DNA polymerase ε leading strand signature mutations result from defects in its proofreading activity

The evidence that purified pol2-M644G DNA polymerase (Pol)ε exhibits a highly elevated bias for forming T:dTTP mispairs over A:dATP mispairs and that yeast cells harboring this Polε mutation accumulate A > T signature mutations in the leading strand have been used to assign a role for Polε in replicating the leading strand. Here, we determine whether A > T signature mutations result from defects in Polε proofreading activity by analyzing their rate in Polε proofreading defective pol2-4 and pol2-M644G cells. Since purified pol2-4 Polε exhibits no bias for T:dTTP mispair formation, A > T mutations are expected to occur at a much lower rate in pol2-4 than in pol2-M644G cells if Polε replicated the leading strand. Instead, we find that the rate of A > T signature mutations are as highly elevated in pol2-4 cells as in pol2-M644G cells; furthermore, the highly elevated rate of A > T signature mutations is severely curtailed in the absence of PCNA ubiquitination or Polζ in both the pol2-M644G and pol2-4 strains. Altogether, our evidence supports the conclusion that the leading strand A > T signature mutations derive from defects in Polε proofreading activity and not from the role of Polε as a leading strand replicase, and it conforms with the genetic evidence for a major role of Polδ in replication of both the DNA strands.

The evidence that purified pol2-M644G DNA polymerase (Pol)ε exhibits a highly elevated bias for forming T:dTTP mispairs over A:dATP mispairs and that yeast cells harboring this Polε mutation accumulate A > T signature mutations in the leading strand have been used to assign a role for Polε in replicating the leading strand. Here, we determine whether A > T signature mutations result from defects in Polε proofreading activity by analyzing their rate in Polε proofreading defective pol2-4 and pol2-M644G cells. Since purified pol2-4 Polε exhibits no bias for T:dTTP mispair formation, A > T mutations are expected to occur at a much lower rate in pol2-4 than in pol2-M644G cells if Polε replicated the leading strand. Instead, we find that the rate of A > T signature mutations are as highly elevated in pol2-4 cells as in pol2-M644G cells; furthermore, the highly elevated rate of A > T signature mutations is severely curtailed in the absence of PCNA ubiquitination or Polζ in both the pol2-M644G and pol2-4 strains. Altogether, our evidence supports the conclusion that the leading strand A > T signature mutations derive from defects in Polε proofreading activity and not from the role of Polε as a leading strand replicase, and it conforms with the genetic evidence for a major role of Polδ in replication of both the DNA strands.
The "division of labor" model and designation of DNA polymerase (Pol) ε as the leading strand replicase and of Polδ as the lagging strand replicase has been derived from studies involving mutator alleles of yeast Polε and Polδ and their effects on the distribution of leading or lagging strand mutations. For instance, yeast cells harboring the Polε pol2-M644G allele, whose encoded polymerase generates dTTP:T mispairs with an 40-fold bias over dATP:A mispairs, exhibit an increased incidence of spontaneous A to T signature mutations in URA3 integrated near ARS306 (1) that can be ascribed to T:dTTP mispair formation in the leading strand. A similar study with the Polδ pol3-L612M allele indicated the prevalence of lagging strand signature mutations consistent with the mispair formation bias exhibited by this Pol3 allele (2). However, in extensive genetic studies in different yeast strains, we subsequently provided evidence contradictory to the "division of labor" replication model, wherein L612M-Polδ generated errors occur on both the leading and lagging DNA strands in pol3-L612M msh2Δ strains (3). We postulated that a more proficient removal of errors by mismatch repair (MMR) from the leading strand accounts for the lack of L612M-Polδ specific errors on this strand and concluded from these studies that Polδ replicates both the leading and lagging DNA strands (3).
The four subunit yeast Polε holoenzyme is comprised of the Pol2 catalytic subunit and the Dpb2, Dpb3, and Dpb4 accessory subunits. While Dpb3 and Dpb4 are not essential (4,5), deletion of either the Pol2 or Dpb2 subunits leads to cell inviability (6,7). Within the Pol2 protein, the N-terminal half encompasses the active polymerase and the extreme C-terminus harbors a zinc-finger motif that is involved in binding the Dpb2 subunit. Importantly, the essential role of Pol2 lies in its ability to bind Dpb2, whereas the N-terminal catalytic polymerase domain of Pol2 is dispensable, although cells grow slowly (8). The Dpb2 subunit also binds directly to GINS (9, 10), a component of the CMG helicase that encircles and travels on the leading strand in the 3 0 →5 0 direction, unwinding the replication fork. Thus, via assembly of the CMG complex, the Pol2 C-terminus plays an essential role in replication by promoting origin firing and DNA unwinding (9,11,12).
Extrapolating from our genetic evidence that Polδ replicates both the leading and lagging DNA strands (3), we hypothesized that leading strand A > T signature mutations in pol2-M644G reflect Polδ misinsertions which escape proofreading by Polε 3'→5 0 exonuclease. To verify this hypothesis, in this study, we determine the rate of A > T signature mutations in Polε proofreading defective pol2-M644G and pol2-4 mutants wherein the pol2-4 mutation abolishes Polε proofreading, and the pol2-M644G mutation impairs mispair recognition (13) rendering proofreading ineffective. However, compared to the highly elevated bias of purified pol2-M644G Polε for forming T:dTTP mispair over the reciprocal A:dATP mispair, purified pol2-4 Polε exhibits no bias for T:dTTP mispair formation (14,15). Hence, if A > T signature mutations in the leading strand resulted from the role of Polε as a leading strand replicase, A > T signature mutations would occur at a much lower rate in pol2-4 cells than in pol2-M644G cells. However, if A > T signature mutations were derived from a role of Polε proofreading activity, then these mutations would occur at nearly the same rate in the pol2-4 strain as in pol2-M644G. Furthermore if A > T signature mutations were due to Polε role in leading strand replication, then there would be no need for the PCNA ubiquitination-dependent recruitment of Polζ for their formation-given the very high proficiency of pol2-M644G Polε for promoting synthesis from T:dTMP mispairs. Our evidence that A > T signature mutations in URA3 occur at the same rate in pol2-M644G and pol2-4 strains and that PCNA ubiquitination and Polζ are required for their formation supports the conclusion that the prevalence of leading strandspecific mutations does not arise from a role of Polε in replication of this strand; rather, it derives from the role of Polε proofreading activity in the removal of Polδ misinsertions on the leading strand.

Results
Leading strand signature mutations in pol2-M644G are dependent upon PCNA ubiquitination and Polζ In both lacZ and steady-state kinetic DNA polymerase fidelity assays, mutant pol2-M644G Polε has been shown to exhibit an 40-fold bias for the misincorporation of dTTP opposite template T than for the complementary dATP opposite template A (1). Since yeast cells that harbor the pol2-M644G mutation exhibit an elevated rate of spontaneously arising A > T hotspot mutations, namely A686T and A279T, in a URA3 reporter gene when integrated into the antisense orientation (OR2) to the left of ARS306 (1-3) (Fig. 1); these A > T mutations have been proposed to arise from T: T mispairs formed during replication of the leading strand by Polε. As shown in Table 1, the pol2-M644G strain exhibits a URA3 mutation rate 24-fold higher than WT cells. To examine the specific effect on rates of the A686T and A279T signature mutations, we determined the rates of these mutations through sequence analysis of ura3 mutations arising in a large number of independent cultures. As shown in Table 2, the rate of A > T mutations is extremely elevated in the pol2-M644G strain compared to WT (1100 fold increase).
Since Polζ is involved in DNA damage-induced and spontaneous mutation generation (16), and since it is a very proficient extender of synthesis from mispaired termini (17), we next examined whether Polζ was required for spontaneous signature mutations generated in the pol2-M644G strain. We find that deletion of the catalytic subunit of Polζ (rev3Δ) in pol2-M644G cells results in an 4-fold reduction in the URA3 spontaneous mutation rate compared to that in the pol2-M644G strain (Table 1). When examined for specific A > T signature mutations, rev3Δ reduces the rate of A686T mutations in pol2-M644G by 4-fold, as was the reduction in the overall A > T mutation rate ( Table 2). Since PCNA ubiquitination is required for Polζ function in cells (16), we next examined the effect of the pol30-119 mutation, which harbors an Arg mutation at Lys164 and thus inhibits PCNA ubiquitination (18,19). Although the overall drop in the spontaneous mutation rate of URA3 in pol2-M644G pol30-119 was similar to that found in the pol2-M644G rev3Δ strain (Table 1), there was a more pronounced effect on the signature A > T mutations. For instance, signature A686T mutation rates in pol2-M644G pol30-119 dropped by nearly 8-fold, and the overall rate of A > T mutations was also reduced by 8-fold in pol2-M644G pol30-119 (Table 2). When we examined signature mutation rates in pol2-M644G cells harboring both the rev3Δ and pol30-119 mutations, the rates were similar to those in the pol2-M644G pol30-119 strain, indicating that rev3Δ and pol30-119 act epistatically in pol2-M644G dependent A > T hotspot mutation formation (Table 2). Altogether, we deduce from our data (Table 2) that the formation of leading strand signature mutations in URA3 in pol2-M644G entails a major PCNA ubiquitination and Polζ dependent pathway (Fig. 2), and suggest that an alternative Polζ and PCNA ubiquitination independent pathway would account for the residual A > T signature mutations that remain in the absence of PCNA ubiquitination or Polζ.
The exonuclease defective pol2-4 mutation confers a similar rate of signature mutations as pol2-M644G We and others have previously observed A686T and A279T hotspot mutations occurring in the URA3-OR2 reporter gene in strains harboring the pol2-4 mutation, defective in Polε 3'→5 0 proofreading exonuclease (3,20). This was unexpected since purified Pol2-4 Polε does not exhibit a bias for the generation of dTTP:T mispairs over dATP:A mispairs (14,15). To examine this further, we determined the rates of A > T signature mutations in the pol2-4 strain. The spontaneous forward mutation rate in URA3 in the pol2-4 strain was 44fold higher than the wild type strain (Table 3). Remarkably, the rate of specific A > T signature mutations was similar to that in the pol2-M644G strain. For instance, the rate of A686T formation was 15.8 × 10 −8 in the pol2-M644G strain ( Table 2) and 14.3 × 10 -8 in the pol2-4 strain ( Table 4). The A279T mutation rate in the pol2-M644G and pol2-4 strains was 4.0 × 10 −8 and 6.0 × 10 −8 , respectively (Tables 2 and 4). Overall, compared to the WT strain, A > T mutations were elevated   1100-fold in the pol2-M644G strain, and 1300-fold in the pol2-4 strain (Tables 2 and 4).
A>T signature mutations in pol2-4 are dependent upon PCNA ubiquitination and Polζ Since the formation of pol2-M644G dependent A > T signature mutations requires PCNA ubiquitination and Polζ, we next examined whether PCNA ubiquitination and Polζ were also required for pol2-4 dependent signature mutations. As shown in Table 3, the spontaneous URA3 forward mutation rate in pol2-4 was lowered 7 to 8-fold by either the rev3Δ, pol30-119, or the rev3Δ pol30-119 double mutation. The overall rate of A > T mutations dropped by 13-fold in the pol2-4 rev3Δ pol30-119 strain, similar to that in the pol2-4 rev3Δ or in the pol2-4 pol30-119 strains (Table 4). Our results that the overall rate of A > T mutations in the pol2-4 rev3Δ pol30-119 strain is reduced to the same extent as in the pol2-4 rev3Δ or pol2-4 pol30-119 strains concur with an epistatic interaction of rev3Δ with pol30-119 in pol2-4 Polε dependent mutation generation (Table 4). Altogether, we infer from these data that A > T signature mutation formation observed in the pol2-4 strain occurs via a pathway involving PCNA ubiquitination and Polζ (Fig. 2); and another pathway that operates independently of PCNA ubiquitination and Polζ would account for the mutations that remain. The sequence data for the various strains are shown in Figures 3-6.

Discussion
Signature mutations in pol2-M644G do not signify Polε role in leading strand replication Polε has been implicated as the leading strand replicase, in part from the evidence that the elevated rate of A > T signature mutations observed in pol2-M644G yeast strains correlates with an extreme bias of M644G Polε for the formation of dTTP:T mispairs that would occur in the leading strand. During replication, M644G Polε would therefore have a high propensity for dTTP:T mispair formation and for proficiently extending synthesis from those mispairs, rather than proofread them. However, we find that these signature mutations are Polζ-dependent and they require ubiquitination of PCNA. If A > T mutations were generated by pol2-M644G Polε as the leading strand replicase via the formation and extension of synthesis from dTMP:T mispairs, then there would have been no need for Polζ. Thus, by that measure, i.e. the formation of leading strand signature mutations, the requirement of Polζ would suggest that it too is a major replicase for the leading strand, which it is not. Furthermore, the reduction in URA3 signature mutations by pol30-119 implies that their formation depends upon the ubiquitination of PCNA, a process not required for replication of the leading strand. Thus, the   Polε proofreads Polδ errors on the leading strand high incidence of spontaneously arising A > T signature mutations in the pol2-M644G yeast strain is not an indicator of the role of Polε as the major leading strand replicase.
Leading strand signature mutations result from lack of removal of Polδ misinsertions in the absence of proofreading by Polε Remarkably, the yeast pol2-4 mutation confers a nearly identical increase in the rate of A > T signature mutations in the URA3 reporter gene as the pol2-M644G mutation. Thus, the A > T mutations in pol2-M644G cells which were thought to have resulted from the 40-fold bias of M644G Polε for dTTP:T mispair formation (1) arise at the same high rate in pol2-4 cells, despite the fact that this exonuclease deficient polymerase exhibits no bias for generating dTTP:T mispairs (14,15). Hence, these pol2-4 dependent leading strand-specific A > T signature mutations in URA3 must derive from a process that is not dependent upon Polε mispair insertion, but are rather dependent upon the lack of removal of dTTP:T mispairs already present in the leading strand. The only way to explain these results is that A > T mutations in pol2-M644G and pol2-4 cells derive from a major role of Polδ in the replication of the leading strand (3), and that they reflect Polδ mis-insertions which escape proofreading by its own 3'→5 0 exonuclease and which are recalcitrant to removal by MMR (21). Thus, A > T signature mutations would accumulate on the leading strand in these Polε mutants because of the reduced ability of pol2-M644G Polε to recognize (13) and the inability of pol2-4 Polε to proofread such Polδ generated   T  T  TT   T   TTT   T  TT  TTT   T   T   T  T  A  A  A   T   T   T   A   A   A  TT  T   T   T  +  G  A   A   T  TT   T  T   T   A   TTTTTTTTT  TTTTTTTTTT  TTTTTTTTTTTTT  TTTTTTTTTTTTTT   T   T   T  TTTTTTT  TTTTTTTT  TTTTTTTT  Ty  mispairs, and not because mutant Polε generates dTTP:T mispairs at a high rate during replication.

Somatic Polε proofreading domain mutations in cancers
The conclusions of this study imply that the high prevalence of mutations that occur in a large variety of cancers harboring somatic Polε proofreading domain mutations (22)(23)(24)(25)(26)(27)(28)(29) derive from PCNA ubiquitination and Polζ dependent extension of synthesis from Polδ generated mispairs on the leading strand that do not get removed in the absence of Polε proofreading function. Furthermore, the indispensability of Polδ for replication of both the DNA strands (3) explains the dearth of somatic Polδ proofreading domain mutations; and the requirement of Polε proofreading activity for the removal of specific Polδ generated mispairs on the leading strand explains the high prevalence of somatic Polε proofreading domain mutations that occur in cancer genomes (29).

Dispensability of Polε polymerase activity for viability
In striking contrast to the indispensability of Polδ polymerase activity for viability (30)(31)(32)(33), the lack of N-terminal Polε polymerase domain supports viability, although cell growth is affected (8). Nevertheless, the observation that the lethality of pol2Δ cells is efficiently rescued by the pol2 mutation that is defective in its polymerase activity and in its PCNA binding PIP domain (34) reinforces the dispensability of Polε polymerase activity for cell survival. These results and the evidence that Polδ signature mutations occur on both DNA strands in pol3-L612M msh2Δ (3,35) and that defects in Polε proofreading activity account for Polε leading strand signature mutations in pol2-M644G or pol2-4 cells (this study) can be explained only if Polδ replicated both the DNA strands and Polε contributed primarily to DNA repair roles on the leading strand.

Yeast strains
All genetic experiments were carried out in isogenic derivatives of the S288C-based yeast strain BY4741 (MATa his3Δ1 leu2Δ0 met15Δ0 ura3Δ0) (36). The pol2-4 and pol2-M644G mutations were integrated into the yeast genome by direct replacement of the wild-type POL2 gene using either pPOL550 or pPOL520, respectively (3). The pol2-pip (FF1199,1200AA) mutation was generated by PCR using mutagenic oligonucleotides, and the resulting PCR fragment was subcloned into the Pol2 direct replacement vector, generating pPOL551. The pol2-M644G, pip double mutant replacement plasmid, pPOL779, was constructed similarly. Yeast strains harboring the pol2 M644G, pol2 pip, and pol2-M644G pip mutations were generated by transformation with the respective plasmids digested with FspI/SwaI restriction endonucleases, and selected for growth on synthetic complete (SC)-uracil media. Excision of the URA3 selectable marker integrated into the 5 0 UTR of pol2 was selected by plating on media containing 5-fluoro-orotic acid (FOA) and confirmed by PCR analysis of yeast genomic DNA. To generate yeast harboring the pol2-4 pip double mutation, the pol2 pip yeast strain YPO-861 was transformed with pPOL550 digested with EcoRI, which integrates the pol2-4 mutation while leaving the pol2 pip mutation intact. The rev3Δ mutation was generated by transformation with plasmid pRev3.75 digested with EcoRI/BamHI and the pol30-119 mutation was integrated into the genome by gene replacement with plasmid pPCNA1.44 digested with Asp718/XbaI. Loss of the URA3 geneblaster was selected by plating cells on 5-FOA media. All genomic mutations were confirmed by either restriction enzyme digestion and/or by sequence analysis of PCR products amplified from yeast genomic DNA.

URA3 forward mutation analysis
To monitor spontaneous forward mutations of URA3 integrated near ARS306, the various yeast strains were transformed to URA3 + with pBJ2176 digested with XhoI/SalI, which targets the integration of the URA3 gene in the antisense orientation (OR2) 1100 bp to the left of ARS306, between the FUS1 and HBN1 genes, in chromosome 3. We previously showed that integration of URA3 at this genomic position in the yeast genome does not alter the firing of ARS306 (3).
URA3 to ura3 mutation rates and spectra Spontaneous forward mutation rates of URA3 OR2 were determined for each yeast strain using the method of the median (37). For each strain, 9 to 15 independent cultures, each starting from 100 URA3+ cells were grown in 3 ml of YPD medium for 3 days. Cells were sonicated, harvested by centrifugation, and then washed and resuspended in sterile water. To determine the median number of mutations arising in the cultures, appropriate cell numbers were plated on SC complete media containing 5-FOA. To determine cell culture viability, appropriate dilutions were plated on SC complete media (Sunrise Science Products). Experiments were repeated 3 to 4 times. For sequence analyses, additional independent cultures were grown as described above, washed, and plated on media containing 5-FOA. A single FOA r colony arising from each culture was patched onto YPD and genomic DNA was extracted. The ura3 gene was amplified via PCR and the products were sequenced using oligos LP2221 and LP2222 (3).

Data availability
All of the study data are included in the article.
Funding and additional information-This study was supported by National Institutes of Health (NIH) grant R01-GM129689 (to S. P.). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.