Model systems to study the parameters determining the success of phage antibody selections on complex antigens

Phage antibody display technology offers a powerful tool foe the isolation of specific antibodies to defined target antigens. Most selection strategie s described to date have relied on the availability of purified and often r ecombinant antigen, providing the possibility tb perform selections on a we ll-defined antigen source. However, when the target antigen cannot be purif ied (e.g., an integral membrane protein), or if the antigen is unknown (e.g ., when searching for novel markers on cells or tissues), panning of phage antibody libraries has to be performed on complex antigen sources such as c ell surfaces or tissue sections, or even by in vivo selection methods. This provides a series of technical and experimental challenges. One focus of o ur research is to select antibodies directed to novel cancer-induced antige ns expressed by tumours and by the tumour vasculature. To understand the pa rameters governing selection on complex antigen sources and to assess the e fficiency of these phage library selections, we have set up two model selec tion systems in which, both tumour cells and vascular endothelial cells ser ve as target "antigen". We describe a model based on phage antibodies direc ted to the tumour antigen epithelial glycoprotein-2, to compare phage antib ody selections on a range of different antigen sources including purified a nd recombinant antigen, whole live cells, tissue cryosections and in vivo g rown solid tumours. Secondly, we describe a model based on a phage antibody directed against the endothelial cell inducible adhesion molecule E-select in. We compare selections on cultured cell monolayers with selections on ce ll suspensions immobilised on columns, to determine which selection approac h is most suitable for the identification of novel tumour endothelial cell markers. Our data provide insight into the efficiency and thus potency of d ifferent selection strategies and show that there are very large difference s in the recovery and enrichment of binding phage between the different met hods tested. Our results further demonstrate the feasibility of phage antib ody selections on whole, intact cells and show that these may sometimes com pare favourably to selections on purified antigen. Selections on endothelia l cells immobilised on columns compare favourably with selections on cell-m onolayers; the most favourable conditions for both selection procedures are described. The implications of our data for phage antibody selections on t hese different complex antigen sources using either non-immune or immune ph age antibody repertoires are discussed. The use of model systems such as th e ones described here will help to determine optimal experimental condition s for phage library selections on complex antigens and aid in developing mo re powerful selection procedures for target discovery. (C) 1999 Elsevier Sc ience B.V. All rights reserved.