In many proteins, mutations distal from the active site are known to drastically re-tune specificity. These mutations do not occur randomly in space—they are associated with a conserved network of amino acids that can be learned from large sequence alignments. In families with diverging functions, it is unclear how to reconcile the family-wide conservation of this network with its apparent role in tuning homolog-specific functions. We have generated comprehensive double-mutational libraries across 9 SH3 homologs, encompassing 3 paralog groups with similar but non-overlapping ligand specificity. This set of ∼3.5 million barcoded variants is among the largest mutational libraries ever comprehensively analyzed. Function is assessed by a high-throughput bacterial two-hybrid assay that links binding to growth rate. By seeing how the epistatic networks of these proteins vary across orthologs (same function) and paralogs (different function), we are teasing apart sequence features that determine global SH3-like function from those that determine ligand specificity.
Andrews et al. (Sun,) studied this question.