T-Cell Receptor/HLA Humanized Mice Reveal Reduced Tolerance and Increased Immunogenicity of Posttranslationally Modified GAD65 Epitope

Despite decades of effort, the mechanisms leading to loss of tolerance and subsequent β-cell destruction in type 1 diabetes have not yet been fully resolved. Accumulating evidence suggests that neoantigens, including posttranslationally modified (PTM) islet autoantigens, are key targets of anti–β-cell autoimmunity (1). Several PTM β-cell antigens have been tied to β-cell autoimmunity, including hybrid insulin peptides (2), defective ribosomal insulin gene products (3), and citrullinated and deamidated peptides (4,5). Because the process of PTM β-cell antigen generation is presumed to be limited to β-cells and, in many cases, linked to β-cell stress (6–8), the hypothesis is that PTM antigen presentation is primarily restricted to peripheral tissues and is not transported to the thymus in a sufficient amount to affect thymic selection (9). Therefore, PTM antigen-specific T cells could avoid thymic-negative and agonist selection, resulting in increased effector responses and reduced Foxp3⁺ regulatory T cells (Tregs).

PTM autoantigens are highly heterogeneous, as are the potential mechanisms of how they contribute to type 1 diabetes. Some modifications lead to drastic changes in peptide sequences (i.e., hybrid peptides), whereas others, like citrullination and deamidation, lead to discrete single-residue modifications within the target peptide epitope. Nevertheless, deamidation and citrullination can alter the epitope-MHC binding affinity and provide a favorable charge for T-cell receptor (TCR) interaction, which is postulated to increase immunogenicity and diabetogenicity of the modified peptides (10–12). Importantly, T cells reactive to deamidated and citrullinated β-cell epitopes are increased in patients with type 1 diabetes, implicating them in disease pathogenesis (4). It is unknown, however, whether partial recognition of the native epitope sequence is sufficient to induce thymic-negative selection or Treg development of thymocytes that preferentially recognize the PTM version of the epitope. It is also unclear to what extent T-cell responses to the native versus PTM epitope exhibit cross-reactivity and whether cross-reactivity could support their activation in the periphery.

The NOD mouse is a highly used, tractable, and historically important model for type 1 diabetes. However, the NOD model has its limitations, including expressing a single MHC class II molecule (H2-IA^g7), which cannot present the whole spectrum of antigenic targets in type 1 diabetes (13). To counter this problem, a number of HLA-humanized strains have been developed directly on the NOD background (14). To further improve the versatility of HLA-humanized mice, we have established a human TCR retrogenic approach (Hu-Rg) for in vivo expression of human TCRs (15). In this study, we used the Hu-Rg method to express patient-derived TCRs specific for the native or deamidated GAD65_115–127 epitope in NOD.HLA-DR4^Tg.H2Ab1^−/−.Rag1^−/− (DR4Tg.RagKO) mice. We compared polyclonal tetramer⁺ responses, monoclonal thymic development, and islet infiltration of native and deamidated GAD65_115–127-reactive T cells. Our results show increased immunogenicity of the deamidated epitope and reduced thymic Treg (tTreg) generation in PTM-specific Hu-Rg mice. Interestingly, although we observed T-cell cross-reactivity between the native and PTM epitopes, the level of cross-reactivity was not sufficient to elicit tTreg development in PTM-specific T cells but was sufficient to trigger islet infiltration. Overall, the results of our study outline differences in T-cell responses to the native and deaminated GAD65 epitope that provide mechanistic insight into the role of PTM autoantigen-reactive T cells in type 1 diabetes.

NOD.HLA-DR4^Tg.H2Ab1^−/−.Rag1^−/− mice were generated by crossing NOD.Cg-Rag1^tm1Mom IL2rg^tm1Wjl Tg(HLA-DRA,HLA-DRB1*0401)39–2Kito/ScasJ with NOD.DQ8, H2-Ab0^null,Rag1^null mice. Both strains were purchased from The Jackson Laboratory. NOD.H2Ab1^−/−.DR4^Tg mice were generated by crossing NOD.HLA-DR4^Tg.H2Ab1^−/−.Rag1^−/− mice with NOD mice. All mice were housed under specific-pathogen-free conditions in Baylor College of Medicine and University of Utah facilities. All experiments were performed in accordance with institutional animal care and use committee protocols at Baylor College of Medicine and University of Utah.

Flow cytometry analyses were performed on an LSRFortessa II (BD Biosciences) or Aurora (Cytek) cell analyzer. Data were analyzed with FlowJo software (Tree Star). Antibodies and dyes used in flow cytometry are listed in Supplementary Table 1. Tetramers presenting native and deamidated GAD65_115–127 were generated with the help of the National Institutes of Health Tetramer Core Facility.

TCRs were isolated from GAD65 tetramer⁺ CD4⁺ T-cell lines sorted and expanded from peripheral blood mononuclear cells of patients with type 1 diabetes (15). Chimeric TCR-expressing retroviral vectors were generated using human variable regions of TCRα and TCRβ and mouse TCR constant regions, as previously described (15). The wild-type (WT) GAD65_115-127 reactive T1D4 (WT-TCR) was previously described (15); 120E-TCR was isolated from the deamidated GAD65_115-127 tetramer⁺ cell line.

Hu-Rg mice were generated as previously described (16). Briefly, bone marrow (BM) hematopoietic precursors were harvested from NOD.HLA-DR4^Tg.H2Ab1^−/−.Rag1^−/− mice and transduced retrovirally with human TCR expression vector harboring an ametrine fluorescent reporter by spinoculation. Transduced BM cells were transferred intravenously into sublethally irradiated (500 rad) NOD.HLA-DR4^Tg.H2Ab1^−/−.Rag1^−/− recipients.

Dendritic cells (DCs) were generated using the method described by Mach et al. (17). Splenocytes (40–50% HLA-DR⁺) harvested from NOD.HLA-DR4^Tg.H2Ab1^−/−.Rag1^−/− mice harboring Flt3 ligand (Flt3l)-secreting B16 tumors were incubated with 25 μmol/L peptide at 37°C in an incubator for 1 h and subsequently transferred (intravenously) into Hu-Rg recipients.

TCR functional avidities were measured using TCR-transduced 4G4.CD4⁺ thymoma (18). TCR.4G4.CD4⁺ cells were transduced with TCR expression vectors by two spinoculations (1,800 rpm, 20°C, 90 min) and purified by CD3ε⁺ MACS enrichment. A total of 50,000 TCR⁺ 4G4 cells were stimulated with 100,000 NOD.HLA-DR4^Tg.H2Ab1^−/−.Rag1^−/− Flt3l-induced DCs. Twenty-four hours after stimulation, interleukin-2 (IL-2) secretion was measured by ELISA.

T cells that infiltrate pancreatic islets were isolated after intrabile duct injection and digestion with collagenase IV (Worthington Biochemical Corporation) (18). Perfused pancreata were incubated at 37°C for 30 min and washed twice with 5% FBS + Hanks’ balanced salt solution (Corning). Islets were handpicked and dissociated at 37°C in 1 mL of cell dissociation buffer (Gibco) for 15 min. Dissociated islets were washed with Hanks’ balanced salt solution before proceeding to analysis.

NOD.H2Ab1^−/−.DR4^Tg mice were immunized subcutaneously with GAD65 peptide emulsified in complete Freund adjuvant (CFA). Twelve days later, inguinal lymph nodes and spleens were harvested. For tetramer-based detection, cells were costained with phycoerythrin (PE)- and allophycocyanin-conjugated tetramers. Splenocytes were enriched by anti-PE MACS (Miltenyi Biotec). For polyclonal T-cell expansion, lymph nodes and spleens were mixed and cocultured with 10 μmol/L GAD65_115–127 peptides. Lymphocytes were isolated after 1 week with Ficoll-Paque PLUS (Cytiva). Isolated lymphocytes were supplemented with 20 units/mL human IL-2 (PeproTech), antigen-presenting cells (APCs) (NOD.HLA-DR4^Tg.H2Ab1^−/−.Rag1^−/− splenocytes or 2,500-rad irradiated NOD.H2Ab1^−/−.DR4^Tg splenocytes) and cultured with the priming peptide for another week. Lymphocytes were purified again with Ficoll-Paque PLUS before analysis or adoptive transfer (intraperitoneally).

All data analyses were performed using GraphPad Prism 10 (GraphPad Software). Statistical significance between sample means was set at P < 0.05.

Data generated and/or analyzed during the current study are available from the corresponding author upon reasonable request.

Native and deamidated GAD65_115–127/DR4 tetramers (tetWT and tet120E, respectively) were previously used to track T-cell responses during progression to type 1 diabetes in human patients (4). GAD65_115–127 epitope was 100% conserved between mouse and human (Supplementary Fig. 1A). To determine whether T cells of the same specificity can be identified in HLA-DR4 humanized NOD mice, we performed tetramer staining followed by magnetic bead enrichment of pooled peripheral secondary lymphoid organs. We identified native and deamidated GAD65_115–127 (p115.WT and p115.120E, respectively)–specific T cells in NOD.H2Ab1^−/−.DR4^Tg (DR4Tg) mice directly ex vivo but in relatively small numbers (10–80) (Supplementary Fig. 1B). Therefore, to effectively compare the phenotypes of T cells that recognize these two epitopes, we immunized DR4Tg mice with p115.WT/CFA or p115.120E/CFA to increase repertoire accessibility. T-cell expansion was observed in response to both peptides, but tet120E⁺ T cells expanded to significantly larger numbers in response to cognate peptide and unlike tetWT⁺ T cells, were also cross-reactive and increased after p115.WT/CFA immunization (Fig. 1A and B). Both tetWT⁺ and tet120E⁺ CD4⁺ effector T cells (Teffs) contained mostly antigen-experienced (CD44⁺) cells (Fig. 1C). Strikingly, Treg frequencies were reduced in the tet120E⁺ population expanded with p115.120E/CFA compared with the tetWT⁺ population with p115.WT/CFA (8.5% vs. 3.6%) (Fig. 1D–F), albeit with similar total Treg numbers between the groups, suggesting reduced capacity of Tregs to suppress p115.120E-induced Teff expansion. Overall, the higher number and lower Treg frequency of the tet120E⁺ CD4⁺ T-cell repertoire suggests that deamidation increases the immunogenicity of the GAD65_115–127 epitope.

To determine the functional capacity of T cells recruited in response to peptide immunization, intracellular interferon-γ (IFN-γ) was measured in T-cell lines expanded ex vivo. Consistent with tetramer-based analysis, p115.WT-primed CD4⁺ T cells upregulated IFN-γ in response to both p115.WT and p115.120E peptides, while CD4⁺ T cells from p115.120E-primed mice responded preferentially to p115.120E (Fig. 1G and Supplementary Fig. 1C). Tetramer analysis of the expanded cell lines confirmed the stability of relative tetramer frequencies in expanded cultures, which was reflective of tetramer frequencies observed directly ex vivo (Supplementary Fig. 1D–G). In combination, these data suggest that prolonged in vitro culture did not eliminate cross-reactivity to the native epitope in p115.120E-reactive CD4⁺ T cells. Interestingly, despite tet120E⁺ T-cell ability to expand in response to both peptides, very few tetWT⁺tet120E⁺ double-positive (DP) T cells were detected (Supplementary Fig. 1D–G). Given that tetramers bind only the very high-affinity T cells (19,20), we reasoned that tet120E⁺ T cells recognize the native epitope with a lower affinity insufficient for tetramer binding.

Negative selection of self-reactive thymocytes and development of tTregs depend on the expression of T-cell cognate self-antigens in the thymus (21,22). While PTM is a part of the normal cellular process, generation of nongermline-encoded autoimmune antigens through PTM is postulated to be induced locally in peripheral tissue and likely absent from the thymus. Consistent with this hypothesis, islet-infiltrating Tregs in NOD mice contain cells specific for insulin epitopes but not hybrid insulin peptides (23,24). Hybrid insulin peptides represent a drastic deviation from germline-encoded sequences. Our observation that deamidation of GAD65_115–127 leads to increased immunogenicity with reduced Treg frequency suggests potential differences in thymic development of p115.WT- and p115.120E-reactive T cells (Fig. 1D and E). To determine whether this single–amino acid change in GAD65_115–127 results in altered thymic selection, we used the Hu-Rg system. We cloned p115.WT- and p115.120E-specific TCRs from tetramer-enriched, ex vivo–expanded T-cell clones generated from peripheral blood mononuclear cells of patients with type 1 diabetes (WT-TCR and 120E-TCR, respectively). Expression and specificity of cloned TCRs were verified by tetramer staining of HEK293T cells cotransfected with TCR and CD3 complex (Fig. 2A and Supplementary Fig. 2A and B). Then, TCR functional avidities were determined by measuring IL-2 secretion from TCR-transduced 4G4 thymoma cell lines stimulated with peptide and APCs (Fig. 2B). Cell surface expression of CD3 was comparable between the two GAD65-specific TCR⁺ cell lines (Supplementary Fig. 2C). Deamidation modestly increases the binding affinity of GAD65_115–127 to HLA-DR4 (IC₅₀ 0.7–0.2 μmol/L) (4). However, WT-TCR–expressing 4G4s responded almost exclusively to p115.WT- and 120E-TCR-expressing 4G4s to p115.120E, with cross-reactivity between GAD65_115–127 peptides visible only at the highest concentrations, showing preferential TCR binding to either p115.WT or p115.120E epitope.

For in vivo expression of human-derived GAD65_115–127-specific TCRs, we used the Hu-Rg system and generated mice expressing a single TCR using DR4Tg.RagKO mice (15,16,25). Thymic analysis showed similar frequencies and numbers of ametrine⁺ (TCR vector fluorescent reporter) cells (Fig. 3A), suggesting comparable engraftment. Both sets of mice showed successful T-cell development, with predominant development of CD4 single-positive (SP) thymocytes (WT-TCR 16.9%, 120E-TCR 4.4%) (Fig. 3B and C). The ability of GAD65-specific T cells to avoid negative selection in these mice was consistent with the prevalence of GAD65-specific T cells in the peripheral blood of patients with type 1 diabetes (26–28). This finding also suggests that the level of GAD65 presentation in the thymus of either epitope was not sufficient to eliminate GAD65-reactive thymocytes. Importantly, we observed differences in the thymic development of Foxp3⁺ Tregs, which were reduced in the thymi of 120E-TCR compared with WT-TCR Hu-Rg mice (1.3% vs. 0.26%) (Fig. 3D–F). This observation was consistent with our analysis of polyclonal tetramer⁺ populations (Fig. 1D–F). Our data indicate that even a single–amino acid change in the epitope has the potential to skew the self-reactive T-cell balance between Tregs and Teffs.

To further demonstrate that the lack of p115.120E-specific Tregs directly depends on the amount of antigen in the thymus, we used exogenous delivery of peptide-pulsed DCs for thymic antigen delivery to increase the presentation of p115.120E (17) (Supplementary Fig. 3A and B). DCs were isolated from spleens of B16.Flt3l tumor-bearing DR4Tg.RagKO mice. Splenic DCs were pulsed with p115.WT, p115.120E, or negative control hemagglutinin (HA) peptide before being transferred intravenously into WT-TCR or 120E-TCR Hu-Rg mice. BM (femurs and tibiae) of Hu-Rg mice were harvested from recipients to measure engraftment. At the time of analysis, BM of WT-TCR and 120E-TCR Hu-Rg mice contained similar numbers and frequencies of ametrine⁺Sca-1⁺Lin⁻ hematopoietic stem cells (Supplementary Fig. 3C and D). We did not observe a change in frequency or number of immature ametrine⁺ thymocytes (MHCI^lo/−) (29) after cognate antigen-pulsed DC transfer (Supplementary Fig. 3E and F). However, CD69 expression increased in CD4SP thymocytes of both WT-TCR and 120E-TCR Hu-Rg mice in response to cognate antigen, indicating increased TCR signaling (Fig. 4A–C). However, activated caspase 3 expression, an indicator of negative selection, was only increased in WT-TCR Hu-Rg mice that received cognate antigen, suggesting 120E-TCR resistance to negative selection (Fig. 4D and E and Supplementary Fig. 3G). Peripheral recirculating mature Tregs have been shown to accumulate in the thymus with age (30). After excluding CD73⁺ recirculating Tregs (31), we still observed a twofold increase in WT-TCR tTregs and a fourfold increase in 120E-TCR tTregs generated in mice that received DCs pulsed with cognate peptides (Fig. 4F and G and Supplementary Fig. 3H). Interestingly, despite the difference in tTreg generation, Treg frequencies were not significantly different in spleens of WT-TCR and 120E-TCR Hu-Rg mice (Supplementary Fig. 4A and B). Cognate antigen transfers also did not induce a significant increase in splenic Treg frequencies (Supplementary Fig. 4H and I). This was potentially caused by the limited peripheral Treg niche in single TCR mice (32). Together, our data suggest that 120E-TCR thymocytes do not have an endogenous defect that impairs Treg development. Instead, deamidated GAD65_115–127-specific tTreg development is limited by the availability of the PTM epitope in the thymus.

Although GAD65 is one of the earliest identified, clinically relevant autoantigens associated with type 1 diabetes, the role of GAD65-specific CD4⁺ T cells in type 1 diabetes development remains unclear. Several clones of mouse and human GAD65-specific CD4⁺ T cells have been reported to infiltrate pancreatic islets to initiate insulitis/peri-insulitis (33–35). However, few have been shown to be diabetogenic (36). In other studies, GAD65-specific CD4⁺ T cells were associated with increased tolerance and protection against diabetes (37,38). While it is possible that T cells specific for different GAD65 epitopes have pathogenic or regulatory roles, given the preponderance of PTM β-cell antigen-reactive T cells after diagnosis (4), PTM GAD65 epitopes might be important autoimmune targets. To compare the peripheral function of native and deamidated GAD65_115–127-specific CD4⁺ T cells and their potential contribution to β-cell autoimmunity, we analyzed CD4⁺ T cells in spleen, nondraining lymph nodes (ndLNs), pancreatic lymph nodes (pLNs), and pancreatic islets of single TCR Hu-Rg mice. Although ametrine⁺CD4⁺ T cells were readily detected in all peripheral lymphoid organs, few were observed to spontaneously infiltrate pancreatic islets (Supplementary Fig. 4C–E). However, in vivo activation of WT-TCR and 120E-TCR T cells with peptide-pulsed DCs resulted in robust islet infiltration (Fig. 5A–C). Because DC-mediated antigen delivery can induce global CD4⁺ T-cell expansion, which was observed in 120E-TCR Hu-Rg mice (Supplementary Fig. 4F and G), we normalized the number of islet-infiltrating CD4⁺ T cells to the number of splenic CD4⁺ T cells. After normalization, the cognate antigen-induced islet infiltration remained significant for both WT-TCR and 120E-TCR Hu-Rg mice (Fig. 5B and C). Significantly elevated islet infiltration was also observed in WT-TCR Hu-Rg mice that received p115.120E cross-reactive antigen (Fig. 5A and B). This was not caused by a change in TCR specificity or differences in ability to upregulate effector functions after TCR expression in Hu-Rg mice since both Hu-Rg TCRs were dominated by production of IFN-γ in response to cognate peptide (Supplementary Fig. 5A–C). T-cell activation, quantified by the expression of CD69, was increased in draining pLNs and islets, reservoirs of endogenous GAD65 antigen, compared with ndLNs (Fig. 5D and E). These data suggest that initial priming is necessary to break GAD65-reactive T-cell tolerance, and after activation, both native and PTM epitope-reactive T cells are sensitized to endogenous antigen and infiltrate pancreatic islets.

GAD65 Hu-Rg T-cell dependence on priming before islet infiltration suggests the requirement for a trigger that releases antigen or alters the context of antigen presentation. Although DR4Tg mice are not susceptible to spontaneous diabetes (14), we observed an age-dependent increase in islet lymphocyte infiltration (Fig. 6A). To determine whether GAD65-reactive T cells are spontaneously recruited into the autoimmune response, we followed tetramer⁺ T cells in DR4Tg mice with age. Endogenous tetWT⁺ and tet120E⁺ CD4⁺ T cells accumulated in pancreatic islets of DR4Tg mice as they aged, while their numbers remained unchanged in secondary lymphoid organs (Fig. 6B and C and Supplementary Fig. 1B). Interestingly, although numbers of islet-infiltrating tetWT⁺ and tet120E⁺ CD4⁺ T cells were not significantly different in older DR4Tg mice, tet120E⁺ Teffs expressed significantly higher levels of T-cell activation markers (CD69, PD-1, and CD25) compared with tetramer⁻ Teffs, while no significant difference in exhaustion and anergy markers were observed in tetramer⁺ and tetramer⁻ populations (Fig. 6D–H). These data suggest that both native and deamidated GAD65-reactive CD4⁺ T cells can be spontaneously recruited into an anti-islet response, albeit their robust activation and expansion in DR4Tg mice depend on secondary triggers. Interestingly, deamidated GAD65_115–127-reactive T cells exhibited a higher level of activation than the native GAD65-reactive counterpart (Fig. 6E and F), further supporting our hypothesis that the deamidated epitope is more immunogenic.

Deamidation is an important form of PTM in autoimmunity and other diseases (39–42). One of the ways deamidation contributes to increased epitope immunogenicity is by addition of a negative charge to improve peptide binding within an MHC groove through favorable anchoring within a positively charged binding pocket, such as autoimmune-associated HLA-DQ8, HLA-DQ2, and H2-IA^g7 (11). Alternatively, deamidation of the TCR contact residue could generate an altered epitope that is relatively “foreign” to the immune system and recruits high-affinity T cells. In the case of GAD65_115–127, deamidation increases peptide/MHC binding affinity by 3.5-fold and can select for TCRs that preferentially recognize either one of the epitopes. The lack of thymic selection against PTM neoantigens has been demonstrated previously (9,24). Such results reinforce the hypothesis that generation of PTM antigens is highly restricted to peripheral tissues, leading to a deficiency in thymic presentation of PTM antigenic epitopes. However, deamidation can be spontaneous or catalyzed. Tissue transglutaminase, which mediates deamidation, is universally expressed, including the thymus (43). Activation of tissue transglutaminase requires a high calcium concentration and is tied to endoplasmic reticulum stress. Thymic epithelial cells, like β-cells, are also burdened with high protein production and susceptible to stress-induced damage (44,45). Therefore, it is possible for some mechanisms of posttranslational modifications, like deamidation, to happen in the thymus. Alternatively, it is conceivable that islet-generated PTM antigens could be brought into the thymus by migratory DCs (46,47). To directly address relative development and peripheral activation of native and deamidated GAD65 epitope–specific T cells in vivo, we used DR4 tetramers and a humanized TCR model.

Our observations suggest that in HLA-DR4 humanized NOD mice, deamidation of GAD65_115–127 generates a more immunogenic epitope. However, it is worth noting that p115.WT-specific T cells were also expanded after immunization with the native epitope and exhibited robust IFN-γ responses, suggesting that tolerance to the native GAD65 epitope is incomplete. Indeed, in Hu-Rg mice that expressed either the p115.WT-specific WT-TCR or the p115.120E-specific 120E-TCR, we observed efficient CD4⁺ T-cell development and minimal indications of negative selection. While tTreg development was reduced in 120E-TCR Hu-Rg mice, exogenous DC-mediated antigen delivery partially restored tTreg frequencies. These data support the role of cognate epitope in Treg development while simultaneously exposing the deficiency of the native epitope normally presented in the thymus in driving PTM Treg development.

Reactivity to PTM antigen could also influence positive selection. The nature of positively selecting ligands is less defined but is generally accepted to be low-affinity self-ligands. If native autoantigen is the positive selection ligand for GAD65 epitope–reactive T cells, high specificity to the PTM antigen and low cross-reactivity to the native antigen could hinder positive selection. This could potentially explain the reduced frequency of CD4SP and higher frequency of DP thymocytes in 120E-TCR thymi compared with WT-TCR (Fig. 3B and C).

Despite the limited and sometimes contradicting observations surrounding GAD65 T-cell reactivity in type 1 diabetes pathogenesis, observations that deamidated and citrullinated GAD65 epitopes are recognized by elevated numbers of T cells in patients with type 1 diabetes suggest that pathogenic T-cell responses are targeted against PTM epitopes (4). It is possible that in type 1 diabetes, GAD65 epitopes need to be “activated” by posttranslational modifications. However, we did not observe substantial spontaneous islet infiltration in 120E-TCR Hu-Rg mice. Instead, for both WT-TCR and 120E-TCR, islet infiltration required priming by exogenous antigen delivery, which could reflect an insufficiency of antigen presented in mouse pancreatic islets (48). In contrast, CD69 upregulation was also observed in pLN and islet CD4⁺ Teffs that received cross-reactive peptide or the control HA peptide (Fig. 5D and E), suggesting that endogenous GAD65 was sufficient to induce CD69 upregulation on pLN and islet-infiltrating WT-TCR and 120E-TCR T cells. An important question then remains: How are GAD65-reactive T cells recruited into the anti–β-cell immune response during progression to type 1 diabetes? It is possible that efficient release and presentation of GAD65 epitopes requires preexisting β-cell stress or damage. For example, streptozotocin-caused DNA damage drives GAD65 upregulation in β-cells. However, this would imply that GAD65-specific T cells are unlikely to be initiators of type 1 diabetes but contributors to later-stage disease progression (49).

Although there are clear differences regarding Treg development between native and PTM epitopes, effector T cells appear to be promiscuous in their recognition of the two epitopes in the periphery (Fig. 1B and C and Supplementary Fig. 1B–G). Despite the low level of reactivity to p115.120E in vitro, WT-TCR Hu-Rg T cells were efficiently primed and infiltrated islets after delivery of p115.120E. It is possible that administered peptides were subject to modifications, and we cannot rule out deamidation of p115.WT peptide in vivo. However, subsequent in vitro expansion with p115.WT failed to reduce populational cross-reactivity. Therefore, in vivo peptide modifications were unlikely to contribute in a major way to the promiscuity in epitope recognition we observed.

In our study, we have expanded on the NOD.DR4 humanized model by combining it with tetramer analysis of polyclonal T-cell responses and expression of human TCRs to improve the translational aspect of the model. Continuous utilization of humanized HLA and TCR in mice will provide new opportunities to dissect the role of various antigens and HLAs in human type 1 diabetes. Importantly, our observations further expand the growing list of type 1 diabetes epitopes that do not support efficient Treg development. The lack in Treg development could be compensated by exogenous delivery of the cognate epitope, which might become a viable therapeutic approach to expand the repertoire of β-cell antigen-specific Tregs. Overall, our data emphasize the deficiency of Treg development against PTM diabetogenic epitopes as a potential key mechanism contributing to perpetual β-cell destruction and epitope spreading in type 1 diabetes. To dissect these underlying mechanisms, our study demonstrates the value of combining HLA Hu-Rg mice to study the intricacies of T-cell–mediated human autoimmune diseases.

Acknowledgments. The authors thank the National Institutes of Health Tetramer Core Facility (Atlanta, GA) for providing the GAD65_115-127/DR4 tetramers. The authors also thank the Cytometry and Cell Sorting Core at Baylor College of Medicine and the Flow Cytometry Core at the University of Utah for technical support.

Funding. This work was supported by the National Institutes of Health (grant AI125301 to M.B. and DK114456 to M.L.B.) and the Robert and Janice McNair Foundation.

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

Author Contributions. Y.J., M.L.B., E.A.J., and M.B. designed and/or performed the experimental work. Y.J. and M.B. wrote the manuscript. Y.K., J.M., G.B.-F., T.L., and S.O.-F. performed the experimental work. M.L.B., E.A.J., and M.B. provided reagents and resources. All authors revised the manuscript and gave their approval of the version to be published. M.B. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.