Clostridium difficile Toxins May Augment Bacterial Penetration of Intestinal Epithelium FREE

ANNUALLY IN the United States, an estimated 10% of hospitalized patients develop nosocomial infections involving more than 2 million people (with 58,000 deaths), and costing more than $4.5 billion.¹ Many nosocomial infections have an undefined focus and seem to be caused by translocating enteric bacteria.^{2 - 4} These infections, which include gram-negative bacteremia (primarily Escherichia coli and other Enterobacteriaceae), are often life-threatening.^{1 ,5 - 6} Septicemia, often acquired in the nosocomial setting, is the 13th leading cause of death in the United States, and its incidence is increasing.⁷ Despite appropriate antimicrobial therapy, the mortality associated with gram-negative bacteremia is 20% to 50%.^{1 ,5}

Clostridium difficile is an anaerobic, gram-positive, spore-forming bacillus found in the normal intestinal flora of approximately 3% of healthy adults, 15% to 30% of hospitalized patients, and up to 50% of neonates.⁸ The incidence of C difficile–associated diarrhea is increasing⁹ and is highly correlated with the use of broad-spectrum antibiotics, particularly third-generation cephalosporins.¹⁰ C difficile produces tissue-degrading enzymes (proteases, collagenases, hyaluronidase, heparinase, and chondroitin-4 sulfatase) as well as 2 toxins, A and B, the widely recognized etiologic agents of antibiotic-associated diseases ranging from diarrhea to pseudomembranous colitis. These toxins, recovered from 90% to 100% of patients with C difficile–associated diarrhea, are known to alter the integrity of confluent cultured intestinal epithelial cells.^{11 - 12}

A growing literature documents that intestinal barrier function is compromised in the diverse clinical conditions associated with translocation of intestinal bacteria; these conditions include enteric bacterial overgrowth, surgery, burn wounds, and other trauma.^{13 - 16} It may not be coincidental that patients at highest risk for systemic infection caused by enteric flora are the same patient populations with a high incidence of C difficile intestinal overgrowth.¹⁷ In vivo, the pathologic effect of the C difficile toxins can include disruption of enterocyte brush border membranes followed by extensive mucosal damage, leukocytic infiltrate into the lamina propria, toxic megacolon, and intestinal perforation.⁸ Formation of pseudomembranes (consisting of fibrin, mucus, and leukocytes) can occur from 2 to 10 days after first use of antibiotics to 1 to 2 weeks after cessation of antibiotics. Resultant mortality may approach 10%.⁸

Following initial experiments with injecting C difficile toxins into rodent ligated intestinal loops, investigators concluded that toxin A alone was responsible for pathogenic activity because it causes a florid inflammatory response including hemorrhagic fluid secretion, tissue necrosis, and mucosal inflammation as well as increased intestinal permeability.¹⁸ Toxin B was termed a cytotoxin, because it caused a cytopathic effect in tissue culture while having no detectable activity in ligated intestinal loops.¹⁸ However, data from experiments using human colonic explants indicated that toxin B was more potent than A in causing mucosal damage, suggesting that both toxins play a role in pathogenesis of C difficile–associated enteric disease.¹⁹ Both toxins induce release of several proinflammatory monokines from cultured human monocytes.²⁰ Toxins A and B are similar in size (308 kd and 269 kd, respectively), with 63% amino acid homology. Both toxins contain 3 recognizable domains, namely the amino terminus containing the active site, the carboxy terminus with repeating units responsible for binding, and a hydrophobic domain important for translocating the toxin across the vesicular membrane into the cytoplasm after internalization.²¹ There is evidence that the intracellular mechanism of action is enzymatic,^{21 - 22} with both toxins acting as glucosyltransferases to glucosylate the small guanosine 5C-triphosphatases (GTPases) Rho, Rac, and Cdc42.^{23 - 25} Glucosylation inactivates these GTPases, which are intimately associated with maintenance of the cellular cytoskeleton. Inactivation leads to cytoskeletal alterations²⁶ and the resultant cytopathic effect characteristic of these toxins on cultured cells. Although much is known about the epidemiology and activity of C difficile toxins, the effects on bacteria-enterocyte interactions (adherence, internalization, and paracellular transmigration) are unstudied.

Because C difficile can be recovered from many high-risk hospitalized patients receiving broad-spectrum antibiotic therapy,¹⁷ because C difficile toxins A and B have been associated with increased intestinal permeability in vitro,^{11 - 12} and because there is growing evidence that increased intestinal permeability may be a common mechanism whereby enteric bacteria penetrate the intestinal epithelium, we formulated the following hypothesis. C difficile–induced alterations in the intestinal barrier facilitate microbial penetration of the intestinal epithelium, which in turn facilitates the translocation of intestinal bacteria, a likely initial event in the pathogenesis of systemic infection in many high-risk patients. Herein we report effects of C difficile toxins A and B on bacterial interactions with cultured enterocytes. The data indicate that intestinal colonization with C difficile may play a heretofore unrecognized role in facilitating bacterial transmigration across the intestinal epithelial barrier.

Salmonella typhimurium ATCC 14028 was obtained from the American Type Culture Collection, Rockville, Md. Proteus mirabilis M13 and E coli M21 are rodent isolates. These species of enteric bacteria have been used in several in vivo studies of oral infectivity in rodents²⁷ as well as in vitro studies of bacterial internalization by cultured enterocytes.²⁸

Clostridium difficile toxin B (0.24 mg/mL) and toxin A (1 mg/mL) were purified as described,^{29 - 30} maintained in sterile normal saline at 4°C, and diluted in tissue culture medium for use in experiments. Toxin activity was confirmed by observation of characteristic cytopathic effect in human foreskin fibroblasts.³⁰ After a 16-hour incubation, C difficile toxins A and B exhibited a cytopathic effect at 10 and 1 ng/mL, respectively.

Caco-2 cells, which are derived from a human colon cancer, have many features of differentiated enterocytes including apical microvilli, tight junctions, structural polarity, ion conductance, and permeability. Caco-2 cells are spontaneously differentiated and in vitro interactions have been shown to have in vivo relevance.^{31 - 33} Caco-2 cells, obtained from the American Type Culture Collection, were cultivated in the absence of antibiotics in 24-well plastic dishes (Falcon; Becton Dickinson Co, Lincoln Park, NJ), seeded at 2 × 10⁴ cells per well, and incubated at 37°C in 9.5% carbon dioxide for 15 to 18 days, at which time these enterocytes are considered polarized and differentiated.^{28 ,31} Viability was determined by the vital dyes trypan blue (0.36%) and propidium iodide (20 mg/mL).

To study the effect of C difficile toxins A and B on bacterial internalization, enterocytes were pretreated for 16 hours with varying toxin concentrations. Bacteria were then added and bacterial internalization was assayed as described,²⁸ with C difficile toxin present throughout the assay. Briefly, overnight tryptic soy broth (Difco Laboratories, Detroit, Mich) cultures of individual bacterial strains were washed twice and diluted in sterile isotonic sodium chloride solution, and 100 µL containing 10⁸ viable bacteria was added to each tissue culture well. Bacterial concentrations were determined by densitometry and confirmed by serial dilution, followed by viable plate counts on appropriate agar media. Bacteria were incubated with enterocytes for 1 hour at 37°C. Enterocytes were then washed 5 times with Hanks balanced salt solution (HBSS). Tissue culture medium containing 50 µg/mL of gentamicin sulfate was added to kill residual viable extracellular bacteria. After 2.5 hours, enterocytes were washed 5 times with balanced salt solution and lysed for 5 minutes with 1% Triton X-100 (Sigma-Aldrich Corp, St Louis, Mo). Viable intracellular bacteria were quantified following serial dilution and incubation on appropriate agar media.²⁸

Experiments were designed to test the ability of C difficile toxins to modulate bacterial transmigration between Caco-2 cells. Using 3.0-µm membrane filter inserts with a 0.3-cm²surface area (Becton Dickinson Co), Caco-2 enterocytes were seeded at 3 × 10³ per membrane and cultivated as described earlier. Caco-2 cultures had varying concentrations of toxin A or B added to the apical chamber 16 hours prior to the migration assay. Bacteria (10⁸ per well) were then apically introduced, and enterocytes were incubated for 1 hour. The basal chamber was then sampled, serially diluted, and incubated on appropriate agar media.²⁸

Transepithelial electrical resistance (TEER) was used to monitor changes (ohms per square centimeter) in epithelial cell culture integrity, presumably caused by loosening of the tight junctions; decreasing values reflect increasing epithelial permeability.³⁴ The TEER of Caco-2 enterocytes was studied with the Millicell Electrical Resistance System (Millipore Corp, Bedford, Mass). All electrical resistance readings were recorded after subtracting the average resistance of 2 membranes in the absence of enterocytes; ie, membranes equilibrated overnight in tissue culture medium. Because TEER values often vary among individual enterocyte cultures, the electrical resistance value was recorded for each membrane before and after experimental treatment, and the percentage decrease from baseline was calculated for each membrane.

Bacterial adherence to cultured enterocytes was assayed as previously described.³⁵ Briefly, 10⁸ bacteria (prepared as described above) were incubated with 10⁶ confluent Caco-2 cells. Washed enterocytes (with adherent bacteria) were lightly fixed with 0.01% paraformaldehyde, incubated with primary rabbit antibody (diluted in 5% goat serum), washed, incubated with horse radish peroxidase–conjugated goat anti-rabbit IgG (Organoteknika-Cappel, West Chester, Pa), washed, the blue color developed with tetramethylbenzidine-hydrogen peroxide color development reagent (Sigma-Aldrich Corp), the reaction stopped with sulfuric acid, and the color intensity read at 450 nm in an enzyme-linked immunosorbent assay plate reader.³⁵ Resulting data are optical density units, a measure of relative adherence. This method does not determine absolute numbers of adherent bacteria, but the goal is to compare differences in bacterial adherence between untreated and toxin-treated enterocytes. Wright-Giemsa stains were also used to qualitatively assess relative bacterial adherence.

Statistical analyses were performed using StatView 4.5 (Abacus Concepts Inc, Berkeley, Calif). Experiments were repeated at least twice on separate days to verify reproducibility. On a given day, for bacterial internalization, assays were performed in triplicate tissue culture wells and the average was considered 1 assay value. For bacterial migration and adherence assays, each well was considered an assay value. Bacterial numbers were log₁₀ transformed and analyzed by analysis of variance followed by the Fisher test for significant difference. Dose responses were analyzed by correlation coefficient (r). Statistical significance was P<.05.

Phalloidin is a phallotoxin that preferentially binds actin filaments as opposed to actin monomers.^{36 - 37} The distribution of filamentous actin in the epithelial cell cytoskeleton was observed using the method of Howard and Meyer,³⁸ with minor modifications. Epithelial cultures were incubated at 37°C for 30 minutes with 0.8-µmol/L fluorescein-labeled phalloidin (Sigma-Aldrich) suspended in 5% buffered formalin containing 0.1 mg of lysophosphatidylcholine (Sigma-Aldrich) per milliliter, then washed 3 times with HBSS, mounted in phosphate-buffered saline and glycerin (1:9) containing 0.1% p-phenylenendiamine (Sigma-Aldrich) at pH 8, and viewed by epifluorescence microscopy.

High-resolution, low-voltage scanning electron microscopy was used to observe the effect of C difficile toxins A and B on confluent Caco-2 cells. Bacterial surface interactions with toxin-treated enterocytes were also observed. Caco-2 cells were cultivated in wells containing a 12-mm-diameter glass coverslip. Caco-2 cells were pretreated with either toxin A or B, then incubated with or without bacteria as described earlier for bacterial internalization. Enterocytes were then washed, fixed, and processed as described.³⁵ Fixed samples were dehydrated in ethanol, critical-point dried with carbon dioxide, and sputter-coated with a 1-nm discontinuous layer of platinum. A modified YAG crystal scintillator was used for backscatter electron imaging at 3 or 4 kV in a Hitachi S-900 (Hitachi, Tokyo, Japan) field emission scanning electron microscope.

The effects of varying concentrations of C difficile toxins A or B on Caco-2 internalization of S typhimurium, P mirabilis, and E coli are presented in Figure 1. For internalization experiments, the maximum toxin A concentration was 10 ng/mL rather than 100 ng/mL, because the higher concentration resulted in enterocyte sloughing, likely facilitated by the multiple washes required by this protocol. Neither toxin had a noticeable effect on internalization of any of the bacterial species tested (Figure 1).

Both toxin A and toxin B caused a dose-dependent decrease in TEER (r=0.9 and 0.8, respectively) (Figure 2), indicating an opening of enterocyte tight junctions. Correspondingly, toxins A and B were also associated with dose-dependent increases in bacterial transmigration through mature Caco-2 enterocyte cultures (Figure 3).

To determine if increased bacterial transmigration (Figure 3) was related to increased bacterial adherence, enterocytes were pretreated with either 100 ng/mL of toxin B or 10 ng/mL of toxin A and bacterial adherence was assayed by enzyme-linked immunosorbent assay. Both toxins were associated with increased bacterial adherence and statistical significance was achieved with P mirabilis and E coli (P<.05) (Figure 4). Wright-Giemsa stains of adherent bacteria (viewed by light microscopy) also confirmed that both toxins were associated with increased bacterial adherence.

Pretreatment of Caco-2 cultures with 0, 1, 10, or 100 ng/mL of either toxin A or B had no apparent effect on viability, and enterocyte cultures were consistently 95% viable or higher. Wright-Giemsa staining showed noticeable separation of individual enterocytes, an effect that increased with increasing toxin concentrations (not shown). Ultrastructural observations of untreated (control) Caco-2 cultures showed confluent enterocytes with tightly apposed borders (Figure 5, A). Although ultrastructural alterations appeared more pronounced with toxin A, both toxins caused distortions in apical microvilli, enterocyte rounding, and individual cells separated from each other (Figure 5, B). Bacteria appeared preferentially adherent to distorted apical microvilli (Figure 5, C and D). Following incubation of enterocytes with toxin A or toxin B, staining with fluorescein-labeled phalloidin revealed alterations in distribution of filamentous actin (Figure 6), and these cytoskeletal alterations appeared more prominent with toxin A than toxin B.

Although exact mechanisms whereby C difficile toxins A and B damage intestinal epithelium are not completely understood, there is evidence that both toxins have direct and indirect effects in vivo. Direct effects, manifested as cytoskeletal damage and increased paracellular permeability, likely result from toxin binding, internalization, and intracellular enzymatic activity.^{21 - 26} Directly, toxins A and B have been shown to damage human colonic explants¹⁹ as well as cultured enterocytes.^{11 - 12} Indirectly, studies involving ligated ileal loops in experimental animals have revealed that toxin A can activate lamina propria neuroimmune cells³⁹ and cause infiltration of the submucosa by neutrophils.⁸ Chaves-Olarte et al⁴⁰ recently suggested that the initial event in C difficile pathophysiology may be toxin A–induced activation of the enteric nervous system. This hypothesis assumes that toxin A binding to the intestinal mucosa results in signal transduction, resulting in secretion of proteins that signal neuroimmune cells to become effectors of inflammatory damage. Intracellular enzyme-mediated (glucosyltransferase) cytotoxicity then becomes important as a secondary mediator of damage.^{23 - 24}

Whether by a direct or indirect effect, both toxins increase intestinal permeability in explanted colonic tissue¹⁹ and in cultured enterocytes,^{11 - 12} and toxin A increases permeability in ligated intestinal loops.⁸ Hecht et al^{11 - 12} have shown that toxins A and B can increase permeability of cultured intestinal epithelium (T84 enterocytes) in the absence of inflammatory cells or their products. Using dual mannitol/sodium flux studies, this group also suggested that the toxin-induced permeability defect may be mediated by tight junctional disruption.¹² Defects in this barrier potentially enhance transepithelial movement of peptides or macromolecules, including bacterial cell wall fragments and toxins that are known to induce intestinal inflamation.⁴¹

Multiple mediators of inflammation, including vasoactive amines and kinins, are known to increase the permeability of vascular endothelium directly, or indirectly by recruiting neutrophils and monocytes, the products of which can contribute to disruption of tight junctions.⁴² Similarly, in clinical conditions associated with increased intestinal permeability (eg, shock, trauma, and sepsis), intercellular tight junctions are a likely target for injury. We have speculated that in these high-risk patients, intestinal bacterial toxins may affect tight junctional integrity, facilitating bacterial translocation.²⁸

We have previously shown that exposure of the lateral membranes of cultured enterocytes (using low extracellular calcium,⁴³ Bacteroides fragilis enterotoxin,²⁸ or cytochalasin⁴⁴ ) promotes preferential bacterial adherence to the exposed lateral enterocyte membranes, and also augments bacterial internalization. These observations have led us to speculate that, in compromised patients, translocation of enteroinvasive bacteria may occur via a similar mechanism. However, in the present study, toxin A– or toxin B–induced exposure of the enterocyte lateral membrane was associated with increased bacterial adherence but not increased internalization, although bacterial paracellular (between enterocytes) migration was augmented in a dose-dependent fashion.

Our working hypothesis links C difficile intestinal overgrowth with compromised intestinal epithelial barrier function and with bacterial penetration of the intestinal epithelium. Herein we report that C difficile toxins A and B caused confluent Caco-2 enterocytes to pull apart, resulting in decreased TEER and increased transmigration of enteric bacteria. This cytopathic effect was accompanied by disruptions in enterocyte filamentous actin. Others have also noted similar toxin-induced actin disruptions in cultured enterocytes, namely T84 cells.^{11 - 12} However, our study is the first to correlate these effects of C difficile toxins with bacteria-enterocyte interactions. If the present in vitro results can be extrapolated to the in vivo situation, the data presented herein suggest that C difficile toxins could act in vivo to compromise the intestinal epithelial barrier and facilitate bacterial penetration into extraintestinal tissues.

Presented at the 19th Annual Meeting of the Surgical Infection Society, Seattle, Wash, April 29, 1999.

This work was supported in part by Public Health Service grant AI23484 from the National Institutes of Health, Bethesda, Md.

Surgical patients are considered high risk for both intestinal colonization with C difficile and for systemic infection caused by normal enteric flora. Data from this in vitro study (with enteric bacteria and cultured enterocytes) indicate that the intestinal toxins produced by C difficile may facilitate bacterial penetration of the intestinal epithelial barrier. Thus, intestinal colonization with C difficile may contribute to a heretofore unrecognized increased risk of septic complications in this patient population.

Reprints: Brad Allen Feltis, MD, Department of Surgery, University of Minnesota, Box 195 UMHC, 420 Delaware St SE, Minneapolis, MN 55455 (e-mail: felti001@tc.umn.edu).

Donald Fry, MD, Albuquerque, NM: Thank you very much for the opportunity to discuss this well-presented paper. I find an analogy in this proposed mechanism of increased gut permeability as being quite analogous to what one sees with the increased vascular permeability of the changes that endothelial cells undergo; specifically, endothelial cells undergo cytoskeletal changes and some degree of globular formation as vascular permeability is increased by any of a number of different specific chemical signals. So the changes of the enterocyte that permit between-cell or paracellular movement of bacteria has a lot more appeal to me than the traditional Dr Wells' explanation of transcellular movement of bacteria. So I find this intellectually a little more satisfying as well.

I guess the question that comes up at this point is about the remodeling of the cytoskeleton. Is the change of permeability a specific response to the toxin, or is this a nonspecific response of the cell to any of a host of different noxious stimuli? You have studied 2 very specific toxins. I would like to beg the question of whether you have looked at other potentially noxious stimuli to your study model here? Is it specific to the toxin or is it potentially a nonspecific response?

One of the other interesting things that you actually allude to in your manuscript is that, in addition to direct effects of the toxin, there are potentially indirect effects. The release of histamine by mucosa-associated mast cells is probably one of the best-studied in vivo models of changes in gut permeability. So I am curious as to whether you have looked at agents such as histamine in terms of whether it potentially changes the cytoskeletal structure of the enterocyte as yet another potential indirect mechanism that might speak to cytoskeletal changes of the enterocyte cells?

I guess my last question is whether these cytoskeletal changes are permanent. Is the cell that has been attacked by the specific toxin irreversibly changed into the globular configuration, or can the cell really recover its normal native morphology and function?

One of the things that is evident to me is that this particular research effort is opening up another Pandora's box of a large number of signals that may be responsible for changing the cytoskeletal structure of the enterocyte. I somehow suspect this will increase indirect cost recovery at the University of Minnesota over the next few years.

Dr Feltis: In answer to your first question about whether or not C difficile toxigenic activity is a specific or nonspecific response, although toxin interactions with enterocytes have not been rigorously studied, some data are available. Previously, we have studied the effects of Bacteroides fragilis enterotoxin on enteroctyes and we noted morphologic changes similar to what we have observed with C difficile toxins, notably, rounding up, and pulling apart of adjacent cells. This argues that these effects may be seen with other intraluminal toxins as well.

Your second question concerns indirect effects of C difficile toxins. As you point out, a similar phenomenon of increased permeability due to tight junctional disruption is observed in vascular endothelium in response to inflammatory cytokines. One current hypothesis on the mechanism of C difficile toxin activity describes the initial event as toxin A–induced activation on the enteric nervous system. This hypothesis assumes that toxin A binding to the intestinal mucosa results in signal transduction, followed by secretion of proteins that signal neuroimmune cells to become effectors of inflammatory damage. A secondary effect of the toxin may be the direct effect on disruption of the actin cytoskeleton. Therefore, similar to inflammatory changes in vascular endothelium, there may be a direct effect on the enterocytes by cytokines from neuroimmune inflammatory cells.

To answer your last question, we do not yet know if the changes in the cell cytoskeleton are reversible. We do know that in cultured fibroblasts, if you remove the toxin stimulus, the cells return to their normal morphologic state after a few hours.

Ed Deitch, MD, Newark, NJ: I would like to compliment the authors on a very nice study. As you know, we have been working with this model of Caco cells since about 1990 and have had a lot of fun with this model. Now, one of the problems with studying the gut in a jar is that you miss the entire unstirred layer. In vivo there is an unstirred layer on top of the enterocytes that's made up of mucous and a number of other factors that prevents many of the things that might injure the enterocyte from reaching it. So, clearly, the gut is wearing a bulletproof vest under normal circumstances and what you have done in this study is make it naked.

So my question is, have you done any studies to look and see whether the same sort of injury would occur if you add mucous back? We have done that and it can be done relatively easily to see if the gut is protected from these toxins by the addition of low levels of virus.

Second, what happens if you put a mixed flora in or try to recreate the competitive environment that resides in the gut? There are a lot of good things that bacteria do as well as bad.

You have made a very important observation. You have dissected out in a reductionist mode a very interesting phenomenon. Have you put the pieces back together to see how it fits when we have a whole picture?

Dr Feltis: The short answer is that we aren't sure what effect an unstirred mucous layer might have on bacteria-enterocyte interactions. Preliminary results from similar experiments in our lab with mucous-secreting HT-29 enterocytes indicate that in vitro the presence of mucous does not alter the morphologic effects of C difficile toxin.

Henri Ford, MD, Pittsburgh, Pa: Very interesting study. We have been also interested in trying to figure out the preferred route of bacterial migration across the epithelium. When we take the isolated piece of colonic mucosa and mount it in the Ussing chamber, we are able to see that the mere presence of bacteria is able to induce similar changes with regards to the TEER. It is clearly associated with a paracellular defect; however, we are not able to demonstrate any increase in bacterial migration across that mucosa, even though when you look at the mannitol flux, it is clearly increased. So clearly, there must be some divergent properties here.

I was wondering if, number one, you looked at mannitol flux within your system and is there any evidence besides what you have surmised that this is indeed a paracellular defect that is allowing the bacteria to get across? Second, have you been able to demonstrate that by either transmission electron microscopy or by scanning electron microscopy that the bacteria are indeed going through the defect that you presume is there?

Dr Feltis: No, we have not performed ion flux studies in our system. We relied on electrical resistance measurements to represent disruption of the enterocyte tight junctions caused by C difficile toxins. Interestingly, in regard to colonic explant models, if the primary activity of the toxins is induction of the gut neuroimmune system via transcellular signaling, having an intact myenteric plexus might be important to elicit the full effect. This is another reason that an in vivo model is essential for further study.