HLA-A2–Matched Peripheral Blood Mononuclear Cells From Type 1 Diabetic Patients, but Not Nondiabetic Donors, Transfer Insulitis to NOD-scid/γc^null/HLA-A2 Transgenic Mice Concurrent With the Expansion of Islet-Specific CD8⁺ T cells

Mice were housed in facilities approved by the Association for Assessment and Accreditation of Laboratory Animal Care at the University of North Carolina (UNC) Animal Facility and handled according to the UNC Office of Animal Care and Use. All experimentation was approved by the UNC Institutional Animal Care and Use Committee. The original breeding pairs of NOD.Cg-Prkdc^scid Il2rg^tm1Wjl Tg(HLA-A2.1)Enge/Sz and NOD.Cg-Prkdc^scidEmv30^b-Tg(HLA-A/H2D/B2M)1Dvs/DvsJ mice were previously described (23,24). Animals were housed in autoclaved microisolator cages and fed autoclaved food and chlorinated water. Mice were used at 8–14 weeks old and defined as diabetic if two consecutive weekly blood glucose measurements were 250 mg/dL.

All protocols involving the collection and use of PBMCs were approved by the UNC Office of Human Research Ethics and conducted according to Institutional Review Board Approval of Human Subjects Research.

Peripheral blood from patients diagnosed with type 1 diabetes or nondiabetic donors (NDD) was collected under sterile conditions in heparinized tubes. PBMCs were isolated on Ficoll-Hypaque. After isolation and purification, cells were tested for expression of HLA-A2 by flow cytometry using fluorescein isothiocyanate–labeled BB7.2 monoclonal antibody (Abcam, Cambridge, MA). HLA-A2⁺ PBMCs from type 1 diabetic patients or NDDs were adoptively transferred into mice.

NOD-scid/γc^null/A2 mice were intraperitoneally injected with 17–52 × 10⁶ HLA-A2⁺ PBMCs in 400 μL of PBS from type 1 diabetic patients or NDD. Spleen, PLN, blood, and islets were removed from recipient mice 2 to 6 weeks after transfer. Islets were isolated as previously described (25). Briefly, pancreata were perfused with a 2 mg/mL solution of collagenase P (Sigma-Aldrich, St. Louis, MO), dissected, and incubated at 37°C for 20 min. Islets were purified using a Ficoll PM 400 (Sigma-Aldrich) gradient, handpicked, counted, and then cultured overnight in RPMI-1640 containing 10% FBS and 4 ng/mL recombinant human interleukin (IL)-2 (Peprotech, Rocky Hill, NJ). Infiltrating lymphocytes were collected and filtered through a 40-µm nylon filter. Each sample constituted all of the islet cells isolated from an individual mouse. Human cell populations were obtained by gating on human CD45⁺/mouse CD45⁻. Cell suspensions were stained using fluorescently-labeled mouse antibodies to human surface markers: CD45 (PerCp), CD19 (PeCy7), CD8 (PB), CD3 (antigen-presenting cell [APC]), and CD4 (APC-Cy7) (ebioscience, San Diego, CA), or CD3 (PO), CD3 (PTR), and CD4 (PO) (Invitrogen, Carlsbad, CA). Samples were analyzed on a Beckman-Coulter (Dako, Glostrup, Denmark) CyAn ADP using Summit V4.3.01 software. Mice that had >2 × 10⁶ human CD45⁺ cells in their spleen were included in our analysis.

Peptides IGRP_265–273 (VLFGLGFAI), IA-2_797–805 (MVWESGCTV), IAPP_5–13 (KLQVFLIVL), and insulin_B10–18 (HLVEALYLV) were purchased from GenScript (Piscataway, NJ) at >90% purity, dissolved in DMSO, and stored at −20°C. A2/IGRP, A2/IA-2, A2/IAPP, and A2/insulin tetramers were assembled with these peptides as described (26). Briefly, HLA-A2 and human β₂-microglobulin were produced in Escherichia coli and refolded with peptide in vitro. Refolded peptide major histocompatibility complex (MHC) monomer was purified by high-performance liquid chromatography, biotinylated using Biotin Protein Ligase, and assembled into tetramers by conjugation with UltraAvidin-PE (Leinco, St. Louis, MO).

Cells (2–3 × 10⁶) were incubated overnight in complete RPMI-1640 containing 10% FBS and 4 ng/mL recombinant human IL-2 (Peprotech). After 24 h, the cells were incubated with IGRP, IA-2, IAPP, and insulin peptide pool or individually at 40 µg/mL of each peptide. A2/IGRP, A2/IA-2, A2/IAPP, and A2/insulin PE-conjugated tetramers (2.5 μL each) along with anti-CD28 (1 μg/mL) was added. Cells were incubated at 37°C, 5% CO₂ for 2 h. Brefeldin A was added to inhibit protein secretion, followed by an additional 4-h incubation, placed on ice, stained with fluorescently labeled surface antibodies, fixed, permeabilized, and stained with human anti-interferon (IFN)-γ monoclonal antibody (ebioscience).

Previous studies using human HSCs have shown NOD-scid/γc^null mice support growth, survival, and differentiation of CD34⁺ HSC (17). HLA-A2*0201 is the most commonly expressed HLA class I allele in Europeans (45%) and is permissive for the development of type 1 diabetes in humans and transgenic mice. Therefore, we evaluated the ability of HLA-A2–matched PBMCs from NDD and type 1 diabetic patients to successfully engraft in NOD-scid/γc^null/A2 mice. Flow cytometry analyses confirmed the presence of human CD45⁺ cells in the spleen of mice receiving PBMCs from NDD or type 1 diabetic patients (Fig. 1A–C). Human CD45⁺ cells were also found in the blood and PLN of NOD-scid/γc^null/A2 mice (Fig. 1D and F), albeit at lower levels. The number of human CD45⁺ cells in both groups of mice that received NDD or type 1 diabetic PBMCs appear to be similar in the spleen (P = 0.206, Fig. 1G) and PLN (P = 0.3177, Fig. 1H). Therefore, we determined that both NDD and type 1 diabetic patient PBMCs survive and home to the spleen (primary site of engraftment) and PLN of NOD-scid/γc^null/A2 mice after transfer. Our data are consistent with previously reported findings that the primary site of engraftment by human cells in NOD-scid/γc^null mice is the spleen (27). Interestingly, the level of engraftment appears to increase markedly in mice receiving >20 × 10⁶ PBMCs from patients with type 1 diabetes. Successful engraftment was defined as recovering ≥2 × 10⁶ human CD45⁺ cells from the spleen of mice that received PBMCs. In our studies, transfer of ∼30 × 10⁶ PBMCs allows for successful engraftment. This demonstrates the ability of NOD-scid/A2/γc^null mice to support survival of PBMCs from patients with type 1 diabetes after transfer.

To evaluate proliferation potential of type 1 diabetic PBMCs in NOD-scid/γ^null/A2 mice, we labeled PBMCs with carboxyfluorescein succinimidyl ester (CFSE) and confirmed the presence of human CD45⁺ cells in the peripheral blood, spleen, PLN, and islets of mice after transfer. Human CD45⁺ cells reconstituted in all compartments tested. Most of the human CD45⁺ cells in the peripheral blood, spleen, PLN, and islets of NOD-scid/γc^null/A2 mice were CFSE low, implying that many of cells underwent at least five rounds of cell division (Fig. 2A–L).

Human cells were observed in the spleen and PLN after adoptive transfer in NOD-scid/γc^null/A2 mice. CD3⁺ T cells and CD19⁺ B cells were observed in the spleen and PLN (Fig. 3). In fact, the predominant human lymphocyte subset in the spleen and PLN of mice was CD3⁺ T cells. CD3⁺ T cells in the spleen of mice receiving NDD or type 1 diabetic PBMCs were 97 ± 1.8% and 82 ± 17.4%, respectively, of the human CD45⁺ cell population (Supplementary Fig. 1A). The level of CD3⁺ T cells in the spleen and PLN was similar in mice receiving cells from NDDs or type 1 diabetic patients (Fig. 3A and B). The number of CD19⁺ B cells in the spleen and PLN was also similar in both groups of mice.

We compared the level of CD4⁺ and CD8⁺ T cells in the spleen and PLN after transfer. Cells were gated from hCD45⁺/CD3⁺ populations. The number (Fig. 3C and D) and percentage (Supplementary Fig. 1C and D) of CD4⁺ and CD8⁺ T cells in the spleen and PLN was similar between both groups. In addition, both groups of mice had similar CD4/CD8 ratios in the spleen (Supplementary Fig. 1E) and PLN (Supplementary Fig. 1F) after transfer.

We determined the level of islet infiltration in NOD-scid/γc^null/A2 mice receiving PBMCs from NDD or type 1 diabetic patients. We isolated islets from individual mice after engraftment and determined the presence of cellular infiltrates. The total number of human CD45⁺ cells was significantly higher in mice that received type 1 diabetic PBMCs (P = 0.0006, Fig. 4A). Among these infiltrates, although not widely appreciated, B cells are found in the islets of both diabetic mice (25) and humans (28). The level of CD3⁺ T cells was significantly higher in mice that received type 1 diabetic PBMCs (P = 0.0238, Fig. 4B), whereas the level of CD19⁺ B cells was similar (P = 0.1310, Fig. 4B). The percentage of CD3⁺ T cells and CD19⁺ B cells was similar between the two groups (Supplementary Fig. 1G).

On a per-islet basis, the number of human CD45⁺ cells (P = 0.0006, Fig. 4C) and CD3⁺ T cells (P = 0.0175, Fig. 4D) was also increased. The level of CD19⁺ B cells (P = 0.0256, Fig. 4E) was also significantly increased. This suggests that although the percentages of infiltrating lymphocytes are similar, the actual numbers of T and B cells are increased in mice that received PBMCs from type 1 diabetic patients.

We determined whether CD4⁺ and CD8⁺ T cells were present in the islets of mice after transfer because both are required for the development of type 1 diabetes in mice and the issue cannot be tested in humans. The number of CD4⁺ T cells per islet was significantly higher in mice that received type 1 diabetic PBMCs compared with NDD PBMCs (P = 0.0175, Fig. 4F). Although the level was not statistically significant, CD8⁺ T cells were also increased in mice that received type 1 diabetic PBMCs (P = 0.1820, Fig. 4G). This shows that a higher number of CD4⁺ and CD8⁺ T cells from the PBMCs of type 1 diabetic patients infiltrate the islets of NOD-scid/γc^null/A2 mice compared with NDDs. Interestingly, the CD4/CD8 ratio was higher in mice that received PBMCs from type 1 diabetic patients (Supplementary Fig. 1I).

The significant advantage of this transfer model is the potential to identify antigen-specific CD8⁺ T cells, which are important for the development of diabetes in patients with type 1 diabetes. To understand the specificity of T cells in the islets of NOD-scid/γc^null/A2 mice transferred with PBMCs from type 1 diabetic patients, we used IGRP, IAPP, insulin, and IA-2/A2 tetramers to identify epitope-specific T cells. Lymphocytes from the spleen and islets of NOD-scid/γc^null/A2 mice were incubated with pooled HLA-A2 tetramers individually loaded with IGRP, IAPP, insulin, and IA-2 peptides after transfer to maximize our sensitivity of detection. Figure 5A shows the number of tetramer⁺ CD8⁺ T cells in the islets and spleen. In fact, the level of CD8⁺/tetramer⁺ T cells was significantly higher in the islets (17–80%) compared with the spleen (5–24%) of mice engrafted with type 1 diabetic PBMCs. These results demonstrate that A2-epitope–specific CD8⁺ T cells primarily infiltrate the islets of mice that received type 1 diabetic PBMCs. Unfortunately, the sample size was not sufficient to examine the cells recovered from the islets for individual A2-tetramer specificity.

To determine if epitope-specific human CD8⁺ T cells from the spleen and islets of mice after transfer are functionally active, cells were incubated with pooled A2 tetramers and their respective peptides and assessed for IFN-γ production after short-term stimulation. Figure 5A shows that CD8⁺/tetramer⁺ T cells are present in the spleen (12%) and islets (15%). In addition, these epitope-specific cells from the spleen and islets produce IFN-γ after peptide stimulation (Fig. 5B). These results demonstrate that the islets of NOD-scid/γc^null/A2 mice contained functional, human, epitope-specific CD8⁺ T cells that are capable of producing IFN-γ after restimulation. Although overt diabetes did not develop in any the mice transferred with PBMCs from patients with type 1 diabetes, we did observe by histology islet infiltration (Supplementary Fig. 2A) and elevated blood glucose levels up to 235 mg/dL (Supplementary Fig. 2D). Mice that received NDD or type 1 diabetic PBMCs displayed no clinical signs of graft-versus-host disease, as determined by histologic analysis of the liver and kidney (data not shown).

To determine whether epitope-specific T cells can continue to expand from one mouse to the next, we isolated splenocytes from a NOD-scid/γ^null/A2 mouse that received 20 × 10⁶ type 1 diabetic PBMCs (mouse 1) and transferred 18 × 10⁶ splenocytes intraperitoneally into a second NOD-scid/γ^null/A2 mouse (mouse 2). We compared the level of CD8⁺/tetramer⁺ T cells isolated from a patient with type 1 diabetes (A21) to 1) observe the level of epitope-specific CD8⁺ T cells after each transfer and 2) determine if they expanded. The number and percentage of CD8⁺/tetramer⁺ T cells increased from 3 × 10³ to 7 × 10⁵ cells and in percentage from 0.06 to 6.85% (Fig. 5C and D), in the spleen of the mouse transferred with type 1 diabetic PBMCs. In fact, these antigen-specific CD8⁺ T cells were not only in the spleen but also in the islets of mouse 2 (Fig. 5D), demonstrating the selective expansion of epitope-specific cells.

NOD-scid/γc^null/A2 mice have been widely used to study human immunology because these mice serve as a host for human immune cells (13). In addition, immune-deficient mice are being used to study various human diseases, including graft-versus-host disease after the injection of MHC-mismatched human PBMCs (29) and also islet transplantation rejection by mismatching MHC class I molecules of donor PBMCs to recipient islet MHC class I molecules (12,30).

Our goal is to study human autoimmunity, specifically type 1 diabetes, using a humanized mouse model by transferring HLA-A2 matched PBMCs from type 1 diabetic patients and NDDs into NOD-scid/γc^null/A2 mice. We observed proliferation and survival of T cells (CD3⁺, CD4⁺, and CD8⁺) and B cells for at least 6 weeks in these mice. Similar levels of T cells in the spleen and PLN were found in mice engrafted with PBMCs from type 1 diabetic patients and NDDs; however, more B cells were found in the spleen and PLN in mice that received PBMCs from type 1 diabetic patients. This demonstrates the ability of human immune cell subsets to engraft secondary lymphoid organs after transfer. Moreover, the higher levels of B cells in the PLN and spleen may reflect what occurs in the PLN and spleen of type 1 diabetic patients. NOD-scid mice that express a chimeric A2/K^b gene and were wild-type at the IL2r-γ chain locus were poorly engrafted (data not shown).

Through this model we have revealed an increased level of islet infiltration in mice that received type 1 diabetic PBMCs compared with NDD PBMCs. Previous studies using NOD-scid mice and HLA-A1⁺ PBMCs from type 1 diabetic patients have demonstrated the presence of islet-reactive T cells in pancreatic tissue (31), but these T-cell clones were not capable of intraislet infiltration. We have shown here that T and B cells preferentially infiltrate the islets of NOD-scid/γc^null/A2 mice that received type 1 diabetic PBMCs and that a significant number of CD8⁺ T cells from the islets and spleen are specific for IGRP, IA-2, IAP, and insulin (known diabetogenic epitopes). These cells were found at a higher frequency in the islets compared with the spleen, indicating the islets are a preferred site of autoreactive T-cell expansion and accumulation. Therefore, the extravasation and accumulation of human autoreactive islet-specific T cells from circulation into the islets suggest that PBMCs from type 1 diabetic patients are capable of infiltrating the target organs necessary for the induction of diabetes.

Circulating cytotoxic T lymphocytes isolated from HLA-A2⁺ type 1 diabetic patients have been shown to kill β-cells through recognition of a glucose-regulated preproinsulin epitope (18). In our model, higher frequencies of diabetogenic epitope-specific CD8⁺ T cells from type 1 diabetic patients infiltrate the islets of NOD-scid/γc^null/A2 mice. Furthermore, these antigen specific cells are capable of producing IFN-γ after peptide stimulation.

Until now, no reported studies have demonstrated islet infiltration of NOD-scid/γc^null/A2 mice by antigen-specific T cells using HLA-A2–matched PBMCs from type 1 diabetic patients. This may be the result of poor engraftment in previous models that is often associated with natural killer cell activity or the lack of human costimulatory molecules such as human MHC molecules (32,33). We identified an engraftment threshold and determined that 20–30 × 10⁶ PBMCs is ideal for maximum engraftment in NOD-scid/γc^null/A2 mice. This signifies a size limitation for transferring human cells into mice, which is similar to previous findings using NOD-scid/γc^null mice (34).

Although many genetic loci contribute to the development of diabetes, initial genome scans attribute particular MHC class I and II alleles as major contributors to the development of autoreactive T-cell responses in both humans and NOD mice (35,36). The process of developing diabetes is accelerated in NOD HLA-A2.1 MHC class I transgenic mice and is mediated by pathogenic A2-restricted T-cell responses (37). In humans, two islet-associated epitopes, IA-2 and GAD65, have been identified as being recognized by HLA-A2–restricted CD8⁺ T cells, thus indicating that CD8⁺ T cells that are HLA-A2 restricted may contribute to β-cell death. Although it is not clear whether CD8⁺ T cells are involved during the early phases of type 1 diabetes, we do know that they are involved in the development of diabetes because diabetes does not develop in NOD mice lacking MHC class I. The importance of IGRP, IAPP, IA-2, and insulin-specific CD8⁺ T cells is still unknown in humans. However, GAD65 and IA-2 are major islet antigens targeted in human type 1 diabetes. The presence of an enriched population of epitope-specific CD8⁺ T cells in the islets after transfer may indicate the expansion of epitope-specific immunodominant T cells from patients with type 1 diabetes in these humanized mice.

Overt diabetes has not yet developed in any of the NOD-scid/γc^null/A2 mice that received PBMCs from HLA-A2⁺ type 1 diabetic patients, although one of the nine recipients of type 1 diabetes HLA-A2⁺ PBMCs did show elevated blood glucose levels up to 235 mg/dL. It is likely that in addition to the HLA-A2 molecule, this mouse model may be improved by the addition of human MHC class II molecules or other costimulatory molecules that would allow interaction of the donor cells with mouse cells in the spleen and lymph node and perhaps provide a better environment for engraftment in these tissues. In the current model, CD4⁺ T cells can interact with engrafted donor APCs via human class II only, whereas the CD8⁺ T cells can receive stimulation from both donor and recipient (NOD-scid/γc^null/A2 mouse) cells.

It is difficult to isolate large numbers of diabetogenic CD8⁺ T cells directly from the peripheral blood of patients with type 1 diabetes. Therefore, one of the advantages of this model is the 1,000-fold increase in epitope-specific CD8⁺ T cells observed in mice engrafted with type 1 diabetic PBMCs. Presumably, this will allow more studies of human diabetogenic T cells; for example, we can test the cross-reactivity of diabetogenic T cells to common pathogens. This model also allows us to identify, directly from type 1 diabetic patients, islet epitope-specific CD8⁺ T cells. The four epitopes tested here comprised 63–88% of the islet-infiltrating CD8⁺ T cells in three of six mice that received type 1 diabetic PBMCs. In the islets of the remaining mice engrafted with type 1 diabetic PBMCs, these epitopes were recognized by 19–27% of CD8⁺ T cell infiltrates. In the future, we plan to use larger pools of peptides derived from islet antigens, and this may lead to the identification of other islet-infiltrating T cells.

This model of autoimmunity may also be useful in predicting islet cell transplantation by analyzing the level of pre-existing diabetogenic T cells in candidate patients with type 1 diabetes. Therefore, with future developments of this model, researchers may investigate more directly the mechanisms underlying T cell responses. We can gain valuable insight into the pathogenesis of various autoimmune diseases and predict immunotherapeutic responses.

This work was partially supported by NC TraCS grant 50KR10963, National Institutes of Health Grant 5-R01-AI-52435-07, Juvenile Diabetes Research Foundation grants 1-2008-24 and 7-2008-848, and the Beta Cell Biology Consortium. F.W.-L. was supported by a mentor-based minority postdoctoral fellowship from the American Diabetes Association grant 7-09-MI-03.

No potential conflicts of interest relevant to this article were reported.

F.W.-L. and E.F.Y. researched data, contributed to discussion, wrote the manuscript, and reviewed and edited the manuscript. G.T. researched data. E.F. provided patient samples and reviewed the manuscript. A.A. and S.P. provided patient samples, contributed to discussion, and reviewed and edited the manuscript. J.L., W.B., J.B., and A.S.C. provided patient samples and reviewed the manuscript. L.D.S. contributed to discussion and reviewed and edited the manuscript. J.A.F. contributed to discussion, wrote the manuscript, and reviewed and edited the manuscript.

The authors thank Eric Scott, Lucinda Hensley, and Shaun Steele of the University of North Carolina at Chapel Hill, Department of Microbiology and Immunology, for providing technical assistance, and Dr. Charles Jennette of the University of North Carolina at Chapel Hill, Department of Medicine, for providing renal histologic interpretations.