Citrullinated Glucose-Regulated Protein 78 Is an Autoantigen in Type 1 Diabetes

Type 1 diabetes is an autoimmune endocrine disease in which loss of central and peripheral tolerance toward β-cell antigens is proposed as the underlying mechanism. However, β-cells themselves also contribute to trigger and/or propagate the autoimmune attack, leading to a dialogue with immune infiltrating cells that may amplify local inflammation (insulitis) in genetically predisposed individuals (1). Insulin (or proinsulin) is probably the primary autoantigen in type 1 diabetes (2), but antigen spreading occurs as the autoimmune assault progresses, with autoantibodies appearing against several non–β-cell-specific autoantigens, such as GAD65 (3), islet antigen 2 (IA2) (4), heat shock protein 60 (HSP60) (5), and chromogranin A (ChgA) (6).

During insulitis, local production of inflammatory mediators, such as the cytokines interleukin (IL)-1β and interferon-γ (IFNγ), triggers β-cell oxidative and endoplasmic reticulum (ER) stress. These, and other signals, may lead to alternative splicing and misfolding of β-cell proteins as well as posttranslational modifications (PTMs) (7–9). In other autoimmune diseases, like rheumatoid arthritis (RA), multiple sclerosis, and celiac disease, such posttranslationally modified proteins behave as autoantigens (10,11), but their relevance in type 1 diabetes is only starting to be explored (12–15).

Building on our previous observation that the ER chaperone 78 kDa glucose-regulated protein (GRP78; also named binding immunoglobulin protein [BiP]) is posttranslationally modified in cytokine-exposed insulin-producing INS-1E cells (9), we now identify this modification as citrullination and show that inflammation-induced citrullinated GRP78 is an autoantigen in type 1 diabetes. These findings suggest a novel role for GRP78 beyond its well-known function in the ER, leading to the loss of tolerance to β-cells in type 1 diabetes.

Western blotting, mass spectrometry, GRP78 cloning, expression, purification, and in vitro citrullination are available in the Supplementary Data.

Primary antibodies were as follows: mouse anti-poly(ADP-ribose) monoclonal antibody (mAb) (Enzo Life Sciences, Antwerp, Belgium), rabbit anti-GRP78 and anti-CHOP polyclonal antibody (pAb) (Santa Cruz Biotechnology, Santa Cruz, CA), rabbit anti-eIF2α pAb, anti–p-eIF2α(Ser51) pAb, anti-PERK mAb and anti–p-PERK(Thr980) mAb (Cell Signaling, Beverly, MA), and mouse anti-actin mAb (Sigma-Aldrich, Diegem, Belgium) for Western blotting; rabbit anti–β-catenin mAb (Cell Signaling Technology) and mouse anti-GRP78 pAb (Abcam, Cambridge, U.K.) for immunocytochemistry. Anti–cit(510)-GRP78 was raised in rabbits against the following peptides: C-aminohexanoic acid-IDVNGIL[citrulline]VTAEDKG-amide and acetyl-IDVNGIL[citrulline]VTAEDKG-aminohexanoic acid-C-amide, through five subsequent injections (21st Century Biochemicals). Specificity of the antibody was confirmed by dot blot against the citrullinated peptide and its native counterpart. Secondary antibodies were as follows: donkey anti-rabbit horseradish peroxidase, donkey anti-mouse Alexa Fluor 488, and donkey anti-rabbit Alexa Fluor 555 (Invitrogen, Merelbeke, Belgium). Ovalbumin and rabbit peptidylarginine deiminase (PAD) enzyme were from Sigma-Aldrich.

Rat INS-1E cells, a gift from Prof. Wollheim (CMU, Geneva, Switzerland), were cultured as previously described (9). INS-1E cells were exposed to recombinant rat IFNγ (500 units/mL; R&D Systems), recombinant human IL-1β (10 units/mL; R&D Systems), thapsigargin (Tg) (15 and 50 nmol/L; Sigma-Aldrich), tunicamycin (Tn) (2 and 5 µg/mL; Sigma-Aldrich), high glucose (HG) (25 mmol/L; Sigma-Aldrich), and palmitate (Pa) (0.5 mmol/L; Sigma-Aldrich). Mouse MIN6 cells, a gift from Dr. Miyazaki (Osaka University, Osaka, Japan), were cultured in DMEM (Invitrogen) containing 15% (volume for volume) FCS, 100 units/mL penicillin, 100 µg/mL streptomycin, and 70 µmol/L β-mercaptoethanol. MIN6 cells were exposed to recombinant mouse IFNγ (500 units/mL; R&D Systems) and human recombinant IL-1β (10 units/mL).

The percentage of living and apoptotic cells was assessed as previously described (9).

C57Bl/6 mice were obtained from Harlan Laboratories (Horst, the Netherlands) and nonobese resistant (NOR) mice from The Jackson Laboratory (Bar Harbor, ME). Nonobese diabetic (NOD) mice have been inbred in our animal facility since 1989 and are kept under semibarrier conditions. For all experiments, a mix of male and female mice was used. All animal manipulations were in compliance with the principles of laboratory care and approved by the Institutional Animal Ethics Committee of KU Leuven.

Pancreatic islets were isolated from 3- or 10-week-old C57Bl/6, NOD, and NOR mice. Islet isolation and culture were performed as previously described (16). C57Bl/6 islets were exposed to recombinant mouse IFNγ (1,000 units/mL) and recombinant human IL-1β (50 units/mL).

Immunofluorescence on isolated islets was performed as previously described (17). Fixed INS-1E cells or sectioned islets were incubated with primary antibody in 1% BSA for 1 h, followed by four washes in PBS before incubation with the secondary antibody in 1% BSA for another hour. Nuclei were detected with DNA-binding dye DRAQ5TM (Biostatus Ltd., Leicestershire, U.K.). Specificity was confirmed by including negative controls with secondary antibodies alone. INS-1E samples were observed under a Zeiss LSM 510 microscope using a Plan-Neofluar 40×/1.3 oil DIC lens. Images were acquired and processed using LSM 510 software (Carl Zeiss AG, Jena, Germany). Mouse islet sections were observed under a Nikon Eclipse Ti microscope using a Plan-Fluor 40×/0.75 DIC lens, and images were acquired and processed using Nis-Elements Viewer 4.20 software (Nikon Instruments Inc.).

INS-1E or MIN6 cells were incubated with the indicated stressors for 12–15 h and then treated as previously described (18).

To determine GRP78 concentration in conditioned media from INS-1E cells, the GRP78 ELISA kit (Enzo Life Sciences, Antwerp, Belgium) was used, according to the manufacturer’s protocol.

Two-dimensional gel electrophoresis (2D-GE) analysis was performed as previously described (9).

Quantitative RT-PCR was performed as previously described (19).

To determine PAD activity levels in islets and pancreata from C57Bl/6, NOR, and NOD mice, the antibody-based assay for PAD activity (ABAP) (ModiQuest Research, Oss, the Netherlands) was used, according to the manufacturer’s protocol.

Serum autoantibodies against two native and citrullinated GRP78 peptides (amino acids 500–519 TFEIDVNGILRVTAEDKGTG and amino acids 295–314 AKRALSSQHQARIEIESFYE) were determined by ELISA as previously described (20). For the citrullinated forms, arg510 or arg306 were replaced by citrulline (synthesized by PolyPeptide Laboratories, Strasbourg, France).

Splenocytes from 10-week-old and new-onset diabetic NOD mice and age-matched C57Bl/6 and NOR mice were cultured in flat-bottom 96-well plates (1 × 10⁶ cells/well) in the absence or presence of the indicated stimuli. IFNγ levels were measured in cell culture supernatant after 48 h of culture, using Meso Scale Discovery technology (Rockville, MD), according to the manufacturer’s protocol.

Statistical analyses of data were performed using GraphPad Prism 6 (GraphPad Software, San Diego, CA). Data are expressed as means ± SEM and were analyzed by a Kruskal-Wallis test followed by a Dunn multiple comparisons test, unless stated otherwise in the figure legend. P values <0.05 were considered significant.

We investigated cytokine-mediated regulation of GRP78 at both transcriptional and translational levels after 12–15 h exposure and at cytokine concentrations that induced only minor apoptosis (Fig. 1A). No significant increase in GRP78 mRNA (Fig. 1B) and total protein expression (Fig. 1C, center panel, and Fig. 1D) were observed as compared with control INS-1E cells. However, a detailed analysis of the plasma membrane fraction revealed a limited presence of GRP78 at the cell surface under basal conditions, which was increased upon IL-1β+IFNγ exposure, but not upon single cytokine exposure (Fig. 1C, bottom panel, and Fig. 1E). This was confirmed in MIN6 cells after 12 h treatment (Fig. 1F), a preapoptotic time point (Fig. 1G). In parallel with the increased plasma membrane translocation of GRP78, we also observed an increase in GRP78 secretion upon cytokine exposure of INS-1E cells (Fig. 1H).

These observations were confirmed by immunocytochemistry showing increased GRP78 staining on the plasma membrane of nonpermeabilized cytokine-exposed INS-1E cells (Fig. 2A), and a clear increase in colocalization between GRP78 and the plasma membrane marker β-catenin in cytokine-exposed INS-1E cells (Fig. 2B). Importantly, a similar membrane translocation was observed in cytokine-exposed C57Bl/6 mouse islets (Fig. 2C), again at an early and preapoptotic time point (Fig. 1I). Thus, surface expression of GRP78 upon cytokine exposure is not solely taking place in clonal β-cell models but also in primary β-cells, and is not a consequence of nonspecific changes associated with cell death, increasing the relevance of these findings.

Cytokine exposure of INS-1E cells and primary rat, mouse, and human β-cells is known to induce ER stress–dependent apoptosis (21–25). We further evaluated the contribution of ER stress to the observed cytokine-induced GRP78 membrane translocation by investigating the effect of the chemical ER stressors Tg and Tn. Cytokines (12–15 h) induced a clear activation of the PERK-eIF2α-CHOP pathway, at the dose tested (Supplementary Fig. 1). For Tg, there was a dose-response effect both in terms of apoptosis (Fig. 3A) and ER stress induction (Fig. 3B and Supplementary Fig. 1), with minor ER stress at 15 nmol/L but a clear induction of the PERK-eIF2α-CHOP branch at 50 nmol/L Tg, toward levels similar to those observed upon cytokine exposure. Of note, Xbp1 splicing was induced by Tg at 50 nmol/L, which was not the case upon cytokine exposure (Supplementary Fig. 1B). INS-1E cells were more resistant to Tn, with less apoptosis (Fig. 3A) and an intermediate activation of PERK-eIF2α-CHOP and Xbp1 splicing, both at 2 and 5 µg/mL (Supplementary Fig. 1), as compared with the higher doses of Tg and cytokines. As illustrated in Fig. 3C–E, 15 nmol/L Tg and 2 µg/mL Tn did not induce membrane translocation of GRP78. Importantly, the higher dose of 50 nmol/L Tg led to a clear membrane translocation, which was paralleled by more marked expression of ER stress markers (see above). A similar effect was observed with 5 µg/mL Tn, although less pronounced.

Finally, upon metabolic stress (Pa [0.5 mmol/L] or the combination of HG [25 mmol/L] + Pa), most of the ER stress markers were increased, although the increase in CHOP mRNA and protein was less marked than that observed with chemical ER stressors or cytokines (Supplementary Fig. 1). On the other hand, the total and membrane-associated GRP78 protein levels remained unaltered (Fig. 3F–H). Taken together, these findings indicate that GRP78 membrane translocation occurs upon inflammation- and severe chemical–induced ER stress, but not upon metabolic stress.

Besides the above described membrane translocation of GRP78, we observed extensive PTM of GRP78 upon cytokine exposure of INS-1E cells (Fig. 4A and previously described [9]) and C57Bl/6 mouse islets (Fig. 4B). This was also the case, although to a lesser degree, for IL-1β or IFNγ exposure alone (Fig. 4A). Of particular interest, 2D-GE analysis of the plasma membrane fraction of control and cytokine-exposed INS-1E (Fig. 4C) and MIN6 cells (Fig. 4D) not only demonstrated the presence of the three different cytokine-responsive GRP78 isoforms, but also showed a cytokine-mediated upregulation of numerous acidic GRP78 isoforms as compared with control cells. On the other hand, metabolic- and chemical-induced (both at low and high concentrations) ER stress did not induce detectable PTMs of GRP78 (Fig. 4E), confirming the specific effects of proinflammatory cytokines.

In order to identify the nature and site of cytokine-induced PTMs in GRP78, we subjected the three different GRP78 isoforms observed in INS-1E cells to mass spectrometry (MS) analysis. Sequence coverage ranged from 62.54 to 74.46% (n = 2). When comparing the resulting peptide profiles of isoform 1 (I1) versus isoform 2 (I2) and I1 versus isoform 3 (I3), one specific peptide, VTAEDKGTGNK (AA511–521), was identified exclusively in I1. This was confirmed by a quantitative differential analysis using trypsin digestion in combination with differential N-butyrylation and endoproteinase Lys-C (endo Lys-C) digestion combined with differential N-propionylation (Fig. 5A and B). As the peptide upstream of VTAEDKGTGNK could not be identified using both methods, we hypothesized the presence of a PTM in this region, possibly on Arg510, which would prevent tryptic digestion in I2 and I3, thereby rendering the resulting peptide (AA493–521) too long for retrieval and detection by MS. Of note, also a second peptide of interest was found, which was identified almost exclusively in I2 and I3, whereas hardly in I1 (AA307–324) (Fig. 5A and B). Since this would point to the presence of a PTM in the basic but not acidic forms, we did not further analyze the relevance of this peptide.

Based on these findings, we then investigated the nature of PTM present in the identified region. We did not succeed in retrieving by MS the longer, Arg510-containing peptide (AA493–521) or a spiked tryptic peptide with heavy label (AA493–521) used as internal control. This is probably caused by insolubility and inability to analyze on a C18 column and forced us to use instead specific enzymatic and antibody-based assays. We focused on three potential PTMs consistent with an acidic shift in isoelectric point without change in molecular weight and described to occur in GRP78 in other cell types, namely ADP ribosylation, phosphorylation, and citrullination (26–29). Both ADP ribosylation and citrullination occur on Arg residues, whereas phosphorylation may occur on Thr 518.

We initially investigated ADP ribosylation and phosphorylation of GRP78, but the absence of a positive signal in 2D Western blots of cytokine-exposed INS-1E cells with anti–poly(ADP-ribose) (Fig. 6A) and the absence of GRP78 staining by Pro-Q Diamond in control- and cytokine-exposed INS-1E cells (Fig. 6B), as well as persistence of the modified isoforms upon treatment with calf intestinal alkaline phosphatase (Fig. 6C) or λ-phosphatase (data not shown), argued against both modifications.

To verify the implication of citrullination, an antibody that specifically recognizes citrullinated GRP78 at Arg510 (selection based on the obtained MS data, see Fig. 5) was raised in rabbits. 2D Western blots from control and cytokine-exposed INS-1E cells with anti–cit510-GRP78 clearly indicated that the cytokine-induced acidic isoforms of GRP78 correspond to citrullinated GRP78 at residue Arg510, whereas no reactivity was observed against the most basic, nonmodified GRP78 isoform 1 (Fig. 6D).

Next, we investigated the potential role of citrullination in the diabetes-prone NOD mouse. Elevated Padi2 mRNA expression was observed in islets of 3- and 10-week-old prediabetic NOD mice as compared with islets from age-matched C57Bl/6 and NOR control mice (Fig. 7A and B). No differences between the strains were observed for Padi1, Padi3, Padi4, and Padi6 expression, which were either low or undetectable. Further analysis in other tissues revealed an overall very low expression of Padi2 in the immune-related tissues thymus, lymph nodes, and spleen. Except for kidney, no elevated levels of Padi2 were observed in the other tissues analyzed in NOD as compared with C57Bl/6 mice. In addition, NOR mice showed even lower/undetectable Padi2 expression in most of the tissues analyzed (Fig. 7C). Furthermore, elevated Padi2 mRNA expression in NOD islets corresponded to higher PAD activity in total pancreases (Fig. 7D and E) and islets (Fig. 7F and G) of 3- and 10-week-old NOD mice, compared with both C57Bl/6 and NOR mice, indicating a marked increase of PAD activity in islets of NOD mice immediately before and during insulitis. Of note, 3-week-old NOD mice did not show any sign of inflammation in the islets, as measured by IL-1β (Fig. 7H) and IFNγ (Fig. 7I) mRNA expression. In 10-week-old NOD mice, on the other hand, clear signs of immune infiltration were observed, as evidenced by high expression of IFNγ and IL-1β mRNA, confirming previous findings from our group (30).

To evaluate whether citrullinated GRP78 contributes to the autoimmune response in NOD mice, serum samples from prediabetic and new-onset diabetic NOD mice and age-matched C57Bl/6 and NOR mice were analyzed for the presence of autoantibodies against the native and citrullinated peptide containing the epitope of interest (p500–519). Serum levels of anti-GRP78 antibodies recognizing this citrullinated epitope were significantly higher in new-onset diabetic NOD mice as compared with age-matched C57Bl/6 or NOR mice (Fig. 8A, right panel). Furthermore, in diabetic NOD mice, significant higher serum antibody levels to the citrullinated peptide were found compared with those of the native peptide. No such differences were observed in the case of another irrelevant (citrullinated) GRP78 peptide (p295–314) tested (Supplementary Fig. 2). These findings provide evidence that this specific citrullinated epitope is important for autoantibody generation during type 1 diabetes development in NOD mice.

Next, to determine if NOD mice have autoreactive T cells against native or citrullinated GRP78, freshly isolated splenocytes from prediabetic and new-onset diabetic NOD mice and age-matched C57Bl/6 and NOR mice were stimulated with various concentrations of native and in vitro citrullinated recombinant mouse GRP78 protein. Secretion of IFNγ was used as a measure of effector T-cell activation. Whereas little to no effector T-cell activation was observed in the three different strains when culturing splenocytes with different concentrations of native GRP78 (Fig. 8B, red line and top right graph), a clear, dose-responsive increase in IFNγ secretion was observed when splenocytes from prediabetic and diabetic NOD mice were cultured in the presence of citrullinated GRP78 (Fig. 8B, blue line and bottom right graph). C57Bl/6 splenocytes were unresponsive to citrullinated GRP78, whereas a minor IFNγ response was detected in NOR splenocytes. Absence of an IFNγ response against both the PAD enzyme alone and the control protein ovalbumin, either native or in vitro citrullinated (at 0.1, 1, and 5 µg/mL) (data not shown), suggests that the observed autoreactive T-cell response is specifically generated against citrullinated GRP78.

The role of posttranslationally modified proteins is well established in several human autoimmune diseases, and evidence for similar phenomena in the development of type 1 diabetes is accumulating (12–14). Most importantly for this study, McGinty et al. (15) recently demonstrated the relevance of citrullination in patients with type 1 diabetes, by showing an increased response to citrullinated GAD65 peptides. We show that the ER chaperone GRP78 is citrullinated specifically upon exposure of β-cells to inflammatory stress. This is paralleled by translocation of GRP78 to the β-cell plasma membrane and eventually its secretion. Under these circumstances, a specific cross-talk between the β-cell and the immune system is initiated, resulting in the generation of autoantibodies and induction of T-cell autoreactivity against citrullinated GRP78 (Fig. 8C). Importantly, we also observed a marked upregulation of Padi2 in islets of NOD mice, providing a strong argument for PADI2 being the diabetes susceptibility gene in the recently identified Idd25 locus on mouse chromosome 4 (31) and adding to a potential role for citrullination in type 1 diabetes (15).

In previous studies, we have shown that GRP78 is posttranslationally modified in INS-1E cells exposed to cytokines (9), a PTM that we now identified as citrullination. Cytokines contribute to β-cell dysfunction and death at least in part through inducing ER stress (23–25). However, citrullination of GRP78 is not induced by chemical ER stressors (Tg or Tn) or by metabolic stress via exposure to HG and/or Pa, suggesting that cytokine-induced GRP78 citrullination occurs through a mechanism independent of its ER stress–inducing capacities. Preliminary data suggest that this citrullination is not mediated by direct upregulation of the PADI2 enzyme by cytokines (data not shown) but needs to be the consequence of increased PAD activity in β-cells exposed to cytokines. Since PAD activity is highly Ca²⁺ dependent, changes in Ca²⁺ fluxes induced in cytokine-exposed β-cells might play a role. The present findings, together with the recent report on posttranslationally modified GAD65 (15), add type 1 diabetes to the list of autoimmune diseases involving citrullination, i.e., RA, multiple sclerosis, and systemic erhythematosus (32). This suggests that citrullination is not a specific disease-related event but rather an inflammation-dependent process occurring preferentially in autoimmune target tissues. The role of citrullination in the induction of autoantigenicity has been best described in RA, with several citrullinated autoantigens, including GRP78, already identified (20,33). These citrullinated peptide epitopes are better accommodated in the HLA pocket of HLA-DR4–type individuals, determining the strength of the immune response to citrullinated peptides and providing a molecular basis for the genetic predisposition of HLA-DR4 individuals to RA (34). This mechanism has recently also been described in type 1 diabetes (15), where HLA-DR4 is an important risk allele for the disease (35).

In addition to citrullination, cytokine-induced translocation of GRP78 to the β-cell plasma membrane may be a crucial step for GRP78 to become an autoantigen. This translocation is shown to be an early event in response to inflammation, suggesting that it is an active process at least in part independent from “protein leakage” by apoptotic β-cells. GRP78 membrane translocation, but not citrullination, was found to be ER stress dependent and paralleled to the increased CHOP expression. Membrane-associated GRP78 has been described in different tumor cell types (36–38) as well as in exocrine pancreatic cells (39) and proliferating endothelial and monocytic cells (40,41). In these models, membrane-associated GRP78 acts, among other functions, as a cell surface signaling receptor for different ligands such as activated α₂-macroglobulin (42,43) and coxsackie A9 virus (44,45) and is found associated with the major histocompatibility complex class I (MHC-I) (44). Of interest, changes in the topography of membrane-associated GRP78, caused by PTMs, may convert GRP78 into a receptor with autoantigenic properties. This has been observed in cancer cells where GRP78 is modified by O-linked glycosylation (38), leading to the generation of GRP78 autoantibodies. The present observations that inflammation induces extensive GRP78 citrullination, translocation to the plasma membrane, and secretion underscore its putative function as an autoantigen in type 1 diabetes. Furthermore, based on the proposed transmembrane model for GRP78 (46), the reactive p500–519 epitope, against which GRP78 autoantibodies are generated in NOD mice, is located in the extracellular domain, thus being exposed to infiltrating immune cells.

The possible role for citrullinated GRP78 as an autoantigen in type 1 diabetes is supported by the present observations showing the generation of autoantibodies as well as CD4⁺ T-cell autoreactivity against citrullinated GRP78 in NOD mice. No T-cell reactivity is observed in C57Bl/6 mice, whereas NOR mice show a minor T-cell response. This supports the idea that inflammation is necessary to initiate this process, as low levels of IFNγ and IL-1β are detected in islets of 10-week-old NOR mice, a phenomenon referred to as protracted, noninvasive insulitis (47). A marked upregulation of PAD2, a key enzyme for protein citrullination, is observed exclusively in islets from NOD mice, as compared with NOR and C57Bl/6 islets. PAD activity is further increased in NOD islets with increasing age and insulitis, suggesting that inflammation plays a role for this phenomenon, perhaps by increasing cytosolic Ca²⁺ concentrations due to cytokine-mediated calcium depletion from the ER (24). Interestingly, expression of Padi2 was very low in NOD thymus and not different from C57Bl/6 and NOR thymus. This may explain the escape of citrullinated GRP78 from thymic tolerization in developing thymocytes, thereby clarifying why citrullinated GRP78 can be recognized as a β-cell–specific autoantigen whereas native GRP78 is ubiquitously expressed. A similar mechanism has been proposed for chromogranin A, where exposure of a naturally occurring cleavage product (peptide WE14) to transglutaminase, expressed in β-cells, forms high- and low-molecular weight aggregates, thus rendering the peptide highly antigenic in NOD mice (13).

Whether these observed processes are also applicable to other ER chaperones or heat shock proteins, such as HSP60, requires further investigation. It will be of utmost importance to determine whether similar processes are involved in human type 1 diabetes. The knowledge that inflammation-mediated β-cell stress is taking place in human type 1 diabetes (48) and the high overlap between autoantigens identified to date in NOD mice and type 1 diabetes patients (49) are in support of this hypothesis.

In conclusion, local inflammation in the islets induces citrullination of GRP78 in the stressed β-cells, turning citrullinated GRP78 into an autoantigen. This modified GRP78 is recognized by both B and T cells, thus propagating and amplifying the ongoing autoimmune attack against the β-cells. Our findings support and provide mechanistic evidence for the concept that inflammation-induced β-cell stress initiates a specific communication between the β-cell and the immune system, which will aggravate and accelerate the development of type 1 diabetes. A genetic predisposition for increased citrullination in the islets, as shown here for NOD mice, is expected to further exacerbate this process. This proposed mechanism (Fig. 8C), implicating tissue-specific and inflammation-induced protein modification, cell surface translocation, and secretion, would also explain why different tissue-specific autoimmune diseases can have similar autoantigens.

Acknowledgments. The technical assistance of Frea Coun, Martine Gilis, Jos Laureys, Willem Van Den Berghe, Wim Werckx, and Farah-Deborah Lok (Laboratory of Clinical and Experimental Endocrinology, KU Leuven) is greatly appreciated. The authors thank Katleen Lemaire (Gene Expression Group, KU Leuven) and Monique Beullens (Laboratory for Biosignaling and Therapeutics, KU Leuven) for advice on GRP78 cloning in pET vector. The authors thank the Cell Imaging Core (KU Leuven) for providing technical assistance with confocal microscopy.

Funding. This work was supported by Juvenile Diabetes Research Foundation International (17-2012-129 and 17-2013-515), the European Community’s Seventh Framework Programme NAIMIT (Natural Immunomodulators as Novel Immunotherapies for Type 1 Diabetes) under grant agreement 241447, the KU Leuven (Geconcerteerde Onderzoeksactie GOA 12/24 and an F+ fellowship for D.R.), and the Flemish Research Foundation (G.0619.12, a postdoctoral fellowship for G.B.F. and W.D. and a clinical research fellowship for C.M.).

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

Author Contributions. D.R. and I.C. designed and performed research, analyzed data, and wrote the manuscript. W.D., G.B.F., A.S., and A.D.G. performed research. D.L.E. analyzed data and wrote and edited the manuscript. P.A. and K.G. edited the manuscript. L.O. and C.M. designed research and wrote and edited the manuscript. L.O. and C.M. 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.