Assimilation of formic acid and CO₂ by engineered Escherichia coli equipped with reconstructed one-carbon assimilation pathways

Significance

While biological utilization of one-carbon (C1) compounds has attracted much attention, previous studies have focused mainly on the utilization of CO₂. Here, we report development of Escherichia coli strains capable of assimilating formic acid (FA) and CO₂ through the C1 assimilation pathway, synthesizing pyruvate from FA and CO₂ by establishing the reconstructed tetrahydrofolate cycle and the reverse glycine cleavage pathway. To generate energy and redox while using less glucose, a heterologous formate dehydrogenase was introduced together with the C1 assimilation pathway. The resulting strain could utilize FA and CO₂ as sole carbon sources for sustaining growth. This work demonstrates that the combined use of the C1 assimilation pathway and formate dehydrogenase allows E. coli to utilize FA and CO₂ efficiently.

Abstract

Gaseous one-carbon (C1) compounds or formic acid (FA) converted from CO₂ can be an attractive raw material for bio-based chemicals. Here, we report the development of Escherichia coli strains assimilating FA and CO₂ through the reconstructed tetrahydrofolate (THF) cycle and reverse glycine cleavage (gcv) pathway. The Methylobacterium extorquens formate-THF ligase, methenyl-THF cyclohydrolase, and methylene-THF dehydrogenase genes were expressed to allow FA assimilation. The gcv reaction was reversed by knocking out the repressor gene (gcvR) and overexpressing the gcvTHP genes. This engineered strain synthesized 96% and 86% of proteinogenic glycine and serine, respectively, from FA and CO₂ in a glucose-containing medium. Native serine deaminase converted serine to pyruvate, showing 4.5% of pyruvate-forming flux comes from FA and CO₂. The pyruvate-forming flux from FA and CO₂ could be increased to 14.9% by knocking out gcvR, pflB, and serA, chromosomally expressing gcvTHP under trc, and overexpressing the reconstructed THF cycle, gcvTHP, and lpd genes in one vector. To reduce glucose usage required for energy and redox generation, the Candida boidinii formate dehydrogenase (Fdh) gene was expressed. The resulting strain showed specific glucose, FA, and CO₂ consumption rates of 370.2, 145.6, and 14.9 mg⋅g dry cell weight (DCW)⁻¹⋅h⁻¹, respectively. The C1 assimilation pathway consumed 21.3 wt% of FA. Furthermore, cells sustained slight growth using only FA and CO₂ after glucose depletion, suggesting that combined use of the C1 assimilation pathway and C. boidinii Fdh will be useful for eventually developing a strain capable of utilizing FA and CO₂ without an additional carbon source such as glucose.

Gaseous one-carbon (C1) compounds such as CO₂ and CH₄ are major greenhouse gases, which are largely responsible for global warming and climate change. With the aim of reducing greenhouse gases, biological conversion of C1 compounds has attracted much attention due to several advantages such as low energy requirements, environmental friendliness, and the possibility of converting greenhouse gases directly to value-added products (1–3). However, a major limitation of biological C1 conversion is the inefficiency of the natural C1 assimilation pathway such as the Calvin–Benson–Bassham (CBB) cycle (4). To address this problem, extensive studies have been performed to enhance biological C1 assimilation by developing de novo pathways using efficient carboxylases (5–9), introducing the CBB cycle into Escherichia coli (10–13), or by integration with an electrochemical method of reducing CO₂ to formic acid (FA) (14, 15). FA can be easily and efficiently synthesized from CO₂ by various chemical processes employing metal catalysts (16) or electricity (17). The use of FA as a fermentation substrate is advantageous as it can be easily stored and transported, it is soluble in water, and its biological assimilation is faster than that of CO₂ (18).

Due to these advantages of FA as a substrate, various studies have been conducted to biologically convert FA to value-added chemicals. For example, there has recently been a report on utilizing FA as a secondary carbon source to produce succinic acid (19). Numerous studies on natural and synthetic pathways for FA and CO₂ assimilation have suggested several promising candidate pathways including the serine pathway, the reductive acetyl-CoA pathway, the reductive glycine pathway, the serine–threonine pathway, the pyruvate formate lyase (Pfl)–threonine pathway, the formate tetrahydrofolate (THF) ligase-Pfl bicycle, the reductive glycine pathway with Pfl, the Pfl-phosphoketolase pathway, the formate reduction with ribulose monophosphate pathway, the 3-oxopropionyl-CoA pathway, and the synthetic enzyme formolase pathway (20–22). Among these candidate pathways, the THF cycle and glycine cleavage (gcv) pathway have several advantages, such as the oxygen tolerance of the enzymes involved, their high energy efficiency, and their independent operation with little overlapping of the central metabolism (23–25).

In a previous report, the THF cycle was established in E. coli by the expression of Methylobacterium extorquens formate-THF ligase (Ftl). FA assimilation through the THF cycle was verified using ¹³C-labeled FA by analyzing ¹³C-labeling patterns in amino acids, including serine, glycine, and methionine (26). However, two major problems in using the THF cycle to efficiently assimilate FA by E. coli remained unsolved. First, the E. coli bifunctional enzyme Fol (encoded folD) possessing 5,10-methylene-THF dehydrogenase and 5,10-methylene-THF cyclohydrolase activities is allosterically inhibited by its reactant, 10-formyl THF (10-CHO-THF) (Fig. 1 and ref. 27). Second, the THF cycle requires glycine to sustain the assimilation of FA since it consumes one glycine molecule per cycle (Fig. 1). Thus, the THF cycle cannot operate continuously without an external glycine supply. To solve the glycine consumption problem, utilization of the gcv pathway was proposed in a previous review paper (20). In E. coli, the gcv pathway degrades glycine and produces 5,10-methylene THF (5,10-CH₂-THF), CO₂, NADH, and NH₃. However, in vitro enzyme assays suggested that the gcv reaction was reversible (28, 29), and thus it might be possible to synthesize glycine from 5,10-CH₂-THF, CO₂, NADH, and NH₃. Furthermore, a recent study reported a newly isolated uncultivated dissimilatory phosphite oxidation bacterium, Candidatus phosphitivorax, that could grow utilizing only CO₂ and [HPO₃]²⁻ by utilizing the reverse gcv pathway (30). The metagenomics-guided analysis discovered that the bacterium lacked key genes of the known CO₂ assimilation pathways, but did contain genes involved in the gcv pathway, suggesting the potential relevance of the gcv pathway for the CO₂ assimilation process.

Fig. 1.

The rTHF cycle and reverse gcv pathway established in E. coli. Red arrows indicate heterologous pathways, black arrows indicate native pathways, and blue arrows indicate overexpressed native pathways. The red Xs indicate genes knocked out. Carbon-labeling patterns in glycine, serine, and pyruvate based on FA and CO₂ carbons are indicated. Blue- and orange-colored carbons originated from FA and CO₂, respectively. Numbers in each carbon represent the carbon number in each compound. Recombinant plasmids designed and used in this study are also shown.

In this study, the aforementioned two key problems of efficiently utilizing the THF cycle in FA assimilation were solved by the reconstruction of the THF cycle and reversal of gcv pathway. The THF cycle was reconstructed using E. coli native serine hydroxymethyltransferase (GlyA) and three heterologous enzymes, Ftl, methenyl-THF cyclohydrolase (Fch), and methylene-THF dehydrogenase (Mtd) (Fig. 1), from M. extorquens. After confirming that the reconstructed THF (rTHF) cycle assimilated FA more efficiently than the previously reported THF cycle, reversal of the gcv pathway was conducted by knocking out the gcv pathway repressor gene (gcvR) to provide glycine from CO₂. An engineered E. coli strain harboring the rTHF cycle and the reverse gcv pathway was able to synthesize up to 96% of proteinogenic serine from FA and CO₂ in a medium also containing glucose. Serine was further converted to pyruvate by the native serine deaminase (Sda). After additional engineering, FA and CO₂ assimilation could be enhanced. To reduce the amount of glucose required for energy and redox generation, Candida boidinii formate dehydrogenase (Fdh) was additionally employed. The final engineered strain was able to sustain growth using only FA and CO₂ after glucose depletion.

Results

Use of the rTHF Cycle Improves FA Assimilation.

The THF cycle assimilates FA utilizing THF as an intermediate by four enzymatic conversion steps catalyzed by the M. extorquens Ftl, E. coli Fol, and GlyA, which are encoded by ftl, folD, and glyA, respectively. Ftl assimilates FA together with THF, producing 10-CHO-THF. The bifunctional enzyme Fol first converts 10-CHO-THF to 5,10-methenyl THF (5,10-CH=THF) and eventually produces 5,10-CH₂-THF. Then GlyA converts 5,10-CH₂-THF and glycine to THF and serine, respectively. THF regenerated by GlyA is utilized again to form 10-CHO-THF by assimilating another FA in a new THF cycle (Fig. 1). However, the E. coli Fol is inefficient in FA assimilation due to allosteric inhibition by 10-CHO-THF. On the other hand, the M. extorquens Fch converts 10-CHO-THF to 5,10-CH=THF, and M. extorquens Mtd converts 5,10-CH=THF to 5,10-CH₂-THF (24). No allosteric inhibition has been reported from the studies on enzyme activities and reaction mechanisms of Fch and Mtd (31–33). Thus, the THF cycle was reconstructed by overexpressing the M. extorquens ftl, fch, and mtd genes encoding Ftl, Fch, and Mtd, respectively, and utilizing E. coli native GlyA (Fig. 1).

The rTHF cycle was capable of efficiently assimilating FA to cellular molecules and consumed more FA than the previously reported THF cycle in E. coli. To compare FA assimilation through the THF cycle and the rTHF cycle, the THF1, THF2, and THF4 strains equipped with the THF cycle, rTHF cycle, and THF cycle with additional E. coli folD overexpression, respectively, were constructed by introducing p100FTL, p100THF, and p100FTLFOL, respectively, in the DH5α strain (SI Appendix, Fig. S1 B, C, and E and Table S1). FA assimilation through the rTHF cycle was analyzed by performing ¹³C isotope analysis and monitoring FA consumption profiles during cultivation. Since FA is toxic at levels above 4 g⋅L⁻¹ (SI Appendix, Fig. S2), the MR minimal medium containing glucose was supplemented with ¹³C-labeled sodium formate (sodium FA) at a concentration of 2.7 g⋅L⁻¹ (the equivalent of 1.84 g⋅L⁻¹ FA) to prevent FA toxicity. One glycine molecule is consumed per single THF cycle reaction. Thus, the culture medium was supplemented with 2 g⋅L⁻¹ of glycine to ensure continuous operation of the THF cycle. ¹³C isotope analysis was carried out by measuring ¹³C labeling in proteinogenic serine and methionine. The carbon originating from FA (hereafter “FA carbon”) is placed at the third carbon position of serine. Also, a portion of 5,10-CH₂-THF is used to synthesize methionine, and in this case the FA carbon is placed at the fifth carbon position of methionine (SI Appendix, Fig. S1 A–E). Thus, the relative FA assimilation flux through the rTHF cycle can be analyzed by determining the ratio of ¹³C-labeled proteinogenic serine or ¹³C-labeled proteinogenic methionine. In the THF1 strain, less than 10% of proteinogenic methionine and serine were labeled with ¹³C (Fig. 2 A and B). The FA concentration at the end of cultivation did not change much from the initial concentration (Fig. 2C). On the other hand, most of the methionine and serine in the THF2 strain were labeled with ¹³C (Fig. 2 A and B), and about half of the supplemented FA was consumed at the end of cultivation. In the THF4 strain, ¹³C-labeled proteinogenic methionine and serine ratios were higher than in the THF1 strain but were lower than in the THF2 strain (Fig. 2 A and B).

Fig. 2.

¹³C isotope analysis of serine and methionine and FA consumption in the WT, THF1, THF2, THF3, and THF4 strains. The THF1 strain is equipped with the THF cycle, the THF2 strain is equipped with the rTHF cycle, the THF3 strain is equipped with a stronger rTHF cycle, and the THF4 strain is equipped with the THF cycle with additional overexpression of the native E. coli folD gene. The MR minimal medium containing glucose was supplemented with 2.7 g⋅L^{−1 13}C-labeled sodium FA (the equivalent of 1.84 g⋅L⁻¹ FA) and 2 g⋅L⁻¹ glycine. (A and B) The number of labeled carbons and their ratios in proteinogenic methionine (A) and serine (B). (C and D) FA concentration profile (C) and growth profile (D) of the THF1, THF2, THF3, and THF4 strains during FA assimilation through the THF cycle. Data are shown as average values with error bars representing the SD obtained in duplicate experiments (n = 2).

To investigate the possibility of enhanced FA assimilation, the rTHF-cycle enzymes were further overproduced by expressing the genes under a strong trc promoter (SI Appendix, Fig. S3 A and B). The THF3 strain equipped with a stronger rTHF cycle overexpressing Ftl, Fch, and Mtd was constructed by introducing plasmid pTrcTHF in the DH5α strain (SI Appendix, Fig. S1E). However, the THF3 strain showed no significant change in the ¹³C-labeled proteinogenic methionine and serine ratios compared with those of the THF2 strain (Fig. 2 A and B). FA consumption in the THF3 strain was also similar to that of the THF2 strain at the end of cultivation (Fig. 2C). Thus, further overexpression of the ftl, fch, and mtd genes was not effective in improving FA assimilation. Growth retardation was not observed in any of the strains (THF1, THF2, THF3, and THF4) equipped with the THF cycle or the rTHF cycle; a prolonged lag phase was observed for the THF2 strain (Fig. 2D).

To further enhance FA assimilation through the rTHF cycle, the effect of increased THF availability in FA assimilation was examined using the THF2 strain. THF availability was increased by the addition of folic acid, which can be converted to THF in vivo by the native dihydrofolate reductase encoded by the folA gene. However, no improvement in FA assimilation through the rTHF cycle was observed (SI Appendix, Fig. S4 A–D).

Increasing in Vivo Glycine Biosynthesis from FA and CO₂.

In native E. coli, glycine is synthesized from serine by GlyA and is degraded to 5,10-CH₂-THF and CO₂ through the gcv pathway encoded by the gcv complex (gcvTHP). To examine possible reversal of the gcv pathway in vivo, reactants utilized in reverse reaction, including 5,10-CH₂-THF, CO₂, and NADH, were additionally supplied. The change in the metabolic flux in the gcv pathway was monitored by analyzing the ¹³C-labeled ratio of the first carbon in proteinogenic glycine when the medium was supplemented with ¹³C-labeled NaHCO₃, since CO₂ assimilated through the gcv pathway is positioned at the first carbon of glycine. However, the additional supply of reactants did not improve the reverse reaction flux (SI Appendix, Fig. S5 A–C). It has been known that gcvTHP genes, which encode the enzymes involved in the gcv reaction, are negatively regulated mainly by gcvR, which inhibits the transcription of gcvTHP genes (34, 35). Thus, gcvTHP genes were overexpressed by introducing the plasmid p184GCV, which harbors gcvTHP genes under the constitutive BBa 23100 promoter, into DH5α to make the GCV strain. Also, the RG1 strain was constructed by additionally knocking out the gcvR repressor gene in the GCV strain. The WT, GCV, and RG1 strains, which are expected to give low, medium, and high levels of gcvTHP expression, were cultured, and glycine biosynthesis through the reverse gcv pathway was analyzed. The ¹³C-labeled first carbon ratios in proteinogenic glycine were gradually increased successively in the WT, GCV, and RG1 strains (SI Appendix, Fig. S5D). The ¹³C-labeled ratio of the first carbon in proteinogenic glycine reached up to 52% of proteinogenic glycine in the RG1 strain by enhancing the reverse gcv reaction (Fig. 3A).

Fig. 3.

¹³C isotope analysis of glycine-, serine-, and pyruvate-forming flux in the WT, RG1, RG2, RG3, RG4, RG5, and RG6 strains which have the rTHF and/or reverse gcv pathway. (A) ¹³C isotope-labeled fraction in glycine carbon number one of the WT, RG1, and RG2 strains.¹³C-labeled NaHCO₃ (8.4 g⋅L⁻¹) was added to supply ¹³C-labeled CO₂. (B and C) Ratio of proteinogenic serine synthesized only from C1 sources (FA and CO₂) (red) or glucose (yellow) in the RG2, RG3, RG4, RG5, and RG6 strains (B). Pyruvate-forming flux from the C1 pathway in the RG2, RG3, RG4, RG5, and RG6 strains. Pyruvate-forming flux from C1 pathway represents the percentage of carbon flux of the C1 assimilation pathway in total pyruvate-forming flux from the C1 pathway and glucose (C). For the RG2, RG3, and RG4 strains, 2.7 g⋅L⁻¹ of ¹³C-labeled sodium FA (the equivalent of 1.84 g⋅L⁻¹ of FA) and 8.4 g⋅L⁻¹ of ¹³C-labeled NaHCO₃ were added to supply ¹³C-labeled FA and ¹³C-labeled CO₂, respectively. For the RG5 and RG6 strains, 2.7 g⋅L⁻¹ of ¹³C-labeled sodium FA (the equivalent of 1.84 g⋅L ¹ of FA) and 8.4 g⋅L⁻¹ of unlabeled NaHCO₃ were added. Data are shown as average values with error bars representing ± SD obtained in duplicate experiments for the WT, RG1, RG2, and RG3 strains (n = 2) and in triplicate for the RG4, RG5, and RG6 strains (n = 3).

The rTHF cycle was additionally introduced in the above strains to supply 5,10-CH₂-THF from FA for synthesizing glycine from FA and CO₂ only; the THF2, THFGCV, and RG2 strains were constructed by introducing p100THF in the WT, GCV, and RG1 strains, respectively. Similar to the above results, the ¹³C-labeled first carbon ratios of proteinogenic glycine were increased successively in the THF2, THFGCV, and RG2 strains. The ratios of ¹³C-labeled first carbon in proteinogenic glycine increased much more upon coexpression of the rTHF cycle (SI Appendix, Fig. S5E). In the RG2 strain, the ¹³C-labeled ratio of the first carbon in glycine reached as high as 94% of proteinogenic glycine (Fig. 3A). These results suggest that the RG2 strain is capable of synthesizing most of the glycine solely from FA and CO₂ by the combined use of rTHF cycle and the reverse gcv pathway.

Biosynthesis of Serine and Pyruvate from FA and CO₂ Through the rTHF Cycle and Reverse Gcv Pathway.

By the combined use of the rTHF cycle and reverse gcv pathway, glycine could be synthesized solely from FA and CO₂ in a glucose-containing medium. Also, serine could be synthesized from two FA molecules and one CO₂ molecule and sequentially converted to pyruvate by the E. coli Sda. Thus, one pyruvate (or serine) molecule was synthesized from two FA and one CO₂ molecules by consuming two ATPs, two NADPHs, and one NADH (Fig. 1). To confirm serine and pyruvate biosynthesis from FA and CO₂, the ¹³C-labeled proteinogenic serine and alanine ratios were measured in the RG2 strain after culturing in a medium containing ¹³C-labeled sodium FA and ¹³C-labeled NaHCO₃. The pyruvate-forming flux from FA and CO₂ in a glucose-containing medium can be calculated from the ¹³C-labeled proteinogenic alanine ratio, since alanine is synthesized from pyruvate by a one-step reaction. The RG2 strain synthesized 86% of the proteinogenic serine from FA and CO₂. On the other hand, only 4.5% of proteinogenic alanine was synthesized from FA and CO₂, which suggested that 4.5% of the total pyruvate-forming flux came from FA and CO₂ (Fig. 3B).

Compared with the ratio of proteinogenic serine synthesized from FA and CO₂, far less pyruvate-forming flux was observed (Fig. 3C). Thus, native sdaA was overexpressed to possibly improve pyruvate-forming flux from FA and CO₂. To examine the effect of overexpressing the E. coli sdaA gene, the THF2 strain was used rather than the RG2 strain because the former strain is capable of generating more serine and can clearly show the effect of sdaA overexpression on the serine-to-pyruvate conversion. Thus, E. coli sdaA was overexpressed in the THF2 strain to make the SDA strain. The THF2 and SDA strains were grown in a medium containing ¹³C-labeled sodium FA and glycine. Unexpectedly, the SDA strain exhibited lower pyruvate-forming flux from FA and glycine than the THF2 strain; the THF2 and SDA strains contained 26% and 17% of ¹³C-labeled proteinogenic alanine, respectively (SI Appendix, Fig. S6A).

To find the reason, the ratios of proteinogenic serine synthesized from glucose and from the THF cycle and glycine in the THF2 and SDA strains were determined (detailed results are presented in SI Appendix, Text S1). It was found that overexpression of sdaA gene promoted serine biosynthesis pathway flux from glucose (SI Appendix, Fig. S6B), causing a decrease in the proteinogenic serine ratio coming from the rTHF cycle. Consequently, the pyruvate-forming flux from the rTHF cycle and glycine was also decreased. Therefore, the bottleneck in improving the pyruvate-forming flux from FA and CO₂ seems to be insufficient flux through the reverse gcv pathway rather than the conversion of serine to pyruvate. When the pyruvate-forming flux from the C1 assimilation pathway became dependent only on the FA assimilation flux through the rTHF cycle by supplying excessive amounts of glycine, the pyruvate-forming flux from FA and glycine was 26% of the total pyruvate-forming flux in the THF2 strain. Glycine needs to be synthesized from CO₂ and 5,10-CH₂-THF, and half of the 5,10-CH₂-THF needs to be used for glycine biosynthesis. One mole of FA is converted to 0.5 mol of pyruvate, and thus 13% of the total pyruvate-forming flux can potentially come from FA and CO₂ if the reverse gcv flux matches the rTHF cycle flux.

Reinforcing the Reverse Gcv Pathway and Further Metabolic Engineering to Increase FA and CO₂ Assimilation into Pyruvate.

To increase the pyruvate-forming flux from FA and CO₂, reverse gcv flux was enhanced by the optimized overexpression of gcvTHP genes. In the RG2 strain, the reverse gcv flux was enhanced by plasmid p184GCV-based overexpression of the gcvTHP operon. However, the use of multiple plasmids negatively affected cell growth and yielded large colony-to-colony variation. Thus, chromosomal overexpression of the gcvTHP operon was examined to determine whether cell growth retardation can be prevented and colony-to-colony variation can be reduced while maintaining strong glycine biosynthesis. For this, the native promoter of the gcvTHP operon in the E. coli genome was changed to a strong trc promoter (DH5α_GT strain) (SI Appendix, Table S1). Then, the rTHF cycle genes were introduced to the DH5α_GT strain to make the RG3 strain (SI Appendix, Table S1). The RG3 strain showed better growth with less colony-to-colony variation, improved serine biosynthesis (95.5% of proteinogenic serine) (Fig. 3B), and higher pyruvate-forming flux (7.6% of total pyruvate-forming flux) from FA and CO₂ compared with the RG2 strain (Fig. 3C). These results suggest that strong chromosomal overexpression of the gcvTHP operon allows better glycine biosynthesis and is more beneficial in increasing FA and CO₂ assimilation into pyruvate. To further improve pyruvate-forming flux from FA and CO₂, the pyruvate formate lyase gene (pflB) was knocked out to block undesired pyruvate degradation to FA. Also, the native lipoamide dehydrogenase (Lpd) gene (lpd) was overexpressed since Lpd supplies NADH to the enzyme complex of the reverse gcv reaction. The resulting RG4 strain, which is the RG3 strain with pflB knockout and lpd overexpression, showed similarly high serine biosynthesis (96.3% of proteinogenic serine) from FA and CO₂ (Fig. 3B). Also, 9.4% of the total pyruvate-forming flux came from FA and CO₂ (Fig. 3C). However, it should be noted that lpd overexpression has the negative consequence of releasing CO₂ as Lpd converts pyruvate to acetyl-CoA. Reflecting this, the acetic acid concentration increased slightly upon lpd overexpression. The RG3, RG3-1 (the RG3 strain with pfl knockout only), and RG4 strains produced 1.34 ± 0.11, 1.18 ± 0.01, and 0.83 ± 0.02 g⋅L⁻¹ of acetic acid, respectively. The acetic acid yields on glucose were 1.25 ± 0.01, 0.92 ± 0.01, and 1.49 ± 0.04 mol acetic acid produced per mole consumed glucose, respectively (SI Appendix, Table S2). The RG4 strain generated 0.57 mol more CO₂ per mole consumed glucose than the RG3-1 strain, since one CO₂ molecule is generated during the conversion of pyruvate to acetyl-CoA. On the other hand, the amounts of CO₂ assimilated through the C1 assimilation pathway per consumed glucose were calculated to be 0.23 mol and 0.25 mol in the RG4 and RG3-1 strains, respectively. These results suggest that lpd overexpression allows increased pyruvate-forming flux from FA and CO₂, but the total CO₂ production per consumed glucose became higher.

To further improve pyruvate biosynthesis from FA and CO₂, we designed another strategy. d-3-phosphoglycerate dehydrogenase encoded by serA is essential for E. coli growing in M9 minimal medium containing glucose since the serA-deleted strain cannot synthesize the 5,10-CH₂-THF needed for the biosynthesis of metabolites such as purines and methionine (5). Thus, we reasoned that the serA-deleted strain equipped with the C1 assimilation pathway would synthesize 5,10-CH₂-THF only through the C1 assimilation pathway, eventually allowing enhanced FA and CO₂ assimilation through the C1 assimilation pathway. The effect of serA knockout on FA and CO₂ assimilation was first examined using the RG4-1 strain (the serA-deleted RG4 strain). However, cell growth was severely retarded even in the complex Luria–Bertani medium. Thus, the effect of serA knockout on C1 assimilation was examined in the RG5 strain (the serA-deleted RG3-1 strain). It was found that knocking out serA was effective in improving FA and CO₂ assimilation into pyruvate. Pyruvate-forming flux from the C1 assimilation pathway reached up to 12.9% of the total pyruvate-forming flux in the RG5 strain, compared with 7.3% in the RG3-1 strain.

In the above studies, a two-plasmids-based expression system was employed. It has been known that in some cases a two-plasmids expression system can be detrimental to cell growth, depending on the genes on the plasmid. Thus, we tested a single-plasmid-based expression system. Plasmid p100THFGcvlpd was constructed to express the rTHF cycle, gcvTHP, and lpd genes from a single plasmid. The pyruvate-forming flux from FA and CO₂ in the resulting RG6 strain was increased to 14.9% of the total pyruvate-forming flux (Fig. 3C). In addition, the specific FA consumption rate reached 42.4 ± 3.6 mg⋅gram dry cell weight (gDCW)⁻¹⋅h⁻¹, which was higher than the consumption rates obtained with the RG4 (20.7 ± 1.8 mg⋅gDCW⁻¹⋅h⁻¹) and RG5 (37.9 ± 5.1 mg⋅gDCW⁻¹⋅h⁻¹) strains (Table 1). The specific CO₂ consumption rate through the C1 assimilation pathway in the RG6 strain was calculated to be 20.3 mg⋅gDCW⁻¹⋅h⁻¹, which was also higher than those obtained with the RG4 (9.9 mg⋅gDCW⁻¹⋅h⁻¹) and RG5 (18.1 mg⋅gDCW⁻¹⋅h⁻¹) strains (Table 1). These results suggest that the use of single-plasmid-based overexpression of the rTHF cycle, gcvTHP, and lpd genes in the DH5α_GTPS strain allowed more efficient assimilation of FA and CO₂.

Table 1.

Specific glucose, FA, and CO₂ consumption rates and C1 utilization ratios of the RG4, RG5, RG6, and RG8 strains

Strain	Specific glucose consumption rate, mg⋅gDCW−1⋅h−1	Specific FA consumption rate, mg⋅gDCW−1⋅h−1	Specific CO2 consumption rate*, mg⋅gDCW−1⋅h−1	C1 utilization ratio†, mol/mol
RG4	249.8 ± 31.4	20.7 ± 1.8	9.9	0.076 ± 0.015
RG5	319.4 ± 5.1	37.9 ± 5.1	18.1	0.104 ± 0.013
RG6	319.7 ± 1.9	42.4 ± 3.6	20.3	0.115 ± 0.008
RG8	232.2 ± 6.0	28.6 (by C1 pathway)‡ 80.3 (by Fdh)‡	13.9	0.076§
RG8¶	370.2	31.0 (by C1 pathway)‡ 114.6 (by Fdh)‡	14.9	0.076§

^*The specific CO₂ consumption rate is not a measured value. It is calculated from the specific FA consumption rate using the stoichiometry of the C1 assimilation pathway; the C1 assimilation pathway assimilates two FA molecules and one CO₂ molecule, thus, the specific CO₂ consumption rate was first calculated as mole basis and later converted to mass basis.

^†The C1 utilization ratio is calculated by (carbon moles from FA and CO₂)/(carbon moles from FA, CO₂, and glucose). The theoretical maximum value of the C1 utilization ratio is 0.627.

^‡The RG8 strain consumes FA through the C1 assimilation pathway (C1 pathway) as well as Fdh. Thus, the specific FA consumption rates through the C1 pathway and Fdh are indicated separately.

^§The C1 utilization ratio is a predicted value. The C1 utilization ratio of the RG8 strain is assumed to be similar to that of the RG4 strain because these two strains showed a similar percentage of the total pyruvate-forming flux coming from FA and CO₂.

^¶Values were obtained in the bioreactor cultivation.

It was also confirmed by ¹³C isotope analysis that pyruvate synthesized through the C1 assimilation pathway was converted to other cellular molecules by the native E. coli metabolism. Valine, leucine, isoleucine, and glutamate were selected as the target cellular metabolites as they are synthesized utilizing pyruvate and/or acetyl-CoA (SI Appendix, Fig. S7A). The ¹³C isotope-labeling patterns observed were as expected (SI Appendix, Fig. S7B), and the ¹³C-labeled proteinogenic amino acids ratios (SI Appendix, Fig. S7 C–F) were consistent with those expected from the ¹³C-labeled pyruvate ratio. Detailed explanations are presented in SI Appendix, Text S2.

When the reaction is reversible, it is important to examine both the forward and reverse reaction fluxes. The major reversible reaction that affects the C1 assimilation pathway flux is the gcv reaction. The fluxes of forward and reverse gcv reactions in the RG6 strain were examined from the number of ¹³C-labeled carbons, the ¹³C-labeled position, and the ratio of proteinogenic glycine (detailed methods and procedures are presented in SI Appendix, Text S3). The flux of the reverse gcv reaction increased as the intracellular glycine pool decreased, while the flux of the forward gcv reaction decreased as intracellular glycine pool decreased (except for feeding 0.7–0.5 g⋅L⁻¹ of 2-¹³C–labeled glycine). These labeling experiments suggested that the forward gcv reaction flux must be very low (e.g., lower than that obtained with 0.01 g⋅L⁻¹ of 2-¹³C–labeled glycine supplementation) (SI Appendix, Fig. S8) because our experiments in assimilating FA and CO₂ to form pyruvate were performed without glycine supplementation. Thus, in the RG6 strain the reverse gcv pathway operates like an almost irreversible reaction, and labeling patterns are little, if at all, affected by the potential reversibility of the reaction.

Introduction of Fdh Further Increased FA Utilization with Less Glucose Consumption.

Although thus far we were able to achieve the most efficient FA and CO₂ assimilation in engineered E. coli equipped with the rTHF cycle and reverse gcv pathway, the need to use relatively large amounts of glucose remained a problem. Cells still need to use glucose to generate energy and redox, although theoretically all the carbons can come from FA and CO₂. Thus, we reasoned that the generation of energy and redox from FA would reduce the amount of glucose needed. To achieve this goal, a heterologous NAD⁺-utilizing Fdh, which converts FA to CO₂ while producing NADH, was introduced into E. coli. Two Fdhs from M. extorquens and C. boidinii reported to have outstanding enzyme activities (36, 37) were chosen in this study. Plasmids p184FDH and p184FDHcbo expressing the fdh genes of M. extorquens and C. boidinii, respectively, were transformed into the RG6 strain to construct the RG7 and RG8 strains, respectively. The RG7 strain exhibited instability and larger colony-to-colony variations, but the RG8 strain did not. The instability of the RG7 strain seems to be due to the inconsistent expression of a large M. extorquens Fdh complex (∼185 kDa) comprising four subunits encoded by an operon of large size (5.3 kb). On the other hand, the C. boidinii Fdh is a homodimeric protein (∼74 kDa) encoded by a 1.1-kb gene (38). Thus, we used the RG8 strain for further experiments.

To confirm that generation of energy and redox from FA would reduce the amount of glucose needed and potentially sustain cell growth only from FA and CO₂, three independent flask cultures were performed in two stages; this was because FA was completely consumed during the flask cultivation as described below. At the first stage of cultivation, the RG8 strain was cultivated in M9 minimal medium supplemented with 5 g⋅L⁻¹ glucose and 3.7 g⋅L⁻¹ sodium FA (the equivalent of 2.5 g⋅L⁻¹ FA). The RG8 strain consumed FA rapidly, and FA was completely depleted when glucose was depleted (Fig. 4A). In contrast, the RG6 strain consumed FA more slowly, with FA remaining after glucose depletion (Fig. 4B). Flask culture of the RG8 strain showed specific glucose and FA consumption rates of 232.2 ± 6.0 and 108.9 ± 3.2 mg⋅gDCW⁻¹⋅h⁻¹, respectively. Also, the pyruvate-forming flux from FA and CO₂ was measured after supplementation with ¹³C-labeled FA as described earlier for the RG2–RG6 strains (Fig. 3C). In the RG8 strain 9.6% of the total pyruvate-forming flux came from FA and CO₂, which was similar to the percentage (9.4%) in the RG4 strain. However, it was lower than the percentage (14.9%) in the RG6 strain due to the consumption of FA by Fdh, which lowered the FA flux toward C1 assimilation pathway. It was assumed that the C1 utilization ratio of the RG8 strain would be similar to that of the RG4 strain (0.076) (Table 1) because these two strains showed similar percentages of the total pyruvate-forming flux coming from FA and CO₂. The specific CO₂ consumption rate of the RG8 strain was calculated to be 13.9 mg⋅gDCW⁻¹⋅h⁻¹. The C1 assimilation pathway and Fdh reaction consumed 26.3 wt% and 73.7 wt% of FA, respectively, suggesting that three times more FA was consumed for energy and redox generation. Nonetheless, the specific FA consumption rate was increased by 160%, while the specific glucose consumption rate decreased by 27% compared with the RG6 strain. Thus, the generation of energy and redox from FA by employing the Fdh reaction indeed reduced the amount of glucose needed by cells. These results led us to examine whether cells can sustain growth utilizing only FA and CO₂.

Fig. 4.

(A and B) Cell growth and glucose and FA consumption profiles of the RG8 strain (A) and RG6 strain (B) in flask cultures. Cells were grown using M9 minimal medium initially supplemented with 5 g⋅L⁻¹ of glucose and 3.7 g⋅L⁻¹ of sodium FA (the equivalent of 2.5 g⋅L⁻¹ of FA). For the RG8 strain, the medium was supplemented with 3.7 g⋅L⁻¹ of ¹³C-labeled sodium FA (the equivalent of 2.5 g⋅L⁻¹ of FA) at the points indicated by the red arrows. Cells for ¹³C isotope analysis were taken at the P1, P2, and P3 points. (C, Center) Biosynthesis route from FA and CO₂ to various amino acids during cell growth from only FA and CO₂. Solid lines and arrows represent single-step pathways; dashed lines and arrows represent multiple-step pathways. 2KG, alpha-ketoglutarate; Ac-CoA, acetyl-CoA; Ala, alanine; E4P, erythrose 4-phosphate; FBP, fructose 1,6-bisphosphate; Glu, glutamate; Gly, glycine; Ile, isoleucine; Leu, leucine; OAA, oxaloacetate; PEP, phosphoenolpyruvate; Phe, phenylalanine; PYR, pyruvate; Ser, serine. (Left, Right, and Bottom Row) ¹³C-labeled proteinogenic amino acids ratios at three different points obtained from flask cultivation (red-outlined graphs) and bioreactor cultivation (blue-outlined graphs). Data from the flask cultivation are shown as average values with error bars representing ± SD obtained in triplicate (n = 3). (D and E) Cell growth and glucose and FA consumption profiles from the bioreactor cultivations of the RG8 strain (D) and the RG6 strain (E). Cells were grown in a bioreactor using M9 minimal medium initially supplemented with 5 g⋅L⁻¹ glucose and 3.7 g⋅L⁻¹ sodium FA (the equivalent of 2.5 g⋅L⁻¹ FA). For the RG8 strain, the medium was supplemented with a mixture of ¹³C-labeled FA and unlabeled FA (1:1 ratio; mole/mole) when the initially supplemented FA was almost depleted, and then the FA mixture was supplemented once again. After that, the medium was supplemented twice with unlabeled FA at the indicated time points. Cells for ¹³C isotope analysis were taken at the F1, F2, and F3 points. Blue lines with closed triangles indicate cell growth. Yellow lines with closed circles indicate glucose concentration. Gray lines with closed squares indicate FA concentration.

After the depletion of the initially supplied glucose, the second stage of cultivation was started from the P1 point (Fig. 4A). To examine whether cells can sustain growth utilizing only FA and CO₂, the M9 minimal medium was manually supplemented with FA when FA was depleted during the cultivation. Initially, the M9 minimal medium was supplemented with 3.7 g⋅L⁻¹ of ¹³C-labeled sodium FA (the equivalent of 2.5 g⋅L⁻¹ of FA) at the P1 point and again at the P2 point when the added ¹³C-labeled FA had been consumed (Fig. 4A). Flask culture was ended at the P3 point. Cells were sampled at the P1, P2, and P3 points; the P1 and P2 samples were taken right before supplementation with ¹³C-labeled FA (Fig. 4A). FA assimilation into amino acids synthesized from the C1 assimilation pathway (glycine and serine), TCA cycle (glutamate and isoleucine), pentose phosphate pathway (phenylalanine), and pyruvate (alanine, valine and leucine) (Fig. 4C) were analyzed by ¹³C isotope analysis. The ¹³C-labeled proteinogenic amino acids ratios at the P1 point should be zero, but some showed slight labeling. This was due to the incomplete resolution of compounds, hydrogen abstraction, and incomplete resolution in the time or m/z domain, causing inaccurate isotopomer ratios in proteinogenic amino acids (Fig. 4C and ref. 39). At the P2 point, the ¹³C-labeled ratios of the above proteinogenic amino acids were increased compared with those at the P1 point (Fig. 4C). The OD₆₀₀ at the P2 point was 10% higher than that at the P1 point. Cells taken at the P3 point showed slightly higher ¹³C-labeled proteinogenic amino acid ratios than those at the P2 point, while the OD₆₀₀ at P3 was slightly lower than that at the P2 point; however, the difference between the OD₆₀₀ values at P2 and P3 is within the error range in flask cultivation (see below for bioreactor cultivation giving clear results). These results suggest that cells were able to sustain growth on FA and CO₂ as the sole carbon sources by the combined use of the C1 assimilation pathway and C. boidinii Fdh even after glucose was depleted.

To examine whether continuous gas supplementation with an air–CO₂ mixture would enhance cell growth and FA consumption, the RG8 strain was cultivated in a 1.3-L bioreactor (initial working volume of 300 mL) with continuous sparging of the air–CO₂ mixture (SI Appendix, Materials and Methods), since both of O₂ and CO₂ are required for cell growth (to produce ATP through aerobic respiration and to synthesize glycine through the C1 assimilation pathway, respectively). The bioreactor cultivation was also performed in two stages, as in flask cultivation described above. At the first stage, the RG8 strain was cultivated in M9 minimal medium supplemented with 5 g⋅L⁻¹ glucose and 3.7 g⋅L⁻¹ sodium FA (the equivalent of 2.5 g⋅L⁻¹ FA). Cell growth of the RG8 strain was improved in bioreactor cultivation, resulting in a shorter lag phase than in flask cultivation (Fig. 4 A and D). The RG8 strain consumed FA rapidly, and FA was almost depleted when glucose was depleted (Fig. 4D). In contrast, the bioreactor-cultured control RG6 strain consumed FA much more slowly, and FA remained present after glucose depletion (Fig. 4E). Bioreactor cultivation of the RG8 strain showed the specific glucose and FA consumption rates of 370.2 and 145.6 mg gDCW⁻¹⋅h⁻¹, respectively, which were 59% and 34% higher than those obtained in flask cultivation. The specific CO₂ consumption rate was calculated to be 14.9 mg gDCW⁻¹⋅h⁻¹. The C1 assimilation pathway and Fdh reaction consumed 21.3 wt% and 78.7 wt% of FA, respectively. The percentage of FA consumed by the Fdh reaction was slightly higher than that observed in flask cultivation, suggesting that supplementation with the air–CO₂ mixture improved FA consumption by the Fdh reaction.

After the depletion of initially supplied glucose, the second stage of bioreactor cultivation was started from the F1 point (Fig. 4D). The M9 minimal medium was supplemented with a mixture of ¹³C-labeled FA and unlabeled FA (in a 1:1 molar ratio) at the F1 point; the FA mixture was used in the bioreactor experiment to reduce the cost of the rather expensive ¹³C-labeled FA. Then the FA mixture was supplemented once more when FA was almost depleted. Unlabeled FA was supplemented at the F2 point and once more afterward, as indicated in the fermentation profile (Fig. 4D); this experiment supplying unlabeled sodium FA was designed to examine decreased ¹³C labeling in proteinogenic amino acids, which would clearly prove that FA was consumed to sustain the growth of the RG8 strain in the absence of glucose. Samples for ¹³C isotope analysis were taken at the F1, F2, and F3 points; the F1 and F2 samples were taken right before FA mixture or unlabeled FA was supplemented (Fig. 4D).

The ¹³C-labeled ratios of proteinogenic amino acids at the F1 point should be zero, but, as in the flask cultivation results, some amino acids showed 1–2% of ¹³C-labeled proteinogenic amino acids ratios resulting from inaccurate isotopomer ratios in proteinogenic amino acids (see above and Fig. 4C). Compared with the samples at the F1 point, the ¹³C-labeled ratios of proteinogenic amino acids were slightly higher in the samples at the F2 point (Fig. 4C). The OD₆₀₀ at the F2 point was 9% higher than that at the F1 point (Fig. 4D). The ¹³C-labeled proteinogenic amino acids ratios were not as high as those in flask cultivation, since the FA mixture rather than pure ¹³C-labeled FA was used. Also, all the ¹³C-labeled ratios of proteinogenic amino acids, except for glutamate (which did not change), were decreased at F3 compared with those at the F2 point (Fig. 4C); the absence of change in glutamate labeling is likely due to the initial low ¹³C labeling (less than 2% at the F2 point). The OD₆₀₀ was 11% higher at the F3 point than at the F2 point (Fig. 4D). Except for glutamate, all the ¹³C-labeled ratios of proteinogenic amino acids were increased by ¹³C-labeled FA supplementation and were decreased by unlabeled FA supplementation. On the other hand, cell growth in the RG8 strain decreased when not supplemented with additional FA after glucose depletion (SI Appendix, Fig. S9A). Thus, it can be concluded that cell growth after glucose depletion was due to carbon and energy metabolism utilizing only FA and CO₂.

Next, we wanted to confirm that cell growth of the RG8 strain in M9 minimal medium without glucose was truly due to the use of FA and CO₂ rather than cellular materials (glycogen, among others.). The RG8 strain was cultivated in M9 minimal medium containing glucose, washed with fresh M9 minimal medium, and transferred to M9 minimal medium with or without FA supplementation. When cells were cultured in M9 minimal medium with FA supplementation, the OD₆₀₀ was increased by 57% in 66 h, consuming 1.65 g⋅L⁻¹ FA (SI Appendix, Fig. S9B). On the other hand, in cells cultivated in M9 minimal medium without FA supplementation the OD₆₀₀ increased by 21% in 25 h, after which no more cell growth was observed (SI Appendix, Fig. S9B). For comparison, in cells cultivated in M9 medium containing FA the OD₆₀₀ increased by 32% in 25 h (the time point when cells stopped growing in the absence of FA), suggesting that cell growth occurred using FA in addition to cellular materials. Also, as mentioned above, cell growth continued after 25 h, and the OD₆₀₀ increased by 57% in 66 h. These results clearly suggest that cells grew using FA and CO₂ (in addition to cellular materials) as the sole carbon and energy sources.

To further elaborate on some growth of cells in the absence of FA, glycogen concentrations in the RG8 strain cultivated with and without FA were measured. However, glycogen was not detected in either case. It should be noted that glycogen accumulation was not observed unless the glycogen biosynthetic pathway was amplified (40), even though the culture condition is not exactly the same as ours. Thus, the 21% increase in OD₆₀₀ seems to be due to the turnover of RNA, proteins, and other cellular materials, including cellular metabolites. In summary, the growth of the RG8 strain observed in a bioreactor in the presence of FA and CO₂ but without glucose was truly due to the consumption of FA and CO₂ as the sole carbon and energy sources (in addition to some use of cellular material turnover).

Discussion

Although previous studies demonstrated the use of THF cycle for FA assimilation, the efficiencies of FA assimilation have been rather low. In this study, we report the development of an engineered E. coli strain equipped with the rTHF cycle and reverse gcv pathway for efficient FA and CO₂ assimilation into serine and pyruvate. The THF cycle was reconstructed using E. coli native GlyA and three heterologous enzymes, Ftl, Fch, and Mtd, from M. extorquens. After confirming that the rTHF cycle assimilated FA more efficiently than the previously reported THF cycle, a reverse gcv pathway was established by knocking out gcvR and overexpressing gcvTHP genes to synthesize glycine from CO₂. An engineered E. coli strain equipped with the rTHF cycle and reverse gcv pathway was able to synthesize up to 96% of proteinogenic serine from FA and CO₂ in a medium also containing glucose. Serine was further converted to pyruvate by the native Sda. After additional engineering, the pyruvate-forming flux from FA and CO₂ was increased up to 14.9% of total pyruvate-forming flux. Establishment of the C. boidinii Fdh reaction reduced the amount of glucose required for energy and redox generation. The final engineered RG8 strain showed specific glucose, FA, and CO₂ consumption rates of 370.2, 145.6, and 14.9 mg⋅gDCW⁻¹⋅h⁻¹, respectively. It seemed possible that the RG8 strain was capable of sustaining growth from only FA and CO₂ in the absence of glucose.

Then, the C1 consumption efficiency of the strain developed in this study was compared with those of other studies. The total specific C1 consumption rate was determined by adding the specific FA consumption rate and specific CO₂ consumption rate; since FA can be considered as a liquefied form of CO₂, FA and CO₂ are assumed to be equivalent. The RG6 strain showed a total specific C1 consumption rate of 62.7 mg⋅gDCW⁻¹⋅h⁻¹, which was 2.6 and 2.8 times higher than the specific CO₂ consumption rates of cyanobacteria [24.1⋅mg gDCW⁻¹⋅h⁻¹ (41)] and E. coli equipped with the CBB cycle [22.5 mg⋅gDCW⁻¹⋅h⁻¹ (11)], respectively. Consequently, the RG6 strain equipped with the C1 assimilation pathway assimilates FA and CO₂ more efficiently than cyanobacteria or E. coli equipped with the CBB cycle. In the case of the RG8 strain showing reduced glucose dependence, the total specific C1 consumption rate was 160.5 mg⋅gDCW⁻¹⋅h⁻¹ (45.9 mg⋅gDCW⁻¹⋅h⁻¹ for the C1 assimilation pathway and 114.6 mg⋅gDCW⁻¹⋅h⁻¹ for the Fdh reaction), which shows much higher C1 consumption efficiency than in previous reports.

While we were revising this paper, two studies on the use of the THF cycle and reverse gcv pathway to synthesize pyruvate or serine from FA and CO₂ appeared (42, 43). In one study (42), Clostridium ljungdahlii Ftl, Fch, and Fol were used to construct the THF cycle. In addition, gcvTHP and sdaA were overexpressed, and native serA was deleted. However, the efficiency of FA and CO₂ assimilation into serine or pyruvate (∼10% of proteinogenic serine was synthesized only from FA and CO₂) was much lower than that reported here (∼98% of proteinogenic serine was synthesized only from FA and CO₂), likely because of the lower efficiencies of C. ljungdahlii enzymes in FA assimilation. In the other report (43), the same M. extorquens Ftl, Fch, and Mtd were utilized to construct the THF cycle. In addition, gcvTHP and lpd were overexpressed, and native serA was deleted. The efficiency of serine synthesis from FA and CO₂ (∼90% of proteinogenic serine) was high but was slightly lower than that reported in our study (98% of proteinogenic serine). It seems that the additional metabolic engineering strategies, including gcv operon repressor (gcvR) and pflB knockout, the change of the native promoter of gcv operon to a strong trc promoter, and overexpressing ftl, fch, mtd, gcvTHP, and lpd using a single plasmid, were effective in improving serine synthesis from FA and CO₂ in our study.

Theoretically, pyruvate can be synthesized from only FA and CO₂ if the C1 assimilation pathway continues to function. This was why glucose was used to supply the cellular energy and redox needed. To synthesize 1 mol of pyruvate from 2 mol of FA and 1 mol of CO₂, 0.297 mol of glucose is required (the detailed calculation is presented in SI Appendix, Text S4). Thus, the C1 utilization ratio can reach up to 0.627, since 3 mol of carbons come from FA and CO₂, and 1.782 mol of carbons come from glucose. However, the highest C1 utilization ratio achieved with the RG6 strain (0.115) is still much lower than the theoretical maximum (0.627) because the C1 assimilation pathway is much less efficient than the glycolytic pathway, as shown by the significantly lower specific C1 consumption rate (62.7 mg⋅gDCW⁻¹⋅h⁻¹) compared with the specific glucose consumption rate (319.7 mg⋅gDCW⁻¹⋅h⁻¹). This is the reason for the large gap between the percentage of serine biosynthesis (98%, corresponding to 100% as the labeled FA was 98% ¹³C-labeled) and pyruvate-forming flux (14.9%) from the C1 assimilation pathway; a large portion of pyruvate is produced from glycolysis, while serine is synthesized only from FA and CO₂ (for example in the RG5 and RG6 strains).

To develop an E. coli strain capable of growing solely from FA and CO₂ without glucose supplementation, further improvement of FA and CO₂ assimilation into pyruvate is needed. In our study, the C1 assimilation pathway flux has been improved mainly by enhancing the reverse gcv pathway. Thus, future study is needed to further improve FA and CO₂ assimilation into pyruvate by enhancing the FA and CO₂ assimilation efficiency of the rTHF cycle. One possible approach is achieving balanced overexpression of the M. extorquens ftl, fch, and mtd genes, since fch and mtd were not overexpressed as much as ftl (SI Appendix, Fig. S3 A and B). Also, it is known that enzymes involved in the reverse gcv reaction have low affinities toward CO₂ and NH₃ (21, 28, 29), which are the reactants of the reverse reaction. Thus, improvement of enzyme affinities toward CO₂ and NH₃ can be another promising solution to enhance FA and CO₂ assimilation by further improving the reverse gcv pathway flux. In addition, the E. coli central carbon metabolism needs to be modified for better growth on FA and CO₂. For example, improving metabolic flux from pyruvate to phosphoenolpyruvate might be a solution to better synthesize intermediates of upper-glycolysis utilized for synthesizing various cellular molecules, including nucleotides, amino acids, and cell walls.

In conclusion, we report the development of engineered E. coli strains capable of efficiently utilizing FA and CO₂ by introducing the rTHF cycle and reverse gcv pathway. The possibility of reducing the amount of glucose needed was also demonstrated by establishing the Fdh reaction. The strategies and E. coli strains developed in this study will be useful for further advancing C1 biorefinery toward the production of chemicals and materials from FA and CO₂.

Materials and Methods

All the materials and methods conducted in this study are detailed in SI Appendix, Materials and Methods, including reagents, media compositions, plasmid construction, gene knockout and native promoter change of the genomic DNA, protein expression experiments, FA assimilation analyses through the rTHF cycle, analysis of FA and CO₂ assimilation using ¹³C isotopes, flux analysis of the serine synthesis pathway, the FA tolerance test, determination of specific consumption rates of glucose, FA, and CO₂, bioreactor cultivation, and other analytical procedures. The data supporting the findings of this study are available in SI Appendix and Datasets S1–S3.