Inflammation-Induced Citrullinated Glucose-Regulated Protein 78 Elicits Immune Responses in Human Type 1 Diabetes

Type 1 diabetes is an autoimmune disease characterized by T-cell–mediated destruction of insulin-producing pancreatic β-cells, leading to insulin deficiency (1). Although it is well-established that loss of central and peripheral tolerance toward native self-proteins leads to the escape and accumulation of autoreactive T-cells, the exact mechanisms that trigger this break in self-tolerance are not entirely understood (2).

Growing evidence suggests that the β-cell is not a passive target of autoimmunity but actively contributes to its own destruction by triggering the immune system (2). Many of the triggers associated with type 1 diabetes, such as inflammatory cytokines, free radicals, and viral infections, result in endoplasmic reticulum (ER) or oxidative stress (3–11). These stresses can lead to modifications of β-cell proteins, including posttranslational modifications (PTMs) (10,12–17), mRNA alternative splicing, hybrid peptide formation (18,19), and nonconventional translation, potentially generating neoepitopes against which no tolerance exists in the immune system (20). In several autoimmune diseases, such as rheumatoid arthritis, multiple sclerosis, systemic lupus erythematosus, and celiac disease, PTM proteins are established autoantigens (21,22), but their relevance in type 1 diabetes is only starting to be explored.

One PTM that received recent attention in type 1 diabetes is citrullination, which involves the deimination of arginine to citrulline by peptidyl arginine deiminase (PAD) enzymes (15,17,23). Our group previously demonstrated that the ER chaperone glucose-regulated protein 78 (GRP78), classically known as an important regulator of the unfolded protein response (21), is citrullinated in β-cell lines when exposed to inflammatory stress (17). To our knowledge, we provide here the first direct evidence by two-dimensional difference gel electrophoresis (2D-DIGE) and liquid chromatography-tandem mass spectrometry (LC-MS/MS) that GRP78 is citrullinated in human islets after cytokine exposure. Together with the identification of higher CD4⁺ T-cell frequencies directed against citrullinated GRP78 (citGRP78) epitopes and significantly elevated autoantibody titers in patients with type 1 diabetes, these results suggest that citrullination of β-cell proteins contribute to loss of tolerance toward β-cells in human type 1 diabetes and identify citGRP78 as a novel islet autoantigen.

Human islets were obtained from the Alberta Diabetes Institute Islet Core and Pisa University, with the agreement of the local ethics committee from the University of Alberta (Edmonton, AB, Canada) (Pro00013094; Pro00001754) and Pisa University (#2615) (24,25) (Supplementary Table 1). These were cultured in DMEM supplemented with 1% l-glutamine (Gibco), 10% FBS, and 100 units/mL penicillin-streptomycin. Islets were exposed to human interleukin-1β (IL-1β) (50 units/mL; R&D Systems), murine tumor necrosis factor-α (TNF-α) (1,000 units/mL; R&D Systems), and human interferon-γ (IFN-γ) (1,000 units/mL; Peprotech) for 24 or 72 h (26), as indicated in the figure legends.

Blood samples for plasma and peripheral blood mononuclear cell (PBMC) collection were obtained from individuals with type 1 diabetes and healthy control subjects (Supplementary Tables 2–4). All procedures were approved by the ethics committee UZ Leuven (S52697; Leuven, Belgium) and conducted in accordance with the principles of the Declaration of Helsinki, with written informed consent obtained from all participants before inclusion.

Quantitative RT-PCR was performed as previously described (17). Primers used were CHOP forward: 5′-GAACGGCTCAAGCAGGAAATC-3′; CHOP reverse: 5′-TTCACCATTCGGTCAATCAGAG-3′; GRP78: Hs.PT.58.22715160 (Integrated DNA Technologies); and ATF3: Hs.PT.56a.41052403.g (Integrated DNA Technologies).

Islets were incubated with propidium iodide (Invitrogen) and Hoechst 33342 (Invitrogen) for 15 min at 37°C. Imaging was performed with a Zeiss Colibri LED microscope. At least 15 islets were analyzed by two independent researchers (one blinded to sample identity).

Analysis with 2D-DIGE was performed with protein lysates from 1,000 islets/condition, as previously described (5).

The 2D gels were blotted onto a polyvinylidene fluoride membrane (GE Healthcare) and probed with anti-GRP78 antibody (Santa Cruz) and goat anti-rabbit horseradish peroxidase (Dako) as previously described (17).

Human islets were lysed in lysis buffer (7 mol/L urea, 2 mol/L thiourea, 4% weight for volume 3-[(3-cholamidopropyl) dimethylammonio]-1-propanesulfonate, 40 mmol/L Tris base, 1% weight for volume dithiothreitol [DTT], and a mixture of protease inhibitors [cOmplete protease inhibitor; Roche Diagnostics]). Cell lysates (10 μg) were reduced in 5 mmol/L DTT for 30 min at 37°C followed by alkylation with 25 mmol/L iodoacetamide and quenching of excess iodoacetamide with 25 mmol/L DTT, both by incubating for 30 min at 37°C in the dark. Reduced and alkylated cell lysates were protein precipitated using the Wessel-Flügge method (27) and were digested with 1 μg modified trypsin (Pierce) or 1.5 μg Glu-C (Promega) at pH 8 in the presence of 5% acetonitrile and 0.01% ProteaseMAX (Promega), subjected to desalting with C18 ZipTip pipette tips (Millipore) (28), and analyzed by unbiased LC-MS/MS or targeted data-dependent acquisition with inclusion list on a hybrid quadrupole-Orbitrap mass spectrometer (Q Exactive; Thermo Fisher Scientific). Peptides were identified by Mascot (Matrix Science) using SwissProt (Homo sapiens, 169,779 entries) as a database through Proteome Discoverer 2.2, incorporating Percolator for peptide validation. Oxidation (M), deamidation (N/Q), and deamidation (R) (referring to citrullination) were included as variable modifications, and carbamidomethylation (C) was included as a fixed modification. Two missed cleavages were allowed for both trypsin and Glu-C digestion. Peptide tolerance was set at 15 parts per million and MS/MS tolerance at 20 millimass units. Only peptides displaying high and medium confidence (posterior error probabilities ≤ 0.01 and 0.05) were retained.

Because citrullination and deamidation result in a mass increment of 0.98402 Da, whereas a ¹³C isotope of a native peptide results in a mass increment of 1.003355 Da, we manually checked differences in MH⁺ of native versus modified peptide. Moreover, we checked differences in retention time between modified and native peptides, as previously described (29). To discriminate between citrullination and deamidation, we manually checked the MS/MS spectra of the putative citrullinated peptides, inspecting the isotopic distribution of peptide fragments containing N, Q, and R. Synthetic peptides (Synpeptide) IDVNGILRVTAE, IDVDGILRVTAE, and IDVNGILXVTAE were used to distinguish between citrullinated and deamidated peptides, using LC-MS/MS fragmentation data and retention time as discriminating criteria.

Native and citGRP78 epitope candidates with predicted binding affinity for HLA-DRB1*04:01 were identified using the in silico approach outlined in previous work (30) and synthesized by Synpeptide. Binding affinity to HLA-DRB1*04:01 was measured by incubating increasing concentrations of peptides in competition with 0.02 μmol/L biotinylated influenza hemagglutinin (HA) 306–318 in wells coated with DRB1*04:01 protein. After washing, residual biotin-HA 306–318 was detected using europium-conjugated streptavidin and quantified using a Victor2D time-resolved fluorometer (PerkinElmer). IC₅₀ values were calculated as the concentration needed to displace 50% of the biotin-HA 306–318 peptide.

Recombinant HLA-DRB1*04:01 was purified from insect cell culture supernatants by affinity chromatography and dialyzed against phosphate buffer (pH 6.0). To prepare tetramers (Tmrs), DRB1*04:01 was biotinylated in vitro and incubated with 0.2 mg/mL peptide at 37°C in 0.2% n-octyl-β-d-glucopyranoside and 1 mmol/L Pefabloc SC. Monomers were conjugated into Tmrs using R-phycoerythrin (PE) streptavidin at a molar ratio of 8:1.

PBMCs from HLA-DRB1*04:01⁺ donors (Supplementary Tables 2 and 3) were isolated from heparinized blood using Lymphoprep and frozen in AIM V Medium (Gibco) containing 10% DMSO.

PBMCs were expanded in vitro as previously described (31). Briefly, 5 × 10⁶ PBMCs were stimulated with 6 μmol/L pooled citGRP78 peptides. After 14 days of in vitro stimulation, PE-labeled Tmrs were used to stain T cells specific for the GRP78 peptides (75 min at 37°C). Analysis was performed by FACSCalibur (BD Biosciences) using FlowJo software (TreeStar). Positivity was defined as the presence of a distinct population of CD4 bright cells at a percentage more than twofold above background (staining of cells from the unstimulated well set to 0.1% for most experiments).

Cryopreserved PBMCs (20–30 × 10⁶) were treated with 50 nmol/L of dasatinib (Cell Signaling) for 7 min at 37°C and stained with Tmrs for 2 h at room temperature (32). Cells were labeled with anti-PE and anti-allophycocyanin magnetic beads for 20 min at 4°C. Next, 1/40th of the cells (preenriched fraction) was reserved to determine the total cell number used in the assay. The remaining fraction was enriched using Miltenyi Biotec MS magnetic columns (enriched fraction) according to the manufacturer’s instructions. Both fractions were stained for 15 min at 4°C with CD4-BV421 (BioLegend), CCR7-APC-Cy7 (BioLegend), CD45RA-AF700 (BD Bioscience), SYTOX Green (Thermo Fisher Scientific), and CD20/CD14 fluorescein isothiocyanate (eBioscience). Samples were fully collected on an LSR II flow cytometer (BD Bioscience) and analyzed using FlowJo version 10.1 software. Data were expressed as the frequency of Tmr⁺ cells per 10⁶ living CD4⁺ T cells and considered positive when >5 (33).

Immunoreactivity of human plasma to native and citGRP78 was performed by ELISA (Supplementary Table 4). Briefly, 0.125 μg of recombinant human native or citGRP78 protein (constructed and produced as previously described [17]) in 0.05 mol/L carbonate-bicarbonate buffer (pH 9.6) (Sigma) was coated onto Immulon 2 HB plates (Thermo Fisher Scientific) overnight at 4°C. The plate was blocked with 1% BSA in PBS containing 0.05% Tween 20 for 2 h at room temperature. Plasma was diluted 1:100 in PBS-Tween 20 containing 0.3% BSA and incubated for 1 h. Mouse anti-GRP78 monoclonal antibody (Santa Cruz) served as a positive control. Species-specific goat anti-human IgG (Southern Biotech) or rat anti-mouse IgM alkaline phosphatase (Santa Cruz) was added as a secondary reagent. Color was developed with para-nitrophenylphosphate substrate (Sigma). All readings were normalized to nonspecific plasma binding from blank wells (no antigen added). In parallel assays, wells coated with PAD enzymes only confirmed specific binding to citGRP78.

Data are shown as mean ± SD. Significant differences between experimental conditions were assessed by paired Student t test, Wilcoxon rank sum test, Kruskal-Wallis test with Mann-Whitney U comparison, or MANOVA as indicated (GraphPad Prism version 7.02 and SPSS version 22 software). P < 0.05 was considered statistically significant.

To study the effect of inflammation on the modification of GRP78, islets were exposed in vitro to IL-1β, IFN-γ, and TNF-α for 24 or 72 h followed by 2D-DIGE analyses. Cytokine exposure significantly induced cell death after 72 h (11.39 vs. 4.41% in control islets; P < 0.05, n = 5) (Fig. 1A). Whereas after 24 h no increased cell death was observed (Fig. 1A), upregulation of CHOP, GRP78, and ATF3 mRNA (P < 0.05, n = 3) suggested that increased ER stress preceded the induction of cell death (Fig. 1B–D).

Analysis of the islet 2D-DIGE pattern in control islets revealed the presence of five GRP78 spots (I1–I5) differing in isoelectric point but not in molecular weight (Fig. 1E). This pointed to the presence of at least five GRP78 isoforms of which the identity was confirmed by 2D Western blotting (Fig. 1F). Quantitative 2D-DIGE analysis of cytokine-exposed versus control islets (n = 5; donors D1–D5) (Supplementary Table 1) revealed a significant upregulation of the two most abundant GRP78 isoforms (isoform I2: 1.5-fold upregulated [P < 0.05]; isoform I3: 1.64-fold upregulated [P < 0.05]). In three of five islet donors (D1–D3), this upregulation was paralleled by a shift in relative abundance toward the isoform with the lower isoelectric point (isoform I3), which is indicative of an increased acidic nature consistent with citrullination or other PTMs, such as phosphorylation, deamidation, and ribosylation, among others (Fig. 1G and H).

To identify which GRP78 amino acid residues were modified, we investigated the human islet proteome by unbiased LC-MS/MS (n = 5; donors D6–D10) (Supplementary Table 1). On the basis of our earlier observation that GRP78 is citrullinated in INS-1E cells (17), we specifically focused on the identification of citrullinated arginine (R) residues. GRP78 contains 28 arginines, all potential sites for citrullination. Combined LC-MS/MS analysis of trypsin and Glu-C–digested islets from five islet donors, control or cytokine-exposed, resulted in 67.58% and 71.71% coverage of GRP78, respectively, including 22 of the 28 arginines (Fig. 2A). However, a conversion to citrulline could be unambiguously identified for none of these covered arginine residues.

We hypothesized that the inability to detect citrullination of GRP78 could indicate that either citrullination does not occur, the citrullinated arginine residues are not covered in the current approach, or the citrullinated arginine residues are not abundant enough to be detected by unbiased LC-MS/MS. We therefore applied a targeted MS/MS strategy (data-dependent acquisition) specifically focusing on R510, the arginine residue we have previously identified as being citrullinated in INS-1E upon cytokine-exposure and against which diabetic NOD mice have elevated levels of circulating autoantibodies and autoreactive T cells (17). Because citrullination often is difficult to distinguish from deamidation, we first analyzed a native, citrullinated, and deamidated version of a synthetic Glu-C peptide harboring R510 by LC-MS/MS. We measured a clear difference in retention time of the various peptides: 182.53 ± 1.87, 209.92 ± 1.91, and 189.18 ± 0.66 min, respectively (Fig. 2B). Unbiased LC-MS/MS analysis of five different human islet samples, control or cytokine-exposed, revealed the presence of precursor ions with a mass, charge, and retention time corresponding to those of the synthetic native and citrullinated peptides. By subsequent targeted LC-MS/MS, fragmentation data were obtained that enabled us to unambiguously assign these precursor ions to the native and citrullinated R510-harboring peptides and to distinguish them from the deamidated peptide, the latter on the basis of the diagnostic b⁺-ion of charge/mass ratio (m/z) 612.34 (Fig. 2C). A combination of correct m/z, retention times, and diagnostic fragmentation data led us to conclude that R510 is citrullinated in human islets.

On the basis of the finding that GRP78 is citrullinated in cytokine-exposed human islets, we investigated whether GRP78 citrullination could result in the generation of neoepitopes recognized in patients with type 1 diabetes. To this end, we first identified GRP78 peptide sequences containing possible HLA-binding motifs. Because we could not cover the full-length GRP78 protein sequence by MS analysis, thus possibly missing other citrullinated arginine residues, we screened the entire GRP78 sequence for potential HLA-DRB1*04:01 binders. We specifically focused on HLA-DRB1*04:01 because binding to this allotype has been shown to be enhanced by citrullination at certain amino acid positions (34,35). A previously developed algorithm was used to predict in silico peptide sequences with a high affinity for HLA-DRB1*04:01 in their native or citrullinated forms. Thirteen peptides were identified as potential binders (Table 1). Of note, among these peptides were 498–512X and 500–514X citrullinated at R510, the residue identified here by LC-MS/MS as citrullinated in cytokine-exposed islets and previously identified as an autoantigenic epitope in NOD mice (17). In vitro binding affinity assays demonstrated that 3 of the 13 citrullinated peptides bound to DRB1*04:01, namely 195–209X (citrullinated on R197, covered by LC-MS/MS, but not modified), 498–512X (citrullinated on R510), and 500–514X (citrullinated on R510) (Table 1). Two other peptides (58–72R, 65–79R) bound DRB1*04:01 only in their native form but with much lower affinity.

We assessed the ability of the citrullinated peptides 195–209X, 498–512X, and 500–514X to elicit CD4⁺ T-cell responses in PBMCs from patients with type 1 diabetes and healthy subjects (Supplementary Table 2) stimulated in vitro for 14 days followed by detection with DRB1*04:01 Tmrs. Despite the low binding affinity of 498–512X and 500–514X (Table 1), stable Tmrs could be generated. All peptides elicited positive CD4⁺ T-cell responses in multiple subjects (Fig. 3). One epitope, 498–512X, was exclusively recognized by CD4⁺ T cells from patients with type 1 diabetes. Peptide 195–209X elicited T-cell responses in one of six healthy subjects and in one patient with type 1 diabetes. In two of six healthy subjects, a response against peptide 500–514X was observed, whereas only one patient with type 1 diabetes was responsive (Fig. 3).

On the basis of this first evidence that citGRP78 epitopes are able to elicit T-cell responses, we assessed the frequencies of CD4⁺ T cells that recognize citGRP78 epitopes in PBMCs from patients with new-onset type 1 diabetes (<1 year, n = 4), patients with long-standing type 1 diabetes (≥1 year, n = 11), and healthy subjects (n = 8) by direct ex vivo Tmr assays (Fig. 4A–H) (donor characteristics listed in Supplementary Table 3). We focused on peptides that elicited a similar or higher T-cell response in patients with type 1 diabetes compared with healthy subjects on the basis of the in vitro Tmr assay results (Fig. 3). The 498–512X reactive CD4⁺ T cells were more frequent in patients with type 1 diabetes than in healthy subjects, with a positive response in 7 of 15 patients (Fig. 4J). Although the native variant 498–512R also elicited a positive response in 5 of 15 patients, the frequency of 498–512R-reactive CD4⁺ T cells was not significantly higher in patients with type 1 diabetes than in healthy subjects (Fig. 4I). For peptide 195–209X (Fig. 4L) and its corresponding unmodified sequence (Fig. 4K), a positive response was observed in a subset of patients, but the frequency of Tmr⁺ cells was not significantly higher than in healthy subjects. Although no significant difference was observed when comparing T-cell frequencies recognizing the native and citrullinated form of each peptide in each individual (Fig. 4M–P), a trend toward higher frequencies of Tmr⁺ T-cells directed against 498–512X was observed (P = 0.058) (Fig. 4N).

In samples from patients with type 1 diabetes with >10 Tmr⁺ cells, we further examined the activation phenotype of 498–512R- and 498–512X-experienced Tmr⁺ T cells on the basis of CCR7 and CD45RA (Fig. 5A–C). In patients with long-standing type 1 diabetes, the 498–512X Tmr⁺ fraction displayed a higher percentage of central memory and a lower percentage of effector memory cells compared with the 498–512R Tmr⁺ fraction (Fig. 5D and E). No differences were observed in the frequency of naive T cells specific for 498–512R and 498–512X (Fig. 5F).

Next, we assessed autoantibody reactivity against citGRP78 and its native counterpart in plasma of patients with new-onset type 1 diabetes (n = 47), long-standing type 1 diabetes (n = 61), and of healthy subjects (n = 89) (Fig. 6A and Supplementary Table 4). With 33% of the patients with long-standing type 1 diabetes being autoantibody positive, these patients had significantly higher anti-citGRP78 autoantibody titers than healthy subjects (Fig. 6A and B). When comparing the reactivity against native and citGRP78 of the same subjects, patients with new-onset and long-standing type 1 diabetes had significantly higher titers of anti-citGRP78 autoantibodies. Of note, 4 of 5 citGRP78 autoantibody-positive patients with new-onset type 1 diabetes and 15 of 20 citGRP78 autoantibody-positive patients with long-standing type 1 diabetes also showed reactivity against the native epitope (Fig. 6C–E).

Autoantibodies against native and citGRP78 primarily were present in patients with multiple autoantibodies against insulin, GADA, and/or IA-2 antigen (Fig. 6F and G). When comparing autoantibody positivity for native and/or citGRP78 in relation to diabetes duration, autoantibodies against native and/or citGRP78 most often were detected in patients diagnosed within 1–5 years (Fig. 6H).

We report on the citrullinated ER chaperone GRP78 as a novel autoantigen in human type 1 diabetes. We demonstrate that citrullination of GRP78 occurs in human islets exposed to inflammatory stress. With some citGRP78 peptides displaying strong binding to HLA DRB1*04:01, we identified the presence of CD4⁺ T cells and circulating autoantibodies reactive against citGRP78 in a subset of patients with type 1 diabetes. We thus provide the first direct evidence in our knowledge that β-cell proteins can be citrullinated and that islet cell–derived GRP78 epitopes may be involved in the pathogenesis of human type 1 diabetes through the formation of citrullinated neoantigens that evoke CD4⁺ T-cell and B-cell responses.

In recent years, PTMs of β-cell proteins have emerged as a potential source for neoantigens in type 1 diabetes, with many studies focusing on their role in inducing T-cell responses (12–16,18,19,36). Here, we provide a more comprehensive insight by covering not only the interaction between the modified proteins and the immune system but also the role of inflammation in inducing these modifications. Given that the PTM citrullination is closely associated with inflammatory conditions and that inflammation plays a pivotal role in the induction, amplification, and maintenance of type 1 diabetes (2), this modification also may be important in this disease (37). The observation that T-cell responses are detected against citrullinated peptides derived from GAD65 (15), islet amyloid polypeptide (14), and now GRP78 is an indication that citrullination indeed plays a role in human type 1 diabetes. Of note, we also show for the first time in our knowledge that β-cell proteins can be citrullinated when human islets are exposed to inflammatory cytokines (IL-1β, IFN-γ, and TNF-α), as exemplified by the citrullination of GRP78. This finding reinforces the concept that inflammation modulates the cross-talk between the β-cell and the immune system not only by mediating β-cell dysfunction and death (3) but also by inducing β-cell neoantigen formation. Recent data on islet-reactive CD8⁺ T cells have suggested that the autoimmune repertoire between patients with type 1 diabetes and healthy control subjects is largely overlapping (38) and that the critical ingredient leading to the priming of this repertoire may be the availability of its target islet antigens in the inflammatory milieu of insulitis. In light of the evidence that GRP78 citrullination is induced by proinflammatory cytokines, we hypothesize that this paradigm also may apply to citGRP78.

We have used a combination of IL-1β, IFN-γ, and TNF-α, where TNF-α may be crucial to evoke cytokine-induced PTMs in human islets, because the combination of IL-1β and IFN-γ alone does not induce GRP78 modifications in human islets (39). Of importance, PTMs of GRP78 were not induced in all islet donors investigated; in two islet preparations, the levels of modified GRP78 were already elevated in control islets and could not be enhanced further by cytokine exposure. This may be explained by higher basal levels of stress that are intrinsically associated with the islet isolation procedure, such as a more harsh enzymatic digestion (40) or a longer cold ischemia time (41). On the other hand, it may also reflect the inherent heterogeneity in humans, with not every individual being prone to citrullination even when exposed to inflammation. Whether the difference in induction of GRP78 PTMs between islet donors is due to variations in genetic background, possibly leading to abnormal expression levels of PAD enzymes, or to differences in susceptibility to environmental triggers, which may result in increased Ca²⁺ fluxes necessary for the activation of PAD enzymes, remains to be elucidated.

With the use of MS, we identified R510 as a citrullination site in human GRP78. The 2D-DIGE profile showed, however, at least five different isoforms, suggesting modifications on several residues. With a sequence coverage of 68–72% identifying 22 of the 28 arginines within GRP78, we potentially missed other citrullination sites. The use of alternative digestion enzymes, such as Lys-C or elastase, or the use of a more sensitive targeted approach as performed here for the R510 region, may result in a more complete picture of GRP78 citrullination. Next to this, other PTMs, such as deamidation, also may result in the multiple isoform spots observed in the 2D-DIGE (42).

Alterations in peptide epitopes are known to affect immune recognition by modulating HLA binding or influencing T-cell recognition (34,35,43). Of the 13 citGRP78 peptides predicted to bind HLA-DRB1*04:01, 2 with a R510 citrullination site (498–512X and 500–514X) were confirmed to be preferential binders compared with their native counterparts. This observation supports the notion that some HLA allotypes, such as HLA-DRB1*04:01, prefer citrulline over arginine at key peptide-binding positions (34,35). This may reflect either a direct interaction of citrulline with HLA anchor positions or an indirect effect mediated by a change in peptide orientation within the binding groove. The binding motif of HLA-DRB1*04:01 displays a strong preference for aliphatic and aromatic residues in the first binding pocket. According to this motif, the canonical nonamer motif of epitope 498–512X most likely to bind to HLA-DRB1*04:01 would be FEIDVNGIL because it contains an optimal F residue at P1, a highly favorable D at P4, and a preferred N at P6 (44–46). This places citrulline in position 10, which according to Yassai et al. (47) can modulate peptide-binding affinity and induce an alternative downward configuration of the peptide in the binding groove, thus shifting the T-cell receptor contact positions.

Although the binding affinity of peptide 195–209 was improved upon citrullination, the frequency of Tmr⁺ T cells directed against this epitope was not significantly increased in patients with type 1 diabetes compared with healthy subjects. Together with our observation that R197 is not citrullinated in cytokine-treated human islets (Fig. 2A), these results may indicate that citrullination of R197 does not readily occur in human type 1 diabetes.

R510 is exactly the same arginine residue as previously found to be citrullinated in INS-1E cells upon cytokine exposure and against which diabetic NOD mice have elevated levels of circulating autoantibodies and autoreactive T cells (17). Using ex vivo Tmr assays and ELISAs, we now demonstrate elevated T-cell and autoantibody responses against 498–512X in patients with type 1 diabetes, which strongly indicates that the same epitope as in NOD mice is recognized in human type 1 diabetes. The observation that there are patients with CD4⁺ T-cells and autoantibody responses to the native, the citrullinated, or both epitopes may suggest that epitope spreading occurs (48). There is great interest in knowing to which epitope the immune system initially responds, following the hypothesis that this epitope may behave as a disease driver (49) and may be a potential target for antigen-specific treatments (48).

That B- and T-cell responses were more readily observed in patients with long-standing (≥1 year) type 1 diabetes than in patients with new-onset type 1 diabetes underscores our hypothesis that citGRP78, rather than being a primary autoantigen, may be involved in the amplification of islet autoimmunity once inflammation is already ongoing. The activation of PAD enzymes requires high levels of cytosolic Ca²⁺ (11), which is favored by ER dysfunction in combination with excessive insulin demand. A similar process has been described for the Ca²⁺-dependent PTM enzyme tissue transglutaminase 2 in which activity is increased by ER stress, leading to β-cell stress-dependent immunogenicity (10). In this scenario, our observation that GRP78 is translocated to the plasma membrane upon cytokine exposure in rodent β-cells (17) and human islets (S. Vig, L.O., unpublished observations) is noteworthy. Although the mechanisms have not yet been elucidated, the combination of membrane translocation and increased cytosolic Ca²⁺ concentrations may be a prerequisite for GRP78 citrullination to occur.

Similar to other autoantigens (12,15), we show that both B- and T-cell responses to citGRP78 are not homogeneous across the patient population, which may reflect the heterogeneity of the disease or the sequestration of autoimmune T cells in diseased organs (38). This heterogeneity suggests that it is unlikely that a single marker can predict disease onset or progression, emphasizing the importance of identifying other autoantigens and to map T-cell responses and autoantibody generation in at-risk individuals and patients with type 1 diabetes during disease initiation and progression.

In conclusion, this study reinforces the notion that β-cells are actively involved in their own destruction by generating citrullinated β-cell protein epitopes when exposed to inflammatory stress, as shown here for GRP78. We demonstrate that one of these citGRP78 epitopes is able to elicit CD4⁺ T-cell and autoantibody responses primarily in patients with long-standing type 1 diabetes. These results may suggest that immune responses against citGRP78 contribute to the acceleration of disease rather than to the precipitation of its development. On the other hand, it may also be possible that GRP78-specific immune cells were mainly sequestered in the pancreas and pancreatic lymph nodes until after disease onset, after which they emerged in the periphery. Our findings further suggest that epitope spreading occurs, with detectable T- and B-cell responses directed against both the native and citrullinated epitope. Given the heterogeneity of type 1 diabetes, the identification of PTM β-cell proteins as potential neoantigens is critical for a better understanding of disease pathology, the discovery of new markers for earlier diagnosis and patient stratification, and the development of novel disease interventions, especially in the view of more personalized therapies. Determination of the stage of disease during which T- and B-cell reactivity toward modified and native self-antigens occurs will shed more light on the question of whether responses to modified epitopes are primary responses and thus drivers of the disease process. This will be of crucial importance for disease prediction, prevention, and development of antigen-specific therapies.

Acknowledgments. The authors thank Frea Coun and Dries Swinnen (KU Leuven, Leuven, Belgium) for technical assistance, Patrick MacDonald (Alberta Diabetes Institute, University of Alberta, Edmonton, AB, Canada) for the supply of human islets, Dr. Sebastien Carpentier (SyBioMa Mass Spectrometry, KU Leuven) for helpful advice, and Hilde Morobé (UZ Leuven) for patient recruitment and collection of blood samples. The authors also thank all the patients and healthy volunteers for participating in this study.

Funding. This work was supported by the Innovative Medicines Initiative Joint Undertaking under grant agreement 115797 (INNODIA). This joint undertaking receives support from the European Union’s Horizon 2020 research and innovation program and the European Federation of Pharmaceutical Industries and Associations, JDRF, and The Leona M. and Harry B. Helmsley Charitable Trust; KU Leuven (GOA14/010); JDRF (2-SRA-2015-52-Q-R, 1-INO-2018-638-A-N); a JDRF postdoctoral fellowship 3-PDF-2018-593-A-N (to M.Bui.); and Fonds Wetenschappelijk Onderzoek (for a PhD fellowship for A.C. [1189518N] and D.P.C. [11Y6716N]).

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

Author Contributions. M.Bui., A.C., F.M.C.S., and I.C. contributed to the design, conduct, analysis, and interpretation of the data and wrote and edited the manuscript. M.Bui., A.C., F.M.C.S., I.C., G.B.-F., M.-L.Y., M.Bug., D.A.-L., M.M., D.P.C., E.W., R.M., J.D.P., P.M., M.J.M., R.D., E.A.J., C.M., and L.O. revised the article and gave their final approval of the version to be published. A.C., E.W., and R.D. conducted, analyzed, and interpreted the proteomics data. G.B.-F., D.A.-L., D.P.C., and E.A.J. contributed to the design, analysis, and/or interpretation of the peptide-binding affinity and Tmr data. M.-L.Y. and M.J.M. designed and performed the ELISA experiments as well as contributed to the analysis and interpretation. M.Bug. and P.M. provided human islets. M.M., R.M., and J.D.P. provided intellectual input. C.M. and L.O. designed research, interpreted data, and wrote and edited the manuscript. C.M. and L.O. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented in abstract form at the Immunology of Diabetes Society Congress, London, U.K., 25–29 October 2018, and the Human Proteome Organization World Congress, Orlando, FL, 30 September–3 October 2018.