kinase families in cases where the two residues share an overlapping binding site. For example, crystal structures of tyrosine kinasepeptide complexes have revealed a single binding pocket accommodating both the +1 and +3 residues. Such information should be considered when designing future phosphorylation site prediction algorithms. Previous studies have shown that specificity-shifting GW 501516 site mutations tend to lead to large losses in enzyme activity. As such, the evolution of new specificity often proceeds by first acquiring permissive amino acid substitutions that stabilize the protein conformation and then next acquiring substitutions that shift specificity. For example, in the case of the glucocorticoid receptors, it was inferred that neutral mutations that stabilized a new conformation must have been acquired before the specificity-shifting mutation could arise in the receptor’s active site. In another example, it was shown that permissive mutations were required in influenza neuraminidase prior to acquisition of drug resistance mutations that subtly altered binding specificities. In directed evolution experiments, additional compensating mutations were required to restore wild-type levels of activity to proteases mutated in their specificity pockets. This requirement for multiple epistatic mutations is likely to slow the evolution of specificity in these systems, while also significantly reducing the chance of convergent Howard et al. eLife 2014;3:e04126. DOI: 10.7554/eLife.04126 13 of 22 Research Article Biochemistry Genomics and Evolutionary Biology specificity evolution. The evolution of CMGC kinase specificity at the +1 site, however, is an outlier to this paradigm. CMGC kinases seem to be relatively tolerant to modulation of +1 specificity by mutation of the DFGx residue: this mutation did not lead to a significant loss of activity in any of the six kinases we tested. This is an unusual case where a single amino acid mutation can drive divergence of specificity without the need for additional stabilizing mutations. Perhaps this tolerance explains the repeated convergent evolution of the DFGx residue. There has been considerable evolutionary diversification to the primary specificities of the CMGC kinases such that paralogs diverged by more than two billion years have almost no overlap in their preferences. In order to make progress studying a tractable period of evolutionary change, we focused our analysis on the sub-family rooted at the common ancestor of Cdk1 and Ime2. PubMed ID:http://www.ncbi.nlm.nih.gov/pubmed/19825302 Following AncCMGI, a gene duplication occurred, and the specificities of the two descendant paralogs almost completely diverged. Cdk1 recognizes a -P-x- motif, while its paralog Ime2 recognizes R-P-x–R. We studied the mutational trajectory of the paralog leading to Ime2, and a similar analysis awaits for the paralog leading to Cdk1. The mechanism by which both paralogs were retained is unclear, although previous work has revealed how other gene paralogs were retained according to various evolutionary behaviors, including sub-functionalization, neo-functionalization, and avoidance of paralog interference. Future work studying the Cdk1 paralog will reveal to what extent these models fit the evolution of CMGC kinases. The evolution of phosphoregulatory networks is analogous to the evolution of gene transcription regulatory networks. In both cases, changing the specificity of the regulator gene–either a kinase or a transcription factor–can potentially lead to the los
