The specificity of protein kinases is usually examined using synthetic peptide substrates, either designed variants, or, more recently random peptide libraries. However not all protein kinases utilize synthetic peptides efficiently as substrates. Even among those that do, these approaches neglect effects caused by three-dimensional protein conformation, or the existence of determinants remote from the phosphorylation site. To follow up our previous peptide studies on the specificity of the AMP-activated protein kinase (AMPK) [Dale, S., Wilson, W. A., Edelman, A. M., & Hardie, D. G. (1995) FEBS Lett. 361, 191-1951, we have expressed the C-terminal, catalytic domain of Chinese hamster hydroxymethylglutaryCoA reductase in Escherichia coli. The domain was expressed with an N-terminal His, tag which allowed rapid purification on Ni'+-agarose. The purified protein retained full enzymic activity, and, as with the native enzyme, was totally inactivated by phosphorylation by AMPK at a single site corresponding to Ser871. Using a novel modification of the unique-site elimination method (which allowed direct mutagenesis of the double-stranded expression vector using a single oligonucleotide primer) we expressed 18 mutations involving residues around Ser871. The results broadly confirmed the recognition motif previously proposed on the basis of peptide studies. Three of the mutants were better substrates for AMPK than the wild type, and one of these (K872A) had hydroxymethylglutaryl-CoA reductase kinetic parameters virtually indistinguishable from the wild type. This suggests that hydroxymethylglutaryl-CoA reductase may have been selected to be a sub-optimal substrate for AMPK.Keywords: AMP-activated protein kinase; hydroxymethylglutaryl-CoA reductase ; protein-kinase specificity ; site-directed mutagenesis ; unique-site elimination.Protein kinases and phosphatases catalyze cycles of phosphorylation and dephosphorylation which appear to represent the major switching mechanism in eukaryotic cells [l 1. Predictions from current partial genome sequences suggest that vertebrate genomes encode of the order of 2000 protein kinases 121. Some protein kinases appear to be specific for a single protein target, but many phosphorylate multiple targets, although even in the latter cases only specific serine, threonine or tyrosine side chains are modified. A key question therefore is, how do protein kinases recognize the correct side chains on the correct protein targets?Previous studies on the specificity of protein kinases have generally used variant synthetic peptide substrates whose sequences were based on phosphorylation sites on known target proteins (e.g. . While this approach has been very informative (reviewed in [6]), it does require prior identification of protein substrates, and sequencing of their phosphorylation sites.