Harnessing Escherichia coli for Bio-Based Production of Formate under Pressurized H₂ and CO₂ Gases

The first obstacle to overcome for exploiting E. coli FHL-1 as a carbon fixing technology was the natural expression regime of the enzyme, which is geared naturally toward environmental conditions favoring the forward reaction. Thus, physiological FHL-1 biosynthesis is controlled by the presence of formate and acidic pH (13–15). The expression of the fdhF gene and hyc operon is coordinated and regulated by a formate-responsive transcriptional regulator FhlA (16) and the repressor HycA (17). Thus, it is clear that strategies to remove native, especially formate-dependent, control of FHL-1 biosynthesis are needed in order to produce active FHL-1 under all growth regimens.

First, an E. coli strain (MR87.5) was constructed in which other hydrogenases (ΔhyaB, ΔhybC) and potential formate production and utilization pathways (ΔpflA, ΔfdhE) were inactivated (Table 1). As previously observed in the context of bio-H₂ production, mutant strains unable to synthesize active pyruvate formate lyase (ΔpflA), which generates formate from pyruvate during fermentation, should only produce FHL-1 when exogenous formate is supplemented to the growth medium (18). This phenotype was observed here for strain MR87.5 (hycE^His, ΔhyaB, ΔhybC, ΔpflA, and ΔfdhE) where H₂-dependent CO₂ reductase (HDCR, reverse FHL-1) activity was only observed in MR87.5 after the cells had been pregrown in exogenous formate at 0.2% (wt/vol) final concentration (Fig. 1B).

Next, the FhlA regulator was specifically targeting for mutagenesis. It has been shown that FhlA variant E183K exhibits a constitutively active phenotype on hyc transcription, maintaining high expression levels even in the complete absence of externally added formate (19). In our work, the MR87.5 strain was modified by the inclusion of the FhlA^E183K allele on the chromosome to give E. coli strain MR93.25 (Table 1). This new strain demonstrated some improved HDCR activity when initially cultured in the absence of exogenous formate (Fig. 1B); however, surprisingly, HDCR activity in the FhlA E183K variant remained strongly inducible by pregrowth in external formate (Fig. 1B). This strain was deemed not suitable for further engineering.

Harnessing production of the entire FHL-1 enzyme is further complicated by the fact that the formate dehydrogenase and hydrogenase genes are located at separate loci on the chromosome. In an initial attempt to stabilize coproduction of the entire FHL-1 complex, a previous study engineered an E. coli strain in which the FDH-H moiety was physically tethered to HycB, resulting in the production of a fully assembled and functional complex (20). Keeping with this strategy here, the fdhF gene was first deleted from the parental strains before the hycA gene was replaced by a version of the fdhF gene fused to hycB using a hemagglutinin (HA) epitope tag (Fig. 1A). This new strain (MR40) was then further modified by the inclusion of alternative transcriptional promoter regions upstream of the ϕfdhF::hycB fusion allele (Table 1). The promoters T5, proD, tatA, and ynfE were chosen as various examples of strong, constitutive or anaerobically induced promoter sequences. The four new strains carrying these promoters (Table 1) were then analyzed for the production of the FdhF^HA-HycB fusion protein by Western immunoblotting against the HA tag (Fig. 1C) and for HDCR activity using intact whole cells (Fig. 1B). As shown in Fig. 1B, when the expression of FHL-1 was left under the control of what remains of the native hyc promoter (P_hyc) in MR40, the cells exhibited no HDCR activity when cultivated in the absence of formate, and no protein could be detected by Western blot analysis using an antibody raised against the HA epitope tag. Moreover, the MR40 strain yielded only low levels of the FdhF^HA-HycB fusion protein when cells were grown with extra formate in the medium (Fig. 1C). As a result, the HDCR activity was only partially restored when formate was supplemented in the growth medium (Fig. 1B).

Next, the synthetic promoter constructs were tested. Among the promoters screened, T5 is a strong promoter used in plasmid-based expression systems (21), and the ynfE promoter was proposed to be highly inducible under anaerobic conditions (22). Another promoter, termed proD, which is constitutive (23), was also tested. Of these, the MR60 strain, which contained the T5 promoter upstream of the ϕfdhF^HA::hycB allele fusion, showed the most convincing protein production yield in the absence of exogenous formate (Fig. 1C). The MR60 strain also displayed strong HDCR activity when grown in the complete absence of exogenous formate (Fig. 1B). In order to demonstrate that the observed HDCR activity was dependent upon active FHL-1, a further control strain was constructed. MR60 was modified by the addition of a ΔhycCD allele that would remove the membrane arm of FHL-1, thus rendering the enzyme inactive (11). The new strain, TOM001 (Table 1), displayed negligible HDCR activity in the small-scale assays (Fig. 1B). Taken together, it is clear that the strategy of coproducing a formate dehydrogenase-hydrogenase fusion protein under a single constitutive promoter, and by removing any requirement for formate for expression, has been successful in generating an E. coli strain harboring active HDCR.

It is notable that the HDCR/formate production yield in the engineered strain MR60 matched, but never exceeded, that observed using the parental strain (Fig. 1B). This suggested that the expression regime was ultimately not the limiting factor in formate production activity. Further genetic engineering was employed to further boost HDCR activity. Maturation of FHL-1 is a multistep process that depends on accessory proteins involved in the biosynthesis of the [Fe-S] clusters, the molybdenum cofactor (MoCo) of the formate dehydrogenase, and the [NiFe] active site cluster of the hydrogenase (9). Previous strategies to stimulate hydrogenase expression and activity have involved the deletion of the iscR gene (24, 25). Here, a version of the MR60 strain carrying a ΔiscR allele (E. coli strain MR62) was constructed (Table 1). Attention was also given here to the MoCo biosynthesis pathway, which is highly conserved and involves a series of accessory proteins and cosubstrates (26). Here, we focused on the deregulation of synthesis of the S-adenosyl methionine (SAM) radical, which plays a critical role in the first step of the pathway (27). Increasing cellular availability of SAM may remove a potential bottleneck in this highly complex biosynthetic pathway. To do this, the MR62 strain was further engineered by the inclusion of a ΔmetJ deletion to yield strain MR94.5 (Table 1). Finally, it was considered worthwhile to attempt to boost the [NiFe] cofactor biosynthesis capability in the strains, and to do this, these cells were transformed with a multicopy vector carrying a synthetic version of the hypA1B1C1D1E1X operon from Ralstonia eutropha (Cupriavidus necator) (28, 29).

Strains with engineered cofactor biosynthesis pathways were analyzed for H₂ production activity catalyzed by the FHL-1 forward reaction. As shown in Fig. 2, H₂ production activity initially decreased in the MR40 parent strain carrying the ϕfdhF^HA::hycB allele fusion, while the incorporation of a T5 promoter upstream of the fusion in the MR60 strain restored the activity to beyond native levels. The ΔhycCD derivative of MR60 (TOM001) was found to be devoid of hydrogen production activity (Fig. 2). This mirrored the behavior of all three strains in the HDCR assay (Fig. 1B). Subsequent deletions of the iscR or metJ or inclusion of extra [NiFe] cofactor accessory genes added no material improvements to FHL-1 activity (Fig. 2). This clearly shows that, under these growth conditions, the metal cofactor biosynthesis, insertion, and maturation pathways of the enzyme were not a limiting factor.

One major obstacle to consider is the thermodynamics and reversibility of the FHL-1 system. The standard redox potentials of the two half-reactions of FHL-1 are very close together; thus, the directionality of the enzyme is very strongly influenced by environmental conditions, with low-pH/high-formate/low-H₂ partial pressure favoring FHL activity and with high-pH/low-formate/high-H₂ partial pressure favoring HDCR activity. Clearly, it would be desirable, if possible, to minimize any tendency toward the FHL forward reaction while HDCR activity is ongoing.

The activities of metal-dependent formate dehydrogenases for formate oxidation and CO₂ reduction vary greatly depending on the originating biological system, the composition of the active-site metal (molybdenum or tungsten), and the nature of its coordinating ligand (either cysteine or selenocysteine amino acid side chains) (30–32). Molybdenum and tungsten are closely related and share an identical organic cofactor when found in enzymes. Overall, tungsten-containing formate dehydrogenases have been suggested to be more efficient at reducing CO₂ because of the lower midpoint potential of the active site metal (33). Thus, it was considered here whether E. coli FHL-1 could be produced as a variant containing tungsten.

One simple approach used to substitute molybdenum for tungsten in enzymes is the growth of E. coli in the presence of increasing amounts of tungsten salts (34, 35). Here, E. coli MR60 cells were first grown under anaerobic conditions using a rich medium without any further supplementation. The metal content of the FHL-1 enzyme was then analyzed by inductively coupled plasma mass spectrometry (ICP-MS) (Table 2). In theory, a perfectly assembled FHL-1 should contain one molybdenum atom and one nickel atom per mol of enzyme (10). The FHL-1 fusion protein complex was purified via a His tag present on the hydrogenase subunit as previously described (20, 36), and, as shown Table 2, the isolated protein was found to contain clearly detectable amounts of nickel ions in a ratio of 1:0.4 with molybdenum. Under these conditions, the enzyme contained essentially no tungsten (Table 2). Clearly, this experiment does not return a perfect 1:1 ratio for molybdenum to nickel. The suggestion is that, under this growth and purification regimen, half of the formate dehydrogenase component is either not stably bound to the complex or not correctly assembled. Note, however, that the fusion protein approach does confer some extra stability on the enzyme, as a similar purification of native FHL-1 resulted in a ratio of only 0.27 mol molybdenum for every mole of nickel (36).

Next, the growth medium was supplemented with tungstate salts before the FHL-1 was again purified and analyzed. In this instance, the ICP-MS data revealed the ratio of molybdenum to nickel in the enzyme had dropped to 0.01 (Table 2). However, the growth in the presence of tungstate salts had simultaneously boosted the tungsten present in the enzyme to a ratio of 0.7 for every nickel atom (Table 2). This demonstrates that tungsten can be incorporated into the FHL-1 enzyme in place of molybdenum when supplied in the growth medium as a tungstate salt.

Next, the MR60 cells growing in the presence of increasing quantities of tungsten salts were analyzed for both forward and reverse reactions of the FHL-1 enzyme (Fig. 3). Both FHL forward and HDCR reverse activities tended to decrease as the concentration of tungstate ions increased in the growth medium. However, the trend profiles of the inhibitions were strikingly different (Fig. 3). Notably, at 1 μM tungstate in the growth medium, a 50% loss of FHL forward (H₂ production) activity was observed, while the same cells retained full HDCR (CO₂ reduction) activity under these conditions (Fig. 3). This result strongly suggests that the substitution of the molybdenum atom at the active site of FDH-H by a tungsten atom can either subtly shift the catalytic bias toward CO₂ reduction or simply inhibit the forward reaction. This simple way to change the kinetic properties of the enzyme could be a useful discovery if FHL-1 is ever going to be exploited fully as a carbon capture technology. Note that, however, both forward and reverse activities were lost when cells were grown with the highest concentration of tungstate ions (Fig. 3). It has been shown that the expression of fdhF and hyc is normally regulated by molybdate concentration in the cell through the action of the transcriptional regulator ModE (37), but this is unlikely to be an issue in our engineered strain. However, there could be wider, global effects of tungstate on cofactor biosynthesis, especially through expression of the biosynthetic genes themselves, which are controlled by a riboswitch in E. coli (38). Indeed, previous studies showed that the incorporation of molybdenum or tungsten at the active site of formate dehydrogenases in Desulfovibrio species is regulated not only by different selectivities in metal incorporation but also at the level of gene expression (32, 39, 40).

The ultimate goal of this research was to generate a host stain, and define some growth conditions, that will perform HDCR throughout the growth phase. Next, we employed a bioXplorer P400 laboratory-scale bioreactor with a gas sparging system to allow a constant and efficient delivery of H₂, CO₂, and/or N₂ to the culture. The engineered MR60 strain, which can only generate formate via engineered FHL-1, was chosen and grown in the presence of tungstate salts to maintain unidirectional HDCR activity. Following inoculation, the oxygen present in the growth medium was observed to be rapidly consumed by the bacteria. When the residual dissolved oxygen had reduced to 0%, only then were H₂ and CO₂ sparged through the cell culture at 50 ml min⁻¹. In this first experiment, no overpressure was applied. Initially, a concomitant drop of pH in the growth medium was observed as CO₂ was added. Hence, sodium hydroxide was automatically pumped into the growth medium to maintain the pH at 7.0 (Fig. S1 in the supplemental material). Bacterial growth was observed in the first 8 h of the run, until the glucose was fully consumed, reaching a maximum turbidity optical density at 600 nm (OD₆₀₀) of 1.8 (i.e., 0.45 g cell dry weight [g_CDW] liter⁻¹) (Fig. 4). Under these conditions, a maximum of ∼8 mmol liter⁻¹ formate was produced after 24 h, with a maximum rate of formate production of 2.4 mmol liter⁻¹ h⁻¹.

These results were considered promising since, even at ambient pressure, the performances of the bioprocess using the optimized E. coli MR60 strain were considered to outcompete comparable systems in which microorganisms naturally produce formate, such as Desulfovibrio sp. (41). Indeed, while formate produced by the MR60-optimized strain of E. coli is in the same range as that produced by Desulfovibrio desulfuricans, the maximum rate of formate production is 4 times higher in this E. coli system. Furthermore, while formate production in D. desulfuricans was observed to begin upon entry into stationary phase, in this system, formate production started immediately upon H₂ and CO₂ sparging in the cell culture. Moreover, as formate production is deregulated in this genetic background, formate production continued even after cells entered stationary phase (Fig. 4). This clearly emphasizes the potential of an E. coli optimized strain for formate production from the hydrogenation of CO₂, even at ambient pressure.

To investigate the effect of elevated pressure on E. coli MR60 cell growth, glucose consumption, and formate production, the bioreactor was next pressurized at 2, 4, or 6 bar with H₂ and CO₂ at a flow rate of 50 ml min⁻¹. Under these conditions, formate production yield (amount of formate produced per unit of cell density) could be increased with gas partial pressure up to 4 bar pressure (Fig. 5); however, no further enhancement of yield of formate produced was observed above 4 bar pressure (Fig. 5). Strikingly, however, above 2 bar pressure, the absolute formate content in the bioreactor was seen to decrease drastically (Fig. 6C). This was accompanied by a clear inhibition of cell growth under elevated pressures of H₂/CO₂ (Fig. 6A) and a concomitant drop in glucose consumption rates (Fig. 6B). To determine whether the elevated pressure per se or the composition of the gas mixture itself was detrimental to the cells, the E. coli MR60 strain was subsequently placed in the pressurized bioreactor under 10 bar pressure of 100% nitrogen gas (Fig. 7A). Strikingly, neither cell growth rate nor the final cell density was impacted negatively by elevated N₂ pressure (Fig. 7A). The cells also produced normal levels of lactate during fermentation (Fig. 6D) but were unable to generate formate (Fig. 6D). This strongly suggests that inhibition of cell growth under elevated H₂/CO₂ pressure is either linked to increasing concentrations of molecular hydrogen and carbon dioxide themselves, or perhaps due to the FHL-1-catalyzed hydrogen-dependent CO₂ reductase activity the cell is being forced to carry out. Indeed, increasing CO₂ concentrations could conceivably induce reversal of certain central metabolic reactions (42) or perhaps interfere with the function of the endogenous carbonic anhydrases (43).

In order to address these questions directly, the TOM001 strain (MR60 ΔhycCD) was chosen for further experimentation. This strain has no FHL (forward) activity (Fig. 2) nor any HDCR (reverse) activity (Fig. 1). To investigate the effect of elevated pressure on E. coli TOM001 growth, the strain was introduced into the pressurized bioreactor set at 6 barG (absolute pressure) with H₂ and CO₂ (Fig. 7B). In the absence of active FHL-1, the TOM001 strain exhibited no significant growth defects when growing under pressurized CO₂/H₂ (Fig. 7B).

These data suggest that reverse FHL-1 activity (maximally forced in our engineered strains when growing under pressurized CO₂/H₂) induces a growth defect. This is most likely related to the complex membrane biology of FHL-1. Sequence analysis suggests the enzyme activity could be coupled to proton or ion transport across the membrane, although the experimental evidence for this in studies of E. coli FHL-1 is not strong (11). Likewise, formate itself must be secreted from the cell after production, and the activity of the channel involved could also be intimately linked with that of the transmembrane proton motive force (44). One solution to overcoming the observed growth defects could involve engineering a water-soluble version of FHL-1 that is not dependent upon attachment to the cytoplasmic membrane for activity.

It was recently demonstrated that bacteria from the Desulfovibrio genus can produce formate from H₂ and CO₂ (41, 45), involving a periplasmic HydAB [FeFe]-hydrogenase (in H₂ oxidation mode) and a cytoplasmic molybdenum-dependent formate dehydrogenase enzyme, both most likely connected via the periplasmic tetraheme cytochrome c₃ network (41). Moreover, it was proposed that D. desulfuricans was able to grow during formate production. As a result, the concentration of formate in the bioreactor increased for 64 h before a maximum steady-state value of 30 mM formate production was achieved. However, although this result was outstanding, the whole bioprocess clearly suffered from low biomass yield (41). In contrast, here, we demonstrated that E. coli can be harnessed for formate production by optimizing the native FHL-1 enzyme complex to behave as an HDCR enzyme. At pressures below 6 barG, the optimized strain itself showed comparable performances to D. desulfuricans cells grown in batch reactor. In addition, operating the bioreactor at moderate pressure (e.g., 2 barG) H₂/CO₂ led to a doubling in the formate concentration in the cell suspension. This clearly demonstrates the potential of engineering the E. coli strain as host for bio-based production of formate while fixing CO₂.

New biocatalysts with remarkable kinetic properties for CO₂ reduction have been recently discovered, such as the HDCR from Thermoanaerobacter kivui (46) or the formate dehydrogenase from Rhodobacter aestuarii (47). Indeed, the HDCR enzymes from Gram-positive acetogenic bacteria share many similarities with formate hydrogenlyases. The HDCR from Acetobacterium woodii is a soluble cytoplasmic enzyme that, while reversible, has the dedicated physiological role of reducing CO₂ to formate using H₂ as reductant (48). Unlike E. coli, A. woodii can utilize fixed formate as the sole carbon source; however, the organism can also be harnessed to perform as an efficient whole-cell carbon capture system (49). Unlike FHL-1, A. woodii HDCR is not membrane attached, which may be considered an advantage in biotechnological applications, but contains an [FeFe]-hydrogenase, which is of a class usually very oxygen sensitive and not naturally found in E. coli. Nevertheless, a heterologous expression system has recently been developed to allow production of the active enzyme in E. coli (50). Thus, developing whole-cell biocatalysis using engineered E. coli remains a very promising option and offers the potential for large-scale and low-cost production (51).

In this study, an E. coli host strain was designed and built for the bio-based production of formate from H₂ and CO₂ in a batch bioreactor. This involved the rational genetic engineering of formate hydrogenlyase-1, a complex bidirectional metalloenzyme, in its native cellular chassis. Chemical substitution of tungsten for molybdenum in FHL-1 was found to instill a catalytic bias toward H₂-dependent CO₂-reductase activity, which promises to be a key finding in future process development. The engineered strain showed an ability to grow under pressure up to 10 bar N₂. Under carbon capture conditions, the modified E. coli was observed to grow at 2 bar pressure of an H₂/CO₂ mixture and to successfully couple bacterial growth to formate production from carbon dioxide. At higher substrate concentrations, the FHL-1 activity became toxic to the cell. Hence, capitalizing on the rational design of the host strain combined with growth medium supplements and innovative reactor technology, the potential of our E. coli host to be employed in the bio-based production of formate from CO₂ was demonstrated. This initial process could serve as a basis for the development of a continuous gas fermentation bioreactor.

This work was based on E. coli K-12 MG059e1, which carries an hycE^HIS allele on the chromosome (36). Strains constructed and employed in this study are listed in Table 1. Gene deletions were carried out by transduction with bacteriophage P1 (52) combined with the lambda red protocol (53) or by homologous recombination using the pMAK705 protocol (54). For those constructed using lambda red, strains from the “Keio” collection of kanamycin-marked mutations in nonessential genes of the E. coli K-12 (BW25113) were used as deletion allele donors (55). Deleted genes were hyaB, encoding the catalytic subunit of Hyd-1; hybC, encoding the catalytic subunit of Hyd-2; pflA, encoding pyruvate formate lyase-activating protein; fdhE, encoding an accessory protein specific for Tat-dependent respiratory formate dehydrogenases; iscR, encoding a transcriptional repressor of the isc operon; and metJ, encoding a transcriptional repressor of the met regulon. Once target genes were replaced with a kanamycin resistance-containing deletion allele, the resultant strains were transformed with plasmid pCP20 (55), which encodes a recombinase, before the plasmid was cured at 42°C. The strain producing the FhlA^E183K variant was obtained by introducing a single-base-pair substitution in the fhlA gene. First, the 500 bp upstream and downstream of the fhlA gene were amplified by PCR and assembled in pMAK705 (54). The mutant allele was then incorporated by QuikChange PCR (Agilent) and then transferred onto the chromosome by homologous recombination (54).

Strains producing FdhF as an N-terminal fusion protein to HycB, joined by a linker sequence containing a hemagglutinin (HA) tag flanked by three glutamines on each side, were assembled using a gene fusion construct in pMAK705 that has been described previously (20). The first strain made here using that construct was MR40 (Table 1). In attempts to upregulate expression of ϕfdhF^HA-hycB, alternative promoter regions were used to modify MR40. First, an existing construct bearing the synthetic T5 promoter region (20) was used to assemble strain MR60 (Table 1). Note that MR60 differs from strain FZBup (pREP4) (20) in that it does not harbor the pREP4 plasmid needed to repress transcription from the engineered T5 promoter. Next, the E. coli proD promoter and the promoter of the E. coli tatA and ynfE genes were amplified and cloned separately into pMAK705 as EcoRI-BamHI DNA fragment (oligonucleotide sequences shown in Table 3). Then, ∼500 bp of sequence upstream of hycA and ∼500 bp of sequence downstream of the 5′ end of the ϕfdhF^HA::hycB allele were amplified and cloned in the same vectors as the KpnI-EcoRI and BamHI-HindIII DNA fragments, respectively. The promoter alleles were then transferred to the chromosome of MR40 as described (20, 54), resulting in three new strains, MR36, MR41, and MR72 (Table 1). Finally, strain TOM001 lacking the membrane arm of FHL-1 (MR60 ΔhycCD; Table 1) was prepared using a previously described pMAK705 plasmid carrying an unmarked ΔhycCD allele (11).

Cells that were grown under anaerobic fermentative conditions (5 liters) were harvested by centrifugation and suspended with lysis buffer containing 50 mM Tris HCl, pH 8.0, with 10 μg ml⁻¹ DNase I (Sigma), 50 μg ml⁻¹ lysozyme (Sigma), and a protease inhibitor cocktail (Roche) at 1 g wet cell weight per 10 ml of buffer. Cells were lysed using a high-pressure cell homogenizer (Homogenising Systems Ltd.) at 1,000 bar before unbroken cells and debris were removed by centrifugation. Membrane proteins were solubilized for 1.5 h at room temperature by adding N-dodecyl-β-d-maltopyranoside (DDM) 1% (wt/vol) directly to the crude extract. Then, the solubilized fraction was loaded onto a 5 ml HisTrap HP column (GE Healthcare) that had been equilibrated in 50 mM Tris HCl (pH 7.5), 150 mM NaCl, 50 mM imidazole, and 0.02% (wt/vol) DDM. Bound proteins were eluted with a 6-column-volume linear gradient of the same buffer containing 500 mM imidazole. Fractions were analyzed by SDS-PAGE (56), and those containing FHL-1 were pooled and concentrated in a Vivaspin (Millipore Inc.) filtration device (50-kDa-molecular-weight cutoff). Metal content was determined by inductively coupled plasma mass spectrometry (ICP-MS) performed as a service by the University of Edinburgh, United Kingdom. For Western immunoblotting, samples were first separated by SDS-PAGE before transfer to nitrocellulose (57). Blots were developed using a mouse anti-HA monoclonal antibody (Invitrogen) and a goat anti-mouse horseradish peroxidase (HRP)-conjugated secondary antibody (Bio-Rad) and visualized with an ImageQuant LAS-4000 imager (GE Healthcare).

Cells were grown under anaerobic fermentative conditions in rich LB medium containing 0.4% (wt/vol) glucose. When mentioned, the medium was supplemented with 0.2% (wt/vol) sodium formate and/or 1 μmol liter⁻¹ sodium tungstate. The culture was harvested by centrifugation (Beckman J6-MI centrifuge) for 30 min at 5,000 × g and 4°C. The cell paste was washed twice in 20 mmol liter⁻¹ 3-(N-morpholino)-propanesulfonic acid (MOPS) buffer, pH 7.4, before the cell pellet was suspended in the same buffer, and volume was adjusted to an OD₆₀₀ of 0.5 (∼0.125 g_CDW). For the FHL forward activity, 5-ml aliquots of washed whole cells were transferred into Hungate tubes before 0.2% (wt/vol) sodium formate was added. The tubes were flushed with nitrogen for 5 min before being incubated at 37°C for 20 h. H₂ content in the headspace was quantified by gas chromatography (GC). For the HDCR reverse reaction, 3 ml of washed whole cells was transferred into Hungate tubes. Tubes were flushed with N₂ for 5 min and then with H₂ for another 5 min before 5 ml CO₂ was added to the tubes. The cells were incubated at 37°C for 20 h. The cell suspension was then passed through a 0.2-μm filter syringe (Sartorius), and the formate content of the cell-free supernatant determined by high-performance liquid chromatography (HPLC).

Growth under pressure was performed in a commercially available bioXplorer 400P system (HEL Ltd., UK). The working culture volume was 250 ml. The bioreactor was set with three gas inlets, H₂, CO₂, and N₂. Each gas inlet was controlled by a gas mass flow controller. The bioreactor was equipped with pH, dissolved oxygen, temperature, and pressure controllers. The WinISO software handled the on-line monitoring and control systems of the reactor. The pH was maintained at 7.0 by addition of a 5-mol liter⁻¹ NaOH solution. The steel bioreactor chamber containing LB-rich medium was first autoclaved before 0.8% (wt/vol) glucose and 1 μmol liter⁻¹ sodium tungstate were added prior to inoculation. A 10% (vol/vol) culture grown in LB-rich medium prepared under aerobic conditions for at least 16 h was used to inoculate the bioreactor. After inoculation, O₂ present in the medium was observed to be rapidly consumed by the culture via the in-line O₂ electrode. The time taken to deplete the residual oxygen content was typically 30 min. After this point, the H₂, CO₂, and/or N₂ gases were sparged through the medium at a gas flow rate of 50 ml min⁻¹. Once the desired back pressure had been reached, the system was maintained at 37°C for 16 h. Aliquots were removed at regular time intervals where the OD₆₀₀ was recorded before cells were removed by passage through a 0.2-μm filter syringe. The resultant supernatants were analyzed for glucose, formate, and other fermentation products by HPLC.

Bacterial growth was monitored by following the OD₆₀₀, and the biomass yield was estimated from the OD₆₀₀ of the culture and the assumption that 1 liter culture with an OD₆₀₀ of 1 contains 0.25 g_CDW. Formate and fermentation metabolite analysis and quantification were determined by HPLC using an UltiMate 3000 uHPLC system (Thermo Fisher Scientific) equipped with an Aminex HPX-87H column (Bio-Rad) using a RefractoMax521 refractive index detector and a VWD-3100 variable-wavelength detector set at A₂₁₀. Typically, samples of 10 μl that were previously clarified through 0.2-μm filters were applied to the column equilibrated in 5 mmol liter⁻¹ H₂SO₄ with a flow rate of 0.8 ml min⁻¹ at 50°C.

Hydrogen gas in the culture headspace was quantified using a GC-2014 gas chromatograph (Shimadzu) equipped with a molecular sieve 5A capillary column and thermal conductivity detector. Nitrogen was used as carrier gas with a 25 ml min⁻¹ flow rate. Typically, 1-ml samples were collected using a gas-tight syringe with Luer lock valve (SGE) and used to manually fill a 500-μl loop.