The spectrum of spontaneous mutation of an endogenous mammalian cell gene has been determined at the DNA sequence level. Thirty independent spontaneous APRTmutations were cloned and subsequently completely sequenced. Twenty-seven contained single base substitutions. Of these, 22 were G-C to A-T transitions, suggesting a major role for the deamination of cytosine in spontaneous mutagenesis of Chinese hamster ovary cells. The remaining mutants included a tandem double substitution, a -1 frameshift, and a 17-base-pair deletion flanked by a 2-base-pair direct repeat. Many of the independently recovered mutants were clustered at sites of multiple occurrence (hot spots). One site accounted for >25% of all independently recovered events. Mutations were generally located within the coding sequence, although two mutations occurred within the consensus sequence for a 3' splice site.A fundamental understanding of mutagenesis in mammalian cells ultimately requires specific knowledge of the DNA sequence alterations involved. A complete analysis would describe the spectrum of mutational alterations for a random collection of independent mutants at a given locus. Such analyses have recently been reported in bacteria (1-3), but logistical difficulties have prevented such an analysis of an endogenous mammalian cell gene. Other endpoints have, by necessity, been used to construct a preliminary view of mutation in mammalian cells. These include comparative mutation induction at several selectable loci (4, 5), twodimensional gel electrophoresis analysis of mutant proteins (6), and Southern blot analysis of restriction site polymorphisms (7-10). Recently, recombinant shuttle vectors, which can be propagated in both prokaryotic and eukaryotic cells, have been developed for the analysis of mutant DNA sequences in mammalian cells (reviewed in ref. 11). Although the use of shuttle vectors facilitates the recovery and sequence analysis of mutations, this experimental approach has several drawbacks, often including an extraordinarily high mutation frequency and a high percentage of deletions and rearrangements among the recovered plasmids (12-16). This is true even under conditions in which the shuttle vector has been incorporated into the chromosome, although some inserts appear to approach the stability expected for natural genes (17, 18). These technical problems and the artificial nature of such constructs make it questionable whether these systems accurately represent mutagenesis at endogenous cellular genes.Our efforts have therefore concentrated on the development of a practical approach to the cloning and sequencing of mutant alleles of an endogenous gene. Our system involves the analysis of mutants at the adenine phosphoribosyltransferase (APRT) locus of Chinese hamster ovary (CHO) cells. This locus is well suited for mutational studies. It encodes an enzyme in the nucleotide salvage pathway and is thus nonessential for cell survival in ordinary growth medium. Consequently, all classes of mutational events can in principl...