Microbiota-Derived Metabolites Suppress Arthritis by Amplifying Aryl-Hydrocarbon Receptor Activation in Regulatory B Cells

1Rheumatoid arthritis (RA) is a disease caused by the malfunction of white blood cells. The cause of this malfunction in unclear; however, recent data suggest that changes to the gut bacteria or “microbiota” impact normal immune system function in RA. Researchers at University College London have discovered that supplementation of mice with butyrate, a molecule produced by the microbiota following breakdown of complex dietary starch, suppresses arthritis. Butyrate supplementation acts tosupportthe function of a population of inflammation-suppressing white blood cells called regulatory B cells by changing the composition of microbiota and increasing the production of a serotonin-derived metabolite, 5-Hydroxyindole-3-acetic acid. These data demonstrate that supplementing the diet with certain microbiota-derived molecules may be a promising treatment for arthritis.

2Regulatory B cells (Bregs) are immunosuppressive cells that contribute to the maintenance of immunological tolerance. Bregs suppress a variety of immune pathologies including autoimmune diseases through the production of interleukin (IL)-10, IL-35, and transforming growth factor beta 1 (TGFβ1). They inhibit the expansion of pathogenic T cells and other pro-inflammatory lymphocytes, and promote regulatory T cell (Treg) differentiation. Toll-like receptor (TLR) agonists, including lipopolysaccharide (TLR4) and CpG oligo-deoxynucleotides (TLR9), in combination with low grade levels of inflammatory cytokines, for example IFNα and/or IL-1β and IL-6, induce IL-10-producing Breg differentiation. The strength of these inflammatory signals is key in determining whether immature B cells develop into Bregs or into mature B cells and antibody-producing plasma cells. We have recently shown that low-grade inflammatory signals that drive the differentiation of immature B cells into Bregs are provided in the gut-associated lymphoid tissue (GALT) as a result of the interaction between the gut microbiota and the innate immune system. Mice depleted of endogenous bacteria following administration of broad-spectrum antibiotics do not develop arthritis or Bregs, suggesting an intricate relationship between microbiota, inflammation, and Breg differentiation. The question of whether inflammatory signals produced in response to the microbiota control Breg development alone or whether microbial factors also play a direct role remains unanswered.

3Among different gut-microbiota-derived metabolites, the most well-characterized are the end products of dietary fiber fermentation, the short-chain fatty acids (SCFAs). SCFAs serve as an important source of nutrients for intestinal epithelial cells supporting barrier function and act as important cellular mediators regulating gene expression, cell differentiation, and gut tissue development. SCFAs are a potent class of immune-modulatory compounds with the capacity to modulate Treg, T helper 17 (Th17) cells, and macrophage differentiation in the gut and periphery. In addition to processing dietary-derived material into potentially immunomodulatory compounds, the gut microbiota shapes host responses to xenobiotics, suggesting that transcription factors implicated in xenobiotic metabolism may closely interact and be influenced by the microbial composition and their products.

4The aryl-hydrocarbon receptor (AhR) is an environmental sensor that binds to a variety of ligands, including xenobiotic ligands such as environmental pollutants (e.g., dioxin) and to physiological compounds derived from the digestion of dietary components by commensal microbiota. AhR plays a pleiotropic role in the maintenance of both the innate and adaptive immune systems in multiple organs and has been shown to be a transcriptional regulator for the development and function of several immune cells including Tregs, Th17, dendritic cells, and more recently, stem cells. In comparison to T cells, we currently lack understanding of how AhR regulates B cell responses. It is known that the levels of AhR expression vary over the lifetime of a B cell and that activation of AhR-dependent gene transcription following ligand binding contributes to the processes controlling B cell differentiation and lineage commitment. Activation of AhR prevents the differentiation of mature B cells into plasmablasts. Recently, we have shown that AhR is highly expressed in IL-10+CD19+CD21hiCD24hiB cells, a subset containing virtually all splenic Bregs, and that AhR promotes and preserves the immunosuppressive state of splenic Bregs by silencing a pro-inflammatory transcriptional program. The nature of the ligands that activate AhR and thus control the balance between pro-arthritogenic and regulatory B cell differentiation remains unknown.

5Here, we show that supplementation with the SCFA butyrate changes the availability of AhR’s endogenous ligands amplifying AhR-dependent gene transcription in CD19+CD21hiCD24hiB cells. Butyrate supplementation augments AhR ligand availability by supporting the growth of tryptophan bacteria that influence the metabolism of tryptophan, which increases the production of the main metabolite of serotonin, 5-hydroxyindole-3-acetic acid (5-HIAA). 5-HIAA, in turn, activates AhR-dependent gene transcription in B cells supporting Breg function and inhibiting germinal center (GC) B cell and plasma cell differentiation. This results in an amelioration of arthritis.

6Recent research has established that dysbiosis of the gut microbiota may be a contributing factor to RA pathogenesis. How dysbiosis contributes to abnormal immune cell function in human arthritic disease remains unknown. We have previously reported that the frequency of CD19+CD24hiCD38hiB cells, which contain the highest proportion of IL-10+Bregs, is inversely correlated with clinical disease severity in RA. Based on evidence that microbiota-derived SCFAs are essential for the maintenance of immunological homeostasis, we hypothesized that dysbiosis in RA could affect SCFAs, resulting in abnormal B cell homeostasis and a reduction in Breg frequency. We collected stool and paired serum samples from RA patients, and SCFA levels were compared to age- and sex-matched healthy controls (HCs) (patient demographics are reported in Table S1). Due to the limitations imposed by the low number of Bregs present in RA patients with active disease, only inactive patients were recruited to this study. We identified that there was a significant reduction in butyrate and propionate, and no difference in acetate, in stool samples of RA patients compared to HCs (Figure 1A). In serum samples, there was no difference in propionate or butyrate but a significant increase in acetate levels in RA patients compared to HCs (Figure S1A). As previously reported, we also found a trend for a negative correlation between age and butyrate or acetate levels in HCs but not RA patients (Figure S1B).

7To investigate whether changes in SCFA levels were associated with abnormal B cell and Breg homeostasis, peripheral blood B cell subset and IL-10+B cell frequency were enumerated in paired blood samples from RA patients and correlated with stool acetate, propionate, and butyrate levels. In RA patients, we reported a significant positive correlation between butyrate levels and the frequency of total CD19+CD24hiCD38hiB cells and with IL-10+B cells (Figures 1D and 1G). There was no significant correlation between propionate or acetate and CD19+CD24hiCD38hiB cells or IL-10+Breg frequency (Figures 1B–1F). There was no significant correlation between stool SCFAs and CD19+CD24hiCD38int (mature naive) or CD19+CD24hiCD38neg (memory) B cell frequency (Figures S1C and S1D). Despite these data only being correlative, they may suggest a role for butyrate in supporting Breg homeostasis in arthritic disease.

8To elucidate the mechanisms by which butyrate may influence the Breg compartment, we utilized the antigen-induced model of arthritis (AIA). In this model, both Breg function and arthritis severity are dependent upon the gut microbiota. Analysis of SCFA levels in the stool of arthritic mice revealed a reduction in butyrate and acetate during the acute and remission phase of arthritis compared to pre-arthritic mice (Figures 2A and S2A). Propionate was significantly reduced during the remission phase of disease, but there were no differences in the acute phase of disease compared to pre-arthritic mice (Figure S2A). These results suggest that the observed defect in SCFA production in arthritic mice, once established, cannot be reversed in spite of the reduced inflammation observed during disease remission. In line with reduced SCFAs, the bacterial families Lactobacillaceae, Rikenellaceae, and Bacteroidaceae were reduced in the stool of arthritic mice compared to naive mice (Figure S2B). Members of these bacterial families form a common functional group of bacteria that metabolize non-digestible carbohydrates into the immunogenic SCFA. Conversely, we detected an increase in Desulfovibrionaceae, Deferribacteraceae, Sutterellaceae, and Prevotellaceae families in the stool of arthritic versus naive mice (Figure S2B).

9Previously published research has demonstrated that supplementation with SCFAs, and in particular butyrate, has an immunosuppressive effect in diseases including diabetes and colitis. To evaluate the contribution of each individual SCFA in controlling the severity of arthritis and to determine the possible role of B cells in mediating suppression, acetate, propionate, and butyrate were supplemented in the drinking water of wild-type (WT) mice and B-cell-deficient (μMT) mice prior to disease induction. Control mice for both genotypes received drinking water that was salt and pH balanced (hereafter referred to as the control group). Only supplementation with butyrate, but not acetate and propionate, reduced arthritis in WT mice compared to control mice (Figures 2B, S2C, and S2D). Butyrate supplementation failed to suppress disease in B-cell-deficient mice (μMT) (Figure S2C), demonstrating that under these experimental conditions, B cells are key in mediating the beneficial effects of butyrate supplementation. As seen in WT mice, supplementation with acetate or propionate did not affect disease severity in μMT mice (Figures S2D and S2E). Furthermore, butyrate supplementation failed to suppress disease in mixed bone marrow chimeric mice lacking IL-10-producing B cells (Figure 2B), pinpointing the requirement of Bregs in the butyrate-mediated suppression of arthritis. Butyrate suppressed the severity of collagen-induced arthritis, further strengthening the anti-inflammatory role of butyrate in arthritic disease (Figure S2F).

10To investigate the effect of butyrate supplementation on both pro-arthritogenic cells and Bregs, we next took advantage of IL-10eGFP reporter mice, allowing the visualization of B cells actively transcribing IL-10. Amelioration of disease in butyrate-supplemented IL-10eGFP reporter mice was similar to WT mice (Figures 2C and 2D respectively showing clinical score and histological changes of the joints). There was no increase in serum butyrate levels in butyrate-supplemented versus control mice (Figure S2G). Suppression of disease by butyrate supplementation was associated with a reduction in TNFα, IL-6, and MCP-1 (CCL2) production by total lymphocytes isolated from the draining lymph node (LN) (Figure S3A) and by a decrease in IL-1β+CD11b+ and IL-6+CD11b− splenocyte frequency compared to control mice (Figures S3B and S3C). Analysis of the T cell compartment revealed a reduction in CD4+IL-17+T cell frequency and IL-17 production by total lymphocytes in the draining LN of butyrate-supplemented versus control mice (Figures S3D and S3E). Butyrate supplementation did not affect IFNγ production or CD4+IFNγ+T cell frequency, nor Foxp3+Treg frequency or Foxp3+Treg number (Figures S3D–S3F). However, butyrate-mediated suppression was reduced in mice when Tregs were depleted following anti-CD25 depleting antibody treatment (Figures S3G and S3H). Therefore, Tregs also play a role in mediating the suppression of arthritis by butyrate supplementation, supporting data in the literature showing the pleiotropic immunomodulatory effect of this SFCA.

11We next assessed the effect of butyrate supplementation on the differentiation of Bregs, identified here as IL-10eGFP+CD19+CD21hiCD24hiB cells. We found that IL-10eGFP+CD19+CD21hiCD24hiBreg number and frequency were similar between butyrate-supplemented and control mice (Figure 2E). The frequency and total number of IL-10eGFP+CD5+B cells, Tim-1+B cells, and CD5+CD1d+B10 cells, all of which have previously been ascribed with regulatory capacity, also remained unchanged following butyrate supplementation (Figures S4A–S4F). There was no difference in the frequency and/or number of splenic follicular (FO) B cells, transitional-1 (T1) B cells, or total CD19+CD21hiCD24hiB cells (Figures S4G–S4J). However, there was a significant reduction of CD19+CD138+Blimp-1+plasmablast and CD19+CD95+GL7+GC B cell frequency and number between butyrate-supplemented and control mice (Figures 2F and 2H). Blinded histological analyses further confirmed a reduction in the number of GCs per B cell follicle and in the size of GCs in the spleens of butyrate-supplemented versus control mice (Figure 2J). Thus, butyrate supplementation had increased the ratio of IL-10eGFP+CD19+CD21hiCD24hiBregs to plasmablasts and IL-10eGFP+CD19+CD21hiCD24hiBregs to GC B cells compared to control mice (Figures 2G and 2I).

12To determine whether butyrate supplementation affects the immunosuppressive function of Bregs, an equal number of IL-10eGFP+CD19+CD21hiCD24hiBregs was isolated from butyrate-supplemented or control IL-10eGFP reporter mice and transferred into syngeneic arthritic hosts. IL-10eGFP+CD19+CD21hiCD24hiBregs from butyrate-supplemented mice displayed enhanced suppressive capacity upon adoptive transfer compared to IL-10eGFP+CD19+CD21hiCD24hiBregs from control mice (Figure 2K). These results demonstrated that butyrate supplementation concurrently increases Breg suppressive capacity and limits GC B cell and plasmablast differentiation.

13We have recently demonstrated that there is higher expression of the AhR in IL-10eGFP+CD19+CD21hiCD24hiBregs and that activation of AhR, which can be used as a proxy for Il10 transcription, contributes to the induction of a transcriptional program that controls IL-10eGFP+CD19+CD21hiCD24hiBreg suppressive function. This, taken together with previous findings demonstrating that AhR suppresses plasmablast differentiation, led us to hypothesize that butyrate supplementation suppresses arthritis and alters B cell subset composition either directly or indirectly by activation of AhR in B cells. In line with this hypothesis, the expression of Cyp1a1, a prototypical reporter gene of AhR activation, was significantly upregulated in B cells isolated from butyrate-supplemented mice compared to control mice (Figure 3A). To confirm a role for AhR in the immune-modulatory effect of butyrate supplementation on the B cell compartment and arthritis severity, we took advantage of Ahrfl/−Mb1cre/+ mice, which have a B cell specific deletion of AhR. Similarly to WT mice, butyrate supplementation suppressed arthritis severity and CD4+IL-17+T cell frequency only in AhR-sufficient Mb1cre/+ mice but not in Ahrfl/−Mb1cre/+ mice (Figures 3B and S5A). As recently shown, B cells from Ahrfl/−Mb1cre/+ mice released less IL-10 compared to those isolated from Mb1cre/+ mice; a reduction that was not restored following butyrate supplementation (Figure 3C). Butyrate supplementation did not alter CD19+CD21hiCD24hiB cell frequency and number in Mb1cre/+ or Ahrfl/−Mb1cre/+ mice compared to control groups (Figure 3D). Corroborating the results in Figures 2F and 2H, butyrate supplementation reduced CD19+CD138+Blimp-1+plasmablast and CD19+CD95+GL7+GC B cell frequency and number in Mb1cre/+ mice but failed to suppress CD19+CD138+Blimp-1+plasmablast and CD19+CD95+GL7+GC B cell frequency and number in Ahrfl/−Mb1cre/+ mice (Figures 3E and 3F). Although Treg frequency and number were unaffected by butyrate supplementation (Figures S5B and S5C), CD3+CD4+CD25+Tregs isolated from butyrate-supplemented Mb1cre/+ mice displayed enhanced suppressive capacity upon adoptive transfer into WT mice (Figure S5D). In contrast, Tregs isolated from both control and butyrate-supplemented Ahrfl/−Mb1cre/+ mice failed to suppress disease on adoptive transfer (Figure S5D).

14As inflammation is a driver of Breg differentiation and function, and because Ahrfl/−Mb1cre/+ mice develop an exacerbated arthritic inflammation compared to Mb1cre/+ mice, we next tested the effect of butyrate supplementation in chimeric mice reconstituted with a 1:1 mix of bone marrow cells from CD45.1 WT and CD45.2 AhR−/− mice (Figure S6A). Under these conditions, WT and AhR−/− B cells are exposed to identical inflammatory signals following arthritis induction. Confirming the results in Figure 3, the frequency and number of WT CD45.1+CD45.2−IL-10+CD19+CD21hiCD24hiB cells was unaffected by butyrate supplementation, whereas AhR−/−CD45.1−CD45.2+CD19+CD21hiCD24hiB cells failed to differentiate into IL-10+CD19+CD21hiCD24hiBregs in both control and butyrate-supplemented mice (Figure S6B). In addition, butyrate supplementation reduced the frequency and number of plasmablasts and GC B cells within CD45.1 WT-derived cells but not in CD45.2 AhR−/−-derived cells (Figures S6C and S6D).

15To understand how butyrate supports Breg suppressive function and suppresses GC B cell and plasmablast differentiation, we compared the gene expression profiles and chromatin accessibility of CD19+CD21hiCD24hiB cells isolated from butyrate-supplemented and control Mb1cre/+ and Ahrfl/−Mb1cre/+mice. This population was chosen as we have shown that differentiation of CD19+CD21hiCD24hiB cells into IL-10+Bregs is dependent upon AhR (Figure S6). There were 412 significantly differentially expressed genes (DEGs) between control and butyrate-supplemented Mb1cre/+CD19+CD21hiCD24hiB cells (Figure 4A). There were more changes (566 significantly DEGs) in butyrate-supplemented versus control Ahrfl/−Mb1cre/+CD19+CD21hiCD24hiB cells. This suggests that, as well as being necessary to mediate some of butyrate’s effects on gene expression, AhR also represses the expression of a number of genes that would otherwise be altered by butyrate treatment (Figure 4A). Signaling pathway impact analysis (SPIA) revealed that the “protein processing in the endoplasmic reticulum” pathway, previously associated with the differentiation of B cells into plasma cells, was significantly downregulated by butyrate supplementation in Mb1cre/+CD19+CD21hiCD24hiB cells and significantly upregulated Ahrfl/−Mb1cre/+CD19+CD21hiCD24hiB cells (Figure 4B). Based on this observation, we interrogated DEGs in the B cell differentiation Gene Ontology term (GO:0030183) and compared the effect of butyrate supplementation on gene expression in both genotypes. B cell lymphoma 6 protein (Bcl6), a master regulator of GC B cell differentiation, and the orphan G protein-coupled receptor (Gpr183), important in extra-follicular plasmablast differentiation, were among the genes reduced in CD19+CD21hiCD24hiB cells from Mb1cre/+ mice compared to Ahrfl/−Mb1cre/+ mice after butyrate supplementation (Figure 4C). Conversely, the expression of Id2, a negative regulator of B cell maturation, was upregulated in CD19+CD21hiCD24hiB cells from Mb1cre/+ mice, but not from Ahrfl/−Mb1cre/+ mice after butyrate supplementation (Figure 4C). To investigate whether there was an AhR-independent mechanism in the Breg-mediated regulation of arthritis by butyrate supplementation, we performed a four-way comparison analysis among all the groups (Figure S7A). This analysis also highlights the baseline transcriptional changes between control Mb1cre/+ versus Ahrfl/−Mb1cre/+ mice, which we have defined in a previously published manuscript. We found that 71 significantly DEGs were regulated in both Mb1cre/+ versus Ahrfl/−Mb1cre/+ mice by butyrate supplementation (Figure S7A; Table S2). There were 195 significantly DEGs observed only in Mb1cre/+ mice but not Ahrfl/−Mb1cre/+ mice following butyrate supplementation after genotype confounding genes had been removed; the majority of these genes were structural proteins (Figure S7A; Table S3).

16Similarly to the baseline transcriptional changes between Mb1cre/+ versus Ahrfl/−Mb1cre/+ mice observed during the transcriptome analysis, there were clear differences in chromatin accessibility as measured by assay for transposase-accessible chromatin using sequencing (ATAC-seq) between control Mb1cre/+ mice and Ahrfl/−Mb1cre/+ mice (Figure S7B). However, corroborating the results in Figure 4C, there was decreased accessibility in several B cell maturation genes, including the Bcl6 and Gpr183 loci, upon butyrate supplementation exclusively in Mb1cre/+CD19+CD21hiCD24hiB cells (Figure 4D). ATAC-seq analysis also revealed that butyrate supplementation did not alter accessibility of the AhR:ARNT specific binding motifs, but did increase accessibility at binding motifs for transcription factors that have been identified to function alongside the AHR:ARNT heterodimer, including Esrra (estrogen receptor alpha), Creb1, and Rarb/Rarg (Retinoic acid receptor) (Figure 4E). We confirmed that, similarly to Tregs and monocytes, butyrate acted as a histone deacetylase inhibitor (HDACi) on splenic B cells in vitro, providing a partial explanation of its effect on the transcriptional and epigenetic landscape of CD19+CD21hiCD24hiB cells (Figure S7C).

17To investigate whether changes in the epigenetic and transcriptional profile of AhR+CD19+CD21hiCD24hiB cells and AhR−CD19+CD21hiCD24hiB cells following butyrate supplementation had altered their stability and ability to differentiate into IL-10 competent Bregs, we followed the fate of adoptively transferred CD19+CD21hiCD24hiB cells isolated from butyrate-supplemented and control WT or global AhR−/− in congenic CD45.1 recipient WT mice. A higher number of donor CD45.2+CD19+CD21hiCD24hiB cells were recovered post-transfer, and more transferred cells were IL-10+ when cells were isolated from butyrate-supplemented WT mice compared to control WT mice (Figures 5A–5D). The rate of cell recovery was not altered by butyrate supplementation when cells were isolated from AhR−/−mice and there was a failure of CD45.2+CD19+CD21hiCD24hiB cells to differentiate into IL-10+Bregs (Figures 5A–5D).

18The microbiota is an important contributor to the pool of endogenous AhR ligands, and we and others have previously shown that changes in the composition of the gut microbiota alters the differentiation of CD19+CD21hiCD24hiB cells into functionally suppressive Bregs. Having excluded a direct effect for butyrate in activating AhR, as butyrate did not upregulate the marker of AhR-activation Cyp1a1 compared to vehicle-treated B cells in vitro (Figure S7D), we investigated whether the endogenous microbiota or their metabolites are important in the butyrate-mediated suppression of arthritis and Breg maintenance. To address this, broad-spectrum antibiotic (ABX)-treated mice were given butyrate by oral gavage; this combination of antibiotics is known to ablate the majority of the gut microbiota. We found that the suppressive activity of butyrate depended upon the presence of the endogenous gut microbiota, as butyrate supplementation was ineffective at suppressing arthritis in ABX-treated mice (Figure 6A). In support of our previously published results showing that commensal microbiota is important in Breg differentiation, B cells isolated from ABX-treated mice expressed less Il10 mRNA compared to untreated controls, and this defect was not recovered after butyrate supplementation (Figure 6B).

19Having established that commensal microbes are required for butyrate to suppress arthritis, we compared the relative abundance of bacteria phyla in the stool of naive, control, and butyrate-supplemented arthritic mice using 16S rDNA amplicon sequencing. Butyrate supplementation induced a shift in the stool microbiota of arthritic mice, favoring a profile that was more similar to naive mice (Figure 6C). A detailed analysis of the bacterial composition revealed an increase in the abundance of the bacterial genera Allobaculum, Bifidobacterium, and Rhodosprillaceae_unclassified in butyrate-supplemented versus control mice (Figure 6D). Members of these bacteria genera have a previously described role in influencing the generation of tryptophan-derived metabolites, a family of ligands implicated in the activation of AhR. To understand whether changes in bacterial composition following butyrate supplementation altered the level of tryptophan-derived metabolites, we measured these metabolites in the stool of butyrate-supplemented and control mice. There were no differences in the amount of tryptophan, tryptamine, indole, and L-kynurenine in stool samples from butyrate-supplemented compared to control mice (Figures 6E and 6F). Indole-3-acetate and Indole-3-propionate levels were also measured but found to be below the limit of detection in all samples. There was, however, a significant increase in 5-HIAA, the main metabolite of serotonin (Figure 6G), and a significant reduction in the level of the kynurenine-derived metabolite kynurenic acid (KYNA) (Figure 6F).

20To directly address how the changes in 5-HIAA and KYNA levels affect AhR-dependent gene transcription in B cells, WT B cells were isolated from naive mice and stimulated with 5-HIAA and KYNA in vitro. Unlike KYNA, which only induced Cyp1a1 induction in B cells, 5-HIAA increased both Cyp1a1 and Il10 expression in B cells compared to vehicle-control-treated B cells (Figure 7A). Most importantly, treatment of WT mice with these AhR ligands in vivo demonstrated that 5-HIAA, but not KYNA, suppressed arthritis development and increased both Cyp1a1 and Il10 transcription in B cells ex vivo (Figures 7B and 7C). To examine the role for AhR in the immunosuppressive effect of 5-HIAA, we gavaged Mb1cre/+ mice and Ahrfl/−Mb1cre/+ mice with 5-HIAA. 5-HIAA suppressed arthritis in Mb1cre/+ mice but not in Ahrfl/−Mb1cre/+ mice (Figure 7D). Finally, to explore the role of 5-HIAA in the ability of butyrate supplementation to suppress arthritis, mice were treated with the tryptophan hydrolase (TPH) inhibitor L-para-chlorophenylalanine (PCPA), which is known to reduce 5-HIAA and serotonin biosynthesis. In mice treated with PCPA, butyrate supplementation lost its ability to suppress arthritis when compared to vehicle-treated control mice (Figure 7E). Collectively, these data demonstrate that butyrate supplementation increases the production of 5-HIAA, a newly identified AhR ligand in B cells, which mediates the suppressive effect of butyrate supplementation in vivo.

21Bregs are generated in the periphery in response to bacterially derived and inflammatory signals. Whereas more is understood regarding how inflammation and inflammatory cytokines drive Breg differentiation, the participation of microbiota in Breg biology remains underexplored. We report that RA patients and arthritic mice have decreased levels of the SCFA butyrate, a microbial metabolite produced by commensal bacteria after dietary fiber fermentation. We also report that butyrate supplementation attenuates disease severity in mice by supporting AhR+Breg function and suppressing mature B cell subset differentiation. Based on our results, we propose the hypothesis that butyrate overcomes the need for inflammatory stimuli in Breg differentiation by increasing the availability of the 5-HIAA, which directly activates AhR and Il10 transcription. This hypothesis needs further exploration and could offer important immunological and physiological groundwork for future therapeutic interventions.

22Recent literature has demonstrated that butyrate can alter the function of a wide variety of immune cells. In agreement with this, we found that butyrate supplementation of arthritic mice lowered effector cytokine production by CD11b+ splenocytes and enhanced Treg suppressive function upon adoptive transfer. Butyrate has been shown to induce Foxp3+ Tregs both directly, by acting as an HDACi, and indirectly, by promoting anti-inflammatory properties in macrophages and dendritic cells by engaging G protein-coupled receptors GPR43 and GPR109A. Building on these findings, we found that Tregs only displayed enhanced suppressive function when isolated from butyrate-supplemented mice with a fully functional Breg compartment. This supports published data demonstrating that Treg homeostasis is altered in mice lacking IL-10-producing B cells. Here, we also describe a previously unappreciated role for butyrate in altering B cell differentiation and function in both mice and humans with arthritic disease. We found that butyrate levels are reduced in RA patients compared to HCs, and that in RA patients, butyrate levels correlate with IL-10-producing B cell frequency. Interestingly, we observed that in HCs there is a trend for a reduction in butyrate levels in older individuals, which may contribute to changes in autoimmune disease susceptibility in later life. Unlike in humans, there was not a direct correlation between butyrate levels and Breg frequency in mice, likely due to differences in disease chronicity, treatment, and tissue analyzed. However, interrogation of the interaction between butyrate and B cells using a murine model of arthritis demonstrated a sophisticated system whereby butyrate alters AhR-dependent gene transcription, including key B cell differentiation genes and immunoregulatory genes serving to support Breg suppressive function and inhibit B cell maturation. Notably, we found that μMT mice, which lack both regulatory and inflammatory (e.g., GC B cells and plasmablasts) B cells, have equivalent disease severity to WT mice in this model of arthritis, but that chimeric mice, which exclusively lack IL-10-producing B cells, develop exacerbated disease compared to chimeric mice with WT B cells. This demonstrates a fundamental role for IL-10-producing B cells in suppressing arthritic severity following butyrate supplementation in this model. These data demonstrate for the first time that a microbially derived metabolite can control the balance between regulatory and mature B cell subsets.

23Our data determined that butyrate supplementation requires a fully competent endogenous microbiota to exert its anti-arthritogenic capabilities on the B cell compartment. Butyrate supplementation shifted the microbiota to increase relative abundance of Allobaculum, Bifidobacterium, and Rhodosprillaceae_unclassified, genera which have been shown to influence tryptophan metabolism. One possible explanation for this shift is that butyrate possesses antimicrobial activity that targets pathobionts, creating a niche for the growth of tryptophan-metabolizing species. In the agricultural industry, butyrate is an established component of chicken feeds used to control the growth of pathogenic bacteria. In addition to a direct bactericidal effect, butyrate enhances the microbicidal function of macrophages by altering their metabolism and eliciting the production of anti-microbial peptides, which may control out-growth of pathogenic components of the gut microbiota. Another complementary explanation justifying the observed shift in bacterial communities is that butyrate acts as a nutrient for beneficial bacteria. At present, we cannot exclude that changes observed in the gut microbiota following butyrate supplementation could be the result of reduced inflammation. Unfortunately, due to the intertwined response between microbiota and inflammation, it is difficult to extricate whether the butyrate effect on bacteria is direct, or mediated by cells or other anti-inflammatory mediators. Future studies will be performed to investigate if the effect reported here is due to changes in inflammation or to direct effect on the microbiota. Our findings support the notion that prebiotics supplementation could be used to restrain inflammation in systemic autoimmune disease with no obvious gut-related pathogenesis.

24AhR is an environmental sensor detecting both xenobiotic ligands and physiological compounds generated by host cells, microbiota, and diet. Among the microbiota-derived ligands for AhR, an important family are tryptophan-derived metabolites. For example, it has been previously demonstrated that tryptophan is endogenously metabolized into tryptamine and indole-3-acetic acid, which directly bind to AhR. More recently, expression of tryptophanase by certain microbiota species has been shown to process tryptophan into indoles and its 3-substituted derivatives, which also act as agonists for AhR. In addition, L-kynurenine and kynurenic acid, which are produced following metabolism of tryptophan by indoleamine 2,3-dioxygenase (IDO), can also activate AhR in immune cells. In this study, we did not detect any variation in the levels of tryptophan, tryptamine, L-kynurenine, or indole, yet we observed a reduction in kynurenic acid (KYNA). We also found the levels of indole-3-substituted derivatives to be below the limit of detection, suggesting that these pathways are unaffected by butyrate supplementation. Rather, our data suggest an additional mechanism by which 5-HIAA, the main metabolite of serotonin (5-HT), activates AhR in B cells following butyrate supplementation. Similarly to T cells, where it has been shown that different AhR ligands drive either Treg or Th17 differentiation, we show that both KYNA and 5-HIAA can activate AhR-dependent gene transcription in B cells, but only 5-HIAA concomitantly induces AhR signaling and Ill0 transcription in B cells. The production of the tryptophan-derived neurotransmitter 5-HT in the gut is intimately connected with the presence and species of the gut microbiota. As well as regulating diverse physiological processes in both the brain and gut, 5-HT also has a proposed immune-modulatory function, including the promotion of B cell proliferation, induction of cytokine release by monocytes, and changing in capabilities of dendritic cells to present antigen and activate T cells. Here, we determine that 5-HT’s main metabolite 5-HIAA upregulates AhR-dependent gene transcription and ll10 transcription in B cells and is immunoregulatory in arthritis. Our data support a recently established link between the serotonergic and AhR pathways, showing the efficacy of 5-HT in inducing Cyp1a1 expression via AhR in intestinal epithelial cells. It also adds to accumulating evidence that butyrate can induce 5-HT release by neural enterochromaffin cells in the gut. We suggest that, as well as regulating gut homeostasis and peristalsis, the butyrate-serotonin-AhR axis also acts to influence Breg homeostasis.

25The data in this study suggest that gut-microbiota-derived metabolites control many aspects of B cell development and Breg function. Moreover, it suggests that the threshold for Breg induction in response to inflammatory stimuli could potentially be lowered in AhR-activating ligand-rich environments. Future work will be used to establish whether the dynamics of Breg activation is orchestrated by the complementary effects of inflammation and certain microbial-derived stimuli. To date, due to the heterogenous nature of the Breg response, researchers have been unable to ascertain how to harness the suppressive function of these cells for therapeutic intervention. This report addresses this gap and reveals that manipulation of microbial end-products that can be supplemented during dietary interventions could serve this purpose well.

26Here, we demonstrate that reduced butyrate levels in RA patients correlate with reduced Breg frequency in peripheral blood. However, as for many human studies, our results are only correlative. One major caveat of our human study was the inability to include patients with active disease (as they have a very low levels of Bregs) as well as patients with other autoimmune diseases. In future studies, we will aim to include RA patients with both active and inactive disease as well as patients with other inflammatory diseases to ascertain whether changes in SCFA are unique to RA or a common feature of all autoimmune diseases. It would also be important to validate our results in another RA cohort of patients, as microbiota composition changes according to geographical region and diet. Moreover, it has been previously shown that the microbiota is not standardized in every animal house, which based on our results, is likely to have significant impacts on the efficacy of butyrate supplementation to suppress severity of arthritis and support Breg function. Future work is therefore necessary to identify particular 5-HIAA-inducing bacterial species that mediate the effect of butyrate supplementation to circumvent these limitations.

27Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Professor Claudia Mauri (c.mauri@ucl.ac.uk). This study did not generate new unique reagents.

28Peripheral blood (PB) and stool samples from healthy adult controls and rheumatoid arthritis patients were obtained with fully informed consent as approved by the London-Bloomsbury and Riverside Research Ethics Committees (IRAS: 191626 and 46584) in accordance with the Declaration of Helsinki. Clinical and demographical data are shown in Table S1.

29B6(Cg)-Il10tm1.1Karp/J (IL-10eGFP) mice were as described and given courtesy of Professor Chris Karp; B6.129S2-Ighmtm1Cgn/J (μMT) mice were purchased from Jackson, USA; 129(B6)-Il10tm1Cgn/J (Il10−/−) mice were kindly given courtesy of Professor Fiona Powrie (Kennedy Institute of Rheumatology Oxford University); B6.SJL-Ptprca Pepcb/BoyJ (CD45.1) mice were kindly given courtesy of Professor Derek Gilroy (University College London); B6.129-Ahrtm1Bra/J (Ahr−/−), Mb1cre/+ (given courtesy of Prof. M. Reth) and Ahrfl/-Mb1cre/+ mice were kindly provided by Professor Brigitta Stockinger (Francis Crick Institute): (all mice were on a C57BL/6 background). For detailed information please see Key Resources Table. C57BL/6 WT mice and DBA/1J WT mice were purchased from Envigo, UK. Sex-matched male and female mice between 6-12 weeks of age were used for antigen-induced arthritis experiments. Analysis of sex-dependent effects was not carried out as there is no reported sex-bias in this model. Male DBA/1J mice 8 weeks of age were used for collagen induced arthritis experiments. Mice were housed at 20-24°C, 45%–64% humidity, and at a 12- light/dark cycle. Experimental mice were feed Harlan Teklad pellets 2018 (18% protein) and breeding mice were feed Harlan Teklad pellets 2010 (19% protein). Mice were specific-pathogen free (Health screening (Full-FELASA profile) was performed annually) and bred and maintained at the animal facility at University College London. All experiments were approved by the Animal Welfare and Ethical Review Body of University College London and authorized by the United Kingdom Home Office.

30AIA was induced and assessed as previously described. Briefly, mice were injected subcutaneously (s.c.) at the tail-base with 200 μg of methylated BSA (mBSA; Sigma-Aldrich) emulsified in 100 μL of complete Freund’s adjuvant (CFA); incomplete Freund’s adjuvant (IFA) (Sigma-Aldrich) containing 3mg/mL of Mycobacterium Tuberculosis (Difco). After seven days, mice received an intra-articular (i.a.) injection of 10 μL of phosphate-buffered saline (PBS) containing 200 μg mBSA in the right knee and 10 μL PBS alone in the left knee as a control. Joint size was measured using calipers (POCO 2T; Kroeplin Längenmesstechnik), and swelling was calculated as a percentage increase in size between inflamed and control knee. Affected joints were removed post-mortem, fixed in 10% (w/v) buffered formalin, and decalcified in 5% EDTA. The joints were subsequently embedded in paraffin, sectioned, and stained with hematoxylin and eosin Y (H&E). Briefly, the sections were rehydrated in PBS, stained with hematoxylin, washed, counterstained with eosin Y, then washed and dehydrated in sequentially higher concentrations of ethanol from 75% to 100%. The sections were scanned using the NDP NanoZoomer (Hamamatsu) at 20x magnification and analyzed with the NDP view software. The joint sections were scored blinded: 1, Normal = no damage; 2, mild = minimal synovitis, cartilage loss, and bone erosion limited to discrete foci; 3, moderate = synovitis and erosion present, but normal joint architecture intact; and 4, severe = extensive erosion and joint architecture disrupted.

31CIA was induced and assessed as previously described. Briefly, 8-week-old male DBA/1J mice were injected intra-dermally (i.d.) at the tail-base with 100 μg of Bovine Collagen Type II (Chondrex) emulsified in 50 μL complete Freund’s adjuvant (CFA); incomplete Freund’s adjuvant (IFA) (Sigma-Aldrich) containing 4mg/mL of Mycobacterium Tuberculosis (Difco). The clinical severity of arthritis was graded as follows: 0, normal; 1, slight swelling and/or erythema; 2, pronounced edematous swelling; 3, pronounced edematous swelling plus joint rigidity; and 4, laxity. Each limb was graded, allowing a maximal clinical score of 16 for each animal. All clinical evaluations were performed in a blinded manner.

321 week prior to the induction of arthritis the drinking water of mice was supplemented with sodium acetate, sodium propionate or sodium butyrate (all 150mM; Sigma-Aldrich) and changed every 3 days as described. A control group received sodium chloride. Mice were maintained on short-chain fatty acids (SCFAs) throughout the duration of the experiment. For RNA-seq and ATAC-seq analysis, mice were gavaged daily with 500mg/kg of sodium butyrate to reduce variation caused by individual differences in daily water intake. Control mice received a gavage of 500mg/kg of sodium chloride. For antibiotic-treated experiments, one week prior to induction of arthritis, vancomycin (500mg/L;Sigma-Aldrich), Neomycin (1g/L; Sigma-Aldrich), and Metronidazole (1g/L;Sigma-Aldrich) were added to drinking water as described. Untreated and treated mice were then gavaged daily with 500mg/kg of sodium butyrate or sodium chloride as a control. Mice were maintained on antibiotics throughout the duration of the experiment. For L-para-chlorophenylalanine (PCPA) experiments, mice were supplemented with butyrate as described above and gavaged daily with PCPA (4mg per mouse) in a suspension of 0.5% methyl cellulose and 0.01% Tween 80. Control mice received vehicle alone. For Treg depletion experiments, mice were supplemented with butyrate as described above and injected intra-peritoneally with 250 μg of anti-CD25 (PC-61.5.3) or appropriate isotype control on two days prior to and on the day of the commencement of butyrate-supplementation, two days prior and on the day of the subcutaneous injection, and two days prior to and on the day of intra-articular injection.

33Mice were gavaged daily from 1 week prior to the induction of arthritis and throughout the experiment with either 5-Hydroxyindole-3-acetic acid (5-HIAA, 0.5mg per mouse) or kynurenic acid (KYNA, 0.125mg per mouse) dissolved in oil. Control mice received vehicle alone.

34Chimeric mice, and appropriate controls, were generated as previously described. Recipient μMT mice received 800cGy gamma-irradiation via a caesium source. 5 h following irradiation, recipients received 10x106 donor bone marrow cells. To generate mice in which the absence of IL-10 was exclusively restricted to B cells, WT mice were reconstituted with mixture of bone marrow consisting of 80% from μMT (B cell deficient) with 20% from Il10−/− mice. Control mice received 80% from μMT and 20% bone marrow from WT mice (to give a normal B cell compartment). To generate CD45.2+AhR−/−CD45.1+ congenic chimeric mice, WT mice were reconstituted with 10x106 donor bone marrow cells containing 50% from CD45.1+ WT mice and 50% from CD45.2+ Ahr−/− mice. Chimeras were left to fully reconstitute over at least 8 weeks before use in AIA experiments.

35Spleens and inguinal lymph-nodes were dissected post-mortem and mashed through a 70 μm cell strainer (BD Biosciences), as previously described. For splenocytes suspensions, erythrocytes from spleens were lysed using Red Cell Lysis Buffer (Sigma-Aldrich). Total B cells were negatively purified by magnetic separation, according to manufacturer’s instructions (Miltenyi Biotec). IL-10eGFP+CD21hiCD24hi Bregs were isolated from IL-10eGFP reporter mice and CD19+CD21hiCD24hi B cells were isolated Mb1cre/+, Ahrfl/-Mb1cre/+, WT or Ahr−/− mice by cell sort. Lymphocytes were analyzed at day 7 post-disease onset unless otherwise stated. Peripheral blood mononuclear cells (PBMC) from RA patients and healthy controls were prepared by density gradient centrifugation using Ficoll Plaque™ plus (GE Healthcare).

36For assessment of murine IL-10 production, B cells from either male or female mice were negatively purified by magnetic separation (Miltenyi Biotic) (purity < 95%–98%) and cultured for 48 h with LPS (1 μg/mL; Sigma-Adrich), with anti-IgM added for the last 24 h of culture (10 μg/mL; Jackson ImmunoResearch) and supernatants from B cell cultures were harvested and analyzed for cytokines using standard sandwich ELISA kit (IL-10; R&D systems) and performed according to manufacturer’s instructions. For assessment of Cyp1a1 and Il10 induction by PCR, B cells were negatively purified by magnetic separation and cultured for 6 h with 5-HIAA (10 μM; Sigma-Adrich) or sodium butyrate (500 μM; Sigma-Adrich). For assessment of inflammatory cytokine production, total lymphocytes were stimulated overnight with 50ng/mL PMA and 250ng/mL Ionomycin. Supernatants were harvested and analyzed for cytokine production using the LEGENDplex Mouse Inflammation Panel (Biolegend) and performed according to manufacturer’s instructions.

37RNA from isolated B cells was extracted using Arcturus Picopure RNA isolation kit (ThermoFisher Scientific) and RNA was reverse transcribed using an iScript cDNA synthesis kit (Bio-Rad), according to manufacturer’s instructions. qPCR was carried out on the cDNA samples using iQ SYBR® Green Supermix (Bio-Rad), according to manufacturer’s instructions. Primers were used at a final concentration of 0.5 μM. A Quantitect primer for Cyp1a1 was purchased from QIAGEN. Primers for β-Actin were custom designed with the following sequences: Act Forward (5′-AGATGACCCAGATCATGTTTGAG-3′); Act Reverse (5′- AGGTCCAGACGCAGGATG-3′) and Il10 as previously described. qPCR data were calculated as the ratio of gene to β-Actin expression by the relative quantification method.

38Flow cytometry was performed as previously described using directly conjugated monoclonal antibodies (abs) and dead cells were excluded using a live/dead discrimination dye (Thermo Scientific) prior to analysis, as previously described. Briefly, for extracellular multi-color analysis, cells were stained at 4°C for 20min as described. For ex vivo analysis of intra-nuclear transcription factors, following extracellular staining, cells were fixed for 20min with Fix Perm (eBioscience) and following permeabilization were incubated with intracellular Abs. For detection of IL-17 and IFN-γ cells were suspended at 5x106 cells/mL in complete medium with PMA (50ng/mL; Sigma-Aldrich), Ionomycin (250ng/mL; Sigma-Aldrich) and GolgiPlug (BD Biosciences) for 4 h. For detection of IL-10 by murine B cells, splenocytes were suspended at 5x106 cells/mL in complete medium with PMA (50ng/mL; Sigma-Aldrich), Ionomycin (250ng/mL; Sigma-Aldrich) and Brefeldin A (BD Biosciences) for 4 h. For detection of IL-10 by human B cells, total PBMC were cultured with CpGB ODN2006 (1 μM; Invivogen) for 72 h prior to stimulation with PMA (50ng/mL), Ionomycin (250ng/mL) and Golgiplug for 4 h. Cells were then stained with surface markers followed by permeabilization and incubation with intracellular Abs. IL-10eGFP expression, was analyzed ex vivo without fixation, and 4,6-diamidino-2-phenylindole (DAPI) (1 μg/mL; Sigma-Aldrich) was added just before acquisition. Splenocytes from WT mice were used as a control to set gates for IL-10eGFP+ cells. Flow cytometric data were collected on an LSRII or LSR Fortessa (BD PharMingen) using FACS Diva software. Data were analyzed using Flowjo (Tree Star). For adoptive transfer experiments, B cell subsets were stained with antibodies against CD19, CD21, and CD24, dead cells were excluded by the use of DAPI (1 μg/mL) and cells were sorted on a FACS aria (BD PharMingen). For RNA-seq and ATAC-seq experiments, a dump channel using antibodies against CD3, CD4, CD8a, CD11b, CD11c, F4/80, Ly6C/G, erythroid cells (TER-119) and TCRγδ (H57-597) was included. Sort purity of B cell subpopulations was routinely > 95%.

39B cells were isolated from the spleen of either butyrate-supplemented mice or control IL-10eGFP reporter mice seven days post-disease onset and stained with antibodies against CD19, CD21, and CD24. B cell subsets were FACS sorted using gates drawn according to previous reports. IL-10eGFP+CD19+CD21hiCD24hiBregs (2.3 × 105) were transferred intravenously into recipient mice on the day of intra-articular injection. The control group (no transfer) received a PBS injection. For congenic transfer experiments, CD45.2+CD19+CD21hiCD24hi B cells (0.75 × 106) were transferred from control mice or mice that have received butyrate-supplementation into recipient CD45.1+ mice.

40CD3+CD4+CD25+ were isolated from butyrate-supplemented and control Mb1cre/+ and Ahrfl/-Mb1cre/+ mice seven days post-disease onset and CD3+CD4+CD25+Tregs (2x105) were transferred intravenously into recipient WT mice on the day of intra-articular injection. The control group (no transfer) received a PBS injection.

41Spleens were dissected and embedded into optimal cutting temperature compound (OCT, Tissue-Tek) and snap-frozen for cryo-sectioning (6 μm). Slides were incubated in 100% ethanol to fix for 5-10min (4°C), followed by rehydration in PBS for 5min (4°C). The sections were blocked with 10% normal goat serum and 0.3% TX-100 (20 min at RT) and then incubated with primary antibodies for 2 h at RT. Primary antibodies: rat anti-mouse GL7 (Thermo Fisher), B220-PE (BD). Secondary antibody: AF647 – conjugated anti-rat IgM (1 h at RT). The slides were mounted in Vectashield with DAPI (Vector Labs). Whole slide fluorescent images (20x) were taken on a Zeiss Axio Scan Z1 microscope using the 365 nm LED to detect DAPI staining in the nuclei, the 470 nm LED for GFP detection (IL-10), the 555 nm LED for PE detection (B220) and the 625 nm LED for the detection of Alexa fluor 647 (GL7). Scans were analyzed using Leica Software and Fiji (ImageJ).

42Individual stool pellets were weighed into clean Eppendorf tubes and homogenized in 1ml of 50% methanol. After centrifugation at 13,000xg for 5mins to remove particulate matter, 200 μL of the clear supernatants were derivatized as previously reported. Briefly, the clear supernatants were spiked with 2-ethylbutyric acid as an internal standard and the mixture incubated with 2-Nitrophenylhydrazine hydrochloride (NPH) at 60°C for 20mins, with 1-Ethyl-3-(3-dimethylaminopropyl) carbodiimide (EDC) in pyridine as catalysts. The reaction was then terminated and color allowed to develop by the addition of potassium hydroxide in methanol, followed by incubation at 60°C for a further 20mins. After cooling, the mixture was acidified by the addition of phosphoric acid and the derivatized fatty acids extracted into diethyl ether. After drying down the ether extracts under a gentle stream of nitrogen gas, the resulting fatty acid hydrazides were dissolved in methanol for high performance liquid chromatography (HPLC) analysis.

43Separation of short-chain fatty acid (SCFA) hydrazides was performed by injecting 25 μL onto on a C8 Hypersil MOS2 column (250 × 4.6mm, 5 μm particle size) and eluting using a linear gradient of 20%–90% acetonitrile against water over 17min at a flow rate of 1.5ml/min. Compounds eluting from the column were monitored by UV/Vis absorption using a measurement wavelength of 400nm, and quantitated by integration of peak area. Standard curves (20.0-0.1 μM) were constructed using pure compounds as follows: Succinic acid (Rt 3.92 min), Lactic acid (Rt 4.72 min), Acetic acid (Rt 5.067 min), propionic acid (Rt 6.173 min), iso-butyric acid (Rt 7.387 min), butyric acid (Rt 7.587 min), 2-methylbutyric acid (Rt 8.733 min), isovaleric acid (Rt 8.947 min), n-valeric acid (Rt 9.24 min), hexonoic acid (Rt 10.707 min), 2-ethylbutyric acid (Rt 9.88 min) hexanoic acid (Rt 10.707 min), Pyruvic acid (Rt 12.773).

44Plasma samples were analyzed as previously described with modifications to measure concentration of unlabelled acetate propionate and butyrate. Internal standard solution (1,2-13C-sodium acetate 40 μmoles, d5-sodium propionate 4 μmoles, 1,2,3,4-13C-sodium butyrate 2 μmoles), was added to 200 μL plasma. Samples were deproteinized with 20ul 20% w/v sulfosalicylic acid and centrifuged for 5mins at 15000 rpm. A 1cm Carboxen/PDMS coated solid-phase microextraction fiber (SPME, Supelco) was inserted into 120 μL supernatant, and analytes adsorbed for 5min. The SPME fiber was then analyzed by gas chromatography mass spectrometry GC/MS, (Thermo Trace GC/DSQ II or Agilent 6890 with 5973N). The inlet temperature was 270°C, and samples were desorbed for 6min and analyzed on HP-FFAP capillary column (25 m x 0.2mm, 0.33 μm film thickness, 0.9ml/min helium carrier flow) at a 1:10 split ratio. The temperature program was: 80°C for 3min, increased to 170°C at 15°C·min−1 and 170°C held for 1min. Ions at m/z 60 were monitored for acetate and butyrate, m/z 62 for 1,2-13C- acetate and 1,2,3,4-13C-sodium butyrate, m/z 74 for propionate and m/z 79 for d5- propionate. Concentrations were calculated by peak area ratios with the respective internal standards.

45Individual faecal pellets were weighed into clean Eppendorf tubes and homogenized in 200 μL of methanol. After centrifuging at 13,000xg for 5min to remove particulate matter, the clear supernatants were diluted 1 in 10 in methanol and subject to high performance liquid chromatography (HPLC) analysis.

46Separation of indoles was performed by injecting 20 μL onto on an ODS Hypersil column (150 × 4.6mm, 3 μm particle size) and eluting using a linear gradient of 5%–100% acetonitrile in 10mM ammonium formate buffer, pH 3.5 over 20min at a flow rate of 0.8ml/min. Compounds eluting from the column were monitored by both fluorescence detection (λex: 275nm, λem: 352nm) as well as by UV/Vis absorption using an online PDA detector (scanning 200-650nm), and quantitated by integration of peak area. Standard curves (20.0-0.1 μM) were constructed using pure compounds as follows: tryptophan (Rt 6.97 min), tryptamine (Rt 13.04 min), indole (Rt 14.41 min), indole-3-acetic acid (Rt 11.42 min), indole-3-propionic acid (Rt 13.7), 3-methylindole (Rt 15.87 min), indole 3-carboxaldehyde (Rt 10.96 min), 5-Hydroxyindole-3-acetic acid (Rt 8.09 min).

47Separation of L-Kynurenine and Kynurenic Acid was performed by injecting 25 μL onto on an ODS Hypersil column (150 × 4.6mm, 3 μm particle size) and eluting using a linear gradient of 1%–10% acetonitrile in 10mM ammonium formate buffer, pH 3.5 over 17min at a flow rate of 0.8ml/min. Compounds eluting from the column were monitored by both fluorescence detection (λem: 364nm, λem: 480nm for 7.5 min followed by λem: 330nm, λem: 390nm for the remainder of the run) as well as by UV/Vis absorption using an online PDA detector (scanning 200-650nm), and quantitated by integration of peak area. Standard curves were constructed using pure compounds as follows: L-Kynurenine (Rt 5.97 min) and Kynurenic Acid (Rt 10.6 min).

48B cells were negatively purified from WT mice and culture for 18 h with 500 μM of butyrate (Sigma-Aldrich). B cells were the lysed and protein from 10x106 cells was resolved by SDS-PAGE, transferred to polyvinylidene fluoride (PVDF) membranes (Amersham), and blotted using anti-H3K27ac (Abcam) and anti-pan-H3 (Abcam). Bound antibodies were revealed with HRP-conjugated species-specific secondary antibodies using ECL substrate (Amersham).

4920-50mg of faecal material was extracted using the QIAmp DNA mini kit (QIAGEN). Extraction was carried out as per the manufacturer’s protocol with an additional cell disruption step by bead beating using lysing matrix E (MP Biolmedicals) at 50 Hz for 1min (Tissuelyser LT, QIAGEN). Two negative extraction controls were included. Barcoded primers spanning the V3-V4 region of the 16S rRNA gene were designed according to to include Illumina adaptor, an 8 nucleotide barcode sequence, a 10 nucleotide pad sequence, a 2 nucleotide linker, and a gene-specific primer:: 341F-CCTACGGGNGGCWGCAG or 805R-GACTACHVGGGTATCTAATCC. (Sigma-Aldrich, Dorset, UK). Extracted DNA samples were amplified with different barcode combinations using the Taq Core PCR kit (QIAGEN) as per manufacturer’s instructions with forward and reverse primers at 0.5μM each. A Microbial Community Standard (Zymo Research) of known bacterial composition was also amplified to assess any bias and error rates. The PCR cycling conditions were as follows: initial denaturation at 95°C for 3min, 30 cycles of 95°C for 30sec, 54°C for 30sec, 72°C for 10min and a final extension at 72°C for 10min. PCR products were purified with AMPure beads (0.7x, Beckman Coulter) and quantified using the Qubit dsDNA High Sensitivity Assay Kit (ThermoFisher). Samples were then pooled to create libraries with approximately equal concentrations of 16S rRNA amplicons from each sample. The pooled library was quality and quantity checked using the High Sensitivity D1000 ScreenTape assay (Agilent Technologies) and a NEBNEXT library quantification kit (New England Biolabs). The pooled library was spiked with 10% PhiX (Illumina) and sequenced on an Illumina MiSeq using the Reagent Kit V2 with 500 cycles (Illumina) and custom primers as previously described. The open-source software Mothur V1.35.13 was used for initial bioinformatic analysis of the sequencing data. Raw sequencing data was demultiplexed and processed according to the online Mothur SOP. Sequences were trimmed and those with ambiguous bases were discarded. Suspected chimeric sequences were identified using VSEARCH and removed. Phylogenetic identification of each OTU was achieved by aligning sequences to the SILVA 16S alignment database (v128). Sequences that did not meet a 97% similarity threshold were discarded of. Sample reads were rarefied to 50,000 reads prior to further analysis. OTU values generated by Mothur were further analyzed using R-studio (phyloseq) or GraphPad Prism (la Jolla, USA) v. 400 Software for Apple Mac.

50Splenic CD19+CD21hiCD24hiB cells were isolated from butyrate-supplemented and control Mb1cre/+ and Ahrfl/-Mb1cre/+ mice at day 7 post-disease onset. Dead cells were excluded using DAPI. Total RNA was isolated from these populations using the Picopure RNA isolation kit (ThermoFisher Scientific), according to manufacturer’s instructions. 60bp single reads were sequenced on 3 lanes of an Illumina hiseq. 130-500ng of total RNA was fragmented followed by reverse transcription and second strand cDNA synthesis. The double strand cDNA was subjected to end repair, A base addition, adaptor ligation and PCR amplification to create libraries. Libraries were evaluated by Qubit and TapeStation. Sequencing libraries were constructed with barcodes to allow multiplexing of samples in 3 lanes. Around 23-43 million single-end 60-bp reads were sequenced per sample on an Illumina HiSeq 2500 V4 instrument. Poly-A/T stretches and Illumina adapters were trimmed from the reads using cutadapt. Resulting reads < 30bp were discarded. Reads were mapped to the Mus musculus GRCm38 reference genome using STAR. Gene annotations were applied from Ensembl (EndToEnd option and outFilterMismatchNoverLmax was set to 0.04). Gene expression levels were quantified using htseq-count (“HTSeq,” n.d.), using the gtf above. Transcripts per million (TPM) values were estimated independently using Kallisto.

51Differential expression analysis was carried out using the default settings of the edgeR algorithm. Genes were considered differentially expressed when the false discovery rate (FDR) adjusted p-values were < 0.05. Signaling Pathway Impact Analysis (SPIA) was used to detect significantly over- or under-expressed pathways, with the Kyoto Encyclopedia of Genes and Genomes (KEGG) Pathways database employed as a reference. The full mouse genome was used as background for enrichment.

52Splenic CD19+CD21hiCD24hiB cells were isolated from butyrate-supplemented and control Mb1cre/+ and Ahrfl/-Mb1cre/+ mice as above for RNA-seq. After sorting, 40,000 cells were washed with 1xPBS (10% FCS). The cell pellet was prepped for sequencing by using the Nextera DNA library preparation kit (Illumina). Briefly, 10.5 μL nuclease free water, 12.5 μL 2x Transposase buffer, 2 μL transposase and 0.25 μL digitonin (0.05%) per reaction were added to the cell pellets. Cells were incubated at 37°C for 30mins. DNA was then purified using a MinElute PCR purification kit (QIAGEN), according to manufacturer’s instructions. Following DNA purification, 1μl of eluted DNA was used in a qPCR reaction to estimate the optimum number of amplification cycles. Library amplification was followed by solid phase reversible immobilization (SPRI) size selection to exclude fragments larger than 1,200bp. DNA concentration was measured with a Qubit fluorometer (Life Technologies). Library amplification was performed using custom Nextera primers. The libraries were sequenced by the Biomedical Sequencing Facility at CeMM using the Illumina HiSeq4000 platform and the 50bp single-end configuration.

53Sequenced reads were trimmed for adaptor and Nextera sequences with skewer and reads were mapped to mm10 reference genome using bowtie2 v2.2.4 with the “-very-sensitive” parameter. Duplicate reads were marked with sambamba and reads were filtered for quality above 30. Peaks for ATAC-seq samples were called with MACS2 version 2.1.1.20160309 using the “-nomodel -extsize 147” parameters. A consensus region set was generated by creating a union of all samples’ summits extended by 250bp, from where blacklisted regions defined by the ENCODE project for mm10 were filtered out. Quantification of chromatin accessibility for each sample was performed by counting the number of reads overlapping each regulatory element for each sample. DESeq2 was used to compare either the effect of genotype or butyrate treatment on the chromatin landscape, and for visualization we created a normalized chromatin accessibility matrix by normalizing for the regulatory elements’ length and GC content using the R package cqn. To assess transcription factor activity, we employed ChromVar by fixing a 500bp window on the center of the regulatory elements and using the JASPAR2016 database, computing deviations and their scores, followed by differential variability for genotype and treatment with default parameters.

54Heatmap analyses for microarray, RNA-seq and ATAC-seq datasets were performed using Multiple Experiment Viewer (MeV_4_8) software. Hierarchical clustering was applied to genes using average linking clustering with the Euclidean distance metric. All data is expressed as Mean ± SE, representative of one experiment. Total n number is indicated in the figure legend. For in vivo studies power calculations were performed on data showing mean maximum wild type arthritic knee swelling of 2mm with a standard deviation of 0.39 mm, and an expected test group arthritic knee swelling of 1.4mm. Group sizes of 3 mice or above were sufficient to reach a statistical power of at least 80%. Mice were assigned at random to treatment groups for all mouse studies, and where possible mixed among cages. Clinical and histological scoring was performed in a blinded fashion. Mice that developed adverse reactions to protocols were excluded from datasets. Statistical significance was determined: using unpaired t tests (comparison of two groups), using Mann-Whitney tests (comparison of two groups, non-parametric data), Spearman’s correlation, one-way ANOVA (comparison of three or more groups) or two-way ANOVA (murine clinical data only). All data met the assumption of statistical tests and variance was similar between groups that were statistically compared and were corrected for multiple comparisons or differences in variance. Results were considered significant at p ≤ 0.05. Statistical tests were carried out using GraphPad Prism (la Jolla, USA) v. 400 Software for Apple Mac.

55The RNA-seq and ATAC-seq datasets generated during this study are available at ArrayExpress (accession numbers E-MTAB-7345; E-MTAB-7525). The 16S datasets generated during this study are available at NCBI Sequence Read Archive (accession number PRJNA603680).