Pak2, a member of the p21-activated protein kinase (Pak) family, is activated in response to a variety of stresses and is directly involved in the induction of cytostasis. At the molecular level Pak2 binds Cdc42(GTP), translocating Pak2 to the endoplasmic reticulum where it is autophosphorylated and activated. Pak2 is autophosphorylated at eight sites; Ser-141 and Ser-165 in the regulatory domain and Thr-402 in the activation loop are identified as key sites in activation of the protein kinase. The function of phosphorylation of Ser-141 and Ser-165 on the activation was analyzed with wild-type (WT) and mutants of Pak2. With S141A, the level of autophosphorylation was reduced to 65% as compared with that of WT and S141D with a concomitant 45% reduction in substrate phosphorylation, indicating that phosphorylation at Ser-141 is required for optimal activity. Autophosphorylation inhibited the interaction between WT Pak2 and Cdc42(GTP). In 293T cells, WT Pak2, S141A, and S141D formed a stable complex with the constitutively active mutant Cdc42 L61, but not with the dominant negative Cdc42 N17. As shown in glutathione S-transferase pull-down assays, S141A bound to Cdc42(GTP) at a 6-fold higher level than that of S141D. In contrast, the S165A and S165D mutants had no effect on autophosphorylation, binding to Cdc42, or activation of Pak2. In summary, autophosphorylation of Ser-141 was required for activation of Pak2 and down-regulated the interaction of Pak2 with Cdc42. A model is proposed suggesting that binding of Cdc42 localizes Pak2 to the endoplasmic reticulum, where autophosphorylation alters association of the two proteins.