Chapter 153

Protein Toxins of Escherichia coli and Salmonella

ALISON D. O’BRIEN and RANDALL K. HOLMES

INTRODUCTION

Bacterial toxins are cell-associated or secreted proteins that damage the eukaryotic target cell. The older literature distinguished "endotoxins" from "exotoxins" based on the location of the toxin in the bacterium. Because many toxins of gram-negative organisms are found in the periplasmic space and are not actively secreted, we will use the term "toxin" to describe the proteinaceous poisonous product of a bacterium and "endotoxin" in reference to biological activities of enterobacterial lipopolysaccharide.

Bacterial toxins harm host cells by a variety of mechanisms, and a number of excellent reviews have been published that define these categories (147, 173, 198). One of the more recent overviews (198) groups bacterial toxins by mode of action into those that (i) help bacteria spread in tissues (e.g., hyaluronidase, collagenase, elastase); (ii) lyse host cells (alpha-toxin of Staphylococcus aureus, Escherichia coli alpha [α] hemolysin [to be discussed below], streptolysin O); (iii) block protein synthesis (e.g., diphtheria toxin, Pseudomonas aeruginosa exotoxin A, Shiga toxins from Shigella dysenteriae and E. coli [to be discussed below]); and (iv) act pharmacologically by elevating or depressing normal cell functions. Examples of the latter toxins include those that elevate cyclic AMP, such as cholera toxin and E. coli labile toxins (to be discussed below), and those that inhibit the release of neurotransmitters, such as tetanus toxin and botulinum toxin.

The toxins produced by E. coli and Salmonella spp. are the focus of this chapter. For some of these toxins, a role in pathogenesis of human and/or animal disease has been experimentally established, whereas for other toxins the link is based on epidemiological data. Wherever possible, we will summarize the experiments describing the relationship between toxin production and clinical manifestations. As discussed by Schlessinger and Schaechter (198), there are several approaches by which a role for a toxin in disease can be established. These include determining whether toxin alone can cause lesions or disease in an appropriate model, assessing whether virulence is correlated with the amount of toxin produced by an organism, testing whether antitoxin is protective, and analyzing the effect of a mutation that prevents toxin production on disease caused by the microbe. The last approach is required to fulfill "molecular Koch’s postulates" as defined by Falkow (59).

The majority of this chapter will be devoted to the toxins of E. coli because considerably less information is available about Salmonella toxins. The toxins produced by E. coli that cause intestinal as well as extraintestinal infections will be discussed. In 1987, Levine described several categories of E. coli that cause gastroenteritis (123), and subsequently two additional categories have been added. These six types of diarrheagenic E. coli include enterotoxigenic E. coli (ETEC), enterohemorrhagic E. coli (EHEC), enteroinvasive E. coli (EIEC), enteroaggregative E. coli (EAggEC), enteropathogenic E. coli (EPEC), and diffusely adherent E. coli (DAEC). Four of these groups of pathogenic E. coli (ETEC, EHEC, EIEC, and EAggEC) produce toxins that will be considered here. Wherever possible, the toxins of diarrheagenic E. coli will be described under the pathogenic category of the bacterium.

TOXINS OF DIARRHEAGENIC E. COLI

ETEC

ETEC are strains of E. coli that produce heat-labile enterotoxin (LT), heat-stable enterotoxin (ST), or both (123) (Table 1). They cause acute, watery diarrhea in humans and animals. ETEC are responsible for significant morbidity and mortality, particularly among young children and elderly persons in developing countries, and they are one of the most common causes of diarrhea among travellers from developed to developing countries (100, 123). Virulence of ETEC typically requires production of species-specific colonization factor antigens in addition to enterotoxins. Vaccines against colonization factor antigens are used frequently in veterinary medicine to protect livestock against ETEC diarrhea (149), but vaccines against ETEC have not yet been developed for human use.

Table 1E. Coli and Salmohella toxins

Heat-Labile Enterotoxins.

  Structure-function. The LTs of E. coli are closely related to cholera toxin (CT) of Vibrio cholerae in structure and function (2, 87, 93, 214). Several gram-negative bacteria other than E. coli also produce enterotoxins related to CT and LT. Collectively these toxins make up the V. cholerae/E. coli family of heat-labile enterotoxins (93). CT and LT have been studied in the greatest detail, and this discussion will focus on the LTs of E. coli.

The LTs are classified in two serogroups, designated LT-I and LT-II, on the basis of neutralization tests (85, 93, 95). Polyclonal antiserum against a particular LT-I neutralizes CT and other LT-Is, but it does not neutralize LT- IIs. Conversely, polyclonal antiserum against a particular LT-II neutralizes other LT-IIs, but it does not neutralize type I enterotoxins. Variants of LT-I and LT-II have been identified both by immunochemical methods and by comparing the deduced sequences of their constituent polypeptides. LT-I produced by ETEC from pigs is designated LTp-I, and LT-I produced by ETEC from humans is designated LTh-I (73, 94, 220). LTp-I and LTh-I exhibit partial antigenic cross-reactivity. Three amino acid sequence variants of LTh-I have been identified, but variation among LTp-Is from different E. coli isolates has not been reported (48). Two variants of LT-II, designated LT-IIa and LT-IIb, have been characterized, and preliminary evidence for the existence of additional variants has been reported (85, 95, 182).

Each LT has both unique epitopes and epitopes that are shared with other variants in the serogroup. In neutralization tests, each polyclonal anti-LT serum exhibits the highest titer against the LT variant used as immunogen, a lower titer against another variant in the same serogroup, and no activity against an LT from the other serogroup (94, 138). Among monoclonal antibodies against type I enterotoxins, some recognize common determinants and react in immunoassays with two or more variants, whereas others recognize unique determinants and react with only one variant (9, 10, 64, 185). All monoclonal antibodies against LT-IIa or LT-IIb characterized to date recognize unique determinants (R. K. Holmes, F. M. J. Petitjean, T. D. Connell, and E. M. Twiddy, unpublished data). Some monoclonal antibodies neutralize enterotoxicity, and others do not.

Each of the heat-labile enterotoxins is a heterohexameric protein with one A polypeptide (240 to 243 amino acids) and five identical B polypeptides (99 to 103 amino acids) (76, 93). The operons that encode these enterotoxins are in plasmids for LT-I (86), in the E. coli chromosome for LT-II (80), and in the V. cholerae chromosome for CT (215). In each operon the gene for polypeptide B is downstream, overlaps the 3' end of the gene for polypeptide A, and is translated in a different reading frame (131, 142, 182, 239, 240). Factors that may contribute to the required overexpression of polypeptide B relative to polypeptide A include higher efficiency of the ribosome binding site for polypeptide B and relatively stable secondary structure in the message that decreases translation for polypeptide A (148, 238). The ctx operon in V. cholerae is positively regulated by the toxR/toxS signal transduction pathway, but no comparable regulatory system exists in E. coli (140). For LT-I production in E. coli, chromosomal mutations confer hypertoxinogenic phenotypes by unknown mechanisms (18), and there are conflicting reports concerning cyclic AMP-dependent regulatory effects on synthesis of LT-I (75, 160). Subinhibitory concentrations of antibiotics can also increase expression of CT in V. cholerae and LT-I or LT-II in E. coli (95, 125, 244).

Representative operons that encode CT, LTp-I, LTh-I, LT-IIa, and LT-IIb have been cloned and sequenced; the amino acid sequences of the polypeptide chains that they encode have been deduced and compared; and the corresponding proteins have been purified to homogeneity and characterized (48, 93, 214). The A polypeptides exhibit 81.3% to 99.2% homology between members of serogroup I, 75.5% homology between LT-IIa and LT-IIb, and 50.5% to 854.2% homology between members of serogroups I and II. The B polypeptides exhibit 80.6% to 97.1% homology between members of serogroup I, 56.4% homology between LT-IIa and LT-IIb, but only 10.7 to 13.6% homology between members of serogroups I and II. Complementation tests in E. coli demonstrated that the A and B polypeptides of type I and type II enterotoxins could assemble to form detectable amounts of hybrid holotoxins, notwithstanding the striking differences in the primary sequences of the B polypeptides from toxins in these two serogroups (38).

The three-dimensional structures of LTp-I and CT were determined by X-ray crystallography and shown to be very similar (74, 206, 208). The five B polypeptides are arranged symmetrically to form a planar, donut-shaped ring. The A polypeptide has two domains, a globular A1 domain that extends outward from one face of the B pentamer and an A2 domain with an extended carboxyl-terminal tail that penetrates through the center of the B pentamer. Structures have also been determined for LTp-I in complex with lactose or galactose and for the B subunit of CT in complex with the oligosaccharide from ganglioside GM1 (144, 145, 207). Each polypeptide in the B pentamer has one oligosaccharide-binding site that is located on the opposite face of the B pentamer from the position of the A1 domain. The structure for a representative LT-II has not yet been determined.

Treatment of any of the heat-labile enterotoxins with trypsin cleaves the surface-exposed loop between A1 and A2 to produce "nicked toxin," with the A1 and A2 fragments joined by a disulfide bond (36, 95, 141). Reduction of the disulfide bond generates fragment A1, which exhibits NAD:Gsα ADP-ribosyltransferase activity that mediates toxicity (155). Although reduction is essential for the enzymatic activity of fragment A1, nicking of the A polypeptide enhances, but is not required for, enzymatic activity and toxicity (79). The GTP-binding protein Gsα is a regulatory component for the adenylate cyclase complex in the plasma membrane of eukaryotic cells that participates in hormone-induced signal transduction. Enterotoxin-mediated ADP ribosylation inhibits the intrinsic GTPase activity of Gsα, persistently activates adenylate cyclase, causes accumulation of intracellular cyclic AMP, and results in activation of cyclic AMP-dependent cellular functions (214). The physiological consequence of intoxication of enterocytes by LT is active secretion of Cl^–, accompanied by passive loss of water, into the lumen of the small intestine, causing the watery diarrhea (62, 63). The cystic fibrosis transmembrane conductance regulator (CFTR) is required for the Cl^–-secretory response to CT in transgenic mice, prompting speculation that partial resistance to diarrheal diseases caused by V. cholerae and ETEC might be a positive selective factor for the high frequency in human populations of individuals heterozygous for cystic fibrosis (72).

The NAD:Gsα ADP-ribosyltransferase activity of the A1 fragment of heat-labile enterotoxins is allosterically activated in vitro by 20-kDa guanine nucleotide-binding proteins from mammalian cells, called ADP-ribosylating factors (157), but whether these factors are required or important for the activity of heat-labile enterotoxins in intoxicated humans or animals has not been established. The A1 fragments of LT-I and LT-II have similar specific activities for Gsα as the ADP ribose acceptor, but with agmatine as an alternative acceptor for ADP ribose the A1 fragment from LT-II is about 200-fold less active than the A1 fragment from LT-I (121).

The crystal structure of LTp-I reveals that the NAD-binding and catalytic sites of its A1 domain share a common motif with the corresponding sites of other ADP-ribosylating toxins (47). The elements of this motif in LTp-I exhibit extensive to limited homology with the corresponding elements in CT, pertussis toxin, exotoxin S of P. aeruginosa, and the mosquitocidal toxin of Bacillus sphaericus, but they show no detectable homology with the corresponding elements in diphtheria toxin and exotoxin A of P. aeruginosa. The proposed mechanism of catalysis involves one strictly conserved glutamate (residue 112 in LT-I) that interacts with the acceptor protein and a second arginine or histidine residue (Arg-7 in LT-I) that interacts with the nicotinamide amido group of bound NAD and positions the N-glycosidic bond for nucleophilic attack by the incoming acceptor protein (47). LT-I, CT, pertussis toxin, and the mosquitocidal toxin (but not diphtheria toxin or exotoxin A) also have a conserved histidine residue (His-44 in LT-I) that is positioned to interact with oxygen in the ribose ring and further increase the susceptibility of the adjacent anomeric carbon atom to nucleophilic attack by the acceptor protein. Mutational analysis of the A polypeptides of LT-I and/or CT (23, 89, 183, 226) identified multiple positions at which single amino acid substitutions had the following effects: a decrease in enzymatic activity and/or toxicity (Arg-7, Val-53, Ser-63, Val-97, Tyr-104, His-107, Glu-110, Glu-112, and Ser-114); no effect on toxicity (Arg-54, Tyr-59, Ser-68, Ala-72, and Arg-192) (but see reference 79 for differing results with Arg-192); and failure to produce immunoreactive A polypeptide detectable in Western blots (Leu-41, Ala-45, Val-60, and His-70).

The B pentamers mediate binding of heat-labile enterotoxins to oligosaccharide moieties of glycolipid receptors on the plasma membranes of target cells (71) and trigger internalization of toxin by a poorly characterized endocytotic pathway (171). Each LT-I or LT-II has a KDEL, RDEL, or RNEL sequence at the carboxyl terminus of its A polypeptide (48) that may function as an intracellular routing signal, consistent with the observed inhibitory effect of brefeldin A on the toxicity of CT and LT for cells in culture (50, 122, 158, 174, 179). These findings suggest that trafficking of enterotoxins through the Golgi and possibly the endoplasmic reticulum is essential for their toxicity.

The B subunits of type I enterotoxins bind strongly to ganglioside GM1 and weakly to ganglioside GD1b (71); and LT-I, but not CT, also binds to specific intestinal glycoproteins (81, 96, 97). The amino acid residues of LT-I that interact directly with the GM1 oligosaccharide (e.g., Glu-11, Tyr-12, His-13, Glu- 51, Gln-56, Trp-88, Asn-90, and Lys-91) are all located in the same B polypeptide (144, 145, 207). The Gly-33 residue of the n+1 polypeptide in the B pentamer forms hydrogen bonds via a water molecule with both the sialic acid and the terminal galactose residue of the GM1 oligosaccharide bound by the n polypeptide. Substitution of Asp for Gly-33 in the B polypeptide of LTp-I (225) or substitution of a negatively charged or large hydrophobic residue for Gly-33 in CT-B (104) abolishes ganglioside-binding activity, as does substitution of a Lys or Glu residue for Trp-88 (104) or modification of Trp-88 by formylation (135) in CT-B. Recent experiments with CT-B demonstrated that triple substitution of Asp for Tyr-12, Lys for Glu-51, and Asp for Lys-91 caused complete loss of ganglioside-binding activity, although the CT-B variants with the corresponding single and double substitutions retained at least partial activity (M. G. Jobling and R. K. Holmes, unpublished data). Complementation experiments demonstrated that the Asp-33 variant and the Asp-12, Lys-51, Asp-91 triple variant formed mixed CT-B pentamers with partial ganglioside-binding activity, although homopentamers consisting of either subunit alone lacked ganglioside-binding activity. These findings are fully consistent with the structure of the ganglioside-binding site determined by X-ray crystallography as summarized above.

The B polypeptides of LT-II, which differ dramatically in primary sequence from the B polypeptides of LT-I (48, 93), also exhibit striking differences in specificity for binding to gangliosides (71). LT-IIa binds best to ganglioside GD1b, followed by GD1a, GT1b, GQ1b, GM1, GD2, and other monosialogangliosides. LT-IIb interacts with a more restricted set of gangliosides, binding best to GD1a, followed by GT1b and, weakly, by GM3. Mutagenesis experiments demonstrated that the Thr-13, Thr-14, and Thr-34 residues of the B polypeptide of LT-IIa are important for ganglioside GD1b-binding activity (37), whereas Thr-13 and Thr-14, but not Thr-34, are important for ganglioside GD1a-binding activity of the B polypeptide of LT-IIb (39). The effects of substitutions of various amino acids at these positions indicated that the hydroxyl groups of these threonine residues are important for ganglioside-binding activity of the B polypeptides of LT-IIa and LT-IIb.

  Animal and cell culture models. Bioassays in animals and cell cultures have had an important role in the discovery of enterotoxins and analysis of their mode of action. CT and related LT-Is were originally characterized by their ability to elicit secretory responses at sub-microgram doses in the small intestine of infant rabbits or in ligated ileal segments of adult rabbits (40, 55), but they also exhibit enterotoxicity in ileal segment assays in other species including pigs, calves, and lambs (151, 212). LT-IIs do not elicit intestinal secretory responses in ligated ileal segments in rabbits (85, 95), but preliminary ileal segment assays in other species showed that LT-IIs exhibited enterotoxic activity in rats, lambs and calves (S. C. Whipp and R. K. Holmes, unpublished data). CT, LT-Is, and LT-IIs activate adenylate cyclase, cause increases in intracellular cyclic AMP, and induce various cyclic AMP-mediated responses in many types of cultured cells (51, 91, 137). Cell lines which exhibit morphologic changes that can be scored by direct microscopic observation include mouse Y1 adrenal tumor cells (52), which become rounded, and Chinese hamster ovary (CHO) cells (83), which become elongated, following exposure to heat-labile enterotoxins at doses from approximately 1 to 25 pg in microtiter assay systems.

  Role in pathogenesis. CT is the principal toxin responsible for the massive watery diarrhea of cholera (65), although mutants of V. cholerae that cannot make CT still make several other toxins (zonula occludens toxin [8], accessory cholera enterotoxin [224], SLT [162]) and retain the capacity to cause mild diarrhea. The mode of action of LT-I is comparable to that of CT (93, 156). Watery diarrhea associated with ETEC infections may be caused by LT, by ST, or by both, and the diarrhea caused by ETEC is typically milder than the diarrhea of cholera. Factors that might account for the lesser severity of ETEC diarrhea compared to cholera include the periplasmic retention of LT-I in E. coli versus the extracellular secretion of CT by V. cholerae, which could favor access to enterocytes by CT versus LT-I (161), and the degree of alkalinity of enterotoxin-elicited small intestinal contents, which appears to be more optimal for growth of V. cholerae than for growth of E. coli (1). Although no definite role is established for LT-IIs in diarrheal diseases, LT-II-producing strains of E. coli are isolated more frequently from animals (including water buffalo, pigs, cattle, etc.) than from humans or the human food supply (21, 30, 84, 203), and a role for LT-II-producing strains of E. coli in diarrheal diseases of animals remains plausible.

STs.

  Structure-function. STs are heat-stable, low-molecular-weight, extracellular bacterial polypeptides that alter the movement of fluid and electrolytes across the intestinal epithelium of susceptible animal hosts (211). There are two unrelated kinds of ST, designated STI (STa) and STII (STb), that differ in structure and mechanism of action (2, 87). STI, STII, or both are encoded by plasmids that vary widely in molecular size and that may also encode LT, fimbrial colonization factor antigens, antibiotic resistance determinants, and colicins (86, 87). Genes for STI and STII have been shown to be present in transposable elements designated Tn1681 (213) and Tn4521 (99, 119), respectively.

STIs are produced by E. coli and several other gram-negative bacteria including Yersinia enterocolitica, V. cholerae O1, and V. cholerae non-O1 (178). Two variants, designated STIa (also called STp or porcine STI) and STIb (STh or human STI), consist of 18 and 19 amino acid residues, respectively, with homologous amino acid sequences and six conserved cysteine residues that form three intramolecular disulfide bonds (178, 205). STIs also exhibit homology with the endogenous intestinal peptide hormone guanylin (126, 195), but the role of guanylin in normal intestinal physiology has not yet been defined. STIa and STIb share antigenic determinants associated with the conserved amino acid sequences at their carboxyl termini (87). Native STIs are poorly immunogenic, but they may be conjugated to carrier proteins in order to prepare antisera and monoclonal antibodies for specific immunoassays (3, 19, 132, 223).

Cloning and analysis of the genes for STIa and STIb in E. coli revealed that the initial gene products are pre-pro-STI molecules each consisting of 72 amino acid residues (154, 213). Processing of the intracellular pre-pro-STIs to form the mature extracellular enterotoxins appears to require secA-dependent transport of pre-pro-STI across the cytoplasmic membrane, removal of the "pre" sequence by signal peptidase I; formation of the correct intramolecular disulfide linkages by periplasmic protein disulfide isomerase (DsbA)-dependent oxidation of the cysteine residues and cleavage of the pro-STI by an undefined protease, and release of mature STI by diffusion across the outer membrane (118, 241, 243). The Cys-39 residue in the "pro" sequence is important for correct recognition by DsbA and proteases during the maturation of STI (241).

Early studies demonstrated that STIs can bind noncovalently with high affinity or low affinity to several different proteins in the plasma membranes of susceptible eukaryotic cells (54, 231). Cloning and expression of the gene for guanylate cyclase C led to the demonstration that STIs bind to the extracellular domain of guanylate cyclase C and directly stimulate its intracellular catalytic activity (201). Activation of guanylate cyclase C in enterocytes of the small intestine initiates a cascade that involves the intracellular accumulation of cyclic GMP, the cyclic GMP-dependent activation of protein kinase A, the protein kinase A-dependent phosphorylation and activation of CFTR (the cystic fibrosis transmembrane conductance regulator), and the resulting CFTR-dependent secretion of Cl^– ions into the intestinal lumen (31, 77, 114). The enterotoxic effects of STI and LT, therefore, both involve the CFTR-dependent pathway of Cl^– secretion by the small intestine. In suckling mice, STI-stimulated increases in intestinal secretion exhibited a close quantitative correlation with increases in occupancy of the low-affinity, but not the high-affinity, STI receptors (232).

The three-dimensional structure of the toxic domain of STI was determined by X-ray crystallography, and the function of STI was analyzed by comparing native enterotoxin with synthetic oligopeptide analogs and with mutant forms of the enterotoxin (92, 178, 195). STI exhibits a right-handed spiral structure with three β turns. The three intramolecular disulfide bonds in STIa link cysteines 5 and 10, 6 and 14, and 9 and 17; in STIb the homologous disulfide bonds link cysteines 6 and 11, 7 and 15, and 10 and 18. Oligopeptides that contain only the 13 amino acids from the amino-terminal cysteine residue through the carboxyl-terminal cysteine residue of STI have full enterotoxic activity, indicating that this segment constitutes the active domain. Reduction of the disulfide bonds in STI abolishes enterotoxicity. STI analogs without the second disulfide bond lack toxicity, but analogs without either the first or the third disulfide bond exhibit partial toxicity. The disulfide bond between cysteines 6 and 14 of STIa or 7 and 15 of STIb is therefore essential for biological activity. The smallest toxic analog of STI is the decapeptide corresponding to conserved residues 5 to 13 of STIa and 6 to 14 of STIb. Carba-analogs of this decapeptide with either one of the two internal disulfide bonds replaced by a CH₂-S linkage that cannot undergo disulfide exchange retain enterotoxic activity. Taken together, these findings indicate that the role of the intramolecular disulfide bonds is to stabilize the active tertiary structure of STI, that binding of STI to its receptor occurs by noncovalent interactions, and that disulfide exchange reactions between STI and its receptor are not required for enterotoxic activity.

The Asn-11, Pro-12, and Ala-13 residues of STIa form an exposed hydrophobic surface that is important for receptor-binding activity and enterotoxicity (195). Replacement of Asn-11 by Val has little effect on activity, but substitution of a charged Asp, Glu, Arg, or Lys residue for Asn-11 decreases activity by 100- to 1,000-fold. The exchange of Val or Ala for Pro-12 causes a slight decrease in activity, but replacement of Pro-12 by Gly has a marked effect on both enterotoxicity and receptor-binding activity. Substitution of Ala-13 by an amino acid with a larger side chain such as Leu results in loss of activity, but Gly-13 or Ser-13 substitutions cause only a moderate decrease in activity. The methyl side chain of Ala-13 is directed outward and is believed to interact directly with a hydrophobic pocket on the surface of the STI receptor. It is noteworthy that the homology between guanylin, an endogenous ligand for GC-C, and the carboxyl-terminal segment of STIa includes residue Ala-13 in the receptor-binding domain of STIa.

STII production has been demonstrated primarily by ETEC isolates from pigs, although DNA probes for STII hybridize with DNA from some E. coli isolates from humans (87). The gene for STII encodes a 71-amino-acid polypeptide with a 23-amino-acid signal sequence (120, 180). The 48-amino-acid mature protein is strongly basic and has a pI of 9.7. Site-directed mutagenesis studies demonstrated that variants with amino acid substitutions for Lys-22 and Lys-23 had markedly decreased toxicity, variants with replacements for Lys-18 and Lys-46 had slightly decreased toxicity, and variants with exchanges for Lys-6 and Lys-7 had nearly wild-type activity (70). Each of the variants was purified before it was tested for enterotoxicity and was shown to be cleaved by signal peptidase at the same position as native STII.

STII does not cause an increase in the intracellular concentration of cyclic AMP or cyclic GMP, and the intestinal secretory response to STII appears to be mediated by secretion of bicarbonate rather than chloride (87). Recent evidence showed that STII causes an increase in the concentration of intracellular calcium ions in intoxicated cells (53), but the secondary messengers that presumably mediate STII action have not yet been clearly defined.

  Animal and cell culture models. STI elicits intestinal secretory responses following intragastric administration to newborn mice or inoculation into ligated ileal loops of pigs (2, 211). Initially STII was shown to be active in ileal loops in pigs but inactive in infant mice. Subsequent studies showed that STII is much more susceptible to proteolysis than STI and that STII can elicit secretory responses in the intestines of mice and rats if it is protected from proteolytic inactivation (234, 235). Increased production of cyclic GMP in response to treatment with STI has been demonstrated in cells from several animal sources. No convenient cell culture assay for STI or STII is currently available for diagnostic or investigative purposes.

  Role in pathogenesis. ETEC that produce STIa are important causes of diarrheal disease in pigs and calves and have also been shown by volunteer studies to cause diarrhea in humans (87). ETEC that produce STIb are isolated most often from humans (153). Use of DNA probes revealed that the gene for STII was the most frequent among enterotoxin genes of ETEC isolates from pigs (150).

EHEC

EHEC cause food-borne hemorrhagic colitis that can sometimes lead to the development of the hemolytic-uremic syndrome (HUS) (109, 187). There are several serotypes of EHEC that cause these syndromes, but EHEC O157:H7 is the single most common cause of large-scale outbreaks such as those that occurred in 1983 and 1993 in association with ingestion of contaminated undercooked hamburgers (167, 187). EHEC of certain serotypes and EPEC, a group of E. coli strains responsible for diarrhea in infants, share a number of pathogenic characteristics. These traits include expression of genes for attaching and effacing of intestinal microvilli (the genes are called eae) and large, albeit distinct, plasmids. However, the key feature that distinguishes all EHEC from EPEC is that EHEC produce one or more types of Shiga-like toxins (SLTs).

SLTs.

  Structure-function. The Shiga-like toxins (SLTs), which are also called Vero toxins (115, 165, 169), are members of the Shiga toxin family (reviewed in references 163 and 169). The family members share a common protein and operon structure, mode of action, and related or identical eukaryotic receptors (163). Individual family members carry the slt operon with the A subunit gene proximal to the B subunit gene on the bacterial chromosome or on toxin-converting phages (163, 166, 168). The prototype toxin is Shiga toxin, which is produced by S. dysenteriae type 1. Shiga toxin has a molecular weight of about 70,000 and is composed of a single A polypeptide of 32 kDa that is noncovalently associated with a pentamer of 7.7-kDa B subunits (49, 169). The A subunit is an N-glycosidase that removes a single adenine residue from 28S rRNA of eukaryotic ribosomes, a cleavage which results in inhibition of protein synthesis in the target cell (57). This mode of action is identical to that of the plant lectin ricin (56), and the A chains of ricin and Shiga toxin share regions of high homology (24, 103). The glutamic acid at position 167 of the SLT-I A subunit (deemed synonymous with the Shiga toxin A subunit, see below) is considered to be within the catalytic domain of this toxin (98). Other amino acids in the A subunit of SLT-I or related ribosome-inactivating proteins that appear to be critical for N-glycosidase function include Tyr-77, Tyr-114, Arg-170, and Trp-203 (42, 103, 199, 242). The pentamer of Shiga toxin B subunits binds to globotriaosylceramide (Gb₃) receptors on the eukaryotic cell membrane (101, 128). The binding of Shiga toxin to Gb₃ is necessary but not sufficient to intoxicate cells (102).

Two immunologically non-cross-reactive groups of SLTs have been isolated from EHEC, and a single EHEC strain can produce SLT-I, SLT-II, or both toxins (169). The SLT-I of EHEC is essentially the same toxin as Shiga toxin (218); the deduced amino acid sequence indicates only a single conservative amino acid change in the A subunit between SLT-I and Shiga toxin. Moreover, no significant sequence or antigenic variations have been reported within the Shiga toxin/SLT-I group. By contrast, members of the SLT-II group, which share about 55% and 57% amino acid sequence homology in the A and B subunits, respectively, with Shiga toxin/SLT-I, exhibit sufficient sequence divergence in the B subunit to display some variability in the cross-neutralizing capacity of polyclonal and monoclonal anti-SLT-II antibodies as well as differences in avidity and specificity of binding to glycolipid receptors. For example, SLT-IIe (formerly called SLT-IIv), which is produced by E. coli strains responsible for edema disease of pigs (87, 136), is only partially neutralized by anti-SLT-II sera elicited against SLT-II isolated from EHEC O157:H7 strain 933 (139). Additionally, SLT-IIe uses globotetraosylceramide as its functional receptor (41, 190) although it can bind to Gb₃ (190). This difference in receptor preference affects the organ-specific tropism of the toxin and is related to a difference in amino acid content of the B subunits of SLT-II and SLT-IIe (17, 227).

The three-dimensional structure of the B oligomer of SLT-I (designated VT-1 in the cited article) has been determined (209, 217). The crystal structure resembles that of the larger E. coli LT B subunit oligomer, a finding which is surprising because SLT-I and LT B subunits are unrelated molecules. Each 69-amino-acid SLT-I B monomer consists of six β-sheets and an α-helix, and adjacent monomers are held together by interactions between β-strands. The cleft between the neighboring B subunits of the pentamer has been proposed as one of five potential Gb₃-binding sites per toxin molecule (217). A more recent X-ray crystallographic examination of Shiga holotoxin revealed that the B subunits form a pentameric ring that encircles the helical carboxy terminus of the A subunit (69).

One model for how cells become intoxicated by Shiga toxin and the SLTs, as outlined by Sandvig and colleagues (193, 194), is as follows. The B pentamer binds to glycolipids on the target cell surface, and the holotoxin is endocytosed through coated pits. The toxin is then transferred both to lysosomes and to the Golgi apparatus. Transport to the Golgi apparatus appears to be a requisite for intoxication. From the trans-Golgi, Shiga toxin is transported in a retrograde manner to the endoplasmic reticulum, and translocation of the A subunit to the cytoplasm occurs in this organelle. The A subunit can be processed by proteolytic nicking and reduction (5, 172) into an approximately 27-kDa amino-terminal A₁ fragment and an approximately 4-kDa carboxyl-terminal A2 fragment. The A1 fragment is enzymatically more active than the intact A polypeptide (5, 172), and A₂ is required for holotoxin integrity (5). The specific site at which A is cleaved into the A₁ and A2 fragments remains to be determined (22, 189).

  Animal and cell culture models. Shiga toxin is paralytic for rabbits and mice and lethal for a variety of animals (29, 163). Shiga toxin is also enterotoxic, i.e., induces fluid secretion in ligated rabbit ileal segments (112), and cytotoxic for selected epithelial cell lines (229) as well as endothelial cells (170, 222). Indeed, endothelial cells are believed to be the target for Shiga toxin and the SLTs in vivo (170, 186), and the location of the endothelial cells targeted by these toxins may determine the clinical presentation of disease (hypothesis discussed in reference 221). Examples of sites of possible Shiga toxin/SLT endothelial cell targets include renal glomeruli in HUS (130), the stomach, small intestine, brain, and eyelid in edema disease of swine (17), and the colon in hemorrhagic colitis (68). Whether damage to these cells is a direct consequence of Shiga toxin/SLT action or indirect and mediated via one or more cytokines (such as tumor necrosis factor) remains to be determined (88, 133, 134, 222, 228).

A number of animal models have been described for study of the role of Shiga toxin and SLTs in the pathogenesis of disease (reviewed in reference 169). Only two of these models closely mimic the type of disease seen in the typical host for the organism. These two models include monkeys orally infected with S. dysenteriae type 1 (68) and pigs parenterally inoculated with the edema disease toxin, SLT-IIe (136). Other animal systems have been developed that permit focus on a particular facet of the disease process. Thus, the rabbit is useful for the study of vascular lesions induced after parenteral administration of SLT-I (186) or intragastric inoculation of SLT-I-producing E. coli (210), and the streptomycin-treated mouse is helpful for the evaluation of the virulence and kidney-damaging potential of EHEC administered orally (129, 230). Unfortunately, no animal model for the development of HUS has as yet been developed.

  Role in pathogenesis. The evidence that Shiga toxin and the SLTs play roles in disease production by S. dysenteriae type 1 and E. coli, respectively, was recently reviewed (169, 221). The key findings in support of this concept are: (i) S. dysenteriae type 1 disease is less severe in humans and monkeys challenged orally with Shiga toxin-negative mutants than with the wild-type strain (68, 124); (ii) parenteral injection of purified SLT-IIe causes edema disease of swine (136); (iii) SLT-producing E. coli strains are epidemiologically associated with hemorrhagic colitis and HUS (reviewed in reference 108); (iv) transduction of E. coli RDEC-1, an enteroadherent rabbit diarrheal pathogen, with an SLT-I converting phage renders the transductant capable of causing colitis in rabbits (210); and (v) death and kidney damage in streptomycin-treated mice fed certain EHEC strains are completely ablated by passive administration of anti-SLT antibodies (129, 230).

EHlys.

Enterohemolysins (EHlys) lyse washed but not unwashed erythrocytes and are characteristically but not exclusively synthesized by SLT-producing E. coli (13). Three types of EHlys have been described (13, 14, 200). Production of EHly1 is associated with a temperate phage from the non-SLT-producing E. coli O26:H1 strain C3888. The toxin is expressed as a 60-kDa protein in the outer membrane of the wild-type strain as well as an EHly1-producing recombinant (12). However, expression of the cloned gene for EHly1 results in a 33- to 35-kDa protein that is immunologically related to the larger protein (219). Thus, EHly1 may be a dimer in vivo. The biochemical characteristics of EHly2 have not been described. A third hemolysin is encoded on the large plasmid of EHEC O157:H7 strains and is related to the α-hemolysin of E. coli (200; also see below). What role if any these EHlys play in the pathogenesis of EHEC disease remains to be determined.

EIEC

Enteroinvasive E. coli strains cause occasional outbreaks of food-associated diarrhea and dysentery in adults in developed nations and provoke similar syndromes in children in lesser-developed nations (reviewed in reference 123). These organisms share both antigenic and pathogenic traits with Shigella spp. Like Shigella spp., EIEC are able to penetrate and multiply within colonic epithelial cells, and both groups of organisms harbor a large plasmid that encodes several products responsible for the invasive phenotype.

The possibility that an enterotoxin and a cytotoxin may play roles in the pathogenesis of EIEC-mediated diarrhea and dysentery, respectively, was proposed by Fasano and colleagues (61), who demonstrated the presence of such activities in culture filtrates and cell lysates of EIEC strains. The enterotoxin described by these authors induces fluid secretion in rabbit loops and causes increases in short circuit current in Ussing chamber experiments (61). The enterotoxin is 68 to 80 kDa, and its production is repressed by iron but not dependent on the presence of the large plasmid (61). The Vero cell cytotoxic activity of EIEC that is detected at low levels is distinct from the enterotoxic activity and is not neutralized by anti-SLT-I or anti-SLT-II antibodies. No experimental data on the function of the enterotoxin or the cytotoxin(s) in EIEC disease have been reported.

EAggEC

EAggEC, which do not produce the LT, ST, or SLT toxins described above, are named for their pattern of adherence to HEp-2 cells in tissue culture (159). EAggEC are isolated significantly more frequently from infants and young children with diarrhea than from age-matched normal controls, and their predilection to cause chronic diarrhea of more than 2 weeks in duration approaches statistical significance (15, 16).

Culture filtrates from several EAggEC isolates were shown to contain a low-molecular-weight, plasmid-determined, partially heat-stable, protease-sensitive enterotoxin, designated EAST1, that induced changes in potential difference and short circuit current in adult rabbit ileal mucosa in an Ussing chamber model but did not react in immunoassays for ST (196). Cloning and sequencing identified a gene, designated astA, that encodes a 38-amino-acid polypeptide and determines EAST1 activity (197). The deduced amino acid sequence of EAST1 exhibits homology with the receptor-binding domains of STI and guanylin, but, like guanylin, EAST1 contains only four cysteine residues. Although EAST1 has not yet been purified, culture filtrates containing the expressed recombinant EAST1 caused an increase in the concentration of cyclic GMP in rabbit ileal mucosa. EAST1 is viewed, therefore, as a member of the STI enterotoxin family, and it is presumed to activate guanylate cyclase C by a mechanism similar to STI or guanylin.

OTHER TOXINS OF E. COLI ISOLATED FROM INTESTINAL AND EXTRAINTESTINAL DISEASE

CNF1 and CNF2

Structure-Function.

Two types of cytotoxic necrotizing factors (CNFs), CNF1 and CNF2, have been detected in extracts of E. coli strains isolated from humans or animals with diarrhea, septicemia, or urinary tract infections (26, 27, 43-45). The CNFs are proteins of 110 to 115 kDa (26, 45) that are lethal for mice (both toxins [44]), chickens (CNF2 [44]), and lambs (CNF2 [46]) and necrotic for rabbit skin (both toxins [43]) or mouse footpads (CNF2 only [45]). These toxins cause multinucleation and actin polymerization of tissue culture cells (26, 67, 175, 177). CNF1 and CNF2 are cross-reactive in neutralization assays (43), although the neutralization titers against the homologous toxins are greater than against the heterologous ones. CNF1 is chromosomally encoded, whereas the determinant for CNF2 is located on F-like "Vir" plasmids (175).

Recently, the genes for CNF1 and CNF2 were cloned and sequenced (58, 177). The two proteins are composed of 1,014 amino acids and share 85% identical and 99% conserved residues. The N-terminal halves of the CNFs contain stretches of amino acids that are homologous to amino acids in the dermonecrotic toxin of Pasteurella multocida (PMT [58, 177]), whereas a small stretch of the C termini of the two toxins shares homology with the dermonecrotic toxin of Bordetella pertussis (233). That the CNFs and PMT are related is particularly intriguing because the CNFs cause multinucleation of tissue culture cells (26, 175), and PMT also affects cell division, as indicated by its potent mitogenic activity (188).

The mechanism of CNF2-induced actin polymerization appears to involve the modification of Rho, a group of small GTP-binding proteins. This modification was demonstrated by a shift toward a higher molecular weight of labeled Rho in cells treated with CNF2 (177). A similar shift in Rho was observed when CNF2 and RhoA were coexpressed in E. coli (177). The latter finding suggests that a direct interaction occurs between CNF2 and the Rho proteins and that the alteration of Rho proteins by CNF2 does not require prior modification and processing of Rho protein by enzymes in eukaryotic cells. Because the effects of CNF2 on the actin cytoskeleton mimic those observed when a constitutively activated mutant RhoA is microinjected into cells, CNF2 may covalently modify the structure of Rho in such a way that the modified Rho protein is functionally hyperactive as compared to the native form (177).

Association with Disease in Humans and Animals.

About 99% of CNF1-positive E. coli strains are hemolytic (28, 45). These CNF1-positive, α-hemolysin-positive strains are not only associated with intestinal and extraintestinal infections in humans but are also isolated from diarrheal stools and blood of calves and pigs (44, 78, 236). CNF2-producing strains carry virulence determinants for colonization of the calf and lamb intestine and, consequently, are only isolated from these domestic animals (176). CNF2-producing strains have been incriminated as causes of septicemia and diarrhea in these animals (46, 176, 237). More recently, CNF1 was shown to induce HEp-2 cells to phagocytose noninvasive bacteria, and this stress-fiber-related activity of CNF1 may permit noninvasive CNF-producing bacteria to be taken up by epithelial cells in the gut (60). However, neither CNF1 nor CNF2 has as yet been proven to play a role in the pathogenesis of intestinal or extraintestinal diseases in humans or animals.

CLDT

Some E. coli strains isolated from humans with diarrhea produce cytolethal distending toxin (CLDT) that causes progressive distention and ultimately death of Vero, HeLa, HEp-2, or CHO cells (107). The distending activity is generally evident by 48 to 72 h of culture, and the pattern is distinct from the effects of other E. coli toxins that affect the particular cell line (107). Furthermore, the E. coli strains that produce CLDT do not belong exclusively to any one of the above diarrheagenic E. coli groups. Similar CLDT activity has been described for certain Shigella isolates (105) as well as Campylobacter jejuni strains (106).

The genes for CLDT have been cloned and sequenced in two laboratories (181, 202). The two groups started with different E. coli strains and each identified three adjacent or slightly overlapping genes, cdtA, cdtB, and cdtC, whose products appear to be required for CLDT synthesis. The predicted sizes of the cdtA, cdtB, and cdtC products are similar in the two reports: 25.5 and 27.7, 29.8 and 29.5, 20.3 and 19.9 kDa, respectively. Nonetheless, there is considerable predicted amino acid sequence divergence between the three Cdt proteins from the two E. coli strains (181). The predicted Cdt proteins do not share homologies with other known proteins (181, 202).

CLDT of E. coli transiently increases cylic AMP levels in CHO cells (107) but does not increase vascular permeability in rabbit skin (107), a trait typically associated with cyclic AMP-elevating toxins. Furthermore, E. coli CLDT is negative in the rabbit ligated ileal loop and suckling mouse assays. Thus, there is no biological evidence to date that CLDT plays a role in the pathogenesis of E. coli diarrheal disease.

α-Hemolysin

α-Hemolysin is produced by pathogenic strains of E. coli that belong to a restricted set of O serogroups and are associated with extraintestinal infections (e.g., urinary tract, blood, etc.) (11). E. coli strains that make α-hemolysin form colonies on blood agar that are surrounded by a zone of clear lysis. α-Hemolysin has a broad spectrum of activity against erythrocytes from several species, but it also causes cytolytic and toxic effects on other cell types. The activity of α-hemolysin is neutralized by specific antiserum. α-Hemolysin is distinguished from enterohemolysin (described above) and from several other partially characterized hemolysins of E. coli (e.g., β-hemolysin, τ-hemolysin, HLY1, HLY2, etc.) by several criteria, including extracellular localization, biochemical and immunochemical properties, and specificity of action for particular animal species and cell types.

Structure-Function.

α-Hemolysin is the prototype for the RTX family (name explained below) of pore-forming, cytolytic toxins that are produced by several genera of gram-negative bacteria, including Enterobacter, Proteus, Morganella, Pasteurella, Bordetella, and Actinobacillus (143). The name RTX (for repeats in toxin) derives from a common structural motif in these toxins, i.e., tandem arrays of a glycine-rich, nine-amino-acid sequence, L-X-G-G-X-G-(N/D)-D-X, that is discussed below. Individual RTX toxins differ in target specificity for cell types and animal species; therefore, some were named as hemolysins and others as leukotoxins. Some enteroaggegative E. coli strains produce a nonhemolytic toxin that is antigenically related to α-hemolysin (7).

The hlyCABD operon of E. coli determines production and extracellular secretion of biologically active α-hemolysin (117). It includes the structural gene hlyA for α-hemolysin, the gene hlyC required for posttranslational modification and activation of α-hemolysin, and genes hlyB and hlyD, which are necessary for the signal sequence-independent export of α-hemolysin to the extracellular milieu by a pathway that bypasses the periplasm. The HlyB and HlyD proteins that mediate export of α-hemolysin belong to the ABC superfamily of bacterial transporters. In some strains of E. coli hlyCABD is carried by plasmids that are heterogeneous with respect to size, incompatibility group, and conjugational capacity, but in other strains hlyCABD is located on the bacterial chromosome (11). The high frequency of spontaneous loss of hlyCABD and the presence of insertion sequences flanking the operon suggest that hlyCABD may be part of a transposon in E. coli.

Regulation of the hlyCABD operon in plasmids is influenced by hlyR, an upstream cis-acting element that acts in an orientation-dependent manner to enhance transcription and transcriptional readthrough beyond hlyD (20). Additional regulatory factors include the chromosomal gene hlyT (allelic with rfaH and srfB), which may encode a transcriptional activator for the hlyCABD operon, and the outer membrane protein TolC, which is necessary for secretion of α-hemolysin. In some strains iron-dependent repression of α-hemolysin production occurs and is mediated by the product of the fur gene (82).

Translation of the nucleotide sequence of hlyA reveals that α-hemolysin is a 110-kDa polypeptide consisting of 1,024 amino acid residues without a classical amino-terminal signal sequence (117). Further analysis shows several distinct structural and functional domains arrayed as follows along the polypeptide chain: an amino-terminal amphipathic helix; a set of 10 amphipathic or hydrophobic putative transmembrane helices between residues 130 and 450 that participate in pore formation in target membranes; residues Lys-564 and Lys-690, which serve as sites for covalent modification of α-hemolysin by fatty acylation of their ε amino groups; a calcium-binding RTX region that contains 11 glycine-rich repeat sequences between residues 739 and 849 and is necessary for receptor-independent, initial binding of α-hemolysin to eukaryotic membranes; and a carboxyl-terminal sequence of about 50 residues that is essential for export of α-hemolysin (20, 117, 143, 216).

The immediate translation product of hlyA is a protoxin that lacks hemolytic activity. Fatty acylation of residues Lys-564 and Lys-690 occurs by a pathway that requires acylated acyl carrier protein as the fatty acyl donor plus large and possibly stoichiometric amounts of HlyC; and modification of both Lys-564 and Lys-690 is necessary to activate the hemolytic activity of α-hemolysin (216). The export process for the hemolysin is complex (20, 117). Both the protoxin form and the activated form of α-hemolysin are export competent. The carboxyl-terminal region of HlyB, apparently located in the cytoplasm, has a typical ATP-binding motif, and HlyB is believed to recognize cytoplasmic α-hemolysin and transport it across the cytoplasmic membrane in a process driven by hydrolysis of ATP. HlyD and TolC are also required for export of α-hemolysin, but their roles are not as clearly defined. α-Hemolysin apparently does not enter the periplasm during the export process, and HlyD may function as a bridge and participate in the delivery of α-hemolysin directly from the inner membrane to the outer membrane.

Role in Pathogenesis.

Although α-hemolysin is an important virulence factor for some pathogenic strains of E. coli, none of the signs and symptoms characteristic of infections is attributable solely to the activity of α-hemolysin (11, 143). Expression of α-hemolysin by nonpathogenic strains of E. coli does not render them fully virulent, and production of α-hemolysin is only one of several traits required for full virulence of E. coli. In a rodent model of peritonitis, strains of E. coli that do not make α-hemolysin are 100- to 1,000-fold less virulent than isogenic strains that do make α-hemolysin. In a perfused rabbit lung model, small amounts of α-hemolysin induce thromboxane-mediated pulmonary hypertension and cause intravascular release of leukotrienes. α-Hemolysin also exhibits a wide variety of effects on isolated cells and subcellular organelles in vitro.

In epidemiological studies of E. coli from humans, production of α-hemolysin is most prevalent among isolates from extraintestinal sites of infection and is often linked with expression of other virulence-associated traits such as P fimbriae, serum resistance, particular capsular antigens, and a highly toxic phenotype for mice (11). In uropathogenic strains of O4 and O6 serogroups, determinants for P fimbriae, CNF1, and α-hemolysin may be present as part of a single mobile genetic element (58).

SALMONELLA TOXINS

Cytotoxins

Several laboratories have described cytotoxic activity associated with Salmonella species (4, 6, 111, 113, 116, 146, 164). In the most extensive survey of cytotoxin production by Salmonella spp., Ashkenazi and colleagues (4) reported that cell lysates of all 133 strains tested exhibited some level of toxicity for HeLa cells. The toxic material(s) was 56 to 78 kDa and could not be neutralized by antibodies to SLT-I or SLT-II, nor did slt probes hybridize to total DNA from the salmonellae. Cytotoxic activity was heat labile and trypsin sensitive. Whether the activity observed by Ashkenazi and colleagues represents one or more toxins was not determined, but the authors did note significant differences in cytotoxic levels among the species of Salmonella tested.

The relationship between the cytotoxin(s) identified by Ashkenazi and the heat-labile, protein-synthesis-inhibiting Salmonella enteritidis cytotoxin studied by Koo and Peterson (116) is unclear, although the latter toxin is also not neutralized by anti-Shiga toxin. By contrast, other laboratories reported that cytotoxic material from three strains of S. enteritidis and one strain of Salmonella typhimurium was neutralized by anti-Shiga toxin antibodies (111, 164). No function for any of these cytotoxins in pathogenesis has been defined.

Enterotoxins

Several investigators have reported that heat-labile enterotoxins are produced by S. typhimurium, S. enteritidis, S. typhi, or other Salmonella species (6, 25, 66, 110, 152, 192). Few have been purified; little information is available concerning their mode(s) of action or role(s) in pathogenesis; and they appear to be heterogeneous. Single reports about a decade ago described a cytotonic enterotoxin from S. typhimurium that is antigenically and structurally similar to CT (66) as well as enterotoxin(s) and cytotoxin(s) from S. typhimurium and S. enteritidis that are not neutralized by hyperimmune antisera against CT or LT (6). Recent reports from India described the purification of high-molecular-weight enterotoxic polypeptides (>100 kDa) from Salmonella stanley and Salmonella newport that are neutralized by homologous antisera but not by antisera prepared against CT, LT, or Shiga toxin (90, 204).

The most extensively studied S. typhimurium enterotoxin, Stn, is a heat-labile protein that resembles CT and LT in biological activities, is neutralized by anti-CT antibodies, but appears to be structurally different from CT and LT (191, 192). The stn gene, which partially overlaps the hydHG operon for regulation of labile hydrogenase activity in S. typhimurium, encodes a 29-kDa polypeptide that exhibits a small region of similarity with other ADP-ribosylating bacterial toxins (34). Recombinant Stn partially purified from E. coli causes secretion in ligated rabbit ileal segments and increases intracellular cyclic AMP accumulation in CHO cells, and its activity is blocked by ganglioside GM1 or by affinity-purified antibody against CT (32). Antiserum raised against synthetic oligopeptides corresponding with the deduced Stn sequence cross-react with a 29-kDa antigen in cell lysates from isolates of Salmonella, Klebsiella, Enterobacter, and Citrobacter (35). Wild-type stn uses the rare initiation codon TTG, and expression of stn or various stn gene fusions in E. coli is improved by using an ATG start codon, an optimal ribosome-binding site, and the strong phage T7 promoter (33, 34, 184). Purification of recombinant Stn or related fusion proteins to homogeneity and detailed analysis of its mode of action and structure-function relationships have not yet been accomplished.

Salmolysin

Libby and colleagues made the observation that low-passage clinical isolates of Salmonella are occasionally hemolytic on blood agar (127). These investigators cloned, sequenced, and expressed the slyA (salmolysin) gene that is responsible for this hemolytic activity. The deduced amino acid sequence of slyA indicated a protein of 16.7 kDa that is not highly related to any other sequences in GenBank. The slyA gene is present in every strain of Salmonella tested, and homologs are detected by Southern hybridization in Shigella serotypes and enteroinvasive E. coli, but not in other members of the Enterobacteriaceae. Partially purified salmolysin is heat sensitive and fully active in the presence of EDTA (unlike the α-hemolysin of uropathogenic E. coli) and does not contain detectable phospholipase or cholesterol oxidase activity. Salmolysin may be a pore-forming toxin, as suggested by its general lytic effect on cultured mammalian cells (cited in reference 127 as data not shown). That salmolysin may play a critical role in Salmonella virulence in the mouse is indicated by the observations that the median lethal dose of an slyA mutant of S. typhimurium is 10³- to 10⁴-fold higher (depending on the route) than that of the wild-type parent and that this mutant survives less well in protease-peptone-elicited mouse peritoneal macrophages than does the parent strain (127).

CONCLUSIONS

Several of the toxins discussed in this chapter have been unequivocally shown to play a key role in the pathogenesis of disease. These include E. coli LTh-I, LTp-I, and STI. For other toxins, such as the SLTs produced by EHEC and the E. coli α-hemolysin, the data are highly supportive that the toxins are critical for the development of specific diseases. However, the information on the pathogenic potential of the other toxins described here is limited or nonexistent. Generally, these gaps in knowledge reflect either the lack of suitable animal models to test hypotheses about virulence mechanisms or the absence of purified protein toxins with which to conduct experiments.

Significant strides in the molecular analyses of several of the toxins have occurred in the last 10 years. The genes have been cloned and sequenced for E. coli LT, ST, SLT, some of the enterohemolysins, EAST1, CNF, CLDT, and α-hemolysin, as well as Salmonella enterotoxin and salmolysin. Purification schemes to obtain homogeneous preparations of the LTs, STs, SLTs, and E. coli α-hemolysin have been detailed in the literature, and monoclonal and/or monospecific antibodies are available for structure-function analysis of most of these latter toxins. Indeed, in the last 2 years, the crystal structures of LTp-I, STI, the B pentamer of SLT-I, and the Shiga holotoxin (synonymous with SLT-I) have been reported. Furthermore, our understanding of the mode of action of some of the E. coli toxins is quite advanced. We know how LT, STI, and SLT intoxicate cells at the molecular level as well as the intracellular target(s) for the CNFs.

In spite of the progress that has been made in dissecting the genetics and biochemistry of several of the toxins presented here, efforts at using this information for development of vaccines and therapeutic intervention strategies have not kept pace. Indeed, several laboratories have tried and continue to try to design effective vaccines to prevent ETEC-induced diarrhea in humans. Research is also under way to engineer antibodies or receptor analogs to halt the development of HUS after EHEC infection. However, the most practical and far-reaching outcome from studies of how bacterial toxins work is the formulation of oral rehydration solutions to counter the intestinal secretion of fluid and electrolytes induced by CT, LT-I, and STI. The use of oral rehydration solutions has had an enormous impact on the health of children in lesser-developed countries and serves as an example of the power of applying fundamental observations on pathogenesis and pathophysiology toward developing more effective methods to treat or prevent infectious diseases.

ACKNOWLEDGMENTS

Research in the authors’ laboratories on toxins described in this chapter was supported by National Institutes of Health grants R01-AI14107 and R01-AI31940 to R.K.H. and by National Institutes of Health grant R01-AI20148 and Department of Agriculture grant 9203162 to A.O’B.

NOTE ADDED IN PROOF

The authors strongly support the proposal for a new, simplified, unified nomenclature for the SLTs (letter, ASM News, 1996, in press). This proposed change reflects the familial relationship of the SLTs to the prototypic Shiga toxin of S. dysenteriae type 1. With this nomenclature, stx will replace slt in all gene designations. The new protein terms will be as follows: Shiga toxin = Stx; SLT-I = Stx1; SLT-II = Stx2; SLT-IIc = Stx2c; SLT-IIe = Stx2e. To avoid confusion between past and current nomenclature, the Vero toxin nomenclature for the proteins may be used providing there is cross reference to Shiga toxin.