Identification of the Minimal Intracellular Vacuolating Domain of the Helicobacter pylori Vacuolating Toxin*

Helicobacter pylori secretes a cytotoxin (VacA) that induces the formation of large vacuoles originating from late endocytic vesicles in sensitive mammalian cells. Although evidence is accumulating that VacA is an A-B toxin, distinct A and B fragments have not been identified. To localize the putative catalytic A-fragment, we transfected HeLa cells with plasmids encoding truncated forms of VacA fused to green fluorescence protein. By analyzing truncated VacA fragments for intracellular vacuolating activity, we reduced the minimal functional domain to the amino-terminal 422 residues of VacA, which is less than one-half of the full-length protein (953 amino acids). VacA is frequently isolated as a proteolytically nicked protein of two fragments that remain noncovalently associated and retain vacuolating activity. Neither the amino-terminal 311 residue fragment (p33) nor the carboxyl-terminal 642 residue fragment (p70) of proteolytically nicked VacA are able to induce cellular vacuolation by themselves. However, co-transfection of HeLa cells with separate plasmids expressing both p33 and p70 resulted in vacuolated cells. Further analysis revealed that a minimal fragment comprising just residues 312–478 functionally complemented p33. Collectively, our results suggest a novel molecular architecture for VacA, with cytosolic localization of both fragments of nicked toxin required to mediate intracellular vacuolating activity.


From the Department of Biology and Biochemistry, University of Houston, Houston, Texas 77204-5513
Helicobacter pylori secretes a cytotoxin (VacA) that induces the formation of large vacuoles originating from late endocytic vesicles in sensitive mammalian cells. Although evidence is accumulating that VacA is an A-B toxin, distinct A and B fragments have not been identified. To localize the putative catalytic A-fragment, we transfected HeLa cells with plasmids encoding truncated forms of VacA fused to green fluorescence protein. By analyzing truncated VacA fragments for intracellular vacuolating activity, we reduced the minimal functional domain to the amino-terminal 422 residues of VacA, which is less than one-half of the full-length protein (953 amino acids). VacA is frequently isolated as a proteolytically nicked protein of two fragments that remain noncovalently associated and retain vacuolating activity. Neither the amino-terminal 311 residue fragment (p33) nor the carboxyl-terminal 642 residue fragment (p70) of proteolytically nicked VacA are able to induce cellular vacuolation by themselves. However, cotransfection of HeLa cells with separate plasmids expressing both p33 and p70 resulted in vacuolated cells. Further analysis revealed that a minimal fragment comprising just residues 312-478 functionally complemented p33. Collectively, our results suggest a novel molecular architecture for VacA, with cytosolic localization of both fragments of nicked toxin required to mediate intracellular vacuolating activity.
The secreted peptide is often proteolytically nicked into aminoand carboxyl-terminal fragments, which remain associated by noncovalent interactions and retain vacuolating activity (5).
VacA exhibits cellular behavior similar to that of protein toxins with intracellular targets (24 -26). Most intracellularly acting toxins possess an overall structure corresponding to the A-B family of bacterial exotoxins, which includes diphtheria, cholera, and anthrax toxins (27). In general, the B component of an A-B toxin binds to specific mammalian cell-surface receptors and facilitates the membrane translocation of an enzymic A moiety into the cytosol. The A moiety covalently modifies a specific intracellular target molecule, resulting in the disruption of important cellular functions. VacA has recently been shown to bind to specific, high affinity cell surface receptors (28 -30). The toxin is internalized inside vesicular-type compartments by an unknown mechanism (26) and subsequently mediates the formation of large acidic vacuoles comprising hybrid endocytic vesicles originating from late endosomal and lysosomal compartments (31)(32)(33)(34). By transfecting HeLa cells with plasmids expressing VacA and thereby eliminating the necessity for receptor binding and membrane translocation, de Bernard and co-workers (24) demonstrated that VacA functions from within the host cell cytosol. Significantly, vacuolation of host cells was observed before VacA was detectable in the cells, indicating that the potency of the toxin may be attributed to a discrete enzymatic moiety, which amplifies the effects of the toxin within the cell (27,35). Although these data strongly suggest that VacA is an A-B toxin, distinct A and B fragments have not been identified.
Because a catalytic function has not yet been identified for VacA, it is not currently feasible to identify the putative A-fragment by testing truncated toxin fragments for enzymatic activity. However, we hypothesized that the A-fragment would induce vacuolation from the host cell cytosol in the absence of the B fragment of the toxin. To localize the putative VacA A-fragment, we genetically constructed amino and carboxyl-terminal truncations of VacA. Plasmids encoding these fragments were used to transfect HeLa cells, which were analyzed for vacuolation. These investigations revealed that a discrete VacA fragment can function from the host cell cytosol to induce vacuoles with properties similar to those caused by full-length toxin added externally to cells. In addition, we discovered that determinants from both fragments of proteolytically nicked toxin can function in trans within the host cell cytosol to induce degenerative vacuolation. Finally, we discuss the implications of these findings with respect to the molecular organization of VacA. Importantly, the identification of a discrete A fragment may be a critical step preceding discovery of the enzymatic activity of the toxin, because a number of bacterial toxins, including diphtheria and cholera, are secreted as proenzymes, which require activation to elaborate their catalytic activities (36,37).

EXPERIMENTAL PROCEDURES
Materials-Cell culture medium, fetal bovine calf serum, and neutral red were purchased from Life Technologies, Inc. Restriction enzymes and T4 DNA ligase were purchased from New England Biolabs (Beverly, MA). High fidelity Pfu DNA polymerase and deoxyribonucleotides were obtained from Stratagene (La Jolla, CA). The plasmid pET20b was obtained from Novagen (Madison, WI). Vaccinia virus expressing T7 polymerase (ATCC 2153-VR) and H. pylori strain 60190 (ATCC 49503) were purchased from American Type Culture Collection (Manassas, VA). Bafilomycin A1 and monensin were obtained from Sigma and Calbiochem, respectively.
Preparation of Plasmids Encoding VacA and Fragments of VacA-Standard protocols were utilized for isolation of plasmid DNA, restriction endonuclease digestions, the polymerase chain reaction, subcloning, and transformation of Escherichia coli (38). The gene for VacA was amplified directly from H. pylori strain 60190 genomic DNA using the polymerase chain reaction and cloned into pTRE (CLONTECH; Palo Alto, CA), which was the parent plasmid used as template for all the VacA-GFP fusions and VacA fragments generated for these investigations.
The gene encoding enhanced green fluorescence protein (GFP) was extracted from pEGFP-N3 (CLONTECH; Palo Alto, CA) and subcloned into pET20b replacing the BamHI and NotI fragment in the polyclonal region. All the VacA fragments were cloned into the modified pET20b-GFP, or pET-20b alone, which harbors a T7 promoter, replacing the NdeI-BamHI fragment of the polyclonal region.
Polymerase chain reaction primers were designed for the two fragments of nicked VacA based on previous reports describing the cleavage site between the two fragments of the nicked protein (5,39). These two fragments comprise residues 1-311 and 312-953, and are referred to as p33 and p70 respectively based on their calculated molecular weights of approximately 33 and 70 kDa. Plasmids were recovered from transformed E. coli XL1-blue, purified, and used for HeLa cell transfections as described below. Full-length VacA and all VacA fragments were sequenced across the length of the entire open reading frame using the Thermo Sequenase dye terminator cycle sequencing pre-mix kit from Amersham Pharmacia Biotech.
Cell Culture-HeLa cells were cultured as monolayers in 25-ml plastic flasks (Corning; Cambridge, MA) in Dulbecco's modified minimal essential medium (DMEM) 1 containing 10% fetal bovine calf serum under 5% CO 2 at 37°C. Twenty-four h before experiments, cells were seeded in 96-well titration plates in DMEM, 10% fetal bovine calf serum at a density of 1 ϫ 10 5 cells/ml.
Transfection of Cells-HeLa cells were plated (200 l) at a density of 1.0 ϫ 10 5 cells/ml in 96-well tissue culture plates (Corning; Cambridge, MA) in DMEM supplemented with 2.5% fetal bovine calf serum, 100 units penicillin/ml, and 100 mg of streptomycin/ml. HeLa cells were first infected with recombinant vaccinia virus (vT7) bearing the T7 RNA polymerase gene (39). Vaccinia virus stock was trypsinized at 37°C for 30 min and added to HeLa cells (24). After infection for 30 min, virus stock was removed, and the HeLa cells were transfected using the calcium phosphate method (38). After the transfection procedure, the cells were incubated in DMEM ϩ 5 mM NH 4 Cl at 37°C for 20 h before analysis.
Analysis of Transfected HeLa Cells-We quantified relative vacuolation based on the HeLa cell uptake of the dye neutral red as described previously (40). The experiments were performed in 96-well plates, and neutral red uptake was determined using a Dynatech MR5000 microtiter plate reader to measure the absorbances at 530 nm (minus the absorbance at 410 nm).
To confirm green fluorescence protein (GFP) fluorescence in transfected HeLa cells, the cells were visualized using a Nikon Diaphot inverted microscope outfitted with a Xenon lamp and fluorescence filter combinations for fluorescence microscopy.
Pharmacological Reagent Effects on Vacuolating Activity-Two h after transfection of HeLa cells, transfection reagents were replaced with DMEM and 5 mM NH 4 Cl supplemented with the indicated concentrations of bafilomycin A1 or monensin and incubated for 20 h under 5% CO 2 at 37°C.

Expression of VacA Fragments in HeLa Cells under Control
of the T7 Promoter-The minimal VacA vacuolating domain was mapped by transfecting HeLa cells with pET-20b harboring genes encoding either the mature VacA peptide (residues 1-953) cloned from the 60190 toxigenic strain of H. pylori or truncated fragments of VacA fused to GFP. To improve protein expression, HeLa cells were first infected with recombinant vaccinia virus (vT7) bearing the gene for phage T7 RNA polymerase followed by transfection with pET-20b encoding VacA fragments fused to GFP (24,39). Using this system, 50 -70% of the cells clearly demonstrated GFP fluorescence (data not shown), which appeared essentially identical to previously published data (24). In HeLa cells transfected with VacA-GFP, vacuolation was observed only in those cells demonstrating GFP fluorescence. VacA-GFP fusions could be detected by GFP fluorescence 5-8 h after transfection, whereas vacuoles were already clearly visible after 4 h, also confirming previously reported results that vacuolation was detected before GFP fluorescence could be visualized in the cell cytosol (24).
Carboxyl-terminal Truncations of VacA-Plasmids were constructed to express multiple VacA fragments lacking portions of the carboxyl terminus. These deletions yielded VacA aminoterminal fragments of 349, 373, 394, 422, 478, 538, 670, and 741 amino acids with or without GFP fusions (Fig. 1B). Purified plasmids expressing either full-length toxin or truncation mutants were used to transfect HeLa cells, which had been previously infected with vT7. Twenty h after transfection, HeLa cells were analyzed for both GFP fluorescence and uptake of neutral red.
Analysis of the truncation mutants revealed that when VacA is directly expressed in the host cell cytosol, the carboxyl terminus is not essential for inducing vacuole formation (Fig. 1). Residues 1-478 (475 carboxyl-terminal deletion of VacA) fused to GFP induced vacuolation of HeLa cells, whereas a fragment comprising VacA residues 1-422 fused to GFP (531 carboxylterminal deletion of VacA) was unable to induce detectable vacuolation (Fig. 1A). To test if the carboxyl-terminal GFP fusion interfered with the vacuolating activity of the 422-residue VacA fragment, this fragment was also expressed in the HeLa cell cytosol without the GFP fusion. Although the 422residue VacA fragment with a free carboxyl terminus was sufficient to mediate vacuolation, neither a shorter VacA fragment comprising residues 1-394 nor any smaller fragment induced vacuolation (Fig. 1B).
Amino-terminal Truncations of VacA-To determine the importance of the VacA amino terminus in the intracellular activity of the toxin, a series of amino-terminal truncation mutants fused to GFP were generated comprising residues 18 -741, 57-741, and 198 -741 (Fig. 2B). The plasmids encoding these fragments were used to transfect HeLa cells after infection of the cells with vT7. The extent of vacuolation was determined by quantifying neutral red uptake. The expression of VacA fragments fused to GFP was confirmed using fluorescence microscopy. None of the amino-terminal truncations were able to induce intracellular vacuolation (Fig. 2). Deletion of only 17 residues resulted in the loss of VacA-mediated vacuolating activity ( Fig. 2A), indicating that in contrast to the carboxyl terminus, the VacA amino terminus appears to be essential for inducing intracellular vacuolation.
Co-transfection of HeLa Cells with Multiple VacA Fragments-VacA is often isolated as a proteolytically nicked protein of two fragments comprising residues 1-311 and 312-953 of the mature protein (p33 and p70, respectively) (5, 41). As previously reported, HeLa cells transfected with plasmids encoding either of these fragments separately were not vacuolated (24). To determine whether p33 and p70 can induce vacuolation when introduced together into cells, we co-transfected HeLa cells with separate plasmids encoding these two fragments. These co-transfection experiments revealed that the two fragments were able to functionally complement each other to induce fully vacuolated cells (Fig. 3). Notably, GFP fused to the carboxyl terminus of p33 did not interfere with p70 complementation. If VacA is an A-B toxin, this would be the first example in which two distinct fragments functionally complement each other from within the host cell cytosol to mediate toxin activity.
Having demonstrated that more than 500 amino acids at the carboxyl terminus of nonnicked VacA are not required for intracellular vacuolation (Fig. 1), we next tested whether the entire p70 carboxyl terminus was required when co-expressed with p33. Plasmids encoding smaller fragments of p70 were designed based on the carboxyl-terminal truncations described above. Co-transfection experiments revealed that a 167-residue VacA fragment comprising amino acids 312-478 with or without GFP fusions functionally complemented inactive p33 to induce intracellular vacuolation (Fig. 3A). However, the smaller 111-residue p70 fragment comprising VacA residues 312-422 did not complement p33, regardless of whether or not the fragment was expressed as a fusion protein with GFP.

Vacuoles Induced by Cytosolic VacA Fragments Respond to the Same Reagents as Vacuoles Induced by Externally Added
Toxin-Ammonium ions potentiate VacA-mediated cellular vacuolation regardless of whether toxin is added externally or introduced directly into the cytosol of HeLa cells (42). The potentiation mechanism is believed to involve osmotic swelling of vacuolar vesicles, which have taken up NH 4 Cl subsequent to a VacA-induced perturbation of cellular endocytic trafficking (42). As shown in Fig. 4, when HeLa cells were transfected with the plasmid encoding the minimal functional VacA fragment (residues 1-422), vacuole formation was potentiated by the presence of 5 mM NH 4 Cl. Likewise, in co-transfection experiments in which VacA fragments were expressed from separate plasmids, vacuolation of HeLa cells was also potentiated by 5 mM NH 4 Cl (Fig. 4). In the absence of NH 4 Cl, VacA fragments Cell vacuolation induced by VacA is strictly dependent on the function of vacuolar-type ATPase proton pumps (34). To determine whether vacuoles formed by the fragments of VacA expressed in the cytosol are similar to those produced by fulllength VacA, we transfected HeLa cells with plasmids expressing full-length VacA fused to GFP or the active VacA fragment comprising residues 1-422. In addition, we co-transfected HeLa cells with plasmids expressing p33-GFP and p70-GFP. We introduced bafilomycin A1 (2-25 nM), which targets vacuolar-type ATPase proton pumps and has been shown to inhibit VacA-induced cellular vacuolation (43). As shown in Fig. 5A, 5-10 nM bafilomycin A1 inhibited neutral red uptake by approximately 50%. Increasing, the drug to 25 nM nearly abolished vacuole formation, indicating that the vacuoles formed intracellularly by minimal VacA fragments also require active vacuolar-type ATPase proton pumps.
Monensin was also reported in earlier studies to inhibit vacuole formation when the toxin was added externally to the cells (42). To compare whether monensin inhibits the intracellular activity of VacA as well as the activity of externally added toxin, we tested the effects of monensin (0.2-5.0 g/ml) on these same VacA fragments expressed from the cytosol of HeLa cells. Monensin addition yielded a concentration-dependent blockage of VacA-induced vacuolation (Fig. 5B); 0.75-2.0 g/ml monensin inhibited neutral red uptake by approximately 50%, and 5 g/ml almost completely inhibited neutral red uptake. Collectively, these results indicate that vacuoles induced by VacA fragments functioning directly from within the host cell cytosol are similar to those induced by full-length toxin.

DISCUSSION
In the absence of a defined biochemical activity for VacA, it is not currently possible to localize an isolated catalytic domain using enzyme activity assays to analyze smaller toxin fragments. However, we expressed VacA fragments in the cytosol of HeLa cells to determine the minimal domain, which induces vacuolation similar to whole toxin added externally to cells. By eliminating the necessity for receptor binding and translocation, we hypothesized that the B fragment of the toxin could be entirely removed without affecting the ability of the toxin to act intracellularly. Other toxin A fragments have been shown to be entirely functional when translocated into host cells by alternative means in the absence of the B fragment of the toxin. For example, catalytically active A fragments of diphtheria toxin, Pseudomonas exotoxin A, and shiga toxin have all been successfully introduced into mammalian cells using the B fragment of anthrax toxin (44 -46).
Our investigations demonstrated that an amino-terminal fragment of approximately 422 residues comprises a minimal VacA domain that can mediate vacuole formation from the host cell cytosol. More than 50% of the 953 amino acids can be deleted from the VacA carboxyl terminus without affecting intracellular vacuolating activity, supporting the hypothesis that VacA contains a discrete A domain responsible for catalytic activity. Furthermore, truncations of greater than 17 residues from the amino-terminal end of VacA abolished vacuolating activity, emphasizing the importance of the amino terminus to vacuolating activity. Although HeLa cells transfected with plasmids encoding truncated VacA-GFP fusion proteins exhibited fluorescence, it cannot be ruled out at this time that the lack of detectable cellular vacuolation is because of amino-terminal loss or degradation by way of selective proteolysis. Based on our data, we propose that the putative A-fragment of VacA is located in the amino-terminal half of the toxin, whereas the B-fragment receptor binding and translocation determinants comprise the carboxyl-terminal half of the toxin.
We isolate VacA from H. pylori primarily as two distinct fragments (p33 and p70) that remain noncovalently associated and retain vacuolating activity subsequent to proteolytic nicking at amino acid 311 (5,47). Although it was initially speculated that p33 and p70 may constitute A and B fragments of the toxin, neither p33 nor p70 alone were able to induce vacuolation when HeLa cells were transfected with plasmids separately encoding these genes. This suggests that the A fragment of the toxin may include domains from both fragments (25). Our results support this proposal, indicating that both p33 and the amino terminus of p70 are required to intracellularly induce vacuolation. In single transfection experiments, 422 residues of VacA were essential for vacuolating activity. However, in co-transfection experiments, a larger fragment of p70 (residues 312-478) was required to complement inactive p33. It is not known whether residues 423-478 are required for direct interaction with p33 or alternatively, if the relatively short 111-amino acid polypeptide (residues 312-422) is unable to fold into a functional domain.
Our results suggest an interesting and perhaps unique molecular structure for VacA among the A-B toxins. Complementation of inactive p33 with a 167-residue fragment of p70 indicates that both fragments are essential. However, it is unclear whether both fragments comprise essential domains of the putative catalytic core of the toxin. We also show that GFP fused to the p33 carboxyl terminus does not interfere with functional complementation by p70, suggesting that the essential determinants of toxin activity are not contiguous across the p33-p70 cleavage site at residue 311. In addition, these results are consistent with a recently published report that 46 residues at the interface between the two VacA fragments could be eliminated without affecting toxin activity (48). Thus, if VacA possesses a discrete catalytic domain, it is unlikely that it spans the p33-p70 interface. It is possible that a putative VacA catalytic domain may be composed of noncontiguous domains requiring assembly before elaborating the catalytic activity of the enzyme. Although some enzymes can be dissected and the fragments functionally reassembled (23,19,49), this has not previously been demonstrated for any toxin A-fragment. Alternatively, p33 and p70 could conceivably possess independent functions, both of which are required for mediating an undefined mechanism to induce formation of vacuoles within the host cell cytosol. FIG. 5. Inhibition of intracellular vacuolating activity with bafilomycin A1 and monensin. Assays were performed as described under "Experimental Procedures." The data were averaged from three separate experiments performed at least in triplicate and are illustrated as a percentage of neutral red uptake of HeLa cells transfected with plasmids in the absence of bafilomycin A1 or monensin. HeLa cells were transfected with plasmids expressing VacA-GFP (q), VacA (residues 1-422) (E), and co-transfected with separate plasmids expressing p33-GFP and p70-GFP (ϫ).
In summary, we have provided the first direct evidence for localization of the putative VacA A-fragment to the aminoterminal half of the toxin. Furthermore, determinants from both fragments of proteolytically nicked VacA are required for the elaboration of toxin activity from within the host cell cytosol. Investigations are ongoing to further localize a discrete catalytic core. Mapping this functional domain as well as the receptor binding and translocation domains, will not only be important to begin understanding the mechanism of VacA cytotoxicity and toxin structure-function relationships but will also be critical to identify stable, nontoxic subdomains as immunogens for incorporation into protective vaccines against H. pylori.
Acknowledgments-We thank Drs. Amy Sater and Patrick Callaerts for their assistance with the fluorescence microscopy to confirm expression of VacA-GFP fusions in transfected HeLa cells. In addition, special thanks goes to Drs. Art Vailas and Daniel Martinez for the generous use of their laboratory's Dynatech MR5000 microtiter plate reader. We also acknowledge Donald Mahoney for helpful discussions in the early phases of this work.