Epitope Selection for HLA-DQ2 Presentation: Implications for Celiac Disease and Viral Defense

This article is featured in In This Issue, p.2515

Presentation of peptides to CD4⁺ T cells by MHC class II (MHCII) proteins (HLA-DR [DR], HLA-DQ [DQ], and HLA-DP [DP] in humans) is a key initiating step in immune responses. Allelic variations in MHCII proteins influence Ag presentation of both self- and foreign proteins, with consequences for autoimmunity/tolerance and host defense (1). HLA-DQ2 (DQA1*05:01/DQB1*02:01) (DQ2; also called DQ2.5) is an allele of particular interest in this regard. This allele confers genetic risk to several autoimmune diseases, including type 1 diabetes and celiac disease (CD) (2, 3). CD is a disorder caused by immune intolerance to ingested cereal gluten proteins of wheat (consisting of gliadin and glutenin subcomponents), barley, and rye. In susceptible individuals, DQ2 proteins preferentially bind and present to CD4⁺ T cells (4, 5) certain proline-rich gluten peptides that have been posttranslationally modified (i.e., deamidation of glutamine) by the enzyme transglutaminase 2 (TG2) (6–9). Although DQ2 is associated with vigorous antigluten T cell responses, DQ2 also is associated with poor responses to several vaccines (10–12) and failure to control hepatitis virus C (13) and hepatitis virus B (14).

Peptide presentation by MHCII, including DQ2, is influenced by interaction with Ag presentation cofactors, invariant chain (Ii), and HLA-DM (DM). Ii, a class II chaperone, directs nascent MHCII/Ii oligomers from the endoplasmic reticulum to MHCII-containing compartments (MIIC) where it is processed to a nested set of class II–associated Ii peptides (CLIP) that bind the groove (15). For most class II allelic proteins, the displacement of CLIP in MIIC is catalyzed by DM that transiently associates with MHCII, protects peptide-receptive MHCII from degradation by peptide exchange when peptides are available, and edits bound peptides in favor of high-affinity ones for cell surface presentation (16). Work from our laboratories (17, 18), confirmed by others (19), showed that DQ2 has reduced interaction with DM, compared with most other alleles.

DM can enhance or suppress the presentation of specific MHCII–peptide complexes. In general, MHCII–peptide complexes with lower intrinsic stability are DM susceptible (20–22), but not all high-stability complexes are DM resistant (23). Immunodominant epitopes preferentially survive DM editing within endosomal compartments and trigger CD4⁺ T cell activation (16). In contrast, other potential epitopes are removed from MHCII through DM catalysis and fail to elicit T cell responses (24), except when they get presented through pathways that avoid interaction with DM (25). We hypothesized that the DM-resistant feature of DQ2 likely contributes to the escape of gliadin peptides from extensive DM editing. In addition, DQ2 has the special ability to stably bind proline-rich gliadin peptides (5) that use TG2-deamidated residues as DQ2-binding anchors (26); together, these unique features of DQ2 may allow gliadin presentation to disease-driving CD4⁺ T cells. In vitro, DQ2 resistance to DM can be overcome by increased DM concentrations or prolonged exposure to DM (17). We produced transfectants expressing high levels of DM and showed that intracellular DM/DQ2 interaction occurs in these cells (18). In addition, DQ2 has a natural deletion in the region involved in interaction with DM (5); we found that insertion of arginine or glycine at α53 (DQ2+53R, DQ2+53G) confers DM sensitivity to the mutant DQ2 proteins, providing another tool to test the effects of increased DM/DQ2 interaction on Ag presentation (18).

In this study, we analyze DM effects on DQ2 presentation of four gliadin epitopes to T cell clones isolated from CD patients. We also test the response of DQ2-restricted CD4⁺ T cells specific for viral peptides from three different viruses. Strikingly, we found different consequences of DM activity for these two groups of T cell clones, with suppression of gliadin presentation and enhancement of viral peptide presentation. These results imply key differences in DQ2 Ag presentation pathways operating in CD compared with host defense against viral infection.

The TxB hybrid APC cell lines T2 (MHCII⁻/DM⁻) and T2DM (MHCII⁻/DM⁺) stably expressing wild-type (WT) DQ2 (T2.DQ2 and T2.DQ2.DM) or mutant DQ2 (T2.DQ2α+53G.DM) have been established previously (18). They were cultured in complete IMDM (Thermo Fisher Scientific, Waltham, MA) supplemented with 10% (v/v) FBS and 2 mM l-glutamine and maintained at 37°C in a humidified atmosphere of 5% CO₂. DQ2-restricted T cell clones specific for different gliadin epitopes were isolated, as described (27). These include TCC819.392 (specific for α1a, see below for a description of different gliadin epitopes), TCC820.250 (specific for α2), TCC820.270 (specific for γ1), and TCC820.59 (specific for γ4d); the various epitopes are described in the 4Ags section below. 1A.B.25, a DQ2-restricted T cell clone recognizing aa 431–440 (EVDMTPADAL) of VP16 (gene UL48) of HSV type 2 (HSV-2), isolated for a previous study, was used (28). DQ2-restricted influenza-specific T cell clones (clone 5) that recognize aa 97–113 of hemagglutinin (HA) protein from A/New Caledonia/20/99 (H1N1) were isolated as described (29). DQ2-restricted human papillomavirus type 16 (HPV16)–specific T cell clone (clone 60) recognizing E7_35–50 was from van der Burg (30).

A high molecular weight fraction of gluten was prepared using a previously described method (18). Briefly, gluten proteins were digested with pepsin and trypsin/chymotrypsin. The pepsin/trypsin-treated gluten (PT-gluten) was further deamidated with TG2 in the presence of CaCl₂. HSV-2 was UV inactivated and prepared as described previously (28). H1 HA protein with C-terminal histidine tag from influenza A/New Caledonia/20/1999 (H1N1) was obtained from BEI Resources (NR-48873). Recombinant HPV16 protein E7 was purchased from CUSABIO (Wuhan, China). An α-gliadin fragment (α1/α2: LQLQPFPQPELPYPQPELPY), containing the overlapping T cell epitopes α1a (underscored) and α2 (bolded) (31, 32), was synthesized and provided by Dr. Xi Jin in the laboratory of C. Khosla, Department of Chemistry, Stanford University, Palo Alto, CA. Synthetic peptides containing the gliadin T cell epitopes (underscored) (27, 33) γ1 (P1213: pyroglutamic acid [pyroE]PEQPQQSFPEQERP) and γ4d (P1936: PFPQPEQPFCEQPQR) were synthesized by GL Biochem (Shanghai, China) and at the University of Oslo, Norway, respectively. Glutamine (Q) at the N terminus of peptide P1213 is unstable and tends to spontaneously convert to pyroE; therefore, P1213 was synthesized with pyroE at the N terminus. The following peptides were synthesized by Genscript (Piscataway, NJ): HA_97–113 (YPGYFADYEELREQLSS), biotinylated MHCIα_49–63 (APWIEQEGPEYWDQE) (18), biotinylated Ii_81–104 (CLIP1: LPKPPKPVSKMRMATPLLMQALPM), biotinylated Ii_92–107 (CLIP2: RMATPLLMQALPMGAL) (17), biotinylated P1269 (QLQPFPQPELPY) containing α1a, biotinylated PS1200 (PQPELPYPQPQS) containing α2 (34), biotinylated P1213 (pyroEPEQPQQSFPEQERP) containing γ1 (31), and biotinylated P1936 (PFPQPEQPFCEQPQR) containing γ4d (35). HPV16 E7_22–56 (LYCYEQLNDSSEEEDEIDGPAGQAEPDRAHYNIVT) was provided by van der Burg (30).

Pulse/chase analysis was performed as previously described (36). Briefly, 6 × 10⁷ T2.DQ2 or T2.DQ2.DM cells were washed with complete Cys/Met-free RPMI medium (containing 10% FBS and 2 mM l-glutamine) and then resuspended in complete Cys/Met-free RPMI medium for starving at 37°C for at least 1 h. After starving, cells were pulsed with 150 μCi/ml Express ³⁵S labeling mix (PerkinElmer, Waltham, MA) for 3 h. One half of the pulsed cells was washed and pelleted without chasing (t = 0 time point), and the other half of pulsed cells was chased overnight in complete RPMI medium and then washed before pelleting (t = 24 h time point). Cell pellets were lysed with lysis buffer (50 mM Tris-HCl, 150 mM NaCl, 1% IGEPAL CA630 from Sigma-Aldrich, St. Louis, MO; PMSF and complete protease inhibitors from Thermo Fisher Scientific). DQ2 proteins in clear cell lysates were then immunoprecipitated using protein G beads (GE Healthcare, Chicago, IL) that were precoupled with anti-DQ mAb SPV-L3. After incubation with cell lysate overnight at 4°C, protein G beads were pelleted, washed five times with cold PBS, and boiled in reducing sample buffer containing 62.5 mM Tris-HCl (pH 6.8), 1% SDS, 3% glycerol, 0.007% bromophenol blue, and 1% 2-ME. DQ2 proteins eluted from the beads were separated by 12% SDS-PAGE gel electrophoresis, and separated bands representing target proteins were visualized by exposing dried gels to radiography film (Kodak, Rochester, NY).

Soluble DM, DQ2-CLIP1, and DQ2α+53G-CLIP1 were generated as previously described (37). For the time course of spontaneous peptide loading, 50 nM thrombin-cleaved sDQ2-CLIP1 or sDQ2α+53G-CLIP1 was incubated with 50 μM biotinylated peptides (CLIP1, CLIP2, MHCIα, P1269, PS1200, P1213, or P1936) in the reaction buffer (100 mM acetate buffer [pH 4.6], 150 mM NaCl, 1% NaCl, 1% BSA, 0.5% Nonidet P-40, 0.1% NaN₃) supplemented with 1× EDTA-free protease inhibitor mixture (Thermo Fisher Scientific). The peptide-loading reaction took place at 37°C for the indicated time. At each time point, two volumes of neutralization buffer (100 mM Tris-Cl [pH 8], 150 mM NaCl, 1% NaCl, 1% BSA, 0.5% Nonidet P-40, 0.1% NaN₃) were added to terminate the reaction. 100 μl of the neutralized mixture was then transferred to a 96-well plate precoated with DQ-specific mAb SPV-L3 and incubated at room temperature for 1 h. After five washes with wash buffer (0.05% Tween 20 in PBS), 100 μl of PBS containing 1% BSA and 1:1000 diluted europium-labeled streptavidin (PerkinElmer) was added to each well and left at room temperature for 1 h. After another five washes, 100 μl of enhancement solution (PerkinElmer) was then applied, and the time-resolved fluorescence signal associated with europium in each well was measured using a plate reader (Tecan, Männedorf, Switzerland). To examine DM effects, 50 nM thrombin-cleaved sDQ2-CLIP1 or sDQ2α+53G-CLIP1 was incubated with 10 μM biotinylated peptides (CLIP2 or MHCIα) in the presence of 50, 200, or 800 nM soluble DM at 37°C for 1 h and then captured by SPV-L3 for quantification of biotinylated peptides, as detailed above.

Frozen T cell clones were thawed, washed, and rested in IMDM with 10% FBS, 2 mM l-glutamine, 2% human serum, and 1% penicillin/streptomycin at 37°C in a humidified atmosphere of 5% CO₂ for at least 1 h. APC including T2.DQ2, T2.DQ2.DM, T2.DQ2α+53G.DM, 9.5.3, and 9.5.3.DM were irradiated (12,000 rad) or fixed with 1% paraformaldehyde for 5 min on ice, followed by two PBS washes. After irradiation/fixation, APC (50,000 cells) were incubated with Ag and T cell clones (50,000 cells). [³H]Thymidine (1 μCi per well; PerkinElmer) was added after 48 h of incubation. Cells were harvested after another 16–18 h, using the Tomtec Harvester (Hamden, CT), and thymidine incorporation (measured as cpm) was measured by Wallac 1450 Microbeta Trilux Liquid Scintillation and Luminescence Counter (PerkinElmer).

For CFSE staining, thawed T cell clones were rested in RPMI 1640 with 10% human serum and 1% penicillin/streptomycin for 3 d after thawing. Cells were washed with PBS and then labeled with 1 μM CFSE in 5% FBS containing PBS for 5 min at 37°C. Labeling reaction was quenched by washes with ice-cold 20% FBS in PBS, twice. CFSE-labeled T cells were resuspended in RPMI 1640 with 5% human serum and 1% penicillin/streptomycin and cocultured with irradiated T2.DQ2 and T2.DQ2.DM at a 1:1 ratio (APC/T cell). The cocultures were stimulated with indicated peptides at 10 μg/ml or with Ag and incubated for 10 d. Cells were washed with flow buffer (2% BSA in PBS) and stained with propidium iodide before analysis on an LSR II cytometer at the Stanford Shared FACS Facility. Data were analyzed by FlowJo version 10 for Windows (FlowJo, LLC).

Surface DQ2 was directly stained by PE-conjugated anti-DQ mAb (Ia3; Leinco Technologies, St. Louis, MO) or indirectly stained using a primary Ab SPV-L3 (mouse anti-DQ) or 2.12.E11 (mouse anti-DQ2) followed by a secondary Ab, PE-conjugated goat anti-mouse IgG (Thermo Fisher Scientific). CLIP1 peptide (Ii aa 81–103, LPKPPKPVSKMRMATPLLMQALP) that is associated with DQ2 was stained by FITC-conjugated anti-CLIP mAb (CerCLIP; BD Biosciences). Directly labeled mAbs were used in case of costaining of multiple surface proteins. DM was detected using PE-conjugated anti-DM mAb (MaP.DM1; BD Biosciences) either on the surface of untreated cells or in cells that were fixed and permeabilized using the Cytofix/Cytoperm kit (BD Pharmingen, San Jose, CA). Fluorescently labeled cells were analyzed using a FACScan flow cytometer (BD Biosciences), and data were analyzed using the FlowJo software (FlowJo, LLC).

Data were analyzed for statistical significance using Prism software (version 4.0; GraphPad Software, San Diego, CA) and are expressed as mean ± SD. Statistical significance and p values are indicated on the figures, where appropriate. The p values <0.05 were considered statistically significant.

We previously constructed DM^null and DM^high APC cell lines expressing DQ2 as the only MHCII allele, without or with DM (T2.DQ2 and T2.DQ2.DM, respectively) (18). The amount of DM available for mediating interaction with DQ2 in T2.DQ2.DM cells is higher than in, for example, activated B cells, as modeled by B lymphoblastoid cell lines. This is not only because the transfected cells were selected to overexpress DM but also because other MHCII allelic proteins, which could compete with DQ2 for interaction with DM, are not expressed in these cells. In physiologic APC, codominantly expressed MHCII molecules with higher affinity for DM compromise DQ2 access to DM. The effects of increased DM/DQ2 ratios and consequent increased DM/DQ2 interaction in the T2.DQ2.DM transfectant include DM chaperoning of DQ2, which increases surface DQ2 levels [(18), Fig. 1A, Supplemental Fig. 1A, 1B], and DM catalysis of CLIP removal from DQ2, which reduces CLIP-associated DQ2 during biosynthesis. To observe the release of DQ2-associated CLIP, we immunoprecipitated DQ2 from T2.DQ2.DM and T2.DQ2 cells in a pulse-chase analysis (Fig. 1B). Freshly synthesized, metabolically labeled DQ2/CLIP complexes were observed within 1 d in T2.DQ2 but are undetectable in T2.DQ2.DM. In a related result, we also found that DQ2/CLIP complexes are nearly absent at the cell surface of T2.DQ2.DM (Fig. 1C, Supplemental Fig. 1C, 1D). Together, these results indicate that increased DM abundance can overcome its poor reactivity to DQ2.

DM is a lysosomal resident and is typically undetectable on the surface of APCs at physiological conditions because of the presence of a lysosomal sorting motif at DMβ cytoplasmic tail (38). To test whether the overexpression of DM in the transfected T2 cells led to the presence of DM at the cell surface, we used anti-DM mAb to stain the surface of T2.DQ2.DM. Flow cytometric analysis showed negligible levels of staining, compared with control lines that were DM deficient (Fig. 1D). This implied that CLIP removal occurred intracellularly, likely in MIIC, rather than at the cell surface.

Given the intracellular activity of DM in T2.DQ2.DM, these cells provided a tool to analyze the effect of DM on DQ2-restricted presentation to T cells of various Ags that require intracellular processing. To test gliadin epitope presentation, we used PT-gluten (see 2Materials and Methods), which is a mixture of relatively large (molecular mass >10 kDa) proteolytic fragments of gluten protein. PT-gluten was also treated with TG2, which converts selected glutamines to glutamic acid, mimicking the naturally occurring deamidation of gluten in the small intestine. As responding CD4⁺ T cells, we used a panel of four gliadin-specific T cell clones. These clones were isolated from the small intestine of CD patients, and each is specific for one of the following epitopes: α1a, α2, γ1, and γ4d [short for DQ2.5-glia-α1a, DQ2.5-glia-α2, DQ2.5-glia-γ1, and DQ2.5-glia-γ4d, respectively, as named elsewhere (35)]. Also, please see Supplemental Table I. These gliadin epitopes are implicated in CD pathogenesis, as they are recognized by intestinal T cells of the majority of adults with CD (39).

Previous work by our group (Sollid group) has shown that TG2-treated PT-gluten requires further processing to generate robust presentation of the relevant T cell epitopes for clones with the specificities we used in this study. This was demonstrated by the ability of irradiated cells to significantly outperform fixed cells as APC (31, 40). Of note, however, PT-gluten Ag preparations also include a low level of fragments that are stimulatory without further processing, presumably binding to surface DQ2 of fixed cells (31). We confirmed the need for processing for our PT-gluten preparation and also the low level of T cell stimulation by fixed cells, presumably from the availability of some smaller fragments in the preparation (Supplemental Fig. 2).

We next tested the influence of DM on the presentation of intracellularly processed gliadin, using irradiated DQ2⁺ T2 transfectants. The gliadin-specific human T cell clones were cocultured with irradiated T2.DQ2 or T2.DQ2.DM cells in the presence of deamidated PT-gluten, and the T cell proliferation was quantified. Both irradiated T2.DQ2 and T2.DQ2.DM cells were able to process PT-gluten and present the corresponding gliadin epitopes to activate the T cell clones in a dose-dependent manner. However, T cell proliferation for all four clones was diminished significantly in the presence of DM (Fig. 2), despite the higher level of DQ2 in T2.DQ2.DM cells (Fig. 1A).

To eliminate the possibility that TxB hybrid T2 cells provided an environment that uniquely allowed DM action on resistant MHCII alleles, we repeated the DQ2-restricted presentation of PT-gluten using another pair of DQ2⁺ APC lines: DM^null EBV-transformed B cell line 9.5.3 (DR3⁺/DQ2⁺/DP4⁺/DM⁻) and its repaired DM^high transfectant 9.5.3.DM^high (41). Similar to the differences observed using DQ2⁺ T2 cells as APC, the expression of DM increased the surface level of DQ2 in 9.5.3.DM^high compared with that of 9.5.3 (Fig. 3A), and 9.5.3.DM^high substantially lacked cell surface CLIP1/MHCII complexes (Fig. 3B), despite the presence of other MHCII alleles (DR3, DP4). Also, similar to the DM localization in T2.DQ2.DM, very little, if any, DM was detected on the surface of 9.5.3.DM^high (Fig. 3C). We then compared the capacity of irradiated 9.5.3.DM^high and 9.5.3 to present PT-gluten and to stimulate the proliferation of gliadin-specific T cell clones. Consistent with our previous findings, significant attenuation of T cell proliferation was observed when DM^high APC were used (Fig. 3D), despite the higher levels of DQ2 in 9.5.3.DM^high cells. Taken together, these findings argue that presentation of gliadin epitopes α1a, α2, γ1, and γ4d, after intracellular processing from gluten-derived products by gut APC, would be limited by functional interaction with DM. Consequently, the poor interaction of DQ2 with DM may contribute to CD pathogenesis.

As another approach to evaluate the suppressive effect of DM/DQ2 interaction on gliadin peptide presentation, we compared gluten presentation by WT DQ2 with presentation by the DQ2α+53G mutant with increased DM susceptibility (18). T cell proliferation of all four gliadin epitope-specific clones was significantly diminished when the T2 transfectant expressing DQ2α+53G was used as the APC (Fig. 4A). T2.DQ2.DM^high and T2.DQ2α+53G.DM^high express comparable levels of intracellular DM (18), ruling out an effect caused by DM abundance in the mutant. In addition, the decrease in T cell proliferation stimulated by T2.DQ2α+53G.DM^high did not result from lower DQ expression, as T2.DQ2α+53G.DM^high cells express significantly higher levels of surface DQ compared with T2.DQ2.DM^high cells (Fig. 4B). Notably, in the absence of DM, T2.DQ2α+53G expressed less surface DQ than T2.DQ2 (18). Therefore, the increased DQ expression on the surface of T2.DQ2α+53G.DM versus T2.DQ2.DM likely reflects another restored DM action: the enhanced DM/DQ2α+53G interaction allows DM to stabilize and rescue these MHCII molecules in MIIC (42, 43).

In addition to the increased DM/DQ interaction, the potential impaired binding of DQ2 mutant to gliadin peptides may also contribute to the DM-mediated suppression of T cell proliferation in response to gliadin peptides presented by T2.DQ2α+53G.DM^high APCs. To test this hypothesis, we ranked four gliadin peptides and three reference peptides based on their relative binding to soluble, recombinant DQ2α+53G as compared with WT DQ2, using a peptide-loading assay. In this assay, biotinylated peptides in excess will competitively replace CLIP1 peptide, engineered to be preloaded in the peptide-binding groove but exchangeable after cleavage of the covalent linker. Reference peptides include a DQ2-binding peptide derived from the MHCIα protein (18), CLIP1 (Ii_81–104), and CLIP2 (Ii_92–107). The relative binding of peptides followed the order CLIP2 > MHCIα ≈ CLIP1 ≈ P1269 (α1a) ≈ P1936 (γ4d) > PS1200 (α2) ≈ P1213 (γ1) for WT DQ2 and CLIP2 > MHCIα > CLIP1 > P1936 (γ4d) > P1269 (α1a) ≈ PS1200 (α2) ≈ P1213 (γ1) for DQ2α+53G (Fig. 5A). Unlike WT DQ2, which is capable of binding to all four gliadin peptides, DQ2α+53G barely bound gliadin peptides, except for a poor binder, P1936 containing the γ4d epitope. These findings confirmed that reduced binding of gliadin epitopes to mutant DQ2 coordinated with increased DM/DQ interaction for DM-mediated suppression of gliadin presentation by T2.DQ2α+53G.DM^high.

An interesting finding of the in vitro study is that although WT DQ2 bound gliadin peptides, none of these epitopes associated with DQ2 better than CLIP1(≈MHCIα < CLIP2). The intermediate binding capability (reflecting lower relative affinities compared with MHCIα) of tested gliadin peptides for DQ2 is the likely basis of reduced presentation of these peptides by DM^high APC and the associated attenuation of T cell proliferation. We next used the in vitro binding assay to test whether a moderate-affinity peptide like CLIP1 bound to DQ2 can be efficiently edited by DM and replaced by higher-affinity competitor peptides (e.g., MHCIα or CLIP2) when DM is abundant, regardless of DM susceptibility of DQ2 proteins. The catalytic effect of DM on peptide loading is best observed at early time points (37); therefore, we performed binding experiments with DM at 1 h and measured the fold change in binding to binding in the absence of DM (Fig. 5B). When present at lower ratios to DQ2 (1:1–4:1), DM had little effect on peptide loading to WT DQ2, in contrast to DQ2α+53G, which is DM susceptible. However, at a DM/DQ2 ratio of 16:1, DM enhanced peptide loading, especially of CLIP2, onto WT DQ2 preloaded with CLIP1 (Fig. 5B). These results agree with our findings using DM^high APCs and support the idea that with sufficient DM editing, the presentation of gliadin peptides would be suppressed, as their relative binding capacities to DQ2 (≤ CLIP1) are not as strong as one would expect for a DM-resistant epitope (see 18Discussion). Notably, the substantially increased DM/DQ2 ratio needed to drive interaction of soluble WT DQ2 and DM molecules is higher than that required for interaction of membrane-bound molecules (44). In our T cell assays, the DM/DQ2 ratio in DM^high APC is sufficient to replace intermediate-affinity gliadin peptides with higher-affinity peptides available in the late endosomal peptide-loading compartments, although this would not happen at constitutive DM/DQ2 ratios in professional APC.

To distinguish effects of gliadin peptide binding to DQ2 and DM/DQ2 interaction and to confirm that both effects underlie the DM-mediated suppression observed in DM^high APCs, we also tested presentation of exogenously loaded gliadin peptides, which bypass intracellular processing and DM editing. Unlike results with PT-gluten, which depends on intracellular processing and peptide loading, the presentation of all four gliadin peptides by T2.DQ2.DM^high increased T cell stimulation to varying degrees (Fig. 6A). This is consistent with peptide loading occurring at the cell surface or in early endosomes. At these sites, DM action is typically low because of lower steady-state DM levels and higher pH than in lysosomes, and DQ2 expression levels primarily contribute to the amount of gliadin peptide that is loaded and presented to T cells. Exogenous peptide presentation also allowed us to evaluate the contribution of DQ2-peptide binding affinity to gliadin presentation. We compared T2.DQ2α+53G.DM to T2.DQ2.DM for the capacity to present the four gliadin peptides to T cell clones. In this comparison, the decreased T cell stimulation by the mutant APC (Fig. 6B) likely results from reduced peptide loading or less-stable binding DQ2α+53G. An efficiently DM-edited peptide repertoire, consisting of mostly stably bound peptides in association with surface DQ2α+53G, also could contribute to poor peptide exchange for exogenous gliadin peptides. Together with analyses on the presentation of PT-gluten, these findings implicate increased DM abundance, the restoration of DM interaction with DQ2, and reduced peptide binding as sources of suppressed stimulation of DQ2-restricted, gliadin-specific T cells from CD patients.

The association of DQ2 with suboptimal responses to some viruses (10, 13, 14, 45) raised the possibility that its reduced interaction with DM might also lead to presentation of moderate-affinity (DM-sensitive) viral peptides, whose unstable binding to DQ2 would reduce the surface t_1/2 of the DQ2/peptide complex and compromise CD4⁺ T cell responses (20, 46). We therefore assessed the effect of DM on the activation of DQ2-restricted T cell clones specific for epitopes from viral proteins. The T cell clone 1A.B.25 is specific for a DQ2-restricted VP16 epitope (aa 431–440) derived from HSV-2 proteins. Allogeneic APC presentation of the epitope determined that the MHCII restriction of this clone matched only at DQ2 allele; this DQ2 restriction was further validated by selective inhibition using anti-DQ (but neither anti-DR nor anti-DP) mAb and strong peptide binding to both purified DQ2 proteins and DQ2-expressing cell lines (38, 47). In contrast to PT-gluten presentation to gliadin-specific T cell clones, the presentation of UV-inactivated HSV-2 by irradiated T2.DQ2.DM^high significantly increased the proliferation of 1A.B.25 (>5-fold) compared with the stimulation using T2.DQ2 (Fig. 7A). This dramatic increase was unlikely due to only the modest (∼2×) DM-dependent increase in surface DQ2 expression. Rather, this result suggested improved presentation of the high-affinity VP16-derived peptide (48) by DM^high APC. We next tested a DQ2-restricted T cell clone specific for HA (HA_97–113). Using native HA from influenza A/New Caledonia/20/1999 (H1N1) as Ag, we observed increased T cell proliferation with T2.DQ2.DM^high compared with T2.DQ2 cells as APC, again with the difference not likely to be fully explained by different levels of surface DQ2. The comparable levels of T cell proliferation in response to peptide (HA_97–113) presentation despite higher DQ2 on the DM^high APC is consistent with reduced peptide loading in the presence of an edited DQ2-associated peptide repertoire in these cells (Fig. 7B). This result also argues that DM^null APC are comparable to DM^high APC when the intracellular processing and presentation pathway is bypassed. Last, we tested a DQ2-restricted CD4 T cell clone specific for an epitope within the E7 protein of HPV16 (E7_22–56). Similar to the other viral peptide-specific T cells, proliferation was enhanced using DM^high APC when Ag was provided as intact protein, particularly at lower doses of Ag, in which DM editing has a greater effect on the level of MHCII/peptide complex. T cell proliferation was not significantly different using extracellular peptide-pulsing of DM^null and DM^high APC lines (Fig. 7B, 7C, Supplemental Fig. 3A, 3B). Thus, DM enhanced presentation of three different DQ2-restricted viral epitopes after intracellular processing and peptide loading.

In this paper, we report differential proliferation of DQ2-restricted, Ag-specific CD4⁺ T cells in response to Ag presented by paired DM^high/DM^null APC. Each pair is composed of isogenic lines differing only in the expression of DM. Also, the lines are polyclonal, so that an unusual clonal phenotype arising from a retroviral insertion site is unlikely to influence the assay results. The use of these model APC provides a robust approach to isolate the effect of DM on the presentation of DQ2-restricted epitopes.

Using a panel of DQ2-restricted T cell clones specific for different non–self-epitopes, we observe a striking pattern: gliadin epitopes associated with developing autoimmunity in CD are DM sensitive (suppressing their presentation), and viral epitopes related to host defense are DM resistant (promoting their presentation). Results broadly similar to ours, although not specifically related to presentation by DQ2, show DM antagonism for DR*04:01-restricted epitopes from glutamate decarboxylase (GAD_273–285), a type 1 diabetes autoantigen (49), and type II collagen (CII_261–273), a rheumatoid arthritis autoantigen (50), whereas numerous pathogen-derived peptides are DM resistant (51, 52). Interestingly, in an in vitro Ag presentation assay system, distinct paths for peptide processing and selection were observed for pathogen-derived proteins compared with autoantigens (23).

Expression of DM at higher than physiological levels in cells allowed us to determine that, after intracellular generation, the DQ2/gliadin complexes we studied are DM sensitive. Some disease-associated gliadin epitopes, for example, the γ1 epitope, are generated by intracellular processing from larger gliadin fragments produced by digestive enzymes (53). These observations raise the question as to how such gliadin peptides get presented, as they must, because T cells specific for these complexes can be isolated. Unlike our DM^high APC lines, professional APC (dendritic cells, macrophages, and B cells) express substoichiometric DM levels relative to MHCII molecules (54), and the interaction time for DM and class II molecules in endosomal compartments is thought to be limited (55). In addition, in professional APC, access to DM is further compromised for DQ2 by competition with other alleles, including DR3, which is linked in the DR3/DQ2 haplotype (56), and interacts efficiently with DM (17). Indeed, in B cells expressing DR3, DQ2, and DM, DQ2 is primarily associated with CLIP peptides, even at the cell surface, whereas DR3 is effectively edited and carries heterogeneous (non-CLIP) peptide cargo (17). Constitutive DM levels in APC in the small intestine are likely insufficient for effective intracellular editing of DQ2–gliadin complexes, providing a path for presentation of these DM-sensitive complexes to T cells.

We also addressed the impact of DM/DQ2 interaction on DM-sensitive epitopes using a mutant, DQ2α+53G, which is a better DM substrate than WT DQ2. Unlike most MHCII, DQ2 has a deletion in the α-chain α helix that leads to loss of a hydrogen bond with the peptide backbone that normally stabilizes the complex in the region of the P1 pocket (57). Gliadin-derived epitopes place a proline residue in the P1 pocket of DQ2 and stabilize binding with alternative hydrogen bonds elsewhere in the binding groove (5, 58). However, crystal structures show that the DQ2α53 deletion affects a region predicted to interact with DM (5, 59) and thus likely contributes to diminished DM susceptibility of DQ2. DQ2α+53G rescues DM susceptibility but, as shown in this study, also reduces gliadin peptide binding. Thus, for WT DQ2, the combination of impaired DM interaction and sufficient gliadin peptide binding leads to efficient presentation of DM-sensitive gliadin peptides.

Our analysis using APC to directly present epitope peptides also showed that DM-sensitive gliadin peptides may be presented to T cells after loading by secondary pathways that avoid the DM/DQ2 interaction. Some gliadin peptides generated in the digestive tract can activate specific CD4⁺ T cells without further proteolytic processing in intracellular, late endosomal MHCII compartments (31). Thus, one scenario is binding of such peptides at the plasma membrane or in early endosomes. In these locations, DM editing of peptide/MHC complexes is inefficient because of low steady-state DM levels and pH (6.5–7.4), which is unfavorable for DM action (37). In addition, reduced efficiency of DM editing of the DQ2-bound peptide repertoire during biosynthesis yields surface DQ2/peptide complexes that are particularly susceptible to exchange and presentation of extracellular peptides (60).

Dramatically, we demonstrated that, in contrast to gluten-derived gliadin peptides, three viral Ags triggered significantly increased proliferation of epitope-specific T cell clones in response to DM^high APC presentation of the corresponding epitopes. Another DQ2-restricted T cell clone specific for EBV gp350 peptide was also found (by one of us) to be stimulated 8-fold more effectively by DM^high, compared with DM^null, APC after incubation with native protein Ag (52). In vivo, the presentation of these DQ2-restricted high-affinity epitopes could occur even at low DM levels through competition with CLIP1/2 and/or endosomal peptides. More likely, perhaps, is that during viral infection, DM activity is increased in activated APC (61, 62). In the specific case of immune activation by viruses, this can arise by direct infection of APC, such as dendritic cell infection by influenza (63) or B cell infection by EBV (64). Notably, EBV gp42, which is essential for viral entry and binds MHCII, preferentially binds to DQ2 (65). In addition, and especially in the case of viruses like HSV, which may not infect professional APC, detection of viral Ag by professional APC or the microenvironment/cytokine milieu of infected cells may be routes to APC activation and increased DM levels (66–70).

In summary, we propose that in CD, insufficient DM accessibility by DQ2, impaired DM/DQ2 interaction, moderately increased DQ2/peptide affinity, and bypassing of DM peptide editing all contribute to the uniquely selective DQ2 presentation of DM-sensitive gliadin epitopes. In contrast, presentation of DM-resistant epitopes that form more-stable complexes with DQ2 likely relies less on the above mechanisms, as DM editing positively affects presentation of these epitopes. Our findings suggest that the elevation of DM expression in peripheral APC (particularly during infection) may benefit self-tolerance by attenuating presentation of DM-sensitive epitopes while boosting presentation of DM-resistant pathogen-derived epitopes and aiding in host defense.

We thank Dr. Robert Belshe at the Center for Vaccine Development, Saint Louis University, for providing inactivated A/New Caledonia/20/99 (H1N1) influenza. We also thank Dr. Xi Jin from the laboratory of Dr. Chaitan Khosla, Department of Chemistry, Stanford University, for synthesizing gliadin peptides. Also, we thank the Vaccine and Infectious Diseases Division, Fred Hutchinson Cancer Research Center, for providing inactivated HSV-2 Ag and T cell clone.

The authors have no financial conflicts of interest.