Deciphering evolutionary instructions for specifying protein fold and function

Classical studies show that proteins have evolved to fold into functional native states that are, at best, only marginally stable through weak non-covalent interactions encoded by their primary sequences. How such fold and functional information is stored in a single amino acid sequence remains elusive. Using the statistical analysis of covariation between pairs of amino acids at all positions in a protein, here we identify groups of a few key physically-interconnected residues, which we term sectors. What information about the fold and function is captured by sectors? Using simulated-annealing Monte Carlo, we introduce variation in the sequence of a single member of the PDZ family in a manner that either preserves or disrupts sector correlations. Experimentally we show that function is specifically retained in designed proteins that obey sector correlations, and strikingly, even in the absence of a native state. Thus, we suggest that native-state stability is not a fundamental requirement for function, and is encoded in the sequence in an idiosyncratic manner in the PDZ family.