The uracil DNA glycosylases (EC 3.2.2.3) characterized to date remove uracil from DNA irrespective of whether it is base paired with adenine or mispaired with guanine in double-stranded substrates or whether it is found in single-stranded DNA. We report here the characterization of uracil glycosylase activity that can remove the base solely from a mispair with guanne. It does not recognize uracil either in A-U pairs or in single-stranded substrates. The enzyme, a 55-kDa polypeptide, was previously characterized as a mismatch-specific thymine DNA glycosylase and was thought to be responsible solely for the correction (to GaC) of G T mispais that arise as a result of spontaneous hydrolytic deaination of 5-methylcytosine to thymine. Given the broader substrate specificity of the enzyme (in addition to uracil and thymine, the protein can also remove 5-bromouracil from mispairs with guanine), we propose that its biological role in vivo may also incinde the correction of a subset of GNU mips inefficiently removed by the more abundant ubiquitous uraci glycosylases, such as those arising from cytosine deatnation in G+C-rich regions of the genome.Uracil residues arise in DNA either by cytosine deamination (1) or by incorporation into the nascent strand during replication as dUMP (2). The former process gives rise to mutagenic G-U mispairs, whereas the latter yields A-U "pairs." Uracil is removed from both these lesions and from singlestranded DNA by ubiquitous uracil DNA glycosylases (EC 3.2.2.3) (2, 3). A recent study addressing the removal ofuracil from a large number ofdifferent sequence contexts suggested that the major uracil glycosylases of both mammals and bacteria may require local strand separation for efficient removal of the base from DNA (4). Correspondingly, G-U mispairs in a G+C-rich sequence context were found to be poor substrates for these enzymes in vitro, although this may not be so for all uracil glycosylases (5). Interestingly, analysis of mutational spectra of several mammalian genes published to date revealed no evidence for an increased rate of C --T transition mutations that could be associated with G+C-rich regions (6, 7). This would imply either that the ubiquitous enzyme is so abundant that it manages to repair even the less accessible substrates or that a second enzyme may be able to compensate for the inefficiency ofthe ubiquitous activity. We describe here another glycosylase activity that, due to its different substrate recognition requirements, efficiently removes uracil from GNU mispairs in double-stranded DNA and may thus be a candidate for such a back-up role.
MATERIALS AND METHODSAll the reagents and solvents used in this study were of analytical purity. The Preparation of Whole Cell Extracts. These extracts were prepared from 180 g offrozen HeLa cells in three 60-g batches as described (9).Fractionation of the Cell Extracts. The extract was diluted 1:4 with HE buffer [25 mM Hepes, pH 7.8/1 mM EDTA/1 mM dithiothreitol/10% (vol/vol) glycerol] and incubated batchwise for 1 h w...