The role of bacterial outer membrane vesicles for intra- and interspecies delivery

Authors


For correspondence. E-mail mauer@lbl.gov; Tel. (+1) 510 486 7702; Fax (+1) 510 486 6488; E-mail jeberleman@lbl.gov; Tel. (+1) 510 486 7702; Fax (+1) 510 486 6488.

Summary

An increasing number of Gram-negative bacteria have been observed to secrete outer membrane vesicles (OMVs). Many mysteries remain with respect to OMV formation, the regulation of OMV content and mode of targeting and fusion. Bacterial OMVs appear to serve a variety of purposes in intra- and interspecies microbial extracellular activities. OMVs have been shown to mediate cell-to-cell exchange of DNA, protein and small signalling molecules. The impact of such material exchanges on microbial communities and pathogenic processes, including the delivery of toxins at high concentration through OMVs, is discussed. This rather recent aspect of microbial ecology is likely to remain an important area of research as an in-depth understanding of OMVs may allow new approaches for combating bacterial infections and provide new routes for selective drug delivery.

Introduction

In the course of evolution bacteria, just as their eukaryotic competition, have evolved elaborate mechanisms to colonize and survive in their respective niches, be it in an aqueous, terrestrial or in a host organism environment (van Elsas et al., 2011). Bacteria must defend themselves in their niche from the takeover of other species or from immune responses of their host. Encounters between different species likely provoke territorial disputes, with either mutual tolerance, hostile stand-offs or annihilation of one species by its competitor (Vetsigian et al., 2011). Hence one expects the co-evolution of a wealth of measures and countermeasures in interspecies disputes (Ferrer and Zimmer, 2012). Pathogenesis is one such example of a complex host–bacteria interplay, where a bacterial species not only attacks the host tissue, but also evades immune response mechanisms.

Intraspecies communication is an important part of microbial community life (Ng and Bassler, 2009), enabling the coordination of behaviour through cell-to-cell signalling, allowing bacterial biofilms to act as a multicellular organism, either through predation or by turning on virulence factors once a sufficient quorum has been sensed (Berleman and Kirby, 2009; Li and Tian, 2012). Homoserine lactone secretion into the extracellular space has been shown to regulate gene expression across bacterial populations (Schuster and Greenberg, 2006), but the mechanism of extracellular secretion of soluble small molecules has a number of disadvantages, including the fast dilution of the signal, and its possible degradation by secreted enzymes (Boyer and Wisniewski-Dyé, 2009). Therefore the danger of significant material waste is real and open broadcasting of one's presence could alert competing organisms. Outer membrane vesicles (OMVs) are an alternative vehicle for cell–cell signalling in bacteria that have recently gained recognition (Mashburn-Warren et al., 2008; Remis et al., 2010; Whitworth, 2011).

Outer membrane vesicles formed by bacteria can mediate intercellular exchange events including cell–cell signalling, protein and DNA exchange and have even been shown to kill cells (Kadurugamuwa and Beveridge, 1996; Shetty et al., 2011; Evans et al., 2012; J.P. Remis, D. Wei, A. Gorur, J. Haraga, S. Allen, H.E. Witkowska, J.W. Costerton, J.E. Berleman and M. Auer, submitted). While these structures were identified several decades ago (Mayrand and Grenier, 1989), their importance has gained recognition in the past 15 years (Beveridge, 1999; Mashburn-Warren et al., 2008). Membrane vesicles provide a useful vector for vaccines and may elucidate new forms of drug delivery (MacDonald and Beveridge, 2002; Collins, 2011; Sanders and Feavers, 2011).

OMV function

Outer membrane vesicles appear to be a scaffold for a variety of extracellular activities, rather than mediate a single specific process (Mashburn and Whiteley, 2005). In this respect, they are quite distinct from other extracellular organelles, such as Type III needles of Pseudomonas aeruginosa or Yersinia pestis, which transfer specific toxins into a host cell (Büttner, 2012), or the Type IV system in Agrobacterium tumefaciens that functions so efficiently in transferring the T-DNA between bacteria and plant cells (Tomlinson and Fuqua, 2009). So while there is ample evidence for OMVs as a cell–cell transport vehicle, the details of their cargo and the mechanism of biogenesis from the budding of the vesicle from the membrane, to targeting and fusion of the vesicle to the recipient cell is much less understood.

We endeavour here to highlight a few of the systems that have been studied and examine them in terms of the type of molecule exchanged, bound or transferred. While doing so it is important to keep in mind the functions that can be provided by exchanging proteins, DNA, carbohydrate or lipid molecules through OMV. Only DNA is likely to provide a heritable change in the receiving cell and only protein is likely to provide enzymatic function, whereas all of these molecules can potentially act as a cell–cell signal.

Regulation of vesicle biogenesis

There are several reports indicating that vesicles form as a response to membrane stress (Kulp and Kuehn, 2010; Collins, 2011). Examination of a number of stressors in Pseudomonas putida indicates that the vesicle formation response is similar in all cases and leads to increased OM hydrophobicity and cell–cell packing (Baumgarten et al., 2012). In P. aeruginosa, 2-heptyl-3-hydroxy-4-quinolone (PQS) biosynthesis has been shown to stimulate OMV production, through promoting increased membrane curvature (Fig. 2A) (Schertzer and Whiteley, 2012). This indicates that in some instances vesicle production is regulated by factors other than stress. PQS also regulates the production of P. aeruginosa virulence factors as a cell–cell signal and it has the same vesicle-inducing impact on outer membranes when it is supplied exogenously as it does when synthesized by cells (Dubern and Diggle, 2008). Schertzer and Whiteley (2012) propose a model where interactions of PQS with the LPS leaflet of the outer membrane, but not the inner phospholipid leaflet drives the process vesicle biogenesis.

Mechanism(s) of vesicle biogenesis

To date no proteins have been identified to be clearly involved in vesicle formation. Three mechanisms have been proposed for vesicle formation: (i) the loss of outer membrane contact with the underlying peptidoglycan allowing excess lipid formation to induce vesicle formation, (ii) local build-up of periplasmic proteins, literally pushing out the outer membrane, thus trapping large amounts of protein inside the resulting vesicle, and (iii) the involvement of integral membrane proteins or possibly small molecules inside the outer membrane that induce curvature of the membrane and hence vesiculation (reviewed in Kulp and Kuehn, 2010). The first mechanism would imply a rather low protein occupancy of the interior of the vesicles, and suggest that the vesicle protein composition is somewhat similar to the outer membrane, whereas the second mechanism suggests high protein occupancy in the vesicles, and also a somewhat similar protein composition of vesicles compared with outer membrane, whereas the third mechanism would predict a rather low protein occupancy in the interior of the vesicles and an enriched integral membrane protein and membrane-associated composition. The latter mechanism would also suggest that most likely a comparison between different species could yield candidate proteins, as conserved biogenesis proteins would be expected to show up in proteomics analysis of OM vesicles from many different species. Thus far no conserved components can be identified through comparison of various proteomics studies (Kahnt et al., 2010; Elmi et al., 2012; Mendez et al., 2012). In addition to proteomics, better genetic studies are required to help identify if there are any conserved loci for OMV formation genes, such as the magnetosome island responsible for forming intracellular organelles in magnetotactic bacteria (Komeili, 2012).

Ultrastructural tomographic 3D studies of faithfully preserved vesicles in Myxococcus xanthus biofilms revealed that the vesicle lumen was relatively empty (Fig. 1) , with proteins being found relatively sparse and close to the membrane (Palsdottir et al., 2009), suggesting that the second mechanism with its predicted dense array of proteins is unlikely to be correct. One prediction of the third but not necessarily the first mechanism is that vesicles have defined sizes, as any kind of integral membrane protein or vesicular lumen protein is likely to pack best in certain vesicle sizes. Interestingly, geometric analysis of vesicles both in 2D and 3D suggest that while there is not a single size for all vesicles, the size seems to be regulated (Palsdottir et al., 2009, J.P. Remis, D. Wei, A. Gorur, J. Haraga, S. Allen, H.E. Witkowska, J.W. Costerton, J.E. Berleman and M. Auer, submitted). It remains to be determined whether there is a prokaryotic equivalent to the eukaryotic vesicle-inducing proteins such as clathrin, COPI/II and/or caveolins (Zanetti et al., 2012), and thus whether there is a set of proteins that could allow for different vesicle diameters via compositional and/or conformational variability. While it is possible that some molecular machinery is responsible for pinching off vesicles only to stay behind in the outer membrane, one would expect such molecular machineries to be part of the vesicle proteome. Comparative analysis of vesicle proteomics data has not yet yielded a clear candidate, but more mass spectrometry together with subsequent genetic analysis may reveal such candidate proteins. Interestingly it was found that certain proteins are enriched or excluded from the OM, arguing for the third proposed mechanism (Bernardini et al., 2007).

Figure 1.

Ultrastructure of OMVs.

A. Vesicle biogenesis observed at the cell pole in M. xanthus with TEM. Both the inner and outer membrane of the cell are observed, with discrete vesicles and membrane-blebbing captured.

B. Cryo-tomography showing the arrangement of vesicles along the cell membrane in M. xanthus, with the electron dense vesicle exterior surrounding the lower density core. (Images courtesy of Remis & Auer; Palsdottir et al., 2009.)

Vesicles as a vector for DNA exchange

Vesicles produced by some bacteria can harbour DNA fragments derived from the chromosome and plasmids as well as RNA (Yaron et al., 2000; Kesty et al., 2004; Mashburn and Whiteley, 2005). A study of vesicle-packaged DNA from Ruminococcus species resulted in the observation of linear dsDNA varying in size up to 90 kbp (Klieve et al., 2005). Importantly for the ecological role of this genus in rumen microbial communities, transformation of vesicle DNA resulted in the transfer of cellulolytic genes to non-cellulolytic strains of Ruminococcus. Yaron et al. determined the transformation efficiency of enteropathogenic Escherichia coli vesicle-packaged DNA to other E. coli strains to be 103 transformants per μg of DNA. A study examining transformation of environmental vesicle DNA restored growth to an E. coli amino acid auxotroph at a frequency of ∼ 10−5 (Velimirov and Hagemann, 2011). This study also indicated that the upper limit of vesicle DNA size could be as high as 370 kbp.

Thus, the incorporation of DNA into vesicles can occur from three routes. Cytoplasmic DNA can be packaged into vesicles and moved out of the cell through specific transport mechanisms. While the details of such a mechanism have not to our knowledge not been deciphered in any system, there is sufficient evidence to warrant further investigation. However, in addition to questions of transport there is the question of which DNA gets selected for replication and transport. It is worth reminding ourselves that not all organisms replicate through binary fission and if processes such as budding (Jogler et al., 2012) can be a successful mechanism for reproduction, than intermediate mechanisms are likely of use as well. Alternatively, DNA could conceivably end up in vesicles as a result of cell death. Cell lysis is a common occurrence in any bacterial population, and the wide range of genomic fragments that end up packaged in vesicles seems to indicate general nuclease activity. In this scenario, bacterial cell death may play a beneficial role to the community, through some storage of heritable information. The third route is that extracellular DNA may be passively taken up and stored in extracellular vesicles. In this case, there are still many questions remaining on the process of uptake. Is incorporation with the OM a passive process? How does DNA move across the much more tightly regulated cytoplasmic membrane?

In addition to vesicles, membrane ‘nanotubes’ have been reported to mediate the exchange of DNA in between cells of Bacillus subtilis, which does not have an outer membrane (Dubey and Ben-Yehuda, 2011). The authors observed exchange of plasmid DNA and subsequent transfer of antibiotic resistance genes. However, it is unclear how common this mechanism of horizontal gene transfer occurs outside of lab conditions. We presume the rate to be extremely low for two reasons: (i) the DNA exchanged was a very high-copy-number plasmid, and (ii) B. subtilis is one of only a few organisms that displays natural transformation, the nanotubes observed may be a specific structure that mediates a cell–cell variation of natural transformation, rather than a general bacterial structure. Further research is needed to determine if this mechanism is more broadly utilized.

Vesicles as a scaffold for protein activities

While the exact determinants of extracellular vesicle production are unknown, some organisms produce vesicles more often than others. In M. xanthus for instance (Fig. 1A) prolific vesicle production has been observed (Palsdottir et al., 2009). Vesicles were observed with tethers to the OM and to each other in highly organized arrangements (Fig. 1B). In addition, there is evidence for prolific transfer of lipids and lipoproteins between cells of M. xanthus that requires the OmpA-like TraAB proteins (Fig. 2B) (Pathak et al., 2012). Two functions for vesicles are proposed for M. xanthus: (i) as an intraspecies mechanism for mediating signalling and social movement, and (ii) as an interspecies mechanism for exchange of cell–cell killing factors (discussed in the next section). Myxococcus xanthus moves on surfaces by gliding motility (Mauriello et al., 2010). While many mutations cause defects in motility, motility mutants at the tgl and cgl loci are conditional (Rodriguez and Spormann, 1999; Nudleman et al., 2005; Pathak and Wall, 2012). These strains can regain motility when incubated adjacent to strains that produce the missing protein. It was initially suggested that OM from two colliding bacteria would merge and thus transfer proteins (Nudleman et al., 2005); however, it seems more likely that extracellular vesicles fulfil the role of exchange of lipoproteins between neighbouring cells (Palsdottir et al., 2009; Remis et al., 2010). Further investigations of the process of vesiculation and in the existence of membrane bridges between cells are needed.

Figure 2.

OMV–membrane fusion.

A. Interspecies transfer of the quinolone cell signal produced by P. aeruginosa, PQS, alters the membrane curvature of red blood cells (right), relative to untreated cells (left).

B. Intraspecies transfer of lipid dyes (red) between cells of M. xanthus (recipient cells, green) is shown to require the TraAB protein system. (Images adapted from Pathak et al., 2012; Schertzer and Whiteley, 2012.)

Extracellular vesicles have also been implicated in exo-enzyme activity as well as cell-to-cell exchange of proteins. Uptake and cytoplasmic incorporation of GFP-labelled extracellular proteins was observed in the planctomycete, Gemmata obscuriglobus (Fuerst and Sagulenko, 2010). While bacterial endocytosis was suggested as the mechanism, currently there is not enough evidence to support such a claim. In Bartonella henselae, a mammalian pathogen, haem toxicity is mediated in part through the expression and packaging of haem-binding hbpC protein into OMV (Roden et al., 2012). Overexpression of the HbpC protein was shown to cause increased OMV-bound haem in B. henselae. Interestingly, B. henselae requires haem as its sole iron source, yet it is not clear if the haem bound in OMVs is only a protective mechanism or if it is also a transport mechanism.

Vesicles as a vehicle for cell killing and pathogenesis

A number of pathogenic organisms have been shown to produce vesicles and these vesicles may play crucial roles in pathogenesis. Pseudomonas aeruginosa forms biofilms in chronic infections and was also observed in acute burn wounds, where vesicles are observed outside of cells in the extracellular matrix (Fig. 3) (Schaber et al., 2007). Acinetobacter baumannii vesicle fractions were shown to mediate host cell apoptosis, which depends on the presence of the OmpA channel protein presumably damaging the membrane gradient of the host cell (Jin et al., 2011). Cytolethal distending toxin from Aggregatibacter actinomycetemcomitans was shown to depend on vesicle transport into both HeLa cells and human gingival fibroblasts (Rompikuntal et al., 2012). Haemolytic activity was detected from vesicle fractions of the pathogen Kingella kingae, as well as detection of the RtxA toxin in both vesicles and the extracellular fraction (Maldonado et al., 2011). In each case the toxic factor differs, indicating that OMVs can mediate the transmission of a variety of factors.

Figure 3.

Vesicles in pathogenesis. Images of P. aeruginosa in infected mouse tissue viewed with TEM. Numerous bacteria (PaB, P. aeruginosa biofilm) are visible surrounding the blood vessel wall (VW); red blood cells (RBC) are apparent in the vessel lumen. Pseudomonas aeruginosa vesicles are present mingled with extracellular matrix that appears as dark fibres between cells (arrow). Images are at the following magnifications: ×5500 (A), ×16 500 (B), ×68 750 (C) and ×110 000 (D) (Images adapted from Schaber et al., 2007).

In addition to bacteria–host pathogenicity, OMVs have been shown to play a role in cell–cell inhibition and killing among competing bacterial species. There is evidence that P. aeruginosa OMVs disrupt biofilm formation, swarming and iron uptake in P. putida (Fernández-Piñar et al., 2011). Our own results (J.P. Remis, D. Wei, A. Gorur, J. Haraga, S. Allen, H.E. Witkowska, J.W. Costerton, J.E. Berleman and M. Auer, submitted) suggest that vesicles from M. xanthus have a cytotoxic effect on E. coli cells. Similar results were obtained by Evans and colleagues (2012) who observed increased cell killing by addition of fusogenic GAPDH enzyme, indicating that integration of OMVs from M. xanthus into the E. coli OM promotes delivery. This also indicates that the uptake of OMVs may be regulated by production of fusogenic enzymes such as GAPDH. In the cases of intraspecies transfer the exchange of useful proteins would help promote competitiveness, whereas during interspecies exchange of toxic cargo, the production of GAPDH by the host cell would be expected to reduce survival.

Specificity of OMV content

With so much variety observed in the content and function of OMVs, the issue that requires more research is whether bacterial OMVs contain specifically addressed messages dependent on neighbouring species. It is conceivable that M. xanthus cells when among non-competitors focuses on delivering ‘benign’ messages, such as social motility proteins. However, the OMV cargo may change when faced with stress, competition or prey. It seems reasonable that friend–foe interactions may occur simultaneously, within a microbial community such as soils and detritus. To discriminate between messages sent by friend and foe OMVs would require the equivalent of a bar code-like distinction. Such a role could be played either by a unique protein composition (specific receptors) or by altering the precise lipopolysaccharide (LPS) composition. It has been reported in other systems and our own data also suggest that there is exquisite control of what makes it into OM vesicles and it wouldn't be surprising if both mechanisms occur simultaneously. It would be remarkable if one could show that vesicle content is tailored to the specific community situation, e.g. that the vesicle content transitions from friendly neighbourly help to hostile toxin release. How such a transition could be regulated, if it indeed exists, would be indeed a mystery.

OMV–OM recognition and fusion

If OM vesicles are to be effective in material or signal transfer they must interact specifically and finally fuse with their respective target cells. Fine 3D ultrastructural studies of high-pressure frozen, freeze-substituted stained biofilms (Palsdottir et al., 2009) revealed tethers between OM vesicles and the M. xanthus OM that kept the vesicles bound, yet 10–20 nm away from the OM surface. Since carbohydrates are not visible under such heavy metal stain sample preparation conditions, we have employed cryo-EM of M. xanthus whole mount bacteria and found OM vesicles and OM to contain a ∼ 5- to 10-nm-thick layer of what appears to be LPS (J.P. Remis, D. Wei, A. Gorur, J. Haraga, S. Allen, H.E. Witkowska, J.W. Costerton, J.E. Berleman and M. Auer, submitted) that could account for the observed distance, suggesting that LPS could play a role in the OMV–cell OM recognition. However, while specific cell–cell recognition seems plausible via LPS and cell surface proteins, what triggers fusion is much less clear. None of the published proteomics studies has yet revealed a clear candidate for OMV–OM fusion.

OMV as nano-sized drug delivery vehicles

Outer membrane vesicles have been successfully used for vaccination purposes as they are much more safe as an antigen compared with the entire bacterial cell and may contain a sufficient number of cell-surface markers (LPS and proteins) to stimulate an immune response (Bernardini et al., 2007; Collins, 2011; Sanders and Feavers, 2011). Another application of OMVs has received much less attention, namely its potential for cell- and species-specific drug delivery. Given the prevalence of this delivery vehicle in bacterial biofilms, one could utilize this system for the undiluted, high concentration delivery of cell toxins, which would only affect the intended target. If we had a detailed understanding of the target cell specificity of OMVs, we could mimic the design and synthetically produce OMV or even better, use appropriately engineered bacteria themselves to produce large quantities of secreted OMV with a content and specificity of choice.

Concluding remarks

The field of OMVs, like many important subjects in modern biology, has taken its journey from original disbelief of the validity of the presented data to acceptance that bacterial outer membrane are not only real but likely have a central function in microbial physiology and community life. The documented delivery of vesicles for exchange of DNA, proteins and small molecules warrants closer attention. The fact that vesicles in some cases can fill most of the extracellular space in a biofilm (Palsdottir et al., 2009) suggests that they serve an important function, as the production of OMVs surely is energetically very costly. If vesicles are used for intraspecies cell–cell communication, material sharing, and possibly even horizontal gene transfer, they immediately become a high-value target for pharmacological research. Knocking out a species-specific communication system and keeping antibiotics-sensitive strains from becoming resistant via horizontal gene transfer and exchange of genetic material (and/or proteins) would seem highly desirable. Since there is no obvious evolutionary link/protein similarity between prokaryotic and eukaryotic vesicle biogenesis systems, one could argue that effective strategies against bacterial vesicle biogenesis are less likely to have side-effects in eukaryotes. As we continue to make the shift in thinking about human health in terms of the microbial ecosystems, understanding vesicle production and exchange by bacteria is one critical aspect to a better understanding of maintaining healthy microbial communities and preventing pathogenesis.

Acknowledgements

This work is part of ENIGMA, a Scientific Focus Area Program supported by the US Department of Energy, Office of Science, Office of Biological and Environmental Research, Genomics: GTL Foundational Science through contract DE-AC02-05CH11231 between Lawrence Berkeley National Laboratory and the US Department of Energy, and LDRD support by the Director, Office of Science, of the US Department of Energy under contract DE-AC03-76SF00098 to M.A. and J.B.

Ancillary