|Home | About | Journals | Submit | Contact Us | Français|
Recently, we constructed and characterized isogenic tcdA and tcdB mutants of a virulent Clostridium difficile strain and used a hamster model of disease to demonstrate that toxin B, not toxin A, is essential for virulence of this emerging pathogen. Earlier studies had shown that purified toxin A alone was able to induce C. difficile disease pathology and that purified toxin B was not effective unless it was co-administered with toxin A, suggesting that the toxins act synergistically. In this addendum we discuss this paradigm-shifting conclusion in the context of current strain epidemiology, particularly with respect to naturally occurring toxin A-negative, toxin B-positive isolates and the NAP1/027 epidemic isolates. The role of toxin receptors and how variant toxins might exert their effects is also discussed in relation to the published data. We conclude that it is critical to use the natural infection process to dissect the role of toxins in disease, and that future studies are contingent on such work. The impact and importance of animal models of C. difficile virulence are therefore considered within this frame of reference.
Clostridium difficile is regarded as an emerging pathogen and in recent years has been recognized as an increasingly important cause of nosocomial disease.1 It is a Gram positive spore-forming anaerobe and is the causative agent of several antibiotic-associated diarrheal diseases that are induced by treatment with antibiotics or by disruption of the normal gastrointestinal flora. These syndromes are collectively known as C. difficile infections or CDI, and include pseudomembranous colitis, toxic megacolon, and may lead to chronic infection, shock and death.2,3 C. difficile is unique among enteric pathogens in that disease is almost always associated with prior antimicrobial therapy or with an alteration of endogenous intestinal flora. Its spore-forming ability and resultant persistence has allowed C. difficile to become common in the hospital environment and it is now the leading cause of infectious diarrhea in hospitals worldwide.1,4
Recently, morbidity and mortality resulting from CDI have increased significantly as a result of changes in the virulence of the causative strains, coupled with changes in antibiotic usage patterns.1,4,5 Since 2002 we have seen the emergence of epidemic or hypervirulent strains of C. difficile, which are toxinotype III NAP1/027 isolates that are also resistant to fluoroquinolones.1,4 These strains reportedly produce more toxin than reference strains and encode a third toxin, binary toxin or CDT.2 Significantly, these isolates appear to cause epidemics more readily than other C. difficile strains.1,4 In the USA, the mortality rates from C. difficile-associated disease have increased from 5.7 per million in 1999 to 23.7 per million in 2004, a rate of death that is greater than deaths caused from all other intestinal infectious diseases combined.5 The cost of treating this disease in the USA alone is estimated as $3.2 billion per annum.6
The major virulence factors of C. difficile are thought to be toxin A and toxin B. These toxins are encoded by the tcdA and tcdB genes, respectively, which are located within a 19.6 kb pathogenicity locus, the PaLoc7 (Fig. 1A). This locus also has three other genes: tcdR, encoding an alternative sigma factor that is required for toxin production,8 tcdC, encoding an anti-sigma factor that interacts directly with TcdR9 and tcdE, which encodes a putative holin that is proposed to be involved in toxin release from the cell.10
Both toxin A and toxin B are proinflammatory and cytotoxic, causing disruption of the actin cytoskeleton and impairment of tight junctions in human intestinal epithelial cells, with resulting fluid accumulation and extensive damage to the large intestine.7 More recently, the toxins were shown to cause the release of inflammatory cytokines from mast cells and macrophages as well as from epithelial cells, leading to further fluid secretion and intestinal inflammation. Interestingly, in addition to damaging the gut, toxin B is also cardiotoxic to zebrafish embryos11 and can cause hemorrhage in the lungs of hamsters,12 suggesting that this toxin may have more systemic effects than previously thought.
Toxin A and toxin B are both members of the large clostridial glucosylating toxin family, which also includes lethal toxin (TcsL) and hemorrhagic toxin (TcsH) from Clostridium sordellii,13 alpha toxin (TcnA) from Clostridium novyi14 and TpeL from Clostridium perfringens.15 The toxins are very large, being between 250 and 308 kDa in size, and have a high degree of sequence identity.16,17
Toxins A and B have three major domains: a catalytically active N-terminal domain, a centrally located translocation domain and a C-terminal receptor binding domain7,16 (Fig. 1B). These toxins are monoglucosyltransferases and have similar substrate specificity; both catalyzing the transfer of a glucose moiety onto Rho family GTPases (Rho, Rac and Cdc42) inside their target cells (Fig. 2). Rho family GTPases control numerous cellular functions, including the organization of the actin cytoskeleton. They also control epithelial barrier function and are involved in the signaling and motility of immune cells.16 Glucosylation of Rho GTPases by toxins A and B locks these proteins into an inactive conformation, thereby blocking all downstream signaling pathways in the cell. The result is disruption of the actin cytoskeleton, cell rounding and eventually apoptosis and death of the intoxicated cell.7,16
Toxins A and B have been purified and studied extensively. However, the role of each toxin in CDI has remained difficult to elucidate because, until very recently, it was not possible to generate isogenic C. difficile mutants. This problem was recently overcome in our laboratory and by others.18,19 Using the method of O'Connor et al.18 we generated multiple independent isogenic mutants in a derivative of the virulent C. difficile isolate, strain 630, by the specific insertion of a recombination plasmid targeted to the tcdA or tcdB gene. The resultant strains were defective in the production of either toxin A or toxin B. The virulence of each mutant then was compared to the wild-type strain to directly determine the relative contribution of each toxin to disease.20
The genotype of the mutants was confirmed by PCR and Southern blotting before a thorough in vitro analysis was performed. Initially qRT-PCR was performed to rule out possible polar effects on the accessory genes of the PaLoc upon insertion of the recombination vector. These experiments showed that expression of the PaLoc genes was the same in the mutant strains as the wild-type parent strain. Interestingly, whilst inactivation of tcdA had no discernible effect on the expression of tcdB, approximately two to three fold higher levels of tcdA expression were observed in the tcdB mutants. This observation was subsequently confirmed by toxin A-specific western blots, which showed higher levels of toxin A in the toxin B mutants, although the explanation for this observation is not known.20
We next confirmed that the mutants no longer produced the appropriate toxin by performing cytotoxicity assays with HT29 and Vero cells, which are sensitive to toxin A or toxin B, respectively (Fig. 3). When HT29 cells were used, supernatants collected for the toxin A mutants were significantly less toxic than those of the wild-type strain. These assays also confirmed that the toxin A produced by the wild-type strain and, importantly, by the tcdB mutants, was functional, since cytotoxicity could be neutralized with toxin A-specific antibodies. As expected, when Vero cells were used the toxin A mutants were as cytotoxic as the wild-type strain. In these assays the cytopathic effects could be neutralized only with toxin B-specific antibodies, demonstrating that toxin B is primarily responsible for Vero cell cytotoxicity. By contrast, supernatants collected from toxin B mutants were significantly less cytotoxic to Vero cells than wild-type, confirming that these strains no longer produced toxin B.20
The virulence of the mutants and the wild-type strain was assessed using the Syrian golden hamster model of infection (Fig. 4). Groups of 10 hamsters were used for each mutant as well as for the wild-type strain. Upon challenge with the wild-type strain, all ten hamsters were colonized and 9/10 animals died (90%). Challenge with two independently derived tcdA mutants resulted in the overall colonization of 17/20 animals. Unexpectedly, 16 (94%) of these animals died providing clear evidence that the tcdA mutants were still as virulent as the wild-type strain even though they no longer produced toxin A. This observation indicates that toxin A is not essential for disease. By contrast, the data obtained with the two tcdB mutants demonstrated that toxin B is essential for disease. Colonization with the tcdB mutants was achieved in 18/20 hamsters and of these only four died (22%). Subsequent analysis of three of these hamsters revealed that the fecal pellets were strongly cytotoxic in Vero cell cytotoxicity assays indicating the presence of toxin B, which suggests that in these animals reversion of the mutants to wild-type had occurred.20
The finding that toxin B and not toxin A is essential for C. difficile disease is in stark contrast to earlier studies performed using purified toxin preparations, which had led to the hypothesis that toxin A was the major virulence factor of C. difficile.7,12,17 In these experiments intragastric challenge of hamsters with toxin A alone resulted in the onset of all of the symptoms typical of C. difficile infection, including fluid accumulation, inflammation and necrosis of intestinal tissues. Purified toxin B was unable to cause any symptoms of disease unless intestinal damage was present, or the toxin was co-administered with sublethal concentrations of toxin A.12 This result suggested that the toxins worked synergistically, with toxin B eliciting an effect only after toxin A had caused prior tissue damage.
The isolation of naturally occurring isolates lacking toxin A, but which still are capable of causing disease, reinforces our findings that toxin A is not essential for the virulence of C. difficile and that toxin B does not require the presence of toxin A for its in vivo activity. The worldwide prevalence rates for these toxin-variant isolates vary considerably, most often from 0.2–3% but reaching as high as 97.9%.21 Interestingly, toxin A−B+ strains cause the same spectrum of disease as toxin A+B+ isolates, ranging from asymptomatic carriage through to pseudomembraneous colitis. There is a propensity towards more severe disease in patients infected with toxin A−B+ strains.21 Although it is not clear why increased severity is associated with these strains it is possible that the absence of a toxin A-specific immune response, which is a key determinant in controlling disease severity, might be important.21
Further evidence that toxin B is able to elicit characteristic symptoms of C. difficile infection in the absence of toxin A comes from experiments with mice carrying human intestinal explants. In this study, toxin B was shown to be as potent an enterotoxin as toxin A, causing severe intestinal epithelial cell damage, increased mucosal permeability and an acute inflammatory response characterized by neutrophil recruitment and tissue damage.22 The results from this study also questioned the traditional view that toxin B acts as a cytotoxin and toxin A as an enterotoxin.
The most obvious explanation for the different responses observed with purified toxin B in earlier animal models compared to the “humanized” mouse model, is the absence of toxin B-specific receptors in the rodent intestine. Indeed, previous studies suggest a strong correlation between the sensitivity of cells towards the action of the C. difficile toxins and the presence of toxin-specific receptors.23 For example, experiments with radiolabelled toxin A and B showed that only toxin A, and not toxin B, could specifically bind to the intestine of hamsters. Interestingly, this tissue is highly sensitive to toxin A, but insensitive to the action of toxin B.12,24 Furthermore, the sensitivity of rabbits of different age to the biological effects of toxin A has been shown to correspond with a developmental increase of toxin A receptors, with 24 day old rabbits much more susceptible than newborns. This observation further highlights the importance of toxin receptors in disease.25
These observations are perhaps not surprising since the mechanism employed by toxin A and toxin B dictates the need for specific receptors to which the toxins can bind as an initial step in the intoxication process. It is only after binding to a cellular receptor that endocytosis of the toxins can occur. Once in the acidified compartment of the endosome the toxins undergo a conformational change that facilitates their insertion into the endosomal membrane and subsequent pore formation. The toxins then undergo a cysteine protease-mediated autocatalytic cleavage event,26 which for toxin B occurs in the presence of inositol hexakisphosphate.27 The N-terminal catalytic domain of the toxins is then released into the cytosol, resulting in the subsequent intoxication of the cell.16 Binding of the toxin to a cellular receptor is therefore critical to the downstream effect that the toxin exerts on the host.
The specific receptors to which the toxins bind are not currently known. However, evidence suggests that toxin A binds to the apical surface of enterocytes, whilst the receptors to which toxin B binds are on the basolateral surface.28 To gain access to basolateral receptors, tissue damage that disrupts the intestinal mucosa must first occur. This may explain why purified toxin B, which does not bind to the apical surface,29 was unable to elicit symptoms in the hamster unless prior intestinal damage was present. The fact that toxin A compromises the epithelial barrier from the apical side might explain why sublethal concentrations of toxin A were found to potentiate the activity of toxin B. If this hypothesis is correct, then the observation in our study that toxin A mutants are as virulent as the wild-type suggests that another protein or compound produced by C. difficile may damage or compromise the intestinal epithelium sufficiently to make the basolateral surface accessible to toxin B.
The isolation of clinical toxin A−B+ strains and the studies with “humanized” mice support our findings that toxin A is not essential for the virulence of C. difficile. The observation that purified toxin A alone was able to induce all of the classical symptoms of disease in the hamster cannot, however, be dismissed. One important factor to consider might be the amount of toxin A being produced by the infecting strain. The strain in which our isogenic mutants were constructed is an erythromycin sensitive derivative of strain 630.18 This strain is known to produce low levels of both toxin A and toxin B.30 By contrast, relatively high titres of purified toxin were administered intragastrically to hamsters in the studies discussed before. It is therefore possible that when higher concentrations of toxin A are present, the toxin is able to more efficiently elicit the symptoms of disease. In this context it would be of interest to assess the virulence of isogenic toxin A and toxin B mutants in a C. difficile strain that produces higher levels of toxin. Perhaps of most importance will be the construction of such mutants in an epidemic NAP1/027 strain. If the role of toxin A in the pathogenesis of disease is dependent on the quantity of toxin produced, then it is possible that a toxin B mutant of a NAP1/027 strain would still be virulent.
The emergence of the NAP1/027 isolates appears to coincide with the increasing incidence of C. difficile infections throughout the developed world, particularly in the US, Canada and Europe.1,2 NAP1/027 strains are thought to produce 16 and 23 fold more toxin A and toxin B respectively than reference strains.2 This increase in toxin production is thought to be the result of a frameshift mutation within the tcdC gene of PaLoc. This results in the production of a truncated and non-functional TcdC protein that is unable to downregulate the expression of the toxin genes. Increased levels of toxin production are hypothesized to be one of the reasons why NAP1/027 strains are more often associated with severe disease and with higher rates of mortality.1,2
The NAP1/027 isolates produce a third toxin, CDT,1,2 which is an ADP-ribosyltransferase that also disrupts the actin cytoskeleton of its target cells.31 The exact role of CDT in disease, if any, is not known. Conflicting data on the role of this toxin exists, since naturally occurring isolates of C. difficile that encode binary toxin but not toxin A or toxin B were shown to be avirulent in the hamster model, despite the observation that purified CDT caused fluid accumulation and hemorrhage in ligated ilieal loops.32 Note that most animal isolates of C. difficile produce functional CDT, suggesting that this toxin may be an important factor in animal associated disease.33 Generation of isogenic mutants in the CDT-encoding genes of a NAP1/027 isolate, or another binary toxin producing strain, together with virulence testing in animals will determine the role that this toxin plays in C. difficile associated disease. Such studies are currently underway in our laboratory.
C. difficile isolates are known to be very heterogeneous across the PaLoc region. This heterogeneity has led to the development of a C. difficile typing scheme known as toxinotyping, which classifies strains into toxinotypes that are defined by specific changes within the PaLoc region.34 There are currently 24 variant toxinotypes in addition to the reference strain VPI 10463, which is classified as type 0. The impact of variant toxinotypes on disease is not known and, with the exception of toxinotype III (NAP1/027 strains are classified as toxinotype III), there does not appear to be any one toxinotype that is associated with a particular disease outcome.34,35 It is clear, however, that mutations within the PaLoc can result in significant changes to the toxins produced. The most drastic effects are seen in the toxin A−B+ strains (toxinotype VIII and X) and the A−B−strains, such as toxinotype XIa and XIb, in which changes in the PaLoc have resulted in the loss of one or both functional toxins.21 Other variants have been documented that produce toxins that differ in size, antigenicity and substrate specificity.
A potentially important toxin B variant is produced by NAP1/027 strains, in which toxin B has extensive alterations in the C-terminal receptor binding domain in comparison to toxin B from the reference strain VPI 10463.36 These changes could potentially have increased the affinity of the toxin for its cellular receptor, or changed its receptor specificity. Although specific binding studies are yet to be performed, in vitro cytotoxicity assays have highlighted differences in both toxicity and cell line specificity of the variant toxin B produced by NAP1/027 strain R20291 compared to toxin B of strain 630 (a toxinotype 0 strain), with the NAP1/027 toxin being more toxic on a broader range of cell types.37 It seems feasible that these differences may contribute to the increased virulence of these strains. Sequence analysis of toxin genes in a number of other strains has led to the identification of other potential receptor binding and translocation domain variants,38 but the clinical relevance of these toxin variants is not known.
Variant toxins with altered substrate specificity might also be of clinical significance. Toxinotypes VIII, X and XIV for example have mutations within toxin B that affect the target Rho family GTPases. Toxin B from the reference strain VPI 10463 glucosylates Rho, Rac and Cdc42, whilst toxin B from toxinotype VIII and X strains no longer modifies Rho. All three variant toxins now glucosylate Rap, Ral and R-Ras.39 Interestingly, the variant toxin B produced by strain 8864 (a toxinotype X strain) was found to be more toxic in the mouse model than wild-type toxin B, having a lethal dose approximately 8-fold lower than the VPI 10463 toxin.40 Whether increased toxicity plays a role in disease outcome is not known.
It is important to consider the role of the animal model used when trying to decipher the roles of any putative virulence factors in human disease. This may be an important factor in the context of CDI. The Syrian golden hamster model is thought to most closely resemble the human form of disease, especially with respect to the susceptibility of the animal to infection following the administration of antimicrobial agents.41 There are, however, some very significant differences. Hamsters are exquisitely sensitive to the toxins and are rapidly killed following exposure, whilst in humans severe cases of CDI and death are less common. Furthermore, the site of infection in the hamster is the caecum and, to a lesser extent the ascending colon, whilst in humans it is restricted to the descending colon.42 The fact that toxin B alone elicits severe enterotoxic effects only in a humanized rodent model again highlights that important differences exist between humans and the animal model, which may make the roles of the individual toxins in human disease more difficult to elucidate.
In the past, clinical diagnosis of C. difficile was performed predominantly using toxin A specific EIA, until the discovery of clinically important toxin A−B+ C. difficile isolates that could not be detected using these assays. In the absence of culturing for the organism, our recently published data20 emphasizes the need to use techniques such as EIA or cytotoxicity assays that will detect the presence of both toxins A and B, not just toxin A.
Recently, several new therapeutics, including synsorb 90 and telovamer, which block the action of the C. difficile toxins rather than eradicating the organism, have been developed as potential new treatments for CDI. The development of synsorb 90 was abandoned following phase III clinical trials,43 presumably because of poor efficacy, and phase III clinical trials using telovamer have recently shown the drug to have poor in vivo efficacy.43 Whether the poor performance of these drugs reflects the fact that both preferentially target toxin A29,44 is not known. Nevertheless, our findings suggest that any new members of this class of drug should be designed to preferentially target toxin B if it is not possible to simultaneously target both toxins.
Our study clearly demonstrated the importance of using the natural infection process when studying disease pathogenesis. Although, studying the action of the purified toxins has been critical to dissecting their biochemical function and providing an insight into the role of these toxins in disease, we have shown that the importance of each toxin in the context of infection with the whole organism cannot be predicted exclusively from such studies. Our studies have highlighted the importance of using isogenic mutants when studying pathogenesis and disease.20
Research in our laboratory was supported by grants from the Australian National Health and Medical Research Council and grant AI057637 from the United States National Institute of Allergy and Infectious Diseases.
Previously published online: www.landesbioscience.com/journals/gutmicrobes/article/10768