Heat-Labile Enterotoxins

MICHAEL G. JOBLING AND RANDALL K. HOLMES^*

[SECTION EDITOR: ALISON O’BRIEN]

Posted April 24, 2006

Department of Microbiology, University of Colorado Health Sciences Center, MS 8333, PO Box 6511, 12800 E. 19th Ave., Aurora, CO 80010

* Corresponding author. Phone: (303) 724-4224, Fax: (303) 724-4226, E-mail: This e-mail address is being protected from spambots. You need JavaScript enabled to view it .

Heat-labile enterotoxins (LTs) ofEscherichia coli are closely related to cholera toxin (CT), which was originally discovered in 1959 in culture filtrates of the gram-negative bacterium Vibrio cholerae, the causative agent of the severe diarrheal disease cholera (30). Several other gram-negative bacteria also produce enterotoxins related to CT and LTs, and together these toxins form the V. cholerae-E. coli family of LTs. Strains of E. coli causing a cholera-like disease were first recognized in the 1950s (31) and were subsequently designated enterotoxigenic E. coli (ETEC) chapter ( Enterotoxigenic Escherichia coli ) strains, one of several subtypes of pathogenic E. coli (see chapter Overview of Escherichia coli as an Intestinal Pathogen ) (68, 109). ETEC strains produce either LTs, heat-stable enterotoxin (discussed in chapter Heat-Stable Enterotoxins (ST)), or both types of toxins. The first LT was described in 1969 (78) as a cholera-like enterotoxin made by a diarrheagenic porcine isolate. Soon after, LTs were identified in strains of E. coli isolated from humans with severe diarrhea (42, 185), and ETEC strains are now recognized as major etiologic agents of diarrhea in young children and infants in developing countries (4) and travelers to tropical areas (102). The majority of LT-positive ETEC strains produce LT that can be neutralized by antisera against CT and are categorized as belonging to the type I serogroup. Additional LT-producing ETEC strains that have been isolated from human stools, cattle, other livestock, and food sources (75, 77, 192) produce related LTs that cannot be neutralized by anti-CT sera, and these belong to the type II serogroup.

SUBUNIT STRUCTURE

CT was first purified in 1966 (48) and was shown to consist of two subunits, A and B. Active CT, also called choleragen, is an 84-kDa complex containing both A and B subunits, whereas a 56-kDa immunologically related but inactive complex called choleragenoid contains only B subunits (51). CT is a heterohexameric complex of one A polypeptide and five identical B polypeptides with sizes of 28 and 11.6 kDa, respectively. The A (or enzymatically active) subunit has latent ADP-ribosyltransferase activity, and the B (or binding) subunit in the pentameric form is responsible for binding to cell surface receptors on sensitive cells. The A subunit in CT purified from V. cholerae consists of two polypeptides, a 22-kDa A1 fragment and a 5.5-kDa A2 fragment, linked by a disulfide bond, which are the products resulting from cleavage of the intact A subunit by the hemagglutinin/protease, HA/P (11).

Purification of an LT showed a similar arrangement of subunits, except that it contains an unnicked A subunit that can be nicked by trypsin (23). Similarly, type II LTs (LT-IIs) isolated from E. coli also show a heterohexameric structure and have unnicked A subunits that can be nicked by treatment with trypsin (89). Activation of the latent enzymatic activity of the A1 domain of these toxins requires both nicking of the A polypeptide and reduction of the disulfide bond between fragments A1 and A2.

IMMUNOLOGY

The division of LTs into two serogroups is based on neutralization tests with antisera against purified reference toxins (47, 77). Antisera raised against any particular type I LT (LT-I) neutralize all other LT-Is but not LT-IIs, and antisera against any LT-II will neutralize other LT-IIs but not LT-Is (77, 132, 205).

LT-Is are highly related and have been isolated from humans (LTh-I) and pigs (LTp-I). E. coli isolates from chickens have been variously reported to produce LTs determined by DNA sequencing to be identical to LTh-I from strain H10407 (100) and to produce LTs determined by immunoreactivity to be identical to LTp-I (221). Variants have been identified by both immunochemical methods and DNA sequencing of cloned loci. Although three variants of the mature A polypeptide of LTh-I have been identified, all LTp-I A polypeptides are identical (38) (Table 1). There are four variants of LTh-I mature B polypeptides and a single LTp-I mature B polypeptide described in the literature (Table 2), although an LTp-I variant of the B polypeptide with the LTh-I-specific residue glutamate at position 102 is also reported in GenBank (accession number AF242418). The LT-II family is more diverse than the LT-I family, and although only two variants of LT-II (termed LT-IIa and LT-IIb) have been well characterized (176, 177), genetic evidence shows that additional variants exist among other E. coli isolates (176). LT-II-producing strains are frequently found among E. coli isolates from healthy pigs (20) and were detected by PCR in each of 33 farm lagoon and waste samples (114). LT-II-producing strains are also frequently isolated from cattle and pigs but rarely from humans (192).

Table 1Amino acid heterogeneity among LT-I toxin A subunits.

Table 2Amino acid heterogeneity among LT-I toxin B subunits.

Serogroup I toxins include CT, LTp-I, LTh-I, and other cholera-like enterotoxins produced by several other gram-negative bacteria such as Aeromonas hydrophila (3) and Plesiomonas shigelloides (63). In Ouchterlony diffusion tests with purified antisera, immunoprecipitates form lines of partial identity, indicating that CT, LTh-I, and LTp-I express both common and unique epitopes, with LTp-I and LTh-I being more closely related to each other than to CT and CT being more closely related to LTh-I than to LTp-I (47, 69, 88, 204, 222). Immunodominant epitopes are associated with the B subunits, and monoclonal antibodies that have potent neutralizing activity against the B subunits show conformation-dependent epitopes (129). More than 20 different epitopes among the B subunits of type I enterotoxins have been defined (50, 179), and where these have been mapped they appear to be highly conformational, recognizing noncontiguous residues both within and among adjacent B monomers in the B pentamer (107).

GENETICS AND REGULATION

The majority of LT-I genes (elt) are located on large, self-transmissible (79) or mobilizable (238) plasmids, although there are instances of LT-I genes’ being located on chromosomes (99) or carried by a lysogenic phage (206). The genes encoding CT in V. cholerae are also phage carried (230). Both LT-IIa and LT-IIb are chromosomally located (176, 177). A study of several LT-producing clinical isolates found the elt genes to be flanked by conserved 236- and 280-bp regions, surrounded by highly variable sequences that contain mostly partial IS elements (190). This finding supports the conclusion of an earlier study that IS elements are involved in dissemination of elt operons among plasmids (155). Genes encoding representatives of all type I and type II enterotoxins have been cloned and sequenced (29, 138, 170, 175, 176, 177, 239), and in all cases the A and B genes are organized similarly, with the A gene preceding the B gene in an operon and both genes being transcribed as a single mRNA molecule. The A and B gene sequences also overlap by 4 nucleotides in the CT and LT-I genes and by 11 nucleotides in the LT-IIa and LT-IIb genes (Fig. 1), suggesting translational coupling of A and B subunit production. Interestingly, a recent paper described the isolation of two operons from clinical isolates of E. coli and Citrobacter freundii, the products of which have high amino acid homology with the LT and CT B subunits, and which have A subunits that are highly homologous to each other but show no homology with the CT or LT A subunit (110).

Fig. 1Overlap of genes encoding the A and B polypeptides of LTs. Fifty bases of sequence (uppercase letters) is shown for the A2-B region of each enterotoxin gene. eltA2 and eltB, LTp-I; toxA2 and toxB, LTh-I; ctxA2 and ctxB, CT; IIa-A2 and IIa-B, LT-IIa; IIb-A2 and IIb-B, LT-IIb. Single-residue notations for amino acids are shown in lowercase letters below their respective codons. Overlaps between the DNA sequences of the A2 and B genes are shown in bold and include the ATG start codon of each B gene and the TGA or TAA stop codon of each A gene. Selected restriction sites are identified in italics. The overlap of these genes suggests translational coupling of the syntheses of the A and B polypeptides of these LTs.

Comparison of the deduced amino acid sequences of the mature proteins show that the various LT-I A and CT A polypeptides share 81 to 99% identity with each other (Fig. 2). In the A1 subunit, porcine LT-I has arginine at residue 4 and human variants have lysine. A further five residues located within the disulfide-linked loop joining A1 and A2 and within the more distal part of A2 also differ among the LT-Is from various isolates. The sequence divergence between LT-I A and CT A polypeptides is greatest in the disulfide loop and amino-terminal half of the A2 polypeptide. The type IIa and IIb A polypeptides share 75% identity, and the overall identity between the A polypeptides of type I and II enterotoxins ranges between 50 and 54% (Fig. 1). The B subunits exhibit greater sequence divergence (Fig. 3A and B), with from 80 to 97% identity between CT and LT-I B polypeptides and 56% identity between type IIa and IIb B polypeptides but only 11 to 14% identity between B subunits of type I and type II enterotoxins.

Fig. 2Clustal W sequence comparison of type I, type II, and CT A polypeptides. Sequences are labeled according to type followed by the strain designation. Residues distinguishing between type I variants are shown in red (porcinelike), green (humanlike), or magenta (CT specific). The consensus shows residues that are absolutely (*), strongly (:), or weakly (.) functionally conserved between all enterotoxins. The position of the disulfide bond between Cys187 and Cys199 (type I numbering) is also shown.

Fig. 3-01Clustal W sequence comparison of type I and CT B polypeptides (excluding types IIa and IIb). Red or green residues are specific to porcine or human LT-I variants, respectively. Residues that differ between CT B polypeptides from Classical (569B) and El Tor (2125) biotypes of V. cholerae are colored magenta. The consensus shows residues that are absolutely (*), strongly (:), or weakly (.) functionally conserved between B subunits.

Fig. 3-02Limited amino acid conservation among representatives of all B subunits. Only residues completely conserved among all B subunits are colored magenta. The consensus shows residues that are absolutely (*), strongly (:), or weakly (.) functionally conserved among all enterotoxins.

The expression of CT in V. cholerae is controlled by a complex regulatory cascade involving two-component signal transduction transcriptional activators ToxRS and TcpPH; a LysR-like activator, AphB (with AphA); and an AraC-like regulator ToxT and is thus responsive to many environmental signals. These signals include quorum sensing through LuxO-regulated expression of HapR ( 245), which in turn represses AphA (115, 116). In contrast, there appears to be much less control over enterotoxin production in E. coli. Mitomycin C has been reported to induce production of LTs in clinical isolates (101), but this induction is most likely due to temperate-phage-induced lysis that releases cell-bound premade toxin (64). Sublethal concentrations of lincomycin or tetracycline are reported to stimulate both LT (240) and CT (126) production. Genetic fusions to lacZ were used to show that cyclic AMP (cAMP) and cAMP receptor protein positively modulate LT production in E. coli (66), but others found no effect (159). In E. coli, LT production is highest at 37°C and decreases with decreasing temperatures, an effect that is due to repression by the binding of the host nucleoid-associated protein H-NS to a control element located within the LT A subunit gene (212). Genetic studies using lacZ fusions as a reporter suggest that LT expression may also be osmoregulated, with expression being maximal under conditions likely to be found in the gut environment (37°C and microaerophilic with relatively high osmolarity) (213). Independent chromosomal mutations that increase LT production up to fivefold provide some evidence for genetic regulation (13), but the corresponding genes have not been identified or characterized. Levels of production of LTs from either of two plasmids in different Enterobacteriaceae varied over a 50-fold range (159), but the mechanisms governing this regulation are unknown. Similarly, the levels of LT-II made by different E. coli isolates vary greatly but by unknown mechanisms (77).

SYNTHESIS AND ASSEMBLY

The stoichiometry of A and B subunits in holotoxin requires the production of five B monomers for every A subunit. One proposed mechanism is a more efficient ribosome binding site for the B gene than for the A gene (138), increasing the rate of initiation of translation of the B gene independently from A gene translation. Biochemical studies of LTs show that there is an initial overproduction of the A subunit relative to that of the B subunit, but increased degradation of free periplasmic A subunit results in two A subunits per five B subunits synthesized (87). An alternative model suggests that the increased synthesis of the B subunit relative to that of the A subunit results both from different translational efficiencies of A and B genes and from increased mRNA secondary structures in the A subunit genes, conserved between CT and LT, that decrease the rate of synthesis of A polypeptides (237).

The B subunits of CT and LT are able to assemble into pentameric structures in the absence of the A subunit, although the presence of the A subunit can accelerate the rate of formation of the B pentamer holotoxin, perhaps by acting as a nidus for recruitment and assembly of the B subunit monomers (85). The A subunits of CT and LT-I do not interact spontaneously with preformed B pentamers (49, 82). This model is supported by the discovery that pentamer assembly occurs via the formation of ordered oligomeric intermediates (125). Nevertheless, when mixed together, periplasmic extracts from strains separately producing the A and B subunits of the type II enterotoxins are able to assemble active holotoxin (25). A rate-limiting step in the assembly process in vivo appears to be the cis isomerization of Pro93 of the CT B or LT B monomer to enable the monomeric subunits to attain an assembly-competent state (183). This cis-proline is conserved in the B subunits of CT, LT-I, and LT-IIs. This isomerization process also dramatically affects the in vitro reassembly of LT B pentamers from monomers obtained by dissociation of purified pentamers (22). Acid-denatured monomers assume a predominantly reassembly-competent state (in phosphoric acid buffer) or a predominantly reassembly-incompetent state (in KCl-HCl buffer), which is consistent with a higher cis/trans equilibrium ratio for this proline in the phosphate-buffered system. Replacement of this proline with alanine or glycine in an attempt to relieve the strain on the peptide backbone does not, however, alleviate the requirement for a cis-peptide bond between residues 92 and 93 (21).

The A2 polypeptide is the main determinant of interactions between the A polypeptide and the B pentamer, and it can be found still associated with the B pentamer after removal of the A1 peptide by reduction of nicked CT and incubation with 4 to 6 M urea (137). Additional evidence that the A2 polypeptide is sufficient to associate the A1 polypeptide with the B pentamer came from genetic fusion of the CT A2 domain onto several proteins normally found in the periplasm, which was sufficient to enable them to assemble with CT B pentamers into holotoxin-like chimeras (104). Crystal structures of CT, LT-I, and LT-IIb holotoxins all confirm that the A2 domains tether the A1 domains to the B pentamers by inserting into the pores of the B pentamers. Interactions between residues of the A2 domain and the pore of the pentamer are electrostatic within the pore but exclusively hydrophobic in nature at the upper end of the pore, and these hydrophobic interactions are critical for efficient assembly of holotoxin (210).

There is a single intramolecular disulfide bond in both the A and B subunits of CT, LT-I, and LT-IIs, and each is required for the correct folding, assembly, and function of the holotoxins. The disulfide bond is formed between Cys187 and Cys199 in the CT and LT-I A subunits, between Cys185 and Cys197 in the LT-IIa and LT-IIb A subunits, and between Cys9(10) and Cys86(81) in the CT and LT-I B subunits (residue numbers for the type II enterotoxin LT-IIa and LT-IIb subunits are in parentheses). Replacement of any of these cysteine residues with serine in the B polypeptide abolishes the ability to assemble pentamers (103) or results in defects in holotoxin assembly or stability (161). Creation of a reducing environment in the periplasm of E. coli by addition of dithiothreitol also immediately prevents assembly of a new LT B pentamer while not affecting pentamers that have already been formed (82). Oxidation of these disulfide bonds can occur spontaneously in vitro (49, 210) but in vivo is catalyzed by DsbA, a periplasmic thiodisulfide oxidoreductase. Loss of this protein by inactivation of the dsbA gene greatly diminishes the amount of the LT B subunit formed by V. cholerae expressing etxB, and a dsbA mutant of E. coli produces greatly diminished amounts of LT-I holotoxin or B pentamers (241).

The LTs are differentiated from the heat-stable enterotoxins (65) of E. coli by being inactivated by boiling or incubation at temperatures as low as 60°C. The A subunit of CT is irreversibly denatured at 51°C, although it does not separate from the B pentamer until the temperature reaches 74°C, the denaturation temperature of the pentamer (73). B pentamers are stable in the presence of 1% sodium dodecyl sulfate (67) or 6 M urea (136), but they disassemble at pHs below 3.2 for the CT B subunit (49) and below 2.0 for the LT B subunit (184).

All enterotoxins are produced in an enzymatically inactive (latent) form. Whereas LT-I and LT-II holotoxins are usually isolated from E. coli with an unnicked A polypeptide (24, 89) and treatment of these holotoxins with trypsin cleaves this loop to generate disulfide-linked A1 and A2 fragments, CT is usually isolated from V. cholerae with the A polypeptide cleaved within the disulfide loop between the A1 and A2 polypeptides (136). Reduction of the disulfide bond in the nicked A subunit is required to activate the latent enzymatic activity of the A1 subunits, whereas nicking enhances but is not absolutely required for enzyme activity and toxicity (74).

THREE-DIMENSIONAL STRUCTURES

The three-dimensional crystal structures of representative members of the LT family (CT, LTp-I, and LT-IIb) have all been determined by X-ray crystallography and found to be highly similar, despite the low amount of sequence identity, especially between the B subunits (162, 195, 227, 243, 244). The overall structure of a planar donut-shaped ring of the B pentamers with the A1 polypeptides noncovalently tethered by their A2 domains to the B pentamers confirms and refines the model proposed from earlier biochemical data and is summarized in Fig. 4.

Fig. 4Overall structure of LTs. The A subunit sits on and penetrates a donut-shaped B pentamer responsible for binding to cell surface receptors. The single A subunit is divided into two polypeptides by proteolytic cleavage in the disulfide-bonded loop (S-S) between A1 and A2. The enzymatic A1 polypeptide is formed from an N-terminal globular enzymatic A1₁ subdomain (red) wrapped by an extended-chain A1₂subdomain (yellow) to a second hydrophobic globular C-terminal A1₃subdomain (blue) that links to the A2 polypeptide (orange). The A1 polypeptide has extensive contacts with the N-terminal alpha-helical segment of the A2 polypeptide, and the C-terminal portion of the A2 polypeptide penetrates into the central pore of the B pentamer (shown as ribbons). A KDEL sequence (or a closely related sequence; in black) present at the C terminus of the A2 polypeptide functions as an ER retention sequence.

The B monomers have a short amino-terminal and a longer central alpha-helix, flanked by beta-sheets. In the pentamer, adjacent monomers each contribute three different beta-sheets to form a six-stranded beta-sheet at the interface among them, and five of these six member sheets together form the ring of the pentamer, with the central pore being formed from the long alpha-helices of the five monomers (195) (Fig. 5). This fold is also found in the Shiga toxin B pentamer (196) where the main-chain carbon atoms can be superimposed with less than 1 Å root mean square (rms) deviation over 47 residues, despite a lack of sequence homology. This fold has been identified in other oligosaccharide or oligonucleotide binding (OB) proteins and is thus termed the OB fold (156) (Fig. 6).

Fig. 5Schematic secondary-structure diagram of LT B pentamer. Shown is a projection view of the pentamer viewed from below (with respect to the holotoxin) with beta-sheets drawn as large blue arrows and alpha-helices drawn as black barrels. The beta-sheets form a single 30-ladder circle, and the large alpha-helices line the central pore of the pentamer. Redrawn from reference 195.

Fig. 6OB fold of the CT B monomer showing bound oligosaccharide (OS)-GM1. Beta-sheets are colored to correspond to those in the OB fold (156). OS-GM1 is shown in stick format.

The A polypeptide sits on the upper surface of the B pentamer, and the lower surface of the pentamer forms the receptor binding pockets. Each A1 subunit has three subdomains with an amino-terminal globular domain (A1₁) forming the core of the enzyme linked by an extended sheet (A1₂) to a second hydrophobic carboxyl-terminal globular subdomain (A1₃) to which residues from the A1₁ subdomain also contribute (243) (Fig. 4). The A1₃ subdomain leads into the disulfide-linked loop connecting A1 and A2. The A1 fragment packs against a long alpha-helix at the amino terminus of the A2 fragment, and the carboxyl terminus of fragment A2 extends into the pore of the B pentamer as an extended coil. Although the original model of CT holotoxin structure showed the A2 domain passing into the pore of the B pentamer as being almost entirely helical (243), a more recent study of two independent crystal forms of wild-type CT and three forms of a Y30S A subunit variant holotoxin showed the A2 tail extending into the B pentamer pore to be in an extended coil, more closely resembling the LT A2 tail (162).

TOXIN SECRETION AND MEMBRANE VESICLES

The A and B polypeptides of CT and all LTs are synthesized in the cytoplasm as precursors with N-terminal signal sequences, and they are then secreted into the periplasmic space (with concomitant processing into the mature forms by removal of the signal sequence), where they assemble spontaneously into holotoxin molecules (82). CT and LTs are both secreted across the outer membrane and into the external milieu by V. cholerae (159), yet both remain periplasmic when produced in laboratory strains of E. coli (86), suggesting a specific secretion system present only in V. cholerae. This idea was supported by the isolation of mutants of V. cholerae unable to secrete CT or LTs (90). By complementation of these mutants, the genes responsible were identified (28, 188), and these genes form a large operon (eps, for extracellular protein secretion) unlinked to the ctx genes (166). They form the terminal branch of the general secretion pathway or type II secretion system (187). These systems permit the secretion of folded proteins across the outer membranes of gram-negative bacteria.

In contrast to CT, most ETEC strains do not secrete LTs into the extracellular milieu, and instead the toxins remain cell associated (86, 98), although LTs can be found in the supernatants from some isolates (117). LTs have been observed in particulate fractions of supernatants from some strains (62, 232), corresponding to outer membrane vesicles (229). LTs are present both in the interiors and on the exteriors of these vesicles (96) and may catalyze the uptake of vesicles by target cells (113). Recently, a functional type II secretion system was found in ETEC strain H10407 that does secrete LTs into the medium (207). LTs secreted by the general secretion pathway remain vesicle associated and bound to lipopolysaccharide (LPS) via the B subunit (97). Further work shows that LTs and CT both bind to E. coli LPS in a manner that requires the 3-deoxy-D-manno-octulosonic acid core (95). Neither LTs nor CT bind to Vibrio LPS, where the 3-deoxy-D-manno-octulosonic acid is phosphorylated. The corresponding differences in extracellular locations of the toxins in ETEC (vesicle associated) and V. cholerae (freely secreted) may contribute to the differences in severities of the diseases caused by these enterotoxigenic bacterial pathogens.

RECEPTORS AND MUTAGENESIS

The B pentamers mediate the binding of LTs to oligosaccharide moieties of glycolipid receptors in the plasma membranes of sensitive cells (58). CT and LT-I bind strongly to the monosialoganglioside GM1 {Gal-(β1→3)-GalNAc-(β1→4)-[NANA(α2→3)]-Gal-(β1→4)-Glc-(β1→1)-ceramide}, with five binding sites per pentamer ( 53). Early biochemical data suggested that the binding pocket was formed at the interface between monomers (34, 35). The crystal structures of LTs with lactose (194) and later that of the CT B subunit with the oligosaccharide of GM1 (144) revealed the binding site to be a pocket formed primarily within a monomer, with small but significant contributions from the adjacent monomer. Involvement of adjacent monomers in formation of the receptor binding pocket was also suggested by experiments demonstrating regeneration of receptor binding activity by formation of mixed hybrid pentamers of different chemically inactivated B polypeptides (36). This involvement is also indicated by genetic complementation of the GM1 binding ability of hybrid pentamers by coexpression of different CT B alleles that alone make pentamers with no or greatly reduced GM1 binding (M. G. Jobling and R. K. Holmes, unpublished data).

The single tryptophan at position 88 in the B subunit is closely associated with GM1 binding, as its intrinsic fluorescence is altered upon binding of GM1 (152). Formylation of tryptophan also abolishes receptor binding (129). Replacement of Trp88 with charged residues abolishes GM1 binding (103) but also affects assembly of the pentamer since all Trp88 replacement mutants are made in subnormal amounts. The crystal structure of the CT B subunit with the oligosaccharide of GM1 (144) shows that Trp88 is intimately involved in GM1 binding, having strong hydrophobic interactions with the NANA moiety. Replacement of Gly33 with Asp in LTp-I B also abolishes receptor binding (219). For the CT B subunit, other amino acid substitutions affecting GM1 binding are located both within the Trp88 pocket and in the adjacent monomer (103), which can be ascribed to steric hindrance when the Gly33 of the adjacent monomer is replaced with aspartate or arginine (141, 142).

Other gangliosides also function as receptors for CT and LTs, such as fucosyl-GM1 (134), GM1 with N-glycolylneuraminic acid in place of NANA (54), and GM1(NeuGc)-Glc-ceramide (157). Derivatives of GM1 with substitutions at the C(1) (carboxyl) and C(7) (acetyl) positions of the NANA can be bound by the CT B subunit with 10- to 100-fold reduction in affinity (119, 120). In vitro, CT and LTs also bind weakly to disialoganglioside GD1b but have little or no affinity for asialo-GM1, GM3, GD2, GD3, GQ1b, or GT3 and LTs but not CT have low but measurable binding to GM2 (58). LTs but not CT can bind to functional receptors other than gangliosides, such as cell surface glycoproteins in rabbit (92) and human (93) intestines. LTp-I and LTh-I may also differ in their abilities to bind paragloboside (lacto-N-neotetraosyl ceramide, Gal-β1→4-GlcNAc-β1→3-Gal-β1→4-Glc-β1→1-ceramide) (111). Blood group A-active glycosphingolipids are also able to act as functional receptors for LTs in HT-29 cells (61), and a hybrid CT-LT B subunit was shown to specifically bind A and B blood group glycolipids at a site on the B polypeptides proposed to be distinct from the GM1 binding site (2). This is the same site that has recently been proposed as the LPS binding site in LTs (95).

The type II enterotoxins LT-IIa and LT-IIb bind preferentially to the disialogangliosides GD1b and GD1a, respectively (58), but they can also bind other gangliosides. LT-IIa also binds GD1b and GM1 with lower binding activities and exhibits weak but detectable binding to GD2, GT1b, and GQ1b, whereas LT-IIb is more restricted and among those gangliosides tested showed lower binding activity only for GT1b (58). Genetic analyses of mutations of the cloned genes have identified three residues in the B subunit of LT-IIa that are important for binding to gangliosides (26). Replacement of any one of three threonines at positions 13, 14, and 34 with a variety of amino acids abolishes the binding of the LT-IIa B subunit to GD1b, but position 34 is not required for binding to GM1. Mutants with serine substitutions at these positions retain some GD1b binding activity. In contrast, in the LT-IIb B subunit, which also has threonines at positions 13, 14, and 34, only Thr13 and Thr14 are required for binding GD1a (27). Analysis of the crystal structure of LT-IIb suggests that these residues are part of a ganglioside binding pocket similar to those in CT and LTs.

ENZYME ACTIVITY AND MUTAGENESIS

CT and all LTs ADP-ribosylate the stimulatory alpha subunit (Gsα) of the heterotrimeric regulatory G protein complex, and this inactivates the intrinsic GTPase activity of the alpha subunit. Accumulation of the ADP-ribosylated GTP-bound Gsα leads to prolonged stimulation of adenylate cyclase activity and increased cAMP levels, resulting in fluid and electrolyte secretion from intestinal epithelial cells and concomitant diarrhea when secretion exceeds the absorptive capacity of the gut. In vitro, the A1 subunits have broad substrate specificity and can also use arginine, other guanidine compounds such as agmatine, other proteins, or themselves as ADP-ribose acceptors. With water as the acceptor, they show NAD-glycohydrolase activity (149). The most striking difference between these toxins is the much lower activity of the LT-II toxins with agmatine as the acceptor, less than 1% of the activity of type I toxins (121).

The three-dimensional structures of the A1 subunits of CT and LTs reveal core catalytic domains similar to those of other ADP-ribosylating toxins (ADPRTs). In a comparison among the Cα backbone residues of the LT-I A subunit, the diphtheria toxin (DT) A subunit, the Pseudomonas exotoxin (PE) A subunit, and the pertussis toxin (PT) S1 subunit (195), 44 residues could be aligned with an rms of only 1.5 Å but only 3 of these residues were fully conserved, Tyr6, Ala69, and Glu112 of the LT-I A subunit. Glu112 corresponds to the catalytic active-site residues Glu553 of PE, Glu148 of DT subunit A (18), and Glu129 of PT subunit S1 (7). Site-directed mutagenesis has identified many other residues in the CT and LT A subunits, including His44, Val53, Ser63, Val97, Glu110, and Glu112 (106, 178), that are critical for the structures and enzymatic activities of these enterotoxins. An enzymatically inactive mutant of LTp-I has a Lys substituted for Glu112 (220). Arg7 corresponds to the important His21 and His440 residues of DT and entertoxin A, respectively, and this region shares more extensive identity to a critical region of the PT S1 subunit, which ADP-ribosylates Giα (15). The Cα backbone coordinates of PT subunit S1 and CT subunit A-LT subunit A show an rms difference of 2 Å over 116 residues, with 33 identical residues (199). All secondary-structure elements within this region are also conserved, except for an alpha-helix (no. 5) in PT subunit S1 which is much shorter in LTs and CT. In addition, some elements are longer in the PT S1 subunit than in the LT A subunit and there are significant differences in the connecting loops. Although Arg7 is part of a beta-sheet predicted to form the floor of a pocket forming the NAD binding site of LT, the crystal structure of an enzymatically inactive R7K variant of LT revealed that this pocket had widened by only 1 Å but that major structural changes had occurred in an active-site occluding loop visible in the wild-type toxin structure but absent in this mutant (225). The authors of this study postulated that the mobility of this loop is critical in enabling NAD to access the active site. Mutants designed to alter flexibility of this loop point to its having a prominent role in cytotoxicity, since mutants designed to increase or decrease flexibility are all considerably less active than the wild type (46). LT-IIs differ from LT-I at four positions within this loop. Stepwise transplantation of the LT-II residues into LT-I holotoxin produces variants that have decreased assembly levels and reduced toxicity compared to the wild type yet show increased enzyme activity in vitro with an artificial substrate, diethylaminobenzylidine-aminoguanidine (45). Since LT-II toxins have very little activity against diethylaminobenzylidine-aminoguanidine, these mutants most likely probe NAD binding and show that the loop of LT-I, although sensitive to individual substitutions, can be replaced with the same loop of LT-II, which has a very similar structure (193, 227).

CT has a K_m for NAD of about 5 mM, whereas PT, DT, and PE have K_ms for NAD in the micromolar range. The crystal structures of DT and PE with NAD show two tyrosine residues (at positions 54 and 65 and 470 and 479, respectively) involved in van der Waals interactions with the nicotinamide-ribose moiety of NAD, and PT subunit S1 has a tryptophan residue (Trp26) that may be positioned to interact with NAD. Replacement of the structurally analogous residue in CT subunit A, Arg25, with tryptophan causes a decrease in toxin activity (106) as does introduction of tyrosines at positions 68 and 72 in CT subunit A, which correspond to positions 59 and 63 in PT subunit S1, which may potentially interact with NAD (9). Reciprocal exchange of structurally analogous residues between CT subunit A and DT subunit A also causes loss of catalytic function (37). The lack of aromatic residues can thus explain the low affinity that CT subunit A1 has for NAD, but the geometry of its active site and its mode of NAD binding must differ subtly from those of the other toxins as the corresponding residues in CT and LTs are not interchangeable.

Unfortunately, until very recently all the available crystal structures of the LTs were of the catalytically latent holotoxin forms. Comparison of the structures of the unnicked and nicked (partially activated) forms of LT-I (140) reveals only minor differences, consistent with these toxins’ requiring both nicking and reduction to attain the active state. Structures of two other nontoxic LT-I mutants, those with V97K and S63K substitutions, also both appear to be in essentially the same conformation as that of the wild type (143, 226), with changes only in the region of the proposed NAD binding site. Substitution of lysine for Ser63 or Val97 may sterically hinder the access of NAD, and additionally, the V97K substitution displaces a water molecule and introduces a hydrogen bond to the catalytic Glu112.

The structure of the R7K variant of LT-I led to the development of a model for the conformational changes presumed to be required to attain an enzymatically active conformation (162, 225) (Fig. 7). The R7K substitution causes a loss of interaction of residue 7 with the active-site occluding loop residues 47 to 56, causing the loop to become too mobile to give clear density and also increasing the mobility of nearby loops of residues 25 to 27 and 33 to 36, which have interactions with each other and the latter of which also has contacts with the LT subunit A2 helix holding LT subunit A1 onto the LT B pentamer. The model predicts that reduction of the nicked form results in either dissociation of LT subunit A1 from LT subunit A2 (as happens in the endoplasmic reticulum [ER] prior to translocation of the active enzyme to the cytosol) or reorientation of the two subunits while they remain associated (in vitro). This positional shift of LT subunit A2 releases contact with residues 30 to 33, which restrains loops of residues 25 to 27 and 33 to 36, causing the entire loop of residues 25 to 36 to move. Displacement of the residue 25 to 36 loop disrupts bonds between the side group and backbone atoms of Arg25 and Tyr55 and causes displacement of the loop of residues 47 to 56, which interrupts interactions between Arg7 and Val53 and Arg54. This model predicts that activation leads to movement of the active-site occluding loop, allowing access by NAD to the cleft, with possible Arg7 interactions with its adenosine moiety and also potential NAD-loop interactions. There is evidence that the loop retains structure since NAD-bound CT subunit A1 is more resistant to trypsin degradation than unbound CT subunit A1 (60) or its R7K variant holotoxin, which shows increased sensitivity (106).

Fig. 7Model for activation of enterotoxins upon cleavage and reduction. (Top panel) A1 and A2 prior to activation, with covalent connections via their mutual peptide chain (black loop) and disulfide bond (orange line) intact. FY: Phe31,Tyr30. (Middle panel) Activation is initiated by cleavage of the peptide loop between A2 and A2 and reduction of the disulfide bond (i), which causes an indirect structural change in the environment of Tyr30 and Phe31 or in the hydrophobic patch (blue) contacted by these residues (ii). This change results in activation loop (red) disorder (iii), which in turn predisposes the active-site loop (green) toward disorder upon binding of NAD. (Bottom panel) In vivo, it is likely that A1 and A2 are physically separated shortly after reduction of the A1-A2 disulfide bond in the ER (iv), simplifying the concept of Tyr30/Phe31 pocket disruption. Reprinted from reference 162 with permission from the American Chemical Society.

This model gained experimental support with a series of mutants of CT designed to mimic this activation pathway and thus show enzymatic activity in the holotoxin molecule without nicking and reduction (E. I. Amaya, K. J. Kim, M. G. Jobling, and R. K. Holmes, unpublished data). One of these CT variant holotoxins, the Y30S variant, that shows enzymatic activity in the unnicked and oxidized state was crystallized in several forms and found to have a disordered residue 25 to 36 activation loop and various degrees of order in the residue 47 to 56 active-site occluding loop (162). This structure suggests that disorder in the activation loop of residues 25 to 36 is both necessary and sufficient for toxin activation and suggests that the active-site loop becomes disordered only in the presence of a substrate.

ENTRY AND TRAFFICKING

For the enzymatically active A1 fragment to reach its substrate, receptor-bound holotoxin must gain access to the cytosol of target cells. Major advances have recently been made in unraveling the complex steps involved in the process of toxin internalization, trafficking, and entry into the cytosol (124) (Fig. 8).

Fig. 8Retrograde toxin trafficking from membrane surface to cytosol. For a full description, see the text. An idealized enterotoxin is shown (B pentamer, dark blue; A2, dark green; A1, yellow). ?, unknown unfolding effector; magenta hexagons, ADP-ribosyl group; blue circle, Gsα; white ellipses, Gβ and Gγ; ADPrGsα, ADP-ribosylated Gsα; AdCy*, activated adenylate cyclase. Adapted and redrawn from reference 124.

There is an initial lag phase of 10 to 15 min after toxin is added before cells show signs of intoxication (increase in cAMP levels), although the toxin rapidly becomes inaccessible to external neutralizing antibodies (52). This parallels the generation of free CT subunit A1 (112), suggesting that reduction of the disulfide bond between CT subunit A1 and CT subunit A2 is the rate-limiting step. GM1-bound CT rapidly disappears from the surfaces of cells and is internalized by clathrin-dependent and -independent means in a variety of cell types (168). GM1 binding results in association with specialized membrane structures termed lipid rafts, a process essential for toxin function (235). Holotoxin then traffics in a retrograde manner from the cell surface to the Golgi apparatus and ER in a Brefeldin A-sensitive process (123, 158, 165). Brefeldin A is a fungal metabolite that disrupts the Golgi apparatus and interferes with Golgi apparatus-to-ER transport ( 146). Tight binding to GM1 may be required to enable the toxins to traffic to the ER, since a mutant B subunit having His57 replaced by Ala forms pentamers that are able to bind GM1 but are no longer toxic (1). Further work demonstrates an inability to bind to receptors localized to lipid rafts, suggesting that tight binding to GM1 in lipid rafts is a necessary sorting motif to permit retrograde trafficking of toxin in the process of toxin internalization (181). All enterotoxins have C-terminal motifs on their A subunits that are recognized by a luminal ER membrane protein, ERD2, that is responsible for retrieving resident ER proteins from the Golgi apparatus (131). Both CT and LT-IIb have A2 subunits with tails ending in KDEL, while LT-I and LT-IIa have a similar sequence, RDEL (RNEL for some variants of LTh-I). The location of the A2 tail protruding from the B pentamer pore places the tail in a position to interact with membrane-bound ERD2. These sequences are not essential for toxin function, but their replacement lengthens the lag period and increases the toxic dose by 10-fold (122). The KDEL sequence is also required for correct assembly of the holotoxin, perhaps to lock the A2 tail into the pore, since its removal prevents holotoxin assembly while still allowing A2 to stimulate the rate of B pentamer assembly (200).

Why traffic to the ER? The location of the receptor binding sites in the B subunits predicts that the holotoxins bind to cell membranes with their catalytic subunits away from the membranes. Results of an electron microscopy study of CT bound to GM1 in a supported lipid bilayer support this conclusion (16). The A subunit is situated on the upper surface of the pentamer, away from the membrane and not correctly positioned for direct membrane penetration. Additionally, the A1 subunit is disulfide linked to the A2 peptide, which is strongly tethered to the B pentamer, and it is unlikely that the whole A subunit can separate from the pentamer. Thus, the A1 peptide likely must separate from the A2—B5 complex. The redox environment of the ER is sufficient to break the disulfide bond holding A1 and A2 together (130). With CT, purified protein disulfide isomerase (PDI), an ER chaperone, is capable of reducing the disulfide bond (164), and reduced PDI was also identified as an ER protein capable of disassembling the already reduced holotoxin (215). In this context, PDI is not acting to reduce the disulfide bond but is acting as a redox-dependent chaperone to facilitate unfolding of A1. Upon oxidation of PDI (complexed with CT subunit A1) by Ero1, A1 was released (214).

Although the delivery of nicked holotoxin to the ER is sufficient to cause the release of reduced A1 from the holotoxin, the enzymatic polypeptide must still cross the ER membrane to gain access to its target. Early studies using membrane-restricted photoreactive glycolipids (211, 234) suggested that CT subunit A1, but not CT subunit B, is capable of inserting itself into lipid bilayers. More recent data indicate that crossing the membrane is an active process whereby the toxin subverts the cellular machinery to transport A1 into the cytosol. Hazes and Read (83) drew together work on the ER quality control system (the ER-associated degradation [ERAD] pathway) and toxin biology to propose that the enzymatic subunits of several A-B type toxins disguise themselves as misfolded proteins to enter the ERAD pathway and get exported into the cytoplasm by the Sec61 machinery. Normally, proteins exported from the ER by this process are targeted for ubiquitination and degradation by the proteasome. The paucity of lysines in the active but not the receptor binding subunits of these toxins (127) may decrease or prevent ubiquitination and may provide a possible explanation for the escape of the active subunits from degradation by the proteasome when these subunits reach the cytosol. Introduction of additional lysines into the wild-type toxin does in fact increase the rate of degradation of the A subunit by the proteasome (182).

Evidence that Sec61 is required for the export of CT subunit A1 from the ER (191) supports this hypothesis. The involvement of the ERAD pathway in enterotoxin trafficking is supported by results of recent studies of cell lines selected for resistance to increased levels of multiple lethal toxins that have different receptors and targets but that are proposed to share a common trafficking pathway. Some of these cell lines that also acquired unselected resistance to CT showed alterations in ERAD activity (208, 209).

How and whether these toxins traffic within the cell once they have gained access to the cytoplasm is currently unknown. Subcellular fractionation of rabbit enterocytes revealed that adenylate cyclase is located at the basolateral membrane (154) whereas the G protein target is enriched in the apical brush border membranes (39, 128), although a more recent study of rat enterocytes concluded that proteins ADP-ribosylated by CT are found only in the crude and basolateral membrane preparations and not in the apical membranes (167). ADP-ribosylated Gsα released from the apical membrane may be able to diffuse or traffic across the cell to activate basolaterally located adenylate cyclase.

COFACTORS

The ADP-ribosyltransferase activity of the A1 polypeptides can be allosterically activated in vitro by 20-kDa GTP binding proteins present in mammalian cytosol called ADP-ribosylation factors, or ARFs (108). ARFs are ubiquitous in eukaryotes and are essential for survival, being involved in the regulation of membrane trafficking and the actin cytoskeleton (180). There are six known mammalian ARFs (five in humans) grouped into three classes, designated class I (ARF1, ARF2, and ARF3), class II (ARF4 and ARF5), and class III (ARF6). Sequence homology has identified a family of ARF-like, or ARL, proteins that are structurally homologous to ARFs but do not activate CT (169) or do so to a much lesser extent (94).

Recent evidence suggests that ARFs may also be involved in toxin action in vivo, although it is currently unknown if the ARF interaction in vivo is essential for toxin activity. LT subunit A1 expressed intracellularly in CHO cells was found to bind to Golgi membranes and alter Golgi morphology (246). Overexpression of dominant-negative (T31N) or constitutively active (Q71L) variants of ARF1 were found to affect the morphological changes seen in CHO cells exposed to CT (148), suggesting a requirement for ARF1 cycling in the mechanism of CT action. Thus, ARF1 may be required for trafficking of CT from the plasma membrane to the Golgi apparatus and ER, although a requirement for direct ARF-CT subunit A1 interaction has yet to be demonstrated in vivo. In vitro, CT subunit A1 has a temperature optimum of 20 to 25°C (153) but attains its maximal activity at physiologic temperatures in the presence of phospholipids or detergents and ARFs (150).

For ARFs to activate CT subunit A1, they must be in the GTP-bound conformation (217). ARFs lowered the K_ms for NAD and agmantine (as substrates) without affectingthe V_max (160), and in the presence of 0.003% sodium dodecyl sulfate, the K_ms were lowered further and the V_max increased. These effects are consistent with ARF’s being a direct allosteric activator of CT subunit A1-catalyzed ADP-ribosylation. Gel filtration studies have provided direct evidence for complex formation (216).

By using recombinant protein from genetic hybrids of ARF1 and ARL1, the C terminus of ARF was shown to be critical for activation of CT subunit A1 (242). Recent genetic analysis of the ARF-CT subunit A1 interaction using a bacterial two-hybrid system defined a region of CT subunit A1 that interacts with ARF6 (105). Amino acid substitutions in CT subunit A1 that eliminate the ARF6 interaction but do not affect the assembly of CT subunit A into holotoxin were found in all three subdomains of CT subunit A1 yet clustered together in the CT subunit A1 structure. These substitutions included residues 41, 42, 44, 97, 99, 104, 167, and 171. These residues are partly buried in the A1 domain of holotoxin, and they suggest that a conformational change in CT subunit A1 occurs during its interaction with ARF. All of these substitutions also greatly reduce CT subunit A1 enzymatic activity and holotoxin toxicity. A similar study using a yeast two-hybrid approach with LT subunit A1 and ARF3 (247) identified the same regions in LT subunit A1 and further concluded that some LT subunit A1 residues involved in its interaction with ARF3 are also involved in its interaction with the A2 domain in holotoxin. These conclusions were recently confirmed and extended with the solving of the crystal structure of a complex of CT subunit A1 and ARF6-GTP (163) (Fig. 9). Although the structures of ARF6-GTP and CT subunit A2 are completely different, 70% of the ARF6-GTP binding surface of CT subunit A1 also interacts with CT subunit A2 and 50% of the CT subunit A2 binding surface of CT subunit A1 also interacts with ARF6-GTP. The CT subunit A1-ARF6-GTP complex also shows why ARF6-GDP cannot bind CT subunit A1 since the conformations of the switch 1 and 2 regions would cause steric clashes with structural elements of CT subunit A1 and residues of ARF6-GTP that contact CT subunit A1 in the complex are buried in ARF6-GDP. The activation loop model (162, 225) proved to be prescient indeed, with residues 25 to 40 forming an ordered coil bound to CT subunit A2 in the latent holotoxin, but upon binding of CT subunit A1 to ARF6-GTP, residues 25 to 33 form an amphipathic helix that interacts with ARF6-GTP. The structural shifts accompanying this rearrangement cause the active-site loop that occludes the proposed active site to swing out, opening the active-site crevice and exposing active-site residues Arg7, Ser61, Glu100, and Glu112 (Fig. 10). Soaking of the CT subunit A1-ARF6-GTP complex with NAD⁺ yields a 2.0-Å crystal structure of the quaternary complex revealing how NAD⁺binds in the active site (163) (Fig. 11). While the nicotinamide mononucleotide moiety binds in a compact, energetically unfavorable conformation similar to that seen in other ADPRT-NAD⁺complexes (9, 43, 80, 139, 218), the adenosine-monophosphate moieties show variable conformations in the different toxin-NAD⁺ complexes (Fig. 12).

Fig. 9Comparison of CT subunit A1 interactions with that of CT subunit A2 and ARF6. (A) The CT subunit A1-ARF6-GTP heterodimer. Secondary structural elements of CT subunit A1 are labeled according to reference 195. CT subunit A1 is gray, with important loop regions shown in gold (activation loop) and red (active-site loop). The mutated active-site residues, Asp110 and Asp112, are green sticks. ARF6 is yellow, with the bound GTP molecule as sticks. (B) CT subunit A1 in complex with CT subunit A2 (blue) from holo-CT. Conformational changes are limited to loop regions, colored as in panel A. Active-site loop residue Thr50 was disordered in the latent holo-CT structure (2); the analogous residue in other inactive holo-CT and LT structures displays increased levels of B factors relative to surrounding residues (162, 195). (C) CT subunit A1 and ARF6-GTP interface residues in a ‘‘butterfly’’ representation. Most interactions of CT subunit A1 with ARF6-GTP occur within switch 1, interswitch, and switch 2 regions, but hydrogen bonds are also made to N-terminal residues Glu13 and Arg15 (gray). Each interacting residue is marked with a sphere at the Cα position, and CT subunit A1 residues are colored according to the ARF6-GTP region contacted. Reprinted from reference 163 with permission from the American Association for the Advancement of Science.

Fig. 10Changes to CT subunit A1 upon binding of ARF6. Changes in loop regions of CT subunit A1 between holo-CT (left) and CT subunit A1-ARF6-GTP (right) lead to opening of the active site (green). Loops are colored as in Fig. 9. Reprinted from reference 163 with permission from the American Association for the Advancement of Science.

Fig. 11Binding of NAD by the CT subunit A1-ARF6 complex. ARF6-GTP-bound CT subunit A1 in complex with its substrate, NAD⁺. (A) A σA-weighted omit map contoured at 2.5 σ, showing density surrounding the NAD⁺ molecule in the CT subunit A1 active site. (B) Surface representation of the NAD⁺-occupied CT subunit A1 active site (green), viewed from the top. The large active-site cleft is open to solvent from the top and from both ends. ARF6-GTP, in yellow, binds to CT subunit A1 (gray) far from the active site (green). The knob (red), formed by active-site loop residues 48 to 52 when CT subunit A1 is ARF bound, and the ARTT motif (brown) (40) form a surface for potential Gsα recruitment. (C) Stereoview of contacts between NAD⁺ and active-site residues (green sticks). Hydrogen bonds are indicated as black dashes. NAD⁺ makes extensive direct interactions with protein atoms but also recruits three water molecules and a glycerol molecule to mediate hydrogen bonds with the active site. Reprinted from reference 163 with permission from the American Association for the Advancement of Science.

Fig. 12Conformation of NAD bound by ADPRTs. Bound NAD⁺conformations observed in ADPRTs. Views were created by superimposing structurally conserved active-site elements (rms deviations, 0.9 to 1.2 Å over ~20 Cα residues). Toxins are grouped according to quaternary-structure classification (10). The NMN moiety binds in a similarly compact fashion in all toxins, but the AMP moiety shows variation. Remarkably, in ADPRTs in the same structural family as CT, such as DT and exotoxin A, the adenine ribose is oriented differently, leading the adenine base to bind in a completely different manner from that in CT subunit A1. Reprinted from reference 163 with permission from the American Association for the Advancement of Science.

There is no direct information available regarding the regions of CT subunit A1 or LT subunit A1 responsible for their interaction with G proteins. Several ADPRTs share two crucial features near their active sites, despite only very limited structural homology: an active-site loop and the ADPRT turn-turn (ARTT) motif (81). In DT and ExoS and ExoT, the active-site loop changes conformation to allow NAD⁺ access (as it is also now known to do in CT [163]), and this altered loop conformation is also required to recognize host-protein substrates (9, 201). In the complex of CT subunit A1 with ARF6-GTP (163), the altered active-site loop creates a new knob near both the active site and the ARTT motif, which potentially defines a region that may interact with Gsα. Studies have shown that ARF-bound or free A1 polypeptides interact preferentially with the heterotrimeric Gs and ADP-ribosylate an arginine in the α subunit (149) first identified in transducin (228) as a homologue of Arg187 (or Arg201 in the long splice variant) of Gsα. This arginine is centrally located near the GTP binding site. Comparisons of the crystal structures of the isolated GTP-bound form of Gα (203) and the heterotrimer (118) show that βγ binding dramatically alters the conformation of the Gα but not the Gβγ subunit. These changes involve switch 1 and switch 2 regions, including the target arginine, providing a possible explanation for the requirement for the heterotrimeric complex for Gα to be recognized efficiently as a substrate by enterotoxins.

ASSAYS

Bioassays with animals and cell cultures have played an important role in the discovery of enterotoxins and the elucidation of their mode of action. CT (30) and LT-I were originally identified based on the secretory responses elicited in ligated ileal segments of adult rabbits. They also exhibit enterotoxicity and elicit secretory responses in ileal segments of other species, including pigs, calves, and lambs (147, 197). LT-IIs were originally identified (75) by their activity in the mouse Y1 adrenal cell assay for enterotoxins, and although they do not elicit secretory responses in ligated rabbit ileal segments, they do show activity in other species such as rats, lambs, and calves (R. K. Holmes and S. C. Whipp, unpublished data). Cell lines that exhibit morphological changes in response to LTs that can be scored by microscopic examination include Y1 adrenal tumor cells (40) and Vero cells (198), which round up, and CHO cells (76), which elongate at low toxin doses from 1 to 25 pg in microtiter assay formats.

ADJUVANTICITY

CT has been proposed to show immunomodulatory activity active during disease, down-regulating the inflammatory component of the immune response, producing the noninflammatory diarrhea characteristic of cholera (59), and both CT and LTs have profound effects as adjuvants (56, 91, 186). Immunization at mucosal surfaces usually induces a state of immunological tolerance to the administrated antigen (231), but CT and LTs are distinct in that they stimulate strong immune responses both to themselves and to coadministered antigens (41). The development of these toxins as mucosal adjuvants is hampered by their toxic effects, and much progress has been made recently in the development of variants of these toxins that have much reduced or no detectable enzymatic activity yet still retain significant adjuvanticity (171). Isolated B subunits of these toxins also show adjuvanticity and immunomodulatory effects themselves ( 202), and although early studies required that the antigen be covalently coupled to CT subunit B to be effective in oral immunization (135), nasal administration proved to be an effective route to obtain adjuvant effects by coadministration of CT subunit B with other antigens (84).

Analysis of the contribution of enzymatic and receptor binding activity to adjuvanticity (33) demonstrated that while enzymatic activity is critical for the blockade of oral tolerance (5), full activity is not required for adjuvanticity (236). Tight receptor binding activity is critical for the adjuvant effects of the isolated B subunits but, surprisingly, not for the adjuvant effects of active holotoxin (32). The reduced ability of an H57A CT subunit B variant to bind cell surface GM1 is enough to abolish its immunomodulatory activity (1). Isolated CT subunit A has also proven effective as a nasal adjuvant (17). Despite being highly homologous, CT and LT-I can have different effects on the immune system, with CT inducing Th2-type responses and LT-I up-regulating a Th1 response (236). These effects may be due to differences in the receptor binding subunits of the toxins (145), which notion is supported by results of studies with CT-LT-I chimeric toxins (12). These toxins also show promise as adjuvants for use in immunization by the transcutaneous route, with wild-type CT and LTs as well as enzymatically inactive or nonnickable variants and free B subunits all being effective (70, 71, 189). The LT-II toxins have also been tested and found to be effective as adjuvants (133). The use of toxins as adjuvants is producing promising results (72) despite recent concerns regarding potential adverse effects of nasal administration of LTs (57).

PATHOGENESIS

The major effects of CT and LTs on the human intestine are clearly due to the opening of apical chloride channels due to the increase in intracellular cAMP in intoxicated enterocytes. CT also stimulates arachidonic acid release by activating phospholipase A2 (174). Neuronal pathways contribute to the pathological effects of CT and LTs (44), since local anesthetics lidocaine and tetrodotoxin both reduce CT-induced diarrhea (19). CT induces the release of secretagogues into the intestinal milieu, including 5-hydroxytryptamine (5-HT) (8), vasoactive intestinal peptide (151), and prostaglandins (173). Prostaglandins are derived from arachidonic acid via the cyclooxygenase pathway (172). Interestingly, LTs do not appear to induce 5-HT release (224), which may in part account for differences in clinical severities between CT and LT diarrheas. The effect of 5-HT may be mediated via substance P (a tachykinin)’s activating neurokinin receptors 1 and 2 (223). The CT-induced neuronal release of acetylcholine, which results in the opening of basolateral potassium channels, may also act synergistically with CT’s direct effect of opening apical chloride channels and may further increase intestinal secretion (6).

There is no definite role for LT-IIs in human disease, and although LT-II-producing strains have been isolated from human stools (77, 192), they are much more commonly isolated or identified in association with animals or food (14, 20, 55, 192). Since purified LT-II toxins show little (LT-IIa [233]) or no (LT-IIb [235]) toxic activity on polarized human intestinal epithelial cells, this may explain their apparent inability to cause human disease.

ACKNOWLEDGMENT

This work was supported by an NIH research grant (AI31940) to R.K.H.