Tolerogenic Delivery of a Hybrid Insulin Peptide Markedly Prolongs Islet Graft Survival in the NOD Mouse

In both type 1 diabetes (T1D) and the NOD mouse model, autoreactive CD4⁺ and CD8⁺ T cells specifically attack and destroy the insulin-producing β-cells in the pancreatic islets of Langerhans (1). The symptomatic onset of T1D usually occurs very early in life or during adolescence, but can sometimes occur much later (2), and the incidence is rapidly increasing globally (3). For ∼100 years, insulin replacement therapy has been the only treatment for T1D, but significant effort is required to maintain control of blood glucose (BG) levels in the normal range while avoiding either hyperglycemia or hypoglycemia (4).

A viable strategy to achieving euglycemia in T1D is through β-cell replacement therapy in the form of transplantation with isolated pancreatic islets (5,6). Patients with T1D who receive transplanted islets can achieve insulin independence (5,6) and have decreased diabetic complications (7). However, a significant challenge to this approach is the reactivation of memory T cells specific for islet antigens (Ags), resulting in the recurrence of disease in transplant recipients (8,9). Diabetic NOD mice transplanted with syngeneic islets succumb to graft destruction within 2 weeks on average due to autoimmunity (8). In patients with T1D who receive allogeneic islets, circulating islet-specific T cells can be detected upon the recurrence of disease (10), suggesting that autoreactive T cells may also contribute to allograft rejection.

Our group has had a long-standing interest in elucidating the role of CD4⁺ T cells in T1D pathogenesis, and we recently found that unusual hybrid insulin peptides (HIPs) are targets for islet-reactive T cells (11). HIPs are posttranslationally modified Ags consisting of fragments of proinsulin fused to peptides from other β-cell secretory granule proteins. We have shown that the widely studied diabetogenic CD4⁺ T-cell clone BDC-2.5, isolated from a diabetic NOD mouse (12), recognizes an insulin C-peptide/chromogranin A hybrid peptide (11). The BDC-2.5 HIP (2.5HIP) is targeted by at least five additional distinct T-cell receptor (TCR) clonotypes from NOD mice (13), revealing the diversity of pathogenic T cells targeting this Ag. Importantly, the 2.5HIP was confirmed to be present in NOD islets by mass spectrometry (14), and it has been established that endogenous 2.5HIP-specific T cells with a proinflammatory Th1 phenotype are present at a very high frequency in the islets (13,15).

Using a disease model in which BDC-2.5 TCR-transgenic CD4 T cells are adoptively transferred into NOD.scid mice, we recently showed that tolerogenic poly(lactide-co-glycolide) (PLG) nanoparticles (NPs) coupled with the 2.5HIP (hereafter referred to as 2.5HIP NPs or 2.5HIP-PLG) prevented the adoptive transfer of diabetes with BDC-2.5 T cells (16). Given the critical role of 2.5HIP-reactive T cells in the pathogenesis of autoimmune diabetes in the NOD mouse model, the question we wanted to ask in this study is whether tolerance induction targeting polyclonal 2.5HIP-reactive CD4 T cells prolongs the survival of transplanted syngeneic islets.

In this study, we demonstrate that 2.5HIP-reactive CD4⁺ T cells infiltrate syngeneic islet grafts in diabetic NOD mice and that treatment with 2.5HIP NPs significantly protects graft survival. In contrast, treatment with PLG-NPs loaded with insulin (Ins)B_9-23, an epitope that has been considered to be a major Ag in both NOD mice and humans (17), showed no protection. Although tolerogenic delivery of numerous autoantigens, including InsB_9-23, can successfully prevent the onset of spontaneous diabetes in the NOD mouse model (18), very few of these can curb the autoimmune process after the onset of hyperglycemia, highlighting the importance of our findings.

NOD and NOD.scid mice were bred in-house or obtained from The Jackson Laboratory. All mice were housed at the University of Colorado Anschutz Medical Campus in specific pathogen-free conditions. All experiments performed were conducted under protocols approved by the Institutional Animal Care and Use Committee.

Female NOD mice >12 weeks of age were monitored for hyperglycemia by urine glucose testing (Diastix; Bayer), which was confirmed by BG testing with a OneTouch Ultra glucometer (LifeScan). Two consecutive BG readings >250 mg/dL were used for enrollment in our study. All mice used were diabetic for 1–3 weeks prior to transplant and had a BG >300 mg/dL on day −7 before transplant, with most being >450 mg/dL. All mice maintained a BG >250 mg/dL until the day of the transplant.

Islets were isolated from female NOD.scid or NOD mice by infusion through the bile duct with CIzyme RI (VitaCyte), followed by Lympholyte 1.1 (Cedarlane Laboratories) gradient centrifugation and handpicking of islets. Diabetic recipient mice received 500 islet equivalents beneath the left kidney capsule. After transplantation, BG levels were monitored daily for 1 week and then three times a week thereafter. Mice (<10%) that did not achieve normoglycemia (<180 mg/dL) within 4 days posttransplant were discarded from the study. Two consecutive BG readings >250 mg/dL were considered as recurrence of disease.

All peptides were obtained from CHI Scientific or Synpeptide at a purity >95%. Peptides used in this study were hen egg lysozyme (HEL)_11-25 (AMKRHGLDNYRGYSL), InsB_9-23 (SHLVEALYLVCGERG), and water-soluble 2.5HIP (RGG-LQTLALWSRMD-GGR).

Carboxylated single-emulsion PLG-NPs were synthesized in the laboratory of S.D.M. as previously described (19,20), lyophilized, shipped to the laboratory of K.H., and stored at −20°C until use. NPs ranged from an average of 500 to 800 nm in diameter. The laboratory of S.D.M. has previously published the specific details on NP size distribution and purity (19–21). Before Ag coupling, NPs were suspended and washed three times in 1× PBS, pH 7.4 (Life Technologies). NPs were then resuspended at 50 mg/mL in PBS with peptide (4 mg/mL) and ethylcarbodiimide (16 mg/mL) at room temperature (21°C) for 1 h with mixing by pipetting every 10 min. Ag-coupled NPs were washed three times with PBS, and a total of 2.5 mg (in PBS) was administered by intravenous injection into recipient mice via the retro-orbital sinus.

The amount of peptide coupled to PLG-NPs was determined by the laboratory of S.D.M., as previously reported (22). Peptide was coupled to NPs in five separate reactions by the standard ethylcarbodiimide coupling protocol, as described in the section above; 0.5 mg of Ag-coupled NPs from each reaction was then dissolved in 10 μL of DMSO. The resulting solutions were analyzed using a 3-(4-carboxybenzoyl) quinoline-2-carboxaldehyde assay (Thermo Fisher Scientific) and read on a fluorescent microplate reader with excitation at 465 nm and emission at 550 nm. The mean peptide load as well as the SD were determined for the five separate reactions.

Within 24–72 h upon recurrence of disease or on the indicated day of analysis, islet grafts were resected from the kidney capsule and digested with 2.4 units/mL of Dispase II (Roche) and 10 μg/mL of DNase I (Roche) for 30 min at 37°C. Any remaining tissue was homogenized in a glass tissue homogenizer to prepare single-cell suspensions. This protocol allowed for recovery of several million leukocytes from the islet graft.

Allophycocyanin- or phycoerythrin (PE)–conjugated MHC class II (I-A^g7) tetramers loaded with HEL_11-25 (AMKRHGLDNYRGYSL), InsB_9-23 mimotope (Insp8G; SHLVEALYLVCGGEG), 6.9HIP (LQTLALNAARDP), and 2.5HIP (LQTLALWSRMD) were obtained from the National Institutes of Health (NIH) Tetramer Core. Allophycocyanin- or PE-conjugated MHC class I (H2-K^d) tetramer loaded with islet-specific glucose-6-phosphatase catalytic subunit-related protein (IGRP)_206-214 (VYLKTNVFL) was also obtained from the NIH Tetramer Core.

Spleen samples were homogenized, and red blood cells were removed using red blood cell lysing buffer Hybri-Max (Sigma-Aldrich). Single-cell suspensions from spleen or islet grafts were stained with I-A^g7 or H-2K^d tetramers for 40 min at 37°C and then counterstained with antibodies at room temperature for 20 min. Cells were stained with eBioscience fixable viability dye eFluor 780 (Thermo Fisher Scientific) to discriminate live cells. For intracellular staining, cells were fixed using the eBioscience Foxp3 transcription factor staining buffer set (Thermo Fisher Scientific) at 4°C overnight or for 30 min. Cells were permeabilized using the permeabilization reagent included in the buffer set and stained with intracellular antibodies at room temperature for 1 h. The dump gate anti-mouse antibodies used were anti-CD11b:BB700 (M1/70; BD Biosciences), anti-CD11c:BB700 (HL3; BD Biosciences), anti-GR1:BB700 (1A8; BD Biosciences), and anti-CD19:BB700 (1D3; BD Biosciences). Other antibodies used included anti-CD45:BUV395 (30-F11; BD Biosciences), anti-CD4:BV711 (GK1.5; BioLegend), anti-CD4:PECy5 (GK1.5; BioLegend), anti-CD8:BV510 (53–6.7; BioLegend), anti-CD8:BV711 (53–6.7; BD Biosciences), anti-CD25:BB515 (PC61; BD Biosciences), anti-CD25:PECy5 (PC61; BioLegend), anti–CTLA-4:PE-Cy7 (4C10–4B9; BioLegend), anti-GITR:BV510 (DTA-1; BD Biosciences), anti-CD73:BV605 (TY/11.8; BioLegend), antifolate receptor 4 (FR4):PE-Cy7 (12A5; eBioscience), anti-CD44:BV786 (IM7; BD Biosciences), CD62L:BV605 (MEL-14; BD Biosciences), and anti–PD-1:BV785 (29F.1A12; BioLegend). Intracellular antibodies used included anti-Foxp3:eFluor 450, PECy7, and PECy5 (FJK-16s; eBioscience), anti–Ki-67:FITC (SolA15; eBioscience), anti–Ki67:PE-Cy7 (16A8; BioLegend), and anti-Helios:PECy5 (22F6; eBioscience). Samples were run on an LSRFortessa X-20 (BD Biosciences) flow cytometer, and data were analyzed using FlowJo software (Tree Star). For all studies, gates were set on CD45⁺, live, dump⁻, CD4⁺, or CD8⁺ T cells.

Single-cell suspensions from spleen or islet grafts were stained with I-A^g7 or H-2K^(d) tetramers in the presence of 1 μg/mL GolgiPlug (BD Biosciences), 0.1 μg/mL phorbol 12-myristate 13-acetate (PMA; Sigma-Aldrich), and 1 μg/mL ionomycin (Sigma-Aldrich) for 2.5 h at 37°C. Cells were washed, surface stained with antibodies, fixed, permeabilized, stained with intracellular antibodies, and analyzed by flow cytometry, as described above. Antibodies used to detect intracellular cytokines were anti–IFN-γ:eFluor450 (XMG1.2; eBioscience), anti–IFN-γ:PECy7 (XMG1.2; BD Biosciences), and anti–TNF-α:FITC (MP6-XT22; eBioscience).

GraphPad Prism version 8.0 software was used for data analysis. Incidence of diabetes was compared using the Mantel-Cox test (log-rank test) and Gehan–Breslow–Wilcoxon test. For flow cytometry data, unpaired two-tailed t tests and one-way ANOVA were used to determine statistical significance. Statistical significance was defined as: *P < 0.05; **P < 0.01; ***P < 0.001; and ****P < 0.0001.

The data sets during this study are available from the corresponding author upon reasonable request. The NPs that support the findings of this study are available from S.D.M., but restrictions apply to the availability of these data, which were used under license for the current study and therefore are not publicly available. NPs may be available from the authors and/or S.D.M. upon reasonable request and with the permission of S.D.M.

In our previous study, analysis with MHC class II (I-A^g7) tetramers indicated that 2.5HIP-reactive CD4⁺ T cells are present in the pancreas of diabetic NOD mice in greater abundance than InsB_9-23–reactive cells (13). In this study, we determined if these epitopes are targeted during the autoimmune attack of an islet transplant and whether tolerizing 2.5HIP- or insulin-reactive T cells could prevent acute graft destruction. For these studies, spontaneously diabetic female NOD mice were transplanted with 500 syngeneic NOD.scid islet equivalents, and the recipients were injected with 2.5 mg of Ag-coupled NPs, once before transplant and twice posttransplant. The HEL_11-25 peptide was used as an irrelevant Ag control. To ensure that the Ags being tested coupled to the NPs efficiently, a carboxybenzoyl) quinoline-2-carboxaldehyde assay was used to quantify the amount of peptide loaded, and results indicated 24, 36, or 70 μg of peptide/mg of NPs for HEL, 2.5HIP, and InsB_9-23, respectively (Supplementary Table 1). Although there was some variation in Ag loading, previous work has shown that Ag-coupled NPs containing as little as 6 μg of peptide/mg of NP are effective for tolerance induction in the experimental autoimmune encephalomyelitis model of multiple sclerosis (20).

As expected, untreated (control) NOD mice displayed a median graft survival of only 14 days posttransplant due to the autoimmune attack of the grafted islets (Fig. 1A and B). There were similar kinetics of rapid disease recurrence in HEL and InsB_9-23 NP-treated mice (Fig. 1A and B). In contrast, treatment with 2.5HIP NPs significantly prolonged graft survival and led to delayed recurrence of diabetes, with a median graft survival observed at day 58 (Fig. 1A and B). Importantly, there were no major differences in initial BG values of mice at the time of transplant that would account for these findings (Supplementary Table 2); in fact, the average values for all groups were markedly elevated (>450 mg/dL) compared with the commonly used cutoff of 250 mg/dL for diabetes onset in the NOD mouse model.

We used various peptide/I-A^g7 tetramers to analyze the composition of CD4⁺ T cells in untreated transplant recipients compared with those treated with Ag-loaded NPs. The insulin tetramer contained a B_9-23 mimotope (Insp8G), modified to improve binding to I-A^g7 (23), and was used as a means of assessing insulin-specific CD4⁺ T cells in the graft. The 6.9HIP tetramer was also included to assess another significant HIP-reactive T cell population found in NOD mice. The 6.9HIP Ag consists of the same insulin C-peptide fragment present in the 2.5HIP fused to a peptide from islet amyloid polypeptide and is the ligand for the diabetogenic BDC-6.9 T-cell clone (24). Similar to 2.5HIP tetramer-positive (tet⁺) T cells, endogenous 6.9HIP tet⁺ cells infiltrate the islets of prediabetic and diabetic NOD mice (13,24). When transplant recipient mice were analyzed upon disease recurrence, we found that the percentage of splenic 2.5HIP tet⁺ cells in 2.5HIP NP-treated mice (∼0.7%) was increased ∼10-fold compared with other groups (Fig. 2A and B). Additionally, while HEL tet⁺ cells were nearly absent from the spleen of control, InsB_9–23 NP, and 2.5HIP NP-treated mice, there was an ∼100-fold expansion in the percentage of these cells in HEL NP-treated mice (∼0.3%) (Fig. 2B). A similar expansion in the absolute number of splenic 2.5HIP tet⁺ or HEL tet⁺ T cells was observed in 2.5HIP NP and HEL NP-treated mice, respectively (Fig. 2C). The percentages of Insp8G tet⁺ and 6.9HIP tet⁺ cells were unchanged in mice treated with 2.5HIP or HEL NPs (Fig. 2B). These findings are similar to what we have observed in prediabetic NOD mice treated with Ag-loaded NPs (B.L.J., B.B., and K.H., unpublished observations) and suggest that clonal expansion of cognate Ag-specific CD4⁺ T cells occurs in the spleen. We did not detect an expansion of splenic Insp8G tet⁺ cells in mice treated with InsB_9-23 NPs (Fig. 2B), possibly due to inefficient priming of insulin-reactive T cells specific for this mimotope.

We also analyzed Ag-specific T cells at the disease site, and the tetramer gating strategy in the islet graft is shown in Supplementary Fig. 1A–C. Unlike what was observed in the spleen, the frequencies and numbers (data not shown) of islet graft-infiltrating Insp8G tet⁺, 6.9HIP tet⁺, and 2.5HIP tet⁺ cells were fairly similar between all groups (Supplementary Fig. 2A and B). Additionally, there were no significant differences between groups in the percentages or numbers of CD4⁺ or CD8⁺ T cells (data not shown). There was a significant increase in the frequency of HEL tet⁺ cells in the graft of HEL NP-treated mice (Supplementary Fig. 2A and B), a surprising observation since this is an irrelevant Ag that is not expressed in the islet transplant itself.

Since the frequency of autoreactive 2.5HIP-specific T cells was not significantly altered in the islet graft of mice in which tolerance was induced with 2.5HIP NPs, we investigated whether treatment resulted in an altered phenotype in these cells. Cell suspensions from grafts of mice euthanized after disease recurrence were stained with I-A^g7 tetramers and stimulated with PMA and ionomycin to detect cytokine production by Ag-specific T cells. The frequency of 2.5HIP tet⁺ cells capable of producing both IFN-γ and TNF-α was between 60 and 80% in untreated mice and in HEL NP and InsB_9-23 NP-treated mice, but was dramatically reduced to 10% on average in mice treated with 2.5HIP NPs (Fig. 3A and B). Unexpectedly, the percentage of IFN-γ⁺TNF-α⁺ cells found in bulk graft-infiltrating CD4⁺ cells was also markedly reduced (to ∼4%) in mice treated with 2.5HIP NPs compared with ∼20–40% found in all other groups (Fig. 3A and B). Moreover, the frequencies of IFN-γ/TNF-α–coproducing CD4⁺ T cells with different autoantigen specificities (InsB_9-23– and 6.9HIP-reactive T cells) were also significantly reduced in 2.5HIP NP-treated mice (Fig. 3A and B). However, it is important to note that the percentage of polyclonal CD4⁺ or islet-specific T cells capable of producing IFN-γ or TNF-α alone was unchanged or even increased in mice treated with 2.5HIP NPs (Fig. 3A). This demonstrates that complete effector function was not lost, only the specific ability to produce both IFN-γ and TNF-α, a likely result of performing the analysis when tolerance to the islet graft had been lost, albeit to a lesser degree in 2.5HIP NP-treated mice.

It has been demonstrated that CD8⁺ T cells recognizing the H-2K^d–restricted peptides 206–214 from IGRP are important in driving spontaneous autoimmunity in NOD mice (25). Additionally, IGRP is targeted by CD8⁺ T cells during the destruction of syngeneic islet grafts (26). Based on these observations, we used an H-2K^d tetramer to analyze the IGRP-specific CD8⁺ T cell population. We found that although the percentage of IGRP tet⁺ cells was not altered in the graft of 2.5HIP NP-treated mice compared with untreated or HEL NP controls (Fig. 4A and B), the production of inflammatory cytokines by these cells was significantly reduced. Roughly 60% of CD8⁺ IGRP tet⁺ T cells in untreated or HEL NP-treated mice produced both IFN-γ⁺ and TNF-α⁺ in response to stimulation, while these levels were reduced to 15% on average in mice treated with 2.5HIP NPs (Fig. 4C and D). The percentage of polyclonal CD8⁺ T cells that coproduced IFN-γ/TNF-α was significantly reduced in 2.5HIP NP-treated mice as well (Fig. 4C and D). Taken together, these data demonstrate that Ag-specific treatment with 2.5HIP NPs negatively regulates effector function of both CD4⁺ and CD8⁺ T cells responding to other islet autoantigens within the graft.

We previously showed that 2.5HIP NPs tolerize adoptively transferred BDC-2.5 TCR-transgenic cells through induction of an anergic program in which cells are unresponsive to antigenic stimulation (16). In T cells, anergy is characterized by a lack of inflammatory cytokine production and impaired cell division (27). Whereas a large percentage of 2.5HIP tet⁺ cells from islet grafts of control, HEL NP, or InsB_9-23 NP-treated mice expressed the proliferation marker Ki-67, 2.5HIP tet⁺ Ki-67⁺ cells were nearly absent in 2.5HIP NP-treated mice (Fig. 5A and B). We also observed that there was a significant reduction in the frequency of Insp8G tet⁺Ki-67⁺ and 6.9HIP tet⁺Ki-67⁺ cells in mice treated with 2.5HIP NPs compared with those treated with HEL NPs (Fig. 5B), but the results were not as striking as those observed for 2.5HIP tet⁺ cells. In addition, there was no change in Ki-67 expression in the polyclonal CD4⁺ population (Fig. 5B), indicating that treatment with 2.5HIP NPs primarily restrains the proliferative capacity of 2.5HIP-specific T cells.

We examined additional markers of putative T-cell anergy in islet-infiltrating T cells. It has been difficult to study anergy in vivo among a heterogeneous population of T cells due to a lack of cell-surface markers to specifically identify anergic cells. However, the finding by Kalekar et al. (28) and Martinez et al. (29) that anergic CD4⁺ T cells can be found in the CD44^hiFoxp3⁻ effector T cell compartment through coexpression of high levels of the ectonucleotidase CD73 and FR4 allowed for tracking of anergic cells upon tolerance induction. Gating on 2.5HIP tet⁺CD44^hiFoxp3⁻ cells, we found that the percentage of CD73^hiFR4^hi cells was very high (∼70%) in 2.5HIP NP-treated mice, compared with untreated controls or those treated with HEL or InsB_9-23 NPs (Fig. 6A and B). Likewise, there was a significant increase in the level of CD73 and FR4 expression as measured by mean fluorescence intensity (Fig. 6C). The anergic 2.5HIP tet⁺ cells were restricted to the Ag-experienced subset, as few CD73^hiFR4^hi high cells were detected in the CD44^loFoxp3⁻ compartment (Fig. 6A). Although there was little increase in the frequency of total CD4⁺, Insp8G tet⁺, or 6.9HIP tet⁺ anergic cells in 2.5HIP NP-treated mice, we observed that a substantial portion (≥40%) of Insp8G tet⁺CD44^hiFoxp3⁻ cells was consistently CD73^hi FR4^hi in all groups of mice (Fig. 6B). These results suggest that 2.5HIP NPs induce anergy, but only in T cells specific for the 2.5HIP Ag and not in those specific for other β-cell epitopes. Furthermore, unlike 2.5HIP-specific T cells, a large portion of InsB_9-23–specific cells naturally have an anergic phenotype upon encountering Ag in islet grafts.

Since our data suggested that treatment with 2.5HIP NPs induces a form of dominant tolerance, it was critical to examine Foxp3⁺ regulatory T cells (Tregs) in treated animals. When 2.5HIP NP-treated mice were analyzed upon disease recurrence, there were very few 2.5HIP tet⁺ Tregs detected (data not shown). Since at this time point tolerance to the graft had failed, it was necessary to investigate potential mechanisms of tolerance induction prior to graft destruction. When mice were analyzed between 1 and 2 weeks posttransplant, we saw a significant increase in the total number of 2.5HIP-specific Foxp3⁺ and CD25⁺Foxp3⁺ Tregs in the spleen of 2.5HIP NP-treated mice in comparison with control mice or mice treated with HEL NPs, which had very few (<50) 2.5HIP-specific Tregs (Fig. 7A and B). In the islet graft of 2.5HIP NP-treated mice, the total number of 2.5HIP tet⁺ Tregs was also higher but not significantly so (Fig. 7A and B). In contrast, there was no increase in the percentage of 2.5HIP tet⁺ Tregs in the spleen or graft of mice treated with 2.5HIP NPs (Fig. 7B). Therefore, treatment with 2.5HIP NPs led to an expansion in the total number but not the percentage of splenic 2.5HIP-specific Tregs. We were not able to compare the phenotype of 2.5HIP tet⁺ Tregs in 2.5HIP NP-treated mice and other groups since there were such low numbers of these cells. However, when 2.5HIP-specific CD25⁺Foxp3⁺ Tregs from the spleen of 2.5HIP NP-treated mice were evaluated, we found they expressed canonical Treg activation markers, such as GITR, CTLA-4, and Helios (Fig. 7C). Additionally, these markers were expressed at higher levels on 2.5HIP tet⁺ Tregs compared with polyclonal CD4⁺ Tregs present in the same mice (Fig. 7C and D), suggesting that the induced Ag-specific Tregs were more activated.

In this study, we showed that 2.5HIP-coupled PLG NPs could be used to delay recurrent autoimmunity in NOD mice and prolong the survival of islet isografts, confirming that this neoepitope is important in both the pathogenesis and regulation of disease in the NOD mouse model. In contrast, we found that NPs loaded with InsB_9-23, a conventional autoantigen long considered to be a critical target in NOD mice and patients with T1D (30), did not have any effect on graft survival. Although the InsB_9-23 NPs were not protective in graft recipients with diabetes, we have observed that these NPs do have some efficacy in delaying onset of diabetes when administered to young prediabetic mice (B.L.J., B.B., and K.H., unpublished observations). Induction of tolerance with InsB_9-23–loaded apoptotic Ag-coupled splenocytes was previously reported to prevent diabetes onset when administered to 5-week-old prediabetic mice, but not in 20-week-old mice, which have more advanced insulitis (31). However, in the same study, the authors also found that treatment with apoptotic Ag-coupled splenocytes loaded with intact insulin showed efficacy in both young and older mice, suggesting that epitope spreading to other portions of insulin may occur as disease progresses (31). The left (N-terminal) portion of both of the mouse HIPs identified contains a fragment from insulin C-peptide (11,24), implying that tolerance against other regions of insulin outside of the B-chain may need to be induced in mice with more established disease. Based on our previous work (13), we suspect that 2.5HIP tet⁺ cells consist of numerous clonotypes that contribute to graft destruction and are therefore targeted by tolerogenic therapy. Moreover, the 2.5HIP is highly antigenic and can stimulate relevant T-cell clones at low nanomolar concentrations (11), whereas InsB_9-23 is a weakly stimulating peptide that binds poorly to I-A^g7 (23), suggesting that enhanced Ag presentation of the 2.5HIP may lead to more efficient priming and tolerance induction. Recently, it was shown that fusion of sequences from InsB_9-23 to fragments from insulin C-peptide results in superagonists that are much more antigenic for insulin-reactive T-cell clones than InsB_9–23 (32). Perhaps tolerance induction with modified forms of InsB_9-23 would be more effective than using the native peptide.

One possible explanation for why 2.5HIP NP-treated mice are protected from disease recurrence could be that the diabetogenic T cells do not traffic to the islet graft. We previously showed in the BDC-2.5 adoptive transfer model that treatment with 2.5HIP NPs leads to T-cell accumulation in the spleen and reduced infiltration of the pancreas (16). For this study, we performed a thorough analysis of Ag-specific T cells using I-A^g7 tetramers in untreated transplant recipients and those treated with HEL, InsB_9-23, or 2.5HIP NPs. We found there was a significant expansion of 2.5HIP tet⁺ cells in the spleen of mice treated with 2.5HIP NPs, indicating that Ag-specific T cells initially undergo clonal expansion in response to the encounter with Ag delivered on PLG-NPs. However, there were no differences in the frequency of InsB_9-23 mimotope (Insp8G) tet⁺, 6.9HIP tet⁺, or 2.5HIP tet⁺ T cells in the graft of mice treated with 2.5HIP NPs versus controls. Further analysis of islet grafts showed that 2.5HIP-specific T cells were dysfunctional in 2.5HIP NP-treated mice upon disease recurrence. Instead of a Th1 phenotype, the graft-infiltrating cells expressed markers associated with T-cell anergy.

Furthermore, our data show that treatment with 2.5HIP NPs, but not InsB_9-23 NPs suppressed the effector function of CD4⁺ and CD8⁺ graft-infiltrating T cells, including those specific for islet autoantigens. The most striking observation was that tolerizing with the I-A^g7–restricted 2.5HIP epitope for CD4⁺ T cells inhibited IFN-γ and TNF-α coproduction by H-2K^d–restricted IGRP-specific CD8⁺ T cells, which have been shown to be pathogenic in the context of islet transplantation (26). Wong et al. (26) reported that inhibiting IGRP-specific CD8⁺ T cell function through the administration of soluble IGRP peptide can delay islet isograft destruction, although not to the extent of that we observed using 2.5HIP NPs. We also found that induction of anergy with 2.5HIP NPs is restricted to T cells specific for 2.5HIP, but that InsB_9-23–specific T cells naturally express anergy markers in islet grafts in the absence of any treatment. It has previously been shown that a large percentage of Ag-experienced insulin tet⁺ cells have a CD73^hi FR4^hi anergic phenotype in prediabetic NOD mice, which develops in the draining pancreatic lymph nodes (33). Extrathymic Ag-presentation of InsB_9-23 likely results in peripheral tolerance via mechanisms such as induction of anergy and peripherally derived Tregs, but this is not the case for the presentation of a neoepitope such as the 2.5HIP. This is substantiated by the fact that we and others have observed that in prediabetic mice, a large percentage of insulin-reactive T cells are Foxp3⁺ Tregs, whereas most HIP-reactive T cells do not express Foxp3 (13,15).

How then does treatment with 2.5HIP NPs promote the suppression of other T-cell populations? Results are most consistent with a form of dominant tolerance in which altered reactivity in 2.5HIP-reactive T cells impacts reactivity to other autoantigens, possibly through a type of linked suppression (34,35). Analysis of 2.5HIP NP-treated mice at an early time point after transplant revealed that there was marked expansion in the numbers of splenic 2.5HIP-reactive CD25⁻Foxp3⁺ and CD25⁺Foxp3⁺ Tregs, which also expressed high levels of markers associated with Treg activation, such as GITR, CTLA-4, and Helios. Upon intravenous injection, Ag-loaded NPs can be found associated with marginal zone macrophages in the spleen (36), an organ that is involved in the clearance of apoptotic cells and maintenance of peripheral tolerance (37). Hence, it is conceivable that the primary site of tolerance induction is the spleen. Our data support a hypothesis in which treatment with 2.5HIP NPs promotes the induction of 2.5HIP-reactive Tregs in the spleen, which regulate effector T cell responses to islet Ags in the graft. Yet, tolerance eventually failed in 2.5HIP NP-treated mice, and upon recurrence of disease, 2.5HIP tet⁺ Tregs could no longer be found (data not shown). We propose that the breakdown in tolerance was caused by a loss of 2.5HIP-reactive Treg stability and/or a corresponding outgrowth of pathogenic T cells specific for other Ags.

The delivery of Ag alone is likely suboptimal to expand a stable Foxp3⁺ Treg population, and we imagine that combining Ag-specific treatment with other nonspecific immunotherapies will be most effective for tolerance induction. For example, NOD mice have defective IL-2 production (38), which could be a barrier to sustaining the long-term survival of induced Tregs and necessitate the need for exogenous supplementation of IL-2. Another promising strategy has been investigated in which Ag was encapsulated within PLG-NPs, and, in addition, transforming growth factor-β was coupled to the NP surface (39). This approach led to more efficient CD25⁺Foxp3⁺ Treg induction in vitro and improved therapeutic outcomes in a mouse model of multiple sclerosis compared with NPs bearing Ag alone. Therefore, the codelivery of Ag and immunomodulatory cytokines may represent a superior strategy to improve Treg induction or survival.

Collectively, this work shows that 2.5HIP PLG-NPs induce tolerance through two distinct pathways, one being the expansion of 2.5HIP-specific Tregs and the other induction of a hyporesponsive state in 2.5HIP-specific effector T cells. Our group has now identified HIPs in human islets (14), and other studies have shown that HIP-reactive T cells are present in the residual islets of deceased organ donors with T1D (11,40), as well as in peripheral blood of new-onset patients (41). There may therefore be potential for targeting HIP-reactive T cells as a means of regulating autoimmunity in humans with T1D.

Acknowledgments. The authors thank M. Coulombe for assistance with the protocol for islet graft dissociation and T.A. Wiles for helpful discussions. The authors also thank the NIH Tetramer Core Facility for providing the MHC tetramer reagents used in this research, the AMC ImmunoMicro Flow Cytometry Shared Resource RRID:SCR_021321, and the University of Colorado Diabetes Research Center (P30-DK116073 grant).

Funding. This work was supported by the NIH/National Institute of Diabetes and Digestive and Kidney Diseases (R01-DK-081166 to K.H., R21-AI-133059 to R.L.B., and F31-DK-113693 to B.L.J.), and JDRF (2-SRA-2018-566-S-B and 2-SRA-2020-907-S-B to K.H.).

Duality of Interest. S.D.M. is a cofounder, a member of the scientific advisory board, a grantee of, and holds stock options in Cour Pharmaceuticals Development Company; a paid consultant for Cour Pharmaceuticals Development Company and Takeda Pharmaceuticals International Co.; a paid consultant and member of the scientific advisory board of NextCure, Inc.; a paid consultant for Kite Pharma; and a paid consultant and member of the scientific advisory board of Myeloid Therapeutics. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. B.L.J. conceptualized the study and designed the experiments. B.L.J., J.E.D., K.S.B., T.N., B.B., and J.G. conducted the investigations. B.L.J. and K.H. wrote the manuscript. R.G.G., S.D.M., R.L.B., and K.H. reviewed and edited the manuscript. K.H. supervised the work. K.H. is the guarantor of this work and, as such, had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. This study was presented at the 103rd Annual Meeting of the American Association of Immunologists, San Diego, CA, 9–13 May 2019.