Outer Membrane Vesicles

AMANDA J. McBROOM AND META J. KUEHN*

[SECTION EDITOR: BRETT FINLAY]

Posted May 12, 2005

Department of Biochemistry, Duke University Medical Center, Box 3711, Durham, NC 27710

*Corresponding author. Phone: 919-684-2545, Fax: 919-684-8885, E-mail: This e-mail address is being protected from spambots. You need JavaScript enabled to view it

Outer membrane vesicles (blebs) are produced by Escherichia coli,Salmonella, and all other gram-negative bacteria both in vitro and in vivo. Most of the research in the field has focused on the properties of vesicles derived from pathogenic bacteria and their interactions with eukaryotic cells. These data indicate that vesicles are able to contribute to pathogenesis. Thus, it appears that pathogenic gram-negative bacteria have co-opted vesicles for the dissemination of virulence determinants. However, the role of vesicle production by nonpathogenic bacteria is less obvious. This section reviews the data demonstrating the mechanistic and physiological basis of outer membrane vesicle production by bacteria.

INTRODUCTION

Production of outer membrane vesicles has been documented for a wide variety of bacterial genera, but a limited number of these vesicles have been characterized thus far. An overview of bacterial strains for which vesicles have undergone initial characterization is found in Table 1. Protein profile analysis and enzymatic assays of vesicles have shown both the presence of outer membrane and periplasmic components and the absence of inner membrane and cytosolic material (13, 26, 27, 42, 45, 56, 58, 104, 116) (S. J. Bauman, S. Villalobos, and M. J. Kuehn, submitted for publication). Electron microscopy studies that have captured images of cells in the process of vesicle release support these biochemical data (Fig. 1 and 2). These micrographs show a bulging away of the outer membrane from the inner membrane and a narrowing of the membranous attachment between the cell and the newly formed vesicle protrusion (12, 17, 21, 22, 48, 56, 72, 73, 83). On severing of the attachment, the vesicle is released into the extracellular milieu and is free to diffuse away from the originating cell. Vesiculation of wild-type cells is seen to occur without signs of lysis or disruption of outer membrane integrity (10, 17, 26, 27, 42) (Fig. 1), which is validated by the observation for many bacteria that maximal production of vesicles occurs during the exponential phase of growth (10, 17, 26, 42, 84) (Bauman et al., submitted). In addition, vesicles possess only a single bilayer membrane, not the double membrane that would be expected if they contained both inner and outer membranes (11, 22, 48, 55, 58, 63, 83, 105).

Fig. 1Thin-section transmission electron microscopy of vesicle production by P. aeruginosa. Arrows point to vesicles of various sizes. Bars indicate 200 nm. Photos courtesy of Alice Dohnalkova, Pacific Northwest National Laboratory, Richland, Wash.

Fig. 2(A) Atomic force microscopy of E. coli and adjacent vesicles (arrows). The irregular surface appearance of the bacteria and vesicles is an artifact of sample drying. Apparent size is increased due to flattening upon adhesion to the imaging substrate. (B) Scanning electron microscopy of E. coli producing vesicles (arrows). Bars indicate size.

Table 1Characteristics of selected outer membrane vesicles.

Investigations of the physical properties, composition, origin, formation, and functions of outer membrane vesicles are discussed below.

PHYSICAL CHARACTERISTICS

The existence of outer membrane vesicles was first widely documented in the 1960s, when the release of small rounded structures from gram-negative bacterial cells was observed by electron microscopy in cultures of Escherichia coli, Vibrio cholerae, and Pseudomonas spp. (12, 61, 110, 111, 113). Subsequent work has demonstrated production of the discrete vesicular structures by a wide variety of gram-negative bacteria (Table 1). Vesicles vary in size, ranging from 10 to 300 nm in diameter (Table 1) (Fig. 1 and 2). Calculations based on atomic force microscopy analysis of Pseudomonas aeruginosa PAO1 vesicles suggest that the average periplasmic mass contained within the lumen of vesicles from this strain is 2 × 10^–10 to 3 × 10^–10 μg (86).

The process of vesiculation is ubiquitous in nature. Vesicles have been observed not only as a phenomenon of planktonic bacterial cultures in vitro, but also in freshwater environments (6), naturally occurring biofilms (7), tissue culture (22, 36), solid agar (15, 17, 40, 64, 70, 105), and human and animal hosts in vivo(8, 15, 20, 22, 23, 33, 34, 38, 39, 56, 98).

Vesicle production is widely variable among genera. For selected strains of E. coli, measurements of protein abundance and radiolabel incorporation have indicated that vesicles typically represent 1% to 5% of total outer membrane material (27, 104, 109) and approximately 0.2% to 0.5% of total cell material in a culture (42, 77). By contrast, Neisseria meningitidis produces an amount of released vesicles corresponding to 8% of total ¹⁴C radiolabel and 12% of the total endotoxin of log-phase cultures (17), significantly more than the quantities released by E. coli. Additionally, even within species vesiculation levels are not constant. Enterotoxigenic E. coli strain ATCC 43886 produces approximately 10-fold more vesicles than the laboratory E. coli strain HB101 (45). We initiated a screen of randomly mutagenized E. coli to analyze the genetic basis for vesicle production (A. J. McBroom, A. P. Johnson, S. Vemulapalli, and M. J. Kuehn, submitted for publication). Gene disruptions that caused vesicle under- and overproduction revealed that vesicle production is affected by genes encoding envelope proteins and lipids, the envelope stress response, and lipid-anchored proteins, as well as global factors. Pathogenic bacteria may express additional factors that affect the amount of vesicles produced.

VESICLE COMPOSITION

Detailed chemical composition studies of vesicles have been published for E. coli and Brucella melitensis (26, 27, 42). Vesicle proteins, lipids (phospholipids, lipopolysaccharide [LPS], and fatty acids), and enzymatic activity closely resemble that of the outer membrane (26, 27, 32, 42, 45, 55, 56, 58, 63, 84, 109, 112, 116) (Bauman et al., submitted), clearly reflecting the outer membrane origin of vesicles.

While vesicle protein profiles are typically very similar to those of the outer membrane, they are not identical. Comparisons of vesicles and the outer membrane fraction of the cells from which they were released have revealed several components whose relative abundance differs (26, 27, 32, 42, 45, 55, 84, 109, 112) (Bauman et al., submitted). Both enriched and depleted protein components have been observed, although few of those components have been identified thus far. Identified outer membrane proteins enriched in vesicles include OmpW and OmpX for enterotoxigenic E. coli (45) and leukotoxin for some strains of Actinobacillus actinomycetemcomitans (55). The heat-labile enterotoxin (LT) of enterotoxigenic E. coli is also enriched; itis present both on the exterior of vesicles and in the interior vesicle lumen (43, 44, 45, 105). Unidentified bands present in higher amounts in vesicle protein profiles than in outer membranes may represent either enriched outer membrane components or packaged periplasmic material.

Identified E. coli outer membrane proteins depleted in vesicles include the major lipoprotein Lpp (42, 109), protein V (109), protein G (42, 77), and OmpA (previously known as protein II*) (42, 71, 77, 109). However, depletion of these components may depend on strain and growth conditions. Reports regarding the depletion of OmpA from E. coli vesicles differ (42, 71, 109). In addition, growth medium affects the incorporation of this protein into vesicles within the same strain. For example, outer membranes and vesicles of E. coli B were compared after growth in minimal media supplemented with various amino acids (70). In the outer membrane, the ratio of OmpA (II*) to the sum of other outer membrane proteins ["Ia" (OmpF) + "a"] was fairly constant; however in vesicles, this ratio depended on amino acid supplementation.

The lipid content of vesicles and outer membranes differs as well. Vesicles contain the major outer membrane lipids of their originating outer membranes (26, 42, 45, 55), but their profiles are not necessarily identical. Modified species of the lipid A component of LPS have been observed in the vesicles but not in the outer membranes of an enterotoxigenic E. coli strain (45), and several lipids of unknown structure detected in A. actinomycetemcomitans vesicles were also not present in corresponding outer membranes (55). These minor lipid components could be present in the outer membrane at low levels and concentrated in vesicles or produced as a result of vesicle formation (55). Enrichment over outer membrane quantities has also been demonstrated for phosphatidylcholine in B. melitensis vesicles (26) and specific LPS O-antigen variants in P. aeruginosa vesicles (7, 48, 80).

Apparently a specific signal is not required for incorporation of proteins into vesicles, because it has been shown that heterologous proteins can be efficiently packaged into the structures. Genes for both the Yersinia pestis Ail outer membrane protein and a periplasmic green fluorescent protein (GFP) from the luminescent jellyfish Aequorea victoria have been introduced into E. coli and subsequently packaged as efficiently as endogenous cargo into vesicles produced by those strains (58). While it is feasible that the Ail protein would contain a signal for vesiculation that could be common to both Yersinia and E. coli if one exists, it is less likely that GFP would contain such a motif.

DNA is associated with vesicles from several bacteria including E. coli. Both chromosomal and plasmid DNA content has been detected that is resistant to external nuclease (19, 63, 114). If DNA enters vesicles during vesicle formation, it must be present in the periplasm of the cell. This has been proposed to occur for E. coli, Neisseria, and Pseudomonas species (19, 48, 63, 87, 114). Alternatively, extracellular DNA could be incorporated into vesicles. For Pseudomonas, studies suggest that free vesicles can take up DNA from the extracellular environment in addition to capturing DNA from the periplasm (87). For A. actinomycetemcomitans vesicles, DNA is associated with the exterior surface and is responsible for the localization of leukotoxin to the vesicle surface (82). Haemophilus influenzae vesicles, produced during conditions that promote competence, contain DNA and are enriched in DNA-binding proteins (14, 16, 51, 52). Unlike other vesicles, H. influenzae vesicles remain bound to the bacteria.

The composition of vesicles, in particular of their proteins and lipids, may reveal functional characteristics of vesicles in the environment and mechanistic requirements that lead to vesicle formation. For instance, particular proteins or lipids may promote membrane curvature or fusion/fission and thus be enriched in the vesicles. Particular enzymatic or nucleic acid cargo included in the structures give insight into the role of vesicles in the transmission of information both between prokaryotes and between prokaryotes and eukaryotes.

VESICLE FORMATION

The process by which vesiculation occurs has been a subject of debate for many years, and several mechanisms of vesicle formation have been proposed. Steps involving physical protrusion of the outer membrane and subsequent release of a spherical body of periplasm bounded by outer membrane have been observed by electron microscopy (10, 17, 21, 22, 48, 56, 72, 73, 83) (Fig. 1 and 2). However, the trigger or signal for vesicle formation at a particular site remains unclear. Central to the proposed models for vesicle formation is the exclusion of the peptidoglycan layer from incorporation into vesicles. Although reports differ as to whether vesicles contain constituent components of peptidoglycan (72, 116), which exist in both free and bound states in the cell, electron micrographs do not reveal an intact peptidoglycan layer within vesicles (23, 48). Numerous interactions occur between peptidoglycan and the outer membrane, mainly through lipoproteins. The formation of an outer membrane bulge and subsequent release of a vesicle that excludes the peptidoglycan layer must occur in regions lacking such a connection. Thus, vesicle formation is predicted to occur at areas where the interactions are sparse or nonexistent, or where connective bonds are actively broken (Fig. 3).

Fig. 3Models of outer membrane vesicle production. (A) Vesicle production occurring at a site of sparse outer membrane-peptidoglycan linkages. Particular outer membrane proteins may accumulate at these sites (diamonds), resulting in their enrichment in vesicles. Outer membrane (OM), peptidoglycan (PG), and inner membrane (IM) are indicated. (B) Vesicle production occurring at a site where the outer membrane-peptidoglycan linkages are broken. Recruitment of specific outer membrane proteins (diamonds) to that site before release may result in enrichment of those proteins in vesicles. (C) Vesicle production occurring due to turgor pressure exerted by a buildup of periplasmic material (circles). This may occur at loosely attached sites of the OM enriched in particular OM proteins, as depicted in panel A.

Hoekstra et al. (42) hypothesized in 1976 that vesicles originate from areas of the cell where there are few outer membrane-lipoprotein connections to hold the outer membrane and peptidoglycan in close association. Extending this concept further, the Witholt laboratory found a reduction in the amount of both free and bound lipoprotein present in vesicles (25 to 35% and 0 to 25% of outer membrane quantities, respectively) (109), and a preferential incorporation of newly synthesized proteins (77). This led to the suggestion of a model where vesicles are produced when the outer membrane of the cell grows faster than the underlying peptidoglycan, resulting in localized detachment of the two layers and a bulge at the site of excess outer membrane (77, 109) (Fig. 3A). However, the conclusion that newly synthesized proteins are incorporated into vesicles at a higher rate than older material was based on the association of radiolabeled leucine with released vesicles. Although there was indeed an increase in total vesicle leucine content as compared with the outer membrane fraction, the study methods did not subtract the contribution of free label in the periplasm, which would have also been present in the vesicles.

Alternatively, targeted degradation of peptidoglycan may be involved in vesicle formation (Fig. 3B). Peptidoglycan-hydrolyzing autolysins are associated with Pseudomonas vesicles (46, 65, 66), and it has been suggested that localized and highly transient hydrolysis of the peptidoglycan layer at the site of vesicle blebbing is required for vesicle release without concomitant cell lysis (48). Such activity would seem to require a complex regulatory system. Perhaps a degradation mechanism specific to vesicle release is not required, since the cell turns over and recycles approximately 50% of its peptidoglycan layer per generation (30). With such frequent dissolution of bonds, there is likely ample opportunity for vesicle formation. In fact, a mutant of Porphyromonas gingivalis with a disruption in a putative autolysin gene produces more vesicles than wild type (35). This does not prove or disprove any particular theory since the strain likely has other autolysin genes, but the observation is consistent with the idea that an imbalance between peptidoglycan and outer membrane growth leads to vesicle overproduction.

Zhou et al. (116)propose another mechanism, hypothesizing that vesicles result from the accumulation of cell wall turnover products. In this model, breakdown products excised from the peptidoglycan layer exert a turgor pressure that causes outer membrane distension and subsequent vesicle release (116) (Fig. 3C). Unlike the Witholt model, this mechanism of vesiculation would not require an imbalance between outer membrane and peptidoglycan growth. The authors point to the presence of muramic acid in P. gingivalis vesicle preparations as evidence in support of their proposal (116).

LPS has also been suggested to play a role in vesicle biogenesis. LPS is a major structural element of the bacterial outer membrane that consists of three components: lipid A, core oligosaccharide, and the O-antigen polysaccharide side chain. The act of vesicle formation requires a sharp change in the membrane from a relatively low-curvature state in the outer membrane of the bacterial cell to the highly curved nature of a small vesicular structure. In studies of P. aeruginosa strain PAO1, which produces two types of O-antigen side chain, it has been suggested that the type and/or length of the O-antigen may affect vesicle formation (65, 80). While the surface of PAO1 contains both A-band and B-band types of O-antigen, vesicles produced from the strain are enriched in the highly charged and longer B-band form (7, 48, 80).

Beveridge and coworkers have proposed that B-band LPS enrichment may occur via preferential budding of vesicles from membrane regions rich in this form of LPS. In their model, charge repulsion between highly electronegative groups in the upper portion of adjacent B-band molecules would compress the hydrophobic segments in the membrane, forcing the region into a higher state of curvature and thereby initiating vesicle formation (46, 65). Regions containing the shorter and more neutral A-band LPS molecules would have little charge repulsion and thus be less likely to form areas of high curvature (46, 65). The results of two independent studies support this theory: an increased amount of material (as measured by total protein and dry weight) is recovered from the culture supernatant of an isogenic mutant expressing only B-band LPS (80), and increased vesicle formation results from PAO1 grown under oxygen stress conditions that increase B-band LPS (91).

Aside from the type of side chain present on the cell surface, the expression of O-antigen also influences vesiculation. Electron micrographs of Salmonella and P. aeruginosa mutants missing the O-antigen side chain show increased vesicle formation (75, 95). This phenomenon has been reported for a range of mutants with disruptions in the core oligosaccharide region (75). Core truncation affects the quantity of vesicle release; however, the phenotype can be attributed entirely to O-antigen loss, and thus these studies do not differentiate which component of LPS causes the phenotype. Recent studies in our laboratory suggest that alteration of the core region itself affects vesiculation. The quantity of released vesicles differs for K-12 E. coli strain DH5α, which lacks O-antigen, and DH5α mutants containing polar disruptions of the waaG (rfaG) core biosynthesis glucosyltransferase gene by using both protein and lipids to quantitate vesicle production (McBroom et al., submitted). Core region LPS mutants have outer membrane defects including decreased OMP expression (4, 92, 95). Therefore, the vesiculation phenotype seen in such mutants could be an effect of altered outer membrane protein composition.

Mutations in the genes encoding the major outer membrane proteins of both Salmonella and E. coli cause an increase in vesicle production without necessarily compromising the integrity or barrier function of the membrane (81) (McBroom et al., submitted). However, without in-depth compositional analyses, it is difficult to interpret whether the specific impact of a major outer membrane protein deletion on vesiculation is direct or indirect.

Other strains observed to release an increased amount of vesicles include mutants with disruptions in the major outer membrane lipoprotein lpp gene and the tolQ, tolR, tolA, tolB, and pal members of the Tol-Pal system in E. coli (5, 9) and Pseudomonas putida (67). Other than involvement in the uptake of colicins and filamentous bacteriophage (106), the only currently known function of the Tol-Pal system is in maintaining outer membrane stability (5, 9, 67, 68). In addition to increased vesiculation, the lpp and tol-pal mutant strains have significant defects in outer membrane integrity demonstrated by their detergent sensitivity and leakage of periplasmic material into the culture medium (5, 9, 67).

Similarly leaky E. coli lpo and Salmonella enterica serovar Typhimurium lkyD mutants deficient in outer membrane lipoprotein function and cell division produce extremely large membrane blebs over the septal region when grown under low Mg²⁺ conditions (24, 25, 41, 100, 107, 108). The lpo mutation is thought to be a deletion encompassing the lpp gene (41), while the lkyD mutation has been suggested to cause a defect in the formation of lipoprotein-peptidoglycan bonds (108). The size and localization of the bulging structures in these mutants to the site of cell division indicate that they are very different from normally produced vesicles. Formation of the bulges at cell division sites is reduced in the E. coli lpo mutant by the addition of Mg²⁺ to the culture media (24, 100), but the increase in Mg²⁺ concentration has no effect on the Salmonella lkyD mutant (24). A double mutant strain of E. coli with disruptions in both lpp and the outer membrane protein gene ompA requires high Mg²⁺ conditions for growth and produces a large number of vesicles over the entire surface of the cell in a manner more consistent with normally produced vesicle structures (96).

The nature of these mutations in peptidoglycan-interacting outer membrane lipoproteins and in components of the periplasm-spanning Tol-Pal system that links the outer membrane, peptidoglycan layer, and inner membrane may support hypotheses of vesicle formation that center on a loosening of the stabilizing interactions in the cell envelope. However, these strains exhibit significant outer membrane instability, thus the abundant vesicles they produce may differ from normally produced vesicles from a wild-type cell.

The behavior of a lysine auxotroph grown under lysine-limiting conditions provides support for the Witholt theory that vesicle formation occurs due to imbalanced growth of the outer membrane and peptidoglycan. During growth under these conditions, both protein and peptidoglycan synthesis are inhibited. However, production of LPS would be unaffected. This continued outer membrane growth while the peptidoglycan layer remains static would lead to bulges of excess outer membrane material and potential release as vesicles. In accordance with this theory, electron microscopy of the mutant shows that it produces numerous vesicles under these conditions (61, 113). Similar increased release of material was observed in studies using chloramphenicol to inhibit protein synthesis in strains of E. coli and serovar Typhimurium (69, 90), and the LPS-phospholipid-protein nature of the released complexes was confirmed by compositional analysis (90).

Regardless of the proposed model, it is uncertain whether vesicle production is an active process requiring energy or coordinated activity. Though this question has not yet been resolved, recent work in our laboratory suggests that outer membrane instability alone is neither necessary nor sufficient for vesicle production. In a screen for mutations causing altered vesiculation levels in an E. coli K-12 strain, we identified several gene disruptions that promote or restrict vesicle production (McBroom et al., submitted). These mutants include both vesicle-overproducing strains with little to no observable envelope stability defects and vesicle-underproducing strains with significant loss of outer membrane integrity. From these data, we conclude that vesiculation is a process not necessarily tied to the state of the outer membrane. Several of the mutants discovered in our screen have disruptions in genes related to peptidoglycan synthesis or degradation. These observations may provide support for a role of vesicle production in maintaining balanced growth of the envelope, or for a link between vesicle production and the strength of the interactions between the outer membrane and peptidoglycan. We also identified mutants in the tol/pal gene cluster, supporting previous reports of increased vesiculation in such strains (5, 9).

A particularly interesting subset of mutants was identified in the sigmaE envelope stress-response pathway (reviewed in references 1 and 2). Our data suggest that cells respond to the disruption or hyperactivation of this response pathway by producing a large quantity of vesicles (McBroom et al., submitted). Without the ability to sense a buildup of misfolded protein products and degrade them, bacteria may relieve the envelope stress by exporting the stress products via vesicles. Hyperactivation of the pathway would drive sigmaE-dependent expression of folding factors to the periplasm and may lead to envelope stress that results in the observed increase in vesiculation. This leads to our proposal of a generalized model for vesicle formation under stress that is in line with the previous peptidoglycan-based turgor pressure hypothesis (116) (Fig. 3C). However, our model operates under a hypothesis that the turgor pressure may be induced not only by peptidoglycan breakdown products, but also by an excess of protein in the periplasm. The recent demonstration that overexpression of the periplasmic domain of certain proteins leads to increased vesicle production fits with our theory (40).

This product-elimination model suggests that vesiculation serves as a means for the cell to eliminate excess envelope material that may be detrimental to its health, in particular, during times of imbalanced growth and stress. It remains to be determined whether this elimination can occur in a directed fashion. For example, it would be most beneficial for the cell to selectively release misfolded or otherwise damaged proteins into vesicles while retaining properly functioning components. Whether vesicles are enriched in this type of specific cargo has not yet been analyzed.

Considering that the outer membrane components in vesicle composition profiles do not simply reiterate the banding pattern of the outer membranes of their originating cells, some selective mechanism must exist for the incorporation of outer membrane material. The mechanism leading to enrichment or depletion could be highly complex, involving specific protein interactions and a specialized machinery for the formation and release of a vesicle; or more simple, indicating a preference for vesicle budding at sites rich or deficient in certain components. Release of vesicles at preferential sites in the latter theory could indicate areas of the membrane that are weaker in attachment to underlying layers of the cell, or perhaps regions that, by virtue of their composition, are more amenable to the drastic change in surface curvature required for vesicle formation (Fig. 3). Preferential incorporation of periplasmic proteins that are lumenal vesicle components, if shown to occur, would take place by a different means. Such components could be incorporated at a higher rate due to a spatial orientation in the cell that places them in closer proximity to sites of vesicle formation than other periplasmic proteins, by a loose association with the outer membrane, or via an as-yet unknown means of selective packaging.

A caveat of this analysis is that vesicles within the population may be heterogeneous in composition. Assessments of enriched or depleted components have been performed by analysis of total purified vesicle samples. Recent and ongoing research points to a complex differential distribution of outer membrane proteins in bacterial cells (29). Considering this heterogeneity of the outer membrane, heterogeneity in outer membrane vesicle composition is likely unless vesicle formation is restricted to occur only at sites of specific composition.

The rate or extent of vesicle production by one bacterial strain can vary more than three orders of magnitude. We have characterized E. coli mutants that produced from 70 to 80% fewer vesicles than wild type and mutants that produced roughly 200-fold more vesicles than wild type (McBroom et al., submitted). In addition, we noted that several strains of E. coli produce more vesicles in rich broth than when grown in minimal media (unpublished data), an observation supported by electron microscopy studies of Pseudomonas showing increased vesiculation in rich media at higher temperatures (111). This plasticity in the degree of vesicle production suggests that vesicle production can be a highly regulated process for bacteria in response to metabolic, environmental, or growth stimuli.

BIOLOGICAL ACTIVITY OF VESICLES

The formation of vesicles offers the cell a means for elimination and remodeling of envelope components, in particular, during times of stress. Bacteria experience many stresses: change in growth phase, temperature, pH, nutrient availability, redox state, and exposure to harmful agents. Vesiculation can be seen as a mechanism for cells to react to conditions in the surrounding environment by carrying away unnecessary components and allowing rapid modification of the outer membrane composition. In addition, vesicles can transmit biological activities distant from the originating cell.

Vesicles may serve a number of growth-enhancing and protective functions for bacteria. P. gingivalis vesicles act as bridging factors in vitro, mediating the coaggregation of a wide range of oral bacteria that probably translates to the promotion of biofilm formation and colonization in vivo (32, 53, 54, 72). In fact, vesicle structures are observed by electron microscopy in gingival plaque samples (34); thus, this aggregative activity may contribute to the establishment of infection in the oral cavity.

Vesicles may also aid a strain’s survival in mixed population infections by eliminating competing bacterial strains. Vesicles package periplasmic peptidoglycan hydrolases (48, 65), and when the vesicles from one species (self) contact other gram-negative and gram-positive cells (non-self), the activity of those enzymes can cause lysis (46, 50, 66). Self-cells have the appropriate regulatory systems to control the delivered autolysins, but non-self-cells do not. Bacteriolytic interaction of vesicles with non-self-cells occurs for numerous vesicle and strain combinations (46, 50, 66). Under poor nutritional conditions, release of nutrients from lysed bacteria and removal of those organisms from the competitive environment could be crucial to survival of the vesicle-producing strains.

Lysis of competing organisms is not the only way in which vesicles have been implicated in scavenging nutrients for the donor cell. Pseudomonas fragi, a meat spoilage bacterium, undergoes induction of vesicle formation and extracellular proteolytic activity upon growth on a protein substrate, while the same strain forms few vesicles in media without protein (21, 101). The protease has been localized to the cell envelope and outer membrane vesicles (102), leading to the hypothesis that vesiculation may function as a mechanism for concentrated secretion of enzymes to hydrolyze proteins into smaller products for uptake by the cell (21, 102).

The potential digestive role for vesicles is not limited to protein substrates. Bacteroides succinogenes grown with cellulose as a carbohydrate source releases vesicles containing the hydrolytic enzymes cellulase and xylanase, and these vesicles seem to be produced in higher numbers by cells that are in physical contact with the cellulose substrate than those that are unattached in the media (23). Free vesicles also adhere to the cellulose fibers in vitro and in vivo. B. succinogenes is one of the major rumen bacteria responsible for forage digestion, and release of vesicles with polymer digestive activity may provide a sugar nutrient source for cellulolytic and noncellulolytic bacteria in that microbial population (23). Finally, P. gingivalis increases vesicle production when it is grown under hemin limitation (74), suggesting vesicles could also play a role in iron acquisition.

Vesiculation may be a protective function for the bacterium both during and after vesicle formation. The physical act of vesicle budding and release offers a mechanism for removing harmful factors from the exterior of the cell. For example, binding of bacteriophage to the cell surface has been shown to increase the release of outer membrane material composed of protein, phospholipid, and LPS at a dramatic rate (69, 71). This high rate of release is sustained for only a short period, possibly as a means for the cell to rid itself of attaching phage particles by sequestering them into vesicles for expulsion. Another example of sequestration and removal of toxic agents via vesicles is exhibited by P. putida. A toluene-resistant strain of P. putida grown in the presence of toluene eliminates the organic solvent from its outer membrane through vesicle production (62).

Vesicles in the area surrounding the cell may also provide the cell protection inside a human or animal host. Vesicles produced by β-lactam-resistant strains of P. aeruginosa contain the β-lactam antibiotic-inactivating enzyme β-lactamase (13). These vesicles are permeant to antibiotics, and drug that enters the vesicle lumen would be inactivated (13). Thus, release of β-lactamase-containing vesicles into the extracellular milieu could enhance the survival of neighboring nonresistant bacteria by decreasing the local antibiotic concentration. P. gingivalis vesicles inhibit antigen presentation by eukaryotic cells, a mechanism that may allow the bacteria to escape detection by the immune system (97). Vesicles have the potential to act as decoy targets, presenting an antigenic surface that draws the resources of the immune system, thereby allowing the survival of the bacteria. Serum-resistant strains of Neisseria gonorrhoeae appear to produce more vesicles than serum-susceptible strains, and these vesicles significantly decrease serum-mediated killing in bactericidal assays (85). Vesicles produced by a serum-resistant strain have been seen to provide more protection and compete more efficiently for binding to serum antibodies than vesicles from a serum-susceptible strain (85). E. coli grown in the presence of serum release a larger quantity of LPS-protein complexes (38) that appear identical in composition to vesicles. These findings suggest that vesicles could act to bind and deplete host immune factors at the site of infection that would otherwise attack the bacteria.

The observation that vesicles contain DNA leads to the intriguing notion that they can serve as transforming agents, facilitating the transfer of genetic material between strains of bacteria. In the case of H. influenzae, vesicles contain proteins with DNA-binding activity and have been suggested to be specialized structures that remain attached to the bacteria that facilitate the uptake of DNA from the environment (14, 16, 51, 52). However, vesicles also can serve as remote vehicles that carry and deliver DNA to recipient cells. DNA contained within vesicles is protected from exonucleases (19, 63, 87, 114), and the fusogenic nature of vesicles (3, 47) could lead to the direct delivery of the DNA into recipient bacteria. Vesicle-mediatedtransformation has been reported in vitro between strains of Neisseria (19) and from E. coli O157:H7 to Salmonella and other E. coli strains (63, 114), along with subsequent expression of the transferred genes (19, 114). However, not all transformation attempts have been successful, which may indicate strain dependence of the mechanism.

Vesicle release is a unique form of secretion that provides a distinct environment for protein folding. Sequestration of proteins into vesicles releases them from the control of the disulfide bond oxidation and isomerization system present in the periplasm of the bacteria. Removed from the recharging inner membrane components of those systems, the periplasmic oxidoreductases lose their ability to act on substrates efficiently. This sequestration has been shown to be especially important in the case of E. coli ClyA, a cytotoxin that exhibits maximal activity when packaged into vesicles (104). ClyA in the periplasm exists in an oxidized, monomeric, and relatively inactive state. Upon secretion as a vesicle component, the redox status of ClyA changes. The monomers become reduced and oligomerize within the vesicles, thus forming highly active pore assemblies as they leave the bacterial cell.

The concept of vesicles as virulence factors has received considerable attention, and they are likely to play a significant role in the pathogenesis of gram-negative bacteria. Their interactions with host defense systems were described above. In addition, vesicles produced by several bacterial strains have been detected during in vivo infection of human and animal hosts and the interactions of vesicles with eukaryotic cells (including delivery of active toxin and internalization of vesicles) have been extensively reported (8, 15, 20, 22, 33, 34, 38, 39, 56, 59, 88, 98). By virtue of their small size, vesicles are able to interact with eukaryotic host cells and tissues in ways that whole bacteria are incapable of achieving. The role of vesicles in virulence is reviewed elsewhere (N. C. Kesty and M. J. Kuehn, submitted for publication).

Based on the aggregate of data from many different laboratories studying many different bacterial species, it is no longer tenable to presume that bacterial outer membrane vesicles are artifacts of microscopy studies or of in vitro growth. By analysis of their composition, mechanism of formation, regulation, and physiological function, progress is being made in understanding the ubiquitous nature of outer membrane vesicles produced by gram-negative bacteria.