Directed evolution and optimization of antibodies for enhanced affinity and stability

Abstract

Antibodies have emerged as the dominant class of diagnostic and therapeutic agents in today’s biotech industry due to their high binding affinity, specificity, stability and manufacturability. In vitro antibody selection using synthetic libraries offers many advantages over in vivo immunization, including higher antibody affinity and greater control over the antibody selection process. However, antibodies generated in vitro often display trade-offs between increased affinity and reduced stability and/or specificity. We have made multiple key discoveries to overcome some of these shortcomings. First, we have developed a yeast surface display method for co-selecting sets of mutations that collectively enhance antibody binding affinity and thermal stability. Our approach uses conformational ligands specific for stably folded antibodies to simultaneously probe for stability and affinity during in vitro antibody sorting. This approach has led to highly evolved and stable human antibody domains with significantly increased affinity.

Mutational analysis of these evolved antibodies reveals that many affinity-enhancing mutations within the antibody binding loops are destabilizing and that compensatory mutations are necessary to maintain antibody stability. Interestingly, our finding that it is necessary to accumulate compensatory mutations during affinity maturation to maintain thermodynamic stability for antibodies generated in vitro is similar to findings for antibodies generated in vivo, which suggests that the formation of the antigen-binding site is generally a destabilizing process. We have used this knowledge to develop refined mutagenesis and selection techniques for generating and sorting antibody libraries. Our libraries have mutations focused in the most important antibody binding loop and follow patterns of amino acid diversity found across thousands of natural antibody sequences. We have used these novel libraries as well as strong positive and negative selections to identify interesting antibody variants that target aggregated forms of the Alzheimer’s Aβ peptide with high binding affinity and specificity. Further, the stability and specificity of the selected antibodies are similar to those observed for natural antibodies and far superior to the properties of other antibodies we have identified using conventional in vitro methods. Collectively, our work provides new strategies to improve upon the design, evolution and selection of in vitro antibody libraries to yield high affinity antibody variants with favorable biophysical properties.