Ganglioside structure dictates signal transduction by cholera toxin and association with caveolae-like membrane domains in polarized epithelia.

In polarized cells, signal transduction by cholera toxin (CT) requires apical endocytosis and retrograde transport into Golgi cisternae and perhaps ER (Lencer, W.I., C. Constable, S. Moe, M. Jobling, H.M. Webb, S. Ruston, J.L. Madara, T. Hirst, and R. Holmes. 1995. J. Cell Biol. 131:951-962). In this study, we tested whether CT's apical membrane receptor ganglioside GM1 acts specifically in toxin action. To do so, we used CT and the related Escherichia coli heat-labile type II enterotoxin LTIIb. CT and LTIIb distinguish between gangliosides GM1 and GD1a at the cell surface by virtue of their dissimilar receptor-binding B subunits. The enzymatically active A subunits, however, are homologous. While both toxins bound specifically to human intestinal T84 cells (Kd approximately 5 nM), only CT elicited a cAMP-dependent Cl- secretory response. LTIIb, however, was more potent than CT in eliciting a cAMP-dependent response from mouse Y1 adrenal cells (toxic dose 10 vs. 300 pg/well). In T84 cells, CT fractionated with caveolae-like detergent-insoluble membranes, but LTIIb did not. To investigate further the relationship between the specificity of ganglioside binding and partitioning into detergent-insoluble membranes and signal transduction, CT and LTIIb chimeric toxins were prepared. Analysis of these chimeric toxins confirmed that toxin-induced signal transduction depended critically on the specificity of ganglioside structure. The mechanism(s) by which ganglioside GM1 functions in signal transduction likely depends on coupling CT with caveolae or caveolae-related membrane domains.

Targeting of cholera toxin and Escherichia coli heat labile toxin in polarized epithelia: role of COOH-terminal KDEL.

Vibrio cholerae and Escherichia coli heat labile toxins (CT and LT) elicit a secretory response from intestinal epithelia by binding apical receptors (ganglioside GM1) and subsequently activating basolateral effectors (adenylate cyclase). We have recently proposed that signal transduction in polarized cells may require transcytosis of toxin-containing membranes (Lencer, W. I., G. Strohmeier, S. Moe, S. L. Carlson, C. T. Constable, and J. L. Madara. 1995. Proc. Natl. Acad. Sci. USA. 92:10094-10098). Targeting of CT into this pathway depends initially on binding of toxin B subunits to GM1 at the cell surface. The anatomical compartments in which subsequent steps of CT processing occur are less clearly defined. However, the enzymatically active A subunit of CT contains the ER retention signal KDEL (RDEL in LT). Thus if the KDEL motif were required for normal CT trafficking, movement of CT from the Golgi to ER would be implied. To test this idea, recombinant wild-type (wt) and mutant CT and LT were prepared. The COOH-terminal KDEL sequence in CT was replaced by seven unrelated amino acids: LEDERAS. In LT, a single point mutation replacing leucine with valine in RDEL was made. Wt and mutant toxins displayed similar enzymatic activities and binding affinities to GM1 immobilized on plastic. Biologic activity of recombinant toxins was assessed as a Cl- secretory response elicited from the polarized human epithelial cell line T84 using standard electrophysiologic techniques. Mutations in K(R)DEL of both CT and LT delayed the time course of toxin-induced Cl- secretion. At T1/2, dose dependencies for K(R)DEL-mutant toxins were increased > or = 10-fold. KDEL-mutants displayed differentially greater temperature sensitivity. In direct concordance with a slower rate of signal transduction. KDEL-mutants were trafficked to the basolateral membrane more slowly than wt CT (assessed by selective cell surface biotinylation as transcytosis of B subunit). Mutation in K(R)DEL had no effect on the rate of toxin endocytosis. These data provide evidence that CT and LT interact directly with endogenous KDEL-receptors and imply that both toxins may require retrograde movement through Golgi cisternae and ER for efficient and maximal biologic activity.

Synthesis and secretion of the plasmid-coded heat-labile enterotoxin of Escherichia coli in Vibrio cholerae.

Both cholera toxin and heat-labile enterotoxin were made and secreted into culture supernatants by Vibrio cholerae containing the enterotoxin plasmid pCG86. Several regulatory mutations in V. cholerae that increased or decreased the synthesis of cholera toxin did not affect production of heat-labile enterotoxin. In contrast, a mutation in V. cholerae that interfered with the secretion of cholera toxin also decreased the secretion of heat-labile enterotoxin, indicating that they are processed by a common secretory pathway. Vibrio cholerae should be useful as a model system for analyzing the secretion of true extracellular proteins by Gram-negative bacteria.

Conjugal transfer of a chromosomal gene determining production of enterotoxin in vibrio cholerae.

Matings between strains of Vibrio cholerae differing in toxinogenicity, nutritional requirements, and antibiotic susceptibilities were performed in order to determine the location of the gene tox that controls production of cholera enterotoxin. Segregation analysis shows that tox is linked to a gene required for histidine biosynthesis. Our data indicate that the tox gene is located on the bacterial chromosome and not on a plasmid in the strains of V. cholerae studied.

Shiga-like toxin-converting phages from Escherichia coli strains that cause hemorrhagic colitis or infantile diarrhea.

Escherichia coli K-12 acquired the ability to produce a high titer of Shiga-like toxin after lysogenization by either of two different bacteriophages isolated from a highly toxinogenic Escherichia coli O157:H7 strain that causes hemorrhagic colitis. One of these phages and another Shiga-like toxin-converting phage from an Escherichia coli O26 isolate associated with infantile diarrhea were closely related in terms of morphology, virion polypeptides, DNA restriction fragments, lysogenic immunity, and heat stability, although a difference in host range was noted. These phages are currently the best-characterized representatives from a broader family of Shiga-like toxin-converting phages.

Cloning of Shiga-like toxin structural genes from a toxin converting phage of Escherichia coli.

The genes controlling high-level production of Shiga-like toxin (SLT) in Escherichia coli were cloned from the SLT converting phage 933J. This phage was isolated from a strain of E. coli that caused a foodborne outbreak of hemorrhagic colitis. The genes that convert normal E. coli to organisms producing high levels of toxin were cloned into the plasmid pBR328 and expressed in E. coli HB101. DNA restriction mapping, subcloning, examination of the cloned gene products by minicell analysis, neutralization, and immunoprecipitation with antibodies to SLT were used to localize the toxin converting genes and identify them as structural genes for SLT. Southern hybridization studies established that the DNA fragment carrying the cloned toxin structural genes had homology with the DNA of Shigella.

Studies on toxinogenesis in Vibrio cholerae. III. Characterization of nontoxinogenic mutants in vitro and in experimental animals.

Spontaneous and chemically induced mutants with reduced ability to produce cholera enterotoxin (choleragen) as an extracellular protein were isolated from Vibrio cholerae strains 569B Inaba, a classical cholera vibrio, and 3083-2 Ogawa, an El Tor vibrio. By qualitative and quantitative immunological assay in vitro such mutants could be separated into different classes characterized either by production of no detectable choleragen (tox minus), or of small quantities of extracellular choleragen, or of large quantities of cell-associated choleragen but little extracellular choleragen. Analysis of proteins in concentrated culture supernates by electrophoresis in polyacrylamide gels showed that cultures from tox minus strains lacked proteins with electrophoretic mobilities corresponding with choleragen or the spontaneously formed toxoid (choleragenoid). Infant rabbits infected with the tox minus strains remained asymptomatic or developed milder symptoms than rabbits infected with the tox+ parental strains. When symptoms of cholera developed after inoculation with tox minus mutants, detectable numbers of tox+ revertants could be isolated from the intestines of the infected animals. Two tox minus strains, designated M13 and M27, caused no sumptoms and showed no evidence of reversion to tox+ during single passage in infant rabbits, and mutant M13 also remained avirulent and stably tox minus during six cycles of serial passage in infant rabbits. Strains M13 and M27 were also noncholeragenic in acult rabbit ileal loops. Quantitative cultures of the intestines from infected infant rabbits demonstrated that the avirulent mutant M13 can multiply in vivo and can persist in the intestinal tract for at least 48 h.

Adherence of bacteria to heart valves in vitro.

The abilities of 14 strains of aerobic gram-positive cocci and gram-negative bacilli to adhere in vitro to human or canine aortic valve leaflets were compared. 2-mm sections of excised valve leaflets were obtained by punch biopsy and were incubated under standardized conditions in suspensions of bacteria. Valve sections were subsequently washed and homogenized, and quantitative techniques were used to determine the proportions of bacteria from the initial suspensions that had adhered to the valve sections. Comparable results were obtained when these adherence ratios were determined by two independent methods based either on measurements of bacterial viability or of radioactivity in 51Cr-labeled bacteria. For each bacterial strain, the adherence ratio was constant over a wide range of concentrations of bacteria in the incubation medium. Strains of enterococci, viridans streptococci, coagulase-positive and coagulase-negative staphylococci and Pseudomonas aeruginosa (adherence ratios 0.003-0.017) were found to adhere more readily to valve sections than strains of Escherichia coli and Klebsiella pneumoniae (adherence ratios 0.00002-0.00004). The organisms that most frequently cause bacterial endocarditis were found to adhere best to heart valves in vitro, suggesting that the ability to adhere to valvular endothelium may be an important or essential charcteristic of bacteria that cause endocarditis in man.

Activation of Escherichia coli heat-labile enterotoxins by native and recombinant adenosine diphosphate-ribosylation factors, 20-kD guanine nucleotide-binding proteins.

Escherichia coli heat-labile enterotoxins (LT) are responsible in part for "traveler's diarrhea" and related diarrheal illnesses. The family of LTs comprises two serogroups termed LT-I and LT-II; each serogroup includes two or more antigenic variants. The effects of LTs result from ADP ribosylation of Gs alpha, a stimulatory component of adenylyl cyclase; the mechanism of action is identical to that of cholera toxin (CT). The ADP-ribosyltransferase activity of CT is enhanced by 20-kD guanine nucleotide-binding proteins, known as ADP-ribosylation factors or ARFs. These proteins directly activate the CTA1 catalytic unit and stimulate its ADP ribosylation of Gs alpha, other proteins, and simple guanidino compounds (e.g., agmatine). Because of the similarities between CT and LTs, we investigated the effects of purified bovine brain ARF and a recombinant form of bovine ARF synthesized in Escherichia coli on LT activity. ARF enhanced the LT-I-, LT-IIa-, and LT-IIb-catalyzed ADP ribosylation of agmatine, as well as the auto-ADP ribosylation of the toxin catalytic unit. Stimulation of ADP-ribosylagmatine formation by LTs and CT in the presence of ARF was GTP dependent and enhanced by sodium dodecyl sulfate. With agmatine as substrate, LT-IIa and LT-IIb exhibited less than 1% the activity of CT and LT-Ih. CT and LTs catalyzed ADP-ribosyl-Gs alpha formation in a reaction dependent on ARF, GTP, and dimyristoyl phosphatidylcholine/cholate. With Gs alpha as substrate, the ADP-ribosyltransferase activities of the toxins were similar, although CT and LT-Ih appeared to be slightly more active than LT-IIa and LT-IIb. Thus, LT-IIa and LT-IIb appear to differ somewhat from CT and LT-Ih in substrate specificity. Responsiveness to stimulation by ARF, GTP, and phospholipid/detergent as well as the specificity of ADP-ribosyltransferase activity are functions of LTs from serogroups LT-I and LT-II that are shared with CT.

Three-dimensional structure of the diphtheria toxin repressor in complex with divalent cation co-repressors.

BACKGROUND: When Corynebacterium diphtheriae encounters an environment with a low concentration of iron ions, it initiates the synthesis of several virulence factors, including diphtheria toxin. The diphtheria toxin repressor (DtxR) plays a key role in this iron-dependent, global regulatory system and is the prototype for a new family of iron-dependent repressor proteins in Gram-positive bacteria. This study aimed to increase understanding of the general regulatory principles of cation binding to DtxR. RESULTS: The crystal structure of dimeric DtxR holo-repressor in complex with different transition metals shows that each subunit comprises an amino-terminal DNA-binding domain, an interface domain (which contains two metal-binding sites) and a third, very flexible carboxy-terminal domain. Each DNA-binding domain contains a helix-turn-helix motif and has a topology which is very similar to catabolite gene activator protein (CAP). Molecular modeling suggests that bound DNA adopts a bent conformation with helices alpha 3 of DtxR interacting with the major grooves. The two metal-binding sites lie approximately 10 A apart. Binding site 2 is positioned at a potential hinge region between the DNA-binding and interface domains. Residues 98-108 appear to be crucial for the functioning of the repressor; these provide four of the ligands of the two metal-binding sites and three residues at the other side of the helix which are at the heart of the dimer interface. CONCLUSIONS: The crystal structure of the DtxR holorepressor suggests that the divalent cation co-repressor controls motions of the DNA-binding domain. In this way the metal co-repressor governs the distance between operator recognition elements in the two subunits and, consequently, DNA recognition.

Surprising leads for a cholera toxin receptor-binding antagonist: crystallographic studies of CTB mutants.

BACKGROUND: Because agents which inhibit the receptor binding of cholera toxin constitute possible lead compounds for the structure-based design of anti-cholera drugs, detailed investigation of the toxin's receptor-binding site is of key importance. The substitution Gly-->Asp at residue 33 of the cholera toxin B subunit (CTB) has been reported to abolish receptor-binding ability. The substitution Arg35-->Asp has been reported to result in deficient assembly of the AB5 holotoxin. The molecular basis for these effects was not readily apparent from analysis of an earlier crystal structure of the wild-type toxin B pentamer in a complex with the receptor pentasaccharide. RESULTS: We now report at a resolution of 2.0 A the crystal structure of a recombinant CTB pentamer containing the Gly33-->Asp substitution. The observed conformation of the Asp33 side chain suggests that the loss in binding affinity is due to a steric clash with atoms C9 and O9 of the sialic acid moiety of the receptor, ganglioside GM1. The crystal structure also reveals an unexpected mode of pentamer-pentamer interaction in which pairs of toxin pentamers are joined by reciprocal insertion of the imidazole ring of His13 from one subunit of each pentamer into one of the receptor-binding sites on the other. The surface of interaction at each pentamer-pentamer interface is on the order of 500 A2, and primarily involves contact of residues 10-14 with the receptor-binding site on the associated pentamer. This same pentamer-pentamer interaction is also present in the crystal structure of a second recombinant CTB containing an Arg-->Asp substitution at residue 35, which we have determined at 2.1 A resolution. CONCLUSIONS: These structures suggest that analogs to all or part of the pentapeptide Ala-Glu-Tyr-His-Asn, corresponding to residues 10-14 of CTB, may constitute lead compounds for the design of binding-site inhibitors.

Crystal structure of a new heat-labile enterotoxin, LT-IIb.

BACKGROUND: Cholera toxin from Vibrio cholerae and the type I heat-labile enterotoxins (LT-Is) from Escherichia coli are oligomeric proteins with AB5 structures. The type II heat-labile enterotoxins (LT-IIs) from E. coli are structurally similar to, but antigenically distinct from, the type I enterotoxins. The A subunits of type I and type II enterotoxins are homologous and activate adenylate cyclase by ADP-ribosylation of a G protein subunit, G8 alpha. However, the B subunits of type I and type II enterotoxins differ dramatically in amino acid sequence and ganglioside-binding specificity. The structure of LT-IIb was determined both as a prototype for other LT-IIs and to provide additional insights into structure/function relationships among members of the heat-labile enterotoxin family and the superfamily of ADP-ribosylating protein toxins. RESULTS: The 2.25 A crystal structure of the LT-IIb holotoxin has been determined. The structure reveals striking similarities with LT-I in both the catalytic A subunit and the ganglioside-binding B subunits. The latter form a pentamer which has a central pore with a diameter of 10-18 A. Despite their similarities, the relative orientation between the A polypeptide and the B pentamer differs by 24 degrees in LT-I and LT-IIb. A common hydrophobic ring was observed at the A-B5 interface which may be important in the cholera toxin family for assembly of the AB5 heterohexamer. A cluster of arginine residues at the surface of the A subunit of LT-I and cholera toxin, possibly involved in assembly, is also present in LT-IIb. The ganglioside receptor binding sites are localized, as suggested by mutagenesis, and are in a position roughly similar to the sites where LT-I binds its receptor. CONCLUSIONS: The structure of LT-IIb provides insight into the sequence diversity and structural similarity of the AB5 toxin family. New knowledge has been gained regarding the assembly of AB5 toxins and their active-site architecture.

Single mutation in the A domain of diphtheria toxin results in a protein with altered membrane insertion behavior.

The insertion of the A domain of diphtheria toxin into model membranes has been shown to be both pH- and temperature-dependent (Hu and Holmes (1984) J. Biol. Chem. 259, 12226-12233). In this report, the insertion behavior of two mutant proteins of diphtheria toxin, CRM197 and CRM9, was studied and compared to that of wild-type toxin. Results indicated that both CRM197 and CRM9 resembled toxin with respect to the pH-dependence of binding to negatively-charged liposomes at room temperature. However, CRM197 differed from toxin with respect to both the pH- and temperature-dependence of fragment A insertion; fragment A197 inserts more readily into the bilayer at 0 degrees C and low pH or at neutral pH and room temperature than does wild type fragment A under these same conditions. This result indicates that the single amino acid substitution in the A domain of CRM197 facilitates entry of fragment A197 into the membrane, suggesting that CRM197 may be conformationally distinct from native toxin. In fact, the fluorescence spectra of CRM197 and wild-type toxin as well as their respective tryptic peptide patterns indicate that, at pH 7, CRM197 more closely resembles the acid form of wild-type toxin than the native form of toxin. These data suggest that CRM197 may be naturally in a more 'insertion-competent' conformation. In contrast, the mutation in the B domain of CRM9 which results in a 1000-fold decrease in binding affinity for plasma membrane receptors apparently does not cause a change in either the insertion of fragment A9 or the lipid-binding properties of CRM9 relative to toxin.

Membrane traffic and the cellular uptake of cholera toxin.

In nature, cholera toxin (CT) and the structurally related E. coli heat labile toxin type I (LTI) must breech the epithelial barrier of the intestine to cause the massive diarrhea seen in cholera. This requires endocytosis of toxin-receptor complexes into the apical endosome, retrograde transport into Golgi cisternae or endoplasmic reticulum (ER), and finally transport of toxin across the cell to its site of action on the basolateral membrane. Targeting into this pathway depends on toxin binding ganglioside GM1 and association with caveolae-like membrane domains. Thus to cause disease, both CT and LTI co-opt the molecular machinery used by the host cell to sort, move, and organize their cellular membranes and substituent components.

Crystal structure of a cobalt-activated diphtheria toxin repressor-DNA complex reveals a metal-binding SH3-like domain.

The diphtheria toxin repressor (DtxR) is the prototype of a family of iron-dependent regulator (IdeR) proteins, which are activated by divalent iron and bind DNA to prevent the transcription of downstream genes. In Corynebacterium diphtheriae, DtxR regulates not only the expression of diphtheria toxin encoded by a corynebacteriophage, but also of components of the siderophore-mediated iron-transport system. Here we report the crystal structure of wild-type DtxR, a 226 residue three-domain dimeric protein, activated by cobalt and bound to a 21 bp DNA duplex based on the consensus operator sequence. Two DtxR dimers surround the DNA duplex which is distorted compared to canonical B -DNA. The SH3-like third domain interacts with the metal at site 1 via the side-chains of Glu170 and Gln173, revealing for the first time a metal-binding function for this class of domains. The SH3-like domain is also in contact with the DNA-binding first domain and with the second, or dimerization, domain. The DNA-binding helices in the first domain are shifted by 3 to 5 A when compared to the apo-repressor, and fit into the major groove of the duplex bound. These shifts are due to a hinge-binding motion of the DNA-binding domain with respect to the dimerization domains of DtxR. The third domain might play a role in regulating this hinge motion.

Crystal structure of the iron-dependent regulator (IdeR) from Mycobacterium tuberculosis shows both metal binding sites fully occupied.

Iron-dependent regulators are a family of metal-activated DNA binding proteins found in several Gram-positive bacteria. These proteins are negative regulators of virulence factors and of proteins of bacterial iron-uptake systems. In this study we present the crystal structure of the iron-dependent regulator (IdeR) from Mycobacterium tuberculosis, the causative agent of tuberculosis. The protein crystallizes in the hexagonal space group P62 with unit cell dimensions a=b=92.6 A, c=63.2 A. The current model comprises the N-terminal DNA-binding domain (residues 1-73) and the dimerization domain (residues 74-140), while the third domain (residues 141-230) is too disordered to be included. The molecule lies on a crystallographic 2-fold axis that generates the functional dimer. The overall structure of the monomer shares many features with the homologous regulator, diphtheria toxin repressor (DtxR) from Corynebacterium diphtheriae. The IdeR structure in complex with Zinc reported here is, however, the first wild-type repressor structure with both metal binding sites fully occupied. This crystal structure reveals that both Met10 and most probably the Sgamma of Cys102 are ligands of the second metal binding site. In addition, there are important changes in the tertiary structure between apo-DtxR and holo-IdeR bringing the putative DNA binding helices closer together in the holo repressor. The mechanism by which metal binding may cause these structural changes between apo and holo wild-type repressor is discussed.

The 1.25 A resolution refinement of the cholera toxin B-pentamer: evidence of peptide backbone strain at the receptor-binding site.

Crystals of the 61 kDa complex of the cholera toxin B-pentamer with the ganglioside GM1 receptor pentasaccharide diffract to near-atomic resolution. We have refined the crystallographic model for this complex using anisotropic displacement parameters for all atoms to a conventional crystallographic residual R=0.129 for all observed Bragg reflections in the resolution range 22 A to 1.25 A. Remarkably few residues show evidence of discrete conformational disorder. A notable exception is a minority conformation found for the Cys9 side-chain, which implies that the Cys9-Cys86 disulfide linkage is incompletely formed. In all five crystallographically independent instances, the peptide backbone in the region of the receptor-binding site shows evidence of strain, including unusual bond lengths and angles, and a highly non-planar (omega=153.7(7) degrees) peptide group between residues Gln49 and Val50. The location of well-ordered water molecules at the protein surface is notable reproduced among the five crystallographically independent copies of the peptide chain, both at the receptor-binding site and elsewhere. The 5-fold non-crystallographic symmetry of this complex allows an evaluation of the accuracy, reproducibility, and derived error estimates from refinement of large structures at near-atomic resolution. We find that blocked-matrix treatment of parameter covariance underestimates the uncertainty of atomic positions in the final model by approximately 10% relative to estimates based either on full-matrix inversion or on the 5-fold non-crystallographic symmetry.

Comparison of high-resolution structures of the diphtheria toxin repressor in complex with cobalt and zinc at the cation-anion binding site.

The diphtheria toxin repressor (DtxR) from Corynebacterium diphtheriae is a divalent-metal activated repressor of chromosomal genes responsible for siderophore-mediated iron-uptake and of a gene on several corynebacteriophages that encodes diphtheria toxin. Even though DtxR is the best characterized iron-dependent repressor to date, numerous key properties of the protein still remain to be explained. One is the role of the cation-anion pair discovered in its first metal-binding site. A second is the reason why zinc exhibits its activating effect only at a concentration 100-fold higher than other divalent cations. In the presently reported 1.85 A resolution Co-DtxR structure at 100K, the sulfate anion in the cation-anion-binding site interacts with three side chains that are all conserved in the entire DtxR family, which points to a possible physiological role of the anion. A comparison of the 1.85 A Cobalt-DtxR structure at 100K and the 2.4 A Zinc-DtxR structure at room temperature revealed no significant differences. Hence, the difference in efficiency of Co2+ and Zn2+ to activate DtxR remains a mystery and might be hidden in the properties of the intriguing second metal-binding site. Our studies do, however, provide a high resolution view of the cationanion-binding site that has most likely evolved to interact not only with a cation but also with the anion in a very precise manner.

Structural studies of receptor binding by cholera toxin mutants.

The wide range of receptor binding affinities reported to result from mutations at residue Gly 33 of the cholera toxin B-pentamer (CTB) has been most puzzling. For instance, introduction of an aspartate at this position abolishes receptor binding, whereas substitution by arginine retains receptor affinity despite the larger side chain. We now report the structure determination and 2.3-A refinement of the CTB mutant Gly 33-->Arg complexed with the GM1 oligosaccharide, as well as the 2.2-A refinement of a Gly 33-->Asp mutant of the closely related Escherichia coli heat-labile enterotoxin B-pentamer (LTB). Two of the five receptor binding sites in the Gly 33-->Arg CTB mutant are occupied by bound GM1 oligosaccharide; two other sites are involved in a reciprocal toxin:toxin interaction; one site is unoccupied. We further report a higher resolution (2.0 A) determination and refinement of the wild-type CTB:GM1 oligosaccharide complex in which all five oligosaccharides are seen to be bound in essentially identical conformations. Saccharide conformation and binding interactions are very similar in both the CTB wild-type and Gly 33-->Arg mutant complexes. The protein conformation observed for the binding-deficient Gly 33-->Asp mutant of LTB does not differ substantially from that seen in the toxin:saccharide complexes. The critical nature of the side chain of residue 33 is apparently due to a limited range of subtle rearrangements available to both the toxin and the saccharide to accommodate receptor binding. The intermolecular interactions seen in the CTB (Gly 33-->Arg) complex with oligosaccharide suggest that the affinity of this mutant for the receptor is close to the self-affinity corresponding to the toxin:toxin binding interaction that has now been observed in crystal structures of three CTB mutants.