Microencapsulated G3C Hybridoma Cell Graft Delays the Onset of Spontaneous Diabetes in NOD Mice by an Expansion of Gitr⁺ Treg Cells

Type 1 diabetes (T1D) is an autoimmune disease that is caused by the selective killing of pancreatic islet β-cells, which results in the shutdown of insulin production. Presently, lifelong exogenous insulin replacement is the only available therapy for these patients (1). More than 10 million people suffer from T1D worldwide, and disease-related complications impose a heavy burden on society (2). The inflammatory process and immune response are key factors in the development of T1D (3). In particular, autoreactive T cells and islet cell-reactive B lymphocytes play an important role in the development of this disorder (4,5). It has been reported that insulin-secreting cells can fully recover under conditions that restrain their destruction; this has led to the exploration of immunomodulatory approaches for T1D therapy (6–12). One of the first approaches used was immune cell depletion to prevent insulin-producing β-cells from being attacked and destroyed. However, this attempt did not meet with much success (13). Therefore, there is a clear need for alternative immunomodulatory methods that restrain pancreatic islet destruction in T1D.

Emerging evidence indicates that pancreatic resident regulatory T (Treg) cells exert immunosuppressive and anti-inflammatory effects, thereby suppressing overexuberant immune responses in T1D. Thus, an effective treatment for patients with T1D may involve the potentiation of immune tolerance via increasing the number of Treg cells (14). Indeed, engineered Treg cells expressing islet-specific antigens have shown good efficacy in T1D preclinical models (15–17). Thus, agents that can induce Treg-cell expansion should be explored further for their potential therapeutic benefits in T1D.

Glucocorticoid-induced TNFR-related (Gitr) protein, also known as TNFRSF18, is mainly expressed by T cells and plays a role in immune system homeostasis and immune-related disorders such as autoimmune diseases and cancer progression (18). A large number of Treg cells express high levels of Gitr in the microenvironment of human cancers (19,20), and Gitr expression is correlated with more aggressive cancer and worse prognosis (21,22). Therefore, Gitr is one of the activated Treg-cell markers that are crucial in determining the suppressive microenvironment of tumors. Indeed, some Gitr-triggering reagents promote tumor rejection by inducing antibody-dependent cellular cytotoxicity (ADCC) of Treg cells (23,24).

Transfer of Gitr-depleted T cells into BALB/c nude mice results in the death of 90% of the mice due to autoimmune disease development, thus indicating that Gitr⁺ cells play a crucial role in immune homeostasis (25). Furthermore, in systemic lupus erythematosus and Sjögren syndrome, the expansion of Cd4⁺Cd25^−/lowFoxp3^lowGitr^int/high (GITR single-positive [sp]) Treg cells was found in patients with low disease activity, thus confirming that Gitr⁺ cells play a role in immune homeostasis in humans too (26,27).

Gitr is a costimulatory receptor, and its triggering favors the activation and expansion of T cells, including conventional Cd4⁺ T cells, effector Cd8⁺ T cells, and Treg cells (18). The resulting in vivo effects depend on the context, the type of disease, and the way by which Gitr is stimulated (28). Despite the context-dependent effect of Gitr triggering, we can assume that the main effects of Gitr triggering favor the expansion of Treg cells based on the reported evidence (19). For example, the number of Treg cells is lower in mice knockout for Gitr and Gitr ligand (GitrL) (29–31) and higher in GitrL transgenic mice (32,33). Moreover, some anti-Gitr monoclonal antibodies (mAbs; e.g., G3c mAb) and GitrL fusion proteins do not promote tumor rejection because they stimulate Treg-cell expansion without promoting ADCC (34,35). In the context of T1D, this could mean that treatment with anti-Gitr mAbs may favor a Gitr⁺ Treg-cell expansion, which prevents β-cell destruction by locally restraining inflammatory responses. To explore this possibility, in the current study, we treated diabetes-prone NOD mice with the anti-Gitr G3c mAb and explored its potential antidiabetic effects, with promising results.

G3C hybridoma cells were cultured in RPMI medium (Gibco) with 10% FBS. When purification of IgM mAbs was needed (ELISA standard curve), cells were cultured in CD Hybridoma serum-free chemical medium (CD Hybridoma Medium, supplemented with 250× Cholesterol Lipid Concentrate) (Gibco). To fabricate G3c-secreting microcapsules that would allow for the outflow of big molecules (such as an IgM), we had to modify the standard microencapsulation procedure; we changed stoichiometric molar ratios of ultrapurified alginate (AG) and poly‐l‐ornithine, used for microencapsulation. We also changed the composition of the capsule’s multilayered membrane, by reframing the layering sequence to fulfill the desired membrane’s molecular weight cutoff. The functionality test and the details on the preparation of the G3c-secreting microcapsules have been provided elsewhere (36). We used 1.5 × 10⁶ G3C hybridoma cells/1 mL AG; such an AG-to-cell ratio would avoid the formation of empty capsules.

G3C hybridoma cells were cultured for 14 days in a serum-free chemical medium (CD Hybridoma Medium, supplemented with 250× Cholesterol Lipid Concentrate) (Gibco). G3c-derived mAb was purified using HiTrap IgM Purification HP column (1 mL size) for affinity purification (GE Healthcare) following the manufacturer’s instructions. The buffer containing purified G3c mAb was exchanged with PBS through dialysis by using the Slide-A-Lyzer Dialysis Cassette (10,000 molecular weight cut-off ) (Thermo Fisher Scientific). The purified G3c mAb was quantified using the Bradford method. We developed an ELISA to quantify the level of G3c mAb produced by G3C cells, as described elsewhere (36).

For the quantization of Igs in the serum of SV129 and Gitr^−/− mice, we used commercially available kits following the manufacturer’s instructions (Mouse Uncoated ELISA kits for IgG1 [cat. 88-50410-22], IgG2a [cat. 88-50420-22], IgG2b [cat. 88-50430-22], IgG2c [cat. 88-50670-22], and IgG3 [cat. 88-50440-22]; all kits were purchased from Thermo Fisher Scientific).

NOD mice were purchased from The Jackson Laboratory (Bar Harbor, ME). Female SV129 and Gitr^−/− SV129 mice were available in our laboratory. The animals were housed in a controlled environment, provided with standard rodent chow and water ad libitum, and kept under specific pathogen-free conditions. Animal care complied with regulations in Italy (D.M. 116192), Europe (O.J. of E.C. L 358/1 12/18/1986), and the U.S. (Animal Welfare Assurance No. A5594–01). The encapsulated G3C hybridoma cells were grafted intraperitoneally into mice under general anesthesia. For each mouse, 1.5 × 10⁶ G3C hybridoma cells in 1 mL of microcapsules (G3C/cps) or 1 mL of empty microcapsules (e/cps) as a control, were grafted. All procedures were approved by the University of Perugia Committee of Animal Research and Ethics.

SV129 and Gitr^−/− SV129 mice (11 weeks old) were grafted with e/cps (n = 5 wild type [WT] and n = 5 Gitr^−/− mice) and G3C/cps (n = 6 WT and n = 6 Gitr^−/− mice) and sacrificed 3 weeks after. Five WT and five Gitr^−/− untreated mice (14 weeks old) were used as control groups. The protocol is summarized in Supplementary Fig. 1.

In the short-term experiment, NOD mice (8 weeks old) were grafted with e/cps (n = 8 mice) and G3C/cps (n = 8 mice) and sacrificed 3 weeks after. Four mice (8 weeks old at baseline) were used as the control group. The protocol is summarized in Supplementary Fig. 1.

In the long-term experiment, NOD mice (11 weeks old) were grafted with e/cps (n = 24 mice) and G3C/cps (n = 22 mice) and sacrificed 3 months after (when they were 24 weeks old) or when they had a persistent high blood glucose (BG) level (2 BG >500 mg/dL). Control groups comprised 1) the baseline control group (n = 4 mice sacrificed when they were 11 weeks old) and 2) the untreated control group (n = 10 mice, 11 weeks old) that were left untreated and sacrificed when 24 weeks old or when they had a persistent high BG level. The protocol is summarized in Supplementary Fig. 1.

During the long-term experiment, one e/cps-treated NOD mouse and one G3C/cps-treated NOD mouse showed weight loss, severe hypoglycemia, and salivary gland abscess, and they were sacrificed. These mice were not considered in the long-term experiment.

After the sacrifice of the NOD mice, histological evaluation of the pancreas was performed using a score adapted from one previously described (37). Briefly, 3–4 μm of paraffin-embedded sections from half of the pancreas were used to prepare 24 slides. Then, 5 among 15 consecutive histological sections were placed on each slide, and among the 24 slides, 8 slides were stained with hematoxylin and eosin by skipping 2 of 3 consecutive slides. Only one of five sections on the slide was analyzed by light microscopy and scored. Insulitis scoring was defined according to the following criteria: 1) intact without insulitis, an islet that shows absence of cell infiltration; 2) intact with insulitis, an islet that is damaged (<50% of the islet area infiltrated or infiltration is restricted to the periphery of the islet); or 3) disrupted, an islet with severe intrainsulitis in which >50% of the islet area is infiltrated.

In the short-term experiment, 591 islets were scored (a mean of 27.6 ± 6.6 islets for baseline mice, 24.7 ± 8.6 islets for e/cps-treated mice, and 22.0 ± 6.9 islets for G3C/cps-treated mice). In the long-term experiment, 986 islets were scored (a mean of 20.8 ± 9.3 islets for baseline mice, 18.3 ± 11.3 islets for untreated control mice, 16.7 ± 13.3 islets for e/cps-treated mice, and 18.5 ± 10.8 islets for G3C/cps-treated mice).

BG levels in NOD mice were evaluated twice weekly by blood sampling from the tail of the animals and measurement of glucose levels with a standard BG meter (37).

Cells from the spleen, pancreas, and pancreatic lymph nodes were recovered as previously described (37).

All cytofluorimetric evaluations were performed on the Attune NxT Flow Cytometer system (Thermo Fisher Scientific) using fluorochrome-conjugated monoclonal antibodies against B220 (clone RA3-6B2), Cd3 (clone 145-2C11), Cd4 (clone GK1.5), Cd8 (clone 53-6.7), Foxp3 (clone FJK-16s), Gitr (clone DTA-1), Cd25 (clone PC61.5), Cd44 (clone IM7), and CD62L (clone MEL-14). All antibodies were provided by Thermo Fisher Scientific. Before Foxp3 staining, cells were permeabilized with the Foxp3/Transcription Factor Staining Buffer Set (Thermo Fisher Scientific) following the manufacturer’s instructions.

For cell proliferation studies, cells were labeled with the CellTrace Violet Cell Proliferation Kit for flow cytometry (Thermo Fisher Scientific) following the manufacturer’s instructions. After staining, cells were cultivated for 3 days in the presence of anti-Cd3 (1:500, functional grade, clone OKT3) (Thermo Fisher Scientific) and anti-CD28 (1:5,000, functional grade, clone 37.51) (Thermo Fisher Scientific) cross-linked to U-shaped wells (96-well plate). Before flow cytometry evaluation, splenocytes were stained with fluorochrome-conjugated anti-B220, -Cd3, -Cd4, and -Cd8 antibodies, and the proliferation rate of CellTrace Violet-labeled B220⁻Cd3⁺Cd8⁻Cd4⁺ cells was evaluated.

For the evaluation of the antigen-specificity of Treg-cell populations, we stained cells with fluorochrome-conjugated tetramers following the manufacturer’s instructions (IGRP Tetramer-VYLKTNVFL-APC, cat. TB-M552-2; NRP Tetramer-KYNKANVFL-BV421, cat. TB-M553-4; and InsB Tetramer-LYLVCGERL-PE, cat. TS-M554–1; all tetramers were purchased from MBL International). To evaluate the antigen-specificity of Treg-cell populations, splenocytes or cells from the pancreas and pancreatic lymph nodes were stained with fluorochrome-conjugated tetramers together with fluorochrome-conjugated anti-B220, -Cd3, -Cd4, -Cd8, -Gitr, and -Cd25 antibodies. Isolation of immune cells from the pancreas and pancreatic lymph nodes resulted in a relatively low number of cells, so that the events detected in the B220⁻Cd3⁺Cd8⁻CD4⁺ gate when tetramers and antibodies against surface markers were stained ranged between 1,500 and 10,000 events, considered to be sufficient for analysis. When intracellular staining of Foxp3 cells was performed, two samples were excluded due to the above reason. In not excluded samples, the mean events in B220⁻Cd3⁺Cd8⁻CD4⁺ gate were equal to 1,596.

Total RNA from the pancreas of NOD mice was isolated using Trizol reagent (Thermo Fisher Scientific), and the generation of cDNA was performed using a PrimeScript RT Reagent Kit (Takara Bio Europe). Real-time quantitative (q)PCR was done with an ABI-7300 Real-time PCR system (Applied Biosystems) using specific FAM/MGB dye-labeled TaqMan probes (Mm00442754_m1 for Cd4, Mm01182107_g1 for Cd8, Mm00475162_m1 for Foxp3, Mm00437136_m1 for Gitr, and Mm01340213_m1 for Cd25). Gene expression was quantitated relatively to the expression of endogenous control GAPDH (4352339E) VIC/MGB probe amplified in the same tube of the investigated genes. All probes were purchased from Applied Biosystems (Thermo Fisher Scientific). All experiments were repeated twice, and the ΔΔ threshold cycle method was used to determine the expression of the genes of interest, as previously described (38).

Kolmogorov-Smirnov (KS) normality test was performed on data from each group before statistical evaluation. To evaluate the differences between groups, P values were calculated using ordinary one-way ANOVA (Tukey) test when the KS normality test was passed and Kruskal-Wallis (Dunn) test when KS normality test failed. In the evaluation of the percentage of diabetic mice in the long-term experiment, the Fisher exact contingency test was performed. In the evaluation of the posttreatment day at which mice developed diabetes, the unpaired Student t test was performed. The log-rank Mantel-Cox test was used to compare Kaplan-Meier curves. The hazard ratio (95% CI) was calculated using the Mantel-Haenszel method. All statistical analyses were conducted using Prism software (GraphPad Software).

No data sets were generated or analyzed during the current study. The hybridoma G3C that supports the findings of this study is available from Kyoto University Medical Science and Business Liaison Organization, but restrictions apply to the availability of this cell line, which was used under license for the current study and therefore is not publicly available. The hybridoma G3C may be available from the authors and/or Kyoto University Medical Science and Business Liaison Organization upon reasonable request and with permission of Kyoto University Medical Science and Business Liaison Organization.

To administer high-quality fully active antibodies, we encapsulated the G3C hybridoma cells in AG-based microcapsules. Microcapsules have been previously used by other groups and by us to deliver cell-derived products in vivo (39–41). The microcapsules used in this study were modified so that the G3c mAb (an IgM-type antibody) could exit the capsule. The microencapsulated hybridoma cells survived for several days in the in vitro culture and for several weeks when transferred into the peritoneum of mice. Details about how the capsules were manufactured, their properties, and IgM delivery are reported elsewhere (36).

We grafted empty microcapsules (e/cps) or the G3C-containing microcapsules (G3C/cps) in the peritoneum of WT and Gitr^−/− mice. Mice were euthanized 3 weeks after microcapsule implantation (Supplementary Fig. 1). Notably, the microcapsules did not elicit an inflammatory response (36), and a detectable amount of the G3c mAb was found in the blood of the mice (Supplementary Fig. 2). However, serum concentrations of the G3c mAb were higher in Gitr^−/− than in WT mice, probably due to the target-dependent peripheral distribution of the G3c mAb in WT mice.

Then, we investigated the effects of both surgical procedure and treatment by evaluating the main cellular (Supplementary Fig. 3) and humoral immunity (Supplementary Fig. 4) features. G3c treatment showed some effects on the immune system of both WT and Gitr^−/− mice, as demonstrated by the increase of spleen weight and the number of splenocytes and B lymphocytes in the spleen. Furthermore, the number of splenic Cd3⁺ and Cd4⁺ T lymphocytes showed a significant increase in WT but not in Gitr^−/− mice after the treatment. Conversely, G3c treatment did not affect the number of Cd8⁺ T splenocytes. Thus, the antibody has some aspecific actions on both WT and Gitr^−/− mice and specific effects (reasonably mediated by Gitr triggering) in WT mice concerning the number of Cd4⁺ cells and, consequently, Cd3⁺ cells.

G3c inoculation affected also humoral immunity. IgG3 and IgG2c increased in both WT and Gitr^−/− mice after G3c treatment; meanwhile, IgG2a significantly increased in G3C/cps-treated compared with e/cps-treated only in WT mice, and the increase was relevant (approximately threefold). The increase of Cd3⁺ T cells, Cd4⁺ T cells, and IgG2a in G3C/cps-treated WT animals may suggest that the effect of G3C treatment is restricted to a specific immune response that promotes an immune-activating effect.

Based on the known expression and effects of Gitr on Treg cells, we investigated whether the number of splenic Foxp3⁺ Treg cells changed in response to G3c treatment. As shown in Fig. 1A, the number of splenic Foxp3⁺ Treg cells was higher in G3C/cps-treated than in untreated WT mice, but no difference was observed between treated and untreated Gitr^−/− mice. Thus, the increase in the number of Foxp3⁺ Treg cells in the G3C/cps-treated WT mice may depend on Gitr triggering. This is supported by the finding that Gitr levels were higher in Foxp3⁺ Treg than in conventional T splenocytes (Fig. 1B). In addition, the number of Foxp3⁺Cd25⁺Gitr^high Treg cells is higher in G3C/cps-treated mice compared with e/cps-treated and untreated mice (Fig. 1C).

We recently demonstrated the existence of GITRsp cells in humans (42,43), their regulatory activity, and their suppressive role in systemic lupus erythematosus and Sjögren syndrome (26,27). Murine GITRsp unconventional Tregs are Foxp3^−/low and Cd25^−/low (Supplementary Fig. 5). Here, we found that after in vivo treatment with G3c mAb, the number of GITRsp cells was more than twice the number in G3C/cps-treated WT mice compared with e/cps-treated and untreated WT mice (Fig. 1C). Furthermore, the proliferation rate of conventional Cd4⁺ T splenocytes isolated from G3C/cps-treated WT mice was lower than in Cd4⁺ cells isolated from untreated and e/cps-treated WT mice (Fig. 1D). On the contrary, the proliferation rate of Cd4⁺ conventional T splenocytes from untreated, e/cps-treated, and G3C/cps-treated Gitr^−/− mice showed no substantial difference, thus indicating that the expanded Foxp3⁺Cd25⁺Gitr^high and GITRsp Treg cells were fully active.

Taken together, these results indicate that G3c mAb induces conventional and nonconventional Treg-cell expansion via Gitr triggering.

We grafted G3C/cps and e/cps into the peritoneum of nondiabetic NOD mice (Supplementary Fig. 1) and evaluated the number of immune cells in the spleen at 3 weeks after the implant.

Similar to SV129 mice, also in NOD mice, G3c mAb had some effects on the growth of splenocytes, B cells, and T cells (Supplementary Fig. 6). Moreover, not only the number of Cd4⁺ but also of Cd8⁺ cells was increased.

As seen in SV129 mice, the percentages and the absolute numbers of splenic Foxp3⁺ and Foxp3⁺Cd25⁺Gitr^high increased after G3c treatment, and the increase was at least partly due to the increase of Gitr^high cells (Fig. 2A and B). Similarly, nonconventional GITRsp Treg cells were higher in the spleen of G3C/cps-treated NOD mice than in the spleen of untreated and e/cps-treated NOD mice (Fig. 2A and B). The surgical procedure and treatment with e/cps had some effects on the percentage but not in the absolute number of cell subsets. In any case, the effects were less relevant than G3C/cps treatment.

Interestingly, we found a higher number of memory T cells in Foxp3⁺Cd25⁺Gitr^high cells than in Foxp3⁺Cd25⁺Gitr^−/low cells and in GITRsp cells than in Foxp3⁻Cd25⁻Gitr^−/low T cells (Supplementary Fig. 7). This suggests that the expanded conventional and nonconventional Gitr⁺ Treg cells might be antigen-specific cells. We therefore evaluated the percentage of GITRsp and Cd25⁺Gitr^high cells recognizing three antigens—Igrp, Nrp, and Insb—considered to be implied in T1D development. Results shown in Fig. 2C show that the percentages of tetramer⁺Cd25⁺Gitr^high and tetramer⁺GITRsp cells in the spleen of NOD mice were higher after treatment compared with untreated mice as concerns Igrp⁺ and Insb⁺ cells (∼1.5 fold). The increase was observed in e/cps- and in G3C/cps-treated mice, although the increase of tetramer⁺Cd25⁺Gitr^high was higher in G3C/cps-treated than in e/cps-treated mice. Moreover, the increase of the percentage of antigen-specific subpopulations within the GITRsp and Cd25⁺Gitr^high cells was not observed (Supplementary Fig. 8), suggesting that the increase of T1D-related antigen-specific cell subpopulations was similar to the increase of other antigen-specific cells that may or may not be restricted to T1D-inducing antigens.

We next evaluated the presence of Tregs in the pancreatic lymph nodes and pancreatic immune infiltrate at 3 weeks after the implantation. The data we obtained were similar to the data found in splenocytes but more relevant from a quantitative point of view (Fig. 3A). Interestingly, we observed no effect on the percentages of Foxp3⁺, Foxp3⁺Cd25⁺Gitr^high, and GITRsp Treg cells in e/cps mice, suggesting that the increase of the above-specified subpopulations in the pancreas is exclusively due to Gitr triggering.

The percentage of Igrp-, Nrp-, and Insb-specific Treg cells in the pancreatic lymph nodes and the pancreatic immune infiltrate at 3 weeks after the implantation was substantially increased compared with untreated NOD mice and/or e/cps-treated mice (Fig. 3B). The only exception was Igrp-specific GITRsp cells. Interestingly, Insb-specific Cd25⁺Gitr^high and GITRsp cells showed a relevant increase after only G3C/cps treatment (6- to 11-fold compared with the untreated animals). Of note, Insb-specific subpopulations increased within the GITRsp cells (Supplementary Fig. 9), suggesting that the increase of these subpopulations was driven by antigen-specific cells. The increase of Insb⁺Cd25⁺Gitr^high was similar to that of Insb⁺ GITRsp cells, even if not significant.

Furthermore, we evaluated the effects of G3c treatment on pancreas histology. Pancreatic islets were divided into three groups according to their histological appearance (Fig. 3C): intact without insulitis (an islet that shows absence of cell infiltration), intact with insulitis (an islet that is damaged in which <50% of the islet area is infiltrated or infiltration is restricted to the periphery of the islet), or disrupted (an islet with severe intrainsulitis in which >50% of the islet area is infiltrated). As shown in Fig. 3C, the percentage of islets that were intact without insulitis and disrupted was different between the pancreas of G3C/cps- and e/cps-treated NOD mice. In both cases, the pancreas from G3C/cps-treated NOD mice showed comparable values to untreated 8-week-old mice (baseline mice).

These findings indicate that Treg-cell expansion via Gitr triggering inhibits islet attack by the immune system during the development of diabetes.

We implanted G3C/cps and e/cps in the peritoneum of nondiabetic NOD mice and evaluated the BG level for each untreated and cps-treated mouse every 48−72 h (Supplementary Fig. 1). Figure 4A shows that several untreated and e/cps-treated NOD mice developed diabetes since BG levels were higher than normal in at least two consecutive evaluations. However, in all except two G3C/cps-treated NOD mice, the BG level remained <200 mg/dL during the whole observation period, with the exception of sporadic, nonconsecutive evaluation; this means that the treatment delays diabetes development. Indeed, one G3C/cps-treated NOD mouse developed diabetes 77 days after the beginning of the treatment and another showed a BG of 300 mg/dL on day 89. Not having a second BG measurement, we could not assess the onset of diabetes in the latter mouse, but we considered this mouse diabetic. On the contrary, it was considered as alive in the Kaplan-Meyer survival curve, because unlike the others, it was not euthanized due to diabetes.

In conclusion, we found that G3C/cps-treated NOD mice had a higher survival rate than untreated and e/cps-treated NOD mice (Fig. 4B). These results suggest that G3c treatment prevents or delays the onset of diabetes in NOD mice.

Figure 5A shows the percentage of islets that were “intact without insulitis” and “disrupted” differed between the pancreas of G3C/cps- and e/cps-treated NOD mice. A similar difference was observed between the pancreas of G3C/cps-treated and untreated NOD mice, but it resulted in a nonsignificant difference, presumably due to the lower number of studied mice. G3C/cps-treated NOD mice showed comparable values of “disrupted” islets to those of baseline, 11-week-old mice (13 weeks younger than G3C/cps-treated NOD mice). Figure 5B shows that the “disrupted”-to-“intact without insulitis” ratio was <1 in baseline young mice and G3C/cps-treated mice and >1 in untreated and e/cps-treated mice, likely indicating that the beneficial effects of the G3C/cps treatment is related to the protection of the islets from autoimmune attack.

When we compared the histology of the islets in mice that were nondiabetic at the end of the 90-day period (Fig. 5A, right panel), no significant differences were observed among G3C/cps-treated, untreated, and e/cps-treated mice, possibly suggesting that G3c treatment potentiates the immune homeostatic attempts that occur in NOD mice not developing the disease.

To understand whether G3c treatment protects pancreatic islets from destruction through the same homeostatic mechanisms activated in NOD mice spontaneously remaining diabetes-free or through other mechanisms, we evaluated whether the effects observed at 3 weeks postmicrocapsule implantation were still observable after 3 months in G3c-treated mice and whether they were similar to untreated NOD mice remaining diabetes free.

First, we evaluated the splenocyte cell composition. Figure 6 shows that in all mice (left panels) and in nondiabetic mice (right panels), the three Treg subsets that underwent expansion 3 weeks after G3c treatment show a less relevant expansion compared with that observed after the 3-week treatment. In particular, Cd25⁺Gitr^high and GITRsp cells did not show significant expansion compared with e/cps-treated mice but showed significant expansion compared with untreated animals. On the contrary, the increase of Cd25⁺Gitr^high in Foxp3⁺ cells is still present both in all and in nondiabetic mice.

Then, we evaluated the infiltration of Cd4⁺, Cd8⁺, and conventional and nonconventional Treg cells by real-time qPCR in nondiabetic and diabetic NOD mice that were implanted with empty and G3C microcapsules or left untreated. Figure 7A shows that similar levels of Cd4⁺ and Cd8⁺ T cells were present in the pancreas of all the animals, thus suggesting that the treatment does not affect T-cell infiltration. Independently from treatment, no differences were observed between diabetic and nondiabetic mice, implying that the development of diabetes was not dependent on the infiltration levels of Cd8⁺ T cells.

Next, we evaluated the mRNA levels of Foxp3, Gitr, and Cd25. Figure 7B shows that the level of Foxp3 expression is much higher in G3c-treated mice than in the other groups and that all nondiabetic mice have Foxp3 values higher than diabetic mice. However, the significance is seen only between G3c-treated mice and diabetic mice (both untreated and e/cps-treated). A similar reasoning can be done observing Gitr expression but not Cd25 expression. The nonsignificant difference among the groups in the expression of Cd25 may be due to the expression of Cd25 even in activated T cells.

The findings may suggest that G3c treatment favors the infiltration of the pancreas and/or homing in pancreatic lymph nodes of conventional and nonconventional Tregs even after months from the beginning of treatment. Moreover, a tendency of infiltration/homing of Tregs is evident in nondiabetic untreated and e/cps-treated mice, suggesting that activated Tregs play a crucial role in the prevention of diabetes in NOD mice.

In the current study, we have demonstrated that prolonged treatment with the anti-Gitr G3c mAb promotes the expansion of conventional and nonconventional Treg-cell subsets in the spleen of healthy animals and diabetes-prone NOD mice and in the pancreas of NOD mice at 3 weeks after the beginning of treatment. G3c-induced Treg-cell expansion was correlated with the prevention/delay of diabetes development in NOD mice and a decrease in the number of disrupted islets.

When we evaluated the animals at the end of the observation period (3 months after the beginning of the treatment), the number of expanded splenic Treg cells was comparable between e/cps-treated and G3C/cps-treated NOD mice. Since Nti et al. (44) demonstrated that diabetes development is associated with an increase in Foxp3⁺ Treg cells in the spleen of NOD mice along with an almost complete disappearance of these cells in pancreatic lymph nodes, we investigated the expression levels of Foxp3, Gitr, and Cd25 in the pancreas of diabetic and healthy NOD mice at the end of the 3-month observation period or when we had to sacrifice them due to continued high levels of BG. We used for gene expression analysis pancreatic sections surrounded by small pancreatic lymph nodes present on the surface of the pancreas. Therefore, our results take into account both pancreatic-infiltrating T cells as well as T cells in pancreatic lymph nodes. Our results confirm and support the findings of Nti et al. (44), who suggested that the destruction of islets correlates with a relevant decrease of Foxp3⁺ Treg cells in the pancreatic lymph nodes. Interestingly, in the pancreas of G3C/cps-treated NOD mice, the levels of Foxp3 and Gitr were similar to those found in untreated and e/cps-treated NOD mice that did not develop diabetes. These findings together indicate that G3c-induced Treg-cell expansion provides protection against the development of diabetes and that a high level of Treg in the pancreas protects from the development of diabetes, independently from the treatment. Accordingly, Bluestone et al. (45) performed a phase I study to treat T1D using expanded polyclonal cells, suggesting that Treg cells have the potential of treating T1D.

Data by Nti et al. (44) and our data demonstrate that Treg cells are present at low levels in the pancreas of NOD mice with diabetes, although Treg cells are present in a sufficient number in the lymphoid organs. The absence of pancreatic Treg cells may be due to a defect in the migration of these cells into the organ, death of resident Treg cells, and/or their instability in a proinflammatory microenvironment, which led them toward a Th-effector phenotype such as Th17 (46). G3c-induced Treg-cell expansion could have a stable Treg phenotype or migrate to the pancreas due to their antigen specificity. Several pieces of evidence suggest that Treg cells expressing high levels of Gitr are activated and function as memory Treg cells (19,47), as confirmed in this context (Supplementary Fig. 7). We can thus hypothesize that Gitr triggering leads to the preferential expansion of a certain type of Treg cells that tend to migrate into the pancreas because they are specific for pancreatic antigens. Indeed, antigen-specific conventional and nonconventional Tregs are expanded after Gitr triggering (Figs. 2C and 3B and Supplementary Fig. 9). The hypothesis that expanded antigen-specific Tregs migrate from the periphery to the pancreas is confirmed by the fact that the increase of antigen-specific Treg cells (in particular, Insb⁺ cells) after G3c treatment is much more relevant in the pancreas (Fig. 3B) than in the spleen (Fig. 2C). Accordingly, Spence et al. (48) demonstrated that Tregs specific for insulin, proinsulin, and preproinsulin are present in the pancreas of NOD mice and are expanded in the islets.

The increase of some Treg subsets after GITR triggering may suggest that the treatment causes a generalized immunosuppression of the mice, even if the increase of B cells and Cd4⁺ T cells in the spleen may suggest that it is not the case. Moreover, in SV129 mice, we observed an increase of IgG2a concentration in the serum, an Ig with stimulating properties, and in NOD mice we observed also the increase of Cd8⁺ splenocytes. Therefore, even if the hypothesis of G3c-induced immunosuppression cannot be excluded, some data seem to suggest that Gitr triggering dictates more immunomodulation than immunosuppression. Future studies will investigate this crucial matter evaluating the response of treated animals to viruses, bacteria, and fungi.

Besides the immunosuppression issue, at least other four points should be addressed in the hypothesis of translating our findings to humans: 1) an anti-human GITR antibody able to have the same effects of G3c must be found; 2) it is necessary to find the optimal way to purify, stabilize, and administer the antibody, provided that administration of microencapsulated hybridoma cells may have a low safety profile; 3) the best dose of the antibody must be found; and 4) the length of the effect of the treatment has to be determined. The appearance of diabetic G3C/cps-treated mice at the end of the study may suggest that the effect of the treatment in NOD mice lasts 3–4 months. The resulting in vivo effects of Gitr triggering is dependent on the context, the disease, and the way by which Gitr is stimulated (28). The anti-Gitr DTA-1 mAb activates Cd8⁺ T cells and promotes tumor rejection via the ADCC of Treg cells only when it is administered some days after tumor implantation (18,23,24). A GitrL fusion protein containing the Fc domain that was designed to maximize valency and the potential to agonize Gitr (GitrL-Fc) has been reported to have similar effects (49). On the contrary, the anti-Gitr G3c IgM mAb and another kind of GitrL fusion protein (a pentameric GitrL extracellular domain that does not contain Fc fragments) do not promote tumor rejection because not only T cells but also Treg cells are stimulated (34,35). Despite the context-dependent and the tool-dependent effect of Gitr binding, we can assume that in the healthy mouse, the main effect of Gitr triggering is the expansion of Treg cells (19,47). This is supported by the finding that the number of Treg cells is lower in GitrL^−/− and Gitr^−/− mice (29–31) and higher in GitrL transgenic mice (32,33). In the presence of interleukin 2 and anti-Cd3 mAbs, Gitr activation by the GitrL-Fc fusion protein promotes the proliferation of both conventional T cells and Treg cells; however, Foxp3⁺ cells exhibit a stronger response to the ligation of Gitr than Foxp3⁻ cells (50). One of the reasons for this effect is that Gitr is expressed at the highest levels by Treg cells (47,51). Our study confirms that the effects of Gitr triggering depend on the context and the way through which Gitr is triggered. In this model, the agonist anti-Gitr G3c IgM mAb promotes more Treg-cell expansion than conventional T-cell expansion.

A recent study demonstrated that treatment of NOD mice with purified IgM reverses new-onset diabetes and enhances Treg-cell numbers (52). Therefore, it can be hypothesized that Gitr triggering and an IgM-specific effect work synergistically to prevent the development of diabetes. The more relevant of these two mechanisms needs to be investigated in future studies using purified G3c mAb and polyclonal IgM antibody. However, in the current study, the observed expansion of Gitr⁺ T cells after treatment suggests that Gitr triggering plays a determinant role in the prevention/delay of diabetes.

The current study, however, has some limitations, and the main one is the low number of NOD mice that developed diabetes after e/cps treatment. This could be the result of the surgical procedure used to implant the microcapsules in the peritoneum as well as the microencapsulation procedure itself. Moreover, our study lacks isotype control because AG microcapsules do not allow testing of polyclonal IgM antibodies; therefore, it is impossible to assess whether the IgM itself or Gitr triggering play the most relevant effect. Another limitation of our study is the length of the observation period of G3C-treated mice (3 months); thus, it is impossible to state whether the treatment prevents or delays diabetes development. In addition, it is important to account for the products released by the hybridoma that, along with the antibody, may have influenced the effects of the anti-Gitr IgM.

In conclusion, the findings of the current study showed that the treatment of NOD mice with G3c mAb microcapsules delayed/prevented diabetes and are proof of concept that support the idea that Gitr signaling can induce Treg cells in vivo, thus improving the outcome of T1D. Therefore, further treatment modalities with antibodies or fusion proteins that are able to stimulate human GITR should be tested in humanized murine models. Such biotechnological drugs that are able to trigger GITR may represent a promising therapeutic approach to inhibit diabetes development by inducing Treg-cell expansion. Induction of Treg-cell expansion in early-onset T1D, when 20−30% of islet β-cells are damaged but not completely destroyed, may enable recovery of the pancreatic β-cell fraction to a level that is sufficient to maintain fairly efficient BG levels in the absence of exogenous insulin supplementation. This therapeutic approach might be eventually successful in the early treatment of patients with newly diagnosed T1D.

Acknowledgments. We thank Yehuda Shoenfeld of Tel-Aviv University, Tel-Aviv, Israel, and the Zabludovicz Center for Autoimmune Diseases, The Chaim Sheba Medical Center, Tel-Hashomer, Israel, for taking some time to share his knowledge with us a few years ago. This gave us the confidence to pursue our research project and demonstrate the clinical usefulness of GITR triggering in treating autoimmune diseases.

Funding. G.N. was supported by Fondo di Ateneo per la ricerca di base 2015. C.R. was supported by the Ministero dell'Istruzione, dell'Università e della Ricerca (MIUR) under grant number 2015ZT9MXY_001.

Duality of Interest. G.N. is a consultant for Pieris Pharmaceuticals, Inc. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. L.C. set up and performed ELISA experiments. L.C. created the figures. L.C. performed immune characterization of WT, Gitr^−/−, and NOD mice, including flow cytometry, proliferation experiments, and qPCR, with the help of M.G.P. and E.R. L.C. and G.N. analyzed cytofluorimetric data. P.M. and G.B. conducted the histological analysis on islets. P.M., G.B., T.P., and A.G. microencapsulated the hybridoma. P.M., T.P., and A.G. monitored NOD mice, evaluating glycemia. G.B. and S.C. grafted microcapsules in WT, Gitr^−/−, and NOD mice. M.G.P. and E.R. studied the best growth conditions for G3C hybridoma. M.G.P., J.S., G.M., R.C., and C.R. contributed to writing the manuscript. J.S. developed G3C hybridoma. G.M, R.C., and C.R. contributed to conception and supervision of the project. G.N. conceived and supervised the project. G.N. wrote the manuscript. G.N. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Data from this study were presented at the 78th Scientific Sessions of the American Diabetes Association, Orlando, FL, 22–26 June 2018; IPITA (International Pancreas and Islet Transplant Association) 2017 – 16th International Congress, Oxford, U.K., 20–23 June 2017; and the 8th International Conference on Autoimmunity: Mechanisms and Novel Treatments, Rhodes, Greece, 17–22 September 2017.