Biochemical Journal

Review article

Super toxins from a super bug: structure and function of Clostridium difficile toxins

Abigail H. Davies, April K. Roberts, Clifford C. Shone, K. Ravi Acharya

Abstract

Clostridium difficile, a highly infectious bacterium, is the leading cause of antibiotic-associated pseudomembranous colitis. In 2009, the number of death certificates mentioning C. difficile infection in the U.K. was estimated at 3933 with 44% of certificates recording infection as the underlying cause of death. A number of virulence factors facilitate its pathogenicity, among which are two potent exotoxins; Toxins A and B. Both are large monoglucosyltransferases that catalyse the glucosylation, and hence inactivation, of Rho-GTPases (small regulatory proteins of the eukaryote actin cell cytoskeleton), leading to disorganization of the cytoskeleton and cell death. The roles of Toxins A and B in the context of C. difficile infection is unknown. In addition to these exotoxins, some strains of C. difficile produce an unrelated ADP-ribosylating binary toxin. This toxin consists of two independently produced components: an enzymatic component (CDTa) and the other, the transport component (CDTb) which facilitates translocation of CDTa into target cells. CDTa irreversibly ADP-ribosylates G-actin in target cells, which disrupts the F-actin:G-actin equilibrium leading to cell rounding and cell death. In the present review we provide a summary of the current structural understanding of these toxins and discuss how it may be used to identify potential targets for specific drug design.

  • large clostridial toxin (LCT)
  • Toxin A (TcdA)
  • Toxin B (TcdB)
  • ADP-ribosyltransferase (ADPRT)
  • Clostridium difficile binary toxin (CDT)
  • pseudomembranous colitis (PMC)

INTRODUCTION

Clostridium difficile is an anaerobic Gram-positive spore-forming bacterium that was first described in 1935. It was given this name as it grew slowly in culture and was difficult to isolate [1]. C. difficile was not thought to be a particularly harmful pathogen until a rise in the number of cases of PMC (pseudomembranous colitis) in the 1970s. Infection with C. difficile accounts for 90–100% of antibiotic-associated PMC, with a 6–30% mortality rate [2]. The majority of patients diagnosed with PMC are treated with broad-spectrum antibiotics such as metronidazole and vancomycin, which can be very effective. Metronidazole is the drug of choice as it is a lower cost drug than vancomycin. Metronidazole and vancomycin have similar efficacy in mild disease, however there is growing concern regarding treatment failure with the use of metronidazole in severe disease. Vancomycin has been shown to be superior in treatment of severe disease [3]. The use of broad-spectrum antibiotics to treat other diseases can disturb the balance of the normal bacterial flora in the gut, resulting in host susceptibility to colonization or overgrowth of C. difficile [1,46]. It is estimated that over 50% of infants are colonized with C. difficile. However, these infants are usually asymptomatic and do not develop PMC. At present there is no clear understanding of why the infection does not manifest in infants [7]. It is thought that the infants' intestinal cells are not fully developed and so are deficient in toxin-specific carbohydrate surface receptors [5,8]. Patients infected with C. difficile initially experience mild diarrhoea, abdominal pains and a fever, and are diagnosed by the presence of C. difficile toxins in stool samples. In the case of mild infection, patients are usually advised to discontinue taking any antibiotics to prevent imbalance of flora in the gut. However, in extreme cases of C. difficile infection leading to PMC, metronidazole and vancomycin are the suggested treatments. Relapse is a common problem associated with antibiotic treatment (occurring in up to 25% of cases), in which case patients are required to continue with prolonged courses of these drugs [4]. C. difficile is a highly infectious bacterium, and its spores can survive on surfaces for long periods of times, therefore it is a difficult organism to control in hospital environments and so the majority of infections are contracted nosocomially.

VIRULENCE FACTORS

C. difficile has a number of virulence factors that contribute to its pathogenicity, but this review will focus on its secreted toxins. The pathogenic strains of C. difficile produce two potent exotoxins, Toxin A and Toxin B (often called TcdA and TcdB), which induce mucosal inflammation and diarrhoea [1,6,9]. In addition to these exotoxins, some C. difficile strains produce an ADP-ribosylating binary toxin (CDT), however, the role of this toxin in disease is unclear [10,11]. In the present review, we focus on the current structural knowledge of clostridial toxins and relevant data regarding the roles of these structures within the host will be discussed in detail.

LARGE CLOSTRIDIAL TOXINS (LCTs)

Biology of LCTs

Bacteria from the genus Clostridium produce a wide range of toxins, the most well known being the clostridial neurotoxins such as the BoNTs (Clostridium botulinum neurotoxins) which can cause flaccid muscular paralysis known as botulism, and the TeNT (Clostridium tetani neurotoxin) which causes spastic paralysis [12]. Toxins A and B, however, are part of the LCT group, owing to their high molecular mass: Toxin A is 308 kDa and Toxin B is 269.6 kDa [13]. All of the members of the LCT family, which includes Toxins A and B from C. difficile, lethal toxins from Clostridium sordellii and α-toxin from Clostridium novyi, act on target cells by modifying small GTPases [14,15]. Toxins A and B are encoded by the genes tcdA and tcdB, which are located in a 19.6 kb locus, known as PaLoc (pathogenicity locus). There are three other genes in the PaLoc that encode positive and negative regulators of the production of these toxins, in addition to a gene encoding a holin-like protein which allows secretion of the toxins [16,17]. C. difficile strains are grouped on the basis of the variations in the structure of the PaLoc [16,18]. There are a number of reports that suggest Toxin B is essential for pathogenicity, whereas the role of Toxin A is less clear. This is reflected in the reported outbreaks of C. difficile strains that are Toxin A negative/Toxin B positive. Universal gene knock-out systems developed for the genus Clostridium [19,20] were used in studies that indicate Toxin A is not essential for virulence [21,22]. However, a recent report suggests that Toxin A is able to cause disease in the absence of Toxin B [23].

Toxins A and B have a multi-modular domain structure described by Jank and Aktories [13] as the ABCD model. This comprises: (A) the biologically active N-terminal domain; (B) the C-terminal binding domain; (C) the cysteine protease domain; and (D) the hydrophobic domain (for a schematic diagram see Figure 1A). The C-terminal binding ‘B’ domain, which consists of polypeptide repeats, is involved in receptor binding to specific cell-surface carbohydrate receptors, but other domains may also play a direct or indirect role in binding [24]. The precise mechanism of toxin uptake into the cells is unclear; however, there is evidence to suggest that it is a four step procedure, summarized in Figure 2: (1) highly specific binding of the ‘B’ domain to receptors on the cell surface followed by receptor-mediated endocytosis of the toxin into the endosomal compartment; (2) a decrease in pH within the endosomal compartment causes conformational changes within the LCTs, allowing pore formation and the subsequent translocation of part of the toxin into the cytosol, which has been shown by the use of 86Rb ions [25,26]. The ‘D’ domain is a large hydrophobic region that makes up almost 50% of the total size of the toxin and is thought to be involved in pore formation prior to translocation of the ‘A’ domain into the cytosol [25,26]. (3) The toxins undergo autoproteolysis allowing only the enzymatic ‘A’ domain to be released into the cytosol. It is thought that the cysteine protease domain ‘C’ is involved in this autoprocessing, and requires InsP6 [2729]; and finally (4) Rho proteins are targeted for glucosylation by the biologically active ‘A’ domain in the cytosol. Toxins A and B target Rho GTPases (Rho, Ras and Cdc42), which are molecular switches involved in numerous signal processes, in particular, the regulation of the actin cytoskeleton (Figure 2). Once the toxins enter the cytosol, they catalyse the addition of UDP-Glc (UDP-glucose) to Thr37 (monoglucosylation) in Rho GTPase leading to depolymerization of actin filaments, disruption of the cytoskeleton and eventually cell rounding and cell death [3032]. C. difficile is a harmful pathogen and the main cause of the rise in cases of PMC in hospitals, therefore it is of great interest to understand the function of LCTs. Structural studies have given us an insight into the molecular mechanism of the LCTs, and we will discuss these below.

Figure 1 Domain organization of C. difficile toxins

(A) The LCTs are composed of four domains, the glucosylating enzymatic domain ‘A’ (dark blue), the autocatalytic processing domain ‘C’ (green), the translocating domain ‘D’ (magenta) and the binding domain ‘B’ (purple). The current solved crystal structures include the ‘A’ domain of Toxin B (Figure 3), the ‘C’ domain of Toxin A (Figure 5) and the ‘B’ domain of Toxin A (Figure 4). The colour scheme for the domain organization of the LCTs is maintained throughout all of the Figures. (B) The domain organization of CDT is shown for the two independently produced components, CDTa and CDTb. The N-terminal signal peptides are displayed in green for both components. The active CDTa' domain is coloured in accordance with Figure 6, with the N-terminal domain (purple) and the C-terminal domain (cyan). The CDTb component has an activation domain (yellow) that must be cleaved to give the CDTb' activated domain (red).

Figure 2 Cartoon representation of the processing machinery of Clostridial toxins

The LCT is displayed using the three-dimensional structures known for the four domains (note: the D domain structure is unknown) from both Toxins A and B, and is colour co-ordinated to the primary domain structure shown in Figure 1. The toxins bind to the target cells via the B domain (purple) and are taken up by receptor-mediated endocytosis. Following a change in the endosomal pH, conformational changes occur within the toxin and the D domain (magenta) forms a pore in the endosomal membrane allowing translocation of the A domain (dark blue) into the cytosol. This then undergoes autocatalytic cleavage via the C domain (green) on binding of InsP6. Upon autocatalytic cleavage, the A domain is released into the cytosol where it glucosylates target Rho GTPases. In the GDP-bound conformations, Rho GTPases are inactive. GDIs keep GTPases in the cytosol. Rho GTPases are activated by GEFs, enabling interaction with effectors that control a number of signalling pathways. GAPs facilitate the hydrolysis of GTP to GDP, converting GTPases back to their inactive form. The A domain (dark blue) glucosylates Rho GTPases, which prevents effector coupling thereby inhibiting specific signalling pathways. In addition, it blocks GEF and GAP activity. GTPases are no longer able to bind GDI and so are found at the plasma membrane. An animated version of this Figure is available at http://www.BiochemJ.org/bj/436/0517/bj4360517add.htm.

Structural aspects of LCTs

Negative stain electron microscopy

Recent studies have shown that Toxins A and B (which share 68% sequence similarity and 47% sequence identity) both have similar native structures to each other using negative stain electron microscopy [33]. Toxins A and B both have an elongated structure containing a ‘head’ domain, a long ‘tail’ domain and a short inner ‘tail’ domain. As Toxin A was considerably more homogenous in structure than Toxin B, mapping studies have focused on Toxin A. The three-dimensional structure of Toxin A was constructed using the random conical tilt approach where the modular domains were mapped into the corresponding structure [33]. After imaging the fragments of the individual domains using electron microscopy and antibody labelling, Pruitt et al. [33] concluded that the long ‘tail’ fragment corresponds to the ‘B’ domain, the short ‘tail’ corresponds to the ‘A’ domain and the ‘head’ corresponds to the ‘D’ domain. The cysteine protease domain was not located. The same structure was then reviewed using a lower pH environment and, as predicted, conformational changes were observed in the ‘head’ shape, confirming the location of the ‘D’ domain in the structure, and also in the glucosyltransferase ‘A’ domain, providing a framework for the molecular mechanism of translocation [33]. Although negative stain electron microscopy provides images of the overall toxin shape, it is beneficial to understand the toxins at a molecular level using X-ray crystallography. Owing to the size and nature of the LCTs, purification and crystallization of the whole toxins is difficult and so a significant amount of work by researchers is focused on structure determination of the four individual domains.

Small angle X-ray scattering

Recently reported data using SAXS (small-angle X-ray scattering) have provided an ab initio, low resolution surface model of the native Toxin B [34]. When combined with the known structural information obtained and models of the unknown domains, this information has allowed visualization of the actual organization of the four individual domains of LCTs for the first time. The data revealed four distinct domain boundaries, into which the structural information and model structures can be aligned. The four domains are organised as demonstrated in Figure 1(A). The hydrophobic ‘D’ domain contains a highly solvent-exposed region which protrudes away from the core structure [34]. The SAXS data were collected at a neutral pH, so the low pH-induced conformational change that is predicted to occur within this ‘D’ domain has not been shown in this structure.

X-ray crystallographic studies of individual LCT domains

‘A’ domain: glucosyltransferase N-terminal domain

The glucosyltransferase domain, which is responsible for glucosylation of Rho proteins, is located at the N-terminal region known as the ‘A’ domain. As there is limited structural data for both Toxins A and B, examples have been taken from the individual domains, where available, for analysis in the present review. Figure 3(A) displays the crystal structure of the ‘A’ domain of Toxin B which gives some insights into the mechanistic implications of Toxin B activity within target cells (PDB code 2BVL) [35]. The structure of the ‘A’ domain of Toxin A is yet to be determined. This catalytic fragment of the LCT is delivered to the cytosol where it glucosylates small GTPases. The ‘A’ domain of Toxin B consists of residues 1–543 in the VPI 10463 C. difficile strain and was co-crystallized with UDP-Glc and manganese ions that are essential for catalysis [36,37]. UDP-Glc was cleaved during the crystallization process owing to the hydrolytic activity of the toxin, but the products α-D-glucose and UDP were identified in the electron density [35]. The crystal structure shares the same common fold (consisting of 234 residues) seen among the glucosyltransferase type A family. The remaining 309 residues are arranged predominantly in α-helices and are thought to contribute to the specificity of the toxin [35]. As shown in Figure 3(B), the catalytic pocket in this Toxin B domain is made up of the β2, β5 and β10 strands, in addition to α12 and α18 helices, and a 510–523 residue loop [38]. The DXD motif involved in the binding of UDP-Glc and Mn2+ is composed of Asp286 and Asp288. The Asp288 residue binds directly to Mn2+, whereas Asp286 forms a hydrogen bond with a water molecule of the Mn2+ co-ordination sphere, in addition to binding to both the 3′-hydroxy group of UDP-ribose and glucose, making it a key residue for catalysis [35]. Mutational analysis by alanine scanning revealed five important residues involved in enzyme activity: Asp270, Arg273, Tyr284, Asn384 and Trp520. In addition, four residues, Arg455, Asp461, Lys463 and Glu472, were found to be important for UDP-Glc recognition [39].

Figure 3 The glucosyltransferase ‘A’ domain of Toxin B

(A) Crystal structure of the glucosyltransferase domain of C. difficile Toxin B (dark blue), where UDP-Glc (green) is bound to the catalytic cleft via a manganese ion (magenta), with the residues involved in the active site displayed (teal) (PDB code 2BVL) [34]. (B) Stereo view of the catalytic cleft of Toxin B with water molecules (blue) and manganese ion (magenta). Residues involved in enzyme activity (teal) and UDP-Glc recognition (raspberry) are displayed, in addition to the DxD motif (orange). Images were created using PyMOL (http://www.pymol.org).

The LCTs target and glucosylate the Rho family of GTPase proteins. These are molecular switches that are predominantly responsible for regulating the actin cytoskeleton. These Rho proteins play other roles in cellular functions such as cell polarity, gene transcription and cell cycle progression. The Rho GTPase family of proteins consists of the Rho, Rac and Cdc42 subfamilies. The proteins exist in two forms: the GTP-bound form which is active and the GDP-bound form which is inactive, as displayed in Figure 2. These two forms are inter-convertible and are regulated by GEFs (guanine-nucleotide-exchange factors), GDIs (guanine-nucleotide-dissociation inhibitors) and GAPs (GTPase-activating proteins) (Figure 2). GDIs bind to Rho GTPases at a switch I region, rendering them soluble in the cytosol and in the inactive form. Activation of GTPases to their GTP-bound form is induced by GEFs, and this activation allows interaction with numerous effector molecules that control a number of signalling pathways. GAPs hydrolyse GTP-bound Rho GTPases into inactive GDP-bound Rho GTPases [40]. In the GDP-bound conformation, Rho GTPases have an exposed Thr37 residue which is targeted by LCTs for glucosylation. Upon glucosylation, actin filaments depolymerize, leading to disruption of the cytoskeleton and cell death. The Rho proteins are not glucosylated when bound to GTP or when their switch I region is complexed with GDIs, as the Thr37 residue is not exposed. Rho is glucosylated in the GDP-bound conformation as the Thr37 residue is exposed and allows an oxygen acceptor atom to be in position for attack from the C1 donor atom of the LCTs [31,32]. Toxin B binds to RhoA in the Rho-GTP form [41].

Two possible mechanisms have been suggested for the catalytic activity of the ‘A’ domain; a double displacement reaction (SN2) and an internal return stereo-specific (SN1-like) reaction. The double displacement reaction firstly requires a nucleophile in a position to attack the β-side of the C1 atom in glucose. However, as this atom is surrounded by non-polar residues, there is no nucleophile in position to attack the C1 atom, hence a double displacement reaction is unlikely. An internal return mechanism SN1 is possible, as the oxocarbenium intermediate would be stabilized by the surrounding negative phosphate and carboxylate groups from Asp270, Asp286, Asp288 and Glu515, which are in turn compensated by the positively charged Arg273 and a Mn2+ ion [35]. LCTs exhibit high substrate/co-substrate specificity, with Toxins A and B only binding UDP-Glc. The two residues Ile383 and Glu385, which are located in the active site, are responsible for this specificity as shown by mutagenesis studies [39,41].

‘B’ domain: binding C-terminal domain

The initial step in the toxication process of C. difficile is the binding of LCTs to the cell-surface of intestinal epithelial cells, which is carried out by the ‘B’ domain. The ‘B’ domain of LCTs interacts with carbohydrate structures such as Gal-α-(1,3)-Gal-β-(1,4)-GlcNAc on the host epithelial cells [42]. Although a functional carbohydrate receptor in humans has yet to be identified, attempts to gain more insights into carbohydrate binding have been made. The crystal structure of the C-terminal binding domain of Toxin A has been determined for two different fragment sizes. The first fragment (127 residues), named TcdA-f1, from strain 48489, toxinotype VI was obtained by expressing a longer fragment and isolating a smaller cleaved fragment (PDB code 2F6E) [43]. The second fragment, TcdA-f2, which is a slightly longer fragment (255 residues), is in complex with a synthetic derivative of a carbohydrate receptor, CD-grease [α-Gal-(1,3)-β-Gal-(1,4)-β-GlcNAcO(CH2)8CO2CH3] (Figure 4A) (PDB code 2G7C) [44]. In this complex there are two carbohydrate-binding regions. However, in the full-length C-terminal fragment there are seven of these potential binding domains that are highly conserved, giving it a high binding capacity. Although there is little information about the binding domain of Toxin B, it is believed that Toxin B uses different receptors to bind to target cell surfaces than Toxin A [41]. However, considering Toxins A and B share 47% sequence identity, the structural information regarding the Toxin A–carbohydrate interactions may provide some clues as to how Toxin B might interact with carbohydrate surface receptors.

Figure 4 The binding ‘B’ domain of Toxin A

(A) Ribbon structure of the binding domain of C. difficile Toxin A (purple) with two carbohydrate synthetic derivatives of CD-grease bound, shown as stick models (pink) (PDB code 2G7C) [43]. (B) Cross-section through the binding domain of Toxin B showing the β-hairpin 31 screw axis, where the β-hairpins are related to the adjacent hairpin by 120°. Images were created using PyMOL (http://www.pymol.org). Adapted from Jank et al. [30] with permission © 2007 Oxford University Press.

Sequence analysis of the Toxin A gene shows that 31.5% of the gene is in 38 contiguous repeating units and these repeats are located in the C-terminal region. These repeats can be split into two categories: class I (long), of which there are seven types, each repeat being 30 amino acids long; and class II (short), of which there are 31 types, each consisting of either 20 or 21 amino acids [45]. TcdA-f2 contains nine small repeats and two long repeats, whereas TcdA-f1 consists of four short repeats and one long repeat [43,44]. Each short and long repeat forms β-hairpins that are connected by loops of 7–10 residues in short repeats and 18 residues in long repeats [41]. The β-hairpins are related to each other by a 31 screw-axis transformation, meaning that each β-hairpin is related to the adjacent hairpin by 120° (Figure 4B) [44]. This results in a β-solenoid left-handed helix, which is a common conformation of many bacterial cell-surface-binding proteins, and is thought to increase surface area for optimized binding to the target cells (Figure 4B) [44]. These results are consistent with that of the TcdAC26–39 structure of the ‘B’ domain from Albesa-Jové et al. [34].

‘C’ domain: cysteine protease domain and the ‘D’ domain: hydrophobic region

The CPD (cysteine protease ‘C’ domain) is thought to be located between residues 543 and 769 in Toxin A and between residues 543 and 767 in Toxin B [27]. The crystal structure of the CPD (543–809) from Toxin A was determined at 1.6 Å (1 Å=0.1 nm) in the presence of InsP6 (Figure 5A) (PDB code 3HO6) [29]. This sits between the enzymatic domain and the delivery domain, playing a role in proteolytic cleavage of the toxin. Toxins A and B both undergo auto-catalytic cleavage in the presence of InsP6. Once the target cell has taken up the LCT via receptor-mediated endocytosis at the ‘B’ domain, the toxin undergoes autoproteolysis in order to allow the ‘A’ domain to pass across the endosomal membrane into the cytosol [46]. The structure reveals a central nine-stranded β-sheet flanked by five α-helices. The InsP6-binding site and the proposed active site are located on opposite sides of the central β-sheet and are separated by a three-stranded β-hairpin, known as the ‘β-flap’. The N-terminal domain extends around the exterior of the domain from the InsP6-binding site to the proposed active site [29]. Cys700, His655 and Asp589 have been identified as the catalytic triad. However, owing to the distances between these residues, hydrogen bond formation is not possible as with other cysteine proteases. Therefore the mechanism of catalysis may differ from other cysteine proteases. Pruitt et al. [29] have shown that the CPD changed conformation on binding of InsP6 to a more stable form, which increases resistance to chymotrypsin digestion. InsP6 is a highly negatively charged ligand and binds to a number of positively charged residues that span the entire domain including Arg753, Tyr579 and seven lysine residues: Lys577, Lys602, Lys649, Lys754, Lys766, Lys777 and Lys794 (see Figure 5B, with these positively charged residues in blue). The ‘β-flap’ is thought to be involved in transmitting structural changes from the InsP6-binding site across to the active site. The CPD uses InsP6 to activate an intramolecular autoproteolytic cleavage event which allows the correct processing of the toxins to transfer the enzymatic domain into the cytosol. Shen et al. [46a] recently solved the structure of the CPD of Toxin B (PDB code 3PEE) which shares 56% sequence identity with the CPD of Toxin A and a highly similar structure [rmsd (root mean square deviation) of 1.3 Å]. Comparison of the structures show a completely conserved active site, an almost identical InsP6-binding site (with the exception of two residues) and a well-conserved β-flap. Using chemical probes, Shen et al. [46a] showed that the CPD of Toxin B can adopt the activated conformation without the presence of InsP6, and InsP6 in fact shifts the equilibrium in favour of the active conformation. Through the use of tryptophan fluorescence assays, mutagenesis and structural analysis. Shen et al. [46] were able to show that a conserved group of interconnected residues in the β-flap region are responsible for communicating the InsP6-binding signal to the active site.

Figure 5 The cysteine protease ‘C’ domain of C. difficile Toxin A

(A) Ribbon structure of the ‘C’ domain (green) displaying the proposed residues in the active site (cyan) and the InsP6 (red) bound on the opposite side of the central nine-stranded β-sheet (PDB code 3HO6) [28]. (B) The InsP6-binding site on the ‘C’ domain displaying the electrostatic potential of the binding site (blue represents positively charged residues). InsP6 is displayed as a ball and stick molecule (red). The InsP6-binding site is highly positive as shown by the abundance of blue colour. Images were created using PyMOL (http://www.pymol.org).

Following receptor-mediated endocytosis of the LCTs, the ‘A’ domain is translocated into the cytosol. The hydrophobic ‘D’ domain is thought to be involved in the translocation, although little is known about its structure. LCTs undergo conformational changes stimulated by a decrease in endosomal pH. Qa'Dan et al. [26] used bafilomycin A1, which is a potent inhibitor of the endosomal vacuolar ATPase pump that controls the acidity of the endosome, to assess the effect of endosomal acidification. Translocation was measured using TNS [2-(p-toluidinyl) napthalene-6-sulfonic acid, sodium salt] fluorescence. For Toxin B, as the acidity of the endosome increased, an increase in TNS fluorescence was seen, suggesting a conformational change within Toxin B and exposure of the hydrophobic ‘D’ domain, thereby confirming its role in translocation [26]. Owing to the hydrophobic nature of the ‘D’ domain, purification and crystallization are difficult, which may explain why the structure has not yet been determined.

CLOSTRIDIUM DIFFICILE BINARY TOXIN

Biology of the binary toxin

In addition to the LCTs, some C. difficile strains can produce an additional toxin called the ADP-ribosyltransferase binary toxin, CDT, which is a member of the ADPRT family [11]. The ADPRT family is divided into four classes on the basis of domain organization and their targets [47]. The first group is the AB5 group that includes cholera toxin. Within this group, the toxins are composed of one A subunit and five B subunits, and they target small regulatory G-proteins. The second group is the AB three domain group which includes diphtheria toxin. Toxins of this group ribosylate a diphthamide residue on elongation factor 2, and have a binding domain, a transmembrane domain and a catalytic domain. The third group is the single polypeptide group, which includes C3 toxin from C. botulinum, and have unknown roles in pathogenesis. The final group is the AB binary toxin group, which are made up of two subunits and target actin. This group includes the C2 toxin from C. botulinum and Iota toxin from Clostridium perfringens [47]. CDT is a member of the AB binary toxin group and is made up of two independently produced components: the enzymatic component, CDTa (mature length 48 kDa), and the transport component, CDTb (mature length 74 kDa); see Figure 1(B) for a schematic representation of the domain organization [10,48].

In one study of CDT genes in the United States, 15.5% of C. difficile strains tested had both the genes for the CDT components (cdtB and cdtA), 8.7% of which did not have the LCT genes [49]. The precise pathogenesis of CDT is yet to be established due to the dominant production of LCTs in most virulent strains of C. difficile and lack of CDT [10]. However, it has been shown that a CDTa–CDTb (CDT) complex is toxic to Vero cells [50]. A recent study has suggested a role for CDT in increasing adherence of bacteria to target cells, by the formation of microtubule protrusions. Using time-lapse and immunofluorescence microscopy, Schwan et al. [51], have shown that CDT forms dynamic microtubule protrusions on the surface of human colon carcinoma cells (Caco-2) concomitantly with ADP-ribosylation of actin and depolymerization of microfilaments. In addition to protrusion formation, cellular microtubule structures were also altered to increase bundling of microtubules. The CDT-induced formation of protrusions caused a ~5-fold increase in adherence of C. difficile in anaerobic conditions. The protrusions were shown to form a dense mesh-work in which the bacteria were caught, contributing to the colonization of C. difficile. Similar results were also demonstrated for the homologues C. botulinum toxin C2 and C. perfringens Iota toxin [51].

The transport component, CDTb, is essential for entry of CDTa (enzymatic component) into the cytosol [52]. The proposed mechanism of uptake of CDTa is similar to other members of the ADPRT family such as C. botulinum C2 toxin and Iota toxin from C. perfringens [48]. The CDTb component must be activated via cleavage, after which it then can form heptamers at the cell surface and bind to specific cell-surface receptors. Subsequently, CDTa binds to CDTb and is taken up into the cell by receptor-mediated endocytosis. In a similar mechanism to the LCTs, low pH-induced conformational changes occur leading to potential heptameric pore formation and translocation of CDTa into the cytosol [48,52]. The N-terminus of CDTa is responsible for interaction with CDTb, whereas the C-terminus harbours the enzymatic activity (Figure 1B) [53]. We can predict that CDTa will irreversibly ADP-ribosylate monomeric G-actin at the Arg177 residue from the structural evidence of Iota toxin (closest homologue of CDTa). This ADP-ribosylation will block polymerization of G-actin to F-actin and subsequently disrupt the F-actin:G-actin equilibrium [54,55]. Disruption of the actin equilibrium results in cell rounding and cell death.

Structural aspects of CDT

The crystal structure of CDTa was solved at three different pH values, 4.0, 8.5 and 9.0 (PDB codes 2WN8, 2WN4 and 2WN5 respectively) [56]. In addition, the structure was solved in complex with ADP ribose donors NADPH and NAD at pH 9.0 (PDB codes 2WN6 and 2WN7) [56]. The structure in complex with NAD was determined at 2.25 Å and detailed mechanistic implications of CDTa have been proposed on the basis of this structural data [56]. It is known that CDTa transfers the ADP ribose group of NAD/NADPH to monomeric G-actin at Arg177, blocking polymerization of actin and therefore leading to the collapse of the cell cytoskeleton. This prediction was made on the basis of the mechanism of C. perfringens Iota toxin, of which CDTa shares 84% sequence identity [54,55].

The structure of CDTa in complex with NAD is displayed in Figure 6(A). We can observe the N-terminal domain, which consists of five α-helices and eight β-strands, extending from residues 1 to 215, and the C-terminal domain, which extends from residues 224 to 240, also consisting of five α-helices and eight β-strands. These domains are linked by a loop extending from residues 216 to 223, shown in red in Figure 6(A) [56]. The C-terminal domain harbours the enzymatic activity, whereas the N-terminal domain is predicted to interact with the CDTb domain. Both NAD and NADPH bind to the catalytic cleft of CDTa via the interacting residues Arg302, Arg303, Gln307, Asn342 and Ser345 [56]. Figure 6(B) displays these catalytic residues in addition to a number of other defining features of the ADPRT family of toxins, for example the PN loop (orange) and the Arg-motif (yellow). The highly flexible ARTT loop can be seen in blue, which is an ADP-ribosyl turn-turn loop that is important for substrate binding, along with the catalytic residues of the EXE motif, which has been shown to be crucial for activity in ADPRTs. In CDTa, the EXE motif is composed of the residues Glu385 and Glu387, which is thought to be involved in stabilizing the substrate–enzyme complex as seen with the corresponding residues Glu378 and Glu380 in Iota toxin. However, in this structure both of these residues are not in direct contact with NAD or NADPH, which might suggest that the EXE motif in CDTa is not necessary for ligand binding and stabilization of this complex, although there is no experimental evidence to support this [56]. However, this could be due to the location of both Glu385 and Glu387 on the flexible ARTT loop. The final feature shown in Figure 6(B) is the STS motif, which includes the residues Ser345 and Ser347. Ser345 forms a strong hydrogen bond with Glu387, and also directly with NAD and NADPH, therefore suggesting a role for the STS motif in ligand binding and catalysis [56].

Figure 6 The active CDTa component of CDT from C. difficile

(A) Ribbon structure displaying the N-terminal domain (purple), the C-terminal domain (cyan) and the connecting loop between the C-terminal and N-terminal domains (red), and the active site is shown within the black box (PDB code 2WN7) [55]. (B) Stereo view of the active site of CDTa with key residues involved in NAD binding: STS-motif (green), the PN-loop (orange), the EXE motif and ARTT loop (blue), the Arg-Arg residues (yellow), and any remaining residues involved in binding are shown in light blue and the NAD molecule is in pink. Images were created using PyMOL (http://www.pymol.org).

The proposed mechanism of ADP-ribosylation of actin in Iota toxin was suggested to be an SN1 reaction after formation of an actin–Iota complex and also by the use of site-directed mutagenesis [57]. The SN1 reaction occurs via two intermediates: first an oxocarbenium ion intermediate and secondly a cationic intermediate. This mechanism is thought to be a common mechanism amongst the ADPRT family of toxins [47]. As Iota toxin is the closest homologue to CDTa, a similar SN1 reaction has been proposed [56]. In following the SN1 mechanism, the catalytic glutamate Glu387 in CDTa should form a H-bond with the 2′-OH of ribose forming an oxocarbenium intermediate and rendering the ribose group vulnerable to nucleophilic attack (Figure 7). Following cleavage of NAD and formation of an oxocarbenium ion, it is thought the ARTT loop is rearranged to bring the Glu385 residue to the reaction centre to stabilize the transfer of ADP-ribose to the Arg177 of actin [56]. Further structural and experimental evidence is required to validate this hypothesis, such as site-directed mutagenesis of catalytic residues in addition to solving the structure of a CDTa–actin complex.

Figure 7 Predicted mechanism of ADP-ribosylation in CDT

The catalytic glutamate residue (blue) forms a hydrogen bond with the 2'-OH of the ribose group in NAD (black). This stabilizes the oxocarbenium intermediate leaving the glycosidic bond susceptible to nucleophilic attack from the target arginine residue from actin (green).

CONCLUDING REMARKS AND PERSPECTIVES

Some of the more dominant and pathogenic strains of C. difficile produce two large glucosylating toxins (Toxin A and Toxin B) and an ADPRT toxin (CDT) such as that from the CD196 strain [10]. However, there are strains of C. difficile that have variations in their ability to produce these toxins, and there is a great uncertainty over the roles of each of these toxins individually in pathogenesis [2123]. As some of the most pathogenic strains of C. difficile produce CDT in addition to the LCTs, it is tempting to question whether or not CDT plays an adjunctive role to the LCTs in pathogenesis.

In the present review we have summarised the research into clostridial toxin structures. The full-length structures of the LCTs are yet to be determined, but the structures of some of the individual domains of these toxins have been solved. From the current structures we now have a better understanding of the binding, the autoproteolysis and their mechanism of action once the enzymatic domain has entered the cytosol. These findings have opened up a number of avenues that are of great value for use in drug therapeutics. These structures have provided multiple targets for drug design such as the binding domain to prevent uptake of the toxins, the cysteine protease domain to prevent cleavage of the toxin, or targeting the enzymatic domain to inhibit its activity. Determination of the crystal structures of the remaining unsolved domains is highly desirable, in addition to solving the structure of an enzyme–substrate complex. This will provide an understanding of how the toxins function as a whole. The crystal structure of CDT has provided a great insight into the active site of the toxin and its mechanistic implications. Further work using mutational analysis and enzyme–substrate complex formation will improve our understanding of the mechanism of ADP-ribosylation and help provide more specific drug targets for therapeutic design.

OUTSTANDING QUESTIONS

  1. What renders Toxin B more cytotoxic than Toxin A?

  2. How do the LCTs bind to cell surfaces, do their specificities differ and does this affect their differences in pathogenicity?

  3. How does the cysteine protease ‘C’ domain cleave the toxin and why does it require InsP6?

  4. By what mechanism do the LCTs glucosylate Rho GTPases?

  5. What is the precise role of CDT in pathogenesis?

  6. Could CDT play an adjunctive role to the LCTs in pathogenesis?

  7. By what mechanism does CDT ADP-ribosylate actin?

FUNDING

A.H.D. is supported by a joint post-graduate studentship by the BBSRC (U.K.) and Health Protection Agency (UK).

Abbreviations: CD-grease, α-Gal-(1,3)-β-Gal-(1,4)-β-GlcNAcO(CH2)8CO2CH3; CDT, Clostridium difficile ADP-ribosylating binary toxin; CPD, cysteine protease ‘C’ domain; GAP, GTPase-activating protein; GDI, guanine-nucleotide-dissociation inhibitor; GEF, guanine-nucleotide-exchange factor; LCT, large clostridial toxin; PaLoc, pathogenicity locus; PMC, pseudomembranous colitis; SAXS, small-angle X-ray scattering; TNS, 2-(p-toluidinyl) napthalene-6-sulfonic acid, sodium salt; UDP-Glc, UDP-glucose

References

  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
  15. 15.
  16. 16.
  17. 17.
  18. 18.
  19. 19.
  20. 20.
  21. 21.
  22. 22.
  23. 23.
  24. 24.
  25. 25.
  26. 26.
  27. 27.
  28. 28.
  29. 29.
  30. 30.
  31. 31.
  32. 32.
  33. 33.
  34. 34.
  35. 35.
  36. 36.
  37. 37.
  38. 38.
  39. 39.
  40. 40.
  41. 41.
  42. 42.
  43. 43.
  44. 44.
  45. 45.
  46. 46.
  47. 46a.
  48. 47.
  49. 48.
  50. 49.
  51. 50.
  52. 51.
  53. 52.
  54. 53.
  55. 54.
  56. 55.
  57. 56.
  58. 57.
View Abstract