Moorella thermoacetica ATCC 39073 and its derivatives were used in this study (Table 1). Modified ATCC1754 PETC medium comprising 1.0 g of NH₄Cl, 0.1 g of KCl, 0.2 g of MgSO₄·7H₂O, 0.8 g of NaCl, 0.1 g of KH₂PO₄, 0.02 g of CaCl₂·2H₂O, 2.0 g of NaHCO₃, 10 ml of trace elements, 10 ml of Wolfe’s vitamin solution (Tanner, 1989), and 1.0 mg of resazurin/L of deionized water was used as the basal medium (Tanner et al., 1993). The pH of the solution was adjusted to 6.9. The medium was prepared anaerobically by boiling and cooling under an N₂–CO₂ (80:20) mixed-gas atmosphere. After cooling, the medium was dispensed into 125 ml glass culture vials (serum bottles) under an N₂–CO₂ mixed-gas atmosphere. The vials were crimp-sealed and autoclaved.

Before starting the culture, fructose, yeast extract, and L-cysteine·HCl·H₂O were added to reach final concentrations of 2.0, 1.0, and 1.2 g/l, respectively. The final volume was adjusted to 50 ml with water. To provide H₂, the headspace pressure in the vials was adjusted to 0.12 MPa by using N₂–CO₂ (80:20) mixed-gas. H₂ gas was then injected at the desired pressure. For example, when 0.01 MPa of H₂ was tested, the total pressure was adjusted to 0.13 MPa by the H₂ gas injection. Cells were grown at 55°C with shaking at 180 rpm.

We sampled and analyzed 1 ml of the culture medium at each time point and calculated the dry cell weight using the optical density (OD) at 600 nm [OD₆₀₀; 1 g (dry cell weight)/L = 0.383 OD₆₀₀; Iwasaki et al., 2017]. The culture supernatant was analyzed for the amount of fructose, formate, acetate, ethanol, and acetone using high-performance liquid chromatography (HPLC; LC-2000 Plus HPLC; Jasco, Tokyo, Japan) equipped with a refractive index detector (RI-2031 Plus; Jasco), Shodex RSpak KC-811 column (Showa Denko, Kanagawa, Japan), and Shodex RSpak KC-G guard column (Showa Denko) at 60°C. Ultrapure water containing 0.1% (v/v) phosphoric acid was used as the mobile phase at a flow rate of 0.7 ml/min, and crotonate was used as the internal standard (Miura et al., 2014). The gas composition in the headspace of the culture vials was analyzed using GC-8A gas chromatograph (Shimadzu, Kyoto, Japan) equipped with a thermal conductivity detector and a stainless steel column packed with activated carbon at 70°C. Argon was used as the carrier gas (Miura et al., 2014). The total gas pressure in the headspace was measured using a differential pressure gauge (DMC-104 N11; Okano Works, Tokyo, Japan).

Strains were grown to reach the exponential phase between 0.5 and 0.7 OD₆₀₀. The culture was immediately filtered to collect cells equivalent to a total count of 20 OD₆₀₀ (volume [mL] × OD₆₀₀ ≈ 20). Filtration was performed using hydrophilic PTFE, 1 μm pore size, and a 90-mm-diameter filter disk (Omnipore; Merck KGaA, Darmstadt, Germany). Harvested cells were immediately immersed in pre-chilled methanol containing 100 μM ribitol and 100 μM (+)-10-camphorsulfonate to quench the metabolic activity. This procedure was quickly performed, within 45 s after opening culture vials, to avoid metabolites from artifacts, such as those caused by oxygenation, degradation, and other modifications. Subsequently, intracellular metabolites were extracted using the chloroform–water–methanol method (Bolten et al., 2007). The supernatant was then concentrated using a centrifugal concentrator (CC-105; Tomy, Tokyo, Japan). According to a previous study, pre-treatment and analysis of the dried samples were performed (Wada et al., 2022).

Moorella thermoacetica can convert one hexose molecule to three acetate molecules in theory (Figure 1A; Fontaine et al., 1942; Schuchmann and Müller, 2014, 2016). The engineered strains were designed to produce ethanol from acetyl-CoA in two steps (reducing reactions; Rahayu et al., 2017; Table 1). The reducing equivalents provided by glycolysis were assumed to be properly consumed (Figures 1A,B). In a model, the Ech complex, HydABC complex, and NfnAB complex would convert the reducing equivalents to NADH and NADPH. These NADH and NADPH would be consumed by the Wood–Ljungdahl pathway to convert CO₂ to acetate in the wild-type strain, whereas reduction of acetyl-CoA would consume the NADH and NADPH in the ethanol-producing strains. Therefore, the redox conditions in the engineered strains producing ethanol should be balanced. In fact, we previously observed that all engineered strains (Table 1) grew and produced ethanol on hexose sugars. However, one molecule of CO₂ is released to produce one molecule of ethanol because of the requirement for extra reducing equivalents (approximately 33% of the carbon is released from hexose sugars in theory). Therefore we supplied H₂ to increase carbon utilization by mixotrophy.

We used a culture containing fructose as the carbohydrate substrate. H₂ was added to the headspace of the culture vial at a partial pressure of 0.08 MPa (equivalent to 40% of the gas phase). The gas phase also contained CO₂, and therefore extra CO₂ could be incorporated in addition to the released CO₂. Of the injected gas, CO₂ was 12% of the total, and the medium contained NaHCO₃ to supply CO₂. First, we tested the effect of H₂ on the wild-type strain. Acetate was produced as the end product and the carbon molar yield improved from 0.74 to 0.82, as expected (Figures 2A,B). The optical density increased similarly while fructose was consumed, and decreased after the complete consumption of fructose in both conditions, indicating no significant effect on the growth (Figure 2C). We then tested an ethanol-producing strain, Mt-aldh, in which the aldh gene encoding aldehyde dehydrogenase was expressed by a constitutive promoter. The main product was acetate, accompanied by a small amount of ethanol. This trend was similar in the H₂-supplied culture, and yield improvement was only observed for acetate production from 0.78 to 0.88 (Figure 2A). The change in ethanol production was not significant (0.02 and 0.03; Figure 2B). No significant effect on the growth was observed (Figure 2D). Although we expected to enhance the yield of reduced products, we reasoned that the abundant activity of the acetate production pathway in the Mt-aldh strain decreased the effect of H₂. We then tested another ethanol-producing strain, Mt-ΔpduL2::aldh, which showed less acetate production due to deletion of one of the two genes (pduL2) encoding phosphoacetyl transferase in the acetate production pathway. Despite the dominant production of ethanol over acetate, product yield enhancement was observed with only acetate from 0.19 to 0.25 (Figures 2A,B). The carbon molar yield for ethanol did not change (0.37). The rest was released as CO₂. The strain grew similarly in both conditions (Figure 2E). The effect of H₂ was in contrast to the results of a previous study, in which the supplementation of reducing power with H₂ was reflected in the production of more-reduced chemicals (Jones et al., 2016).

We tested the Mt-ΔpduL1ΔpduL2::aldh strain, which produces almost exclusively ethanol, because of the complete knockout of both two genes (pduL1 and pduL2) encoding phosphoacetyl transferase. In this case, we observed an enhancement in the ethanol yield. The carbon molar yield for ethanol was increased from 0.47 to 0.54 (Figure 2B). Approximately 50% of the carbon from fructose was released as CO₂ in the absence of H₂, and H₂ supplementation supported capture and conversion of CO₂. However, the supplemented fructose was not completely consumed in the H₂-supplied condition after 60 h of cultivation, whereas the same strain in the H₂-unsupplied condition, or the other strains in both conditions, consumed fructose completely (Figures 2C–F). Only 40.2% of the supplemented fructose was consumed in the H₂-supplied condition, and the growth was significantly reduced in a correlated manner (Figure 2F). The volumetric amount of ethanol was also significantly less in the H₂-supplied condition, showing only 45.5% against the no-H₂ condition (17.6 mM in the absence of H₂ and 8.0 mM in the presence of H₂), reflecting the amount of consumed fructose. Therefore, although the Mt-ΔpduL1ΔpduL2::aldh strain showed increased carbon molar yield for ethanol under H₂-supplemented mixotrophic conditions, growth inhibition emerged as an unexpected bottleneck.

We also tested the addition of different amounts of H₂ to the Mt-ΔpduL1ΔpduL2::aldh strain. In addition to the condition of partial pressure 0.08 MPa, we tested 0.04, 0.02, and 0.01 MPa, because the pressure of H₂ is correlated with dissolved H₂ in the culture medium. All cases with H₂ at any concentration showed growth inhibition effects. Interestingly, the effect of growth inhibition was dose-dependent, showing more potent inhibition by a higher concentration of H₂ in the culture medium, rather than by a certain threshold. This tendency was clear when the growth rate was plotted against the H₂ pressure, showing a linear correlation (Figure 2H).

To investigate the H₂-dependent growth inhibition mechanism of the ethanol-producing strain, we assessed intracellular metabolism by metabolome analysis. We used H₂ at 0.02 MPa, because a high dose of H₂ (such as 0.08 MPa) inhibited the growth almost completely, and hence might have significant effects on multiple metabolic pathways. We sampled the cells from the exponential phase and analyzed intracellular metabolites using GC–MS and LC–MS following the sample preparation method we developed. We succeeded in quantifying 19 intracellular metabolites, including seven cofactors (Figures 3A,B).

We then compared the Mt-ΔpduL1ΔpduL2::aldh strain and the wild-type strain, with or without H₂ supplementation. Among all analyzed metabolites, NADH levels were striking in the Mt-ΔpduL1ΔpduL2::aldh strain under H₂-supplied conditions. The NADH level increased by approximately four times compared to that in the no H₂ condition, whereas H₂ supplementation did not affect the NADH level in the wild-type strain. There was no such significant difference specific to H₂ supplementation in the other metabolites in the Mt-ΔpduL1ΔpduL2::aldh strain or metabolites in the wild-type strain (Figure 3B).

When the overall metabolite profiles were compared in the Mt-ΔpduL1ΔpduL2::aldh and wild-type strains in the no H₂ condition, the levels of glucose-6-phosphate, fructose-6-phosphate, fructose-1,6-bisphosphate, 3-phosphoglyceric acid, glutamate, and ATP were lower, and the levels of acetyl-CoA, AMP, and NADPH were higher in the Mt-ΔpduL1ΔpduL2::aldh strain. The ATP level was lowered in the Mt-ΔpduL1ΔpduL2::aldh due to the knockout of acetate production coupled with substrate-level phosphorylation, but the ATP level was enough to maintain the growth (Figure 2F). In contrast, the AMP level was higher in the Mt-ΔpduL1ΔpduL2::aldh, and this may be related with the change of ATP level. The higher level of acetyl-CoA probably reflects a difference of conversion rate of acetyl-CoA to ethanol and acetate. The higher level of NADPH in the Mt-ΔpduL1ΔpduL2::aldh strain suggests that the redox balance in this strain was altered by metabolic engineering. Metabolome analysis indicated that the Mt-ΔpduL1ΔpduL2::aldh strain suffered redox imbalances due to both metabolic engineering and H₂ supplementation.

Metabolomic analysis suggested a strong relationship between growth inhibition by H₂ and increased levels of intracellular NADH. In M. thermoacetica, NAD⁺ is reduced to NADH by the electron-bifurcating hydrogenase HydABC complex (Wang et al., 2013). The HydABC complex reduces NAD⁺ and ferredoxin using electrons from H₂. This reaction is reversible and produces H₂ from NADH and reduced ferredoxin in vitro. Therefore, we measured the amount of H₂ in the headspace of culture vials (Figure 4). The amount of H₂ in the headspace was traced in conditions with or without H₂ supplementation, and the wild-type strain and the Mt-ΔpduL1ΔpduL2::aldh strain were compared. H₂ was supplied at 0.02 MPa of a partial pressure, in addition to fructose as a carbohydrate substrate, which was the same condition for our metabolome analysis. There was almost no H₂ production by the wild-type strain, and the supplied H₂ was consumed over time (Figure 4A). In contrast, the Mt-ΔpduL1ΔpduL2::aldh strain apparently did not consume H₂ under the same conditions (Figure 4B). Moreover, when H₂ was not supplied, the H₂ level increased in the culture vial of the Mt-ΔpduL1ΔpduL2::aldh strain, in contrast to that in the wild-type strain. H₂ production is usually attributed to the disposal of excess electrons from metabolism. In this case, this was most likely due to the increased level of NADH. However, H₂ formation would require a sufficiently low level of H₂ as the product. Therefore, the Mt-ΔpduL1ΔpduL2::aldh strain would have produced H₂ for the clearance of the excess electrons from catabolizing fructose, and H₂ supplementation would inhibit the H₂ production, causing the growth inhibition due to the redox imbalance.

The results of the H₂ measurement strongly indicated that the Mt-ΔpduL1ΔpduL2::aldh strain produced H₂ using excess reducing equivalents and balanced the intracellular redox. Our metabolome analysis showed imbalanced redox in the Mt-ΔpduL1ΔpduL2::aldh strain, manifested as an increased level of intracellular NADH. If a high level of intracellular NADH is the direct cause of growth inhibition, the Mt-ΔpduL1ΔpduL2::aldh strain should recover its growth in the presence of H₂ by lowering the intracellular NADH level. M. thermoacetica uses dimethyl sulfoxide (DMSO) as an electron acceptor, and the reported cases of bacterial DMSO reduction are NADH-dependent reactions (Zinder and Brock, 1978; De Bont et al., 1981; Drake and Daniel, 2004; Takemura et al., 2021b; Rosenbaum et al., 2022). We attempted to oxidize intracellular NADH via DMSO reduction by supplementing the culture medium with DMSO. We set up a culture of the Mt-ΔpduL1ΔpduL2::aldh strain with H₂ supplied at 0.08 MPa of partial pressure in addition to fructose, which showed strong growth inhibition (Figures 2F–H). At 24 h, the culture was supplied with 10 mM DMSO. The Mt-ΔpduL1ΔpduL2::aldh strain showed very slow growth with H₂ supplementation, but the growth rate dramatically increased upon DMSO supplementation (Figure 5A). The fructose supplied was completely consumed after 55 h (Figure 5B). The supplied DMSO was readily consumed within 20 h after the DMSO addition, consistent with the recovery of growth and fructose consumption (Figure 5C). The addition of DMSO also improved total ethanol production, whereas acetate production remained minor (Figures 5D,E).

We analyzed the effect of DMSO on the intracellular metabolome in the presence of H₂ (Figure 3B). We used the same culture conditions as in the metabolome analysis (H₂ partial pressure = 0.02 MPa), except for DMSO supplementation, which was provided when the cells entered the exponential phase. As expected, the intracellular level of NADH was lower than that in the H₂ condition. Therefore, growth recovery correlated with intracellular NADH levels.

We found that the ethanol production pathway designed to balance the redox reaction requires tuning the imbalanced redox by producing H₂. This means that our ethanol-producing strains need to be re-engineered to balance the redox reaction to benefit from H₂-supplemented mixotrophy. However, it is possible that artificial modifications in the genome can affect metabolic activities in an unpredictable manner. We previously succeeded in engineering M. thermoacetica to produce acetone (Kato et al., 2021; Table 1). The acetone synthesis pathway does not require any oxidoreductases to convert acetyl-CoA to acetone, which has the same redox balance as that of the native acetate pathway (Figure 1C). Therefore, acetone production has completely the same redox balance as acetate production, and the redox balance should not be affected. On the other hand, the pduL2::acetone strain has the same elements of genetic modification as an ethanol-producing strain, Mt-ΔpduL2::aldh, using a pyrF marker for selection and a constitutive G3PD (glyceraldehyde 3-phosphate dehydrogenase) promoter to express aldh, disrupting the acetate pathway. Therefore, we examined whether the introduction of an oxidoreductase affected the redox balance by testing the H₂-supplemented mixotrophic acetone production of the pduL2::acetone strain.

We set up cultures of the pduL2::acetone strain with fructose as the carbohydrate substrate and tested the effect of H₂ supplementation, as was performed for ethanol-producing strains. H₂ was supplied at 0.08 MPa of partial pressure, the highest dose used for the ethanol-producing strains. The pduL2::acetone strain grew in the presence of H₂ in the same manner as in its absence (Figure 6A). The optical density increased while fructose was consumed, and decreased after the complete consumption of fructose in both cases. Furthermore, the produced acetone and acetate increased with H₂ supplementation (Figure 6B), in contrast to ethanol production. Acetone production was enhanced by 13%, and the total carbon molar yield (the sum of acetate and acetone) was greater than one, indicating that extra CO₂ was converted to these metabolites. CO₂ was externally provided in the headspace gas and from NaHCO₃ in the medium, in addition to CO₂ released from metabolism. Engineering to introduce a non-reductive pathway did not erase the effect of mixotrophy. Therefore, we concluded that introducing oxidoreductase reactions to convert acetyl-CoA to ethanol caused H₂ production by extra electrons and H₂ inhibition.