Functional Properties of the p33 and p55 Domains of the Helicobacter pylori Vacuolating Cytotoxin*

Helicobacter pylori secretes an 88-kDa vacuolating cytotoxin (VacA) that may contribute to the pathogenesis of peptic ulcer disease and gastric cancer. VacA cytotoxic activity requires assembly of VacA monomers into oligomeric structures, formation of anion-selective membrane channels, and entry of VacA into host cells. In this study, we analyzed the functional properties of recombinant VacA fragments corresponding to two putative VacA domains (designated p33 and p55). Immunoprecipitation experiments indicated that these two domains can interact with each other to form protein complexes. In comparison to the individual VacA domains, a mixture of the p33 and p55 proteins exhibited markedly enhanced binding to the plasma membrane of mammalian cells. Furthermore, internalization of the VacA domains was detected when cells were incubated with the p33/p55 mixture but not when the p33 and p55 proteins were tested individually. Incubation of cells with the p33/p55 mixture resulted in cell vacuolation, whereas the individual domains lacked detectable cytotoxic activity. Interestingly, sequential addition of p55 followed by p33 resulted in VacA internalization and cell vacuolation, whereas sequential addition in the reverse order was ineffective. These results indicate that both the p33 and p55 domains contribute to the binding and internalization of VacA and that both domains are required for vacuolating cytotoxic activity. Reconstitution of toxin activity from two separate domains, as described here for VacA, has rarely been described for pore-forming bacterial toxins, which suggests that VacA is a pore-forming toxin with unique structural properties.

Helicobacter pylori secretes an 88-kDa vacuolating cytotoxin (VacA) that may contribute to the pathogenesis of peptic ulcer disease and gastric cancer. VacA cytotoxic activity requires assembly of VacA monomers into oligomeric structures, formation of anion-selective membrane channels, and entry of VacA into host cells. In this study, we analyzed the functional properties of recombinant VacA fragments corresponding to two putative VacA domains (designated p33 and p55). Immunoprecipitation experiments indicated that these two domains can interact with each other to form protein complexes. In comparison to the individual VacA domains, a mixture of the p33 and p55 proteins exhibited markedly enhanced binding to the plasma membrane of mammalian cells. Furthermore, internalization of the VacA domains was detected when cells were incubated with the p33/p55 mixture but not when the p33 and p55 proteins were tested individually. Incubation of cells with the p33/p55 mixture resulted in cell vacuolation, whereas the individual domains lacked detectable cytotoxic activity. Interestingly, sequential addition of p55 followed by p33 resulted in VacA internalization and cell vacuolation, whereas sequential addition in the reverse order was ineffective. These results indicate that both the p33 and p55 domains contribute to the binding and internalization of VacA and that both domains are required for vacuolating cytotoxic activity. Reconstitution of toxin activity from two separate domains, as described here for VacA, has rarely been described for pore-forming bacterial toxins, which suggests that VacA is a pore-forming toxin with unique structural properties.
Helicobacter pylori is a Gram-negative bacterium that chronically infects the human stomach (1)(2)(3). A major secreted protein produced by H. pylori is an 88-kDa cytotoxin, known as VacA (4 -7). VacA is considered an important virulence factor in the pathogenesis of peptic ulceration and gastric cancer (6 -9). Over the past decade, it has become clear that VacA causes many different effects on mammalian cells (7,10). The most prominent effect of VacA is its capacity to induce vacuo-lation in epithelial cells (4,5). Other reported effects of VacA include depolarization of the membrane potential (11)(12)(13), apoptosis (14,15), detachment of epithelial cells from the basement membrane (16), interference with the process of antigen presentation (17), activation of mitogen-activated protein kinases (18,19), and inhibition of activation-induced proliferation of T lymphocytes (19 -21). Many of these effects are dependent on the capacity of VacA to form anion-selective membrane channels (13,15,(21)(22)(23)(24).
VacA is expressed in H. pylori as a 140-kDa protoxin that undergoes amino-and carboxyl-terminal processing, yielding a mature 88-kDa secreted VacA toxin (5,6,(25)(26)(27). The 88-kDa VacA monomers can assemble into large water-soluble flowershaped structures comprising 6 -14 monomers (28 -30). The assembly of VacA into oligomeric structures is likely to be required for membrane channel formation, as well as for many effects of VacA on mammalian cells (31).
Several cell-surface receptors for VacA have been reported, including two receptor protein tyrosine phosphatases (RPTP-␣ and -␤) (16,(32)(33)(34), the epidermal growth factor receptor (35), and heparan sulfate (36). VacA also localizes to lipid raft domains on the surface of mammalian cells (12,37,38). Following binding of VacA to the plasma membrane, the toxin is internalized by cells (39,40). Internalized VacA can localize to late endocytic compartments (from which vacuoles arise) as well as mitochondria (24,41). It has been proposed that VacA causes cell vacuolation by forming anion-selective channels in the membrane of endocytic compartments (11,41) and that it increases mitochondrial membrane permeability by forming channels in the inner membrane of mitochondria (24).
Partial proteolytic digestion of the mature secreted 88-kDa VacA toxin yields two fragments that are ϳ33 and 55 kDa in mass (p33 and p55, respectively) (see Fig. 1A). This proteolytic cleavage occurs primarily between amino acids 311 and 312 of the mature secreted toxin from H. pylori strain 60190 (and possibly several adjacent sites) (27), which are predicted to comprise a hydrophilic surface-exposed loop of VacA. It has been suggested that the p33 and p55 fragments represent two domains or subunits of VacA (6,28,31,42). Several lines of evidence indicate that amino acid sequences located within a hydrophobic region near the amino terminus of the p33 domain are required for the formation of anion-selective membrane channels (13,23,43,44), and that the p55 domain is responsible for VacA binding to mammalian cells (39,(45)(46)(47). However, detailed analysis of the functional roles of p33 and p55 domains has not been performed. In this work, we describe the expression and functional analysis of recombinant p33 and p55 domains. Our data indicate that neither the p33 nor the p55 recombinant VacA domain exhibits detectable vacuolating cytotoxic activity when added individually to the surface of mammalian cells. In contrast, we demonstrate that when combined, the p33 and p55 domains form p33⅐p55 protein complexes and complement each other to permit VacA internalization into cells and reconstitution of vacuolating cytotoxic activity.

EXPERIMENTAL PROCEDURES
Bacterial Strains and Growth Conditions-Escherichia coli DH5␣ was used for plasmid propagation and was grown in Luria-Bertani (LB) broth or on LB agar at 37°C. For expression of recombinant proteins, expression plasmids were transformed into E. coli strain JM109 (DE3) (Promega), which encodes an isopropyl-␤-D-thiogalactopyranoside-inducible copy of the RNA polymerase gene from bacteriophage T7. Transformants were grown in Terrific broth (Invitrogen) supplemented with 25 g of kanamycin/ml (TB-KAN). H. pylori wild-type strain 60190 (American Type Culture Collection 49503; Manassas, VA) and strain VT330 (encoding a c-Myc-tagged VacA protein) (48) were grown on Trypticase soy agar plates containing 5% sheep blood at 37°C in ambient air containing 5% CO 2 . H. pylori liquid cultures were grown in sulfite-free brucella broth supplemented with 0.5% activated charcoal (49).
Purification of VacA from H. pylori-VacA was purified in an oligomeric form from culture supernatants of H. pylori, as described previously (28). In all experiments, purified VacA preparations from H. pylori were acid-activated before the addition of VacA to the cells, as described previously (40,50).
Plasmid Construction-VacA-expressing plasmids were constructed by cloning vacA sequences from H. pylori strain 60190 into pET-41b (Novagen) using procedures similar to those described previously (51). A vacA sequence encoding the VacA p33 domain (amino acids 1-312 of the mature, secreted H. pylori VacA toxin) with a His 6 tag at the carboxyl terminus (p33His) was PCR-amplified from the pMM592 plasmid (51) using primers BA9146, 5Ј-CCCACTAGTAAGAGGAGAC-GCCATGTTTTTTACAACCGTG-3Ј, and OP6228, 5Ј-CCCCTGCAGCT AGTGATGGTGATGGTGATGTTTAGCACCACTTTGAGAAGG-3Ј. The PCR product was digested with SpeI and SalI and ligated into XbaIand SalI-digested pET-41b (conferring kanamycin resistance) (Novagen). We also generated a plasmid that encoded a VacA p33 domain with two tags (c-Myc and His 6 ), each at the carboxyl terminus of the protein (p33Myc-His). A vacA sequence encoding the p55 domain (amino acids 312-821 of the mature, secreted H. pylori VacA toxin) with a c-Myc tag (p55Myc) at the amino terminus was PCR-amplified from H. pylori VT330 genomic DNA using primers OP6229, 5Ј-CC-CACTAGTAAGAGGAGACGCCATGGCAAACGCCGCACAGG-3Ј and AND515a, 5Ј-CCCCGTCGACTTAAGCGTAGCTAGCGAAACGCG-3Ј. Also, a vacA sequence encoding the p55 domain with a His 6 tag at the amino terminus was generated using primers AND7265, 5Ј-CCCAC-TAGTAAGAGGAGACGCCATGCATCACCATCACCATCACAAAAAC-GACAAACAAGAGAGC-3Ј and the AND515a primer. PCR products were digested and cloned into pET-41b, as described above. The use of these primers resulted in a modification of the ribosomal binding site of pET-41b and encoded a methionine at the amino terminus of each VacA protein. The entire vacA fragment in each plasmid was analyzed by nucleotide sequence analysis to verify that no unintended mutations had been introduced.
Expression of Recombinant VacA Proteins-VacA expression plasmids were transformed into the E. coli expression strain JM109 (DE3), and transformants were then inoculated into TB-KAN and grown at 37°C overnight with shaking. These cultures were diluted 1:100 into TB-KAN and incubated at 37°C until they reached an absorbance (A 600 ) of 0.5. Cultures were then induced with a final isopropyl-␤-D-thiogalactopyranoside concentration of 0.5 mM and incubated at 25°C for 16 -18 h (p55 proteins) or at 37°C for 2 h (p33 proteins). These varying conditions for isopropyl-␤-D-thiogalactopyranoside induction were selected to optimize expression and activity of the two different VacA domains.
E. coli soluble extracts were generated as described previously with minor modifications (51). Briefly, 50 ml of isopropyl-␤-D-thiogalactopyranoside-induced cultures were pelleted, washed in 0.9% NaCl, and resuspended in a solution (1 ml) that contained 10 mM Tris (pH 7.5), 100 mM NaCl, 1 mM EDTA, protease inhibitors (Complete Mini; Roche Applied Science), and 20,000 units of ReadyLyse lysozyme (Epicenter)/ ml. Bacterial cells were incubated on ice for 30 min with periodic mixing, after which a solution (3 ml) containing 50 mM Tris (pH 8.0), 2.67 mM MgCl 2 , and 74 units of Omnicleave Nuclease (Epicenter)/ml was added. The samples were then mixed briefly, subjected to four successive rounds of freezing (in a dry ice methanol bath) and thawing at 37°C, and then the insoluble debris was pelleted. The E. coli soluble extracts containing the VacA proteins were collected and stored at Ϫ20°C until use.
Immunoblot Analysis of Recombinant VacA-Proteins in E. coli soluble extracts were separated by SDS-polyacrylamide gel electrophore-sis, transferred to nitrocellulose membrane, and immunoblotted using a polyclonal anti-His antibody (Santa Cruz Biotechnology), a monoclonal anti-c-Myc (9E10) antibody, or a polyclonal anti-VacA serum (958) (12), followed by secondary antibodies conjugated with horseradish peroxidase (Bio-Rad). Signals were generated by the enhanced chemiluminescence reaction (Amersham Biosciences) and detected using x-ray film. In experiments requiring the use of multiple recombinant VacA proteins, the relative concentrations of recombinant VacA in different E. coli soluble extracts were analyzed by immunoblotting with an anti-His antibody, and the extracts were then normalized such that the relative molar concentrations of VacA in different preparations were approximately equivalent.
Cell Culture and Vacuolating Assay-HeLa cells were grown in minimal essential medium (modified Eagle's medium containing Earle's salts) supplemented with 10% fetal bovine serum in a 5% CO 2 atmosphere at 37°C. AZ-521 cells, a human gastric adenocarcinoma cell line (Culture Collection of Health Science Research Resources Bank, Japan Health Sciences Foundation) were grown in minimal essential medium supplemented with 10% fetal bovine serum and 1 mM non-essential amino acids (Invitrogen). AGS human gastric epithelial cells (American Type Culture Collection CRL 1739) were grown in RPMI 1640 medium supplemented with 10% fetal bovine serum. For vacuolating assays, the cells were seeded at 1.2 ϫ 10 4 cells/well into 96-well plates 24 h prior to each experiment. E. coli soluble extracts, containing recombinant VacA proteins, were normalized as described above and were added to fetal bovine serum-free tissue culture medium overlying the cells (supplemented with 50 g/ml gentamicin and 10 mM ammonium chloride) for 1 h at 37°C. The cells were washed two times with phosphate-buffered saline and then incubated in fetal bovine serum-free tissue culture medium, containing 10 mM ammonium chloride and gentamicin, for 4 -6 h at 37°C. For the p33/p55 complementation assays, E. coli soluble extracts were mixed and incubated for 1 h at 25°C prior to addition to the medium overlying the cells. Purified VacA from H. pylori culture supernatant was routinely acid-activated prior to testing in cell culture assays (40,50), whereas E. coli soluble extracts were not acid-activated (51). After incubation, cell vacuolation was examined by inverted light microscopy and quantified by a neutral red uptake assay (52). Neutral red uptake data are presented as A 540 values (mean Ϯ S.D.).
Immunoprecipitation of VacA Complexes-E. coli soluble extracts containing different recombinant VacA proteins were normalized as described above. Normalized soluble extracts were then mixed and incubated for 1 h at 25°C. VacA complexes were immunoprecipitated with an anti-c-Myc monoclonal antibody (2 g/ml antibody 9E10) and protein G-coated beads (Zymed Laboratories Inc.) (48). To analyze interactions between p33, p55, and p33/p55 mixture with full-length VacA, normalized soluble extracts were mixed with acid-activated c-Myc-tagged VacA (Myc-VacA) purified from H. pylori culture supernatant (2 g/ml) for 1 h at 25°C, and the proteins were immunoprecipitated as described above. Immunoprecipitated proteins were analyzed by immunoblotting with an anti-His (Santa Cruz Biotechnology) antibody or anti-c-Myc monoclonal antibody (9E10), followed by secondary antibodies conjugated with horseradish peroxidase (Bio-Rad), as described above.
Analysis of VacA Binding and Internalization into Mammalian Cells-To analyze interactions of VacA with the surface of cells, E. coli soluble extracts, containing recombinant VacA proteins, were added to HeLa cells grown on cover glasses in 6-well plates for 1 h at 4°C or 37°C. VacA interactions with mammalian cells were then analyzed by indirect immunofluorescence (12,41). Briefly, the cells were washed with Tris-buffered saline (TBS) (10 mM Tris, 150 mM NaCl, pH 7.5) and fixed with 3.7% formaldehyde. Fixed cells were incubated with an anti-c-Myc antibody (1:500) or with an anti-VacA polyclonal antiserum that recognizes the p55 domain for 1 h at 25°C. The cells were then washed and incubated with a Cy3-conjugated secondary antibody (1: 500) for 1 h at 25°C. To analyze VacA internalization into host cells, E. coli soluble extracts containing single recombinant VacA proteins or mixtures of recombinant VacA proteins were incubated with HeLa cells for 1 h at 37°C. Afterward, medium containing unbound proteins was removed, and the cells were incubated in fresh tissue culture medium (without fetal bovine serum or ammonium chloride) for 16 h at 37°C. The cells were then washed with TBS, fixed with 3.7% formaldehyde, and permeabilized with 100% methanol for 30 min at Ϫ20°C (41). The cells were incubated with the anti-VacA polyclonal antiserum or the anti-c-Myc antibody, followed by a Cy3-conjugated secondary antibody. After immunolabeling, cover glasses were washed with phosphate-buffered saline, mounted on slides with Aqua-Polymount (Polysciences, Warrington, PA), and viewed with a LSM 510 confocal laser scanning inverted microscope (Carl Zeiss).
For immunoblot analysis of VacA interactions with cells, HeLa cells were seeded into a 96-well plate and incubated with E. coli soluble extracts, as described above. Cells were then washed three times with TBS and lysed directly in the wells of the 96-well plate by adding SDS-containing buffer. The presence of VacA in these cell lysates was detected by immunoblotting, as described above.

Expression of Recombinant p33 and p55
VacA Domains-To study functional properties of the p33 and p55 VacA fragments (Fig. 1A), we expressed these two putative domains as recombinant proteins using a recently developed system that allowed expression of a functional cytotoxic form of full-length VacA in E. coli (51). Soluble E. coli extracts containing the recombinant p33 and p55 domains were generated as described under "Experimental Procedures." Both the recombinant p33 and p55 proteins were successfully expressed based on immunoblotting analysis (Fig. 1B).
p33 and p55 VacA Domains Complement Each Other for Vacuolating Activity-Previously, it has been shown that E. coli soluble extract containing the full-length 88-kDa recombinant VacA protein exhibits vacuolating cytotoxic activity when added to mammalian cells (51). Therefore, we investigated whether either the p33 or the p55 protein was capable of causing cell vacuolation activity when added to mammalian cells. No detectable vacuolating activity was observed when E. coli extracts containing the p33 or p55 protein were added individually to HeLa cells ( Fig. 2A). In contrast, when extracts containing the p33 and p55 proteins were mixed and then added to HeLa cells, extensive cell vacuolation was detected, based on results of a neutral red uptake assay and also based on light microscopic examination ( Fig. 2A and data not shown). Extensive vacuolating activity was also observed when gastric cell lines (AGS and AZ-521) were intoxicated with the p33/p55 mixture (data not shown). As seen with purified VacA, the vacuolating activity exhibited by the mixture of the recombinant p33 and p55 VacA domains was dose-dependent (Fig. 2B).
These results indicate that a mixture of recombinant p33 and p55 VacA domains is capable of reconstituting vacuolating cytotoxic activity.
p33 and p55 Domains Form Oligomeric Complexes-We hypothesized that the ability of the p33 and p55 proteins to complement each other for vacuolating activity might require the formation of protein complexes comprising these two proteins. To test whether the p33 and p55 VacA domains could physically interact, we mixed different combinations of these epitope-tagged recombinant proteins and then performed a series of immunoprecipitation experiments using an anti-c-Myc antibody, as described under "Experimental Procedures." In these experiments, p33/p55 interactions were detected (Fig. 3A,  lanes 4 and 7), whereas p55/p55 (Fig. 3A, lane 5) and p33/p33 (Fig. 3A, lane 6) interactions were not detected. These data indicate that the recombinant p33 and p55 proteins are capable of interacting in solution to form p33⅐p55 protein complexes. To determine whether the p33 and p55 domains form only simple binary complexes or also higher ordered complexes in solution, we mixed different epitope-tagged p33 and p55 domains (p33His, p55His, and p33Myc-His) and then immunoprecipitated the protein complexes using an anti-c-Myc antibody, as described above. When samples containing the p33His, p55His, and p33Myc-His proteins were mixed, they formed protein complexes comprising all three proteins (Fig. 3B, lane 3). The ability to form these p33Myc-His⅐p33His⅐p55His complexes was dependent on the p33/p55 interaction (Fig. 3B, lane 2), because p33His and p33Myc-His failed to interact when mixed in the absence of p55 (Fig. 3B, lane 1). These data indicate that the p33⅐p55 protein complex can potentially be composed of at least three independent subunits. However, we were unable to detect the assembly of p33 and p55 domains into large 1,000-kDa complexes similar to those formed by the 88-kDa secreted VacA Interactions of p33 and p55 Domains with Full-length 88-kDa VacA-We next investigated whether the recombinant p33 and p55 proteins could physically interact with the wild-type 88-kDa VacA protein purified from H. pylori culture supernatant. For these studies, we used a c-Myc-tagged 88-kDa VacA protein (Myc-VacA) purified from H. pylori supernatant (48) and immunoprecipitation methodology. We did not detect any interaction of p33 or p55 with the 88-kDa Myc-VacA when the p33 and p55 domains were tested independently (Fig. 3C). In contrast, when the p33/p55 mixture was incubated with the 88-kDa Myc-VacA, both p33 and p55 interacted with the 88-kDa Myc-VacA protein (Fig. 3C). These data provide evidence indicating that both the p33 and p55 domains play a role in the process by which VacA assembles into oligomeric structures.
Binding of p33 and p55 VacA Domains to Mammalian Cells-In the next series of experiments, we investigated the capacity of the recombinant p33 and p55 proteins to interact with mammalian cells. As a first approach, interaction of the p33 and p55 domains with mammalian cells was investigated using immunoblotting methodology. Weak binding to HeLa cells was detected when either the p33 or p55 domain was added individually to cells for 1 h at 4°C (Fig. 4A). However, when the p33 and p55 domains were mixed and then added to HeLa cells, the amount of p33 and p55 protein associated with the cells was substantially increased compared with the amount detected when these proteins were tested individually (Fig. 4A). Similar results were observed when the VacA domains were added to cells at 37°C for 1 h instead of 4°C (data not shown), or when cell binding was assessed using gastric cell lines (AGS and AZ-521) instead of HeLa cells (data not shown).
As a second approach to investigate p33 and p55 interactions with mammalian cells, we used indirect immunofluorescence methodology. As expected, full-length 88-kDa c-Myc-tagged VacA purified from H. pylori bound to the surface of HeLa cells, and the binding could be detected with either an anti-c-Myc antibody or a polyclonal anti-VacA antiserum that recognizes the p55 domain (Fig. 4B, panels 1 and 5) (12, 39). In contrast, no immunoreactive signal on the surface of cells was detected with these antibodies following incubation of cells with negative control extracts (Fig. 4B, pET, panels 2 and 6). When recombinant p33 and p55 domains were added individually to cells, binding of p33 (either p33Myc-His or p33His) to the surface of the cells was detectable (Fig. 4B, panel 3, and data not shown), but binding of p55 was not detected (Fig. 4B, panel  7). As shown in Fig. 3A, we were able to detect binding of the p55 domain to the surface of cells by immunoblot methodology, but we were unable to detect interaction of the recombinant p55 protein with the surface of HeLa cells using immunofluorescence assays, despite testing two different forms of this protein (p55His or p55Myc) and multiple antibodies, including the anti-VacA polyclonal antiserum used in panel A. We presume that the relevant epitopes are not accessible to the antibodies under the conditions of the immunofluorescence assay. When the p33 and the p55 proteins were mixed and then added to HeLa cells, both proteins were detected on the surface of HeLa cells (Fig. 4B, panels 4 and 8). Thus, binding of p55 to the cell surface was detected by immunofluorescence assay when the p33 and p55 domains were added together to cells, but not when p55 was added independently to cells. Interestingly, the distribution of p33 on the surface of HeLa cells was punctate when p33 was added alone to cells, whereas it was continuous (non-punctate) when p33 was added to cells together with the p55 domain (Fig. 4B, panels 3 and 4). These data indicate that the interactions of p33 and p55 VacA domains with the surface of cells are substantially altered when both domains are present.
Intracellular Localization of the p33 and p55 VacA Domains-We next investigated whether the p33 and p55 proteins were internalized into mammalian cells. HeLa cells were intoxicated with either purified VacA from H. pylori or E. coli soluble extracts containing the p33 domain, the p55 domain, or the p33/p55 mixture, as described under "Experimental Procedures." Internalized VacA was visualized by indirect immunofluorescence analysis of permeabilized cells. As expected, the 88-kDa VacA purified from H. pylori was internalized into HeLa cells (Fig. 5, panel 1) (39 -41). Little if any internalization of the p33 or p55 protein was detected when these proteins were added individually to cells (Fig. 5, panels 2 and 3). In contrast, when the p33 and the p55 proteins were mixed and then added to cells, both domains were internalized (Fig. 5,  panels 4 and 5). In most experiments, internalization of the p55 protein was more readily detectable than internalization of the p33 protein, perhaps because of limited sensitivity of the antic-Myc antibody for detecting internalized p33. These data indicate that both the p33 and p55 VacA domains are required for toxin internalization into target cells.
Sequential Addition of p55 and p33 VacA Domains-To further investigate the role of p33 and p55 interactions with host cells, we sequentially incubated HeLa cells with E. coli soluble extract containing p33 followed by extract containing p55 (p33 1 , p55 2 ), or p55 followed by p33 (p55 1 , p33 2 ) (superscript numbers indicate the order in which the samples were added to the cells) (Fig. 6). As in all of the preceding experiments, the concentrations of p33 and p55 proteins added to the cells were approximately equivalent. Following incubation of cells with the first VacA protein, the culture medium overlying the cells (containing unbound VacA fragments) was washed away prior to the addition of the second VacA protein. No vacuolating activity was detected when the p33 domain was added first, followed by the p55 domain (Fig. 6A) (p33 1 , p55 2 ). In contrast, extensive vacuolating activity was observed when the p55 domain was added first, followed by the p33 domain (Fig. 6A)  (p55 1 , p33 2 ).
To further investigate the process by which the (p55 1 , p33 2 ) combination induced vacuolation, we analyzed the interactions of sequentially added VacA domains with the surface of host cells. In the sequential addition experiments, the intensity of the immunoblot signals was consistently weaker for the protein added first compared with the protein added second, regardless of the order of addition (p33 1 , p55 2 or p55 1 , p33 2 ). When the p55 domain was bound first, followed by the p33 domain, a condition that resulted in cell vacuolation (p55 1 , p33 2 ) (Fig. 6A), the amount of p33 protein bound to cell membranes was increased compared with when the p33 domain was added alone (Fig. 6B,  bottom panel, lane 2 versus 5). Similar results were obtained when different epitope-tagged p33 and p55 proteins (e.g. p33Myc-His and p55Myc) were used and when gastric cell lines (AGS and AZ-521) were substituted for HeLa cells (data not shown). These data indicate that binding of p55 to cells enhances subsequent binding of p33.
We then investigated the binding of sequentially added VacA domains to the surface of host cells, using indirect immunofluorescence methodology. In these sequential addition experiments, both p33 and p55 domains could be detected on the cell surface, regardless of the order of addition (p55 1 , p33 2 or p33 1 , p55 2 ) (Fig. 6C, panels 2 and 3). In contrast, binding of p55 to cells was not detectable by immunofluorescence in experiments in which p55 was added independently (Fig. 4B). When the p55 and p33 proteins were added sequentially to cells, both proteins localized on the cell surface in a punctate distribution, regardless of the order of addition (Fig. 6C, panels 2 and 3). In contrast, when cells were incubated with the p33/p55 mixture, both proteins localized in a continuous (non-punctate) pattern on the surface of HeLa cells (Fig. 6C, panel 1).
We next tested the hypothesis that there might be differences in the internalization of VacA, depending on the order in which VacA domains are added to cells. When the p55 domain was added first to cells, followed by the p33 domain (p55 1 , p33 2 ), a condition that results in cell vacuolation (Fig. 6A), internalization of the p55 protein was detected (p55 1 , p33 2 ) (Fig. 6D). Internalization of p33 was also detected, but the intensity of the internalized p33 signal was relatively weak compared with the p55 signal (data not shown). When the p33 protein was added first followed by the p55 protein (p33 1 , p55 2 ), a condition that fails to cause cell vacuolation (Fig. 6A), neither p55 nor p33 was detected inside cells (p33 1 , p55 2 ) (Fig. 6D and data not shown). Thus, in these sequential addition experiments, there are marked differences in the internalization of VacA depending on the order in which VacA domains are added to cells.  (panels 1, 3, and 5) and an anti-c-Myc antibody to detect the p33Myc-His protein (panels 2 and 4).

DISCUSSION
The H. pylori VacA toxin produces a wide array of structural and functional alterations in intoxicated mammalian cells (7,10). An important goal is to identify functional domains of VacA that contribute to specific steps in the intoxication process. Previous studies (6,28,31) have demonstrated that the mature secreted VacA toxin undergoes proteolytic degradation to yield two fragments (p33 and p55). It has been suggested that these two fragments represent two domains of VacA (6,31,42,53), but in the absence of a high resolution VacA structure, the relevant structural features of these two putative domains remain poorly characterized. In the current study, we investigated various properties of recombinant p33 and p55 VacA domains. Although our efforts to purify functional forms of p33 and p55 recombinant proteins have thus far been unsuccessful, it has nevertheless been possible to investigate various functional properties of these domains using crude preparations of the p33 and p55 recombinant proteins.
Our data indicate that the p33 and p55 VacA domains interact in solution to form protein complexes (Fig. 3). This finding is consistent with the results of previous studies, which demonstrate that p33 and p55 proteins interact in the yeast twohybrid system (31), as well as in HeLa cells transiently cotransfected with plasmids expressing the p33 and p55 proteins (42,54). In addition, p33 and p55 VacA fragments remain physically associated following proteolysis of the 88-kDa VacA protein secreted by H. pylori (6,31). Notably, in the current study, heterotypic interactions between p33 and p55 were detected, but homotypic interactions (either p33/p33 or p55/p55) were not detected (Fig. 3A). This finding is consistent with the results obtained in a yeast two-hybrid system (31). Although homotypic interactions (p33/p33 or p55/p55) have not been detected in either the yeast two-hybrid system or in the current study, it remains possible that such interactions could occur in the context of a membrane environment (44) or within certain variant forms of VacA oligomers (45).
VacA 88-kDa monomers produced by H. pylori can assemble into large water-soluble oligomeric structures comprising 6 -14 subunits (28 -30). Our data indicate that when mixed together, p33 and p55 domains can form complexes composed of at least three independent subunits (Fig. 3B). However, we were unable to demonstrate assembly of p33 and p55 domains into high molecular mass oligomeric complexes similar to those formed by H. pylori 88-kDa monomers. Moreover, we were not able to detect interactions of either the individual p33 domain or p55 domain with the full-length secreted VacA protein from H. pylori (Fig. 3C). In contrast, when the p33/p55 mixture was combined with full-length VacA, interactions of both the p33 and p55 domains with full-length 88kDa VacA were detected (Fig. 3C). These results suggest that formation of large VacA oligomeric structures proceeds more efficiently via interactions among full-length 88-kDa VacA monomers than via interactions among isolated p33 and p55 domains.
An important conclusion of the current study is that p33 and p55 VacA domains lack detectable vacuolating activity when added individually to cells, but when mixed, the p33 and p55 domains complement each other, resulting in reconstitution of vacuolating activity (Fig. 2). The basis for this phenomenon can FIG. 6. Sequential addition of p33 and p55 domains to HeLa cells. A, HeLa cells were incubated for 1 h at 37°C with the indicated E. coli soluble extracts. Alternatively, HeLa cells were sequentially incubated with E. coli soluble extract containing p33, followed by p55 (p33 1 , p55 2 ), or with p55, followed by p33 (p55 1 , p33 2 ). After incubation of the first VacA domain with cells for 1 h at 37°C, unbound proteins were washed away, and the second VacA protein was then added for an additional 1 h at 37°C. Cells were then washed and incubated for 4 -6 h at 37°C in fresh culture medium. Vacuolating activity was measured by a neutral red uptake assay. Results represent the mean ϩ S.D. from triplicate samples. B, E. coli soluble extracts containing p55His and p33Myc-His were added to HeLa cells as indicated. The capacity of the VacA proteins to interact with host cells was assessed by immunoblot (I.B.) analysis using an anti-VacA polyclonal antibody (top panel) and an anti-His antibody (bottom panel). C, HeLa cells were intoxicated for 1 h at 37°C with E. coli soluble extract containing a p33/p55 mixture (column 1). Alternatively, p33 was bound first, followed by p55 (p33 1 , p55 2 ; column 2), or p55 was bound first, followed by p33 (p55 1 , p33 2 ; column 3), as described above. The capacity of the VacA proteins to interact with the cell membrane of host cells was assessed by indirect immunofluorescence using an anti-VacA polyclonal antibody to detect p55His (p55; top panels) and an anti-c-Myc antibody to detect the p33Myc-His protein (p33; bottom panels). D, HeLa cells were incubated with acid-activated VacA purified from H. pylori or E. coli soluble extracts containing the indicated recombinant VacA proteins, as described above. Thereafter, the cells were incubated in fresh culture medium for an additional 16 h at 37°C. Entry of VacA into cells was analyzed by indirect immunofluorescence (I.F.) of permeabilized cells using an anti-VacA polyclonal antibody to detect the p55 VacA domain. be understood in part by examining the interactions of VacA domains with the cell surface. We demonstrated in the current study that the individual p55 and p33 domains are each capable of binding to the cell surface. This finding is consistent with results of a previous study, which indicated that p55 and p33 domains can each bind to artificial lipid membranes (55,56). It seems likely that individual p55 and p33 domains may interact with lipids on the surface of mammalian cells.
Notably, we demonstrated that VacA interactions with the surface of cells are altered in several ways when both p33 and p55 domains are present, compared with when only a single domain is present. First, in comparison to individual p33 and p55 domains, a p33/p55 mixture binds more avidly to the cell surface. This increased binding is observed for both the p33 and p55 domains (Fig. 4A). Second, when p55 is added to cells in the absence of p33, the binding of p55 is detectable in immunoblot assays but not in immunofluorescence assays. In contrast, when a mixture of p55 and p33 domains is added to cells, the binding of p55 is detectable in both assays (Fig. 4B). This suggests that the conformation or orientation of the p55 domain on the surface of cells may be altered in the presence of the p33 domain. Finally, when added individually to cells, the p33 domain localizes in a punctate distribution on the cell surface, but when added to cells along with p55, p33 localizes in a continuous (non-punctate) distribution on the cell surface (Fig. 4B). Previous studies have reported the existence of multiple cell surface receptors for VacA secreted by H. pylori, and accordingly, it seems likely that recombinant p33⅐p55 complexes may bind to multiple different cell surface components (see model depicted in Fig. 7, A and B).
In the current study, we demonstrated that when added together, the p33 and p55 proteins are both internalized by host cells, whereas internalization is not detectable when the p33 or p55 domains are added individually to host cells (Fig. 5). The failure of p55 to be internalized when added independently to cells is consistent with the results of a previous study, in which a mutant VacA protein consisting mainly of the p55 domain was secreted by H. pylori and bound to the surface of host cells but was not internalized (45). We propose a model in which binding of p33⅐p55 VacA complexes to a specific site on the cell surface (for example, a specific receptor and/or lipid rafts) promotes VacA oligomerization. We propose that the p33⅐p55 oligomeric complex undergoes a conformational change to permit membrane insertion of the p33 domain and that the complex can then be internalized into the cell (Fig. 7,  B-D). The capacity of internalized p33⅐p55 complexes to induce cell vacuolation is consistent with results of a previous study, which showed that intracellular co-expression of p33 and p55 results in vacuolating cytotoxic activity (53).
Further insight into the functional roles of p33 and p55 domains comes from studies in which these domains are added sequentially to cells. We demonstrated that binding of p55 to the cell surface, followed by the addition of p33 (p55 1 , p33 2 ), results in cell vacuolation (Fig. 6A). This result can be explained by the formation of p55⅐p33 complexes on the surface of the cells. We speculate that the isolated p55 domain is able to bind specific cell-surface components that promote oligomerization, membrane insertion, and internalization of VacA (Fig. 7). Thus, our model proposes that internalization of VacA into cells is dependent on an interaction of the p55 domain with specific cell surface components (Fig. 7). In support of this model, several previous studies have provided evidence indicating that amino acid sequences in the p55 domain of p88 VacA contribute to the process of VacA binding to cells (39,(45)(46)(47). In the current study, binding of p33 to the cell surface, followed by the addition of p55 (p33 1 , p55 2 ), did not result in detectable cell vacuolation. We speculate that p33 may not be able to bind certain relevant cell surface components that are required for VacA internalization. Alternatively, sequential addition in this order (p33 1 , p55 2 ) may not permit the formation of the p33⅐p55 complexes or may prevent the formation of p33⅐p55 complexes in the proper conformation required for internalization.
VacA causes numerous effects on intoxicated cells (7,10,57), and many VacA-mediated effects are dependent on the capacity of VacA to form membrane channels (11,14,19,(21)(22)(23)(24). Therefore, it is of interest to view the current results in the context of what is known about functional domains of other pore-forming toxins. Two previous studies have investigated putative func- tional domains of the pore-forming toxins aerolysin (58) and listeriolysin (59). When the two domains of aerolysin and listeriolysin were co-expressed in Aeromonas and Listeria respectively, in both cases the two domains were able to assemble into proteins with hemolytic activity (58,59). In contrast, when the two domains of listeriolysin were expressed separately and then mixed together, no hemolytic activity was detected (59). This suggests that assembly of listeriolysin domains into a functionally active toxin may require interactions between the two domains during early stages of the protein folding process. To the best of our knowledge, VacA is the only pore-forming toxin for which cytotoxic activity has been reconstituted from two independently expressed functional domains. Thus, it seems likely that there may be major differences in the structural organization of VacA compared with most other known pore-forming toxins. Most pore-forming bacterial toxins act primarily on the plasma membrane of host cells, whereas VacA produces several effects that are dependent on its localization in intracellular sites (24,41). It is possible that the unique features of VacA may be related to the capacity of this toxin to act in multiple cellular sites.