Bioconversion of Carbon Monoxide to Formate Using Articially Designed Carbon Monoxide:Formate Oxidoreductase in Hyperthermophilic Archaea

Ferredoxin-dependent metabolic engineering of electron transfer circuits has been developed to enhance redox efficiency in the field of synthetic biology, e.g., for hydrogen production and for reduction of flavoproteins or NAD(P) + . Here, we present the bioconversion of carbon monoxide (CO) gas to formate via a synthetic CO:formate oxidoreductase (CFOR), designed as an enzyme complex for direct electron transfer between noninteracting CO dehydrogenase and formate dehydrogenase using an electron-transferring Fe-S fusion protein. The CFOR-introduced Thermococcus onnurineus mutant strains showed CO-dependent formate production in vivo and in vitro . The formate production rate from purified CFOR complex and specific formate productivity from the bioreactor were 348 ± 34 μ mol/mg/min and 90.2 ± 20.4 mmol/g-cells/h, respectively. The CO-dependent CO 2 reduction/formate production activity of synthetic CFOR was confirmed, indicating that direct electron transfer between two unrelated dehydrogenases was feasible via mediation of the FeS-FeS fusion protein. apoferritin (443 kDa), beta-amylase (200 kDa), albumin (66 kDa) and carbonic anhydrase (29 kDa). Strep-tag purified protein was loaded and eluted at a flow rate of 0.5 ml/min, then the fractions were selectively collected at about 488 kDa of a single peak. The above procedures were carried out under anaerobic conditions. The purified proteins by size exclusion chromatography were used for enzyme assay, and examined via sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) according to the standard methods. The major bands were identified by LC-MS/MS analysis service Proteome Seoul, Korea).


Construction of CO/formate bioconversion mutants via molecular fusion of two Fe-S proteins
The redox proteins CODH and FDH from T. onnrineus NA1 were engineered systematically through molecular fusion of electron-transferring Fe-S proteins to construct a single redox complex. The codh and fdh3 gene clusters included the codhABCD and focA-fdh3ABC genes, respectively 25,29 ( Fig. 1a and Supplementary  22,30,31 . Therefore, Fdh3B was predicted to transfer electrons from Fdh3A to Fdh3C by connecting them in the complex. The amino acid sequences of CodhAB were also homologous to the CooFS proteins (41.7% and 50.3% identity, respectively) in Rhodospirillum rubrum. CooF mediates electron transfer from the CODH catalytic subunit CooS to hydrogenase and interacts directly with CooS 32,33 ; hence, spontaneous enzyme complex formation of CodhA and CodhB is easily predictable. Accordingly, Fdh3C-CodhA and Fdh3B-CodhA were designed and constructed.
Fusion of Fe-S proteins led to the formation of novel protein complexes associated with CODH and FDH, termed synthetic CFOR. Therefore, the fdh3B or fdh3C genes were fused directly to the codhA gene using Gibson Assembly in two possible arrangements, fdh3BC:codhA and fdh3B:codhA ( Fig. 1b and 1c). Structurally, the N-and C-termini of FdnH are located on the distal-end [4Fe-4S] cluster 22 ; thus, the distal-end [4Fe-4S] cluster at each Fe-S protein was expected to be aligned face-to-face in every possible fusion combinations. Predicted models of the synthetic CFORs are presented in Supplementary Fig. 2. Next, the two flexible linkers (GGGGS) 1 and (GGGGS) 2 were designed with the fdh3BC:codhA fusion arrangement. However, (GGGGS) 3 insertion was obtained during the homologous recombination of the (GGGGS) 2 , resulting in three different lengths of linkers, which was confirmed by sequencing analysis. The fdh3BC:codhA fusion constructs pFd3CoL1C1118, pFd3CoL2C1119, and pFd3CoL3C1120 carried three different lengths of flexible linkers, i.e., (GGGGS) 1 , (GGGGS) 2 , and (GGGGS) 3 , respectively (Supplementary Table 1). Notably, the shortest linker, (GGGGS) 1 , showed the highest formate productivity (Supplementary Fig. 6b and 6c) and was selected for fdh3B:codhA fusion.
A fosmid vector was used to facilitate cloning for chromosomal insertion of the synthetic CFOR. Insertion of the expression construct was targeted to a region of the chromosome between convergent genes TON_1126 and TON_1127, as previously described 34 . A 9-kbp DNA fragment containing the fdh3 and codh region and the P 0157 promotor-HMG cassette was inserted into the chromosome of T. onnurineus D02 by transformation of pFd3CoL1C1118, pFd3CoL2C1119, and pFd3CoL3C1120, generating the mutant strains BCF01, BCF02, and BCF03, respectively (Fig. 1b).
Strain BCF13 was then constructed by transformation of the pFd3NStrepCoL1C1149 fosmid, which contains the fdh3B:codhA fusion with (GGGGS) 1 , into T. onnurineus D04 strain with additional deletion of the fdh3 whole gene cluster ( Fig. 1c and Supplementary Fig. 3a). An affinity-purification Strep-tag was also inserted within the operon at the N-terminus of fdh3A to allow easy purification of the synthetic CFOR enzyme at the strain BCF13 (Fig. 1c). Strain D05 was also constructed as a negative control for BCF13, which has no fusion linker between fdh3B and codhA ( Supplementary Fig. 3b).
limit for the parental strain ( Supplementary Fig. 4b). Based on the overall reaction, equivalent amounts of formate and H + are produced, thereby decreasing the pH during formate production. Indeed, the pH of the cell supernatant was significantly decreased from pH 6.9 to 5.0 only in the BCF01 strain ( Supplementary Fig. 4c). This may affect sustainable formate production and cell growth, which are optimal at pH 6.5 under CO supplementation conditions 27 . Thus, 0.1 M bis-Tris propane (pH 6.5) buffer was added to the cell growth medium to prevent the abrupt pH change due to bioconversion of CO to formate, thus stabilizing the pH without growth inhibition ( Supplementary Fig. 5).
Next, we compared the formate production ability of the remaining mutants. Cell growth was identical, and formate production was detected ( Supplementary Fig. 6). BCF01 showed the highest formate production (7.5 mmol/L) after 60 h of incubation ( Supplementary Fig. 6b). The relative formate productivities of BCF02 and BCF03 at the final time point were determined to be 78% and 84%, respectively, compared with BCF01 ( Supplementary Fig. 6c). The (GGGGS) 1 linker showed the highest formate productivity; however, the effect on the electron transfer efficiency was insignificant. BCF13, carrying the fdh3B:codhA fusion with the (GGGGS) 1 linker, showed 6.6 ± 2.1 mmol/L formate production under CO-supplemented cell growth within 24 h (Fig. 2a), which was higher than that of BCF01 (4.3 ± 0.4 mmol/L formate) at the same time. Cell growth was similar to the other strains ( Fig.   2a). Therefore, all subsequent experiments were conducted using the BCF13 strain. Formate production from the cell suspension was also investigated in serum vials using the strain D05 (without fusion linker) and BCF13 (with fusion linker) at an OD 600 of about 0.5, incubated at 80℃ in the presence of CO gas with 2 bar (gauge pressure) CO/CO 2 (50:50 v/v) or CO/N 2 (50:50 v/v) mix gas. Activity and stability of the cell suspensions were confirmed by H 2 productivity that showed similar values among strains and gas conditions (Fig. 2c). In the BCF13, 3.0 ± 0.27 mmol/L formate was produced after 60 min incubation under the CO/CO 2 mix gas (Fig. 2b). In contrast, when the headspace was filled with CO/N 2 , only 0.3 ± 0.04 mmol/L formate was produced due to the low CO 2 partial pressure ( Supplementary Fig.   7). Although the CO oxidation reaction provides the equivalent CO 2 requirement for the formate production, the additionally supplemented CO 2 enhances the CO 2 reduction/formate production reaction. In the D05, formate production was detected under CO/CO 2 mix gas as a concentration of 0.1 ± 0.007 mmol/L at 60 min, which is 30-fold lower than the BCF13. The results indicate that electron transfer between CODH and FDH modules is achievable just by overexpression of the enzymes but extremely enhanced by a flexible fusion of FeS-FeS in the synthetic CFOR complex.

Purification of the synthetic CFOR complex
We then purified the CFOR enzyme complex isolated from strain BCF13 grown in a fed-batch bioreactor with CO-supplemented MM1 medium to activate the expression of genes controlled by the strong promoter P 0157 , which induces robust transcription and translation under CO-supplemented growth conditions 35 . The CFOR complex was purified using a strep-tag fused to the N-terminus of the FDH catalytic subunit, Fdh3A, and then analyzed by SDS-PAGE. Fdh3A, CodhB, and Fdh3B-CodhA fusion subunits, were present, with apparent molecular masses of 76, 67, and 42 kDa, indicating that the FdhB-CodhA fusion protein spontaneously bound to CodhB and FdhA to form the CFOR complex ( Fig. 3a). Protein bands consistent with the calculated molecular weights from deduced amino acid sequences were observed for all three subunits. The calculated size of Fdh3B-CodhA was 43,722 Da.
The CFOR complex was further purified using size-exclusion chromatography, and the CFOR complex from gel filtration was eluted as a single major peak with an apparent mass of around 488 kDa ( Fig. 3b and 3c). The three major bands of purified protein were identified by LC-MS/MS analysis by bands cut from the SDS-PAGE gel ( Fig. 3a and Supplementary Table 4). The 76-kDa and 67-kDa protein bands were identified as FdhA and CodhB, respectively. The 42-kDa protein band was thought to be the FdhB-CodhA fusion protein, which was identified as two different proteins, i.e., Fdh3B and CodhA. The additionally inserted formate transporter FocA 36 could enhance the secretion of intracellular formate synthesized by the CFOR complex. CodhC (29,274 Da) and CodhD (7,678 Da) are hypothetical proteins predicted to be involved in the maturation of the catalytic subunit CodhB 37,38 . However, FocA, CodhC, and CodhD subunits were not detected in both the SDS-PAGE gel and by LC-MS/MS analyses.
Thus, the entire CFOR complex was composed of the CO dehydrogenase catalytic subunit (CodhB), FDH catalytic subunit (Fdh3A), and Fe-S fusion proteins (Fdh3B-CodhA) connecting the two catalytic subunits.

Catalytic properties of the CFOR enzyme complex
The activities of CODH and FDH from purified CFOR complex were determined individually. The isolated CFOR catalyzed CO oxidation with a specific activity of 2,209 ± 159 μmol/mg/min and formate oxidation with a specific activity of 369 ± 181 μmol/mg/min at 80℃ (Supplementary Table 5). The specific activity of FDH was lower than that of CODH; therefore, the reaction of FDH was expected to be a limiting factor for the rate of formate production in the overall CFOR reaction. To confirm direct electron transfer by the Fe-S fusion protein, we investigated whether the purified CFOR could catalyze CO oxidation/formate production without any other additional electron carriers. The enzyme indeed catalyzed formate production from CO/CO2 (50:50 v/v) mix gas with a specific activity of 348 ± 34 μmol/mg/min ( Fig. 4 and Supplementary Table 5). However, formate was below the detection limit under conditions with 100% CO or without CFOR enzyme (Fig. 4). The maximum activity of the purified CFOR was determined at pH 7.5 in 50 mM sodium phosphate buffer ( Supplementary Fig. 8).
The results demonstrated that the simple, flexible fusion of two Fe-S proteins enabled electron transfer between them, leading to the formation of novel enzymes by assembly of noninteracting redox enzymes.
Another formate production enzyme, HDCR in A. woodii, which reduces CO 2 to formate using hydrogen, has a specific activity of 10 μmol/mg/min 28 ; therefore, synthetic CFOR showed a 35-fold higher formate production rate than native HDCR.

Bioconversion of CO to formate in a bioreactor
The formate production potential of the strain was tested in a bioreactor where 100% CO was continuously fed with a flowrate of 0.02-0.122 vvm (CO volume/working volume/min). Supplementary Table 6 summarizes the bioreactor parameters for the T. onnurineus BCF13 strain. Formate production was detected at a concentration of 56.4 ± 6.4 mmol/L after fermentation for 6 h (Fig. 5b). The formate production rate and specific formate productivity were calculated as 13.1 ± 0.9 mmol/L/h and 90.2 ± 20.4 mmol/g-cells/h, respectively. Recently CO-dependent formate production by coupling of CODH, ferredoxin, and HDCR was reported using A. woodii as a whole-cell biocatalyst; the formate production rate was 0.28 mmol/L/h 39 . T. onnurineus NA1 and its derivatives mutants used in this study are a basic hydrogenogenic carboxydotroph that can grow on CO as an energy source via the CO-dependent respiratory gene cluster codh-mch-mnh3 25,27,29 . Thus, the BCF13 strain showed carboxydotrophic properties, such as H 2 and CO 2 production ( Fig. 2c and 5c), and could grow under 100% CO conditions with the maximum specific growth rate (μ max ) of 0.621 ± 0.051 h -1 (Fig. 5a). This spontaneous CO 2 production by the CO-dependent respiration enhances the formate production in the bioreactor. Therefore, T. onnurineus BCF13 could be used as an industrial microorganism for the production of H 2 and formate simultaneously from CO.

Discussion
Fe-S proteins involved in the electron transfer chain have been well characterized, both structurally and functionally. Fe-S proteins in many oxidoreductase complexes are typically associated with a large subunit that has catalytic redox activity and forms an enzyme complex, such as respiratory complex I, FDH-N and formate hydrogenlyase in E. coli 22,30,40 . However, direct electron transfer between CODH and FDH has not yet been found in nature. Thus, if one could construct an electron transfer system between these two oxidoreductases, it could serve as a universal electron transfer system. Accordingly, we generated fusion proteins between CodhA and Fdh3B to create a novel electron transfer path. The small subunits CodhA and Fdh3B specifically interacted with their catalytic large subunits. Therefore, the molecular fusion of Fe-S proteins may spontaneously mediate the formation of a novel CODH-FDH protein complex. However, a simple fusion between Fe-S proteins may not allow electron transfer.
According to the electron tunneling theory, the maximum distance between the distal end [Fe-S] clusters at each protein must provide a distance of at least 14 Å for electron transfer, indicating that tight-binding and a sophisticated rearrangement of the two proteins are essential.
We showed that electron transfer was possible by simple fusion of the flexible (GGGGS) linker between electron-transferring Fe-S proteins. Based on the result, we suggest that the hyperfine magnetic field may affect this phenomenon. Indeed, Johnson et al. reported that the iron nuclei in the [Fe-S] cluster showed a magnetic hyperfine interaction with an electron spin S of 1/2 , producing an effective field of about 180 kG in the reduced state 17 . Therefore, we hypothesized that the [4Fe-4S] clusters located at both distal ends of the Fe-S fusion protein would combine tightly by physical magnetic interaction, providing the electron tunneling condition. Accordingly, we propose the following model for the CO oxidation and formate production-coupled bioactivity of the artificially constructed CFOR complex by simple, flexible fusion of two Fe-S proteins (Fig. 6). First, there was no CO oxidation reaction in the initial; CodhA and FdhB were connected by a flexible linker but could not bind tightly, and hence, there was no electron transfer occurring based on the electron tunneling theory. Additionally, when the CO oxidation reaction began in CodhB, the distal-end [4Fe-S4] cluster at CodhA was reduced by electron transfer, leading to generation of a hyperfine magnetic field on the [4Fe-4S] cluster.
Moreover, another fused Fe-S protein, Fdh3B, was forced into the magnetic field by the flexible linker.
Next, the ferromagnetic Fe atoms in the distal-end [4Fe-4S] cluster at Fdh3B were magnetized based on the influence of the magnetic field, thus binding tightly and rearranging with the [4Fe-4S] cluster of CodhA to provide the electron tunneling distance. Finally, electrons could be transferred from CODH to FDH through CodhA-Fdh3B fusion protein. Therefore, the synthetic CFOR complex catalyzes the CO 2 reduction/formate production reaction. However, no previous reports have described an attempt to transfer electrons directly between noninteracting proteins. Here, we developed a FeS-FeS fusion protein to provide a novel electron transfer path for the production of formate from CO by the synthetic CFOR enzyme complex in hyperthermophilic archaea. The electron transfer strategy endeavored here is based on an elementary natural phenomenon that 'Iron is attracted to a magnet.' This approach also exploited the self-assembly and self-regenerating abilities of live cells to create a novel formate production species. Overall, our results provide important insights into the synthesis of a new electron path using synthetic biology-based metabolic engineering. We focused on the possibility of electron transfer between noninteracting Fe-S proteins in this study. However, the effect of introducing various fusion combinations among Fe-S proteins and oxidoreductases remains to be determined for the broad application of this approach.

Parameter Value
Maximum specific growth rate, μ max (h -1 ) a 0.621 ± 0.051 Formate production rate (mmol/L/h) b 13.1 ± 0.9 Specific formate productivity (mmol/g-cells/h) b 90.2 ± 20.4 a The µmax was determined using the values of the linear regression slope in time windows of 1 to 4.5 hours. b Values were determined by dividing total yield by time difference from 2 to 6 hours.  For the pH-stat batch culture, T. onnurineus strain BCF13 was serially cultured in a 150-ml serum bottle and 7 L bioreactors (Fermentec, Cheongwon, South Korea); the working volumes of which were 80 ml and 5 L of MM1 medium, respectively, at 80°C. The bioreactors were sparged with pure argon gas (99.999%) through a microsparger. The agitation speed was 500 rpm, and the pH was controlled at 6.2 using 2 M KOH in 3.5% NaCl. The inlet gas of 100% CO was supplied by using a mass flow controller (MKPrecision, Seoul, South Korea) at feeding rates of 100 -610 ml/min. E. coli EPI300 TM -T1 R (Epicentre Biotechnologies, Madison, USA) strain was used for fosmid based molecular cloning purposes. Fosmid containing E. coli clones were cultured in LB medium containing 12.5 ug/ml chloramphenicol. General molecular biology manipulations and microbiological experiments were carried out by standard methods 44 .

Cloning and construction of the CFOR expression vector
The cloning strains, plasmids and fosmids used in this study are listed in Supplementary Table 1. The fdh3 gene cluster deletion mutant (strain D04) was constructed by the previously used gene disruption system in T. onnurineus NA1 34 (Supplementary Fig. 3a). To construct fosmid vector backbone, previously constructed complementary insertion site (Left_arm (TON1128-TON_1127) region-P 0157 promotor-HMG cassette-Right_arm (TON_1126) region) 34 was amplified by PCR using the primers listed in Supplementary Table 2. The amplicon and fosmid vector pCC1FOS (Epicentre Biotechnologies, Madison, USA) were assembled into a single vector, pNA1comFosC1096 (Supplementary Table 1 Table 2.

Analytical methods
Cell growth was monitored by measuring optical density at 600 nm (OD 600 ) with a spectrophotometer (Eppendorf, Hamburg, Germany). The unit value of OD 600 corresponded to 0.361 g/L (dry cell weight) as previously determined 35 . The amounts of CO, CO 2 , and H 2 gas were measured using a gas chromatograph (GC) as described previously 35 . Formate was measured by YL9100 high-performance liquid chromatography (HPLC; YL Instrument Co., Anyang, South Korea) with a Shodex RSpak KC-811 column (Showa Denko, Kanagawa, Japan). Ultrapure water containing 0.1% (v/v) phosphoric acid was used as the mobile phase at a flow rate of 1.0 ml/min. All samples were prepared with 1 ml of culture broth and centrifuged to remove cell debris at 4°C, 13,480 × g for 5 min. The supernatants were purified with a syringe filter and analyzed by HPLC.

Cell suspension experiment
To prepare cell suspensions, T. onnurineus strain BCF13 was anaerobically cultured in a 7-L fermentor with a working volume of 3 L as described above. At the end of the culture, the cells were harvested by centrifugation at 5,523 × g for 30 min at 20°C and then washed two times with an anaerobic modified PBS (600 mM NaCl, 2.7 mM KCl, 10 mM Na 2 HPO 4 , 2 mM KH 2 PO 4 , and 2 mM DTT). Finally, cells were resuspended in the same buffer at cell densities of OD 600 = 0.5. Four milliliters of cell suspensions were transferred to a 20-ml serum vial under a headspace of CO/CO 2 (50:50 v/v) mix gas at about 2 bar (gauge pressure), or CO/N 2 (50:50 v/v) mix gas at about 2 bar (gauge pressure), respectively. The cell suspensions were incubated at 80°C, and then the formate concentration and headspace gas composition were determined by HPLC and GC, respectively.

Purification of enzymes
Purification of CFOR enzyme complex was carried out as described previously 46 . Typically 2 to 4 g (wet weight) of fed-batch cultured T. onnurineus strain BCF13 cell pellets were harvested and washed with the modified PBS and then resuspended in buffer W (100 mM Tris-HCl, 150 mM NaCl, pH 8.0). Cells were then disrupted by sonication on ice, and cell debris was removed by centrifugation (15,000 x g for 40 min at 4°C). Affinity column purification was carried out following the manufacturer's protocols with a Strep-Tactin system (IBA-Lifsciences, Göttingen, Germany). The molecular weight and additional purification of CFOR complex were determined by analyzing the purified protein on a calibrated Superdex 200 10/300 GL column equilibrated buffer W using fast protein liquid chromatography (Äkta FPLC System, Amersham Biosciences). The column was calibrated by using these standards: thiroglobulin (669 kDa), apoferritin (443 kDa), beta-amylase (200 kDa), albumin (66 kDa) and carbonic anhydrase (29 kDa). Strep-tag purified protein was loaded and eluted at a flow rate of 0.5 ml/min, then the fractions were selectively collected at about 488 kDa of a single peak. The above procedures were carried out under anaerobic conditions. The purified proteins by size exclusion chromatography were used for enzyme assay, and examined via sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) according to the standard methods. The major bands were identified by LC-MS/MS analysis service (Yonsei Proteome Research Center, Seoul, Korea).

Enzyme assays
CODH activity was assayed at 80°C by a colorimetric assay with methyl viologen (MV) as an electron acceptor (ε 578 = 9.7 mM/cm at 578 nm) 47 , and CO as an electron donor. The assay was conducted with 4.4 ng of purified CFOR complex in 2 ml volume of sodium phosphate buffer (50 mM sodium phosphate, and 2 mM DTT, pH 7.5) containing 5 mM MV in a 4 ml serum-stoppered glass cuvette. 100% CO gas was purged in the headspace at about 2 bar (gauge pressure), and then the reaction was initiated by incubation of the reaction mixture at 80°C. Formate dehydrogenase activity was assayed using 360 ng of purified CFOR complex under the same conditions except that CO was replaced by 10 mM sodium formate. The formate production assay was carried out in 2 ml volume of 50 mM sodium phosphate buffer (pH 7.5) containing 2 mM DTT and 50 ug of purified CFOR in a 20 ml serumstoppered vial. CO gas was purged in the headspace at about 2 bar (gauge pressure) of CO/CO2 (50:50 v/v) mix gas, or at about 0.5 bar of 100% CO gas. The reaction was initiated by incubation at 80°C. Formate in reaction mixture was determined by HPLC.

Thermodynamic calculation
The biological standard Gibbs energy value (ΔG´o) was calculated by the Nernst equation using values of the standard Gibbs energy (ΔG o ) reported by Amend and Shock 48 .