B Cells Drive Autoimmunity in Mice with CD28-Deficient Regulatory T Cells

This article is featured in In This Issue, p.3927

Forkhead box protein 3⁺ regulatory T cells (Tregs) are essential for immune homeostasis and prevention of autoimmunity (1). Constitutive loss of functional Tregs in humans or mice leads to severe autoimmunity shortly after birth (2–4). Moreover, induced loss of Foxp3⁺ Tregs in adult animals leads to rapid autoimmune inflammation in adult animals (5), demonstrating the importance of these cells for both developing and maintaining immune self-tolerance.

Thymic Treg development requires the integration of many signals from cell surface receptors, including the IL-2R and CD28 (6), which are important for effector T cell maturation. Loss of IL-2 or key component(s) of the IL-2R impairs Treg formation and leads to subsequent autoimmunity. Similarly, CD28^−/− mice exhibit a loss of Tregs (7), although the impairment of effector T cells occurring as a result of loss of the CD28 costimulatory pathway prevents the emergence of rampant autoimmunity (8).

As CD28 is required for normal Treg development, it was difficult if not impossible to examine its role in Treg maintenance and function using mice with constitutive deletion of CD28. Thus, we created animals with targeted loss of CD28 in Tregs [termed CD28-ΔTreg mice (9)]. Using these animals, we showed a T cell–intrinsic role for CD28 in mature Tregs, as CD28-ΔTreg mice exhibited a number of autoimmune features, including skin and lung inflammation.

Tregs have important roles in suppressing both cellular and humoral responses (10), and in particular, Tregs may directly suppress B cells and autoantibody generation to prevent autoimmune diseases (11, 12). Tregs are able to kill B cells or migrate to germinal centers (GCs) where they suppress Th cell–dependent B cell responses (13, 14). The likely mechanism for this mode of action is via a newly recognized Treg subset, named follicular Treg (T_FR) cells, which express CXCR5, Bcl-6, ICOS, and PD-1, share developmental cues with T follicular helper (T_FH) cells, and restrain GC reactions (10, 15–17).

The role of CD28 in Treg maintenance and function, together with the importance of T_FR cells in controlling GC reactions and Ab production, led us to determine how the loss of CD28 in Foxp3⁺ cells altered T_FR cell differentiation and function. In this study we show that CD28 is required for full T_FR cell differentiation as well as their optimal suppressive capacity. CD28-ΔTreg mice display elevated numbers of T_FH cells, reduced numbers of T_FR cells, and enhanced responses to Ag immunization. Moreover, in vitro assays that control for T_FR cell numbers demonstrated reduced T_FR function on a per-cell basis. Surprisingly, we found that although genetic ablation of B cells only had a minimal effect on the appearance of activated CD8⁺ T cells in CD28-ΔTreg mice, it prevented both the accumulation of activated CD4⁺ T cells and the occurrence of autoimmunity. These data reveal an unexpected contribution of B cells toward autoimmunity seen when CD28 is targeted on Tregs, and suggest a critical role of the T_FR–T_FH–B cell axis in this process.

Mice with conditional targeting of CD28 in Foxp3⁺ cells (CD28^fl/fl Foxp3^YFP-Cre, termed CD28-ΔTreg mice, and female CD28^fl/fl Foxp3^YFP-Cre/+) were generated and bred in our facility (9). B cell–deficient μMT B6 mice (18) were purchased from The Jackson Laboratory and bred to CD28-ΔTreg mice. All colonies were maintained under specific pathogen-free conditions. All experiments described in this manuscript were approved by the Institutional Animal Care and Use Committee at the Massachusetts General Hospital.

Cell cultures were performed using RPMI 1640 (Mediatech) supplemented with 10% heat-inactivated FBS, 100 U/ml penicillin, 100 μg/ml streptomycin, 2 mM l-glutamine, and 50 mM 2-ME (Sigma-Aldrich). Fluorescent anti-CD3, anti-CD4, anti-CD8α, anti-CD19, anti-CD25, anti-CD38, anti-CD44, anti-CD62L, anti-CD69, anti-CD80, anti-IgM, anti-IgG, anti–IL-4, and anti–IL-17 Abs were purchased from BioLegend. Anti–CTLA-4, anti–PD-1, anti-Fas, anti-ICOS, and anti-CXCR5 were purchased from BD Biosciences-Pharmingen. Anti-MHC class II (MHCII), anti-GL7, anti-IFN-γ, anti-IL21, and an anti-Foxp3 staining kit were purchased from eBioscience. BD Cytofix/Perm buffer was used for intracellular staining. For cytokine detection, cells were stimulated by Leukocyte Activation Cocktail (BD Biosciences) for 4–6 h and then fixed in eBioscience Fix/Perm buffer for intracellular cytokine staining. Cells were run on a BD LSR II flow cytometer or a Beckman Coulter Navios flow cytometer and analyzed by FlowJo software (FlowJo).

Freshly isolated tissues were fixed overnight at 4°C in 10% neutral buffered formalin. Tissues were then rinsed in 70% ethanol and sectioned by the Dana-Farber Harvard Cancer Center Pathology Core for H&E staining. Immunohistochemical analysis was performed at the Dana-Farber Harvard Cancer Center Rodent Histopathology Core described as previously (9). Tissue sections were incubated with Abs against mouse B220 and IgM (BD Biosciences), followed by biotinylated secondary Abs (Vector).

Briefly, 100 μg of nitrophenyl (NP)₁₈-OVA (Bioresearch Technologies) was suspended in a 1:1 H37RA CFA emulsion (Sigma) and injected s.c. into the left and right flanks. After 7 d mice were euthanized and inguinal draining lymph nodes (dLNs) and/or spleen were harvested for flow cytometric analysis or suppression assays (see below).

The in vitro suppression assay was performed as previously reported (19, 20). Briefly, cell populations were sorted to 99% purity on an Aria II flow cytometer. Sorted cells were counted on an Accuri cytometer (BD Biosciences) by gating live cells only. Then 5 × 10⁴ B cells and 3 × 10⁴ T_FH cells [stained by Cell Trace Violet (Invitrogen) in indicated experiments according to the manufacturer’s instructions] with or without 1.5 × 10⁴ T_FR from dLNs were cultured with 2 μg/ml soluble anti-CD3 (2C11; BioXcell) and 5 μg/ml anti-IgM (Jackson ImmunoResearch). After 6 d, cells were harvested and analyzed by flow cytometry. The proliferation of T_FH cells was detected by Ki67 staining or the dilution of Cell Trace Violet.

Comparison of means between groups used the two-tailed Student t test. Differences were considered statistically significant at p < 0.05.

CD28^fl/fl mice were bred with Foxp3^YFP-Cre mice to generate mice with a specific deletion of CD28 in Foxp3⁺ Tregs (CD28-ΔTreg mice) (9). As previously reported, from birth to ∼1 mo of age, CD28-ΔTreg mice had a relatively normal appearance. However, by 8–12 wk of age, CD28-ΔTreg mice developed signs of autoimmunity. This manifested as crusting eyelids, hair loss from the face and trunk, inflammatory infiltrates of the skin and lungs, and generalized lymphadenopathy and splenomegaly characterized by accumulation of activated (CD44^hiCD62L^lo) CD4⁺ and CD8⁺ T cells. These findings clearly implicated a Treg-intrinsic function for CD28 in normal immune homeostasis, but did not define the role of individual immune effector components, i.e., T and/or B cells. The role of the latter was of particular interest, given the recent discovery of T_FR cells and their crucial role in controlling humoral immune responses. Thus we examined the role of B cells in the autoimmune phenotype of CD28-ΔTreg mice.

Immunohistochemical staining revealed elevated numbers of B220⁺ B cells with larger follicle sizes, and enhanced IgG production in the follicles of enlarged LNs of 4 mo old CD28-ΔTreg mice (Fig. 1A–C). As CD28 is only deleted in mature Foxp3⁺ Tregs, we speculated that the abnormal B cell activation in CD28-ΔTreg mice might be attributed to a defect of T_FR cells in the absence of CD28 signaling. The percentage of B cells in LNs, but not spleen, was increased in inflamed 2 mo old CD28-ΔTreg mice (Fig. 2A, 2B). Because of the increased total cellularity in the dLNs of CD28-ΔTreg mice, as well as the relatively normal percentages of most other immune cells such as NK cells, myeloid cells, and dendritic cells (9), the enhanced B cell percentages and numbers indicate an actual and specific expansion of B cells (Fig. 2B). Concurrent with this, the serum IgG1 and IgG2a levels were increased in autoimmune CD28-ΔTreg mice (Fig. 2C), although we did not observe any alterations in anti-dsDNA Abs (Fig. 2D). Nevertheless, GC B cells (GL-7⁺ Fas⁺) were increased in CD28-ΔTreg mice, along with increases in class-switched IgG⁺IgM⁻, IgG⁺CD38⁺ effector/memory B cells, and plasma cells (Fig. 2E–I). Finally, we noted evidence of general B cell activation, as B cells in CD28-ΔTreg mice upregulated their surface expression of CD69, CD80, and MHCII (Fig. 2J). Collectively, the analysis of B cell phenotype indicates that CD28-deficient mice have increased B cell responses.

CD4⁺ T_FH cells migrate to B cell follicles to provide signals for B cell activation and Ab production (21, 22), whereas T_FR cells are the major regulatory population that control T_FH and B cell activation. Compared to control unimmunized wild type (WT) mice (CD28^+/+Foxp3^YFP-Cre mice), the LNs of 2 mo old CD28-ΔTreg mice showed significantly enhanced T_FH cell (CD4⁺Foxp3⁻CXCR5⁺ICOS⁺) percentages and numbers, with a reciprocal but nonsignificant reduction of T_FR percentages (Fig. 3A), suggesting that the reduced T_FR cell percentage and/or T_FR cell function in CD28-ΔTreg mice results in a failure to regulate T_FH cell homeostasis. The low percentage of T_FR cells in the control conditions without immunization prevented a robust comparison of T_FR cell differentiation. However, the ratio of T_FR to T_FH cells is known to be a decisive factor in the magnitude of humoral immune response (10, 19), and the reduced T_FR to T_FH cell ratio in the CD28-ΔTreg points to dysregulation of humoral immune responses. Thus, to further test the role of CD28 in T_FR cell differentiation, WT or CD28-ΔTreg mice were immunized by NP-OVA/CFA s.c. As shown in Fig. 3B, NP-OVA immunization induced expansion of both T_FH and T_FR cells in WT mice, but CD28-ΔTreg mice showed defective induction of T_FR cells in the dLNs.

As CD28-ΔTreg have an autoimmune phenotype, it remained possible that the resultant inflammation was the cause of defective T_FR cell development, rather than a cell-intrinsic role for CD28 in T_FR cells. To address this, we used female mice that were heterozygous for Foxp3-Cre (Foxp3^+/Cre). We have shown previously (Ref. 9 and data not shown), due to random X-chromosome inactivation roughly 50% of Tregs in CD28^fl/fl × Foxp3^+/Cre mice have deleted CD28 (and can be identified by YFP expression), whereas half retain CD28 and are YFP⁻. As shown in Fig. 3C and 3D, we observed significantly higher percentages of CD28⁺ T_FR cells compared with CD28⁻ T_FR cells in both naive and immunized mice. This is consistent with the data in Fig. 3B showing lower numbers of T_FR cells in CD28-ΔTreg mice and confirms a cell-intrinsic role for CD28 in T_FR cell development.

Specific cytokines, such as IL-4 and IL-21, are important in supporting humoral responses. Thus we next examined selected cytokine production in CD28-ΔTreg mice. Compared to WT control mice, we observed significantly elevated percentages of CD4⁺ cells from 3 mo old CD28-ΔTreg mice secreting IL-4 and IL-21, and to a lesser extent IL-17 and IFN-γ (Fig. 3E, 3F). As IL-21 is the major cytokine to direct the generation of T_FH cells, its enhanced expression in CD28-ΔTreg mice matched the increased humoral immune response (23, 24). In addition, B cells and IL-21 can activate and enhance the cytotoxicity of CD8⁺ cells (25, 26). Likewise, CD8⁺ cells in CD28-ΔTreg mice demonstrated enhanced IFN-γ production (Fig. 3G, 3H).

Previously, we demonstrated that global CD28^−/− mice had severely reduced percentages of T_FR cells (19). However, in this study we found that the reduction in T_FR cells with conditional deletion of CD28 on Tregs was less profound, despite the marked failure to regulate Ab production. Therefore, we next determined whether loss of CD28 alters T_FR suppressive capacity. To do this, we performed an in vitro T_FR suppression assay in which WT (CD28^fl/+ × Foxp3^YFP-Cre) and CD28-ΔTreg mice were immunized with NP-OVA. After 7 d, T_FH cells and B cells were sorted from control mice and cocultured in the presence or absence of T_FR cells from WT or CD28-ΔTreg mice along with anti-IgM and anti-CD3 for 6 d. We found that CD28-ΔT_FR cells suppressed the expression of Ki67, a cell cycle protein, and proliferation in T_FH cells to a much lesser degree compared with control T_FR cells (Fig. 4A, 4B). Moreover, T_FR cells from CD28-ΔTreg mice were less effective at suppressing class-switch recombination and proliferation in B cells compared with control T_FR cells, suggesting that CD28 costimulation is essential for the full suppressive capacity of T_FR cells (Fig. 4C). Consistent with previous studies (27, 28). CXCR5⁻ICOS⁻ Tregs from immunized WT (CD28^fl/+ × Foxp3^YFP-Cre) mice were far less efficient at suppressing class-switch recombination than were WT T_FR cells, showing the same modest level of suppression as was observed for CD28-deleted T_FR cells (Fig. 4D).

Previously, we have shown that CD28-deficient Tregs had reduced expression of CTLA-4 (9), a molecule known to regulate the proliferation and function of T_FR cells (29, 30). However, as shown in Fig. 4F, CD28-ΔT_FR cells express comparable levels of CTLA-4, CD44, CD25, and ICOS as WT T_FR cells. Therefore, alterations in CTLA-4 did not appear to be responsible for the reduced suppressive capacity of CD28-deficient T_FR. We have also shown that ICOS expression on T_FR cells can be used to identify T_FR cells that gain access to the B cell follicle (28). We did not observe a difference in ICOS expression in CD28-deficient T_FR cells (Fig. 4F). Using female mice that were heterozygous for Foxp3-Cre (Fig. 4E, see details of experimental system in Fig. 3), we observed diminished CXCR5 staining in CD28-deficient Tregs, with little or no difference in bcl-6 expression. These data show a cell-intrinsic role for CD28 in CXCR5 expression but suggest there is little role for CD28 signals in initiation of lineage-specific transcription factor expression.

To assess whether the skin autoimmunity in CD28-ΔTreg mice was due to enhanced B cell responses allowed by defective T_FR cell differentiation and function, we bred CD28-ΔTreg mice to μMT mice, which lack B cells (18). As shown in Supplemental Fig. 1, and consistent with previous studies (31, 32), loss of B cells led to a block of T_FH cell differentiation in CD28-ΔTreg × μMT mice. Strikingly, the lack of B cells rescued the severe autoimmunity seen in CD28-ΔTreg mice. This included the phenotype of grossly visible skin inflammation (Fig. 5A, indicated by arrows), lymphadenopathy (Fig. 5B), and inflammatory histology in the skin and lung, the main target organs of disease we observed in CD28-ΔTreg mice (9) (Fig. 5C).

Finally, we examined T cell activation in lymphoid organs of CD28-ΔTreg × μMT mice. B cell deficiency completely prevented the activation of CD4⁺ T cells, but not CD8⁺ T cells in the LNs and spleen of CD28-ΔTreg × μMT mice (Fig. 6A, 6B, Supplemental Fig. 2). Interestingly, although CD8⁺ effectors were not prevented, IFN-γ production by both CD4⁺ and CD8⁺ T cells was reduced (Fig. 6C, 6D). Collectively, these results indicate a dissociation between the role of B cells in mediating CD4⁺ versus CD8⁺ T cell activation in CD28-ΔTreg mice, and demonstrate that B cells are required for activation of CD4⁺ T cells and/or Ab production for the autoimmunity observed in these animals.

Previously we reported that cell-specific deletion of CD28 on Foxp3⁺ Tregs (CD28-ΔTreg mice) led to a systemic autoimmune disease characterized by activation of CD4⁺ and CD8⁺ T cells, with skin and lung inflammation (9). In this study, we further characterized these animals and found a striking elevation of serum Igs accompanied by increases in IgG class-switched and memory B cells, as well as GC B cells in the LNs of these animals.

It is well established that B cell activation in GCs is regulated by T_FR cells. This prompted us to examine how loss of CD28 in Tregs might specifically affect T_FR cells. We found a marked reduction in T_FR cells, both in unimmunized animals and following immunization. This suggested that the dysregulation of B cell responses observed in unmanipulated CD28-ΔTreg mice might be due to alterations in T_FR numbers. However, another, non–mutually exclusive possibility was that optimal T_FR function required CD28 and thus the ability of T_FR cells to control humoral responses was intrinsically defective, independent of cell number. To test this possibility, we turned to an in vitro assay of T_FR cell–suppressive capability, and showed that indeed, loss of CD28 led to an impairment in T_FR cell function.

The mechanism by which CD28 is required for optimal T_FR cell function remains under investigation. Previously, CD28 signaling in CD4⁺ T cells has been shown to induce the expression of both CXCR5 and ICOS (33, 34). Compared to WT T_FR, CD28-ΔT_regs did not demonstrate a significant reduction in the expression of ICOS. However, it is still plausible that CD28 signaling may regulate the coexpression of CXCR5 and ICOS in Tregs that might dictate T_FR cell induction upon immune stimulation.

Perhaps the most striking finding in our studies was the loss of aberrant CD4⁺ T cell activation and the prevention of clinical autoimmunity by genetic deletion of B cells, indicating a strong and unexpected B cell contribution to autoimmunity. Interestingly, previous work has demonstrated that the absence of B cells, or therapeutic depletion of B cells, reduces autoimmune pathology and prolongs survival in scurfy mice (35). In a related vein, a number of clinical trials have shown that B cell–depletional therapies ameliorate autoimmune diseases, such as rheumatoid arthritis and multiple sclerosis, which are believed to have predominant, if not exclusive, T cell pathogenesis (36–38). Collectively, our findings demonstrate a clear role for B cells in T cell–driven autoimmune inflammation. It seems likely that this role is via Ag presentation, although this remains to be formally demonstrated and other factors such as Ab or cytokine production cannot be completely excluded.

Interestingly, despite the fact that loss of B cells prevented overt autoimmunity from developing, we found a split effect on T cell activation. Specifically, whereas B cell depletion prevented the activation of CD4⁺ T cells, there was little to no effect on CD8⁺ T cell activation. This may indicate a differential or preferential reliance of CD4⁺ versus CD8⁺ T cells for types of APCs. Alternatively, the functional deficiency of CD28-ΔTregs may induce aberrant activation of CD8⁺ T cells.

In summary, our data reveal an essential role for CD28 in directing the development and function of T_FR cells to control B cell activation and humoral immune responses. In the autoimmune disease model of CD28-ΔTreg mice, reduction in T_FR cell differentiation and suppressive function results in broad dysregulation of B cell homeostasis. Importantly, genetic depletion of B cells in CD28-ΔTreg mice prevented CD4⁺ T cell activation and rescued the autoimmune phenotype. Although other Treg subsets may contribute, our findings suggest that alterations in T_FR cell differentiation and/or function result in systemic autoimmunity and skin inflammation. Future work will be needed to determine how promoting T_FR cell differentiation and function can be used to help treat autoimmune diseases including psoriasis.

We thank members of the Turka and Sharpe laboratories for many helpful discussions over the course of this work.

The authors have no financial conflicts of interest.