Leveraging synthetic phage-antibody libraries for panning on the mammalian cell surface

The vast majority of monoclonal antibodies available in the market as research reagents, diagnostic tools, or therapeutic antibodies have been produced using hybridoma technology, which has changed little since its invention in the 1970s (Kohler and Milstein 1975). This statement is particularly true for cell surface and secreted proteins. Although hybridoma technology is used to generate most monoclonal antibodies for biological research, the technology suffers from several fundamental drawbacks that have limited discovery of functional antibodies for cell surface and secreted proteins. For example, the need for animal immunization means that selections occur in an uncontrolled environment, which may be suitable for stable antigens but not for sensitive antigens, including membrane proteins and conformational epitopes. The animal strongly inuences the antibody repertoire that results from introducing a foreign antigen, so there is no guarantee that immunodominant epitopes will be those of interest. By extension, the mouse immune repertoire is restricted to eliminate antibodies that would recognize mouse proteins. Therefore, it can be a challenge to raise effective antibodies against conserved epitopes in other species, which is highly problematic since many functionally important epitopes, including protein-protein interaction sites, are highly conserved. Moreover, hybridomas provide antibodies and not the encoding DNA. This makes it difcult to alter or improve an antibody of interest without complicated procedures that convert the antibody into a recombinant form that can be genetically engineered. Last, hybridoma technology requires the use of animals, which is cumbersome and expensive. The limits of hybridoma technology have become rate limiting in the postgenomics era as antibodies grow in number and potential as therapeutic agents. More efcient and sophisticated panning procedures using human antibody frameworks that are displayed as libraries on model display systems should help to relieve this bottleneck.