Efficient Production of 3′-Sialyllactose Using Escherichia coli

Abstract

3′-Sialyllactose (3′-SL), a key component of human milk oligosaccharides, provides significant health benefits and immune modulation, and is increasingly used in infant formula and dietary supplements. This study presents a novel approach for the efficient biosynthesis of 3′-SL using Escherichia coli BL21star(DE3)ΔlacZ through genomic integration. We first addressed the issue of metabolic competition by deleting crucial genes, nanA, nanK, nanE, and nanT, that are involved in the degradation of N-acetylneuraminic acid. This strategic gene knockout minimized the flux through competing pathways. The engineered Escherichia coli strain was subsequently transformed with the exogenous genes neuBCA and nST, enabling the de novo synthesis of 3′-SL. A modular metabolic engineering strategy was utilized to optimize the expression of key enzymes within the MSU module, enhancing and balancing the carbon flux distribution. Additionally, a cofactor regeneration strategy was implemented to increase CTP availability, which improved cofactor recycling and fine-tuned the metabolic pathway for maximal 3′-SL production. Transport protein screening was incorporated to further increase the extracellular concentration of 3′-SL, resulting in an unprecedented yield of 56.8 g/L in a 5L bioreactor fermentation, setting a new benchmark in the field.

1. Introduction

Human lactooligosaccharides (HMOs) are a group of oligosaccharides present in human milk, constituting the third major solid component after lactose and fat. They are primarily composed of five monosaccharides: d-glucose, d-galactose, N-acetylglucosamine, l-fucose, and N-acetylneuraminic acid (Neu5Ac) in varying proportions.¹ Over 200 different HMOs have been identified, with concentrations decreasing throughout the lactation period: approximately 20–25 g/L in colostrum and 5–20 g/L later in lactation.² HMOs are categorized into neutral core HMOs, fucosylated HMOs, and sialylated HMOs.³ Among the sialylated oligosaccharides, 3′-sialyllactose (3′-SL) and 6′-sialyllactose (6′-SL) are the most abundant, offering benefits such as improved gut health, antiadhesion properties, antiviral activity, and promotion of brain development. Their potential applications in infant formula and dietary supplements are substantial.

3′-SL biosynthesis can occur through three pathways: the NeuC pathway, the AGE pathway, and the NanE pathway. The common intermediate in these pathways is N-acetylmannosamine (ManNAc), which can be synthesized through three methods: UDP-GlcNAc can be irreversibly converted to ManNAc by UDP-GlcNAc 2-epimerase; GlcNAc can be reversibly converted to ManNAc by GlcNAc 2-epimerase; and GlcNAc-6-phosphate can be reversibly converted to ManNAc-6-phosphate by GlcNAc-6-phosphate 2-epimerase, which is then dephosphorylated to ManNAc.² 3′-SL is formed by the linkage of Neu5Ac to lactose via an α-(2,3) linkage.⁴ Neu5Ac and lactose are direct precursors for 3′-SL synthesis, with Neu5Ac being converted to CMP-Neu5Ac by CMP-Neu5Ac synthase. Thus, Neu5Ac is a critical precursor for efficient 3′-SL synthesis. 3′-SL exhibits significant biological activities and application potential in immune modulation,^5,6 gut microbiota interaction,^7,8 antiviral and antitumor effects,^9,10 and brain development.^11,12 Further research into these aspects will enhance understanding of its mechanisms and promote its use in human health.^7,13,14

Currently, 3′-SL can be synthesized via chemical synthesis or microbial fermentation. Chemical synthesis is complex and suited for laboratory-scale research, while microbial fermentation is promising for industrial-scale production due to its short production cycle and diverse substrate pathways. Microbial fermentation methods are categorized into de novo synthesis and remedial pathways. In 2002, Priem et al. elucidated the 3′-SL biosynthesis process by modifying Escherichia coli (E. coli) JM109, knocking out nanA and lacZ M15, and overexpressing CMP-NeuAc synthase and α2,3-SiaT from Neisseria meningitidis. Under continuous addition of glycerol, sialic acid, and lactose, they achieved a 3′-SL yield of 2.6 g/L.¹⁵ Sprenger et al. discovered that E. coli could utilize its internal metabolism to produce CMP-Neu5Ac from UDP-GlcNAc.¹⁶ By knocking out nanAKET, lacZ, and lacA, and overexpressing neuC, neuB, neuA, and α2,3-SiaT, they achieved a 3′-SL yield of 25.5 g/L without external Neu5Ac addition. Zhang et al. employed a plasmid-integrated dual strategy, introducing genes necessary for CMP-Neu5Ac synthesis and α2,3-SiaT into E. coli BL21(DE3)ΔlacZ, optimizing enzyme expression, and integrating the best α2,3-SiaT. Their strategy achieved a 3′-SL yield of 31.4 g/L.¹⁷

This study utilizes BL21star(DE3)ΔlacZ as the initial host, knocking out the nanAKET gene cluster to reduce competitive pathways. The neuBCA and nST genes are integrated to identify optimal combinations, and metabolic pathways are optimized using MSU modularization. The expression of UDP-GlcNAc-related synthesis genes glmM, glmS*, and glmU is increased, and suitable endogenous and exogenous transport proteins are screened. The regeneration of CTP is promoted through cofactor engineering, the limitation of insufficient supply is improved, the metabolic pathway is fine-tuned, and transporters suitable for 3′-SL are screened to maximize the synthesis of 3′-SL. Finally, the optimized strains are subjected to fermentation in a 5L bioreactor using a continuous feeding strategy, with glucose and lactose supplied to increase the maximum titer of 3′-SL.

2. Materials and Methods

2.1. Strains, Plasmids, and Media

All strains and plasmids used in this study are listed in Supporting Information (SI), Table S1. Gene cloning and plasmid construction were performed using E. coli DH5α, with E. coli BL21star(DE3)ΔlacZ serving as the initial host strain for the production of 3′-SL. Both strains were cultured in Luria–Bertani (LB) medium to prepare seed cultures. E. coli was grown in liquid LB medium containing 5 g/L yeast extract, 10 g/L tryptone, and 10 g/L NaCl at 37 °C with shaking at 230 rpm. When necessary, antibiotics were added to the media at the following concentrations: ampicillin, 100 μg/mL; kanamycin, 50 μg/mL; and spectinomycin, 50 μg/mL.

2.2. Gene Editing

The natural promoters of the genes glmS, cmk, and transport proteins were replaced with alternative promoters using homologous recombination. For example, using cmk as the template, primers cmk-F and cmk-R were employed to amplify the P_T7-cmk homologous arms. The pRSFDuet plasmid was linearized with EcoRI and HindIII, then ligated using homologous recombination enzymes and screened on LB agar containing kanamycin. Colony PCR with primers pRSF-DF and pRSF-DR was performed for identification, and plasmids from correctly sequenced strains were extracted. This plasmid was used as the template to amplify the pSPIN-P_T7-cmk homologous arms with primers T7 cmk-F and T7 cmk-R. The pSPIN plasmid was linearized using XhoI and PstI, ligated with homologous recombination enzymes, and screened on LB agar with kanamycin. Colony PCR with primers pSPIN-DF and pSPIN-DR was conducted for identification, and plasmids from correctly sequenced strains were extracted. The expression cassette containing the target gene was then cloned into pSL1765, which includes an N32-specific sequence. The resulting plasmid was used to transform recipient cells, with positive transformants selected on LB agar containing 25 μg/mL kanamycin. Correctly sequenced single colonies were grown in liquid LB to remove plasmids, resulting in plasmid-free engineered strains. Cloning was performed using 2 × ClonExpress Mix (Vazyme, Nanjing, China) and Gibson Assembly Kit (New England Biolabs, NEB). PCR amplification and validation were conducted using 2 × Phanta Max Master Mix (Vazyme, Nanjing, China). All primers used are listed in SI, Table S1.

Gene knockout in E. coli chromosomes was performed using an improved CRISPR-Cas9 system with pEcCas and pEcgRNA, which include Cas9 and λ-Red, respectively. pEcCas expresses λ-Red recombinase induced by l-arabinose (10 mM) for homologous recombination. The pEcgRNA, designed with sgRNA and N20 sequences using the CHOPCHOP Web site, targets the gene for editing. Next, 100 ng of pEcgRNA and 400 ng of DNA homologous arms were transferred into 100 μL of E. coli competent cells containing pEcCas via electroporation. The cells were then recovered in 1 mL LB at 37 °C, 180 rpm for 1 h and plated on LB agar with kanamycin and spectinomycin. Positive colonies were identified by colony PCR and gene sequencing. Finally, plasmids were lost to obtain plasmid-free engineered strains. The plasmids and primers used for gene editing are listed in SI, Tables S1 and S2.

2.3. Cultivation Conditions

To prepare the seed culture, 0.5 mL of overnight culture was inoculated into 50 mL of fresh LB medium. When the optical density at 600 nm (OD₆₀₀) reached 0.6–0.8, 1 mL of this culture was transferred into a 500 mL shaking flask containing 100 mL of defined medium for 3′-SL biosynthesis. The defined medium (DM) for 3′-SL production, with a pH of 6.80, consisted of 30 g/L glycerol, 17.9 g/L Na₂HPO₄·12H₂O, 3.1 g/L KH₂PO₄, 2.0 g/L NH₄Cl, 1.0 g/L (NH₄)₂HPO₄, 1.7 g/L citric acid, 15 mg/L CaCl₂, 2 g/L yeast extract, 15 g/L Oxoid tryptone, 2.2 g/L C₆H₅Na₃O₇·2H₂O, 1 g/L MgSO₄·7H₂O, 0.3 g/L Triton-X100, 10 mg/L vitamin B1, and 10 mL of trace element solution (containing 25 g/L FeCl₃·6H₂O, 2 g/L CaCl₂·2H₂O, 2.0 g/L ZnCl₂, 1.9 g/L CuSO₄·5H₂O, 0.42 g/L MnCl₂·H₂O, 2 g/L Na₂B₄O₇·10H₂O, and 2 g/L Na₂MoO₄·2H₂O). At an OD₆₀₀ of 0.6–0.8, isopropyl-β-D-1-thiogalactopyranoside (IPTG) and lactose were added to final concentrations of 0.2 mM and 15 g/L, respectively. The culture was induced at 28 °C for 48 h.

2.4. Fed-Batch Production in a 5 L Bioreactor

In a 5 L bioreactor, 1.65 L of defined medium was prepared, consisting of 10 g/L initial glycerol, 4.0 g/L (NH₄)₂SO₄, 9.2 g/L K₂HPO₄, 8.2 g/L KH₂PO₄, 0.3 g/L citric acid, 6.0 g/L tryptophan, 2.0 g/L yeast extract, 10 mg/L thiamine, 2.0 g/L MgSO₄·7H₂O, 0.02 g/L CaCl₂, and 10 mL/L trace element solution (the same as used in the shake flasks). A 0.15 mL aliquot of this culture was grown in a 1 L shake flask containing 150 mL LB medium for 6 h and then transferred to the bioreactor to prepare the seed culture. Throughout the cultivation process, stirring speed was automatically controlled to maintain dissolved oxygen levels between 30% and 50%. During the selection phase, the culture was maintained at 37 °C with a stirring rate of 900 rpm and an aeration rate of 2 VVM. Once the initial glycerol was completely consumed, glucose (800 g/L) was fed into the bioreactor at a constant rate of 3.8 g/L/h. The temperature was then reduced to 29.5 °C. After 4 h, IPTG (final concentration 0.2 mM) and 30 g/L lactose were added to induce 3′-SL biosynthesis. Throughout the fermentation, the pH was maintained at 6.80 using 28% ammonia, and after 12 h of induction, an additional 30 g/L of lactose was supplemented. The culture medium was sampled every 12 h to measure OD₆₀₀, lactose, and 3′-SL levels.

2.5. Analytical Methods

Cell growth was assessed by measuring the optical density (OD) at 600 nm using a UV-5100 spectrophotometer (Metash, China). A 2 mL aliquot of the culture was boiled for 5 min and then centrifuged at 12,000 g for 10 min. The supernatant was filtered through a 0.22 μm membrane for liquid chromatography analysis using HPLC. 3′-SL, lactose, glucose, and glycerol concentrations were measured by high-performance liquid chromatography (HPLC) (LC-16, Shimadzu, Japan), using a TSKgel Amide-80 column (Welch Materials, China) and a refractive index (RID-20A) detector (Shimadzu). A solution of 70% acetonitrile and 30% water was used as the mobile phase with a flow rate of 1 mL/min. The column was heated to 60 °C for the analysis of 100 μL diluted culture broth. All numerical results were repeated three times and are expressed as the mean ± SD. Data were analyzed using Origin 9.0 software (OriginLab, Northampton, MA, USA).

3. Result

3.1. Selecting E. coli Chassis Strain for 3′-SL Production

In the field of biosynthesis, selecting an appropriate host strain is crucial for achieving efficient production. In this study, we aimed to produce 3′-SL using E. coli and performed a detailed screening and comparison of six different strains: BL21(DE3), BL21star(DE3), JM109(DE3), DH5α, MG1655, and BW25113. Among these, BL21(DE3) and BL21star(DE3) are part of the BL21 series, while JM109(DE3), DH5α, MG1655, and BW25113 belong to the K12 series.

BL21(DE3) is renowned for its high expression capacity and stability, making it widely used in industrial production and suitable for large-scale protein expression. BL21star(DE3), an improved strain, has increased mRNA stability due to the LysS mutation, which reduces protein degradation and is suitable for high-level protein production. JM109(DE3) has lower but moderate expression levels, making it suitable for early screening and validation; it is also less sensitive to antibiotics, making it ideal for preliminary experiments. BW25113 has a complex genetic background, which is useful for creating gene knockout mutants and constructing plasmids but may present challenges in large-scale production. Because JM109(DE3), DH5α, and BW25113 have deletions in the chromosomal lac operon, they are unable to metabolize lactose. BL21(DE3) and BL21star(DE3) require the knockout of lacZ to eliminate β-galactosidase activity, thereby preventing the hydrolysis of lactose into galactose, which affects the production of 3′-SL.

This study involved integrating neuBCA from Campylobacter jejuni and nST from Neisseria gonorrheae to explore which E. coli strain is more suitable for 3′-SL production, as illustrated in Figure 1A. The results from shake flask fermentation after 48 h (Figure 1B) revealed that BL21 series strains, after lacZ knockout, performed better in 3′-SL production compared to K12 series strains. Among them, BL21star(DE3)ΔlacZ yielded the highest amount of 3′-SL at 1.88 g/L, which was 41.4%, 45.7%, and 38.2% higher than BL21 (DE3)ΔlacZ, BW25113, and MG1655ΔlacZ, respectively. Ultimately, BL21star(DE3)ΔlacZ was selected as the base strain for further development to achieve higher levels of 3′-SL production.

Figure 1.

Bioproduction of 3′-SL in E. coli strains. (A) Metabolic pathway of 3′-SL biosynthesis in engineered E. coli. Red × marks indicate knockouts; blue color indicates optimized expression; black dotted lines indicate multistep reactions; blue arrows indicate the NeuC pathway; green arrows indicate the AGE pathway; pink arrows indicate the NanE pathway; gray arrows indicate transmembrane transport. DHAP, 3-deoxy-d-arabinoheptulosonate 7-phosphate; GAP, glyceraldehyde 3-phosphate; PEP, phosphoenolpyruvic acid; TCA, tricarboxylic acid cycle; Fru-6P, fructose 6-phosphate; Fru-1,6P, fructose 1,6-diphosphate; GlcN-6P, glucosamine 6-phosphate; GlcN-1P, glucosamine 1-phosphate; GlcNAc-1P, N-acetylglucosamine 1-phosphate; GlcNAc-6P, N-acetylglucosamine 6-phosphate; UDP-GlcNAc, uridine 5′-diphospho-N-acetylglucosamine; GlcNAc, N-acetylglucosamine; ManNAc-6P, N-acetylmannosamine 6-phosphate; ManNAc, N-acetylmannosamine; Man-6P, mannose-6-phosphate; Neu5Ac, N-acetylneuraminic acid; CMP-Neu5Ac, cytidine 5′-monophosphate N-acetylneuraminic acid; 3′-SL, 3′-sialyllactose; pfkA, phosphofructokinase; pfkB, 6-phosphofructokinase II; fbp, fructose-1,6-bisphosphatase; glpx, fructose-1,6-bisphosphatase II; fbaAB, fructose-bisphosphate aldolase; tpiA, triosephosphate isomerase; pykA, pyruvate kinase; ppsA, phosphoenolpyruvate synthetase; lacY, lactose permease; nagA, N-acetylglucosamine-6-phosphate deacetylase; nagB, glucosamine-6-phosphate deaminase; glmS, glucosamine-6-phosphate synthase; glmM, phosphoglucosamine mutase; glmU, N-acetylglucosamine-1-phosphate uridyltransferase/glucosamine-1-phosphate acetyltransferase; GNA1, glucosamine-6-phosphate N-acetyltransferase; nanK, N-acetylmannosamine kinase; nanE, N-acetylmannosamine-6-phosphate epimerase; yqaB, a phosphatase with high specificity for N-acetylglucosamine-6-phosphate; age, N-acetyl-d-glucosamine 2-epimerase; neuC, UDP-GlcNAc 2-epimerase; neuB, Neu5Ac synthase; nanA, Neu5Ac aldolase; neuA, CMP-Neu5Ac synthetase; nanT, Neu5Ac transporter; nST, α2,3-sialyltransferase; manA, mannose-6-phosphate isomerase; lacZ, β-galactosidase; galP, H+ symporter; glpD, glycerol-3-phosphate dehydrogenase; glpF, glycerol facilitator; glpK, glycerol kinase; ptsG, glucose-specific PTS enzyme IIBC component. (B) Comparison of different E. coli host strains for 3′-SL production. The E. coli host strains include BL21(DE3) with lacZ knocked out, BL21star(DE3) with lacZ knocked out, JM109(DE3), DH5α, and MG1655 with lacZ knocked out, and BW25113. Each strain was transformed with neuBCA and nST integrated into the chromosome. Statistical significance was determined by one-way ANOVA followed by Tukey’s posthoc test; * p < 0.05, ** p < 0.01, *** p < 0.001. The highest 3′-SL production was observed in strain BL21star(DE3) with a yield of 1.88 ± 0.0862, which was significantly higher than that of other strains (e.g., BL21(DE3) with a yield of 1.33 ± 0.172, p < 0.05).

3.2. Constructing the de Novo Biosynthetic Pathway of the 3′-SL in the Engineered E. coli BL21star(DE3)

All three biosynthetic pathways for 3′-SL share ManNAc as a common intermediate. Most current research focuses on optimizing the neuC pathway, which uses UDP-GlcNAc as the precursor and includes GlcN-1P → UDP-GlcNAc → ManNAc → Neu5Ac → CMP-Neu5Ac → 3′-SL. Alternatively, the AGE pathway, which begins with GlcNAc, has been less studied,. Zhu et al. used Bacillus subtilis as the AGE pathway in the host to study the biosynthesis of 3′-SL.¹⁸ Thus, our study focuses on the neuC pathway.

The biosynthetic pathway from ManNAc to Neu5Ac can proceed via two principal routes, catalyzed by Neu5Ac synthase (NeuB) and Neu5Ac aldolase (NanA), respectively. It is noteworthy that the reaction catalyzed by NanA is reversible, which may lead to an ineffective cycle of precursor substances during metabolism. The N-acetylneuraminate lyase, encoded by the nanA gene, participates in the metabolic breakdown of Neu5Ac precursors, thereby inhibiting the biosynthesis and accumulation of CMP-Neu5Ac. Knocking out the nanA gene can effectively prevent the degradation of Neu5Ac into ManNAc and pyruvate. Neu5Ac can be transported by the action of the NanT enzyme and can also be converted into NeuAc within the cell, providing nutrition for E. coli. Knocking out the nanT gene to block this transport thereby affects the cell’s utilization and metabolism of Neu5Ac. In the metabolic process of Neu5Ac, NanA, N-acetylmannosamine kinase (NanK), and N-acetylmannosamine-6-phosphate epimerase (NanE) jointly catalyze its degradation to form GlcNAc-6P. Subsequently, under the action of GlcNAc-6P deacetylase (NagA) and GlcN-6P deaminase (NagB), GlcNAc-6P is further degraded into fructose-6P (Fru-6P). NanK promotes the metabolic conversion of ManNAc to ManNAc-6-P. Meanwhile, NanE is involved in the process of utilizing N-acetylneuraminate and N-acetylmannosamine as carbon sources and can act as an allosteric activator of glucosamine-6-phosphate deaminase, further promoting the metabolic process. We used lacZ-deficient E. coli BL21star(DE3) to knock out the genes nanA, nanK, nanE, and nanT involved in Neu5Ac degradation, creating the initial BS0 strain. Subsequently, the neuB, neuC, and neuA genes from Campylobacter jejuni and the nST gene from Neisseria gonorrheae were regulated by P_T7 promoter. Single copies were integrated into the caiB and hlyE sites of E. coli genome, resulting in the BS1 strain, which produced 2.47 g/L of 3′-SL after 48 h of shake flask fermentation (Figure 2B). Building on BS1, we further increased the copy number of neuBCA and nST to find the optimal combination, neuBCA is integrated into yjiP and yjiV sites, and nST is integrated into ydeU and intQ sites. The BS4 strain, with 2 copies of neuBCA and 1 copy of nST, achieved a 3′-SL yield of 3.68 g/L, a 59.1% increase compared to BS1, indicating that this combination was optimal for overexpressing neuBCA and nST (Figure 2A). However, increasing the copies of neuBCA and nST to three each resulted in a notable decrease in 3′-SL yield, likely due to upstream expression flux limitations affecting downstream production. Overall, optimizing the copy numbers of downstream genes led to higher yields and demonstrated the efficacy of this approach for 3′-SL biosynthesis.

Figure 2.

Optimization of 3′-SL production through multicopy number combinations. (A) Different copy number combinations of neuBCA and nST. (B) Comparison of BS1–BS9 in terms of 3′-SL production and cell growth in shake flasks. “neuBCA:nST” denotes the various copy number ratios of neuBCA to nST.

3.3. Enhancing the Biosynthesis of UDP-GlcNAc by Modifying the MS*U Module within the neuC Pathway

To enhance the upstream flux of the neuC pathway, we optimized the MSU module to directly boost the expression of UDP-GlcNAc synthesis genes glmM, glmS*,¹⁹ and glmU, which are crucial for increasing UDP-GlcNAc supply and addressing the issue of ManNAc export from the cell (Figure 3A). UDP-GlcNAc is a key precursor in the 3′-SL biosynthesis pathway.²⁰

Figure 3.

Effect of critical enzyme expression in the precursor pools of the MSU module on 3′-SL. (A) UDP-GlcNAc metabolic supply module within the 3′-SL biosynthetic pathway. (B) Enhanced UDP-GlcNAc supply through the combination and modular optimization of glmM, glmS, and glmU using the P_T7 promoter. (C) Comparison of 3′-SL production and cell growth in shake flasks with BS4T1–BS4T7. (D) Enhanced UDP-GlcNAc supply through the combination and modular optimization of glmM, glmS*, and glmU using the P1 promoter. (E) Comparison of 3′-SL production and cell growth in shake flasks with BS4P1–BS4P11.

Initially, we used P_T7 promoter expression box to integrate glmM, glmS* and glmU genes into pseudogene loci ykgH, XA and AH. While P_T7-driven glmS* showed some positive effects, it was not significant. Subsequent combinatorial approaches demonstrated that the BS4T5 strain, with glmM, glmS*, and glmU expression, produced 5.10 g/L of 3′-SL, a 1.35-fold increase over BS4T2 (Figure 3C).

Liu et al. designed three promoters with varying expression strengths to regulate the GlmM module and the GlmU-GlmSA module and ultimately found that glmM benefits the entire metabolic pathway only when regulated by the low-intensity promoter P1.²¹ Additionally, the team found that when the GlmM module is maintained at a low expression level and the GlmU-GlmSA module is at a medium level, more carbon sources are introduced into the Neu5Ac biosynthetic pathway. To explore the impact of this strategy on the yield of 3′-SL, we compared the low-intensity P1 promoter with the high-intensity P_T7 promoter for overexpressing glmM and also investigated the effects of overexpressing glmS* and glmU individually, co-overexpressing glmSU, and co-overexpressing all three genes glmMSU on 3′-SL (Figure 3D). Inspired by Liu et al.’s work on promoter strengths, we employed a low-strength P1 promoter for glmM and high-strength P_T7 for glmS* and glmU. The results showed that P1-driven glmM expression was beneficial for 3′-SL production (Figure 3E), outperforming high-strength P_T7 promoter combinations. Additionally, individual overexpression of glmS* and glmU was more effective for 3′-SL production than their coexpression, indicating that single-gene overexpression is more advantageous for precursor supply.

We tried to use the medium-strength P_tac promoter expression box to integrate the regulatory glmS* and glmU at the XA and AH sites, consistent with Liu’s findings that high expression is not always optimal. The strain BS4P8, with optimized promoter conditions, achieved a 3′-SL yield of 5.48 g/L, a 1.49-fold increase over BS4. Thus, optimizing the expression of glmM, glmS*, and glmU significantly improves UDP-GlcNAc supply and enhances 3′-SL biosynthesis.

3.4. Promoting CTP Regeneration through Cofactor Engineering

In the biosynthesis pathway of 3′-SL, NEUA-catalyzed enzyme requires CTP as a cofactor. CTP is crucial not only for the de novo synthesis of 3′-SL but also for the synthesis of cellular DNA and RNA.²² Endo et al. established a large-scale production system for CMP-NeuAc and sialic acid oligosaccharides by overexpressing CMP-NeuAc synthetase and CTP synthetase genes in whole-cell reactions.²³ In this study, we optimized the supply of CTP for product synthesis by either overexpressing genes related to CTP synthesis or reducing CTP metabolic consumption, thus constructing a coupled ATP and CTP regeneration cycle to enhance cofactor regeneration and increase 3′-SL fermentation yield (Figure 4A).

Figure 4.

Fine-tuning of 3′-SL synthesis through the optimization of key gene expression levels. (A) The metabolic pathways involved in coupling ATP and CTP regeneration cycles. (B) Enhancement of 3′-SL biosynthesis via a CTP regeneration strategy. The symbol “–” indicates that the gene was not modified; “+” denotes gene overexpression; “++” signifies that the gene was overexpressed with a copy number of 2.

Based on the engineered strain BS4P8, we overexpressed four genes from E. coli: cytidine monophosphate kinase (CMK, EC: 2.7.4.25), which catalyzes the conversion of CMP to CDP; nucleoside diphosphate kinase (NDK, EC: 2.7.4.6), which catalyzes the conversion of CDP to CTP; CTP synthase (PyrG, EC: 6.3.4.2), which catalyzes the ATP-dependent conversion of UTP to CTP; and polyphosphate kinase (PPK1, EC: 2.7.4.1), which catalyzes the conversion of terminal phosphate groups from ATP to long-chain polyphosphates. We tested the effects of individual and combined overexpression of these genes at the motA, intQ, and ydeU sites. We found that overexpression of CMK in strain BS4P12 and NDK in strain BS4P13 increased 3′-SL production to 6.12 and 5.92 g/L, respectively, representing increases of 11.7% and 8.1%. Conversely, strains overexpressing PPK1 (BS4P14) and PyrG (BS4P15) showed reduced 3′-SL yields of 3.69 and 3.38 g/L, respectively, a decrease of 32.7% and 38.3% compared to the control strain BS4P8(Figure 4B). This suggests that the regulation of PPK1 and PyrG did not yield the expected results, possibly due to the need for more precise regulation of the CTP regeneration pathway to minimize interference with host metabolism. Studies have shown that CTP synthase PyrG, a homologue of E. coli CtpS, can form filamentous structures that inhibit its activity, and enhancing multiple gene combinations did not positively impact yield.^24,25

Given that CMK and NDK enhance 3′-SL synthesis, we further explored the effects of overexpressing CMK and NDK in double copies and their combination. However, the results showed that increasing copy numbers or combining the genes did not further enhance 3′-SL production and, within certain ranges, led to inhibition. Overall, overexpression of CMK significantly enhanced 3′-SL production. CMK is involved in the salvage synthesis of pyrimidine nucleotides, catalyzing the transfer of phosphate groups from ATP to CMP to produce CDP, which plays a crucial role in DNA and RNA synthesis.

3.5. Fine-Tuning the Metabolic Intensity of Key Pathways

Starting with the BS4P12 strain, we fine-tuned the metabolic pathways to enhance 3′-SL production. In the 3′-SL biosynthetic pathway, the nagB gene encodes glucosamine-6-phosphate deaminase, which catalyzes the conversion of the precursor GlcN-6-P to Fru-6-P. We attempted to knock out the nagB gene, resulting in the BS4P21 strain. The 3′-SL yield of BS4P21 was 6.24 g/L, showing only a modest increase compared to the control.

The wecB gene encodes UDP-N-acetylglucosamine 2-epimerase, which converts UDP-GlcNAc, a key precursor for NeuAc synthesis, into UDP-N-acetylmannosamine (UDP-ManNAc). The intention behind knocking out the wecB gene was to prevent the accumulation of UDP-GlcNAc in the NeuAc biosynthesis pathway. However, the 3′-SL yield in the BS4P22 strain decreased.

To minimize PEP metabolism losses and avoid excessive carbon flux conversion into organic acids, we knocked out the pykA gene, which encodes pyruvate kinase, in the BS4P12 strain and overexpressed the ppsA gene, resulting in the BS4P23 and BS4P27 strains. Flask fermentation results showed that reducing the conversion of PEP to PYR did not positively affect 3′-SL production. Instead, the 3′-SL yields decreased to 5.01 g/L in BS4P23 and 1.83 g/L in BS4P27, representing reductions of 1.22-fold and 3.34-fold, respectively (Figure 5E). This suggests that these modifications may have led to a redistribution of carbon flux, energy imbalance, and impacts on cell growth.²⁶

Figure 5.

Fine-tuning the metabolic intensity of key pathways. (A) Dynamic control of 3′-SL production in E. coli through the repression of pfkA and activation of glmS. (B) Genome modification at the pfkA gene promoter by integrating negative-response biosensor-controlled promoters, pL19. (C) Comparison of engineered strains with deletions of selected genes. (D) Application of the flux biosensor in 3′-SL biosynthesis. (E) 3′-SL and OD₆₀₀ values of strains with pykA knockout and ppsA overexpression. (F) 3′-SL and OD₆₀₀ values of combination strains. (G) 3′-SL and OD₆₀₀ values of strains with lacY overexpression.

The pfkA and pfkB genes encode 6-phosphofructokinase, which is involved in the conversion of Fru-6-P to FBP. To shift more carbon flux from central carbon metabolism and other competing pathways to the Fru-6-P node in 3′-SL production, we individually knocked out the pfkA and pfkB genes, obtaining strains BS4P24 and BS4P25. The results indicated that knocking out pfkA significantly decreased 3′-SL yield. However, knocking out pfkB led to a substantial increase in 3′-SL production, with the BS4P25 strain achieving a 3′-SL yield of 6.85 g/L, a 1.21-fold increase compared to the control (Figure 5C).

It is important to note that glycolysis is a major metabolic pathway in all organisms. Maintaining a balance between glycolytic flux and the biosynthesis pathway is crucial when constructing microbial cell factories. Regulating glycolytic flux to address metabolic imbalance is rare. Zhu et al. designed and constructed a dual-function glycolytic flux biosensor capable of dynamically adjusting glycolytic flux to achieve high-level production of biochemical substances.²⁷ They created and modified a series of biosensors with positive and negative responses to fit various thresholds and dynamic ranges. These engineered biosensors were validated through multiple experiments to characterize fructose-1,6-bisphosphate levels. By replacing the pfkA promoter, they enhanced the yield of the target product GlcNAc and suggested that increasing glmS expression would further improve results (Figure 5A,B). We applied this strategy to enhance 3′-SL production, resulting in the BS4P26 strain. The results showed no significant change in yield compared to the control, but the OD₆₀₀ value was notably reduced, which helped mitigate the issue of excessive wet weight during fermentation. The 3′-SL production of the BS4P26 strain increased by 10.8% compared to the BS4P12 strain, while the OD₆₀₀ decreased by 11.7%. This indicates that dynamic regulation of pfkA leads to greater flux into the 3′-SL production pathway (Figure 5D).

We further combined the knockouts of nagB and pfkB with the replacement of the pfkA promoter to obtain the strain BS4P30. During flask fermentation, we successfully increased 3′-SL production. The BS4P30 strain produced 7.22 g/L of 3′-SL in shake flasks, which is 1.1 times higher than that of the BS4P12 strain, and the OD₆₀₀ increased by only 6% (Figure 5F). To enhance the utilization efficiency of lactose, we overexpressed the lacY gene under the control of the P_T7 promoter, resulting in the construction of the BS4P31 strain. Notably, during shake flask fermentation, this strain produced 7.55 g/L of 3′-SL, while the OD₆₀₀ value decreased by 13.7%. Although this outcome represents a modest increase in 3′-SL production, the improvement is not substantial. Consequently, we hypothesize that intracellular accumulation of 3′-SL may inhibit its further synthesis. Based on this, we believe that transporter engineering plays a crucial role in facilitating product excretion and enhancing the yield of metabolic products (Figure 5G).

3.6. Screening of Export Genes to Increase 3′-SL Production

An exemplary industrial microbial strain must possess not only robust product synthesis capabilities but also efficient product secretion mechanisms. We postulate that intracellular accumulation of 3′-SL may suppress its own synthesis. Engineering of transporter proteins is pivotal for enhancing product excretion and boosting the yield of metabolic products. Research on the HMOs efflux system remains relatively scarce. Hollands et al. identified CDT2 from Kluyveromyces lactis, which enhances the efflux of 2′-FL in yeast.²⁸ Sugita et al. discovered four endogenous E. coli transport genes (setA, setB, ydeA, and mdfA), whose overexpression markedly increased the production of LNT II.²⁹ Liao et al. demonstrated that the heterologous transport gene CmSET positively affects the efflux of LNnT, achieving an 89.14% increase in yield in a 5L bioreactor.³⁰ However, no studies have yet focused on 3′-SL transport proteins. Consequently, we endeavored to identify efficient 3′-SL efflux proteins in E. coli by leveraging an efflux gene enhancement strategy to augment 3′-SL production. We selected 18 proteins, six of which were heterologous and 12 endogenous, all recognized as transport proteins, and enhanced their expression by replacing their native promoters with a potent T7 promoter. We then investigated the effects of this overexpression on E. coli biomass and 3′-SL yield. Song et al. discovered that inactivation of tolR could lead to membrane instability and the formation of outer membrane vesicles (OMVs).³¹ Microscopic analysis of the mutant strain showed a distinct separation between the cytoplasmic matrix and the cell membrane, along with an increase in cell membrane fluidity.

In genetic engineering, functional genes that encode transport proteins or complexes facilitating the transmembrane transport of target molecules are essential for the genetic modification of bacterial cells. These genes can be either native or introduced from external sources. The transport proteins that mediate the translocation of target substances across the bacterial cell’s outer membrane typically involve porins, which form channels in the outer membrane. These porins are characterized by a ″β-barrel” structure, comprised of β-sheets. The exterior-facing nonpolar amino acid residues allow the porins to integrate seamlessly into the lipid bilayer of the outer membrane, while the polar residues on the interior side form an aqueous channel. Moreover, sugar transport proteins facilitating the transmembrane transport of carbohydrates can be constituted by a single polypeptide or a complex of multiple polypeptide subunits. These subunits assemble into homo- or hetero-oligomeric structures responsible for the transmembrane transport of carbohydrates. These sugar transport proteins orchestrate the movement of target oligosaccharides from the bacterial cytoplasm across the inner membrane to the periplasmic space, a process critical for the cell’s metabolic functions and interactions with its environment.³² Through genetic engineering technology, we successfully express the efflux gene of 3′-SL efficiently to promote the synthesis of oligosaccharides in the cell and enhance its transport ability across the plasma membrane. Specifically, the system includes genes encoding transporters, which are responsible for transporting 3′-SL from the cytoplasm to the periplasm, and genes encoding outer membrane porins, which facilitate further transport of 3′-SL from the periplasm to the fermentation medium. This optimization process significantly increased the concentration of 3′-SL in the fermentation medium, resulting in significant efficiency gains in the production process. We systematically screened both endogenous and exogenous proteins, including seven outer membrane porins (OmpN, OmpC, OmpG, OmpG*(E15H-E174H), OmpA, OmpF, and ChiP) and four multidrug efflux pump proteins (MdfA, MdtM, MdtL, MdtG, YcaD, YajR, and YnfM). The exogenous proteins included two outer membrane porins, namely scrY (AHM80897.1) from Klebsiella pneumoniae and CymA (WP 004136082.1) from Klebsiella oxytoca, as well as four multidrug efflux pump proteins: YbSET (GenBank: EEQ08298.1) from Yersinia bercovieri ATCC 43970, SmSET (WP_197791067.1) from Serratia marcescens, RnSET (WP_241579217.1) from Rosenbergiella nectarea, and CmSET (WP_024910347.1) from Chania multitudinisentens, all selected through the NCBI BLAST database. We expressed these 18 genes using the P_T7 promoter expression cassette, where endogenous proteins’ expression was enhanced by replacing their native promoters with the strong T7 promoter, while exogenous proteins were obtained through genomic integration. This approach led to the construction of strain BS4P32–51, and further, the tolR gene was knocked out to generate strain BS4P52 to investigate the impact of these modifications on 3′-SL production. Results from shake flask fermentation indicated that among the 14 transport proteins screened, most did not significantly enhance 3′-SL production. However, overexpression of the endogenous protein YcaD (BS4P43) and the exogenous protein ScrY (BS4P46) significantly enhanced the efflux of 3′-SL, resulting in the production of 7.71 g/L and 8.12 g/L of 3′-SL, respectively, which was a 7.5% increase compared with the control strain BS4P46. However, no further increase in 3′-SL titer was observed when these two proteins were coexpressed, which may be due to disruption of the transport system or metabolic burden caused by overexpression of the proteins. In comparative analysis, despite successfully integrating the four exogenous proteins YbSET, SmSET, RnSET, and CmSET into the genome, the resulting strains did not produce 3′-SL during shake flask fermentation. We hypothesize that expression of these exogenous proteins may have disrupted the transport of 3′-SL. Furthermore, in the tolR gene knockout strain B34P53, the production of 3′-SL did not significantly improve compared to the control strain BS4P31. We speculate that the knockout of the tolR gene may have induced membrane instability, potentially affecting the yield of 3′-SL. Based on these observations, we further screened for transport proteins and ultimately identified ScrY as the most suitable candidate. The strain BS4P46, which overexpresses scrY, achieved a 3′-SL yield of 8.12 g/L, which represents a 7.5% increase compared to the yield level of BS4P31, indicating a significant enhancement in production efficiency. Concurrently, the OD₆₀₀ values of strain BS4P46 also showed a decrease, which may reflect a trade-off between cell growth and product synthesis (Figure 6B–E).

Figure 6.

Enhancement of strain efflux capacity. (A) Model of the efflux phospholipid bilayer in E. coli. (B) Effect of overexpressing endogenous outer membrane porins. (C) Effect of overexpressing endogenous outer membrane glycoproteins. (D) Effect of overexpressing exogenous transporters. (E) Effects of overexpressing transporters and knocking out tolR on strain efflux capacity.

This finding underscores the importance of considering cellular physiological states in metabolic engineering and highlights the need for careful assessment of genetic modifications that may impact cell health. Our results not only provide a new strategy for the biosynthesis of 3′-SL but also offer valuable insights for optimizing production processes on an industrial scale in the future.

3.7. Production of 3′-SL in a 5 L Bioreactor

To demonstrate the large-scale production potential of the strain BS4P46 of 3′-SL, we cultivated the engineered strain in a 5 L bioreactor with intermittent fed-batch glucose supplementation. Rapid production of 3′-SL was achieved between 25 to 35 h of cultivation, followed by a slower growth phase from 71 to 119 h, culminating in a plateau at 107 h with a yield of 56.8 g/L of 3′-SL, which represents a 7-fold increase compared to shake flask cultivation. The overall productivity and specific productivity of 3′-SL were 0.19 g/L/h and 1.01 g/g DCW, respectively, with a lactose conversion rate of 0.55 mol 3′-SL/mol. In exploring the optimal feeding strategy for the 5 L bioreactor, we compared different ratios of glucose and glycerol feeding regimens and found that the highest 3′-SL yield was achieved with a feeding formula containing only glucose. We speculate that this may be due to the more complex metabolic pathways of glycerol compared to glucose, leading to lower utilization efficiency (SI, Figure S1). To the best of our knowledge, this yield is the highest reported to date for 3′-SL production and represents the first method utilizing an integrated strain strategy for 3′-SL production. Compared to current methods that employ plasmid-bearing strains as chassis cells for 3′-SL production, our approach using an integrated strain offers greater stability and a longer fermentation period, making it more suitable for large-scale fermentation processes.

In summary, this study has successfully developed a green and sustainable process for the biosynthesis of 3′-SL from glucose through systematic metabolic engineering strategies. We constructed the initial strain for the de novo synthesis pathway of 3′-SL using a gene integration strategy for the first time and optimized the multicopy number combination of neuBCA and nST genes. We introduced a modular engineering strategy to efficiently express precursor pool genes and enhanced the regeneration of cofactors to improve the supply of CTP. Furthermore, we blocked key precursor competitive pathways, dynamically regulated pfkA to balance the distribution of carbon flux in the metabolic network, and screened for transport proteins with positive effects to enhance the strain’s excretion capacity. These comprehensive strategies led to the development of a fermentation-stable engineered strain BS4P46, which for the first time overcame the technical bottlenecks of plasmid-containing bacteria in industrial applications. In this study, we employed a starvation fermentation strategy, which resulted in a complete depletion of glucose in the fermentation broth, as detected at 0 g/L. This finding indicates an efficient utilization of glucose by the strain during fermentation. At the scale of a 5-L fermenter, the BS4P54 strain achieved a 3′-SL titer of 56.8 g/L, setting a new record for the highest yield of 3′-SL produced via a synthetic biology approach (Figure 7). Nonetheless, from an industrial production perspective, there is still room for improvement in the production efficiency of 3′-SL. Future research should focus on further reducing the intracellular accumulation of 3′-SL and decreasing the cell concentration to achieve further optimization of the synthesis process and enhance production efficiency. This could be accomplished by optimizing metabolic pathways, enhancing the activity of key enzymes, or improving fermentation conditions to meet the demands of industrial production.

Figure 7.

Production of 3′-SL in a 5-L bioreactor.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.jafc.4c08703.

• Biological strains, plasmids, primers for construction and knockout used in this study and investigated the production of 3′-SL under various growth conditions in a 5L fermenter with different formulations (PDF)

Author Contributions

Xinyang Lv: results analysis, data curation, and writing of theoriginal draft; Yifan Liu: data curation; Jinyong Wu:experimental supervision and formal analysis; Lixia Yuan:software; Xiangsong Chen: supervision and funding acquisition; and Jianming Yao: supervision and conceptualization.

This study was funded by the Anhui Provincial Postdoctoral Researchers Research Activity Grant (2020B440).

The authors declare no competing financial interest.