All strains and plasmids used in this study are listed in Table 1. The host strain in this study was E. coli W (KCTC1039), which was provided by the Korean Collection for Type Cultures (KCTC; Jeongeup, Korea). Sugar transporter mutation was performed by deleting ptsG (ADT74705), and flagella mutation was performed by deleting flhC (ADT75524). All deletion methods utilized λ-red recombinase-based homologous recombination (Datsenko and Wanner, 2000). The name of the ptsG knockout mutant was Wp and that of the flhC and ptsG double mutant was Wpf. All deletions were confirmed via polymerase chain reaction using genomic DNA as a template. All oligonucleotides used in this study were synthesized by Bionics (Seoul, Korea) and are listed in Supplementary Table 1.

During genetic engineering, strains were cultivated and confirmed in Lysogeny broth (10 g/L of tryptone, 10 g/L of NaCl, and 5 g/L of yeast extract). All cultivations for analysis were performed in M9 minimal medium (6 g/L of Na₂HPO₄, 3 g/L of KH₂PO₄, 1 g/L of NH₄Cl, 0.5 g/L of NaCl, and 0.01% Thiamine–HCl) with 20 g/L of glucose and 1 mL of trace elements [2.86 g/L of H₃BO₃, 1.81 g/L of MnCl₂⋅4H₂O, 0.9 g/L of FeCl₃⋅6H₂O, 0.39 g/L of Na₂MoO₄⋅2H₂O, 0.222 g/L of ZnSO₄⋅7H₂O, 0.079 g/L of CuSO₄⋅5H₂O, and 49.4 mg/L of Co(NO₃)₂⋅6H₂O]. The antibiotics used were carbenicillin (100 μg/mL), kanamycin (50 μg/mL), and chloramphenicol (34 μg/mL). All chemicals were purchased from Sigma-Aldrich (St. Louis, MO, United States).

Cultivation was performed in 250 mL Erlenmeyer flasks with 25 mL of working volume at 250 rpm and 37°C under aerobic conditions.

Bacterial cell mass was estimated via measurement of optical density at 600 nm (OD₆₀₀) using a DU730 (Beckman Coulter, Brea, CA, United States). Measurements of glucose and acetate were performed using high-performance liquid chromatography (Waters, Milford, MA, United States) with a Refractive Index Detector 2414 and SH1011 columns (Shodex, Tokyo, Japan). The measurement was carried out at a temperature of 45°C, and 10 mM sulfuric acid was used as the mobile phase and the flow rate was 0.6 mL/min.

Enhanced green fluorescent protein (EGFP) intensity was measured using a microplate reader (Synergy H1; Biotek, Winooski, VT, United States). The cultured bacterial strains were washed with phosphate-buffered saline (PBS) solution and diluted. The measurement result was multiplied by the dilution ratio. The achieved excitation peak was at 479 nm, and the detected emission peak was at 520 nm.

For the ATP assay, BacTiter-Glo^TM (Promega, Fitchburg, United States) was used. The culture broth was harvested in the early exponential phase (OD_600nm ≈ 1.0). First, 100 μL of harvested culture broth was mixed with an equal volume of BacTiter-Glo^TM reagent. The mixed samples were incubated at room temperature for 5 min. Luminescence was measured using a microplate reader (Synergy H1). The luminescence unit was converted to μM using a standard curve, and the result was divided by dry cell weight. Therefore, the ATP concentration was calculated as μmol/g_cell.

For the NADPH/NADP⁺ assay, NADP/NADPH-Glo^TM (Promega, Fitchburg, United States) was used. Harvested cell of early exponential phase in 50 μL of PBS was lysed in 50 μL of 0.2 M NaOH and 1% (w/v) dodecyl trimethyl ammonium bromide (DTAB). Then, 25 μL of 0.4 M HCl was treated at 60°C for 15 min. Next, 25 μL of 0.5 M Trizma base was added for neutralization, and NADP/NADPH-Glo^TM reagents were also added. Samples were incubated at room temperature for 30 min. Luminescence was measured using a microplate reader (Synergy H1). PBS, NaOH, DTAB, HCl, and Trizma base were from Sigma-Aldrich (St. Louis, MO, United States).

For ¹³C-MFA, [1,2-¹³C]-glucose was added to the M9 medium. The 1 mL of cell broth in the early exponential phase (OD_600nm ≈ 1.0) was harvested and centrifuged at 15,000 × g for 10 min at 4°C to remove the supernatant. The obtained pellet was washed twice with distilled water and then fully dried with a freeze dryer (Hail, Gimpo, Korea). Then, 200 μL of 6 N HCl was added and hydrolyzed at 110°C for 24 h. After hydrolysis, sample neutralization was performed by adding 200 μL of 6 N NaOH and filtered using an Amicon Ultra 0.5 mL 10 kDa centrifugation filter (Millipore Corporation, Burlington, MA). Next, each sample was derivatized with 80 μL of N-(tert-butyldimethylsilyl)-N-methyl-trifluoroacetamide and 50 μL of pyridine for 1 h at 70°C.

Samples were measured using gas chromatography-mass spectrometry (Agilent 7890 B GC system) with an HP-5MS column (0.25 μm, 30 m × 0.25 mm; Agilent Technology, Santa Clara, CA, United States). The initial temperature was 80°C, which was raised to 280°C at a rate of 7°C/min and held for 10 min. One microliter of the sample was injected at 270°C in 1:10 split mode, and a flow rate of 1 mL/min of helium was used for the mobile phase. The ion source temperature was 230°C, and the electron impact was ionized at 70 eV. Measurement data were analyzed by single ion monitoring, while MassHunter was used for mass spectra, and mass isotopomer distributions (MIDs) were obtained (Antoniewicz et al., 2007a).

The network model for the flux analysis construction was conducted based on a previous report (Supplementary Table 2) (Leighty and Antoniewicz, 2013). INCA software, based on the elementary metabolite unit for ¹³C-MFA, was used (Antoniewicz et al., 2007b; Young, 2014). Flux estimation was performed by minimization of the differences between the simulated ones and the MIDs of the proteinogenic amino acids from experimental measurements using least squares regression. Flux estimations were performed 10 times to find the global solution. Chi-square statistical tests were performed for goodness of fit (Im et al., 2020). The metabolic fluxes were calculated from the MIDs of proteinogenic amino acids with an acceptable sum of squared residuals (SSR) (Supplementary Table 3). The SSRs were 13.1 (expected SSR range: 5.6–26.1) and 34.6 (expected SSR range: 16.8–47.0) in Wp and Wpf, respectively.

Deletion of flhC was expected to reduce unnecessary protein production and ATP consumption (Shachrai et al., 2010). First, we deleted the flhC gene in the E. coli W strain (KCTC1039) and named it Wf. Previously, we demonstrated that the ptsG knockout mutant from the W strain (KCTC1039), Wp, showed improved protein production (Jung et al., 2019). Therefore, the flhC gene was deleted in the Wp strain to construct the strain named Wpf.

Cultivation was performed in a shake flask under aerobic conditions in glucose M9 minimal medium at 37°C. Growth retardations with increased lag phase were observed in Wp and Wpf compared to W and Wf (Figure 1A), although the final ODs were similar after 24 h. The amount of glucose consumption was higher in W and Wf, whereas it was significantly decreased in Wpf. In the Wpf strain, the specific glucose uptake rate was also decreased compared to other strains (Figure 1B). These results indicate that Wpf experienced significant metabolic burden by knocking out both ptsG and flhC. However, the flhC knockout strain in the wild-type strain, Wf, did not show any phenotypic differences from the wild-type strain W. To identify the burden, intracellular cofactors in each strain were measured.

ATP levels and NADPH/NADP⁺ ratios were measured to see how these factors were affected by the mutations. Slight increases in the ATP concentration and NADPH/NADP⁺ ratio were observed in the Wf strain compared to those in the W strain (Figures 1C,D). However, the same flhC deletion in the ptsG knockout mutant caused a much larger effect. The ATP level was increased by 1.5-fold, and the NADPH/NADP⁺ ratio also increased from 1.73 to 2.08 in Wpf compared to Wp. These differences would be significant enough to cause cofactor imbalance, which could result in growth retardation and reduced glucose uptake. The NADH/NAD⁺ ratios were also measured in these strains, which showed little difference between them. To clarify the cofactor imbalance in more detail, ¹³C MFA was applied to the Wp and Wpf strains.

¹³C-MFA was conducted to investigate the metabolic burden of the Wpf strain. The metabolic fluxes of Wpf were compared to those of Wp. The fluxes of the glycolytic pathway were very similar between the two strains. And the fluxes of Entner Doudoroff pathway from 6-phosphogluconate to 2-Keto-3-deoxy-6-phosphogluconate (KDPG) were zero at both two strains. However, the fluxes of the pentose phosphate pathway were changed (Figure 2A). The flux from 6-phosphogluconate to ribose-5-phosphate, the first step in the pentose phosphate pathway, increased from 35.62 to 40.88 in Wpf compared to Wp, and all other fluxes of the subsequent pentose phosphate pathway were increased. Other significant changes in central carbon metabolic fluxes were observed in the TCA pathway. For example, the flux of acetyl CoA to citrate, the first step of the TCA cycle, increased from 55.88 to 71.55 in Wpf compared to Wp, and the fluxes of all other TCA cycle pathways increased in Wpf compared to Wp (Figure 2A).

Based on the carbon flux results, cofactor generation and consumption were also calculated (He et al., 2014; Long et al., 2017, 2018). The results are also showed in Supplementary Table S4. With the increased TCA cycle and pentose phosphate pathway fluxes, the fluxes of NADH (FADH₂) and NADPH generation also increased 1.28- and 1.2-fold, respectively (Figures 2B,C). Increased NADH levels were used for oxidative phosphorylation. However, despite the increase in NADPH, the NADPH consumption flux for amino acid synthesis and biomass formation decreased by 0.87-fold. Due to this cellular redox imbalance, the flux of transhydrogenase was also changed (Canonaco et al., 2001). In the case of Wp, the flux for converting NADH to NADPH was 38.0173, whereas for Wpf, the flux for converting NADPH to NADH was observed to be 9.1855 (Figures 2B,C). The significant change in transhydrogenase flux was due to cellular redox imbalance, which might be caused by the excessive accumulation of NADPH in the Wpf strain by not forming flagella.

In addition, due to the increased TCA cycle flux, the ATP generation flux of the TCA cycle changed (Figure 2D). It increased 2.27-fold in Wpf compared to Wp. Subsequently, the increased cofactors were converted to ATP by oxidative phosphorylation. The flux of ATP generated by oxidative phosphorylation increased 1.28-fold in Wpf compared to Wp. Finally, at the section of others that are said to be surplus ATP, the flux increased 1.63-fold in Wpf compared to Wp. Large amounts of ATP resulted from deleting flhC.

The results of ¹³C-MFA similarly explain the changes in ATP levels and NADPH/NADP⁺ ratios shown in Figures 1C,D. A previous report showed that artificially increased intracellular ATP retarded cell growth (Na et al., 2015). Thus, we suspected that Wpf with high levels of ATP and NADPH induced growth retardation and reduced glucose consumption.

To reduce the metabolic burden caused by surplus production of ATP and NADPH in the Wpf strain, plasmid replications and protein overexpression were attempted. Plasmids pJKR-H and pZA31 MCS were transformed into the Wpf strain, and the resulting strains were named WpfJ and WpfZ, respectively. In addition, a gene encoding GFP was cloned into pZA31 MCS, and the resulting plasmid was transformed into Wpf, constructing the WpfE strain. The strains were cultivated in a shake flask in M9 glucose medium at 37°C. All three strains, WpfJ, WpfZ, and WpfE, showed growth restoration compared to the Wpf strain (Figure 3A). The growth restoration effect was most significant in the WpfE strain. Interestingly, glucose uptake patterns of those strains were about the same as the Wpf strain, while having much lower glucose uptake than the Wp strain (Figures 1B, 3B).

To investigate the metabolic burden of these strains, ATP levels and NADPH/NADP⁺ ratios were measured. The ATP levels decreased by 0. 88-, 0. 89-, and 0.86-fold in WpfJ, WpjZ, and WpfE, respectively, compared to Wpf (Figure 3C). The NADPH/NADP⁺ ratios also decreased from 2.08 in the Wpf strain to 1.54, 1.64, and 1.46 in WpfJ, WpjZ, and WpfE, respectively (Figure 3D). It is likely that the growth of the Wpf strain was recovered due to the consumption of excess ATP and NADPH by plasmid replication and heterologous protein expression.

The excess ATP and NADPH accumulation in cells can be beneficial for recombinant protein production (Lim et al., 2002; Lee et al., 2007; Kim et al., 2012). For example, excess ATP and NADPH were synthesized with a reduced amount of glucose consumption in the Wpf strain. When heterologous protein was expressed in the Wpf strain, EGFP was produced without increased glucose consumption. This means that the yield of protein synthesis per unit of glucose can be increased significantly in the Wpf strain. To verify the effect of flhC knockout on the yield of protein synthesis, the WpE and WpfE strains were constructed by transforming the plasmid-expressing EGFP into the Wp and Wpf strains, which were cultivated in a shake flask in glucose M9 medium at 37°C.

While growth was restored in WpfE compared to Wpf (Figure 3A and Supplementary Figure 1A), it was decreased in WpE compared to Wp (Supplementary Figure 1B). The growth rate of WpfE was comparable to that of WpE with significantly lower glucose consumption (Figures 4A,B). The glucose uptake decreased from 12.49 g/L to 7.54 g/L in WpfE compared to WpE, despite the similar cell densities reached in both strains. These results suggest that the metabolism of WpfE was much more efficient than that of WpE in order to produce biomass and protein with a unit of glucose. The EGFP yield per consumed glucose increased by 1.81-fold in WpfE compared to WpE (Figure 4C). The WpfE strain showed even higher protein production yield than Wp, which outperformed the wild-type strain in terms of protein yield (Jung et al., 2019).