A recent discovery effort resulted in identification of novel splice variant and secretory granule antigens within the HLA class I peptidome of human islets and documentation of their recognition by CD8+ T cells from peripheral blood and human islets. In the current study, we applied a systematic discovery process to identify novel CD4+ T cell epitopes derived from these candidate antigens. We predicted 145 potential epitopes spanning unique splice junctions and within conventional secretory granule antigens and measured their in vitro binding to DRB1*04:01. We generated HLA class II tetramers for the 35 peptides with detectable binding and used these to assess immunogenicity and isolate T cell clones. Tetramers corresponding to peptides with verified immunogenicity were then used to label T cells specific for these putative epitopes in peripheral blood. T cells that recognize distinct epitopes derived from a cyclin I splice variant, neuroendocrine convertase 2, and urocortin-3 were detected at frequencies that were similar to those of an immunodominant proinsulin epitope. Cells specific for these novel epitopes predominantly exhibited a Th1-like surface phenotype. Among the three epitopes, responses to the cyclin I peptide exhibited a distinct memory profile. Responses to neuroendocrine convertase 2 were detected among pancreatic infiltrating T cells. These results further establish the contribution of unconventional antigens to the loss of tolerance in autoimmune diabetes.
Introduction
Type 1 diabetes is a chronic disease in which pancreatic β-cells are selectively destroyed through autoimmune attack. Although CD8+ T cells dominate the pancreatic immune infiltrates (1), the strong genetic association of type 1 diabetes with susceptible HLA class II alleles and the diagnostic importance of islet-specific antibodies both point to a significant role for CD4+ T cell responses in this disease (2–4). Numerous studies have sought to define the antigens and epitopes recognized by autoreactive T cells in subjects with type 1 diabetes (5). In general, published studies of CD4+ and CD8+ T cell responses have focused on a limited number of “classical” islet antigens, such as preproinsulin (PPI), GAD (GAD65), islet antigen (IA)-2, and zinc transporter 8 (ZnT8), all of which are known to be targeted by autoantibodies. Other work has identified CD4+ T cell epitopes within additional islet-associated proteins that are not standard antibody targets, such as islet-specific glucose-6-phosphatase catalytic subunit-related protein (IGRP) and islet amyloid polypeptide (IAPP) (6,7). T cells that recognize epitopes derived from all of these antigens have been detected in the peripheral blood of subjects with type 1 diabetes with use of HLA class II tetramers (8), and, in some cases, those specificities have been observed among islet-infiltrating T cells (9,10). However, questions remain about the importance and relative immunodominance of this wide array of conventional epitopes. Furthermore, the epitope specificity for the majority of islet-infiltrating T cells that have been cataloged remains unknown (9), leaving open the possibility that additional disease-relevant antigens are yet to be discovered.
It is increasingly appreciated that, as part of the disease process, pancreatic β-cells undergo inflammatory and metabolic insults, which trigger endoplasmic reticulum stress in β-cells (11,12). These insults have been shown to promote the formation of neoepitopes, including enzymatically modified antigens (7,13,14), oxidized antigens (15,16), hybrid peptide sequences (17), and novel polypeptides formed through translation of alternative open reading frames (18). The formation of such neoepitopes has the potential to increase immune recognition of β-cells, eliciting additional inflammation and creating a feed forward loop of dysfunction and destruction (19). Furthermore, it was recently shown that alternative mRNA splicing represents an additional pathway for generating novel antigenic sequences that are underrepresented in the thymus and regarded as nonself. The idea that neoepitope-specific T cells more readily escape negative selection remains an open question, but a recent study demonstrated that a T cell receptor that recognizes a deamidated GAD65 peptide was more readily selected than a T cell receptor that recognizes the corresponding wild-type peptide (20). Recent work demonstrates that a distinct alternative mRNA splicing signature can be induced in β-cells by inflammatory stress (21). Combining a rich HLA class I peptidomics data set obtained from the ECN90 β-cell line with RNA-sequencing data from primary human islets (either untreated or exposed to inflammatory cytokines) led to the identification of tissue-specific mRNA splice variants and other novel secretory granule antigens that are present within human islets (22,23). CD8+ T cells that recognize unique peptide sequences from these proteins were detectable within peripheral blood and pancreatic infiltrates by tetramer staining. Importantly, pancreas-infiltrating T cells reactive to three of these novel epitopes were observed to be enriched in individuals with type 1 diabetes (23). CD8+ T cells reactive to the murine orthologs of granule antigens were also observed in the islet infiltrates of prediabetic NOD mice and were diabetogenic on in vivo transfer into NOD.scid mice, with a potency equivalent to that of PPI-reactive CD8+ T cells.
In light of these compelling findings, we sought to investigate CD4+ T cell responses directed against unique splice variant and secretory granule protein–derived peptides. We screened a library of peptides to identify novel CD4+ T cell epitopes restricted by the prevalent type 1 diabetes–susceptible HLA-DRB1*04:01 allele. We then applied a systematic approach to identify novel immunogenic peptides. We next used the corresponding HLA class II tetramers to evaluate the frequency and phenotype of T cells reactive to these novel epitopes in the peripheral blood of subjects with type 1 diabetes and HLA-matched control subjects. Collectively, our work supports the relevance of these novel antigens and epitopes in HLA-DRB1*04:01+ subjects with type 1 diabetes.
Research Design and Methods
Human Subjects
The study was approved by Benaroya Research Institute’s institutional review board (protocol no. IRB07109). All samples were obtained under approved research protocols and with written informed consent. For epitope-defining studies, peripheral blood was collected from 15 individuals with type 1 diabetes (Supplementary Table 1), immediately processed, and used for in vitro T cell assays. Peripheral blood was also collected and cryopreserved from 10 additional individuals with type 1 diabetes (Supplementary Table 2) and 8 age-matched healthy control subjects (Supplementary Table 3). Cryopreserved samples were thawed in batches for direct tetramer staining. All subjects carried at least one HLA-DRB1*04:01 allele. Subject characteristics are summarized in Supplementary Tables 1–3. Live pancreas slices from Network for Pancreatic Organ donors with Diabetes (nPOD) donor 6472 (Supplementary Table 4) were distributed through the nPOD Pancreas Slice Program under approved research protocols after written consent was obtained for research from organ donor families.
Epitope Prediction and Peptide Synthesis
Putative epitopes from splice variant junction regions and secretory granule proteins were selected with use of a prediction algorithm as previously described (24,25). Briefly, we calculated motif scores by multiplying coefficients corresponding to each anchor residue for all possible core 9-mers within each unique junction region and peptides spanning each secretory granule protein of interest. All peptides with plausible binding motifs were synthesized (Sigma-Aldrich, St. Louis, MO). A more comprehensive peptide set was synthesized for the secretory granule protein proprotein convertase subtilisin/kexin type 2 (PCSK2). As summarized in Supplementary Table 5, a total of 145 peptides were synthesized. All peptides were dissolved in DMSO to a stock concentration of 20 mg/mL and used for binding assays. Those with detectable binding were subsequently used for tetramer production and T cell studies.
Peptide Binding to HLA-DRB1*04:01
Binding was assessed through a competition assay, as previously described (26). Briefly, increasing concentrations of each peptide were incubated in competition with a biotinylated reference influenza hemagglutinin peptide (HA306-318) at 0.02 μmol/L in wells coated with DR0401 protein. After washing, residual biotin–HA306-318 was detected with use of europium-conjugated streptavidin (PerkinElmer, Waltham, MA) and quantified with a VICTOR2 D time-resolved fluorometer (PerkinElmer). Curves were simulated with Prism software (version 7.0; GraphPad Software, San Diego, CA) and IC50 values calculated as the concentration needed to displace 50% of the reference peptide.
Preparation of HLA Class II Protein and Tetramer Reagents
DR0401 protein was purified from insect cell cultures as previously described (27,28). Monomers were loaded with 0.2 mg/mL peptide at 37°C for 72 h in the presence of 0.2 mg/mL n-Dodecyl β-D-maltoside and 1 mmol/L Pefabloc (Sigma-Aldrich). Individual peptide-loaded monomers for all candidate epitopes were conjugated into tetramers with R-PE streptavidin (Invitrogen, Waltham, MA) at a molar ratio of 8:1 and used to stain in vitro expanded T cells. For direct ex vivo staining, R-PE streptavidin–conjugated CCNI-00819–37 (CCNI-008 p4, HTATPLDFLHIMDSSQLIH), phycoerythrin (PE)-CF594 streptavidin (BD)–conjugated PCSK2243–257 (PCSK2 p23, QPFMTDIIEASSISH), PE-Cy5 streptavidin (BD)–conjugated UCN358–72 (UCN3 p5, SFHYLRSRDASSGEE), PE-Cy7 streptavidin (SouthernBiotech, Birmingham, AL)–conjugated PPI76–90 (SLQPLALEGSLQKRG), and BV421 streptavidin (BioLegend, San Diego, CA)–conjugated influenza MP97–116 (MP p54, VKLYRKLKREITFHGAKEIS) tetramers were prepared to provide multicolor, single tube tetramer staining.
In Vitro Tetramer Assays and T Cell Clone Isolation
Peripheral blood mononuclear cells (PBMCs) were isolated by Ficoll underlay, counted, and either frozen in 7% DMSO and subsequently rethawed or immediately used for assays. PBMCs were suspended in T cell medium (RPMI, 10% pooled human serum, 1% penicillin-streptomycin, 1% l-glutamine) at 4 × 106 cells/mL and stimulated with peptides (20 µg/mL total) in 48-well plates for 14 days, with addition of medium and IL-2 starting on day 7. Cells were stained with individual PE-labeled tetramers for 75 min at 37°C, followed by CD3 BV510 (BioLegend), CD4 BUV395 (BD Biosciences, Franklin Lakes, NJ), CD25 FITC (BioLegend), and a combination of CD14 PerCP-Cy5.5 (Thermo Fisher Scientific, Waltham, MA), CD19 PerCP-Cy5.5 (BioLegend), and Viaprobe (BD Biosciences) for 15 min at 4°C, run on an LSR II (BD), and analyzed with FlowJo (Tree Star, Ashland, OR). Clones were isolated by sorting of single tetramer-positive CD4+ T cells with a FACSAria (BD) and expanded in 96-well plates in the presence of 1 × 105 irradiated PBMCs and 2 µg/mL phytohemagglutinin (PHA) (Remel, Lenexa, KS), with addition of media and IL-2 starting on day 10.
T Cell Clone Maintenance and Characterization
Clones specific for each immunogenic peptide were maintained in RPMI medium supplemented with 10% human serum and restimulated as needed with 2 µg/mL PHA. For assessment of specificity, T cell clones were restained with HLA class II tetramers or assayed for assessment of peptide-specific proliferation as follows: 104 T cells/well were plated with 1 × 105 irradiated DR0401+ PBMCs and stimulated in triplicate with 10 µg/mL peptide. After incubation for 48 h at 37°C and pulsing with medium containing 3[H]-thymidine (1 µCi/well), incorporation was measured 18 h later with a scintillation counter.
Ex Vivo Tetramer Staining
Ex vivo tetramer staining and enrichment were accomplished with use of previously published protocols (29,30). A total of 20 million PBMCs were thawed and treated with dasatinib for 10 min at 37°C. PBMCs were then stained with 4.5 µL CCNI-PE, 8 µL PCSK2-PE-CF594, 4.5 µL UCN3-PE-Cy5, 4.5 µL PPI76–90–PE–Cy7, and 4.5 µL influenza MP97–116–BV421–labeled tetramers at room temperature for 90 min. Cells were washed and enriched as previously described (29,30), through incubation with anti-PE and anti-Myc magnetic beads (Miltenyi Biotec, Bergisch Gladbach, Germany) (an Myc tag was incorporated in the influenza MP97–116 tetramer) at 4°C for 20 min and washed again, and a 1/100th fraction was saved for antibody staining (“Pre”). The other fraction was passed through an MS Column (Miltenyi Biotec). Bound or PE-, PE-Cy5–, PE-CF594–, PE-Cy7–, or BV421-labeled cells were flushed and collected. Both enriched (Post) and nonenriched (Pre) fractions were labeled with SYTOX Green, anti–CD4-V500 (BD), anti–CD45RA-AF700 (BD), anti–CXCR3-BV785 (BioLegend), anti–CCR4-BV605 (BioLegend), anti–CCR7-BV650 (BD), anti–CD14-FITC (BioLegend), and anti-CD19-FITC (BioLegend). Samples were run on a BD LSR II flow cytometer, and data were analyzed with FlowJo software, version 10. Single T cells (defined as a SYTOX negative and CD4+/CD14−/CD19−) were gated to exclude cells that were positive for more than one tetramer fluorophore. Each tetramer-positive T-cell population was then gated to determine the number of positive events and analyzed for surface receptor expression. The frequency (F) of epitope-specific T cells per million CD4+ T cells was calculated as follows: F = (1,000,000 × tetramer-positive events from enriched tube)/(100 × number of CD4+ T cells from the Pre fraction). T-cell lineages were first assigned based on CD45RA and CCR7 expression: naive (CD45RA+CCR7+), central memory (CD45RA−CCR7+), effector memory (CD45RA−CCR7−), and terminal effectors (CD45RA+CCR7−). The memory cell populations were pooled and subdivided into Th1-like (CXCR3+, CCR4−, and CCR6−), Th2-like (CXCR3−, CCR4+, and CCR6−), Th17-like (CXCR3−, CCR4+, and CCR6+), Th1/17-like (CXCR3+, CCR4−, and CCR6+), Th1/2-like (CXCR3+, CCR4+, and CCR6−), or Th1*-like (CXCR3+, CCR4+, and CCR6+) lineages (31–33).
Islet-Derived T-Cell Assays
Cell lines from islet-infiltrating T cells were recovered and expanded from nPOD donor 6472 (see Supplementary Table 4). For generation of lines, 150 µm live slices of pancreas were digested with 1 mg/mL collagenase-P (Sigma-Aldrich) in Hanks’ balanced salt solution with Mg2+/Ca2+ (Thermo Fisher Scientific) for 10 min in a 37°C water bath with vigorous shaking. Digestion was confirmed by visual inspection. Digested tissue was washed and placed in medium (AIM-V supplemented with 2 mmol/L l-glutamine, 5 mmol/L HEPES, 100 units/mL penicillin, and 100 μg/mL streptomycin, 0.1 mmol/L of each nonessential amino acid, and 1 mmol/L sodium pyruvate (all from Lonza, Walkersville, MD) with 5% heat-inactivated human male AB serum (Access Biologicals, Vista, CA). Individual islets were handpicked and placed in culture with irradiated PBMCs as feeders, a blocking anti-Fas antibody (1 µg/mL; eBiosciences, now Applied Biosystems, Santa Clara, CA); anti–PD-1 antibody (1 µg/mL; BD Biosciences), and mifepristone (100 nmol/L; Invitrogen). After 12 days of culture, T cell lines were collected and further expanded in culture and then phenotyped by flow cytometry for T cell subsets. Epstein-Barr virus-transformed lymphoblastoid B-cell lines (B-LCL) were generated from autologous splenocytes. Responsiveness was assessed as previously described (17). Briefly, irradiated autologous B-LCL (generated as previously described [9]), were plated (20,000/well) after pulsing with peptides at 50 µg/mL, no peptide, or DMSO volume control and then cultured with T cells (50,000/well) and anti-CD28 (5 µg/mL) for costimulation. After 48 h of coculture, responses were detected by IFN-γ ELISA. The islet T cell line previously reactive with PCSK2 peptide was restimulated with autologous B-LCL pulsed with the peptide and expanded in culture. At rest, the islet T cell line (30,000 viable cells per well) was challenged with autologous B-LCL pulsed with individual peptides and tested for cytokine secretion by IFN-γ ImmunoSpot (all reagents from CTL) after 24 h of coculture.
Statistics
All statistical analysis was performed with GraphPad Prism 9.2.0. The tests used (as appropriate) included unpaired t tests, Wilcoxon signed rank test, Mann-Whitney U tests, and Kruskal-Wallis test with Dunn multiple comparison test. P values <0.05 were considered significant. Marginal P values ≤0.1 are also provided, when applicable. Biological sex was considered as a variable, but no significant associations were observed.
Data and Resource Availability
The data sets generated or analyzed during the current study are available from the corresponding author on reasonable request.
Results
Identification of Immunogenic Splice Variant and Secretory Granule Peptides
Through combined human islet transcriptomics and β-cell HLA peptidomics data sets, we previously identified a set of candidate antigens consisting of 10 mRNA splice antigens (cyclin I splice variant CCNI-008, GAD2 splice variant GAD2-003, guanine nucleotide binding protein variants GNAS-002 and GNAS-036, islet amyloid polypeptide variant IAPP-002, protein tyrosine phosphatase receptor type N variant PTPRN-021, receptor-type tyrosine-protein phosphatase N2 variant PTPRN2-005, RNA exonuclease 2 variant REXO2-020, secretogranin-5 variant SCG5-009, and solute carrier family 30 member 8 variant SLC30A8-002), peptide splice antigens (an IAPP/IAPP cis-spliced antigen and a neuropeptide Y [NPY]/SCG5 trans-spliced antigen), and three conventional secretory granule proteins (SCG5, urocortin-3 [UCN3], and PCSK2) (23). To facilitate efficient screening of the large number of possible peptides from these candidate proteins, we applied a previously described algorithm to identify peptides with predicted HLA-DRB1*04:01 binding motifs (24,25). To predict possible motifs, we considered each unique splice junction within CCNI-008, GAD2-003, GNAS-002, GNAS-036, IAPP-002, PTPRN-021, PTPRN2-005, REXO2-020, SCG5-009, and SLC30A8-002 and the full sequences of SCG5, UCN3, PCSK2, cis-spliced IAPP/IAPP, and trans-spliced NPY/SCG5 (also including N-terminal and C-terminal flanking residues to generate a minimum peptide length of 15). We identified a total of 145 candidate peptides (Supplementary Table 3). These were synthesized and tested for their ability to bind HLA-DRB1*04:01. In total, 35 peptides had detectable binding to HLA-DRB1*04:01 (Supplementary Table 3). The reduced fraction of peptides with confirmed binding was expected because we chose a lower cutoff value than used in our previous study (34). For these 35 peptides, we next assessed their immunogenicity by stimulating CD4+ T cells from PBMCs of HLA-DRB1*04:01+ subjects with type 1 diabetes and staining with each of the corresponding HLA class II tetramers after 2 weeks of in vitro expansion. Among the peptides tested, CCNI-00819–37 (CCNI), PCSK2243–257 (PCSK2), and UCN358–72 (UCN3) were determined to be immunogenic, based on their ability to elicit a detectable population of tetramer-positive CD4+ T cells after expanding T cells from subjects with type 1 diabetes (Fig. 1A). In performing T cell expansions in a total of 15 donors with type 1 diabetes (Supplementary Table 1), 5 showed positive staining for CCNI, 4 showed positive staining for PCSK2, and 8 showed positive staining for UCN3. In addition, we were able to isolate T cell clones for these epitopes (by sorting and expanding single tetramer–positive T cells), which proliferated robustly in response to their cognate peptide but not to an irrelevant peptide (Fig. 1B). The sequences and predicted minimal motifs for these peptides are summarized in Table 1. Among these three peptides, the CCNI splice variant peptide is a neoepitope with a sequence that differs substantially from the corresponding conventional CCNI sequence (sharing only 57.9% identity), resulting in a distinct predicted binding register (Table 2). The PCSK2 and UCN3 peptides are conventional epitopes. The remaining 32 peptides were not advanced for further study either because they failed to elicit measurable responses in more than one subject (29 peptides) or because the tetramer staining pattern was atypical and could not be confirmed by isolating a peptide-responsive T cell clone (3 peptides).
Peptide . | Amino acid sequencea . | Protein source . | IC50 (µmol/L)b . | Response ratec . |
---|---|---|---|---|
CCNI-00819–37 | HTATPLDFLHIMDSSQLIH | Cyclin I splice variant | 7.2 | 90% |
PCSK2243–257 | QPFMTDIIEASSISH | Neuroendocrine convertase 2 | 0.07 | 60% |
UCN358–72 | SFHYLRSRDASSGEE | Urocortin-3 | 0.94 | 100% |
Peptide . | Amino acid sequencea . | Protein source . | IC50 (µmol/L)b . | Response ratec . |
---|---|---|---|---|
CCNI-00819–37 | HTATPLDFLHIMDSSQLIH | Cyclin I splice variant | 7.2 | 90% |
PCSK2243–257 | QPFMTDIIEASSISH | Neuroendocrine convertase 2 | 0.07 | 60% |
UCN358–72 | SFHYLRSRDASSGEE | Urocortin-3 | 0.94 | 100% |
The best predicted minimal epitope is shown in boldface type. Secondary motifs are underlined.
IC50 represents the peptide concentration that displaces one-half of the reference peptide.
Ex vivo responses were examined in a total of 10 subjects with established type 1 diabetes. The response rate (percentage of subjects with a detectable population of tetramer-positive T cells) is listed for each novel epitope.
Peptide . | Amino acid sequencea . |
---|---|
CCNI-00819–37 | HTATPLDFLHIMDSSQLIH |
CCNI143–161 | HTATPLDFLHIFHAIAVST |
Peptide . | Amino acid sequencea . |
---|---|
CCNI-00819–37 | HTATPLDFLHIMDSSQLIH |
CCNI143–161 | HTATPLDFLHIFHAIAVST |
The best predicted minimal epitope is underlined. Conserved amino acids appear in boldface type.
Novel Epitope-Specific CD4+ T Cells Are Detectable in Subjects with Diabetes
We next asked whether T cells that recognize these three novel epitopes are detectable in the peripheral blood of individuals with type 1 diabetes (Supplementary Table 1). We directly assessed the frequency and surface phenotype of epitope-specific T cells using a multiplex HLA class II tetramer staining approach that allows ex vivo enrichment and detection of multiple tetramer specificities in a single peripheral blood sample and costaining with cell surface marker antibodies (10). The flow cytometry panel included tetramers corresponding to three novel immunogenic peptides: CCNI, PCSK2, and UCN3 plus a PPI76–90 reference tetramer that was previously reported as an immunodominant self-epitope (8). Each of the three novel tetramers was effective at labeling a distinct population of T cells (Fig. 2A shows a positive staining result for each tetramer), albeit with lower fluorescence intensity for CCNI. Staining above the limit of detection (previously reported as 1 cell per million for direct HLA class II tetramer staining [35]) was observed in 9 of 10 subjects for CCNI, 6 of 10 subjects for PCKS2, 10 of 10 subjects for UCN3, and 10 of 10 subjects for PPI (Fig. 2B). Individual subjects had variable frequencies of T cells specific for these epitopes (Fig. 2C). For example, subjects T1D no. 1 and no. 6 had comparatively high frequencies of PPI- and PCSK2-specific T cells and lower frequencies of CCNI- and UCN3-specific T cells, whereas T1D no. 3 had high frequencies of CCNI- and PCSK2-specific T cells and lower frequencies of UCN3- and PPI-specific T cells.
T Cells Specific for Novel Epitopes Have Distinct Phenotypes
To draw inferences about T cell lineages, we used cell surface marker expression to characterize tetramer-positive T cells that recognized these novel specificities. The cell surface markers present in our panel included CD45RA and CCR7, which we used to classify T cells as naive, central memory (TCM), effector memory (TEM), and terminal effectors (TEMRA) (Fig. 3A). To draw inferences about the cell subset distribution for T cells that recognized these novel specificities, we used CXCR3, CCR4, CXCR5, and CCR6 expression to define T helper subsets using established lineage definitions: Th1-like (CXCR3+, CCR4−, and CCR6−), Th2-like (CXCR3−, CCR4+, and CCR6−), Th17-like (CXCR3−, CCR4+, and CCR6+), Th1/17-like (CXCR3+, CCR4−, and CCR6+), Th1/2-like (CXCR3+, CCR4+, and CCR6−), or Th1*-like (CXCR3+, CCR4+, and CCR6+) lineages (31–33) (Fig. 3B). For each novel epitope and PPI, approximately one-half of the T cells exhibited an apparently naive phenotype. We observed apparent differences in the memory distribution of CCNI-specific T cells as compared with PCSK2, UCN3, and PPI. In particular, CCNI had a distinct memory profile, with a significantly lower proportion of naive cells than UCN3 and that trended lower than PCSK2 (Fig. 3C). Subset analysis suggested that CCNI-, PCSK2-, and PPI-specific memory T cells all exhibited a predominantly Th1-like surface phenotype. In contrast, UCN3 had a distinct cell subset distribution, with a significantly lower proportion of Th1-like cells than CCNI (P = 0.04) and proportion of Th2-like cells that was significantly higher than PCSK2 (P = 0.03) and trended higher than CCNI (P = 0.1) (Fig. 3C). These results suggest that UCN3-specific T cells may exhibit a different functional phenotype than CCNI-, PCSK2-, and PPI-specific T cells.
T Cells Specific for Novel Epitopes Are Present at Higher Frequencies in Subjects With Diabetes
To further assess the relevance of T cells that recognize these novel epitopes, we assessed their frequency in the peripheral blood of age- and HLA-matched control subjects (demographics in Supplementary Table 2). As expected, some T cells that recognized these epitopes were present in control subjects. However, the combined frequency of novel epitope–reactive T cells in subjects with type 1 diabetes was significantly higher than in control subjects (P = 0.0003) (Fig. 4A). Considering T cell frequencies for individual epitopes, frequencies trended toward being higher in subjects with type 1 diabetes than in control subjects for PCSK2 and UCN3 (Fig. 4B), but differences only reached statistical significance for CCNI (P = 0.02, Šídák multiple comparisons test). Using CD45RA and CCR7, we classified novel epitope–reactive T cells as naive or memory and compared the ratio (the combined percentage of all memory subsets divided by the percentage that were naive) in subjects with type 1 diabetes and control subjects (Fig. 4C). Although subjects with type 1 diabetes showed a very modest trend toward having a higher proportion of epitope-specific memory T cells, there was no significant difference (P = 0.565). We next used CXCR3, CCR4, and CCD6 to compare the surface phenotype of novel epitope–reactive T cells in subjects with type 1 diabetes and control subjects (Fig. 4D). Unlike subjects with type 1 diabetes, the predominant phenotype in healthy control subjects was Th2 like (CCR4 single positive). Thus, subjects with established type 1 diabetes had significantly greater numbers of novel epitope–reactive CD4+ T cells and control subjects lacked the Th1-like bias seen for some specificities in subjects with type 1 diabetes.
Cells That Recognize Novel Epitopes Are Present Among Islet-Infiltrating T Cells
To further support the relevance of these novel epitopes, we next evaluated whether their cognate T cells are present within human islets by probing the specificity of T cell lines from nPOD donor 6472 (Supplementary Table 4). Multiple T cell lines were expanded as delineated in Research Design and Methods. We tested the reactivity of these lines toward CCNI-008, PCSK2, and UCN3 by assaying for IFN-γ secretion stimulated by each peptide. Islet-derived CD4+ T cell line AG (isolated and expanded from nPOD donor 6472) recognized PCSK2 in the context of autologous splenocyte-derived B-LCL by IFN-γ secretion (Fig. 5A) and by enzyme-linked immunosorbent spot (ELISpot) (Fig. 5B). The donor was positive for HLA-DRB1*04:04 but not HLA-DRB1*04:01, indicating that PCSK2 can apparently be presented by both of these HLA-DR4 subtypes. These observations establish that CD4+ T cells that recognize PCSK2 are present among the islet-infiltrating T cells of this subject, further supporting the relevance of these newly discovered epitopes to disease.
Discussion
During the development of type 1 diabetes, there is a failure of self-tolerance that is most clearly manifested by the appearance of islet autoantibodies but is also accompanied by loss of T cell tolerance to insulin and other islet-specific antigens (8,36–39). The sequential acquisition of additional islet autoantibodies suggests mounting waves of autoimmune attack that encompass new antigen specificities (36,40). As a consequence of innate and adaptive immune activation and subsequent metabolic dysregulation, pancreatic β-cells experience diverse stressors (41–45). These have been shown to promote the formation of neoepitopes through enzymatic and nonenzymatic posttranslational modification of proteins and defective ribosomal initiation (7,13,16,18,45–47). Such neoepitopes are thought to represent a profound challenge to central and peripheral tolerance mechanisms (19,48). Indeed, T cells that recognize citrullinated, deamidated, and hybrid peptide neoepitopes are present at higher frequencies in subjects with type 1 diabetes and are also present among islet-infiltrating T cells and in pancreatic lymph nodes in this setting (7,9,10,17,49).
The induction of altered transcripts and splice variants leads to the generation of neoepitopes (and epitopes from previously understudied secretory granule proteins) that were previously shown to be recognized by CD8+ T cells (22,23). In this study, we sought to identify novel epitopes derived from those secretory granule proteins and stress-associated splice variant proteins and to explore CD4+ T cell responses directed against these epitopes. Using a systematic approach, we predicted peptides likely to bind to DR0401, assessed their in vitro binding and immunogenicity, and verified their validity by isolating peptide-responsive T cell clones from subjects with type 1 diabetes. These efforts led us to identify three novel epitopes. One of these epitopes spanned a unique junction from cyclin I splice variant CCNI-008. The other epitopes originated from PCSK2 and UCN3, which are both conventional secretory granule proteins, but which have not previously been appreciated as CD4+ T cell targets. Previously identified CD8+ T cell epitopes from these antigens were restricted by HLA-A2 (CCNI-00814–22, PCSK230–38, and UCN31–9) (22) and HLA-A3 (UCN346–56) (23). The murine orthologs PCSK2 and UCN3 included epitopes (restricted by Kd) that were found in the islet-infiltrating T cells of prediabetic NOD mice (23) (mRNA splice epitopes were not investigated in mice, as splicing rules display significant interspecies differences). The novel CCNI-008 epitope that was recognized by CD4+ T cells (CCNI-00819–37) overlaps with the CD8+ T cell epitope CCNI-00814–22, raising the possibility of linked recognition. The novel PCSK2 and UCN3 CD4+ T cell epitopes (PCSK2243–257 UCN358–72) did not overlap with the previous CD8+ T cell epitopes.
Through direct analysis with HLA class II tetramers, we observed detectable populations of T cells that recognized these peptides in the peripheral blood of subjects with type 1 diabetes that, in aggregate, were present at significantly higher frequencies than in HLA-matched control subjects. In considering the individual epitopes separately, only CCNI-specific T cells were present at significantly higher frequencies. Notably, T cells specific for this splice variant–derived CCNI peptide had the highest overall frequencies and exhibited a distinct memory profile (increased proportions of TCM and TEM) compared with the other specificities examined. As a whole, these observations suggest that T cells that recognize this splice variant–derived epitope might be more effectively primed and expanded than T cells that recognize the epitopes derived from conventional secretory granule antigens. Notably, T cells that recognize the PCSK2 epitope were present among the islet-infiltrating T cells of an nPOD donor, supporting the relevance of this specificity. In accord with this, CCNI- and PCSK2-specific T cells predominantly exhibited a Th1-like phenotype, whereas a significant proportion of T cells specific for the UCN3 epitope exhibited a Th2-like surface phenotype. Based on the paradigm that Th1 cells have a key role in the pathogenesis of type 1 diabetes, as supported by the foundational work of Katz et al. (50), this observation suggests a pathogenic role for these cells in type 1 diabetes.
T cells that recognize these novel epitopes were also present in HLA-matched control subjects, albeit at lower frequencies and lacking the Th1-like phenotype that was predominant in subjects with type 1 diabetes. In line with previous reports on islet-reactive CD8+ T cells (22,23,51), a notable proportion of these novel autoreactive CD4+ T cells were naive in peripheral blood, supporting the possibility that the fraction actively engaged in the autoimmune process may be preferentially sequestered in the pancreas target organ (52,53). These observations indicate incomplete tolerance, consistent with other studies suggesting that having a disease-susceptible HLA is sufficient to cause the selection of a potentially autoreactive T cell repertoire (51,54,55). The observation that self-reactive T cells were not completely limited to a naive phenotype in healthy donors indicates that limited T cell expansion can occur in the absence of autoimmune disease. This is in accord with the concept of “benign” islet autoimmunity, which postulates that self-reactive T cells are present in all individuals and that contributing behavior on the part of pancreatic β-cells is required to potentiate disease progression (52,53).
Our study does have limitations. Notably, although we observed diverse frequencies of T cells specific for novel T cell epitopes, our study was not adequately powered to correlate these differences with clinical parameters. Furthermore, our study was limited to peptides with predicted binding to DR0401 and derived from targets that were selected based on HLA class I peptidome data. Although the known β-cell antigens targeted by CD4+ and CD8+ T cells have a high degree of overlap (5), the approach we followed would omit any self-antigens that are uniquely presented to CD4+ T cells. In addition, given that HLA binding predictions are imperfect, there could be epitopes present among the peptides we chose not to synthesize. Since our methods relied on HLA tetramer staining, which preferentially detects high-affinity T-cell receptors, it is possible that there are additional low-affinity specificities that we failed to discover. Also, it is likely that additional epitopes derived from the unique splice variants and secretory granule proteins that we studied are recognized in human subjects with type 1 diabetes in the context of other disease-associated HLA molecules, such as DQ0302 (DQ8). We have developed a method for predicting DQ8-binding peptides, which we plan to use in our ongoing studies.
Collectively, our findings introduce UCN3 as a potentially relevant CD4+ T cell antigen and strongly suggest that the CCNI splice variant and PCSK2 are important T cell targets in type 1 diabetes. Among these, CCNI-specific T cells appear to be more extensively expanded in the circulation of subjects with type 1 diabetes. We can speculate that pathogenic CD4+ T cells that recognize this tissue-specific splice junction epitope may not be effectively controlled by central and peripheral tolerance mechanisms. Our data support the hypothesis that stress-related changes that occur in human β-cells can generate neoepitopes, eliciting inflammatory T cell responses, which amplify autoimmunity and promote the progression of type 1 diabetes.
This article contains supplementary material online at https://doi.org/10.2337/figshare.21264837.
Article Information
Acknowledgments. The authors thank Virginia M. Green for copy editing and proofreading the manuscript and Cynthia Cousens-Jacobs for administrative support (both at Benaroya Research Institute).
Funding. This work was supported by JDRF grants 1-SRA-2020-978-S-B and 2-SRA-2014-297-Q-R to E.A.J. Work in the laboratory of S.C.K. was supported by National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health, grant UC4DK116284 to S.C.K. Work in the laboratory of D.L.E. was supported by Welbio and Fonds National de la Recherche Scientifique, Belgium, grants CR-2015A-06 and CR-2019C-04, and Innovate2CureType1-Dutch Diabetes Research Fundation (DDRF). Work in the laboratory of R.M. was supported by The Leona M. and Harry B. Helmsley Charitable Trust (1901-03689), Agence Nationale de la Recherche (ANR-19-CE15-0014-01), Fondation pour la Recherche Medicale (EQU20193007831), and IdEx Université Paris 2019. D.L.E. and R.M. received funding from the Innovative Medicines Initiative 2 Joint Undertaking under grant agreements 115797 and 945268 (INNODIA and INNODIA HARVEST), which receive support from the EU Horizon 2020 program, the European Federation of Pharmaceutical Industries and Associations, JDRF, and The Leona M. and Harry B. Helmsley Charitable Trust. The work of C.S. was supported by JDRF under grant 3-SRA-2019-791-S-B. This research was performed with the support of nPOD (RRID:SCR_014641), a collaborative type 1 diabetes research project supported by JDRF (nPOD: 5-SRA-2018-557-Q-R) and The Leona M. and Harry B. Helmsley Charitable Trust (2018PG-T1D053, G-2108-04793). Organ procurement organizations partnering with nPOD to provide research resources are listed at https://www.jdrfnpod.org/for-partners/npod-partners/.
These funders had no role in study design, data collection and analysis, or decisions made to prepare or publish the manuscript. The content and views expressed are the responsibility of the authors and do not necessarily reflect the official view of nPOD.
Duality of Interest. No potential conflicts of interest relevant to this article were reported.
Author Contributions. R.M. and E.A.J. conceptualized and designed the study. R.M. and D.L.E. identified the sequences of candidate splice variant junctions and secretory granule proteins of interest. C.S. and S.L. were responsible for subject recruitment, sample collection, and clinical data collection. P.G. and D.A.-L. performed peptide binding assays and T cell studies. A.M. and S.C.K. performed islet-derived T cell experiments. P.G. and E.A.J. wrote the manuscript with assistance from all co-authors. E.A.J. obtained funding and was responsible for the entire project. E.A.J. 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.