Comprehensive proteomic profiling of outer membrane vesicles from Campylobacter jejuni

Abstract

Gram-negative bacteria constitutively release outer membrane vesicles (OMVs) during cell growth that play significant roles in bacterial survival, virulence and pathogenesis. In this study, comprehensive proteomic analysis of OMVs from a human gastrointestinal pathogen Campylobacter jejuni NCTC11168 was performed using high-resolution mass spectrometry. The OMVs of C. jejuni NCTC11168 were isolated from culture supernatants then characterized using electron microscopy and dynamic light scattering revealing spherical OMVs of an average diameter of 50 nm. We then identified 134 vesicular proteins using high-resolution LTQ-Orbitrap mass spectrometry. Subsequent functional analysis of the genes revealed the relationships of the vesicular proteins. Furthermore, known N-glycoproteins were identified from the list of the vesicular proteome, implying the potential role of the OMVs as a delivery means for biologically relevant bacterial glycoproteins. These results enabled us to elucidate the overall proteome profile of pathogenic bacterium C. jejuni and to speculate on the function of OMVs in bacterial infections and communication.

Introduction

Gram-negative bacteria secrete outer membrane vesicles (OMVs) that consist of lipopolysaccharides (LPS), outer membrane proteins, periplasmic proteins, outer membrane lipids, and virulence factors [1, 2]. In pathogens, OMVs are released extracellularly often attaching to host cells through adhesive molecules then transferring compounds that may modulate the host immune response [3]. They have been used for vaccine development [4] or for delivery of foreign proteins such as beta-lactamase [5]. Antigenic OMVs from Neisseria meningitidis were used to develop an administrative medicine to combat the microbial infection by boosting host immune response [6, 7]. Thus, bacterial OMVs have drawn much attention due both to their biological functions and potential use.

The human food-borne pathogen Campylobacter jejuni is the major bacterial causative agent of food poisoning [8] and infections can lead to Guillain-Barré syndrome (GBS), a disorder affecting the peripheral nervous system [9]. Currently, diverse approaches using genomics and proteomics techniques are being applied to better understand the pathogenesis and physiology of this medically important organism [10–13]. In particular, various proteins (i.e. outer membrane proteins) in C. jejuni have been investigated for the immunogenic properties and potential for use in vaccine development [14–16].

Due to the lack of a clear understanding for the biological roles of OMVs, the last decade has seen proteomic analysis of OMVs from a variety of pathogenic bacterial species such as Escherichia coli, Helicobacter pylori, Legionella pneumophila, Neisseria meningitidis, Pseudomonas aeruginosa etc [17–23]. The vesicular proteome of C. jejuni, however, has yet to be sufficiently investigated, despite the likelihood of their role in pathogenesis and virulence. Currently, the OMVs of C. jejuni have only been examined as the delivery mechanism for certain toxins [24].

In this study, we investigated the proteome profile of OMVs from the clinically isolated C. jejuni NCTC11168 using high-resolution LTQ-Orbitrap mass spectrometry (FT MS). Purified OMVs from bacterial culture supernatants were characterized and prepared for MS analysis. The results of this analysis showed a diverse set of enzymes and known virulence factors were present in the OMVs and may play a meaningful role during host-cell interaction.

Material and methods

Purification of OMVs from bacterial culture supernatants

C. jejuni NCTC11168 (ATCC #700819), originally isolated from a patient, was grown in brain heart infusion (BHI) media under microaerobic conditions (1% O₂, 10% CO₂, 10% H₂ and balanced with N₂) at 37°C. A 10 mL overnight culture was inoculated to 1L of media that was allowed to grow 18 h to mid-log phase. OMVs were purified from culture supernatants, as described previously with some modifications [17]. Briefly, the cells were pelleted using centrifugation at 6,000×g for 20 min at 4°C twice. The supernatants were filtered with 0.22 μm filter device to remove remaining cells. The resultant was centrifuged at 30,000×g for 40 min at 4°C to remove cell debris, large cell bulges or vesicle aggregates, and finally centrifuged with 150,000×g for 3 h at 4°C. The pellets were resuspended with 1.6 mL of 10% OptiPrep iodixanol solution (Sigma-Aldrich)/PBS, then gently overlaid onto the top of 40% (1.9 mL) and 50% (3 mL) of OptiPrep/PBS solutions. The OMVs were isolated from the protein-rich fractions with ultracentrifugation at 150,000×g for 2 h at 4°C, then dissolved with PBS buffer and stored at −80°C prior to use.

Transmission electron microscopy (TEM)

Isolated OMVs were ten-fold diluted with deionized water and 3 μL of this suspension was added to glow discharged, 200-mesh formvar-carbon coated grids (Pella Inc.), then the samples were stained with 1% uranyl acetate for negative staining. Samples were applied for 45s, blotted with filter paper and air dried. Digital images were recorded at 67,000× magnification, on a Tecnai T12 electron microscope (FEI Company, Hillsboro, OR) operating at 120 keV acceleration and equipped with a Gatan Ultrascan 2×2K CCD camera.

Dynamic light scattering (DLS)

The size distribution of the OMVs was determined using a DynaPro NanoStar light scattering instrument (Wyatt Technology, Santa Barbara, CA). DLS data were collected from 100 μL of PBS-dissolved OMVs by ten 5 s acquisitions at 25°C with a laser at a wavelength of 659 nm.

Western Blot analysis

Subcellular fractions of C. jejuni were analyzed by western blotting with soybean agglutinin lectin (SBA, Vector Laboratories Inc, CA). Briefly, the fractionated samples were separated on 12% acrylamide gels, transferred to polyvinylidene fluoride (PVDF) membranes by Trans-Blot Turbo transfer system (Bio-Rad Laboratories, Inc., Hercules, CA), and probed with fluorescein-conjugated SBA lectins (50 nM in PBS) after blocking with 2.5% (w/v) bovine serum albumin in PBS for 1 h at room temperature. After washing unbound lectins, the membranes were scanned with Storm 860 Molecular Imager (Molecular Dynamics) at excitation and emission wavelengths of 450 nm and 520 nm for detection.

Sample preparation for MS analysis

Fifty micrograms of total proteins in OMV sample were digested with trypsin (Promega, Madison, WI) on Amicon ultra centrifugal filter units (30K molecular weight cutoff, Millipore, Billerica, MA) based on a method (FASP; filter-aided sample preparation) developed by Manza et al. [25] and modified by Wisniewski et al. [26]. The OMV and whole lysate (WL) were first dissolved in 1% (w/v) SDS and 0.1 M DTT in 100 mM Tris-HCl (pH 7.6) and incubated at 95°C for 5 min, respectively. Alkylation and digestion were performed according to a published protocol [21]. Desalting of digested peptides was performed on Empore High-Performance extraction disk cartridges C18-SD (3M, St. Paul, MN) as described previously [26]. The final peptide samples were dried and reconstituted with 0.1% (v/v) formic acid/water for MS analysis.

NanoLC-MS/MS analysis

Mass spectrometric experiments were performed using a nanoLC LTQ-Orbitrap mass spectrometer (Thermo Fisher Scientific, Bremen, Germany) equipped with a nano-electrospray ion source and an Easy II nanoLC, as described previously [27] with some modifications. A 5 μL sample (1 μg of total proteins) was loaded on a self-packed 5 μm Reprosil C18AQ (Dr. Maisch GmbH, Ammerbuch-Entringen, Germany) column (75 μm × 150 mm). The mobile phases consisted of acetonitrile/formic acid/water (2/0.2/98, v/v/v) for A and acetonitrile/formic acid/water (80/20/0.2, v/v/v) for B. The FASP processed samples were analyzed on a 150 min gradient from 2 to 30% B. The LTQ-Orbitrap was operated in data-dependent acquisition mode to automatically alternate between a full scan (m/z 300–1700) in the Orbitrap and CID MS/MS scans in the linear ion trap. The ten most intense peptide ions were isolated for fragmentation. Helium was used as collision gas for CID. Normalized collision energy was 35% and activation time was 30 ms. Unless otherwise noted, three replicate measurements were made at each MS setting. Data acquisition was controlled by Xcalibur 2.0.7 and Tune 2.4 software (Thermo Fisher Scientific).

Database searching

Raw files were processed and searched using MaxQuant (version 1.2.2.5). Spectra were searched against the total predicted C. jejuni sequences from UNIPROT (1624 sequences, downloaded on 04/12/09) and common contaminants (262 sequences). The spectra were also searched against an equal number of decoy sequences to estimate the false discovery rate (FDR), similar to as described previously [28]. The specified enzyme was trypsin with up to two missed cleavages. Oxidation of methionine (+15.9949) and acetylation of protein N-terminus (+42.0106) were specified as variable modifications and carbamidomethylation of cysteine (+57.0215) was specified as a fixed modification. All other parameters were the default, including a fragment ion tolerance of 0.5 Da and a maximum precursor ion tolerance of 6 ppm after recalibration. Identifications were filtered to curate a dataset with less than 1% FDR at both the peptide and protein level.

Bioinformatic analyses

Prediction of a signal peptide for the general secretory (Sec) or the twin-arginine translocation (Tat) pathway of the vesicular proteins was performed by PRED-TAT software (http://www.compgen.org/tools/PRED-TAT) [29] and the predicted Tat motif was obtained from TATFIND 1.4 server (http://signalfind.org/tatfind.html) [30]. Additional prediction of the signal peptide for vesicular proteins was done using the SignalP 4.1 server (http://www.cbs.dtu.dk/services/SignalP) [31] and using PrediSi software (http://www.predisi.de) [32]. A consensus motif for N-glycosylation (D/E-X₁-N-X₂-S/T; X₁ and X₂ can be any amino acids, except proline) was searched using an in-house algorithm.

Results

Isolation and characterization of OMVs from C. jejuni NCTC11168

OMVs, secreted by C. jejuni into the extracellular milieu, were prepared from culture supernatants by serial centrifugation to remove cells and larger vesicles. The isolated OMVs were fixed and examined by transmission electron microscopy (TEM). The electron micrographs revealed that the OMVs of C. jejuni are spherical in shape with a generally uniform size distribution (Fig. 1A). This method dehydrates the vesicles resulting in an average diameter of 50 nm. Some of the vesicles have collapsed revealing the bilayer (Fig. 1A, white arrow).

Figure 1.

Characterization of OMVs of C. jejuni. (A) Negative-staining TEM of OMVs. The mean diameter of those OMVs is ca. 50 nm. This TEM image shows the bilayered and spherical shaped vesicle (white arrow). (B) Size distributions of OMVs measured by DLS. The averaged hydrodynamic radius of OMVs is ca. 75 nm. (C) SDS-PAGE profiles of OMVs. M, protein marker; WL, whole-cell lysate; PP, periplasmic proteins; OMP, outer membrane proteins. Filled arrows indicate proteins observed in both OMP and OMVs, while empty arrows show proteins observed in both PP and OMVs. (D) Western Blot result probed with fluorescein-conjugated SBA lectin. Filled arrows indicate the SBA-reactive bands in the OMV sample.

Dynamic light scattering (DLS) was used to get an accurate measure of hydrated OMVs. This method directly measures particle size confirming, the homogeneous size distribution seen by TEM. Hydrated, the mean hydrodynamic radius was 77.5 ± 4.6 nm (Fig. 1B). OMVs prepared in the laboratory may not reflect the natural population of OMVs but the nano-sized OMVs are considered to biologically most relevant due to their ability to access a variety of environments within the host.

Proteome profiling of OMVs using MS

Along with purifying the OMVs, the periplasm and outer membrane were fractionated. These three sample pools were separated by SDS-PAGE and visualized by Coomassie staining (Fig. 1C). Comparing the bands on the gel, OMVs appeared to possess both outer membrane and periplasmic proteins; however, there were distinct differences indicating that the OMVs were unique from both. Soybean agglutinin lectin (SBA)-reactive bands were also observed from the vesicular proteins in Western Blot analysis, supporting the likely existence of glycoproteins in the membrane vesicles (Fig. 1D). It should be pointed out that the intensity of the positive bands from the OMVs was relatively lower than those from other samples, indicating that only a small fraction of glycoproteins are present in OMV when compared to whole cell lysates. OMV proteins were digested by trypsin and the subsequent peptides were analyzed using nanoLC-MS/MS with a high resolution LTQ-Orbitrap. Raw data were searched against a UNIPROT protein database of predicted C. jejuni proteins. The experiment was performed three times. A total of 283 vesicular proteins were observed in the purified OMV samples. 134 of these proteins were identified in two out three experiments with high confidence (Table S1) and these were functionally classified. The vesicular proteins were categorized by the predicted subcellular localizations and biological functions (Fig. 2A and B). We observed 48 membrane-associated proteins (36%), 15 periplasmic proteins (11%) and 43 cytosolic proteins (32%). Twenty eight (21%) of them remain unknown, but five of these were predicted to possess signal peptides, indicating that they are more likely periplasmic proteins. In addition, the functional groups ‘energy production and conversion’, ‘cell wall/membrane/envelope biogenesis’, and ‘cell motility’ were present. We also observed 64 enzymes including 20 oxidoreductases (15%), 18 transferases (13%), 5 hydrolases (4%), 7 lyases (5%), 6 isomerases (5%) and 8 ligases (6%), (Fig. 2C).

Figure 2.

Classification of cellular locations (A), functional annotation (B), and vesicular enzymes (C) of the identified proteins of OMVs. (B) The functional classification of the vesicular proteins was done based on the COG database (http://www.ncbi.nih.gov/COG).

Functional annotation of the genes of the vesicular proteins

To interpret the functional relationship of the diverse set of genes identified from the OMVs, the vesicular proteins were functionally annotated by DAVID tools [33]. The 134 high-quality identifications from the OMV samples were compared to 774 genes that were identified from whole-cell lysate of C. jejuni in this study (data not shown). The lists of functional annotation terms of related genes and clusters of the annotated terms enriched in the OMV samples can be found in Table S2 and S3. Among the terms, OMV proteins were found to be enriched for terms related to ‘energy production’ and ‘respiratory metabolism’, indicating a potential role of OMVs in respiration. Flagellar assemble- and cell adhesion-related genes were also enriched in the OMV samples. Interestingly, some of the known N-linked glycoproteins, which are encoded by cjaA, dsbI, kpsD, napA, Cj0089, Cj0114, Cj0152c, Cj0268c, Cj0404, Cj0511 and Cj0694, were found in the vesicles. The annotation terms, which were not clustered in this enrichment analysis, are listed in Table S4.

Signal peptide and N-glycosylation motifs in vesicular proteins

Signal peptide predictors were used to further characterize the identified vesicular proteins of C. jejuni since most outer membrane and periplasmic proteins are known to have a signal peptide sequence in their N-termini that marks them for localization to the periplasm [34]. Prediction of signal peptides in the vesicular proteins was performed using diverse bioinformatics tools such as SignalP 4.1, PrediSi, PRED-TAT and TATFIND 1.4, then the signal motif for the twin-arginine translocation (Tat) pathway was discriminated from the signal peptide for the general secretory (Sec) pathway [35]. Finally, fifty three vesicular proteins (40% of the total) were predicted to carry a signal peptide or a TM domain. Among those, eight vesicular proteins (6% of the total) were predicted to use the Tat apparatus, while 35 proteins (26% of the total) were assumed to be localized via the Sec pathway (Fig. 3A). The substrates for the Tat pathway were classified in proteins that play a role in electron transports and/or anaerobic respiration. Predicted Tat substrates and their sequences are listed in Table 2.

Figure 3.

Pie charts and a bar graph of the relative percentages of the vesicular proteins regarding a signal peptide (SignalP) and/or an N-glycosylation motif (GlycoMotif). (A) A SignalP for Sec or Tat pathways as well as a TM domain was separated from the proteins not carrying any SignalP. (B) SignalP and/or GlycoMotif-containing proteins were discriminated. (C) Actual glycoproteins were separated from the proteins-containing only GlycoMotif or not at all.

Table 2.

List of predicted Tat signal substrates of the vesicular proteins by TATFIND 1.4.

Uniprot accession	Description	Gene	Predicted Tat motifa
Q9PPD9	Periplasmic nitrate reductase	napA
Q0PCC0	Molydopterin containing oxidoreductase	Cj0005c
Q0PB73	Succinate dehydrogenase flavoprotein subunit	sdhA
* Q0P8Y9	Ni/Fe-hydrogenase small chain	hydA
Q0P8A3	Putative periplasmic oxidoreductase	Cj1516
Q0P8A7	Putative formate dehydrogenase large subunit (Selenocysteine containing)	fdhA
Q0P970	Ubiquinol-cytochrome c reductase iron-sulfur subunit	petA
Q0PBP1	Molybdopterin containing oxidoreductase	Cj0264c

^aSequence including predicted a Tat signal motif and a hydrophobic region. A Tat motif was bold and underlined. A hydrophobic region was italic and shaded.

^*This Tat substrate was not predicted using TATFIND 1.4, but using PRED-TAT. Therefore, the Tat motif sequence was predicted manually.

In addition, the bacterial N-glycosylation motif (i.e. D/E-X₁-N-X₂-S/T) was searched from the list of vesicular proteins. Thirty five proteins had the motif in their sequences, but only 11 of them have been experimentally confirmed [36, 37]. Regardless of the presence or absence of signal peptides, the glycosylation motif was present in 35 proteins and certain proteins seem to be actually glycosylated (Fig. 3B and C).

Discussions

The analysis of the OMVs isolated from C. jejuni unveiled the vesicular protein pool providing insight toward the role of OMVs in bacterial physiology and pathology. We isolated a homogeneous population of spherical OMVs, identifying 134 vesicular proteins by mass spectrometry. These were categorized by their biological function and predicted cellular location. The most enriched terms were related to energy generations (i.e. aerobic respiration and TCA cycle) while the most abundant enzymes belonged to the oxidoreductase family. Flagellum assembly proteins were also highly represented.

Several of the observed enzymes, such as oxidoreductases, are known to be involved in the TCA cycle, generating energy by oxidation of certain substrates during respiration. Some respiratory enzymes such as nitrate reductase (NapA), nitrite reductase (NrfA) and cbb3-type oxidase (CcoN) in C. jejuni are known to play a significant role in host colonization [38]. The napA or nrfA deficient strain showed a dramatically reduced level of colonization, while the ccoN mutant completely lost the ability to colonize in chicks [38]. It is also known that, instead of oxygen, C. jejuni can use nitrate, formate, DMSO and/or trimethylamine N-oxide (TMAO) as a respiratory electron acceptor under anaerobic condition, which implies that the ability of the reductases to use a variety of alternative acceptors is crucial in the pathology and physiology [38, 39]. Recently, Liu and coworkers reported that the change of its respiration mode of C. jejuni by the expression levels of those respiratory chain-associated enzymes is closely associated with the bacterial survival in host environments [11].

In the analysis of signal peptides, eight vesicular proteins were predicted to be localized via the twin-arginine translocation (Tat) pathway and were classified in cofactor-containing proteins involved in electron transport chains. This agreed well with the previous results that the Tat machinery in bacteria is mainly responsible for the transport of cofactor-binding folded proteins across the cytoplasmic membrane [40]. Although only eight vesicular proteins were predicted to possess the Tat signal peptide, the actual Tat substrates being found in OMVs might be more substantial as Tat substrates can be secreted as part of a complex [40]. We also observed that some proteins have Sec signal peptides that contribute to a diverse range of cellular processes such as flagellin assembly, signal transduction, amino acid metabolism and virulence. In contrast to the Sec-protein translocation, the Tat-derived protein transport seems to be more specifically used in adaptation and survival of C. jejuni in diverse environments. Thus, it is possible that OMVs can be used as an export route for these crucial enzymes into the environment, to assist in respiration; alternatively, it could be a route to secrete non-functioning enzymes.

This study also identified some cytoplasmic proteins in OMVs. While Horstman and Kuehn believe that cytoplasmic proteins should not be present in OMVs [41], many proteomics studies on OMVs confirmed their presence [18, 20, 42–45]. Further analyses are certainly necessary to confirm the presence or absence of cytoplasmic proteins in OMVs. It is nevertheless worth pointing out DNA and RNA is also present in OMV, indicating that some transcriptional and ribosomal proteins may be needed in OMVs [46–49].

Cell adherence is the first step in host-pathogen interactions and adhesion factors are important components of a bacterial infection [50]. In OMVs we find flagellar proteins, well-known adhesins [51, 52], such as flagellin A, B and flagellar hook proteins. Along with rhamnolipids, flagellar proteins of Pseudomonas aeruginosa are known to be a major inducing factor of psoriasis in human skin [53]. Therefore, the flagellin assembly proteins found in the OMVs of C. jejuni might have some role in bacterial host interaction.

Outer membrane proteins (OMPs), expected to be found in OMVs, were observed such as pore-forming proteins (PorA and Omp50), fibronectin-binding proteins (CadF and Cj1279c) and a surface antigen (CjaA), each known virulence factors [15, 54–60]. In addition, several putative lipoproteins (Cj0089, Cj0396c, Cj0406c and Cj0497) and putative membrane proteins (Cj0129c, Cj0152c, Cj0268c and Cj0404) were found that are additional virulence factor candidates.

C. jejuni protein N-glycosylation is associated with bacterial host interactions [61] and 11 N-glycoproteins are found in the OMVs suggesting a role in pathogenesis. Among those, the surface-exposed CjaA is a well-known antigenic molecule and its glycosylation is known to be related to virulence [54, 62, 63]. An additional set of periplasmic and transmembrane N-glycoproteins were also identified as OMV components.

Recently, Elmi and coworkers identified 151 vesicular proteins from a hypermotile strain C. jejuni 11168H and showed that the OMVs with functional virulence factors [i.e. cytolethal distending toxin (CDT)] can trigger host-immune responses [64]. We found 64 proteins in common from their protein list, but 70 proteins were unique (Fig. 4). This may represent the dynamic nature of protein expression in C. jejuni regarding strain-dependency or growth conditions [57, 65]. OMVs prepared in the laboratory may not reflect the full proteome of OMVs from C. jejuni found in eukaryotic hosts due to the dynamic protein expression profile in response to environmental stimuli. The combined proteomic data provides valuable information toward understanding the pathogenesis and survival of C. jejuni.

Figure 4.

Venn-diagram showing the number of vesicular proteins identified from this study and the previous study done by Elmi et al. [64]. Sixty four proteins were identified in both studies, while 70 and 87 proteins were observed in each study. Dotted circle presents the vesicular proteins even including low-confidence ones, then numbers in parentheses present proteins including the low-confidence proteins.

We identified some lower confidence vesicular proteins of interest (Table S5), e.g. known toxins (CdtA and CdtB) and antigenic peptides (PEB1, PEB2 and PEB3) as well as multidrug efflux components (CmeA and CmeD). The virulence factor CDT complex consisting of CdtA, CdtB and CdtC is released via OMVs causing cell distending effects on host cells [24]. PEB1, a common antigen and cell-binding factor, can be used as a candidate vaccine [66]. CmeA and CmeD function as a multidrug efflux pump contributing to bacterial resistance to a broad range of antimicrobial agents [67, 68]. Some of these (i.e. CmeA and PEB3) are known to be N-glycosylated as well. As we observed in this study, the OMVs of C. jejuni possess a diverse set of proteins such as enzymes, adhesins, toxins etc., implying their latent functions in biological processes. Some of the vesicular proteins could be novel antibacterial targets, diagnostic markers of C. jejuni-mediated diseases and/or vaccine candidates. Therefore, further biological assays need to be performed to elucidate the biological roles of each protein in this context.

Recently, it was reported that the gram-positive pathogen Stapylococcus aureus produces membrane-derived vesicles (MVs) inducing apoptosis of host cells [69, 70] and a secreted vesicle was also observed from the solid spores of Streptomyces coelicolor, a kind of actinobacteria [71]. Taken together, most bacterial cells likely produce MVs; therefore, the secretion of bacterial MVs is likely a ubiquitously observed phenomenon. As the global vesicular proteome of human pathogens becomes clear we will better understand the various biological functions of OMVs.

Table 1.

List of known N-glycoproteins, which were reported in literature [36, 37], identified from OMVs of C. jejuni.

Uniprot accession	Description	Gene	Subcellular localization	COGa	Signal peptideb	Related toc
Q0PB04	Putative secreted protease	Cj0511	Extracellular	O	Sec	Proteolysis[72]
Q0P8H0	Capsule polysaccharide export system periplasmic protein	kpsD	Membrane	M	Sec
Q0PCA8	Disulphide bond formation protein	dsbI	Membrane	O	TM
Q0PC48	Putative lipoprotein	Cj0089	Membrane	M	Sec
Q0PBY5	Putative membrane protein	Cj0152c	Membrane	X
Q0PBA6	Putative transmembrane protein	Cj0404	Membrane	M	TM
Q0PBN7	Putative transmembrane protein	Cj0268c	Membrane	M	TM
Q0P9S0	Putative amino-acid transporter periplasmic solute-binding protein	cjaA	Outer membrane	E	Sec	Highly immunogenic[54]
Q9PPD9	Periplasmic nitrate reductase	napA	Periplasm	P	Tat	Host colonization[38]
Q0PC23	Putative periplasmic protein	Cj0114	Periplasm	M	Sec
Q0PAI5	Putative periplasmic protein	Cj0694	Periplasm	M	TM

^aCOG, Clusters of orthologous groups of proteins. E, amino acid transport and metabolism; M, cell wall/membrane/envelope biogenesis; P, inorganic ion transport and metabolism; O, post-translational modification, protein turnover, chaperones; X, poorly characterized.

^bSignal peptide motifs of the vesicular proteins were predicted using PRED-TAT. Sec, a signal peptide for the general secretory pathway; Tat, a signal peptide for the twin-arginine translocation pathway; TM, transmembrane domain.

^cKnown functions related to bacterial virulence or survival.