The mechanism of the photodimerization of adjacent adenine bases on the same strand of DNA has been elucidated by determining the structure of one of the two major photoproducts that are formed by UV irradiation of the deoxydinucleoside monophosphate d(ApA). The photoproduct, denoted d(ApA)*, corresponds to a species of adenine photodimer first described by Pörschke (Pörschke, D. (1973) J.Am.Chem.Soc. 95, 8440-8446). From a detailed examination of its chemical and spectroscopic properties, including comparisons with the model compound N-cyano-N1-(1-methylimidazol-5-yl)formamidine, it is deduced that d(ApA)* contains a deoxyadenosine unit covalently linked through its C(8) position to C(4) of an imidazole N(1) deoxyribonucleoside moiety bearing an N-cyanoformamidino substituent at C(5). On treatment with acid, d(ApA)* is degraded with high specificity to 8-(5-amino-imidazol-4-yl)adenine whose identity has been confirmed by independent chemical synthesis. It is concluded that the primary event in adenine photodimerization entails photoaddition of the N(7)-C(8) double bond of the 5'-adenine across the C(6) and C(5) positions of the 3'-adenine. The azetidine species thus generated acts as a common precursor to both types of d(ApA) photoproduct which are formed from it by competing modes of azetidine ring fission.