The 5X In-Fusion^® HD Enzyme Premix was purchased from Takara Bio (Kusatsu, Japan). Strep-Tactin XT 4 Flow high-capacity resin was obtained from IBA Life Sciences (Göttingen, Germany). Disposable PD-10 desalting columns were purchased from Cytiva (Marlborough, MA, United States). Vivaspin 6 centrifugal concentrators with a molecular weight cutoff (MWCO) of 100 kDa were purchased from Sartorius (Göttingen, Germany). A polypropylene column (1 ml) was purchased from Qiagen (Hilden, Germany). The Ziptip C₁₈ resin was purchased from Millipore (Burlington, MA, United States). All other chemical reagents were purchased from Sigma-Aldrich (St. Louis, MO, United States) unless otherwise stated.

To construct the strep-tag II-fused RcFDH expression plasmid, pTrcHis-RcFDH (Choi et al., 2018) was used as a template. Infusion cloning was performed to substitute the hexahistidine-tag for strep tag II. pTrcHis-RcFDH was amplified by PCR with the in-fusion primer (FW: 5′-GCC​ACC​CGC​AGT​TCG​AAA​AAG​GTA​TGG​CTA​GCA​TGA​CGG​ATA​CC-3′, RV: 5′-CGA​ACT​GCG​GGT​GGC​TCC​AAG​AAC​CCC​CCA​TGG​TTT​ATT​CCT​CC-3′). The PCR product was mixed with 5X In-Fusion HD Enzyme Premix to generate pTrcHis-strep-RcFDH. The E. coli MC1061 strain was transformed with pTrcHis-Strep-RcFDH, and the R. eutropha HF210 [pGE771] strain (Lauterbach and Lenz, 2013) was used as the ReSH-expressing strain.

For the expression of ReSH and RcFDH, a 7 L scale fermenter was used. Previously, Lenz described the heterotrophic cultivation of R. eutropha derivatives (Lenz et al., 2018). A 10X H16 buffer (pH 7.0) consisting of 250 mM Na₂HPO₄ and 110 mM KH₂PO₄ was used as the medium. For a 1 L of fructose-ammonium (FN) medium, 100 ml of 10X H16 buffer was mixed with 850 ml of sterilized water (additional 13% (w/v) of Bacto agar in case of solid agar plates) and autoclaved. Next, 10 ml of 20% (w/v) NH₄Cl, 1 ml each of 20% (w/v) NH₄Cl, 20% (w/v) MgSO₄*7H₂O, 1% (w/v) CaCl₂*H₂O, 0.5% (w/v) FeCl₃*6H₂O (in 0.1 N HCl), 1 mM NiCl₂, and 1.25 ml of 40% (w/v) D-fructose were mixed and filled up to 1000 ml with sterile H₂O. A single colony of R. eutropha was pre-cultured in 50 ml of FN medium containing 10 μg ml⁻¹ tetracyclin until the OD436nm reached 1. For the main culture, 5 L of modified fructose-glycerol-ammonium (FGN_mod) with 0.05% (w/v) glycerol, 5 ml of SL6 trace element solution (Lenz et al., 2018), and 5 ml of 1 mM ZnCl₂ (added to the FN medium containing 10 μg/ml tetracycline) were prepared in the fermenter. The pre-culture was inoculated into the FGN_mod medium and subjected to 300 rpm shaking and 1 VVM aeration at 30°C. The pH range was maintained between 6.9 to 7.0 through automatic injection of 1 N NaOH. After 24 h, 5 ml of 1 mM NiCl₂ was added. When the OD at 436 nm reached 9–11, the cells were harvested by centrifugation at 6,000 × g for 10 min before storage at −80°C.

For RcFDH expression, a single-cell colony was pre-cultured in Luria-Bertani (LB) medium containing 150 μg ml⁻¹ ampicillin for 12 h at 37°C. For the main culture, 5 L of LB medium containing 150 μg ml⁻¹ ampicillin, 1 mM sodium molybdate, and 20 μM isopropyl β-D-1-thiogalactopyranoside was prepared in the fermenter. The pre-culture was inoculated into the LB medium and subjected to 100 rpm shaking and 0.1 VVM aeration at 30°C. After 24 h, the cells were harvested by centrifugation at 6,000 × g for 10 min before storage at -80°C.

To purify ReSH and RcFDH, cell pellets were resuspended in 50 mM potassium phosphate buffer (pH 7.0) (Kpi buffer) containing 1 mg/ml lysozyme to a concentration of 1 g/10 ml. The resuspended cells were lysed by sonication (amplitude 28%, on/off 2 s/4 s) for 1 h. Insoluble cell debris was removed by centrifugation at 13,000 × g for 30 min. Strep-Tactin XT 4Flow high-capacity resin (2 ml) was mixed with the clear supernatants and incubated at 4°C for 30 min. The resin was washed with Kpi buffer containing 300 mM potassium chloride on a gravity-flow polypropylene column to remove any impurities. The proteins were eluted with 3 ml of Kpi buffer containing 50 mM biotin and buffer-exchanged with Kpi buffer containing 10 mM potassium nitrate using a PD-10 column. Protein purity was verified by SDS-PAGE (Figure 2). The concentrations of purified proteins were determined by measuring their absorbance at 280 nm using a microplate reader (Synergy, BioTek, Winooski, VT, United States), as previously reported for other proteins (Kim et al., 2019; 2021; Bak et al., 2020). The extinction coefficients of ReSH and RcFDH were calculated to be 165,710 and 350,000 M⁻¹⋅cm⁻¹, respectively, based on their amino acid sequences.

Proteins in buffer were desalted using Ziptip C₁₈ according to the manufacturer’s protocol. Purified ReSH and RcFDH were mixed in a 1:1 (v/v) ratio with a sinapinic acid-saturated matrix solution consisting of 30% acetonitrile, 0.1% trifluoroacetic acid (TFA), and 70% water (v/v). The mixtures were subjected to mass characterization by Autoflex speed (Bruker Corporation, Billerica, United States).

The enzyme reaction kinetics of ReSH were measured for the NAD⁺-dependent oxidation of H₂ to H⁺ in the presence or absence of O₂. The sealing cuvette was filled with 900 μL of Kpi buffer containing NAD⁺ and sealed; then, 100% H₂ and a mixed gas consisting of 10% O₂ and 90% N₂ (or 100% N₂ for anaerobic conditions) were injected simultaneously for 30 min at 10 ml/min. ReSH (2 ml, 80 nM) was purged with 10 ml/min N₂ gas bubbling in a 10 ml sealing vial for 30 min to remove O₂ from the air. The reaction was initiated by mixing 100 μL of 80 nM ReSH with a gas-saturated solution in a sealed cuvette. The final concentration of NAD⁺ was varied from 0 to 2 mM.

The enzyme reaction kinetics of RcFDH were measured for NADH-dependent reduction of CO₂ to formate in the presence or absence of O₂. The sealing cuvette was filled with 900 μL of Kpi buffer containing NADH and sealed; then, 100% CO₂ and a mixed gas consisting of 4% O₂ and 96% N₂ (or 100% N₂ for anaerobic conditions) were injected simultaneously for 30 min at 10 ml/min, respectively. RcFDH (2 ml, 2 μM) was purged with 10 ml/min N₂ gas bubbling in a 10 ml sealing vial for 30 min to remove O₂ from the air. The reaction was initiated by mixing 100 μL of 2 μM RcFDH with a gas-saturated solution in a sealing cuvette. The final concentration of NADH was varied from 0 to 1 mM.

All measurements were performed in triplicate based on the change in the absorbance at 365 nm in the cuvette, measured using a T60 UV-Vis spectrophotometer (PG Instruments Ltd., Lutterworth, UK). The change in absorbance over 1 min was plotted using the Michaelis-Menten equation to calculate the kinetic parameters.

For the cascade reaction in the presence or absence of O₂, the gas content was controlled in a 20 ml polytetrafluoroethylene (PTFE) septa sealing vial. The vials were filled with 500 μL of reaction solution containing 3.2 U/mL ReSH, 0.16 U/mL RcFDH, 1 mM NAD⁺, and 0.5 M Kpi buffer and sealed. A needle was inserted into the septa for gas evacuation. Then, 10 ml/min CO₂ and 20 ml/min N₂/O₂ mixed gas were injected for 30 min (the needle did not enter the reaction solution). The O₂ ratios of the mixed gas varied from 0%–2%–4%; therefore, the final concentrations of O₂ were 0, 1, and 2%. The reaction was initiated by a 10 ml/min H₂ gas injection. Formate production was sampled every 20 min during incubation for 1 h, and 10 μL of 6 N H₂SO₄ was added to the 100 μL sample to inactivate the enzymes immediately. Additionally, 240 μL of distilled water was mixed with the sample, and the aggregate enzymes were removed by centrifugation at 13,000 × g. Formate production was quantified by HPLC (1260, Agilent, CA, United States) equipped with a diode-array detector and an Aminex HPX-87H column (BIO-RAD, CA, United States) with a mobile phase of 5 μM H₂SO₄ at a flow rate of 0.6 ml/min. The retention time of formate was 13.010 min. The formate concentration was calculated using a formate calibration curve (Supplementary Figure S1).

ReSH and RcFDH are expressed in R. eutropha and E. coli, respectively. They were purified using affinity resins, as described in the Materials and methods. Five bands of purified ReSH subunits were observed, which matched the expected molecular weights (HoxF, 68,110 Da; HoxH, 54,863 Da; HoxU, 26,173 Da; HoxY, 22,881 Da; HoxI, 18,567 Da) (Figure 2A). Similarly, three bands of purified RcFDH subunits were observed, which were consistent with the expected molecular weights (FdsA, 104,466 Da; FdsB, 52,699 Da; FdsG, 17,304 Da) (Figure 2B). Both enzymes showed high purity. The identity of the purified enzymes was confirmed by MALDI-TOF mass spectrometry. The experimentally determined masses of ReSH subunits were 67,542, 54,492, 26,038, 22,836, and 18,545 m/z, which matched well with the expected masses (68,111, 54,864, 26,174, 22,882, and 18,568 m/z, respectively) with less than 1% deviation (Supplementary Figures S2A–C). The experimentally determined masses of RcFDH subunits were 104,259, 52,385, and 17,136 m/z, which matched well with the expected masses (104,467, 52,700, and 17,305 m/z, respectively) with less than 1% deviation (Supplementary Figures S2D, E). These results showed that the purified ReSH and RcFDH were successfully prepared.

We investigated the enzymatic activities of ReSH and RcFDH in the presence or absence of O₂. The NAD⁺-dependent H₂ oxidation reaction rate by ReSH was measured, and the Michaelis-Menten curve was fitted to calculate the kinetic parameters using Origin 2022 program (Figure 3A). Both k_cat and K_m values of ReSH showed an insignificant difference under the 0% and 5% O₂ conditions (Table 1). Similarly, The NADH-dependent CO₂ reduction reaction rate by RcFDH was measured, and the Michaelis-Menten curve was fitted to calculate the kinetic parameters (Figure 3B). Likewise, k_cat and K_m values of RcFDH showed an insignificant difference between the 0% and 2% O₂ conditions (Table 2). These results show that purified ReSH and RcFDH retained the enzymatic activity at least under less than 2% O₂.

We determined the NAD⁺, ReSH, and RcFDH contents for the cascade reaction of ReSH and RcFDH. Owing to the relatively low k_cat value (Tables 1, 2), the rate-determining step was the CO₂ reduction by RcFDH. Because the reaction rate of RcFDH was saturated at NADH concentrations above 1 mM (Figure 3B), the NAD⁺ concentration was determined to be 1 mM. For the continuous CO₂ reduction by RcFDH, the concentration of ReSH was determined to maintain a state in which all NAD⁺ was reduced to NADH. The concentration of RcFDH was fixed at 0.08 U/mL and the amount of ReSH was adjusted to 0, 0.08, 0.8, and 1.6 U/mL (U/mL ratio of ReSH:RcFDH = 0:1, 1:1, 5:1, 10:1, 20:1). Reaction solutions were placed in a 20 ml sealing vial, and 10 ml/min CO₂ and 10 ml/min H₂ were injected for 1 h simultaneously, after which formate was measured (Supplementary Figure S3). Formate production was not observed in the reaction solution without ReSH. In contrast, substantial formate production was observed in the reaction solution with the three components (ReSH, RcFDH, and NAD⁺). Formate production was saturated above a 5:1 ratio. At higher ReSH concentrations, NAD⁺ was immediately converted to NADH through H₂ oxidation. This result set the cascade reaction content to 1 mM NAD⁺, and the U/mL ratio of ReSH:RcFDH = 20:1.

We demonstrated H₂ and CO₂ conversion into formate under 0%–2% O₂ conditions. ReSH, RcFDH, and 1 mM NAD⁺ were mixed and placed in a 20 ml sealing vial. Changes in the concentrations of NADH and formate over time were investigated when O₂ (at a controlled concentration), H₂, and CO₂ were simultaneously and continuously injected into the vial. During the injection of the gases, under all O₂ conditions from 0% to 2%, NAD⁺ was reduced to NADH and maintained at 1 mM by H₂ oxidation of ReSH (Figure 4A). Furthermore, the formate concentration increased continuously (Figure 4B) owing to the CO₂ reduction of RcFDH. Approximately 230 μM of formate was produced after 1 h, which showed a statistically insignificant difference at 0, 1, or 2% O₂ conditions (p > 0.05). In order to investigate the O₂-tolerant limit of the system, we tested the formate production in a higher concentration of O₂ (Supplementary Figure S4). We observed a substantial reduction in formate production at 5% O₂ compared to 0%. Therefore, in the specific enzyme systems we chose, the O₂-tolerance limit was between 2% and 5%. The O₂-tolerance of both H₂ase and FDH is attributed to the reduction of O₂ bound to the active site of enzymes, leading to the reactivation of active site. Therefore, we speculated that the substantial loss of enzymatic activities at 5% O₂ results from that O₂ binding to the active site is more favorable than O₂ reduction at the active site. These results demonstrate, as hypothesized, the plausibility of a cascade reaction using ReSH and RcFDH, even in the presence of O_2. Of course, greater O₂-tolerance limit would be beneficial in developing practical processes. We speculate that there are ways to increase the O₂-tolerance limit of enzymes. First, the enzyme concentration can be adjusted to increase O₂-tolerance limit. O₂-tolerance is likely attributed to the reduction mechanism of O₂ to either H₂O or H₂O₂. In this case, O₂ is a co-substrate of these enzymes. Therefore, if the concentrations of enzymes were sufficiently high, the enzymes would quickly reduce O₂, leading to the increased O₂-tolerance limit. Another possible approach to increase O₂-tolerance is engineering enzyme. Recently it was reported that the simple point mutations in the gas tunnel region of O₂-sensitive CO dehydrogenase greatly increased the O₂-tolerance limit (Kim et al., 2022). We speculate that such enzyme engineering strategy can be applied to ReSH and RcFDH to increase O₂-tolerance limit.