Mus81 and Yen1 promote reciprocal exchange during mitotic recombination to maintain genome integrity in budding yeast

Abstract

Holliday junction (HJ) resolution is required for segregation of chromosomes and for formation of crossovers during homologous recombination. The identity of the resolvase(s) that functions in vivo has yet to be established, although several proteins able to cut HJs in vitro have been identified as candidates in yeasts and mammals. Using an assay to detect unselected products of mitotic recombination we found a significant decrease in crossovers in the Saccharomyces cerevisiae mus81Δ mutant. Yen1 serves a back-up function responsible for resolving intermediates in mus81Δ mutants, or when conversion tracts are short. In the absence of both Mus81 and Yen1 intermediates are not channeled exclusively to non-crossover recombinants, but instead are processed by Pol32-dependent break-induced replication (BIR). The channeling of recombination from reciprocal exchange to BIR results in greatly increased spontaneous loss of heterozygosity (LOH) and chromosome mis-segregation in the mus81Δ yen1Δ mutant, typical of the genomic instability found in tumor cells.

Introduction

The primary function of homologous recombination (HR) in mitotic cells is to repair double-strand breaks (DSBs) or single-strand gaps that form as a result of replication fork stalling and/or collapse, from processing of spontaneous damage and from exposure to DNA damaging agents. Spontaneous mutagenesis, chromosome loss and chromosome rearrangements increase in the absence of HR highlighting the importance of HR in the maintenance of genome integrity (Moynahan and Jasin, 2010). During meiosis, the repair of programmed DSBs by HR is essential to establish a physical connection between homologous chromosomes to ensure their correct disjunction at the first meiotic division.

The homology-dependent repair of DSBs initiates with the processing of DNA ends to create 3′ single-stranded DNA (ssDNA) tails, the substrate for binding by the Rad51 strand exchange protein (Mimitou and Symington, 2009). Once formed, the Rad51-ssDNA complex searches for a homologous sequence in dsDNA and promotes invasion of the ssDNA into donor dsDNA to form a joint molecule with a displaced strand (D-loop) (San Filippo et al., 2008). DNA polymerase δ extends the 3′ end from the broken chromosome using the unbroken homologous sequence as a template, restoring those residues lost by end resection (Li et al., 2009). To resolve the intermediate by synthesis-dependent strand annealing (SDSA), the invading strand that has been extended by DNA synthesis is displaced and anneals to complementary sequences exposed on the other side of the break, yielding exclusively non-crossover products (Fig. 1). Srs2, RTEL1 and Mph1/FANCM helicases are implicated in D-loop collapse to form non-crossovers (Mimitou and Symington, 2009).

Figure 1.

Model for repair of DSBs by homologous recombination (see text for details). STR=Sgs1-Top3-Rmi1. Scissors indicate sites of HJ cleavage; dual cleavages in the same plane generate non-crossovers, whereas cleavages in opposite planes result in crossover products.

Alternatively, the displaced strand of the D-loop can anneal with complementary sequence on the other side of the break to form a double Holliday junction intermediate (dHJ) connecting the two parental duplexes. Such intermediates have been detected during meiotic and mitotic recombination in Saccharomyces cerevisiae (Bzymek et al., 2010; Schwacha and Kleckner, 1995). Removal of the two HJs and separation of the recombinant duplexes can occur by dissolution or endonucleolytic cleavage (Fig. 1). In vitro studies have shown dissolution of DNA structures containing two HJs by human BLM, a member of the RecQ helicase family, acting in concert with the type I topoisomerase TopoIIIα and the accessory proteins Rmi1 and Rmi2 (Singh et al., 2008; Wu and Hickson, 2003; Xu et al., 2008). This activity would be predicted to promote repair by a non-crossover mode, consistent with genetic data showing increased mitotic crossovers in the absence of BLM or the yeast ortholog, Sgs1 (Chaganti et al., 1974; Gangloff et al., 1994; Ira et al., 2003). While the helicase pathways are well established to prevent crossovers in mitotic DSBR, the nucleases that function in resolution of recombination intermediates to form crossover products remain to be fully elucidated.

The MUS81 gene was originally found in yeast two hybrid screens by the interaction of its gene product with Rad54 in S. cerevisiae and with Cds1 in Schizosaccharomyces pombe, and by the synthetic lethality of the mus81 mutation with sgs1 (Boddy et al., 2000; Interthal and Heyer, 2000; Mullen et al., 2001). Mus81 interacts with Mms4/Eme1 to form a structure specific nuclease, related to the Rad1/XPF-Rad10/ERCC1 heterodimeric nuclease, which cleaves a variety of branched structures, including 3′ flaps, D-loops and nicked HJs (Kaliraman et al., 2001; Osman et al., 2003). Fission yeast mus81 mutants have very low spore viability and accumulate meiotic recombination intermediates (Boddy et al., 2001; Cromie et al., 2006). Based on the greater activity in cleavage of D-loops and nicked HJs than intact HJs, Osman et al. (Osman et al., 2003) proposed that Mus81–Eme1 processes a strand exchange intermediate generated prior to a fully ligated dHJ intermediate, channeling intermediates into crossover products. In contrast, mus81 mutants of S. cerevisiae exhibit only a two-fold reduction in meiotic crossovers and no defect in mitotic crossovers (Argueso et al., 2004; de los Santos et al., 2003; Ira et al., 2003; Jessop and Lichten, 2008; Oh et al., 2008; Robert et al., 2006). Furthermore, mice lacking Mus81 show near normal fertility (Dendouga et al., 2005; Holloway et al., 2008; McPherson et al., 2004). mus81 mutants exhibit high sensitivity to DNA damaging agents that cause replication fork stalling and fail to process forks in response to replication stress, suggesting an important role for Mus81-Eme1/Mms4 in replication fork repair (Hanada et al., 2007).

In the same screen that identified MUS81 and MMS4 as essential for viability of the sgs1 mutant, alleles of slx1 and slx4 were identified (Mullen et al., 2001). Unlike mus81, the synthetic lethality of slx1 or slx4 with sgs1 is not suppressed by rad51 mutation suggesting the lethality might be due to problems other than, or in addition to, unresolved recombination intermediates (Fricke and Brill, 2003). In vitro, the yeast Slx1-Slx4 complex preferentially cleaves 5′ flap structures; however, recent studies report efficient HJ cleavage by the human SLX1-SLX4 complex (Coulon et al., 2004; Fricke and Brill, 2003; Klein and Symington, 2009). Slx4 also interacts with the Rad1-Rad10 nuclease (Klein and Symington, 2009), and both MEI9/Rad1 and MUS312/Slx4 of Drosophila melanogaster are required for meiotic crossovers (Sekelsky et al., 1995; Yildiz et al., 2002). Similarly, Caenorhabditis elegans HIM-18/Slx4 functions in meiotic crossover formation (Saito et al., 2009). A novel member of the Rad2/XPG family of nucleases called GEN1 (Yen1 in budding yeast) was recently identified as the endonuclease responsible for cleavage of intact HJs in mammalian cell extracts (Ip et al., 2008).

The role of these different structure-specific nucleases in mitotic recombination has not been thoroughly investigated. This is relevant to tumor development because a crossover can lead to LOH of an entire chromosome arm centromere distal to the site of exchange if the recombinant chromosomes segregate to opposite poles during mitosis. It is estimated that 40% of the cases of hereditary retinoblastoma are caused by mitotic recombination leading to loss of the wild type copy of the RB1 gene (Hagstrom and Dryja, 1999). Because budding yeast has the recently identified Yen1/GEN1 Holliday junction resolvase, which is absent from S. pombe, we considered the possibility that the less severe recombination phenotype of the budding yeast mus81 (or mms4) mutant is due to the activity of Yen1 (Ip et al., 2008). To test this hypothesis we generated a mus81Δ yen1Δ double mutant and determined the phenotype in assays to measure DSB-induced mitotic recombination, spontaneous LOH, and chromosome aneuploidy.

Results

Mus81 and Yen1 nucleases promote reciprocal exchange associated with DSB-induced mitotic recombination

To determine whether the Mus81-Mms4 (hereafter Mus81) and/or Yen1 nucleases function in resolution of mitotic recombination intermediates we measured crossovers associated with gene conversion (Fig. 2A). The genes encoding resistance to hygromycin (Hph) or nourseothricin (Nat) were inserted at the his3 locus, 154-kb downstream of the ADE2 locus in diploid strains with two different ade2 alleles, one with an insertion of the I-SceI cut site (ade2-I), the other with a 2-bp deletion at the NdeI site (ade2-n) (Mozlin et al., 2008). In addition, the strains are heterozygous for a URA3 replacement of the MET22 locus on the other chromosome arm to ensure that LOH for Hph or Nat is due to crossing over and not chromosome loss. The diploid strains also contain a P_GAL-I-SCEI cassette integrated at the lys2 locus for regulated expression of the I-SceI endonuclease.

Figure 2.

Genetic assay for DSB-induced gene conversion and associated LOH. (A) Schematic of the diploid strain; the I-SceI cut site was inserted 950 bp upstream of the NdeI site, which is mutated on the other homolog (shown by *). (B) Colonies formed after 1 hr of I-SCEI induction in the wild-type strain. ade2 mutants accumulate a red pigment resulting in red colonies, whereas Ade⁺ recombinants form white colonies. (C) Repair of DSBs in G2 can occur by short tract or long tract gene conversion to produce ADE2 or ade2-n recombinants. (D) Crossovers associated with gene conversion can be detected if the two participating chromatids segregate to opposite poles during mitosis. If the crossover chromatids segregate to the same daughter cell heterozygosity for Nat and Hph is maintained (not shown). Break-induced replication results in homozygosis of the Nat marker and is detected regardless of the chromatid segregations. See also Figs. S1 and S2

A gross defect in DSBR would be expected to result in high frequency chromosome loss or reduced survival following I-SceI induction. Cultures of wild type, mus81,Δ yen1Δ and mus81Δ yen1Δ homozygous diploids with the recombination reporter were diluted and plated on solid medium containing glucose (non-inducing) or galactose (inducing) to determine the plating efficiency. The diploids exhibited a plating efficiency of 0.75–1 (cfu galactose/cfu glucose), with no significant difference between them (Fig. S1). A low frequency of DSB-induced chromosome loss was observed for the mus81Δ yen1Δ mutant (3–4%), but not for the other strains. Thus, all of the strains are proficient for DSBR. To monitor the kinetics of DSB formation, I-SceI was induced by addition of galactose to cultures of wild type, mus81Δ, yen1Δ or mus81Δ yen1Δ diploids. DNA was extracted from cells at different times after I-SceI induction and the cutting efficiency was determined by Southern blot analysis. In the wild type and single mutants >50% of DNA was cut by I-SceI 1.5 hr after induction, but there was a delay in detection of the I-SceI-cut fragments and lower yield in the mus81Δ yen1Δ double mutant (Fig. S2). Furthermore, by plating on selective media the maximal number of Ade⁺ recombinants was detected after a 2 hr induction for the wild type and single mutants, and at 2.5–4 hr in the double mutant (Fig. S1). The analysis of recombinants was conducted after a 1–1.5 hr I-SceI induction for the wild-type strain and single mutants, and 3 hr after I-SceI induction in the double mutant, to optimize recovery of colonies with both daughter cells even though higher levels of recombinants could have been obtained with longer inductions.

Following liquid I-SceI induction cells were plated on non-selective medium containing glucose to prevent further expression of I-SceI and grown for 2–4 days (Fig. 2B). From the wild type strain, 8.7% of the resulting colonies were white (Ade⁺), 24.7% were sectored red/white and 66.6% were red (Table S1). Red colonies that were due to long tract gene conversion were distinguished from those that had not induced I-SceI by a re-induction assay (see Experimental Procedures) (Nickoloff et al., 1999). In the wild type strain more than half of the red colonies (40.3% of the total) were Ade⁻ recombinants and only 26.3% of the total colonies scored had not recombined. If only one sister chromatid were cut by I-SceI in G2 cells, repair would occur preferentially from the sister template (Kadyk and Hartwell, 1992), restoring the I-SceI cut site. Recombinants can only be detected if both sisters are cut, either from replication of a chromosome digested by I-SceI in the G1 phase, or simultaneous cleavage of both sister chromatids in S/G2, followed by repair from the intact chromatids with the ade2–n allele. Even though some DSBs might be induced in G1, we presume their repair occurs in G2 (Aylon et al., 2004; Ira et al., 2004; Lee and Petes, 2010). Accordingly, the solid white colonies arise when both broken sister chromatids are repaired by short tract gene conversion, the red colonies are due to two long tract conversions, and the red/white colonies correspond to repair of one sister by short tract conversion and the other sister by long tract conversion (Fig. 2C). We rarely observed red/white colonies in which the red sector retained the ade2–I allele, indicating both sister chromatids are usually cut and repair occurs in G2. Because the solid red and white colonies could correspond to one of the two daughter cells from a recombinant cell that had divided before plating, the red/white-sectored colonies are the most informative to the mechanism of recombination.

The unselected colonies from the I-SceI induction were replica plated to medium containing hygromycin or nourseothricin to detect LOH events (Fig. 2D), and to medium lacking uracil or methionine to detect chromosome loss. Note that only one of the two possible segregations of the sister chromatids following reciprocal exchange will lead to LOH for the Nat and Hph markers; therefore, the number of these events was doubled to generate percent crossovers and the same number subtracted from the class with no LOH. Forty-nine percent of the red/white-sectored colonies maintained heterozygosity for the Nat and Hph markers as expected for gene conversion without a crossover (NCO class). Two classes of crossover (CO) events were recovered. Eleven percent of events were due to a short tract conversion associated with a crossover forming a white Hyg^S Nat^R sector and a red recombinant sector that was Hyg^R Nat^S. For 38%, the white half of the sectored colony was Hyg^R Nat^S and the red half Hyg^S Nat^R. The two CO classes are summed in Figure 3B. The remaining 2% of red/white-sectored colonies was heterozygous for the drug markers in one sector and homozygous for Nat (or less frequently Hph) in the other sector. Because these events retain heterozygosity for the MET22 and URA3 markers in both sectors they cannot be due to chromosome loss in one sector. These events are best explained by break-induced replication (BIR); however, BIR should yield Nat^R in the homozygous sector, whereas some events recovered were Hyg^R (Table S2). The rare Nat^R Hyg^R/Hyg^R sectored colonies could be due to unresolved recombination intermediates that are broken during cell division and undergo BIR in the next cell cycle (see Discussion). The ability to distinguish between reciprocal crossovers and BIR is a unique feature of this assay.

Figure 3.

Mus81 and Yen1 are required for reciprocal exchange. (A) Distribution of recombinant colony types in wild type, mus81Δ, yen1Δ and mus81Δ yen1Δ diploids. Percent events for each strain among the red/white sectored (B), red (C) and white (D) colonies. See also Tables S1 and S2.

There was no change in the distribution of recombinant colony classes, or in the type of recombination event scored among the red/white sectored colonies, in the yen1Δ mutant compared with wild type (Fig. 3). However, the mus81Δ and the mus81Δ yen1Δ mutants both showed an increase in the percent of white colonies and a decrease in the percent of red/white sectored colonies (Fig. 3A). The proportion of red recombinant colonies was also increased in the mus81Δ yen1Δ mutant compared with mus81Δ or wild type. The 8-fold reduction in red/white sectored colonies observed in the mus81Δ yen1Δ mutant does not appear to be due to more cells dividing before plating because it was also observed when shorter induction times for I-SceI were used or if the I-SceI induction was performed in the presence of nocodazole (data not shown). We favor the hypothesis that in the mus81Δ yen1Δ mutant one recombinant daughter cell frequently dies to account for the increased number of solid red and solid white colonies. The distribution of recombination events among the red/white-sectored colonies was also significantly different in the mus81Δ and mus81Δ yen1Δ mutants compared with wild type (Fig. 3B). The percent COs was reduced from 49% to 34% in the mus81Δ mutant (P=0.009) and the number of BIR events increased from 1.9% to 5.9% (P=0.024). Only 12.8% CO events were recovered from the mus81Δ yen1Δ double mutant, significantly less than wild type (P=0.0001) and the mus81Δ single mutant (P=0.0003). In addition, the number of BIR events in the mus81Δ yen1Δ double mutant was 10-fold higher than wild type (P=0.0001) and 3-fold higher than the mus81Δ single mutant (P=0.0002). The decrease in COs in the mus81Δ mutant compared with wild type is compensated for by an increase in the NCO class, but the additional decrease in COs in the mus81Δ yen1Δ double mutant does not result in a further increase in NCOs over that seen in the mus81Δ single mutant and instead these are recovered as BIR-like events.

Analysis of the colonies displaying a solid white or red color revealed a sectored phenotype for the drug resistance markers in 73–88% of the colonies analyzed from the wild type strain and these were classified as CO or BIR events, depending on the status of the markers. Two percent of the red colonies exhibited LOH for the Nat or Hph markers for the entire colony. These “other LOH” events are most likely due to cell division prior to plating so that only one daughter cell was scored. It is unlikely these are due to two BIR events based on the frequency of BIR in the wild type strain, but BIR or recovery of only one daughter cell could account for the large increase in this class of events in the mus81Δ and mus81Δ yen1Δ mutants. Similar to the red/white-sectored colonies, the red recombinant colonies exhibited a high percent of COs and the number was reduced in the mus81Δ and mus81Δ yen1Δ mutants; notably, only 3.4% of the colonies were due to COs in the mus81Δ yen1Δ mutant, compared with 51.6% for the wild type strain (Fig. 3C). The BIR and other LOH events were increased in the mus81Δ yen1Δ mutant compared with wild type and the mus81Δ single mutant. The same trends were seen for the white colonies except the percent COs was lower than observed for the red and red/white colonies, only 22% for the wild type strain and a significant reduction in all of the mutants, including yen1Δ (P=0.004). No CO events were recovered from the mus81Δ yen1Δ mutant, a significant reduction compared with wild type (P=0.0001) and the mus81Δ single mutant (P=0.03) (Fig. 3D). The complete phenotypic characterization of all classes of recombinants is presented in Table S2.

DSB-induced LOH in the mus81Δ yen1Δ double mutant is suppressed by pol32Δ

The increased BIR observed in the mus81Δ and mus81Δ yen1Δ mutants prompted us to test whether these events are dependent on POL32. POL32 encodes a non-essential subunit of the DNA Polymerase δ complex and mutants are defective for long tract gene conversion and BIR (Deem et al., 2008; Gerik et al., 1998; Jain et al., 2009; Lydeard et al., 2007; Smith et al., 2009). Furthermore, recent studies have shown Pol δ-PCNA extends the 3′ invading end of Rad51-catalyzed strand invasion intermediates in vitro (Li et al., 2009). Because the detection of I-SceI cut fragments and production of Ade⁺ recombinants were delayed in the pol32Δ mutant compared with wild type, cells were induced with I-SceI for 1.5 hr prior to plating (Figs S1 and S2). The first notable difference between the pol32Δ mutant and wild type was an increase in the number of white colonies (61% vs. 12%) and a decrease in red colonies (23% vs. 55%) among the recombinants after I-SceI induction (Fig. 4A, Table S1). In addition, the percent CO events among the red, white and red/white-sectored colonies was reduced by two-fold in the pol32Δ mutant (P=0.0004) (Fig. 4B, Table S2). These observations are consistent with shorter gene conversion tracts, and lower associated crossing over when conversion tracts are short. The same phenotype was reported for the pol3-ct mutant, which harbors a conditional mutation of POL3 encoding the catalytic subunit of Pol δ (Maloisel et al., 2004; Maloisel et al., 2008).

Figure 4.

The elevated BIR and residual crossovers in the mus81Δ yen1Δ mutant are Pol32-dependent. (A) Distribution of recombinant colony types in wild type and pol32Δ diploids. (B) Percent of each type of recombination event among the red/white-sectored colonies for each strain are shown. See also Tables S1 and S2

When pol32Δ was combined with mus81Δ we found a decrease in the percent COs in all colony types compared with mus81Δ and pol32Δ single mutants (Fig. 4B, Table S2). Strikingly, no CO events were recovered from 778 total colonies analyzed from the mus81Δ pol32Δ yen1Δ triple mutant. These results suggest Pol δ, or at least the function contributed by Pol32, is important for driving recombination intermediates to crossovers, or that the few events scored as COs in the mus81Δ yen1Δ mutant are not true reciprocal crossovers and instead some type of BIR event (see Discussion). There was a decrease in the BIR and other LOH events among the red/white sectored and red recombinant colonies from the mus81Δ pol32Δ yen1Δ mutant compared with mus81Δ yen1Δ (P<0.01). Thus, the elevated BIR in the mus81Δ yen1Δ mutant is Pol32 dependent.

DNA damage sensitivity of the mus81Δ mutant is increased by the yen1Δ mutation

The mitotic recombination data suggest Mus81 is the primary resolvase with Yen1 serving a back-up function. To determine whether Yen1 contributes to the DNA damage resistance of the mus81Δ mutant, we measured cell survival of the single and double mutants in response to DNA damaging agents. The yen1Δ mutant was not sensitive to ionizing radiation (IR), hydroxyurea (HU) or methyl methanesulfonate (MMS), but the mus81Δ yen1Δ double mutant exhibited higher sensitivity to all agents than the mus81Δ single mutant, consistent with recent studies (Figs. 5A and S3) (Blanco et al., 2010; Tay and Wu, 2010). The high sensitivity of the mus81Δ yen1Δ mutant to agents that stall replication, compared with the mild sensitivity to IR, suggests a defect in replication fork progression or restart. We also tested the DNA damage sensitivity of mutants with deletions of genes encoding the other structure-specific nucleases implicated in HJ resolution. The mus81Δ rad1Δ yen1Δ triple mutant showed higher sensitivity to IR and MMS than the double mutants (Fig. S3). This effect is mainly due to elimination of nucleotide excision repair because the mus81Δ rad14Δ yen1Δ triple mutant was also more sensitive than the mus81Δ yen1Δ double mutant, but there appears to be a Rad1 specific function as well (Fig. S3). The slx1Δ mutation did not increase the MMS or IR sensitivity of the mus81Δ, rad1Δ or yen1Δ single, double or triple mutants (Fig. S3). Thus the redundancy in DNA repair is only observed for the mus81Δ yen1Δ double mutant.

Figure 5.

MUS81 and YEN1 are required for MMS and HU resistance and prevent accumulation of joint molecules. (A) Overlapping roles of MUS81 and YEN1 in MMS and HU resistance. 10-fold serial dilutions of log-phase cultures were spotted onto plates containing the indicated amounts of MMS or HU. (B) Doubling times of the indicated haploid (1N) and diploid (2N) strains. (C) Cell cycle progression of cells synchronized in G1 with α-factor and released in media containing 0.03% MMS. One hr after damage induction cells were washed and transferred to MMS-free medium with α-factor to arrest at the next cell cycle before resuming replication. (D) Western blot of protein samples from cells harboring an HA-epitope tagged Rad53 after 1hr exposure to 0.03%MMS (+) or to buffer only (−). Rad53 was detected with α-HA. (E) PFGE analysis of samples of DNA from cells recovering after a 0.03% MMS treatment for 1 hr. PFGE gels were transferred to nylon membranes and hybridized with probes against chromosomes III and XV. (F) Quantitation of the percent unresolved chromosomes (indicative of joint molecules); error bars show standard deviation from three trials, * p<0.05 **p<0.005 compared to the amount of unresolved chromosomes in wild type. See also Figs. S3 and S4.

The mus81Δ yen1Δ double mutant exhibits a growth defect, G2 arrest and aberrant nuclear morphology

Although the mus81Δ and yen1Δ single mutants do not exhibit significant growth defects, the growth rate of the mus81Δ yen1Δ double mutant is reduced, and the growth defect is more pronounced in diploids than haploids (Fig. 5B). FACS analysis of synchronized cells did not reveal a significant delay in S phase progression for the mus81Δ yen1Δ mutant, but the FACS profile of unsynchronized cells showed the proportion of G2 cells was increased for the mus81Δ yen1Δ double mutant (Fig. 5C, Fig. S4). The G2 arrest was also apparent from the cell morphology (Fig. S4), as described previously (Blanco et al., 2010). When cells were arrested in G1 and released into medium containing MMS for 1 hr the mus81Δ and mus81Δ yen1Δ mutants both showed the same delay in S-phase progression as the wild-type strain, but a greater delay in the G2 to G1 transition suggesting lesions accumulate that trigger the G2 DNA damage checkpoint or there is a defect in chromosome segregation (Fig. 5C). The DNA damage checkpoint is activated normally in the mus81Δ and mus81Δ yen1Δ cells and there is a low level of Rad53 activation in the absence of exogenous damage in the mus81Δ yen1Δ mutant (Fig. 5D). However, the cells were unable to complete repair and resume cell cycle or they attempted mitosis with persistent toxic intermediates. During the second cell cycle after MMS treatment, the mus81Δ mutant cells accumulated in G2 and this appears to be a terminal arrest (>90% of cells are inviable when plated). The mus81Δ yen1Δ cells undergo lysis during the second cell cycle.

To determine whether unresolved branched structures accumulate in the mus81Δ yen1Δ mutant following MMS treatment, cells that had been arrested with α-factor were released into MMS-containing medium for 1 h, then harvested and resuspended in MMS-free medium containing nocodazole to hold the cells in G2 during repair. Intact chromosomes were analyzed by pulsed-field gel electrophoresis (PFGE) during the recovery period (Fig. 5E). A defect in completion of replication or accumulation of branched DNA molecules is expected to result in retention of chromosomes in the wells of the gel. To quantify recovery of intact chromosomes the gels were blotted and probed for chromosomes III (315 kb) and XV (1091 kb) (Fig. 5F). Full-length chromosomes were recovered within 1 hr after MMS treatment in the wild type and yen1Δ strains. In the mus81Δ mutant, chromosomes were retained in the wells 3 hr after recovery from MMS, especially the longer chromosome XV, and this defect was also apparent in the mus81Δ yen1Δ double mutant. These results are consistent with Mus81 resolving stalled replication forks to allow completion of DNA synthesis and/or resolving recombination intermediates formed in response to MMS.

Increased spontaneous LOH and chromosome aneuploidy in the mus81Δ yen1Δ mutant

The reduced proliferation and G2 arrest of the mus81Δ yen1Δ mutant could be due to unresolved replication and/or recombination intermediates interfering with chromosome segregation. To determine whether the mutants exhibit increased chromosome loss, the rate of spontaneous LOH was measured by selecting for loss of the heterozygous URA3 insertion using the drug 5-fluoroorotic acid (5-FOA) (Fig. 6A). Loss of URA3 can occur by mitotic recombination (crossover between the MET22 locus and CEN15, BIR or gene conversion) or by chromosome loss; to distinguish between them the 5-FOA resistant colonies were replica plated to YPAD/Nat medium to monitor simultaneous loss of the Nat marker. From the percent events that were Nat^S rates of mitotic recombination and chromosome loss could be determined. The rate of chromosome loss was very low in the wild type strain; this could be because XV is a large chromosome and is infrequently lost, or monosomy is poorly tolerated. Thus, most of the LOH in wild type is due to mitotic recombination. The rate of mitotic recombination was increased by 20-fold in the mus81Δ yen1Δ mutant (P=0.035), compared with wild type (Fig. 6B). We found 12 and 568-fold increases in chromosome loss in the mus81Δ (P<0.01) and mus81Δ yen1Δ mutants (P<0.05), respectively (Fig. 6C). There was no difference in the rate of LOH between wild type and the yen1Δ mutant.

Figure 6.

The mus81Δ and mus81Δ yen1Δ mutants exhibit increased spontaneous LOH and chromosome loss/gain. (A) Assay for spontaneous LOH. Spontaneous loss of the URA3 marker can occur by mitotic recombination or by chromosome loss. The rates of mitotic recombination (B) and chromosome loss (C) were calculated for the indicated genetic backgrounds; error bars indicate standard deviations. (D) Assay for chromosome mis-segregation using the ade2 direct repeats. (E) Relative Ade⁺ Ura⁺/Ade⁺ co-segregation rates for the indicated strains; error bars indicate standard deviations. (F) Schematic showing alternative pathways that could lead to the accumulation of joint molecules in G2/M. See also Fig. S5.

Although the rates of mitotic recombination and chromosome loss in the pol32Δ mutant were the same as in wild type (P=0.66 and P=0.28, respectively), the pol32Δ mutation reduced the rate of mitotic recombination in the mus81Δ (P<0.01) and yen1Δ mutants (P<0.05) to below the pol32Δ rate, suggesting both Mus81 and Yen1 promote spontaneous reciprocal exchange, and in the absence of these nucleases intermediates are channeled to Pol32-dependent BIR (Fig. 6B). Furthermore, pol32Δ reduced the mitotic recombination rate of the mus81Δ yen1Δ mutant by 18-fold (P<0.05). The high rate of mitotic recombination in the triple mutant could be due to more recombinogenic lesions resulting from aberrant replication. The pol32Δ mutation failed to suppress the elevated rate of chromosome loss in the mus81Δ or mus81Δ yen1Δ mutants (Fig. 6C); therefore, all of the suppression of LOH by pol32Δ is of recombination, presumably BIR. To confirm the increased LOH in the mus81Δ yen1Δ mutant was due to recombination, rates were measured in rad51Δ derivatives. The rad51Δ diploid showed a 20-fold decrease in mitotic recombination and a 10-fold increase in chromosome loss (Fig. S5). The rad51Δ mutation significantly reduced mitotic recombination and chromosome loss in the mus81Δ yen1Δ mutant indicating that some of the chromosome loss events are due to unresolved recombination intermediates.

To directly assess the mis-segregation of sister chromatids we made use of a construct containing truncated ade2 genes that share 900-bp of homology separated by a copy of the URA3 gene. This strain is Ade⁻ Ura⁺, but a spontaneous deletion results in an Ade⁺ Ura⁻ phenotype. Cells that have an Ade⁺ Ura⁺ phenotype arise by failure to segregate sister chromatids, or an increase in ploidy, coupled with a rare spontaneous deletion (Fig. 6D) (Chan and Botstein, 1993). We measured the rate of spontaneous Ade⁺ recombinants and Ade⁺ Ura⁺ recombinants; an increase in the ratio of Ade⁺ Ura⁺/Ade⁺ events is indicative of a chromatid mis-segregation defect. In the mus81Δ mutant there was a 2.5-fold increase in these events (P<0.01), and a 25-fold increase in the mus81Δ yen1Δ double mutant (P<0.005), consistent with a defect in sister chromatid segregation (Fig. 6E). When grown non-selectively on rich medium, the Ade⁺ Ura⁺ clones exhibited high reversion to Ade⁻ Ura⁺ or Ade⁺ Ura⁻ indicating the presence of an unstable disome.

A rad51Δ mutation was introduced to determine whether the sister-chromatid segregation defect was due to unresolved recombination intermediates. The Ade⁺ recombination rate in all the rad51Δ strains was increased ~2.5 to 4–fold compared to the wild type; however, we assume that the increase in Ade⁺ reversion does not affect the relative rates. A reduction in the mis-segregation rate was observed compared to the mus81Δ yen1Δ mutant (P<0.01), but the mus81Δ yen1Δ rad51Δ mutant still showed a 6-fold increase compared to the wild type (P=0.03). The increased mis-segregation rate of the mus81Δ mutant was unaffected by rad51Δ suggesting the partial suppression of the sister-chromatid segregation defect of the mus81Δ yen1Δ mutant is due to the presence of Mus81-dependent but Rad51-independent intermediates. This is consistent with the observation that rad51Δ synergizes with mus81Δ for MMS sensitivity (Ii and Brill, 2005). The Mus81 function in processing stalled replication forks could contribute to the Rad51-independent accumulation of replication intermediates in mus81Δ, which is also expected to cause mis-segregation of chromatids during mitosis (Fig. 6F).

Discussion

We developed an assay to detect unselected DSB-induced mitotic recombination with the goal of identifying the nucleases responsible for crossover formation in budding yeast. Our data indicate that Mus81 is the primary activity with Yen1 serving a back-up function. Surprisingly, recombination intermediates are not channeled exclusively to non-crossovers in the absence of Mus81 and Yen1, but instead a class of events is observed most consistent with BIR. Furthermore, spontaneous LOH resulting from BIR and chromosome loss/gain is increased in the mus81Δ mutant and to an even greater extent in the mus81Δ yen1Δ double mutant. Thus, resolution of recombination intermediates by endonucleolytic cleavage is essential to generate reciprocal crossovers, for chromosome segregation and maintenance of genome integrity.

Mus81 and Yen1 are required for mitotic crossovers

Diploids with ade2 heteroalleles and dominant heterozygous markers downstream of the ade2 locus were used to detect crossovers associated with DSB-induced gene conversion by LOH for the Hph and Nat markers. These events could be distinguished from chromosome loss using markers on the other chromosome arm. Furthermore, reciprocal and non-reciprocal (BIR) events were readily distinguished, a critical feature of the assay that enabled detection of the crossover defect of mus81Δ and mus81Δ yen1Δ mutants. In wild type cells, conversion is biased to long tract events (>900 bp) and these are more frequently associated with crossovers than short tract events, consistent with previous studies (Inbar et al., 2000; Mitchel et al., 2010). Using this system we found a higher frequency of crossovers associated with gene conversion than reported in previous studies of DSB-induced allelic recombination in budding yeast (Malkova et al., 1996; Nickoloff et al., 1999). The increased number of crossovers recovered in the ade2 assay could be due to marker and/or chromosome specific effects, or the methods used to detect and score events. Here, each sectored colony was considered as one event, but if both sectors are scored as two independent recombination events then the percent COs is lower. Ectopic assays have also been used to determine the percent crossovers associated with DSB-induced or spontaneous gene conversion (Inbar et al., 2000; Ira et al., 2003; Jinks-Robertson and Petes, 1986; Robert et al., 2006). The percentage of crossover events is generally lower in these assays than allelic assays and can be increased by using longer repeats (Inbar et al., 2000).

There was a significant decrease in the percent COs and an increase in the NCO and BIR classes among the red/white sectored colonies in the mus81Δ single mutant (Fig. 3). This trend was also seen in the red and white colonies, with an increase in the BIR and other LOH classes at the expense of COs. These data are consistent with Mus81 nuclease cleaving recombination intermediates to generate crossover products, and in the absence of this function intermediates are resolved as NCOs or are channeled to Pol32-dependent BIR. Based on the preferred substrates for Mus81 in vitro (Osman et al., 2003), it seems likely that crossovers are generated by two cleavage events, one of the 3’ flap formed by the extended D-loop that captures the non-invading end, and the other at the nicked HJ (Fig. 7A). The interaction with Rad54 would position Mus81 to cleave these structures following strand invasion (Interthal and Heyer, 2000). The increase in NCOs could be due to D-loop collapse or maturation to a dHJ intermediate that is subsequently dissolved by Sgs1-Top3-Rmi1 (STR) or resolved in a non-crossover configuration by Yen1. The high levels of crossovers that still form in the mus81Δ mutant are primarily dependent on Yen1. These could arise by Yen1 cleavage of a dHJ intermediate, or of a single HJ intermediate if the D-loop is also cleaved by Yen1 or by another structure-specific nuclease. Ira et al (Ira et al., 2003) reported no decrease in crossovers in the mus81Δ mutant using a physical assay to detect DSB-induced recombination between ectopic repeats. In their system the percent crossovers is very low and this may have precluded detection of a subtle defect in the mus81Δ mutant. Alternatively, an increase in BIR, which would produce the same size restriction fragments in the population as bona fide CO events, would have veiled a crossover defect.

Figure 7.

Models for Mus81 and Yen1 resolution of recombination intermediates. (A) The 3′ end of the strand invasion intermediate is extended by DNA Pol δ and then displaced to form a NCO or cleaved by Mus81 to generate a CO. If the D-loop undergoes reverse branch migration to form a dHJ, dissolution by STR will generate a NCO; resolution by Yen1 (or possibly Mus81) generates NCO or CO products. In the absence of Mus81, Yen1 is predicted to cleave HJs that are fully ligated or formed by reversed branch migration. Cleavage of the HJ without cleavage of the leading end of the D-loop might channel the intermediate to BIR. In the absence of Mus81 and Yen1, BIR might occur by a migrating D-loop intermediate, or the unresolved structures persist until mitosis. (B) An unresolved recombination intermediate (shown as a single HJ) might be broken during mitosis if the centromeres segregate to opposite poles. Segregation of one centric fragment with an intact chromatid might result in BIR in the next cell cycle (this would probably occur in S/G2, but only one chromatid is shown for simplicity) resulting in homozygosis of the Nat or Hph marker. If both chromatids were broken during mitosis then BIR in the two daughter cells would result in products indistinguishable from a reciprocal crossover.

Although the yen1Δ single mutant did not show a decrease in COs among the red/white or red colonies there was a significant decrease in this class of events among the white colonies (Fig. 3). The white colonies result from short tract gene conversion and are less frequently associated with crossovers in the wild type strain. It is possible that this class of crossovers arises when the DNA synthesis tract is short and the strand invasion intermediate is stabilized by reversed branch migration of the D-loop creating an intact HJ, or dHJ intermediate that is subsequently cleaved by Yen1 or dissolved by STR (Allers and Lichten, 2001) (Fig. 7A). If reversed branch migration forming a dHJ intermediate that could be dissolved by STR were responsible for the short tract (Ade⁺) events, it would explain the failure to recover COs among the white colonies formed in the mus81Δ yen1Δ double mutant.

There was a decrease in the percent COs among the red/white and red colonies derived from the mus81Δ yen1Δ double mutant, compared with the mus81Δ single mutant and wild type, consistent with both nucleases acting to resolve recombination intermediates to generate crossovers. Although COs were greatly decreased in the mus81Δ yen1Δ double mutant, there was not a corresponding increase in NCOs for the red/white and red colony classes compared to the mus81Δ single mutant, but instead, more Pol32-dependent BIR-like events were recovered. The BIR events could arise directly from the initial strand invasion if the leading end of the D-loop was not cleaved (Fig. 7A). However, a more likely explanation is that cells attempt to divide with an unresolved intermediate resulting in mechanical or enzymatic breakage of chromosomes and the fragments produced then engage in secondary recombination events in the following cell cycle (Fig. 7B). Such events might be the cause of the prolonged G2 arrest and increased doubling time of the double mutant. If the unresolved intermediate contained two HJs one would predict that it would be dissolved by the STR complex to form non-crossover products. Because events are not channeled exclusively to NCOs in the mus81Δ yen1Δ mutants it seems more likely that a single HJ intermediate persists that cannot be dissolved by the STR complex or resolved in the absence of the Mus81 and Yen1 nucleases. Formation of a single HJ intermediate would require cleavage of the leading end of the D-loop and second end capture. In support of this idea, intermediates containing a single HJ accumulate during meiosis in S. pombe mus81 mutants (Cromie et al., 2006). Furthermore, DSB-induced dHJ intermediates detected by physical monitoring occur at ten-fold lower levels in mitotic cells than during meiotic recombination (Bzymek et al., 2010). A recent study reported only a two-fold reduction in the resolution of a single HJ-containing plasmid substrate following transformation of the mus81Δ yen1Δ mutant (Tay and Wu, 2010). The high residual resolution in their assay could be due to replication of the substrate yielding the expected dimeric product, or cleavage of the substrate during transformation.

Pol32 is required for long tract gene conversion and crossovers

Previous studies suggested that Pol32 has a specific role in BIR (Lydeard et al., 2007). Here we show gene conversion tracts are shorter than wild type in the pol32Δ mutant and two-fold less likely to result in crossover recombinants. This phenotype is similar to the previously characterized pol3-ct mutant, and suggests the pol32Δ and pol3-ct mutations both reduce the processivity of Pol δ during recombinational repair (Maloisel et al., 2004; Maloisel et al., 2008). The DNA synthesis defect of the pol32Δ mutant is not exclusive to recombinational repair; a recent study showed pol32Δ mutants accumulate ubiquitylated PCNA in a non-perturbed cell cycle and activate the DNA damage checkpoint suggesting Okazaki fragments are not always fully extended (Karras and Jentsch, 2010). Thus, pol32Δ mutants have a general defect in extension of DNA synthesis tracts by Pol δ. Because BIR and double-strand gaps spanning >1 kb require extensive polymerization by Pol δ the pol32Δ mutant is extremely defective in these assays (Jain et al., 2009; Lydeard et al., 2007; Payen et al., 2008).

Here we show the increased BIR events observed in the mus81Δ and mus81Δ yen1Δ mutants are dependent on Pol32, consistent with the requirement for extensive DNA synthesis. It was surprising that the residual crossovers detected among the red/white and red colonies in the mus81Δ yen1Δ double mutant were eliminated by the pol32Δ mutation. We consider two explanations for this finding. First, it is possible that the limited extension of the invading strand in the pol32Δ mutant results in more events that perform second end capture by reversed branch migration of the D-loop resulting in dHJ intermediates that are dissolved exclusively by STR in the absence of Mus81 and Yen1 (Fig. 7). Second, an unresolved HJ intermediate might be broken by mechanical stress during mitosis if the two connected non-sister chromatids attempt to segregate to opposite poles. The fragments produced might undergo Pol32-dependent BIR in the following cell cycle. If both chromatids were broken and the centric fragments engaged in BIR then both sectors would become homozygous for the downstream drug markers generating a product indistinguishable from a true CO. Random breakage of one chromatid could lead to BIR in the next cell cycle and result in homozygosis of either the Hph or Nat marker. The consequences of segregating two chromatids connected by a HJ to the same daughter cell are unknown. Yen1 is localized in the nucleus during the G1 stage of the cell cycle and might be important to resolve any residual structures prior to S-phase (Kosugi et al., 2009). The failure to recover COs in the mus81Δ pol32Δ yen1Δ mutant (Fig. 5, Table S2) suggests the Slx1-Slx4 and Rad1-Rad10 nucleases do not resolve mitotic recombination intermediates in S. cerevisiae.

Spontaneous LOH in the mus81Δ yen1Δ mutant occurs primarily by BIR

Most of the spontaneous LOH events in wild type cells are due to mitotic recombination. The rate of LOH in this assay is similar to what has been reported for other chromosomes, but the percent of events due to chromosome loss is much lower than observed for chromosomes III and V that are more commonly used to measure loss rates (Dershowitz and Newlon, 1993; Klein, 2001). Chromosomes III and V are shorter than chromosome XV and might be more frequently lost than larger ones. In addition, monosomy is poorly tolerated for some chromosomes resulting in selection for rare endoreduplication events (Alvaro et al., 2006). These factors may contribute to the low chromosome XV loss observed.

A small increase in LOH was observed in the mus81Δ mutant and this was primarily due to Pol32-dependent BIR. Indeed, in the mus81Δ pol32Δ double mutant the rate of mitotic recombination was 5-fold lower than wild type indicating that Mus81 dependent spontaneous COs are a primary source of LOH. The pol32Δ yen1Δ double mutant also showed a significant decrease in spontaneous mitotic recombination compared with wild type, or the single mutants. These results suggest that Mus81 and Yen1 both function in crossover formation in the pol32Δ background and that in the absence of either activity BIR events increase. The large increase in spontaneous LOH observed in the mus81Δ yen1Δ double mutant is due to both Pol32-dependent BIR and Pol32-independent chromosome loss.

Chromosome aneuploidy in the mus81Δ yen1Δ mutant occurs by recombination dependent and independent mechanisms

Chromosome loss accounts for only a small percentage of the spontaneous LOH events in the wild type strain, but large increases in chromosome loss were found in the mus81Δ and mus81Δ yen1Δ mutants. An increase in spontaneous chromosome aneuploidy was also observed for Mus81⁻/⁻ mouse cells indicating this function of Mus81 is conserved (Dendouga et al., 2005; McPherson et al., 2004). Chromosome mis-segregation is predicted to result from failed resolution of recombination intermediates and could be one source of chromosome loss; however, chromosome loss can occur by a variety of other mechanisms. The rate of chromosome loss was reduced in the mus81Δ rad51Δ yen1Δ triple mutant compared with the mus81Δyen1Δ double mutant consistent with the failure to segregate unresolved recombination intermediates contributing to chromosome loss (Fig. S5). In the haploid assay that detects mis-segregation of sister chromatids a 2.5-fold increase was detected in the mus81Δ single mutant and a 25-fold increase in the mus81Δ yen1Δ double mutant (Fig. 6). The increased rate of chromosome mis-segregation in the mus81Δ mutant was not dependent on Rad51, suggesting these events are due to unresolved replication intermediates. This is consistent with the persistence of MMS-induced joint molecules detected by PFGE in mus81Δ mutant, and the increased MMS sensitivity of the mus81Δ rad51Δ double mutant (Ii and Brill, 2005). The increased chromosome mis-segregation of the mus81Δ yen1Δ strain was partially suppressed by the rad51Δ mutation suggesting most of the mis-segregation events in the double mutant are due to recombination. These data reveal the complexity of the mus81Δ phenotype and suggest the Mus81 nuclease plays two important roles in the cell: first, to resolve stalled replication forks to facilitate restart; and second, to cleave strand invasion intermediates to form reciprocal crossovers. Our results suggest that aberrant processing of recombination and replication intermediates could contribute to the chromosome instability associated with tumor cells (Schvartzman et al., 2010).

Experimental Procedures Summary (full methods are presented in SUPPLEMENTAL DATA)

DSB-induced recombination assay

Diploids (Table S3) were grown in YPR (1% yeast extract: 2% peptone; 2% raffinose) liquid medium until cultures reached an OD₆₀₀ of 0.4, and galactose was added to a final concentration of 2% to induce I-SceI. After 1–3 hrs cells were plated onto YPAD (1% yeast extract; 2% peptone; 2% dextrose; 10mg/l adenine) medium, incubated for 2 to 4 days, then replica-plated onto YPAD/Hyg/Nat, synthetic complete (SC)-Ade, SC-Met, SC-Ura and SCR-Ade+Gal media to classify recombination events. Statistical significance for the number of recombination events between given strains was calculated by Fisher’s exact test.

Spontaneous LOH assay and chromosome mis-segregation assays

Rates were determined by the method of the median from 9 independent cultures of diploid or haploid strains (Table S3) with the indicated reporter. The rates were measured at least three times with independent isolates for each strain and the mean values are presented. 5-FOA resistant colonies recovered from the LOH assay were tested for growth on YPAD/Nat to identify chromosome loss events. Statistical significance for the rates of spontaneous LOH, chromosome loss and chromosome mis-segregation between given strains was calculated using the Student’s unpaired t-test.