Multilamellar and Multivesicular Outer Membrane Vesicles Produced by a Buttiauxella agrestis tolB Mutant

Membrane vesicle (MV) formation has been recognized as a common mechanism in prokaryotes, and MVs play critical roles in intercellular interaction. However, a broad range of MV types and their multiple production processes make it difficult to gain a comprehensive understanding of MVs. In this work, using vesicle separation and electron microscopic analyses, we demonstrated that diverse types of outer membrane vesicles (OMVs) were released from an engineered strain, Buttiauxella agrestis JCM 1090^T ΔtolB mutant. We also discovered a previously undiscovered type of vesicle, multilamellar/multivesicular outer membrane vesicles (M-OMVs), which were released by this mutant using unconventional processes. These findings have facilitated considerable progress in understanding MV diversity and expanding the utility of MVs in biotechnological applications.

ABSTRACT

Outer membrane vesicles (OMVs) are naturally released from Gram-negative bacteria and play important roles in various biological functions. Released vesicles are not uniform in shape, size, or characteristics, and little is known about this diversity of OMVs. Here, we show that deletion of tolB, which encodes a part of the Tol-Pal system, leads to the production of multiple types of vesicles and increases overall vesicle production in the high-vesicle-forming Buttiauxella agrestis type strain JCM 1090. The ΔtolB mutant produced small OMVs and multilamellar/multivesicular OMVs (M-OMVs) as well as vesicles with a striking similarity to the wild type. M-OMVs, previously undescribed, contained triple-lamellar membrane vesicles and multiple vesicle-incorporating vesicles. Ultracentrifugation enabled the separation and purification of each type of OMV released from the ΔtolB mutant, and visualization by quick-freeze deep-etch and replica electron microscopy indicated that M-OMVs are composed of several lamellar membranes. Visualization of intracellular compartments of ΔtolB mutant cells showed that vesicles were accumulated in the broad periplasm, which is probably due to the low linkage between the outer and inner membranes attributed to the Tol-Pal defect. The outer membrane was invaginating inward by wrapping a vesicle, and the precursor of M-OMVs existed in the cell. Thus, we demonstrated a novel type of bacterial OMV and showed that unconventional processes enable the B. agrestis ΔtolB mutant to form unique vesicles.

IMPORTANCE Membrane vesicle (MV) formation has been recognized as a common mechanism in prokaryotes, and MVs play critical roles in intercellular interaction. However, a broad range of MV types and their multiple production processes make it difficult to gain a comprehensive understanding of MVs. In this work, using vesicle separation and electron microscopic analyses, we demonstrated that diverse types of outer membrane vesicles (OMVs) were released from an engineered strain, Buttiauxella agrestis JCM 1090^T ΔtolB mutant. We also discovered a previously undiscovered type of vesicle, multilamellar/multivesicular outer membrane vesicles (M-OMVs), which were released by this mutant using unconventional processes. These findings have facilitated considerable progress in understanding MV diversity and expanding the utility of MVs in biotechnological applications.

INTRODUCTION

Membrane vesicles (MVs) are naturally produced from bacteria and archaea, and those derived from Gram-negative bacteria are called outer membrane vesicles (OMVs). OMVs are liberated from the outer surface, as bilayer spheres between 20 and 400 nm in size, and contain proteins, phospholipids, lipopolysaccharide (LPS), and in some cases DNA (1, 2). They are multifunctional and involved in interkingdom communication, nutrient acquisition, predation, horizontal gene transfer, and the maintenance of biofilm structure (1). In addition, MVs derived from pathogenic bacteria contain toxic compounds, involved in transferring virulence factors to host cells and modulating host immune responses, enabling bacteria to survive in hostile environments (3, 4). Because of their therapeutic potential, MVs have been developed as biotechnological tools in vaccine and drug delivery systems (3, 5).

The components and characteristics of MVs could vary even among the same species, e.g., depending on growth stage (6), culture medium (7), and oxygen conditions (8) for Pseudomonas aeruginosa. The detailed differences in the MVs released under planktonic and biofilm conditions have been documented for several bacterial species, including P. aeruginosa (9–13), Helicobacter pylori (14), and Lactobacillus reuteri (15). MVs released from a strain are usually not homogenous, as they are varied in size and appearance as detected in electron microscopic images. Furthermore, outer-inner membrane vesicles (O-IMVs), which are double-membrane vesicles and OMVs containing inner membrane vesicles, constituted up to 0.1 to 2% of total OMVs released from many bacterial species (16, 17). However, it is difficult to separate a specific group of particles from their crude total MVs; thus, the characteristics of each MV are still largely unknown.

The precise mechanism of OMV biogenesis has yet to be defined and varies among bacterial species and culture conditions (4). In OMV biogenesis analyses using specific or random mutations in individual strains, several important factors and the defects of a range of proteins localized at the outer membrane (OM) were found to result in increased OMV production (18–24). One of the best-characterized factors is Tol-Pal, which is widely conserved across Gram-negative bacteria. The Tol-Pal complex is a cell-division component that aids in the invagination of the outer membrane and maintenance of membrane stability (25, 26). The first observed involvement of this factor in blebbing was that gene disruption within the tol-pal gene cluster results in hypervesiculation in Escherichia coli (18). In E. coli, the Tol-Pal system comprises five proteins, including three inner membrane proteins (TolA, TolQ, and TolR), periplasmic protein TolB, and outer membrane protein Pal (peptidoglycan [PG]-associated lipoprotein) (27). TolA's transmembrane domain interacts with the transmembrane domains of TolQ and TolR (28), the periplasmic domain of TolA interacts with Pal (29) and TolB (30), and TolB interacts with Pal (31). A defective Tol-Pal component causes the dissociation of outer membrane from the underlying peptidoglycan, resulting in membrane curvature and blebbing. Higher OMV production attributed to a deletion of any of the genes within tol-pal was reported in not only E. coli (18, 19) but also Pseudomonas putida (32), Salmonella enterica serovar Typhimurium (23, 33), Erwinia chrysanthemi (30), Caulobacter crescentus (26), Helicobacter pylori (34), Shigella boydii (35), and Shigella flexneri (36). Phenotypic changes caused by Tol-Pal defects are not limited to hypervesiculation but include decreased virulence, altered cell morphologies, and incompetence in cell division (25, 26, 30). OMVs derived from such hypervesiculating mutants have been effectively used as biotechnological applications in vaccine development to induce immune responses (35–38).

Previously, we examined MV production in our laboratory strains and found that Buttiauxella agrestis type strain JCM 1090 produces higher levels of MVs than E. coli MG1655 (39). MVs derived from this strain have a specific function of delivering MV contents to intraspecies bacterial cells in Buttiauxella spp. Buttiauxella strains have been isolated from unpolluted soil and drinking water, surface water, sewage, and soil and fecal samples (40). Here, we investigated the effect of TolB depletion on OMV formation in the high-OMV-producing strain B. agrestis JCM 1090^T to decipher its unique OMV biogenesis. We purified each type of OMV separately to examine their characteristics. A technology for visualizing images of frozen specimens with fracturing and etching processes to analyze OMV biogenesis was used, revealing that several unconventional processes are involved in the blebbing of multilamellar and multivesicular vesicles. Thus, this work demonstrates the formation of a novel type of OMV in the engineered bacterial strain and provides a model for unique OMV biogenesis.

RESULTS

Components of vesicles released by hypervesiculating B. agrestis.

To characterize the MVs released by B. agrestis JCM 1090^T, the protein composition of MVs was analyzed. Purified MVs were examined by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) assay, and major bands were identified by time of flight mass spectrometry (TOF-MS) or tandem mass spectrometry (MS/MS) analyses. Detailed information on the Mascot scoring is listed in Table 1. The proteins localized in the OM (OmpA and OmpC) and extracellular protein (FliC) were identified in MVs (Fig. 1A), consistent with prior studies using E. coli (41, 42), suggesting that vesicles released from this bacterium are also mainly OMVs.

TABLE 1.

Identified proteins in JCM 1090^T vesicles

Band no.	MWa (kDa)	NCBI sequence no.	Analysis method	Mascot score	Protein name
1	54.3	WP_034496216.1	PMF	177	Flagellar filament structural protein
2	42.4	WP_034500311.1	MS/MS	214	Elongation factor Tu, partial
3	46.0	WP_034497136.1	MS/MS	260	Translocational protein TolB
4	40.3	WP_034494098.1	MS/MS	706	Membrane protein OmpC
5	38.3	KFC82962.1	MS/MS	112	Outer membrane protein A
6	28.1	SUW62904.1	MS/MS	278	Tol-Pal system protein YbgF
7	18.9	WP_034492967.1	MS/MS	166	Divisome-associated lipoprotein YraP
8	18.0	WP_034494237.1	MS/MS	384	Outer membrane protein X

^aMW, molecular weight.

FIG 1.

Effect of components on vesicle production in B. agrestis. (A) Purified outer membrane vesicles (OMVs) were separated by SDS-PAGE using a 14% acrylamide gel and stained with Coomassie brilliant blue. Identified main bands are shown on the right. (B) OMV formation in the wild type (WT) and mutants. The amount of vesicles extracted from the supernatants was normalized to cell density, and each value shown is relative to that of WT. The data are shown as the mean ± standard deviation from three replicates. *, P < 0.05; **, P < 0.005 compared to WT. (C) Growth curves of WT and ΔtolB mutant. Data indicate the mean ± standard deviation from three replicates. (D) Organization of the tol-pal gene clusters in B. agrestis JCM 1090^T and E. coli MG1655. Numbers represent the nucleotide lengths.

Deletion of tolB increases OMV formation in B. agrestis.

To further investigate the effect of these proteins on OMV production, we constructed ΔompA, ΔompC, ΔfliC, ΔtolB, and ΔybgF mutants and examined their OMV production. All mutants, other than ΔfliC, showed higher OMV production than the wild type (WT) (Fig. 1B). Particularly, OMV production in ΔtolB was approximately 17-fold higher than that of WT (Fig. 1B) despite slower growth (Fig. 1C). The growth rate of other mutants was not different from that of WT (data not shown). Thus, the disruption of tolB causes hyper-vesicle production in B. agrestis. B. agrestis has homologs of all of the tol-pal genes in E. coli (Fig. 1D), and its TolB is 93.5% identical to that of E. coli at the amino acid level, suggesting similar function.

It has been known that the electronegative charge of the bacterial surface causes charge-to-charge repulsion, resulting in outward membrane blebbing (43). Zeta potential was not significantly different between WT (−4.6 mV ± 0.51 mV) and the ΔtolB mutant (−6.4 mV ± 0.20 mV), suggesting that the increased OMV formation is not likely due to zeta potential in the B. agrestis ΔtolB mutant.

Morphological changes in the ΔtolB strain.

WT and ΔtolB mutant cells were observed by phase contrast and fluorescence microscopy with SYBR green staining. B. agrestis WT was predominantly rod-shaped; the ΔtolB strain showed multiple shapes, including rod-shaped to streptococcus-like forms (Fig. 2A). Negative staining and transmission electron microscopy (TEM) analysis confirmed that both shapes were observed in ΔtolB mutant cells (Fig. 2B), consistent with previous reports (26, 30, 34, 44). Small protrusions from the outer membrane were observed in ΔtolB mutant cells (Fig. 2C) but were absent in WT cells, suggesting that these protuberances contribute to the high OMV production in the ΔtolB strain.

FIG 2.

Cellular morphologies of B. agrestis WT and ΔtolB strain. (A) Cell morphologies of each strain labeled with SYBR green were observed by phase contrast (upper panels) and fluorescence (bottom panels) microscopy. Scale bar = 10 μm. (B) Transmission electron microscopy (TEM) micrographs of negative-stained cells of WT (upper) and ΔtolB mutant (middle and bottom). Scale bar = 500 nm. (C) TEM micrographs of ΔtolB cell blebbing small vesicles, corresponding to red dotted-line boxes in panel B. Scale bar = 100 nm.

Multiple types of vesicles are formed in the B. agrestis tolB mutant.

To characterize OMVs derived from the hyper-vesicle-producing B. agrestis ΔtolB mutant, crude OMVs were extracted from the supernatant and examined by TEM. OMVs are generally released from B. agrestis WT (termed WT-OMVs) and are spherical in shape with a bilayer membrane between 20 and 150 nm in diameter (Fig. 3A); O-IMV-like double-lamellar vesicles were also observed in crude WT-OMVs (Fig. 3B), consistent with previous reports (16, 17). In contrast, more types of OMVs were observed in the ultracentrifuged pellets of the ΔtolB mutant supernatant (Fig. 3C), and some vesicles were multilamellar or multivesicular with diverse shapes. Single bilayer vesicles were between 20 and 150 nm in length, while multilamellar/multivesicular vesicles were mostly 400 nm. Not only double-lamellar membrane structures (Fig. 3D) but also triple-lamellar structures, which appeared as “onion-like vesicles” (Fig. 3E), were observed in the crude vesicles of the ΔtolB strain. Pear-shaped vesicles were also present, some with triple membranes (Fig. 3F), while others contained two individual vesicles (Fig. 3G). Triple-membrane vesicles were relatively large compared to single-membrane vesicles (Fig. 3C). TEM images showed that the proportion of large-sized OMVs (>10,000 nm²) was not significantly different between WT-OMVs and ΔtolB mutant OMVs by number and area (Fig. 3H to K) due to the higher production of small vesicles (<2,000 nm²) in the ΔtolB strain (Fig. 3J and K). All vesicles larger than 10,000 nm² were classified as multilamellar or multivesicular OMVs (Fig. 3C). Thus, the B. agrestis ΔtolB strain produces small vesicles and a new type of multilamellar/multivesicular OMV (M-OMV).

FIG 3.

Various types of vesicles released from ΔtolB strain. (A to G) TEM images of crude vesicles extracted from B. agrestis WT (A, B) and ΔtolB mutant (C to G). Most of the vesicles were composed of a bilayer membrane (A), whereas some were O-IMV-like double-lamellar vesicles (B) in WT-OMVs. (C to G) Various types of vesicles were observed in ΔtolB strain OMVs. Red arrowheads show membranes in double- or triple-lamellar OMVs (multilamellar OMVs) (B, F). Yellow arrowheads show two independent vesicles in one vesicle (multivesicular OMVs) (G). Not only spherical multilamellar vesicles (D, E) but also pear-shaped vesicles were formed by the ΔtolB mutant (F, G). Scale bars = 100 nm. (H to K) Size distribution of outer membrane vesicles. OMV area was measured from TEM images using ImageJ software. The number (H, J) and area (I, K) of vesicles derived from the WT (H, I) and ΔtolB (J, K) strains were measured. Values shown represent fractions of 500 nm², and the right-most bars indicate vesicles larger than 10,000 nm². OMVs of less than 500 nm² were not detected. Data shown represent mean ± standard deviation from three independent images of WT-OMVs (n = 134, 141, and 156) and ΔtolB strain OMVs (n = 504, 929, and 835).

Gradient ultracentrifugation sorts each type of vesicle.

Ultracentrifugation of the filtered supernatant of the ΔtolB mutant was separated into two layers; the upper layer was a transparent semiliquid, while the bottom was a white solid pellet (Fig. 4A and B). The two layers had a different appearance from the ultracentrifuged pellet of the filtered supernatant of WT, suggesting that different OMVs formed. The translucent layers were subjected to iodixanol density gradient ultracentrifugation, and targeted OMVs with densities between 1.15 and 1.20 g/ml were obtained from the fractions. The presence of OMVs was confirmed by TEM (Fig. 4C), and the average hydrodynamic diameter of the OMVs from translucent bands (t-bands) was smaller than that of WT-OMVs (Fig. 4D). A significant correlation between density and size was observed in OMVs derived from the t-band (Fig. 4E), suggesting that membrane abundance affects the weight of these OMVs and the proportion of interior components is extremely low. The fraction with the white solid pellets yields two separate bands containing putative vesicles (Fig. 4A and B). TEM analysis confirmed the presence of OMVs (Fig. 4C); OMVs with similar appearance to WT-OMVs were present in the upper band (u-band), while M-OMVs were present in the lower band (m-band). The average hydrodynamic diameter of the OMVs contained in the m-band was slightly larger than that of the WT-OMVs (Fig. 4D). The OMVs present in the m-band were further analyzed by the quick-freeze deep etching and replica electron microscopy (QFDE-EM) method. M-OMVs were present in the m-band (Fig. 5A to D). The image of OMVs from freeze fracture, without the step of etching, confirms the multilamellar components of OMVs present in the m-band (Fig. 5E). Taken together, the gradient ultracentrifugation method has enabled the sorting of vesicle types and concludes that the B. agrestis ΔtolB strain produces various types of OMVs.

FIG 4.

Separation of several types of vesicles. (A) OMV isolation procedures. Supernatants from WT and ΔtolB strains were filtered and ultracentrifuged. In the ΔtolB mutant sample, the pellets were separated into a translucent gelatinous layer and white solid pellets. Each fraction was subjected to iodixanol density gradient ultracentrifugation. (B) Images of crude OMV suspension (a, pellets of WT sample; b, translucent gelatinous layer from ΔtolB mutant sample; c, white solid pellets from ΔtolB mutant sample). (C) TEM micrographs of the purified OMVs. OMVs purified from WT, translucent gelatinous layer from ΔtolB strain (t-band), upper band of density gradient centrifuge from white solid pellets (u-band), and the lower band containing multilamellar/multivesicular membrane vesicles (m-band) are shown. (D) Hydrodynamic diameter of the OMVs. Data are the average of three independent replicates for the dynamic light scattering analysis. The error bars represent the standard deviations. (E) Correlation between particle size and OMV t-band density. The translucent layer of the ultracentrifuged pellets from the ΔtolB strain supernatant was divided into 20 fractions using gradient ultracentrifugation (fraction 1 being the lightest and fraction 20 the heaviest). OMVs were contained between fraction 4 and fraction 12. Particle size of the OMVs from each fraction was analyzed using dynamic light scattering. Data labels indicate the fraction number.

FIG 5.

Quick-freeze replica electron micrographs of multilamellar membrane vesicles. The replica membrane was coated on a frozen sample (m-band OMVs) treated with (A to D) or without (E) deep etching. Magnified images (B to D) of the areas in white squares on panel A are provided. Scale bars = 100 nm.

The main components of vesicles are OMs.

As it has been reported that O-IMVs are formed in a variety of bacterial species (16, 17), it is theorized that M-OMVs also contain inner membrane and cytoplasm (CP) components. To investigate that possibility, the protein components of OMVs purified from each band were analyzed by SDS-PAGE and silver staining. Significant differences were not observed for each band pattern (Fig. 6A), suggesting that the main components of OMVs derived from the ΔtolB strain are similar to those of WT-OMVs.

FIG 6.

Components of vesicles. (A) Protein profiles of purified vesicles. Each purified OMV was evaluated by SDS-PAGE using a 10% acrylamide gel, and protein bands were detected by silver staining. Levels of KDO (B) and β-NADH (C), which are the biological markers for outer membrane and inner membrane, respectively. Data represent mean ± standard deviation from three replicates.

To further investigate components, 2-keto-3-deoxyoctonate (KDO), β-NADH, and β-galactosidase of each type of OMV were examined as the markers for OM, inner membrane (IM), and cytoplasm (CP) components, respectively. All OMVs derived from the ΔtolB strain showed higher KDO concentrations than WT-OMVs (Fig. 6B). Particularly, that of t-band OMVs was 3-fold higher than that of WT-OMVs, suggesting that the ratio of LPS to protein is high in t-band OMVs. The NADH activity of the t-band OMVs was slightly lower than that of the other OMVs (Fig. 6C). Although the control ΔtolB mutant cells expressed around 2.2 U/g of protein (data not shown), those of OMVs exhibited decreased activity at less than 1.0 U/g of protein (Fig. 6C), indicating very low IM content in all types of OMVs. Furthermore, β-galactosidase activity was not detected in any of the OMV samples, although the control ΔtolB mutant cells showed 6 U/g of protein (data not shown). Thus, the proportion of LPS to protein in ΔtolB strain OMVs was higher than that of WT-OMVs, while the IM and CP contents were low in all types of OMVs.

Membrane vesicles accumulate in the periplasm.

The QFDE-EM method has features of much higher spatial and submillisecond time resolutions (45), and this method is a powerful tool for investigating the spatial structure of bacterial envelopment (46). This method was applied to analyze OMV biogenesis. Using the freeze-fractured section of B. agrestis WT cells (Fig. 7A), the intracellular compartments were clearly visualized as follows: presumptive OM, peptidoglycan (PG), IM, and CP. Many possible precursors of the extracellular multivesicular vesicles were observed within the ΔtolB mutant (Fig. 7B to F). One image of ΔtolB mutant cells shows both spherical and random shapes of vesicles accumulated in the periplasm (Fig. 7B). Another shows double-lamellar vesicles in the periplasm (Fig. 7C). An OM was found invaginating inward and enveloping a vesicle, probably originating from the extracellular milieu or its own outer membrane (Fig. 7D). In addition, loose and curved OM portions were observed in the ΔtolB strain (Fig. 7D), attributed to the decreased stability in the OM-PG-IM cross-link due to the depletion of TolB. Complicated membrane structures in the periplasm were observed in many ΔtolB mutant cells (Fig. 7E and F). Invagination of the OM was observed across the lateral cell surface (Fig. 7C and E) rather than the cell pole. Blebbing was observed at the septa of dividing cells (Fig. 7G) and the surface of nondividing cells, particularly at the cell pole (Fig. 7H), suggesting that OMVs were formed at both sites. Taken together, the results show that vesicles accumulate in the periplasmic space in ΔtolB mutant cells and are released into the extracellular milieu as M-OMVs.

FIG 7.

Quick-freeze deep-etch and replica electron micrographs of B. agrestis cells. (A) WT cell image. Outer membrane (OM), peptidoglycan (PG), inner membrane (IM), and cytoplasm (CP) are shown. Scale bar = 500 nm. (B to H) ΔtolB strain cell images. Large numbers of vesicles were accumulated in a cell (B) and double-lamellar membrane vesicles were localized in the periplasm (C). Panel D shows the colored image of panel C. The yellow arrowhead indicates the outer membrane invaginating inward and enveloping a vesicle. The black arrowheads indicate the loose and curved portions of the outer membrane (OM). (E) Another image of intracellular vesicles is shown. (F) Enlarged images of red squares in panel E. Vesicle is released at the septa of dividing cells (G) and the cell pole (H). Red arrowheads indicate blebbing from the outer membrane. Scale bars = 500 nm.

DISCUSSION

Increased OMV production due to the Tol-Pal defect has been reported in E. coli and other bacteria (18, 19, 23, 30, 32–36). It has been proposed that depletion of one of the Tol-Pal components causes instability of the OM-PG-IM cross-link in the periplasmic space, resulting in hypervesiculation (23). Here, we aimed to analyze the characteristics and biogenesis of OMVs in the high-OMV-producing B. agrestis ΔtolB strain. High OMV production and OM protrusions were observed in the mutants, further suggesting the release of unconventional M-OMVs.

Size, form, and components of released MVs are not uniform even within one bacterial cell; i.e., small OMVs are released from the cellular surface, while relatively large OMVs are formed at the division septa in S. enterica serovar Typhimurium (23). We managed to separate different components of OMVs from the supernatant of the ΔtolB mutant by conventional vesicle purification methods using ultracentrifugation and density gradient fractionation, which yielded significant results in terms of density and configuration for three types of OMVs (t-band, u-band, and m-band).

One of the unconventional OMVs released from the B. agrestis ΔtolB mutant is the M-OMV. M-OMVs are large and multilamellar/multivesicular vesicles. Double-bilayer membrane vesicles are naturally released by Gram-negative environmental and pathogenic bacteria (16, 17) and contain IM and CP components, as well as DNA, implicating the involvement of OMVs in horizontal gene transfer. The percentage of O-IMVs with respect to total MVs was 0.2 to 2% in pathogenic bacteria including P. aeruginosa, Acinetobacter baumannii, and Neisseria gonorrhoeae. From our results, M-OMV-containing large vesicles (>10,000 nm²) were 1.2% by number and 11.4% by two-dimensional area (Fig. 3J and K), suggesting that the ratio of M-OMVs is higher by volume. M-OMVs observed in this study can be classified as a novel class of OMV, which can be differentiated from previously reported O-IMVs for the following reasons. First, the size and configuration of M-OMVs were not similar to those of O-IMVs. Typical O-IMVs are spherical with diameters between 50 and 160 nm (17), while large spheres and pear-shaped M-OMVs contained more triple-lamellar or more than two individual vesicles. Obviously, we cannot deny the possibility that M-OMVs with diameters greater than 200 nm were removed or the sphere-shaped OMVs were altered by the 0.2-μm filtration step. Second, the levels of IM and CP components of OMVs in the m-band were low, similar to those in the WT-OMVs. Third, unlike in O-IMV biogenesis (16), vesicles were accumulated in the periplasmic space in ΔtolB mutant cells.

The blebbing sites seem to be diverse in tol-pal mutants, as observed at the cell pole (19, 26, 33), dividing sites (23, 26, 32), and lateral cell surface (18, 19, 26, 32). In the B. agrestis ΔtolB mutant, larger sizes of OMVs were formed at the cell pole and dividing sites (Fig. 7G and H), while relatively small blebs were observed at the lateral cell surface (Fig. 2C), consistent with previous reports (23, 26).

The OM was abnormally separated from the IM in the ΔtolB mutant (Fig. 7C and D), and such defects in OM-PG-IM integrity would contribute to the accumulation of vesicles in the periplasm (Fig. 7B), as in the case of Caulobacter (26). A model for the underlying mechanism facilitating the release of these multivesicular OMVs is proposed in Fig. 8A. Although OMVs from the m-band were mainly composed of OM, some portions of intracellular vesicles that accumulated in the periplasm might be made of IM. As this strain produces a higher level of OMVs, undefined OMV biogenesis of this bacterium could be attributed to the unconventional multivesicular OMV formation.

FIG 8.

Proposed models for multilamellar/multivesicular outer membrane vesicle formation in the B. agrestis ΔtolB mutant. (A) Model for multivesicular OMV formation. A few vesicles are accumulated, using an unknown mechanism, in the broadened periplasm region due to low linking between outer and inner membranes in the ΔtolB mutant. Intracellular vesicles are made of OM or IM. Curved outer membrane contains one or more intracellular vesicles, and multivesicular OMVs are released. (B) Model for multilamellar OMV formation. The outer membrane is invaginated as an inward-trapping vesicle, although it remains unknown whether trapped vesicles are derived from the same cell or the extracellular milieu. Double-lamellar membrane vesicles accumulate in the periplasm region, and then triple-lamellar membrane vesicles are released. One of the blebbing sites is the septum of dividing cells.

OM was invaginated on the inward side, entrapping another vesicle in the ΔtolB mutant (Fig. 7C and D). The Gram-negative bacteria possess a negatively charged outer surface because of LPS, and the repulsion between adjacent LPS causes membrane curvature on the outward side (47). Conversely, the zeta potential of B. agrestis JCM 1090^T is almost neutral (more than −5 mV), which is much higher than that of other bacterial strains tested (39). Here, the deletion of tolB did not significantly alter the surface charge. Furthermore, OMVs derived from this bacterium effectively interacted with the Buttiauxella strains (39). Taken together, extracellular OMVs interact with the outer surface of ΔtolB mutant cells and OM, without fixed PG or IM, and are integrated into the periplasm. Such endocytosis-like M-OMV formation has only been observed in the B. agrestis ΔtolB strain. A model for such multilamellar OMV production is proposed in Fig. 8B. The inward invagination of OM was observed at the lateral cell surface, particularly in the middle part of cells.

Pear-shaped vesicles were observed in both multilamellar and multivesicular OMVs (Fig. 3C, F, and G). While these unusual bacterial MVs have not been reported before and understanding their novel shape will require some further analysis, we propose one mechanism by which these vesicles may be created. While Gaussian curvature is positive in conventional spherical vesicles, positive and negative Gaussian curvatures are present in pear-shaped OMVs. It is likely that surface tension and increased bending elasticity of the outer surface of these vesicles is maintained by heterogeneous surface chemistry. One proposed mechanism underlying the production of these pear-shaped vesicles relies on the idea that some proteins localized to the blebbing sites may be tethered to pieces of the cell wall during OM curvature, resulting in the production of these awkwardly shaped OMVs. We were not able to uncover the exact molecular mechanism underlying these shapes, but we hope to analyze and understand these pathways in the future.

OMVs present in the t-band ranged from 50 to 90 nm in hydrodynamic diameter (Fig. 4D). Such small sizes of OMVs are common in the supernatant of Gram-negative bacteria, but a gelatinous OMV layer after ultracentrifugation has not been observed even in the E. coli ΔtolB mutant. Small protuberances present in the B. agrestis ΔtolB mutant (Fig. 2C), but not in the WT, suggest that these protrusions contribute to the formation of small OMVs. OMVs present in the t-band had a negative correlation between the density and hydrodynamic diameter (Fig. 4E), indicating low levels of contents other than membranes in those OMVs. Therefore, the ratio of LPS to protein in t-band OMVs was higher than that in other types of OMVs (Fig. 6B). The Tol-Pal complex is important for transferring phospholipids from OM to IM, resulting in OM lipid homeostasis in E. coli (48). Likewise, accumulation of phosphatidylglycerols in OM was observed in S. enterica with Tol-Pal defects (44). Moreover, the VacJ/YrbB defect, which is also conserved in many Gram-negative bacteria, increases OMV formation in Haemophilus influenzae and Vibrio cholerae (24). Thus, the accumulation of phospholipids on the OM caused by defects in phospholipid transport from OM to IM might increase the production of small OMVs in the B. agrestis ΔtolB mutant.

In conclusion, the deletion of tolB in B. agrestis JCM 1090^T causes increased production of various types of OMVs, including types that are as yet undescribed. The ΔtolB mutant produces relatively small OMVs, triple-lamellar vesicles, and multiple vesicle-incorporating vesicles besides the general type of OMV. We propose the name multilamellar/multivesicular outer membrane vesicles (M-OMVs) for these novel types of OMVs. Furthermore, it is pertinent to investigate if bacteria from the natural environment and other engineered strains are capable of producing M-OMVs. As the content of these M-OMVs can be considered stable, they may be useful in the development of novel biotechnological applications.

MATERIALS AND METHODS

Microbial strains and plasmids.

The microbial strains and plasmids used in this study are listed in Table 2. B. agrestis was grown in tryptic soy broth (TSB) medium at 30°C. For genetic manipulations, LB (Luria-Bertani Miller) was used for E. coli culture. When necessary, streptomycin was used at a concentration of 50 μg/ml for E. coli and B. agrestis. DAPA (2,6-diaminopimelic acid) was used at a concentration of 300 μM in LB medium for the growth of E. coli β2163.

TABLE 2.

Strains and plasmids used in this study

Strain or plasmid	Genotype	Reference or source
Escherichia coli
JM109	recA1 supE44 endA1 hsdR17 gyrA96 relA1 thi Δ(lac-proAB) F' (traD36 proAB+ lacIq lacZΔM15)	Laboratory stock
β2163	(F−) PR4-2-Tc::Mu ΔdapA::(erm-pir) [Kmr Emr]	52
Buttiauxella agrestis
JCM 1090T	Wild type (CUETM77167)	BRC-JCM
ΔtolB mutant	JCM 1090T ΔtolB mutant	This study
ΔompA mutant	JCM 1090T ΔompA mutant	This study
ΔompC mutant	JCM 1090T ΔompC mutant	This study
ΔfliC mutant	JCM 1090T ΔfliC mutant	This study
ΔybgF mutant	JCM 1090T ΔybgF mutant	This study
Plasmid
pKNG101	Cloning vector; Smr, mob, sacB	53
pKNG101-ΔtolB	tolB deletion cassette in pKNG101	This study
pKNG101-ΔompA	ompA deletion cassette in pKNG101	This study
pKNG101-ΔompC	ompC deletion cassette in pKNG101	This study
pKNG101-ΔfliC	fliC deletion cassette in pKNG101	This study
pKNG101-ΔybgF	ybgF deletion cassette in pKNG101	This study

Construction of mutants.

The deletion of the tolB, ompA, ompC, fliC, or ybgF gene of B. agrestis was achieved by conjugation and homologous recombination using the suicide vector pKNG101. Fragments flanking the open reading frame were amplified with oligonucleotide pairs Ba_tolB_5F_HindIII/Ba_tolB_5R and Ba_tolB_3F/Ba_tolB_3R_BamHI, listed in Table 3. These fragments were used to perform overlapped extension PCR, creating a DNA fragment lacking the tolB open reading frame. The resulting DNA fragments were digested with HindIII and BamHI and ligated into compatibly digested pKNG101 to yield pKNG101-ΔtolB. Similarly, other pKNG101 based vectors were constructed using primers listed in Table 3 and restriction enzymes BamHI/XbaI. E. coli β2163 harboring this plasmid was used as a donor for conjugation, and marker-exchange mutagenesis was carried out. Gene deletion was confirmed by PCR and sequence analysis.

TABLE 3.

Primers used in this study

Primer name	Sequence (5′ to 3′)a
Ba_tolB_5F_HindIII	CCCAAGCTTAAAGCTGCCGATGCTGAGAAG
Ba_tolB_5R	GCAGGGAATGCGTGCAGCACGGCTGCCCACAG
Ba_tolB_3F	TGCACGCATTCCCTGCCTGGTCGCCGTATCTG
Ba_tolB_3R_BamHI	CGGGATCCCCTGGATTTGACCACGAAGGG
Ba_ompA_5F_BamHI	CGGGATCCCTTTGACTCAGTCATCAGCAGG
Ba_ompA_5R	ACGTCTTTTTGCGCCTCATTATCATCCAAAAAGGC
Ba_ompA_3F	AGGCGCAAAAAGACGTTGTAACTCAGCCTGCG
Ba_ompA_3R_XbaI	GCTCTAGAAGAGCACGCCATTCAAGTCG
Ba_ompC_5F_BamHI	CGGGATCCTATCAGGCACTGGCGTCATCG
Ba_ompC_5R	CCAACACCTGGAACCAGGAGGGACAGTAC
Ba_ompC_3F	CTGGTTCCAGGTGTTGGTCTGGTTTACCAG
Ba_ompC_3R_XbaI	GCTCTAGAACCGACGGAAACCAGATAACG
Ba_fliC_5F_BamHI	CGGGATCCTGGCAGAGATACCACTATCAGC
Ba_fliC_5R	CTCCCTGCGGTTGGGTCAACAGCGACAGGC
Ba_fliC_3F	TGACCCAACCGCAGGGAGTGCTGTCTCTG
Ba_fliC_3R_XbaI	GCTCTAGACTGGTAATAAAGGGTCAGCACC
Ba_ybgF_5F_BamHI	CGGGATCCGTCAACGTTCCTGGATGAAACG
Ba_ybgF_5R	TCGTCCTGCGCTATGCCAACCAGTAACGAC
Ba_ybgF_3F	GGCATAGCGCAGGACGAAGCCATTGCTGC
Ba_ybgF_3R_XbaI	GCTCTAGAGGCGTGCAGGATTCGAACC

^aUnderlining indicates restriction enzyme sites.

Quantification of vesicle production.

B. agrestis was grown in 5 ml TSB medium in each test tube and shaken at 200 rpm at 30°C for 16 h. Bacterial culture was centrifuged, and the supernatant was filtered through a 0.20-μm membrane and ultracentrifuged at 200,000 × g for 1 h at 4°C using Himac CP80WX (Eppendorf Himac Technologies) with a P50A3 angle rotor. The pellet was suspended in 50 mM HEPES (pH 6.8)/0.85% NaCl (HEPES-NaCl buffer). Vesicles were stained with lipophilic styryl dye FM4-64 at 5 μg/ml at 37°C for 30 min, and the fluorescence was measured using a microplate reader (PerkinElmer). Vesicle formation was shown as the amount of vesicles normalized to cell density.

Isolation and purification of vesicles.

Vesicles were extracted using a modified version of a previous report (39). B. agrestis was grown in 100 ml TSB medium in each flask and shaken at 200 rpm at 30°C for 16 h. Bacterial culture was centrifuged at 6,000 × g for 15 min at 4°C. The supernatant was filtered through 0.45- and 0.20-μm membranes and ultracentrifuged at 150,000 × g for 2 h at 4°C using a P45AT angle rotor (Eppendorf Himac Technologies). The pellets were resuspended in HEPES-NaCl. For vesicle purification, extracted vesicles were adjusted to 1 ml of 45% (wt/vol) iodixanol (OptiPrep; Axis-Shield Diagnostics Ltd.) in HEPES-NaCl, transferred to the bottom of ultracentrifuge tubes, and layered with iodixanol-HEPES-NaCl (2 ml of 40, 35, 30, and 25% and 1 ml of 20%). The samples were ultracentrifuged at 100,000 × g for 20 h at 4°C using a swinging bucket rotor (P40ST; Eppendorf Himac Technologies). Then, 500-μl fractions were collected from each gradient. The fraction-containing vesicles were ultracentrifuged and resuspended in HEPES-NaCl.

Microscopic observation.

Cells were observed using the Olympus BX53 (Olympus) microscope, and images were captured with the charge-coupled-device (CCD) camera DP72 and processed by the imaging software cellSens. Bacterial DNA was stained with SYBR green, and cell images for phase contrast or fluorescence (485 nm for excitation and 535 nm for emission) were taken.

For TEM observation, samples were placed on 400-mesh grids Cu (JEOL) pretreated with 0.01% α-poly-l-lysine. Bacterial cells and vesicles were stained with 2% (NH₄)₆Mo₇O₂₄ and observed using a JEM-1010 transmission electron microscope (JEOL) at 80 kV equipped with a FastScan-F214 (T) CCD camera (TVIPS).

For QFDE-EM analysis, bacterial cells were washed twice with HEPES-NaCl buffer and centrifuged following the protocol described earlier (46). Briefly, a rabbit lung slab, mica flakes, and bacterial cell pellets were placed on a paper disk attached to an aluminum disk, and then samples were quickly frozen using liquid helium with a CryoPress (Valiant Instruments). The specimens were placed in a chamber maintained at −180°C using a JFDV freeze-etching device (JEOL), freeze fractured with a knife at −180°C after the sample temperature was increased to −120°C, and freeze-etched at −104°C for 15 min. Subsequently, samples were coated with platinum at a 2-nm thickness at a rotary-shadowing angle of 20° and then coated with carbon at a rotary-shadowing angle of 80°. The replicas were then floated in full-strength hydrofluoric acid, rinsed in water, cleansed with a commercial bleach containing sodium hypochlorite, rinsed in water, and finally placed onto 400-mesh Cu grids. Replica specimens were observed with TEM. The acquired pictures were adjusted and colored using the Adobe Photoshop software.

For freeze fracture, bacterial cells were washed and suspended in HEPES-NaCl buffer containing 20% glycerol. The protocol was the same as aforementioned, excluding the etching step.

Particle analyses.

The area of each vesicle was analyzed using ImageJ. Images contained more than 100 vesicles were used for the calculation of particle size. Hydrodynamic diameters of the purified vesicles were analyzed with a Zetasizer Nano ZS particle analyzer (Malvern Instruments) in phosphate-buffered saline (PBS) at 30°C. The hydrodynamic zeta average diameter was calculated using the dynamic light scattering (DLS) method. The zeta potential of the bacterial cells was determined by applying the Smoluchowski approximation.

Quantification of protein concentration.

SDS (1%) was added to homogenize the purified vesicles. Bacterial cells were suspended in PBS and homogenized with an NR-50M ultrasonic homogenizer (Microtec). Protein concentration was determined by BCA protein assay (Thermo) following the manufacturer’s instruction.

Protein composition analysis.

Proteins were treated with SDS loading buffer (Bio-Rad) containing 5% 2-mercaptoethanol at 100°C for 5 min and separated by SDS-PAGE. Protein patterns were detected by using Coomassie brilliant blue (CBB) G250 (Bio-Rad) or silver stain (Fujifilm Wako). Peptide mass fingerprinting (PMF) was performed as reported previously (49). Briefly, CBB-stained bands were cut out from polyacrylamide gel, and trypsin-based in-gel digestion was performed using a DigestPro96 instrument (Intavis). Fragmented peptides were washed with 0.1% trifluoroacetic acid and eluted with 0.1% trifluoroacetic acid in 50% acetonitrile. Eluted samples were desalted using a ZipTip (Millipore). Then, matrix-assisted laser desorption ionization–time of flight mass spectrometry (MALDI-TOF MS) was performed with an autoflex speed system (Bruker Daltonics). Proteins were identified by PMF analysis according to the entire B. agrestis molecular mass in the NCBI protein database using the MASCOT search engine.

KDO analysis.

KDO analysis was performed based on a modified method of a previous report (50). Briefly, 50 μl of 200 μg/ml proteins (vesicles or cell lysates) were mixed with 50 μl of 0.5 N sulfuric acid and heated at 100°C for 15 min. After cooling to room temperature, 200 μl of 100 mM arsenous acid in 0.5 N HCl was added and mixed. Subsequently, samples were mixed with 800 μl of 4 mM thiobarbituric acid and then heated at 100°C for 10 min. After cooling to room temperature, 1.5 ml of butanol solution (n-butanol/sulfuric acid, 19:1) was added and vortexed for 30 s. After 10 min, 1 ml of the butanol layer was transferred to new microtubes and centrifuged at 10,000 × g for 2 min, and the residual water fraction was removed. Absorbances of the butanol layer at 522 nm and 509 nm were measured, and KDO concentrations (mM/μg of protein) were calculated from the A₅₂₂ − A₅₀₉ value using purified KDO as a standard.

NADH oxidase assay.

The NADH oxidase assay was conducted based on a modified method of a previous report (51). Purified vesicles or cell lysates were suspended in a 2× mixture solution (0.4 mM dithiothreitol [DTT], 100 mM Tris-HCl [pH 7.9], 0.1% NaHCO₃, 0.2 mg/ml β-NADH). Changes in absorbance were measured every 5 s for 1 min with a spectrophotometer at 340 nm. The molar absorbance coefficient of oxidized β-NADH used was 2.83 μM⁻¹ cm⁻¹, and specific activity was defined as micromoles of product formed per minute per gram of protein.

β-Galactosidase assay.

The activity of β-galactosidase was examined by the degradation of o-nitrophenyl-β-d-galactopyranoside (ONPG) to the yellow product of nitrophenol. Five micrograms of protein (vesicles or cell lysate) was suspended in 200 μl of 1 mM ONPG in PBS, and A₃₄₀ was measured over time. The molar absorbance coefficient of ONPG used was 4.2 mM⁻¹ cm⁻¹, and specific activity (unit) was defined as millimoles of product formed per minute per gram of protein.

Bioinformatics analyses.

Genome sequences were obtained from GenBank accession numbers NZ_JMPI00000000.1 and NC_000913.3 for B. agrestis (ATCC 33320, equivalent to JCM 1090^T) and E. coli MG1655, respectively. Amino acid sequences of TolB were compared using CLC Sequence Viewer 8.