Salmonella enterica delivers its genotoxin through outer membrane vesicles secreted from infected cells

Authors


Summary

Cytolethal-distending toxins (CDTs) belong to a family of DNA damage inducing exotoxins that are produced by several Gram-negative bacteria. Salmonella enterica serovar Typhi expresses its CDT (named as Typhoid toxin) only in the Salmonella-containing vacuole (SCV) of infected cells, which requires its export for cell intoxication. The mechanisms of secretion, release in the extracellular space and uptake by bystander cells are poorly understood. We have addressed these issues using a recombinant S. Typhimurium strain, MC71-CDT, where the genes encoding for the PltA, PltB and CdtB subunits of the Typhoid toxin are expressed under control of the endogenous promoters. MC71-CDT grown under conditions that mimic the SCV secreted the holotoxin in outer membrane vesicles (OMVs). Epithelial cells infected with MC71-CDT also secreted OMVs-like vesicles. The release of these extracellular vesicles required an intact SCV and relied on anterograde transport towards the cellular cortex on microtubule and actin tracks. Paracrine internalization of Typhoid toxin-loaded OMVs by bystander cells was dependent on dynamin-1, indicating active endocytosis. The subsequent induction of DNA damage required retrograde transport of the toxin through the Golgi complex. These data provide new insights on the mode of secretion of exotoxins by cells infected with intracellular bacteria.

Introduction

Pathogenic bacteria have evolved sophisticated strategies for delivering to their host cells exotoxins and other effector molecules that promote bacterial survival, replication and spreading (Watson and Spooner, 2006; Hodges and Hecht, 2012). The genotoxic cytolethal-distending toxins (CDTs) produced by several Gram-negative bacteria are targeted to the nucleus of the intoxicated cells where they cause DNA damage and the activation of DNA damage responses (reviewed in Guerra et al., 2011). The CDTs produced by intracellular pathogens that replicate in vacuolar compartments must cross several bacterial and cellular membranes in order to reach neighbouring cells and exert their genotoxic activity (reviewed in Spano and Galan, 2008).

The CDTs consist of three subunits: CdtA, CdtB and CdtC, typically encoded in a single operon (Scott and Kaper, 1994; Lara-Tejero and Galan, 2001). The enzymatically active subunit, CdtB, shares structural and functional homology with the mammalian deoxyribonuclease I (DNase I) (Lara-Tejero and Galan, 2000; Elwell et al., 2001; Nesic et al., 2004), while the CdtA and CdtC subunits contain ricin-like lectin domains (Nesic et al., 2004) that are involved in toxin internalization (Hassane et al., 2003; McSweeney and Dreyfus, 2004; 2005). The holotoxin binds to cholesterol-rich domains of the plasma membrane, is internalized through dynamin-dependent endocytosis and retrogradely transported via the Golgi complex to the endoplasmic reticulum (ER) for subsequent delivery to the nucleus (Cortes-Bratti et al., 2000; Guerra et al., 2005; Boesze-Battaglia et al., 2006).

The CDT encoded by S. Typhi differs from other members of the family in that it is not associated with the CdtA and CdtC subunits (Haghjoo and Galan, 2004). Instead, its toxicity is dependent on transcription of the plt (pertussis-like toxin) A and pltB genes whose products form with CdtB a tripartite complex known also as Typhoid toxin (Spano et al., 2008). S. Typhi produces the Typhoid toxin only upon internalization into the host cell (Haghjoo and Galan, 2004). Based on subcellular fractionation and density gradient centrifugation Galan and co-workers proposed that the toxin is exported from the Salmonella-containing vacuole (SCV) into vesicular structures (Spano et al., 2008). It is surmised that the vesicles traffic to the cellular membrane where they fuse, releasing the Typhoid toxin into the extracellular medium from where it can intoxicate bystander cells (Spano et al., 2008). The molecular details of the events leading to the release of the toxin into the SCV, its incorporation into vesicles, and the delivery of the vesicles to the extracellular milieu remain largely unknown.

In this communication we have detailed the release of Typhoid toxin and its uptake by neighbouring cells. We demonstrate that epithelial cells infected with a recombinant S. Typhimurium strain release the Typhoid toxin in vesicles that are positive for LPS and resemble the outer membrane vesicles (OMVs) produced by extracellular Gram-negative bacteria. Loading of the active CdtB into OMVs is independent on the expression of pltA and pltB. Release of these vesicles into the extracellular environment requires an intact SCV and is dependent on anterograde transport towards the cellular cortex on microtubule and actin tracks. Bystander cells internalize the Typhoid toxin-loaded OMVs by dynamin-dependent endocytosis, followed by the retrograde transport of the toxin to the nucleus via the Golgi complex.

Results

Generation of a recombinant S. Typhimurium strain expressing Typhoid toxin

The pltB–pltA and cdtB genes of S. Typhi (Fig. S1A; Haghjoo and Galan, 2004; Spano et al., 2008) were cloned under the control of their endogenous promoters into the pEGFP-C1 plasmid (Fig. S1B). Each Typhoid toxin subunit was engineered with a unique C-terminal epitope-tag to allow detection of the protein in Western blots (Fig. S1B). Transformation of the plasmids into S. Typhimurium strain MC71 (Clements et al., 2002), generated the recombinant strain MC71-CDT. Isogenic strains lacking cdtB (MC71ΔcdtB) or the pltB/pltA (MC71ΔpltBA) genes were produced as controls (Fig. S1B). The recombinant strains showed levels of entry and replication in epithelial cells comparable to those observed for the wild-type MC71, indicating that expression of the CDT genes does not alter the invasive capacity (Fig. S2 and data not shown).

To assess whether the expression of CDT subunits is comparable to that of S. Typhi, MC71-CDT and MC71ΔcdtB were grown in LB medium or in minimal medium pH 5.8 (MM5.8) that mimic the conditions of the SCV (Ygberg et al., 2006), and protein expression was monitored over time by Western blot (Fig. 1A). The CdtB subunit was detected in MC71-CDT cultured for 3 h in MM5.8 medium, while the expression of the PltA and PltB subunits was delayed until 6 h (Fig. 1A). Interestingly, in absence of the cdtB gene, the PltA subunit was detected as a doublet and PltB was expressed at significantly higher levels (Fig. 1A). Bacteria grown in LB medium did not express any of the three subunits even when the culture was prolonged for up to 9 h (Fig. 1A). The expression of CdtB was mainly dependent on the chemical composition of the medium (e.g. low Mg2+ and phosphate) rather than the pH value, since similar protein levels were detected upon 8 h incubation in minimal medium at pH 7.2 (MM7.2) compared with those observed in MM5.8 medium (Fig. S3). Conversely, the regulation of the PltA and PltB subunits was partially influence by the pH, since the highest levels of expression were observed in bacteria grown in MM5.8 (Fig. S3), suggesting that the two promoters are differentially regulated. This conclusion is further supported by the time kinetics of gene expression shown in Fig. 1A.

Figure 1.

Expression of the CDT subunits by the S. typhimurium MC71 strains.

A. The recombinant MC71ΔcdtB and MC71-CDT strains were grown in LB or MM5.8 medium for the indicated periods of time. Expression of the Typhoid toxin subunits tagged with the indicated epitopes was assessed in total-cell lysates by Western blot. DnaK was used as loading control. One representative experiment out of three is shown.

B. CaCo-2 cells were left uninfected, or infected with MC71ΔcdtB or MC71-CDT strain at the moi 50:1. Expression of the Typhoid toxin subunits was assessed 24 h post infection in total-cell lysate using the indicated tag-specific antibodies. Actin was used as loading control. The asterisk indicates the presence of a non-specific band.

C. CaCo-2 cells were infected with the MC71ΔcdtB or MC71-CDT strains at moi 50:1 for 24 h. Upper panel. Induction of DNA damage was assessed by immunofluorescence using a mouse antibody specific to the phosphorylated form of histone H2AX (γH2AX), followed by a donkey anti-mouse secondary antibody conjugated to Alexa-555 (red). Bacteria were visualized using a rabbit serum anti-LPS followed by a donkey anti-rabbit secondary antibody conjugated to Alexa-488 (green). Nuclei were counterstained with DAPI (blue). Magnification 63×. Lower panel. Analysis of the cell cycle distribution was assessed by PI staining and flow cytometry analysis.

Collectively these data demonstrate that MC71-CDT expresses the Typhoid toxin only in conditions that mimic the vacuolar environment of the host cells, thus validating the model used in this study.

The three Typhoid toxin subunits were detected in lysates of the human epithelial cell line CaCo-2 24 h after infection with MC71-CDT. Also in this setting, we observed increased levels of a high-molecular-weight form of PltA and higher expression of PltB in cells infected with the control strain MC71ΔcdtB (Fig. 1B). CaCo-2 cells infected with MC71-CDT exhibited high levels of DNA damage, assessed by phosphorylation of the histone H2AX (γH2AX), and activation of the DNA damage checkpoint, assessed by accumulation of cells in the G2 phase of the cell cycle (Fig. 1C). These effects were not observed in cells infected with the control MC71ΔcdtB strain.

S. Typhimurium secretes the Typhoid toxin in OMVs

Next, we investigated how the Typhoid toxin is released by the producing bacterium. Haghjoo and Galan have shown that the CdtB of S. Typhi contains a sec-dependent secretion signal that delivers the protein in the periplasmic space (Haghjoo and Galan, 2004). Bioinformatic analysis of the amino acid sequences predicted the presence of N-terminal signal peptides also in PltB and PltA (Fig. S4). Once in the periplasm the holotoxin may be released in OMVs, as previously shown for the CDTs produced by the extracellular bacteria Escherichia coli, Campylobacter jejuni and Actinobacillus actinomycetemcomitans (Berlanda Scorza et al., 2008; Lindmark et al., 2009; Elmi et al., 2012; Rompikuntal et al., 2012). To test this possibility, supernatant of MC71-CDT grown for 8 h in MM5.8 medium were subjected to differential ultracentrifugation. In agreement with previous reports (reviewed in Kulp and Kuehn, 2010), electron microscopy analysis revealed that the bacterium releases a heterogeneous population of OMVs, which stained positive for LPS (Fig. 2A). Based on the size we could identify two subpopulations of vesicles: a major subpopulation with an average size of 50 nm and a smaller fraction (23 ± 2%) of larger vesicles with an average diameter of 130 nm (Fig. 2A). Western blot analysis confirmed the presence of the three toxin subunits as well as LPS in the OMVs preparations. Failure to detect the cytosolic DnaK protein confirmed the absence of contaminating bacterial debris (Fig. 2B). The CdtB subunit was detected by immunogold labelling only after sonication of the OMVs, indicating that the toxin is packed within the vesicles (Fig. 2C). However, due to the sonication procedure, it was not possible to evaluate the size and number of the CdtB-positive vesicles. The presence of functional CDT was confirmed by the ability of OMVs to induce DNA damage in CaCo-2 cells. As shown in Fig. 2D, vesicles purified from the supernatants of MC71-CDT induced cell distension, DNA damage and activation of the G2 checkpoint. These effects were not observed in cells exposed to OMVs purified from the supernatant of MC71ΔcdtB (Fig. 2D).

Figure 2.

Endosomal-like environment promotes secretion of Typhoid toxin-loaded OMVs. The recombinant MC71-CDT strain was grown in MM5.8 medium for 8 h. OMVs were purified by differential centrifugation.

A and C. Electron microscopy analysis of OMVs labelled with a rabbit serum anti-LPS (A) or a mouse monoclonal antibody to the FLAG tag (C) followed by Protein A conjugated to 10 nm gold nanoparticles. When indicated the OMVs preparations were subjected to sonication. Quantification of the OMV size distribution is presented in the right panel of Fig. 2A (one out of two experiments is shown).

B. The presence of LPS and the Typhoid toxin subunits in the OMV preparation was assessed by Western blot. Bacterial cell lysate was used as control.

D. CaCo-2 cells were exposed to 20 μl of purified OMVs. The cell morphology is shown in the representative phase-contrast micrograph 48 h after intoxication. Detection of the γH2AX (green) was assessed 24 h after intoxication by indirect immunofluorescence. Nuclei were counterstained with DAPI (blue). Cell cycle distribution was assessed 48 h after intoxication by PI staining and flow cytometry.

To assess whether loading of the active CdtB subunit into the vesicles is dependent on the expression of PltA and PltB, the abundance of CdtB in OMVs derived from MC71-CDT or the MC71ΔpltBA strain was compared in Western blots. Similar amounts of CdtB were detected in the two OMVs preparations, indicating that PltA and PltB are not required for loading of the active subunit (Fig. S5A). However, the accessory subunits were required to deliver the genotoxic activity of the purified vesicles, as evidenced by the activation of the G2 checkpoint response only in CaCo-2 incubated for 48 h with the MC71-CDT OMVs (Fig. S5B). Collectively these data demonstrated that S. Typhimurium secretes Typhoid toxin-loaded OMVs under conditions that mimic growth within the SCV of infected cells.

Typhoid toxin-loaded OMVs are released from S. Typhimurium-infected cells

Next we investigated how the Typhoid toxin-loaded OMVs are released from the infected cells. To this end, we first used immunofluorescence to assess the localization of CdtB in CaCo-2 cells infected with the GFP-tagged MC71-CDT for 24 h. As shown in Fig. 3A, CdtB was detected in the proximity of bacteria (thin arrows) and in punctae that did not colocalized with the GFP-positive bacteria (thick arrows). Double immunofluorescence analysis further demonstrated that some of CdtB positive but bacteria-negative punctae colocalized with LPS (Fig. 3B), suggesting compartmentalization of the toxin in LPS-positive vesicles. Electron microscopy analysis confirmed the presence of LPS positive vesicles within the SCV (Fig. 3C, black arrowheads). We could not confirm the presence of CdtB in these OMVs due to a strong background of the immunogold staining.

Figure 3.

LPS-positive vesicles are produced in Salmonella-infected cells.

A. Confocal micrographs of CaCo-2 cells infected for 24 h with the MC71SRF430-GFPΔcdtB or MC71SRF430-GFP-CDT strain at the moi 50:1. The CdtB subunit (red) was visualized using an anti-FLAG-specific antibody, followed by a donkey anti-mouse secondary antibody conjugated to Alexa-555. Bacteria are GFP positive (green) and nuclei were counterstained with DAPI (blue). The lower panel shows an ultra-resolution 3D reconstruction (Velocity® software) of the confocal micrographs shown in the upper panel. Thin arrowheads indicate the presence of CdtB in close proximity of bacteria, while the thick arrowheads indicate the presence of bacteria-negative, CdtB-positive punctae.

B. CaCo-2 cells were infected as described in Fig. 3A. The CdtB subunit (red) and LPS (white) were visualized by confocal microscopy. Bacteria are GFP positive (green) and nuclei were counterstained with DAPI (blue). The arrowheads indicate the presence of CdtB- and LPS-positive punctae.

C and D. Electron microscopy analysis of CaCo-2 cells infected as described in Fig. 3A. LPS was detected using a rabbit serum anti-LPS followed by Protein A-conjugated to 10 nm gold nanoparticles. The arrowheads indicated the presence of LPS-positive vesicles in (C) and vesicles containing LPS-positive OMVs in (D). SCV: Salmonella-containing vacuole.

Next we asked how the Typhoid toxin-loaded OMVs are transported from the SCV to the exterior of the cells. We detected the presence of vesicles budding from the SCV containing LPS-positive OMVs, which may be then transported to the cellular membrane and released in the culture medium (Fig. 3D). To test this possibility, extracellular vesicles (EVs) were purified by differential ultracentrifugation from the supernatant of MC71-CDT infected CaCo-2 cells 24 h post infection (p.i.). In accordance with previous reports (reviewed in Raposo and Stoorvogel, 2013), cells secreted a mixture of EVs of different sizes (Fig. 4A). Western blot analysis of the vesicular fraction demonstrated the presence of Major Histocompatibility Class (MHC) class I, a marker of cellular exosomes (reviewed in Chaput and Thery, 2011), CdtB and LPS (Fig. 4B). Failure to detect the cytosolic bacterial protein DnaK or the host cell ER marker BIP confirmed the absence of bacteria debris and microsomal contamination (Fig. 4B). Only the EVs purified from MC71-CDT-infected cells carried the genotoxic activity as demonstrated by their capacity to induce cell distension, DNA damage and activation of the G2 checkpoint response in CaCo-2 cells (Fig. 4C).

Figure 4.

S. enterica-infected cells release LPS- and CdtB-positive vesicles. CaCo-2 cells were infected with the MC71-CDT strain at the moi 100:1, the culture supernatant was collected 24 h p.i., and EVs were purified by differential ultracentrifugation.

A and D. Electron microscopy analysis of EVs labelled with a rabbit serum anti-LPS (A) or a mouse monoclonal antibody to the FLAG tag (D) followed by Protein A conjugated to 10 nm gold nanoparticles. When indicated the OMVs preparations were subjected to sonication. The average diameter for the LPS positive vesicles, indicated with black arrowheads in (A), was 121.5 ± 47 nm. The empty arrowheads indicate LPS-negative vesicles.

B. Western blot analysis of the purified EVs. Total-cell lysate (5 μg) of the cells infected with the MC71-CDT strain was used as positive control.

C. CaCo-2 cells were exposed to 20 μl of EVs purified from CaCo-2 infected either with the MC71ΔcdtB or with the MC71-CDT strain. The cell morphology is shown in the representative phase-contrast micrograph 48 h after intoxication. Detection of γH2AX (green) was assessed 24 h after intoxication by indirect immunofluorescence. Nuclei were counterstained with DAPI (blue). Cell cycle distribution was assessed 48 h after intoxication by PI staining and flow cytometry.

In order to further characterize the CDT containing vesicles, LPS and CdtB were detected by immunogold labelling and electron microscopy analysis. As shown in Fig. 4A, 39 ± 4% of the EV were LPS-positive (black arrows). The LPS-positive vesicles showed an average diameter of 120 nm, comparable to that of an OMVs subset released from bacteria grown in MM5.8 medium (compared Figs 2A and 4A). CdtB could be detected by immunogold labelling only upon sonication of the EVs preparations (Fig. 4D) suggesting that the toxin is inside the vesicles. This was further confirmed by observation that pre-incubation of the vesicular fraction with trypsin did not hamper the induction of cell cycle arrest in cells treated for 48 h (Fig. S6A). As control we demonstrated that: (i) soluble CdtB is readily degraded by trypsin (Fig. S6B), and (ii) the effect was abolished by neutralization of the trypsin activity with soybean trypsin inhibitor (0.25 mg ml−1) (Fig. S6). To exclude that CdtB and LPS might be found in aggregates that co-purify with the vesicular fraction, the EV preparation was further purified in a continuous sucrose density gradient (Fig. 5A). Western blot analysis of seven fractions collected from the gradient demonstrated enrichment of LPS in Fraction 3 (Fig. 5B), which contains vesicles with buoyant density of 1.13 g ml−1. Although we could not detect the CdtB, possibly due to low amount of the toxin, only Fraction 3 induced DNA damage and cell cycle arrest in CaCo-2 cells treated for 24 h and 48 h respectively (Fig. 5C and D). Thus, the Typhoid toxin is contained in vesicles that have both the size and buoyant density of OMVs.

Figure 5.

Characterization of the EVs preparation by sucrose gradient.

A. The density of each of the seven fractions collected was measured using a refractometer.

B. Western blot of the sucrose gradient fractions using a LPS-specific rabbit serum. Total-cell lysate from MC71-CDT infected was used as positive control. The right panel show the optical density of the LPS-specific signal relative to the background. One representative experiment out of two is shown.

C. CaCo-2 cells were exposed to 20 μl of the sucrose density fractions. Untreated cells or cells exposed to the purified CDT from H. ducreyi (2 μg ml−1) were used as negative and positive controls respectively. Detection of γH2AX (green) was assessed 24 h after intoxication by indirect immunofluorescence. Nuclei were counterstained with DAPI (blue). Magnification 63×. Cell cycle distribution was assessed 48 h after intoxication by PI staining and flow cytometry.

D. Quantification of γH2AX-positive cells.

To address whether secretion of the Typhoid toxin-loaded OMVs was dependent on the compartmentalization of Salmonella within the SCV, we used a MC71-CDT strain carrying an inactivating mutation within the sifA gene (MC71-CDTsifA), which impairs the integrity of the vacuole and results in the release of the bacteria in the cytoplasm (Beuzon et al., 2000). We confirmed that deletion of this gene did not alter the levels of the Typhoid toxin expression when the bacteria were cultured for 8 h in MM5.8 medium (Fig. 6A). Furthermore, the MC71-CDTsifA strain showed levels of entry and replication in epithelial cells comparable to those observed for the wild-type MC71-CDT (Fig. 6B). As expected, the sifA mutant showed a cytosolic distribution 24 h post infection (Fig. 6C, upper panel). The different subcellular distribution did not prevent intoxication, as shown by the prominent cell distension induced in the infected CaCo-2 cells 24 h p.i. (Fig. 6C, lower panel). However, in spite of the production of an active Typhoid toxin, CaCo-2 cells infected with the MC71-CDTsifA failed to secrete EVs carrying the genotoxic activity (Fig. 6D).

Figure 6.

SifA is required for secretion of Typhoid toxin-loaded OMVs.

A. The MC71-CDT and MC71-CDTsifA strains were grown in LB or MM5.8 medium for 8 h. Expression of the Typhoid toxin subunits tagged with the indicated epitopes was assessed in total-cell lysates by Western blot. DnaK was used as loading control. The asterisk indicates the presence of a non-specific band. One representative experiment out of two is shown.

B. CaCo-2 cells were infected with the MC71-CDT or MC71-CDTsifA strain at the moi 50:1. The data are presented as number of cfu ml−1 of viable bacteria recovered 2 h (entry) and 24 h (replication) post infection. The input represents the number of cfu ml−1 present in the bacterial suspension used for infection. The mean of two independent experiments is shown in the figure.

C. CaCo-2 cells were infected as described in (B) for 24 h. Upper panel: representative confocal micrographs where bacteria were detected using a rabbit serum anti-LPS followed by a donkey anti-rabbit secondary antibody conjugated to Alexa-488 (green). The cellular cytoskeleton was visualized using a mouse anti-tubulin antibody, followed by a donkey anti-mouse secondary antibody conjugated to Alexa 555 (red). Nuclei were counterstained with DAPI (blue). Magnification 63×. Lower panel: representative phase-contrast micrographs showing the cell morphology 24 h after intoxication (magnification 20×).

D. Cell cycle distribution of CaCo-2 cells left untreated (CTR) or exposed for 48 h to 20 μl of EVs purified from cells infected either with MC71-CDT or with MC71-CDTsifA strain.

These data indicate that the Typhoid toxin-loaded vesicles are produced within an intracellular vacuole, and further packed in a SifA-dependent manner in larger cargoes budding from the SCV, as suggested by the electron microscopy picture presented in Fig. 4D.

Microtubule and actin tracks play an important role in directing the anterograde vesicular trafficking towards the cortex and promoting the release of exocytic vesicles at the plasma membrane (Ross et al., 2008; de Curtis and Meldolesi, 2012). We have therefore investigated whether release of toxin-loaded OMVs was dependent on an intact cellular cytoskeleton. To this end, CaCo-2 cells infected for 3 h with the MC71-CDT strain were left untreated or exposed to nocodazole (25 mM) or cytochalasin D (5 mM) to interfere with microtubule and actin polymerization. EVs were purified 12 h p.i by differential centrifugation. As shown in Fig. 7A and B, both treatments did not impair the recovery of bacteria or the cell viability; however, the release of vesicles carrying an active genotoxin was inhibited of approximately 50% in nocodazole or cytochalasin D-treated cells compared with the levels of secretion observed in the control cells (Fig. 7C).

Figure 7.

Secretion of Typhoid toxin-loaded OMVs requires a functional microtubule and actin cytoskeleton. CaCo-2 cells were infected with the MC71-CDT strain at the moi 100:1. After bacterial entry, cells were left untreated (CTR) or treated with nocodazole (Noc, 25 mM) or cytochalasin D (CytoD, 5 mM). The culture supernatant was collected 12 h p.i., and EVs were purified by differential centrifugation.

A. Number of cfu ml−1 of viable bacteria recovered 12 h p.i.

B. Percentage viable cells assessed by trypan blue exclusion 12 h p.i.

C. CaCo-2 cells were exposed to 20 μl of EVs purified from cells infected with the MC71-CDT strain and left untreated (CTR) or exposed to nocodazole (Noc, 25 mM) or cytochalasin D (CytoD, 5 mM) for 12 h. Detection of γH2AX was assessed 48 h after intoxication by indirect immunofluorescence. Nuclei were counterstained with DAPI. Magnification 63×. The right panel showed the quantification of the γH2AX-positive cells. One out of two experiments is shown in the figure.

Collectively these data demonstrate that Salmonella releases Typhoid toxin-loaded OMVs from infected epithelial cells in a SifA-dependent manner, and this secretion required an intact cellular cytoskeleton. The OMV-nature of these vesicles was further supported by the finding that, similarly to the requirement for loading of CdtB in OMVs (Fig. S5), the incorporation of CdtB did not require PltB and PltA, while presence of these subunits was required for intoxication of bystander cells (Fig. S7). The secretion of these bona-fide Typhoid toxin-loaded OMVs was not restricted to CaCo-2 cells, since similar vesicles were detected in the supernatant of a panel of cell lines of epithelial (CaCo-2 and HeLa), mesenchymal (U2OS) and macrophage (RAW264.7) origin infected for 24 h with MC71-CDT (Fig. S8).

Paracrine uptake of the Typhoid toxin-loaded OMVs

Following the release of Typhoid toxin-loaded OMVs in the extracellular environment, the toxin is delivered to the neighbouring cells. This may occur either by direct fusion of the OMVs with the plasma membrane or by active endocytosis of the vesicles (reviewed in Ellis and Kuehn, 2010). Dynamin-1, a small GTPase that promotes fission of the endocytic vesicles from the plasma membrane, plays a key role in both clathrin-dependent and clathrin-independent endocytosis (reviewed in Ferguson and De Camilli, 2012). To assess whether Typhoid toxin-loaded OMVs are internalized via endocytosis, CaCo-2 cells were pre-incubated with the dynamin-1 inhibitor Dynasore (80 μM for 1 h) before exposure to Typhoid toxin-loaded OMVs pre-labelled with the green fluorescent cell linker dye PKH67. To avoid confounding effects due to the heterogeneity of the cellular EVs, we initially performed this set of experiments with OMVs purified from the MC71-CDT strain grown for 8 h in MM5.8 medium. Pre-incubation with Dynasore significantly reduced the levels of OMV internalization (Fig. 8A–C) and the consequent toxin-induced DNA damage, assessed by γH2AX staining (Fig. 8D and E). Similar results were obtained with EVs purified from infected CaCo-2 cells (data not shown). Thus, dynamin-dependent endocytosis plays an important role in the internalization of Typhoid toxin-loaded OMVs that are delivered to the extracellular space.

Figure 8.

Internalization and trafficking of Typhoid toxin-loaded OMVs requires dynamin and a functional Golgi. CaCo-2 cells were left untreated or pre-treated with the dynamin-1 inhibitor Dynasore (Dyno, 80 μM) for 1 h prior incubation with the PHK67 labelled OMVs for 6 h.

A. Representative Z-projection of untreated or Dynasore-treated CaCo-2 cells labelled with WGA conjugated to Alexa-680 (blue) 6 h post intoxication.

B. The confocal microscopy orthogonal view shows the presence of the PKH67-labelled vesicles within untreated CaCo-2 cells.

C. Quantification of cells carrying PKH67-labelled vesicles.

The genotoxic activity of the Typhoid toxin-loaded vesicles was evaluated by detection of γH2AX (green) 6 h after intoxication by indirect immunofluorescence in cells left untreated or pre-treated with Dynasore (Dyno, 80 μM) or Brefeldin A (BFA, 2 μg ml−1) for 1 h. Nuclei were counterstained with DAPI (blue).

D. Representative micrograph of the γH2AX staining in intoxicated CaCo-2 cells left untreated or pre-treated with Dynasore.

E and F. Quantification of γH2AX-positive cells. Mean ± SEM of three independent experiments.

Upon internalization in endosomes, the Haemophilus ducreyi CDT is retrogradely transported to the nucleus via the Golgi complex and the ER (Guerra et al., 2005). We asked therefore whether the DNA damaging activity of the Typhoid toxin internalized via OMVs endocytosis is also dependent on retrograde transport via this organelle. Pre-treatment of CaCo-2 cells with Brefeldin A (2 μg ml−1 for 1 h), a drug known to disrupt the Golgi, prior exposure to Typhoid toxin-loaded OMVs abolished the induction of DNA damage (Fig. 8F).

Discussion

The production and release of CDT by cells infected with Salmonella enterica is an intriguing issue, since the toxin is produced by the bacterium in an intracellular vacuole and must cross several bacterial and cellular membranes to intoxicate target cells.

This poses first of all the question of the Typhoid toxin secretion from the bacterium. The process does not require the two well-characterized Salmonella type III secretion systems (T3SS), indicating that the toxin is not directly injected into the cytosol of the infected cells (Haghjoo and Galan, 2004). The Typhoid toxin B subunit contains a Sec-dependent translocation signal that mediates secretion into the bacterial periplasm (Haghjoo and Galan, 2004). Using bioinformatics analysis we have identified similar translocation signals also in the PltA and PltB polypeptides (Fig. S4), suggesting that the functional toxin is assembled in the periplasm before secretion. The CDTs produced by bacteria that colonize extracellular environments are delivered from the periplasm via OMVs (Berlanda Scorza et al., 2008; Lindmark et al., 2009; Elmi et al., 2012; Rompikuntal et al., 2012). We have found Typhoid toxin-loaded LPS positive vesicles in the supernatants of S. enterica grown under conditions that mimic the environment of the SCV (Fig. 2), suggesting that a similar route of secretion is adopted by intracellular bacteria. Recent findings indicate that a S. Typhi N-acetyl-β-d-muramidase, named typhoid toxin secretion A (TtsA) digests the peptidoglycan wall and mediates the release of the Typhoid toxin via an unknown mechanism (Hodak and Galan, 2013), which, based on our data, we propose is OMVs production.

A second issue is how the Typhoid toxin-loaded OMVs are released from the SCV into the extracellular environment.

In agreement with Spano et al. (2008), we detected the CdtB subunit in close proximity of the bacterium, as well as in punctae dispersed in the cytoplasm of infected cells (Fig. 4A), and some of these structures colocalized with LPS (Fig. 4B). Conceivably, these two CDT positive clusters represent different mechanisms of secretion: (i) Typhoid toxin-loaded OMVs, which are secreted into SCV-derived vesicles, and (ii) soluble toxin packed into cellular vesicles that then fuse with the plasma membrane releasing the soluble content into the extracellular environment (Spano et al., 2008).

Previous studies have demonstrated the presence of OMVs in the cytoplasm of Salmonella-infected cells (Garcia-del Portillo et al., 1997; Yoon et al., 2011), and we further detected OMVs within structures budding from the SCV (Fig. 4). Vesicular trafficking at the Salmonella vacuole is tightly regulated by bacterial proteins secreted in the host cytoplasm via the Salmonella pathogenicity island (SPI)-2 T3SS (reviewed in Steele-Mortimer, 2008). The T3SS effector SifA and its host interacting partner SKIP (SifA and kinesin interacting protein) promote the fission of vesicles from the SCV and their kinesin-1-dependent anterograde transport (reviewed in Figueira and Holden, 2012). Our data demonstrate that the release of the genotoxin-loaded vesicles is dependent on SifA and on an intact SCV, since OMVs secretion is impaired in cells infected with the sifA mutant (Fig. 6). This scenario is further supported by the requirement of an intact microtubule and actin cytoskeleton for the delivery of this cargo to the plasma membrane and its secretion into the extracellular environment (Fig. 7), as summarized in Fig. 9.

Figure 9.

Proposed model for secretion of Typhoid toxin-loaded OMVs from Salmonella-infected cells. OMVs containing the Typhoid toxin are secreted from Salmonella (1) and packed as cargoes into vesicles budding from the SVC in a SifA-dependent manner (2). The SCV-derived vesicles travel to the cell membrane via a kinesin-dependent anterograde transport on cellular microtubules (3), to be released in the extracellular environment by an exocytosis-like process (4). The secreted Typhoid toxin-loaded OMVs are internalized by dynamin-dependent endocytosis in bystander cells (5), and the toxin requires retrograde transport via the Golgi complex to promote DNA damage (6).

A third issue is how the released toxin reaches its targets. Several mechanisms have been proposed for the intracellular delivery of virulence factors loaded within OMVs (Kulp and Kuehn, 2010). The CDT-loaded OMVs produced by A. actinomycetemcomitans fuse with cholesterol-rich domains of the target cell membrane (Rompikuntal et al., 2012), and disruption of the Golgi by BFA treatment did not impair the internalization of the OMV protein OmpA, confirming direct delivery to the cytoplasm (Rompikuntal et al., 2012), but the requirement for CDT trafficking was not assessed. The Heat Labile enterotoxin (LT) that is exposed on the surface of E. coli secreted OMVs binds to the GM1 receptor in caveolin positive cholesterol-rich domains of the target cell membrane (Horstman and Kuehn, 2002) and is endocytosed in non-acidified vesicles (Kesty et al., 2004). In contrast, internalization of OMVs carrying the VacA toxin of Helicobacter pylori does not require cholesterol, suggesting that binding does not occur within lipid rafts, while intoxication is partially inhibited by chlorpromazine that blocks clathrin-dependent endocytosis (Parker et al., 2010). Our data demonstrate that, similarly to the soluble CDT produced by H. ducreyi (Cortes-Bratti et al., 2000), the Typhoid toxin-loaded OMVs produced by infected cells are internalized in a dynamin-dependent manner, indicating that active uptake plays a major role (Fig. 8). However, since the inhibition of dynamin did not fully prevent induction of DNA damage, other routes, such as direct fusion of the OMVs with the cell membrane or a dynamin-independent internalization, may also contribute to the intracellular delivery of the Typhoid toxin. Independently of the internalization route, intoxication was blocked by treatment with BFA, indicating that the Typhoid toxin is transported to the nucleus via the Golgi complex. It remains unknown how the toxin gains access to this intracellular compartment. Interestingly the presence of the accessory proteins PltA and PltB is not required for the secretion of the toxin-loaded OMVs; however, they are necessary for the delivery of the genotoxic activity to the bystander cells, indicating that they play a key role in the internalization and/or intracellular trafficking of the active CdtB subunit (Figs S5 and S7).

Our findings raise the question of why CDT should be packed in vesicles and not just released as a soluble toxin, as previously proposed (Spano et al., 2008). As reviewed by Kulp and Kuehn (2010) the inclusion of an exotoxin into vesicles offers several advantages: (i) the complex protects the CDT from extracellular proteases (Fig. S6), increasing thereby the half-life and allowing intoxication of distant targets, (ii) higher concentration of the toxin can be delivered to a specific destination, decreasing thereby the effective amount that needs to be produced, thus saving energy, and (iii) thanks to the incorporation of different adhesins, the CDT-loaded vesicles can be delivered to a broad spectrum of host cells.

S. Typhi is the only CDT producing bacterium associated with human cancer (Dutta et al., 2000). In this context, the secretion of CDT-loaded OMVs may have a double effect on tumour promotion and/or initiation in chronic Salmonella carriers: (i) it may improve the delivery of a genotoxic agent thus promoting genomic instability, as previously shown for the Helicobacter hepaticus CDT (Guidi et al., 2013), and (ii) packaging of the genotoxin in vesicles that contain LPS and other pathogen-associated molecular patterns (PAMPs) may contribute to the chronic activation of non-classical oncogenes, such as NFκB and STAT3 (reviewed in Grivennikov et al., 2010), leading to the establishment of an inflammatory environment that could allow intoxicated cells to overcome the tumorigenesis barrier (Bartkova et al., 2006).

Experimental procedures

Cell culture and media

HeLa, CaCo-2, U2OS and RAW264.7 cell lines were purchased from ATCC (LGC Standards Teddington, Middlesex, UK). HeLa, CaCo-2 and U2OS cells were maintained in Iscove's Modified Dulbecco's Medium (IMDM, Sigma-Aldrich, St. Louis, MO, USA), while the Raw264.7 cells were cultured in RPMI (Sigma-Aldrich) containing 2 mM l-Glutamine (Sigma-Aldrich), 10% fetal bovine serum (FBS) (Invitrogen, Grand Island, NY, USA), penicillin 100 U ml−1 and streptomycin 100 mg ml−1 (Sigma-Aldrich), 10 μg ml−1 Ciprofloxacin (Sigma-Aldrich) (complete medium) at 37°C in a humid atmosphere of 5% CO2. The bacteria were routinely grown in Luria–Bertani (LB) medium, supplemented with 50 μg ml−1 chloramphenicol or 50 μg ml−1 kanamycin as indicated in Table 1. Low-pH minimal medium (MM5.8) contained 100 mM Bis/Tris buffer (pH 5.8), 0.1% Casamino Acids, 0.16% glycerol and 10 μM MgCl2 (all reagents were purchased from Sigma-Aldrich). For production and purification of EVs the cells were cultured in exosome-free medium, produced as previously described (Thery et al., 2006).

Table 1. Bacteria strains used in this study
Strain nameResistance geneReference
  1. Kan, kanamycin; Cm, chloramphenicol.
MC71NoneClements et al. (2002)
MC71-CDTKan or CmThis study
MC71ΔcdtBKanThis study
MC71ΔpltBAKanThis study
MC71SRF430-GFPCmThis study
MC71SRF430-GFP-CDTKan, CmThis study
MC71SRF430-GFPΔcdtBKan, CmThis study
MC71-CDTsifACm, KanThis study

Bacteria strains

The bacterial strains used in this study are listed in Table 1. The MC71 was previously described (Clements et al., 2002). Cloning of the genes encoding for the Typhoid toxin was performed as follows. The pltBA operon was amplified by overlap extension PCR of genomic DNA from S. typhi strain CT18 using the primers with 5′ overhang complementary sequences: 5′-CCCCTCGAGATTCTGTAACTGATAAAG-3′ and 5′-ATTTCTTCTCCCCTCTTAGTGGTGGTGGTGGTGGTGCTT-3′; 5′-CACCACCACCACTAAGAGGGGAGAAGAAATAATGAAAAAG-3′ and 5′-CATAAGCTTTTACAGGTCTTCTTCAGAGATCAGTTTCTGTTCTTTAGAAAGTATAAG-3′. The purified PCR products were annealed and further amplified using the primers: 5′-CCCCTCGAGATTCTGTAACTGATAAAG-3′ and 5′-CATAAGCTTTTACAGGTCTTCTTCAGAGATCAGTTTCTGTTCTTTAGAAAGTATAAG-3′. The PCR product was then cloned into the XhoI–HindIII sites of the pEGFP-C1 plasmid, carrying a kanamycin resistance gene (Clontech Laboratories, Mountain View, CA). The cdtB gene was amplified using the primers: 5′-ACGCAAGCTTTGAATAAACTTCATTTTA-3′ and 5′-CGCGGTACCTTACTTGTCATCGTCGTCCTTGTAGTCACAGCTTCGTGCCAA-3′, and the PCR product was ligated into the HindIII–KpnI sites. The cassette encoding for the Typhoid toxin was also cloned into the XhoI–KpnI sites of the pACYCDuet-1 plasmid, carrying a chloramphenicol resistance gene (Merck-Millipore, Darmstadt, Germany).

The GFP expressing strain MC71SRF430-GFP was produced by P22 transduction from the SL1344 PprgH–gfp+ JH 3010 strain (Hautefort et al., 2003). The sifA mutant was initially generated in the 14028 strain, where the sifA gene was replaced by a kanamycin cassette following the one-step replacement method described by Datsenko and Wanner (2000). The mutation was then transferred into the MC71-CDT by P22 transduction.

Expression of the Typhoid toxin subunits was tested by Western blot in bacteria grown in LB medium (non-inducing condition) or MM5.8 medium (inducing condition) at 37°C, in a shaking incubator at 200 r.p.m. At the indicated periods of time, equal volumes of each culture were collected in a 1.5 ml tubes, bacteria were centrifuged for 5 min at 6000 g and lysed in Laemmli buffer (Laemmli, 1970) for Western blot analysis.

Infection of epithelial cells

Infection of epithelial and macrophage cells was performed by gentamicin protection assays, as previously described (Bjur et al., 2006). Activation of the G2 checkpoint was assessed by propidium iodide staining and flow cytometry analysis (Guidi et al., 2013).

Vesicles purification

From bacterial cultures. The indicated MC71 strains were cultured in LB medium for 12 h at 37°C with shaking at 200 r.p.m., and harvested at 2000 g. After washing in PBS the bacteria were resuspended in 50 ml of LB or MM5.8 medium at a final optical density of 0.5 at 600 nm, and cultured for 8 h at 37°C and 200 r.p.m. The bacteria were then pelleted at 2000 g at 4°C for 10 min and the supernatant containing secreted vesicles was centrifuged at 10 000 g for 30 min and filtered with a 0.2 μm filter (Acrodisc 25 mm syringe filter PN4614, Pall Corporation, Port Washington, NY, USA) to remove cell debris. After two rounds of centrifugation at 40 000 g for 70 min, the vesicle containing pellet was resuspended in 300 μl of PBS supplemented with 20% glycerol, stored at 4°C and used within 48 h.

From infected eukaryotic cells. Thirty millions CaCo-2, HeLa or U2OS cells were plated in 10 cm cell culture dishes in complete medium and allowed to adhere for 24 h. Cells were infected with the indicated S. enterica MC71 strains at moi 100:1 in exosome-free media. The culture supernatant was harvested after 24 h and centrifuged at 2000 g for 10 min. The cleared supernatant was further centrifuged at 10 000 g for 30 min and filtered with a 0.2 μm filter. The filtered medium was centrifuged at 40 000 g for 70 min and the pellet was resuspended in 10 ml cold PBS and further centrifuged at 40 000 g for 70 min to remove impurities and proteins not associated with vesicles. EVs were resuspended in 300 μl of PBS supplemented with 20% glycerol and stored at −20°C.

All centrifugation steps at 2000 g were performed at 4°C, with a swing-bucket rotor Hermle 220.86 in a Hermle Z383K refrigerated centrifuge (Hermle-Labortechnik GmbH, Wehingen Germany). The centrifugation steps at 10 000 g and 40 000 g were performed at 4°C, with a swing-bucket rotor Sorvall TH641 in a Sorvall Discovery 90 refrigerated ultracentrifuge (Thermoscientific, Waltham, USA).

Sucrose gradient fractionation

Extracellular vesicles purified by differential centrifugation were resuspended in 500 μl of 2 M sucrose solution in 20 mM Hepes buffer, and deposited at the bottom of an ultracentrifuge polyallomer tube (Cat. # 331372, Backman Coulter, Brea USA). A 2 to 0.2 M sucrose gradient was laid on top using an automatic gradient mixer (Auto Densi-Flow LABCONO, Kansas City, MO, USA). After centrifugation at 210 000 g overnight at 4°C, seven fractions of 500 μl were collected from the top of the tube. Each fractions was diluted in 10 ml PBS and centrifuged at 100 000 g for 2 h at 4°C. The pellets were resuspended in 300 μl PBS supplemented with 20% glycerol and stored at −20°C for further analysis. For Western blot analysis, the fractions were precipitated in 25% trichloroacetic acid (Sigma-Aldrich), washed twice with acetone (Sigma-Aldrich), and the pellet was dissolved in 2× Laemmli buffer (Laemmli, 1970). CDT activity was assessed by treating monolayers of 2 × 105 cells in 12-well plates with 20 μl of each fraction in complete medium for the indicated time. Induction of DNA damage and activation of the checkpoint response were assessed by γH2AX and PI staining respectively.

Electron microscopy

Transmission electron microscopy (TEM) analysis of EVs was performed as previously described (Thery et al., 2006). Briefly, the vesicle preparation was resuspended in 2% paraformaldehyde (PFA) and deposited on glow-discharged formvar/carbon Cu-grids (200 mesh, Electron Microscopy Science, Philadelphia, USA) for 20 min at room temperature. The grids were then washed in PBS and blocked in PBS 1% BSA for 10 min, followed by incubation with monoclonal antibody anti-FLAG (clone M2, Sigma-Aldrich, dilution 1:50), or rabbit polyclonal antibody specific for the S. enterica OMA (Reagensia AB, Solna, Sweden, dilution 1:50) in blocking solution for 1 h at room temperature. The grids were then washed four times in blocking solution and further incubated with Protein A conjugated to 10 nm gold nanoparticles (Wageningen, the Netherlands) at 1:50 dilution in blocking solution for 1 h at room temperature. After six washes in blocking solution and incubation in 1% glutaraldehyde (Agarscientific, Stansted, UK) for 5 min to stabilize the immune reaction, the grids were extensively washed in water. Contrast and embedding was performed with a uranyl-oxalate solution (1:1 ratio of 4% uranyl acetate purchased from Agar Scientific and 0.15 M oxalic acid, purchased from Sigma-Aldrich, at pH 7 adjusted with ammonia) for 5 min, followed by incubation in methyl cellulose-UA solution [9:1 ratio of 2% methyl cellulose 25 centipose (Sigma-Aldrich), and 4% uranyl acetate (Agar Scientific)] for 10 min on ice. After drying at room temperature, immunocomplexes were visualized with a CM120 Transmission Electron Microscope (FEI, Eindhoven, the Netherlands) at 70 kV, and pictures were acquired with a side-mounted CCD MagaView III (Olympus Soft Imaging, GMBH). When indicated, the vesicle preparation was sonicated for 5 min at maximal intensity in a water-bath high power ultrasound sonicator (Bioruptor, Diagenode, Liège, Belgium), before immunostaining.

CaCo-2 cells uninfected or infected with the indicated MC71 strains for 24 h were fixed in 3% PFA in 0.1 M phosphate buffer, embedded in 10% gelatin, treated with 2.3 M sucrose and frozen in liquid nitrogen. Sections were prepared as previously described (Sodersten et al., 2007). Grids were placed directly on drops of 2% BSA (fraction V, Sigma-Aldrich) and 2% Fish gelatin (GE Healthcare, Buckinghampshire, UK) in PBS. Sections were incubated with the rabbit polyclonal antibody specific for the S. enterica OMA (Reagensia AB, 1:500 dilution) in blocking buffer (2% BSA in PBS) overnight in a humidified chamber at room temperature. After extensive washing in blocking buffer, the bound antibodies were detected with Protein A conjugated to 10 nm gold nanoparticles (Biocell, BBInternational, Cardiff, England) in blocking buffer at a final dilution of 1:100. Samples were rinsed in 2% BSA in PBS, fixed in 2% glutaraldehyde (Agarscientific, Stansted, UK), contrasted with 0.05% uranyl-acetate, and embedded in 1% methylcellulose. Immunogold-labelled complexes were visualized using a Tecnai G2 Bio TWIN Transmission Electron Microscope (FEI company, Eindhoven, the Netherlands) at 100 kV. Images were acquired with a Veleta digital camera (Soft Imaging System GmbH, Műnster, Germany).

Western blot

Proteins were fractionated by SDS-polyacrylamide gel electrophoresis using precast gels (Invitrogen, Carlsbad, CA, USA), transferred to polyvinylidene difluoride (PVDF) membranes (Millipore, Billerica, MA, USA) and probed with 1:1000 dilution of the indicated antibodies, followed by the appropriate horseradish peroxidase-conjugated secondary antibody (GE Healthcare, Piscataway, NJ, USA). The blots were developed by enhanced chemiluminescence (GE Healthcare) according to the instructions of the manufacturer. The following antibodies were used: mouse monoclonal antibodies to DnaK (Enzo Life Science, Plymouth, PA, USA), β-Actin and FLAG (clone M2, Sigma-Aldrich), HIS (Cell Signaling Technology, Beverly, MA, USA), c-MYC (clone 9E10, Santa Cruz Biotechnology, Dallas, TX, USA), Ab HC10, which recognizes β2m-unassociated HLA class I heavy chain (Stam et al., 1986) (kindly, provided by Dr Hidde Ploegh, Massachusetts Institute of Technology, Cambridge, MA), the rabbit polyclonal antibodies specific to, BIP (Cell Signaling Technology), and the S. enterica OMA (Reagensia AB).

Immunofluorescence

Phospho-H2AX (γH2AX) staining was performed as previously described (Guidi et al., 2013). For the CdtB and LPS staining, 1 × 105 CaCo-2 cells were seeded on 13 mm diameter slides in a 12-well plate in 2 ml complete medium and allowed to adhere for 48 h prior infection with the indicated S. Typhimurium strains. After 24 h, the cells were fixed in 4% PFA in PBS for 20 min at room temperature. Slides were washed twice in PBS, and further incubated with 0.2% Triton X-100, 3% BSA in PBS for 30 min at room temperature for blocking and permeabilization. The CdtB was visualized using a mouse anti-FLAG monoclonal antibody (clone M2 Sigma-Aldrich, 1:2000 dilution in PBS) for 1 h at room temperature, followed by Alexa-555-conjugated donkey anti-mouse secondary antibody (Invitrogen, Grand Island, NY, USA, 1:1000 dilution in PBS). LPS was visualized using a rabbit polyclonal antibody specific for the S. enterica OMA (Reagensia AB), followed by Alexa-647 or Alexa-488-conjugated donkey anti-rabbit secondary antibody (Invitrogen, 1:1000 dilution in PBS). The nuclei were stained with Vectashield mounting medium for fluorescence with DAPI (Vector Laboratories). The slides were examined with a fluorescence microscope (Leica DMI 6000 B), equipped with Hamamatsu ORCA-R2 digital camera.

OMVs labelling and internalization

Purified OMVs were labelled with the PKH67 green fluorescent cell linker kit (Sigma-Aldrich) according to the instruction of the manufacturer. The labelled vesicles were diluted in 10 ml PBS, centrifuged at 15 000 g for 1 h and resuspended in 200 μl of PBS supplemented with 20% glycerol and stored at 4°C. For the internalization experiments, 1.5 × 105 CaCo-2 cells were grown for 48 h on 13-mm-diameter slides. Cells were pre-treated with the dynamin-1 inhibitor Dynasore (80 μM, Sigma-Aldrich) for 1 h at 37°C prior. Slides were incubated with the PKH67-labelled OMVs on ice for 20 min, and then either fixed in PFA 4% (time 0 h) or further incubated in complete medium in absence or presence of inhibitor for 6 h, prior fixation. Staining of the cellular membranes was performed with Alexa-680-conjugated Wheat germ agglutinin (WGA) (10 μg ml−1 in Hank's Balanced Salt Solution, HBSS, Sigma-Aldrich) for 1 h at room temperature. Slides were washed twice in HBSS, mounted with Vectashield mounting medium (Vector Laboratories), and examined with a Leica DMRXA confocal microscope (63× objective, Wetzlar, Germany) equipped with a CCD camera.

Brefeldin A (BFA) and Dynasore treatments

Three hundred thousand CaCo-2 cells were grown on 13-mm-diameter slides in 12-well plate and allowed to adhere for 48 h. The cells were then pre-treated with BFA 2 μg ml−1 (Sigma-Aldrich) or the dynamin-1 inhibitor Dynasore (80 μM, Sigma-Aldrich) in complete medium for 1 h at 37°C and further exposed to purified EVs for 6 h before γH2AX staining.

Bioinformatic analysis

The amino acid sequences of PltA, PltB and CdtB were retrieved from S. enterica subsp. enterica serovar Typhi str. CT18 (Parkhill et al., 2001) (NCBI ASM19599v1) and analysed with the SignalP 4.1 software (ref. 10.1038/nmeth.1701) to predict the presence of the signal peptide and cleavage site with a D cut-off of 0.57.

Acknowledgements

We are grateful to Sergej Masich for his support for the electron microscopy analysis. This work was supported by grants awarded by Swedish Research Council and the Swedish Cancer Society to T.F. R.G. and L.L. are supported by the Karolinska Institutet doctoral funding (KID). T.F. is a fellow of the Swedish Cancer Society.

Ancillary