Abstract
The DNA polymerases δ and ε are the major replicative polymerases in the yeast Saccharomyces cerevisiae that possess 3′ → 5′ exonuclease proofreading activity. Many errors arising during replication are corrected by these exonuclease activities. We have investigated the contributions of regions of Polε other than the proofreading motifs to replication accuracy. An allele, pol2-C1089Y, was identified in a screen of Polε mutants that in combination with an exonuclease I (exo1) mutation could cause a synergistic increase in mutations within homonucleotide runs. In contrast to other polymerase mutators, this allele specifically results in insertion frameshifts. When pol2-C1089Y was combined with deletions of EXO1 or RAD27 (homologue of human FEN1), mutation rates were increased for +1 frameshifts while there was almost no effect on −1 frameshifts. On the basis of genetic analysis, the pol2-C1089Y mutation did not cause a defect in proofreading. In combination with a deletion of the mismatch repair gene MSH2, the +1 frameshift mutation rate for a short homonucleotide run was increased nearly 100-fold whereas the −1 frameshift rate was unchanged. This suggests that the Pol2-C1089Y protein makes +1 frameshift errors during replication of homonucleotide runs and that these errors can be corrected by either mismatch repair (MMR) or proofreading (in short runs). This is the first report of a +1-specific mutator for homonucleotide runs in vivo. The pol2-C1089Y mutation defines a functionally important residue in Polε.
ERROR avoidance and correction are essential for reducing a species' mutational load. DNA polymerases are intrinsically accurate during replicative synthesis due to both base selectivity and opportunities to correct errors through proofreading (Kunkel 1992). In addition, mismatch repair (MMR) systems provide for postreplicational monitoring and correction of errors.
In all eukaryotes examined nuclear DNA polymerases Polε and -δ possess 3′ → 5′ exonuclease activity (Morrisonet al. 1991). Genetic studies in the yeast Saccharomyces cerevisiae have demonstrated that the 3′ → 5′ exonuclease activities of these polymerases are responsible for proofreading of newly replicated DNA (Morrisonet al. 1991; Simonet al. 1991; Morrison and Sugino 1994) and that proofreading mutants are frameshift and base substitution mutators (Morrisonet al. 1991; Simonet al. 1991). Proofreading by other replicative polymerases can decrease DNA replication error rates up to two orders of magnitude in vitro (Kunkel 1992) and in vivo (Schaaper 1993). Proofreading activity is provided by the conserved amino acid motifs ExoI, ExoII, and ExoIII found at the amino-terminal end of B family polymerases (Morrisonet al. 1991; Simonet al. 1991; see Figure 1).
Proofreading by either Polδ or Polε is ineffective at correcting frameshift errors in long homonucleotide runs during in vitro or in vivo replication (Kroutilet al. 1996; Tranet al. 1997). Such runs are especially prone to frameshift mutations, which are generally acknowledged to arise by replication slippage (reviewed in Gordenin and Resnick 1998). Two models for slippage have been proposed. One model proposes that slippage occurs through disassociation of the polymerase from the template and incorrect reannealing of the template and nascent strands (Streisingeret al. 1966; Kunkel and Soni 1988). Recently an alternative model for slippage leading to insertion frameshifts has been proposed. In this model frameshifts occur during reannealing of the DNA strands after partitioning of the nascent strand between the polymerization and proofreading domains (Fujiiet al. 1999). The in vitro correction of mispaired bases by proofreading is limited to approximately five nucleotides behind the 3′ terminus of the primer (Lamet al. 1999). Genetic studies have revealed similar limitations on proofreading in vivo (Tranet al. 1997).
The 5′ → 3′ exonuclease encoded by EXO1 is involved in recombination, resistance to UV damage, and MMR (Fiorentiniet al. 1997; Tishkoffet al. 1997a; Johnsonet al. 1998; Qiuet al. 1998; Tranet al. 1999b) and complements many defects in rad27 strains (Tishkoffet al. 1997b; Parenteau and Wellinger 1999). Rad27, the yeast homologue of human flap endonuclease (FEN1), is a 5′ → 3′ exo/endonuclease responsible for the maturation of Okazaki fragments during lagging strand DNA synthesis and removal of 5′ flaps (reviewed in Lieber 1997). Rad27 has also been proposed to function in MMR (Johnsonet al. 1995). However, its role in MMR has been questioned by the finding that in forward mutation assays the spectra of Δrad27 and Δmsh2 mutants are very different (Tishkoffet al. 1997b). The absence of RAD27 can result in increased mutation rates including expansion of repeat sequences and large duplications (Tishkoffet al. 1997b; Freudenreichet al. 1998; Kokoskaet al. 1998; Maureret al. 1998). Furthermore, a structure with mispaired nucleotides in close association with a 5′ flap is processed by Fen to remove both the flap and the mispaired nucleotides (Rumbaughet al. 1999). Mutants lacking both Exo1 and Rad27 are inviable (Tishkoffet al. 1997b; Garyet al. 1999). Overexpression of Exo1 in a Δrad27 mutant complements the temperature sensitivity and partially complements the mutator phenotype of the mutants (Tishkoffet al. 1997b; Parenteau and Wellinger 1999) suggesting that Exo1 may also be able to function directly in processing of flaps found in lagging strand replication intermediates.
Error avoidance in long homonucleotide runs is accomplished predominantly by MMR (Kolodner 1996; Tranet al. 1997). The Escherichia coli methyl-directed mismatch repair system has provided a model for the understanding of MMR in eukaryotes. The basic steps of these MMR systems are similar, namely (1) mismatch recognition, (2) incision of the newly replicated DNA strand, (3) nuclease-mediated deletion of the mismatch, and (4) gap filling and ligation. Similar to the E. coli system, in vitro results with human extracts have shown that mismatch excision can occur from either side of a mismatch. In yeast the combination of a 3′ 5′ → Polδ or Polε exonuclease deficiency with a defect in either the EXO1 or the RAD27/FEN1 5′ → 3′ exonucleases caused a synergistic increase in mutations in long homo nucleotide runs (Garyet al. 1999; Tranet al. 1999b). No synergy was found between mutations in other mismatch repair genes (deletion of MSH2, MSH3, MSH6, or PMS1) and deficiencies in proofreading (Polδ or Polε) for mutations occurring in long homonucleotide runs. On the basis of these results it was proposed that the exonuclease activities of Exo1, Rad27, and Polε or Polδ participate in and substitute for each other at the excision step of MMR (Tranet al. 1999b).
Relatively little is known about the regions of eukaryotic DNA polymerases that determine DNA replication accuracy in vivo. One approach is to create mutations within the polymerases and relate structural changes to specific mutation-generating characteristics in vitro. We have taken a similar in vivo approach in the generation of DNA Polε mutations. In addition to functioning in cellular DNA replication and proofreading, Polε has been implicated in various types of DNA repair including nucleotide excision repair (NER), base excision repair (BER), and recombination (Budd and Campbell 1997; Burgers 1998). We have developed a novel mutation detection system, based on interactions with exo1 mutants, to identify functions of Polε other than proofreading that influence genome stability. We have identified a unique DNA polymerase mutator with defects that are not due to a change in proofreading capacity or the ability to function in MMR. Unlike other mutators, it specifically causes +1 frameshift mutations in homonucleotide runs.
MATERIALS AND METHODS
General genetic and molecular methods: Standard yeast media and yeast extract-peptone-dextrose (YPD) media with G418 have been described previously (Roseet al. 1990). Yeast cells were grown at 30° unless otherwise stated. Yeast transformations were performed by the method of Gietz and Schiestl (1991). Preparation of bacterial growth media and molecular methods have been described previously (Tran et al. 1995, 1996).
Strains and plasmids: A series of isogenic strains was constructed from CG379 (Matα ade5-1 his7-2 leu2-3,112 trp1-289 ura3-52) containing deletions of various DNA metabolism genes. Insertion of the InsE element with a homonucleotide run containing varying lengths of A within LYS2 has been described (Tranet al. 1999b). The plasmid used to integrate mutations within POL2, p173, was constructed by subcloning the BamHI-BspEI fragment of POL2 from YCpPol2 (Morrisonet al. 1990) into the BamHI-AvaI sites of pFL34*. pFL34* is identical to pFL34 (Bonneaudet al. 1991) except that the URA3 marker is in the opposite orientation after BglII digestion (K. Lobachev, personal communication). Strains containing pol2-C1089Y were constructed using a site-directed mutant of p173. Site-directed mutants were made using the QuickChange mutagenesis kit from Stratagene (La Jolla, CA) according to the manufacturer's instructions. Primers to make the Cys to Tyr change and to add an RsaI restriction site were 5′-GGTAAAAGATAAAGGTCTACAGTACAAATATATTATTAGTCAAAACC and 5′-GGTTTTGAACTAATAATATATTTGTACTGTAGACCTTTATCTTTTACC.
Mutagenesis of plasmid DNA: One tube containing plasmid p173 (pFL34* + POL2) was treated with 1 m hydroxylamine in 50 mm pyrophosphate, 100 mm NaCl, pH 7.0, for 1 hr at 70°. After stopping the reaction on ice, the DNA (3 μg) was dialyzed against TE and transformed into DH5α cells. DNA was isolated using a QIAGEN (Chatsworth, CA) column as described by the manufacturer. Since cells were grown for multiple generations mutations at the same sites cannot be considered independent.
Screen for A12 homonucleotide run mutators: The mutagenized plasmid DNA was digested with AgeI or PimAI and transformed into cells, which were plated to uracil drop-out media. After 3 days of growth at 30° colonies were replica plated to uracil and to lysine drop-out media. After 3 days of growth, colonies that exhibited four or more papillae on the lysine drop-out medium were picked from the uracil drop-out plate and streaked for single colonies. Mutator colonies were grown on 5-FOA and those that had lost the URA3 gene and were still mutators were used for further study.
Gene replacement and disruptions: We used p173-rsa, described above, for replacement of the wild-type POL2 gene with the C1089Y allele. The plasmid was digested with AgeI or PimAI and transformed into cells with selection on uracil dropout media. Transformants were screened for the presence of the site-directed change by a two-step procedure. After PCR amplification the fragment was digested with RsaI to determine if the change was present. Strains that had the mutation were grown on 5-FOA to select for loss of the URA3 gene. Other strains containing DNA repair/replication mutants were made from previously described strains (Tran et al. 1997, 1999b).
Mapping and DNA sequence analysis of mutants: The locations of the new POL2 mutations were determined by mapping using gap repair with deleted versions of p173. After determining that the mutations were between the AspI and SpeI sites of plasmid p173, DNA sequence analysis was performed on genomic DNA from this area. Six DNA fragments were made by PCR amplification using primers pairs (numbered from the ATG with + indicating 5′ → 3′ strand and − indicating the 3′ → 5′ strand; all 20 nucleotides): 1, pol2+1949 and pol2−3300; 2, pol2+2948 and pol2−3800; 3, pol2+3220 and pol2−4030; 4, pol2+3520 and pol2−4342; 5, pol2+3826 and pol2−5000; 6, pol2+4100 and pol2−5000. Sequencing of the PCR fragments was done using an ABI model 373A DNA sequencer. Sequencing of Lys+ revertants was done by PCR amplification of a fragment of the LYS2 gene that included the InsE insertion that contained the homonucleotide run. The primers used to amplify and sequence this fragment and the his7-2 homonucleotide run have been described (Shcherbakova and Kunkel 1999; Tranet al. 1999b).
Measurement of mutation rates: Mutation rates were determined by a fluctuation test using the method of the median (Lea and Coulson 1949) on at least 12 independent cultures as described (Tranet al. 1999b).
RESULTS
Isolation of Polε mutators: Proofreading is required for accurate replication by Polε. Little is known about the contributions of other regions of this polymerase to replication fidelity. We investigated the ability of randomly generated Polε mutations, external to the known proofreading domain, to increase the mutation rate in the long A12 homonucleotide run of the lys2::InsE-A12 allele. Such runs are highly sensitive to subtle changes in proteins that can affect the appearance of mutations (Tran et al. 1997, 1999a,b; Clarket al. 1999; Drotschmannet al. 1999). The lys2::InsE-A12 allele was used previously in the characterization of various DNA metabolic mutants for their ability to influence +1 frameshift mutations (Tran et al. 1999a,b).
The Polε mutants were examined in a strain containing a deletion of EXO1. Use of such a genetically sensitized background (Perkinset al. 1999) can increase the probability of finding mutants when screening large numbers of random isolates. On the basis of the results with other double mutants (Tranet al. 1999b), we predicted that mutant identification in the Δexo1 strain would be more efficient than in the wild-type strain because of anticipated synergistic interactions between Polε mutations and the deletion allele of EXO1.
(A) Domain structure of the protein coded for by DNA Pol2 cDNA and location of the changed amino acid in the mutator allele. Regions of Polε including the exonuclease domains, the polymerization domain, and the zinc finger domain are shown relative to the sequenced mutation. Shown underneath is the region that was mutagenized in this study. (B) Multispecies alignment of the region containing amino acid C1089. The Cys residue at amino acid 1089 of S. cerevisiae is in a 17-amino acid block that is highly conserved. When four or more species have an identical amino acid in a given location this position is considered conserved and it is high-lighted in this figure.
Screening of >5000 Ura+ transformants yielded 3 with significantly higher Lys+ reversion rates. Their mutator phenotype was complemented by the plasmid YCpPol2 containing the wild-type POL2 gene. Isolates 37a and 45f were characterized further (the third mutant is under study). The role of the EXO1 defect in revealing the POL2 mutator phenotype was determined by introducing EXO1 into cells on the 2μ plasmid pRDK480. The relative mutation rates of both of the pol2 exo1 mutants compared to wild type were increased 33-fold. Addition of the pRDK480 plasmid reduced the relative rates to 2- and 3-fold over wild type, respectively. This established that the pol2 mutants isolated are at best weak lys2::InsE-A12 mutators on their own and require a deficiency in the EXO1 gene for their strong mutator phenotype in this assay.
Identification of the alterations within Polε: The Polε mutator alleles were mapped initially by gap repair. Mutations 37a and 45f were localized to the same 1 kb of DNA. DNA sequencing revealed that both clones had a G-to-A change at base 3266 (C1089Y) of the coding region, suggesting a common origin for this mutation in these strains (Figure 1A). Alignment of Polε homologues from S. cerevisiae, Schizosaccharomyces pombe, Mus musculus, Caenorhabditis elegans, Emericella nidulans, Arabidopsis thaliana, and Homo sapiens identified a common cysteine residue in a 17-amino acid block of homology (Figure 1B) that is between 60 and 100% identical in these organisms designated as C-2 (Huang et al. 1999a,b). The cysteine to tyrosine change at amino acid 1089 is located 83 amino acids downstream from the last of the polymerase domains (Figure 1A) in a region that has been proposed to be involved in subunit interactions (Kestiet al. 1993).
To demonstrate that the C1089Y substitution is the only alteration in POLε required for the mutator phenotype, we made two site-directed mutant constructs of plasmid p173. The first construct had only the change found in the mutator allele. The second construct also contained silent changes giving rise to an RsaI restriction site. Either plasmid was sufficient to yield the mutator phenotype observed in the original mutant (data not shown). In our subsequent analysis we utilized strains containing the p173-Rsa construct; this allele is referred to as pol2-C1089Y. Plating efficiencies or growth rates at 30° or 37° were not altered by this allele alone or in combination with Δexo1 (data not shown).
Impact of the Polε mutator in combination with other mutations: The Polε defect was examined on its own and in combination with mutations in other genes that impact on the maintenance of homonucleotide runs. Isogenic strains that contained A10, A12, and A14 runs within the LYS2 gene or an A7 run in the HIS7 gene (i.e., his7-2) were used. Revertants of the lys2 alleles can occur only in a 79-bp window (Tranet al. 1997); reversion of the his7-2 allele occurs in a 43-bp window (Shcherbakova and Kunkel 1999). Defects in MMR result in all (A10, A12, and A14) or nearly all (his7-2) reversions occurring in these runs (Tran et al. 1997, 1999b; Shcherbakova and Kunkel 1999). Revertants associated with the A12 and A7 runs are due to +1 frameshifts and A14 and A10 revertants arise by −1 frameshifts. As shown in Table 1 the pol2-C1089Y mutation itself does not increase mutation rates in the lys2::InsE-A12, A7, A10, and A14 homonucleotide runs. In contrast, the weak mutator phenotype for forward mutation rate at the CAN1 locus is increased relative to that in POL+ strains. The mutator effect of pol2-C1089Y at CAN1 is greater than that observed for the proofreading mutant pol2-4 (Morrison and Sugino 1994; Tranet al. 1999b; Table 1).
When combined with other DNA replication/repair defects, the pol2-C1089Y mutation was similar to pol2-4 for frameshifts in the A12 run. Both the Exo1 and Rad27 5′ → 3′ exonucleases have been implicated in mismatch repair (Johnsonet al. 1995) although this is controversial (Tishkoffet al. 1997b). The pol2-4 rad27 and pol2-4 exo1 double mutants exhibit synergistic increases in mutation rates for homonucleotide runs, as well as at the CAN1 locus, relative to the single mutants (Tishkoffet al. 1997b; Tranet al. 1999b). Results for double mutants with pol2-C1089Y are presented in Table 1. Double mutants containing pol2-C1089Y and either Δexo1 or Δrad27 also led to synergistic (15- to 20-fold) increases in the mutation rates in the lys2::InsE-A12 homonucleotide run. There was no significant enhancement of the lys2::InsE-A12 mutation rate when pol2-4 or pol2-C1089Y was combined with a mutation in the MMR gene MSH2, possibly due to the already high mutation rate of the msh2 mutant in this assay. The effect of the pol2-C1089Y mutation does not appear to extend to recombination since there was no effect in a plasmid-based recombination assay (J. M. Kirchner and M. A. Resnick, unpublished observations).
The POL2-C1089Y protein causes +1 frameshift mutations in long homonucleotide runs: As shown in Table 1, the combination of pol2-C1089Y with Δexo1 or Δrad27 caused a >25-fold increase in reversion of the his7-2 allele. The proofreading defect pol2-4 leads to similar synergistic increases in −1 frameshifts within long homonucleotide runs (Tranet al. 1999b; Table 1). We therefore examined combinations of Δexo1 or Δrad27 with pol2-C1089Y for their effects on reversion of lys2 alleles containing A10 or A14 runs. Surprisingly, pol2-C1089Y allele did not lead to increased mutation rates in these −1 frameshift mutation detection assays, whereas pol2-4 resulted in ~34- and 4-fold increases in mutation rates, respectively, when combined with Δexo1 or Δrad27 in the lys2::InsE-A14 reversion assay (Table 1).
To determine if the mutations scored in the lys2::InsE-A12 assay were actually +1 frameshifts within the homonucleotide run, we sequenced Lys+ revertants of the pol2-C1089Y Δexo1 strain. As expected, all revertants (12/12) of the pol2-C1089Y Δexo1 strain in the lys2::InsE-A12 assay were due to +1 frameshifts. Similarly, all mutations that occurred in the his7-2 homonucleotide run in the pol2-C1089Y Δexo1 strain were associated with +1 mutations (10/10 sequenced). Thus, when assayed for frameshift mutator activity in long homonucleotide runs, the pol2-C1089Y allele leads to a specific increase in +1 mutations.
The pol2-C1089Y mutator is not due to a proofreading defect: The 3′ → 5′ proofreading exonuclease activity of Polε (as well as Polδ) can greatly reduce the potential for mutations. To determine if the pol2-C1089Y allele alters proofreading, even though the substituted amino acid is outside the known exonuclease domains, this mutant was examined for several phenotypic characteristics common to proofreading mutants. In comparison with the single mutants, the double mutant pol2-4 msh2 exhibits a synergistic increase in CAN1 forward mutation rates (Morrison and Sugino 1994; Tranet al. 1999b). In pol2-C1089Y msh2 strains the mutation rate in the CAN1 forward mutation assay is not significantly different from the msh2 single mutant (Table 1). The combinations pol2-4 exo1 or pol2-4 rad27 also result in synergistic increases in CAN1 mutation rates; however, no synergy is observed for the corresponding pol2-C1089Y double mutants (Table 1).
Inactivating the 3′ → 5′ exonuclease activity of Polε does not increase mutation rates in long homonucleotide runs (Kroutilet al. 1996; Tranet al. 1999b; Table 1). This suggests that Polε acts together with other nucleases, such as Exo1 and Rad27, to reduce mutations in long homonucleotide runs so that double mutants exhibit greatly enhanced mutation rates for both +1 and −1 frameshifts (Garyet al. 1999; Tranet al. 1999b; Table 1). The pol2-C1089Y mutation differs from pol2-4 in that only +1 frameshifts are increased. Haploid pol2-4 pol3-01 double mutants, lacking both proofreading activities, are inviable, presumably due to a catastrophic increase in mutations (Morrisonet al. 1993; Morrison and Sugino 1994). We therefore transformed a strain containing the Polδ proofreading mutation pol3-01 and a plasmid expressing wild-type POL2 with the pol2-C1089Y integrating plasmid. Following selection for the presence of only pol2-C1089Y and loss of the POL2 plasmid, viable double mutants could be isolated that exhibited normal growth rates. The mutation rate of the double mutant pol3-01 pol2-C1089Y at lys2::InsE-A12 was comparable to that of the pol3-01 mutant alone (8.7 × 10−6 vs. 4.0 × 10−6). These results lead us to conclude that the pol2-C1089Y allele's defect is unlikely to result from a deficiency in proofreading.
Interaction of pol2-C1089Y with mutations affecting error avoidance: increases in mutation rates for reversion in homonucleotide runs (A10, A12, A14 of lys2::InsE; A7 of his7-2) and forward mutation in CAN1
Mutation ratesa in short homonucleotide runs (A4 and A5 of lys2::InsE)
POL2-C1089Y protein creates errors in short homonucleotide runs: Although proofreading is not efficient at correcting frameshift errors in long homonucleotide runs, both Polδ and Polε proofreading can greatly reduce errors in short runs (Tranet al. 1999b; Table 2). We investigated the impact of the pol2-C1089Y mutation on the appearance of frameshift mutations in short homonucleotide runs (A5 or A4) using the lys2::InsE-A run system. A defect in MMR (Δmsh2) leads to 39- to 78-fold increases in mutation rates in the A5 (+1 frameshifts) and A4 (−1 frameshifts) runs (Table 2; Tranet al. 1997), respectively. There was a strong synergistic increase in mutation rates when msh2 and pol2-C1089Y were combined in the lys2::InsE-A5 strain, as was also the case for the pol2-4 msh2 strain. No synergistic effect was observed for the double mutant pol2-C1089Y Δmsh2 in a lys2::InsE-A4 strain. In contrast, a pol2-4, Δmsh2 double mutant showed a strong synergistic increase in −1 frameshifts. The pol2-4 mutation also shows synergy with Δexo1 and Δrad27 for both the lys2::InsE-A4 and the lys2::InsE-A5 mutation targets (Tranet al. 1999b; Table 2). In contrast to pol2-4, no increase in mutation rate was observed for double mutants formed between pol2-C1089Y and either Δexo1 or Δrad27 in the +1 short homonucleotide run assay (A5) over that found for either of the single mutants (data not shown). (Since pol2-C1089Y has shown synergy only for +1 frameshifts, −1 frameshifts were not examined in these double mutants.) Synergy for mutation with Δmsh2 in the double mutant strain verified that pol2-C1089Y is not a MMR mutant. Thus, the +1 frameshift mutations generated in the pol2-C1089Y mutant occur in both long and short homonucleotide runs and are correctable by the post-replicational MMR systems. The specificity for +1 frameshifts and the interaction between the Pol2-C1089Y protein and nucleases affecting MMR are inconsistent with it being a proofreading defect. These data lead us to suggest that pol2-C1089Y encodes a novel error-prone DNA polymerase mutant.
The Pol2-C1089Y protein does not appear to affect processing of Okazaki fragments: Since the pol2-C1089Y allele does not lead to a MMR defect, the observation that there is synergy with either the Δrad27 or the Δexo1 mutation in long homonucleotide runs suggests that the pol2-C1089Y allele may cause a defect in the completion of lagging strand replication or it may generate mutations. To rule out involvement of pol2-C1089Y in Okazaki fragment processing, we have examined mutation rates in double and triple mutants of genes affecting flap processing and MMR in combination with pol2-C1089Y. Previously, it was shown that Exo1 as well as Rad27/FEN1 might have a role in removing flaps generated during processing of Okazaki fragments since a Δexo1 Δrad27 double mutant is inviable and overexpression of Exo1 can compensate for the temperature sensitivity and partially compensate for the mutator phenotype of a Δrad27 strain (Tishkoffet al. 1997b; Parenteau and Wellinger 1999).
If the elevated mutation rates in the double mutants formed with Δrad27 or Δexo1 are due to additional impairment of Okazaki fragment processing by the pol2-C1089Y, then the triple mutants pol2-C1089Y Δexo1 Δmsh2 or pol2-C1089Y Δrad27 Δmsh2 would be expected to exhibit a substantial increase in mutation rate over the pol2-C1089Y Δmsh2 mutant (this has the highest mutation rate of any of the double mutant combinations between these genes). This was not the case. The triple mutants and the double mutants exhibited comparable mutation rates for the lys2::InsE-A12 allele: 1.0 × 103 and 1.1 × 103 for the triple mutants formed with exo1 and rad27, respectively, compared to 0.5 × 10−3 for the pol2-C1089Y, msh2 strain.
DISCUSSION
Isolation of Polε mutators: Assays based on long homonucleotide runs are highly sensitive systems for examining changes in mutation avoidance systems (reviewed in Gordenin and Resnick 1998). Combinations of single gene defects, each having only a subtle phenotype separately, can result in synergistic mutation responses on these runs (Garyet al. 1999; Tranet al. 1999b). Homonucleotide runs have been employed in screens to identify mutators, to isolate mutants with subtle changes in MMR, and to characterize the role of some DNA metabolic genes in the mutation process. We initiated a study to identify Polε mutations, external to the known proofreading domains, that could destabilize long homonucleotide runs. Because this polymerase is thought to participate in multiple repair processes (reviewed in Budd and Campbell 1997) as well as replication, such mutations could aid in determining structural and functional roles of the protein. Using a genetically sensitized exo1-deleted background to screen the mutants, we identified a novel Polε mutator that specifically increases +1 frameshift mutations.
Mutator effects of the pol2-C1089Y mutation: The Δexo1 mutation alone is not a mutator for A12 runs. However the Δexo1 mutation was expected to greatly enhance the impact of other mutators in a lys2::InsE-A12 reversion assay based on responses of double mutants formed between Δexo1 and pol2-4 (Table 1) or pol3-01 (in a diploid strain; Tranet al. 1999a). The fragment of POLε that was mutagenized in our screen excluded the exonuclease domains. However, it is possible that alterations in the amino acid sequence in regions distant from the proofreading motifs might physically interact with these motifs and disrupt proofreading ability.
The screen, in a repair-compromised strain, identified the mutation pol2-C1089Y, which markedly increased mutation rates in the lys2::InsE-A12 run. The increase in mutation rates in the new pol2 mutants was dependent on loss of exo1 function. This mutator could not have been revealed through traditional screens since the pol2 single mutant had a barely detectable mutator phenotype for long homonucleotide runs. The lack of a strong phenotype of pol2-C1089Y alone on A7, A10, A12, and A14 is not surprising if the errors generated are corrected by the proofreading and/or the MMR systems.
The approach of using a sensitized strain background (i.e., Δexo1) could be used to isolate additional mutations of DNA polymerases ε or δ. We utilized Δexo1, but other nuclease defects that impact on error avoidance could be considered. For example, the rad27 pol2-C1089Y and exo1 pol2-C1089Y double mutants had similar phenotypes.
pol2-C1089Y is not a proofreading defect, but instead induces mutations that are subject to MMR and proofreading: Several lines of evidence lead us to conclude that pol2-C1089Y is not a proofreading mutation but instead that the altered protein increases errors during DNA synthesis. The combination of a proofreading mutation, in either Polδ or Polε with a deletion of either exo1 or rad27, synergistically increases both +1 and −1 frameshift mutations in homonucleotide runs as well as CAN1 forward mutations (Tranet al. 1999b; Table 1). This is contrary to the results obtained with pol2-C1089Y where only the rate of +1 frameshift mutations was increased and there was little increase in mutations to CAN resistance (Tables 1 and 2). The pol3-01, pol2-C1089Y combination was viable as a haploid and did not exhibit increased mutation rates in long homonucleotide runs. Double mutants lacking polymerase δ and polymerase ε proofreading exhibit extremely high mutation rates as diploids and they are inviable as haploids (presumably due to error catastrophe; Morrison and Sugino 1994). We propose that the pol2-C1089Y allele leads to a change in the DNA polymerase that does not involve error correction, but instead increases the likelihood of errors during replication of homonucleotide runs.
Mutations arising in long homonucleotide runs are subject to MMR (reviewed in Gordenin and Resnick 1998). Previously it was suggested that Polε and Polδ proofreading nucleases as well as Exo1 participate directly in MMR error excision (Tranet al. 1999b). The observation of synergy between the msh2 and pol2-C1089Y defects for reversion of his7-2 and lys2::InsE-A5 suggests that this pol2 mutation can cause a substantial mutational burden, most of which is repaired by the MMR and/or the proofreading system. Proofreading could explain the relatively modest 9-fold mutation rate increase in the Δmsh2 pol2-C1089Y strain over the single Δmsh2 strain for A7 runs. Errors in short homonucleotide runs (A4 and A5) are efficiently corrected by both the MMR system and by proofreading (Kroutilet al. 1996; Tranet al. 1997). The 200-fold increased reversion of lys2::InsE-A5 in the Δmsh2 pol2-C1089Y strain relative to msh2 alone supports the view that MMR corrects errors generated by pol2-C1089Y. Results with reversion systems that focus on changes in short homonucleotide runs also indicate that Pol2-C1089Y generates errors that are correctable by proofreading. Combining a mismatch repair defect with a proofreading defect increases mutation rates in short homonucleotide runs 30- to 300-fold over the rates in a MMR− mutant (Tranet al. 1999b; Table 2). A similar effect was found for +1 frameshift mutations in lys2::InsE-A5 when pol2-C1089Y was combined with Δmsh2. However, −1 frameshifts in lys2::InsE-A4 were not increased. On the basis of these observations we propose that the involvement of pol2-C1089Y in replication or repair results in the induction of mutations and that these mutations are subject to both MMR and proofreading. The mechanism of induction and the reason for the +1 mutator phenotype in homonucleotide runs may be due to differences in ability of the nascent and template strands to form structures that could result in replication slippage by this mutant polymerase (discussed below).
Consequences of the pol2-C1089Y mutation on Pol2 function: The increase in only +1 frameshifts seen in the pol2-C1089Y mutant is unusual in that most DNA polymerases have higher levels of −1 frameshifts when examined for spontaneous mutations (Kunkel 1992). The interaction of T7 DNA polymerase with thioredoxin increases the processivity of the polymerase (Taboret al. 1987) and, in the absence of thioredoxin, there is an increase in frameshift errors (Kunkelet al. 1994; Kroutilet al. 1996). The +1 frameshift mutation rate is nearly 50-fold higher in the absence of thioredoxin and there is a five-fold bias in +1 vs. −1 frameshifts. This specificity is similar to the strong bias toward +1 frameshifts with the pol2-C1089Y mutation.
PCNA is key to DNA polymerase processivity in eukaryotes. PCNA binds to DNA polymerases δ and ε in vitro and stimulates processivity (Tsurimoto 1998). PCNA can interact directly with the Polε protein (Looret al. 1997); however, this interaction could also be stabilized through subunit interactions. The region where the Pol2-C1089Y substitution occurs has been proposed to function in interactions of Polε with its subunits (Kestiet al. 1993). Of the four identified subunits, DPB2 and DPB11 genes are essential, while DPB3 and DPB4 are not (reviewed in Sugino 1995). It is possible that the C1089Y substitution interferes with some interaction of the subunits and/or PCNA. This was tested by creating strains deleted for Dpb3 or Dpb4, or overexpressing Dpb2 or Dpb11, and examining their genetic interactions with Δexo1 or Δrad27 in the presence or absence of pol2-C1089Y. Neither deletion of the Dpb-3 or Dpb-4 subunits nor overexpression of Dpb2 or Dpb11 altered the mutation rate in the lys2::InsE-A12 assay. These data suggest that the pol2-C1089Y defect was not due to altered subunit interactions (data not shown).
Crystal structure data for DNA polymerases have led to a model where the enzyme resembles a partly open right hand with three domains called fingers, palm, and thumb forming a U shape (Joyce and Steitz 1994). Cocrystal structure data for DNA PolI and Taq DNA polymerases with DNA indicate that the thumb subdomain has many interactions with the phosphate backbone of the DNA. Multiple interactions occur between the thumb and the template-primer duplex DNA molecule (Joyce and Steitz 1994). The crystal structure of the class B DNA polymerase from phage RB69 has been determined recently (Wanget al. 1997). The replicative polymerases δ and ε of S. cerevisiae belong to the class B polymerase family. Many polymerases of this family show good amino acid alignment with each other until the start of the thumb subdomain, after which the sequences diverge (Wanget al. 1997). A method for estimating the location of the pol2-C1089Y mutation against the known crystal structure is to start at the last location of alignment, the end of the T region in the RB69 structure described by Wang et al. (1997), and to add 83 amino acids, corresponding to the distance from that position to amino acid 1089 in the yeast protein. This places the amino acid C1089 in the thumb subdomain.
Minnick et al. (1996) showed that a 24-amino acid deletion at the tip of the thumb domain of the large fragment of E. coli PolI (Klenow fragment) causes a specific increase in +1 frameshift mutations in vitro. Their mutant, like pol2-C1089Y, did not increase base substitutions or −1 frameshift mutations. Characterization of the mutant PolI enzyme revealed a fourfold decrease in processivity relative to the wild-type enzyme. Since frameshift mutations in homonucleotide templates are caused by slippage of the template-primer (Kunkel 1992), the processivity defect suggests that the mutator phenotype of the polymerase might be due to incorrect reannealing after disassociation. However, frameshift mutations can also occur during highly processive DNA synthesis in vitro (Kroutilet al. 1998).
An alternative model for insertional frameshifts has been proposed recently by Fujii et al. (1999). These researchers suggest that partitioning of the nascent strand between the polymerization site and the exonuclease site of the polymerase during proofreading allows for incorrect reannealing of the nascent strand and the specific generation of + frameshift mutations. This model implicates proofreading activity itself in the generation of frameshift errors. An attractive feature of this model is that it does not require disassociation of the polymerase from the template strand during the frameshift generation process.
On the basis of the above observations, we suggest that the pol2-C1089Y mutation may affect processivity either directly or due to an altered interaction with the appropriate replication factor(s). Alternatively, more slippage events may occur on the nascent strand compared to the template strand leading to synergy for +1 frameshift mutants according to the model of Fujii et al. (1999). Genetic and biochemical analyses of intragenic pol2-4, C1089Y double mutants that may allow us to distinguish between these two possibilities are underway.
Implications of the pol2-C1089Y mutator: While pol2-C1089Y had a dramatic effect on mutation rates when combined with defects in other genes, there was little if any consequence on its own. Thus, it appears that a neutral Polε mutation can have a dramatic effect in combination with another subtle mutation. The strong synergy for mutations in short homonucleotide runs seen when pol2-C1089Y is combined with Δmsh2 suggests that this polymerase allele could also lead to genome instability when combined with a partial defect in MMR. There are now examples of synergistic interactions between mild mutations in DNA metabolic genes leading to dramatic effects. Recently it was demonstrated that a mutation in the PCNA binding site of Rad27 was genetically neutral. However, cells were inviable when this mutation was combined with a subtle DNA polymerase δ proofreading defect in the heterozygous diploid pol3-01/POL3 (Garyet al. 1999).
The present results, in which we identify an apparently silent DNA polymerase ε mutation, have important implications for human disease. Homonucleotide runs are common in all organisms. MMR defects that destabilize homonucleotide runs have been identified in several tumor cell lines, particularly those from colon cancers. A recent report describes somatic instability associated with the adenomatous polyposis coli (APC) gene variant APC I1307K. This variant, which appears in several examples of familial colorectal neoplasia (Gryfe et al. 1998, 1999), results from a mutation that creates an A8 homonucleotide run from separate A3 and A4 runs. In 42% of carriers (who have the A8 run), compared to 4% in the general population (who have the A3XA4 run), a +1 frameshift occurs in the run resulting in gene inactivation. The tumor predisposition caused by the APC I1307K allele is not associated with a MMR deficiency (Prioret al. 1999). The identification of a Polε mutation that can cause a preponderance of +1 frameshift mutations in homonucleotide runs suggests that altered replication could contribute to the appearance of tumors even if the cells are MMR proficient. This example indicates that some disease susceptibilities resulting from genetic instability may not be due to loss of a single function, such as MMR. These genetic instabilities could arise by interactions between subtle or silent mutations in genes that can impact on mutation avoidance or mutation generation in at-risk motifs (ARMs) such as homonucleotide runs.
Acknowledgments
We are grateful to Drs. A. Greene and D. Gordenin for primers; to Drs. R. Kolodner, H. Araki, and A. Sugino for plasmids; and to Drs. W. Copeland, D. Gordenin, K. Lewis, K. Lobachev, K. A. Street, and Y. Pavlov for critical comments on the manuscript, and especially to Drs. Gordenin, Lewis, and Lobachev for important discussions during the course of this work.
Footnotes
-
Communicating editor: L. S. Symington
- Received January 31, 2000.
- Accepted April 20, 2000.
- Copyright © 2000 by the Genetics Society of America
