Functional Adaptation of BabA, the H. pylori ABO Blood Group Antigen Binding Adhesin

See allHide authors and affiliations

Science  23 Jul 2004:
Vol. 305, Issue 5683, pp. 519-522
DOI: 10.1126/science.1098801


Adherence by Helicobacter pylori increases the risk of gastric disease. Here, we report that more than 95% of strains that bind fucosylated blood group antigen bind A, B, and O antigens (generalists), whereas 60% of adherent South American Amerindian strains bind blood group O antigens best (specialists). This specialization coincides with the unique predominance of blood group O in these Amerindians. Strains differed about 1500-fold in binding affinities, and diversifying selection was evident in babA sequences. We propose that cycles of selection for increased and decreased bacterial adherence contribute to babA diversity and that these cycles have led to gradual replacement of generalist binding by specialist binding in blood group O–dominant human populations.

The gastric pathogen Helicobacter pylori causes chronic inflammation, which may progress to peptic ulceration and gastric cancer (1). H. pylori expresses adhesins that confer intimate adherence to the gastric epithelium, where the bacteria can gain easy access to nutrients from host tissues. Both fucosylated glycoproteins and sialylated glycolipids have been shown to be binding sites for H. pylori in the gastric epithelium (2, 3). The South American H. pylori strain P466 has been shown to bind the fucosylated H1 and Lewis b antigens (blood group O phenotype) but not the A–Lewis b antigen (blood group A phenotype) (2). These fucosylated blood group antigens are highly expressed in gastrointestinal epithelium (4). The H antigen is the carbohydrate structure that defines blood group O phenotype in the ABO blood group system. The Lewis b antigen (Leb), which is difucosylated, is formed by addition of a branched fucose (Fuc) residue to H1. The antigens that define blood group A and B phenotypes and corresponding antigens in the Lewis blood group system are formed by terminal N-acetylgalactosamine (GalNAc) or galactose (Gal) substitutions of H1 and Leb [A and A–Lewis b (ALeb) and B–Lewis b (BLeb) antigens, respectively] (Fig. 1A, fig. S1, and table S1). Epidemiologically, individuals of blood group O phenotype are particularly prone to peptic ulcer disease (5). Binding of H. pylori to H1 and Leb is mediated by the blood group antigen–binding adhesin BabA, an outer membrane protein (6) expressed by most disease-causing H. pylori strains (7).

Fig. 1.

(A) Structures of fucosylated blood group antigens and their relationship to the ABO system. Fuc, green; Gal, pink; GalNAc, blue (table S1). (B) ALeb inhibits adherence of H. pylori. Gastric mucosa stained with hematoxylin and eosin (i) and with Leb-mAb (arrow) (ii). Fluorescein isothiocyanate–labeled cells of strains 17875 (iii) and P466 (iv) adhere to the gastric epithelium. Pretreatment with ALeb eliminated adherence of 17875 (v) but did not affect P466 (vi). In contrast, pretreatment with Leb eliminated adherence of both 17875 (vii) and P466 (viii). Adherent bacteria were quantified in fig. S2. (C) H. pylori bound to soluble 125I-labeled blood group antigens. (D) Leb/ALeb binding ratios distinguish specialist and generalist strains. Sources of strains: SA/S and SA/G, South American specialists (48/77 = 62%) and generalists (29/77 = 38%); A, Alaska Native (24); J, Japanese (25); Sw, Swedish (76); G, German (41); S, Spanish (22). Strains tested are listed in Table 1, column I. Generalists were defined as exhibiting Leb/ALeb ratios of <2.5 (95% confidence interval). This confidence interval has its basis in consideration of Leb/ALeb binding ratios from all populations (table S4), except the South American population and one unusual Spanish specialist, strain S831 (arrow, Leb/ALeb ration > 70). Specialists stand out in this bimodal distribution as strains with Leb/ALeb rations ≥ 2.5.

The present study began with an observation that binding of H. pylori strain CCUG17875 (17875) to the gastric mucosa could be blocked by pretreatment with soluble ALeb in addition to Leb. However, adherence of strain P466 could not be blocked by pretreatment with ALeb even though it was inhibited by Leb (Fig. 1B and fig. S2). The binding specificities of several H. pylori strains were then studied with the use of soluble radiolabeled glycoconjugates (8). The blood group antigens used were all natural oligosaccharides, except for BLeb, which was synthesized chemically. Strains 17875 and J99 bound A, ALeb, BLeb, H1, and Leb, although subsequent tests indicated they did not bind the related Forssman antigen (with GalNAcα1.3 but not Fucα1.2) (9). By contrast, P466 only bound H1 and Leb (Fig. 1C and table S2). Similarly, with glycosphingolipids (GSLs) (8), strains 17875 and J99 each bound the difucosylated Leb, ALeb, and BLeb GSLs, whereas strain P466 only bound Leb GSL. Binding to monofucosylated GSLs was very weak or was observed only inconsistently. None of the strains bound to fucosylated antigens in the closely related type-2 core chain (fig. S3 and table S3). Consequently, we consider P466 to be a specialist strain of H. pylori and the other two to be generalists, able to tolerate bulky Gal and GalNAc end groups.

The distribution of binding specificities was studied further with the use of 373 H. pylori strains from various geographic regions (8) (Table 1 and table S4). Among the 188 strains from Swedes, Germans, Spaniards, Japanese, and Alaska Natives that bound Leb and/or ALeb, 178 (95%) bound both Leb and ALeb. In contrast, South American isolates from Peruvian and Venezuelan Amazon Amerindian populations and also from a Colombian mestizo (mixed Amerindian-European ancestry) population were distinctly different (P < 0.001), because 40% (31 of 77 isolates) bound Leb only. Each of 10 Leb-binding Peruvian strains that did not bind ALeb were also shown to be nonproficient in BLeb binding, and each of 10 Swedish strains that did bind ALeb also bound BLeb (table S5). A few ALeb-only binders were also found in most populations, but these strains were not studied further.

Table 1.

Prevalence of Leb and ALeb binding among 373 H. pylori isolates from different parts of the world. These binding analyses have their basis in the use of soluble 125I-labeled Leb and ALeb conjugates. An amount >1% bound of added conjugate defines a strain positive for binding. The final assignments of strains as specialists versus generalists has its basis in the Leb/ALeb binding ratios in Fig. 1D. Column I indicates total number of adherent isolates (binding Leb and/or ALeb). This total was subdivided into strains that bind both ALeb and Leb (in column II) and more specialized strains (III, binding Leb only, and IV, binding ALeb only). The percent value in column I is given in relation to the total number of strains; the percent values in columns II to IV relate to the total number of adherent isolates (column I).

Populations Total no. of strains I Leb and/or ALeb II ALeb and Leb III Leb only IV ALeb only
Swedish 102 76 (75%) 71 (93%) 2 (3%) 3 (4%)
German 75 41 (55%) 41 (100%) 0 0
Spanish 53 22 (42%) 20 (91%) 2 (9%) 0
Japanese 30 25 (83%) 24 (96%) 0 1 (4%)
Alaska Native 33 24 (73%) 22 (92%) 1 (4%) 1 (4%)
Subtotal: 293 188 178 5 5
South American
Peruvian Amerindian 59 59 (100%) 33 (56%) 24 (41%) 2 (3%)
Colombian Mestizo 14 12 (86%) 7 (58%) 5 (42%) 0
Venezuelan Amerindian 7 6 (86%) 4 (67%) 2 (33%) 0
Subtotal: 80 77 44 31 2
Grand total 373 265 (71%)

Many clinical isolates exhibited intermediate specialist or generalist binding phenotypes. Weighted comparisons were made to better distinguish strains with merely low Leb binding (and even lower ALeb binding) from apparent specialists with borderline phenotypes (strong Leb adherence plus some ALeb adherence). A bimodal distribution of binding ratios was found for South American isolates: 62% (48 of 77) bound to Leb much better than to ALeb (from 2.5-fold to about 100-fold) (specialists), whereas the remaining 38% had Leb/ALeb binding ratios equivalent to those of most Swedish, Spanish, German, Japanese, and Alaska Native isolates (generalists) (95% confidence interval, 0 to 2.5) (Fig. 1D and table S4). In contrast, only 5% (10 of 188) non-Amerindian strains exhibited specialist binding ratios.

The BabA adhesin was next analyzed for blood group A and B antigen binding (8). The babA-null mutant strains 17875babA1A2 and J99babA could not bind Leb, ALeb, or BLeb in soluble conjugates (Fig. 1C) or in GSLs (9). By receptor overlay, Leb and ALeb conjugates bound to the BabA band of about 70 kD from strain 17875, as identified with antibodies against BabA. In comparison, BabA from strain P466 bound only to Leb (fig. S4). Thus, the BabA protein alone is the adhesin that determines either generalist or specialist binding modes.

Soluble Leb and ALeb competed for binding to strain 17875, which suggested that these antigens bind at the very same site in BabA (fig. S5). However, ALeb affinity seemed to be lower than Leb affinity, because five times more bacterial cells were required to bind a given amount of ALeb than Leb conjugate. In accordance with this, Scatchard analysis revealed a sevenfold higher affinity for Leb than for ALeb (Ka values of 3.9 × 1011 M–1 and 5.5 × 1010 M–1, respectively) (fig. S6). A 1500-fold range in Leb affinities (2 × 108 M–1 to 3 × 1011 M–1, with a median of 2.2 × 1010 M–1) and a 300-fold range in ALeb affinities (2 × 108 M–1 to 6 × 1010 M–1, with a median of 2.6 × 109 M–1) were found among 68 Swedish clinical isolates. The Leb and ALeb affinities are strongly correlated (r = 0.74, P < 0.001), although with about fivefold (median) higher affinity for Leb (fig. S7 and table S4). We suggest that the higher affinity for Leb may cause stronger gastric mucosal adherence of H. pylori in blood group O individuals (with higher amounts of Leb in gastrointestinal mucosa compared to individuals of blood group A or B) and may thus contribute to their relatively higher risk of peptic ulcer disease (5). However, specialization does not confer further increases in Leb binding affinities as revealed by comparison of Peruvian specialists and Peruvian, Spanish, and Swedish generalists, because the affinities for Leb did not differ significantly (P > 0.05) (fig. S8). This suggests that binding affinities need to stay balanced for establishment of long-term infection.

To better understand BabA evolution, we analyzed the variable central region of babA phylogenetically (8). Maximum-likelihood (ML) analysis identified five distinct sequence clusters: one Peruvian, two mixed Peruvian and Spanish, one Japanese, and one Alaska Native (Fig. 2). Specialist and generalist alleles and alleles of high and low Leb binding affinities were found in each of the three Peruvian groups. Selection pressures on individual babA codons and in branches of the phylogenetic tree were estimated with the use of codon-based models of sequence evolution and ML (8). This analysis confidently predicted (Bayesian probability > 0.99) 23 sites under diversifying selection (for amino acid change) (fig. S9 and table S7), which suggests that BabA has been subject to heterogeneous selective pressures. Diversifying selection was also evident in several branches of the phylogenetic tree (fig. S10). We conclude that clustering of babA sequences does not depend on large functional motifs, that differences in strength and specificity of binding can each result from subtle differences in amino acid sequence, and that selection for differences in amino acid sequence has contributed to babA divergence and its adaptation to host environments.

Fig. 2.

ML analysis of babA sequences from 66 clinical isolates showing that specialists and generalists are not phylogenetically distinct. Branches with significant bootstrap support (≥ 50) are indicated; origins of H. pylori strains are color coded: red, Peru; green, Spain; blue, Alaska Natives; pink, Japan; and black, Venezuelan Amazon. ♦ indicates Peruvian specialists and ‡ indicates rare specialists from other geographic regions. The A714 babA branch has been truncated about sixfold to conserve space. Bar indicates 0.1 nucleotide substitution per site.

We used DNA transformation to test whether the generalist binding mode could be explained by minor functional motifs in babA itself. Specialist strains S831 and P445 were transformed with chromosomal DNA from generalist strain 17875, and ALeb binding transformants were enriched with the use of ALeb magnetic beads (8). Two independent ALeb binding transformants, S831G and P445G, were characterized (fig. S11 and table S8). First, immunoblotting with an antibody against a peptide specific for 17875 BabA (8) detected BabA in transformants but not in S831 or P445 parent strains (fig. S11B). Second, the gained ALeb affinity of transformant S831G was nearly identical to that of DNA-donor strain 17875, whereas that of transformant P445G was lower (2 × 108 M–1) but still within the range of generalist ALeb affinities (table S8). Third, Southern blot analysis (8) revealed the 17875 babA gene in each ALeb binding transformant (fig. S11C). Fourth, DNA sequencing identified a complete 17875 babA allele in S831G (fig. S12A) and a mosaic allele in P445G (P445 babA with about 1238 base pairs replaced by 17875 babA) (figs. S11D and S12B). Last, the 17875-derived babA allele was specifically deleted from each ALeb binding transformant by a second transformation with the use of the 17875 babA null-mutant vector (8). The resulting mutants neither bound ALeb nor reacted with the 17875 BabA antibody but still expressed BabA and bound only Leb, much as their S831 and P445 specialist ancestors did (fig. S11A and table S8). Thus, these tests showed that the generalist ALeb binding is an inherent property of the babA gene.

We propose that the current distribution of babA alleles reflects long-term adaptation to the types of receptors available in local populations, punctuated by fine-tuning of adherence during persistent infection of any individual host. Although adherence should benefit H. pylori by allowing better access to nutrients and delivery of effector molecules, tight adherence may be deleterious when host responses are robust. Recently, H. pylori infection and gastritis were found to promote gastric mucosal expression of inflammation-associated sialyl-Lewis x/a antigens (3), in competition with the fucosylated antigens that were studied here (10). Changes in BabA adhesins that help strains adapt to host gastric environments could arise by point mutation or short patch recombination between strains, or between divergent alleles in the same strain as illustrated by the ALeb binding transformant P445G [(1113) and this study]. Such new babA alleles will often differ from their parents in affinity, and those that are best suited to the local gastric environment (whether they are of higher or lower affinity) will be selected. Such flexibility should help to ensure rapid adaptation of H. pylori populations to the glyco-phenotype and host response of each infected person.

The postulated cycles of selection for decreased and then increased adherence during chronic infection and transmission from one person to another should result in retention of the ALeb, BLeb, and Leb generalist binding modes in most human populations because of the abundance of A, B, and O blood groups in them. In explaining the abundance of Leb-only specialist strains in South America, we note that Amerindians of this region are unique in being almost entirely of blood group O phenotype (14). One might invoke the idea of selection for dedicated specialists in this population. However, because the distribution of Leb affinities of Peruvian specialists is similar to those of generalists everywhere, we prefer an alternative explanation, which invokes recurrent cycles of selection for loss and restoration of binding activity: Only restoration of Leb binding activity would be selected for in any uniformly blood group O population; thus generalist binding would be lost by attrition.

Most alleles of housekeeping genes in the Peruvian H. pylori strains studied here were closely related to those found in Spanish strains but not those of Asian strains (15, 16). This implies descent of these Peruvian strains mainly from European strains (16). If this is correct, most Peruvian specialist babA alleles may have arisen by mutation and/or recombination over only the last ∼500 years. We propose that such rapid evolvability of the BabA adhesin in response to host mucosal glycosylation patterns fine-tunes H. pylori strains to their individual hosts, helps them to avoid the most deleterious of host responses, and contributes importantly to the extraordinary chronicity of human H. pylori infection worldwide.

Supporting Online Material

Materials and Methods

Figs. S1 to S12

Tables S1 to S8

References and Notes

View Abstract

Stay Connected to Science

Navigate This Article