Roseburia hominis A2-183^T (=DSM 16839^T = NCIMB 14029^T) and other Roseburia species were maintained and grown on synthetic YCFA media as described before (22). Escherichia coli and Salmonella enterica strains were cultivated in LB medium. All Roseburia culture manipulations were performed in a MACS-MG-1000 anaerobic workstation (Don Whitley Scientific) under an atmosphere of 80% N₂, 10% CO₂ and 10% H₂ at 37°C.

Germ-free animal experiments were performed in the gnotobiotic rodent breeding facility of INRA (ANAXEM platform, Institut Micalis, INRA, Jouy-en-Josas, France). The GF C3H/HeN male mice were allocated into the control (N = 8) and treatment (N = 10) groups and caged individually. R. hominis A2-183^T (=DSM 16839^T = NCIMB 14029^T) was grown anaerobically at 37°C in YCFA media. At days 0, 1, and 2, animals in the treatment group were given 10⁹ colony-forming units (CFU) of R. hominis culture by gavage, while control animals were given 100 µL of YCFA medium. The ileal, ascending colonic, and cecal samples were collected at days 14 and 28. GF TLR5KO (C57BL/6 genetic background, N = 3) and C57BL/6 (N = 3) animals (Jouy-en-Josas), and conventional TLR5KO (C57BL/6 genetic background, N = 6) and Boy/J (C57BL/6 congenic, N = 6) animals (Medical Research Facility, University of Aberdeen) were gavaged with the live cultures of R. hominis to evaluate its functional importance. Boy/J (C57BL/6 B6 Cd45.1) mice were the background controls for conventional TLR5K0. They carry a CD45.1 pan leukocyte marker but are otherwise equivalent to C57BL/6 wild-type mice (23, 24). Twelve female C57BL/6 (6 weeks old) mice were used to evaluate the effect of R. hominis during DSS-induced mild colitis. Three mice were dosed daily by direct administration of a culture of R. hominis for a total of 8 days. Six mice were dosed with culture medium. Exposure to DSS was for 4 days (from day 3 to day 6). Three R. hominis-dosed and three control mice were offered sterile water containing DSS (20 g/L) over this period. On day 7, these mice were switched back to sterile water with no DSS. The mice were euthanased and dissected aseptically on day 9. Twenty-two female C57BL/6 mice (6 weeks old) were used to evaluate the effect of R. hominis during DSS-induced pathological colitis. After the acclimatization period of 7–10 days, the mice were dosed daily with 50 µL (10⁹ CFU) of R. hominis in growth culture medium for 14 days. Control animals were given growth culture medium. From day 9, mice were given DSS (MW 50 kDa, 30 g/L) in their drinking water for 6 days. The animals were euthanized on day 15 and tissue sampling was performed as described earlier. The management and experimental procedures with animals were approved by the respective Local Ethical Review Committees.

Bacterial RNA from the mouse cecum contents was isolated and labeled with dCTP-Cy3 or dCTP-Cy5 during cDNA synthesis. PCR products amplified from an R. hominis small fragment library (6,000 clones) were used to create duplicate microarrays on glass slides using a MicroGrid II TAS array-spotting robot (BioRobotics). Microarray hybridization was performed in a GeneTAC hybridization station (Genomic Solutions). Dye-swap and separate RNA purification protocols were used to avoid potential biases.

Murine RNA was extracted from the ileum and ascending colon tissues and hybridized to the GeneChip “NuGO Mouse Array” and GeneChip “Mouse Genome Array” (Affymetrix). Data analysis was performed using R¹ and Bioconductor.² The microarray data were submitted to NCBI GEO (Gene Expression Omnibus) with accession number GSE25544.

The R. hominis-specific primers 5′-CCCACTGACAGAGTATGTAATGTAC-3′ and 5′-GCACCACCTGTCACCAC-3′ were used for qPCR analyses of fecal samples to validate the efficiency of colonization. The analyses were performed using a 7500 Fast Real-Time PCR System (Applied Biosystems) with a Power SYBR Green PCR Master Mix (Applied Biosystems). RT-qPCR analyses were performed on the ileal and colonic RNA samples for host gene expression studies using a QuantiFast SYBR Green PCR Kit (Qiagen) and QuantiTect Primer Assays (Qiagen).

FISH analyses were performed with neutral buffered, formalin-fixed gut tissue sections (2 µm thick) with 16S rRNA probes, specific for the domain of Bacteria (Cy3-labeled Eub338) and specific for the R. hominis strain A2-183 (FITC-labeled GTACATTACATACTCTGTCAGTG). Bacteria were visualized at x630 magnification. Immunolocalization of R. hominis flagellin was examined in methanol-fixed colon content smears using rabbit antisera anti-FlaA1 or anti-FlaA2 (Covalab) and Alexa donkey anti-rabbit 488 (Molecular Probes). T cell markers were examined on sequential intestinal tissue sections (8 µm) using Ly6G-FITC, CD3-FITC, isospecific IgG (BD Biosciences), and double-labeled FoxP3 Alexa Fluor 594 (Abcam) and CD3-FITC. Intestinal tissue sections (4 µm) were also stained with hematoxylin/eosin. A complete transverse cross-section of the colon from each animal was visualized at ×200 magnification. Each field of view was scored from 0 to 4 according to the method based on Ref. (25).

Caco-2 and HT29 cells were grown at 37°C in a 75% humidified atmosphere with 5% CO₂ in 24-well plates. Before any treatment, cells were washed twice with Hanks’ Balanced Salt Solution and kept in DMEM supplemented with l-glutamine, selenium, and transferrin for 24 h. Epithelial cells were harvested after incubation for 2 or 4 h and then stored in RNAlater for further analyses.

Bone marrow cells were harvested from the femur and tibia of C3H/HeN and C57BL/6 mice. Bone marrow-derived dendritic cells (BMDCs) were generated by Flt3L or GM-CSF culture, stimulated with 100 ng/mL flagellin, and analyzed by flow cytometry. Supernatants were collected and tested for IL-10 and IL-12 by cytometry bead arrays (BD Bioscience).

Cells from the intestine and mesenteric lymph nodes (MLNs) were isolated as described previously (26), with minor modifications. In brief, cell suspensions were incubated with 100 U/mL of collagenase VIII (Sigma-Aldrich) in RPMI supplemented with 20% of FBS at 37^oC for 20 min (MLN) or 60 min (intestine). Single cell suspensions were analyzed by flow cytometry. For mixed leukocyte reactions, naïve CD4⁺ T cells were purified from the major lymph nodes and spleens of OTII transgenic mice using MACs Miltenyi CD4⁺ T cell Isolation Kit as per manufactures instruction. Purified CD4⁺ T cells were co-incubated with flagellin activated Flt3L-derived BMDCs in the absence or presence of OVA323–339 peptide (1 µg/ml). After 5 days, T cell differentiation was determined by flow cytometry.

Intestinal lamina propria cells and MLN cells were labeled with CD4-FITC, CD25-APC (eBioscience), CD8-APC-Cy7, CD3-PerCP (Biolegend), and B220-BV570 (BD Biosciences). Labeling of intracellular FoxP3 was performed after extracellular staining and fixation/permeabilization of cells. GM-CSF-derived dendritic cells were labeled with CD11b-PerCP Cy5.5 (BD Biosciences), CD11c-PE-Cy7, I-A/I-E-APC-Cy7, CD80-PE, CD86-APC, CD8-FITC, and B220-BV570 (Biolegend). Flt3L-derived dendritic cells were labeled with CD11c-PE-Cy7, CD11b- or Siglec-H-PerCP Cy5.5 (Biolegend), I-A/I-E-APC-Cy7, CD317-PE, CD40-Alexa Fluor 647, CD103-FITC, and B220-BV570. Coculture T cells were labeled with CD4-PerCP Cy5.5 (BD Biosciences), CD127-PE-Cy7, CD73-APC, CD195 (CCR5)-FITC, CD62L-BV570, and CD25-APC-Cy7 (Biolegend).

Spot intensities on the microarrays were log-transformed and Loess normalization was applied to the microarray results. RT-qPCR data were transformed on a base-2 logarithmic scale and analyzed by one-way analysis of variance with a significance threshold of P < 0.05. Statistical significance between the treatment groups was evaluated using Student’s t-test.

More detailed protocols are available in SI Materials and Methods in Supplementary Material.

Genome analysis (Figure 1A) identified about 5% of all genes were related to chemotaxis and motility function including four different flagellin genes of which one located within the flagellar operon is associated with flagellar motility.

A biomass of R. hominis was given by gavage to GF C3H/HeN mice on three consecutive days. R. hominis-colonized both the ileum and colon but was found in much higher numbers in the colon, up to 1 × 10¹⁰ bacteria/g feces (Figure S1 in Supplementary Material) and was found to be closely associated with the colonic mucosa.

Differential gene expression in the bacterium in response to association with the host was investigated using the R. hominis microarray. Bacterial RNA was isolated from three different experimental conditions to distinguish between the effects of the gut environment and animal dietary components, (i) in vivo, from the cecum of mono-associated mice; (ii) in vitro, from bacteria grown in the presence of dietary components; and (iii) in vitro, from bacteria grown in culture.

Colonization of GF mice with R. hominis correlated with the increased host gene expression, which was highest in the colon, particularly at day 28 post-colonization (Figure 1B). The number of differentially expressed genes in the ileum was very low at day 28, consistent with the reduced bacterial numbers. Differential transcriptome responses in the two tissue sites and time points were illustrated by a clear separation of significant transcripts in the heatmap analysis (Figure 1C).

Differentially expressed genes were identified (in vivo vs. in vitro) with RT-qPCR validation performed on 42 differentially expressed genes (Tables S1 and S2 in Supplementary Material). The mobA- and mobL-like genes that are involved in conjugation/mobilization transfer showed strong upregulation in vivo (Figure 2A). Other subsystems induced by the gut environment included membrane transport, in particular magnesium transport, and motility and chemotaxis including multiple methyl-accepting chemotaxis proteins and genes of the flagellar operon (Figure 2B). R. hominis possesses multiple flagellin genes: flaA1, flaA2, flaA3, and flaB. The expression of flaA1and flaA2 in the murine gut environment was verified by Western–blot and immunocytochemistry (Figure 2C).

Genes involved in innate immunity and gut barrier function of the host were significantly induced by R. hominis colonization [Figure 1C; NCBI GEO (Gene Expression Omnibus), accession number GSE25544]. The GO process “innate immune response” (GO:0045087) was upregulated and included the TLR-related genes Tlr5, Tlr1, and Vnn1. The upregulation of Tlr5 was of particular interest, given the presence of flagellar genes in the genome of R. hominis and the expression of the corresponding proteins in the gut of mono-colonized mice. The flagellins may have a role in this innate signaling pathway to mediate innate and adaptive immune responses. Other innate immunity genes affected by R. hominis colonization included the antimicrobial peptides Defb37, Pla2g3, Muc16, and Itln and the gut barrier function genes Sprr1a, Cldn4, Pmp22, Crb3, and Magi3. Innate immunity genes that were upregulated in the ileum in response to R. hominis included Defcr20, Pcgf2, Ltbp4, Igsf8, and Tcfe2a. We also found a negative regulation of the NF-κB pathway (GO:0043124) by R. hominis, which may contribute to the immune homeostasis by downregulating this inflammatory cascade.

Immune response was a major pathway induced by R. hominis at day 28 in the ascending colon of mono-associated mice. The pathways significantly affected in this category were mostly involved in T cell function, including IL-10 signaling and regulation of T cell function by CTLA-4 (Table S3 in Supplementary Material). The Ly6 gene family was also induced in the ascending colon. In particular, the GPI-anchored gene product of Ly6g6c was upregulated 25-fold, and the related gene Ly6g6e was upregulated twofold at day 28. The majority of hematopoietic cells, including neutrophils and plasmacytoid dendritic cells, express one or more members of the Ly6 family. The increased presence of Ly6G⁺ cells in R. hominis-colonized mice was confirmed by immunocytochemistry (Figure 3A).

While the pathways of T cell regulation were significantly influenced by association with R. hominis, the corresponding effects on T cell differentiation required further investigation. We therefore assessed the number of double-positive CD3⁺FoxP3⁺ cells in the colonic lamina propria of mono-associated C3H/HeN mice. A significant increase in the number of regulatory T cells (CD3⁺FoxP3⁺ cells) was detected in the germfree C3H/HeN and C56BL/6 mice colonized with R. hominis (Figures 3A,B).

Conventional C3H/HeN and C57BL/6 mice were gavaged with R. hominis for 14 days to determine the impact of the bacterium on Treg cells in the lamina propria of animals with a normal microbiota. This treatment resulted in a small but significant increase in the population of CD3⁺CD4⁺CD25⁺FoxP3⁺ T cells in the C3H/HeN (P = 0.04) and a smaller increase in the C57BL/6 mice (P = 0.05) when compared to non-colonized mice (Figure 3C). While the population of Tregs was induced in conventional mice after colonization with R. hominis under homeostatic conditions, it was unknown if this induction could be maintained under conditions of mild colitis. We therefore investigated the impact of the bacterium on Treg cells in the lamina propria of animals that were given low dose DSS. FoxP3-tagged mice of C57BL/6 background were colonized with R. hominis for 2 days before treatment with low dose of DSS for 4 days followed by 2 days of pure drinking water. DSS treatment alone significantly increased the population of CD3⁺CD4⁺CD25⁺FoxP3⁺ T cells (P = 0.02) (Figure 3D). When the animals were additionally treated with R. hominis, there was a further significant increase in the population of CD3⁺CD4⁺CD25⁺FoxP3⁺ T cells (P = 0.02) versus DSS (Figures 3D,E).

The influence of bacteria on the differentiation of T cells may reflect a specific array of TLR ligands displayed by a particular bacterium. TLR5KO (GF and conventional) and WT murine strains were gavaged with the bacterium to evaluate the functional importance of R. hominis and its flagellins. Analysis of the differentially expressed genes in TLR5KO and WT Boy/J mice colonized with R. hominis revealed that although T cell pathways were still influenced by the colonization event in TLR5KO mice, responses were more related to IL-4, IL-5, IL-6, IL-9 pathways and not to IL-10 and CTLA-4 (Table S4 in Supplementary Material). Furthermore, in contrast to mono-associated C3H/HeN and C57BL/6 animals (Figure 3B) and conventional Boy/J animals (Figures 4B,C), the numbers of double-positive CD3⁺FoxP3⁺ cells in the lamina propria of TLR5KO mice were not increased due to R. hominis treatment (Figure 4A).

A set of flagellins from pathogenic and commensal bacteria, including recombinant R. hominis flagellins, was used to compare their effect on activation of signaling responses in IECs and BMDCs. Caco-2 cells were treated with the identical concentrations of different bacterial flagellins (Figure 5A). The flagellin of pathogenic S. enterica serotype Enteritidis (SE) induced a larger gene panel than the flagellins of commensal E. coli K12 or R. hominis FlaA1. E. coli and R. hominis induced a strong response and, with a similar gene complement between the two, formed a separate clade distinct from SE. In contrast, FlaA2 was generally neither pro-inflammatory nor did it activate the conserved gene signature (IL8, CXCL1, CXCL2, and CXCL10) that was induced by the other recombinant bacterial flagellins. In fact, the set of genes affected was closer to that of the control cells and formed a cluster separated from the other flagellins. FlaA1 from R. hominis induced different responses in Flt3L- and GM-CSF-derived BMDCs compared to SE and K12 (Figures 5B,C). In particular, FlaA1 was uniquely able to activate Flt3L-expanded DCs, with the upregulation of I-A/I-E and CD40 and production of IL-10 by BMDCs from both C3H/HeN and C57BL/6 mice. The IL-10/IL-12 ratio was particularly elevated in C57BL/6 DCs (Figure 5D), which were found to be CD103⁺Siglec-H⁺. To test whether flagellin activated BMDCs could differentially modulate naïve T cell differentiation, mixed leukocyte cultures of CD4⁺ OTII cells with flagellin stimulated Flt3-derived BMDC were performed. The data show that the production of IL-10 and IL-12 by the flagellin stimulated Flt3-derived BMDC ratio triggered naïve T cell differentiation toward a Tregs profile (Figure 5E). In contrast to FlaA2, FlaA1 significantly increased the Treg population in the T cell cocultures.

The effects of R. hominis on innate and adaptive immunity prompted us to test its therapeutic efficacy using the murine DSS model (Figure S2 in Supplementary Material; Figure 6). The treatment group was dosed daily with R. hominis (~50 μL, ca. 10⁹ CFU) via gavage for a period of 14 days while the untreated group received the same amount of bacterial growth medium. The two groups were then given DSS in drinking water (MW 50 kDa, 30 g/L) beginning from day 8 onward. On day 15, DSS-treated mice without R. hominis addition had a strong elevation of a panel of pro-inflammatory biomarkers compared to control mice, with gene induction levels ranging from 4- to 49-fold (Figure 6A). Induction of pro-inflammatory genes was significantly lower in the R. hominis-treated mice compared to the control group, indicating a strong therapeutic benefit of oral administration of R. hominis. On day 15, severe inflammation was detected in the ascending colon of DSS-treated mice without an R. hominis supplement, while the colonic mucosa in the R. hominis-supplemented group displayed only low-level inflammation, consistent with the reduced expression of pro-inflammatory genes (Figures 6B,C). Treatment with R. hominis also alleviated the weight loss in DSS-treated mice (Figure S3 and Table S5 in Supplementary Material).

In the next set of experiments, we investigated metabolic changes in the host and bacterium due to host–microbe association. Mice mono-associated with R. hominis displayed significant metabolic changes compared to control GF mice. In particular, the GO processes “negative regulation of response to food” (GO:0032096), “negative regulation of appetite” (GO:0032099), and “regulation of catecholamine secretion” (GO:0050433) were all downregulated in the ascending colon after colonization by R. hominis [NCBI GEO (Gene Expression Omnibus), accession number GSE25544]. The genes involved in these processes were Agt, Cartpt, Cck, and Cxcl12, with the corresponding mRNA changes induced ranging from 2- to 12-fold. Cck, in particular, plays a major role in digestion and satiety as a hunger suppressant. Gcg expression was also downregulated at this gut site.

To establish whether the transcriptional changes of these genes have any physiological relevance in terms of body composition, analyses of dry carcass weight and lipid composition were performed. Dry carcass weights of R. hominis-associated mice were significantly heavier than those of GF animals (Figure 7A). Carcass lipid analysis showed that the total adiposity was also significantly higher in R. hominis-colonized animals at day 14 (Figure 7B).

Significant metabolic changes were also detected on the other side of the host–microbe interaction spectrum. In particular, expression of R. hominis genes involved in butyrate production was upregulated by the gut environment (Figure 7C). These were the genes encoding acetyl-CoA acetyltransferase, 3-hydroxyacyl-CoA dehydrogenase, butyryl-CoA dehydrogenase, and phosphoenolpyruvate carboxykinase. The stimulatory effect of host epithelial cells on production of acetate and butyrate by R. hominis was further confirmed in separate in vitro experiments (Figure S4 in Supplementary Material). Bacterial cells incubated with Caco-2 and HT29 cells produced significantly higher amounts of these metabolites compared to R. hominis control cells incubated under the same conditions but without the epithelial cells.