temperature-controlled gene expression in Synechocystis pFC1K (Table S1) was constructed ( Figure S2) using the vector pFC1, which possesses the lcI 857 -lp R system for temperatureregulated gene expression and replicates autonomously in E. coli and cyanobacteria [5]. In Synechocystis pFC1 replicates at the same 10-20 copies number per cell as the chromosome [5]. The lcI 857 temperature sensitive repressor-encoding gene tightly controls the activity of the otherwise strong lp R promoter that is followed by the lcro ribosome-binding site (59-AGGA-39) and ATG start codon embedded within the unique Nde I restriction site (59-CATATG-39) for in frame-fusion of the protein coding region of the studied genes [5,10]. The Km r gene was PCR amplified from the commercial pUC4K plasmid (Table S1) with the oligonucleotide primers KmHinCFW and KmHinCRV (Table S2), which introduced a flanking HincII restriction site. After HincII cleavage, the Km r gene was cloned in place of the Sm r /Sp r marker of pFC1 opened with NaeI and XmnI, yielding pFC1K which was verified by PCR and nucleotide sequencing (Big Dye kit, ABI Perking Elmer).

RNA isolation and analysis
200 ml of mid-log phase cultures (2.5610 7 cells ml 21 ) were rapidly harvested by vacuum filtration (less than 1 min); resuspended in 4 ml Tris 50 mM pH 8, EDTA 5 mM; immediately frozen in an Eaton press chamber cooled in a dry ice and ethanol bath and disrupted (250 MPa). RNA were extracted and purified with the Qiagen kit RNAeasy as we described [10,17]. RNA concentration and purity (A 260 /A 280 .1.9) were determined with a Nanodrop (Thermo scientific) and migration on agarose gel to verify the absence of RNA degradation. The absence of contaminant DNA was verified with the Taq DNA-dependent DNA-Polymerase (Invitrogen) using primers specific to the control gene rnpB (Table S2). For quantitative RT-PCR the gene specific primers were chosen so as to generate DNA fragment of similar length comprised between 163 bp and 234 bp. Each assay was triplicated, allowing the mean threshold cycle value (C T ) to be calculated from standard curve by the iQ5 optical system software (BioRad). Each gene-specific standard curve was made by 4 fold serial dilution of wild-type strain cDNA (ranging from 9375 to 9.16 ng) against log input cDNA concentration for each primer (data not shown). For each primer tested, the regression value (DC T versus cDNA concentration) was less than 0.1, indicating approximately equal amplification efficiencies. Then, for each studied gene the C T value was converted to gene copy number per ng of template cDNA.
Western blot analysis of the HoxF and HoxH proteins 40 mg of Synechocystis proteins separated on 12% SDS PAGE (Thermo scientific) were transferred (iBlot system; Invitrogen) to   nitrocellulose membrane (Invitrogen), which were blocked for 1 h at room temperature or overnight at 4uC with 5% non-fat milk in phosphate buffered saline (PBS). Immunodetection was performed using the following rabbit antibodies [18] anti-HoxF (dilution 1:2000) and anti-HoxH (dilution of 1:500). R800 goat anti-rabbit IgG (Invitrogen) were used as secondary antibodies (dilution of 1:1000), and immune complexes were visualized by chemiluminescence (ECL from GE Healthcare Amersham).

Hydrogenase activity measurements
Hydrogenase activities were measured on 1 ml aliquots of midlog phase culture concentrated 5.7-fold by centrifugation, using a modified Clark-type electrode (Hansatech, UK), and saturating amounts of Na-dithionite (20 mM) and methylviologen (5 mM) as the electron donor, as we previously described [10].

Results
The hoxEFUYH operon is dispensable to the photoautotrophic growth of Synechocystis To investigate the role of the octacistronic hoxEFUYH operon on the physiology of Synechocystis, we constructed a DhoxEFUYH::Km r deletion mutant ( Figure 1) by replacing the whole hoxEFUYH operon (from 58 bp upstream of the ATG start codon of hoxE to 8 bp downstream of the stop codon TAA of hoxH) by a Km r marker gene ( Figure S3). Therefore, the Km r gene was amplified from the pFC1K plasmid constructed in this study ( Figure S2 and Table S1), using specific oligonucleotide PCR primers (Table S2) that generated SphI and AflII restriction sites for cloning into pGEM-T ( Figure S3). Meanwhile, the two regions of Synechocystis DNA (about 300 bp each) flanking the hoxEFUYH operon were independently amplified with PCR primers (Table S2) that generated SphI or AflII restriction sites ( Figure S3). After cleavage, these Synechocystis DNA segments were cloned on each side of the Km r marker of the pGEM-T derivative to serve as platforms for homologous recombinations promoting targeted gene replacement [19] (i.e. replacement of the hoxEFUYH operon by the Km r gene) upon transformation. The resulting DhoxEFUYH::Km r DNA cassette ( Figure S3) was verified by PCR and nucleotide sequencing, prior to transformation to Synechocystis [19]. Km r mutants were analyzed by PCR ( Figure S4) to verify that the Km r marker had properly replaced the whole hoxEFUYH operon (from 71 bp upstream the hoxE ATG start codon to 19 bp after the hoxH TAA stop codon) in all copies of the chromosome, which is polyploïd [19]. The absence of the hoxEFUYH operon in the DhoxEFUYH::Km r mutant was confirmed upon the analysis of culture grown for a few generations in absence of Km to stop counter-selecting the propagation of possibly remaining wild-type (WT) chromosome copies, prior to the PCR assays ( Figure S4). We also confirmed through quantitative RT-PCR that the DhoxE-FUYH::Km r mutant completely lacks hoxEFUYH transcripts (data not shown), as well as the HoxF and HoxH proteins and hydrogenase activity ( Figure S5). These data, together with the fact that the DhoxEFUYH::Km r mutant grows as healthy as the WT strain under standard photoautotrophic conditions ( Figure S5), showed that the hoxEFUYH operon is dispensable in Synechocystis. This finding is consistent with the previously observed dispens-ability of the Synechocystis hoxHY genes in the otherwise WT strain [20], and the hoxEFUYH operon in the glucose tolerant mutant [2].
Construction and analysis of a mutant for temperaturecontrolled high-level expression of the hoxEFUYH operon: replacement of its weak promoter by the strong promoter lp R controlled by the temperature-sensitive repressor lcI 857 To increase the expression of the hoxEFUYH operon, while caring with the possibility that above a certain level it might become toxic, we decided to replace the weak promoter of the hoxEFUYH operon [10] by a strong and controllable promoter. Therefore, we used the lcI 857 -lp R system that expresses genes proportionally to the temperature of growth [4,5,21]: i.e. absence of expression at 30uC (the standard growth temperature) and strong expression at 39uC (good growth of wild-type cells). Practically, the 2.8 kb Km r -lcI 857 -lp R DNA cassette of the pFC1K plasmid presently constructed ( Figure S2) was amplified by PCR amplified, and introduced in place of the hoxEFUYH operon promoter (the 691 bp region upstream of the ATG start codon of hoxE), as follows ( Figure S6). The 252 bp region of Synechocystis DNA upstream of the hoxEFUYH operon promoter was amplified by PCR with specific oligonucleotide primers (Table  S2), digested with the BamHI and SphI and cloned upstream of the Km r marker of pFC1K opened with the same enzymes. Meanwhile, the 527 bp region of Synechocystis DNA encompassing the hoxE coding sequence was cloned as a NdeI-EcoRI DNA segment downstream of the lp R -promoter of pFC1K opened with the same enzymes. The resulting pTR-hoxEFUYH plasmid was linearized with EcoRV and transformed to Synechocystis, where homologous recombinations introduced the Km r -lcI 857 -lp R DNA cassette in place of the natural promoter of the hoxEFUYH operon. Km r transformant clones growing in standard conditions were analyzed by PCR ( Figure S7) to verify that the Km r -lcI 857 -lp R DNA cassette had properly replaced the natural hoxEFUYH operon promoter, in all copies of the polyploïd chromosome. We verified that this mutant, hereafter designated as TR-hoxEFUYH ( Figure 1), possessed no wild-type (WT) chromosome copies, even when cells were grown in absence of Km that otherwise counterselect the propagation of WT chromosome copies ( Figure S7). The TR-hoxEFUYH mutant strain grew as fit as the WT strain at both 30uC and 39uC ( Figure 2). Then, we verified through quantitative RT-PCR analysis that the lcI 857 -lp R promoter system expressed the TR-hoxEFUYH operon in a true temperature-controlled way. Therefore, total RNAs isolated from the TR-hoxEFUYH mutant and WT strains grown at either 30uC or 39uC were hybridized with specific RT-PCR primers (Table S2) designed to amplify an internal segment of each eight genes of the hoxEFUYH operon (hoxE, hoxF, sll1222, hoxU, hoxY, ssl12420, sll1225 and hoxH; Figure 1). The abundance of all eight transcripts were similar in WT cells grown at 30uC (the standard temperature) or 39uC (Figure 2), showing that the expression of the WT-hoxEFUYH operon from its natural promoter is not affected by these temperatures. By contrast, all eight hoxEFUYH mRNAs were much more abundant in TR-hoxEFUYH cells grown at 39uC as compared to 30uC, in agreement with our previous reports that at 30uC the lcI 857 repressor strongly inhibits the activity of the l PR promoter [4,5]. Furthermore, at 39uC the hoxEFUYH transcript levels were higher in the TR-hoxEFUYH mutant than in the WT strain ( Figure 2). Consistently, the abundance of the HoxF and HoxH proteins were higher in TR-hoxEFUYH cells grown at 39uC as compared to 30uC, or to WT cells grown either at 30uC or 39uC ( Figure 2).
Collectively, these data demonstrated that the hoxEFUYH operon is expressed in a temperature-controlled way in the TR-hoxEFUYH mutant, and that the overexpression of the hoxEFUYH operon at 39uC is not detrimental to cell life. However, the 39uCinduction of the level of active hydrogenase in the TR-hoxEFUYH mutant was modest in comparison to the strong induction of the hoxEFUYH transcripts or HoxF and HoxH proteins ( Figure 2). As the activity of the Ni-Fe hydrogenase can be limited by the availability of Fe and/or Ni atoms [22,23], we have also measured the level of active hydrogenase in cells grown at either 30uC or 39uC in standard medium (MM) with or without the addition of both Fe (17 mM) and Ni (2.5 mM). The data showed that, together, the overexpression of the hoxEFUYH operon and the increased Niand Fe-availabilities led to an eight-fold higher amount of active hydrogenase ( Figure 2). Though significant the gain in active hydrogenase remained lower than what was expected from the strong accumulation of hoxEFUYH transcripts and HoxF and HoxH proteins, suggesting that other limiting factors exist such as the abundance of the hypABCDEF proteins involved in assembly of the pentameric HoxEFUYH hydrogenase enzyme complex [2,9].
Construction and analysis of a mutant for concomitant temperature-controlled overexpression of the hoxEFUYH operon and hypABCDEF genes in Synechocystis To increase the formation of active hydrogenase enzymes, we decided to construct a mutant for temperature-controlled highlevel expression of not only the hoxEFUYH operon but also all six genes hypABCDEF within the same cells (Figure 1). Therefore, the scattered Synechocystis hypABCDEF genes ( Figure S1) were cloned in that order in our replicative plasmid vector pFC1 [5], under the control of the above mentioned lcI 857 -l PR system for tight temperature-controlled expression ( Figure S8). First, the hypA1 and hypB1 genes were PCR amplified from Synechocystis DNA (see Table S2 for the primers), and joined through standard PCRdriven overlap extension [24] in a single DNA molecule flanked by a NdeI restriction site encompassing the ATG start codon of hypA1 and a SalI site downstream of the hypB1 stop codon. After cleavage with both NdeI and SalI, the hypA1-hypB1 cassette was cloned in pFC1 opened with the same enzymes, yielding the pTR-hypAB plasmid. Then, the hypC, hypD and hypE genes were PCR amplified from Synechocystis DNA, joined in that order in a single DNA molecule, and cloned as a SalI and BspeI restriction DNA fragment in pTR-hypAB opened with the same enzymes. The resulting plasmid pTR-hypABCDE was opened with BspEI to clone the BspEI restriction fragment harboring the hypF gene generated by PCR, yielding the plasmid pTR-hypABCDEF ( Figure S8). The structure of the pTR-hypABCDEF plasmid was verified by PCR analyses ( Figure S9) and DNA sequencing of the lcI 857 -l PR cassette and all cloning junctions. Then pTR-hypABCDEF was introduced by conjugation [5] into the above-mentioned TR-hoxEFUYH mutant over-expressing the chromosomal hoxEFUYH operon in a temperature-regulated way. This yielded the Km r -Sm r /Sp r mutant hereafter referred to as TR-hoxEFUYH-hypABCDEF, which also carried the WT alleles of the hypABCDEF genes in its chromosome ( Figure 1). The TR-hoxEFUYH-hypABCDEF cells (abbreviated as TR2) grew as fit as the wild-type strain (WT) at 30uC or 39uC (Figure 3 and Figure S10). Then, we verified through quantitative RT-PCR that this mutant strongly expressed the hoxEFUYH and hypABCDEF genes in a temperature-controlled way, thanks to the lcI 857 and lp R devices. Indeed, the abundance of the hoxEFUYH and hypABCDEF transcripts was much higher in TR-hoxEFUYH-hypABCDEF cells grown at 39uC than at 30uC, or as compared to WT cells grown at 30uC or 39uC. Consistently, the HoxF and HoxH proteins were much more abundant in TR-hoxEFUYH-hypABCDEF cells grown at 39uC, than at 30uC, or as compared to WT cells grown at either 30uC or 39uC (Figure 3). Furthermore, the level of active hydrogenase of the TR-hoxEFUYH-hypABCDEF mutant was higher in cells cultivated at 39uC than at 30uC, or as compared to WT cells grown at either 30uC or 39uC ( Figure S10). Moreover, the gain of active hydrogenase observed at 39uC was significantly higher in the TR-hoxEFUYH-hypABCDEF mutant, which overexpresses the hoxEFUYH operon and the hypABCDEF genes, than in the TR-hoxEFUYH mutant, which overexpresses only the hoxEFUYH operon ( Figure 3). This finding is consistent with the role of the HypABCDEF in assembling a functional HoxEFUYH hydrogenase complex [2,9]. Collectively, our data demonstrated that the strong, temperature-controlled, expression of hoxEFUYH operon and the hypABCDEF genes are not detrimental to cell life. Again, the 13-fold increase in the amount of active hydrogenase observed in these TR-hoxEFUYH-hypABCDEF cells cultivated at 39uC, in the presence of additional Fe and Ni that positively influence hydrogenase activity, was low in comparison to the strong accumulation of the hoxEFUYH transcripts or HoxF and HoxH proteins (Figure 3 and Figure S10).
Construction and analysis of a mutant for strong constitutive expression of the hoxEFUYH operon and hypABCDEF genes in Synechocystis The present finding that the concomitant high-level expression of the hoxEFUYH and hypABCDEF genes was not toxic to the temperature-controlled mutant (TR-hoxEFUYH-hypABCDEF) growing at 39uC, prompted us to attempt further increasing it in a constitutive way as a step towards the engineering of a powerful strain for hydrogen production as well as to facilitate the comparative analysis of such a mutant with the WT and our DhoxEFUYH mutant (DhoxEFUYH::Km r ). Therefore, the plasmids pTR-hoxEFUYH and pTR-hypABCDEF were deleted of most of the lcI 857 gene encoding the temperature-sensitive repressor because it is not totally inactivated at 39uC [5] and Synechocystis grows poorly at higher temperatures. Practically, the pTR-hoxEFUYH and pTR-hypABCDEF plasmids were cleaved with the restriction enzyme PsiI to delete (517 bp) the lcI 857 repressor gene yielding the pCE-hoxEFUYH and pCE-hypABCDEF plasmids (CE for constitutive expression). The pCE-hoxEFUYH plasmid ( Figure  S11) was linearized with EcoRV and transformed to Synechocystis, where homologous recombinations introduced the Km r -lp R DNA cassette in place of the weak [10] hoxEFUYH operon promoter, in all copies of the polyploïd [19] chromosome ( Figure S12). This Km r CE-hoxEFUYH mutant grew as fit as the WT strain in standard (30uC) photoautotrophic conditions ( Figure 4). Then, we verified through quantitative RT-PCR that this mutant strongly expressed all eight genes of the hoxEFUYH operon, thanks to the lp R promoter, and accumulated the HoxF and HoxH proteins ( Figure 4). Consistently, the level of active hydrogenase in the CE-hoxEFUYH mutant was higher than that of the WT strain ( Figure 4). Together, the constitutive overexpression of the hoxEFUYH operon and the increased Ni-and Fe-availabilities led to a fourteenth-fold higher hydrogenase activity (Figure 4), i.e. a higher increase than the eight-fold value observed after heat induction of the temperature-controlled TR-hoxEFUYH mutant ( Figure 2).
To increase the formation of active hydrogenase enzymes, we introduced the above-mentioned pCE-hypABCDEF plasmid (Sm r / Sp r ; Figure S13) by conjugation [5] into the CE-hoxEFUYH mutant. This yielded the Km r -Sm r /Sp r mutant hereafter referred to as CE-hoxEFUYH-hypABCDEF, which also carried the WT alleles of the hypABCDEF genes in its chromosome (Figure 1). This CE-hoxEFUYH-hypABCDEF mutant grew as fit as the WT strain and the CE-hoxEFUYH mutant in standard (30uC) photoautotrophic conditions ( Figure 4). As expected, the CE-hoxEFUYH-hypABCDEF mutant strongly expressed the hoxEFUYH operon and the hypABCDEF genes accumulated the corresponding proteins ( Figure 4; and Table S3). Furthermore, the level of active hydrogenase of the CE-hoxEFUYH-hypABCDEF mutant was higher than those of the CE-hoxEFUYH mutant and the WT strain in that order ( Figure 4). Together, the constitutive overexpression of the hoxEFUYH operon and the hypABCDEF genes, and the increased Ni-and Fe-availabilities led to a seventeenth-fold increase in the level of active hydrogenase as compared to WT cells cultivated in absence of Fe and Ni supplementation (Figure 4). We think this strain with an increased amount of active hydrogenase is a suitable chassis for future gene manipulations required to engineer a powerful hydrogen producer.
The hoxEFUYH operon, but not the hypABCDEF genes, operates in the protection against hydrogen peroxide All aerobic organisms invariably produce reactive oxygen species, such as O 2 -(superoxide anion) and H 2 O 2 through the accidental autoxidation of redox enzymes [25], which occurs when their reduced cofactors accidentally reduces oxygen. This phenomenon is frequent in cyanobacteria, where their active photosynthesis massively produces oxygen and often generates an excess of electrons [26]. As the cyanobacterial hydrogenase enzyme complex has been proposed to act as an electron valve releasing some of the supernumerary electrons [9], we have tested the H 2 O 2 tolerance of the presently reported hydrogenase mutants under anaerobic conditions. The DhoxEFUYH::Km r (deletion) mutant appeared to be more sensitive to H 2 O 2 than the WT strain and the CE-hoxEFUYH mutant in that order ( Figure 5). Similar, but smaller, differences in the H 2 O 2 tolerance of the various strains were observed under standard photoautotrophic (aerobic) conditions (data not shown). These findings support the proposal that the hydrogenase enzyme act as an electron valve without ruling the possibility that the Hox enzyme also participates in the detoxification of H 2 O 2 . By contrast, the overexpression of the hypABCDEF genes actually decreased the tolerance to H 2 O 2 , as shown by the comparison of the WT strain and the CE-hypABCDEF mutant on one hand, and the mutants CE-hoxEFUYH and CE-hoxEFUYH-hypABCDEF on the other hand ( Figure 5). Collectively, these findings showed that the hoxEFUYH operon and the hypABCDEF genes contribute to H 2 O 2 tolerance, positively (hoxEFUYH operon) and negatively (hypABCDEF). Future experiments will be required to test whether the higher H 2 O 2 tolerance directed by the overexpression of the hoxEFUYH operon is due, directly or indirectly, to the increased abundance of (i) the HoxHY hydrogenase enzyme per se, (ii) the HoxEFU diaphorase enzyme, and/or (iii) the Sll1222, Ssl2420 and Sll1225 proteins of as yet unknown function.
The hoxEFUYH operon operates in the defense against the stress (likely redox) triggered by the reduced carbon metabolites glucose and glycerol Several lines of previous evidences prompted us to test the influence of the reduced carbon metabolites glycerol and glucose on the aerobic growth of the hydrogenase mutants presently reported. Glycerol, a cheap surplus of saponification and biodiesel industries [27], and glucose were both shown to stimulate hydrogen photoproduction in the cyanobacteria Cyanothece ATCC 51142 [28] and Arthrospira (Spirulina) maxima CS-328 [23], respectively. These data suggested that hydrogen photoproduction is often limited by reductant availability. In addition, we have previously reported that glucose is toxic to Synechocystis incubated under an otherwise physiological illumination [29], likely because it decreases CO 2 -assimilation that normally consume a large number of photosynthetic electrons, thereby enabling the spared electrons to recombine with O 2 to generate toxic reactive oxygen species. Like glucose, glycerol was found to be toxic to the WT strain growing aerobically under a normal light fluence ( Figure 5), in agreement with a previous report [30]. In contrast, the presently constructed hydrogenase overproducing mutants CE-hoxEFUYH and CE-hoxEFUYH-hypABCDEF were killed by neither glucose nor glycerol ( Figure 5). These findings are consistent with the proposal that the Hox hydrogenase enzyme reoxidizes NAD(P)H during fermentative conditions to allow catabolism of endogenous carbohydrates to proceed [9]. Collectively these findings strengthened the proposal that hydrogenase operates as an electron valve, preventing supernumerary photosynthetic electrons to recombine with O 2 to generate toxic reactive oxygen species.

Discussion
In spite of valuable previous studies, the role of the hydrogen metabolism remains puzzling in cyanobacteria, and the activity of the Ni-Fe hydrogenase, the hydrogen-producing enzyme, remains low [2,8,9]. Therefore, we have studied the hydrogen production machine of the model unicellular strain Synechocystis PCC6803. This was done through deletion of the hydrogenase-encoding hoxEFUYH operon, as well as the overexpression of the hoxEFUYH operon alone, or in combination with the six hypABCDEF genes (Figure 1) which operate in hydrogenase assembly and are scattered in the Synechocystis genome ( Figure S1). The DhoxE-FUYH::Km r deletion mutant (Figures S2-4) grew as fit as the wildtype (WT) strain ( Figure S5), showing that the hoxEFUYH operon is dispensable to Synechocystis. This finding is consistent with the dispensability of the hoxHY genes in the otherwise WT strain [20], and of the hoxEFUYH operon in the glucose tolerant mutant [2]. To increase the expression of the hoxEFUYH operon and the hypABCDEF genes, while caring with the possibility that above a certain level they might become toxic, we used the lcI 857 -lp R system that expresses genes proportionally to the temperature of growth [4,5,21]: i.e. absence of expression at 30uC (the standard growth temperature) and strong expression at 39uC (good growth of wild-type cells). Practically (Figures S6-10), we replaced the weak [10] natural promoter of the hoxEFUYH operon by the Km r -lcI 857 -lp R DNA cassette, and we cloned the hypABCDEF genes under the control of the same lcI 857 -lp R system, in the Sm r /Sp r plasmid pFC1, which replicates at the same 10-20 copies per cell as the chromosome [5]. As expected, the resulting mutants expressed the hoxEFUYH operon alone (TR-hoxEFUYH mutant) or together with the hypABCDEF genes (TR-hoxEFUYH-hypABCDEF mutant) in a temperature-controlled way (no expression at 30uC, strong expression at 39uC; Figures 2-3), and the 39uC-induced strong expression of the hoxEFUYH operon and the hypABCDEF genes was not detrimental to cell fitness. In the presence of higher Ni-and Fe-availabilities, the overexpression of the hoxEFUYH operon alone or in combination with the hypABCDEF genes led to eight-fold or thirteen-fold increased amounts of active hydrogenase, respectively (Figure 2). To our knowledge the TR-hoxEFUYH-hypABCDEF mutant is the first cyanobacterial strain capable of overexpressing simultaneously the endogenous hoxE-FUYH and hypABCDEF genes in vivo. Thanks to the tight temperature-control of the hydrogenase activity of this mutant (none at 30uC, strong at 39uC) it will be possible in the future to use high-throughput ''omics'' techniques to perform a kinetic analysis of the global cell responses to hydrogen production, to attempt distinguishing between rapid (likely specific) responses, and slow (likely indirect) responses.
To further increase the production of active hydrogenase we deleted the lcI 857 repressor-encoding gene from the abovementioned constructions, because it is not totally inactivated at 39uC [5] and Synechocystis grows poorly at higher temperatures. The resulting mutants CE-hoxEFUYH and CE-hoxEFUYH-hypABC-DEF (CE for strong constitutive expression) grew as fit as WT cells ( Figure 4) and exhibited higher levels of active hydrogenase than their temperature-regulated counterparts (fourteen-fold in CE-hoxEFUYH and seventeen-fold in CE-hoxEFUYH-hypABCDEF as compared to WT cells growing in standard medium). All mutants displayed higher levels of expression of the hoxEFUYH and hypABCDEF genes than that of active hydrogenase (Figure 4) indicating that other limiting factors should be dealt with. Therefore, we will pay a particular attention to glutathionylation (the formation of a mixed-disulfide between the cysteines residues of a protein and glutathione (the anti-oxidant tripeptide cglutamyl-cysteinyl-glycine) because we recently reported that the AbrB2 hydrogen regulator, and the mercuric reductase enzyme, can be glutathionylated [15,21].
Meanwhile, the presently constructed mutants proved useful to advance our understanding on the physiological role of the hydrogenase enzyme, which was very limited so far. We compared the H 2 O 2 tolerance of the WT and mutant strains to test whether the cyanobacterial hydrogenase enzyme truly acts as an electron valve releasing excess of photosynthetic electrons [9] to prevent their recombination with O 2 that generates toxic reactive oxygen species (ROS). We found that the hoxEFUYH operon and the hypABCDEF genes contribute to the protection against H 2 O 2 ( Figure 5), positively (hoxEFUYH operon) and negatively (hypABC-DEF). In addition, we tested the influence on our mutants of the reduced-carbon metabolites glucose and glycerol, a cheap surplus of industries [27], because they stimulated hydrogen production in the cyanobacteria Arthrospira (Spirulina) maxima [23] and Cyanothece ATCC 51142 [28]. Both glucose and glycerol were toxic to Synechocystis growing under an otherwise normal light fluence ( Figure 5), probably because these reduced metabolites somehow decline the electrons-consuming CO 2 -assimilation, thereby allowing spared electrons to recombine with O 2 and generate ROS. By contrast, neither glucose nor glycerol killed the hydrogenase overproducing mutants CE-hoxEFUYH and CE-hoxEFUYH-hy-pABCDEF ( Figure 5), in agreement with the proposal that the Hox hydrogenase enzyme reoxidizes NAD(P)H that serves for the catabolism of carbohydrates [9]. Collectively our findings strengthened the proposal that hydrogenase operates as an electron valve, preventing supernumerary photosynthetic electrons to recombine with O 2 to generate toxic ROS. We view the hydrogenase complex as an important enzyme to cyanobacteria as Synechocystis, which likes growing as biofilm, a thick network of autoaggregated cells [31], where the cells are certainly frequently exposed to H 2 O 2 and reduced metabolites released by their neighbours (living, or dying and lysing). This view is conforted by the absence of hydrogenase enzyme in most planktonic cyanobacteria living in open oceans.

Conclusions
Using gene deletion and overexpression we have shown that the hoxEFUYH operon operates in the defence against H 2 O 2 , glycerol and glucose stresses, and that the simultaneous overproduction of the HoxEFUYH and HypABCDEF proteins led to a 20-fold increase in active hydrogenase. We think that our sophisticated mutants with higher hydrogenase contents and a healthy growth will be very useful for the purification of large hydrogenase quantities for structural analyses; to continue deciphering the biological role of the hydrogen production machine; and for further genetic manipulations aiming at increasing the O 2 tolerance of the hydrogenase enzyme for a better hydrogen production. Figure S1 Schematic representation of hydrogen production machine in Synechocystis PCC6803 adapted from [9]. The genes are represented by arrows, which point in the direction of their transcription (http://bacteria.kazusa.or.jp/ cyanobase/), and are colored similarly to their protein products. The green numbers indicate the spacing distance (in kilobases) between the scattered genes. The hoxEFUHY operon is weakly transcribed [10] as the polycistronic mRNA (bent blue arrow), which encodes (i) the hydrogenase sub-complex (made by the HoxY protein and the HoxW-matured HoxH subunit); (ii) the HoxEFU diaphorase sub-complex; and (iii) the three proteins of unknown function (white forms). The electron transfer FMN cofactor, Fe-Ni center, and [4Fe-4S] and [2Fe-2S] clusters of the Hox proteins, are represented by the blue squares, the pink star, dark-red squares and light-red diamonds, respectively. The zincbound to HypA1 and HypB1 proteins is shown as the blue form. CP designates carbamoyl phosphate. The brown lines stand for the reversible inactivation of Hox activity mediated by oxygen. The photosynthetic membrane is represented in green. (TIFF) Figure S2 Construction of the Km r pFC1K plasmid for temperature regulated gene expression in Synechocystis. The genes are represented by large arrows, which point in the direction of their transcription. The red triangle indicates the strong lp R promoter followed by the lcro ribosome-binding site (59-AGGA-39) and ATG start codon embedded within a unique NdeI restriction site (59-CATATG-39) for in frame-fusion of the studied protein-coding regions. They are expressed in a temperature-controlled way thanks to the lcI 857 gene (hatched arrow), which encodes the temperature-sensitive repressor that tightly controls lp R . The transcription and translation stop signals (TT) preventing read-through of gene expression from the antibiotic resistance gene (Sp r /Sm r in pFC1 or Km r in pFC1K) are indicated by the black bars. (TIFF) Figure S3 Construction of the DhoxEFUYH::Km r DNA cassette for the deletion of the Synechocystis hoxEFUYH operon, which comprises the hoxE, hoxF, sll1222, hoxU, hoxY, ssl12420, sll1225 and hoxH genes in that order (Figure 1 and Figure S1). The genes are represented by white (sll1222, ssl2420 and sll1225) or pink (hoxE, hoxF, hoxU, hoxY and hoxH) boxes, which point in the direction of their transcription. The transcription terminator (TT) preventing the read-through of expression from the Km r gene is represented by the vertical grey rectangle. The rectangles designated as hoxup2 and hoxdwn represent the DNA regions flanking the hoxEFUYH operon, which served as platforms for homologous recombinations promoting the targeted replacement of the hoxEFUYH operon by the Km r gene. The strong lp R promoter is represented by the red triangle oppositely oriented to the lcI 857 gene, which encodes the temperature-sensitive repressor that tightly controls lp R . The transcription and translation stop signals (TT), which prevent read-through of gene expression from the Km r marker are indicated by the vertical grey bar. The hoxup region of DNA (purple rectangle) upstream of the hoxEFUYH operon promoter and the hoxE gene served as platform for homolous recombinations, which introduced the Km r -lcI 857 -lp R DNA cassette in place of the weak promoter [10] of the hoxEFUYH operon. (TIFF) Figure S7 PCR verification of the Synechocystis TR-hoxEFUYH mutant (TR1) for temperature regulated high-level expression of the hoxEFUYH operon. (A) Schematic representation of the hoxEFUYH operon locus in the WT strain or the TR1 mutant, which harbors the Km r -lcI 857 -lp R cassette in place of the natural 691 bp-long hoxEFUYH promoter region (starting from the first bp upstream of the hoxE ATG start codon). The oligonucleotides primers represented by small colored triangles (Table S2) Figure S8 Construction of the pTR-hypABCDEF plasmid for temperature regulated expression of the Synechocystis hypABCDEF genes. For the sake of clarity, the four genes hypB1 (sll1432), hypC (ssl3580), hypE (sll1462) and hypF (sll0322) are represented oppositely to their natural orientation (Figure 1 and Figure S1). The small colored arrows indicate the position of the oligonucleotide primers used for the PCR amplification (dashed lines) and assembly (blue arrows) used for cloning the hypABCDEF genes into the pFC1 vector [5], yielding pTR-hypABCDEF. These PCR primers (Table S2) are namely: HypA1NdeIFwd (blue rightward-pointing arrow) and Hy-pA1ASSRv (yellow leftward-pointing arrow) for PCR1; Hyp-B1ASSFwd (yellow rightward-pointing arrow) and HypB1SalIRv (green leftward-pointing arrow) for PCR2; HypCSalIfwdbis (green rightward-pointing arrow) and HypCASSrvbis (orange leftwardpointing arrow) for PCR3; HypEASSfwd (purple rightwardpointing arrow) and HypEBspeIrv (red leftward-pointing arrow) for PCR4; HypDASSfwd (orange rightward-pointing arrow) and HypDASSrv (purple leftward-pointing arrow) for PCR5; and HypFBspeIfwdbis (red rightward-pointing arrow) and HypFB-speIrv (red leftward-pointing arrow) for PCR6. The lp R promoter is represented by the red triangle oppositely oriented to the lcI 857 repressor-encoding gene. The transcription and translation stop signals (TT) preventing read-through of gene expression are indicated by grey bars. (TIFF) Figure S9 PCR verification of the pTR-hypABCDEF plasmid. (A) Schematic representation of the hypABCDEF genes (grey boxes) in the pTR-hypABCDEF plasmid replicating in E. coli (lane C+ for positive control) or in the Synechocystis mutant designated as TR-hoxEFUYH-hypABCDEF (TR2). The oligonucleotides primers (Table S2) used to generate the pTR-hypABCDEF specific DNA segments (dashed lines) of the following sizes: 1.3 kb (PCR1, panel B); 2.6 kb (PCR2, panel C) and 770 bp (PCR3, panel D) are namely: HypA1NdeIFwd (blue arrow) and HypB1SalIRv (green leftward-pointing arrow) for PCR1; HypC-SalIfwdbis (green rightward-pointing arrow) and HypFBspeIFwd-Bis (red arrow) for PCR2; and HypDASSrv (purple leftwardpointing arrow) and HypEASSfwd (purple rightward-pointing arrow) for PCR3. Marker (M F ) = 1 Kb plus DNA Ladder (Fermentas). Note that the PCR1-3 reactions can amplify only the adjacent hypABCDEF genes present in the pTR-hypABCDEF plasmid, not the chromosomal hypABCDEF genes because they are located too far away from each others (see Figure S1 and Figure  S8). This explains the absence of PCR products in the negativecontrol Synechocystis strain TR1 (the TR-hoxEFUYH mutant), which lacks pTR-hypABCDEF.  Figure S11 Construction of the Km r -lp R DNA cassette for constitutive strong expression of the Synechocystis hoxEFUYH operon. The genes are represented by large arrows, while the dark bar indicates the transcription and translation stop signals (TT), which prevent read-through of gene expression from the Km r marker. The strong lp R promoter is represented by the red triangle oppositely oriented to the repressor encoding-gene lcI 857 , which was inactivated by PsiI restriction during the construction of the Km r -lp R DNA cassette. The 252 bp hoxup region of DNA upstream of the hoxEFUYH operon promoter and the hoxE gene, served as platform for homologous recombinations that introduced the Km r -lp R DNA cassette in place of the weak natural promoter of the hoxEFUYH operon. (TIFF) Figure S12 PCR verification of the CE-hoxEFUYH mutant for strong constitutive expression of the hox-EFUYH operon. (A) Schematic representation of the hoxEFUYH operon in the WT strain or the CE-hoxEFUYH mutant (CE1), which harbors the Km r -lp R DNA cassette in place of the natural 691 bp-long hoxEFUYH promoter region (starting from the first bp upstream of the hoxE ATG start codon). The oligonucleotides primers represented by small colored triangles (Table S2) Figure S13 (A) Construction of the pCE-hypABCDEF plasmid for constitutive strong expression of the hypABCDEF genes. pCE-hypABCDEF was generated after the PsiI cleavage and religation of the pTR-hypABCDEF plasmid to inactivate the lcI 857 repressor gene, which normally controls the strong lp R promoter (red triangle). The genes are represented by large arrows while the transcription and translation stop signals (TT) are indicated by dark grey bars. (B) Schematic representation of the hypABCDEF genes in the pCE-hypABCDEF plasmid replicating in E. coli (lane C+ for positive control) or in the Synechocystis mutant designated as CE-hoxEFUYH-hypABCDEF (CE2). The oligonucleotides primers (Table S2) used to generate the pCE-hypABCDEF specific DNA segments (dashed lines) of the following sizes: 1.3 kb (PCR1, panel B); 2.6 kb (PCR2, panel C) and 770 bp (PCR3, panel D) are namely: HypA1NdeIFwd (blue arrow) and HypB1SalIRv (green leftward-pointing arrow) for PCR1; HypCSalIfwdbis (green rightward-pointing arrow) and HypFBspeIfwdbis (red arrow) for PCR2; and HypDASSrv (purple leftward-pointing arrow) and

Supporting Information
HypEASSRwd (purple rightward-pointing arrow) for PCR3. Marker (M F ) = 1 Kb plus DNA Ladder (Fermentas). Note that the PCR1-3 reactions can amplify only the adjacent hypABCDEF genes present in the pCE-hypABCDEF plasmid, not the chromosomal hypABCDEF genes because they are located too far away from each others (see Figure S1 and Figure S8). This explains the absence of PCR products in the negative-control Synechocystis strain CE1 (the CE-hoxEFUYH mutant), which lacks pCE-hypABCDEF. (TIF)