Oral Supplementation With Selected Lactobacillus Acidophilus Triggers Antimicrobial Response, Activation of Innate Lymphoid Cells Type 3 And Improves Colitis

Live biotherapeutic products constitute an emerging therapeutic approach to prevent or treat inammatory bowel diseases. Lactobacillus acidophilus is a constituent of the human microbiota with probiotic potential, that are illustrated by direct and indirect antimicrobial activity against several pathogens and improvement of intestinal inammation. In this study, we evaluated the anti-inammatory properties of the L. acidophilus strain BIO5768 and assessed the underlying mechanisms of action. BIO5768 was able to counteract the acute colitis that is induced by 2,4,6-trinitrobenzene sulfonic acid (TNBS). When administered alone or in combination with Bidobacterium animalis spp. lactis BIO5764 and Limosilactobacillus reuteri, BIO5768 was also able to alleviate intestinal inammation induced by Citrobacter rodentium infection. Supplementation of naïve mice with either strain BIO5768 alone or as mixture, increased the gene expression of several target genes involved in immune signaling, including c-type lectin Reg3 gamma. Consistently, the ability of innate lymphoid cells to secrete IL-22 was enhanced in response to BIO5768. Interestingly, the aforementioned responses were shown to be independent of NOD2 and Th17 signaling in mice that were mono-colonized with BIO5768. In conclusion, we identify a new potential probiotic strain with the ability for the management of inammatory bowel diseases, and provide some insights into its mode of action.


Introduction
Crohn's disease (CD) is traditionally characterized by the development of transmural in ammatory lesions leading to progressive destruction of the intestinal wall. In Europe, the combined prevalence is about 250-300 cases per 100,000 inhabitants, with about 2.2 million people affected. Several epidemiological and experimental evidences indicate that in emerging countries incidence rates are rising due to the in uence of many gene-environmental interactions, such as, amongst others, tobacco smoking. While the lifetime therapeutic management of the disease is far from optimal, CD impairs patient's ability to work while also their social interaction is altered in various ways. Invasive surgery and bowel resection is required in about two-thirds of CD patients, although it does not cure the disease.
Consequently, the annual economic burden of medical care of CD patients ranges between 2.1 and 16.7 billion € in Europe 1 .
The gut microbiota plays a crucial role in many physiological processes from the earliest days of life, including the maturation of the gut barrier and the immune system. Recent progress in high-throughput sequencing revealed that lowered bacterial diversity is commonly observed in the gut microbiota of CD patients 2,3 . Speci cally, a decreased prevalence of Faecalibacterium prausnitzii was associated with a higher risk of postoperative recurrence of ileal CD 4 . Inappropriate interaction between some components of the gut microbiota and the mucosal immune system is thereby thought to in uence disease initiation and progression 5,6 . It is worth noting that clinical remission following anti-TNF therapy was signi cantly associated with an increased bacterial diversity of fecal microbiota and correlated with higher levels of butyrate and substrates involved in butyrate synthesis 7 . Accordingly, prospective studies revealed that early microbiome changes can predict the response of CD patients to Vedolizumab 8 and Ustekinumab 9 .
Therefore, supplementation with health-promoting probiotics is becoming particularly attractive not only to reconstitute the diversity and the functionality of patient's microbiome but also to counteract the exaggerated in ammatory responses. While probiotics showed successful impact against ulcerative colitis, none of the reported clinical trials with probiotics proved some e cacy in CD [10][11][12][13] . A possible explanation for such ine ciency is that their mode of action may rely on certain genes that are found mutated in CD patients. In agreement with this hypothesis, we provided evidence that the protective capacity of a selected strain of Ligilactobacillus salivarius Ls33 (formerly known as Lactobacillus salivarius Ls33) requires an intact nucleotide-binding oligomerization domain 2 (NOD2) signaling 14 , mutations of which occur in more than one third of the CD patients in Europe and North America. We then performed a comprehensive screening of the Bioprox probiotic collection for identifying strains that are exhibiting anti-in ammatory and antimicrobial abilities. This led us to identify a Lactobacillus acidophilus strain inducing antimicrobial responses that do not depend on NOD2 signaling. This strain BIO5768 was originally isolated from the human gastrointestinal tract and is produced and commercialized as dietary supplements by the society Bioprox Healthcare.
In the current study, we report that supplementation with L. acidophilus BIO5768 enhanced the activity of innate lymphoid cells type 3 (ILC3), that are known to play an essential role in the maintenance of the barrier function and tissue repair 12 , 15 . Speci cally, we tested their capacity to modulate the expression of interleukin-22 (IL-22) and of antimicrobial peptides (AMPs) both in vitro and in vivo. Consequently, the expression of several IL-22-targeted genes was enhanced in response to BIO5768, as earlier observed in response to retinoic acid and aryl hydrocarbon receptor 16,17 . Interestingly its ability to induce AMPs responses was shown to be NOD2 and IL-17-independant. Given that BIO5768 exhibited a different mode of action than the previously studied strains Bi dobacterium animalis spp. lactis BIO5764 (referred herein as BlO5764) and Limosilactobacillus reuteri BIO5454 (formerly known as Lactobacillus reuteri BIO5454, referred herein as BIO5454) 18 , we compared its in vivo anti-in ammatory capacity when administered alone or in combination, by using a Citrobacter rodentium infection model and a TNBS colitis model.

Results
L. acidophilus BIO5768 ameliorates the severity of TNBS-induced acute colitis in mice.
The potential ability of the strain BIO5768 to limit the severity of colitis was rst evaluated in an experimental murine model of TNBS-induced acute colitis. While we observed only a moderate effect of the bacterial oral supplementation on weight loss (Fig. 1a), BIO5768 administration dampened the severity of colitis as indicated by a signi cant decrease of the macroscopic Wallace score (Fig. 1b), con rmed by the histological analyses of colon tissue indicated by a decreased Ameho score, albeit not signi cant (Fig. 1c). The expression of genes encoding the pro-in ammatory markers Tnfa and Cxcl2 was also signi cantly downregulated in the BIO5738-treated group and to a lower extend also the genes encoding Il1b and Il6 (Fig. 1d).
L. acidophilus BIO5768 downregulates in ammatory responses in the Citrobacter rodentium infection model.
A C. rodentium infection model in mice was used to mimic the human situation in which enteropathogenic and enterohemorrhagic strains of Escherichia coli contribute to the development of intestinal in ammatory responses in IBD. We evaluated the impact of the oral administration of BIO5768 on the prevention and/or limitation of transient colitis caused by this bacterium, pathogenic for mice. Despite no effect of BIO5768 on the burden of C. rodentium (Fig. 2a), the colon length was signi cantly reduced in mice infected by C. rodentium (p < 0.001), while no signi cant difference was measured in mice treated with BIO5768, in comparison to non-infected control mice (Fig. 2b). Hyperplasia of crypts in C. rodentium-infected mice was signi cantly increased compared to non-infected mice (p < 0.001). There was only a marginal effect of BIO5768 administration on the shortening of crypt length, as compared to C. rodentium-infected mice (Fig. 2c). Infection by C. rodentium signi cantly elevated gene expression in the colon of the pro-in ammatory markers (Tnfa, Il6 and Il1b), as compared to non-infected mice (p < 0.001). BIO5768 supplementation lowered the expression of these aforementioned pro-in ammatory markers, but this effect was not statistically signi cant (Fig. 2d).
L. acidophilus BIO5768 promotes antimicrobial expression in vitro and in vivo, IL-22 production and dendritic cell maturation.
We evaluated the capacity of BIO5768 to regulate the production antimicrobial peptides by using the mouse intestinal cell line m-ICcl2 presenting a crypt phenotype, and is able to differentiate to cells that share similarities with Paneth cells 19 . BIO5768 induced the expression of the transcripts encoding defensin beta 2 (Defb2) and defensin alpha 4 (defa4), although no effect was observed on Regenerating islet-derived protein 3 gamma (Reg3g) expression and on IL-22 (Fig. 3a). We then evaluated the capacity of BIO5768 to promote antimicrobial peptide expression in vivo in BALB/c naïve mice. BIO5768 signi cantly increased the expression of Defb2 (p < 0.05) and to a lower extend Reg3g in the proximal colon while the expression of Defa4 was not affected by BIO5768 administration (Fig. 3b). Similar results were observed in the distal colon (Fig. S1). We then evaluated the capacity of the strain to regulate the expression of the gene encoding for IL-22. Despite the absence of signi cant changes in Il22 and Reg3g expression in the proximal colon of mice treated with BIO5768, we observed a positive correlation between Il22 and Reg3g expression (Fig. 3c). This result suggested to us that BIO5768 may at least modestly regulate the activity of ILC3 cells. While we observed similar frequency of IL-22-producing Natural Cytotoxicity Receptor positive (NCR+) ILC3 cells in the mesenteric lymph nodes (MLN) of treated and untreated mice (Fig. 3d), an increased expression of IL-22 within the MLN of mice supplemented by BIO5768 was observed in the NCR − ILC3 (Fig. 3e), known to initiates IL-17 production upon IL-23 stimulation 20 . However, the BIO5768 was not able to expand the number of regulatory T cells, notably CD4 + CD25 + FoxP3 + (data not shown). Dendritic cells (DCs) are professional antigen presenting cells that are playing a key role in induction of cellular immunity and polarization of T helper cells. To test the capacity of BIO5768 to induce DC maturation, bone marrow derived DCs (BMDCs) were stimulated by BIO5768 for 24 hours and cell surface DC activation markers were analyzed by ow cytometry. BIO5768 signi cantly increased cell surface presence of MHCII (p < 0.01), CD40 (p < 0.05) and CD86 (p < 0.01), demonstrating a strong effect of BIO5768 on the DC maturation (Fig. 3f).
The anti-microbial abilities of L. acidophilus BIO5768 in vivo is independent of NOD2 and IL-17.
To test the possible dependence of NOD2 and IL-17 signaling on the antimicrobial abilities of BIO5768, germ-free (GF) mice de cient for Nod2 (Nod2 −/− ), for the receptor-Interacting Protein 2 (Rip2 −/− ) and for IL-17 Receptor A (IL17ra −/− ) were mono-colonized by BIO5768. Thirty days after mono-colonisation, no signi cant differences on gene expression of Defb2, Ang4, Reg3g, Il22, Il10 and II17a was observed in the colon of the mono-colonized de cient mice, as compared to similarly treated wild type (WT) GF mice, as shown by Fig. 4a. While we observed an enhanced production of IL-22 by either natural cytotoxicity receptor positive or negative subsets of ILC3, the production of IL-22 by total CD4 + was also signi cantly increased in the large intestine of mice supplemented by BIO5768 (Fig. 4b). By contrast, such differences were not observed in the MLN from those mice (data not shown).
The mixture of BIO5768 with two other strains alleviated in ammation in the C. rodentium infection model.
We evaluated the capacity of L. acidophilus BIO5768 in combination (referred as mixture) with the two other strains B. animalis spp. lactis BIO5764 and Li. reuteri BIO5454 that were previously shown to limit the severity of colitis caused by C. rodentium 18 . Mice were treated orally for ve consecutive days by the mixture prior to C. rodentium infection. Similarly to single strain application (for BIO5764 and BIO5454 see reference 18 and for BIO5768 see Fig. 2a), the mixture had no effect on the burden of C. rodentium ( Fig. 5a) nor on the colon length (data not shown). However, the mixture supplementation was able to signi cantly downregulate the gene expression of Il1b, while having moderate effect on Cxcl2, Tnfa and Il6 (Fig. 5b).
The mixture increased antimicrobial peptide expression and IL-22 secretion by innate lymphoid cells type 3.
The potential of the mixture to induce in vivo expression of antimicrobial peptides was compared to the impact of individual administration of BIO5768. Supplementation by the mixture to naïve mice induced the expression of Defb2, Defa4, Reg3g and Il22, but it was only signi cant for Reg3g (Fig. 6a). Interestingly, the effect of the mixture was higher than BIO5768 alone for Reg3b and Defa4, while the Il22 gene expression was higher for BIO5768 alone (p < 0.01). The capacity of the mixture to promote IL-22 secretion by ILC3 isolated from MLN was analyzed by multiparameter ow cytometry. The mixture signi cantly elevated IL-22 production by all three ILC3 subpopulations, both NCR + and NCR − ILC3 as well as lymphoid tissue inducer cells (LTi) (Fig. 6b). Supplementation by the mixture also promoted IL-17A + RORgt + CD4 + T cells (Fig. 6c), whereas BIO5768 failed to do so (data not shown). As what observed with BIO5768 alone, the mixture was not able to increase signi cantly the abundance of CD4 + CD25 + FoxP3 + Tregs in MLN (data not shown), to the same extent as previously observed with BIO5464 18 .

Discussion
Crohn's disease is associated with multiple pathogenic factors including genetic polymorphisms, gut microbiota dysbiosis and broadly dysregulated adaptive immune and antimicrobial responses. Therefore probiotics able to dampen disease severity and to restore the composition of the gut microbiota are promising safe alternatives in IBD. The potential of bacterial strains to limit the severity of in ammation and the degree of intestinal damage in chemically induced colitis, has been largely reported [21][22][23] . Similarly, in several documented studies, different strains of lactobacilli and bi dobacteria were able to alleviate detrimental in ammatory responses in a murine model of infectious colitis induced by C. rodentium 24,25 . On the contrary, Kennedy et al. observed no effect of L. plantarum 299 on TNBS-induced colitis in rat experimental model highlighting the strain-dependent effect of probiotics in different experimental models 26 . It is important to emphasize that probiotics may mediate their bene cial effect in a strain-dependent manner via distinct signaling pathways that overall remain poorly studied.
We previously reported that the anti-in ammatory capacity of selected lactobacilli relied on an intact NOD2 signaling 14 . Genome-wide association studies demonstrated that NOD2 is the most important genetic factor linked to ileal CD. Indeed, over 30 percent of CD patients present loss-of-function polymorphisms in NOD2. NOD2 is involved in the sensing of bacteria and in the secretion of antimicrobial peptides (AMPs), making it a key player in innate and adaptive immune responses and regulation of the gut microbiota. The NOD2-dependent bene cial effect of probiotics could explain the failure of clinical trials using probiotics for CD patients, as compared to successful results observed in ulcerative colitis 10,13 . Therefore the selection of probiotic strains able to exhibit protective effects in a NOD2-independent manner is relevant to alleviate CD outcome in patients carrying NOD2 mutations.
L. acidophilus is one of the main commercial species of lactic acid bacteria available in different types of dairy products or dietary supplements with claimed probiotic effects 27 . In this report, we highlighted the capacity of the strain L. acidophilus BIO5768 to improve the severity of in ammation in two experimental models of colitis, TNBS induced colitis and transient colitis caused by the pathogenic bacterium C. rodentium. Similarly, L. acidophilus NCFM was shown to be effective in inhibiting colitis induced by C. rodentium 28 . Neonatal and adulthood supplementation of mice by L. acidophilus also limited the severity of pathology associated with the presence of C. rodentium 12,29 . The strain-dependent effect among probiotic species is well known, and differences among L. acidophilus strains in improving intestinal epithelial barrier function was reported recently 30 . Notably, the strain of L. acidophilus LA1 uniquely enhanced intestinal tight junction 31 barrier function in a TLR2 dependent manner 30 . L. acidophilus was also reported to exhibit antioxidant and anti-in ammatory potential in an experimental model of arthritis 32 . The protective effect of L. acidophilus against experimental colitis was also shown to be dose-dependent, emphasizing the importance of selecting an optimal dosing regimen 33 . A recent report also indicated that the anti-in ammatory abilities of L. acidophilus LA-5 depend on the matrix in which the bacterium is delivered (capsules versus yogurt) 34 .
Since impaired production of AMPs in patients suffering from CD is a largely reported issue, the capacity of probiotics to promote host defence is highly desirable. We showed both in vitro using a murine cell line mimicking Paneth´s cells (m-ICcl2), and in vivo using naïve mice, that the BIO5768 strain was able to upregulate the expression of some beta-defensins and the production of IL-22. IL-22 is produced by different types of immune cells, notably ILC3 and other immune cells such as Th17, Th22, natural killer cells, γδT cells and lymphoid tissue inducer (LTi). This cytokine is important in maintaining the integrity of the epithelial barrier 31,35,36 . Several reports highlighted promising potential of distinct bacterial strains to renew AMP production both in vitro and in vivo 37 . Different Lactobacillus species and especially Li. reuteri were also shown to activate IL-22 production by ILC3 38,39 . ILC3 is a heterogeneous population consisting of three subpopulations: NCR + ILC3, NCR − ILC3 and LTi. During the early postnatal period when the gut associated lymphoid tissue is developing, LTi regulates the composition of the gut microbiota and contributes to the establishment of bacterial tolerance. In adults, LTi are important for the renewal of damaged gut lymphoid tissue suggesting their critical role in IBD. Since their function is impaired in CD patients 40 , ILC3 has become a potential target of novel therapeutic approaches 31,41 . Under normal conditions, ILC3 is induced by bacterial metabolites such as SCFA or tryptophan metabolites. Recently, a protective effect of Li. reuteri D8 on the epithelial barrier has been documented in vitro using co-cultured system with lamina propria lymphocytes (LPLs) in an organoid model 42  Other studies have reported that supplementation with three Lactobacillus strains with high tryptophanmetabolizing activities were able to restore intestinal IL-22 production 44,45 . Lactobacilli were also reported to maintain healthy gut mucosa by producing L-ornithine able to increase the level of AhR ligand, L-kynurenine, upon tryptophan metabolism in the gut epithelial cells, therefore increasing the expansion of RORγt + IL-22 + ILC3 cells 46 . Here we con rmed that BIO5768 was su cient to increase IL-22 production by different subsets of ILCs and CD4 + T cells. RORγt is involved in Th17 cell development, which produces the key effector cytokine IL-17, playing a dual role in IBD. Pro-in ammatory immune responses mediated by IL-17 are known to contribute to the pathology of IBD on the one hand but on the other hand, mice with impaired IL-17 signaling typically present a worse course of colitis 47,48 . Blocking IL-17 activity also worsen gut in ammation 49 . T H 17 cells have been shown to be capable of regulatory functions and to be crucial in maintaining mucosal immunity against speci c pathogens by promoting mucosal barrier repair through the stimulation of epithelial cells and tight junction protein, as well as the induction of antimicrobial peptides 50 . Other studies also suggest that IL-17 might be a novel class of cytokines which possesses both pro-and anti-in ammatory abilities, suggesting its critical role in setting and maintaining the gut homeostasis either by promoting barrier function 51 or dampening expression of RANTES 52 . Pleiotropism in T h 17-associated responses may be attributed to IL-22. IL-22 was indeed shown to be secreted abundantly by T h 17 cells 53 . Interestingly, L. acidophilus was reported to suppress the activation of the IL-23/Th17 axis associated with DSS-induced colitis 54 . In the present study, we showed that the L. acidophilus BIO5768 strain had a capacity to signi cantly increase IL-22 production together while limiting the severity of in ammation in TNBS and C. rodentium experimental models of colitis.
Importantly, the expression of transcripts encoding for AMPs and IL-22 was not different in BIO5768 mono-associated germ-free mice de cient for Nod2 (Nod2 −/− ) and IL-17 (Il17ra −/− ) signaling. This led us to provide evidence that the selected strain BIO5768 triggers NOD2-independent AMP expression. Since between 30% and 50% of CD patients in the Western countries carry loss-of-function Nod2 polymorphisms 55 , this potential probiotic strain could therefore be an interesting candidate for further testing in preclinical models of chronic colitis, especially in models exhibiting NOD2 de ciency, in order to con rm the e cacy of the strain to treat patients with NOD2 polymorphisms.
Many studies reported improved performance of probiotic mixtures, compared to individual strains. A mixture containing L. acidophilus was shown to alleviate DSS-induced colitis, notably by increasing the expressions of TJs and by upregulating the number of Tregs 56 . To ensure that probiotic supplementation could indeed trigger multiple protective signaling pathways, including the capacity to promote secretion of IL-22, we evaluated the capacity of the BIO5768 strain in combination with the two other strains B. animalis spp. lactis BIO5764 and Li. reuteri BIO5454, previously reported to exhibit anti-in ammatory abilities in experimentally induced colitis, albeit with different modes of action 18 . Indeed, in vitro, strain BIO5454 e ciently triggered IL-22 secretion by ILC3 and CD4 + T lymphocytes and NOD2-independent AMP expression. Likewise, BIO5764 did not require an intact NOD2-dependent signaling pathway for regulating Defb2 and IL-22 18 . In the C. rodentium infectious colitis model, the mixture with the three strains was able to downregulate the expression of in ammatory genes, in a similar manner as the BIO5768 strain alone. The mixture also promoted the expression of antimicrobial peptides in naïve mice, with the level of Reg3g and Defa4 being higher than with strain BIO5768 alone, although it was lower for Il22 and Defb2. Importantly, the mixture increased IL-22 and IL-17A secretion by different ILC3 subsets and by Th17, respectively. The propensity of the mixture to induce cytokines with the capacity to trigger expression of antimicrobial peptides is of high importance in the patients suffering from CD.
In conclusion, we highlighted the health bene cial abilities of L. acidophilus strain BIO5768 to counteract the severity of in ammation in two experimental models of colitis. The bene cial effect of this bacterium was partly mediated by its capacity to trigger antimicrobial responses and to promote the functional activity of ILC3. Interestingly, the antimicrobial capacity of this strain was shown to be NOD2-independent which could be of great interest for the treatment of CD patients with loss-of-function NOD2 polymorphisms. Combining of this strain with two other dairy strains maintained not only the antiin ammatory and antimicrobial potential, but also triggered a Th17 response. Further studies are warranted to clarify the mode of action of L. acidophilus or the mixture, notably if the protective effects are indeed NOD2-independent in vivo. Our results suggest that the selected strains could provide an interesting complementary therapy in maintaining remission and improving the quality of life of patients. It remains necessary to investigate their clinical e cacy and the safety of the strains, however.
For in vivo administration, fresh cultured bacteria were resuspended in PBS at 2.5 x 10 9 CFU/ml and were intragastrically administrated to mice (5×10 8 CFU in 200 µl) as described previously 18 .
Mice. C57BL/6 and BALB/c female mice were purchased from Charles River (L´Arbresle, France) and were housed in speci c pathogen-free condition in the animal facilities of the Institut Pasteur de Lille , and mouse epidermal growth factor (10 ng/ml) (Peprotech, Neuilly-Sur-Seine, France). Cells were differentiated until 10 days-culture by changing the medium every two days. On day 11, cells were stimulated with the bacteria (bacteria/cell ratio 10:1) for 4 hours. After stimulation, RNA was extracted from adherent cells using the Macherey-Nagel NucleoSpin RNA II isolation kit (Düren, Germany) and stored at -80°C until used for further gene expression quanti cation.
Preclinical model of TNBS-induced colitis. TNBS-induced murine model of acute colitis 58 was performed using BALB/c mice as described previously 18 . Brie y, anesthetized mice received an intra-rectal administration of TNBS (Sigma-Aldrich Chemical, France; 110 mg/kg) dissolved in 0.9 % NaCl/ethanol (50/50 v/v). The protective effect of the probiotic strain was evaluated by oral administration (intragastric feeding) of bacteria (5x10 8 CFU/ mice) starting 5 days before colitis induction. Forty eight hours after colitis induction, mice were sacri ced and the severity of colitis was graded according to the macroscopic in ammation based on Wallace scoring method 59 . Histological analysis was performed on May-Grünwald-Giemsa stained 5 µm tissue sections from colon samples and in ammation was graded according to Ameho score. Immediately after sacri ce, colonic samples were taken and stored in RNAlater storage solution (Ambion, Austin, TX, USA) at -80°C until further processed.
Preclinical model of infectious colitis. Citrobacter rodentium infection was performed using the kanamycin resistant DBS 120K strain as described previously 18 . Brie y, a single colony of C. rodentium was cultured overnight in Luria Bertani broth containing 50 µg/ml kanamycin, under agitation. Bacteria suspension was centrifuged, washed, resuspended in PBS and adjusted to 5x10 9 CFU/ml. Mice were infected by oral administration of C. rodentium (10 9 CFU per mice). The potential capacity of the selected probiotic strains to limit in ammation caused by C. rodentium was evaluated upon intragastric administration of the bacteria or the mixture (5x10 8 CFU per mice, all strains were present in equal amount in the mixture) 5 days prior C. rodentium infection and daily following infection until termination of the experiment. Mice were sacri ced 9 days after infection. Level of infection was monitored as described previously 18 . Histological analyses were performed on May-Grünwald-Giemsa stained 5 µm tissue sections from colon samples xed in 10% formalin and embedded in para n and crypt length was measured using ZEN (Zeiss, Oberkochen, Germany). Immediately after sacri ce, proximal and distal colon segments were put in RNAlater® (Ambion, Life Technologies, Foster City, CA, USA) and frozen at -80°C until RNA extraction and qRT-PCR analysis.
RNA extraction and analysis of gene expression using quantitative real-time polymerase chain reaction (qRT-PCR). Colon samples were removed at sacri ce and stored in RNAlater® storage solution (Ambion, Life Technologies, Foster City, CA, USA) at − 80°C until qRT-PCR analysis. Tissue samples were homogenized using Lysing Matrix D (MPbio, Eschwege, Germany) and total RNA from samples was extracted as described previously 18 . Brie y, Macherey-Nagel NucleoSpin RNAII isolation kit (Düren, Germany) was used for RNA extraction according to the manufacturer's recommendation. RNA quantity and quality were checked by Nanodrop (260/280 nm, 260/230 nm) and 1 µg RNA was reverse-transcribed using the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Woolston Warrington, UK). Statistics. GraphPad Prism was employed for graph preparation and statistical evaluation. Statistical signi cance was determined using either non-parametric Mann Whitney test, non-parametric one-way analysis of variance (ANOVA) followed by Dunn multiple comparison posthoc test and non-parametric two-way ANOVA with Bonferroni post-tests (GraphPad Prism software). Data with p values ≤ 0.05 were considered to be signi cant. The Spearman correlation coe cient was used to analyze correlation between Il22 and Reg3g expression. Figure 1 Protective effect of L. acidophilus BIO5768 supplementation in a mouse model of TNBS-induced colitis.

Declarations
BIO5768 (5x108 CFU of each) was administered to BALB/c mice for 5 consecutive days before and 1 day after TNBS induction. After 48 h, mice were sacri ced and the impact of BIO5768 was analyzed on a) weight loss, b) macroscopic grading of in ammation according to Wallace score, c) histological analyses of colon tissue according to Ameho score and d) gene expression of pro-in ammatory markers. Bars represent means ± SEM of 10 mice per group. *p < 0.05.

Figure 2
Impact of L. acidophilus BIO5768 supplementation on the course of colitis in an experimental mouse model of C. rodentium infection. BIO5768 (5x108 CFU) was daily administered to C57BL/6J mice 5 days prior to infection and after the infection until the day of sacri ce. Mice were orally inoculated with 1×109 CFU C. rodentium and sacri ced 9 days after infection. a) Effect of BIO5768 on bacterial load of C. rodentium determined by plating over the course of the experiment, fecal samples dilution on Luria Bertani medium containing 50 µg/ml kanamycin and compared with control infected mice. b) Effect of BIO5768 supplementation on the colon length and c) crypt length. d) Pro-in ammatory genes expression analyzed by qRT-PCR. Results are expressed as means ± SEM. *** p < 0.001.  The anti-microbial abilities of L. acidophilus BIO5768 in vivo is independent of NOD2 and IL-17 signaling. Germ-free female C57BL/6 mice de cient for NOD2 (Nod2-/-), RIP2 (Rip2-/-), and IL-17 Receptor A (Il17Ra-/-) were mono-colonized with strain BIO5768 by a single administration (5x108 CFU/ mice) and compared to monocolonized GF WT mice (n=6 mice per group). After 30 days mono-association, no signi cant difference on gene expression of defb2, ang4, reg3g, IL-22, il10 and IL17a induced by BIO5768 was observed in mono-colonized de cient mice compared to mono-colonized WT mice. Values are expressed as the relative mRNA levels of mono-colonized de cient mice compared with mono-colonized WT mice and expressed as means ± SEM. b) Impact of BIO5768 supplementation on the percentage of cells producing IL-22, determined by ow cytometry in colon and caecum. NCR natural cytotoxicity receptor. Data represent means values of each group (n = 8 mice) ± SEM. * refers to the comparison of BIO5768-mon-colonized mice versus untreated mice. .* p < 0.05; **p < 0.01 Figure 5 Capacity of the mixture to limit Citrobacter rodentium-induced colitis in C57BL/6J mice. Mixture of three bacterial strains (5x108 CFU/ day / mice, all strains were present in equal amount in the mixture) was administered by intragastric gavage for 5 days. a) Effect of mixture supplementation on C. rodentium burden, and b) gene expression of cxcl2, Il6, Il1b and tnfa in proximal colon evaluated by qRT-PCR. Values are expressed as the relative mRNA levels compared with samples from untreated mice and represent a mean of 10 mice per group ± SEM. * p < 0.05 **p < 0.01.