We used the lad forward mutagenesis system to determine the mutational specificity of UV-induced mutation in a repair-proficient (Uvr') and a repair-deficient (AUvrB) strain ofEscherichia coli. The spectra recovered at similar levels of mutagenesis were similar, the exception being a mutational hotspot at site A24 specific to the AUvrB strain. Mutations induced at this hotspot, as well as those induced at other mutational hotspots that were found to be common to both the Uvr' and Uvr-strains, involve G-C -* APT transitions. All of the hotspots are at sites of potential dipyrimidine photoproducts, such as thymine-cytosine and cytosine-cytosine dimers, or of pyrimidine-cytosine photoproduct Py-C* lesions. Each of these hotspots occurs at a site in the potential hairpin loop of quasi-palindromic sequences. These observations suggest an important role for DNA structure in determining the fate of UV-induced premutational lesions.The lacI system of E. coli provides a method for determining UV-induced mutational specificity at a large number of sites (1-3). In contrast, earlier studies in other systems generally relied upon the analysis ofreversion at a rather limited number of sites (4-6). Often the mutants analyzed in such reversion studies were originally induced by the mutagen (7); therefore, the possibility existed that preferentially mutable sites or hotspots had been selected and that these might behave atypically. Alternatively, the original mutation might have removed a critical DNA target sequence rendering the site practically immutable. Moreover, in studies of the reversion of nonsense mutations, the majority of "revertants" often occur not in the structural gene but at suppressor loci, which behave differently in their responses to UV light (4, 8, 9). The lacl system allows an examination of forward mutagenesis at 65 individual sites where nonsense mutations can arise by a single base substitution. Because both the DNA sequence and the location of the nonsense mutations have been established (10), each mutation can be attributed to a specific transition or transversion event.Detailed knowledge ofthe sites at which mutagenesis occurs and of the specific base changes produced may yield important clues about the premutagenic lesions and how they are processed. In the case ofUV irradiation, the mutational mechanism is poorly understood. Photoreactivation experiments suggest that mutagenesis in both excision-proficient and excision-deficient strains requires pyrimidine dimers (11,12). Nevertheless, considerable evidence exists that not all UV-induced mutagenesis occurs at the sites of induced lesions (13). For example, undamaged bacteriophage A plated on UV-irradiated hosts show high levels of recA+-dependent mutagenesis ("indirect mutagenesis") (14). In a current model, UV-induced mutagenesis in E. coli proceeds through a recA+-dependent, inducible error-prone repair pathway called "SOS" repair [reviewed by Witkin (13)]. It is postulated that, where alternative repair pathways are unable to function...