Comparison of the Immunomodulatory Properties of Three Probiotic Strains of Lactobacilli Using Complex Culture Systems: Prediction for In Vivo Efficacy

Background While the use of probiotics to treat or prevent inflammatory bowel disease (IBD) has been proposed, to this point the clinical benefits have been limited. In this report we analyzed the immunological activity of three strains of Lactobacillus to predict their in vivo efficacy in protecting against experimental colitis. Methodology/Principal Findings We compared the immunological properties of Lactobacillus plantarum NCIMB8826, L. rhamnosus GG (LGG), L. paracasei B21060 and pathogenic Salmonella typhimurium (SL1344). We studied the stimulatory effects of these different strains upon dendritic cells (DCs) either directly by co-culture or indirectly via conditioning of an epithelial intermediary. Furthermore, we characterized the effects of these strains in vivo using a Dextran sulphate sodium (DSS) model of colitis. We found that the three strains exhibited different abilities to induce inflammatory cytokine production by DCs with L. plantarum being the most effective followed by LGG and L. paracasei. L. paracasei minimally induced the release of cytokines, while it also inhibited the potential of DCs to both produce inflammatory cytokines (IL-12 and TNF-α) and to drive Th1 T cells in response to Salmonella. This effect on DCs was found under both direct and indirect stimulatory conditions – i.e. mediated by epithelial cells - and was dependent upon an as yet unidentified soluble mediator. When tested in vivo, L. plantarum and LGG exacerbated the development of DSS-induced colitis and caused the death of treated mice, while, conversely L. paracasei was protective. Conclusions We describe a new property of probiotics to either directly or indirectly inhibit DC activation by inflammatory bacteria. Moreover, some immunostimulatory probiotics not only failed to protect against colitis, they actually amplified the disease progression. In conclusion, caution must be exercised when choosing a probiotic strain to treat IBD.


Introduction
The intestine is home to trillions of commensal bacteria that participate in digestive functions while helping to protect the host from the aggression of pathogens [1]. Commensals are not ignored by the immune system, rather they are tolerated via a concerted action of epithelial cells and immune cells [2][3][4]. The interaction of commensals with pattern recognition receptors on the apical surface of epithelial cells is protective against colitis [5][6][7][8], but, commensals can also be colitogenic as several susceptible mice when reared under germ free conditions do not develop colitis [9].
Given this dual role of bacteria -protective versus colitogenicthe use of probiotics as therapeutic agents in IBD has been proposed [10][11][12]. Probiotics can be considered as those microorganisms that are beneficial to the host. According to an FAO/WHO joint report 'there is good evidence that specific strains of probiotics are safe for human use and able to confer some health benefits on the host, but these benefits cannot be extrapolated to other strains without experimentation' [13].
Indeed, given the heterogeneity of probiotics and the two-faced character of bacteria as colitogenic or protective, it is clear that not all microorganisms may have the same effect on the host. Instead, it seems likely that each probiotic species, and within each species each strain, may have distinct activities. In agreement with this, the therapeutic effect of probiotics in patients with IBD depends on the strain of probiotic that is used, on the stage of the disease, and on the analyzed pathology (reviewed in [12]). Hence the correct use of probiotics as therapeutic agents calls for a precise knowledge of their activity. Table 1 encapsulates the previously published studies detailing the activity of different probiotics in protecting animals against experimental colitis. From this it can be appreciated that the activity of certain strains depends on the colitis model used. For instance, in rats, Lactobacillus GG (from now on called LGG) has been shown to have no protective effect against dinitrobenzene sulfonic acid (DNBS)-or trinitrobenzenesulfonic acid (TNBS)induced colitis [14,15], a partially protective effect in the iodoacetamide colitis model [15], and a detrimental effect in the Reduced weight loss, colonic damage, inflammatory cytokine expression in colon. [54] F. prausnitzii Mice: TNBS Yes Antinflammatory effect: able to block NF-kappaB activation and IL-8 production on Caco-2 cells. [20] Reduction of colonic inflammatory cytokines.
LGG or mixture of S. thermophilus, L. acidophilus, and B. longum

Mice: DSS
Yes Improved the DAI, reduced bacterial translocation, and reduced inflammation [17] L. salivarius 433118 DSS model [16]. It is also interesting to note that within the same species, different strains can behave differently. For instance, in mice, Lactobacillus plantarum DSM 15313 [17] has been shown to partially protect against DSS colitis while the strains ACA-DC 287 [18] or Lp115 [19] have no effect in TNBS colitis. The mechanisms of action of probiotics can be quite disparate. They have been shown to modulate the permeability of epithelial barriers, alter the inflammatory potential of epithelial cells, compete with pathogens for mucosal colonization, or directly modify the activity of immune cells (reviewed in [10,11]). However, the principal mechanism of protection against colitis is via a reduction in the release of pro-inflammatory cytokines, but the pathways and cells involved in this mechanism are not yet clear. Some probiotics like Faecalibacterium prausnitzii (whose reduction in the normal gut microbiota has been associated with Crohn's disease (CD) recurrence [20]) can inhibit NF-kB activation and IL-8 production by Caco-2 cells and reduce colonic inflammatory cytokines in colitic mice [20]. In other cases, probiotics like Lactobacillus casei or Lactobacillus gasseri expressing a manganese superoxide dismutase (MnSOD) can reduce inflammation via the inhibition of neutrophil recruitment [21,22]. Another mechanism of action of probiotics has been described for a mixture of eight probiotics named VSL#3 [23] or for Lactobacillus plantarum DSM 15313 [17], which relies upon an attenuation of the increased epithelial barrier permeability caused by DSS and the resulting bacterial translocation. In addition, VSL#3 can induce the differentiation of protective T regulatory cells [24]. This activity of probiotics may be exerted directly on dendritic cells (DCs), which are professional antigen presenting cells, as probiotic-loaded DCs can modulate the TNBS-induced inflammatory reaction through the differentiation of protective T regulatory cells [25].
DCs are pivotal in the initiation of adaptive immune responses and can directly contact and internalize intestinal bacteria [26][27][28]. Further, DCs can receive tissue conditioning by intestinal epithelial cells that control the DC inflammatory potential [29][30][31]. Hence, probiotics can interact either directly with DCs or indirectly, via the action of epithelial cells. In this manuscript we used a technique established in our laboratory to evaluate the ability of three Lactobacilli strains (plantarum, LGG and paracasei B21060) to activate DCs either directly or indirectly via the action of epithelial cells.
LGG and L. plantarum were chosen in order to compare already published reports while L. paracasei has been chosen for its capacity to survive the gastrointestinal tract, temporarily associate with the intestinal wall in humans [32,33] and inhibit T cell proliferation [34]. We found that the activity of the three probiotics was very different. L. paracasei the more immunomodulatory among the three strains, was able to inhibit the inflammatory potential of pathogenic Salmonella and protect against experimental colitis.

Probiotics induce the phenotypical maturation of DCs
The first outcome of the encounter of DCs with bacteria is the upregulation of surface activation markers. We generated DCs from human peripheral blood monocytes (MoDCs) of 4 different donors to assess donor-associated variability. MoDCs were incubated for 1 h with live logarithmic-phase Lactobacilli (L. plantarum, L. paracasei and LGG) or with Salmonella typhimurium in medium without antibiotics at a 10:1 (bacteria:DC) ratio. Cells were extensively washed and the medium was changed to one containing antibiotics. Cells were tested 24 hours later for upregulation of MHC II (HLA-DR) and costimulatory (CD80) molecules. Consistent with a recent report on bone marrowderived DCs [35], the three probiotic strains induced a similar upregulation of HLA-DR and CD80, which was however reduced as compared to the upregulation induced by Salmonella ( Fig. 1A-B, Figure S1). Given the low multiplicity of infection and the short exposure time, the viability of the cells (as seen by the frequency of propidium iodide (PI) negative and annexin V negative cells), was similar after each probiotic treatment, while Salmonella was partially toxic (Fig. 1C).

Probiotics have a different ability to induce cytokine production by DCs
Phenotypical activation of DCs does not necessarily correlate with their functional activation [36] and the type of cytokines released can have an impact on T cell polarization. Therefore, we analyzed the production of IL-12p70, IL-10, TNF-a and IL-12p40 by MoDCs after 24 h treatment with bacteria, as above. Salmonella was a strong inducer of all of the tested cytokines, while the three Lactobacilli elicited differential cytokine release ( Fig. 2A). L. plantarum and LGG induced a cytokine response that was very similar to that of Salmonella, while L. paracasei induced lower levels of IL-12p70, TNF-a and IL-10 when compared to Salmonella. Thus, the only strain displaying a reduced inflammatory potential was L. paracasei.
It has been shown that IL-10 can negatively regulate the expression of IL-12p70 [37]. To test whether an early increase in IL-10 release could impact on IL-12 production, we analyzed IL-10 release during the initial phases of DC activation (4-6 hours). The levels of IL-10 were not higher in any of the cultures of DCs with Lactobacilli during the times at which IL-12p70 was low, thereby suggesting IL-12p70 induction is delayed in comparison to Salmonella and is not controlled by IL-10 ( Fig. 2B).

The difference in cytokine production reflects different T cell polarizing ability
Cytokine release by DCs is important to drive the polarization of T cells towards Th1, Th2, Th17 or T regulatory cells. Given the differences observed in cytokine production we analyzed the capacity of bacteria-treated DCs to activate and polarize T cells. DCs were incubated with live bacteria and then cultured with highly purified allogeneic naïve CD4 + CD45RA + T cells. As shown in Figure 3A, all three Lactobacilli were less potent in inducing T cell proliferation when compared to Salmonella, probably reflecting their reduced ability to upregulate surface activation markers ( Fig. 1A-B). When we analyzed the cytokines produced by T cells we found that T cells activated with paracasei-treated DCs were affected in their ability to release IFN-c, IL-2, IL-10 and IL-6 ( Fig. 3B). In contrast, L. plantarum-treated DCs activated T cells similarly to S. typhimurium-treated cells in terms of IFN-c release but induced less IL-10, while LGG-treated DCs induced the opposite, more IL-10 and less IFN-c (Fig. 3B). There was no difference in IL-17 production (p.0.05). Some probiotics (i.e. L. casei and reuteri) can induce the development of T regulatory cells [24,25,38]. However, the different Lactobacilli and Salmonella displayed a similar ability to drive CD25 + Foxp3 + T regulatory cells ( Figure S2).

L. paracasei is less inflammatory on epithelial cells
It is likely that the first cells interacting with the intestinal flora are the epithelial cells (ECs) that line the intestinal wall, especially in IBD patients which display a reduced mucous layer [39]. For this reason, we studied the response of monolayers of ECs to probiotics (5610 7 CFU/TW) incubated from the apical (luminal) side. One hour after incubation, bacteria were washed out and medium was changed with one containing antibiotics (gentamycin 100 mg/mL). The different bacteria had no effect on the viability of the cells (not shown). We analyzed the release of IL-8, a chemokine involved in the recruitment of neutrophils, and of thymic stromal lymphopoietin (TSLP) and TGF-b, cytokines shown to inhibit the inflammatory potential of DCs [30,31]. We found that while Salmonella induced the release of IL-8 as early as 4 hours after bacterial treatment, all tested Lactobacilli induced negligible IL-8 release (Fig. 4). This probably reflects their inability to invade epithelial cells. L. paracasei displayed a greater ability to induce the release TGF-b whose levels were statistically higher than basal levels only at 4 and 6 h post incubation ( Fig. 4; p,0.05).
In addition, L. paracasei induced TSLP expression that peaked at 48 h from infection, while both L. plantarum and LGG induced a very transient production of TSLP. Salmonella induced sustained TSLP release that did not reach the level of L. paracasei (Fig. 4). Interestingly, TSLP mRNA levels were sustained in ECs treated with all bacteria ( Figure S3), but the meaning of this observation remains to be understood. One could speculate that TSLP protein is either not released at later time points, or is easily degraded in response to LGG and L. plantarum. By contrast, L. paracasei and Salmonella incubation may lead to increased release and/or stabilization of the protein.

L. paracasei inhibits the inflammatory potential of DCs
Having shown that L. paracasei was the least inflammatory among the three Lactobacilli strains, we focused on this strain for further experiments. We utilized three different conditions involving the interaction between bacteria, epithelial cells and DCs. DCs were either incubated with: a. L. paracasei (LP) and Salmonella (SL) individually or together; b. EC supernatant for 24 h and then subsequently with each bacteria; c. supernatants of ECs pre-incubated for 24 h with L. paracasei (Sn caco LP) on the apical side and then (24 h later) with each bacterial strain. As shown in Fig. 5, L. paracasei had a strong anti-inflammatory effect on DCs both when directly co-incubated with Salmonella and indirectly when supernatants of LP-treated ECs were incubated with DCs before exposure to Salmonella. The co-incubation of DCs with LP and Salmonella significantly reduced the ability of Salmonella to induce IL-12p70 and TNF-a, while not altering its ability to    promote IL-10 and IL-6 production (Fig. 5). A similar scenario was observed when DCs were first incubated with supernatants of LPtreated ECs and then infected with Salmonella. However, as we have already described [29], the incubation of DCs with unconditioned EC supernatant also reduced the ability of DCs to release IL-12p70 but not TNF in response to Salmonella. Therefore, the exposure of ECs to LP strongly inhibited the inflammatory response of Salmonella on DCs by inhibiting both IL-12p70 and TNF-a release (Fig. 5). This effect may be mediated either by LP-induced release of anti-inflammatory mediator/s by ECs, or by some component of L. paracasei that is translocated across the monolayer. The involvement of whole LP translocated across the monolayer is unlikely, as we could not detect intact bacteria from the basolateral side (not shown) and supernatants were filtered before incubation with the DCs. We could not detect any effect of LP on IL-6 or IL-10 (Fig. 5).
We then analyzed whether the factor(s) involved in the antiinflammatory effect was a soluble mediator and could be found in the culture supernatant of L. paracasei. DCs were coincubated with Salmonella and either L. paracasei or its culture supernatant (sn LP: derived from the same amount of CFUs used for DC incubation). Interestingly, the LP supernatant alone (7% volume/volume of tissue culture medium) was extremely efficient in inhibiting the DC release of inflammatory cytokines while it was unable to alter the ability of DCs to release IL-10 or IL-12p40 (Fig. 6). When LP was extensively washed before incubation with DCs it lost the ability to inhibit the DC release of inflammatory cytokines in response to Salmonella (Fig. 6), suggesting that the anti-inflammatory effect of L. paracasei is dependent on a soluble metabolite or mediator. It is likely that this mediator(s) is not released during the limited time of LP in culture with the DCs as we could not detect LP growth during the 1 h incubation time with DCs (most likely due to the aerobic culture conditions, not shown). Further diluting the LP culture supernatant 1 to 5 (1.4% volume/volume) but also 1 to 10 (0.7% volume/volume) was still able to inhibit the release of IL-12p70 and TNF-a, indicating the high efficacy of the soluble mediator(s) (Fig. 6).
Coincubation of Salmonella and L. paracasei affects the ability of DCs to activate Th1 T cells Having shown that the coincubation of DCs with LP and Salmonella (SL) drastically reduced the ability of DCs to release IL-12p70 while preserving IL-10 production we analyzed the ability of these DCs to polarize inflammatory T cells. We treated DCs with either LP, or SL or the two together. Cells were then incubated with highly purified naïve T cells and cytokine release in culture supernatants was tested. As shown in Figure 7 (see the situation a), T cells activated with DCs that were incubated with both Salmonella and L. paracasei were highly impaired in their ability to release IL-2, IL-10, IL-6 and IFN-c. We could not observe any difference in IL-17, IL-13 or IL-5 suggesting that LP+SL treated DCs were still capable of inducing Th17 or Th2 polarization (Fig. 7).
DCs incubated with supernatants of L. paracasei treated ECs are affected in their ability to drive Th1 T cells DCs incubated with supernatants of LP-treated ECs are affected in their ability to release IL12-p70 and TNF-a in response to Salmonella. Consequently, we evaluated whether this had an impact  on Th1 T cell polarization. To accomplish this, DCs were preincubated with supernatants from either untreated (sn caco: situation b) or paracasei-treated ECs (sn caco LP: situation c) for 24 h, and then with either LP or Salmonella for an additional 24 h before incubation with naïve T cells for 5 days. As shown in Fig. 7, the preincubation of DCs with LP-treated-EC supernatants prior to Salmonella infection, drastically reduced the DC's ability to activate T cells and drive their polarization to Th1 T cells as evidenced by a decrease in IFN-c, IL-2 and IL-6 production. There was no difference in IL-17 and IL-13 production (p.0.05) while IL-10 and IL-5 levels were also reduced in culture supernatants (Fig. 7). This indicates that the incubation of ECs with L. paracasei has a strong effect on the ability of DCs to activate T cells in response to Salmonella.

The in vitro activity of probiotics predicts their efficacy in vivo
We next compared the activity of the three Lactobacilli in protecting mice against an acute model of colitis. We chose the DSS colitis model as it provokes a strong inflammatory response that is primarily mediated by DCs [40]. Mice were pretreated i.g. for 7 days with 10 10 CFUs of either L. plantarum, LGG or L. paracasei, or with PBS as a control. Then mice received for 5 days 2% DSS in the drinking water and the development of colitis was followed over time by measurement of body weight, stool consistency and presence of blood in the feces. We found that L. plantarum and LGG, consistent with their ability to strongly activate DCs, were not only ineffective in protecting against colitis, but were in fact detrimental. Indeed, LGGand L. plantarum-treated mice displayed an increased disease activity index (DAI) and all died between 10 and 12 days from DSS administration (Fig. 8). In contrast, mice receiving L. paracasei although displaying a similar weight loss as PBS-DSS treated controls, showed a delay in colitis development and a reduced severity of disease (as shown by reduced DAI in Fig. 8B).

Discussion
IBD is a multifactorial disorder caused by both genetic and environmental factors. The host microflora contributes to disease development and perpetuation as colitis-susceptible mice reared under germ free conditions do not develop the disease and IBD patients treated with antibiotics experience amelioration of symptoms. Further, a colitogenic flora able to disseminate the inflammatory condition has been identified in mice [41]. Finally, it has been shown that a reduced proportion of mucosal associated Faecalibacterium prausnitzii is associated with a higher risk of postoperative recurrence of ileal CD [20]. Hence, a lot of effort has been undertaken to find ways to reestablish a protective intestinal flora. In this context, the use of probiotics as therapeutic or preventive agents in IBD has long been proposed, with however limited clinical efficacy [12]. Here we employed both complex in vitro culture systems and an in vivo mouse model to study the immunological effect and anti-inflammatory properties of three different probiotic strains: L. plantarum, LGG and L. paracasei. We show that Lactobacillus strains dramatically differed in their ability to activate DCs and to drive the polarization of T cells. In particular, we found that L. plantarum and LGG were more similar to Salmonella in terms of DC activation and T cell polarization than L. paracasei. The latter proved very poor in activating DCs and in inducing cytokine production (both inflammatory and noninflammatory). Furthermore, all Lactobacilli induced the development of CD25+Foxp3+ T regulatory cells similarly to Salmonella.
When we tested the capacity of the different Lactobacilli to activate epithelial cells we found that L. paracasei induced an increased release of TGF-b and TSLP. Both factors play an important role in driving the differentiation of non-inflammatory human DCs [30,31]. Moreover, L. paracasei affected the ability of DCs to respond to Salmonella typhimurium both directly (when coincubated with Salmonella) and indirectly (after treatment of DCs with supernatants from ECs which had been treated with L. paracasei upon their luminal side). Co-incubation of DCs with Salmonella and L. paracasei together strongly inhibited the release of IL-12p70 while leaving the release of IL-10 untouched. This resulted in reduced differentiation of Th1 T cells. Similarly, the incubation of DCs with supernatants of ECs treated with L. paracasei affected the release of IL-12p70 and TNF-a and the development of Th1 T cells. This suggests that once in the intestine, L. paracasei may act directly on intraepithelial DCs limiting their ability to induce inflammation even to potent inflammatory bacteria, but they may also indirectly inhibit the activity of DCs that do not enter directly in contact with paracasei, via conditioning of ECs. We do not yet know whether this effect is mediated by a soluble factor released by epithelial cells or by a bacterial-derived mediator transcytosed across the epithelial cells. Both possibilities are likely as we found that the direct effect of L. paracasei on DCs is indeed dependent on an LP released soluble mediator. The culture supernatant of L. paracasei was, by itself, sufficient to inhibit the release of IL-12p70 and TNF-a, but not IL-10, by DCs in response to Salmonella. This effect was very strong as it occurred even at very low concentrations of bacterial supernatant (0.7% of the tissue culture volume). Other examples of soluble mediators that are responsible for the anti-inflammatory activity of bacteria have been reported. For instance, Faecalibacterium prausnitzii also inhibits IL-8 production induced by IL-1b in Caco-2 cells via an unknown soluble mediator [20]. Similarly, Lactobacillus acidophilus exerts its activity via the interaction of the surface (S) layer A protein (SlpA) with the C-type lectin DC-SIGN on dendritic cells [42]. However, it is unlikely that an S layer protein is the mediator of the anti-inflamamtory activity of L. paracasei for two reasons. First, L. paracasei's activity is exclusively dependent on a soluble mediator while S layer proteins are also present on the surface of the bacterium. Second, while SlpA alone induces the release of IL-10, as expected through its activation of DC-SIGN, L. paracasei supernatant is unable to induce IL-10 production by DCs. Hence, it remains to be analyzed what is the nature of the soluble mediator released by L. paracasei and what is the mechanism of inhibition of IL-12p70 and TNF-a release while sparing IL-10 release by DCs. Therefore, we have described a new property of probiotics, which is to inhibit cytokine release and T cell activation by DCs in response to strong immunogens like Salmonella. The effect of L. paracasei on DC activation was not limited to Salmonella but was also active against immunogenic Lactobacilli (not shown), suggesting that L. paracasei may also inhibit the activation of the immune system in response to immunogenic commensal bacteria. This is likely the case as when we pretreated mice for 7 days with the three different Lactobacilli before exposure to DSS, only L. paracasei was protective. As DCs have been shown to regulate the outcome of DSS colitis [40] it is likely that the probiotic effect observed in vivo is dependent on the capacity of L. paracasei to inhibit the inflammatory potential of DCs.
Recently, it has been shown that the logarithmic growth phase of Lactobacilli is associated with the induction of anti-inflammatory genes [43]. While all of the bacterial preparations used in the current study were logarithmic, still we observed that LGG and L. plantarum NCIMB8826 were detrimental as they induced the death of DSS-treated mice, suggesting that these strains are inflammatory independently of their growth phase. A recent report has shown that the ratio between IL-10 and IL-12 production after treatment of human peripheral blood mononuclear cells with probiotics, was predictive of in vivo probiotic activity in the TNBS colitis mouse model [44]. The authors showed that a higher IL-10/IL-12 ratio correlated with better in vivo efficacy. Accordingly, in our study, based on MoDCs, we show a worse DAI in mice pretreated with plantarum that had a lower IL-10/IL-12 ratio than the other two strains (plantarum: 0,877; LGG: 1,222; paracasei: 1,176). However, the IL-10/IL-12 ratio would not explain the better outcome of L. paracasei versus LGG that displayed a similar cytokine ratio. It is important to note that L. paracasei was a poor inducer of cytokines (both inflammatory and non-inflammatory) and impacted on the ability of DCs to produce inflammatory cytokines in response to pathogens, suggesting that this could represent a new class of immunomodulatory probiotics that do not act via the induction of tolerogenic responses.
In conclusion, we show that probiotics should be divided into immunostimulatory and immunomodulatory according to their ability to interact with immune and non-immune cells, and their clinical use should be tailored accordingly. For instance, LGG, which is immunostimulatory has been shown to be more appropriate in the prevention of nosocomial rotavirus-dependent diarrhea in infants [45] or in decreasing the incidence of atopic dermatitis [46], than as an additive therapy in children with CD [47] or in CD patients after surgery [48]. In contrast L. paracasei, which is immunomodulatory, may be used to dampen inflammatory responses and may be recommended to maintain the remission phase in IBD. Hence, each probiotic strain should be characterized for their immune activity before being proposed for clinical applications.

Materials and Methods
Mice and Bacterial strains C57/BL6 mice were purchased from Charles River laboratories. All mice were maintained in microisolator cages in a specific pathogen-free animal facility. All experiments were performed in accordance with the guidelines established in the Principles of Laboratory Animal Care (directive 86/609/EEC) and approved by the Italian Ministry of Health.

Bacterial treatments and assessment of MoDC viability
MoDCs were incubated for 1 h with live logarithmic-phase Lactobacilli (L. plantarum, L. paracasei and LGG) or with Salmonella typhimurium in medium without antibiotics at a 10:1 (bacteria:DC) ratio. Cells were extensively washed and the medium was changed to one containing gentamycin (100 mg/ml). Cells were tested 24 hours later for viability after double staining with FITCconjugated Annexin V (BD PharMingen, San Diego, CA) and 1.25 mg/ml propidium iodide (Sigma Chemical Co.), and analyzed by flow cytometry. Annexin V/propidium iodide double negative cells are indicative of viable cells.

Epithelial cell monolayers
Caco-2 cells were seeded in the upper chamber of a transwell filter (Costar 3 mm diameter of pores) for 7-10 days until a transepithelial resistance (TER) of 300 OhmNcm 2 was achieved.
Epithelial cell monolayers were incubated with bacteria (5610 7 CFU/TW) upon the apical surface (top chamber). One hour after incubation, bacteria were washed out and medium was changed with one containing antibiotics (gentamycin 100 mg/mL). Culture supernatants were collected 24 hours later from the bottom chamber (facing the basolateral membrane), filtered through a 0.2 mm filter (Nalgene) and used to activate MoDCs.
MoDCs were incubated for 24 hours in culture supernatant (1:2) and then treated or not with bacteria (10:1 bacteria to DC) for 1 h in medium without antibiotics. Subsequently bacteria were washed out and cells were left in culture for an additional 23 h in medium containing gentamycin 100 mg/mL. Analysis of cytokines released by epithelial cells or DCs was carried out by testing culture supernatants.

MoDC T-cell co-cultures
MoDCs were collected after 24 hours of incubation with the different stimuli and then incubated with purified allogeneic CD4 + CD45RA + T cells (Miltenyi) in 48-well plates (at a ratio of 10 T cells to 1 DC). To measure T cell proliferation: MoDC-T cell were co-cultured for 72 hours, followed by a 16-hour pulse with 1 mCi [ 3 H] thymidine (Amersham, Milan). Cell-associated radioactivity was detected after Cell harvesting (TomTec) on filtermats using a Betaplate Counter (MicroBeta TriLux, PerkinElmer).
To measure cytokine release: After 5 days of MoDC-T cell coculture, supernatants were collected and directly analyzed for cytokine measurements.

DSS colitis
6 mice per group were administered intra gastrically (i.g.) once a day for 7 days with 200 ml PBS containing 10 10 CFUs of bacteria grown as above. Control mice were administered with plain PBS. Mice were then fed with 2% DSS dissolved in the drinking water for 5 days without probiotics, followed by 7 days of plain water and analyzed over time for colitis development. Mice were weighed every day and feces were collected to measure consistency and the presence of blood by HEMOCCULT (BeckmanCoulter, Inc). At day 13 after DSS treatment mice were sacrificed.

Statistical analysis
Student's paired t test was used to determine the statistical significance of the data. Significance was defined as *, p,0.05; **, p,0.01 (two-tailed test and two-sample equal variance parameters). Statistic calculations were performed by JMP 7 software (SAS Cary). transwell filter and incubated with bacteria (56107 CFU/TW) from the apical surface (top chamber). One hour after incubation, bacteria were washed out and medium was changed with one containing antibiotics (gentamycin 100 mg/mL). Cells were collected 3, 5, 23 and 48 hours later from the bottom chamber. mRNA was isolated and retrotranscribed. Quantitative RT-PCR showing TSLP mRNA expression levels normalized to TBP gene are shown. The bars represent normalized TSLP expression values (TSLP/TBP ratios). One of two independent experiments is shown. Found at: doi:10.1371/journal.pone.0007056.s003 (0.14 MB PPT)

Author Contributions
Conceived and designed the experiments: MR. Performed the experiments: EM GM II. Analyzed the data: EM GM II MR. Wrote the paper: MR.