Bacterial outer membrane vesicles: New insights and applications.

Outer membrane vesicles (OMVs) (∼50-250 nm in diameter) are produced by both pathogenic and nonpathogenic bacteria as a canonical end product of secretion. In this review, we focus on the OMVs produced by gram-negative bacteria. We provide an overview of the OMV structure, various factors regulating their production, and their role in modulating host immune response using a few representative examples. In light of the importance of the diverse cargoes carried by OMVs, we discuss the different modes of their entry into the host cell and advances in the high-throughput detection of these OMVs. A conspicuous application of OMVs lies in the field of vaccination; we discuss its success in immunization against human diseases such as pertussis, meningitis, shigellosis and aqua-farming endangering diseases like edwardsiellosis.


Introduction
Outer membrane vesicles (OMVs) are spherical outer membrane (OM) products of gram-negative (Escherichia coli, Pseudomonas aeruginosa, Pseudomonas gingivalis, Salmonella typhi, Yersinia pestis, Vibrio cholerae and Helicobacter pylori) and gram-positive bacteria, archaea and some eukaryotes such as fungi (Amano et al., 2010;Bauman & Kuehn, 2009;Brown et al., 2015;Horstman & Kuehn, 2000;Ismail et al., 2003;Kulp & Kuehn, 2010;Mirlashari et al., 2001;Schwechheimer et al., 2013). Blebbing of the OM of gram-negative bacteria releasing vesicles was observed from late 1960s (Chatterjee & Das, 1967Work et al., 1966), and electron micrographs from these actively growing bacteria (first observed in V. cholerae) showed OMV formation ranging in size between $40 and 110 nm. It was reported that the formation of OMVs was not dependent on the nature of growth media and cells from both active and stationary phases were able to produce OMVs. However, the functional relevance of OMV formation in the bacterial physiology and pathogenesis were elucidated by later works (Deatherage et al., 2009;Horstman and Kuehn, 2000;Kadurugamuwa & Beveridge, 1995;Li et al., 1998).
In this review, we focus on the OMVs produced by gram-negative bacteria. We provide an overview of the OMV structure, various factors regulating their production and their role in modulating host immune response. In light of the importance of the diverse cargoes carried by OMVs, we discuss the advances in the high-throughput detection of these OMVs and application in vaccine development.
OMVs: genesis from bacterial outer membrane and their structure Outer membrane vesicles are derived from bacterial outer membranes entrapping the periplasmic content along with fragments of PG and nucleic acids occasionally. The bacterial OM is unique for its asymmetrical distribution of lipid moieties, wherein the exterior leaflet is enriched in lipopolysaccharides (LPS), and the inner leaflet predominantly contains glycerophospholipid (see Figure 1(a)) (Roier et al., 2016;Schwechheimer & Kuehn, 2015). Endotoxin or lipid-A, a component of LPS, decorates the OM and is responsible for the immune modulation during pathogenic invasions (Raetz & Whitfield, 2002). During the vesicle formation, the asymmetry of the OM is retained in OMVs. The spherical structure of OMVs provides a higher surfaceto-volume ratio, therefore allowing the residence of LPS in high density. LPS plays a major role in hostpathogen interplay and modulation of host immunity.
Lipopolysaccharides is synthesized on the inner membrane (IM) by the "Raetz-pathway" which is a nine-step process (Raetz & Whitfield, 2002;Whitfield & Trent, 2014). The enzymes involved in LPS synthesis extend between the IM and OM of the bacterial cell wall. LPS synthesis is initiated by a group of Lpx enzymes, present on the cytoplasmic face of the IM by the formation of a precursor called kdo2-lipid A (see Figure 1(b)). Kdo2-lipid A is modified by the addition of core-oligosaccharide (genesis of LPS), and this is followed by flipping to the periplasmic face. The major population of lipid A moieties are further modified by ligation of O-antigen units of various lengths (see Figure 1(b)). Finally, the LPS moiety is exported to the OM [reviewed in Whitfield & Trent (2014), Raetz & Whitfield (2002)]. Out of these steps, the ligation of Oantigen confers variability to the structure of LPS (see Figure 1(b)), and these differential modifications cater to the need of the bacterium to adapt to its local environment (Kulp & Kuehn, 2010;Sohlenkamp & Geiger, 2016). For example, OMVs produced by P. aeruginosa in biofilms are different in their composition with respect to both lipids and protein (Metruccio et al., 2016). LPS synthesis is conserved in gram-negative bacteria, and there are no enzyme homologs in eukaryotes. Complete lack of such pathways in eukaryotes renders various steps in the LPS synthesis pathway as a potent drug target. Figure 2 shows the schematic representation of the structure of a typical OMV (Figure 2(a)) and its formation from bacterial OM (Figure 2(b)). The composition of the lipid bilayer of OMVs retains the asymmetry observed in bacterial OM, with complex lipids in the outermost layer mounted on a uniform glycerophospholipid inner layer. Recently, a few examples of double bilayer OMVs have been observed in Bacillus anthracis (Rivera et al., 2010), Shewanella vesiculosa M 7T (Perez-Cruz et al., 2013), Ahrensia kielensis, and Pseudoalteromonas marina (Hagemann et al., 2013) whose detailed discussion is not in the scope of this review.
Outer membrane vesicles are produced by the expansion and bulging of the OM followed by pinching of the OM to form spherical vesicles. Different inducers of OMV production are discussed in the next section.

Regulatory mechanisms of OMV formation
Various mechanisms of production of OMVs have been proposed in the last four decades (Schwechheimer & Kuehn, 2015). Table 1 provides an overview of the different mechanisms involved in OMV production. Production of OMVs is regulated at different physiological stages of bacterial cell growth. Membrane proteins like Lpp and OmpA, located in the periplasmic space (see Figure 1(a)), span the gap between the OM and PG. Loss of association between OM and PG by a reduction in synthesis of Lpp or OmpA, or alteration in PG structure (reducing the binding of these proteins to PG layer), enhances OMV formation (Moon et al., 2012;Schwechheimer & Kuehn, 2015;Schwechheimer et al., 2014). This suggests that the function of these proteins as crosslinkers between OM and PG restrict the initiation of bulging of OM to produce OMVs (Moon et al., 2012; Figure 1. (A) Structure of the gram-negative bacterial cell envelope. The envelope consists of two lipid bilayers and a peptidoglycan layer. The bacterial outer membrane (OM) is unique for its asymmetrical distribution of lipid moieties, wherein the exterior leaflet is enriched in lipopolysaccharides (LPS), and the inner leaflet predominantly contains glycerophospholipid. The inner membrane (IM) is symmetrical in the glycerophospholipid composition in both the leaflets and contains many complex protein assemblies like the YrbDE, Tol component of the Tol-Pal system and NlpA. The region between IM and OM is called as the periplasmic space (PS). The periplasm contains the peptidoglycan (PG) layer along with other protein assemblies such as VacJ/YrbC, Tol B and Pal component of the Tol-Pal system, and Lipoma preferred partner (Lpp). The rest of the Yrb complex components (YrbBF) are juxta-membrane from the cytoplasmic face of IM. (B) The building blocks of the LPS layer. An LPS moiety is composed of the inner and outer core oligosaccharides followed by variable length O-antigens covalently linked to the outer-core. The inner core an LPS moiety contains L-glycerol-D-manno-heptopyranose (L-D-Hep) and Kdo as the major components while the outer core contains D-glucose as the major component linked with L-D-Hep and D-galactose monomers (see text for more details).
Schwechheimer & Kuehn, 2015;Schwechheimer et al., 2014). Interestingly, an IM lipoprotein gene nlpA (see Figure 1) deletion was found to reduce OMV production. However, the deletion of ompA or lpp genes enhanced the production of OMVs (Schwechheimer & Kuehn, 2015;Schwechheimer et al., 2014). Stress inducing factors like osmotic-shock brought about by high saline conditions, dehydration caused by long-chain alcohol, heat-shock (Baumgarten et al., 2012;McBroom & Kuehn, 2007), antibiotic stress (Maredia et al., 2012) and accumulation of misfolded proteins or components of PG (Hayashi et al., 2002;Nakae, 1975) are potent inducers of OMV production. Short fragments of PG are produced from interrupted biosynthesis or imbalance between synthesis and degradation (Hayashi et al., 2002). Local accumulation  Figure 1 for details of bacterial OM). OMVs host a large repertoire of cargoes such as nucleic acids, enzymes, toxins, fragments of peptidoglycans, and fluid from the periplasmic space. (B) Formation of OMV starts with the bulging of the outer membrane at very small region followed by progress in the expansion of the membrane. The increased expansion is established with the spherical structure of blebs. Eventually vesicle pinches off at the joint neck and gets separated from the cell.
of these unwanted products interacts with the membrane and causes bulging that eventually turns into vesicles. Stress induced vesiculation provides benefits to the resident bacteria. For example, hyper-vesiculating E. coli mutants survive better against ethanol stress or the OM-damaging antimicrobial peptide polymyxin B (McBroom & Kuehn, 2007). Another proposed function of OMV production is to shed overproduced LPS load to maintain the membrane integrity (Bonnington & Kuehn, 2016), which has been reviewed earlier (Schwechheimer & Kuehn, 2015).
A gram-negative facultative anaerobe, Shewanella oneidensis, shows hyper-vesiculation upon deletion of the degQ gene (Ojima et al., 2015). The degQ gene product is a periplasmic protease involved in the degradation of periplasmic proteins. Deletion of degQ gene leads to protein accumulation in the periplasmic space, which in turn enhances OMV formation.
The Tol-Pal system (see Figure 1(a)) is an IM protein complex involved in the uptake of iron siderophores, type-A colicins and single stranded DNA phages (Letain & Postle, 1997). The Tol-Pal system consists of 5 proteins TolA, TolB, TolR, TolQ, and Pal. The tolA, tolR and tolQ genes are part of an operon, and their deletion causes a significant increase in vesiculation (Bernadac et al., 1998). This suggests the important role of TolA, TolQ, and TolR in regulating OMV formation. In the case of Serratia marcescens, a gramnegative opportunistic pathogen, the production of OMV is temperature-regulated. Higher amount of OMV is produced at 22 or 30 C whereas reduction in production is observed at 37 C (McMahon et al., 2012).
Cells growing at a lower temperature ($22 or 30 C) produce higher levels of OMVs compared to the cells grown at 37 C. Enhancement in OMV production under thermal stress was due to reduced production of Enterobacterial common antigen (ECA), a glycerophospholipid found in the outer leaflet of OM. This reduction in the ECA was linked to RcsB phosphorelay signaling. Thermal stress causes RcsB signaling activation and reduced transcription of the ECA gene. Eventually, reduced membrane integrity leads to increased OMV production (McMahon et al., 2012).
Another proposed mechanism for inducing OMV production involves the phospholipid transporter VacJ-Yrb ABC (ATP-binding cassette) system located in the bacterial IM (see Figure 1(a)). Low levels of VacJ-Yrb components were shown to increase OMV production by increased accumulation of curvature-inducing phospholipids in the bacterial OM (Roier et al., 2016).
Lipid-A modifications play an important role in OMV production. The enzyme lipid-A deacylase called PagL modifies lipid-A by deacylation. The accumulation of deacylated lipid-A on the outer leaflet creates positive curvature leading to increased vesiculation (Elhenawy et al., 2016).

High-throughput detection of OMVs
Outer membrane vesicles were first discovered and characterized by transmission-electron microscopy (Chatterjee & Das, 1967Work et al., 1966). In the past, OMVs were purified by filtration, ultracentrifugation, and characterized by their protein or lipid  (Elhenawy et al., 2016;Gaillard et al., 2014) content. In these processes, the quantitative detection of OMV size and number was difficult to achieve with high precision. In view of the knowledge gathered from proteomics analysis of OMVs produced by a large group of bacteria and their relevance in immune modulation (see later), new methodologies of detection and characterization of OMVs are required. Recently, a novel method has been used for determination of the size and concentration of extracellular vesicles using tunable resistive pulse sensing (Vogel et al., 2016) which employs extrapolation of scanning ion occlusion spectroscopy (SIOS) which is based on the property of scalable membrane pores (i.e. screening by stretchability of membrane pores of different sizes) to analyze nano-sized particles (Roberts et al., 2010;Sowerby et al., 2007) (see Figure 3). As the particles of different size pass through the membrane pore, it induces changes in the current across the membrane by interfering with the open pore conductivity, which can be detected with high sensitivity and accuracy. The stretchable pores allow detection of a selective range of nano-sized OMVs. The data generated from the measurements can be used for high-throughput determination of vesicle size and number.

Mechanism of OMV entry into host cells
Outer membrane vesicles from various mucosal pathogens are majorly internalized by epithelial cells in the host, and this entry is mainly driven by two endocytic pathways: clathrin-mediated endocytosis, and the lipid raft-mediated pathway (Canas et al., 2016). Figure 4 summarizes the different pathways of entry of OMVs into the host cell. Vesicles from enterohemorrhagic E. coli, H. pylori, A. actinomycetemcomitans and Brucella abortus, enter host cells predominantly via clathrinmediated endocytosis (Bielaszewska et al., 2013;Canas et al., 2016;Cheng et al., 2014;O'Donoghue & Krachler, 2016;Olofsson et al., 2014). Along with clathrin, dynamin has also been shown to play an important role in OMV internalization in the case of E. coli O 104 :H 4 entry (Bielaszewska et al., 2017;Kunsmann et al., 2015). Lipid rafts are dynamic membrane microdomains rich in lipids like cholesterol and sphingolipids, and proteins such as caveolin. Vesicles from ETEC. coli, H. influenza (Kaparakis-Liaskos & Ferrero, 2015), P. gingivalis (Furuta et al., 2009), Moraxella catharralis (Schaar et al., 2011;Vidakovics et al., 2010), C. jejuni (Elmi et al., 2012) and P. aeruginosa are internalized through a lipid raft-mediated pathway (Bomberger et al., 2009;Furuta et al., 2009;. Lipid rafts enriched in G M1 act as molecular beacons for adhering to OMVs from ETEC and the bound complexes are internalized via lipid raft-mediated endocytosis Johnson et al., 2009;Horstman & Kuehn, 2002;Kesty & Kuehn, 2004). Bulk uptake of OMVs has been proposed through actin-driven, ruffled protrusions from the cell membrane (Kaparakis-Liaskos & Ferrero, 2015;Kunsmann et al., 2015). Inhibition of actin polymerization by cytochalasin D or wiskostatin has been observed to reduce the entry of OMVs from P. aeruginosa into the epithelial cells lining the airway tract (Bomberger et al., 2009). Interestingly, a host factor called bactericidal/ permeability-increasing protein (BPI) present on the surface of epithelial cells helps in regulating the internalization of OMV, e.g. in the case of N. meningitides OMVs, expression of BPI enhances the entry of OMVs into the host cells (Elsbach & Weiss, 1993).

Host immune-modulation by OMVs and its potential application as vaccine
Outer membrane vesicles are known to carry cargoes such as endotoxins and RNA (discussed above), and these engage in dynamic interplay with the host immune system (Kuehn & Kesty, 2005). We discuss the host immune response of ETEC, Legionella pneumophila, N. meningitides and P. aeruginosa as representative examples in this review. Enterotoxigenic E. coli colonizes the small intestine, where they produce heat-labile (LT, thermolabile) and/ or heat-stable enterotoxins (ST, thermostable) Roy et al., 2011;Taxt et al., 2010). LT loaded OMVs are released into the intestine where they bind to the G M1 ganglioside receptors located on the host intestinal cell (Johnson et al., 2009;Spangler, 1992). This binding triggers the internalization of the OMV with toxin into the cell cytoplasm. LT acts through the ADP-ribosylation of the Gs subunit of adenylate cyclase, which results in an increase in cAMP levels. cAMP then activates protein kinase A (PKA), which phosphorylates the R domain of cystic fibrosis transmembrane conductance regulator (CFTR). CFTR is an essential component of epithelial Cl À transport systems. Phosphorylation of CFTR manifests as an increase in chloride and water efflux into the intestinal lumen leading to watery diarrhea. The pro-inflammatory response of LT-associated OMV purified from ETEC showed a significant immunogenic response. However, the response was slower compared to soluble LT (Chutkan & Kuehn, 2011). OMVs uptake by caveolin pathways and LT internalization by G M1 may lead to differences in the efficiency of the intracellular processing of LT and therefore the difference in response (Chutkan & Kuehn, 2011).
Legionella pneumophila¸a causative agent of Legionnaire's pneumonia, is a human-and amoebaeresident gram-negative bacterial pathogen. In humans, it infects alveolar macrophages (Jager et al., 2015) and propagates within the phagosome (Fields et al., 2002) by preventing phagosome maturation and fusion with the lysosomes (Fernandez-Moreira et al., 2006). L. pneumophila inhibits phagosome-lysosome fusion by secreting effector proteins into the cytoplasm via the type IV secretion system. This is followed by further modifications which transform the resident phagosome compartment into a vacuole, thereby enabling a protective replication niche (Ge and Shao, 2011;Hubber and Roy, 2010;Isberg et al., 2009). Recent reports show that OMVs produced from L. pneumophila have pro-inflammatory action on macrophages and epithelial cells (Galka et al., 2008;Jager et al., 2015). This is supported by the proteomics analysis of L. pneumophila OMVs which revealed many virulence factors among the 70 OMV-associated proteins identified (Jager et al., 2015).
Neisseria meningitides, another gram-negative pathogen, causes a severe human disease called meningitis. This infection compromises the protective membranes covering the brain and spinal cord leading to symptoms such as stiff neck, high fever, sensitivity to light, confusion, headaches, and vomiting (WHO, 2017). N. meningitides is an extracellular pathogen which is capable of inducing pyroptosis, a process that requires LPS interaction with caspase-11. Interestingly, the LPS released from OMVs encapsulated in endosomes interacts with caspase-11 in the cytosol. This interaction activates caspase-11 followed by acute inflammatory response and eventually sepsis (Vanaja et al., 2016).
Recently, miRNA-sized small RNAs (msRNAs) have been identified in OMVs secreted by P. aeruginosa. The msRNAs are expressed from bacterial cells and are packed into OMVs. Interestingly, these msRNAs could regulate the host gene expression and thereby modulate host immune response (Koeppen et al., 2016). This study has motivated identification of msRNA targets in eukaryotic genomes using bioinformatics tools (Choi et al., 2017).
Immunogenicity of OMVs has been a subject of an experimental vaccine and has culminated in many promising results against a number of bacterial infections (Acevedo et al., 2014;Asensio et al., 2011;Gaillard et al., 2014;Leitner et al., 2015;Mitra et al., 2016;Norheim et al., 2015;Park et al., 2011;Price et al., 2016;Roberts et al., 2008;van der Pol et al., 2015). Small size, immunogenic properties, and nonreplicative behavior of OMVs make them a good choice for vaccination. OMVs have been tested for their ability to enhance host immunity and provide protection against pathogens. Recently, OMVs purified from a genetically engineered nonpathogenic strain expressing desired glycans of interest, were used for vaccination of mice and chickens to confer protection against Streptococcus pneumonae and C. jejuni infection, respectively (Price et al., 2016). Higher yield ($60%) of bioengineered OMVs has been achieved by deletion of the tolA gene (part of the Tol-Pal system described earlier) in E. coli compared to its wild type strain (Bernadac et al., 1998). Similar enhancement in OMV production from Shigella boydii, a dysentery causative pathogen, has been observed by deletion of the tolA gene. Interestingly, immunization by OMVs purified from DtolA S. boydii provided enhanced protection against lethal dosage of S. boydii than offered by wild-type OMVs (Mitra et al., 2016).
A promising use of OMVs as a vaccine has been reported against pertussis (whooping cough), caused by bacterial pathogen Bordetella pertussis, which has a record of causing the deaths of $195,000 children annually (WHO, 2017). Intranasal immunization of mice with OMVs secreted by engineered B. bronchiseptica expressing the pagL gene (involved in OM modification, see Table 1) to enhance OMV production, displayed reduced toxicity with a lower induction of proinflammatory cytokines. Exposure to engineered OMVs conferred around nine months of protection against Bordetella infection with a higher clearance of various B. pertussis isolates (Asensio et al., 2011;Gaillard et al., 2014).
Immunization by OMVs against Edwardsiella tarda infection holds great economic significance in the light of the loss caused by this infection in aquaculture globally. E. tarda often infects fishes, eels, and flounder causing Edwardsiellosis. The infection with E. tarda leads to exophthalmia, the loss of pigmentation, swelling of the abdominal surface, petechial hemorrhage on skin and fin, and rectal hernia. In addition, suppurative hepatitis, suppurative interstitial nephritis, and inflammation in the spleen are the primary histopathological diagnosis (Park et al., 2012). E. tarda infection can be transmitted through fecal contamination to amphibians, reptiles, and mammals (including humans) causing gastroenteritis and diarrhea (Vandepitte et al., 1983). Earlier studies have shown that E. tarda OM proteins can induce an immune response and are effective in the development of protection against E. tarda infection (Castro et al., 2008;Kawai et al., 2004;Kwon et al., 2006). This potential was realized in providing immunization to olive flounder using OMVs purified from E. tarda (Park et al., 2011).

Conclusions
Our understanding of OMVs has come a long way since their discovery in the $1960s. The urgency to understand the molecular details of OMV structure and function comes from the emerging dovetailed relationship between OMVs and biofilms (Wang et al., 2015). The merit for the urgency derives from the mammoth danger posed by microbial biofilms toward equipment damage, product contamination, energy losses and chronic infections (Wolcott et al., 2010;Yang et al., 2012). In this review, we summarize the genesis of OMVs from the host bacterial cell wall and present an overview of the different cues and genetic factors that induce OMV secretion. We discuss a new high-throughput technique to detect, quantitate and characterize the OMVs. We provide a brief outline of different pathways of entry of OMV into the host cells followed by host immune modulation by internalized OMVs, taking a few representative examples. OMVs have been clinically isolated from human blood and other external secretions like saliva and urine (Cheng et al., 2014). OMVs' ability to elicit an immunogenic response has been utilized for vaccine development by engineering pathogenic bacteria to secrete OMVs with desired immunogenic properties. Taken together, exploration of the different aspects of OMVs is a rapidly growing field, and we present some of the recent insights and advances in our understanding of OMVs along with its application in the area of vaccine development.