Although the observation that Gram-negative bacteria produce outer membrane vesicles (MVs) was made over 40 years ago, their biological roles have become a focus of study only within the past 10 years. Recent progress in this area has revealed that bacterial MVs are utilized for several processes including delivery of toxins to eukaryotic cells, protein and DNA transfer between bacterial cells, and trafficking of cell–cell signals. Some of these roles appear to be generalized among the Gram-negative bacteria while others are restricted to specific bacterial species/strains. Here we review the known roles of MVs, propose other roles for MVs in mediating interspecies and inter-kingdom communication, and discuss the mechanism of MV formation.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
A large number of Gram-negative bacteria naturally produce extracellular outer membrane vesicles (MVs). These spherical bilayered MVs are liberated from the outer membrane and range in size from 50 to 250 nm in diameter. Similar to the outer membrane, MVs contain lipopolysaccharide (LPS), outer membrane proteins, and phospholipids; however, it is clear that MVs are not simply small vesicles derived entirely from the outer membrane. Along with outer membrane proteins, MVs also contain periplasmic proteins. In some cases, specific proteins have been shown to be enriched or excluded from MVs suggesting a specific sorting mechanism(s) for these proteins (Wensink and Witholt, 1981; Kato et al., 2002; Wai et al., 2003). In addition, MVs from Porphyromonas gingivalis contain muramic acid (Zhou et al., 1998) suggesting that peptidoglycan is present within vesicles liberated from this bacterium. These findings illustrate that the composition of MVs is complex and distinct from the outer membrane and accentuates the potential for a diversity of biological functions for these vesicles.
Membrane vesicles have been found associated with Gram-negative bacteria growing planktonically and in surface-attached biofilm communities, in solid and liquid media, and in natural environments (Beveridge, 1999). Studies of the biological roles of MVs have primarily focused on their use as trafficking vehicles for bacterial toxins. As MV toxin trafficking has been reviewed recently (Kuehn and Kesty, 2005; McBroom and Kuehn, 2005), this review will primarily focus on other known and hypothesized roles for MVs and discuss potential mechanism(s) of MV formation.
Biological roles of MVs
Interspecies interactions shape bacterial community structure in natural environments and during multispecies infections. To enhance their fitness in polymicrobial communities, bacteria have evolved a multitude of characteristics including the production and secretion of antimicrobial factors that act to eliminate competing microorganisms. The antimicrobial activity of supernatants from the Gram-negative bacterium Pseudomonas aeruginosa has been recognized for over 100 years (Pasteur and Joubert, 1877; Bouchard, 1889); however, the role of individual lytic components and how these components are trafficked outside the bacterium has not been systematically examined. Work by Li et al. (1996) illustrated that MVs isolated from P. aeruginosa possess antimicrobial activity. Studies investigating the basis of this antimicrobial activity identified a small protein with murein hydrolase activity capable of cleaving the covalent bonds of peptidoglycan. Further studies illustrated that through MV interactions with Gram-negative and Gram-positive bacteria, this enzyme is delivered to the peptidoglycan layer where it facilitates degradation and subsequent lysis of the bacterial cells. Recent work in ourlaboratory indicates that in addition to the murein hydrolase, P. aeruginosa MVs contain an array of antimicrobial quinolone molecules that also facilitate killing of Gram-positive bacteria (Mashburn and Whiteley, 2005).
Subsequent studies provided evidence that antimicrobial activity is not restricted to MVs isolated from P. aeruginosa. Li et al. (1998) reported that MVs from 15 different Gram-negative bacteria possess antimicrobial activity against Gram-positive and Gram-negative bacteria, with P. aeruginosa having the broadest killing spectrum. Interestingly MVs showed very little lysis of the donor strain from which they were derived. Due to their antimicrobial activity and their proposed function in polymicrobial competition, antimicrobial MVs have been termed ‘predatory’ MVs (Beveridge et al., 1997). Although the importance of predatory MVs in nature is incompletely understood, it is likely that they mediate lysis of competing bacteria in polymicrobial communities. The advantage of packaging lytic factors within MVs is unknown. In some environments, MVs might facilitate dissemination and protection of antimicrobial factors or might act to concentrate them and allow their mass delivery to surrounding bacteria. However, it is clear that lysis of surrounding bacteria will not only reduce competition within polymicrobial environments but also serve to provide the lytic bacterium with nutrients liberated from lysed bacteria.
Fourth mechanism of DNA transfer
Kolling and Matthews (1999) provide evidence that in addition to proteins, MVs also contain DNA. These studies noted that MVs isolated from Escherichia coli strain O157:H7 harbour DNA that is protected from degradation by extracellular DNases. This study also demonstrated that E. coli O157:H7 MVs are proficient in mediating transfer of DNA from one strain to another, enabling efficient transformation of genes. Yaron et al. (2000) provided evidence that E. coli O157:H7 MV DNA originates from multiple sources including the chromosome, plasmids and phage One plasmid packaged within E. coli O157:H7 MVs was a 3.3 kb plasmid present in some clinical isolates and harbouring genes necessary for plasmid replication, mobilization and partitioning. Interestingly this plasmid does not contain genes important for biosynthesis of the pilus necessary for efficient conjugation of mobilizable plasmids. The authors hypothesized that packaging this plasmid into MVs allows it to be trafficked from donor to recipient cells, thereby eliminating the need for pilus biosynthesis genes and the close proximity of bacteria cells normally required for plasmid conjugal transfer. Thus MVs represent a fourth method of gene transfer distinct from transformation, conjugation and transduction. Plasmid transfer using MVs allows protection and direct delivery of genetic material between cells and could be clinically relevant through trafficking of genes associated with virulence and antibiotic resistance. Further studies revealed that MV DNA packaging is not restricted to E. coli. P. aeruginosa MVs also contain DNA and as observed in E. coli, P. aeruginosa MVs package intact plasmids (Renelli et al., 2004). However, unlike E. coli MVs containing plasmids, MVs from P. aeruginosa are not capable of mediating plasmid transformation into recipient cells under the conditions tested, illustrating that strain/species origin might impact the biological function of bacterial MVs.
As MVs package periplasmic components, how can cytoplasmic DNA be present in MVs? Although the mechanism has not been elucidated, two models have been proposed (Renelli et al., 2004). The first model involves movement of DNA from the cytoplasm into the periplasmic space where it is encapsulated during protrusion and ‘pinching off’ of the MVs; however, it is not known if DNA can move directly from the cytoplasm to the periplasm. Alternatively extracellular DNA (potentially from lysed cells) could enter the periplasm and become encapsulated into MVs. The second model involves internalization of extracellular DNA (again possibly from lysed cells) by MVs. Although the mechanism of MV DNA uptake is unknown, it is clear that exogenously added DNA is sequestered by MVs and protected from DNase treatment. Recent experiments by Renelli et al. (2004) support the notion that both mechanisms are important for incorporation of DNA into MVs.
Roles for MVs in natural competence have also recently been suggested. Haemophilus parainfluenza is a naturally competent Gram-negative bacterium capable of binding and internalizing double-stranded DNA containing unique sequences. Kahn et al. (1982) demonstrated that H. parainfluenza competent cells possess higher levels of membrane blebs on their outer membrane than non-competent cells. Upon addition of linear DNA to these competent cells, membrane blebs on the bacterial surface disappeared. The authors suggested that DNA specifically interacts with the membrane blebs thereby facilitating import of the membrane DNA complex into the bacterium. MVs liberated from the cell were also isolated and shown to contain DNA which was resistant to exogenous DNase treatment. A similar study by Deich and Hoyer (1982) showed that competent Haemophilus influenzae also contain high levels of MVs on their surface. These MVs were liberated from the bacteria after the cells were returned to normal (non-competent) growth conditions resulting in a significant loss of DNA binding capacity by the cells; however, the liberated MVs retained the ability to bind DNA. These studies suggest MVs play an important role in natural competence and might function to protect DNA from degradation as well as mediate DNA delivery during natural transformation in Haemophilus.
Membrane vesicles biology has primarily focused on their use as delivery vehicles for bacterial toxins to eukaryotic cells. The subject has been reviewed in depth recently (Kuehn and Kesty, 2005) and will only be presented briefly here. Several pathogenic bacteria including E. coli, Actinobacillus actinomycetemcomitans, Helicobacter pylori and Shigella dysenteriae (for a complete list see Kuehn and Kesty, 2005) produce MVs with associated toxins that are capable of delivering these toxins to eukaryotic target cells. One particularly interesting study of the E. coli cytolysin A (ClyA) indicated that not only is this toxin present within MVs but that MVs are critical for its assembly into an oligomeric (active) form. ClyA is the most abundant protein found in MVs from E. coli and is present as inactive monomers in the periplasm and as active oligomers in MVs (Wai et al., 2003). ClyA was eightfold more potent when delivered to eukaryotic cells via MVs rather than as purified monomeric protein (Wai et al., 2003). There are numerous other examples of toxin trafficking by MVs but in many cases the contribution of MVs to toxin secretion as opposed to other secretory pathways is not known. It is likely that in some cases MVs could serve a minor role in secretion whereas they may play a larger role in other cases.
Membrane vesicles are not only important for transferring protein toxins to eukaryotic cells but are also important for trafficking proteins between bacterial cells of the same species. Ciofu et al. (2000) demonstrated that P. aeruginosa uses MVs as transport vehicles to traffic the antibiotic resistance protein β-lactamase from one bacterial cell to another. β-Lactam antibiotics are often used to treat chronic P. aeruginosa infections such as those in the lungs of individuals with the heritable disease cystic fibrosis. Trafficking β-lactamase from one cell to another effectively allows bacteria within the population to ‘share’ the antibiotic resistance protein and obviates the need for the gene encoding the β-lactamase protein to be present within all P. aeruginosa cells in a population. It is instead conceivable that high levels of β-lactamase production and liberation by a few bacterial cells might be sufficient to protect a large portion of the bacterial population from the antimicrobial action of β-lactam antibiotics. Thus MVs provide P. aeruginosa communities with a novel method of antimicrobial resistance that might enhance its survival and persistence in clinical settings. It is plausible that other Gram-negative bacteria might also use this mechanism of antibiotic resistance during chronic and acute infections.
Many bacteria use chemical signals to communicate and modulate expression of specific genes in response to these signals. These systems, collectively referred to as quorum sensing (QS) systems, allow bacteria to monitor their population density and co-ordinate their group activities in response to cell density (Parsek and Greenberg, 2000). P. aeruginosa uses QS to control transcription of approximately 5% of all its genes (Schuster et al., 2003; Wagner et al., 2003), and this bacterium serves as a model bacterium for mechanistic studies of QS. P. aeruginosa produces three primary cell–cell signalling molecules including butyryl-homoserine lactone (C4-HSL), 3-oxo-dodecanoyl homoserine lactone (3OC12-HSL), and 2-heptyl-3-hydroxy-4-quinolone (referred to as the Pseudomonas Quinolone Signal or PQS) which collectively constitute an integrated signalling network (Passador et al., 1993; Pearson et al., 1995; Pesci et al., 1999).
A fundamental aspect of QS is the movement, by active or passive means (Kaplan and Greenberg, 1985; Pearson et al., 1999), of signals into the extracellular milieu, where they serve as mediators of cell–cell communication. However, many QS signalling molecules have significant hydrophobic character that hampers their dissemination between bacterial cells. We recently reported that PQS, but not C4-HSL or 3OC12-HSL, is associated with MVs which serve to traffic this signal within a P. aeruginosa population. Not only is PQS packaged into MVs but this signal is also required for MV formation. Although the mechanism of MV formation is currently unknown, it is clear that the PQS molecule but not PQS signalling mediates MV formation (Mashburn and Whiteley, 2005).
The use of MVs as signal trafficking vehicles has several implications for P. aeruginosa group activities. P. aeruginosa is an important constituent of environmental polymicrobial communities such as those observed in the plant rhizosphere (Iswandi et al., 1987; Gupta et al., 1999). MVs might facilitate movement of P. aeruginosa signals within these communities and might also serve to protect signals from degradation by other microorganisms within the community. By packaging small antimicrobial molecules, MVs also provide a trafficking mechanism for delivery of these molecules to surrounding bacteria, thereby reducing competition within the community. Lysis of surrounding bacteria also provides P. aeruginosa with nutrients for growth. Our laboratory recently reported that P. aeruginosa lyses Staphylococcus aureus in vivo and uses the lysate as a source of iron (Mashburn et al., 2005). Thus MVs provide P. aeruginosa with a small molecule delivery mechanism critical for co-ordinating group activities and for enhancing the competitiveness of P. aeruginosa during growth in polymicrobial communities.
Is MV signal packaging restricted to P. aeruginosa? Although PQS production appears to be restricted to P. aeruginosa, a large number of Gram-negative bacteria produce acyl-homoserine lactone signals (AHLs). The observation that the majority of the AHL signals (C4-HSL and 3OC12-HSL) produced by P. aeruginosa are not packaged within MVs does not preclude MVs from being important for AHL trafficking in other bacteria. Several bacteria produce AHLs with significantly more hydrophobic character than C4-HSL and 3OC12-HSL. The photosynthetic bacterium Rhodobacter capsulatus produces N-hexadecanoyl-HSL (C16-HSL) (Schaefer et al., 2002) and the alfalfa symbiont Sinorhizobium meliloti produces several AHLs including N-octadecanoyl-HSL (C18-HSL) (Marketon et al., 2002). Approximately 35–50% of the C16-HSL in R. capsulatus is associated with the bacterial cells (Schaefer et al., 2002), suggesting that AHL hydrophobicity significantly influences its localization. Thus it is not unreasonable to hypothesize that MVs might be important for trafficking these hydrophobic AHL signals between bacterial cells.
Another potential use of MVs is signal trafficking between prokaryote and eukaryote symbionts. One system where this could be important is in establishing symbiosis between Rhizobium spp. and their legume partners. Several species of rhizobia form nitrogen-fixing root nodules on the roots/stems of leguminous plants. Within these nodules, the bacteria convert atmospheric dinitrogen to ammonium for use by the plant and the plant in turn provides the bacterium with growth substrates (Patriarca et al., 2002). Nodule formation requires communication via specific signals produced by the plant and the bacterium. Although varying slightly in structure, the bacterium-produced signal (collectively referred to as Nod factor) is composed of a chitin backbone of three to five β-1,4 linked N-acetylglucosamines and a fatty acyl chain of 16–20 carbons (Lerouge et al., 1990; Price et al., 1992). Nod factor possesses domains with significant hydrophobic character and studies by Goedhart et al. (1999) indicate that Nod factor readily inserts into artificial membranes. These studies also demonstrate that Nod factor-loaded artificial membranes are competent to traffic Nod factor to the host plant. Whether or not Nod factor is naturally trafficked via MVs is not known but it is clear that Nod factor-loaded vesicles are competent to communicate with the plant.
Mechanisms of MV formation
Although interest in MV biology has increased in recent years, very little is known about the mechanism of MV formation. Electron microscopy studies of budding bacteria reveal an outward protrusion of MVs from the surface of the bacterium until they are pinched off and released into the surrounding medium (Chatterjee and Das, 1967; Deich and Hoyer, 1982; Kahn et al., 1982; Pettit and Judd, 1992; Kadurugamuwa and Beveridge, 1995; Beveridge et al., 1997; Li et al., 1998; Beveridge, 1999; Fiocca et al., 1999). Models for the molecular mechanism of MV formation have been proposed primarily based on genetic and/or biochemical experiments using E. coli and P. aeruginosa. Although these models are not mutually exclusive, it is likely that lipoproteins play a role in MV formation. Lipoproteins are covalently attached to peptidoglycan and act to cross-link the outer membrane and peptidoglycan. Studies by Hoekstra et al. (1976) provided evidence that E. coli MVs contain substantially less lipoprotein than the outer membrane of the corresponding bacterial cells. The authors reasoned that MVs must originate from regions of the outer membrane where there are few peptidoglycan–outer membrane linkages and proposed that this could occur at the site of cell division. Although subsequent studies showing hypervesiculation of E. coli mutants lacking the primary lipoprotein Lpp were purported to support this model (Bernadac et al., 1998), these strains have very fragile outer membranes making it difficult to define the role of Lpp in MV formation. A second lipoprotein, YfgL was also shown to be important for MV formation in E. coli; however, in contrast to the Lpp mutant, deletion of yfgL caused a significant decrease in MV formation (Rolhion et al., 2005). Although a lipoprotein, the primary function of YfgL is not peptidoglycan–outer membrane anchoring but rather synthesis/degradation of peptidoglycan (Eggert et al., 2001). This latter study provides evidence that peptidoglycan turnover is critical for MV formation.
Based on these lipoprotein studies and recent studies of MV formation in P. aeruginosa and P. gingivalis, three molecular models of MV formation have been proposed.
Model 1. MVs are formed when the outer membrane expands faster than the underlying peptidoglycan layer (Wensink and Witholt, 1981). This causes localized detachment of the peptidoglycan from the outer membrane due to a lack of lipoprotein linkages (Hoekstra et al., 1976; Wensink and Witholt, 1981). If this asymmetric growth persists, areas of detachment will ‘bulge’ and be released from the outer membrane as MVs. Support for this model is primarily based on the observation that significantly less lipoprotein is present in E. coli MVs and that proteins that are known to interact with lipoprotein are largely excluded from MVs (Hoekstra et al., 1976; Wensink and Witholt, 1981). This model is not predicated on the complete absence of lipoprotein from the MVs because approximately 66% of the lipoprotein in E. coli is not covalently bound to peptidoglycan (Braun, 1975; Inouye, 1975; Hoekstra et al., 1976); thus it is plausible that lipoprotein present within E. coli MVs is primarily this unbound form.
Model 2. MVs are formed when products arising from normal peptidoglycan turnover are not efficiently internalized by the bacteria cell but instead build up locally in the periplasmic space (Zhou et al., 1998). These turnover products then exert a turgor pressure on the outer membrane causing the outer membrane to bulge. Mechanical motion would then cause shearing and release of the MVs into the extracellular medium. The evidence supporting this model is that muramic acid, a component of peptidoglycan, is present within MVs. This muramic acid was not precipitated with trichloroacetic acid suggesting that peptidoglycan within P. gingivalis MVs is primarily in the form of low molecular weight fragments (Zhou et al., 1998). Subsequently, Hayashi et al. (2002) reported that mutations in a P. gingivalis peptidoglycan hydrolase (autolysin) enhance MV formation in P. gingivalis. These authors argue that loss of this autolysin may prevent complete degradation of cell wall components leading to accumulation of peptidoglycan intermediates in the periplasm. Although these experiments can be interpreted to support this model, as yet no study has evaluated whether the levels of peptidoglycan within P. gingivalis MVs are sufficient to exert turgor pressure on the outer membrane. It is also not known if MVs produced by other bacteria contain peptidoglycan. Future studies addressing these questions are important for determining the validity of this model. It should be noted that this model will also require peptidoglycan build-up in regions of the outer membrane containing few peptidoglycan–outer membrane linkages.
Model 3. Several studies have proposed that the LPS chemical composition is also an important mediator of MV formation. Studies performed over 25 years ago with Salmonella and P. aeruginosa indicate that strains containing truncated LPS exhibit enhanced membrane blebbing (Smit et al., 1975; Salkinoja-Salonen and Nurmiaho, 1978). Subsequent studies expanded on these initial observations and provided evidence that MVs from P. aeruginosa are primarily composed of negatively charged B-band LPS instead of the more neutrally charged A-band LPS (Kadurugamuwa and Beveridge, 1995). From these findings, a model for P. aeruginosa MV formation was proposed in which the electronegative charge of the B-band LPS causes charge-to-charge repulsion and membrane instability resulting in outward membrane blebbing and trapping of periplasmic components within MVs (Kadurugamuwa and Beveridge, 1995). This third model of MV formation is supported by the observation that growth under conditions that enrich B-band LPS enhances MV formation by P. aeruginosa (Sabra et al., 2003).
Our recent studies describing the importance of PQS in mediating MV formation provide new evidence for this model. The most compelling hypothesis for how PQS mediates MV formation is via destabilization of the Mg2+ and Ca2+ salt bridges in the outer membrane that are critical for negating charge–charge repulsions between LPS molecules. Several quinolones, including quinolones with significant structural similarity to PQS, have been shown to interact with cations (Marshall and Piddock, 1994); thus PQS may function by sequestering positive charge from Mg2+/Ca2+ in the outer membrane. Ionic interactions between PQS and Mg2+/Ca2+ would reduce the ability of these cations to stabilize charge repulsions between LPS molecules. This is not an unreasonable hypothesis as other divalent chelators (EDTA) also enhance MV formation in P. aeruginosa (Eagon and Carson, 1965). This model for PQS-mediated MV formation is also supported by recent observations in our laboratory that P. aeruginosa MV formation is suppressed by the addition of exogenous Mg2+ (L.M. Mashburn and M. Whiteley, unpubl. data). As with previous models, it is likely that PQS-mediated MVs arise from areas of the outer membrane that have few lipoprotein–peptidoglycan linkages. It is interesting to note that P. aeruginosa possesses significantly less lipoprotein than E. coli (Martin et al., 1972) which may provide an explanation for the observation that P. aeruginosa in general produces significantly higher levels of MVs than E. coli (L.M. Mashburn and M. Whiteley, unpubl. data).
Figure 1 outlines the three potential mechanisms of MV formation. Of course these three models are not mutually exclusive and a plausible model could be proposed where these phenomena or combinations of these phenomena collectively promote vesiculation. The basic mechanism of outer membrane blebbing dictates that MV formation will not occur in regions of the outer membrane containing strong outer membrane–peptidoglycan linkages; thus one component that is likely consistent for any MV formation model is the lack of peptidoglycan-associated outer membrane proteins in the blebbing region. Whether or not these areas of the outer membrane simply have reduced levels of peptidoglycan-binding outer membrane proteins that are at that instance not associated tightly with peptidoglycan or if a specific mechanism exists for destroying these linkages is unknown. However, based on the importance of autolysins in MV formation, it is probable that areas undergoing significant peptidoglycan cleavage/reorganization are likely sites for MV formation.
In recent years interest in MVs from various Gram-negative bacteria has significantly increased, and these studies have provided a view of the diverse biological functions of bacterial MVs (Fig. 2). Now that these biological functions are being recognized, mechanistic studies outlining the molecular details of MV formation will provide new insight into MV biology. Although research has primarily focused on distinct aspects of MV formation in P. aeruginosa and E. coli, it is plausible that these bacteria (and Gram-negatives in general) may utilize a common MV formation pathway. Although we anticipate that there will be some species-specific aspects of MV formation, it is likely that the overlying mechanism of MV formation will be similar. With the diverse roles of MVs in biology, it is conceivable that the ability to modulate MV formation by bacteria in natural settings (either environmental or clinical) would provide a novel means of manipulating bacterial group activities.
In addition to studies examining the mechanism of MV formation, understanding how MVs interact with and deliver their cargo to prokaryotic and eukaryotic cells is critical to fully understanding the biological roles of MVs. As MVs are capable of interacting with Gram-positive bacteria, Gram-negative bacteria, and eukaryotic cells, it will be interesting to determine whether the mechanism of cargo delivery to various cell types is similar. Although most MV studies have focused on toxin delivery to eukaryotic cells, we propose that pathogenic and non-pathogenic bacteria have usurped vesicles for a large number of processes, of which toxin delivery is only one component. Now that the diversity of MV biological functions is appreciated, it is important to examine the ubiquity of MV formation in the microbial world. Most studies have focused on the roles of MVs in pathogenic bacteria; however, it is likely that these studies have only scratched the surface for the biological roles of MVs in the natural environment.
We would like to thank members of the Whiteley lab for critical review of this manuscript. M.W. is supported by grants from the Oklahoma Center for the Advancement of Science and Technology, NIH, and the Cystic Fibrosis Foundation.