The primary advantage of using OMVs as antibiotic delivery vehicles is their inherent ability to deliver their cargo across the cell envelope of Gram-negative bacteria. With a cell envelope that consists of two membranes, Gram-negative bacteria are inherently resistant to many antibiotics (2). Several reports of the OMV-mediated delivery of active proteins or genes across the Gram-negative cell envelope highlight the potential utility of OMVs to enhance antibiotic delivery to these cells. Although detailed mechanisms of this process remain elusive, future work to better understand this delivery processes will further enhance research into the use of OMVs as antibiotic delivery vehicles to overcome current transport limitations.

The first reports of OMV-mediated transport across the Gram-negative cell membrane focused on the delivery of peptidoglycan-degrading hydrolases (19, 22). These “predatory” OMVs were hypothesized to fuse with the target cell’s outer membrane, delivering their enzyme cargo to the periplasm of the target cells (22). This hypothesis was supported by subsequent experiments demonstrating that components of Shigella flexneri and Pseudomonas aeruginosa OMVs are incorporated into the membranes of other bacterial cells (Salmonella typhi, Salmonella typhimurium, and Escherichia coli) (18). More recently, myxobacteria, small Gram-negative soil-dwelling bacteria (28), have been found to produce OMVs encapsulating hydrolytic enzymes (29, 30), which exhibited lytic activity against E. coli (14, 31). Together, these findings demonstrate that OMVs are able to deliver cargo to the periplasm of certain bacteria ( Figure 2A ).

In addition to the delivery of intact, functional proteins, OMVs have been observed to facilitate delivery of DNA to bacterial cells ( Figure 2B ). The first evidence of DNA in OMVs was uncovered in 1995, when Kadurugamuwa and Beveridge observed its presence in the OMVs produced by two strains of P. aeruginosa (H103 and ATCC 19660) (32). Kolling and Matthews later demonstrated the presence of DNA in OMVs produced by E. coli O157:H7. They observed that these OMVs contained DNA encoding certain virulence genes, including stx1 and stx2, which encode for Shiga toxins 1 and 2, respectively. These genes were delivered to noncompetent recipient cells, E. coli JM109 (33). Yaron et al. later demonstrated that E. coli O157:H7 OMVs transfer DNA to E. coli JM109 and Salmonella cells, and the recipient cells were shown to express the virulence proteins encoded in the genes (34).

The role of OMVs in the horizontal transfer of antibiotic resistance genes has also been demonstrated. Rumbo et al. observed that two clinical strains of Acinetobacter baumannii that were resistant to carbapenem released OMVs containing the blaOXA-24 gene, which encodes for a β-lactamase. When these blaOXA-24-containing OMVs were incubated with a carbapenem-susceptible strain of A. baumannii, resistance to several β-lactam drugs was observed. Importantly, this previously susceptible strain was subsequently found to express blaOXA-24 and to release blaOXA-24-containing OMVs (35). Similarly, OMV-mediated horizontal gene transfer has also been identified in the oral pathogen Porphyromonas gingivalis (36), E. coli O104:H4 (37), and S. typhi (38).

Delivery of functional genes and the important role of OMVs in horizontal gene transfer indicates that the OMVs enable delivery of their DNA cargo into the bacterial cytosol. This process was first visualized by Fulsunder et al. by using immunogold labeling of double-stranded DNA to observe movement of DNA from the donor bacterial cells (Acinetobacter bayli, JV26) into OMVs and then recipient cells (both E. coli DH5α and A. bayli JV26) (15).

Together, these observations of OMV-mediated protein and DNA delivery demonstrate that OMVs are able to deliver functional cargo across the Gram-negative cell envelope. This behavior is particularly appealing for the delivery of antibiotics as it suggests that encapsulation of the drugs within OMVs might decrease the transport issues that limit the efficacy of many antibiotics against Gram-negative bacteria.

With their membrane structure, OMVs have a unique property of being able to transport both hydrophilic and hydrophobic molecules simultaneously. This property has been demonstrated in the natural delivery processes of OMVs and holds great importance in the future development of OMVs for antibiotic delivery.

Quorum sensing is the process by which bacteria sense cell population density and as a result, alter gene expression. In bacteria, this process occurs through the release of certain molecules; as cell density increases, the concentration of these quorum sensing molecules increases correspondingly, thus serving as a signal of high population density (39). N-acyl-homoserine lactones (AHLs) are the most common quorum sensing molecules employed by Gram-negative bacteria (40). A long-standing question in the quorum sensing field was how hydrophobic, long-chain containing AHLs were delivered through the aqueous extracellular environment. Toyofuku demonstrated that N-hexadecanoyl-L-homoserine lactone (C16-HSL, Figure 2C ), produced by Paracoccus denitrificans, is packaged into outer membrane vesicles to promote solubility of the molecule (41). Similarly, CAI-1, a long-chain ketone QS molecule ( Figure 2C ), was observed to be released in association with OMVs produced by Vibrio harveyi. The OMV-associated CAI-1 was able to be delivered in an active form to non-CAI-1-producing cells, including Vibrio cholerae (42).

P. aeruginosa produces several quorum sensing molecules, including 2-heptyl-3-hydroxy-4-quinolone (Pseudomonas quinolone signal, PQS). Mashburn and Whiteley demonstrated that a majority of the produced PQS was released in association with OMVs, while less hydrophobic signaling molecules were not. Interestingly, the authors observed that the PQS molecule itself promotes OMV formation (43). Subsequent work by this group found that PQS intercalates into the outer membrane to induce membrane curvature, thereby promoting OMV formation (42, 44, 45) ( Figure 2D ).

Recently, Choi et al. demonstrated that Chromobacterium violaceum delivers the hydrophobic molecule, violacein, to bacterial cells by packaging it in OMVs (46). This process appeared to be regulated, as the OMVs were found to contain more violacein than protein. The OMV-encapsulated violacein retained its activity against the Gram positive organism, Staphylococcus aureus (46).

Thus, the natural ability of OMVs to solubilize hydrophobic molecules could enable improved delivery of lipophilic antibiotics, which often exhibit low bioavailability due to poor solubility, limited absorption, and rapid metabolism (47).

OMVs have been observed to naturally interact with other bacterial cells, both from the same and different species. Selective delivery of their cargo has been observed, but the processes mediating this phenomenon remain unclear.

Tashiro et al. used a classical colloidal science theory, the Derjaguin-Landau-Verwey-Overbeek (DLVO) theory in an attempt to explain the interaction of certain OMVs with specific bacterial cells. This group observed that OMVs produced by Buttiauxella agrestis selectively associated with B. agrestis cells, enabling delivery of plasmid DNA and gentamicin specifically to B. agrestis cells. Because this selective association of OMVs with the bacterial cells did not require the cells to be viable, the authors hypothesized that interaction energies, as calculated using the DLVO theory, might explain this behavior. In this theory, the interaction energy is defined as the sum of the attractive London-van der Waals forces, which depend on OMV radius, and the repulsive electric force, which is a function of the surface charge (zeta potential) of the OMV. The authors observed a correlation between interaction energies and OMV association, which was not entirely linear; therefore, they proposed that this interaction energy is only one factor that regulates OMV specificity for certain bacterial cells, and they hypothesized that surface proteins on both the OMV and bacterial cell surface likely play an additional role in this specific delivery process (48).

Tran and Boedicker investigated whether OMV-mediated DNA transfer is regulated by the relatedness of the OMV donor and recipient cells. The authors observed that E. coli is able to encapsulate plasmids with different replication origins within its OMVs and deliver this genetic cargo to recipient cells. Aeromonas veronii and Enterobacter cloacae exhibited similar behavior. The rate of gene transfer between the three types of OMVs and five types of cells: E. coli, A. veronii, E. cloacae, C. violaceum, and P. aeruginosa was studied, but no relation between the rates of uptake and the relatedness of the donor and recipient cells was observed (49). The authors did observe that the rates of delivery differed depending on the origin of the OMVs, with A. veronii OMVs being the most efficient (49).

Recently, some evidence of the involvement of an OMV surface protein in selective delivery of OMVs was reported. Agrobacterium tumefaciens is a phytopathogen that releases OMVs containing a small lipoprotein called Atu8019. The authors of this study observed that OMVs produced by a ΔAtu8019 deletion mutant were similar in properties to the OMVs released by wildtype cells; however, OMVs from the deletion mutant exhibited an inhibited propensity for cell association, suggesting a role for this protein in selective OMV delivery (50).

The naturally targeted delivery of OMVs to specific bacterial cells holds exciting promise in their development for drug delivery. However, the details of this process have yet to be elucidated. Future research to identify the biological determinants enabling this specificity will enhance the design of targeted delivery systems, both natural and synthetic, for improved antibiotic function.

A final advantage of using OMVs for antibiotic delivery is their extreme stability and their ability to protect their luminal content from enzymatic degradation, thus promoting long-distance delivery.

The inherent ability of OMVs to protect their cargo from degradation has been widely reported, particularly in the transfer of β-lactamases between bacteria. This process has been observed in a number of organisms, including A. baumanii, Moraxella catarrhalis, Stenotrophomonas maltophilia, E. coli, and P. aeruginosa (51–55). The luminal location of these antibiotic resistance enzymes has been demonstrated to protect the proteins from enzymatic degradation (54) as well as serum IgG-mediated neutralization (56). In addition, transfer of various protein toxins via OMVs has been reported to protect them from enzymatic degradation (12, 57–60), which has been hypothesized to enable long-distance delivery in vivo (12).

OMVs are also able to protect their nucleic acid cargo from enzymatic degradation. Koeppen et al. observed that inclusion of RNA within the OMV lumen protected it from RNase digestion (61). Similarly, OMV-encapsulated genes were protected from DNase digestion (34, 35, 62).

In addition to protecting their cargo from enzymatic degradation, OMVs appear to protect their cargo from degradation due to handling and storage. In a systematic study of the stability of OMV-encapsulated cargo, Alves et al. packaged an enzyme, phosphotriesterase (PTE) into the lumen of E. coli OMVs. The authors observed increased stability of the protein cargo relative to free PTE against multiple freeze-thaw cycles (63). Later work by this group demonstrated that encapsulation within OMVs protected long-term enzyme activity under multiple storage conditions, including freezing, heating, and lyophilization (64).

Together these observations demonstrate that encapsulation within the OMV lumen is able to protect the cargo from degradation, both in vivo and during storage. This property is likely to enhance the activity of encapsulated antibiotics, enabling delivery of reduced dosages.

OMVs are naturally produced throughout bacterial growth; however, the resulting yields are too low for biotechnological applications. To overcome this challenge, many groups have looked towards genetic modifications that result in increased vesiculation. Mutations in the Tol-Pal system have been particularly appealing for this purpose. The tol-pal operon consists of seven genes, including the five genes comprising the Tol-Pal system: tolQ, tolR, tolA, tolB, and pal. These genes express proteins that together, form a complex linking the inner membrane, peptidoglycan, and outer membrane (71) ( Figure 3A ). Mutation of any one of the genes results in increased vesiculation in E. coli (72, 73). A similar approach that included mutation of the tolR gene along with the galU gene, which is involved in LPS biogenesis, was found to increase vesiculation in Shigella sonnei (74). In Helicobacter pylori, deletion of tolB but not pal increased vesiculation (75).

Building on the finding of the importance of the Tol-Pal system in vesiculation, Henry et al. demonstrated that genetic modifications to promote production of certain protein domains that interact with elements of the Tol-Pal system can likewise increase OMV production. Specifically, they found that periplasmic production of a TolR domain induced a high level of vesiculation in E. coli. Additionally, periplasmic production of the translocation domain of colicin A, colicin E3, and minor coat protein g3p also increased vesiculation. Finally, the authors demonstrated that the approach could be used in other bacteria, including Shigella flexneri and Salmonella enterica (76).

These multiple genetic approaches to increase OMV production have already enabled the use of OMVs for biotechnological purposes, particularly as vaccines and are likely to enable future development of OMVs as natural antibiotic delivery vehicles.

Another important limitation in the use of OMVs as drug delivery vehicles is their inflammatory nature. Derived from the outer membrane, the surface of OMVs is primarily composed of lipopolysaccharide (LPS). LPS consists of a hydrophobic lipid A molecule, which is tethered to the membrane via six acyl chains, core oligosaccharides, and the O-antigen (77) ( Figure 3B ). The lipid A portion, also called endotoxin, is responsible for the inflammatory response induced by LPS (78). Thus, to use OMVs as drug delivery vehicles, it is imperative that the toxicity of the LPS be reduced.

One strategy to reduce the toxicity of LPS is to genetically modify the genes leading to full acylation of the lipid A moiety. Nine enzymes are required for the biosynthesis of lipid A (77) and knockout of certain genes encoding these enzymes, in particular lpxL and lpxM (also known as msbB) have been demonstrated to result in under-acylated strains with reduced endototoxicity (79–81).

Alternatively, modification of the phosphorylation of the lipid A moiety can be an effective strategy to reduce endotoxicity of LPS. Lipid A is usually diphosphorylated (82). Edgar Ribi observed that monophosphorylated lipid A is significantly less immunogenic (83), and since then, monophosphoryl lipid A (MPL) has been FDA-approved as an adjuvant (84). A promising strategy to create OMVs consisting of monophosphorylated lipid A is to express the Helicobacter pylori Hp0021 gene in the OMV-producing organism. This enzyme removes the 1-phosphate of the lipid A moiety, resulting in monophosphorylated lipid A (85). When this gene was heterologously expressed in E. coli, the 1-phosphate group of lipid A was removed (86). While promising, this approach has not yet been used to develop OMVs for biotechnological purposes.

While LPS toxicity remains a serious concern with using OMVs as drug delivery vehicles, multiple promising approaches have demonstrated the potential to reduce the inflammatory response, thus enabling their future use as antibiotic delivery vehicles.

While certain OMVs naturally possess some antibiotic properties, additional work has focused on encapsulating specific molecules within the OMV lumen to expand on the potential of OMVs as delivery vehicles. Several methods to encapsulate different molecules, including antibiotics, have been explored, as described below. Further optimization of these approaches will enable full realization of the potential of OMVs as drug delivery vehicles.

In order to enhance targeting or functionality of OMVs, several approaches to display particular moieties on the vesicle surface have been employed. These genetic approaches involve the development of fusions between the desired protein and certain surface-localized proteins (42, 100–102), or the use of the SpyCatcher/Tag or SnoopCatcher/Tag systems to promote isopeptide bond formation between the desired protein and a surface-localized protein (103, 104) ( Figure 5 ). These approaches have been primarily developed for vaccine technology; however, targeted delivery of antibiotics specifically to pathogenic cells could likely be accomplished using similar approaches.

The ClyA toxin expressed by many E. coli strains and enriched in OMVs (27) has been commonly employed as a fusion partner to localize specific proteins to the OMV surface. Kim et al. first demonstrated the power of this approach by creating a series of chimeric ClyA fusion proteins, using green fluorescent protein (GFP), β-lactamase, β-galactosidase, organophosphorous hydrolase (OPH), and a single chain Fv antibody fragment (100). The authors observed that each fusion partner retained its activity and was located on the surface of the vesicles (100). In engineering their siRNA-containing OMVs, Gujrati et al. genetically fused a targeting affibody to the ClyA protein and demonstrated that this approach enabled targeted delivery of the siRNA-containing OMVs to HER2-expressing cells (94).

Another fusion protein approach that has been shown to be successful takes advantage of the autotransporter (AT) Hbp. Mycobacterium tuberculosis antigens were localized to the surface of E. coli or Salmonella enterica OMVs through fusion to Hbp. Hbp is one of the most abundant proteins detected in E. coli and S. enterica OMVs, making it a strong fusion partner candidate. In addition, the authors exploited the dispensability of certain side domains of Hbp to simultaneously display multiple heterologous antigens on the OMV surface (101). In subsequent work, the authors developed a strategy to overcome limitations in the size and complexity of proteins that can be displayed on the surface of the OMVs via fusion to Hbp. In this approach, the authors fused the SpyTag protein (105) to Hbp. Upon translocation of the Hbp across the outer membrane, the SpyTag protein was found to be displayed on the surface of the OMVs. Large protein antigens or nanobodies were then conjugated to the SpyCatcher protein to enable efficient ligation to the OMV. The SnoopTag/SnoopCatcher system (106) was used in tandem to facilitate surface display of heterologous proteins (103, 104).

Chen et al. developed an approach to simultaneously localize proteins of interest within and on the surface of E. coli OMVs. To target the OMV lumen, the authors created a fusion protein with SlyB, a native lipoprotein that is localized at the inner leaflet (periplasmic side) of the outer membrane (107). At the same time, they localized an antibody on the surface of the OMV using an ice nucleation protein (INP) anchor (108) tethered to an antibody-binding Z domain via a scaffold assembly consisting of 3 cohesin domains (Scaf3). As proof of concept, the authors encapsulated nanoluciferase for detection purposes and tethered IgG to the surface to target thrombin (102).

The ability to functionalize the surface of the OMV to enable selective delivery holds great promise for the targeted delivery of antibiotics. Although not yet commonly employed, targeted delivery of antibiotics represents a promising approach to limit the development of antibiotic resistance, as it would limit exposure of the healthy microbiota to the drug.