Psoralens are mutagenic compounds of vegetable origin that are used as photosensitizing agents in the treatment of various skin diseases, blood cell cancer, and autoimmune disorders. To study the mechanism of mutagenicity of psoralens in humans, we examined the efficiency and fidelity of simian virus 40 origindependent replication in a human cell extract of M13mp2 DNA randomly treated with the psoralen derivative 4-hydroxymethyl-4,5,8-trimethyl psoralen plus UVA irradiation. Replication of DNA treated with variable amounts of 4-hydroxymethyl-4,5,8-trimethyl psoralen and a fixed UVA fluence was inhibited in a concentration-dependent manner. However, covalently closed monomer-length circular replication products were observed. Product analysis by renaturing agarose gel electrophoresis after cross-linking with 250-to 280-nm UV light indicated that approximately 1 of 9 psoralen monoadducts was bypassed during in vitro replication. Introduction of product DNA into Escherichia coli to score replication errors in the lacZ␣ reporter gene demonstrated that replication of the damaged DNA was more mutagenic than was replication of undamaged DNA. Sequence analysis of lacZ mutants revealed that damage-dependent replication errors were predominantly T ⅐ A3C ⅐ G transitions, transversions at C ⅐ G base pairs, and deletions of single A ⅐ T base pairs, the last occurring most frequently in homopolymeric runs. A comparison of error specificities with two substrates having the replication origin asymmetrically placed on opposite sides of the mutational target suggests that the lagging-strand replication apparatus is less accurate than the leading-strand replication apparatus for psoralen monoadduct-dependent deletion errors. A model is proposed based on the preferential loopout of the monoadducted base from the strand that templates retrograde discontinuous synthesis.DNA replication in eukaryotic cells is a complex process requiring a variety of proteins to synthesize the leading and lagging strands in an asymmetric yet coordinated manner (14). Because of this asymmetry, DNA adducts that escape repair may have different potentials for blocking replication or promoting replication infidelity, depending on which strand the adduct is located on. For example, in mammalian cells deficient in excision repair, mutations were predominantly found on one of the two strands of the hprt gene (6,19,20,50,51), leading to the suggestion that the fidelity of translesion bypass is more error prone on the leading strand (20, 51).To study the processes that determine the fidelity of DNA replication, we have been using the SV40 replication system as a model for human chromosomal replication (for reviews, see references 29 and 37). Using substrates containing the SV40 origin of replication, we (27,30,31,41) and others (11) have shown that the replication apparatus in mammalian cell extracts is quite accurate when replicating undamaged DNA, with an error rate ranging from less than or equal to 6.2 ϫ 10 Ϫ6 to 1 ϫ 10 Ϫ7 error per nucleotide. These observ...