Role of DNA Replication Proteins in Double-Strand Break-Induced Recombination in Saccharomyces cerevisiae

Abstract

Mitotic double-strand break (DSB)-induced gene conversion involves new DNA synthesis. We have analyzed the requirement of several essential replication components, the Mcm proteins, Cdc45p, and DNA ligase I, in the DNA synthesis of Saccharomyces cerevisiae MAT switching. In an mcm7-td (temperature-inducible degron) mutant, MAT switching occurred normally when Mcm7p was degraded below the level of detection, suggesting the lack of the Mcm2-7 proteins during gene conversion. A cdc45-td mutant was also able to complete recombination. Surprisingly, even after eliminating both of the identified DNA ligases in yeast, a cdc9-1 dnl4Δ strain was able to complete DSB repair. Previous studies of asynchronous cultures carrying temperature-sensitive alleles of PCNA, DNA polymerase α (Polα), or primase showed that these mutations inhibited MAT switching (A. M. Holmes and J. E. Haber, Cell 96:415-424, 1999). We have reevaluated the roles of these proteins in G₂-arrested cells. Whereas PCNA was still essential for MAT switching, neither Polα nor primase was required. These results suggest that arresting cells in S phase using ts alleles of Polα-primase, prior to inducing the DSB, sequesters some other component that is required for repair. We conclude that DNA synthesis during gene conversion is different from S-phase replication, involving only leading-strand polymerization.

Repair of double-strand breaks (DSBs) by homologous recombination involves new DNA synthesis. It has been of great interest to elucidate how repair DNA synthesis is related to normal S-phase replication. A clear indication of the interrelatedness of recombination and replication came from studies of late replication in bacteriophage T4 and restarting of stalled or collapsed replication forks in Escherichia coli, which are seen to be recombination-dependent, origin-independent processes (14, 25). Analogous events, termed break-induced replication (BIR), occur in eukaryotes and may also play a key role in reinitiating stalled replication forks and in telomere maintenance (16, 29, 40). But in eukaryotes, DSBs are most often repaired by gene conversion, involving only a short “patch” of new DNA synthesis (36).

During S-phase replication of yeast and other higher eukaryotes, the six minichromosome maintenance (MCM) proteins (Mcm2p to Mcm7p) are believed to comprise the helicase that unwinds the template DNA strands (26). They are assembled at the end of mitosis into prereplicative complexes (pre-RCs) and play a direct role in the activation of DNA synthesis at replication origins (35). Another component of the pre-RC is Cdc45p, a cold-sensitive allele of which, cdc45-1, is defective in initiating DNA replication (1, 60). By constructing conditional degron mutations of the Mcm proteins and Cdc45p, Labib et al. (27) and Tercero et al. (49) have demonstrated that continuous Mcm and Cdc45 activities are required for replication fork progression as well. Chromatin immunoprecipitation analysis also demonstrated that after origin firing, the Mcm complex and Cdc45p dissociate from origin DNA and translocate into adjacent DNA with similar kinetics as the eukaryotic replication fork (2). DNA synthesis is then carried out by three essential DNA polymerases, Polα, Polɛ, and Polδ (55). The Polα-primase complex, encoded by POL1, POL12, PRI1, and PRI2, initiates DNA polymerization on the leading and lagging strands in a coordinated fashion (5). Then, by a process of polymerase switching, Polɛ (POL2) and Polδ (POL3) finish elongation on opposite strands of the replication fork (44). The loading of these polymerases depends on their interaction with PCNA, which is also involved in many other aspects of DNA metabolism, such as nucleotide excision repair, postreplication mismatch repair, and base excision repair (24). The Okazaki fragments of the lagging strand are further processed by the FEN-1 protein and Dna2, which remove the last 5′ ribonucleotides and likely the Polα-synthesized DNA (4, 9, 55). These processed Okazaki fragments are then able to be joined by DNA ligase I (in budding yeast, Cdc9), which is the final and essential step of DNA replication (23, 54, 55). In addition to Cdc9, yeast has one other ligase, Dnl4. Mammalian cells contain four DNA ligases; homologues of only ligase I and IV have been identified in yeast (50). The mammalian DNA ligase IV is essential for V(D)J recombination and nonhomologous end joining (NHEJ) in human lymphocytes (6, 12, 13, 15). Yeast lacking the homologous gene DNL4 are viable and deficient in NHEJ (43, 48, 57).

To understand the roles of these essential replication factors in repair DNA synthesis, we have been studying the developmental switch of the mating type locus (MAT) of the budding yeast Saccharomyces cerevisiae (17, 18). HO endonuclease creates a programmed site-specific DSB at MAT, the repair of which, by gene conversion, leads to a replacement of mating type sequences. One of the two homologous silent donor loci on the same chromosome (HMLα or HMRa) is used as the template for new DNA synthesis. A minimum of about 700 bp of new DNA must be copied from the donor to insert new DNA sequences at MAT. MAT switching has been a useful model of DSB-induced gene conversion in general; for example, repair of HO-induced DSBs created at different DNA sequences appears to be quite similar to spontaneous gene conversion events, both in mitosis and meiosis (3, 32, 36).

By physical monitoring of DNA undergoing MAT switching after the synchronous induction of a DSB in mitotic cells containing conditional-lethal mutations, we found that PCNA is required to synthesize even as little as 30 nucleotides following strand invasion (19). Temperature-sensitive (ts) mutations in either Polɛ or Polδ are also defective in MAT switching, although there is some functional redundancy between these two polymerases in gene conversion (19). Mutants of the lagging-strand replication, pol1-17 (Polα), pri2-1 (primase), and rad27Δ (FEN-1), also greatly impaired the completion of DSB repair. But a ts mutation in the origin recognition complex (ORC), orc5-1, did not affect the efficiency of MAT switching (19). Therefore, these results suggested that gene conversion and BIR may be mechanistically linked, starting with the establishment of a modified origin-independent replication fork after strand invasion, involving both leading- and lagging-strand synthesis from the donor template (19).

However, these conclusions were mainly drawn from experiments carried out on asynchronous mutant cells. It is possible that the mutant phenotype simply arose from arrested DNA replication, sequestering the recombination machinery from reaching the DSBs. Therefore, we have reassessed the roles of these replication components in nocodazole-arrested G₂ cells when DNA replication is completed and the cells are preparing for mitosis. The same approach has also been utilized to investigate the Mcm2-Mcm7 complex, Cdc45p, and DNA ligases in MAT switching. We report that even though the Mcm2-Mcm7 complex, Cdc45p, and DNA ligase I are absolutely necessary for genomic replication, they are not required for DSB-induced gene conversion during MAT switching. Surprisingly, even in a strain lacking Dnl4 and at the restrictive temperature for cdc9-1, MAT switching can still be completed, suggesting that another backup ligase activity may exist in yeast. When a break is formed in the G₂ stage of the cell cycle, pol1-17 and pri2-1 cells are able to complete MAT switching nearly as efficiently as the wild-type cells, suggesting that DSB-induced gene conversion may only involve leading-strand synthesis.

MATERIALS AND METHODS

Strains.

The wild-type strain AMH11 has the genotype hoΔ HMLα MATα HMRa ade1-100 leu2-3 leu2-112 his4-519 ura3-52 ade3::GAL::HO (19). The cdc54-1 cold-sensitive allele was introduced into AMH11 by integration and excision of a YIP5 (URA3-containing) plasmid, p306-cdc54-1, kindly provided by S. Bell (1), to generate the strain yXW7. The mcm7-td strain has the genotype hoΔ HMLα MATa HMRα-BamHI ura3 ade1 ade3::GAL::HO leu2 trp1::hisG ura3-52 mcm7-td (CUP1p-Ub-DHFRts-HA-MCM7::TRP1). It was transformed with the pADH1/UBR1 plasmid provided by A. Varshavsky (31) to obtain the mcm7-td pADH1/UBR1 strain yXW8. The cdc45-td strain yJT125 has the genotype hoΔ HMLα MATa HMRα-BamHI ura3 ade1 ade3::GAL::HO leu2 trp1::hisG ura3-52 cdc45-td (CUP1p-Ub-DHFRts-HA-CDC45::TRP1) (49).

AMH1 (MATα pol1-17), AMH2 (MATα pri2-1), and AMH4 (MATα pol30-52) are isogenic derivatives of AMH11 (19). yXW5 is the MATα derivative of AMH3 (MATa pol3-14), and yXW6 is the MATα derivative of AMH12 (MATa pol2-18), obtained from a mating type switch of these strains through galactose induction of the HO endonuclease at the permissive temperature.

The cdc9-1 strain R177 has the genotype hoΔ HMLα MATa HMRa leu2-3 leu2-112 his4-519 ade1-100 ura3-52 bar1::hisG cdc9-1. A galactose-inducible GAL::HO gene was integrated at ADE3 by using the YIPade3HO plasmid constructed by Sandell and Zakian (42) to obtain the isogenic strain yXW9. The cdc9-1 dnl4Δ strain yXW10 was constructed from yXW9 by the one-step gene replacement method (41), using a 5.7-kb HindIII-SacI dnl4::LEU2 fragment from pJJ252, kindly provided by A. Tomkinson. After transformation, the resulting Leu⁺ colonies were confirmed by Southern analysis.

Induction of MAT switching.

MAT switching was performed as described before (19, 56). The temperature-sensitive strains were pregrown overnight at the permissive temperature in YP-lactate medium. The culture was divided the next morning when it reached 10⁷ cells/ml. One half was maintained at the permissive temperatures, while the other half was shifted to the nonpermissive temperatures. Cells were incubated another 3 h to allow for inactivation of the proteins of interest at the restrictive temperatures; then a 2.0% final concentration of galactose was added for 1 h to induce HO endonuclease, followed by addition of glucose to a final concentration of 2.0% to repress further cutting by HO. For the mcm7-td pADH/UBR1 and the cdc45-td strains, 0.1 mM CuSO₄ was added to the YP-lactate medium at the permissive temperature, and CuSO₄ was subsequently omitted from the medium when cells were shifted to 37°C to induce degradation of the degron-fusion protein (27). For reevaluating other replication components in MAT switching, the cultures were provided with 15 μg of nocodazole/ml when they reached 6 × 10⁶ cells/ml. They were grown for another generation at the permissive temperatures until more than 85% of the cells were arrested at the G₂/M stage and then split. After incubation for 3 h to allow for inactivation of the proteins of interest at the restrictive temperatures, galactose was added to a final concentration of 2.0% for HO induction as described above.

DNA analysis.

Purified genomic DNA was digested with StyI, separated on a 1.4% native gel, and probed with a ³²P-labeled MAT distal fragment (from pJH364) (19, 56). For experiments testing the effect of Mcm7p and Cdc45p in MAT switching (strains yXW8 and yJT125), DNA was digested with StyI and BamHI to distinguish Yα from Yα-BamHI (58). The Southern blots were scanned by PhosphorImager, and the repair efficiency was calculated as described before (19).

Alkaline gel analysis was previously described (56). Genomic DNA was digested with HindIII, separated by gel electrophoresis on a 0.7% NaOH-agarose gel containing 50 mN NaOH, and probed with a ³²P-labeled Yα fragment (from pJH315). The blots were scanned by PhosphorImager, and the ratio of the switched product fragment relative to the corresponding HMLα DNA in the same lane was determined and plotted against time.

Fluorescence-activated cell sorter analysis.

Flow cytometry analysis was done with a Becton Dickinson fluorescence-activated cell sorter analyzer as previously described (53). The DNA content reflects an average of about 15,000 cells.

Immunoblotting.

The Mcm7-td fusion protein contains the influenza virus hemagglutinin (HA) epitope and so can be detected in immunoblots with the monoclonal antibody 12CA5 (Roche) (27). Immunoblot analysis was performed as previously described (49).

RESULTS

Role of Mcm proteins and Cdc45p in gene conversion.

The Mcm2-Mcm7 heterohexamer functions as a complex (52). To test whether Mcm2-7 proteins are required for gene conversion, we first looked at a cold-sensitive allele of MCM4, cdc54-1, in MAT switching (1). Mutant cells grown at the permissive temperature of 30°C were shifted to the nonpermissive temperature of 14°C, following which cells were grown for another generation in order to disrupt the protein activity. Galactose was added to a final concentration of 2% for 1 h to induce a DSB at MAT, followed by the addition of glucose (final concentration, 2%) to repress further HO expression. DNA collected at regular intervals was analyzed on Southern blots to monitor the kinetics and efficiency of recombination. The amount of recombined product at each time point was calculated relative to the amount of HO-cut fragment produced after 1 h of galactose induction, with each sample normalized to the MAT distal DNA in the corresponding lane (see Materials and Methods). At 30°C, it took about 2 h for the product to appear in the wild-type cells, and the process was nearly 100% efficient. The wild-type strain recombined as efficiently at 14°C, but with a 2- to 6-h delay, which implies the existence of some slow steps in gene conversion at low temperature (Fig. 1). However, cdc54-1 was as proficient for repair as the wild-type strain. Normal levels of recombination were found at the restrictive temperature (Fig. 1).

FIG. 1.

Mcm4 protein is not required for DSB repair at MAT. cdc54-1 mutant cells grown at the permissive temperature of 30°C were shifted to the nonpermissive temperature of 14°C, at which the cells were grown another generation to deplete the protein activity (see Materials and Methods). A 2% final concentration of galactose was added for 1 h to induce a DSB at MAT. DNA extracted at intervals after HO cutting was digested with StyI and separated by gel electrophoresis on a 1.4% native gel. Southern blots were probed with a ³²P-labeled MAT distal fragment. Cells switching from MATα to MATa will obtain the StyI restriction site within Ya, yielding a smaller Ya StyI fragment. The 1-h time point represents 1 h of galactose induction of the HO endonuclease. Arrowheads indicate the switched product. Percent switching was calculated from the ratio of the amount of HO cleavage at 1 h compared to the amount of final product, normalized to the MAT distal DNA in each lane (see Materials and Methods).

We were concerned that conventional temperature- or cold-sensitive alleles might retain residual activity at the restrictive temperature. In order to obtain better control of protein inactivation, we used strains in which the only copy of the MCM7 gene was fused to a heat-inducible degron at its N terminus (27) (also see Materials and Methods). At 37°C, the degron cassette is recognized and polyubiquitinated by the Ubr1 protein, resulting in the processive proteolysis of the fused protein through the N-end rule (10). However, it has been shown that degradation of such fusion proteins at higher temperature is sometimes inefficient and incomplete (27). To improve the method, we took advantage of the fact that overexpressing UBR1, the gene encoding ubiquitin ligase, leads to efficient protein inactivation without inhibiting cell growth (27, 31). In cells carrying a pADH/UBR1 overexpression plasmid, Mcm7-td protein was efficiently degraded to below the level of detection within one generation of cell growth (Fig. 2A). Consequently, cells accumulated 1C DNA content due to their inability to enter S phase and replicate their genome (Fig. 2B) (27). However, these mutant cells were still able to complete MAT switching at 37°C (Fig. 2C). Products were formed with the same kinetics as at the permissive temperature and were nearly 100% efficient (Fig. 2D), which strongly suggests that the Mcm2-7 protein complex is not required for DSB-induced MAT gene conversion.

FIG. 2.

Mcm7p and Cdc45p are not required for DSB repair at MAT. (A) An asynchronous culture of the mcm7-td pADH/UBR1 strain was grown at 23°C and then shifted to 37°C. Samples were taken at the indicated hour after the temperature shift to prepare protein extracts. (B) The DNA content of the mcm7-td pADH/UBR1 strain after shifting to 37°C for 3 h was compared by flow cytometry to that of the same culture maintained at 23°C. (C) Southern analysis of the requirement for Mcm7p in MAT switching. Methods are described in the legend for Fig. 1 and in Materials and Methods. The mcm7-td pADH/UBR1 strain contains the donor HMLα and HMRα-B, in which there is a single base pair mutation in Yα that creates a BamHI site about 100 bp from the HO cleavage site (58). Genomic DNA was digested with StyI and BamHI to distinguish Yα from Yα-BamHI (Yα-B). Arrowheads indicate the switched product. (D) Percent switching was calculated as described in the legend for Fig. 1 and in Materials and Methods. Both the Yα and Yα-BamHI products were included. (E) Southern analysis of the requirement for Cdc45p in MAT switching. Methods are described in the legend for panel C. Thecdc45-td strain also contains the donor HMLα and HMRα-B, so genomic DNA was digested with StyI and BamHI to distinguish Yα from Yα-BamHI. Arrowheads indicate the switched product. Percent switching includes both the Yα and Yα-BamHI products. In these experiments, due to the residual expression of the HO endonuclease before galactose induction, a mixed population of MATa and MATα cells existed before the initiation of the time course experiment.

We further tested MAT switching in cells carrying a heat-inducible degron of the essential Cdc45p. It has been shown that Cdc45p is rapidly degraded after the temperature shift and that both the initiation and elongation of DNA replication are prevented (49). However, even after 3 h of incubation at 37°C, MAT switching still occurred normally in the cdc45-td strain, with similar amounts of switched product formed at both 23 and 37°C (Fig. 2E). It is possible, although we think unlikely, that degron-tagged Mcm7 or Cdc45p bound to origin sequences such as those near HML could be particularly resistant to degradation at the restrictive temperature, but we note that we obtained the same results with a cold-sensitive mcm4 mutant.

DSB-induced MAT gene conversion does not require components of lagging-strand replication.

Previously, we reported that mutations of lagging-strand replication, pol1-17 (Polα) and pri2-1 (primase), greatly inhibited the completion of DSB repair (19). To rule out an indirect effect caused by blocking replication in S phase, we tested MAT switching in nocodazole-arrested G₂ cells carrying the pri2-1 mutation. Nocodazole synchronizes cells at prometaphase by inhibiting the assembly of mitotic spindles. Mutant cells grown at the permissive temperature of 23°C were first treated with nocodazole until more than 85% of the cells were arrested at the G₂ stage of the cell cycle, shifted to 37°C to inactivate primase, and then HO induced while maintaining G₂ arrest. MAT switching occurred efficiently in nocodazole-arrested wild-type cells (Fig. 3A). As reported before (19), asynchronous pri2-1 cells were very defective in MAT switching at the nonpermissive temperature, with only 10% product visible at the 6-h time point compared to the wild-type cells (Fig. 3A). However, when the mutant cells were blocked in G₂, they were able to perform MAT switching efficiently, with 76% ± 2% product formed at 37°C (Fig. 3A). Since DNA primase is essential for initiation of Okazaki fragment polymerization (11, 39), this result suggests that DSB-induced MAT gene conversion does not involve lagging-strand replication. To support this idea, we reexamined the role of Polα in MAT switching. Nocodazole-arrested pol1-17 cells were able to recombine with 83% ± 7% efficiency at the restrictive temperature (Fig. 3B). Thus, DSB-induced MAT gene conversion does not require the Polα-primase complex. In these mutants, the observation that there was not 100% repair at 37°C could be attributed to the fact that the nocodazole arrest was not complete, leaving some mutant cells still blocked in S phase.

FIG. 3.

MAT switching does not require primase and Polα. (A) An asynchronous culture of the pri2-1 strain is not able to complete MAT switching at 37°C but repairs the break very efficiently when arrested in G₂. G₂ arrest was accomplished by applying to cultures 15 μg of nocodazole/ml for one generation of cell growth at the permissive temperature of 23°C. MAT switching was carried out as described in the legend for Fig. 1 and in Materials and Methods. (B) pol1-17 cells are almost normal for MAT switching in G₂.

Nevertheless, different from pri2-1 and pol1-17 mutants, cells carrying a cold-sensitive allele of PCNA, pol30-52, were incapable of accumulating switched product under nocodazole arrest. In these experiments, due to a very low level of expression of the HO endonuclease before galactose induction, a small portion of the cells had switched to MATa at the permissive temperature before the initiation of the time course experiment. Both MATα and MATa were cleaved when HO was induced, but no accumulation of switched product was observed even 20 h after HO induction at the restrictive temperature of 14°C, whereas switching at the permissive temperature of 30°C was nearly 100% efficient (Fig. 4A). This implies that the clamp protein is still necessary to recruit or stabilize the proper DNA polymerases at the site of damage. We then examined the roles of the two PCNA-associated DNA polymerases, Polɛ and Polδ, in DSB repair in G₂. MAT switching in a G₂-arrested ts mutant of Polɛ (pol2-18) and in a G₂-arrested ts mutant of Polδ (pol3-14) occurred with 71 and 68% efficiency, respectively, at 37°C without major delays (Fig. 4B), suggesting some functional redundancy between these two polymerases in gene conversion. We note that we could not analyze a pol3-14 pol2-18 double mutant, as this strain is inviable.

FIG. 4.

Southern analysis of PCNA, Polδ, and Polɛ mutants in MAT switching in G₂. Methods are described in the legend for Fig. 3. (A) Cells carrying a cold-sensitive allele of PCNA, pol30-52, are still DSB repair defective in nocodazole arrest at 14°C. (B) ts alleles of either Polɛ (pol2-18) or Polδ (pol3-14) are able to finish MAT switching efficiently without major delays in G₂ at 37°C.

Roles of DNA ligase I and IV in homologous recombination.

DNA ligation is the last and essential step in DNA replication, repair, and recombination. According to the DSB repair model of Szostak et al. (46), or simple synthesis-dependent strand-annealing (SDSA) repair mechanisms (reviewed in reference 36) (see Fig. 6), there should be two necessary ligations. We first analyzed the participation of the essential replication ligase I (CDC9) in this process by introducing the cdc9-1 allele. It has been shown that ts cdc9-1 mutant cells rapidly and irreversibly eliminate DNA ligase I activity at 37°C (23). However, the cdc9-1 strain was proficient for repair even 3 h after shifting to the nonpermissive temperature (Fig. 5A).

FIG. 6.

A molecular model of MAT switching, based on the SDSA model of DSB repair (36). A DSB is induced at the Y-Z junction by HO (A). 5′-to-3′ exonuclease activity creates a 3′ single-stranded tail (B) that invades the homologous silent donor sequence HMLα and initiates new DNA synthesis, using the 3′ end as a primer (C). Branch migration or some helicase activity unwinds the newly synthesized strand from its template and brings it back to the DSB (D), and when the nonhomologous Ya tail has been clipped off by Rad1-Rad10 and maybe Msh2-Msh3 proteins, the second end of the broken DNA can copy the displaced newly synthesized strand, again using the 3′ end as a primer (E). Replication is terminated by capture of the second end of the DSB, while filling-in synthesis and ligation are required as the final step (F).

FIG. 5.

DNA ligases I and IV are not required for DSB repair at MAT. (A) Methods are described in the legend for Fig. 1. Both the cdc9-1 and the cdc9-1 dnl4Δ strains were able to form wild-type levels of recombined products at 37°C as assayed on a native gel. The strains used in these experiments carry HMLα and HMRa. The wild-type strain shows switching from MATα predominantly to MATa, whereas the cdc9-1 and cdc9-1 dnl4Δ strains show switching to MATα, but some recombination with both donors is seen. (B) DNA samples collected from the same time course of the cdc9-1 dnl4Δ strain as shown in panel A were digested with HindIII, separated by gel electrophoresis on a 0.7% alkaline gel, and probed with a ³²P-labeled Yα fragment (see Materials and Methods). Arrowheads indicate the 4.4-kb switched product. The same probe also hybridizes with a 6.1-kb HMLα fragment. The ratio of the switched product fragment, relative to the corresponding HMLα DNA in the same lane, was determined and plotted against time.

The occurrence of gene conversion in the absence of the Cdc9 ligase may be explained by the presence of a second ligase. In budding yeast, the other identified ligase is Dnl4, which is not essential but is necessary for NHEJ (43, 48, 57). Dnl4 cannot complement the growth defects of Cdc9 but might be able to repair nicks made during DSB repair at MAT. Therefore, a cdc9-1 dnl4Δ double mutant was constructed and tested for its ability to complete MAT switching. Without the function of Dnl4, the cells were severely impaired in DNA NHEJ, as measured by the ability of the cells to ligate a linearized, restriction endonuclease-cleaved plasmid transformed into the cells (data not shown).

Surprisingly, the cdc9-1 dnl4Δ strain was still able to carry out recombination at 23 and 37°C, with wild-type amounts of recombined product formed at both temperatures (Fig. 5A). However, the apparent completion of recombination could reflect the fact that the two unligated strands of DNA were still held together by base pairing, thus allowing product to be seen on a native gel. If this were the case, then examining separated strands on a denaturing gel should show that each strand was interrupted by unligated nicks. Thus, DNA samples from the same time course were digested with HindIII, which generates a 4.4-kb fragment around the HO cut site, and separated on a 0.7% alkaline gel to see if each single strand was intact. The gel was probed with a Yα fragment that detects the formation of the switched product. Nevertheless, the same amount of intact product was detected in the cdc9-1 dnl4Δ double mutant at 37°C as at the permissive temperature (Fig. 5B). These results suggest that eliminating the activities of both known DNA ligases in yeast does not prevent MAT switching. We note that it is possible that residual ligase activity may exist in the double mutant or that the nicks on each strand were used as primers for “nick translation,” effectively moving the nick to a more distant location. If this is the case, then the nicks must have been displaced by at least 2 kb.

DISCUSSION

Physical monitoring of DNA undergoing MAT switching provides a powerful tool to dissect the genetic requirement of DSB repair in vivo. Using this biochemical approach, we have analyzed the relationship between DSB-induced MAT gene conversion and DNA replication. We conclude that both processes depend on the presence of PCNA, but repair DNA synthesis is distinct from normal replication in several aspects.

First, DSB-induced gene conversion does not require the lagging-strand components. Cells carrying temperature-sensitive mutations in the Polα-primase complex are able to carry out MAT switching efficiently in the G₂ stage of the cell cycle (Fig. 3). The function of Cdc45p, which is responsible for loading the Polα-primase complex onto chromatin in DNA replication (1), is also dispensable for gene conversion (Fig. 2E). We believe that after strand invasion occurs, new DNA polymerization is initiated through leading-strand synthesis, using the 3′ end that was generated by resection. The Rad51-mediated strand invasion event by this 3′ end provides the primer end and, thus, may eliminate the need for DNA primase activity. The replication “bubble” could remain small, or active branch migration might also occur, as suggested by the SDSA models, with the newly synthesized strand being displaced from the template and returned to the broken molecule. After the removal of the nonhomologous tail by Rad1-Rad10 and probably Msh2-Msh3 proteins (21, 45), the second end of the DSB can also copy the returned strand, again using leading-strand synthesis (Fig. 6). However, the recruitment of the proper leading-strand polymerases onto the 3′ ends still requires PCNA, as the pol30-52 mutation completely impaired MAT switching at 14°C (Fig. 4A).

The conclusion that Polα and primase are not required to complete MAT switching is quite different from what we had concluded earlier (19) by studying repair in cells carrying pol1-17 or pri2-1 after asynchronous cells had been arrested for several hours at their restrictive temperature before inducing HO endonuclease. In these experiments, all cells were arrested in S phase, with stalled replication forks. It is likely that under these conditions some other component of DNA replication becomes sequestered, making it unavailable to participate in repairing the DSB induced after arrest. This component of DSB repair could be PCNA or both Polδ and Polɛ, or another DNA replication factor.

In our group's original study, where we concluded that Polα and primase were important for DSB repair (19), we recognized the possibility that stalling replication might sequester other components; consequently, we attempted to rule out this possibility by examining repair in G₁ cells, arrested by α-factor. Our current understanding of DSB repair is that there is very little repair by HO endonuclease-induced gene conversion in G₁ cells that remain arrested by α-factor, at a point called “start,” which is prior to the activation of the Cdk1, Cdc28 (51; G. Ira, X. Wang, A. Pellicioli, L. Wan, N. M. Hollingsworth, M. Foiani, and J. E. Haber, submitted for publication). The strains used previously (19) did not contain a bar1 mutation that renders cells hypersensitive to α-factor, and it is likely that the PCR product that was seen in wild-type G₁-arrested cells may have come from a minority of cells that exited from G₁ arrest. The amount of product may have been overestimated and was not confirmed by Southern blot analysis. The inhibition of recombination in α-factor-arrested cells does not affect cells that are arrested by inactivating Mcm4, Mcm7, or Cdc45, as these cells have all progressed beyond the start point of the cell cycle and have active Cdc28 and are thus not impaired in DSB repair. For all these reasons, we believe that the best way to confirm that conditional mutations in DNA replication also play a role in DSB repair is to carry out the studies in nocodazole-blocked G₂ cells.

A second conclusion from our work is that, whereas Polδ and Polɛ are both absolutely essential for replication elongation (5), it seems that they can substitute for each other to carry out DNA synthesis needed for gene conversion (Fig. 4B). It remains possible that the residual enzymatic activity in these mutants is sufficient to copy the donor template during DSB repair at the restrictive temperature but not enough to carry out DNA replication.

Third, although the requirements of the Mcm complex and Cdc45p for replication elongation are clear (27, 49), neither of them is necessary for DSB-induced MAT gene conversion, where a minimum of 700 bp of new DNA synthesis must occur (Fig. 1 and 2). One related observation is that during double-strand gap repair, the size of the DNA segment to be copied limits the efficiency of repair (37). As the length of the template for gap repair increases from a few base pairs to 9 kb, the efficiency of repair decreases by fourfold, which implies that DNA repair synthesis is not intrinsically as efficient or processive as normal replication. Based on our results, one explanation for this reduced efficiency is that repair synthesis lacks the processive Mcm helicase and Cdc45p, so that DNA polymerization tends to dissociate from the template, as it has to traverse longer distances.

During pre-RC formation in early G₁, the recruitment of Mcm2-Mcm7 to chromatin depends on Cdc6 and Cdt1, through their binding to the ORC (35). MAT switching does not require the ORC (19). Yeast Cdc6 protein accumulates at the end of mitosis and disappears after the initiation of DNA replication (38). Controlled by Cdc28 kinase activity, S. cerevisiae Cdt1 accumulates in the nucleus during G₁ and is excluded from the nucleus for the rest of the cell cycle to prevent second-round replication (35, 47). These results suggest that without the involvement of the ORC and the unavailability of Cdc6 and Cdt1, the MCM members cannot be recruited to broken DNA ends to form an active hexameric helicase in DSB repair. We also note that MAT switching can occur in G₂ cells (Fig. 3A), where Mcm proteins are largely excluded from the nucleus (34). Although we found no effect of inactivating Mcm4, Mcm7, or Cdc45, it is of course possible that a small amount of residual activity of each of these temperature-sensitive mutations remains at the restrictive temperatures, not enough to allow replication but enough to allow a single DSB repair event.

It remains possible, however, that Mcm proteins and Cdc45p will be important in the related homologous recombination process of BIR. BIR is apparently a processive and efficient process. In a diploid strain, when a broken chromosome end shares homology with the unbroken chromosome only on one side of the DSB, more than half of the cells repaired the DSB by copying 20 kb of sequence from the template chromosome until it reached the end (8). Morrow et al. (33) showed that the duplication events can extend up to 230 kb. But BIR appears to have much slower kinetics compared to gene conversion, with products starting to appear at 4 h after DSB formation (A. Malkova, M. L. Naylor, M. Yamaguchi, G. Ira, and J. E. Haber, submitted for publication). It is possible that the delay reflects the amount of time necessary for the recruitment of the Mcm proteins to the damage and the de novo assembly of a replication fork with both leading and lagging-strand synthesis. It will be interesting to know whether the Mcm proteins, Cdc45p, and the lagging-strand components are necessary for BIR and, if so, how they are recruited. In bacteria, recombination-dependent, origin-independent DNA replication depends on the PriA protein to organize the replication fork (28, 59). No homologue of PriA has been identified in eukaryotes, but it is possible that there is an analogous protein that is required for this role.

Last, there is a surprising apparent lack of requirement for the identified DNA ligases in DSB repair. During DNA repair, when resynthesis is involved, a nick will be left at the 3′ end of the new DNA strand which needs to be sealed by a DNA ligase (Fig. 6E). But MAT switching was unaffected in a ts cdc9-1 mutant. In yeast, Cdc9 and Dnl4 are targeted to different pathways in DNA replication and repair and are only poorly able to substitute for each other. However, it has been suggested that, if the replicative load is low, other ligases can compensate for ligase I in replication (7). But surprisingly, with deletion of Dnl4 and at the nonpermissive temperature for cdc9-1, DSB repair could still be carried out (Fig. 5A), whereas NHEJ was greatly impaired (data not shown). We found full-length DNA strands even when single strands of DNA from a 4.4-kb restriction fragment were examined on denaturing gels (Fig. 5B). This suggests that these two ligases are not required for sealing nicks during MAT switching and that there is another ligase activity involved in gene conversion. Again, we recognize that there could be residual activity of cdc9-1 which is sufficient to repair the two nicks of the repaired DNA but insufficient to allow replication. Alternatively, the nicks could have been nick translated out of the 4.4-kb region that we examined, a displacement of at least 1 kb from the original site in each case.

Mammalian cells carry two other ligases, DNA ligase II and III (50). DNA ligase II appears to be a degradation product of ligase III (20); DNA ligase III has several isoforms, which are believed to play a role in homologous recombination during meiotic prophase in testes (30). But no homologues of these ligases have been found in yeast. Another ligase function has been reported in mammalian cells, DNA ligase V, which has DSB sealing activity (22). No gene encoding ligase V has been identified. But it is possible that some functional homologue of these mammalian ligases exists in yeast and is responsible for ligating DNA strands in DSB repair, which will further distinguish repair DNA synthesis from S-phase replication.