Allelic Variation within Helicobacter pylori babA and babB

A total of 42 H. pylori strains were obtained from patients from several different locales and with differing clinical diagnoses (Table1). Strains were cultured at 37°C in a 5% CO₂ atmosphere on Trypticase soy agar plates with 5% sheep blood (TSAB) and frozen at −70°C until used in this study.

Genomic DNA was prepared from each strain after 48 to 72 h of growth on TSAB plates, as described previously (47). Strains were examined by PCR to determine vacA allelic types and the presence of cagA, iceA1, and iceA2, using established primers (10, 43, 48). PCR products were analyzed by agarose gel electrophoresis (10).

To compare sequences of babA and other homologs from strains J99 and 26695, nucleotide searches were performed using BLAST algorithms (9) from GenBank and The Institute for Genomic Research (http://www.tigr.org/tdb/mdb/hpdb/hpdb.html ) and Astra-Boston (http://scriabin.astrazeneca-boston.com/hpylori/ ) databases. Translations of nucleotide sequences were performed using GCG (Genetics Computer Group; Madison, Wis.) Translate, and pairwise similarities were determined using GCG Gap. Points of division were determined based on an alignment of the paralogues, where the first division point corresponds to the beginning of the HP722 open reading frame (ORF) (nucleotides 1 to 360 of 26695 babA), the second division point corresponds to the beginning of the HP229 ORF (nucleotides 361 to 760), and the third division point is based on the midpoint of the HP229 ORF (nucleotides 761 to 1510 and nucleotide 1511 to the end of each ORF). Secondary structure analysis was performed using Garnier's algorithm (21) athttp: //pbil.ibcp.fr/cgi-bin/npsa_automat.pl?page=/NPSA/npsa_gor.html, transmembrane domains were deduced using the TMPred server athttp://www.ch.embnet.org/software/TMPRED_form.html (24), and isoelectric points were deduced using the server at http://www.expasy.ch/tools/pi_tool.html (46).

babB fragments were amplified using primers AN5036 and AN5037 (Table 2).babA fragments were amplified primarily with primers C6678 and C6679; however, alternate combinations of forward and reverse primers among C6678, C6679, B1999, BA909, and B1998 were used to amplify fragments from strains that did not amplify with the primary primer set. PCR fragments were purified using the QiaQuick PCR purification kit and the QiaQuick Gel Extraction kit. PCR fragments were directly sequenced on both strands, using an automated Applied Biosystems, Inc., sequencer at the Vanderbilt Cancer Center sequencing core facility, and were analyzed using Sequencer 3.1.1 (Gene Codes Corp, Inc., Ann Arbor, Mich.).

Multiple alignments of nucleotide and predicted amino acid sequences were created using GCG Pileup, ClustalW, and GCG Pretty (Wisconsin Package version 9.1; Genetics Computer Group). Similarity plots were created using SimPlot 2.5 (http://www.welch.jhu.edu/ s̃ray/download). Phylograms of each nucleotide alignment were generated using both parsimony and distance matrix methods, using Paup 4.0b2 (Sinauer Associates, Sunderland, Mass.). All phylograms were displayed using Treeview (35) and Paup 3.1 (Illinois Natural History Survey, Champaign), using midpoint rooting. Transitions and transversions were calculated with MEGA 1.01 (Pennsylvania State University, University Park). Nucleotide and amino acid similarities of both babA and babB fragments were determined using BoxShade 3.2. Average GC content was calculated using GeneDoc 2.5 (K. B. Nicholas and H. B. Nicholas, Jr.). Synonymous and nonsynonymous substitution rates were generated using DnaSP 3.0 (37). The amount of homoplasy (H ratio) was determined as the mean of five independent tests using the homoplasy test (33), and all sites containing gaps were excluded from the analysis. The consistency index was determined using MacClade Version 3.0 (Sinauer Associates). Compatibility matrices were created using Reticulate (26). All means, standard deviations, and ratios were determined using Corel Quatro Pro (Corel Corp., Ottawa, Canada).

ThebabA and babB ORFs were amplified from H. pylori strain J99, using primer pairs B9617-AN5954 and C5773-C5774, respectively (Table 2). Each ORF was cloned into pGem-T Easy and transformed into Escherichia coli DH5α. A uniqueXbaI site was used to linearize the resulting plasmids containing babA, and a unique BamHI site was used to linearize babB-containing plasmids. The aphAcassette (kanamycin resistance) was amplified from plasmid pUC4K using primer pair C7491-C7492 or AN5485-AN5486 with either XbaI orBamHI restriction ends, respectively (Table 2). The resulting fragments were then cloned into the respective unique sites in babA and babB, creating plasmids pDP501 and pDP601. Strain J99 was transformed to Kan with both plasmids, and chromosomal DNA was isolated from each resulting strain. Insertion of the aphA cassette into babA or babBwas confirmed using primer pairs AN5045-AN5954 and AN5043-C5774, respectively.

H. pylori strains were cultured for 48 h at 37°C in a 5% CO₂ atmosphere on TSAB plates. Cells were harvested into 50 ml of brucella broth supplemented with 5% newborn calf serum and grown for another 48 h. Cells then were harvested and diluted to an optical density (OD) of 1.0 at 450 nm. One milliliter of each culture was then resuspended in a Phosphate-buffered saline (PBS) solution containing Thimersol and Tween 20 (PBSTT), plus 0.1% gelatin, and incubated at 37°C for 1 h. Cells were washed in PBSTT and incubated with or without 1 μg of Lewis B-human serum albumin glycocongugate (Isosep, Tullinge, Sweden) for 1 h at 37°C in PBS containing 0.5% bovine serum albumin. Cells were washed, followed by incubation for 1 h at 37°C with anti-Lewis B murine monoclonal antibody (Signet Pathology Laboratories, Inc., Dedham, Mass.) in PBS containing 0.1% gelatin and 0.5% bovine gamma globulin. Cells were washed, followed by incubation with anti-mouse immunoglobulin M horseradish peroxidase conjugate (Cappel Laboratories, Cochranville, Pa.) for 1 h at 37°C in a solution containing 0.1% bovine gamma globulin and 1% bovine serum albumin. Cells were washed twice and developed for 30 min in 500 μl of a solution containing 9.4 ml of 0.2 M Na₂HPO₄, 10.6 ml of 0.1 M citric acid, and 32 μl of H₂O₂. Results were expressed in relative units (OD × 1,000) as the OD difference between cells with Lewis B and those incubated without Lewis B. Each experiment was repeated on consecutive days.

babAGenBank accession numbers are AF277904 to AF277942, and babBGenBank accession numbers are AF277943 to AF277981 .

A GenBank BLAST search of the genome in H. pylori strain 26695 usingbabA (HP1243) nucleotides revealed 24 genes with scores of >100 and probabilities ranging from 0.0 to 0.998. For 11 genes, probabilities ranged from 0.0 × 10⁻¹¹⁸ to 1.9 × 10⁻¹¹⁸, whereas for the other 13 genes, probabilities were substantially lower. Using babA(JHP833) from H. pylori strain J99 to search that strain's ORFs, 22 genes were identified with scores of >100, with 10 having probabilities ranging from 0.0 × 10⁻⁹⁵ to 7.5 × 10⁻⁹⁵ and the other 12 genes having substantially lower scores. Therefore, our further analysis concentrated on the strongest paralogues in each strain.

All of the 11 genes from 26695 are members of a paralogous family, previously identified as outer membrane proteins through the characteristic C-terminal alternating hydrophobic residues (40), with at least one N-terminal domain of similarity and at least seven C-terminal domains of similarity. Two genes (HP25 and HP229) are predicted on the basis of sequence homology to encode porins (18, 19). These 11 genes (10 genes from J99) had significant 5′ similarity (over 350 nucleotides) but much greater 3′ similarity (over 750 to 1,500 nucleotides).

These 11 genes (10 genes in J99) share other properties. By Garnier's secondary structure analysis, each protein sequence was predicted to have parallel motifs at similar relative positions (data not shown). All have predicted basic pIs ranging from 8.47 to 9.24 (data not shown), except for two paralogues encoded by JHP212 and JHP1261 from strain J99 (both with pIs of 6.52). Each ORF has at least two predicted transmembrane domains (one each within the 3′ and 5′ regions), and all contain a predicted N-terminal signal peptidase I signal sequence (40). Based on the BLAST search probability scores and the similarity of the gene motifs and specific structures, we have classified these 11 genes (10 genes in J99) as babAparalogues.

Comparison of the 26695babA paralogues with reference gene babA (HP1243) from 26695 (data not shown) showed that the first segment (nucleotides 1 to 360) is well conserved among five genes; however, the other five genes either have a truncated version or do not possess this segment. Pairwise similarity comparisons show that HP227 and HP1342 are identical throughout their ORFs and that HP317 is nearly identical tobabA (HP1243) in this first segment; most of the paralogues have 61 to 100% similarity to each other in this same region (Table3). The next segment (nucleotides 361 to 760) is more variable in that none are >64% similar to one another, except for HP725 and HP722 (73% identical to each other), and the third segment (nucleotides 761 to 1510) also has substantial variability (Table 4). However, the fourth segment (nucleotide 1511 to the end of each ORF) showed similarities ranging from 67 to 100% among the different paralogues. These analyses indicate that there is substantial 5′ and 3′ conservation of these paralogues, with the greatest variation in the midregions. The mean (± standard deviation) (both calculated with Corel Quatro Pro) percent nucleotide similarities for the various regions of the babA paralogues were as follows: nucleotides 1 to 360, (68.0 ± 14.5)%; nucleotides 361 to 760, (42.9 ± 13.2)%; nucleotides 761 to 1510, (48.5 ± 13.7)%; and nucleotide 1511 to the end, (76.0 ± 8.11)%. A parallel analysis of the 10 babA paralogues in strain J99 identified similar relationships (data not shown).

Most of thebabA paralogues are in similar genomic locations in 26695 and J99, based on their flanking genes. Each is flanked by at least one gene that is identical between the strains with the exception of babA, babB, and HP25 (data not shown). Although the genes flanking babA in 26695 flankbabB in J99, the genes that flank babB in 26695 do not immediately flank babA in J99. The only significant identity in the upstream intergenic regions exists betweenbabA (HP1243) and HP317 (74% over 300 nucleotides), its most similar paralogue in strain 26695, and between HP722 and HP725 (88% over 290 nucleotides). In contrast, there is substantial (74 to 98%) identity in the intergenic regions downstream of babA, babB, and HP317. Although babA and HP317 are most similar, the regions downstream of babB and HP317 are 98% identical. Since HP317 and babB are flanked by identical downstream genes (HP316 and HP895), there is virtual identity for 900 bp downstream (data not shown).

To better understand the origins of these genes, the phylogeny of the 21 paralogues from the two strains was determined (Fig.1). Each 26695 ORF is most closely related to an ORF in J99, indicating the substantial interstrain conservation of each paralogue. The identical ORFs in 26695 (HP227 and HP1342) are closely related to the identical ORFs in J99 (JHP212 and JHP1261). ORF HP317 has no J99 counterpart. The phylogeny confirms the expectation that babA, HP317, and babB all are highly related and that HP722 and HP725 are highly related to each other. All bootstrap values on the phylogram are ≥75, and essentially identical phylogenies were produced using maximum parsimony algorithms (data not shown).

Although the genomes of 26695 and J99 have an overall 94% conservation at the nucleotide level (5), the 10 paired babAparalogues ([90.4 ± 2.4]%) are less highly related (Fig. 1). Levels of amino acid identity generally parallel the nucleotide identity for each pair of homologs (data not shown). The average GC content for the H. pylori genome is 39% (5, 40), while for the 11 paralogues in 26695, the mean GC content is (42.9 ± 1.6)% (range from 41 to 46%). For the J99 paralogues, the mean GC content is (43.1 ± 1.6)% (range from 41 to 45%). The GC content of the cluster (Fig. 1, cluster A) of paralogues including hopA is (41.4 ± 0.6)%, whereas for the cluster (Fig. 1, cluster B) of paralogues includingbabA it is (44.2 ± 1.0)%. This significant (P < 0.001, Student's t test) difference is further support for the validity of the phylogram.

Because of the segmental conservation and diversity among these paralogues, we next examined the best-characterized paralogues, babA andbabB, for their diversity among different H. pylori strains. From multiple alignments of the four and three sequences of babA and babB available from different strains, respectively (5, 25, 40; unpublished data), there is strong 5′ and 3′ conservation for both genes, with the greatest variation occurring in the midregions (data not shown). The greatest diversity occurs between nucleotides 612 and 1046 inbabA (86% mean identity) and between 488 and 1010 inbabB (91% mean identity). We then examined thesebabA and babB regions by directly sequencing PCR products from 42 H. pylori strains from around the world (Table 1). The babA sequences from the different strains are highly variable (Fig. 2); however, the segment including predicted amino acids 306 to 334 (nucleotides 915 to 999) shows five distinct families of variants, designated allele groups AD1 (babA diversity allele 1), AD2, AD3, AD4, and AD5 (Table1), as determined through similarity plot analysis (Fig.3A and data not shown). However, there exists a spectrum of similarity among the different babAallele groups, where groups AD3 and AD4 are most similar to one another, followed by group AD2, while groups AD1 and AD5 have little overall similarity to any of the other allele groups (Fig. 3B and data not shown). Phylogenetic analyses of nucleotides 915 to 999 also support the grouping of five separate babA variant families (Fig. 4, segment 1 phylogram), yielding results congruent with those of the similarity plots, while parallel analyses of the flanking nucleotides (830 to 914 and 1000 to 1084) do not show these allelic groupings (Fig. 4 and data not shown). All 42 strains belong to one of these allele groups, with AD5 strain 97-147 resembling each of the other four allele groups (Fig. 2).

Analysis of the translated babB sequences (nucleotides 488 to 1010 of 26695babB) suggests that babB is more conserved than is babA; however, the region between amino acids 216 and 252 (nucleotides 646 to 754) shows three families of variants, designated allele groups BD1 (babB diversity allele 1), BD2, and BD3 (Fig. 5 and Table 1), as initially discovered through similarity plot analysis (Fig. 3C and data not shown). As for babA, there also exists a spectrum of similarity, where allele groups BD2 and BD3 are most similar to one another, while group BD1 has little overall similarity to the other allele groups (Fig. 3D). Phylogenetic analysis also reinforces these groupings (Fig. 6, segment 1 phylogram), but these groupings are not present in the flanking (537 to 645 and 755 to 863) nucleotides. Importantly, strains that are clustered based on particular babA allelic types are not clustered based on their babB allelic types (Fig. 4 and 6).

Phylogenetic analysis of the regions 3′ (segment 3) of the babA and babB allele group regions demonstrates geographic clustering of the strains (Fig. 4 and 6). All but three of the U.S. strains cluster on the upper half of thebabA segment 3 phylogram, while two of the remaining three U.S. strains cluster near three of the four South American strains on the lower half. Most strains from Europe and Oceania cluster with the U.S. strains. Many of the Indian strains form a small cluster on the upper half, and all of the East Asian strains are found on the lower half of the phylogram. In the phylogram of babB segment 3, all of the Japanese strains are present on the upper branch, whereas all but one of the U.S. and South American strains are on the lower half. Thus, for both babA and babB, allele groupings (segment 1) that are not related by geographic origin of the strains and that are 5′ of regions (segment 3) that are largely related according to geographic origin are present. The regions 5′ of segment 1 for both genes also show evidence of geographic clustering (data not shown).

Among the 42 strains studied, the babAdiversity region varied in length from 434 to 488 bp (>10%), whereas there was little babB length heterogeneity (514 to 526 bp). The babB diversity region also was much more conserved at both the nucleotide (87 versus 78%) and amino acid (84 versus 74%) levels than was the babA diversity region. Nearly everybabA and babB fragment was unique, except that Indian strains 98-344 and 97-679 had identical babBfragments while differing in other genotypes (Table 1). The GC contents for the fragments (46 and 43% for babA and babB, respectively) resembled those for the entire genes (Fig. 1). Both transitions and transversions occurred at high levels but more often inbabA than in babB (data not shown), further indicating the diversity in these regions.

To explore the evolutionary basis for variation in thebabA and babB diversity regions, we analyzedK_s (mean number of synonymous substitutions per site), K_a (mean number of non synonymous substitutions per site), and theK_a/K_s ratio for all possible pairs of aligned sequences. Since synonymous base changes do not alter amino acid translations, their usage is less constrained than that of nonsynonymous base changes, and thus K_sshould roughly reflect the divergence time between sequences. The greater value of K_s for babA than forbabB (0.33 versus 0.23) suggests a more distant common ancestor for babA. Since theK_a/K_s ratio controls for differences in divergence time, the ratios' similarity forbabA and babB (0.42 versus 0.51) indicates that these two regions are subject to parallel functional constraints, which are less than previously determined within single alleles for the 5′ region of cagA (42) and the midregion ofvacA (11). Comparisons ofK_a/K_s ratios among thebabA or babB allele groups indicate that the functional constraints for each allele group are similar (data not shown).

Homoplasy is similarity within a locus that is not attributable to common ancestry and may reflect convergent, parallel, or reverse evolution (MacClade). In the absence of specific selective influences, recombination is believed to be the most important mechanism resulting in homoplasies on bacteria. The consistency index measures the amount of homoplasies in a population (from 0.0 to 1.0), where 1.0 indicates the absence of homoplasy. Phylograms of the babA and babB diversity regions have identical consistency indices of 0.34, indicating that there is substantial homoplasy among both these sequences. To determine whether the homoplasy is due to recombination, both the babA andbabB diversity region fragments then were examined by the homoplasy test (33); which measures the frequency with which the same nucleotide changes (homoplasies) occur in different branches of a maximum parsimony tree. The product, the Hratio, represents the frequency of observed synonymous homoplasies relative to the frequencies expected for the observed sequence variation under clonal descent (in the absence of recombination,H = 0) or free recombination (H = 1) (1, 33, 38). Several H. pylori housekeeping genes yield ratios ranging from 0.60 to 0.79, indicating high levels of recombination for these genes (1). AlthoughvacA gene fragments yield an H ratio of 0.69,cagA gene fragments yield H ratios of 0.15 to 0.17, among the lowest ratios yet observed for any species (38). babA and babB fragments yielded H ratios of 0.71 and 0.67, respectively, indicating that recombination is frequent in both of these loci and of a magnitude similar to that in the housekeeping genes.

As a further measure of the extent of recombination withinbabA and babB, compatibility matrices were used. In compatibility matrices, white spaces represent pairs of informative sites that are compatible with a maximal parsimony tree, whereas black spaces show pairs of sites that are incompatible. The mean compatibility score measures the average frequency with which pairs of informative sites are compatible. Both the babA and babB matrices are largely filled with black spaces (Fig. 7), and both yield similarly low mean compatibility scores (0.31 and 0.34, respectively), confirming that repeated recombination events have been involved in the evolution of both gene segments.

To determine if there is any association betweenbabA and babB allele groups and the ability of the strains to bind Lewis B, the Lewis B binding capacity of a group of strains representing each allele group was assessed. There was substantial variation from <10 to 658 units (mean, 346 ± 282) (see Materials and Methods for description of units) among these strains; however, there was no association between the allele group status of strains from each allele group and Lewis B binding (data not shown). Each babB allele group included strains that bound Lewis B well, or not at all, while strains from each babAallele group showed a broad range of values. Isogenic babAand babB mutants also were tested in the assay (Table5). The babA mutants failed to bind Lewis B, whereas the babB mutants bound to a similar degree as did the wild-type strains (Table 5). These results further indicate that, while babA is involved in binding to Lewis B (25), babB is not and that the particularbabA and babB allele groups are not determining factors in Lewis B binding.