Defining the Roles of TcdA and TcdB in Localized Gastrointestinal Disease, Systemic Organ Damage, and the Host Response during Clostridium difficile Infections

ABSTRACT

Clostridium difficile is a leading cause of antibiotic-associated diarrhea, a significant animal pathogen, and a worldwide public health burden. Most disease-causing strains secrete two exotoxins, TcdA and TcdB, which are considered to be the primary virulence factors. Understanding the role that these toxins play in disease is essential for the rational design of urgently needed new therapeutics. However, their relative contributions to disease remain contentious. Using three different animal models, we show that TcdA⁺ TcdB⁻ mutants are attenuated in virulence in comparison to the wild-type (TcdA⁺ TcdB⁺) strain, whereas TcdA⁻ TcdB⁺ mutants are fully virulent. We also show for the first time that TcdB alone is associated with both severe localized intestinal damage and systemic organ damage, suggesting that this toxin might be responsible for the onset of multiple organ dysfunction syndrome (MODS), a poorly characterized but often fatal complication of C. difficile infection (CDI). Finally, we show that TcdB is the primary factor responsible for inducing the in vivo host innate immune and inflammatory responses. Surprisingly, the animal infection model used was found to profoundly influence disease outcomes, a finding which has important ramifications for the validation of new therapeutics and future disease pathogenesis studies. Overall, our results show unequivocally that TcdB is the major virulence factor of C. difficile and provide new insights into the host response to C. difficile during infection. The results also highlight the critical nature of using appropriate and, when possible, multiple animal infection models when studying bacterial virulence mechanisms.

IMPORTANCE

Clostridium difficile is a leading cause of antibiotic-associated diarrhea and an important hospital pathogen. TcdA and TcdB are thought to be the primary virulence factors responsible for disease symptoms of C. difficile infections (CDI). However, the individual contributions of these toxins to disease remain contentious. Using three different animal models of infection, we show for the first time that TcdB alone causes severe damage to the gut, as well as systemic organ damage, suggesting that this toxin might be responsible for MODS, a serious but poorly understood complication of CDI. These findings provide important new insights into the host response to C. difficile during infection and should guide the rational development of urgently required nonantibiotic therapeutics for the treatment of CDI.

INTRODUCTION

Hospital-acquired infection with the Gram-positive, spore-forming bacterium Clostridium difficile is a major global public and veterinary health concern. This pathogen causes severe gastrointestinal illness and death and is the most significant cause of hospital-acquired diarrhea in many countries, which places a considerable economic burden on health care systems (1). The importance of C. difficile occurrence in animals has also recently become apparent, with disease or carriage identified in domestic animals, including pigs, cattle, horses, and companion animals (2, 3). C. difficile causes a spectrum of gastrointestinal diseases, collectively known as C. difficile infections (CDI), which can range from mild diarrhea through moderately serious disease to severe pseudomembranous colitis (1). Unlike the case for other enteric pathogens, disease is almost always associated with antimicrobial therapy. Importantly, the increase in antibiotic-resistant so-called “superbugs” in recent years has led to much higher usage of broad-spectrum antibiotics. Paradoxically, treating these superbugs has left patients vulnerable to infection by opportunistic pathogens, such as C. difficile.

After infection is established, most C. difficile strains produce two major toxins, TcdA and TcdB, which are encoded by the tcdA and tcdB genes and are both members of the large clostridial cytotoxin family. These toxins are potent monoglucosyltransferases that irreversibly modify members of the Rho family of host regulatory proteins, leading to disruption of downstream signaling pathways and cell death (4). Infection with toxigenic C. difficile strains therefore causes extensive colonic inflammation and epithelial tissue damage. The net effect is rapid fluid loss into the intestinal lumen, which manifests as diarrhea (4). C. difficile isolates that produce TcdB but not TcdA have emerged and continue to be isolated; these isolates cause the full clinical spectrum of CDI despite only producing one of the major toxins (5). Many human and animal strains also produce a third toxin, binary toxin or CDT, encoded by the cdtA and cdtB genes (6). The role of this toxin during infection and disease remains to be elucidated; however, recent studies suggest that this toxin may play a role in adherence and colonization of C. difficile in the host (6).

In the absence of methods to genetically manipulate C. difficile, early disease studies involved the intragastric administration of purified toxins to animals. This research suggested that TcdA was the major virulence factor and that TcdB played a less important role in disease (7). This hypothesis was challenged upon the emergence of naturally occurring variant strains that did not produce TcdA but did produce TcdB (TcdA⁻ TcdB⁺) that caused fulminant human CDI (5). Two recent infection studies in hamsters attempted to clarify the roles of TcdA and TcdB in disease by using isogenic toxin mutants constructed in the low-virulence clinical isolate 630 (8, 9). The first study found that TcdB alone resulted in disease (9), while the second concluded that both TcdB and TcdA could individually cause severe disease (8). More recently, a third study using isogenic mutants constructed in strain R20291, an epidemic BI/NAP1/027 strain, together with a hamster infection model also concluded that both toxins could cause fulminant disease independently (10). A number of possibilities could explain the discrepancies in the first two studies, such as genetic differences in the C. difficile 630-derived strains, differences in animal challenge protocols, or variations in the intestinal microbiota of animals at different research facilities, which can substantially influence infection outcomes (11). As a consequence, it was suggested that the same panel of strains should be virulence tested in multiple laboratories to minimize the impact of experimental variation (12). To address these disparities and comprehensively define the contributions of TcdA and TcdB to disease, we used three independent animal models to study the relative virulence of a new group of isogenic C. difficile toxin gene mutants derived from a Canadian epidemic C. difficile BI/NAP1/027 isolate (strain M7404) that had not been extensively laboratory passaged.

RESULTS

Construction and characterization of tcdA, tcdB, and cdtA toxin gene mutants in a BI/NAP1/027 clinical isolate.

Two independent tcdA (TcdA⁻ TcdB⁺) and tcdB (TcdA⁺ TcdB⁻) mutants (mutants 1 and 2 for each) and a tcdAB (TcdA⁻ TcdB⁻) double mutant were constructed in strain M7404. Like all BI/NAP1/027 isolates, this strain encodes a third toxin, CDT, an ADP-ribosyltransferase binary toxin (13), which was disrupted separately by mutagenesis of the cdtA gene (CDT⁻). Before virulence testing, each mutant was genotypically confirmed by PCR (data not shown) and Southern hybridization analysis, which confirmed the specific targetron insertions (see Fig. S1 in the supplemental material). Western immunoblot analyses (see Fig. S2) together with Vero and HT29 cell cytotoxicity and neutralization assays (see Fig. S3 and S4) then confirmed the expected toxin profile of each mutant compared to that of the wild type. No differences in the sporulation frequencies of the mutant and wild-type strains were detected (see Table S1).

TcdB-producing strains are more virulent than tcdB-deficient isogenic mutants in mouse and hamster infection models.

To maximize the robustness of the infection outcomes, we used three different animal models, performed in three independent laboratories. First, an optimized mouse infection model, based on a previously published model (14) and further developed at Monash University (Australia), in which animals consistently develop severe disease was used. Infection with the wild type (TcdA⁺ TcdB⁺) and the TcdA⁻ TcdB⁺ mutants resulted in marked weight loss (Fig. 1A), with 100% and 95% of mice, respectively, dying within 48 h, rising to 100% for the latter group by 72 h (Fig. 1B). These animals also showed other signs of severe disease, such as shallow and labored breathing, profuse diarrhea, and isolation from littermates (Fig. 1C). In contrast, only moderate weight loss was recorded for TcdA⁺ TcdB⁻ strain-infected mice (Fig. 1A); significantly more (80%) of these animals survived than did wild-type strain-infected animals (log rank test, P < 0.0001) (Fig. 1B), and they showed significantly less signs of physiological distress (Mann-Whitney test, P < 0.001) (Fig. 1C). Importantly, a 500-fold-higher infectious dose of the TcdA⁺ TcdB⁻ strain did not elicit more severe disease outcomes (see Fig. S5A and B in the supplemental material) than the lower dose (Fig. 1A and B). Finally, TcdA⁻ TcdB⁻ strain-infected mice lost no weight (Fig. 1A), survived the infection (Fig. 1B), and exhibited no other disease signs (Fig. 1C). As with the TcdA⁺ TcdB⁻ strain-infected mice, the results for survival and physiological distress were significantly different in TcdA⁻ TcdB⁻ strain-infected mice than in mice infected with the wild-type strain (P < 0.0001 for both survival and distress). Statistical analysis using a Mann-Whitney test subsequently showed that mice infected with TcdB-producing strains had significantly shorter colons than mock-infected mice (P < 0.0001 for the wild-type strain, P = 0.004 and P < 0.001 for TcdA⁻ TcdB⁺ mutants 1 and 2, and P < 0.001 and P < 0.001 for CDT⁻ mutants 1 and 2) (Fig. 1D), a common feature in mouse models of ulcerative colitis (15). The colon lengths in the TcdA⁺ TcdB⁻ (P = 0.345 and P = 0.540) and TcdA⁻ TcdB⁻ (P = 0.862) strain-infected mice were statistically the same as in mock-infected mice (Fig. 1D).

FIG 1 .

Toxin B is the primary mediator of fulminant C. difficile disease. (A to D) Monash mice were infected with C. difficile strains. Results for the strains are indicated by different colors as follows: black, wild-type strain M7404 [TcdA⁺B⁺ (WT)] (n = 19); light blue, tcdB mutant 1 (TcdA⁺B⁻1) (n = 14); dark blue, tcdB mutant 2 (TcdA⁺B⁻2) (n = 10); pink, tcdA mutant 1 (TcdA⁻B⁺1) (n = 13); purple, tcdA mutant 2 (TcdA⁻B⁺2) (n = 10); green, tcdA tcdB double mutant (TcdA⁻B⁻) (n = 15); light orange, cdtA mutant 1 (CDT⁻1) (n = 10); dark orange, cdtA mutant 2 (CDT⁻2) (n = 10); and gray, mock-infected control (mock) (n = 5). Results are shown for weight loss (A), survival (B), activity and appearance (C), and colon length (D) of infected Monash mice. (E) Survival of Sanger mice infected with C. difficile strains indicated by different colors as follows: black, wild-type TcdA⁺ TcdB⁺ strain [TcdA⁺B⁺ (WT)] (n = 16); light blue, TcdA⁺ TcdB⁻ mutant 1 (TcdA⁺B⁻1) (n = 16); pink, TcdA⁻ TcdB⁺ mutant 1 (TcdA⁻B⁺1) (n = 13); light orange, cdtA mutant 1 (CDT⁻1) (n = 15); and gray, mock infected (n = 15). (F) Days from infection to death of Hines hamsters infected with C. difficile strains indicated by different colors as follows: black, wild-type TcdA⁺ TcdB⁺ strain [TcdA⁺B⁺ (WT)] (n = 8); light blue, TcdA⁺ TcdB⁻ mutant 1 (TcdA⁺B⁻1) (n = 9); pink, TcdA⁻ TcdB⁺ mutant 1 (TcdA⁻B⁺1) (n = 9); and light orange, cdtA mutant 1 (CDT⁻1) (n = 10). Data represent the mean results ± standard deviations (SD).

Disease resulting from infection with the CDT⁻ mutants was indistinguishable from disease caused by the wild type for each parameter analyzed (Fig. 1A to D). Note that in all subsequent animal infection experiments, only one TcdA⁻ TcdB⁺, TcdA⁺ TcdB⁻, or CDT⁻ mutant was assessed since the infection phenotypes between the independent mutants using the Monash model were identical.

Strain virulence was next tested at the Sanger Institute (United Kingdom), using a different mouse infection model in which animals typically develop self-limiting intestinal inflammation and rarely progress to severe disease or death (16). Unexpectedly, 40% of TcdA⁻ TcdB⁺ strain-infected mice succumbed to acute infection within 48 h, which was a significantly higher rate than for mice infected with the wild-type strain (log rank test, P = 0.015) (Fig. 1E). In contrast, animals infected with the wild-type, TcdA⁺ TcdB⁻, TcdA⁻ TcdB⁻, or CDT⁻ strain all survived the infection (Fig. 1E).

Finally, the hamster model of CDI developed at Hines VA Hospital and used in earlier studies (9) was used here. In contrast to the mouse infection results, all infected hamsters died irrespective of the infecting strain (Fig. 1F). However, a significant delay (log rank test, P = 0.0015) was evident in the time from colonization to death of TcdA⁺ TcdB⁻ strain-infected animals (1.33 ± 0.16 days [mean ± standard deviation]) compared to the time to death for wild-type strain-infected animals (0.43 ± 0.16 days). There was no statistical difference in the time to death of TcdA⁻ TcdB⁺ (0.8 ± 0.13 days) or CDT⁻ (0.28 ± 0.12 days) strain-infected animals compared to the time to death of wild-type strain-infected animals (Fig. 1F). These results indicate that in the hamster model of infection, only the TcdA⁺ TcdB⁻ mutant was attenuated in virulence, which agreed with our previous data using toxin mutants in C. difficile strain 630 (9).

To ensure that each strain used in this analysis germinated and colonized the infected hosts with equal efficiency, the colonization efficiencies of the tcdA (TcdA⁻ TcdB⁺ mutant 1) and tcdB (TcdA⁺ TcdB⁻ mutant 1) mutants in comparison to that of the wild-type M7404 TcdA⁺ TcdB⁺ strain in the Monash mouse model of CDI and the Hines VA Hospital hamster infection model were determined. No differences were seen in either mice or hamsters, confirming that such differences were not responsible for disease attenuation resulting from infection with the TcdA⁺ TcdB⁻ mutants (see Fig. S6A and D in the supplemental material). Competition assays were also performed in which Monash mice were simultaneously infected with equal numbers of the tcdA (TcdA⁻ TcdB⁺ mutant 1) and tcdB (TcdA⁺ TcdB⁻ mutant 1) mutants to determine whether either strain was reduced in fitness in vivo (see Fig. S6E). As expected, both strains were found to be equally fit in vivo, with competitive index (CI) values of 1.178 ± 0.267 and 1.119 ± 0.361 obtained at 24 h and 48 h after infection, respectively. This result confirms that the attenuated virulence of the TcdA⁺ TcdB⁻ mutants is not due to reduced in vivo fitness of these strains. Finally, to confirm that the toxin levels produced by the mutant strains were equivalent to those produced by the wild-type strain in vivo, TcdA- and TcdB-specific cytotoxicity assays were performed on the luminal contents of mice infected with each strain (see Fig. S6B and C). Each strain was found to produce the expected toxin in vivo, at levels that were not significantly different from the levels detected from the wild-type strain.

Infection with TcdB-producing strains causes severe gut and distal organ damage.

Previous studies assessing the roles of TcdA and TcdB in disease did not investigate histological damage arising in the gut as a consequence of CDI. Here, the microscopic effects of TcdA, TcdB, and CDT on the gut—in colonic and cecal tissues from Monash mice and cecal tissues from Sanger mice, collected 2 and 4 days postinfection, respectively—were examined by histopathology. Similar observations were made for both groups, with the wild type and the TcdA⁻ TcdB⁺ mutants all causing severe gut damage associated with eroded and often absent crypts, mucosal ulceration, and goblet cell loss (Fig. 2A). Polymorphonuclear cell (PMN) influx into the lamina propria, enterocyte hyperplasia, and severe submucosal edema associated with hemorrhage were also seen (Fig. 2A). In contrast, TcdB-negative strains caused limited tissue damage that was confined to mild edema and PMN influx (Fig. 2A), even when a 500-fold increased dose of the TcdA⁺ TcdB⁻ strain relative to the dose of the wild type was used (see Fig. S5C to E in the supplemental material). All of the damage was TcdB- or TcdA-induced, since tissues from TcdA⁻ TcdB⁻ strain-infected mice resembled those from mock-infected control mice (Fig. 2A). Independent histological scoring confirmed these observations, with TcdB-producing strains eliciting the greatest injury and strains only producing TcdA causing less damage (Fig. 2B and C). Statistical analysis using a Mann-Whitney test showed significantly less intestinal damage than in wild-type strain-infected mice only in animals infected with the TcdA⁺ TcdB⁻ mutants (P = 0.0022 and P = 0.0043) and the TcdA⁻ TcdB⁻ mutant (P = 0.0022) in the Monash model of CDI and in the TcdA⁺ TcdB⁻ mutant-infected mice (P < 0.0001) in the Sanger model of CDI. Histopathological scoring of tissues from CDT⁻ strain-infected mice resembled the results for wild-type strain-infected mice (Fig. 2B and C).

FIG 2 .

Toxin B causes severe local histopathological damage. Histopathological analyses of tissues collected from mice infected with the wild-type strain M7404 [TcdA⁺B⁺ (WT)], tcdA mutant 1 (TcdA⁻B⁺), tcdB mutant 1 (TcdA⁺B⁻), tcdA tcdB double mutant (TcdA⁻B⁻), or cdtA mutant 1 (CDT⁻1) or mock infected with PBS were performed. (A) Representative images of hematoxylin-and-eosin-stained tissues from Monash mice on day 2 postinfection and from Sanger mice on day 4 postinfection are shown. Scale bars are shown (200 µm). (B) Histopathological scoring of damage to tissues from Monash mice infected with the wild-type and mutant C. difficile strains. Scores are shown for tissues from mice infected with the TcdA⁺ TcdB⁺ (WT) strain (n = 12), TcdA⁺ TcdB⁻ mutant 1 (n = 6), TcdA⁺ TcdB⁻ mutant 2 (n = 5), TcdA⁻ TcdB⁺ mutant 1 (n = 6), TcdA⁻ TcdB⁺ mutant 2 (n = 6), TcdA⁻ TcdB⁻ double mutant (n = 6), CDT⁻ mutant 1 (n = 5), and CDT⁻ mutant 2 (n = 5) and from mock-infected mice (n = 6). All values are mean results ± SD. (C) Histopathological scoring of damage to tissues in Sanger mice infected with wild-type and mutant C. difficile strains. Scores are shown for tissues from mice infected with the TcdA⁺ TcdB⁺ (WT) strain (n = 12), TcdA⁺ TcdB⁻ mutant 1 (n = 12), TcdA⁻ TcdB⁺ mutant 1 (n = 8), and CDT⁻ mutant 1 (n = 12) and from mock-infected mice (n = 12). All values are mean results ± SD.

C. difficile infection can cause multiple organ dysfunction syndrome (MODS) (17, 18), but the role of toxins is unknown. Organs were therefore collected from Monash CDI model mice and their histopathology examined; 88% of wild-type strain-infected and 80% of TcdA⁻ TcdB⁺ strain-infected mice had damage to the thymus, sagittal lymph nodes, spleen, or kidneys (Fig. 3A). No organ damage was detected in mice infected with the TcdA⁺ TcdB⁻ or TcdA⁻ TcdB⁻ strain or mock-infected mice (Fig. 3A). To determine whether an increased amount of TcdA could elicit systemic effects similar to those observed with TcdB-producing strains, mice infected with a 500-fold higher infectious dose of the TcdA⁺ TcdB⁻ strain were also examined; however, no organ damage was detected (see Fig. S5C in the supplemental material). Infection with TcdB-producing strains was particularly devastating to the thymus, as evidenced by a complete loss of medullary and cortical junctions and widespread lymphocyte apoptosis (Fig. 3B). In severe cases, there were also signs of lymphoid necrosis (lymphorrexhis) found outside the secondary follicles and associated with phagocytic macrophages. Similarly, in wild-type strain- and TcdA⁻ TcdB⁺ strain-infected mice, sagittal lymph nodes showed a reactive phenotype categorized by the presence of lymphocytic apoptosis, areas of lymphoid depletion, and focal regions of macrophage aggregates and multinucleated giant cells, as well as a decrease in the number and size of lymphoid follicles and a marked absence of germinal centers (Fig. 3B). Mild lymphocyte apoptosis was also observed in splenic tissue isolated from mice infected with the wild type and the TcdA⁻ TcdB⁺ mutant (Fig. 3B), with focal areas of tissue showing loss of red and white pulp morphology, as well as an increase in tingible body macrophages and periarteriolar lymphoid sheaths (PALS). Kidney damage was only seen in wild type-infected mice, where capsular lesions were observed.

FIG 3 .

C. difficile infection with TcdB-producing strains is associated with multiorgan damage. (A) Organs were collected from Monash mice and assessed for damage compared to the state of control tissues, and the data collated. (B) Representative images of organ tissues collected from Monash CDI mice infected with wild-type strain M7404 [TcdA⁺B⁺ (WT)], tcdA mutant 1 (TcdA⁻B⁺), tcdB mutant 1 (TcdA⁺B⁻), or tcdA tcdB double mutant (TcdA⁻B⁻) or mock infected with PBS. Note the loss of structure in medullary and cortical regions in the thymus (M, medulla, C, cortex) and in the germinal centers (Gc) of the sagittal lymph nodes (SG Lns). Scale bars are shown (thymus, 500 µm; spleen, 200 µm; SG Lns and kidney, 100 µm).

Infection with C. difficile toxin gene mutants elicits unique host response signatures.

To further explore the effects of TcdA and TcdB on the host, we studied the immune response transcriptomic signatures seen in the large intestinal tissues collected from Monash and Sanger wild-type, TcdA⁻ TcdB⁺, and TcdA⁺ TcdB⁻ strain- and mock-infected mice (see Table S2 in the supplemental material). Tissues collected from CDT⁻ strain-infected mice were not included in this analysis, since CDT appears to have little effect on virulence under the conditions tested in this study. Hierarchical clustering of samples showed that most TcdA⁻ TcdB⁺ strain-infected tissues displayed expression profiles similar to those of wild-type strain-infected tissues, whereas TcdA⁺ TcdB⁻ strain-infected samples clustered closely with mock-infected tissues, consistent with the attenuated virulence phenotype of the TcdA⁺ TcdB⁻ strains (Fig. 4). Infection with TcdB-producing strains (wild-type and TcdA⁻ TcdB⁺ strains) induced expression changes in a diverse group of host genes involved in the innate immune response, inflammation, and cellular apoptosis. These included s100a8, lcn2, cxcl1, il1b, and duox2, which encode the S100 calcium binding protein, lipocalin 2, chemokine ligand 1, interleukin 1β, and dual oxidase 2, respectively. In the absence of TcdB, these genes did not appear to be differentially regulated. Importantly, however, the expression of many of these genes was higher following infection with the wild-type strain than with the TcdA⁻ TcdB⁺ mutant, suggesting that TcdA, as well as TcdB, is involved in modulating the host innate immune and inflammatory response to C. difficile infection. Many genes involved in cellular metabolism were also differentially regulated in response to CDI, including genes encoding enzymes such as synthases, methyltransferases, oxidases, carboxypeptidases, and dehydrogenases. In addition, a profound effect on cellular detoxification pathways was noted, with a number of UDP-glucuronosyltransferase genes (ugt1a6b, ugt1a6a, ugt1a1, ugt1a7c, ugt1a10, ugt2b36, and ugt2b35), cytochrome P450 family protein-encoding genes (cyp2c55, cyp2c65, cyp4b1, cyp4f14, cyp2d12, cyp3a13, cyp2d22, and cyp2d26), and carboxylesterase genes (ces1f, ces2, ces1d, and ces2e) significantly downregulated. Although unrelated, these protein families play a key role in protecting host cells from damage caused by endogenous and exogenous toxins, as well as being involved in the removal of xenobiotic substances (19–21). Finally, several genes encoding solute carrier (SLC) family proteins (slc37a1, slc26a3, slc17a4, slc39a5, slc30a10, slc26a2, slc40a1, slc16a5, slc9a2, slc02b1, slc16a9, and slc5a6) were also found to be downregulated, predominantly in response to infection with the wild-type strain and the TcdA⁻ TcdB⁺ mutant. These proteins play a key role in many central metabolic cellular processes since they are involved in the transport of a variety of substrates, such as sugars, inorganic ions, nucleotides, and amino acids (22). Collectively, these data highlight the profound effect that C. difficile infections have on central processes within host epithelial cells, particularly in response to TcdB. Furthermore, unique signatures of the host response to each toxin mutant strain were identified, suggesting that clinical C. difficile strains might elicit differential host responses depending on the combination of toxins produced by the infecting strain.

FIG 4 .

Transcriptomic analysis of the host response during C. difficile infection. Heat maps of the transcriptomes of colonic tissues collected from mice following infection with the wild-type strain M7404 [TcdA⁺B⁺ (WT)], tcdB mutant 1 (TcdA⁺B⁻), or tcdA mutant 1 (TcdA⁻B⁺) or mock infection with PBS using the Monash mouse model (A) or the Sanger mouse model (B) of CDI. Scaled expression values within the heat maps are color coded according to the color range shown, with red being the lowest and green being the highest level of transcription. The dendrogram shown above each heat map depicts hierarchical clustering of the transcriptomic response of each mouse following infection and was derived using a Pearson hierarchical-clustering algorithm. The infecting strains are indicated underneath the individual heat maps.

DISCUSSION

Over the last 30 years, most studies defining the roles played by the C. difficile major toxins in disease utilized purified TcdA or TcdB and their direct administration to animals or application to cell lines, which provided invaluable insights into the mechanism of action of each of the toxins. It has only been in the last decade, however, that genetic manipulation technology has allowed toxin gene mutants to be constructed, facilitating hamster infection studies that have allowed the roles of the toxins to be more accurately defined (8, 9, 23). These infection studies all suggested that TcdB plays a more important role in disease than suggested by the earlier work involving purified toxins (8–10, 24). The study presented here has further defined the roles of TcdA and TcdB and has also assessed the role of CDT binary toxin in CDI, using three different animal models of disease. This work comprehensively shows that both TcdA and TcdB play a role in disease pathogenesis but that TcdB is the major virulence factor and causes severe host damage and disease in all animal models. In contrast, the CDT binary toxin appears to play a minor role in CDI under the experimental conditions used in this study, although other studies have suggested that this toxin can enhance colonization (25) or cause tissue damage in a small number of animals in a hamster CDI model (10). It is clear from this and other studies that CDT is not a major virulence factor, with a minority of virulent strains producing this toxin (13). Furthermore, a previous study has shown that TcdA- and TcdB-negative but CDT-positive strains are avirulent in the hamster infection model (26). The design of future experiments to clarify the role of CDT during infection will need to be carefully considered, since they will need to capture the subtle and synergistic effects that this toxin is likely to be exerting on the host. At this time, however, the role of CDT in C. difficile infection and disease pathogenesis remains undefined.

Although many of the recent infection studies have unquestionably clarified the role of TcdB in CDI and showed that TcdB function does not depend on the presence of TcdA, the role of TcdA itself is less apparent (8–10, 24). Here, we clearly show in all three infection models that TcdA plays a less important role than TcdB in CDI caused by BI/NAP1/027 C. difficile isolates. Histopathological analysis of cecal and colonic tissues collected from infected mice showed that TcdB was responsible for the majority of intestinal damage arising during infection, with TcdA causing more superficial and localized damage. This finding contrasts with previous suggestions that TcdA is the primary cause of damage to the colonic niche during the course of a C. difficile infection (8), although it agrees with other research that showed that TcdB is a more potent enterotoxin than TcdA in human intestinal explants (27). Unexpectedly, this work has also identified a role for TcdA in modulating the severe effects of TcdB in the Sanger murine CDI model. Using this model, mice infected with the TcdA⁻ TcdB⁺ strain succumbed to infection, while those infected with the wild-type or the TcdA⁺ TcdB⁻ mutant strain did not. This observation aligns with previous clinical observations suggesting that naturally occurring TcdA⁻ TcdB⁺ strains are capable of causing more severe disease than some TcdA⁺ TcdB⁺ strains (5). Furthermore, a recent study showed that the administration of TcdA neutralizing antibodies in piglets with CDI resulted in more severe disease than in untreated animals (28). While further work is needed to determine the mechanism by which TcdA modulates the effects of TcdB, the simplest explanation may be one of competition: TcdA intoxication of cells may prevent TcdB from acting on the same cells, thereby limiting the pool of cells on which TcdB can act and consequently reducing disease severity. Regardless, these observations have important implications for the development of next-generation, nonantibiotic, toxin-directed CDI therapeutics, because they suggest that these approaches should focus on both TcdA and TcdB or TcdB alone since targeting only TcdA may return adverse clinical outcomes. The increasing isolation of clinical and veterinary TcdA⁻ TcdB⁺ variants also reinforces the need for the development of TcdB-based therapies.

The severity of disease caused by TcdB-producing strains may result from the extraintestinal organ damage that is caused by infection with these strains. These systemic effects were absent in mice challenged with C. difficile strains that did not produce TcdB. These observations align with previous findings showing that TcdB is cardiotoxic in a zebrafish embryo model of intoxication (29) and a study showing that the administration of TcdB-specific antibodies lessened the systemic effects of CDI in a piglet infection model (28). Furthermore, another study showed that both TcdA and TcdB could disseminate from the gut during infection with a strain that produced both toxins, leading to systemic toxemia (18). Our results suggest that TcdB is the factor responsible for systemic dissemination and damage and may therefore be responsible for the onset of MODS, a severe complication of CDI that is associated with high mortality rates. Although this study cannot discern whether extraintestinal damage is directly mediated by TcdB or is a consequence of toxin-induced “leaky gut” that compromises gut integrity and allows the systemic dispersal of microbial components (30), including other toxins, it is clear that TcdB alone is associated with systemic and severe disease.

The systemic effects of infection with TcdB-producing C. difficile strains extended to many organs, including lymphoid tissues and, in particular, the thymus. Thymic injury during infection with other pathogens has been described (14–16) and can be associated with poor outcomes postrecovery (31). Thymic atrophy associated with infectious diseases can result in temporary lymphocyte depletion and altered circulating T-cell populations, which is of little consequence in healthy individuals where the thymus and immune functions are restored rapidly following infection. However, age-related thymic involution can result in an inability to restore immune function following infection (31). CDI effects on the thymus may influence immune function postinfection in a similar way, resulting in an increased risk of disease relapse or secondary infections, such as funguria or vancomycin-resistant Enterococcus (VRE) infections, which are associated with CDI patients (32, 33). This finding is particularly relevant to elderly patients, a group at significant risk for CDI (34) who have already undergone thymic atrophy (31).

Transcriptomic analysis of mice infected with the wild-type and mutant strains revealed that TcdB is the major factor inducing host innate immune and proinflammatory responses. These responses were generally stronger in mice infected with the wild-type strain than in those infected with the TcdA⁻ TcdB⁺ derivative, suggesting that TcdA also plays a role in upregulating these responses, which agrees with similar studies using purified toxins (35, 36). Of note, TcdB was recently shown to induce cellular apoptosis via NADPH oxidase-mediated production of reactive superoxide species, and a reduction in reactive oxygen species (ROS) activity protected colonic explants from TcdB-induced damage (37). Although upregulation of NADPH oxidase genes was not detected in our study, one of the most upregulated pathways following infection was found to be Duox2, which produces ROS in a way similar to NADPH oxidase. Duox2 is predominantly expressed in gastrointestinal epithelial cells, where it plays an important role in host defense against invading pathogens, such as Listeria monocytogenes, Salmonella enterica serovar Typhimurium, and Helicobacter spp., through the inducible production of hydrogen peroxide and ROS (38). As with NADPH oxidase (37), it appears that TcdB might induce aberrant activation of the Duox2 pathway within the intestinal epithelium during infection. Since ROS is known to cause organ damage and is also implicated in the pathogenesis of gastrointestinal conditions, such as small intestine ischemia and ulcerative colitis (39–41), it is possible that the uncontrolled Duox2-mediated burst of H₂O₂ and ROS resulting from C. difficile infection with TcdB-producing strains contributes to gut damage. Although further work is needed to understand the role that upregulation of Duox2 may play in C. difficile-induced gut damage, this pathway may represent a novel therapeutic target for reducing the level of gastrointestinal damage during CDI by reducing ROS production. Similar approaches have been used to develop a promising class of novel therapeutics that inhibit NADPH oxidases for the treatment of numerous inflammatory diseases and cancers (42, 43).

Overall, our results have important implications for the development and validation of treatment agents for CDI and for disease pathogenesis studies. Genetically, C. difficile is a heterogeneous species, with new disease-causing variants emerging regularly. Although many cases of CDI are acquired within hospitals, recent molecular epidemiological studies have shown that many other environmental sources may contribute to the infection reservoir. Regardless of the environmental source, TcdB appears to be the only toxin common to all current and emerging human and animal disease-causing isolates (44–46), reinforcing the outcomes presented here and the need for TcdB-targeted therapeutics. Our results also support the use of multiple animal infection models for the purpose of comprehensive evaluation of therapeutics. Finally, despite being an infection that is confined to the gut, C. difficile infection is clearly shown by this work to cause damage to organs distal to the infection site, and this work also provides new insights into the systemic host damage that can occur as a consequence of a localized gastrointestinal infection.

MATERIALS AND METHODS

Construction and characterization of toxin gene mutants.

The tcdA, tcdB, and tcdAB C. difficile mutants were made using targetron technology and utilizing appropriately retargeted derivatives of plasmid pDLL1 and conjugative matings from a B. subtilis BS34A donor strain, as previously described (23). For isolation of the double toxin mutant, antibiotic selection could not be used, and so this strain was detected by PCR screening for both the tcdA- and tcdB-specific insertion of the targetron. The targetron element inserted after nucleotide 4068 of the sense strand for tcdA, nucleotide 1587 of the sense strand for tcdB, and nucleotide 421 of the sense strand for cdtA. The tcdAB double toxin mutant was constructed using tcdB mutant 1 as the parent strain. To verify that the targetron insertions had occurred as anticipated, PCR using the following oligonucleotide primers were used: for tcdA, EBS-universal (sigma) and JRP3602 (TAAATGTACTACCTACAATAACAGAGGG); for tcdB, EBS-universal and JRP1592 (GTGGCCCTGAAGCATATG); and for cdtA, EBS-universal and JRP1744 (GGGAAAGAAAAGAAGCAGAAAG). Strain numbers were assigned as follows: for tcdA mutant 1, DLL3043; for tcdA mutant 2, DLL3045; for tcdB mutant 1, DLL3101; for tcdB mutant 2, DLL3102; and for the tcdAB mutant, DLL3121.

Each mutant was also confirmed by Southern hybridization analysis as described previously, using an ermB-specific PCR product (47), a tcdB- or tcdA-specific PCR product (9), or a cdt-specific PCR product amplified using primers JRP1746 (GGAAGACGAAGATTTGGATACA) and JRP2505 (GGTTTTAGCTCAGACATAGGGA).

To confirm the correct toxin production profiles in each mutant and wild-type strain, TcdA-, TcdB-, and CdtA-specific Western blot analyses were performed as described previously (9, 13), except that toxins were precipitated from culture supernatants using chloroform-methanol (8), and Vero and HT29 cell cytotoxicity and neutralization assays were also performed as described previously (9).

Complementation of the tcdA and tcdB genes was attempted; however, despite multiple attempts, it did not prove possible to clone the intact tcdA or tcdB gene into an appropriate shuttle plasmid that would facilitate conjugative transfer into C. difficile.

Animal infection trials.

Virulence trials using the Monash and Sanger mouse models and the Hines hamster model were performed as described previously (2) and are described in detail in the supplemental material. For the Monash mouse model, all monitoring was carried out in accordance with Victorian State Government regulations and the Monash University Animal Ethics guidelines. All experiments were approved by the Monash University SOBS B Animal Ethics Committee. All Sanger mouse model experiments were performed in accordance with the United Kingdom Home Office Animals (Scientific Procedures) Act of 1986. All hamster experiments were approved by the Institutional Animal Care and Use Committee (IACUC) at the Hines VA Hospital.

Histopathological scoring and staining.

All histopathological analysis was performed by an independent, certified pathologist at the Australian Phenomics Network, University of Melbourne. Monash and Sanger colonic and cecal sections were assessed using a scoring system based on previously established parameters (26), described in detail in Table S3 in the supplemental material. Terminal deoxynucleotidyltransferase-mediated dUTP-biotin nick end labeling (TUNEL) staining was performed on thymus sections using an in situ cell death detection kit with fluorescein (Roche), following the manufacturer’s instructions.

Microarrays.

Total RNA was extracted from colonic tissues isolated from mice at Monash University and the terminal cecal tissues of mice from the Sanger Institute, using the Qiagen RNeasy kit according to the manufacturer’s instructions. RNA quality was assessed using the Agilent 2100 bioanalyzer (Agilent Technologies), and microarray gene expression analyses performed on an Illumina GEX platform. Microarray data were analyzed using GeneSpring software (Agilent Technologies). Genes showing differential expression were selected with a Benjamini-Hochberg-corrected P value of <0.05 and fold change (compared to mock-infected controls) of >2 for Monash tissues or >1.5 for Sanger tissues. The transcriptomic responses of samples were arranged as a heat map (Fig. 4) by applying a Pearson hierarchical-clustering algorithm based on both entities and conditions. Analyses of enriched Gene Ontology terms and KEGG pathways associated with differentially expressed gene sets were subsequently performed using the DAVID (Database for Annotation, Visualisation and Integrated Discovery) (27) functional classification tool and the InnateDB database (28) for innate immune response.

Note that additional methods are provided in the supplemental material.

SUPPLEMENTAL MATERIAL

Text S1 

Supplemental results and methods. Download

Figure S1 

Analysis of toxin gene mutants by Southern blotting. The wild type and isogenic mutant strains were examined by Southern blotting using HpaI-/SpeI-digested chromosomal DNA and a tcdA-specific probe (A), XbaI-digested chromosomal DNA and a tcdB-specific probe (B), HpaI-/EcoRI-digested DNA and a cdt-specific probe (C), or XbaI-digested chromosomal DNA and an ermB-specific probe that is specific for insertion of the targetron (D). Legend: TcdA⁺ TcdB⁺ (WT), strain M7404; avirulent, pathogenicity locus (PaLoc)-negative control strain CD37; TcdA⁻ TcdB⁺1/2, independent tcdA mutants; TcdA⁺ TcdB⁻1/2, independent tcdB mutants; TcdA⁻ TcdB⁻, tcdA tcdB double mutant; CDT⁻1/2, independent cdtA mutants. HindIII-digested lambda DNA was used for molecular size markers (sizes in base pairs are indicated). Download

Figure S2 

Comparative analysis of toxin production. The wild-type and isogenic mutant strains were examined by Western immunoblotting using precipitated supernatant proteins from the strains indicated and TcdA-specific antibodies (A), TcdB-specific antibodies (B), or iota A-specific antibodies, which cross-react with CDTa (C). Legend: TcdA⁺ TcdB⁺ (WT), strain M7404; avirulent, PaLoc-negative control strain CD37; TcdA⁻ TcdB⁺1/2, independent tcdA mutants; TcdA⁺ TcdB⁻1/2, independent tcdB mutants; TcdA⁻ TcdB⁻, tcdA tcdB double mutant; CDT⁻1/2 independent cdtA mutants. The molecular weight size markers are shown in kDa. Download

Figure S3 

Vero cell cytotoxicity and neutralization assays. Serial doubling dilutions of C. difficile culture supernatants were made and used in cytotoxicity assays. (A) The morphological changes of Vero cells were observed and scored by microscopy after 24 h. The endpoint was recorded as the last dilution when 100% cytopathic effect (CPE) was observed. The toxin titer is the reciprocal of the endpoint dilution. Legend: TcdA⁺ TcdB⁺ (WT), strain M7404; avirulent, PaLoc-negative control strain CD37; TcdA⁻ TcdB⁺1/2, independent tcdA mutants; TcdA⁺ TcdB⁻1/2, independent tcdB mutants; TcdA⁻ TcdB⁻, tcdA tcdB double mutant; CDT⁻1/2, independent cdtA mutants. The mean values from triplicate assays are shown together with the standard errors of the means (SEM). (B) Neutralization of cytotoxicity. C. difficile culture supernatants were added to Vero cells and morphological changes observed after 24 h; supernatants were untreated (no antibody) or pretreated for 90 min with neutralizing antibodies to TcdB (anti-TcdB), TcdA (anti-TcdA), or iota A (anti-Ia) prior to addition to the Vero cells. Representative images are shown. Scale bar indicates 100 µm. Download

Figure S4 

HT29 cell cytotoxicity and neutralization assays. Serial doubling dilutions of C. difficile culture supernatants were made and used in cytotoxicity assays. (A) The morphological changes of HT29 cells were observed and scored by microscopy after 24 h. The endpoint was taken as the last dilution when 100% CPE was observed. The toxin titer is the reciprocal of the endpoint dilution. Legend: TcdA⁺ TcdB⁺ (WT), strain M7404; avirulent, PaLoc-negative control strain CD37; TcdA⁻ TcdB⁺1/2, independent tcdA mutants; TcdA⁺ TcdB⁻1/2, independent tcdB mutants; TcdA⁻ TcdB⁻, tcdA tcdB double mutant; CDT⁻1/2 independent cdtA mutants. The mean values of triplicate assays are shown together with the SEM. (B) Neutralization of toxicity. C. difficile culture supernatants were added to HT29 cells and morphological changes observed after 24 h; supernatants were untreated (no antibody) or pretreated for 90 min with neutralizing antibodies to TcdB (anti-TcdB), TcdA (anti-TcdA), or iota A (anti-Ia) prior to addition to the HT29 cells. Representative images are shown. Scale bar indicates 100 µm. Download

Figure S5 

Infection with a higher infectious dose of a TcdA⁺ TcdB⁻ strain does not result in more severe disease or multiorgan damage. (A, B) Weight loss (A) and survival (B) of Monash mice infected with 1 × 10⁶ spores of the wild-type strain (TcdA⁺ TcdB⁺) (n = 6) (green line) or 5 × 10⁸ spores of tcdB mutant 1 (TcdA⁺ TcdB⁻) (n = 12) (blue line) or mock infected with PBS (n = 6) (grey line). (C) Representative images of organ tissues collected from Monash CDI mice infected with 1 × 10⁶ spores of wild-type strain M7404 (WT) or 5 × 10⁸ spores of tcdB mutant 1 (TcdA⁺B⁻) or mock infected with PBS (Mock). (D, E) Histopathological scoring of damage to cecal tissues (D) and colonic tissues (E) from Monash mice infected with 1 × 10⁶ spores of the wild-type strain (WT) (n = 6) or 5 × 10⁸ spores of tcdB mutant 1 (TcdA⁺B⁻) (n = 12) or mock infected with PBS (Mock) (n = 6). All values are the geometric means ± SEM. Download

Figure S6 

The TcdA⁺ TcdB⁻ and TcdA⁻ TcdB⁺ toxin mutants are equally fit in vivo and do not have a colonization defect in comparison to the wild-type strain. (A) Colonization efficiencies of the wild-type strain (TcdA⁺ TcdB⁺) (black bars) and the TcdA⁺ TcdB⁻ (blue bars) and TcdA⁻ TcdB⁺ (pink bars) isogenic toxin mutants at 12, 24, 36, and 48 h following infection of mice using the Monash model of CDI, shown as the number of C. difficile cells (CFU/ml) within the feces of infected mice at each time point. All values are the geometric means ± SEM. Note that C. difficile numbers were also enumerated from feces collected from mock-infected mice; viable counts were zero, as anticipated, and so cannot be seen on the graph. (B) In vivo TcdA toxin assays performed on HT29 cells using luminal contents collected from the gastrointestinal tracts of mice (n = 3) infected with the wild-type strain [TcdA⁺B⁺ (WT)] (black bar), tcdB mutant 1 (TcdA⁺B⁻1) (blue bar), tcdA mutant 1 (TcdA⁻B⁺1) (pink bar), or the tcdAB double toxin gene mutant (TcdA⁻B⁻⁾ (grey bar) or mock infected (white bar). The mean values are shown together with the SEM. (C) In vivo TcdB toxin assays performed on Vero cells using luminal contents collected from the gastrointestinal tracts of mice (n = 3) infected with the wild-type strain [TcdA⁺B⁺ (WT)] (black bar), tcdB mutant 1 (TcdA⁺B⁻1) (blue bar), tcdA mutant 1 (TcdA⁻B⁺1) (pink bar), or the tcdAB double toxin gene mutant (TcdA⁻B⁻) (grey bar) or mock infected (white bar). The mean values are shown together with the SEM. (D) Colonization efficiencies of the wild-type strain [TcdA⁺B⁺ (WT)] (black circles), tcdB mutant 1 (TcdA⁺B⁻1) (blue squares), and tcdA mutant 1 (TcdA⁻B⁺1) (pink triangles) isogenic toxin mutants using the hamster model of CDI, shown as the time from infection to the isolation of C. difficile colonies from feces of infected animals using CCFA growth medium (9). (E) Competitive index (CI) values for the TcdA⁺ TcdB⁻ toxin gene mutant versus the TcdA⁻ TcdB⁺ toxin gene mutant at 24 and 48 h following infection using the Monash model of CDI. Individual CI values are represented by closed black circles, with the horizontal lines representing the geometric means. Download

Table S1 

Comparison of the sporulation frequencies of the wild-type strain and isogenic toxin gene mutants.

Table S2 

List of genes differentially upregulated and downregulated by infection with C. difficile wild type and toxin mutants, using the Monash and Sanger infection model.

Table S3 

Parameters used for histopathological scoring of colon and cecum tissues.