HLA Class II molecules on haplotypes associated with type 1 diabetes exhibit similar patterns of binding affinities for coxsackievirus P2C peptides

Abstract

Enteroviruses such as coxsackievirus B4 (CVB4) are proposed as possible environmental triggers or accelerants of the autoimmune process that leads to type 1 diabetes mellitus. One putative mechanism to account for this association is mimicry between virus components and islet autoantigens. Particular interest has focused on the CVB4 non-structural protein P2C, which we previously showed to be a major target of the effector memory anti-CVB4 CD4 T-cell response, and which harbours a region of sequence similarity with the islet autoantigen, glutamic acid decarboxylase (GAD65). Since several distinct human leucocyte antigen (HLA) Class II molecules are associated with development of type 1 diabetes, we hypothesized that for functional mimicry to be important, any potential region(s) of mimicry in P2C should bind to each of these susceptibility molecules. In the present study therefore we examined the affinity of 20-mer overlapping P2C peptides for soluble HLA-DR4, -DR3, -DQ2 and -DQ8. We identified one discrete region of P2C with high binding affinities for all of these HLA Class II molecules. Moreover, the binding affinity of P2C peptides was significantly correlated between HLA molecules present on the same susceptibility haplotype (e.g. DR4 and DQ8, P =0.0076; DR3 and DQ2 P = 0.002). We conclude that possession of these haplotypes favours restricted presentation of viral epitopes, and speculate that this could promote the potential for mimicry between microbial proteins and islet autoantigens.

Figure 1

Examples of peptide-HLA binding curves, IC₅₀ calculation and assay reproducibility. Two HLA-DR4 binding assays performed 2 years apart are shown. (a) In the first assay, the solid curve (closed boxes) represents competition of increasing concentrations of non-biotinylated peptide (CLIP) for binding of the biotinylated reporter CLIP (bio-CLIP) peptide used at a fixed concentration of 2·5 μm. The dashed curve (open boxes) represents binding of a high-affinity P2C peptide, competing for binding with bio-CLIP (2·5 μm). In (b) an identical assay is performed 2 years later. Calculated IC₅₀ values are similar in both assays, indicating excellent assay reproducibility characteristics. In (c), the effect of DMSO on enhancing signal to noise ratio is shown by increased concentration of DMSO in the reaction mixture: (○) 0% (×) 10% (▵) 15% and (□) 20% (v/v). Error bars represent SEM from three separate assays.

Figure 2

Overlapping peptide-binding maps for HLA Class II molecules associated with type 1 diabetes. For each HLA-molecule tested, the 10 peptides with highest binding affinities (as identified in competitive inhibition binding assays) are indicated in a map of all overlapping peptides from the 329-amino-acid P2C molecule (represented by boxes numbered 1–40). The five highest affinity peptides are shown by black shading, and next five highest by grey shading. At the bottom of the figure is the linear sequence of P2C, with the position of the PEVKEK motif indicated.

Figure 3

Correlation of binding affinities for different HLA Class II molecules of the P2C overlapping peptide set. Non-parametric (Spearman ranking) correlation (n = 40 for all data sets, except for plots with correlation for HLA-DQ8 where n = 34) is shown. Significant correlations are seen for peptide-binding affinities for HLA II molecules encoded on the same haplotype (i.e. between HLA-DR3 and -DQ2, and between -DR4 and -DQ8). In addition, there is a strong correlation between binding affinities for the two HLA-DQ molecules. A weaker, but significant correlation is seen for binding to HLA-DQ8 and -DR3. In contrast, no significant correlation in binding affinities was observed between HLA-DR3 and -DR4, or between -DR4 and -DQ2.

Figure 4

Receiver-operator characteristic (ROC) plots to determine the optimal predictive value. ROC were plotted to assess the ability of different computer algorithms to predict HLA molecule binding (maximizing specificity and minimizing false predictions). (a) shows ROC plot for HLA-DR4 and (b) for HLA-DR3. The programs used were syfpeithi (SYF), rankpep (RP) and propred (PP), and combinations of each as shown. The upper-left corner of each graph represents the point of highest specificity and sensitivity. High scoring sequences by at least two of the three programs (SYF/RP/PP) match high-affinity peptides with high specificity and sensitivity.