Discovery of a Selective Islet Peptidome Presented by the Highest-Risk HLA-DQ8trans Molecule Free

Type 1 diabetes (T1D) is an autoimmune disease characterized by autoreactive T-cell–mediated destruction of the insulin-producing pancreatic β-cells (1–4). The search for naturally processed and presented epitopes (NPPE) for high-risk HLA class II as a target for autoreactive CD4 T cells in T1D has been focus of attention over the years. Most attention was given to HLA-DR–binding epitopes (5,6), whereas HLA-DQ–binding epitopes deserve investigation, too. Indeed, subjects heterozygous for HLA-DQ2 and -DQ8 are endorsed with by far the highest risk for development of T1D, but what functional consequences explain this synergistically increased risk compared with a double dose of HLA-DQ2 or -DQ8 remain unclear. We previously revealed the unique peptide binding properties of HLA-DQ molecules composed of the products of DQA1*0201 (coding for the α-chain of DQ2) and DQB1*0302 (coding for the β-chain of DQ8), the so-called HLA-DQ8trans molecule (7–9). The islet epitopes presented by HLA-DQ8trans are largely unknown. Importantly, HLA-DQ8cis/trans–restricted CD4 T-cell clones have been isolated from human insulitis lesions, and T-cell autoreactivity was confirmed for several proinsulin peptides, underscoring the potential relevance of preproinsulin (PPI) peptides presented by HLA-DQ in diabetogenesis (10–13). We contend that knowledge of the HLA-DQ8trans islet peptidome provides insight in the mechanism by which HLA-DQ2/8 heterozygosity imposes excessive risk for T1D. In turn, HLA-DQ–restricted islet epitopes could yield key reagents for biomarker assays (enzyme-linked immunospot [EPISPOT], DQ tetramers) and feed the pipeline of islet tissue–specific reagents being developed for peptide immunotherapy (8,14–16).

Islet epitopes for T1D predisposing HLA class II molecules have been identified by in silico prediction of T-cell epitopes (17–19), using overlapping islet peptides (14,15,20–22) or by pulsing B cells with islet antigens (23,24). Although these approaches can identify CD4 T-cell epitopes, epitopes derived from islet autoantigens are generally presented by HLA class II molecules after a sequence of events termed naturally processing and presentation by professional antigen-presenting cells (APC) (25). Dendritic cells (DC) are the master regulators of the immune orchestra and initiate and shape the innate and adaptive immune responses. Here, we investigated presentation of islet epitopes by DC expressing T1D highest-risk HLA-DQ. For this purpose, DC expressing HLA-DQ2 and/or -DQ8 were pulsed with three islet-autoantigens, namely PPI, islet tyrosine phosphatase IA-2 (insulinoma-associated antigen-2), and GAD (65 kDa isoform; GAD65). After HLA-DQ isolation and peptide elution, using an HLA-elution method that we optimized for low cell numbers, eluted NPPE were analyzed using high-resolution tandem mass spectrometry (MS/MS). To verify the relevance of these peptides derived from the islet-autoantigens in the context of T1D immunopathology, we analyzed peripheral blood of HLA- and age-matched patients with T1D and patients without diabetes for the presence of autoreactive T cells. We report the first islet-epitopes that are uniquely processed and presented by the T1D highest-risk HLA-DQ2/8 molecules expressed on DC. These islet epitopes are preferentially presented by HLA-DQ8trans. Such novel HLA-DQ restricted epitopes can be used as biomarkers for disease progression and/or contribute to the development of novel immunotherapeutic strategies (e.g., tolerogenic DC therapy).

After informed consent, peripheral blood was drawn from 30 patients diagnosed with T1D (age: 17 ± 6; disease duration: 0–43 years) and from 36 healthy control subjects matched for age and HLA-DQ (age: 27 ± 6). Peripheral blood mononuclear cells (PBMC) were freshly isolated by Ficoll centrifugation and resuspended in culture medium (Iscove's Modified Dulbecco's Medium; Gibco BRL, Paisely, U.K.) containing 10% pooled human heat-inactivated serum. PBMC were subsequently tested for the presence of autoreactive CD4 T cells using ELISPOT.

Recombinant proteins were produced as previously described (26). Briefly, PPI, IA-2, and GAD65 were amplified by PCR from human islet cDNA. PCR products were cloned by Gateway Technology (Invitrogen, Carlsbad, CA) in a bacterial expression vector containing a histidine tag at the N-terminus. Proteins were overexpressed in Escherichia coli BL21(DE3) and affinity purified using anti-His antibody (Invitrogen). Size and purity of recombinant proteins were analyzed by gel electrophoresis and Western blotting using anti-His antibody. Endotoxin contents were below 50 IU/mg recombinant protein, which is below the detection threshold, as tested using a Limulus Amebocyte Lysate Assay (Cambrex, East Rutherford, NJ). All proteins were tested in lymphocyte stimulation assays to exclude antigen-nonspecific T-cell stimulation. Peptides were synthesized according to standard fluorenylmethoxycarbonyl chemistry using a Syro II peptide synthesizer (MultiSynTech, Witten, Germany). The integrity of the peptides was checked using ultra-performance liquid chromatography-MS and matrix-assisted laser desorption/ionization time-of-flight MS. The following biotinylated indicator peptides were used in the cell-free HLA-DQ peptide binding studies: CLIP: KMRMATPLLMQAL (DQ2cis), AAEAALEAEEWAA (DQ2trans), and AAPHTTQPAVEAA (DQ8trans) and HSV-2: EEVDMTPADALDDFD (DQ8cis).

Isolation and generation of DC from homozygous HLA-DQ2, homozygous HLA-DQ8, or heterozygous HLA-DQ2/8 healthy blood donors was performed as described previously (27). PBMC isolated from each buffy coat were separately cultured and pulsed with islet antigens. For each elution, 40 × 10⁶ pulsed mature DC (mDC) were obtained from three donors per HLA-DQ genotype that were pooled after pulsing with islet autoantigen before the HLA-peptide elutions. PBMCs were isolated by Ficoll gradient from three HLA-typed buffy coats per HLA-DQ typing, and CD14⁺ monocytes were isolated and cultivated with granulocyte macrophage colony-stimulating factor (800 units/mL) and interleukin (IL)-4 (500 units/mL) (Invitrogen, Breda, the Netherlands) for 6 days to obtain immature DC (iDC). The iDC were pulsed with PPI, IA-2, and GAD65 for 6 h, after which iDC were matured by incubating 0.5 × 10⁶ DC/well in a 24-well plate with lipopolysaccharide (100 ng/mL) for 24 h in the continuous presence of the three islet autoantigens. After 30 h, pulsed mDC were harvested, washed three times with PBS to remove excess of islet autoantigens, and lysed in 1 mL lysis buffer (50 mmol/L Tris, 150 mmol/L NaCl, 5 mmol/L EDTA, 0.5% zwitterion, 10 mmol/L iodoacetamide, and protease inhibitors [complete inhibitor mix; Roche]); they were centrifuged at high speed for 60 min at 10,000g to remove nuclei and insoluble material.

Islet autoantigens pulsed mDC were lysed, and proteins were digested using the filter-aided sample preparation method (28). Briefly, 100 μg of protein was loaded on a 30-kDa filter. SDS was removed in three washes by 8 mol/L urea. The proteins were alkylated using iodoacetamide, and the excess reagent was washed through the filters by three additional washes with 8 mol/L urea. Proteins were digested overnight using endoproteinase LysC, followed by a 4-h digestion using trypsin at room temperature. Tryptic peptides were desalted on C18 SepPak. Peptides were subsequently fractionated by strong cation exchange on an Agilent 1100 gradient high-performance liquid chromotography (HPLC) system (Agilent, Waldbronn, Germany) equipped with an in-house packed strong cation exchange-column (320 µm ID, 15 cm, polysulfoethyl A 3 μm; Poly LC), run at 4 μL/min. The gradient started with 10 min at 100% solvent A 70/30/0.1 (water/acetonitrile/formic acid), after which a linear gradient was started to reach 100% solvent B (250 mmol/L KCl, 35% acetonitrile, and 0.1% formic acid) in 15 min, followed by 100% solvent C (500 mmol/L KCl, 35% acetonitrile, and 0.1% formic acid) in the following 15 min. The eluent was held at 100% solvent C for 5 min to clean the column, then switched back to 100% solvent A. Fifteen fractions were collected in 1-min intervals, lyophilized, and reconstituted in 30 μL 95/3/0.1 (water/acetonitrile/formic acid). Dissolved fractions were analyzed by online nano–HPLC MS with a system consisting of an Agilent 1100 gradient HPLC system and a LTQ FT Ultra MS (Thermo Fisher Scientific, Bremen, Germany). Fractions (5 μL) were injected onto a homemade precolumn (100 μm × 15 mm; Reprosil-Pur C18-AQ 3µm, Dr. Maisch, Ammerbuch, Germany) and eluted via a homemade analytical nano-HPLC column (15 cm × 50 µm; Reprosil-Pur C18-AQ 3 μm). The gradient was run from 0 to 30% solvent B (10/90/0.1 water/acetonitrile/formic acid) in 10–155 min. A tip of ∼5 μm was drawn from the tip of the nano-HPLC column to act as electrospray needle. Full-scan MS spectra were acquired in the Fourier transform-ion cyclotron resonance-MS with a resolution of 25,000 at a target value of 5 × 10⁶. The five most intense ions were selected and fragmented in the linear ion trap using collision-induced dissociation at a target value of 10,000. For MS/MS spectral matching, Mascot 2.2.04 (Matrix Science) was used, with 2 ppm precursor and 0.5-Da fragment accuracy. Variable modifications included N-terminal protein acetylation and methionine oxidation. Carbamidomethylation of cysteine was selected as a fixed modification. The false discovery rate was set to 1%.

Affinity purification of HLA-DQ molecules from mDC and subsequent peptide elutions were performed as follows. We optimized our existing peptide-elution protocol (8) for low DC numbers. All HLA-DQ isolation and washing steps were performed using a 100-μL pipet tip. Lysate was precleared by running it through a 100-μL pipet tip containing a small filter and packed with 100 μL Sepharose beads. The precleared lysate was collected, and HLA-DQ molecules were subsequently isolated using a pan-DQ (SPV-L3) antibody coupled to Sepharose beads and packed in a 100-μL pipet tip; the lysate was passed through the SPV-L3 microcolumn to isolate HLA-DQ molecules by gravity force. Columns were washed with four bed volumes of lysis buffer, followed by four bed volumes of low-salt buffer (120 mmol/L NaCl and 20 mmol/L Tris-HCl, pH 8.0), high-salt buffer (1 mol/L NaCl and 20 mmol/LTris-HCl, pH 8.0), no-salt buffer (20 mmol/L Tris-HCl, pH 8.0), and low-Tris buffer (10 mmol/L Tris-HCl, pH 8.0). The HLA-peptide complexes were eluted with two bed volumes of 10% acetic acid. HLA-DQ eluates (containing peptides and HLA) were fractionated with an HPLC system. The material was eluted using a gradient of 0–50% acetonitrile supplemented with 0.1% trifluoroacetic acid.

MS analysis of HLA-eluted peptides was performed as described previously (8), with some modifications. After immune precipitation, proteins and HLA-peptides in the unfiltered eluate were separated by selective elution from a small C18 column in two fractions with 20% and 30% acetonitrile, respectively (29). The HLA-peptides were analyzed via online C18-nano-HPLC-MS with a system consisting of an Easy-nLC 1000 gradient HPLC system (Thermo Fisher Scientific) and a Q Exactive MS (Thermo Fisher Scientific). Fractions were injected onto a homemade precolumn (100 μm × 15 mm; Reprosil-Pur C18-AQ 3 μm, Dr. Maisch, Ammerbuch, Germany) and eluted via a homemade analytical nano-HPLC column (15 cm × 50 μm; Reprosil-Pur C18-AQ 3μm). The gradient was run from 0 to 30% solvent B (10/90/0.1 water/acetonitrile/formic acid) in 120 min. The nano-HPLC column was drawn to a tip of ∼5 μm and acted as the electrospray needle of the MS source. The Q Exactive MS was operated in top 10 mode. Parameters were resolution 70,000 at an automatic gain control target value of 3 million maximum fill time of 100 ms (full scan) and resolution 35,000 at an automatic gain control target value of 1 million/maximum fill time of 128 ms for MS/MS at an intensity threshold of 78,500. Apex trigger was set to 1 to 5 s, and allowed charges were 1–3. In a postanalysis process, raw data were converted to peak lists using Proteome Discoverer 1.4. For peptide identification, MS/MS spectra were submitted to the human IPI 3.87 database using Mascot 2.2.04 software with the following settings: 10 ppm and 20 millimass units deviation for precursor and fragment masses, respectively; no enzyme was specified. All reported hits were assessed manually.

Binding of identified islet peptides to all four HLA-DQ molecules was studied in cell-free HLA-DQ peptide-binding assays. As the source of HLA-DQ, Epstein-Barr virus–transformed B lymphocyte cells lentiviral transduced to express a single HLA-DQ molecule were used (8). These assays are based upon competition between a fixed concentration biotinylated reporter peptide (0.6 µmol/L) and an unbiotinylated islet peptide (0–300 µmol/L); once the islet peptide competes with the indicator peptide a drop in signal (counts/min) is observed representing binding of the islet-peptide. Affinities are subsequently calculated using GraphPad 5 software; the concentration of islet peptide required for half-maximal inhibition of binding of the reporter peptide indicate the half-maximal effective concentration (EC₅₀) value. Although binding of the islet peptides to different HLA-DQ molecules cannot be compared accurately due to the amino acid sequence of indicator peptides differing between the HLA-DQ assays (different HLA-DQ molecules with different properties), a difference in EC₅₀ value ≥10 times was considered substantial.

Detection of IFN-γ and IL-10 production by CD4 T cells in response to the identified NPPE (10 µg/mL) was performed using ELISPOT, as described previously (30). Data are expressed as the simulation index (SI) = total number of spots per triplicate/total number of spots in triplicate in the presence of diluent alone. An SI ≥3 is considered as positive.

Because our islet antigens were not labeled with a fluorescent dye and antigen uptake and processing cannot be measured by flow cytometry or fluorescent microscopy, we performed proteome analysis of DC pulsed with PPI, IA-2, and GAD65 to confer intracellular uptake of the islet antigens. Lysates of islet antigen pulsed mDC (∼40 × 10⁶) were digested with trypsin, and tryptic peptides were analyzed by MS. Identified peptides were screened against a human protein database containing the amino acid sequences of PPI, IA-2, and GAD65. We independently analyzed the proteome of two lysates pulsed with PPI/IA-2/GAD65. Of all three islet antigens, we retrieved peptides derived from the antigens in the two lysates. Peptides of both the N- and C-terminus were identified in two independent proteome analysis covering 32.0 ± 0% of PPI, 23.0 ± 1.0% of GAD65, and 26.5 ± 0.5% of IA-2 (Supplementary Fig. 1). In addition, nested sets of peptides were retrieved, indicating accurate MS analysis. These data show that islet antigen pulsed mDC efficiently take up whole antigens.

To decipher the HLA-DQ islet peptidome, iDC homozygous for HLA-DQ2 or HLA-DQ8 or heterozygous for HLA-DQ2/8 were pulsed with PPI, IA-2, and GAD65 and subsequently matured with lipopolysaccharide and IFN-γ. Complete maturation of DC was confirmed by phenotype (data not shown). A total of 353 unique peptides (derived from 56 proteins) were eluted from HLA-DQ2/8. More islet peptides were retrieved from heterozygous HLA-DQ2/8 (19 of 353 [5.4%]) than from homozygous HLA-DQ8 DC (6 of 459 [1.3%]). No islet peptides were eluted from HLA-DQ2. PPI and IA-2 peptides were retrieved from HLA-DQ2/8 and HLA-DQ8. GAD65 islet peptides were not identified. From PPI, peptides with a mean length of 13 amino acids (range 8–16) were retrieved solely from DC expressing the highest-risk HLA-DQ2/8 molecules (Fig. 1 and Table 1). Three PPI peptides were eluted from HLA-DQ2/8 encompassing two distinct core regions: PPI_17–24 (part of the PPI signal-sequence) and PPI_54–69. Intriguingly, PPI_17–24 has been identified as a CD8 T-cell epitope that is naturally processed and presented by HLA-A2 (31). Now, for the first time, PPI_17–24 is identified as a naturally processed and HLA-DQ–presented peptide. We recently identified PPI_54–69 (TRREAEDLQVGQVELG) as a CD4 T-cell epitope that preferentially is presented by HLA-DQ8trans and becoming highly immunogenic after posttranslational modification (30). Here, this same PPI_54–69 epitope is naturally processed and presented by heterozygous HLA-DQ2/8–expressing DC. Peptides from the C-terminus of PPI were not retrieved.

From HLA-DQ2/8, a total of 16 IA-2 peptides (Fig. 1 and Table 1) were eluted with a mean peptide length of 14 amino acids (range 11–19). All peptides are members of nested peptide sets covering three distinct core regions of the N-terminus of IA-2. Similarly nested peptide sets were retrieved from HLA-DQ8. Peptides from the C-terminus of IA-2 were not retrieved. To validate MS analysis of the identified peptides, we synthesized PPI- and IA-2–eluted peptides (PPI_17–24, IA-2_296–311, and IA-2_319–333); MS spectra of the eluted peptides and the synthesized peptides fully overlapped (Supplementary Fig. 2).

HLA-DQ2/8 heterozygous cells can express four types of HLA-DQ molecules: HLA-DQ2cis, HLA-DQ2trans, HLA-DQ8trans, and HLA-DQ8cis (11). Therefore, HLA-DQ binding of the PPI- and IA-2–eluted peptides was validated in competitive HLA-DQ/peptide–binding assays. Binding of PPI_54–69 to HLA-DQ was already validated in our previous study (30). PPI_17–24 only bound to HLA-DQ8trans (Fig. 2 and Table 1). IA-2 peptides, encompassing the same core regions, were retrieved from HLA-DQ2/8 and HLA-DQ8. Binding of IA-2_142–159, IA-2_293–311, and IA-2_318–333 was observed for HLA-DQ8trans and -DQ8cis (Fig. 3 and Table 1). Strong binding of these IA-2 peptides was observed for HLA-DQ8trans and weak binding for HLA-DQ8cis. Binding of the IA-2–eluted peptides to HLA-DQ2cis/trans was not observed.

The minimal binder registers (MBRs) in the eluted peptides responsible for HLA-DQ binding were in silico predicted by using the predictions algorithm software MOTIFS and the peptide-binding motifs of all four HLA-DQ molecules (8). Predicted MBRs in the IA-2 peptides for HLA-DQ8cis/trans (Table 1) were consistent with the observed binding of the eluted IA-2 peptides to HLA-DQ8cis/trans. MBRs in the eluted IA-2 peptides for HLA-DQ2cis/trans were not predicted. Most of the eluted IA-2 peptides contained three to four anchor residues (crucial for HLA-peptide binding) in the predicted MBRs for HLA-DQ8trans and two anchor residues for HLA-DQ8cis. Indeed, stronger binding of the IA-2 peptides to HLA-DQ8trans was observed compared with HLA-DQ8cis. To validate HLA-DQ binding of the putative registers in the IA-2 peptides, synthetic peptides spanning the MBRs (Table 1) were synthesized with two extra alanines both N- and C-terminally necessary for proper HLA-DQ binding. Strong binding to DQ8trans was observed for five predicted DQ8trans MBRs, and two predicted DQ8cis MBRs bound with low affinity to DQ8cis (data not shown). These in silico predictions support the results from the DQ elution studies; the number of eluted IA-2 peptides DQ8trans > DQ8cis.

We examined proliferative responses in fresh peripheral blood of patients with T1D and HLA-DQ– and age-matched control subjects without T1D against the DQ NPPEs derived from PPI and IA-2 using ELISPOT. ELISPOT studies with patients with T1D and the identified PPI_54–69 natural epitope were performed previously (30). Collectively, T-cell responses in 30 patients with T1D could be observed against three DQ naturally processed epitopes ex vivo in 12 (40%) for PPI_17–24, 16 (53%) for IA-2_293–311, and 16 (53%) for IA-2_318–333 (Fig. 3 and Supplementary Tables 1 and 2). Only a few patients (2 of 21 [9.5%]) responded to IA-2_142–159 by producing IL-10 (data not shown). Healthy individuals showed less frequent responses to the tested DQ natural epitopes (15–28%), with most of the healthy individuals responding with IFN-γ to PPI_17–24 and with IL-10 to the IA-2 epitopes. Patients responding to PPI_17–24 produced IFN-γ (31%), IFN-γ+IL-10 (8%), or IL-10 (61%). Because we recently reported that the PPI_17–24 epitope is also presented by HLA-A2 to autoreactive CD8 T cells (31), we checked the HLA class I typing when available but did not see a trend in that direction; all five cases in which PBMCs from patients responded to PPI_17–24 by production of IFN-γ were HLA-A2 negative. This excludes the possibility that this responsiveness can be attributed to HLA-A2–restricted CD8 T cells.

Patients showed predominantly IFN-γ to the two IA-2 epitopes (75% and 88%, respectively); patients responding to IA-2_293–311 produced IFN-γ (62.5%) or IFN-γ+IL-10 (12.5%), and 25% of patients responded solely with IL-10. Patient responses against IA-2_318–333 showed even a more proinflammatory phenotype, because 75% produced IFN-γ and 12.5% produced IFN-γ+IL-10. A decrease in patients responding to IA-2_318–333 exclusively with IL-10 (12.5%) was observed compared with IA-2_293–311. Our data demonstrate that autoreactive CD4 T-cell responses can be detected against these novel PPI and IA-2 naturally processed epitopes that preferentially bind the T1D highest-risk HLA-DQ8trans molecule. Furthermore, the quality and prevalence of these T-cell responses both differ between patients with T1D and case control subjects without T1D. Although our cohort remains insufficiently sized to draw firm conclusions regarding the influence of HLA status on the presence of CD4 T-cell responses, these data demonstrate the presence of autoreactive T-cell responses against the identified novel DQ epitopes.

We provide evidence on the functional consequences of presentation of islet proteins by the highest-risk HLA-DQ8trans that may contribute to the association of this molecule with the highest genetic predisposition to T1D. Qualitative and quantitative differences are both observed between the islet peptidomes of HLA-DQ2/8 heterozygosity versus homozygosity. First, the total number of peptides derived from PPI and IA-2 presented by highest-risk HLA-DQ2/8 is greater compared with homozygous HLA-DQ8, whereas no islet peptides were retrieved from homozygous HLA-DQ2. Second, HLA-DQ2/8 DC generate an exclusive islet peptidome uniquely presented by the highest-risk HLA-DQ8trans molecule. Finally, we provide evidence that DC selectively present islet autoantigens PPI and IA-2 epitopes but not GAD65 epitopes by high-risk HLA-DQ.

The identified PPI naturally processed peptides that were retrieved solely from the highest-risk HLA-DQ2/8 were confirmed to bind HLA-DQ8cis/trans selectively. The unusually short DQ-binding PPI_17–24 peptide is located in the signal-sequence of PPI and has been described as a naturally processed and HLA-A2–restricted epitope of diabetogenic CD8 T cells (31). We now report islet autoreactive CD4 T-cell responses against the same PPI epitope in patients with T1D but not in healthy age- and HLA-matched subjects. The PPI_17–24 epitope as an 8-mer does not fulfill the high-risk HLA-DQ peptide-binding motif (8). Yet, binding of PPI_17–24 was confirmed for HLA-DQ8trans. Unconventional length and binding of peptides to HLA class II molecules has been reported (32,33). Our finding that PPI_17–24 as an 8-mer is naturally presented by T1D highest-risk HLA-DQ8trans extends the notion that CD4 T cells can respond to unusually presented self-peptides. This signal peptide was generated by DC pulsed with whole PPI. The question that emerged is how peptides from the signal-sequence of PPI can be presented by HLA-DQ on DC. In the NOD mouse model of autoimmune diabetes, insulin-secretory granules from β-cells and even whole β-cells are ingested by APC and transported to the pancreatic lymph node (34,35), and T cells from patients with T1D respond to insulin secretory granules derived from β-cells (36). Subsequently, intact or partially synthesized PPI that is embedded in the endoplasmic reticulum can become a natural source of signal peptides presented by HLA-DQ8trans on the surface of DC. Pancreatic β-cells do not secrete whole PPI as an extracellular source for uptake by DC. However, a small proportion of whole PPI (including the signal sequence) is present in the cytosol of human islets that may become accessible to DC under pathological conditions in T1D, such as β-cell stress, thus ending up in HLA-DQ for presentation to the immune system (37,38). We previously reported PPI_54–69 as preferentially presented by the highest-risk HLA-DQ8trans molecule (30). The proportion of patients responding against this epitope doubled after posttranslational modification, with most patients showing a proinflammatory immune response. This PPI_54–69 epitope is now confirmed as a processing product presented solely by HLA-DQ2/8 heterozygous DC.

IA-2 peptides were presented both by HLA-DQ2/8 and to a lesser extent HLA-DQ8–expressing DC and encompassed identical core regions that were derived solely from the IA-2 extracellular domain. Binding of the IA-2 peptides was validated for HLA-DQ8cis/trans. Because humoral responses have been detected against the IA-2 intracellular domain, several immune studies hitherto focused on CD4 T-cell epitopes derived from the IA-2 intracellular domain. Uptake of IA-2 by pulsed DC was confirmed, but peptides from the IA-2 intracellular domain were not retrieved, which may imply a limited relevance of these peptides as a target for CD4 T cells in the context of high-risk HLA-DQ. Preliminary data show that peptides retrieved from the IA-2 extracellular domain were derived solely from DQ2/8 heterozygous DC that were deamidated at one Q residue, suggesting posttranslational modification of islet-peptides presented by the highest-risk DQ2/8 expressed on DC.

The naturally processed PPI and IA-2 epitopes identified proved to be targets of CD4 T cells in approximately half of the patients with T1D. Patient T cells responding to the IA-2 epitopes showed a proinflammatory phenotype (IFN-γ), whereas the IA-2–specific T cells from a few responding donors without diabetes showed an anti-inflammatory response (IL-10). Intriguingly, immune responses to PPI_17–24 showed an inverse pattern of cytokine production that we speculate reflects differential regulatory mechanisms against different islet epitopes in T1D patients versus healthy donors. T-cell responses to peptides eluted from HLA-DQ2/8 DC were not exclusively recognized by HLA-DQ2/8 heterozygous patients with T1D. Although it is certainly conceivable that other HLA molecules than those tested here may act as an alternative restriction element in such cases (notably HLA-A2) (31), this discrepancy between the HLA-DQ ligandome generated by processing of the whole PPI protein versus exogenous pulsing with an excess of synthetic PPI peptide only in the T-cell assay may also relate to the notion that the amount of PPI peptides in the latter case may overcome the threshold required for T-cell activation in donors with T1D expressing HLA-DQ2 or -DQ8. This peptide was repeatedly undetectable in elution studies from HLA-DQ2 or -DQ8 homozygous donors, but exclusively retrieved from HLA-DQ2/8 heterozygous DC upon processing of PPI protein. Yet, our HLA-DQ–binding study demonstrates that once a peptide is generated, it may also bind to other HLA-DQ molecules, be it to a lesser extent.

GAD65 peptides were not retrieved from HLA-DQ, although uptake of the protein was confirmed by proteome analysis of pulsed DC. Peptide loading pathways in APC are diverse; peptide loading into HLA class II molecules in the MHC class II compartment requires CLIP derived from the invariant chain and HLA-DM for presentation in DC, but in B cells, a modifier of HLA-DM is expressed (HLA-DO) that associates with HLA-DM and restricts HLA-DM activity to more acidic compartments in B cells. Although the exact role of HLA-DO in different APCs remains largely unknown, DO can function as a competitive and irreversible inhibitor of HLA-DM in subsets of APCs (39,40); when present, DO subtly alters the repertoire of HLA class II–bound peptides displayed at the surface of APCs. DO is mainly expressed in B cells, and its effect on HLA class II presentation would therefore be mostly observed in B cells. Epitopes recognized in the context of DQ display a DM-sensitive phenotype, whereas for DR molecules, a tendency toward DM-resistant epitopes is observed (41); presentation of DM-sensitive (DQ) antigens benefited more from maturation of DC compared with DM-resistant (DR) antigens. Thus, peptide binding to HLA class II molecules is modulated differently in B cells compared with DC, providing a plausible explanation why DC do not present GAD65 peptides in HLA-DQ. Because GAD65 epitopes have been identified to be naturally processed and presented by B cells as target of CD4 T cells, this implies that DC, uniquely involved in priming of naïve CD4 T cells, present a relatively small set of islet epitopes in HLA-DQ to autoreactive T cells derived from a limited number of islet autoantigens.

The recombinant islet proteins used in this study are not necessarily in their native conformation. For DC, the three-dimensional structure is not crucial for protein uptake. This may differ for B cells, for which the three-dimensional structure of an antigen is important for uptake via the B-cell receptor and subsequent processing. For identification of B-cell epitopes, proper folding of the recombinant proteins will therefore be more important. Nonetheless, the actual processing occurs in lysozymes, in which the acidic milieu will affect the natural confirmation of proteins. Considering the time of antigen processing that we chose, the possibility remains that additional peptides may generated and presented by DC. Studies in mice show that DC are able to retain and present antigens long after uptake (42) and that the same peptides are presented by DC at early and later time points. Rapidly generated peptides from proximal parts of antigens may be overrepresented in our approach. Additional peptides presented by (subsets of) DC at later time points after uptake in vivo may conceivably arise.

It is conceivable that our identified PPI and IA-2 DQ-NPPEs represent the tip of the iceberg of the total islet peptide ligandome. The identified peptides in this study might impose a bias regarding the most favorable binding characteristics, and peptides with unfavorable properties are more challenging to identify due to technical limitations. Nevertheless, our observations that the particular islets eluted from HLA-DQ proved immunogenic in patients with T1D, in particular, compared with subjects without diabetes, even if their binding affinity to HLA was relatively weak, implies that our strategy of epitope discovery with potential relevance to the disease is suitable and rewarding.

Our identification of naturally processed and HLA-DQ–presented epitopes by DC may bear relevance to the selection of islet peptides for prevention of T1D to avoid loss of tolerance to the triggering epitopes, affecting the choice of potential therapeutic agents for the prevention of T1D as well as immune monitoring of islet autoreactive CD4 T-cell responses in prediction, disease progression, and, possibly, as surrogate end points in immunotherapeutic trials and as part of the developing area of HLA-DQ tetramer technology in T1D.

Acknowledgments. The authors thank Kees Franken, from the Department of Immunohematology and Blood Transfusion of Leiden University Medical Center, for excellent technical advice during the production of recombinant proteins.

Funding. B.O.R. is a member of the Danish Diabetes Academy. This work was supported by JDRF grant 17-2012-547, the European Union’s 7th Framework Programme (FP7/2007-2013) under grant agreement No. 241447 (NAIMIT), a VICI grant of the Netherlands Organisation for Scientific Research grant 918.86.611, and an Expert Center Grant from the Dutch Diabetes Research Foundation.

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

Author Contributions. M.v.L. conceived experiments, researched data, and wrote and reviewed the manuscript. A.H.d.R., J.P., S.L., A.J., and I.G.-T. researched data. P.A.v.V., T.N., J.W.D., S.A., H.J.A., and M.P. contributed to discussion and reviewed and edited the manuscript. B.O.R. supervised this study and contributed to discussion and writing the manuscript. B.O.R. 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.

Prior Presentation. Parts of this study were presented in abstract form at the Immunology of Diabetes Society 14th International Congress, Munich, Germany, 12–16 April 2015.