The generally accepted mechanism for the formation of etheno derivatives upon reaction of adenosine or cytidine with haloacetaldehydes involves two intermediates. The first, a primary addition to the exocyclic amino group, has not been experimentally verified. The second, a cyclic form of the first intermediate, has been described in monomers but presumed to be too unstable to exist in polynucleotides since such derivatives would be readily dehydrated to other derivatives at pHs below neutrality. We have found that the cyclic intermediates of adenosine and cytidine are the predominant products in polynucleotides, even upon extensive reaction with chloroacetaldehyde at neutrality. The hydrated compounds have half-lives at pH 7, 37 degrees C, of 1.4 h and 13 h for adenosine and cytidine, respectively. Two types of evidence are presented for the existence of the first intermediate, a (1-hydroxy-2-chloroethyl)-substituted exocyclic amino group. Firstly, poly d[A-T] cannot form etheno derivatives (except when denatured) and the observed cross-linking is therefore attributed to alkylation by the chlorinated sidechain of the adenine residue (A), acting on the N6 of A on the opposite strand. Secondly, our results show that blocking of the acceptor nitrogen, needed for cyclization, leads to the formation of relatively stable derivatives of adenosine and cytidine. Guanosine, as a monomer, is modified extensively, but in synthetic polymers no reaction was detected, possibly due to secondary structure.