1. GTP, but not p[NH]ppG (guanosine 5'-[betagamma-imido]triphosphate), abolishes the sensitivity of glucagon-stimulated adenylate cyclase to the lipid-phase separations occurring in the outer half of the bilayer in liver plasma membranes from rat. 2. When either GTP or p[NH]ppG alone stimulate adenylate cyclase, the enzyme senses only those lipid-phase separations occurring in the inner half of the bilayer. 3. Trypsin treatment of intact hepatocytes has no effect on the basal, fluoride-, GTP- or p[NH]ppG-stimulated adenylate cyclase activity. However, (125)I-labelled-glucagon specific binding decays with a half-life matching that of the decay of glucagon-stimulated adenylate cyclase activity. 4. When GTP or p[NH]ppG are added to assays of glucagon-stimulated activity, the half-life of the trypsin-mediated decay of activity is substantially increased and the decay plots are no longer first-order. 5. Trypsin treatment of purified rat liver plasma membranes abolishes basal and all ligand-stimulated adenylate cyclase activity, and (125)I-labelled-glucagon specific binding. 6. Benzyl alcohol activates the GTP- and p[NH]ppG-stimulated activities in an identical fashion, whereas these activities are affected differently when glucagon is present in the assays. 7. We suggest that guanine nucleotides alter the mode of coupling between the receptor and catalytic unit. In the presence of glucagon and GTP, a complex of receptor, catalytic unit and nucleotide regulatory protein occurs as a transient intermediate, releasing a free unstable active catalytic unit. In the presence of p[NH]ppG and glucagon, the transient complex yields a relatively stable complex of the catalytic unit associated with a p[NH]ppG-bound nucleotide-regulatory protein.