Gut Commensal Segmented Filamentous Bacteria Fine-Tune T Follicular Regulatory Cells to Modify the Severity of Systemic Autoimmune Arthritis

Mice

K/BxN mice were generated by crossing KRN TCR transgenic mice on the C57BL/6 (B6) background with NOD mice. BxN mice were generated by crossing B6 mice with NOD mice. Nur77 deficient mice (B6;129S2-Nr4a1^tm1Jmi/J) were obtained from JAX and crossed to KRN mice to generate Nur77^+/−.KRN mice. TCRα^−/−.BxN mice were generated by crossing TCRα^−/−.B6 with TCRα^−/−.NOD mice as previously described (31). KRN.FoxP3Cre.Bcl-6^fl/fl (Tfr KO) mice are generated by crossing KRN mice to Foxp3Cre mice (Stock No: 016959, JAX). Then, KRN.Foxp3Cre mice are further crossed to Bcl-6^fl/fl mice (Stock No: 023727, JAX) to generate KRN.Foxp3Cre.Bcl6^fl/fl mice with deletion of Bcl-6 specifically in Foxp3⁺ T cells. Ankle thickness was measured with a caliper (J15 Blet micrometer) as described previously (40). Ankle thickness indicates the average value of two ankles from one mouse while ankle thickening indicates the delta; i.e. the ankle thickness of a later time point minus the initial baseline thickness in any individual mouse. All mice were housed at the animal facility at the University of Arizona. All experiments were conducted according to the guidelines of the Institutional Animal Care and Use Committee at the University of Arizona.

Antibodies and flow cytometry

For surface staining, fluorophore-conjugated monoclonal Abs (mAbs) specific for CD3ε(17A2), TCRβ(H57–597), CD4(GK1.5 or RM4–5), PD-1 (RMP1–30), GITR (DTA-1), and CTLA-4 (UC10–4B9) were obtained from BioLegend. The mAb recognizing CXCR5 (2G8) was from BD Pharmingen. The mAb recognizing TCR Vβ6(RR4–7) was from eBioscience. For intranuclear staining, buffers from a Foxp3 Staining Buffer Set (eBioscience) were used to stain with Abs recognizing Foxp3 (FJK-16s, eBioscience), Helios (22F6, Biolegend), and/or Nur77 (12.14, eBioscience). Cells were run on an LSRII (BD Biosciences), and analyses were performed with FlowJo (TreeStar) software.

Microbiota reconstitution and antibiotic treatment

Our mouse colony and SFB colonization was described previously (31). We matched gender and age of K/BxN mice for the experiments. Briefly, to generate SFB positive (SFB+) mice, SFB negative (SFB−) mice were weaned at 21 days old, then orally gavaged for 3 consecutive days starting at 22 days old with SFB-containing feces collected in house. The SFB− mice were ungavaged littermate controls. The SFB colonization status was examined on day 10 after SFB gavage or equivalent by SFB-specific 16S rRNA quantitative PCR as previously described (31). For experiments involved both SFB− and SFB+ K/BxN mice, mice were analyzed at 5–7 week old (2–4 weeks after gavage). For the antibiotic treatment and 2-DG experiments, as all experimental mice were inherently SFB+, we directly used littermates that were maintained as an SFB+ K/BxN colony by vertical transmission from SFB gavaged mothers, rather than individually gavaged experimental mice. For the antibiotic treatment, SFB+ K/BxN mice were treated with 0.5g/L vancomycin (Hospira) in drinking water containing sweetener (Equal, 2.5 g/300 mL) or left untreated, starting at 22 days of age for 2 weeks before analysis taken place at 5 week old (41).

2-deoxy-D-glucose and 2-NBDG administration

For 2-deoxy-D-glucose (2-DG) administration, SFB colonized K/BxN mice were treated i.p. with 2-DG (Sigma) at a concentration of 15–30 mg/mouse in PBS, every other day for 4–5 injections starting at 28 days of age before analysis took place at 5–6 weeks old. Control mice received PBS alone. To determine cellular glucose uptake, 2-(N-(7-Nitrobenz-2-oxa-1,3-diazol-4-yl)Amino)-2-Deoxyglucose (2-NBDG, Invitrogen) was administered retro-orbitally at 100μg/mouse 30 minutes prior to tissue collection.

ELISA

Anti-GPI Ab titers were measured as described (40). In brief, ELISA plates were coated with recombinant mouse GPI at 5 μg/ml, and diluted mouse sera were added. Subsequently, plates were washed and alkaline-phosphatase (AP)-conjugated anti-mouse IgG Abs were added. After the final wash, AP substrate was added and titers were quantified as optical density values via an ELISA reader. Ab titers were expressed as arbitrary units, which were calculated from serial dilutions of sample serum and defined as the reciprocal of the highest dilution that gave a background Optical Density (OD₄₀₅) value set as 0.1 (42).

Adoptive transfer

For the K/BxN transfer model, splenic and lymph node-derived CD4⁺ T cells from SFB− Nur77^+/−.KRN, KRN.FoxP3Cre.Bcl-6^fl/fl (Tfr KO), or KRN mice were enriched using CD4-conjugated MACS beads (Miltenyi) and adoptively transferred by retro-orbital (r.o.) injection into SFB− and/or SFB+ TCRα^−/−.BxN recipients as indicated in the corresponding figures. Two weeks after transfer, organs were collected for flow cytometry and sera were collected and analyzed by ELISA.

In vitro stimulation

TCR stimulation for Nur77 expression was performed by adding a final concentration of 5 μg/ml GPI (282–294) peptide (LSIALHVGFDHFE, GenScript) to splenocytes at 10⁶ cells/well in 96 well plate for a total duration of 0, 2, 4, or 6 hours at 37°C before FACS analysis.

Statistical analysis

Differences were considered significant when p < 0.05 by Student’s t test (two-tailed, unpaired) or two-way analysis of variance (ANOVA) (Prism 6, Graph-Pad Software). *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001.

The impact of autoimmunity and the gut commensal SFB on Tfr cell responses

By TCR analysis, Maceiras et al. demonstrated that Tfr cells expressed potentially autoreactive TCRs, as the TCR repertoire of Tfr cells is close to that of Treg cells (43). However, in this study, no direct evidence was shown to demonstrate that Tfr cells recognize self-Ag. Moreover, we have previously demonstrated that SFB enhance the Tfh cell response in both systemic (spleen) and gut-associated lymphoid tissue (Peyer’s patches/PPs) of K/BxN mice (31). We are thus interested in examining whether SFB also impact self-reactive Tfr cells to indirectly affect the Tfh cell response. We used the autoimmune TCR transgenic strain K/BxN and its non-transgenic background B6xNOD (BxN) strain in both SFB− and SFB+ conditions to ask how the SFB and self-Ag may impact the Tfr cell responses. We observed that the percentage of splenic Tfr cells was highly upregulated in K/BxN mice compared to BxN mice regardless of SFB status (Fig 1A, gating scheme and representative data depicted in Suppl. Fig. 1A and 1B). While this may be due to a lack of active GCs in the non-autoimmune BxN mice, this phenotype still holds in Peyer’s Patches (PPs), a site with active GCs. Interestingly in both BxN and K/BxN mice, SFB reduced the percentage of Tfr in total CD4⁺ T cells from PPs but not spleen. Based on our previous study, this is likely a secondary effect due to the dramatic increase of Tfh cells in PPs which reduces the percentage of non-Tfh cell types (31). Using Treg depletion and adoptive transfer, it was reported that Tfr cells are mostly derived from Tregs (18, 44, 45). Therefore, we examined the possibility that SFB may alter the proportion of Tfr relative to total Foxp3⁺ Tregs. We found that SFB did not alter the proportion of Tfr cells in the total CD4⁺Foxp3⁺ T cell population (Fig. 1B). This was true regardless of autoimmune condition or tissue site as we found SFB did not alter Tfr/CD4⁺Foxp3⁺ T cell ratio in either spleen and PPs of BxN and K/BxN mice. However, there was a much higher Tfr/CD4⁺Foxp3⁺ ratio in autoimmune K/BxN than control BxN strains in both spleen and PPs. In addition to the percentage data, we also show the cell number data for CD4⁺ T cell, Tfr, Tfh and CXCR5^loPD-1^lo Treg (Suppl. Fig. 1C and 1D). However, many reports have shown that it is not the absolute numbers of Tfr or Tfh cells and rather it is the ratio of Tfr/Tfh cells dictates the magnitude of Ab responses with higher Tfr/Tfh ratio correlated with lower Ab responses. Therefore, we also calculate the ratio of Tfr/Tfh and found that SFB reduced the Tfr/Tfh ratio in spleen, and this reduction was even more obvious in PP (Fig. 1C) (46–48). Foxp3 levels of the cells in Tfrs and CXCR5^loPD-1^lo Tregs were also examined and no difference was observed between SFB− and SFB+ groups in both cell types in the K/BxN mice (Suppl. Fig. 2A).

The use of Helios as a marker to define thymus-derived Tregs has been controversial (49–51). Nevertheless, Helios has been shown to regulate effector Treg fitness and to be required by Tfr cells to control Tfh cell responses in an immunization study (52). Therefore, we examined Helios expression in the Tfr cells of autoimmune K/BxN mice. As expected, Helios was highly expressed (~80%) in splenic Tfr cells similar to the level observed in CXCR5^loPD-1^lo Treg cells (Fig. 1D and Suppl. Fig. 2B). However, unlike CXCR5^loPD-1^lo Tregs whose expression of Helios was reduced to between 40–50% in PPs, the expression of Helios on Tfr cells remained relatively high (~75%) in PPs. Finally, SFB did not impact the expression of Helios except in the PP CXCR5^loPD-1^lo Treg group. Interestingly, there was an increase of Helios⁺ cell numbers in Tfrs and CXCR5^loPD-1^lo Tregs in spleen and PPs of SFB+ compared to SFB− K/BxN mice (Suppl. Fig. 2C). The auto-Ab isotype in the K/BxN model is IgG1 (31). Previously, we have demonstrated that spleen, but not PPs, is the major producing site of anti-GPI IgG1 Abs and that SFB-induced PP Tfh cells did not induce anti-GPI of the IgA isotype (31). Here, we also measured serum anti-GPI in IgG1 and IgA isotype and plotted these against ankle thickness (Fig 1E). There was a positive correlation between anti-GPI IgG1 auto-Ab and ankle thickness but the IgA titers were undetectable even in SFB+ group. Hence, despite the fact that SFB generally induce stronger immunophenotypes in PPs, we focused on spleen as the target organ for the Tfr cell studies for the rest of this paper.

Surface CTLA-4 expression on systemic Tfr cells can be modulated by gut microbiota

Cytotoxic T lymphocyte-associated protein 4 (CTLA-4) is a co-inhibitory molecule expressed by T cells (53). It has previously been shown that Tfr cells potently suppress GC B cell responses through CTLA-4 (33). We therefore investigated whether gut SFB may modulate systemic Tfr cell function by altering the expression of CTLA-4 on systemic Tfr cells. It is known that most CTLA-4 is localized in intracellular compartments (54). We have chosen to focus on surface CTLA-4 expression because it is the optimal expression of surface CTLA-4 that is crucial for the balance of maintaining immunotolerance and preventing autoimmunity (55). For example, one interesting study shows that surface CTLA-4 can remove CD80 and CD86 from dendritic cells. After removal, these costimulatory ligands are degraded inside CTLA-4 expressing cells, resulting in impaired costimulation via CD28 (56). Throughout this paper, the CTLA-4 examined are surface CTLA-4 except for Suppl. Fig 4A and 4B which show intracellular CTLA-4 expression. Notably, we found that gut SFB reduced CTLA-4 expression on systemic (splenic) Tfr cells (Fig. 2A). Interestingly, this effect was more significant in Tfr cells than CXCR5^loPD-1^lo Tregs. The SFB-induced CTLA-4 reduction on Tfr cells was associated with an increase of arthritis development (Fig 2B). To confirm our finding using a different approach, we treated SFB+ mice with vancomycin, an antibiotic targeting gram-positive bacteria such as SFB. Vancomycin is poorly absorbed from the gut (57), thus its effect when delivered by the oral route is mainly limited to the gastrointestinal tract. Indeed, SFB depletion by vancomycin rescued CTLA-4 expression on Tfr cells (Fig. 2C) and reduced arthritis severity (Fig. 2D). CTLA-4 is highly expressed on Tfr cells, and this expression is required for Tfr cells to negatively regulate humoral immune responses (33, 34). In the K/BxN mice, higher CTLA-4 expression on Tfr cells also strongly correlated with reduced arthritis in the gavaging (Fig. 2E) and antibiotic treatment (Fig. 2F) experiments.

SFB reduces CTLA-4 surface expression in Tfr cells by promoting a glycolytic metabolic program

Recently, metabolic status in T cells has been shown to play a key role in their function (58, 59). Additionally, the germinal center is known for its hypoxic environment, which promotes glycolysis (60). However, little is known regarding whether glycolysis will impact effector molecules such as CTLA-4 on Treg or Tfr cells. As SFB is known to induce a prominent GC response in both spleen and PPs (31, 61), we tested whether the Tfr cells in SFB+ mice shift their metabolism to a pro-glycolytic program. 2-DG is a glucose analog; however, unlike glucose, 2-DG cannot be fully metabolized, and as a result 2-DG acts to inhibit glycolysis (62–64). Fluorescently-labeled forms of 2-DG such as 2-NBDG can accumulate in cells and serve as a good marker for tissue glucose uptake and glycolytic activity (62–64). To assess acute glucose uptake, we injected SFB− and SFB+ K/BxN mice with 2-NBDG 30 minutes prior to tissue collection. As 2-NBDG staining requires intact cells, we used GITR, a Treg marker, as the surface marker to identify Foxp3⁺ cells, and confirmed that GITR⁺ cells were 92–96% Foxp3⁺ in the Tfr and CXCR5^loPD-1^lo Treg cell populations (Suppl. Fig. 3A). Our data showed that, compared to CXCR5^loPD-1^lo Tregs, Tfr indeed displayed an increase in 2-NBDG uptake in both SFB− and SFB+ groups, suggesting that Tfr cells are more glycolytic than CXCR5^loPD-1^lo Tregs (Fig. 3A). Interestingly, SFB colonization further enhanced 2-NBDG uptake in Tfr cells, suggesting that Tfr in SFB+ mice have further shifted their metabolism to a pro-glycolytic program.

Given the prior observation that SFB reduced CTLA-4 expression on Tfr cells (Fig. 2A), we hypothesized that the SFB-mediated increase in glycolytic metabolism may contribute to reduced CTLA-4 expression in the Tfr population. To examine the effect of glycolysis on Tfr cell CTLA-4 levels, we treated SFB+ K/BxN mice with 2-DG in vivo to inhibit glycolysis. Our data showed that there was a significant reduction of arthritis and anti-GPI titers in 2-DG treated compared to PBS treated SFB+ K/BxN mice (Fig. 3B and 3C). However, systemic inhibition of glycolysis can impact many cell types, and we cannot therefore directly correlate the reduction in arthritis and autoantibodies to Tfr cells. In this regard, we found that there was no significant difference in lymphocyte numbers, including both Tfr and CXCR5^loPD-1^lo Tregs, and there was some reduction in non-lymphocytes in 2-DG treated mice compared to PBS treated controls (Suppl. Fig. 3B and 3C). To directly examine whether the glycolytic inhibition may impact the effctor molecules on Tfr cells, we examined the CTLA-4 expression and found that inhibition of glycolysis by 2-DG caused an increase of CTLA-4 expression on Tfr cells(Fig. 3D). These results, together with data from Fig 2 suggest that SFB colonization, reduced the CTLA-4 expression on Tfr cells by promoting their glycolytic activity.

SFB decrease T cell receptor (TCR) signal strength indicated by the increase of Nur77^lo Tfr cells

Next, we searched for early events during T cell activation that would explain the observed CTLA-4 reduction on Tfr cells in SFB colonized mice. It has been reported that CTLA-4 expression on the T cell surface is proportional to TCR signal strength (39). Therefore, we tested whether the SFB-dependent CTLA-4 reduction on Tfr cells is due to reduced TCR signaling in SFB+ compared to SFB− Tfr cells. Nur77 (a.k.a. Nr4a1) is a transcription factor whose expression is induced rapidly upon T cell activation, in a manner representative of TCR signaling intensity (65). Thus, Nur77 reporter mice, as well as direct detection of endogenous Nur77 using antibodies, have been used as methods to evaluate TCR signal strength (65, 66). To our surprise, TCR signal strength in Tfr cells was increased relative to CXCR5^loPD-1^lo Tregs ex vivo (the levels of Nur77 were measured in freshly isolated T cells without further TCR stimulation in vitro) (Fig. 4A). Importantly, there was a significant increase of Nur77^lo cells in the SFB+ group of both splenic Tfr and CXCR5^loPD-1^lo Treg cell populations. We found there was a similar trend in PPs, though the significance in Tfr cell remained and the difference between SFB− and SFB+ groups in CXCR5^loPD-1^lo Treg cells became non significant, suggesting this SFB effect is more pronounced in Tfr cells (Suppl. Fig. 3D). Moreover, we have shown that, similar to the study by Egen et al. showing CTLA-4 surface expression is proportional to the TCR signal strength, there was a strong positive correlation of surface CTLA-4 with Nur77 in autoimmune K/BxN mice (Fig. 4B) (39). However, when we performed the same experiments by staining intracellular CTLA-4, there was actually a negative correlation between intracellular CTLA-4 and Nur77, and SFB further increased the retention of CTLA-4 within intracellular compartments (Suppl. Fig. 4A and 4B). Thus, SFB reduce surface CTLA-4, and increase intracellular CTLA-4 expression. These data suggest that SFB may condition Tfr cells to express a reduced level of CTLA-4 through modifying TCR signaling strength.

K/BxN mice express a transgenic TCR, and therefore all T cell populations express the same TCR toward self-antigens. Accordingly,we hypothesized that the mechanism controlling Nur77 levels in these cell types is the relative amount of TCR expressed on the cell surface. To quantify the TCR level, we stained T cells with antibodies against TCR Vβ6, the β chain for the KRN TCR transgene. We found that Tfr cells express the highest TCR Vβ6 level, followed by Tregs, and conventional T cells (Tconv) with the least TCR expression (Fig. 4C). Indeed, our data demonstrated that the TCR Vβ6 levels on Tconv, Treg and Tfr cells positively correlated with these cells’ Nur77 level in both SFB− and SFB+ hosts (Fig. 4D). However, between SFB− and SFB+ groups within the same cell type, we did not observe any reduction of TCR Vβ6 levels by SFB (Fig 4B). Since Nur77 expression was reduced by SFB in these populations, these data raise the question of how SFB reduces TCR signal strength independently of TCR expression.

Nur77 haplodeficiency increases glycolysis and decreases CTLA-4 expression on Tfr cells, enhancing arthritis development

After its rapid induction post-TCR activation, Nur77 gets down-regulated after just a few hours (66, 67). We therefore wondered if SFB colonization promotes faster Nur77 degradation after TCR stimulation, leading to reduced Nur77 detection ex vivo seen in Fig. 4A. To test this, we examined the temporal characteristics of endogenous Nur77 protein expression in GPI peptide stimulated T cells (containing Tconv, Treg and Tfr cells). In Tfr cells, Nur77 protein levels increased similarly between the SFB− and SFB+ groups up to 4 hours after stimulation with GPI peptide (Fig. 5A). However, there was a small but significant reduction of Nur77 in SFB+ Tfr cells by the 6 hour time point. There is no difference in cell numbers of Tfr cells at 6 hrs compared to 0 hrs (Suppl. Fig. 5A). No significant difference was observed in Tconv cells and CXCR5^loPD-1^lo Treg cells (Fig. 5A). Overall, these data suggest that SFB colonization facilitates the decline of Nur77 levels (Fig 4A and 5A).

Next we addressed whether SFB’s reduction of Nur77 in Tfr and CXCR5^loPD-1^lo Tregs would explain the reduction of CTLA-4 in these populations. To mimic the Nur77 reduction by SFB, we transferred control KRN T cells or Nur77^+/−.KRN T cells that that were heterozygous for deletion of Nur77 into SFB− TCRα^−/−.BxN hosts (T cell deficient mice on a B6xNOD background and thus, bearing NOD MHC class II, I-A^g7, for self-Ag presentation). Notably, this K/BxN T cell transfer model allows for manipulation of the gene of interest specifically in T cells, and host mice spontaneously develop disease symptoms after T cell transfer (35, 36). The haplodeficiency of Nur77 in Nur77^+/−.KRN T cells reduce Nur77 gMFI level to around 50% when compared to WT KRN T cells in both Tfr and CXCR5^loPD-1^lo Treg groups (Suppl. Fig. 5B). When KRN T cells were transferred into SFB− TCRα^−/−.BxN recipients, only minimal ankle thickening was observed (Fig. 5B). Importantly, when Nur77^+/−.KRN T cells were transferred, arthritis and auto-Ab development was enhanced. Additionally, CTLA-4 expression was reduced in both CXCR5^loPD-1^lo Treg and Tfr cells in the Nur77^+/− deficient background compared to WT (Fig. 5C). Therefore, this suggests that Nur77 levels act upstream of CTLA-4 expression in both CXCR5^loPD-1^lo Tregs and Tfr cells, and that the sped-up degradation of Nur77 in SFB+ Tregs may underlie their reduced CTLA-4 expression, leading to an exacerbated disease state.

Similar to Fig 2E and 2F, in the Nur77^+/−.KRN T cell transfer system, we also found a negative correlation of Tfr’s CTLA-4 expression and arthritis severity, and the reduction of CTLA-4 in Nur77^+/−.KRN Tfr cells is correlated with a stronger arthritis development (Fig. 6A). Additionally, there is also a negative correlation between surface CTLA-4 expression of Tfr versus anti-GPI titers (Fig. 6B). Nur77 expression regulates metabolic activity of T cells including glycolysis (68). As inhibition of glycolysis increases CTLA-4 expression levels in Tfr (Fig. 3D), and Nur77 deficiency results in the reduction of CTLA-4 expression, we investigated whether Nur77 deficiency may increase glycolytic activity that could explain the reduction of CTLA-4 in Nur77 deficient T cells. Our data demonstrated an increase of 2-NBDG⁺ Tfr cells in Nur77 deficient mice compared to WT controls (Fig. 6C). This data is inconsistent with a previous report showing that the general CD4+ T cell population in Nur77 KO mice display increased glycolytic activity (68). It is interesting that we found a similar result with only Nur77^+/− haplodeficiency and in Tfr cells, a cell type derived from Treg, which are traditionally viewed as cell type relies more on oxidavtive phosporylation than the glycolysis pathway.

Tfr deficiency enhances autoimmune arthritis and autoantibody development dependent on microbiota status

As mentioned earlier, many studies support the hypothesis that Tfr cells are a negative regulator of humoral immunity that act by restraining excessive Tfh and GC B cell responses (17–19, 45). In support of this idea, the loss of CTLA-4 on Tfr cells resulted in defective suppression of antigen-specific antibody responses (34). However, recent studies have challenged the suppressor model of Tfr cell function. In two of these studies, which were conducted in murine immunization and infection models, Tfr cells were found to promote GC B cell responses and Tfr deficiency actually resulted in reduced IgG Ab titers (24, 69). We have thus far demonstrated that alteration of CTLA-4 expression on Tfr cells by SFB, antibiotics, and Nur77 deficiency generate a negative correlation between Tfr CTLA-4 levels and arthritis severity, which is suggestive of a negative regulatory role for Tfr cells in autoimmune K/BxN mice (Figs. 2E, 2F and 6A). Finally, we would like to definitively determine the positive or negative effect of Tfr cells on autoimmune arthritis and auto-Ab responses in K/BxN mice, and how SFB may further modulate this positive or negative effect. To address these questions, we first crossed KRN to Foxp3Cre and Bcl-6^fl/fl mice to generate KRN.Foxp3Cre.Bcl-6l^fl/fl mice that are deficient in Tfr cells (to simplify, referred to as Tfr KO). Similar to many reported cell-targeted deletions, the Tfr in our Tfr KO model is not depleted by 100% (only ~75% reduction) (Suppl. Fig. 6). Nevertheless, as we had achieved a large reduction of Tfr cells, we tested our hypothesis by transferring CD4⁺ T cells from Tfr KO or KRN mice into SFB− and/or + TCRα^−/−.BxN recipients. Our results showed an increase in arthritis and autoantibodies in SFB− mice receiving Tfr KO T cells compared to mice receiving control KRN T cells (Fig. 7A and 7B). However, we did not observe differences between the WT and Tfr KO groups in SFB+ mice. Interestingly, we also observed a significant induction of Tfh cells in the hosts receiving Tfr KO T cells compared those receiving WT KRN T cells under SFB− but not SFB+ conditions (Fig. 7C). This indicates that, under SFB− conditions, Tfr cells act to restrain the Tfh response, thereby inhibiting the autoimmune state and that Tfr cells are unable to assert their repressive function under SFB+ conditions. Overall, our data show that Tfr-mediated control over autoimmune arthritis and autoantibodies is operated in a microbiota dependent manner.