Genome-scale metabolic model (GEM) analysis to simulate production of fuculose-1-phosphate using E. coli

For GEM analysis, the core E. coli metabolic model was used to simulate growth [13]. The following reactions were included in the model to enable the production of l-fuculose: l-fuculose transport, l-fuculose production, l-fuculose phosphatase, l-fuculose kinase, fuculose-1-phosphate aldolase, 1,2-PDO oxidoreductase, and 1,2-PDO transport. MATLAB (version R2019b) and Cobra Toolbox (version 2.37.3) were used for analysis [14].

Construction of mutants

To selectively remove the target genes from the genome of the MG1655 strain, a single-step gene knockout procedure was conducted [15]. The primers used in this study are described in S1 Table. Annealing regions (20 bp) homologous to the kanamycin resistance cassette template plasmid, pKD4, were included in addition to 50 bp of a sequence homologous to regions flanking the target genes. The purified polymerase chain reaction (PCR) product obtained was electroporated into MG1655, which had previously been transformed with the lambda Red recombinase expression vector, pKD46, bearing a heat-labile origin. To facilitate recombination of the electroporated fragment, expression of lambda Red recombinase in MG1655 cells was induced by treatment with 50 mM arabinose at 30 °C. After the transformation of the PCR fragment into MG1655 harboring pKD46, the cells were recovered in Luria-Bertani (LB) broth at 37 °C. Kanamycin-resistant transformants were grown at 37 °C to induce the loss of the heat-labile pKD46 plasmid. Loss of the pKD46 plasmid was confirmed by loss of resistance to ampicillin. Final knockout isolates were screened for the absence of target genes and the presence of the kanamycin cassette using PCR. The kanamycin cassette was removed via flanked FRT recombination sites to obtain a markerless mutant using pCP20.

Construction of pCDF-fucA plasmid

The fucA fragment was amplified using PCR, with primer pairs fucA F/fucA R. The amplified PCR fragments were designed to share 15-nt homologous bases at each end. The pCDF Duet-1 plasmid was linearized using NdeI/XhoI restriction enzymes. The PCR fragments were fused with the plasmid DNA using the In-Fusion HD Cloning Kit (Clontech, Mountain View, CA, USA) to generate pCDF–fucA. Specifically, the In-Fusion cloning reactions were completed with 200 ng PCR fragment, 150 ng linearized plasmid DNA, 2.0 μL 5× In-Fusion HD Enzyme Premix, and double-distilled water (ddH₂O) for a final volume of 10 μL. The reaction mixture was incubated at 50 °C in a water bath for 15 min and then placed on ice for the transformation of competent DH5α E. coli cells. The pCDF–fucA plasmid was screened using colony PCR with the primer pair, pCDF V-F/V-R, and then sequenced.

Culture conditions

Engineered E. coli strains were cultivated in 50-mL baffled flasks containing 17 mL of LB medium supplemented with 10 g/L glucose, 5 mM LAD, and 50 μg/mL streptomycin. Cultures were incubated at 30 °C with shaking at 180 rpm. Gene expression was induced with 0.2 mM isopropyl β-d-1-thiogalactopyranoside.

Analyses of extracellular metabolites

Extracellular metabolites, including glucose, LAD, 1,2-propanediol (1,2-PDO), and acetic acid, were quantified using an Agilent 1100 HPLC system (Agilent Technologies, Santa Clara, CA, USA) equipped with a refractive index detector and an Aminex HPX-87H organic acid column (Bio-Rad, Hercules, CA, USA). Chromatographic separation was performed at 65 °C with 0.01 N H₂SO₄ as the mobile phase at a constant flow rate of 0.6 mL/min. The retention times were 8.8, 12.9, 14.8, and 17.1 min for d-glucose, LAD, acetic acid, and 1,2-PDO, respectively. Quantification was carried out using calibration curves generated from authentic analytical standards.

Identification and quantification of l-fuculose

To identify and quantify l-fuculose, an authentic l-fuculose standard and supernatants of engineered E. coli cultures were analyzed using gas chromatography–mass spectrometry (GC/MS) after derivatization steps, such as methoximation and silylation. For methoximation, 20 μL of methoxyamine hydrochloride solution (40 mg/mL in pyridine) was added to the dried metabolite samples, which were then incubated at 30 °C for 90 min. Subsequently, silylation was carried out by adding 50 μL of N-methyl-N-trimethylsilyl-trifluoroacetamide to the mixtures, which were then incubated at 37 °C for 30 min.

Metabolite analysis was performed on an Agilent 7890A gas chromatograph coupled to a 5975C quadrupole mass spectrometer (Agilent Technologies). A 1.0-μL aliquot of the derivatized sample was injected in splitless mode into an RTX-5Sil MS column (30 m × 0.25 mm i.d.; 0.25 μm film thickness; Restek, PA, USA) equipped with a 10-m guard column. The oven temperature program was: initial hold at 50 °C for 1 min, ramping to 330 °C at 20 °C/min, and final hold at 330 °C for 5 min. Mass spectra were acquired at 85–550 m/z with electron impact ionization at 70 eV.

In silico (GEM) analysis of E. coli for l-fuculose production

GEM analysis was conducted to predict the optimal flux distribution of E. coli for l-fuculose biosynthesis when LAD and glucose were used as substrates. Under aerobic conditions, the in silico model predicted that E. coli would simultaneously utilize LAD and glucose via FucA activity to generate l-fuculose-1-phosphate, which would subsequently be dephosphorylated to l-fuculose by innate sugar phosphatases (Fig 1A). The simulation further indicated that glucose concentrations equal to or greater than those of LAD were required to sustain l-fuculose biosynthesis (S2 Table), and that increasing oxygen availability would enhance both biomass formation and l-fuculose production (Fig 1B). In contrast, under anaerobic conditions, all LAD would be diverted to 1,2-PDO through FucO activity, resulting in no flux toward l-fuculose production (S2-S3 Table). These results suggest that sufficient oxygen supply is essential for efficient flux toward l-fuculose synthesis.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. In silico genome-scale metabolic model (GEM) of Escherichia coli for l-fuculose production.

(A) Simplified metabolic pathways of LAD and DHAP utilization leading to l-fuculose biosynthesis. Competing pathways and targeted gene deletions (fucI, fucK, fucO, aldA, tpiA) are indicated. (B) Predicted relationship between oxygen uptake rate, l-fuculose synthesis, and growth rate based on GEM analysis.

https://doi.org/10.1371/journal.pone.0349374.g001

GEM analysis also identified several competing metabolic pathways that could interfere with l-fuculose biosynthesis. Although LAD can be converted to l-fuculose-1-phosphate by FucA, it can also be diverted to 1,2-PDO by FucO or to lactic acid by AldA. The other substrate, DHAP, can be converted to glyceraldehyde-3-phosphate (G3P) by TpiA and directed into glycolysis. Furthermore, the synthesized l-fuculose itself can be isomerized to l-fucose by FucI or rephosphorylated to l-fuculose-1-phosphate by FucK. Although a previous study demonstrated that innate sugar phosphatases were capable of dephosphorylating l-fuculose-1-phosphate to l-fuculose, low phosphatase activity could lead to the reconversion of the former into LAD and DHAP through the reverse FucA reaction [1]. Collectively, these results highlight the potential metabolic sinks that limit the efficient channeling of LAD and DHAP toward l-fuculose production.

Strain engineering of E. coli and l-fuculose production

Based on the aforementioned GEM analysis, gene deletions were introduced to reduce competing pathways and retain l-fuculose as the final product. First, the fucI gene, which encodes an enzyme that converts l-fuculose into its isomer (l-fucose) and thereby generates a mixture of both sugars, was deleted. Second, the fucK gene, which encodes an enzyme that otherwise diverts l-fuculose into central metabolism by phosphorylating it to l-fuculose-1-phosphate, was deleted to enable the retention of l-fuculose as the final product. In addition, fucA was overexpressed under the control of the T7 promoter to catalyze the aldol condensation of LAD and DHAP to L-fuculose-1-phosphate. Overexpression of fucA was confirmed by SDS-PAGE analysis following IPTG induction (S1 Fig), showing a distinct band at the expected molecular weight. The resulting strain, designated Fuc 1 (ΔfucI ΔfucK, fucA overexpression), was cultivated in glucose–LAD medium for 96 h.

During fermentation, glucose was steadily consumed, cell density increased until 24 h and then entered the stationary phase, and acetic acid accumulated up to 24 h before stabilizing (Fig 2A). LAD was rapidly depleted within the first 12 h, whereas 1,2-PDO began to accumulate from 12 h, with its concentration increasing gradually until the end of fermentation. L-fuculose production was detected after 12 h and gradually increased throughout fermentation, reaching 38.92 ± 5.19 mg/L (Fig 3A). GC/MS analysis of the culture supernatant confirmed the presence of L-fuculose (Fig 3B, C).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Fermentation profiles of E. coli strains engineered for l-fuculose production.

(A) Fuc 1 (ΔfucI ΔfucK; fucA::T7). (B) Fuc 2 (ΔfucI ΔfucK ΔtpiA; fucA::T7). (C) Fuc 3 (ΔfucI ΔfucK ΔtpiA ΔfucO; fucA::T7). (D) Fuc 4 (ΔfucI ΔfucK ΔtpiA ΔfucO ΔaldA; fucA::T7). For each strain, cell growth (OD₆₀₀), glucose consumption, and acetic acid production are shown in the left panels, and LAD consumption with 1,2-propanediol (1,2-PDO) production is shown in the right panels. Error bars indicate standard deviations based on three independent experiments.

https://doi.org/10.1371/journal.pone.0349374.g002

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. L-fuculose production during fermentation and confirmation by mass spectrometry.

(A) L-Fuculose production by engineered E. coli strains (Fuc 1–4). Data are presented as mean ± standard deviation. (B) Mass spectrum of L-fuculose produced by the engineered strain. (C) Mass spectrum of the analytical L-fuculose standard.

https://doi.org/10.1371/journal.pone.0349374.g003

To increase intracellular availability of DHAP, the tpiA gene, which encodes an enzyme that converts DHAP to G3P for entry into glycolysis, was deleted in addition to fucI and fucK. The resulting strain, designated Fuc 2 (ΔfucI ΔfucK ΔtpiA, fucA overexpression), was cultivated under the same conditions as those for Fuc 1. Compared with Fuc 1, Fuc 2 exhibited markedly reduced growth, with glucose being consumed only gradually over 96 h, accompanied by lower biomass formation and acetic acid production (Fig 2B). LAD consumption was also delayed, requiring up to 48 h for complete depletion, whereas conversion into 1,2-PDO occurred steadily throughout the fermentation. L-fuculose production increased gradually during fermentation (Fig 3B), but the final concentration was lower than that obtained with Fuc 1, reaching 32.15 ± 3.62 mg/L, corresponding to a yield of 4.96 ± 0.94%. Deletion of tpiA can lead to DHAP accumulation and toxicity, thereby impairing cell growth [16,17]. Consistent with this observation, the severe growth defect observed in Fuc 2 likely resulted from elevated intracellular DHAP levels. Although the deletion of tpiA was expected to enhance flux toward l-fuculose, actual l-fuculose production decreased compared to that achieved using Fuc 1, suggesting that DHAP accumulation exerted a toxic effect that outweighed the potential benefits of increased substrate availability.

To further reduce LAD diversion into byproducts, the fucO and aldA genes were also deleted. The fucO encodes a 1,2-PDO oxidoreductase, which converts LAD into 1,2-PDO, whereas aldA encodes an aldehyde dehydrogenase, which converts LAD into lactate. The resulting strains were designated as Fuc 3 (ΔfucI ΔfucK ΔtpiA ΔfucO, fucA overexpression) and Fuc 4 (ΔfucI ΔfucK ΔtpiA ΔfucO ΔaldA, fucA overexpression). Both Fuc 3 and Fuc 4 displayed growth and glucose consumption patterns similar to those of Fuc 2, but LAD depletion was further delayed, with complete consumption occurring only after 72 h (Fig 2). Despite this delay, most LAD was still converted into 1,2-PDO. The L-fuculose production improved substantially compared to that achieved using Fuc 1 and Fuc 2. L-fuculose production increased during fermentation (Fig 3A). Fuc 3 produced 45.30 ± 7.37 mg/L L-fuculose (yield: 6.11 ± 0.99%), whereas Fuc 4 yielded 50.25 ± 4.30 mg/L (yield: 6.78 ± 0.58%). Compared to Saccharomyces cerevisiae (used in a previous study), which produced only 1.55 mg/L of l-fuculose [1], the Fuc 4 strain achieved more than a 30-fold improvement, suggesting the potential of E. coli as a host for L-fuculose biosynthesis. Although aldA deletion was intended to block the conversion of LAD to lactate, lactate levels were not directly quantified in this study. Therefore, the contribution of aldA deletion to reducing lactate formation remains to be further validated.

These results indicate that elimination of fucO and aldA reduced LAD catabolism to some extent and prolonged substrate availability, thereby enhancing l-fuculose production. The majority of LAD was still diverted to 1,2-PDO (Fig 2), suggesting that oxidoreductases other than that encoded by fucO may contribute to this conversion. In addition, the accumulation of acetate observed during fermentation likely reflects overflow metabolism in E. coli under high carbon flux conditions, which can divert carbon away from the desired product and reduce L-fuculose production efficiency [18]. The overall yields remained below 7% of the theoretical maximum, highlighting insufficient FucA activity and limited dephosphorylation efficiency of sugar phosphatases as critical metabolic bottlenecks. If l-fuculose-1-phosphate is not rapidly dephosphorylated, it can be reconverted into LAD and DHAP by the reverse FucA reaction, further restricting net l-fuculose production. Collectively, these findings suggest that further optimization—such as enhancing FucA and sugar phosphatase activities and identifying additional enzymes involved in LAD-to-1,2-PDO conversion—is required to achieve higher yields of l-fuculose from E. coli. Although this study was performed in shake-flask cultures, future studies involving bioreactor-scale fermentation will be important to validate the scalability and industrial relevance of the proposed strategy.