The commensal Escherichia coli CEC15 reinforces intestinal defences in gnotobiotic mice and is protective in a chronic colitis mouse model

Escherichia coli is a regular inhabitant of the gut microbiota throughout life. However, its role in gut health is controversial. Here, we investigated the relationship between the commensal E. coli strain CEC15 (CEC), which we previously isolated, and the intestine in homeostatic and disease-prone settings. The impact of CEC was compared to that of the probiotic E. coli Nissle 1917 (Nissle) strain. The expression of ileal and colonic genes that play a key role in intestinal homeostasis was higher in CEC- and Nissle-mono-associated wild-type mice than in germfree mice. This included genes involved in the turnover of reactive oxygen species, antimicrobial peptide synthesis, and immune responses. The impact of CEC and Nissle on such gene expression was stronger in a disease-prone setting, i.e. in gnotobiotic IL10-deficient mice. In a chronic colitis model, CEC more strongly decreased signs of colitis severity (myeloperoxidase activity and CD3+ immune-cell infiltration) than Nissle. Thus, our study shows that CEC and Nissle contribute to increased expression of genes involved in the maintenance of gut homeostasis in homeostatic and inflammatory settings. We show that these E. coli strains, in particular CEC, can have a beneficial effect in a chronic colitis mouse model.

pro-inflammatory adherent-invasive E. coli (AIEC) 12 , colorectal cancer 13 , and, to a lesser extent, ulcerative colitis (UC) 14 . Similarly, the abundance of intestinal Escherichia also increases in several mouse models of inflammatory bowel disease (IBD) 15,16 . A more direct effect of certain indigenous strains of E. coli on intestinal inflammation has been demonstrated in gnotobiotic mouse models with a predisposition to inflammation. Thus, mono-association of mice genetically prone to inflammation with E. coli strains, originally isolated from mice gut microbiota, results in intestinal inflammation [17][18][19] .
However, the role of indigenous E. coli toward gut health is far from clear and requires further investigation. Indeed, another commensal E. coli isolate failed to induce disease in antibiotic-pretreated IBD-susceptible mice, despite robust colonization 20 . The inflammation-prone HLA B27 transgenic rat model responds only very moderately to E. coli strains isolated from CD patients, whereas other bacteria induce severe colitis 21 . Studies in a mono-associated IL2 −/− mice model report divergent effects of indigenous E. coli strains in the induction of colitis 19 . Furthermore, E. coli strains originally isolated from human gut microbiota are the basis of at least two commercially available probiotic products, commercialized under the names Mutaflor and Symbioflor2. The E. coli Nissle 1917 (Nissle) strain, the active component of Mutaflor, is one of the most thoroughly investigated and documented probiotics 22 . Clinical trials have shown a beneficial effect of Nissle for the maintenance of remission in UC, similar to that of mesalazine 23 .
We previously isolated a primo-colonizing E. coli strain called CEC15 (CEC) from freshly pooled fecal samples of 15-day-old suckling rodents as a major representative of this environment. We reported that this E. coli strain elicits sequential remodelling of the colonic epithelium in gnotobiotic rodent models, affecting different arms of the intestinal epithelial defences required to achieve a microbiota-accommodating homeostasis 5 .
Here, we aimed (1) to gain further insight into the relationship between CEC and host intestinal health and (2) to compare the intestinal host response to CEC to that of the probiotic E. coli Nissle. We explored the ileal and colonic host response to these two strains under three different conditions: (i) disease-free mono-associated wild-type (WT) mice, (ii) mono-associated IL10 deficient (IL10 −/− ) mice, predisposed to intestinal inflammation; (iii) conventional IL10 −/− mice exposed to inflammatory challenge as a chronic colitis model.
In summary, our data show that the host response to commensal strains of E. coli mobilized key genes that are involved in sustaining the symbiotic relationship with the gut microbiota. We further demonstrate that the host can over-mobilize its defence mechanisms when mono-associated with commensal E. coli in a setting in which the animals are predisposed to inflammation. Finally, we found that the impact of CEC is beneficial to the host in a chronic colitis IL10 −/− mouse model.

Methods
Ethics approval. All procedures involving animal experimentation were carried out according to the European guidelines for the care and use of laboratory animals under the authority of a license issued by the French Veterinary Services (authorization number 78-122 specific to CC) and were approved by the French "Ministère de l'Enseignement Supérieur et de la Recherche" (authorization number APAFIS#3441-2016010614307552). Experiments involving gnotobiotic WT or IL10 −/− mice were performed at the Anaxem facility of the MICALIS Institute (INRA, Jouy-en-Josas, France), which is accredited by the French "Direction Départementale de la Protection des Populations (DDPP78)", accreditation number A78-322-6. Experiments involving conventional (CV) WT or IL10 −/− mice were performed at the IERP facility (INRA, Jouy-en-Josas, France), which is accredited by the French "Direction Départementale de la Protection des Populations (DDPP78)", accreditation number DDPP-VET-13-0124. GF WT mice and GF IL10 −/− mice (generated from the CV IL10 −/− mice (B6.129P2-Il10tm1Cgn/J; see below) using the standard procedure of cesarean delivery), were purchased from the GF rodent breeding facilities of the CNRS-TAAM (transgenesis, archiving and animal models) center (Orléans, France). They were delivered to Anaxem under sterile conditions and immediately transferred to the experimental isolator. After reception, GF mice were left undisturbed for eight days before starting the experiment. Conventional WT and IL10 −/− (B6.129P2-Il10tm1Cgn/J) 24 mice were born and bred at the IERP under standard conditions. The absence of IL10 gene expression was verified in IL10 −/− GF and IL10 −/− CV mice. All WT GF and CV mice used in experiments were C57BL/6. The genetic background of IL10 −/− mice is a mix of C57BL/6 and 129/Ola. All animals used in this study were males. Gnotobiotic mice. Six groups of gnotobiotic mice were studied ( Supplementary Fig. 1A,B): either GF WT (GF WT ), or IL10 −/− (GF IL10 ); two groups of mono-colonized WT mice inoculated with either E. coli CEC or E. coli Nissle (CEC WT and Nissle WT groups, respectively) and two groups of mono-associated IL10 −/− mice inoculated with either E. coli CEC or E. coli Nissle (CEC IL10 and Nissle IL10 groups, respectively). The absence of microbes was verified in GF mice by microscopic observation of fresh feces and culturing of fecal material on various bacterial culture media. All mice were maintained in Trexler type isolators and received the same standard diet (ad libitum, R03-40 SAFE sterilized by gamma irradiation at 45 kGy). CEC and Nissle inocula were prepared from fresh overnight cultures. Bacterial pellets were obtained by centrifugation (4 °C; 20 min; 4,700 × g), re-suspended in sterile PBS, immediately introduced into the isolator and 100 µL (containing 10 8 bacteria) used to inoculate GF mice. All mice were 8 to 10 weeks old at the time of inoculation and were sacrificed 21 days post-inoculation. Each experiment was carried out independently at least three times.
Experimental colitis in CV IL10 −/− mice. Experimental chronic colitis was induced in six-to-eight-week-old CV IL10 −/− mice by rectal injection with dinitrobenzene sulfonic acid (DNBS) (Supplementary Fig. 1C). The protocol was carried out according to 25 , except for the dose of DNBS, which was lower in our study. Prior to DNBS administration, mice were anesthetized with Xelamine/Ketamine and a 10 cm long piece of PE-90 tubing was attached to a syringe and inserted 3.5 cm into the colon. On the first day of the protocol, mice received one rectal dose of 150 mg/kg DNBS (ICN Biomedical Inc., Santa Ana, CA) in 30% ethanol. All mice received a subcutaneous injection of 1 ml saline solution (0.9% NaCl) for three days to prevent dehydration. Mice were allowed to recover for 21 days and then received a second DNBS injection at day 21, reactivating inflammation. Ten days before sacrifice, DNBS treated CVIL10 −/− mice were given a dose of 1 × 10 9 CFU of CEC (DNBS IL10 + CEC) or Nissle (DNBS IL10 + Nissle) daily by oral gavage, or the same volume of PBS as a positive control for the disease (DNBS IL10 ) ( Supplementary Fig. 1C). Mice were sacrificed 24 days after the first DNBS injection, i.e. three days after the second. Run in parallel, CV IL10 /− mice received PBS instead of DNBS and were supplemented daily with PBS (Control), as a negative control group.
Sample collection. The day of sample collection, at 9:00 AM, mice were anesthetized with isoflurane, killed by cervical dislocation, and the ileum and colon promptly removed. An unflushed section of ileum and colon was kept and fixed in CARNOY (four weeks at 4 °C) for mucus layer thickness measurements as in 5 . The remainder of the intestinal tissues was quickly washed with PBS and immediately used either for ex vivo permeability measurements in an Ussing chamber, or fixed in 4% paraformaldehyde (PFA; 6 h at room temperature) for further histological analyses, or frozen at −80 °C for RNA extraction. Cecal contents were recovered and frozen at −80 °C before DNA extraction for 16S rRNA gene sequencing.
Histology and Immunostaining. CARNOY-and PFA-fixed ileum and colon were dehydrated and embedded in paraffin according to a standard protocol 26 . Staining was performed with Dako reagents according to the recommendations of the manufacturer. For PFA fixed samples, antigen retrieval was performed by boiling the slides for 40 min in 10 mM tri-sodium citrate pH 6.0. The primary antibodies used and corresponding dilutions were: anti-cadherin1 (Cdh-1, Invitrogen; 1/50), anti-CD3 (Abcam, 1/500), anti-mucin-2 (Muc2, Santa Cruz; 1/500), anti-mucin-13 (Muc13, Santa Cruz; 1/500), and anti-Ki67 (Dako; 1/50). Negative controls were performed by omitting the primary antibody from the reaction. For each section, Ki67+ cells were counted for 10 crypts and the results expressed as the percentage of total crypt cells. CD3 + cells were counted by microscopic field (x200). For mucus thickness analyses, CARNOY-fixed cuts were stained with the Muc-2 antibody. Tissues were observed with a 3DHistech Panoramic Scan and signal quantification was performed using Panoramic Viewer ® or ImageJ software. Gene expression profiling using the TaqMan OpenArray System. For studies in gnotobiotic WT or IL10 −/− mice, gene expression analysis was performed with a customized TaqMan OpenArray Real-time PCR System (Life Technologies). We designed two TaqMan OpenArrays, one dedicated to the ileum, the other to the colon, following a thorough review of the literature and searching the NCBI public repository Gene Expression Omnibus. For each TaqMan OpenArray, 220 genes were selected, plus four endogenous control genes (actb, gapdh, ubc, and tpb). The candidate genes selected for this study survey various functions of ileal and colonic cells, mainly involved in mucosal defence: immunological responses, intestinal barrier, oxidative stress, anti-microbial peptides, cellular signalling, regulation of cell proliferation and differentiation, detoxification, DNA-damage detection, growth factors, inflammasome, inflammation pathway, lipid synthesis and metabolism, pattern recognition, and solute transport (Supplementary Table 1-2, for the list of genes included in the ileum and colon array cards, respectively). For studies in DNBS treated Il10 −/− mice, gene expression profiling was performed using the TaqMan OpenArray Mouse Inflammation Panel plate (Life Technologies) designed by the manufacturer, that consists of 632 gene targets selected for their involvement in inflammatory responses. The cDNA (10 µl) was mixed with the TaqMan OpenArray Real-Time PCR Master Mix and loaded onto the cards using the AccuFill ™ System. The cards were cycled in an OpenArray NT Cycler System (Applied Biosystems) at the integrative microgenomic platform (@BRIDGe, INRA, Jouy-en-Josas) following the manufacturer's protocol. The same sample was systematically loaded in each TaqMan OpenArray and used to check the reproducibility between the plates. Data were extracted using OpenArray Real-Time qPCR Analysis software (Applied Biosystems). The fold-change in gene expression (Rq or relative quantification) was calculated using the comparative 2 −ΔΔCq method with global normalization of all gene expression data using GenEx software (Multid Analyses) 27 . The GF group was used as a calibrator for the gnotobiotic experiments and the Control group for gene expression profiling of the DNBS treated CV IL10 −/− mice.

Principal component analysis, heatmap representation, and statistical analysis of the TaqMan OpenArray data. The R statistics environment was used for data analyses. Data obtained from TaqMan
OpenArray experiments are represented either by Principal Component Analysis (PCA) plots or heatmaps. Rq values were used to build the PCA plots and heatmaps. Prior to PCA, we filtered the data using a one-way permutation test (oneway_test, R package "coin") to remove genes from the datasets for which the variation in expression was less informative (threshold set to p < 0.01). PCA was then carried out on the filtered dataset using the R packages FactoExtra and FactoMineR. The number of genes included in the PCA analysis is specified in the figure legend for each comparison. Heatmaps were generated for genes showing a significant difference between the mono-associated mice and the GF group (p < 0.05; Man-Whitney test). Clustering on the gene expression profile was applied to the genes in the heatmap.
Single real-time quantitative PCR analyses of gene expression. Single real-time quantitative PCR assays were used to confirm the results obtained on the TaqMan Open Arrays system using the corresponding TaqMan assay. All gene expression results are expressed with the 2 −∆∆Ct method (Rq), using gapdh as the housekeeping gene and values from the GF (gnotobiotic experiments) or Control (Chronic colitis experiments) mice as calibrators 27 .

Ex vivo intestinal permeability measurements.
After removal, segments of the ileum and colon were immediately mounted in Ussing chambers, as previously described 5 . Para-cellular permeability was further assessed by measuring mucosal-to-serosal flux of 4 kDa non metabolizable fluorescein isothiocyanate-labeled dextran (FD4) or 0.4 kDa FITC-sulfonic acid (FSA) for 90 min. Molecules were added to the mucosal side of the chamber at a final concentration of 0.4 mg/mL and the fluorescence intensity determined at the serosal side. Trans-epithelial conductance was measured by clamping the voltage and recording the change in the short-circuit current (Isc). At the end of the experiment, tissues were challenged with the cholinergic analog carbachol (CCh) on the serosal side (100 mM) and the ΔIsc was recorded to check the viability of the tissue.
Myeloperoxidase activity. Myeloperoxidase (MPO) activity was assayed according to 25 . MPO activity is expressed as the units per milligram of total protein. Lowry's method was used for protein quantification.
16S rRNA gene sequencing and analysis. Cecal DNA of CV IL10 −/− mice was extracted as described in 28 and the V3-V4 hyper-variable region of the 16S rRNA gene amplified with the primers F343 (CTTTCC CTACACGACGCTCTTCCGATCTTACGGRAGGCAGCAG) and R784 (GGAGTTCAGACGTGTGCTCTTCC GATCTTACCAGGGTATCTAATCCT). The PCR amplicons were sent to the GeT-PlaGe platform (INRA, Toulouse) for sequencing using Illumina Miseq technology. Single multiplexing was performed and the multiplexes purified and loaded onto the Illumina MiSeq cartridge according to the manufacturer's instructions. Raw sequences were analyzed using the FROGS pipeline to obtain the OTU (operational taxonomic units or phylotypes) abundance table, phylogenetic tree, and taxonomic table using the default parameters 29 . Statistical analyses of the 16S rRNA sequences were performed from the FROGS-generated outputs using R and environment version 3.2.3. Calculations of within-community diversity (α-diversity), between-community diversity (β-diversity), and relative abundance taxonomic summaries were performed using the add-on package "Phyloseq". Statistical analyses. R 3.3.1 and GraphPad 5 software were used to produce graphs and for statistical analyses. Data are expressed as the mean ± standard deviation in scatter plots. Differentially expressed genes from OpenArray data were obtained with non-parametric multiple comparison tests using false discovery rate (FDR) corrected p-values (the threshold for significance of differential expression was set to p < 0.05). Comparisons of other quantitative variables were performed using the non-parametric Mann-Whitney test (p < 0.05 was considered significant).

Results
Ileal and colonic gene expression profiles of CEC or Nissle mono-associated mice differ from those of GF mice. We carried out gene expression profiling by high-throughput open-array qPCR of the ileum and colon of CEC WT , Nissle WT , and GF WT mice ( Fig. 1) using two customized gene-expression array plates. One was designed for the ileum and the other the colon, with a selection of candidate genes mainly involved in mucosal defence of the intestine (Supplementary Tables 1, 2). PCA revealed a clear separation of the gene expression profile between mono-associated and GF WT mice, both in the ileum (Fig. 1A) and colon (Fig. 1B). Accordingly, in the ileum, PC1 separated the mono-associated CEC WT and Nissle WT groups from the GF WT group, with both mono-associated groups overlapping. In the colon, PC1 also separated the gene expression profiles of the GF WT and mono-associated groups. The mono-associated CEC and Nissle groups tended to slightly diverge along PC2, revealing a stronger effect of the CEC strain on the colonic gene expression profile. We also included CV WT mice for comparison with the CEC WT , Nissle WT , and GF WT mice ( Supplementary Fig. 2). As expected, PCA first separated the CV WT from all gnotobiotic groups along PC1. In the ileum and colon, the CEC WT and Nissle WT groups clustered together and were distinct from the GF WT and CV WT groups ( Supplementary Fig. 2).
We next selected a subset of genes from high-throughput qPCR data that were differentially expressed between the GF WT and mono-associated groups and exhibited a Rq value > 1.5. Gene expression variation were plotted in a heatmap (Fig. 1C for the ileum and Fig. 1D for the colon). Again, there was a clear impact of the CEC and Nissle strains on the gene expression profile at both sites of the intestine (Fig. 1C,D). Differences in gene expression were confirmed in single TaqMan assays (Examples given in Supplementary Fig. 3 for the ileum and Supplementary Fig. 4 for the colon). Genes involved in several key intestinal functions in the ileum and colon of mice mono-associated with CEC or Nissle were upregulated. Indeed, genes were upregulated for: i) enzymes involved in ROS/RNS turnover, such as duox2 (dual oxidase 2), duoxa2 (dual oxidase activator 2), and nos2 (nitric oxide synthase 2) for both the ileum and colon; gpx2 (glutathione peroxidase 2) for the ileum only; and nox1 (NADPH oxidase 1) for the colon only; ii) antimicrobial peptide production and barrier functions: reg3-γ, reg3-β (regenerating islet-derived-γ and -β) and pla2g2a (secretory phospholipase A group IIA) for both the ileum and colon, ang4 (angiogenin-4) for the ileum only, and fut2 (α(1,2)-fucosyltransferase) for the colon only; and iii) factors involved in the immune response, such as IL18 (interleukin-18), cxcl10 (C-X-C motif chemokine ligand 10), tnfα (tumor necrosis factor α) and tap1 (transporter associated with antigen processing) for both the ileum and colon; IL2rb, IL4ra and IL2rg for the ileum only; and IL1β, IL6, IL17d, thpo (thrombopoietin), and ifnγ (interferon-γ) for the colon only (Fig. 1C,D) and ( Supplementary Figs 3 and 4). Furthermore, CEC or Nissle modulated genes involved in the transport function of the intestine, such as the fructose transporter slc2a5 (solute carrier family 2 member 5; previously called glut5) and members of the aquaporin water channel family aqp3 and aqp7, which were dramatically upregulated in the ileum of CEC WT and Nissle WT mice, and the gene clca4 (Ca 2+-activated chloride channel 4), which was upregulated in the colon (Fig. 1C,D) and ( Supplementary Figs 3 and 4). www.nature.com/scientificreports www.nature.com/scientificreports/ CEC and Nissle maintain intestinal barrier function and epithelial integrity in gnotobiotic WT mice. We further investigated the impact of the CEC and Nissle strains on the barrier formed by the network of intestinal mucins, as it is a key player in physico-chemical protection. We examined the thickness of the mucus layer formed by the highly glycosylated secreted mucin2 (Muc2), and the abundance of membrane-associated mucins, Muc13 and Muc4. Muc2 staining showed the ileal mucus layer to be thicker in mono-colonized CEC WT and Nissle WT mice than GF WT mice, tending to become similar to that of CV WT mice ( Fig. 2A,B). Similarly, Muc13 staining in the ileum of both CEC WT and Nissle WT mice was stronger than that of control GF WT mice (Fig. 2C), but not in the colon (Supplementary Fig. 5A). There were also no modifications in Muc4 staining for any of the groups (Supplementary Fig. 5B). We also investigated the effect of the two strains on expression of the epithelial proliferation marker Ki67 (Fig. 2D). In the Nissle WT group, the percentage of Ki67 + cells was higher than that in the GF WT group in both the ileum and colon (Fig. 2D), whereas there was a significant increase in the percentage of proliferative epithelial cells in only the colon in the CEC WT group (Fig. 2D).
We then assessed the effect of E. coli strains on intestinal permeability in the ileum and colon. We monitored the para-cellular passage from the mucosal to serosal side of high-and low-molecular weight molecules, FD4 (Fig. 2E) and FSA (Fig. 2F), respectively, in an Ussing chamber. There were no differences in permeability between the groups. Among the panel of tight junction or cell adhesion proteins we tested [claudins-4,-5,-7,-8,-12,-15, -17, ZO-1 and F11r (data not shown), Cld2 and Cld3 ( Supplementary Fig. 5C,D)], we only observed greater immunostaining of cadherin-1 (Cdh1) in the CEC WT than the GF WT mice (Fig. 2G,H). These data show that intestinal barrier function and epithelial integrity are preserved in gnotobiotic WT mice.
Key genes involved in mucosal defense and immune response are strongly mobilized in IL10 −/− mono-associated mice. Gnotobiotic IL10 −/− mice have been previously used as a model to investigate the colitis-inducing potential of individual bacterial strains, including indigenous strains of E. coli 17,18 . The colitis that IL10 −/− mice develop requires microbial exposure, as GF mice are protected from disease 17,18 . We investigated the effect of CEC and Nissle in mono-associated IL10 −/− mice, three weeks post-colonization, a time interval previously used to observe the effect of colitogenic indigenous E. coli strains in 129S6/SvEv IL10 −/− mice 18 . There was no weight loss nor weight differences between GF IL10 , CEC IL10 , and Nissle IL10 mice (data not shown). There was also no histological mucosal damage to the intestines of CEC I1L0 or Nissle IL10 mice ( Supplementary Fig. 6).
We further investigated the effect of CEC and Nissle on the ileal and colonic gene expression profile in gnotobiotic IL10 −/− mice. We first randomly sampled four of the 10 individuals belonging to the GF IL10 and CEC IL10 groups to assess the ileal gene expression profile using the customized gene-expression array plates. The data obtained were then compared to those previously obtained for GF WT and CEC WT mice. PCA showed the CEC IL10 group to strongly diverge from the others along PC1 (Fig. 3A). There was a striking difference of the gene expression profile from that of the CEC WT group, and the difference between the CEC IL10 and GF IL10 groups was higher than that between the CEC WT and GF WT groups. This initial screening suggests that the intestinal responses to E. coli of IL10 −/− mice are different from those of WT mice. Indeed, CEC and Nissle may have a stronger effect on the expression of ileal genes in IL10 −/− than WT mice, such as shown for ileal reg3β and nox1 by qPCR single assays (Fig. 3B).
We investigated whether this difference in the effect of CEC and Nissle between IL10 −/− and WT mice also occurred in the colon by targeting key genes involved in mucosal defense using single qPCR assays (Fig. 3C). Thus, we investigated colonic expression of genes involved in the ROS/RNS turnover (duox2, nox1, nos2), antimicrobial peptide production (reg3γ and pl2g16), the mucosal intestinal barrier (fut2), immune response (IL1β,IL22), as well as stat1, which orchestrates antimicrobial responses and nod-like receptors 5 (nlrc5), an intracellular protein involved in the detection of microbes. These genes were more highly expressed in IL10 −/− than WT mice, when they are mono-associated with CEC or Nissle (Fig. 3C).
The differential effect of the strains on IL10 −/− and WT mice was not due to colonization levels, as both strains similarly colonized the WT and IL10 −/− mice at a level of 10 10 bacteria/g of stool ( Supplementary Fig. 7). Overall, these results show that genes involved in intestinal defense mechanisms are strongly mobilized in the presence of CEC and Nissle in IL10 −/− mice.

Intestinal barrier function is not altered in Il10 −/− mice mono-colonized with CEC and Nissle.
We assessed different markers to investigate whether intestinal barrier function, at the ileal and colonic level, is modified in Il10 −/− mice mono-colonized with CEC and Nissle: Ki67 (Fig. 4A,B), cadherin1 (cdh1), a protein involved in cell adhesion (Fig. 4C,D) and the thickness of the mucus layer (Fig. 4E,F). The number of proliferative epithelial cells in the ileum and the colon of IL10 deficient mono-colonized mice was higher than that of the GF IL10 group (Fig. 4A,B for the ileum and colon respectively). Ileal Cdh1 staining was greater in mono-colonized IL10 deficient mice than in GF I1L0 mice (Fig. 4C,D). Furthermore, the mucus layer in the ileum was thicker in the 2 mono-colonized groups than in GF IL10 mice, similar to our observations for WT mice (Fig. 4E,F). But in contrast to WT mice, this increase also occurred in the colon of IL10 −/− mice (Fig. 4E,F). Intestinal permeability was assessed in vitro at the level of the ileum and the colon. No difference in FSA passage among the groups was observed (Fig. 4G). In contrast, the ileal transepithelial conductance of CEC IL10 and Nissle IL10 mice was lower that than of the GF IL10 mice (Fig. 4H), suggesting a tendency to a lower intestinal permeability. Taken as a whole, this data reveals that CEC and Nissle do not alter intestinal permeability in a model of predisposition to inflammation. These strains can even impact positively on some markers of epithelial barrier function.
CEC and Nissle partially reduce the severity of inflammation of DNBS-treated CV IL10 −/− mice. We next investigated the effect of CEC on inflammation using DNBS-treated CV IL10 −/− mice as a model of a chronic intestinal colitis. Although the Ameho score was unexpectedly close to zero (Supplementary Fig. 8 www.nature.com/scientificreports www.nature.com/scientificreports/ a decrease in body weight (Fig. 5A), shortening of the small intestine and colon (Fig. 5B), increased myeloperoxidase (MPO) activity, a marker of neutrophil infiltration (Fig. 5C), infiltration of immune cells (CD3 + T lymphocytes) (Fig. 5D,E), a pro-inflammatory gene expression profile (Supplementary Fig. 9A,B). www.nature.com/scientificreports www.nature.com/scientificreports/ www.nature.com/scientificreports www.nature.com/scientificreports/ Although CEC and Nissle were unable to induce recovery of the body weight in the model (Fig. 5A), both strains rescued the inflammation-associated reduction of ileal and colon length and MPO activity to control values (Fig. 5B,C). CEC had an additional positive effect on the number of ileal and colonic CD3 + T lymphocytes (Fig. 5D,E). Moreover, the permeability of the intestinal barrier did not increase after CEC treatment ( Supplementary Fig. 10).
The gene expression profile of E. coli DNBS-treated groups was skewed towards an anti-inflammatory profile, close to that of the control group. We then analysed the inflammatory profile in Control (negative control for the disease), DNBS IL10 , and DNBS IL10 + CEC/Nissle mice by focusing on a panel of genes involved in inflammation. Genes for which the resulting Rq values showed them to be significantly modified by DNBS treatment and rescued by E. coli strains, were selected and the corresponding Rq values plotted in a heatmap for the ileum and colon (Figs 6A and 7A).
The gene expression profile in the ileum of the control group was partially restored when DNBS IL10 mice were treated with the commensal E. coli strains CEC or Nissle (Fig. 6A). Thus, single TaqMan assay experiments confirmed that increased expression of interleukins such as IL6, IL17α and ltf (lactotransferrin) observed for DNBS IL10 relative to those of the control group was counteracted when mice received CEC or Nissle (Fig. 6B). www.nature.com/scientificreports www.nature.com/scientificreports/ Similarly, the gene expression profile observed in the colon of DNBS IL10 + CEC was close to that of the control group (Fig. 7A). The effect of CEC was stronger than that of Nissle (Fig. 7A), with the recovery of Control expression levels for genes, such as the AMPs (camp, pla2g2e, reg3α, and reg3β) and several cytokines (tnfα, cxcl11, IL22 and IL6), following E. coli treatment. Single TaqMan assays confirmed that the increased expression of the a2m (alpha-2-macroglobulin), IL6, IL17α, orl1 (oxidized low density lipoprotein receptor 1), clu (clusterin) and cfi (complement component factor I) genes following DNBS treatment decreased to Control values when the DNBS treated mice received CEC (Fig. 7B).
The effect of CEC on gene expression is independent of an effect on gut microbiota composition. We investigated whether the effect of the E. coli strains involved modifications of the gut microbiota by analyzing the intestinal bacterial composition of the different experimental groups. We found no differences in the α-diversity, which measures the taxonomic richness of the cecal microbiota communities, between groups (Fig. 8A). In addition, β-diversity analysis, which measures the degree of similarity between the gut microbial communities, revealed no clustering of the mice according to the DNBS or E. coli treatments (Fig. 8B). At the phylum level, there was a decrease in the relative abundance of Firmicutes and an increase in that of Proteobacteria in the DNBS IL10 mice relative to that of the Control group, revealing dysbiosis in our model (Fig. 8C). Neither CEC nor Nissle treatment corrected the changes in microbiota composition induced by DNBS treatment.  www.nature.com/scientificreports www.nature.com/scientificreports/

Discussion
Most studies demonstrating that several human diseases are associated with an expansion in the gut of Enterobacteriaceae, including E. coli, are based on sequencing data. However, this approach has insufficient resolution to detect genetically similar organisms of differing ecophysiology and impact on health. Considering phenotypic diversity at the strain level would aid the assessment of how resident E. coli may affect gut health and disease outcomes, as demonstrated by a recent study 30 .
Here, we extend our knowledge on a commensal E. coli strain that we previously isolated from suckling rodents, that we named CEC15, showing dynamic transcriptional responses of the ileum and the colon to CEC15. The expression of genes that play a key role in mucosal defence and the maintenance of mutualistic host-microbiota interactions was elevated in CEC mono-associated mice, similar to observations made with the probiotic Nissle 1917. Indeed, we show a core-response to both CEC and Nissle strains, in both the ileum and colon. This includes a set of genes involved in ROS turnover: duox2 and nox1, both generating extracellular H 2 O 2 ; the expression of duoxa2, required for the maturation of functional duox2, also increased, consistent with increased duox2 expression. Expression of the AMPs, reg3β and reg3ϒ, ang4, and pla2g2a also increased, as well as that of fut2. We further demonstrated that components of the mucosal immune system recognize and react to www.nature.com/scientificreports www.nature.com/scientificreports/ the presence of commensal E. coli by increasing their expression, consistent with previous results that showed that GF mice colonized with members of the gut microbiota undergo immune system activation and development 31 .
Previous studies have shown that the intestinal expression of duox2 32 , members of the reg3 family 33 , and fut2 34 are upregulated by the gut microbiota in mice and that certain single commensal or probiotic bacterial strains can contribute to this effect [33][34][35] . These are key components of the innate immune system that help to fight against pathogens [36][37][38] while preserving the symbiotic nature of the relationship between the gut microbiota and the host. In mice, duox2-generated H 2 O 2 and members of the reg3 family play a key role by spatially segregating indigenous bacteria, thereby dampening microbiota-induced mucosal immune responses 35,36 . Fut2 is involved in the fucosylation of glycol-conjugates expressed on epithelial cells, providing a source of host-derived complex carbohydrates for the gut microbiota. The induction of fut2 plays a role in colonization resistance against pathogens by restoring commensal diversity 38 .
Genes modulated in mono-associated mice include those that encode Ca 2+ -activated chloride channel 4, (clca4), the fructose transporter Slc2a5 (previously called GLUT5), and members of the water and glycerol channel aquaglyceroporin family (AQP -3 and -7), suggesting that the primary functions of the intestine may also be modulated in mono-associated mice 39 . CLCA4 expression is downregulated in UC, suggesting that the control of electrolyte balance may be part of the defence mechanism against luminal microbes 40 . As recently shown, AQP3 transports H 2 O 2 generated at the cell surface by NOX1 and DUOX2 to mediate signal transduction in colonic epithelia 41 . This further argues for a central role of E. coli strains on ROS turnover and signalling.
We tested the effect of the CEC strain in IL10 −/− mice, a well-recognized model for immune-mediated colitis, given its relevance to human IBD 42 . We speculate that the impact of the genetic background was low in the models we used. Indeed, the genetic background of wild type GF mice was C57BL/6 and that of GF IL10 −/− mice was a mix of C57BL/6 and 129/Ola. In our study, we found that, in the absence of IL10, the host response to commensal CEC and Nissle was stronger 21 days post-inoculation than in WT mice, as observed for duox2, duox2a, nox1, nos2, reg3β, reg3ϒ, and fut2 expression. We also observed strong activation of immune responses, as observed for cxcl10, tap1, il1β, and IL22. In particular, neither the expression of stat1 nor that of nod-like receptor 5 (nlrc5), were modified in WT mice by the presence of the two bacterial strains, whereas they were four to eight fold higher in mono-associated IL10 −/− than GF IL10 mice. A previous study suggested that IL10 may be involved in the control of the homeostatic relationship between indigenous strains of E. coli and the host 43 . In our study, intensification of the host response may have compensated for the absence of IL10.
We found that nos2 gene expression can be altered by the bacterial status of mice as seen in the difference between the CV WT and GF WT models. Both CEC and Nissle also upregulated nos2 expression in the WT mice but to a lesser extent than in the CV WT group. This E. coli related increase of nos2 is intensified in Il10 −/− mono-colonized mice. The role of intestinal nos2 in E. coli growth has been demonstrated in an inflammation setting. Previous studies have shown an elevated intestinal expression of nos2 during the inflammation process and that the host-nos2-derived by-products of reactive nitrogen species contribute to the proliferation of E. coli 44 . However, the physiological role of nos2 in this context has not been fully investigated and needs further studies.
We observed no evidences for intestinal inflammation in the mono-associated IL10 −/− mice, neither in the ileum or colon, despite a high level of colonization. Furthermore, the onset of inflammation was previously shown to be preceded by increased ileal and colonic permeability in the IL10 −/− mouse model 45 . In contrast, we found that some parameters of ileal and colonic permeability and integrity were improved in the CEC and Nissle mono-associated IL10 −/− groups, although we did not find any major differences in the passage of molecules. Thus, we found decreased electrical conductance associated with stronger staining of cadherin-1 in the ileum of IL10 −/− mice mono-colonised with E. coli. In addition, we found the epithelial integrity to be preserved, based on the positive effect of the strains on an epithelial proliferative marker. In addition, the ileal expression of Muc13 increased. Finally, we showed that CEC and Nissle can induce an increase in the thickness of the mucus layer in the ileum, which extended to the colon in the IL10 −/− mice.
We further investigated the effect of strain CEC in a disease setting using a model of conventional Il10 −/− mice exposed to DNBS. As previously mentioned, the IL-10 deficient mice model shares similarities with human IBD patients 42 . As the onset of colitis can vary among mice according to several factors 46 , a DNBS administration protocol, previously validated in our lab 25 , was used to synchronize colitis in this model. Our data show a beneficial role of CEC in promoting gut homeostasis upon mucosal injury in IL10 −/− mice. Although CEC was unable to reverse DNBS-associated weight loss, it attenuated several types of DNBS-induced damage, and reversed the gene expression profile towards that of the control group, both in the ileum and colon. The beneficial effect of CEC appeared to be stronger than that of the Nissle strain in the colon, based on the gene expression profile.
Several studies have reported that Nissle has an inhibitory effect on other E. coli species [47][48][49] . Regarding our 16S data, we did not observe any impact of CEC or Nissle on the gut microbiota composition. However, as 16S data analysis has insufficient resolution to distinguish genetically close organisms, we cannot exclude a remodelling of E. coli population by the administration of CEC15 or Nissle.
Our results that show a beneficial effect of CEC compared to those of other studies, strongly suggest that different E. coli strains of the gut microbiota may differ dramatically in their colitogenic or probiotic potential. For example, indigenous E. coli strains have a colitis-inducing potential in a genetically predisposed model for inflammation, whereas others have no negative impact 19 . The genetic determinants underlying such divergent behaviour are currently far from being understood. Preliminary data show that the genome of the CEC strain does not contain the genomic pks island (data not shown). This is in contrast to the Nissle strain, for which the pks island was shown. This gene cluster encodes non-ribosomal peptide synthetases (NRPS) and polyketide synthetases (PKS) and produces colibactin (a peptide-polyketide genotoxin) that can induce DNA damage by inducing double-strand breaks 50 . However, the role of pks in gut health is a matter of debate. Indeed, the probiotic properties of Nissle are linked to the pks island, more precisely to a lipopeptide encoded in this this region 51 . Sequencing of the CEC15 genome and genomic comparison of other commensal E. coli strains with a known