Tfr cells lack IL-2Rα but express decoy IL-1R2 and IL-1Ra and suppress the IL-1–dependent activation of Tfh cells

See allHide authors and affiliations

Science Immunology  08 Sep 2017:
Vol. 2, Issue 15, eaan0368
DOI: 10.1126/sciimmunol.aan0368

Immune regulation in the germinal center

Unregulated production of antibodies may contribute to the development of autoimmunity. Follicular regulatory T (Tfr) cells are thought to limit the germinal center (GC) reaction and to reduce antibody production within B cell follicles in both humans and mice, yet how Tfr cells control the GC reaction remains unclear. Ritvo et al. closely characterize Tfr cells and identify these cells as a rare population of CD4+CXCR5+PD-1+Foxp3+ cells that do not express CD25 and do not respond to interleukin-2. When compared with follicular helper T (Tfh) cells and regulatory T cells, Tfr cells clustered with Tfh cells. Moreover, they expressed decoy molecules for the interleukin-1 signaling pathway, suggesting a mechanism for the suppression of Tfh cells.


Follicular regulatory T (Tfr) cells from lymph node germinal centers control follicular helper T (Tfh) cell–dependent B cell activation. These scarce cells, often described and purified as CD25+ cells, are thought to be derived from thymic regulatory T (Treg) cells. However, we observed that mouse Tfr cells do not respond to interleukin-2 (IL-2), unlike Treg cells. Stringent immunophenotyping based on B cell lymphoma 6 (Bcl6), programmed cell death protein 1 (PD-1), and CXCR5 expression revealed that Tfr cells are actually CD25, in mice and humans. Moreover, Tfr cell characterization based only on CXCR5 and PD-1 high expression without excluding CD25+ cells resulted in contamination with Treg cells. Transcriptome studies of CD4+CXCR5+PD-1+Bcl6+Foxp3+CD25 Tfr cells revealed that they express the IL-1 decoy receptor IL-1R2 and the IL-1 receptor antagonist IL-1Ra, whereas Tfh cells express the IL-1R1 agonist receptor. IL-1 treatment expanded Tfh cells in vivo and activated their production of IL-4 and IL-21 in vitro. Tfr cells suppressed the IL-1–induced activation of Tfh cells as efficiently as the IL-1 receptor antagonist Anakinra. Altogether, these results reveal an IL-1 axis in the Tfh cell control of B cell responses and an IL-2/IL-1 dichotomy for Treg cell control of effector T cells versus Tfr cell control of Tfh cells.


Germinal centers (GCs) form within peripheral lymphoid organs to foster T cell–dependent stimulation of B cells that results in the production of high-affinity antibodies (1, 2). To produce such antibodies, GC B (GCB) cells interact with follicular helper T (Tfh) cells to enable their activation and differentiation into plasma cells (35). The Tfh cell lineage is imprinted by expression of B cell lymphoma 6 (Bcl6), a transcription factor that represses alternative fates and promotes Tfh cell differentiation and function (6). Bcl6 promotes the expression of CXCR5 (79), which controls the homing of Tfh cells to the B cell follicle and GCs (10, 11). Interleukin-6 (IL-6) and IL-21 contribute to Tfh cell differentiation (12, 13) when naïve T cells are stimulated through their T cell receptor (TCR) and costimulatory molecules, such as ICOS (inducible costimulator) and CD28 (7, 14, 15). Tfh-mediated CD40L/CD40 signaling in the presence of IL-21 and IL-4 induces proliferation, isotype switching, and differentiation of B cells (1620).

The generation of autoantibodies and the development of autoimmune diseases can result not only from the activation of autoreactive T and B cells but also from a dysregulation of GC formation and maintenance by Tfh cells (21, 22). Limiting the number of Tfh cells within GCs is critical to warding off the emergence of autoantibodies (2325). Thus, regulation of Tfh cells is essential for limiting GC reactions against self-antigens and preventing autoimmunity (26, 27).

Follicular regulatory T (Tfr) cells were found to limit the GC reaction and to reduce antibody production within B cell follicles in human tonsils (28) and in mice (29). These cells were initially reported to express Bcl6, programmed cell death protein 1 (PD-1), and CXCR5, the canonical markers of Tfh cells, as well as forkhead box P3 (Foxp3) and CD25, the canonical markers of regulatory T (Treg) cells (30, 31). Tfr cells were first described as CD25hi, similar to Treg cells (29, 32), and Tfr cells are characterized/purified according to this marker in most studies. Except for mice expressing a reporter gene of Foxp3 expression, there is no other physiological surface marker that allows one to separate live Tfr cells from Tfh cells. Tfr cells were proposed to develop not only from thymus-derived Treg cells but also from Foxp3 precursors in a PD ligand 1–dependent manner (33).

How Tfr cells control the GC reaction remains unclear. In vitro suppression assays showed that Tfr cells suppress B and Tfh cell responses, inhibiting B cell activation and class switch recombination and decreasing Tfh cell production of cytokines such as IL-4 (34, 35). Tfr suppression of Tfh cells has been reported to be dependent on cytotoxic T lymphocyte–associated protein 4 (36, 37). Depletion of Tfr cells in Bcl6fl/fl × Foxp3CRE mice had no impact on the size of the Tfh cell or GCB cell compartments (38). However, vaccinated Tfr-depleted mice have altered antigen-specific antibody responses, with significantly increased immunoglobulin A (IgA) levels and decreased avidity of IgGs against the immunogen, suggesting a qualitative role of Tfr cells in antibody production (38).

The antigen specificity of Tfh and Tfr cells is not well known, and diverging results have been reported regarding Tfr cells. Concordant results showed that Tfh cells are specific for the immunizing antigen (33, 39, 40). In contrast, a previous study showed that Tfr and Tfh cells share TCRs specific for the immunizing antigen (33), whereas a recent one reported oligoclonal expansions in Tfh cells and a broad TCR usage in Tfr cells from the same GCs (41).

In this work, we identified Tfr cells as a rare population of CD4+CXCR5+PD-1+Foxp3+ cells that do not express CD25 and likewise do not respond to IL-2. On the basis of a stringent characterization of Tfr cells, we analyzed their transcriptome and compared it with that of Tfh and Treg cells. Transcriptome studies clustered Tfr cells with Tfh cells rather than Treg cells and revealed a previously uncharacterized IL-1 axis in Tfr cell suppression of Tfh cells. Together, our work should prompt a reassessment of the biology of Tfr cells based on their better characterization and highlights an IL-2/IL-1 dichotomy for Treg cell control of effector T (Teff) cells versus Tfr cell control of Tfh cells.


Tfr and Tfh cell response to IL-2

We investigated the in vivo response of Tfh and Tfr cells to IL-2. Mice expressing the green fluorescent protein (GFP) as a reporter gene of Foxp3 expression were injected with an adeno-associated virus coding for IL-2 (AAV–IL-2) that allows continuous production of IL-2 (42). Twenty-one days after AAV–IL-2 injection, mice were immunized or not immunized with ovalbumin (OVA), and their spleens were analyzed 10 days after the immunization. CD19+CD4 B cells, which are known to be CXCR5hi and PD-1, were used to select the appropriate gating for defining CD4+CXCR5hiPD-1hi follicular T (Tfol) cells (fig. S1). GFP expression was then used to identify the subsets of Foxp3+ Tfr cells and Foxp3 Tfh cells within Tfol cells.

IL-2 increased the proportion (Fig. 1A) and numbers (fig. S2) of splenic Treg cells, as previously reported (4244), irrespective of the immunization. The marked increase of Tfh cells induced by immunization was reduced by IL-2 (Fig. 1B). In contrast, whereas immunization had no impact on the proportion of Tfr cells, IL-2 induced a significant decrease of their proportion in CD4+ cells (Fig. 1C). The combined effects of IL-2 on Tfh and Tfr cells led to a decreased Tfr/Tfh ratio for both immunization and IL-2 treatment, alone or in combination (Fig. 1D).

Fig. 1 Tfr and Tfh cell response to IL-2.

(A to D) Proportions of Treg (A), Tfh (B), and Tfr (C) cells in CD4+ T cells and the Tfr/Tfh ratio (D) of untreated mice (filled circles), IL-2–treated mice (filled squares), immunized mice (empty circles), and immunized IL-2–treated mice (empty squares). n = 5; *P < 0.05, **P < 0.01, Mann-Whitney U test. Data are representative of three independent experiments.

Tfr cells do not express CD25

This negative impact of IL-2 on Tfr cells led us to carefully assess their expression of CD25 (IL-2Rα, a component of the high-affinity receptor for IL-2) by flow cytometry. We first stringently defined Tfol cells on the basis of high CXCR5 and PD-1 expression within CD4+ cells. This gating defined Tfol cells as all CXCR5hiPD-1hi cells expressing Bcl6 (6), whereas CXCR5int/loPD-1int/lo non-Tfol cells did not (Fig. 2A). Within each of these two cell populations, we could detect the presence of a subset of Foxp3+ cells, thus defining CD4+CXCR5hiPD-1hiBcl6+Foxp3+ Tfr cells and CD4+CXCR5int/loPD-1int/loBcl6Foxp3+ Treg cells (Fig. 2B). Tfr cells did not express CD25, unlike Treg cells that expressed it highly (Fig. 2C). Similar observations were made with cells obtained from C57BL/6, BALB/c, and NOD (nonobese diabetic) genetic backgrounds (fig. S3). Expression levels of Bcl6 and CD25 were comparable among CXCR5hiPD-1hi Tfh and Tfr cells (fig. S4).

Fig. 2 Tfr cells do not express CD25.

(A) Flow cytometry contour plots (left) showing CXCR5hiPD-1hi Tfol (orange) and CXCR5int/loPD-1int/lo non-Tfol (blue) cells within CD4+ T cells and histograms showing their Bcl6 expression status (right). (B) Foxp3 expression of Tfol and non-Tfol cells defining Treg cells and Tfr cells (both Foxp3+). (C) Mean fluorescence intensity (MFI) of CD25 expression levels in Treg cells and Tfr cells (n = 5). Data are representative of three independent experiments (left). Representative histogram of the expression of CD25 in Tfr and Treg cells (right). (D and E) IL-4 (D) and IL-21 (E) production by Tfh cells from immunized mice, in the presence (n = 4) or absence (n = 4 to 5) of Tfr cells. Results are representative of two independent experiments. (F) Flow cytometry plot identifying different populations according to their PD-1 and CXCR5 expression within CD4+Foxp3+ T cells (left) and their expression of Bcl6 (middle) and CD25 (right). For (C) to (E), *P < 0.05, **P < 0.01, Mann-Whitney U test.

To confirm the bona fide nature of CXCR5hiPD-1hiFoxp3+CD25 Tfr cells, we studied their suppressive activity in a classic in vitro assay investigating the inhibition of IL-4 and IL-21 production by Tfh cells from immunized mice cocultured with B cells (34). Adding Tfr cells to these cocultures significantly decreased IL-4 production (Fig. 2D) and almost completely suppressed IL-21 production (Fig. 2E).

Because Tfr cells were initially described as CD25+ (32), we speculated that this could be due to contamination by CXCR5int/loPD-1int/lo Treg cells. We thus analyzed CD25 and Bcl6 expression on a gradient of subsets defined by the intensity of CXCR5 and PD-1 expression among CD4+Foxp3+ cells (Fig. 2F, left). Whereas CXCR5hiPD-1hi Tfr cells homogeneously express Bcl6 and are CD25, the progressive decrease in the intensity of CXCR5 and PD-1 expression was associated with a parallel loss of Bcl6 expression and an increase in CD25 expression (Fig. 2F, middle and right). These results indicate that a nonstringent characterization of Tfr cells leads to the inclusion of Treg cells.

Tfr cells’ transcriptomic profile distinguishes them from Treg and Tfh cells

We analyzed the expression profile of a set of 545 immune-related genes in Tfh, Tfr, and Treg cells from NOD and BALB/c mice using NanoString technology. Hierarchical clustering based on the entire set of genes clustered Treg cells samples apart from Tfol cell samples (fig. S5). We further used the 81 genes that better separate the three cell subsets (see the Supplementary Materials) to perform an additional hierarchical clustering (Fig. 3A) that again identified two main clusters of Tfol and Treg cells, regardless of the genetic background. Within the Tfol cluster, Tfr and Tfh cells were well separated. We evaluated the accuracy and statistical robustness of this clustering by multiscale bootstrap resampling using the pvclust R package (Fig. 3B) (45, 46). This statistical process attributes accuracy to each branch of the clustering by calculating an approximately unbiased (AU) P value as a percentage (45). The higher the percentage, the higher the accuracy of a given branch of the clustering. In our data set, the AU P values for the separation of Treg and Tfol cells, as well as for the separation of Tfr and Tfh cells within Tfol cells, were 100%, indicating the robustness of the cell subset identification.

Fig. 3 Tfr cells’ transcriptomic profile distinguishes them from Treg and Tfh cells.

(A) Heat map comparing the gene expression profiles of Treg (green), Tfr (blue), and Tfh (black) cells from two genetic backgrounds after immunization. Red, high gene expression; blue, low gene expression. (B) Dendrogram of Treg, Tfr, and Tfh cells representing the hierarchical cluster analysis performed with the pvclust R package.

Differential gene expression in Treg, Tfh, and Tfr cells

We next analyzed the genes differentially expressed between Tfh, Tfr, and Treg cells using Ingenuity Pathway Analysis (IPA). The IPA’s Upstream Regulator analytic tool identified NFATc1, a subunit of the NFAT (nuclear factor of activated T cells) complex known to interact with Foxp3 (47, 48), as a key component differentiating the transcriptome profiles of our three cell populations. NFATc1 is down-regulated (blue) in Treg cells and up-regulated (orange) in Tfh and Tfr cells, in accordance with (i) the observed expression levels of genes regulated by NFATc1 and (ii) their expected positive or negative regulation by NFATc1 (49), as modeled by IPA. Foxp3 expression aside, the pattern of NFATc1-dependent gene expression was identical between Tfh and Tfr cells. In contrast, each of the NFATc1 up-regulated genes in Treg cells was down-regulated in Tfr and Tfh cells, and vice versa (Fig. 4A).

Fig. 4 Gene and protein expression reveals a unique functional profile for Tfr cells compared with Treg and Tfh cells.

(A) NFATc1 expression and its downstream regulated genes. (B and C) Variations in functions (B) and expression (C) of indicated molecules in Treg (white), Tfh (blue), and Tfr (pink) cells. (D and E) MFI of IL-1R2 (D) (n > 20) and OX40 (E) (n = 8) in Tfr, Treg, Teff, and Tfh cells. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001, Mann-Whitney U test.

IPA then identified the pathways differentially modulated between the cell subsets. Tfr cells appeared as very active cells, with markedly up-regulated pathways related to movement, survival, viability, and proliferation. In contrast, pathways related to death and mortality were up-regulated in Tfh cells (Fig. 4B).

Supervised analyses then identified ILs and their receptors as markedly differentially expressed among the three cell subsets. IL-2Rα, IL-2Rβ, and IL-2Rγ are highly expressed by Treg cells, whereas Tfr and Tfh cells only express IL-2Rβ and IL-2Rγ (Fig. 4C, left). IL-1Rs were not expressed by Treg cells and had distinct expression profiles for Tfol subsets (Fig. 4C, right). Tfh cells only expressed the agonist IL-1 receptor IL-1R1, which transmits the inflammatory message of IL-1. In contrast, Tfr cells expressed lower levels of IL-1R1 than Tfh cells but high levels of IL-1R2, the IL-1 receptor decoy (50), and, to a lesser extent, IL-1Rn, the IL-1R antagonist (51).

IL-4 and IL-21, cytokines involved in Tfh-mediated B cell help (16, 20, 52), were mostly expressed by Tfh cells (fig. S6). IL-2 expression could not be detected in any of the subsets, whereas IL-1β expression was at the limit of detection (number of transcripts ≈ 100) and was similar in each subset (fig. S6). Costimulation molecules involved in B-T cell cooperation were also differentially expressed in Tfol cells. CD40L was mostly expressed on Tfh cells, whereas OX40 (fig. S6) was mostly expressed on Tfr cells. Thus, the IL-1 receptors CD40L and OX40 are the main molecules that discriminate Tfh and Tfr cells, as also confirmed by flow cytometry for IL-1R2 and OX40 (Fig. 4, D and E, and fig. S7).

IL-1–dependent Tfh cell activation

To assess the functional relevance of our IL-1–related gene expression profile, we investigated the role of IL-1β in Tfh cells. We treated immunized mice with recombinant IL-1Ra (Anakinra), which blocks the IL-1β response in humans (53) and mice (5456), or with murine IL-1β.

Compared with untreated immunized mice (OVA + Alum), inhibition of IL-1β during immunization led to a significantly reduced proportion of Tfh cells (Fig. 5A), whereas injection of IL-1β induced a significant increase of Tfh cell proportion. This increase was observed even when immunization was performed without Alum, indicating the autonomous effect of IL-1β on Tfh cells. The increased Tfh cell response induced by IL-1β translated into an increased production of anti-OVA–specific IgG (Fig. 5B).

Fig. 5 IL-1β induces Tfh cell activation that is inhibited by Tfr cells.

(A) Fold change in Tfh cell proportion among CD4+ cells from indicated conditions in comparison with OVA + Alum control condition (dotted line) (n = 6 to 7). Data shown are from one experiment and representative of two independent experiments. (B) Serum anti-OVA–specific IgG production by mice immunized with OVA, treated (n = 3) or not treated (n = 5) with IL-1β. (C) IL-4 production by B cells and Tfh cells from immunized mice, in the presence or absence of B cells or IL-1β. (D and E) IL-4 and IL-21 production by Tfh cells from immunized mice cocultured with B cells, with or without Tfr cells or Anakinra (D) and with or without IL-1β and Tfr cells (E). (C to E) n = 4 to 5; *P < 0.05, **P < 0.01, Mann-Whitney U test.

We then tested the contribution of IL-1β to the production of IL-4 by Tfh cells from immunized mice (Fig. 5C). B cells alone did not produce IL-4, whereas baseline Tfh cell IL-4 production was around 50 pg/ml. Coculture of B and Tfh cells led to a doubling of IL-4 production, whereas the addition of IL-1β rather than B cells led to an even more pronounced increase of IL-4 production by Tfh cells. Thus, Tfh cells are directly activated by IL-1β.

We next investigated the IL-1β dependence of the Tfh cell suppression by Tfr cells. When Tfr cells (at a Tfh/Tfr ratio of 2) were added to a coculture of Tfh and B cells, we observed a 50% reduction in IL-4 production (Fig. 5D), which corresponds to the value of basal IL-4 production by Tfh cells (Fig. 5C). This inhibition was equivalent to the inhibition observed when adding Anakinra (Fig. 5D). This also suggests that some cells in the culture produced IL-1β. We confirmed by flow cytometry that this production was mainly from B cells (fig. S8). Similar results were observed for the production of IL-21, leading to a sevenfold decrease of IL-21 production when Tfr cells are added to the culture (Fig. 5D).

Tfr cells could suppress the IL-1β–induced Tfh cell activation (Fig. 5E). Whereas IL-1β led to a 10-fold increase of IL-4 and IL-21 production by Tfh, the addition of Tfr cells markedly inhibited this IL-1β–dependent IL-4 and IL-21 production (Fig. 5E). Together, these results indicate that there is an IL-1β-dependent activation of Tfh cells that can be suppressed by Tfr cells.

Tfr cell response to immunization with self-antigens or foreign antigens

We compared the response of Tfr cells to antigenic stimulation with two representative antigens. OVA immunization led to a higher increase in Tfh cells compared with insulin (INS) immunization (Fig. 6A), and vice versa, for the proportion of Tfr cells (Fig. 6B). The resulting Tfr/Tfh ratio was thus about fourfold higher after immunization with a self-antigen compared with a foreign antigen (Fig. 6C). This translated well in a higher proportion of CD19+Foxp3Bcl6+GL7+IgD GCB cells among CD19+ B cells (fig. S9) after OVA immunization compared with INS immunization (Fig. 6D). There was an inverse correlation between the proportions of Tfr cells and GCB cells (Fig. 6E). When IL-1β was added during INS immunization, GCB cell values were in the range of those obtained after OVA immunization (Fig. 6F), indicating that the lower expansion of GCB cells after INS immunization was not intrinsically limited by a lower frequency of INS-specific B cells. There was again an inverse correlation between the proportions of Tfr cells and GCB cells (Fig. 6G)

Fig. 6 Tfr cells respond better to immunization with self-antigens than with foreign antigens.

(A to C) Tfh (A) and Tfr (B) cell proportions within Tfol cells and Tfr/Tfh ratio (C) after INS (n = 6) or OVA (n = 7) immunization. (D and E) GCB cell proportions within CD19+ B cells (D) and correlation between the percentage of Tfr and GCB cells from INS-immunized (filled circles) or OVA-immunized (empty circles) mice (E). (F and G) GCB cell proportions within CD19+ B cells (F) and correlation between the percentage of Tfr and GCB cells from INS-immunized mice receiving IL-1β (empty circles) or not (filled circles) (G). (A to C and F) **P < 0.01, Mann-Whitney U test. (E and G) Each symbol represents one mouse, and the Spearman rank correlation value is shown.

Phenotype of Tfr cells in humans

We investigated the phenotype of Tfh and Tfr cells in humans by analyzing these cells in spleen or mesenteric lymph node cells obtained from healthy organ donors, or from tonsils. By using the same gating strategy that we used for mice (Fig. 7A), we could show the absence of CD25 expression on human Tfr cells (Fig. 7B). As for mice, we analyzed CD25 expression on a gradient of subsets defined by the intensity of CXCR5 and PD-1 expression among CD4+Foxp3+ cells (Fig. 7C, left). Although CXCR5hiPD-1hi Tfr cells are homogeneously CD25, the progressive decrease in the intensity of CXCR5 and PD-1 expression was associated with a parallel increase in CD25 expression (Fig. 7C, right). As for mice, IL-1R2 was expressed only on Tfr cells (Fig. 7D).

Fig. 7 Human Tfr cells do not express CD25.

(A) Flow cytometry contour plots showing CD4+Foxp3 and CD4+Foxp3+ T cells (left); CXCR5hiPD-1hi Tfh cells (blue) and CXCR5int/loPD-1int/lo Teff cells (cyan) within CD4+Foxp3 T cells (middle); and CXCR5hiPD-1hi Tfr cells (red) and CXCR5int/loPD-1int/lo Treg cells (green) within CD4+Foxp3+ T cells (right). (B) Histograms showing CD25 expression level (left) and MFI (right) on gated subsets from (A). (C) Flow cytometry contour plot identifying different subsets according to PD-1 and CXCR5 expression within CD4+Foxp3+ T cells (left) and the corresponding expression of CD25 (right). (D) Histograms showing the IL-1R2 expression level of the subsets mentioned above (left) and their MFI (right). (B and D) n = 5; *P < 0.05, **P < 0.01, Mann-Whitney U test.


IL-2 is now widely developed for stimulating Treg cells in the context of inflammatory and autoimmune diseases (57). To better understand the consequences of an IL-2 treatment, and because Tfr cells are thought to derive from Treg cells, we assessed Tfr cell response to IL-2. We found that Tfr cell proportion was reduced rather than augmented during IL-2 treatment. This intriguing observation led us to thoroughly assess their CD25 phenotype and to determine that Tfr cells do not express CD25, neither in mice nor in humans. Contrary to Treg cells, which comprise both CD25+ and CD25 subsets (58), careful investigations assessing expression of the canonical Bcl6 marker (Fig. 2F) led us to conclude that Tfr cells are homogeneously CD25 and CXCR5hiPD-1hi. These results initially obtained in C57BL/6 mice were further confirmed in BALB/c and NOD mice (fig. S3) and also observed for human Tfr cells (Fig. 7C). Although a recent study on Tfr cell conditional depletion in Bcl6fl/flFoxp3CRE mice suggested that Tfr cells are CD25lo rather than CD25+ (38), we believe that they are truly CD25, as also supported by their lack of expansion (Fig. 1C) after treatment with IL-2. IL-2 treatment not only expands CD25+ Treg cells but also induces an increase of CD25 expression by CD25lo/− lymph node Treg cells (59, 60). Our results now define Tfr cells as a very small population of CD4+CXCR5hiPD-1hiFoxp3+Bcl6+CD25 cells. Therefore, when analyzing Tfr cells, which are scarce, if a very stringent gating on the highest CXCR5 and PD-1 expression and/or an exclusion of CD25+ cells is not implemented, there will be significant contamination of the studied population by Treg cells that we estimated at >50% with commonly used gating. Thus, many of the properties initially assigned to Tfr cells have likely been attributed to a mixed population of Treg and Tfr cells. Our results suggest that there is a need to revisit these properties in more stringently selected Tfr cells. It should be realized that this will be difficult, because Tfr cells are reduced to a minute population.

With this phenotypic characterization, we were able to sort only 2000 to 6000 Tfr cells per spleen of immunized mice. Nonetheless, we could confirm the functionality of Tfr cells. In vitro, they suppress Tfh-mediated production of IL-4 and IL-21 (Fig. 2, D and E), necessary cytokines for Tfh-mediated differentiation of antibody-producing B cells (61). Such a reduction of IL-21 and IL-4 production has been described as the main mechanism by which Tfr cells inhibit antibody responses (34). The in vivo suppressive activity of CD25 Tfr cells is also supported by the observation that the sole elimination of Bcl6/Foxp3 cells led to highly abnormal Tfh cell and GCB cell responses (38). As we show that our cells are the only Bcl6+Foxp3+ cells (Fig. 2F), this work strongly supports our conclusions. In vivo, the increase in Tfr and Tfh cells upon immunization was reciprocally dependent on the self/nonself nature of the antigen, suggesting different antigenic specificities, as recently reported (41).

To further revisit Tfr cell functionality on the basis of this previously unknown characterization, we studied the transcriptomic profile of Tfr cells and compared it with that of Tfh and Treg cells. We used the NanoString technology, which is reported to be more quantitative for small cell populations (62). These analyses first showed that Tfr cells were clustered with Tfh and not with Treg cells (Fig. 3A and fig. S5). This clustering was shown to be robust by a multiscale bootstrap sampling statistical validation (Fig. 3B). This suggests that the transcriptional program controlled by Bcl6 imprints the transcriptome landscape of Tfr cells more than the transcriptional program controlled by Foxp3. The first study describing Tfr cells (29) showed a clustering of Tfr cells with Treg cells; we believe that this could be due to contamination of the Tfr cells by Treg cells.

Our observation does not give clues as to the origin of Tfr cells as cells either derived from thymic Treg cells or induced from Tfh cells. The Bcl6-imprinted program could be secondarily imprinted over an initial imprinting by Foxp3, or vice versa. Lineage study and possibly TCR repertoire studies should help address this important issue.

We found that murine Tfr cells highly express the decoy receptor IL-1R2 (Fig. 4C). IL-1R2, an IL-1β and IL-1β negative regulator, has been observed on neutrophils, macrophages, monocytes, B cells, and CD4+ T cells (6366). Tfr cells also express IL-1Ra, the IL-1 receptor antagonist. We confirmed IL-1R2 expression on most, if not all, mouse and human Tfr cells by flow cytometry (Figs. 4D and 7D). IL-1R2 has also been described in some Treg cells after in vitro activation (65, 66) and in pancreatic (67) and infiltrating Treg cells from colorectal tumors, but not from lung tumors or normal tissues (68). We could not detect significant expression of IL-1R2 in Treg cells from immunized mouse splenocytes or human lymph nodes. In contrast, Tfh cells express the IL-1R1 agonist receptor but do not express IL-1R2 or IL-1Ra. These notable observations suggest that there could be an IL-1 axis in the regulation of Tfh cells by Tfr cells.

This IL-1 axis in the control of Tfh cell activation is directly supported by in vitro and in vivo results. The following observations collectively highlight the importance of IL-1β in Tfh cell activation and argue for a direct action of IL-1β on Tfh cells: (i) Tfh cells express IL-1R1 agonist receptor; (ii) IL-1β alone activates Tfh cell–dependent production of IL-4 and IL-21 in vitro; (iii) IL-1β alone (without adjuvant) triggers the expansion of Tfh cells in vivo in response to immunization; and (iv) Anakinra blocks the expansion of Tfh cells in vivo in response to immunization. Furthermore, the following observations demonstrate the biological relevance of antagonist IL-1Ra expression on Tfr cells: (i) Tfr cells suppress the production of both IL-4 and IL-21 by Tfh cells cocultured with B cells to the same extent as Anakinra, and (ii) Tfr cells suppress the IL-1β–triggered production of both IL-4 and IL-21 by Tfh cells.

Last, in the literature, the IL-1 axis in Tfh cell control of B cell response is indirectly supported by numerous observations: (i) administering IL-1 during immunization leads to enhanced antibody production (6971); (ii) antibody production is significantly reduced in IL-1–deficient mice (72, 73) but enhanced in mice lacking the expression of the IL-1 receptor antagonist IL-1Ra (72, 73); (iii) the effect of IL-1 on antibody production works through induction of CD40L (73), which is highly expressed on Tfh cells (fig. S6); and (iv) many adjuvants used for immunization are IL-1 inducers (74, 75). These effects have been ascribed to IL-1β, rather than to IL-1α (72). These results, together with the observation that IL-1β alone activates Tfh cells in vitro and in vivo, support the idea that our results should be relevant to all immunization procedures increasing IL-1β production.

Tfh and Tfr cells are important for the regulation of humoral responses, in health and disease. Our results support that (auto)antibody production depends on IL-1β availability for Tfh cells, regulated by IL-1 antagonists expressed by Tfr cells. However, how general the contribution of the IL-1 axis is in the Tfh/B cell response to various antigenic stimulations (such as against other foreign antigens or self-antigens and during infection or autoimmune diseases) remains to be investigated, because this was not explored in this study. The mechanisms at work for its regulation also remain to be dissected. Selective knockout of IL-1R1 and IL-1R2/IL-1Ra in Tfh and Tfr should help address these questions. Last, the lineage of Tfr cells remains to be elucidated. Together, our results indicate a dual regulation of T cells in lymph nodes, one between Treg and Teff cells regulated by IL-2 outside GCs and the other between Tfh and Tfr cells regulated by IL-1β inside GCs.


Study design

For human and mice flow cytometry assays, the sample sizes were of at least five individuals per experiment. For enzyme-linked immunosorbent assay (ELISA) and NanoString, sample sizes were of at least four per condition for which cells were sorted from pooled splenocytes from nine mice to ensure a sufficient number of Tfr cells. Sample size was determined on the basis of experimental feasibility and for statistical significance. The experiments were not randomized. The investigators were not blinded to the allocation during experiments and analyses.


Male and female NOD Foxp3-GFP and BALB/c Foxp3-GFP (C.129X1-Foxp3tm3Tch/J) mice (8 to 14 weeks old) expressing GFP under the control of the promoter of Foxp3 gene were provided by V. Kuchroo of Brigham and Women’s Hospital (Boston, MA). C57BL/6 Foxp3-GFP mice expressing GFP under the control of the promoter of Foxp3 gene were provided by B. Malissen of the Centre d’Immunologie de Marseille-Luminy (France). All animals were maintained at the University Pierre and Marie Curie (UPMC) Centre d’Expérimentation Fonctionnelle animal facility (Paris, France) under specific pathogen–free conditions in agreement with current European legislation on animal care, housing, and scientific experimentation (agreement number A751315). All procedures were approved by the local animal ethics committee.

IL-2 treatment in mice

AAV8-CAG-IL2 recombinant vectors (rAAVs) were generated by triple transfection of human embryonic kidney 293T cells, as described previously (42). Mice were injected intraperitoneally once before immunization at day −21 (D−21) with 1010 viral genomes of rAAVs diluted in 100 μl of 1× phosphate-buffered saline (PBS1×).


Mice were either immunized once (D0) and sacrificed at D10 for Tfr/Tfh phenotypic and transcriptomic studies or immunized twice (D0 and D14) and then euthanized at D28 for the analysis of GCB cells. Intraperitoneal injection was performed with 100 μg of OVA (OVA A5503, Sigma-Aldrich) mixed with 500 μg of aluminum hydroxide (Alum) gel (AlH303, Sigma) or with 4.5 IU of human INS (Umuline Rapide, Lilly) mixed with 500 μg of Alum.

In vivo treatment

Mice treated with recombinant human IL-1Ra received Anakinra (1 mg per mouse; Amgen) every 24 hours from D−1 before immunization to D4 after immunization. Mice treated with IL-1β received recombinant mouse IL-1β (0.5 μg per mouse; BioLegend) at D1 and D2 after immunization.

Flow cytometry analysis of mouse cells

Fresh total cells from lymph nodes and spleens were isolated in PBS1×–3% fetal bovine serum (FBS) and stained for 20 min at 4°C with the following monoclonal antibodies at predetermined optimal dilutions: CD121b-BV421, CD19-PeCF594, CD4-V500, CD8a-AF700, Bcl6-APC, CD278-BV421, CXCR5-Biotin (BD Biosciences), GL7-e450, Foxp3-AF488, PD-1–PE (PD-1–phycoerythrin), CD134-APC (CD134-allophycocyanin) (eBioscience), streptavidin-APC (eBioscience) or streptavidin–APC-Cy7 (BD Biosciences), CD25-PC7 (eBioscience), or CD25–eFluor 660 (BD Biosciences). CXCR5 staining was performed using biotinylated anti-CXCR5 for 30 min at 20°C followed by APC- or APC-Cy7–labeled streptavidin at 4°C. Intracellular detection of Foxp3 was performed on fixed and permeabilized cells using appropriate buffer (eBioscience), following the manufacturer’s recommendations. Cells were acquired on an LSR II flow cytometer (Becton Dickinson) and analyzed using FlowJo software (TreeStar Inc.). Dead cells were excluded by forward/side scatter gating.

Human tissues

Human tissues were obtained from prospective organ donors through an approved research protocol (no. 2014-108) authorized by the French Biomedicine Agency and the Ministry of Education and Research. Relatives of the donors provided informed consent for the collection of samples. Tissues were collected after the organs were flushed with cold preservation solution.

Tissue samples were maintained in cold PBS and brought to the laboratory within 2 to 4 hours of organ procurement where they were rapidly processed using mechanical digestion, resulting in high yields of living lymphocytes. For the spleen, an additional step was carried out to isolate mononuclear cells by density gradient centrifugation with Lymphoprep (STEMCELL Technologies).

Flow cytometry analysis of human cells

Cell pellets were resuspended in a 106 cells/50 μl of FBS concentration and then treated using the PerFix-nc kit (Beckman Coulter), following the manufacturer’s recommendations. Cells were stained using a panel of fluorescence-conjugated fluorescein isothiocyanate (FITC), PE, PE–Texas Red (PE-TR), AF647, PE-Cy7, AF405, Pacific Blue, or BV510 monoclonal antibodies to detect the following cell proteins: CD4, CD25, CD127, Foxp3 (Beckman Coulter), CD3 (BD Biosciences), PD-1, CXCR5 (Ozyme), and IL-1R2 (R&D Systems). Flow cytometry data acquisition was performed on BD FACSAria II (BD Biosciences). Control samples included unstained and single fluorochrome-stained compensation beads (UltraComp eBeads, eBioscience). Flow cytometry data were analyzed using FlowJo software (TreeStar Inc.).

Cell sorting

Splenocytes from immunized mice were stained with Ter-119–biotin and B220-biotin antibodies for 20 min at 4°C and labeled with anti-biotin magnetic beads (Miltenyi Biotec) for 15 min at 4°C. B cells and erythrocytes were depleted on an autoMACS separator (Miltenyi Biotec), following the manufacturer’s procedure. Enriched T cells were stained as described in the “Flow cytometry analysis of human cells” section, and the following subsets were sorted on BD FACSAria II (BD Biosciences), with a purity of >98%: CD4+CD8CXCR5hiPD-1hiFoxp3 Tfh cells, CD4+CD8CXCR5hiPD-1hiFoxp3+ Tfr cells, and CD4+CD8CXCR5int/loPD-1int/loFoxp3+ Treg cells.

Suppression assays

Tfh, Tfr, and CD19+ B cells were sorted from D−10 OVA-immunized mice, following a previously described “Immunization” protocol. For all the conditions, 2 × 104 Tfh cells and 5 × 104 B cells were cultured for 96 hours in 96-well plates (Nunc) in complete RPMI 1640 (Thermo Scientific) with anti-IgM (5 μg/ml; clone eB121-15F9, eBioscience) and three CD3/CD28 beads for one T cell (Dynabeads Mouse T-Activator, Thermo Scientific). We then added either (i) 1 × 104 Tfr cells with (“Tfh + B + Tfr + IL-1”) or without (“Tfh + B + Tfr”) 1 μg of recombinant mouse IL-1β (BioLegend), (ii) only 1 μg of recombinant mouse IL-1β (“Tfh + B + IL-1”), or (iii) Anakinra (500 ng/ml; Amgen) (“Tfh + B + Anakinra”). IL-4 and IL-21 secretion levels were measured by ELISA (eBioscience) in supernatants of cultured cells, according to the manufacturer’s recommendations.

Gene expression analysis based on a NanoString immunology panel

Sorted cells were washed in PBS1× and stored in RNAqueous kit lysis buffer (Ambion Inc./Life Technologies) at −80°C. Total RNA was extracted according to the manufacturer’s instructions, and quality was assessed on a bioanalyzer using the Pico RNA Reagent Kit (Agilent Technologies). Gene expression was analyzed using a NanoString mouse immunology panel, following the manufacturer’s recommendation (see Supplementary Materials and Methods).

Statistical analysis

Flow cytometry, cytokine production, and gene expression data were analyzed using nonparametric Mann-Whitney U test on GraphPad Prism v5 [P values, such as P > 0.05 (not significant), *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001, are indicated in the figures]. Exact P values are provided in table S1. Nonparametric correlation analyses were performed by Spearman rank correlation coefficient calculation. NanoString data were analyzed using IPA (QIAGEN). Clustering analysis and multiscale bootstrap resampling were performed using heatmap.2 and pvclust R packages on R version 3.1.3 (cluster method, average; distance, correlation).

See the Supplementary Materials for additional Materials and Methods information.


Materials and Methods

Fig. S1. Representative flow cytometry gating for Tfol cells.

Fig. S2. IL-2 increases the number of splenic Treg cells.

Fig. S3. Expression of CD25 on Tfr and Tfh cells in three different genetic backgrounds.

Fig. S4. Tfh and Tfr cells have similar expressions of Bcl6 and CD25.

Fig. S5. Clustering of Treg, Tfr, and Tfh cells based on the expression of the entire 545-gene set from the NanoString mouse immunology panel.

Fig. S6. Cytokines, CD40L, and OX40 expression of Tfr, Treg, and Tfh cells.

Fig. S7. IL-1R2 and OX40 expression of Tfr, Treg, Teff, and Tfh cells.

Fig. S8. IL-1β production in coculture of B and Tfh cells.

Fig. S9. GCB representative gating after INS or OVA immunization.

Table S1. Exact P values of the asterisk symbols shown in figures.

Source data (Excel file)


Acknowledgments: We are grateful to B. Gouritin for help in cell sorting, to P.-A. Vinot for advice regarding ELISA assays, and to W. Chaara for advice regarding gene expression analysis. We thank the Institut Curie Genomic Platform for advising and performing the NanoString experiment, G. Lebreton and colleagues from the Cardio-Thoracic Surgery Department of the Pitié-Salpétrière Hospital for providing the human tissues, F. Tankere from the Otolaryngology Department of the Pitié-Salpétrière Hospital for providing the human tonsils, and V. Kuchroo and B. Malissen for the mice provided. Funding: P.-G.G.R. is a doctoral fellow of the “Ecole de l’Inserm Liliane Bettencourt” and was sponsored by Servier. L.F. was funded by a “DIM Région Ile de France” doctoral fellowship. The work of D.K., E.M.-F., G.C., V.Q., G.F. and F.B. was funded by Assistance Publique–Hôpitaux de Paris, INSERM, Sorbonne Université–UPMC (Paris 6), as well as by LabEx Transimmunom (ANR-11-IDEX-0004-02) and European Research Council Advanced Grant TRiPoD (322856) (to D.K.). Author contributions: P.-G.G.R., G.C., F.B., G.F., and L.F. performed the mouse experiments. P.-G.G.R. performed analyses, including statistical analysis. V.Q. performed the experiment on human samples and analyses, including statistical analysis. P.-G.G.R., G.C., E.M.-F., and D.K. conceived the experiments. P.-G.G.R., E.M.-F., and D.K. wrote the manuscript, with input from all authors. D.K. conceived, supervised, and obtained funding for the entire study. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. Additional data related to this paper may be requested from the authors.
View Abstract

Navigate This Article