Incorporation of Heterologous Outer Membrane and Periplasmic Proteins into Escherichia coli Outer Membrane Vesicles*

Gram-negative bacteria shed outer membrane vesicles composed of outer membrane and periplasmic components. Since vesicles from pathogenic bacteria contain virulence factors and have been shown to interact with eukaryotic cells, it has been proposed that vesicles behave as delivery vehicles. We wanted to determine whether heterologously expressed proteins would be incorporated into the membrane and lumen of vesicles and whether these altered vesicles would associate with host cells. Ail, an outer membrane adhesin/invasin from Yersinia enterocolitica, was detected in purified outer membrane and in vesicles from Escherichia coli strains DH5α, HB101, and MC4100 transformed with plasmid-encoded Ail. In vesicle-host cell co-incubation assays we found that vesicles containing Ail were internalized by eukaryotic cells, unlike vesicles without Ail. To determine whether lumenal vesicle contents could be modified and delivered to host cells, we used periplasmically expressed green fluorescent protein (GFP). GFP fused with the Tat signal sequence was secreted into the periplasm via the twin arginine transporter (Tat) in both the laboratory E. coli strain DH5α and the pathogenic enterotoxigenic E. coli ATCC strain 43886. Pronase-resistant fluorescence was detectable in vesicles from Tat-GFP-transformed strains, demonstrating that GFP was inside intact vesicles. Inclusion of GFP cargo increased vesicle density but did not result in morphological changes in vesicles. These studies are the first to demonstrate the incorporation of heterologously expressed outer membrane and periplasmic proteins into bacterial vesicles.

Bacterial outer membrane vesicles interact with both eukaryotic cells and other bacteria via surface-expressed factors to deliver vesicle components and virulence factors (5,6,8,(11)(12)(13)(14)(15)(16). 2 For example, LT associated with lipopolysaccharide on the surface of enterotoxigenic E. coli (ETEC) vesicles triggers internalization via caveolae and delivers not only catalytically active LT, which intoxicates the eukaryotic cell, but also other bacterial vesicle components. 2 Other studies have suggested that outer membrane invasins IpaB, C, and D may catalyze the internalization S. flexneri vesicles (16).
To date, vesicle components have not been altered by genetic manipulation. Previous studies demonstrated that vesicles could be generated containing periplasmic gentamicin by treatment of cells with gentamicin, but they differed from native vesicles in their composition and size (15)(16)(17). Because vesicles are composed of outer membrane and periplasmic components, we hypothesized that expressed heterologous outer membrane and periplasmic proteins should be packaged in vesicles. Furthermore, we wanted to determine whether vesicle properties could be altered by the expression of proteins into the periplasm and outer membrane of bacteria. For instance, green fluorescent protein (GFP) transported to the periplasm and packaged in vesicles could be used as a lumenal vesicle marker, whereas vesicle incorporation of an outer membrane adhesin/ invasin, Ail from Yersinia enterocolitica, could alter the adhesion and internalization properties of the vesicles. In this study, we demonstrate that Ail and periplasmic GFP are packaged in vesicles and that these altered vesicles can be used to track vesicle interactions with mammalian cells.

EXPERIMENTAL PROCEDURES
Reagents and Cell Culture-E. coli strains DH5␣ (Stratagene), MC4100, and ETEC (ATCC strain 43886) were grown in LB or CFA broth (1% casamino acids, 0.15% yeast extract, 0.005% MgSO 4 , and 0.005% MnCl 2 ), respectively. HB101/pVM102 (Ail) was kindly provided by Dr. Virginia Miller (18). Strains were grown in the presence of kanamycin (10 g/ml) and/or ampicillin (100 g/ml) as required. Human colorectal HT29 cells (ATCC HTB-38) were grown in McCoy's 5a media supplemented with 10% bovine calf serum. CHO-K1 cells (ATCC CCL-61) were grown in Ham's F12K media supplemented with 10% bovine calf serum. All cell lines were grown in the presence of penicillin/ streptomycin/amphotericin B antibiotic-antimycotic solution (Sigma) * This work was supported by a Burroughs Wellcome Fund Investigator in Pathogenesis of Infectious Disease Award (to M. J. K.), a National Institutes of Health Grant, and an American Lung Association Research Grant. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Plasmid Construction-For pNKTG2 (Tat-GFP), the Tat signal sequence and mutGFP3 sequence were cut out of pJDT1 (19) and inserted into pTrc99A (Amersham Biosciences) using EcoRI and HindIII. The mutGFP3 sequence was replaced with mutGFP2 (20) using MscI and HindIII. pNKTAT (TatABCE) was constructed by partial digestion of pTatABCE (21) using EcoRI and HindIII and insertion into the multiple cloning site of pK187 (22). To induce Tat-GFP or TatABCE, overnight cultures were diluted 1:10 and grown at 37°C for 6 h, unless otherwise mentioned, in the presence of 0.1 mM IPTG. Plasmids were transformed by electroporation into bacteria using a modified CaCl 2 protocol (23).
Ail Confirmation-To confirm that the 17-kDa band detected in DH5␣/Ail vesicles by Coomassie stained SDS-PAGE was Ail, purified DH5␣/Ail vesicles (15 g) were loaded on a 15% SDS-PAGE and transferred to polyvinylidene difluoride, and the N-terminal sequence of the 17-kDa band was determined to be Ala-Ser-Glu-Asn-Ser-Ile-Ser-Ile (Tufts Core Facility).
Cargo Quantitation-To determine the percent of GFP or Ail packaged in vesicles, bacteria were pelleted from a culture grown either overnight (Ail) or induced for 6 h (Tat-GFP/TatABCE). Vesicles were pelleted as described above and filter-sterilized. Bacteria and vesicles were precipitated with trichloroacetic acid (20% final concentration; 4°C, 30 min), and the proteins were pelleted (10 min, 16,000 ϫ g) and washed with acetone. Samples were then resuspended in 1% SDS in phosphate-buffered saline. Vesicle samples and dilutions of the bacterial preparation (1:300 for Ail; 1:1000 for GFP) were analyzed by SDS-PAGE and Coomassie staining (Ail) or immunoblotting (GFP). Densitometry results were quantified using NIH Image.
Outer membranes were purified using the Sarkosyl method as described previously (24). To purify outer membrane from DH5␣ strains with and without Ail, spheroplasts collected from the periplasm preparation were resuspended in ice-cold 10 mM Tris (pH 8) and sonicated, and the cells were pelleted (5 min, 8,000 ϫ g). Whole membranes were pelleted from the supernatants (60 min; 40,000 ϫ g) and washed in 10 mM Tris, pH 8, before being resuspended in distilled water and freezethawed. The membranes were then incubated in 0.5% Sarkosyl (sodium N-lauroylsarcinosinate; 25°C, 20 min), and the outer membrane was pelleted (90 min, 40,000 ϫ g). Equal volumes of sample were separated by SDS-PAGE and immunoblotted for GFP or MutL.
Negative Staining Electron Microscopy-To visualize vesicles, samples were applied to carbon-coated 400-mesh copper grids (Electron Microscopy Sciences) and stained with 2% uranyl acetate.
Quantitative Fluorescence Assay-To quantitate the amount of vesicles bound by cells, a 96-well fluorescent assay was used. CHO cells were plated at a concentration of 8 ϫ 10 4 cells/well and incubated overnight (37°C, 5% CO 2 ) to allow cell adherence. Cells were washed with Hanks' buffer (2ϫ) and incubated with FITC-labeled vesicles (1 g) in serum-and antibiotic-free media (100 l). Following incubation, wells were washed with Hanks' (2ϫ) to remove unbound vesicles, and 100 l of Hanks' containing 1% Triton X-100 was added. Fluorescence was quantitated using a 96-well plate FLUOstar Galaxy fluorometer (BMG Lab Technologies) with excitation at 485 nm and emission at 520 nm. Relative fluorescence units were converted to micrograms of vesicle protein using separate standard curves that established the relative fluorescence units per microgram values of FITC-DH5␣ vesicles and FITC-DH5␣ Ail vesicles (r ϭ 0.99), which were present on each microtiter plate tested.
Fluorescence and Immunofluorescence Microscopy-CHO or HT29 (1.6 ϫ 10 5 cells/well) cells were plated on Permanox microwell chamber slides (Nunc Inc.) and incubated (37°C) with vesicles (1-5 g/well) for various times in serum-and antibiotic-free media. Unbound vesicles were then removed, and the cells were washed. Cells were fixed with 4% paraformaldehyde. For immunofluorescence, cells were then permeabilized and blocked for 30 min with 0.1% Triton-X, 5% goat serum, and 0.1% bovine serum albumin. Cells were then incubated for 1 h with rabbit anti-GFP antibody (2.5 g/ml) and visualized with rhodamine red-X-labeled goat anti-rabbit antibody (2.5 g/ml; Jackson ImmunoResearch Laboratories). Slides were mounted with a coverslip and Pro-Long Antifade kit (Molecular Probes), and the cells were observed using a Nikon Eclipse TE200 laser-scanning confocal microscope.

RESULTS
Ail Is Present in E. coli Vesicles-To determine whether exogenous outer membrane proteins are secreted via vesicles, purified vesicles were prepared from several laboratory strains of E. coli transformed with a plasmid encoding the Y. enterocolitica outer membrane protein Ail under the control of its native promoter (18). Ail was detected in purified outer membrane preparations and in vesicles produced by DH5␣/Ail (Fig.  1, A and B), MC4100/Ail, and HB101/Ail (data not shown), and its presence in vesicles was confirmed by N-terminal sequencing. Co-fractionation in Optiprep density gradients of Ail with OmpA and OmpF/C, outer membrane components of native vesicles (Fig. 2, A and B), showed that Ail was shed into the culture supernatant in association with other outer membrane vesicle components. Electron microscopy examination of vesicles in low density Optiprep fractions revealed that Ͼ90% were closed vesicles (Fig. 2, C and D), not membrane whirls or fragments of lysed vesicles. These data demonstrated that the heterologous outer membrane proteins expressed in E. coli are present in native-like vesicles.
To investigate the effect of a heterologous protein on yield, we compared vesicle production as a function of protein concentration per colony forming unit. After overnight growth, DH5␣/Ail produced 2.3-fold more vesicles per colony forming unit than DH5␣. A difference in vesicle density was not detected, because vesicles with Ail migrated to the same density fraction as DH5␣ vesicles (see peak vesicle fractions 4 and 5, Fig. 2, A and B). In addition, no distinguishable difference was detected in the vesicles by negative staining electron microscopy: DH5␣ vesicles ranged from 22-90 nm (Fig. 2C), and DH5␣/Ail vesicles ranged from 22-77 nm (Fig. 2D) in diameter. Therefore, outer membrane vesicle density and size were unaltered due to the incorporation of a heterologous protein into the membrane, but vesicle yield increased.
To determine whether Ail was incorporated differently into vesicles than endogenous outer membrane cargo, we compared the amounts of Ail, OmpF/C, and OmpA packaged into vesicles (Fig. 3). Of the total amount in the bacteria, 0.32% of the Ail, 0.23% of OmpF/C, and 0.14% of the OmpA were packaged into vesicles, which falls in the previously reported range of protein packaged in E. coli vesicles (26,27). In addition, the Omp to Ail ratio appeared constant in each fraction (0.5 ϩ 0.03 S.E., fractions 2-7; Fig. 2B). Therefore, heterologous Ail is neither selectively enriched nor selectively excluded from vesicles, and Ail appears to be included in every vesicle.
Ail Induces Vesicle Internalization by Eukaryotic Cells-Our previous work has demonstrated that LT, which is externally bound to vesicles because of its association with lipopolysaccharide, induces the association and internalization of vesicles by eukaryotic cells (2, 28). 2 Because Ail is a known adhesin/invasin that can confer an invasive phenotype to HB101 (18), we wanted to determine whether Ail is able to catalyze the internalization of DH5␣ vesicles. We used fluorescently labeled vesicles to study cell association. Wild type DH5␣ vesicles displayed minimal eukaryotic cell association (Fig. 4A); however, the presence of Ail in vesicles increased cell association dramatically (Fig. 4B). Quantitation of cell-associated fluorescence revealed that the presence of Ail increased DH5␣ vesicle-cell association 10-fold (Fig.  4C) and that Ail-dependent vesicle-cell association was vesicle concentration-dependent (data not shown). Interestingly, the unbound DH5␣/Ail vesicles removed from the eukaryotic cells contained Ail as shown by Coomassie staining (data not shown). This further supported the finding that Ail was present in every vesicle, because Ail would be expected to be depleted from this fraction if the vesicles contained heterogeneous cargo populations. These data demonstrate that an GFP Transport to the Periplasm Is Limited by Endogenous Levels of TatABCE-Next we wanted to determine whether an exogenous periplasmic protein, which could be utilized in future studies as a lumenal vesicle marker, would also be packaged in vesicles. Due to the reducing environment of the periplasm, GFP transported via the Sec pathway is unable to fold in the periplasm (29); however, GFP folded in the cytoplasm can be transported to the periplasm by the twin arginine transporter system (Tat) using the Tat signal sequence (19,31).
DH5␣ and 43886, a pathogenic E. coli strain, were transformed with a plasmid encoding IPTG-inducible GFP (20) fused to the Tat signal sequence (Tat-GFP). As shown previously for Tat-GFP expressed in MC4100 (19,31), we found that induction of Tat-GFP expression increased the concentration of periplasmic GFP (data not shown). Although GFP-associated fluorescence was detectable, the amount of periplasmic GFP was saturable and could not be increased with longer induction (data not shown). The Tat machinery (TatABCE) has previously been shown to limit the transport of Tat substrate SufI into the periplasm (21, 32). To determine whether this was the limiting factor in Tat-GFP transport, bacteria expressing Tat-GFP were transformed with a plasmid encoding IPTG-inducible TatABCE. Both periplasmic fluorescence (Fig. 5A) and immunoblot analysis for GFP (data not shown) revealed a significant increase of periplasmic GFP when exogenous TatABCE expression was induced.
Periplasmic GFP was detectable in DH5␣ and 43886 after induction of Tat-GFP and TatABCE (Fig. 5, B and C). In the periplasm, the Tat signal sequence is cleaved from Tat-GFP to yield the mature 27-kDa form of GFP. Bands with slightly higher molecular weights were seen in the spheroplasts, which may represent immature forms of GFP (Fig. 5B). After induction for 6 h, 91 and 86% of the GFP was periplasmic in DH5␣/ Tat-GFP/TatABCE and 43886/Tat-GFP/TatABCE, respectively. Immunoblot analysis for MutL, a cytoplasmic protein, confirmed that the spheroplasts were intact and did not leak cytoplasmic components into the periplasmic preparation ( GFP in the Periplasm Is Packaged into Vesicles-Because Tat-GFP was transported and properly folded in the periplasm, we wanted to determine whether it was packaged into vesicles. Vesicles were purified from cultures of strains expressing Tat-GFP with and without TatABCE after 6 and 12 h of induction. GFP was detectable in both DH5␣-and 43886-derived vesicles when TatABCE was expressed (Fig. 6, A and B). Low levels of GFP were detected in the vesicles of 43886 cells expressing only endogenous levels of TatABCE (Fig. 6B). Furthermore, GFP cofractionated with vesicles on an Optiprep gradient during vesicle purification, as observed by the presence of OmpF/C, OmpA, and GFP in the same fractions (Fig. 7A, fractions 4 -6). In contrast, soluble GFP and maltose-binding protein (MBP) (Fig. 7B, fractions 8 and 9) or GFP and MBP from lysed vesicles (data not shown) remained in the heavy fractions in an Optiprep gradient. These data showed that heterologously expressed periplasmic GFP copurified during vesicle purification is associated with intact E. coli vesicles.
43886 vesicles containing GFP were slightly more dense than those without GFP, because GFP and OmpF/C and OmpA were detected in more dense fractions of the gradient (Fig. 7A; compare Omp peak in fractions 4 and 5, middle panel, with peak in fraction 3, lower panel). Similar results were seen with DH5␣ vesicles containing GFP (data not shown). Although the difference in density due to the presence of lumenal GFP in vesicles suggested that the lipid to protein ratio was altered (Fig. 7A), we did not detect a significant difference in the size of vesicles purified from strains with and without Tat-GFP/ TatABCE (Fig. 7, C and D). 43886 vesicles ranged from 43-165 nm and 43886/Tat-GFP/TatABCE ranged from 20 -165 nm in diameter, whereas DH5␣ vesicles ranged from 22-90 nm and DH5␣/Tat-GFP/TatABCE ranged from 25-90 nm in diameter. Note that native DH5␣ vesicles are approximately one-half the size of 43886 vesicles (Fig. 7, C and D, left panels). Unlike the effect of a heterologous outer membrane protein, inclusion of GFP did not affect bacterial vesicle yield.
We determined that GFP in the culture supernatant was not present as a result of reduced membrane integrity. Unlike bacterial cultures of strains expressing cytoplasmic GFP (data not shown), an RNase detection assay confirmed that bacteria expressing Tat-GFP/TatABCE were not leaking periplasm into the culture supernatant. We had previously determined that vesicles remain intact and do not leak periplasmic components when stored in 0.2 M NaCl or when osmotically buffered in Optiprep (data not shown). To establish whether GFP was a lumenal component of vesicles rather than attached to their extracellular surface, Pronase was added to intact and solubilized vesicles. GFP was protected from degradation in unsolubilized vesicles, demonstrating that periplasmic GFP is in the lumen of 43886 vesicles (Fig. 8A). Similar results were observed with the endogenous lumenal vesicle component, MBP (Fig. 8A).
To determine whether GFP was incorporated into vesicles to the same extent as endogenous periplasmic cargo, we compared the amounts of GFP and MBP packaged in vesicles (Fig. 8B). Of the total amount in the bacteria, 0.1% of both GFP and MBP were packaged into 43886 vesicles. As with Ail, we were unable to detect any vesicles that did not contain GFP, and subsets of vesicles from 43886/Tat-GFP/TatABCE fractionated by density (lanes 3-6; Fig. 7A) all contained the same ratio of Omps to GFP. Therefore, periplasmic GFP is neither selectively enriched nor selectively excluded from vesicles, and GFP appears to be included in every vesicle.
GFP-containing 43886 Vesicles Interact with Eukaryotic Cells-In previous studies, we have examined the internalization of ETEC vesicles by human colorectal (HT29) cells. 2 ETEC vesicles interact with HT29 cells via LT, and FITC-labeled outer membrane vesicle components other than LT were internalized. One goal of packaging GFP inside vesicles was to utilize GFP as a lumenal marker during internalization and trafficking experiments of ETEC vesicles. GFP-containing vesicles were fluorescent; however, the cell-associated fluorescence of GFP-containing vesicles was barely above background, unlike the amount of cell-associated fluorescence of vesicles externally labeled with FITC used in the quantitative assay for Ail-containing vesicles. The fluorescence of the GFP was also too low for confocal microscopy experiments, although punctate green staining was visible (data not shown). Nevertheless, we wanted to determine whether GFP, a lumenal vesicle compo-nent, was associated with cells. Therefore, immunofluorescence was used to enhance the GFP signal. Bright punctate staining was observed in HT29 cells incubated with purified vesicles from 43886/Tat-GFP/TatABCE (Fig. 9A), but not from DH5␣/ Tat-GFP/TatABCE (Fig. 9B). These results demonstrated that lumenal contents are present in cell-associated vesicles and that GFP can be utilized as a lumenal vesicle marker for intracellular trafficking.

DISCUSSION
Vesicles are ubiquitously shed by Gram-negative bacteria and have been shown to be vehicles for intercellular transport. They are particularly significant to the study of pathogenic bacteria because they contain and deliver virulence factors to host cells. To understand vesicle formation and vesicle-mediated transport and trafficking inside the host cell, we need to manipulate both vesicle adhesins and vesicle content. Here we show that both membrane and lumenal protein contents of vesicles produced by laboratory and pathogenic strains of E. coli can be manipulated such that native vesicles contain active heterologous factors.
Ail is a protein from Y. enterocolitica that is abundantly expressed in the outer membrane of E. coli. We demonstrate that Ail is also packaged in E. coli vesicles without altering the size or shape of the vesicles. Ail expression allows HB101, a normally non-invasive E. coli strain, to invade CHO cells (18). Similarly, we found that Ail increased DH5␣ vesicle association with CHO cells and that the DH5␣/Ail vesicles appeared by microscopy z-sectioning to be internalized. Thus, engineered vesicles can be used to investigate host cell association and trafficking mediated by specific outer membrane protein adhesins that are active and presented in their native membrane environment. Although whole bacteria pathogens may activate potentially interesting regulatory effects, vesicles from engineered strains can provide a valuable, novel tool in studying a specific outer membrane protein in its native, membrane context. Vesicles provide a "constant expression" level of an Omp, which can be particularly useful in its initial characterization, whereas changes in the bacterial outer membrane composition induced in response to host cells could confuse results. Thus, we propose that the initial interaction and subsequent cellular response due to a specific surface component can be studied using heterologous proteins incorporated in vesicles.
In addition to modifying vesicle surface components, we were interested in changing the soluble, lumenal cargo of vesicles to contain a fluorescent marker, GFP. Previously, GFP fused to the Tat signal sequence has been shown to be transported to the periplasm by the Tat pathway in the E. coli laboratory strain MC4100 (19,31). We demonstrate here that Tat-GFP can also be used in pathogenic E. coli strains. The transport of the SufI Tat substrate to the periplasm was shown to depend on the level of Tat expression in the cell (21,32). Likewise, we show that the Tat transporter is limiting for the Tat-GFP substrate in both the lab strains and the pathogenic strains of E. coli. We further observed that DH5␣ expressing Tat-GFP without TatABCE overexpression appeared filamentous, much like the ⌬TatC mutants (33), perhaps because overloading the Tat transporter with the fusion protein resulted in the lack of a functional Tat transporter for natural Tat substrates.
Interestingly, the DH5␣/Tat-GFP/TatABCE spheroplasts did not have any detectable fluorescence over that of background, very little cytoplasmic Tat-GFP was detectable by immunoblotting, and, at endogenous (low) levels of TatABCE, spheroplasts contained little to no cytoplasmic GFP (data not shown). However, we did detect fluorescence and GFP in spheroplasts of 43886/Tat-GFP/TatABCE. Because 43886 is more difficult to lyse than DH5␣, we used polymyxin B to purify periplasm. Thus, whereas the spheroplasts' content did not leak into the periplasmic preparation (as determined by MutL immunoblotting), it was not determined whether the spheroplast fraction contained unlysed cells that could explain the GFP detected in this fraction.
To determine whether periplasmic GFP entered vesicles, we first showed that crude and gradient-purified vesicle preparations from TatABCE-overexpressing bacteria contained GFP. However, it was possible that this associated GFP cofractionated with vesicles because it was bound to their exterior. This was of particular concern, because bacteria expressing cytosolic GFP were highly osmotically sensitive (20). We found that periplasmic GFP did not decrease outer membrane integrity and that vesicle GFP was protected from Pronase unless the vesicles were solubilized with SDS. These results proved that GFP was lumenally packaged in vesicles.
GFP-containing vesicles were denser; however, the average vesicle size did not change. This result suggested that the protein-lipid ratio increased, presumably as a result of the inclusion of heterologous soluble cargo. Unlike the GFP-containing vesicles, there did not appear to be a density difference in DH5␣/Ail vesicles. Both DH5␣ and 43886 package outer membrane proteins into vesicles to the same extent. Of total cellular OmpF/C, DH5␣ and 43886 package 0.23% and 0.17% in vesicles, respectively, whereas 0.14% of OmpA is packaged by DH5␣, and 0.15% by 43886. These data correspond to previously reported values (26,27). Therefore, there may be an optimal protein-lipid ratio within the outer membrane, even when an abundant heterologous membrane protein is present. Although the lipid to protein ratio appeared unaltered, the vesicle yield was slightly increased with the incorporation of Ail. However, induction of GFP, which altered vesicle density, did not change vesicle yield. We conclude that heterologous lumenal and outer membrane cargo may affect vesiculation differently.
Unfortunately, the intrinsic fluorescence of vesicles containing GFP was barely above background fluorescence and could not be used to detect cell-associated vesicles. However, GFP inside 43886 vesicles could be detected in vesicle-treated HT29 cells by amplifying GFP detection with antibodies to GFP, showing that GFP can be used as a lumenal marker protein.
More importantly, this result demonstrates that ETEC vesicles carry lumenal components into eukaryotic cells.
In eukaryotic cells, signal sequences or protein modifications are necessary signals for protein sorting into different vesicles and organelles. For example, the endoplasmic reticulum retention signal sequence, KDEL, signals for retrograde transport from the Golgi to the ER (34), whereas exosomes are enriched in ubiquitinated proteins (35,36). Bacterial protein secretion through the inner membrane, such as by the Tat and the sec pathways, also utilizes signal sequences (19,37,38). Currently, little is known about protein sorting into outer membrane vesicles produced by bacteria. Previous studies have reported that certain proteins such as LT, leukotoxin, and ClyA are enriched in vesicles, whereas other periplasmic proteins, such as DsbA, are excluded, suggesting that vesiculation is an active and directed process (2,6,8). In this study, the percentage of Ail or GFP packaged in vesicles corresponded to the amount of endogenous outer membrane and periplasmic components packaged in vesicles as well as to previous reports of the percentage of protein packaged in E. coli vesicles (26,27). Presumably, GFP would not contain a vesicle "sorting sequence," since it is produced by the jellyfish Aequorea victoria and is not a native bacterial protein. These data suggest that a cargo tag is not necessary for protein sorting into bacterial vesicles. Instead, we propose that the previously described toxin enrichment in vesicles may be due to cell envelope "hot spots" for vesicle budding. These enriched areas may occur due to the location of protein secretion machinery, such as the type II secretion machinery necessary for LT secretion through the inner and outer membranes (28,30,39).
In conclusion, heterologous outer membrane and periplasmic proteins, such as Ail and periplasmic GFP, are packaged in vesicles and can be used to modify the characteristics of the vesicle. Outer membrane and external proteins such as Ail and LT can be used to direct the interactions of vesicles with eukaryotic cells, and GFP and other lumenal vesicle cargo can be used to track soluble transported cargo. Together, these tools provide valuable methods to study the nature of vesicle-mediated transport.