Adenylate cyclase from bovine brain cortex was solubilized with 14 mM cholate and 1 M (NH4)2SO4. Gel filtration over a column of Sepharose 6B separated the catalytic unit (CU) from a factor (G/F) that confers responsiveness to 5'-guanylyl imidophosphate (p[NH]ppG) or fluoride. The separated CU, which elutes with a Kav, of 0.48 +/- 0.01 (n=5), is not responsive to p[NH]ppG or fluoride and is relatively inactive when Mg . ATP is the substrate but activated 8-15-fold by Mn2+. The separated G/F elutes with a Kav of 0.70 +/- 0.02 (n=4). It restores the responsiveness of the CU to p[NH]ppG and fluoride. Activation of the enzyme by p[NH]ppG before solubilization does not decrease the amount of G/F eluting with a Kav of 0.7. Therefore, the G/F is probably present in brain cortex in excess over the CU. p[NH]ppG stabilizes the G/F but not the CU against thermal inactivation, suggesting that it interacts with G/F and not with CU. Incubation of the G/F with p[NH]ppG before addition of CU markedly increases the rate of activation of the reconstituted enzyme by p[NH]ppG. We propose, therefore, that the rate-limiting step in adenylate cyclase activation is a process in G/F alone and not a slow conformational change in CU or a slow association of G/F with CU. Binding of p[NH]ppG to the isolated G/F appears to be readily reversible; the ability of fully activated G/F to stimulate CU can be blocked if GDP is added before CU. In contrast, after the CU has been activated by interaction with G/F, GDP cannot reverse the activation. This suggests that association with the CU increases the affinity of G/F for p[NH]ppG.