A Critical Insulin TCR Contact Residue Selects High-Affinity and Pathogenic Insulin-Specific T Cells

The two susceptibility alleles most highly associated with developing type 1 diabetes (T1D) are the HLA and the insulin promoter region (1). This combination of HLA and antigen alleles suggests a primary role for the T-cell receptor (TCR)/peptide/MHC trimolecular complex in the initiation and development of a T cell–mediated disease. Position 57 of the HLA-DQ β-chain is particularly important, as susceptibility correlates with uncharged residues (Val, Ser, or Ala), while a negatively charged aspartic acid (Asp) residue found at this position is protective (2). MHC haplotype is similarly important in T1D susceptibility of NOD mice, where 57 of the MHC β-chain also has a non-Asp residue (Ser). It is thought that HLAs with non-Asp residues form a shallow groove that destabilizes binding of self-peptides and alters negative selection during thymocyte development (3,4). Additionally, the ability of the insulin epitope B:9-23 (InsB_9-23) to slide within the mouse MHC class II groove allows for the positively charged residue within the I-Ag7 to repel a positive residue (Arg) in position 9 of insulin peptide InsB_9-23 (5,6). Together, these molecular interactions at the interface of key antigen insulin and susceptible HLA alleles suggest an environment for unstable trimolecular complex formation and weak thymocyte selection pressures.

In humans, the insulin gene variable nucleotide tandem repeat allele is associated with T1D susceptibility, while the role of insulin expression in mice is less clear. This is due in part to mice expressing two insulin genes, with Ins1 differing from Ins2 by just two amino acids (in the β-chain at position 9 and position 29) (7,8). However, the Ins2 product preproinsulin 2 is the predominant isoform expressed in the mouse thymus (8,9). Importantly, it has been shown that a majority of the islet-infiltrating CD4 and CD8 T cells are reactive to insulin epitopes (10). Interrogation of the antigenic InsB_9-23 peptide by an alanine scan revealed the tyrosine residue at position 16 as a critical residue for CD4 T-cell clone recognition (11). Importantly, transgenic (Tg) substitution of endogenous insulin genes with the mutated form (Y16A) resulted in complete protection of NOD mice from diabetes development, due in part to ignorance of peripheral antigen (12,13). However, how and whether insulin antigen expression shapes autoreactive T-cell repertoire remain unknown.

Foxp3-expressing regulatory T cells (Tregs) specific for self-antigen are known to express high levels of CD5 and Nur77 compared with non-Tregs, indicative of their preferentially high affinity for self-antigens (14–16). We have previously shown that most insulin-specific (InsB_9-23) TCRs can generate a number of Foxp3⁺ T cells in TCR retrogenic (Rg) mice (17,18); however, it is not known whether altering the TCR contact residue at position 16 affects Treg development. To determine the role of insulin epitope in shaping pathogenic and Treg repertoires, we analyzed T-cell development, TCR affinity, and T-cell pathogenic potential using tetramer analysis and single TCR NOD and NOD.Ins^Y16A mutant mice.

NOD/ShiLtJ (NOD), NOD CD45.2 (NOD.C-[Ptprc-D1Mit262]/WehiJ), and NOD.CB17-Prkdcscid/J (NOD.scid) mice were purchased from The Jackson Laboratory, bred, and maintained under specific pathogen-free conditions in accordance with the Institutional Animal Care and Use Committee at Baylor College of Medicine. NOD.Cg-Tg(Ins2*Y16A)3Ell Ins1^tm1Jja Ins2^tm1Jja/GseJ (NOD.Ins^Y16A Tg) mice were a gift from Dr. Maki Nakayama (University of Colorado Denver).

InsB_9-23 (C19A) (SHLVEALYLVAGERG; weak agonist peptide) was used in immunization, expansion, and restimulation of CD4⁺ T-cell lines. Tetramers presenting HLVERLYLVCGGEG and HLVERLYLVCGEEG peptides were generated with the help of the National Institutes of Health (NIH) Tetramer Core Facility.

Flow cytometry analyses were performed on an LSRFortessa II (BD Biosciences), and data were analyzed with FlowJo software (Tree Star, Inc.). Monoclonal antibodies against the following molecules were used: Foxp3 (FJK-16s) from eBioscience and CD3 (145-2C11), CD4 (GK1.5), CD5 (53-7.3), CD8 (53-6.7), PD-1 (29F.1A12), FR4 (12A5), CD73 (TY/11.8), CD44 (IM7), TCRβ (H57-597), IFNγ (XMG1.2), CD45.2 (104), and CD45.1 (A20) from BioLegend.

CD4⁺ T cells were expanded from the spleens of 6- to 8-week-old female NOD or NOD.Ins^Y16A Tg mice immunized with insulin C19A. T-cell lines were generated by plating 5 × 10⁶ splenocytes/well in a 24-well plate with complete RPMI containing 10% FCS and 10 μmol/L peptide. After 2 weeks, CD4⁺ T cells were purified with magnetic-activated cell sorting using the manufacturer’s protocol. The average purity was >95% and <1% Foxp3⁺. Five million purified CD4⁺ T cells were transferred to 8- to 10-week-old NOD.scid female recipients.

Diabetes development was monitored weekly with Diastix (Bayer), and positive readings were confirmed with the Breeze2 glucometer (Bayer). Mice were considered diabetic if their blood glucose was >400 mg/dL.

Pancreatic islets were isolated after intrabile duct injection and digestion with collagenase IV (Worthington Biochemical). Islets were handpicked and incubated at 37°C for 15 min in 1 mL cell-dissociation buffer (Invitrogen) and further dissociated by vortexing. Cells were then washed in 10 mL 5% FBS/Hanks’ balanced salt solution and analyzed by flow cytometry.

RNA was isolated from sorted CD4⁺ T cells (CD4⁺CD3⁺). All samples were sorted with >97% purity. cDNA was synthesized using the SMARTer Ultra Low Input RNA Kit (Clontech Laboratories). Library preparation was performed with the Illumina Nextera XT kit before paired-end RNA sequencing (RNA-seq) using the Illumina NextSeq 500 platform for 150 cycles (NextSeq 500 Mid-Output Kit). Sequencing reads were aligned to the mouse genome (RefSeq mm10) using Spliced Transcripts Alignment to a Reference (STAR), and transcript assembly and fragments per kilobase of transcript per million mapped reads estimates were obtained. DESeq2 was used to compute differential expression between groups, with differentially expressed genes defined by q < 0.05 with Benjamini-Hochberg correction for multiple testing, followed by filtering of data to remove values with log₂ fold changes greater than 4 and less than −4. Heat maps and principal component analysis were generated in R (version 3.2.3) using the Heatmaps package (version 1.0.12) with viridis (version 0.4.0). All original RNA-seq data were deposited in the National Center for Biotechnology Information’s Gene Expression Omnibus (GEO) database.

Retroviral transduction of mouse bone marrow cells was performed as previously described (17,19–22). Transduced NOD.scid.CD45.2⁺ bone marrow cells were collected, washed, and resuspended in PBS with 0.5% FBS. Bone marrow cells were injected via the tail vein into sublethally irradiated (600 rad) CD45.1⁺ NOD wild-type (WT) or NOD.Ins^Y16A Tg (Ins1^−/−Ins2^−/−) mice. Mice were analyzed 6 weeks after bone marrow transplant.

The details of the micropipette adhesion frequency assay have previously been described (23–25). In brief, peptide–MHC (pMHC)–coated red blood cells (RBCs) and T cells were placed on opposing micropipettes and mechanically brought into contact in a controlled contact area (A_c) for 2 s. The T cell was retracted at the end of the contact period, and the presence of adhesion (indicating TCR–pMHC ligation) was observed microscopically by elongation of the RBC membrane. This contact–retraction cycle was performed 50 times per T cell–RBC pair to calculate an adhesion frequency (P_a). For each cell pair, two-dimensional (2D) affinity (A_cK_a) was then calculated using the following equation: A_cK_a = −ln[1 − P_a]/(m_rm_l), where m_r and m_l reflect the receptor (TCR) and ligand (pMHC) densities, respectively.

A 60–base pair sequence of the rearranged TCRβ CDR3 region was amplified and sequenced for all samples using the immunoSEQ assay (Adaptive Biotechnologies) as previously described (19). This is a multiplex PCR and high-throughput assay for the rearranged DNA of T and B cells. This assay uses a combination of adjusted primer concentrations and computational corrections to correct for PCR bias common to multiplex PCR. Raw sequence data were filtered based on the TCRβV, -D, and -J gene definitions provided by the ImMunoGeneTics information system database (www.imgt.org) and binned using a modified nearest-neighbor algorithm to merge closely related sequences and remove both PCR and sequencing errors (26).

Unless otherwise described, all analysis was performed using Prism 5 (GraphPad Software). Outlier tests were performed on all data sets and, where identified, are noted in the figure legends.

The data sets generated during the current study are available from the corresponding author upon reasonable request. The data sets generated and analyzed during the current study are available in the GEO repository and can be found under the identifier GSE139544.

Previous work has shown that pathogenic T cells develop and persist in NOD.Ins^Y16A Tg diabetes-free mice and cause diabetes upon transfer into NOD.scid mice that have WT insulin genes (12,13). However, the effect that the mutated epitope could have on an insulin-reactive repertoire was not investigated. Therefore, we sought to first quantify and compare the number of thymic tetramer⁺ (21G22E; HLVERLYLVCGGEG) (27) cells in NOD and NOD.Ins^Y16A Tg mice. The InsB_{10-23(21G22E)} tetramer includes insulin peptide modified at p8(InsB:21) and p9(InsB:22) to enhance pMHC complex stability (27). To measure thymic tetramer cell numbers, magnetic column tetramer enrichment of three pooled thymi was performed as previously described (28). Similar numbers of tetramer⁺ CD4 single positive (CD4SP) cells were found within the enriched pooled thymic samples from NOD and NOD.Ins^Y16A mice, indicating comparable thymic output (Fig. 1A). Additionally, based on tetramer mean fluorescence intensity (MFI), there did not appear to be a difference in TCR affinity for insulin peptide between NOD and Ins^Y16A mice (Supplementary Fig. 1A). Due to the normally low frequency of insulin-specific T cells in peripheral lymphoid organs of unprimed mice (Supplementary Fig. 1B), we used an immunization strategy to expand the population to levels that allow a more in-depth analysis. NOD and NOD.Ins^Y16A Tg mice were first immunized with InsB_9-23 and complete Freud’s adjuvant (CFA), and InsB_{10-23(21G22E)} tetramer⁺ cells were enriched from the spleens and analyzed by flow cytometry 12 days postimmunization. Both the frequency and absolute number of tetramer⁺ cells were significantly increased in Ins^Y16A Tg mice compared with NOD WT mice after immunization, suggesting increased starting number and/or affinity of T cells in Ins^Y16A mice (Fig. 1B). Although we expected that CD5 expression would be decreased on insulin-reactive T cells that develop in the absence of cognate peptide in Ins^Y16A mice, the level of CD5 expression was comparable between the two groups, suggesting tetramer⁺ insulin-reactive T cells that escape deletion in WT mice are ignorant of insulin epitope in the thymus (Fig. 1C). While there was no difference between markers associated with exhaustion and anergy (PD-1 and FR4/CD73) (29,30), Ins^Y16A tetramer⁺ cells had increased expression and number of CD44⁺ cells, suggesting increased activation of Ins^Y16A T cells (Fig. 1C). Importantly, we observed a significant reduction in the ratio of tetramer⁺Foxp3⁺ Tregs in Ins^Y16A mice compared with controls, although total numbers of Foxp3⁺ Tregs were comparable (Fig. 1D).

To determine whether the increase in InsB_9-23–reactive T cells is a result of defective negative selection resulting in a unique TCR repertoire, T cells from immunized mice were expanded in vitro with InsB_9-23 peptide for 2 weeks and then analyzed by high-throughput, next-generation TCR sequencing. Comparing Ins^Y16A Tg and WT T-cell lines revealed a significant decrease in diversity and elevated oligoclonality (higher/restricted clonality) of Ins^Y16A TCRs (Fig. 2A). The average CDR3 length was slightly longer in Ins^Y16A T cells, suggesting perhaps alterations in the TCR repertoire (Supplementary Fig. 2A). However, Vβ usage showed similar distribution between the two groups, albeit there was clear focusing of the repertoire toward few select Vβ families, most prominently Vβ2 (Fig. 2B). The overall diversity of the InsB_9-23 repertoire was high, leading to the lack of significant sharing of TCR sequences among Ins^Y16A samples (Supplementary Table 1); however, several high-expressing clones were shared among Ins^Y16A and WT samples (Supplementary Fig. 2B and C). Together, these data indicate that expression of mutated epitope during thymic selection primarily affects Treg development rather than causing global alterations in the TCR repertoire.

Our data thus far suggest the Y16A mutation may not have a significant effect on the selection of self-reactive TCR repertoire. However, in a polyclonal system, it is difficult to determine how the altered insulin epitope confers selection pressures on a specific TCR. In order to address the impact of the Ins^Y16A mutation on the development of single insulin-reactive TCR, we used TCR Rg technology (17,19–22) to generate mice expressing a well-characterized insulin-reactive TCR (18) in either the presence of Y16A insulin or WT insulin (Fig. 3A, top). Sublethally irradiated CD45.1 NOD WT and Ins^Y16A Tg (Ins1^−/−Ins2^−/−) mice received NOD.scid.CD45.2⁺ bone marrow transduced with 4-8 TCR retrovirus. Six weeks after bone marrow transfer, we analyzed the development of CD45.2⁺ T cells expressing insulin-reactive TCRs. Sufficient reconstitution of host thymus by donor cells was seen in all mice (Supplementary Fig. 3). Total thymocyte cellularity of the CD45.2⁺ 4-8 Rg cells was similar between mice expressing Y16A and WT insulin (Fig. 3A, bottom). Additionally, there was no difference in CD4⁺ T-cell selection as measured by CD4SP ratio and number, TCR MFI, or CD5 MFI (Fig. 3B). As expected, 4-8 insulin-reactive T cells accumulated to similar numbers in peripheral lymphoid organs but were not able to infiltrate pancreatic islets of epitope mutant mice (Fig. 3C). Although these data indicate that insulin epitope does not play a role in the negative selection of the 4-8 TCR thymocytes, subtle changes in TCR reactivity to pMHC could alter the strength of signal necessary for Foxp3 expression and Treg development (17,31,32) (Fig. 1D). Therefore, we measured the ratio and number of Foxp3-expressing CD4SP thymocytes and peripheral Foxp3-expressing Tregs in the pancreatic lymph nodes (PLNs) and islets. The analysis revealed a dramatic loss of Foxp3⁺ cells within the CD4SP thymocytes of Ins^Y16A Tg mice (Fig. 4A). Additionally, Foxp3-expressing Tregs were almost completely void in the PLNs and islets (Fig. 4B and C), indicating the InsB amino acid residue 16Y is a critical contact and binds 4-8 TCRs to mediate InsB_9-23–specific Treg lineage commitment.

Although TCR repertoire analysis showed limited alterations in Ins^Y16A Tg insulin-reactive T cells, we postulated that there might be intracellular rather than clonotypic changes within the insulin-reactive T cells that develop in the absence of cognate self-antigen. Therefore, we assessed the functional capacity of insulin-reactive T cells that developed on the mutant background. To test whether InsB_9-23–specific CD4⁺ Ins^Y16A T cells are more pathogenic, T-cell lines from Ins^Y16A and WT NOD mice were generated by immunization with InsB_9-23 in CFA, as before. Splenocytes were harvested 12 days later and expanded with 10 μmol/L insulin epitope for 14 days, after which CD4⁺ T cells were purified with magnetic-activated cell sorting and transferred to NOD.scid mice. NOD.scid mice that received Y16A T cells began developing diabetes just 3 weeks posttransfer and had an overall increased rate of diabetes compared with NOD.scid mice receiving WT T cells (Fig. 5A). Because there were negligible frequencies of Foxp3⁺ T cells in transferred cell lines, the result suggested an intrinsic difference in the function of effector T cells that develop in the absence of cognate antigen. To identify the enhanced effector mechanism behind Ins^Y16A T-cell pathogenicity, we performed RNA-seq analysis on Ins^Y16A and WT T cells from prediabetic transferred NOD.scid mice 5 weeks posttransfer. From this analysis, 91 genes were differentially expressed (q < 0.05) between Ins^Y16A and WT islet-infiltrating T cells (Supplementary Fig. 4). Gene ontology pathway analysis revealed alterations in numerous immune pathways, including T-cell differentiation and cell-to-cell adhesion pathways with particular upregulation in Trat1, Ctla4, Fos, and Junb associated with Ins^Y16A cells (Supplementary Fig. 4A). Interestingly, there also appeared to be a decrease in the production of chemokine ligands expressed by Ins^Y16A T cells that are responsible for recruitment of Tregs (CCL4 and CCL5) (Supplementary Fig. 4A).

Given the increased expression of genes associated with TCR signaling in islet-infiltrating Ins^Y16A cells, we wanted to assess whether there was a preferential expansion of high-affinity clones at the tissue site. Therefore, we compared T-cell reactivities and biophysical TCR affinities of the two cell populations in the spleens and pancreatic islets. Prior to transfer, Ins^Y16A cells exhibited increased reactivity to insulin peptide, as measured by IFNγ secretion, suggesting an overall higher sensitivity for antigen (Fig. 5B). Next, we compared insulin tetramer binding using two insulin tetramers expressing 22E- and 21G22E-modified versions of InsB_10-23 peptide. While both modifications allow for the insulin peptide to be presented in an unfavorable but dominant register 3, InsB_10-23(22E) tetramer is thought to bind type A insulin–specific T cells that respond to naturally processed insulin, while InsB_{10-23(21G22E)} tetramer binds type B T cells that are not activated by naturally processed insulin (27,33). While overall low frequencies of tetramer⁺ T cells in unprimed NOD and Ins^Y16A mice did not show obvious differences (Fig. 1A and Supplementary Fig. 1), insulin peptide immunization and expansion revealed a significant increase in the percentage of insulin tetramer⁺ T cells in T-cell lines derived from immunized Ins^Y16A mice (Fig. 5C). Interestingly, the increase in tetramer staining was more pronounced for InsB_10-23(22E) compared with InsB_{10-23(21G22E)} epitope (4.6-fold vs. 1.9-fold increase). This observation is consistent with the hypothesis that type A insulin–reactive T cells are more sensitive to thymic expression of naturally processed insulin antigen, and high-affinity type A cells are more likely to escape selection in Ins^Y16A mutant mice (33). To assess the contribution of high-affinity clones to β-cell destruction in vivo, transferred CD4⁺ T cells were isolated directly from the spleens and islets of NOD.scid recipient mice, and their TCR affinities were compared by 2D Biophysical Affinity measurements (18,24). Use of 2D allows quantification of TCR affinities in individual T cells by measuring the frequency of productive T contacts with RBCs coated with InsB_9-23(21G22E):H-2A^g7 monomers (24). The effective affinities of T cells isolated from both the spleen and islets of Ins^Y16A mice were significantly higher than those of WT mice (Fig. 5D). Interestingly, spleens of mice that received Ins^Y16A T cells exhibited two distinct subpopulations of cells: a lower-affinity subpopulation, comparable to those observed in WT T cells, and a higher-affinity subpopulation (19.4% of Ins^Y16A T cells, >0.5 × 10⁻⁴ μm⁴) that was absent from spleens of mice that received WT T cells. Within the islets, both the WT and Ins^Y16A T cells contained high-affinity T cells (12% and 22%, respectively), although the average affinity was still significantly higher in Ins^Y16A T cells. Overall, our data show that insulin expression in the periphery is a critical component of self-tolerance. Interestingly, high-affinity T cells were able to escape thymic selection in both WT and Ins^Y16A mice, although insulin-reactive T cells from Ins^Y16A mice exhibited increased pathogenic capacity. Together, these observations suggest that high-affinity pathogenic insulin-reactive T cells are subject to regulation on multiple levels, including partial deletion during thymic selection, suppression by Tregs, and cell-intrinsic peripheral regulatory mechanisms, all of which are mediated by self-antigen.

MHC class II–restricted insulin epitopes can “slide” within the MHC groove and be presented in different “registers” or positions (34,35). This phenomenon has been attributed to the InsB_9-23 epitope, in which three different potential registers have been identified (34,36). The proportion of T cells that recognize each of the three registers is unknown; however, infiltrating CD4⁺ T cells in NOD mice can bind tetramers presenting the core insulin epitope in all three registers (27,37,38). While cross-reactivity and degeneracy of the insulin-reactive polyclonal T-cell population have not been fully defined, we surmise that our findings will primarily be applicable for register 3 T cells for two main reasons. Firstly, register 3 is considered the most unstable, and thus, it is likely that T cells specific for InsB_9-23 bound in register 3 are more likely to escape thymic selection and comprise the majority of the InsB_9-23–reactive population in the periphery. Secondly, the generation of the Y16A insulin mutant and the InsB_{10-23(E21G/R22E)} tetramer was designed with register 3 in mind (35). Although we cannot definitively exclude that the InsB_{10-23(E21G/R22E)} peptide is presented in more than one register within the tetramer reagent, we think it is unlikely, as the negatively charged E will be more likely to bind the positively charged pocket of I-Ag7 in register 3 conformation. Because our study used unmodified full-length InsB_9-23 WT peptide for priming and T-cell line expansion, our functional analysis did not exclude cells that are specific for a particular register. Our tetramer analysis shows that register 3–reactive T cells are affected by the tyrosine residue, and the majority of cells that are expanded on WT peptide are specific for register 3 (48–58% in WT and 79–96% in Ins^Y16A) (Fig. 5D). Based on the recently solved crystal structure of register 1 and 2 T-cell clones, it seems likely that recognition of InsB_9-23 is dependent on Y16 for all three registers (39); therefore, we cannot exclude the possibility that register 1– or 2–reactive clones would be similarly affected in Ins^Y16A mice. Nevertheless, the enrichment of InsB_{10-23(E21G/R22E)}–reactive T cells in primed Ins^Y16A mice suggests preferential expansion of register 3–specific T cells in the absence of Y16.

Although the original studies by Nakayama et al. (12,13) showed that splenocytes from Ins^Y16A mice are less pathogenic compared with WT NOD splenocytes in a transfer model, subsequent study showed that this was likely due to lack of priming, and exposure of these cells to WT insulin led to increased pathogenicity of splenocytes obtained from mutant mice. Because the studies confirmed the presence of diabetogenic T cells in diabetes-resistant Ins^Y16A mutant mice, we aimed to dissect the developmental differences and pathogenic potential of insulin-reactive T cells that develop in mutant mice. We found that in the absence of recognizing their cognate antigen in the thymus, a subset of pathogenic insulin-reactive T cells with high affinity escaped negative selection and displayed an increased sensitivity to insulin peptide (Figs. 1C, 2, and 5). Although similar numbers of insulin-reactive thymic and peripheral CD4 T cells were seen in both NOD WT and Y16A insulin mutant mice, this is mostly likely due to the absence of the InsB_9-23 epitope within Ins^Y16A mice. Indeed, the insulin-specific Ins^Y16A T cells expanded robustly in response to InsB_9-23 in vitro and were highly pathogenic in vivo after transfer (Figs. 1B and 5). These data suggest that thymic insulin expression restricts the TCR repertoire to low-affinity effector T cells while at the same time allowing for InsB_9-23–specific Treg development.

The development of thymically derived Tregs is critical to the maintenance of immune homeostasis, including the prevention of autoimmune diseases such as T1D (17,40–42). Using CD5 and Nur77 as surrogates for TCR reactivity, it has been shown that stable Foxp3 expression is dependent upon higher avidity TCR/pMHC interactions (15,17,43,44). Previously, we increased the overall TCR avidity of insulin-reactive T cells and altered Treg development by ectopically expressing insulin B in bone marrow–derived antigen-presenting cells (17). The increase in thymic Treg development completely protected high-affinity 4-8 and low-affinity 12-4.1 TCR Rg NOD mice from developing T1D (17). In our current studies, we used the 4-8 TCR Rg system to determine whether disruption of an important TCR contact residue (InsB_9-23 Y16) will negatively affect the development of insulin-specific Tregs. Treg development was greatly diminished within the Ins^Y16A mice reconstituted with 4-8 TCR bone marrow compared with NOD mice (Fig. 4). We did not observe any other developmental distinctions between the two backgrounds, indicating a certain level of antigen-specific TCR reactivity is necessary for positive selection and survival of insulin-reactive Treg cells compared with conventional CD4⁺ T cells. However, our results also indicate positive selection and survival of a high-affinity and pathogenic effector T-cell population. The explanation for why high-affinity effector T cells survive, but Foxp3-expressing Tregs do not develop, may be due to the dependence of Foxp3 expression upon relatively higher levels of TCR stimulation (17,44–46), while positive selection and survival of high-affinity TCRs only require positive selection with low-affinity ligands (15,47). Indeed, our data suggest that the lack of a stable TCR/pMHC complex formation results in an altered TCR repertoire with significantly higher-affinity TCRs and an activated transcriptional profile. Moreover, in the absence of selection pressures by the p16 tyrosine residue on the TCR repertoire, highly pathogenic T-cell clones cause rapid islet destruction in the presence of their cognate ligand. Together, our data describe a critical role for the insulin peptide position 16 tyrosine residue in restricting the pathogenic TCR repertoire as well as the development of insulin antigen-specific Tregs.

Acknowledgments. The authors thank Dr. Maki Nakayama (University of Colorado Denver) for the kind gift of the NOD Y16A Ins1^−/−Ins2^−/− mice. The authors also thank the NIH Tetramer Core Facility (Atlanta, GA) for the InsB_{10-23(21G22E)} tetramer and the InsB_10-23(22E) tetramer, the Cytometry and Cell Sorting Core at Baylor College of Medicine (Houston, TX) for support and funding, and Joel M. Sederstrom (Baylor College of Medicine) for assistance.

Funding. This work was supported by the NIH (grants R01-AI-125301 to M.Be. and K22-AI-104761, R56-DK-104903, and R01-DK-119352-01 to M.L.B.), the Cytometry and Cell Sorting Core at Baylor College of Medicine with funding from the NIH (National Institute of Allergy and Infectious Diseases grant P30-AI-036211, National Cancer Institute grant P30-CA-125123, and National Center for Research Resources grant S10-RR-024574), the New York Stem Cell Foundation to M.A.S., American Diabetes Association Junior Faculty Awards 1-15-JF-07 to M.Be. and 1-17-JDF-013 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. M.Be., M.A.S., B.L., B.D.E., and M.L.B. designed and/or performed the experimental work. M.Be., M.L.B., and M.Bo. provided reagents and resources. M.Be., M.A.S., B.L., and M.L.B. wrote the manuscript. E.K. and L.G. performed the experimental work. M.L.B. is the guarantor of this work and, as such, had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.