In silico Prediction and Experimental Confirmation of B-cell Epitopes for Molecular Serology of Chlamydia spp.

Abstract

Urgently needed species-specific ELISAs for detection of antibodies against Chlamydia (C.) spp. have been elusive due to high cross-reactivity of chlamydial antigens. To identify Chlamydia species-specific B-cell epitopes, we ranked potential epitopes of immunodominant chlamydial proteins that are polymorphic among all Chlamydia species. High-scoring peptides were synthesized with N-terminal biotin followed by a serine-glycine-serine-glycine spacer, immobilized onto streptavidin-coated microtiter plates, and tested with mono-specific mouse hyperimmune sera against each Chlamydia species in chemiluminescent ELISAs. For each of nine Chlamydia species, 3-9 dominant polymorphic B-cell epitope regions were identified on OmpA, CT618, PmpD, IncA, CT529, CT442, IncG, Omp2, TarP, and IncE proteins. Sixteen-40 amino acid-long peptides of these epitopes reacted highly and with absolute specificity only with homologous, Chlamydia mono-species-specific sera. The probability of cross-reactivity of closely related peptide antigens correlated with percent sequence identity, and declined to zero at less than 50% sequence identity. In the course of this investigation, a failure to accurately predict B-cell epitopes was observed for such prediction algorithms. Using our database of confirmed chlamydial B-cell epitopes and non-epitopes, we sought to understand the reasons for this failure. In our investigation, short 7-12aa peptides of B-cell epitopes bound antibodies poorly, thus epitope mapping with short peptide antigens would have falsely classified many of these epitopes as non-epitopes. We also show in published datasets of confirmed epitopes and non-epitopes a direct correlation between length of peptide antigens and antibody binding. Elimination of short, ≤11aa epitope/non-epitope sequences improved public datasets for evaluation of in silico B-cell epitope prediction. Following evaluation of a comprehensive set of algorithms for prediction of protein properties, we show that protein disorder tendency best describes B-cell epitopes. For B-cell epitope identification from a protein with 86% accuracy, we recommend using the 25aa moving average plot of the IUPred-L disorder score, and selecting 16-30aa peptides of peak regions for laboratory testing.

In conclusion, we have developed an accurate approach for B-cell epitope prediction and have applied it to identification of highly specific peptide antigens for molecular serology of Chlamydia spp. The combined approach also lends itself to identification of relevant epitopes of other microbial pathogens.