The development and application of in silico mapping techniques for the study of carbohydrate-protein interactions

Carbohydrate-protein interactions are exquisitely specific, and underpin many biological and biochemical processes. In particular, carbohydrate-antibody recognition is the initial step in transplant rejection across ABO blood group and species barriers. Such interactions are difficult to structurally characterize, due to the high flexibility of carbohydrates. In silico techniques can be used to fill the gaps in structural knowledge, but must be thoroughly validated prior to use. In this project, novel in silico mapping techniques were developed and validated for investigating carbohydrate-protein recognition, with a focus on carbohydrate-antibody recognition. Anti-αGal antibodies, which recognize the αGal epitope on porcine tissues and are involved in the rejection of pig-to-human xenografts, were used as a test system throughout the project. A selection of docking programs were evaluated for their ability to reproduce the carbohydrate binding modes of a series of high resolution carbohydrate-antibody crystal structure complexes. Of the programs investigated, it was found that Glide performed the best at this task. The results also highlighted that interaction-based approaches could be useful in identifying likely binding modes. A “site mapping” technique was developed, which utilizes information from a given docking ensemble to identify protein residues likely to be involved in ligand recognition. The technique was validated for carbohydrate-antibody recognition using poses generated by Glide. Application of the technique to a panel of anti-αGal antibodies established the structurally conserved nature of carbohydrate recognition by these antibodies. Since site mapping alone cannot be used to infer likely binding modes, ligand-based mapping approaches were developed. The “epitope mapping” technique is used to determine ligand atoms likely to be involved in protein interaction. The “conformation mapping” technique is used to identify likely torsion angles of carbohydrate glycosidic linkages. Both of these techniques utilize the same docking ensemble as the site mapping technique. By combining the output from the three mapping techniques, likely carbohydrate binding modes of the anti-αGal antibodies were determined. These binding modes demonstrated the structural basis of the observed carbohydrate selectivity by two of these antibodies. The site mapping technique was applied to investigate peptide-antibody recognition, highlighting its potential application in the design of carbohydrate-mimetic peptides. The technique was validated using a series of high resolution peptide-antibody crystal structure complexes. The recognition of peptide mimics of the αGal epitope by the anti-αGal antibodies was investigated using the site mapping technique. By comparing the carbohydrate- and peptide-derived site maps for each of the antibodies, it was determined that the peptides largely act as structural mimics of the carbohydrates. Carbohydrate-lectin recognition was also investigated using molecular docking and the site mapping technique. Although molecular docking could usually identify the crystal bound carbohydrate conformation, it was rarely ranked highly. This highlights the need for alternative scoring approaches when studying carbohydrate-lectin recognition using molecular docking. Site mapping was shown to identify lectin residues involved in carbohydrate recognition with improved consistency and accuracy over the top ranked docking pose, and thus, mapping techniques may be useful for these structurally investigating these challenging systems. The in silico mapping techniques developed in this project are likely to be generally useful for studying ligand-protein recognition, as well as being valuable tools for drug design.