Stimulation of CD25⁺CD4⁺ regulatory T cells through GITR breaks immunological self-tolerance

Abstract

CD25⁺CD4⁺ regulatory T cells in normal animals are engaged in the maintenance of immunological self-tolerance. We show here that glucocorticoid-induced tumor necrosis factor receptor family–related gene (GITR, also known as TNFRSF18)—a member of the tumor necrosis factor–nerve growth factor (TNF-NGF) receptor gene superfamily—is predominantly expressed on CD25⁺CD4⁺ T cells and on CD25⁺CD4⁺CD8⁻ thymocytes in normal naïve mice. We found that stimulation of GITR abrogated CD25⁺CD4⁺ T cell–mediated suppression. In addition, removal of GITR-expressing T cells or administration of a monoclonal antibody to GITR produced organ-specific autoimmune disease in otherwise normal mice. Thus, GITR plays a key role in dominant immunological self-tolerance maintained by CD25⁺CD4⁺ regulatory T cells and could be a suitable molecular target for preventing or treating autoimmune disease.

Main

Accumulating evidence suggests that, apart from clonal deletion and anergy, T cell–mediated control of self-reactive T cells contributes to the maintenance of natural immunological self-tolerance^1,2,3. For example, depletion of CD25⁺CD4⁺ T cells, which constitute 5–10% of peripheral CD4⁺ T cells in normal naïve animals, leads to the spontaneous development of various autoimmune diseases in genetically susceptible animals; reconstitution of the depleted population prevents the development of autoimmunity^4,5,6. Elimination of CD25⁺CD4⁺ T cells enhances immune responses to nonself-antigens such as allogeneic tissue grafts⁴. It also provokes effective tumor immunity in otherwise nonresponsive animals⁷. On the other hand, expansion of CD25⁺CD4⁺ regulatory T cells or augmentation of their activity can suppress allograft rejection and even induce allograft tolerance^2,3,8. Thus, this naturally occurring immunoregulatory T cell population is not only engaged in preventing autoimmune disease but can also be exploited to evoke effective tumor immunity or induce transplantation tolerance.

CD25⁺CD4⁺ regulatory T cells have unique immunological characteristics compared with other regulatory or suppressor T cells induced by certain routes of exogenous immunization or tolerance induction⁸. For example, they do not proliferate in response to antigenic stimulation in vitro (that is, they are naturally anergic) and can potently suppress the activation and proliferation of other CD4⁺ or CD8⁺ T cells in an antigen-nonspecific manner through cell-to-cell interactions^9,10. CD25⁺CD4⁺ regulatory T cells are phenotypically similar to “memory” or “primed” T cells and less dependent on CD28-mediated costimulation for their activation^9,10,11. However, ligation of CD28 expressed on CD25⁺CD4⁺ regulatory T cells, in conjunction with T cell receptor (TCR) stimulation, can break their anergic and suppressive state^9,10,11. In addition, they constitutively express cytotoxic lymphocyte–associated antigen 4 (CTLA-4)^12,13,14. Blockade of CTLA-4 abrogates CD25⁺CD4⁺ T cell–mediated suppression in vitro, elicits autoimmune disease and evokes tumor immunity in vivo. This indicates that CTLA-4 may be a costimulatory molecule for activating CD25⁺CD4⁺ regulatory T cells⁸. Upon activation, they also express membrane-bound tumor growth factor-β (TGF-β), which might mediate suppression through cell-cell contact¹⁵. They are produced, at least in part, by the normal thymus as a functionally mature (that is, anergic and suppressive) form via the thymic selection process^11,16. Although these findings indicate that several key molecules are involved in the generation and function of CD25⁺CD4⁺ regulatory T cells, the molecular basis for their generation or function still remains unknown.

To investigate the molecular basis of T cell–mediated self-tolerance, we attempted to characterize cell surface molecules that are expressed on CD25⁺CD4⁺ regulatory T cells and are engaged in their activation and proliferation or exertion of suppression. We raised monoclonal antibodies (mAbs) capable of neutralizing in vitro CD25⁺CD4⁺ T cell–mediated suppression and characterized the molecules recognized by the mAbs. With these techniques we show here that the glucocorticoid-induced tumor necrosis factor receptor family–related gene (GITR, also known as TNFRSF18)—a member of the tumor necrosis factor–nerve growth factor (TNF-NGF) receptor family of proteins^17,18,19—is predominantly expressed on CD25⁺CD4⁺ regulatory T cells in the thymus and periphery. We also show that signaling through GITR can abrogate natural immunoregulation, thereby inducing autoimmune disease in otherwise normal mice. Among the TNF–TNF receptor (TNFR) superfamily of proteins—which regulate diverse biological functions such as cell proliferation, differentiation, activation and death—some members, including Fas–Fas ligand (FasL), CD40-CD40L and OX40-OX40L, contribute to the maintenance of self-tolerance by controlling survival or activation of self-reactive T cells^20,21,22,23. The role of GITR in T cell–mediated suppression of self-reactive T cells is another mode by which TNF-TNFR superfamily proteins contribute to maintaining immunological self-tolerance. In addition, the role of GITR could potentially be exploited to control immune responses to nonself or quasi-self antigens.

Results

A mAb recognizing GITR can abrogate suppression

A rat was immunized with an anergic and suppressive CD25⁺ CD4⁺ T cell line, which was established by repeatedly stimulating CD25⁺CD4⁺ T cells from normal BALB/c mice with anti-CD3 and a high dose of interleukin 2 (IL-2)^9,11 (data not shown). Hybridoma cells were generated by fusing the spleen cells from the immunized rat with mouse myeloma cells, assessed for secreting antibodies capable of neutralizing in vitro CD25⁺CD4⁺ T cell–mediated suppression of anti-CD3–stimulated T cell proliferation⁹, then cloned. One clone, designated DTA-1, secreted a monoclonal rat immunoglobulin G2a (IgG2a) that showed high neutralizing activity (see below). Another clone secreted antibodies specific for mouse CD28, ligation of which abrogated suppression (data not shown)^9,11.

Immunoprecipitation of cell surface molecules from a DTA-1^hi T cell hybridoma, 18.3.6, with the use of the mAb DTA-1 revealed that the molecule the mAb recognized was a 66–70-kD molecule composed of two 40-kD components (Fig. 1). Two-dimensional electrophoresis of the immunoprecipitate formed a single spot, which provided further evidence that the molecule was a homodimeric protein (data not shown).

Figure 1: DTA-1 mAb recognizes GITR.

(a) DTA-1 recognizes GITR expressed on COS-7 cells transfected with GITR cDNA. COS-7 cells were transfected with pME18S-GITR (thick line) or empty vector (thin line), and stained with DTA-1 and FITC-anti–rat IgG. (b) Immunoprecipitation of GITR from GITR-expressing cells by DTA-1. Surface-biotinylated 18.3.6 T cell hybridoma cells (lanes 1, 4 and 7), untreated COS-7 cells (lanes 2, 5 and 8) or COS-7 cells transfected with pME18S-GITR (lanes 3, 6 and 9) were immunoprecipitated by DTA-1. Precipitates were subjected to SDS-PAGE under nonreduced (lanes 1–3), reduced (lanes 4–9) or N-glycosidase–treated and reduced (lanes 7–9) conditions and blotted onto a nitrocellulose membrane. Blotted proteins were detected with avidin-peroxidase and appropriate substrates.

Expression cloning of the cDNA encoding the DTA-1–recognized molecule from a cDNA library constructed with mRNA from the 18.3.6 hybridoma cells revealed that the nucleotide sequence was identical to that encoding GITR^17,18,19. Cytofluorometric analysis revealed that COS-7 cells transfected with a vector expressing the GITR portion of the cloned gene were stained with DTA-1, whereas COS-7 cells transfected with empty vector were not (Fig. 1a). Immunoprecipitation from the GITR-transfected COS-7 cells with DTA-1 mAb revealed 70-kD and 37-kD bands (which presumably corresponded to dimer and monomer molecules, respectively) in nonreduced conditions, and a 37-kD band in reduced conditions (Fig. 1b). Although the 37-kD band differed slightly in molecular size from the band precipitated from 18.3.6 cells, the difference could be attributed to the degree of glycosylation. This is because N-glycosidase treatment of the immunoprecipitate from GITR-transfected COS-7 cells or 18.3.6 cells produced bands of the same size (Fig. 1b). Thus, the molecule recognized by DTA-1 mAb was GITR, a 66–70-kD homodimeric glycoprotein.

Tissue distribution of GITR-expressing cells

Flow cytometric analyses of the spleen and lymph node cells of normal BALB/c mice revealed that GITR is highly expressed on CD25⁺CD4⁺ T cells (Fig. 2a). In the thymus, CD25⁺CD4⁺ single positive (SP) thymocytes expressed GITR at equivalent amounts to peripheral CD25⁺CD4⁺ T cells (Fig. 2a). The majority of peripheral GITR^hi T cells were CD5^hi, CD45RB^lo and largely CD62L^hi (Fig. 2b). In contrast, CD25⁻CD4⁺ T cells, CD8⁺ T cells, B cells (B220⁺ cells) and macrophages (F4/80⁺ cells) expressed low amounts of GITR (Fig. 2c). Both CD11b⁺ and CD11c⁺ cells were GITR^lo, which indicated that macrophages and dendritic cells also constitutively express GITR in low amounts. Activation of CD4⁺ and CD8⁺ T cells by stimulation with anti-CD3 and IL-2 induced high expression of GITR (Fig. 2c), whereas activation of B cells or macrophages with lipopolysaccharide (LPS) or interferon-γ (IFN-γ), respectively, only slightly enhanced expression. Spleen cells from other strains of mice—including C57BL/6 (B6), C3H and nonobese diabetic mice—also showed a similar expression pattern before and after in vitro activation (data not shown).

Figure 2: Expression of GITR on mouse lymphoid cells.

(a) Thymocyte suspensions (depleted of CD8⁺ cells by anti-CD8 and complement treatment) or spleen cell suspensions (depleted of B cells by panning on anti-Ig–coated dishes) prepared from a 2-month-old BALB/c mouse were stained with anti-CD4, anti-CD25 and DTA-1. Expression of GITR on CD25⁺ or CD25⁻CD4⁺ thymocytes or T cells (gated in the left panels) is shown in the histograms (right panels). (b) Normal BALB/c spleen cells were stained with DTA-1 + anti-CD25, anti-CD45RB, anti-CD5 or anti-CD62L. (c) Freshly prepared naïve BALB/c spleen cells (left panels) and activated cells stimulated with anti-CD3 + IL-2, LPS or IFN-γ (right panels) were stained with anti–DTA-1 and antibodies to CD4, CD8, B220 or F4/80. GITR expression by CD4⁺ and CD8⁺ T cells after stimulation with anti-CD3 + IL-2, by B220⁺ cells (B cells) after LPS stimulation or by F4/80⁺ cells (macrophages) after IFN-γ stimulation is indicated by the thick lines. Thymocyte suspensions prepared from a normal BALB/c mouse were stained with anti–DTA-1, anti-CD8 and anti-CD4; expression of GITR by the indicated cell fractions is indicated by the thick lines. Red blood cell–lysed normal BALB/c bone marrow cells were stained with DTA-1. In each case, the thin lines represent control staining with an irrelevant antibody. One representative of three independent experiments is shown in a–c.

In contrast to CD25⁺CD4 SP thymocytes, CD25⁻ CD4⁺ SP, CD8⁺ SP and CD4⁻ CD8⁻ thymocytes expressed low amounts of GITR and CD4⁺CD8⁺ thymocytes scarcely expressed it at all. The majority of bone marrow cells also expressed low amounts of GITR (Fig. 2c).

Thus, high expression of GITR in the T cell compartment is confined to CD25⁺CD4⁺ SP thymocytes and CD25⁺CD4⁺ peripheral T cells in normal naïve mice, although other T cells—which normally show low GITR expression—can become GITR^hi upon activation.

Abrogation of suppression by anti-GITR

DTA-1 mAb abrogated in vitro CD25⁺CD4⁺ T cell–mediated suppression in a dose-dependent manner when added to a culture of CD25⁺CD4⁺ T cells and CD25⁻CD4⁺ T cells (Fig. 3a) or CD8⁺ T cells (Fig. 3b) stimulated with soluble monoclonal anti-CD3 and x-irradiated antigen-presenting cells (APCs). In these experiments, the mAb did not break the anergic state of CD25⁺CD4⁺ T cells even at high concentrations (up to 400 μg/ml, data not shown).

Figure 3: DTA-1 mAb abrogates suppression mediated by CD25⁺CD4⁺ T cells or thymocytes.

(a) CD25⁺CD4⁺ (□) or CD25⁻CD4⁺ (○) spleen cells (1×10⁴) purified from 2-month-old BALB/c mice (gated as in Fig. 2a), or these two populations mixed in equal amounts (▵), were stimulated for 3 days with anti-CD3 with the use of x-irradiated BALB/c spleen cells (5×10⁴) as APCs in the presence of graded amounts of DTA-1. (b) CD25⁺CD4⁺ (▪) or CD8⁺ (●) spleen cells (1×10⁴), or these two populations mixed in equal amounts (▴), were similarly stimulated in the presence of graded amounts of DTA-1. (c) CD25⁻CD4⁺ T cells from 2-month-old CD28-deficient BALB/c mouse were cultured for 3 days with Con A (3 μg/ml) in the presence of graded amounts of DTA-1 or rat IgG. (d) CD25⁺ or CD25⁻CD4⁺ SP thymocytes purified by cell sorting (as enclosed in Fig. 2a) from 2-month-old BALB/c mice, or these two populations mixed in equal amounts, were cultured for 3 days with anti-CD3 and X-irradiated APCs in the presence of DTA-1 (100 μg/ml) or normal rat IgG. In these experiments, background counts in the wells containing APCs were only <1000 cpm. One representative of three independent experiments is shown in a–d. The means of duplicate cultures are shown in each figure; in each case the s.e.m. was within 10% of the mean.

It is plausible that, like other TNFR family members, GITR may act as a costimulatory molecule for TCR-stimulated T cells^24,25. Indeed, when T cells from CD28-deficient BALB/c mice were stimulated with anti-CD3 and increasing concentrations of DTA-1, T cell proliferation increased threefold (Fig. 3c). In addition, although DTA-1 did not enhance the proliferation of CD25⁻CD4⁺ T cells stimulated with a high concentration of anti-CD3 (∼5 μg/ml) (Fig. 3a), it slightly (two- to threefold) enhanced the proliferative responses of CD4⁺ T cells and CD8⁺ T cells at lower concentrations (<0.5 μg/ml) of anti-CD3 (data not shown). These findings indicate that GITR can function as a costimulatory molecule in TCR-mediated stimulation of naïve T cells.

Upon TCR stimulation, CD25⁺CD4⁺ SP thymocytes from normal naïve mice also suppress the activation and proliferation of other thymocytes¹¹. The DTA-1 mAb abrogated such CD25⁺CD4⁺ SP thymocyte-mediated suppression, which indicated that GITR on thymocytes (Fig. 2a) is functionally equivalent to GITR on peripheral T cells (Fig. 3d).

To determine whether other TNFR or TNF family members can contribute to CD25⁺CD4⁺ T cell–mediated suppression, mAbs specific for TNFR type I, TNFR type II, 4-1BB, CD27, CD30, CD40 or OX40 were added to anti-CD3–stimulated cultures of CD25⁺CD4⁺ T cells, CD25⁻CD4⁺ T cells and APCs (Fig. 4). At concentrations ranging from 1 μg/ml to 100 μg/ml, none of them showed neutralizing activity. Monoclonal anti-LTβR, polyclonal anti-TRAIL (antibody to TNF-related apoptosis-inducing ligand) or polyclonal anti-RANKL (antibody to receptor-activator of NF-κB ligand) (at concentrations up to 100 μg/ml) and lymphotoxin-β receptor–Ig (100 μg/ml) or human TNF-α (1–100 ng/ml) also failed to neutralize CD25⁺CD4⁺ T cell–mediated suppression (data not shown). Interference with Fas-FasL interactions or neutralization of TNF-α fails to abrogate CD25⁺CD4⁺ T cell–mediated suppression⁹. Thus, together these results showed that unlike GITR, other TNF-TNFR family members cannot abrogate CD25⁺CD4⁺ T cell–mediated suppression.

Figure 4: Effects of antibodies specific for TNFR family proteins on CD25⁺CD4⁺ T cell–mediated suppression.

An equal number (1×10⁴) of CD25⁺CD4⁺ T cells and CD25⁻CD4⁺ T cells were mixed and stimulated with anti-CD3 (10 μg/ml) and X-irradiated spleen cells (as APCs) in the presence of the indicated concentrations of monoclonal or polyclonal (P) antibodies specific for TNFR family proteins. The proliferative response of each culture was expressed as the percentage response; the proliferation of CD25⁻CD4⁺ T cells in response to the addition of antibody was designated as a 100% response. One representative of two independent experiments is shown.

Signal through GITR attenuates suppression

Comparison of the ability of the DTA-1 mAb and its Fab fragment to neutralize CD25⁺CD4⁺ T cell–mediated in vitro suppression showed that Fab fragments were unable to neutralize suppression even at high concentrations sufficient to fully stain GITR-expressing cells (Fig. 5a and Web Fig. 1 on the supplementary information page of Nature Immunology on line). In addition, both soluble and plate-coated DTA-1 up-regulated NF-κB transcription in GITR cDNA-transfected HEK 293 cells, although overexpression of mouse GITR in HEK 293 cells per se activated NF-κB to some degree (an approximate twofold increase compared with HEK 293 cells transfected with empty vector) (Fig. 5b). These findings indicated that the DTA-1 mAb has agonistic activity; that is, it may actively transduce an anti-suppressive signal to CD25⁺CD4⁺ T cells, CD25⁻CD4⁺ T cells and/or APCs by ligating GITR molecules.

Figure 5: Requirement for an active signal through GITR to CD25⁺CD4⁺ regulatory T cells for suppression.

(a) In the presence of graded concentrations of anti–DTA-1 Fab fragments, CD25⁺CD4⁺ or CD25⁻CD4⁺ spleen cells (1×10⁴) (designated CD25⁺ or CD25⁻, respectively) purified from 2-month-old BALB/c mice, or these two populations mixed in equal amounts (CD25⁺ + CD25⁻), were cultured for 3 days with anti-CD3 and x-irradiated BALB/c spleen cells (as APCs). Closed symbols denote the effect of intact DTA-1 (100 μg/ml). (b) NF-κB reporter luciferase activities in transfected HEK 293 cells (see Methods) were measured when cells were stimulated with soluble (SL) or plate-bound (PB) DTA-1. Data are shown as the fold increase in activity compared with cells transfected with empty vectors in triplicate experiments; they are representative of four independent experiments. (c) CD4⁺ T cells from a normal mouse or rat were stained with biotinylated DTA-1 (bold lines) or an irrelevant mAb (thin lines) and PE-streptoavidin. (d) CD25⁺ or CD25⁻CD4⁺ T cells (1×10⁴) prepared from BALB/c mice were cultured on x-irradiated rat spleen cells (as APCs) in the presence or absence of DTA-1 (100 μg/ml). (e) Mouse CD25⁺CD4⁺ T cells prepared as in d were cultured with rat T cells (1×10⁴) and rat APCs (5×10⁴) in the presence or absence of DTA-1. (f) The responses of the cell mixtures shown in e are presented as percentages of the responses of rat T cells without mouse CD25⁺CD4⁺ T cells. For the four independent experiments, lines connect responses with or without DTA-1 in the same experiments.

CD25⁺CD4⁺ T cells suppress proliferation of other T cells by a cognate interaction on the surface of APCs⁹. Given that GITR is expressed to various degrees on CD25⁺CD4⁺ T cells, CD25⁻CD4⁺ T cells and APCs (Fig. 2), we attempted to determine on which cells GITR is required for abrogating the suppression. In contrast to neutralization of suppression by adding DTA-1 to the culture of these three cell populations (Fig. 3), DTA-1 preincubation of each of three populations, every combination of the two or all three populations failed to abrogate suppression in the subsequent culture (data not shown). The result suggests that the metabolism of GITR molecules on the cell surface may be rapid; alternatively, continuous ligation of GITR molecules may be required for attenuating the suppression. Assuming that the continuous presence of DTA-1 in the culture is required for the abrogation, we then used mouse and rat lymphocytes in combination to discriminate between the cell populations. This is because DTA-1, which is a rat antibody, did not react with rat lymphocytes (Fig. 5c). On rat APCs, mouse CD25⁺CD4⁺ T cells effectively suppressed mouse CD25⁻CD4⁺ T cells stimulated with concanavalin A (Con A), and DTA-1 effectively neutralized the suppression (Fig. 5d). This indicated that GITR expressed on CD25⁺ or CD25⁻ T cells or both is required for the neutralization, but that GITR expression on APCs is not. In addition, mouse CD25⁺CD4⁺ T cells suppressed the proliferation of rat T cells on rat APCs and DTA-1 neutralized suppression (Fig. 5e,f), which indicated that stimulation of the GITR expressed on CD25⁺CD4⁺ T cells suffices to neutralize suppression.

These results indicate that ligation of GITR molecules expressed on CD25⁺CD4⁺ regulatory T cells actively transduces signals to the cells, thereby abrogating their suppressive activity on the activation and proliferation of other T cells.

CD25⁻CD4⁺ T cell–derived GITR^hi cells are not suppressive

Because GITR^loCD4⁺ T cells up-regulated surface expression of GITR after in vitro activation (Fig. 2b), it is possible that any CD4⁺ T cells may become suppressive once they express GITR at amounts equivalent to CD25⁺CD4⁺ T cells naturally present in naïve mice. To examine this possibility, we prepared T cell lines from CD25⁺ or CD25⁻CD4⁺ T cells from normal naïve BALB/c mice by repeated in vitro stimulation as described^9,11. The resulting cell lines—designated as CD25(+) or CD25(–), respectively—expressed GITR in higher amounts than CD25⁺CD4⁺ T cells freshly prepared from normal naïve mice did (Fig. 6). Functionally, the CD25(+) line maintained the anergic and suppressive properties, whereas the CD25(−) line was neither anergic nor suppressive. This indicated that the enhancement of GITR expression per se on T cells is not sufficient for rendering T cells anergic and/or suppressive.

Figure 6: Failure of activated GITR^hi T cells derived from CD25⁻CD4⁺ T cells to exert suppressive activity.

(a) CD25⁺ or CD25⁻CD4⁺ T cells were purified from normal 2-month-old BALB/c mice and stimulated several times with anti-CD3 along with APCs and recombinant IL-2 (50 U/ml). Eight days after the last stimulation, these stimulated T cells derived from CD25⁺ or CD25⁻CD4⁺ T cells—designated as the CD25(+) (thick line) or CD25(–) line (dotted line)—were collected, washed, stained with biotinylated DTA-1 and PE-streptavidin and GITR expression assessed. An irrelevant antibody was used for control staining (thin line). (b) CD25(+) line, CD25(–) line or freshly prepared CD25⁻CD4⁺ T cells (designated CD25⁻ cells) were mixed in the combinations indicated in equal amounts and stimulated for 3 days with anti-CD3 and x-irradiated splenic APCs. One representative of five independent experiments is shown.

GITR^hi T cell elimination or stimulation in vivo

Because of the involvement of GITR in CD25⁺CD4⁺ T cell–mediated suppression, we examined whether depletion of GITR^hi T cells would lead to the activation of self-reactive T cells and consequently the development of autoimmune disease. When BALB/c nu/+ spleen cell suspensions treated with DTA-1 and rabbit complement were transferred to BALB/c nu/nu mice, the majority of recipient mice developed autoimmune gastritis when the mice were examined histologically and serologically 3 months later (Fig. 7a). The gastritis was characterized histologically as loss and damage of gastric parietal cells and chief cells by the infiltration of mononuclear cells into the gastric submucosa and epithelium in a manner similar to that observed in autoimmune gastritis in humans (Fig. 7b,c)²⁶. The incidence of autoimmune gastritis and the titers of autoantibodies specific for gastric parietal cells in these nu/nu mice were comparable with those observed after adoptive transfer of anti-CD25 and complement-treated spleen cells into nu/nu mice (72.7% versus 87.5%) (Fig. 7a). Some mice (∼40%) in both groups also developed histologically evident oophoritis.

Figure 7: Development of in vivo and in vitro autoimmunity by eliminating GITR^hi T cells or administration of DTA-1.

(a) Development of histologically evident gastritis associated with anti–parietal cell autoantibody in BALB/c nu/nu mice that received 3×10⁷ BALB/c spleen cells treated with anti–DTA-1 + complement or anti-CD25 (7D4) + complement. (b) Histology of autoimmune gastritis in a BALB/c nu/nu mouse that received DTA-1 + complement–treated spleen cells. (c) Histology of normal gastric mucosa in a nu/nu mouse that received BALB/c spleen cells treated with complement only (original magnification: ×100). Spleen cells treated with DTA-1 and complement contained <0.1% CD25⁺ cells and 1.6±0.6% (n=5) anti–rat Ig⁺ cells. This indicated the survival of a small number of GITR^lo cells coated with DTA-1 (see also Fig. 2a,b). (d) BALB/c mice were inoculated three times, once a week, from 2 weeks of age with DTA-1 or rat IgG (1 mg per inoculation) and histologically examined 3 months later for the development of gastritis and serologically measured for the titers of anti–parietal cell autoantibody. For histological grading of gastritis, see^4,5,43; briefly here grade 2, macroscopically and microscopically evident gastritis (closed circles); grade 1, histologically evident mild gastritis (gray circles); histologically intact gastric mucosa (open circles). (e) Staining of spleen cells from DTA-1–inoculated BALB/c mice 3 days after the last inoculation. Cells were stained with anti-CD4 and anti-CD25 or anti-CD4 and anti–rat IgG. One representative of three independent experiments is shown. (f) Spontaneous proliferation of spleen cells (1×10⁶/ml) from normal BALB/c mice in the presence of DTA-1 (100 μg/ml) or rat IgG. Data are representative of three independent experiments.

Administration of DTA-1 to young BALB/c mice for a limited time-period (once a week for 3 weeks from 2 weeks of age) also induced, by 3 months of age, histological evidence of autoimmune gastritis accompanied by the appearance of anti-parietal cell autoantibodies in the circulation (Fig. 7d). Other autoimmune diseases were not histologically observed. Inoculated DTA-1 bound to GITR expressed on T cells, as shown by staining lymphocytes with anti–rat IgG (Fig. 7e). It did not, however, reduce the number of T cells or CD25⁺CD4⁺ T cells, which showed that DTA-1 did not eliminate GITR-expressing cells.

Removal of CD25⁺CD4⁺ T cells from normal spleen cells leads to spontaneous proliferation of the residual T cells in vitro, and reconstitution of CD25⁺CD4⁺ T cells suppresses this autologous (or syngeneic) mixed lymphocyte reaction (MLR)^7,8. Addition of DTA-1 to the culture of normal spleen cells elicited the autologous MLR (Fig. 7f). This further supported the data showing that stimulation of GITR by DTA-1 mAb, not elimination of GITR-expressing cells, can elicit autoimmunity.

Thus, not only elimination of GITR-expressing CD25⁺CD4⁺ T cells, but also stimulation of GITR in vivo or in vitro, can cause autoimmunity in otherwise normal mice.

Discussion

We have shown that GITR is predominantly expressed as a 70-kD homodimeric glycoprotein on CD25⁺CD4⁺ peripheral T cells and CD25⁺CD4⁺ SP thymocytes in normal naïve mice. In addition, stimulation of GITR with a specific mAb abrogated CD25⁺CD4⁺ T cell–mediated suppression in vitro and in vivo, which induced autoimmunity.

GITR is structurally similar to other TNFR superfamily members, such as OX40, 4-1BB and CD27, which lack the intracellular death domain required for induction of apoptosis and mediate intracellular signaling by recruiting TRAF proteins to their cytoplasmic tails^{20,24,25,27,28}. These molecules are highly expressed after lymphocyte or T cell activation^24,25,27,28. In addition, like OX40, 4-1BB and CD27, GITR stimulation showed a costimulatory activity for TCR-stimulated naïve resting T cells. In contrast with these common properties, ligation of GITR with DTA-1 mAb neutralized CD25⁺CD4⁺ T cell–mediated suppression, whereas ligation of OX40, 4-1BB and CD27 with available specific antibodies did not. GITR is thus distinct from other TNFR family proteins in the role it plays in CD25⁺CD4⁺ T cell–mediated suppression, although further study is needed to assess the costimulatory activity of GITR on T cells at different activation stages, on T cell subsets and on T cell effector functions (including cytokine production).

Although GITR is expressed on various lymphoid cells at different amounts, GITR expression on CD25⁺CD4⁺ regulatory T cells is essential for abrogating CD25⁺CD4⁺ T cell–mediated suppression. This is because mouse CD25⁺CD4⁺ T cells suppressed the proliferation of rat T cells on rat APCs and the antibody specific for mouse GITR neutralized the suppression. In addition, active signaling through GITR, not mere blockade, seems to be required for abrogating suppression: intact DTA-1 mAb neutralized suppression, but Fab fragments could not. Indeed, the DTA-1 mAb induced NF-κB up-regulation in a GITR-expressing cell line, as engagement of human GITR with its ligand reportedly activates transcription of the gene encoding NF-κB^18,19. Taken together these findings suggest that stimulation of GITR on CD25⁺CD4⁺ regulatory T cells may antagonize the suppressive signals delivered through the TCR and CTLA-4, which may act as a costimulatory molecule for activation of CD25⁺CD4⁺ regulatory T cells^8,14. The results, however, do not exclude the possibility that the costimulatory activity, if any, of GITR may activate other T cells (CD25⁻CD4⁺ T cells or CD8⁺ T cells) and contribute to abrogating CD25⁺CD4⁺ T cell–mediated suppression.

It remains to be determined how GITR stimulation modulates CD25⁺CD4⁺ T cell–mediated suppression. Ligation of CD28 or blockade of CTLA-4 expressed on CD25⁺CD4⁺ regulatory T cells can also abrogate suppression^9,10,11,14. GITR, however, differs from CD28 or CTLA-4 in the mode by which it attenuates suppression. For example, strong ligation of CD28 breaks the anergic state of CD25⁺ CD4⁺ regulatory T cells, whereas GITR ligation does not^9,11. In addition, in contrast with the suppression-attenuating effect of GITR-mediated signals, active signals via CTLA-4 appear to induce suppression by activating CD25⁺CD4⁺ regulatory T cells and blockade of CTLA-4–mediated signaling dampens suppression¹⁴. Although CTLA-4 is constitutively expressed on CD25⁺CD4⁺ regulatory T cells^12,13,14, GITR stimulation did not down-regulate CTLA-4 expression (data not shown). CD25⁺CD4⁺ regulatory T cells express a membrane form of TGF-β, which might mediate the suppression by delivering a suppressive signal to other T cells a cognate manner¹⁵. GITR stimulation, however, did not down-regulate membrane expression of TGF-β (data not shown). Taken together, these results indicate that, although GITR is not an effector molecule mediating suppression, it can play an immunoregulatory role by modulating the suppressive activity of CD25⁺CD4⁺ regulatory T cells. It remains to be shown how GITR-mediated signaling is connected to the CD28 or CTLA-4 signaling pathway, or to signals leading to expression of effector molecules that mediate suppression, such as the membrane form of TGF-β.

High surface expression of GITR is confined to CD25⁺CD4⁺ T cells and CD25⁺CD4 SP thymocytes in normal naïve mice, although activation increases GITR expression in other T cells, B cells and APCs, which are GITR^lo in the resting state. However, activated CD25⁺CD4⁺ T cells or T cell lines derived from CD25⁻CD4⁺ T cells are unable to suppress other T cells irrespective of the increase in expression of GITR. This indicates that high expression of GITR per se is unable to endow T cells with regulatory activity. CTLA-4 also shows a similar expression pattern; that is, CD25⁺CD4⁺ SP thymocytes and CD25⁺CD4⁺ T cells in normal naïve mice constitutively express CTLA-4 and any T cells can express CTLA-4 upon activation^{12,13,14,29,30}. In addition to the similar patterns of GITR and CTLA-4 expression, both CD25⁺CD4⁺ T cells and CD25⁺CD4 SP thymocytes are CD45RB^loCD44^hiCD54^hi(ICAM-1^hi) and CD11a-CD18^hi (LFA-1^hi) in normal naïve mice, indicating that they may be in an "activated, "memory" or "primed" state in the normal internal environment^9,10,11,31. In addition, GITR stimulation can break suppression in both thymocytes and peripheral T cells. These findings suggest that GITR may play a common role in thymocytes and peripheral T cells. They also suggest that CD25⁺CD4⁺ SP thymocytes and CD25⁺CD4⁺ peripheral T cells may be derived from the same lineage, constituting a functionally and phenotypically distinct subpopulation of T cells.

Besides its role in T cell–mediated suppression, GITR may also contribute to the thymic generation and selection of CD25⁺CD4⁺ regulatory T cells. Given that GITR expression makes T cells resistant to TCR-induced apoptosis^17,19, it may, for example, render CD25⁺CD4⁺ thymocytes resistant to thymic negative selection. This process may contribute to the thymic production of regulatory T cells with high-affinity TCRs for self-antigens. Such CD25⁺CD4⁺ regulatory T cells may be easily activated by self-antigens in the periphery and efficient in controlling self-reactive T cells^8,9,11,16. High GITR expression on CD25⁺CD4⁺ regulatory T cells may also contribute to their survival by rendering them resistant to TCR-mediated apoptotic signals. The possible roles of GITR in the thymus and periphery are currently under investigation.

We showed that not only physical elimination of GITR^hi T cells, but also modulation of GITR, can cause autoimmune disease in normal mice. The autoimmune diseases thus induced are immunopathologically similar to those induced by elimination of CD25⁺CD4⁺ regulatory T cells³². This indicates that modulation of GITR-mediated signal transduction to CD25⁺CD4⁺ regulatory T cells can induce autoimmune disease in normal animals presumably by affecting CD25⁺CD4⁺ T cell–mediated control of self-reactive T cells. The results also suggest that blockade of GITR signaling may be able to enhance CD25⁺CD4⁺ T cell–mediated suppression and thereby augment self-tolerance. This could be a useful treatment for autoimmune diseases.

Given that GITR stimulation can attenuate T cell–mediated immunoregulation, what is the physiological role of GITR? Although tissue distribution of GITR ligand (GITRL) expression is not well characterized, the gene encoding GITRL is reportedly transcribed in various tissues, including endothelial cells, but not in lymphocytes^17,18,19. For example, LPS or IFN-γ stimulation of endothelial cells enhances their GITRL expression¹⁷ (and data not shown). It is therefore plausible that, upon local infection by microbes, up-regulation of GITRL expression on endothelial cells may transduce signals to CD25⁺CD4⁺ regulatory T cells and attenuate their suppressive activity, thereby augmenting immune responses to the invading microbes. It is also likely that costimulatory activity of GITR to naïve or activated effector T cells may augment the immune responses.

Elimination or reduction of the number of CD25⁺CD4⁺ regulatory T cells can induce effective tumor immunity in otherwise nonresponding mice by activating tumor-specific as well as nonspecific effector cells⁷. Based on the role of GITR in attenuating the suppressive activity of CD25⁺CD4⁺ regulatory T cells, it may be feasible to use anti-GITR treatment as an immunotherapy for cancer either alone or in combination with anti-CD25 and/or anti–CTLA-4 treatment^7,8,29,30. In addition, anti-GITR treatment in combination with antibodies specific for other members of the TNF-TNFR family—such as CD40, 4-1BB or OX40—may also augment anti-tumor immune responses by activating effector anti-tumor lymphocytes^33,34,35. In addition to tumor immunity, there is accumulating evidence that CD25⁺CD4⁺ regulatory T cells play a crucial role in maintaining transplantation tolerance^2,3,36,37,38. Modulation of GITR-mediated signal transduction may break allograft tolerance as well as self-tolerance; alternatively blockade of GITR-mediated signals may enable CD25⁺CD4⁺ regulatory T cells to maintain allograft tolerance more stably. Thus, GITR may be a suitable molecular target for preventing or treating autoimmune disease, inducing tumor immunity or establishing transplantation tolerance in humans.

Methods

Mice and rats.

Six-week-old female BALB/c, BALB/c athymic nude (nu/nu), BALB/c euthymic (nu/+) and B6 mice were from Japan SLC Co. (Shizuoka, Japan). CD28-deficient mice were from Jackson Laboratory (Bar Harbor, ME)³⁹. Wistar rats were from Charles River Japan (Yokohama, Japan). All mice and rats were maintained in our animal facility; they were treated in accordance with the institutional guidelines for animal care.

Cells.

COS-7 and HEK 293 cells were from T. Suda (Cancer Research Institute, Kanazawa University, Japan) and T. Saitoh (Kyoto University, Kyoto, Japan), respectively. The 18.3.6 T cell hybridoma was established by fusing a BALB/c-derived CD25⁺CD4⁺ T cell clone (data not shown) with TCR-α/β⁻ BW5147 cells with the use of standard procedures.

Preparation of mAbs.

Wistar rats were intraperitoneally immunized three times every 2 weeks with 5×10⁶ cells of the CD25⁺CD4⁺ T cell line (see Results) then intravenously injected with 5×10⁶ cells of the cell line 1 month later. Spleen cells were fused with P3X63Ag8.653 myeloma cells (from the American Type Culture Collection, Rockville, MD) 3 days after the final immunization. For details of screening of the mAbs see the Results. Fab fragments were prepared by digesting antibodies with immobilized papain (Pierce Chemical Co., Rockford, IL) according to the manufacturer's instruction.

Proliferation assays.

Spleen cell suspensions prepared from 8-week-old BALB/c mice were stained with fluorescein isothiocyanate (FITC)–anti-CD25 (7D4) and phycoerythrin (PE)–anti-CD4 (H129.19) or PE–anti-CD8 (53–6.7) (all from PharMingen, San Diego, CA) and sorted by flow cytometry (Epics-ELITE, Coulter Electronics, Miami, FL) as described⁹. Purity of the sorted CD25⁺CD4⁺ T cells, CD25⁻CD4⁺ T cells or CD25⁻CD8⁺ T cells was >94%, >98% or >98 %, respectively. Sorted cells (1×10⁴) and x-irradiated (20 Gy) BALB/c spleen cells (5×10⁴), as APCs, were cultured for 3 days in 96-well round-bottomed plates (Corning Coster, Cambridge, MA) in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum, penicillin (100 U/ml), streptomycin (100 μg/ml) and 50 μM 2-mercaptoethanol (2-ME). Anti-CD3 (mAb 145-2C11) at a final concentration of 5% supernatant, which induced maximum proliferation, or Con A at 3 μg/ml was added to the culture to stimulate cells. Incorporation of [³H]thymidine (1 μCi/well) by proliferating cells was measured during the last 6 h of culture.

To stimulate splenic T cells, B cells or macrophages, spleen cells were stimulated for 3 days with 10 μg/ml of anti-CD3 and 50 U/ml of murine recombinant IL-2 (3.89×10⁶ U/mg, Shionogi, Osaka, Japan), LPS (25 μg/ml, Sigma, St. Louis, MO) or murine recombinant IFN-γ (200 pg/ml, R&D Systems, McKinley, ME), respectively.

T cell lines were prepared from sorted CD25⁺CD4⁺ T cells or CD25⁻CD4⁺ T cells (1γ10⁴/well) and were stimulated weekly with anti-CD3 (mAb 145-2C11) at a final concentration of 5% of supernatant. x-irradiated (20 Gy) syngeneic spleen cells (5×10⁴/well) were used as APCs in the presence of IL-2 (50 U/ml).

Flow cytometric analysis.

FITC–anti-CD25 (7D4), PE-labeled antibodies to CD4 (H129.19), CD8 (53-6.7) and B220 (RA3-6B2) and biotinylated antibodies to CD25 (7D4), CD45RB (16A), CD5 (53-7.3), CD62L (MEL-14), CD4 (GK1.5), CD80 (16-10A1) and CD86 (GL1) were all from PharMingen; PE-labeled anti-F4/80 (A3-1) was from Serotec (Oxford, UK). PE-streptavidin (Becton Dickinson, San Jose, CA), RPE-Cy5–conjugated streptavidin (DAKO A/S, Denmark) and FITC-conjugated F(ab′)2 fragment of anti–rat IgG (Jackson ImmunoResearch, West Grove, PA) were used as secondary reagents. Anti-CD16/32 (PharMingen) was used to block Fc receptors. Anti–TNFR type I (55R-286), anti–TNFR type II (TR75-89), anti–4-1BB (1AH2), anti-CD30 (mCD30.1) and anti-CD27 (LG.3A10) were from PharMingen. Anti-OX40 (OX86) was from Coulter. Polyclonal anti-GITR was from Genzyme-Techne (Cambridge, MA).

Cell surface biotinylation, immunoprecipitation and N-glycosidase treatment.

18.3.6 T cell hybridoma cells (1×10⁷), washed twice with Hank's buffered salt solution (HBSS) and suspended in PBS were incubated with sulfosuccinimidobiotin (Amersham Pharmacia Biotech, Little Chalfont, UK) for 30 min at 4 °C were occasionally shaken, washed and stained with mAb or irrelevant control antibody for 30 min on ice. After washing, cells were lysed with RIPA buffer (50 mM Tris-HCl (pH 7.5), 150 mM NaCl, 1% Triton X-100, 50 mM iodoacetamide, 1 mM PMSF, 10 mg/ml of soybean trypsin inhibitor and 0.1% sodium azide) for 30 min on ice and centrifuged. The cell lysates were incubated with protein G Sepharose for 1 h at 4 °C, then washed four times with 0.5 M NaCl-containing RIPA buffer. Bound proteins were then subjected to SDS-PAGE, blotted onto a PVDF membrane (Millipore, Bedford, MA) and visualized with avidin-peroxidase (Amersham Pharmacia Biotech) and enhanced chemiluminescence (Amersham Pharmacia Biotech). The immunoprepicipates were also treated with 1.5 mM N-glycosidase F (Calbiochem-Novabiochem, San Diego, CA), 167 mM Tris-HCl, 13 mM 1,10-phenanthrorine and 1% Triton X-100 and incubated overnight at 37 °C to remove oligosaccharides.

cDNA cloning.

A plasmid cDNA expression library was constructed with mRNA from 18.3.6 hybridoma cells. Briefly, first-strand cDNA primed with oligo(dT)primer was synthesized using Superscript II reverse transcriptase (Gibco-BRL, Gaithersburg, MD). After treating with RNase H (Takara, Kyoto, Japan), second-strand DNA was synthesized with DNA polymerase I (Takara). After blunting the ends, BstXI adapters were added and cDNA larger than 1 kb was recovered from the agarose gel. The cDNA was ligated to the BstXI site of pME18S vector⁴⁰. Escherchia coli DH10B cells (Gibco-BRL) were transformed with the library DNA by electroporation, which yielded ∼1×10⁷ independent clones. COS-7 cells were transfected with this library using Lipofectamine 2000 (Gibco-BRL). DTA-1–expressing COS-7 cells were enriched by panning on anti–DTA-1–coated dishes; episomal DNA was recovered, amplified in E. coli and retransfected to COS-7 cells⁴¹. After four rounds of these treatments, several clones were sequenced and subjected to BLAST search.

NF-κB reporter assay.

With the use of Lipofectamine 2000 (Gibco-BRL), HEK 293 cells (5×10⁵) were transfected with 0.1 μg of the reporter plasmid expressing firefly luciferase, pNFκB-Luc (Stratagene, La Jolla, CA), 0.1 μg of the internal control vector expressing sea pansy luciferase, pRL-SV40 (from H. Kubota, Kyoto University), together with either pME18S-GITR, empty vector or pMEKK-1 (from E. Nishida, Kyoto Univeristy). Forty-eight hours after transfection, luciferase activity was determined by dual luciferase assay (Promega, Madison, WI), stimulated with soluble anti–DTA-1, control antibody or 100 ng/ml of human TNF-α for 4 h⁴². Some cells were passaged on anti–DTA-1– or control antibody–coated wells 24 h after transfection. The activity of firefly luciferase was normalized against that of sea pansy enzyme.

In vivo transfer of spleen cells.

Spleen cells of 8-week-old female BALB/c nu/+ mice were treated twice with anti-CD25 (mAb 7D4) or DTA-1 mAb and rabbit complement as described^4,5. After washing twice with HBSS, 3×10⁷ cells were intravenously injected into 8-week-old female BALB/c nu/nu mice, which were histologically examined 3 months later.

Histology.

Various organs were fixed with 10% formalin, processed for hematoxylin and eosin staining and histologically graded, as described^4,5,43.