All bacterial strains and plasmids used in this study were listed in Supplementary Table 1. E. coli K12 BW25113 (used as WT) and its derivatives (KEIO collection) were obtained from National BioResource Project (National Institute of Genetics) (Baba et al., 2006). Plasmid pCP20 was introduced into each mutant harboring FRT-Km-FRT cassette and the kanamycin (Km) cassette was eliminated using flippase/flippase recognition target (Flp/FRT) recombination. Plasmid pUC19 was introduced into non-marker mutants and the transformants were used for further experiments. E. coli cells were grown in lysogeny broth (LB) Miller (1% w/v tryptone, 0.5% w/v yeast extract and 1% w/v NaCl) at 37°C with shaking at 200 rpm or on solid LB agar plates. When required, antibiotics were added to the medium at 100 μg/mL ampicillin or 50 μg/mL kanamycin. 10% glycine was added to be a final concentration of 1%, when necessary.

Vesicle extraction from E. coli was performed as previously described with some modifications (Ojima et al., 2015). Precultures were inoculated into 100 mL of LB broth with or without 1% glycine at an initial OD₆₀₀ of 0.01, and they were grown with shaking for 16 h at 37°C. Bacterial cells were removed by centrifugation, and the supernatants were filtered through 0.45 μm and 0.2 μm-pore-size filters. Ammonium sulfate was added to the filtrates (final concentration of 400 g/L), and the samples were incubated at room temperature for 1 h. Vesicles were recovered from the suspensions by centrifugation at 9,500 × g for 30 min at 20°C. Pellets were resuspended in 15% (v/v) glycerol and concentrated by ultracentrifugation at 150,000 × g for 1 h at 4°C using an Himac CP80WX and a P50A3 rotor (Eppendorf Himac Technologies, Hitachinaka, Japan), and they were then resuspended in 50 mM HEPES (pH 6.8)-0.85% (w/v) NaCl (HEPES-NaCl buffer). The vesicles were stored at −80°C until further use.

For quantification of vesicles, extracted vesicles were incubated with 5 μg/mL of FM4-64 for 30 min on a 96-well black plate at 37°C in the dark. The fluorescence intensities of the fluorescently labeled samples were measured using a microplate reader (PerkinElmer, Waltham, MA, United States) at 506/750 nm (excitation/emission wavelength). Water-soluble linoleic acid was used as a standard. Vesicle formation is presented as the amount of vesicles normalized to the cell density.

Extracellular and vesicle-associated DNA was quantified using a Quant-iT PicoGreen assay (Thermo Fisher Scientific, Waltham, MA, United States). For quantification of extracellular DNA, cell cultures were centrifugated and the supernatants were filtered through 0.45 μm and 0.2 μm-pore-size filters to remove bacterial cells. If necessary, vesicle-free supernatants were prepared by the ultracentrifugation at 150,000 × g for 1 h at 4°C. DNA concentration of the supernatant was measured using the PicoGreen. A Tris-EDTA (TE) buffer (pH 8.0) was used to dilute the PicoGreen and lambda-DNA. A standard curve was constructed by serial dilution of DNA based on quantitation by the commercial provider and then incubated for 5 min at room temperature while protected from light. After incubation, the sample fluorescence was measured using a microplate reader (PerkinElmer) at 480 nm for excitation and 520 nm for emission.

For quantification of vesicle-associated DNA, extracted vesicles were treated with GES lysis reagent (5M guanidinum thiocyanate, 100 mM EDTA, 0.5% w/v sodium N-lauroylsarcosinate) for prior to labeling. DNA concentrations were measured by PicoGreen, and the values were normalized to the vesicle protein concentration. Vesicles were lysed with 1% (w/v) SDS and protein concentration of vesicles were determined using a Micro BCA Protein Assay Reagent Kit (Thermo Fisher Scientific).

Internal vesicle-associated DNA was obtained by the treatment of vesicles (20 μg protein) with 2 U DNase I (Fujifilm Wako Pure Chemical Co., Osaka, Japan) at 37°C for 30 min according to the manufacturer’s directions, followed by release of the internal DNA by lysis of vesicles with GES reagent. External and internal DNA associated with vesicles derived from E. coli grown with and without 1% glycine was quantified using PicoGreen.

Hydrodynamic diameters of vesicles were measured by a Zetasizer Nano ZS particle analyzer (Malvern Panalytical Ltd., Malvern, United Kingdom) in HEPES-NaCl buffer at 30°C. The hydrodynamic zeta average diameter was calculated using the dynamic light scattering (DLS) method. The concentration of vesicles was measured by nanoparticle tracking analysis (NTA) using a NanoSight LM10 instrument (Malvern) equipped with an sCMOS camera (Andor, Belfast, United Kingdom) as described previously (Tashiro et al., 2017). Briefly, the samples were diluted in HEPES-NaCl buffer, five replicate videos were collected from each sample, and particle movement was analyzed using NTA software (Version 3.1, Malvern). The velocity of particle movement was used to calculate the particle size according to the two-dimensional Stokes-Einstein equation.

The plasmid copy number was determined by real-time PCR analysis. For quantification of internal vesicle-associated plasmid, extracted vesicles were treated with DNase I at 37°C for 30 min to remove external DNA, DNase I was inactivated at 75°C for 15 min, and internal DNA was released from vesicles at 100°C for 10 min. DNA fragments were amplified and quantified using LightCycler FastStart DNA Master SYBR Green I (Roche, Basel, Switzerland) with specific primer pairs (M13-F/M13-R for pUC19 and dxs-F/dxs-R for chromosome listed in Supplementary Table 1) on a LightCycler 2.0 (Roche) according to the manufacturer’s instructions.

The permeability of bacterial membrane was analyzed by using BacLight Live/Dead bacterial viability staining kit (Thermo Fisher Scientific). Bacterial cells grown at the late exponential phase was centrifuged and washed twice with 0.85% NaCl. Cell suspension was treated with SYTO9 and propidium iodide (PI) following the manufacturer’s instruction. Cells were observed using the Olympus IX73 (Olympus, Tokyo, Japan) microscope, and images were captured with the charge-coupled-device (CCD) camera DP73 and processed by the imaging software cellSens.

The permeability of the outer membrane was analyzed by using fluorescent probe 1-N-phenylnaphthylamine (NPN) as previously described (Soh et al., 2020). Briefly, bacterial cells were centrifuged and washed twice with 5 mM HEPES buffer (pH 7.2) containing 1 mM sodium azide. Cell suspension was adjusted to OD₆₀₀ = 0.5 and kept at room temperature for 30 min. 200 μL of samples were applied to black microtiter plate and 4 μL of 500 mM NPN was mixed to achieve a final concentration of 10 μM. The NPN fluorescence was measured using a microplate reader (PerkinElmer) at 350 nm for excitation and 420 nm for emission.

For observation of negatively stained vesicles, samples were placed on Cu400 mesh grids (JEOL, Tokyo, Japan) that were pretreated with 0.01% α-poly-L-lysine. Bacterial cells and vesicles were stained with 2% (NH₄)₆Mo₇O₂₄ and observed using a JEM-1010 (JEOL) at 80 kV that was equipped with a FastScan-F214 (T) CCD camera (TVIPS, Gauting, Germany).

For quick freeze and replica electron microscopic (QFDE-EM) observations, the protocol was as described previously (Takaki et al., 2020). Briefly, bacterial cells were centrifuged, washed, mixed with a rabbit lung slab and mica flakes, and then placed on a paper disk attached to an aluminum disc. The samples were quickly frozen in liquid helium using a CryoPress (Valiant Instruments, St. Louis, MO, United States). The specimens were placed in a chamber maintained at −180°C using a JFDV freeze-etching device (JEOL). The samples were freeze-fractured with a knife and freeze-etched. Subsequently, the samples were coated with platinum and then coated with carbon. The replicas were floated in full-strength hydrofluoric acid, rinsed in water, cleaned with a commercial bleach containing sodium hypochlorite, and rinsed in water. Replica specimens were placed onto grids and observed using a JEM-1010 (JEOL).

The incorporation of plasmid DNA into OMVs has previously been reported in E. coli (Tran and Boedicker, 2017, 2019); however, the mechanism by which plasmid DNA moves from the cytoplasm to OMVs has not been fully elucidated. To gain further insight regarding this process, we focused our studies on examining the relationship between OMV biogenesis and the incorporation of plasmids into MVs. We used E. coli BW25113 WT and three high OMV-producing strains (ΔnlpI, ΔrseA, or ΔtolA) that possessed differing hypervesiculating mechanisms. The non-conjugative, non-mobilized and high copy number plasmid pUC19 was used in this experiment, as the copy number per vesicle was the highest among plasmids as previously tested by Tran and Boedicker (2017). These mutants harboring pUC19 exhibited higher OMV formation compared to that of the control strain WT/pUC19 (Figure 1). The average hydrodynamic diameter of OMVs derived from ΔnlpI was slightly less than that from WT, and those from ΔrseA and ΔtolA did not differ from that of WT (Supplementary Figure 1).

To investigate the association of DNA with OMVs, DNA concentration was investigated using the PicoGreen assay that detects double-stranded DNA. First, we measured the DNA concentration of the supernatant before and after ultracentrifugation to examine the extent of DNA associated with OMVs. Concentrations of extracellular DNA (eDNA) in ΔnlpI and ΔrseA were higher than those in WT, while the eDNA concentration was significantly decreased in the OMV-free supernatant (Figure 2A), suggesting that most of eDNA is associated with OMVs. Next, we investigated the DNA concentration associated with the extracted OMVs. As PicoGreen reagent is not permeable through the cellular membrane in the absence of additional treatments, OMVs were lysed with guanidium thiocyanide prior to labeling with PicoGreen. The DNA associated with OMVs derived from ΔnlpI was slightly increased compared to levels derived from WT (Figure 2B), thus indicating that the total of external and internal DNA associated with OMVs was highest in OMVs from ΔnlpI.

To analyze the plasmid DNA concentration in OMVs, extracted OMVs were treated with DNase I to remove externally associated DNA from OMVs, and the ratio of plasmid to vesicle was evaluated using real-time PCR and nano-tracking analysis. The amount of plasmid DNA in vesicles was increased threefold in ΔnlpI/pUC19 and was significantly decreased in ΔrseA/pUC19 and ΔtolA/pUC19 compared to that in WT/pUC19. By contrast, the relative plasmid copy number in bacterial cells was not significantly different among the samples (Supplementary Figure 2). These results suggest that defects in peptide crosslinks in PG can increase the incorporation of plasmids into OMVs in ΔnlpI even though the average size of these OMVs was smaller than that of the OMVs from the other strains.

To understand the relationship between defects in peptide crosslinks in PG and the association of DNA with OMVs, we investigated the impact of glycine on OMV formation. Glycine inhibits PG synthesis by substituting glycine for the D- or L-alanine of PG during growth in E. coli (Hammes et al., 1973), and the addition of glycine significantly enhances OMV formation in E. coli Nissle 1917 (Hirayama and Nakao, 2020). As described in a previous report, the addition of a final concentration of 1% (w/v) glycine enhanced OMV formation in E. coli BW25113 harboring pUC19 (Figure 3A). The presence of OMVs surrounding bacterial cells with 1% glycine was confirmed by TEM (Figure 3B), which are thought to be vesicles that did not leave the cell surface due to increased formation. The size distribution of OMVs was not significantly different between OMVs in the presence and absence of 1% glycine, although OMVs derived from ΔnlpI/pUC19 did exhibit smaller diameters (Figure 3C).

To investigate the association between glycine-induced PG deficiency and DNA secretion, we examined the DNA content within the supernatant and in OMVs. The amount of eDNA released per cell was quantified using the PicoGreen assay and according to cellular density, and the results revealed that eDNA release in the presence of 1% glycine was fourfold higher than that in the control (Figure 4A). To compare DNA content inside and outside of OMVs, DNA concentrations were measured in OMVs treated with or without DNase. The percentage of internal DNA relative to total DNA associated with OMVs was 13% in OMVs derived from E. coli in LB medium and 70% in OMVs with glycine (Figure 4B), thus suggesting that the addition of glycine enhances the amount of internal DNA in OMVs. Furthermore, real-time PCR analysis indicated that the ratio of plasmid to OMVs induced by glycine were higher than were those in the control (Figure 4C). These results support the idea that PG defects increase the incorporation of plasmid DNA into OMVs.

As glycine inhibits the proper alignment of PGs in E. coli (Hammes et al., 1973; Jonge et al., 1996), we evaluated the growth and membrane integrity of E. coli in the absence and presence of glycine. We analyzed membrane permeability using SYOT9 and PI, both of which are frequently used to test bacterial viability. The growth curve revealed that the presence of 1% glycine slightly repressed E. coli growth in LB medium (Supplementary Figure 3). At the late exponential phase (3 h culture), the percentage of PI-labeled cells was increased (Figures 5A,B); however, the colony forming unit (CFU) did not change significantly in the presence or absence of glycine (Figure 5C), suggesting that glycine increases the membrane permeability.

To further analyze the effect of glycine on membrane permeability, we used the fluorescent probe 1-N-phenylnaphthylamine (NPN) that is a small hydrophobic molecule (219 Da) and cannot cross the OM. When the membrane is damaged, the hydrophobic molecule NPN can enter into the phospholipid layer, resulting in prominent fluorescence (Muheim et al., 2017). At the late exponential phase, the membrane permeability of OM showed a high level (approximately threefold) in the presence of 1% glycine (Figure 5D). Thus, membrane remodeling already occurred at the late exponential phase, although the bacterial viability was not altered at that point.

O-IMVs have been suggested as a route for MV-mediated gene transfer in several bacterial species (Pérez-Cruz et al., 2013, 2015). To evaluate the relationship between DNA incorporation and the appearance of OMVs, OMVs derived from WT (with and without 1% glycine), and ΔnlpI were analyzed by negative staining using transmission electron microscopy (TEM). Most OMVs were single spherical vesicles in all samples; however, double lamellar vesicles were observed in OMVs of WT grown with glycine or from ΔnlpI (Figure 6 and Supplementary Figure 4). This suggested that these double lamellar vesicles are O-IMVs and may contain cytoplasmic components, including DNA.

To investigate the spatial structure of the bacterial envelope after exposure to glycine, we used quick-freeze deep-etch and replica electron microscopy (QFDE-EM) to observe bacterial cellular surface with high spatial and sub-millisecond time resolutions (Tulum et al., 2019; Takaki et al., 2020). When E. coli cells were grown without glycine, abnormally altered cellular surfaces were not observed in the freeze-fractured sections (Figures 7A,B). The QFDE-EM method enables visualization of the intracellular compartments such as the presumptive OM, IM, and cytoplasm (CP) (Figures 7B,C). The addition of glycine caused the separation of OM and IM and resulted in curvature of the OM (Figure 7D). Furthermore, a large amount of OMVs were observed on the cellular surface in the presence of glycine (Figure 7E). Interestingly, vesicles composed of double membranes were observed in the sample containing glycine (Figure 7F). The freeze-fractured section of the blebbing cells revealed that the multilamellar vesicles were O-IMVs and contained cytoplasmic components (Figures 7G,H), thus suggesting that O-IMVs are a possible route for incorporation of DNA into OMVs in the presence of glycine.