Adhesive and Chemokine Stimulatory Properties of Potentially Probiotic Lactobacillus Strains

Transcription

1 125 Journal of Food Protection, Vol. 70, No. 1, 2007, Pages Copyright, International Association for Food Protection Adhesive and Chemokine Stimulatory Properties of Potentially Probiotic Lactobacillus Strains MARÍA G. VIZOSO PINTO, 1 TOBIAS SCHUSTER, 1 KARLIS BRIVIBA, 2 BERNHARD WATZL, 2 WILHELM H. HOLZAPFEL, 1 AND CHARLES M. A. P. FRANZ 1 * 1 Institute of Hygiene and Toxicology and 2 Institute of Physiology of Nutrition of the Federal Research Centre for Nutrition and Food, Haid-und-Neu-Strasse 9, D Karlsruhe, Germany MS : Received 17 May 2006/Accepted 28 July 2006 ABSTRACT Five Lactobacillus plantarum strains and two Lactobacillus johnsonii strains, stemming either from African traditionally fermented milk products or children s feces, were investigated for probiotic properties in vitro. The relationship between the hydrophobic-hydrophilic cell surface and adhesion ability to HT29 intestinal epithelial cells was investigated, and results indicated that especially the L. johnsonii strains, which exhibited both hydrophobic and hydrophilic surface characteristics, adhered well to HT29 cells. Four L. plantarum and two L. johnsonii strains showed high adherence to HT29 cells, generally higher than that of the probiotic control strain Lactobacillus rhamnosus GG. Most strains with high adhesion ability also showed high autoaggregation ability. The two L. johnsonii strains coaggregated well with the intestinal pathogens Listeria monocytogenes Scott A, Staphylococcus aureus ATCC 25923, Escherichia coli ATCC 25922, and Salmonella enterica serovar Typhimurium ATCC The L. plantarum BFE 1685 and L. johnsonii 6128 strains furthermore inhibited the adhesion of at least two of these intestinal pathogens in coculture with HT29 cells in a strain-dependent way. These two potential probiotic strains also significantly increased interleukin-8 (IL-8) chemokine production by HT29 cells, although modulation of other cytokines, such as IL-1, IL-6, IL-10, monocyte chemoattractant protein-1 (MCP-1), tumor necrosis factor alpha (TNF- ), and transforming growth factor (TGF- ), did not occur. Altogether, our results suggested that L. plantarum BFE 1685 and L. johnsonii BFE 6128 showed good adherence, coaggregated with pathogens, and stimulated chemokine production of intestinal epithelial cells, traits that may be considered promising for their development as probiotic strains. Lactobacillus spp. are normal inhabitants of the human oral cavity and of the gastrointestinal tract of humans and animals (33, 46). Not only the resident lactobacilli, but also the exogenous bacteria with functional properties can be helpful to the host, the latter by promoting a beneficial balance of the autochthonous microbiota of the gut. Strains with such health-promoting characteristics have been defined as probiotics (11, 17, 18). Probiotic strains furthermore have beneficial effects on the host by controlling undesirable microorganisms and by modulating the immune system (12, 32). However, despite efforts to elucidate their mechanism of action, it is still not well understood how probiotics work (30, 42). The ability to adhere to the intestinal epithelium is one of the main criteria for selecting new probiotic strains, as this property allows strains to remain at least transiently in the intestinal tract and exert their probiotic effects, such as excluding pathogenic bacteria by competing for adhesion sites (2). Adhesion of bacteria to animal cells is a much more complex phenomenon than adhesion of bacteria to inanimate surfaces. This is because of the complexity of both microbial cell surfaces and eukaryotic membranes and the ability of living cells to regulate the expression of molecules on their surface in response to changes in the environment. Thus, this process involves nonspecific as well as * Author for correspondence. Tel: ; Fax: ; charles.franz@bfel.de. specific ligand-receptor mechanisms (14). Hydrophobic interactions contribute in the initial adhesion of numerous pathogens to tissues (8). As early as 1924, Mudd and Mudd (34) demonstrated that bacteria can vary considerably in their degree of hydrophobicity. Furthermore, it was suggested that bacteria with high hydrophobic surfaces reversibly adhere to intestinal cells (9). If this is followed by a specific ligand-receptor bond, the adhesion may become irreversible (14). In addition, Del Re et al. (7) and Kos et al. (24) suggested that autoaggregation can be correlated to adhesion to intestinal epithelial cells. Cesena et al. (6) have shown that the gastrointestinal persistence in vivo, as well as the adhesion to epithelial cells in vitro, was higher for a Lactobacillus crispatus strain with an aggregating phenotype than for its nonaggregating mutant. Autoaggregation and coaggregation have also been related to the ability to interact closely with undesirable bacteria (9, 15). Epithelial cells are the first site of entry of intestinal pathogenic bacteria and may provide early signals for the acute mucosal inflammatory response by releasing cytokines and inflammatory mediators. In response to several invasive pathogens, epithelial cells have been shown to secrete interleukin-1 alpha (IL-1 ), IL-6, IL-8, granulocytemacrophage colony-stimulating factor, growth-regulated oncogene alpha, monocyte chemoattractant protein-1 (MCP-1), and tumor necrosis factor alpha (TNF- ) (29). To a lesser extent, some nonpathogenic bacteria also seem to

2 126 VIZOSO PINTO ET AL. J. Food Prot., Vol. 70, No. 1 TABLE 1. Adhesion of lactobacillus strains to HT20 cells and to solvents a % adhesion ( SE) to c : Strains Origin Adhesion index b n-hexadecane Chloroform Ethyl acetate L. plantarum BFE 1684 Infant feces * d BFE 1685 Infant feces * BFE 5878 Kule naoto * BFE 5092 Kule naoto BFE 5759 Kwerionik ATCC 8014 ATCC * L. johnsonii BFE 6128 Kule naoto 100 e ND* BFE 6154 Kule naoto 100 e ND* Probiotic control strains L. johnsonii BFE 663 f LA-1, Nestlé * L. rhamnosus GG f Valio * L. paracasei BFE 675 g Actimel, Danone L. casei Shirota BFE 688 g Yakult a ATCC, American Type Culture Collection (Manassas, Va.); ND, not determined. b Adhesion index: mean standard error (n 4). Number of bacteria adhering to 100 HT29 cells. c Means standard errors of two measurements of three separate experiments. d *, significantly different (ANOVA, Dunnet test, P 0.05) from the negative control (L. casei Shirota BFE 688). e More than 100 bacteria per 100 HT29 cells were not possible to count; therefore, they were given an adhesion index of 100 and no standard error was determined. f Positive control strains. g Negative control strains. be able to stimulate the production of proinflammatory and anti-inflammatory mediators (25). In this study, we investigated the adhesion of selected Lactobacillus strains, isolated from traditional fermented products and infant feces, to intestinal epithelial cells (HT29 cell line). Previously, the in vitro adhesion ability, together with ph tolerance, coaggregation ability, and antimicrobial properties of probiotic strains, was verified to correlate with colonization data in vivo (6, 21). Thus, the in vitro assays in this study served as a first step in evaluating the possible adhesion ability for the development of probiotic strains. The hydrophobicity and electron-donor properties of their surfaces, as well as their abilities to autoaggregate and coaggregate with pathogenic strains, were screened to determine whether a relationship with adhesion to enterocyte-like cells could be explained on the basis of these properties. The capacity of these potential probiotic strains to induce the production of IL-8 of epithelial cells in vitro was tested by coculturing the lactic acid bacteria with HT29 cells. MATERIALS AND METHODS Bacterial strains and growth conditions. The Lactobacillus strains used in this study (Table 1) were isolated from children s feces and from two traditional fermented milk products: Kule naoto (Kenya) and Kwerionik (Uganda). They were previously identified by phenotypic and genotypic methods and selected for their in vitro probiotic characteristics (48). Lactobacillus plantarum ATCC 8014 and the known probiotic strains Lactobacillus johnsonii LA-1 (Nestlé), Lactobacillus paracasei Actimel (Danone), Lactobacillus casei Shirota (Yakult), and Lactobacillus rhamnosus GG (Valio) were also used in this study. Bacteria were grown in deman Rogosa Sharpe broth (MRS broth; Merck, Darmstadt, Germany) at 37 C overnight before use. For long storage, bacteria were maintained at 20 C in 15% (vol/vol) glycerol. Listeria monocytogenes Scott A, Enterococcus faecium DSM 13590, Salmonella enterica serovar Typhimurium (Salmonella Typhimurium) ATCC 14028, Staphylococcus aureus ATCC 25923, and Escherichia coli ATCC were cultured in Standard I (Merck) broth at 37 C. Probiotic reference strains and new isolates were maintained as multiple stocks to prevent laboratory adaptation, and excessive subculturing was thus avoided. HT29 cell culture and coculture with bacteria. The enterocyte-like HT29 cell line was obtained from the German collection of microorganisms and cultures (DSMZ, Braunschweig, Germany). Cells were routinely cultured in Dulbecco modified Eagle medium (DMEM; Invitrogen, Karlsruhe, Germany) supplemented with 10% (vol/vol) heat-inactivated (56 C, 30 min) fetal calf serum (FCS; PAA Laboratories GmbH, Cölbe, Germany) that contained penicillin G (100 U/ml; Invitrogen) and streptomycin sulfate (100 g/ml; Invitrogen) in T 80 culture flasks (Nunclon Surface, Nunc, Roskilde, Denmark). Cultures were incubated at 37 C under 5% CO 2. The medium was changed every other day, and the cells were subcultured every 4 to 8 days by treatment with trypsin-edta (0.25% trypsin with EDTA 4Na; Invitrogen) for 3 to 5 min at 37 C. Cells were diluted 1:3 with trypan blue solution (1 part trypan blue, 2 parts bidistilled water) (Sigma, Taufkirchen, Germany) to check for vitality and then counted with a hemocytometer. For coculture experiments, HT29 cells per well were seeded in six-well plates (Falcon, VWR International, Bruchsal, Germany) and allowed to attach and grow for 48 h. Subconfluent monolayers were washed twice with endotoxin-free phosphate-buffered saline (PBS; ph 7.4) (Dulbecco, Invitrogen) before

3 J. Food Prot., Vol. 70, No. 1 LACTOBACILLUS SPP. ADHESION AND CHEMOKINE STIMULATION 127 adding new culture medium without FCS or antibiotics to the control wells and suspensions of probiotic candidates containing CFU/ml to trial wells. In another experimental set, cells were treated with lipopolysaccharide (0.1 g/ml) (LPS; no. L4516, Sigma). These cocultures were incubated for 24 h, after which supernatants were collected, centrifuged, and stored at 80 C until assayed. Cytokine ELISA. Culture supernatants were analyzed for IL-1, IL-8, IL-6, MCP-1, and transforming growth factor (TGF- ) by means of the enzyme-linked immunosorbent assay (ELISA) DuoSet (R&D Systems Europe, Abingdon, UK) according to the manufacturer s specifications. In addition, culture supernatants were analyzed for IL-10 and TNF- by means of ELISA CytoSets (Biosource Europe S.A., Nivelles, Belgium) according to the manufacturer s instructions. Cytotoxicity assays. Lactate dehydrogenase is a stable cytoplasmatic enzyme in eukaryotic cells that is released to the cell culture supernatant only upon damage of the plasma membrane. The activity of this enzyme was measured in the collected supernatants with the cytotoxicity detection kit (lactate dehydrogenase) from Roche (Mannheim, Germany) with a spectrophotometric microtiter plate reader. In a parallel experiment, the vitality of HT29 cells treated with bacterial suspensions was studied by means of the MTT assay (49), which relies on the reduction of MTT [3- (4,5-dimethyldiazol-2-yl)-2,5-diphenyltetrazolium bromide], Sigma) by metabolically active mitochondrial dehydrogenases to formazan salts. HT29 cells grown in 96-well plates to 80 to 90% confluence were treated with 100 l of a suspension of each probiotic strain ( CFU/ml) for 24 h. At the end of the treatment period, cells were washed with PBS three times, and 100 l of MTT (0.5 mg/ml in DMEM) was added to each well at 37 C for 4 h, protected from light. The resulting formazan salts were dissolved in 100 l of a solution containing 10% (vol/vol) sodium dodecyl sulfate and 0.3% (vol/vol) HCl and measured colorimetrically at 560 nm. Bacterial adhesion to HT29 cells. Overnight cultures were washed three times by centrifuging at 7,500 g for 10 min and resuspending in PBS. The bacterial cells were suspended and diluted to a concentration of CFU/ml in DMEM. The bacterial concentration was adjusted for each individual strain by correlating the A 580 nm with the CFU per milliliter as determined in previously performed growth curves. The inoculum density was confirmed by plate counting after the addition of the bacterial cells to DMEM. Prior to the adhesion assays, confluent HT29 cell monolayers grown on eight-well chamber slides (Falcon, VWR) were washed twice with PBS (ph 7.4) to remove cellular debris, FCS, and antibiotics. Bacterial suspensions ( CFU/ml, 363 l per chamber) were added to the monolayers (approximately HT29 cells, resulting in a ratio of 100 bacterial cells per HT29 cell) and incubated for 1 h at 37 C and 5% CO 2. After incubation, the cells were washed three times with PBS (ph 7.4) to remove nonadherent bacteria. Slides were allowed to air dry for 30 min at room temperature, fixed with methanol (100%; Merck) for 10 min, and Gram stained (7, 13, 35). Each chamber was 28 microscopic fields in length and 18 microscopic fields in width. Nine photographs were taken with a Leitz Camerasystem for automatic microscopy (Leitz, Wetzlar, Germany) fitted to a Leitz Aristoplan microscope. Photographs from nine different length-width coordinates (for different chamber wells, the same coordinates were used) were taken at 1,000 magnification. All HT29 cells, as well as attached bacteria, were counted in the nine microscopic fields, and the number of adherent bacteria per 100 HT29 cells was calculated (adhesion index). The adherence of each bacterial strain was determined in four chamber wells. For each well, nine microscopic fields were analyzed, resulting in a data set of 36 adhesion counts per strain. The adherence of strains was classified according to the criteria reported by Del Re et al. (7) as follows: nonadhesive, fewer than 5 bacteria per 100 cells; adhesive, 6 to 40 bacteria per 100 cells; strongly adhesive, more than 40 bacteria per 100 cells. Adhesion inhibition assays. Changes in adherence to epithelial cells of L. monocytogenes Scott A, E. faecium DSM 13590, Salmonella Typhimurium ATCC 14028, and E. coli ATCC were determined with an in vitro epithelial binding assay. HT29 cells were grown to 100% confluence on six-well plates (Falcon, VWR) and washed twice with PBS (ph 7.4). One milliliter of a suspension of lactobacilli (approximately to CFU/ml) in DMEM (without FCS and without antibiotics) was added to the wells and incubated for 1 h at 37 C and 5% CO 2 to allow adhesion of the strains. Control wells without lactobacilli were treated in the same way and incubated with DMEM without bacteria. Following the 1-h incubation, L. monocytogenes Scott A, E. faecium DSM 13590, Salmonella Typhimurium ATCC 14028, and enterotoxigenic E. coli ATCC were inoculated to the wells ( to CFU per well) and incubated under the same conditions. After a further 1-h incubation, nonadherent bacteria were removed by washing three times with PBS (ph 7.4). Cells with adherent bacteria were lysed with 1.5 ml of 1% (vol/ vol) Triton X-100 (Merck), serially diluted, and plated onto differential media (Palcam agar for L. monocytogenes with Listeria selective supplement [Oxoid, Basingstoke, UK]; MacConkey agar [Oxoid] for E. coli; Salmonella-Shigella agar [Oxoid] for Salmonella; and MRS [Merck] supplemented with erythromycin [64 g/ ml; Sigma] for E. faecium). Microbial adhesion to solvents. Microbial adhesion to solvents was measured by the method described by Rosenberg et al. (38) and Bellon-Fontaine et al. (1) with slight modifications. Overnight cultures of lactobacilli were harvested by centrifugation at 7,500 g for 5 min, washed twice in PBS (ph 7.4), resuspended, and diluted in PBS to reach an absorbance (A 0 ) of about 0.5 at A 580 nm (measured with a SmartSpec Photometer, Biorad, Munich, Germany). Two milliliters of test solvent was added to 2 ml of the bacterial cell suspension and mixed by vortexing for 1 min. The three solvents used were n-hexadecane (Sigma), which is an apolar solvent; chloroform (Merck), which is a monopolar and acid solvent; and ethyl acetate (Merck), which is a monopolar and basic solvent. The aqueous phase was taken after a 20-min incubation at room temperature, and its A 580 nm was measured (A 1 ). The percentage of bacterial adhesion to solvent was calculated as (1 A 1 /A 0 ) 100. Aggregation. The autoaggregation assay was performed as described by Del Re et al. (7) with modifications. Overnight cultures of bacterial cells were washed as described above, and the cells were suspended in PBS at ph 7.4, PBS at ph 4.0, a supernatant of the overnight bacterial cultures in MRS, or a neutralized supernatant (adjusted to ph 7.4 by the addition of NaOH). Two milliliters of each of these suspensions was vortexed for 10 s, and autoaggregation was determined after 2 h at room temperature. For determination of autoaggregation, 100 l was removed from the top of the suspension and transferred to a cuvette containing 900 l of PBS (ph 7.4). The absorbance (A 1 ) was measured at 580 nm. The autoaggregation percentage is expressed as follows: (1 A 1 /A 0 ) 100, where A 0 represents the absorbance at time zero.

4 128 VIZOSO PINTO ET AL. J. Food Prot., Vol. 70, No. 1 FIGURE 1. Scoring system for the coaggregation assay. Organisms used to illustrate the different scores were Salmonella Typhimurium ATCC together with L. plantarum BFE 1685 (A), Salmonella Typhimurium ATCC together with L. plantarum BFE 5092 and 10% supernatant from L. plantarum BFE 5092 grown in MRS broth (B), L. johnsonii BFE 6128 together with L. monocytogenes Scott A (C), and Salmonella Typhimurium ATCC together with L. johnsonii BFE 6128 and 10% supernatant from L. johnsonii BFE 6128 grown in MRS broth (D). Photographs were taken on a Zeiss inverted microscope ( 200 magnification). The photographs show scores 1 to 4 (1 for small and dispersed clumps (A), 2 for medium-sized and dispersed clumps (B), 3 for abundant and medium-sized clumps with turbid supernatant (C), and 4 for very big clumps and clear supernatant (D)). Coaggregation. Overnight cultures of lactobacilli and pathogen strains were washed twice with PBS (ph 7.4). To each well of a 24-well microtray (Costar, VWR), 500 l of the Lactobacillus suspension (10 9 CFU/ml) and 500 l of the intestinal pathogen suspension (10 9 CFU/ml) were added. After mixing, the trays were incubated at 37 C with shaking at 100 rpm for 2 h. A scoring system from 0 to 4 was used: 0, no coaggregation; 1, small and dispersed clumps; 2, medium-sized and dispersed clumps; 3, abundant and medium-sized clumps (C); and 4, very big clumps and clear supernatant (Fig. 1) (37). Each suspension was examined for aggregation microscopically at 200 magnification with an inverted microscope, though clumping at score levels of 3 and 4 could be seen macroscopically. Statistical analysis. Data are expressed as means standard errors of the mean of three experiments done in duplicate. Statistical significance between groups was assessed by a two-way analysis of variance (ANOVA), followed by the Dunnett test for multiple comparisons to controls. Probability values P of 0.05 or 0.01 were taken as criteria for a significant difference, as indicated in each case. Statistical analyses were performed by SAS 8.1 software (SAS Institute, Cary, N.C.). RESULTS Microbial adhesion to solvents. The use of three solvents allowed the evaluation of the hydrophobic-hydrophilic cell surface properties of lactobacilli and their Lewis acid-base (electron-donor and -acceptor) characteristics (5, 36). As shown in Table 1, L. plantarum strains BFE 5878, BFE 5092, BFE 5759, and ATCC 8014 and L. rhamnosus GG had a low partitioning percentage in the apolar solvent n-hexadecane, indicating that this group of strains possesses a hydrophilic surface. L. plantarum BFE 1684 and BFE 1685, L. johnsonii BFE 6128, BFE 6154, and LA-1 (BFE 663), L. paracasei BFE 675, and L. casei Shirota BFE 688 were characterized by a high affinity to n-hexadecane, indicating the hydrophobic nature of their surfaces (Table 1). The results of the microbial adhesion to chloroform showed that four L. plantarum strains (BFE 5878, BFE 5092, BFE 5759, and ATCC 8014) had a low affinity for this acidic solvent, whereas all other strains had a strong affinity for it. As n-hexadecane has the same van der Waals properties as chloroform, the differences observed in the adhesion properties are a result of the Lewis acid-base interactions resulting from the electron-donor properties of the surface of the strains tested (36). When ethyl acetate, an electron donor, was employed for determining the microbial adhesion to solvents, adhesion values were noticeably high (ranging from 36.3 to 79.2%) for L. johnsonii strains BFE 663, BFE 6128, and BFE 6154, which indicates that these bacteria also have an acidic surface character. L. plantarum 1684 and L. rhamnosus GG also showed a relatively high adherence value (30%) to ethyl acetate. The rest of the strains adhered poorly to this strongly basic solvent, with values ranging from 0 to 16.8%, reflecting the nonacidic character of these strains. Adhesion of Lactobacillus strains to HT29 cells. Eight Lactobacillus strains were investigated for their ability to adhere to the human intestinal epithelial HT29 cells. In addition, four commercial probiotic strains, previously well characterized for their adhesion properties, were used as reference strains (13, 45). L. rhamnosus GG and L. johnsonii (LA-1) served as positive control strains, while L. paracasei (Actimel, Danone) and L. casei Shirota (Yakult) were used as negative control strains and were described as being adhesive and as having a low and nonspecific binding capability, respectively (13, 45). After Gram staining of the slides, quantification of adherent bacterial strains, clearly visible as dark purple rods on a pale pink cell background, was possible. L. plantarum BFE 5759 could be characterized as nonadhesive, as its adherence values did not differ significantly from those of the negative control (L. casei Shirota, P 0.05). L. plantarum ATCC 8014 showed moderate adherence, with values differing signifi-

5 J. Food Prot., Vol. 70, No. 1 LACTOBACILLUS SPP. ADHESION AND CHEMOKINE STIMULATION 129 FIGURE 2. Effect of Lactobacillus spp. strains on the adherence of pathogens to intestinal epithelial-like cells (HT29). A to CFU per well inoculum of each pathogenic strain was added 1 h after the addition of lactobacilli ( to CFU per well). One hour after the addition of E. coli ATCC 25922, E. faecium DSM 13590, L. monocytogenes Scott A, or Salmonella Typhimurium ATCC adherent bacteria were quantified by determining the number of CFU. Results are expressed as means standard errors of at least four experiments. * Significantly decreased binding when compared with binding on wells without lactobacilli (P 0.05, ANOVA, Dunnett t test). Binding of the pathogen alone (indicated by darkly shaded square) or after the pretreatment of HT29 cells with L. rhamnosus GG (less darkly shaded square), L. casei Shirota BFE 688 (less darkly shaded square than for L. casei), L. johnsonii BFE 6128 (less darkly shaded square than for L. johnsonii), or L. plantarum BFE 1685 (square with no shading). cantly from those of the negative control (L. casei Shirota, P 0.05). All other probiotic candidates were classified as strongly adhesive, and they all showed significantly higher adherence values than the negative control (P 0.05). Interestingly, L. johnsonii strains BFE 6128 and BFE 6154 isolated from Maasai milk (Kule naoto) showed a much higher adhesive index than the positive control strains L. johnsonii LA-1 and L. rhamnosus GG, which are described in the literature as adhering well to intestinal epithelial cells (26). Effect of adherent Lactobacillus strains on the adhesion of pathogens to HT29 cells. L. plantarum BFE 1685 and L. johnsonii BFE 6128 were selected for adherence studies with pathogens because of their strong adhesion to HT29 cells. As control strains, the adhesive L. rhamnosus GG and the nonadhesive L. casei Shirota (BFE 688) were used. In these experiments, HT29 cells were preincubated with lactobacilli ( CFU/ml) before the addition of E. coli ATCC 25922, E. faecium DSM 13590, L. monocytogenes Scott A, or Salmonella Typhimurium ATCC at a concentration of CFU/ml, corresponding to a ratio of 100:1:1 (probiotic bacteria:pathogenic bacteria:ht29 cells). Adhesion of E. coli was significantly (P 0.01) reduced when HT29 cells were pretreated with L. rhamnosus GG or L. johnsonii BFE 6128 and was completely abolished when cells were pretreated with L. plantarum BFE On the other hand, only L. plantarum BFE 1685 was able to completely inhibit (P 0.01) E. faecium DSM adhesion, whereas none of the other Lactobacillus test or adherent control strains were able to influence the adhesion of this strain (Fig. 2). L. monocytogenes Scott A adhesion was significantly (P 0.01) reduced by all tested strains, except for the nonadherent L. casei Shirota (BFE 688). This latter strain did not influence FIGURE 3. Autoaggregation of Lactobacillus spp. strains. Overnight cultures were washed and resuspended in PBS at ph 7.4 (rectangle with no shading), in PBS at ph 4 (rectangle with diagonal shading), or in MRS spent culture supernatant (darkly shaded rectangle). A 580 nm was measured at the beginning (A 0 ) and after a 2-h incubation (A 1 )at23 C. Autoaggregation percent is expressed as (1 A 1 /A 0 ) 100. the adhesion of the E. coli, E. faecium, or L. monocytogenes strains used in this study, but it slightly reduced (P 0.05) the adherence of Salmonella Typhimurium ATCC cells. Salmonella Typhimurium adhesion was also significantly reduced by the presence of L. plantarum BFE 1685 cells (Fig. 2). Autoaggregation of Lactobacillus strains. Autoaggregation was studied on the basis of the sedimentation characteristics of the strains, because as cells aggregated, they sedimented and cleared the supernatant. The nonadherent reference strains L. paracasei BFE 675 and L. casei BFE 688 showed aggregation values of less than 20% (Fig. 3), whereas the adherent reference strains L. johnsonii LA-1 (BFE 663, Fig. 3) and L. rhamnosus GG autoaggregated with values of more than 20%. Although these strains are both known to adhere well to intestinal epithelial cells, a clear difference in autoaggregation could be observed, as L. johnsonii BFE 663 showed much higher autoaggregation values, i.e., more than 80%, while those of L. rhamnosus GG were below 40% (Fig. 3). A variation in aggregation values obtained with acidic (ph 4.0) and neutral (ph 7.4) PBS was not noted. Only for L. johnsonii strains BFE 6154 and BFE 6128 was the autoaggregation enhanced when tested in their overnight supernatants, and it was noticeably higher than that observed in PBS at ph 4.0 (Fig. 3). Measurements of autoaggregation on neutralized supernatants was not possible because most of the strains continued growing, which led to an increase in A 580 nm, resulting in negative autoaggregation values (data not shown). Coaggregation with foodborne pathogens. E. coli ATCC and Salmonella Typhimurium ATCC were able to autoaggregate with scores of 4 and 3, respectively. When lactobacilli were present, the scores were reduced by at least one unit in all cases, except for L. johnsonii BFE 6128 and BFE 6154 strains, which coaggregated with Salmonella Typhimurium with a score of 3 (Table 2). After Gram staining, it was verified by light microscopy

6 130 VIZOSO PINTO ET AL. J. Food Prot., Vol. 70, No. 1 TABLE 2. Coaggregation of potential probiotic Lactobacillus strains with intestinal pathogens a Strain L. monocytogenes Scott A S. aureus ATCC E. coli ATCC Salmonella Typhimurium ATCC L. plantarum BFE BFE BFE BFE BFE L. johnsonii BFE BFE BFE L. rhamnosus GG L. casei Shirota BFE L. paracasei BFE L. plantarum ATCC L. monocytogenes Scott A 2 S. aureus ATCC E. coli ATCC Salmonella Typhimurium ATCC a Plates were observed with an inverted microscope with a 20 objective. Scoring system adopted for the coaggregation assay; 0, no clumps; 1, small and dispersed clumps in some microscopic fields; 2, medium-sized and dispersed clumps; 3, abundant and medium-sized clumps with turbid supernatant; 4, very big clumps with a clear supernatant. that aggregates were composed of gram-positive as well as gram-negative bacteria. Three probiotic reference strains (L. rhamnosus GG, L. casei Shirota, and L. paracasei BFE 675) as well as L. plantarum ATCC 8014 and L. johnsonii BFE 6128 and BFE 6154 were able to coaggregate with L. monocytogenes Scott A with a higher score than that obtained when L. monocytogenes autoaggregated (Table 2). Some variation was observed for the coaggregation of L. plantarum strains with S. aureus ATCC Two of the reference probiotic strains (L. johnsonii LA-1 and L. casei Shirota) did not coaggregate at all, whereas L. johnsonii BFE 6128 and BFE 6154 coaggregated with S. aureus with a score of 4 (Table 2). When the spent culture supernatant of overnight cultures of lactobacilli was added to the coaggregation system (1:10) with L. monocytogenes and Salmonella Typhimurium, the coaggregation with L. plantarum strains BFE 5878, BFE 5092, BFE 5759, BFE 1685, and BFE 1684 and L. johnsonii BFE 6128 increased by one unit (data not shown). In that way, L. johnsonii BFE 6128 reached the highest score (4) with these foodborne pathogen strains. HT29 cell viability in coculture experiments. None of the tested bacterial strains significantly modified (P 0.05) the vitality of HT29 cells after a 24-h incubation FIGURE 4. Induction of IL-8 secretion in HT29 epithelial cells by lactobacilli. An inoculum of CFU per well was given to HT29 cell monolayers grown in six-well plates and incubated for 24 h. Supernatants were collected and centrifuged, and IL-8 levels were determined by ELISA. Values are means standard errors of at least three separate experiments. * Significantly different from the control (P 0.01, ANOVA, Dunnett test). when compared with the negative control (cells cultivated with DMEM alone), as assessed by the release of lactate dehydrogenase into the supernatant. Bacterial lactate dehydrogenase was not released into the medium, as corroborated with supernatants of bacteria grown on DMEM without HT29 cells. The same results were obtained with the MTT assay in separate experiments (data not shown). Modulation of cytokine secretion in HT29 cells. Ten strains were tested for their ability to induce cytokine secretion in intestinal epithelial cells (Fig. 4). IL-1, IL-6, MCP-1, IL-10, and TNF- could not be determined in the supernatants treated with either bacteria or LPS (0.1 g/ ml). TGF- was detected; however, it was secreted constitutively, and no significant (P 0.01) difference could be demonstrated between the 24-h supernatant of the control (cells grown with DMEM without serum and antibiotics) and the probiotic-treated samples ( CFU/ml in DMEM without FCS and antibiotics). Two L. plantarum strains (BFE 5759 and BFE 1685) and one L. johnsonii strain (BFE 6128) significantly (P 0.01) induced the secretion of IL-8 when compared with the control without bacteria (Fig. 4). Nevertheless, when cells were incubated for 24 h with LPS (0.1 g/ml), the levels of this chemokine were much higher ( pg/ml; data not shown). The two well-known probiotics L. rhamnosus GG and L. paracasei BFE 675 (Actimel, Danone), on the other hand, down-regulated the release of IL-8, as shown in Figure 4. The other Lactobacillus strains neither reduced nor enhanced the secretion of this chemokine by HT29 cells significantly. DISCUSSION Bacterial attachment to the intestinal mucosa is considered a very important criterion for selection of probiotic strains, as it allows bacterial strains to at least prolong their transit time through the intestine, permitting them to exert their beneficial interactions and possibly to colonize the intestine (2, 27). Intestinal epithelial cell lines have been extensively used to perform comparative studies of adhesion properties of probiotic strains (2, 26, 27). Although mucus-

7 J. Food Prot., Vol. 70, No. 1 LACTOBACILLUS SPP. ADHESION AND CHEMOKINE STIMULATION 131 producing epithelial cells in tissue culture are available, in our studies, we wanted to investigate the cell-cell interaction of eukaryotic cells with bacteria from two different points of view, i.e., the binding interaction and the elicitation of a cytokine response as a result of this interaction. For this reason, mucus-secreting cells were not used, as is the case in most other binding studies (2, 7, 26, 27), although the role of the mucus in inhibiting the binding of bacteria to the epithelial cells should not be forgotten. There are several direct and indirect methods for studying adhesion potential in vitro, but there is still no consensus about the optimal method that can actually predict microbial adhesion in vivo (2). Nevertheless, it is widely accepted that the level of adhesion of a probiotic strain under a given assay condition (often expressed as the percentage of adhesion) does not constitute an absolute value and has to be evaluated in relation to the nonadherent strains tested under the same conditions. In this and other studies (44, 45), the L. casei strain Shirota, known to adhere weakly to intestinal epithelial cells, was used as a negative control. Another relevant point when comparing the binding of different strains is to maintain the amount of bacteria added to the system at similar levels, because the number of bound bacteria is influenced by the number of bacteria added to the wells (45). Both factors, i.e., the inclusion of a negative control and the maintenance of the same inoculum of probiotic bacteria or probiotic candidates (as determined by plating), were considered in the adhesion assay in this study. Not all the strains tested here could be described as adherent strains. However, two L. johnsonii strains (BFE 6128 and BFE 6154), isolated from Maasai milk, were strongly adherent, as shown by percent values that were much higher than those observed for known commercial probiotic strains (L. rhamnosus GG and L. johnsonii LA- 1). The values of the microbial adhesion to solvents test obtained with n-hexadecane reflect the hydrophobicity of the bacterial surface, whereas the values obtained with chloroform and ethyl acetate can be regarded as a measure of the electron-donor (basic) and electron-acceptor (acidic) characteristics of the cell walls, respectively (1). These two L. johnsonii strains showed a high affinity for both the hydrophobic solvent n-hexadecane and the polar solvents chloroform and ethyl acetate used in the assays for microbial adhesion to solvents. This indicates that the surface of these strains is able to interact simultaneously with charged (hydrophilic) and noncharged (hydrophobic) molecules. It appears, therefore, that this bivalent nature of the cell surface of these L. johnsonii strains may point toward a high adhesion to intestinal epithelial cells, as indicated by our adhesion assay results. However, adsorption to surfaces is a nonselective process that only partially explains the phenomenon of adhesion. Nevertheless, it is still important, as it brings two surfaces close enough to permit possible adhesins and cell receptors to interact with each other. A great variability in adhesion to HT29 cells was found among L. plantarum strains. One of the strains (L. plantarum BFE 5759) was nonadhesive, another was moderately adhesive (L. plantarum ATCC 8014), and four strains (L. plantarum BFE 5878, BFE 5092, BFE 1685, and BFE 1684) were strongly adhesive. Strains that showed poor adhesion to HT29 cells (L. paracasei BFE 675, L. casei Shirota BFE 688, and L. plantarum BFE 5759) showed variable values of adhesion to n- hexadecane and chloroform, but all had low adhesion values to ethyl acetate. The nonadhesive Lactobacillus strains in this study, therefore, appeared to have a weak acidic surface character characterized by an affinity for a basic solvent (ethyl acetate), a variable hydrophobicity, and a moderate-to-strong basic character (characterized by an affinity for an acidic solvent such as chloroform). Pelletier et al. (36) reported similarities among the physicochemical surface properties of strains of the same species and suggested the use of these properties in a taxonomic perspective for microbial classification. We disagree with this suggestion, as important variations in the surface properties of bacteria of the same species (e.g., L. plantarum) were shown to occur in this study. Strains isolated from traditional fermented products were shown to be capable of adhering to intestinal epithelial cells (HT29) in our study, indicating that not only human isolates possess this property. Strains of the same species showed different adhesion capabilities, confirming that adhesion is a strain-specific property, as has been suggested before (21, 43). For L. johnsonii BFE 6128 and BFE 6154, this could be because they not only adhered individually, as for L. johnsonii LA-1, but also that they formed big clusters. Some of the L. plantarum strains adhered individually, and others adhered in pairs or short chains. In this study, we showed that strains of L. johnsonii autoaggregated well and also adhered to intestinal epithelial cells. Strains of L. plantarum generally autoaggregated to a lesser extent than L. johnsonii strains. While most of the L. plantarum strains (BFE 1684, BFE 1685, BFE 5878, and BFE 5092) adhered well to HT29 cells, L. plantarum strain BFE 5759 was nonadherent, even though its autoaggregation ability was similar to the other highly adherent L. plantarum strains. The nonadherent strains L. paracasei BFE 675 and L. casei Shirota (BFE 688) also did not autoaggregate well. Del Re et al. (7) stated that the autoaggregation of Bifidobacterium spp. is strongly related to adhesion. Generally, our results also suggested that the majority of bacteria able to autoaggregate also adhered well to epithelial cells, although not for L. plantarum strain BFE Thus, it may be speculated that autoaggregation is indeed related to adhesion, but this relationship may not be exclusive, because exceptions were noted. Since a general trend between autoaggregation and the level of adherence to intestinal epithelial cells was suggested by our results, the autoaggregation test may serve as an indicator of potentially probiotic strains with good adherence ability in initial screening experiments. However, since we clearly showed that there might be exceptions to this indicative ability, adherence studies in cell cultures remain of great importance and cannot be replaced by autoaggregation tests. Apart from possibly serving as an indicator to adherence ability, the autoaggregation phenotype is an interesting probiotic property, as it plays an important role in the col-

8 132 VIZOSO PINTO ET AL. J. Food Prot., Vol. 70, No. 1 onization of the oral cavity (23) and the urogenital tract (4) as well as in the gastrointestinal persistence of the organisms in vivo (6). For L. johnsonii strains BFE 6154 and BFE 6128, autoaggregation was enhanced when they were tested in their own overnight supernatant (at ph values of approximately 4.0), and it was considerably higher than that observed in PBS at ph 4.0. This indicated that autoaggregation was not a ph-dependent effect but relied on the presence of an autoaggregating factor in the supernatant. It has also been hypothesized that one of the mechanisms through which lactic acid bacteria protect the host from infections is their ability to coaggregate with intestinal pathogens and uropathogenic bacteria (20). A strong inclination to autoaggregation does not imply a strong coaggregation property, but it has been observed that strains with high coaggregation ability also show high autoaggregation (9). This was also supported by our findings, because the strains with the highest coaggregation scores also autoaggregated strongly. However, exceptions were noted, because L. johnsonii BFE 663, which had high autoaggregation values, did not coaggregate with S. aureus and only weakly coaggregated with L. monocytogenes. Previous coaggregation studies have been limited to bacteria from human origin. Schachtsiek et al. (40) were the first to report the coaggregation of Lactobacillus coryniformis, a food- or feed-associated bacterium, with pathogens. Thus, L. coryniformis was able to coaggregate with E. coli K-88, Campylobacter jejuni, and Campylobacter coli but not with Salmonella Typhimurium, Clostridium perfringens, or L. monocytogenes. To our knowledge, we report in the present study the first L. johnsonii strain of food origin able to coaggregate with human pathogens (L. monocytogenes, S. aureus, E. coli, and Salmonella Typhimurium), even to a higher degree than the commercial probiotic strains L. rhamnosus GG, L. johnsonii LA-1 (BFE 663), and L. casei Shirota (BFE 675). Strains with this property could be of special interest to the food industry, because coaggregates may be formed in the food matrix and thereby prevent the entrapped pathogens from adhering to host cells (40). It is expected that a sufficient amount of a probiotic cell suspension must be consumed by the host to enable the probiotic to exert its beneficial effect (26). It has also been shown that some probiotic strains able to inhibit the adhesion of pathogens do so only when present at a higher concentration than the pathogen itself (26, 31). For these reasons, we also used higher concentrations of lactobacilli in the adhesion inhibition assays. Corresponding to a previous report for a model with Caco-2 cells (28), L. rhamnosus GG did not inhibit Salmonella Typhimurium adhesion to HT29 cells in this study. Two strongly adherent strains, L. plantarum BFE 1685 and L. johnsonii BFE 6128, however, were able to significantly reduce the adhesion of the human pathogens E. coli and L. monocytogenes in HT29 cell culture. In addition, L. plantarum BFE 1685 reduced the adhesion of the E. faecium and Salmonella Typhimurium strains used in this study. We also observed a slight inhibition of Salmonella Typhimurium ATCC by the nonadhesive L. casei Shirota, which seems not to depend on competitive exclusion for adhesion sites, because L. casei adheres poorly to HT29 cells. Instead, it might depend on another antagonistic mechanism. The spectrum and magnitude of the inhibition of adhesion, as well as the autoaggregation and coaggregation properties of L. plantarum or L. johnsonii, suggest that the mechanisms involved in the inhibition of adhesion are different. More studies are needed to elucidate if this is due to unspecific steric hindrance or a specific mechanism involving adhesins. Adhesion of pathogenic bacteria to mucosal surfaces is considered the first step of intestinal infections (10). Therefore, lactobacilli, such as the strains selected in this study, that inhibit the adhesion of certain pathogens to intestinal epithelial cells are very interesting probiotic candidates. The intestinal mucosa is the main interface between the immune system and the intestinal lumen. It not only functions as a physical barrier, which is constantly exposed to exogenous compounds, such as food, microorganisms, and their metabolites, but can also generate signals for communication with underlying mucosal immune cells (16, 39). Thus, antigen processing and signaling by epithelial cells are primary steps of innate and adaptive defense mechanisms in the earliest phases after invasion by pathogens. Nonpathogenic bacteria also elicit a response, but their mechanism of immune stimulation is even less elucidated. There are contradictory reports about the spectrum of cytokines secreted by intestinal epithelial cells (41). According to our view, this reflects that studies reporting the secretion of IL-10 included intestinal mucosa isolated from biopsies. Such intestinal mucosa not only contains epithelial cells, but also different immune cell populations from the epithelial lining and the lamina propria, which are difficult to separate from the thin enterocyte layer (41). Vidal et al. (47) detected very low levels of TNF- (e.g., 14 pg/ml) in the supernatants of LPS-stimulated HT29 cells. In contrast, we could not detect TNF-, and this could be explained by the difference in the sensitivity of the detection assays used that had a detection limit of pg/ml. Alternatively, this could be explained by variations that may occur in a cell line itself and that depend on the degree of cell differentiation (3). IL-6 has been shown to be secreted by intestinal epithelial Caco-2 cells stimulated with bacteria (19), but IL- 6 was not detected in the supernatants of HT29 cells upon stimulation with Lactobacillus strains. To our knowledge, IL-6 secretion by HT29 cells has not been reported yet. Therefore, the absence of IL-6 in our study may be because of inherent variations from line to line in the repertoire and levels of cytokine secretion or to species- or strain-specific effects. Jung et al. (22) showed that HT29 cells secreted MCP-1 in response to gram-negative invasive pathogens and to L. monocytogenes, but not in response to LPS. None of the Lactobacillus strains of this study were able to induce the secretion of this cytokine. Under the conditions investigated in this study, we could detect variations only in the IL-8 levels secreted by HT29 cells when stimulated with certain lactic acid bacteria, showing that this effect was strain-specific. Thus, L. johnsonii BFE 6128 and L. plantarum BFE 1684 and BFE 5759 could enhance IL-8 secretion, but this was considered

9 J. Food Prot., Vol. 70, No. 1 LACTOBACILLUS SPP. ADHESION AND CHEMOKINE STIMULATION 133 a weak enhancement when compared to the stimulatory effect observed with LPS under the same conditions. A similarly weak IL-8 induction in cell culture was also previously reported for the nonpathogenic Bacillus subtilis (natto) (19) and Saccharomyces cerevisiae (39) strains isolated from foods. It is well known that IL-8 secreted by intestinal epithelial cells recruits and activates neutrophils and monocytes in the adjacent lamina propria. This might lead to a proinflammatory state, which could be harmful when there is already a background of inflammation, as, for example, with intestinal bowel disease. On the other hand, in healthy individuals, such a weak immune stimulation might result in keeping the immune system alert and in a condition of readiness to react with an inflammatory response if challenged by pathogenic bacteria (39). However, the cytokine responses of epithelial cells appear to be influenced by immune cells (16). Therefore, the modulation of a further immune reaction, which involves innate and adaptive components, is generally undertaken by professional immune cells of the lamina propria. Accordingly, it is hard to predict the outcome of the observed IL-8 stimulation on the further development of an immune response when such interactions are tested in cell culture in the absence of immune cells of the lamina propria. As an example, the probiotic strain E. coli Nissle 1917, which exhibits anti-inflammatory properties in animal studies, actually induced proinflammatory IL-8 production by HT29 cells in vitro (25). Therefore, our data strongly suggest that enterocytes interacted specifically with L. plantarum BFE 1685 and BFE 5759 and L. johnsonii BFE 6128 strains without assistance from other specialized cells of the immune system. However, the further course of this immune stimulation should be determined in animal experiments or human volunteer studies. In conclusion, the L. plantarum strain BFE 1685 and L. johnsonii strain BFE 6128 studied here displayed interesting probiotic properties. In a previous investigation, these strains also survived under simulated gastrointestinal conditions, inhibited the growth of pathogens in vitro, and showed bile salt hydrolase activity (48). In this study, it was shown that these strains were highly adherent to HT29 cells and prevented the adhesion of intestinal pathogens. Measurements of adhesion ability in vivo require animal experiments or human volunteer studies, which are not only expensive but are fraught with strict ethical regulation, making these types of measurements difficult for routine investigation. In this study, the in vitro adherence assays were performed to select potentially in vivo adhesive strains, as previous studies indicated a relationship between in vitro adhesion assays and in vivo colonization ability (6, 21). In addition, they showed potential for aggregation with specific pathogens and were capable of cytokine stimulation of HT29 cells. Therefore, these strains appear to fulfill important requirements for probiotics. Clearly, however, the colonization ability and immune stimulatory properties should be validated in further in vivo studies. ACKNOWLEDGMENTS We thank Mr. L. Korn for his invaluable advice in statistics. M. G. Vizoso Pinto acknowledges the Konrad-Adenauer-Stiftung for a Ph.D. scholarship. REFERENCES 1. Bellon-Fontaine, M. N., J. Rault, and C. J. van Oss Microbial adhesion to solvents: a novel method to determine the electron-donor/electron-acceptor or Lewis acid-base properties of microbial cells. Colloids Surf. 7: Blum, S., R. Reniero, E. J. Schiffrin, R. Crittenden, T. Mattila-Sandholm, A. C. Ouwehand, S. Salminen, A. von Wright, M. Saarela, M. Saxelin, K. Collins, and L. Morelli Adhesion studies for probiotics: need for validation and refinement. Trends Food Sci. Technol. 10: Boecker, U., A. Schottelius, J. M. Watson, L. Holt, L. L. Licato, D. A. Brenner, R. B. Sartor, and C. Jobin Cellular differentiation causes a selective down-regulation of IL-1 mediated NF-kB activation and IL-8 gene expression in intestinal epithelial cells. J. Biol. Chem. 275: Boris, S., J. E. Suarez, F. Vazquez, and C. Barbes Adherence of human lactobacilli to vaginal epithelial cells and interactions with uropathogens. Infect. Immun. 66: Briandet, R., T. Meylheuc, C. Maher, and M. N. Bellon-Fontaine Listeria monocytogenes Scott A: cell surface charge, hydrophobicity, and electron donor and acceptor characteristics under different environmental growth conditions. Appl. Environ. Microbiol. 65: Cesena, C., L. Morelli, M. Alander, T. Siljander, E. Tuomola, S. Salminen, T. Mattila-Sandolm, T. Vilpoen-Salmela, and A. von Wright Lactobacillus crispatus and its nonaggregating mutant in human colonization trials. J. Dairy Sci. 84: Del Re, B., B. Sgorbati, M. Miglioli, and D. Palenzola Adhesion, auto-aggregation and hydrophobicity of 13 strains of Bifidobacterium longum. Lett. Appl. Microbiol. 31: Doyle, R. J Contribution of the microbial effect to microbial infection. Microbes Infect. 2: Ehrmann, M. A., P. Kurzak, J. Bauer, and R. F. Vogel Characterization of lactobacilli towards their use as probiotics adjuncts in poultry. J. Appl. Microbiol. 92: Finlay, B. B., and S. Falkow Common themes in microbial pathogenicity revisited. Microbiol. Mol. Biol. Rev. 61: Fuller, R Probiotics in plants and animals. J. Appl. Bacteriol. 66: Fuller, R., and G. Perdigón Conclusions, p In R. Fuller and G. Perdigón (ed.), Probiotics 3. Immunomodulation by the gut microflora and probiotics. Kluwer Academic Publishers, Dordrecht, Germany. 13. Gopal, P. K., J. Prasad, J. Smart, and H. S. Gill In vitro adherence properties of Lactobacillus rhamnosus DR20 and Bifidobacterium lactis DR10 strains and their antagonistic activity against an enterotoxigenic Escherichia coli. Int. J. Food Microbiol. 67: Gordon, D. C., W. A. Simpson, and E. H. Beachey Adhesion of bacteria to animal tissues, p In D. C. Savage and M. Fletcher (ed.), Bacterial adhesion. Plenum Press, New York. 15. Gusils, C., A. Pérez-Chaia, S. González, and G. Oliver Lactobacilli isolated from chicken intestines: potential use as probiotics. J. Food Prot. 62: Haller, D., C. Bode, W. P. Hammes, A. M. A. Pffeifer, E. J. Schiffrin, and S. Blum Non-pathogenic bacteria elicit a differential cytokine response by intestinal epithelial cell/leukocyte co-cultures. Gut 47: Holzapfel, W. H., P. Haberer, R. Geisen, J. Björkroth, and U. Schillinger Taxonomy and important features of probiotic organisms in food and nutrition. Am. J. Clin. Nutr. 73:365S 373S. 18. Holzapfel, W. H., and U. Schillinger Introduction to pre- and probiotics. Food Res. Int. 35: Hosoi, T., R. Hirose, S. Saegusa, A. Ametani, K. Kiuchi, and S. Kaminogawa Cytokine responses of human intestinal epithelial-like Caco-2 cells to non-pathogenic bacterium Bacillus subtilis (natto). Int. J. Food Microbiol. 82: Huis in t Veld, J. H., R. Havenaar, and P. Marteau Establishing a scientific basis for probiotics R&D. Trends Biotechnol. 12: Jacobsen, C. N., N. Rosenfeldt, A. E. Hayford, P. L. Moller, K. F.