INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Cancer can be viewed as a genetic disease whereby the accumulation of mutations leads to the eventual activation of cellular oncogenes or inactivation of tumor suppressor genes (53). These events, in turn, confer growth or survival advantages to tumor cells. This framework predicts that genetic defects causing increased mutation rates should predispose cells to carcinogenesis. Such a prediction has been fulfilled by the discovery that inherited defects in DNA mismatch repair genes underlie the etiology of hereditary nonpolyposis colon carcinoma, a cancer predisposition syndrome (22, 44). Given this example, it seems likely that other genetic defects giving rise to mutator phenotypes should also contribute to carcinogenesis. One potential class of such mutator genes is the group of genes encoding DNA replication accessory factors, since these factors modulate the fidelity of DNA replication as well as participate in various aspects of DNA repair.

Proliferating cell nuclear antigen (PCNA) is a replication accessory factor encoded by the essential gene POL30 in Saccharomyces cerevisiae (4). PCNA forms a homotrimeric ring around the DNA that serves as a binding platform for DNA polymerases. In doing so, PCNA enhances the processivity of DNA polymerases (26, 51). PCNA homologues from various eukaryotes share a high degree of sequence identity (19), and despite the lack of significant primary sequence homology between the Escherichia coli DNA polymerase III processivity factor and PCNA homologues, the three-dimensional crystal structures of these two proteins are superimposible (24, 26). This strong evolutionary conservation suggests that information obtained about PCNA in lower eukaryotes may be applicable to more complex organisms.

In addition to serving as a polymerase processivity factor, PCNA modulates the fidelity of DNA synthesis in vitro (41) as well as interacts with factors involved in nucleotide excision repair (9), mismatch repair (MMR) (18, 52), and base excision repair (42). PCNA also participates in the processing of branched DNA structures, including those formed during lagging-strand DNA synthesis (31, 56). The abundance of PCNA-interacting proteins led to the suggestion that PCNA may serve as a "docking bay" for many proteins and, in the process, coordinate various aspects of DNA synthesis and repair (19). Of the above-mentioned processes, DNA polymerase fidelity, MMR, and branch structure processing are mutation suppressing mechanisms pertinent to the results presented in this report.

Three major factors determine the ability of a polymerase to conduct faithful DNA replication, and all three are modulated by PCNA (5, 16, 30, 41): the selection of nucleotides that properly pair with the template (nucleotide discrimination), the 3'5' exonuclease activity that removes misincorporated nucleotides (proofreading), and adherence to the template during DNA replication (processivity). Defects in nucleotide discrimination or proofreading could lead to base substitution formation via nucleotide misincorporation. Failure of the DNA polymerase to adhere to the template properly could result in polymerase slippage, causing frameshifts, deletions, or duplications (27). Since specific types of mutations are associated with particular DNA polymerase defects, the defects of a mutant polymerase may be inferred by an analysis of the types of mutations that it caused. For instance, the pol2-4 allele in S. cerevisiae encodes a proofreading-deficient form of DNA polymerase  that causes a roughly 10-fold increase in mutation rate. Not surprisingly, the mutation spectrum of pol2-4 is notable for an increased accumulation of base substitutions (40). Another example is the pol3-t allele, which encodes a temperature-sensitive (ts) form of DNA polymerase . This mutation facilitates the formation of deletions flanked by short repeats, suggesting that pol3-t increases polymerase slippage (48, 49).

In addition to modulating polymerase fidelity, PCNA is also required for MMR in a human crude extract system (52). MMR is an evolutionarily conserved process which corrects misincorporated bases that escape DNA polymerase proofreading. Cells completely defective in MMR exhibit a 10- to 1,000-fold increase in the rate of mutation accumulation. Most of these mutations occur as frameshift mutations in mononucleotide runs (13, 35, 47). In S. cerevisiae, MMR initiates with the binding of mismatches by the MSH2-MSH6 or the MSH2-MSH3 heterodimer. Subsequent to this binding, other factors are recruited to the mismatch, leading to the eventual nicking, degradation, and resynthesis of the DNA strand containing the mutagenic nucleotide (23). In E. coli, this strand discrimination is mediated by (i) transient undermethylation of newly synthesized GATC sites by the Dam methylase and (ii) activation of MutH, the MMR-initiating endonuclease, to cleave the unmethylated DNA strand at hemimethylated GATC sites (38). The mechanism of strand discrimination in eukaryotes, where DNA methylation is probably not involved, is less clear. The observation that PCNA functions in MMR at a step prior to strand excision (14, 52) raises the possibility that PCNA functions at the initiation stages of MMR and that this may involve coupling of mismatch repair proteins to the replication machinery by PCNA (11, 22, 28, 39).

In vitro studies suggest that PCNA may participate in the processing of branched DNA structures via its interaction with RAD27 (31, 56). RAD27 (also known as RTH1, FEN1, MF-1, and exonuclease IV) is an evolutionarily conserved protein that has both a 5' flap endonuclease and a 5'-to-3' exonuclease activity. These activities are crucial for the processing of Okazaki fragments during DNA replication (32, 55). S. cerevisiae rad27 mutants accumulate double-strand breaks (DSBs) and exhibit a strong mutator phenotype (47). The mutations that arise in rad27 mutants are predominately duplications where the region duplicated is flanked by short stretches of imperfect direct repeats. These duplications are believed to result from mutagenic repair of DNA strand breaks involving misannealing of short single-stranded DNA tails (21, 47). Similar duplications have been reported as inactivating somatic mutations in the adenomatous polyposis coli as well as the p53 gene (12, 47).

Since PCNA participates in many processes that normally contribute to mutation suppression, we screened a panel of previously isolated S. cerevisiae pol30 mutants for mutator phenotypes. Eleven pol30 mutator mutants were identified. Mutation rate, spectrum, and epistasis analyses suggest that PCNA functions in multiple mutation-suppressing processes, ranging from polymerase fidelity to MMR. Detailed characterization of one MMR-defective mutant (pol30-104) yielded genetic evidence consistent with the notion that PCNA plays an important role in strand discrimination during MMR.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

General genetic methods. YPD (yeast extract-peptone-dextrose), sporulation, SD (synthetic dextrose), 5FOA (5-fluoroorotic acid), and canavanine media were prepared as previously described (6). Transformations were also performed as previously described (6). All strains were propagated at 37°C. Chromosomal DNA preparations were isolated by using glass bead lysis (18a) or Puregene kits (Gentra). PCR was performed in 50- to 100-µl reactions containing 0.5 U of Klentaq DNA polymerase (Ab Peptides), 0.08 U of Pfu DNA polymerase (Stratagene), 20 ng of genomic DNA, and 10 pmol of each primer in 1× PCII buffer (Ab Peptides). Primers were synthesized by the Dana-Farber Cancer Institute Molecular Biology Core Facility or Cybersyn Corp (Lenni, Pa.). Unless otherwise noted, standard three-temperature PCR was performed. Prior to DNA sequencing, PCR products were purified by using QIAquick Spin PCR purification kits (Qiagen). All DNA sequencing was performed with a Perkin Elmer/Applied Biosystems 377 DNA sequencer and dye terminator chemistry according to the manufacturer's instructions.

Strain and plasmid construction. All haploid strains used were derivatives of S288C parental strains provided by Fred Winston (Harvard Medical School). Five series of strains, all derived from RDKY3023 (MATa ura3-52 leu21 trp163 his3200 lys2Bgl hom3-10 ade21 ade8), were constructed. The first series (RDKY3546 to RDKY3548) was derived by transforming RDKY3023 with TRP1 ARS-CEN plasmids bearing pol30-41, -45, and -52 (pBL245-41, pBL230-45, pBL230-52; reagents kindly provided by Peter Burgers, Washington University, St. Louis, Mo. [3]). The POL30 gene in the first series was disrupted with a hisG-URA3-hisG cassette by transforming with MluI-KpnI-digested pBL243 (also provided by Peter Burgers) to generate the second series of strains, consisting of RDKY3549 to RDKY3551. Proper disruptions were verified by PCR amplification with 5'-TCA CAC TGA TGT AGT GGT GG and 5'-CCT GTT CCA CCG GCC CAG GG. The third series (RDKY3552 to RDKY3561) was derived by replacing POL30 in RDKY3023 with LEU2-tagged pol30 alleles by transformation with SacI-digested pCH1572 (POL30), pCH1573 (pol30-100), pCH1575 (pol30-102), pCH1576 (pol30-103), pCH1577 (pol30-104), pCH1579 (pol30-105), pCH1580 (pol30-106), pCH1578 (pol30-108), pCH1637 (pol30-126), and pCH1638 (pol30-114) (2, 37). Proper integrants were confirmed by PCR amplification with 5'-CGA ATT GAC CTT CTA CTG GGA and 5'-TGT CGC CGA AGA AGT TAA GA. The fourth series (RDKY3543 and RDKY3562 to RDKY3569) was derived by disrupting the MSH2 gene of the third series with a hisG-URA3-hisG cassette by transformation with AatII-PvuII-digested pRDK351 (35). Proper integrants were confirmed by PCR amplification with 5'-TCA AGT CAA TAC TTA AAC GCC and 5'-GAT TCG GTA ATC TCC GAA CAG AAG G. The fifth series (RDKY3544 and RDKY3570 to RDKY3577) was derived by disruption the MSH6 gene of the third series with a hisG-URA3-hisG cassette by transformation with EcoRI-SphI-digested pEAI108 (provided by Eric Alani, Cornell University). Proper integrants were confirmed by PCR amplification with 5'-CCA TTG ATG CGG ACG AAA ACT CGG and 5'-ATT TGC AAA GGG AAG GGA TG. RDKY3588 was constructed by transforming RDKY3560 with pCH1071 (URA3 GAL). RDKY3578 was constructed by transforming RDKY3560 with pRDK762 (URA3 GAL-RAD27). RDKY3668 was constructed by transforming RDKY3552 with pRDK762 (URA3 GAL-RAD27).

MAT leu2 can1-51 hom3

MAT leu2 can1-51 hom3 LEU2.pol30-104

In vivo MMR assay. The phagemids and most methods used to construct the mismatch-containing heteroduplexes were as previously reported (25). The heteroduplexes were purified by either benzoylated naphthoylated DEAE-cellulose chromatography followed by exonuclease V digestion (33) or high-pressure liquid chromatography (1). Comparable results were obtained in assays using substrates purified by using either protocol. The heteroduplex DNA was transformed into appropriate S. cerevisiae strains via lithium acetate or electroporation transformation. The transformants were plated onto SD Ura plates supplemented with 6 mg of adenine per liter, incubated at 37°C, and scored after 5 days.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

POL30 mutants exhibit variable increases in spontaneous mutation rates. We screened a panel of previously isolated S. cerevisiae pol30 mutants (2, 3) for mutator phenotypes. Since many of the pol30 mutants were cold sensitive, mutation rates were assessed at 37°C. Three mutator assays were used: an assay that scores for arginine permease inactivation (CAN1), an assay that detects reversion of a +4 insertion in the LYS2 gene (lys2-Bgl), and an assay that detects reversion of a +1 insertion in the HOM3 gene (hom3-10) (13, 35, 47). Whereas the lys2-Bgl and the hom3-10 assays are particularly sensitive indicators of defective MMR, the CAN1 assay is less specific for defective MMR and more sensitive to other mutagenic pathways (6, 13, 35, 47, 50). Eleven of the twelve mutants examined exhibited elevated mutation rates in all three assays (Table 2, set A). Of these mutants, pol30-52 and -104 exhibited the most dramatic mutator phenotypes, particularly in assays sensitive to defective MMR (41- and 21-fold increases in lys2-Bgl, respectively; 243- and 324-fold increases in hom3-10, respectively). However, these increases were lower (Mann-Whitney test, P < 0.05) than those caused by an msh2 mutation, a mutation thought to completely inactivate MMR (82-fold in lys2-Bgl; 500-fold in hom3-10). These observations suggest that pol30-52 and -104 did not completely inactivate MMR. The comparable CAN1 mutation rates of the pol30-52, pol30-104, and msh2 mutants (15-, 26-, and 18-fold increases, respectively), then, suggests that pol30-52 and -104 caused mutagenic defects that are not related to MMR in addition to causing partial defects in MMR. The mutator phenotypes of the other pol30 mutants were more subtle. pol30-100, -108, -126, and -114 caused modest rate elevations that were most evident in the hom3-10 assay. pol30-41, -45, -102, -103, and -105 caused weak rate elevations in all three assays.

POL30 mutants can be divided into distinct categories based on CAN1 mutation spectra. Although the assays used in this study are differentially sensitive to various mutagenic defects, they are not completely specific. Thus, it is difficult to determine the mutagenic defects of pol30 mutants based solely on mutation rate analyses. Because the analysis of mutations accumulated in mutator mutants have been instrumental in deciphering the mechanism of mutagenesis (6, 8, 13, 35, 47, 48), we determined the mutation spectra of the various pol30 mutants. The CAN1 assay was selected for this purpose because it is sensitive to a variety of mutational events, including base substitutions, frameshifts, deletions, duplications, inversions, and translocations (6, 35, 47). Given the large number of pol30 alleles studied and the absence of rationale for focusing on a specific allele, we sequenced 20 to 30 CAN1-inactivating mutations isolated from each pol30 mutant (Tables 3 and 4).

8

10

The lys2-Bgl reversion spectrum of pol30-104 suggests a defect in DNA MMR. Since mutants that are completely defective in MMR (e.g., msh2, mlh1, and pms1) exhibit a characteristic lys2-Bgl reversion spectrum (8, 13, 35), we wished to determine whether such a spectrum could be seen in any of the group A mutants. pol30-104 was selected as a representative allele for this purpose. As shown in Fig. 1, all of the pol30-104 Lys⁺ revertants resulted from 1 frameshifts; 65% of these frameshifts occurred in a run of 6 A's (nucleotides 664 to 669), 20% occurred in a run of four C's (nucleotides 697 to 700), and 10% occurred in a run of five T's (nucleotides 689 to 693). The distribution of these frameshift mutations is similar to those observed in msh2, mlh1, and pms1 strains, further supporting the notion that pol30-104 causes defective MMR.

View larger version (22K): [in this window] [in a new window]   FIG. 1.   Lys2-Bgl reversion spectra from wild-type, msh2, rad27, and pol30 mutants. The sequence shown represents the reversion window (nucleotides 650 to 798) of the lys2-Bgl assay. The location of the nucleotide alteration resulting in functional restoration of LYS2 is indicated by |. Each independent 1 frameshift is indicated by a  above the deleted nucleotide. Duplications are indicated by (+) followed by the nucleotides inserted. Deletions of greater than one base pair are marked by () followed by the nucleotides deleted. The ratio indicated represents the proportion of each particular class of events. The wild-type and msh2 spectra are from reference 8; the rad27 spectra is from reference 47.

An in vivo MMR assay indicates that pol30-104, -108, and -126 cause defects in MMR. To directly assess the defects of group A mutants in MMR, various pol30 ura3-52 ade2 ade8 strains were transformed with either an A/C or a G/T mismatch-containing URA3 ADE8 plasmid (25). The mismatch resides in the ADE8 gene such that only one strand of the plasmid encodes a functional protein. If the mismatch is corrected prior to DNA replication, the resulting colony will be nonsectored and either red or white. However, if the mismatch is not corrected, then the two ADE8 alleles will segregate after replication, resulting in a red/white sectored colony. Thus, the efficiency of MMR in various pol30 mutants can be assessed by scoring for Ura⁺ sectored colonies. For comparison, we selected pol30-103 from group B and pol30-100.

lys2-Bgl duplications in pol30-126 mutants differ from those observed in rad27 mutants. The duplication mutations that arose in group B mutants were similar to those observed in rad27 mutants (47). Since PCNA is known to stimulate the endo/exonucleolytic activity of RAD27 in vitro (31, 56), one interpretation of these observations is that group B mutants are defective in stimulating RAD27 activity. To explore this possibility, we tested whether galactose-induced RAD27 expression in a pol30-126 strain would suppress duplication formation. pol30-126 was selected because it exhibited the highest rate of duplication accumulation in group B. For this analysis we used the lys2-Bgl assay because the small mutation window in this system (0.5 kb versus 2.1 kb for CAN1) minimizes the sequencing effort necessary to obtain a large sample size (13, 35).

Epistasis analyses with msh2 and msh6 mutations suggest that pol30 mutations also cause mutagenic defects that are distinct from defects in MMR. Except for group A mutants, the mutation spectra of pol30 mutants differed from that of a msh2 mutant, suggesting that pol30 mutations cause defects in processes distinct from MMR. Additionally, mutation rate analyses of some group A mutants (pol30-52 and -104) suggest that defective MMR constitutes only one of the mutagenic defects in these mutants. To explore these possibilities, we conducted an epistasis analysis between the various pol30 mutations and a msh2 mutation. In at least one mutator assay, double mutants containing a combination of msh2 and pol30-100, -103, -104, or -108 exhibited mutation rates significantly higher than that of a msh2 or a pol30 single mutant (Mann-Whitney test, P < 0.05) (Table 2, set b). This result indicates that pol30-100, -103, -104, and -108 caused defects in processes distinct from MMR.

A msh2 mutation rescues the synthetic lethality between pol30-104 and rad52. A series of genetic observations in E. coli raised an interesting hypothesis regarding the defect of group A mutants in MMR. The initial proposal of methyl-directed MMR in E. coli was based on the genetic observation that dam mutations were lethal in combination with recA mutations and that this lethality was suppressed by MMR-inactivating mutations, such as mutS (36). These observations were interpreted to mean that in the absence of a strand discriminating signal, the MMR machinery (e.g., MutH) aberrantly nicks both the parental and the daughter DNA strands, leading to DSBs which require RecA-mediated recombinational repair. When MMR is inactivated by a mutS mutation, DNA breaks are not generated, and RecA is no longer required for viability. The mechanism of daughter strand discrimination has yet to be described in eukaryotes. However, the recent observation that PCNA functions in MMR at a step prior to strand excision raises the possibility that it can facilitate strand discrimination (14, 52). If PCNA does facilitate strand discrimination in MMR, then defects leading to aberrantly initiated MMR may cause DNA breaks. Interestingly, a number of pol30 mutations, including the group A mutations, were reported to be synthetically lethal with mutations in the RAD52 series of genesgenes required for recombinational repair of DNA breaks (37). Thus, by analogy to the genetics of MMR in E. coli, a msh2 mutation might rescue the synthetic lethality between rad52 and group A mutations. Moreover, if the above analogy is appropriate, then msh2 should not rescue the synthetic lethality between rad52 and other pol30 mutations.

View larger version (83K): [in this window] [in a new window]   FIG. 2.   A msh2 mutation rescues the synthetic lethality between pol30-104 and rad52. (a and b) pol30-104 (RDKY3583), pol30-104 msh2 (RDKY3584), pol30-104 rad52 (RDKY3585) and pol30-104 msh2 rad52 (RDKY3586) strains bearing a URA3 POL30 plasmid were grown at 30°C in liquid SD Ura to saturation; 10-fold serial dilutions of each culture were spotted onto an SD Ura plate to determine viability (a) and an SD +5FOA plate to determine whether the URA3 POL30 plasmid was required for viability (b). (c and d) pol30-104 msh2 (RDKY3669) and three independent isolates of pol30-104 msh2 rad52 (RDKY3587), all bearing URA3 POL30 and TRP1 MSH2 plasmids, were grown at 30°C in liquid SD Ura Trp media to saturation; 10-fold serial dilutions of each culture were spotted onto an SD Ura Trp plate to determine viability (c) and an SD Trp +5FOA plate to determine whether the URA3 POL30 plasmid was required for viability in the presence of the TRP1 MSH2 plasmid (d).

A msh2 mutation decreases the hyper-rec phenotype caused by pol30-104. Since DNA breaks are known to be recombinagenic (43) and since pol30-104 strains accumulate DNA breaks (37), pol30-104 might be expected to cause a hyperrecombination (hyper-rec) phenotype. Moreover, if the DNA strand breaks in pol30-104 mutants occurred as a result of aberrantly initiated MMR, then the hyper-rec phenotype should be suppressed by a msh2 mutation. To test this, we used a diploid strain carrying two recessive alleles (can1-51 and hom3) on one copy of chromosome V; the other copy of chromosome V is wild type for both alleles (15, 17). Phenotypically, the strain is canavanine sensitive and prototrophic. However, if gene conversion at CAN1 occurs or if there is a crossover between CAN1 and the centromere followed by proper segregation, canavanine-resistant and homoserine prototrophic progeny will be obtained. On the other hand, loss of the wild-type chromosome V will lead to canavanine-resistant but homoserine auxotrophic progeny.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

In this report, we described the effects of 12 pol30 alleles on the rate of mutation accumulation. Eleven of the twelve pol30 mutations studied caused increased accumulation of base substitutions, frameshifts, deletions, duplications, or some combination of these types of mutations. Although the pol30 alleles studied here were previously shown to cause defects in DNA replication and to cause sensitivity to various DNA-damaging agents (2, 3, 37), we could not find any correlation between these phenotypes and the mutator phenotype of pol30 mutants. The lack of such correlation suggests that the pol30 mutants are simultaneously defective in multiple cellular processes. Consistent with this interpretation, many pol30 mutations caused more than one mutagenic defect.

Based on their mutator phenotypes, the pol30 mutants were divided into three groups. Because many pol30 mutants harbor multiple mutagenic defects, they were placed in more than one group (Table 5). Group A mutations (pol30-52, -104, -108, and -126) caused MMR defects. Two of the group A mutations (pol30-52 and -104) were previously reported to cause defects in MMR because they induce mutator phenotypes that were epistatic to MMR-inactivating mutations (18, 52). Our results extend these observations by demonstrating that pol30-52 and -104 strains exhibited mutation spectra indistinguishable from mutants completely inactivated for MMR, such as msh2 (35, 47). Moreover, we showed that pol30-104 strains are defective in the in vivo repair of plasmids containing mispaired bases. Using mutation spectra analysis and in vivo MMR defect as criteria, two additional MMR-defective POL30 alleles, pol30-108 and -126, were identified. pol30-104 and -108 were mutated at the same amino acid (A₂₅₁V for pol30-104 and A₂₅₁T for pol30-108). Although these two alleles were similarly defective in terms of cold sensitivity and cell cycle arrest, pol30-104 caused a significantly more severe MMS^s phenotype (2, 37) and a stronger mutator phenotype.

Although the mutator phenotypes of the pol30-52 and -104 strains were initially thought to result solely from defective MMR, two recent studies indicate that polymerase slippage also contributes to the pol30-52 mutator phenotype (20, 57). Our mutation spectra analyses as well as the msh2 and msh6 epistasis analyses of pol30-104, -108, and -126 suggest that these MMR-defective pol30 alleles, like pol30-52, are additionally defective in some aspect of replicative fidelity. Contrary to the results presented here, Johnson et al. (18) reported that pol30-104 caused a mutator phenotype that is epistatic to msh2, mlh1, and pms1. This discrepancy may be explained if the plasmid based dinucleotide instability assay used by Johnson et al. is more specific to MMR-related defects than our assays. The disagreement between our results and the reported epistatic relationship between pol30-104 and mlh1 in the CAN1 assay (18) may have resulted from differences in strain background or from differences between a mlh1 and a msh2 mutation.

Group B mutants (pol30-45, -103, -105, -126, and -114) exhibited increased accumulation of either deletions alone or a combination of deletions and duplications. All of the CAN1 duplications in pol30 mutants, like those seen in rad27 mutants, involved short repeated sequences at the breakpoint. Since PCNA stimulates the activity of RAD27 in vitro (31, 56), one interpretation of these observations is that group B mutants are defective in stimulating RAD27 activity. This interpretation suggests that group B mutants, like rad27 mutants, accumulate DSBs and that these breaks are repaired by misannealing of single-stranded DNA tails (21, 47). However, two lines of evidence are inconsistent with this idea. First, the lys2-Bgl duplications in pol30-126 mutants differed from those observed in rad27 mutants. Second, the CAN1 duplications in pol30-126 mutants were, on average, shorter than those observed in rad27 mutants. Finally, the duplications in pol30-126 could not be suppressed by the expression of RAD27 under a GAL10 promoter. Although the third observation could be explained by insufficient RAD27 expression, this explanation fails to account for the first two observations. One explanation that takes into account all three observations is that defective PCNA-DNA polymerase or PCNA-DNA interactions decreases the adherence of the polymerase to the template, thereby facilitating polymerase slippage. Consistent with this explanation, recent studies suggest that pol30-52, an allele defective in MMR, also induced replication slippage in the context of simple repeated sequences such as dinucleotide repeats (20, 57). Also, DNA polymerase mutants have been shown to accumulate repeat-mediated deletions and duplications (34, 48). While the above hypothesis is appealing, we cannot exclude the possibility that in the absence of wild-type PCNA, RAD27 may mediate the removal of some but not all of a flap structure. In vitro, however, RAD27 removes the entire flap structure in the absence of PCNA (32). We also cannot exclude that possibility that the pol30 mutants are defective in a yet to be identified repair system that specifically prevents deletion/duplication formation.

Group C mutants (pol30-100, -103, -105, -108, and -114) exhibited increased accumulation of base substitutions (Table 5) and synergistic mutator effects in the CAN1 assay when combined with msh6 mutations (Table 2, set c). The synergistic effect with msh6, a mutation thought to completely inactivate the repair of base-base mispairs, suggests that group C mutations increased polymerase misincorporation. Another interpretation of this synergy with msh6 is that the pol30 mutants are defective in MSH3-dependent MMR. This interpretation is unlikely because the mutator phenotype of msh3 msh6 double mutants differed from those of group C msh6 double mutants; msh3 msh6 double mutants were potent mutators in the hom3-10 and the lys2-bgl assay but were modest mutators in the CAN1 assay (increases of roughly 500-, 100-, and 30-fold, respectively, over wild-type rates) (35). The group C msh6 double mutants exhibited CAN1 mutation rates comparable to that of msh3 msh6 double mutants. However, the hom3-10 and lys2-bgl reversion rates of group C msh6 double mutants (ranging from increases of 6- to 15-fold over wild-type rates) were significantly lower than those of msh3 msh6 double mutants.

Since the crystal structure of PCNA has been solved (26), we wished to determine whether group A, B, and C mutations were localized to any particular region of PCNA. The amino acids altered by group A mutations (pol30-52, -104, -108, and -126) were located in three distinct regions. pol30-52 changed a residue located in the monomer-monomer interface region. pol30-104(-108) changed a residue in the beta sheets connecting the two monomer domains (i.e., the interdomain region). pol30-126 changed a residue in one of the alpha helices contacting DNA. Therefore, we could not define a single region of PCNA where amino acid substitutions caused MMR defects. Likewise, group B and C mutations were not localized to any particular region of the PCNA structure. We were also unable to find any correlation between the known biochemical defects of pol30 mutants and the severity of their defects in MMR. For instance, pol30-52 trimers exhibited a notable decrease in stability relative to pol30-104 trimers (7a). However, the two mutants were equally defective in MMR by the criteria established here.

Suppression of the pol30-104 rad52 synthetic lethality and the pol30-104-induced hyper-rec phenotype by MSH2 inactivation provides some insights into the mechanistic details of MMR. One interpretation of these results is that PCNA mediates a signal that targets newly synthesized DNA strands for excision during MMR. By analogy to the genetic interaction between mutations in dam, recA, and mutS in E. coli (36), in the absence of such a signal (as may be the case in a pol30-104 mutant), the MMR machinery might nick both the template and the daughter DNA strands. Alternatively, in the absence of a strand discrimination signal, the MMR machinery can nick the template strand. In the presence of the many nascent strand breaks induced by pol30-104 (37), these template strand nicks lead to the formation of DSBs. Another explanation is that PCNA couples MMR proteins to replication proteins. In this scenario, pol30-104 causes replication fork stalling when MMR is initiated such that the stalled replication fork is converted to a DSB (46). These models are not mutually exclusive. We are currently testing these possibilities by isolating additional suppressors of the pol30-104 rad52 synthetic lethality.

The observations reported here have a number of implications regarding the genetics of cancer susceptibility. First, the observation that many pol30 mutants are mutators suggests that missense mutations in the human PCNA could increase cancer susceptibility. This possibility is particularly intriguing when one takes into consideration the partial dominant mutator effect of some pol30 alleles, such as pol30-52 and -104. Second, though many pol30 mutants were by themselves weak mutators, their mutator effects were dramatically elevated in the presence of a msh6 mutation. This observation suggests that weak pol30 mutator alleles, which might exist as natural polymorphic variants in a population, could act as modifiers of partially defective MMR alleles, such as partial loss of function MSH2 alleles or MSH6 null alleles. Finally, pol30-induced duplications may lead to trinucleotide expansion, a process underlying the pathogenesis of many human neurological disorders (45). All of these possibilities await further examination.