Rho/Ras-GTPase-dependent and -independent activity of clostridial glucosylating toxins

Clostridial glucosylating toxins alter the actin cytoskeleton and intercellular junctions

By inactivating Rho proteins, clostridial glucosylating toxins induce cell rounding, which is accompanied by a loss of actin stress fibres, reorganization of the cortical actin, disruption of the intercellular junctions and subsequently increase in cell barrier permeability. These effects are also reported as cytopathic effects, in contrast to the cytotoxic effects, which result in toxin-induced cell death (Just & Gerhard, 2004; Rupnik & Just, 2006). Differences in actin cytoskeleton modification are observed between the clostridial glucosylating toxins. TcdB induces loss of actin stress fibres and reorganization of focal adhesions accompanied by cytoplasmic retraction and cell rounding, whereas long protrusions radiating around the cell yield a particular ‘actinomorphic’ morphology (Ottlinger & Lin, 1988). TcdA and TcnA cause similar cell alterations, whereas TcsL produces cell rounding with loss of actin stress fibres but without branched membrane protrusions. These different effects might result from the different Rho proteins targeted by the toxins: TcdA, TcdB and TcnA inactivate the three main Rho-GTPases (Rho, Rac and Cdc42) whereas TcsL only modifies Rac (Table 1). However, Rac1 rather than RhoA or Cdc42 has been shown to be the critical target for the cytopathic effects induced by TcdB (Halabi-Cabezon et al., 2008). Although Rac inactivation seems to be the main mechanism responsible for the cytopathic effects, the additional Rho-GTPases modified by TcdA, TcdB and TcnA are probably involved in different actin cytoskeleton alterations from those induced by TcsL.

TcdA and TcdB alter the barrier function of polarized intestinal cells such as Caco-2 and T84 cells by increasing paracellular permeability (Hecht et al., 1988, 1992; Johal et al., 2004). Concomitant with the permeability increase, TcdA and TcdB disrupt apical and basal actin filaments with a subsequent disorganization of the ultrastructure and component distribution (i.e. ZO-1, ZO-2, occludin, claudin) within tight junctions (TJs). TcdB decreases actin–ZO-1 association and disperses ZO-1 and occludin from lipid-rich membrane microdomains, without changing the occludin phosphorylation status (Chen et al., 2002; Nusrat et al., 2001). Since Rho plays an important role in TJ assembly, the effects of TcdB and TcdA on TJs presumably result from Rho glucosylation (Jou et al., 1998; Popoff & Geny, 2009). In contrast, TcsL-82, which inactivates Rac, and TcsL9048, which inactivates both Rac and Cdc42 (Table 1), induce a specific depolymerization of basolateral actin filaments, an increased cell barrier permeability, and a drastic perturbation of adherens junctions (AJs), whereas apical actin and TJs are preserved. Thereby, TcsLs preferentially depolymerize actin filaments in the basolateral cell compartment and alter AJs. TcsLs cause a dramatic redistribution of E-cadherin and α- and β-catenins from the plasma membrane to the cytoplasm, with a reduced and discontinuous distribution of these molecules at the junctional ring and a diffuse localization into the cytosol. Simultaneously, E-cadherin is reduced at the cell surface and is increased in the detergent-soluble fraction. Thereby, TcsLs remove the whole E-cadherin–catenin complex from the membrane and increase paracellular permeability of cell monolayers (Boehm et al., 2006; Richard et al., 1999).

TcsL is a highly potent lethal toxin but its mechanism of lethality is still elusive. Intraperitoneal injection of TcsL in mice induces an increased vascular permeability, mainly at the level of lung vessels, leading to a massive extravasation of blood fluid in the thoracic cavity, dehydration, increase in haematocrit, hypoxia, and finally cardio-respiratory distress. Alterations of lung endothelial cells are apparent by electron microscopy, showing modifications of AJs, which appear fainter and more discontinuous than in controls. VE-cadherin immunostaining is markedly altered, with an interrupted and locally diffuse staining reflecting a redistribution of VE-cadherin from the junctional ring to the cytoplasm (Geny et al., 2007). Thereby, TcsL, which inactivates Rac among the Rho-GTPases, seems to preferentially target endothelial cells in the cardio-respiratory area, to alter the distribution of VE-cadherin, and subsequently to increase the endothelial barrier permeability, leading to local oedema and haemodynamic perturbations probably responsible for lethality.

Which cell signalling pathways downstream of Rac controlling actin polymerization and intercellular junctions are altered by clostridial glucosylating toxins?

Several Rac effectors controlling the actin cytoskeleton and intercellular junctions have been identified (Fig. 1) (Burridge & Wennerberg, 2004; Van Aelst & D’Souza-Schorey, 1997). Multiple signalling pathways downstream of Rac have been characterized. The main role of Rac is to coordinate the formation of lamellipodia, but it is also involved in the control of other actin structures. Rac activates actin polymerization by stimulating PI4P5-kinase (PI4P5K) and through the WAVE/IRSp53 and PAK (P-21-activated kinase) pathways. In addition, Rac and Cdc42 directly act on the activity and assembly of the E-cadherin–catenin complex, forming AJs through a common effector IQGAP (IQ motif containing GTPase-activating protein) (Fukata & Kaibuchi, 2001). IQGAP1 has been shown to interact with β-catenin and the cytoplasmic domain of E-cadherin. A model has been proposed in which Rac and Cdc42 regulate E-cadherin junctions through IQGAP1. Active Rac and Cdc42 in the GTP-bound form interact with IQGAP1 and thus prevent the association of IQGAP1 with β-catenin, which results in the stabilization of E-cadherin–catenin complex. In the GDP-bound form, non-active Rac and Cdc42 do not interact with IQGAP1, which associates with β-catenin, thereby displacing α-catenin from its binding to β-catenin. This leads to a dissociation of α-catenin linked to actin filaments from the E-cadherin–catenin complex, conferring a weak adhesive activity (Bashour et al., 1997; Fukata & Kaibuchi, 2001; Fukata et al., 1999; Kaibuchi et al., 1999a, b). Further work with knockdown Rac and/or IQGAP1 supports the idea that Rac positively regulates cell–cell adhesion through IQGAP1 by promoting the accumulation of actin filaments, E-cadherin and β-catenin at MDCK cell–cell contact sites (Noritake et al., 2004).

The various pathways regulated by Rac have been investigated in response to TcsL treatment. In polarized epithelial cells, TcsL induces a redistribution of the E-cadherin–catenin complexes from the membrane (detergent-insoluble fractions) to the cytoplasm (soluble fractions). In contrast to 12-O-tetradecanoylphorbol 13-acetate (TPA), which induces cell–cell dissociation via inhibition of Rac, resulting in increased association of IQGAP1 with β-catenin and dissociation of α-catenin from E-cadherin (Fukata & Kaibuchi, 2001), TcsL does not change the ratio of catenin–E-cadherin–IQGAP1 complexes, indicating that this toxin does not modify AJ integrity through the Rac/IQGAP pathway (Boehm et al., 2006). Therefore, TcsL-mediated dissociation of E-cadherin from AJs does not seem to involve the Rac-IQGAP1 pathway directly, but rather Rac-dependent cortical actin depolymerization. Indeed, disruption of actin filaments induced by iota toxin or cytochalasin D, which inhibit actin polymerization by directly interacting with actin molecules, also leads to similar AJ alterations including removal of the E-cadherin–catenin complex from the cell surface (Boehm et al., 2006). But how does TcsL-dependent Rac inactivation disorganize cortical actin filaments and E-cadherin junctions? A Rac pathway involves modifications of PAK-LIM-kinase and cofilin phosphorylation, thereby inhibiting actin depolymerization (Arber et al., 1998; Chen & Macara, 2006; Yang et al., 1998). Since Tcs-82 does not change the expression or phosphorylation of cofilin, this pathway is probably not involved in the toxin process (Geny et al., 2010). However, we have found that TcsL-82 induces a rapid dephosphorylation of paxillin and thus decreases the number of focal adhesions (FAs). This effect is Rac dependent, since paxillin dephosphorylation is prevented in TcsL-82-treated cells overexpressing constitutively active Rac1 (Rac^G12V), and cells overexpressing activated Rac1 exhibit a higher number of FAs, a level which is not modified upon intoxication. Paxillin dephosphorylation does not appear to result from inactivation of the Rac major tyrosine kinase, Src or FAK (FA kinase), but rather from an earlier event. Rac inactivation might activate some phosphatase and thereby dephosphorylate proteins from FA complexes such as paxillin (Geny et al., 2010). Several studies have reported the involvement of Rac1 in the control of AJ regulation, probably through reorganization of actin cytoskeleton (Braga et al., 1997, 1999; Fukata & Kaibuchi, 2001; Popoff & Geny, 2009; Takaishi et al., 1997). However, the molecular mechanism(s) leading to the loss of cell adherence and cell–cell interaction upon Rac1 inactivation is (are) not yet fully deciphered. We observed an interaction between FA and AJ complexes as paxillin, talin and β-catenin are co-immunoprecipitated in resting cells. Upon cell intoxication with TcsL-82, and thus Rac1 inactivation, interactions between β-catenin, paxillin and talin are markedly decreased, indicating that FA and AJ complexes and their interactions are loosened. Moreover, protein interactions inside FA complexes also appear to be loosened, as the level of talin, which co-immunoprecipitates with paxillin, is decreased noticeably after cell intoxication (Geny et al., 2010). These findings are in agreement with the critical role of Rac1 in FA complex assembly (Guo et al., 2006) and the recent work of the group of Birukov, which demonstrates such an interaction by showing that FA and AJ proteins can be co-immunoprecipitated and that interactions between FA and AJ complexes via paxillin and β-catenin association involve Rac1 as a regulator (Birukova et al., 2007, 2008).

In addition, we have shown that TcsL-82 induces a rapid decrease (within 1–2 h) in the cellular content of PI4P (phosphatidylinositol 4-phosphate), PI4,5P₂ (phosphatidylinositol 4,5-bisphosphate) and PI3,4,5P₃ (phosphatidylinositol 3,4,5-triphosphate). Polyphosphoinositides regulate numerous cellular processes, including FA formation and modulation of the actin cytoskeleton, two processes modified by TcsL-82 intoxication. Indeed, polyphosphoinositides play a role in membrane anchorage of several proteins or protein complexes, in particular proteins containing a PH (pleckstrin homology) or a PX (phox homology) domain (Ferguson et al., 1995; Lemmon et al., 2002; Simonsen & Stenmark, 2001) as well as complex organization in lipid raft microdomains. Lipid rafts are enriched not only in cholesterol and sphingolipids but also in polyphosphoinositides such as PI4,5P₂ and PI3,4,5P₃ (Bodin et al., 2001; Caroni, 2001; Doughman et al., 2003; Golub & Caroni, 2005; Logan & Mandato, 2006; van Rheenen et al., 2005). FA and AJ molecules also belong to highly organized membrane microdomains (Gaus et al., 2006), and lipid raft content in AJ and FA proteins is modified upon TcsL-82 intoxication. Indeed, the level of talin, a protein that belongs to the protein complex forming FA, decreases in detergent-insoluble membranes of TcsL-82-intoxicated cells compared to that present in control cells. Several AJ proteins, including E-cadherin, β-catenin and p120 catenin, as well as PI4,5P₂, are also decreased in detergent-insoluble membrane structures of TcsL-82-intoxicated cells compared to control resting cells (Geny et al., 2010). These changes in lipid raft protein content are probably related to modifications in polyphosphoinositides of these membrane structures induced by TcsL-82. Moreover, the TcsL-82-dependent decrease in PI4,5P₂, and possibly also in PI3,4,5P₃, is likely to be responsible for actin depolymerization induced by the toxin. Indeed several actin-associated and/or actin-regulating proteins, including gelsolin (Janmey & Stossel, 1987), profilin (Lassing & Lindberg, 1985), α-actinin (Fukami et al., 1994) and vinculin (Burridge & Mangeat, 1984; Menkel et al., 1994), have been reported to interact with PI4,5P₂. For example, association of PI4,5P₂ with vinculin induces a conformational change in vinculin allowing it to interact with talin, which binds actin (Gilmore & Burridge, 1996). It has been proposed that PI4,5P₂ enhances actin polymerization by uncapping actin filaments, thereby increasing the number of free barbed ends, which are the actin polymerizing filament ends (Hartwig et al., 1995).

Rac1 appears to play a central role in controlling polyphosphoinositide cell content since this small GTPase has been shown to associate with and regulate the PI4P5K, which generates PI4,5P₂ (Carpenter et al., 1999; Hartwig et al., 1995; Tolias et al., 1995, 2000). PI4P5K was reported to be present in FAs (Di Paolo et al., 2002; Ling et al., 2002). The molecular mode of action of TcsL-82 in decreasing polyphosphoinositides has not yet been elucidated. The main effect of glucosylation of small GTPases by clostridial glucosylating toxins has been reported to be modified interaction with different effectors (Herrmann et al., 1998; Sehr et al., 1998; Vetter et al., 2000). However, no significant difference was found in the association between PI4P5K and non-glucosylated or glucosylated Rac1 loaded either with GDP or with GTPγS (Geny et al., 2010), but this does not rule out the possibility that TcsL-82 could inactivate PI4P5K via Rac1 at a step other than Rac-PI4P5K interaction. Also, it cannot be excluded that the observed decrease in cellular phosphoinositide content could result from the activation of lipid phosphatases upon TcsL intoxication and Rac inactivation.

Rac seems to be the main target of clostridial glucosylating toxins, and mediates the toxin effects on actin cytoskeleton reorganization. Thereby, TcsL, like TcsL-82, only glucosylates Rac among the Rho-GTPases, and TcdB, which inactivates Rho, Rac and Cdc42 alters the actin cytoskeleton mainly via Rac inactivation rather than RhoA inactivation (Halabi-Cabezon et al., 2008). Therefore, it is likely that clostridial glucosylating toxins trigger a common pathway downstream of Rac including a decrease in phosphoinositides, to induce the actin cytoskeleton disorganization and the subsequent pathological effects, such as the alteration of epithelial and endothelial barrier integrity.

It is noteworthy that TcsL-dependent Rac inactivation through glucosylation at Thr35 does not impair all the Rac downstream signalling pathways (Fig. 2). Rac interacts with more than 31 potential downstream effectors, and it is now known that these effectors use distinct residues within the main docking domains, including switch-I and switch-II regions, and in some cases within the C-terminal area (Bustelo et al., 2007). Indeed, mutation of specific residues results in impaired recognition of only a subset of effectors and prevents the activation of only the corresponding signalling pathways. For instance, Rac amino acids 123–135, known as the insert region, are involved in binding to IQGAP but not to PAK. Tyr40 but not Phe37 from the switch-I is critical for Rac binding to PAK, whereas Phe37 is important for the Rac-dependent change in the actin cytoskeleton (Bishop & Hall, 2000; Bustelo et al., 2007). This could account for the fact that TcsL, which modifies Thr35, impairs a limited number of effectors. However, Rac glucosylation at Thr-35 is reported to block the switch-I domain in its inactive conformation, thus impairing Rac interaction with all the downstream effectors (Genth & Just, 2010; Genth et al., 2008). Since Rho-GTPases including Rac are distributed in various subcellular compartments (Michaelson et al., 2001; Spiering & Hodgson, 2011), it is conceivable that TcsL, as well as the other clostridial glucosylating toxins, preferentially inactivates certain subcellular pools of Rac, which coordinate a restricted number of effectors.

TcdA has also been reported to alter intestinal cell TJs through dephosphorylation of paxillin and FAK (Kim et al., 2009). However, the TcdA-dependent mechanism of paxillin and FAK dephosphorylation seems to be different from that of TcsL. TcdA was reported to directly interact with the catalytic domain of Src, inhibiting its kinase activity towards paxillin and FAK. This process is independent of Rho-GTPase glucosylation, since the glucosyltransferase inhibitor UDP-2′,3′-dialdehyde did not prevent dephosphorylation of paxillin and FAK by TcdA, and also independent of tyrosine phosphatases based on experiments with sodium orthovanadate, a tyrosine phosphatase inhibitor, which was not found to impair TcdA-induced dephosphorylation activity (Kim et al., 2009). Further work is required to better understand the clostridial glucosylating toxin activity downstream of Rho-GTPase and to identify possible alternative mechanisms to Rho-GTPase inhibition in alteration of the actin cytoskeleton and intercellular junctions. In contrast, the recently described TcdA activity consisting of α-tubulin deacetylation leading to microtubule depolymerization, loss of TJs, and proinflammatory cytokine production was found to be Rho-GTPase-dependent (Nam et al., 2010).

Cellular effects induced by clostridial glucosylating toxins independently of Rho/Ras-GTPase inactivation

Clostridial glucosylating toxins have been found to activate the mitogen-activated protein (MAP) kinase (MAPK) pathways, leading to an inflammatory response. TcdA, which glucosylates, and thus inactivates, small GTPases from the Rho family, Rho, Rac and Cdc42, was also reported to activate MAPKs in a monocytic cell line. Activation of ERK1/2 and SAPK/JNK by TcdA was shown to be transient (less than 30 min) whereas that of p38 MAPK, starting within minutes of cellular intoxication, is sustained, lasting for at least 1 h (Warny et al., 2000). The latter MAPK activation was demonstrated to be involved in the inflammatory response accompanying intoxication by TcdA (Kim et al., 2005b; Warny et al., 2000), a response that was not observed during TcsL-82 intoxication in mice (Geny et al., 2007). TcdA-induced activation of p38 MAPK was also shown to lead to p53 activation and finally to apoptosis mostly independently of Rho inactivation in human untransformed NMC460 colonocytes (Kim et al., 2005a). In human mast cells, TcdA and TcdB activate p38 MAPK and ERK1/2, leading to a p38 MAPK-dependent release of interleukin-8 (IL-8) and prostaglandins D₂ and E₂. This process was found to result from toxin entry into cells, but not from a receptor-mediated effect. Although cytoskeleton rearrangement participates in mast cell degranulation, increased secretion of IL-8 and prostaglandins was found to be dependent on the TcdB-mediated p38 MAPK activation and not derived from its actin depolymerization effect (Meyer et al., 2007). This does not rule out the possibility that a Rho-GTPase inactivation-dependent pathway distinct from that of actin depolymerization might be involved. However, ERK1/2-p38 MAPK activation and subsequent IL-8 production induced by TcdA was found to be independent of its glucosyltransferase activity based on the different kinetics of the two activities (Kim et al., 2005b; Warny et al., 2000). MAPK activation possibly results from an initial induction of reactive oxygen species (ROS) by TcdA (Kim et al., 2005b). TcdA also increases IL-8 secretion via activation of the NF-κB pathway (Kim et al., 2006), which is probably mediated by ROS generation from mitochondria independently of Rho inactivation (He et al., 2000, 2002). In addition, a possible role of RhoB activation (Gerhard et al., 2005) subsequent to TcdA-dependent Rho inactivation in NF-κB activation has been hypothesized (Kim et al., 2006).

We have also observed that TcsL-82 activates the three MAPK pathways ERK1/2, p38 MAPK and SAPK/JNK and subsequently the phosphorylation of transcriptional factors such as c-Jun and ATF2. This effect is transient. In HeLa cells, TcsL-82-dependent activation of MAPK pathways occurs within about 1 h, reaches a maximum after 2 h and then decreases (Geny & Popoff, 2009). The kinetics of TcsL-82-dependent activation of MAPK pathways is similar to that of in vivo enzymic activity towards Rac/Ras-GTPases. This raises the question whether MAPK activation is dependent on Rac/Ras-GTPase glucosylation by Tcsl-82. A derivative cell line (DonQ) from the cell line Don has a low content of UDP-glucose and thus is resistant to TcsL-82 intracellular activity. Although glucosylation of Rho/Ras-GTPases and cell rounding are prevented in DonQ cells in contrast to the parental Don cells, TcsL-82 induces MAPK activation in both cell lines (Geny & Popoff, 2009), indicating that glucosylation of Rac/Ras-GTPases is not required for the toxin-dependent MAPK activation. Similar findings have been obtained with TcdA: activation of MAPK by TcdA was reported to be independent of Rho-GTPase. The early p38 MAPK activation by TcdA precedes Rho glucosylation, suggesting that the two effects are not related. It has been proposed that a toxin interaction with an as yet unknown cell-surface receptor and/or intracellular entry of the toxin might trigger p38 MAPK activation (Kim et al., 2005b; Warny et al., 2000). We have tested whether the C-terminal part of TcsL-82, which interacts with a cell-surface receptor, might induce MAPK activation. No effect on MAPK activation has been found with a recombinant TcsL C-terminal protein. The whole toxin is required to induce this effect. In cells pretreated with bafilomycin A1, which blocks the translocation of the N-terminal part of the toxin from acidified endosome to the cytosol, TcsL-82 is unable to trigger the MAPK activation (Geny & Popoff, 2009). Therefore, the N-terminal part of TcsL-82, which contains the enzymic site, has to be translocated into the cytosol to induce the MAPK activation, but independently of Rac/Ras-GTPase glucosylation.

Interestingly, a specific JNK inhibitor, JNK inhibitor II, prevents the TcsL-82-dependent cytoskeleton modifications and actin depolymerization but in a transient manner, whereas inhibitors of ERK1/2 and p38 MAPK are without effect (Geny & Popoff, 2009). JNK inhibitor II delays the TcsL-82-dependent glucosylation of Rac and Ras proteins. As mentioned in the previous paragraph, TcsL-82 induces stress fibre depolymerization and cell rounding, most probably as a consequence of Rac inactivation. Retardation of Rac glucosylation by JNK inhibitor II accounts for the delayed effects of TcsL-82 on actin depolymerization and cell morphology modification. JNK inhibitor II does not directly impair Rac glucosylation by TcsL-82, since the kinetics of in vitro Rac glucosylation by TcsL-82 is not modified in its presence, but rather acts indirectly. It is presumed that JNK activation by TcsL-82 induces a facilitation of Rac and Ras glucosylation, possibly mediated by another, as yet unknown, cofactor. In contrast, changes in cell morphology and actin depolymerization resulting from TcdB intoxication are not prevented by JNK inhibitor II, although this intoxication is also accompanied by an activation of the SAPK/JNK pathway (Geny & Popoff, 2009). This further argues that Tcsl-82-dependent glucosylation of Rac/Ras-GTPases and activation of the SAPK/JNK pathway proceed by two independent mechanisms. Cell stress, such as cytoskeleton alteration, might induce MAPK activation. Indeed, cytochalasin D, which caps actin and thus induces its depolymerization, activates the JNK pathway. But the specific JNK inhibitor II does not prevent cytochalasin D-dependent actin depolymerization (Geny & Popoff, 2009). Therefore, the SAPK/JNK pathway seems to be connected to Rac/Ras-GTPases and subsequently to the control of actin polymerization, but not directly to the actin cytoskeleton, and this interplay between the SAPK/JNK and Rac/Ras-GTPase pathways seems to be specifically targeted by TcsL and not by the other toxins active on Rho-GTPases.

Another consequence of TcdA and TcdB intoxication independent of Rho-GTPase inactivation is apoptosis. Contradictory mechanisms of apoptosis induced by TcdA have been reported, including p38 MAPK-dependent p53 activation leading to cytochrome c release and caspase-3 activation in a Rho-GTPase independent manner (Kim et al., 2005a), and a caspase-3/caspase-8 activation via Rho-GTPase glucosylation and independently of p53 (Brito et al., 2002; Nottrott et al., 2007). The TcdB apoptotic effect has been found to interact with mitochondria, leading to membrane permeability alteration, mitochondrial swelling and cytochrome c release independently of Rho-inhibiting activity (Matarrese et al., 2007).

Alteration of intercellular junctions by clostridial glucosylating toxins as a consequence of actin depolymerization following Rho-GTPase inactivation is well documented (Just & Gerhard, 2004; Popoff & Stiles, 2006). However, an additional mechanism not requiring Rho-GTPase has been reported. Indeed, TcdA seems to disturb TJs and increase cell monolayer permeability through redistribution of ZO-1 from detergent-resistant membrane fractions to soluble fractions via a protein kinase C (PKC) pathway (Chen et al., 2002).

From clostridial glucosylating toxin-dependent cellular effects to pathology

A major difference between C. difficile and C. sordellii infections is that these two pathogens modulate the host inflammatory response differently. C. difficile can grow in the intestinal lumen and secrete toxins which alter the intestinal mucosa and induce a strong inflammatory response, whereas C. sordellii, which is also an agent of haemorrhagic enteritis and enterotoxaemia, mainly observed in animals, can in addition, most often as a result of a wound, enter and multiply in deeper tissues, like muscle and connective tissues, without inducing or triggering only a modest inflammatory response of the host. The absence of or a weak inflammatory response is observed in natural cases of C. sordellii infections as well as in experimental animals (Aldape et al., 2006; Geny et al., 2007). Thus C. sordellii avoids the host innate defence, proliferates and locally produces toxins, which diffuse and induce a toxic shock syndrome. C. sordellii and C. difficile toxins, although they share some common mechanisms of action, exert different cellular effects leading to alteration of epithelial/endothelial barrier integrity accompanied by a strong inflammatory response or by killing inflammatory cells. The different set of Rho/Ras-GTPases inactivated by C. sordellii and C. difficile toxins accounts in part for these different effects. Inactivation of Rho-GTPases mainly induces disorganization of the actin cytoskeleton and subsequent modification of cell morphology and loss of cell–cell contacts, whereas inhibition of Ras-GTPases causes cell killing by cell cycle arrest and/or apoptosis through several signalling pathways including the PI3/AKT, RalGEF/Ral and Raf/ERK pathways (reviewed by Genth & Just, 2010). A subtle balance between these two main effects, called cytopathic and cytotoxic, is controlled by each toxin according to the cell type. TcdB and TcdA, via inactivation of Rho, Rac and Cdc42, profoundly disorganize the actin cytoskeleton and tight and basolateral intercellular junctions, leading to severe dysfunctions and lesions of the intestinal mucosa (diarrhoea, necrosis). TcsL, by inactivating only Rac from the Rho-GTPases, more specifically alters basolateral intercellular junctions, leading to alteration of endothelial barriers and oedema, notably in the cardio-respiratory system, and by inactivating Ras induces lymphocyte and phagocytic cell killing (Fig. 3). Moreover, Rac inactivation in lymphocytes and phagocytic cells probably impairs their migration and phagocytic function via actin filament depolymerizatioin.

Most of the intracellularly active toxins unleash their cytotoxic programme by interacting with a unique intracellular mechanism of action most often consisting in an enzymic modification of a unique or a set of related targets. But some toxins and type III secretion virulence factors from invasive bacteria, which differ from the toxins by their mode of entry into cells based on a direct injection into the cytosol through a bacterial secretion machinery, are multifunctional proteins. They contain several enzymic domains able to modify more than one substrate (Sansonetti, 2004). This strategy ensures a greater efficiency of the overall toxic activity towards a wide set of cell types. The Rho/Ras-GTPase-independent effects of C. sordellii and C. difficile toxins contribute again to the different effects of these toxins. Indeed TcdA attacks TJ integrity by Rho-GTPase inactivation and alternative pathways (Chen et al., 2002; Kim et al., 2009). TcsL-82-dependent SAP/JNK activation facilitates Rac and Ras glucosylation, thus reinforcing the effects of Rac/Ras inhibition in target cells, namely epithelial/endothelial barrier dysfunction as well as phagocytic and immune cell alteration (migration and phagocytosis impairment, apoptosis). Inhibition of the inflammatory response possibly also results from TcsL activation of other signalling cascades subsequent to MAPK activation, since TcsL-82 has been found to activate several transcriptional factors (Geny & Popoff, 2009). In contrast, TcdB-dependent activation of ERK, p38 MAPK or SAP/JNK does not influence Rho-GTPase glucosylation, and activation of the MAPK and NF-κB pathways in a Rho/Ras-GTPase-independent manner, mainly p38 MAPK activation by TcdA, results in a strong inflammatory response (Geny & Popoff, 2009; Kim et al., 2005a; Warny et al., 2000). The fact that C. sordellii is recognized by the innate immune system through Toll-like receptors (TLRs) like other pathogens further supports the idea that the absence of an inflammatory response is not related to inadequate immune detection but rather to a specific activity of C. sordellii toxins (Aldape et al., 2010). An additional Rac/Ras-independent effect of clostridial glucosylating toxins might include the repression of glucocorticoid receptor and reduced glucocorticoid suppression of proinflammatory cytokine secretion, since this effect correlates with a decreased activation of p38 MAPK. Prevention of p38 MAPK activation occurs at lower toxin concentrations than those used in the studies showing an activation of p38. TcsL was shown to prevent the glucocorticoid suppression of TNFα (tumour necrotizing factor) secretion in ex vivo splenocytes and to amplify the effects of the glucocorticoid antagonist RU-486, which has been associated with dramatic C. sordellii toxic shock in women (Miech, 2005; Tait et al., 2007). Interruption of the glucocorticoid response by TcsL probably contributes to the toxic shock syndrome induced by this toxin. Therefore, clostridial glucosylating toxins display complex sets of small GTPase-dependent and -independent activities in cells, leading to two main pathological effects: epithelial/endothelial barrier alteration and cell killing, mainly of immune and inflammatory cells. This results in two distinct pathologies, although produced by closely related toxins: C. difficile colitis, associated with a strong inflammatory response; and C. sordellii infection, characterized by bacterial proliferation in deeper tissues via an inhibition of the inflammatory response and associated with a toxic shock syndrome.