Caco-2 cells, obtained from INSERM (Paris, France), were subcultured at low density and used between passage 90 and 105, according to Natoli et al. (24). Cells were maintained at 37 °C in a 95% air/5% CO₂ atmosphere at 90% relative humidity in Dulbecco's modified Eagle's medium (DMEM) containing 25 mM glucose, 3.7 g/l NaHCO₃, 4 mM L-glutamine, 1% non-essential amino acids, 1 × 10⁵ U/l penicillin, 100 mg/l streptomycin, 10% heat inactivated fetal bovine serum (FBS, Euroclone, Milan, Italy). Cell culture media and reagents were purchased from Corning (Milan, Italy), unless otherwise stated. For experiments, cells were seeded in 24-well plates (Becton Dickinson, Milan, Italy) or polyethylene terephthalate permeable filters (Falcon™ 10.5 mm diameter, 0.4 μm pore size, Becton Dickinson, Milan, Italy), which allow separation between apical (AP) and basolateral (BL) compartments. After confluency, cells were left for 17–21 days to allow complete differentiation. Medium was changed 3 times a week.

The GOS product, synthesized by β-galactosidase through the lactose transgalactosylation reaction, was provided as a powder by Royal Agrifirm Group (START+; Royal Agrifirm Group, Drongen, Belgium). For all experiments, GOS powder was weighed, resuspended in experimental medium (bacterial or cell growth medium) at 2% final concentration, and filtered with 0.45 μm syringe filters (Merck Millipore, Darmstadt, Germany). The experimental design is outlined in Figure 1.

Caco-2 cells were differentiated on permeable filters (1 × 10⁶ cells/filter), as described above. Cell membrane permeability was assayed by measuring the transepithelial electrical resistance (TEER), according to Srinivasan et al. (25), with a Millicell Electrical Resistance system (Merck Millipore, Darmstadt, Germany). TEER values were checked before each experimental assay, to verify correct TJ formation and complete cell differentiation into mature enterocytes. TEER measurements were carried out at different time points, depending on the experiments, as detailed below. For each time point, the TEER values of treated samples were expressed as % of the TEER values of control (untreated) cells, set at 100%. Cell permeability was also measured at the end of treatments by measuring the phenol red passage, as reported by Monastra et al. (26). Briefly, after three washes with phosphate buffered saline containing Ca⁺⁺ and Mg⁺⁺ (PBS⁺⁺), 0.5 ml 1 mM phenol red was added to the AP compartment of cell monolayers, whereas 1 ml PBS⁺⁺ was added in the BL compartment. After 1 h incubation at 37 °C, 0.9 ml BL medium was collected, added with 0.1 ml 0.1 N NaOH, and read at 560 nm (Infinite M200 microplate reader, Tecan, Milan, Italy) to determine the phenol red concentration passed from AP to BL compartment. This concentration was used to calculate the phenol red apparent permeability coefficient (Papp) by applying the following formula: Papp = Ct × V_BL/Δt × ·C0 × A, where V_BL is the volume of the BL compartment (cm³), A is the filter area (cm²), Δt is the time interval (s), Ct is the phenol red concentration in the BL compartment at the end of time interval, and C0 is the phenol red concentration in the AP compartment at the beginning. The TJ were considered open and indicative of an absence of cell monolayer integrity when the phenol red Papp values were above 1 × 10⁻⁶ cm s⁻¹, as determined by previous reports in the literature (27). Differences observed among samples below this threshold value were thus considered biologically irrelevant, irrespective of statistical significance.

In order to assess the potential toxicity of GOS, Caco-2 cells, differentiated on permeable filters, were apically treated with 2% GOS. TEER was recorded for up to 72 h, and phenol red Papp was assayed at the end of treatment. Two independent experiments, carried out in triplicate, were performed.

The intestinal pathogen F4⁺ enterotoxigenic Escherichia coli strain (ETEC F4⁺, O149:F4ac⁺, LT⁺, STa⁺, STb⁺), provided by the Lombardy and Emilia Romagna Experimental Zootechnic Institute (Reggio Emilia, Italy) was grown in Luria-Bertani (LB) broth at 37 °C. The ETEC F4⁺ strain GIS26 (O149:F4ac⁺, LT⁺, STa⁺, STb⁺) was grown on brain heart infusion (BHI) agar plates for 16 h at 37 °C.

Lactobacillus amylovorus ATCC 33198 (LMG9434), Lactobacillus amylovorus DSM 16698 and Limosilactobacillus reuteri subsp. porcinus DSM 110571 were grown in De Man Rogosa Sharpe (MRS) medium (supplemented with 0.5 g/l L-cysteine for DSM 110571), for 24 h at 37 °C under microaerophilic conditions.

All media and supplements were provided by Oxoid (Thermo Fisher, Waltham, MA, USA), unless otherwise stated.

A liquid broth assay was performed to evaluate if ETEC F4⁺ growth could be affected by the presence of GOS in LB or antibiotic- and FBS-free DMEM media. Overnight cultures of ETEC F4⁺ were diluted 1:50 in the two media containing or not 2% GOS and then dispensed into 96-well plates at a final volume of 200 μl per well. Bacterial growth was monitored by recording the OD₆₀₀ of the bacterial cultures for 22 h at 1 h intervals using an automated plate reader (Infinite M200, Tecan, Milan, Italy). The OD₆₀₀ values were normalized with respect to the medium alone. One independent experiment was conducted in triplicate.

The adhesion and invasion assays were carried out on Caco-2 cells, differentiated in 24-well plates (1 × 10⁶ cells/well). Cells were placed in antibiotic- and FBS-free DMEM 16 h before the experiment. On the day of the assay, overnight bacterial cultures of ETEC F4⁺ were diluted 1:20 in LB broth and grown for approximately 3 h up to the exponential growth phase, as monitored by OD₆₀₀ reading. Appropriate amounts of bacterial cells were harvested by centrifugation at 5,000 rpm for 10 min, resuspended in antibiotic- and FBS-free cell culture medium, pre-incubated with GOS or DMEM for 10 min at 37 °C and then added to cell monolayers at a concentration of 1 × 10⁸ colony forming units (CFU)/well (approximately 100:1 bacteria to cell ratio). Co-cultures of bacteria and Caco-2 cells were incubated at 37 °C for 1.5 h. In the adhesion assay, non-adhering bacteria were removed by 5 washes with Hanks' Balanced Salt solution (HBSS: 137 mM NaCl, 5.36 mM KCl, 1.67 mM CaCl₂, 1 mM MgCl₂, 1.03 mM MgSO₄, 0.44 mM KH₂PO₄, 0.34 mM Na₂HPO₄, 5.6 mM glucose) and cell monolayers were lysed with 1% Triton-X-100, according to Roselli et al. (28).

In the invasion assay, after the 1.5 h infection period, co-cultures of bacteria and Caco-2 cells were washed twice with HBSS and then incubated with antibiotic- and FBS-free DMEM containing gentamicin sulfate (50 mg/L; Merck, Darmstadt, Germany) for additional 2.5 h, to kill residual viable extracellular bacteria. After gentamicin incubation, Caco-2 cells were washed and lysed as described above. For both adhesion and invasion assays, viable bacterial cells (adhered and invading, respectively) were quantified by plating appropriate serial dilutions of Caco-2 lysates on LB agar. CFU were counted on plates after overnight incubation at 37 °C.

For both assays, at least two independent experiments were performed in duplicate.

The villus adhesion assay was performed as described by Rasschaert et al. (29). Small intestinal villi were collected from the mid-jejunum of a healthy piglet (6 weeks old female, Belgian Landrace x Piétrain, weaned at 4 weeks), euthanized for an unrelated experimental trial, approved by the Ethical Committee of the Faculties of Veterinary Medicine and Bioscience Engineering at Ghent University (approval number EC2019/085). No additional animals were used or sacrificed for this study. The piglet was genotyped for the F4 receptor presence using the MUC13 SNP assay (30). Small intestinal villi were incubated with 4 × 10⁸ CFU of the ETEC F4⁺ strain GIS26 for 1 h at room temperature. To assess the ability of GOS to inhibit adhesion of ETEC F4⁺ to the small intestinal epithelium, the bacteria were pre-incubated with 0.5 mg/ml GOS for 1 h prior to performing the adhesion assay. As a positive control for inhibition of adhesion, the bacteria were incubated with a FaeGac-specific monoclonal antibody (mAb in house, clone IMM01). The adhesion of the bacteria to the villus epithelium was then evaluated by counting the number of bacteria adhering along a 50 μm villous brush border length at 5 randomly selected places, with a phase-contrast microscope at a 600 × magnification. The number of bacteria per 250 μm villous brush border length was then calculated. This procedure was performed three times, in a single experiment. The % inhibition was calculated by dividing the number of adherent bacteria in the presence of GOS or mAb by the number of adherent bacteria in the absence of GOS.

Caco-2 cells differentiated on permeable filters (1 × 10⁶ cells/filter) were apically treated with ETEC F4⁺ (1 × 10⁸ CFU/well), prepared as described above. Appropriate amounts of bacterial cells were centrifuged, resuspended in antibiotic- and FBS-free cell culture medium, pre-incubated with GOS or DMEM for 10 min at 37 °C and then apically added to cell monolayers. TEER was recorded every 30 min for up to 2.5 h, and phenol red Papp was assayed at the end of treatments. Three independent experiments, carried out at least in duplicate, were performed.

To assess the effect of GOS on ETEC-induced TJ protein delocalization, Caco-2 cells, differentiated on glass coverslips in 24-well plates (1 × 10⁶ cells/well), were treated with 2% GOS in the presence or absence of ETEC F4⁺ (1 × 10⁸ CFU/well), prepared as described above, for 1.5 h at 37 °C. At the end of the experiment, Caco-2 cells were washed with cold PBS⁺⁺, fixed in ice-cold methanol for 3 min, and then incubated with rabbit polyclonal anti-ZO-1 or mouse monoclonal anti-occludin (Zymed Laboratories, San Francisco, CA, USA) or rabbit polyclonal anti-P-p65 (Cell Signaling Technology, Danvers, MA, USA) for 1 h. For secondary detection, cells were incubated with tetramethylrhodamine isothiocyanate (TRITC) conjugated secondary antibodies (Jackson Immunoresearch, Milan, Italy), for 1 h. 4′,6-diamidino-2-phenylindole (DAPI) was used to label DNA in nuclei. Stained monolayers were mounted on glass slides using the Prolong Gold antifade reagent (Molecular Probes, Invitrogen, Milan, Italy) and analyzed using a confocal fluorescence microscope (LSM 700, Zeiss, Jena, Germany). Three independent assays were conducted.

To assess the effect of GOS on ETEC F4⁺-induced p65 phosphorylation, Caco-2 cells differentiated on permeable filters (1 × 10⁶ cells/filter) were apically treated with 2% GOS in the presence or absence of ETEC F4⁺ (1 × 10⁸ CFU/well), prepared as above described, for 1.5 h at 37 °C. At the end of treatments, cells were washed twice with cold PBS and lysed in cold radioimmunoprotein assay buffer (RIPA: 20 mM Tris-HCl pH 7.5, 150 mM NaCl, 0.1% SDS, 1% Na deoxycholate, 1% Triton X-100), supplemented with 1 mm phenylmethylsulphonyl fluoride, protease inhibitor (Complete Mini, Roche, Milan, Italy) and phosphatase inhibitor (PhosSTOP, Roche, Milan, Italy) cocktails. Cell lysates (50 μg total proteins, quantified by Lowry Assay) were dissolved in sample buffer (50 mM Tris-HCl pH 6.8, 2% SDS, 10% glycerol, 100 g/L bromophenol blue, 10 mm β-mercaptoethanol), heated for 5 min, fractionated by SDS-polyacrylamide gel (4–20% gradient) electrophoresis and transferred to nitrocellulose filters (Trans-Blot Turbo, Biorad, Milan, Italy). Membranes were incubated with the following primary antibodies: rabbit monoclonal anti-human NF-κB total p65 (clone D14E12) or P-p65 (Ser536, clone 93H1) or mouse monoclonal anti-human α-tubulin (clone DM1A), all from Cell Signaling Technology (Danvers, MA, USA). Proteins were detected with horseradish peroxidase-conjugated secondary antibodies (Cell Signaling Technology, Danvers, MA, USA) and enhanced chemiluminescence reagent (ECL kit Lite Ablot Extend, Euroclone, Milan, Italy), followed by the analysis of chemiluminescence with the charge-coupled device camera detection system Las4000 Image Quant (GE Health Care Europe GmbH, Milan, Italy). Relative levels of the phosphorylated p65 protein were normalized to their corresponding unphosphorylated form. To verify equal protein loading, the housekeeping protein α-tubulin was used as internal control (Supplementary Figure S1). Three independent experiments, carried out in triplicate, were performed.

The experiments were conducted using the probiotic strains Lactobacillus amylovorus ATCC 33198 (LMG9434), L. amylovorus DSM 16698 and Limosilactobacillus reuteri subsp. porcinus DSM 110571. Each probiotic strain, pre-grown overnight in MRS at 37 °C, was inoculated at 1:500 dilution (L. amylovorus ATCC 33198 and DSM 16698) or diluted to OD = 1 and inoculated at 1% (v/v; L. reuteri subsp. porcinus DSM 110571) in MRS medium without glucose (MRSnoG, Liofilchem, Italy) or added with 2% (w/v) glucose (used as reference carbon source) or added with 2% (w/v) GOS (used as test carbon source), in separate tubes. The MRSnoG medium served as negative control. Each inoculum was dispensed into 96-well plates at a final volume of 200 μl (L. amylovorus ATCC 33198 and DSM 16698) or 300 μl (L. reuteri subsp. porcinus DSM 110571) per well, in triplicate. Bacterial growth was monitored by recording the OD₆₀₀ of the cultures for 22 h at 1 h intervals using an automated plate reader (Infinite M200, Tecan, Milan, Italy; or Multiskan FC, Life Technologies, Merelbeke, Belgium). The OD₆₀₀ values were normalized with respect to the MRS medium alone, representing the blank. One independent experiment was conducted in triplicate, for each strain.

The statistical significance of the differences was evaluated by one-way ANOVA followed by post hoc Tukey HSD test, after verifying normality and homogeneity of variance by Shapiro–Wilk's and Levene's tests, respectively. For ETEC F4⁺ growth curve and adhesion/invasion assays, Student's t-test was applied. In the figures, mean values with different superscript letters or asterisks (for ANOVA or t-test, respectively) significantly differ. Statistical significance was set at P values < 0.05. The statistical analyses were executed with Microsoft Office Excel 2011 upgraded with XLSTAT (version 4 March 2014).

GOS was preliminarily tested for its potential to directly impact on ETEC F4⁺ growth in LB and DMEM media (Supplementary Figure S2). In both media, the presence of 2% GOS enhanced pathogen growth during the exponential phase (2 h, 4 h and 8 h time points, P < 0.01 within each time point), while at late stationary phase (22 h time point), growth kinetics resulted to decrease in LB (Supplementary Figure S2A, P < 0.01), or be unchanged in DMEM (Supplementary Figure S2B), with respect to the corresponding media alone. We therefore concluded that GOS had no relevant direct inhibitory effect on ETEC F4⁺ growth.

In order to evaluate if GOS exposure could perturb intestinal epithelial permeability, TEER and phenol red Papp were measured in differentiated Caco-2 cells after treatment with 2% GOS, apically added to cell monolayers for up to 72 h. The results show that no significant differences could be observed between treated and untreated (control) cells, in terms of TEER measurements (Figure 2A) and phenol red Papp (Figure 2B), indicating that monolayers integrity was not affected by 2% GOS treatment. Indeed, permeability coefficients lower than the 1 × 10^–6 cm s^–1 threshold are considered indicative of intact cell monolayers, irrespective of differences among samples that are below this threshold.

The ETEC F4⁺ strain is known to be able to adhere to and being internalized by Caco-2 cells, with a number of recovered cells in the order of magnitude of 10⁶ CFU/ml and 10³ CFU/ml after adhesion and invasion, respectively, with initial bacterial loads of 1 × 10⁸ CFU/ml (28). The potential capacity of 2% GOS to counteract ETEC F4⁺ adhesion and invasion was investigated in Caco-2 cells. Figure 3A shows the results of the adhesion assay, indicating that the presence of 2% GOS significantly reduced pathogen adhesion, with a decrease of approximately 0.5 log CFU/ml, as compared to pathogen alone (P < 0.001). A similar reduction activity was also observed in the pathogen invasion assay, as shown in Figure 3B, where the number of invading ETEC F4⁺ cells was significantly reduced by approximately 0.5 log CFU/ml, as compared to pathogen alone (P < 0.01).

To evaluate the ability of GOS to inhibit the adhesion of ETEC F4⁺ to the porcine small intestinal epithelium, an ex vivo adhesion inhibition assay was performed. As shown in Figure 4, GOS reduced by 55% the amount of ETEC F4⁺ adhering to the villus epithelium. This suggests that GOS can inhibit the binding of ETEC F4⁺ to the small intestinal epithelium in pigs.

ETEC F4⁺ is known to affect Caco-2 cell monolayer integrity, with a TEER decrease starting at 90 min after AP infection (28). The potential GOS ability to counteract the ETEC F4⁺-induced permeability increase was investigated in Caco-2 cells. As expected, ETEC F4⁺ infection induced a significant TEER decrease (Figure 5A, P < 0.05), as compared to untreated (control) and GOS-treated cells, either alone or in combination with pathogen infection (Figure 5A). This decrease was associated with a biologically relevant phenol red Papp increase, as the corresponding values were in the order of magnitude of 2.2 × 10⁻⁶ cm s⁻¹, indicating that the TJ were open, while the cells co-incubated with ETEC F4⁺ and GOS showed phenol red Papp values comparable to untreated and GOS-treated cells (Figure 5B, P < 0.001).

ETEC F4⁺ has been already shown to damage cell junctions, as previously reported (31). Specifically, TJ proteins of Caco-2 cells are affected by ETEC F4⁺ infection. To evaluate the potential of GOS to protect TJ proteins from pathogen-induced damage, immunolocalization of occludin and ZO-1 was evaluated by immunofluorescence in Caco-2 cells treated with ETEC F4⁺ and GOS, either alone or in combination. As expected, ETEC F4⁺ infection induced occludin and ZO-1 delocalization and disappearance from cell boundaries (Figures 6, 7 for occludin and ZO-1, respectively), indicating cell junction rupture, as also suggested by TEER and phenol red Papp measurements. GOS treatment did not affect TJ proteins, which remained properly distributed on cell boundaries similarly to untreated (control) cells. When Caco-2 cells were treated with GOS in combination with ETEC F4⁺, both occludin and ZO-1 remained correctly localized on cell boundaries, indicating an ability of GOS in preventing cell membrane damage induced by the pathogen.

In a previous study we showed that ETEC F4⁺ infection induced translocation of phosphorylated (P)-p65 into the nucleus of Caco-2 cells, indicating activation of the NF-κB inflammatory pathway (32). To investigate the ability of GOS to inhibit this inflammatory activity, the phosphorylation status of p65 was assessed by immunofluorescence and Western blot analysis in Caco-2 cells treated with ETEC F4⁺ and GOS, either alone or in combination. The results confirmed that ETEC F4⁺ infection induced a migration of P-p65 into the nucleus, and a significant increase of the phosphorylated form of the p65 subunit (P < 0.05), as compared to the phosphorylation level of p65 observed into untreated (control) cells, as shown in Figures 8, 9, respectively. Cell treatment with GOS alone did not induce any change in p65 phosphorylation as compared to control, indicating that GOS did not induce NF-κB activation (Figures 8, 9 for Western blot and immunofluorescence, respectively). In Caco-2 cells treated with ETEC F4⁺ in combination with GOS, phosphorylation level of p65 did not differ from control cells, indicating that GOS were able to prevent the pathogen-induced migration of p65 into the nucleus (Figure 8). The immunolocalization analysis of P-p65 supports the Western blot results, showing that no nuclear localization of P-p65 could be observed in cells treated with ETEC F4⁺ in combination with GOS or treated only with GOS (Figure 9).

The potential prebiotic activity of GOS was assessed by testing its effect on the growth of three probiotic strains of porcine origin. The experimental procedure consisted in replacing the carbon source normally used in bacterial culture medium, namely glucose, with the compound to be tested (i.e. GOS), followed by a comparative evaluation of bacterial growth on the different carbon sources. The growth curves of L. amylovorus ATCC 33198, L. amylovorus DSM 16698 and L. reuteri subsp. porcinus DSM 110571 (reported in Figures 10A–C, respectively) clearly showed the ability of all the three probiotic strains to successfully use GOS as a carbon source. Indeed, the respective growth kinetics during 22 h incubation displayed a similar trend between GOS-supplemented and glucose-supplemented conditions, with the typical lag, exponential and stationary phases, differently from the negative control (no carbon source), where no growth at all could be observed (Figures 10A–C). In particular, the overall OD₆₀₀ values of L. amylovorus DSM 16698 (8 h, 10 h, 12 h and 22 h time points) and L. reuteri subsp. porcinus DSM 110571 (10 h, 12 h, 14 h and 22 h time points) resulted significantly higher as compared to glucose-supplemented conditions (P < 0.05, Figures 10B, C, respectively). On the other hand, the early and mid exponential phase (8 h and 10 h time points) of the L. amylovorus ATCC 33198 growth curve in GOS-supplemented medium overlapped with that in the presence of glucose, although reaching OD₆₀₀ values significantly lower than those observed in the presence of glucose at the late exponential (12 h time point) and at the late stationary (22 h timepoint) phase (P < 0.05, Figure 10A).