Modulating Protective and Pathogenic CD4⁺ Subsets via CD137 in Type 1 Diabetes

NOD and NOD-scid colonies were maintained under specific pathogen-free conditions in our animal facilities. Mice were handled in accordance with the institutional animal care guidelines of the University of Pittsburgh School of Medicine. Urinary glucose analysis was performed using Tes-tape (Shionogi, Osaka, Japan) once a week. Plasma glucose levels were determined when glycosuria was detected, and mice with a blood glucose level above 300 mg/dl were considered diabetic. Antibodies against mouse CD3, CD4, CD8, CD25, CD28, CD49b, and T-cell β-receptor were purchased from BD Bioscience. Anti-Foxp3 antibody was purchased from eBioscience. Anti-mouse CD137 antibody (3H3) was produced by R.S. Mittler as previously described (22).

The pancreas of NOD mice was removed at the indicated times, fixed in 10% formaldehyde, and embedded in paraffin. Thin sections at six levels, 150 μm apart, were cut for staining with hematoxylin-eosin to evaluate islet-infiltrating cells by light microscopy. At least 35 islets from each recipient mouse were observed and scored using the following criteria: grade 0, islets free of insulitis; grade 1, peri-insulitis; grade 2, intra-insulitis with mononuclear cell infiltration of <50% of the area of each islet; and grade 3, intra-insulitis with mononuclear cell infiltration of >50% of the area of each islet.

The spleen from each mouse was removed aseptically and minced. After lysing erythrocytes, the cells were washed three times with PBS. To sort CD4⁺, CD4⁺CD25⁺, CD4⁺DX5⁻, or CD8⁺ cells, splenocytes were prepared by magnetic separation using a magnetic sorting system (Miltenyi Biotec, Auburn, CA) according to the manufacturer’s instructions. The purities of sorted cells were >90%. In other experiments, purified CD4 populations were obtained by sorting using flow cytometry. To stimulate sorted splenocytes, cells were suspended in RPMI 1640 supplemented with 10% heat-inactivated fetal bovine serum (Gibco-BRL) and penicillin/streptomycin (Gibco-BRL). The CD4⁺ cells (1 × 10⁶) resuspended in 2 ml culture medium were transferred to each well of a 24-well anti-CD3 antibody precoated plate with or without soluble anti-CD28 antibody (1 μg/ml). The cells were cultured for 72 h at 37°C in a humidified 5% CO₂ incubator, and the supernatant was collected at the end of culture.

For suppression experiments, CD4⁺CD25⁺ cells, 5 × 10⁴ of effector or suppressor cells, or both were seeded to 96-well plate with soluble anti-CD3 antibodies and irradiated splenocytes as APCs.

Cells were incubated with Fc blocker (Pharmingen) and stained with labeled antibodies for 25 min at 4°C. Samples were analyzed on an FACS Caliber (Becton Dickinson, Miami, FL). Intracellular cytokine staining of Foxp3 was performed according to the manufacturer’s instructions.

Cytokines (interferon-γ [IFN-γ], IL-4, and IL-10) were measured by enzyme-linked immunosorbent assay (ELISA) as previously described (31). Briefly, flat-bottom 96-well plates were coated with anti–IFN-γ, anti–IL-4, or anti–IL-10 antibodies (Pharmingen), and supernatants were added. After incubation, biotinylated anti-cytokine antibodies and then europium-avidin solution (Perkin Elmer, Turku, Finland) were added. After adding enhancement solution (Perkin Elmer), plates were measured with an ELISA reader (Perkin Elmer).

Results are presented as means ± SE. Differences between the two groups were analyzed using Mann-Whitney U test for nonparametric unpaired observations. Log-rank tests were used to compare the incidence of diabetes between two groups using JMP IN software (SAS Institute) and Graph-pad (Graph-pad Software).

To test the effect of anti-CD137 on the course of type 1 diabetes in NOD mice, we treated 6-week-old female NOD mice with a total of three 200-μg doses of anti-CD137 antibody (clone 3H3) (22) administered over a 9-week interval every 3rd week. Control mice received PBS alone; mice were followed for glycosuria. The administration of anti-CD137 antibody significantly suppressed type 1 diabetes (Fig. 1; P = 0.0005).

It has previously been published that when placed on the NOD background, the Idd9.3 region (containing the B10 allele of Cd137) decreased diabetes incidence but did not result in decreased insulitis (6). We tested whether anti-CD137 treatment ameliorated insulitis; as shown in Fig. 2A, anti-CD137 did not suppress pancreatic islet infiltration (“insulitis”) in mice protected from diabetes. To further evaluate the progression of insulitis in treated mice, NOD mice were treated at 6 and 9 weeks of age with 200 μg anti-CD137 antibody or with PBS alone, and at 10 weeks of age the pancreas was scored for insulitis. As shown in Fig. 2B, the insulitis score between treated and control NOD mice was no different. Therefore we conclude that anti-CD137 treatment prevents type 1 diabetes but not by preventing insulitis.

The presence of insulitis in the treated mice raised the question of whether anti-CD137 directly or indirectly affected pathogenic T-cells, for example, by inducing deletion or regulation of autoreactive T-cells. We addressed this possibility by transferring 20 × 10⁶ splenocytes from anti-CD137–treated nondiabetic NOD mice, aged over 200 days, into NOD-scid recipients. Although the donor mice were protected from diabetes, seven of nine NOD-scid recipients developed type 1 diabetes, indicating that the donors retained pathogenic T-cells (Fig. 2C). This result demonstrated that pathogenic T-cells were not destroyed or inactivated in anti-CD137–treated NOD mice.

The above experiments raised the question: why did splenocytes from aged, but nondiabetic, treated mice transfer disease? We reasoned that, because some anti-CD137–treated mice eventually develop diabetes (Fig. 1), the regulatory effect likely decreased over time. As a result, when splenocytes from aged treated mice were transferred into NOD-scid, homeostatic expansion of autoreactive cells in the NOD-scid lymphopenic environment (32) might overwhelm any remaining regulatory effect. The hypothesis resulting from this scenario is that regulatory cells from recently treated mice should be able to ameliorate disease in the NOD-scid transfer model. To test this hypothesis we used a pre-diabetic NOD transfer system, as previously described (33), to investigate the protective cell populations in the pre-diabetic mice. CD4⁺ cells from anti-CD137–treated 10-week-old pre-diabetic NOD mice, but not from PBS-treated age-matched controls, significantly delayed disease transfer into NOD-scid recipients (Fig. 3; P = 0.0016). (Note the donor population used in this model is pre-diabetic; as shown in Fig. 3, diabetic splenocytes transfer disease more rapidly.) Therefore freshly recruited CD4⁺ regulatory cells can counteract the disease process even in the NOD-scid transfer model. Notably, however, in the NOD-scid transfer model, the protective effect of anti-CD137 antibody treatment was less than in “normal” NOD mice (Fig. 1), likely as a result of rapid homeostatic effector expansion in the NOD-scid environment. It has previously been reported that anti-CD137 treatment can induce CD8⁺ regulatory cells as a protective mechanism (17). To test the hypothesis that regulatory CD8⁺ cells could have suppressor function in treated NOD mice, CD8⁺ cells from either anti-CD137–treated or PBS-treated control NOD mice were transferred to NOD-scid recipients with diabetic CD4⁺ cells. CD8⁺ cells from treated NOD mice, however, did not have protective ability, indicating that CD8⁺ regulatory cells are unlikely to participate in anti-CD137–induced diabetes protection in NOD mice (data not shown). These results are in agreement with data of Foell et al. (22), which showed that anti-CD137–mediated reversal of systemic lupus erythematosus (SLE) was independent of CD8⁺ T-cells.

The co-transfer experiment shown above encouraged us to evaluate the functional properties of CD4⁺ cells induced by anti-CD137 antibody treatment. As shown in Fig. 4, sorted CD4⁺ splenocytes from anti-CD137–treated mice produced significantly more IL-4 and IL-10 and less IFN-γ than CD4⁺ cells from untreated mice.

To determine the source of IL-4/IL-10–producing cells, we examined Dx5⁺ NKT cells, because it has been recently shown that CD137^−/− mice have decreased numbers of CD3⁺/Dx5⁺ NK T-cells, and it is well established that NKT cells produce substantial quantities of IL-4 and IL-10 (34,35). We hypothesized that one mechanism of anti-CD137–mediated protection occurred through its effects on NK T-cell numbers and function. To test this hypothesis, we isolated Dx5⁺ cells from anti-CD137–treated and untreated NOD mice (at 10 weeks of age as above) and examined the numbers of positive cells. The number of Dx5⁺ T-cells increased in anti-CD137–treated NOD spleen and peripheral lymph node (P = 0.0045 and 0.0012, respectively; Fig. 5A).

Next, we asked to what extent the change in cytokine production in the CD4⁺ population was due to Dx5⁺ cells. Stimulated CD4⁺ cells from anti-CD137–treated spleens produced more IL-4 and IL-10 than CD4⁺ cells from PBS-treated mice (Fig. 5B, solid bars), and removing DX5⁺ cells from the CD4⁺ population eliminated the enhanced level of cytokine production; the CD4⁺DX5⁻ cells from treated mice had the same production of IL-4 and IL-10 as untreated CD4⁺ cells (Fig. 5B, hatched bars).

Finally, to test whether Dx5⁺ NKT cells had a protective effect in vivo, we used two approaches. First, CD4⁺Dx5⁺ cells were purified using sorting by flow cytometry. NOD-scid recipients then received either CD4⁺DX5⁻ populations alone or CD4⁺DX5⁻ cells plus CD4⁺Dx5⁺ cells from anti-CD137–treated pre-diabetic mice, along with diabetic CD8⁺ cells. As shown in Fig. 5C, there was no significant benefit in diabetes prevention obtained by adding CD4⁺Dx5⁺ to CD4⁺Dx5⁻ cells, compared with CD4⁺Dx5⁻ cells alone. In the second approach, NOD-scid mice received either whole CD4⁺ (including Dx5⁺) cells from untreated pre-diabetic NOD mice or CD4⁺DX5⁻ cells, along with diabetic CD8⁺ cells. Again, as shown in Fig. 5D, there was no significant difference in diabetes onset between the two groups, indicating no treatment effect of depleting Dx5⁺ cells in the transfer system. Both CD4⁺Dx5⁺ and CD4⁺Dx5⁻ transfers from anti-CD137–treated mice delayed diabetes compared with CD4⁺ or CD4⁺Dx5⁻ populations from untreated mice (Fig. 5C compared with D, P = 0.0023), but neither Dx5⁺ nor Dx5⁻ fractions from untreated mice delayed diabetes onset (Fig. 5D; compare with PBS-treated CD4 cells in Fig. 3). This result showed that there was no protective effect of the CD4⁺Dx5⁺ population in the NOD-scid transfer system (although it remains possible that a protective effect of Dx5 cells exists and is not shown by our experiments) but confirmed that there must be at least one protective CD4⁺ cell population in the anti-CD137–treated CD4⁺Dx5⁻ cell population.

The transfer of type 1 diabetes from splenocytes of anti-CD137–treated nondiabetic mice and the protective capability of anti-CD137–treated CD4⁺ subsets indicated that anti-CD137 treatment did not destroy/inactivate pathogenic CD4⁺ T-cells but acted on regulatory cells. The presence of a protective effect in the treated CD4⁺Dx5⁻ subset led us to investigate other regulatory CD4 subsets. It has been well established that CD4⁺CD25⁺ regulatory cells can modulate the development of type 1 diabetes in the presence of pathogenic T-cells. Quantitative and qualitative defects of CD4⁺CD25⁺ Treg cells have been reported in NOD mice, and CD4⁺CD25⁺ Tregs can prevent type 1 diabetes in NOD mice (25,26). Therefore, we investigated the quantitative and qualitative phenotype of CD4⁺CD25⁺ Treg cells in anti-CD137–treated NOD mice.

As shown in Fig. 6A, we found that naïve CD4⁺CD25⁺ cells in NOD mice upregulate CD137 compared with CD25⁻ cells, supporting two previous reports, which indicated that CD4⁺CD25⁺ cells upregulated CD137 compared with CD4⁺CD25⁻ cells (27,28). The upregulation of CD137 on CD25⁺ cells suggested a mechanism whereby CD137 antibody treatment could preferentially affect CD4⁺CD25⁺ cells. As shown in Fig. 6B, anti-CD137 antibody treatment increased the number of CD4⁺CD25⁺ T-cells in NOD mice.

To confirm that increased CD4⁺CD25⁺ T-cells were regulatory cells and not polyclonally activated T-cells upregulating CD25, intracellular Foxp3 expression was analyzed. CD25⁺ cells in anti-CD137–treated mice were Foxp3⁺ (Fig. 6C), indicating that the administration of anti-CD137 antibody induced Foxp3⁺CD4⁺CD25⁺ Treg cells associated with diabetes protection. To prove the regulatory function of CD4⁺CD25⁺ Treg cells in vitro, we performed suppression studies, which showed that CD4⁺CD25⁺ cells from anti-CD137–treated NOD mice had suppressive activity on CD25⁻ cells from both naïve and anti-CD137–treated mice (Fig. 6D). These findings indicated that CD137 antibody–induced CD4⁺CD25⁺ cells have the Treg cell phenotype.

To directly prove that treatment-induced CD4⁺CD25⁺ cells have protective ability, we transferred CD4⁺CD25⁺ or CD4⁺CD25⁻ cells from anti-CD137–treated mice, along with diabetic CD8⁺ cells, to NOD-scid recipients. As shown in Fig. 6E, all NOD-scid recipients transferred with CD4⁺CD25⁺ cells were completely protected from diabetes compared with those that received CD4⁺CD25⁻ cells (P < 0.0001). Moreover, mice receiving anti-CD137–treated CD4⁺ cells from which CD25⁺ cells were depleted (CD4⁺CD25⁻) showed no significant difference in disease onset compared with CD4⁺ T-cells transferred from NOD mice treated with PBS alone (Fig. 6E, P = 0.91), which strongly supports the hypothesis that anti-CD137 therapy acts on CD4⁺CD25⁺ cells. Interestingly, we found that some of the mice receiving CD25⁻ cells developed inflammatory bowel disease (IBD) (manifested by wasting and rectal prolapse) or neuropathy, consistent with previous reports (data not shown; 36,37). Of mice receiving the CD25⁻ cells, five of nine developed diabetes, whereas four of nine developed either IBD or neuropathy. The finding that diabetes and neuropathy/IBD were mutually exclusive is consistent with the report of Setoguchi et al. (36), who had similar results.

To test whether anti-CD137 had a preferential cell target in vivo, we harvested splenocytes from NOD mice treated twice with 3H3, 1 week after the final treatment. We stained for CD4 and CD25 and used the second-step (anti-IgG2a-bio) antibody specific for 3H3, followed by strep-avidin phycoeythrin. As shown in Fig. 6F, 3H3 was bound only to CD4⁺CD25⁺ cells, not CD4⁺CD25⁻ cells in vivo, consistent with a mechanism of action directed to this Treg subset. There was not significant in vivo binding of 3H3 to CD8 or Dx5⁺ cell subsets.

Finally, we tested directly whether anti-CD137 antibody could act on pathogenic CD4⁺ cells by using a different cell transfer system, in which CD4⁺ and CD8⁺ cells were sorted from diabetic NOD splenocytes and transferred into NOD-scid recipients. At the time of transfer, recipient NOD-scid mice were treated with a single dose of 200 μg anti-CD137 or PBS. Treatment of NOD-scid recipients with anti-CD137 at the time of adoptive transfer did not prevent diabetes compared with PBS-treated control NOD-scid recipients and slightly but significantly increased the onset of diabetes (Fig. 7). This result confirms that the protective effect of anti-CD137 did not act by inactivating pathogenic CD4⁺ and CD8⁺ T-cells and, in fact, shows that activating CD137 on autoreactive T-cells could enhance disease onset. This result agrees with the data of Sun et al. (23), who showed that although anti-CD137 therapy can prevent EAE, it could not prevent EAE after adoptive transfer of a pathogenic T-cell population. The lack of effect of anti-CD137 antibody at time of transfer into NOD-scid is likely due to an imbalance of regulatory cells compared with pathogenic cells in diabetic NOD spleen populations.

We have demonstrated that treatment specifically directed to the Idd9.3 candidate gene product, CD137, can prevent diabetes in NOD mice and showed that anti-CD137 therapy increased the number of CD4⁺Dx5⁺ and CD4⁺CD25⁺ Treg cells. By using transfer models into NOD-scid mice, we showed that the effect of anti-CD137 treatment is not directly on the CD4⁺ pathogenic autoreactive T-cell population (because these cells persist in treated mice and can transfer disease) but is on Treg cells that control the pathogenicity of the autoreactive cells.

Anti-CD137 therapy has previously been demonstrated to prevent murine SLE, EAE, and collagen arthritis (17,21–23). Ours is the first report, to our knowledge, to show that it prevents autoimmune diabetes. The only previous report of the effect of CD137 on the diabetes disease process is that overexpression of a Cd137 transgene increased the incidence of diabetes (38). Although our results indicate a protective effect of CD137 stimulation in type 1 diabetes of NOD mice, our results combined with the results of Sytwu et al. (38) suggest that there can be too much of a good thing. CD137 is a member of the TNF receptor superfamily; in this regard, it is useful to recall that TNF-α had dual effects in type 1 diabetes: anti–TNF-α treatment at 9–10 weeks prevented diabetes, whereas TNF-α given to neonates increased diabetes severity (39–41). Our results show that anti-CD137 is most effective at preventing diabetes when given early in the disease process. Once a cadre of autoreactive T-cells is fully developed, anti-CD137 cannot prevent disease and may even worsen it (as shown by the earlier onset of diabetes when anti-CD137 is given at the time of transfer of NOD CD4⁺ diabetic cells into NOD-scid; Fig. 7). Moreover, splenocytes taken from aged anti-CD137 mice can transfer diabetes, suggesting a finite regulatory capacity of the anti-CD137 induced Treg cells. Because autoimmune, activated pathogenic CD4⁺ T-cells upregulate CD137, it is possible also that once the autoimmune repertoire has expanded, anti-CD137 activates the autoreactive cells, opposing the protective effect of CD4⁺CD25⁺ Treg cells. In this regard, it is noteworthy that we find that anti-CD137 treatment of lymphocytic chonomeningitis virus–infected mice leads to sustained high-level production of TNF-α (Mittler lab, unpublished data).

Nevertheless, our results are important for several reasons. Cd137 is a strong candidate gene for Idd9.3; i.e., it is highly likely that CD137 has a significant role in the pathogenesis of spontaneous type 1 diabetes in NOD mice. Lyons et al. (6) showed that although the Idd9.3 region containing the B10 allele of Cd137 decreased diabetes incidence on an NOD background, it did not prevent insulitis, consistent with our result. Moreover, these authors showed that the insulitic lesions in the Idd9 congenic upregulated IL-4, consistent with our demonstration that anti-CD137–treated CD4⁺ cells upregulate IL-4 production (6). Cannons et al. (10) have recently followed up this work by demonstrating that T-cells from NOD mice congenic for the B10 allele of Cd137 proliferate more vigorously to anti-CD3 plus anti-CD137 stimulation than NOD cells, i.e., that the NOD Cd137 allele may have relatively defective signaling capability. The authors speculate that decreased signaling through NOD CD137 could result in defective regulatory cell production. Our results are consistent with this hypothesis, because we show that agonistic stimulation through CD137 can prevent diabetes in NOD mice by the induction of a regulatory cell population. Although our results do not prove that Cd137 is the Idd9.3 gene, they are entirely consistent with this conclusion and supportive of prior studies defining the role of CD137 in type 1 diabetes.

Our results show that anti-CD137 mediates protection from diabetes via effects on CD4⁺CD25⁺ Treg cells. It has previously been shown that NOD CD4⁺CD25⁺ Treg cells numbers and functional capacity are decreased and that boosting NOD CD4⁺CD25⁺ Treg cell function can prevent diabetes (25,26). What we show for the first time is that NOD CD4⁺CD25⁺ Treg cells function may be linked to the Idd9.3 candidate gene, Cd137. We show that CD4⁺CD25⁺ Treg cells upregulate CD137 compared with the CD4⁺CD25⁻ subset and that transfer of the CD25⁺ subset completely prevented diabetes, whereas the CD4⁺CD25⁻ cell subset offered no significant protection. These findings are consistent with previous reports showing that CD4⁺CD25⁺ Treg cells upregulate CD137 (27,28) and with the increase of CD4⁺CD25⁺ cells in anti-CD137–treated mice in various models, although regulatory function was not proved in the latter reports (21,23). The effects of anti-CD137 antibody on CD4⁺CD25⁺ Treg cells are still controversial. 4-1BB ligand induces proliferation of CD4⁺CD25⁺ Treg cells and maintains their suppressor function (29), in contrast to no observable effect of anti-CD137 antibody on CD4⁺CD25⁺ Treg cells in vitro (30). This study also reported that regulatory cells from CD137^−/− mice could control CD45RB^high T-cell–induced colitis, implying that CD4⁺CD25⁺ Treg cells may exist independent of CD137 signaling (30). Moreover, in the setting of experimental autoimmune thyroiditis, the administration of anti-CD137 antibody breaks tolerance and worsens the disease (42). The double-edged effect of CD137 signaling likely depends on the balance of pathogenic and protective subsets at the time of treatment in any given model system, as we showed in Fig. 7. None-theless, our findings of specific binding of anti-CD137 antibody only to CD4⁺CD25⁺ cells in vivo plus our demonstration that these cells mediate the protective effect of the CD4 population in anti-CD137–treated mice strongly support the hypothesis that the CD4⁺CD25⁺CD137⁺ Treg cell subset is a critical protective subset in type 1 diabetes.

We confirmed the observation recently published by Sakaguchi and colleagues (36,37) that CD25-depleted CD4 cells transfer several autoimmune diseases. In particular, we also confirmed their observation that the transfer of diabetes versus another organ-specific autoimmune syndrome, such as IBD or peripheral neuropathy, is mutually exclusive in this system (36). The mechanism for this striking organ specificity and exclusivity of autoimmune attack remains unknown. However, we have recently published a similar effect using genetic manipulation, in which we showed that modification of the NOD background by addition of Idd loci from B6/B10 on chromosomes 3 and 4 completely prevented diabetes while producing an autoimmune biliary disease (43,44). The remarkable “switching” of the tissue focus of autoimmune attack via environmental (CD25 depletion) or genetic (introgression of congenic regions) mechanisms is as yet unexplained but is clearly of great importance to understanding the pathogenesis of autoimmunity and type 1 diabetes.

In conclusion, we, for the first time, report that anti-CD137 antibody therapy suppresses type 1 diabetes in NOD mice and that this effect is dependent on induction of Treg cells. Treatment must be initiated before the development of a cadre of autoreactive T-cells, suggesting competing effects on protective and pathogenic cell subsets in type 1 diabetes. These results implicate a role for the Idd9.3 candidate gene, Cd137, in type 1 diabetes pathogenesis and identify CD137 as a potential therapeutic target.

W.M.R. has received National Institutes of Health Grant RFA-A102-006.