Female NOD/LtJ mice and BALB/c mice were purchased from the Jackson Laboratories and maintained in the Biological Resource Center at National Jewish Health (Denver, CO). Prediabetic NOD mice used in this study were 12 weeks of age and had normal blood glucose readings before experiments (<250 mg/dl). Mice were considered diabetic when blood glucose levels were >250 mg/dl for two consecutive days. Animal husbandry and experimental procedures were conducted under protocols approved by the Institutional Animal Care and Use Committee of the National Jewish Health.

Fluorescently labeled mAbs for flow cytometry were used as follows: FITC-B220, eFluor 450-F4/80, APC eFluor 780-CD8, Alexa Fluor 700-CD4, APC-Foxp3, and PerCP-Cy5.5-CD44 (eBioScience), and Foxp3/Transcription Factor Fixation/Permeabilization (eBioscience, 501129060). Unlabeled TCR Cβ-specific H57-597 was expressed and purified from the B-cell hybridoma (a gift from Kappler/Marrack lab). All peptides were synthesized by Peptide 2.0 Inc.

The first 20 amino acids (SCNTATCVTHRLAGLLSRSG, SG20) of mouse CGRP peptide (equivalent to KS20 peptide) are similar to the first 20 amino acids (ACDTATCVTHRLAGLLSRSG, AG20) of human CGRP peptide (5). The IA^g7 α and β chains were inserted in the baculovirus transfer vector under P10 and PH promoters, respectively. The β chain contained the wild-type (WT) CGRP SG20 peptide, or a CGRP variant SG20S17E (SCNTATCVTHRLAGLL E RSG), or the IAPP KS20V17E peptide (KCNTATCATQRLANFL E RSS) or the truncation of IAPP KS20 peptide from 1 to 5, KS20△1–5 peptide (TCATQRLANFL E RSS), with a C-terminal flexible linker (GGGSLVPRGSGGGGS) inserted between a signal peptide and the N terminus of the β1 domain as previously described (24, 25). All constructs contained this flexible linker attaching the peptide to the N terminus of the IA^g7 β chain and a stabilizing acid–base leucine zipper attached to the C terminus of the IA^g7 (26). The crystal structure of the IA^g7-KS20 complex showed that the position of 17 amino acids occupied the P9 binding pockets of IA^g7 (data not shown). Hence, we introduced a mutation at position 17 to E (underlined) to enhance the binding between the peptide and IA^g7 without adverse effects on the peptide activity. The soluble IA^g7-SG20, IA^g7-KS20, and IA^g7-KS20△1–5 complexes were produced and purified as previously described (25). Briefly, Hi5 cells were infected separately with IA^g7-SG20, IA^g7-KS20, or IA^g7-KS20△1-5 recombinant baculovirus in spinner flasks at 19°C for 6 days. The supernatants were harvested, and the debris was removed by centrifugation. Soluble IA^g7-SG20, IA^g7-KS20, and IA^g7-KS20△1–5 were purified using zipper-specific 2H11 mAb (27) and coupled with CNBr-activated Sepharose 4B resins (GE Healthcare). As previously described, the purified proteins were biotinylated with the BirA enzyme (28). After biotinylation, the IA^g7-SG20 protein was applied to a Superdex 200 Increase column (GE Healthcare) to remove the aggregate or other contaminating proteins, and peak 2 was collected for the tetramer generation ( Supplementary Figure S1A ). The purified biotinylated protein was analyzed by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) with and without reduction ( Supplementary Figure S1B ). These biotinylated proteins were mixed with streptavidin-PE for 1 h at 4°C and then were purified again with a Superdex 200 Increase column ( Supplementary Figure S1C ). The IA^g7-KS20 tetramers and IA^g7-KS20△1–5 tetramers were produced with the same method.

The plasmids of IA^g7-KS20 and IA^g7-SG20 containing the transmembrane-cytoplasmic tail of the baculovirus gp64 protein were used as templates for mutation. The primers were used for mutation in Supplementary Table S1 . Site-directed mutagenesis experiments were carried out using PfuUltra II fusion DNA polymerase (Agilent Technologies). The PCR reaction system is 50 μL: 5 μL of 10× reaction buffer, 50 ng of DNA template, 1 μL of 10 μM forward and reverse primer, 1 μL of dNTP, 1 μL of PfuUltra II, 3 μL of dimethyl sulfoxide (DMSO), and water added to the final volume of 50 μL. The temperature program used was 1 min at 95°C followed by 18 cycles of 50 s at 95°C, 50 s at 60°C, and 8 min at 68°C. The PCR product was digested with 1 μL DpnI (R0176L, NEB) for 2 h and transformed to competent DH5α.

For antigen-presenting cells (APCs), we used two forms of the M12.C3 B-cell lymphoma, one expressing IA^g7 (M12C3^G7) and the other expressing DQ8 (M12C3-DQ8) (29). T-cell hybridomas (1×10⁵ cells/well) were mixed with an equal amount of APCs and cultured overnight with various concentrations of KS20 or SG20 peptides in a volume of 250 µL. Twenty-four hours later, culture supernatants were then harvested, and secreted IL-2 was measured with a functional assay, following the growth and survival of the HT-2, which is an IL-2-dependent cell line (30). For insect cell surface expression, the different viruses encoding WT KS20, KS20 variants, WT SG20, or SG20 mutants were used to infect mouse ICAM⁺B7⁺ SF9 insect cells (28). Three days later, cells in each well were washed with balanced salt solution (BSS) two times and used as APCs to stimulate BDC-5.2.9 or CGRP-reactive T-cell hybridomas (1×10⁵ cells/well). Twenty-four hours later, supernatants were collected for HT-2 assay.

Briefly, 96-well plates were first coated with 100 μg/ml streptavidin overnight at 4°C. After blocking with 30% fetal bovine serum (FBS) and washing three times with PBS-T (0.1% Tween in phosphate-buffered saline, PBS), 5 μg/mL biotinylated protein IA^g7-KS20, IA^g7-SG20, or IA^g7-SG20C7S was added and incubated for 1 h at 37°C. After washing with PBS-T, 100 μL of different concentrations of mAb LD96.24 was added and incubated for 1 h at room temperature. Then, IgG-alkaline phosphatase (1:30,000) was added and incubated for another hour at room temperature. After five times washing with PBS-T, p-nitrophenyl phosphate (PNPP) substrate solution was added and incubated for 30–60 min, and then, the OD₄₀₅ was measured.

Pancreatic cells were prepared following the protocol as described previously (31). Briefly, freshly isolated NOD pancreases were cut into small pieces and digested in 25 mL of BSS containing 5% (vol/vol) FBS plus 5 μM CaCl₂ and 100 μg/mL collagenase (C9407; Sigma) at 37°C for 15 min. The digested mixture was then washed with 10% FBS in BSS, crushed, and passed through a 100-μm nylon mesh screen. The cells were washed multiple times in 10% FBS in BSS until the supernatant was clear and then resuspended in 10% FBS in spinner-minimum essential medium (S-MEM) (Gibco).

For tetramer staining, we followed the protocol as previously described (28). Briefly, approximately 20 μg/mL MHC II-Peptide tetramers and 1 μg/mL H57-597 were added at 25 μL/well of 96-well U plates and then incubated at 37°C in a 5% CO₂ incubator for 2 h. After tetramer staining and cell surface staining with fluorescent anti-B220, anti-F4/80, anti-CD8, anti-CD44, and anti-CD4 mAbs, cells were fixed and intracellular staining with Foxp3-APC using the eBioscience Foxp3/Transcription Factor Staining Buffer Set (Thermo Fisher Scientific) according to the manufacturer’s instructions. Stained single-cell suspensions were analyzed using a Fortessa flow cytometer running FACSDiva (BD Biosciences, US). FSC 3.0 files were analyzed with Flowjo 10.0 Software.

CGRP-specific T-cell hybridomas used in these studies were generated as described previously (30). Briefly, three female NOD mice were immunized with 10 μg lipopolysaccharide (LPS) and 100 μg CGRP SG20 peptide by intraperitoneal injection (i.p) twice at 2-week intervals. Four days after the final challenge, pancreatic lymph nodes (PLNs) and splenocytes were isolated and re-stimulated with 10 μg/mL CGRP SG20 peptide for 4 days. Cells were purified with lymphocyte separation medium (MP Biomedicals) and cultured in a complete tumor medium (CTM) containing IL-2 for 3 days. T-Cell hybridomas were established by fusion of the purified T-cell blasts to the thymoma BW5147α⁻β⁻ thymoma cells using polyethylene glycol 1450 and cloned by limiting dilution under HAT (Sigma, H0262) selection as described previously (30). For CGRP-specific T-cell screening, T-cell hybridomas were stimulated by M12C3^G7 with or without 1 μg/mL SG20S17E, and supernatants of each well were collected for IL-2 assay 24 h later. The positive wells were CGRP-specific T-cell hybridomas, and they were used for the following experiments.

Total RNA of each CGRP-reactive T-cell hybridomas was isolated, and cDNA was prepared using PrimeScript™ RT Reagent Kit (RR037A, Takara). We first used the FITC-anti-TCR Vβ 1-14 (BioLegend) to stain CGRP-specific T-cell hybridomas to determine TCR Vβ segment usages. Then, TCR Vβ gene fragments were PCR amplified by applying a TCR C region primer of the β chain and a TRBV primer according to the flow results. We used a TCR C region primer of the α chain and a panel of TRAV primers (n=13) to determine TRAV gene segment usage. Upon determination of specific TRAV and TRBV gene segment usages, larger-scale PCR reactions were purified with the Gel and PCR Clean-UP kit (Takara Bio USA, Cat. No. 740609.250). The PCR products were ligated to pCR™2.1 Vector using TA Cloning™ Kit (Invitrogen, Cat. No. K202020). After transforming DH5α cells, the single clone was selected for sequencing. All TRAV and TRBV sequences were analyzed using ImMunoGeneTics (IMGT)/HighV-QUEST (http://www.imgt.org).

Comparisons between two groups were performed with a two-tailed Student’s t-test, and more than two groups were compared by ANOVA. Each experiment was repeated at least once to assess reproducibility. *p < 0.05, **p < 0.01, ***p < 0.001.

CGRP family peptides all share a common motif CXXXXC near the N-terminus both in human and mice ( Figure 1 ) (32). The mouse CGRP and mouse IAPP contain the same CNTATC motif. There is only one amino acid difference between human CGRP and mouse CGRP CXXXXC motifs. The processed CGRP peptides are present in the Ins-secreting β cells and are similar to IAPP, with about 50% amino acid sequence homology. Both peptides are 37 amino acids long and have an N-terminal disulfide bridge between residues Cys2-Cys7. Although IAPP and CGRP peptides have been shown to have distinct functions, they have similar effects on peripheral Ins resistance. Diabetogenic T cells recognize the antigenic peptides derived from β-cell proteins presented by MHC proteins on APCs (33). IAPP and CGRP peptides are processed similarly, and CGRP may cross-activate IAPP-reactive CD4 T cells in the pancreas of NOD mice. The disulfide-containing IAPP KS20 peptide is a target antigen for highly diabetogenic CD4 T cell, BDC-5.2.9 (5, 22). The typical role of IAPP and CGRP in the immunomodulation of CD4 T-cell responses might be linked to the sequence identity of the N-terminal portion of peptides. To confirm our hypothesis, we used the B-cell line M12C3^G7, which expresses IA^g7 to stimulate BDC-5.2.9 in the presence of CGRP or IAPP peptides (29). ICAM⁺B7⁺ SF9 cells expressing IA^g7-linked CGRP or IAPP peptides were also used as APCs to test their ability to stimulate BDC-5.2.9 (29). The results showed that neither CGRP SG20 peptide nor truncated IAPP KS20△1-5 peptide stimulated T-cell BDC-5.2.9 ( Figures 2A, B ). Similarly, we expressed the soluble IA^g7-CGRP proteins and generated IA^g7-SG20 tetramers. The IA^g7-KS20△1-5 tetramers completely lost the binding ability to BDC-5.2.9. Like KS20△1-5, IA^g7-SG20 tetramers did not stain BDC-5.2.9 ( Figure 2C ). These results confirmed that the KS20 disulfide bridge plays an important role in the activation of T cell BDC-5.2.9, but there is no cross-reactivity between IA^g7-SG20 and T cell BDC-5.2.9.

There are seven amino acid variations between mouse IAPP KS20 and CGRP SG20 peptides. Although the disulfide loop of CGRP at the N terminal is very similar to IAPP ( Figure 3A ), the CGRP peptide did not activate the KS20-reactive T cell BDC-5.2.9. Recently, we generated a mAb, LD96.24, which binds IA^g7-KS20 complex but not other IA^g7-peptide complexes (23). We tested if mAb LD96.24 could bind to the IA^g7-SG20 complex. The ELISA results showed that LD96.24 has a weaker binding affinity to IA^g7-SG20 than IA^g7-KS20. After mutating the Cys7 of SG20 to Ser to break the disulfide bridge, there was no binding between LD96.24 and IA^g7-CGRP, which suggested that LD96.24 interacted primarily with the CGRP disulfide loop ( Figure 3B ). However, the same disulfide bridges present in SG20 N terminus were insufficient for BDC-5.2.9 activation. This outcome indicated that the amino acids other than the CXXXXC motif also seemed to contribute to BDC-5.2.9 T-cell stimulation. want to test which amino acids of the CGRP SG20 peptide are critical for BDC-5.2.9 activation. The differences between IAPP KS20 and CGRP SG20 peptides were at amino acid positions 1, 8, 10, 14, 15, and 17 ( Figure 3A ). Variations of these amino acids can reveal the key amino acids of IAPP and CGRP peptides affecting BDC-5.2.9 activation. We used ICAM⁺B7⁺ SF9 cells expressing IA^g7 linked with WT KS20, and KS20 mutants of K1S, KS20 A8V, Q10H, N14G, F15L, and V17S, based on SG20 sequence as APCs to stimulate BDC5.2.9 T cells (22). The results showed that A8, Q10, and F15 were necessary for BDC-5.2.9 activation aside from the disulfide bond. The amino acid at these positions may be involved in TCR binding. Meanwhile, K1, N14, and V17 had no effects on T-cell activation. These amino acids could either be anchor residues, whose side chains were buried in the peptide-binding pockets of the IA^g7 (34) or not in the TCR binding footprint. We also mutated CGRP SG20 to the KS20 amino acids at the same positions, including SG20 V8A, H10Q, and L15F, to see which amino acids of SG20 mutation could restore the stimulation of BDC-5.2.9. The results confirmed that amino acids of A8, Q10, and F15 were essential for T-cell BDC-5.2.9 stimulation ( Figure 3C ). However, none of these single mutations alone were able to stimulate BDC-5.2.9. The SG20 S1K, G14N, and S17V mutations were also not sufficient to restore the stimulation of SG20 to BDC-5.2.9. The data explained why SG20 did not activate T cell BDC-5.2.9.

The IA^g7-SG20 proteins were biotinylated and incorporated into phycoerythrin (PE) streptavidin fluorescent tetramers. We used IA^g7-SG20 tetramers to investigate whether CGRP-reactive CD4 T cells are present in the pancreas of prediabetic and diabetic NOD mice. IA^g7-KS20 tetramers and IA^g7-Ins B:9-23 P8G tetramers were used as positive controls (28) because these Ins and KS20-reactive pathogenic T cells were known to be present in the pancreases of prediabetic NOD mice (22, 26, 31). We examined the IA^g7-SG20 tetramer⁺CD4⁺CD44^high T cells in the pancreases and PLNs of 12-week-old prediabetic NOD mice and diabetic NOD mice. Pancreatic and PLN lymphocytes were analyzed by flow cytometry, first gating on live, B220⁻, F4/80⁻, CD8⁻, CD44^high, and CD4⁺ T cells; then, the tetramer positive cells were examined as previously described (31). The results showed that the high avidity SG20 tetramer-positive cells were easily detected in the CD44^high CD4⁺, but not CD8⁺ T cells with the IA^g7-SG20 tetramer from pancreases of prediabetic and diabetic NOD female mice ( Figure 4A ). The percentages of CGRP-reactive T cells in the pancreas of diabetic mice were significantly higher than those in prediabetic NOD mice ( Figure 4B ). There are fewer CGRP-reactive CD4 T cells in PLNs than in the pancreas of prediabetic and diabetic NOD mice. The percentages of CGRP-reactive T cells were significantly higher in the PLNs of diabetic than in prediabetic NOD mice ( Supplementary Figures S2A, B ). Using the same gating strategy, KS20 and Ins B:9-23 P8G-reactive T cells in the pancreas of diabetic NOD mice were recognized as positive controls ( Supplementary Figure S3 ). CGRP-reactive CD4 T cells accumulated in the pancreas of prediabetic and diabetic NOD mice, which indicated that CGRP is a potential autoantigen in NOD mice. In the meantime, we did not find any CGRP-reactive T cells in BALB/c mice. We found that CGRP-specific regulatory T cells (Tregs) exist in the pancreas of some prediabetic and diabetic NOD mice ( Supplementary Figure S4 ). Tregs play an important role in suppressing T1D by controlling autoreactive T cells (35). The CGRP-reactive T cells can be divided into Foxp3⁺ and Foxp3⁻ T cells, which may play different functions in the development of T1D.

To further characterize CGRP-reactive T cells, we generated 20 CGRP-reactive T-cell hybridomas (20% of total hybridomas) from CGRP peptide immunized NOD mice. Most of these T-cell hybridomas could be stained by IA^g7-SG20S17E tetramers but not IA^g7-KS20V17E tetramers ( Figure 5A ), which indicated that these SG20-specific T cells could not cross-react with the KS20 peptide. Due to the similarity of SG20 and KS20 N termini, we evaluated if the SG20-specific T-cell clones 22, 37, and 155 respond to IAPP KS20 peptide with different TCR V segment usages. ICAM⁺B7⁺ SF9 expressing WT SG20, KS20, or SG20S17E peptides were used for T-cell stimulation assay. The results affirmed that the 22, 37, and 155 CGRP T cell hybridomas did not cross-react with the KS20 peptide. Both WT and S17E mutations of SG20 peptides were able to stimulate these T cells; however, S17E significantly enhanced T-cell activation as expected ( Figure 5B ). We then used different concentrations of synthetic peptide SG20S17E to stimulate these T-cell hybridomas, and the results showed that 0.1–1 μg/mL SG20S17E could activate all of them ( Figure 5C ). These results confirmed that the CGRP-reactive T cells are present in NOD mice, and IA^g7-SG20S17E tetramer is an efficient tool to detect CGRP-reactive T cells in vivo.

Sequencing analysis of these CGRP-reactive T-cell hybridomas showed that TRAV13 and TRBV13 were the most utilized TCR V gene segments, being expressed in 69% and 54% of all CGRP-reactive T-cell hybridomas, respectively ( Figure 6A ). The complementarity-determining region (CDR) 1 and CDR2 loops of the TCR β chain contact the MHC alpha-helices, while the hypervariable CDR3 regions interact mainly with the peptide (29). TCRAV and TCRBV gene segment usages of the CGRP-responsive T-cell hybridomas are summarized in Figure 6B . In both TCR α and TCR β chains, CDR3 loops have the highest sequence diversity and are the principal determinants of receptor binding specificity. After alignment of all these TRAV13 CDR3 and TRBV13 CDR3 amino acids sequences, we found the common amino acid sequence motifs in CDR3 regions. For example, 131 and 146 T-cell hybridomas share the same TRAV and TRBV usage, while CDR3 sequences of TRBV are different. 60, 155, and 160 TCRs share the same TRAV usage, while TRBV usage is different ( Supplementary Figure S5A ). 131,134, and 135 TCRs share the same CDR3 loops in TCR Vβ regions. Gln in all CDR3 regions of TRBV13 and Asn in the most CDR3 region of TRAV13 are highly conserved, indicating that these amino acids may interact with SG20 peptide similarly ( Supplementary Figure S5B ).

The high sequence homology between SG20 and human AG20 suggested that AG20 peptides may cross-react with these CGRP-reactive T cells ( Figure 1 ). We stimulated these T-cell hybridomas with M12C3^G7 loading with different concentrations of human AG20 peptide. The results confirmed that the human CGRP AG20 peptide could be presented by mouse IA^g7 molecules and cross-react with these CGRP-reactive T-cell hybridomas ( Figure 7A ). HLA-DQ8 molecules not only share structural similarities with the mouse homolog IA^g7 but can also cross the species barrier and functionally replace IA^g7 molecules to stimulate diabetogenic T cells and produce diabetes (29). Meanwhile, we tested if HLA-DQ8/AG20 could activate these mouse CGRP-reactive T-cell hybridomas. The M12C3-DQ8 cells incubated with different concentrations of human AG20 and SG20 were used to stimulate CGRP-reactive T-cell hybridomas. HLA-DQ8/AG20 did not activate the mouse CGRP-reactive T-cell hybridomas. This implied that MHC molecules were crucial for CGRP T-cell activation besides the disulfide loops ( Figure 7B ).