Inflammation and Lymphopenia Trigger Autoimmunity by Suppression of IL-2–Controlled Regulatory T Cell and Increase of IL-21–Mediated Effector T Cell Expansion

Regulatory T cells (T_regs) are essential for immune homeostasis and the maintenance of peripheral self-tolerance. Accordingly, perturbations of T_regs are associated with inflammatory (1) and autoimmune diseases (2). T_reg development and function is controlled by the transcription factor Foxp3. Its mutation or depletion results in fatal autoimmune lymphoproliferative disease in mice (3) and immunodysregulation polyendocrinopathy enteropathy X-linked syndrome in humans (4, 5). Thus, factors that adversely affect T_reg numbers, stability, function, or qualitative features and result in an imbalance of increased pathogenic effector T cells (T_effs) over T_regs may contribute to the development of autoimmune disease (1, 2).

A negative correlation between T_reg numbers and/or function and human autoimmune rheumatoid arthritis (RA) has been shown in previous studies (6). Likewise, studies in murine models of RA clearly indicate a beneficial role for T_regs in reducing disease development and severity (7–9). In humans and mice, T_regs infiltrate the synovial tissue of inflamed joints, as well as secondary lymphoid organs, and display suppressor functions (9–13).

The protective role of T_regs has also been demonstrated in the well-defined murine K/BxN ([KRNtgxNOD]F1) model of RA (11–13). K/BxN mice spontaneously develop arthritis at ∼4 wk of age. A break of tolerance in this model (14) leads to the activation of KRNtg (KRN TCR transgenic C57BL/6) expressing CD4⁺ T cells upon recognition of the autoantigen GPI that is presented in a complex with the MHC class II molecule Ag7. Interactions between KRNtg CD4⁺ T cells and GPI-specific B cells then result in the production of anti-GPI–specific IgG1 autoantibodies, which trigger synovial inflammation (15–17). T_regs are also expanded in joints and lymphoid organs of arthritic K/BxN mice, proliferate well, and display full functionality in suppressing T_effs in vitro (11, 13, 14). Moreover, the absence of T_regs in K/BxN mice carrying a scurfy mutation led to a more rapid and aggressive arthritis (11). However, despite T_reg enrichment at sites of inflammation, and their full functionality, K/BxN mice still experience development of disease. This may be because numerical and/or functional imbalances within the T_reg and T_eff compartments tip the balance toward autoimmunity. This has been partially explained by increased T_eff resistance toward T_reg-mediated suppression, as well as shorter proliferative bursts and higher apoptosis rates in T_regs (13). However, the factors and molecular mechanisms that influence T_reg and T_eff homeostasis in an autoimmune setting are poorly understood. Inflammation and lymphopenia are prevalent in autoimmunity and may influence T_reg/T_eff stability, functionality, and homeostasis (18). Given that both these conditions are present in arthritic K/BxN mice (19, 20), we used this model to study whether and how inflammation and lymphopenia influence T_reg stability and function and T_reg/T_eff homeostasis, and modulate autoimmunity.

We observed full functionality of T_regs isolated from inflamed organs of arthritic mice in in vitro suppression assays and were able to exclude the possibility of an unstable T_reg phenotype that would result in a conversion of T_regs into T_effs. However, we also observed that T_reg expansion was delayed relative to T_eff expansion (T follicular helper cells [T_FHs] and Th17 cells) in inflamed lymphoid tissues in early stages of disease. This likely resulted in acute unhindered T_eff activity, which may induce disease that, once initiated, rapidly becomes uncontrolled. Inflammation and lymphopenia in arthritic mice exacerbated this delay and further increased T_eff proliferation. Arthritic mice also had decreased availability of IL-2 and increased levels of IL-21. Neutralization and supplementation with IL-2 showed that T_reg expansion was dependent on this cytokine. Transfer of wild-type and IL-21R^−/− KRNtg T_effs and T_regs in various combinations showed that IL-21R^−/− deficiency mainly affected T_eff maintenance. Increases in thymic IL-7 and CD127 expression on thymic Foxp3⁺ single-positive (SP) CD8⁻CD4⁺ T cells indicated that enhanced T_reg frequencies may be because of increased IL-7–mediated thymic output and recruitment to sites of inflammation, where T_regs, however, fail to expand and exert their suppressive function effectively.

KRNtg mice were obtained from D. Mathis and C. Benoist (Harvard Medical School, Boston, MA), CD28^−/−.CD45.1 mice from C. Vinuesa (Australian National University, Canberra, ACT, Australia), and B6.H-2g7/g7 mice were purchased from The Jackson Laboratory, NOD/ShiLtJArc and CD45.1 congenic C57BL/6 (B6.SJL/ptprc^a) mice were obtained from the Animal Resources Centre (Perth, Australia), IL-21R^−/− mice from C. King (Garvan Institute of Medical Research), and Foxp3.gfp mice from C. Vinuesa. IL-21R^−/−.KRNtg and Foxp3.gfp KRNtg mice were generated by crossing IL-21R^−/− and Foxp3.gfp mice with KRNtg mice, respectively. CD28^−/−.Ag7^+/− (congenic for CD45.1) mice were generated by crossing CD28^−/−.CD45.1 mice with B6.H-2g7/g7 mice. Crossing KRNtg with NOD/Lt mice generates either KRNtg-expressing arthritic K/BxN (KRNtgxNOD)F1 mice or KRNtg⁻ and healthy littermates (BxN). Crossing B6.SJL/ptprc^a with NOD mice generates BxN.45.1 mice (B6.SJL/ptprc^a × NOD)F1. Foxp3.gfp KRNtg mice were bred to NOD mice to generate Foxp3.gfp K/BxN mice; because Foxp3 is on the X chromosome, only male Foxp3.gfp K/BxN mice were analyzed. When indicated, some of the strains were further crossed with B6.SJL/ptprc^a mice to generate CD45.1 or CD45.1.2 congenic mice. Genotypes were assessed by genomic PCR or FACS. Experiments were approved by the Garvan-St. Vincent’s and the Monash Animal Ethics Committees.

Cell suspensions were prepared from pooled spleens and lymph nodes (LNs). CD4⁺ T cells and subsets were isolated to a purity of >98% by negative selection using MACS microbeads (Miltenyi Biotech) according to the manufacturer’s instructions followed by FACS sorting. Purified CD4⁺ T cells were injected i.v. into recipient mice as outlined in the respective experiments. Neutralizing anti–IL-2 Abs (clone S4B6.1; WEHI, Melbourne) were injected i.p. at a concentration of 1 mg at day 0 of adoptive transfer; 250 μg neutralizing anti–programmed cell death-1 (anti-PD1; RMP1-14) and anti-PD1 ligand (anti–PD-L1; MIH5) Abs were injected i.p. at 0, 12, and 36 h after adoptive transfer; 2 μg recombinant mouse IL-2 (rmIL-2; Peprotech) was injected i.p. at −12, 0, 12, 24, and 36 h after adoptive transfer; 20 μg rmIL-21 (Peprotech) was injected i.p. at 2, 4, 6, 8, 10, and 12 d after adoptive transfer; and 100 μg LPS was injected i.p. in BxN.45.1 mice.

Measurement of arthritis development was done as previously described (21). In brief, clinical severity of arthritis was assessed every 1–2 d for all 4 paws on a scale from 0 to 3 and indicated as cumulative score: 0, normal; 1, erythema, swelling limited to individual digits or mild ankle swelling insufficient to reverse the normal V shape of the foot; 2, swelling sufficient to make the ankle and midfoot approximate in thickness to the forefoot; 3, reversal of the normal V shape of the foot, swelling of the entire paw including multiple digits.

Cells were collected from crushed spleen, peripheral LNs (pool of inguinal, brachial, axillary, and cervical LNs), or thymus. For intracellular cytokine and nuclear Foxp3 staining, cells were fixed and permeabilized with Foxp3 Fix/Perm Buffer Set (eBioscience) according to the manufacturer’s instructions. Intracellular cytokine expression was detected after restimulation of cells with 50 ng/ml PMA and 1 μg/ml ionomycin (Sigma-Aldrich) in the presence of brefeldin A and monensin (eBioscience). Cells were analyzed with BD LSRII and FACSCanto II. The following Abs were used: anti-CD4 V450 (BD), anti-CD4 PE-Cy7 (Biolegend), anti-CD103 FITC (BD), anti-CD25 PE-Cy7 (Biolegend), anti-Foxp3 allophycocyanin/FITC (eBioscience), anti-CD45.2 V450 (BD), anti-CD45.1 allophycocyanin-eFluor 780 (eBioscience), anti-CD162 PE (BD), anti-CD62L FITC (BD), anti-PD1 PE (eBioscience), anti-CXCR5 Biotin (BD), anti-streptavidin PerCP-Cy5.5 (eBioscience), anti-IgG1 FITC (BD), anti-CD45R/B220 allophycocyanin-eFluor 780 (eBioscience), anti-GL7 (T and B cell activation Ag) FITC (BD), anti-CD95 PE (BD), anti-CD196 (CCR6) allophycocyanin (BD), anti–IL-17 PE (BE), anti–IL-2 PE (eBioscience), anti–IL-21 biotin (R&D), streptavidin allophycocyanin (BD), anti–neuropilin-1 (anti–Nrp-1) allophycocyanin (R&D), anti-CD127 allophycocyanin (Biolegend), and anti-CD8 PerCP (BD).

Anti-GPI IgG1 levels were determined in serum by ELISA as previously described (22).

Total RNA was extracted using TRIzol reagent (Invitrogen), and the QuantiTect Reverse Transcription Kit (Qiagen) was used for cDNA synthesis according to the manufacturer’s instructions. Quantitative real-time PCR was performed using an ABI Prism 7900HT Real-Time PCR system (Applied Biosystems). For each sample, mRNA abundance was normalized to the amount of the housekeeping gene Gapdh and results expressed as arbitrary units.

FACS-purified T cell subsets were cultured in RPMI 1640 medium supplemented with 1% (v/v) l-glutamine, 1% (v/v) nonessential amino acids, 1% (v/v) sodium pyruvate, 1% (v/v) penicillin (10,000 U) and streptomycin (10,000 U; all from Life Technologies), 10% (v/v) FBS (Hyclone), and 55 μM 2-ME (Life Technologies). T_effs were stained with CellTrace Violet Component (Invitrogen) and 50,000 T_effs cocultured with T_regs at indicated ratios in the presence of 200,000 irradiated feeder splenocytes from NOD mice and anti-CD3 (clone 145-2C11, 2 μg/ml). Cell proliferation was determined 4 d later by FACS analysis.

For immunoblot analysis, nuclear extracts of thymi were prepared with NE-PER Nuclear and Cytoplasmic Extraction reagents (Thermo Scientific) according to the manufacturer’s instructions. Samples were analyzed by SDS-PAGE and transferred to a nitrocellulose membrane that was probed with either anti-histone 2B (Millipore, Billerica, MA) or anti–NF-κB p65 (Cell Signaling). The membranes were then probed with HRP-conjugated secondary Ab and developed with Western Lightning chemiluminescence reagent (Perkin Elmer, Boston, MA).

Statistical significance was determined by calculating p values using an unpaired t test on Instat software (GraphPad Software, San Diego, CA). All data are means ± SD and are representative of at least two independent experiments. The p values ≤0.05 were considered significant (*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001).

The concept that T_regs represent a stable T cell lineage has recently been challenged by several studies and remains controversial (23, 24). Conditions such as lymphopenia and inflammation have been suggested to drive T_reg instability and their conversion into T_effs (i.e., Th17 and T_FHs) (23–25). We investigated whether an unstable T_reg phenotype and conversion into T_effs could explain why increased numbers of T_regs in arthritis are unable to prevent the development of disease.

We first confirmed that T_reg frequencies are increased in secondary lymphoid organs (spleen and peripheral LNs) of arthritic K/BxN mice compared with KRNtg⁻ nondiseased BxN littermates (Fig. 1A). Higher expression of CD103 was found on arthritic K/BxN compared with naive BxN Foxp3⁺ T_regs, which indicates that they were effector/memory–like T_regs (Fig. 1A) (26). Interestingly, arthritic K/BxN T_regs had a higher proportion of CD25⁻Foxp3⁺ T_regs (Fig. 1A). This subset was recently reported to be less stable in a lymphopenic environment (23, 27). Moreover, it was shown previously that a proportion of CD25^lo Foxp3⁺ cells lose Foxp3 expression and acquire a pathogenic Th17 phenotype within the inflammatory milieu in arthritis (2). We therefore compared nonarthritic with arthritic T_regs in BxN and K/BxN mice, respectively. In arthritic T_regs we did not find increased IL-17 or IFN-γ expression (Fig. 1B), whereas CCR6 was increased (Fig. 1B).

To examine the functionality of arthritic versus nonarthritic T_regs, we isolated Foxp3.gfp⁺ CD4⁺ cells from arthritic Foxp3.gfp K/BxN and naive Foxp3.gfp KRNtg mice, respectively, and used them in suppression assays. As shown in previous studies (13), arthritic effector T_regs were as efficient as naive nonarthritic T_regs in suppressing T_eff proliferation (Fig. 1C).

To address whether T_regs are unstable and can readily convert into T_effs, we performed adoptive transfer experiments of Foxp3.gfp⁺ CD4⁺ cells isolated from Foxp3.gfp KRNtg mice. Because lymphopenia and inflammation have been shown to influence T_reg stability, we assessed whether this occurs in arthritis by using the following recipient animals: (i) lymphoreplete naive BxN.45.1 mice, (ii) quasi-lymphopenic CD28^−/−.Ag7^+/− mice, and (iii) inflammatory arthritic K/BxN.45.1 mice. Mice were sacrificed 14 d after transfer, and adoptively transferred cells were examined for expression/downregulation of GFP in spleen and LN. Under all conditions, a small fraction (1.27–9.44%) of transferred Foxp3.gfp⁺ KRNtg T_regs had become GFP⁻. The greatest frequency of GFP⁻ cells was found in the LN of quasi-lymphopenic CD28^−/−.Ag7^+/− mice (Fig. 2A, 2B). Moreover, the examination of absolute numbers of recovered T_regs showed that T_reg expansion is generally hampered in the respective lymphopenic and/or inflammatory milieu of CD28^−/− and K/BxN mice (Fig. 2C).

We next addressed whether Foxp3.gfp⁺ T_regs were able to adopt a T_eff (P-selectin glycoprotein ligand 1 [PSGL1]^lo, CD62L^lo, PD1^hi, CXCR5^hi) or effector/memory T_reg (CD103) phenotype and to exert T_eff function (germinal center [GC] formation, anti-GPI IgG1 production, and arthritis development). Highly purified Foxp3.gfp⁻ KRNtg T_effs or Foxp3.gfp⁺ KRNtg T_regs from Foxp3.gfp KRNtg mice were transferred into BxN.45.1 or CD28^−/−.Ag7^+/− recipients. CD103 was almost exclusively upregulated on transferred Foxp3.gfp⁺ KRNtg T_regs (Fig. 3A, 3B). In contrast, only transferred Foxp3.gfp⁻ KRNtg T_effs displayed significant downregulation of CD62L and PSGL1, and higher levels of the T_FH markers CXCR5 and PD1 (Fig. 3A, 3B). That CXCR5 and PD1 were also expressed on some Foxp3.gfp⁺ KRNtg T_regs is in accordance with the recent description of a subset of follicular T_regs sharing T_FH phenotypic characteristics (28). Apart from CXCR5 and PD1 expression, T_regs or T_effs displayed a comparable phenotype when transferred into CD28^−/−.Ag7^+/− or BxN.45.1 mice (Fig. 3A, 3B). Generally, in CD28^−/−.Ag7^+/− mice, a greater proportion of transferred T_effs adopted a CXCR5^hiPD1^hi T_FH phenotype (Fig. 3A, 3B). Adoptively transferred Foxp3.gfp⁺ KRNtg T_regs did not induce an autoaggressive B cell response (GL7^hiFas^hi GC B cells and IgG1 class-switched B cells; Fig. 3C), anti-GPI IgG1, or the development of arthritis (Fig. 3D). Collectively, our data show that a small subset of cells may have converted from T_regs to T_effs in arthritis but did not induce an autoaggressive B cell response. The proportion of converted cells among transferred T_regs, if conversion did indeed occur, is likely to be too small to induce disease.

We next assessed whether the observed conversion of small numbers of T_regs into T_effs may have been, as hypothesized recently (24), simply an artifact due to the outgrowth of a small number of contaminating non-T_regs. This seems feasible because it is impossible to sort GFP⁺ CD4⁺ T cells to a purity of 100%. Thus, we adoptively transferred highly purified CD45.1.2⁺ Foxp3.gfp⁻ KRNtg T_effs together with CD45.2⁺ Foxp3.gfp⁺ KRNtg T_regs into CD45.1⁺ BxN.45.1 recipients at a ratio of 4:1 (CD45.1.2⁺GFP⁻:CD45.2⁺GFP⁺). Replicate groups of mice received ∼99% pure CD45.2⁺ Foxp3.gfp⁺ KRNtg or ∼94% pure CD45.2⁺ Foxp3.gfp⁺ KRNtg cells. Four days after transfer, T_effs had proliferated to a greater extent than T_regs (2.49 and 4.33 times more in spleen and LN, respectively) as determined by the ratio of CD45.1.2⁺ Foxp3.gfp⁻ KRNtg T_effs:CD45.2⁺ Foxp3.gfp⁺ KRNtg T_regs, compared with day 0 (Table I and Supplemental Fig. 1A, 1B). Mathematically the increased proliferation rate of T_effs making up ∼1 or 6% of CD45.2⁺ Foxp3.gfp⁺ KRNtg T_regs would translate into a frequency of ∼2.5 or ∼15% CD45.2⁺ GFP⁻ cells in spleen and ∼4.3 or ∼26% CD45.2⁺ GFP⁻ cells in LN. These values were close to the actual percentage of CD45.2⁺ GFP⁻ cells determined in spleen (∼4.5 ± 0.36 or ∼16.1 ± 2.28%) and LN (∼5.8 ± 0.91 or ∼29.1 ± 3.46%) by FACS. This shows that a small population of outgrowing non-T_regs can proliferate and be misinterpreted as unstable and converted T_regs. Thus, these results suggest that the contamination of GFP⁻ cells could account for the observed unstable phenotype of T_regs. Conversion of T_regs into T_effs is not supported by our data and does not explain why numerically expanded T_regs are unable to prevent the development of arthritis.

Although there is an increase in T_regs in arthritic 8-wk-old K/BxN mice (Fig. 1A), it is possible that a T_reg/T_eff imbalance resulting in a numerical T_eff superiority may occur at presymptomatic stages and induce the development of disease. Thus, we examined the induction of T_regs (CD4⁺Foxp3⁺ cells) and T_effs (CCR6⁺IL-17⁺ Th17 cells and CXCR5^hi PD1^hi T_FH) during different stages of disease development in arthritic K/BxN mice.

T_regs and T_effs (T_FH and Th17 cells) were assessed: (i) when they arrive in the periphery at 17 d of age and as clonal depletion becomes incomplete; (ii) in prearthritic mice at 22 d and when peripheral CD4⁺ T cells in spleen and LN appear in significant numbers; (iii) during acute arthritis and robust inflammation at 5 wk; (iv) through transition from acute into chronic arthritis at 8 wk; and (v) during chronic arthritis at 12 wk of age (20). T_reg frequencies remained the same (spleen) or were reduced (LN) at day 17 during disease onset, and increases at both sites only occurred from day 22 (Fig. 4A, 4B, and Supplemental Fig. 2A, 2B). From then T_reg frequencies plateaued. In line with T_reg frequencies, CD103 upregulation on Foxp3⁺ CD4⁺ T also occurred from day 22 (Fig. 4C). In contrast, T_effs (Th17 and T_FH cells) were markedly increased at both 17 and 22 d in spleen and LN, and then plateaued from 5 wk of age compared with healthy BxN mice (Fig. 4A, 4B, and Supplemental Fig. 2A, 2B). These results show that at early preclinical disease stages, altered T_reg/T_eff ratios result in T_eff superiority.

We next assessed whether delayed T_reg compared with T_eff expansion would also occur in adoptively transferred KRNtg CD4⁺ T cells and how the proliferation of T_effs is influenced by a lymphopenic environment. To do this, we injected naive KRNtg CD4⁺ T cells containing ∼10% naive Foxp3⁺ T_regs into lymphoreplete BxN.45.1 (Fig. 4D, 4E, and Supplemental Fig. 2C) or quasi-lymphopenic CD28^−/−.Ag7^+/− mice (Fig. 4F and Supplemental Fig. 2D), and examined T_eff (T_FH and Th17 cells) and T_reg differentiation at 4, 8, and 14 d after transfer. Th17 differentiation was barely detectable in adoptively transferred cells (data not shown); therefore, we focused on T_FH differentiation (CXCR5^hiPD1^hi). As in K/BxN mice, increases in T_regs and CD103⁺ T_regs were not observed until later (day 8) in BxN.45.1 mice, whereas T_FHs increased early (day 4; Fig. 4D, 4E, and Supplemental Fig. 2C). The discrepancy between T_reg and T_FH expansion was even more dramatic in quasi-lymphopenic CD28^−/−.Ag7^+/− mice (Fig. 4F, Supplemental Fig. 2D), which may partly account for the observed more severe disease course (Supplemental Fig. 2E). Contrary to truly lymphopenic mice (RAG^−/−, CD3e^−/− mice), CD28^−/− mice have normal lymphoid cellularity and tissue organization. Nevertheless, it is likely that transferred TCR-responsive CD4⁺ T cells behave in a similar way when transferred into quasi-lymphopenic CD28^−/− mice or truly lymphopenic animals. Lack of competition by TCR-unresponsive endogenous CD4⁺ T cells, reduced IL-2 levels, and a general T_reg defect (29, 30) may account for a generally more vigorous expansion of transferred Foxp3⁻ T_effs over T_regs in CD28^−/− mice.

Thus, decreased T_reg compared with T_eff (T_FH and Th17 cells) expansion was observed during early disease stages in K/BxN mice, as well as in adoptive transfer experiments, and was exaggerated under quasi-lymphopenic conditions. We propose that an increased autoreactive T_eff to T_reg expansion during early disease results in the uncontrolled outgrowth and activity of T_effs that dominate T_regs, and is sufficient to tip the balance of control and induce autoimmunity. The delayed increase of T_regs may then only suppress, but no longer contain, disease progression.

Although T_reg expansion was delayed at early disease stages compared with T_effs, the frequencies of both cell types were still substantially increased in secondary lymphoid organs of arthritic K/BxN compared with nondiseased BxN mice (Fig. 1A). Thus, we next examined whether and how the inflammatory milieu might influence T_eff and T_reg homeostasis. CD4⁺ T cells were isolated from KRNtg mice, CFSE-labeled, and adoptively transferred into naive BxN.45.1 or arthritic (i.e., inflamed) 6-wk-old K/BxN.45.1 recipient mice, and proliferation was assessed 60 h later. The proliferation of both T_effs and T_regs was substantially reduced on transfer into arthritic K/BxN mice (Fig. 5A–C). However, T_regs were more affected than T_effs (Fig. 5A, 5B, 5D).

We then assessed the role of IL-2 within the inflamed milieu, because this cytokine is particularly required for T_reg homeostasis in the periphery (31). Its availability may be decreased during inflammation and lymphopenia, which would account for the reduced T_reg/T_eff proliferation and particularly the inhibition of T_reg expansion in inflamed tissues. We first examined IL-2 expression by splenocytes of arthritic K/BxN compared with healthy BxN mice (Fig. 6A). IL-2 levels displayed a trend toward decreased expression by arthritic K/BxN CD4⁺ T cells as examined by FACS and quantitative PCR on sorted CD4⁺ T cells (Fig. 6A). We were unable to detect IL-2 in serum or plasma of arthritic K/BxN or nonarthritic BxN mice by ELISA. In view of the lymphopenia (Fig. 6B) and reduced IL-2 expression by CD4⁺ T cells in arthritic K/BxN mice (Fig. 6A), an IL-2 deficit appears, however, to be conceivable. To further assess the effects of IL-2 on early T_eff and T_reg proliferation, we transferred CFSE-labeled KRNtg CD4⁺ T cells into arthritic K/BxN.45.1 mice treated with 2 μg rmIL-2 at −12, 0, 12, 24, and 36 h. Treatment of recipient mice markedly increased the proliferation of T_regs and, to a lesser extent, T_effs (Fig. 5A–C). We next assessed whether IL-2 availability could also influence the disease course. KRNtg CD4⁺ T cells were adoptively transferred into nonarthritic BxN.45.1 recipients that were treated with 1 mg neutralizing anti–IL-2 Abs at day 0 (clone S4B6.1). In accordance with the favorable effects of IL-2 on T_reg proliferation, neutralization of IL-2 resulted in more severe disease (Fig. 6C). Unfortunately, we were unable to directly mitigate disease courses in K/BxN mice by injection of rIL-2 (data not shown). This may be the result of a whole range of explanations, including suboptimal dose and route of administration, the IL-2 not reaching the correct sites or cells, or that disease in K/BxN mice was too strong to be reversed with this treatment. Additional studies will be required to explore these issues. Regardless, IL-2 promotes T_reg proliferation and the suppression of arthritis. However, its levels are reduced in arthritis and, therefore, other pathways presumably account for the increased numbers of T_regs in this disease model.

In contrast with IL-2, IL-21 levels may be increased because of the infiltration of the main IL-21 producers, T_FH, into inflamed organs of K/BxN mice (Fig. 4A, 4B) (32). This cytokine is important for T_eff homeostasis in K/BxN mice (19), and it has inhibitory effects on T_reg function (33–37). Thus, the role of IL-21 in T_eff and T_reg homeostasis in arthritis was investigated. We found that IL-21 expression by splenocytes of arthritic K/BxN mice was increased compared with healthy BxN mice as examined by FACS and quantitative PCR on sorted CD4⁺ T cells (Fig. 6D). To further assess the effects of IL-21, we adoptively transferred CFSE-labeled IL-21R^−/− Foxp3⁺ and Foxp3⁻ KRNtg CD4⁺ T cells into nonarthritic BxN.45.1 and arthritic K/BxN.45.1 mice, and the proliferation of both transferred cell types was examined 60 h later (Fig. 5A–C). IL-21R deficiency had minimal effects on the proliferation of both transferred Foxp3⁻ T_effs and Foxp3⁺ T_regs (Fig. 5A–C). We then assessed whether IL-21 could modulate the maintenance of T_effs or T_regs, rather than early proliferation. T_effs and T_regs were sorted from CD45.2⁺ IL-21R^−/− KRNtg and CD45.1.2⁺ KRNtg mice as CD25⁻CD4⁺ T_effs and CD25^hiCD4⁺ T_regs, respectively, because these lines were not backcrossed on Foxp3.gfp reporter mice. We then transferred KRNtg T_effs and IL-21R^−/− KRNtg T_regs, KRNtg T_effs and KRNtg T_regs, or IL-21-R^−/− KRNtg T_effs and KRNtg T_regs (1:10 T_reg/T_eff ratio) into BxN.45.1 mice. After 14 d posttransfer there was a marked increase in the T_reg/T_eff ratio among transferred cells in mice that had received IL-21R^−/− T_effs compared with other groups (Fig. 6E). In accordance with this, these animals also displayed minimal disease (Fig. 6E). That IL-21 influences disease progression by modulation of T_reg and T_eff homeostasis was further supported by the observation that injection of IL-21 into BxN.45.1 mice transferred with KRNtg CD4⁺ T cells triggered disease. This was accompanied by increases in the percentage of recovered Foxp3⁻ T_effs at the expense of Foxp3⁺ T_regs (Fig. 6F).

Thus, we observed that the opposing regulation of IL-2 and IL-21 within inflamed tissue combines to impair T_reg proliferation and increase T_eff maintenance, which promoted arthritis.

Because IL-2 does not fully reconstitute hampered T_reg/T_eff proliferation and IL-21R deficiency only slightly affected the early proliferation of T_effs and T_regs, other factors within the inflammatory milieu may influence their expansion. Recently, PD1 and PD-1 ligand (PD-L1) have been described as negative regulators of T_reg function (38). In chronically infected hepatitis C virus patients, T_regs expressed higher levels of PD1 than T_effs; moreover, PD-L1 levels were increased at inflammatory sites. This inhibited T_reg expansion to a greater extent than T_eff expansion and made them more susceptible to PD1-dependent exhaustion (38). We therefore assessed whether, in our model of autoimmune arthritis, enhanced PD1–PD-L1 signaling could play a similar role. PD1 expression was higher on T_regs than T_effs both in naive BxN and arthritic K/BxN mice (Supplemental Fig. 3A). CD11b⁺ macrophages/monocytes and CD11c⁺ DCs, in arthritic K/BxN mice, as well as CD19⁺ B cells were largely expanded (Supplemental Fig. 3B). DCs and B cells displayed a significant upregulation of PD-L1 expression (Supplemental Fig. 3C). We therefore propose that PD-L1 levels are generally increased within inflamed organs of arthritic mice and may contribute to the hampered expansion of T_regs because of their high PD1 expression. However, we were not able to improve T_reg proliferation in K/BxN recipients by treatment with neutralizing anti-PD1 or anti–PD-L1 Abs in vivo (Supplemental Fig. 3D). This does not fully exclude a role for PD1–PD-L1; however, it is likely to be minor relative to IL-2/IL-21 or other as yet unidentified factors.

The proliferation of T_regs was decreased in the inflammatory and lymphopenic environment of arthritic mice, and it remains unknown how the numbers of these cells are increased in this disease. We investigated whether enhanced thymic output followed by recruitment of T_regs to inflamed organs could account for the increase. Foxp3 expression was assessed on SP CD8^-CD4⁺ thymocytes of 16-d-, 22-d-, and 8-wk-old arthritic K/BxN compared with naive BxN mice. In naive BxN mice, Foxp3 expression on SP CD4⁺ thymocytes was comparable between young (16 and 22 d) and adult mice (8 w; Fig. 7A). In contrast, in arthritic mice, Foxp3 expression on SP CD4⁺ thymocytes was significantly increased at 22 d and further increased at 8 wk (Fig. 7A). Hence T_reg frequencies in the thymus (Fig. 7A) correlate well with T_reg frequencies in lymphoid organs (Fig. 4A, 4B). Nrp-1 is a surface molecule that was recently identified as a reliable marker to distinguish natural T_regs (nT_regs) arising in thymus from inducible T_regs (iT_regs) generated in the periphery through the induction of Foxp3 (39). We found a higher percentage of Nrp-1–expressing T_regs in secondary lymphoid organs of arthritic K/BxN compared with healthy BxN animals (Fig. 7B). Moreover, the absolute cell number per lymphoid organ of Nrp-1⁻ iT_regs was comparable between arthritic K/BxN and healthy BxN mice, whereas Nrp-1⁺ nT_regs were increased in arthritic animals (Fig. 7C). This supports our assertion that the increase in T_regs may be mainly because of an increased thymic T_reg output.

We next investigated which factors may mediate increased thymic T_reg production. The common γ-chain cytokines IL-2, IL-7, and IL-15 are important for thymic T_reg homeostasis (31, 40–42). Increased NF-κB signaling in thymocytes may function as sensor of peripheral inflammation and increase the secretion of TNF-α, inducing the development of T_regs in the thymus (43). Thus, the mRNA expression of IL-2, IL-7, IL-15, and TNFα was determined in the thymi of 8-wk-old healthy BxN and arthritic K/BxN mice (Fig. 8A). To examine NF-κB signaling, extracts were isolated from thymi and assessed by immunoblot analysis (Fig. 8B). There were no differences in the expression of IL-2, IL-15, or TNFα or in the production of NF-κB between healthy BxN and arthritic K/BxN mice. However, there was an upregulation of IL-7 in the thymi of arthritic mice (Fig. 8A, 8B). In accordance with this, Foxp3⁺ SP CD8⁻CD4⁺ thymocytes in arthritic K/BxN mice expressed higher levels of IL-7R (CD127) compared with naive BxN mice (Fig. 8C).

These observations may point toward a potential role of IL-7 signaling in the regulation of thymic T_reg homeostasis (42). Because IL-7 signaling may be influenced by inflammatory cells or lymphopenia (44, 45), we compared 3-wk-old prearthritic and noninflamed animals with 8-wk-old arthritic mice. The arthritic 8-wk-old animals showed a substantially less pronounced lymphopenia in secondary lymphoid organs (Fig. 8D). They also exhibit a trend toward thymic upregulation of IL-7 mRNA, and elevated CD127 levels (Fig. 8D) were detected on their SP CD4⁺CD8⁻Foxp3⁺ thymocytes (Fig. 8D). The fact that 8-wk-old animals display less severe lymphopenia but arthritic inflammation compared with 3-wk-old prearthritic mice suggests that inflammation rather than lymphopenia may drive those changes (Fig. 9).

To further examine possible effects of inflammatory signals, we injected nonarthritic BxN.45.1 animals with LPS and analyzed thymi after 12 h by FACS (Supplemental Fig. 4A) and after 4 h by quantitative PCR (Supplemental Fig. 4B). In support of our previous data, we detected increased CD127 expression on SP Foxp3⁺CD8⁻CD4⁺ thymocytes (Supplemental Fig. 4A). Changes in thymic IL-7 mRNA expression levels (Supplemental Fig. 4B) or increases in the percentage of Foxp3⁺ among SP CD4⁺CD8⁻ thymocytes (Supplemental Fig. 4A) were not detectable. The increased CD127 expression on CD4⁺CD8⁻Foxp3⁺ thymocytes further indicates a possible regulatory effect of inflammatory signals on thymic IL-7–mediated T_reg differentiation that is mediated by the receptor for this cytokine. That LPS treatment did not affect thymic IL-7 expression or increase the percentage of Foxp3⁺ among SP CD4⁺CD8⁻ thymocytes may be because of a whole range of explanations such as the choice of time point, inflammatory stimulus, or the duration of inflammation, which needs to be thoroughly examined in future studies. Moreover, it will be interesting to address mechanistic details of thymic T_reg regulation, output, and migration to inflamed sites in future studies, as well as the influence of inflammatory stimuli and common γ-chain cytokines.

In this article, we show that dysregulation of thymic and peripheral T_reg and T_eff homeostasis drives the induction and progression of autoimmunity in the KRN model of murine RA. We show that a short period of insufficient T_reg control of autoreactive T_effs (T_FH and Th17 cells) may allow the onset of autoimmunity that then becomes uncontrolled. The inflammatory milieu generated after disease initiation may then further deregulate T_reg/T_eff homeostasis and exacerbate the disease course. We identified the cytokines IL-2 and IL-21 as key players in disturbed peripheral T_reg and T_eff homeostasis. A lymphopenia-induced IL-2 deficit particularly affected T_reg expansion, and at the same time, increased IL-21 responses in the inflammatory milieu supported the maintenance of T_effs. Moreover, our data indicate that inflammation may mediate an enhanced IL-7–mediated thymic nT_reg output that could explain their increased frequencies at inflammatory sites. However, once they arrive in the periphery, T_regs fail to expand adequately, which limits their suppressive effects that are necessary to control expanding autoaggressive T cells (Fig. 9).

Our study highlights the importance of IL-2 and IL-21 as modulators of peripheral T_reg and T_eff homeostasis, and clearly demonstrates how opposing IL-2/IL-21 regulation promotes autoimmunity. We found decreased IL-2 and increased IL-21 responses in arthritic CD4⁺ T cells of K/BxN compared with healthy BxN mice. IL-2 deficiency particularly affected T_reg proliferation because IL-2 administration triggered T_reg expansion, whereas IL-2 neutralization promoted more severe disease. Hence our data support the concept that IL-2 is particularly required for the homeostasis and survival of T_regs in the periphery (46). Accordingly, another study showed that T_regs are the first to respond to IL-2, and IL-2 signaling particularly favored T_reg activity rather than promoting T_eff expansion (47).

Like IL-2, IL-21 is primarily produced by activated CD4⁺ T cells, but its activity opposes the function of IL-2 (32). In terms of T_regs, IL-21 has been postulated to counteract T_reg suppression (33, 34), inhibit their de novo production (35, 36, 48), and negatively regulate T_reg homeostasis through a feedback loop where it reduces IL-2 production from T_effs (37). In addition, IL-21 has been shown to promote T_eff activity (37). We did not detect any inhibitory effects of IL-21 signaling on T_reg homeostasis. Also, IL-21R deficiency did not affect the early proliferation of T_effs; instead, it did reduce late T_eff expansion and mitigated disease. Accordingly, IL-21 administration triggered arthritis and increased the ratio of T_effs to T_regs among transferred autoaggressive cells. Mechanistically it is possible that the inductive effects on T_eff homeostasis are due to the fact that IL-21 may confer survival signals to T_effs or promote their increased resistance to T_reg-mediated suppression (33, 34), as reported in animal (13, 33, 34, 49) and human (50, 51) studies of autoimmunity. Data from in vitro suppression assays that the functionality of arthritic T_regs was equal (our studies) or increased (13) compared with naive T_regs may argue against direct inhibitory effects of IL-21 on the function of these T_regs. It seems most likely that the autocrine secretion of IL-21 enables activated CD4⁺ T cells to expand and eventually overwhelm T_reg suppression. Given the frequent involvement of IL-21 in autoimmunity (32) and its predominant secretion by infiltrating Th17 and T_FH cells (35, 36, 52), such a scenario may be a general phenomenon that contributes to the pathogenesis of autoimmunity.

Inflammatory cytokines, especially IL-21, are also implicated in the induction of an unstable T_reg phenotype. That Foxp3-expressing T cells represent a stable, terminally differentiated lineage has recently been questioned by a number of studies, where it was proposed that T_regs, or a subpopulation of them, retain developmental plasticity (2, 23, 27, 53, 54). In an autoimmune setting, inflammatory signals or a lymphopenic environment could destabilize the T_reg phenotype and allow for functional plasticity and reprogramming into Th1, Th17, or T_FH effector cells (23, 27, 53–59). T_reg instability was, however, not supported by all studies (24) and remains controversial. In our study, we did not observe a noticeable conversion of T_regs into T_effs, nor were T_regs able to adopt T_eff (T_FH) function that would result in GC formation, autoantibody production, and disease development. Using mixed adoptive transfer studies, we showed that a relatively stronger expansion of T_effs (T_FH and Th17 cells) together with the possibility of the outgrowth of T_effs containing isolated T_regs can be misinterpreted as T_reg to T_eff conversion. Factors, such as inflammatory cytokines or lymphopenia that support a stronger T_eff expansion, increase the number of “pseudo-converted cells.” Accordingly, other studies indicated that T_reg instability was particularly observed in lymphopenic hosts (53) and within the gut microenvironment (Peyer’s patches and mesenteric LN) (54, 55). This is in accordance with our own observations where the greatest numbers of pseudo-converted cells were in CD28^−/− recipients. It is likely that reduced IL-2 availability under lymphopenic conditions hampers T_reg proliferation, whereas high levels of IL-21 in gut-associated lymphoid organs may support expansion of T_effs, and hence contaminating T_effs outproliferate T_regs.

Although it has been shown that CD25^hi T_regs retain their Foxp3 expression (2, 24), CD25^loFoxp3⁺ T_regs were identified as an unstable subset (23, 27). Moreover, under the inflammatory conditions of collagen-induced arthritis, such a cell population was able to acquire a Th17 phenotype induced by IL-6 (27). However, this study also showed that these exT_reg Th17 cells may originate from iT_regs rather than nT_regs, with only the latter being increased in inflamed lymphoid organs in our studies. We did not find increased IL-17 or IFN-γ, but CCR6 was elevated on arthritic T_regs, what they may have acquired to facilitate migration to inflamed tissues.

In our studies of T_reg stability, it has to be taken into consideration that Foxp3.gfp reporter mice were used. For our questions, the Foxp3.gfp reporter mouse model has been a critical tool because sorting of T_reg cells at high purity was required. However, recent reports have indicated that the reporter may change autoimmune severity (60, 61). For our study, we would anticipate the reported phenotypic, functional, and transcriptional perturbations (60, 61) to have little or no impact, although this remains untested.

Under normal circumstances, the scenario that T_reg expansion follows that of T_effs with some delay has physiological importance. To boost their fitness and expansion, T_regs rely on adequate T_eff activation (62, 63). For this reason, it has been proposed that T_reg responses do not arrest a primary immune response, but rather track the T_eff response (64) to starve them of survival factors, such as IL-2 or TNF-α (47, 62, 63). This enables necessary immune responses, which are then controlled to prevent overexuberant immune activation. Under conditions that favor T_eff proliferation (inflammation and lymphopenia), there is a dysregulation in the T_reg/T_eff balance, and the T_reg response may be overwhelmed by T_effs, resulting in autoimmunity. This would rely on a more potent expansion of T_effs in response to inflammation or lymphopenia at different disease stages. At early or predisease stages, the resulting unopposed T_eff activity may tip the balance toward autoimmunity; during established disease, it may underpin phases of remittent or relapsing disease.

In summary, we demonstrate how imbalances in T_reg and T_eff homeostasis control the onset and progression of disease in RA. Given that states of inflammation or lymphopenia are a relatively common occurrence throughout life, and may also result from therapeutic regimens, it is necessary to elucidate factors and mechanisms that restore homeostasis or augment T_reg activity. Because IL-2 acts as a T_reg enhancer outweighing its simultaneous fueling of the T_eff pool, and IL-21 confers maintenance signals to T_effs, they may become new targets for novel attempts to manipulate immunological tolerance, or effector responses, to subdue or enhance immune responses. Such a concept may not only have significant implications for therapeutical use in RA, but also in other autoimmune and inflammatory diseases and cancer immunotherapy.

We thank D. Mathis and C. Benoist for KRNtg, A. Rudensky and C. Vinuesa for Foxp3.gfp reporter, J. Cyster and C. King for IL-21R^−/− mice, the MLC Garvan Flow Cytometry Facility and Monash Flow Cytometry Facility for help with cell sorting, staff of the Garvan and Monash animal facilities and the Australian Bioresources facility, the Garvan Australian Cancer Research Foundation Facility for help with genotyping, and Linda Mason and Florence Lim for help with animal care and transfer and technical support.

The authors have no financial conflicts of interest.