Discriminative T cell recognition of cross-reactive islet-antigens is associated with HLA-DQ8 transdimer–mediated autoimmune diabetes

See allHide authors and affiliations

Science Advances  21 Aug 2019:
Vol. 5, no. 8, eaaw9336
DOI: 10.1126/sciadv.aaw9336


Human leukocyte antigen (HLA)–DQ8 transdimer (HLA-DQA1*0501/DQB1*0302) confers exceptionally high risk in autoimmune diabetes. However, little is known about HLA-DQ8 transdimer–restricted CD4 T cell recognition, an event crucial for triggering HLA-DQ8 transdimer–specific anti-islet immunity. Here, we report a high degree of epitope overlap and T cell promiscuity between susceptible HLA-DQ8 and HLA-DQ8 transdimer. Despite preservation of putative residues for T cell receptor (TCR) contact, stronger disease-associated responses to cross-reactive, immunodominant islet epitopes are elicited by HLA-DQ8 transdimer. Mutagenesis at the α chain of HLA-DQ8 transdimer in complex with the disease-relevant GAD65250–266 peptide and in silico analysis reveal the DQ α52 residue located within the N-terminal edge of the peptide-binding cleft for the enhanced T cell reactivity, altering avidity and biophysical affinity between TCR and HLA-peptide complexes. Accordingly, a structurally promiscuous but nondegenerate TCR-HLA-peptide interface is pivotal for HLA-DQ8 transdimer–mediated autoimmune diabetes.


Major histocompatibility complex (MHC) genes, also known as human leukocyte antigen (HLA) genes in humans, are the prominent susceptibility factor for many autoimmune diseases, including multiple sclerosis (MS), autoimmune diabetes (type 1 diabetes or T1D), and rheumatoid arthritis (RA). Several hypotheses for MHC-linked susceptibility in autoimmune diseases have been proposed, including preferential accommodation of self-peptides or altered self-peptides derived from the target organ of each disease by MHC molecules, promotion of autoimmunity through molecular mimicry between self- and microbial antigens, and unstable trimolecular interactions among the T cell receptor (TCR)–peptide-MHC (pMHC) complexes that facilitate the escape of self-reactive T cells from negative selection [recent reviews in (1, 2)]. However, because of limited knowledge of T cell epitopes and scarcity of self-reactive T cells, the molecular and structural association between susceptible alleles and T cell autoreactivity is not clearly understood.

T1D is a chronic disease that causes severe loss of insulin-producing β cells in the pancreatic islets of Langerhans. T1D is predominantly linked to genetically predisposed individuals carrying HLA-DQ8 (encoded by HLA-DQA1*0301-DQB1*0302 and hereafter referred to as DQ8cis) or HLA-DQ2 (encoded by HLA-DQA1*0501-DQB1*0201 and hereafter referred to as DQ2cis) haplotypes (3, 4). HLA-DQ8/DQ2 heterozygous individuals confer a synergistically higher risk for disease (57). Epidemiological studies link this high risk to the naturally occurring trans-complementary molecule HLA-DQ8 transdimer (hereafter referred to as DQ8trans), consisting of the DQ2cis α chain (DQA1*0501) and the DQ8cis β chain (DQB1*0302) (57). The other trans-complementary molecule HLA-DQ2 transdimer, consisting of the DQ8cis α chain (DQA1*0301) and the DQ2cis β chain (DQB1*0201), also confers T1D risk but to a lesser extent (7).

Despite conferring exceptionally high risk in T1D, little is known about the DQ8trans-restricted epitope repertoire and T cell recognition. DQ8cis and DQ8trans molecules have an identical β chain (DQB1*0302). Protein sequence alignment by BLAST (Basic Local Alignment Search Tool) reveals a high similarity between the α chain of DQ8cis (DQA1*0301) and the α chain of DQ8trans (DQA1*0501), but with distinct residues, namely, α23, α31, α37, α44, α47, α48, α50, α52, α72, and α73, that are proximal to or within the peptide-binding cleft (8, 9). A special feature of the DQ susceptibility alleles in T1D is the absence of an aspartic acid residue at position 57 of the β chain (4), which leads to a highly skewed peptide repertoire favoring a negative charge at the p9 anchor in DQ8cis (10, 11). Although the differences in α chain may lead to distinct peptide-binding specificities between DQ8cis and DQ8trans (12), cross-recognition of both DQ8cis and DQ8trans by CD4+ T cell clones specific for some self-epitopes has been reported (1316), suggesting convergence of TCR footprints onto both alleles (9) due to “missing the differences” form of molecular mimicry. In light of this, elucidating the molecular basis of T cell recognition of DQ8cis/trans molecules may advance the understanding of DQ8trans-specific immunity and the mechanism by which DQ8trans increases the risk for T1D.

In this study, we investigate the structural and functional attributes of DQ8cis/trans that are critical to T1D progression by an integrated in vitro, direct ex vivo, biophysical, and in silico analysis of specific CD4+ T cells using DQ8cis, DQ2cis, and DQ8trans tetramer reagents. We find a high degree of T cell promiscuity between DQ8cis and DQ8trans for self- and nonself-antigens. T cells specific for cross-reactive, disease-relevant islet epitopes show an enhanced responsiveness toward DQ8trans-peptide (pDQ8trans) complexes compared to DQ8cis-peptide (pDQ8cis) complexes, as a result of increased TCR-pDQ8trans affinity. Modified TCR recognition toward immunodominant and disease-relevant epitopes is associated with DQ α chain residues at the N-terminal end of peptide-binding cleft that potentially involves reorientation of the epitopes under the context of different DQ α chains. Thus, preferential TCR cross-recognition of pDQ8trans complexes through specific DQ α chain residues provides a molecular explanation for the extremely high risk of developing T1D that is conferred by DQ8trans.


A high degree of epitope sharing and T cell promiscuity is observed between T1D-susceptible DQ8cis and DQ8trans alleles

We assessed the structural and functional properties of DQ8/DQ2 cis and trans molecules by examination of T cells specific for common Influenza A epitopes, as well as multiple T1D-associated islet antigen epitopes. Tetramer-guided epitope mapping (17) was applied to identify influenza-specific epitopes. Briefly, total CD4+ T cells from multiple DQ8+DQ2+ participants were stimulated using pooled peptides spanning Influenza A/New Caledonia/20/1999 Hemagglutinin (H1HA or HA) and Influenza A/New York/318/2003 Nucleoprotein (H1NP or NP). After culturing for 14 days, antigenic epitopes were identified with DQ8cis, DQ2cis, and DQ8trans tetramers. A representative DQ8trans tetramer staining for pooled and individual epitope mapping of H1HA is shown in fig. S1. We observed extensive epitope sharing between DQ8cis and DQ8trans for Influenza A antigens (Fig. 1A and Table 1). Most identified epitopes were stained by both tetramers (Table 1, group I). Peptides HA102–118/HA108–124, HA274–290/HA280–296, HA398–414/HA404–420, NP288–304/NP294–310, and NP480–496/NP486–498 were single overlapping epitopes as specific T cells were stained by tetramers with either peptide. In vitro expanded primary T cells specific for HA186–202, HA340–356, HA422–438, HA440–456, HA452–468, NP175–191, and NP240–256, by contrast, were visualized by either DQ8cis or DQ8trans tetramer only (Table 1, group II, group III, unique epitopes for DQ8cis, and inconclusive). To confirm restriction of each cis- or transdimer, human embryonic kidney (HEK) 293 cells, which are commonly used for antigen presentation studies (18, 19), were transfected with DQcis or DQtrans expression constructs. DQ8cis-, DQ2cis-, DQ8trans-, and DQ2trans-transfected HEK293 cells had comparable surface DQ expression, as determined by the mean fluorescence intensity of anti-DQ (SPVL3) antibody staining (fig. S1C). Responses of T cell clones specific for HA186–202, HA440–456, NP175–191, and NP240–256 indicated that these cells could be activated by either DQ8cis or DQ8trans in T cell proliferation assay. On the other hand, multiple DQ8cis-restricted HA452–468-specific clones from two individuals failed to respond to pDQ8trans in T cell proliferation assay, confirming HA452–468 as a unique DQ8cis epitope. We were unable to isolate DQ8cis-restricted HA340–356- and HA422–438-specific lines or clones to attribute the recognition of these specificities to pDQ8trans. In contrast, only one DQ2cis-restricted HA peptide (HA102–118/HA108–124) was accommodated by DQ8cis/trans (Fig. 1A and table S1).

Fig. 1 A high degree of epitope sharing and T cell promiscuity is observed between DQ8cis and DQ8trans for influenza and islet antigens.

(A) A high degree of epitope overlap (highlighted in red) was observed between DQ8cis- and DQ8trans-restricted influenza-specific T cells. In contrast, only one DQ2cis-restricted epitope (HA102–118/HA108–124) was accommodated by DQ8cis/trans. Numbers indicate epitopes shared between DQ8cis, DQ2cis, and DQ8trans or exclusively presented by DQ8cis, DQ2cis, or DQ8trans. Each overlapping epitope (HA102–118/HA108–124, HA274–290/HA280–296, HA398–414/HA404–420, NP288–304/NP294–310, or NP480–496/NP486–498) was viewed as one epitope. (B and C) Cross-recognition of islet-specific clones was confirmed by T cell reactivity in the presence of 500 nM peptides. Black and white bars represent T cell reactivity stimulated by DQ8cis- and DQ8trans-transfected HEK293 cells, respectively. Error bars represent SDs among three experiments. SI, stimulation index.

Table 1 List of unique and shared epitopes derived from common Influenza A and islet antigens for DQ8cis and DQ8trans.

Ins, insulin; ICA69, islet cell antigen 69.

View this table:

The identification of overlapping Influenza A epitope repertoires between DQ8cis and DQ8trans prompted us to investigate whether cross-recognition of peptides presented by both restriction elements also occurs in islet antigens. We screened a total of 130 possible DQ8cis/trans-binding peptides from important islet antigens, specifically from chromogranin A, cyproheptadine, islet autoantigen 69 (ICA69), islet-specific glucose-6-phosphatase catalytic subunit–related protein (IGRP), γ-aminobutyric acid (GABA) receptor–associated protein, glutamic acid decarboxylase 65 (GAD65), phogrin, preproinsulin, tyrosine phosphatase–related islet antigen–2 (IA-2), secretory granule proteins, synaptic adhesion molecules, and zinc transporter 8 (ZnT8) (for peptide selection, see Materials and Methods). The list included several peptides exclusively presented by DQ8trans (12). We identified eight previously unknown peptides (Table 1) that elicited in vitro responses from peripheral blood mononuclear cells (PBMCs) of T1D participants. No T cell responses to peptides exclusively presented by DQ8trans (12) were detected. Cross-reactivity to these eight epitopes together with InsB11–24, GAD65121–140, and GAD65250–266, which were published in our previous studies (20, 21), was examined by tetramer staining and functional responses of T cell clones. To avoid any bias in TCR selection due to the presence of DQ8trans, we set up stimulation in DQ8+DQ2 participants and isolated clones using DQ8cis tetramers. GAD65250–266-, InsB11–24-, IA-2761–777–, IA-2957–972–, IGRP306–320-, and ZnT815–29-specific clones from several participants were stained by both DQ8cis and DQ8trans tetramers (fig. S2A). Cross-recognition of these six epitopes was confirmed by T cell reactivity (Fig. 1, B and C). Clones specific for GAD65121–140, IA-2168–182, IA-2273–287, ICA69297–311, and Phogrin323–337 were not effectively stained by DQ8trans tetramers (fig. S2B) but proliferated in response to pDQ8trans (Fig. 1B). Together, 16 of 19 (84%) influenza epitopes and 11 of 11 islet epitopes (100%) were cross-recognized.

Cross-recognition of islet-specific T cells occurs in ex vivo unmanipulated, polyclonal populations

As the aforementioned cross-reactivity was observed by using T cell clones/lines, we scrutinized the cross-recognition of polyclonal and unmanipulated islet-specific T cells by direct ex vivo analysis in DQ8+DQ2 participants. GAD65250–266- and IA-2957–972–specific cells were investigated as these two epitopes elicited in vitro responses in more than half of T1D participants screened. T cell cross-recognition to polyclonal DQ8cis/GAD65250–266-specific and DQ8trans/GAD65250–266-specific cells was validated by double staining of ex vivo enriched cells with DQ8cis and DQ8trans tetramers [Fig. 2A; the specificity of ex vivo staining for DQ8/GAD65250–266 has been demonstrated in our previous study (20)]. In contrast, IA-2957–972–specific T cells could only be effectively visualized by DQ8trans tetramers; however, functional cross-reactivity and specificity of the staining was revealed when ex vivo sorted DQ8trans/IA-2957–972–specific clones from DQ8 (DQA1*0301-DQB1*0302) homozygous patients were stimulated with the specific peptide presented by DQ8cis or DQ8trans (Fig. 2B). A representative ex vivo staining from one T1D and one healthy participant is shown in fig. S2. Results of ex vivo staining experiments are summarized in table S2. Higher frequencies of GAD65250–266- and IA-2957–972–specific cells were observed in both DQ8+DQ2 and DQ8+DQ2+ T1D patients compared to DQ8+ control participants, but we did not observe differences between DQ8+DQ2 and DQ8+DQ2+ patients (Fig. 2, C and D). Overall, our in vitro and ex vivo data suggested that GAD65250–266- and IA-2957–972–specific T cells are immunodominant and disease relevant in human T1D, and T cell promiscuity in the context of DQ8cis/trans for these two epitopes was present in unmanipulated, polyclonal T cell populations.

Fig. 2 Direct ex vivo cross-recognition analysis of disease-relevant islet-specific T cells.

(A) Double staining by DQ8cis and DQ8trans tetramers was observed for ex vivo polyclonal GAD65250–266-specific cells in DQ8+DQ2 participants. DQ8trans/HA102–188 tetramer was used as a control. The double staining shown here was performed in DQ8 (DQA1*0301-DQB1*0302) homozygous individuals. (B) Ex vivo sorted DQ8trans/IA-2957–972–specific clones cross-recognized DQ8cis and DQ8trans. (C and D) Comparison of total (left) and memory (right) CD4+ T cell frequencies for GAD65250–266-specific (C) and for IA-2957–972–specific (D) T cells among DQ8+DQ2 patients (closed circles), DQ8+DQ2+ patients (open circles), and DQ8+ healthy controls (gray circles). The y axis in (C) and (D) indicates T cell frequencies. Statistical significance was evaluated by one-way analysis of variance (ANOVA) followed by Bonferroni’s test. *P < 0.05, **P < 0.01, ***P < 0.001; ns, not significant (P > 0.05). Tmr, tetramer.

DQ8cis- and DQ8trans-restricted epitopes elicit T cell responses by conserved putative TCR contact residues

We next investigated whether DQ8cis/trans cross-reactivity to InsB11–24 (the primary autoantigen in the nonobese diabetic mouse model and relevant in human T1D (21, 22)), GAD65250–266, and IA-2957–972 epitopes was mediated by the same TCR recognition sites using single-site alanine scanning mutagenesis. Several clones from different participants were investigated, and a representative result was shown in fig. S3B. Previous and current data demonstrated that the minimal T cell epitopes for GAD65250–266, InsB11–24, and IA-2957–972 are GAD65253–261 (23), InsB14–22 (21), and IA-2961–969 (fig. S3A), respectively. TCRs interacted with GAD65250–263, InsB11–24, and IA-2957–972 across the entire peptide region, as mutations at both middle and end residues abolished the recognition of T cell clones. T cell reactivity profiles for these epitopes were similar between DQ8cis and DQ8trans as substitutions at given positions of the epitope of the DQ8cis/trans complexes have very similar to identical effects on the cognate T cell clones. These data indicate that both molecules elicit T cell responses via the same TCR recognition motif (fig. S3B). Studies by Eerligh et al. (24) suggest that the DQ8 cis- and transdimers present InsB6–22 in a different binding register. This discrepancy is likely accounted for by a single amino acid shift in the antigenic motif recognized by our insulin clones [14ALYLVCGER22 (21), the predominant form recognized in DQ8+ individuals (25)] versus those of Eerligh et al. (13EALYLVCGE21).

Islet autoimmunity is promoted by DQ8trans

The functional implication of T cell cross-reactivity to immunodominant InsB11–24, GAD65250–266, and IA-2957–972 epitopes was explored using islet-specific clones isolated with DQ8cis tetramers from in vitro expanded cultures of memory CD4+ T cells in multiple DQ8+ T1D participants. All six GAD65250–266 clones proliferated with pDQ8cis and pDQ8trans, but pDQ8trans elicited greater responses across all but one of the clones (Fig. 3A). Similarly, all IA-2957–972–specific clones proliferated more strongly in response to DQ8trans when stimulated with specific peptide (Fig. 3B). pDQ8trans also induced greater cytokine production from cross-reactive T cell clones that recognized pDQ8trans better, including significantly more interferon-γ (IFN-γ) secretion from pDQ8trans-induced GAD65250–266-specific clones [Fig. 3C and table S3, P < 0.05 by paired t tests, compared to 0.2574 for interleukin-4 (IL-4) and 0.1862 for interleukin-10 (IL-10)]. Similar IFN-γ, IL-4, and IL-10 secretion results as to GAD65250–266-specific clones were seen for IA-2957–972–specific clones (Fig. 3C and table S3). Although InsB11–24-specific clones did not proliferate strongly in response to peptide stimulation, higher levels of IFN-γ secretion were induced (from three different InsB11–24 clones) by pDQ8trans, compared to the levels of IFN-γ induced by pDQ8cis (Fig. 3C). A summary of the T cell clones, their proliferative capacity, and cytokine secretion is summarized in table S3. Thus, islet-specific cross-reactive T cells respond more strongly to DQ8trans-presented peptides than DQ8cis-presented peptides and produce higher amounts of proinflammatory cytokines associated with T1D pathogenesis in the former compared to the latter case, regardless of the DQ status (DQ8+DQ2 or DQ8+DQ2+) of the host.

Fig. 3 pDQ8trans most often elicit stronger disease-associated responses from cross-reactive islet-specific T cell clones.

(A) All but one GAD65250–266-specific clones were activated more vigorously by DQ8trans in the presence of 50 and 500 nM peptides. (B) DQ8trans induced stronger responses from IA-2957–972–specific clones in the presence of 50 and 500 nM peptides. (C) Secretion of IFN-γ was significantly higher from cross-reactive T cell clones when stimulated with pDQ8trans. Peptide concentrations used in (C) were as follows: 5 nM (GAD65250–266), 50 nM (IA-2957–972), 2.5 μM (InsB11–24), which were the minimal concentrations that cytokine secretion could be detected for specific clones. Error bars in (A) and (B) represent SDs among three experiments. Comparison of T cell responses to DQ8cis and DQ8trans was evaluated by paired t tests. Statistical analysis in (A) and (B) was not performed as SIs for some clones were ≤3.

Enhanced T cell recognition of DQ8trans/GAD65250–266 is correlated with biophysical interactions and functional avidity

A direct correlation exists between two-dimensional (2D) affinity and T cell functional responses (26, 27). We next investigated the 2D affinity and cross-recognition of pDQ8cis and pDQ8trans to GAD65250–266-specific CD4 T cells using purified covalent DQ8cis- or DQ8trans-GAD65250–266 protein complexes. We found that clones 03-C1, 05-C1, and 07-C1 all had significantly higher affinity (>2-fold) for the pDQ8trans molecule (Fig. 4A), which would be in agreement with the observed increases in the stimulation index (SI) (Fig. 3A). In contrast, the 09-C1 clone had higher affinity for and was more responsive to pDQ8cis (Figs. 4A and 3A, respectively). We also evaluated the 2D affinity of influenza matrix protein–specific clone and its 2D affinity also correlated with T cell function. Dose-response curves of T cells have been used to generate functional avidity comparisons between T cells. The functional avidity of GAD65250–266-specific T cell clones was also correlated with pMHC potency as the dose-response curves suggested that higher pDQ8cis concentrations were required to reach a similar response to that of pDQ8trans (Fig. 4B). On the other hand, no apparent variation in functional avidity between DQ8cis and DQ8trans for GAD65250–266-specific T cell clone 09-C1, which exhibited stronger responses and biophysical interactions for DQ8cis, was observed (Fig. 4B). The lack of difference might be associated with the unusual high affinity of cognate TCR as TCR with affinities beyond the natural range can initiate T cells responses faster than TCR within the natural range, although both may endow T cells with the same functional avidity (28).

Fig. 4 Altered T cell avidity for DQ8trans/GAD65250–266 is associated with the absence of Arg52α residue in DQA1*0501.

(A) 2D affinity of five antigen-reactive clones to DQ8cis (closed shapes) or DQ8trans (open shapes). 2D affinity correlates with T cell potency. MP, matrix protein. (B) Stronger responses of GAD65250–266-specific clones to pDQ8trans were correlated with higher TCR-pMHC avidity. Arg52α and, to a lesser extent, α44–50 residues for clone 07-C1 were critical to the enhanced T cell avidity toward DQ8trans/GAD65250–266. Experiments were run in duplicates with the error bars denoting the SD. Statistical significance was evaluated by extra sum-of-squares F test and listed in table S4. WT, wild-type; CI, confidence interval. (C) Amino acid sequence alignment of the core peptide-binding cleft region between the α chain of DQ8cis (DQA1*0301) and the α chain of DQ8trans (DQA1*0501). The parallel lines in DQA1*0501 indicate identical amino acid residues between two α chains. The numbers above the amino acid sequence are the sequence number of the residue in the α chain, as presented in (44), ensuring structural equivalence of the respective residues and their counterparts in HLA-DRA. The regions in which mutations were made in DQA1*0501 are highlighted in red. **P < 0.01, ****P < 0.0001.

To assess the contribution of distinct α chain residues to altered T cell functional avidity, a set of DQ8trans mutants in complex with GAD65250–266 were made in which specific α chain residues adjacent to or within the peptide-binding pockets [namely, α31 and α52 in pocket 1, α44–50 at the lateral side of pocket 1, and α72–73 in pocket 9 (8, 9)] were replaced with those of DQ8cis (table S4 and Fig. 4, B and C). The corresponding residues Glu31α, Arg52α, Ile72α, and Val73α in DQ8cis were Gln31α, a deletion in position 52, Ser72α, and Leu73α in DQ8trans (Fig. 4C). We were not able to produce DQ8trans mutant protein with addition of Arg52α alone (the residue absent in the pocket 1 of DQ8trans) and had to generate this mutant protein in combination with other mutations (DQ8trans αQ31E, αF51_αR53insR, and DQ8trans ∆α44–50, αW43_αF51insDQ8cis α44–50, αF51_αR53insR in table S4 and Fig. 4B). T cell functional avidity was not affected with substitution of α31 or α72–73 (DQ8trans αQ31E and DQ8trans αS72I, αL73V in table S4 and Fig. 4B) and was only slightly reduced for GAD65250–266 clone 07-C1 with the substitution of α44–50 (DQ8trans ∆α44–50, αW43_αF51insDQ8cis α44–50 in table S4 and Fig. 4B). Inclusion of Arg52α with α31 or α44–50 substitutions in DQ8trans, however, significantly reduced responses of T cell clones to a level similar to that of DQ8cis (table S4 and Fig. 4B). As substitution of either α31 or α44–50 alone had a minimal effect, these results highlight the importance of inserting an additional Arg in between Phe51α and Arg53α in DQ8trans diminishing T cell responses to pDQ8trans. The pivotal role of the Arg52α residue in the regulation of TCR reactivity was further confirmed by the observation that T cell functional avidity for DQ8cis was enhanced by replacing the α44–52 domain in DQ8cis with the corresponding domain from DQ8trans (DQ8cis ∆α44–50, ∆α52, αW43_αF51insDQ8trans α44–50 in table S4 and Fig. 4B). In total, these data suggest that the presence or the absence of Arg52α in DQA1 chain plays an important role in regulating DQ8cis/trans-restricted T cell responses.

Altered orientation of important TCR-contact residues in the bound peptides by the absence of Arg52α residue in DQA1*0501

Although Arg52α can potentially dictate peptide-binding preference at pocket 1 (9), the differences in T cell reactivity to pDQ8cis/pDQ8trans cannot be adequately explained by pMHC affinity as the shared GAD65250–266 and IA-2957–972 epitopes bound comparably to both molecules (fig. S4A). Thus, altered T cell avidity by Arg52α might reflect a differential display of the bounded peptide toward TCR. This scenario was investigated by in silico modeling analysis. The side view of the superimposed GAD65251–263 bound by DQ8cis and DQ8trans revealed almost identical positioning for anchor residues p4Phe, p6Met, p7Phe, and p9Glu (Fig. 5A). However, a slightly different orientation at the p1Ile anchor led to a considerable divergence at the TCR contact surface, including the important TCR contact residues p3Arg, p5Lys, and p11Lys (Fig. 5A and fig. S3B). A similar observation was found in the superposition of IA-2957–972, with distinct orientation at potential TCR-contacting p3Leu and p8Glu residues (Fig. 5A and fig. S3B). Replacement of the α44–50 residues of DQ8trans with those of DQ8cis and insertion of Arg52α into DQ8trans (DQ8trans ∆α44–50, αW43_αF51insDQ8cis α44–50, αF51_αR53insR) resulted in a nearly matched peptide orientation between DQ8cis and DQ8trans in the context of GAD65250–266 and IA-2957–972 epitopes (Fig. 5B), in contrast to little change of peptide orientation in DQ8trans when only the α44–50 substitutions were in place (DQ8trans ∆α44–50, αW43_αF51insDQ8cis α44–50).

Fig. 5 In silico analysis of epitope presentation in the context of different DQ molecules.

(A) Superimposed side view of GAD65251–263, InsB12–23, and IA-2959–971 shown in stick form in the grooves of DQ8cis (nitrogen, blue; oxygen, red; sulfur, yellow; all other atoms, magenta) and DQ8trans (same atom color convention, for N, O, and S; all other atoms, green). (B) A nearly matched peptide orientation was seen in the superimposed side view of GAD65251–263, InsB12–23, and IA-2959–971 in the grooves of DQ8cis (magenta) and DQ8trans ∆α44–50, αW43_αF51insDQ8cis α44–50, αF51_αR53insR (green). Same depiction conventions and atom color codes are used as in (A). (C) TCR view of pocket 9 in complex with InsB12–23 revealed differences in residue interaction (listed in table S5). The pocket 9 view of the DQ8cis-Ins complex is rotated −60o (top half away from the viewer; bottom half toward the viewer), and the DQ8trans-Ins complex is also rotated −90o, with respect to the x axis. DQ residues are shown in stick form with the same atom color codes for N, O, and S, and carbon is in orange. The p9Arg anchor of InsB12–23 is in ball-and-stick format with carbon atoms in magenta (DQ8cis) or green (DQ8trans). All DQ residues shown in stick form are at a distance <5 Å (any atom to any atom) from the p9Arg anchor residues, with the exception of Tyr32β in DQ8trans that is shown just for comparison. Only the major interactions (highlighted in red in table S5) are shown here. The two major putative hydrogen bonds for the DQ8cis molecule and six major putative hydrogen bonds for the DQ8trans molecule are plotted. The hydrogen bond between Arg76αNη2 and Ser72αOγ of DQ8trans is not plotted because of an overlap with a covalent bond. The line indicating the cation-π interaction between p9ArgCζ and Tyr32βCγ of DQ8cis is only visible close to the latter atom and obscured around the former atom because of the depiction of covalent bonds of p9Arg. All relevant information is provided in table S5.

The presence or absence of Arg52α was also associated with the conformation of some MHC residues (fig. S4B), possibly linked to peptide orientation, that leads to remarkably different modeled interactions with the adjacent, prominent TCR contact residues of the bound epitopes. In the DQ8cis/GAD65251–263 modeled complex, p3Arg interacted via hydrogen bonds with the OH group of Thr61α and the side chain amide group of Asn62α and was away from Asp55α. p5Lys pointed toward the β chain, interacting via a salt bridge with Glu74β. By contrast, in the DQ8trans/GAD65251–263 modeled complex, p3Arg cannot interact with Thr61α and Asn62α as they were distantly located. The terminal guanidine group of p3Arg formed a salt bridge with Asp55α in the DQ8trans complex. p5Lys pointed upward and away from p3Arg, Arg70β, and Glu74β; the latter two residues were closer to each other than in the DQ8cis modeled complex.

Molecular modeling of InsB12–23, which contains the same register as InsB11–24, revealed stronger interactions between the p9Arg anchor of InsB12–23 (InsB22Arg) and surrounding pocket 9 residues of the DQ8trans compared to that of DQ8cis (Fig. 5C), in addition to a reorientation of the N and C termini of the insulin peptide in the groove (Fig. 5A). Besides the interactions observed for both molecules, p9Arg anchor of InsB12–23 formed four hydrogen bonds with Ser72α of DQ8trans (albeit one of them weak), in contrast to only weak van der Waals interactions between the p9Arg anchor of InsB12–23 and Val73α of DQ8cis (table S5). The extra hydrogen bonds between Arg76α and Ser72α (Fig. 5C and table S5) in the DQ8trans/InsB12–23 complex, but not in the DQ8cis/InsB12–23 complex, would further stabilize pocket 9 and potentially contribute to the stability of the DQ8trans/InsB12–23 complex. Thus, for the nonacidic p9 anchor InsB12–23 epitope, the expanded network of residue interaction in pocket 9 might enhance accommodation of epitopes in DQ8trans. Similar to GAD65250–266 and IA-2957–972, substitution of only α44–50 with the respective residues from DQ8cis (A1*0301) causes trivial deviations in the orientation of anchor residues p-2 to p2 and p8 to p10 but hardly any change from anchor residues p3 to p7. Together, the modeling results suggested that Arg52α residue of DQ α chain regulates TCR-pDQ8cis/trans intramolecular interactions through reorientation of bound peptides.


Although it is well established that DQ8/DQ2 cis and trans molecules play a pivotal role in T1D, little is known regarding the cognate islet-epitope repertoire and the interaction between pathogenic T cells and DQ molecules. In contrast to the notion that DQ8trans increases susceptibility to T1D by presentation of a unique T cell epitope repertoire, in this study, we discovered a high degree of epitope overlap and T cell promiscuity between DQ8cis and DQ8trans (Table 1 and Fig. 1), and cross-reactive T cells from homozygous DQ8 participants often recognized the trans-complementary molecule (Fig. 2, A and B). Despite the similarity in epitope recognition and pMHC association (figs. S3B and S4A), increased affinity of TCR-pMHC complexes by distinct amino acid residues at DQ2cis α chain leads to enhanced functional recognition of T cells to DQ8trans in complex with immunodominant and disease-relevant GAD65250–266 (Fig. 4), correlating with the increased risk of T1D of DQ8/DQ2 heterozygosity. Epistatic interaction of HLA genes in other autoimmune diseases includes shared epitope–containing DRB1 alleles in RA (29, 30) and DRB1*1501/DRB5*0101 in MS (31) but with undefined or a small number of T cell specificities. Our study provides a molecular and functional basis whereby disease risk may be altered primarily by increased T cell promiscuity driven by HLA epistatic interaction.

Our data suggest a promiscuous but nondegenerate TCR-peptide-DQ8cis/trans interaction, in which the pMHC/TCR affinity/avidity is governed mainly by the presence or absence of the potential non-TCR contact residue DQ Arg52α (8, 9) within the peptide-binding cleft. T cell cross-reactivity can occur by preservation of structural similarity at specific TCR-pMHC contact sites without the need for TCR degeneracy. In cross-recognition of pDQ8cis and pDQ8trans, the differences in the pDQ8cis/trans α chain are distal to or within the peptide-binding cleft (8, 9) and do not overlap with TCR footprints of canonical docking (9). The increased 2D affinity and functional avidity for DQ8trans/GAD65250–266 (Fig. 4) cannot be attributed to GAD65250–266 affinity for MHC (fig. S4A). Thus, TCR discrimination of DQ8cis/trans-GAD65250–266 complexes must involve differences in peptide or MHCII conformation. Distinct peptide orientation patterns at critical TCR contact residues and distinct conformations of several conserved α chain residues at the TCR/pMHCII interface, due to the presence or absence of DQ Arg52α, were seen in the modeled structures of DQ8cis and DQ8trans in complex with GAD65250–266 (Fig. 5, A and B). DQ α52-mediated peptide reorientation was also observed in the context of two other disease-relevant InsB11–23 and IA-2957–972 epitopes (Fig. 5, A and B), implying a common architecture for TCR recognition of both DQ8cis and DQ8trans-restricted islet epitopes. In contrast to the narrowed TCR usage in cross-reactive, gliadin-specific T cell clones (9, 32), TCR usage in cross-reactive, GAD65250–266-specific T cell clones was quite diverse (fig. S5A). The divergent TCR repertoire suggests an unconfined flexible TCR docking zone enabling the promiscuous TCR discrimination of pDQ8cis/trans complexes.

The T cell clones generated in this study were tetramer-positive and would likely have high affinity for their respective pMHC molecules. In general, we have found tetramers to require 2D affinity values of 10−4 to 10−5 μm4 for their avidity reaction (33). The highly sensitive 2D micropipette assay confirmed their high affinity (10−3 μm4) yet was still able to discern significant affinity differences between the DQ8cis and DQ8trans that directly connect to the functional responses. It should be noted that twofold differences in affinity can lead to major functional and fate decisions including differentiation and tissue trafficking of T follicular helper cells (34). Thus, we demonstrate that not only all clones tested are all of high affinity but also, when comparing DQ8cis to DQ8trans, the pMHC with the highest 2D affinity correlates with the most effective T cell response and subsequent susceptibility to T1D (Figs. 3A and 4A).

T cell clones derived by DQ8cis or DQ8trans tetramer were only activated by pDQ8cis and pDQ8trans but not by pDQ2cis or pDQ2trans (fig. S5B), raising the possibility that the shared β chain is more critical to T cell recognition. The dominance of one MHC chain in T cell recognition has been observed for myelin basic protein peptide–specific T cells, including the highly cross-reactive Hy.1B11 (3537). The few available crystal structures of autoimmune TCRs in complex with their specific pMHCIIs typically exhibit an extremely focused and narrow peptide footprint (3537). However, our T cell reactivity profiles for GAD65250–266, InsB11–23, and IA-2957–972 suggest that specific T cells recognized these epitopes across the entire peptide region (fig. S3B), arguing against the case of unconventional TCR docking for these epitopes.

The enhanced functional recognition via pDQ8trans of immunodominant GAD65250–266- and IA-2957–972–specific T cells at the molecular level (Fig. 3) did not lead to significantly higher frequencies of specific T cells in DQ8+DQ2+ patients, compared to DQ8+DQ2 patients (Fig. 2, C and D), and phenotypically, most GAD65250–266- and IA-2957–972–specific T cells demonstrated a T helper 1–like phenotype (table S3). A number of homeostatic properties could be involved in control of T cell numbers, e.g., proliferation of these disease-related cells in participants with T1D could be influenced by the suppressive immune responses from other T cells, an anergic state of islet-specific memory cells (38) might well keep the majority of cross-reactive cells in quiescence, or differences in surface expression of DQ molecules could impact susceptibility to T1D (39). Thus, at the cellular level, the role of DQ8trans in islet autoimmunity may be an active component of the dynamic temporal nature of the autoimmune process (40, 41).

It has been suggested that the absence of aspartic acid residues at β57 in DQ8/DQ2 cis and trans molecules leads to a highly skewed peptide repertoire favoring a negative charge at the p9 anchor (1012). Our T cell epitope data agreed with these previous reports as more than 90% epitopes identified in this study have a negative charge at the C-terminal end. In contrast, two influenza antigen–derived epitopes with a nonacidic p9 anchor residue (HA186–202 and NP175–191) were identified in this study, and both of them bound to DQ8cis/trans with a moderate affinity (fig. S4A). Recently, an oncogenic self-peptide with an atypical DQ8cis/trans-binding motif was found to be restricted by DQ8cis (42). These observations highlight the presence and importance of atypical DQ8cis/trans-binding epitopes in immune responses to foreign and self-antigens. Therefore, identification of T cells specific for atypical DQ8cis/trans-binding epitopes might further advance the understanding of DQ8cis/trans-specific T cell repertoire and their implication in progression of T1D.

In summary, the promiscuous TCR/pMHCII interactions reported in this study provide functional and structural insights into TCR recognition of T1D–associated DQ8cis/trans complexes. The reported mechanism is distinct from selective presentation of autoimmune peptides or altered T cell repertoires as the primary drivers of autoimmune pathology. The increased risk of T1D in DQ8/DQ2 heterozygous individuals is attributable, in part, to epistatic interactions between HLA alleles that lead to promiscuous TCR recognition of pDQ8cis and pDQ8trans, with amplified functional responses to the latter. The exquisite discrimination of these two molecules in complex with islet-derived epitopes by TCR results from modest differences in the α chain at the N-terminal end of the pocket binding cleft. The mode of TCR discrimination to DQ8cis/trans complexes reported in this study is likely to be shared by other DQ8cis/trans-restricted islet epitopes. Thus, the structural and functional characteristics of genetically linked T cell promiscuity demonstrated in this study may be important for the immunopathology and immunomodulation of T1D. Arg52α is not surface exposed, but its presence may serve as a functional switch while preserving the structural similarity at specific TCR-pMHC contact sites without much change in the pMHC conformation. Modulation of TCR/pDQ8cis/trans affinities via mutagenesis at Arg52α may lead to more effective induction of anergy/apoptosis or conversion of naïve low-avidity autoreactive T cells into memory-like autoregulatory cells, e.g., by gene editing or nanoparticle-based immunotherapy.



These studies were approved by the Institutional Review Board of Benaroya Research Institute (BRI, Seattle, WA) and Seattle Children’s Hospital. All DQ8+DQ2 and DQ8+DQ2+ participants were volunteers of Caucasian descent and were recruited with written consent from participating individuals or assent and the consent of their guardians if participants were minors (tables S2 and S3). HLA typing was conducted by the BRI sequencing and genotyping core facilities. Autoantibodies were measured at the Barbara Davis Center for Childhood Diabetes.

HLA-DQ proteins and tetramers

Recombinant DQ8cis, DQ2cis, DQ8trans, and mutated DQ8cis/trans proteins were produced as previously described (20). Briefly, soluble DQ proteins were purified from insect cell culture supernatants by affinity chromatography. For the preparation of HLA class II tetramers, DQ proteins were in vivo biotinylated in Drosophila S2 cells before harvest and exchanged to citric/phosphate buffer (pH 5.4) (or to phosphate-buffered saline for covalent pDQ8cis or pDQ8trans complexes). The biotinylated monomer was loaded with peptide (0.2 mg/ml) by incubating at 37°C for 72 hours in the presence of n-dodecyl-β-maltoside (0.2 mg/ml) and 1 mM Pefabloc SC (Sigma-Aldrich, St. Louis, MO). Peptide-loaded monomers were subsequently conjugated into tetramers using R-PE (phycoerythrin)–streptavidin (Biosource International, Camarillo, CA) at a molar ratio of 8:1. The quality control of DQ monomers, empty or covalently linked with GAD65250–266, was carried out by staining T cell clones with serially diluted DQ tetramers.

Selection of islet-specific peptides and peptides synthesis

Multiple T1D-associated islet antigens, IA-2, ICA69, Phogrin, and Zn T8, were screened using the DQ8cis binding prediction algorithm in NetMHCIIpan (43) to identify potential DQ8cis/trans-restricted peptides. The top 20 to 25 ranking peptides for each protein were selected for T cell experiments. IGRP-specific epitopes was identified directly by Tetramer Guided Epitope Mapping. Peptides derived from influenza antigens, islet antigens, and a panel of modified GAD65250–266, InsB11–24, and IA-2957–972 peptides with alanine, glycine, or phenylalanine substitution were synthesized by Mimotopes (Clayton, Australia).

Tetramer-guided epitope identification

PBMCs were prepared from the blood of T1D participants or healthy participants by Ficoll underlay. CD4+ T cells were isolated using the Miltenyi CD4+ T cell isolation kit, plated in 48-well plates with adherent antigen-presenting cells (APCs), and stimulated with the selected islet-peptide set or a pool of five consecutive peptides of 17 amino acids in length with an 11-residue overlap spanning the H1HA and H1NP protein sequences. The cells were cultured for 2 weeks in the presence of T cell medium and IL-2 (Roche; added every 24 to 48 hours starting on day 7) at 37°C and stained with PE tetramers. Subsequently, cells were stained with CD4 PerCP (BD Biosciences), CD3 FITC (fluorescein isothiocyanate), and CD25 APC monoclonal antibodies (eBioscience) and analyzed by flow cytometry. Cells from pools that gave positive staining were analyzed again with the corresponding individual peptide tetramers to identify each antigenic peptide. Our criterion for positivity was distinct staining that labeled a compact tetramer-positive CD4+ population and was more than threefold above background (cells from an unstimulated well) in the same experiment.

T cell clone isolation

T cell clones were generated by staining cultured T cells with tetramers, sorting gated tetramer-positive CD4+ cells using FACSAria (at single-cell purity, applying a singlet gate to the lymphocyte population), and expanding in a 96-well plate in the presence of 1.0 × 105 irradiated PBMCs and phytohemagglutinin (2 μg/ml; Remel Inc., Lenexa, KS). After expansion of each T cell clone into a single well of a 48-well plate, clones were stained again with tetramer and also stimulated in parallel with the corresponding peptides (500 nM), adding HLA-DQ–matched irradiated PBMCs as APCs and measuring thymidine uptake to verify epitope specificity.

T cell proliferation, cytokine enzyme-linked immunosorbent assay, and half-maximal effective concentration determination

Proliferation assays were performed in triplicate in 96-well round-bottom plates. T cell clones were stimulated with irrelevant or specific peptides in the presence of irradiated nontransfected or DQ-transfected HEK293 cells as APCs. After 72-hour stimulation, cultures were pulsed with 1 μCi of [3H] thymidine and harvested 18 hours later. SI was calculated by dividing the net counts of specific peptides (counts per minute of DQ8cis- or DQ8trans-stimulated cells subtracted from counts per minute of nontransfected cells) by the counts per minute of irrelevant peptide. For cytokine enzyme-linked immunosorbent assay (ELISA), IFN-γ (clone MD-1), IL-4 (clone 8D4–8), and IL-10 (clone JES3-19F1) capturing antibodies (BioLegend) were coated onto 96-well round-bottom plates. Supernatants (50 μl) from cultures of T cell clones were collected after 48 hours of stimulation and added to each well. After overnight incubation, bound cytokines were detected by biotinylated anti–IFN-γ (clone 4 s.B3), anti–IL-4 (clone MP4-25D2), and anti–IL-10 (clone JES3-12G8) antibodies (BioLegend) and quantified using a Victor2 D time-resolved fluorometer (PerkinElmer). Unless otherwise stated, peptide concentrations were everywhere 2.5 μM for the wild-type insulin peptide and 500 nM for all other peptides. For half-maximal effective concentration (EC50) determination in Fig. 4B, secreted IFN-γ was measured by ELISA after co-incubation of T cell clones with serial-diluted mutated DQ8cis/trans proteins for 48 hours. For EC50 determination in fig. S4A, various concentrations of each biotinylated peptide were incubated in wells coated with DQ8 and DQ8trans proteins. After washing, the remaining biotinylated peptides were labeled using europium-conjugated streptavidin (PerkinElmer) and quantified using the Victor2 D time-resolved fluorometer (PerkinElmer). EC50 values were calculated via a nonlinear regression model in Prism software. All experiments except EC50 determination in fig. S4A were performed in the presence of anti-CD28 antibodies (1 μg/ml).

Statistical analysis

Statistical analysis was performed with Prism software (GraphPad Software Inc.). One-way analysis of variance (ANOVA) followed by Bonferroni’s test was used to perform multigroup comparisons of T cell frequency. An unpaired or paired t test was used to compare the T cell responses between pDQ8cis and pDQ8trans. The extra sum-of-squares F test was used to compare EC50 between different DQ8cis/trans proteins.

Binding motif determination for GAD65250–266, InsB11–24, and IA-2957–972

A panel of 17 GAD65250–266-, 13 InsB11–24-, and 16 IA-2957–972–derived peptides with a single alanine substitution was synthesized to determine the antigenic motif. Glycine or phenylalanine substitution was used if the original residue was an alanine. GAD65250–266, InsB11–24, and IA-2957–972 T cell clones obtained from T1D patients were stimulated with 500 nM (for GAD65250–266 and IA-2957–972) or 2.5 μM (for InsB11–24) of wild-type or modified peptides in the presence of irradiated DQ8cis- or DQ8trans-expressing HEK293 cells as APCs. For GAD65250–266 and IA-2957–972, cultures were pulsed after 72 hours of stimulation with 1 μCi of [3H] thymidine and harvested 18 hours later. For InsB11–24, culture supernatants were collected after 48 hours of stimulation and assayed for IFN-γ secretion by cytokine ELISA. The % activity was calculated by dividing the SI of mutated peptides by the SI of the wild-type peptide. All stimulation experiments were performed in the presence of anti-CD28 antibodies (1 μg/ml).

Ex vivo analysis of islet-specific T cells

Thirty million PBMCs in culture medium at a concentration of 150 million/ml were incubated with 50 nM dasatinib for 10 min at 37°C. The cells were next stained with 250 nM PE-labeled tetramers at room temperature for 120 min, followed by antibody staining for 20 min at 4°C with CD4 APC (clone RPA-T4, eBioscience), CD45RO FITC (clone UCHC1, eBioscience), and a combination of CD14 PerCP (clone MφP9, BD Pharmingen) and CD19 PerCP (clone SJ25C1, BD Pharmingen) to exclude B cells and monocytes from the analysis. Cells were washed twice and incubated with anti-PE magnetic beads (Miltenyi Biotec, Bergisch Gladbach, Germany) at 4°C for another 20 min. The cells were washed again and enriched with a Miltenyi magnetic column. Samples were labeled with ViaProbe (BD Biosciences) and analyzed on a BD FACSCalibur flow cytometer. Frequencies were calculated by dividing the number of tetramer-positive cells in the bound fraction by the number of total CD4+ T cells in the sample. For ex vivo costaining of GAD65250–266-specific cells with DQ8cis and DQ8trans tetramers, PBMCs from DQ8+DQ2 participants were first stained with PE-labeled DQ8trans/GAD65250–266 tetramers. After enrichment, tetramer-positive cells were stained again with APC-labeled DQ8cis/GAD65250–266 tetramers at 37°C for 1 hour.

2D micropipette adhesion frequency assay

The 2D affinity of cultured HLA-DQ CD4+ T cell clones specific for DQ8cis (DQA1*0301/DQB1*0302) and DQ8trans (DQA1*0501/DQB1*0302) were measured using the previously characterized 2D micropipette adhesion frequency assay (33). Briefly, red blood cells (RBCs) were coated with Biotin-LC-NHS (BioVision), followed by streptavidin (Thermo Fisher Scientific) and with either covalently linked, biotinylated DQ8cis/GAD65250–266 or DQ8trans/GAD65250–266 monomers. In this 2D assay, the adhesion frequency between TCR on the T cell and ligand (pMHC on an RBC) aspirated on opposing pipettes was observed using an inverted microscope. An electronically controlled piezoelectric actuator repeated 50 T cell contact and separation cycles with the pMHC-coated RBC while keeping time (t) constant. Following retraction of the T cell, adhesion (binding of TCR:pMHC) was observed as distention of the RBC membrane, allowing for quantification of the adhesion frequency (Pa). Surface pMHC (ml) ligand and TCR (mr) densities were determined by flow cytometry and BD QuantiBRITE PE beads for standardization (BD Biosciences). The calculation of molecules per area was determined by dividing the number of TCRs and pMHCs per cell by the respective surface areas. The relative 2D affinities were calculated using the following equation: AcKa = −ln [1 − Pa (1)]/mrml. Normalized adhesion frequency was calculated using the equation [−ln(1 − Pa(s))/ml (pMHC)]. Geometric means of all measured single-cell affinities are reported ± SEM.

Molecular modeling

Models of DQ8cis and DQ8trans in complex with GAD65251–263, InsB12–23, and IA-2959–971, using the antigenic motifs shown in fig. S3, were prepared on a Silicon Graphics Fuel workstation using the program Discover, version 2005 (Accelrys, San Diego, CA) as previous reported (12, 15, 24). Briefly, the basis molecule for pDQ8cis complexes was the DQ8cis-InsB11–23 complex, without any solvent molecules [(8); pdb code: 1jk8]. To generate pDQ8trans complexes, we superimposed the DQ2cis-α gliadin complex (pdb code: 1s9v; also used free of solvent) onto this DQ8cis-InsB11–23 complex, as previously described (12, 24). The subsequent availability of the crystal structure of a DQ8trans molecule complexed with a gluten peptide and the cognate TCR (9) allowed us to compare the so-obtained model pDQ8trans structure with its crystal counterpart. The root mean square deviation in the alpha carbons (Cαs) of the two structures was very small, both in the histocompatibility molecules and the bound antigenic peptides, confirming the validity of the approach. In the case of molecular simulation of mutated DQ8trans structures containing the Arg52α insertion, the DQ8-InsB11–23 crystal structure above was used as the base molecular complex (8). We assumed that the amino acid deletion in the HLA-DQ2 α chain occurs at position 52, as this results in the least deviation from structural equivalence of neighboring residues among DQ molecules with this deletion and those without [e.g., Arg53α, as evidenced by all the crystal structures of DQ8cis, DQ2cis, and DQ8trans (9)]. Amino acid substitutions both in the antigenic peptide and the αβMHCII molecules were performed manually by selecting from a library of available rotamers of the software to arrive at residues with orientations consistent with the environment in which they were placed. Molecular simulation was performed via energy minimization, as previously described (12, 15, 24), at a physiological pH of 7.4. Minimization proceeded by one thousand steps of the steepest gradient approach, followed by another thousand steps of the conjugated gradient approach, as previously described (12, 15). Figures were drawn with the aid of WebLab Viewer version 3.5 and DS Viewer Pro version 6.0 (both from Accelrys, San Diego, CA). The .pdb coordinates of the pDQ8cis/trans complexes shown and discussed here will be available to interested researchers upon request to A.K.M. or G.K.P.


Supplementary material for this article is available at

Fig. S1. Representative tetramer-guided epitope mapping for H1HA and surface expression of DQ molecules on HEK293 cells.

Fig. S2. Cross-reactivity of islet-specific T cell clones was investigated by staining of DQ8cis and DQ8trans tetramers and a representative ex vivo staining for DQ8trans/GAD65250–266 and DQ8trans/IA-2957–972 from one T1D and one healthy participant.

Fig. S3. DQ8cis and DQ8trans elicit T cell responses by the same antigenic motif.

Fig. S4. Affinity and presentation of shared epitopes.

Fig. S5. TCR usages for GAD65250–266-specific T cell clones and investigation of cross-reactivity of islet-specific T cell clones by different DQ restriction elements.

Table S1. List of common Influenza A epitopes presented by DQ2cis.

Table S2. Summary of ex vivo results for GAD65250–266 and IA-2957–972.

Table S3. List of the T cell clones, their proliferative capacity, and cytokine secretion.

Table S4. Comparison of functional avidity EC50 between wild-type DQ8trans and other DQ8cis/trans variants in complex with GAD65250–266.

Table S5. Predicted interactions between p9Arg/Arg76α with residues at pocket 9 of DQ8cis or DQ8trans.

This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial license, which permits use, distribution, and reproduction in any medium, so long as the resultant use is not for commercial advantage and provided the original work is properly cited.


Acknowledgments: We thank C. Pihoker and research coordinators at Seattle Children’s Hospital and coordinators from BRI Diabetes and Translational Research Programs as well as the BRI clinical core for obtaining and processing samples from participants. We appreciate the assistance from BRI genomic and bioinformatics cores for determination of TCR α chain and β chain usage. Funding: This work was supported by DP3 DK097653 and DK106909 grants from the National Institutes of Health and by the JDRF-2-SRA-2019-678-S-B grant from JDRF. The Silicon Graphics Fuel instrument and the accompanying software were obtained via grant no. MIS 91949 from the Epirus Regional Development Programme to the Technological Educational Institute of Epirus, through the 3rd Community Support Framework of the European Union (80% European Union funds, 20% Hellenic state funds). Author contributions: I-T.C. and W.W.K. designed the research. I-T.C., G.T.N., B.D.E., and W.W.K. wrote the manuscript. W.W.K. performed islet-specific peptide selection. I-T.C., T.J.G., E.M.K., R.J.N., J.W.M., E.A.J., and N.T.-C. performed the research. G.K.P. and A.K.M. performed molecular modeling. C.G. assisted with participant selection and the study design. All authors reviewed the manuscript. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. Additional data related to this paper may be requested from the authors.

Stay Connected to Science Advances

Navigate This Article