The Repertoire of Glycosphingolipids Recognized by Vibrio cholerae

The binding of cholera toxin to the ganglioside GM1 as the initial step in the process leading to diarrhea is nowadays textbook knowledge. In contrast, the knowledge about the mechanisms for attachment of Vibrio cholerae bacterial cells to the intestinal epithelium is limited. In order to clarify this issue, a large number of glycosphingolipid mixtures were screened for binding of El Tor V. cholerae. Several specific interactions with minor complex non-acid glycosphingolipids were thereby detected. After isolation of binding-active glycosphingolipids, characterization by mass spectrometry and proton NMR, and comparative binding studies, three distinct glycosphingolipid binding patterns were defined. Firstly, V. cholerae bound to complex lacto/neolacto glycosphingolipids with the GlcNAcβ3Galβ4GlcNAc sequence as the minimal binding epitope. Secondly, glycosphingolipids with a terminal Galα3Galα3Gal moiety were recognized, and the third specificity was the binding to lactosylceramide and related compounds. V. cholerae binding to lacto/neolacto glycosphingolipids, and to the other classes of binding-active compounds, remained after deletion of the chitin binding protein GbpA. Thus, the binding of V. cholerae to chitin and to lacto/neolacto containing glycosphingolipids represents two separate binding specificities.


Introduction
Diarrheal disease caused by Vibrio cholerae remains a major health problem in many parts of the world, leading to 100 000 deaths annually [1]. Infecting V. cholerae adhere to the small intestinal epithelium, and cause diarrhea primarily by the production of cholera toxin (CT). CT consists of one A-subunit with enzymatic activity, and five B-subunits mediating binding of the toxin to the small intestinal epithelium. In the early 1970s the GM1 ganglioside was identified as the receptor for CT [2]. Since then a large collection of data about the molecular details of the GM1 binding by CT, and the mechanisms for induction of diarrhea, has accumulated [3].
There are more than 200 O-antigen serogroups of V. cholerae known to date, but epidemic cholera is caused only by strains belonging to the O1 and O139 serogroups. The V. cholerae O1 strains are divided into two biotypes, designated classical and El Tor, that have evolved from separate lineages [4]. The first six pandemics were caused by the classical biotype, but after 1961 it has been displaced by the El Tor biotype. Recently, an evolution among El Tor strains with emergence of hybrid biotype strains with altered cholera toxin has been observed [5].
In contrast to the advanced information available about the interaction between CT and the GM1 ganglioside, the knowledge about the mechanisms for attachment of V. cholerae bacterial cells to the intestinal epithelium is sparse. Several hemagglutinins with roles in adherence have been isolated, but their precise binding specificities have not been defined [4]. Recently, a chitin-binding protein (GbpA), which mediates attachment of V. cholerae to chitin surfaces, human intestinal cells and mouse intestine, was characterized [6,7]. These interactions are inhibited by GlcNAc, but since chitin per se is not present on the intestinal epithelium, the carbohydrate sequences required for GbpA-mediated intestinal attachment have not been elucidated.
In the present study we searched for V. cholerae carbohydrate recognition by binding of V. cholerae to a large number of glycosphingolipid mixtures. Several specific interactions with minor non-acid glycosphingolipids were thereby detected, and after isolation of binding-active glycosphingolipids, and characterization by mass spectrometry and 1 H NMR, three distinct modes of glycosphingolipid binding were defined.

Construction of gbpA Deletion Mutant Strains
The gbpA genes in two different V. cholerae strains were disrupted by insertional mutagenesis using a suicide plasmid. One was the El Tor strain JBK70, and the other was the classical strain JS1569, a rifampicin resistant mutant of CVD103 [9].
A 520 base-pair fragment from the central region of the gbpA gene was amplified from strain JS1569 using primers gbpAf2 (59-GGGGAGATCTAAACTCGCGTGTTTGATAACGAG-39) and gbpAr2 (59- . The regions of the primers in italics are non-homologous tails that contain a recognition site for the restriction endonuclease XbaI. The underlined nucleotide in primer was changed from C to A in order to introduce a stop codon in the gbpA gene in addition to the insertion of suicide plasmid. The amplified DNA was digested with XbaI and ligated into the synthetic R6K-based suicide vector pMT-suicide1 [11]. Ligated DNA was transformed into the E. coli strain S17-1 [12], and positive clones were isolated on the basis of restriction analysis of plasmids isolated from chloramphenicol resistant transformants. The resulting plasmid was then transferred into the recipient strains of V. cholerae by transconjugation accomplished by patch mating and selection of transconjugants by resistance to chloramphenicol and either rifampicin (JS1569) or polymyxin B (JBK70). The presence of the insert in the chromosomes of the resulting transconjugants was confirmed by PCR using primers gbpAf1 (59-GCCAACCACGT-CACAAAGGATTCC-39) and gbpAr1 (59-GAGTGGAGAGG-TAGCCACTGGAG-39) which gave a fragment 2.5 kb larger than the 1.2 kb wild-type due to insertion of the pMT-suicide1 plasmid and a repeat of a 520 bp fragment of the gbpA gene. The resulting V. cholerae strains were given the strain numbers 1382 (El Tor) and 1375 (classical). Culture and labeling of these gbpA deletion mutant strains was done using the same conditions as above, with addition of chloramphenicol 12.5 mg/ml to the agar plates and the AKI medium.

Pretreatment of Bacterial Cells
Suspensions of radiolabeled V. cholerae El Tor were prepared as described above, and suspended in PBS. Prior to incubation on thin-layer chromatograms the bacteria were subjected to one of the following treatments: (I) incubation at 37uC for 60 min; (II) incubation at 65uC for 60 min; (III) treatment with trypsin (Sigma) 1 mg/ml at 37uC for 60 min.
Binding of radiolabeled bacteria to glycosphingolipids on thinlayer chromatograms was done as described [14]. Dried chromatograms were dipped in diethylether/n-hexane (1:5 v/v) containing 0.5% (w/v) polyisobutylmethacrylate for 1 min. The chromatograms were blocked with BSA/PBS/TWEEN for 2 h at room temperature. Then the plates were incubated for 2 h at room temperature with 35 S-labeled bacteria (1-5610 6 cpm/ml) diluted in BSA/PBS/TWEEN. After washing six times with PBS, and drying, the plates were autoradiographed for 12-36 h using XAR-5 x-ray films (Eastman Kodak, Rochester, NY). Chromatogram binding assays with 125 I-labeled lectin from Erythrina cristagalli were done as described [15].

Isolation of El Tor Binding Glycosphingolipids
Total acid and non-acid glycosphingolipid fractions were isolated by standard methods [13]. The non-acid glycosphingolipid fractions were separated by repeated silicic acid chromatography, and final separation was achieved by HPLC or by chromatography on Iatrobeads (Iatrobeads 6RS-8060; Iatron Laboratories, Tokyo) columns, eluted with chloroform/methanol/water 65:25:4 (by volume), followed by chloroform/methanol/water 60:35:8 (by volume), and finally chloroform/methanol/ water 40:40:12 (by volume). The fractions obtained were analyzed by thin-layer chromatography and anisaldehyde staining, and the V. cholerae El Tor binding activity was assessed using the chromatogram binding assay. The fractions were pooled according to the mobility on thin-layer chromatograms and their El Tor binding activity.

Endoglycoceramidase Digestion and Liquid Chromatography/Electrospray Ionization Mass Spectrometry
Endoglycoceramidase II from Rhodococcus spp. [16] (Takara Bio Europe S.A., Gennevilliers, France) was used for hydrolysis of glycosphingolipids, and the oligosaccharides obtained were analyzed by capillary-liquid chromatography mass spectrometry and tandem mass spectrometry [17]. In brief, the oligosaccharides were separated on a column (20060.180 mm) packed in-house with 5 mm porous graphite particles (Hypercarb, Thermo Scientific), and eluted with an acetonitrile gradient (A: 10 mM ammonium bicarbonate; B: 100% acetonitrile). The saccharides were analyzed in the negative ion mode on an LTQ linear quadrupole ion trap mass spectrometer (Thermo Electron, San José, CA). The IonMax standard ESI source on the LTQ mass spectrometer was equipped with a stainless steel needle kept at -3.5 kV. Compressed air was used as nebulizer gas. The heated capillary was kept at 270uC, and the capillary voltage was -50 kV. Full-scan (m/z 380-2 000, 2 microscans, maximum 100 ms, target value of 30 000) was performed, followed by data dependent MS 2 scans of the three most abundant ions in each scan (2 microscans, maximum 100 ms, target value of 10 000). The threshold for MS 2 was set to 500 counts. Normalized collision energy was 35%, and an isolation window of 3 u, an activation q = 0.25, and an activation time of 30 ms, was used. The conditions for MS 3 and MS 4 were the same, except that the thresholds were set to 300 and 100 counts, respectively.

Proton NMR Spectroscopy
1 H NMR spectra were acquired on a Varian 600 MHz spectrometer at 30uC. Samples were dissolved in dimethyl sulfoxide/D 2 O (98:2, by volume) after deuterium exchange.

Screening for V. cholerae Carbohydrate Recognition
In the initial screening for carbohydrate recognition by V. cholerae, mixtures of glycosphingolipids from various sources were used in order to expose the bacteria to a large number of potentially binding-active carbohydrate structures. Thus, the binding of the bacteria to acid and non-acid glycosphingolipid mixtures isolated from the small intestine of different species (human, rat, cat, rabbit, dog, monkey and pig), erythrocytes of different species (human, cat, rabbit, dog, horse, chicken and sheep), human cancers (lung cancer, kidney cancer, colon cancer, liver cancer and gastric cancer), and rabbit thymus, was tested. In most non-acid glycosphingolipid fractions a binding of both the El Tor strain and the classical strain to compounds migrating in the monoand diglycosylceramide regions was observed (exemplified in Figure 1, lanes 2,7,8,10). In addition, a number of slow-migrating minor glycosphingolipids in non-acid fractions of human small intestine, dog erythrocytes, cod intestine, rabbit thymus, rabbit erythrocytes, and chicken erythrocytes were distinctly recognized by the El Tor strain (Figure 1, lanes 1-3, 8-10).

Characterization of the El Tor Binding Slow-migrating Glycosphingolipid of Human Small Intestine
After separation of the non-acid fraction from human small intestine, the subfractions obtained were pooled according to El Tor binding activity, giving a subfraction containing the slowmigrating El Tor binding compound (designated fraction HI-I).
Based on the observations listed below fraction HI-I was characterized as Galb3GlcNAcb4Galb3(Fuca4)GlcNAcb3-Galb4Glcb1Cer, i.e. a Le a pentaosylceramide substituted with a terminal lacto moiety.

I)
On thin-layer chromatograms the El Tor binding glycosphingolipid migrated in the hexa2/heptaglycosylceramide region ( Figure 1, lane 1; Figure 2A H + ] 2 ion at m/z 1217 gave a C-type fragment ion series (C 2a at m/z 382, C 3a at m/z 544, C 4a at m/z 893, and C 5a at m/z 1055) demonstrating a Hex-HexNAc-Hex-(Fuc)Hex-NAc-Hex-Hex sequence ( Figure 2E). Intense cross-ring 0,2 Atype fragments are diagnostic for carbohydrates substituted at C-4 [17][18][19]. The 0,2 A 6 ion at m/z 1157, and the 0,2 A 6 -H 2 O ion at m/z 1139, were derived from cross-ring cleavage of the 4-substituted Glc of the lactose unit at the reducing end. However, no 0,2 A 2 fragment ion at m/z 281 (not shown), or 0,2 A 4 fragment ion at m/z 646, were observed, suggesting that the HexNAcs were substituted at 3-position. III) 1 H NMR of fraction HI-I shows a mixture of three compounds, all with Le a signatures ( Figure 2F). The two major structures (,35% and 60%, respectively) are Galb 3(Fuca4)GlcNAcb4Galb3(Fuca4)GlcNAcb 3-Galb4Glcb1Cer previously characterized by NMR [20], and its precursor Galb3GlcNAcb4Galb3(Fuca4)Glc-NAcb3Galb4Glcb1Cer in which the terminal Fuca4 is lacking resulting in a lacto terminal, as found above by mass spectrometry. This structural difference results in drastic shifts of the anomeric resonances arising from the lacto terminal relative to the same residues in the Le a terminal: downfield by +0.034 ppm for GlcNAcb3 and upfield for Galb3 by 20.169 ppm, thus yielding shifts corresponding to those seen for the lacto-terminated hexaosylceramide at 4.794 ppm and 4.14 ppm (Table 1). All other resonances of the lacto-terminated compound overlap completely with those of the dimeric Le a structure ( Figure 2F; Table 1). The third minor compound (,5%) is the Le a pentaosylceramide (Galb3(Fuca4)GlcNAcb3Galb4Glcb1Cer).

Characterization of the El Tor Binding Slow-migrating Glycosphingolipids of Rabbit Thymus
After separation of the non-acid glycosphingolipid fraction from rabbit thymus, and pooling of the subfractions obtained according El Tor binding activity, four El Tor binding fractions were obtained, designated fraction TH-I, TH-II, TH-III, and TH-IV, respectively. These fractions were characterized by mass spectrometry, 1 H NMR, and binding of the Galb4GlcNAcb-binding E. cristagalli lectin [14], as follows.
A. Fraction TH-I. Fraction TH-I was characterized as GlcNAcb3Galb4GlcNAcb3Galb4Glcb1Cer. This conclusion is based on the following observations:

I)
On thin-layer chromatograms fraction TH-I migrated in the pentaglycosylceramide region ( Figure 3B, lane 2).

II)
No binding of the E. cristagalli lectin to fraction TH-I was obtained ( Figure 3C, lane 2).

III)
LC-ESI/MS of the oligosaccharides obtained by hydrolysis of fraction TH-1 gave a major [M-H + ] 2 ion at m/z 909, demonstrating an oligosaccharide with two HexNAc and three Hex ( Figure 4A). MS 2 of the ion at m/z 909 resulted in a series of C-type fragment ions (C 2 at m/z 382, C 3 at m/z 585, and C 4 at m/z 747) identifying a pentasaccharide with HexNAc-Hex-HexNAc-Hex-Hex sequence ( Figure 4B). The 0,2 A 3 fragment ion at m/z 484, and the 0,2 A 3 -H 2 O fragment ion at m/z 466, indicated a 4-substituted internal HexNAc. The 0,2 A 5 ion at m/z 849, and the 0,2 A 5 -H 2 O ion at m/z 831, were derived from cross-ring cleavage of the 4-substituted Glc of the lactose part.

IV)
The proton NMR spectrum of fraction TH-1 in Figure 4C reveals a single species showing two GlcNAcb3 residues at 4.603 ppm and 4.641 ppm, two overlapping Galb4 residues at 4.25 ppm and a Glcb1 at 4.15 ppm, consistent with the structure GlcNAcb3Galb4GlcNAcb3-Galb4Glcb1Cer [21] (Table 1).

B. Fraction TH-II. Characterization of the El Tor binding fraction
TH-II demonstrated neolactohexaosylceramide (Galb4GlcNAcb3Galb4GlcNAcb3Galb4Glcb1Cer) as the major compound. This conclusion was based on the following properties:

II)
Fraction TH-II was recognized by the E. cristagalli lectin ( Figure

IV)
The 1 H NMR spectrum of fraction TH-II (not shown) reveals a mixture of neolactohexaosylceramide, as evi-    C. Fraction TH-III. The main compound of fraction TH-III was a neolactoheptaosylceramide with a terminal Galili epitope (Gala3Galb4GlcNAcb3Galb4GlcNAcb3Galb4Glcb1Cer). This conclusion was based on the following observations:

I)
Fraction TH-III migrated in the heptaglycosylceramide region on thin-layer chromatograms ( Figure 3D-E, lane 6).

II)
Fraction TH-III was not recognized the E. cristagalli lectin ( Figure 3F, lane 6).      cleavage of the 4-substituted Glc of the lactose unit at the reducing end.

IV)
The 1 H NMR spectrum of fraction TH-III (not shown) shows a structure having a Galili terminus elongated by an internal Galb4GlcNAcb3 yielding Gala3Galb4Glc-NAcb3Galb4GlcNAcb3Galb4Glcb1Cer, having almost identical shift values as the B5 pentaosylceramide (Table 1). A minor second species is also present, seen by a GlcNAcb3 doublet at 4.794 ppm and a Galb3 doublet at 4.13 ppm, which indicate a lacto-terminated hexaglycosylceramide Galb3GlcNAcb3Galb4Glc-NAcb3Galb4Glcb1Cer [20], where the remaining anomeric signals strongly overlap with those of the other structure (Table 1).
D. Fraction TH-IV. El Tor binding fraction TH-IV was characterized as neolactooctaosylceramide (Galb4GlcNAcb3-Galb4GlcNAcb3Galb4GlcNAcb3Galb4Glcb1Cer). This conclusion was based on the following properties:

IV)
The 1 H NMR spectrum of TH-IV ( Figure S1D) is very similar to TH-II revealing a neolactoocta compound and the seven-sugar Galili-terminated structure described above. The chemical shifts are thus identical to those described for fraction TH-II (Table 1).
Thus, the slow-migrating El Tor binding glycosphingolipids of rabbit thymus were characterized as GlcNAcb3-neolactotetraosylceramide, neolactohexaosylceramide, the B7 heptaosylceramide and neolactooctaosylceramide. In fraction TH-II, the B5 pentaosylceramide was also present, but this compound was not recognized by of V. cholerae (see below).
The El Tor binding glycosphingolipids of cod intestine (Figure 1, lane 3) and chicken erythrocytes (Figure 1, lane 10) were also identified as neolactohexaosylceramide by the same set of methods (data not shown). Furthermore, the uppermost El Tor binding glycosphingolipid of dog erythrocytes (Figure 1, lane 2) was characterized as Gala3Galb4GlcNAcb6(Gala3Galb4Glc-NAcb3)Galb4GlcNAcb3Galb4Glcb1Cer ( Figure S2), i.e. a branched neolactodecaglycosylceramide with terminal Galili epitopes.
Binding to neolactotetraosylceramide and lactotetraosylceramide required 4 mg of the compound on the thin-layer plate, and was only found in approximately 70% of the experiments, indicating that the binding epitope was not optimally exposed in these relatively short glycosphingolipids. For neolactotetraosylceramide an increased binding is obtained by the addition of a terminal GlcNAcb3 (No. 2), and inspection of the El Tor binding compounds in Figure 5 reveals that the GlcNAcb3Galb3/ 4GlcNAc sequence is the common denominator of all the El Tor binding lacto/neolacto based glycosphingolipids with more than four carbohydrate units.

Characterization of a Novel Glycosphingolipid of Rabbit Erythrocytes Recognized by V. cholerae
The non-acid glycosphingolipids of rabbit erythrocytes has been characterized in detail by mass spectrometry and NMR, demonstrating a number of Gala3Galb4GlcNAc-terminated linear and di2/triantennary oligo-and polyglycosylceramides [22][23][24][25]. The major non-acid glycosphingolipid of rabbit erythrocytes is the B5 pentaosylceramide, and the El Tor binding glycosphingolipid was found just below this compound ( Figure 1B, lane 9; Figure 6B, lane 4). Separation of a non-acid glycosphingolipid fraction from rabbit erythrocytes and pooling of the subfractions obtained according to El Tor binding activity gave a fraction containing the El Tor binding compound (designated fraction RE-I). Structural characterization by mass spectrometry and proton NMR, as detailed below, identified two glycosphingolipids in fraction RE-I; the B5 pentaosylceramide and a novel glycosphingolipid structure, Gala3Gala3Galb4GlcNAcb3Galb4Glcb1Cer, i.e a B5 pentaglycosylceramide elongated by Gala3.  Figure 6F). The spectrum also had three a-signals clustered around 4.90 ppm, one of which coincides precisely with an internal Gala3 (4.893 ppm) substituted with a terminal Gala3, as found earlier for Gala3Gala3Galb4Glcb1Cer [26]. Thus, a structure with the sequence Gala3Gala3Galb4GlcNAcb3-Galb4Glcb1Cer is present, in accord with the mass spectrometry data above, and where the shifts for the remaining sugar residues are very close to those found for the B5 pentaosylceramide ( Table 1). The two a-signals seen on either side of the internal Gala3 resonance just mentioned could not be assigned.

Binding to Reference Glycosphingolipids
Finally, the binding of the El Tor strain to a number of reference glycosphingolipids was evaluated (summarized in Table 2). Like many other bacteria, the El Tor strain bound to lactosylceramide with phytosphingosine and hydroxy fatty acids (No. 32 in Table 2; Figure 7B When reference glycosphingolipids with terminal a3-linked Gal were examined for V. cholerae El Tor binding activity ( Figure 7D) the bacteria bound to Gala3Gala3Galb4Glcb1Cer (lane 3), in addition to the Gala3Gala3Galb4GlcNAcb3Galb4Glcb1Cer from rabbit erythrocytes (lane 2). Again, no binding to the B5 pentaosylceramide (lane 1) occurred, and the B type 2 hexaosylceramide (Gala3(Fuca2)Galb4GlcNAcb3Galb4Glcb1Cer; lane 4) was also non-binding.

Glycosphingolipid Binding by Classical V. cholerae
The glycosphingolipid binding of the El Tor strain and the classical strain were tested in parallel throughout the experiments described above. Although the classical strain often gave a high background, the glycosphingolipid binding pattern of the classical strain ( Figure 7F) correlated well with the glycosphingolipid recognition by the El Tor strain. Thus, the classical strain also bound to neolacto-(lane 1), lacto-(lane 2), and Gala3Gala3Galterminated glycosphingolipids (lane 3), and also to lactosylcer-

Glycosphingolipid Recognition by a GbpA Deletion Mutant Strain
Thereafter, the role of the chitin binding protein GbpA in the binding of V. cholerae to glycosphingolipids was investigated by using a strain with inactivation of the gbpA gene. This strain displayed the same binding activities as the parental strain, i.e. lacto/neolacto glycosphingolipids, Gala3Gala3Gal-terminated glycosphingolipids and lactosylceramide/gangliotriaosylceramide were still recognized (exemplified in Figure 7H).

Pretreatment of Bacterial Cells
In order to investigate the bacterial structures involved in V. cholerae glycosphingolipid binding the bacterial cells were subjected to pretreatments prior to the chromatogram bindng assays. Trypsin treatment and heating of the bacteria each decreased the binding to neolacto, lacto, Gala3Gala3Gal-terminated, and lactosylceramide-related glycosphingolipids ( Figure S3).

Discussion
Adhesion of microbes to target cells is considered to be an important step in the infection process, allowing an efficient delivery of toxins and other virulence factors close to the cell surface. Potential host receptors, the majority of which are glycoconjugates, have been identified for a large number of bacteria [27,28]. However, little or no data exists regarding the potential adhesion strategies of V. cholerae bacterial cells. In this study, the carbohydrate binding specificity of one El Tor strain and one classical strain was examined by binding to a large number of variant glycosphingolipids. Thereby, highly specific interactions with a number of non-acid glycosphingolipids were obtained. We focused on minor complex El Tor binding compounds, which were isolated and characterized by mass spectrometry and 1 H NMR. After comparison with the binding of the bacteria to a number of reference glycosphingolipids, three modes of El Tor glycosphingolipid recognition could be discerned.

I. Lacto/Neolacto Binding
Most of the V. cholerae binding glycosphingolipids isolated and characterized in this study have lacto or neolacto core chains ( Figure 5, Table 2). Lacto and neolacto glycosphingolipids are present in the human small intestinal epithelium [29][30][31], thus may contribute to V. cholerae colonization of the human small intestine. The V. cholerae binding lacto-Le a glycosphingolipid (Galb3Glc-NAcb4Galb3(Fuca4)GlcNAb3Galb4Glcb1Cer) was isolated from human small intestine, and the binding of the Galb4GlcNAcbinding E. cristagalli lectin to slow-migrating glycosphingolipids of human small intestine demonstrates the presence of complex unsubstituted neolacto glycosphingolipids in the human small intestine [32]. In contrast, the N-linked glycans of human small intestinal glycoproteins have terminal neolacto chains which are heavily substituted with fucose [33], and thus should not be recognized by V. cholerae.
For high affinity binding to lacto/neolacto based glycosphingolipids a GlcNAcb3Galb3/4GlcNAc sequence is required ( Figure 5). However, certain substitutions of the GlcNAcb3Galb3/4GlcNAc sequence abolished the binding of the bacteria. Thus, addition of a terminal NeuGca3 to the terminal Gal of neolactohexaosylcer- The main difference between the lacto core and the neolacto core glycosphingolipids is that substitution of the innermost GlcNAc by a Fuca is tolerated for the compound with lacto core (No. 25), while it abolishes V. cholerae binding in the case of the glycosphingolipid with neolacto core (No. 15). This difference could be caused by conformational effects of the Fuc residues on the exposure of the binding epitope. It may also be that the lacto and neolacto sequences are recognized by two separate adhesins. The latter suggestion is somewhat supported by the fact that the classical strain fails to bind to the lacto terminated glycosphingolipid from human intestine, although it binds to the complex neolacto glycosphingolipids. Resolution of this issue must await the identification of the adhesin(s) involved in the lacto/neolacto binding.
There is a strong correlation between the severity of cholera and the blood group ABO phenotype of infected individuals, with blood group O individuals being more prone to develop severe diarrhea upon contracting V. cholerae infection than individuals with blood groups A, B or AB [34,35]. However, in this study no binding of V. cholerae bacterial cells to glycosphingolipids with terminal blood group A, B or H determinants was obtained. A tempting speculation is that the bacteria instead might benefit from the binding to lacto and neolacto core chains, since they thereby will avoid individual variations.
Recognition of glycoconjugates with neolacto sequences is a common theme in microbial carbohydrate binding, and has also been described for e.g. Streptococcus pneumoniae [36], Helicobacter pylori [21], and relapsing fever Borrelia [37]. In each case the detailed structural requirements for binding differs. Interestingly, the V. cholerae binding GlcNAcb3Galb4GlcNAc sequence is also the minimum sequence recognized by H. pylori. However in contrast to V. cholerae, H. pylori binds to sialyl-neolactohexosylceramide, indicating a different architecture of the carbohydrate binding site of the neolacto binding adhesins.
In summary, glycosphingolipids with repetitive lacto or neolacto elements are preferentially recognized by V. cholerae. An increased binding is obtained by addition of a terminal b3-linked GlcNAc to neolactotetraosylceramide, indicating that the preferred binding epitope is the GlcNAcb4Galb4GlcNAc sequence. Since minute amounts of several of the binding-active glycosphingolipids were obtained the relative binding affinities could, unfortunately, not be accurately determined.

II. Gala3Gala3Gal Binding
The binding to Gala3Gala3Galb4GlcNAcb4Galb4Glcb1Cer and Gala3Gala3Galb4Glcb1Cer is the second binding specificity of V. cholerae. Since the bacteria do not recognize the B5 glycosphingolipid (Gala3Galb4GlcNAcb4Galb4Glcb1Cer), the minimum binding epitope in this case is the Gala3Gala3Gal moiety. This sequence could be the product of the a1,3galactosyltransferase or the iGb3 synthase [38,39]. However, functional forms of these two enzymes have not been found in humans, and thus it is not likely that glycoconjugates with the Gala3Gala3Gal sequence have a role in the adhesion of V. cholerae to the human small intestine.

III. Lactosylceramide Binding
The third carbohydrate binding specificity of V. cholerae is represented by the binding to lactosylceramide, with a concomitant binding to galactosylceramide, isoglobotriaosylceramide, gangliotriaosylceramide and gangliotetraosylceramide. The binding in the mono-, di-and triglycosylceramide region in Figure 1, Figure 7F and 7H, and Figure S3B and S3D, is thus most likely due to recognition of these compounds. The same glycosphingolipid binding pattern has previously been reported for many other bacteria, both pathogens and members of the indigenous flora [27]. While minute amounts of diglycosylceramides, most likely lactosylceramide, are present in the human intestinal epithelium [29], isoglobotriaosylceramide, gangliotri-and gangliotetraosylceramide have not been found in these cells and, thus, it is unlikely that these compounds are critical receptors for V. cholerae.
V. cholerae chitin binding protein GbpA has been proposed to mediate attachment of the bacteria to human intestine by binding to GlcNAc-containing glycoconjugates [6,40]. We hypothesized that GbpA was involved in the binding of V. cholerae to the lacto/ neolacto containing glycosphingolipids. However, the lacto/ neolacto binding of V. cholerae remained after inactivation of the gbpA gene. Thus, the binding of V. cholerae to chitin and to lacto/ neolacto containing glycosphingolipids represents two separate binding specificities. This is supported by the glycan array data in a recent crystalization study of GbpA, where only binding to chitin oligosaccharides was observed [7]. The GbpA deletion mutant strain also bound to the Gala3Gala3Gal-terminated compounds and the set of reference glycosphingolipid recognized by the wild type El Tor strain.
One important virulence factor of V. cholerae is the toxincoregulated pilus (TCP). Expression of TCP is essential for V. cholerae colonization of the human small intestine [41], and it has been suggested that TCP mediates binding of the bacteria to the intestinal epithelium [42]. The culture conditions mainly used in the present study (AKI medium at 30uC) favors TCP production [43]. However, the binding-active glycosphingolipids were also recognized after culture in CFA broth at 27uC, i.e. under conditions when little or no TCP is expressed, indicating that the glycosphingolipid binding capacity of V. cholerae does not reside in the toxin-coregulated pilus.
The glycosphingolipid binding was decreased by treatment of the bacteria with trypsin and heat, indicating that bacterial cell surface proteins were involved in the interaction with carbohydrates. However, Schild et al. have reported that the attachment of lipopolysaccharide and capsule mutants of V. cholerae to the brush borders of the mucus-producing human intestinal cell line HT29-Rev MTX is reduced, demonstrating that lipopolysaccharides are important for V. cholerae adherence [44]. In order to investigate the role of lipopolysaccharides in V. cholerae glycosphingolipid binding, the bacterial cells were incubated with polyclonal antibodies to V. cholerae O1 antigen prior to the chromatogram binding assays. Occasionally a reduced binding to some glycosphingolipids was thereby obtained (data not shown), but the patterns were not consistent. One possible explanation for this variable reduction is that the coating of the bacterial cell surface with antibodies causes a general steric hindrance of cell surface components and adhesins that promote glycosphingolipid binding.
In this study we have for the first time defined the carbohydrate binding properties of V. cholerae, which is important for understanding the molecular interactions between the bacteria and its target host cells. The next step is identification, isolation and molecular cloning of the adhesins involved in the different carbohydrate binding specificities, in order to evaluate their roles in the infection process. The designation J refers to Table 1. The 1 H NMR spectrum reveals an essentially pure compound, which is characterized by two overlapping Gala3 resonances at 4.827 ppm, two GlcNAcb3 at 4.652 ppm, one GlcNAcb6 at 4.395 ppm, three Galb4 around 4.28 ppm, a fourth Galb4 at 4.254 ppm and lastly Glcb1 at 4.161 ppm (Table 1). These data are practically identical to previously published findings for the branched Gala3Galb4Glc-NAcb6(Gala3Galb4GlcNAcb3)Galb4GlcNAcb3Galb4Glcb1Cer allowing for temperature-induced differences [24].