Determinants of non-toxicity in the gastric pathogen Helicobacter pylori.

The Helicobacter pylori vacuolating cytotoxin gene, vacA, is naturally polymorphic, the two most diverse regions being the signal region (which can be type s1 or s2) and the mid region (m1 or m2). Previous work has shown which features of vacA make peptic ulcer and gastric cancer-associated type s1/m1 and s1/m2 strains toxic. vacA s2/m2 strains are associated with lower peptic ulcer and gastric cancer risk and are non-toxic. We now define the features of vacA that determine the non-toxicity of these strains. To do this, we deleted parts of vacA and constructed isogenic hybrid strains in which regions of vacA were exchanged between toxigenic and non-toxigenic strains. We showed that a naturally occurring 12-amino acid hydrophilic N-terminal extension found on s2 VacA blocks vacuolating activity as its removal (to make the strain s1-like) confers activity. The mid region of s2/m2 vacA does not cause the non-vacuolating phenotype, but if VacA is unblocked, it confers cell line specificity of vacuolation as in natural s1/m2 strains. Chromosomal replacement of vacA in a non-toxigenic strain with vacA from a toxigenic strain confers full vacuolating activity proving that this activity is entirely controlled by elements within vacA. This work defines why H. pylori strains with different vacA allelic structures have differing toxicity and provides a rational basis for vacA typing schemes.

The Helicobacter pylori vacuolating cytotoxin gene, vacA, is naturally polymorphic, the two most diverse regions being the signal region (which can be type s1 or s2) and the mid region (m1 or m2). Previous work has shown which features of vacA make peptic ulcer and gastric cancer-associated type s1/m1 and s1/m2 strains toxic. vacA s2/m2 strains are associated with lower peptic ulcer and gastric cancer risk and are non-toxic. We now define the features of vacA that determine the nontoxicity of these strains. To do this, we deleted parts of vacA and constructed isogenic hybrid strains in which regions of vacA were exchanged between toxigenic and non-toxigenic strains. We showed that a naturally occurring 12-amino acid hydrophilic N-terminal extension found on s2 VacA blocks vacuolating activity as its removal (to make the strain s1-like) confers activity. The mid region of s2/m2 vacA does not cause the non-vacuolating phenotype, but if VacA is unblocked, it confers cell line specificity of vacuolation as in natural s1/m2 strains. Chromosomal replacement of vacA in a nontoxigenic strain with vacA from a toxigenic strain confers full vacuolating activity proving that this activity is entirely controlled by elements within vacA. This work defines why H. pylori strains with different vacA allelic structures have differing toxicity and provides a rational basis for vacA typing schemes.
Gastric colonization by Helicobacter pylori is the main cause of peptic ulceration, gastric carcinoma, and gastric mucosaassociated lymphoid tissue lymphoma (1)(2)(3). However, although more than half the world's population is chronically infected with this Gram-negative bacterium, most people remain asymptomatic. Who develops disease depends on strain virulence, host genetic susceptibility, and environmental factors. Several bacterial virulence factors have been linked with disease: the active form of the vacuolating cytotoxin, VacA (4,5); presence of a pathogenicity island encoding a type IV secretion system, cag (6 -8); and possession of a specific adhesin, BabA (9).
The active form of VacA induces extensive cytoplasmic vacuolation in epithelial cells (10), causes gastro-duodenal damage in a mouse model (11), and increases gastric ulcer risk in H. pylori-infected Mongolian gerbils (12). The vacA gene encodes a preprotoxin of ϳ139 kDa (13)(14)(15)(16). This includes an N-terminal signal peptide and a ϳ50-kDa C-terminal autotransporter domain, both of which are cleaved during toxin secretion through the bacterial membranes (15). The ϳ90-kDa mature toxin may undergo further processing at an exposed protease-sensitive loop into ϳ37-kDa N-and ϳ58-kDa C-terminal fragments (p37 and p58, respectively) (15,17). The toxin binds to cell surface receptors through the p58 domain (18 -20). However, when expressed in epithelial cells, only p37 and an N-terminal fragment of p58 are required for vacuolation (21). Vacuolation is dependent on the insertion of VacA multimers into cell membranes to form anion selective pores (22,23). The mechanism of subsequent vacuole formation and the cellular origin of the vacuoles from late endosomes have been studied extensively (24 -30).
Research has focused on the most active form of VacA, but most H. pylori clinical isolates express less vacuolating or nonvacuolating forms (31). The gene, vacA, is naturally polymorphic and differences are most marked in two areas: the signal region, encoding the signal peptide and the N terminus of the mature protein (which may be type s1 or s2) and the mid region, encoding part of the p58 domain (type m1 or m2) (31). vacA signal and mid regions from all clinical isolates of H. pylori can be classified as one of these types, and all combinations of signal and mid region occur naturally, although the s2/m1 structure is rare (31,32). The vacA allelic type is associated with vacuolating activity in vitro: strains with s1/m1 vacA cause more extensive vacuolation in HeLa cells than those with s1/m2 vacA, and vacA s2/m2 strains are invariably non-vacuolating (31). The existence of stable polymorphisms affecting toxin function is of major biological interest; it may form a paradigm for varying functionality of proteins in other bacteria with high levels of genetic recombination in which mosaic genes are common. However, for H. pylori, vacA polymorphism is also potentially of major clinical importance: vacA s2/m2 strains are less frequently associated with both peptic ulceration and gastric carcinoma than vacA s1/m1 or s1/m2 strains (33)(34)(35)(36)(37)(38).
In this report, we define the determinants of non-toxicity in H. pylori with type s2/m2 vacA. Work by us and others on the most toxic s1/m1 type of vacA has guided our approach. We have previously shown that vacA transcription is higher for some toxic than for some non-toxic strains (39). We and others have also shown that the N terminus of mature s1/m1 VacA in toxic strains is one important determinant of toxicity and that adding an s2-like N-terminal extension blocks activity (21, 40 -43). However, one cannot extrapolate from this to imply that the N terminus of natural s2/m2 strains is the cause of their non-toxicity, so we aimed to determine whether this was indeed the case. Naturally occurring vacA type s1/m1 strains cause vacuolation in a wider range of cell lines than s1/m2 strains (19,44), so we also planned to show whether the m2 mid region was the cause of non-toxicity in natural s2/m2 strains.
Finally, it is unclear whether elements outside vacA in s2/m2 strains contribute to non-toxicity, and we aimed to resolve this issue.
Construction of Isogenic vacA Signal Region Hybrid Strains-The effect of vacA signal region type on toxin production and activity was studied by constructing H. pylori isogenic vacA hybrid strains in which different extents of the signal region had been exchanged between the s1 and s2 forms on the H. pylori chromosome. Such hybrids were constructed using plasmids pA153::cat and pCTB2::cat, containing the promoter and 5Ј terminus of vacA, including the signal region, from strains Tx30a (vacA s2/m2; ToxϪ) and 60190 (vacA s1/m1; Toxϩ), respectively, together with the 3Ј terminus of the upstream gene cysS with a chloramphenicol resistance marker (chloramphenicol acetyltransferase; cat) inserted immediately downstream (see Table I). These pBluescript-derived plasmids act as suicide vectors in H. pylori, thus chromosomal recombinants were constructed by natural transformation, allelic exchange, and chloramphenicol rescue as previously described (41) (see Fig. 1). Transformation of pA153::cat and pCTB2::cat into their respective homologous strains resulted in the isolation of control strains Tx30a/CAT and 60190/CAT, containing the insertion of cat upstream of the vacA promoter. Transformation of Tx30a with pCTB2::cat and 60190 with pA153::cat gave rise to multiple hybrid strains. DNA extracted from single colonies (45) was typed for vacA signal and mid region by allele-specific PCR, as previously described (31,46). Only vacA type s1 recombinants were identified from the transformation of Tx30a with pCTB2::cat, and one was selected and termed Tx30a/P1S1. For the transformation of 60190 with pA153::cat, type s1 and s2 recombinants were identified and termed 60190/P2S1 and 60190/P2S2, respectively. The precise extent of recombination was determined for these mutants by nucleotide sequencing. pA153::cat and pCTB2::cat were similarly transformed into the toxigenic strain 84-183 (vacA s1/m1) to give hybrids 84-183/P2S2 and 84-183/P1S1 60190 , respectively.
Construction of vacA Mid Region Hybrid Strains-The effect of vacA mid region type was studied in a similar manner. Plasmids pNV5 and pNV2, containing 3Ј terminal 3576 bp and 3587 bp of vacA with 1 kb and 1.5 kb of downstream sequence from strains Tx30a and 60190, respectively, with a kanamycin resistance marker (aminoglycoside phosphotransferase; aphA) inserted 0.5 kb 3Ј to vacA, within the fecE gene, were introduced into strain Tx30a by natural transformation, allelic exchange, and kanamycin marker rescue. Recombinants were typed for vacA signal and mid region by allele-specific PCR (31,46). Transformation with pNV5 and pNV2 gave rise to the control strain Tx30a/KAN and the hybrid strain Tx30a/M1 (vacA s2/m1), respectively. Similarly, transformation of strain 60190 with pNV2 and pNV5 created the control strain 60190/KAN and the hybrid strain 60190/M2 (vacA s1/m2), respectively. Tx30a/M1 was further transformed with pCTB2::cat to produce the hybrid strain Tx30a/P1S1M1 (vacA s1/m1).
Site-directed Mutagenesis of vacA-The vacA mutant Tx30a/N1, containing an in-frame deletion of the region encoding the s2-specific N-terminal extension, was constructed by first deleting this 36-bp region from the cloned vacA fragment in plasmid pA153::cat using inverse PCR with primers NdelF (5Ј-GCTTTTTTCACAACCGTGATCATTCCA-GCC-3Ј) and NdelR (5Ј-AGCCCCCAGTTCGGTGCCCATTAACACCC-3Ј), which bound to nucleotides 472-501 and 407-435 in the vacA sequence of Tx30a (31), respectively. Template DNA was removed by DpnI restriction endonuclease digestion (New England BioLabs (UK) Ltd.), and the PCR product was end-polished using Pfu DNA polymerase (Stratagene Europe), 5Ј-phosphorylated with T4 polynucleotide kinase in the presence of 1 mM ATP, and blunt-end-religated. Following transformation into Escherichia coli strain DH5␣, plasmid DNA was extracted, and the presence of the 36-bp deletion screened by PCR using primers VA1F and VA1R (31), and confirmed by nucleotide sequencing. The vacA mutation was introduced into chromosomal vacA of H. pylori strain Tx30a by natural transformation, allelic exchange, and chloramphenicol marker rescue, and the presence of the vacA mutation was confirmed by PCR as before.
Quantification of VacA-VacA production was quantified by antigen detection ELISA as previously described (47). Briefly, duplicate 48-h broth culture supernatant samples of each strain were adsorbed to a microtiter plate overnight at 4°C, blocked with 3% bovine serum albu- This study min (Sigma), and bound antigen incubated with a 1:5,000 dilution of rabbit antiserum to either purified VacA from the type s1/m1 strain 60190 (Ab123) (47), or the recombinant VacA p58 fragment from either this strain (Ab929) (18) or the m2 type strain Tx30a (Ab927), all kindly donated by Dr. T. L. Cover, Nashville, TN. For detection we used anti-rabbit IgG-horseradish peroxidase conjugate (Sigma) then orthophenyldiamine/hydrogen peroxide. ELISA values were expressed as A 492 nm , corrected for bacterial density (A 600 nm ). Detection of VacA by Immunoblotting-H. pylori broth culture supernatants or water extracts were separated by SDS-PAGE and transferred to nitrocellulose by electroblotting. Nitrocellulose blots were blocked in 5% (w/v) milk in PBS-Tween 20 (PBS-T) then incubated with rabbit anti-VacA antiserum (Ab123, 927, or 929) diluted 1:10,000. Blots were washed three times in PBS-T, incubated in anti-rabbit IgG-horseradish peroxidase conjugate, washed six times in PBS-T, and visualized by ECL TM detection (Amersham Biosciences).
Determination of VacA Activity-Water extracts were prepared by harvesting 48-h growth from a single plate into 1 ml of sterile distilled water, vortexing for 30 s, and incubating at room temperature for 20 min. Cells were removed by microcentrifugation, and the supernatant containing VacA was filter-sterilized. Vacuolating activity was determined using three different epithelial cell lines: HeLa, AGS (a human gastric adenocarcinoma cell line), and RK13 (a rabbit kidney cell line). Assays were performed by adhering 10 4 epithelial cells in RPMI 1640 media supplemented with 10% FCS (both Invitrogen) to a microtiter plate overnight. The media was then replaced with fresh media containing 10 mM ammonium chloride, and a 5-fold dilution of water extract. Cells were incubated overnight and then visually assessed for vacuolation by light microscopy.

RESULTS
The Type s2 Promoter/Signal Region of vacA Determines the Non-toxic Status of the ToxϪ Strain Tx30a-Our previous work had shown that vacA s1/m1 strains are usually toxic and vacA s2/m2 strains invariably non-toxic (31). Our first aim was to define whether the promoter/signal region of a candidate nontoxic vacA s2/m2 strain, Tx30a, was directly responsible for its non-vacuolating phenotype. To do this, we used allelic exchange to replace the promoter and signal region of vacA on the Tx30a chromosome with that of strain 60190 (toxϩ; vacA s1/ m1) to make an artificial hybrid s1/m2 vacA construct in a non-toxigenic strain background, which we called Tx30a/P1S1 (see "Materials and Methods" and Fig. 1). Nucleotide sequencing of this hybrid showed the crossover point to be between 588 and 763 in the Tx30a sequence (31) and proved that the promoter region was identical to that in the s1/m1 strain 60190. To ensure that insertion of the chloramphenicol resistance marker (cat), used for the allelic exchange experiment, was not influencing VacA production or activity, we also constructed the control strain Tx30a/CAT with cat inserted between cysS and vacA (see "Materials and Methods"). Insertion of the chloramphenicol cassette in Tx30a did not affect VacA levels in broth culture supernatants, as determined by antigen detection ELISA, and all later comparisons used Tx30a/CAT as control. Replacement of the signal and promoter regions with those from 60190 increased VacA production as determined by ELISA using antiserum Ab927 (Tx30a/CAT mean VacA production 0.03 Ϯ 0.01, n ϭ 10, versus Tx30a/P1S1 mean 0.17 Ϯ 0.05, n ϭ 8, p Ͻ 0.005, t test). This finding was confirmed by immunoblot using the same antiserum (Fig. 2). As expected, water extracts of the non-toxigenic control strain, Tx30a/CAT, did not cause vacuolation of any of the three cell lines tested ( Fig. 3A and Table II). However, Tx30a/P1S1 water extracts induced extensive cytoplasmic vacuolation of RK13 cells following overnight incubation (Fig. 3B). Interestingly, the vacuolating activity observed for this strain was cell line-specific, because the same water extracts were unable to induce vacuolation of HeLa and AGS cells (Table II). This is consistent with the described phenotype of a naturally occurring vacA s1/m2 strain (19). Thus, naturally occurring differences in vacA signal regions are directly responsible for VacA production and activity.
Confirmation That the s2 Promoter/Signal Region of vacA Determines Non-toxic Status Using Toxϩ Strain Backgrounds-Using identical methodology, we confirmed our results by performing reciprocal experiments in which we replaced the signal and promoter regions of vacA in strain 60190 (toxϩ; vacA s1/m1) with those of Tx30a (toxϪ; vacA s2/m2) to create the hybrid strain 60190/P2S2. Nucleotide sequencing confirmed that the sequence up to nucleotide 1070 had been replaced with that of Tx30a. As for Tx30a, we constructed a control strain of 60190 with the chloramphenicol cassette inserted just after the cysS stop codon; this did not affect VacA production or activity, and 60190/CAT was used as the control in all further experiments. Insertion of the signal and promoter region from Tx30a reduced VacA production as determined by ELISA using Ab123 (60190/CAT mean VacA production 1.01 Ϯ 0.12, n ϭ 8, versus 60190/P2S2 mean 0.19 Ϯ 0.02, n ϭ 4, p Ͻ 10 Ϫ3 ). As expected, 60190/CAT water extracts caused extensive cytoplasmic vacuolation of all three cell lines studied (Fig. 4A and Table II). In contrast, no vacuolation of HeLa, AGS, or RK13 cells was observed with water extracts of the hybrid strain 60190/P2S2 (Fig. 4C and Table II). Next, we aimed to confirm our results in a third independent strain background. We selected strain 84-183, an easily naturally transformable strain of vacA type s1/m1, which we have previously shown to transcribe VacA less strongly than strain 60190 and to be less strongly vacuolating (39). Replacement of the signal and pro- These experiments show that replacing the type s1 promoter and signal region in a toxϩ strain with a type s2 region from a toxϪ strain reduced VacA production and abolished activity. Interestingly, strains 60190/P2S2 and 84-183/P2S2, which both have a hybrid s2/m1 vacA structure, grew similarly to their respective controls on blood agar and in broth. Thus the reason that strains of vacA type s2/m1 are uncommon in nature is not that they have obvious self-toxicity or a growth disadvantage, at least in vitro.
The vacA Promoter Region Determines Differences in VacA Production, but the Signal Region Determines Differences in Vacuolating Activity-Having shown directly that the vacA signal and promoter regions in vacA s2/m2 strain Tx30a were together an important determinant of its non-toxigenic status, we now aimed to define the role of each region individually. were separated by SDS-PAGE, transferred to nitrocellulose by electroblotting for 1 h at 140 mA, and detected using a 1:10,000 dilution of antiserum to the recombinant p58 subunit of Tx30a VacA (Ab927). An immunoreactive band of ϳ93 kDa was detected for the control strain Tx30a/CAT as reported previously (31). A more intense, but slightly smaller band was detected for Tx30a/P1S1 confirming that VacA production is increased in this strain. The difference in size is in agreement with the removal of the VacA N-terminal extension resulting from swapping the s2 signal region for the s1 type.
a Vacuolating activity was assessed by incubating either HeLa, AGS, or RK13 epithelial cells overnight with a water extract of the appropriate strain. Vacuolation was recorded as positive if more than 50% of the cells within a randomly chosen field were vacuolated and negative if the number of vacuolated cells was the same or less than that observed for untreated cells.
b The signal region of this strain is s2 type for the signal peptide including the cleavage site, and s1 type for the mature N terminus. From the previous experiments it was unclear whether the low level of VacA production in strains Tx30a and 60190/P2S2 was the reason these strains were non-toxic or whether there were other determinants of non-toxicity. We aimed to address this by screening for transformants from our previous experiments where the recombination point was between the promoter region and signal sequence coding region. Screening of mutants generated by transforming strain Tx30a with the promoter and signal regions of vacA from 60190 yielded no such transformants. However, the transformation of 60190 with pA153::cat yielded a transformant, which typed as s1 by allele-specific PCR, and we called this 60190/P2S1. Sequence analysis showed that this hybrid contained a promoter region and ribosomal binding site identical to that of the type s2 strain Tx30a, but the signal sequence remained 60190 vacA. The exact 3Ј crossover point was between nucleotides 795 and 835 in the 60190 vacA sequence (13). Studying this hybrid allowed us to determine the relative effects of the s2 promoter and signal regions in the 60190 strain background. Replacing the promoter region only, significantly reduced VacA ELISA levels using Ab123 (60190/P2S1 mean 0.34 Ϯ 0.09, n ϭ 4, p Ͻ 0.005, versus 60190/CAT, not different from 60190/P2S2). However, in contrast to 60190/P2S2, which did not express vacuolating activity, 60190/P2S1 caused vacuolation in all three cell lines ( Fig. 4B and Table II). Thus the lack of vacuolating activity observed for 60190/P2S2 was not simply due to reduced VacA production as 60190/P2S1 produced similar levels of VacA but still retained vacuolating activity. This shows that, as expected, the promoter region, not the signal region, determines VacA production. However, the signal region is responsible for determining VacA activity.
The N-terminal Hydrophilic Extension on Mature Type s2 VacA Determines Its Non-vacuolating Phenotype-A striking difference in the signal region between vacA type s1 and s2 strains is the presence of a 12-amino acid extension to the non-toxigenic, s2 form of VacA (31). The mature N terminus of s1 type VacA is markedly hydrophobic, whereas the s2-specific, N-terminal amino acid extension is strongly hydrophilic (Fig.  5). We hypothesized that the presence of this hydrophilic extension blocks the vacuolating activity of VacA produced by s2 strains such as Tx30a. To test this, we made an in-frame deletion in Tx30a vacA of the 36 nucleotides encoding this extension, thus removing it from the mature VacA product (see "Materials and Methods"). We named the resulting isogenic mutant strain Tx30a/N1. Removal of the N-terminal amino acid extension reversed the non-vacuolating phenotype of Tx30a, such that water extracts of Tx30a/N1 caused extensive vacuolation of RK13 cells (Fig. 3C). Tx30a/N1 did not vacuolate AGS or HeLa cells (Table II). Note that Tx30/N1 has a type m2 mid region, so this result is consistent with the described phenotype of a naturally occurring vacA s1/m2 strain (19). To ensure that the vacuolating activity of Tx30a/N1 was not due to a greater VacA production compared with Tx30a/CAT, we de-termined the amount of VacA in water extracts by ELISA using Ab123. VacA levels were no greater for Tx30a/N1 than Tx30a/ CAT (means 0.11 Ϯ 0.02 for both, n ϭ 10 and 4, respectively, p ϭ ns). Thus, the non-toxigenic phenotype of the vacA s2/m2 strain Tx30a is due to the hydrophilic extension on the N terminus of mature VacA.

Replacing the m2 Mid Region of a Non-toxigenic Strain with an m1 Mid Region Is Not Sufficient to Render It Toxigenic-
Having shown that non-toxic vacA s2/m2 strains could be rendered toxic by replacing the signal region with an s1 region or by removing the N-terminal amino acid extension on the mature VacA protein, we now aimed to assess whether these strains could also be rendered toxic by changing the mid region to an m1 type. To do this we replaced the vacA mid region of strain Tx30a with that of strain 60190 (toxϩ; vacA s1/m1) to make a vacA s2/m1 construct in a non-toxigenic strain background, which we termed Tx30a/M1 (see "Materials and Methods"). Sequence analysis confirmed that the vacA mid region sequence had been exchanged for the 60190 sequence downstream of nucleotide 828. This strain also gained the 3Ј region of vacA from 60190, but this region, which encodes the Cterminal bacterial outer membrane transporter, which is cleaved from mature VacA after export, is very similar between all vacA alleles (48). As the kanamycin resistance marker used for allelic exchange was located 3Ј to vacA within fecE, we also constructed the control strain Tx30a/KAN containing just the marker insertion.
Insertion of the kanamycin cassette in fecE did not affect growth on blood agar or VacA activity (Table II), and all later comparisons used Tx30a/KAN as the control. Inactivation of fecE, encoding the ATP-binding protein component of an Fe 3ϩdicitrate ABC transporter, has previously been shown not to affect Fe 2ϩ or Fe 3ϩ transport, or growth on brain heart infusion-FCS medium (49). Replacing the vacA m2 mid region with an m1 type did not render the strain toxigenic: neither Tx30a/ KAN nor Tx30a/M1 caused vacuolation in any of the three cell lines tested (Table II and Fig. 3D). VacA levels in water extracts of Tx30a/M1 were not significantly different to those of Tx30a/KAN as determined by ELISA using an m2-specific antibody (Ab927) (means 0.32 Ϯ 0.01, n ϭ 4 and 0.35 Ϯ 0.004, n ϭ 2). However, using an m1-specific antibody (Ab929) VacA levels appeared 2-fold higher for Tx30a/M1 (means 0.45 Ϯ 0.01, n ϭ 4 and 0.22 Ϯ 0.01, n ϭ 2, p Ͻ 10 Ϫ3 ). Although it is not possible to directly compare VacA production for these strains by ELISA, owing to antigenic differences in the mid region, it is worth noting that, even if VacA levels were higher in Tx30a/M1 water extracts, no vacuolating activity was observed. This shows that merely replacing the m2 mid region with an m1 mid region is insufficient to render a non-toxigenic vacA s2/m2 strain toxigenic. Presumably, the type s2 N-terminal extension is sufficient to block toxic activity. Interestingly, strain Tx30a/ M1, which has a hybrid s2/m1 vacA structure, grew similarly to Tx30a and Tx30a/CAT on blood agar and in broth. This con- FIG. 5. The N-terminal protein sequence of mature VacA. A, the N terminus of mature VacA from the s2-type strain Tx30a contains a 12-amino acid extension compared with that of the s1-type strain 60190. B, a mean hydrophobicity plot shows that in contrast to the hydrophobic mature N terminus of 60190 VacA, the N-terminal extension found on Tx30a VacA is hydrophilic. firms, in a Tx30a background, our findings in the 60190 and 84-183 backgrounds, that possession of vacA with the naturally uncommon s2/m1 structure is not obviously disadvantageous in vitro.
vacA from a Toxigenic Strain Is Sufficient to Render a Nontoxigenic Strain Fully Toxigenic-The failure of an m1 vacA mid region to render toxϪ strain Tx30a toxigenic could have been because the m1 mid region was non-functional in this Tx30a/M1 hybrid. To refute this, we next replaced the s2 signal and promoter regions with s1 regions derived from strain 60190, to make a vacA s1/m1 construct in a non-toxigenic Tx30a background. Our second and more important aim in making this hybrid was to replace type s2/m2 vacA in Tx30a with type s1/m1 vacA from 60190 to assess to what extent single copy chromosomal complementation with toxigenic vacA conferred activity in a non-toxigenic strain background.
To replace the s2 signal and promoter regions in Tx30a/M1 with s1 regions from 60190, we transformed Tx30a/M1 with pCTB2::cat to produce the hybrid strain Tx30a/P1S1M1. Sequence analysis of vacA from this hybrid confirmed that the gene had been replaced with that of the toxigenic strain 60190 with the exception of a 190-bp region from nucleotides 639 to 828 encoding part of the p37 domain of the VacA protein, which was still derived from Tx30a. Water extracts of Tx30a/P1S1M1 induced extensive vacuolation of all three cell types studied, similar to 60190 and controls 60190/CAT and 60190/KAN (Table II and Fig. 3E). VacA levels in the Tx30a/P1S1M1 and 60190/KAN water extracts were also similar as determined by ELISA using Ab929 (means 0.50 Ϯ 0.10 and 0.43 Ϯ 0.01, respectively, n ϭ 4 for both, p ϭ ns). This confirms that the m1 mid region is functional in a Tx30a background and that vacA itself from a toxigenic strain is sufficient to confer full toxigenic activity. Thus differences between toxϩ and toxϪ strains elsewhere on the chromosome do not contribute to differences in VacA toxicity between strains.
The m2 Mid Region from a Non-Toxigenic vacA s2/m2 Strain Confers Cell-line Specificity of Vacuolating Activity to a Toxigenic s1/m1 Strain-The type m2 mid region of a naturally occurring vacA s1/m2 strain has previously been shown to confer specificity of vacuolation such that the strain vacuolates RK13 cells but not HeLa or AGS cells (19). Our previous experiment described above in which we created Tx30a/P1S1 with its s1/m2 structure, suggests that the m2 mid region from an originally non-toxigenic vacA s2/m2 strain also confers cell specificity. To prove that the m2 vacA mid region is directly responsible for this effect, we replaced the type m1 mid region in strain 60190 with the m2 mid region from Tx30a to create the vacA s1/m2 hybrid strain 60190/M2. Sequence analysis confirmed that the mid region had been replaced with the m2 form downstream of nucleotide 1696. As before we also constructed the control strain 60190/KAN containing the kanamycin marker insertion in fecE. Water extracts of 60190/M2 and 60190/KAN vacuolated RK13 cells equally (Fig. 4D). However, in contrast to the control, which also caused extensive vacuolation of HeLa and AGS cells, 60190/M2 did not cause vacuolation of either of these cell types (Table II). The amount of VacA in water extracts of 60190/M2 appeared nearly 2-fold lower than those of the control 60190/KAN, as determined by ELISA using the m1-specific antibody, Ab929 (means 0.23 Ϯ 0.01 and 0.43 Ϯ 0.01, n ϭ 4 for both, p Ͻ 10 Ϫ4 ), but over 3-fold higher using the m2-specific antibody, Ab927 (means 0.83 Ϯ 0.04 and 0.25 Ϯ 0.005, n ϭ 4 for both, p Ͻ 10 Ϫ5 ). Immunoblotting confirmed these findings (data not shown). Given that the ELISA values obtained with each antibody would have been underestimated for VacA alleles of the opposite mid region type, it is likely that the actual VacA amounts for the hybrid and control strains were similar. This experiment shows that the m2 mid region from a non-toxigenic vacA s2/m2 strain is functional in determining cell line specificity. Notably also, the vacuolating phenotype of 60190/M2 was the same as that obtained for Tx30a/P1S1, showing that the activity of the s1/m2 form of vacA is independent of strain background. DISCUSSION Compared with toxigenic strains of H. pylori, non-toxigenic vacA s2/m2 strains are associated with a much lower risk of peptic ulceration and gastric adenocarcinoma (31,36,37,50). In this study we have shown why vacA s2/m2 strains are non-toxigenic, the primary determinant being a 12-amino acid N-terminal extension on the VacA protein, which blocks toxin activity regardless of production level or p58 binding region type. Removal of this extension confers toxicity. The m2 vacA mid region in non-toxigenic s2/m2 strains is not the cause of non-toxigenicity, but is functional in conferring cell specificity. Chromosomal complementation with single copy vacA from a toxigenic vacA s1/m1 strain shows for the first time that elements outside vacA are not needed for full vacuolating activity.
Mature VacA from toxigenic vacA type s1 strains has a hydrophobic N-terminal region, which can insert into lipid bilayers (51). In non-toxigenic vacA type s2 strains this region is preceded by a 12-amino acid hydrophilic N-terminal extension, which we have shown blocks vacuolating activity. Partially deleting the hydrophobic N terminus of type s1 VacA (for example, amino acids 6 -27) also blocks vacuolating activity, and, although VacA still forms pores in artificial membranes, they form more slowly and are less anion selective (42). We have previously shown that adding the s2-specific hydrophilic extension to the N terminus of type s1 VacA abolishes vacuolating activity (41). However, although this slows pore formation in artificial membranes, it does not abolish it nor change the anion selectivity of these pores (43). Thus, it remains unclear whether the hydrophilic N-terminal extension of s2 VacA blocks membrane insertion but that this is not necessary for pore formation in artificial lipid membranes, or whether it acts through another mechanism, for example, through changing the conformation of the active p37 subunit of VacA.
Our key finding is that type s2/m2 VacA is functionally vacuolating once the hydrophilic N-terminal extension is removed. Why H. pylori should possess a functional "blocked" form of VacA is a fascinating enigma. One possibility is that the toxin becomes activated in vivo through cleavage of this extension. However, if this occurs, it is insufficient to render such strains pathogenic. A second possibility is that VacA possesses an important biological function other than inducing vacuolation, which is not blocked by the s2 extension. For example, it may perform other functions ascribed to VacA such as increasing epithelial permeability, stimulating epithelial cell apoptosis (52), inhibiting antigen presentation (53), or binding to cytoskeletal proteins (54).
Wild-type H. pylori strains of vacA type s2/m1 have only rarely been isolated (32,55,56), and we have identified only one that expresses VacA. Population genetic analyses show that vacA structure is characterized by frequent recombination events between vacA from different strains (57)(58)(59)(60), so vacA s2/m1 structures would be expected to arise in vivo as frequently as s1/m2 structures. In this study, we constructed the vacA s2/m1 structure in both originally toxigenic and originally non-toxigenic strain backgrounds, and these strains grew indistinguishably under laboratory conditions from other strains. Thus, although we speculate that s2/m1 vacA offers a selective disadvantage or fails to offer a selective advantage in vivo, the nature of this remains unclear.
An important finding in our study is that single copy chromosomal replacement of s2/m2 vacA in a non-toxigenic strain with s1/m1 vacA from a toxigenic strain confers full vacuolating activity similar to that of the parent toxigenic strain. This shows that production and vacuolating activity of VacA are not dependent on chromosomal elements outside vacA. Furthermore, as full vacuolating activity was conferred without amino acids 639 -828 (which are fairly well conserved between vacA alleles) being replaced, we can infer that these residues do not contribute to reduced VacA activity.
Pathogenic strains with the s1/m1 type of VacA have been extensively characterized, including the determination of the complete genome sequence for two such strains (61)(62)(63). However, such strains comprise only about 40% of strains isolated from patients undergoing endoscopy (31)(32)(33)(34)(35)(36)(37)(38), and this proportion would be expected to be smaller in an unselected population. We have now concentrated on the s2/m2 type of VacA found in non-pathogenic, non-toxigenic strains and the common hybrid s1/m2 type and uncommon hybrid s2/m1 type (32,55,56). We have shown that strains with the type s2 signal region are non-vacuolating due directly to the N-terminal extension on the mature toxin of these strains. The most commonly used test for these strains is an allelic type-specific PCR assay based on detection of this region (31,46), and our study confirms that this test has a rational biological basis. We have also shown that the mid region of both s2/m2 strains and s1/m2 hybrid strains is responsible for cell-specific vacuolation, and this provides validity to the widespread use of mid region allelic type-specific PCR. However, the full significance of vacA polymorphism is unlikely to be defined until studies are undertaken in large patient series, and this will not be possible until serum tests are developed which can accurately differentiate between infection with strains expressing the various VacA types.