N-2-Acetylaminofluorene (AAF), a well-known chemical carcinogen, when covalently linked to guanine residues constitutes a premutagenic lesion that is converted in vivo into frameshift mutations. In Escherichia coli, it is thought that -AAF adducts block the replication fork and that the mutagenic processing of the -AAF adducts is mediated by the SOS response. The construction in vitro of plasmids containing -AAF adducts in one strand only of a double-stranded DNA molecule enabled us to investigate the segregation of the strands and the mutagenicity of the lesions in vivo. The two DNA strands were "genetically labelled" by means of a single base-pair mismatch in the tetracycline-resistance gene, one strand carrying the wild-type allele and the other strand a mutant tetracycline-sensitive allele. The two strands contained either no -AAF adducts, -AAF adducts in one strand or -AAF adducts in both strands. When such constructions are used to transform bacterial cells the following are found. When no -AAF adducts are present on either strand of the DNA, a mixture of plasmids having information from both parent strands is found in 80% of the transformed bacterial clones. With -AAF adducts present in one strand only, in 90% of the transformants there is a consistent loss of the parent strand information that contained the -AAF adducts. In the constructions having -AAF adducts in both strands, the transformed bacteria carry either one or the other allele in a pure form. Our results suggest that when blocking lesions (-AAF adducts) are present in one strand only, they trigger the specific loss of that strand. The forward mutation frequency (i.e. the tetracycline-resistance gene inactivation frequency) was found to be more than ten times lower when the -AAF adducts are bound to one strand only compared with the situation where both strands carry the premutagenic lesions. Moreover, when the isolated mutants were sequenced, the mutations were found to consist of a mixture of true -AAF-induced mutations (i.e. -1 or -2 frameshift mutation at previously determined mutation hot spots) and of mutations that are not targeted at -AAF adducts. We suggest that these "background" mutants arose from the mutagenic processing of cryptic lesions present in our DNA. The low mutagenic efficiency of -AAF adducts, when present in one strand only of a duplex DNA, most probably results from the above-described loss of the damaged strand.(ABSTRACT TRUNCATED AT 400 WORDS)