We have used thymine glycol and dihydrothymine as representative ring saturation products resulting from free-radical interaction with DNA pyrimidines, and urea glycosides and beta-ureidoisobutyric acid (UBA) as models for pyrimidine-ring fragmentation products. We have shown that thymine glycol and the ring-fragmentation products urea and beta-ureidoisobutyric acid, as well as abasic sites, are strong blocks to DNA polymerases in vitro. In contrast, dihydrothymine is not a block to any of the polymerases tested. For thymine glycol, termination sites were observed opposite the putative lesions, whereas for the ring-fragmentation products, the termination sites were primarily one base prior to the lesion. These and other data have suggested that thymine glycol codes for an A, and that a base is stably inserted opposite the damage, whereas when a base is inserted opposite the non-coding lesions, it is removed by the 3-->5 exonuclease activity of DNA polymerase I. Despite their efficiency as blocking lesions, thymine glycol, urea and UBA can be bypassed at low frequency in certain specific sequence contexts. When the model lesions were introduced individually into single-stranded biologically active DNA, we found that thymine glycol, urea, beta-ureidoisobutyric acid, and abasic sites were all lethal lesions having an activation efficiency of 1, whereas dihydrothymine was not. Thus the in vitro studies predicted the in vivo results. When the survival of biologically active single-stranded DNA was examined in UV-induced Escherichia coli cells where the block to replication was released, no increase in survival was observed for DNA containing urea or abasic sites, suggesting inefficient bypass of these lesions. In contrast, beta-ureidoisobutyric acid survival was slightly enhanced, and transfecting DNA containing thymine glycols was significantly reactivated. When mutation induction by unique lesions was measured using f1-K12 hybrid DNA containing an E. coli target gene, thymine glycols and dihydrothymine were found to be inefficient as premutagenic lesions, suggesting that in vivo, as in vitro, they primarily code for A. In contrast, urea and beta-ureidoisobutyric acid were efficient premutagenic lesions, with beta-ureidoisobutyric acid being about 4-5-fold more effective than urea glycosides, which have approximately the same rate of mutation induction as abasic sites from purines. Sequence analysis of the mutations resulting from these ring-fragmentation products shows that the mutations produced are both lesion and sequence context dependent. The possible roles that bypass efficiency and lesion-directed misinsertion might play in mutagenesis are discussed.