Homeostatic maintenance of natural Foxp3⁺ CD25⁺ CD4⁺ regulatory T cells by interleukin (IL)-2 and induction of autoimmune disease by IL-2 neutralization

CD25⁺ CD4⁺ CD8⁻ cells in the thymus and spleen of normal naive mice constitutively expressed IL-2Rβ (CD122) as well as IL-2Rγc chain (CD132), whereas CD25⁻ CD4⁺ CD8⁻ cells expressed only the latter, indicating that CD25⁺ CD4⁺ T cells constitutively express the high affinity IL-2R even in the thymus (Fig. 1 A).

To determine the effects of IL-2 neutralization on the composition of CD25⁺ or CD25⁻ CD4⁺ T cells in the thymus and the periphery, we i.v. administered 1 mg of neutralizing anti–IL-2 mAb to normal BALB/c mice and assessed the number of CD25⁺ or CD25⁻ CD4⁺ CD8⁻ thymocytes/T cells. The treatment reduced the percentages and absolute numbers of CD25⁺ CD4⁺ T cells (Fig. 1 B) and GITR^high CD4⁺ T cells (Fig. S1) in the thymus, spleen, and lymph nodes (not depicted), confirming a previous result by others for splenocytes (25). Reduction was significant from 3 d after anti–IL-2 injection for nearly 2 mo, gradually recovering to normal levels by 3 mo after treatment. Recovery in the thymus was faster than in the periphery. In contrast, the treatment did not significantly alter the number of CD25⁻ CD4⁺ T cells in the thymus and spleen.

Because IL-2 regulates the expression of CD25 (26), it is possible that IL-2 neutralization might impair CD25 expression by CD25⁺ CD4⁺ T cells rather than reducing their cell number. To examine this possibility, we assessed the expression levels of Foxp3 14 d after anti–IL-2 treatment (10, 27, 28). Splenic CD4⁺ T cells purified from anti–IL-2–treated mice exhibited significantly (3.4-fold; P < 0.01) reduced amounts of Foxp3 mRNA compared with those from control saline-treated mice (Fig. 1 C). CD4⁺ CD8⁻ thymocytes from anti–IL-2–treated mice also showed reduced Foxp3 mRNA expression, but the reduction was not so marked as in splenic CD4⁺ T cells (2.2-fold; P < 0.06), presumably due to physiologically lower expression levels of Foxp3 in thymic CD25⁺ CD4⁺ CD8⁻ thymocytes than in CD25⁺ CD4⁺ splenic T cells (Fig. S2 A).

To determine then whether anti–IL-2 treatment reduces natural T reg cells with autoimmune-preventive activity, we transferred spleen cells from BALB/c mice 14 d after treatment to syngeneic nude mice and examined possible development of autoimmune disease (8). Fig. 1 D shows that all of the recipients developed histologically and serologically evident autoimmune gastritis, whereas nude mice cell transferred from control BALB/c mice did not. In addition, cotransfer of anti–IL-2–treated spleen cells with CD25⁺ CD4⁺ T cells from untreated normal animals inhibited the development of gastritis.

Taken together, these results indicate that IL-2 neutralization can selectively reduce the number of autoimmune-preventive Foxp3-expressing natural CD25⁺ CD4⁺ T reg cells in the immune system, especially in the periphery.

To determine whether IL-2 neutralization affects the thymic generation, peripheral maintenance (survival and/or growth), or both, of CD25⁺ CD4⁺ T reg cells, we thymectomized (Tx) 5–6-wk-old BALB/c mice and injected 1 mg anti–IL-2 mAb or saline 3 wk later. 14 d after anti–IL-2 mAb treatment, the percentage and the number of CD25⁺ CD4⁺ T cells decreased equally in Tx and sham-Tx mice (Fig. 2, A and B). This indicates that IL-2 neutralization reduces peripheral CD25⁺ CD4⁺ T reg cells whether there is a thymic supply of T reg cells or not.

Despite in vitro anergy of CD25⁺ CD4⁺ T reg cells, they proliferate on transfer to lymphopenic mice (29, 30). Furthermore, in normal nonlymphopenic mice, peripheral CD25⁺ CD4⁺ T cells show more active proliferation than CD25⁻ CD4⁺ T cells in the physiological state (Fig. 3 A; references 9, 31, and 32). Based on these findings, we first determined whether IL-2 neutralization in normal nonlymphopenic mice blocked natural proliferation of CD25⁺ CD4⁺ T cells. Normal BALB/c mice were injected with anti–IL-2 mAb or saline and subsequently with 1 mg 5-bromo-2′-deoxyuridine (BrdU) every 12 h for 3 d. IL-2 neutralization reduced incorporation of BrdU in CD25⁺ CD4⁺ T cells to one third of control mice, with no reduction in CD25⁻ CD4⁺ T cells, indicating that CD25⁺ CD4⁺ T cells require IL-2 for their physiological proliferation in the periphery (Fig. 3, A and B).

To determine then whether IL-2 is also essential for lymphopenia-driven expansion of CD25⁺ CD4⁺ T cells, we transferred highly purified (>99%) 5- and 6-carboxyfluorescein diacetate succinimidyl ester (CFSE)-labeled CD25⁺ CD4⁺ T cells from BALB/c-Thy1.1 congenic mice to BALB/c–RAG-2^−/− recipients and treated them with anti–IL-2 mAb to exclude a possible contribution of IL-2 derived from the host (for example, produced by the host DCs; reference 33) or transferred CD25⁺ CD4⁺ T cells themselves. Assessment 4 d later revealed that transferred Thy1.1⁺ T cells had vigorously proliferated, and this proliferation was not affected by anti–IL-2 (Fig. 4 A). This IL-2–independent proliferation was further confirmed by a similar level of proliferation of CD25⁺ CD4⁺ T cells transferred to BALB/c–RAG-2^−/− IL-2^−/− mice (not depicted). Furthermore, when these recipient mice were examined 4 wk later, the number of Thy1.1⁺ splenic T cells was not significantly different among the recipients, whether they were IL-2^+/+ or IL-2^−/−, cotransferred with CD25⁻ CD4⁺ T cells or not, or treated with anti–IL-2 or saline (Fig. 4 B).

Taken together, these results indicate that IL-2 is an indispensable growth factor for physiological proliferation and survival of CD25⁺ CD4⁺ T reg cells in normal nonlymphopenic conditions but dispensable for lymphopenia-induced homeostatic expansion or survival.

Next, we attempted to determine which cells produce the IL-2 required for the growth/survival of CD25⁺ CD4⁺ T reg cells in normal naive nonlymphopenic mice. We quantitatively assessed IL-2 mRNA expression by real-time PCR in various cell populations prepared from the spleens of normal naive BALB/c mice. The analyzed populations included macrophages (CD11b⁺), DCs (CD11c^high MHCII⁺), B cells (CD19⁺), NKT cells (TCRβ⁺ α-GalCer/CD1d-tetramer⁺), NK cells (DX5⁺ TCRβ⁻), CD8⁺ T cells and CD4⁺ T cells, which were sorted to CD25⁻, CD25^low, and CD25^high cells (Fig. 5 A). High IL-2 transcription was detected in CD4⁺ T cells, especially CD25^low CD4⁺ T cells, and to lesser degrees in NKT, NK, and CD8⁺ T cells. In accord with this, CD25^low CD4⁺ T cells secreted IL-2 in significantly higher amounts than CD25^high or CD25⁻ CD4⁺ T cells during 24 h in vitro culture (Fig. 5 A). Similar results were obtained with DO11.10 TCR transgenic mice. CD25^low CD4⁺ T cells showed an approximately five- or twofold higher transcription of IL-2 than CD25^high or CD25⁻ CD4⁺ T cells, respectively, whereas CD4⁺ T cells from RAG^−/− DO11.10 mice, which lack CD25⁺ CD4⁺ T reg cells (34), scarcely transcribed the gene (Fig. 5 B). This CD25^low CD4⁺ population was mainly composed of KJ1-26^negative-low T cells similar to CD25^high CD4⁺ T cells, indicating that CD25^low CD4⁺ T cells expressed endogenous TCR α chains paired with transgenic TCR β chains. In contrast to IL-2 transcription, Foxp3 expression was higher (approximately threefold) in the CD25^high population than in the CD25^low CD4⁺ population in BALB/c mice and DO11.10 transgenic mice, exhibiting an inverse correlation between IL-2 and Foxp3 expression (Fig. 5 C; reference 10). Furthermore, consistent with their low Foxp3 expression status, CD25^low CD4⁺ T cells failed to suppress the anti-CD3–driven proliferation of CD25⁻ CD4⁺ T cells in vitro, contrasting with potent suppression by CD25^high CD4⁺ T cells (Fig. 5 D). Notably, however, not only CD25^high but also CD25^low CD4⁺ T cells proliferated poorly upon anti-CD3 stimulation (Fig. 5 D). Judging from low level expression of Foxp3 by CD25^low CD4⁺ T cells (10-fold less than CD25^high cells), their hypo-responsiveness could be attributed to suppression by a small number of contaminating Foxp3^high CD25^high T reg cells due to the limit of fidelity in cell sorting. Alternatively, Foxp3^low CD25^low CD4⁺ T cells themselves might be hypoproliferative but not suppressive. Likewise, small amounts of IL-2 mRNA detected in the CD25^high or CD25⁻ fraction could be due to contaminating IL-2–secreting CD25^low cells or production of a small amount of IL-2 by these populations.

CD25^low CD4⁺ CD8⁻ thymocytes were present in the thymus of normal naive mice and expressed Foxp3 mRNA at levels one third of thymic CD25^high CD4⁺ T cells (Fig. S2 A). The IL-2 mRNA levels of CD25^low CD4⁺ CD8⁻ thymocytes were, however, much lower than the levels of the peripheral counterpart and equivalent to those of CD25^high CD4⁺ CD8⁻ thymocytes (Fig. S2 B). In addition, in contrast with splenic CD25^low CD4⁺ T cells, which were higher in CD44 expression than peripheral CD25^high or CD25⁻ CD4⁺ T cells, CD25^low CD4⁺ CD8⁻ thymocytes were intermediate in CD44 expression between CD25^high and CD25⁻ CD4⁺ thymocytes (Fig. S2 C). These data indicate that splenic CD25^low CD4⁺ T cells are activated in the periphery and therefore actively transcribe IL-2.

Taken together, the CD25⁺ CD4⁺ T cell population in normal naive mice is functionally heterogeneous, containing activated nonregulatory CD25^low T cells, which might be the main source of IL-2 in normal naive mice. Assuming that CD25⁺ CD4⁺ T cells that express endogenous TCR α chains in TCR transgenic mice are more self-reactive than other T cells (34), such CD25^low CD4⁺ T cells might be more self-reactive than CD25⁻ CD4⁺ T cells and therefore activated by self-antigens in the physiological state and actively forming IL-2.

Based on the result that anti–IL-2 mAb administration reduced the number of functional T reg cells (Fig. 1), we examined whether autoimmune disease would develop in normal BALB/c mice after anti–IL-2 mAb injection on days 10 and 20 after birth. Such treatment indeed induced by 3 mo of age histologically evident gastritis accompanying the appearance of anti-parietal cell autoantibodies in the circulation (Fig. 6, A and B). Furthermore, splenocytes from gastritis-bearing BALB/c mice adoptively transferred gastritis to BALB/c athymic nude mice, indicating that the lesion was a bona fide T cell–mediated autoimmune disease (Fig. 6 C).

This anti–IL-2 mAb treatment of young mice significantly reduced the number of CD25⁺ CD4⁺ T cells for nearly 2 mo. The number gradually recovered to the control level by 3 mo after treatment (Fig. S3 A), similar to the result with adult mice (Fig. 1 B). The recovered CD25⁺ CD4⁺ T cells were equally anergic and exhibited comparative degrees of suppressive activity and similar levels of Foxp3 mRNA expression as those from control mice (Fig. S3, B and C), indicating that the majority of them were T reg cells but not activated autoimmune T cells. Thus, IL-2 neutralization and the resulting transient T reg cell deficiency is sufficient to elicit autoimmune disease in otherwise normal mice.

Diabetes-prone NOD mice are known to be genetically susceptible to other organ-specific autoimmune diseases such as thyroiditis and sialadenitis (35, 36). Therefore, we examined whether neutralization of IL-2 would exacerbate diabetes and produce autoimmune disease in a wider spectrum of organs compared with nontreated NOD mice. A similar course of anti–IL-2 treatment as shown in Fig. 6 for BALB/c mice triggered an earlier onset of diabetes, resulting in a higher incidence of diabetes (Fig. 7 A) and more severe forms of insulitis at the time of examination (Fig. 7, B and C). Furthermore, the incidence and histological severity of gastritis and thyroiditis increased in these anti–IL-2–treated mice. Histologically evident gastritis and thyroiditis in the treated mice was 90.9 and 54.5%, respectively, compared with 0% in control NOD mice at 20 wk of age (Fig. 7, B, D, and E, and see Fig. 9). These mice with gastritis or thyroiditis developed high titers of anti-parietal cell or anti-thyroglobulin autoantibodies, respectively (Fig. 7, D and E). Histologically evident inflammation in lacrimal glands and salivary glands was also observed in 36.4 and 9%, respectively, of anti–IL-2–treated NOD mice but absent in control mice (Fig. 7 B and see Fig. 9).

Notably, some mice developed ataxia and paralysis of the limbs around 15 wk after anti–IL-2 treatment (Fig. 8, A and B). Histological examination of the nervous system revealed lymphocytic infiltration and demyelination in peripheral nerves (Fig. 8 C) but not in the central nervous system (not depicted). Anti–IL-2 treatment elicited neuropathy in 56.3% of NOD mice and triggered diabetes in 50% of NOD mice assessed at 3 mo of age. Interestingly, neuropathy and diabetes were apparently mutually exclusive: only 1 of 11 mice developed both diseases (Fig. 9). The peripheral neuropathy could be adoptively transferred to naive NOD.SCID mice by spleen cell suspensions from neuropathic NOD mice. Both CD4⁺ and CD4⁻ T cells from neuropathic NOD mice induced histologically evident neuritis in the SCID recipients (100% by CD4⁺ T cells and 40% by CD4⁻ T cells), accompanying clinical signs of neuropathy such as trembling limbs (80% by CD4⁺ T cells and 40% by CD4⁻ T cells; Fig. 8, D and E).

Thus, anti–IL-2 mAb treatment enhances diabetes and produces a wide spectrum of organ-specific autoimmune diseases and most notably, peripheral neuritis in autoimmune-prone NOD mice.

The following findings in this report support IL-2 being critically required for the peripheral maintenance of natural CD25⁺ CD4⁺ T reg cells and consequently the immunologic self-tolerance sustained by them. First, injection of anti–IL-2 mAb to Tx mice resulted in a marked reduction of the number of peripheral CD25⁺ CD4⁺ T cells, which corresponded to Foxp3-expressing natural T reg cells. The number of CD4⁺ T cells expressing high levels of GITR, which is more stably expressed by Foxp3-expressing natural T reg cells than CD25 (37), was also reduced after anti–IL-2 neutralization (Fig. S1), supporting the reduction of T reg cells rather than conversion of T reg cells from CD25⁺ to CD25⁻. Second, neutralization of IL-2 in normal mice inhibited the physiological proliferation of peripheral CD25⁺ CD4⁺ T cells. Third, IL-2 neutralization in normal mice for a limited period induced autoimmune disease similar to the one produced by depletion of CD25⁺ CD4⁺ T reg cells. Furthermore, transfer of spleen cells from BALB/c mice with low numbers of CD4⁺ T cells expressing CD25 and Foxp3 after anti–IL-2 treatment produced in nude mice autoimmune diseases similar to those seen after the transfer of CD25⁺ cell–depleted BALB/c spleen cells, and cotransfer of normal CD25⁺ CD4⁺ T cells prevented the autoimmune development (8). Taken together, these findings indicate passive cell death of CD25⁺ CD4⁺ T reg cells as a major mechanism of T reg cell reduction upon IL-2 deprivation, although our results do not exclude the possibility that IL-2 neutralization may reduce CD25 expression by natural T reg cells, thereby impairing their activation or down-regulating their Foxp3 transcription, leading to the attenuation of their suppressive activity.

Previous studies with IL-2^−/− or IL-2R^−/− mice also demonstrated a critical role for IL-2 and IL-2R in homeostasis of peripheral CD25⁺ CD4⁺ T reg cells. For example, not only IL-2^−/− mice but also IL-2Rβ^−/− mice bear few CD25⁺ CD4⁺ T cells in the thymus and periphery (19, 38). In chimeric RAG-2^−/− mice reconstituted with IL-2^−/− and CD25^−/− BW cells, the normal number of peripheral CD25⁺ CD4⁺ T cells was established solely from IL-2^−/− BM cells (39). This indicates that IL-2 produced by CD25^−/− cells promoted the differentiation of IL-2^−/− T cells to T reg cells. In IL-2Rβ^−/− mice, the number of CD25⁺ CD4⁺ T cells was restored in the thymus and the periphery by thymic expression of an IL-2Rβ transgene, indicating a critical role of the IL-2/IL-2R pathway for the thymic generation of CD25⁺ CD4⁺ T cells (38). Thus, our present results, together with these findings by others, indicate that IL-2 is crucially required for the maintenance of peripheral CD25⁺ CD4⁺ T reg cells. IL-2 may also be required for the activation of CD25⁺ CD4⁺ T reg cells to exert suppression (23). It is likely that severe inflammation/autoimmunity in IL-2^−/−, CD25^−/−, or CD122^−/− mice could be attributed to a synergistic effect of the reduction of natural T reg cells and their impaired function and differentiation in addition to possible altered activation-induced cell death of self-reactive T cells.

In contrast to the requirement of IL-2 for physiological expansion of CD25⁺ CD4⁺ T reg cells in normal nonlymphopenic mice, homeostatic proliferation in a lymphopenic environment appears to be IL-2 independent. It has been shown that CD25⁺ CD4⁺ T cells as well as conventional CD25⁻ CD4⁺ T cells undergo extensive homeostatic expansion when transferred into lymphopenic recipients (29, 30). It is therefore likely that strong TCR stimulation of CD25⁺ CD4⁺ T reg cells by self-peptide/MHC, together with costimulation via CD28, may suffice to sustain their homeostatic proliferation in a T cell–deficient environment. It may correspond to the finding that TCR and CD28 costimulation can abrogate the anergic state of CD25⁺ CD4⁺ T reg cells and evoke their proliferation in vitro, as TCR stimulation in the presence of high dose IL-2 has a similar in vitro effect (40, 41).

We have shown that splenic CD25⁺ CD4⁺ T cells are phenotypically and functionally heterogeneous and can be divided to two functionally distinct populations according to the level of CD25 expression: one is the CD25^high population, which exhibits suppressive activity, expresses a high level of Foxp3, and produces little IL-2, the other being the CD25^low population, which has little suppressive activity, expresses a low level of Foxp3, and produces IL-2. The CD25⁺ CD4⁺ T cell population thus contains both CD25^low CD4⁺ activated non–T reg cells and Foxp3-expressing CD25^high CD4⁺ T reg cells. TCR transgenic mice also harbor IL-2–transcribing Foxp3^low CD25^low CD4⁺ T cells, whereas RAG^−/− TCR transgenic mice did not. These findings collectively indicate the following. First, some of the IL-2–transcribing CD25^low CD4⁺ T cells might be activated by self-antigens because this population in TCR transgenic mice predominantly expresses endogenous TCR α chains paired with transgenic β chains and appears to be positively selected because of high self-reactivity of such TCRs, similar to positive selection of CD25^high CD4⁺ T reg cells (34). Second, the presence of activated CD25^low CD4⁺ non–T reg cells within the CD25⁺ CD4⁺ T cell population explains the finding by others that CD25⁺ CD4⁺ T cells isolated from normal naive mice were able to induce colitis when transferred to SCID mice and subsequently treated with anti–IL-10R mAb (42). Such colitis-inducing CD25⁺ CD4⁺ T cells may well have contained pathogenic CD25^low CD4⁺ T cells in addition to CD25⁺ CD4⁺ T reg cells, and the IL-10R blockade may have abrogated the suppression exerted by the latter, allowing the former to cause colitis. Third, our results substantiate the finding in humans and mice that the CD25^high CD4⁺ population predominantly contains T reg cells (43, 44). Thus, assuming that CD25^low CD4⁺ T cells secrete IL-2 in the physiological state by responding to self-antigens or commensal bacteria, IL-2 can be a mediator of negative feedback control by which activated CD25^low CD4⁺ nonregulatory T cells contribute to the maintenance and activation of natural CD25⁺ CD4⁺ T reg cells, which in turn limit the expansion of the former. This may occur locally where both natural T reg cells and effector T cells are recruited to APCs presenting particular self- or nonself-antigens.

IL-2 neutralization enhanced autoimmunity in NOD mice presumably by reducing natural CD25⁺ CD4⁺ T reg cells. The treatment facilitated the onset of T1D, produced more severe insulitis with higher incidence of diabetes, and provoked other autoimmune diseases in more severe a form than control mice. Interestingly, anti–IL-2 mAb-treated NOD mice also developed T cell–mediated autoimmune neuropathy similar to the one that developed in B7-2–deficient or anti–B7-2 mAb-treated NOD mice (45). It was shown that treatment with pertussis toxin alone, which impairs blood brain barrier, evoked EAE in NOD mice (46). In addition, allelic variants of the IL-2 gene contribute to the development of T1D and EAE in NOD congenic mice (6). Taken together, it is likely in NOD mice that genetic variations of IL-2 may contribute to determining the genetic susceptibility to these autoimmune diseases by altering the development and function of natural T reg cells (47). Furthermore, such genetic variation/anomalies of IL-2 may cause different autoimmune disease in combination with other genes (47, 48), as illustrated by the development of autoimmune gastritis in BALB/c mice and autoimmune neuropathy and other autoimmune diseases in NOD mice (9).

In conclusion, IL-2 is an indispensable cytokine for the peripheral maintenance of natural CD25⁺ CD4⁺ T reg cells. Alteration in the formation of IL-2 by other T cells, structural abnormality of the IL-2 molecule itself (49), or anomalies in intracellular signal transduction through IL-2R (50–52) might be predisposing to, or even causative of, various autoimmune diseases (6, 53). Furthermore, not only genetic modification but also environmental alteration of IL-2 formation may cause autoimmune disease. For example, cyclosporin A can cause autoimmune disease similar to the one produced by IL-2 neutralization, presumably by inhibiting IL-2 formation and consequently reducing natural CD25⁺ CD4⁺ T reg cells or impairing their functions (54). On the other hand, IL-2 can be used for strengthening natural self-tolerance and preventing autoimmune disease or stabilizing transplantation tolerance by maintaining the survival/expansion and augmenting the suppressive activity of natural T reg cells.

BALB/c, BALB/c athymic nude, C.B-17 SCID, NOD, and NOD.SCID mice were purchased from CLEA. IL-2–deficient mice were purchased from The Jackson Laboratory and backcrossed to BALB/c mice more than eight times. BALB/c-Thy1.1 congenic mice were established in our laboratory (55). DO.11.10 TCR transgenic mice and RAG-2–deficient BALB/c mice were provided by D. Loh (Washington University, St. Louis, MO) and Y. Shinkai (Kyoto University, Kyoto, Japan), respectively. All mice were maintained in our animal facility and treated in accordance with the guidelines for animal care approved by the Institute for Frontier Medical Sciences, Kyoto University.

The rat anti–IL-2 mAb (IgG2a, clone S4B6; reference 56) was purified from ascites using protein G columns (Amersham Biosciences). The following reagents were purchased from BD Biosciences: purified anti-CD3ε (145-2C11), anti-CD16/CD32 (2.4G2), and anti–IL-2Rβ (TM-β1); and FITC-labeled or biotinylated anti-CD25 (7D4), FITC–anti-CD4 (RM4-5), PE–anti-CD4 (H129.19), FITC–anti-CD19 (1D3), PE–anti-CD11c (HL3), PE–anti-CD8 (53–6.7), PE–anti-Thy1.1 (OX-7), PE–anti-CD122 (TM-β1), and PE–anti-CD132 (4G3). PE–anti-CD3ε (145-2C11), PE–anti-CD11b (M1/70), and PE–anti-pan NK cells (DX-5) were purchased from eBioscience. α-GalCer (KRN7000) was provided by Kirin Brewery Co. α-GalCer/CD1d tetramer was provided by H. Wakasugi and Y. Ikarashi (National Cancer Institute, Tokyo, Japan). Anti-GITR mAb (DTA-1) was produced, purified, and labeled with Alexa in our laboratory (57).

Lymphocyte suspensions were prepared from thymi, spleens, and lymph nodes. Erythrocytes were lysed with ACK buffer. Thymic CD4⁺ CD8⁺ and CD8⁺ cells were depleted by treatment with anti-CD8 mAb (3.155) and rabbit complement (Cedarlane Laboratories). For flow cytometric analysis, 10⁶ cells were incubated with FITC- or PE-labeled or biotinylated mAbs and then with CyChrome- or PE-streptavidin (BD Biosciences) as a secondary reagent for biotinylated antibodies. Cells were then analyzed by flow cytometry (Epics-XL; Beckman Coulter). To sort peripheral CD4⁺ T cell subpopulations, CD4⁺ T cells were first enriched by depleting B cells, CD8⁺ cells, and adherent cells by panning as described previously (8). After staining with fluorescent antibodies, cells were sorted by an Epics Altra cell sorter (Beckman Coulter). The purity of the resulting CD25⁺ or CD25⁻ CD4⁺ population was ⩾99%. In some experiments, CD25⁺ CD4⁺ T cells were purified using the MACS system (Miltenyi Biotec). The enriched CD4⁺ T cell fraction was stained with biotinylated anti-CD25 followed by streptavidin PE. Cells were then labeled with anti-PE microbeads and passed through LS columns twice to obtain CD25⁺ cells. The purity of the resulting CD25⁺ CD4⁺ T cells was 90–95%.

Organs were fixed with 10% formalin and processed for hematoxylin and eosin staining. Autoantibodies specific for the gastric parietal cells and thyroglobulin were assayed by ELISA (8). Gastritis and thyroiditis were graded 0–2+ as described previously (8).

Foxp3 or IL-2 mRNA levels were quantified by using real-time PCR as described previously (10). IL-2 primer sequences used in this study were: 5′-tccagaacatgccgcagag-3′ and 5′-cctgagcaggatggagaattaca-3′. The IL-2 probe used was 5′-FAM-actccccaggatgctcaccttcaaattt-3′. The primers were designed to hybridize on an intron–exon junction to prevent amplification of contaminating genomic DNA.

Purified CD25^high, CD25^low, and CD25⁻ CD4⁺ T cells (8.5 × 10⁵ per well in a flat-bottomed 96-well plate) were cultured in 80 μl of complete RPMI medium for 24 h. To block the binding of IL-2 to IL-2R on these populations, 10 μg/ml anti–IL-2Rβ was added in the culture. IL-2 concentration in culture supernatants was determined by sandwich ELISA (eBioscience), which has a detection limit of 2 pg/ml.

Sorted CD25⁻ CD4⁺ cells (2.0 × 10⁴ per well in a U-bottomed 96-well plate) and 4.0 × 10⁴ irradiated APCs per well were cultured for 72 h in the presence of 1.0 μg/ml anti-CD3ε mAb (145-2C11; BD Biosciences). CD3⁺ cell–depleted spleen cells were used as APCs. [³H]thymidine (1 μCi/well; DuPont/NEN Life Science Products) was added during the last 6 h of culture. Background counts in the wells with APCs alone were always <200 cpm (40).

Mice were injected i.p. with 1.0 mg BrdU (Sigma-Aldrich) every 12 h for 3 d. 10⁷ pooled LN and spleen cells were surface stained with PE-CD4 and biotin-CD25 followed by RPE-Cy5-streptavidin (DakoCytomation). BrdU staining was performed as reported by others (58) using FITC-labeled anti-BrdU mAb (DakoCytomation).

Purified CD25⁺ CD4⁺ T cells were incubated with 3 μM CFSE in PBS for 10 min at room temperature. An equal volume of normal mouse serum was then added to stop the reaction. After washing twice, 10⁶ labeled cells were i.v. injected.

Female 5–6-wk-old mice were anesthetized by 10% nembutal and Tx or sham operated as previously described (8).

Blood glucose levels were measured every week with a Medisafe reader (TERUMO). Mice were considered diabetic when blood glucose levels were >300 mg/dl after two consecutive measurements.

All statistical analyses were performed by Student's paired t test.

Fig. S1 shows that IL-2 neutralization decreases the number of CD4⁺ GITR^high cells in the thymus and periphery. Fig. S2 shows possible activation of CD25^low CD4⁺ T cells in the periphery. Fig. S3 shows that the size of CD25⁺ CD4⁺ T cells decreases and then recovers in BALB/c mice treated with anti–IL-2 on days 10 and 20 after birth. Figs. S1–S3 are available.

We thank Z. Fehervari for critically reading the manuscript; K. Hirota and R. Ishii for assistance with cell sorting; T. Matsushita for histology; M Kakino for maintaining mice; and H. Wakasugi, Y. Ikarashi, and Kirin Brewery Co. for reagents.

This work was supported by grants-in-aid from the Ministry of Education, Sports and Culture of Japan.

The authors have no conflicting financial interests.