Hierarchy of Interactions in Protein Evolution

Deciphering the relationship between genotype and phenotype is complicated by the sheer number of possible cooperative interactions amongst the parts that make up biological systems. For even small systems such as individual protein domains, it has been difficult to comprehensively obtain high quality empirical data of amino acid interactions to distinguish different models for the global pattern of cooperativity. The statistical coupling analysis (SCA) - one approach for studying the co-evolution of amino acid positions in homologous sequences - provides a model for this pattern that is distinct from spatial proximity in tertiary structure, positional conservation, or even other forms of co-evolution. Here, we use an extension of deep mutational scanning to analyze nearly 50,000 single and double mutations in several homologs of a model protein - the PDZ family of protein interaction domains. Across the domains queried experimentally, the distributions of couplings between pairs of positions from all possible double mutants are well-approximated by unimodal distributions such that their average provides an estimate of the intrinsic coupling between them. Importantly, the SCA provides the best representation of this experimental pattern of couplings conserved among the homologs. These results highlight the heterogeneous pattern of couplings in protein structures and motivate the re-focus of efforts to understand protein folding and function toward the study of the origin of the co-evolving network of amino acids.