Dynamics of Lewis b binding and sequence variation of the babA adhesin gene during chronic Helicobacter pylori infection in humans.

UNLABELLED: Helicobacter pylori undergoes rapid microevolution during chronic infection, but very little is known about how this affects host interaction factors. The best-studied adhesin of H. pylori is BabA, which mediates binding to the blood group antigen Lewis b [Le(b)]. To study the dynamics of Le(b) adherence during human infection, we analyzed paired H. pylori isolates obtained sequentially from chronically infected individuals. A complete loss or significant reduction of Le(b) binding was observed in strains from 5 out of 23 individuals, indicating that the Le(b) binding phenotype is quite stable during chronic human infection. Sequence comparisons of babA identified differences due to mutation and/or recombination in 12 out of 16 strain pairs analyzed. Most amino acid changes were found in the putative N-terminal extracellular adhesion domain. One strain pair that had changed from a Le(b) binding to a nonbinding phenotype was used to study the role of distinct sequence changes in Le(b) binding. By transformations of the nonbinding strain with a babA gene amplified from the binding strain, H. pylori strains with mosaic babA genes were generated. Recombinants were enriched for a gain of Le(b) binding by biopanning or for BabA expression on the bacterial surface by pulldown assay. With this approach, we identified several amino acid residues affecting the strength of Le(b) binding. Additionally, the data showed that the C terminus of BabA, which is predicted to encode an outer membrane β-barrel domain, plays an essential role in the biogenesis of this protein. IMPORTANCE: Helicobacter pylori causes a chronic infection of the human stomach that can lead to ulcers and cancer. The bacterium can bind to gastric epithelial cells with specialized outer membrane proteins. The best-studied protein is the BabA adhesin which binds to the Lewis b blood group antigen. Since H. pylori is a bacterium with very high genetic variability, we asked whether babA evolves during chronic infection and how mutations or recombination in babA affect binding. We found that BabA-mediated adherence was stable in most individuals but observed a complete loss of binding or reduced binding in 22% of individuals. One strain pair in which binding was lost was used to generate babA sequences that were mosaics of a functional allele and a nonfunctional allele, and the mosaic sequences were used to identify amino acids critically involved in binding of BabA to Lewis b.

INTRODUCTION

The human gastric pathogen Helicobacter pylori infects more than half of the world’s population. Infection is usually acquired during early childhood, persists lifelong, and results in chronic inflammation of the gastric epithelium, which remains asymptomatic in most cases. However, long-lasting chronic infection can lead to severe sequelae including gastric and duodenal ulcers, gastric cancer, or mucosa-associated lymphoid tissue (MALT) lymphoma (1).

H. pylori resides within the mucous layer in close proximity to the gastric epithelial cells (2–4). Dynamic adhesion to the mucosal epithelium plays an important role in establishing persistent colonization and induction of gastric inflammation (5). The H. pylori genome encodes a large number of outer membrane proteins (OMPs) (6). While the functions of most OMPs are still unknown, some of them have been shown to be involved in bacterial adhesion to host cells (7–12). The best-characterized OMP of H. pylori is the blood group antigen-binding adhesin (BabA) that mediates binding to fucosylated Lewis b [Le(b)] and related histo-blood group antigens located on the surfaces of gastric epithelial cells and mucins (10, 11). BabA expression has been shown to be associated with more severe inflammation and clinical disease (13). Despite this, not all H. pylori strains possess a babA gene, and not all BabA-expressing strains are able to bind to Le(b). In addition, BabA proteins from diverse H. pylori strains can vary substantially in their specificity and binding strength (14–16). The molecular basis of these different binding properties is not known, and so far no structural analysis of BabA has been published.

babA belongs to the Hop group of OMPs which show remarkable homology at their 5′ and 3′ ends (6, 12). The two most closely related babA paralogs are babB and babC for which no function has been identified. The extensive sequence homology between babA, babB, and babC has been shown to enable intrachromosomal recombination between the three genes (17). babA and its two closely related paralogs have been found in three different chromosomal locations, referred to as locus A, locus B, and locus C, respectively (14, 16). Experimental infection of different animal models with a Le(b) binding H. pylori strain resulted in the frequent loss of the Le(b) binding ability in the course of chronic infection (18–20). Based on these findings, it has been suggested that the dynamic modulation of adherence might facilitate adaptation of H. pylori to the changing conditions in the gastric environment (5).

In the present study, we aimed to characterize the dynamics of Le(b) binding and sequence variation of babA during chronic infection of humans. We analyzed a total of 47 sequential H. pylori isolates obtained from 23 individuals during two clinical trials in Louisiana in the United States (21) and Colombia (22). We observed a relatively stable Le(b) binding phenotype among isolates from the same individual. The BabA sequences of sequential isolates differed mostly in the putative N-terminal extracellular adhesion domain. Studies of strains harboring mosaic babA genes with different Le(b) binding properties demonstrated the importance of the C terminus for expression of a stable BabA protein and identified specific amino acids important for ligand binding.

RESULTS

Lewis b binding and BabA expression in sequential H. pylori isolates.

Previous studies performed in different animal models have shown that H. pylori frequently loses its ability to bind Le(b) during infection (18–20). To analyze the dynamics of Le(b) adherence during chronic H. pylori infection in humans, we determined the Le(b) binding properties and BabA expression for a total of 47 sequential H. pylori isolates obtained from 23 individuals participating in two clinical trials in Louisiana in the United States (21) and Colombia (22), respectively (Table 1). Twenty-three pairs of strains were isolated at intervals ranging from 3 months to 4 years after the initial biopsy specimen; for one individual, a third isolate was obtained 16 years after the initial biopsy specimen. Altogether, 27 strains (57%) exhibited Le(b) binding, with marked differences in binding efficiencies, while 20 strains (43%) did not bind to Le(b). As expected, all Le(b) binding strains expressed the BabA adhesin, while in most of the strains that did not bind Le(b) (16/20 [80%]), no BabA protein was detectable (summarized in Table 1; also see Fig. S1 in the supplemental material). The differences in Le(b) binding were significantly associated with the origin of the strain (P = 0.014 by the chi-square test). Almost all strains from Colombia (NQ strains; 13/15 [87%]) displayed a Le(b) binding phenotype, whereas less than half of the strains from the United States (LSU strains; 14/32 [44%]) showed binding to Le(b).

TABLE 1 

Le(b) binding and bab genotyping of sequential H. pylori isolates

Stain set no.[a]	Strain[b]	Le(b) binding[c]	Stat. sign.[d]	Protein expression		bab genotype[e]		babA full-length sequencing[f]
				BabA	BabB	Locus A	Locus B	Locus	babA length (bp)	BabA length (aa)
1	NQ267	4.9 ± 1.8	ns	+	−	B	A	B	2,253	750
	NQ1624	3.5 ± 0.8		+	+	B	A	B	2,241	746
2	NQ315	3.3 ± 1.0	ns	+	+	A	A	B	2,229	742
	NQ1712	4.5 ± 2.5		+	+	A	A	B	2,229	742
3	NQ331	2.3 ± 0.7	ns	+	+	AB	A	B	2,235	744
	NQ1832	4.6 ± 1.7		+	+	AB	A	B	2,235	744
4	NQ352	6.9 ± 2.6	P ≤ 0.05	+	−	A	B	A	2,211	736
	NQ1701	1.3 ± 0.7		+	+	A	B	A	2,208	735
5	NQ366	3.7 ± 0.7	P ≤ 0.01	+	+	AB	A	B	2,247	748
	NQ1790	2.1 ± 0.4		+	+	AB	A	B	2,247	748
6	NQ367	4.2 ± 0.8	ns/P ≤ 0.01	+	+	A	B	A	2,250	749
	NQ1671	3.3 ± 1.3	P ≤ 0.05	+	+	A	B	A	2,250	749
	NQ4191[g]	0.9 ± 0.1		−	+	B	B
7	NQ372	1.8 ± 0.6	ns	+	+	A	A	B	2,232	743
	NQ1886	1.7 ± 0.7		+	+	A	A	B	2,232	743
8	LSU1010-1[g]	0.8 ± 0.3	ns	−	+	B	B
	LSU1010-3[g]	0.9 ± 0.2		−	+	B	B
9	LSU1013-2	4.4 ± 0.7	ns	+	+	A	AB	ND
	LSU1013-6	5.6 ± 2.1		+	+	A	AB	A	2,235	744
10	LSU1014-1	3.8 ± 1.0	P ≤ 0.05	+	+	A	B	A	2,235	744
	LSU1014-6	0.7 ± 0.1		−	+	A	AB	A	2,214	737
11	LSU1016-1	6.6 ± 0.5	P ≤ 0.001	+	−	A	AB	A	2,238	745
	LSU1016-5	2.4 ± 0.5		+	+	A	AB	A	2,238	745
12	LSU1021-6	2.9 ± 1.0	ns	+	+	A	AB	A	2,220	739
	LSU1021-7	4.4 ± 2.3		+	+	A	AB	A	2,220	739
13	LSU1027-1	1.0 ± 0.1	ns	−	+	A	AB	A	759	252
	LSU1027-3	1.0 ± 0.1		−	+	A	AB	A	759	252
14	LSU1037-1[g]	0.9 ± 0.2	ns	−	+	B	B
	LSU1037-5[g]	0.9 ± 0.1		−	+	B	B
15	LSU1040-1[g]	0.8 ± 0.1	ns	−	+	B	B
	LSU1040-6[g]	1.0 ± 0.1		−	+	B	B
16	LSU1054-1	3.1 ± 1.8	ns	+	−	A	AB	A	2,226	741
	LSU1054-5	1.1 ± 0.2		+	+	A	AB	A	2,226	741
17	LSU1062-1	0.9 ± 0.2	ns	+	+	A	B	A	2,220	739
	LSU1062-3	0.9 ± 0.2		+	−	A	B	A	2,220	739
18	LSU1067-1	2.5 ± 1.1	ns	+	+	A	B	A	2,247	748
	LSU1067-5	3.2 ± 1.2		+	+	A	B	A	2,247	748
19	LSU1074-1	0.9 ± 0.2	ns	−	−	A	B	A	1,638	545
	LSU1074-4	0.7 ± 0.4		−	−	A	B	A	1,638	545
20	LSU2002-1[g]	0.8 ± 0.2	ns	−	+	B	B
	LSU2002-8[g]	0.7 ± 0.2		−	+	B	B
21	LSU2003-1	4.0 ± 1.4	ns	+	+	A	AB	ND
	LSU2003-7	5.0 ± 1.8		+	−	A	A	ND
22	LSU3001-1	1.9 ± 0.6	ns	+	−		A	B	2,214	737
	LSU3001-6	1.7 ± 0.6		+	−	A	A	B	2,217	738
23	LSU3005-1[g]	0.9 ± 0.0	ns	−	+	B	B
	LSU3005-4[g]	1.0 ± 0.4		−	+	B	B

Sequential isolates were obtained from 23 individuals.

The origin of H. pylori isolates is indicated as follows: NQ for the isolates from Colombia and LSU for isolates from Louisiana in the United States.

Values represent mean ± SD of Le(b)/BSA binding ratios determined in at least three independent experiments. Strains are considered nonadherent if the Le(b)/BSA ratio was ≤1.5.

Stat. sign., statistical significance. The values for Le(b) binding of sequential H. pylori isolates were compared by Welch’s t test. The P value of statistically significant differences is indicated. ns, no statistically significant difference.

A, babA; B, babB; AB, babA and babB.

babA genes at the indicated locus were completely sequenced. The length of the encoded BabA proteins (in amino acids [aa]) was deduced from the corresponding gene sequences. ND, no complete babA sequence was determined.

No babA gene was identified for this strain.

Most sets of sequential isolates (18/23 [78%]) did not exhibit significant differences in Le(b) binding or BabA expression; this included nine pairs with binding to Le(b) and eight pairs without binding. In five strain sets, we observed a statistically significant reduction (NQ366/1790, LSU1016-1/-5) or even complete loss (NQ352/1701, NQ367/1671/4191, LSU1014-1/-6) of Le(b) binding in the course of infection. One additional strain pair (LSU1054-1/-5) showed a loss of Le(b) binding in the later strain, but this effect was not statistically significant due to the unusually variable Le(b) binding of the first strain, LSU1054-1. In two cases, the loss of Le(b) binding was accompanied by a complete loss of BabA expression (NQ4191 and LSU1014-6) as detected by Western blotting.

bab genotyping of sequential H. pylori isolates.

To shed light on the genetic basis responsible for the observed differences in Le(b) binding and BabA expression (Table 1), we next aimed to determine the bab genotype of the sequential isolates and to identify babA-containing loci for the subsequent sequencing of babA genes. The three known chromosomal bab loci termed locus A, locus B, and locus C were analyzed for the presence of babA, babB, and babC by PCR and partial sequencing (summarized in Table 1). bab genotypes were highly diverse in our strain collection (Fig. 1). In total, 23% (11/47) of the isolates had a J99-like genotype (AB) with babA in locus A and babB in locus B. Only two strains (4%) showed a 26695-like genotype (BA) with the genomic location of babA and babB inverted compared to that in J99. Copies of only babA (AA) or only babB (BB) were present at both loci in 6 (13%) and 11 strains (23%), respectively. The BB genotype and the apparent lack of babA genes in these strains were in agreement with the lack of BabA expression and their inability to bind to Le(b). In one of the strains, we identified only a single babA copy in locus B but could not amplify any bab gene in locus A. We did not detect any bab gene in locus C and no babC gene in any of the analyzed loci. Finally, although all genomic DNAs used for bab genotyping were derived from low-passage-number single colonies of the respective H. pylori strains, we observed a mixed bab genotype (AB A and A AB) in 16 isolates (34%), a phenomenon that has been previously reported (18, 23). The same chromosomal location of babA and babB suggests a high frequency of recombination between the bab loci A and B, which may result in the rapid emergence of differing subpopulations.

FIG 1 

Distribution of bab genotypes. Frequency of babA and babB genes at the chromosomal loci A and B. None of the strains (n = 47) possessed babC in any of the three loci or had a bab gene in locus C. AB, babA in locus A and babB in locus B; BA, babB in locus A and babA in locus B; AA, babA in loci A and B; BB, babB in loci A and B; AB A, babA and babB in locus A and babA in locus B; A AB, babA in locus A and babA and babB in locus B; - A, no bab gene in locus A and babA in locus B.

Comparisons of the sequential isolates identified no change of bab genotype in most pairs of strains (19/23 [83%]). The remaining four strain sets differed in their bab genotype at one locus (LSU1014, LSU2003, LSU3001, and NQ367 triplet). Most interestingly, the third strain of the triplet (NQ4191) had replaced babA in locus A by a second babB copy, which explains both the observed absence of BabA expression and loss of Le(b) binding.

Sequence variation of babA.

bab genotyping identified 36 strains in our strain collection that harbored at least one copy of babA. Nine of these strains did not bind to Le(b), while the other 27 strains displayed Le(b) binding to various extents. To correlate the sequence diversity of BabA proteins with the observed differential Le(b) binding capacity, we next aimed to determine babA nucleotide sequences of these isolates. For strains with a mixed bab genotype, we restricted our analysis to loci exclusively containing babA. Altogether, we successfully obtained complete babA gene sequences from 33 of the 36 strains (Table 1). For three of the 27 Le(b) binding strains, we did not succeed in sequencing a babA gene. However, the observation of BabA expression and Le(b) binding indicates the presence of a functional babA gene in these strains.

Pairwise comparisons of the babA nucleotide sequences of 16 pairs of sequential isolates identified four pairs with identical babA genes (Fig. 2 and Table 2). The remaining 12 strain pairs showed various levels of differences including isolated single nucleotide polymorphisms (SNPs) and clusters of nucleotide polymorphisms (CNPs) indicative of recombination. The latter could be the result of an intragenomic recombination event or due to recombination with a different H. pylori strain in case of coinfection with more than one strain. Altogether, the sequence similarity of babA ranged between 97 and 100%. Only strain pair NQ267/1624, where nucleotide polymorphisms were distributed throughout the sequence, indicating that the complete gene had been exchanged by a single recombination event, displayed a markedly lower sequence similarity (91.4%). Five of the nine Le(b) nonbinding strains (LSU1027-1/-3, LSU1074-1/-4, and LSU1014-6) contained babA genes encoding premature stop codons at different positions (Fig. 2), which abolished BabA expression, while the other four strains harbored full-length babA genes encoding detectable BabA protein (Table 1; see Fig. S1 in the supplemental material). The deduced BabA amino acid sequences differed in 11 strain pairs (Fig. 2 and Table 2). Most differences were located in the predicted extracellular adhesion domain. The highest number of amino acid changes was observed in strain pairs NQ267/1624, LSU1014-1/-6, and NQ352/1701.

FIG 2 

babA sequence variation between sequential isolates. Pairwise alignments of babA nucleotide sequences and deduced BabA amino acid sequences were performed using MUSCLE (45). The polymorphic sites in each nucleotide alignment were determined using the program PSFIND and graphically displayed with the program HAPPLOT. Polymorphic sites are displayed by vertical lines. Corresponding plots for the protein alignments were generated manually. Blue vertical lines indicate differences between sequential isolates. The signal sequence (S) as well as the predicted domain structure based on Phyre2 analysis are indicated. aa, amino acids.

TABLE 2 

Comparison of babA sequences between pairs of sequential isolates

Strain set no.[a]	Strain	Interval between isolates (mo)	babA gene						BabA protein
			Length (bp)	Sequence similarity (%)	No. of PS[b]	No. of SNPs and/or CNPs[c]	S[d]	N[d]	Length (aa)	No. of aa changes	Sequence similarity (%)
1	NQ267	36	2,253	91.41	215	1 CNP	35.5	139.5	750	70	92.94
	NQ1624		2,241						746
2	NQ315	36	2,229	99.91	3	1 CNP	3	0	742	0	100
	NQ1712		2,229						742
3	NQ331	36	2,235	99.95	1	1 SNP	0	1	744	1	99.86
	NQ1832		2,235						744
4	NQ352	36	2,211	97.01	71	3 CNPs	46.75	21.25	736	19	98.09
	NQ1701		2,208						735
5	NQ366	36	2,247	99.73	6	1 CNP	5	1	748	1	99.86
	NQ1790		2,247						748
6	NQ367	36	2,250	99.15	19	1 SNP	14	5	749	4	99.59
	NQ1671		2,250			1 CNP			749
7	NQ372	36	2,232	100	0		0	0	743	0	100
	NQ1886		2,232						743
8	LSU1014-1	24	2,235	98.03	42	3 CNPs	12	11	744	23	96.63
	LSU1014-6		2,214						737
9	LSU1016-1	12	2,238	99.91	2	1 CNP	0	2	745	2	99.86
	LSU1016-5		2,238						745
10	LSU1021-6	12	2,220	100	0		0	0	739	0	100
	LSU1021-7		2,220						739
11	LSU1027-1	3	759	100	0		0	0	252	0	100
	LSU1027-3		759						252
12	LSU1054-1	12	2,226	99.86	3	1 CNP	1	2	741	2	99.73
	LSU1054-5		2,226						741
13	LSU1062-1	3	2,220	99.95	1	1 SNP	0	1	739	1	99.86
	LSU1062-3		2,220						739
14	LSU1067-1	12	2,247	98.26	40	2 CNPs	24	16	748	13	98.52
	LSU1067-5		2,247						748
15	LSU1074-1	6	1,638	100	0		0	0	545	0	100
	LSU1074-4		1,638						545
16	LSU3001-1	24	2,214	98.15	44	1 CNP	15	19	737	17	98.1
	LSU3001-6		2,217						738

babA sequences were obtained from 16 pairs of sequential isolates.

The number of polymorphic sites (PS) was determined with the program PSFind based on pairwise sequence alignments (MUSCLE) and includes substitutions, insertions, and deletions.

SNP, single nucleotide polymorphism; CNP, cluster of nucleotide polymorphisms (polymorphisms are separated by <200 bp).

The average number of synonymous sites (S) or nonsynonymous sites (N) was determined with the program MEGA5.

Next, we performed a multiple-sequence alignment (MSA) of BabA amino acid sequences from all BabA-expressing strains. After removal of duplicate sequences, the data set included 21 Le(b) binding strains and four Le(b) nonbinding strains (see Fig. S2 in the supplemental material). The highest sequence diversity was found in a region corresponding to amino acids 218 to 229 in the BabA sequence of strain J99. Based on this MSA, the sequence variability for each site was determined using the Protein Variability Server (PVS) (24). The highest diversity was observed in the putative N-terminal extracellular adhesion domain of BabA, while the C terminus, predicted to encode an outer membrane β-barrel, showed markedly less sequence variation (Fig. 3).

FIG 3 

Sequence variability of BabA from BabA-expressing NQ and LSU strains. The Shannon diversity index (H) for each site in the multiple-sequence alignment of BabA (see Fig. S2 in the supplemental material) was determined using the Protein Variability Server (PVS) (24). H can range from 0 (the same amino acid is present at that position in all sequences) to 4.322 (all 20 amino acids are equally represented in that position). The predicted domain structure based on Phyre2 analysis is indicated on top of the figure.

Characterization of strains with mosaic babA alleles.

We next aimed to study the correlation of distinct sequence changes in babA with the Le(b) binding ability in an isogenic strain background. Strain pair LSU1014-1/-6, which had changed from a Le(b) binding to a nonbinding phenotype, was selected for the generation of isogenic H. pylori strains harboring mosaic babA genes. The loss of Le(b) binding in strain LSU1014-6 was accompanied by the absence of BabA expression on Western blots (see Fig. S1 in the supplemental material). However, the level of babA transcript expressed by LSU1014-6 was similar to that expressed by LSU1014-1 (data not shown), indicating that the difference in the Le(b) binding phenotype was indeed related to distinct protein properties. The predicted BabA protein sequences of both strains differed in 11 amino acids scattered along the sequence, plus the C terminus (Fig. 4A). The deletion of a G nucleotide at position 2172 in a stretch of 5 G’s caused a frameshift in babA of strain LSU1014-6, resulting in 12 exchanged amino acids and a truncation of seven residues by a premature stop codon. To characterize the specific influence of these sequence differences on ligand binding and/or protein expression, we generated mosaic babA genes by transformation of strain LSU1014-6 with babA PCR products amplified from either the adherent strain LSU1014-1 or the nonadherent strain LSU1014-6 (for details, see Materials and Methods). H. pylori transformants were enriched for recombinants with restored Le(b) binding using a biopanning protocol or for expression of surface-located BabA via a pulldown assay. A total of 11 clones with mosaic babA alleles were single colony purified, subjected to babA gene sequencing (Fig. 4B), and characterized for BabA protein expression (Fig. S3) and Le(b) binding (Fig. 4C).

FIG 4 

Analysis of babA mosaic alleles. (A) Schematic representation of differences in the deduced BabA protein sequence of strain pair LSU1014-1/-6. The signal sequence (S) is indicated as well as the domain structure as predicted by Phyre2 analysis (44). BabA of strain LSU1014-6 differs in 11 amino acids and in addition contains a distinct and shortened C terminus due to a frameshift mutation. (B) Schematic representation of BabA sequences from the generated mosaic alleles based on strain pair LSU1014-1/-6. The numbers above the schematic representation of BabA in strain LSU1014-1 indicate the positions of amino acid changes. The length of the BabA protein (in amino acids [aa]) and result of BabA expression determined by Western blotting (WB) are indicated to the right of the schematic representations of the Bab sequences. (C) Le(b) binding was determined by ELISA and is represented as mean + SD from at least four independent experiments. All strains except LSU1014-6 and clone 135 had Le(b)/BSA ratios of >1.5. Le(b) binding of recombinant clone 1 was set at 100%, and statistical significance versus the value for clone 1 was determined by Welch’s t test. Statistical significance is indicated as follows: *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001; ns, not statistically significant.

Clone 1, which possessed a babA gene identical to the gene in strain LSU1014-1, expressed BabA and showed restored binding to Le(b) compared to strain LSU1014-6. Compared to LSU1014-1, binding was even significantly increased (1.6-fold). Le(b) binding of all other recombinant clones was quantitated and expressed in relation to that of clone 1. These clones contained mosaic babA genes with combinations of LSU1014-1 and LSU1014-6 specific residues, and with the exception of clone 135, they displayed both BabA expression and distinctly different binding to Le(b).

Clone 135 lacked both BabA protein expression and Le(b) binding, while reverse transcription-PCR analysis detected babA-specific mRNA as in strain LSU1014-6 (data not shown). BabA of clone 135 was identical to BabA of LSU1014-1 except for the distinct and truncated C terminus. In contrast, clone 522, whose babA allele was identical to LSU1014-6 except for the C terminus, expressed BabA and bound to Le(b), albeit to a reduced extent. This indicated that the C terminus is important for expression of stable BabA protein, whereas the 11 amino acid differences along the BabA sequence contribute to differences in Le(b) binding. Analysis of the remaining eight clones, which all contained the intact C terminus of strain LSU1014-1 allowed a closer inspection of the influences of these 11 amino acid differences on Le(b) binding. For example, clones 12 and 743 showed no increase in Le(b) binding compared to clone 522, indicating that the specific residues at positions 421 and 630 in strain LSU1014-1 were probably less important for ligand binding. In contrast, the presence of further LSU1014-1-specific residues in clone 362 (residues 378 and 385) and clone 5 (residues 212, 213, 215, 220, 225, 378, and 385) restored Le(b) binding back to the level of clone 1, suggesting that these residues play an important role in Le(b) binding. This was further confirmed in the remaining clones, which all showed diminished Le(b) binding.

To further analyze the importance of the C terminus for protein biogenesis, we performed structure prediction analyses of the BabA sequences using Phyre2 (Fig. 5). The structural model of the N-terminal adhesion domain of BabA was based on the previously published extracellular adhesion domain of H. pylori SabA (25) and was predicted with high confidence for all strains (100% confidence; >21% identity). In contrast, the confidence of the structure prediction of the C-terminal β-barrel membrane domain differed considerably between Le(b) binding and nonbinding strains (Fig. 5B). While the outer membrane domain of BabA from strain LSU1014-1 was predicted with high confidence (95.3%), confidence values for the membrane domains of strain LSU1014-6 (74.6%) and clone 135 (76.8%) were markedly lower, suggesting that the missing C terminus affects the correct folding of the β-barrel and may thereby influence stability.

FIG 5 

Phyre2 structure prediction of BabA. The BabA sequences of strains LSU1014-1 and LSU1014-6 and clone 135 were subjected to Phyre2 analysis for structure prediction (see Materials and Methods). (A and B) The models are colored showing the secondary structure (A) or confidence of the predicted structure (B), ranging from high (red) to low (blue). (C) Structural model generated with the BabA sequence of strain LSU1014-1 [Le(b) binding] with the amino acids that differed in strains LSU1014-1 and LSU1014-6 (not binding) indicated in orange. Additionally, the location of the distinct C terminus is colored in pink.

DISCUSSION

In this study, we analyzed the dynamics of Le(b) binding and sequence variation of babA during chronic H. pylori infection in humans. In contrast to reports from experimental infections in animal models (18–20), the Le(b) binding phenotype was relatively stable during human infection, as Le(b) binding ability changed in only 5 out of 23 sets of sequential isolates. This stable BabA phenotype might be due to the already long-standing infection and a stable physiological situation in the infected individuals when the first isolate was collected.

The reported loss of Le(b) binding in animal infection experiments has been the result of either gene conversion, frameshifts due to variation in length of repeat sequences, or mutations leading to amino acid substitutions or truncated proteins (18–20). We observed a loss of Le(b) binding due to replacement of babA with a second copy of babB in one case. In the other isolates, changes in Le(b) binding ability were caused by mutations leading to amino acid changes in BabA, which were mainly located in the predicted N-terminal adhesion domain of BabA. We did not observe loss of Le(b) binding due to phase variation in any of these isolates.

Since we analyzed only one purified H. pylori clone per biopsy, we cannot assess potential intrastrain variation of Le(b) binding within a stomach or between different parts of the stomach niche. Two recent studies of our group have analyzed multiple single colony isolates from two stomach biopsies for hopZ gene status and detected only minimal intrastrain variation (26, 27). Another study analyzing the bab genotype of more than one single colony isolate from each of three patients observed intrastrain genotype variation in two of the three cases (14).

Our experimental data and protein structure predictions suggest that the C terminus of BabA may be important for protein folding and possibly stability. The C terminus is predicted to encode an outer membrane β-barrel (28), and correct assembly into the outer membrane is essential for protein function. Folding and membrane insertion of OMPs are evolutionarily highly conserved and mediated by the BAM (β-barrel assembly machinery) complex (29, 30). The central and essential component of this multiprotein complex, BamA, is itself an OMP and recognizes substrate OMPs by a C-terminal signature sequence (31). This sequence is found in the majority of bacterial OMPs and is highly conserved among Proteobacteria (32, 33). Changes of this sequence have been shown to diminish or completely disturb OMP assembly (31, 33, 34). For example, deletion of the C-terminal phenylalanine of the Escherichia coli porin PhoE disabled BamA activity, while a synthetic peptide consisting of the last 12 amino acids was sufficient to activate BamA (31). So far, the BAM complex has not been analyzed in H. pylori. However, BamA is conserved in H. pylori (35), and it has been found in a recent analysis of the cell surface proteome of H. pylori (28). Thus, we assume that recognition of the truncated BabA protein by the BAM complex is hampered due to the missing C terminus, which might subsequently lead to rapid protein degradation to avoid accumulation in the periplasmic space. Interestingly, BabA belongs to the Y-Hop subgroup of H. pylori OMPs that encode a tyrosine residue at the final position (6). Future studies should concentrate on the impact of this amino acid on OMP assembly.

The BabA sequence comparison from Le(b) binding and nonbinding strains displayed a high level of variation in the predicted adhesion domain of BabA, while the C terminus was less variable. However, we were not able to identify specific amino acids essential for Le(b) binding. These finding are in agreement with a previous sequence comparison of Le(b) binding and nonbinding strains (15). On the basis of mosaic babA alleles, we identified some amino acids that affected the Le(b) binding strength but did not completely disable adherence. Further studies are needed to correlate protein sequence diversity and capacity of Le(b) binding.

The significantly increased binding of clone 1 compared to strain LSU1014-1 despite identical babA sequences is likely to be the result of integration of this allele in the different LSU1014-6 strain background of this transformant. Since only the coding babA sequence was used for transformation, the babA promoter region might differ and differentially influence gene transcription. Recently, it was demonstrated that variation in length of the thymine (T) nucleotide repeat located in the sabA promoter adjacent to the −35 element resulted in different transcriptional activity (36, 37). These differences led to distinct protein expression and receptor binding activity (36). The babA promoter of H. pylori strain 26695 contains a homopolymeric adenine (A) repeat between the −35 and −10 promoter elements (36), which might similarly affect gene transcription or transcript stability. Since the incorporation of BabA into the outer membrane is essential for its functional activity, differences in the functionality of the BAM complex, which is responsible for the correct assembly and membrane insertion of OMPs, might present an alternative explanation for the observed differences in Le(b) binding.

Our data demonstrate that the ability to bind to Le(b) is a relatively stable phenotype during human infection but that microevolution within babA occurs commonly, driven by mutation and recombination. These sequence changes can result in a complete loss of protein expression, such as the one observed in the sequence encoding the C terminus, or in gradual changes in binding properties, such as those caused by the amino acid replacements in the coding sequence other than the C terminus (Fig. 5C). It seems likely that these changes are the result of selection in vivo, reflecting changing conditions in the stomach. Our study is a step toward a better understanding of the correlations between BabA sequence and Le(b) binding phenotype, which will be greatly enhanced once BabA crystal structures together with its ligand Le(b) become available, permitting a mapping of the functional information provided here to specific parts of the interaction surface in the BabA structure. A detailed understanding of the critical amino acid residues involved in Le(b) binding will also be essential in guiding efforts toward pharmacological inhibition of BabA-Le(b) interactions in a therapeutic or prophylactic setting.

MATERIALS AND METHODS

H. pylori strains and culture conditions.

The sequential H. pylori isolates analyzed in this study were obtained during two clinical trials performed in Louisiana in the United States (21) and Colombia (22). These trials compared the development of gastric lesions in groups treated with antibiotics, sucralfate, or a dietary supplement. Only five of the 16 individuals from whom the LSU strains were isolated (LSU1013, LSU1014, LSU1021, LSU1040, and LSU1067), and none of the Colombian trial participants (NQ strains) had received antibiotics that might have caused a selective bottleneck in the H. pylori population despite the failure to achieve eradication. The genetic relationships of the sequential isolates were previously analyzed by extended multilocus sequence analysis (MLSA) (38). H. pylori strain J99 (39) and an isogenic J99babA mutant were used as controls in the Lewis b binding assay. For a complete list of strains, see Table S1 in the supplemental material. Bacteria were cultured from frozen stocks on blood agar plates (Columbia agar base II; Oxoid, Germany) supplemented with 10% horse blood (Oxoid), and the antibiotics vancomycin (10 mg/liter), polymyxin B (2500 U/liter), amphotericin B (4 mg/liter), and trimethoprim (5 mg/liter). Plates for the J99babA mutant carrying a chloramphenicol resistance cassette were additionally supplemented with chloramphenicol (20 mg/liter). Strains were incubated under microaerobic conditions (5% O2, 10% CO2, and 85% N2) at 37°C.

Generation of an isogenic J99babA mutant.

The J99babA mutant was constructed by natural transformation-mediated allelic exchange as described previously (40). Oligonucleotide sequences are provided in Table S2 in the supplemental material. Briefly, babA was amplified by PCR and ligated into pUC18 (pSUS2703). This plasmid was used as a template for an inverse PCR, designed to almost completely delete the babA gene. Finally, a chloramphenicol (cat) resistance cassette (41) was introduced. The resulting plasmid (pSUS2705) was used as a suicide plasmid for natural transformation of H. pylori strain J99. The successful chromosomal replacement of babA with the cat cassette via allelic exchange (double crossover) was verified by PCR using suitable primer combinations. PCR with a pUC18-specific primer was performed to exclude possible integration of the complete plasmid into the genome (single crossover).

bab genotyping and full-length sequencing of babA.

Genomic DNA from low-passage-number single colonies of bacterial strains was isolated using the QIAamp DNA minikit (Qiagen, Germany). PCRs were performed according to standard protocols, and amplification products were purified using the QIAquick PCR purification kit (Qiagen). PCR amplicons were sequenced bidirectionally by the Sanger method using the BigDye terminator v1.1 cycle sequencing kit and a 3130xl genetic analyzer (Applied Biosystems). Sequence data were analyzed using BioNumerics v6.01 (Applied Maths NV, Sint-Martens-Latem, Belgium). bab genotypes in loci A and B were determined using combinations of locus- and bab gene-specific primers as previously described (14). Analysis of locus C was performed by PCR amplification of the complete locus (16). Using BlastX search, the partial gene sequences were identified as either babA or babB, while no babC genes were found (42). Identified babA genes were completely sequenced using overlapping PCR fragments of locus-specific primers and bab gene-specific primers located in the middle variable region. See Table S2 in the supplemental material for a complete list of primer sequences. The identity of sequenced babA genes to their orthologue in strain 26695 based on BlastX search ranged between 85 and 92%. Pairwise sequence alignments were analyzed for the numbers of isolated single nucleotide polymorphisms (SNPs) and of clusters of nucleotide polymorphisms (CNPs), which comprise polymorphisms that are <200 bp apart from each other and are an indication of recombination.

Analysis of BabA and BabB expression in H. pylori strains.

H. pylori strains were grown on blood agar plates for up to 48 h. Bacterial cells were harvested in phosphate-buffered saline (PBS) by centrifugation (5,000 × g, 4°C, 10 min). Equal amounts of bacteria were resuspended in sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) sample buffer, incubated for 10 min at 95°C, separated by SDS-PAGE (8%), and analyzed by Western blotting. Primary antibodies against BabA and BabB raised in rabbits were kindly provided by Thomas Borén. Peroxidase-labeled AffiniPure goat anti-rabbit IgG (H+L) (Jackson ImmunoResearch Laboratories) was used as a secondary antibody. Western blots were developed with SuperSignal West Pico chemiluminescent substrate (Thermo Scientific, Rockford) and exposed to Amersham Hyperfilm ECL film (GE Healthcare, Little Chalfont, United Kingdom).

Analysis of Lewis b binding ability of H. pylori strains.

H. pylori strains were tested for their Le(b) binding ability using an enzyme-linked immunosorbent assay (ELISA) as previously described (43). Briefly, bacterial strains were grown on plates for 24 h and harvested in PBS by centrifugation (2,795 × g, 5 min, 4°C). For surface labeling, 2 × 108 live bacteria were incubated with 125 µg/25 µl NHS-LC-biotin [succinimidyl-6-(biotinamido)hexanoate] for 50 min at room temperature (RT) in the dark. Labeled bacteria were washed twice in PBS, collected by centrifugation (2,795 × g, 10 min, 4°C), and finally resuspended in 1 ml of PBS with 5% bovine serum albumin (BSA). The wells of a 96-well Universal Covalent microtiter plate (Corning Costar, Cambridge, MA) were coated in two replicate plates for each experimental condition with either 50 µl PBS, 250 ng BSA, or 250 ng Le(b)-BSA for 1 h at RT in the dark. After removal of supernatants, the plate was exposed to UV light for 30 s in a Stratalinker (Stratagene, Germany). The remaining binding sites on the plates were blocked by incubation with 5% BSA in PBS for 1 h at RT. Biotinylated bacteria (50 µl/well) were coincubated in the coated plates for 1 h at RT in the dark. The plates were washed three times with PBS, and adherent bacteria were fixed with 100 µl paraformaldehyde (2% in 100 mM potassium phosphate [pH 7]) for 1 h at RT. After three washing steps with 0.05% Tween 20 in PBS, the plate was blocked with 10% fetal calf serum (FCS) in PBS for 1 h at RT. The plate was washed five times with 0.05% Tween 20 in PBS and incubated with neutravidin-horseradish peroxidase (HRP)-conjugate in PBS plus 10% FCS for 90 min at RT to detect bound biotinylated bacteria. After seven washes with PBS containing 0.05% Tween 20, the plate was incubated with 100 µl/well 3,3′,5,5′-tetramethylbenzidine (TMB) (BD Biosciences) for 30 min at RT. The reaction was stopped by the addition of 1 M H3PO4 (50 µl/well), and the extinction at 450 nm was measured using a microplate reader. A Le(b)/BSA binding ratio of >1.5 was indicative of Le(b) binding. Le(b) binding of all strains in duplicate samples was determined in at least three independent experiments.

Generation of babA mosaic alleles.

Strain pair LSU1014-1/-6 was selected for generation of mosaic babA genes. The nonadherent strain LSU1014-6 was transformed with PCR products of babA from the adherent strain LSU1014-1 or from strain LSU1014-6 [restoration of intact C terminus with primer babA_1014-1_2186(5G)] by natural transformation or electroporation (see Table S3 in the supplemental material). Transformants were enriched for either Le(b) binding or BabA expression. Enrichment of strains with restored Le(b) binding was performed using a biopanning procedure as previously described (17). Briefly, transformed LSU1014-6 was mixed with biotinylated Le(b) conjugate and streptavidin-coated magnetic beads. Le(b) binding transformants were captured with a magnet and subsequently cultured on plates. BabA-expressing clones were enriched using a pulldown assay. Briefly, protein A magnetic beads were incubated with 10 µl anti-BabA antiserum for 1 h at 4°C. The beads were then incubated with 1.5 × 108 bacteria for 1 h at 4°C. BabA-expressing transformants were captured with a magnet and subsequently cultured. Recombinant clones were single colony purified and then analyzed for babA gene sequence, BabA expression, and Le(b) binding as described above.

Structure prediction by Phyre2.

Deduced BabA sequences of strain pair LSU1014-1/-6 and the generated babA mosaic alleles were subjected to the Phyre2 web server (44) for structure prediction using the intensive modeling mode. The structural model of the adhesion domain of BabA was based on similarities to the extracellular adhesion domain of SabA from H. pylori (25). The modeling of the β-barrel domain was based on similarities to different templates (c2k0lA, d1g90a, and c2x27X).

Nucleotide sequence accession numbers.

The babA sequences have been deposited in the European Nucleotide Archive (ENA) and are available from the ENA browser at http://www.ebi.ac.uk/ena/data/view/LN650055-LN650087.

BabA and BabB protein expression in sequential H. pylori isolates. Whole-cell extracts of H. pylori strains were analyzed for protein expression of BabA (A) and BabB (B), respectively. Strain J99, which expresses both BabA and BabB, was used as a positive control. The deletion mutant J99babA and the BabB-negative strain NCTC11637 were used as negative controls. Download

Figure S1, TIF file, 0.9 MB

Sequence alignment of BabA amino acid sequences of sequential H. pylori isolates. Multiple-sequence alignment of BabA sequences from Le(b) binding (n = 21) and nonbinding (n = 4; highlighted in gray) strains excluding duplicate sequences were performed using MUSCLE (45). The BabA sequence of the Le(b) binding H. pylori strain J99 was additionally included. The alignment was visualized by CLC Genomics Workbench. Amino acid residues are marked depending on the level of conservation ranging from 0% (blue) to 100% (red). The consensus sequence is based on the majority of sequences. Download

Figure S2, PDF file, 0.2 MB

BabA protein expression in strains with mosaic babA genes. Whole-cell extracts of H. pylori strains were analyzed for protein expression of BabA. Strain J99 was used as a positive control. The deletion mutant J99babA was used as a negative control. The band running slightly below BabA that is also detected in the babA mutant represents BabB (compare with Fig. S1 right panel, where BabB in J99babA is revealed by anti-BabB antiserum). Download

Figure S3, TIF file, 1.1 MB

H. pylori strains used in this study

Table S1, DOC file, 0.1 MB.

Primer sequences used in this study

Table S2, DOC file, 0.3 MB.

Generation of H. pylori strains with mosaic babA genes

Table S3, DOC file, 0.1 MB.