Paradoxical Effect of Reduced Costimulation in T Cell-Mediated Colitis¹²

The B7-1 (CD80) and B7-2 (CD86) costimulatory molecules play overlapping roles by binding to the same two molecules. One binding partner, CD28, is required for costimulation (1, 2, 3). B7-1 and B7-2 also bind to CTLA-4, which has been shown to down-regulate immune responses (4). For example, mice deficient for CTLA-4 exhibit a profound alteration of the immune system characterized by uncontrolled CD4⁺ T cell expansion and infiltration of multiple organs (5, 6, 7). Additionally, a series of studies have suggested an in vivo role for CTLA-4 in peripheral tolerance and anergy induction (8, 9). Doyle et al. (10) reported on the role of CTLA-4 in restricting T cell clonal expansion, which could be a major mechanism for the intrinsic restriction of T cell proliferation and cell cycle-driven differentiation of activated T cells. Furthermore, polymorphisms in the ctla4 gene leading to alteration in its expression are associated with the susceptibility to human autoimmune diseases (11).

Although B7/CD28 costimulation is required for the pathogenesis of autoimmune diseases (12, 13, 14, 15), and even for the lethal lymphoproliferative disease in CTLA-4-deficient mice, up-regulation of the inhibitory receptor CTLA-4 is also dependent on B7/CD28 costimulation (16). B7/CD28 costimulation enhances production of the potent T cell growth factor IL-2 (1). Besides driving proliferation, IL-2 can contribute to cell death, and it has been reported to be important for the induction of CTLA-4 expression (17). Therefore, optimal CD4⁺ T cell activation and IL-2 production might be key elements for the effective regulation of T cell responses through CTLA-4.

Despite sharing the same ligands, the expression patterns of B7-1 and B7-2 differ, and the results from several studies suggest that B7-1 and B7-2 may not play redundant roles (18). For example, in some reports, B7-2 blockade in mice has been associated with decreased Th2 cytokine production, while B7-1 blockade has been associated with decreased Th1 cytokine production (14, 19). B7-2 but not B7-1 was found to be important for oral tolerance induction (20). These differential effects, however, have not been universally observed. For example, the effects of a single B7-1 and a single B7-2 knockout were similar in experimental allergic encephalomyelitis (EAE)⁴ (21).

In this study, we studied the role of B7 costimulation in an experimental mouse model of colitis induced by T cell transfer. In this model, syngeneic CD4⁺ T cells with a naive phenotype (CD45RB^high) are injected into immune-deficient RAG^−/− or SCID mice (22, 23). This causes a colitis several weeks later that bears some features in common with human Crohn’s disease, as differentiated Th1 cells are required for pathogenesis (24), and it is characterized by chronic transmural inflammation in the large intestine. In this study, we have investigated the effect of altering B7 molecule expression in recipients on colitis induction in this transfer model. Although the absence of both B7 molecules abrogated disease, surprisingly, the absence of only one B7 molecule, either B7-1 or B7-2, greatly accelerated disease. Additionally, B7-2-deficient recipients developed colitis even without CD40-CD40L signals that were essential to induce colitis in B7 wild-type recipients. Our data further suggest that this disease acceleration may be mediated by decreased IL-2 synthesis, which may act by causing decreased induction of CTLA-4 expression.

C57BL/6J (B6), B6 background deficient mice for RAG1, B7-1, B7-2, B7-1/B7-2, IL-2, and CD40L were purchased from The Jackson Laboratory and maintained at the La Jolla Institute for Allergy and Immunology. B7-deficient mice were crossed onto the RAG1^−/− or IL-2^−/− background at the La Jolla Institute for Allergy and Immunology. Sentinel mice were found to be free of the Helicobacter species by PCR analysis by the University of Missouri Research Animal Diagnostic Laboratory (Columbia, MO). The Institutional Animal Care and Use Committee approved the experiments.

mAbs for CD4 (GK1.5), CD25 (7D4 and 3C7), CD45RB (16A), CD28 (37.51), CTLA-4 (UC10-4F10-11), IL-2 (JES6-5H4 and S4B6), IL-4 (11B11), IFN-γ (XMG1.2), and TNF-α (MP6-XT22) were purchased from BD Biosciences. Abs for CD8α (53-6.7), CD90.2 (Thy1.2) (53-2.1), NK1.1 (PK136) and TCRβ (Η57-597) were purchased from eBioscience. Hamster anti-CD3ε (145-2C11) for stimulating T cells and anti-CTLA-4 (UC10-4F10-11) were purified from culture supernatant of hybridoma cells purchased from the American Type Culture Collection.

For disease induction experiments, CD4⁺CD45RB^high cells (99.9% pure, containing <0.1% CD4⁺CD25⁺) were sorted on a FACSVantage SE with FACSDiVa option (BD Biosciences) after the preisolation of CD4⁺ cells using CD4 (L3T4) microbeads (Miltenyi Biotec) according to the manufacture’s protocol. Groups of recipients were injected i.v. with 5 × 10⁵ donor lymphocytes. Diseased animals were sacrificed for analysis between 12 and 16 wk after transfer. For detecting CTLA-4 expression in vivo, CD4⁺CD25⁻ cells obtained from B7-1^−/−B7-2^−/− mice were used as naive donor cells (>97% pure CD4⁺CD45RB^high cells containing <0.05% CD4⁺CD25⁺ cells; data not shown). In some experiments, to mildly block IL-2, 3 μg of anti-IL-2 mAb (clone S4B6) in 500 μl of PBS was injected i.p. immediately after the transfer of donor T cells.

Three samples of 2–3 mm obtained from the distal, middle, and proximal portions of the large intestine were processed for H&E staining. Samples were coded and scored by a pathologist blinded to the conditions under which the experiment was conducted. A previously described scoring system (maximum score, 14) was used for the tissue sections (25). Scores (maximum, 14) from three parts were averaged to represent the severity of disease.

For priming of CD4⁺ T cells, 5 × 10⁵ sorted B6 CD4⁺CD45RB^high T cells or B7-1^−/−/B7-2^−/− CD4⁺CD25⁻ cells (described above) were cultured with anti-CD3 mAb and 1 × 10⁷ APC in RPMI 1620 medium supplemented with 10% FCS, 2-ME, and penicillin-streptomycin-glutamine (Invitrogen Life Technologies) in 24-well plates. In some experiments, responder cells were labeled with CFSE according to the protocol published before (26). As APC, T cells were depleted from each mouse spleen using CD90.2 microbeads (Miltenyi Biotec). The flow-through containing CD90.2⁻ cells was gamma irradiated before use. The proliferation index (the average number of divisions that those cells that divided underwent) was calculated using FlowJo software (Tree Star). In some experiments, to block IL-2 signal, anti-IL-2 (clone S4B6) or anti-CD25 (clone 3C7) was added at the beginning of the culture. To analyze costimulation and/or IL-2-dependent CTLA-4 expression, 1 × 10⁶ CD8⁻NK1.1⁻ spleen cells isolated from IL-2^−/−B7-1^−/−B7-2^−/− mice (containing <0.05% CD4⁺CD25⁺ cells) were stimulated with 1.0 μg/ml anti-CD3 mAb. Three different concentrations of anti-CD28 mAb (37.51) and serially diluted rIL-2 (BD Biosciences) were added. For cytokine analysis upon restimulation, primed CD4⁺ T cells were washed and rested in fresh medium for 48 h. Viable cells purified with Lympholyte-M (Cedarlane Laboratories) were restimulated with C57BL/6 macrophage cell line IC-21 (purchased from American Type Culture Collection) and anti-CD3 mAb. In the other experiment, primed T cells were washed with medium and 1 × 10⁵ primed T cells were directly cultured with 1.0 μg/ml anti-CD3 mAb in the presence of 1 × 10⁵ total spleen cells from B6 RAG1^−/− mice in a 96-well round-bottom plate. After 3 days of culture, supernatants were collected for conventional ELISA.

To obtain large intestinal lymphocytes, large intestines were opened longitudinally, washed, and minced on a bacterial-grade plate and digested by incubating with HBSS (Invitrogen Life Technologies) containing 5% FCS and 1.5% collagenase type IV (Sigma-Aldrich) at 37°C for 40 min with shaking (200 rpm). Cells were washed and resuspended with 20 ml of 35% Percoll gradient followed by a centrifugation at 1700 rpm for 20 min. Pellets were further run over a 70–44% discontinuous Percoll gradient. Cells at the interface between 70 and 44% Percoll were collected as large intestine cells. In vitro-cultured cells were washed and used directly for the analysis. For CTLA-4 staining, cells were stained for surface molecules, fixed/permeabilized with Cytofix/Cytoperm (BD Biosciences), and stained with PE-conjugated anti-CTLA-4 Ab (5 μg/ml). For detecting apoptotic cells, biotinylated annexin V was used with allophycocyanin-conjugated streptavidin (BD Biosciences) according to the manufacturer’s protocol. For ex vivo intracellular cytokine staining, CD4⁺ cells were isolated from spleens of the recipients using CD4 microbeads. CD4⁺ cells were restimulated with PMA (2 ng/ml) and ionomycin (0.5 μg/ml) for 5 h with brefeldin A for the last 2 h. After the stimulation, cells were stained for intracellular cytokines using a Fixation/Permeabilization Solution kit from BD Biosciences.

DNA content was analyzed using propidium iodide. In brief, CD4⁺ cells were enriched from the spleens of the recipients using CD4 microbeads, followed by surface staining for TCRβ. The cells were rinsed with PBS and fixed with ice-cold 70% ethanol for >1 h followed by RNase treatment with PBS containing 100 μg/ml RNase and 40 μg/ml propidium iodide at 37°C for 30 min. Samples were analyzed by flow cytometry.

To investigate the role of B7/CD28 costimulation in the pathogenesis of experimental colitis in mice, we transferred sorted, CD4⁺CD45RB^high T cells from wild-type C57BL/6 (B6) spleens into B6 RAG1-deficient (RAG1^−/−) mice. RAG1^−/− recipients included mice that express both B7-1 and B7-2 (B6 RAG1^−/−) and mice that express only a single type of B7 molecule (B7-1^−/− RAG1^−/− or B7-2^−/− RAG1^−/−) or neither B7 molecule (B7-1^−/−B7-2^−/− RAG1^−/−). As has been reported, B6 RAG1^−/− recipients that were reconstituted with CD4⁺CD45RB^high T cells developed the symptoms of chronic colitis, such as body weight loss and diarrhea, 6–10 wk after the T cell transfer (Fig. 1,A). A critical role of B7/CD28 costimulation in the pathogenesis of colitis is illustrated by the finding that B7-1^−/−B7-2^−/− RAG1^−/− recipients in the experiments shown did not exhibit colitis symptoms in 16 wk of observation. Surprisingly, however, recipients lacking either B7-1 or B7-2 developed the symptoms of wasting disease as early as 10–14 days after T cell transfer (Fig. 1,A), although in some sets of mice the acceleration was even more dramatic in B7-2^−/− RAG1^−/− recipients. Histological analyses of the large intestines 14 days after transfer demonstrated massive cell infiltration and mucosal hyperplasia in both B7-1^−/− RAG1^−/− and B7-2^−/− RAG1^−/− mice, whereas only focal cell infiltration was seen at this relatively early time point in B6 RAG1^−/− recipients (Fig. 1 B). The inflammation at 2 wk in the single B7-deficient RAG1^−/− recipients was characterized by an infiltrate of mononuclear cells and epithelial hyperplasia. Residual mucin-producing cells in the markedly inflamed mucosa suggested rapid onset of inflammation induced by donor T cells. Severe inflammation was restricted to the large intestine, although as in B6 RAG1^−/− mice, mild cell infiltration was seen in other organs such as the small intestine, pancreas, stomach, and skin (data not shown). Interestingly, despite the rapid onset of disease, B7 single-deficient RAG1^−/− recipients survived until 12 wk after transfer when they were sacrificed for analyses. The symptoms of chronic colitis were consistently observed, however, throughout the entire 12-wk observation period.

The CD40-CD40L receptor-ligand pair plays important roles in enhancing T cell function, especially Th1 differentiation. In multiple colitis models, including the T cell transfer model that we are using in this study, the CD40-CD40L interaction has been reported to be indispensable for the induction of disease (27, 28, 29, 30). We transferred CD40L-deficient CD4⁺CD45RB^high cells into B7 wild-type and B7-2-deficient RAG1^−/− recipients. As has been reported, B7 wild-type RAG1^−/− recipients did not develop wasting disease by 9 wk after transfer (Fig. 2,A), although histological analysis demonstrated scattered mononuclear cell infiltration (Fig. 2,B). In contrast, all seven B7-2^−/−RAG1^−/− recipients that were transferred with the same donor cells developed severe chronic colitis quickly, as early as when they received wild-type donor T cells (Fig. 2). These data demonstrate a qualitative difference in colitis induction in the B7-2^−/−RAG1^−/− recipients, such that not only is disease accelerated, but it can occur in the absence of signals that are absolutely required for pathogenesis in B7 wild-type recipients.

Despite the rapid proliferation and activation of the donor T cells (31), colitis takes weeks to develop in B6 RAG1^−/− recipients. This suggested that the induction of disease was slowed from the maximum possible rate in B6 RAG1^−/− mice due to regulation, and that this regulation apparently was reduced or absent in B7 single-deficient recipients. We further hypothesized that insufficient costimulation with either B7-1 or B7-2 might fail to fully induce CTLA-4, which could be responsible for the postulated regulation. Consistent with this, CD28 cross-linking, along with TCR stimulation, has been reported to play a role in up-regulating CTLA-4 (16).

To address these issues, we first used an in vitro culture system in which CFSE-labeled, CD4⁺CD45RB^high T cells from B7-1^−/−B7-2^−/− mice were stimulated with anti-CD3 mAb and T cell-depleted spleen cells as APC. APC from B6, B7-1^−/−, B7-2^−/−, or B7-1^−/−B7-2^−/− mice were tested in parallel, representing the wild-type and various degrees of reduced B7-dependent costimulation. After 4 days of culture, responder CD4⁺ T cells that were stimulated with wild-type APC divided two to seven times with 1.0 μg/ml anti-CD3 mAb. In contrast, most of the CD4⁺ T cells that were stimulated without any B7 costimulation failed to divide and underwent apoptotic cell death, as evidenced by the majority that were annexin V positive (Fig. 3,A). Cells that remained alive under these conditions may be refractory to stimulation, because they did not proliferate despite TCR stimulation. By contrast, responder T cells that were stimulated with either B7-1^−/− or B7-2^−/− APC exhibited comparable proliferation to those stimulated with wild-type APC according to the proliferation indexes calculated from CFSE dilution (Fig. 3,B), without evidence for significantly increased populations of apoptotic cells or cells refractory to CD3 stimulation. Interestingly, however, actively proliferating T cells in the culture with either B7-1^−/− or B7-2^−/− APC did not up-regulate CTLA-4 as much (Fig. 3,C), even with a higher concentration of anti-CD3 Ab (Fig. 3 D). The effect was even more pronounced with B7-2^−/− than with B7-1^−/− APC. This indicates that although costimulation with either B7-1 or B7-2 alone was sufficient for permitting cells to respond and rescuing primed T cells from undergoing apoptosis, costimulation with both B7-1 and B7-2 was required for optimal expression of CTLA-4.

Previous reports have suggested that B7/CD28 costimulation and IL-2 are two important elements for the increase in CTLA-4 expression upon T cell activation (16, 17). Because IL-2 production is highly enhanced by B7/CD28 costimulation (1, 2, 32, 33), we reasoned that costimulation by only one B7 molecule might result in impaired induction of CTLA-4 expression in part due to suboptimal IL-2 production. To test this hypothesis, we first analyzed how B7/CD28 costimulation and IL-2 cooperatively up-regulated CTLA-4 upon activation. We stimulated naive splenic CD4⁺ T cells obtained from IL-2, B7-1 and B7-2 triple gene knockout mouse (containing <0.1% CD4⁺CD25⁺ T cells; data not shown) in the presence of three different concentrations (5, 50, and 500 ng/ml) of anti-CD28 mAb, as a source of costimulation. In each condition, serially diluted rIL-2 was also added. As shown in Fig. 4,A, costimulation and IL-2 independently but cooperatively up-regulated CTLA-4 expression. This was especially evident in the lower range of IL-2 concentrations. We also tested the role of IL-2 in up-regulating CTLA-4 expression by endogenous B7 molecules by Ab blocking. Addition of either anti-IL-2 or anti-IL-2Rα mAb dramatically reduced CTLA-4 expression, although blocking IL-2 signals only slightly decreased cell division (Fig. 4 B).

Because optimal IL-2 production itself requires CD28 costimulation, we measured IL-2 production of T cells that are cultured with B7 wild-type or deficient APC. Despite comparable proliferation, as shown in Fig. 3,A, IL-2 in the supernatant was profoundly impaired in the culture with the B7 single-deficient APC (Fig. 4 C). IL-2 production was augmented by adding a blocking anti-CTLA-4 mAb to cocultures with wild-type APC, confirming the role of CTLA-4 in down-regulating IL-2 production. These data indicate that costimulatory signals to T cells by a single B7 molecule profoundly impair CTLA-4 expression not only because of insufficient costimulation itself, but also because of a critical reduction in IL-2 production.

Despite a systemic priming and activation of naive donor CD4⁺ T cells in the transfer model of colitis, the most significant inflammation is observed in the large intestine of the immune-deficient recipient mice. Systemically primed T cells probably require additional stimulation by Ags derived from bacterial flora in the large intestine to become effector cells and to induce inflammation locally. To address how T cells primed by B7 single-deficient APC can cause more rapid colitis induction, we performed in vitro experiments to analyze IL-2 and IFN-γ production upon secondary stimulation by anti-CD3 mAb, in the presence of the macrophage cell line IC-21 (Fig. 5,A), or fresh spleen cells from B6 RAG1^−/− mice (Fig. 5,B). After 3 days of secondary culture, T cells that were primed with B7 single-deficient APC produced more IFN-γ in both experiments, compared with those that had been primed with wild-type APC (Fig. 5). Interestingly, they also produced more IL-2 upon restimulation, although, as noted above, IL-2 production was decreased during the priming phase with B7 single-deficient APC. Addition of a blocking anti-CTLA-4 mAb (Fig. 5,A) or anti-IL-2 mAb during priming with wild-type APC (Fig. 5 B) caused a similar enhancement of cytokine production upon restimulation. These data suggest that primed T cells that failed to up-regulate CTLA-4, because of reduced costimulation by a single B7 molecule, can respond with increased cytokine production upon restimulation.

In the majority of B7 single-deficient RAG1^−/− recipients that received naive CD4⁺ T cells, thickening of the large intestine was found by 10 days after T cell transfer. At the same time, a characteristic feature that correlated with the rapid induction of colitis was an enlarged spleen and mesenteric lymph nodes (LN) as early as 7–10 days after transfer. Intracellular cytokine staining of restimulated spleen cells that were recovered 7 days after transfer demonstrated that the expanded spleen T cells from B7 single-deficient RAG1^−/− recipients exhibited increased percentages of Th1 cytokine producing cells, especially TNF-α and IL-2, compared with those from B6 RAG1^−/− recipients (Fig. 6,A; TNF-α, 36.0 and 49 compared with 14.9%; IL-2, 49 and 27.5 compared with 19.8%). In both types of recipients, many of the TNF-α highly positive CD4⁺ T cells also were IL-2 positive. The increased IL-2 production in B7 single-deficient RAG1^−/− recipients was similar to the increase observed in vitro when T cells primed with single B7-deficient APC were restimulated. Furthermore, DNA content analysis using propidium iodide demonstrated that 2- and 6-fold more of the TCRβ⁺ cells in B7-2^−/−RAG1^−/− recipients were in the S-G₂-M phases of the cell cycle in the spleen and mesenteric LN, respectively (Fig. 6 B). These data indicate that T cells primed in B7-2^−/−RAG1^−/− recipients were capable of more active proliferation upon restimulation compared with those in B6 RAG1^−/− mice.

The increased in vivo T cell activation in B7-deficient RAG1^−/− recipients could be due to reduced induction of CTLA-4, as we observed in vitro. To determine whether the induction of CTLA-4 by activated CD4⁺ T cells is affected by the absence of one B7 molecule in vivo, we analyzed TCRβ⁺ donor cells that were recovered from LN at an early time point during priming, at 2 days after transfer. Donor T cells in peripheral and mesenteric LN of B7-2^−/−RAG1^−/− recipients expressed significantly lower levels of CTLA-4 than those in B6 RAG1^−/− recipients (Fig. 6 C).

To support our hypothesis that insufficient production of IL-2 during priming, because of reduced CD28 costimulation, causes a subsequent uncontrolled IL-2 synthesis and acceleration of pathogenesis, we blocked IL-2 signaling by injecting anti-IL-2 mAb immediately after the transfer of CD4⁺CD45RB^high cells. After a single injection of a limited amount (3 μg/mouse) of anti-IL-2 mAb, B6 RAG1^−/− recipients developed accelerated colitis demonstrated by an enhanced accumulation of total cells and TCRβ⁺ donor cells in the spleen and the large intestinal lamina propria at 2 wk (Fig. 7,A). Histological scores of the colon confirmed the induction of accelerated colitis with injection of anti-IL-2 mAb (Fig. 7, B and C). Injection of a higher amount of anti-IL-2 mAb (100 μg/mouse) or later anti-IL-2 mAb injections did not accelerate colitis induction (data not shown). This suggests that blocking IL-2 during priming is important for disease acceleration and a possible later requirement for IL-2 in disease acceleration.

In the present study, we have demonstrated the indispensable roles of costimulation by both B7-1 and B7-2 in the regulation of activation-induced T lymphocyte proliferation following cell transfer. Reduced costimulation in either B7-1- or B7-2-deficient recipients resulted in a dramatic acceleration of colitis induction and Th1 cytokine production following transfer of CD4⁺CD45RB^high cells, and consistent with this, T lymphocytes stimulated in vitro with single B7-deficient APC produced increased levels of cytokines following in vitro restimulation.

The transfer experiment with CD40L-deficient T cells emphasized the difference between B7 wild-type recipients and B7-2^−/−RAG^−/− recipients in the regulation of activated T cells. We hypothesize that the critical role of CD40-CD40L interaction might be to break the suppression that is induced during activation of CD4⁺ T cells by up-regulated CTLA-4 expression. It has been reported that CTLA-4 and CD40L are reciprocally regulated in the Th1 response, as CTLA-4 blockade up-regulated CD40L expression by T cells. Moreover, cross-linking of CD40 on CD8⁺ dendritic cells in the culture down-regulated surface CTLA-4 expression of T cells. We speculate that the CD40-CD40L interaction may be less essential for T cell differentiation in the B7 single-deficient recipients, because T cells that were activated with a single B7 molecule exhibited a reduced expression of CTLA-4. Regardless of the exact mechanism, the data are consistent with an altered state of the activated T cells and the APC in the B7 single-deficient recipients that develop rapid colitis. Although it has been understood that the CD40-CD40L interaction is critical for macrophage activation (34) and Th1 CD4⁺ T cell differentiation by enhancing IL-12 expression by APC (29), Strom and colleagues (35) demonstrated that CTLA-4 signaling was required for the optimal induction of allograft tolerance by the anti-CD40L treatment in the combined donor-specific transfusion model (36 ). This suggests that CTLA-4 is still required for regulating T cell activation even in the absence of CD40-CD40L interaction.

Although the disease-accelerating effects of a single B7 deficiency were dramatic in the colitis model, this effect was not so evident in other cases, such as EAE (21). Furthermore, differential effects of blocking B7-1 or B7-2 have been reported to give divergent effects in some different experimental systems, such as EAE (14), diabetes (19), and also colitis (20). In the transfer model of colitis, while the absence of either B7-1 or B7-2 caused an acceleration of disease, a deficiency of B7-2 caused a more stable and profound phenotype compared with a B7-1 deficiency. Some of these divergent outcomes may reflect the different experimental systems that were studied. In the colitis model, unlike EAE, naive T cells are primed in the absence of adjuvant, which should increase B7 levels. The absence of adjuvant in colitis could render priming and subsequent CTLA-4 induction more sensitive to the absence of one B7 molecule. In one previous report (15), however, it was found that colitis caused by CD4⁺CD45RB^high T cell transfer could be inhibited by anti-B7-1 but not by anti-B7-2 Abs. The amount of B7-2 Ab used may not have been sufficient and it is possible that some of the Abs have agonistic properties when administered in vivo.

We considered that CTLA-4, whose induction is dependent on B7-CD28-mediated signals (16), might provide one means to regulate T cell expansion and delay the onset of disease. A crucial role for CTLA-4 in regulating CD4⁺ T cells has been demonstrated in gene-deficient mice (5, 6, 7). In addition to its negative regulatory effects in vitro on IL-2 transcription (37, 38) and cell cycle progression (39, 40), a series of studies have demonstrated the important in vivo functions of CTLA-4 in anergy induction (8, 41) and restriction of clonal expansion of T cells (10). Furthermore, restriction of continuous cell cycle progression and proliferation by CTLA-4 seems to inhibit activated T cells from differentiating into effector T cells (10, 42).

In accord with a role for CTLA-4 in limiting the progression of pathogenesis, even in the absence of CD25⁺ regulatory T cells, naive CD4⁺ T cells proliferated but did not fully increase their expression of CTLA-4 when cultured with APC that lacked either B7-1 or B7-2. Additionally, donor CD4⁺ T cells that were recovered from the LN of B7-2^−/−RAG1^−/− mice 2 days after transfer exhibited reduced levels of CTLA-4 expression. Although both B7 costimulatory molecules are required, their mechanisms of action might not be the same. Effective CTLA-4 signaling has been reported to require the formation of a multivalent lattice (43, 44), depending upon the homodimerization of B7 molecules. A higher tendency for the homodimerization of B7-1 molecules has been observed, and, consistent with this, it has been reported that B7-1 is more effective at recruiting CTLA-4 into the immunological synapse (45). Therefore, constitutive B7-2 expression by APC in B7-1^−/− mice may allow some CTLA-4 induction during priming but insufficient CTLA-4 engagement. By contrast, because B7-1 expression by APC is inducible rather than constitutive, in B7-2^−/− mice, the efficiency of CTLA-4 induction during priming may be less effective. These differences also could account for the more consistent and pronounced activation and disease acceleration in B7-2^−/−RAG1^−/− recipients.

In the present study, costimulation by the single B7-deficient APC was sufficient for near-normal proliferation and survival of the primed T cells in vitro, but IL-2 production was seriously compromised. Our data indicate that IL-2 is likely to be a critical factor in the up-regulation of CTLA-4 expression, allowing the activated T cells to escape regulation. IL-2 has complex effects, some of which are independent of CTLA-4, since it influences both T lymphocyte growth and activation-induced cell death. Despite decreased IL-2 production following priming by B7 single-deficient APC, activated T lymphocytes that accumulated in the spleen of B7 single-deficient RAG1^−/− recipients 7 days after transfer were actively proliferating and capable of producing increased levels of IL-2, in addition to TNF-α, which is associated with colitis induction in this model (24). Similarly, when restimulated in vitro after priming with single B7-deficient APC, CD4⁺ T cells produced increased rather than decreased levels of IL-2. Consistent with a critical role for IL-2 during priming in limiting pathogenesis through CTLA-4 induction, and perhaps through other pathways as well, blocking of IL-2 signaling during the priming phase with a limited amount of anti-IL-2 mAb dramatically accelerated colitis induction. In contrast, in vivo blocking with anti-CTLA-4 mAb did not have a similar effect on colitis induction (data not shown). The previous results from in vivo blocking with CTLA-4 mAbs have not been consistent, with one study finding increased weight loss after CTLA-4 Ab treatment (15) and another reporting no effect when pathogenic T cells were transferred in the absence of regulatory T cells (46). We found that blockade of CTLA-4 with mAb up-regulated IL-2 in our culture system during priming. Furthermore, donor T cells were greatly decreased in the LN 7 days after transfer when mice were treated with anti-CTLA-4 mAb (data not shown). It is therefore possible that CTLA-4 blocking caused overexpression of IL-2, with the increased IL-2 sensitizing the activated T cells for activation-induced cell death (47, 48). Regardless, although the combined in vitro and in vivo data strongly suggest that the reduced IL-2 produced following costimulation in the absence of a single B7 molecule acts in part through reduced CTLA-4 expression to accelerate colitis, our in vivo data only demonstrate a correlation among single B7 costimulation, reduced CTLA-4, and accelerated disease.

Regulating autoreactive CD4⁺ T cells is critical for the maintenance of peripheral tolerance. Apart from regulatory T cells that act in trans, activation-induced regulation of CD4⁺ T lymphocytes is an additional important system for the maintenance of peripheral tolerance. IL-2 production by activated T cells could play a role as a sensor for the host to trigger multiple activation-induced regulatory systems, including activation-induced cell death and negative regulation through CTLA-4. Consistent with its important role, CTLA-4 blockade therapy in patients with metastatic melanoma causes grade III/IV autoimmunity, including colitis, dermatitis, and others (49). Insufficient costimulation allows activated T lymphocytes to escape from this system because of impaired secretion of IL-2 by primed T cells, however, which could be a critical mechanism for the pathogenesis of autoimmune diseases such as inflammatory bowel disease. Therefore, the results here provide a caution for strategies for immune therapy mediated by blockade of costimulation. Although a complete block, as observed in the B7-1^−/−B7-2^−/−RAG1^−/− recipients, will decrease T cell priming, a partial block may allow primed, autoreactive T cells to escape some forms of immune regulation, especially under lymphopenic conditions such as those following immune suppression of transplant patients.

We thank Drs. Michael Croft and Timothy Denning for suggestions and discussions and Drs. Olga Turovskaya and Raziya Shaikh and Chris Lena for technical help.

The authors have no financial conflict of interest.