Fluorescence intensity and anisotropy measurements using the fluorescent adenosine cyclic 3',5'-phosphate (cAMP) analogue 1,N6-ethenoadenosine cyclic 3',5'-phosphate (epsilon-cAMP) are sensitive to the dissociation of epsilon-cAMP which occurs when either the type I or the type II regulatory subunit (RI or RII) of cAMP-dependent protein kinase associates with the catalytic subunit. Studies using epsilon-cAMP show that MgATP has opposite effects on the reconstitution of both types of protein kinase: MgATP strongly stabilizes the type I holoenzyme while it slightly destabilizes the type II holoenzyme. The synthetic substrate Kemptide has a small inhibitory effect on the reconstitution of both holoenzymes when tested at 10 microM concentration. The protein kinase inhibitor has a larger effect which is especially pronounced in the reassociation of the type I enzyme. The diminished relative ability of the type I regulatory subunit to compete with the protein kinase inhibitor suggests that the combined effects of the two opposing equilibria (epsilon-cAMP and catalytic subunit binding) are different for the two types of regulatory subunits. Displacement experiments show that cAMP and epsilon-cAMP bind about equally well to the type I subunit. Slow conformational changes accompanying the binding of epsilon-cAMP by both regulatory subunits are greatly accelerated with the holoenzymes, suggesting that dissociation of the holoenzymes occurs via ternary complexes. The time courses of epsilon-cAMP binding also show the heterogeneity of binding characteristics of RII. The 37 000-dalton fragment of type II subunit retains the epsilon-cAMP binding properties of the native subunit. However, only a fraction of the fragment preparation (approximately 32% estimated from sedimentation measurements) binds the catalytic subunit well, suggesting heterogeneity of cleavage.