Abstract
PAKs (p21 activated kinases) are an important class of Rho effectors. These contain a Cdc42-Rac1 interaction and binding (CRIB) domain and a flanking auto-inhibitory domain (AID) which binds the C-terminal catalytic domain. The group II kinases PAK4 and PAK5 are considered significant therapeutic targets in cancer. Among human cancer cell lines we tested, PAK5 protein levels are much lower than those of PAK4, even in NCI-H446 which has the highest PAK5 mRNA expression. Although these two kinases are evolutionarily and structurally related, it has never been established why PAK4 is inactive whereas PAK5 has high basal activity. The AID of PAK5 is functionally indistinguishable from that of PAK4, pointing to other regions being responsible for higher activity of PAK5. Gel filtration indicates PAK4 is a monomer but PAK5 is dimeric. The central region of PAK5 (residues 109-420) is shown here to promote self-association, and an elevated activity, but has no effect on activation loop Ser(602) phosphorylation. These residues allow PAK5 to form characteristic puncta in cells, and removing sequences involved in oligomerization suppresses kinase activity. Our model suggests PAK5 self-association interferes with AID binding to the catalytic domain, thus maintaining its high activity. Further, our model explains the observation that PAK5 (1-180) inhibits PAK5 in vitro.