145. Using Phage Display in Functional Genomics 

Peter Pavlik, Rob Segal, Daniele Sblattero, Vittorio Verzillo, Roberto Marzari, and Andrew Bradbury 
Los Alamos National Laboratory, Los Alamos, New Mexico and Scuola Internazionale Superiore di Studi Avanzati (SISSA), Trieste, Italy 
bradbury@icgeb.trieste.it 

Phage display offers the possibility of selecting polypeptides (and the genes which encode them) from libraries of 1e10 or more different polypeptides on the basis of their abilities to bind target proteins and subdomains. This diversity far surpasses the estimated number of total genes in the human genome. The application of this technology to the Human Genome Project will powerfully accomplish a central goal: the derivation of ligands that recognize protein products of all human genes, such ligands being either antibodies, or protein fragments. 

Where the recognition ligands derived from this relatively new technology are antibody binding regions (single chain Fv) they can be employed in the same way as traditional antibodies. As such, they can play essential roles in assigning gene function, including the characterization of spatiotemporal patterns of protein expression and the elucidation of protein-protein interactions. Where the recognition ligands are protein fragments, they can be considered to be potential protein-interaction partners for the immobilized polypeptide and so a starting point for further biochemical studies. This project has concentrated on trying to find a general way to isolate antibodies against gene products, preferably starting from gene sequence and using peptides to avoid the need for cloning and expression. A new method to make phage antibody libraries has been developed and a new large library using this method is presently under construction. A library provided by Jim Marks, UCSF, has been used to select antibodies against a number of cell cycle and DNA repair proteins. We have succeeded in miniaturising selection on proteins to a 96 pin format. Should gene products be available this is a very efficient way to select antibodies in a high throughput format. We have also used scanned peptides (180 in total) derived from five different proteins (ubiquitin, cdk2, human serum albumin, cyclin D, transglutaminase) to select antibodies. Some of the antibodies selected are able to recognise the native protein. We are attempting to derive rules, based on physicochemical characteristics and other predictive algorithms, which predict which peptide sequences will select antibodies recognising the full length protein.