Novel Carbohydrate Binding Site Recognizing Blood Group A and BDeterminants in a Hybrid of Cholera Toxin and Escherichia coliHeat-labile Enterotoxin B-subunits*

(Received for publication, June 25, 1999, and in revised form, October 12, 1999)

Jonas Ångstrom‡, Malin Backstrom§, Anna Berntsson‡, Niclas Karlsson‡, Jan Holmgren§,Karl-Anders Karlsson‡, Michael Lebens§, and Susann Teneberg‡¶

From the ‡Institute of Medical Biochemistry, Goteborg University, P. O. Box 440, SE 405 30 Goteborg, Sweden,and the §Department of Medical Microbiology and Immunology, Goteborg University, Guldhedsgatan 10,SE 413 46 Goteborg, Sweden

The B-subunits of cholera toxin (CTB) and Esche-richia coli heat-labile enterotoxin (LTB) are structur-ally and functionally related. However, the carbohy-drate binding specificities of the two proteins differ.While both CTB and LTB bind to the GM1 ganglioside,LTB also binds to N-acetyllactosamine-terminated gly-coconjugates. The structural basis of the differences incarbohydrate recognition has been investigated by asystematic exchange of amino acids between LTB andCTB. Thereby, a CTB/LTB hybrid with a gain-of-functionmutation resulting in recognition of blood group A andB determinants was obtained. Glycosphingolipid bind-ing assays showed a specific binding of this hybrid B-subunit, but not CTB or LTB, to slowly migrating non-acid glycosphingolipids of human and animal smallintestinal epithelium. A binding-active glycosphingo-lipid isolated from cat intestinal epithelium was charac-terized by mass spectrometry and proton NMR asGalNAca3(Fuca2)Galb4(Fuca3)GlcNAcb3Galb4GlcNAcb3Galb4Glcb1Cer. Comparison with reference gly-cosphingolipids showed that the minimum bindingepitope recognized by the CTB/LTB hybrid wasGala3(Fuca2)Galb4(Fuca3)GlcNAcb. The blood group Aand B determinants bind to a novel carbohydrate bind-ing site located at the top of the B-subunit interfaces,distinct from the GM1 binding site, as found by dockingand molecular dynamics simulations.

Enterotoxins produced by Vibrio cholerae and enterotoxi-genic Escherichia coli are causative agents of diarrheal dis-eases leading to millions of deaths annually (1). Both choleratoxin (CT)1 and E. coli type I heat-labile enterotoxin (LT) are

oligomeric proteins with one A-subunit and five B-subunits (2).The A-subunits have ADP-ribosyltransferase activity, whilethe B-subunits mediate binding to receptors on the eukaryoticcell surface. LT type I B-subunits originating from porcine andhuman isolates of enterotoxigenic E. coli (pLTB and hLTB,respectively) share 96% sequence identity with each other, butonly approximately 80% with CTB (3).

Despite great similarity between the carbohydrate bindingsites of the B-subunits, evidenced by recent crystal complexes(4–6), the carbohydrate binding specificities of CTB and LTBdiffer. Both B-subunits bind with high affinity to GM1 (7),whereas only LTB interacts with N-acetyllactosamine-termi-nated glycoconjugates (8–10). Binding of LTB, but not CTB, togangliotetraosylceramide and the GD1b ganglioside has alsobeen reported (8, 10, 11). The structural basis of the differencesin carbohydrate binding between CTB and hLTB have beeninvestigated by construction of a number of CTB/hLTB hybrids,having CTB amino acids substituted with heterologous aminoacids of hLTB (12). By introducing hLTB residues in the 1–25region and at positions 94 and 95 of CTB, a hybrid B-subunit(designated LCTBH) was created, with N-acetyllactosamine-binding properties almost indistinguishable from hLTB.

These studies have now been extended by re-substitution ofsingle amino acids of LCTBH back to the original CTB resi-dues. By re-substituting from Ser4 to Asn, a daughter hybridwith reduced N-acetyllactosamine-binding capacity was ob-tained. However, this daughter hybrid had a novel carbohy-drate binding specificity, as described in the present paper.

EXPERIMENTAL PROCEDURES

Plasmids and DNA Manipulations—All recombinant B-subunitswere produced from plasmids derived from pML-CTBtac, essentially asdescribed (13). pML-LCTBtacK was obtained by introduction of syn-thetic oligonucleotides between the unique SacI and PstI sites in pML-LCTBtacH. The mutated plasmid was electroporated into the classical01 V. cholerae strain JS1569 (DctxA). The structure of the hybrid genewas confirmed by DNA sequencing using Sequenase 2.0 from USB(Amersham Pharmacia Biotech, United Kingdom). Oligonucleotideswere from KEBO Lab, Spånga, Sweden, and restriction enzymes werefrom Roche Molecular Biochemicals or New England Biolabs, and usedaccording to the manufacturer’s instructions.

Production, Purification, and Characterization of B-subunits—Re-combinant B-subunits were produced and purified as described (12).Purified B-subunits were analyzed by SDS-polyacrylamide gel electro-phoresis. Protein concentrations were determined using Bradford’s pro-tein assay (14) (Bio-Rad) with bovine serum albumin as standard.Sequential Edman degradation was performed on a Procise 492 proteinsequencer (Perkin Elmer).

Glycosphingolipid Binding Assays—Glycosphingolipids were iso-

* This work was supported by Swedish Medical Research CouncilGrants 12628, 3967, and 10435; Swedish Technical Research CouncilGrant 97-296; the Swedish Cancer Foundation; and the WallenbergFoundation. The costs of publication of this article were defrayed in partby the payment of page charges. This article must therefore be herebymarked “advertisement” in accordance with 18 U.S.C. Section 1734solely to indicate this fact.

¶ To whom correspondence should be addressed: Inst. of MedicalBiochemistry, Goteborg University, P. O. Box 440, SE 405 30 Goteborg,Sweden. Tel.: 46-31-773-34-92; Fax: 46-31-413-190; E-mail: Susann.Teneberg@medkem.gu.se.

1 The abbreviations used are: CT, cholera toxin; CTB, cholera toxinB-subunit; Fuc, fucose; Hex, hexose; HexNAc, N-acetylhexosamine; LT,E. coli heat-labile enterotoxin; LTB, E. coli heat-labile enterotoxinB-subunit; HPLC, high performance liquid chromatography. The glyco-sphingolipid nomenclature follows the recommendations by the IUPAC-IUB Commission on Biochemical Nomenclature (CBN) (CBN for Lipids(1977) Eur. J. Biochem. 79, 11–21; CBN for Lipids (1982) J. Biol. Chem.257, 3347–3351; CBN for Lipids (1987) J. Biol. Chem. 262, 13–18). It isassumed that Gal, Glc, GlcNAc, GalNAc, NeuAc, and NeuGc are of the

D-configuration, Fuc of the L-configuration, and all sugars present in thepyranose form.

THE JOURNAL OF BIOLOGICAL CHEMISTRY Vol. 275, No. 5, Issue of February 4, pp. 3231–3238, 2000© 2000 by The American Society for Biochemistry and Molecular Biology, Inc. Printed in U.S.A.

This paper is available on line at http://www.jbc.org 3231

by guest on February 2, 2018http://w

ww

.jbc.org/D

ownloaded from

lated and characterized by mass spectrometry, 1H NMR, and degrada-tion studies, as outlined in (8, 10).

Mixtures of glycosphingolipids (20–40 mg/lane) or pure compounds(0.1–4 mg/lane) were separated on aluminum-backed silica gel 60 high-performance thin-layer chromatography plates (Merck), using chloro-form/methanol/water (60:35:8, by volume) as solvent. Chemical detec-tion was done with anisaldehyde (15).

Binding of B-subunits to glycosphingolipids on thin-layer chromato-grams or adsorbed in microtiter wells were performed as described (8,10), using 125I-labeled B-subunits diluted in phosphate-buffered saline,pH 7.2, containing 2% (w/v) bovine serum albumin and 0.1% (w/v)NaN3, to approximately 5 3 106 cpm/ml.

Chromatogram binding assays with monoclonal antibodies directedagainst blood group A, B, and H determinants (Dakopatts a/s, Glostrup,Denmark) were done as described (16) using 125I-labeled anti-mouseantibodies for detection.

Isolation of an LCTBK-binding Non-acid Glycosphingolipid fromEpithelial Cells of Cat Small Intestine—A total non-acid glycosphingo-lipid fraction (64 mg) was obtained from pooled epithelial cell scrapingsfrom the small intestines of 12 cats by standard methods (17). Thenon-acid glycosphingolipids (50 mg) were first separated on a silicic acidcolumn, stepwise eluted with increasing amounts of methanol in chlo-roform. The fractions containing triglycosylceramides and larger glyco-sphingolipids were pooled, giving 45 mg, and further separated byHPLC on a Kromasil 5 Silica column (2.12 3 25 cm, inner diameter;particle size, 5 mm; Skandinaviska Genetec, Kungsbacka, Sweden)eluted with a linear gradient of chloroform/methanol/water 80:20:1 to40:40:12 (by volume) during 180 min with a flow rate of 4 ml/min. Each4-ml fraction was analyzed by thin-layer chromatography using anisal-dehyde for detection. The fractions containing tetraglycosylceramidesand larger glycosphingolipids were tested for binding of LCTBK usingthe chromatogram binding assay. The binding-active compound elutedin tubes 151–158, and after pooling of these fractions 0.4 mg wasobtained.

Mass Spectrometry—For electron ionization mass spectrometry, ali-quots of the isolated glycosphingolipid were permethylated (18), orpermethylated and reduced with LiAlH4 (19). The samples were ana-lyzed on a JEOL SX-102A mass spectrometer (JEOL, Tokyo, Japan)using the in-beam technique (20). The analyses of both derivatives wereperformed with an electron energy of 70 eV, trap current of 300 mA, andacceleration voltage of 10 kV. The temperature was raised from 150 °Cto 410 °C, by increases of 10 °C/min.

Proton NMR Spectroscopy—1H NMR spectra were obtained on aVarian 500-MHz spectrometer at 30 °C. Samples were dissolved indimethyl sulfoxide-d6/D2O (98/2, by volume) after deuterium exchange.

Molecular Modeling and Dynamics Simulations—Docking and mo-lecular dynamics simulations were conducted on a Silicon GraphicsIndigo2Extreme workstation using the Quanta97/CHARMm22 softwarepackage (Molecular Simulations Inc., Waltham, MA), whereas the Apentasaccharide (GalNAca3(Fuca2)Galb4(Fuca3)GlcNAcb) initiallywas constructed using Biograf software package (Molecular Simula-tions Inc.) before transferring the structure to the aforementioned pro-gram. The glycosidic dihedral angles of the CHARMm-refined structuredeviated insignificantly from literature values (21). All protein struc-tures were constructed from the crystal structure of the pLT-lactosecomplex (Ref. 4; Protein Data Bank entry 1LTT): two whole B-subunits(G and H) were thus used in the construction of corresponding subunitsof hLTB and the hybrid mutants listed in Table I following proceduresoutlined earlier (22). The root-mean-square deviations for the mainchain atoms of energy-refined hLT, LCTBK, and LCTBK in complexwith the A pentasaccharide (snapshot at 180 ps) relative to the pLTcrystal structure were 0.55, 0.86, and 0.94 Å, respectively. A compari-son between LCTBK and its complex with the A pentasaccharideyielded root-mean-square values of 0.38 Å for the backbone atoms and0.62 Å when all atoms were considered. Of the several starting struc-tures for subsequent dynamics runs that were generated through man-ual docking of the A pentasaccharide into the putative LCTBK bindingsite, only a few very similar orientations of the pentasaccharide gave asatisfactory surface complementarity as well as involving the aminoacid side chains implicated in the binding studies. Vacuum dynamicssimulations at 300 K using a distance-dependent dielectric constant(e 5 6r) and a 2-fs time step were performed as described (22). The final1-ns run lasted for 380 CPU h. In these runs, 70 residues surroundingthe binding site were allowed freedom of movement, whereas in energyminimization the whole dimer was movable.

RESULTS

Generation and Characterization of Hybrid CT/hLT B-sub-units—The amino acid sequences of the B-subunits utilized inthis study are summarized in Table I. The daughter hybriddesignated LCTBK was obtained by re-substituting from Ser(found in hLTB) to Asn (found in CTB) at position 4. SDS-polyacrylamide gel electrophoresis analysis of LCTBK indi-cated that it formed stable pentamers. It also appeared identi-cal to LCTBH in terms of reactivity with CTB- and LTB-specificmonoclonal antibodies (23). However, identification of the 14N-terminal amino acids of LCTBK by sequential Edman deg-radation confirmed the presence of Asn at position 4.

Binding to GM1 Ganglioside and N-Acetyllactosamine-termi-nated Glycosphingolipids in Microtiter Wells—The microtiterwell assay showed that all B-pentamers bound to GM1 withsimilar affinities (Fig. 1, A), demonstrating that the mutationsintroduced had not affected the ability to interact with the GM1ganglioside. Half-maximal binding of all B-subunit proteinsoccurred at approximately 5 pmol/well.

As reported previously hLTB and LCTBH, but not CTB,bound to the N-acetyllactosamine-terminated glycosphingo-lipids neolactotetraosylceramide (Galb4GlcNAcb3Galb4Glcb1Cer; Fig. 1B) (12). Linear neolactohexaosylceramide (Galb4GlcNAcb3Galb4GlcNAcb3Galb4Glcb1Cer; data not shown)and branched neolactohexaosylceramide (Galb4GlcNAcb6(Galb4GlcNAcb3)Galb4Glcb1Cer; Fig. 1C) were also bound byhLTB and LCTBH. However, the level of binding of LCTBK toN-acetyllactosamine-terminated compounds was reduced vir-tually to the level of CTB. Furthermore, while hLTB andLCTBH bound to gangliotetraosylceramide (Galb3GalNAcb4Galb4Glcb1Cer), no binding of CTB or LCTBK to thisglycosphingolipid was observed (data not reproduced).

Binding to Glycosphingolipids on Thin-layer Chromato-grams—In accordance with the results from the microtiter wellassay, the hybrid B-subunits bound to the GM1 ganglioside onthin-layer chromatograms (Fig. 2, lane 9, and present in lane8). By binding of B-pentamers to glycosphingolipids from smallintestinal epithelium of single human individuals on thin-layerchromatograms, an aberrant behavior of LCTBK was detected.Unlike the other B-pentamers, LCTBK bound selectively toslowly migrating non-acid glycosphingolipids present in some

TABLE IAmino acid sequences of the recombinant CTB, human LTB, and

hybrid B-subunits

Positiona rCTBb hLTBc LCTBH LCTBK

1 Ala Ala Ala Ala4 Asn Ser Ser Asnd

7 Asp Glu Glu Glu10 Ala Ser Ser Ser18 His Tyr Tyr Tyr20 Leu Ile Ile Ile25 Phe Leu Leu Leu31 Leu Met Leu Leu38 Ala Val Ala Ala44 Asn Ser Asn Asn75 Ala Thr Ala Ala80 Ala Thr Ala Ala82 Val Ile Val Val83 Glu Asp Glu Glu94 His Asn Asn Asn95 Ala Ser Ser Ser

102 Ala Glu Ala Alaa Only the amino acids which are non-homologous between CTB and

LTB are shown.b The recombinant CTB has an alanine instead of a threonine at

position 1 (39).c The sequence of human LTB is from strain H74–114 (40).d The CTB amino acids in the hybrids LCTBH and LCTBK are in

bold.

CTB/LTB Hybrid Binding to Blood Group A and B Determinants3232

by guest on February 2, 2018http://w

ww

.jbc.org/D

ownloaded from

individuals (Fig. 2C, lanes 3 and 4). LCTBK-specific binding toslowly migrating glycosphingolipids was also detected in thenon-acid fractions of rabbit (Fig. 2C, lane 1), rat, dog, pig, andcat intestine and human meconium (data not shown).

Isolation and Characterization of an LCTBK-binding Glyco-sphingolipid from Epithelial Cells of Cat Small Intestine—Thenon-acid glycosphingolipid fraction from epithelial cells of catsmall intestine was separated by chromatography on a silicicacid column, followed by HPLC on straight-phase silica gel.The fractions containing the LCTBK-binding compound werepooled, giving 0.4 mg, which was used for structural character-ization by electron ionization mass spectrometry and protonNMR spectroscopy.

The mass spectrum of the permethylated LCTBK-bindingglycosphingolipid isolated from cat intestinal epithelial cells(Fig. 3A) has ions characteristic of a terminal blood group Adeterminant: terminal HexNAc (m/z 260), terminal fucose(m/z 189), and A trisaccharide (HexNAc(Fuc)Hex; m/z 638 and606). The next ions that can be attributed to the carbohydratechain are seen at m/z 1056 and 1024, indicating a HexNAc-(Fuc)Hex(Fuc)HexNAc pentasaccharide. The next carbohy-drate units toward the reducing end is a Hex (hexasaccharideions at m/z 1261 and 1229), followed by a HexNAc (heptasac-charide ions at m/z 1506 and 1474), a Hex (octasaccharide ionsat m/z 1710 and 1678), and a Hex (nonasaccharide ions at m/z1914 and 1882). The concluded sequence was supported by theion at m/z 1988, containing the whole carbohydrate chain andpart of the fatty acid, and by molecular ions at m/z 2596, 2624and 2652 (nonasaccharide with phytosphingosine and hydroxy20:0, 22:0 and 24:0 fatty acids, respectively).

The major long chain base is phytosphingosine (m/z 396),and the ceramide composition is given by the ions at m/z 694and 722, indicating phytosphingosine combined with hydroxy22:0 and 24:0 fatty acids, respectively.

The mass spectrum of the permethylated and reduced glyco-sphingolipid (Fig. 3B) has a series of prominent immonium ions(F-fragments), formed by loss of part of the long chain base, atm/z 2200–2312. These ions give information about the numberand type of sugars, and the fatty acid composition, and in thepresent case demonstrate the presence of a saccharide com-posed of two fucoses, three N-acetylhexosamines, and four hex-oses, combined with hydroxy 16:0 to 24:0 fatty acids. Carbohy-drate sequence ions are found at m/z 189 (terminal fucose),m/z 246 (terminal HexNAc), m/z 624 (terminal A trisaccha-ride), m/z 1028 (HexNAc(Fuc)Hex(Fuc)HexNAc pentasaccha-ride), m/z 1233 and 1201 (hexasaccharide), and m/z 1668(octasaccharide).

A prominent ion at m/z 182 is present in the mass spectra ofboth derivatives. This ion is due to rearrangement of GlcNAc,and is only found in spectra of glycosphingolipids with a type 2core (Hexb4HexNAc) (24, 25). The proposed carbohydrate se-quence has HexHexNAc at two positions, either of which couldhave a type 2 core. However, further information may be ob-tained from the spectrum of the permethylated and reducedderivative. In glycosphingolipids having a Hex in 1–3 linkage toa HexNAc (type 1 core), a series of rearrangement ions derivedfrom the immonium ions are found (26). These ions are ob-tained by fragmentation of the Hex1–3HexNAc linkage andrearrangement of the GlcNAc with loss of the acetamido group

FIG. 1. Intact GM1 ganglioside binding capacity, but reducedbinding of N-acetyllactosamine-terminated glycosphingolipidsby the hybrid B-subunit LCTBK, demonstrated by binding of125I-labeled B-subunits to serial dilutions of glycosphingolipidsin microtiter wells. Data are expressed as mean values of triplicatedeterminations.

FIG. 2. Selective binding of the hybrid B-subunit LCTBK toslowly migrating non-acid glycosphingolipids of human andrabbit small intestinal epithelium. Glycosphingolipids were chro-matographed on aluminum-backed silica gel plates using chloroform/methanol/water (60:35:8, by volume) as solvent system, and visualizedwith anisaldehyde (A). Duplicate chromatograms were incubated with125I-labeled LCTBH (B) and LCTBK (C), followed by autoradiographyfor 12 h, as described under “Experimental Procedures.” The laneswere: non-acid glycosphingolipids of rabbit small intestinal epithelium,40 mg (lane 1); non-acid glycosphingolipids of human small intestinalepithelium of three separate individuals, 40 mg/lane (lanes 2–4); acidglycosphingolipids of human small intestinal epithelium, 40 mg (lane 5);GM1 ganglioside, 0.4 mg (lane 6). The arrow marks the migration ofresidual GM1 ganglioside in the samples.

CTB/LTB Hybrid Binding to Blood Group A and B Determinants 3233

by guest on February 2, 2018http://w

ww

.jbc.org/D

ownloaded from

of 58 mass units. A type 1 linkage at the non-reducing Hex-HexNAc would give rise to a series of ions at m/z 1502–1614(F 2 640 2 58), while a type 1 linkage of the HexHexNAc closeto the reducing end would produce a series at m/z 893–1005(F 2 1249 2 58). The absence of both ion series thus suggeststhat the LCTBK-binding glycosphingolipid has type 2 corechains at both positions.

Thus, by mass spectrometry, the LCTBK-binding glycosphin-golipid was tentatively identified as a nonaglycosylceramidewith a terminal A trisaccharide and a HexNAc(Fuc)Hex(Fu-c)HexNAcHexHexNAcHexHex sequence, with two type 2 link-ages (Hexb4HexNAc).

The 1H NMR spectrum at 30 °C of the fraction containing thenonaglycosylceramide with a terminal A trisaccharide (notshown) displayed anomeric proton signals, which could easilybe assigned on the basis of earlier published spectra. Thus, thesignals at 5.10 ppm (Fuca2), 5.08 ppm (GalNAca3), 4.86 ppm

(Fuca3), 4.66 ppm (GlcNAcb3), and 4.45 ppm (Galb4) confirmthe presence of an ALey determinant as in the A7 type 2glycosphingolipid (27) whereas the signals at 4.26, 4.72, 4.28,and 4.22 ppm are consistent with the internal sequenceGalb4GlcNAcb3Galb4Glcb1 (28, 29). Combined with the datafrom mass spectrometry the identity of the LCTBK-bindingglycosphingolipid can thus be established as GalNAca3(Fuca2)Galb4(Fuca3)GlcNAcb3Galb4GlcNAcb3Galb4Glcb1Cer,i.e. the A9 type 2 glycosphingolipid.

Binding of B-subunits to Blood Group-active Glycosphingo-lipids in Microtiter Wells—Binding to the A9 type 2 glycosphin-golipid from cat small intestine in microtiter wells (Fig. 4A)confirmed that this glycosphingolipid was preferentially recog-nized by LCTBK, with a half-maximal binding at approxi-mately 20 pmol/well.

Next, the binding of LCTBK to a number of glycosphingolip-ids with structures related to the A9 type 2 glycosphingolipidwas tested in microtiter wells (summarized in Table II).LCTBK, but not the other B-subunits, bound to the A7 type 2glycosphingolipid (GalNAca3(Fuca2)Galb4(Fuca3)GlcNAcb3Galb4Glcb1Cer) with a half-maximal binding at approximately40 pmol/well (Fig. 4B). No binding of LCTBK to the A7 type 1glycosphingolipid (GalNAca3(Fuca2)Galb3(Fuca4)GlcNAcb3Galb4Glcb1Cer), the A6 type 2 glycosphingolipid (Gal

FIG. 3. Electron ionization mass spectra of the permethylated(A), and permethylated and LiAlH4-reduced (B), LCTBK-bindingglycosphingolipid isolated from the epithelial cells of cat smallintestine. Above the spectra are simplified interpretation formulaerepresenting the species with phytosphingosine and hydroxy 22:0 fattyacid. The analytical conditions were: electron energy, 70 eV; trap cur-rent, 300 mA; and acceleration voltage, 10 kV. The temperature wasraised from 150 °C to 410 °C, by increases of 10 °C/min. Both spectrawere recorded at 380 °C.

FIG. 4. Selective binding of LCTBK to glycosphingolipids withterminal GalNAca3(Fuca2)Galb4(Fuca3)GlcNAcb sequences. A,binding of LCTBK, but not CTB, hLTB or LCTBH, to the A9 type 2glycosphingolipid (GalNAca3(Fuca2)Galb4(Fuca3)GlcNAcb3Galb4GlcNAcb3Galb4Glcb1Cer) adsorbed in microtiter wells. B, bindingLCTBK to the A7 type 2 glycosphingolipid, but not to glycosphingolipidswith related structures in microtiter wells. Open circles,GalNAca3(Fuca2)Galb4(Fuca3)GlcNAcb3Galb4Glcb1Cer (A7–2); opensquares, GalNAca3(Fuca2)Galb3(Fuca4)GlcNAcb3Galb4Glcb1Cer(A7–1); filled circles, GalNAca3(Fuca2)Galb4GlcNAcb3Galb4Glcb1Cer(A6–2); filled squares, Fuca2Galb4(Fuca3)GlcNAcb3Galb4Glcb1Cer(Y-6). Data are expressed as mean values of triplicate determinations.

CTB/LTB Hybrid Binding to Blood Group A and B Determinants3234

by guest on February 2, 2018http://w

ww

.jbc.org/D

ownloaded from

NAca3(Fuca2)Galb4GlcNAcb3Galb4Glcb1Cer), or the Y6 gly-cosphingolipid (Fuca2Galb4(Fuca3)GlcNAcb3Galb4Glcb1Cer)was obtained. Furthermore, the binding of LCTBK to the B7type 2 glycosphingolipid (Gala3(Fuca2)Galb4(Fuca3)GlcNAcb3Galb4Glcb1Cer of human erythrocytes (Fig. 5, lane 2) indi-cated that the acetamido group of the terminal GalNAc was notessential for the interaction, and thus the minimal structuralelement involved in the recognition process was Gala3(Fuca2)Galb4(Fuca3)GlcNAcb.

Molecular Modeling and Dynamics Simulations—Inspectionof Table I and the pLT/CT crystal structures (4–6) show thatthe non-conserved amino acids at positions 7, 25, 83, and 102are in close proximity to Asn4 of LCTBK, strongly suggesting abinding site location within this perimeter. Docking a bloodgroup A pentasaccharide (GalNAca3(Fuca2)Galb4(Fuca3)GlcNAcb) into the LCTBK hybrid and ensuing molecular dy-namics simulations (1 ns) of the complex yields the pictureshown in Figs. 6–9. The pentasaccharide lies in a shallowdepression of the protein surface at the subunit interfaces withthe fucoses exposed to the solvent. In addition to an excellentprotein-saccharide surface complementarity, critical hydrogenbond interactions with the side chains of Gln3, Asn4, Ser26,Thr28, Thr41, Thr47, Glu83, and Lys84 as well as with the pep-

tide backbone are found involving all five sugars (Figs. 7 and 9).The electrostatic potential energy surfaces for the LCTBKbinding site (Fig. 7), generated by a water probe, and thecorresponding surface for the pentasaccharide (partially shownin Fig. 8), reveal several regions of advantageous complemen-tary potentials of opposite sign. This is particularly evident forthe interactions of the GalNAca3 and Fuca3 residues, whichare essential for binding to occur (see Table II). Due to its rigidnature (21), conformational changes of the pentasaccharide arevery minor. Protein conformational changes are small as well,being restricted to re-orientation of the Leu25 and Glu83 sidechains in order to accommodate the pentasaccharide. Otherchanges include movement of Ser26 approximately 1.2 Å to-ward Galb4 and an average downward movement of the 42–46loop by 2 Å compared with the pLT crystal structure (4).

The significance of the Ser4 3 Asn mutation in LCTBK lies

FIG. 5. Recognition of blood group A- and B-active heptagly-cosylceramides and larger glycosphingolipids by LCTBK. Glyco-sphingolipids were chromatographed on aluminum-backed silica gelplates using chloroform/methanol/water (60:35:8, by volume) as solventsystem, and visualized with anisaldehyde (A). Duplicate chromato-grams were incubated with 125I-labeled LCTBK (B), hLTB (C), andmonoclonal antibodies directed against the blood group A determinant(D) and the blood group B determinant (E), followed by autoradiographyfor 12 h, as described under “Experimental Procedures.” The laneswere: non-acid glycosphingolipids of human blood group A erythrocytes,40 mg (lane 1); non-acid glycosphingolipids of human blood group Berythrocytes, 40 mg (lane 2); non-acid glycosphingolipids of human bloodgroup O erythrocytes, 40 mg (lane 3); neolactotetraosylceramide(Galb4GlcNAcb3Galb4Glcb1Cer), 1 mg (lane 4); A7 type 2 glycosphin-golipid (GalNAca3(Fuca2)Galb4(Fuca3)GlcNAcb3Galb4Glcb1Cer), 1mg (lane 5); GM1 ganglioside (Galb3GalNAcb4(NeuAca3)Galb4Glcb1Cer),0.5 mg (lane 6). The numbers to the left of A indicate the approxi-mate number of the carbohydrate units in the bands.

FIG. 6. Ribbon representation of two subunits of the LCTBKB-pentamer showing the location of the blood group A pentasac-charide binding site at the subunit interface (top center) rela-tive to the classic GM1 pentasaccharide binding site (bottomright). The B-dimer is shown with the top slightly tilted toward theviewer and with the a-helices lining the central pore of the B-pentamerfacing away from the viewer. The terminal GalNAc residue of the Apentasaccharide is pointing to the right, whereas the GlcNAc at thereducing end is seen to point to the left. The terminal Gal residue of theGM1 oligosaccharide is seen at the top whereas the sialic acid is seenpointing to the left.

TABLE IISummary of results from glycosphingolipid binding assays

No. Trivial name StructureBinding

CTB hLTB LCTBH LCTBK

1. GM1 Galb3GalNAcb4(NeuAca3)Galb4Glcb1Cer 1 1 1 12. LacCer Galb4Glcb1Cer 2 2 2 23. Isoglobotri Gala3Galb4Glcb1Cer 2 2 2 24. Gangliotetra Galb3GalNAcb4Galb4Glcb1Cer 2 1 1 25. Neolactotetra Galb4GlcNAcb3Galb4Glcb1Cer 2 1 1 26. H5–2 Fuca2Galb4GlcNAcb3Galb4Glcb1Cer 2 2 2 27. B5 Gala3Galb4GlcNAcb3Galb4Glcb1Cer 2 2 2 28. Forssman GalNAca3GalNAcb3Gala4Galb4Glcb1Cer 2 2 2 29. Neolactohexaa Galb4GlcNAcb3Galb4GlcNAcb3Galb4Glcb1Cer 2 1 1 2

10. A6–2 GalNAca3(Fuca2)Galb4GlcNAcb3Galb4Glcb1Cer 2 2 2 211. B6–2 Gala3(Fuca2)Galb4GlcNAcb3Galb4Glcb1Cer 2 2 2 212. Y-6 Fuca2Galb4(Fuca3)GlcNAcb3Galb4Glcb1Cer 2 2 2 213. NeuGc-neolactohexa NeuGca3Galb4GlcNAcb3Galb4GlcNAcb3Galb4Glcb1Cer 2 2 2 214. A7-2 GalNAca3(Fuca2)Galb4(Fuca3)GlcNAcb3Galb4Glcb1Cer 2 2 2 115. A7-1 GalNAca3(Fuca2)Galb3(Fuca4)GlcNAcb3Galb4Glcb1Cer 2 2 2 216. B7-2 Gala3(Fuca2)Galb3(Fuca3)GlcNAcb3Galb4Glcb1Cer 2 2 2 117. B7-1 Gala3(Fuca2)Galb3(Fuca4)GlcNAcb3Galb4Glcb1Cer 2 2 2 218. A9-2 GalNAca3(Fuca2)Galb4(Fuca3)GlcNAcb3Galb4GlcNAcb3Galb-

4Glcb1Cer2 2 2 1

a Glycosphingolipid no. 9 was produced from no. 13 by mild acid hydrolysis (1% acetic acid at 100 °C for 1 h).

CTB/LTB Hybrid Binding to Blood Group A and B Determinants 3235

by guest on February 2, 2018http://w

ww

.jbc.org/D

ownloaded from

partly in the additional hydrogen bonds formed between theAsn4 side chain and the Galb4 4-OH and the Fuca3 3-OH, andpartly in the very snug fit of this side chain against the carbo-hydrate surface (Fig. 8). The orientation of the amide of theAsn4 side chain differs by 180° from that observed in CTB (5, 6).This is due to the nearby Glu7 in LCTBK, as opposed to theshorter Asp7 in CTB, which favors the opposite configuration.This also results in a preference of the hydroxyl group of Thr6

side chain to point downward to hydrogen-bond to Asn4 (Figs.7 and 8).

Another significant difference between LCTBK and CTB isLeu25, which in CTB is a phenylalanine. In CTB Phe25 forms ahydrophobic patch with the methyl group of Thr41 as doesLeu25 of pLT (4). However, in LCTBK Leu25 is too far fromThr41 due to a re-orientation of the Leu25 side chain thatappears necessary when the pentasaccharide complex isformed. In CTB, Phe25 is locked in the crystal conformation,suggesting that its side chain would partially block access tothe binding site by sterically interfering with the Fuca2 residueand also unfavorably interact with the 6-OH of the GlcNAcb

FIG. 7. Stereo views of the blood group A pentasaccharide and its interactions with surrounding LCTBK amino acid side chains.Both panels were generated from an energy-refined snapshot 180 ps into a 1-ns molecular dynamics run. In the upper panel the backbones of thetwo different subunits are colored red and blue, respectively, whereas the side chains are in light blue. The a carbon atoms of several amino acidsare numbered as they appear in the sequence and can be identified from Fig. 9. The blood group A pentasaccharide is shown in yellow with theterminal GalNAc residue at the top and the GlcNAc at the reducing end at the bottom. Hydrogen bonds are shown as white dashed lines. Theglycosidic dihedral angles (F, C) of the pentasaccharide differed insignificantly from those of the isolated pentasaccharide. The lower panel showsthe electrostatic potential energy surface of LCTBK in the blood group A pentasaccharide binding site, generated by a water probe with a 1.4-Åradius, where blue represents the most negative potential and red the most positive one (630 kcal/mol). The complementary surface generated forthe pentasaccharide is almost identical in shape but reveals potentials of opposite sign in several significant parts of the surface.

CTB/LTB Hybrid Binding to Blood Group A and B Determinants3236

by guest on February 2, 2018http://w

ww

.jbc.org/D

ownloaded from

residue. The combination of these factors and the unfavorableorientation of Asn4 probably account for the absence of bindingof CTB to the blood group A/B determinant.

The model also accounts for the relatively weak binding ofA9–2 by hLTB and LCTBH since the Ser4 side chain present inthese molecules is unable to provide the favorable carbohydrateinteractions observed for the Asn4 side chain of LCTBK.

DISCUSSION

The binding of the B-subunits of CT and LT to receptorglycoconjugates on the small intestinal epithelial cells is a

prerequisite for the following steps in toxin action leading todiarrhea. In addition, both CT and LT elicit strong immuneresponses, and are among the most potent mucosal adjuvantsyet identified (30, 31). The immunogenicity, and to some extentthe adjuvant activity, are also dependent on receptor binding,as shown by recent studies using the Gly33 [arrow] Asp mutantof LT, which is devoid of GM1 binding capacity (32–34).

The binding of the B-subunits of cholera toxin to the GM1ganglioside is a paradigm for protein-carbohydrate interac-tions. However, in the CTB/LTB hybrid LCTBK, with bloodgroup A- and B-binding capacity, the blood group determinantsare accommodated in a novel carbohydrate binding site, dis-tinct from the GM1 binding site. This novel binding site islocated at the top of the B-subunit interfaces as found bydocking and molecular dynamics simulations. Changes of car-bohydrate binding specificities by substitutions of specificamino acids within the carbohydrate binding sites have previ-ously been reported for E-selectin, P-selectin, and mannose-binding protein (35–37), but this is the first report of thecreation of a novel binding site in a carbohydrate-binding pro-tein. The reason for the lost ability of LCTBK to bind to N-acetyllactosamine-terminated glycosphingolipids and gan-gliotetraosylceramide is, however, at present not apparent.

An additional observation is that also CTB exhibits a weakbinding to branched neolactohexaosylceramide, but does notbind to linear to N-acetyllactosamine-terminated glycosphingo-lipids (Fig. 1). Docking studies and molecular dynamics simu-lations suggest that this is due to additional interactions be-tween the the b6-linked branch and amino acid residuesoutside the GM1 binding site, i.e. van der Waals interactionsbetween the -CbH2 group of His13 and the terminal galactose,and between Asn14 and the 3-OH of this galactose, in additionto the interactions described earlier for neolactotetraosylcer-amide (10).2 The Galb4GlcNAcb6(Galb3GlcNAcb3)Galb ele-ment of branched neolactohexaosylceramide is found also inthe carbohydrate chains of glycoproteins. However, CTB does

2 S. Teneberg, A. Berntsson, and J. Ångstrom, manuscript inpreparation.

FIG. 8. Close-up view of the interac-tions of the Asn4 residue of LCTBKwith the blood group A pentasaccha-ride. The electrostatic potential energysurface of the pentasaccharide, generatedby a 1.4-Å water probe, is color-coded cor-responding to the LCTBK surface shownin Fig. 7, whereas the van der Waals sur-face of the amide moiety of Asn4 is coloredusing standard atom colors. Note thegroove, formed by the GalNAca3 acet-amido moiety, the Galb4 4-CH, 4-OH, and6-CH2 groups and the Fuca3 3-OH (thetwo latter groups are located just belowthe border of the figure), into which theamide moiety of Asn4 fits precisely. Forclarity, the contribution of the Asn4 CbH2group to the van der Waals surface wasomitted but makes the side chain fit eventighter.

FIG. 9. Schematic view of the blood group A pentasaccharideand its interactions with surrounding amino acid side chains ofLCTBK as found from molecular dynamics simulations. Hydro-gen bonds are indicated by dashed arrows.

CTB/LTB Hybrid Binding to Blood Group A and B Determinants 3237

by guest on February 2, 2018http://w

ww

.jbc.org/D

ownloaded from

not bind to human small intestinal glycoproteins on blottingmembranes, and hLTB binds only after de-sialylation (38),indicating that certain substitutions on the Galb4GlcNAcb6(Galb3GlcNAcb3)Galb core obliterates the binding.

An important further step will be to analyze whether thenovel mode of binding of LCTBK, with recognition of carbohy-drate receptors different from the GM1 ganglioside, will allowthe toxin to exert its biological effects. The location of thebinding site and the direction of the carbohydrate residues atthe reducing end suggest that LCTBK may bind “upside down”with the GM1 binding sites directed away from the surface.The next step will thus be to produce an LCTBK holotoxin,where the presence of the A-subunit will prevent the upsidedown binding. This holotoxin should not bind to the relativelyshort glycosphingolipids tested here, but may bind to longerglycosphingolipids or A/B determinants on glycoproteins, andwill be a novel tool for dissection of the relation of the bindingevent to the biological functions of the toxins. Thereafter, aholotoxin with an LCTBK/G33D mutation in the B-subunitswill be constructed, giving a toxin that binds only to bloodgroup A and B determinants. This hybrid toxin will allowfurther insights into how the biological activities of the toxinsare related to recognition of different carbohydrate receptors.

In addition, efforts to crystallize LCTBK, alone and in com-plex with A pentasaccharide, are currently under way.

Acknowledgment—We gratefully acknowledge the use of the Varian500-MHz machine at the Swedish NMR Center, Hasselblad Laboratory,Goteborg University.

REFERENCES

1. Black, R. E. (1985) in Proceedings of the 11th Nobel Conference (Holmgren, J.,Lindberg, A., and Mollby, R., eds) pp. 23–32, Studentlitteratur, Lund,Sweden

2. Spangler, B. D. (1992) Microbiol. Rev. 56, 622–6473. Domenighini, M., Pizza, M., Jobling, M. G., Holmes, R. K., and Rappuoli, R.

(1995) Mol. Microbiol. 15, 1165–11674. Sixma, T. K., Pronk, S. E., Kalk, K. H., van Zanten, B. A. M., Berghuis, A. M.,

and Hol, W. G. J. (1992) Nature 355, 561–5645. Merritt, E. A., Sarfaty, S., van den Akker, F., L’Hoir, C., Martial, J. A., and

Hol, W. G. J. (1994) Protein Sci. 3, 166–1756. Merritt, E. A., Kuhn, P., Sarfarty, S., Erbe, J. L., Holmes, R. K., and Hol,

W. G. J. (1998) J. Mol. Biol. 282, 1043–10597. Holmgren, J. (1973) Infect. Immun. 10, 851–8598. Ångstrom, J., Teneberg, S., and Karlsson, K.-A. (1994) Proc. Natl. Acad. Sci.

U. S. A. 91, 11859–118639. Orlandi, P. A., Crithley, D. R., and Fishman, P. H. (1994) Biochemistry 33,

12886–1289510. Teneberg, S., Hirst, T. R., Ångstrom, J., and Karlsson, K.-A. (1994) Glycoconj.

J. 11, 533–540

11. Fukuta, S., Magnani, J. L., Twiddy, E. M., Holmes, R. K., and Ginsburg, V.(1988) Infect. Immun. 56, 1748–1753

12. Backstrom, M., Shahabi, V., Johansson, S., Teneberg, S., Kjellberg, A., Miller-Podraza, H., Holmgren, J., and Lebens, M. (1997) Mol. Microbiol. 24,489–497

13. Lebens, M., Johansson, S., Osek, J., Lindblad, M., and Holmgren, J. (1993)Bio/Technology 11, 1574–1578

14. Bradford, M. M. (1976) Anal. Biochem. 72, 248–25215. Waldi, D. (1962) in Dunnschicht-Chromatographie (Stahl, E., ed) pp. 495–515,

Springer-Verlag, Berlin16. Hansson, G. C., Karlsson, K.-A., Larson, G., McKibbin, J. M., Blaszczyk, M.,

Herlyn, M., Steplewski, Z., and Koprowski, H. (1983) J. Biol. Chem. 258,4091–4097

17. Karlsson, K.-A. (1987) Methods Enzymol. 138, 212–22018. Larson, G., Karlsson, H., Hansson, G. C., and Pimlott, W. (1987) Carbohydr.

Res. 161, 281–29019. Karlsson, K.-A. (1974) Biochemistry 13, 3643–364720. Breimer, M., Hansson, G. C., Karlsson, K.-A., Larson, G., Leffler, H., Pascher,

I., Pimlott, W., and Samuelsson, B. E. (1980) in Advances in Mass Spec-trometry (Quayle, A., ed) Vol. 8, pp. 1097–1108, Heyden & Son, London

21. Imberty, A., Mikros, E., Koca, J., Mollicone, R., Oriol, R., and Perez, S. (1995)Glycoconj. J. 12, 331–349

22. Moreno, E., Teneberg, S., Adar, R., Sharon, N., Karlsson, K.-A., and Ångstrom,J. (1997) Biochemistry 36, 4429–4437

23. Lebens, M., Shahabi, V., Backstrom, M., Houze, T., Lindblad, M., andHolmgren, J. (1996) Infect. Immun. 64, 2144–2150

24. Karlsson, K.-A. (1976) in Glycolipid Methodology (Witting, L. A., ed) pp.97–122, American Oil Society, Champaign, IL

25. Karlsson, K.-A. (1978) Prog. Chem. Fats Other Lipids 16, 207–23026. Karlsson, K.-A., and Larson, G. (1979) J. Biol. Chem. 254, 9311–931627. Elson, C. O. (1996) in Mucosal Vaccines (Kiyono, H., Ogra, P. L., and McGhee,

J. R., eds) pp. 59–72, Academic Press Inc., San Diego28. Levery, S. B., Nudelman, E. D., Andersen, N. H., and Hakomori, S. (1986)

Carbohydr. Res. 151, 311–32829. Clausen, H., Levery, S. B., McKibbin, J. M., and Hakomori, S. (1985) Biochem-

istry 24, 3578–358630. Hakomori, S., Nudelman, E., Levery, S. B., and Kannagi, R. (1984) J. Biol.

Chem. 259, 4672–468031. Dickinson, B. L. & Clements, J. D. (1996) in Mucosal Vaccines (Kiyono, H.,

Ogra, P. L., and McGhee, J. R., eds) pp. 73–87, Academic Press Inc., SanDiego

32. Nashar, T. O., Webb, H. M., Eaglestone, S., Williams, N. A., and Hirst, T. R.(1996) Proc. Natl. Acad. Sci. U. S. A. 93, 226–230

33. Guidry, J. J., Cardenas, L., Cheng, E., and Clements, J. D. (1997) Infect.Immun. 65, 4943–4950

34. de Haan, L., Verweij, W. R., Feil, I. K., Holtrop, M., Hol, W. G. J., Agsteribbe,E., and Wilschut, J. (1998) Immunology 94, 424–430

35. Kogan, T. P., Revelle, B. M., Tapp, S., Scott, D., and Beck, P. J. (1995) J. Biol.Chem. 270, 14047–14055

36. Revelle, B. M., Scott, D., Kogan, T. P., Zheng, J., and Beck, P. J. (1996) J. Biol.Chem. 271, 4289–4297

37. Blanck, O., Iobst, S. T., Gabel, C., and Drickamer, K. (1996) J. Biol. Chem. 271,7289–7292

38. Karlsson, K.-A., Teneberg, S., Ångstrom, J., Kjellberg, A., Hirst, T. R.,Bergstrom, J., and Miller-Podraza, H. (1996) Bioorg. Med. Chem. 4,1919–1928

39. Sanchez, J., and Holmgren, J. (1989) Proc. Natl. Acad. Sci. U. S. A. 86,481–485

40. Leong, J., Vinal, A. C., and Dallas, W. S. (1985) Infect. Immun. 48, 73–77

CTB/LTB Hybrid Binding to Blood Group A and B Determinants3238

by guest on February 2, 2018http://w

ww

.jbc.org/D

ownloaded from

Karl-Anders Karlsson, Michael Lebens and Susann TenebergJonas Ångström, Malin Bäckström, Anna Berntsson, Niclas Karlsson, Jan Holmgren,

B-subunits Heat-labile EnterotoxinEscherichia coliin a Hybrid of Cholera Toxin and

Novel Carbohydrate Binding Site Recognizing Blood Group A and B Determinants

doi: 10.1074/jbc.275.5.32312000, 275:3231-3238.J. Biol. Chem. 

  http://www.jbc.org/content/275/5/3231Access the most updated version of this article at

 Alerts:

  When a correction for this article is posted• 

When this article is cited• 

to choose from all of JBC's e-mail alertsClick here

  http://www.jbc.org/content/275/5/3231.full.html#ref-list-1

This article cites 35 references, 14 of which can be accessed free at

by guest on February 2, 2018http://w

ww

.jbc.org/D

ownloaded from