Foxp3⁺ follicular regulatory T cells control T follicular helper cells and the germinal center response

Abstract

Follicular helper (T_FH) cells provide crucial signals to germinal center B cells undergoing somatic hypermutation and selection that results in affinity maturation. Tight control of T_FH numbers maintains self-tolerance. We describe a population of Foxp3⁺Blimp-1⁺CD4⁺ T cells constituting 10-25% of the CXCR5^highPD-1^highCD4⁺ T cells found in germinal center after immunization. These follicular regulatory T cells (T_FR) share phenotypic characteristics with T_FH and conventional Foxp3⁺ regulatory T cells (T_reg) yet are distinct from either. Similar to T_FH cells, T_FR development depends on Bcl-6, SAP, CD28 and B cells; however T_FR originate from thymic-derived Foxp3⁺ precursors, not naïve or T_FH cells. T_FR are suppressive in vitro and limit T_FH and germinal center B cell numbers in vivo. In the absence of T_FR, an outgrowth of non-antigen-specific B cells in germinal centers leads to fewer antigen-specific cells. Thus, T_reg cells use the T_FH differentiation pathway to produce specialized suppressor cells that control the germinal center response.

Germinal centers are clusters of rapidly-dividing B cells formed in secondary lymphoid tissues in response to T-dependent antigens. Within germinal centers, mutation of the B cell receptor V-region genes together with subsequent selection results in the production of high affinity plasma cells and memory B cells¹. Defective selection can result in the production of autoantibodies and a break in self-tolerance^{2, 3}. Germinal center B cell selection can in part be mediated by a specialized helper T cell subset, CXCR5^highPD-1^high T follicular helper (T_FH) cells⁴. T_FH cells develop in a Bcl-6-dependent manner and provide germinal center B cells with survival and selection signals. Limiting the numbers of T_FH cells within germinal centers has been shown to be critical to prevent the emergence of autoantibodies^5,6. Little is known about T_FH control; in mice, Qa-1-restricted CD8 T cells can regulate the T_FH compartment⁷ and in humans, CD4⁺CD25⁺CD69⁻ T cells with a suppressive function in vitro have been found in germinal centers^{8, 9}. Regulatory T Cells (T_reg) have also been shown enter the primary B cell follicle in mice, but their phenotype, ontogeny and ability to control T_FH cells remain unknown¹⁰.

T_reg that develop in a Foxp3-dependent manner repress the growth and function of CD4⁺ effector T cells. Humans and mice lacking Foxp3 cannot form T_reg and develop fatal autoimmunity^11-15. In order to repress T_H1-, T_H2- and T_H17- mediated immune responses, T_reg have been shown to co-opt selective aspects of the differentiation programs required for these T_H subsets: Tbet/Stat1, IRF-4 or Rorγt signaling respectively^16-18. Here we show that Foxp3⁺ T_reg can be diverted to become T_FH repressors via expression of Bcl6 and SAP-mediated interaction with B cells. The resulting follicular regulatory T cells (T_FR) share features of both T_FH and T_reg cells, localize to germinal centers, and regulate the size of the T_FH cell population and germinal centers in vivo.

Foxp3⁺ follicular regulatory T cells are distinct from T_FH and T_reg

After immunization with a T-dependent antigen we observed that ~10-25% of CD4⁺CXCR5^highPD-1^high ‘T_FH’ cells expressed the transcriptional regulator of the T_reg lineage, Foxp3 (Figure 1a). These cells followed the same formation and resolution kinetics as conventional T_FH cells (Figure 1b). Foxp3⁺ cells could be visualized within germinal centers identified by immunofluorescence staining of frozen spleen sections from immunized mice (Fig. 1c); 17% ± 8% of germinal center CD3⁺ cells also expressed Foxp3.

Figure 1. A proportion of CXCR5^highPD-1^highCD4⁺ cells express the transcription factor Foxp3.

(a,b) After SRBC immunization Foxp3⁺ cells were identified in the CXCR5^highPD-1^highCD4⁺ ‘T_FH’ compartment, these cells follow the same kinetics as classic T_FH cells. (c) Foxp3⁺ cells (red) are present within the Bcl6⁺ germinal center area (green) following SRBC immunization. Scale bar represents 100μm. (d) Heat map comparing the gene expression profiles of different CD4⁺ T cell subsets from Foxp3^GFP mice seven days after immunization. Red: high gene expression; blue: low gene expression. The cells were sorted using the following markers and for simplicity will be referred by the abbreviations in parentheses throughout: CD4⁺CD44^lowFoxp3⁻ naïve (T_N) cells, CD4⁺CD44^highCXCR5^int/lowPD-1^int/lowFoxp3⁻ effector/memory (T_EM) cells, CD4⁺CD44^intCXCR5^int/lowPD-1^int/lowFoxp3⁺ regulatory T cells (T_reg), CD4⁺CXCR5^highPD-1^highFoxp3⁻ T follicular helper (T_FH) cells and CD4⁺CXCR5^highPD-1^highFoxp3⁺ follicular regulatory (T_FR) cells. (e) Il21 and Il4 mRNA measured by quantitative PCR from sorted cells using the strategy described in (d) normalized to Gapdh. Heights of the bars represent the mean and error bars represent the range of expression from 3 biological replicates. nd: gene expression not detected. (f) Left: Intracellular expression of CD40L as determined by flow cytometry in T_reg (blue), T_FH (green) and T_FR (red) cell populations; the grey histogram represents a staining control from an immunized CD40L-deficient mouse. Right: Cxcl13 mRNA measured by quantitative RT-PCR as described in (e). (g) Cell surface expression of GITR, CD25 and intracellular CTLA4 in T_reg (blue), T_FH (green) and T_FR (red) cell populations; grey histograms represent the isotype control. (h) Relative Gzmb and Gzma mRNA determined by quantitative RT-PCR as described in (e). (i) Percentage of CD103⁺ cells within the T_reg, T_FH & T_FR populations, each symbol represents one mouse. (j) Left: Cell surface expression of ICOS as determined by flow cytometry in T_reg (blue), T_FH (green) and T_FR (red) cell populations; the grey histogram represents staining level of an isotype control. Right: Il10 mRNA detected by quantitative RT-PCR of as described in (e). Flow cytometric and RT-PCR data are representative of at least three independent experiments. In (e)-(i): Statistical significance was determined using a one-way ANOVA analysis with Bonferroni’s multiple testing correction; * P<0.05; **P<0.01; ***P<0.001.

To obtain information about the identity and function of CD4⁺CXCR5^highPD-1^high Foxp3⁺ cells - designated T_FR, we performed microarray expression profiling on sorted populations from Foxp3^GFP mice¹⁹ seven days after SRBC immunization. T_reg, T_FH, non-T_FH effector/memory cells (T_EM) and naïve (T_N) T cells were also included (sorting strategy is depicted in Supplementary Fig. 1). T_FR more closely resembled T_reg than T_FH, T_EM or T_N (Fig. 1d and Supplementary Table 1), with elevated expression of many T_reg associated genes including Foxp3, Ctla4, Gitr, Klrg1 and Prdm1. Nevertheless, T_FR also expressed high amounts of the prototypic T_FH genes Cxcr5, Pdcd1, Bcl6, Cxcl13, and Icos. T_FR did not express the helper cytokines IL-21 or IL-4 (Fig. 1e) or the costimulatory ligand CD40L (Fig. 1f), but expressed comparably high levels of the ligand for CXCR5, CXCL13²⁰, as T_FH cells (Figure 1f). Differential expression of T_FH or T_reg associated molecules was confirmed by flow cytometry and/or real time PCR (Fig. 1g-i and Supplementary Fig. 2).

T_FR expressed numerous molecules characteristically expressed by T_reg, such as GITR and CTLA-4, but at higher levels than T_reg, consistent with an activated T_reg phenotype²¹ (Figure 1g). Gzma expression was comparable to T_reg (Figure 1h) but Gzmb, a reported target of Bcl-6 repression²² was barely detectable. T_FR also expressed high levels of Il10 mRNA and surface ICOS protein, which are common to both T_FH and T_reg cells (Figure 1j and Supplementary Fig. 2). The elevated levels of GITR, IL-10 and ICOS on T_FR compared with the rest of the T_reg pool is consistent with an effector T_reg phenotype²³, which suggests T_FR have a regulatory function. The phenotypic features shared by T_FH cells and T_FR may account for their common germinal center localization.

T_FR and T_FH cells require similar differentiation cues for their formation and maintenance

As both T_FH cells and T_FR co-localize in germinal centers, we sought to determine whether T_FR formation was dependent on similar developmental cues. T cell priming through CD28 is one of the first signals required for T_FH development^{24, 25}. Enumeration of T_FH and T_FR in mixed Cd28^−/− CD45.2: CD45.1 Cd28^+/+ bone marrow chimeras immunized seven days previously with SRBC revealed a complete absence of both T_FH and T_FR cells in the absence of CD28 signaling (Fig. 2a-c). Consistent with previous reports^{26, 27}, CD28-deficiency moderately reduced peripheral T_reg numbers (Supplementary Fig. 3a).

Figure 2. T_FR require the same differentiation cues as T_FH cells for their development.

Flow cytometric contour plots (a, d, g) and dot plots of T_FH (b, e, h) and T_FR (c, f, i) cells in the groups of mice described below, seven days after SRBC immunization. (a-c) Mixed bone marrow chimeras generated by sub-lethally irradiating Rag2^−/− mice and reconstituting their immune system with a 1:1 ratio of bone marrow cells from CD45.1 Cd28^+/+ and CD45.2 Cd28^−/− mice or control CD45.1 Cd28^+/+ and CD45.2 Cd28^+/+. (d-f) C57BL/6 (BL/6) and B-cell deficient μMT mice. (g-i) Sh2d1a^+/+ and Sh2d1a^−/− mice. Each symbol represents one mouse and horizontal bars represent median values. Figures represent one of 3 independent experiments with similar results. Statistical significance was determined using a Mann-Whitney Test: *P<0.05, **P<0.01.

SAP-dependent interactions of T_FH precursors with B cells are required for T_FH formation and/or maintenance^{5, 28-30}. We therefore investigated whether interactions with B cells and/or SAP-mediated signals are essential for T_FR formation. Neither T_FH nor T_FR cells formed after SRBC immunization of B cell deficient μMT mice (Fig. 2d-f) whereas T_reg formed normally (Supplementary Fig. 3b). In addition, immunization of Sap-deficient (Sh2d1a^−/−) mice revealed that, similar to T_FH cells (Figure 3g, h), T_FR cells are dependent on SAP for their formation (Figure 3i). T_reg were only slightly reduced in the absence of SAP (Supplementary Fig. 3c). These data demonstrate that the developmental requirements of T_FR are similar to those that govern T_FH formation and dispensable for the generation of T_reg.

Figure 3. T_FR cells express Bcl-6 and Blimp-1.

(a) Bcl6 and Prdm1 mRNA normalized to Gapdh determined by quantitative RT-PCR from sorted cells using the strategy described in Fig. 1d and Supplementary Fig. 1. Heights of the bars represent the mean and error bars represent the range of expression from 3 biological replicates. Statistical significance was determined using a one-way ANOVA analysis with Bonferroni’s multiple testing correction; * P<0.05; **P<0.01; ***P<0.001. Bar graphs are representative of 3 experiments. (b) Immunofluorescence of frozen spleen sections from mice immunized seven days previously with SRBC. The germinal center is demarcated by the white dotted line in the three consecutive sections. Upper panel: AID (red) and CD3 (green); middle panel: Foxp3 (red) and Bcl-6 (green); lower panel: Foxp3 (red) and Blimp1 (green). Scale bar represents 100μm. (c) Flow cytometric contour plots of T_FH (upper panels) & T_FR (lower panels) formation in the draining (mediastinal) lymph node ten days after intranasal influenza infection of mixed fetal liver chimeras reconstituted with a 1:1 ratio of fetal liver cells from E14.5 CD45.2 Prdm1^gfp/gfp : CD45.1 Prdm1^+/+ embryos, E14.5 CD45.2 Bcl6^−/− : CD45.1 Bcl6^+/+ embryos or control E14.5 CD45.2 Prdm1^gfp/+ : CD45.1 Prdm1^+/+ embryos.

Coordinated Bcl-6 and Blimp-1 expression in T_FR cells

Bcl-6, the transcriptional regulator of the T_FH subset, regulates key molecules required for follicular localization and function^31-33 in a process thought to be counteracted by the transcriptional repressor Blimp-1³¹; Bcl-6 and Blimp-1 also mutually repress each other during B cell differentiation^{34, 35}. We asked whether, similar to T_FH cells, T_FR cells also expressed Bcl-6 and would be devoid of Blimp-1 expression. Quantitative RT-PCR revealed that Bcl6 was expressed in T_FR cells (Fig. 3a). Of note, T_FR co-expressed Prdm1, the gene encoding Blimp-1, and its expression on T_FR was higher than in any other CD4 T cell subset (Fig. 3a). Expression of Bcl-6 and Blimp-1 protein in Foxp3⁺ cells within germinal center identified was also confirmed by immunofluorescence staining of spleen sections from SRBC-immunized mice. All Foxp3⁺ T_Fr within AID⁺ germinal center expressed Bcl-6, albeit at low levels (Fig. 3b) and 75% stained positive for Blimp-1 7 days after immunization; this proportion was reduced to 50% by day 14.

Bcl-6 is required for T_FR formation and Blimp-1 regulates T_FR homeostasis

To determine whether Blimp-1 and/or Bcl-6 play a role in T_FR cell formation or homeostasis, we reconstituted sub-lethally irradiated CD45.1 mice with a 1:1 ratio of fetal liver cells from congenically-marked Prdm1^+/+ and Prdm1^gfp/gfp embryos, Bcl6^−/− and Bcl6^+/+ embryos or control Prdm1^+/+ and Prdm1^gfp/+ embryos. Eight weeks after reconstitution the mice were infected intranasally with influenza virus (HKx31), and 10 days later T_FR formation was assessed in the mediastinal lymph node. In contrast to published data³¹, loss of Blimp-1 did not alter the proportion of T_FH cells but caused T_FR to double (Fig. 3c), suggesting that Blimp-1 limits the size of the T_FR population. This is consistent with a recent report showing Blimp-1 limits the numbers of effector T_reg through a Bcl-2-dependent mechanism²³.

Mixed Bcl6^−/−: Bcl6^+/+ chimeras confirmed previous reports that T_FH cells do not form in the absence of Bcl-6 (Fig. 3c, upper panel). Cells lacking Bcl-6 expression did not give rise to T_FR cells despite the presence of germinal centers in the mice (Fig. 3c, lower panel). Similar results demonstrating the requirement for Bcl-6 in splenic T_FR generation were obtained after SRBC immunization (Supplementary fig. 4a-c). As reported previously³³, Bcl-6 was dispensable for T_reg formation (Supplementary fig. 4d). Together, this suggests that Bcl-6 is essential for T_FR formation and Blimp-1 expression regulates the size of the T_FR population.

Although co-expression of Bcl-6 and Blimp-1 seems paradoxical, there are precedents in which both Prdm1 and Bcl6 are co-regulated; for example in both effector and memory CD8⁺ subsets³⁶. Blimp-1 has been recently shown to influence T_reg function inducing an effector phenotype^23,37. It is expressed by T_reg at mucosal sites and by a small (8-12%) subset of splenic T_reg, which produce IL-10 in a Blimp-1-dependent manner³⁷. Blimp-1⁺ T_reg and T_FR cells also share expression of high amounts of IL-10, GITR and ICOS²³. T_FR are thus likely to be the follicular counterparts of the Blimp-1⁺ IL-10⁺ effector T_reg found at mucosal surfaces.

T_FR cells derive from T_reg precursors

The observation that T_FR require the same cues as T_FH for their differentiation raised a critical question: do T_FR represent induced T_reg that arise from T_FH cells that switch on Foxp3 in the germinal center, or do they derive from Foxp3⁺ T_reg? Plasticity of CD4 helper T cell subsets is well documented³⁸, as is the adoption of T_H transcriptional programs by T_reg^16-18.

In order to test whether T_FR derive from T_FH cells that turn on Foxp3 expression we transferred 1×10⁵ naïve cells (CD4⁺CD44^lowCD25⁻) from CD45.1 mice expressing the 3A9 TCR transgene (TCR^HEL), which recognizes hen egg lysozyme (HEL) peptide presented by I-A^k, into congenic CD45.2 B10.BR mice. Seven days after immunization with HEL in alum, 6-10% of T_FH cells derived from the donor HEL-TCR T cells, but no donor-origin T_FR could be identified; all T_FR derived exclusively from the recipients’ cells (Fig. 4a). Between 1-2% of the transferred naïve TCR^HEL donor population developed into inducible T_reg (Fig. 4b). To exclude that an idiosyncratic effect of the 3A9 transgene had precluded development of T_FR cells, this experiment was repeated with OT-II transgenic T cells: Again, OT-II T cells could form T_FH but not T_FR (Fig. 4c, d), suggesting that T_FR do not derive from T_FH cells.

Figure 4. T_FR derive from Foxp3⁺ precursors.

Flow cytometric contour plots of splenic CD4⁺CXCR5^highPD-1^high cells (a) or CD4⁺ cells (b) seven days after 1×10⁵ transferred transgenic TCR^3A9 HEL-specific CD45.1 T cells were adoptively transferred into congenically distinct CD45.2 B10.Br mice and immunized with HEL in alum. Flow cytometric contour plots of splenic CD4⁺CXCR5^highPD-1^high cells (c) or CD4⁺ cells (d) seven days after adoptive transfer of 1×10⁵ OT-II OVA-specific Thy1.2 T cells into congenically distinct Thy1.1 C57BL6 mice and immunization with OVA in alum. (e) Flow cytometric contour plots of splenic CD4⁺ T cells from CD45.1 C57BL/6 mice seven days after adoptive transfer of 1×10⁶ sorted naïve CD4⁺CD44^intFoxp3⁺ T_reg (top panel) or CD4⁺CD44^lowFoxp3⁻ naïve T cells (lower panel) from unimmunized CD45.2 Foxp3^GFP mice and KLH in Ribi immunization. Transferred CD45.2 cells are shown in red, the endogenous CD45.1 cells are represented by the grey contour plots. Histograms showing Foxp3-GFP expression in transferred CD45.2⁺CD4⁺CXCR5^highPD-1^high cells. (f) Contour plots of splenic CD4⁺ T cells and quantification of T_FR cells (g) from Foxp3^DTR mice six days after SRBC immunization and administration of either 0.9% saline (top panel) or DT (lower panel). Histograms show Foxp3⁺ cells within the CD4⁺CXCR5^highPD-1^high compartment. (h) Flow cytometric contour plots of splenic CD4⁺ cells from Foxp3-cre x ROSA-Stop-flox-YFP mice immunized seven days previously with SRBC (left panel). Enumeration of the proportion of CD4⁺CXCR5^highPD-1^high cells that expressed YFP and/or Foxp3 (right panel). Each symbol represents one mouse and horizontal bars represent median values. Figures are representative of 2-4 independent experiments.

To test whether T_FR derive from Foxp3⁺ precursors, 1×10⁶ naïve CD4⁺CD44^lowFoxp3⁻ T cells or CD44^intFoxp3⁺ T_reg from unimmunized Foxp3^GFP mice were adoptively transferred into congenically-marked mice. Seven days after immunization with KLH in Ribi, ~1-2% of both donor-origin T_reg and donor-origin naïve cells had upregulated CXCR5 and PD-1 to high levels (Figure 4e). More than 90% of donor-origin CXCR5^highPD-1^high T_reg cells retained Foxp3 expression, but none of the transferred naïve T cells that became CXCR5^highPD-1^high after immunization switched on Foxp3 to become T_FR cells (Fig. 4e).

We then asked whether thymic T_reg (nT_reg) could become T_FR cells. Thymic Foxp3⁺CD4^SP or Foxp3⁻CD4^SP from Foxp3^GFP CD45.2 mice were adoptively transferred into CD45.1 mice. Seven days after SRBC immunization only Foxp3⁺ CD45.2 cells had become T_FR; Foxp3⁻ CD45.2 cells had become T_FH but not T_FR (Supplementary Fig. 5). Furthermore, ~97% of T_FR cells expressed Helios (Supplementary Fig. 5), a transcription factor which has been reported to be expressed by thymic-derived nT_ref but not T_ref induced in the periphery³⁹.

In an alternative strategy to confirm that T_FR derive from Foxp3⁺ precursors we used mice in which the gene encoding for the diphtheria toxin receptor (DTR) has been inserted in the Foxp3 locus (Foxp3^DTR), so that treatment with diphtheria toxin (DT) selectively ablates all T_reg within 48 hours ⁴⁰. Foxp3^DTR mice were immunized with SRBC and treated with either DT or saline immediately afterwards. Six days after immunization, T_FR cells had formed normally in mice that did not receive DT but were absent in DT-treated mice (Fig. 4f, g) demonstrating T_FR cannot form if Foxp3⁺ cells are absent at the time of immunization.

Within Peyer’s patches, but not in the spleen, it has been demonstrated that a proportion of Foxp3⁻ T_FR cells derive from Foxp3⁺ precursors⁴¹. To investigate whether splenic T_FR cells stably express Foxp3, we immunized Foxp3-Cre⁴² x ROSA-Stop-flox-YFP mice, in which any cell that has expressed Foxp3 is permanently marked. Seven days after SRBC immunization, the majority (97%) of YFP⁺ CD4⁺CXCR5^highPD-1^high T_FR cells coexpressed Foxp3 and less than ~0.6% of CXCR5^highPD-1^high cells were Foxp3⁻YFP⁺ (Fig. 4h), consistent with the observation that Foxp3 expression is stable in the natural T_reg population⁴³. These data suggest that T_FR cells derive from Foxp3⁺ thymic T_reg that co-opt the T_FH cell differentiation program to migrate to germinal center, where they maintain Foxp3 expression.

T_FR cells suppress T cell proliferation in vitro and repress T_FH cells in vivo

Expression of Foxp3 by CD4⁺ T cells initiates a transcriptional program that confers suppressor function^11-13. T_FR and T_reg sorted from immunized Foxp3^GFP mice displayed comparable suppressive ability in vitro (Supplementary Fig. 6a). To determine whether T_FR cells are suppressive in vivo, we sought to ablate T_FR cells after the germinal center response had been established. For this, DT was first administered to Foxp3^DTR mice 5d after SRBC immunization, when T_FH and T_FR cells have already formed, but the response has not yet reached its peak⁴⁴. After 3 days of DT treatment – 8 days after immunization – only 1% of T_FR cells and 5% of T_reg cells were present compared with vehicle-only-treated mice (Supplementary Fig. 7), and DT-treated mice displayed a significant increase in both the proportion and total number of T_FH cells compared to controls (Fig. 5a). At this time point, germinal center B cell numbers were comparable between DT- and vehicle-treated groups (Figure 5b).

Figure 5. T_FR regulate the size of the T_FH population.

Flow cytometric contour plots and graphs of T_FH cells (a) and germinal center B cells (b) from the spleens of Foxp3^DTR mice immunized eight days previously (d0) with SRBC. Five days after immunization the mice were treated with either DT or saline. (c-f) Analysis of mixed bone marrow chimeras generated by sub-lethally irradiating Rag2^−/− mice and reconstituting their immune system with either a 1:1 ratio of Sh2d1a^−/− CD45.2 : Foxp3^DTR CD45.1 bone marrow or control Sh2d1a^+/+ CD45.2 : Foxp3^DTR CD45.1 bone marrow. Eight weeks after reconstitution chimeric mice were immunized with SRBC and treated with 50μg/Kg of DT on one day prior to immunization and d2 and d5 thereafter. Splenocytes were analyzed on d8 for the proportion and total number of CD4⁺CXCR5^highPD-1^highFoxp3⁺ T_FR cells (c), CD4⁺Foxp3⁺ T_reg (d), CD4⁺CXCR5^highPD-1^high T_FH cells (e) and B220⁺ GL-7^highCD95^high germinal center B cells (f). Each symbol represents one mouse and horizontal bars represent median values. Statistical significance was determined using a Mann-Whitney Test: *P<0.05, **P<0.01.

To confirm the in vivo regulatory role of T_FR, it is necessary to deplete them while leaving the T_reg and T_FH compartments intact. To achieve this we generated mixed bone marrow chimeras with a 1:1 ratio of either congenically marked Sh2d1a^−/−:Foxp3^DTR or Sh2d1a^+/+:Foxp3^DTR marrow. After immunization and DT treatment, CD45.2⁺ Sh2d1a^−/− cells should not able to form T_FH or T_FR cells, but should form T_reg cells normally, whereas Foxp3^DTR cells should form T_FH cells, but lack all Foxp3⁺ cells. Thus, T_FH and T_reg populations should still be present in both chimeras, but T_FR cells will be selectively reduced in Sh2d1a^−/− : Foxp3^DTR mice (Supplementary Table 2).

Such chimeric mice were generated and treated with DT 1 day prior to, and 2 and 5 days after SRBC immunization. At day 8 after immunization >90% of T_FH cells in Sh2d1a^−/−:Foxp3^DTR chimeras derived from Foxp3^DTR cells compared to ~50% in the control Sh2d1a^+/+:Foxp3^DTR mice (Supplementary Fig. 8a). The size of the T_reg population was comparable in both groups of mice (Fig. 5c) with the majority (97%) of T_reg being CD45.2⁺ (Supplementary Fig. 8b). Critically, the number of T_FR was reduced by 5 times in Sh2d1a^−/−: Foxp3^DTR mice compared with the control chimeras (Fig. 5d, Supplementary Fig. 7c). There was also an increase in both the number and proportion of T_FH cells (Fig. 5e) and of germinal center B cells (Fig. 5f), indicating that T_FR limit the germinal center response during a T-dependent immune reaction in vivo.

T_FR limit the outgrowth of non-antigen-specific germinal center B cells

To further characterize the role of T_FR during the germinal center response, T_FR were transiently depleted at the peak of the germinal center response. For this, Foxp3^DTR and control Foxp3^WT mice were immunized with NP-KLH in alum, and treated with DT six days later. Ten days after immunization, the fraction of germinal center B cells was not significantly different between the two groups (Fig. 6a, b) but combined T_reg and T_FR depletion led to a reduction in the proportion of germinal center B cells specific for the dominant epitope of the immunizing antigen, the hapten NP (Fig. 6a, c).

Figure 6. T_FR restrict the outgrowth of non-antigen specific clones in the germinal center.

Flow cytometric contour plots (a) and graphs (b) of total GL-7⁺CD95⁺ germinal center B cells and (c) NP⁺ germinal center B cells ten days after immunization of Foxp3^WT and Foxp3^DTR mice that have been treated with DT 6 days after NP-KLH immunization. Statistical analyses performed using Mann Whitney U-test. Experimental outline (d) of immunization and DT or saline treatment scheme of Foxp3^DTR mice (n=8 per group) to examine the antigen specific immunoglobulin response over time, mice were bled prior to, and d10, d15, d20 and d28 after primary immunization. Mice were given a booster immunization 24 days after the primary immunization. (e) ELISA analysis of NP12 and NP2 antibodies in the experiment outlined in (d). Error bars represent the standard error of the mean from eight individual mice from one experiment, representative of two experiments. Statistical analyses in (e) were performed using a two-way ANOVA with Bonferroni post test to compare differences at each time point. Graphs and flow cytometric contour plots of NP⁺ germinal center B cells (f), total GL-7⁺CD95⁺ germinal center B cells (g) and NP⁺ bone marrow plasma and memory cells (h, i) 21 days after NP-CGG immunization of chimeric mice generated by reconstituting Rag2^−/− mice with a 1:1 mix of Sh2d1a^−/−:Foxp3^−/−, Sh2d1a^+/+:Foxp3^+/+, Sh2d1a^+/+:Foxp3^−/− and Sh2d1a^−/−:Foxp3^+/+ fetal liver. Statistical analyses in (f, g, h and i) were performed using a one-way ANOVA with Bonferroni post test correction. Each symbol represents one mouse and horizontal bars represent median values. *P<0.05, **P<0.01, ***P<0.001.

In order to determine whether this had any long-term impact on the antigen-specific antibody response, Foxp3^DTR mice were immunized with NP-KLH in alum and treated with either DT or saline at 7, 10 and 13 days after immunization, then boosted with NP-KLH in alum 24 days after primary immunization (Fig. 6d). High (anti-NP2) & low (anti-NP12) affinity anti-NP antibody titers were assessed by ELISA prior to and d10, 15, 10 and d28 after immunization. Titers of high and low affinity antibodies were comparable until d20, when both were reduced in DT-treated mice compared with controls (Fig 6e). Four days after secondary immunization (day 28 after primary challenge), high affinity antibodies remained lower in DT-treated mice (Fig 6e). This suggests that depletion of T_FR and T_reg during the germinal center response does not increase antigen-specific antibody production.

To test whether the reduction in the antigen-specific germinal center response was a consequence of T_FR depletion rather than general T_reg depletion, we generated Sh2d1a^−/−:Foxp3^−/− mixed fetal liver chimeras that selectively lack T_FR cells. Three groups of control chimeras were generated in parallel: Sh2d1a^+/+:Foxp3^+/+, Sh2d1a^+/+:Foxp3^−/− and Sh2d1a^−/−:Foxp3^+/+. Eight weeks after reconstitution, chimeric mice were immunized with alum-precipitated NP-chicken gammaglobulin (NP-CGG). 21 days post-immunization, Sh2d1a^−/−:Foxp3^−/− chimeras had a reduction in T_FR cells compared to all control groups (Supplementary Fig. 9a and b) and an expanded T_FH population (Supplementary Fig. 9c). The proportion of Foxp3⁺ T_reg was comparable amongst all 4 groups (Supplementary Fig. 9d). Consistent with a selective defect in T_FR and not T_reg, circulating T_H1 and T_H2 cells were not expanded in Sh2d1a^−/−:Foxp3^−/− mice 14 days post-immunization (Supplementary Fig. 10a and b) nor in the spleen 21 days after immunization (Supplementary Fig. 10c and d). This also confirms that T_FR cells are specialized in the regulation of follicular responses, while other T_reg effectors control T_H1 and T_H2 cells.

As observed in Foxp3^DTR mice, the proportion and absolute number of antigen-specific (NP⁺) germinal center B cells was reduced in Sh2d1a^−/−:Foxp3^−/− mice compared with control chimeras (Fig. 6f), despite formation of abundant germinal centers (Fig. 6g). NP-specific splenic memory B cells and NP-specific bone marrow plasma cells appeared reduced in Sh2d1a^−/−:Foxp3^−/− chimeras compared with controls 21 days after immunization (Figure 6h and i) and the differences were statistically-significant against all control groups except for the Sh2d1a^−/−:Foxp3^+/+ group, in which there was greater variability. At this time point there was no difference in anti-NP antibody titers between Sh2d1a^−/−:Foxp3^−/− mice compared with control chimeras (Supplementary Fig. 11), probably due to a large component being of extrafollicular origin given that abundant NP⁺ plasma cells were still detectable in the spleen (Supplementary Fig. 11). Together, this suggests that T_FR act to limit the outgrowth of non-antigen specific clones in the germinal center.

Discussion

We have shown here that in response to T-dependent antigens a proportion of naïve T_reg can turn on Bcl-6, which allows them to adopt the T_FH differentiation program and express the follicular homing receptor CXCR5 to localize to the germinal center. Here they exert suppressive functions on T_FH cells and the germinal center response. Unlike T_FH cells, T_FR cells express Blimp-1, which is required to control their numbers in the germinal center. In response to other extracellular stimuli, naïve Foxp3⁺ cells can turn on Tbet or increase activity of IRF4 or STAT3, required for T_H1, T_H2 and T_H17 cell formation respectively^16-18. The specialized regulatory program determined by each of these transcription factors is likely to operate via modification of the T_reg chemokine receptor profile in order to allow migration into an anatomical location where T_reg are poised to regulate specific T cell responses to prevent autoimmunity or inflammation-associated tissue damage.

Our data suggests T_FR cells are specialized in controlling the germinal center reaction through limiting the numbers of T_FH cells and inhibiting selection of non-antigen-specific B cells including those carrying self-reactive receptors. Furthermore, the ability of T_FR cells to ensure dominance of antigen-specific clones over the germinal center response appears essential for formation of normal numbers of long-lived plasma cells and memory B cells. Although it is likely that T_FR control germinal center B cells indirectly, through their ability to limit T_FH numbers, it is also possible that T_FR also negatively regulate germinal center B cells directly. This would be akin to the description of direct regulation of antigen-presenting cells by T_reg^{45, 46} and reports that T_reg can directly regulate B cell function⁹. Germinal center B cells are the predominant APCs within the germinal center microenvironment, and this makes them attractive candidates for T_FR-mediated inhibition.

Selection of cognate germinal center B cells by T_FH cells is one of the key mechanisms by which germinal center tolerance is regulated⁴⁷. Dysregulation of the T_FH population has been previously demonstrated to result in autoimmunity^{3, 5, 7, 48}, highlighting the need to tightly control positive selection in germinal centers. Understanding the mechanisms by which T_FR are regulated and their TCR specificity will be important for dissecting the pathogenesis of the increasing number of pathologies in which T_FH cells appear to play a role, including disease-associated ectopic germinal center formation seen in many autoimmune diseases³, atherosclerosis⁴⁹, and chronic allograft rejection⁵⁰. We therefore postulate that T_FR cells may represent a critical peripheral tolerance mechanism, essential for preventing germinal center-derived autoimmunity.