13C-Metabolic Flux Analysis Reveals the Metabolic Flux Redistribution for Enhanced Production of Poly-γ-Glutamic Acid in dlt Over-Expressed Bacillus licheniformis

Poly-γ-glutamic acid (γ-PGA) is an anionic polymer with various applications. Teichoic acid (TA) is a special component of cell wall in gram-positive bacteria, and its D-alanylation modification can change the net negative charge of cell surface, autolysin activity and cationic binding efficiency, and might further affect metabolic production. In this research, four genes (dltA, dltB, dltC, and dltD) of dlt operon were, respectively, deleted and overexpressed in the γ-PGA producing strain Bacillus licheniformis WX-02. Our results implied that overexpression of these genes could all significantly increase γ-PGA synthetic capabilities, among these strains, the dltB overexpression strain WX-02/pHY-dltB owned the highest γ-PGA yield (2.54 g/L), which was 93.42% higher than that of the control strain WX-02/pHY300 (1.31 g/L). While, the gene deletion strains produced lower γ-PGA titers. Furthermore, 13C-Metabolic flux analysis was conducted to investigate the influence of dltB overexpression on metabolic flux redistribution during γ-PGA synthesis. The simulation data demonstrated that fluxes of pentose phosphate pathway and tricarboxylic acid cycle in WX-02/pHY-dltB were 36.41 and 19.18 mmol/g DCW/h, increased by 7.82 and 38.38% compared to WX-02/pHY300 (33.77 and 13.86 mmol/g DCW/h), respectively. The synthetic capabilities of ATP and NADPH were also increased slightly. Meanwhile, the fluxes of glycolytic and by-product synthetic pathways were all reduced in WX-02/pHY-dltB. All these above phenomenons were beneficial for γ-PGA synthesis. Collectively, this study clarified that overexpression of dltB strengthened the fluxes of PPP pathway, TCA cycle and energy metabolism for γ-PGA synthesis, and provided an effective strategy for enhanced production of γ-PGA.

Bacillus species have been proven as the efficient γ-PGA producers, and a number of metabolic engineering strategies have been developed to improve γ-PGA yield. For example, knocking out glutamate dehydrogenase genes rocG and gudB improved glutamic acid accumulation, which led to a 38% increase of γ-PGA yield in Bacillus amyloliquefaciens . The synthesis of extracellular polysaccharide and lipopolysaccharide were blocked to decrease by-product yields for γ-PGA production in B. amyloliquefaciens LL3 . In addition, strengthening of NADPH and ATP supplies all benefited γ-PGA production (Cai et al., 2017. Recently, cell surface engineering was proven to be an effective strategy for enhancement production of metabolics. Overexpression of phosphatidylserine synthase gene pssA could enhance the cell membrane integrity and hydrophilicity, and further improved the cell tolerance and biorenewable yields (short-chain fatty acids, organic alcohols, organic acids and other aromatic compounds, etc) in E. coli (Tan et al., 2017). Elevation of membrane cardiolipin levels via overexpressing cardiolipin synthase gene clsA significantly increased hyaluronic acid titer by 204% in Bacillus subtilis (Westbrook et al., 2018). However, no research has been focused on the relationship between cell surface engineering and γ-PGA synthesis.
The biochemical reactions in cells were difficult to be profoundly described with the existing techniques. While, acting as a new method, 13 C metabolic flux analysis ( 13 C-MFA) is a promising tool to quantify cellular metabolism reaction rates in a network via isotopomer tracer experiments (Long et al., 2016). Moreover, 13 C-MFA is a powerful tool to quantify the energy generation and consumption rates in the metabolic pathways (He et al., 2014;Yao et al., 2016), and it also applied to comprehend and analyze the changes in central carbon metabolism under aerobic and anaerobic conditions. For instance, the optimal tracers [1,2-13 C] glucose, [1,6-13 C] glucose, [1,2-13 C] xylose, and [5-13 C] xylose were applied in Escherichia coli under aerobic and anaerobic conditions, and their results demonstrated that the fluxes of EMP pathway and TCA cycle were all increased under anaerobic conditions, which generated more energy, formate, alcohol and succinate to adapt the adverse environments (Gonzalez et al., 2017).
In this study, the genes dltA, dltB, dltC, and dltD of dlt operon were, respectively, deleted and overexpressed in B. licheniformis WX-02, and 13 C-MFA was performed to expound the effect of dltB overexpression on metabolic flux redistribution. In addition, the transcriptional level, by-products contents were also measured during γ-PGA synthesis. The aim of this study is to illustrate the relationship between dlt operon overexpression and γ-PGA synthesis by 13 C-MFA, and provides an efficient strategy of strain improvement for γ-PGA production.

Strains and Culture Conditions
Strains and plasmids used in this study were provided in Table 1. B. licheniformis WX-02 was acted as the original strain for constructing recombinants, and E. coli DH5α was served as the host strain for plasmid construction. B. licheniformis and E. coli were grown in LB medium with responsible antibiotic (20 µg/mL kanamycin, 100 µg/mL ampicillin, or 20 µg/mL tetracycline), when required.

Strain Construction
The construction procedures of dltB deletion and overexpression strains were served as the examples. For dltB deletion strain, the upstream and downstream arms of gene dltB were, respectively, amplified by the primers dltB-F1/R1 and dltB-F2/R2 (Supplementary Table S1), based on the genomic DNA of B. licheniformis WX-02, and fused by Splicing Overlapping Extension PCR (SOE-PCR) with primers dltB-F1/R2. The fused fragment was inserted into the plasmid T 2 (2)-Ori at SacI/XbaI, colony PCR and DNA sequence confirmed that the dltB deletion vector was constructed successfully, named as T 2 -dltB. Then, T 2 -dltB was electro-transferred into B. licheniformis WX-02, and the positive transformant was cultivated in LB medium with 20 µg/mL kanamycin at 220 rpm and 45 • C, and sub-cultured for three times to obtain the single-crossover recombinants. The recombinants were grown in LB medium at 37 • C with six subcultures, and the kanamycin sensitive colonies were further confirmed by colony PCR and DNA sequence, and the dltB deletion strain was named WX-02 dltB.

Analytic Methods
The cell biomass was monitored by measuring OD 600 using a UV-spectrophotometer-752 N (Shanghai Instrument Analysis Instrument Co., Ltd., Shanghai, China), glucose concentrations were determined via using the Enzyme Electrode Analyzer SBA-40E (Shandong Academy of Sciences, Shandong, China). The γ-PGA yield was measured according to our previously reported method . Briefly, the volume of 2 mL fermentation broth was mixed with 4 mL distilled water, and cells were separated by centrifugation at 12000 rpm for 6 min after adjusting the pH to 2.5∼3.0. While, three volumes of absolute ethanol were added into the supernatant after adjusting pH to 7.0, and the precipitate was dried at 80 • C to a constant weight for measuring γ-PGA yield. The net negative charge was measured by determining the cation binding rate of cell surface in LB medium, according to the previously described method Chen et al., 2018).
The concentrations of by-products, acetoin, 2,3-butanediol and acetic acid, were measured by GC/MS (Gas chromatograph Trace, Thermo; Triple Quadrupole Mass Spectrometer, Thermo; column: DB-WAXMS, 30 m × 0.25 mm × 0.25 µm, Thermo; United States). Briefly, the volume of 0.5 mL fermentation broth was mixed with 1.5 mL absolute ethanol, and the supernatant was separated by centrifugation at 12000 rpm for 10 min. Equal volume of ethyl acetate was added into the supernatant for extracting acetoin, 2,3-butanediol and acetate. The parameters of GC-MS were set as follow: Gas chromatographic inlet temperature was 220 • C, the injection volume was 1 µL, carrier gas flow rate was 1 mL/min and the split ratio was 20:1. The temperature program: hold at 40 • C for 1 min, and increased to 160 • C at 6 • C/min, hold for 1.5 min, and then raised to 220 • C at 30 • C/min, hold for 2 min. Solvent delay was set as 4 min. The range of mass to charge ratio (m/z) in MS was set between 35 and 500.
The gene transcriptional level were measured with TRIzol R Reagent (Invitrogen, United States) and PrimeScript TM II 1st Strand cDNA Synthesis Kit (TaKaRa, Japan) according to our previously reported method (Cai et al., 2017). The primers used for RT-qPCR were listed in Supplementary Table S2, and the housekeeping gene 16S rRNA was used to normalize the gene expression data. The data were averaged and presented as the mean ± SD.

Mass Isotopomer Distribution Analysis of Amino Acids
To analyze the mass isotopomer distributions (MIDs) of amino acids, [1,2-13 C] glucose was served as the sole carbon source for cell growth and γ-PGA synthesis. The cell pellets at the mid-logarithmic phase (7 h) were collected and hydrolyzed by 6 M HCl at 100 • C for 24 h. The supernatant of hydrolyzate was air-dried, and the precipitate was subsequently derivatized with N-tert-butyldimethylsilyl-N-methyltrifluoroacetamide (TBDMS). Then, the derivatization products were quantified by GC/MS (Gas chromatograph Trace, Thermo; Triple Quadrupole Mass Spectrometer, Thermo; column: TG-5MS, 30 m × 0.25 mm × 0.25 µm, Thermo; United States). The injection volume and injection split ratio were 1 µL and 1:10, respectively, and carrier gas helium was 1.2 mL/min. GC temperature program was set as follows: hold at 150 • C for 2 min, and increased to 280 • C at 3 • C/min, and increased to 300 • C at 20 • C/min, and then hold for 5 min. Solvent delay was set as 5 min, and the range of mass to charge ratio (m/z) in MS was set between 60 and 500 (You et al., 2012). The software WuFlux-Ms Tool was used to analyze and correct amino acid MS data ( (Wahl et al., 2004). Isotopomer labeling fractions (M0, M1, M2, etc.) represent fragments containing unlabeled, singly labeled, doubly labeled amino acids, etc (Wan et al., 2017).

Statistical Analyses
All samples were analyzed in triplicate, and the data were presented as the mean ± SD for each sample point. Significant differences were determined by one-way analysis of variance (ANOVA). Statistical significance was defined as P < 0.05.
Previously, the dlt operon was proved to play the important role on the net negative charge of cell surface. As shown in Figure 2, the negative charges of cell surface were all increased in the dlt deletion strains, and decreased in the dlt overexpression strains. The negative charge was reduced by 48.93% in the dltB overexpression strain WX-02/pHY-dltB.

C-Metabolic Flux Analysis
The metabolic flux distributions were determined based on the measured mass isotopomer distributions of proteinogenic amino acids and synthetic rates of γ-PGA, acetic acid, acetoin and 2,3butanediol. The measured mass isotopomer distributions (MIDs) of proteinogenic amino acids coincided well with the simulated MIDs (Supplementary Table S3), indicating the good fit and high flux precision between the measured and simulated data (Yao et al., 2016). The measured biomass formation rates were not employed as the constraints of 13 C-MFA model ( Table 2). The flux values and exchange coefficients were estimated with the 13 C MFA model listed in Supplementary Table S4.
As shown in Figure 4, the fluxes in the central metabolism of B. licheniformis were redistributed in WX-02/pHY-dltB. Firstly, the flux from pyruvate to acetyl-CoA was increased by 11.12%, and the flux distributed into tricarboxylic acid cycle was enhanced by 38.38% in WX-02/pHY-dltB. The flux for acetic acid production was increased slightly, whereas, the flux of overflow metabolism (from pyruvate to acetoin and 2,3-butanediol) was decreased by 26.93%. Secondly, the flux of PP pathway was increased by 7.82% in the dltB overexpression strain, which might generate more NADPH for γ-PGA synthesis. Thirdly, the flux through pyruvate carboxylation (PYR + CO 2 → OAA), the major anaplerotic flux into TCA cycle, was increased by 25.40% in WX-02/pHY-dltB. The increase of pyruvate carboxylation could enable more carbon flow from glucose to TCA cycle, and further increased the demand for oxaloacetate based biomass synthesis ( Table 2; He et al., 2014). Fourthly, the fluxes from α-ketoglutaric acid to glutamic acid and γ-PGA were increased by 19.47 and 57.92%, respectively. Fifthly, the flux of EMP pathway was decreased by 4.13%, which further reduced the NADH generation and by-products (acetoin, 2,3-butanediol and acetic acid) syntheses. Lastly, the estimated biomass biosynthesis flux was slightly higher (about 6.40%) in WX-02/pHY-dltB (Figure 4), which positively correlated with the results of cell biomass ( Table 1).
The cofactors (such as NADH, NADPH, etc.) and ATP play the important roles in metabolic production (Kind et al., 2013), as well as for γ-PGA synthesis (Cai et al., 2017. Based on our metabolic flux model in B. licheniformis, the NADPH and ATP formation rates in WX-02/pHY-dltB were increased by 12.50 and 3.54%, which were beneficial for γ-PGA synthesis ( Table 3).

DISCUSSION
Cell wall and cell membrane are the selective permeation barrier of bacteria, and engineering of cell surface negative charge, membrane phospholipid head and fatty acid hydrophobic tail composition could affect bacterial growth and product production (Hyyrylainen et al., 2007;Ghorbal et al., 2013;Tan et al., 2017). Previous researches demonstrated that the increases of negative charge on the cell surface could improve the secretion efficiency of target protein, and the increase rate of target protein with lower isoelectric points (PI) was higher than that of protein with higher PI Chen et al., 2018). While, based on our results, the increases of cell surface negative charge led to the decrease of γ-PGA yield in the dlt operon mutants.
On the one hand, overexpression of dltB slowed down the absorption and utilization rate of glucose (Figure 3A), and further decreased the flux of EMP pathway. Previous research implied that reduction of glycolysis flux in E. coli THRD could reduce the accumulation of acetate and increase NADPH and L-threonine generations (Xie et al., 2014), which was consistent with our results ( Figure 3A and Table 2). Moreover, the alanylation of teichoic acids could modulate the negative charge of cell wall to protect secretory or cell wall-associated proteins against degradation during the post-translocational folding (Hyyrylainen et al., 2000), and enhancement of cell surface negative charge could increase the binding capacity of autolysin, and accelerated autolysis of bacteria (Steen et al., 2005). Figure 3A showed that the logarithmic period of WX-02/pHY-dltB was extended by 2∼3 h, and the growing status of WX-02 dltB was obviously badness in post-fermentation compared with WX-02 (Supplementary Figure S1). This results indicated that overexpression of dltB could promote the cell growth by reducing the autolysis of cells. Therefore, overexpression of dltB reduced the absorption and utilization rate of glucose, which decreased the flux of EMP pathway and synthesis capabilities of by-products (acetoin, 2,3-butanediol and acetic acid), and extended logarithmic growth phase, and all these phenomenons were beneficial for γ-PGA synthesis.
Then, 13 C-labeled isotope tracer and 13 C-MFA were applied to further evaluate the metabolic flux redistributions in the dltB overexpression strain. Based on our results, the flux of EMP pathway is weakened, which reduced the NADH generation and by-products synthesis, and our results were positively consisted with the previous research (Lee and Oh, 2016). Secondly, the flux of PPP pathway and NADPH supply were all enhanced in the dltB overexpression strain. Since the conversion of α-ketoglutarate to glutamic acid requires NADPH in B. licheniformis WX-02, enhancement of NADPH supply contributes to γ-PGA synthesis (Cai et al., 2017;Tian et al., 2017). In addition, the flux of TCA cycle was enhanced by 38.38% in the dltB overexpression strain, and the increase of α-ketoglutaric acid synthesis flux was beneficial for cell growth (Wan et al., 2017) and γ-PGA production. Besides, the flux of complement-deficient pathway (PYR → OAA) was enhanced by 25.40%, which is beneficial for oxaloacetate accumulation and cell growth (Sauer et al., 1999). Finally, the flux from α-ketoglutarate to glutamate and further generate γ-PGA were, respectively, increased by 19.47 and 57.92%, which were consistent with fermentation characters (Figure 1). Furthermore, overexpression of dlt operon led to the decrease of cell surface negative charge (Figure 2), and further changed the extracellular microenvironment (Hyyrylainen et al., 2000), which might promote the recruitment of signal molecular ComP to activate two component system ComP-ComA for γ-PGA synthesis. Based on our results, the transcriptional levels of comP/comA and degU/degS were all increased in the dltB overexpression strian, which further affected the transcription levels of genes involved in γ-PGA synthesis and carbon metabolism ( Figure 5). Moreover, the decrease of cell surface negative charge could reduce the binding of cationic (e.g., Mg 2+ and Ca 2+ , etc.) and cationic antimicrobial peptides on cell surface (Neuhaus and Baddiley, 2003), and the low levels of cationic or cationic antimicrobial peptides could activate PhoQ(R)∼PhoP system (Garcia Vescovi et al., 1996;Ren et al., 2017). In this study, the concentration of phosphate in the culture medium was too high to activate PhoQ(R)∼PhoP system (Devine, 2018) in WX-02/pHY300, however, due to the low level of Mg 2+ and Ca 2+ on cell surface of dltB overexpression strain, the PhoQ(R)∼PhoP might be activated in WX-02/pHY-dltB, which was consistent with our unpublished results that strengthening PhoQ(R)∼PhoP system benefits γ-PGA synthesis. Finally, γ-PGA is an anionic polypeptide, and the decrease of cell surface negative charge might reduce the strength of electrostatic repulsion (Cafiso et al., 2014) between TAs and γ-PGA, which further benefited γ-PGA secretion. To test this hypothesis, the negatively charged products (such as lichenysin, bacitracin, pulcherrimin, et al.) were analyzed, and our results showed that the yields of those products with negative charge were all increased in dlt overexpression strains (Supplementary Figure S3).

CONCLUSION
Cell surface engineering is a promising tactic for enhanced production of metabolites. This study demonstrated that overexpression of dlt operon could increase γ-PGA production for the first time. Furthermore, 13 C-MFA was applied to illustrate the affect of dltB overexpression on γ-PGA synthesis. Our results showed that overexpression of dltB could reduce the negative charge of cell surface, which further reduced the absorption and utilization rate of glucose and the flux of EMP pathway. Meanwhile, the increases of TCA cycle flux, NADPH and ATP supplies, glutamic acid formation benefited γ-PGA synthesis. This work demonstrated that 13 C-MFA provided the distinct metabolic insights in engineered microbes for the changes in central carbon metabolism.

AUTHOR CONTRIBUTIONS
PH and SC designed the study. PH, SH, and YC carried out the molecular biology studies and construction of recombinant strains. PH, DC, SH, and YC carried out the fermentation studies. PH and NW carried out the metabolic flux analysis. PH, DC, SL, and SC analyzed the data and wrote the manuscript. All authors read and approved the final manuscript.