Cytotoxic Necrotizing Factors: Rho-Activating Toxins from Escherichia coli

GUDULA SCHMIDT AND KLAUS AKTORIES*

[SECTION EDITOR: ALISON O'BRIEN]

Posted April 12, 2004

Institut für Experimentelle und Klinische Pharmakologie und Toxikologie der Albert-Ludwigs-Universität Freiburg, Albertstr. 25, D-79104 Freiburg, Germany

Phone: 0049 761 203 5301, Fax: 0049 761 203 5311, E-mail: This e-mail address is being protected from spambots. You need JavaScript enabled to view it

Many bacterial protein toxins, which are important virulence factors and causative agents of human and/or animal diseases, target low-molecular-mass GTP-binding proteins of the hosts. Remarkably, toxins affect these GTPases in a bidirectional manner; e.g., some toxins inhibit and some activate the GTPases. Here, we review the Escherichia coli toxins called cytotoxic necrotizing factors (CNFs), which cause activation of Rho GTPases. We describe their modes of action, structure-function relationships, and roles in disease. Excellent reviews of Rho-activating toxins have been published (3, 23).

Rho GTPases

Rho GTPases, the targets of CNFs, belong to the Ras superfamily of low-molecular-mass GTPases and act as molecular switches in various signaling pathways. Rho GTPases are ubiquitously expressed. The family contains [mt]15 members, including RhoA, -B, -C, -D, -E, and -G; Cdc42; and Rac1, -2, and -3, which show [mt]50% sequence identity (for a review, see reference 54). Like all GTP-binding proteins, they are active in the GTP-bound form and inactive with GDP bound. Activation occurs by nucleotide exchange catalyzed by guanine nucleotide exchange factors (GEFs) and is inhibited by the binding of guanine nucleotide dissociation inhibitors. Inactivation by hydrolysis of the bound GTP is catalyzed by GTPase-activating proteins (GAPs) (53, 55, 57).

Low-molecular-mass GTPases of the Rho family are known as master regulators of the actin cytoskeleton. Moreover, they are involved in various signal transduction processes, from transcriptional activation, cell cycle progression, and cell transformation to apoptosis. Many Rho GTPase effectors involved in these processes have been identified, including protein and lipid kinases, phospholipase D, scaffolding proteins, and numerous adaptor proteins (2, 14). Rho GTPases are the eukaryotic targets of a variety of bacterial protein toxins that either inhibit or activate the Rho proteins. CNF1 and -2 from E. coli are Rho-activating toxins. CNFs deamidate a specific glutamine (e.g., Gln63 in RhoA), which is crucial for GTP hydrolysis. Thus, Rho GTPases are persistently activated, and signaling processes, which are governed by Rho GTPases, are disturbed (1).

Rho PROTEINS AS TARGETS OF BACTERIAL TOXINS

In addition to their cell-biological importance, Rho proteins have been recognized as major eukaryotic targets for a variety of bacterial protein toxins, including activating and inactivating toxins. Some toxins catalyze a covalent modification of the GTPases, whereas others modulate their function (summarized in Fig. 1). For example, C3-like toxins from Clostridium botulinum, Clostridium limosum, and Staphylococcus aureus ADP-ribosylate small GTPases of the Rho family (e.g., at Asn41 of RhoA [51]), thereby inactivating them. A second family of Rho protein-inactivating toxins, the large clostridial cytotoxins, modify the GTPases by glucosylation (6, 26, 27). Members of this toxin family are Clostridium difficile toxins A and B, including various isoforms; the lethal and the hemorrhagic toxins from Clostridium sordellii; and the alpha toxin from Clostridium novyi. These toxins modify a highly conserved threonine residue (e.g., Thr37 in RhoA) in the switch 1 region of the GTPases, which is involved in Mg²⁺, nucleotide, and effector binding. Thus, modification of this threonine causes inactivation of the GTPases (20, 50).

Fig. 1Bacterial toxins acting on Rho GTPases. A large variety of bacterial toxins target small GTPases of the Rho family. C3-like exoenzymes ADP ribosylate RhoA at asparagine 41, inhibiting the biological functions of the GTPase. All members of the Rho subfamily are glucosylated by large clostridial cytotoxins (e.g., C. difficile toxins A and B). Recently, it was shown that the Yersinia outer protein T (YopT) is a member of a new family of cysteine proteases that cleave RhoA in front of the isoprenylated cysteine at the C terminus, blocking membrane binding and inactivating the GTPase. Rho is constitutively activated by CNF1 and -2 of E. coli. The CNF1-related DNT from Bordetella species, which transglutaminates glutamine 63 of RhoA, activates the GTPase. Rho proteins are not exclusively covalently modified by bacterial toxins. Some bacterial effectors, like Salmonella SptP, Pseudomonas exoenzyme S (ExoS), and Yersinia outer protein E (YopE), modulate the activities of Rho GTPases by acting as regulatory proteins with GAP activity or with GEF functions ( Salmonella SopE).

Rho proteins are also activated due to covalent modification catalyzed by bacterial protein toxins like CNF1 and CNF2 from E. coli (described below). Moreover, they are activated by the dermonecrotic toxin (DNT) from Bordetella species. DNT transaminates Rho GTPases at the same glutamine residue that is deamidated by CNFs (24, 37). Recent studies indicate that Rho proteins are not exclusively covalently modified by bacterial toxins. Some bacterial effectors, like the Salmonella SopEs and SptP, modulate the activity of Rho GTPases by acting as regulatory proteins with GAP (SptP) or GEF (SopEs) functions (19, 52).

CNFS AFFECT A WIDE VARIETY OF CELLS—MECHANISM OF UPTAKE

Twenty years ago, Caprioli and coworkers isolated a toxin from E. coli obtained from enteritis-affected children. Because of the necrotizing effects on rabbit skin, they called the toxin CNF (7). CNF1, and later the homologue CNF2, first named Vir cytotoxin (40) (90% sequence identity [41]), were found in various pathogenic E. coli strains isolated from animals (e.g., piglets and calves [12]) and humans (11).Whereas CNF1 is chromosomally encoded, CNF2 is located on transmissible plasmids (39). Recently, the gene for a CNF of Yersinia pseudotuberculosis (CNF_Y) was identified, with 65% sequence identity to CNF1 over the whole molecule and located on the chromosome of the bacteria (35).

All three toxins are single-chain proteins with molecular masses of ~115 kDa. They are constructed like AB toxins, with a cell-binding domain at the N terminus (amino acids 53 to 190) and a C-terminal catalytic domain (amino acids 720 to 1014) (31). The central part contains two hydrophobic helices (amino acids 350 to 412), which are separated by a short loop (amino acids 373 to 386). This part appears to be involved in membrane translocation (43).

CNFs are cytotoxic for a wide variety of cells, including 3T3 fibroblasts, Chinese hamster ovary cells, Vero cells, HeLa cells, and cell lines of neuronal origin. This implies that a commonly expressed receptor is responsible for the uptake of CNF1. Recently, the receptor for the entry of the toxin into cells was suggested to be the 37-kDa laminin receptor precursor, which is ubiquitously expressed and interacts with the N-terminal part of the toxin in the yeast two-hybrid screen (8). The uptake of CNF appears to occur by clathrin-dependent and -independent endocytosis, followed by cell entry from an acidic endosomal compartment (9).

MODE OF ACTION OF CNFS

CNFs lead to the enlargement and flattening of culture cells and the formation of filopodia, membrane ruffles, and a dense network of actin stress fibers, indicating activation of RhoA proteins (15). Moreover, CNFs change the migration behavior of RhoA in sodium dodecyl sulfate-polyacrylamide gel electrophoresis (39), a finding that, quite early, suggested that the toxins catalyze a covalent modification of RhoA. Two groups independently showed that CNF1 deamidates glutamine 63 of RhoA (17, 48). This amino acid is essential for the GTP hydrolysis catalyzed by the GTPase. Thus, GTP hydrolysis is also blocked after treatment of Rho with CNF1 or -2 or CNF_Y in the presence of a GAP, and RhoA is held constitutively active. Similarly, Rac and Cdc42 are deamidated by CNF (34). Deamidation occurs at the equivalent amino acid residue, glutamine 61. CNF is highly specific for Rho GTPases. The minimal Rho sequence necessary for deamidation or transglutamination by CNF1 is a peptide covering mainly the switch II region (D59 to D78) of RhoA (16, 33).

Besides their deamidating activity, CNFs have been shown to catalyze the transglutamination of the same glutamine residue of Rho. This modification also leads to permanent activation of the GTPases (49). A similar mechanism of Rho modification has been reported for the CNF1-related DNT from Bordetella species (23, 24, 28, 46).

DNT, which is produced by Bordetella pertussis, Bordetella parapertussis, and Bordetella bronchiseptica,got its name from dermonecrotic effects caused by intradermal injection (25). DNT is a large, 160-kDa heat-labile protein which shows significant sequence similarity with CNF only at the C terminus of the toxin (the catalytic domain). It has been shown that DNT causes the polyamination of glutamines 63 and 61 of Rho GTPases in the presence of primary amines. However, it appears that DNT prefers transglutamination whereas CNF is primarily a deamidase (Fig. 2) (46). Recently, putrescine, spermidine, and spermine have been identified as invivo substrates for transglutamination (37, 47). At least in vitro, lysine is a preferred substrate for transglutamination by DNT (47).

Fig. 2Molecular mechanism of Rho activation by CNF and DNT. Glutamine 63 of RhoA (Gln 61 of Rac and Cdc42) is essential for the hydrolysis of bound GTP. CNF and DNT deamidate this glutamine residue, creating glutamic acid, and the GTP-hydrolyzing activity of the GTPase is blocked. In the presence of primary amines, the toxins can polyaminate Rho at the same residue, thereby also blocking hydrolysis of GTP. Polyamination is the preferred activity of DNT, whereas CNF is a much better deamidase.

CELL-BIOLOGICAL EFFECTS

Cultured mammalian cells treated with CNFs are characterized by dramatic changes in actin- containing structures, including stress fibers, lamellipodia, and filopodia. Most striking is the formation of multinucleation in these cells (40). It is not clear whether activation of all Rho GTPases is needed for this effect or which GTPases are involved. Multinucleation might be caused by the blocking of cell division without any change in nuclear cycling or by an increased rate of nuclear cycles without cleavage of the cells (10). Recent data show that CNF2 uncouples S phase from mitosis. It induces the synthesis of cytoplasmic material and allows the start of another S phase without first completing mitosis.

Surprisingly, c-Jun kinase activity, which is controlled by Rac, was found to be only transiently increased after CNF treatment of cells, although Rho GTPases are constitutively activated (8). Recently, the reason for the transient activation was identified. It was shown that Rac (but not [or much less] RhoA or Cdc42), which is involved in c-Jun kinase activation, is markedly degraded by a proteasome-dependent pathway in CNF1-treated HEK293 cells (13, 32). Thus, it appears that the targeted cell is able to block the persistent activation of deamidated Rac by rapid degradation. At first glance, this coupling of activation and inactivation is comparable to the tight regulation of the GTPases by Salmonella strains producing a GEF (Rac activation) as well as a GAP (Rac inactivation) for Rho GTPases. So far, it is not clear whether degradation of activated Rac is part of a general mechanism in mammalian cells to limit "overactivation" of GTPases (see below).

STRUCTURE-FUNCTION RELATIONSHIP

Recently, the crystal structure of the enzyme domain of CNF1 was solved (5). Although the folding of the catalytic domain is unique, the crystal structure analysis nicely confirmed the previous suggestion that CNF belongs to the catalytic-triad family. In fact, it was shown that a cysteine (Cys 866) and a histidine (His 881) residue are essential for enzyme activity (49). As identified from the crystal structure, the third catalytic residue appears to be a valine (Val 833), a finding which is rather unusual among catalytic-triad enzymes (Fig. 3) (5). A reason for the high specificity of CNF may be the existence of a deep cleft in the molecule, with the catalytic cysteine at the bottom (5). The catalytic domain of CNF1 exhibits a protein fold different from those of mammalian transglutaminases (42).Thus, the structure of the CNF1 catalytic region contributes to the idea of a convergent evolution of the toxin and mammalian enzymes with the same mechanism rather than gene transfer.

Fig. 3Sequence homology between CNF and DNT.CNF and DNT consist of 1,014 and 1,451 amino acids, respectively. The toxins show homology within their catalytic domains (ΔCNF, amino acids 709 to 1014, and ΔDNT, amino acids 1136 to 1451), which are located at the C termini of the proteins. The highest sequence similarity is observed in a stretch of 64 amino acid residues which covers residues 1250 through 1314 of DNT, showing ~45% sequence identity, whereas the amino acid sequence of the whole catalytic domain of DNT is only ~13% identical with the sequence of the catalytic domain of CNF1. Cysteine 866 and histidine 881 are essential for enzyme activity in CNF1. As deduced from the crystal structure, a third "catalytic" residue appears to be valine 833.

CNFS AS VIRULENCE FACTORS

Recently, it was suggested by multiple studies that CNFs, which are produced by several but not all enteropathogenic and uropathogenic E. coli strains (7), are important virulence factors for E. coli-caused diseases. However, the roles of CNFs in inflammation and tissue damage seem to be tissue dependent. Using CNF1-deficient E. coli strains, it was shown that the colonization of and tissue damage to the urinary tracts of mice was less pronounced than with the isogenic wild-type strains (45). Moreover, it was reported that tissue damage to rat prostates is more severe with CNF1-producing uropathogenic E. coli strains than after infection with isogenic toxin-negative strains (44). In contrast, no difference in lung and serosal inflammation induced by CNF-producing and -deficient strains has been observed (18). However, the involvement of the toxin in the traversal of the blood-brain barrier in a animal meningitis model has been shown (29). Taken together, several findings support the notion that CNFs are major virulence factors in various infection models.

So far, the pathophysiological mechanisms underlying a potential role as virulence factors are not entirely clear. Several reports of the effects of CNFs obtained in studies with cell cultures and isolated tissues provide initial hints. In Caco-2 cells, CNF1 increases intestinal permeability (21). It causes a decrease in transepithelial electrical resistance, which is accompanied by increased paracellular permeability. Similarly, a marked reduction in tight-junction gate function, accompanied by displacement of the tight-junction proteins occludin and zonula occludens-1 away from the tight-junction membrane, was observed in intestinal T84 cells (22). Thus, the barrier function of epithelial cells is greatly affected by CNF. Surprisingly, a recent finding about the CNF-induced degradation of Rac points in the same direction. It was shown by Doye and coworkers that degradation of Rac, which is induced by CNF, enhances the motility of target cells (13). Moreover, CNF significantly increased the internalization of bacteria by 804G rat bladder carcinoma cells. Again, this effect was observed after the degradation of Rac, indicating that not only the activation of Rho GTPases but also the specific degradation, occurring subsequent to Rho GTPase activation, might be essential for the ability of CNF to act as a virulence factor. Without doubt, the effects of CNFs on immune cells are similarly important for their role as virulence factors. The actin-and-microtubule cytoskeleton is known to play a major role in the activities of cell-mediated immune responses. By activation and degradation of Rho GTPases, CNFs greatly interfere with innate and acquired immunity. For example, it was recently reported that CNF1 increases the cytotoxicity of NK cells (36) and activates T lymphocytes to release several cytokines, which are known to augment inflammatory processes (4).

Rho GTPases are increasingly recognized as essential factors in the development of cancer and metastasis. This fact has initiated a discussion as to whether activation of Rho proteins by CNFs might be involved in tumorigenesis. Moreover, CNF1 increases the expression of the cyclooxygenase 2 (Cox-2) gene in fibroblasts (56). Increased expression of Cox-2 is observed in some types of tumors, e.g., colon carcinoma. Lipid mediators produced by the enzyme are suggested to be responsible for tumor progression (38). Thus, it is feasible that chronic infections with CNF-producing E. coli might carry a tumorigenic potential important for the pathogenesis of some types of cancer (30).