To help elucidate the mechanisms involved in spontaneous mutagenesis, DNA sequencing has been applied to characterize the types of mutation whose rates are increased or decreased in mutator or antimutator strains, respectively. Increased spontaneous mutation rates point to malfunctions in genes that normally act to reduce spontaneous mutation, whereas decreased rates are associated with defects in genes whose products are necessary for spontaneous mutagenesis. In this article, we survey and discuss the mutational specificities conferred by mutator and antimutator genes in the budding yeast Saccharomyces cerevisiae. The implications of selected aspects of the data are considered with respect to the mechanisms of spontaneous mutagenesis.

SPONTANEOUS mutations play a fundamental role in evolution and have been implicated in aging, carcinogenesis, and human genetic disease (Harmon 1981; Kirkwood 1989; Cooper and Krawczak 1990; Arber 1991; Drake 1991a; Loeb 1991, 1994; Wintersberger 1991; Caskeyet al. 1992; Strauss 1992). They are thought to originate as a consequence of intracellular events, including the formation of DNA lesions, the occurrence of DNA synthesis errors during replication, repair and recombination, and the movement of transposable elements (Sargentini and Smith 1985; Smith and Sargentini 1985; Ramel 1989; Loeb and Cheng 1990; Drake 1991b; Kunkel 1992; Smith 1992; Amariglio and Rechavi 1993; Ameset al. 1993; Lindahl 1993).

DNA lesions can arise naturally through intracellular metabolism or the intrinsic instability of DNA. For example, there is evidence consistent with spontaneous alkylation, deamination, and loss of DNA bases, as well as their modification by reactive oxygen species (Loeb and Preston 1986; Ameset al. 1993; Lindahl 1993; Friedberget al. 1995). Because a substantial fraction of endogenous DNA damage may involve the alteration or loss of nucleotide bases, many spontaneous mutations may be generated during replication past miscoding or noninstructional lesions. The accuracy of DNA synthesis on undamaged templates is attributed in large measure to DNA polymerase (pol)-mediated nucleotide selection and proofreading of replication errors, as well as to postreplicative mismatch correction of errors that escape proofreading (Loeb and Cheng 1990; Kunkel 1992; Goodmanet al. 1993; Modrich 1994). The topography of the DNA molecule itself also is an important factor, however, and roles have been proposed for DNA secondary structures and misaligned template primers in the production of single and multiple base pair alterations (Ripley and Glickman 1983; Kunkel 1990). Insertions of transposable elements generally constitute larger sequence changes in DNA. Although transposon movement can be a relatively infrequent event, transposition rates may be increased by the presence of DNA damage, including that which occurs spontaneously (Bradshaw and McEntee 1989; Kunz et al. 1990, 1994a).

Two general strategies have been used in attempts to better understand the mechanisms responsible for spontaneous mutagenesis. The first is to characterize strains that have enhanced spontaneous mutation rates. The rationale for this route is that mutator phenotypes are expected to result from defects in genes whose products act to minimize genetic instability. Indeed, as expected, such studies have revealed that spontaneous mutations can arise through failure of DNA repair or processes that maintain the accuracy of DNA replication (Haynes and Kunz 1981; Sargentini and Smith 1985; Krameret al. 1989a; Rebeck and Samson 1991; Michaels and Miller 1992; Smith 1992; Xiao and Samson 1993; Kunzet al. 1994a and references therein). The second approach is to isolate antimutator mutants with the aim to identify genes whose functions are required for spontaneous mutagenesis. A number of these mutants have been recovered, mainly in prokaryotic systems, and where characterized, they have been found to have alterations primarily in genes that encode DNA polymerases (Sargentini and Smith 1985; Morrisonet al. 1989; Smith 1992; Drake 1993 and references therein; Fijalkowskaet al. 1993).

Early studies of spontaneous mutagenesis in mutator and antimutator strains relied on genetic analysis of reversion, suppression, or forward mutation (for reviews, see Lawrence 1982; Sargentini and Smith 1985; Friedberget al. 1995). Because of technological restrictions, the systems used did not identify the DNA sequence alterations involved, or they were unable to detect all types of mutation that occurred. These limitations were overcome by applying DNA sequencing techniques to examine forward mutations in prokaryotic and then eukaryotic organisms (Friedberget al. 1995). The resulting data on the types of DNA sequence changes that occur, the rates at which they arise, their site specificity, and how these parameters are influenced by mutator and antimutator alleles have yielded insights into processes that act to maintain genetic stability and ensure accurate transmission of the hereditary material.

In this article, we review the modulation of spontaneous mutagenesis in mutator and antimutator strains of the budding yeast Saccharomyces cerevisiae, as revealed by DNA sequence analysis. First, we discuss briefly the selected aspects of systems used in the analysis of mutagenesis. Next, sequence changes arising spontaneously in a wild-type background are considered to establish the types of mutations that do occur and provide a basis for comparison. We then survey the effects of eliminating genes whose inactivation leads to a mutator or antimutator phenotype on the magnitude and specificity of spontaneous mutagenesis. Finally, we focus on a select number of findings with implications for the mechanisms of spontaneous mutagenesis in eukaryotic cells.

Assay systems: In vivo mutagenesis assay systems can be based on the characterization of forward or reverse mutations on a target gene (assay gene). Sequence analyses using forward systems have been reported for the CAN1, LYS2, SUP4-o, and URA3 genes. Commonly used reversion systems are based on ade2, his7, hom3, or lys2 alleles. Mutant strains are usually constructed by deleting part or all of the gene, disrupting the gene sequence by inserting additional DNA into the gene, or by introducing point mutations into the gene. Deletion of a significant portion or all of the gene is the preferable method because this usually results in complete loss of the gene's function. In contrast, disruptions and point mutations may result in partial retention or alteration of gene function, and different point mutations especially may produce different phenotypic effects. Assay genes can be located on a chromosome or a centromeric plasmid. This type of plasmid mimics chromosome behavior with respect to replication during S phase, copy number (predominantly single copy), and replicative stability (Campbell and Newlon 1991). Potential differences in chromatin structure and overall size, however, may influence metabolic processes that involve the target gene. This should be considered when interpreting data for plasmid-borne genes.

Forward systems are preferable in the study of mutagenesis because they can detect a wide variety of alterations at many sites within a gene. In contrast, reversion systems only detect specific changes and cannot be used to obtain mutational spectra or to map distributions of mutations. The advantage of a reversion system is that once the relevant change is known, sequencing is not usually required, thereby allowing rapid screening of a large number of samples. In this case, however, genotypic and phenotypic reversion to wild type are assumed to involve the same sequence change, but this may not necessarily be so. For example, intragenic suppression may be indistinguishable phenotypically from locus reversion. Reversion systems will not be considered further in this review.

Mutational spectra derived from different assay systems should be compared with caution because of the inherent characteristics of the systems. Ideally an assay system should involve a target gene large enough to allow all possible types of mutation to occur and be recovered. While a larger target gene should be more representative of the genome, practical difficulties in sequencing could negate this advantage. Multiple sequencing sessions may be required to identify mutations in a large target. Even if a mutation is identified near the sequencing primer, the entire target has to be sequenced to verify that there are no additional alterations. Another desirable feature is a wide variety of sequence contexts, which helps maximize detection of sequence-specific effects. The proportion of intronic sequences in the assay system, however, should be low because many intronic mutations may not produce phenotypic changes. If nondetectable mutations are not randomly distributed with respect to class or location throughout the target gene, then mutational specificity will be biased and the mutational data will not be representative of the true spectrum. Thus, if possible, the ability of the system to detect all possible substitutions at all positions within the target should be characterized. If this requirement is not met, the interpretation of the spectral data should acknowledge this limitation. Irrespective of the size or sequence content of the assay target, a statistically significant number of mutants must be examined to construct a meaningful spectrum. Small data sets increase the probability that spectral features are caused by chance.

Spontaneous DNA sequence alterations: To date, mutations that occur spontaneously in wild-type cells have been characterized using the yeast centromere plasmid-borne SUP4-o allele, the URA3 gene integrated at the chromosomal HIS3 locus, and the endogenous CAN1 gene (Girouxet al. 1988; Leeet al. 1988; Kunzet al. 1990; Kanget al. 1992; Rocheet al. 1994; Tishkoffet al. 1997). The endogenous LYS2 and URA3 genes also have been used, but only to study insertions of the yeast transposable element Ty (Natsouliset al. 1989). The most extensive analysis has involved SUP4-o. In total, eight different classes of mutation were detected at this locus (Table 1), nine if the double base pair substitution category is expanded to show individual tandem and nontandem double changes, and 13 if the complex changes are further subdivided. Single base pair substitution and deletion, multiple base pair deletion, duplication, and more complex replacements also were found at URA3, and single base pair substitution, deletion and insertion, and complex changes were detected at CAN1. Despite the large difference in the numbers of mutations analyzed (SUP4-o: 582; URA3: 70; CAN1: 20), the single base pair substitutions and deletions constituted roughly similar fractions of the total events characterized and occurred at roughly comparable rates for the three genes. Although this points to common mutational mechanisms for genes on yeast centromere plasmids and chromosomes, the very small number of ura3 and can1 mutations analyzed mandates caution when making comparisons. In addition, the spontaneous mutation rate per base pair for SUP4-o is an order of magnitude higher than expected on the basis of comparison to mutation rates for other microorganisms, perhaps because SUP4-o encodes a tRNA or resides on a plasmid (Drake 1991a).

Spontaneous single base pair substitutions have been recovered at 65 of the 66 exon sites, and one of the two intron positions where such changes can be detected in SUP4-o with the genetic screen used. However, only 75% of the substitutions detectable in the exons (129 out of 174) and intron (three out of four) of the gene have been found to occur spontaneously so far, and some positions are mutated much more frequently than others. Both types of transition and all four transversions have been identified in SUP4-o (Table 2), but there is a slight excess of transversions (transversion:transition ratio = 1.44:1). Furthermore, events at G·C pairs outnumber those at A·T pairs by nearly 4:1 (367:97), a ratio that is almost threefold greater than that expected (1.5:1) on the basis of random occurrence of mutations at the G·C (41) and A·T (28) sites that can be detectably mutated. Collectively, the data suggest that the occurrence and repair/correction/editing of spontaneous DNA damage or replication errors in yeast is influenced by sequence context. In addition to single events, a tandem double and a tandem triple substitution, as well as two nontandem double substitutions, have been found in SUP4-o.

Some differences are apparent for substitution mutagenesis at the URA3 and CAN1 genes. The relative proportions of A·T → G·C transitions and A·T → T·A transversions at URA3 differ, and transversions outnumbered transitions 3.3:1 compared with SUP4-o, but there were also fourfold more substitutions at G·C pairs, as observed for SUP4-o. No substitutions were detected at A·T pairs in CAN1, and the fractions of the three substitutions at G·C pairs differ from the corresponding values for SUP4-o, although the transversion:transition ratios (1.6) are similar for SUP4-o and CAN1. These differences may reflect experimental dissimilarities, the very small numbers of mutations analyzed for URA3 and CAN1, or limitations caused by the redundancy of the genetic code on the types of substitutions that can be selected at individual sites in a protein-encoding gene.

Single and multiple base pair deletions (the latter ranging from eight to 807 base pairs in length) or insertions account for 18% of the SUP4-o mutations in the wild-type background. The majority (44 out of 49) of single base pair losses occurred in runs of two, three, or five base pairs with many of the deletions (40), as well as all (five) of the single base pair additions, taking place in a tract of five G·C pairs. The seven deletions and insertions at URA3 and CAN1 also occurred mainly in base pair runs. This pattern is consistent with a mutational mechanism involving misalignment of template and primer strands during replication through runs of repeated base pairs (Streisingeret al. 1966). The much greater number of deletions than insertions within the five-base pair run in SUP4-o implies either that the template strand “slips” and is stabilized more frequently than the primer strand or that the mutational intermediate resulting from primer strand slippage is corrected more efficiently. A number of the multiple base pair deletions retained single copies of short (two to six base pairs), repeated sequences originally present at the deletion termini and so may have been generated by nonhomologous recombination occurring between the repeats. All of the multiple base pair insertions involved integration of the yeast retrotransposon Ty. In at least one instance, the insertion may have been the remnant of a Ty element that excised from SUP4-o by homologous recombination between its flanking delta sequences. Interestingly, most (34 out of 39) of the Ty elements inserted between positions 37 and 38 of the tDNA sequence. The reason for this hot spot is not known, but it may depend in part on functional RAD6 and REV3 genes because deletion of either gene reduces Ty insertions at the hotspot by about one-third (Kanget al. 1992; Rocheet al. 1994). The tendency for Ty elements to insert near tRNA genes (Boeke and Sandmeyer 1991) may also be involved, but it is unlikely to account for the hotspot by itself.

A duplication of seven base pairs and several more complex events were also detected. The complex events include a mutant with nontandem deletions of one and 32 base pairs, two mutants each having a single base pair substitution adjacent to a single base pair deletion, and five mutants having the sequence 5′-GATCTCA-3′ replaced with 5′-CCGGG-3′. The latter change was most likely caused by processing of a DNA secondary structure that was formed by pairing within a quasipalindromic sequence (Kohalmiet al. 1991). This possibility is strongly supported by finding in a separate study (Kohalmiet al. 1991) the other sequence replacement (5′-CCCGG-3′ converted to 5′-TGAGATC-3′) that could be templated by the same secondary structure.

The available data provide evidence that a large variety of DNA sequence alterations can occur spontaneously in yeast, pointing to the complexity of spontaneous mutagenesis in this organism. They also support hypotheses that implicate roles for DNA sequence–directed events in the generation of spontaneous mutations in eukaryotic cells.

View this table:

Sequence alterations identified in mutator and antimutator strains

View this table:

Single base pair substitutions detected in mutator or antimutator strains

Specificity of mutator and antimutator alleles

In the next sections, we describe briefly the mutational specificities of defects that confer mutator or antimutator phenotypes. The genes involved are cataloged according to the processes that they influence: proofreading, mismatch correction, DNA repair [nucleotide excision repair (NER), base excision repair, RAD6-dependent repair, and recombinational repair], and nucleotide metabolism. The data are summarized in detail in Tables 1 and 2, which also provide the relevant references.

Proofreading: DNA pols δ, ε, and γ are the proofreading-proficient polymerases identified in yeast (Campbell 1993; Morrison and Sugino 1993). Exonucleolytic proofreading–deficient (Exo) mutants of pols δ and ε exhibit a mutator phenotype with a mutation rate increase of up to 130-fold for pol δ Exo over wild type. The pol ε Exo mutator confers a much less pronounced mutator effect that is consistent with mammalian DNA pol ε being more accurate than pol δ during in vitro replication (Thomaset al. 1991). The magnitude of the pol ε Exo mutator effect (~10-fold increase in mutation rate) also suggests that the higher fidelity of pol ε results more from presynthetic error control mechanisms than from proofreading. Analysis of mutations generated in SUP4-o and URA3 by the pol δ Exo mutator, as well as mutations in URA3 by the pol ε Exo mutator, revealed that increases in single base pair events (substitutions, deletions, and insertions) accounted for virtually the entire increase in the mutation rates. (No sequence data were presented for spontaneous mutagenesis at URA3 in the wild-type background, so the magnitude of the rate increases for the individual classes of mutation could not be determined. Single base pair mutations, however, were the only types of event recovered at URA3 in the mutator strains.) Among the single base pair substitutions detected in SUP4-o, the G·C → C·G and A·T → T·A transversions exhibited the smallest and greatest rate increases, respectively. C/C mismatches are corrected postreplicatively much less efficiently in Saccharomyces cerevisiae than are G/G mismatches (Bishopet al. 1989; Krameret al. 1989b). Thus, the data suggest that pol δ produces relatively few C/C mispairs when replicating SUP4-o. This interpretation is supported for pol δ and is extended to pol ε by the failure to detect any G·C → C·G transversions among 91 single base pair substitutions induced within a chromosomal copy of the URA3 gene by a pol δ Exo or pol ε Exo mutator. The SUP4-o data also suggest that pol δ proofreads A/A and/or T/T mismatches more efficiently than other base mismatches, a possibility that is consistent with the relative inefficiency with which T/T mismatches in SUP4-o are corrected (Ramachandran 1996; Yanget al. 1996). Interestingly, the magnitude and specificity of the pol δ Exo mutator was influenced by the orientation of the SUP4-o gene with respect to the origin of replication (Ramachandran 1996). The implications of this observation are discussed below. Although an attempt was made to examine the effect of URA3 orientation on the specificities of the pol δ Exo or pol ε Exo mutators (Morrison and Sugino 1994), too few substitutions were characterized (20–30 per URA3 orientation) to arrive at any statistically meaningful conclusion. Nonetheless, it is worth noting that some differences in the relative fractions of substitutions upon inversion of URA3 in the pol δ Exo background are similar to certain orientation-dependent differences detected for SUP4-o.

The mut7-1 allele confers a temperature-sensitive defect in pol δ that results in a mutator phenotype. It is not known, however, whether the mutation influences the polymerizing or proofreading activities of the polymerase. Several observations argue against a proofreading defect. Elimination of proofreading by pol δ is not lethal, whereas mut7-1 confers temperature-sensitive lethality. Furthermore, it is clear that the mutator effect is much smaller than that observed for the proofreading deficiency, although viability at 30° indicates that the full effect of the mut7-1 allele is not apparent at this temperature. In addition, multiple base pair deletions, complex changes, and G·C → C·G transversions were recovered at URA3 in the mut7-1 background, but not at URA3 for the pol δ Exo mutator. One must be cautious in interpreting the latter findings because the data necessary to determine the magnitude of the mut7-1 mutator effect at 32° and 34° were not presented, and at 30°, as many as 25% of the mutations analyzed could have been spontaneous in origin.

Mismatch correction: The specificities of mutators caused by inactivation of the PMS1, MSH2, and MLH1 or MSH3 plus MSH6 mismatch repair genes have been characterized. The mutation rates were increased in all mismatch correction-deficient strains relative to the wild type, reflecting the importance of mismatch correction in maintaining genetic stability. At SUP4-o, for which specificity data for the wild-type background were provided, only the rates of single base pair events were increased in pms1, msh2, and mlh1 strains, with deletions/insertions showing the greatest increases. Recently, MSH2 was reported to be required for correction of a 26-base pair insertion mismatch formed in cells undergoing meiosis (Kirkpatrick and Petes 1997). No multiple base pair insertions/deletions were recovered in the msh2 strains, however, suggesting either that such mismatches form much less frequently in mitotic cells or that the role of MSH2 in large loop resolution may be meiosis specific. Interestingly, the SUP4-o specificity data indicate that the mlh1 mutator features a larger proportion of single base deletions/insertions than the spectra for pms1 and msh2. It has been suggested that the Pms1p, Msh2p, and Mlh1p proteins might function in the same pathway, possibly with interaction between Pms1p and Mlh1p (Prollaet al. 1994). If so, one might expect the mutational specificities of all three mutators to be identical. There may be, however, different substrate-dependent mismatch repair complexes leading to differences in spectra, depending on the mismatch repair gene affected (Marsischkyet al. 1996). Although inactivation of MSH2 resulted in a much greater proportion of deletions/insertions at CAN1 than observed for SUP4-o, relatively few mutations were sequenced, so the significance of this finding remains to be established. The msh6 mutator produced mainly base pair substitutions in CAN1, with a much smaller fraction of single base pair deletions than observed for any other mismatch correction defect. Simultaneous inactivation of MSH3 and MSH6, however, resulted in a spectrum of can1 mutations closer to that for the msh2 mutator. With the exception of the MSH6 strain, more transversions than transitions were recovered in all of the mismatch correction-deficient mutants. Changing the orientation of SUP4-o within its plasmid replicon did not alter the spectrum of mutations obtained in a pms1 strain, suggesting that for naturally occurring mismatches, gene orientation influences neither the direction nor efficiency of mismatch repair (Ramachandran 1996).

Nucleotide excision repair: RAD1 encodes an endonuclease that is thought to function in a complex with Rad10p, incising the DNA backbone 5′ to the lesion (Friedberg 1988; Friedberget al. 1995). The mutation rate of the rad1 mutant was moderately higher than the wild-type strain, with increments in the rates of single base pair substitution and deletion and Ty insertion accounting for the entire rate increase. The rates of all six types of base pair substitutions were increased in the rad1 background. Most of the single base pair deletions occurred in runs of two or more base pairs, consistent with the mutational intermediate being a nucleotide loop that arose through strand slippage during replication. It was suggested that the increase in the rate of single base pair events might reflect a decrease in mismatch correction (Kunzet al. 1990). RAD1, however, is not required for correction of heteroduplex plasmids that carry base mismatches (Kang and Kunz 1992). Furthermore, the increased rates of substitutions and deletions in the rad1 background are dependent on REV3, which encodes a DNA polymerase (ξ) that is believed to function in translesion synthesis (Nelsonet al. 1996b). This implicates translesion synthesis in the generation of spontaneous mutations that arise in the rad1 mutant. Although there is evidence that Rad1p functions in the mismatch repair of a 26-base pair nucleotide loop formed during meiosis (Kirkpatrick and Petes 1997), only three multiple base pair deletions/insertions involving just two or three base pairs were detected among 249 SUP4-o mutations that were analyzed for the rad1 strain.

The RAD3 gene encodes an ATPase/helicase that is not only necessary for NER, but that is also involved in transcription (Friedberg 1988: Friedberget al. 1995). RAD3 is essential because of its latter role and, consequently, only mutators caused by point mutations in RAD3 have been characterized. rad3-1 and rad3-102 increased the SUP4-o mutation rate by 3.2- and 23-fold, respectively. The substantial difference in the magnitude of the mutation rate increases most likely reflects the different effects of the two-point mutations on Rad3p activity. For example, rad3-1 renders cells very sensitive to UV radiation, whereas rad3-102 slightly decreases UV resistance. In both cases, the mutation rate enhancements were caused primarily by increases in the rates of single base pair events, and the relative fractions of base pair substitutions, deletions, and insertions were similar. The transversion:transition ratio, however, was 3.8-fold greater in rad3-102 (7.2) than in rad3-1 (1.9), and the distributions of substitutions within SUP4-o were quite different (Monteloneet al. 1992; Yanget al. 1996). Interestingly, the transversion:transition ratio for the latter strain was closer to that observed for the rad1 mutant (2.2), which also is highly UV sensitive. Despite some similarities in the spectra for rad3-1 and rad1, there are indications that the mutational mechanisms associated with the two mutators may not be identical. Ty insertions were increased only in the rad1 strain, and rad3-1 but not rad1 has been found to increase the efficiency of base mismatch correction (Yanget al. 1996). If the enhanced efficiency of mismatch repair offsets the rad3-1 mutator effect to some extent, then the Rad3 protein might make a more substantial contribution to genetic stability than that suggested by the magnitude of the rad3-1-mediated increase in the SUP4-o mutation rate alone. The status of mismatch repair in the rad3-102 mutant has not been assessed so far.

RAD6-dependent repair: Members of the RAD6 epistasis group for UV sensitivity are thought to function in postreplication repair or damage tolerance via translesion synthesis (Haynes and Kunz 1981; Friedberg 1988; Friedberget al. 1995; Nelson et al. 1996a,b). The influence of inactivating the RAD6, RAD18, and REV3 genes on the specificity of spontaneous mutagenesis has been determined using single- and double-mutant strains.

The RAD6 gene product is an ubiquitin-conjugating enzyme that is capable of ubiquitinating histones H2A, H2B, and H3 in vitro (Jentschet al. 1987; Haaset al. 1991). The rad6 mutator increased the SUP4-o mutation rate by about fivefold, specifically promoting the two-base pair transitions, G·C → T·A transversions, and Ty insertions. The substitutions were not caused by a failure to correct mismatches that could have given rise to the detected base changes. Not only were the Ty insertions increased in number, but they also occurred at considerably more locations throughout SUP4-o than observed for the wild-type strain, suggesting that protein ubiquitination influences both the rate and target site specificity of Tyretrotransposition. Promotion of the four different types of mutation by the rad6 mutator was REV3 dependent, but the REV3 requirement was not uniform for each of the three mutational classes, with REV3 function necessary for 60–98% of the changes, depending on the class of mutation.

RAD18 encodes a protein that contains DNA- and nucleotide-binding motifs and that has an ATPase activity (Baillyet al. 1997). The rad18 mutator is unique among eukaryotic mutators identified to date in that it displays a highly specific phenotype. Only the rate of G·C → T·A transversions is increased, by about eight-fold. Again, this effect is not caused by a mismatch correction deficiency. Intriguingly, extracts of rad18 cells are defective in nicking supercoiled plasmid DNA containing methylene blue plus light-induced DNA lesions (T. Y.-K. Chow and B. A. Kunz, unpublished results). These lesions include 8-hydroxyguanine (Kasai and Nishimura 1984), which can give rise to G·C → T·A transversions during DNA replication (Shibutaniet al. 1991; Woodet al. 1992; Moriya 1993). Rad18p and Rad6p form a heterodimeric complex with ubiquitin-conjugating, DNA-binding, and ATP hydrolytic activities (Baillyet al. 1997). The mutational specificity data suggest that this complex might be required to limit the production of G·C → T·A transversions by DNA damage, and the function of RAD18 in spontaneous mutagenesis might be to direct the complex to the relevant lesions. Thus, it is not surprising that the production of G·C → T·A transversions by the rad18 mutator also is REV3 dependent, and that the degree of REV3 dependency is similar to that observed for rad6-mediated G·C → T·A transversion.

RAD27 apparently functions in a novel mutation avoidance pathway that minimizes the occurrence of duplications of sequences flanked by short direct repeats. The rad27 mutator greatly increased the rate of these events and promoted complex sequence changes. In contrast to all yeast mutators characterized to date, the rate of single base pair events did not appear to be increased. Because only a small number of mutations were analyzed, however, it remains possible that the rate of single base pair substitutions, deletions, or insertions is enhanced by rad27, but to a much lesser extent than the rates for the events detected.

As mentioned above, the REV3 gene product is a DNA polymerase that is likely to be required for translesion synthesis (Nelsonet al. 1996b). Defects in this gene confer an antimutator phenotype (Cassieret al. 1980; Quahet al. 1980) whose specificity has been characterized. The rates of all classes of spontaneous mutation detected in an isogenic wild-type strain were reduced in the rev3 background, although the major portion of the overall mutation rate decrease was associated primarily with single base pair substitutions, deletions, and insertions. This observation, plus the fact that the rate reductions for other events were small or few mutations were recovered, makes it difficult to say whether the data indicate a very broad specificity for the antimutator. It has been suggested that RAD6 might influence other members of the RAD6 epistasis group via ubiquitination of their gene products (Sunget al. 1990). The manifestation of the rev3 antimutator phenotype in the rad6 background, however, implies that ubiquitination of Rev3p by Rad6p is not necessary for the role of Rev3p in spontaneous mutagenesis.

Inactivation of the DNA damage-inducible gene DDR48 also has been reported to confer an antimutator phenotype (Treger and McEntee 1990). Subsequent analyses of locus reversion and suppression of lys2-1, forward mutation at CAN1 and SUP4-o, and the specificity of spontaneous mutation at SUP4-o in ddr48 strains, however, did not confirm the antimutator phenotype (Rocheet al. 1995b).

Recombinational repair: The rad52 gene has been implicated in the recombinational repair of DNA double-strand breaks and in postreplication repair of UV-induced DNA damage in NER-deficient cells (Friedberg 1988; Friedberget al. 1995). Disruption of RAD52 results in a small increase in the SUP4-o mutation rate, with the mutator primarily affecting substitutions at G·C pairs. Unlike the situation for the rad1, rad6, and rad18 mutators, REV3 is essential for the rad52 mutator effect. These results suggest that RAD52 is involved in the repair of spontaneous DNA damage that, if left unrepaired, can give rise to spontaneous mutation only via REV3-dependent processing. This may be true for other members of the RAD52 epistasis group (Haynes and Kunz 1981) because the Rad51p and Rad52p proteins interact in vivo (Milne and Weaver 1993), and because enhanced reversion of the lys1-1 and his1-7 alleles by the rad51 mutator also exhibits complete REV3 dependence (Quahet al. 1980). The nature of the DNA damage involved is not known. It is not clear, however, how processing a DNA double-strand break might generate primarily transitions and transversions at G·C pairs. Perhaps RAD52 is required for the recombinational repair of daughter-strand gaps that are formed opposite certain types of spontaneous lesions. Inactivation of RAD52 might then leave only REV3-dependent translesion synthesis to fill the gaps.

Base excision repair: APN1 encodes the major apurinic/apyrimidinic (AP) endonuclease involved in base excision repair in S. cerevisiae (Popoffet al. 1990). Consistent with endogenous AP sites being mutagenic lesions, deletion of APN1 resulted in a mutator, and characterization of SUP4-o mutations revealed increased rates mainly for single base pair substitution and Ty insertion. Among the substitutions, the rate of total transversions involving replacement by a G·C pair was increased 10-fold compared to a threefold rate increase for total transversions to an A·T pair. In particular, the rate of A·T → C·G events was 59-fold greater in the apn1 strain. Elimination of the N3-methyladenine glycosylase (MAG) in the apn1 mutant decreased the A·T → C·G transversion rate by 83% relative to the apn1 strain and reduced the rate of total substitutions to that for the wild-type parent. These results suggest that replication past unrepaired AP sites resulting from spontaneous DNA alkylation is largely responsible for the substitution specificity of the apn1 mutator. The data also point to activation of Ty retrotransposition by spontaneously occurring DNA damage.

UNG1 encodes the yeast uracil-DNA-glycosylase (Impellizzeriet al. 1991). Inactivation of UNG1 increased the SUP4-o mutation rate by 10-fold, with an increase in the rate of G·C → A·T transitions accounting for 93% of the mutator effect. This result is consistent with failure to remove uracil resulting from spontaneous deamination of cytosine.

Nucleotide metabolism: There is considerable evidence that regulation of intracellular DNA precursor levels is an important factor in minimizing spontaneous genetic instability (Kunzet al. 1994b). Accordingly, a defect in the gene (DCD1) encoding deoxycytidylate deaminase greatly increased the intracellular dCTP: dTTP ratio and conferred a modest mutator effect. Only the rate of single base pair events, mainly substitution, was increased. As might be expected on the basis of the elevated dCTP pool, ~80% of the substitutions could have involved replacement with C compared to 37% for the wild-type strain. Given the DNA precursor imbalance, a decrease in the proportion of G·C → A·T transitions and a corresponding increase in the fraction of A·T → G·C transitions might have been expected. Although the proportion of G·C A·T events decreased by 50%, the fraction of A·T→ G·C events did not change significantly, and there was an increase in the proportion of G·C → C·G and A·T → C·G transversions. Rather than misincorporation during replicative DNA synthesis, these substitutions are likely to be caused by misincorporation of dCTP during repair of apurinic sites because the dcd1 mutator effect is eliminated by inactivation of APN1 (B. A. Kunz, unpublished results).

Insights from DNA sequence analysis: Characterization of the specificities of yeast mutator and antimutator alleles has revealed that defects in various genes elicit a range of distinct spontaneous mutational spectra. Furthermore, dissimilarities in spectra have been observed for alleles of the same gene (rad3-1 vs. rad3-102) or for genes thought to be involved in the same repair process (rad1 and rad3, rad6, and rad18). Such findings point to unexpected subtleties in the control of genetic stability in S. cerevisiae and presumably in other eukaryotic organisms. Next, we highlight several observations that bear on the mechanistic aspects of spontaneous mutation.

The A rule vs. endogenous abasic sites in eukaryotic cells: The mutational specificity of AP sites was first explored in transfection experiments in which single-stranded bacteriophage DNA was introduced into E. coli, with the specific changes established by analysis of the DNA recovered from the cells (Schaaperet al. 1983; Kunkel 1984). The mutations found were mainly transversions that apparently resulted from the insertion of dAMP residues opposite the AP sites during replication. Analysis of the template specificity of abasic sites during in vitro DNA synthesis by a variety of DNA polymerases indicated a preference for the insertion of dAMP (Sagher and Strauss 1983; Randallet al. 1987; Takeshitaet al. 1987). These and other results gave rise to the so-called “A rule” (noninstructional insertion of adenine by DNA polymerase), which has since been applied to explain the mutational specificity of a variety of DNA lesions that lack obvious coding capacity (Strauss 1991). Analysis of the apn11 mutator effect indicates that unrepaired endogenous AP sites have significant mutagenic potential in eukaryotes but, surprisingly, with a mutational specificity different from that expected on the basis of the A rule (Kunzet al. 1994a). In particular, a large increase in the rate of transversion to G·C base pairs was observed (Table 2), with most of these changes attributed to the loss of DNA purines. Interestingly, the magnitude of the apn11 mutator effect is dramatically reduced by a deletion in DCD1 (B. A. Kunz, unpublished results), which results in a decrease in the intracellular dGTP pool (Kohalmiet al. 1991). Consequently, AP sites in yeast appear to trigger the insertion of dGMP during translesion DNA synthesis. This preference did not arise from a selection bias in the mutational assay system used because, in SUP4-o, the insertion of dAMP opposite purines can be detected at more sites than insertion of dGMP (63 vs. 52 sites, respectively) (Kohalmi and Kunz 1992). Experiments involving the transfection of AP site-containing DNA into eukaryotic cells have also suggested that the mutational specificity of these lesions might be different in nucleated cells (Gentilet al. 1990; Cabral Netoet al. 1992; Kamiyaet al. 1992). More recently, the product of the yeast REV1 gene has been shown to insert cytosine opposite a site-directed AP lesion in vitro (Nelsonet al. 1996a). Whether the Rev1 protein is required for mutagenesis at endogenous AP sites, however, remains to be determined. Nonetheless, the foregoing observations argue that the A rule cannot be readily extrapolated to eukaryotic cells.

Roles of pol δ in DNA replication: DNA pols α, δ, and ε are implicated in the replication of eukaryotic cellular DNA (Bambara and Jessee 1991; Wang 1991; So and Downey 1992; Campbell 1993; Morrison and Sugino 1993; Hübscher and Spadari 1994; Stillman 1994). Polymerase α most likely functions in the initiation of leading strand and Okazaki fragment synthesis, but the precise roles of pols δ and ε are still under debate. The mutational specificity data for yeast DNA pol δ that lacks 3′ → 5′ exonuclease activity (Tables 1 and 2) might be interpreted to suggest that pol δ replicates only one of the two DNA strands at a yeast replication fork, leaving open the possibility that pol ε replicates the other strand. This interpretation of the specificity data for pol δ is based on the following rationale. The type and location of errors that result from replication by DNA polymerases depend in part on the sequence of the template DNA (Kunkel 1992; Izutaet al. 1995). Thus, the pattern of mutations that are produced when a DNA polymerase replicates one strand of a gene might differ from the mutation pattern for replication of the other strand by the same polymerase. If only a single replication fork traverses the gene in question, the sequences of the templates within the gene will depend on the orientation of that gene with respect to the origin from which the replication fork initiates. Should a DNA polymerase replicate, for example, only the leading strand template, then for one orientation, the polymerase would produce a mutational spectrum dictated in part by the sequence of the transcribed strand. For the other orientation, the sequence of that template strand would be changed so that a different spectrum, directed in part by the sequence of the nontranscribed strand, might result. If the DNA polymerase replicates both strands to the same extent, then there might be little difference in the mutational spectrum that is produced upon gene inversion. In either orientation, the polymerase would encounter the sequences of both the transcribed and nontranscribed strands.

A single DNA replication fork does traverse the plasmid-borne SUP4-o gene (Ramachandran 1996). Consequently, the orientation of this gene within the plasmid dictates whether an individual strand is part of the leading or lagging strand template. Inversion of SUP4-o to change the sequence context of these templates alters the magnitude and specificity of the mutator due to elimination of proofreading by DNA pol δ (Tables 1 and 2), but it does not influence mismatch correction in this gene (Ramachandran 1996). Thus, the specificity data are consistent with pol δ encountering a different sequence context upon inversion of SUP4-o, therefore replicating just one template strand at the DNA replication fork. Two other interpretations of these data, however, must be considered. First, the leading and lagging strand replication complexes most likely involve at least some different accessory factors. Dissimilarities (aside from polymerases) in the constitutions of the complexes might lead to different fidelities of DNA replication, and the failure of pol δ to proofread might amplify these differences in accuracy. Consequently, even if pol δ replicated both DNA templates, the pattern of errors produced for the two orientations of SUP4-o might still differ, although it is difficult to say whether this would be true for the overall mutation rate. Second, it is conceivable that replication in vivo by eukaryotic DNA polymerases is dissociated from proofreading by the same enzymes. In this case, the specificity data might suggest that it is the proofreading rather than polymerizing activity of pol δ that operates on only one of the two template strands. If so, then one would have to question why pol δ corrects errors on only one of the two strands it replicates and why a different 3′ → 5′ exonuclease corrects the other strand. Regardless of such conundrums, the differences in mutational specificity upon inversion of SUP4-o also suggest that for any gene, the spectrum and distribution of spontaneous mutations may depend in part on the position and orientation of the gene with respect to the nearest active origin of replication.

Specificity of mismatch repair: The yeast mismatch correction genes MLH1, MSH2, and PMS1 have been shown to participate in a system that recognizes single base pairs and deletion/insertion mispairs (Krameret al. 1989b; Yang 1995; Marsischkyet al. 1996). Comparison of the mutational specificity data for a wild-type strain and isogenic mismatch correction mutants reveals that the rate of single base pair deletion/insertion in SUP4-o is increased much more than the rate of single base pair substitution (Table 1). This observation indicates that deletion mispairs are recognized more efficiently than substitution mispairs by the mismatch correction system, consistent with the situation in E. coli (Schaaper and Dunn 1987, 1991). The rate increase is the same for transitions and transversions (Table 2), however, indicating that there is little difference in the efficiencies with which naturally occurring transitions and transversion mispairs are recognized in yeast. In contrast, E. coli data point to much more efficient recognition of transition than transversion mispairs (Schaaper and Dunn 1987, 1991), suggesting that there may be fundamental differences in base mispair recognition by the general mismatch correction systems in prokaryotes and eukaryotes. Collectively, these observations also indicate that differences in the relative proportions of spontaneous single base pair substitutions and deletions/insertions may not only reflect different frequencies of mutational intermediate formation, but may also reflect differences in the repair of specific mismatches by MLH1/MSH2/PMS1-dependent mismatch correction.

Contribution of translesion synthesis to spontaneous mutagenesis: Evidence has been presented that bacteriophage T4 and E. coli antimutators are very specific and, therefore, may influence only particular pathways for error discrimination (Drake 1993; Fijalkowskaet al. 1993). On the other hand, the rev3 antimutator has relatively broad specificity that affects at least all base pair substitutions as well as single base pair deletions/insertions (Tables 1 and 2). Indeed, the magnitude of the antimutator effect suggests that ≤60% of the spontaneous single base pair substitutions and deletions might be caused by translesion synthesis. If so, the Rev3 DNA polymerase might be involved in processing many different lesions, and, therefore, a substantial fraction of spontaneous mutagenesis in yeast might reflect mechanisms for tolerance rather than error-prone repair of DNA damage. The fact that the rev3 antimutator counteracts several rad mutators, although to different extents (Roche et al. 1994, 1995a), implies that the relevant DNA lesions are normally substrates for error-free repair. Furthermore, certain of the data suggest that if left unrepaired by error-free mechanisms, some endogenous DNA lesions may give rise to mutations almost exclusively via REV3-dependent processing (e.g., see data for RAD52).

Inducibility of Ty transposition by spontaneous DNA damage: Insertion of the Ty retrotransposon into SUP4-o was increased in strains having apn1, rad1, or rad6 deletions (Table 2). APN1 is required for the repair of abasic sites and free radical-induced strand breaks, and defects in RAD1 (NER) and RAD6 (postreplication repair?) sensitize cells to a variety of DNA-damaging agents. Thus, spontaneously occurring DNA lesions might mobilize Ty elements (Kunz et al. 1990, 1994a). Consistent with this hypothesis, treatment with UV or 4-nitroquinoline-1-oxide, mutagens that produce DNA damage subject to NER, has been reported to activate Ty transposition (Bradshaw and McEntee 1989). Spontaneous transposition rates, however, are not increased by mutator alleles of RAD3 or RAD18, which greatly sensitize cells to UV (Kunzet al. 1991; Yanget al. 1996). In addition, the enhancement of Ty transposition by inactivation of RAD6, which encodes a ubiquitin-conjugating enzyme (Jentschet al. 1987), might not necessarily be related to defective DNA repair. Picologlou et al. (1990) suggested that either failure to ubiquitinate Ty proteins or altered chromatin structure in rad6 strains, resulting in increased access of transposition complexes to DNA, might be responsible. Such effects also might account for the broadened target site specificity for Ty insertion into SUP4-o in the rad6 background. On the other hand, Ty target site specificity also is expanded by deleting REV3 (Rocheet al. 1994). Although this again points to a role for spontaneous DNA damage in Ty transposition, it is clear that the situation is not straightforward.


We thank our collaborators and the many investigators who provided the plasmids that were used to make the strains described in this article. Work from B.A.K.'s laboratory was supported by the Natural Sciences and Engineering Research Council of Canada, the Medical Research Council of Canada, and the Australian Research Council.


View Abstract