ABSTRACT
DNA polymerases play a central role during homologous recombination (HR), but the identity of the enzyme(s) implicated remains elusive. The pol3-ct allele of the gene encoding the catalytic subunit of DNA polymerase δ (Polδ) has highlighted a role for this polymerase in meiotic HR. We now address the ubiquitous role of Polδ during HR in somatic cells. We find that pol3-ct affects gene conversion tract length during mitotic recombination whether the event is initiated by single-strand gaps following UV irradiation or by site-specific double-strand breaks. We show that the pol3-ct effects on gene conversion are completely independent of mismatch repair, indicating that shorter gene conversion tracts in pol3-ct correspond to shorter extensions of primed DNA synthesis. Interestingly, we find that shorter repair tracts do not favor synthesis-dependent strand annealing at the expense of double-strand-break repair. Finally, we show that the DNA polymerases that have been previously suspected to mediate HR repair synthesis (Polε and Polη) do not affect gene conversion during induced HR, including in the pol3-ct background. Our results argue strongly for the preferential recruitment of Polδ during HR.
Homologous recombination (HR) is a process that allows genetic exchange between DNA sequences sharing homology and leads to gene conversion or crossovers (COs). The ingenuity of this process is underscored by its conservation from bacteria to humans and its implication in a variety of unrelated nuclear processes. HR is implicated in the restart of stalled replication forks (24). It serves in the repair of DNA damage, such as single-strand gaps, double-strand breaks (DSBs) and interstrand cross-links. It is implicated in mating type switching in yeast strains and in the diversification of immunoglobulin-variable genes in vertebrates (6). In meiosis, the primary function of HR is to establish a physical connection between homologous chromosomes to ensure their correct disjunction at the first meiotic division. In addition, meiotic HR contributes to diversity by creating new linkage arrangements between genes or parts of genes (35).
In the past few years, it has become evident that these multiple roles of HR are achieved with variations in the process. Two primary models allow for gene conversion as it is observed in different contexts (Fig. 1). In the seminal DSB repair (DSBR) model of Szostak et al. (41), supported mainly by molecular meiotic studies in yeast, the formation of DSBs is followed by exonucleolytic degradation of the 5′ ends of the broken duplex to expose single-stranded tails with 3′ termini (40) (Fig. 1). DSB formation and 5′-end resection are followed by the invasion of an intact nonsister chromatid by only one of the two single-stranded tails (13). Single-end invasion (SEI) results in hybrid DNA (hDNA), in which the two strands in a duplex are of different parental origin. If the two parental duplexes are genetically different within the region of strand exchange, the resulting hDNA contains mismatched base pairs and is referred to as heteroduplex DNA. Then, DNA synthesis, capture of the second end, providing a new primer for a second round of DNA synthesis, and ligation generate a double Holliday junction (dHJ) intermediate with asymmetric hDNA (i.e., hDNA on only one of the two duplexes) on each side of the DSB and on each chromatid (Fig. 1). If an HJ undergoes branch migration, then hDNA will be formed on both duplexes (symmetric hDNA). Eventually, dHJ intermediates are resolved by cutting, at each junction, either both outside strands or both inside strands. Cutting of the two junctions in opposite directions generates COs, while cutting in the same direction generates noncrossovers (NCOs). Recently, a dissolution mechanism processing dHJs and leading exclusively to NCOs has been proposed (45). In this process, a helicase merges the HJs, creating a hemicatenated intermediate that is resolved by a type IA topoisomerase (Fig. 1).
Models of HR. Diagrammed are DSBR by the DSBR pathway (left) and by SDSA (right). Shown are two recombining DNA duplexes, one indicated in red and the other in blue (see text for details). In the DSBR model, arrowheads depict cleavage of dHJs; only two of four possible resolution products are shown.
The synthesis-dependent strand annealing (SDSA) model differs from the DSBR model at the step following SEI intermediate formation (31). According to the SDSA model (Fig. 1), the invading strand in an SEI intermediate is extended by DNA synthesis and then subsequently displaced. The newly synthesized DNA strand then anneals to the single-stranded tail on the other side of the break. DSBR is completed by DNA synthesis and ligation. If extension of the invading strand creates a single-stranded tail that is longer than the exposed complement on the opposite side of the break, a flap structure with an exposed 3′ end will result after the two strands anneal, and this flap will need to be removed to allow ligation. In this model, asymmetric hDNA occurs only on one side of the break and only on the chromatid that suffered the DSB. The SDSA repair pathway does not lead a priori to dHJ formation and, therefore, produces only NCO products. Evidence supporting SDSA has come from phage (9) and from tripartite HR experiments in yeast and flies in which sequences homologous to each DSB end are located on different chromosomes (27, 38). Moreover, in vegetative yeast cells, during MAT gene switching in Saccharomyces cerevisiae, the recipient locus receives all newly synthesized DNA (14), a specific feature of the SDSA model. Finally, evidence that SDSA is a meiotic pathway comes from the finding that two kinetically and genetically distinct mechanisms of DSBR are at work during meiosis, with one leading specifically to NCOs and one to COs (3). In addition, molecules with newly synthesized DNA on both sides of a meiotic DSB have been detected by the incorporation of thymidine analogs into chromosomes (42).
The study of DNA polymerase(s) implicated in repair synthesis might provide important information on HR pathways. However, analysis of DNA synthesis during HR is complicated by the fact that essential processive DNA polymerases, DNA polymerases δ (Polδ) and ε (Polε) are implicated (8, 21, 44). Moreover, it is possible that they can compensate for each other with redundant roles (44). Finally, for each role that HR plays during different DNA transactions, it is possible that a specific DNA polymerase is recruited. For example, it is striking to see the role that the translesion synthesis polymerase η (Polη) plays in immunoglobulin gene conversion tracts in chicken B lymphocytes (17).
We reported previously that an allele of the POL3 gene of S. cerevisiae coding for the catalytic subunit of Polδ exhibits lower levels of HR. The mutant allele, pol3-ct, does not display obvious defects in DNA replication and repair. However, the pol3-ct allele leads to a significant decrease of meiotic gene conversion, along with a decrease of COs, a defect associated with shorter hDNAs (21). This effect demonstrates a direct participation of Polδ in meiotic HR and suggests that this role is nonredundant. Interestingly, although it has been suggested that the decision to yield a meiotic CO or NCO is determined before DNA synthesis occurs (3-5), the pol3-ct defect in CO formation raises the possibility that there are more events channeled toward the SDSA pathway in the mutant.
In the current study, we took advantage of the pol3-ct mutation to address several central issues concerning the role of DNA polymerases during recombinational repair in vegetative cells. We found that (i) Polδ is the preferred DNA polymerase for recombinational repair regardless of the nature of the initiating lesion, (ii) the pol3-ct defect is independent of mismatch correction in hDNA, (iii) the pol3-ct mutation results in shorter stretches of repair synthesis in somatic cells, and (iv) reduced repair synthesis does not favor SDSA.
MATERIALS AND METHODS
Yeast strains, plasmids, and media.All media were prepared as previously described (36). Strains used in the present study are listed in Table 1. Induced HR frequencies, spontaneous CO frequencies, and the role of Polε and Polη in induced HR were determined in isogenic derivatives of strain W303-1A (43). pol2-16 strains were obtained after mating, sporulation, and dissection from the parental strain ASY102 (W303-1A MATapol2-16 bar1Δ::hisG) kindly provided by A. Sugino (29). The parental rad30Δ C10-10A (W303 MATα rad30Δ::HIS3) is a gift of R. Woodgate (23). The role of mismatch repair (MMR) in the pol3-ct defect was investigated in strain FF18733-derived cells (1). Isogenic derivatives of YPH250 cells (37) were used to study HO-induced and UV-induced gene conversion tract lengths. DY3438 and DY3439 parental strains were kindly provided by J. Nickoloff (28). Replacement of the wild-type (WT) copy of POL3 at the endogenous locus by the pol3-ct allele is achieved by using plasmid PLMδ1 as previously described (21).
Strain list
Irradiation and recombination.Cells in stationary phase (UV) or exponential growth (γ rays) were plated at appropriate dilutions on yeast-peptone-dextrose (YPD) and synthetic plates. We performed UV irradiation using a 264-nm source delivering 1 J/m2/s. The cells were treated for 0, 15, 30, 45, and 60 seconds. The γ irradiation was performed using a 137Cs source delivering 50 Gy/min. Survival was determined as the number of cells forming colonies on YPD medium after a given irradiation divided by the number of colonies from cells not irradiated. Similarly, we determined HR frequencies by dividing the number of recombinant colonies growing on selective medium by the number of unselected colonies subjected to the same dose of irradiation. The values obtained were corrected by subtracting the number of recombinants present on the nonirradiated plates.
DSB-induced gene conversion tract length assay.Two-day-old colonies of DY3515-13 and LMN1/2 diploid strains were inoculated into 2 ml of YP-glycerol (3%) medium for 24 h. Dilutions were plated onto nine plates of YPD medium and nine plates of YP-galactose (YPGal) medium and incubated for 4 days at 30°C. GAL-HO is specifically induced on YPGal plates, leading to DSB cleavage by HO endonuclease at the ura3-HO site (see Fig. 4). DSB-dependent cell killing was calculated as the number of cells forming colonies on YPGal medium divided by the number of cells forming colonies on YPD. Next, YPD and YPGal plates were replica plated on synthetic medium lacking uracil (SC-Ura). Repair of DSBs by HR leads to the formation of Ura+ prototrophs by using the unbroken allele as a donor. No Ura+ prototroph among around 2,000 replica-plated colonies was recovered from the nine YPD (HO noninduced) plates. Inversely, most (95%) colonies on YPGal yield Ura+ prototrophs after replica plating. Since a number of colonies are sectored on SC-Ura, all HO-induced DSBs at the URA3 locus do not occur at the level of the first plated single cell. Since DSBs at URA3 could occur after the first mitotic division from single cells, Ura+ cells issued from the same colony could be the result of different DSB-induced HR events. Ninety-five Ura+ replica-plated colonies for both DY3515-13 and LMN1/2 were streaked out on SC-Ura plates to obtain 95 newly isolated single Ura+ colonies representing 95 independent HR events.
Genomic DNA was extracted to perform PCR and restriction analysis on heterozygous restriction markers (see Fig. 4). Primers p4 and p7 amplify a 1,935-bp fragment containing the EcoRI site 2,656 bp upstream of the HO site. Primers p10 and p12 amplify a 1,093-bp fragment containing the AseI and PstI sites, 412 and 227 bp upstream of the HO site, respectively. Primers p8 and p9 amplify a 1,047-bp fragment containing the NcoI site at the same position as HO and the XbaI site 332 bp downstream of HO. Loss of heterozygosity (LOH) is observed at the marker sites when the PCR fragment is totally cleaved or, inversely, totally refractory to cleavage. The frequencies of the HO-induced recombination classes were compared using the chi-square test (≤0.05 or ≤0.01).
Selection of COs.CO selection and analysis were performed as described previously (34).
RESULTS
pol3-ct leads to a deficit of radiation-induced recombinants.The pol3-ct allele leads to a decrease of meiotic heteroallelic HR (21). To investigate the effect of pol3-ct on vegetative cells, we analyzed radiation-induced HR in diploid isogenic strains. We used two mutant alleles of ARG4: arg4ΔRV, a 2-bp deletion ablating the EcoRV site at position +258 in ARG4, and arg4ΔBg, a 4-bp insertion by fill-in of a BglII site at position +1,274. These mutations are separated by 1,016 bp. In addition, we used the lys2-NdeI-5′ and lys2-NdeI-3′ alleles of LYS2 carrying mutations corresponding to 2-bp insertions after fill-in of NdeI sites. These two sites within the gene are at positions +1,190 and +3,230 and are separated by 2,040 bp. DLM1, a POL3 homozygous strain, and DLM2, a pol3-ct homozygous strain (Table 1), were used to measure the frequencies of induced Arg+ and Lys+ recombinants. As previously observed (21), the presence of the pol3-ct allele makes yeast strains sensitive neither to UV nor to γ-ray irradiation at the doses that trigger repair by HR (Fig. 2), indicating that the efficiency of recombinational repair is not affected by pol3-ct. On the other hand, although induction of intragenic HR is apparent in the pol3-ct/pol3-ct diploid strains, this induction is decreased at least threefold (Table 2). A decrease of recombinants among survivors is observed at both loci tested, irrespective of the irradiation type (Fig. 2). These observations show that pol3-ct indeed affects mitotic HR. Additionally, we found that the pol3-ct allele is recessive, since the heterozygous POL3/pol3-ct DLM12 strain behaves similarly to the WT strain (Fig. 2).
Effects of irradiation on survival and heteroallelic HR. (A) Effects of UV light on survival and HR between arg4 and lys2 heteroalleles. (B) Effects of γ irradiation on survival and HR between arg4 and lys2 heteroalleles. All the curves correspond to the mean values from at least six independent experiments.
Effect of pol3-ct on induced heteroallelic recombinationa
pol3-ct defect is independent of MSH2 MMR pathway.Msh2 is a MutS homologue and participates in the repair of base-base mismatches, one-base insertions, deletions, and small (1 to 14 bases) loops that can be formed in hDNA (39). Induced heteroallelic HR frequency is increased in an msh2Δ strain (2, 7) (Fig. 3), revealing that some events are “erased” by the action of Msh2. Msh2 is also involved in the rejection of hDNA formed between divergent DNA sequences in single-strand annealing assays (11). Thus, the hyporecombinogenic (hyporec) phenotype of pol3-ct could be due to an increased efficiency of MMR that would leave less unrepaired hDNA that forms a prototroph after DNA replication (Fig. 3A). Alternatively, rejection by Msh2 of hDNA containing a single mismatch in Polδ-ct mutants could be favored. Therefore, we expected to observe a milder hyporec phenotype of pol3-ct in an msh2Δ background.
MMR-independent induced HR defect of pol3-ct. (A) Model for recombinant Arg+ prototroph formation between arg4ΔRV (RV) and arg4ΔBg (Bg) heteroalleles. The formation of hDNA that includes the site of one mutation leads to a mismatch. Arg+ recombinants arise either from MMR toward the WT or from mitotic allelic segregation of unrepaired mismatches following DNA replication. (B) Survival and induced HR between arg4ΔRV and arg4ΔBg heteroalleles following UV irradiation, observed in WT, pol3-ct, msh2Δ, and pol3-ct msh2Δ strains. (C) Model for recombinant Arg+ prototroph formation between arg4ΔRV (RV) and arg4ΔAg (Ag) heteroalleles. These heteroalleles, carrying mutations located 69 bp away from each other, when involved in HR lead to hDNA that includes the two mutation sites and therefore has two close mismatches. The formation of Arg+ prototrophs requires independent repair of at least one of the two mismatches toward the WT by the short-patch MMR pathway (SP-MMR). (D) Survival and induced HR between arg4ΔRV and arg4ΔAg heteroalleles following UV irradiation, observed in WT, pol3-ct, msh2Δ, and pol3-ct msh2Δ strains.
We analyzed UV-induced HR between arg4ΔRV and arg4ΔBglII heteroalleles in pol3-ct, msh2Δ, and pol3-ct msh2Δ strains. As shown in Fig. 3B, the msh2Δ diploid strain shows the expected elevated frequency of induced Arg+ recombinants compared to the frequency in the WT. Conversely, the pol3-ct strain shows a decreased frequency of Arg+ recombinants. The pol3-ct msh2Δ double mutant shows a clear decreased frequency of Arg+ recombinants compared to the frequencies in both the WT and the single msh2Δ mutant strain. Comparing the msh2Δ and pol3-ct msh2Δ strains, we found that the decrease is threefold, which corresponds to the same decrease observed between pol3-ct and the WT in a MSH2 background. Therefore, the pol3-ct defect does not depend on the MSH2-driven MMR pathway.
Randomly initiated recombination events cover shorter DNA regions in pol3-ct strains.The observation that, following irradiation, pol3-ct does not reduce DNA repair efficiency and shows an MSH2-independent hyporec phenotype argues in favor of a role for Polδ in hDNA extension. However, the loss of induced prototroph formation could be due either to shorter (as expected in the case of pol3-ct) or, alternatively, to longer hDNA. Indeed, the probability of an hDNA spanning the site of one mutation should decrease in the case of shorter hDNA randomly initiated after irradiation. On the other hand, if longer hDNA is formed, one might expect more coconversion of heteroallelic markers, i.e., more hDNA with two mismatches and a decreased probability of repair toward a WT allele. To rule out the long-tract hypothesis, we used heteroalleles at ARG4 carrying mutations only 69 bp away from each other (7) (Fig. 3C). The probability of them being included in the same hDNA is elevated. In this situation, the MSH2-dependent MMR pathway gives long excision tracts that prevent the formation of Arg+ recombinants regardless of the strand that is excised. In addition, if the two close mismatches are left unrepaired in msh2Δ mutants, no strand will yield an Arg+ allele after DNA replication (Fig. 3C). Thus, in this scenario, the most likely possibility for forming induced Arg+ recombinants is to repair at least one of the mismatches, via a short-patch MMR pathway that is enhanced in the absence of Msh2 (7) (Fig. 3C). We recovered a smaller number of induced recombinants when the mutations were close, a result highlighting the efficiency of long-patch MMR. However, the absence of Msh2 leads to a similar increased frequency of Arg+ in the pol3-ct strain (4.8-fold at 60 J/m2) and in the WT (3.2-fold at 60 J/m2), clearly indicating that the short-patch mismatch repair is functional in both backgrounds (Fig. 3D). Importantly, we still observe a twofold decrease of induced Arg+ frequency in the pol3-ct background, a result that can be best explained by shorter conversion tracts (see Discussion, Fig. 3D, and Table 2).
HO-induced DSBs are repaired with shorter conversion tracts in pol3-ct.Because DNA repair is not impaired and the hyporec phenotype of pol3-ct upon irradiation is not linked to mismatch repair, we used a system designed by Nickoloff and coworkers (28) to determine the molecular effect of pol3-ct on mitotic gene conversion tract length. In this system, both allelic URA3 copies are abrogated, one by the insertion of a 39-bp recognition site for the HO endonuclease and the other by a frameshift mutation generating an XbaI restriction site. Moreover, the HO recognition site is flanked by phenotypically silent mutations, absent on the homologous chromosome and creating restriction site polymorphisms (Fig. 4A). Allelic HR is initiated by a DSB after induction of the HO endonuclease gene on galactose-containing plates. Independent Ura+ recombinants can be easily selected, and conversion of the silent flanking markers can be monitored after PCR amplification and restriction analyses.
LOH of RFLP markers following HO-induced DSBs within URA3 on chromosome V. (A) Map of the URA3 region containing RFLP markers and an HO site. Both parental chromosome Vs are represented. Following the induction of HO endonuclease with galactose, a DSB is formed at the HO site (arrow). Repair by homologous HR using the donor unbroken chromosome (in black) gives Ura+ recombinants, as long as XbaI, a +1 frameshift-inactivating URA3, is not included in the hDNA. The recipient chromosome suffering the DSB at the HO site is represented in white. The silent RFLP markers used in this study are indicated (E, EcoRI; A, AseI; P, PstI; N, NcoI). (B) Observed LOH in Ura+ recombinants following HO-induced DSBs. After HO-induced DSB formation, LOH for each marker in independent Ura+ recombinants can be monitored by PCR amplification and restriction analyses. Gray circles indicate markers still heterozygous after DSBR. Black circles are for markers that became homozygous after DSBR, with the genotype of the donor molecule. White circles are for markers that became homozygous but with the “recipient” genotype. Seven Ura+ genotypes (A to G) are observed. We distinguish three categories of events: continuous events (A to D), discontinuous events (E), and complex events (F and G) in which the initial recipient chromosome becomes the donor. The numbers of Ura+ belonging to each class are given for the WT and for the pol3-ct mutant. The frequency of each class is given in parentheses. (C) The LOH frequency for each marker is given as a function of the distance separating the marker and the site of the HO-induced DSB. For each marker, the LOH frequency was calculated as the sum of the genotypes converted for the marker divided by the total number of events analyzed (for example, for EcoRI in the WT, 36 + 2 + 2 + 5/140 = 32%).
In this system, plating cells on medium containing either glucose or galactose allows us to determine the viability of cells after the induction of a DSB. We found that 87% of WT cells (DY3515-13) and 89% of pol3-ct cells (LMN1/2) survive HO induction, indicating that cells harboring the pol3-ct allele repair HO-induced DSBs efficiently.
After DSB induction, we analyzed LOH at five heterozygous restriction markers in independent Ura+ recombinants for both the WT and pol3-ct strains (Fig. 4). The LOH patterns show that the parental HO site-bearing chromosome is always the recipient of information, confirming that these recombinants are the result of an initial break at HO. In addition, these patterns are indicative of predominantly continuous tracts (no interruption in the continuity of the information transferred; around 95% of all events) (classes A, B, C, and D in Fig. 4B). Interestingly, the frequency of class D, corresponding to continuous LOH tracts reaching the farthest-upstream restriction fragment length polymorphism (RFLP) marker, is 2.5-fold higher in the WT than in the mutant (P < 0.01). Conversely, the A, B, and C classes, corresponding to shorter conversion tracts, are more abundant in the pol3-ct mutant than in the WT (P < 0.001). Therefore, the pol3-ct mutation leads to a loss of long tracts of LOH and an enrichment of short tracts.
The LOH frequency of each RFLP can be analyzed separately (Fig. 4C). We found a decrease of LOH as a function of the distance from the HO site that is comparable in both the WT and the pol3-ct mutant, up to a distance of 412 bp (AseI site). However, when we analyze LOH at the EcoRI site located 2,656 bp from the break, the frequency drops to 32% in the WT and to a significantly lower (P < 0.01) 13% in the pol3-ct strain. Thus, the frequency of DSB-induced LOH is decreased 2.5-fold in pol3-ct for the farthest marker, while there is no decrease at the immediate vicinity of the break. The effect of pol3-ct is therefore dependent on the distance between the markers and the initial DSB, a result reminiscent of the effect of pol3-ct at the HIS4 meiotic hot spot (21).
It should be noted that the PCR-based strategy that we decided to follow does not allow differentiation between LOH events due to gene conversion events or to COs occurring during the G2 phase of the cell cycle (28). Therefore, the pol3-ct defect observed in this system could potentially reflect a defect in CO formation instead of a defect in gene conversion tract length. However, G2 COs in this system represent at most 10% of all events (see Fig. 2 in reference 20). In addition, it is only half of G2 COs that can lead to class D events, depending on the mitotic segregation of the recombinant chromatids. Therefore, we can estimate that the frequency of class D corresponding to long gene conversion tracts rather than G2 COs in the WT is around 20% instead of 25% and, thus, still at least twofold higher than in the pol3-ct mutant. Moreover, as shown below, pol3-ct does not influence the frequency of spontaneous mitotic COs. Thus, we interpret the steeper gradient of DSB-induced LOH found in the pol3-ct strain as the result of shorter gene conversion tracts.
pol3-ct does not lead to a decrease of spontaneous mitotic COs.To determine the possible effect of pol3-ct on mitotic CO formation, we used the spontaneous CO recovery assay designed by Robert et al. (34). The system is based on arg4ΔRV and arg4ΔBglII heteroalleles carrying mutations separated by 1 kb. The arg4ΔBglII heteroallele is located at its endogenous locus on chromosome VIII, and arg4ΔRV is located between a WT and a mutated allele of URA3 on chromosome V, in the same orientation with respect to the centromere. An HR event between these ectopically located arg4 heteroalleles can generate a functional copy of the ARG4 gene by gene conversion of a maximum tract length of 1.5 kb associated or not associated with a CO. A CO leads to a reciprocal translocation that separates the duplicated URA3 and ura3-1 alleles to individual chromosomes. It is therefore possible to directly infer CO events in a secondary screen based on replica plating onto a medium containing 5-fluoroorotic acid (5-FOA) that kills Ura+ yeast cells specifically. Thus, Arg+ colonies resulting from an HR event associated with a CO will not form many papillations when replica plated onto 5-FOA, whereas those resulting from a simple gene conversion event that retains the URA3 direct repeats will. We previously found that, in WT cells, 11.2% of the conversion events yielding Arg+ prototrophs are associated with a CO. The percentage of COs among Arg+ recombinants, determined from three pol3-ct-independent segregants, was found to be 11.3%. Thus, the formation of shorter gene conversion tracts is not associated with a decrease of spontaneous CO in the pol3-ct mutant.
Polε is dispensable for induced recombination in the WT and in the pol3-ct mutant.The pol3-ct phenotype implicates Polδ in the formation of long gene conversion tracts but does not exclude a role for other polymerases in short-event formation. Polε, the other processive polymerase responsible for DNA replication, has been implicated in gene conversion at the MAT locus (44). To determine the role of Polε during induced heteroallelic HR, we took advantage of the pol2-16 allele of POL2. POL2 encodes the catalytic subunit of Polε, and the pol2-16 mutation eliminates both the polymerase and exonuclease activities of Polε (18). Interestingly, this allele neither leads to cell lethality nor confers hypersensitivity to γ-ray irradiation or HO-induced DSBs (18). Therefore, the role of Polε during HR is either marginal or can be fully carried out by another polymerase.
We measured induced HR frequencies at ARG4 (Fig. 5A and B) and LYS2 (data not shown) and found no effect in pol2-16 strains. Thus, unlike what is observed in pol3-ct, HR is as efficient in pol2-16 mutants as in the WT. Additionally, if Polε were recruited more frequently during HR to substitute for a defective Pol3-ct, the sensitivity to irradiation in the pol3-ct pol2-16 double mutant should be increased. As shown in Fig. 5, this is not observed. Furthermore, since the hyporec phenotype of pol3-ct is conserved in the pol2-16 context, we conclude that short conversion tracts do not require the activity of Polε.
Effects of irradiation on Polε and Polη mutants in POL3 and pol3-ct homozygous backgrounds. (A) Effects of UV light on the pol2-16 and pol3-ct single mutants and on the pol2-16 pol3-ct double mutant. Frequencies of survival and of UV-induced HR between arg4 heteroalleles are plotted. (B) Effects of γ irradiation on the pol2-16 and pol3-ct single mutants and on the pol2-16 pol3-ct double mutant. Frequencies of survival and of UV-induced HR between arg4 heteroalleles are plotted. All the curves correspond to the mean values from at least four independent experiments. (C and D) Same as panels A and B, respectively, except that pol2-16 is replaced with rad30.
Induced recombination is not promoted by translesion polymerase Polη in yeast.Recently, Polη has been implicated in short conversion tracts in vertebrates (17). Therefore, we investigated induced HR in a strain with a deletion of the RAD30 gene encoding yeast Polη. Polη operates in a subpathway of postreplication repair, and therefore, rad30Δ cells are only mildly sensitive to UV irradiation and are resistant to γ rays (Fig. 5C and D). Interestingly, we found that the rad30Δ mutant shows elevated HR frequencies compared to those in the WT specifically after UV treatment (Fig. 5C). This result can be interpreted as an increase in long gene conversion tracts in the rad30Δ mutant. However, this hyperrecombinogenic phenotype is not observed after γ-ray irradiation.
Interestingly, we found that the pol3-ct rad30Δ double mutant does not show an increased sensitivity to irradiation compared to the sensitivities of the single mutants (Fig. 5C and D). This implies that DNA synthesis during HR is still efficient in the double mutant. In addition, pol3-ct leads to a threefold decrease of UV-induced Arg+ regardless of the presence of Polη (Fig. 5C and D). This indicates that, if Polη is implicated in HR in yeast, its role is either minor or fully redundant.
DISCUSSION
The analysis of the pol3-ct mutation allowed us to address several key issues concerning the role of DNA polymerases during HR in vegetative cells.
Polδ-ct does not impair DNA replication.Two novel genetic interactions involving pol3-ct are in agreement with the conclusion that DNA replication is not impaired in the mutant. The pol2-16 mutation is a deletion in the catalytic domain of Polε that eliminates its polymerase and exonuclease activities (18). Interestingly, a haploid strain harboring both pol2-16 and pol3-01 (the 3′ to 5′ exonuclease-deficient mutant of Polδ) is dead (26, 29). Similarly, pol2-16 is synthetic lethal with thermosensitive alleles of POL3 at permissive temperatures (29). Hence, the pol2-16 allele is extremely sensitive to impaired DNA synthesis occurring during chromosomal DNA replication, because of a defective Polδ. We found that pol2-16 pol3-ct strains are alive and do not show any negative interaction in the growth of dissected spores (data not shown). Additionally, unlike the pol3-01 mutator allele of POL3, pol3-ct msh2Δ double mutants could be easily recovered from diploid strains heterozygous for msh2 and pol3-ct.
Polδ-ct defect in HR is not restricted to meiosis and to DSB-initiated events.Spontaneous heteroallelic HR is not decreased in pol3-ct strains. However, the defect of pol3-ct is not restricted to meiosis (21), since pol3-ct exhibits shorter gene conversion tract lengths after HO-induced DSBs and is defective in UV-induced HR, which is mainly initiated by single-stranded gaps (19). Thus, an additional important conclusion from the present work is that the pol3-ct defect does not depend on the nature of the initiating lesion (DSBs versus single-strand breaks or gaps). The absence of an effect on spontaneous recombination could indicate that WT Polδ is recruited merely in response to DNA damage and not for the DNA synthesis required during spontaneous HR. However, spontaneous HR is not altered in pol2-16 mutants, nor is it decreased in pol3-ct pol2-16 double mutants (data not shown), strongly suggesting that Polδ does indeed participate in spontaneous HR. Alternatively, it is possible that HR initiated during DNA replication is monitored by proteins present at the replication fork, like Mrc1 (34), that affect gene conversion tracts and abolish the defect of Polδ-ct, a hypothesis that is currently under investigation.
Polδ-ct defect results from shorter HR intermediates rather than from defective MMR.The pol3-ct allele leads to a deficit of induced recombinants. This might result from increased lethality due to inefficient HR. This hypothesis can be ruled out since pol3-ct mutants are not sensitive to DSBs formed after γ-ray irradiation at the doses for which we see an effect on recombination or the expression of the HO endonuclease. Alternatively, this might result from an altered MMR. Indeed, gene conversion is the result of MMR, and it has been proposed that the exonuclease activity of Polδ participates in the removal of a mismatch-containing DNA strand (16). Yet, in meiosis, we found that a palindromic insertion at the HIS4 locus that usually escapes MMR, when included in hDNA, shows a decrease of non-Mendelian segregation in the pol3-ct background that is comparable to the decrease found with well-repairable markers. We now show that the decrease of induced HR in the pol3-ct mutant is independent of both the long-patch and the short-patch MMR pathways. In addition, since pol3-ct is not a mutator allele (21 and our unpublished data), it is unlikely that pol3-ct exhibits a defect in its exonuclease activity. Thus, low levels of induced recombinants are not a consequence of an altered MMR. Alternative hypotheses can account for our observations. On one hand, reduced recovery of induced recombinants could reflect a persistence of the mutant cells in the G2 phase of the cell cycle, thus favoring the repair off the sister chromatid. On the other hand, the nature of the HR intermediates could be different in the mutants. We favor the second hypothesis for the following reasons. (i) Physical analysis of gene conversion tract length linked to the repair of a targeted DSB indicates that the tracts are shorter. (ii) We performed a UV-induced recombination experiment using the URA3 system depicted in Fig. 4 and looked at the coconversion of XbaI and a BamHI centromere proximal marker located at 426 bp that is not shown in Fig. 4 (unpublished data). We found that, among the XbaI convertants yielding a uracil prototroph, 20% did include the neighboring BamHI site in the mutant, versus 50% in the control. This experiment confirms the presence of shorter UV-induced conversion tracts in the mutants. In conclusion, the pol3-ct defect leads to shorter intermediates, indicating that Polδ participates in hDNA elongation.
Polδ-ct yields shorter repair synthesis but does not bias the choice of the repair pathway.In meiosis, the recombination events are equally distributed among the DSBR and the SDSA pathways (3, 22, 42). The three- to sixfold reduction in intragenic conversion can only be explained if the pol3-ct allele affects both pathways. In mitosis, however, the SDSA pathway is predominant, as suggested by the strong bias toward NCO events under the control of the Srs2 helicase (15, 34). Intriguingly, the pol3-ct allele leads to a similar decrease in gene conversion in meiosis and in vegetative cells, indicating that its defect is probably independent of the repair pathway.
In the SDSA pathway, maximal hDNA length corresponds to the full amount of nucleotides incorporated to compensate for the initial end resection (Fig. 1). However, a synthesis that produces a shorter stretch of DNA is still sufficient for reannealing and subsequent repair out of the context of a D-loop. In the pol3-ct mutant, shorter conversion tracts could reflect a lesser amount of nucleotide incorporation after strand invasion, but they do not impede repair. In the DSBR pathway, the incorporation of fewer nucleotides could result from the capture of the D-loop by the second end before the full amount of resected DNA is synthesized, thus yielding shorter conversion tracts.
Interestingly, a shorter length of DNA synthesis is expected to facilitate the displacement of the invading strand and favor SDSA. However, we observed the formation of complex events following either HO induction (Fig. 4B) or UV irradiation (our unpublished results) which likely reflect symmetric hDNA produced by branch migration followed by MMR (Fig. 1) (28). These events occur as frequently in the pol3-ct mutant as in the WT. Similarly, spontaneous COs resulting from the formation of ligated dHJs also occur as frequently in the pol3-ct mutant as in the WT. These results indicate that the length of the repair synthesis does not favor the SDSA pathway at the expense of DSBR.
Polδ is the major polymerase implicated in HR.If two or more DNA polymerases were equally implicated in HR, we would expect modest effects in single mutants. Yet, the pol3-ct defect leads to a three- to sixfold decrease of induced heteroallelic HR (Table 2). Similarly, following an HO-induced DSB, the LOH frequency of the 2.6-kb upstream marker is decreased 2.5-fold and could be underestimated by the presence of COs formed in G2 (Fig. 4). In meiosis, the decrease of gene conversion in the pol3-ct mutant reaches sixfold (21). Such an important decrease could be interpreted as a replacement of the normal polymerase(s) used for HR by Polδ-ct, yielding mainly shorter tracts of gene conversion. However, this possibility is unlikely, given that pol3-ct is recessive. Therefore, the pronounced effect of pol3-ct on HR shows that the role of Polδ in hDNA extension is nonredundant and indicates that Polδ is preferentially recruited for long HR events.
We tested the pol2-16 allele, which neither synthesizes DNA nor exhibits sensitivity to γ rays (18), which indicates that if Polε is participating in HR, its involvement can be efficiently replaced. Interestingly, pol2-16 does not affect induced heteroallelic HR, suggesting that gene conversion tracts are unchanged in this mutant. In addition, we found that the pol3-ct pol2-16 double mutant does not lead to increased sensitivity to irradiation, a phenotype expected when DNA synthesis during HR is impaired. Moreover, the defect in induced HR observed in pol3-ct is independent of pol2-16. Therefore, the pol3-ct defect in induced recombination is not the result of shorter intermediates formed by Polε. Thus, the role of Polε in HR remains an open question.
RAD30 encodes the translesion polymerase Polη which has been implicated in the short extension of 3′ strands during HR (17, 25). In addition, mutants for Polη show longer gene conversion tracts during immunoglobulin gene conversion in chicken DT40 cells (17). If yeast Polη were also responsible for short tracts of gene conversion, we would expect the mutant to exhibit longer conversion tracts and, hence, elevated recombination levels. Indeed, we found that rad30Δ mutants show a hyperrecombinogenic phenotype after UV. However, this effect is not observed after γ-ray irradiation, a result which suggests a frequent channeling of UV lesions into the HR pathway rather than a general increase in the formation of long gene conversion tracts. In addition, pol3-ct does not alter the sensitivity to irradiation of the rad30Δ strain, while rad30Δ has no effect on the tract length observed in the pol3-ct strain. Therefore, either Polη does not participate in DNA synthesis during HR or it plays a role that is shared with another polymerase.
Because the defect of pol3-ct does not depend on Polε or Polη, it is likely that DNA synthesis during HR is performed principally by Polδ. In agreement with this idea, heteroallelic thermosensitive mutant alleles of POL3 do not recombine to form a WT copy of POL3 following irradiation in diploid cells (8), a result indicating that induced HR is rare in the context of an inactive Polδ. Thus, we propose that Polδ is preferentially recruited to the 3′ invading end to achieve recombinational DNA synthesis in yeast.
pol3-ct, a new tool with which to study the late steps of HR.We propose a decreased processivity to explain the pol3-ct defect in HR. It could be due to an unstable loading of Polδ-ct by the proliferating-cell nuclear antigene (PCNA). Indeed, DNA synthesis in HR requires PCNA (44), and posttranslational ubiquitin and SUMO modifications of PCNA provide a mechanism by which the pathway choice for the processing of DNA lesions can be made (12, 30, 32). Similarly, it is possible that a specific posttranslational modification of PCNA is necessary to recruit and stabilize Polδ for origin-independent DNA synthesis during HR. Polδ-ct would be stabilized inefficiently by the modified PCNA and would unload rapidly as soon as a stalling occurs (5′-end contact or heterology to include into hDNA). The recessivity of pol3-ct would be explained by a competition strongly in favor of the WT Polδ, firmly loaded onto PCNA. In addition, a specific interaction between Polδ and a modified PCNA would also explain a preferential recruitment during HR.
The defect of Polδ-ct may also reveal a loss of interaction with a DNA helicase responsible for the unwinding of the donor molecule. Yeast strains lacking Sgs1, a member of the recQ helicase family that includes human BLM and WRN helicases, exhibit the same hyporec phenotype following UV and γ-ray irradiation as the pol3-ct mutant (10). It is therefore possible that Sgs1 assists Polδ in its role in hDNA elongation, perhaps by displacing the D-loop ahead of Polδ.
The study of pol3-ct has allowed us to demonstrate not only the preferential involvement of Polδ in HR but also its crucial participation in hDNA extension. Hence, the pol3-ct mutant is a powerful tool with which to study the late steps of the HR process (DNA repair synthesis, hDNA extension, capture of the 5′ end of the break, and junction migration) and find partners involved in these processes.
ACKNOWLEDGMENTS
We thank J. Nickoloff, A. Sugino, and R. Woodgate for kindly providing yeast strains. We also thank E. Coïc and S. Marcand for critical reading of the manuscript and B. Asselain for his enlightened advice on statistical analysis.
This work was supported by a grant of the European Community (LSHG-CT-2003-503303), a grant from the INCa (PL003), and a grant from the Agence Nationale de la Recherche (ANR-07-BLAN-0350-01) and by the CNRS and the CEA.
FOOTNOTES
- Received 6 September 2007.
- Returned for modification 26 October 2007.
- Accepted 5 December 2007.
- Copyright © 2008 American Society for Microbiology
