|Home | About | Journals | Submit | Contact Us | Français|
The Saccharomyces cerevisiaeEXO1 gene encodes a 5′ exonuclease that participates in mismatch repair (MMR) of DNA replication errors. Deleting EXO1 was previously shown to increase mutation rates to a greater extent when combined with a mutator variant (pol3-L612M) of the lagging strand replicase, DNA polymerase δ (Pol δ), than when combined with a mutator variant (pol2-M644G) of the leading strand replicase, DNA polymerase ε (Pol ε). Here we confirm that result, and extend the approach to examine the effect of deleting EXO1 in a mutator variant (pol1-L868M) of Pol α, the proofreading-deficient and least accurate of the three nuclear replicases that is responsible for initiating Okazaki fragment synthesis. We find that deleting EXO1 increases the mutation rate in the Pol α mutator strain to a significantly greater extent than in the Pol δ or Pol ε mutator strains, thereby preferentially reducing the efficiency of MMR of replication errors generated by Pol α. Because these mismatches are closer to the 5′ ends of Okazaki fragments than are mismatches made by Pol δ or Pol ε, the results not only support the previous suggestion that Exo1 preferentially excises lagging strand replication errors during mismatch repair, they further imply that the 5′ ends serve as entry points for 5′ excision of replication errors made by Pol α, and possibly as strand discrimination signals for MMR. Nonetheless, mutation rates in the Pol α mutator strain are 5- to 25-fold lower in an exo1Δ strain as compared to an msh2Δ strain completely lacking MMR, indicating that in the absence of Exo1, most replication errors made by Pol α can still be removed in an Msh2-dependent manner by other nucleases and/or by strand displacement.
Because the two strands of duplex DNA are anti-parallel and DNA polymerases only copy DNA in the 5′ to 3′ direction, coordinated replication of the eukaryotic nuclear genome is intrinsically asymmetric, with a leading strand replicated first and a lagging strand replicated slightly thereafter as a series of ~200 base Okazaki fragments (reviewed in ). We are interested in understanding relationships between this asymmetry and the fidelity of DNA replication. To investigate this, we are using Saccharomyces cerevisiae strains containing mutant alleles of the POL1 (Pol α), POL2 (Pol ε) and POL3 (Pol δ) genes. These alleles, pol1-L868M [2, 3], pol2-M644G  and pol3-L612M ( and references therein), encode replicases with single amino acid replacements at the polymerase active site that retain high replicative capacity but have reduced fidelity. Strains harboring these alleles have elevated spontaneous mutation rates and a specific error bias signature, thereby identifying the polymerases responsible for generating the majority of replication errors in vivo. The patterns of mutagenesis in these strains [4–7] and the strand-specific incorporation of ribonucleotides during replication by Pol ε [6, 7], suggest that under normal circumstances, Pol ε is the primary leading strand replicase, while Pol δ primarily participates in lagging strand replication after Pol α-primase initiates the synthesis of Okazaki fragments.
DNA replication fidelity is determined by the nucleotide selectivity of pol α, δ and ε, by the 3′ exonucleolytic proofreading activities intrinsic to Pols δ and ε (but not Pol α), and by mismatch repair (MMR) of replication errors that escape proofreading. MMR begins when a mismatch generated during nuclear DNA replication is recognized either by Msh2-Msh6 (MutSα), which recognizes single base-base and small insertion-deletion (indel) mismatches, or by Msh2-Msh3 (MutSβ), which recognizes indel mismatches with a specificity that is partially redundant with Msh2-Msh6 ([8–10] and reviewed in ). Mismatch recognition initiates a series of downstream steps that ultimately remove the replication error from the nascent strand and allow new DNA to be correctly synthesized. In addition to the mismatch, the MMR machinery requires a signal to direct excision to the newly synthesized strand containing the replication error. Studies in vitro show that the signal can be a nick or gap located either 3′ or 5′ to the mismatch, with the protein requirements for MMR differing somewhat depending on the location of the DNA ends relative to the mismatch [12–14]. The origin and exact nature of the strand discrimination signal used for MMR in vivo remains uncertain. Attractive possibilities for this signal include the 3′ ends of growing chains at the replication fork and/or the 5′ ends of Okazaki fragments that are transiently present during lagging strand replication.
To probe the enzymology of MMR in vivo and gain insight into the nature of the strand discrimination signal, we are examining the extent to which MMR specificity and efficiency varies as a function of several variables, most especially the polymerase that generates the error. This focus is motivated by a specific prediction that emerges from several related observations. Early studies in E. coli demonstrate that MMR most efficiently corrects those mismatches generated at the highest rates during replication, i.e., transition and indel mismatches [15–17]. If reciprocity between production of mismatches and their repair is evolutionarily conserved in eukaryotes, then MMR should not only more efficiently correct transition and indel mismatches, it may also more efficiently correct mismatches made by Pol α as compared to those made by Pol δ, because Pol α is naturally exonuclease deficient and generates mismatches at substantially higher rates than does proofreading-proficient Pol δ . Early studies also showed that the efficiency of MMR in E. coli is inversely proportional to the distance between the mismatch and the strand discrimination signal . In the initial study showing that PCNA was required for eukaryotic MMR at a step prior to mismatch excision, we proposed  that PCNA could physically link MMR and replication in a manner that would allow DNA ends associated with replication to serve as strand discrimination signals. That study was followed by another  which led to the more specific hypothesis that the 5′ ends of Okazaki fragments could signal for MMR of replication errors in the nascent lagging strand. If the relationship between distance and bacterial MMR efficiency is conserved in eukaryotes, and if the 5′ ends of Okazaki fragments can serve as strand discrimination signals, this too predicts that MMR might more efficiently correct mismatches made by Pol α as compared to mismatches made by Pol δ. This is because mismatches made as Pol α initiates Okazaki framents will be closer to DNA 5′ termini than would more internal mismatches made by Pol δ. We initially tested this prediction by comparing mutation rates in L612M Pol δ and L868M Pol α strains that were either MMR proficient or deleted for MSH2 (msh2Δ) and therefore lacked both Msh2-Msh6-dependent and Msh2-Msh3-dependent MMR activity. Pairwise comparisons of specific single base mutation rates in these strains provided an apparent MMR efficiency for errors made by each polymerase. We found that mismatches made by Pol δ are repaired efficiently, but the equivalent single base-base mismatches made by Pol α appear to be corrected even more efficiently . Possible explantions for this higher efficiency include use of the 5′ ends of Okazaki fragments to initiate mismatch removal, either by strand displacement  or by a 5′ exonuclease.
Based on genetic evidence , we previously proposed that one candidate for an exonuclease to perform mismatch excision initiated at the 5′ end of an Okazaki fragment is exonuclease 1 (Exo1), which is well known to participate in eukaryotic MMR [24, 25]. In support of this idea, Kolodner and colleagues  recently reported that deletion of EXO1 increases the mutation rate of the pol3-L612M mutator strain to a greater extent than for the pol2-M644G mutator strain. Here we extend this effort by examining the mutagenic consequences of deleting EXO1 (exo1Δ) in all three strains, pol1-L868M, pol2-M644G and pol3-L612M. The hypothesis that the 5′ DNA ends of Okazaki fragments are used for MMR predicts that deleting Exo1 should reduce the efficiency of MMR of errors made by Pol α even more than for errors made by Pol δ. This is exactly what we observe. We also compare the results to those seen upon complete loss of all Msh2-dependent MMR in the same polymerase mutator strains [6, 22]. The results lead to three interpretations. (I) Exo1 has a much greater role in repairing lagging strand replication errors as compared to leading strand replication errors, supporting a similar conclusion by Kolodner and colleagues  and extending it to the second essential lagging strand replicase. (II) The 5′ ends of Okazaki fragments are entry points for 5′ to 3′ excision of mismatches by Exo1 during MMR in vivo, and they may also serve as strand discrimination signals. (III) Like DNA replication, MMR enzymology in vivo can differ on the lagging and leading strands.
All strains used in this study are isogenic derivatives of strain Δ|(2)|-7B-YUNI300 (MATa CAN1 his7-2 leu2-Δ::kanMX ura3-Δ trp1-289 ade2-1 lys2-ΔGG2899-2900) . Mutator alleles; pol1-L868M, pol2-M644G, and pol3-L612M have been described previously [2–5]. Heterozygous EXO1/exo1Δ diploids were generated by PCR based targeted gene-disruption. Deletion of EXO1 was verified by phenotype and by PCR across the disrupted region, and haploids were obtained from tetrad dissection. Measurements of spontaneous mutation rates by fluctuation analysis were as described previously [4, 5]. For each ura3 mutant sequenced, an independent colony was patched to YPDA and then replica plated to media containing 5-FOA. Genomic DNA from a single 5-FOA-resistant colony from each patch was isolated, and the URA3 gene was amplified by PCR and sequenced.
To determine if differences in mutation rates between strains are significant we used a one-sided Mann-Whitney t-test.
Mutation rates for resistance to 5-fluoro-orotic acid (5-FOA) were determined in EXO1 versus exo1Δ strains that encode either wild type DNA polymerases or mutator variants of Pol α (pol1-L868M), Pol δ (pol3-L612M) or Pol ε (pol2-M644G). Compared to the wild type (POLEXO1) strain, the mutation rate of the exo1Δ single mutant strain was elevated by 12-fold, and as previously observed, the mutation rates in the pol1-L868M, pol2-M644G and pol3-L612M [4, 5] single mutant strains were each elevated by a few-fold (Fig. 1A). The mutation rate in the double mutant pol1-L868M exo1Δ strain was increased by 140-fold, to 51 ×10−7. This is greater than the 64-fold increase observed in the pol3-L612M exo1Δ strain (rate = 23 × 10−7). The rates in these double mutant strains are much more than the sum of rates in the individual single mutant strains, consistent with the idea that two processes are acting in series, i.e., Exo1-dependent MMR is correcting replication errors that escape the fork. In contrast, the mutation rate in the pol2-M644G exo1Δ strain was 5.2 × 10−7, which is remarkably close to the sum of the increases observed for the pol2-M644G and exo1Δ single mutants (5.7 × 10−7). When mutation rates were measured at the CAN1 locus in these same strains, similar relative differences among the double mutant strains were obtained (Fig 1B). At both URA3 and CAN1, the rates in the pol1-L868M exo1Δ strain were significantly higher than the rates in the pol3-L612M exo1Δ and pol2-M644G exo1Δ strains (p values < 0.05, Fig. 1A).
The above interpretation depends on a demonstration that the mutations arising in the pol1-L868M exo1Δ and pol3-L612M exo1Δ strains are characteristic of partial loss of MMR of DNA replication errors rather than a defect in other DNA transactions in which Exo1 participates . One advantage of the polymerase variants used here is that, in the absence of Msh2-dependent MMR, they have mutational signatures characteristic of common replication errors . To determine if these signatures are observed in the exo1Δ strains, we examined the DNA sequence changes in the URA3 gene responsible for 5-FOA-resistance, in independent mutants collected from the pol1-L868M exo1Δ and pol3-L612M exo1Δ strains. The majority of mutants did indeed contain single base substitutions and deletions (Table 1), and these were non-randomly distributed throughout the URA3 open reading frame (Figure 2).
Focused on the hypothesis that deleting Exo1 would reduce the efficiency of MMR of lagging strand replication errors made by Pol α more than for errors made by Pol δ, the data in Figures 1 and and22 were used to calculate specific mutation rates (Table 1 and Fig. 3A) and MMR correction factors (Fig. 3B) for three classes of mutations that are signatures of a complete loss of MMR in pol1-L868Mmsh2Δ strains . These classes are (i) total single base errors (base substitutions and indels), (ii) total G to A transitions, which comprise 74% of all single base errors in pol1-L868Mmsh2Δ and pol3-L612M msh2Δ strains and are generated by misincorporation of dTTP opposite template G during lagging strand replication by Pol α (pol1-L868M) or Pol δ (pol3-L612M) and (iii) G to A transitions at characteristic hotspots in the URA3 reporter gene (Figure 2 and and3B)3B) . Mutation rates for these events are shown on a log scale in Figure 3A, where rates for MMR-proficient strains are depicted as open bars, rates for exo1Δ strains are represented by open + grey bars, and rates for msh2Δ strains are represented by open + grey + black bars. Dividing the rates in the exo1Δ strains by the rates in the MMR-proficient strains provides the correction factor (CF values in Fig. 3B) for Exo1-dependent MMR of each type of error. Correction efficiencies for these mismatches in the pol1-L868M strain range from 43 to 490. These values are consistently higher than the 9- to 15-fold correction efficiencies for the same classes of mismatches generated in the pol3-L612M strain (Fig. 3B), thus supporting the hypothesis.
Dividing the mutation rates in the pol1-L868M exo1Δ strain by the rates in the pol1-L868M msh2Δ strain that is completely defective in MMR reveals yields values of 4% to 19% (Fig. 3C). Thus efficient MMR still occurs in the absence of Exo1.
About 10% of Okazaki fragment synthesis is catalyzed by Pol α. This presents a strong challenge to nuclear genome stability because Pol α is a naturally proofreading-deficient replicase that is substantially less accurate than is proofreading-proficient Pol δ. An early idea for eliminating errors made by Pol α was that the mismatches it generates are removed by strand displacement synthesis during normal Okazaki fragment maturation, a process that historically has not invoked participation of MMR proteins . However, canonical Okazaki fragment maturation proteins alone are insufficient to remove many Pol α errors, as clearly illustrated by the fact that the mutation rate of the pol1-L868M strain is very strongly elevated by deleting Msh2 (see [2, 5] and Fig. 3). Now, in addition to Msh2, the present study demonstrates the involvement of Exo1 in MMR of Pol α errors. That Exo1 is indeed correcting Pol α replication errors, rather than affecting mutation rates via another DNA transaction, is clearly illustrated by the increase in single base mutations characteristic of loss of MMR in the Pol α mutator strain (Figs. 2 and and3).3). Because mismatches generated by Pol α are closer to the 5′ ends of Okazaki fragments than are mismatches made by Pol δ, the stronger mutator effect of deleting Exo1 in the pol1-L868M strain strongly support the hypothesis that the 5′ ends of Okazaki fragments are entry points for Exo1-dependent excision of a mismatch, and may possibly serve as strand discrimination signals for MMR. The fact that deleting EXO1 is much less mutagenic than deleting MSH2indicates that Pol α errors can also be corrected by an Exo1-independent MMR process. This repair may involve partially redundant nucleases, and/or a strand displacement mechanism to remove mismatches [13, 14, 23, 30–32]. The possibility that Exo1 is redundant with other exonucleases has been suggested before [11, 32–34], and our data extend this idea to include replication errors made by both lagging strand replicases. Finally, the fact that Exo1 influences MMR of errors made by the two lagging strand replicases to a greater extent than the leading strand replicase illustrates that, like DNA replication, eukaryotic DNA mismatch repair is enzymologically asymmetric.
We thank Kasia Bebenek and Dmitry Gordenin for thoughtful comments on the manuscript, and the NIEHS Molecular Genetics Core Facility for sequencing mutants. This work was supported by Project Z01 ES065089 to TAK from the Division of Intramural Research of the National Institutes of Health, National Institute of Environmental Health Sciences.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.