|Home | About | Journals | Submit | Contact Us | Français|
Vibrio ssp. are associated with infections caused by contaminated food and water. A Type III Secretion System (T3SS2) is a shared feature of all clinical isolates of V. parahaemolyticus and some V. cholerae strains. Despite being responsible for enterotoxicity, no molecular mechanism has been determined for the T3SS2-dependent pathogenicity. Here we show that although Vibrio ssp. are typically thought of as extracellular pathogens, the T3SS2 of Vibrio mediates host cell invasion, vacuole formation and replication of intracellular bacteria. The catalytically active effector VopC is critical for Vibrio T3SS2 mediated invasion. There are other marine bacteria encoding VopC homologues associated with a T3SS and, therefore, we predict that these bacteria will also likely to use T3SS mediated invasion as part of their pathogenesis mechanisms. These findings suggest a new molecular paradigm for Vibrio pathogenicity and modify our view for the roles of T3SS2 effectors translocated during infection.
A Type III Secretion System (T3SS) is used by bacterial pathogens to inject effector proteins into the cytoplasm of their host cells. While the T3SS machinery is often conserved among Gram-negative pathogens, the effectors from each system differ widely in their mechanism of action. These effectors are typically potent proteins that mimic or capture an endogenous eukaryotic activity to disrupt the cellular response to infection (Broberg and Orth, 2010; Ham et al., 2011). Comparative genome analysis has demonstrated that a common Vibrio progenitor gave rise to V. parahaemolyticus, V. cholerae and other Vibrio species (Okada et al., 2009). The acquisition of a T3SS similar to that found in Yersinia species (causal agent of the Plague and gastroenteritous), herein referred to as T3SS1, is one of the distinguishing features between V. parahaemolyticus and V. cholerae (Okada et al., 2010a).
The more recent acquisitions by V. parahaemolyticus of a second Type III Secretion System (T3SS2), thermostable direct hemolysins (TdhA/S) and TDH related hemolysin (TRH) have given rise to a diverse set of pathogenic strains, enabling the bacteria to adapt a host-associated lifestyle (Makino et al., 2003; Nishibuchi, 1995). T3SS2 is located on a pathogenicity island that is hypothesized to be encoded on a mobile genetic element and is a shared feature of all characterized clinical isolates of V. parahaemolyticus and some V. cholerae strains (Okada et al., 2010b). Despite its discovery more than sixty years ago and suggestions that some V. parahaemolyticus strains are invasive, no molecular mechanism has been determined for the pathogenicity linked to clinical presentation of gastroenteritis (Akeda et al., 1997; Broberg et al., 2011; Makino et al., 2003). Herein, we have discovered that the enterotoxic T3SS2 mediates invasion of V. parahaemolyticus into non-phagocytic cells using an effector, VopC (Park et al., 2004a). This mechanism also appears to be shared by some epidemic V. cholerae non-O1 strains that do not contain cholera toxin. This VopC-containing T3SS is found in a number of marine bacteria, suggesting a previously uncharacterized pathogenesis mechanism for these bacterial pathogens.
Based on G-C content, the T3SS1 of V. parahaemolyticus was ancestrally acquired, while the clinically associated T3SS2 was obtained through a relatively recent lateral gene transfer of a pathogenicity island (Makino et al., 2003). Although the evolutionary history of V. parahaemolyticus subsequently diverged from that of V. cholerae, in which some strains acquired either a T3SS2 or a phage-encoded cholera toxin, respectively, the two species contain remarkable similar T3SS2s and the effectors associated with this secretion system are well conserved (Okada et al., 2010b). Among known T3SS2 effectors, there is VopC (VPA1321 in V. parahaemolyticus RIMD2210633) that shows sequence similarity to the catalytic domain of cytotoxic necrotizing factor (CNF) toxins, including those secreted by Yersinia ssp., Bordetella ssp. and Escherichia coli (Supplemental Table 1). V. parahaemolyticus appears to have hijacked the catalytic domain from the toxins, since VopC encodes for a T3SS2 secretion signal linked to that enzymatic domain. By contrast, the toxins encode distinct domains for them to be secreted by bacteria, be taken up by the host cells and modify host cell targets. These toxins induce changes in cell shape and facilitate invasion of the pathogens into the host cell (Aktories and Barbieri, 2005). Therefore, it is logical to ask if Vibrio T3SS2 uses VopC to mediate bacterial invasion into a host cell.
Our studies into the role of V. parahaemolyticus T3SS2 and initial examination of Vibrio invasion involve the CAB2 strain, derived from the pathogenic RIMD strain. CAB2 contains an active T3SS2 but does not express the hemolysins and T3SS1, which allows for T3SS2 activity to be studied independent of other virulence factors. As a control we used the CAB4 strain, which does not contain active hemolysins, T3SS1 or T3SS2 (Supplemental Table 2). Initial infection and cytotoxicity assays demonstrated that CAB2 induced with bile salt was able to cause changes in cell shape followed by cell lysis at approximately 4–5 hours after infection (Gotoh et al., 2010)(Figure 1A–D). Surprisingly, our observations also suggested that the T3SS2 only CAB2 was invading host cells. When invasion of non-phagocytic HeLa cells was analyzed over time, the amount of intracellular CAB2 increased in the first few hours of infection and then precipitously dropped at approximately 4–5 hours, coincident with the timing of cell lysis (Figure 1E). The numbers of intracellular bacteria over time after initial infection were also quantitated by incubating the infected cells in medium containing gentamicin to kill extracellular bacteria and eliminate the possibility of re-infection. An increase in the number of surviving (intracellular) bacteria followed by a dramatic decrease coincident with cell lysis at end of the time course, suggesting Vibrio was able to replicate inside host cells (Figure 1F). The invasion efficiency of CAB2 was comparable to that of intracellular pathogen Shigella flexneri (Figure 1G), and we observed similar results with differentiated epithelial colorectal CaCo2 cells (Supplemental Figure 1). These studies support the hypothesis that V. parahaemolyticus uses a T3SS2-mediated mechanism for invasion, intracellular replication and lysis of infected cells and is consistent with an early, isolated observation for a few invasive pathogenic V. parahaemolyticus strains (Akeda et al., 1997).
To further characterize host cell invasion, we analyzed HeLa cells infected with the same V. parahaemolyticus strains by transmission electron microscopy (TEM). HeLa cells infected with CAB2, but not CAB4, contained intracellular bacteria (Figure 1H–I).
We next investigated the role of VopC in T3SS2 mediated invasion. The catalytic domain shared by VopC and the CNF toxins is predicted to have deamidase/transglutaminase activity that has been associated with modification of Rho family GTPases, including Rac, Rho and/or CDC42 (Flatau et al., 1997). The conserved catalytic residues in VopC are cysteine 220 and histidine 235 (866 and 881 in CNF1, respectively) (Figure 2A). Modified by toxins CNF or DNT (Bordetella dermonecrotizing toxin) through deamidation or transglutamination, respectively, a Rho GTPase becomes constitutively active, triggering changes in the actin cytoskeleton of the infected cell (Masuda et al., 2000; Schmidt et al., 1997). These changes in cell shape facilitate invasion of the pathogens into the host cell (Doye et al., 2002).
Therefore, we tested whether Vibrio invasion was dependent on the presence of VopC. CAB2 and CAB2ΔvopC+VopC, but not CAB2ΔvopC, were able to invade HeLa cells (Figure 2B–D), suggesting a critical role for VopC in invasion. While VopC is necessary for invasion, the cytotoxicity of T3SS2 is independent of VopC and likely due to the delivery of other effectors (Supplemental Figure 2A–B). To assess whether the enzymatic activity of VopC was critical, we analyzed invasion mediated by the VopC deletion strain complemented with the putative catalytically inactive VopC C220S. In contrast to the CAB2ΔvopC+VopC strain, the CAB2ΔvopC+VopC C220S strain was unable to mediate invasion (Figure 2B), while VopC C220S was expressed and secreted at a similar level to wild VopC (Supplemental Figure 2C). Cytoskeletal inhibitors were reported to be able to block invasive V. parahaemolyticus (Akeda et al., 1997). In the presence of inhibitors nocodazole or cytochalasin D, CAB2 was indeed no longer invasive, supporting the hypothesis that VopC induces changes in cytoskeleton shape to facilitate invasion (Figure 2B). To further investigate the invasion process, we analyzed HeLa cells infected with various strains by TEM. HeLa cells infected with CAB2ΔvopC+VopC, but not CAB2ΔvopC, contained intracellular bacteria (Figure 2C–D). The number of intracellular bacteria varied and in some cases, intracellular bacteria appeared to be replicating (Figure 2D).
We next assessed if VopC is using a mechanism similar to CNF toxins by analyzing the active state of host GTPases. Over the course of infection with CAB2, but not with CAB2ΔvopC, CDC42 was activated as demonstrated by its interaction with its downstream target, the binding domain of p21-activated kinase 3 (PAK PBD) (Figure 3A). Infections with a deletion strain reconstituted with wild type VopC (CAB2ΔvopC+VopC), but not with the deletion strain reconstituted with a putative catalytically inactive VopC (CAB2ΔvopC+VopC C220S), also caused CDC42 activation without affecting the total level of this GTPase (Figure 3A). These studies implicate the catalytic activity of VopC in the activation of Rho GTPases during infection.
To further test the activity of VopC, we transfected HeLa cells with VopCΔ67, an N-terminal truncation that retains the catalytic domain but removes the putative signal sequence to aid expression. Transfection of HeLa cells with VopCΔ67, but not GFP or VopCΔ67 C220S, induced the formation of both actin ruffles and filopodia (Figure 3B–G). These changes in cell morphology appeared to be due to the enzymatic activity of VopC, because transfection of the predicted catalytically inactive VopCΔ67 C220S did not induce changes in the actin cytoskeleton when the proteins were expressed at a similar level in HeLa cells (Supplemental Figure 3A).
Further biochemical analysis of the transfected cells revealed that the expression of VopCΔ67, but not VopCΔ67 C220S, caused the activation of both Rac and CDC42, but not RhoA, as measured by the ability of the activated GTPases to interact with their downstream targets, PAK PBD or the binding domain of Rhotekin, respectively (Figure 3H). The specificity of VopC for Rac and CDC42 but not RhoA is distinct from the activities observed with the CNF secreted toxins, where theyCNF1 and DNT (dermonecrotizing toxin) activate Rac, CDC42 and RhoA by deamidation and/or transglutamination (Horiguchi, 2001).
To test the biochemical activities of VopC, we first performed a transglutamination assay using the fluorescent amine-donor dansylcadaverine and Rho GTPases as substrates. MBP-VopCΔ67, but not MBP-VopCΔ67 C220S, was able to transglutaminate Rac and CDC42, but not RhoA with dansylcadaverine, which is consistent with the selective activation of these GTPases observed during transfection (Figure 3H–I, Supplemental Figure 3B). CNF1 and DNT toxins modify Q61 in the switch 2 region of Rac and CDC42 resulting in constitutive activation of these small G proteins (Aktories and Barbieri, 2005). We observed that VopC is unable to transglutaminate a Rac Q61L mutant, supporting the hypothesis that VopC also targets this residue specifically (Figure 3I). We further observed that MBP-VopCΔ67, but not MBP-VopCΔ67 C220S, could also act as a deamidase. Ammonia release resulting from a deamidation reaction was detected when VopC was incubated with Rac, but not Rac Q61L, further suggesting that Q61 is the target residue of VopC activity (Figure 3J). To confirm the activity and residue specificity VopC exerts on Rho family GTPases, we expressed GST-Rac in the presence or absence of MBP-VopCΔ67 and determined the total mass of purified GST-Rac by mass spectrometry (MS). No significant difference was observed between Rac and Rac co-expressed with VopC. Since a tranglutaminase reaction would add substantial mass, this result suggests that the reaction does not occur when these proteins are co-expressed (Supplemental Figure 4). We analyzed chymotryptic digests of the two Rac samples using liquid chromatography followed by MS to determine whether Q61 had been modified. MS analysis demonstrated that only the deamidated form of Q61 was detected when Rac was co-expressed with VopC, while no deamidated form of Q61 was detected in the control sample, demonstrating that VopC activated Rac by deamidation of Q61 (Figure 3K; Supplemental Figure 4).
Some pathogenic V. cholerae non-O1 strains contain a T3SS2-like gene cluster encoding a similar repertoire of effectors, including VopC. Interestingly, two of these clinical isolates, V. cholerae 1587 and 623-39, in contrast to the well-studied V. cholerae serotypes O1 and O139, do not contain a cholera toxin, but are correlated with sporadic cholera epidemics, pathology in animal models and limited, albeit unexplained cell invasion (Dalsgaard et al., 1995). Similar to V. parahaemolyticus, we observed cytotoxicity induced by V. cholerae 1587. Using time-course invasion experiments and LDH release assays, we obtained results that suggested V. cholerae 1587 could replicate inside HeLa cells (Figure 4A–C). We observed that V. cholerae 1587, but not the V. cholerae O1 El Tor strain, which contains cholera toxin but not a T3SS, were able to invade non-phagocytic HeLa cells in a manner similar to that observed with V. parahaemolyticus (Figure 4D). Furthermore, V. cholerae 1587ΔvopC was no longer able to invade HeLa cells, and actin accumulation inhibitor cytochalasin D was able to block V. cholerae 1587 invasion, suggesting VopC has a similar role for invasion in V. cholerae as in V. parahaemolyticus. Thus, these V. cholerae non-O1 strains contain an active T3SS that mediates their invasion into host cells.
We have discovered that T3SS2 enables V. parahaemolyticus to act as an invasive pathogen. This T3SS2-mediated invasion is dependent on the presence of the effector VopC that has a deamidase/transglutaminase activity. Upon translocation into the host cell, VopC activates small GTPases Rac and CDC42 to induce changes in the actin cytoskeleton and facilitate the entry of V. parahaemolyticus into non-phagocytic host cells. It is possible that other known and unknown effectors secreted by T3SS2 may facilitate pathogen survival and replication inside the host cell, as well as cell lysis when a critical intracellular bacterial burden has been reached. Other known T3SS2 effectors include: VopP/A, a Ser/Thr acetyltransferase; VopL/F, an actin nucleator; and VopV, an actin bundling protein (Hiyoshi et al., 2011; Liverman et al., 2007; Okada et al., 2010b; Trosky et al., 2007). We propose that these effectors are likely to be involved in the intracellular survival, replication and spreading of these Vibrio species within the infected host. Although a direct connection between T3SS2 mediated invasion and T3SS2 mediated enterotoxicity has not been established, future studies will no doubt elucidate the molecular mechanisms of other yet-to-be-characterized T3SS2 effectors and how T3SS2 might contribute to possible intracellular pathogenesis.
Herein, we also observed that another pathogenic bacterium, a V. cholerae non-O1, non-139 strain, contains a T3SS2 with a similar repertoire of effectors including VopC. This strain of V. cholerae does not encode a cholera toxin but is involved in limited epidemic outbreaks (Dalsgaard et al., 1995). We propose that this pathogenicity for V. cholerae may be mediated by the recently acquired T3SS2 that allows for invasion. Our observations suggest that the V. cholerae T3SS2 functions independently of cholera toxin and could be sufficient for pathogenicity of V. cholerae during infection of its animal host.
Finally, the implications for invasive bacteria most likely extend beyond Vibrio, as there are other species of bacteria from marine environments that have acquired a T3SS encoding a VopC homologue (Supplemental Table 1). These observations present a new molecular paradigm for Vibrio pathogenicity, and future studies on invasion, survival and spreading of the bacteria within the host will undoubtedly result in better understanding on signaling mechanisms used by both the host and the pathogen.
V. parahaemolyticus CAB strains were derived from POR1 (RIMD 2210633 ΔtdhAS), generously provided by Drs. Tetsuya Iida and Takeshi Honda (Park et al., 2004b). They were made by deleting the transcription factors ExsA and/or VtrA that regulate the two T3SSs (Kodama et al., 2010; Kodama et al., 2007; Zhou et al., 2008). To induce T3SS2, media was supplemented with 0.05% bile salt and bacteria were grown at 37°C for 2 hours (Gotoh et al., 2010). V. cholerae strains 1587 were generously provided by Dr. John Mekalanos and it was grown in LB at 37 °C.
LDH release at each indicated time point was measured in triplicate using a Cytotoxicity Detection kit (Takara) according to the manufacturer’s instructions. Results are presented as cytotoxicity as calculated from percent of total lysis.
HeLa cells were infected at a multiplicity of infection (MOI) of 10 for the indicated time points. At each time point, 100 µg/ml gentamicin was added to kill extracellular bacteria. Cells were lysed with 0.5% Triton X-100 and the intracellular bacteria were assessed. For the gentamicin incubation experiments, cells were infected for 2 hours, washed with DMEM and incubated in DMEM containing 100 µg/ml gentamicin. At each time point, cells were washed and lysed and the intracellular bacteria were assessed.
HeLa cells were infected for two hours and fixed in 2.5% glutaraldehyde in 0.1M sodium cacodylate, followed by 1% osmium tetroxide in 0.1M sodium cacodylate. Cells were embedded in epoxy resin (Electron Microscopy Sciences) and polymerized at 60 °C. Ultrathin sections were cut at 80 nm and stained with uranyl acetate and lead citrate. Sections were examined at 120KV using a Tecnai G2 Spirit transmission electron microscope (FEI Company) and images recorded on a Gatan USC1000 2k CCD camera (Gatan Inc).
The nucleotide sequence 1kb upstream and 1kb downstream of VopC were cloned into pDM4, a CmR OriR6K suicide plasmid (kindly provided by Dr. Doug Call). The resulting plasmid was conjugated into CAB2 and transconjugants were selected on media containing 25µg/ml chloramphenicol. Bacteria were counter-selected by growing on media containing 15% sucrose.
The ΔvopC deletion strain was reconstituted using pBAD (Invitrogen) containing vopC FLAG or vopC C220S FLAG preceded by 1kb of its upstream sequence and a kanamycin resistance gene.
pBAD vopC-FLAG or pBAD vopC C220S-FLAG were mated into different strains. Bacteria were induced with 0.05% bile salt at 37 °C (Gotoh et al., 2010). 3h after induction, the culture was centrifuged to separate supernatants and pellets. The supernatant was filtered and ice-cold trichloroacetate was added to a final concentration 17 of 10%. Samples were kept on ice overnight and then centrifuged. The precipitates and pellets were analyzed by western blotting using mouse anti-FLAG antibody (Sigma).
HeLa cells were cultured in DMEM (Invitrogen) supplemented with 10% heat-inactivated fetal bovine serum (Sigma) at 37 °C with 5% CO2. Cells were transfected using Fugene HD (Roche) transfection agent according to manufacturer’s protocol.
HeLa cells were fixed in 3.2% paraformaldehyde after treatment. Nuclei were stained with Hoechst (Sigma) and actin cytoskeleton was stained with rhodamine-phalloidin (Molecular Probes). Cells were viewed on a Zeiss LSM 510 scanning confocal microscope and images were converted using ImageJ and Adobe Photoshop.
HeLa cells were infected at an MOI of 10. At each time point, cells were washed and lysates were collected by scraping into Mg2+ lysis buffer (20mM Tris HCl pH 7.5, 10mM MgCl2, 150mM NaCl, 1% Triton X-100). Cellular debris was spun down at 12,000g for 20 minutes and 500µg of cleared lysates was added to 30µg of GST-PAK PBD on glutathione agarose beads and incubated for one hour at 4°C. Samples were separated by SDS-PAGE and immunoblotted with anti-CDC42 (Cell Signaling). As a loading control, total cell lysates were immunoblotted with anti-aldolase (Santa Cruz Biotechnology). Transfected HeLa cells were collected and treated in a similar fashion as described above. Activated RhoA was pulled down using a RhoA activation kit (Cytoskeleton) according to manufacturer’s instructions.
pGex-KG-PAK-PBD (provided by Dr. Neal Alto), pGex-KG-RhoA, pGex-KG-Rac, pGex-KG-Rac Q61L, pGex-KG-CDC42 (provided by Dr. Paul Sternweis), pET28a-MBP-VopCΔ67 and pET28a-MBP-VopCΔ67 C220S were transformed into BL21 (DE3) (Novagen). His-tagged proteins (MBP-VopCΔ67 and MBP-VopCΔ67 C220S) were purified using Ni2+ affinity purification (Qiagen). GST-tagged proteins (GST-PAK-PBD, GST-RhoA, GST-Rac, GST-Rac Q61L and GST-CDC42) were purified using glutathione agarose beads (Sigma).
3 mg/ml of GST-RhoA, GST-Rac, GST-Rac Q61L or GST-CDC42 were incubated with 1:20 molar ratio of MBP-VopCΔ67 or MBP-VopCΔ67 C220S and 5 mM dansylcadaverine (Sigma) at 37 °C for 2 hours. Samples were boiled and separated on 12% SDS-PAGE. Gels were examined with UV exposure using an AlphaImager 2200 (Alpha Innotech Corporation) and images were recorded. The gels were stained with Coomassie Brilliant Blue (Biorad) for protein bands.
120 µM of GST-Rac or GST-Rac Q61L were incubated with a 1:20 molar ratio of MBP-VopCΔ67 or MBP-VopCΔ67 C220S at 37°C for 2 hours. The released NH4+ from glutamine deamidation was measured using an ammonia assay kit (Sigma) according to the manufacturer’s instructions.
pGex-KG-Rac alone or pGex-KG-Rac and pET28a-MBP-VopCΔ67 together were transformed into BL21 (DE3) (Novagen) for protein expression. GST-Rac was purified using glutathione agarose beads (Sigma). Proxeon nano-tips (Denmark) were used to inject the samples into a QStar XL Q-TOF mass spectrometer (Applied Biosystems, Framingham, MA). Spectra were acquired with mass range m/z 500–2000. The molecular weights of proteins were calculated using the Bayesian Protein Reconstruct tool of the Analyst QS1.1 software.
Ten micrograms of each sample were fractionated by SDS-PAGE and bands containing GST-Rac were cut out, treated with DTT and iodoacetamide and digested with chymotrypsin. Samples from the digests were analyzed by nano-LC/MS/MS using a system in which a Dionex LC-Packings HPLC (Sunnyvale, CA) was coupled with an LTQ Orbitrap Velos (Thermo Scientific).
The V. cholerae 1587ΔvopC strain was generated essentially as described (Merriam et al., 1997; Walker and Miller, 2004). Briefly, regions 1 kb upstream and downstream of VopC were cloned into pSR47S, a KanR OriR6K suicide plasmid (kindly provided by Dr. Virginia Miller). pSR47s was then conjugated into V. cholerae 1587 and transconjugants were selected by replating twice on Vibrio-selective medium (VSM) containing 60 µg/ml kanamycin. Bacteria were counter-selected by growing on media containing sucrose.
We thank for insightful discussions, critical reading and/or generous supply of reagents N. Alto, T. Iida, T. Honda, L. McCarter, M. Phillips, V. Miller, J. Mekalanos and the Orth lab. K.O., L.Z., C.A.B., and A.M.K. are supported by grants from NIH and the Welch Foundation (I-1561). K.O is a Beckman Young Investigator, Burroughs Wellcome Investigator, and W.W. Caruth Biomedical Scholar.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.