Binding of Clostridium difficile to Caco-2 epithelial cell line and to extracellular matrix proteins


*Corresponding author. Tel.: +39 (6) 49902343; Fax: +39 (6) 49387112, E-mail address:


Adhesion of Clostridium difficile to Caco-2 was examined as a function of monolayers polarization and differentiation. The number of adherent C. difficile C253 bacteria per cell strongly decreased when postconfluent 15-day-old monolayers were used (1.7 bacteria per cell versus 17.3 with 3-day-old monolayers). Following disruption of intercellular junctions by ethylene glycol-bis(β-aminoethyl ether)-N,N,N′,N′,-tetraacetic acid, a significant rise in the level of bacterial adhesion was observed, above all in postconfluent monolayers. Immunofluorescence studies of bacteria and transferrin receptor, a marker of basolateral pole of polarized monolayers, showed that C. difficile C253 adheres mainly to the basolateral surface of differentiated and undifferentiated polarized Caco-2 cells. Furthermore, binding of C. difficile C253 to several extracellular matrix proteins in vitro was demonstrated by an ELISA-based assay.


Clostridium difficile is the aetiological agent of antibiotic-associated pseudomembranous colitis (PMC) and of most cases of colitis or diarrhea (AAD) in patients undergoing antibiotic therapy [1]. Although the role of toxins A and B is known [2], the exact sequence of pathological events leading to disease is not well understood. In particular, the establishment of C. difficile on human intestinal mucosa has yet to be elucidated. To this aim, we studied the adhesion of C. difficile to epithelial cells in vitro by using the human colonic epithelial cell line Caco-2. This cell line closely resembles small intestinal epithelial cells and has played a major role in studies on mechanisms of adherence and invasion of many pathogenic bacteria, including Salmonella typhimurium, Shigella flexneri, Listeria monocytogenes, Pseudomonas aeruginosa and enterotoxigenic Escherichia coli[3–7]. Grown in vitro under standard culture conditions, Caco-2 cells spontaneously exhibit signs of structural and functional differentiation and polarization [8,9]. Whereas polarization is characterized by the presence of both tight junctions and cell polarity, differentiation corresponds to a maturation process resulting in formation of functionally differentiated brush-border microvilli. The differentiation of Caco-2 cells is total at late confluency (14–21 days), but the polarization is established earlier and, after confluency, involves the whole monolayer. Nonconfluent Caco-2 monolayers (3–4 days old) consist mainly of large islands containing several cells, in which polarization and differentiation start from the central cells of the island. In polarized monolayers, the basolateral surface of Caco-2 cells may be artificially exposed by disruption of intercellular junctions after treatment with Ca2+ chelators [10,11].

In the present study, the adhesion of C. difficile to Caco-2 cells was examined as function of monolayer polarization and differentiation. The Caco-2 cell surface area involved in the adhesion process was explored by using the transferrin receptor as a marker of the basolateral area of polarized monolayers [12]. Furthermore, the ability of C. difficile to bind in vitro to isolated extracellular matrix (ECM) proteins was investigated.

2Materials and methods

2.1Bacterial strains and growth conditions

C. difficile strain C253, a serogroup C, toxigenic clinical isolate obtained from a patient with AAD [13] was used in this study. When indicated, the toxigenic C. difficile strain Cd 79685, isolated from a patient with PMC and belonging to serogoup S3, and C. difficile ATCC 43597 (both provided by P. Bourlioux, University of Paris Sud, Châtenay-Malabry, France) were also used. The latter strain was not toxigenic and it belonged to serogroup D. Bacteria were grown anaerobically in brain heart infusion broth (Oxoid S.p.A., Milan, Italy), added with 0,5% yeast extract (Oxoid) and 0.1% hemin (Sigma, St. Louis, MO, USA), for 18 h at 35°C.

2.2Cell culture

The human colon carcinoma cell line Caco-2 (ATCC HTB37) was used between passages 48 and 65. Cells were routinely grown in minimum essential medium with Earle's salts (MEM) (Gibco BRL, Life Technologies, Inc., MD, USA) supplemented with 10% inactivated fetal bovine serum (FBS) (Gibco) and 1% MEM nonessential amino acids (Gibco) in a 5% CO2 atmosphere at 37°C. Confluency occurred after 5–6 days of incubation. For the adhesion assay, propagated cells were seeded at 5×104 per well onto 12-mm-diameter glass cover slips in a 24-well tissue culture plate (Falcon, Becton Dickinson Company, Paramus, NJ, USA). The culture medium was changed every other day. Monolayers were used 3 (nonconfluent monolayers) and 15 days (postconfluent monolayers) after seeding. The state of polarization and morphological differentiation was monitored by scanning electron microscopy (SEM).

2.3Disruption of intercellular junctions

Caco-2 monolayers were washed twice with Ca2+-free MEM (S-MEM) (Gibco), with 2% FBS added with 0.1 mM ethylene glycol-bis(β-aminoethyl ether)-N,N,N′,N′,-tetraacetic acid (EGTA) (Roche Diagnostic GmbH, Mannheim, Germany), and treated with the same medium for 1 h prior to infection in a 5% CO2 atmosphere at 37°C. Untreated monolayers were in parallel washed and incubated with MEM–2% FBS without EGTA.

2.4Adhesion assay

Bacterial cells were harvested by centrifugation at 5800×g for 20 min, washed once with phosphate-buffered saline (PBS) and diluted 1:10 (1×108 bacteria ml−1) in MEM–2% FBS, for the adhesion assay with the untreated monolayers, or in S-MEM–2% FBS with 0.1 mM EGTA, for the assay with the monolayers previously treated with EGTA. 1 ml of these bacterial suspensions was added to each tissue-culture plate well previously treated or not with EGTA. After 1.5 h of incubation in anaerobic chamber (Don Whitley Scientific Limited, Shipley, UK), cells were washed three times with PBS and finally incubated with 1% Triton X-100 (M-Medical, Florence, Italy) for 5 min at 37°C. The counts of adherent bacteria released from the cells were determined by plating appropriate dilutions on Columbia blood agar (Oxoid). To express the results as number of adherent bacteria per cell, Caco-2 cells from noninfected monolayers were collected by trypsinization and enumerated in parallel. Control wells, where C. difficile cells in MEM–2% FBS with or without EGTA were added in absence of any Caco-2 cells, were also included. Each experiment was carried out in duplicate and repeated three times. Infected monolayers were fixed and stained with Gram and May-Grünwald's–Giemsa stains for microscopic observation.


Infected monolayers were fixed with 2.5% glutaraldehyde in 0.1 M Millonig's phosphate buffer (MPB) at 4°C for 1 h. After washing in MPB, samples were post-fixed with 1% OsO4 in the same buffer for 1 h at 4°C and dehydrated in ascending acetone concentrations. The specimens were critical-point dried using liquid CO2 and sputter-coated with gold before examination on a Stereoscan 240 scanning electron microscope (Cambridge Instruments, Cambridge, UK).

2.6Double fluorescence labeling of bacteria and transferrin-receptor observed by confocal laser scanning microscopy (CLSM)

Infected monolayers were washed twice in PBS and fixed with 4% paraformaldehyde in PBS for 10 min at room temperature. Bacteria were labeled using a 1:10 dilution of antiserum against whole cells of C. difficile C253, prepared as previously described [13], and revealed by a 1:300 dilution of goat anti-rabbit rhodamine-conjugated (TRITC) immunoglobulin G (Sigma). The receptor for transferrin was stained with a 1:20 dilution of monoclonal antibody to human transferrin receptor (Roche Diagnostic GmbH) and revealed with a 1:30 dilution of goat anti-mouse fluorescein-conjugated (FITC) immunoglobulin G (Sigma).

The labeled preparations were observed in a confocal fluorescent imaging system using a CLSM LEICA TCS 4D (Leica Instruments, Heidelberg, Germany) supplemented with an argon/krypton laser and equipped with 40×1.00–0.5 and 100×1.3–0.6 oil immersion lenses. Confocal sections were generally taken at intervals of 1 μm. The excitation and emission wavelengths employed were 488 and 510 nm for FITC-labeled transferrin, and 568 and 590 nm for TRITC-labeled bacteria, respectively. The acquisitions were recorded, employing pseudo-color representation.

2.7Binding of C. difficile C253 to immobilized matrix proteins

Microtiter plates (Greiner, Frickenhausen, Germany) were coated with ECM proteins (fibronectin, laminin, types I, III, IV, V collagen, vitronectin and fibrinogen) (Sigma) by incubating 100 μl of each protein (diluted in 0.25 M NaHCO3, 0.25 M Na2CO3 buffer, pH 9.6) at concentrations ranging from 0.62 to 80 μg ml−1 in assay wells overnight at 4°C. After coating, wells were washed three times with PBS containing 0.05% Tween 20 (PBST), blocked with 300 μl of 5% skim milk in PBST for 2 h at 37°C, and washed three times with PBST. C. difficile C253 cells were pelleted by centrifugation, washed with PBS, and adjusted to a specific turbidity (OD600=0.750) in PBST. A 100-μl suspension was added to each well and incubated for 2 h at 37°C in anaerobic chamber. After removing unbound bacteria by washing wells five times with PBST, C. difficile cells were detected by subsequent incubation with a 1:1000 dilution of rabbit antiserum against whole C. difficile C253 cells [13], with alkaline-phosphatase-conjugated goat anti-rabbit IgG secondary antibody (Sigma) diluted 1:30,000 and with p-nitrophenylphosphate (Sigma). Absorbance at a wavelength of 405 nm was measured in an ELISA plate reader (Bio-Rad Laboratories, Hercules, CA, USA). Results are expressed as average of three independent assays, carried out in duplicate. Bovine serum albumin (BSA) (Sigma) and the highly glycosylated fetuin (Sigma) were used as controls for nonspecific binding. Control wells coated with the different ECM proteins without adding C. difficile cells were also tested to rule out antibody reactivity with ECM proteins.


3.1Adhesion of C. difficile to Caco-2 cells

The ability of C. difficile C253 to adhere to Caco-2 monolayers grown for 3 and 15 days was examined. The mean number of adherent C. difficile C253 bacteria per cell was 17.3 (standard deviation 5.50) using 3-day-old monolayers and 1.7 (standard deviation 0.10) utilizing 15-day-old monolayers (Table 1). By microscopic observation, 3-day-old monolayers were nonconfluent and consisted of islets containing few or several cells. Rare individual cells and small islets appeared totally surrounded by adherent bacteria (data not shown). Interestingly, in large islets, bacteria were located mainly at the outer edge (Fig. 1) and less inside the islets. In the postconfluent monolayers grown for 15 days, a very small number of bacteria interacted with the cells (data not shown).

Table 1.  Effect of disruption of intercellular junctions on adhesion of C. difficile to Caco-2 cells
  1. a3- and 15-day-old monolayers untreated or previously treated with S-MEM plus EGTA were infected with 1×108 CFU of C. difficile. After 1.30 h of incubation, adherent bacteria were determined on agar plates.

StrainAdherent bacteria to 3-day-old monolayersaAdherent bacteria to 15-day-old monolayersa
 No pretreatmentPretreatment with S-MEM+EGTANo pretreatmentPretreatment with S-MEM+EGTA
ATCC 4359724.7±8.6028.0±6.551.8±0.2316.0±2.64**
Values are mean numbers of adherent bacteria per cell ± standard deviations of three separate experiments. Statistical analysis between values for EGTA-pretreated cells versus those for untreated cells was performed by Student's t-test. *P<0.05, **P<0.01
Figure 1.

Caco-2 cells infected with C. difficile C253. Light micrograph of a Gram and Giemsa stained 3-day-old monolayer showing bacteria closely interacting with the periphery of a large islet. Bar: 20 μm.

To make the basolateral cell surface accessible to bacteria, the intercellular junctions of the Caco-2 cells were disrupted by Ca2+ depletion in the extracellular medium. Treatment of monolayers with S-MEM added with EGTA caused a two-fold rise in the adhesion level of strain C253 to 3-day-old monolayers and a 20-fold increase in that to 15-day-old monolayers (Table 1). As shown in the same table, C. difficile strains 79685 and ATCC 43597 exhibited an adherence pattern similar to that of C253, although the increase was less pronounced. For all the C. difficile strains examined, very few bacteria were visible in control wells without Caco-2 cells, indicating that bacterial binding to glass cover slips was almost absent.


SEM observations of infected 3 day-old monolayers showed Caco-2 cells flat, with moderate microvillous structures and some cell-to-cell junctions. Adhesion of C. difficile C253 to nonconfluent monolayers was not homogeneous, in agreement with observations by conventional microscopy. Bacteria were mainly located along the outer edge of cell islets (data not shown), while some bound randomly to the Caco-2 cells surface in the middle areas of the islets (Fig. 2a). Bacteria directly interacting with the ECM fibers, particularly in EGTA-treated samples, could be observed (Fig. 2e). Observation at higher magnification confirmed that the visible structures were ECM fibers and not cell extensions (data not shown). Examination of 15-day-old monolayers showed Caco-2 cells displaying typical morphology of differentiated state, with increased basolateral polarity and appearance of protruding villous structures at the apical surface (brush border). In infected 15-day-old monolayers, most of the adhering bacteria could be seen along the intercellular junctions in gaps between cells (Fig. 2b), while few bacteria (0–2 per cell) appeared to bind to the apical dome. After opening of intercellular junctions by EGTA, an evident increase of bacteria in close contact with the basolateral surface of Caco-2 cells was observed (Fig. 2d). Clusters of bacteria were also associated with the ECM fibers (Fig. 2f). Finally, those bacteria which were not apparently connected with fibers (Fig. 2c–e) were very likely interacting with residual ECM generated by fibers destroyed during the experimental procedures.

Figure 2.

Caco-2 cells infected with C. difficile C253. SEM micrographs of 3- (a,c,e) and 15-day-old monolayers (b,d,f), untreated (a,b) or pretreated with EGTA (c–f). Following disruption of intercellular junctions by EGTA, bacteria are shown to interact with the Caco-2 cells’ basolateral pole (c,d). Bacteria directly interacting with the ECM fibers are visible in panels e and f (arrows); inset in panel e shows a detail at higher magnification. Clusters of bacteria, which are not apparently associated with fibers, are probably interacting with residual ECM (c–e). Bars: a–f=10μm, inset=3μm.

3.3Immunofluorescence study and confocal microscopy

Observations of untreated 3-day-old monolayers showed that fluorescent signals of transferrin receptor and bacteria colocalized on the peripheral cells of the islets composing the nonconfluent monolayers (Fig. 3a). Untreated 15-day-old monolayers showed a very weak signal of transferrin receptor along the intercellular junctions only on few cells (Fig. 3b) and the rare presence of labeled bacteria mainly in the upper focal planes. When postconfluent monolayers were treated with EGTA, a dramatic increase in the transferrin receptor-derived fluorescent signal was observed (Fig. 3c,d) as well as in the number of fluorescent bacteria, which were mainly located in the mid-focal planes of the cells, immediately before the planes showing the transferrin receptor's fluorescent signal (Fig. 3e1–6).

Figure 3.

Confocal imaging micrographs of sites of C. difficile C253 adhesion to Caco-2 monolayers and cell distribution of the transferrin receptor. a,b: Double immunofluorescence labeling of bacteria (red) and transferrin receptor (green) in untreated 3- (a) and 15-day-old monolayers (b). The yellow color corresponds to colocalization of the two fluorescence signals; adhesion sites of bacteria and transferrin receptor colocalize on the peripheral cells of the islet in the 3-day-old-monolayer, no colocalization is detectable in the 15-day-old-monolayer. c,d: Immunofluorescent labeling of transferrin receptor in untreated (c) and EGTA-treated (d) 15-day-old monolayers; the transferrin receptor, located at the basolateral cell surface, is made accessible by EGTA treatment. e1–e6: Double immunofluorescent labeling of bacteria (red) and transferrin receptor (green) in EGTA treated 15-day-old monolayers. Confocal section series starts at the mid-focal plane of the monolayer and ends at the bottom. The top image (e1) corresponds to the first detectable fluorescent signal (mid-focal plane) and the other images (e2–e6) follow consecutively. The yellow color corresponds to colocalization of the two fluorescence signals; adhesion sites of bacteria are located just above the distribution of the transferrin receptor. Bars: 25 μm.

3.4Binding of C. difficile C253 to specific ECM proteins

As shown in Fig. 4, C. difficile C253 bound efficiently to fibronectin and collagen types I, III, IV and V in a dose-dependent manner, while binding to fibrinogen and vitronectin was uniformly high, regardless of the amount of protein used. Adhesion to collagen types IV and V was higher than that to the other collagen types. Binding to laminin was as low as that to BSA and fetuin. No antibody reactivity with the different ECM proteins was detected in the control wells.

Figure 4.

C. difficile C253 binding to immobilized ECM proteins. Microtiter plate wells were coated with increasing concentrations of fibrinogen, vitronectin, fibronectin and laminin (A), collagen types I, III, IV and V (B) and incubated with C. difficile cells. After washing, bacterial binding to ECM proteins was spectrophotometrically quantified by an ELISA, using rabbit antiserum against C. difficile C253 whole cells. BSA and fetuin were used as controls for nonspecific binding. Data shown are means of three independent assays, carried out in duplicate. Error bars indicate standard deviations.


Results obtained in this study indicated that the adhesion of C. difficile to Caco-2 cells was a function of monolayer age. In postconfluent 15-day-old monolayers the level of adhesion fell 10-fold compared to nonconfluent 3-day-old monolayers. Moreover, microscopic observations of nonconfluent monolayers revealed that bacteria adhered mainly to the outer edge of peripheral cells of the islets. This pattern of adherence resembled that of some enteroinvasive bacteria that enter through the basolateral surface of differentiated and nondifferentiated polarized Caco-2 cells [4,5].

To investigate the hypothesis that C. difficile binds mainly to the basolateral surface of Caco-2 cells, monolayers were pretreated with EGTA to open intercellular junctions. These treatments appeared more effective with postconfluent than with nonconfluent monolayers, supporting our original assumption. In fact, the basolateral pole is normally exposed at the outer edge of the islands in nonconfluent monolayers, whereas it is made accessible by EGTA treatment in postconfluent monolayers. Furthermore, SEM observations of postconfluent monolayers clearly showed direct adhesion of bacteria to cells in intercellular gaps of untreated monolayers and to the basolateral cell surface of those treated with EGTA.

In our in vitro model of adhesion, the three C. difficile strains tested showed a similar adherence pattern, irrespective of their toxigenic status. As discussed below, toxins may however play a role in adhesion in vivo.

Since the expression of the transferrin receptor can be considered a specific marker of the basolateral area of polarized Caco-2 cells [12], the immunofluorescence studies confirmed the basolateral adhesion of C. difficile on both nonconfluent and postconfluent monolayers. Actually, results on differentiated and polarized postconfluent monolayers suggested that an adhesion site for C. difficile was located in the mid-focal planes of Caco-2 cells, just above the planes in which transferrin receptor was detected. Recently, some pathogens have been found to use as a receptor the zonula adherence protein E-cadherin, which is distributed on the lateral membrane just below the tight junctions [14,15]. It might be supposed that C. difficile also interacts with that cell–cell adhesion molecule, but further studies are necessary to support this hypothesis.

Other authors have reported that C. difficile interacts with apical microvilli of differentiated Caco-2 cells, even if the number of adhering bacteria per cell was quite low [16]. In this study, SEM and confocal observations of intact postconfluent monolayers showed that the prevalent adhesion site of C. difficile was located along the intercellular junctions in gaps between cells. However, some bacteria were found to bind also to the apical dome and this latter interaction may be crucial during the first steps of host infection in vivo.

Since SEM observations showed several bacteria interacting directly with the ECM fibers, an ELISA-based assay was used to assess the binding of C253 strain to immobilized matrix components. Results indicated that C. difficile is able to specifically bind to several ECM proteins in vitro, suggesting that it might use them as attachment sites in vivo. In recent years bacterial pathogens have been shown to adhere to components of both the basement membranes and the interstitial tissue of eukaryotic cells [17–20]. Since in our study several ECM components seem involved, C. difficile might possess numerous ECM-binding adhesins or, like Yersinia enterocolitica, a multifunctional adhesin with different domains for different ECM proteins [21].

In conclusion, the present study shows that C. difficile preferentially uses the basolateral surface of differentiated and undifferentiated polarized Caco-2 cells as adhesion site and it specifically binds to several ECM proteins in vitro. Although there is no clear evidence, our data seem to suggest the involvement of two distinct adhesion mechanisms: (i) a direct binding to a yet unknown cell receptor located on the lateral surface of Caco-2 cells, and (ii) an interaction with some components of ECM fibers. Further studies in this respect are necessary and the recent identification in C. difficile of a gene cluster including genes encoding surface proteins with adhesive properties [22] supplies new prospects.

By extrapolating these results to the in vivo setting, a two-step process might be supposed. C. difficile might initially interact with the apical microvilli of the intestinal epithelial cells and begin to release toxins A and B. It has been shown that toxins A and B disrupt epithelial barrier function [23], probably acting through the Rho proteins on apical actin structure and on tight junctions organization [23,24]. The basolateral pole of epithelial cells becomes thus accessible and a larger number of bacteria might interact with a receptor therein located and with the ECM proteins. Previous observations in the hamster model that adhesion of avirulent nontoxigenic strains is facilitated by co-administration of toxin A [25] seems in agreement with this sequence of events.


We are grateful to Tonino Sofia for editorial assistance.