In this study, the HT-29 human colonic epithelial cell line (American Type Culture Collection, Manassas, VA) was chosen as this IEC model has been extensively used by many groups investigating epithelial responses to bacteria.^11,13,36,37 HT-29 cells were cultured in modified McCoy's 5A medium (Gibco-BRL, Grand Island, NY) supplemented with 10% heat-inactivated fetal calf serum (FCS) (Sigma-Aldrich) in the presence or absence of 100 U/ml penicillin G and 100 µg/ml streptomycin (Gibco-BRL). The HeLa (human cervix epithelial-like) cell line (ATCC) was cultured in minimal essential medium (Gibco-BRL) supplemented with 0·1 mm non-essential amino acids (Gibco-BRL), 1·5 mm l-glutamine, 10% heat-inactivated FCS, 100 U/ml penicillin G and 100 µg/ml streptomycin. The cells were routinely propagated in 75-cm² tissue culture flasks at 37° in a humidified, 5% CO₂ incubator until they approached 80–90% confluency. Subsequently, the cells were trypsinized and used in experimental investigations as specified below.

For all assays, cell viability was determined by trypan blue exclusion, and a known number of HT-29 cells were seeded into 25-cm² culture flasks, 9·6-cm² six-well plates, or 3·8-cm² 12-well plates. In some experiments, after 24 hr HT-29 cells were incubated for varying times with B. infantis, L. salivarius or S. typhimurium at a bacterial to epithelial cell ratio of 10 : 1, a dose used previously by others.¹³ In other investigations, HT-29 cells were grown to confluence, and confluent monolayers were treated with 1 × 10⁵, 1 × 10⁶ or 1 × 10⁷ CFU/ml B. infantis or L. salivarius. Dose–response studies were performed to determine the optimal concentrations of flagellin and TNF-α to use for stimulation of IECs and, subsequently, HT-29 cells were treated with 0·5 µg/ml purified S. typhimurium flagellin (InvivoGen Corp., San Diego, CA) or 5 ng/ml TNF-α (R & D Systems, Minneapolis, MN). Bacterial survival following 1, 2, 6 or 11 hr of incubation with HT-29 cells was assessed by plating serial dilutions of cell culture supernatants on MRS agar or TSA. Following a 24-hr incubation (for S. typhimurium and L. salivarius) or a 48-hr incubation (for B. infantis), the CFU were quantified. In some experiments, HT-29 cells were pretreated for 2 hr with a known dose of commensal bacteria and subsequently were infected with an equivalent dose of S. typhimurium or treated with 0·5 µg/ml flagellin or 5 ng/ml TNF-α for varying times.

IEC viability was assessed using propidium iodide. Briefly, cell culture supernatants were removed from untreated and bacteria-treated HT-29 cells. The cells were scraped, washed in PBS, and resuspended in 50 µg/ml propidium iodide (Sigma-Aldrich) and 5 Kunitz units/ml ribonuclease A in PBS (Sigma-Aldrich). The stained cells were analysed using an Epics Elite flow cytometer (Beckman Coulter, Inc., Fullerton, CA), and propidium iodide fluorescence was detected at 675 nm. The pH of the corresponding cell culture supernatants was measured using a PHM61 Laboratory pH Meter (Radiometer A/S, Copenhagen, Denmark).

The 2-hr mRNA samples from the gene array experiments were selected for RT-PCR analysis of IL-8 mRNA. mRNA (250 ng) was reverse-transcribed to yield cDNA using 40 U/µl RNase inhibitor (Ambion) and the Expand Reverse Transcriptase system (Roche Diagnostics Ltd, East Sussex, UK) according to the manufacturer's protocol. Real-time RT-PCR for human IL-8 was performed in a LightCycler 1·5 (Roche) using a previously described IL-8 primer set³⁸ (MWG-Biotech Ltd, Ebersberg, Germany) and a Fast Start DNA Master SyBr Green I kit as specified by the manufacturer (Roche). Calculations were performed using GAPDH as a relative standard, and the GAPDH primers (forward, 5′-ACCACAGTCCATGCCATCAC-3′; reverse, 5′-TCCACCACCCTGTTGCTGTA-3′) were synthesized by Clontech (BD Biosciences, San Jose, CA).

In humans, MUC3 is among the predominant ileocolonic mucins⁴ and E-cadherin is a cell surface protein involved in cell adhesion.³⁹ We used real-time RT-PCR to examine the expression of E-cadherin and two MUC3 genes, MUC3A and MUC3B,⁴⁰ in commensal-treated HT-29 cells. In the absence of antibiotics, confluent HT-29 monolayers were treated with 1 × 10⁷ cells/ml B. infantis or L. salivarius for 1 or 2 hr. The cells were harvested by trypsinization, and total RNA was isolated using the Absolutely RNA® Miniprep kit (Stratagene, La Jolla, CA) according to the manufacturer's protocol. Samples were treated twice with RNase-free DNase I. Total RNA (1 µg) was reversed-transcribed from random primers (Roche) using ImProm-II RT (Promega, Madison, WI) and RNasin Plus RNase inhibitor (Promega). Real-time RT-PCR analysis was performed on 5 µl of the resulting cDNA using LightCycler TaqMan Master (Roche), and LightCycler Uracil-DNA glycosylase (Roche) as specified by the manufacturer, together with Universal ProbeLibrary probes (Exiqon, Vedbaek, Denmark) in a final reaction volume of 20 µl. Primers designed using the Universal ProbeLibrary Assay design centre (http://www.roche-applied-science.com/sis/rtpcr/upl/adc.jsp) were synthesized by MWG Biotech Ltd. PCR amplification was performed using a LightCycler 1·5 instrument (Roche). Thermal cycling conditions comprised 2 min at 40° and 10 min at 95°, followed by 45 amplification cycles at 95° for 10 seconds, 60° for 30 seconds, and 72° for 1 second, and a 40° cooling cycle for 30 seconds. Calculations were performed using GAPDH as the endogenous control reference gene. Fold difference in gene expression was calculated according to the standard formula 2^{(Ec–Rn) – (Et–Rt)}, where Ec is the crossing threshold (ct) of the experimental gene in untreated control samples, Rn is the ct of GAPDH in untreated samples, Et is the ct of the experimental gene in treated samples, and Rt is the ct of GAPDH in treated samples. Data are presented as the mean ± standard error (SE) of three independent experiments.

HeLa cells were selected as an epithelial cell model to visualize intracellular NF-κB localization as they have an optimal nucleus-to-cytoplasm ratio and are epithelial cells.¹³ Intracellular NF-κB activation in the model was assessed using immunohistochemistry as described previously.¹³ Briefly, HeLa cells (1 × 10⁷ cells/ml) were grown on coverslips in six-well plates for 3 days, and incubated with 1 × 10⁷ B. infantis or 20 ng/ml TNF-α for 30 min. The coverslips were washed, fixed, blocked, and incubated with 6 µg of mouse anti-NF-κB antibody (BD Pharmingen, Mississauga, Ontario, Canada), followed by fluorescein isothiocyanate-conjugated goat anti-mouse immunoglobulin G (IgG) (1 : 64) (Sigma-Aldrich) as described previously.¹³ The coverslips were mounted in glycerol-PBS, and slides were viewed using a LSM510 (Carl Zeiss, Jena, Germany) confocal microscope.

HT-29 cells (3 × 10⁶) were seeded in 25-cm² tissue culture flasks in 5 ml of antibiotic-free medium. After 24 hr of incubation, the medium was replaced with antibiotic- and serum-free medium, and the cells were treated with S. typhimurium, flagellin or TNF-α in the presence or absence of probiotic pretreatment for varying times. Nuclear proteins were extracted using the Active Motif Nuclear Extract kit (Active Motif Europe, Rixensart, Belgium) according to the manufacturer's instructions, and the total protein concentration of the lysates was determined by Bradford assay (Bio-Rad, Hercules, CA). Activation of the NF-κB p65 subunit in 5 µg of HT-29 nuclear extracts was determined using an NF-κB p65 enzyme-linked immunosorbent assay (ELISA)-based transcription factor assay kit (TransAM assay) (Active Motif Europe) according to the manufacturer's protocol. The NF-κB detecting antibody recognizes an epitope on p65 that is accessible only when NF-κB is activated. The positive control Jurkat nuclear extract provided with the kit was used to assess assay specificity.

Subconfluent or confluent HT-29 cells grown in antibiotic- and serum-supplemented media were treated with known doses of probiotic bacteria or pro-inflammatory stimuli for 6 or 24 hr. Following treatment, immunoreactive IL-8 protein levels in cell-culture supernatants were quantified using an ELISA DuoSet kit (R & D Systems) according to the manufacturer's protocol. In the gene array assays, the IL-8 DuoSet kit was used also to assess extracellular IL-8 protein levels in cell-culture supernatants and intracellular IL-8 in these cells following lysis with ice-cold water.

In accordance with a protocol approved by the Ethics Committee of Cork University Hospital, Cork, Ireland, peripheral blood from healthy volunteers (n = 3) was collected by venepuncture into sterile ethylenediaminetetraacetic acid (EDTA) vacutainer tubes. PBMCs were purified from buffy coats in histopaque tubes (Greiner Bio-one, Inc., Longwood, FL) using Ficoll-Hypaque (Pharmacia, Dübendorf, Switzerland) gradient centrifugation (400 g for 30 min). PBMCs were taken from the interface and washed four times with Ca²⁺- and Mg²⁺-free PBS, and were finally resuspended in Ca²⁺- and Mg²⁺-free degassed column buffer (PBS, pH 7·2, supplemented with 0·5% BSA and 2 mm EDTA). Subsequently, monocytes were isolated from mononuclear cells by negative selection using the Monocyte Isolation Kit II (Miltenyi Biotec) with MACS Column and MACS Separator (Miltenyi Biotec) according to the manufacturer's instructions. To generate DCs, monocytes were incubated for 5 days in DMEM containing 10% FCS, 100 000 IU/ml granulocyte–monocyte colony-stimulating factor, and 40 000 IU/ml IL-4 (BD Biosciences). Cell purity was assessed by cell surface staining and flow cytometry using antibodies obtained from BD Biosciences, and viability was determined by trypan blue exclusion. Cell purity as determined by HLA-DR-positive and CD3/CD14/CD16/CD19/CD20/CD56-negative cells was consistently >85%, and cell viability was consistently >98%. Subsequently, PBMC-derived DCs (1 × 10⁶ cells/ml) were seeded in 1·9-cm² 24-well plates in DMEM supplemented with 10% heat-inactivated FCS, 100 U/ml penicillin G, 100 µg/ml streptomycin, and 2·5 µg/ml fungizone (Gibco-BRL). Cells were treated with medium alone (negative untreated controls), or with 10 : 1 B. infantis or L. salivarius for 48 hr. Subsequently, IL-10 and TNF-α protein levels in cell-culture supernatants were quantified using commercially available ELISA kits (R & D Systems) according to the manufacturer's protocols. In order to determine the levels of endotoxin in the commensal bacterial preparations we used a Limulus amebocyte lysate (LAL) gel-clot assay with a sensitivity of 0·03 Endotoxin units/ml (Charles River Laboratories, Inc., Charleston, SC). The assay was used according to the manufacturer's instructions and Escherichia coli control standard endotoxin (Charles River Laboratories, Inc.) was used to confirm LAL-reagent sensitivity. Tap water was used as a positive control and LAL-reagent water (Charles River Laboratories, Inc.) as a negative control.

Human Cytokine Expression Arrays were used to examine inflammatory gene expression in bacteria-treated HT-29 cells. Treatment with B. infantis or L. salivarius for 0·5, 2 or 6 hr did not augment the expression of any of the 847 immune-related genes assayed, whereas infection with S. typhimurium increased expression of 36 genes associated with pro-inflammatory responses (Table 1). These included a repertoire of signal transducers, immunoreceptors, chemokines, and transcription factors, among them NF-κB and the chemokine IL-8.

In order to confirm the gene array results, we used HeLa cervical epithelial cells as an independent epithelial model to assess NF-κB activation by immunohistochemistry. Untreated HeLa cells expressed NF-κB p65 constitutively in the cytoplasm (Fig. 2a), and treatment with 20 ng/ml TNF-α caused a rapid cytoplasmic to nuclear translocation of NF-κB, and increased levels of NF-κB p65 were detected in the nucleus within 30 min. In contrast, treatment with B. infantis for the same time did not induce nuclear translocation and NF-κB remained localized to the cytoplasm (Fig. 2a). Similarly, in HT-29 cells, treatment with B. infantis or L. salivarius for 30 min or 1 hr did not stimulate the DNA-binding activity of NF-κB p65 compared with untreated cells, whereas infection with S. typhimurium or treatment with TNF-α significantly increased NF-κB DNA-binding activity at 1 hr or 30 min, respectively (Fig. 2b). NF-κB DNA-binding activity was detected in the positive control Jurkat nuclear extract, and the specificity of NF-κB binding in the assay was confirmed by competition with free wild-type NF-κB consensus oligonucleotide or mutated NF-κB oligonucleotide (Fig. 2b). The data demonstrate that epithelial cells respond differently to various antigens, and, in contrast to S. typhimurium and TNF-α, the commensal bacteria used in this study did not induce NF-κB nuclear translocation or DNA-binding activity.

In subconfluent HT-29 cells treated with B. infantis or L. salivarius, the levels of IL-8 mRNA or protein detected after 2 or 24 hr, respectively, were similar to those found at baseline in untreated cells (Figs 3a and b). Conversely, infection with S. typhimurium stimulated a significant increase in IL-8 mRNA expression after 2 hr (13·3-fold) and IL-8 protein secretion after 24 hr (7-fold) compared with untreated cells. As shown in Fig. 3(c), S. typhimurium quickly induced IL-8 mRNA and maximum levels were detected within 2 hr of infection. Infection also resulted in a time-dependent accumulation of extracellular IL-8 that was associated with a corresponding decrease in the amount of detectable intracellular IL-8 protein (Fig. 3c).

We next examined whether treatment with increasing doses of B. infantis or L. salivarius (1 × 10⁵, 1 × 10⁶, or 1 × 10⁷ CFU/ml) affected IL-8 secretion by confluent HT-29 cell monolayers. IL-8 protein levels were measured after 6 hr. A dose-dependent inhibition of baseline IL-8 secretion by confluent HT-29 monolayers was observed (Fig. 4). At a concentration of 1 × 10⁷ CFU/ml, which under the experimental conditions was equivalent to approximately 10 bacteria per epithelial cell, both B. infantis and L. salivarius significantly inhibited basal IL-8 secretion by 38 and 44%, respectively. This suggests that these commensal bacteria exert anti-inflammatory effects on the epithelium by down-regulating the secretion of IL-8.

We next determined whether the modulation of S. typhimurium-induced pro-inflammatory responses by the probiotic bacteria was attributable to interference with pathogen binding to IECs. Binding of S. typhimurium to HT-29 cells was evaluated in the presence or absence of probiotic pretreatment using biofluorescence and the plate dilution method. The two methods yielded complementary data and, as shown in Table 2, pretreatment with neither B. infantis nor L. salivarius affected the number of IEC-associated S. typhimurium detected 2 hr after infection. These data indicate that the immunoregulatory effects of the probiotic bacteria observed in this study are not attributable to competitive inhibition of S. typhimurium association with HT-29 cells.

Flagellin is a key activator of pro-inflammatory responses to Salmonella in IECs.^22,42 Therefore, we explored whether the probiotic bacteria could inhibit IEC responses to Salmonella flagellin. Firstly, in dose–response studies, we incubated various doses of flagellin from S. typhimurium (0·1, 0·5 and 1·0 µg/ml) with confluent HT-29 monolayers, and found that, at a concentration of 0·5 µg/ml, flagellin significantly induced NF-κB activation and IL-8 secretion. Treatment with 0·5 µg/ml flagellin for 1 hr resulted in a 3·15 (± 0·3)-fold increase in NF-κB binding activity compared with untreated cells, and 1876 (± 262) pg/ml IL-8 was detected after 6 hr (P < 0·05) (data not shown). HT-29 monolayers were treated for 2 hr with either B. infantis or L. salivarius (1 × 10⁶ or 1 × 10⁷ CFU/ml), followed by 0·5 µg/ml flagellin for 6 hr. As shown in Fig. 6, pretreatment with L. salivarius did not affect flagellin-induced IL-8 at either dose tested. However, flagellin-induced IL-8 was significantly inhibited in cells that were pretreated with 1 × 10⁷ CFU/ml B. infantis. The results suggest strain-specific antagonistic effects on inducible IL-8 expression in IECs.

It has been proposed that some strains of probiotic bacteria exert beneficial effects on the intestinal epithelium by up-regulating the expression of mucin genes.^36,43 We used real-time RT-PCR to determine the expression of MUC3A and MUC3B mRNA in confluent HT-29 cells treated for 1 or 2 hr with 1 × 10⁷ cells/ml probiotic bacteria. MUC3A and MUC3B mRNA was detected in untreated HT-29 cells, and, as shown in Fig. 7, treatment with B. infantis or L. salivarius had no significant effect on the expression of these genes. Similarly, no difference was observed in E-cadherin mRNA expression following treatment with the commensal bacteria (data not shown).