Nanoparticles Containing an Insulin-ChgA Hybrid Peptide Protect from Transfer of Autoimmune Diabetes by Shifting the Balance between Effector T cells and Tregs

Abstract

CD4 T cells play a critical role in promoting the development of autoimmunity in Type 1 Diabetes (T1D). The diabetogenic CD4 T cell clone BDC-2.5, originally isolated from a non-obese diabetic (NOD) mouse, has been widely used to study the contribution of autoreactive CD4 T cells and relevant antigens to autoimmune diabetes. Recent work from our lab has shown that the antigen for BDC-2.5 T cells is a hybrid insulin peptide (2.5HIP) consisting of an insulin C-peptide fragment fused to a peptide from chromogranin A (ChgA), and that endogenous 2.5HIP-reactive T cells are major contributors to autoimmune pathology in NOD mice. The objective of this study was to determine if poly(lactide-co-glycolide) (PLG) nanoparticles (NPs) loaded with the 2.5HIP antigen (2.5HIP-PLG) can tolerize BDC-2.5 T cells. Infusion of 2.5HIP-PLG NPs was found to prevent diabetes in an adoptive transfer model by impairing the ability of BDC-2.5 T cells to produce pro-inflammatory cytokines through induction of anergy, leading to an increase in the ratio of Foxp3⁺ regulatory T cells to IFN-γ⁺ effector T cells. This work is the first to use a hybrid insulin peptide, or any neoepitope, to re-educate diabetogenic T cells and may have significant implications for the development of an antigen-specific therapy for T1D patients.

Introduction

Type 1 Diabetes (T1D) is an organ-specific autoimmune disease in which β-cells of the pancreatic islets are destroyed, leading to decreased insulin production (1). Currently, there is no cure for T1D and treatment consists of lifelong insulin replacement therapy, which does not address the underlying mechanisms of disease. Much of our understanding of the pathogenesis of T1D has come from studies in the non-obese diabetic (NOD) mouse model and it is well established that autoreactive CD4⁺ and CD8⁺ T cells contribute to the destruction of β-cells in both NOD mice and T1D patients. Different approaches have been investigated to regulate the autoimmune response, including administration of anti-CD3 antibody (2, 3) and blockade of pro-inflammatory cytokines (4, 5). However, these treatments are non-specific and thus can have a wide range of side effects. The ideal therapy for T1D would be induction of antigen-specific tolerance, specifically targeting self-reactive T cells while leaving the remainder of the immune system intact. The major hurdle in developing antigen-specific therapies is often a lack of understanding about the precise role of disease-associated antigens and their epitopes in the disease process. Insulin is considered to be a major autoantigen in both NOD mice and humans (6) suggesting it may be a potential therapeutic target. Administration of insulin peptides to young NOD mice can prevent the development of diabetes (7, 8) and intact insulin can reverse new-onset diabetes in a portion of mice when delivered on antigen-coupled apoptotic splenocytes (Ag-SP) (9). While preclinical studies in mice were promising, development of an antigen-specific therapy for human T1D has not yet been successful (10, 11). For example, insulin given orally, intranasally, or intravenously has been tested in large prevention studies without any evidence of clinical benefit (12–14). Although lack of success with the insulin trials could be attributed to many factors, it may be that alternative methods for delivery of T cell epitopes in a tolerogenic fashion will be required.

The focus of our lab has been to identify antigens for diabetogenic T cells with one goal being discovery of novel peptides that can be used to induce antigen-specific tolerance. The diabetogenic CD4⁺ T cell clone BDC-2.5, derived from the NOD mouse, has been a valuable tool for understanding the role of CD4 T cells in autoimmune diabetes (15–17). BDC-2.5 T cells have a Th1 phenotype and produce the pro-inflammatory cytokines IL-2, IFN-γ, and TNF-α upon stimulation. We recently discovered that the antigen for BDC-2.5 T cells is a neoepitope formed by fusion of a sequence from insulin C-peptide and WE14, a natural cleavage product of the secretory granule protein chromogranin A (ChgA) (18). The resulting BDC-2.5 hybrid insulin peptide (2.5HIP) is very agonistic for BDC-2.5 T cells and endogenous 2.5HIP-reactive CD4⁺ T cells are present at a high frequency in the islets of NOD mice (18–20). The goal of this study was to determine if 2.5HIP could be used to induce tolerance in BDC-2.5 T cells. We developed a strategy whereby 2.5HIP is delivered on carboxylated poly(lactide-co-glycolide) (PLG) nanoparticles (NPs), based on prior work showing that PLG NPs can serve as surrogates for Ag-SP and are effective antigen carriers to induce tolerance and protect from disease in the experimental autoimmune encephalomyelitis (EAE) model of multiple sclerosis (21, 22). In the EAE system, antigen-loaded PLG NPs (Ag-PLG) localize to the splenic marginal zone and liver when infused intravenously (i.v.) and are thought to induce T-cell anergy and regulatory T cells (Tregs) in an IL-10- and PD-L1-dependent fashion. Recent work from the Miller lab in the NOD mouse model has shown that PLG NPs encapsulating either the MHC II-restricted p31 mimotope or the MHC I-restricted IGRP NRPA7 mimotope can induce tolerance in BDC-2.5 CD4⁺ and NY8.3 CD8⁺ T cells, respectively (23). Although the PD-L1/PD-1 axis and CTLA-4 are critical for tolerance induction in diabetogenic T cells, the mechanisms by which tolerance is induced and maintained with Ag-PLG, are incompletely understood. Here we show that 2.5HIP-loaded PLG NPs protect from transfer of autoimmune diabetes primarily by decreasing the effector function of pathogenic CD4⁺ T cells through anergy induction and by shifting the balance between Tregs and effector T cells.

Materials and Methods

Mice

NOD, NOD.scid, and BDC-2.5TCR-Tg.NOD mice were bred in house. NOD.Foxp3^{DTR/ EGFP} mice were obtained from the Jackson Laboratory and crossed to BDC-2.5TCR-Tg.NOD mice. All mice were housed in specific pathogen-free conditions at the University of Colorado School of Medicine. Experimental mice were monitored for development of diabetes by urine glucose testing (Diastix, Bayer) and hyperglycemia was confirmed by blood glucose testing with a OneTouch Ultra glucometer (LifeScan). Mice were considered diabetic when blood glucose levels were >15 mmol/l (270 mg/dl) for two consecutive readings. All experimental procedures were conducted in accordance with a protocol approved by the Institutional Animal Care and Use Committee.

Peptides

Peptides were obtained commercially from CHI Scientific or Synpeptide at a purity >95%. Peptides used in this study were HEL_11–25 (AMKRHGLDNYRGYSL) and water-soluble 2.5HIP (RGG-LQTLALWSRMD-GGR).

Activation and adoptive transfer of BDC-2.5 T cells

Single cell suspensions were prepared from the spleen and lymph nodes (inguinal, mesenteric, brachial, axillary, pancreatic) of BDC-2.5TCR-Tg.NOD or BDC-2.5TCR-Tg.NOD.Foxp3^DTR/EGP+ mice by homogenization in a glass tissue homogenizer. The cells were activated in vitro with 2.5HIP (2.5 μM) in complete medium (CM) containing 2% EL-4 supernatant as a source of IL-2. CM is Dulbecco’s Modified Eagle’s Medium supplemented with 44 mmol/L sodium bicarbonate, 0.55 mmol/L L-arginine, 0.27 mmol/L L-asparagine, 1.5 mmol/L L-glutamine, 1 mmol/L sodium pyruvate, 50 mg/L gentamicin sulfate, 50 mmol/L 2-mercaptoethanol, 10 mmol/L HEPES, and 10% FCS. Cells were plated at a final concentration of 1 × 10⁶ cells/mL in 162 cm² flasks and incubated at 37 °C in a humidified chamber containing 10% CO₂. After 3 days cells were harvested, washed, and 10 × 10⁶ viable T cells were injected i.v. into 8–12 week old NOD.scid recipients.

Preparation and administration of antigen-coupled PLG nanoparticles

Nanoparticles (500 nm carboxylated single emulsion poly(lactide-co-glycolide) (PLG) particles) were synthesized in the Miller laboratory as previously described (21, 22). The particles were lyophilized and stored at −20 °C. Before use, the particles were resuspended and washed 3 times in 1X PBS pH 7.4 (Gibco), suspended at 50 mg/ml in PBS and coupled with peptide (4 mg/ml) using ethylcarbodiimide (ECDI) (16 mg/ml) at room temperature (21 °C) for 1 hr with intermittent shaking. Antigen-coupled particles (Ag-PLG) were washed 3 times with PBS. A total of 1.75 mg of Ag-PLG was administered i.v. in a volume of 150 μl PBS one day after T cell transfer. Specific details on PLG nanoparticle size distribution and purity may be found in previous publications (21, 22, 24).

Administration of soluble peptide

A 2.5HIP stock solution was diluted in PBS to the desired concentration and administered i.v. in a volume of 100 μl one day after T cell transfer.

Quantification of peptide coupled to PLG nanoparticles

The amount of 2.5HIP coupled to PLG nanoparticles was measured using a CBQCA (3-(4carboxybenzoyl) quinoline- 2-carboxaldehyde) assay (Thermo Fisher Scientific) as previously described (25). Briefly, 2.5HIP was coupled to single emulsion particles in five separate tubes by a standard coupling protocol using ECDI; 0.5 mg of 2.5HIP-coupled nanoparticles were then dissolved in 10 μl of DMSO. The resultant solution was analyzed using the CBQCA assay and read on a fluorescent microplate reader with emission at 550 nm and excitation at 465 nm. The mean peptide load was determined for the 5 independent coupling reactions.

Histology

The pancreas was removed from NOD.scid mice and fixed with 10% formalin overnight; samples were then transferred to 70% ethanol. Samples were paraffin embedded, sectioned, and stained with hematoxylin and eosin (H&E) by the Morphology and Phenotyping core at University of Colorado Anschutz Medical Campus. Saffron Scientific Histology Services performed the aldehyde-fuchsin staining on sections cut from the same tissue samples.

Co-transfer of Tregs and diabetic spleen cells

Single cell suspensions were prepared from the spleen and lymph nodes of 2.5HIP-PLG-treated NOD.scid mice euthanized at 8 weeks post adoptive transfer and pooled together. CD4⁺ CD25⁺ Tregs were isolated by magnetic enrichment using the EasySep Mouse CD4⁺ CD25⁺ Regulatory T Cell Isolation Kit II (Stemcell Technologies) according to the manufacturer’s guidelines. Isolated Tregs (0.6 × 10⁶ viable cells) were mixed with 10 × 10⁶ splenocytes from diabetic NOD mice and injected i.v. into individual NOD.scid recipients, which were then followed for development of diabetes as described above.

Ex vivo flow cytometry analysis

APC or PE-conjugated I-A^g7 tetramers loaded with 2.5HIP (LQTLALWSRMD) were obtained from the NIH tetramer core. Pancreas and spleen were harvested from NOD.scid recipient mice. Spleen samples were homogenized and pancreata were digested in 5 mg/ml collagenase from Clostridium histolyticum (Sigma) and 0.01 mg/ml DNase I (Roche) for 15 min at 37°C to yield single cell suspensions. Cells were stained with tetramer for 1 hr at 37°C and then counterstained with antibodies at room temperature. For intracellular staining, cells were fixed and permeabilized using the eBioscience Foxp3/transcription factor staining buffer set (Invitrogen). Fixable viability dye eFluor780 (eBioscience) was used to discriminate live cells. The dump gate anti-mouse antibodies used were anti-CD11b:BB700 (M1/70, BD), anti-CD11c:BB700 (HL3, BD), anti-GR1:BB700 (1A8, BD), anti-CD19:BB700 (1D3, BD), and anti-CD8:BB700 (53–6.7, BD). Other antibodies used included: anti-CD45:BUV395 (30-F11, BD), anti-CD4:BV711 (GK1.5, Biolegend), anti-Foxp3:PE and eFluor450 (FJK-16s, eBioscience), anti-CD25:BB515 (PC61, BD), anti-CTLA-4:PE-Cy7 (4C10–4B9, Biolegend), anti-GITR:BV510 (DTA-1, BD), anti-ICOS:BV605 (7E.17G9, BD), anti-CD127:PE-Cy7 (SB/199, eBioscience), anti-CD103:eFluor450 (2E7, eBioscience), anti-CD44:BV510 (IM7, BD), anti-CD73:BV605 (TY/11.8, Biolegend), anti-FR4:PE-Cy7 (12A5, eBioscience), anti-IFN-γ:APC (XMG1.2, BD), anti-TNF-α: FITC (MP6-XT22, eBioscience), and anti-T-bet: BV605 (4B10, Biolegend). Samples were run on a Fortessa X-20 (BD) flow cytometer. Data analysis was performed using FlowJo software V10 (Tree Star).

Stimulation of cells and intracellular cytokine staining

Single cell suspensions were prepared from the spleen and pancreas of NOD.scid mice as described above. CD4⁺ T cells were isolated by magnetic enrichment using the EasySep Mouse CD4⁺ T Cell Isolation Kit (Stemcell Technologies) according to the manufacturer’s guidelines. Spleens from NOD mice (non-diabetic) were digested in 200 Mandl units/ml Collagenase D (Roche) and 0.25 mg/ml DNase I (Roche) for 30 mins at 37°C to isolate dendritic cells. The NOD splenocytes were then depleted of CD4⁺ T cells and irradiated at 3,500 rads. CD4⁺ T cells from NOD.scid mice were co-cultured overnight with the irradiated splenocytes and 1 μg/ml of 2.5HIP. Golgi-Plug (BD) was then added at a final concentration of 1 μg/ml for 5 hours. Cells were then washed, surface-stained with tetramer and antibodies, fixed, permeabilized, stained with intracellular antibodies, and analyzed by flow cytometry as described above.

RNA-sequencing

NOD.scid mice received adoptive transfer of activated BDC-2.5TCR-Tg.NOD.Foxp3^DTR/EGP+ T cells followed by treatment with 2.5HIP-PLG NPs one day after T cell transfer. On day 6 post adoptive transfer, spleen and lymph nodes (inguinal, mesenteric, brachial, axillary, pancreatic) were collected from 4 recipient mice per group, made into cell suspensions, and pooled together. CD4 T cells were first enriched by magnetic cell separation using the EasySep Mouse CD4⁺ T cell Isolation Kit (Stemcell Technologies) and then stained with dump antibodies in BB700 (see above), anti-CD45:BV510 (30-F11, Biolegend), anti-CD4:BV711 (GK1.5, Biolegend), and SYTOX red viability dye (Thermo Fisher Scientific). Live CD4⁺ GFP⁻ (Foxp3⁻) T cells were sorted using an Aria Fusion Cell Sorter (BD). RNA was extracted using an RNeasy Plus Mini Kit (Qiagen). RNA quality was verified using a High Sensitivity ScreenTape Assay on the Tape Station 2200 (Agilent Technologies) and measured with a NanoDrop 1000 (Thermo Fisher Scientific). Library construction was performed using the Universal Plus mRNA Library Kit (NuGen Technologies) and sequencing was performed on the NovaSeq 6000 Instrument (Illumina) using paired-end sequencing (2×150bp) by the University of Colorado Cancer Center Genomics and Microarray Core. Illumina adaptors were trimmed from the fastq, raw sequence read files using BBDuk (BBMap - Bushnell B. - sourceforge.net/projects/bbmap/). Adaptors were removed from both paired-end reads from the 3’ end and all paired-reads <50bp after trimming were removed. Reads were aligned to the Ensembl annotation of the mouse transcriptome (GRCm38.p6, release 94 (26)) using the STAR RNA-seq aligner (v2.6.0a) (27). Transcript quantification was done using RSEM (v1.3.1) (28). Expected transcript counts were compiled and the edgeR R package (v3.24.2) (29, 30) was used for normalization across samples. Differential expression was calculated for the comparison of 2.5HIP-PLG-treated to untreated mice for genes with a mean CPM >1. The anergy pathway (31) gene set enrichment analysis was performed using the fgsea R package (v1.8.0) (32). A heatmap displaying Log2 fold-change was generated using the ComplexHeatmap R package (v1.20.0) (33). The sequencing data are available in GEO under data repository accession no. GSE126553 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE126553).

Statistics

Data was analyzed using GraphPad Prism Version 7.0a software. The Mantel-Cox test (log-rank test) and the Gehan-Breslow-Wilcoxon test were used to compare incidence of diabetes. Statistical significance was determined using two-tailed Student’s t tests and one-way ANOVA. Statistical significance was defined as: * = p < 0.05, ** = p < 0.01, *** = p < 0.001, and **** = p < 0.0001.

Results

PLG nanoparticles loaded with 2.5HIP prevent transfer of diabetes by BDC-2.5 T cells

In order to determine if the 2.5HIP antigen for BDC-2.5 T cells could be used to induce antigen-specific tolerance we used a well-described system, developed by Miller and colleagues, in which disease-relevant autoantigens are coupled to the surface of poly(lactide-co-glycolide) (PLG) nanoparticles (NPs) (21, 22). This is achieved by chemical-cross linking of peptide to carboxyl groups on the particle surface with ethylcarbodiimide (ECDI) and builds upon previous work using ECDI to couple antigens to apoptotic splenocytes as a tolerance induction strategy in NOD mice (7). Here we used an aggressive adoptive transfer system wherein activated NOD.BDC-2.5 T cell receptor (TCR)-transgenic T cells (10 × 10⁶) were transferred into NOD.scid mice. The T cells were stimulated in vitro with 2.5HIP prior to transfer as this pre-activation step results in the induction of hyperglycemia in 100% of control untreated (i.e., untreated with NPs) recipients within 10 days or less (Fig. 1A), which is not the case for unstimulated transgenic T cells (data not shown). We administered 1.75 mgs of 2.5HIP-coupled PLG NPs (2.5HIP-PLG) intravenously (i.v.) one day after T cell transfer and found that treatment with 2.5HIP-PLG NPs prevented diabetes for up to 2 months in around 80% of NOD.scid mice, whereas injection of NPs coupled with an irrelevant peptide (HEL_11–25-PLG) had no effect (Fig. 1A). Histology was performed on untreated and 2.5HIP-PLG-treated (tolerized) mice to assess the degree of insulitis and β-cell granularity. Tolerized mice had significantly less islet infiltration by leukocytes and there were many β-cells containing intact granules (Supplementary Fig. 1A–C).

Figure 1. 2.5HIP-PLG NPs prevent transfer of diabetes and infiltration of the pancreas by BDC-2.5 T cells.

A: NOD.scid mice were injected i.v. with 10 × 10⁶ 2.5HIP-activated BDC-2.5 TCR-Tg T cells. Mice were either untreated (control) or injected i.v. the next day with HEL_11–25-coupled NPs (HEL-PLG), 2.5HIP-coupled NPs (2.5HIP-PLG), or 2.5HIP alone without NPs (64 or 200 μg 2.5HIP). Data in each panel is representative of 2–10 independent experiments. **P < 0.01, ****P < 0.0001. B: Single-cell suspensions were prepared from the pancreas and spleen of control and 2.5HIP-PLG-treated mice euthanized at 2 weeks post adoptive transfer, followed by staining with I-A^g7/2.5HIP tetramer and antibodies. Gates were set on live, CD45⁺, lineage-negative (lin⁻), CD4⁺, 2.5HIP tet⁺ cells, as shown in Supplementary Fig. 2. Data are representative of 3 independent experiments (n = 6). **P < 0.01

We also tested whether high dose soluble peptide alone could induce protection, based on literature showing that 200 μg of p63 mimotope peptide given i.v. was somewhat effective for inducing tolerance in activated BDC-2.5 T cells (34). In contrast to treatment with 2.5HIP-PLG, injecting 200 μg of soluble 2.5HIP i.v. without NPs induced protection that lasted for around 3 weeks before failing (Fig. 1A). To compare the amount of 2.5HIP that was delivered on NPs versus soluble peptide given i.v., the peptide load of 2.5HIP on NPs was quantified using a CBQCA assay. We found that an average of 36.4 μg of 2.5HIP was coupled per mg of NPs, resulting in around 64 μg of total peptide administered per mouse on PLG NPs. Similar to what was observed using 200 μg of soluble peptide, administering 64 μg of 2.5HIP i.v. alone induced transient protection only (Fig. 1A), demonstrating that loading of 2.5HIP on a NP carrier is critical for inducing long-lasting tolerance intravenously.

To investigate how mice treated with 2.5HIP-PLG were protected from T cell-induced diabetes, we analyzed T cells ex vivo from the spleen and pancreas of control (diabetic) and tolerized mice euthanized at 2 weeks post adoptive transfer. Mice treated with 2.5HIP-PLG that were protected long-term were also euthanized at 8 weeks to investigate mechanisms of how tolerance is maintained. To detect transferred CD4⁺ BDC-2.5 T cells by flow cytometry, an I-A^g7 tetramer containing the 2.5HIP sequence was used (Supplementary Figure 2). Quantification of 2.5HIP tetramer⁺ (tet⁺) T cells at the 2-week time point revealed that they were significantly decreased in the pancreas of tolerized mice compared to controls, but accumulated in the spleen, indicating that T cell trafficking to the islets is impaired (Fig. 1B).

2.5HIP-PLG nanoparticles alter characteristics of BDC-2.5 regulatory T cells

It is critical to determine how treatment with 2.5HIP-PLG NPs modulates the phenotype of activated BDC-2.5 T cells and reduces their pathogenicity in order to understand the mechanisms of antigen-specific tolerance induction. To perform a detailed examination of the fate of both BDC-2.5 Foxp3⁺ regulatory T cells (Tregs) and effector T cells (Teffs), we compared these populations ex vivo. To analyze Tregs, we used either expression of the transcription factor Foxp3 alone or expression of the IL-2 receptor alpha subunit (CD25) together with Foxp3 since bonafide Tregs are known to express CD25 and depend on IL-2 for survival. As expected, only about half of Foxp3⁺ CD4 T cells expressed CD25 (Fig. 2A). We found that the percentage of Foxp3⁺ and CD25⁺ Foxp3⁺ Tregs was increased in the spleen and pancreas of tolerized mice analyzed at both 2 and 8 weeks in comparison to controls (Fig. 2B). Accordingly, the absolute number of Tregs in the spleen of tolerized mice was elevated compared to controls (Fig. 2C), indicating that Tregs accumulate in the periphery upon treatment. However, the number of Tregs in the pancreas was not elevated (Fig. 2C) due to the fact that the total number of BDC-2.5 T cells was reduced in the pancreas of tolerized mice (Fig. 1B).

Figure 2. BDC-2.5 Treg percentages, but not numbers, are elevated in the pancreas upon 2.5HIP-PLG treatment.

A–C: Single-cell suspensions were prepared from the pancreas and spleen of control and 2.5HIP-PLG treated mice euthanized at 2 or 8 weeks post adoptive transfer, followed by staining with I-A^g7/2.5HIP tetramer and antibodies. Gates were set on live, CD45⁺, lineage-negative (lin⁻), CD4⁺, 2.5HIP tet⁺ cells. A: Representative example of Foxp3 and CD25/Foxp3 staining from the pancreas of a 2.5HIP-PLG treated mouse euthanized at 8 weeks. B: Summary of 2–3 independent experiments for control 2 wks (n = 6), 2.5HIP-PLG 2 wks (n = 6), and 2.5HIP-PLG 8 wks (n = 4). Each symbol represents an individual mouse. *P < 0.05, **P < 0.01. C: Data are representative of 3 independent experiments (n = 6) from mice euthanized at 2 weeks. **P < 0.01, ***P < 0.001.

We next examined the expression of Treg activation markers on BDC-2.5 Tregs to determine if 2.5HIP-PLG NPs might alter their functional capacity. As anticipated, compared to bulk 2.5HIP tet⁺ T cells from tolerized mice, CD25⁺ Foxp3⁺ tet⁺ Tregs expressed significantly higher levels of CTLA-4, GITR, and ICOS (Fig. 3A). However, when the expression of CTLA-4, GITR, and ICOS on Tregs was compared between untreated and 2.5HIP-PLG-treated mice, no differences were observed (Fig 3B). In control and tolerized mice analyzed at the 2-week time point, the relative level of expression as measured by mean fluorescence intensity was also very similar (Fig 3C). It is likely that tolerance induction did not lead to changes in expression of these markers because Tregs already expressed high levels after in vitro stimulation prior to adoptive transfer (data not shown). Of note, almost all Tregs from tolerized mice analyzed at 8 weeks also expressed CTLA-4, GITR, and ICOS (Fig. 3B), suggesting that the cells remain activated for a long period of time. To determine whether Tregs were functional long-term after treatment, CD4⁺ CD25⁺ T cells were sorted from tolerized mice euthanized at the 8-week time point and co-transferred with polyclonal diabetogenic T cells into NOD.scid recipients. Whereas transfer of 10 × 10⁶ splenocytes isolated from diabetic NOD mice alone causes disease within around 6 weeks, significant protection was observed when Tregs from tolerized mice were co-transferred (Supplementary Fig. 3). Hence, tolerance induction with 2.5HIP-PLG NPs allows for the long-term survival and function of BDC-2.5 Tregs that are capable of suppressing Teff cells of multiple specificities.

Figure 3. BDC-2.5 CD25⁺ Foxp3⁺ Tregs from both control and 2.5HIP-PLG-treated mice express CTLA-4, GITR, and ICOS.

A–C: Single-cell suspensions were prepared from the pancreas and spleen of control and 2.5HIP-PLG-treated mice euthanized at 2 or 8 weeks post adoptive transfer, followed by staining with I-A^g7/2.5HIP tetramer and antibodies. Gates were set on live, CD45⁺, lineage-negative (lin⁻), CD4⁺, 2.5HIP tet⁺ cells, CD25⁺ Foxp3⁺ cells. A. Representative example of CTLA-4, GITR, and ICOS staining on bulk 2.5HIP tet⁺ or tet⁺ Treg cells from the pancreas of a 2.5HIP-PLG-treated mouse euthanized at 8 weeks. B: Summary of 2 independent experiments for control 2 wks (n = 4), 2.5HIP-PLG 2 wks (n = 3–4), and 2.5HIP-PLG 8 wks (n = 3–4). Each symbol represents an individual mouse. C: Geometric mean fluorescence intensity (gMFI) data from mice euthanized at 2 weeks from 1 experiment is shown (n = 2).

In contrast to what we observed for expression of CTLA-4, GITR, and ICOS, the markers CD127 and CD103 were differentially expressed on Tregs. CD127 is the IL-7 receptor alpha subunit and similar to humans, mouse Tregs are often characterized by low or absent expression of CD127 relative to effector T cells (35). We found that CD127 was down-regulated on Tregs from tolerized mice compared to controls (Fig. 4A & B). The expression of CD127 remained low on Tregs in the pancreas of tolerized mice for up to 8 weeks (Fig. 4B). Another marker that was found to be different was integrin alpha E, also known as CD103. Expression of CD103 has often been used as a marker of effector-memory Treg cells (36, 37). We found that CD103 is expressed on around 20–40% of Tregs from both control and tolerized mice analyzed at 2 weeks (Fig. 4D). In contrast, when evaluated at 8 weeks, almost 100% of the CD25⁺ Foxp3⁺ T cells expressed CD103 in the 2.5HIP-PLG-treated group; this was not the case with the bulk 2.5HIP tet⁺ T cell population (Fig. 4C & D).

Figure 4. BDC-2.5 CD25⁺ Foxp3⁺ Tregs from control and 2.5HIP-PLG-treated mice differ in expression of CD127 and CD103.

A–D: Single-cell suspensions were prepared from the pancreas and spleen of control and 2.5HIP-PLG-treated mice euthanized at 2 or 8 weeks post adoptive transfer, followed by staining with I-A^g7/2.5HIP tetramer and antibodies. Gates were set on live, CD45⁺, lineage-negative (lin⁻), CD4⁺, 2.5HIP tet⁺ cells, CD25⁺ Foxp3⁺ cells. A: Representative example of CD127 staining on tet⁺ Treg cells from the pancreas of a control (CTR) or 2.5HIP-PLG-treated (TOL) mouse euthanized at 2 weeks. C: Representative example of CD103 staining on bulk 2.5HIP tet⁺ or tet⁺ Treg cells from the pancreas of a 2.5HIP-PLG-treated mouse euthanized at 8 weeks. B&D: Summary of 2–3 independent experiments for control 2 wks (n = 4–6), 2.5HIP-PLG 2 wks (n = 4–6), and 2.5HIP-PLG 8 wks (n = 3–4). Each symbol represents an individual mouse. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Pro-inflammatory cytokine production by BDC-2.5 effector T cells is impaired upon induction of tolerance with 2.5HIP-PLG nanoparticles

One potential explanation for why percentages, but not absolute numbers, of Tregs increase in the pancreas of tolerized mice is that the frequency of Teffs is decreased. To determine whether treatment with 2.5HIP-PLG NPs diminishes the pathogenic BDC-2.5 Teff cell population, we tested the ability of T cells from control and tolerized mice to produce pro-inflammatory Th1 cytokines in response to re-stimulation with 2.5HIP antigen. The spleen and pancreas of tolerized mice examined at 2 weeks contained considerably fewer T cells capable of producing IFN-γ or both IFN-γ and TNF-α in comparison to control mice (Fig. 5A, C, D). T cells from the pancreas of tolerized mice analyzed at 8 weeks were also defective in cytokine production, indicating that the function of Teffs is impaired long-term (Fig 5C & D). The percentage of TNF-α⁺ IFN-γ⁻ T cells was also lower although not significantly (Fig. 5B). Our data indicate that tolerance induction with 2.5HIP-PLG NPs dramatically reduces the percentage of BDC-2.5 Teff cells in both the spleen and pancreas.

Figure 5. BDC-2.5 T cells from 2.5HIP-PLG-treated mice are impaired in their ability to produce pro-inflammatory cytokines.

A–D: Single-cell suspensions were prepared from the pancreas and spleen of control and 2.5HIP-PLG-treated mice euthanized at 2 or 8 weeks post adoptive transfer. CD4 T cells were stimulated with 2.5HIP peptide in the presence of antigen-presenting cells followed by I-A^g7/2.5HIP tetramer staining and intracellular cytokine staining. Gates were set on live, CD45⁺, lineage-negative (lin⁻), CD4⁺, 2.5HIP tet⁺ cells. A: Representative example of IFN-γ and TNF-α staining from the pancreas of a control and 2.5HIP-PLG-treated mouse euthanized at 2 weeks. B–D: Summary of 2–3 independent experiments for control 2 wks (n = 5), 2.5HIP-PLG 2 wks (n = 5–6), and 2.5HIP-PLG 8 wks (n = 5). Each symbol represents an individual mouse. *P < 0.05, **P < 0.01, ****P < 0.0001.

2.5HIP-PLG nanoparticles cause a shift in the ratio of BDC-2.5 regulatory T cells to effector T cells

Although expansion in the number of Tregs present in the pancreas of tolerized mice was not observed, we did detect a striking decrease in the frequency of Teff cells. It was therefore important to determine if the balance between Tregs and Teffs was altered. We found that the ratio of Foxp3⁺ Tregs to IFN-γ⁺ Teff cells was significantly increased in the spleen and pancreas of 2.5HIP-PLG-treated mice compared to untreated mice examined at 2 weeks (Fig. 6A & B). In the spleen of long-term tolerized mice there was also a significant increase overall. In the pancreas there was only a modestly increased ratio in three mice compared to controls, but in two mice there was a very high ratio of 2 or above, signifying a higher frequency of Tregs than Teffs (Fig. 6B). Since some mice treated with 2.5HIP-PLG eventually became diabetic (Fig. 1A), these results are likely a consequence of heterogeneity between mice. Of note, the tolerized mice with a low Treg/Teff ratio had nearly double the number of total pancreas-infiltrating 2.5HIP tet⁺ T cells compared to the mice with a high ratio (data not shown) revealing that a higher Treg/Teff ratio correlates with less overall T cell infiltration. We also observed that fewer T cells from the spleen of tolerized mice express the Th1 cell ‘master’ transcription factor T-bet upon re-stimulation (Fig. 7A & B), but there were no differences in cells from the pancreas (data not shown). The T-bet⁺ cells consisted of both IFN-γ⁺ Teffs and Foxp3⁺ Tregs in accordance with other reports (38, 39), suggesting that T-bet expression is important for the function of both pathogenic and suppressive T cells in NOD mice. In 2.5HIP-PLG-treated mice, the percentage of T-bet⁺ IFN-γ⁺ Teff cells was decreased while the percentage of T-bet⁺ Foxp3⁺ Tregs was increased (Fig. 7 C & D), demonstrating that the balance of these specialized cell types was altered upon tolerance induction.

Figure 6. 2.5HIP-PLG-treated mice have an increased BDC-2.5 Treg to effector T cell ratio.

A–B: Single-cell suspensions were prepared from the pancreas and spleen of control and 2.5HIP-PLG-treated mice euthanized at 2 or 8 weeks post adoptive transfer. CD4 T cells were stimulated with 2.5HIP peptide in the presence of antigen-presenting cells followed by I-A^g7/2.5HIP tetramer staining and intracellular cytokine staining. Gates were set on live, CD45⁺, lineage-negative (lin⁻), CD4⁺, 2.5HIP tet⁺ cells. A: Representative example of IFN-γ and Foxp3 staining from the pancreas of a control and 2.5HIP-PLG-treated mouse euthanized at 2 weeks. B: Summary of 2–3 independent experiments for control 2 wks (n = 5), 2.5HIP-PLG 2 wks (n = 5–6), and 2.5HIP-PLG 8 wks (n = 5). Each symbol represents an individual mouse. *P < 0.05, ***P < 0.001.

Figure 7. Fewer BDC-2.5 T cells from the spleen of 2.5HIP-PLG-treated mice express T-bet.

A–B: Single-cell suspensions were prepared from the spleen of control and 2.5HIP-PLG-treated mice euthanized at 2 weeks post adoptive transfer. CD4 T cells were stimulated with 2.5HIP peptide in the presence of antigen-presenting cells followed by I-A^g7/2.5HIP tetramer staining and intracellular cytokine staining. Gates were set on live, CD45⁺, lineage-negative (lin⁻), CD4⁺, 2.5HIP tet⁺ cells. A: Representative example of T-bet staining on unstimulated or peptide-stimulated T cells from a control and 2.5HIP-PLG-treated mouse. B: Summary of 2 independent experiments for T-bet, IFN-γ (gated on T-bet⁺), and Foxp3 (gated on T-bet⁺) staining of stimulated T cells from control (n = 3) and 2.5HIP-PLG-treated (n = 4) mice. Each symbol represents an individual mouse. *P < 0.05, **P < 0.01, ***P < 0.001.

2.5HIP-PLG nanoparticles induce an anergic phenotype in BDC-2.5 effector T cells

Anergy is a mechanism of peripheral tolerance in which T cells fail to produce pro-inflammatory cytokines and proliferate in response to antigen (40). Previous studies using antigen-loaded PLG NPs have suggested that induction of anergy in Teff cells is an important mechanism in promoting antigen-specific tolerance (41), although the exact contribution of anergy to induction and long-term maintenance of tolerance has not been clear due to a lack of surface markers to identify anergic CD4⁺ T cells. Using procedures described by Mueller and colleagues (42), we investigated whether anergic CD4⁺ T cells could be found in the CD44^hi Foxp3⁻ Teff compartment, by co-expression of high levels of the ecto-nucleotidase CD73 and folate receptor 4 (FR4). When tissues from mice euthanized at 2 weeks were evaluated ex vivo, the frequency of CD73^hi FR4^hi T cells was significantly higher in tolerized mice compared to controls (Fig. 8A & B). The CD73^hi FR4^hi T cell population was nearly absent after in vitro stimulation prior to adoptive transfer (data not shown), demonstrating that anergy markers were specifically up-regulated as a result of 2.5-PLG treatment. In tolerized mice analyzed at the 8-week time point, the average of CD73^hi FR4^hi T cells (~20%) in the pancreas was higher than at 2 weeks, implying the frequency of anergic T cells increases over time (Fig. 8B).

Figure 8. BDC-2.5 effector T cells from 2.5HIP-PLG-treated mice have an anergic phenotype.

A&B: Single-cell suspensions were prepared from the pancreas and spleen of control and 2.5HIP-PLG-treated mice euthanized at 2 or 8 weeks, followed by staining with I-A^g7/2.5HIP tetramer and antibodies. A: Representative example of CD73 and FR4 staining from the pancreas of a 2.5HIP-PLG-treated mouse euthanized at 8 weeks. Top panel: gates were first set on on live, CD45⁺, lineage-negative (lin⁻), CD4⁺, 2.5HIP tet⁺, CD44^hi Foxp3⁻ T cells. Bottom panel: gates were set on CD44^hi Foxp3⁻ T cells then on CD73^hi FR4^hi cells. B: Summary of 2–3 independent experiments for control 2 wks (n = 6), 2.5HIP-PLG 2 wks (n = 6), and 2.5HIP-PLG 8 wks (n = 4). Each symbol represents an individual mouse. *P < 0.05, **P < 0.01, ***P < 0.001. C&D: GSEA was performed on sorted CD4⁺ Foxp3⁻ T cells from 2.5HIP-PLG-treated versus untreated mice (n = 4) using a previously published anergy gene set (GSE2323 (ref. (31)). A heatmap is shown to display the genes of this set and the Log2 fold-change of 2.5HIP-PLG-treated versus untreated mice. RNA-seq data are available in Supplementary Table I. Data are from 1 experiment.

It is known that recognition of antigen by the TCR in the absence of co-stimulation leads to NFAT (but not AP-1) activation, promoting the expression of negative regulators of T cell activation (43). These counter-regulatory products include the transcription factors EGR2 and EGR3, as well as the E3 ubiquitin ligases CBL-B, ITCH and GRAIL, which are all critical for T cell anergy. To further support the surface marker expression data suggesting that 2.5HIP-PLG NPs induce an anergic phenotype, we investigated differences in gene expression in Teff cells from tolerized and control mice using RNA-sequencing. For the adoptive transfer into NOD.scid mice, T cells from BDC-2.5TCR-Tg.NOD.Foxp3^DTR/EGFP+ mice were used. These mice express a diphtheria toxin receptor (DTR) sequence fused to an enhanced green fluorescent protein (EGFP) from Foxp3 promoter/enhancer regions on a BAC transgene (44) allowing us to FACS-sort CD4⁺ Foxp3⁻ Teffs. Analysis of genes with a fold change equal to or greater than 3 showed that treatment with 2.5HIP-PLG NPs led to the over-expression of Egr2 and Egr3, Rnf128 (encodes GRAIL), the Egr3 target gene Spry1, and the cell surface anergy marker Izumo1r (encodes FR-4) (Supplementary Table I). Gene set enrichment analysis (GSEA) revealed that a molecular program of T cell anergy (31) was significantly up-regulated in Teffs from mice treated with 2.5HIP-PLG (Fig. 8C & D). Other genes that were found to be up-regulated were the inhibitory receptors Pdcd1 (encodes PD-1) and Lag3 (Supplementary Table I), implying a transcriptional network may be shared by the anergic T cells with exhausted T cells (45).

Discussion

Our discovery that the ligand for diabetogenic BDC-2.5 T cells is a hybrid insulin peptide (2.5HIP) (18) suggests that post-translationally modified proteins are important autoantigens in T1D. This is supported by findings from our lab and others (19, 20) showing that endogenous 2.5HIP-reactive CD4⁺ T cells are present at a higher percentage and have a more pathogenic phenotype in NOD mice than do insulin B:9–23-reactive T cells, which are thought be important contributors to autoimmune diabetes (46). Since hybrid insulin peptides (HIPs) are likely to be uniquely created in β-cells and not expressed in the thymus, this may provide an explanation of how diabetogenic T cell clones escape thymic deletion and raises the question of whether HIPs can be used to re-educate pathogenic T cells.

In this study we found that 2.5HIP loaded on PLG nanoparticles (NPs) can be successfully used to tolerize BDC-2.5 T cells and prevent transfer of diabetes into NOD.scid mice (Fig. 1A). Administration of 2.5HIP-coupled NPs (2.5HIP-PLG) was considerably more effective in the induction of long lasting tolerance than intravenous injection of soluble peptide alone (Fig. 1A). This is likely due to more efficient uptake and tolerogenic presentation of the peptide by antigen-presenting cells (APCs). The particles are highly carboxylated, yielding an anionic surface charge that has been shown to be critical for binding and tolerance via the MARCO scavenger receptor (21). By delivering 2.5HIP on the NPs, MARCO-expressing APCs in tolerogenic organs such as the spleen and liver are specifically targeted whereas administering peptide in the absence of a carrier probably leads to poor uptake and presentation, even if large amounts of antigen is used.

Consistent with a recent publication from the Miller group using a similar BDC-2.5 transfer system and p31-loaded NPs (23), we found that tolerance induction led to accumulation of T cells in the spleen and reduced trafficking to the islets (Fig. 1B). In the study by Prasad et al., treatment with p31-loaded NPs led to splenic T cell accumulation as early as 3 days post adoptive transfer, indicating that effects on migration begin rapidly after transfer and persist for at least 2 weeks. Since we found fewer T cells from the spleen of tolerized mice express T-bet (Fig. 7A & B), a potential mechanism that may prevent T cell trafficking to the islets is the down-regulation of T-bet upon encounter with antigen-containing MARCO⁺ APCs in the spleen. This in turn could lead to reduced expression of the Th1 cell-associated chemokine receptor CXCR3, which is regulated by T-bet (47). We also observed that the percentage of Foxp3⁺ Tregs was increased after tolerance induction with 2.5HIP-PLG (Fig. 2B). However, similar to what was found with absolute numbers of bulk BDC-2.5 T cells (Fig. 1B), tolerized mice had lower numbers of Tregs in the pancreas and higher numbers in the spleen (Fig. 2C). This indicates that, in the adoptive transfer system used here, Ag-PLG NPs do not actually expand the number of Tregs at the site of inflammation.

To determine whether treatment with 2.5HIP-PLG alters the function of BDC-2.5 Tregs, we investigated the phenotype of Tregs ex vivo at 2 and 8 weeks post tolerance induction. Two markers were differentially expressed on Tregs from control and experimental mice, one of which was the IL-7 receptor alpha subunit, also known as CD127. When Tregs from untreated mice were examined, 20–60% of the cells expressed CD127 (Fig. 4B). In contrast, surface expression of CD127 was nearly absent on Tregs from 2.5HIP-PLG-treated mice (Fig. 4A & B). BDC-2.5 Tregs did not express CD127 after in vitro stimulation (data not shown), indicating that this marker was up-regulated only when cells were transferred into mice and the recipients were left untreated. It is known that CD127 is down-regulated on all T cells after activation and remains low on Foxp3⁺ Tregs, whereas CD127 is re-expressed on the majority of effector and memory T cells (35). Perhaps in untreated diabetic mice in which there is a large degree of inflammation, Tregs become unstable and begin to take on properties of effector-memory T cells by expressing CD127 (48). In the future, it will be important to determine if stable expression of Foxp3 is achieved upon Ag-PLG treatment through epigenetic modification of the Treg-specific demethylated region (49). We also observed that almost all Tregs from long-term tolerized mice expressed the Treg effector-memory marker CD103 (Fig. 4C & D). CD103⁺ Tregs have been found to be more suppressive in vivo than CD103^- Tregs due to enhanced trafficking to inflammatory sites (50, 51). We propose that CD103⁺ cells are long-lived Tregs, which are responsible for maintenance of peripheral tolerance to islet autoantigens. Our finding that Tregs isolated from long-term tolerized mice are capable of suppressing a polyclonal T cell population confirms that they remain functional for an extended period of time (Supplementary Fig. 3).

When the effect of 2.5HIP-PLG NPs on pathogenic BDC-2.5 T cells was examined, we found that Teff cells from tolerized mice were significantly compromised in their ability to produce pro-inflammatory cytokines upon re-challenge with peptide (Fig. 5A–C). Additionally, Treg percentages were increased and Teff cell percentages were decreased in tolerized mice, which was reflected by an increase in the ratio of Tregs to Teffs (Fig. 6A & B). However, at the 8-week time point some tolerized mice had a high Treg/Teff ratio while in others it was low (Fig. 6B). We expect that mice with a low ratio in the pancreas would eventually become diabetic, whereas mice with a high ratio would be protected indefinitely. Our findings imply that an increase in the ratio of antigen-specific Foxp3⁺ Tregs to IFN-γ⁺ Teff cells is an indicator of whether tolerance induction occurs. Analysis of the cell-surface anergy markers CD73 and FR4 revealed that significantly more Teff cells from tolerized mice were CD73^hi FR4^hi, indicative of an anergic state (Fig. 8A & B). Upon treatment with 2.5HIP-PLG NPs, the frequency of CD73^hi FR4^hi T cells increases over time, suggesting that anergy may be most important for long-term maintenance of tolerance. Transcriptome analysis revealed that a molecular signature of T cell anergy is in fact enriched in Teff cells from tolerized mice (Fig. 8C & D). Although the mechanisms responsible for induction of T cell anergy in vivo remain unclear and will require further investigation, previous studies have demonstrated that APCs cultured with PLG NPs in vitro exhibit a down-regulation of the co-stimulatory molecules CD80 and CD86, reduced production of IL-12, and high expression of the co-inhibitory molecule PD-L1 (52, 53). Thus, it is likely that uptake of Ag-PLG NPs by APCs through the scavenger receptor MARCO leads to tolerogenic reprogramming and promotes antigen-presentation in the absence of co-stimulation and cytokines responsible for Th1 cell differentiation.

This study establishes that carboxylated PLG NPs loaded with a HIP are effective for inducing tolerance in a rapid transfer model of T1D. However, since human T1D is a complex heterogenous autoimmune disease involving reactivity to several autoantigens, the critical test will be to determine whether our findings translate to the NOD spontaneous model of disease. One of the major issues with translation of antigen-specific therapies for T1D to the clinic is that it is still not completely understood which human CD4 T cell epitopes are important for triggering autoimmunity. Although the initial discovery of HIPs was made in NOD mice, HIP-reactive CD4 T cells have also been isolated from the islets of deceased T1D donors (18, 54). Furthermore, we have since found that HIP-reactive CD4 T cells with a pathogenic phenotype can be isolated from the peripheral blood of T1D patients (Baker et al., unpublished), demonstrating the relevance of these neoepitopes to human T1D. While it is not yet clear whether there are dominant HIPs in T1D patients, such as 2.5HIP in NOD mice, it is encouraging that novel autoantigens other than insulin are being discovered and may potentially be used to induce tolerance in humans. The use of PLG nanoparticles, loaded with HIP antigens, is a promising therapeutic approach that addresses the underlying mechanisms of autoimmune pathology and may have the ability to prevent or reverse T1D.

Key Points.

• A hybrid insulin peptide (HIP) neoepitope can tolerize diabetogenic T cells.

• Tolerance induction promotes T cell anergy and increases Treg/Teff ratio.

• HIPs are potential therapeutics in Type 1 Diabetes.