Modulatory Role of DR4- to DQ8-restricted CD4 T-Cell Responses and Type 1 Diabetes Susceptibility

HLA-DQ8, -DR4, and -DQ8/DR4 transgenic mice devoid of murine endogenous MHC class II were generated and fully characterized by previous studies (17,24). The animals were bred and housed under pathogen-free conditions in the animal facility of the Rangos Research Center at the Children’s Hospital of Pittsburgh. The animal experiments were conducted in compliance with the animal research care and committee of the Children’s Hospital of Pittsburgh.

Peptides hGAD65_206-220 (TYEIAPVFVLLEYVT), hGAD65_209-217 (IAPVFVLLE), hGAD65_536-550 (RMMEYGTTMVSYQPL), and hGAD65_554-566 (VNFFRMVISNPAA) were synthesized by Chemicon (Temecula, CA) (21,25).

Human B-LCLs BM92 and KT17 were obtained from International Histocompatibility Working Group (Seattle, WA). WT51 and FS were maintained at our laboratory. The DQ-DRB haplotypes of these B-LCLs are shown in Table 1.

Fluorescein isothiocyanate (FITC)-conjugated anti-mouse T-cell receptor (TCR) Vβ panel, anti–HLA-DQ (Leu-10), anti-mouse CD86 (GL1), phycoethrin-conjugated anti-mouse CD3 (17A2), anti-mouse CD4 (GK1.5), anti–HLA-DR (L243), anti-mouse CD80 (16-10A1), APC-conjugated anti-mouse CD45R/B220 (RA3-6B2), purified anti-mouse CD16/CD32 (mouse Fc block) (2.4G2), and isotype control antibodies were purchased from BD Pharmingen (San Diego, CA). For a typical surface staining, 5 × 10⁵ cells were stained with fluorescence-conjugated antibodies. The data were collected using a BD FACS Calibur flow cytometer (Becton Dickinson, Palo Alto, CA) and analyzed by CellQuest software (Becton Dickinson).

Primary T-cell lines were generated according to previous protocols (26–28). Briefly, T-cells from peptide-primed HLA transgenic mice were maintained in a biweekly restimulation culture containing 10 μg/ml cognate peptide, 2.5–5 units/ml mouse recombinant interleukin (IL)-2 (Roche, Indianapolis, IN), and irradiated (25 Gy) DQ8⁺ or DR4⁺ splenocytes. After 6–8 rounds of restimulation, each T-cell culture was characterized for TCR Vβ usage and antigenic specificity. T-hybridomas were generated by fusing CD4 T-cells from day-4 restimulation cultures with BW5147α⁻β⁻ lymphoma (provided by Dr. Willi Born from University of Colorado Health Science Center) in the presence of PEG-1500 (Sigma, St. Louis, MO). Successfully fused T-hybridomas were selected by hypoxanthine/aminopterin/thymidine supplement (Invitrogen, Carlsbad, CA) and recloned by limiting dilution according to a standard protocol (29).

Total RNA was prepared from 2 × 10⁶ T-cells with TRIzol Reagent (Invitrogen). Total RNA (1.5 μg) was used for cDNA preparation with the SuperScript II first-strand cDNA synthesis system (Invitrogen). PCR was performed with Advantage cDNA polymerase (Clontech, Palo Alto, CA). The PCR product was cloned into pDrive vector (Qiagene, Valencia, CA) and sequenced using M13/M13Reverse primers by 3130 Genetic Analyzer (Applied Biosystems, Fost City, CA). The sequences were analyzed using Sequencher software (version 4.5; Gene Codes, Ann Arbor, MI).

T-cells (20,000 cells/well) from 2-week restimulation cultures were cocultured (in triplicate) with irradiated HLA transgenic mouse splenocytes (500,000 cells/well) and peptides in round-bottom 96-well plates. In some experiments, MACS microbead-conjugated anti-CD90 monoclonal antibody and MACS LS column were used to deplete T-cell subsets from spleen cells before performing the assay (Miltenyi, Auburn, CA). Supernates were harvested to determine the production of γ-interferon (IFN-γ) (72 h). For the T-hybridoma assay using human B-cell lines as APC, B-cells were inactivated by 50 μg/ml of Mitomycin C (Sigma) treatment at 37°C for 20 min. After extensive washing, B-cells (40,000 cells/well) were cocultured with T-hybridomas (20,000 cells/well) in the presence of peptide. Supernates were harvested 24 h later for IL-2 detection. Enzyme-linked immunosorbent assay (ELISA) was applied to determine the cytokine concentration with a Europium-Streptavidin detection system (Perkin Elmer). Antibodies and recombinant protein standards for IL-2 and IFN-γ ELISA application were purchased from BD Pharmingen.

Cytokine ELISA results were analyzed using Prism 4 software (GraphPad Software, San Diego, CA). Differences in T-cell responses elicited by DQ8 versus DQ8/DR4 transgenic splenocytes were assessed with the two-tailed unpaired t test. The results of T-hybridoma responses against a panel of B-LCLs were analyzed by one-way ANOVA. A P value <0.05 was considered statistically significant.

HLA-DQ8 and -DR4 (0401) transgenic mice were immunized with hGAD65_206-220 and hGAD65_536-550. One week later, the lymphocytes from draining lymph nodes and spleens of immunized mice were harvested and in vitro T-cell cultures established. Four specific CD4 T-cell lines were obtained. T206 was generated from HLA-DQ8 transgenic mice and specifically responded to DQ8⁺ splenocytes by secreting IFN-γ in the presence of 10 μg/ml cognate peptide hGAD65_206-220 (Fig. 1A). T536.1 was obtained from HLA-DQ8 transgenic mice primed with hGAD65_536-550. On cognate peptide (10 μg/ml) hGAD65_536-550 recall stimulation, T536.1 produced IFN-γ (Fig. 1B). Neither T206 nor T536.1 had cross-reactivity to DR4. However, they responded to the splenocytes from HLA-DQ8/DR4 double-transgenic mice. Another two CD4 T-cell lines, DR4p206 and DR4p536, were obtained from hGAD65_206-220–and hGAD65_536-550–primed HLA-DR4 transgenic mice, respectively. Both secreted IFN-γ in response to cognate peptide (10 μg/ml) pulsed DR4⁺ or DQ8⁺DR4⁺ APC (Fig. 1C and D). The data of T206 versus DR4p206 and T536.1 versus DR4p536 unambiguously demonstrated that both hGAD65_206-220 and hGAD65_536-550 could be bound by HLA-DQ8 and -DR4, yielding credence to the proposition of peptide competition.

We planned to use T206 and T536.1 as reagents to examine peptide presentation efficiency by DQ8, since IFN-γ production was correlated with the concentration of cognate peptide (data not shown). We therefore analyzed the TCR β-chain expression to examine the clonality of these DQ8-restricted T-cell lines. As revealed by flow cytometric analysis, T206 exclusively expressed Vβ6 TCR (Fig. 1A). T536.1 exclusively expressed Vβ1 TCR, which was identified by RT-PCR using a panel of TCR Vβ-specific primers (this panel consists of 24 specific Vβ elements, including some rare Vβ genes like Vβ1 that does not have a corresponding monoclonal antibody) (30). To further confirm that these lines exclusively expressed a single Vβ element, the β-chain cDNAs of T206 and T536.1 were also cloned. Sequencing of 20 colonies confirmed the clonal phenotype of each cell line (GeneBank accession no. DQ421779 and DQ421780). In summary, these T-cell lines were assessed to manifest clonal-like phenotype and were stable in culture after multiple restimulation. Thus, they were considered to be reliable reagents to evaluate peptide presentation efficiency. The DR4-restricted CD4 T-cell lines DR4p206 and DR4p546 were only used to demonstrate that the same epitopes could also be presented by DR4, a requisite condition for peptide competition between DR4 and DQ8. The clonality of these lines was not critical to our main purpose of evaluating DQ8 peptide presentation efficiency.

Since DQ8 and DR4 could bind the same peptide, they would potentially compete for the peptide when they were coexpressed on the same APC. To test this hypothesis, we evaluated the relative peptide occupancy of DQ8 by comparing the CD4 T-cell responses elicited by DQ8⁺ APC (in the absence of DR4 expression) with DQ8⁺DR4⁺ APC. Splenocytes from DQ8 and DQ8/DR4 transgenic mice that expressed a similar level of DQ8 were used as APC (Fig. 2A and B). In addition, no difference in expression of CD80/CD86 costimulatory molecules was observed (data not shown). T-cells were depleted from the splenocytes before assaying; thus, the activities of radiation-resistant T_reg-cells in the crude APC population were excluded. In the presence of DR4, IFN-γ secreted by T206 or T536.1 was decreased under different Ag concentrations (Fig. 2C and D). It indicated that DR4 coexpression diminished DQ8-restricted T-cell responses.

We also examined the T-cell responses elicited by DQ8⁺DR4⁺ mouse splenocytes and mixed APC composed of DQ8⁺ and DR4⁺ splenocytes in the presence of cognate peptide (2 μg/ml). IFN-γ production by T206 (Fig. 3B) and T536.1 (Fig. 3C) indicated that the presence of DR4 on separate APC only slightly reduced DQ8-restricted T-cell responses (∼18% reduction for T206 and 3% reduction for T536.1), whereas the coexpression of DR4 and DQ8 on the same APC resulted in a profound decrease of T206 and T536.1 responses (72 and 47% reduction, respectively). Analysis by flow cytometry demonstrated that cell surface expression of DQ8 and DR4 was similar on the splenocytes used in this experiment (Fig. 3A). Taken together, these results show that DR4 modulates DQ8-restricted T-cell responses effectively only when it was coexpressed with DQ8 by the same APC.

We next examined the effects of other DR4 subtypes (DRB1*0401, 0402, 0403, 0404, and 0406) on DQ8-restricted CD4 T-cell responses. It is known that DQ8-DRB1*04 haplotypes carrying specific DRB1*04 alleles are associated with variable susceptibility to type 1 diabetes (15,16). The sequence alignment indicated that these DRB1*04 alleles differred from each other mainly at the β86, β74, β71, β70, and β57 positions (Fig. 4A). These residues are responsible for the size and polarity of the P1, P4, P6, and P9 pockets (31). This implies that distinct DR4 subtypes bind peptides with different affinities and manifest variable peptide competition potentials.

Since transgenic mice expressing other DRB1*04 alleles have not been developed, the most available sources of antigen-presenting cells are Epstein-Barr virus–transformed human B-cells. B-LCLs, including WT51, FS, BM92, and KT17, homozygously express identical DQ8 and DRB4 but different DRB1*04 alleles (Table 1). As the only exception, KT17 heterozygously expresses DRB1*0403 and 0406 at the DRB1 locus. However, since the amino acid sequence within the DR peptide–binding groove is identical between 0403 and 0406, it is presumed that DRA1*0101/DRB1*0403 and DRA1*0101/DRB1*0406 heterodimers manifest an identical affinity for peptide binding. Considering human B-LCLs that did not express murine costimulatory molecules, we used T-hybridomas TH206 and TH536.1 instead of primary T-cell lines as detection reagents, since T-hybridoma responses were less dependent on costimulation than primary T-cells (32). These T-hybridomas were generated by fusing primary T-cell lines, T206 and T536.1, with CD3/TCR-deficient BW5147α⁻β⁻ lymphoma (33). They expressed identical TCRs as their parental T-cell clones and responded to their cognate peptide recall by secreting IL-2. TH206, however, responded to a short truncated peptide hGAD65_209-217 better than hGAD65_206-220 (data not shown).

All DQ8⁺ human B-cell lines were able to stimulate TH206 and TH536.1 responses (Fig. 4B and C). However, the magnitude of hybridoma responses elicited by human B-LCLs was variable and ranked from high to low in the order of FS > WT51 > BM9 > KT17. This was not simply due to the level of DQ8 expression because the weak T-cell responses induced by high-DQ–expressing KT17 and BM92 were in contrast to the strong T-cell responses triggered by relatively low DQ expression of WT51 and FS (Table 2). Although the B-LCLs utilized for this purpose may display variable amounts of DQ and DR molecules so that the relative amount of DR4 may also be involved, we found that the results obtained were suggestive of a consistently decreasing peptide presenting efficiency of DQ8 in the presence of distinct DRB1*04 alleles. The DRB1*0403/0406⁺ KT17 always required 5- to 10-fold more peptide to achieve an equivalent T-cell response in comparison with DRB1*0402⁺ FS and DRB1*0401⁺ WT5. These results are consistent with the interpretation that DRB1*0403/0406 and DRB1*0404 alleles may have a stronger competition for these peptides than DRB1*0402 or DRB1*0401.

To confirm the effect of DR4 on DQ8-restricted T-epitope presentation, we thus investigated the hybridoma responses in the presence of a blocking peptide, hGAD65_554-566, which was previously shown to bind DR4 (DRA1*0101/DRB1*0401) with a high affinity but only weakly to DQ8 (21). DQ8-DRB1*0401–expressing WT51 was used as APC. Results showed that responses of TH206 and TH536.1 were increased in the presence of as low as 2 μg/ml peptide blockade (Fig. 5A and B). When the blocking peptide was increased, T-hybridoma responses were increased accordingly. This dose-dependent increase of the T-hybridoma response indicated that the reduced response without blocking peptide was related to the availability of the DR4 peptide–binding groove.

The rationale of studying DR4 peptide competition is to have a better understanding of DQ8, DR4, and autoimmune diabetes. Coexpression of specific DRB1*04 alleles with type 1 diabetes–suspceptible DQ8 in humans modulates the risk of type 1 diabetes development, while the mechanisms have not been fully unveiled. Our study compared the DQ8-restricted T-cell responses elicited by different DQ8⁺ APC (HLA transgenic mice splenocytes and human B-cells). The major variances were DR4 alleles. The diminished Th1 responses (Fig. 2C and D) not only demonstrated the reduced DQ8 peptide presentation in the presence of DRB1*0401 but also implicated more profound biological significance regarding the role of Th1 cells in the progression of diabetes. It has been well acknowledged that IFN-γ has multiple impacts on the process of β-cell destruction, such as recruiting pathogenic CD4/CD8 T-cells to the islets and sensitizing β-cells via upregulating antigen processing/presentation machinery (34,35). Therefore, the decreased IFN-γ production from Th1 cells due to peptide competition partially contributes to the reduced CD4/CD8 T-cell infiltration observed in the islets of HLA-DQ8/DR4 transgenic mice compared with HLA-DQ8 transgenic mice (both HLA transgenic lines are in the mII⁻/RIP-B7.1-C57BL/6 background) (17). By examining T-hybridoma responses triggered by human B-LCLs, we tentatively concluded that specific DRB1*04 alleles were associated with variable competition potential in the order of DRB1*0402 < DRB1*0401 < DRB1*0404 < DRB1*0403/0406. This order is consistent with the degrees of protection conferred by these DRB1*04 alleles. It has been broadly discussed that the variable degrees of protection are correlated with the polymorphisms in the DR peptide binding site (36). However, no evidence has been shown that the protective DRB1*0403 is more capable of deleting autoreactive CD4 T-cells or selecting regulatory T-cells in the thymus than susceptible DRB1*0401 alleles. This lack of evidence thus favors peripheral tolerance mechanisms such as the peptide competition model. By using our data, we now suggest that the DQ8-DRB1*0403 haplotype is more protective than the DQ8-DRB1*0401 haplotype because the DRB1*0403 allele can compete for diabetogenic peptides better than the DRB1*0401 allele. We are further exploring the effect of protective DRB1*04 alleles in comparison with other susceptible alleles in alternative systems because the variable expression levels of DQ8 and DR4 in those B-LCLs leave the question that the weak T-hybridoma response observed in KT17 is not only due to the competition by DRB1*0403/0406 alleles for the antigens but also to higher relative abundance of DR over DQ molecules per cell.

The results that compared T-cell responses elicited by DQ8⁺DR4⁺ APC and DQ8⁺ APC mixed with DR4⁺ APC showed that the variation of T-cell activity could not be due to the binding by DR4 on the APC surface. These data actually helped us better understand to what extent extra- and intracellular peptide competition contributes to the decreased T-cell responses. Extracellular peptide loading occurs by peptide replacement or direct loading into an empty peptide-binding groove. Intracellular loading, mediated by endocytosis and H2-M (or HLA-DM in human) (37–39), is not only applied to whole protein antigen but also to short peptides (40). Whether the extra- or intracellular process prevails may be related to the specific peptide (40). In our cases, DR4 can certainly compete for both extracellular peptides and engulfed peptides since it is intrinsically able to bind the peptides (Fig. 1). However, the competition for the extracellular peptides is insufficient to affect DQ8-restricted T-cell responses dramatically. Because excessive amounts of peptide were added into the medium to ensure a detectable T-cell response, the amount of peptide “stolen” by DR4 on the cell surface is probably insignificant compared with what is left in the medium. This notion was confirmed by the results that DR4 expressed on separate APCs was less capable of diminishing DQ8-restricted T-cell responses (Fig. 3). In contrast, coexpression of DR4 with DQ8 on the same APC had a very different effect (Fig. 3). These results indicated that the surface peptide loading of DR4 (at about pH 7) does not contribute to the reduction of DQ8-restricted T-cell responses. It thus suggested that the diminished T-cell responses due to DR4 coexpression was mainly caused by peptide competition within intracellular compartments where the pH environment was low (about pH 5), and the quantity of peptide was limited by the small volume of internalized medium.

An alternative explanation for the diminished DQ8-restricted T-cell responses is that the presence of DR4 interferes with the mobility of peptide-DQ8 complexes to form immunological synapses. This explanation also requires the coexpression of HLA-DQ8 and -DR4 on the same cell surface. However, the data from our DR4 blockade experiment revealed that the availability of DR4 binding site, instead of DR4 itself, was relevant to the variation of T-cell responses (Fig. 5). Therefore, our study does not support this possibility.

It is noteworthy that our results were obtained from in vitro T-cell assays using activated/effector T-cells to evaluate peptide occupancy. The decreased T-cell responses caused by peptide competition implies a more profound biological significance under in vivo situations, where the autoantigen resource is limited, the frequency of autoreactive T-cells is low, and most T-cells express naïve phenotypes. Mature/effector T-cells and naïve/immature T-cells have been known to be different regarding the requirements for activation. For the naïve T-cells, the quantity of TCR-peptide-DQ8 engagement had to reach a threshold (∼8,000 TCR/cell) to obtain activation (41,42); otherwise, these autoreactive T-cells are “silenced” or anergic. Although costimulation lowers the threshold (∼2,000 TCR/cell), the number of available self-peptide–DQ8 complexes on the APC is still a crucial parameter to direct naïve autoreactive T-cells to differentiate into harmful effector Th1 cells or to become nonresponsive. It is more critical for the cells that naturally have lower TCR surface density. They might require an unattainably high concentration of antigen to engage enough TCRs for a successful activation. Indeed, T-cells that escaped thymic selection with potentially autoreactive TCRs often intrinsically had a lower TCR density on their surface (43).

Many previous studies have addressed the peptide competition issue in different ways. The evidence of peptide elution studies suggested that there was no competition between different MHC alleles (44). However, those peptides eluted from different MHC molecules represent the most abundant self-proteins and may not be diabetogenic. In addition, since only strongly bound peptides can be copurified with MHC before analysis, the pools of eluted peptide actually exclude the weakly bound peptides that are more likely the real candidates of type 1 diabetes–relevant peptide competition. Examining the affinity between synthesized peptides and purified MHC class II molecules is another useful approach to provide chemical evidence to show that different MHC molecules can bind to the same peptide (21). However, different detergent concentrations and reference peptides were used to stabilize DQ8 and DR4 heterodimers in the solution. The approach of directly comparing binding affinity is hence less capable of revealing the outcome of competition.

Our observations provided in vitro evidence for peptide competition as an explanation of the modulating role of DR4 in the T-cell response to diabetes-related autoantigens. Our T-cell response data are consistent to the hierarchical association of different DQ8-DRB1*04 haplotypes with type 1 diabetes susceptibility and bridges protein structure studies with genetic analysis. It thus refined our understanding toward the role of DR4 to the risk of type 1 diabetes in the context of DQ8-conferred primary susceptibility. There is no doubt that DR4 expression directed positive selection for TCRs with novel specificities (45). Some of those might potentially recognized islet antigen–derived peptides. However, this experimental model did not address whether and how these DR4-restricted T-cells contribute to the progression of diabetes.

This work was supported in part by the National Institutes of Health Grants AIO56374 and RO1 DK024021 (to M.T.).

We are grateful to Dr. Willi Born for the generous gift of the BW5147α⁻β⁻ T lymphoma.