Synergetic Fermentation of Glucose and Glycerol for High-Yield N-Acetylglucosamine Production in Escherichia coli

N-acetylglucosamine (GlcNAc) is an amino sugar that has been widely used in the nutraceutical and pharmaceutical industries. Recently, microbial production of GlcNAc has been developed. One major challenge for efficient biosynthesis of GlcNAc is to achieve appropriate carbon flux distribution between growth and production. Here, a synergistic substrate co-utilization strategy was used to address this challenge. Specifically, glycerol was utilized to support cell growth and generate glutamine and acetyl-CoA, which are amino and acetyl donors, respectively, for GlcNAc biosynthesis, while glucose was retained for GlcNAc production. Thanks to deletion of the 6-phosphofructokinase (PfkA and PfkB) and glucose-6-phosphate dehydrogenase (ZWF) genes, the main glucose catabolism pathways of Escherichia coli were blocked. The resultant mutant showed a severe defect in glucose consumption. Then, the GlcNAc production module containing glucosamine-6-phosphate synthase (GlmS*), glucosamine-6-phosphate N-acetyltransferase (GNA1*) and GlcNAc-6-phosphate phosphatase (YqaB) expression cassettes was introduced into the mutant, to drive the carbon flux from glucose to GlcNAc. Furthermore, co-utilization of glucose and glycerol was achieved by overexpression of glycerol kinase (GlpK) gene. Using the optimized fermentation medium, the final strain produced GlcNAc with a high stoichiometric yield of 0.64 mol/mol glucose. This study offers a promising strategy to address the challenge of distributing carbon flux in GlcNAc production.


Introduction
N-acetylglucosamine (GlcNAc) is the monomer unit of chitin, which is the second most abundant polysaccharide on Earth and can be commonly found in crustaceans, fungi and insects [1]. It is also a basic component of various heterologous biopolymers, such as hyaluronic acid and chondroitin sulfate, which play important roles in cartilage and joint health [2,3]. Furthermore, the GlcNAc molecule can be frequently observed in glycoproteins, mammalian growth factors and hormones, which are directly involved in a broad range of physiological functions [4]. Due to its unique characteristics, GlcNAc and its derivatives have received extensive attention for their commercial applications in the healthcare, cosmetics and pharmaceutical industries [5].
Traditionally, GlcNAc is produced through chemical and enzymatic hydrolysis of crustacean shells [6,7]. However, there are several drawbacks to these extraction processes, such as the limitation of raw material supply and severe environmental pollution. Moreover, GlcNAc is difficult to extract from crab and shrimp shells without allergenic risk for individuals who suffer from shellfish allergies. In recent years, microbial production of Glc-NAc has drawn tremendous attention, as it is a promising alternative to the production of non-shellfish-derived GlcNAc in a low-cost and environmentally compatible manner [8][9][10]. Several microorganism species have been evaluated for GlcNAc production, including Escherichia coli [11], Bacillus subtilis [12], Saccharomyces cerevisiae [13], Lactobacillus plantarum [14] and Corynebacterium glutamicum [15]. The biosynthesis pathway of GlcNAc from the precursor fructose-6-phosphate (F-6-P) involves three crucial enzymes, glucosamine-6phosphate synthase (GlmS), glucosamine-6-phosphate N-acetyltransferase (GNA1) and GlcNAc-6-phosphate phosphatase (Figure 1). Various metabolic engineering strategies have been applied to improve GlcNAc production and the current efforts are focused largely on enhancing the GlcNAc biosynthesis pathway through key enzyme screening and overexpression, deleting by-product biosynthetic pathways, blocking catabolism of intracellular GlcNAc and engineering transcription factors [15][16][17]. The key precursor for GlcNAc biosynthesis is F-6-P, which is also an essential intermediate for the Embden-Meyerhof-Parnas pathway (EMP). Furthermore, sufficient supplies of glutamine and acetyl-CoA, which act as amino and acetyl donors, are also important for GlcNAc production ( Figure 1). Therefore, modulation of the balance between cell growth and GlcNAc biosynthesis is crucial for high-level GlcNAc production. However, there are relatively few studies focusing on this strategy, which may limit the further improvement of microbial production of GlcNAc. Red arrows and crosses indicate gene deletions; green arrows indicate the GlcNAc biosynthesis pathway from glucose; purple arrows indicate glycerol utilization pathway. G-3-P, glycerol-3-phosphate; DHAP, glycerone phosphate; G-6-P, glucose-6-phosphate; F-6-P, fructose-6-phosphate; F-1,6-BP, fructose-1,6-bisphosphate; EMP, Embden-Meyerhof-Parnas pathway; TCA, tricarboxylic acid cycle; Glu, glutamic acid; Gln, glutamine; PPP, pentose phosphate pathway; GlcN-6-P, glucosamine-6phosphate; GlcNAc, N-acetylglucosamine; glpK, glycerol kinase gene; zwf, glucose-6-phosphate dehydrogenase gene; glmS, glucosamine-6-phosphate synthase gene; gna1, glucosamine-6-phosphate N-acetyltransferase gene; yqaB, GlcNAc-6-phosphate phosphatase gene.
Cell metabolism can be rationally divided into growth and production modules by using mixed substrates with direct access to multiple pathways. Based on this strategy, production enhancements have been widely reported. For example, in myo-inositol fermentation, a creative strategy has been exploited for efficient inositol production (reaching as high as 106.3 g/L) by synergetic utilization of glucose and glycerol as carbon sources [18]. Additionally, the productivity of lycopene was significantly improved in the fed-batch culture of glycerol supplemented with glucose and arabinose, which was 11.7-fold higher than that without auxiliary carbon sources [19]. Furthermore, it was reported that by controlled cofeeding of ATP and NADPH generators, such as glucose and gluconate, CO 2 reduction and CO 2 -derived lipid production were dramatically accelerated compared to the CO 2 -only control [20]. Thus, the synergistic substrate cofeeding strategy represents a good option to modulate carbon flux distribution in GlcNAc biosynthesis.
In this study, our aim was to achieve high GlcNAc production by modulation of cell growth and GlcNAc biosynthesis using the synergistic substrate cofeeding strategy with glucose and glycerol. First, the glucose utilization pathways of E. coli, including EMP and the pentose phosphate pathway (PPP), were blocked by deleting the 6-phosphofructokinase genes (pfkA and pfkB) and glucose-6-phosphate dehydrogenase gene (zwf ). Second, the glycerol consumption pathway and the GlcNAc biosynthesis pathway were enhanced. Consequently, glucose would be conserved for GlcNAc production while glycerol would be used to support cell growth and supply glutamine and acetyl-CoA for GlcNAc biosynthesis. Finally, the fermentation medium was optimized and GlcNAc production reached 2.62 g/L with a stoichiometric yield of 0.64 mol GlcNAc/mol glucose in shake flask fermentation. The results from this study provide valuable guidance and an essential reference for achieving rational distribution of carbon flux for the production of other value-added biochemicals.

Construction and Characterization of an E. coli Platform Strain with High F-6-P Supply
To achieve a high yield of GlcNAc from glucose, carbon flux distribution at principal branch points (glucose-6-phosphate (G-6-P) and F-6-P) in the central metabolic network of E. coli must be significantly modified from that observed during balanced growth, so that the GlcNAc precursor F-6-P can be synthesized in the optimal stoichiometric ratio. For E. coli strains, G-6-P could be driven toward PPP through ZWF, while F-6-P was mainly broken down to pyruvate in EMP via PFK encoded by the pfkA and pfkB genes ( Figure 1). Therefore, the zwf, pfkA and pfkB genes were successively knocked out in E. coli MG1655(DE3) using the CRISPR-Cas9 system to block PPP and EMP [21], generating mutants MG1655(DE3)∆pfkA, MG1655(DE3)∆pfkB, MG1655(DE3)∆pfkA∆pfkB, MG1655(DE3)∆zwf and MG1655(DE3)∆pfkA∆pfkB∆zwf.
The growth profiles of the metabolically engineered strains and the E. coli MG1655(DE3) wild-type strain were then compared on various media, including an M9s medium with different carbon sources (glucose, glycerol or glucose+glycerol) ( Figure 2). Glucose and glycerol, which enter the EMP upstream or downstream of F-6-P, were chosen. As shown in Figure 2A-C, the growth of mutants with a single deletion of the pfkA or pfkB gene (MG1655(DE3)∆pfkA or MG1655(DE3)∆pfkB) was almost unaffected compared to the wild-type strain under all tested culture conditions. However, deletion of the zwf gene (MG1655(DE3)∆zwf ) led to a slight decrease of growth rate, which might be due to the inefficient supply of NADPH (Figure 2A,B). Double-deletion mutant (MG1655(DE3)∆pfkA∆pfkB) and triple-deletion mutant (MG1655(DE3)∆pfkA∆pfkB∆zwf ) showed increased maximum OD600 on a glycerol medium ( Figure 2B). Surprisingly, the mutant MG1655(DE3)∆pfkA∆pfkB∆zwf with a blocked PPP and EMP could grow well under the culture condition where glucose was used as the sole carbon source (Figure 2A). The glucose and glycerol consumptions of different MG1655(DE3) mutants on various media are given in Figure 2D-G. Although the triple-deletion mutant MG1655(DE3)∆pfkA∆pfkB∆zwf could grow on a glucose medium, the glucose consumption was significantly lower than those of the wild-type strain and the single-or double-deletion mutants ( Figure 2D), which would result in the intracellular accumulation of F-6-P. As expected, the presence of glucose inhibited the consumption of glycerol for the MG1655(DE3) wild-type strain and three single-deletion mutants via carbon catabolite repression. In contrast, the carbon catabolite repression was mildly alleviated for mutants MG1655(DE3)∆pfkA∆pfkB and MG1655(DE3)∆pfkA∆pfkB∆zwf with 0.33 and 0.45 g/L glycerol consumed, respectively, when glucose and glycerol were used in a mixed carbon source ( Figure 2F,G). This phenomenon coincides with the result reported by Shiue and co-workers that significantly reduced glucose transportation and utilization in the cell would result in the alleviation of carbon catabolite repression [22]. However, only a small amount of glycerol was consumed in our mixed-carbon-source fermentation. This might be because mutants MG1655(DE3)∆pfkA∆pfkB and MG1655(DE3)∆pfkA∆pfkB∆zwf still have a weak glucose-utilization ability.
Recently, the construction of an E. coli ∆pfkA∆pfkB∆zwf triple-deletion mutant has been reported by several groups, and the published data showed that this strain has a severe growth defect when glucose is the sole carbon source [23][24][25][26]. However, our results demonstrated that mutant MG1655(DE3)∆pfkA∆pfkB∆zwf could still consume glucose through an unknown pathway. Font et al. reported that the strain E. coli LJ110∆pfkA∆pfkB∆zwf showed no growth on a glucose medium but could form small colonies on fructose agar plates [26]. Because the intracellular F-6-P could accumulate after deletion of the pfkA, pfkB and zwf genes, we assumed that F-6-P might enter the lower glycolytic trunk via an F-1-P bypass ( Figure 3B), whereby (1) F-6-P is dephosphorylated by phosphatase to form fructose and (2) fructose is then subsequently converted to fructose-1,6-bisphosphate (F-1,6-BP) by enzyme II fru of the phosphoenolpyruvate (PEP)-dependent phosphotransferase system (PTS) and 1-phosphofructokinase encoded by fruA and fruK, respectively.

Medium Optimization for Enhanced Growth of MG1655(DE3)∆pfkA∆pfkB∆zwf-pKGGY
It is well known that cell biomass production during the cell growth phase is usually important to the end-product's biosynthesis. Although GlcNAc was detected in the fermentation broth of MG1655(DE3)∆pfkA∆pfkB∆zwf-pKGGY, the production was still very low, mostly due to the growth restriction of this strain ( Figure 4B). Two reasons may account for the growth defect of a triple-deletion mutant harboring plasmid pKGGY: (1) carbon flux distribution for cell growth is not sufficient; (2) the biosynthesis of essential nutrients is restricted. Therefore, to enhance cell growth and biomass production, various concentrations of pyruvate (EMP intermediate, providing additional carbon source for cell growth), citric acid (tricarboxylic acid (TCA) cycle intermediate, providing additional carbon source for cell growth) and LB broth (providing essential nutrients) were added into the fermentation medium ( Figure 5). The results demonstrated that the addition of pyruvate and citric acid did not increase the growth rate and cell biomass of MG1655(DE3)∆pfkA∆pfkB∆zwf-pKGGY ( Figure 5A,B), indicating that carbon flux distribution might not be the cause of the growth defect of this strain. However, the cell growth of MG1655(DE3)∆pfkA∆pfkB∆zwf-pKGGY was reinstalled when more than 10% LB broth was added to the medium ( Figure 5C), suggesting that the growth was probably limited by some essential nutrients. Thus, the medium M9s + 10 g/L glucose + 5 g/L glycerol + 10% LB was used as the fermentation medium for the following experiments.

Discussion
Metabolic engineering aims to achieve high-yield production of value-added chemicals in engineered strains, making them economically feasible in commercial production. To achieve this goal, the target pathway is usually boosted, while competing pathways are eliminated or attenuated. However, when the competing pathways are related to the central metabolism, especially the EMP and PPP, application of this strategy becomes challenging due to their important effects on cell growth. Carbon cofeeding has been successfully used to balance growth and production metabolism, demonstrating the effectiveness of this strategy [18,19]. In this study, the carbon cofeeding strategy was successfully adopted for high-yield GlcNAc production.
The main challenge for high-level GlcNAc biosynthesis using microbial cell factory is the sufficient supply of F-6-P, which is the precursor for GlcNAc biosynthesis and the important intermediate for EMP and PPP. Disruption or attenuation of EMP and PPP could increase the yield of bioproducts derived directly from F-6-P. To improve the GlcNAc yield from glucose, a triple-deletion mutant, MG1655(DE3)∆pfkA∆pfkB∆zwf, was constructed with the EMP and PPP of this mutant blocked. Our results showed that this mutant could still utilize glucose, probably through an F-1-P bypass catalyzed by F-6-P phosphatase, FruA and FruK ( Figure 3B). However, results from other groups showed that the tripledeletion mutant of E. coli (∆pfkA∆pfkB∆zwf ) could not grow on a glucose medium [26]. This phenomenon might be caused by the existence of phage DE3 on the genome of E. coli MG1655, which can alter gene expression and regulation of the host. Although the tripledeletion mutant in our study could consume glucose, fermentation results revealed that most of the glucose in the fermentation broth of MG1655(DE3)∆pfkA∆pfkB∆zwf-pKGGY was converted to GlcNAc (0.64 mol GlcNAc/mol glucose) when glucose and glycerol were used as the mixed carbon source ( Figure 6A,C), indicating that blocking the EMP and PPP can favor GlcNAc production. In addition, repression of glycerol utilization by glucose was alleviated in mutant MG1655(DE3)∆pfkA∆pfkB∆zwf ( Figure 2F,G), which facilitated the coutilization of glucose and glycerol for GlcNAc production and cell growth. Moreover, the glycerol utilization was further enhanced via overexpression of glpK gene from P. pastoris.
A growth defect was observed for mutant MG1655(DE3)∆pfkA∆pfkB∆zwf-pKGGY, which badly influenced GlcNAc production. The GlcNAc production of MG1655(DE3)-∆pfkA∆pfkB∆zwf-pKGGY was much lower than that of MG1655(DE3)∆pfkA∆pfkB-pKGGY when M9s supplemented with glucose and glycerol was used as the fermentation medium. Yet, the addition of LB broth (>10% v/v) could dramatically increase the triple-deletion mutant's growth and GlcNAc production (over 200-fold rise) ( Figure 5C). Finally, the GlcNAc production of MG1655(DE3)∆pfkA∆pfkB∆zwf-pKGGY reached 2.62 g/L using the optimized medium, representing an increase of about 3.2-fold compared to that of mutant MG1655(DE3)∆pfkA∆pfkB-pKGGY. These results implied that the biosynthesis of some essential nutrients was limited due to the expression of genes related to GlcNAc production and glycerol utilization. Further investigations are required to determine the physiological causes of the observed phenomena.
Industrial fermentations of E. coli strains are usually plagued by unproductive conversion of glucose to acetate, which leads to low product yields and inhibition of cell growth [33,34]. In this study, most of the acetic acid generated in the fermentation of MG1655(DE3)∆pfkA∆pfkB∆zwf-pKGGY was re-assimilated at the end of fermentation caused by the upregulated expression of the acs gene ( Figure 6D,E). The acetate production of MG1655(DE3)∆pfkA∆pfkB∆zwf-pKGGY was reduced by more than 90% compared to that of MG1655(DE3)-pKGGY. These results indicated that blocking the EMP and PPP could promote acetate re-assimilation in E. coli, which further increased the substrate conversion efficiency of the mutant strain.
In conclusion, an E. coli platform strain with high F-6-P supply was constructed by blocking the EMP and PPP. Through introduction of glycerol consumption pathway and the GlcNAc biosynthesis pathway, the synergistic glucose and glycerol cofeeding strategy was successfully applied for GlcNAc production in this study. Ultimately, the fermentation medium was optimized in order to enhance the growth of the final mutant, resulting in a high GlcNAc yield of 0.64 mol/mol glucose. The mutant MG1655(DE3)∆pfkA∆pfkB∆zwf developed here is a promising host with minimal accumulation of acetate byproduct, which could be further engineered for other forms of valuable biochemical production.

Plasmid Construction
All the primers and plasmids used in this study are listed in Supplementary Table S1  and Table 1, respectively. For the construction of plasmid pRed_Cas9_recA, the gRNA expression cassette and homologous arms for poxb gene deletion on the plasmid pRed_Cas9_recA_ ∆poxb300 (MolecularCloud plasmid# MC_0000001) were eliminated through the modular assembly method [21]. To construct CRISPR-Cas9-assisting donor plasmids (harboring gRNA expression cassette and homology arms for target gene deletion), pEASY-T3 vector (Trans-Gen, Beijing, China), which is a high copy-number plasmid, was chosen as the mother vector. For the attempts to delete the pfkA, pfkB and zwf genes in E. coli MG1655(DE3), the J23119 promoter fused with a 20-nt guiding sequence (pfkA: GTGTCTGACATGATCAACCG, pfkB: CACGTACATGTGGAAGCAAG, zwf : GCGTGCTGACTGGGATAAAG) was integrated into pEASY-T3 via TA cloning, and then the homology arms (~500 bp each) were inserted into the Sbf I and NdeI sites, generating plasmids p∆pfkA, p∆pfkB and p∆zwf, respectively.

Mutant Screening
Genome editing in E. coli MG1655(DE3) was performed following the procedure described previously with some modifications [21,36]. In brief, plasmid pRed_Cas9_recA was transformed into E. coli MG1655(DE3) by electroporation, followed by plating on LB+Kan plates and culturing at 30 • C. For genome editing, the donor plasmid containing gRNA and homology arms was then transformed into MG1655(DE3)-pRed_Cas9_recA competent cells. The resulting cells were spread onto LB+Kan+Amp plates and incubated overnight at 30 • C. The transformant colonies were picked and inoculated in a LB+Kan+Amp liquid medium. The obtained cultures were then diluted serially and plated onto LB+Kan+Amp plates supplemented with 2 g/L D-arabinose to induce the expression of Cas9 nuclease and the λ-Red system. After the colonies were observed, the putative mutants were screened by colony PCR and then confirmed by Sanger sequencing (Supplementary Figures S1 and S2). To cure the pRed_Cas9_recA and donor plasmid in the newly obtained mutant, successive transferring at 37 • C was performed in an LB liquid medium without antibiotics. After five transfers, the culture was then diluted serially and spread onto LB plates for colony development. The pure mutant was obtained through colony PCR and then further verified by Kan and Amp selection.

Batch Fermentation
Batch fermentation with various GlcNAc-producing strains was carried out in a 1 L shake flask with 100 mL reaction volume at 37 • C and 200 rpm. M9s supplemented with different carbon sources (glucose and/or glycerol) and growth factors (pyruvate, citric acid or LB broth) were used as the fermentation medium. The strains were initially cultured in an LB medium to generate the seed culture (OD600 reached about 1.5). Then, a 2% inoculum of seed culture was used for all batch fermentations. Samples were taken every 6 h for the analysis. All fermentations were performed in triplicate.

Analytical Methods
Cell density was measured using a microplate reader at 600 nm (OD 600 ). The glucose, glycerol and acetate concentrations in the fermentation broth were quantified using a highperformance liquid chromatography system (LC-20A, Shimadzu, Kyoto, Japan) equipped with a Sugar-ParkI column (Waters, Milford, MA, USA) and refractive index detector (RID). The mobile phase was ddH 2 O at a flow rate of 0.6 mL/min at 80 • C. The GlcNAc concentration was determined by an LC-MS system (Sciex TripleTOF 6600 interfaced with the UHPLC Agilent 1290 Infinity I) equipped with an ACQUITY UPLC BEH Amide column (21 mm × 100 mm, 1.7 µm). The mobile phases were a blend of solvent A (0.1% formic acid in water) and solvent B (0.1% formic acid in acetonitrile) at a flow rate of 300 µL/min at 45 • C.

RT-qPCR Analysis
The primers for RT-qPCR are shown in Supplementary Table S1. Cells cultivated in various media were harvested during the exponential growth phase (OD600 reached about 1.5). Total RNA extraction (RNA-easy Isolation Reagent, Vazyme, Beijing, China) and the reverse transcription of cDNA (HiScript III RT SuperMix for qPCR, Vazyme, Beijing, China) were conducted according to the manufacturer's instructions. RT-qPCR was carried out using a QuantStudio 6 Flex system (Applied Biosystems, Foster City, CA, USA) and ChamQ Universal SYBR qPCR Master Mix (Vazyme, Beijing, China). RT-qPCR were performed following the procedure described by Lu et al. [37]. The transcription levels of the target genes were analyzed by the 2 −∆∆Ct method [38], where the 16S rRNA gene was used as the internal standard. Each sample was run in triplicate.