Ethics Statement

All animal experiments were approved by the Laboratory Animal Care and Use Committee of the Institute of Microbiology v.v.i., Academy of Sciences of the Czech Republic, approval ID: 244/2009.

Gliadin fragments

Peptic fragments of gliadin (Sigma, St Louis, MO) were prepared on a pepsin agarose gel (ICN, Biomedicals, Ohio) as described [27], [28]. Protein concentrations were measured by bicinchoninic acid assay (BCA Protein assay, Pierce, Rockford, IL). All reagents were tested by E-toxate test for lipopolysaccharide (LPS) (Sigma, St. Louis, MO) and were below the limit of detection (2 pg/ml).

Cell culture and stimulation

The rat normal small intestine epithelial cell line (IEC)-6 was purchased from American Type Culture Collection (ATCC, CRL-1592, Manassas, VA) and was grown in 45% Dulbecco's Modified Eagle Medium (DMEM) Glutamax and 45% RPMI 1640, containing 10% inactivated (56°C for 30 min) fetal bovine serum (FBS) (Gibco, Barcelona, Spain), 100 µg/ml streptomycin, 100 U/ml penicillin, and 0.1 U/ml insulin (Sigma, St. Louis, MO) at 37°C and 5% CO₂. Cells were stimulated with bacteria and gliadin fragments (100 µg).

Bacterial strains and culture conditions

The following strains were used: Bifidobacterium bifidum IATA-ES2 (CECT 7365), Shigella CBD8, and Escherichia coli CBL2. B. bifidum IATA-ES2 was isolated from the feces of healthy babies and identified as described previously [29], [30]. E. coli CBL2 and Shigella CBD8 were isolated from celiac patients and identified as described by Sanchez et al. [26]. The Shigella strain was included to exemplify the possible effects of an actual intestinal pathogen in this disease context.

Bifidobacteria were grown routinely in de Man, Rogosa, and Sharpe (MRS) broth (Scharlau Chemie SA, Barcelona, Spain) with 0.05% (w/v) cysteine and incubated at 37°C under anaerobic conditions (AnaeroGen; Oxoid, Basingstoke, UK) for 22 h. Enterobacteria were grown routinely in Violet Red Bile Dextrose (VRBD) agar (Scharlau Chemie SA, Barcelona, Spain) at 37°C for 24 h under aerobic conditions.

Cells were harvested by centrifugation (6000× g for 15 min) at the stationary growth phase, washed 2 times with PBS, and resuspended in PBS that contained 20% glycerol. Aliquots of these suspensions were frozen in liquid nitrogen and stored at −80°C until use. The number of live cells after storage was determined as colony-forming units (CFUs) on MRS-C, Schadler, or VRBD agar after 48 h incubation under optimal conditions. For all strains, more than 90% cells were alive on thawing, and no significant differences were observed during storage (4 months). One fresh aliquot was thawed for each new experiment to avoid variabilities in live bacterial cell numbers between experiments.

Bacterial adhesion assay

Rat epithelial cells (IEC-6 line) were grown in 24-well plates in DMEM to confluence; the monolayers were washed twice with PBS, and 250 µl of labeled bacterial cell suspension (at an absorbance of 0.5 (10⁶ CFU/ml) at 600 nm) was added to each well.

Bacterial staining was performed with 10 mM 5-CFDA (5-carboxyfluorescein diacetate) (Sigma, St. Louis, MO) as described by Izquierdo et al. [29]. Briefly, labeled bacterial suspensions were added to IEC-6 cultures at A₆₀₀ 0.50. The epithelial cells and labeled bacteria were incubated together at 37°C for 1 h. IEC-6 cells were washed 2 times with PBS to remove nonadherent bacteria, and adherent cells were lysed in 200 µL 1% SDS (Sigma, St. Louis, MO) in 0.1 M NaOH at 37°C for 1 h [29].

Supernatants were collected in Costar black round-bottom 96-well plates (Corning Inc., Corning, NY, USA), and the fluorescence was measured on a microplate fluorometer (Fluoroskan Ascent, Labsystem, Oy, Finland) with excitation and emission wavelengths of 485 nm and 538 nm, respectively. Adhesion was expressed as the percentage of fluorescence that was recovered from adherent bacteria, relative to the initial fluorescence of the bacterial suspension per well.

Experimental animals

Wistar-AVN germ-free (GF) rats were reared in Trexler-type plastic isolators under controlled sterile conditions. Granulated gluten-free diet 02 (maize 57%, soya meal 30%, sunflower oil 1.5%, linseed oil 1.5%, salt, DL-lysine 0.5%, DL-methionine 1%, mineral, and vitamin mixture) was sterilized regularly by irradiation (59 kGy, Bioster, Czech Republic) [31], [32].

Rat intestinal loops

The bacterial strains and gluten were tested in ligated ileal loops of GF rats. Two-month-old GF inbred AVN rats (approximately 200 grams) were deprived of food for the 24 h before surgery (with free access to water). The rats were premedicated intramuscularly with 1 ml of a mixture of ketamine (10 mg/ml) and xylazine (2 mg/ml).

The three ligated loops (each approximately 2 cm long) were created with nylon ligatures in the jejunum and proximal ileum, beginning approximately 3 cm from the ileocecal junction. Each loop was followed by a short intervening segment (2 cm) that was not inoculated [32]. Five hundred microliters of inoculum, containing 10⁶ CFU of bacteria alone or with gliadin (250 µg) and/or IFN-γ (250 U, AbD Serotec), was injected into the intestinal loops. After inoculation, the jejunum was returned to the abdomen, and the laparotomy incision was closed. After 8–9 h, the rats were euthanized by severing of the carotid artery. Tissue samples and contents of the loops were collected for further analysis.

Immunohistology

Tissue from the loop was fixed immediately in 10% neutral buffered formalin or Carnoy's solution. The fixed tissues were cut and processed using routine methods. Paraffin sections (5 µm) were deparaffinized in xylene, rehydrated through an ethanol gradient to water, and stained in periodic acid-Schiff (PAS) to evaluate mucin-secreting goblet cells. The villi (10–15) in these sections were examined by light microscopy to determine the number of PAS-positive goblet cells per 100 enterocytes in the intestinal tissue, expressed as the medians and quartiles from 5–10 independent measurements.

Gliadin was detected in the intestinal loops by immunolocalization. Briefly, snap-frozen intestinal loop samples, embedded in OCT (Tissue-tek, Sakura Fine Tek, Torrance, CA, USA), were cryosectioned at 6 µm, air-dried, fixed for 5 min in acetone, and stored at −20°C. The sections were washed and endogenous peroxidase blocked by 1% H₂O₂. Then, the sections were incubated with peroxidase-labeled monoclonal anti-gliadin antibodies (Elisa Development Prague, Czech Republic) overnight at 4°C, washed, and incubated with Tyramide Signal Amplification - TSA™ Plus Fluorescence system (PerkinElmer, USA) for 30 min. The samples were counterstained with Evans blue and Hoechst to visualise tissue cells and nuclei. Afterwards the sections were embedded in Vectashield mounting medium (Vector Laboratories, UK). All speciemens were examined using a confocal microscope Olympus FV 1000 SIM.

ZO-1 expression was measured by incubating the sections with rabbit polyclonal anti-ZO-1 or anti-claudin-1 antibodies (Zymed Laboratories Inc., San Francisco, CA) at 4°C overnight, washing them, and incubating them with goat anti-rabbit IgG-FITC (in 10% PBS- normal goat serum, Zymed Laboratories Inc.) for 2 h.

Control sections were treated similarly, except that they were incubated with secondary antibodies only. Images of the specimens were viewed under an Olympus BX 40 microscope that was equipped with an Olympus DP 70 digital camera.

Western blot of tissue lysates

Intestinal tissue from the loops was homogenized on ice in protein extract buffer (Pierce, Rockford, IL) with a protease inhibitor cocktail (Pierce) for 10 min and sonicated. Samples were centrifuged at 10,000× rpm for 10 min at 4°C and stored at −80°C until use. Protein concentrations were measured using the BCA Protein Assay Kit (Pierce).

Proteins were denatured with sample buffer (106 mmol/L Tris-HCl, 141 mmol/L Tris base pH 8.5, 0.51 mmol/L EDTA, 10% glycerol, 2% SDS, 0.22 mmol/L SERVA blue G250, 0.175 mmol/L phenol red, 0.1 mmol/L 2-mercaptoethanol) for 5 min at 100°C, separated on a 10% (for claudin-1) or gradient 5% to 20% (for ZO-1) polyacrylamide gel and blotted onto 0.2-mm PVDF membranes (Serva, Germany).

The membranes were blocked with 2% (w/v) dry milk in 0.05% PBS-Tween-20 for 1 h at room temperature and incubated overnight at 4°C with antibodies against claudin-1 (1∶1000), ZO-1 (1∶1000) (ZYMED Laboratories Inc.), and β-actin (1∶5000) (Abcam, Cambridge, MA, USA). After incubation with the respective primary antibodies, the membranes were washed 3 times for 5 min in 0.05% PBS-Tween-20 and exposed to species-specific horseradish peroxidase-labeled secondary antibodies (1∶1000) (ZYMED Laboratories) for 1 h at room temperature. The reactions were developed using the ECL Plus Western blotting reagent (Pierce), and the signal intensities were measured on the LAS-1000 luminescence detector (Fujifilm, Tokyo, Japan) and processed with AIDA 1000/1D Image Analyzer software, version 3,28 (Raytest Isotopenmmessgeraete GmbH, Straubenhardt, Germany). After stripping, all membranes were re-probed with antibodies against β-actin (1: 5000, Abcam), to document the same protein concentration in all samples.

Scanning electron microscopy

Terminal ileum tissue was fixed in 3% glutaraldehyde in 0.1 M cacodylate buffer (pH 7.2), dehydrated in an ethanol series (50, 70, 80, 90, 96, 100, and 100% v/v), transferred to 100% (v/v) acetone, and dried in a BALZERS CPD 010 critical-point dryer (Balzers, Lichtenstein). The samples were then sputter-coated with gold and examined on an Aquasem electron microscope (Tescan, Czech Republic) in the SEM mode.

Rat cytokine array

The cytokine spectra in the rat intestinal loop washes were measured using the semiquantitative RayBio™ Rat Cytokine Antibody Array 1 (RayBiotech, Norcross, GA, USA), which detects 19 growth factors, cytokines, and chemokines, following the manufacturer's recommendations. The signal intensity was measured on an LAS-1000 luminescence detector (Fujifilm), and the resulting images were analyzed using AIDA software (version 3.28; Raytest) to quantify spot densities. The background staining was subtracted, and the data were normalized as described [33].

Statistical analysis

Statistical analysis was performed using SPSS, version 17.0 (SPSS Inc., Chicago, IL, USA). To establish the homogeneity of variances and the distribution of the data, the Levene test was run. As a result of the non-normal distribution of the data and the nonhomogeneity of the variances, Mann-Whitney U-test was used to assess the effect of each variable. The data were expressed as medians and quartiles. Different letters (a–e) mean statistically significant differences between stimuli, the identical letters correspond to non-significant differences. P<0.05 was considered statistically significant.

1. Goblet cell population in the jejunum is influenced by gliadin and intestinal bacteria

The effect of gliadin and the proinflammatory cytokine IFN-γ on epithelial cells in the presence or absence of various bacterial strains was examined in vivo using loops of small intestine that were ligated surgically from rats kept on a gluten-free diet and reared under germ-free (GF) conditions. As shown in Figure 1, the various stimuli led to changes in the number of PAS-positive goblet cells (examples A–F). To evaluate these changes, the number of PAS-positive goblet cells per 100 epithelial cells was counted (as summarized in Figure 1G). The addition of gliadin into the loops decreased the number of PAS-positive goblet cells compared with PBS used as a control. A similar effect was observed after applying IFN-γ alone and with gliadin (Figure 1B,G). The number of goblet cells after combination of gliadin with E. coli CBL2 (Figure 1E,G) or Shigella CBD8 (Figure 1C,G) was even lower. The addition of IFN-γ to above mentioned samples slightly increased the number of goblet cells (Figure 1D,F,G).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. Effects of bacterial strains and gliadin on goblet cells.

Histological staining of PAS-positive goblet cells in rat intestinal loops exposed to: B. bifidum IATA-ES2+gliadin (200 µg) (A), IFN-γ (225 U) (B), Shigella CBD8+gliadin (C), Shigella CBD8+gliadin +IFN-γ (D), E. coli CBL2+gliadin (E) and E. coli CBL2+gliadin+IFN-γ (F). Bacteria were applied at 10⁶/loop. Changes in goblet cells are expressed as medians and interquartile ranges (25% to 75%) of the number of PAS-positive goblet cells/100 epithelial cells (G). These values were for B. bifidum IATA-ES2 (39, 35–41), E. coli CBL2 (38, 35–41) and Shigella CBD8 (25, 20–27) when applied alone to the loops. Different letters (a–e) indicate statistically significant differences between medians as calculated by Mann-Whitney U test (P<0.05). Identical letters correspond to non-significant differences. The separate dots or asterisks indicate outliers. The pictures were obtained for the specimens viewed under an Olympus BX 40. Scale bar, 50 µm.

https://doi.org/10.1371/journal.pone.0016169.g001

When gliadin was combined with B. bifidum IATA-ES2 (Figure 1A), the number of PAS-positive goblet cells increased, attaining the same value as in PBS-treated loops (Figure 1G). Moreover, the combination of B. bifidum IATA-ES2 with Shigella CBD8, gliadin, and IFN-γ increased the PAS-positive goblet cell population. The effect of B. bifidum IATA-ES2 was less evident when loops were exposed to E. coli CBL2 (Figure 1G).

By scanning electron microscopy, the addition of B. bifidum IATA-ES2 did not affect mucin secretion and did not evoke any changes in intestinal loop architecture (Figure 2A). Mucin secretion was slightly higher after the addition of gliadin (data not shown). IFN-γ, however, induced mucin release (Figure 2B), and higher effect was observed when gliadin and E. coli CBL2 were injected with IFN-γ (Figure 2D). The combination of Shigella CBD8, gliadin, and IFN-γ boosted mucin secretion into the lumen and impacted the architecture of the epithelial layer, as shown in Figure 2F. Interestingly, addition of B. bifidum IATA-ES2 to ”harmful agents” (gliadin and IFN-γ with/without E. coli) slightly decreased the mucin secretion as compared to those agents alone (Figure 2C,E).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. Mucin production by goblet cells in rat intestinal loops.

Mucin production after application of B. bifidum IATA-ES2 (A), IFN-γ (B), gliadin+IFN-γ+ B. bifidum IATA-ES2 (C), E. coli CBL2+gliadin+IFN-γ D), E. coli BL2+gliadin+ IFN-γ+B. bifidum IATA-ES2 (E) and Shigella CBD8+gliadin+IFN-γ (F). Bacteria were applied at 10⁶/loop. The samples were coated with gold and examined by Aquasem electron microscopy (Tescan, Czech Republic) in the SEM mode. Scale bar, 20 µm.

https://doi.org/10.1371/journal.pone.0016169.g002

2. Translocation of gliadin into intestinal villi is influenced by intestinal bacteria

We determined whether intestinal epithelial layer permeability and gliadin peptide translocation was affected by bacterial strains. Using mouse anti-gliadin antibody, we monitored the transfer of gliadin peptides through the epithelial layer after exposure of the intestinal loops to IFN-γ and bacterial strains.

Gliadin, when applied with B. bifidum IATA-ES2 and IFN-γ, was observed only in low amounts inside the lamina propria - forming foci mainly on the apical section of certain villi (Figure 3A). In contrast, the combination of E. coli CBL2, gliadin and IFN-γ induced small, local changes (crypt widening), and gliadin was detected primarily below the epithelial layer (Figure 3B). The combination of Shigella CBD8, gliadin, and IFN-γ increased gliadin translocation, and gliadin was detected primarily inside the lamina propria (Figure 3C). Furthermore, using differential interference contrast (Figure 3D–F corresponding to samples in upper row), we confirmed the decrease in a number of goblet cells in loops treated with E. coli CBL2 and their loss after treatment with Shigella CBD8, documented in Figure 1.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 3. Changes in intestinal permeability induced by gliadin and various bacterial strains.

Intestinal loops were exposed to: gliadin fragments with IFN-γ and B. bifidum IATA-ES2 (A,D), gliadin+IFN-γ+E. coli CBL2 (B,E) and Shigella CBD8+gliadin+IFN-γ (C,F); The white arrows indicate gliadin fragments found by immunofluorescence (A–C) using mouse peroxidase-labeled monoclonal anti-gliadin antibody and TSA™ Plus Fluorescence systems and black arrows indicate goblet cells (D, E). Bacteria were applied at 10⁶/loop. The specimens were viewed under confocal microscope Olympus FV 1000 SIM using differential interference contrast (D–F). Scale bar, 20 µm.

https://doi.org/10.1371/journal.pone.0016169.g003

These data are also consistent with our fluorescence microscopy results, which demonstrated the distribution of TJ components, claudin-1, and ZO-1 in intestinal loops that were treated with gliadin, IFN-γ, and/or various bacterial strains (Figure 4A–J). Gliadin alone or with IFN-γ downregulated ZO-1 expression (Figure 4A,C) compared to PBS-exposed loops (Figure 4I). On the other hand, simultaneous addition of B. bifidum IATA-ES2 with IFN-γ and gliadin upregulated ZO-1 expression (Figure 4E). When the loops were simultaneously exposed to E. coli CBL2, gliadin and IFN-γ, ZO-1 fluorescence was reduced (Figure 4G).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 4. Distribution of claudin-1 and ZO-1 in rat intestinal loops.

Exposure of intestinal tissue to gliadin digest alone (A) or with IFN-γ (C) led to reduced ZO-1 expression at the periphery of the villi. The combination of gliadin+IFN-γ+B. bifidum IATA-ES2 (E) maintained the original level of ZO-1 as in PBS-treated loops (I). When gliadin+IFN-γ were applied with E. coli CBL2 (G) ZO-1 fluorescence was weaker. No changes in claudin-1 localization (B, D, F, H) were detected after any stimulus in comparison with PBS control (J). Representative pictures of three experiments are shown. The specimens were viewed under an Olympus BX 40 microscope. Scale bar, 20 µm. Western blot of tissue lysates (K) from intestinal loops stimulated with: 1. gliadin, 2. gliadin+IFN-γ, 3. B. bifidum IATA-ES2+gliadin, 4. E. coli CBL2+gliadin +IFN-γ, 5. E. coli CBL2+B. bifidum IATA-ES2+gliadin+IFN-γ, 6. B. bifidum IATA-ES2+gliadin+IFN-γ, and 7. PBS. The separated proteins on membranes were stained with anti ZO-1 or claudin-1 antibodies and re-probed with antibodies against β-actin to document the same protein concentration in all samples.

https://doi.org/10.1371/journal.pone.0016169.g004

In contrast, the typical pattern of claudin-1 expression at the periphery of intercellular (enterocyte) contacts was unaffected by addition of gliadin alone or with IFN-γ, B. bifidum IATA-ES2 or E. coli CBL2 (Figure 4B,D,F,H) compared to PBS-treated loops (Figure 4J). Nevertheless, the combination of gliadin, IFN-γ, and Shigella CBD8 nearly extinguished ZO-1 and claudin-1 signals (data not shown).

To support the fluorescence microscopy findings, intestinal tissue from the stimulated loops was extracted, and changes in TJ proteins were measured by western blot. As shown in Figure 4K, ZO-1 expression was more sensitive to various stimuli than claudin-1. Gliadin, IFN-γ, and, particularly, their combination with E. coli CBL2 reduced ZO-1 levels in tissues. The addition of B. bifidum IATA-ES2 to this mixture increased ZO-1 levels, confirming the fluorescence microscopy data. When B. bifidum IATA-ES2 was added with gliadin, ZO-1 levels approximated to those of the PBS control. When Shigella CBD8 was used, the fragmentation of TJ proteins was detected (data not shown).

3. Interaction of bacteria with the epithelial layer in vitro

The different effects of bacterial strains on gliadin translocation and expression might be a consequence of differences in the adhesion properties of individual bacterial strains that determine host-microbe interactions.

The interaction of various bacterial strains from celiac patients or healthy subjects (which comprise potentially beneficial and pathogenic bacteria) with epithelial cells was analyzed in vitro using the IEC-6 rat cell line; the adherence of bacteria to IEC-6 cells and the impact of gliadin were measured. As shown in Figure 5, the percentage of adhered bacteria varied only slightly, and the differences between E. coli CBL2, Shigella CBD8 and B. bifidum IATA-ES2, were not statistically significant. The simultaneous addition of gliadin fragments and bacteria to cell cultures had an insignificant effect on bacterial adhesion.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 5. Adhesion of different bacterial strains to IEC-6 cells.

The highest percentage of adhered bacteria was observed for E. coli CBL2 and Shigella CBD8. The differences between tested bacterial strains, as well as the effect of simultaneously added gliadin fragments were non-significant as established by applying the Mann-Whitney U-test. Data are expressed as medians and interquartile ranges (25% to 75%) of adhesion of four independent experiments. None of the differences was found to be statistically significant (P<0.05). The separate dot indicates an outlier.

https://doi.org/10.1371/journal.pone.0016169.g005

4. Cytokine secretion into the gut lumen

Cytokine production in response to administration of food and bacterial antigens and IFN-γ to rat intestinal loops was measured in intestinal washes by cytokine array (Figure 6A–H). The secretion of cytokines, such as chemotactic factor for monocytes and neutrophils (MCP)-1, tissue inhibitor of metalloproteinase (TIMP)-1, vascular endothelial growth factor (VEGF), and beta-nerve growth factor β-(NGF), increased.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 6. Cytokine array analysis of rat intestinal loops washes.

Layout of the arrays (A), cytokine profiles from loops treated with PBS (control) (B), gliadin (C), gliadin+IFN-γ (D), B. bifidum IATA-ES2+gliadin+IFN-γ (E), E. coli CBL2+gliadin+IFN-γ (F), and E. coli CBL2+B. bifidum IATA-ES2+gliadin+IFN-γ (G). The data are expressed as relative levels of selected cytokines (percentage of positive controls). Cytokine-induced neutrophil chemoattractant (CINC)-2 and -3, monocyte chemoattractant protein (MCP)-1, macrophage inflammatory protein (MIP)-3α, nerve growth factor β-(NGF), tumor necrosis factor (TNF)-α, vascular endothelial growth factor (VEGF). The signal intensity was measured using the LAS-1000 luminescence detector (Fujifilm, Tokyo, Japan).

https://doi.org/10.1371/journal.pone.0016169.g006

The most abundant cytokines, MCP-1 and TIMP-1, which play a role in tissue protection, were induced by B. bifidum IATA-ES2 in a mixture of gliadin and IFN-γ. The addition of E. coli CBL2 to this mixture decreased MCP-1 and TIMP-1 release into the intestinal loops. VEGF secretion rose, particularly by the addition of E. coli CBL2 to gliadin and IFN-γ but was unaffected by simultaneous addition of B. bifidum IATA-ES2 to this mixture.

The spontaneous production of β-NGF was independent of any stimulus. Further, cytokine-induced neutrophil chemoattractant (CINC)-3, IFN-γ, IL-10, IL-1α, IL-1β, IL-6, macrophage inflammatory protein (MIP)-3α, and TNF-α levels were low. Although it was difficult to determine the effect of the stimuli on low cytokine production, CINC-3 was detected only in loops that were inoculated with E. coli CBL2. In PBS treated loops the cytokine IL-10, IL-1α, IL-1β, and TNF-α were undetectable (as summarized in Figure 6H).

When Shigella CBD8 replaced E. coli CBL2, cytokine levels increased markedly. Nevertheless, the high background of the microarrays, reflecting the impact of Shigella CBD8 on intestinal tissue, rendered the precise evaluation of these data impossible.