Efficient CO2-Reducing Activity of NAD-Dependent Formate Dehydrogenase from Thiobacillus sp. KNK65MA for Formate Production from CO2 Gas

NAD-dependent formate dehydrogenase (FDH) from Candida boidinii (CbFDH) has been widely used in various CO2-reduction systems but its practical applications are often impeded due to low CO2-reducing activity. In this study, we demonstrated superior CO2-reducing properties of FDH from Thiobacillus sp. KNK65MA (TsFDH) for production of formate from CO2 gas. To discover more efficient CO2-reducing FDHs than a reference enzyme, i.e. CbFDH, five FDHs were selected with biochemical properties and then, their CO2-reducing activities were evaluated. All FDHs including CbFDH showed better CO2-reducing activities at acidic pHs than at neutral pHs and four FDHs were more active than CbFDH in the CO2 reduction reaction. In particular, the FDH from Thiobacillus sp. KNK65MA (TsFDH) exhibited the highest CO2-reducing activity and had a dramatic preference for the reduction reaction, i.e., a 84.2-fold higher ratio of CO2 reduction to formate oxidation in catalytic efficiency (k cat/K B) compared to CbFDH. Formate was produced from CO2 gas using TsFDH and CbFDH, and TsFDH showed a 5.8-fold higher formate production rate than CbFDH. A sequence and structural comparison showed that FDHs with relatively high CO2-reducing activities had elongated N- and C-terminal loops. The experimental results demonstrate that TsFDH can be an alternative to CbFDH as a biocatalyst in CO2 reduction systems.


Introduction
Reducing the atmospheric CO 2 level has received a great deal of attention recently as an approach to combat global warming and fossil-fuel shortages, but this process remains challenging. Biological CO 2 fixation is one of the most important approaches to solving these problems. Enzymatic CO 2 reduction has been examined extensively as a promising approach to greenhouse gas fixation and the production of renewable fuels and chemicals [1][2][3]. The enzymatic reduction of CO 2 using FDHs has been widely studied for the production of valuable chemicals, such as formic acid and methanol [4,5]. Formic acid is considered to be a promising replacement for methanol in miniature fuel cells [6]. Formic acid has been produced by the hydrolysis of methyl formate, which is synthesized via methanol carbonylation in commercial processes. Therefore, it would be environmentally attractive to prepare formic acid from CO 2 gas by enzymatic biotransformation.
Many efforts have been made to develop CO 2 -reducing chemical catalysts, and recent research on chemical catalysts has led to improved rates for CO 2 reduction [7,8]. However, chemical catalysts require harsh reaction conditions and/or expensive metals, such as ruthenium, rhodium, and iridium [9][10][11]. In contrast to a chemical CO 2 reduction, CO 2 can be reduced by enzymes under mild conditions. There are few biocatalysts capable of biological CO 2 fixation, e.g. pyruvate decarboxylase (EC 4.1.1.1), carbonic anhydrase (EC 4.2.1.1), and FDH (EC 1.2.1.2). Pyruvate decarboxylase can catalyze the reversible conversion of pyruvate into CO 2 and acetaldehyde and thus requires equimolar acetaldehyde for the conversion of CO 2 into pyruvate [3]. It should be noted that carbonic anhydrase can catalyze the rapid interconversion of CO 2 and bicarbonate but this is not a real CO 2 reduction reaction but a CO 2 hydration reaction [12,13]. However, FDH can reduce CO 2 to formate without any other organic chemicals, and formate can be sequentially reduced to formaldehyde and methanol by coupling aldehyde dehydrogenase and alcohol dehydrogenase reactions [4]. Therefore, FDH has been widely adopted in CO 2 reduction reactions [4,[14][15][16]. FDH can be divided into two groups, NAD-independent or NADdependent. NAD-independent FDHs have a high CO 2 -reducing activity but include extremely oxygen-labile catalytic components, such as metal ions (tungsten or molybdenum), iron-sulfur clusters, and selenocysteine, making these FDHs unsuitable for industrial applications [17][18][19][20]. Recently, NAD-dependent FDHs have been utilized in CO 2 reduction systems as an alternative to NADindependent FDHs. In particular, CbFDH is commercially available and has been widely adopted as a CO 2 -reducing biocatalyst in electrochemical, photochemical, and enzymatic reactions [4,14,16] as well as a NADH-regenerating biocatalyst in enzyme-coupled reaction systems [21][22][23][24]. However, the CO 2reducing activity of CbFDH is still very low for practical applications, and thus it is necessary to discover more efficient FDHs than CbFDH.
In this study, we report superior CO 2 -reducing performance of TsFDH. We selected five FDHs based on their biochemical properties, e.g. acidic optimum pH, specific activity, and stability, and investigated their feasibility as CO 2 -reducing biocatalysts. Enzyme activities in formate oxidation and CO 2 reduction were measured, and the ternary complex model was applied to understand the characteristics of FDHs. Finally, the concentration of formate produced form CO 2 gas using TsFDH and CbFDH was compared. Based on these experimental results, TsFDH can be a good substitute for CbFDH as an efficient CO 2 -reducing biocatalyst.

Materials
The fdh genes used in this study were synthesized with an additional C-terminal hexa-histidine sequence by GenScript (USA). Restriction enzymes were purchased from Takara (Japan). Pfu DNA polymerase and ligase were purchased from New England Biolabs (USA). The pET-23b(+) vector and competent cell E. coli DH5a/BL21 (lDE3) were purchased from Novagen (USA). Ni-NTA resin was purchased from Qiagen (USA). SYPRO Orange dye was purchased from Invitrogen (USA). CbFDH and all chemicals used in enzyme reactions were purchased from Sigma-Aldrich (USA).
Cloning, expression, and purification of FDHs. The fdh genes containing a C-terminal hexa-histidine sequence in the pUC57 vector were PCR-amplified using primers (Table S1). The amplified fragments were digested by Nhe I and EcoR I and then cloned into a pET-23b(+) vector. The sequence of recombinant plasmids was confirmed by DNA sequencing (Cosmogenetech, Korea). The recombinant plasmids were chemically transformed into competent E. coli BL21 (lDE3). Expression and purification of the FDHs were performed as previously described [25]. Ni-NTA resin was used to purify the FDHs with a C-terminal hexahistidine tag. To obtain pure enzymes, the recombinant FDHs were washed with 50 ml of wash buffer (50 mM NaH 2 PO 4 pH 7.0, 300 mM NaCl, 40 mM imidazole), and the purity of the FDHs was determined by 12% SDS-PAGE.
Determination of conformational stability. The conformational stability of the FDHs was determined by differential scanning fluorimetry (DSF) using a real-time quantitative PCR thermal cycler (LightCycler 480, Roche, USA). The DSF method can measure the temperatures at the midpoint of a protein's melting transition (T m ) by monitoring their thermal unfolding. SYPRO Orange (20x) and 10 to 30 mM of purified FDHs were used in 96-well qPCR microplates. The T m values were determined by heating from 20 to 95uC at a 1uC/min scan rate as previously described [26].
Activity assay of FDHs. The FDH activity was determined by monitoring the absorbance changes at 340 nm during the redox reactions catalyzed by the FDHs at 25uC and different pH values (5.5,7.0). One unit of oxidation activity was defined as the amount of enzyme required to produce 1 mmol of NADH per minute under standard conditions. The oxidation of formate was conducted using an assay solution (2 ml of a 100 mM sodium phosphate buffer) containing 20 mg of FDH, 200 mM sodium formate, and 2 mM NAD + . One unit of reduction activity was defined as the amount of enzyme required to consume 1 mmol of NADH per minute under standard conditions. The reduction of CO 2 was conducted using an assay solution (2 ml of a 100 mM sodium phosphate buffer) containing 3.0 mg of FDH, 50 mM NaHCO 3 , and 0.15 mM NADH. In the case of the CO 2 reduction reaction, NaHCO 3 was used as a substrate to supply CO 2 because the concentration of gaseous CO 2 cannot be accurately determined. Therefore, the substrate concentration for CO 2 reduction was represented as that of NaHCO 3 . A relatively large amount of enzyme was used for CO 2 reduction due to the very low CO 2 -reducing activity of CsFDH. All activities were calculated by subtracting the decomposition rate of NADH in the absence of the enzyme.
Determination of kinetic parameters. The saturation concentration of NADH required for CO 2 reduction was too high to adopt the steady-state model and thus the ternary complex model [27] was applied with low concentrations of substrate and cofactors. The kinetic parameters were determined according to the sequential mechanism model proposed by Cleland [28]. This mechanism was described as follows (1): where V max is the maximum reaction rate, K A and K B are the Michaelis constants for NADH (reduction)/NAD + (oxidation) and NaHCO 3 (reduction)/formate (oxidation), respectively, in the presence of saturating concentrations of the other substrate. K iA is the dissociation constant for the enzyme-cofactor complex. The model was fitted to the experimental data using SigmaPlot (Version 10.0, Systat Software Inc., USA). The kinetic parameters of the FDH-catalyzed oxidation and reduction reactions were determined in 100 mM sodium phosphate buffer (pH 7.0) with a 1.1 ml reaction volume at 25uC. To perform the formate oxidation reaction, the content of FDHs was fixed at 5 mg, while the concentration of NAD + varied from 0.1 to 0.3 mM and that of sodium formate varied from 10 to 50 mM. The reduction reaction of CO 2 was achieved with 10 and 100 mg of TsFDH and CbFDH, respectively, and the ranges of concentrations for NADH and sodium bicarbonate were same as those of NAD + and sodium formate for the formate oxidation reaction. The initial rates of the reactions were determined by measuring the change in NADH.
Formate production from CO2 gas. Sodium phosphate buffer (100 mM, pH 7.0) was purged with CO 2 for 1 h and adjusted to pH 7.0 with NaOH. Formate production reactions were performed by 4.4 mM FDHs (1.0 mg of TsFDH or 0.9 mg of CbFDH) with 10 mM NADH. The total volume of reaction solution was 5 ml. The reaction solution was mixed with a magnetic stir bar at 200 rpm and was purged continuously with CO 2 at a flow rate of 50 ml/min during the CO 2 reduction reactions. The amount of formate produced was determined by high-pressure liquid chromatography (HPLC) equipped with an Aminex HPX 87-H ion exclusion column (30067.8 mm) and a refractive index detector (RID). The temperature of the column and the RID detector was set to 35uC. H 2 SO 4 solution (5 mM) was used as the mobile phase at a flow rate of 0.6 ml/min. Forty microliters of the sample was injected into HPLC for analysis.
The conformational stability of the FDHs was determined by evaluating their thermal unfolding using the DSF method as previously described [25], and all FDHs were sufficiently stable (T m = 42.3,61.8uC) under ambient conditions (Fig. S1).

Enzyme activities for formate oxidation and CO 2 reduction
The activities of the recombinant FDHs, including CbFDH, in the oxidation and reduction reactions are shown in Fig. 2. All enzyme activities were determined by measuring the initial rates of reaction. All FDHs except for CsFDH had higher enzyme activities in the formate oxidation reaction than CbFDH. CsFDH activity was 1.0,1.3 U/mg enzyme at all tested pH ranges ( Fig. 2A). AaFDH exhibited the highest activity, and its enzyme activity was 5.2-fold (20.0 U/mg enzyme) and 3.0-fold (18.3 U/ mg enzyme) higher than those of CbFDH at pH 5.5 and 7.0, respectively. The FDHs did not exhibit a drastic change in formate oxidation activity at the tested pH ranges. The CO 2 -reducing activities of FDHs are shown in Fig. 2B. Compared to the formate oxidation activity, all FDHs exhibited low enzyme activities (0.5,12.2 mU/mg) in the CO 2 reduction reaction and had an acidic optimum pH (pH 5.5 or 6.0). Four screened FDHs (AaFDH, MsFDH, PsFDH, and TsFDH) were more active over the pH range of 5.5 to 7.0 than CbFDH, and TsFDH had the highest CO 2 -reducing activity. Its activities were at least 7.3-fold higher (12.2, 9.5, 8.7, and 4.0 mU/mg enzyme) than those of CbFDH (1.6, 1.3, 0.7, and 0.5 mU/mg enzyme) at pH 5.5, 6.0, 6.5, and 7.0, respectively.

Enzyme kinetics
The kinetic parameters of TsFDH and CbFDH were determined at pH 7.0 because NADH is slowly degraded at acidic pHs [33], resulting in unreliable kinetic values at acidic pHs. Initial rates were measured with respect to cofactors at various fixed concentrations of substrates. Kinetic parameters for the forward and reverse reaction were obtained by fitting the initial rates to equation (1) with none-linear regression (Table 1). Regression coefficients (r 2 ) for all reactions were over 0.937 (Fig. 3). In the formate oxidation reaction, TsFDH showed a lower substrate binding affinity (K B , 16.24 mM) and a higher turnover number (k cat , 1.769/s) than CbFDH (8.55 mM and 1.081/s, respectively), and both enzymes exhibited similar catalytic efficiencies (k cat /K B ). In the CO 2 reduction reaction, however, TsFDH had a 3.4-fold higher substrate-binding affinity (K B , 9.23 mM), a 21.2-fold higher turnover number (k cat , 0.318/s), and a 85-fold higher catalytic efficiency (k cat /K B , 0.034/mM?s) than CbFDH (31.28 mM, 0.015/s, and 0.0004/mM?s, respectively). TsFDH exhibited a dramatic preference for the reduction reaction with a 84.2-fold higher catalytic efficiency ratio of CO 2 reduction to formate oxidation than that of CbFDH. The recombinant  CbFDH expressed in E. coli and commercial CbsFDH (Sigma-Aldrich) showed similar catalytic efficiencies for the forward and reverse reaction (data not shown) and thus only the data from the recombinant CbFDH was reported in this paper.

Formate production through enzymatic CO 2 reduction
Formate was produced from CO 2 gas using TsFDH and CbFDH. The amount of formate produced through the enzymatic CO 2 reduction is shown in Fig. 4. TsFDH and CbFDH produced 0.74 mM and 0.13 mM formate in 120 min, respectively. The formate production rate of TsFDH and CbFDH was linear, with a constant rate of 0.023/s and 0.004/s, respectively. TsFDH showed a 5.8-fold higher formate production rate than CbFDH, which is consistent with the results shown in Fig. 2B.

Discussion
FDHs can catalyze the conversion of CO 2 and formate and thus are of great interest as CO 2 -reducing biocatalysts for CO 2 sequestration and for the production of formate as a source of fuels and commodity chemicals. NAD-independent FDHs can drive the CO 2 reduction reaction with electrons supplied from an electrode and artificial electron mediators, such as methyl viologen, exhibiting very high CO 2 -reducing catalytic efficiency [20,34]. Despite this advantage, the use of NAD-independent FDHs in CO 2 reduction systems does not appear to be practical due to the requirement for complicated catalytic components, such as molybdopterin cofactor, iron-sulfur clusters, and selenocysteine, in addition to their oxygenlabile activity, which results in insoluble and inactive expression in E. coli [35][36][37]. Recently, K. Schuchmann and V. Müller reported that a hydrogen-dependent carbon dioxide reductase (HDCR) from Acetobacterium woodii can catalyze reduction of CO 2 to formate with very high activity [38]. However, it is also very unstable under aerobic conditions as it has the catalytic components (Table S2). In contrast to NAD-independent FDHs and HDCR, NAD-dependent FDHs are oxygen-stable and can be highly expressed in E. coli as demonstrated in this study, but their practical applications in CO 2 -reduction systems are still limited due to their low CO 2 -reducing activities.
In this study, we attempted to identify FDHs that are superior to a conventional CO 2 -reducing biocatalyst, i.e., CbFDH. FDHs suitable for CO 2 reduction were screened from BRENDA (BRaunschweig ENzyme Database, http://www.brendaenzymes.org) [39] based on their optimum pH. The catalytic mechanism of formate oxidation by NAD-dependent FDHs has been demonstrated to involve direct hydride transfer from formate to the C4 atom of the nicotine amide ring of NAD + [40,41]. However, it remains unclear whether NAD-dependent FDHs use a proton-relay system in the CO 2 reduction reaction. The abundance of protons would be favorable for the reduction of many chemicals [42,43]. Moreover, Peacock and Boulter reported that FDH from Phaseolus aureus (PaFDH) had 19.7-fold higher CO 2 -reducing activity at pH 6.3 than at pH 8.0 (710 pmol/min vs. 36 pmol/min) with approximately equivalent concentrations of enzyme and substrate and showed a 19.6-fold lower ratio of the rates of the forward (formate oxidation) and reverse (CO 2 reduction) reaction (2,300 at pH 6.3 vs. 45,000 at pH 8.0) [27]. These results imply that FDHs with an acidic optimum pH would be more efficient for CO 2 reduction than FDHs with neutral or alkaline optimum pH. In the present study, the preliminary experiment also revealed that CbFDH has a higher CO 2 -reducing activity at an acidic pH (1.6 mU/mg enzyme at pH 5.5) than at a neural pH (0.5 mU/mg enzyme at pH 7.0), whereas it exhibited higher formate-oxidation activity at a neutral pH (6.1 U/mg enzyme at pH 7.0) than at an acidic pH (3.8 U/mg enzyme at pH 5.5) (Fig. 2). In addition, it has been reported that CO 2 , bicarbonate (HCO 3 2 ), and carbonate (CO 3 22 ) are dominant at acidic, neutral, and alkaline pHs, respectively [44]. As mentioned above, CO 2 reduction at an acidic pH has apparent advantages over reduction under neutral or alkaline pH conditions, including the presence of abundant protons, a high CO 2 fraction [42], and higher CO 2 -reducing activity of FDHs. Therefore, we selected five NAD-dependent FDHs (AaFDH, CsFDH, MsFDH, PsFDH, and TsFDH) with an acidic optimum pH and comparable activity and stability to those of the reference enzyme CbFDH. All FDHs were successfully expressed and purified with Ni-NTA resin (Fig. 1). The five FDHs characterized in this study showed an acidic optimum pH for formate oxidation and CO 2 reduction and moderate stability under ambient conditions (Table S3).
As shown in Fig. 2, AaFDH had the highest enzyme activity in the oxidation reaction, which was at least 3-fold higher than that of CbFDH. High formate oxidation activity and the DSF results (Table S3) indicate that AaFDH could be used as an alternative NADH-regenerating enzyme to CbFDH over a broad pH range. In the reduction reaction, all enzymes, including CbFDH, exhibited an acidic pH optimum, indicating that an acidic pH is more favorable than neutral pH for CO 2 reduction. The lowest values for the ratio of the rates of the oxidation and reduction reaction were calculated as 4,600 at pH 6.5, 2,400 at pH 5.5, 1,400 at pH 5.5, 3,600 at pH 6.5, 1,800 at pH 6.5, and 820 at pH 5.5 for AaFDH, CbFDH, CsFDH, MsFDH, PsFDH, and TsFDH (obtained from Fig. 2), which are comparable to the value of PaFDH [27]. All FDHs showed better reduction activities at acidic pH than at neutral pH. TsFDH was the best biocatalyst in terms of CO 2 -reducing activity and the ratio of the rates of the oxidation and reduction reactions.
Previous enzyme kinetics revealed that FDHs have a sequential mechanism in which both substrates bind to the enzyme in a defined or random order before the products are released. Eukaryotic FDHs catalyze formate oxidation by an ordered kinetic mechanism, but bacterial FDHs follow a random mechanism with a rapid equilibrium [27,40]. The Michaelis-Menten plots were obtained (Fig. S2). Saturation of NaHCO 3 and NADH is very difficult for the CO 2 reduction reaction. Saturation of NADH for enzyme-catalyzed CO 2 reduction could not be achieved by high degradation rate of NADH and inhibitory effect Kinetic parameters for the forward (formate oxidation) and reverse (CO 2 reduction) reactions were calculated by fitting the initial rates to equation (1). [b] Formate oxidation (A: NAD + , B: sodium formate).    of degraded compounds. The reaction rate was decreased with increasing NADH concentration of over 0.4 mM [45]. Low solubility of CO 2 in buffer at atmospheric pressure also caused the difficulty of CO 2 saturation for enzyme-catalyzed CO 2 reduction. Therefore, typical Michaelis-Menten saturation plot which shows the convergence of velocity to v max could not be obtained. However, kinetic constants could be obtained on the basis of generally acceptable rapid equilibrium assumption for enzymesubstrate complex. Double reciprocal plots of eukaryotic CbFDH and bacterial TsFDH were linear and gave intersecting patterns in the forward and reverse reaction (Fig. S3), indicating that the kinetic mechanism of both FDHs is sequential. Both FDHs exhibited a similar binding affinity for formate, which is comparable to that of typical NAD-dependent FDHs (1,15 mM) [40]. Both FDHs had a similar catalytic efficiency (k cat /K B ) in the oxidation of formate, but TsFDH showed a dramatic preference for CO 2 reduction due to the 21.2-fold higher turnover number compared to CbFDH. These catalytic properties enable TsFDH to produce formate from CO 2 gas more efficiently than CbFDH without the saturation of the reaction rate (Fig. 4). Conventional CO 2 reduction systems using commercial CbFDH for the production of formate or methanol require in situ regeneration of NADH to drive CO 2 reduction [4]. The formate production rate of TsFDH can be further improved by incorporating a NADH-regeneration system e.g., chemical, electrochemical, photochemical, or enzymatic method [46]. At an acidic pH, despite high TsFDH-catalyzed CO 2 reduction reaction, it is evident that formate productivity will be gradually decreased due to degradation of NADH under acidic conditions [47]. Recently, various NAD analogs including thio-NAD, APAD, PAAD, and NAAD were found to be more efficient and stable than NAD in electrochemical regeneration systems [48]. Thus, these analogs need to be investigated as an alternative cofactor to overcome the instability NADH at an acidic pH. Tishkov and Popov performed structural and multiple sequence alignment of eukaryotic FDHs and bacterial FDHs and found that bacterial FDHs have an addition loop in the N-terminal end, which may have an important role in the discrimination of the kinetic mechanism of the two different groups of FDHs [40]. In this study, it should be noted that four bacterial FDHs showed higher CO 2 -reducing activities than two eukaryotic FDHs. Multiple sequence alignment of six FDHs was conducted using CLUSTAW2 and ESPript 2.2 [49,50] (Fig. 5). The FDHs tested in this study can be classified into two groups (bacterial FDHs, AaFDH, MsFDH, PsFDH, and TsFDH, vs. eukaryotic FDHs, yeast CbFDH and fungal CsFDH) as previously described [40]. Bacterial FDHs had at least 82.5% sequence identity, and CbFDH and CsFDH shared 61.2% sequence identity. Although all the amino acids critical for catalysis or cofactor binding are highly conserved [51] in both FDH groups, bacterial FDHs contained an elongated N-terminal loop and C-terminal loop compared to eukaryotic FDHs (Fig. 5). The N-terminal loop may be involved in the determination of the kinetic mechanism, i.e., an ordered or a random model, and the C-terminal loop may contribute to the better CO 2 -reducing activities of bacterial FDHs than those of eukaryotic FDHs. To understand the amino acid differences of FDHs at molecular level, the structural alignment of TsFDH and CbFDH was performed. The structure of TsFDH was modeled using SWISS-MODEL homology modeling [52]. The holo-crystal structure of NAD-dependent FDH from Pseudomonas sp. 101 (PdFDH, pdb code: 2NAD) was used as a template (92.0% sequence identity with TsFDH) for homology modeling of TsFDH structure because the C-terminal loop, which covers the substrate channel, is only present in the holo-structure: the loop may be largely fluctuated in the apo-structure (pdb code: 2NAC). The Nterminal loop covers a significant part of the enzyme (Fig. 6A), and some amino acids in the loop interact with other amino acids of the subunit or the other chain in a dimeric form (structures not shown). However, the N-terminal loop does not have direct interactions with the substrate binding pocket. It was reported that PdFDH has a narrow substrate channel, and Arg284 on the wall of the substrate channel provides conformational mobility for binding and delivery of substrates [53]. In addition, Arg284 has close contacts with an inhibitor i.e. azide in the ternary complex structure (enzyme-NAD-azide, pdb code: 2NAD). Based on this structure information of PdFDH, it can be speculated that upon sequential binding of cofactor and substrate the C-terminal loop can be formed and then, contribute to conformational changes of the substrate channel including Arg284 for enzyme catalysis. The C-terminal loop, which is not present in the structure of CbFDH, in the modeled structure of TsFDH also covers the substrate binding channel, including Arg284 (Fig. 6B). This structural feature of TsFDH may be associated with CO 2 accessibility or binding to the active site given that the kinetics data revealed that TsFDH exhibited a better K B value than CbFDH (Table 1). However, we do not yet know whether the loops play important roles in CO 2 binding or catalytic motion in the bacterial FDHs. Although there are many crystal structures and abundant biochemical information on NAD-dependent FDHs, the functions of these loops remains unclear. We plan to prepare a TsFDH Cterminal loop deletion mutant to test the hypothesis.
In summary, five FDHs with acidic optimum pH identified from biochemical data were tested for CO 2 reduction. The superior CO 2 -reducing activity of TsFDH was confirmed by enzyme kinetics and formate production from CO 2 gas. We propose that TsFDH is an alternative to the conventional CO 2 -reducing biocatalyst CbFDH. However, further experiments, including protein engineering and the development of NADH-regeneration systems, will be required to improve the CO 2 -reducing efficiency of TsFDH.