Phase Variation in H Type I and Lewis a Epitopes of Helicobacter pylori Lipopolysaccharide

Abstract

Helicobacter pylori NCTC 11637 lipopolysaccharide (LPS) expresses the human blood group antigens Lewis x (Le^x), Le^y, and H type I. In this report, we demonstrate that the H type I epitope displays high-frequency phase variation. One variant expressed Le^x and Le^y and no H type I as determined by serology; this switch was reversible. Insertional mutagenesis in NCTC 11637 of JHP563 (a poly(C) tract containing an open reading frame homologous to glycosyltransferases) yielded a transformant with a serotype similar to the phase variant. Structural analysis of the NCTC 11637 LPS confirmed the loss of the H type I epitope. Sequencing of JHP563 in strains NCTC 11637, an H type I-negative variant, and an H type I-positive switchback variant showed a C14 (gene on), C13 (gene off), and C14 tract, respectively. Inactivation of strain G27, which expresses Le^x, Le^y, H type I, and Le^a, yielded a transformant that expressed Le^x and Le^y. We conclude that JHP563 encodes a β3-galactosyltransferase involved in the biosynthesis of H type I and Le^a and that phase variation in H type I is due to C-tract changes in this gene. A second H type I-negative variant (variant 3a) expressed Le^x and Le^a and had lost both H type I and Le^y expression. Inactivation of HP093-HP094 resulted in a transformant expressing Le^x and lacking Le^y and H type I. Structural analysis of a mutant LPS confirmed the serological data. We conclude that the HP093-HP094 α2-fucosyltransferase (α2-FucT) gene product is involved in the biosynthesis of both Le^y and Le^x. Finally, we inactivated HP0379 in strain 3a. The transformant had lost both Le^x and Le^a expression, which demonstrates that the HP0379 gene product is both an α3- and an α4-FucT. Our data provide understanding at the molecular level of how H. pylori is able to diversify in the host, a requirement likely essential for successful colonization and transmission.

The human gastric pathogen Helicobacter pylori displays molecular mimicry with the gastric epithelial cells of the human host (3). Often, H. pylori lipopolysaccharide (LPS) expresses both Lewis x (Le^x) (see Fig. 1) and Le^y human blood group antigens, but strains expressing sialyl-Le^x, H type I, Le^a, Le^b, and the nonfucosylated polylactosamine chain (i-antigen) have been described (16, 17); strains expressing H type 2 have not been found yet. For biosynthesis of Le^x/y antigens, the activity of a variety of glycosyltransferases is required: α2- and α3-fucosyltransferases (α2- and α3-FucT), β4-galactosyltransferases (β4-GalT) and β3-N-acetylglucosaminyltransferases (β3-GlcNAcT). Two α3-fucT genes, homologues of HP0379 and HP0651 of H. pylori strain 26695, have been identified (13, 15). The products of these genes have different fine specificities (5). For this reason, the homologues of HP0379 and HP0651 are designated futA and futB, respectively; α2-fucT (the homologue of gene HP093-HP094 in strain 26695) is designated futC (21, 22). A β4-galT gene has recently been identified in H. pylori (the homologue of HP0826 in strain 26695 [14]). It is unknown which gene codes for β3-GalT and β3-GlcNAcT. In vitro, the recombinant futC gene product is able to form H type I from a synthetic Galβ1.3GlcNAc acceptor (22), but formal proof of the involvement of this gene in biosynthesis of H type I in H. pylori LPS is lacking.

FIG. 1.

Structure of Lewis blood group antigens and glycosyltransferases required for biosynthesis. Gal, d-galactose; Fuc, l-fucose; GlcNAc, N-acetyl-d-glucosamine.

A striking feature of H. pylori Lewis antigens is their ability to phase vary (4, 5). Phase variation is defined as the high-frequency, reversible switching of phenotype. For instance, a strain expressing Le^x may yield phase variants expressing Le^y or the i-antigen. Phase variation in other bacteria like Haemophilus influenzae and Neisseria spp. has been shown to contribute to bacterial virulence and host adaptation (13). We have shown that populations of H. pylori LPS phase variants can be isolated from the human stomach, which provides evidence that LPS phase variation contributes to strain diversity in the human host (6). The molecular mechanism of phase variation in Le^x/y has been investigated recently (5). Long homopolymeric C tracts present in the open reading frames of futA, futB, and futC may change length during replication due to DNA slippage (“slipped-strand mispairing”). The result is a reversible (translational) frame switch that leads to either a full-length active gene product (fucosyltransferase enzyme) or an inactive truncated form.

Phase variation in H type I or Le^a has not been documented yet. In this paper, we show that the H type I and Le^a epitopes are also phase variable, and we identify a hitherto unrecognized gene (β3-galT) required for biosynthesis of these epitopes. We also demonstrate that phase variation in this gene is due to length changes in a C tract. We also show that the futA gene product is capable of acting as an α4-FucT, which is required for Le^a biosynthesis, and that the futC gene product is essential for biosynthesis of H type I.

MATERIALS AND METHODS

Bacterial strains and LPSs.

Strains NCTC 11637, G27, and phase variant 3a have been described before (4, 8, 11). The LPSs of strains J223 (expresses H type I [16]) and UA948 (expresses Le^a [16]) were purified as described previously (16).

MAbs, synthetic glycoconjugates, and serological procedures.

The monoclonal antibodies (MAbs) used in this study and their specificities are shown in Table 1. Determination of the epitope fine specificity of MAbs 4D2 and 218/2B3 was done by enzyme-linked immunosorbent assay (ELISA) by testing their reactivity with synthetic glycoconjugates as described before (2) (Table 2). Two types of conjugates were used, linked to polyacrylamide (PAA; Syntesome, Moscow, Russia) or linked to protein, i.e., either human serum albumin (HSA), bovine serum albumin, or keyhole limpet hemocyanin (KLH) (all three prepared at Isosep, Tullinge, Sweden). LPS phase variants were isolated by colony blot and serotyped by ELISA as previously described (4). The frequency of phase variation is defined as the percentage of colonies expressing a given variant serotype (4).

TABLE 1.

MAbs used in this study

MAb	Specificity	Isotypea	Source (reference)
4D2	H type I	IgM	R. Negrinib (2)
54.1F6A	Lex	IgM	G. J. van Damc (20)
Hp151	Ley	IgG	R. Negrini (2)
NAM61-1A2	i-Antigen	IgM	D. Blanchardd (10)
218/2B3	Lea, Leb	IgM	Isosepe

^aIg, immunoglobulin. 

^bLaboratory Unit, City Hospital, Brescia, Italy. 

^cDepartment of Parasitology, University of Leiden, Leiden, The Netherlands. 

^dRegional Blood Transfusion Service, Nantes, France. 

^eTullinge, Sweden; the supplier documents this MAb as specific for Le^a, while our own data (Table 2) indicate an additional specificity for Le^b and LNT. 

TABLE 2.

Epitope specificities of MAbs 4D2 and 218/2B3^a

Antigen	OD of MAb tested
	4D2	318/2B3
Ley-HSA	—	—
Lex-HSA	—	—
Lea-HSA	—	>2.5
Leb-HSA	—	—
H1-HSA	1.6	—
H11-BSA	—	—
Lea-KLH	—	>2.5
Leb-KLH	—	—
H1-KLH	>2.5	—
LNT-KLH	—	>2.5
Ley-PAA	—	—
Lex-PAA	—	—
Lea-PAA	—	1.3
Leb-PAA	—	1.5
H1-PAA	—	—
H11-PAA	—	—
Galβ1.3GlcNAc	ND	—
Galβ1.4GlcNAc	ND	—
Fucα1.2Gal	ND	—
Fucα1.3GlcNAc	ND	—
Fucα1.4GlcNAc	ND	—

^aReactivity of MAbs tested for by ELISA with synthetic glycoconjugates expressed as OD₄₉₂. ND, not done; BSA, bovine serum albumin; —, OD₄₉₂ < 0.3. 

Insertional mutagenesis of glycosyltransferase genes.

Plasmids for inactivation of glycosyltransferase genes were constructed as follows. A DNA fragment containing futA (the homologue of HP0379 of strain 26695) was amplified from H. pylori 26695 genomic DNA by PCR with primers SM119704 (TTCTAAAGTGGATCCTGAAAT) and SM119706 (GAGTGGGCGAAAGAGAGATTG), positioned approximately 1 kb from the 5′ and 3′ ends of HP0379, respectively. The construction of plasmid pHP0379::Km^r, carrying the futA gene interrupted with a kanamycin cassette, has been described before (5, 15). This plasmid was used to inactivate futA in strain 3a by natural transformation.

A plasmid for inactivation of futC (the homologue of HP093-HP094 of strain 26695) was constructed as follows. A DNA fragment containing futC was amplified with primers FTα1-2F (TCTAATACGCCTGTGCTGTT) and FTα1-2R (CCAATACGCCTCTTCTTCTT) and ligated into pGEM-T, cut out with SacII-SpeI, and ligated into SacII-SpeI-opened plasmid pBCα3, carrying a kanamycin cassette (9). This plasmid was used to inactivate futC in strain NCTC 11637 by natural transformation.

A plasmid for the inactivation of β3-galT (the homologue of JHP563 in strain J99 and HP619 in strain 26695, respectively) was constructed as follows. Primers J563-F1 (CGGGGTACCAGCTCGTTTCAGAATCCAATGAG), J563-F2 (TCGACTGCAGCCAAAGAACTACCTGAGTCTTGC), J563-R1 (TCGACTGCAGTCTTGTTGCGTTTCTTGTGTGGG), and J563-R2 (ATTTGCGGCCGCTTAAAGGAGCGTATCGTCTGCTG), containing 5′ restriction sites for KpnI, PstI, PstI, and NotI, respectively, were used to amplify two DNA fragments, CM1 and CM2, of the β3-galT gene (calculated length, 540 and 627 bp, respectively) that were cloned separately into plasmid pSK to yield pCM1 and pCM2; these were digested with KpnI-PstI and PstI-NotI, respectively. The gel-purified fragments were cloned together into linearized pSK containing compatible ends (KpnI and NotI) to yield pCM12. A PstI-restricted kanamycin cassette was cloned into the unique PstI site of pCM12. The resulting plasmid was used to mutate β3-galT in strains NCTC 11637 and G27; the H. pylori transformants were selected by kanamycin resistance.

PCR analysis of transformants.

H. pylori transformants were analyzed by PCR to confirm that the recombination had occurred at the intended location. For each of the three inactivated glycosyltransferase genes, one primer was located outside the DNA fragment cloned and the other primer was located in the kanamycin cassette. For futA in strain 3a, KO379 primer N-379 (AGCAGCCCCAATAAAGAAAT, located upstream from SM119704 in HP0378) was used in combination with KanaR-Gl (TTTAGACATCTAAATCTAGG); the calculated product length is 3,097 bp. For futC, primer N2-93/94 (GAACGCTTGCTAGAACACTC, located upstream from FTα1.2F in HP093-HP094) was used in combination with KanInvF (TTACCTATCACCTCAAATGG); the calculated product length is 873 bp. For β3-galT, primer N-619 (AAGTCAAAGGCGTTTGGATA, located upstream from J563F-1 in JHP563) was used in combination with KanInvF; the calculated product length is 601 bp.

DNA sequencing.

C-tract sequencing of α3-fucT genes HP0379 and HP0651 was performed as previously described (5). C-tract sequencing of JHP563 was carried out on chromosomal DNA purified by cesium chloride centrifugation and on gene fragments amplified by PCR with primers F1 and R2 (1,215 bp). The sequencing of both the PCR products containing the C tract and of genomic DNA was performed with primer CM4F (ATGCAAGCCTTAGAAGATTGCTTG); likewise, G tracts in the opposite strand were sequenced with primer CM3R (TCCTACGATATTCTTATACACTTC). Sequencing of both C and G tracts was carried out thrice: twice with template DNA from separate PCR amplifications and once with chromosomal DNA.

Structural analysis of LPS.

LPSs of NCTC 11637-derived mutants in futC and β3-galT were methylated and subjected to fast atom bombardment-mass spectrometry (FAB-MS) as previously described (5).

RESULTS AND DISCUSSION

Characterization of MAbs 4D2 (anti-H type I) and 218/2B3 (anti-Le^a).

MAb 4D2 reacted with H type I linked to KLH or HSA but not with H type I linked to PAA; it did not react with any of the other glycoconjugates tested. In addition, this MAb reacted strongly (optical density at 492 nm [OD₄₉₂] > 2.5) with LPS of strain J223, which by structural analysis has been shown to contain H type I (16). MAb 218/2B3 reacted with HSA- and KLH-linked Le^a, with the Galβ1-3GlcNAcβ-1-3Galβ1-4Glc-containing blood group related neoglycoprotein lacto-N-tetraose (LNT)-KLH, and with PAA-linked Le^a and Le^b. In addition, this MAb reacted strongly (OD₄₉₂ > 2.5) with LPS of strain UA948, which has been shown by structural analysis to contain Le^a (16). We concluded (Table 2) that MAb 4D2 reacts exclusively with H type I and that serotyping results obtained with 218/2B3 need to be interpreted with care.

Characterization of H type I-negative phase variants and identification of genes involved in H type I biosynthesis.

We first isolated H type I-negative variants of NCTC 11637. After probing colony blots of strain NCTC 11637 with anti-H type I MAb 4D2, H type I-negative colonies were isolated and serotyped by ELISA (Table 3). The H type I-negative variant 3a has been described earlier (4). Three types of variants were obtained: variant H1, which was H type I negative, i positive, and Le^x/y positive (frequency of phase variation, 0.45%); variant H12, which had a strongly decreased H type I expression, was Le^y negative, and was Le^x positive (frequency of phase variation, 0.6%); and strain H13, which was Le^y positive and Le^x negative (frequency of phase variation, 0.3%). Strain 3a (frequency of phase variation, 0.3%) represented a fourth type of H type I-negative variant; this strain was Le^y negative, had a strongly decreased H type I expression, expressed Le^x, and in addition was Le^a/b positive.

TABLE 3.

Serological characterization of H. pylori strains, phase variants, and knockouts^a

Strain	MAbs tested (specificity)b
	54.1F6a (Lex)	NAM61-1A2 (i-antigen)	Hp151 (Ley)	4D2 (H type I)	218/2B3 (Lea/b)
NCTC 11637	2.5	—	1.4	2.3	—
H11	>	0.9	>	—	—
H12	>	—	—	0.5	—
H13	—	—	2.1	—	—
3a	>	—	—	1.4c	1.4
H111	>	—	1.1	>	—
NCTC 11637 KO563	>	1.2	>d	—	—
G27	>	>	>	2.2	0.4
G27 KO563	>	>	>	—	—
LeA	>	—	—	1.1	1.3
3a1	>	—	2.0	>	—
NCTC 11637 KO93-KO94	>	—	—	—	>
3a KO379	—	1.1	—	1.6	—

^aReactivities of MAbs were determined by ELISA and are expressed as OD₄₉₂. 

^b—, OD₄₉₂ < 0.3; >, OD₄₉₂ > 2.5. 

^cTitration of MAb 4D2 demonstrated an eightfold-higher reactivity with NCTC 11637 than with 3a. 

^dTitration of MAb Hp151 showed a 32-fold-higher reactivity with NCTC KO563 than with NCTC 11637. 

We first investigated whether the loss of H type I expression was reversible: H11 (H type I negative) was colony blotted with the anti-H type I MAb 4D2, and a 4D2-positive colony (named variant H111) was isolated and serotyped; the LPS serotype of H111 (frequency of phase variation, 1.1%) was identical to that of NCTC 11637 (Table 3). Thus, strain NCTC 11637 was able to phase vary reversibly in the H type I epitope without a simultaneous loss of Le^y expression. Based on the structures shown in Fig. 1, we hypothesized that phase variation in a β3-galT gene would explain the reversible switch in phenotype observed: when this gene was off, no H type I can be formed, with synthesis of Le^x and Le^y still being possible. As yet, no β3-galT gene has been annotated for either strain 26695 or J99; however, several other genes with sequence homology to glycosyltransferases have been identified (1, 19). We investigated one of these genes, JHP563 from strain J99 (the homologue of gene HP0619 in strain 26695), because it has sequence similarity to lic2b, a gene necessary for LPS biosynthesis in H. influenzae; in addition, JHP563 and HP0619 contain homopolymeric C tracts, signatures of phase variation. We inactivated the JHP563 homologue of NCTC 11637 with conventional recombinant techniques; the knockout mutant NCTC 11637 KO563 indeed lacked the H type I epitope and had an increase in Le^y expression (Table 3); inactivation of the gene in another strain (G27) also yielded a mutant that lacked H type I (Table 3). PCR analysis (observed product length of 600 bp versus calculated length of 600 bp) demonstrated that the mutation indeed had taken place in a homologue of JHP563. Of interest, G27 also expresses Le^a, while the isogenic JHP563 knockout lacked this antigen; this suggests that the JHP563 homologue plays a role in the synthesis of both H type I and Le^a.

The LPS structure of NCTC KO563 was compared with that of its parent NCTC 11637. Recently, NCTC 11637 was shown to express, in addition to Le^x (8), two other Lewis determinants (17), Le^y [m/z 812→606(812-206)] and H type 1 [638→228(638-410)]. The FAB-MS of the intact LPS of methylated NCTC 11637 KO563 yielded strong ions at m/z 189 for terminal Fuc, m/z 812→606 for Le^y, m/z 1435→1229 for Le^y→Le^x, m/z 682 for Fuc-(1→3)-GlcNAc-(1→7)-DDHep, m/z 886 for Fuc-(1→3)-GlcNAc-(1→7)-[Glc-1→2]-DDHep, and m/z 1090 for Fuc-(1→3)-GlcNAc-(1→7)-[Glc-1→6-Glc-1→2]-DDHep. No ions indicative of H type 1, 638→228, were observed in this FAB-MS. These data confirmed the loss of the H type I epitope and the increased expression of Le^y that were also seen in serological examinations (Table 3).

Our findings demonstrate that JHP563 determines the biosynthesis of the H type I epitope and Le^a and suggest that JHP563 codes for a β3-GalT. Evidence for a role of this gene in phase variation in the H type I epitope was obtained by sequencing this β3-galT in strains NCTC 11637, H11, and H111. Poly(C) tract lengths were as follows: C14 for NCTC 11637, C13 for H11, and C14 for H111. No other differences in DNA sequence of β3-galT in these strains were observed (data not shown). A β3-galT with a C₁₄ tract yields a full-length polypeptide, while a C₁₃ tract yields a truncated product; this is in agreement with the serological data that show that the NCTC 11637 and H111 express H type I, while this epitope is absent in strain H11. These data confirm the role of JHP563 homologues in H type I phase variation and confirm that they encode a β3-GalT.

The second class of H type I-negative variants we isolated is exemplified by variant H13, which expresses Le^y only. We described variants of the H13 phenotype in a previous paper; this phenotype is due to phase variation in β3-GlcNAcT (4).

The third class of H type I-negative variants was exemplified by the two similar yet distinctive strains, H12 and 3a (Table 3). Both were Le^x positive, with a decreased expression of H type I and Le^y as compared to NCTC 11637; in addition, 3a expressed Le^a/b. A colony blot of strain NCTC 11637 probed with MAb 218/2B3 (anti-Le^a/b) yielded a variant (strain LeA, frequency of variation, 0.6%) with a serotype indistinguishable from that of strain 3a (Table 3). This simultaneous loss of H type I and Le^y in strain 3a was reversible. An H type I-positive switchback variant (strain 3a1; frequency of variation, 1.0%) with a serotype similar to that of strain NCTC 11637 was isolated from strain 3a by colony blotting after probing with MAb 4D2 (Table 3). The simultaneous loss of H type I and Le^y can be explained by phase variation in futC, the gene encoding α2-FucT (encoded by a homologue of HP093-HP094 of strain 26695). Inactivation of futC in strain NCTC 11637 yielded a mutant (NCTC KO93-KO94) with an LPS phenotype similar to that of strain 3a (Table 3); PCR analysis of NCTC 11637 KO93-KO94 (observed product length of 850 bp versus calculated length of 873 bp) demonstrated that the mutation had indeed taken place in futC. We believe that the phenotype of the futC knockout is due to inactivation of futC and not to polar effects, because in both strains 26695 and J99, futC is not in an operon and the genes downstream have an orientation opposite to that of futC. In addition, the genes downstream code for a restriction-modification system, not likely to influence LPS phenotype; finally, the observed phenotype is as expected, since earlier in vitro studies had shown that FutC is able to fucosylate Galβ1.3GlcNAc to form H type I (22). The reaction of this knockout mutant with MAb 218/2B3 indicates that it expressed Le^a and not Le^b, because synthesis of Le^b requires an active α2-FucT, while this enzyme is not required for Le^a biosynthesis (Fig. 1). The LPS structure of NCTC KO93-KO94 was compared with that of its parent, NCTC 11637. The FAB-MS of the methylated intact NCTC 11637 KO93-KO94 indicated the presence of the sole blood group antigen, Le^x [638→432(638-206)], along with m/z 682 for Fuc-(1→3)-GlcNAc-(1→7)-DDHep. These data demonstrate that Le^y and H type I, both present in parent strain NCTC 11637 (17), had been lost upon mutagenesis of futC. These findings demonstrated the role of the HP093-HP094 homologue in biosynthesis of both Le^y and H type I and extended findings that purified, recombinant HP093-HP094 is able to α2-fucosylate the synthetic acceptor Galβ1.3GlcNAc to yield H type I (22). The molecular basis of phase variation in futC has already been elucidated (21) and is due to length changes in C tracts, as well as to ribosomal slippage.

Finally, we investigated which gene encodes the α4-FucT that is required for biosynthesis of the Le^a epitope. Previously, we have shown that two α3-FucTs are involved in biosynthesis of Le^x (5); the genes encoding the two enzymes are futA and futB. The gene futB is off in NCTC 11637 (5), and sequencing data showed that this is also the case in strain 3a; both strains have a C₉ tract which leads to biosynthesis of a truncated polypeptide. In contrast, futA is on in both strains (a C₁₀ tract). We tested the hypothesis that FutA is also involved in Le^a biosynthesis and able to transfer fucose to the C₄ position of GlcNAc. PCR analysis of the futA knockout mutant of strain 3a (3a KO379) demonstrated that the mutation had taken place at the intended location (observed product length of 3.2 kb versus calculated length of 3,097 bp). As determined by serology, 3a KO379 indeed lacked Le^a and expressed only the nonfucosylated polylactosamine (i-antigen) and H type I. This demonstrated that FutA is both an α3- and an α4-FucT. These findings extend the recently published data of Rasko et al. (18) that show that also FutB is both an α3- and α4-FucT.

In summary, we have identified the β3-galT gene that is essential for biosynthesis of H type I and Le^a and demonstrated that this gene phase varies through DNA slippage in a poly(C) tract. In addition, we have shown that futC and futA are involved not only in biosynthesis of Le^x/y but also in that of H type I and Le^a. These findings complement earlier studies of phase variation in Le^x and Le^y (4, 5, 21) and show that H. pylori has a great potential to diversify. The biological role of H. pylori Lewis antigens remains largely unknown. Recent data, however, show that Lewis antigens might be adhesins (12); phase variation in LPS might allow attachment and detachment of bacteria by this on-off expression of Lewis antigens and hence facilitate transmission and colonization (7).