Inhibition of hydrogen uptake in Escherichia coli by expressing the hydrogenase from the cyanobacterium Synechocystis sp. PCC 6803

Background Molecular hydrogen is an environmentally-clean fuel and the reversible (bi-directional) hydrogenase of the cyanobacterium Synechocystis sp. PCC 6803 as well as the native Escherichia coli hydrogenase 3 hold great promise for hydrogen generation. These enzymes perform the simple reaction 2H+ + 2e- ↔ H2 (g). Results Hydrogen yields were enhanced up to 41-fold by cloning the bidirectional hydrogenase (encoded by hoxEFUYH) from the cyanobacterium into E. coli. Using an optimized medium, E. coli cells expressing hoxEFUYH also produced twice as much hydrogen as the well-studied Enterobacter aerogenes HU-101, and hydrogen gas bubbles are clearly visible from the cultures. Overexpression of HoxU alone (small diaphorase subunit) accounts for 43% of the additional hydrogen produced by HoxEFUYH. In addition, hydrogen production in E. coli mutants with defects in the native formate hydrogenlyase system show that the cyanobacterial hydrogenase depends on both the native E. coli hydrogenase 3 as well as on its maturation proteins. Hydrogen absorption by cells expressing hoxEFUYH was up to 10 times lower than cells which lack the cloned cyanobacterial hydrogenase; hence, the enhanced hydrogen production in the presence of hoxEFUYH is due to inhibition of hydrogen uptake activity in E. coli. Hydrogen uptake by cells expressing hoxEFUYH was suppressed in three wild-type strains and in two hycE mutants but not in a double mutant defective in hydrogenase 1 and hydrogenase 2; hence, the active cyanobacterial locus suppresses hydrogen uptake by hydrogenase 1 and hydrogenase 2 but not by hydrogenase 3. Differential gene expression indicated that overexpression of HoxEFUYH does not alter expression of the native E. coli hydrogenase system; instead, biofilm-related genes are differentially regulated by expression of the cyanobacterial enzymes which resulted in 2-fold elevated biofilm formation. This appears to be the first enhanced hydrogen production by cloning a cyanobacterial enzyme into a heterologous host. Conclusion Enhanced hydrogen production in E. coli cells expressing the cyanobacterial HoxEFUYH is by inhibiting hydrogen uptake of both hydrogenase 1 and hydrogenase 2.


Background
Cyanobacteria are diverse, ancient (present 3.5 billion years ago), photosynthetic, and photoautotrophic, and it is believed that these bacteria evolved to become chloroplasts in plant cells [1]. Cyanobacteria have at least three enzymes involved in hydrogen synthesis/metabolism: (i) nitrogenase which produces hydrogen as nitrogen is converted to ammonia, (ii) uptake hydrogenase which consumes hydrogen produced by nitrogenase, and (iii) a bidirectional hydrogenase which can both consume and produce hydrogen [1]. The bi-directional hydrogenase was employed here since hydrogen production via the nitrogenase requires substantially more energy from the cell (16 ATP per mole of hydrogen) so it would be a lessenergy-efficient system (N 2 + 8H + + 8e -+16ATP → 2NH 3 + H 2 + 16ADP + 16P i ) [1].
The reversible (bi-directional) hydrogenase enzyme of Synechocystis sp. PCC 6803 produces hydrogen via the reaction 2H + + 2e ↔ H 2 (g) [1]; the source of the two electrons is NADH. The genes (hoxEFUYH) encoding this enzyme were identified and indicate that HoxFU are ironsulfur proteins that bind NADH (diaphorase) and that the large hydrogenase subunit HoxH contains six conserved sites for binding the Ni-Fe cofactor [2]. The small hydrogenase subunit HoxY may bind a [4Fe-4S] cluster [1]. The function of HoxE is not clear but it may be a bridging subunit in the membrane [3]. We chose this bacterium since it is well-characterized with the complete genome (3,573,470 bp) sequenced in 1996 [4]; hence, the hydrogenase is readily cloned. Transcription of HoxEFUYH is regulated by the LexA transcription activator, which specifically binds to the promoter region of the hox operon [5,6]. Note that the hydrogenase enzyme is sensitive to oxygen [7], so the assays are performed anaerobically.
Hydrogenase enzymes in E. coli are involved in two distinct modes of hydrogen metabolism: hydrogen production via hydrogenase 3 and hydrogen uptake by hydrogenases 1 and 2 [8]. Hydrogenase 1 (encoded by hyaABCDEF), hydrogenase 2 (encoded by hybOABC-DEFG), and hydrogenase 3 (encoded by hycABCDEFGHI) have nickel, iron, and three non-protein diatomic ligands (cyanide and carbon monoxide) in the active site which rely on the auxiliary proteins HypABCDEF (metalochaperones for NiFe insertion) and SlyD (nickel insertion) for maturation as well as may possibly rely on the chaperones GroEL/GroES [9]. Hydrogenase 1 and hydrogenase 2 are αβ heterodimers of a small subunit and a Ni-Fe containing catalytic large subunit and are present in the inner membrane facing the periplasmic space [10][11][12]. In E. coli, hydrogen is produced by hydrogenase 3 in the formate hydrogenlyase system (FHL) [13]. hycE encodes the large subunit of hydrogenase 3, and hycA encodes the repressor gene of the FHL system including the hyc operon [14].
HycI protease catalyses a C-terminal proteolytic cleavage of the HycE large subunit, and HypA, HypB, HypC, HypD, HypE, and HypF are required for metallocenter assembly [15]. Ordinarily, cyanobacteria employ photosynthesis fueled by light energy to produce hydrogen. However, if an active hydrogenase from a cyanobacterium may be expressed in E. coli, it is possible to use the energy from simple sugars (e.g., from agricultural products and wastes) to produce hydrogen. Other advantages of using E. coli are that the use of energy from sugar rather than light avoids relying on the availability of light and avoids the production of oxygen as occurs during photosynthesis. Oxygen as an impurity in hydrogen arising from photosynthetic activity is undesirable for fuel cells based on enzyme electrodes [16] and is undesirable as a fire hazard [17]. Hence, large production of hydrogen is more advantageous via fermentation rather than photochemical production [17].
Hydrogen is a 100% renewable fuel that burns cleanly, is efficient, and generates no toxic by-products [7]. Hydrogen is also the preferred choice for fuel cells. Not only is H 2 a clean fuel, producing only water as its by-product, it actually has a higher energy content than oil (142 MJ/kg for H 2 vs. 44.2 MJ/kg for oil), and is thus more efficient. Most of the H 2 now produced globally is by the process of steam reforming and the water-gas shift reaction, or as a by-product of petroleum refining and chemicals production [18]. Use of biological methods of H 2 production promises significant energy reduction costs, as these processes do not require extensive heating (or extensive electricity as in electrolysis plants). Here we report the cloning of an active cyanobacterial enzyme complex into E. coli to enhance hydrogen production primarily by limiting hydrogen uptake by the native E. coli hydrogenases.

Enhanced E. coli hydrogen production by HoxEFUYH
To create a recombinant system which produces hydrogen via fermentation, we cloned the hydrogenase locus (hoxE-FUYH) of Synechocystis sp. PCC 6803 into the well-studied bacterium E. coli. DNA sequencing and restriction enzyme digests showed the correct locus was cloned; our plasmid was designated pBS(Kan)Synhox ( Figure 1A). Native E. coli TG1 produces hydrogen via the FHL system during mixed-acid fermentations [19], and TG1 expressing the cyanobacterial hydrogenase produced 3-fold more hydrogen than cells which lacked the hoxEFUYH locus (22 ± 1 vs. 7 ± 1 μmol/mg protein) after 6 h in complex medium. More hydrogen was measured at least 23 times for cells expressing hoxEFUYH relative to the negative control that lacks the cyanobacterial locus so the effect is reproducible. The negative controls of both autoclaved TG1/pBS(Kan)Synhox and autoclaved E. aerogenes HU-101 did not produce hydrogen. Note that co-elution with pure hydrogen confirmed that hydrogen was produced by the E. coli cells and also our retention time for hydrogen was consistent with literature values (22.8 vs. 22 sec) [20]. In addition, the recombinant E. coli expressing the cyanobacterial hydrogenase produced about 2-fold more hydrogen after 6 h in complex medium than the positive control E. aerogenes HU-101 (11 ± 0.3 μmol/mg protein), a wellstudied producer of hydrogen [21,22]. Furthermore, hydrogen gas bubbles were clearly more visible in the recombinant strain compared to the host which lacked the cyanobacteria locus and also visible in the positive control but were not visible with autoclaved samples (Figure 2). Therefore, the recombinant E. coli strain produces significantly higher quantities of hydrogen gas.

Optimization of medium and time course of hydrogen production
The cyanobacterial genes were fused to a lac promoter in pBS(Kan); therefore, the expression of hoxEFUYH will be suppressed by catabolite repression if glucose is included in complex medium. To search for an optimal medium for producing hydrogen, glucose was replaced with fructose, galactose, maltose, lactose, glycerin, citrate, and succinate in complex medium. Hydrogen production in complex medium with fructose, galactose or maltose was 20% more than that with glucose in TG1/pBS(Kan)Synhox whereas the other carbon sources did not improve hydrogen production.
To investigate hydrogen production with TG1/ pBS(Kan)Synhox in more detail, a hydrogen time course experiment was performed. Hydrogen produced by E. coli TG1 cells with or without HoxEFUYH was maximum within 1.5 h ( Figure 3). Furthermore, from 1.5 h to 6 h, hydrogen produced in the absence of HoxEFUYH Hydrogen production with E. coli TG1/pBS(Kan)Synhox and E. aerogenes HU-101 (hydrogen bubbles shown) Figure 2 Hydrogen production with E. coli TG1/pBS(Kan)Synhox and E. aerogenes HU-101 (hydrogen bubbles shown). Representative samples shown from hydrogen assay experiments in complex medium with glucose (repeated 3 times).    Figure 3); hence, the hydrogen formed in the presence of HoxEFUYH was more stable. This suggested that hydrogen uptake is inhibited by expression of active HoxEFUYH. Note that after 18 h, the hydrogen yield from cells expressing hoxEFUYH was over 41 times more than that of the wild-type strain ( Figure 3).

Enhanced hydrogen production depends on native hydrogenase 3
To ascertain if elements of the native E. coli host FHL system impact the cyanobacteria hydrogenase system, hydrogenase activities of the cyanobacteria and the E. coli FHL were assayed in a series of mutants that lack either E. coli hydrogenase 3 of the FHL (HD705) or lack the maturation machinery required for assembling hydrogenases in E. coli (SE1497 and PMD23). We also examined a mutant that lacked the transcriptional activator FhlA, and thus would not express genes encoding the FHL complex (SE1174) as well as assayed a mutant that cannot form selenoproteins (WL400) and thus is unable to produce active formate dehydrogenase H (formate dehydrogenase H is sole electron donor for hydrogenase 3). Since the parent strain MC4100 with pBS(Kan)Synhox produced hydrogen but the FHL mutants (HD705, SE1174, SE1497 PMD23, and WL400) harboring pBS(Kan)Synhox did not produce hydrogen gas, active expression of the cyanobacteria hydrogenase system relies on both an active E. coli hydrogenase 3 as well as the maturation proteins of the host and cannot simply be due to HoxEFUYH acting as an electron donor to HYD3.

Role of the cyanobacterial proteins HoxEFUYH
To observe the expression of the recombinant enzymes, SDS-PAGE was performed. As shown in Figure 4, HoxU (26 kDa) from the cyanobacterium was clearly expressed in TG1/pBS(Kan)Synhox that produced hydrogen gas in both complex medium and LB medium while the other proteins (HoxE, 18 kDa; HoxF, 62 kDa; HoxY, 23 kDa; HoxH, 51 kDa) were not observed. As expected, HoxU was not observed in TG1/pBS(Kan) (negative control).

HoxU
Also, the expression of HoxU was greater in complex medium relative to LB medium and complex medium produced more hydrogen. Hence, HoxU of Synechocystis sp. PCC 6803 may play a major role in the elevated hydrogen production in E. coli.
To discern if only HoxU is required for producing more hydrogen, hoxU was cloned under the lac promoter (designated as pBS(Kan)HoxU, Figure 1B), and hydrogen production in TG1 harboring pBS(Kan)HoxU was examined. TG1/pBS(Kan)HoxU produced more hydrogen than TG1/ pBS(Kan) but only 44.6 ± 7.3% of that of TG1/ pBS(Kan)Synhox. These results indicate that the additional hydrogen produced by expressing HoxEFUYH is not solely the result of active HoxU; hence, other active HoxEFUYH proteins are required for producing hydrogen.

Mechanism of enhanced hydrogen production is via inhibition of hydrogen uptake
Since the hydrogen time course experiment ( Figure 3) showed the hydrogen produced by cells expressing hoxE-FUYH was more stable than hydrogen from cells which lacked this locus and given that the greater hydrogen production seen upon cloning the cyanobacterial locus was dependent on both the native hydrogenase maturation proteins and active hydrogenase 3, we theorized that Hox-EFUYH may be influencing hydrogen uptake. Recall that hydrogenase 1 and 2 are involved in hydrogen uptake only [10,11]. Hence, hydrogen uptake activity was measured three ways to investigate this hypothesis. As shown in Table 1, E. coli TG1 expressing HoxEFUYH had 3.3 times less hydrogen uptake compared to the negative control TG1/pBS(Kan) and E. coli MC4100 had similar results. These hydrogen uptake results were corroborated by using a plate assay for reversible hydrogenase activity which showed 10-fold less hydrogen uptake upon expressing HoxEFUYH and by a GC-based hydrogen uptake assay which indicated that H 2 uptake activity in TG1/ pBS(Kan)Synhox is 2.1 ± 0.2-fold less than TG1/pBS(Kan) over 0 to 6 h. Hence, the active cyanobacterial enzymes (HoxEFUYH) inhibit hydrogen uptake consistently in E. coli.
To determine by which of the three native E. coli hydrogenases that hydrogen uptake was affected by HoxEFUYH, a series of isogenic mutants of E. coli BW25113 was used ( Table 1). Using two mutants for hydrogenase 1 (hyaA and hyaB), hydrogen uptake was measured and found to increase consistently 2-fold upon removing hydrogenase 1 in the presence of HoxEFUYH; hence, HoxEFUYH decreases hydrogen uptake via hydrogenase 1. Similarly, using two mutants for hydrogenase 2 (hybB and hybC), hydrogen uptake increased consistently 2.5-to 2.8-fold (Table 1) upon removing hydrogenase 2 in the presence of HoxEFUYH; hence, HoxEFUYH decreases hydrogen uptake via hydrogenase 2. In contrast, upon adding Hox-EFUYH, isogenic mutations that eliminate hydrogenase 3 activity (hycE and hycG) ( Table 1) decreased hydrogen uptake (rather than increasing it). In addition, hydrogen uptake in the hyaB hybC double mutant that eliminated hydrogenase 1 and 2 activity was identical to that of the wild-type strains expressing hoxEFUYH and the expression of hoxEFUYH had no effect in the double mutant (Table  1). Corroborating these results, hydrogen production by BW25113 hyaB hybC/pBS(Kan)Synhox was the same as BW25113 hyaB hybC/pBS(Kan) (Figure 3). On the other hand, a triple mutant (hyaB hybC hycE; defective in hydrogenase 1, 2, and 3) with pBS(Kan) or pBS(Kan)Synhox did not produce hydrogen (data not shown), indicating that hydrogenase 3 is essential for producing hydrogen. Therefore, HoxEFUYH works through hydrogenase 1 and hydrogenase 2 rather than through hydrogenase 3, and the decrease in hydrogen uptake seen with the hydrogenase 3 mutants is due to HoxEFUYH inhibition of the remaining active hydrogenase 1 and 2 enzymes.

DNA microarrays
To investigate whether cloning of the cyanobacterial hox-EFUYH merely increased hydrogen production by up-regulating the native E. coli hydrogenase system, we examined differential gene expression upon expression of hoxEFUYH from pBS(Kan)Synhox. The microarrays showed that gene expression for the hya, hyb, hyc, and hyp operons was not altered between TG1/pBS(Kan)Synhox and TG1/pBS(Kan); hence, functional HoxEFUYH is necessary for enhanced hydrogen production in E. coli (Additional file 1). Surprisingly, the differential gene expression indicated that primarily biofilm-related genes are regulated by expressing HoxEFUYH as shown in Table 2

Discussion
To date, no cyanobacterial hydrogenase protein has been actively expressed in E. coli. Functional expression here of the cyanobacterial hydrogenase components in E. coli allows for both mutagenesis for structure/function determinations as well as for enhanced hydrogen production via saturation mutagenesis and DNA shuffling [23]. Also, E. coli cells offer two advantages over normally photosynthetic microbes regarding protein evolution. First, the transformation efficiency of E. coli (greater than 10 9 transformants per microgram of plasmid) is at least three orders of magnitude greater than those of photosynthetic bacteria [24]. A rapid E. coli-based genetic selection method for hydrogenase activity would enable the sampling of thousands of different hydrogenase mutants in a single day. Second, hydrogen production by cyanobacteria requires light, and oxygen is produced through photosynthesis; oxygen production is undesirable for fuel cells; therefore, it is more desirable to clone the hydrogenase into E. coli rather than cyanobacteria since E. coli is a facultative anaerobe.
In cyanobacterium Synechocystis sp. PCC 6803, HoxFU are iron-sulfur proteins that bind NADH (diaphorase) [25]. Also, HoxU is thought to serve as the bridging unit in the link between respiration and the hydrogenase [26]. In addition, cyanobacteria Anacystis nidulans SAUG 1402-1 and Anacystis sp. PCC 7942 showed reduced hydrogenevolution catalyzed by the bidirectional hydrogenase upon mutation of hoxU [27]; these indicate that HoxU is important for hydrogenase activity. Our SDS-PAGE results demonstrated that the HoxU protein (26 kDa) was clearly expressed in TG1/pBS(Kan)Synhox and that the expression of HoxU increased according to increasing hydrogen production. Since only HoxU was seen with SDS-PAGE, we surmised that it may have its own promoter. Corroborating this, between the stop codon of hoxF and the start codon of hoxU there is a gap of 705 bp that contains six putative promoters upstream of hoxU based on promoter prediction software [28]. The expression of HoxU is probably regulated independently by at least one of these promoters; this suggested that controlling expression of HoxU may provide significant insight for elevating hydrogen production, which proved correct since cloning only hoxU accounted for about half of the hydrogen produced by cloning hoxEFUYH. This shows the other cyanobacterial proteins (HoxEFYH), although not clearly observed with SDS-PAGE, are beneficial for producing hydrogen.
Since expressing HoxUYH in E. coli TG1 yields nearly the same effect as expressing all of HoxEFUYH via TG1/ pBS(Kan)Synhox, HoxUYH are clearly important for enhanced hydrogen production.
Hydrogen production was enhanced as much as 41-fold via production of the active cyanobacterial HoxEFUYH, Standard deviations shown from one representative experiment with 2 replicates. a IPTG was added for hydrogen uptake (1 mM) assays activity for 6 h in complex medium. and the hydrogen produced using TG1/pBS(Kan)Synhox was more stable in comparison to that of TG1/pBS(Kan); this indicated that TG1/pBS(Kan)Synhox has reduced hydrogen uptake activity. Since any mutation related to hydrogenase 3 eliminated the benefit of expressing the cyanobacterial hydrogenase, it is clear that hydrogenase 3 is necessary for the HoxEFUYH effect since only hydrogenase 3 produces hydrogen in E. coli whereas HoxEFUYH maintains this hydrogen that is produced by limiting hydrogen uptake by hydrogenase 1 and 2. In accordance with this interpretation, hydrogen uptake was found to be 5 to 10 times lower upon expression of HoxEFUYH. Also, the microarray analysis shows that native hydrogenase gene expression is not affected by enhanced hydrogen production with TG1/pBS(Kan)Synhox, so there are no transcriptional effects related to cloning hoxEFUYH. Taken together, these results show expression of HoxEFUYH increases hydrogen production by reducing hydrogen uptake of the native E. coli hydrogenases, but hydrogenase 3 is required to produce the hydrogen in the first place. Note that hydrogenase 1 and 2 have hydrogen uptake activity [10,11], and hydrogenase 3 has hydrogen produc-tion activity (hydrogenase 3 is the primary source of hydrogen gas production in E. coli [29] as hydrogenase 4 is inactive [30]) In contrast to the native hydrogenase gene expression, the DNA microarrays indicated the expression of many biofilm-related genes were altered upon expression of HoxE-FUYH (Table 2), and this altered expression led to an increase in biofilm formation. Interestingly, our study of temporal gene expression in E. coli K-12 biofilms [31] shows that the genes related to hydrogenase 1 (hyaAB-CDE) and to hydrogenase 2 (hybBC) are transiently repressed, that the genes related to hydrogenase 3 (hycBF) and hydrogenase 4 (hyfBC) are up-regulated in the process of biofilm formation, and that some hya or hyf mutants produce 3-to 7-fold more biofilm. So the fact that expression of HoxEFUYH affects biofilm genes ( Table 2) and that hydrogenase genes are routinely found in biofilm studies [31][32][33][34] suggest that biofilm formation is related to enhanced hydrogen production. Hence, it appears that biofilm formation may repress hydrogen uptake activity or induce hydrogen production, either of which would  [32][33][34] result in enhanced hydrogen production. In the future, we will need to ascertain whether biofilm formation is directly related to enhanced hydrogen production and how HoxEFUYH represses hydrogen uptake activity in E. coli.
To the best of our knowledge, on a protein basis,

Conclusion
E. coli TG1 cells with pBS(Kan)Synhox yielded 41 times more hydrogen after 18 h than those with empty vector pBS(Kan) due to active HoxEFUYH from Synechocystis sp. PCC 6803 (primarily through active HoxUYH). The mechanism for this enhanced hydrogen production is that hydrogen is formed first by hydrogenase 3 (so the HoxEFUYH effect relies on active hydrogenase 3 and its maturation proteins), then HoxEFUYH inhibits hydrogen uptake by E. coli native hydrogenase 1 and hydrogenase 2.
In effect, a novel way to reduce reversible hydrogen formation has been discovered using a cyanobacterial locus.

Bacterial strains, growth, and total protein
Strains are shown in Table 3. E. coli cells containing pBS(Kan) and derivatives were initially streaked from -80°C glycerol stocks on Luria-Bertani (LB) agar plates [40] containing 100 μg/mL kanamycin and incubated at 37°C. After growth on LB agar plates, these strains were cultured from a fresh single colony in LB medium [40] or complex medium [22] supplemented with 100 μg/mL kanamycin at 37°C with shaking at 250 rpm (New Brunswick Scientific Co., Edison, NJ). Wild-type E. coli K-12 BW25113 was obtained from the Yale University CGSC Stock Center, and its isogenic deletion mutants (Keio collection) were obtained from the Genome Analysis Project in Japan [41].

Eliminating kanamycin resistance and P1 transduction
Plasmid pCP20 [43] was used as described previously [44] to eliminate the kanamycin resistance gene (kan R ) from the isogenic BW25113 mutants (Keio strains) defective in hydrogenase 1, hydrogenase 2, and hydrogenase 3 so that pBS(Kan)Synhox could be added and so that a double and triple mutant could be constructed (Table 3). P1 transduction [45] and pCP20 were used to create E. coli MW1000 (hyaB hybC Δkan) from BW25113 hybC Δkan by transferring hyaB kan R via P1 transduction and using pCP20 to eliminate the kanamycin resistance marker.

Hydrogen uptake assays
Hydrogen uptake activities were measured as the increase in absorbance as oxidized methylviologen (MV) (ε 604 = 13.9 mM -1 cm -1 [21]) is reduced [MV +2 + 1/2H 2 → MV +1 + H + ] as reported previously [46] except whole cells were used rather than lysed cells. Cells were prepared as for the hydrogen assay except 1 mM IPTG was added to induce expression of the cyanobacterial hydrogenase system for 6 h, then the cell pellets from 1.5 mL were resuspended with 1 mL Tris buffer (50 mM, pH 8.0) in the anaerobic glove box. Oxidized MV (MV +2 , colorless) solution (1 mL, 0.8 mM in 50 mM Tris buffer, pH 8) was sparged first with nitrogen gas for 10 min to remove oxygen to prevent residual oxygen from oxidizing any reduced MV (MV +1 , purple) that is formed by the hydrogenases, was poured into cuvettes that were sealed with rubber stoppers, and was sparged with pure hydrogen gas for 10 min. Whole cell suspensions (0.5 mL) were mixed into the cuvettes and the change in absorbance during 5 min was monitored using a spectrophotometer (Varian, Walnut Creek, CA). Two independent cultures were used.
To corroborate the MV uptake assay, a filter paper assay for reversible hydrogenase activity was used based on a method described previously [47]. Filter paper (Whatman 541; Whatman plc, 27 Great West Road, UK) was immersed in MV solution (18 mM in 6 mM Tris buffer, pH 7.5), was air-dried with a hair dryer, and was firmly pressed on agar plates containing equal-size of colonies. The filter paper was then incubated at room temperature in a moist atmosphere of pure hydrogen in a Gas-Pak anaerobic chamber. The cells and adjacent white filter turned from white to blue-purple in approximately 10 min.
For the GC-based hydrogen uptake assay, cells were prepared as for the hydrogen assay except 1 mM IPTG was added to induce expression of the HoxEFUYH enzymes and after 6 hours, the cell pellets from 100 mL were resuspended with 25 mL phosphate buffer (100 mM, pH 7.5) including 1 mM IPTG in the anaerobic glove box. The cell suspension (20 mL) was added to sealed crimp-top vials (27 mL). The bottles were sparged with hydrogen for 10 min, were incubated at 37°C with shaking for 3 to 6 h, and the amount of hydrogen in the head space was measured as described above.

Microarray analysis
To isolate RNA from hydrogen producing cells, TG1/ pBS(Kan)Synhox and TG1/pBS(Kan) were cultured in complex medium with fructose as for the hydrogenase assay. RNA was isolated as described previously [48] with the RNeasy kit (Qiagen, Inc.). To inhibit RNase and ensure high-quality RNA, β-mercaptoethanol, which acts as a reducing agent to irreversibly denature RNase, and guanidinium isothiocyanate contained in the RLT buffer (RNA Lysis Tissue, RNeasy mini kit: Qiagen, Inc.), which is a strong but temporary RNase-denaturing agent, were utilized. The E. coli GeneChip antisense genome array was used (part no. 900381, Affymetrix, Inc., Central Expressway, Santa Clara, CA), and contains probe sets for all 4,290 open reading frames, rRNA, tRNA, and 1,350 intergenic regions. Analysis of the microarray data was as previously described [32], and the data have been deposited in the NCBI Gene Expression Omnibus [49] and are accessible through accession numbers GSM129630 and GSM129631.

Ninety-six-well biofilm assay
Biofilm formation was quantified in 96-well polystyrene plates as reported previously [50]. Biofilm formation of E.
coli TG1 with pBS(Kan)Synhox and pBS(Kan) was measured in LB supplemented with 0.2% glucose under anaerobic conditions using a Gas-Pak system. Thirty replicate wells were averaged to obtain each data point. Three independent cultures were used.