Genetic Factors Affecting the Impact of DNA Polymerase Proofreading Activity on Mutation Avoidance in Yeast

Corresponding author: Michael A. Resnick, Laboratory of Molecular Genetics, National Institute of Environmental Health Sciences, 111 TW Alexander Dr., P.O. Box 12233, Research Triangle Park, NC 27709., resnick{at}niehs.nih.gov (E-mail)

Communicating editor: L. S. SYMINGTON

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Base selectivity, proofreading, and postreplication mismatch repair are important for replication fidelity. Because proofreading plays an important role in error correction, we have investigated factors that influence its impact in the yeast Saccharomyces cerevisiae. We have utilized a sensitive mutation detection system based on homonucleotide runs of 4 to 14 bases to examine the impact of DNA polymerase proofreading on mutation avoidance. The contribution of DNA polymerase proofreading on error avoidance was found to be similar to that of DNA polymerase proofreading in short homonucleotide runs (A₄ and A₅) but much greater than the contribution of DNA polymerase proofreading in longer runs. We have identified an intraprotein interaction affecting mutation prevention that results from mutations in the replication and the proofreading regions, resulting in an antimutator phenotype relative to a proofreading defect. Finally, a diploid strain with a defect in DNA polymerase proofreading exhibits a higher mutation rate than a haploid strain. We suggest that in the diploid population of proofreading defective cells there exists a transiently hypermutable fraction that would be inviable if cells were haploids.

IN the bacterium Escherichia coli, the accuracy of replication is controlled by at least three steps, acting serially to ensure high fidelity: base selection, exonucleolytic proofreading, and postreplication mismatch repair (MMR). The main replicative polymerase in E. coli is polymerase III holoenzyme. The Pol III subunit, encoded by the dnaE gene, is the catalytic subunit that is responsible for base selectivity. The proofreading exonuclease (subunit ) is encoded by the dnaQ gene. The , , and subunits (the subunit has an unknown function) are tightly bound together to form the polymerase III core (MCHENRY and CROW 1979 ). Replication errors are recognized and corrected by the multiprotein mutHLS mismatch repair system. The base selectivity, proofreading, and mismatch repair systems reduce the errors by 10⁵-, 10²-, and 10³-fold, respectively. The combined efficiency of these three steps reduces replication errors to 10^-10 errors per replicated nucleotide (SCHAAPER 1988 , SCHAAPER 1993 ).

There are three DNA polymerases required for chromosomal replication in eukaryotes, polymerases (Pol) , , and , which are encoded by the POL1, POL3, and POL2 genes, respectively. Polymerase is responsible for synthesis of primers for Okazaki fragments in the lagging strand, and the Pol and Pol have been proposed for lagging and leading DNA strand replication, although their relative roles have not been established (SUGINO 1995 ). Unlike Pol , which has only a polymerase catalytic function, the Pol and Pol also have a 3' 5' proofreading exonuclease activity in their N-terminal region (KESTI and SYVAOJA 1991 ; MORRISON et al. 1991 ; SIMON et al. 1991 ). In the yeast S. cerevisiae, point mutations (pol3-01 and pol2-4) in the exonuclease-conserved domains eliminate proofreading activity of Pol and Pol , respectively, and result in a frameshift and a base substitution mutator phenotype (MORRISON et al. 1991 , MORRISON et al. 1993 ; SHCHERBAKOVA and PAVLOV 1996 ). As in E. coli, replication errors are checked by an MMR system, which is composed of several proteins homologous to E. coli MutS and MutL (MODRICH and LAHUE 1996 ). Combined defects in proofreading and MMR can lead to mutation synergism or cell death in haploid strains. The lethal effect is possibly due to excessive mutation rates (MORRISON et al. 1993 ; MORRISON and SUGINO 1994 ; TRAN et al. 1997B , TRAN et al. 1999 ).

While the interaction between proofreading and MMR in mutation prevention is well established, little is known about the potential interaction between proofreading and other DNA polymerase activities. On the basis of the model proposed by SCHAAPER 1993 , polymerase, proofreading, and MMR exert their functions independently. Data supporting this model are provided by the E. coli dnaE antimutator alleles, which are antimutators either alone or in combination with defective proofreading (FIJALKOWSKA and SCHAAPER 1995 ) or defective MMR (FIJALKOWSKA et al. 1993 ). The dnaQ926 proofreading deficiency appears to cause lethality as a result of the loss of proofreading and subsequent saturation of DNA MMR (error catastrophe). However, dnaQ926 strains are viable if they carry a dnaE antimutator allele or a multicopy plasmid carrying the E. coli mutL gene (FIJALKOWSKA and SCHAAPER 1996 ). As the polymerase () and proofreading () subunits are bound tightly together (MCHENRY and CROW 1979 ), a mutation in the subunit could affect the editing ability of the DNA polymerase holoenzyme complex. The dnaE173 mutation in the polymerase subunit leads to a 1,000- to 10,000-fold increase in mutation rate (MAKI et al. 1991 ), and it was proposed that this allele affects the proper interaction between and subunits, resulting in a defect in the proofreading capacity of the Pol III holoenzyme complex.

The interactions between the proofreading and polymerase regions and their potential role in mutation prevention are not well understood in eukaryotes. The yeast pol2-18 mutation, located in the polymerase region of Pol , leads to a weak mutator phenotype as well as temperature sensitivity (ARAKI et al. 1992 ; SHCHERBAKOVA et al. 1996 ). In combination with the Pol proofreading mutation (pol2-4; MORRISON et al. 1991 ), pol2-18 exhibits mutation frequency synergy (SHCHERBAKOVA et al. 1996 ) consistent with the model that the polymerase and proofreading activities act in series. Moreover, pol2-18 is an antimutator with respect to mutagenesis induced by the base analog N⁶-hydroxylaminopurine (HAP), which is proposed to occur via a replicative misincorporation mechanism (SHCHERBAKOVA and PAVLOV 1996 ). The authors also found pol2-18 to be a HAP-specific antimutator in combination with the proofreading defect pol2-4. There have been no reports concerning interaction between polymerase and proofreading regions within DNA Pol .

According to the replication slippage model of STREISINGER et al. 1966 , the incidence of frameshift mutations is expected to increase with the length of a homonucleotide run. Using a sensitive system for detecting mutations in long homonucleotide runs, we previously found that both MMR and Pol proofreading provide efficient mutation prevention in short runs, but that only MMR is capable of preventing frameshift mutations in runs 8 nucleotides (TRAN et al. 1997B ). However, the impact of the Pol proofreading activity in correcting errors in homonucleotide runs could not be investigated, because the combination of Pol proofreading (pol3-01) and MMR (msh2 or pms1) defects is lethal in a haploid strain (MORRISON et al. 1993 ; TRAN et al. 1999 ). Because homozygous pol3-01 pms1 or pol3-01 msh2 diploids are viable (MORRISON et al. 1993 ; TRAN et al. 1999 ) the efficiency of Pol proofreading and comparison with Pol proofreading can be addressed in diploids (see below).

In this study we examine the impact of Pol proofreading on mutation avoidance. Specifically, we have investigated (i) the interaction between mutations in the polymerase and the proofreading regions of Pol , (ii) relative contributions of Pol proofreading and postreplication MMR to mutation avoidance, (iii) the relative effectiveness of Pol and Pol proofreading on homonucleotide run templates, and (iv) the mutational consequences of a Pol proofreading defect in haploid vs. diploid strains. Our results show that a Pol proofreading defect can be influenced by a change elsewhere in the protein. Specifically, the temperature-sensitive pol3-t mutation, which is located in the replicative region (TRAN et al. 1997A ) and exhibits a deletion mutation phenotype, acts as an antimutator when combined with a Pol proofreading defect. Overall it appears that Pol proofreading has a much greater impact on mutation avoidance than Pol proofreading and has a larger contribution to error avoidance in homonucleotide runs. The mutator effect of Pol proofreading deficiency is further increased in diploid strains compared to the corresponding haploid strains. We suggest that a transiently hypermutable fraction exists that is revealed when diploid cells are defective in Pol proofreading.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

General genetic and molecular methods:
Yeast standard media (SHERMAN et al. 1986 ) and YPD containing G418 (WACH et al. 1994 ) were used. Yeast cells were grown at 30°, although strains with the temperature-sensitive polymerase mutation pol3-t were grown at 25° for mutation studies. Yeast transformations were performed according to GIETZ and SCHIESTL 1991 . The preparation of bacterial media and general molecular methods has been described (SAMBROOK et al. 1989 ).

Strains and plasmids:
A series of isogenic strains were constructed from the original CG379, MAT ade5-1 his7-2 leu2-3,112 trp1-289 ura3-52 (MORRISON et al. 1991 ) and from its pol3-01 or pol3-t or pol3-t, 01 derivatives. These strains contain modified InsD (Figure 1A and see below) or InsE inserts in the chromosomal LYS2 gene, where the A₄ run was changed to A₅, A₇, A₈, A₁₀, A₁₂, and A₁₄ (Figure 1B; TRAN et al. 1997B ). The mutations pol3-01 and pol3-t are point mutations in the exonuclease and the polymerase domains of the POL3 gene, respectively (MORRISON et al. 1993 ; TRAN et al. 1997A ; and see Figure 2). The mutation in pol3-01 alters the aspartate and glutamate residues in the essential exonuclease motif, FDIEC, at amino acids 320–324; the D321A and E323A double mutation in pol3-01 eliminates the proofreading activity of the polymerase (MORRISON et al. 1993 ). The pol3-t allele is due to a single base substitution in the context GAAGAC to GAAAAC leading to loss of the MboII site. The pol3-t allele encodes a D641N substitution (TRAN et al. 1997A ) in the vicinity of the conserved polymerase motif VI of DNA Pol (BOULET et al. 1989 ). [In the revised nucleotide sequence of POL3 (MORRISON and SUGINO 1992 ) this amino acid is at position 643 of Pol (Figure 2)]. The pol3-t allele produces a mutant polymerase that induces replication slippage between distant short repeats as well as frameshift mutations in homopolymeric runs (TRAN et al. 1995 , TRAN et al. 1996 ). The pol3-01 mutation causes a mutator phenotype due to increased base substitution and frameshift mutation rates (MORRISON et al. 1993 ). Strains with the double mutation pol3-t, 01 were constructed from pol3-01 strains. The pol3-t allele was introduced to the chromosomal pol3-01 allele using plasmid p171 and a gene replacement technique described previously (KOKOSKA et al. 1998 ). The presence of both pol3-01 and pol3-t mutations was confirmed by restriction digest of appropriate PCR products (MORRISON et al. 1993 ; KOKOSKA et al. 1998 ). These strains also contain modified InsE inserts in the chromosomal LYS2 gene, where the A₄ run was changed to A₅, A₇, A₈, A₁₀, A₁₂, and A₁₄ homonucleotide runs (TRAN et al. 1997B ). Strain 1036 lys2-BX: (MATa ade2-1 arg4-8 leu2-3,112 lys2-BX thr1-4 trp1-1 ura3-52 cup1-1) contains a deletion of the BamHI-XhoI region covering the InsE inserts in the LYS2 gene. The pol3-01 mutation was introduced into this strain using YIpAM26 plasmid as described previously (MORRISON et al. 1993 ). The MSH2, RAD27, and EXO1 genes were disrupted in strains using the gene disruption technique as described below.

View larger version (17K): In this window In a new window Download PPT slide   Figure 1. (A) The InsD insert described previously (TRAN et al. 1995 ) was modified to InsLD by site-directed mutagenesis to generate three stop codons (). The insert is flanked by two direct repeats (underlined); one belongs to the insert and the other belongs to the LYS2 gene. The insert causes a +1-nt shift of the LYS2 reading frame, generating a TGA stop codon (). Three additional stop codons in insLD block all three possible open reading frames so reversion can occur only via extended deletions that restore the original LYS2 reading frame. (B) Series of homonucleotide runs in the lys2::InsE insert in the LYS2 gene described previously (TRAN et al. 1997B ). The InsE insert shifts the LYS2 reading frame, generating a TGA stop codon (). The A4 homonucleotide run hot spot for -1 deletions (****) was increased up to A14. The hot spot for -1-nt deletions in the pol3-01 haploids and diploids is denoted (**).

View larger version (13K): In this window In a new window Download PPT slide   Figure 2. Location of the proofreading (pol3-01) and polymerase (pol3-t) mutations in polymerase of S. cerevisiae. The EXO, POL, and ZnF boxes identify exonuclease, polymerase, and Zn-finger conserved motifs, respectively, in the polymerase alignment (HINDGES and HUBSCHER 1997 ). The proofreading mutation pol3-01 is due to a change from FDIEC to FAIAC in the EXO1 domain at positions 320–324 (MORRISON et al. 1993 ). The pol3-t mutation is due to a nucleotide change G to A that eliminates the MboII recognition site, producing amino acid substitution D643N in the vicinity of the polymerase motif VI of S. cerevisiae polymerase . [Note that previously the pol3-t mutation was reported as a D641N substitution (TRAN et al. 1997A ), where the amino acid position was determined on the basis of the initially published sequence (BOULET et al. 1989 )]. The new D643N amino acid substitution position is based on a revised POL3 gene sequence [EMBL accession no. X61920 (MORRISON and SUGINO 1992 ) as well as the Stanford University S. cerevisiae database at http://genome-www2.stanford.edu/cgi-bin/SGD/getSeq?map=pmap=YDL102W=0=0=]. S.c., Saccharomyces cerevisiae; S.p., Schizosaccharomyces pombe; C.a., Candida albicans.

lys2::InsLD construction:
The InsD insert in the plasmid p92 (TRAN et al. 1995 ) was modified to InsLD (locked InsD) containing three additional stop codons (Figure 1A). Changing InsD to InsLD on the plasmid was done using site-directed, double-stranded mutagenesis (Chameleon mutagenesis kit; Stratagene, La Jolla, CA). Double-stranded p92 plasmid was annealed with a mix of a mutation oligonucleotide 5'-CTGACTCTTATACACTAGTAGCTACCTGACGTGGGCAAAC-3' (underlined nucleotides indicate differences of InsLD from InsD) and the oligonucleotide 5'-GTAATTGCAAGTGGATATCTGAACCAGTCC-3' as a helper primer (underlined nucleotides change the unique EcoRI site in p92 to an EcoRV site). Replacement of the chromosomal wild-type LYS2 gene with the lys2::InsLD allele was done by two-step gene replacement using a HpaI-digested integrative version of the plasmid (without the ClaI-ClaI ARS-CEN cassette).

Gene replacement and disruption:
The following genes were disrupted: MSH2, RAD27, and EXO1. For MSH2 disruption we used a SacI-PstI msh2::LEU2 fragment from p203 (TRAN et al. 1997A ). The RAD27/RTH1 gene was disrupted using an EcoRI-SalI rth1::URA3 fragment (JOHNSON et al. 1995 ). The entire open reading frame of EXO1 was deleted using PCR disruption with the kanMX module (WACH et al. 1994 ) and primers described below. The lowercase type indicates nucleotide sequences that belong to the kanMX cassette; DNA sequences belonging to the genes are written in uppercase. For the EXO1 gene we amplified the kanMX cassette using EXO1-kanMX-3': 5'-TTGGCTTGACTTAGTAGTTTCGATGTCCCTTTTCTTACTTatcgatgaattcgagctcg-3' and EXO1-kanMX-5': 5'-AGGTATGAAGGAGAAGTGTTAGCCATTGATGGCTATGCATcgtacgctgcaggtcgac-3' and verified by PCR using the following primers: EXO1-test-3, 5'-ATTGGGAAAGCAAGGAGATAG-3' and EXO1-test-5, 5'-TCTTCTTCCTCAGTTAAAGC-3'. The RAD27/RTH1 gene disruption transformants were identified by MMS sensitivity (340 µl MMS/liter of YPD), induction of illegitimate mating, and by PCR using RTH1-1F, 5'-GGACACCGGAAGAAAAAAT-3' and primer RTH1-2R, 5'-AACTTCAGGGTCAAGAAACAGCA-3'.

Construction of msh2 pol3-01 diploid strains:
The combination of pol3-01 mutation with null mutations in MMR genes PMS1 or MSH2 is lethal in haploids, but diploid strains are viable (MORRISON et al. 1993 ; TRAN et al. 1999 ). To study the impact of the DNA Pol proofreading in mutation prevention we constructed a series of isogenic diploid strains msh2 pol3-01 containing homonucleotide runs of various sizes in the lys2::InsE insert. The derivative strain 1036 lys2-BX pol3-01 was obtained from 1036 lys2-BX using YIpAM26 (MORRISON et al. 1993 ). The 1036 MATa lys2-BX pol3-01 strain was transformed with the replicative plasmid pBL304 (MORRISON and SUGINO 1994 ) carrying the POL3 and the URA3 genes. Then the MSH2 gene in the transformant was deleted using a PstI-SacI msh2::LEU2 fragment from plasmid p203 (TRAN et al. 1997A ). The resulting strain 1036 lys2-BX pol3-01 msh2 pBL304 was mated with a series of MAT pol3-01 strains isogenic to CG379, containing homonucleotide runs (A₄, A₅, A₇, A₈, A₁₀, A₁₂, and A₁₄) in the lys2::InsE insert. After loss of plasmid pBL304 on 5-fluoroorotic acid (5-FOA) medium, the second copy of the chromosomal MSH2 gene in the diploid was deleted using the SpeI-SpeI fragment msh2::URA3 from plasmid pII2-Tn10-LUK7-7 (REENAN and KOLODNER 1992 ). Deletion of MSH2 was verified by PCR as described previously (TRAN et al. 1996 ). The resulting diploid strain is pol3-01/pol3-01 msh2::LEU2/msh2::URA3 lys2::InsE-A_n/lys2.

Construction of isogenic homozygous diploid strains:
To investigate the impact of a ploidy on mutation rates we constructed a series of isogenic homozygous diploid strains. Haploid leu2 strains derived from CG379 were transformed with plasmid YEpHO (gift from Dr. Chernoff) carrying the LEU2 and the HO endonuclease genes. The HO endonuclease induces mating type switching in haploid strains (MATa to MAT or vice versa). Haploid strains with the opposite mating types then form MATa/MAT homozygous diploid strains. Transformants with YEpHO were grown on YPD media to allow loss of the plasmid. Single Leu^- clones were isolated. Diploid clones were identified as being nonmaters with MATa his3 or MAT his3 testers as well as giving a low forward mutation rate to Can^r because of the presence of two CAN1 gene copies. For strains where the LEU2 marker could not be used, we utilized plasmid pGHO-TRP1 (BENNETT et al. 1993 ) containing the TRP1 and the HO gene under control of the GAL1-10 promoter. Mating type switching was induced by incubation in galactose media for 8 hr and diploid clones were identified as described above.

Mutation analysis:
Mutation rates were determined in at least 12 independent cultures by a fluctuation test using the median method described by LEA and COULSON 1949 . The nature of the Lys⁺ revertants was identified by sequencing the reversion window of the lys2::InsE insert, as described previously (TRAN et al. 1995 ). Sequencing was performed using ABI 373 automated sequencer.

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Experimental systems:
To investigate genetic controls of extended deletions and small (-1 nt, +1 nt) frameshift mutations, we used the previously described InsD (31-bp) and InsE (61-bp) insertion mutations in the chromosomal LYS2 gene (TRAN et al. 1995 ). These insertions at a common position in the LYS2 gene are flanked by 7- and 6-bp direct repeats, respectively. (One direct repeat belongs to the LYS2 gene, and another belongs to the insert.) Both inserts shift the reading frame of the LYS2 gene and generate a TGA stop codon (Figure 1). Reversions can arise by either extended or small (-1- or +2-bp) deletions/insertions that restore the original LYS2 gene reading frame (TRAN et al. 1995 , TRAN et al. 1996 ). Extended deletions appear to occur via replication slippage (TRAN et al. 1995 ). We generated another sensitive mutation detection insert, InsLD, which is derived from InsD by site-directed mutagenesis by adding three stop codons in all three possible reading frames (Figure 1A). This InsLD mutation detection system allows the identification of only extended deletions, because reversion is observed only if all three stop codons are removed.

To investigate the mutational consequences of a DNA Pol proofreading defect (pol3-01), a DNA Pol polymerase mutation (pol3-t), and a defect in MMR, we used a series of homonucleotide runs in the InsE insert described previously (TRAN et al. 1997B ). The longest homonucleotide run, A₄ in InsE, was changed to A₇, A₁₀, and A₁₄ resulting in +1-bp frameshifts and to A₅, A₈, and A₁₂ leading to -1-bp frameshifts (Figure 1B). To evaluate the nature of the mutations in the homonucleotide runs, a sample of Lys⁺ revertants from the +1 and -1 frameshift allele detection systems was sequenced in the reversion window region. The mutator phenotype was also examined using a his7-2 reversion assay. While the nature of the his7-2 mutation is not known, it has been used in several studies investigating proofreading and MMR defects. It exhibits high reversion rates with either MMR or proofreading mutators (MORRISON et al. 1991 , MORRISON et al. 1993 ; MORRISON and SUGINO 1994 ).

The instability of the homonucleotide runs and the deletion of the insertion InsLD were examined in strains with the proofreading (pol3-01) defect or with the putative polymerase defect (pol3-t) or with defects in both functions pol3-t, 01. (Construction of the double mutation pol3-t, 01 is described in MATERIALS AND METHODS.) The locations of the pol3-01 and pol3-t mutations in DNA Pol are shown in Figure 2. Since only in the absence of MMR can errors generated during replication be accurately measured, the role of DNA Pol proofreading in mutation prevention in homonucleotide runs was examined in MMR^- (msh2) strains. Because the double mutation pol3-01 msh2 is lethal for haploid strains (TRAN et al. 1999 ), the impact of the proofreading defect pol3-01 on mutation rates was investigated in diploid strains lacking postreplication MMR.

Effect of DNA Pol proofreading and polymerase mutations on extended deletions:
Previously, we described the replication defective mutant of DNA Pol , pol3-t, whose defect is due to a mutation in the vicinity of the conserved polymerase VI motif (Figure 2). At the permissive temperature, this mutant polymerase exhibits increased replication slippage, resulting in both extended deletions and -1- and +2-bp frameshifts (TRAN et al. 1995 , TRAN et al. 1996 ). As shown in Table 1, the pol3-t strain has a >50-fold higher rate of reversion of the InsLD mutation than the isogenic wild type and has only a modest effect (up to 11-fold increase) on frameshift mutations in homonucleotide runs (Table 1 and Table 2). The InsLD revertants were due entirely to extended deletions. In contrast, there was no increase in extended deletions in the pol3-01 proofreading mutant, although it is a strong mutator for CAN1 forward mutations (Table 1). All but three reversions of the lys2::InsLD allele obtained in the proofreading mutant pol3-01 were due to precise deletions between the distant short repeats identified in Figure 1A. (Those three reversions were due to point mutations that eliminated only one stop codon. Because the original reading frame of the LYS2 gene was not restored in these revertants, the revertants might be due to secondary suppressor mutations that suppress a second stop codon in InsLD.)

View this table: In this window In a new window   Table 1. Effect of the DNA Pol proofreading pol3-01 and the polymerase pol3-t mutations on reversion of the lys2::InsLD mutation by replication slippage between distant short repeats

The double mutant pol3-t, 01 provided the opportunity to investigate possible interactions between the replication and proofreading defects of Pol in the same polypeptide. The double mutant pol3-t, 01 demonstrates growth properties similar to the pol3-t mutant and it has the same deletion rate as a pol3-t mutant (Table 1). However, the high canavanine resistance mutation rate of the pol3-01 mutant is reduced (Table 1). On the basis of this and results described below, we suggest that pol3-t is an antimutator when combined with the pol3-01 mutator allele.

Mutator and antimutator phenotype of the polymerase mutation pol3-t:
In the E. coli Pol III holoenzyme, proofreading and polymerase activities are encoded by the dnaQ and dnaE genes, respectively. These two activities presumably act in series (SCHAAPER 1993 ); therefore, defects in both activities may result in mutator synergism relative to the single mutator phenotypes. We investigated the possible interaction between the proofreading and replication activities in yeast using the single and double mutants described in the previous section.

In an MMR⁺ background pol3-t has only a modest influence on the frameshift mutation rate in runs of various lengths (Table 2 and TRAN et al. 1995 , TRAN et al. 1996 ). However, on the basis of previous results (TRAN et al. 1996 ) and the data in Table 2, the pol3-t mutant generates a large number of premutational changes that are efficiently corrected by MMR. Similar to the observation with a Pol proofreading mutant (TRAN et al. 1997B ), the impact of the pol3-t mutator decreases with increased length of the homonucleotide run. For -1 frameshifts, the mutation rate increases of pol3-t msh2 mutants in the A₄ and A₁₄ runs are, respectively, 7- and 3-fold those of msh2 mutants alone; similarly, for +1-bp frameshifts, the mutation rate increases of pd3-t msh2 mutants in the A₅ and A₁₂ runs are, respectively, 8.6- and 2.2-fold those of msh2 mutants alone (Table 2).

As described previously (TRAN et al. 1997B ) and in Table 1 and Table 2, the proofreading (pol3-01) and the polymerase (pol3-t) mutations increase the Can^r forward mutation rate and the mutation rate in homonucleotide runs. Surprisingly, the pol3-t, 01 double mutant does not exhibit a synergistic mutator phenotype. Instead of the expected mutation enhancement, there is a 2- to 33-fold decrease in the mutation rate for various homonucleotide runs in the double mutant, relative to strains defective only in proofreading (Table 2). There also appears to be an antimutator effect on the Can^r forward mutation rate of pol3-t in the pol3-01 background (Table 1).

Impact of DNA Pol proofreading defect on mutation in homonucleotide runs:
Previously we examined the interaction between mutations in the Pol proofreading function and MMR on instability of long homonucleotide runs. Both Pol proofreading and MMR are efficient in preventing errors in short runs (A₄ and A₅), while only MMR prevents frameshift mutations in runs of 8 nucleotides (TRAN et al. 1997B ). It was not possible to investigate the interaction between Pol proofreading defect (pol3-01) and MMR in the haploid double mutant pol3-01 msh2, because it is inviable (TRAN et al. 1999 ). Here we analyze the interaction between Pol proofreading and MMR on homonucleotide runs in diploid isogenic strains. The diploid strains have one copy of the chromosomal LYS2 gene deleted in the BamHI-XhoI region covering the InsE insert and the other copy of the LYS2 gene contains homonucleotide runs of various lengths in the InsE (see MATERIALS AND METHODS).

Similar to observations in haploid strains, the msh2 and pol3-01 diploid strains also generally exhibit an exponential increase in the mutation rate with increased length of homonucleotide run (Table 2 and Table 3). Increasing the homonucleotide run length greatly increased the incidence of -1-nt and +1-nt mutations for all diploid strains tested (Table 3 and Table 4). For wild-type strains, the proportion of reversions that were specifically due to deletions or additions in the homonucleotide runs increased from 18% (6/32) for the A₄ and 20% (9/44) for the A₅ to almost 100% for the A₁₀, A₁₂, and A₁₄ homonucleotide runs (Table 3). The wild-type and msh2 strains exhibited comparable homonucleotide run mutation rates in both haploid (TRAN et al. 1997B ) and diploid strains.

Unlike the wild-type and msh2 strains, the pol3-01 diploid exhibits an important difference from the haploid. As shown in Table 5, the diploid has a much higher mutation rate (3- to 19-fold) as compared with the haploid strain. There are also differences in the mutation spectra between pol3-01 haploid and diploid strains. Among 37 Lys⁺ revertants of the lys2::InsE-A₄ allele in the haploid pol3-01 strain, only 7 were at the GG hotspot position (indicated in Figure 1B) and 4 were in the A₄ run. However, in the pol3-01/pol3-01 diploid strain 37 of 42 Lys⁺ revertants were at the GG hot spot and none were in the A₄ run.

As MMR effectively corrects errors generated during replication, replication infidelity can be measured adequately only in the absence of the postreplication MMR. Similar to previous reports for DNA Pol proofreading (TRAN et al. 1997B ), there is a synergistic interaction between the msh2 and DNA Pol proofreading deficiencies (Table 3 and Table 4). However, there are differences in the mutation spectra between pol2-4 msh2 and pol3-01 msh2 strains. Among 39 revertants examined from the pol3-01 msh2 lys2::InsE-A₄/pol3-01 msh2 lys2 diploid strain, 14 were at the GG hot spot and 6 revertants occurred in the A₄ run. This differs from the spectrum of the haploid pol2-4 msh2 lys2::InsE-A₄ strain where all but 3 among 29 revertants were in the A₄ run (TRAN et al. 1997B ). [We did not investigate the pol2-4 msh2/pol2-4 msh2 mutation spectrum because the pol2-4/pol2-4 and msh2/msh2 diploids did not exhibit any differences in mutation rates in comparison with pol2-4 and msh2 haploids, respectively (Table 6).] It is interesting that for the msh2/msh2 diploid strain, there were no revertants because of changes at the GG hot spot and 12 among 18 Lys⁺ revertants occurred in the A₄ run. These results are similar to those found for the msh2 haploid, in which 20 of the 30 Lys⁺ revertants arose in the A₄ run (TRAN et al. 1997B ). The high mutation frequency at the GG hot spot in the pol3-01/pol3-01 diploid strain is likely dependent on the DNA sequence context because mutation hot spots are not observed at two other GGG sites within the reversion window.

In the MMR^- background, the efficiency of DNA Pol proofreading for -1-nucleotide (nt) frameshift mutations is sharply reduced as the homonucleotide run length is increased from A₇ to A₁₀. The A₄ and A₇ mutation rates in the double mutant pol3-01 msh2/pol3-01 msh2 are more than 63-fold higher than those in the msh2/msh2 strain, but less than 3-fold higher for the A₁₀ and A₁₄ runs (Table 4; last column). For +1-nt insertions the efficiency of Pol proofreading is also decreased with increasing length of the homonucleotide run. The mutation rate in the A₅ run in msh2 pol3-01/msh2 pol3-01 is 70-fold higher than in the single msh2/msh2 mutant; in contrast, when the mutation rate in the A₁₂ run in the same two strains is compared, the difference is 9-fold (Table 4; last column). Thus, despite the reduced proofreading efficiency of Pol within longer homonucleotide runs, DNA Pol proofreading is still active during replication of the A₁₂ run, strongly affecting the mutation rate for +1-nt insertions. This is different from the influence of Pol proofreading, which is eliminated as the run length is extended from A₅ to A₈. The mutation rate in msh2 pol2-4 is 321-fold higher than in msh2 for the A₅ run, but there are only 2.3- and 1.6-fold increases for the A₈ and A₁₂ runs, respectively (see Table 2 in TRAN et al. 1997B , TRAN et al. 1999 ).

Differences in hypermutability of the Pol proofreading mutant between diploid and haploid strains:
We found that the diploid pol3-01 mutant exhibits a higher mutation rate than the haploid pol3-01 strain. This ploidy effect appeared specific to pol3-01 and was not observed in wild-type, msh2, or exo1 mutants. As shown in Table 5, the differences in rate were observed for homonucleotide runs of different lengths. No differences were found between haploid and diploid wild-type, msh2, or exo1 mutants.

Because the diploid strains were constructed by mating strains from different backgrounds (Table 5), some of the differences between haploid and diploid strains could have arisen from strain background variation. Therefore, we constructed a series of homozygous diploid strains by HO endonuclease-induced autodiploidization of haploid strains (see MATERIALS AND METHODS). Spontaneous mutation reversion rates were measured for the his7-2, lys2::InsE-A₁₂, and lys2::InsE-A₁₄ mutants. (Note that because the diploid strains have two alleles, the mutation rate in the diploid strains might be expected to be twofold higher than in haploids.) As shown in Table 6, there is no significant difference in the haploid and diploid mutation rates for the three loci (his7-2, lys2::InsE-A₁₂, or lys2::InsE-A₁₄) in wild-type strains and in mutator strains pol2-4, msh2, exo1, and rad27. However, the pol3-01 diploid strain exhibits an 8.5- to 48-fold higher mutation rate than the haploid strain. The EXO1 and RAD27 genes code for 5' to 3' exonucleases. The former is implicated in the excision of mismatches (TRAN et al. 1999 ) and the latter is required for the removal of flaps during lagging strand replication.

In a haploid strain, a pol3-01 mutation combined with the Pol proofreading defect pol2-4 (or with either mutation exo1, msh2, or pms1) is lethal, while the equivalent diploid strain is viable (MORRISON et al. 1993 ; MORRISON and SUGINO 1994 ; TRAN et al. 1999 ). The inviability of these mutant combinations in the haploid is considered to be due to the accumulation of excessive mutation. It is possible that the increased mutation rate in diploid pol3-01 mutants is due to the existence of a highly mutable fraction of cells that would be inviable if the cells were haploid (see DISCUSSION). If this is true, then the coincidence of reversion of two independent mutations should be higher than expected on the basis of the single reversion rates. In the pol3-01 diploid, the reversion rate for lys2::InsE-A₁₂ to Lys⁺ is 1.58 x 10^-4 [95% confidence interval (CI) of the mutation rate: 0.55 x 10^-4–2.31 x 10^-4] and for his7-2, it is 5.9 x 10^-6 (CI: 2.03 x 10^-6–17.7 x 10^-6) to His⁺. If reversion to His⁺ and to Lys⁺ were independent events, then the expected rate of simultaneous reversion to Lys⁺ His⁺ would be the product of the two reversion rates or 9.3 x 10^-10 (CI: 1.1 x 10^-10–4.1 x 10^-9). Instead, we observed that the rate of appearance of double mutants was 3.39 x 10^-7 (CI: 1.14 x 10^-7–4.37 x 10^-7), or 365-fold higher than the expected reversion rate.

If variation of MMR protein expression or activity is the source of a hypermutable cell fraction in the population, then loss of MMR should result in a homogeneously mutable cell population. We therefore measured the reversion rate to Lys⁺, to His⁺, and to His⁺ Lys⁺ in the homozygous strain pol3-01 msh2/pol3-01 msh2. The Lys⁺ and His⁺ reversion rates are 1.25 x 10^-3 (CI: 0.7 x 10^-3–2.0 x 10^-3) and 2.7 x 10^-4 (CI: 1.8 x 10^-4–5.9 x 10^-4), respectively. The double His⁺ Lys⁺ reversion rate is 9.7 x 10^-6 (5.0 x 10^-6–16.4 x 10^-6), which is only 29-fold higher than the expected reversion rate of 3.4 x 10^-7 (CI: 1.3 x 10^-7–1.2 x 10^-6) as compared with a 365-fold increase in the MSH2 strain. Thus, while elimination of MMR reduced much of the proposed mutational heterogeneity, these results suggest that MMR is not the sole source of heterogeneity in the pol3-01 cell population.

If there is a hypermutable cell fraction in the population, the effect may be only transient. We therefore examined 48 independent Lys⁺ revertants from the diploid strain pol3-01 lys2::InsE-A₁₂ his7-2 for increased mutability. On the basis of a replica-plating assay for His⁺ reversion, these isolates and the original strain had comparable His⁺ reversion rates indicating that the proposed hypermutability is transient.

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Replication fidelity is dependent on many factors that include base selectivity, proofreading, and postreplication MMR as well as the DNA sequence being replicated. We have developed systems to address the impact of MMR and DNA Pol proofreading and polymerase defects during the replication of a variety of DNA templates that include at-risk motifs (i.e., ARMs; GORDENIN and RESNICK 1998 ), which are prone to the generation of errors. Because replication and postreplication MMR complexes are highly conserved from yeast to human, and ARMs such as homonucleotide runs are found in the genomes of all organisms, the present data provide insight into mutation avoidance mechanisms during replication in higher eukaryotes. Because of its importance to replication fidelity and its possible role in MMR (LONGLEY et al. 1997 ; TRAN et al. 1999 ), we have analyzed the DNA Pol proofreading function in relation to several factors that influence replication accuracy. (While the pol3-01 mutation results in loss of proofreading, it is conceivable that some aspect of the results are due to this mutation disturbing an as-yet-undefined function of this domain.) Because a Pol proofreading defect (pol3-01) is haploid lethal in combination with an msh2 mutation while an msh2 pol2-4 double mutant is viable, it appears that Pol plays a greater role in mutation avoidance than DNA Pol .

Interaction between Pol proofreading and polymerase domains:
Using the construct lys2::InsLD, which allows for specific detection of large deletion mutations, we show here that the polymerase proofreading defect pol3-01 did not increase replication slippage over long distances (Table 1). Moreover, this mutation did not alter replication slippage induced by the polymerase mutation pol3-t when these two mutations were combined in the same gene. Because pol3-01 is a mutator for both frameshifts and base substitutions (MORRISON et al. 1993 ) we suggest that the proposed misalignment induced by nucleotide misincorporation (KUNKEL and SONI 1988 ) does not play a large role in replication slippage between distant repeats. Also, the observation that DNA polymerase proofreading does not repair loops formed between 7-nt repeats separated by 24 bp indicates that proofreading does not act on a loop that is located 7 nt from the 3' end of a replication fork.

While pol3-01 does not affect pol3-t-associated replication slippage, pol3-t acts as an antimutator with respect to pol3-01 (Table 2), indicating an interaction between the corresponding regions. This was surprising, because synergy might be expected from the combination of these two mutators, both of which affect frameshift mutations in homonucleotide runs. The reason for the antimutator effect of the pol3-t mutation is not clear, but could relate to replication processivity. As suggested in our previous studies, which demonstrate that the pol3-t mutation induces deletions in inverted repeats (GORDENIN et al. 1993 ), this mutant polymerase may replicate DNA more slowly or with decreased processivity, resulting in more single-stranded DNA regions in the lagging strand. Earlier studies demonstrate that decreased polymerase processivity can influence the fidelity of replication. For example, in the absence of its processivity cofactor, thioredoxin, T7 DNA polymerase creates more insertions (mutator) and fewer deletions (antimutator) in homopolymeric runs in vitro (KUNKEL et al. 1994 ). The phenotype of the pol3-t mutant, induction of replication slippage over long distances, is consistent with the idea that this mutant polymerase dissociates from the DNA template often. It is possible that dissociation at a site of misalignment in a homonucleotide run results in replication arrest and loss of the premutational event. Previously, it was shown that the Pol can participate in MMR, and we proposed that its exonuclease is directly involved in the mismatch excision step (LONGLEY et al. 1997 ; TRAN et al. 1999 ). One possible explanation is that in the pol3-t, 01 mutant MMR can be more efficient than in the pol3-01 strain. Another possibility is that the pol3-t mutation partially restores the proofreading defect. For example, several mutations have been identified (Y. PAVLOV and A. SUGINO, personal communication) in the Pol polymerase region that act as antimutators in the pol3-01 background and restore viability to pol3-01 msh2 and pol3-01 pms1 double mutant haploid strains.

In summary, our analysis and comparison of pol3-t and pol3-01 mutants demonstrates that they have strikingly different mutator effects. The pol3-t mutation has a large impact on replication slippage between separated small repeats, but a relatively small effect on frameshift mutations (Table 1 and Table 2). In contrast, the proofreading exonuclease-deficient pol3-01 polymerase does not induce replication slippage between distant repeats, but it greatly increases the frameshift mutation rate in both short and long homonucleotide runs (Table 3 and Table 4).

Interaction between MMR and Pol proofreading and polymerase activities in mutation avoidance:
MMR has an important role in preventing mutations in homonucleotide runs and particularly in long runs, because the longer the homonucleotide run, the greater the role it plays. The pol3-t mutation results in mostly large deletions and a small number of frameshift mutations in a MMR⁺ background (TRAN et al. 1996 ). As the postreplication MMR effectively corrects errors generated during replication, replication infidelity can be measured adequately only in a MMR^- background. The fraction of frameshift mutations increases dramatically in a MMR^- background, which indicates a synergistic interaction between pol3-t and msh2 with respect to frameshifts. However, this applies primarily to shorter homonucleotide runs, because with increased length of the run the mutator impact of pol3-t decreases (Table 2). A similar pattern is observed for the pol3-01 proofreading mutant. Thus, for both the polymerase (pol3-t) and the proofreading (pol3-01) mutations in an MMR^- background, the impact on mutation rate is greatest in shorter homonucleotide runs. Synergy between polymerase defects and postreplication MMR was also demonstrated previously in studies of haploid pol2-4 mutants (MORRISON and SUGINO 1994 ; TRAN et al. 1997B ) and in diploid double mutants pol3-01 pms1 and pol3-01 msh2 (MORRISON et al. 1993 ; TRAN et al. 1999 ).

The impact of homonucleotide run length on mutation rates in proofreading and polymerase mutants:
The accuracy of DNA replication is dependent not only on MMR, proofreading, and base selectivity, but also on the sequence of the DNA template. Using an in vitro system, KROUTIL et al. 1996 showed that proofreading prevents many frameshifts in short homonucleotide runs, but the proofreading effect decreases with length of the homonucleotide run. This same effect was confirmed for Pol proofreading in vivo (TRAN et al. 1997B ). In the present work, we examined the impact of Pol proofreading on the mutation rate in homonucleotide runs in diploid strains. We observed that in pol3-01 msh2 double mutant strains, the frameshift mutation rate (-1 and +1 frameshifts) in shorter runs (4–8 nucleotides) was 40- to 70-fold higher than in the msh2 single mutant. This difference is reduced to 2- to 9-fold in runs 10, 12, or 14 nucleotides in length. It is possible that frameshift intermediates in longer runs have a greater chance to escape Pol proofreading during replication. This result is consistent with earlier studies carried out both in vitro (KROUTIL et al. 1996 ) and in vivo (TRAN et al. 1997B ) for other polymerases.

In general, it appears that Pol proofreading has less of an impact on the mutation rate than Pol proofreading. This conclusion is based on (i) pol2-4, but not pol3-01, being haploid viable in combination with an MMR defect; (ii) differences in rates for various DNA templates in the presence of either Pol or proofreading mutations; and (iii) Pol proofreading having effects over a shorter distance than Pol (Table 4 and Table 2 in TRAN et al. 1997B ). For short runs, A₄ and A₅, pol2-4 msh2 (TRAN et al. 1997B ) and pol3-01 msh2 mutation rates are comparable. However, unlike pol3-01, the pol2-4 has little effect as the run increases to A₇ or A₈. Furthermore, Pol , but not Pol , proofreading can act on +1-nt frameshift intermediates in the A₁₂ run (Table 4, last column).

The difference between the mutation rates for -1 and +1 frameshifts in pol3-01 mutants is interesting and could reflect differences in the interaction of the -1 or +1 frameshift intermediate with the mutant polymerase during replication of long homonucleotide runs. Previous results showed that defective MMR increases both -1-nt and +1-nt frameshifts (SIA et al. 1997 ; TRAN et al. 1997B ); however, the relative increase of -1-nt frameshifts is much greater than that of +1-nt frameshifts. It is possible that the efficiency of MMR is greater for repair of -1-nt frameshift intermediates. Alternatively, -1-nt frameshifts may be generated more often than +1 frameshifts during replication. Our results favor the second possibility. It is possible that wild-type DNA Pol may correct +1 frameshift intermediates with higher efficiency than it corrects -1 frameshift intermediates, because -1 frameshift errors appear to be insensitive to pol3-01 in long runs (Table 4). In the double mutant pol3-01 msh2 (Table 4) we have observed comparable mutation rates for both -1-nt and +1-nt.

The pol3-t mutation also increases the frameshift mutation rate nearly 8-fold in A₄ and A₅ homonucleotide runs, when strains are msh2 defective. With longer runs, the effect is reduced; the frameshift mutation rate increases only 2.2- and 2.9-fold for the A₁₂ to A₁₄ runs, respectively (Table 2). Possibly the pol3-t polymerase mutation increases DNA misalignment during replication. Its overall impact may become less for long homonucleotide runs, where misalignment events would greatly increase [as suggested by the model of STREISINGER et al. 1966 ]. Another possibility is that the pol3-t mutation in the polymerase region partially impairs the proofreading activity of Pol . As described above for the Pol proofreading defect pol3-01, there is a reduced impact on mutation with increasing homonucleotide run length.

Hypermutability of diploid pol3-01 strains:
For several strains examined, the mutation rate is similar for both haploid and diploid cells (there is a generally small increase in diploids as expected for the additional second allele copy; Table 6). While no ploidy dependence was observed for wild-type, msh2, exo1, rad27, or pol2-4 strains, the pol3-01 strain was an exception to this pattern. The diploid pol3-01 mutant has a much higher frameshift mutation rate than a haploid (Table 5 and Table 6) for homonucleotide runs of various lengths as well as for the his7-2 allele. Similar observations were also made for the base substitution mutation rate (P. SHCHERBAKOVA and Y. PAVLOV, personal communication).

Hypermutability in diploids as compared to haploids was also found for mutagenesis by the base analog N⁶-hydroxylaminopurine (HAP; PAVLOV et al. 1988 , PAVLOV et al. 1991 ). The HAP-induced forward mutation rate at the LYS2 gene in a diploid strain was nearly 100-fold higher than expected on the basis of the mutation rate in a haploid strain. As the lys2 mutations induced by HAP were in most cases different for each of the two lys2 alleles in the diploid, the two mutations must be due to independent mutational events. It is possible that many cells treated with HAP as haploids die because of multiple mutations that inactivate essential genes. Because most mutations in yeast are recessive, multiple mutants would be viable in diploid cells, resulting in more lys2 mutants being recovered. Thus, the full impact of HAP mutagenesis can only be fully revealed in diploid cells (PAVLOV et al. 1988 , PAVLOV et al. 1991 ).

To explain the differences in mutability between haploid and diploid pol3-01 strains, we have proposed that the pol3-01 diploid population includes a fraction of hypermutable cells that is absent in the haploid pol3-01 population. This could occur if the cell population includes cells with transiently or permanently reduced expression of MMR genes or any other gene that would create hypermutability in combination with pol3-01. Several observations provide support for this idea. The pol3-01 mutation when combined with either a MMR defect (pms1 or msh2), an exo1 mutation, or a DNA Pol proofreading defect is inviable in a haploid because of excessive mutation errors, whereas the diploid double mutants are viable and exhibit hypermutability (MORRISON et al. 1993 ; MORRISON and SUGINO 1994 ; TRAN et al. 1999 ). Thus, a hypermutable fraction of pol3-01 cells with transient inactivation of one of these functions would be eliminated in a haploid, but would survive if the cells are diploid. By analogy with proofreading-defective E. coli strains, which display variable amounts of MMR deficiencies because of saturation (FIJALKOWSKA and SCHAAPER 1995 , FIJALKOWSKA and SCHAAPER 1996 ; SCHAAPER 1988 ), the yeast pol3-01 mutants may have reduced MMR capacity. This is supported by the significantly lower mutator effect of msh2 in pol3-01 strains than in POL⁺ strains (e.g., 20-fold vs. 14,000-fold for the A₁₄ run; Table 4). Complete loss of MMR in E. coli dnaQ926 is associated with loss of viability (FIJALKOWSKA and SCHAAPER 1996 ) as in the yeast pol3-01 msh2 and pol3-01 pms1 double mutants (MORRISON et al. 1993 ; TRAN et al. 1999 ).

The concept of a hypermutable cell fraction in pol3-01 mutants is also supported by the high frequency of coincident mutations in separate loci. The rate of simultaneous reversion for the two alleles his7-2 and lys2::InsE-A₁₂ in a homozygous pol3-01 diploid is 365 times higher than expected if the two events occur independently. A hypermutable state might also be revealed as a fraction of cells that exhibit higher frequencies of recessive lethals when diploids undergo meiosis. This could be examined in the diploid cells that exhibited multiple mutations. If a reduced level of MMR activity is responsible for the hypermutable cell fraction, then loss of MMR should render the population homogenous with regard to mutation. In the diploid pol3-01 msh2 strain the rate of simultaneous reversion for the two loci was much closer to the expected rate for two independent events. The lack of complete independence in coincident mutations in a MMR^- background may indicate that factors other than MMR could also play a role in the formation of the hypermutable cell population. Alternatively, the pol3-01 msh2 diploid strain, like the pol3-01 diploid, may also demonstrate the ability, although less severe, to accumulate a subfraction of hypermutable cells that increase the occurrence of coincident double reversion events in the population.

Because the Lys⁺ revertants from a pol3-01 diploid were no more mutable than the original strain, the proposed hypermutability is likely to be a transient phenomenon, and could be due to epigenetic changes in a portion of the population. The possibility that epigenetic change may cause hypermutability, on either a transient or permanent basis, is relevant to understanding the etiology of cancer in mammalian cells. For example, analysis of the lacI mutation spectrum from thymic tumor DNA of mouse Msh2-/- revealed a fraction of lacI genes that had multiple mutations (BAROSS et al. 1998 ). It was suggested that an additional mutator activity, such as an error-prone DNA polymerase, leads to increased genomic instability in these MMR-deficient tumors. Epigenetic changes were observed in several tumor suppressor genes and hMLH1. For these genes, an epigenetic process involving promoter hypermethylation-induced repression of transcription has been demonstrated in association with cancer development (HERMAN et al. 1994 , HERMAN et al. 1995 , HERMAN et al. 1998 ; MERLO et al. 1995 ).

Note added in proof:
The sequence of the his7-2 allele used in this study was recently described (P. V. SHCHERBAKOVA and T. A. KUNKEL 1999, Mutator phenotypes conferred by MLH1 overexpression and by heterozygosity for mlh1 mutations. Mol. Cell Biol. 19: 3177–3183). This mutation is due to a deletion of one A nucleotide in a run of eight adenines, so that reversions of the his7-2 allele can arise by -1 or +2 frameshifts.

	FOOTNOTES

¹ Present address: Lifesensors, Inc., 271 Great Valley Pkwy., Malvern, PA 19355.
² Present address: Lineberger Comprehensive Cancer Center, University of North Carolina, Chapel Hill, NC 27599.

	ACKNOWLEDGMENTS

We are grateful to Drs. P. Shcherbakova, Y. Pavlov, and A. Sugino for providing unpublished data, and to Drs. Y. Pavlov, Yong Hwan Jin, M. Longley, R. Schaaper, and M. Sander for comments on the manuscript.

Manuscript received October 8, 1998; Accepted for publication December 21, 1998.

	LITERATURE CITED

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

ARAKI, H., P. A. ROPP, A. L. JOHNSON, L. H. JOHNSTON, and A. MORRISON et al., 1992  DNA polymerase II, the probable homolog of mammalian DNA polymerase epsilon, replicates chromosomal DNA in the yeast Saccharomyces cerevisiae.. EMBO J. 11:733-740[Medline].

BAROSS, F. A., S. E. ANDREW, J. E. PENNEY, and F. R. JIRIK, 1998  Tumors of DNA mismatch repair-deficient hosts exhibit dramatic increases in genomic instability. Proc. Natl. Acad. Sci. USA 95:8739-8743[Abstract/Free Full Text].

BENNETT, C., A. LEWIS, K. BALDWIN, and M. RESNICK, 1993  Lethality induced by a single site-specific double-strand break in a dispensable yeast plasmid. Proc. Natl. Acad. Sci. USA 90:5613-5617[Abstract/Free Full Text].

BOULET, A., M. SIMON, G. FAYE, G. BAUER, and P. BURGERS, 1989  Structure and function of the Saccharomyces cerevisiae CDC2 gene encoding the large subunit of DNA polymerase III. EMBO J. 8:1849-1854[Medline].

FIJALKOWSKA, I. J. and R. M. SCHAAPER, 1995  Effects of Escherichia coli dnaE antimutator alleles in a proofreading-deficient mutD5 strain. J. Bacteriol. 177:5979-5986[Abstract/Free Full Text].

FIJALKOWSKA, I. J. and R. M. SCHAAPER, 1996  Mutants in the Exo I motif of Escherichia coli dnaQ: defective proofreading and inviability due to error catastrophe. Proc. Natl.