Novel Binding Epitope for Helicobacter pylori Found in Neolacto Carbohydrate Chains STRUCTURE AND CROSS-BINDING PROPERTIES*

Helicobacter pylori is a bacterium that colonizes the stomach of a majority of the global human population causing common gastric diseases like ulcers and cancer. It has an unusually complex pattern of binding to various host glycoconjugates including interaction with sialylated, sulfated, and fucosylated sequences. The present study describes an additional binding epitope comprising the neolacto internal sequence of GlcNAc (cid:1) 3Gal (cid:1) 4GlcNAc (cid:1) . The binding was detected on TLC plates as an interaction with a seven-sugar ganglioside of rabbit thymus. The glycolipid was purified and characterized as Neu5Gc (cid:2) 3Gal (cid:1) 4GlcNAc (cid:1) 3Gal (cid:1) 4GlcNAc (cid:1) 3Gal (cid:1) 4Glc (cid:1) 1Cer with less than 10% of the fraction carrying a repeated lacto (type-1) core chain, Gal (cid:1) 3GlcNAc (cid:1) 3Gal (cid:1) 3GlcNAc (cid:1) . After stepwise chemical and enzymatic degradation and structural analysis of products the strongest binder was found to be the pentaglycosylceramide GlcNAc (cid:1) 3Gal (cid:1) H. pylori cells grown in liquid cultures. This growth environment elim- inates binding to simple sialylated glycolipids (29), thus providing bet-ter conditions for investigation of the neolacto epitope. Other Analytical Methods— Hexose content in glycolipids was deter-mined according to Dubois et al. (30).

Helicobacter pylori colonizes the stomach of a majority of the global human population and is implicated in several diseases of the gastrointestinal tract including chronic gastritis, duodenal and gastric ulcers, and gastric adenocarcinoma (1,2). In 1994 this pathogen was classified by the World Health Organization as a class I carcinogen (3). Like many other microbes, H. pylori recognizes carbohydrates, probably mediating essen-tial attachment to host cells (4,5). However, H. pylori is unusually complex in its binding to carbohydrates as shown by interaction with sialylated oligosaccharides (6), gangliotetraosylceramide (7), Lewis b antigen (8), monohexosylceramide (9), lactosylceramide (10), lactotetraosylceramide (11), sulfatide (12), and heparan sulfate (13). The inhibition of binding of H. pylori to gastric cells by glycoconjugates (14) and free oligosaccharides (15) has been reported, and sialyllactose was shown to eradicate bacteria or decrease bacterial density in rhesus monkeys experimentally infected with H. pylori (16). Two bacterial adhesins, which are outer membrane proteins, have so far been identified, BabA, which recognizes Lewis b antigen (17), and SabA, which binds to sialylated glycoconjugates (18).
The present study describes an additional carbohydrate epitope recognized by H. pylori. The activity was first detected as a binding to a seven-sugar ganglioside in a mixture of rabbit thymus gangliosides separated on TLC plates. However, the interaction was independent of sialic acid because activity remained after treatment with mild acid. Therefore the ganglioside was purified and structurally characterized in detail followed by partial degradation studies that showed a pentaglycosylceramide, GlcNAc␤3Gal␤4GlcNAc␤3Gal␤4Glc␤1Cer, 1,2 to be the most active among the products. On screening for binding using a reference library of glycolipids three other pentaglycosylceramides differing in terminal monosaccharide substitution were also found to be bound by H. pylori.

EXPERIMENTAL PROCEDURES
Isolation and Purification of Glycolipids from Rabbit Thymus-Total mixtures of neutral and acid glycosphingolipids were isolated from acetone powder of 1 kg of rabbit thymus (Pel-Freez, North Arkansas, AR). The powder was extracted in a Soxhlet apparatus with C/M (2:1) 3 for 24 h followed by C/M/H 2 O (8:1:1) for 36 h. Dry residue of the combined extracts (240 g) was subjected to Folch separation (19), and the material in the upper hydrophilic phase was fractionated by ionexchange gel chromatography on DE23 cellulose (DEAE, Whatman). These isolation steps gave 2.5 g of the crude fraction containing acid glycosphingolipids. The gangliosides were then separated according to number of sialic acids using open tubular chromatography on a glass column filled with DEAE-Sepharose (CL6, Amersham Biosciences). The column was connected to an HPLC pump producing a concave gradient (preprogrammed gradient number 4, System Gold chromatographic software, Beckman) starting with methanol and ending with 0.5 M * This work was supported by Swedish Research Council Grant 06X-12628, the Swedish Cancer Foundation, Wilhelm and Martina Lundgrens Research Foundation, the Swedish Medical Society, the Adlerbertska Research Foundation, Ingabritt and Arne Lundberg Foundation, and Symbicom Ltd. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ These authors contributed equally to this work. § To whom correspondence should be addressed. CH 3 COONH 4 in methanol. About 350 mg of the ganglioside fraction were applied for each separation using 500 g of Sepharose (50-mminner diameter column; bed height, 130 mmm) and a flow rate of 4 ml/min. The monosialylated gangliosides (218 mg) were further separated by HPLC on a silica column (SH-044.10, Yamamura Chemical Laboratories Co. Ltd., Kyoto, Japan) using a linear eluting gradient of C/M/H 2 O (60:35:8 to 10:10:3) with a flow rate of 4 ml/min. The final yield of the seven-sugar ganglioside fraction was 23 mg/1000 g of the acetone powder. Other glycosphingolipids used in these studies were prepared according to Karlsson (20) unless otherwise specified.
Chemical Desialylation of Gangliosides-Chemical desialylation was performed in 1.5% CH 3 COOH in water at 100°C after which the material was neutralized with NaOH and evaporated under nitrogen after addition of excess of methanol.
De-N-acylation-Conversion of the acetamido moiety of GlcNAc residues into primary amine was accomplished by treating various glycosphingolipids with anhydrous hydrazine as described previously (10).
Partial Acid Degradation of Core Chain-For partial degradation of the core carbohydrate chain, the seven-sugar ganglioside was hydrolyzed in 0.5 M HCl for 7 min in a boiling water bath. The material was then neutralized with NaOH and partitioned in C/M/H 2 O (8:4:3) (19). The lower phase was collected and evaporated under nitrogen, and the residue was used for analysis.
Endoglycoceramidase Digestion of Glycolipids-The reaction mixture contained 200 g of a glycolipid, 80 g of sodium taurodeoxycholate, and 0.8 milliunit of the enzyme (endoglycoceramidase from Rhodococcus species, Genzyme Corp.) in 160 l of 50 mM acetate buffer, pH 6.0 (21). The sample was incubated overnight at 37°C after which water (140 l) and C/M (2:1, 1.5 ml) were added. Subsequently the sample was shaken and centrifuged. The upper phase was dried under nitrogen, redissolved in a small volume of water, and desalted on a Sephadex G-25 column (0.4 ϫ 10 cm), which had been equilibrated in H 2 O, and eluted with the same solvent. Fractions of about 0.1 ml were collected and tested for the presence of sugars (using TLC plates and anisaldehyde).
Preparation of Pentaglycosylceramide Using ␤-Galactosidase from Escherichia coli-The six-sugar glycolipid fraction (2 mg), obtained from the seven-sugar ganglioside of rabbit thymus, was dissolved in 1.5 ml of 0.1 M potassium phosphate buffer, pH 7.2, containing sodium taurodeoxycholate (1.5 mg/ml), MgCl 2 (1 mM), and E. coli ␤-galactosidase (Roche Applied Science; 500 units when assayed with 2-nitrophenyl-␤-D-galactoside as substrate) and incubated overnight at 37°C. The material was next partitioned in C/M/H 2 O (10:5:3), and the glycolipid in the lower phase was purified using silica gel chromatography (0.4 ϫ 5-cm column). The column was packed in C/M (2:1) and eluted with C/M/H 2 O (60: 35:8). Fractions of about 0.2 ml were collected and tested for the presence of carbohydrates. The sugar-containing eluate was evaporated under nitrogen. To remove all contaminating detergent the chromatography was repeated twice. The final recovery of the pentaglycosylceramide fraction was 0.7 mg.
Analytical Enzymatic Tests-Oxford GlycoSystems (Oxford, UK) enzymatic tests were performed according to the manufacturer's recommendations except that Triton X-100 was added to each incubation mixture to a final concentration of 0.3%. When a mixture of sialidase (Arthrobacter ureafaciens) and ␤4-galactosidase (Streptococcus pneumoniae) was added for digestion, buffer from the ␤-galactosidase kit was used. If ␤-N-acetylhexosaminidase (S. pneumoniae) was present in the digestion mixture the buffer from the ␤-hexosaminidase enzyme kit was used. The enzyme concentrations in the incubation mixtures were 80 milliunits/ml for ␤-galactosidase, 120 milliunits/ml for ␤-hexosaminidase, and 1 unit/ml for sialidase. The substrate concentration was about 20 M. Enzymatic hydrolysis was performed overnight at 37°C. After digestion the samples were dried and desalted using small columns containing 0.3 g of Sephadex G-25 (22)  Fast Atom Bombardment Mass Spectrometry (FAB-MS)-FAB-MS analyses were performed on a Jeol SX-102 mass spectrometer (Jeol, Tokyo, Japan). Negative FAB spectra were recorded in the negative ion mode at 10-kV accelerating voltage using xenon atom bombardment (6 keV) and triethanolamine as matrix.
Methylation of Saccharides-Methylation was performed according to Larson et al. (23) except that sodium hydroxide was added to samples before methyl iodide as suggested by Needs and Selvendran (24). In some experiments the saccharides were reduced with NaBH 4 before methylation. In the latter case the amount of methyl iodide was increased to a final Me 2 SO/methyl iodide proportion of 1:1 (25).
Analysis of Methylated Oligosaccharides-Gas chromatography of methylated oligosaccharides was carried out on a Hewlett-Packard 5890A Series II gas chromatograph equipped with an on-column injector and a flame ionization detector. The oligosaccharides were analyzed on a fused silica capillary column (11 m ϫ 0.25-mm inner diameter) coated with cross-linked PS264 (Fluka; film thickness, 0.03 m). Samples were dissolved in ethyl acetate and injected on-column at 80°C. The temperature was programmed from 80 to 390°C at a rate of 10°C/min with a 2-min hold at the upper temperature. Gas chromatography (GC)/electron ionization MS of the methylated oligosaccharides was as described in detail previously (25). The analysis was performed on a Hewlett-Packard 5890A Series II gas chromatograph interfaced to a Jeol SX-102 mass spectrometer.
Proton NMR Spectroscopy-All samples for 1 H NMR spectroscopy were first deuterium-exchanged and then dissolved in 0.5 ml of Me 2 SO/ D 2 O (98:2). The spectrum of the seven-sugar ganglioside of rabbit thymus was recorded at 400 MHz on a modified Varian XL400 spectrometer, whereas spectra of the five-and six-sugar glycolipids of rabbit thymus origin were recorded at 500 MHz on a Jeol Alpha 500 spectrometer. All spectra were obtained at 30°C and were referenced via the internal solvent signal.
Molecular Modeling-Minimum energy conformers of glycosphingolipids were calculated within the Quanta2000/CHARMm25 molecular modeling package (Accelrys Inc.) on a Silicon Graphics Indigo 2 extreme work station as described previously for toxin A from Clostridium difficile (26).
TLC Bacterial Overlay Assay-Thin-layer chromatography was performed on glass-or aluminum-backed silica gel 60 HPTLC plates (Merck) essentially as described by Karlsson and Strömberg (27). TLC plates with separated glycolipids were treated for 1 min with 0.3% (weight by volume) polyisobutylmethacrylate Plexigum (Röhm GmbH, Darmstadt, Germany) in diethyl ether/hexane (3:1) and blocked for 2 h by incubation in 2% (w/v) bovine serum albumin in phosphate-buffered saline containing 0.1% (weight by volume) Tween 20. The plates were then overlaid with a suspension of radiolabeled bacteria (diluted in phosphate-buffered saline to 1 ϫ 10 8 colony-forming units/ml and 1-5 ϫ 10 6 cpm/ml) and incubated at room temperature for 2 h. Finally the plates were washed three to four times in phosphate-buffered saline, dried, and exposed to XAR-5 x-ray films (Eastman Kodak Co.) for 12-72 h.
Bacterial Cultivation and Labeling-The strains were stored at Ϫ80°C in tryptic soy broth containing 15% glycerol (by volume). The bacteria were initially cultured on gonococcal agar base-Campylobacter medium (28) agar under humid (98%) microaerophilic conditions (5-7% O 2 , 8 -10% CO 2 , and 83-87% N 2 ) for 48 -72 h at 37°C. For growth in liquid medium the cells were transferred to Ham's F-12 nutrient mixture supplemented with 10% fetal calf serum (SERA-Lab) using an inoculum of 1 ϫ 10 5 colony-forming units/ml. [ 35 S]Methionine was then added to the inoculated medium (50 Ci/10 ml), and the bacteria were allowed to grow for 24 h under mild shaking. The bacterial cells were harvested by centrifugation (3500 ϫ g for 5 min), washed twice in phosphate-buffered saline (pH 7.3), resuspended in the same buffer (1 ϫ 10 8 cells/ml), and used for overlay assay. The bacterial motility and purity were checked by phase-contrast microscopy. For cultivation of H. pylori on agar medium, a semisolid Brucella agar (Difco) supplemented with 10% heat-inactivated fetal calf serum and enriched with 0.5% IsoVitaleX was used. The agar medium was inoculated by streaking with bacterial material from the agar plate, and the new plate was sprinkled with [ 35 S]methionine (100 Ci/plate). The incubation conditions and preparation of bacterial cells were done as described above.
Most of the experiments of the present work were performed on H. pylori cells grown in liquid cultures. This growth environment eliminates binding to simple sialylated glycolipids (29), thus providing better conditions for investigation of the neolacto epitope.
Other Analytical Methods-Hexose content in glycolipids was determined according to Dubois et al. (30).

Screening for H. pylori Carbohydrate Binding-To expose
H. pylori to a larger number of potentially binding-active carbohydrate structures, mixtures of glycolipids isolated from various species and tissues were utilized in a chromatogram binding assay. The binding pattern shown in Fig. 1 was thereby frequently obtained. Thus, although most compounds were not recognized by the bacteria, a selective binding to some glycosphingolipids was detected. The binding in lanes 1 and 3 corresponds to the previously reported recognition of gangliotetraosylceramide (7) and lactotetraosylceramide (11), respectively. In addition, H. pylori bound to a slow migrating compound (arrow) in the acid glycosphingolipid fraction of rabbit thymus (lane 5). This fraction was subsequently isolated and purified as described under "Experimental Procedures" and examined using various analytical approaches.
Characterization of the H. pylori-binding Fraction of Rabbit Thymus-The negative ion FAB mass spectrum of the most strongly binding fraction from rabbit thymus fraction is shown in Fig The 1 H NMR spectrum of the heptaglycosylceramide fraction is shown in Fig. 3A. The analysis revealed features very similar to those found for the same structure having a terminal Neu5Ac (32) instead of Neu5Gc. Six ␤-doublets were seen in the anomeric region of the spectrum readily identifiable as Glc␤ H-1 at 4.18 ppm, two overlapping Gal␤ H-1 signals at 4.28 ppm originating from two internal residues, and a third Gal␤ H-1 signal at 4.24 ppm stemming from the residue linked to sialic acid whereas two overlapping Glc-NAc␤ H-1 signals were found at 4.67 and 4.68 ppm, respectively. The chemical shifts of the Gal and GlcNAc residues were consistent with a type-2 structure having a repetitive Gal␤4GlcNAc␤3 core. Sialic acid was identified by the quartet from H-3 eq at 2.77 ppm (not shown), a value indicating an ␣3 linkage. The amide substitution of the sialic acid was almost exclusively N-glycolyl according to the pseudodoublet seen at 3.88 ppm. Only a minor signal from an N-acetyl group was seen at 1.89 ppm (not shown). Thus, the major component in the binding fraction was identified as Neu5-Gc␣3Gal␤4GlcNAc␤3Gal␤4GlcNAc␤3Gal␤4Glc␤1Cer (Number 1 in Table I). Additionally a minor GlcNAc␤3 signal at 4.80 ppm indicated the presence of a small amount of the corresponding lacto structure, Neu5Gc␣3Gal␤3GlcNAc␤3-Gal␤3GlcNAc␤3Gal␤4Glc␤1Cer.
Epitope Dissection: Delineation of the Parent Heptaglycosylceramide Followed by Binding Studies-The heptaglycosylceramide fraction was degraded using either mild acid hydrolysis or enzymatic digestion, and the products were further investigated for H. pylori binding. The glycolipids with six, five, and four monosaccharides per ceramide were active, whereas the triglycosylceramide was not (Fig. 4, Chemical degradation). A relatively strong binding was noted to the generated pentaglycosylceramide fraction.
To evaluate the relative binding strengths of different neolacto fragments, the hexa-, penta-, and tetraglycosylceramide fractions (prepared as described under "Experimental Procedures") were applied on silica gel TLC plates in equimolar and decreasing amounts. The plates were then overlaid with 35 S-labeled H. pylori, and the binding was visually evaluated using autoradiography (Fig. 5). As shown, the pentaglycosylceramide was detectable in amounts down to 0.039 nmol/spot (n ϭ 7, S.D. d n Ϫ 1 ϭ 0.016 nmol), whereas hexa-and tetraglycosylceramides were active in amounts down to 0.2 and 0.3 nmol of glycolipid/spot, respectively. The combined results showed that binding of H. pylori to the heptaglycosylceramide of rabbit thymus was not dependent on the presence of sialic acid and that the most potent binder among the hydrolytic products was the pentaglycosylceramide, which has GlcNAc as a terminal monosaccharide.  1. Detection of an H. pyloribinding glycolipid in the acid glycolipid fraction of rabbit thymus. The glycolipids were separated on aluminumbacked silica gel plates using C/M/H 2 O (60:35:8) as a solvent system, and the binding assay was performed as described under "Experimental Procedures." The right plate represents autoradiography after binding of radiolabeled H. pylori strain CCUG17875, and the left plate shows glycolipids detected with anisaldehyde (Anis). Lane 1, non-acid (N) glycolipids of mouse feces, 20 g; Lane 2, non-acid glycolipids of rat small intestinal (int) epithelium, 40 g; Lane 3, non-acid glycolipids of human meconium, 40 g; Lane 4, acid (Ac) glycolipids of human blood group O erythrocytes (erythr.), 40 g; Lane 5, acid glycolipids of rabbit thymus, 20 g; Lane 6, bovine brain gangliosides, 40 g. Autoradiography was for 12 h. Structural Analysis of the Binding Glycolipid Products-The highly active hexa-and pentaglycosylceramide fractions were carefully investigated for the presence of structures other than neolacto structures using a combination of various analytical approaches.
The negative ion FAB spectra of these fractions (not shown) revealed pseudomolecular ions [M Ϫ H] Ϫ at m/z 1590.9 and 1428.8 and a series of Y i fragment ions in agreement with Hex-HexNAc-Hex-HexNAc-Hex-Hex-Cer and HexNAc-Hex-HexNAc-Hex-Hex-Cer, respectively.
The 1 H NMR analysis of the Hex-HexNAc-Hex-HexNAc-Hex-Hex-Cer revealed four major doublets in the anomeric region with ␤-couplings having an intensity ratio of 2:2:1:1 (Fig. 3B). The signals at 4.66 ppm (GlcNAc␤3), 4.26 ppm (internal Gal␤4), 4.20 ppm (terminal Gal␤4), and 4.17 ppm (Glc␤1) were in agreement with the results previously published for this compound (32,33). As expected, the dominating component was identified as the neolacto glycolipid Gal␤4GlcNAc␤3Gal␤4GlcNAc␤3Gal␤4Glc␤1Cer (Number 2 in Table I). The presence of a low intensity doublet at 4.80 ppm together with a small acetamido methyl signal at 1.81 ppm (seen as a shoulder on the large neolacto methyl resonance, not shown), confirmed the presence of a small fraction of lacto structure. The total amount of lacto linkage was estimated to be less than 10%. 1 H NMR analysis of the penta fraction (Fig. 3C) revealed five major ␤-doublets in the anomeric region at 4.65 ppm (internal GlcNAc␤3), 4.62 ppm (terminal GlcNAc␤3), 4.26 ppm (twoproton intensity, internal Gal␤4), and 4.17 ppm (Glc␤1), consistent with a GlcNAc␤3Gal␤4GlcNAc␤3Gal␤4Glc␤1Cer structure (Number 3 in Table I) and in agreement with the six-sugar compound having been stripped of its terminal Gal␤4. Again there was a small ␤-doublet at 4.79 ppm corresponding to 3-substituted GlcNAc␤3 (type-1 chain). The expected acetamido methyl signal was also seen as a shoulder on the much larger neolacto methyl signal at 1.82 ppm (not shown).
The carbohydrate part of the hexaglycosylceramide was additionally analyzed using GC/MS. The carbohydrate chains were released from the six-sugar-containing fraction by ceramide glycanase, an enzyme that hydrolyzes the ␤1-1 glycosidic linkage between the reducing sugar and the ceramide. The free oligosaccharides were then methylated and analyzed as shown in Fig. 6. Two carbohydrate fractions were identified in the material by GC/MS, corresponding to peaks A and B in the gas chromatogram (Fig. 6, upper panel). Both had the expected Hex-HexNAc-Hex-HexNAc-Hex-Hex carbohydrate sequence as confirmed by the oxonium B i fragment ions at m/z 187, 464 (432), 668, 913, and 1118, shown in the mass spectra in Fig. 6, middle and lower panels. As indicated in the figure, some of these ions could overlap with ions from the Z i series (31). The predominant oligosaccharide (peak B in the gas chromatogram), which accounted for more than 90% of the total material (93% according to peak area), was characterized by a strong fragment ion at m/z 182, confirming the presence of 4-substituted GlcNAc␤, typical of neolacto (type-2) carbohydrate chains (34). The minor oligosaccharide (peak A in the gas chromatogram, 7% according to peak area) gave, in accordance with NMR analyses, a spectrum representative of a lacto (type-1) carbohydrate chain with a very weak fragment ion at m/z 182 and a strong fragment ion at m/z 228.
In summary, the main components of the hexa-and pentaglycosylceramide fractions were of the neolacto type having structures of Gal␤4GlcNAc␤3Gal␤4GlcNAc␤3Gal␤4Glc␤1Cer (Number 2, Table I) and GlcNAc␤3Gal␤4GlcNAc␤3Gal␤4Glc␤1Cer (Number 3, Table I), respectively. Both contained in addition small amounts of lacto (type-1) structures with Gal␤3GlcNAc unit(s). Fucose-containing carbohydrates (possible H. pyloribinding blood group antigens, Refs. 8 and 17) were not detected in the mixtures. It should be noted that the ␤-galactosidase from E. coli used for preparation of pentaglycosylceramide from hexaglycosylceramide was not specific for the ␤1-4 linkage, thus hydrolyzing both lacto and neolacto units.

FIG. 3. Anomeric region of the 1 H NMR spectra of the H. pylori-binding neolacto (type-2) penta-, hexa-, and heptaglycosylceramides of rabbit thymus origin. The upper trace
shows the heptaglycosylceramide (NeuGc␣3Gal␤4GlcNAc␤3-Gal␤4GlcNAc␤3Gal␤4Glc␤1Cer) recorded at 400 MHz, whereas the middle and lower traces represent the hexaglycosylceramide (Gal␤4GlcNAc␤3Gal␤4GlcNAc␤3Gal␤4Glc␤1Cer) and pentaglycosylceramide (GlcNAc␤3Gal␤4GlcNAc␤3Gal␤4Glc␤1Cer), respectively, recorded at 500 MHz. Assignments of anomeric resonances are indicated above by the respective traces. All three samples were dissolved in Me 2 SO/D 2 O (98:2), and spectra were obtained at 30°C. moniae). The products were then tested for the interaction with H. pylori (Fig. 4, Enzymatic degradation). As expected, the reaction converted the pentaglycosylceramide to lactosylceramide (Gal␤4Glc␤1Cer) thereby abolishing the bacterial binding (Fig. 4, lanes 2). Reappearance of the activity in the tetraglycosylceramide region was not observed due to immediate enzymatic decomposition of this compound confirming the presence of the ␤1-4 glycosidic bond between Gal and the internal GlcNAc. The fate of the lacto component could not be followed on the plate because the amount of this fraction was below the chemical detection level. However, control analysis using lactotetraosylceramide, Gal␤3GlcNAc␤3Gal␤4Glc␤1Cer (Fig. 4,  lanes 4 and 5), confirmed that the ␤1-4-galactosidase used in this experiment did not hydrolyze Gal␤3GlcNAc units. Thus, the binding activity observed for the pentaglycosylceramide was dependent on the presence of the neolacto component, GlcNAc␤3Gal␤4GlcNAc␤3Gal␤4Glc␤1Cer (Number 3, Table I).
In a similar experiment the neolacto character of the epitope was proven for the hexaglycosylceramide fraction (not shown).
The interaction of H. pylori with neolacto carbohydrate b Glycosphingolipids derived from structure Number 1 by chemical or enzymatic degradation. c Glycosphingolipids de-N-acylated by use of anhydrous hydrazine.

FIG. 4. Examples of degradation studies: binding of H. pylori (S-032 strain) to glycolipid products obtained through chemical (left) or enzymatic (right) degradations.
Degradation products were separated on TLC silica gel plates and visualized either by chemical staining using anisaldehyde (Anis) or by overlay with 35 S-labeled bacterial cells (H. pylori). Chromatographic conditions were as described in Fig. 1. For chemical digestion, the parent heptaglycosylceramide was degraded using mild acid hydrolysis. The picture shows the chromatographic region comprising glycolipids with three to six monosaccharides per ceramide as indicated. For enzymatic digestion, the pentaglycosylceramide fraction was digested using a mixture of ␤1-4-galactosidase (S. pneumoniae) and ␤-hexosaminidase (S. pneumoniae). Lane 1, control pentaglycosylceramide; Lane 2, pentaglycosylceramide after digestion; Lane 3, reference glycolipids of human erythrocytes, from top, lactosylceramide (double band), trihexosylceramide, and globotetraosylceramide; Lane 4, control lactotetraosylceramide (human meconium); Lane 5, lactotetraosylceramide after digestion. chains should represent a new binding activity since no systematic correlation was observed between binding to lacto and neolacto carbohydrates using a number of bacterial strains with different binding specificities. On the other hand, a strong binding at lower picomole levels was independently found for both neolacto (Fig. 5) and lacto (11) structures.
H. pylori Binding to Related Glycolipids-In another series of experiments, a panel of various glycolipids with related structures was investigated for binding of H. pylori on TLC plates. (Fig. 7 and Table I). In addition to the neolacto glycolipids of rabbit thymus origin (Numbers 1-3 and 7 in Table I) and lactotetraosylceramide (Number 22 in Table I) previously described as binding-active for H. pylori (11), four other compounds were shown to interact with the bacterium (Numbers 4 -6 and 19). Structures Numbers 4, 5, and 6 are analogues of Number 3, having as terminal saccharide GalNAc␤, GalNAc␣, and Gal␣, respectively, instead of GlcNAc␤. Furthermore the only extension of GlcNAc␤3Gal␤4GlcNAc␤3Gal␤4Glc␤Cer tolerated by the bacteria was Gal␤4 as shown by binding to structures Numbers 1 and 2 (Table I). Other elongated structures such as Neu5Ac␤3-X-2 (Number 26) and GalNAc␤3-B5 (Number 17) were not bound by the bacteria. It may further be noticed that the acetamido group of the internal GlcNAc␤3 in B5 appears essential for binding since de-N-acylation of this side group to -NH 2 by treatment with anhydrous hydrazine (10) leads to a complete loss of binding (Number 12) as is also the case when the type-2 tetraglycosylceramide structure is similarly treated (Number 11).
Molecular Modeling and Cross-binding-To understand the binding characteristics of the different glycolipid molecules the conformational preferences of active as well as inactive structures were investigated by molecular modeling. Fig. 8 shows the minimum energy conformers of the terminal trisaccharide of the type-2 GlcNAc␤3-terminated pentaglycosylceramide (Number 3 in Table I) together with the minimum energy conformers of corresponding fragments of the three other binding-positive isoreceptors containing terminal GalNAc␤3 (Number 4), GalNAc␣3 (Number 5), and Gal␣3 (Number 6), respectively. As shown, despite differences in anomerity, the absence or presence of an acetamido group, axial or equatorial position of the 4-OH of the terminal sugar, and the fact that the ring plane of the terminal ␣3-linked compounds is raised somewhat above the corresponding plane of the ␤3-linked compounds, a substantial topographical similarity exists between these structures. These observations indicate that H. pylori binding to these four isoreceptors represents a case of molecular mimicry rather than being due to four separate specificities. Of interest is that H. pylori binds to the same four sugar sequences as found for toxin A from C. difficile (26) (Table II). However, binding by these two microbes is not identical because H. pylori is able to bind to internally located binding epitopes, whereas toxin A displays a definite preference for a terminally located epitope in the binding-active isoreceptor structures.
The relative binding strength of H. pylori to the rabbit thymus heptaglycosylceramide and to its six-, five-, and four-sugar daughter compounds indicates that the three-sugar fragment GlcNAc␤3Gal␤4GlcNAc␤3 constitutes the sequence required for maximal activity. Thus, in the hexaglycosylceramide an inhibitory effect from the terminal Gal␤4 is expected, whereas for tetraglycosylceramide lack of a terminal GlcNAc␤3 reduces the binding strength because only two of three sugars in the epitope are present. The essentiality of the internal GlcNAc␤3 is clearly shown by the loss of bacterial binding both to the tetraglycosylceramide and B5 following de-N-acetylation of the acetamido group to an amine (Numbers 11 and 12, Table I). In view of these results the absence of binding to e.g. the P1 antigen, the H5 type-2 glycolipid, and the Neu5Ac␣3/6-tetraglycosylceramide structures is easily rationalized (Numbers 13, 14, 24, and 25) because these extensions interfere directly with the proposed binding epitope. Also the glycolipid from bovine buttermilk (35), which has a ␤6-linked lactosamine FIG . 5. Binding of 35 S-labeled H. pylori (S-032 strain) to a dilution series of glycolipids of rabbit thymus origin on silica gel thin-layer plates. The figure shows relative strength of binding to tetra-and pentaglycosylceramides (upper panel) and to penta-and hexaglycosylceramides (lower panel). The glycolipids (Numbers 2, 3, and 7 in Table I)  branch attached to the internal Gal␤4 of the tetraglycosylceramide (Number 18), is non-binding due to blocked access to the binding epitope by the ␤6-linked extension. Furthermore addition of either GalNAc␤3 to B5 or Neu5Ac␣3 to X-2 results in complete loss of binding (Numbers 17 and 26). It is further seen that the negative influence of a Fuc␣2 unit as in H5 type-2 is confirmed by H. pylori non-binding to the A6 type-2 and B6 type-2 structures (Numbers 15 and 16).
Reproducibility of Binding by H. pylori Cells-About 60 different H. pylori strains were tested for binding to neolacto type-2 structures, and most of them (91%) turned out to be binding-active. The binding frequency on TLC plates by a selected strain (S-032) as recorded for hexa-and pentaglycosylceramides (Numbers 2 and 3 in Table I) was about 90% (n ϭ 100). A lower reproducibility (ϳ50%) was observed for the parent seven-sugar ganglioside (Number 1 in Table I) and tetraglycosylceramide (paragloboside, Number 7). The frequency at the level of 90% should be interpreted as a reliable and efficient binding because it is known that H. pylori is variable regarding expression of binding activities (4). It has been shown that the interaction with carbohydrate epitopes may change depending on bacterial strains and growth conditions (29,36) and that differences may appear even for different batches of the same strain grown in the same medium. 4 H. pylori strains are classified according to their binding abilities, and, for example, binding in a sialic acid-dependent way was observed for only about 30% of known cultured strains (37). DISCUSSION The novel specificity of H. pylori described in this study was first detected as a binding on TLC plates to a seven-sugar ganglioside fraction of rabbit thymus. Based on NMR and MS analyses the fraction was found to be a mixture of neolacto (type-2) and lacto (type-1) species, namely Neu5Gc␣3Gal␤4-GlcNAc␤3Gal␤4GlcNAc␤3Gal␤4Glc␤1Cer (repeated type-2 core sequence, more than 90%) and Neu5Gc␣3Gal␤3Glc-NAc␤3Gal␤3GlcNAc␤3Gal␤4Glc␤1Cer (repeated type-1 core sequence), respectively. Since the binding was not dependent on sialic acid, a detailed structural characterization of the active sequence was performed. It should be noted that we have earlier described a binding of H. pylori to lactotetraosylcera-mide (11), Gal␤3GlcNAc␤3Gal␤4Glc␤1Cer (type-1 chain). However, based on enzymatic degradation and extensive structural analysis of the generated glycolipids, the most active product was found to be the pentaglycosylceramide GlcNAc␤3Gal␤4-GlcNAc␤3Gal␤4Glc␤1Cer (Number 3, Table I).
In addition we found that three other pentaglycosylceramide analogues, having terminal GalNAc␤3, GalNAc␣3, and Gal␣3, respectively, interact with the bacterium. Molecular modeling showed that all four binding pentaglycosylceramides (Numbers 3, 4, 5, and 6, Table I) may have the terminal trisaccharide in a similar conformation (Fig. 8). The combined results indicate that the binding epitope requires a three-sugar sequence to obtain maximal activity. The interchangeability of the terminal saccharide of the four five-sugar glycolipids that bind H. pylori has also been observed for toxin A produced by C. difficile involving the same glycolipids (Table II), and furthermore, the same trisaccharide unit was found to represent the minimal binding epitope (26). Therefore it is likely that a common or similar structural motif is utilized for the binding domains of these two pathogenic agents. However, the observed difference is that toxin A does not tolerate any extensions terminally, indicating that the terminal saccharide in the five-sugar compounds is somewhat more important for binding of toxin A than for H. pylori. Quantitative studies regarding binding strength of the four isoreceptors were not performed due to shortage of the purified glycolipids.
Concerning human gastric mucosa, available glycoconjugate analyses do not allow a conclusion on the existence of the neolacto epitope to be drawn. Glycolipids isolated from mucosa scrapings apparently lacked the binding-active neolacto epitope (11), but glycoproteins and mucins have not yet been tested for this specificity.
A binding epitope similar to one described here has been found to be recognized by pneumococci and E. coli. The binding of several strains of S. pneumoniae to human nasopharyngeal epithelial cells was inhibited by oligosaccharides and glycoconjugates containing GlcNAc␤3Gal (49). Free GlcNAc had a strong inhibitory activity on hemagglutination of bovine erythrocytes by fimbrial components of diarrhea-associated E. coli (50), and GlcNAc␤3Gal was concluded to be the minimal sequence for binding to F17 fimbriae of E. coli (51).
As summarized in the Introduction a large number of carbohydrate-binding specificities have been reported for H. pylori, altogether nine specificities including the present neolacto epitope. The potential meaning of this complex set of properties has been discussed (4). Only two bacterial adhesins belonging to the unique H. pylori family of 30 outer membrane proteins have been identified, namely the Lewis b-binding adhesin, BabA (17), and the adhesin recognizing sialylated saccharides, SabA (18). It is likely that several of the other reported specificities are based on adhesins of this outer membrane protein family, and only when these are identified can more precise studies be performed to find out the biological relevance of all these potential bacterium-host protein-carbohydrate interactions. One possibility is that the bacterium is able to regulate the expression of individual specificities to target separate microniches or adapt adhesion to changes in microecology (4). In this respect it is of interest that the Lewis b epitope (8), the lactotetraosylceramide epitope (11), and sulfatide epitope are all present in gastric epithelium but apparently are lacking in neutrophils. On the other hand, sialylated glycoconjugates are practically lacking in normal gastric epithelium (4,18) but are abundant in neutrophils (45,46). Therefore, H. pylori may selectively attach to one of these two target cells, the epithelial cells or the neutrophils, to optimize the chances of bacterial persistence. In accordance with this it has been reported that experimental infection by H. pylori in monkeys is associated with modifications of expression of the bacterial outer membrane proteins, including BabA (54). Apparently neutrophil activation is needed to induce a mild inflammation, which provides H. pylori with products for nutritional purposes (4,52), and it is possible to inhibit this activation in vitro by sialylated oligosaccharides (53). A recent study (55) documented that SabA is essential for neutrophil activation and oxidative burst, whereas neither BabA, the neutrophil-activating protein HP-NAP, nor the vacuolating cytotoxin VacA are required. Furthermore it was shown that the link between binding to sialylated receptor and oxidative burst involves a G-protein-linked signaling pathway. Therefore, a main function of SabA is to target H. pylori to neutrophils and induce inflammation, whereas BabA and other potential adhesins rec-  Table  II and Numbers 3-6 in Table I). Despite differences in anomerity, the absence or presence of an acetamido group, axial or equatorial position of the 4-OH of the terminal sugar, and the fact that the ring plane of the terminal ␣3-linked compounds is raised somewhat above the corresponding plane of the ␤3-linked ones, a substantial topographical similarity exists between these structures, thus explaining their similar affinities for the bacterial adhesin.
ognizing lactotetraosylceramide or sulfatide may be necessary for adhesion to epithelial cells. It was also shown (18) that H. pylori colonization of human and monkey stomach stimulated expression of sialylated glycoconjugates of the gastric epithelium, improving bacterial adhesion, and the number of attached bacteria was directly related to the level of inflammation.
The biological relevance of the GlcNAc␤3Gal␤4GlcNAc␤ binding specificity described in the present study is still unclear. The binding is expressed by 90% of tested H. pylori strains; this is high compared with the expression of Lewis b or sialic acid-dependent binding specificities, which were found for about one-third of tested strains. Identification of the corresponding potential adhesin with affinity for neolacto structures and other studies including knock-out experiments may be necessary to find out the function of this novel specificity of H. pylori.