es. We used a positional scanning peptide library to characterize the full primary specificity of these kinases. Briefly, we used a set of 182 peptide mixtures, in which a central phosphoacceptor position was surrounded by random sequence. Within each mixture, one of nine positions was fixed to a single amino acid residue. Peptides were subjected in parallel to a radiolabelled kinase assay, and the extent of radiolabel incorporation indicates which residues are preferred or disallowed by the kinase at each position within the peptide sequence. As previously reported for S. cerevisiae Ime2 and mouse ICK, PSPL analysis revealed that all kinases assayed share a strong preference for arginine at the 3 position and proline at the 2 position. However, we found that selectivity for residues C-terminal to the phosphoacceptor was more variable. Specifically, the preferred residue at the +1 position varied between arginine for the S. cerevisiae, Candida glabrata, and Yarrowia lipolytica Ime2 homologs and proline for the three mammalian RCK kinases. Neurospora crassa Ime2 and Naegleria get 1022150-57-7 gruberi LF4 phosphorylated peptides with +1R and +1P relatively equally. All kinases also tolerated alanine relatively equally. Additional biochemical characterization using four consensus peptides that were varied at the phosphoacceptor and +1 positions revealed differences in steadystate kinetics underlying the +1 specificity switch. Ancestral reconstruction of the CMGC group of kinases We sought to reconstruct the evolutionary events that led to the modern diversity of IME2/RCK/LF4 specificities. To achieve this goal, we curated a library of 329 amino acid sequences sampled broadly from acrross the CMGC group and then reconstructed their evolutionary history using maximum likelihood phylogenetic methods. The resulting phylogeny and reconstructed ancestral sequences were strongly supported by 
the evolutionary model. These N-terminal determinants correspond to the conserved motif found in the IME2/RCK/LF4 kinases as well as the DYRK kinases. Interestingly, AncCMGI could phosphorylate peptides having either a proline or an arginine residue at the +1 position, though it displayed a 5.6-fold preference for proline. Thus AncCMGI appeared to have a modest +1 proline preference, in contrast to the more stringent proline requirement of the extant CDK and MAPK families. Thus, the specificity of AncCMGI contains elements of the diverged specificities of many major sub-families of the CMGC group. Notably, the domain architecture of AncCMGI is most similar to IME2/RCK/LF4 kinases. That is, AncCMGI contains the canonical CMGC insert loop, but it lacks any C-terminal extension. Furthermore, AncCMGI does not appear to require cyclin for activity: we observed no significant co-purifying proteins, and E. coli does not encode any cyclin orthologs. These data indicate that the cyclin dependence of CDKs and the requirement for an PubMed ID:http://www.ncbi.nlm.nih.gov/pubmed/19826170 additional C-terminal extension to stabilize the Ca helix in MAPKs are characteristics that arose later during evolution. In addition, AncCMGI contains a MAPK-like TXY motif in the activation loop. Phosphorylation of this motif is required for the activation of MAPKs. Because E. coli lack endogenous kinases capable of phosphorylating this TXY motif, AncCMGI is likely activated through autophosphorylation, similar to extant mammalian DYRK and GSK family kinases. +1 specificity evolved from modest proline preference to strong arginine preference via an expande