Each regulatory subunit of cAMP-dependent protein kinase has two tandem cAMP-binding sites, A and B, at the carboxyl terminus. Based on sequence homologies with the cAMP-binding domain of the Escherichia coli catabolite gene activator protein, a model has been constructed for each cAMP-binding domain. Two of the conserved features of each cAMP-binding site are an arginine and a glutamic acid which interact with the negatively charged phosphate and with the 2'-OH on the ribose ring, respectively. In the type I regulatory subunit, this arginine in cAMP binding site A is Arg-209. Recombinant DNA techniques have been used to change this arginine to a lysine. The resulting protein binds cAMP with a high affinity and associates with the catalytic subunit to form holoenzyme. The mutant holoenzyme also is activated by cAMP. However, the mutant R-subunit binds only 1 mol of cAMP/R-monomer. Photoaffinity labeling confirmed that the mutant R-subunit has only one functional cAMP-binding site. In contrast to the native R-subunit which is labeled at Trp-260 and Tyr-371 by 8-N3cAMP, the mutant R-subunit is convalently modified at a single site, Tyr-371, which correlates with a functional cAMP-binding site B. The lack of functional cAMP-binding site A also was confirmed by activating the mutant holoenzyme with analogs of cAMP which have a high specificity for either site A or site B. 8-NH2-methyl cAMP which preferentially binds to site B was similar to cAMP in its ability to activate both mutant and wild type holoenzyme whereas N6-monobutyryl cAMP, a site A-specific analog, was a very poor activator of the mutant holoenzyme. The results support the conclusions that 1) Arg-209 is essential for cAMP binding to site A and 2) cAMP binding to domain A is not essential for dissociation of the mutant holoenzyme.