Antibodies can evolve to recognize essentially any protein with high specificity and affinity. While natural antigen binding sites utilize all 20 natural amino acids to some extent, analysis of functional antibodies reveals clear biases for or against some amino acids. Most significantly, tyrosine and serine are highly abundant in antigen binding sites in general and at antigen contact sites in particular.

As it is now possible to construct antibody-phage libraries with synthetic diversity, we have used synthetic antibody libraries to investigate the roles of different chemical diversity in antigen recognition. Using a tetranomial genetic code that allows for only four amino acids (tyrosine, serine, alanine, and aspartate) we were able to generate antibodies against vascular endothelial growth factor (VEGF) that bound with high affinity and specificity. Structural and mutational analyses indicated that tyrosine was the major mediator of binding energy at the antigen binding sites of the anti-VEGF Fabs, and the results suggested that it might be possible to further simplify the code for antigen recognition. To this end, we constructed Fab libraries in which combining sites were randomized with a binary genetic code that allowed for only tyrosine and serine. We envisioned that in these libraries, tyrosine might act as an effective “functional” amino acid with a large side chain that could provide significant binding contacts. In contrast, serine might act as an “auxiliary” amino acid with a small side chain that could provide space for the tyrosine side chains and also contribute to an overall hydrophilic surface.

Remarkably, naïve Fab libraries constructed with the binary code were extremely effective in generating highly specific antibodies against a wide array of antigens. Furthermore, the binary Fabs exhibited exquisite specificity in cell-based assays. The structures of binary Fabs in complex with antigen reveal that the highly homogenous binding surfaces are dominated by tyrosine, but nonetheless, recognize diverse chemical groups on the antigen. In ongoing studies, we are now adding back chemical and conformational diversity into the minimalist binary background in a precisely defined manner, and these highly controlled studies will enable the accurate assessment of the consequences of expanded diversity.

Our results suggest that certain amino acids possess features that increase the likelihood that they will be able to make productive contributions to binding affinity and specificity. Thus, biased libraries that favour such amino acids are likely to be much more effective in generating antibodies with high affinity and specificity. As a corollary, it seems that other amino acids are ill suited for productive contacts, and the absence or depletion of such amino acids will likely improve naïve antigen recognition. The results should aid the design and use of synthetic antibody libraries. In addition, and perhaps more importantly, the findings have significant implications for the fundamental principles and mechanisms that mediate molecular recognition at protein-protein interfaces.

Extended Abstract Status: Not Uploaded