key: cord-0018142-ra0b73b2
authors: van der Lelie, Daniel; Oka, Akihiko; Taghavi, Safiyh; Umeno, Junji; Fan, Ting-Jia; Merrell, Katherine E.; Watson, Sarah D.; Ouellette, Lisa; Liu, Bo; Awoniyi, Muyiwa; Lai, Yunjia; Chi, Liang; Lu, Kun; Henry, Christopher S.; Sartor, R. Balfour
title: Rationally designed bacterial consortia to treat chronic immune-mediated colitis and restore intestinal homeostasis
date: 2021-05-28
journal: Nat Commun
DOI: 10.1038/s41467-021-23460-x
sha: 07bab85329d4990f39407998aaa1279d2c810cd7
doc_id: 18142
cord_uid: ra0b73b2

Environmental factors, mucosal permeability and defective immunoregulation drive overactive immunity to a subset of resident intestinal bacteria that mediate multiple inflammatory conditions. GUT-103 and GUT-108, live biotherapeutic products rationally designed to complement missing or underrepresented functions in the dysbiotic microbiome of IBD patients, address upstream targets, rather than targeting a single cytokine to block downstream inflammation responses. GUT-103, composed of 17 strains that synergistically provide protective and sustained engraftment in the IBD inflammatory environment, prevented and treated chronic immune-mediated colitis. Therapeutic application of GUT-108 reversed established colitis in a humanized chronic T cell-mediated mouse model. It decreased pathobionts while expanding resident protective bacteria; produced metabolites promoting mucosal healing and immunoregulatory responses; decreased inflammatory cytokines and Th-1 and Th-17 cells; and induced interleukin-10-producing colonic regulatory cells, and IL-10-independent homeostatic pathways. We propose GUT-108 for treating and preventing relapse for IBD and other inflammatory conditions characterized by unbalanced microbiota and mucosal permeability.

C hronic intestinal inflammation can be induced by multiple exogenous and endogenous signals and is mediated by immune and nonimmune cells in genetically susceptible hosts with defects in epithelial barrier function, immunoregulation, or bacterial killing. Exogenous substances including dietary products, pathogenic microorganisms, xenobiotics including antibiotics, or various combinations thereof, can trigger initial mucosal injury and/or dysbiosis to initiate acute intestinal inflammation that is perpetuated by the antigenic activities of a subset of resident microbiota 1, 2 . Examples of these conditions are inflammatory bowel diseases (IBD), which encompass two main clinical disorders: Crohn's disease and ulcerative colitis. Current IBD treatments primarily control inflammation through anti-inflammatory and immunosuppressive mechanisms. Some of the most successful drugs for treating IBD include infliximab, adalimumab, vedolizumab, ustekinumab, and tofacitinib, which target specific immune components to control the inflammatory process. However, these and other immune modulating drugs induce sustained, steroid-free remission in only a small subset of patients and can have multiple serious side effects, including an increased risk for serious and potentially life-threatening infections and neoplasia. In addition, these drugs do not correct upstream conditions that contribute to the chronic inflammatory mechanisms, including the leaky mucosal barrier, a pro-inflammatory gut microbiome and immunoregulatory defects.

As an alternative to anti-inflammatory and immunosuppressive therapies, microbiome-inspired live biotherapeutic products (LBPs) are being developed to treat conditions linked to chronic intestinal inflammation and increased permeability. The traditional approach for LBP discovery has been to compare the microbiomes of healthy subjects and patients suffering from a specific condition, such as IBD, to identify microorganisms that are lacking or under-represented in large databases, such as the HMP2 project 3 . This information, further enforced by the results from microbial association studies, is used to propose a therapeutic formulation to replenish the microorganisms that are lacking or under-represented 2, 4 . In the case of IBD, early efforts have focused on the use of strains belonging to the Clostridium clusters IV and XIVa, which were found to successfully decrease inflammation in rodent IBD models [5] [6] [7] . Using germ-free (GF) mice inoculated with healthy human fecal material pretreated with chloroform to enrich for spore-forming bacteria, a stable 17strain consortium was enriched from a single donor based on their ability to induce colonic regulatory T cells (Tregs) 7, 8 . This consortium was comprised of spore-forming Clostridium cluster IV, XIVa, and XVIII strains that produced butyrate and decreased the severity of several colitis models 7 .

Open label application of Fecal Microbiome Transplants and enrichment-based approaches have several disadvantages. The outcome is defined by the stool sample used for the enrichment, with different samples representing different consortia with variable efficacy; 9 undesirable strains/functions associated with safety risks including virulence factors and transferable antibiotic resistance functions might be present, such as the presence of enteropathogenic and Shigatoxin-producing Escherichia coli strains 10 and antibiotic-resistant E. coli strains 11 in FMTs, or the presence of transferable vancomycin resistance elements as found in the genome of the VE202 consortium strain Blautia coccoides VE202-06 (GenBank Accession Number Accession: PRJDB525). Furthermore, consortium modeling, as presented in this study, shows that other bacterial species besides spore-forming Clostridium bacteria provide metabolic support and additional therapeutic functions required for optimal engraftment and therapeutic performance of live biotherapeutic products in the hostile gut environment of patients with intestinal inflammation. These shortcomings can be addressed by a bottom-up rational consortium design approach that is rigorously informed by mechanistic modeling and insights from microbiome ecology and disease pathogenesis. We used this approach to combine wellcharacterized strains isolated from many healthy human stool samples into a consortium of metabolically interdependent strains with a variety of therapeutic functionalities being distributed in a redundant way between strains. Initially a 17-strain consortium, GUT-103, was designed around publicly available strains. GUT-103 rapidly colonized mice, restored normal function to the inflamed colon, and prevented and reversed established experimental colitis in gnotobiotic mice. Based on these proof of concept studies, a refined 11-member consortium, GUT-108, was designed around a panel of proprietary human bacterial strains that strongly engrafted and provided similar redundant protective functions. Therapeutically applied GUT-108 corrected functional dysbiosis of the inflamed gut microbiome and treated established colitis in a humanized mouse colitis model while decreasing opportunistic pathogenic bacteria, increasing resident protective bacterial groups, and restoring immunologic and metabolic homeostasis.

Rational design of the GUT-103 consortium. Using a bottomup approach, LBPs were rationally designed to provide key therapeutic functions involved in mucosal homeostasis and immune modulation defective in the gut microbiome of IBD patients 2,3 . These functions include synthesis of short-chain fatty acids (SCFA) including butyrate and propionate, synthesis of indole, indole-3 propionic acid (IPA) and indole-3 acetic acid (IAA) from tryptophan, and deconjugation of bile salts and their conversion into therapeutic secondary bile acids, especially deoxycholic acid (DCA) and lithocholic acid (LCA). Additional functions included to control pathobionts were the synthesis of antimicrobials and siderophores and the uptake of heterologous siderophores produced by Enterobacteriaceae including Escherichia, Klebsiella, and Shigella species, allowing bacteria to compete with these opportunistic pathogens for the essential nutrient iron. The importance of the gut microbiome to provide these specific functions [12] [13] [14] and the role of specific microbes encoding key activities, such as the 7α/β-dehydroxylation bile acid conversion pathway encoded by Clostridium scindens 15 has been extensively reviewed.

Next, commensal human gut bacteria that can provide these desired functions were identified. Reference isolate genome sequences for common human gut microbes were collected based on published gut 16 S rRNA studies and isolate source metadata (https://www.vmh.life/#home). Isolate and metagenome assembled genome (MAG) sequences were gathered from RefSeq 16 , PATRIC 17 and the Human Microbiome Project 18 . Consistently annotated genomes were used to determine which genomes were likely to encode proteins with the desired therapeutic functions. Comparative analysis of annotations from these genomes also enabled the identification of genera where selected therapeutic functions were highly conserved. A key finding from this analysis was that members of the Clostridium clusters IV and XIVa families could not provide several critical functionalities required to restore the proinflammatory microbiome of IBD patients, including the synthesis of propionate and indole-based derivatives from tryptophan. Therefore, the rational design of LBPs to restore these critical functionalities should also include members of the families Bacteroidaceae, Eubacteriaceae, Akkermansiaceae, and non-Clostridium clusters IV and XIVa family members. Ultimately, a set of the most desirable genera were selected from which to compose the rationally designed LBPs, including isolates belonging to the genera Akkermansia (Supplementary Data 1), Bacteroides (Supplementary Data 2), Clostridium (Supplementary Data 3), and Faecalibacterium (Supplementary Data 4). Multi-genome analysis for the genera Akkermansia and Faecalibacterium combined with whole-genome DNA average nucleotide identity (ANI) showed that species previously classified as Akkermansia muciniphila and Faecalibacterium prausnitzii are comprised of several species, distinguished by over 10% difference in whole-genome DNA sequence compared to their type strains.

Although this comparative analysis identified desirable genera, it was still necessary to identify specific strains within each genus to comprise the consortia. In this selection process, it is most important to identify strains with complementary auxotrophies for essential amino acids, vitamins, and co-factors. This was based on previous analyses of natural microbiomes, which revealed that co-evolved strains in these systems often have complementary auxotrophies 19 , and addresses the risk of one strain outcompeting the other members of the consortium. Furthermore, redundancy in metabolic dependencies between strains is important to avoid that loss of one strain will not lead to the collapse of the consortium. Auxotrophic analysis of isolate strain genomes and MAGs from our selected genera revealed that auxotrophies were typically conserved among most members of the same species, but with enough variation to make this analysis important to include in the design process.

Overall, using redundancy in key functionalities combined with a distribution of complementary auxotrophies to create a network of metabolically interdependent strains, a consortium of 17 publicly available human strains was selected, referred to as GUT-103 (Table 1) . For each of the strains, individual strain models were constructed that were subsequently combined into a consortium model. Flux Balance analysis was used to predict in silico that the GUT-103 strains form a network of metabolic dependencies complementing each other's metabolic needs, indicating that GUT-103 can function as a consortium without the need for auxotrophies being complemented by resident bacteria from the imbalanced gut microbiome of IBD patients 3 . The GUT-103 functional characteristics, auxotrophy reactions, flux balance analysis, and strain interaction table to predict the growth of the individual GUT-103 strains and the GUT-103 consortium under defined conditions are shown in Supplementary Data 5.

GUT-103 successfully prevents and treats experimental colitis. Preliminary testing using quantitative PCR with species-specific rpoB gene primers showed that 2 weeks after gavage GUT-103 successfully colonized GF 129 WT and 129 interleukin-10 deficient (Il10 −/− ) mice (see Supplementary Fig. 1a , b and Supplementary Table 1 ). Three human bacteria, Escherichia coli LF82, Enterococcus faecalis OG1RF, and Ruminococcus gnavus ATCC29149 (EER strains) were used based on their ability to colonize GF 129 WT and 129 Il10 −/− mice ( Supplementary  Fig. 2a , b and Supplementary Table 2 ), to induce chronic colitis and to activate strain-specific Th1 and Th17 mucosal immune responses in gnotobiotic Il10 −/− mice 20, 21 . Selective colonization of gnotobiotic Il10 −/− mice by EER showed increased fecal lipocalin-2 concentrations (149.8 ± 80.2 ng/g feces, a 2 log increase from baseline values) at 2 weeks, with further progression over 4 weeks ( Supplementary Fig. 3 ). Based on these kinetic studies showing mildly active colitis 2 weeks after EER colonization, we chose 2 weeks for GUT-103 therapeutic intervention. The ability of GUT-103 to colonize, compete with IBD-relevant pathobionts and to both prevent and treat establish Th1/Th17mediated colitis was subsequently tested (Fig. 1a) . GUT-103 was successfully established one week after gavage in EER Il10 −/− mice, (Fig. 1b) with a composition similar to that in ex-GF mice ( Supplementary Fig. 1 ). All strains were detected but concentrations varied. Three dominant strains, Blautia producta DSM2950, Akkermansia muciniphila ATCC BAA-835 and Bacteroides massiliensis DSM17679, with complementary auxotrophies, suggest the formation of an integrated metabolic network. Relatively high levels of Bacteroides stercoris DSM19555, which has similar auxotrophies and functions as Bacteroides massiliensis DSM17679, were also observed. These predominant community strains are predicted to synthesize SCFA and indole, but lack synthesis of siderophores, bacteriocin synthesis and bile acid conversion. GUT-103 decreased concentrations of each of the EER strains in aggregate to less than 20% of the total bacterial population, even in the therapeutic protocol where the three colitogenic strains were established for 2 weeks (Fig. 1b) . The preventive protocol blocked R. gnavus concentrations to below 1%.

Colonization with GUT-103 did not induce inflammation, but both the preventive and therapeutic protocols significantly decreased colitis in EER-inoculated Il10 −/− mice compared with PBS controls, as measured by lipocalin-2 levels (Fig. 1c) , crypt hyperplasia and lamina propria (LP) mononuclear cell infiltration (Fig. 1d) , interaction table predict the growth of the individual GUT-108 and blinded histology scoring (Fig. 1e) . Gavage with GUT-103 also increased synthesis of butyrate and propionate that was deficient in EER-inoculated Il10 −/− mice (Fig. 1f ) and decreased inflammation measured by cecal mRNA expression for IFNγ, IL-12p40, and TNFα (Fig. 1g) .

Bile acid profiles were analyzed to further evaluate restoration of a functional gut microbiome by engraftment with GUT-103 strains. Significantly higher levels of multiple unconjugated bile acids were observed in cecal contents from Therapy, Prevention and GUT-103 groups compared with EER-inoculated Il10 −/− mice (Fig. 2a) . Compared to EER, therapeutic application of GUT-103 increased cecal luminal concentrations of the unconjugated primary bile acids βMCA, CA, and CDCA. Cecal contents of the GUT-103 group had higher levels of the secondary bile acids UDCA and LCA than EER-inoculated Il10 −/− mice, but low DCA levels were not significantly different between these groups (Fig. 2a) . In contrast to free bile acids, the cecal luminal levels of multiple conjugated bile acids, including TβMCA, TCDCA, TUDCA, and GCA, in all GUT-103 groups were significantly lower than those in EER (Fig. 2b) . However, the level of glycineconjugated GLCA was significantly higher in the GUT-103colonized group than in the EER group (Fig. 2b) . Overall, compared to EER, GUT-103 decreased levels of conjugated bile acids especially for taurine-conjugated bile acids. Together, these results demonstrate both protective and therapeutic activities of GUT-103 and in vivo validation of the anticipated replacement of pathobionts and protective metabolic functions of this rationally designed consortium. It should also be noted that in preliminary studies we demonstrated that two independent consortia of 14 and 19 bacterial isolates, respectively, from two healthy human volunteers induced colitis when colonized into germ-free 129 Il10 −/− mice 22, 23 .

Design of the optimized GUT-108 consortium. Informed by the GUT-103 outcomes, we rationally designed an optimized 11-strain consortium referred to as GUT-108 (Table 2) . Compared to GUT-103, GUT-108 included improved redundancy for synthesis of the therapeutic secondary bile acids LCA and DCA covered by Extibacter sp. GGCC_0201; yersiniabactin synthesis provided by Clostridium symbiosum GGCC_0272; and several strains with functions antagonistic against opportunistic pathogens into the Enterobacteriaceae family. The GUT-108 functional characteristics, auxotrophy reactions, flux balance analysis, and strain interaction table predict the growth Functions involved in the synthesis of siderophores, the uptake of heterologous siderophores, and the deconjugation of bile salts and the conversion to secondary bile acids are indicated (+). 3α-HSD 3α-hydroxy steroid dehydrogenase, 7α-HSD 7α-hydroxy steroid dehydrogenase, 7β -HSD 7β-hydroxy steroid dehydrogenase, 7α-DH 7α-dehydratase.

of the individual GUT-108 strains and the GUT-108 consortium under defined conditions and are shown in Supplementary Data 6.

GUT-108 successfully establishes in gnotobiotic mice models.

GUT-108 was gavaged into two different GF models, Il10 +/eGFP VertX reporter and Il10 −/− mice, with engraftment in both Il10 + / eGFP VertX reporter (Fig. 3a) and Il10 −/− mice (Fig. 3b ). Colonization levels of individual strains were determined using quantitative PCR. Colonization with GUT-108 did not induce inflammation as documented by representative distal colonic photomicrographs of H&E-stained tissue (Fig. 3c) histologic scores (Fig. 3d) . Importantly, GUT-108 colonization promoted a healthy gut microbiome with key functionalities for SCFA, especially synthesis of acetate, butyrate and propionate, and IAA synthesis (Fig. 3e ). Four GUT-108 strains, Clostridium butyricum GGCC_0151, Intestinimonas butyriciproducens GGCC_0179, Bitterella massiliensis GGCC_0305, and Barnesiella sp. GGCC_0306 were established at levels below 1% in both Il10 + / eGFP VertX reporter (Fig. 3a) and Il10 −/− mice (Fig. 3b) , with Clostridium butyricum GGCC_0151 failing to get established in Il10 + / eGFP VertX reporter mice. However, due to redundancies in metabolic interdependencies and therapeutic functions, low luminal concentrations of these strains did not prevent restoring key metabolic phenotypes of a healthy gut microbiome. The ability of the GUT-108 consortium to stimulate regulatory (protective) IL-10-mediated immune responses was determined by comparing the type and levels of IL-10-producing immune cells in Il10 + / eGFP VertX reporter mice. Two weeks after being introduced, GUT-108 stimulated higher numbers of colonic LP IL-10-producing CD4 + T cells, B cells, and dendritic cells (DC) and increased numbers and percentages of regulatory T cells, including inducible Tregs (IL-10 + RoRγT + FoxP3 + CD4 + cells and IL-10 + FoxP3 + CD4 + cells) (Fig. 3f) . Although showing an upward trend, no statistical difference was observed for FoxP3 neg IL-10 + T cells (TR 1 ).

GUT-108 reverses gut microbiome dysbiosis, restores functionality. Therapeutic application of GUT-108 was tested in a newly developed experimental colitis model using GF Il10 −/− mice (129SvEv background) humanized with a fecal transplant from a healthy human donor (Fig. 4a) . Experimental colitis was induced by inoculating the GF Il10 −/− mice with stool from the healthy donor that was previously verified to induce aggressive colitis in ex-GF Il10 −/− mice 24 . Over time the normal microbiota from a healthy donor are altered in an inflammatory environment 21, 25, 26 . We documented in Fig. 3c and Fig. 3d that gnotobiotic Il10 −/− mice selectively colonized with the GUT-108 strains did not develop colitis. The difference in outcomes of colitis in Il10 −/− mice colonized with the complex microbiota of the healthy human stool versus protection with GUT-108 is likely because GUT-108 contains only protective, commensal bacteria, in contrast to the presence of pathobiont strains present at low levels in healthy human stool, but which expand in a proinflammatory environment. Therapeutic application of GUT-108 beginning 2 weeks after human fecal colonization changed the community composition ( Supplementary Fig. 4 ) and reduced the expansion of opportunistic pathogenic bacteria belonging to the Enterobacteriaceae family that occurred in mice receiving PBS (Supplementary Table 3 ). Metagenome analysis showed that treatment with GUT-108 resulted in engraftment of all GUT-108 strains except Clostridium scindens GGCC_0168, restoring a gut microbiome that promoted the abundance of beneficial resident Clostridium (Clusters IV and XIVa) strains not part of the GUT-108 consortium. Compared to PBS therapy and consistent with the GUT-103 therapeutic results, GUT-108 decreased by 90% the abundance of Enterobacteriaceae, including Klebsiella pneumoniae, Salmonella enterica, Escherichia coli, and Shigella species, from 5% to below 0.5% (Fig. 4b) .

The progression of colitis in humanized Il10 −/− mice was evaluated by serial fecal lipocalin-2 ( Fig. 4c ) after human fecal gavage. Two weeks after human fecal inoculation, elevated levels of fecal lipocalin-2 were observed (compared to basal lipocalin-2 levels of~3 ng/g feces material), indicating the onset of moderate to severe colitis. Therapeutic application of GUT-108 at 2 weeks after human fecal transplantation decreased inflammation compared to the PBS treatment as measured by fecal lipocalin-2 ( Fig. 4c ) and blinded histologic scoring (Fig. 4d) . GUT-108 decreased histologic inflammation in the cecum and distal colon, as well as the summed colonic score, which we used as an overall inflammatory index 20, 21, 27 . Representative photomicrographs of the distal colon demonstrated decreased inflammation following GUT-108 therapy with reduced mononuclear cell infiltration in the LP and submucosa, mucosal thickening and crypt hyperplasia (Fig. 4e) . Metabolite analysis confirmed that the therapeutic application of GUT-108 restored a physiologically healthy gut microbiome that produced increased levels of anti-inflammatory bacterial metabolites associated with mucosal homeostasis, including the Fig. 1 Effect of gavage with GUT-103 on the gut microbiome community composition, lipocalin-2 levels, histology of the gut epithelium, SCFA synthesis and tissue cytokine expression of Il10 −/− mice where experimental colitis was induced by inoculation with Escherichia coli LF82 (E), Enterococcus faecalis OG1RF (E) and Ruminococcus gnavus ATCC29149 (R). a Schematic overview of experimental design. Gnotobiotic Il10 −/− mice were inoculated with GUT-103 (N = 5), EER (N = 9), GUT-103 plus EER (preventive protocol) (N-7), and EER plus GUT-103 (therapeutic protocol) (N = 7). After 5 or 6 weeks, animals were killed and tissue samples were collected for analysis; b Composition of the microbiome in fecal pellets of gnotobiotic 129 Il10 −/− mice inoculated with EER, GUT-103 plus EER (preventive protocol), and EER plus GUT-103 (therapeutic protocol) determined by qPCR. Il10 −/− mice were inoculated via gavage with GUT-103 (2 × 10 +7 cfu/strain) or EER (2 × 10 +7 cfu/strain). The composition of the gut microbiome was determined on fecal pellets or after sacrifice on cecal contents using qPCR with species-specific primers. The average community composition (pie diagrams) for seven animals per treatment are presented. . Il10 −/− mice were inoculated via gavage with GUT-103 (2 × 10 +7 cfu/strain) or EER (2 × 10 +7 cfu/strain). a Unconjugated bile acids: αMCA α-muricholic acid, βMCA β-muricholic acid, CA cholic acid, CDCA chenodeoxycholic acid, LCA lithocholic acid, UDCA ursodeoxycholic acid, DCA deoxycholic acid. b Conjugated bile acids: TβMCA taurine-conjugated β-muricholic acid, TCA taurocholic acid, TCDCA taurochenodeoxycholic acid, TUDCA tauroursodeoxycholic acid, GUDCA glycoursodeoxycholic acid, GCA glycine-conjugated cholic acid, GLCA glycine-conjugated lithocholic acid. 1e-way ANOVA, Mean ± SE, *P < 0.05, **P < 0.01, ***P < 0.005, ****P < 0.001. Source data and calculated P-values are provided as a Source Data file. Genes involved in the deconjugation of bile salts and the conversion to secondary bile acids are indicated (+).

CGH choloylglycine hydrolase, LCD L-cantinine hydratase, 3-oxo-5α 3-oxo-5-alpha-steroid-4-dehydrogenase, 7α-HSD 7α-hydroxy steroid dehydrogenase, 3α-CHD 3α-hydroxycholate dehydrogenase, 3α-DH 3α-hydroxy bile acid-CoA-ester-3 dehydrogenase, 7α-DH 7αdehydratase, 7β-DH 7β-dehydratase, SBS sodium-bile acid symporter system.

Il10 -/- Fig. 3 Effect on the gut microbiome community composition, colonic histology, and fecal metabolomics of gnotobiotic mice inoculated with GUT-108 or treated with PBS for 2 weeks. Il10 + / eGFP VertX reporter mice (a) or Il10 −/− mice (b) were inoculated via gavage with GUT-108 (2 × 10 +7 cfu/ strain), and after 2 weeks community composition of the gut microbiome was determined on cecal contents using qPCR with species-specific primers. The average community composition (circle diagram) as well as the individual community composition for six Il10 + / eGFP VertX reporter mice and five (Fig. 4f) . Likewise, therapeutic application of GUT-108 significantly increased levels of the secondary bile acids LCA and DCA (Fig. 4f) , a key objective in the rational design of GUT-108. In contrast to propionate, treatment did not change butyrate levels, consistent with the findings of the selectively colonized gnotobiotic mouse models, where stool butyrate levels were 30fold lower than those for propionate (Fig. 3e) .

Therapeutic application of GUT-108 reduces inflammatory cytokines. Reduced severity of inflammation and improved mucosal immune homeostasis after therapeutic application of GUT-108 was characterized by decreased levels of colonic LP effector CD4 + T cells, including Th1 and Th17 cells that produce the inflammatory cytokines IFNγ and IL-17α, respectively, or both IFNγ and IL-17α (Fig. 5a) ; reduced distal colonic expression levels of innate and Th1 and Th17 pathway inflammatory cytokines, Fig. 4 GUT-108 strains decreased colitis in Il10 −/− mice colonized with human fecal microbiota. a Schematic overview of experimental design. On day 1, mice were inoculated with a human stool previously verified to induce aggressive colitis in gnotobiotic Il10 −/− mice. Therapeutic application of GUT-108 or PBS control started after 2 weeks. After 4 weeks necropsies were performed, and tissue samples were collected for analysis. b Bacterial species whose abundance significantly changed in the gut microbiome of mice that received therapeutic application of GUT-108 compared to PBS control mice. Differences in relative species abundance between the two treatments is expressed on a Log10 scale. Strains with decreased (red) or increased (blue) relative abundance are indicated. GUT-108 strains are indicated in black. *P < 0.05 value using the two-sided Benjamini-Hochberg corrected Wilcoxon test. c Serial fecal lipocalin-2 levels were measured using ELISA after 2 (before start of GUT-108 or PBS therapy), 3 and 4 weeks. Dots indicate individual mice data. Mann-Whitney unpaired two-tailed t-test. Bars indicate Mean ± SE, *P < 0.05, ***P < 0.001, N.S. indicates not significant. N = 11/group. d Blinded histologic scoring assessed the level of inflammation for different parts of the intestine, including the distal ileum (Ile), cecum (Ce), proximal-(Pc) and distal-(Dc) colon, rectum (Re), and combined (Ce + Pc + Dc + Re). Dots indicate individual mice data. Mann-Whitney unpaired two-tailed t-test. Bars indicate Mean ± SE. Unpaired two-tailed t-test was performed on the data. *P < 0.05, ***P < 0.001, NS not significant. N = 11/group. e Representative distal colonic photomicrographs of H&E-stained tissue showing colonic tissues 2 weeks after starting therapeutic treatment of humanized Il10 −/− mice with GUT-108 or PBS. The experiment was repeated independently with similar results for each animal in the GUT-108 (N = 11) or PBS (N = 11) group. The bar in the picture indicates 100 µm; Hu + GUT-108 refers to humanized Il10 −/− mice treated with GUT-108; Hu + PBS refers to humanized Il10 −/− mice that received PBS as a placebo control. f Induction of protective fecal metabolites by therapeutic GUT-108 treatment beginning at 2 weeks after the induction of inflammation by human stool. Week 4 represents 2 weeks of therapeutic treatment. Hu+GUT-108 refers to humanized Il10 −/− mice treated with GUT-108; Hu + PBS refers to humanized Il10 −/− mice that received PBS buffer. DCA deoxycholic acid, LCA lithocholic acid, IPA indole-3-propionic acid. Twoway ANOVA and adjusted P-values were calculated by the multiple comparisons test. Bar indicates Mean ± SE. N.S. not significant; *: adjusted P < 0.05; **: adjusted P < 0.01; ***: adjusted P < 0.001. Source data and calculated P-values are provided as a Source Data file. including IL-1β, IL-12p40, IL-13, IL-17α, IFNγ, and TNFα (Fig. 5b) , and increased colonic expression levels of receptors and pathways implicated in mucosal healing (Fig. 5c, Supplementary Fig. 5 ). Several protective pathway components that are decreased in IBD were upregulated 2 weeks after application of GUT-108, further supporting its therapeutic role and providing mechanisms for protection (Fig. 5c) . Expression of the Ahr pathway genes, including the aryl hydrocarbon receptor (Ahr), the aryl hydrocarbon receptor repressor (AhrR), and the Cytochrome P450 Family 1 Subfamily A Member 1 (Cyp1A1) gene 28 , were significantly upregulated. Protective pathways with increased gene expression levels included the genes for the defensins DefCR1 and DefA, whose altered production is suggested to be integrally involved in IBD pathogenesis 29 , and the genes for aldehyde dehydrogenases Aldh1A1 and Aldh1A2, whose expression is reduced in colonic macrophage and DC subsets of patients with ulcerative colitis regardless of inflammation 30 (Fig. 5c ). Together these results confirm, by multiple readouts, that administration of GUT-108 treats established colitis in humanized mice, out-competes pathobionts, inhibits effector immune responses, and restores homeostatic metabolic and protective immune responses.

Resident bacterial strains of the GUT-103 and GUT-108 consortia, designed to interdependently restore normal function to the inflamed colon, rapidly colonized gnotobiotic mice as well as ex-germ-free humanized mice that had developed a proinflammatory gut microbiome. These GUT-103 and 108colonized mice exhibited the desired functional properties consistent with the rational design of the two consortia.

The two separate rationally designed consortia of intestinal bacterial strains with redundant functionalities that were previously found to be under-represented in the gut microbiome of IBD patients with active disease 3 were shown to promote homeostatic immune functions and bacterial metabolism and prevent onset or progression of intestinal inflammation. We believe that this approach has the potential to maintain long term remission in a physiologic and safe manner, which should be tested in a Phase 1 clinical trial.

Testing GUT-103 against EER-induced colitis in gnotobiotic Il10 −/− mice is clinically relevant to human Crohn's disease because adherent-invasive Escherichia coli and Ruminococcus gnavus are increased in active Crohn's disease 3,31,32 and the adherent-invasive E. coli strain used was isolated from the ileum of a Crohn's disease patient 33 . GUT-108 showed therapeutic efficacy to reverse established colitis in Il10 −/− humanized mice by intervening 2 weeks after onset of moderate-severe intestinal inflammation, supporting a potential role for GUT-108 to treat active IBD. Moreover, the Il10 −/− model of chronic pathobiontdriven T-cell-mediated chronic inflammation is more predictive of therapeutic responses in IBD patients than acute epithelial injury models such as dextran sodium sulfate 2 .

GUT-103 and GUT-108 were designed based on human data showing that protective functions provided by commensal bacteria are under-represented in the gut microbiome of IBD patients. These functions include the synthesis of SCFAs, indole and its derivatives, bile acid deconjugation and conversion, and competition for the critical nutrient iron and the synthesis of antagonistic molecules to control opportunistic pathogens 3, 34 . Several animal studies have highlighted the role for SCFAs, especially propionate and butyrate, in regulatory T cell recruitment and function 8, 12, 13, [34] [35] [36] . The recruitment in the colon and extrathymic conditioning of regulatory T cell response by SCFA make these molecules an important link in the crosstalk between the gut microbiome and the immune system. Therefore, commensal bacteria identified to produce propionate and butyrate were selected in the rational design of GUT-103 (Table 1) and GUT-108 (Table 2) .

Several studies have shown the role of indole, a metabolite produced from tryptophan, and its metabolites in reducing attachment of pathogenic E. coli to epithelial cells 37 , strengthening the mucosal barrier and mucin stimulating production 38 . Therefore, commensal bacteria identified to produce indole and its derivatives were selected in the rational design of GUT-103 (Table 1) and GUT-108 (Table 2 ).

An inflammatory gut microbiota can result in the inefficient microbial conversion of bile salts into their primary and secondary bile acids 3 . IBD patients with an unbalanced gut microbiome due to inflammation have lower fecal and circulating concentrations of secondary bile acids and higher conjugated fecal bile acid concentrations than do healthy subjects 39 . Thus, activities essential for the conversion of primary bile acids, specifically the conversion of CA and TCDCA acid via a multistep process that includes 7-alpha-dehydroxylation by 7-alpha dehydratase (7-α-DH) or 7-alpha-hydroxy steroid dehydrogenase (7α-HSD) activity 15 , are included in the strain selection for GUT-103 (Table 1) and GUT-108 (Table 2) .

Competition for iron helps drive the competitiveness and establishment of microorganisms 40 . Therefore, GUT-103 (Table 1) and GUT-108 (Table 2 ) include several strains that synthesize one or more siderophores under iron-limiting conditions. Ideally, these siderophores are insensitive to inhibition by Lipocalin-2, a peptide that inhibits specific siderophores and their uptake, and is a major colonic defense system triggered by bacterial infections. Lipocalin-2 levels were increased after induction of inflammation in Il10 −/− mice colonized with the EER consortium (Fig. 1c) or human fecal microbiota (Fig. 4c) .

Bacteriocins, of which lantibiotics are considered a specific class, have shown great promise as new antibiotics for therapeutic application, as reviewed by Field et al. 41 . Thus, bacteriocin synthesis was included as a key functionality in strain selection as part the rational design process of GUT-103 (Table 1) and GUT-108 ( Table 2) .

The optimized 11-strain GUT-108 consortium was rationally designed to build on the proof-of-concept results obtained with GUT-103. GUT-108 went beyond members of the Clostridium clusters IV and XIVa strains, including Bacteroides and Akkermansia species. Furthermore, based on their genome analysis, strains with undesirable properties including the presence of transferable antibiotic resistances or putative virulence factors were excluded. We also omitted species that are fastidiously anaerobic, such as Faecalibacterium prausnitzii, from the GUT-108 consortium. Compared to GUT-103, GUT-108 strains provide additional redundancy for the synthesis of the protective secondary bile acids LCA and DCA, plus multiple mechanisms to compete with opportunistic pathogenic Enterobacteriaceae including synthesis of the siderophore yersiniabactin, and lantibiotics. In the inflammatory gut environment of humanized Il10 −/− mice, beneficial Lachnospiraceae and Ruminococcaceae family members are decreased while opportunistic Enterobacteriaceae are increased 42 , as reported for Crohn's disease patients 32, 43 . Therapeutic application of GUT-108 reduced levels of colitogenic Enterobacteriaceae and increased beneficial resident Clostridium (Clusters IV and XIVa) species, especially Lachnospiraceae including Dorea species and Lachnoclostridium species that are not GUT-108 constituents (Fig. 4b) . This altered community composition increased cecal luminal propionate concentrations, but not butyrate levels (Fig. 4f) . Previous studies demonstrated that butyrate levels are not necessarily an indicator of a healthy gut microbiome, as butyrate synthesis from fermentation of amino acids such as lysine can contribute to inflammation under conditions associated with mucosal permeability 44 .

GUT-108 increased expression of Gpr41 and showed an upward trend for Gpr43 in Il10 −/− mice humanized with a fecal transplant ( Supplementary Fig. 5 ). As previously reported, SCFA produced by gut bacteria stimulate Tregs 8 with propionate's effect mediated through GPR43 36 . SCFA also mediate the function of GPR41, a key regulator that controls host energy balance 45 .

With its functional redundancy and metabolically interdependent auxotrophies, GUT-108 is designed to engraft and perform under a wide range of conditions. When applied to Il10 −/− mice humanized with a fecal transplant, all GUT-108 strains except Clostridium scindens GGCC_0168 were established for at least 2 weeks. Clostridium scindens has been previously described as one of the essential strains necessary to convert primary bile acids into LCA and DCA. However, despite the absence of this strain, the established functional multi-strain network produced secondary bile acids, with Extibacter sp. GGCC_0201 providing the 7α-dehydratase activity required to convert CA and CDCA into the therapeutic secondary bile acids DCA and LCA, respectively. Normalizing the intestinal bile acid profile can restore intestinal epithelial stem cell function 46 , and increase colonic RORγ + Treg cell counts that ameliorate host susceptibility to colitis 47 , while LCA stimulates Treg differentiation and inhibits Th17 cells 48 consistent with GUT-108's ability to restore secondary bile acid metabolism (Fig. 4f) and activate inducible IL-10 + RORγ FoxP3 + CD4 + Treg cells (Fig. 3f) . GUT-108 stimulated regulatory (protective) immunity by increasing numbers of colonic LP IL-10-producing CD4 + T cells, B cells and DC and numbers and percentages of regulatory T cells, including inducible Tregs (IL-10 + RORγT + FoxP3 + CD4 + cells) and IL-10 + Tregs (Fig. 3f) . We further documented the antiinflammatory effects of therapeutic GUT-108 in Il10 −/− mice humanized with a fecal transplant by demonstrating that GUT-108 decreased IFN-ɣ + , IL-17α + , and IFN-ɣ + IL-17α + synthesizing colonic LP CD4 + TH 1 and TH 17 cells (Fig. 5a) and reduced expression levels of innate and Th1 and Th17 pathway cytokines, including IL-1β, IL-12p40, IL-13, IL-17α, IFNγ, and TNFα (Fig. 5b) . Interestingly, GUT-108 treatment increased expression of IL-15 mRNA, a homeostatic cytokine that controls T cell inflammatory responses. Exogenous IL-15 treatment decreases IL-17α expression by Th17 cells in vitro through STAT5 enrichment at the IL-17 locus 49 , consistent with the ability of GUT-108 therapy to increase IL-15 gene expression and decrease IL-17α mRNA expression in colonic LP cells (Fig. 5b) .

Increased intestinal bacterial metabolism of tryptophan, especially indole and its derivatives IAA and IPA, activates the Ahr pathway. AHR acts as a sensor of the microbiota community and, through its established role of modulating immune functions, maintains host-microbe homeostasis 50 . IPA is also a pregnane X receptor (PXR) agonist mediating its responses through TLR4 51 . GUT-108 therapy increased both IAA and IPA levels in stool (Fig. 4f) and colonic Ahr gene expression (Fig. 5c ). AHR is a critical mediator of anti-inflammatory responses to infection by bacterial pathogens and of the differentiation and function of immune cells including T cells, innate lymphoid cells, macrophages and DC 28 . AHR promotes the expression of the anti-inflammatory cytokine IL-10 and inhibits macrophage apoptosis, decreases the expression of inflammatory cytokines (IL-6 and TNF-α) and inhibits activation of NF-κB. Therefore, the Ahr pathway is critical to protect from excessive inflammatory cytokine expression and septic shock. In addition, Ahr pathway activation protects the mucosa during inflammation 52 .

Efficacy of GUT-108 in Il10 −/− mice is mediated by IL-10independent mechanisms. Further insights into these possible protective mechanisms include increased expression of metabolite sensors and mediators (Gpr41, Gpr43, Fxr, Pxr, Pparg, Fgf15, Fgf21) and pathways mediating differentiation of immune cells including Treg and Breg cells (cMaf, Il5, April, Aid, Bcl6) (Supplementary Fig. 5 ). For example, intestinal epithelial FGF15 is activated by bile acids serving as ligands for the nuclear receptor farnesoid X receptor (FXR). FXR/FGF15 signaling regulates bile acid homeostasis and protects against experimental colitis 53 . APRIL impacts immune regulatory T cells by stimulating their proliferation and survival, and directly contributing to their immune suppression 54 . Bcl6 induces IL-10 and follicular T helper cells and regulates the balance of innate lymphoid cells subsets 55 . Decreased expression of Nos2 ( Supplementary Fig. 5 ), a part of the Ace2 -Nos2-IFNɣ biosynthesis gene cascade 56 , could lead to lower levels of ACE2, a key regulator to control intestinal inflammation induced by epithelial damage 57 . Certain viruses, including the coronaviruses SARS-CoV-1 and SARS-CoV-2, use the ACE2 protein for infecting respiratory and intestinal epithelial 58 . Therefore, chronic gut inflammation, as seen in type-2 diabetes and obesity 59 might trigger elevated gut epithelial ACE2 levels, and therefore patients suffering from these conditions could be more sensitive for Coronavirus infection and at higher risk for complications, highlighting the importance of therapeutically targeting the pro-inflammatory gut microbiome as the underlying cause of chronic inflammation 60 .

In addition to analyzing fecal material, we measured cecal microbiota and in some cases cecal luminal metabolites to represent bacterial communities and function within the cecum/ colon, since the cecum is one of the sites of most active inflammation (Fig. 4d) . Previous rodent studies have shown broadly similar microbial patterns in cecal luminal and fecal samples 61 . Furthermore, we demonstrate similar cecal (Fig. 2) and fecal (Fig. 4f ) secondary bile acid (LCA and DCA) responses to GUT-103 and GUT-108, respectively. Although no animal model completely replicates all clinical features of human Crohn's disease or ulcerative colitis, we believe that our use of a human fecal transplant in germ-free mice to initiate chronic TH 1 /TH 17 -mediated colitis and our treatment protocol of administering GUT-103 and 108 to mice with established mild to moderate inflammation replicates IBD as closely as feasible for preclinical studies. Proof of efficacy in human IBD will have to await a clinical trial.

GUT-103 and GUT-108 combine multiple modes of action to treat the upstream causes of inflammation by correcting the abnormal microbiome environment, activating various IL-10 synthesizing immune cells, lowering inflammatory responses, and restoring bacterial metabolic profiles to levels found in stool samples of healthy individuals. These overlapping protective mechanisms are predicted to maintain long term remission of IBD in a physiologic and safe manner, in contrast to most biologicals, which block downstream immune effector responses by neutralizing a single cytokine or molecule and induce immunosuppression that can be associated with increased infection and neoplasms. These integrated protective mechanisms make GUT-108 a promising novel therapy to treat a range of conditions whose pathogenesis is characterized by dysbiosis-mediated chronic intestinal inflammation and increased mucosal permeability. Besides IBD this could include graft versus host disease, hepatic encephalopathy, alcoholic liver disease, atherosclerosis, hypertension, obesity, metabolic syndrome, and type-2 diabetes mellitus. The human fecal sample for transplant to mice was collected in compliance with all relevant ethical regulations for work with human participants, and with informed consent from the participant under the approved University of North Carolina Institutional Review Board protocol #17-1528.

Microbiology techniques. All bacterial strains were grown on LYH-BHI medium (Brain-heart infusion medium supplemented with 0.5% yeast extract (Difco)), except Bitterella massiliensis GGCC_0305 and Barnesiella sp. GGCC_0306, which were grown on Peptone Yeast Glucose broth (Anaerobe Systems). Strains were cultivated at 37°C under strict anaerobic conditions (H 2 (5%), CO 2 (10%)-, and N 2 (85%) v /v/v) . After growth, cultures were concentrated at 10 +9 cfu/ml in phosphate buffered saline (PBS). Cultures were tested for contamination with aerobic bacteria by streaking on LB medium, followed by incubation at 37°C under aerobic conditions. For quality control, a culture was considered pure if no aerobic growth was observed and the sequenced 16 S rRNA gene of each culture matched the anticipated composition.

DNA sequencing. Bacterial genome sequencing libraries were generated using the ThruPLEX DNA-seq Kit (Rubicon Genomics). Individual strain libraries were combined in equimolar proportions in one pool and sequenced by 125 bp pair end read sequencing on the Illumina HiSeq2500. Onboard image processing and base calling were performed. The sequence data quality score (Q score) was used as a quality control metric with the specification that ≥80% of bases must have a Q score of ≥30. After trimming, sequencing reads were assembled with SPAdes (version 3.13.0) using default parameters 62 and annotated using Prokka (version 1.14-dev) 63 and the RAST server [64] [65] [66] . Genome sequences for GUT-108 strains were submitted to GenBank (Table 2) .

DNA from fecal and cecal material was extracted with AllPrep PowerViral DNA/RNA Kit (Qiagen). Metagenome sequence libraries were constructed using the KAPA Hyper DNA library prep. DNA libraries were multiplexed and loaded on an Illumina HiSeq4000 instrument per manufacturer's instructions. Metagenome sequencing was performed using a 2 × 150 paired-end configuration; image analysis and base calling were conducted by the HiSeq Control Software (v2.0.12) on the HiSeq instrument. Sequencing reads were trimmed, and mouse reads were filtered out using Trimmomatic (version v0.39) 67 . Species abundance was determined directly on the metagenome sequencing reads with Kaiju (version v1.7.2) 68 . The raw sequencing reads from the fecal samples have been deposited with links to BioProject accession number PRJNA703330 in the NCBI BioProject database (https://www.ncbi.nlm.nih.gov/bioproject/).

Multi-genome analysis and modeling. Genome sequences for GUT-108 strains or reference strains, downloaded from GenBank files (National Center for Biotechnology Information, NCBI) or Pathosystems Resource Integration Center (PATRIC) 17 were consistently annotated using Prokka (version 1.14-dev), the RAST server, antiSMASH (version 5.1.1) 69 , and Blast against custom-built protein databases. Models were reconstructed, gapfilled, and Flux Balance Analysis 70 (FBA) was run using the KBase fbatools module v2.0, which is currently deployed in beta mode in the KBase narrative interface (https://narrative.kbase.us/). Reconstruction was performed using the "Build metabolic model" app; gapfilling was performed using the "Gapfill Metabolic Model" app; all flux balance analysis was performed using the "Run Flux Balance Analysis" app; community models were assembled using the "Merge Metabolic Models into Community Model" app; and community flux balance analysis was performed using the "Run Flux Balance Analysis" app. As a last step, all models were manually curated and refined prior to final analysis. Auxotrophies were determined by reviewing the Pathway table to determine if the compounds were created via a reaction or combination of reactions. '+' indicates the compound is produced. "AA" indicates the compound is not produced. Average nucleotide identity was determined via pyani (version 0.2.7.dev) 71 . The results are presented in Supplementary Data files 1to 6. Quantitative (RT)-PCR of bacterial DNA. Q-PCR using strain-specific primers against the single copy RpoB gene were used to quantify the composition of the GUT-103 and GUT-108 consortia after gavage in gnotobiotic mice. RpoB primers were validated to ensure there was no significant cross-reactivity between their target bacterium and other bacteria that were present in the consortia (Supplementary Fig. 6 ). The strain-specific RpoB primers are provided in Supplementary Data 7. The genomic DNA was extracted from fecal or culture media using AllPrep PowerViral DNA/RNA Kit (Qiagen). Q-PCR were performed with QuantStudio3 (Thermo Fisher Scientific, PA, USA) using SYBR No-ROX reagents (Bioline, TN, USA) and 10 ng of genomic DNA per strain with the following PCR settings: 95°C, 3 min; 40 cycles of (95°C, 5 sec; 60°C, 20 sec); melting curve analysis: 95°C, 15 sec; 60°C, 15 sec; 95°C, 15 sec. The data were created by comparative Ct method (2 -Ct ). Melting curve analysis confirmed the presence of single products with expected melting temperatures.

Targeted metabolomics. Targeted metabolomics were conducted to determine concentration levels of bile acids, short-chain fatty acids (SCFAs), and tryptophan/ indole metabolites in mouse feces and/or cecal contents. Organic solvents and reagents of LC/MS-grade (Optima TM ), including methanol (MeOH), isopropanol (IPA), acetonitrile (ACN), water, and formic acid (FA), were purchased from Fisher Scientific (Waltham, WA, USA). Reagents of analytical grade, including 1propanol (PrOH), pyridine (Py), propyl chloroformate (PCF), alongside semiconductor-grade sodium hydroxide (NaOH), and anhydrous sodium sulfate (Na 2 SO 4 ), were purchased from Sigma-Aldrich (St. Louis, MO, USA).

For SCFA analysis, authentic reference chemical standards of acetic acid (AA), propionic acid (PA), and butyric acid (BA) were purchased from Sigma-Aldrich (St. Louis, MO, USA); stable isotope-labelling (SIL) d4-AA, d2-PA, and d2-BA were procured from Fisher Scientific (Waltham, MA, USA). Fecal SCFA extraction and measurement methods were performed as previously described with minor modifications 72 . In brief,~40 mg sample aliquots were extracted on a TissueLyzer (Qiagen, Hilden, Germany) into 0.005 M NaOH aqueous solution spiked with 50 μg/mL of d4-AA, 10 μg/mL of d2-PA, and 10 μg/mL of d2-BA, followed by centrifugation at 13,200 rpm for 20 min. The supernatant was transferred to glass tubes and derivatized by sequential addition of water, PrOH:Py (3:2, v/v), and PCF. Upon capping, brief vortexing and two-min sonication was performed. The propyl derivatives of SCFAs were extracted twice with hexane, combined, and transferred to Na 2 SO 4 -containing vials for GC-MS quantitation. Instrumental analysis was performed on an Agilent 7820 A gas chromatography system coupling to a 5977B mass spectrometric detector (MSD) (Santa Clara, CA, USA), where 1 µL of derivatives was injected under 1:10 split mode for electron ionization and separated by a DB-5ms column (30 × 0.25 × 0.25 mm) (Santa Clara, CA, USA) with 1 mL/ min helium flowing through. The GC oven temperature program was set as follows: initial oven 175 temperature held at 50°C for 2 min, ramped to 70°C by 10°C/min, to 85°C by 3°C/min, 176 to 110°C by 5°C/min, to 290°C by 30°C/ min, and finally held at 290°C for 8 min; the MSD was operated at full scan mode with scan range of m/z 30-600. The relative standard deviations of intra-and interday precision of analysis were below 10%. Lower limits of quantitation (LLOQ) were determined to be 1 ng for AA (on column), 1 ng for PA, and 0.1 ng for BA.

For bile acid analysis, authentic reference standards (in salt forms) of 23 bile acids, including dihydroxycholestanoic acid (DHCA), glycoursodeoxycholic acid (GUDCA), α-hyocholic acid (HCA), taurochenodeoxycholic acid (TCDCA), cholic acid (CA), taurodeoxycholic acid (TDCA), glycodeoxycholic acid (GDCA), chenodeoxycholic acid (CDCA), deoxycholic acid (DCA), lithocholic acid (LCA), ursodeoxycholic acid (UDCA), tauroursodeoxycholic acid (TUDCA), murideoxycholic acid (MDCA), glycocholic acid (GCA), glycohyocholic acid (GHCA), glycochenodeoxycholic acid (GCDCA), tauro-ω-muricholic acid (TωMCA), tauro-α-muricholic acid (TαMCA), tauro-β-muricholic acid (TβMCA), taurocholic acid (TCA), α-muricholic acid (αMCA), β-muricholic acid (βMCA), ωmuricholic acid (ωMCA) were purchased from Cayman Chemical (Ann Arbor, MI, USA) and Steraloids Inc. (Newport, R.I., USA). Synthetic SIL internal standards, including d4-GCA, d6-GDCA and GDHCA, d4-GLCA, d4-GUDCA, d4-TβMCA, d4-CA, d4-TUDCA, d9-TCDCA, d6-TDCA, d4-UDCA, d9-CDCA, d6-DCA, and d4-LCA were purchased from Cambridge Isotope Laboratories, Inc. (Tewksbury, MA, USA). Bile acids in feces and/or cecal content were determined using LC-MS/ MS based on previously published procedures with slight modifications 73, 74 . Briefly,~10 mg cecal contents were extracted with ACN and MeOH. Twenty microliter of 1 µM internal standard mixture of bile acids was spiked in prior to the extraction. The samples were homogenized on a TissueLyzer (Qiagen, Hilden, Germany) followed by centrifugation at 16,000 rpm for 10 min. Both supernatants were combined, centrifuged at 16,000 rpm for 10 min again, SpeedVac dried, reconstituted with MeOH, and diluted to 1.2 mL with 0.1% FA in water. The mixture extracts were passed through Biotage ISOLUTE ® C18 cartridges (100 mg/1 mL) (Uppsala, Sweden) which had been washed with MeOH and conditioned with MeOH:water (with 0.1% FA) (1:5, v/v) prior to analysis. Upon completion of sample loading, the cartridges were washed with MeOH:water (with 0.1% FA) and subject to vacuum drying. The analytes were eluted afterwards with MeOH, SpeedVac dried at 8°C, and resuspended in 50% MeOH upon analysis. The instrumental analysis was conducted on a Vanquish UHPLC system coupling to a TSQ Quantis triple quadrupole mass spectrometer (Thermo Scientific, Waltham, MA, USA) interfacing with an electrospray ionization (ESI) source operated in negative ion mode. Chromatographic separation was carried out on an ACQUITY Histology and scoring. Fixing, staining, and blind histology scoring were described previously 20 .

Statistical analysis. GraphPad Prism (version 8) software was used for statistical analysis. For the metabolite analyses, the two-way ANOVA in Graph Pad Prism 8 software was applied to determine the statistical significance of SCFAs and bile acids among groups. Dunnett's multiple comparisons test was performed to determine whether a specific SCFA or bile acid is significantly different between EER control and other groups. An adjusted p-value (calculated by Graph Pad Prism 8 software, version 8.0.0) of <0.05 was considered as significant difference. Results were adjusted based on the amount of stool used.

Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

The relevant data are available from the authors upon reasonable request. Metagenome sequencing data that support the findings of this study have been deposited in GenBank with the BioProject ID: PRJNA703330. The genome sequences for the GUT 21 ; 478749.5. The genome sequences for the novel strains used in the GUT-108 consortium that support the findings of this study have been deposited in GenBank with the accession codes: JABFCE000000000, JABFCF000000000, JABFCG000000000, JABFHK000000000, JABFAG000000000, JABFCH000000000, JABFCI000000000, JABFCJ000000000, JABFCK000000000, JABFCL000000000, and JABFCM000000000. Raw data for the targeted metabolomics are available from Metabolomics Workbench repository, https://doi.org/10.21228/M8XM6R, project ID PR001120. The bacterial strains used in the GUT-103 consortium are commercially available from ATCC (https:// www.ATCC.org) and DSMZ (https://www.DSMZ.de). The bacterial strains used in the GUT-108 consortium can be obtained from the corresponding author through a Material Transfer Agreement with the restriction that they can only be used for academic research purposes. Source data are provided with this paper.

Received: 3 June 2020; Accepted: 29 April 2021;

The bile acids are eluted at a flow rate of 400 µL/min with a gradient elution of mobile phase starting at 25% B, held for 2 min; increased to 40% B in 13 min; increased to 98% B at 20 min, held for 2 min; decreased to 25% B at 22.5 min, and held for 3.5 min, for a total run of 26 min. The source parameters were: spray voltage, 3500 V; sheath gas, 50 arbitrary unit; aux gas, 25 arbitrary unit; sweep gas, 6 arbitrary unit; ion transfer tube temperature, 325°C; vaporizer temperature, 350°C . Quantification of bile acids was based on an isotope dilution method using the responses of d4-GCA (for GCA and GHCA), d6-GDCA (for GDCA and GDHCA), d4-GLCA (for GLCA), d4-GUDCA (for GCDCA, GHDCA, and GUDCA), d4-TβMCA (for TαMCA

The data were corrected for differences in the amount of stool analyzed. Regression coefficients R 2 > 0.99 for all calibration curves. LODs for all bile acids were determined as below 0.5 nM. For analysis of tryptophan and indole metabolites, reference chemical standards of tryptophan (Trp), kynurenine (Kyn), indole, indole-3-acetate (IAA), and indole-3-propionate (IPA), were purchased from Sigma-Aldrich

) was used; injected analytes underwent 15-min chromatographic separation on an ACQUITY UPLC HSS T3 column (Waters, Milford, MA) and positive electrospray ionization (ESI), and ions were detected under parallel reaction monitoring (PRM) mode. Internal standard quantitation was performed disparately to determine the concentration levels of the five metabolites, using calibration curves with at least five points and a fitted linearity regression of R 2 > 0.99. LLOQ were determined as 50 femtomole (fM) on column for Kyn and Trp, 100 fM for IAA and IPA, and 0.5 picomole (pmol) for indole

Germ-free and gnotobiotic mice were maintained in positive-pressure isolators and housed in separate polycarbonate cages at constant room temperature (22°C ± 10%), air humidity (50 ± 20%), and a light/dark cycle of 12 h. Mice had free access to food and water. Standard mouse chow (TD2020SX; Teklad Diets, Madison, WI) was sterilized by irradiation at 25 kGy. GF mice were colonized with EER or human feces by oral gavage (200 μl) and 3-4 mice/cage were housed within gnotobiotic Trexler isolators (EER/GUT-103 experiments and gnotobiotic GUT-108 experiments)

Animal use protocols were approved by the Institutional Animal Core and Use Committee of the University of North Carolina at Chapel Hill

The stool was derived from a single healthy donor, Donor-Y, and was previously found to induce moderate to severe colitis in Il10 −/− mice 24 . One gram stool was diluted 100-fold with anaerobic PBS, and vigorously mixed for 5 min under anaerobic conditions. For application of GUT-103 and GUT-108, 300 µl resuspended strain mixture in anaerobic PBS was applied per mouse by oral gavage. GUT-103 and GUT-108 strains were grown individually, subsequently mixed to equal concentrations (cfu/ml), and provided at a dose of 2.0 × 10 +7 cfu/strain in a total volume of 300 µl. The strain mixture was provided four times via oral gavage on days 15

Fresh murine feces (2-5 pieces/mouse) were collected and immediately snap-frozen on dry ice and stored at −80°C. The human fecal sample used for murine fecal transplant was isolated from a healthy volunteer and stored at −80°C

Colonic tissues were opened longitudinally, washed twice with 1×

Denuded tissue samples were washed twice with HBSS containing 2.5% FBS and 1% penicillin-streptomycin and incubated with a stirrer at 450 r.p.m. in HBSS containing 2.5% FBS, 0.5 mg/ml of collagenase (Sigma) for 30 min at 37°C. Cell preparations were filtered through 100-μm nylon mesh to achieve single-cell suspensions. LP cells were purified using a 40-70% discontinuous Percoll gradient

For the GFP-positive gate, GFP-negative colonic LP cells from C57BL/6 wild-type mice were stained with all antibodies used in the experiment as a fluorescenceminus-one control. The flow cytometry gating strategies used to collect the data presented in Fig. 3f and Fig. 5a are provided in Supplementary Fig. 7 and Supplementary Fig. 8, respectively. The following antibodies were used for flow cytometry: anti-CD16/CD32 monoclonal antibody (Fc block, BD Biosciences, Cat# 553141), Live/Dead fixable dead cell stains (APC-Cy7, Invitrogen, Cat# L10119) at dilution of 1:2400, anti-CD45 antibody (Pacific Orange, Invitrogen, Cat# MCD4530) at dilution of 1:300, anti-B220 antibody (Pacific Blue, Invitrogen, Cat# 48-0452-82) at dilution of 1:200, anti-CD19 antibody (BV605, Biolegend, Cat# 115539) at dilution of 1:300, anti-CD4 antibody (eFluor450, eBioscience, Cat# 48-0041-80) at dilution of 1:200, anti-CD4 antibody (APC, Biolegend, Cat# 100516) at dilution of 1:400, anti-CD3 antibody (PE-Cy5, eBioscience, Cat# 15-0031-81) at dilution of 1:300, anti-TCRb antibody (Alexa Fluor 700, BD Biosciences, Cat# 560705) at dilution of 1:300, anti-CD8 antibody (PE-Cy7, eBioscience, Cat# 25-0081-81) at dilution of 1:300, anti-CD11b antibody (PE-Cy7, BD Biosciences, Cat# 552850) at dilution of 1:300, anti-CD11c antibody (Alexa Fluor 700, BD Biosciences, Cat# 560583) at dilution of 1:400, anti-CD64 antibody (PE, Biolegend, Cat# 139304) at dilution of 1:300, anti-MHC classII antibody (BV650, Biolegend, Cat# 107641) at dilution of 1:300, anti-IFNg antibody (BV421, Biolegend, Cat# 505829) at dilution of 1:300, anti-IL-17a antibody (BV605, Biolegend, Cat# 506927) at dilution of 1:300, anti-Foxp3 antibody (PE, eBioscience, Cat# 12-5773-82) at dilution of 1:300, anti-RORgt antibody

Immunopathogenesis of IBD: current state of the art

Roles for intestinal bacteria, viruses, and fungi in pathogenesis of inflammatory bowel diseases and therapeutic approaches

Multi-omics of the gut microbial ecosystem in inflammatory bowel diseases

Microbial-based and microbial-targeted therapies for inflammatory bowel diseases

5th Probiotics, Prebiotics and New Foods Congress

Butyric acid-producing anaerobic bacteria as a novel probiotic treatment approach for inflammatory bowel disease

Treg induction by a rationally selected mixture of Clostridia strains from the human microbiota

Commensal microbe-derived butyrate induces the differentiation of colonic regulatory T cells

A defined commensal consortium elicits CD8 T cells and anti-cancer immunity

FDA Safety Alert Regarding the Use of Fecal Microbiota for Transplantation and Risk of Serious Adverse Events Likely due to Transmission of Pathogenic Organisms

Drug-resistant E. coli bacteremia transmitted by fecal microbiota transplant

The gut microbiota at the service of immunometabolism

Gut microbiota-derived metabolites as key actors in inflammatory bowel disease

Bile salt biotransformations by human intestinal bacteria

Clostridium scindens baiCD and baiH genes encode stereo-specific 7alpha/ 7beta-hydroxy-3-oxo-delta4-cholenoic acid oxidoreductases

NCBI prokaryotic genome annotation pipeline

PATRIC, the bacterial bioinformatics database and analysis resource

The human microbiome project

Experimental evolution of metabolic dependency in bacteria

Variable phenotypes of enterocolitis in interleukin 10-deficient mice monoassociated with two different commensal bacteria

Induction of bacterial antigen-specific colitis by a simplified human microbiota consortium in gnotobiotic interleukin-10 −/− mice

Combined in vivo/in vitro identification of human resident bacterial strains that selectively induce IL-10 and promote mucosal homeostasis

Protective and potentially aggressive bacterial species occur simultaneously in healthy hosts

Tu1844-IL-10-independent protective activities of humanderived Clostridium strains in experimental colitis

Analysis of the large bowel microbiota of colitic mice using PCR/DGGE

Host-mediated inflammation disrupts the intestinal microbiota and promotes the overgrowth of Enterobacteriaceae

Microbiota maintain colonic homeostasis by activating TLR2/MyD88/PI3K signaling in IL-10-producing regulatory B cells

The role of the aryl hydrocarbon receptor (AHR) in immune and inflammatory diseases

Defensins and inflammation: the role of defensins in inflammatory bowel disease

Macrophage and dendritic cell subsets in IBD: ALDH + cells are reduced in colon tissue of patients with ulcerative colitis regardless of inflammation

High prevalence of adherent-invasive Escherichia coli associated with ileal mucosa in Crohn's disease

The treatment-naive microbiome in new-onset Crohn's disease

Invasive ability of an Escherichia coli strain isolated from the ileal mucosa of a patient with Crohn's disease

Strategies to dissect host-microbial immune interactions that determine mucosal homeostasis vs. intestinal inflammation in Gnotobiotic mice

Metabolites produced by commensal bacteria promote peripheral regulatory T-cell generation

The microbial metabolites, short-chain fatty acids, regulate colonic Treg cell homeostasis

Differential effects of epinephrine, norepinephrine, and indole on Escherichia coli O157:H7 chemotaxis, colonization, and gene expression

The bacterial signal indole increases epithelial-cell tight-junction resistance and attenuates indicators of inflammation

Connecting dysbiosis, bile-acid dysmetabolism and gut inflammation in inflammatory bowel diseases

Nutritional iron turned inside out: intestinal stress from a gut microbial perspective

Bioengineering lantibiotics for therapeutic success

A healthy human fecal specimen includes multiple E. Coli, Klebsiella Pneumoniae and Enterobacter strains with variable abilities to activate Th1 cells and induce colitis

Gut microbiota composition and functional changes in inflammatory bowel disease and irritable bowel syndrome

Immune response modulation using live biotherapeutics, for conditions such as allergy desensitization. USA patent

Regulation of energy homeostasis by GPR41

Mitochondrial impairment drives intestinal stem cell transition into dysfunctional Paneth cells predicting Crohn's disease recurrence

Microbial bile acid metabolites modulate gut RORgamma(+) regulatory T cell homeostasis

Bile acid metabolites control TH17 and Treg cell differentiation

The role of IL-15 in activating STAT5 and fine-tuning IL-17A production in CD4 T lymphocytes

Indole and tryptophan metabolism: endogenous and dietary routes to Ah receptor activation

Symbiotic bacterial metabolites regulate gastrointestinal barrier function via the xenobiotic sensor PXR and Toll-like receptor 4

Regulation of the immune response by the aryl hydrocarbon receptor

PPARalpha-UGT axis activation represses intestinal FXR-FGF15 feedback signalling and exacerbates experimental colitis

APRIL signaling via TACI mediates immunosuppression by T regulatory cells in multiple myeloma: therapeutic implications

Bcl6 promotes follicular helper T-cell differentiation and PD-1 expression in a Blimp1-independent manner in mice

Differentially expressed genes in the caecal and colonic mucosa of Landrace finishing pigs with high and low food conversion ratios

ACE2 links amino acid malnutrition to microbial ecology and intestinal inflammation

ACE2 expression by colonic epithelial cells is associated with viral infection, immunity and energy metabolism

Intestinal microflora and metabolic diseases

COVID-19 and the gut microbiome: more than a gut feeling

Us vs. Them" pair housing: effects on body weight, open field behavior, and gut microbiota in rats selectively bred on a taste phenotype

Assembling single-cell genomes and mini-metagenomes from chimeric MDA products

Prokka: rapid prokaryotic genome annotation

The RAST Server: rapid annotations using subsystems technology

The SEED and the rapid annotation of microbial genomes using subsystems technology (RAST)

RASTtk: a modular and extensible implementation of the RAST algorithm for building custom annotation pipelines and annotating batches of genomes

Trimmomatic: a flexible trimmer for Illumina sequence data

Fast and sensitive taxonomic classification for metagenomics with Kaiju

antiSMASH 5.0: updates to the secondary metabolite genome mining pipeline

High-throughput generation, optimization and analysis of genome-scale metabolic models

Genomics and taxonomy in diagnostics for food security: soft-rotting enterobacterial plant pathogens

A targeted metabolomic protocol for short-chain fatty acids and branched-chain amino acids

Quantitative-profiling of bile acids and their conjugates in mouse liver, bile, plasma, and urine using LC-MS/MS

Systemic gut microbial modulation of bile acid metabolism in host tissue compartments

We would like to thank Jeremy Herzog and Lisa Holt for technical assistance. The work was funded by Gusto Global LLC. Germ-free mice were provided by NIH grants P40OD010995; P30DK034987; P01DK094779, and the Crohn's and Colitis Foundation. Instrumentation and resources for metabolomics work was partially supported by NIH grants R01ES024950; P30ES010126; R03ES032067, and P42ES031007.

Supplementary information The online version contains supplementary material available at https://doi.org/10.1038/s41467-021-23460-x.Correspondence and requests for materials should be addressed to D.v.d.L.Peer review information:Nature Communications thanks the anonymous reviewers for their contribution to the peer review of this work.

Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/ licenses/by/4.0/.