β-Cell Stress Shapes CTL Immune Recognition of Preproinsulin Signal Peptide by Posttranscriptional Regulation of Endoplasmic Reticulum Aminopeptidase 1

Type 1 diabetes (T1D) results from selective and progressive destruction of insulin-producing cells by autoreactive CD8⁺ T cells (CTLs) (1,2). Immunohistochemistry of insulitic pancreases obtained through the Network for Pancreatic Organ Donors with Diabetes program (nPOD) have shown infiltration of immune cells and an increased expression of markers characteristic of the unfolded protein response (UPR) (3–6). Altogether, these results suggest an association between endoplasmic reticulum (ER) stress and the increased visibility of β-cells to the immune system (7–9). While several proteins have been identified as potential autoreactive T-cell targets, evidence from mouse and human studies have suggested that insulin protein itself could be the main and primary autoantigen targeted by infiltrating CTLs (10–12). The posttranslational processing pathway that generates insulin from its precursor molecule preproinsulin (PPI) is well established: the signal peptidase Sec11 cleaves off the signal peptide (SP) cotranslationally upon translocation of the protein into the ER via the translocon Sec61 (13). After folding and formation of disulfide bonds, proinsulin is transported via the Golgi system into immature secretory vesicles where mature insulin is generated by the combined action of prohormone convertases and release of the C-peptide central region (13–15). Although being the least studied domain of the PPI molecule, increasing evidence highlights the importance of the 24 amino acids–long insulin SP sequence as a major source of insulin-derived class I epitopes (16). Peptide elution experiments performed on HLA-A2 purified from surrogate β-cells have led to the identification of prominent HLA class I binders derived from the SP domain (17), and PPI_15–24–directed CTLs were highly prevalent in patients with recent-onset T1D (18). Recently, the importance of the combined action of the SP peptidase and ER aminopeptidase 1 (ERAP1) in the trimming of the PPI SP and in the generation of PPI SP–derived epitopes was demonstrated using cell-free translocation assay and CRISPR/Cas technology (19). While these data match with the substrate preference of ERAP1 (20,21), the link between pathophysiological features of T1D and the immunoreactivity against PPI_15–24 remains unclear.

Here, we investigated the effect of inflammation and ER stress on ERAP1 gene expression in human β-cells and show that proinflammatory cytokines modulate ERAP1 expression both at transcriptional and posttranscriptional level though interaction with miR-17. More importantly, we show in primary human islets that during inflammation, the specific inhibition of IRE1α endoribonuclease activity reduces ERAP1 expression and limits SP-derived epitope presentation to PPI autoreactive CTLs. Altogether, these data establish a direct connection between ER stress and β-cell immunogenicity and underline the importance of ER sensors in shaping antigenic peptide presentation to diabetogenic CTLs.

PPI cDNA was obtained from reverse transcriptase reaction of total human islets RNA extraction using the following primers: PPI-Full (forward): ATG GCC CTG TGG ATG CGC CTC CTG CCC; PPI-Full (reverse): GTT GCA GTA GTT CTC CAG CTG GTA GAG GGA GCA. Lentivirus (LV)-cytomegalovirus (CMV)-PPI was generated by insertion of the coding region of the PPI cDNA into pLV-CMV-bcGFP. For miRNA reporter construct, the following primers were annealed and cloned into pMIR-REPORT (Thermo Fisher Scientific) open with PmeI. ERAP untranslated region (UTR) forward: 5′-GTA ATT TGA ATA TAG ACA CAA TGC ACT TTA TTG CAC TTT CAA TTC TTA TAA AGC; ERAP UTR reverse: 5′-GCT TTA TAA GAA TTG AAA GTG CAA TAA AGT GCA TTG TGT CTA TAT TCA AAT TAC; ERAP UTR mutated (mut) forward: 5′-GTA ATT TGA ATA TAG ACA CAA TGC ACT TTA TTG TGC TTT CAA TTC TTA TAA AGC; and ERAP UTR mut reverse: 5′-GCT TTA TAA GAA TTG AAA GCA CAA TAA AGT GCA TTG TGT CTA TAT TCA AAT TAC. The ERAP1 promoter reporter construct was generated by inserting a XhoI/HindIII fragment containing the −1,325/+60 region of the ERAP1 promoter into pGL3 vector (Promega, Madison, WI). The promoter region was cloned from human genomic DNA using the following primers: ERAP promoter forward: 5′-CCC TCG AGG TCA CAG AAT GAG ATA GAA GGT AGG CAC AAG-3′ and ERAP promoter reverse: 5′-GGA AGC TTC CTA CCC GCG GCT CGA GCG CGC TGT ACC TGG-3′. The underlined sequences represent the restriction sites used for cloning. The constructs were verified by sequencing.

HEK 293T cells were grown in high-glucose DMEM supplemented with 10% (v/v) heat-inactivated FBS (Gibco BRL) and penicillin/streptomycin at 37°C, 5% CO₂. PPI_15–24–directed CTLs (17) were maintained in Iscove's Modified Dulbecco's Medium supplemented with 10% human serum, interleukin (IL)-2 and IL-15 and restimulated every 14 days with irradiated JY cells (pulsed with 2 μg/mL PPI_15–24 peptide) at a 1:1 ratio and irradiated peripheral blood mononuclear cells of five different donors (at a ratio of 1:5) in the presence of IL-2, IL-15, IL-l7, IL12, and leuco-A. PPI_15–24 was synthesized using solid-phase fluorenylmethyloxycarbonyl (Fmoc) chemistry and analyzed by reverse-phase high-performance liquid chromatography (RP-HPLC) mass spectrometry for purity and identity. HEK293T cells are HLA-A2⁺ and routinely used in our laboratory as a target for HLA-A2–restricted CD8 clones (22). Although many differences could be expected between cell line and primary β-cells, EndoC-βH1 cells remain the best material available to study human β-cell biology (23). EndoC-βH1 cells, obtained from Dr. Raphael Scharfmann (Paris Descartes University, France) (24), were maintained in low-glucose DMEM supplemented with 5.5 μg/mL human transferrin, 10 mmol/L nicotinamide, 6.7 ng/mL selenite, 50 μmol/L β-mercaptoethanol, 2% human albumin, 100 units/mL penicillin, and 100 μg/mL streptomycin. Cells were seeded in extracellular matrix, fibronectin precoated culture plates. Inflammatory stress was induced by a mixture of 1,000 units/mL interferon-γ (IFN-γ) and 2 ng/mL IL-1β for 24 h. Staurosporine was used at 100 nmol/L for 1 h and thapsigargin was used at 100 nmol/L for 24 h. MKC3946 inhibitor was used at 10 μmol/L for 24 h in our assay, unless differently stated. The inhibitor was added simultaneously to other treatments.

Transient transfection of miRNA mimics and DNA vectors were performed using Lipofectamine 2000 (Invitrogen, Karlsruhe, Germany) according to manufacturer’s protocol. EndoC-βH1 cells were transfected in a 12-well plate, using a final concentration of 200 nmol/L total miRNA precursor. The Homo sapiens (hsa)-miR-17-5p precursor (PM12412; Ambion), premiR #1 (negative control 1, AM17110; Ambion) and hsa-miR-17-5p inhibitor (AM12412; Ambion) were used. Experiments were continued 24 h after transfection. For the validation of miRNA targeting sites, 293T cells were cotransfected in a 96-well plate with 125 ng pMiR-luciferase (luc)-ERAP1-wild-type (wt)UTR or pMIR-luc-ERAP1-mutUTR and 50 nmol/L miRNA precursor hsa-miR-17-5p precursor or premiR hsa-miR-1 per well. Transfections were performed in triplicate, and cells were analyzed 24 h after transfection. For the evaluation of ERAP1 transcriptional regulation, EndoC-βH1 cells were cotransfected with the ERAP1 promoter reporter construct and a lentiviral vector including a GFP gene under the control of a CMV promoter to estimate transfection efficiency. Transfections were performed in a 96-well plate with 150 ng total DNA per well. Transfected cells were treated 24 h later with thapsigargin (100 nmol/L) or 1,000 units/mL IFN-γ and 2 ng/mL IL-1β for 24 h, and luciferase activity was measured.

Cells were lysed in luciferase lysis buffer (125 mmol/L Tris/HCl [pH 7.8], 10 mmol/L cyclohexane-1,2-diaminetetraacetic acid, 10 mmol/L dithiothreitol, 50% [v/v] glycerol, and 5% [v/v] Triton X-100]. Luciferase activity was determined by luminometry using the Luciferase Assay Reagent (Promega). β-Galactosidase activity was determined by luminometry using the Galacto-Light Dual-Light kit (Tropix). Light emission was determined using the Lumat LB9501 luminometer (Berthold, Bad Wildbad, Germany).

The vectors were produced as described previously (25). Briefly, the lentiviral backbone containing the gene of interest and the three helper plasmids (encoding HIV-1 gag-pol, HIV-1 rev, and the VSV-G envelope protein) were cotransfected overnight using the calcium phosphate method into 293T cells. The medium was refreshed, and viruses were harvested after 48 and 72 h, passed through 0.45-µm filters, and stored at −80°C. Virus was quantified by antigen capture ELISA measuring HIV p24 levels (ZeptoMetrix, New York, NY) as described (26). Viral supernatants were added to fresh medium supplemented with 8 mg/mL Polybrene (Sigma-Aldrich), and the cells were incubated overnight. The next day, the medium was replaced with fresh medium. Transduction efficiency was analyzed 3–6 days after transduction.

shRNA lentiviral constructs for ERAP1 knockdown were obtained from the Mission shRNA library (clones TRCN060539, TRCN060540, TRCN060541, and TRCN060542; Sigma-Aldrich). Based on preliminary assays to assess knockdown efficiency (data not shown), we selected the TRCN060542 clone for further use. The shERAP1-encoding lentivirus was produced as described above.

Total RNA was extracted from cultured cells using TRIzol reagent following the manufacturer’s instructions. Isolated RNA was quantified using a Nanodrop 1000 spectrophotometer. Approximately 500 ng RNA was reverse transcribed using the Superscript RT II kit (Invitrogen). Expression of the genes interest was detected using the following primers: insulin forward: GCA GCC TTT GTG AAC CAA CA, insulin reverse: CGG GTC TTG GGT GTG TAG AAG; ERAP1 forward: GAA AAC CAT GAT GAA CAC TTG G, ERAP1 reverse: CCA CCT CTT CTG GGA GGA TGA G; GAPDH forward: ACA GTC AGC CGC ATC TTC TT, GAPDH reverse: AAT GAA GGG GTC ATT GAT GG; XBP1s forward: 5′-CTG AGT CCG CAG CAG GTG-3′, XBP1s reverse: 5′-GAG ATG TTC TGG AGG GGT GA-3′; and ERAP2 forward: GGG GCT TTC CCA GTA GCC ACT AAT GG, ERAP2 reverse: GAA TCT TCC TCT GAC TGA AGG GTG GC. PCRs were performed on a PTC-200 (Biozym, Landgraaf, the Netherlands) using the following conditions: 94°C for 5 min; 35 cycles of 30 s at 94°C, 30 s at 60°C, and 1.5 min at 72°C; 10 min at 72°C. Real-time PCRs were performed in triplicate using the SybrGreen master mix kit (Applied Biosystems, Nieuwerkerk a/d IJssel, the Netherlands) and an Applied Biosystems Step One Plus. Comparative ΔΔ cycle threshold values were performed using the GAPDH gene as reference. Values are represented as mean ± SE.

For miRNA quantification, total RNA was reverse transcribed using TaqMan MicroRNA Reverse Transcription Kit (Applied Biosystems) and detected using hsa-miR-17 TaqMan miRNA assays (PN4427975; Applied Biosystems) and TaqMan 2× Universal PCR Master Mix, no uracil N-glycosylase (Applied Biosystems) according to the manufacturer’s instructions. miRNA expression was normalized to RNU6 using the following primers: RNU6 forward: 5′-CTC GCT TCG GCA GCA CA-3′; and RNU6 reverse: 5′-AAC GCT TCA CGA ATT TGC GT-3′.

Cells were lysed in buffer containing 50 mmol/L Tris-HCl (pH 7.4), 250 mmol/L NaCl, 0.1% Triton X-100, and 5 mmol/L EDTA. The lysis buffer was supplemented with protease inhibitor cocktail (Roche). Protein quantification was performed with the BioRad protein assay reagent. For ERAP1 analyses, 50 μg proteins extracts were loaded on 12% acrylamide/bis acrylamide SDS page gel. After electrophoresis, protein transfer was performed on a nitrocellulose membrane (GE Healthcare). Membranes were stained with anti-ERAP1 overnight at 4°C (B10; Sc-271823; Santa Cruz), and goat anti-mouse IgG horseradish peroxidase (sc-2005; Santa Cruz Biotechnology) was used for the detection. The loading control was β-actin (MAB1501; EMD Millipore).

For intracellular insulin staining, cells were fixed at 4°C with 2% paraformaldehyde PBS for 20 min, followed by 10-min incubation at 4°C with permeabilization buffer (0.5% saponin, 2% BSA, PBS). Rabbit anti-insulin (sc-9168; Santa Cruz Biotechnology) and anti–rabbit-phycoerythrin (1:500 dilution) (#111,116,144; Jackson ImmunoResearch) were used. All antibody incubations were performed in the permeabilization buffer for 30 min at 4°C. For surface staining, cells were incubated on at 4°C for 30 min with monoclonal mouse anti-human HLA-ABC Antigen/RPE Clone W6/32 (R7000; DAKO) or with mouse anti-human CD8 allophycocyanin (555369; BD Pharmigen) in the flow cytometry buffer (0.5% human albumin, 0.01% Na azide, PBS). Analysis was performed using BD LSR II (BD Biosciences) and Flowjo software. A total of 10,000 events were recorded.

Pancreatic islets were obtained from human organ donor pancreata. Human islets were isolated from organ donors. Islets were only studied if they could not be used for clinical purposes and if research consent had been obtained. According to the national law, ethics approval is not required for research on donor tissues that cannot be used for clinical transplantation. The isolations were performed in the Good Manufacturing Practices-facility of Leiden University Medical Center according to the previously described protocol (27). For experimental use, human islets were maintained in ultra-low attachment plates (Corning, Corning, NY) in low-glucose DMEM supplemented with 10% FBS, 100 units/mL penicillin, and 100 μg/mL streptomycin. Dispersed islet cells were treated with 1,000 units/mL IFN-γ and 2 ng/mL IL-1β for 24 h to induce ER stress. All methods were performed in accordance with relevant guidelines and regulations.

Islet preparations were washed with PBS and dispersed by trypsinization in an ultra-low attachment 6-well plate. The day after, cells were counted and seeded in a 96-well plate at a concentration of 50,000 (for donor R153 and 615) or 200,000 (for donor R155) islet cells per well. Treatment with proinflammatory cytokines and MKC3946 was performed as described for 24 h. On the 3rd day, HLA-A2-PPI_15–24–specific CTLs were added at a ratio 5E:1T (for donor R153 and 615) or 1E:2T (for donor R155) in the presence of mouse anti-human CD107a- FITC (11-1079-42; Thermo Fisher Scientific) was added. Cocultures were incubated at 37°C for 2 h. Cells were washed with flow cytometry buffer (0.5% human albumin, 0.01% Na azide, PBS), stained for CD8, and analyzed by flow cytometry, as mentioned above. The absolute degranulation (D) capacity of T cells was calculated as a ratio of (percentage of CD107a⁺ cells/percentage of total CD8⁺ cells) (data not shown). The relative degranulation was estimated by calculating the percentage of change of the absolute degranulation of the treated samples compared with their respective controls. More specifically for the cytokine (CYT)-treated sample: ([]/⁠) ∗ 100. For the cytokine plus MKC3946 sample: ([]/⁠) ∗ 100.

The supernatant was used for detection on MIP-1β production by the T cells, using the MIP-1β ELISA kit (BMS2030INST; Thermo Fisher Scientific), according to the manufacturer’s protocol.

miRNA interaction with ERAP1 mRNA has been evaluated by crossing information available at miRbase v21.0 (https://www.mirbase.org/) and Tarbase v8 (https://carolina.imis.athena-innovation.gr/diana_tools/web/index.php?r=tarbasev8/index) using β-cell–specific miRNAs databases, as previously published (28,29). A Venn diagram was generated using an online Venn diagram plotter (https://omics.pnl.gov/software/venn-diagram-plotter). Data are presented as mean ± SEM. Calculations were performed using GraphPad Prism 7 software. Unpaired t tests were performed for all comparisons, except from the experiments of T-cell activation upon coculture with dispersed islets (Fig. 4E and F), where a paired t test between samples of the same islet donor was performed. P values of ≤0.05 were considered statistically significant.

The data sets generated during and/or analyzed during the current study are available from the corresponding author upon reasonable request.

The localization of insulin SP within the ER and the TAP (Transporter associated with antigen processing) independent routing of the PPI_15–24 have suggested proteasome-independent degradation mechanisms (17). The presence of three alanine residues at the C-terminal part of the insulin SP represents a high-affinity binding motif for the hydrophobic pocket of ERAP1, and the two leucine residues in position 13–14 make this region a suitable substrate for ERAP1 trimming (Fig. 1A) (20,30). To test for the implication of ERAP1 on PPI SP processing and to evaluate the consequences for antigenic peptide recognition by CTLs, surrogate β-cells were generated by genetically modifying HEK 293T cells with a lentiviral vector containing the full-length PPI cDNA, driven by a CMV early enhancer and promoter. After verification of the increased insulin gene expression in these cells (Fig. 1B), we used an ERAP1-directed shRNA containing lentivirus to specifically downregulate ERAP1 gene expression. Transduction of ERAP1 shRNA (multiplicity of infection = 1) led to a 90% reduction in the ERAP1 mRNA level compared with nontarget control shRNA (Fig. 1C). Interestingly, ERAP1 downregulation in these experiments increased significantly ERAP2 gene expression (Fig. 1D) but had no impact on (pro)insulin expression or HLA class I surface expression (Fig. 1E and F). Yet, coculture of modified cells with PPI_15–24–specific, HLA-A0201-restricted CTLs showed a reduced T-cell activation as measured by T-cell degranulation (exposure of CD107a at the cell surface) (Fig. 1E) and a reduced MIP-1β secretion (Fig. 1F). These data support the role of ERAP1 in the processing of the PPI_15–24 epitope from human PPI (14).

In the course of diabetes development, proinflammatory cytokines secreted by infiltrating immune cells are believed to promote autoimmunity by inducing ER stress in β-cells (7,31–33). To evaluate the impact of the T1D pathophysiological condition on ERAP1 gene expression, EndoC-βH1 cells were exposed to a mixture of proinflammatory cytokines (i.e., IFN-γ and Il-1β) or to a sarcoplasmic reticulum Ca²⁺-ATPase pump blocker (i.e., thapsigargin) as an ER stress inducer. Although the stress response of EndoCBH1 and primary human β-cells differs (34), cytokine treatment and thapsigargin stimulation led to increased expression of marker from the unfolded protein response in EndoCBH1 (Supplementary Figs. 1 and 2). In these conditions, the increased stress status, as measured by XBP1 splicing (Fig. 2A), correlated with an upregulation of ERAP1 expression both at the gene and protein level (Fig. 2B and C). To characterize the underlying regulatory mechanisms, we generated a luciferase reporter construct to assess the transcriptional regulation of the ERAP1 gene by introducing the ERAP1 promoter region (−1,325/+60) (35) upstream of a luciferase encoding sequence (Fig. 2D). After transfection of the reporter construct in EndoC-βH1 and stress treatment induction, a threefold increase in light emission after IFN-γ/IL-1β treatment was observed, which confirms the transcriptional regulation by IFN-γ (36). However, the absence of transcriptional activation of the ERAP1 promoter after thapsigargin treatment in our assay suggested additional regulatory mechanisms controlling ERAP1 gene expression during ER stress.

To evaluate a possible posttranscriptional control of ERAP1 mRNA by miRNAs, we searched for potential miRNA binding sites within the ERAP1 3′-UTR region. Using β-cell–specific miRNA data sets (37) and two different algorithms for prediction of miRNA binding sites (miRbase and TarBase [28,38]) (Supplementary Table 1), six putative miRNAs were identified that could be expressed by β-cells and destabilize ERAP1 mRNA expression (Table 1 and Fig. 2E). Notably, these miRNAs belong to the miR-17 family, implying a conserved mechanism of regulation. To confirm these in silico findings, luciferase reporter constructs were generated harboring a 54-nucleotide-long fragment of the ERAP1–3′-UTR carrying the predicted miR-17 binding site as well as a mutant construct containing 2 nucleotides substitution preventing the formation of the miRNA-induced silencing complex. Transient transfection of miR-17 in HEK293T luciferase reporter cells reduced luciferase activity from the native UTR reporter construct but had no effect on the luciferase mutant construct (Fig. 2F). To validate that miR-17 expression negatively affects ERAP1 mRNA in β-cells, we forced expression of miR-17 in EndoC-βH1 and determined ERAP1 gene expression by quantitative (q)PCR. Transient transfection of miR-17 resulted in a 30% reduction of ERAP1 gene expression, while cotransfection of the complementary anti-miRNA (anti–miR-17) reversed the inhibitory effect of miR-17 on ERAP1 gene expression (Fig. 2G). Of note, insulin gene expression in these assays remained stable (Fig. 2H). Altogether, these results reveal that miR-17 contributes to ERAP1 regulation by direct interaction with the ERAP1 3′-UTR region. To connect inflammation-induced ER stress and miRNA regulation, EndoC-βH1 cells were treated with cytokines, and miR-17 expression was determined by TaqMan assay. Consistently with previous reports (39), we observed a >20% decrease in miR-17 expression after 24-h treatment, indicating that these cytokines may modulate ERAP1 expression by affecting the miR-17 level (Fig. 2I). These results have been confirmed after chemically induced stress by thapsigargin (Supplementary Fig. 2B). In line with these results, transfection of miR-17 in EndoC-βH1 blunted proinflammatory cytokine induction of ERAP1 (Fig. 2J) without affecting insulin gene expression (Fig. 2K).

Previous studies have identified miR-17 as an important link between the endoribonuclease IRE1α and the induction of β-cell death via activation of the thioredoxin-interacting protein (TXNIP) (39). To determine whether the ERAP1 posttranscriptional regulation in stress conditions was mediated by IRE1α, EndoC-βH1 β-cells were treated with cytokines and XBPs, and miR-17 and ERAP1 gene expressions were monitored for 24 h in a time-course experiment (Supplementary Fig. 2A). As anticipated, IFN-γ/IL-1β stimulation led to decreased expression of miR-17 and increased expression of both XBPs and ERAP1. Interestingly, specific inhibition of IRE1α, by MKC3946 treatment (40) increased expression of miR-17 both at 6 h and 24 h in the homeostatic condition and prevented its degradation in cytokine conditions (Fig. 3A–C). Under these conditions, cytokines-induced ERAP1 gene expression was partially reduced by MKC3946 cotreatment at 6 h (Fig. 3B). This observation was more pronounced after 24 h of treatment, where inhibition of the cytokine-induced ERAP1 gene expression reached 40% (Fig. 3D) without any effect on cells viability (Supplementary Fig. 3A). IRE1α inhibitor dose-response experiments performed in EndoC-βH1 illustrated the direct correlation between IRE1α activity (as assessed by spliced XBP1 expression) and ERAP1 gene expression (Fig. 3E and F). Chemically induced stress led to similar changes in gene expression (Supplementary Fig. 2B); also in these conditions, the effect of thapsigargin on ERAP1 expression was inhibited by MKC3946 cotreatment (Supplementary Fig. 2C). More importantly, similar experiments performed in freshly isolated human islets confirmed that the increased ERAP1 expression mediated by cytokines or thapsigargin could be counteracted by MKC3946 treatment (Fig. 4A and B).

To determine the consequences of our findings on the insulin SP trimming, presentation, and subsequent CTLs activation, we examined whether IRE1α inhibition affects immune recognition of the PPI_15–24 SP–derived epitope presented by primary human β-cells. After dispersion, isolated human islet cells (HLA-A2) were maintained for 24 h in the presence or absence of IFN-γ/IL-1β and treated with MKC3946. After treatment, islet cells were cocultured with the PPI_15–24–specific, HLA-A0201–restricted CTL clone for 2 h before the analyses (Fig. 4C). We verified that MKC3946 had no deleterious effect on T-cell function, no direct effect on CD107a surface expression (Supplementary Fig. 3B–D), and no effect on HLA class I surface expression on islet cells (Fig. 4D), and determined T-cell activation in the different conditions by measuring T-cell degranulation and MIP-1β secretion. We demonstrated in islet preparations from three different pancreas donors that CD107a surface expression in CD8⁺ cells was significantly increased after coculture with islet cells treated with cytokines, illustrating an increased surface density of the specific peptide HLA ligand at the β-cell surface. Furthermore, we showed that MKC3946 cotreatment abrogates the deleterious effect of cytokines on T-cell degranulation (Fig. 4E). Quantification of the MIP-1β production in the cell supernatant confirmed the effect of the IRE1α inhibition on PPI_15–24 peptide processing (Fig. 4F), supporting the notion that ER stress is involved in shaping CTL responses in T1D.

We demonstrate that ERAP1 is critical for the generation of a prevalent PPI-derived autoantigenic peptide and present a new regulatory mechanism connecting ER stress and increased β-cell visibility to the immune system.

The upregulated ERAP1 expression observed both in isolated human islets and in a human β-cell line after cytokine stimulation or chemically induced stress extends proof for the implication of ERAP1 in PPI SP processing/presentation (19) and its participation in β-cell destruction during inflammation in patients with T1D. Several leucine amino peptidases induced by INF-γ have been shown to participate in peptide trimming (41); however, few have been shown to be located specifically within the ER, where processing of the insulin SP is likely to occur. Without excluding the participation of other ER resident endopeptidases to the process, the reduced T-cell activation and degranulation of PPI_15–24 CTL observed after coculture with ERAP1 knockdown surrogate β-cells or human islet cells treated with the IRE1α inhibitor is in line with a previous study targeting ERAP1 expression by specific siRNA (19). Yet, whether the observed reduction in T-cell activation is sufficient for protecting β-cells from CTL-mediated attack remains to be established. As in a previous study (42), the reduced expression of ERAP1 led to increased expression of ERAP2 in our assays. On the view of the complementarity role of these two aminopeptidases (43), this compensatory effect may explain the absence of effect of the ERAP1 knockdown on HLA class I surface expression. Similar experiments performed in human and mouse cells have shown a maximal 10% reduction after ERAP1 downregulation by siRNA (44,45). Although the effect of the ER stress inhibitor on adhesion and costimulatory molecules has not been studied in depth, the lack of effect of MKC3946 on HLA-A/-B/-C expression in human islet cells suggests that the differential T-cell activation observed is a direct consequence of the relative density of PPI_15–24/HLA complexes at the cell surface.

ERAP1 has been initially described as an IFN-γ–responsive gene (46). Although we cannot exclude the presence of an ER stress responsive element in an enhancer region more upstream to the promoter core region chosen in our transcriptional-based assays, our results are pointing to two distinct regulatory mechanisms during inflammation: a transcriptional regulation mediated by IFN-γ–responsive elements (i.e., IRF1 binding sites) within the ERAP1 promoter proximal region and an ER stress-dependent posttranscriptional regulation. In fact, our data using the sarcoplasmic reticulum Ca²⁺-ATPase pump blocker thapsigargin show additional regulatory mechanisms of ERAP1, initiated by the ER stress sensor IRE1α and involving a posttranscriptional control of ERAP1 mRNA. The results obtained in human islet cells exposed to IRE1α inhibitor, where MKC3946 treatment completely abolishes the chemically induced ER stress–mediated ERAP1 expression but only partially impacts (∼50%) cytokine-induced expression, perfectly illustrate this dual regulation.

Several miRNAs have been implicated in the response to stress (47); among these, miR-17 has been shown to be a master regulator of β-cell apoptosis by controlling the thioredoxin-interacting protein for type 1 and type 2 diabetes (39,48). Interestingly, TXNIP regulation was demonstrated to be dependent of both PERK and IRE1α sensors (49). The incomplete downregulation of ERAP1 in our experiments after treatment with MKC3946 inhibitor may also imply that, as for TXNIP, multiple arms of the unfolded protein response could be involved in ERAP1 regulation.

In this study, we confirmed the complexity of antigenic peptide generation originating from the SP domain of the human PPI. The implication of the IRE1α/miR-17 pathway in regulating the resident ER protein trimming underscores the key role played by ER stress in the development of autoimmunity. We propose that inflammatory cytokines released by infiltrating autoreactive immune cells during insulitis increase β-cell visibility to the immune system by increasing peptide processing and presentation (increase HLA). The increased density of the PPI_15–24 peptide-HLA complex at the β-cell surface would be the consequence of a combined direct transcriptional activation mediated by binding of IFN regulatory factors to the ERAP1 promoter and posttranscriptional control of the ERAP1 mRNA mediated by the IRE1a/miR-17 axis (Fig. 4G). Supported by recent work performed in NOD mice highlighting the potential of IRE1a kinase inhibitors in maintaining β-cell integrity and preserving β-cell function (50), our study designates IRE1α as a relevant therapeutic target to reduce β-cell visibility to the immune system and to prevent T cell–mediated destruction in human diabetes pathogenesis.

Acknowledgments. The authors thank Steve Cramer and Martijn Rabelink, Cell and Chemical Biology, Leiden University Medical Center, for technical help.

Funding. This work is supported by JDRF, Stichting Diabetes Onderzoek Nederland (DON), and the Diabetes Fonds (Dutch Diabetes Research Foundation) and by the Innovative Medicines Initiative 2 Joint Undertaking (IMI2-JU) under grant agreement No. 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. B.O.R. is supported by the Wanek Family Project for Type 1 Diabetes.

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

Author Contributions. S.T. and M.J.L.K. performed the experiments and wrote the manuscript. A.v.d.S. and S.L. performed the experiments. E.J.d.K and F.C. provided the human islets and wrote manuscript. R.C.H. wrote the manuscript. B.O.R. and A.Z. supervised the project, designed the experiments, and wrote the manuscript. A.Z. 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.