In Vivo Physiological and Transcriptional Profiling Reveals Host Responses to Clostridium difficile Toxin A and Toxin B

INTRODUCTION

The toxins TcdA and TcdB are two key virulence factors of Clostridium difficile, an intestinal, opportunistic pathogen responsible for more than 300,000 infections in the United States per year (2009 data), with several estimates of annual cost between $433 million and $8.2 billion (1–4). Clinical manifestations include leukocytosis and diarrhea. The importance of TcdA and TcdB is underlined by the facts that strains without either toxin colonize but do not cause disease and that intoxication causes manifestations similar to those of infection (5–7). TcdA and TcdB are similar in size, amino acid sequence, and enzymatic specificity yet exhibit different enzymatic activities and in vivo potencies (8–10). Furthermore, much remains unknown about common and divergent cellular pathways leading to toxin-mediated host responses (11, 12).

Determining the relative roles of TcdA and TcdB in pathogenesis has proven difficult in part because of variable findings within and between animal models as well as species-specific responses. Clinically, strains lacking TcdA are commonly isolated from infected patients, and no TcdA⁺/TcdB⁻ clinical strain has ever been reported (13). Toxin effects in the context of infection have typically been studied using animal models in which an antibiotic regimen and subsequent disruption of intestinal flora must precede infection with C. difficile (14, 15). By generating mutant strains, Lyras et al. found that TcdB but not TcdA was essential for hamster infection, yet Kuehne et al. found, in a similar hamster infection model, that either toxin was sufficient (5, 6). Investigating toxin effects more directly, multiple intoxication models have demonstrated TcdA to be enterotoxic, while TcdB caused little to no pathology (7, 10, 16). However, epithelial damage in human xenografts in mice is greater with TcdB than TcdA, suggesting that many differences in toxin effects are species specific (17). The ability of either toxin to bind, enter, and/or activate intestinal cells may also explain differential effects of TcdA and TcdB. The sequences differ most in the C-terminal binding domain. TcdB has been shown to be incapable of binding the brush border membranes of hamsters, although TcdB has been found to further damage bruised ceca, synergize with TcdA, and contribute to pathogenesis during infection (5, 16, 18). Multiple receptors for TcdA have been proposed or identified, yet the roles of these receptors in different organisms, animal models, and cell types are unclear (19–24). TcdB weakly binds various trisaccharides and oligosaccharides, yet no functional receptor for TcdB has been identified (25). It is also possible that differences in intracellular actions of TcdA versus TcdB are responsible for differences in the host response. Although a similar dose of TcdA or TcdB may result in different gross pathologies, it is unclear if entirely different pathways are activated or repressed or if the same overall functions are affected to different degrees. We previously analyzed the transcriptional response of a human, ileocecal, epithelial cell line (HCT8) to TcdA and TcdB and showed that the toxins induce very similar transcriptional signatures, yet the effects of TcdB occurred earlier (26). In addition, we found altered regulation of many genes involved in cell growth and division but no overwhelming expression of inflammatory markers or other genes associated with physiological changes in vivo. The in vivo effects of these toxins have not been investigated by measuring genome-wide responses, and many of the links between cellular responses and physiological changes remain unknown. Therefore, we used an in vivo system, intracecal injection of toxin into mice, and collected samples to characterize the genome-wide cellular responses and gross physiological effects of each toxin over a 16-h time course.

It has been difficult to tease apart the aspects of the host response to TcdA and TcdB because of the important interactions among the many tissues, cell types, and signals involved (27, 28). The intestinal epithelium, the initial barrier to these toxins, continuously interacts with surrounding cells throughout the development and resolution of disease. Therefore, we focused on the transcriptional response of epithelial-layer cells to toxin and other toxin-related effects. Given the importance of surrounding tissues and with recent evidence of systemic dissemination of toxins, we chose cecal injection of toxin, an open system, as opposed to closed ileal loop models or ex vivo systems that may restrict toxin to a limited area (29). Additionally, previous studies have focused on separate facets of the host response, typically with only one toxin per study (30–36). To address these deficits in the knowledge of this illness, we measured genome-wide expression from epithelial-layer cells exposed to TcdA and TcdB to simultaneously capture effects of each toxin.

Using this approach, we identify several genes differentially expressed after toxin treatment that serve as specific candidates to investigate further. Additionally, we employ currently available bioinformatic methods and also introduce novel methods to identify groups of regulated genes associated with known biological functions. Our measurements were taken on several biological levels to link changes in one set of variables (e.g., gene expression) to changes in others (e.g., pathology and blood counts). These linkages serve as tools to validate previous findings as well as identify novel functions affected by TcdA or TcdB. Of the many linkages that could be explored, we further investigated chemokine expression and the role of two chemokines in the response to TcdA cecal injection with respect to changes in pathology, neutrophil recruitment, and survival. In addition to the comparison between toxins and identification of differentially expressed genes, these associations and concepts serve as a basis for further probing the host response to these toxins in the context of C. difficile infection.

MATERIALS AND METHODS

RESULTS

DISCUSSION

This is the first study to characterize the genome-wide transcriptional response to TcdA and TcdB in vivo. Additionally, several other parallel measurements were assessed to quantify changes at the cellular and tissue levels. The overall dynamics of the host responses to TcdA include rapid changes in cecal pathology, neutrophil infiltration, and gene expression. Conversely, TcdB elicits a delayed transcriptional response and causes significantly less pathology yet still recruits neutrophils and induces histopathological changes by 16 h. The combined effects of TcdA and TcdB (20 μg/toxin) on histopathology at 2 h and the overall transcriptional response at 2 and 6 h are not additive. However, two of three mice injected with TcdA+B did not survive to our 6-h time point, so we do not rule out potential synergism at later times. For example, Hirota et al., using lower doses (5 μg/toxin), found that TcdA and TcdB may act synergistically 4 h after both toxins are introduced intrarectally (46). As for TcdB alone, several studies have found that TcdB does not damage hamster or mouse intestines, and TcdB does not bind to hamster brush border membranes (16, 18). However, Lyerly et al. showed that when the cecum was bruised before intragastric administration of TcdB, all mice became ill (16). Hence, it is possible that the experimental procedure of cecal injection allows or enhances the TcdB-induced pathology we observe at 16 h. In line with our findings that TcdB has pathological effects, Libby et al. found that 16 of 16 hamsters died within 36 h of cecal injection of 60 μg of TcdB and found that 35 μg of TcdA resulted in epithelial lesions, edema, and neutrophil infiltration (7). We were able to quantify separate aspects of these host responses over time, revealing the relative responses to TcdA and TcdB, individually altered genes and markers of intoxication, and regulated gene sets associated with pathways that function at the intracellular and extracellular levels.

This study builds on our previous analysis of the transcriptional response of a human ileocecal epithelial cell line (HCT8) to TcdA or TcdB (2, 6, and 24 h after toxin treatment). After mapping all orthologous genes, we found that the overall transcriptional response of HCT8 cells is poorly correlated with the responses of the cecal epithelial-layer cells in vivo (r² < 0.03 for all comparisons; see Fig. S9 in the supplemental material). Some of these dissimilarities presumably are due to the different experimental systems and mRNA sources (in vitro versus in vivo and human versus mouse). Many differences may also represent important responses primarily observed in an in vivo experimental system. For instance, transient expression of chemokines and increased expression of several other cytokines was not observed in HCT8 cells. Also, the altered expression of many metabolic genes did not occur in HCT8 cells. Conversely, cell cycle- and DNA damage-associated gene sets were not enriched in vivo as they were with HCT8 cells. Selecting for expression that is similar between the data sets, 10 genes are commonly upregulated (Rhob, Klf2, Klf6, Jun, Dusp1, Gdf15, Hspa1a, Dusp1, Bcl2l15, and Gpx2) and a few are commonly downregulated (Edn1, Alpi, and Bmp2) (see Fig. S10). In another high-throughput analysis of the host cell response to TcdA, Zeiser et al. analyzed the changes in the proteome of Caco2 cells 24 h after TcdA exposure (47). By mapping transcripts to proteins, we found that the proteomics data were poorly correlated with transcriptional changes in HCT8 cells and our in vivo data (r² < 0.01 for all comparisons). However, similar to our previous study and the current study, Zeiser et al. did note changes in the amount of many cell cycle-associated proteins and several proteins involved in lipid and cholesterol metabolism.

Aside from the quantitative and temporal differences of the physiological responses, many transcriptional similarities exist between the host response to TcdA and TcdB. For both toxins, the immediate transcriptional response, indicative of initial or acute toxin effects, is represented by altered expression of a small set of genes. By 6 h, the number of differentially expressed genes for both toxins increases ∼200-fold, coinciding with changes in pathology, including neutrophil infiltration. TcdA challenge leads to approximately 10-fold more differentially expressed genes at 2, 6, and 16 h; however, there is significant overlap in the genes affected by TcdA and TcdB (Fig. 2). Although TcdA-induced changes were greater in magnitude, correlation coefficients (which are scale invariant) between TcdA and TcdB demonstrate strong overall similarity in gene expression signatures (Fig. 2E). This difference in scale between the toxin responses may result from differences in molecular functions and/or the number, type, and sensitivity of cells affected. Which and how many cells are affected might also originate from the differential abilities of TcdA and TcdB to bind and enter intestinal cells. In line with our results showing that TcdB caused significantly less pathology than TcdA, Rolfe found little to no TcdB adsorption relative to TcdA on hamster brush border membranes (18). The location and transport of toxins within the gut, which is very poorly understood, may also partly explain the extent of pathology in intoxicated mice. After cecal injection of TcdA or TcdB, we did not observe any significant pathology in the colon. Hence, the effects of the toxins which we measured were restricted to the cecum. Nevertheless, other systemic effects emanating from local insult may contribute to overall pathology. Multiple animal studies have observed increased mucosal permeability after TcdA intoxication, and Steele et al. demonstrated systemic dissemination of both TcdA and TcdB during severe infection in mice and piglets (29). Despite the various explanations for the lesser effects we observe with TcdB, a change in transcription after injection of TcdB is distinct even at 2 and 6 h, when changes in histopathology and many other variables are not apparent. Furthermore, this transcriptional response is highly correlated with the response to TcdA. Hence, the transcriptional analysis reveals that overall intracellular responses of epithelial-layer cells to TcdA and TcdB are largely similar, although the magnitudes of the gross observed pathologies may differ.

Beyond piecemeal identification of genes with altered expression, the transcriptional data as a whole reflects the cellular responses to underlying molecular interactions. Our analyses identified the upregulation of several genes encoding Rho binding proteins and small GTPases that are known to be affected by TcdA and TcdB. We also identified strong upregulation of many interferon-inducible GTPases. These interferon-inducible GTPases have been implicated in several mechanisms of cell-autonomous immunity, such as inflammasome activation, recognition of pathogens in vacuoles, assembly of defense complexes, and autophagy (48). Although the transcription of interferon genes is unaltered in the epithelial-layer cells we isolated, the GTPase upregulation suggests the functional presence of interferons. Consistent with this, Ishida et al. found increased transcription of IFN-γ in whole tissue and increased production of IFN-γ by infiltrating neutrophils after injection of TcdA into ligated ileal loops of mice (33). In the same analysis, beta-tubulin binding and histone deacetylase activity were also enriched. Although the relevance of these associations may seem unclear at first, Nam et al. have recently shown that inhibition of histone deacetylase 6 blocks TcdA-induced tubulin acetylation and subsequent mucosal damage (49). Hence, our analysis corroborates previous data and suggests that our transcriptional data reflect other functions that are affected in parallel with or prior to transcription. For example, a novel set of toxin-induced genes identified in our analysis includes several metabolic genes. At 2 h, multiple genes associated with cholesterol and steroid synthesis (specifically the mevalonate pathway) are slightly upregulated, whereas at 6 and 16 h, genes associated with fatty acid metabolism and detoxification of xenobiotics are downregulated. The expression of several of these genes is known to be controlled directly or indirectly by nuclear receptors. In light of the finding that gp96 (a paralogue to heat shock protein 90 [Hsp90]) serves as a receptor for TcdA in vitro, these transcriptional changes may also result from Hsp90 interactions (e.g., Hsp90 proteins bind xenobiotic response elements and steroid hormone receptors) (24).

In addition to the intracellular effects described above, several genes previously associated with toxin-mediated pathophysiology are altered. For instance, our self-contained enrichment test identified the interleukin-1 (IL-1)-mediated signaling pathway as the most enriched gene set 2 h after TcdA injection. More specifically, TcdA increases proinflammatory Il1b expression by 80% at 2 h and then over 500% at 16 h; TcdB causes no such change. In addition to Il1b, concomitant increases in the IL-1 receptor antagonist, Il1rn, suggest a natural feedback mechanism. These findings are in line with a previous study demonstrating that recombinant IL1RN pretreatment attenuated TcdA+B-induced inflammation (50). Expression of Il33, an IL-1 family cytokine involved in mucosal signaling, also dramatically increases in response to TcdA (28- and 95-fold at 6 and 16 h, respectively). Similarly, in mice infected with C. difficile (VPI 10463), intestinal levels of IL-33 increase at the peak of infection (data not shown), suggesting that this cytokine is responsive in part to toxin effects. In another study, Hirota et al. measured cytokine levels in colonic tissue lysates 4 h after intrarectal instillation of TcdA+B. The data published by Hirota et al. are correlated with our measured cecal, epithelial-layer gene expression (r = 0.5 for TcdA+B at 6 h) and intracellular protein expression (r = 0.36 for TcdA at 6 h) (see Fig. S5 in the supplemental material). Six hours after cecal injection of TcdA, serum cytokine concentrations correlate even more strongly with Hirota et al.'s data (r = 0.68). Hence, our data from cecal cells and serum after cecal injection of TcdA and/or TcdB are in agreement with findings from colonic tissue after intrarectal instillation of TcdA+B. Additionally, our data show the expression changes of many other genes over a 16-h time course for each toxin individually.

Although the expression and/or release of chemokines and cytokines are known responses to C. difficile toxins, the full profile of the transcription of cytokines in response to toxins has not been investigated. Previously, the role of many chemokines and proinflammatory mediators in TcdA-induced enteritis has been studied individually with ileal loops or in vitro (27). For example, Morteau et al. found increased Ccl3 and Ccl5 transcription in whole tissue 1 h after TcdA injection and showed that Ccl3 knockout mice were less susceptible to TcdA (30). Castagliuolo et al. identified increased expression of Cxcl2 in rat epithelial cells after injection of TcdA into ligated ileal loops (34). Our results identify additional chemokines which are transiently expressed and provide insight into the response of the epithelial layer over a time course covering the development of toxin-mediated pathogenesis. Additionally, our data include the expression of all cytokines in response to TcdA and TcdB separately and in combination. We found that overall cytokine expression in response to TcdA and TcdB is correlated, yet TcdA-induced changes are more pronounced. However, TcdB does not increase the expression of several inflammation-associated chemokines as TcdA does. This difference between toxins is particularly interesting in light of the fact that TcdB cecal injection causes less tissue damage and inflammation than injection of TcdA. Our data also show that the chemokine expression in response to TcdA is transient, yet many other cytokines continue to be expressed even after chemokine expression returns to basal levels. The early chemokine expression may be involved in the acute toxin effects. However, since C. difficile infections typically last several days, the continued expression of several other cytokines is an unexplored and potentially interesting area of research.

In order to characterize the timing and the effects of the acute response after toxin exposure, we investigated the early effects of CXCL1 and CXCL2 in the same cecal injection system. The increased expression of chemokine genes involved in neutrophil recruitment (e.g., Cxcl1 and Cxcl2) was coincident with neutrophil infiltration. Local epithelial damage and neutrophil infiltration were not attenuated by neutralization of CXCL1 and CXCL2. Similar findings in pathology between neutralizing antibodies and isotype controls could be due to insufficient levels of antibody near the site of toxin injection or the method of antibody administration. In another neutralization study, Castagliuolo et al. injected anti-CXCL2 intravenously into fasted rats 15 min prior to injection of TcdA into ileal loops; this neutralization attenuated TcdA-induced fluid secretion, mucosal permeability, and MPO activity (34). Other ileal loop experiments with TcdA have revealed that pathology and neutrophil infiltration become evident 1 to 3 h after intoxication (33, 34). Hence, it is unlikely that the effects we measured in this study would have been discernible prior to our earliest 2-h time point. On the other hand, few studies have examined the host responses after the initial acute response; most have focused on responses within the first 6 h (30–33, 35, 36, 51). Our latest time point, 16 h, did reveal attenuation of the TcdA-induced increase in systemic neutrophils for anti-CXCL1- and anti-CXCL2-treated animals. Moreover, serum levels of CXCL1 and CXCL2 were predictors for survival. This result also suggests that surviving mice are poor responders in terms of Cxcl1 and Cxcl2 expression and production. Related to this, a recent study by El Feghaly et al. showed that inflammatory markers in stool, not the number of C. difficile CFU, correlated with clinical outcomes (52). In a hamster infection model, Steele et al. showed that serum levels of CXCL1, IL-6, tumor necrosis factor α, and IL-1β are increased in cases of severe, systemic infection (29). Our results demonstrating that high CXCL1 and CXCL2 correlate with mortality are supportive of these infection studies and suggest that the release of these and perhaps other inflammatory markers in serum is primarily toxin mediated. Thus, our experiments with anti-CXCL1 and anti-CXCL2 emphasize the importance of the locale of toxin effects and chemokine expression systemically and/or throughout the intestine.

The multilevel measurements and our analysis have revealed important aspects of the relative host responses to TcdA versus TcdB intoxication, novel changes in transcription underlying observed physiological changes, and many aspects of the early time course and dynamics of the host response near the site of injection and in circulating blood. These extensive data and experimental framework also provide a basis for comparisons and future investigations with mutant toxins and mutant mouse strains. Furthermore, these data may be used to identify diagnostic markers or novel targets to attenuate host responses to TcdA and TcdB.

Supplementary Material