Type 1 diabetes is a T cell–mediated autoimmune disease that is characterized by Ag-specific targeting and destruction of insulin-producing β cells. Although multiple studies have characterized the pathogenic potential of β cell–specific T cells, we have limited mechanistic insight into self-reactive autoimmune T cell development and their escape from negative selection in the thymus. In this study, we demonstrate that ectopic expression of insulin epitope B:9–23 (InsB9–23) by thymic APCs is insufficient to induce deletion of high- or low-affinity InsB9–23–reactive CD4+ T cells; however, we observe an increase in the proportion and number of thymic and peripheral Foxp3+ regulatory T cells. In contrast, the MHC stable insulin mimetope (InsB9–23 R22E) efficiently deletes insulin-specific T cells and prevents escape of high-affinity thymocytes. Collectively, these results suggest that Ag dose and peptide–MHC complex stability can lead to multiple fates of insulin-reactive CD4+ T cell development and autoimmune disease outcome.
Thymocyte selection is essential for maintaining immune homeostasis and establishing central tolerance to self-peptides through a process known as clonal deletion (1). Early thymocyte progenitors endure strict selection pressures within the thymic cortex and medulla through positive and negative selection, respectively (1, 2). During negative selection, efficient deletion of high-affinity autoreactive T cells involves medullary thymic epithelial cells (mTECs) and thymic dendritic cells (DCs) (3). mTECs can present a wide range of tissue-restricted Ags driven by the endogenous expression of autoimmune regulator (Aire) (4, 5). DCs serve as a second self-antigen source by capturing and cross-presenting mTEC-derived tissue-restricted Ags or transporting tissue Ags from the periphery to the thymus (6, 7). Why mechanisms of central tolerance can fail in autoimmune disorders, such as T1D, is still unclear.
Type 1 diabetes (T1D) is a chronic T cell–mediated autoimmune disorder that specifically targets and destroys β cells in the pancreatic islets of Langerhans; however, the mechanisms involved in breaking self-tolerance appear to be multifaceted and poorly understood (8). Whereas T1D diagnosis is based in part on islet-specific Abs detected within serum (9), autoreactive T cells are the primary population that drives disease development and progression (10). Genome-wide association studies have revealed a strong linkage between certain HLA alleles and diabetes susceptibility, indicating a role for the TCR–peptide–MHC (TCR–pMHC) trimolecular complex (11). The susceptible allele HLA-DQ8 is associated with relatively unstable binding and self-Ag presentation upon the shallow HLA groove, presumably reducing TCR T cell avidity and lowering negatively selecting ligands in the thymus (12, 13). Interestingly, the MHC class II (MHC II) allele that confers most of the genetic susceptibility in NOD mice has similar structural homology to DQ8 and has been shown to display unstable binding to the dominant insulin epitope B:9–23 (InsB9–23) (14, 15). This instability of the self-peptide–MHC complex allows peptide shifting upon MHC II into different registers or positions, warranting reduced TCR avidity and net concentration of certain peptide–MHC (pMHC) complexes to thymocytes (12, 13, 16). In support of this concept, studies have shown that different register shifts of insulin B exist between the thymus and pancreas, resulting in neoantigen formation (11–13).
Aside from register shifting, tissue-specific neoantigens can arise in β cells because of oxidative and endoplasmic reticulum stress during high demands of insulin production (17–19). For instance, it has been shown that T cells specific for deamidated β cell Ags are present at higher frequencies in T1D patients (20). Moreover, recent work has demonstrated that posttranslational modifications (PTMs) occur in β cell granules, where the N terminus of chromogranin A or IAPP2 peptides fuses with the C terminus of insulin peptide to generate highly immunogenic neoantigens (17, 19). These modifications likely affect the C-terminal MHC anchor residue, resulting in stable binding of peptide in a register that is normally unstable (12, 13, 18, 21). In fact, modifying the MHC anchor residue of InsB9–23 peptide at position 9 from an arginine to a glutamic acid (R22E) makes it highly stimulatory for insulin-reactive T cells, and it has been used to generate pMHC tetramers to track these cells in vivo (22, 23).
Genome-wide association studies have also revealed a strong link between T1D susceptibility and allelic variants in the variable number of tandem repeats (VNTRs) of the insulin promoter region. Individuals carrying the VNTR-III polymorphism express more thymic insulin transcript, which correlates with a 3- to 4-fold protection from developing T1D compared with the VNTR-I polymorphism (24–26). In support of the notion that increased levels of thymic insulin improve central tolerance, NOD mice that transgenically overexpress mouse proinsulin II are protected from spontaneously developing diabetes (27–29). However, it was not determined whether full-body transgenic expression of insulin provides protection through peripheral and/or central tolerance mechanisms (27–29). Thymically derived regulatory T cells (tTregs) play a central role in peripheral tolerance and protection against developing autoimmune diseases, including T1D (30–32). Moreover, studies have shown that Treg frequencies may be lower or functionally altered in humans with T1D and NOD mice (33, 34). In addition, higher-affinity interactions with self-peptide–MHC and TCR have been shown to be required for the expression and maintenance of Foxp3 (35–37). To address the impact of thymic Ag dose and pMHC stability on the selection of insulin-specific T cells, we used TCR retrogenic (Rg) technology (38–40) to generate mice that coexpress either low- or high-affinity insulin-reactive TCRs (on T cells) and either insulin or insulin mimetope R22E (on APCs). Our studies reveal that ectopic insulin Ag expression during thymocyte development of high- or low-affinity insulin-reactive TCRs does not result in the deletion of these T cells; however, all mice were protected from developing T1D long term. This was due in part to the increased number and ratio of Foxp3+ Tregs found in thymus, peripheral lymphoid organs, and pancreas, because ectopic insulin expression in the absence of Foxp3 was not protective. However, ectopic expression of the R22E mimetope promoted negative selection of only high-affinity insulin-specific T cells because of an increase in TCR signaling during thymocyte development. Our data highlight the physiological necessity of stable TCR–pMHC interactions that promote negative selection of autoreactive T cells and the importance of insulin-specific Treg generation in controlling T1D.
Materials and Methods
NOD.CB17-Prkdcscid/J (NOD.scid), NOD.129P2(C)-Tcratm1Mjo/DoiJ (NOD.TCRa−/−), and NOD.Cg-Foxp3sf/DoiJ (NOD.scurfy) mice were obtained from The Jackson Laboratory and maintained at our facility. B6-Tg(Nr4a1-EGFP/cre)820Khog/J (Nur77GFP) mice (37) were originally donated by Kristin Hogquist (University of Minnesota) and, after backcross to the NOD, were crossed with NOD.scid to generate NOD.scid-Nur77GFP mice. All mice were housed in specific pathogen-free conditions. Animal protocols were approved by the Baylor College of Medicine Institutional Animal Care and Use Committee.
Generation of TCR and Ii-80 fusion peptide multicistronic vectors
Previously published insulin-specific TCR retroviral constructs 4-8 and 12-4.1 were modified to include antigenic or control peptides (41). In brief, DNA sequence containing the first 80 aa of the invariant chain (Ii-80) (42) fused with either full-length insulin B, mimetope insulin B R22E, or hen egg lysozyme cDNA was inserted into the TCR vector (Fig. 1A).
TCR expression on HEK-293T cells
Transient transfection of HEK-293T cells was performed as previously described (38–40). In brief, cells were cultured in six-well plates at 2 × 105 cells per well overnight at 37°C. Cells were transfected with TCRαβ-Ii-80 peptide plasmid (1 μg) and CD3 plasmid (1 μg) using 6 μl of TransIT LT1 transfection reagent (Mirus). Cells were harvested 48 h later and stained with CD3 and TCRvβ murine Abs for flow cytometric analysis.
I-Ag7+ M12 B cell hybridoma line was transduced with our vector of interest and ametrine+ cells were purified by FACS. A total of 100,000 TCR transfectants (P2-Insulin TCR and 14H4-HEL) were stimulated with 50,000 FACS purified I-Ag7+ M12 cells. After 24 h, a standard sandwich ELISA was performed using HRP (SA:HRP) complex (GE Health).
Retroviral-mediated stem cell gene transfer to generate Rg mice
Retroviral transduction of murine bone marrow cells was performed as previously described (38–41). Transduced bone marrow cells were collected, washed, and resuspended in PBS with 0.5% FBS. Bone marrow cells were injected via tail vein into sublethally irradiated NOD.scid (300 rad) or NOD.Tcra−/− (500 rad) mice. Mice were monitored for diabetes incidence or analyzed 5–8 wk after bone marrow transplant.
Flow cytometric analysis and intracellular staining
Rg mice were harvested 6–8 wk after adoptive transfer of transduced bone marrow cells; peripheral organs consisting of spleen, thymus, pancreatic lymph nodes (PLNs), and pancreata were harvested from each Rg mouse for analysis. Pancreata were digested with collagenase IV (Worthington, Lakewood, NJ), and single islets were isolated for further analysis as previously described (41). For flow cytometric analysis, murine Abs against the following molecules were used: CD4 (GK1.5), CD8 (53-6.7), TCRvβ (H57-597), CD5 (53-7.3), Foxp3 (FJK-16s), CD69 (H1.3F3), CD73 (TY/11.8), folate-like receptor 4 (FR4; 12A5), Helios (22F6), Ki67 (B56), F4/80 (BM8), CD45R-B220 (RA3-6B2), CD11c (N418), CD11b (M1/70), and I-Ag7 (39-10-8). LSR Fortessa (BD Biosciences) was used for flow cytometric analysis, and collected data were analyzed using FlowJo software.
All analyses were performed using Prism 5 GraphPad Software. All pairwise comparisons were performed using nonparametric Mann–Whitney U test. Group comparisons were done using a two-way ANOVA. Diabetes incidence curves were compared using the log-rank test.
Development of TCR-2A–peptide retroviral constructs
To study the effect of ectopic insulin Ag expression on the development of thymocytes with different TCR affinities, we chose two TCRs that have been previously characterized to possess different biophysical and functional affinity for InsB9–23 (41). The two TCRs selected for this study, relatively high affinity (4–8) and low affinity (12-4.1), were isolated from the pancreatic islets of prediabetic mice, ensuring their diabetic potential and physiological relevance (41, 43, 44). Even though 4-8 and 12-4.1 TCRs have distinct affinities, both cause spontaneous diabetes development after re-expression in vivo, albeit 12-4.1 TCR mice exhibited slower kinetics of disease development (41). To ectopically express insulin epitope, we modified the TCR retroviral constructs to include Ii-80 fused with either the insulin B chain (Ii-80 INS), insulin B chain modified at position 22 (Ii-80 R22E), or hen egg lysozyme (Ii-80 HEL) control peptide connected by a 2A linker (23, 42) (Supplemental Fig. 1A). The Ii-80 sequence of the fusion protein targets the peptide to the late endosomal MHC II compartment, where it is efficiently loaded onto MHC II for presentation (42). The R22E insulin B amino acid substitution allows for the negatively charged glutamic acid to stably interact with the positively charged ninth binding pocket of MHC II I-Ag7, forming a stable insulin–MHC II complex for peptide presentation to insulin-specific TCRs (11, 13, 23). Subimmunogenic treatment of NOD mice with insulin R22E results in the generation of induced Foxp3+ peripheral Tregs (pTregs) and subsequent protection from developing T1D (23). However, the impact of the R22E peptide on Ag presentation within the thymus and subsequent thymocyte development of insulin-reactive T cells has not been determined.
To verify TCR cell surface expression of both 4-8 and 12-4.1 insulin-specific TCRs with each retroviral construct, we transduced HEK 293T cells with each TCR construct, along with the CD3εδγζ chains. All constructs resulted in similar cell surface expression of TCRs (Fig. 1A). Next, we evaluated peptide presentation of each invariant chain fusion peptide by transducing M12.C3 (M12) I-Ag7+ B cell lymphoma line with retroviral constructs. Each transduced M12 cell line was cocultured with a 4G4 thymoma T cell line expressing either an insulin-specific TCR or the control 14H4 HEL11–25–specific TCR. After 24 h of coculture, supernatants were collected and measured for IL-2 secretion by ELISA (Fig. 1B). HEL-specific 4G4 T cells secreted detectable levels of IL-2 only when cocultured with M12 cells transduced with the construct containing Ii-80 HEL, but not Ii-80 INS or R22E fusion peptides (Fig. 1B, left panel). As expected, insulin-specific 4G4s secreted IL-2 in the presence of both Ii-80 INS transduced M12 cells and Ii-80 R22E transduced M12 cells, although response to Ii-80 R22E agonist was increased (Fig. 1B, middle panel). To correlate the concentration of IL-2 secreted by 4G4 T cells stimulated with Ii-80 fusion peptides (INS and R22E) to a soluble peptide concentration, we incubated I-Ag7+ M12 APCs with decreasing concentrations of soluble InsB9–23 peptide, ranging from 100 to 1 μM. The Ii-80 INS fusion peptide stimulated insulin-specific 4G4s to a similar level as did 6–12 μM soluble InsB9–23 peptide, whereas the levels of IL-2 induced by Ii-80 R22E exceeded our highest insulin InsB9–23 peptide concentration (Fig. 1B, right panel). Therefore, our in vitro data indicate that each retroviral vector can lead to simultaneous and efficient expression of an Ag-specific TCR, as well as presentation of Ii-80 fusion peptides by APCs for T cell recognition.
Ectopic expression of insulin B protects low- and high-affinity insulin-specific TCR Rg mice from developing autoimmune diabetes
We next wanted to address, in vivo, whether increasing Ag availability during the development of insulin-reactive thymocytes could influence diabetes onset. To test this, we generated TCR Rg mice by transducing NOD.scid bone marrow with retroviral vectors expressing insulin-specific TCRs 2A-linked to either Ii-80 HEL, Ii-80 INS, or Ii-80 R22E. As a negative control, Rg mice were generated to express 14H4 HEL11–25–specific TCR with Ii-80 HEL. After transduction, bone marrow cells were i.v. transferred into sublethally irradiated NOD.scid recipients. Insulin-specific 4-8 TCR Rg mice expressing either Ii-80 INS or Ii-80 R22E peptides were completely protected from the onset of the autoimmune disease, whereas Rg mice expressing Ii-80 HEL peptide developed diabetes at a similar rate as was previously described for this TCR (41) (Fig. 2A). Disease protection was similarly observed in low-affinity 12-4.1 insulin-specific TCR Rg mice coexpressing Ii-80 INS or Ii-R22E, but not Ii-80 HEL (Fig. 2B). However, due to the ability of both peripheral and thymic APCs to express the fusion peptides, the observed protection could be attributed to either central or peripheral tolerance induction. To investigate potential mechanisms of protection, we next performed the analysis of the thymus and peripheral lymphoid organs.
Insulin B pMHC complex stability determines the fate of insulin-specific thymocytes
To address the mechanism of protection in Ii-80 INS- and Ii-R22E–expressing TCR Rg mice, we harvested thymi and spleen 5–8 wk after bone marrow transfer. All 4-8 TCR Rg mice (Ii-80 INS, Ii-R22E, and Ii-80 HEL) had similar ratios of ametrine+ APCs that are normally found in the thymic and splenic compartments (Supplemental Fig. 1B, 1C). We then performed flow cytometric analysis of the thymi to determine how the presence of peptide-expressing APCs might alter the development of high- (4–8) and low- (12-4.1) affinity insulin-reactive thymocytes. Surprisingly, overexpression of wild type insulin peptide by Ii-80 INS in either high- or low-affinity TCR mice did not change the frequencies nor the numbers of CD4+CD8+ double-positive (DP) and CD4+ single-positive (SP) thymocytes (Fig. 3A–C, top panels). However, expression of strong agonist Ii-80 R22E significantly reduced the frequencies and cell number of CD4 SP 4-8 thymocytes in comparison with Ii-80 HEL (Fig. 3A, 3C, top panels), which was consistent with a significant increase in Annexin V staining, indicating an increase in negative selection (Fig. 3D, left panel). Furthermore, there was a reduction in the ratio of CD4 SP-to-DP cells and CD69hi TCR+ DP thymocytes within 4-8 Rg mice expressing Ii-80 R22E compared with Ii-80 HEL or Ii-80 INS, signifying a block in the transition from DP to SP development in the presence of R22E (Fig. 3E, 3F, left panels) (42). Interestingly, ectopic expression of Ii-80 R22E did not affect the ratio or cell numbers of the low-affinity 12-4.1 thymocytes (Fig. 3A–C, bottom panels). Generally, low-affinity 12-4.1 thymocytes were less sensitive to ectopic expression of INS or R22E, and exhibited a less significant increase in the Annexin V staining compared with 4-8 thymocytes (Fig. 3A, bottom panels, Fig. 3D, right panel). Likewise, we also observed no differences in the CD4 SP-to-DP ratios and frequencies of positively selecting Ii-80 12-4.1 CD69hi TCR+ DP thymocytes (Fig. 3E, 3F, right panels).
To directly assess the strength of TCR signaling in response to exogenous insulin peptides in vivo, we used the Nur77GFP transgenic reporter activated downstream of Erk phosphorylation (30). High-affinity 4-8 DP thymocytes exhibited significantly higher levels of Nur77GFP in response to R22E (Supplemental Fig. 2A, left panel), whereas 4-8 CD4 SP thymocytes had slightly elevated levels of Nur77GFP expression (Supplemental Fig. 2B, left panel), indicating that the increase in negative selection was due to stronger TCR signaling. As another indicator of stronger TCR signals received in R22E-expressing mice, we observed a decrease in levels of TCR expression at the SP stage of thymocyte development (Supplemental Fig. 2C, left panel). These data reinforced the idea that insulin-specific thymocytes survive and undergo positive selection in mice expressing Ii-80 HEL or INS, whereas the few surviving Ii-80 R22E selected T cells downregulate TCR to persist. Similarly to the high-affinity T cells, the selection of low-affinity 12-4.1 insulin-specific thymocytes showed significant increase in Nur77GFP expression in response to Ii-80 R22E compared with Ii-80 INS (Supplemental Fig. 2A, right panel), which was consistent with the corresponding increase in Annexin V staining seen at the SP stage of Ii-80 R22E 12-4.1 Rg mice (Fig. 3D, right panel). However, the TCR and Nur77GFP expression was not significantly different between the groups, in line with our observations that selection of 12-4.1 thymocytes was not affected by the ectopic insulin peptide (Fig. 3A–C, bottom panels, Supplemental Fig. 2B, 2C, right panels). Collectively, our data show that high-affinity thymocytes are more sensitive to overexpression of strong agonists and are deleted in the presence of R22E, whereas ectopic expression of wild type self-antigen had minimal effect on negative selection of either high- and low-affinity T cells.
Subimmunogenic treatment of adult NOD mice with R22E peptide results in the generation of induced Foxp3+ Tregs and protection from T1D onset (23). However, we have shown that insulin-specific thymocytes developing in the presence of Ii-80 R22E peptide are efficiently deleted. To determine whether there is an increase in the development of Foxp3+ Tregs in the few surviving CD4 SP cells in Ii-80 R22E Rg mice, we performed intracellular staining for Foxp3 in TCR Rg mice expressing Ii-80 R22E, HEL, or INS. Surprisingly, there was no increase in Foxp3+ cells within the SP thymocytes of 4-8 or 12-4.1 TCR Rg mice expressing the R22E agonist (Fig. 4). However, there was a small but significant increase in the ratio and number of Foxp3+ CD4 SP Tregs in both 4-8 and 12-4.1 Ii-80 INS mice (Fig. 4). Therefore, our data suggest that increasing insulin availability promotes the development of Foxp3+ insulin-specific T cells, while increasing the stability of the trimolecular complex allows for efficient deletion of insulin-reactive thymocytes.
High- and low-affinity insulin-specific T cells infiltrate the pancreatic islets in the presence of ectopic insulin peptide
To determine whether the insulin-specific T cells survive and persist after thymic emigration in Ii-80 peptide-expressing TCR Rg mice, we analyzed the frequencies of CD4 T cells from the spleen, PLN, and pancreatic islets of 4-8 and 12-4.1 Rg mice expressing Ii-80 HEL, INS, and R22E. T cells were readily observed in the periphery of 4-8 Rg mice that expressed Ii-80 HEL and INS, whereas Ii-80 R22E-expressing 4-8 Rg mice contained virtually no T cells within the spleen and pancreatic islets (Fig. 5). Interestingly, a small percentage of CD4+ T cells could be found within the PLN of mice expressing Ii-80 R22E; however, TCR expression levels were reduced on these T cells, as well as on the few CD4+ T cells found within the spleen and islets, indicating possible compensatory TCR downmodulation in response to strong activation (Supplemental Fig. 3, left panels). In contrast with 4-8 Rg mice, 12-4.1 Ii-80 INS and Ii-80 R22E Rg mice showed a significant reduction in peripheral T cells within the spleen, PLN, and islets (Fig. 5). Furthermore, the TCR expression of 12-4.1 TCR Rg mice decreased in proportion to the potency of the agonist insulin peptide within all organs (Supplemental Fig. 3, right panels). The overall reduction in peripheral 12-4.1 TCR Rg CD4+ T cells in the presence of Ii-80 INS and Ii-80 R22E, but relatively normal thymocyte development, indicates the importance of peripheral tolerance mechanisms for regulation of this low-affinity TCR. Overall, our data show that the relative contribution of central and peripheral tolerance mechanisms to regulation of T cell development and accumulation in periphery is different between high- and low-affinity insulin-specific lymphocytes.
Ectopic expression of insulin increases thymic Tregs
Due to the fact that we observed abundant accumulation of insulin-specific peripheral T cells within the PLNs and islets of Ii-80 INS mice in the absence of autoimmunity, we wanted to assess whether this was due to an increase in regulatory Foxp3+ T cells or perhaps another tolerance mechanism. Analysis of Rg mice revealed that Foxp3+ Tregs were significantly increased in the spleens, PLNs, and islets of Ii-80 INS 4-8 and Ii-80 INS 12-4.1 TCR Rg mice (Fig. 6, Supplemental Fig. 4A). Although there was a significant overall reduction in peripheral CD4+ T cells in 4-8 Ii-80 R22E Rg mice, we did observe a significant increase in the ratio of Foxp3+ T cells within the PLN compared with Ii-80 HEL (Fig. 6, Supplemental Fig. 4A). These data suggest that increased development of Foxp3+ Tregs accounted for the prevention of autoimmunity in the Ii-80 INS mice. However, it is unclear whether similar mechanisms are involved in protection of mice expressing Ii-80 R22E, or whether the robust negative selection induced by R22E is sufficient to protect insulin TCR Rg mice from diabetes.
Several studies indicate a dominant role for Tregs in protecting against the development of autoimmune-mediated diabetes in NOD mice (33, 34, 45). However, the relative influence of thymically derived compared with peripherally induced Tregs in autoimmune diabetes is not fully understood and is difficult to ascertain. Therefore, we wanted to determine whether increasing insulin Ag systemically preferentially leads to the generation of thymically derived or peripherally induced Tregs. The transcription factor Helios and the cell surface molecule Neuropilin-1 (Nrp-1) are two markers that are associated with tTregs and Foxp3 stability (46–49). The overwhelming majority of 4-8 and 12-4.1 TCR Rg mice expressing Ii-80 INS Foxp3+ Tregs expressed Helios within the spleen, PLN, and islets (Supplemental Fig. 4B). It has been suggested that the coexpression of Helios and Nrp-1 on Tregs is a more faithful marker for tTregs (46, 47). We found that similar percentages of peripheral Foxp3 Tregs from 4-8 Ii-80 INS and HEL coexpress Nrp-1 and Helios (Supplemental Fig. 4C). In addition, Foxp3+ Tregs from Ii-80 INS and HEL 12-4.1 TCR Rg mice express similar Nrp1+ Helios+ percentages, and these percentages are nearly the same as an NOD WT mouse (Supplemental Fig. 4C). This indicates both thymically derived and peripherally induced Tregs are present (Supplemental Fig. 4C). We also evaluated the expression of Ki67, a marker of cellular proliferation, to determine whether the increase in Foxp3+ T cells within Ii-80 INS Rg mice is due in part to peripheral expansion. Although Ki67+ Foxp3+ Tregs were found at a slightly higher frequency in the spleens of Ii-80 INS 4-8 Rg mice, the frequencies of Ki67+ Ii-80 INS and Ii-80 HEL Tregs were similar within the PLNs and islets (Supplemental Fig. 4D, left panels). Ki67+ Foxp3+ Tregs were equivalent across all organs in 12-4.1 Rg mice (Supplemental Fig. 4D, right panels). Together these data indicate that ectopic insulin expression promotes the development of both peripheral and thymic Tregs, and the increase in Tregs found within the PLNs and islets is not due to local Ag-induced proliferation.
In addition to Treg induction, self-antigen expression in periphery can also induce a state of nonresponsiveness or anergy (50). Anergic T cells can be identified based on coexpression of FR4 and CD73 (22). When we analyzed the relative frequencies of FR4+ CD73+ T cells in INS and HEL mice, we observed a slight, but not significant, increase in the frequency of FR4+CD73+ anergic T cells in the spleens and islets of Ii-80 INS 4-8 TCR Rg mice, whereas similar ratios of anergic T cells were found in the PLNs (Supplemental Fig. 4E, left panels). The low-affinity 12-4.1 TCR Rg mice expressing Ii-80 INS did not display any relative increase in anergic T cells compared with Ii-80 HEL (Supplemental Fig. 4E, right panels). These data indicate that the dominant peripheral tolerance mechanism was orchestrated by Treg development and not anergy. To test the hypothesis that Tregs are sufficient for the autoimmune protection in Ii-80 INS 4-8 TCR Rg mice, we used NOD.scid.scurfy donor bone marrow, which has a mutation in Foxp3 and cannot lead to the development of functional Tregs. We transferred NOD.scid.scurfy bone marrow transfected with either Ii-80 HEL or INS 4-8 TCR vectors into NOD.scid recipients and monitored mice for diabetes. In contrast with mice generated with NOD.scid bone marrow, where Ii-80 INS 4-8 TCR Rg mice were fully protected from developing autoimmune diabetes (Fig. 2), NOD.scid.scurfy Ii-80 INS 4-8 TCR Rg mice developed accelerated diabetes onset with complete penetrance by 13 wk (Fig. 7). However, there was a significant delay in diabetes onset in comparison with NOD.scid.scurfy Ii-80 HEL mice, suggesting a contribution of another regulatory mechanism responsible for protection of Ii-80 INS mice from diabetes. Together, our data indicate increasing the stability of the trimolecular complex allows for efficient deletion of insulin-reactive thymocytes, whereas increasing insulin availability promotes the development of Foxp3+ insulin-specific T cells, which are maintained in the periphery and efficiently protect mice from autoimmunity.
The tightly controlled processes of positive and negative selection allow for the survival of a robust T cell repertoire that is both nonautoimmune and protective against pathogens (1, 2). For selection to occur, developing thymocytes must interact with an array of thymic APCs (cortical thymic epithelial cells, mTECs, B cells, macrophages, and DCs) that present thymically derived and peripherally transferred antigenic epitopes (3, 6, 7). In our study, we find that ectopic insulin (InsB9–23) and agonist insulin (R22E) expression restricted to bone marrow–derived APCs protect high- and low-affinity insulin-specific TCR Rg mice from autoimmune diabetes. However, in the presence of ectopically expressed InsB9–23, there was no change in the percentage of 4-8 or 12-4.1 DP cells undergoing positive selection or transitioning into CD4 SP developmental stage, suggesting that autoimmune T cells escape negative selection irrespective of Ag concentration presented in the thymus. Our results indicate that increasing insulin Ag availability in the thymus does not increase negative selection of insulin-specific T cells, but may instead promote the development of Foxp3+ Tregs, leading to an increase in pTregs. It is also possible that recirculating Foxp3+ Tregs contribute to the overall increase in thymic Tregs, and further studies will need to be performed to address the relative contribution of tTreg and pTregs. Nevertheless, the increase in negative selection observed in the presence of insulin R22E offers strong evidence that the stability of the TCR–pMHC trimolecular complex is critical in the establishment of central tolerance to insulin Ag.
Studies have shown that bone marrow–derived APCs, including DCs, play a critical role in tTreg development by both AIRE-dependent and -independent mechanisms (7). Foxp3+ tTreg development occurs in a restricted thymic niche regulated in part by the presence of APCs presenting cognate Ag (5, 51). Recent work has also demonstrated that up to 50% of thymic Foxp3+ Tregs at 10 wk of age are actually pTregs that have circulated back to the thymus (52). In our studies, we observe an approximately 3-fold increase in thymic Foxp3+ Tregs at 6–8 wk after bone marrow reconstitution, indicating that the increase in thymic Tregs may be a combination of both recirculating pTregs and tTregs. Importantly, in the presence of the agonist R22E peptide, high-affinity 4-8 insulin-specific thymocytes were efficiently deleted, whereas thymic development of low-affinity 12-4.1 insulin-specific thymocytes remained unchanged. Given the relatively low affinity of 12-4.1 thymocytes and their poor selection, it is likely that the instability of the trimolecular complex has a negative effect on the development of low-affinity self-reactive T cells (43). The increased affinity of the insulin agonist R22E did not appear to favor either positive or negative selection of 12-4.1 T cells; however, there was a significant reduction in the number of peripheral T cells, resulting in limited T cell infiltration of the pancreatic islets. Interestingly, peripheral Ii-80 R22E 12-4.1 TCR Rg T cells did not show higher expression of anergy markers, FR4 and CD73, and it remains to be determined whether the reduction in peripheral T cells is the result of activation-induced cells death or T cell exhaustion.
Recent studies indicate that the inherent instability of the insulin peptide–HLA–TCR trimolecular complex allows for thymic escape of low-affinity autoreactive T cells, which subsequently become activated by recognizing more stable neopeptide Ags (11, 17–19). Indeed, pancreatic islet insulin-containing granules are highly immunogenic compared with artificially synthesized protein (21). Moreover, PTMs can form fusion peptides containing pieces of different proteins, such as insulin and chromogranin A (17). Therefore, formerly ignorant T cells can be activated by the de novo synthesis of antigenic peptides not seen during thymocyte development. However, the role of PTMs in thymocyte development has not been explored. Our studies strongly indicate that the introduction of PTMs during thymocyte development, modeled here by the synthetic insulin R22E Ag, dramatically alters positive and negative selection, and therefore could be a potential therapeutic target.
Peripheral tolerance is maintained by a collection of T cell subsets, including tTregs and pTregs, which arise from the thymus and peripheral secondary lymphoid organs. tTregs mediate their suppressive effects similarly to pTregs through multiple contact-dependent and -independent mechanisms (53). In our study, we find that ectopic insulin presentation by APCs, compared with mice expressing ectopic HEL peptide, promotes the development of insulin-specific Tregs, which mediate suppression of autoimmune diabetes. Similar Helios and Nrp-1 expression found on Foxp3+ pTregs within the Ii-80 HEL and INS groups implies that tTregs and pTregs are readily present in the periphery; however, Ii-80 INS pTregs are not preferentially expanding in the presence of ectopic insulin expression. Therefore, it is likely that both pTregs and tTregs are important in the inhibition of islet-specific T cells. Together, our data suggest that Ag-specific Tregs play a dominant role in preventing autoimmune diabetes. Nonetheless, our results do not exclude a contributing role for anergy in maintaining peripheral tolerance.
Studies have shown that the insulin B epitope binds poorly within the groove of the MHC II molecule I-Ag7 in NOD mice (11). Moreover, HLA-DQ2/DQ8 has structural similarities to I-Ag7 and may inadequately present insulin B in humans, resulting in thymic escape of autoreactive T cells (11, 54, 55). Interestingly, insulin B peptide can bind to the MHC groove in different binding registers that allow different levels of TCR reactivity (11, 13, 21). A recent study found that a modified diabetogenic peptide (R22E) can form relatively stable pMHC complexes and elicit protection from autoimmune diabetes via the induction of Foxp3+ Tregs within the pancreata (23). Our results suggest an important contribution of thymic insulin R22E peptide presentation, where insulin-specific TCR Rg mice efficiently delete InsB9–23–specific thymocytes because of the peptides’ strong agonistic activity. Because Ii-80 R22E expression protected against autoimmune diabetes onset similarly to that of Ii-80 INS, it suggests that ectopic expression of insulin B peptide can influence different T cell fates during thymic selection and alter their pathogenicity based on pMHC stability. Although R22E is not a naturally derived peptide, it allows us to model PTM Ags targeted in T1D.
In conclusion, we demonstrated that by increasing insulin presentation through a retroviral-mediated system in NOD mice, we were able to alter the development of pathogenic T cells with varying reactivity for insulin peptide. Our data highlight the role of cognate Ag availability (InsB:9–23) and peptide stability (insulin R22E) in the establishment of tolerance. First, APC availability of insulin B:9–23 did not delete high-affinity 4-8 TCR or low-affinity 12.4-1 TCRs, but instead promoted the development of Foxp3+ Tregs. Second, presentation of the insulin agonist R22E efficiently deleted high-affinity 4-8 T cells, whereas both 4-8 and 12.4-1 T cells were significantly reduced in the periphery, with limited infiltration into pancreatic islets. These data suggest that the establishment of tolerance mechanisms, such as negative selection and Treg development, is not only dependent on the availability of β cell–specific Ags, but also on the stability of the pMHC during development. Overall, this emerging insight may lead to novel therapeutic strategies to establish central tolerance to insulin, and hence prevent or delay the onset of autoimmune diabetes.
We thank Kristin Hogquist for Nur77GFP mice (30); Natalie Tully and Samuel Blum for mouse maintenance, breeding, and genotyping of mouse colonies; Baylor College of Medicine Cytometry and Cell Sorting Core; the staff of the Center for Comparative Medicine, Alkek Building for Biomedical Research facility for mouse rederivation, veterinary, and technical assistance; the Baylor College of Medicine MHC Tetramer Production Facility for peptides; and the Biology of Inflammation Center and the Diabetes and Endocrinology Section for helpful discussions.
This work was supported by National Institutes of Health Grants K22 AI104761 and R56 DK104903 (to M.L.B.), an American Association of Immunologists Careers in Immunology fellowship (to M.L.B. and T.L.), and the Robert and Janice McNair Foundation (to M.L.B.).
The online version of this article contains supplemental material.
Abbreviations used in this article:
folate-like receptor 4
the first 80 aa of the invariant chain
insulin epitope B:9–23
- MHC II
MHC class II
medullary thymic epithelial cell
pancreatic lymph node
type 1 diabetes
regulatory T cell
thymically derived regulatory T cell
variable number of tandem repeat.
The authors have no financial conflicts of interest.