Sequence and Antigenic Variability of the Helicobacter mustelae Surface Ring Protein Hsr

Abstract

We have identified an array of more than 500 repetitive sequences flanking the hsr gene, which encodes the major surface protein of the ferret pathogen Helicobacter mustelae. The repeats show identity exclusively to the amino-terminal half of Hsr. Analysis of Hsr from three strains indicated variability of exposed epitopes. Characterization of an hsr mutant showed that Hsr is not an adhesin.

Helicobacter mustelae infection is prevalent in North American and other ferret populations (4, 6). It produces in infected animals a pathology very similar to that caused by Helicobacter pylori in humans (2, 5, 11). The ferret infection model offers some distinct advantages (10), including the ability to induce gastric carcinoma (7). Despite similarities in pathologies caused by H. pylori and H. mustelae, the latter organism is distinguished by an array of ring structures covering the cell surface, composed of a single protein, designated Hsr (for Helicobacter surface ring). This 150-kDa protein forms rings of 8.5 nm in diameter, comprised of monomers extending 6 nm from the cell surface, anchored by the hydrophobic carboxy terminus (12). The Hsr protein belongs to the autotransporter family (8), but it lacks a serine protease catalytic triad (N. T. Forester and P. W. O'Toole, unpublished data), consistent with its retention on the cell surface. The protein is a strong immunogen in rabbits (12) and elicits antibody production in infected ferrets (4). The biological function of Hsr is currently unknown. We have now investigated the hsr flanking sequences as part of a study seeking to explain how the Hsr protein is highly expressed and efficiently exported to the cell surface.

H. mustelae strain 4298, from which the structural hsr gene was originally cloned, and eight recently isolated H. mustelae strains have been described elsewhere (4, 12). Bacteria were cultured as described previously (12). As part of a wider study to identify expression signals of hsr in strain 4298, the upstream flanking sequences (7,481 bp) and subsequently the downstream flanking sequences (2,880 bp) were subcloned from λEMBL3 clones into plasmids and the DNA sequences were determined. Escherichia coli strain ER2206 (12) was used as host for molecular cloning, following standard protocols (15). DNA sequence data were obtained using custom primers and an ABI Prism 377 automated sequencer and were analyzed with the Geneworks package (IntelliGenetics, El Camino, Calif.). The repetitive nature of the DNA (see below) placed restrictions on primer-walking. The physical structures and positions with respect to the hsr gene of all subclones were confirmed by restriction mapping and PCR to be the same as those of the genome (not shown). Subclones were confirmed by fine detail restriction mapping and specific PCR across clones. Sequences were compiled in contigs by manual alignment and assembly based on physical map data.

The 14,919-bp region incorporating the hsr gene and hsr-related flanking sequences is referred to as the hsr locus (Fig. 1). Apart from hsr (4,557 bp), only one other significant open reading frame (ORF) was present (ORF2, 552 bp). The hypothetical gene product has 40% residue identity with a putative gene product of H. pylori with unknown function (TIGR HP0785; JHP0722). This ORF contains no sequences similar to hsr. Analysis of the rest of the locus showed extensive tracts of sequence identical to hsr, which were strictly inverted with respect to the hsr gene orientation (Fig. 1). The 557 perfectly repeated sequences in the hsr locus varied in length from 12 bp to 741 bp, with no mismatches. Longer elements composed of combinations of shorter repeats were also present. The flanking repeat sequences were based exclusively on the first 2,362 bp of the Hsr coding sequence.

FIG. 1.

Physical organization and repetitive nature of the hsr locus. (A) Schematic diagram with nucleotide coordinates. Double-headed arrows indicate: (1) probe used for Southern hybridization (from within hsr gene), and PCR products (511 bp) from variable region (all strains) and upstream (801 bp; strain F6); (2) PCR product from nonvariable region; (3) PCR product from β-barrel region. Shaded regions of hsr indicate the signal peptide (black), variable (hatched), nonvariable (white), and β-barrel (gray) regions. USR, upstream repeats; DSR, downstream repeats. (B) Dot-matrix analysis of the hsr locus sense strand versus the sense strand (left box) or its complement (right box). The alignment was produced with the Geneworks program, using a k-tuple size of 8, and an output compression factor of 32. Lines have been drawn on the matrices to delineate regions of the hsr locus.

The repetitive flanking sequence around the hsr gene could be involved in, or arise from, rearrangement in hsr. Three regions of the hsr gene (indicated in Fig. 1) were amplified by PCR from eight H. mustelae strains and directly sequenced for comparison with the sequence of strain 4298. Multiple sequences were aligned with the Clustal V program (9) using the default parameters. Comparison of sequences encoding the central region of Hsr or the carboxy-terminal end (Fig. 1) showed that they were 99.1 or 99.2% identical, respectively (data not shown). Multiple alignment of the sequences amplified from near the beginning of hsr showed significant sequence variation (Fig. 2). Comparison of a second block, from residues 230 to 460 of Hsr, also showed variability (not shown). This region of the hsr gene was therefore termed the variable region (Fig. 1A), although we cannot formally confirm that this variability extends completely through the repetitive region because of difficulties in designing specific primers. The rest of the protein comprised the nonvariable region and the β-barrel domain which was inferred from sequence identity to autotransporters (Fig. 1).

FIG. 2.

Multiple alignment of part of the variable region of the Hsr proteins of nine H. mustelae strains, corresponding to residues 37 through 183 of the strain 4298 molecule. Manually aligned translations, with hsr locus (HSRL) coordinates, of regions of the hsr locus of strain 4298 are shown in the lower rows. Variant sequences present in the amplified regions of the ferret isolates, which are also found in the translated flanking sequences of strain 4298, are indicated by underlining. The bottom row (§) shows part of the translated 811-bp PCR product of the upstream flank of strain F6, which corresponds physically to positions 1001 to 1172 of the hsr locus. Arrowhead, signal peptidase cleavage site of Hsr of strain 4298.

Some of the variable sequence stretches were flanked in frame, on each side, by blocks of highly conserved sequence. For example, for the four underlined divergent motifs in Fig. 2, DNA sequences corresponding to the “conserved-variable-conserved” motifs in the Hsr proteins from the ferret isolates were identified in the upstream flanking sequence of strain 4298. An 801-bp fragment from the upstream flank was also amplified by PCR of the variable regions, and this product from strain F6 was sequenced. The sequence was largely identical to the corresponding upstream flank from strain 4298 (Fig. 2), except for the presence of a variant sequence in the strain F6 fragment encoding a DANAKN motif. This protein sequence motif is also found in the variable region of the Hsr protein of strain F6 (Fig. 2). It therefore appears that the repetitive elements in the hsr locus could contribute to antigenic variability, for example by gene conversion or homologous recombination as demonstrated in other bacteria (3, 16). Helicobacter spp. undergo allelic exchange at high frequency, as evidenced by the ease with which many genes have been inactivated by marker exchange, and recombination occurs more frequently in H. pylori than in any other organism studied to date by the homoplasy test (17). However, we cannot currently distinguish whether the presence of the repeats provides a driving force for rearrangement of hsr or is merely a consequence of it. We investigated the extent of hsr-related DNA by Southern hybridization of genomic DNA from the type strain 4298 and six recent isolates, digested with DraI. The probe for this experiment was a 511-bp fragment coding for part of the signal sequence and variable region of the hsr gene as indicated in Fig. 1. In the case of strain 4298, for which the hsr locus sequence had been determined, the hybridization pattern was consistent with the DraI cleavage map of the hsr locus (data not shown). In the other strains, there was minor restriction fragment length polymorphism in this pattern, but the hybridizing fragments added up in size to a genomic region equivalent to that of the sequenced hsr locus. Rearrangement events involving the hsr gene, if they indeed occur, are therefore apparently restricted to the DNA within the hsr locus.

To investigate antigenic variability, the Hsr proteins from three strains were first purified from total membrane preparations (12) by preparative electrophoresis in a 1.5% agarose-0.5% sodium dodecyl sulfate (SDS) gel, prepared in SDS-polyacrylamide gel electrophoresis running buffer (15), followed by dialysis and concentration. Purity, adjudged by silver staining, was greater than 95%. Unfortunately, purified Hsr bound weakly to enzyme-linked immunosorbent assay (ELISA) plates under widely varied conditions, so a whole-cell ELISA (14) was performed with antiserum JA6. This was raised against gel-purified Hsr protein of H. mustelae strain 4298 by immunization of rabbits (12) and was used in immunological procedures as previously described (12). The optimal antigen ratio for bacterial cells was determined by serial fivefold dilution. The coating efficiencies of strains were monitored by absorbance (600 nm) of unbound cells and did not differ significantly. ELISA data (Fig. 3) showed that the antiserum against the strain 4298 Hsr protein reacted most strongly with cells of that strain, despite the conservation of the Hsr sequence outside the amino-terminal half of the molecule. Preadsorption by membranes of strain 4298 reduced reactivity with all three strains to the same low level. Adsorption by membranes of either strain F15 or F21 had relatively little effect on serum reaction with strain 4298. Reactivity with cells of strain F15 or F21 was less affected by preadsorption by membranes of a heterologous strain, except for strain 4298.

FIG. 3.

ELISA analysis of immunoreactivity of Hsr on the surface of H. mustelae cells. Cells of the indicated strain were used to coat the plates. Sera employed were JA6 antisera preadsorbed by membranes of Hsr-deficient strain 4298-C3A (open bars), strain 4298 (light gray bars), strain F15 (dark gray bars), and strain F21 (black bars). Values graphed are means of quadruplicate measurements, and error bars are standard deviations.

Immunoreactivity of Hsr was further analyzed by slot blot (Bio-Rad, Calif.). The concentration of purified, detergent-free, Hsr protein was determined (Bio-Rad assay), and 52 ng was applied per slot. A dilution series of membrane sample was separated by SDS-polyacrylamide gel electrophoresis, stained with Coomassie blue, and Hsr was quantified by densitometric analysis (NIH Image software version 1.58) relative to purified protein. Membrane preparations applied contained 270 ng of Hsr protein per slot, a loading established by trial to allow chromogenic development of the immunoblotted pure protein at a similar rate. All samples were applied in a 100-μl volume of 10 mM phosphate-buffered saline, pH 7.4. Hsr proteins from all strains reacted at similarly high levels, whereas membrane preparations reacted differently, being weaker for strains F15 and F21 (Fig. 4). This suggested that there were numerous common epitopes between the purified Hsr proteins but that they were not accessible on membrane preparations (Fig. 4). Preadsorption by membranes reduced the reactivity with membrane preparations but had little effect on reactivity with the purified proteins (Fig. 4). Adsorption by membranes of an Hsr-deficient strain (see below) had little effect on reactivity of purified Hsr proteins, though it emphasized the differences in reaction strength of the membrane samples (Fig. 4).

FIG. 4.

Slot immunoblot analysis of Hsr. Rows contained membrane preparations of strains 4298 (A), F15 (B), or F21 (C), or purified Hsr protein from strain 4298 (E), F15 (F), or F21 (G). Row D contained membrane preparation from the Hsr-deficient strain 4298-C3A. Strips were reacted with either unadsorbed JA6 antiserum (column 1), or JA6 antiserum adsorbed by membrane of strain 4298 (column 2), strain F15 (column 3), strain F21 (column 4), or an hsr mutant of strain 4298 (column 5).

Although there is sequence variability in Hsr, the conservation of large blocks suggests functional constraints on the molecule. A possible role for this abundant surface protein might be adhesion, as described for other autotransporter proteins (recently reviewed [8]). To test this possibility, mutants defective for Hsr production were constructed by insertion of a cat marker (18) following published protocols (13), either in the junction of the nonvariable region and the β-barrel domain (mutant 219) or in the variable region (mutant 4298-C3A) of Hsr. Both mutants were verified genotypically by PCR to confirm single insertion of the cat marker into hsr by double crossover. They grow at the same rate in laboratory culture as the wild-type strain, and neither produces truncated Hsr detectable by Western immunoblotting (with anti-Hsr antibody JA6), presumably because the truncated protein is unstable.

Tissue culture and cell adhesion assays followed standard protocols (1). The gastric adenocarcinoma cell lines Kato III and AGS cells were from the American Type Culture Collection (Rockville, Md.). Ferret gastroduodenal cells were isolated as previously described (1). When the isogenic hsr-knockout strain 219 was compared with the wild-type strains for adhesion to two human gastric epithelial cell lines and ferret primary gastric epithelial cells, it adhered as well as the wild-type strains in all three assays (not shown). The mutant reproducibly adhered 15 to 20% better than the wild type to the human cell lines, but this effect was not seen with ferret epithelial cells and therefore can probably be discounted. These findings collectively suggest that Hsr is not an adhesin, and the function of this protein remains unclear. Experimental infection of ferrets is planned to explore the role of Hsr in pathogenesis; long-term colonization will probably be required to determine if antigenic variation in Hsr occurs.

Nucleotide sequence accession number.

The DNA sequence of the hsr locus has been deposited in the GenBank nucleotide sequence database under accession no. AF254134.