Enterobacter Ludwigii Protects Experimental Colitis Through CD103+DC and Treg Cells

Background: Intestinal bacteria are closely related with inammatory bowel disease (IBD), and regulatory cell-mediated immune tolerance is important to inhibit IBD. Commensal intestinal bacteria play key roles in regulating immune tolerance cell, however, bacterial strains directly involved in this regulation remain to be identied. Results: In the present study, metronidazole, among nine antibiotics, was found to have the best effect on protecting mice against DSS-induced colitis. Enterobacter ludwigi, abundant in mouse feces after metronidazole treatment, was identied to decrease mice susceptibility to DSS-induced colitis with or without the presence of complex intestinal bacteria. E. ludwigii gavage increased CD103 + DCs and Foxp3 + Treg cells in intestinal microenvironment, and effects of E. ludwigii on diminishing colitis were lost in DC or Treg depletion mice. CD103 + DCs isolated from E. ludwigii-treated mice showed enhanced ability to promote the Treg differentiation from naive T cells. DCs, directly stimulated by live E. ludwigii strain or its culture supernatant, had increased immune tolerance ability for Treg differentiation in vitro. Conclusions: Overall, our ndings identify a facultative anaerobe bacterial strain E. ludwigii, which directly enhances CD103 + DC and Treg-mediated immune tolerance, resulting in protecting mice against DSS-induced colitis.


Introduction
In ammatory bowel disease (IBD), including Crohn's disease (CD) and ulcerative colitis (UC), spreads globally with an incidence rate of approximately 0.3% [1]. Pathogenesis of IBD involving genetic factors and environmental triggers such as microbiome, remains to be elucidated [2].
Gut microbiota alteration is closely associated with IBD occurrence [3], and gut bacterial diversity is reduced in feces and colonic mucosa of IBD patients compared with that of patients in remission [4,5]. Several kinds of bacteria including Bi dobacterium species, Lactobacillus species, Clostridium species, Bacteroides species, Faecalibacterium prausnitzii, Roseburia species, Suterella species, Akkermansia muciniphila, and Eubacteriumhallii, most of which are obligate anaerobic bacteria, have been reported to play roles in reducing in ammation during IBD [6,7]. Aerobes or facultative anaerobes are always hypothesized to contribute to IBD pathogenesis.
Antibiotics decrease the number of luminal bacteria and change the composition of intestinal ora [8].
Rifaximin is effective at inducing CD remission, which increases bene cial bacteria such as Bi dobacteria and F. prausnitzii [9]. The combination of metronidazole and cipro oxacin improved symptoms of perianal diseases [10], meanwhile metronidazole and ornidazole successfully prevented clinical postoperative recurrence of Crohn's disease [11,12]. The antibiotic combinations, such as metronidazole/tobramycin, metronidazole/amoxicillin/tetracycline, and metronidazole/tobramycin/vancomycin/rifaximin, show bene t of remission induction in moderate UC [13,14]. However, as antibiotics have side effects on patients and impact on intestinal ora, use of antibiotics is limited in IBD patients [15].

In ammatory response during IBD is caused by induced Th1/Th17 effector cells and impaired regulatory cells including regulatory T cells (Treg), B cells (Breg), M2 macrophages, dendritic cells (CD103 + DCs) and
innate lymphoid cells (ILCs) [6,16,17]. Regulatory cell-mediated immune tolerance to foreign antigens and autologous proteins in the colon is important to inhibit IBD [18]. Tregs play a key role in immunological tolerance through secretion of inhibitory cytokines, and Tregs could disrupt metabolic processes in T effector lymphocytes (Teffs), induce apoptosis of Teffs, trigger cytotoxic activity against Teffs, neutralize DC function, and produce amphiregulin to repair tissue [19]. CD103 + DC promotes Treg differentiation and inhibits Th1/Th17 immune response in mesenteric lymph nodes (MLN) by producing TGF-β, retinoic acid, indoleamine-2,3-dioxygenase (IDO), AhR ligands and carbonic anhydrate I epitope peptide [6,20]. Only a few bacteria have been identi ed to directly induce immune tolerance cells. Due to the presence of a large number of intestinal bacteria, strains involved in immune tolerance and their mechanism remain to be clari ed.
In the present study, a facultative anaerobe Enterobacter ludwigii isolated from metronidazole-treated mouse feces, was found to enhance the ability of CD103 + DC to differentiate Treg cells, which increased the CD103 + DC/Treg-depended tolerance response to reduce mice susceptibility to DSS-induced colitis.

Mice susceptibility to DSS-induced colitis is in uenced by different kinds of antibiotics
To study the role of intestinal bacteria in in ammatory bowel disease, we used a mouse model of DSSinduced acute colitis combined with antibiotic treatment. Male C57BL/6J mice were pretreated with different antibiotics for 7 days (ampicillin, chloramphenicol, cipro oxacin, clindamycin, metronidazole, neomycin, polymycin, tetracycline, vancomycin, a broad-spectrum antibiotic cocktail composed of ampicillin, vancomycin, neomycin and metronidazole (ABX), and no antibiotic treatment as the control), and subsequently exposed to the same antibiotic plus 3% DSS for another 7 days (Fig. S1A, additional le 1). The colitis severity was evaluated through body weight loss, disease activity index (DAI) scores, colon lengths and histopathological scores. ABX treatment, which is widely used to clear normal intestinal bacteria, greatly reduced mice susceptibility to DSS treatment as reported previously [21] (Fig.   S2A to D, additional le 2). Among the antibiotics used, metronidazole, vancomycin and clindamycin were observed to decrease mice susceptibility to DSS-induced colitis signi cantly, with less body weight loss, lower DAI scores, longer colon lengths and lower histopathological scores compared with the other antibiotic treatment and the control group ( Fig. S1B to E, Fig. S2A to H, additional le 1 and 2). Metronidazole treatment, compared with vancomycin or clindamycin treatment, had the best effect in reducing colitis ( Supplementary Fig. 2I). In addition, other antibiotic treatment was not found to signi cantly attenuate the severity of DSS-induced colitis ( Fig. S1B to E, Fig. S2A to H, additional le 1 and 2).
Taken together, these data indicate that different kinds of antibiotics play different roles in mice susceptibility to DSS-induced acute colitis, and metronidazole treatment greatly attenuated DSS-induced colitis. As different antibiotic treatment could shape different intestinal bacteria composition, we hypothesized that metronidazole-induced intestinal bacteria may play roles in protecting mice against experimental colitis.
Metronidazole treatment increases CD103 + DCs and Foxp3 + Treg cells in intestinal microenvironment CD103 + DCs are reported to induce Treg cell differentiation, resulting in intestinal homeostasis [22]. Therefore, we examined the effect of metronidazole treatment on CD103 + DCs and Treg cells in mesenteric lymph nodes (MLN) and colonic lamina propria (cLP) using ow cytometry. In MLN, the percentage of CD103 + DCs in total DCs was signi cantly increased in the metronidazole group compared with that in the control and neomycin group (Fig. 1A), and a trend toward higher percentage of CD103 + DCs in total DCs was observed for the metronidazole group without statistical signi cance in cLP ( Fig. 1B). The percentage of CD25 + Foxp3 + Tregs in CD4 + T cells in the metronidazole group was increased compared with that in the control and neomycin group in MLN, and compared with that in the control group in cLP ( Fig. 1C to D). In addition, we examined the IFN-g + or IL-17A + CD4 + cells in cLP, which are reported to enhance intestinal in ammation and can be regulated by Treg cells [23,24]. The percentage of either IFN-g + or IL-17A + CD4 + in total CD4 + cells was decreased in the metronidazole group compared with that in the neomycin group in cLP, and a trend toward lower percentage without statistical signi cance was observed for the metronidazole group compared with the neomycin group in MLN ( Fig.   S3A to D, additional le 3). These results suggest that metronidazole treatment induces CD103 + DCs and Foxp3 + Treg cells in intestinal microenvironment, which is correlated with reduced experimental colitis in mice.

Analysis of bacteria remodeled by metronidazole treatment
To characterize the potential bacterial population related with the reduced susceptibility to DSS-induced acute colitis in mice treated with metronidazole, we performed 16S rRNA gene sequencing of bacterial DNA collected from mice feces on the zero and the fth day during DSS treatment, and bacterial DNA collected from colon mucosal tissues on the seventh day, in the mouse model of DSS-induced acute colitis combined with antibiotic treatment described above (Fig. S1A, additional le 1).The 44 samples generated a total of 1888804 high-quality 16S rRNA sequences, and the OTUs were obtained based on an equal sequencing depth (27528 reads per sample) to be used for further analysis. The bacteria of mice treated with metronidazole had decreased alpha diversity (based on Shannon's index) compared with those of mice in the neomycin or control group ( Fig. 2A). As lower bacterial alpha diversity is related with increased intestinal in ammation [25], it is suggested that bacterial alpha diversity is not involved in different mice susceptibility to DSS-induced colitis in this mouse model. The beta diversity of bacteria based on Bray-Curtis revealed that bacteria in mice treated with metronidazole were different from those in mice of the neomycin or control group, and similar bacteria in feces and colon mucosal tissues were examined. Male C57BL/6J mice were given 10 9 CFU of each strain via oral gavage for 7 days, and were subsequently given the same strain plus 3% DSS for another 7 days (water gavage instead of bacteria was used in the control group) (Fig. S4B, additional le 4). Among the six strains, E. ludwigii gavage showed lower DAI scores, longer colon lengths and lower histopathological scores in mice treated with DSS compared with the control group, and L. johnsonii showed lower histopathological scores ( Fig. S4C to F, additional le 4). These results suggest that E. ludwigii plays a role in protecting mice against DSSinduced acute colitis.
We also check whether E. ludwigii mixed with other isolated bacterial strains could enhance its effect on DSS-induced acute colitis. Male C57BL/6J mice were given strain mixture (the percentage of each strain was same, and a total of 10 9 CFU was used) for 14 days, and 3% DSS were added in drinking water from the seventh day (Fig. S5A, additional le 5). However, no obvious protection effect of the mixture of different strains with E. ludwigii was found compared with the control group ( Fig. 5B to E, additional le 5), indicating that a large number of other bacterial strains in the mixture attenuate the effect of E. ludwigii.
In addition, male C57BL/6J mice were given 10 9 CFU of E. ludwigii and 3% DSS at the same time for 8 days (Fig. S5F, additional le 5), and mice in the E. ludwigii group had less body weight loss, lower DAI scores, longer colons and lower histological scores compared with the control and E. gallinarum group ( Fig. S5G to J, additional le 5), indicating that E. ludwigii could reduce the severity of DSS-induced colitis without pretreatment.
In order to examine the major effect of E. ludwigii on DSS-induced acute colitis, mice were pretreated with ABX for ve days, and were given 10 9 CFU of E. ludwigii, E. gallinarum, or L. johnsonii via oral gavage one day after ABX treatment for 5 days, and were subsequently given the same strain in addition with 3% DSS and LPS challenge (8 mg/kg body weight, LPS is used to stimulate colitis in the absence of commensal bacteria [26]) for 7 days (Fig. 3A). E. ludwigii gavage showed less weight loss compared with the E. gallinarum and L. johnsonii group (Fig. 3B). Notably, E. ludwigii gavage showed a signi cantly lower DAI score compared with the E. gallinarum, L. johnsonii, and the control group, beginning at the fth day of DSS challenge and lasting until the mice were euthanized (Fig. 3C). In addition, mice with E. ludwigii gavage had longer colon lengths and lower histopathological scores with preserved crypt structures and less in ammatory cells in ltration, compared with the other three groups ( Fig. 3D to E). We noticed that mice with L. johnsonii gavage had moderately higher histopathological scores compared with the control group ( Fig. 3E), although L. johnsonii gavage showed lower histopathological scores in mice without ABX treatment (Fig. S4F, additional le 4), which suggests that L. johnsonii decreases DSS-induced histopathological scores in mice depending on other bacteria.
Taken together, these results indicate that E. ludwigii abundance induced by metronidazole treatment could protect mice against DSS-induced acute colitis with or without the presence of other intestinal bacteria.

E. ludwigii increases CD103 + DCs and Foxp3 + Tregs in intestinal microenvironment
As metronidazole treatment induces CD103 + DCs and Foxp3 + Treg cells in intestinal microenvironment, we examined the effect of E. ludwigii on CD103 + DCs and Foxp3 + Treg cells using the same mouse model for evaluating single bacterial strain (Fig. 3A). In MLN, the percentage of CD103 + DCs in total DCs cells was signi cantly increased in the E. ludwigii group compared with that in the L. johnsonii and control group ( Fig. 4A). In cLP, the percentage of CD103 + DCs in total DCs cells was obviously increased in the E. ludwigii group compared with that in the E. gallinarum and L. johnsonii group (Fig. 4B). A trend toward higher percentage of CD103 + DCs without statistical signi cance was also observed for the E. ludwigii group compared with the E. gallinarum group in MLN (Fig. 4A), and compared with the control group in cLP (Fig. 4B). In both MLN and cLP, the percentage of CD25 + Foxp3 + Tregs in CD4 + T cells was detected to be higher for the E. ludwigii group than that for the control and E. gallinarum group (Fig. 4C to D). The percentage of CD25 + Foxp3 + Treg cells was higher for the E. ludwigii group compared with that for the L. johnsonii group in MLN (Fig. 4C), and a higher trend without statistical signi cance was observed for the E. ludwigii group compared with that for the L. johnsonii group in cLP (Fig. 4D). In immuno uorescence assays, the proportion of CD103 + DCs in total DCs cells and Foxp3 + Treg cell number were increased in the E. ludwigii group compared with that in the control group in cLP ( Fig. 5A to B).
We also examined the effect of E. ludwigii on CD103 + DCs and Foxp3 + Tregs using the mice that were pretreated with ABX and subsequently given 10 9 CFU of single strain for 5 days without DSS and LPS challenge (Fig. S10A, additional le 10). We found a signi cant increase in CD103 + DC frequency in MLN and cLP of the E. ludwigii group compared with the control and E. gallinarum group ( Fig. S10B to C, additional le 10), and Foxp3 + Treg frequency in cLP was also increased in the E. ludwigii group compared with the control and E. gallinarum group ( Fig. S10D to E, additional le 10). These results suggest that E. ludwigii contributes to maintenance of tolerant immune cells even in intestinal homeostasis.
E. ludwigii gavage induces CD103 + DCs' ability to promote Foxp3 + Treg differentiation To examine the function of CD103 + DCs and Foxp3 + Tregs induced by E. ludwigii in vivo, CD103 + DCs and Foxp3 + Tregs were puri ed from MLNs of mice in the E. ludwigii, E. gallinarum, L. johnsonii and control group in the above mouse model (Fig. 3A) using uorescence-activated cell sorting and were analyzed using quantitative reverse transcription PCR (qRT-PCR). The Tgfb1, Tgfb2, Aldh1a2 and Pdl1 mRNA level of CD103 + DCs, which are usually used to evaluate CD103 + DCs' ability in tolerogenic responses [27][28][29], were increased in the E. ludwigii group compared with those in the control, L. johnsonii or E. gallinarum group (Fig. 5C). The Tgfb1,Tgfb2 and Il-10 mRNA level of Foxp3 + Tregs, which are used to evaluate Tregmediated T cell suppression ability [30], were increased in the E. ludwigii group compared with that in the control, L. johnsonii or E. gallinarum group (Fig. 5D). To further validate that E. ludwigii-induced CD103 + DCs play a role in Foxp3 + Treg differentiation, CD103 + DCs puri ed from MLNs of mice in the E.
ludwigii, E. gallinarum, L. johnsonii or control group were co-cultured with naive CD4 + T cells from spleens of untreated Foxp3-GFP-DTR mice. CD103 + DCs from the E. ludwigii group enhanced conversion of naive CD4 + T cells into Foxp3 + Tregs, compared with those from the other three groups (Fig. 5E). These results indicate that E. ludwigii gavage enhances the ability of CD103 + DCs in immune tolerance to drive Foxp3 + Treg differentiation, resulting in decreased susceptibility of mice to DSS-induced acute colitis.
E. ludwigii directly enhances DCs' ability to promote Foxp3 + Treg differentiation To examine if E. ludwigii had direct effect on DC, we pulsed DCs isolated from C57BL/6J mice MLN with E. ludwigii or E. gallinarum for 2 hours (MOI = 10). The Tgfb1, Tgfb2, Aldh1a2 and Pdl1 mRNA levels of DCs, were increased in the E. ludwigii-challenged group compared with that in the control or E.gallinarumchallenged group (Fig. 6A). It was shown that E. ludwigii culture supernatant, but not heat-inactivated bacteria, could increase Tgfb1, Tgfb2, Aldh1a2 and Pdl1 mRNA levels of DCs ( Fig. 6B to C). We further assessed the regulatory action of E. ludwigii-stimulated DC by coculturing with naive CD4 + CD62L + CD44 -CD25 -T cells for 72h, and the phenotype of polarized T cells was then analyzed by ow cytometry and immuno uorescence. E. ludwigii culture supernatant-stimulated DCs, compared with the control and E. gallinarum group, obviously promoted the differentiation of Treg cells from naive T cells (Fig. 6D to E). Therefore, we conclude that E. ludwigii interacts with DCs directly to increase DCs' immune tolerance ability, which drive Foxp3 + Treg differentiation.

Discussion
The intestinal ora in balance is mainly composed of obligate anaerobes, due to a low level of oxygen tension in the distal part of the gastrointestinal tract [31]. Gut microbiota dysbiosis is usually characterized by expansion of aerobes or facultative anaerobes belonging to Proteobacteria and Fusobacteria phylum, and by reduction of obligate anaerobes belonging to Firmicutes phylum [31,32].
Therefore, aerobes or facultative anaerobes are usually hypothesized to contribute to IBD pathogenesis. In the present study, we found that a facultative anaerobic bacterial strain isolated from metronidazoletreated mouse feces, played a role in reducing mice susceptibility to DSS-induced colitis through enhancing CD103 + DC-mediated immune tolerance.
Metronidazole, as a nitroimidazole preventing nucleic acid synthesis, limits anaerobic bacteria, protozoans and microaerophiles [33]. Metronidazole is medicative in active Crohn's colitis patients and preventive of post operative recurrence of CD in combination with other antibiotics, but has less bene t for ulcerative colitis [34]. Metronidazole is also effective in treating Clostridium di cile-associated colitis [35]. However, prolonged high doses of metronidazole have side effects, such as gastrointestinal disturbance, peripheral neuropathy and encephalopathy, which limited its widespread use in clinical practice [36]. In the acute DSS mouse colitis model, metronidazole treatment could prevent colonic in ammation, but has no therapeutic effect on established colitis [37]. Metronidazole in combination with cipro oxacin treatment resulted in expansion of Enterococcaceae and Lactobacillaceae in mice [33]. In our study, we found Enterobacteriaceae was also abundant after metronidazole treatment, and a bacterial strain E. ludwigii was identi ed to promote immune tolerance through increasing CD103 + DCs' ability to drive Treg differentiation independent of other intestinal ora.
The interactions between commensal bacteria and immune cells are very important to maintain intestinal homeostasis. However, only a few bacteria have been reported to directly interact with intestinal immune cells or epithelia cells to enhance immune tolerance. Segmented lamentous bacteria (SFB) strongly activates non-in ammatory homeostatic Th17 responses in the small intestine [38,39]. B. fragilis could directly induce IL-10-producing Treg differentiation in the gut by activating TLR2 [40,41]. A 15 kDa protein isolated from F. prausnitzii acts on intestinal epithelial cells to inhibit NF-κB pathway to prevent colitis [42]. In the present research, we proved that E. ludwigii directly promoted DCs' tolerance function to enhance Treg differentiation.
E. ludwigii is a motile rod-shaped gram-negative bacterium, and some E. ludwigii isolates with antibiotic resistance genes have been reported to be associated with surgical site infections or nosocomial bloodstream infections [43,44]. Therefore, E. ludwigii strains should be carefully considered for direct application. Our data demonstrate that E. ludwigii culture supernatant has effect on DC, providing the possibility that the secreted components or metabolites of E. ludwigii play direct roles in immune tolerance, which may lead to novel strategies to treat IBD.

Conclusions
E. ludwigii, abundant in metronidazole-treated mouse feces, decreases mice susceptibility to DSS-induced colitis with or without the presence of complex intestinal bacteria, and has both preventive and therapeutic bene t for DSS-induced colitis. E. ludwigii gavage increases CD103 + DCs and Foxp3 + Tregs in intestinal microenvironment with or without DSS stimulation, and effects of E. ludwigii on diminishing colitis depend on DCs and Tregs. E. ludwigii gavage induces CD103 + DCs' ability to promote Foxp3 + Treg differentiation, and direct interactions between E. ludwigii and DCs enhances DCs' immune tolerance ability.

Animals
Wild type male C57BL/6J mice, aged 6-8 weeks, were purchased from Academy of Military Medical Science (Beijing, China). CD11c-GFP-DTR mice were a gift from Professor Xuetao Cao of Nankai University. Foxp3-GFP-DTR mice were a gift from Professor Xiaoming Feng of Institute of Hematology and Hospital of Blood Disease, Chinese Academy of Medical Sciences. All mice were bred and maintained under speci c pathogen free condition and a standard chow diet ad libitum in the animal facility at Tianjin Medical University. Animal procedures and protocols were performed with 6 to 8 weeks male mice and were approved by Animal Care and Use Committee, Tianjin Medical University.

DSS-induced colitis mouse model
To study antibiotic effect on mice susceptibility to DSS-induced colitis, male C57BL/6J mice were pretreated with antibiotics for seven days, and exposed with the same antibiotics in addition with 3% DSS (0216011090, M.W. = 36,000-50,000 Da; MP Biomedicals, Solon, OH, USA) in drinking water for another seven days [46]. Feces were collected on the zero and fth day, and tissues were collected on the seventh day when mice were euthanized.
To test the effect of individual bacterial strain on DSS-induced colitis, the strain was inoculated into 10 ml of sterile BHI broth (E. gallinarum, K. huaxiensis, E. ludwigii, E. lactis and K. michiganensis) or MRS broth (L. johnsonii) containing 2.5% L-cysteine, grown in anaerobic penicillin bottles at 37 °C for 24 h. Cultured bacteria were pelleted by centrifugation (5000 × g for 5 min at 4 °C) and resuspended in PBS to obtain a density of 5 × 10 9 CFU/ml. Each male C57BL/6J mouse was gavaged with 200 μl liquid bacterial strains (10 9 CFU) once per day for fourteen days, and 3% DSS was added to the drinking water of the mice from the seventh day to the fourteenth day during the bacteria administration.
To check the effect of bacterial strain mixture on DSS-induced acute colitis, individual strain was cultured separately in anaerobic penicillin bottles containing BHI broth at 37 °C for 24h and mixed together in the same proportion before gavage. Each male C57BL/6J mice were pretreated with 10 9 CFU strain mixture for seven days, and exposed with the same strain mixture plus 3% DSS in drinking water for another seven days.
To evaluate the therapeutic effect of E. ludwigii on experimental colitis without pretreatment, male C57BL/6J mice were treated with 3% DSS in drinking water for 7 days and gavaged with single bacterial strain once a day during the same period.
To study the major effect of individual bacterial strain on DSS-induced acute colitis, male C57BL/6J mice were pretreated with ABX for 5 days and water for 1 day, then mice were gavaged with 10 9 CFU bacterial strains once per day for twelve days. 3% DSS was added to the drinking water of mice, and mice were gavaged with LPS (8 mg/kg body weight/day, L2880, Sigma-Aldrich, St. Louis, MO, USA) simultaneously from the fth day to the twelfth day during the bacteria administration.

Processing of sequencing data
Illumina sequencing generated raw fastq les, and the sequence reads were then merged, trimmed, ltered, aligned, and clustered by operational taxonomic unit (OTU) using UPARSE (version 7.1 http://drive5.com/uparse/). The taxonomy of each 16S rRNA gene sequence was analyzed by RDP Classi er algorithm (http://rdp.cme.msu.edu/) against the Silva (SSU123) 16S rRNA database using con dence threshold of 70%. OTU sequences were clustered, and those with > 97% similarity were binned into the same OTU. Alpha diversity metrics at the OTU level based on Shannon's index [48] were mainly calculated by MOTHUR program (https://www.mothur.org/wiki/Main_Page). Beta diversity at the OTU level based on Bray-Curtis [49], were qualitatively examined by MOTHUR program and visualized in Principal Coordinate Analysis (PCoA) using the vegan R packages [50]. Bacterial taxonomic distributions of each group communities at the genus level were visualized using the vegan R packages. A heatmap with a base-10 logarithmic transformation of the absolute abundance at the genus level was generated through R package vegan. Analyses of alpha diversity, beta diversity, bacterial taxonomic distributions, and generation of the heatmap were performed at Majorbio I-Sanger Cloud platform (https://cloud.majorbio.com). 16S rRNA gene sequencing data have been submitted to NCBI Sequence Read Archive (SRA) with the accession number of PRJNA657382.

Isolation of bacterial strains
Fresh feces pellets collected from mice with metronidazole treatment were homogenized in 1 ml sterile phosphate buffered saline containing 2.5% L-cysteine and serially diluted with PBS, seeded onto multiple agar plates (hopebiol, Qingdao, China), including Man Rogosa Sharpe agar (MRS) and Brain Heart Infusion agar (BHI). After incubated under aerobic conditions or anaerobic conditions at 37 °C for 48 h to 72 h, individual colonies with distinct morphologies were picked up and the selected colonies were stream plated in the same media for another 2 or 3 days for further puri cation and identi cation.

Preparation of lymphocytes
Collected colons were opened longitudinally after excising fat tissues and washed with PBS to remove luminal feces, followed by shaking in PBS containing 2 mM dithiothreitol (DTT, D1070, Solarbio, Beijing, China) for 10 min once and PBS containing 5 mM EDTA for 10 min thrice at 37 °C to remove epithelial cells. Lamina propria tissues were sliced into small pieces and digested with PBS containing 2% fetal bovine serum, 100 U/ml penicillin, 100 U/ml streptomycin, 1 mg/ml collagenase IV (C5138,Sigma-Aldrich) and 100 ng/ml DNaseI for 30 min at 37 °C in a shaking water bath. The digested cell tissues were then ltered through a 70 μm nylon cell strainer (352350, BD Biosciences, San Jose, CA, USA) to obtain single cell suspensions and resuspended in 40% percoll to perform percoll gradient separation for obtaining lamina propria lymphocytes. Spleens and mesenteric lymph nodes were ground directly into a single cell suspension mechanically, in which spleen cells were subjected to lyse erythrocytes on ice for 15 min. For Treg cell analysis, after staining of cell surface with CD3, CD4 and CD25 antibodies, lymphocyte suspensions were xed, permeabilized using Foxp3/transcription factor staining buffer sets (00-5523-00,

Identi cation of knockout mice
Genomic DNA was extracted from mouse tail, and the mouse genotype was identi ed by PCR with speci c primers (Table S1, additional le 11). To investigate the role of DC and Treg deletion in E. ludwigii-mediated protection, CD11c-GFP-DTR transgenic mice were injected intraperitoneally (i.p.) with 200 µl PBS containing 100 ng of Diphtheria toxin (4 ng DT/g body weight, D0564-1MG, DTx, Sigma-Aldrich) to rapidly deplete CD11c + DC, and transgenic Foxp3-DTR-GFP mice were given intraperitoneal injections of 1 µg DTx (50 ng DT/g body weight) to deplete Treg. DTx was injected on one day prior to bacteria gavage, and subsequent doses were given every other day to maintain DC and Treg depletion throughout the whole phase of bacteria gavage. When the mice were sacri ced, the spleens, MLNs, and colons were removed and analyzed using ow cytometry to verify the elimination e ciency of DC and Treg cells (FACS Canto II Flow Cytometer, BD Biosciences; the FlowJo software, FlowJo).

Tissue immuno uorescence analysis
Small segment colons were embedded in OCT compound with liquid nitrogen and cut into 5-µm sections followed by xing with cold acetone for 10 min. After blocked with 5% BSA for 1 h, sections were incubated with CD103 antibody (1:200, ab224202, Abcam, Cambridge, MA, USA) in blocking buffer overnight at 4 ℃. After that, slides were incubated with Alexa Fluor 594 -labeled second antibody (1:200, SA00006-8, Proteintech, Chicago, IL, USA) for 1 h and counterstained with DAPI to stain nuclei. Images were acquired using a confocal uorescence microscope (Leica TCS-SP8, Leica Microsystems, Germany).

Statistical analysis
Data were presented as the mean ± SEM. The statistical signi cance of the differences between three or more groups was tested using one-way ANOVA or two-way ANOVA. All operations are performed in commercially available software (Graphpad Prism 6, San Diego, CA) and values of P < 0.05 were considered statistically signi cant. and the percentage of CD25+Foxp3+Tregs in CD4+ T cells in MLN (C) and cLP (D) of mice in each group. n = 9 in each group. Data are the mean ± SEM, one-way ANOVA. *p < 0.05; **p < 0.01; ***p < 0.001.   gallinarum, L. johnsonii or sterile water one day after ve days of ABX treatment. After ve days of single bacterial strain administration, mice were given 3% DSS and LPS (8 mg/kg body weight) simultaneously to induce colitis in addition with single bacterial strain administration for another 7 days, and mice were euthanized. (B-C) Body weight (B) and disease activity index (C) were examined every day during the course of DSS treatment. (D-E) Colon length (D) and histopathological scores by H&E staining (E) in each group. n = 9 (B to D) and n = 8 (E) in each group. Data are the mean ± SEM, two-way ANOVA (B to C) or one-way ANOVA (D to E). *p < 0.05; **p < 0.01; ***p < 0.001.