Incorporation of the available data on rac in neutrophils, CDC42 in yeast, and rho in fibroblasts suggests a general model for the function of rho-like GTPase (Figure 1). Conversion of an inactive cytoplasmic rho-related p21GDP/GDI complex to active p21. GTP occurs by inhibition of GAP and/or stimulation of exchange factors in response to cell signals. p21.GTP is then able to interact with its target at the plasma membrane. This could result in a conformational change in the target, enabling it to bind cytosolic protein(s). Alternatively, p21.GTP could be actively involved in transporting cytosolic protein(s) to the target. A GAP protein, perhaps intrinsic to the complex, would stimulate GTP hydrolysis allowing p21.GDP to dissociate. Solubilization of p21GDP by interaction with GDI would complete a cycle. What about the nature of the final complex? The rac-regulated NADPH oxidase complex in neutrophils is currently the best understood and most amenable to further biochemical analysis. Two plasma-membrane bound subunits encode the catalytic function necessary for producing superoxide, but the two cytosolic proteins, p47 and p67, are essential for activity. Why the complexity? Production of superoxide is tightly coordinated with phagocytosis, a membrane process driven by rearrangement of cortical actin. This is not unrelated to the membrane ruffling and macropinocytosis that we observe in fibroblasts microinjected with p21rac. It is tempting to speculate, therefore, that in neutrophils rac is involved not only in promoting the assembly of the NADPH oxidase but also in the coordinate reorganization of cortical actin leading to phagocytosis. For CDC42 controlled bud assembly in yeast, the components of the plasma-membrane complex are not so clear. By analogy with rac in neutrophils, it seems likely that CDC42 is involved in promoting the assembly of cytosolic components at the bud site on the plasma membrane. These putative cytosolic proteins have not yet been identified, but BEM1 and ABP1 are two possible candidates. The biochemical basis for the stimulation of adhesion plaques and actin stress fibers by p21rho in fibroblasts is also unclear. However, components of the adhesion plaque such as vinculin and talin are known to be cytosolic when not complexed with integrin receptors, and rho could be involved in regulating their assembly into the adhesion plaque. Several things are still difficult to incorporate into this model. First the target for CDC42, the bud site, although not yet structurally defined requires the activity of another small GTPase, BUD1. Similarly, in activated neutrophils, the NADPH oxidase is found in a complex with rap1, the mammalian homologue of BUD1 (BoKoch et al., 1989). It seems likely, therefore, that the target is not simply a plasma-membrane protein but may be a complex of proteins whose formation is under the control of the rap1/BUD1 GTPase. The other black box in this model is the actin connection: activation of bud assembly by CDC42 is followed by actin polymerization, activation of NADPH oxidase in neutrophils occurs concomitantly with phagocytosis, a cortical actin-dependent process, and p21rho in fibroblasts couples the formation of adhesion plaques to actin stress fibers. One possible link between the GTPase-driven assembly of a plasma-membrane complex and actin polymerization could involve the SH3 domain. Interestingly, both p47 and p67 and yeast ABP1 and BEM1 have SH3 domain. If rho-like GTPases recognize plasma-membrane targets already associated with cortical actin, then this could promote an interaction with a subset of SH3-containing proteins. The result of this would be a GTPase-regulated …