We have constructed single-stranded, M13-based vectors that contain a specifically located thyminethymine pyrimidine-pyrimidone(6-4) UV photoproduct and have used these to estimate the frequency and accuracy of DNA replication past this adduct in uvrA6 cells of Escherichia coli. Both the normal and the Dewar valence photoisomer of the (6-4) adduct were studied. In the absence of SOS induction, vectors carrying the photoproducts were rarely replicated; relative to the lesion-free control, 1.9% of vectors carrying the normal (6-4) isomer produced plaques, and with the Dewar valence isomer the proportion was 0.4%. In SOS-induced cells, these frequencies rose to 22.1% and 12.3%, respectively. The error frequency of replication past the normal isomer in SOS-induced cells was high; in a random sample of 185 progeny phage analyzed, 169 (91%) contained mutations, all of which were targeted. Equally striking, a high proportion of the mutations (158/169; 93%) were of only one type, namely 3' T C transitions. Both the error frequency and the specificity were much reduced with the Dewar valence isomer; overall, 74/140 (53%) of the phage analyzed were mutant, and of these only 34 (46%) entailed the 3' T -+ C transition. We speculate that the high error frequency and specificity arise from the formation of a stable TOG base pair, involving hydrogen bonds at 0-2 and N-3 in the pyrimidone ring. Potential hydrogen bonds at these sites are coplanar in the normal but not in the Dewar isomer, perhaps explaining the reduced specificity of mutagenesis with the latter adduct.Irradiation of DNA with germicidal UV light produces a variety of photoproducts, the chief of which are the cyclobutyl dipyrimidines (dimers) and the pyrimidine-pyrimidone(6-4) adducts (1, 2). We were interested in studying the thymine-thymine (6-4) adduct [5-hydroxy-6-4'-(5'-methylpyrimidin-2'-one)-thymine] as an example of the latter type of photoproduct, for two reasons. (i) We anticipated that it might provide insights into the mechanism determining base selection during translesion synthesis-that is, during chain elongation past the site of a template lesion-and, in general, help in understanding the relation between the structure of a lesion and its mutagenic potential. Previous experiments with cis-syn (3, 4) and trans-syn (5) cyclobutane dimers, and also with abasic sites (6), each located at a T-T site in the same sequence context, suggested that the UV photoproducts often form correct A'T base pairs during translesion synthesis, and that when mutations occur, they might result from base mispairing. We were therefore interested to examine other lesions for evidence of such correct or aberrant base pairing.(ii) We wished to study the T-T (6-4) adduct because ofconflicting conclusions concerning the relative proportions of mutations induced by dimers and adducts in Escherichia coli. Although adducts are collectively 5-to 10-fold less abundant than dimers in UV-irradiated DNA, some evidence (7-9, 25) suggests that they are responsible for the majorit...