An amino acid motif in HLA-DRß1 distinguishes patients with uveitis in juvenile idiopathic arthritis

A.J.W. Haasnoot; M.W. Schilham; S.S.M. Kamphuis; P.C.E. Hissink Muller; A. Heiligenhaus; D. Foell; R.A. Ophoff; T.R.D.J. Radstake; A.I. Den Hollander; T.H.C.M. Reinards; S. Hiddingh; N. Schalij-Delfos; E.P.A.H. Hoppenreijs; M.A.J. van Rossum; C. Wouters; R.K. Saurenmann; N. Wulffraat; R. ten Cate; J.H. de Boer; S.L. Pulit; J.J.W. Kuiper

Proteome data from human iris tissues (2,959 nonredundant proteins) was used as a representative source of proteins present in iris tissue. 52 JIA-uveitis patients commonly have antinuclear antibodies (ANA) directed to iris tissues, thus, we focused on nuclear iris proteins to generate a potentially disease-relevant dataset. Protein accession numbers were extracted to filter in UniProt (Universal Protein Resource) for proteins with a subcellular location annotation limited to the nucleus. 79 We selected 147 proteins (Supplementary Table 4 ) that fulfilled these criteria and their full length amino acid sequences were fed into the neural network of the netMHCIIpan3.1 server. The netMHCIIpan algorithm 38 was selected for its high quality performance in peptide binding prediction for any MHC class II allele for which the sequence is identified, including those for whom only few experimental peptide binding affinities (limited allele-specific training data) are available. We ascertained this by testing the predicted binding affinity of several hundreds of experimentally identified 15-mer peptide ligands eluted from HLA-DRβ1:01, 80,81 which revealed that >95% were correctly picked up by netMHCIIpan as ligands for DRβ1:01 (IC 50 <500nM, data not shown).