Engineering the Outer Membrane Could Facilitate Better Bacterial Performance and Effectively Enhance Poly-3-Hydroxybutyrate Accumulation

ABSTRACT

Poly-3-hydroxybutyrate (PHB) is an environmentally friendly polymer and can be produced in Escherichia coli cells after overexpression of the heterologous gene cluster phaCAB. The biosynthesis of the outer membrane (OM) consumes many nutrients and influences cell morphology. Here, we engineered the OM by disrupting all gene clusters relevant to the polysaccharide portion of lipopolysaccharide (LPS), colanic acid (CA), flagella, and/or fimbria in E. coli W3110. All these disruptions benefited PHB production. Especially, disrupting all these OM components increased the PHB content to 83.0 wt% (PHB content percentage of dry cell weight), while the wild-type control produced only 1.5 wt% PHB. The increase was mainly due to the LPS truncation to Kdo₂ (3-deoxy-d-manno-octulosonic acid)-lipid A, which resulted in 82.0 wt% PHB with a 25-fold larger cell volume, and disrupting CA resulted in 57.8 wt% PHB. In addition, disrupting LPS facilitated advantageous fermentation features, including 69.1% less acetate, a 550% higher percentage of autoaggregated cells among the total culture cells, 69.1% less biofilm, and a higher broken cell ratio. Further detailed mechanism investigations showed that disrupting LPS caused global changes in envelope and cellular metabolism: (i) a sharp decrease in flagella, fimbria, and secretions; (ii) more elastic cells; (iii) much greater carbon flux toward acetyl coenzyme A (acetyl-CoA) and supply of cofactors, including NADP, NAD, and ATP; and (iv) a decrease in by-product acids but increase in γ-aminobutyric acid by activating σ^E factor. Disrupting CA, flagella, and fimbria also improved the levels of acetyl-CoA and cofactors. The results indicate that engineering the OM is an effective strategy to enhance PHB production and highlight the applicability of OM engineering to increase microbial cell factory performance.

IMPORTANCE Understanding the detailed influence of the OM on the cell envelope and cellular metabolism is important for optimizing the E. coli cell factory and many other microorganisms. This study revealed the applicability of remodeling the OM to enhance PHB accumulation as representative inclusion bodies. The results generated in this study give essential information for producing other inclusion bodies or chemicals which need more acetyl-CoA and cofactors but less by-product acids. This study is promising to provide new ideas for the improvement of microbial cell factories.

INTRODUCTION

Poly-3-hydroxybutyrate (PHB), a kind of biodegradable polyester synthesized by a wide variety of bacteria as carbon and energy storage inclusions, has been determined to be an environmentally friendly bioplastic also with promising applications in biofuels, chemicals, and pharmaceuticals (1). Ralstonia eutropha or Cupriavidus necator cells naturally accumulate PHB to enhance the survival of the cells in times of starvation (2, 3). The PHB biosynthesis pathway leading from central intermediates to PHB requires only three enzymatic steps, starting with the condensation of two molecules of acetyl coenzyme A (acetyl-CoA) (thiolase reaction, PhaA) to acetoacetyl-CoA and the subsequent reduction of acetoacetyl-CoA (reductase reaction, PhaB) to the monomeric precursor of PHB, 3-hydroxybutyryl-CoA (3HBCoA), which is polymerized to PHB by PHB synthase (PhaC) (Fig. 1a) (2). Under conditions of unbalanced growth (e.g., high C:N ratio of nutrients), PHB accumulates more easily (4). Gram-negative bacteria are usually used as industrial producers for PHB. In Escherichia coli, PHB could not be synthesized naturally but could be synthesized from acetyl-CoA by heterologous overexpression of the phaCAB operon cloned from R. eutropha with the supply of cofactor NADPH (Fig. 1a) (5, 6).

FIG 1.

LPS and PHB are synthesized from glucose, and 25 genes are relevant to the biosynthesis of LPS polysaccharide in E. coli K-12. (a) PHB is derived from acetyl-CoA and requires the cofactor NADPH. Glc, glucose; G6P, glucose-6-phosphate; 3PG, 3-phosphoglycerate; PEP, phosphoenol pyruvate; PP, pentose phosphate pathway; TCA, tricarboxylic acid cycle. The cell envelope of E. coli is composed of the outer membrane (OM), periplasm, and inner membrane (IM). On the cell surface, there are approximately 2 × 10⁶ LPS residues, much CA, flagella, and fimbria in E. coli cells. LPS is located in the outer leaflet of the OM and typically contains Kdo₂-lipid A and polysaccharides. Kdo, 3-deoxy-d-manno-oct-2-ulosonic acid; Hep, heptose; Glc, glucose; Gal, galactose; Rha, rhamnose. (b) Structure of LPS and related genes. Enzymes responsible for turning Kdo₂-lipid A into core-lipid A are shown in red, enzymes responsible for turning core-lipid A into LPS are shown in blue, and two gene clusters are required for synthesizing the polysaccharide of LPS in the E. coli W3110 genome. One cluster contains 14 genes (shown in red), and these genes encode enzymes that sequentially add monosaccharides to Kdo₂-lipid A, forming core-lipid A. The other cluster contains 11 genes (shown in blue). These genes encode enzymes that synthesize and polymerize the O-antigen and link the O-antigen repeats to core-lipid A, forming LPS. Gene wbbL contains an insH insertion.

Previous studies showed that E. coli strains JM109, DH5α, XL1-Blue, B (ATCC 11303), and HB101 are good hosts for PHB production, but E. coli strains ATCC 9631 (strain W) and ATCC 23716 (strain K-12) accumulated much less PHB when plasmid pSYL105 harboring the phaCAB operon was introduced; the considerable filamentation of cells accumulating PHB was observed for all these strains except K-12 and W (7). Furthermore, our previous study also showed that JM109 and DH5α could accumulate much more PHB than W3110 (ATCC 27325) when plasmid pDXW-8-phaCAB was introduced (8). Engineering better microbial cell factories has become more and more important for PHB production. Many effective efforts have been made in E. coli to increase PHB biosynthetic efficiency, including pathway engineering, synthetic biology approaches, process optimization, and cell shape engineering (9, 10). For metabolic pathway engineering, the enhancement of the acetyl-CoA pool (11, 12) and the cofactor supply are efficient (13, 14); for cell size engineering, modifying cell division (15, 16), reducing cell wall thickness (17), and prolonging cell life span (18) to increase cell volume also efficiently enhance the accumulation of many inclusion bodies. In summary, PHB accumulation depends on intracellular metabolism and cell size.

Indeed, the outer membrane (OM) in Gram-negative bacteria has been proven to be the most important factor in determining the stiffness and strength of Escherichia coli cells (19), which would influence the cell size when PHB accumulates. Lipopolysaccharide (LPS) forms the outer layer of the OM in Gram-negative bacteria (Fig. 1a) (20, 21). In addition, LPS is critical for the formation of other molecules on the OM, including colanic acid (CA) (22), flagella (23), and Omps (24). There are approximately 2 × 10⁶ LPS molecules in an E. coli cell occupying 25% of cell envelope components. LPS typically consists of Kdo₂ (3-deoxy-d-manno-octulosonic acid)-lipid A and polysaccharides (Fig. 1b) (20). Kdo₂-lipid A is the minimal LPS for the growth of E. coli, and polysaccharides could help cells resist environmental stress (25, 26). The biosynthesis of polysaccharides requires two gene clusters (Fig. 1b). One cluster contains 14 genes. These genes encode enzymes that sequentially add monosaccharides of precursors for heptose, glucose, and galactose to Kdo₂-lipid A, forming core-lipid A; the other cluster contains 11 genes. These genes encode enzymes that synthesize and polymerize the O-antigen and link the O-antigen repeats to core-lipid A, forming LPS. The synthesis of monosaccharides for LPS is derived from the glucose 6-phosphate (G6P) formed in the glycolytic pathway (Fig. 1a) (10). The rigidity and low fluidity of LPS are mainly attributed to the negatively charged phosphate groups and carboxyl groups on the LPS molecule, which can be mediated by divalent cations such as Ca²⁺ and Mg²⁺ to form electrostatic cross-linking interactions (27). Apart from LPS, exopolysaccharide, flagella, and type I fimbria are also components of the cell envelope in E. coli (28) (Fig. 1a). In E. coli, the major exopolysaccharide (EPS) is CA (Fig. 1a). The monosaccharides of CA are also derived from G6P (10). These molecules also have long biosynthetic pathways (10). As numerous and important molecules in the cell envelope, the LPS, CA, flagella, and fimbria not only function as cell barriers but also consume many carbon (C), nitrogen (N), and energy sources.

Our previous studies suggested that the removal of flagella and pili from Pseudomonas putida could enhance polyhydroxyalkanoate (PHA) production (29). Our recent study also confirmed that truncating LPS by deleting rfaD to disrupt l-d-heptose formation for LPS can efficiently enhance the production of PHB to 67.8 wt% (PHB content percentage of dry cell weight) in E. coli W3110 (8). Remarkably, the results identified the close link of OM with PHB accumulation; however, the mechanism is still unclear, and the influences on PHB accumulation by disrupting all the other unnecessary genes relevant to LPS, and other unnecessary components of OM, still remain under investigation.

In this study, we engineered the OM by blocking all the additions of monosaccharides and modifier groups to Kdo₂-lipid A of LPS, CA, flagella, and/or fimbria in E. coli W3110. To study PHB accumulation, we used the common plasmid pBHR68 harboring the phaCAB operon cloned from R. eutropha (5). pBHR68 is derived from a pBluescript-derived multicopy vector with ampicillin resistance and was constructed by B. Rehm in the Steinbuchel laboratory (5). When plasmid pBHR68 was introduced, in comparison with E. coli W3110/pBHR68, OM-engineered WJW08/pBHR68 cells accumulated 83.0 wt% PHB, similar to that of LPS-engineered WJW02/pBHR68 cells, which are completely filled with PHB to expand the cell volume up to 25-fold and to produce PHB content of 82 wt%, much greater than that of a mutant with only RfaD removed (67.8 wt%). Furthermore, we also found that disrupting gene clusters relevant to CA, flagella, or fimbria also benefited PHB accumulation, especially CA. Moreover, we systematically studied the mechanism by analyzing in detail global metabolic and envelope regulations with the assistance of transcriptomics and metabolomics analysis. The results indicated that engineering the OM is an efficient strategy to improve cell factory performance to accumulate PHB.

RESULTS

Disrupting LPS, CA, flagella, and fimbria could influence cell characteristics, and LPS truncation had the most significant impacts.

To better understand the influences on cell characteristics and PHB accumulation by OM engineering, we constructed a series of OM mutants. The details of deletions in the different mutants are shown in Fig. S1 in the supplemental material. LPS is the most important component of the OM. We blocked all the unnecessary monosaccharides and modifier groups in LPS by deleting 25 genes (gmhD-waaQ and wbbL-rmlB) relevant to the polysaccharide portion of LPS in E. coli W3110, resulting in mutant WJW02. Then, the 119 genes responsible for the biosynthesis of CA (28 genes, including galF-wza and yccC-ymcD), flagella (57 genes, including flhE-motA, fliY-fliR, and flgN-flgL), and fimbria (9 genes, including fimB-fimH) were further deleted in WJW02, resulting in WJW08. To better study the influences of other OM structures, we also constructed separate mutants lacking CA, flagella, or fimbria by disrupting 28 genes, 57 genes, and 9 genes, respectively, resulting in mutants WJW09, WJW010, and WJW011.

LPS was extracted from WJW02 cells and determined to be Kdo₂-lipid A by SDS-PAGE and electrospray ionization-tandem mass spectrometry (ESI-MS/MS). An SDS-PAGE image showed that WJW02 LPS migrated much faster than W3110 LPS, suggesting that WJW02 synthesizes shorter LPS (Fig. 2a). Furthermore, the ESI-MS/MS analysis showed that peaks around m/z 1,796, 2,016, and 2,236 were observed for WJW02 (Fig. 2b). The peak at m/z 2236.4 should be the [M-H]⁻ ion of Kdo₂-lipid A (30). The peaks at m/z 2,016.4 and 1,796.6 should be derived from the Kdo₂-lipid A ion by the loss of one and two Kdo residues, respectively. This confirmed that the shorter LPS made by WJW02 cells is Kdo₂-lipid A.

FIG 2.

Influences on E. coli OM and cell characteristics of different deletions of LPS, exopolysaccharide (EPS), flagella, and type I fimbria. (a) SDS-PAGE analysis of LPS isolates from W3110 and WJW02 cells. WJW02 LPS migrates faster than W3110 LPS. (b) ESI-MS/MS determination of Kdo₂-lipid A produced by WJW02. Peaks around m/z 1,796, 2,016, and 2,236 should represent lipid A (lack of two Kdos), Kdo-lipid A (lack of one Kdo group), and Kdo₂-lipid A, and the peaks of m/z 1,796.6 and 2,016.4 should be derived from Kdo₂-lipid A. (c) TEM images of E. coli W3110, WJW02, WJW08, WJW09, WJW010, and WJW011 cells. W3110 cells show many secretions (secreted substrates like EPS), flagella, and fimbria around the cells, but they were not found around WJW02 and WJW08 cells. Flagella are indicated by black arrows; the flagella and fimbria could be observed for WJW09 cells, and fimbria/secretions and flagella/secretions could be observed for WJW010 and WJW011, respectively. (d) CA formation of different E. coli strains. (e) Motility comparison of different E. coli W3110, WJW02, WJW09, WJW010, and WJW011 cells. (f) Growth curves of different OM mutants with the control E. coli W3110 incubated in LB, at 37°C, 200 rpm. (g) Growth curves of different OM mutants with the control E. coli W3110 incubated in M9 with the addition of 20 g/liter glucose, at 37°C, 200 rpm. Error bars show the standard deviation from the mean of three independent experiments.

The influences on E. coli cell morphology by different disruptions on the OM were then studied by transmission electron microscopy (TEM) analysis. The TEM images showed that W3110 had secretions (secreted substrates like extracellular polysaccharides), flagella, and fimbria around the cells (Fig. 2c); WJW09 cells were similar to W3110 cells, WJW010 cells lacked flagella, and WJW011 cells lacked fimbria (Fig. 2c). It was interesting that the morphology of WJW02 cells was quite different from that of W3110 cells (Fig. 2c). First, W3110 cells had spindle-shaped ends, while WJW02 cells had flat blunt ends. Second, W3110 cells had flagella and fimbria (black arrows) around them, but WJW02 cells did not. Third, W3110 cells had secretions around them, but WJW02 cells did not.

According to a CA detection assay, there is almost no CA in WJW09, and a small amount of CA (0.0052 g/liter) was detected in E. coli W3110 in M9G (M9 medium supplemented with 20 g/liter glucose) at 37°C (Fig. 2d). Compared to CA formation in W3110, CA formation was decreased in WJW02 culture, similar in WJW011 culture, but increased in WJW010 culture (Fig. 2d). The motility assay showed that WJW02 cells had much weaker motility than W3110 (Fig. 2e), WJW010 cells lacked motility, WJW011 cells showed slightly weaker motility, and WJW09 showed motility similar to that of W3110 cells (Fig. 2e), consistent with the above-described TEM analysis.

In addition, we studied the effect of the deletions on growth on different media (complex/mineral salts) (Fig. 2f and g). WJW02 cells did not impact growth rate during log phase, showing only approximately 6% lower final cell optional density (optical density at 600 nm [OD₆₀₀]) either for LB or M9G medium (Fig. 2f and g). Even though the additional removals of CA, flagella, and fimbria in WJW02 could slightly increase the growth rate, the final biomass of WJW08 was similar to that of WJW02. This indicated that the LPS truncation could slightly disturb cell growth, while the disruption of flagella and type I fimbria facilitated approximately 5% and 6% higher final OD₆₀₀, respectively, than that of W3110 in M9G. This indicates that the different OM disruptions did not cause significant inhibition in cell growth. These results indicate that the different disruptions in OM in E. coli achieved corresponding structural changes and had different influences on cell characteristics. Remarkably, LPS truncation had the most significant impacts.

Disrupting LPS, CA, flagella, and fimbria in OM could enhance PHB production in E. coli, and LPS engineering facilitates the most significant enhancement.

With these mutants, we studied the applicability of engineering OM to improve PHB as an inclusion body representative. Plasmid pBHR68 was introduced into the mutants and control strain W3110, resulting in WJW02/pBHR68, WJW08/pBHR68, WJW09/pBHR68, WJW010/pBHR68, WJW011/pBHR68, and W3110/pBHR68, respectively. After 24 h of fermentation in M9G, the cell morphology of and PHB granule formation by these E. coli cells were analyzed by electron microscopy analysis. According to ultrathin section TEM (ultra-TEM) images (Fig. 3a), WJW02/pBHR68 cells were completely filled with large PHB particles, but almost no PHB particles were observed in W3110/pBHR68 cells. WJW02/pBHR68 cells became enlarged and increased up to 25-fold compared with W3110/pBHR68 cells. Unlike other PHB producers whose cells become longer when accumulating PHB, i.e., by the filamentation of cells accumulating PHB (7), there was no filamentation of cells for WJW02/pBHR68. But almost all of the WJW02/pBHR68 cells showed very robust expansion and were completely filled with PHB granules. According to ultra-TEM images, a large amount of PHB accumulated in WJW02/pBHR68 cells, but almost no PHB granules were observed in W3110/pBHR68 cells (Fig. 3b). PHB stained with Nile red emitted green fluorescence under the conditions of excitation at 488 nm and emission at 530 nm. W3110/pBHR68 and WJW02/pBHR68 cells stained with Nile red were observed by fluorescence microscopy. Fluorescent foci were not detected in W3110/pBHR68 cells, but stronger green fluorescent foci completely filled WJW02/pBHR68 cells; furthermore, WJW02/pBHR68 cells were significantly enlarged, with many green bulges on the cell surface (Fig. 3b). For other OM mutants, most WJW09/pBHR68 cells also became larger (about 4-fold) and accumulated more PHB than W3110/pBHR68 cells, and many WJW010/pBHR68 cells also produced more PHB than W3110/pBHR68 cells but less than WJW09/pBHR68, while WJW011/pBHR68 cells produced a small amount of PHB, slightly greater than that produced by W3110/pBHR68 cells (Fig. 3a). WJW09/pBHR68, WJW010/pBHR68, and WJW011/pBHR68 cells were smaller in size than WJW02/pBHR68 cells. This indicates that disrupting LPS may benefit membrane elasticity to create more space for accumulation of PHB.

FIG 3.

Electron microscopy images, flask fermentation profiles, and cell phenotypes of WJW02/pBHR68 and W3110/pBHR68 cells. (a) Ultrathin section TEM images of W3110/pBHR68, WJW02/pBHR68, WJW09/pBHR68, WJW010/pBHR68, and WJW011/pBHR68 cells. (b) Fluorescence microscopy images of W3110/pBHR68 and WJW02/pBHR68 cells. DIC, differential interference contrast; FITC, fluorescence (fluorescein isothiocyanate) contrast; merged, overlapping contrast of both DIC and FITC. (c) Growth curves of different recombinants during PHB fermentation. (d) Dry cell weight (DCW). (e) PHB contents (wt%). (f) Residual biomass (DCW-PHB) of W3110/pBHR68, WJW02/pBHR68, WJW08/pBHR68, WJW09/pBHR68, WJW010/pBHR68, and WJW011/pBHR68 cells. (g) Conversion efficiency from glucose to PHB. (h) Culture pH. (i) Extracellular acetate determined by HPLC. (j) Broken cell ratio. (k) Autoaggregation after standing for 60 min. (l) Biofilm formation in shaking situation at 200 rpm after 48 h. Error bars show the standard deviation from the mean of three independent experiments.

After 24 h of flask (batch) fermentation, in comparison to W3110/pBHR68, WJW02/pBHR68 showed 2.1-fold higher optional density (Fig. 3c). The OD₆₀₀ values of WJW02/pBHR68 and WJW08/pBHR68 showed significant increases from 0 to 12 h, and those of WJW02/pBHR68 and WJW08/pBHR68 showed further increases from 12 to 24 h, but the OD₆₀₀ values of W3110/pBHR68 showed an obvious increase only from 0 to 6 h and almost no increase from 6 to 24 h (Fig. 3c). Further observations by microscope for WJW02/pBHR68 cells after 6 h, 12 h, 18 h, and 24 h of fermentation showed that almost every cell was completely filled with PHB granules, and the cells became larger with fermentation time (see Fig. S2 in the supplemental material). This suggests that the PHB granules could be well accumulated in WJW02/pBHR68 cells during the whole 24 h of fermentation. Hence, WJW02/pBHR68 cells showed much greater dry cell weight (DCW) (4.62 g/liter) and PHB content (82.0 wt%) than W3110/pBHR68 cells, which showed only 1.32 g/liter DCW and 1.5 wt% PHB (Fig. 3d and e). WJW02/pBHR68 cells converted 34.5% glucose into PHB, while W3110/pBHR68 cells converted only 0.2% glucose into PHB in (Fig. 3g). Compared with WJW02/pBHR68, WJW08/pBHR68 cells showed similar optional density, DCW (4.7 g/liter), PHB content (83.0%), and glucose conversion efficiency (35.2%) (Fig. 3c to e and g), much greater than W3110/pBHR68 cells, despite that their residual biomass was smaller than that of W3110/pBHR68 cells (Fig. 3f). The reduced biomass may be caused by LPS truncation and PHB stress, which leads to decreased soluble proteins for biomass (31). Indeed, E. coli strain K-12 is not a good host for PHB accumulation (7), and its production of PHB is generally much less than that of other E. coli strains, such as DH5α and JM109 (8). The disadvantage of E. coli W3110 for producing PHB further strongly indicated that LPS disruption is an efficient strategy to improve PHB production.

We further studied the influences on PHB production by separate disruptions of CA, flagella, and fimbria. After 24 h of fermentation, WJW09/pBHR68, WJW010/pBHR68, and WJW011/pBHR68 showed 140%, 47%, and 28% higher OD₆₀₀ values (Fig. 3c). WJW09/pBHR68 showed increasing OD₆₀₀ from 0 to 12 h, but mainly from 0 to 6 h for WJW010/pBHR68 and WJW011/pBHR68 with growth trends similar to that of W3110/pBHR68. CA-engineered mutant WJW09/pBHR68 cells produced 2.92 g/liter of DCW and accumulated 57.2% PHB (wt%) with 19.6% conversion efficiency from glucose to PHB, and WJW010/pBHR68 cells produced 1.9 g/liter of DCW and accumulated 19.2% PHB (wt%) with 5.6% conversion efficiency, while WJW011/pBHR68 cells accumulated 2.8% PHB (wt%) with only 0.64% conversion efficiency, slightly higher than that of W3110/pBHR68 (Fig. 3d, e, and g). The residual biomasses of WJW010/pBHR68 and WJW011/pBHR68 were higher than that of W3110/pBHR68 (Fig. 3f), which further suggested that the deletion of flagella and fimbria benefits cell growth. The results suggested that blocking CA could significantly contribute to the enhancement of PHB accumulation, that lacking flagella is also beneficial for accumulating PHB, and that deleting type I fimbria has a slight influence on PHB accumulation. Taken together, the results solidly prove that engineering LPS, CA, flagella, and fimbria all benefit PHB accumulation and that LPS truncation facilitates the most significant enhancement of PHB production.

Moreover, we found that the culture pH of WJW02/pBHR68 cells was 6.5, while that of W3110/pBHR68 cells was 4.7, and WJW09/pBHR68 (pH 5.1), WJW010/pBHR68 (pH 4.87), and WJW011/pBHR68 (pH 4.88) showed slightly higher pH values than W3110/pBHR68 (Fig. 3h). Culture pH is closely related to acetate accumulation as a by-product acid. WJW02/pBHR68 synthesized 1.47 g/liter acetate, 69.1% lower than W3110/pBHR68, which synthesized 4.76 g/liter acetate, and WJW09/pBHR68, WJW010/pBHR68, and WJW011/pBHR68 produced slightly lower acetate levels of 4.1, 4.4, and 4.2 g/liter, respectively (Fig. 3i). This suggested that the OM engineering could decrease the formation of by-product acetate for PHB.

To better understand the influence on bacterial fermentation features, we also analyzed the broken cell ratio, autoaggregation, and biofilm formation of these recombinant E. coli mutants by using W3110/pBHR68 as a control. As shown in Fig. 3j, more WJW02/pBHR68 cells were broken than W3110/pBHR68 cells at the same time, suggesting that WJW02 cells have less pressure durability due to LPS disruption. The autoaggregation ratio represents cellular autoaggregation ability, defined as the percentage of aggregated cells among the total culture cells. The autoaggregation ratio of WJW02/pBHR68 cells was significantly increased (Fig. 3k), possibly because LPS truncation in WJW02 exposed antigen 43 (Ag43) on the cell surface, leading to more autoaggregation (32). Ag43 is an OM protein closely related to the autoaggregation ability of E. coli cells, and the expression of Ag43 plays an important role in the process of bacterial autoaggregation in recognizing cells and adherence between cells (33). Other mutants showed broken cell ratios and autoaggregation abilities similar to those of W3110/pHBR68 cells. The biofilm formation ability of WJW02/pHBR68 cells was significantly decreased, by 69.1%, suggesting that LPS plays an important role in the complex biofilm formation process, and WJW09/pBHR68 and WJW011/pHBR68 cells showed 9.8% and 50.8% lower biofilm formation (Fig. 3l). These features are all beneficial to industrial applications.

Engineering LPS weakens cell envelope thickness and rigidity, leading to elastic cells.

To better investigate the mechanism of influence on PHB accumulation by disrupting LPS, we studied the cell membrane morphology of E. coli WJW02 and W3110. The above-described TEM analysis (Fig. 2c) suggested that engineering LPS could reduce the formation of flagella and fimbria on the OM, and further transcriptome analysis also showed that the genes responsible for membranes and cell wall components, especially flagella, were downregulated in WJW02 compared to W3110 (Fig. S3), which also explained why PHB accumulation in WJW08/pBHR68 cells was similar to that in WJW02/pBHR68 cells. Furthermore, ultra-TEM analysis suggested that the envelope of WJW02 cells became thinner than that of W3110 cells. The images in Fig. 4a show that the thickness of the cell wall with OM (white arrows) of WJW02 cells was 10 to 11 nm, which was approximately 20% lower than that that of W3110 cells (13 to 15 nm) (Fig. 4a). In addition, the periplasmic distance between peptidoglycan and the inner membrane (IM) was lower in WJW02 cells than in W3110 cells (Fig. 4a and b). But the cell envelope and thickness ofWJW09, WJW010, and WJW011 cells were similar to those of W3110 cells (Fig. 4a). Interestingly, we found that the transcriptional level of the lpp gene, responsible for lipoprotein Lpp, was significantly downregulated, by 85%, in WJW02 cells compared to W3110 cells. The decrease in Lpp is also beneficial in reducing periplasmic size (34, 35), and it could be repressed by SlrA small RNA (sRNA), which could be activated by the global regulator σ^E (36, 37) (Fig. 4b). σ^E can be activated to the highest level when LPS is truncated to Kdo₂-lipid A (38). A thinner cell envelope could enhance PHB accumulation (17). The results suggest that LPS engineering is beneficial for weakening the cell envelope thickness and rigidity, leading to elastic cells beneficial for inclusion body accumulation (Fig. 4b).

FIG 4.

Influences on cell membrane morphology by disrupting LPS, CA, flagella, or type I fimbriae. (a) Ultra-TEM images of E. coli W3110, WJW02, WJW09, WJW010, and WJW011 cells. The cell wall and OM (white arrows), IM (cyan blue arrows), and periplasmic space are all visible, the outside layer is LPS (red arrows) in W3110, WJW09, WJW010, and WJW011 and Kdo₂-lipid A (brown arrow) in WJW02, and the innermost layer of the cell wall is peptidoglycan (PGN) (black arrows). (b) The mechanistic model of LPS truncation facilitates better membrane characteristics. Cell wall stiffness is attributed to LPS polysaccharide interactions through salt bridge connections with negatively charged phosphate groups on LPS. The lipoprotein Lpp repressed by SlrA sRNA has been proven to control the periplasmic size, and Lpp and periplasmic size are decreased in WJW02 cells; flagella and type I fimbria are significantly reduced in WJW02 cells. The lack of interactions between LPS molecules by salt bridge connection could facilitate weakened cell wall thickness and rigidity, beneficial to the enlargement of cell volumes more easily when PHB granules accumulate.

Engineering OM facilitates higher levels of acetyl-CoA and cofactors NADP, NAD, and ATP.

PHB biosynthesis needs precursor acetyl-CoA and cofactor NADPH. We found that LPS truncation could significantly improve the levels of acetyl-CoA and cofactors, including not only NADPH but also NADH and ATP (Fig. 5). Ultraperformance liquid chromatography-mass spectrometry (UPLC-MS) analysis showed that WJW02 produces 6.45-fold more acetyl-CoA than W3110 (Fig. 5a). And the acetyl-CoA levels of WJW09, WJW010, and WJW011 increased 25.5%, 19.2%, and 7.7% relative to that of W3110 (Fig. 5a). WJW02 cells showed 126% and 236% higher ATP levels at early and late log phase, respectively, than W3110 cells (Fig. 5b), much higher than that of WJW09, WJW010, and WJW011 cells. At early log phase, the ATP levels of WJW09, WJW010, and WJW011 increased 16.2%, 24.2%, and 10.7% compared with that of W3110; at late log phase, WJW010 and WJW011 increased 55.9% and 38.7%, respectively, while WJW09 showed no change in comparison with W3110 (Fig. 5b). This result suggests that the lack of flagella or fimbria could save cellular energy but is still much less significant than LPS simplification. With respect to NAD determination, WJW02 and WJW010 showed 120% and 153% higher NAD_total levels, respectively (Fig. 5c). With respect to NADH/NAD⁺ ratios, WJW02 showed 6.66-fold higher ratios than W3110, and those of WJW09, WJW010, and WJW011 increased 79.8%, 84.4%, and 34.2% relative to W3110, respectively (Fig. 5c). With respect to NADP determination, WJW02 and WJW010 showed 77.0% and 17.6% higher NADP_total levels, respectively, than W3110 (Fig. 5d). Despite the fact that WJW09 showed a 62.1% lower NADP_total level than W3110, it had a 142% higher NADPH/NADP⁺ ratio than W3110 (Fig. 5d). In summary, the results demonstrate that engineering the OM benefits the biosynthesis of acetyl-CoA and cofactors to produce more PHB. The mutant WJW02 facilitates the most acetyl-CoA and cofactors, and WJW09 and WJW010 could also increase the levels of them.

FIG 5.

Influences on cellular metabolism by engineering LPS, CA, flagella, and type I fimbria. (a) Determination of intracellular acetyl-CoA for different mutants. (b) Relative levels of ATP. (c) NAD levels and NADH/NAD⁺ ratio. (d) NADP levels and NADPH/NADP⁺ ratio. Error bars show the standard deviation from the mean of three independent experiments. E. coli cells were fermented in minimal medium supplemented with 20 g/liter glucose (M9G) at 37°C, 200 rpm. For acetyl-CoA determination, E. coli cells were grown to an OD₆₀₀ of 1.5; for cofactor NADP, NAD, and ATP determination, the cells were grown in M9G to early log phase (OD₆₀₀, 1.5) and mid-log phase (OD₆₀₀, 2.5). The unit “μmol/g Protein” refers to micromoles of NADP or NAD per gram of total cellular protein.

To investigate whether and how LPS engineering influences metabolic pathways, we analyzed key intracellular intermediate metabolites involved in carbon metabolism of WJW02 by transcriptomics, metabolomics, and LC-MS with W3110 as the control. Transcriptomics analysis showed that the glycolytic pathway (Embden-Meyerhof-Parnas [EMP]) and tricarboxylic acid (TCA) cycle were both upregulated in WJW02 (Fig. 6a), especially the genes responsible for glucose uptake and acetyl-CoA formation. Notably, the important regulator fnr that is responsible for regulating many reaction steps of the EMP pathway and TCA cycle is significantly upregulated by 2^19.9-fold (978,356-fold), and the reaction steps producing NADH and ATP are also upregulated. For the pentose phosphate (PP) pathway, the genes tktA and talB responsible for forming sedoheptulose-7-phosphate (S7P, a precursor of heptose in LPS) were downregulated, but the genes tktB and talA for recycling back to the EMP pathway were upregulated, and the genes zwf, pgl, and gnd responsible for forming NADPH were also upregulated (Fig. 6a). We also determined the metabolites derived from the EMP, TCA, and PP pathways using metabolomics analysis (Fig. 6b). Consequently, the levels of acetyl-CoA were significantly increased 6.45-fold in WJW02 compared to W3110 (Fig. 6c). As shown in Fig. 6b, compared to W3110, the key intermediate metabolites glucose-1-phosphate (G1P), 3-phosphoglyerate (3PG), and phosphoenolpyruvate (PEP) of the EMP pathway and the intermediate metabolites citrate (CIT), succinate (SUCC), fumarate (FUM), and malate (MAL) of the TCA cycle were all improved, but G6P, fructose-6-phosphate (F6P), and pyruvate of the EMP pathway were decreased in WJW02. In addition, the cofactors NADPH, NADH, and ATP mainly formed from the EMP, TCA, and PP pathways were significantly increased in WJW02, and the related cofactor precursors NAAD and AMP were also increased. This result suggests that engineering LPS facilitated more carbon flux into acetyl-CoA and more cofactor formation.

FIG 6.

Carbon metabolic pathway analysis of E. coli WJW02 compared to W3110. (a) Transcriptional analysis of glucose metabolism regulation. Upregulated genes with expressed ratios are colored red, and downregulated genes with expressed ratios are colored blue. Reactions shown with blue arrows in addition to black arrows are regulated by the regulator Fnr. (b) Metabolomics analysis of intracellular metabolites and cofactors. The determination of the cofactors NAD, NADP, NAAD, ATP, and AMP was performed using LC-QQQMS, and the others were determined using GC-Q-TOF-MS. Error bars show the standard deviation from the mean of six independent experiments. (c) UPLC-MS analysis for acetyl-CoA. (d) Other sugar biosynthesis from glucose and related genes. The solid black circle represents a glucose molecule. Blue numbers represent different genes. Abbreviations: EMP, Embden-Meyerhof-Parnas pathway; TCA, tricarboxylic acid cycle; PP, pentose phosphate; GLC, glucose; G1P, glucose-1-phosphate; G6P, glucose-6-phosphate; F6P, fructose-6-phosphate; FDP, fructose-1,6-bisphosphate; DHAP, dihydroxyacetone phosphate; GA3P, glyceraldehyde-3-phosphate; GL3P, glycerate-phosphate; DPG, glycerate-1,3-phosphate; 3PG, 3-phosphoglycerate; 2PG, 2-phosphoglycerate; PEP, phosphoenol pyruvate; PYR, pyruvate; LAC, lactate; AcCoA, aectyl-CoA; ETH, ethanol; AC, acetic acid; CIT, citrate; ICIT, isocitrate; AKG, α-ketoglutarate; SuccCoA, succinyl-CoA; SUCC, succinate; FUM, fumarate; MAL, malate; OAA, oxaloacetate; 6PG, 6-phosphogluconate; 6PGA, 6-phosphoglucono-δ-lactone; RU5P, ribulose-5-phosphate; X5P, xylulose-5-phosphate; R5P, ribose-5-phosphate; S7P, sedoheptulose-7-phosphate; GA3P, glyceraldehyde-3-phosphate; E4P, erythrose-4-phosphate.

We also found that the carbon flux directed toward other metabolites, including UDP-GlcNAc, mannose-6-phosphate, sucrose, hexose, trisaccharide, glycerol-3-galactoside, N-acetyl-mannosamine, N-acetyl-d-hexosamine, β-mannosyl-glycerate, and lactitol, was decreased in WJW02. Indeed, UDP-GlcNAc is the main component of LPS, and the biosynthesis of other monosaccharides derived from the EMP (G6P) or PP (S7P) pathway for LPS is also decreased (Fig. 6b). The decreases in G6P and F6P in WJW02 corresponded with the weakened carbon flux to other sugar metabolism pathways identified by transcriptomic analysis (Fig. 6d). Therefore, LPS engineering could effectively improve the carbon flux toward acetyl-CoA, form cofactors, and reduce the carbon flux to other pathways.

Engineering LPS regulates the formation of acetate and GABA by activating acid resistance systems.

During the 48-h culture of E. coli W3110 and WJW02 incubated in M9G, the pH value of WJW02 was higher than that of W3110 (Fig. 7a). We then determined the extracellular organic acid and amino acid levels and found that WJW02 cells accumulated much less acetate but much more gamma-aminobutyric acid (GABA). In a comparison with W3110, WJW02 produced 2.4 g/liter acetate, 45.3% lower than that produced by W3110 (Fig. 7b), but WJW02 produced 0.583 g/liter GABA, 42-fold more than W3110 (Fig. 7c). This suggests that the higher pH of WJW02 culture was due to not only the decreased acetate but also the increased GABA. These differences in WJW02 not only reduce by-products to save more carbon flux for acetyl-CoA but also consume more N sources. However, with respect to pH and acetate formation, WJW09, WJW010, and WJW011 showed no obvious differences from control strain W3110 (Fig. 7a and b), while they produced slightly more GABA than W3110 (Fig. 7c). Therefore, the enhancement of acetyl-CoA and cofactors including NADP, NAD, and ATP by LPS engineering may be partly due to the lack of CA, flagella, and fimbria in WJW02 but not to the pH difference. It also suggests that engineering CA, flagella, and fimbria could influence the acetyl-CoA and cofactor levels but had no obvious influences on other features.

FIG 7.

pH changes and acetate and GABA synthesized in E. coli OM mutants, and activation of AR systems analysis in WJW02 compared to W3110. (a) Comparison of culture pHs during 48 h of flask fermentation. (b) Comparison of extracellular acetate levels determined after 24 h of fermentation. (c) Comparison of extracellular γ-aminobutyric acid (GABA) levels after 24 h of fermentation. For panels a to c, error bars show the standard deviation from the mean of three independent experiments. (d) Metabolomics analysis of intracellular metabolites derived from nitrogen metabolism according to LC-QQQMS. “w” represents W3110, and “2” represents WJW02. Error bars show the standard deviation from the mean of six independent experiments. (e) Signal model regulation by Kdo₂-lipid A. Kdo₂-lipid A activates σ^E by combining with RseB, which is necessary for inactivated σ^E. Regulator σ^E further activates stress resistance genes. (f) Transcriptional regulation of genes relevant to AR systems in WJW02 using W3110 as the control. AR1 and AR2 are the main AR systems in E. coli cells. Genes marked with red asterisks are positive regulators, and genes marked with blue asterisks are negative regulators of acid resistance. Red bars denote upregulated genes, and blue bars denote downregulated genes.

Remarkably, the more highly secreted GABA is beneficial for the conversion of intracellular available N sources to extracellular unavailable N products. In addition, metabolomics analysis showed that N metabolites to form urea were also increased in WJW02 cells, including adenine, hypoxanthine, xanthine, and urate (Fig. 7d). Urea is also an unavailable N and benefits the higher pH. The results suggest that engineering LPS could significantly upregulate the N metabolism pathway to reduce the available N, which would contribute to N limitation inside WJW02 cells. The C savings and N consumption both contribute to the higher C:N ratio in WJW02 than in W3110.

In E. coli, outer membrane integrity is essential for cell survival and is monitored by the σ^E stress response system (37) (Fig. 7e). In WJW02, LPS was truncated to Kdo₂-lipid A, which could activate the global regulator σ^E to the highest levels (38). The activated regulator RpoE could further activate stress resistance systems, including membrane repair systems and acid resistance (AR) systems (Fig. 7e). Indeed, the regulator rpoE was upregulated by 2.08-fold in WJW02 compared with W3110. As a result, the genes and regulators involved in the AR systems (39) were significantly upregulated in WJW02, especially the acid fitness island for AR1 (Fig. 7f) (40). The genes gadA, gadB, and gadC responsible for AR2 were also significantly upregulated (Fig. 7f). The decarboxylases GadA and GadB convert glutamate to GABA, and GadC transports GABA out in exchange for glutamate, which explains why a large amount of GABA accumulates in WJW02. This result indicates that engineering LPS could activate the global regulator σ^E to upregulate AR systems and further regulate organic acid accumulation and the N metabolism pathway, which is beneficial for establishing a better intracellular metabolism environment for higher C:N ratios.

DISCUSSION

In this study, we highlighted the applicability of engineering the OM to improve cell factories for PHB accumulation. We found that LPS engineering contributes the most to the OM for enhanced PHB production, due to the absence of formation of CA, flagella, and fimbria and the best metabolism and cell membrane features for improved PHB production. Compared to the control W3110/pBHR68 cells with only 1.5 wt% PHB, the OM-engineered WJW08/pBHR68 cells significantly improved the PHB content to 83.0 wt% and showed a much higher glucose conversion efficiency of 35.2%, and LPS-engineered WJW02/pBHR68 cells completely filled with PHB significantly increased the cell volume up to 25-fold and increased the PHB content to 82.0 wt% with a glucose conversion efficiency of 34.5%, similar to WJW08/pBHR68. This enhancement is much more significant than that of the mutant with only rfaD removed, which facilitated 67.8 wt% PHB and 22% conversion efficiency. Moreover, separate disruption of CA and flagellum engineering are also very beneficial to PHB production. Remarkably, previous studies showed that E. coli strain JM109(pSYL105) accumulated PHB to the highest extent (85.6 wt%), and E. coli XL1-Blue(pSYL105) yielded the highest glucose conversion efficiency of 0.369 g PHB/g glucose (7). In this study, in which we engineered the OM of E. coli K-12 W3110 (ATCC 27325) by deleting the gene clusters responsible for the polysaccharide of LPS, CA, flagella, and type I fimbria, PHB production was increased up to 83 wt% from 1.5 wt% and with 0.35 g/g conversion efficiency, close to that of E. coli strain JM109 or E. coli XL1-Blue, which further suggested that the OM engineering for E. coli W3110 was an efficient strategy to realize high PHB production. The results broaden engineering strategies to improve cell factories for PHB accumulation as an inclusion body.

Our results also indicated that truncating LPS is beneficial for PHB accumulation. Disrupting LPS could result in a similar enhancement of PHB production of the OM-engineered mutant by disrupting not only LPS but also the CA, flagella, and fimbria. Our further mechanism investigations could well explain the phenomenon. The decrease in flagella and fimbria observed in WJW02 cells contributes to PHB accumulation. Our previous studies showed that deleting genes relevant to LPS polysaccharide biosynthesis, such as waaC, gmhD, waaL, waaF, and waaG, have a significant influence on flagellum formation (22, 23, 41). We also determined that blocking CA, flagella, or fimbria could enhance PHB production to different degrees, especially CA. Our recent study demonstrated that the deletion of waaC could block the glycosyl precursor supply for CA formation (22); moreover, another recent study suggested that CA is a main competitor for carbon sources during the accumulation of PHB granules (42). Indeed, our results proved that blocking CA could enhance acetyl-CoA and cofactors, including NADP, NAD, and ATP. For flagella and fimbria, previous studies have indicated that the flagella could consume 2% of the energy of the entire cell (43). Our results also proved that blocking flagella could increase cofactors, including NADP, NAD, and ATP. However, the much less significant influence of these cofactors than of LPS further verified that disrupting LPS of the OM is the most efficient strategy, and the optimized cell performance by engineering LPS is interesting. Therefore, the OM-engineered mutant WJW08/pBHR68 with the additional block of 87 genes relevant to the formation of CA, flagella, and fimbria in WJW02 cells did not show obvious further improvement in PHB production.

Notably, cell sizes and cell envelope rigidity are critical for inclusion body accumulation. Many excellent studies have demonstrated that engineering strategies to enlarge cell sizes and weaken the cell wall could efficiently improve inclusion body accumulation (17, 18). Importantly, a recent study published in Nature proved that the stiffness and strength of E. coli cells are largely due to the OM, and LPS mostly contributes to the rigidity and stiffness of the OM (19). LPS engineering could decrease cell envelope rigidity in E. coli. Our studies also demonstrate that the cell wall combined with the OM of WJW02 cells is thinner than that of W3110 cells and more easily expanded. This difference is related to the downregulated Lpp, which is also due to the activated σ^E (35). The decreased rigidity and thickness of the cell envelope contribute to the enlargement of cell volume and to the accommodation of more PHB granules (15–17). The results demonstrate that blocking all formation of the polysaccharide portion of LPS is an efficient strategy to improve cell elasticity and enlarge cell volume when inclusion bodies accumulate.

Remarkably, engineering the components of the OM leads to better bacterial intracellular metabolisms for the biosynthesis of PHB, especially truncating the LPS molecule. Sufficient carbon flux to acetyl-CoA and the cofactor NADPH supply in our engineered mutants are also important for the accumulation of PHB, because many reports have suggested that the enhancement of acetyl-CoA levels could significantly improve inclusion body production (11, 12) and improving the levels of cofactor NADPH could also enhance the accumulation of PHAs (13, 14). To increase the acetyl-CoA biosynthetic efficiency, a series of metabolic engineering strategies have been developed, including expressing hydrogenase 3 and acetyl-CoA synthetase (11), as well as appropriately employing the EP-bifido pathway (12); to increase the cofactor NADH or NADPH supply, overexpressing glucose-6-phosphate dehydrogenase (14, 44), making glutamate an N source (14), and blocking the etf operon responsible for an electron transport pathway (13) could enhance PHA or PHB accumulation. In our study, remodeling the OM increased the acetyl-CoA and cofactor NADP supply, especially LPS engineering. Transcriptomics and metabolomics analyses both demonstrated that carbon metabolism flux to acetyl-CoA and the formation of cofactors were upregulated in WJW02 cells. And the improvements in acetyl-CoA and cofactors are more significant than that of mutant WJW00 with removal of RfaD (8). Moreover, the carbon flux directed to other sugar biosynthetic pathways was significantly downregulated. Notably, we believed that the effects of engineering LPS are not only due to the saving of nutrients and energy in WJW02 cells but also closely related to the significantly upregulated regulator fnr activated by LPS damage. Fnr is responsible for upregulating metabolic pathways under stressful environmental conditions (45). Coincidentally, LPS damage significantly upregulates fnr. As a result, LPS engineering could efficiently regulate cellular metabolism to enhance the flux to acetyl-CoA and cofactor supply.

The higher C:N ratio and limited N sources are also important for PHB accumulation (6, 46, 47). The C:N ratio in E. coli depends on glucose consumption and acetate production (45). For C sources, increasing acetyl-CoA and reducing by-products are beneficial for increasing the C:N ratio; for N sources, reducing the N source and converting intracellular available N to extracellular unavailable N could also contribute to a higher C:N ratio. Interestingly, engineering LPS could not only reduce by-product accumulation but also upregulate N source consumption and secretion to the extracellular space as the unavailable N form of GABA. Remarkably, these metabolic regulations are due to the activation of AR systems in WJW02 cells, and AR systems are indeed regulated by the regulator rpoE, which is also upregulated in WJW02, corresponding with the highest levels of σ^E being activated by Kdo₂-lipid A (37). Hence, to resist the danger caused by LPS damage to protect itself, AR systems are significantly upregulated in WJW02 cells, leading to stronger pH regulation. Surprisingly, this coincidental regulation is an efficient strategy to reduce by-products and increase the C:N ratio for PHB production in WJW02 cells.

MATERIALS AND METHODS

Construction of bacterial strains.

The bacterial strains used in this study are listed in Table 1, and the detailed deletions for different mutants are shown in Fig. S1 in the supplemental material. The plasmids and primers are listed in Tables S1 and S2, respectively.

TABLE 1.

Strains used in this study

Strain	Description	Source
W3110	Wild-type E. coli; F− λ−	ATCC 27325
W3110/pCas	Harboring pCas in W3110 for gene editing	This study
WJW02	Derived from W3110 by disrupting 25 genes (gmhD-waaQ, wbbL-rmlB) relevant to LPS polysaccharide biosynthesis	This study
WJW08	Derived from W3110 by disrupting 119 genes relevant to biosynthesis of LPS polysaccharide (25 genes; gmhD-waaQ, wbbL-rmlB), EPS (28 genes; galF-wza and yccC-ymcD), flagella (57 genes; flhE-motA, fliY-fliR, and flgN-flgL), and fimbria (9 genes; fimB-fimH)	This study
WJW09	Derived from W3110 by disrupting 21 genes (galF-wza) relevant to colanic acid and 7 genes (yccC-ymcD) for type IV EPS biosynthesis	This study
WJW010	Derived from W3110 by disrupting 57 genes (flhE-motA, fliY-fliR, and flgN-flg) relevant to flagellum biosynthesis	This study
WJW011	Derived from W3110 by disrupting 9 genes (fimB-fimH) relevant to type I fimbria biosynthesis	This study
W3110/pBHR68	Harboring plasmid pBHR68 in W3110	This study
WJW02/pBHR68	Harboring plasmid pBHR68 in WJW02	This study
WJW08/pBHR68	Harboring plasmid pBHR68 in WJW08	This study
WJW09/pBHR68	Harboring plasmid pBHR68 in WJW09	This study
WJW010/pBHR68	Harboring plasmid pBHR68 in WJW010	This study
WJW011/pBHR68	Harboring plasmid pBHR68 in WJW011	This study

WJW02 was constructed by site-specific recombination (48). The plasmid pDTW202-cat was constructed from pDTW202 (49) by replacing kan with cat. Briefly, the waaQ::FRTkan mutant allele was moved into strain WJW00/pKD46 (30) by transduction from W3110ΔwaaQ::FRTkan to construct strain WJW00ΔwaaQ::FRTkan. Then, Flp was expressed by pCP20 to delete the cluster (gmhD-waaQ) with FRTkan in WJW00ΔwaaQ::FRTkan by recombination of the longest FRT sites to form strain WJW01. The wbbL::lox L kan and rfbB::lox R cat mutant alleles were amplified from pDTW202 and pDTW202-cat, respectively, and transformed into WJW01/pKD46 to construct the strain WJW01ΔwbbL::lox L kanΔrfbB::lox R cat. Then, Cre was expressed by pKD46-cre (50) to delete the cluster (wbbL-rfbB), and pKD46-cre was removed by 42°C incubation. The LPS-engineered mutant was named WJW02.

For other gene clusters deletion, the CRISPR-Cas9 system (51) was used. For galF-wza, the plasmid pTargetF-galF-wza containing an N₂₀ sequence (20-base gene fragment complementary to the deletion target gene sequence) for targeting the galF-wza locus was extracted from an overnight culture using a TIANprep mini plasmid kit (Tiangen Biotech, Beijing, China). The editing template fragment with two homologous arms corresponding to the upstream and downstream regions of the galF-wza locus was obtained by overlap extension PCR. Then, pTargetF-galF-wza together with the template fragment was electroporated and cotransformed into W3110/pCas. One milliliter of cell culture concentrated by centrifugation was plated on LB agar containing kanamycin (50 mg/liter) and spectinomycin (50 mg/liter) for counterselection. The galF-wza knockout strain was finally obtained using colony PCR with the verified primers (galF-wza-U-F and galF-wza-D-R), and both pTargetF-galF-wza and the temperature-sensitive plasmid pCas were subsequently cured by the addition of 0.5 mM IPTG (isopropyl-β-d-thiogalactopyranoside) and incubation at 42°C with shaking. The cluster disruption for CA involved 21 genes, including galF to wza, that for flagella involved 57 genes distributed in 3 clusters, i.e., motA to flhE, flgN to flgL, and fliR to fliY, and that for type I fimbria involved 9 genes, fimB to fimH, resulting in WJW09, WJW010, and WJW011, respectively. The mutant WJW08 was constructed by further blocking CA, flagella, and type I fimbria in WJW02.

Growth medium and cultivation conditions.

All of the strains were grown at 37°C in Luria-Bertani (LB) medium, except for the strains containing plasmids pKD46, pCP20, pKD46-cre, and pCas, which were grown at 30°C. The pBHR68 harboring phaCAB is used for PHB production (5). For flask fermentation, the OM-engineered mutants and wild-type W3110 were all freshly transformed with pBHR68, and then the three correct recombinants for every strain were grown to mushroom lawn on fresh LB agar plates for 12 h, and the lawns were directly cultivated in LB medium for seed cultures. For PHB accumulation and the analysis of organic acids and amino acids produced by W3110 and WJW02, the seed cultures were incubated with 25 ml of LB medium in 100-ml flasks at 200 rpm. For W3110/pBHR68, WJW02/pBHR68, WJW08/pBHR68, WJW09/pBHR68, WJW010/pBHR68, and WJW011/pBHR68, 100 μg/ml ampicillin was added; after the seed culture OD₆₀₀ reached 1.8 to 2.0, the cultures were inoculated into M9 medium with 20 g/liter glucose. With an original OD of 0.25, the cultures were cultivated in 500-ml flasks at 200 rpm. M9 medium (pH 7.0) contains 17.1 g/liter Na₂HPO₄·12H₂O, 3 g/liter KH₂PO₄, 0.5 g/liter NaCl, 1 g/liter NH₄Cl, 1 mM MgSO₄, 0.1 mM CaCl₂, and 1% of vitamin B₁ (VB₁) solution (10 mg/ml).

LPS structure analysis.

The LPSs of W3110 and WJW02 were isolated according to the hot phenol-water extraction method (30). Briefly, the cell pellets were harvested from 1 ml of cultures (OD₆₀₀, 1.0) and then resuspended in 100 μl of Tris-acetate-EDTA (TAE) buffer and mixed with 200 μl of solution I, consisting of 3% SDS, 0.6% Tris, and 6.4% of 2 M NaOH, and heated at 100°C for 15 min. The LPSs in the mixture could be extracted with 250 μl of phenol-chloroform (1:1, vol/vol) by centrifugation at 12,000 rpm for 5 min. Two hundred microliters of supernatant was transferred to a new centrifuge tube, and the LPS sample was mixed with 200 μl of H₂O and 50 μl of sodium acetate (3 M and pH 5.2). Then, 500 μl of absolute ethanol was added and centrifuged at 12,000 rpm for 5 min. The precipitate was dissolved in 200 μl of solution II, containing 50 mmol/liter Tris-hydrochloride (pH 8.0) and 100 mmol/liter sodium acetate, and then mixed with 400 μl of absolute ethanol. Finally, LPS samples were collected after centrifugation at 12,000 rpm for 5 min and then dissolved in 50 μl of sterilized double-distilled water (ddH₂O). Five microliters of each LPS sample was separated on 15% SDS-polyacrylamide gels and visualized by silver staining, as described by Tsai and Frasch (52).

To further elucidate the LPS structure of WJW02, Kdo₂-lipid A was extracted from 400-ml LB cultures (OD, 1.0) using the modified Bligh-Dyer method according to our previous study (30, 53). Briefly, Kdo₂-lipid A was first extracted with a single-phase Bligh-Dyer mixture: the cells were first collected at 4,000 × g for 10 min, the cell pellets were resuspended in 16 ml of water, and then 20 ml of chloroform and 40 ml of methanol were added to form a 76-ml single-phase mixture (chloroform-methanol-water, 1:2:0.8, vol/vol/vol). The mixture was stirred for 1.0 h, and most Kdo₂-lipid A existed in the supernatant after centrifugation at 2,000 rpm for 30 min. The supernatant was added to 20 ml chloroform and 20 ml water to convert to a two-phase Bligh-Dyer system. After centrifugation at 2,000 × g for 30 min, the lower organic phase containing Kdo₂-lipid A was recovered and dried by rotary evaporation and stored at −20°C.

Kdo₂-lipid A extracted from WJW02 was analyzed by electrospray ionization-mass spectrometry (ESI/MS) and ESI-tandem mass spectrometry (ESI-MS/MS) (30). The mass spectra were acquired on a Waters Synapt Q-TOF mass spectrometer (Waters, Milford, MA, USA) equipped with an ESI source. Lipid samples were dissolved in a solvent of chloroform and methanol (2:1, vol/vol) and immediately infused into the ion source and scanned in the negative-ion mode at 0.2 μl/min. The negative-ion ESI/MS was carried out at −150 V, and the collisional activation of ions was performed at −6 V. The collisional activation of ESI-MS/MS was performed at −50 V. Data acquisition and analysis were performed using MassLynx V4.1 software (Waters Corp., Milford, Ma, USA).

CA detection and motility assays.

For CA detection, the quantitative detection method was used according to our previous study (22). Briefly, E. coli strains were grown in M9G liquid medium to the mid-log phase with a normalized number of cells and then heated at 100°C for 15 min to deactivate the enzymes and release CA. Next, the samples were cooled to room temperature and centrifuged at 12,000 × g for 10 min to discard the cells or cell debris. The supernatant was then added with 3 volumes of ice-cold anhydrous ethanol, allowed to rest without motion on ice, and incubated for 4 h. After the solution became clear, the sample was centrifuged at 10,000 × g for 15 min and dried to remove the ethanol. The dried samples were prepared for the determination of CA production. Each CA sample was dissolved fully in 1 ml deionized water, mixed with 5 ml sodium tetraborate-sulfuric acid solution (0.475 mg/liter), and then heated at 100°C for 5 min. After cooling, 100 ml hydroxyl diphenyl (1.5 g/liter) solution was added to the sample and mixed well. Finally, the absorbance was measured at 526 nm. The liquid medium without cells was used as the control, and different concentrations of glucuronic acid were used to make a standard curve. For the motility assay, 3-μl aliquots of LB-grown cultures with an OD₆₀₀ of 1.0 were dropped on LB supplemented with 0.3% agar plates and incubated for 48 h.

Quantitative RT-PCR analysis and transcriptome analysis.

Total RNA from strains W3110 and WJW02 was isolated using the Qiagen RNeasy total RNA kit. Reverse transcription was carried out with the RevertAid first-strand cDNA synthesis kit (Fermentas, Vilnius, Lithuania). Real-time PCR (RT-PCR) analysis was performed with SYBR Premix Ex TaqII (TaKaRa, Japan) in accordance with the protocol of the ABI Step One real-time PCR system (Applied Biosystem, Foster City, CA). All of the measurements were performed in triplicate. Whole-genome transcriptional analyses of E. coli strains W3110 and WJW02 were established according to a published method (23). Cultures of E. coli strains were grown at 37°C to the early exponential phase with an OD₆₀₀ of approximately 1.3. The libraries were sequenced using an Illumina HiSeq 2000 (BGI Shenzhen, China).

Metabolomics assay using GC-Q-TOF-MS and LC-QQQMS.

The E. coli cells were cultured in M9G to an OD₆₀₀ of 1.5, and then the same amounts of cells from six independent culture samples for both E. coli W3110 and WJW02 were collected at 4°C. The samples were immediately washed with phosphate-buffered saline (PBS) (pH 7.4, 4°C) twice and collected for liquid nitrogen freezing. The prepared samples were broken to prepare for gas chromatography (GC) or LC by three freeze-thaw cycles using liquid nitrogen. We carried out GC-time of flight mass spectrometry (GC-TOF/MS) experiments in our laboratory at Jiangnan University. The LC-QQQMS (liquid tandem chromatography triple quadrupole coupled to time of flight mass spectrometry) experiment was performed by Shanghai Biotree Biotech Co., Ltd.

For GC-quadrupole (Q)-TOF/MS, three steps, including derivatization, GC-TOF/MS analysis, and metabolic profiling analysis, were performed. Briefly, first, for 37°C normal-temperature derivatization, analytical samples were freeze dried. O-Methylhydroxylamine reagent and N-methyl-N-trimethylsilyltrifluoroacetamide (MSTFA) with 1% trimethylchlorosilane (TMCS) and 40 mg/ml methoxyamine hydrochloride (MeOX) were used to derivatize samples, which were incubated at 30°C for 90 min and at 37°C for 30 min. A fatty acid methyl ester (FAME) mixture (1 mg/ml; C₈-C₁₆, 1 mg/ml; C₁₈-C₃₀, 0.5 mg/ml in chloroform) was added to the quality control (QC) samples as an internal standard before derivatization. For GC-TOF/MS analysis, a Pegasus BT gas chromatography time-of-flight mass spectrometer (LECO, USA) was used with a DB-5 MS column (30 m by 250 μm inside diameter, 0.25-μm film thickness; Agilent, USA). Third, the raw data were converted twice with Shimadzu GC-MS Post Run software and ABF converter software. Files in “abf” format were analyzed using MS-DIAL software with the Fiehn library. The retention index (RI) tolerance was 5000. We deducted the blank sample peaks and all known artifact peaks (i.e., peaks caused by column bleeding) from the cell samples. We also removed the systematic error by random forest (SEFFR) to normalize the QC samples, correct the data, and obtain the final results.

For LC-QQQMS analysis, 1 ml of extraction solvent (precooled at −20°C, acetonitrile-methanol-water, 2:2:1) was added to the metabolite extraction sample, vortexed for 30 s, homogenized at 45 Hz for 4 min, and sonicated for 5 min in an ice-water bath. This was repeated 2 times, and the sample was incubated at −20°C for 1 h and centrifuged at 12,000 × g and 4°C for 15 min. Then, we transferred 825 μl of clear supernatant to a new EP tube and dried it under a gentle nitrogen flow. The residue was reconstituted with 100 μl of 50% acetonitrile/water and centrifuged at 12,000 × g and 4°C for 15 min. A 60-μl aliquot of the clear supernatant was transferred to an autosampler vial for UHPLC-MS/MS analysis. An equal aliquot (15 μl) from all of the samples was mixed to form the pooled sample for quality control, and a 60-μl aliquot was used for UHPLC-MS/MS analysis. UHPLC separation was carried out using an Agilent 1290 Infinity series UHPLC system (Agilent Technologies) equipped with a Waters Acquity UPLC BEH amide column (100 by 2.1 mm, 1.7 μm).

UPLC-MS analysis for intracellular acetyl-CoA and determination of intracellular levels of cofactors NADP, NAD, and ATP.

E. coli cells were fermented in minimal medium supplemented with 20 g/liter glucose (M9G) to an OD₆₀₀ of 1.5. The determination of acetyl-CoA was performed according to our previous study (8, 54). The data were analyzed and visualized with Xcalibur software.

For NADP and NAD levels, the cells were grown to early log phase (OD₆₀₀, 1.5) and mid-log phase (OD₆₀₀, 2.5), and for ATP, cells were grown to early log phase (OD₆₀₀, 1.5) and mid-log phase (OD₆₀₀, 2.5) and then collected and determined using an NADP⁺/NADPH assay kit with WST-8 (S0179; Beyotime), an NAD⁺/NADH assay kit with WST-8 (S0175; Beyotime), and an Enhanced ATP assay kit (S0027; Beyotime) according to our previous study (8).

TEM, ultra-TEM, and LSCM analysis.

According to our previous study (8), to analyze cell morphology and cell membrane morphology, different E. coli cells were subjected to TEM and ultra-TEM analysis. TEM and ultra-TEM experiments were entrusted to the Electron Microscopy Center of East China Normal University. The cells were negatively stained with 2.5% valeric acid phosphate, and specimens were imaged with a JEOL JEM 2100 (JEOL Ltd., Tokyo, Japan). For cell morphology analysis of E. coli W311, WJW02, WJW08, WJW09, WJW010, and WJW011, overnight (10-h) LB solid culture cell colonies were used. For analysis of PHB accumulation, 24-h culture cells were harvested by centrifugation at 10,000 × g for 1 min, washed three times with PBS at a pH of 7.4, and fixed with 2.5% glutaraldehyde for ultra-TEM analysis. For laser scanning confocal microscopy (LSCM), first, 1-ml cultures were harvested and washed twice with PBS (pH 7.4) and resuspended in100 μl PBS (pH 7.4). The cell resuspensions were added to 1 μmol Nile red dissolved in dimethyl sulfoxide (1 μg/μl). The mixtures were incubated in the dark at 37°C for 30 min after vigorous mixing. Finally, stained cells were washed three times with double-deionized water to remove any residual fluorescent dye prior to resuspension in PBS (pH 7.4), and 10 μl of the suspension was prepared on slides with a cover glass and nail polish for visualization by use of a fluorescence microscope (Leica TCS SP8; Leica, Germany). Cell excitation was accomplished using a 488-nm argon laser with emission at 530 nm. Under the conditions of excitation of 488 nm and emission of 530 nm, the PHB granules stained with Nile red could emit green fluorescence. Photographs were captured with Leica TCS SP8 software (CellSens Standard 1.9).

Analytical methods for PHB, organic acids, and amino acids.

For PHB analysis and fermentation product determination, according to our previous study (8), briefly, cells were harvested by centrifugation at 4,000 × g for 10 min. The cell dry weight was measured after vacuum lyophilization. PHB contents were analyzed by use of a GC-2010 plus gas chromatograph (Shimadzu, Japan) (55). Subsequently, the supernatant was filtered to analyze fermentation products. The concentrations of glucose were determined using an SBA-40C biosensor (Biology Institute of Shandong Academy of Sciences, Shandong, China). The fermentation product amino acids were determined by HPLC (Thermo Scientific, USA) equipped with an ion exchange column (Aminexs HPX-87H; Bio-Rad) and a refractive index detector (RI-150; Thermo Scientific, USA). Organic acid concentrations were analyzed by an HPLC instrument (GC112A; Shanghai Precision & Scientific Instrument Co., Ltd., China) equipped with a thermal conductivity detector with a flow rate of 10 μl/min. The temperatures of the injector, detector, and column were maintained at 120, 120, and 80°C, respectively.

Broken cell ratio, autoaggregation ability, and biofilm formation.

The autoaggregation ability and biofilm formation ability of cells grown in LB medium were determined according to our previous study (30). The autoaggregation ability was expressed as the autoaggregation percentage, and the value of [(A₀ − A_i)/A₀] × 100 represents the autoaggregation ability. The OD₆₀₀ of the original culture was recorded as A₀, and the OD₆₀₀ values at different times of cultures at 1 cm from the liquid surface were recorded as A_i. For analysis of the broken cell ratio, we determined the ultrasonic breaking ratio. Twenty milliliters of cells at an OD₆₀₀ of 1.0 was collected and resuspended in 5 ml of PBS (pH 7.4) to an OD₆₀₀ of A₀. The resuspensions were subjected to an ultrasonic disruptor (Scientz Co. Ltd., China). The OD₆₀₀ values at different times for ultrasonic samples were recorded as A_i, and the value (A₀ − A_i)/A₀ was represented as the ultrasonic broken cell ratio.

Data availability.

Raw transcriptome sequence data are available in the Sequence Read Archive (SRA; https://www.ncbi.nlm.nih.gov/sra) under accession number SRP160437.