Design, synthesis and evaluation of high-affinity binders for the celiac disease associated HLA-DQ2 molecule

Celiac disease is caused by uncontrolled CD4 T-cell responses directed to wheat-derived gluten peptides bound to the disease predisposing HLA-DQ molecules. The only available treatment is a life-long gluten-free diet which is complicated by the widespread use of wheat-derived gluten in the food industry. As the binding of gluten-derived peptides is a prerequisite for the induction of the inflammatory T-cell response, blockers that would prevent gluten peptide binding to the HLA-DQ molecules might be used as an alternative to the gluten-free diet. In the present study we have analyzed the binding properties of a set of previously identified natural ligands for HLA-DQ2, the primary disease predisposing allele. An in silico method, Epibase, ranked these peptides and the top one, a peptide with a nine amino acid core FVAEYEPVL, was measured among these peptides as the peptide with the highest binding affinity for HLA-DQ2. In a stepwise approach we subsequently tested the impact of N-terminal extensions and systematic single amino acid substitutions within the core of this peptide which revealed that an N-terminal extension with the tripeptide sequence ADA increased binding affinity 5- to 6-fold. In addition the substitution analysis indicated which amino acids were most preferred at anchor residues in the lead peptide, generally leading to an increase of binding affinity with a factor of 2. Next we tested which combinations of such preferred amino acids yielded the best results. The combined results indicate that a peptide with sequence ADAYDYESEELFAA (core in bold) had superior binding properties. This peptide was chosen as a lead peptide for further optimization with non-natural amino acids at the p1 position, since molecular modeling indicated that none of the natural amino acids is able to optimally occupy the p1 pocket. A set of 8 non-proteinogenic amino acids was designed, synthesized and incorporated in the lead peptide (and in two control peptides) and tested for binding to HLA-DQ2. The results indicate that the effect of the incorporation of these non-proteinogenic amino acids depended on the peptide in which they were incorporated and that the maximum increase in binding affinity obtained was approximately 2-fold. Altogether lead sequences were obtained that have a binding affinity for HLA-DQ2 that is 100- to 200-fold higher compared to that of the gluten-derived peptide that has the highest affinity for HLA-DQ2. Such peptides are candidate lead peptides for further optimization. Our results, however, also indicate that in order to obtain further significant increases in binding affinity alternative approaches will have to be explored.