Host-probiotic interaction: new insight into the role of the endocannabinoid system by in vivo and ex vivo approaches

The endocannabinoid system plays an important role in regulating inflammation in several chronic or anomalous gut inflammatory diseases. In vivo and ex vivo studies showed that 30 days treatment with a probiotic mix activated the endocannabinoid system in zebrafish. These results highlight the potential of this probiotic mixture to regulate immune cell function, by inducing gene expression of toll-like receptors and other immune related molecules. Furthermore, TUNEL assay showed a decrease in the number of apoptotic cells, and this finding was supported by a reduction in pro-apoptotic factors and an increase in anti-apoptotic molecules. The results presented here strengthen the molecular mechanisms activated by probiotic mix controlling immune response and inflammation.

intestinal and blood dendritic cells and inhibits the generation of pro-inflammatory T helper cells 16 . It appears the protective effects of VSL#3 may be mediated by the DNA isolated from bacterial components of VSL#3 rather than by their metabolites or ability to colonise the colon. It was demonstrated that the anti-inflammatory effect of VSL#3 was dependent on non-methylated bacterial DNA (CpG) signalling acting via the toll-like receptor (TLR)-9 17 . The aim of the study was to provide, by in vivo and ex vivo experiments, new insights into the molecular mechanisms through which VSL#3 probiotics can be recognized as non-pathogen microbes and can activate the immune response without causing inflammation. The zebrafish, Danio rerio, has gained considerable attention as an experimental model in biomedical research 18 . The availability of germ-free species, the resemblance of its physiology to that of mammals, the high similarity of immune system molecules to those in other vertebrates including mammals [19][20][21] , together with the high throughput capability in genetic and chemical screens, give zebrafish research a significant impact in promoting the development of personalized and precision medicine. Zebrafish colonization is characterized by a condition very similar to that of human. To investigate the effect of microbial colonization on the zebrafish host, Rawls and colleagues 22 reared germfree zebrafish and investigate gene expression profiles. They showed that colonization altered the expression of 212 genes. Of these 212 genes, 59 responses were conserved between human, mice and zebrafish. These conserved genes were mainly involved in epithelial proliferation, promotion of nutrient metabolism, and innate immune responses. This indicates that the response towards microbes is in part highly conserved in all these species, especially the pathways here studied.

Results
Effects of VSL#3 on intestinal epithelium integrity. Healthy fish exposed to VSL#3 for 4 weeks showed similar behaviour and survival to controls. Morphological study of intestines of VSL#3-treated zebrafish revealed substantial integrity of gut mucosa comparable to that of controls, without any inflammatory infiltrate or reactive status to probiotics (Fig. 1A,B).
The histological analysis showed an intact epithelial barrier with a columnar uniform epithelial and regular apical brush border without signs of damage in the intestines exposed to probiotic. In addition, after VSL#3 treatment significant differences were seen in enterocyte length, villus length and crypt depth (Table 1S).
Moreover, TUNEL assay showed a high apoptotic rate among epithelial and mesenchymal cells in gut from control fish with respect to treated ones (Fig. 1C,D). The reduction in the apoptotic process in intestinal epithelium after probiotic administration was also confirmed at molecular level by the lower expression of pro-apoptotic signals such as casp3 and baxa, concomitantly with an increase in anti-apoptotic signals such as bcl2a (Fig. 1E-G).
Effects of VSL#3 on the endocannabinoid system: an in vivo approach. Figures 2 and 3 show modulation of the ECS after VSL#3 administration. IHC analysis of gut sections from treated animals highlighted strong presence of Cnr1 in epithelial cells, in some mesenchymal cells, and in the ganglia of the muscular layer of the intestine wall. In contrast, gut sections from control animals showed a strong stain only in parietal ganglia, with markedly low levels of expression in enterocytes or other cell types ( Fig. 2A-E). At molecular level, the involvement of VSL#3 administration on endocannabinoids synthesis and degradation was shown firstly in the gut. A significant increase of abhd4 gene expression was observed, coding for a key enzyme involved in AEA synthesis. Concomitantly, gene expression of faah and mgll, which code for the enzymes involved in degradation of AEA and 2-AG respectively, was significantly decreased after VSL#3 treatment. Moreover, molecular changes to endocannabinoid receptors cnr1 and cnr2 in zebrafish gut as consequence of VSL#3 administration were also observed while, the trpv1 gene was not affected (Fig. 3A-F).
Effects of VSL#3 on toll-like receptor activation. In zebrafish intestine, 30 days of probiotic administration induced a significant (p < 0.001) increase in gene expression levels of key molecules that can recognize microbes in the intestine such as tlr1, tlr2 and tlr9, with respect to control animals (Table 1A). The increase in gene expression was consistent with the increase at protein level found by immunohistochemistry with the exception of Tlr2, which showed no significant changes (Table 1B).
IHC analysis showed intense and diffuse presence of Tlr3 in the mucosal epithelium of treated animals. In sections from these animals it is evident that the stain was present in a continuous and constant pattern and almost exclusively localized along the superficial margin of enterocytes. In contrast, sections from control animals were characterized by weak and irregular Tlr3 presence in the enterocyte, without any staining of cell free pole membranes (Fig. 4A,B).
Effects of VSL#3 on immune response activation. Figure 4 and Table 1A,B illustrate modulation of the immune response after VSL#3 administration. In particular, gut from treated animals showed an increase in both gene and protein levels of molecules implicated in the innate immune response such as Il1b, Tnfa and Myd88 (Table 1A,B). In addition, both mRNA of il10 and casp1, and their protein levels, were significantly (p < 0.001) higher in VSL#3-treated fish than controls (Table 1A,B). Using IHC analysis on gut from treated animals, Nos2a protein was localized specifically in endothelial cells or in macrophages, and to a lesser extent in epithelial cells. In control gut there was substantial negativity for the presence of Nos2a, both in the epithelium and in endothelial cells. (Fig. 4C,D); conversely, an increase of both protein and mRNA level for Nos2a was seen in VSL#3-treated fish (Table 1A,B). Concomitantly, the increase at both gene and protein level of signals involved in antigenic "tolerance" such as Tgfb1a and Nfkb was registered (Table 1A,B). In VSL#3-treated fish, IHC analysis showed higher Tgfb1a protein signal that appeared diffusely localized throughout the epithelium and in mononuclear cells with respect to control ones (Fig. 4E,F). Finally, the p65 subunit of the Nfkb heterodimer was found to be localized in a large proportion of enterocytes and in many mononuclear cells infiltrating the lamina propria of intestinal mucosa of VSL#3-treated fish, while in control gut it was restricted to scattered epithelial cells (Fig. 4G,H). Effects of VSL#3 on the endocannabinoid system: an ex vivo approach. Figures 5 and 6 illustrate modulation of ECS, immune response and inflammation after ex vivo exposure of zebrafish intestine. Eight hours of in vitro VSL#3 treatment on ex vivo intestine samples significantly increases cnr1, cnr2, faah, tlr2, il1b and casp3 gene expression, confirming the stimulatory role of the probiotic. In contrast, 8 hours of exposure to AM251, a Cnr1 antagonist, induced in these samples a significant reduction of cnr1, cnr2, and il1b expression, without affecting faah and casp3 gene expression or the apoptotic rate evidenced by TUNEL. Finally, the concomitant exposure of intestine to VSL#3 and AM251 in the ex vivo assay showed a significant decrease in cnr1, tlr2 and il1b gene expression and in apoptotic rate, without affecting cnr2, faah and casp3 gene expression (Figs 5 and 6).

Discussion
Chronic or anomalous gut inflammation is the primary cause of many gut inflammatory diseases, including ulcerative colitis and Crohn's disease, and it has been clearly shown that the endocannabinoid system plays an important role in modulating this inflammation 9, 23-25 . In particular, Di Marzo and Izzo 26 demonstrated that cnr1 and cnr2 activation may limit inflammation on well-established models of IBD in rodents. Data from several trials (reviewed by Chapman et al. 13 ) suggested that VSL#3 has clinical potential in the treatment of many gut inflammatory diseases, but the mechanisms by which it exerts its beneficial role are not fully elucidated.
Until now no publications are reported in literature, that indicate studies regarding endocannabinoid system organization in gastrointestinal tract and its modulation by the microbiome in zebrafish. Meanwhile, given that several gut functions and immune genes are conserved between zebrafish and mammals, the zebrafish is an interesting model organism to investigate fundamental processes underlying intestinal inflammation and injury 27 . In this context, the similarities of this model with mammal and rodent strongly suggest to consider relevant our results regarding distribution and role of endocannabinoid receptors in ZF. These results are relatable to that which would be the endocannabinoid expression in humans or mice GI tract. The results obtained here in zebrafish intestine are in agreement with the previous reports from Rousseaux and collaborators 11 done both in vitro in human HT-29 epithelial cells and both in vivo in Balb/c mice and Sprague Dawley rats, showed the ability of Lactobacillus acidophilus to induce intestinal cannabinoid receptors, and from Distrutti and co-workers 28 who reported increased cnr1 gene expression in mouse colon after VSL#3 administration.
In the present study, we offer evidence that in control intestine, cnr1 presence was only detectable in ganglia, while in VSL#3-treated gut, Cnr1 was also present in epithelial cells and in some mesenchymal cells. Interestingly, as suggested by Sibaev et al. 29 it should be noted that in our experiments, spindle-shaped cnr1 immunopositive cells, encountered within the outer muscle layer, were systematically ckit/GFAP/S100 immunonegative (data not shown), demonstrating that interstitial cajal cells (ICC) are not involved in this VSL#3-induced cnr1 overexpression.
The anti-inflammatory capability of VSL#3 was indicated by the positive modulation of cnr2 concomitantly with the constant level of trpv1 genes on zebrafish intestine. In fact, the importance of Cnr2 receptor activation in the immune-modulatory effects of endocannabinoids has been well recognized, and its anti-inflammatory role is evident in many pathological conditions such as inflammatory pain and gastrointestinal inflammatory disorders 30 . In addition, it has been shown that Trpv1 binding AEA induces inflammation in the gut 31 .
In the present study, it was demonstrated that in zebrafish, probiotics included in the VSL#3 formulation communicate with the host by activating Tlrs as demonstrated in others species 17 . Chu and Mazmanian 32 noted in a review that Tlrs, being pattern-recognition receptors (PRR), may have evolved to mediate the bi-directional cross talk between "good" bacteria and their hosts. In particular, in zebrafish the PRR system is highly similar to that in other vertebrates and tlr orthologs have been identified and well characterized in this fish species 33,34 . Key features of the fish Tlrs and the factors involved in their signalling cascade have high structural similarity to the mammalian Tlr system. The ligand specificity of several Tlr receptors is highly conserved in all vertebrates 35 . Most vertebrate genomes are recognized to have at least one gene representing each of the seven major tlr1, tlr2, tlr3, tlr4, tlr5, tlr7 and tlr11 families 36 but the most conserved among zebrafish and human and other mammals models are tlr1, 2 3 and 9 34 . In particular, after microbial identification, all Tlrs with the exception of Tlr3 trigger the activation of intracellular pathways, which is myeloid differentiation factor 88 (Myd88)-dependent 37 . In this case, Myd88, through the activation of nuclear factor Nfkb, induces production of pro-inflammatory cytokines including Il1b and Tnfa 37 . The interaction of a probiotic, such as E. coli Nissle with Tlr2 and Tlr4 was previously found in mammals 38 , but here for the first time we have established the ability of VSL#3 to communicate with the host by activating several tlrs in zebrafish. In particular, VSL#3 administration activated Tlr1 and 2, respectively involved in the recognition of cell-surface bacterial peptidoglycans and lipoproteins both in mammal and fish 39 . Moreover, Tlr3 and 9, which are nucleic acid-sensing and recognize nucleotides from both viruses and bacteria 17,40 , were activated after VSL#3 administration. The activation of these receptors by VSL#3, in particular Tlr2 and Tlr9, has been related to immune system activation without triggering a negative inflammatory response. Tlr2 is widely described as a PRR, mediating host protective immune responses 41,42 . Interestingly, not all Lactobacilli strains can activate Tlr2, suggesting that strain-specific immunoregulatory effects may be partially mediated via Tlr2 activation 43,44 .
Results obtained with our zebrafish model, when treated with VSL#3, show on the one hand an over-stimulation of tlr2 and tlr9 and respective proteins, and on the other hand increased expression of some inflammatory cytokines, but without histological signs of mucosal inflammation and damage. These results are in line with previous observations in macrophages on the ability of some probiotics, such as L. rhamnosus GG, to moderately activate the production of inflammatory cytokines and the transcription factors involved in humans [45][46][47][48] 49,50 . Moreover, anti-inflammatory genes such as il10 and tgfb1a and genes for transcription factors such as nfkb, were affected by probiotic administration 51 . The anti-inflammatory effect exerted by VSL#3 highlighted in our study was supported by the maintenance of intestinal epithelial integrity, shown by histological analysis. In addition, the reduced apoptosis rate shown by TUNEL assay, as well as the reduced level of pro-apoptotic signals such as casp3 and baxa, concomitantly with the increase in anti-apoptotic signals such as bcl2a 52-54 support the anti-inflammatory role of VSL#3. Although single epithelial cell apoptosis and elimination without loss of barrier permeability are normal physiological events in the gastrointestinal tract 55 , an increased epithelial apoptotic ratio might be cause of relevant leaks in the epithelial barrier. Uncontrolled cell cycle or apoptosis are often the main mechanisms in pathogenesis of cancers, autoimmune diseases or chronic inflammation [56][57][58] . Our results are in agreement with some previously obtained in a murine model after VSL#3 treatment showed a decreased apoptosis in intestinal epithelial cells 59 .
In our results, the induction by VSL#3 administration of il1 and tnfα associated with casp1 expression resulted in an anti-apoptotic final outcome. These results are in accordance with the recent observation that casp1 activation fails to trigger pyroptosis in all cell types, and somewhat surprisingly, epithelial cells use casp1 activation to prevent cell death 60 . The function of casp1 is analogous to the activities of other apoptotic caspases (Casp3 and 8) in modulating the fate of certain cell types 61 . Low levels of casp1, as probably occurred in our case, stimulate cell survival responses, control intracellular bacterial growth and mediate inflammatory cytokine production. When casp1 activation reaches a critical threshold level, cells undergo pyroptosis and release inflammatory intracellular contents. To our knowledge, our study is the first to conduct an ex vivo study of zebrafish intestine. We compared histological integrity and apoptotic signals in control intestine before and after 8 hours of in vitro culture and no significant differences were found, clearly indicating that such culture methods are a valid and reliable tool to study the direct effects of bacteria on intestine.
The results achieved by the ex vivo studies clearly showed that the immune stimulant and anti-inflammatory role of VSL#3 is mediated by the activation of the endocannabinoid system. After 8 hours of in vitro VSL#3 exposure, a clear induction of cnr1, cnr2, tlr2 and il1b genes was found, concomitantly with a reduction in casp3 gene expression and TUNEL positive cells, confirming the in vivo study results. At the same time, the lower cnr1 levels induced by AM251 exposure in cultured intestine were related to a decrease in immunerelated gene expression and an increase in apoptotic signal (casp3 and TUNEL-positive cells). The simultaneous addition of VSL#3 and AM251 to the culture medium did not re-establish the levels of signals found in the intestine exposed to VSL#3, strengthening the concept that it acts through the activation of endocannabinoid system.
Concluding, the endocannabinoid system is recognized as an anti-inflammatory agent, with identified protective roles in many inflammatory diseases. In the present study, we report in vivo and ex vivo results that provide a novel molecular network through which probiotics such as VSL#3 activate the endocannabinoid system and induce Tlr signalling. These results further highlight the potential of this probiotic mixture to regulate immune cell functioning, decreasing detrimental effects and consequent inflammatory events in zebrafish. Considering

Methods
In vivo fish maintenance and VSL#3 treatment. Adult male zebrafish specimens purchased from a local supplier (Acquario di Bologna, Bologna, Italy) were acclimated to laboratory conditions and their health state was monitored for four weeks prior to the beginning of the experiments. Fish were divided into a control group (CTRL), which was fed with commercial food, and a probiotic-treated group (VSL#3), that received a commercial diet containing the lyophilized probiotic at a final concentration of 10 9 CFU/g for 30 days. The probiotic formulation utilized was VSL#3 (produced before January 31, 2016; VSL Pharmaceuticals, Inc., Gaithersburg, MD, USA The probiotic strain was mixed into the diet prior to feeding. All fish were served with a quantity of food ranging from 1.5% to 2% of their bodyweight per day. The experiment was conducted in triplicate and for each replicate tank, the final housing density was 20 males. All the procedures involving animals were conducted in accordance with the Italian law on animal experimentation and were approved by the Ethics Committee of Università Politecnica delle Marche (Prot #63/INT/ CESA12-16). All efforts were made to minimize suffering.
After 30 days of treatment, ten males from each experimental group were sacrificed by a lethal overdose of anesthesia (500 mg/L MS-222 [3-aminobenzoic acid ethyl ester] buffered to pH 7.4; Sigma).
Five intestines from each experimental group were removed and fixed in PFA fixative. The remaining intestines were sampled for histology, immunohistochemistry and q-PCR assay.
Ex vivo tissue culture. Twenty-four control fish were sacrificed after a 12-hr fast by a lethal overdose of anesthesia (as above). Intestines were excised and placed in a tissue culture dish of 35 × 10 mm (Falcon). The medium employed in these cultures consisted of 9 parts Dulbecco's modified medium (DMEM)/Ham's F12 (Sigma) and 1 part fetal calf serum (Sigma). In addition, 10,000 U of crystalline penicillin G and 0.01 g of streptomycin sulfate (Sigma) were added to each 100 ml of medium. The intestines were cultured in the presence of antibiotics for 4 hours and then placed in a new tissue culture dish in the same medium without the antibiotics. For experimental purposes, incubations were carried out separately in Dulbecco's Modified Eagle Medium (DMEM) (Thermo Fisher Scientific) (CTRL), DMEM + VSL#3 (final concentration 10 9 CFU/ml) (VSL#3), DMEM + AM251 (Cayman Chemical) (10 nM) (AM251), and DMEM + VSL#3 + AM251, (VSL#3 + AM251). AM251 concentration was chosen on the basis of previous studies 63, 64 .
The dishes were covered, placed on a steel rack set in a McIntosh jar (Scientific Products, Inc., Los Angeles, CA) and maintained at 37 °C. The jar was gassed with 95% 0 2 and 5% C0 2 for 20 min and then sealed. Intestines from each experimental group were sampled for histology, immunohistochemistry and q-PCR assay after 8 hours of culture.
A 5 mm-long piece from the anterior part of each ex-vivo cultured gut was used also for morphological evaluations, after dissection and a 10% buffered formalin fixing procedure and immunohistochemistry analysis.
Histology. PFA-fixed samples were processed for routine histology to obtain a transversal section, which was stained with hematoxylin and eosin (H&E). Sections were evaluated under a light microscope (Carl Zeiss, Jena, Germany) for any degenerative and/or inflammatory changes.
Villus length, crypt depth, villus length/crypt depth ratio and enterocyte length were determined from three fish per group; 50 images were taken for each intestine, for a total of 150 images per group, and 12 enterocytes were measured per image. Enterocyte lengths were measured starting from their basal membrane connected with the lamina propria up to the free, luminal pole of the cell. All measurements were taken from micrographs using the Software Image J. In all cases, measurements were made using sections in which villi, cryptae, and enterocytes were cut along their entire length; only crypts or villi that were sagittally sectioned along their entire length were scored. For enterocytes, the cell was considered complete when the nucleus was visible and aligned, to ensure that the section plane was appropriate. Scientific RepoRts | 7: 1261 | DOI:10.1038/s41598-017-01322-1 peroxide for 5 minutes followed by rinsing for 5 minutes in distilled water. For antigen retrieval, slides were incubated in three antigen retrieval solutions: citrate buffer (pH 6.0) for tgfb1a, nfkb, and tlrs; EDTA (pH 8.0) for myd88; and 0.01 M Tris-EDTA buffer (pH 9.0) for anti-cnr1, anticnr2, anti-casp3, and anti-casp1 in a steamer (Black & Decker, Towson, MD) for 20 minutes. Non-specific binding was blocked by incubation of slides for 10 minutes with a proteinblocking agent (Dako, Carpinteria, CA) before application of the primary antibody. All antibodies used were specific for zebrafish, or previously tested and reactive also for this species. Slides were incubated overnight in a moist chamber with the following primary antibodies: polyclonal (pAb) rabbit anti-cnr1 The immunoreaction with streptavidin-immunoperoxidase (Streptavidin-immunoperoxidase, Black & Decker, Towson, MD) was visualized with 3,3′-diaminobenzidine substrate (Vector, Burlingame, UK). Tissues were counterstained with Mayer's hematoxylin. For negative immunohistochemical controls the primary antibodies were omitted. Sections of zebrafish and killifish spleen and kidney served as positive control tissues for TUNEL, casp1 and 3, tgfb1a, and myd88 cell staining. Positive control tissues for tlrs staining consisted of fish skin and gills sections. For scoring of intestinal parameters such as cnr1-and cnr2-receptor positive cells, tlrs, tgfb1a+, myd88+, nos2a+, and nfkb+ cells, the cells were quantified in the small intestinal compartment of the GI tract (anterior portion of the intestine: villi, basal crypt area, villuscrypt junction) as previously reported 21 . Similar evaluations were performed in an attempt to evaluate the Caspase3, Caspase1 and TUNEL+ cells. All cell types were evaluated using a light microscope (Carl Zeiss, Jena, Germany), a × 40 objective, a × 10 eyepiece, and a square eyepiece graticule (10 × 10 squares, having a total area of 62,500 μm 2 ). Ten appropriate fields were chosen for each compartment and arithmetic means were calculated for each intestinal region. Results were expressed as IHC positive cells per 62,500 μm 2 . For all parameters, cells on the margins of the tissue sections were not considered for evaluation to avoid inflation of positive cell numbers. For the evaluation of different cnr1 and 2 and tlrs subsets in the same histological sections, consecutive 3-μm-thick bioptic cross-sections were cut. Sections were placed consecutively on each of eight separate slides, after which the ninth section was placed on the first slide, next to the first section, continuing for 48 sections. A single slide, upon which were six bioptic cross-sections from each fish, was analysed for any given immunostain. Numbers of cnr1+, cnr 2+, tlrs (1,2,3,9)+, tgfb1+, casp3+ and cas1+ epithelial and/or immunocompetent cells were quantified by using an image analysis system consisting of a light microscope (Carl Zeiss, Jena, Germany) attached to a Javelin JE3462 high-resolution camera and a personal computer equipped with a Coreco-Oculus OC-TCX frame grabber and high-resolution monitor. Computerized color image analysis was performed using Image-Pro Plus software (Media Cybernetics). The area of each biopsy in all six cross-sections in every fish was recorded, as was the total number of epithelial positive cells determined by immunostaining as previously described. For each fish, the total immunostained cells were counted per section, and stained cell densities were expressed as the number of immunocompetent/epithelial cells per square millimeter of analyzed bioptic area 65 . RNA extraction and cDNA synthesis. Total RNA extraction from intestinal tissues was performed following Falcinelli and coworkers 66 . Total RNA extracts were eluted in 30 μl of RNAse-free water. Final RNA concentrations were determined with a NanoDrop TM 1000 Spectrophotometer (Thermo Scientific) and the RNA integrity was verified by ethidium bromide staining of 28 S and 18 S ribosomal RNA bands on 1% agarose gel. RNA was stored at −80 °C until use. Total RNA was treated with DNAse (10 IU at 37 °C for 10 min, MBI Fermentas). A total amount of 1 µg of RNA was used for cDNA synthesis, employing the iScript cDNA Synthesis Kit (Bio-Rad).

Real time PCR.
PCRs were performed with the SYBR green method in an iQ5 iCycler thermal cycler (Bio-Rad Laboratories) following Gioacchini and coworkers 67 . Actin beta (act1b) and ribosomal protein, large, P0 (rplp) were used as internal standards in each sample in order to standardize the results by eliminating variation in mRNA and cDNA quantity and quality. No amplification products were observed in negative controls and no primer-dimer formations were observed in the control templates. The data obtained were analysed using the iQ5 optical system software version 2.0 (Bio-Rad) including GeneEx Macro iQ5 Conversion and genex Macro iQ5 files. The primer sequences used are reported in Table 2S.
Statistical analysis. Data are presented as mean ± SD. Student's t-test was used for comparison between the two experimental groups for in vivo experimentation. Two-Way ANOVA followed by the Tukey test as a multiple comparisons test was used for comparison among experimental groups for ex vivo experimentation. Statistical evaluation of enterocyte length, villus height, crypt depth and villus height/crypt depth ratio were tested by variance analysis and confirmed by nonparametric methods (Wilcoxon). All statistical analyses were performed using the statistical software package Prism5 (Graphpad Software, Inc. USA) with significance accepted at P < 0.05.