Mutants Defective in Rad1-Rad10-Slx4 Exhibit a Unique Pattern of Viability During Mating-Type Switching in Saccharomyces cerevisiae

Abstract

Efficient repair of DNA double-strand breaks (DSBs) requires the coordination of checkpoint signaling and enzymatic repair functions. To study these processes during gene conversion at a single chromosomal break, we monitored mating-type switching in Saccharomyces cerevisiae strains defective in the Rad1-Rad10-Slx4 complex. Rad1-Rad10 is a structure-specific endonuclease that removes 3′ nonhomologous single-stranded ends that are generated during many recombination events. Slx4 is a known target of the DNA damage response that forms a complex with Rad1-Rad10 and is critical for 3′-end processing during repair of DSBs by single-strand annealing. We found that mutants lacking an intact Rad1-Rad10-Slx4 complex displayed RAD9- and MAD2-dependent cell cycle delays and decreased viability during mating-type switching. In particular, these mutants exhibited a unique pattern of dead and switched daughter cells arising from the same DSB-containing cell. Furthermore, we observed that mutations in post-replicative lesion bypass factors (mms2Δ, mph1Δ) resulted in decreased viability during mating-type switching and conferred shorter cell cycle delays in rad1Δ mutants. We conclude that Rad1-Rad10-Slx4 promotes efficient repair during gene conversion events involving a single 3′ nonhomologous tail and propose that the rad1Δ and slx4Δ mutant phenotypes result from inefficient repair of a lesion at the MAT locus that is bypassed by replication-mediated repair.

IN the baker's yeast Saccharomyces cerevisiae, spontaneous and induced DNA double-strand breaks (DSBs) are primarily repaired by homologous recombination (reviewed in Pâques and Haber 1999). In the initial steps of repair, DSBs are acted upon by a 5′–3′ exonuclease activity to yield two 3′ single-stranded ends. These ends interact with RPA, Rad51, Rad52, Rad54, Rad55, and Rad57 to allow strand invasion into a homologous double-stranded donor sequence. DNA synthesis initiating from the 3′ invading end results in copying of DNA sequence from the donor locus, and recombination is completed either by resolution of a Holliday junction intermediate or by synthesis-dependent strand annealing (SDSA). Homologous recombination can also occur by nonconservative mechanisms including single-strand annealing (SSA) and break-induced replication (BIR). During SSA, a DSB located between repeated sequences is processed by 5′–3′ exonuclease activity and the 3′ single-stranded ends anneal at homologous sequences, resulting in deletion of the intervening sequence. In BIR, strand invasion of one 3′-end into a homologous sequence is followed by replication that continues along the chromosome arm (reviewed in Pâques and Haber 1999).

Mating-type switching in S. cerevisiae is a unidirectional gene conversion event in which a DSB created at the MAT locus is repaired using one of two silent mating-type cassettes, HMRa or HMLα (reviewed in Haber 1998). This programmed recombination event is initiated by HO endonuclease cleavage within MAT, and donor preference is such that cells preferentially repair the DSB using the donor sequence of the opposite mating type (Wu and Haber 1995, 1996; Wu et al. 1997; Haber 1998). Crossovers, which would lead to intrachromosomal deletions, are rarely associated with mating-type switching (Klar and Strathern 1984), and mating-type switching is thought to occur by a SDSA mechanism (McGill et al. 1989; Haber 1998; Pâques and Haber 1999; Ira et al. 2006).

The HO cleavage site at the MAT locus is located at the junction between homologous and nonhomologous sequence with respect to the donor cassette. Strand invasion is thought to be initiated by the 3′ tail that is homologous to the donor sequence, leaving the second 3′-end as a nonhomologous tail following annealing of the repaired invading strand (Figure 1A). Thus, a single 3′ nonhomologous tail must be removed to complete repair. Previous genetic studies have shown that 3′ nonhomologous tail removal depends on the activity of the Rad1-Rad10 endonuclease, as well as the Msh2-Msh3 DNA mismatch recognition complex (Fishman-Lobell and Haber 1992; Ivanov and Haber 1995; Kirkpatrick and Petes 1997; Saparbaev et al. 1996; Sugawara et al. 1997).

Figure 1.—

Synthesis-dependent strand annealing model for mating-type switching in S. cerevisiae (adapted from Pâques and Haber 1999). (A) Only the MATa and HMLα loci are shown. Mating-type switching is initiated by a DSB formed by HO endonuclease at the MATa locus near the Y/Z1 junction. This is followed by 5′–3′ resection to create 3′ single-stranded ends, and the 3′-end with homology to the HMLα donor sequence initiates strand invasion and primes DNA synthesis off of the donor template. Strand displacement from the donor sequence followed by annealing onto the broken chromosome results in the formation of a 3′ single-stranded nonhomologous tail that must be excised prior to the subsequent DNA synthesis and ligation steps. Rad1-Rad10-Slx4 is hypothesized to act in 3′ nonhomologous tail removal at this step. (B) Mating-type switching involving two nonhomologous ends due to insertion of KANMX sequence on the distal side of the break. A 3′ nonhomologous tail removal step is required to allow priming of DNA synthesis off of the invading strand. Repair then proceeds as above.

Rad1-Rad10 is a structure-specific endonuclease that cleaves DNA at the junction of double-stranded and 3′ single-stranded DNA (ssDNA) and has been characterized in its role during nucleotide excision repair (NER) as well as in the removal of 3′ nonhomologous tails and blocked 3′ termini, including Top1-associated DNA (Sung et al. 1993; Bardwell et al. 1994; Vance and Wilson 2002; Guzder et al. 2004). The importance of Rad1-Rad10 for its non-NER DNA processing functions is highlighted by the fact that mice lacking the mammalian homolog of Rad1-Rad10, ERCC1-XPF, exhibit features of premature aging, including a very reduced life span (20–38 days), severe runting, and abnormalities of the liver, skin, kidney, and spleen, while mice lacking other NER factors develop normally and have a normal life span (McWhir et al. 1993; Weeda et al. 1997).

In plasmid-based studies, both Rad1-Rad10 and Msh2-Msh3 are required for recombinational repair when two 3′ nonhomologous tails are present (Sugawara et al. 1997; Colaiácovo et al. 1999). Repair events involving only one nonhomologous end are also hypothesized to require Rad1-Rad10 and Msh2-Msh3, although a second, less efficient pathway involving the 3′–5′ proofreading activity of DNA polymerase δ has been shown to remove 3′ ssDNA <30 nucleotides long (Pâques and Haber 1997; Colaiácovo et al. 1999). The Haber lab previously reported that mating-type switching in G₁-arrested cells is significantly less efficient in rad1Δ mutants, but stated no further defects (Holmes and Haber 1999b).

Rad1-Rad10 and Msh2-Msh3 are also required during SSA, which involves two nonhomologous tails. The requirement for Msh2-Msh3 depends on the length of the annealed region; annealed regions >1 kb in length are repaired independently of Msh2-Msh3. Thus, Msh2-Msh3 is thought act by binding and stabilizing the double-strand/single-strand junctions to promote Rad1-Rad10-dependent cleavage of 3′-ends (Sugawara et al. 1997; Pâques and Haber 1999). Consistent with this, in vitro biochemical studies have shown that purified Msh2-Msh3 binds specifically to double-strand/single-strand junctions and opens up the junction, possibly providing a more suitable substrate for Rad1-Rad10 cleavage (Surtees and Alani 2006). Recent work from Flott et al. (2007) has also implicated the Slx4 protein in Rad1-Rad10-dependent 3′ nonhomologous tail removal. The authors found that Slx4 forms a complex with Rad1-Rad10 that is mutually exclusive of the interaction with its endonuclease partner, Slx1. Slx4 was found to be required for Rad1-dependent DSB repair by single-strand annealing, presumably at the 3′ nonhomologous tail removal step (Flott et al. 2007).

A single unrepaired DSB is sufficient to trigger G₂/M cell cycle arrest in S. cerevisiae (Sandell and Zakian 1993). Arrest at the G₂/M transition can be elicited by the DNA damage or spindle checkpoints. While cell cycle checkpoints are not normally activated during mating-type switching, the DNA damage response is activated in strains lacking both donor sequences, which are thus unable to repair the DSB by gene conversion (Pellicioli et al. 1999, 2001; Lee et al. 2003). Activation of the DNA damage checkpoint has also been shown to occur during DSB repair at MAT when the donor locus is on a separate chromosome, most likely because the repair process takes longer to occur (Vaze et al. 2002). A role for the spindle checkpoint during mating-type switching has not been reported.

In this study, we used a variety of techniques to examine the importance of the Rad1-Rad10-Slx4 complex in 3′ nonhomologous tail removal during mating-type switching. We show that mutants defective in the Rad1-Rad10-Slx4 complex exhibited a RAD9-dependent, partially MAD2-dependent cell cycle arrest and decreased cell survival during mating-type switching. A third of rad1Δ and slx4Δ cells induced for mating-type switching showed a unique viability profile during pedigree analysis, with one switched and one dead daughter cell arising from the same DSB-induced cell. We hypothesize that this phenotype arises from replication bypass of an inefficiently repaired DNA lesion at MAT. This work indicates that the Rad1-Rad10-Slx4 complex promotes the efficient repair of DSBs involving a single 3′ nonhomologous tail intermediate.

MATERIALS AND METHODS

Strains and plasmids:

All strains used in this study are shown in Table 1. Parental strains EAY745 (MATa to MATα), EAY 744 (MATa to MATa), and EAY742 (donorless) were created by single-step gene replacement with SphI- and PvuII-digested pEAI118 to integrate MSH2-HA₄∷LEU2 at the endogenous MSH2 locus in JKM161, JKM160, or JKM139, respectively, and were kindly provided by J. Haber. Insertion of the HA₄ epitope into Msh2 did not disrupt gene function (Goldfarb and Alani 2004). All strains contain an HO endonuclease gene under control of the galactose-inducible GAL10 promoter to allow for inducible mating-type switching. To create the parental strain EAY1042 used in the double nonhomology experiments (Figure 1; supplemental Figure 1; Table 1), EAY745 was transformed with a PCR-generated fragment containing 57 bp of Ya sequence proximal to the MAT HO cut site, 1428 bp of KANMX sequence, and 52 bp of sequence distal to the HO cut site. Integration of the KANMX-containing fragment (MATa∷KANMX4) was confirmed by both PCR and Southern blot analysis. Yeast were transformed with the appropriate DNA fragments using the lithium acetate method (Gietz and Schiestl 1991), and integrations were confirmed by PCR followed by phenotype testing.

TABLE 1.

Strains used in this study

Name	Genotype	Strain notes and deletion constructs
MATa to MATα
EAY745	Wild type	Derived from JKM161 (Δho, HMLα, MATa, Δhmr∷ADE1, ade1-100, leu2-3,112, lys5, trp1∷hisG, ura3-52, ade3∷GAL10∷HO)
EAY853	rad1Δ	pWS1510 (rad1Δ∷URA3; E. Friedberg)
EAY969	msh2Δ	pEAI99 (msh2Δ∷TRP1; this lab)
EAY854	msh3Δ	pEAI88 (msh3Δ∷hisG-URA3-hisG; this lab)
EAY2087	slx4Δ	slx4Δ∷KANMX
EAY1788	pol3-01	YIpAM26 (pol3-01∷URA3; from A. Sugino)
EAY1332	mus81Δ	mus81Δ∷KANMX
EAY1730	rad9Δ	rad9Δ∷KANMX
EAY1968	mad2Δ	mad2Δ∷KANMX
EAY1562	mms2Δ	mms2Δ∷KANMX
EAY1778	mph1Δ	mph1Δ∷KANMX
EAY2125	rad1Δrad10Δ	rad10Δ∷KANMX; see above
EAY2090	rad1Δslx4Δ	See above
EAY1803	rad1Δ pol3-01	See above
EAY797	rad1Δmus81Δ	See above
EAY1726	rad1Δrad9Δ	See above
EAY1973	rad1Δ mad2Δ	See above
EAY1725	rad1Δmms2Δ	See above
EAY1776	rad1Δ mph1Δ	See above
MATa to MATa
EAY744	Wild type	Derived from JKM160 (Δho, Δhml∷ADE1, MATa, HMRa, ade1-100, leu2-3,112, lys5, trp1∷hisG, ura3-52, ade3∷GAL10∷HO)
EAY1356	rad1Δ	pWS1510 (rad1Δ∷URA3; E. Friedberg)
EAY2084	slx4Δ	slx4Δ∷KANMX
MATa∷KANMX to MATα
EAY1042	Wild type	Derived from JKM161; see materialsandmethods
EAY1115	rad1Δ	pWS1510 (rad1Δ∷URA3; E. Friedberg)
EAY1040	msh2Δ	pEAI99 (msh2Δ∷TRP1)
EAY1118	msh3Δ	pEAI88 (msh3Δ∷hisG-URA3-hisG)
EAY1407	rad51Δ	pJH683 (rad51Δ∷URA3; from J. Haber)
donorless
EAY742	Wild type	Derived from JKM139 (Δho, Δhml∷ADE1, MATa, Δhmr∷ADE1, ade1-100, leu2-3,112, lys5, trp1∷hisG, ura3-52, ade3∷GAL10∷HO)

All strains used in this study are derived from JKM161, JKM160, and JKM139, kindly provided by J. Haber. Gene disruptions and mutant alleles were made by transforming S. cerevisiae strains with restriction-digested plasmids for rad1Δ, msh2Δ, msh3Δ, rad51Δ, and pol3-01 as listed above. All other disruptions were made by integrative transformation of PCR products generated by amplification of KANMX sequences from either plasmid pFA6-KanMX4 (Wach et al. 1994) or genomic DNA from the Saccharomyces Genome Deletion Project knockout strains (http://www-sequence.stanford.edu/group/yeast_deletion_project/deletions3.html). See materials and methods for details.

Media and culture conditions:

For time-course experiments, dilutions of stationary phase cultures were made in yeast–peptone (Difco) medium, pH 6.8, containing 2% (w/v) lactate and grown at 30° until mid-log phase (1–2 × 10⁷ cells/ml). Cultures were induced with galactose (U.S. Biological) to 2% (w/v) final concentration and samples were collected at relevant time points. HO expression was suppressed after 30 min by the addition of glucose (U.S. Biological) to 2% (w/v) final concentration. To maintain a consistent number of cells at each time point throughout the time course, individual samples were diluted to the same cell density as the time zero sample.

Cell survival assays:

Asynchronous cultures were grown to mid-log phase and induced with galactose for 30 min. Uninduced controls were diluted similarly with water. Both induced and uninduced cultures were diluted 2500-fold and plated in triplicate on YPD plates immediately following the addition of glucose to the media. After growth for 3 days at 30°, the percentage survival was calculated as the number of colonies arising from induced relative to uninduced cultures. At least four independent cultures were used for each strain (Table 2). Results are shown as the mean ± SEM and were statistically analyzed using an unpaired two-tailed Student's t-test (http://www.physics.csbsju.edu/stats/t-test_bulk_form.html; see results).

TABLE 2.

Viability and mating-type switching efficiency of wild-type and mutant strains

	% survival	% switched
One nonhomologous end (MATa to MATα)
Wild type	76.1 ± 3.1	85.7 ± 2.7
rad1Δ	58.6 ± 2.1*	70.6 ± 2.8*
slx4Δ	57.1 ± 2.5*	67.6 ± 3.9*
rad1Δ slx4Δ	54.4 ± 2.2*	69.3 ± 4.3*
msh2Δ	67.6 ± 4.0	80.6 ± 2.5
msh3Δ	67.9 ± 2.0	81.5 ± 3.0
mus81Δ	72.6 ± 1.7	88.7 ± 5.9
pol3-01	65.1 ± 5.1	83.2 ± 3.4
rad1Δ mus81Δ	56.1 ± 3.8*	74.5 ± 7.5
rad1Δ pol3-01	55.6 ± 5.2*	70.6 ± 3.8*
rad9Δ	75.1 ± 3.1	82.5 ± 2.6
rad1Δ rad9Δ	56.3 ± 2.7*	74.7 ± 3.3
Two homologous ends (MATa to MATa)
Wild type	99.7 ± 2.2	NA
rad1Δ	95.3 ± 4.5	NA
slx4Δ	93.8 ± 4.3	NA
Two nonhomologous ends (MATa∷KANMX to MATα)
Wild type	62.7 ± 3.7	72.8 ± 7.4
rad1Δ	26.8 ± 1.2*	0.8 ± 0.8*
msh2Δ	23.5 ± 2.9*	9.8 ± 2.3*
msh3Δ	29.9 ± 1.7*	10.0 ± 3.1*
rad51Δ	25.2 ± 2.7*	0.0 ± 0.0*

Percentage cell survival (induced/uninduced) was determined by examining the viability of cells plated after a 30-min induction of HO expression. Surviving cells were assayed to determine the percentage that had switched mating type from MATa to MATα as described in materials and methods. Data are presented as the mean ± SEM of at least four independent experiments. Asterisks indicate values significantly different from wild-type with P < 0.01, Students t-test. For one nonhomologous end (standard mating-type switching), nonhomologous sequence (Ya) is present on only the proximal side of the DSB. For two homologous ends, strains were induced for MATa to MATa switching; thus, “% switched” is not applicable (NA). For two nonhomologous ends, the indicated strains contain nonhomologous sequences on both sides of the DSB due to insertion of KANMX on the distal side of the break (see Figure 1B).

Mating-type switching assay:

To determine mating types, individual colonies from cell survival assays (20–40/replicate) were crossed with arg4 MATa and MATα tester strains (EAY759 and EAY760; from N. Sugawara, Haber laboratory) and replica plated onto synthetic complete plates (Rose et al. 1990) lacking both arginine and lysine to select for diploids. The percentage of switched cells was determined for each cell survival experiment and is shown in Table 2 as the mean ± SEM.

Southern blot analysis:

Chromosomal DNA was isolated during time-course experiments as described (Holmes and Haber 1999a; Goldfarb and Alani 2004) following a 30-min galactose induction. DNA was then digested with StyI (New England Biolabs) for single nonhomology strains or with AvaII, BanI, and BlpI (New England Biolabs) for double nonhomology strains and electrophoresed on 1% TAE–agarose gels with 1× TAE buffer. Southern blot transfer and hybridizations were performed essentially as described by the manufacturer (Amersham) using the Church and Gilbert method (1984).

All probes used for Southern blot analysis were amplified by PCR using EAY745 yeast genomic DNA and ³²P-labeled using the NEBlot kit (New England Biolabs) according to the manufacturer's description. To probe MAT-specific bands, we radiolabeled a 638-bp PCR product beginning 67 bp downstream of the MAT Z2 region using pJH364 forward and reverse primers (5′-ACGAATTGGCTATACGGGAC and 5′-GTCCAATCTGTGCACAATGAAG, respectively, from the Haber lab). Efficient DSB formation was detected 30 min after galactose induction by Southern blot analysis (Figure 2). To visualize mating-type switching in double nonhomology strains, probes were produced from a 277-bp PCR product amplified using primers AO585 (5′-CTTAGCATCATTCTTTGTTCTTAT) and AO586 (5′-CAAGAAGGCGAATAAGATAAAGA). Loading control probes for blots of the double nonhomology strains were created by amplifying a 235-bp PCR product with primers AO583 (5′-CTCGTATTGGAGAAATAAGTTTTCGT) and AO584 (5′-GGTAGAGTCTTATTGGCAAGATAG) (supplemental Figure 1). Ya-specific probes (supplemental Figure 2) were created by labeling a 539-bp PCR fragment made using primers AO1425 (5′-GGACAACATGGATGATATTTGTAGTATGGCGG) and AO1049 (5′-CTGTTGCGGAAAGCTGAAAC), both located within Ya. Blots were visualized using the Phosphor Imaging system and quantified using the ImageQuant program (Molecular Dynamics). Quantification of repair efficiency in Figure 2C was done as described previously (Wang et al. 2004), with product bands set relative to the first HO cut band and normalized relative to the MAT distal band in each lane. Ya loss was quantified by setting the Ya proximal band in each lane relative to the value at t = 0 (supplemental Figure 2).

Figure 2.—

Southern blot analysis of mating-type switching in wild-type and rad1Δ strains: (A) Diagram of the MAT locus showing the restriction sites used for Southern blot analysis, expected fragment lengths, and location of the probes used for detection of mating-type switching. (B) Analysis of digested DNA for wild-type and rad1Δ mutants induced for mating-type switching. Experiments were performed at least three times, with representative time courses shown. (C) Quantification of repair efficiency as described in materials and methods.

Chromatin immunoprecipitation:

Samples from time-course experiments were chromatin immunoprecipitated as described previously (Goldfarb and Alani 2004). Msh2-HA₄ was immunoprecipitated from yeast cell extracts using the 12CA5 monoclonal antibody, and expression of Msh2p-HA₄ was confirmed by Western blot (Goldfarb and Alani 2004, 2005). All strains used in the chromatin immunoprecipitation (ChIP) experiments contain a deletion of the HMRa donor so that the MATa locus could be specifically amplified by PCR. PCR reactions, electrophoresis conditions, and quantification were similar to those described in Evans et al. (2000), but with different primer sets. To detect sequences proximal to the DSB, a 267-bp fragment containing the Ya sequence was amplified from immunoprecipitated and input chromosomal DNA using AO1048 (5′-TCACCCCAAGCACGGGCATT) and AO1049 (5′-CTGTTGCGGAAAGCTGAAAC), which are adjacent to the HO recognition site (Figure 3). Samples were run on 1.5% TAE–agarose gels and bands were quantified relative to the maximal signal using Scion Image (Scion). Since the input signal decreases during mating-type switching as the Ya sequence is removed, the data are presented as the amount of chromatin-immunoprecipitated Ya PCR product detected after HO induction relative to that at t = 0. A 163-bp CRY1 control band was also amplified from the chromosomal input DNA using primers AO1106 (5′-CGCCAGAGTTACTGGTGGTATGAAGG) and AO1107 (5′-GGAGTCTTGGTTCTAGTACCACCGG). The PCR signal was quantified within the linear range of detection, and ChIP was specific to both the epitope tag and formaldehyde crosslinking (Goldfarb and Alani 2004).

Figure 3.—

Msh2 localization to the DSB in wild-type, rad1Δ, and donorless mutants. (A) Location of primers used for semiquantitative PCR following Msh2 chromatin immunoprecipitation. (B) Representative chromatin immunoprecipitation and PCR detection of Msh2 localization to MAT during mating-type switching in wild-type, rad1Δ, and donorless mutants. Since the Ya sequence is removed during mating-type switching, the input signal is also shown using primers to an unrelated locus (CRY1). (C) For each time point, the Msh2 ChIP signal was set relative to the t = 0 signal, with the maximum signal for each time course set as 1.0 to compare the relative the timing of Msh2 localization. Each data point represents the mean of three to four experiments ± SEM.

FACS analysis:

Cells were collected at various times after HO induction as described above. Aliquots of cells were pelleted at the relevant time points, fixed in 70% ethanol, and stored at 4° for up to 7 days. Cell samples were resuspended in 50 mm NaCitrate, pH 7.4, sonicated briefly, and treated for 1 hr with RNase A at 37°, followed by a 1-hr treatment with Proteinase K at 37°. DNA was stained with 1 nm final concentration of Sytox Green (Invitrogen), and samples were analyzed at the Cornell University Biomedical Sciences Flow Cytometry Core Laboratory (Ithaca, NY). Percentage of cells in G₁, S, or G₂/M phases was determined by gating according to 1n and 2*1n DNA content. A representative FACS profile for wild-type cells at t = 0 is shown in Figure 4B, with vertical gates for G₁, S, and G₂/M phases. Values shown in Figure 4A reflect the mean of three or more samples per time point ± SEM.

Figure 4.—

FACS analysis of cells undergoing mating-type switching. (A) Bar graphs of the percentage of cells in G₁, S, or G₂/M phases of the cell cycle at 0, 2, 4, and 6 hr following induction of mating-type switching in wild-type, rad1Δ, and donorless strains (average of at least three experiments ± SEM). See materials and methods for details. The increase in the percentage of G₂/M cells in rad1Δ mutants relative to wild type is statistically significant (P < 0.01 at t = 2 and t = 4 hr, Student's t-test). (B) Representative FACS profile for wild-type cells at t = 0, with vertical gates separating the 1n (G₁) and 2n (G₂/M) DNA content.

Pedigree analysis:

Cells were induced for HO cleavage at MAT as described above. Following addition of glucose to the medium at t = 0.5 hr, 15 μl of culture was dropped down the center of a YPD plate and single, unbudded cells were separated at 1-cm intervals under the light microscope using a microdissection needle. Cells were visualized beginning at t = 0.5 hr, incubated at 30° between manipulations, and monitored every 20–30 min until daughter cells were able to be separated from each other (t = ∼4–10 hr). Cells that did not complete cell division within 10 hr were not scored. The length of time required for completion of cell division is reported in Table 4 as the mean of all cells in each category ± SEM. Plates were incubated for 3 days at 30°, and colonies were tested for mating type as described above. Cells were categorized by viability and mating type as shown in Table 3. Pairs of daughter cells scored as both unswitched were not included because we cannot rule out the failure to form a DSB in these cells. The number of cells present in dead cell clusters was also recorded and is visualized in supplemental Figure 3B. Photographs of representative cells (supplemental Figure 3) were taken under the light microscope using a Fuji FinePix S5000 digital camera.

TABLE 4.

Average length of cell cycle in cells undergoing mating-type switching

	Average time required for cell division (hr)
	Uninduced	Two switched	One switched, one dead	Two dead
Wild type	4.5 ± 0.2	4.5 ± 0.1	NA	NA
donorless	4.9 ± 0.2	NA	NA	10 ± 0.2*
rad1Δ	4.7 ± 0.2	7.0 ± 0.3*	8.4 ± 0.3*	8.2 ± 0.4*
slx4Δ	4.7 ± 0.1	5.8 ± 0.2*	7.0 ± 0.4*	8.6 ± 0.4*
rad1Δ slx4Δ	4.5 ± 0.1	5.5 ± 0.4*	6.8 ± 0.4*	9.0 ± 0.4*
msh2Δ	4.3 ± 0.1	5.6 ± 0.1*	NA	6.5 ± 0.3*
rad9Δ	4.9 ± 0.1	4.5 ± 0.1	4.8 ± 0.2	NA
mad2Δ	4.9 ± 0.2	4.9 ± 0.2	4.2 ± 0.3	NA
rad1Δ rad9Δ	5.2 ± 0.1*	4.8 ± 0.2	4.8 ± 0.1	5.3 ± 0.1
rad1Δ mad2Δ	5.5 ± 0.2*	5.6 ± 0.3*	6.7 ± 0.3*	6.3 ± 0.2*
mms2Δ	4.5 ± 0.1	5.1 ± 0.1*	NA	6.1 ± 0.5*
mph1Δ	4.5 ± 0.1	4.7 ± 0.1	5.9 ± 0.6*	6.5 ± 0.8*
rad1Δ mms2Δ	4.5 ± 0.1	5.8 ± 0.3*	6.4 ± 0.1*	6.3 ± 0.1*
rad1Δ mph1Δ	5.3 ± 0.1*	6.2 ± 0.3*	8.1 ± 0.3*	8.3 ± 0.3*

Cell cycle duration was determined during pedigree analysis (see Table 3 and materials and methods) and is shown as the mean length of time required for division (hr) ± SEM. “NA” denotes categories containing <10% of cells, and thus cell cycle lengths are not reported. Asterisks denote statistical significance from wild type with P < 0.01; values were compared to wild type “uninduced” or “two switched” as appropriate.

TABLE 3.

Pedigree analysis of wild type and mutants induced for mating-type switching (MATa to MATα) or mock induced

	% two switched	% one switched, one dead	% two dead	% other	N
A. Induced
Wild type	96	4	0	0	81
donorless	0	0	96	4	51
rad1Δ	38	32	28	2	97
slx4Δ	51	33	12	4	49
rad1Δ slx4Δ	29	44	19	8	59
msh2Δ	78	7	11	4	81
rad9Δ	82	12	6	0	90
mad2Δ	84	10	4	2	49
rad1Δ rad9Δ	20	41	39	0	102
rad1Δ mad2Δ	33	43	15	8	60
mms2Δ	79	7	12	2	58
mph1Δ	69	16	10	5	70
rad1Δ mms2Δ	19	54	26	1	99
rad1Δ mph1Δ	28	38	30	3	81

	% two alive	% one alive, one dead	% two dead	N
B. Mock induced
Wild type	99	0	1	91
donorless	100	0	0	41
rad1Δ	98	2	0	119
slx4Δ	94	3	3	36
rad1Δ slx4Δ	94	5	2	62
msh2Δ	84	9	7	76
rad9Δ	98	0	2	51
mad2Δ	100	0	0	26
rad1Δ rad9Δ	89	3	8	72
rad1Δ mad2Δ	91	2	8	53
mms2Δ	90	7	3	71
mph1Δ	100	0	0	52
rad1Δ mms2Δ	87	9	4	92
rad1Δ mph1Δ	96	3	1	75

Following induction of mating-type switching in liquid culture, single cells were separated on YPD medium under the light microscope using a microdissection needle. Cells were monitored at regular intervals, and daughter cells were separated after completion of the first cell division. Cells that formed colonies were tested for mating type as described in materials and methods. The number of cells (N) tested for each strain is shown. For cells induced for mating-type switching, “% other” includes pairs of daughter cells scored as “one unswitched and one dead,” or “one unswitched and one switched.” Pairs of daughter cells scored as “both unswitched” were not included because we cannot rule out the failure to form a DSB in these cells. For the uninduced experiments, viability is shown for cells mock induced with water.

RESULTS

Decreased mating-type switching in the absence of the Rad1-Rad10-Slx4 complex:

Rad1-Rad10 and Msh2-Msh3 are proposed to act during mating-type switching in steps involving the removal of a single 3′ nonhomologous tail on the non-invading strand as depicted in Figure 1 (Holmes and Haber 1999b; Pâques and Haber 1999). Previous work examining the role of Rad1-Rad10 during gene conversion primarily utilized plasmid-based assays in which DNA sequence on one or both sides of a DSB site contained nonhomologous sequence with respect to a donor sequence, also present on the plasmid (Sugawara et al. 1997; Colaiácovo et al. 1999). To examine the coordination of repair and checkpoint signaling factors during gene conversion on the chromosome, we analyzed roles for Rad1-Rad10-Slx4 and Msh2-Msh3 in mating-type switching in S. cerevisiae, which is hypothesized to involve removal of a single 3′ nonhomologous tail following the annealing step of SDSA (Figure 1A; Haber 1998; Pâques and Haber 1999; Ira et al. 2006).

Mating-type switching was induced in MATa strains expressing HO endonuclease from the galactose-inducible GAL10 promoter (materials and methods). As shown in Table 2, cell viability following DSB induction was high in wild type (76% ± 3%) but reduced in rad1Δ (59% ± 2%; P < 0.01, Student's t-test) and rad1Δ rad10Δ double mutants (63% ± 1%; data not shown). The decrease in cell viability was specific to strains induced for MATa to MATα switching; no significant decrease was observed in strains induced for completely homologous switching (MATa to MATa) that does not involve 3′ nonhomologous tails (Table 2). In addition, the percentage of surviving cells that had switched mating type was reduced in rad1Δ strains relative to wild type (71% ± 3% vs. 86% ± 3%; P < 0.01). This decrease in gene conversion may be due to an increase in repair of the break by nonhomologous end joining to yield MATa cells or could be indicative of aberrant repair or more disruptive nonhomologous end joining that disrupts the MAT locus and yields an “a-like faker” phenotype, since cells lacking a functional MAT locus phenocopy MATa by default (Strathern et al. 1981).

Recently, Flott et al. (2007) reported that the Slx4 protein forms a complex with Rad1-Rad10 and is critical for its 3′ nonhomologous tail removal activity during repair by SSA. As predicted from this work, Slx4 also functions with Rad1-Rad10 in mating-type switching. slx4Δ and rad1Δ slx4Δ mutants showed viability (57% ± 3% and 54% ± 2%, respectively) and switching phenotypes (68% ± 4% and 69% ± 4%, respectively) similar to rad1Δ strains (Table 2). In contrast, msh2Δ and msh3Δ strains displayed only a subtle decrease in viability (68%), and the percentage of switched cells was similar to wild type (Table 2). Thus, Msh2-Msh3 appears nearly dispensable for nonhomologous tail removal during mating-type switching, where the 3′ tail is on the non-invading strand.

We hypothesized that the gene conversion observed in the absence of Rad1-Rad10-Slx4 could be facilitated by the action of redundant nucleases that remove the 3′ Ya nonhomologous tail. However, disruption of Mus81-Mms4 or the polymerase δ 3′–5′ proofreading activity did not have a significant effect on the viability of rad1Δ mutants following mating-type switching (Table 2). Since mating-type switching can occur in rad1Δ mutants, albeit less efficiently, it is likely that unknown nucleases or multiple redundant nucleases are able to remove 3′ nonhomologous tails when Rad1-Rad10-Slx4 is absent. A recent study identified Saw1, a protein that interacts with Rad1-Rad10 and is thought to recruit Rad1-Rad10 to recombination intermediates (Li et al. 2008). It is possible that Saw1 may recruit other nucleases as well, allowing for completion of mating-type switching in the absence of Rad1-Rad10-Slx4.

Southern blot analysis was used to examine product formation in wild-type and rad1Δ strains during mating-type switching (Figure 2). Efficient DSB formation was observed at the MAT locus within 30 min of induction in all strains and products were detectable by 1 hr post-induction in wild type, consistent with previous studies (White and Haber 1990; Colaiácovo et al. 1999). rad1Δ mutants displayed a ∼10% reduction in product formation relative to wild type. This result is much more subtle than that seen in an analysis of MATα-to-MATa switching in G₁-arrested rad1Δ cells (Holmes and Haber 1999b), but is consistent with the viability data presented above. The defects exhibited by mutants lacking Rad1-Rad10-Slx4 are more apparent in the pedigree, FACS, and chromatin immunoprecipitation studies described below and may indicate that, while MATα product formation appears to be only mildly reduced in rad1Δ mutants, the gene conversion at MAT might be associated with BIR, aberrant recombination, or disrupted signaling.

Previous studies have shown a strict requirement for both Rad1-Rad10 and Msh2-Msh3 when both sides of a DSB contain nonhomologous sequence (Sugawara et al. 1997; Colaiácovo et al. 1999). In repair of such breaks, a 3′ nonhomologous tail must be removed during the strand invasion step for repair DNA synthesis to proceed, in addition to 3′ nonhomologous removal at the later synthesis-dependent annealing step (Figure 1B). To confirm that Rad1-Rad10 and Msh2-Msh3 are required for removing 3′ nonhomologous tails on the invading strand during chromosomal mating-type switching, we inserted the KANMX sequence on the distal side of the HO cut site at the MAT locus (Figure 1B). In wild-type strains containing the KANMX insertion, gene conversion was delayed but completed with little loss of viability (Table 2; supplemental Figure 1). Consistent with previous studies, we found that both Rad1-Rad10 and Msh2-Msh3 complexes were required for gene conversion involving two 3′ nonhomologous tails. No gene conversion product was detected by Southern blot in rad1Δ and msh3Δ mutants (supplemental Figure 1), and the viability and switching efficiency of rad1Δ mutants was comparable to that of rad51Δ mutants completely defective in gene conversion, as shown in Table 2 (Sugawara et al. 1995). While the viability of msh2Δ and msh3Δ strains was equivalent to that of rad1Δ mutants, both msh mutants exhibited a greater percentage of switched cells (10%; Table 2), consistent with the idea that Msh2-Msh3 plays a supporting role that may be less critical than the role of Rad1-Rad10. The residual viability in rad1Δ, msh2Δ, msh3Δ, and rad51Δ strains is likely due to nonhomologous end joining, as seen in strains completely lacking donor sequences (Moore and Haber 1996).

Prolonged Msh2 localization to the DSB in rad1Δ mutants:

Because Rad1-Rad10 is predicted to remove 3′ nonhomologous tails on the non-invading strand following the annealing step (Figure 1), we reasoned that rad1Δ mutants would exhibit a delay in removal of the 3′ Ya sequence. Using Ya-specific probes, we were unable to detect a difference in the loss of Ya between wild-type and rad1Δ strains (supplemental Figure 2). Detection of any delay is confounded by the fact that the initial resection of the break should lead to loss of the 5′ strand of Ya with similar kinetics in both strains. Thus, we additionally performed chromatin immunoprecipitation using HA-tagged Msh2, α-HA antibody, and PCR primers located within the Ya sequence as described previously (Goldfarb and Alani 2004). Our lab previously showed that the Msh2 protein localizes rapidly to DSBs (Evans et al. 2000).

As shown in Figure 3, Msh2 localized immediately to the MAT locus following DSB formation, peaked at 1 hr post-induction, and then decreased, consistent with the kinetics of product formation shown in Figure 2 and a role for Msh2-Msh3 in DSB repair. A similar pattern was seen using primers specific to the X-Ya junction; however, peak levels were achieved at a slightly later time point (1.5 hr; data not shown). While the input signal is lost over time due to conversion to MATα, the input signal at the unrelated CRY1 locus was constant throughout the time course.

In rad1Δ mutants, Msh2 localized to MAT following DSB formation, but in contrast to wild type, Msh2 remained near the break for ∼3 hr (Figure 3). We observed a similar Msh2 localization pattern for donorless mutants unable to complete mating-type switching, where the 3′-ends are thought to be stable despite a complete inability to perform homologous repair (Vaze et al. 2002; Aylon et al. 2003). Msh2 localization was also prolonged at the X-Ya junction in rad1Δ compared to wild type (data not shown). Thus, while we were unable to detect a delay in loss of the Ya sequence in rad1Δ mutants, the prolonged presence of Msh2 at the break during mating-type switching suggests that at least a subset of rad1Δ mutants contain recombination intermediates at later time points.

rad1Δ mutants induced for mating-type switching exhibit G₂/M arrest:

The above observations encouraged us to examine the cell cycle progression of rad1Δ mutants during mating-type switching. Mutants lacking both donor sequences have previously been shown to exhibit a prolonged G₂/M cell cycle delay due to an inability to repair the DSB by homologous recombination (Toczyski et al. 1997; Lee et al. 1998). We used FACS analysis to measure the DNA content of wild-type, rad1Δ, and donorless mutants following DSB induction. As shown in Figure 4, wild-type strains showed little variation in the percentage of cells in G₁, S, or G₂/M phase during the course of mating-type switching. Consistent with the known arrest phenotype, the majority of cells from a strain lacking both HMLα and HMRa sequences (donorless) were present in G₂/M phase at 4 hr (83.5% ± 1.5%) and 6 hr post-induction (71.8% ± 3.9%). rad1Δ strains showed a significant increase in the percentage of G₂/M cells at 2 hr (58.3% ± 4.0%) and 4 hr post-induction (65.1% ± 1.6%) relative to wild type (41.9% ± 1.5% and 30.1% ± 3.3%, respectively; P < 0.01), but returned to wild-type levels by 6 hr, suggesting that the absence of Rad1-Rad10 leads to a G₂/M arrest that is both shorter and earlier than observed in donorless mutants. This is consistent with gene conversion occurring in rad1Δ mutants, although inefficiently, in contrast to donorless mutants, which can survive only by nonhomologous end joining (Moore and Haber 1996).

Mutants lacking Rad1-Rad10-Slx4 show unique viability profiles in pedigree analysis following mating-type switching:

To further analyze the viability and cell cycle phenotypes seen in rad1Δ and slx4Δ mutants during mating-type switching, we performed pedigree experiments in which single, unbudded (G₁) cells were isolated after DSB formation and monitored through the cell cycle. Daughter cells were separated following the first cell division (materials and methods). Cells that grew into colonies were subsequently assayed for mating type. As shown in Table 3, 96% of wild-type cells yielded two viable daughter cells that had both switched mating type. In contrast, only 38% of rad1Δ mutants formed two switched colonies and, of the remaining cells, 32% formed one switched colony and one dead cell cluster and 28% formed two dead cell clusters. slx4Δ and rad1Δ slx4Δ strains exhibited phenotypes similar to rad1Δ mutants (Table 3). No such decrease in viability was seen in these strains in the absence of the DSB, nor in rad1Δ strains induced for completely homologous MATa to MATa switching (Table 3B; data not shown). The “one switched, one dead” category is particularly intriguing, since repair and death arise from the same induced cell, and it is unique to cells undergoing gene conversion. Thus, the effect of the rad1Δ and slx4Δ mutations on mating-type switching is much more severe than was apparent in liquid culture assays, where asynchronous cells were induced for mating-type switching and the fate of daughter cells could not be assessed.

We also measured the length of the first cell division following DSB induction during the pedigree experiments. As shown in Table 4, completion of cell division was delayed by 3 hr in rad1Δ mutants compared to wild type (P < 0.01, Student's t-test), consistent with the FACS analysis presented above (Figure 4). Strains lacking donor sequences exhibited an even longer delay (10 ± 0.2 hr to complete division compared to 4.5 ± 0.1 hr in wild type; Table 4). During this extended period, rad1Δ and donorless cells displayed a large-budded morphology suggestive of G₂/M arrest (supplemental Figure 3A).

After completion of the first cell division, ∼45% of rad1Δ cells failed to form colonies in the pedigree analysis (the dead cells from both the “two dead” and “one switched, one dead” categories), but divided several times before forming dead cell clusters (average of 8 ± 1 cells; supplemental Figure 3B). This phenotype is consistent with the phenomenon of break adaptation, in which cells exit the cell cycle arrest despite the continued presence of unrepaired DNA and differs from the death seen in cells undergoing DSB repair that fail to exit a G₂/M arrest (Toczyski et al. 1997; Lee et al. 1998, 2003; Pellicioli et al. 2001). This adaptation phenotype is consistent with a significant proportion of rad1Δ cells induced for mating-type switching being unable to complete repair of the break. Inviable cells from donorless strains exhibited a more severe phenotype following checkpoint exit and died with one large-budded cell or two cells (adaptation for only one cycle) as documented previously (Lee et al. 1998), most likely due to the presence of more extensive DNA damage due to prolonged 5′–3′ resection.

Consistent with the cell survival assays described above, msh2Δ and msh3Δ mutations had little effect on viability during mating-type switching in pedigree experiments (Table 3, data not shown). Viability was reduced equally in both the induced and uninduced states, with ∼80% of msh2Δ cells forming two viable colonies, ∼10% forming one alive and one dead cell cluster, and ∼10% with two inviable cells (Table 3). Thus the absence of MSH2 confers a general decrease in viability that appears unrelated to the formation of an HO-induced DSB. A more subtle decrease in viability (5%) was observed for strains lacking SLX4 in the uninduced state.

G₂/M delay in rad1Δ mutants is dependent upon both the DNA damage response and the spindle checkpoint:

To test whether the cell division delay observed in rad1Δ mutants was mediated by the DNA damage checkpoint, we measured cell viability and cell cycle duration in rad1Δ mutants defective for the Rad9-dependent DNA damage response. rad1Δ rad9Δ double mutants exhibited cell cycle lengths comparable to wild-type and rad9Δ mutant cells (∼5 hr; Table 4), in contrast to ∼8 hr for rad1Δ mutants. Thus, the G₂/M cell cycle arrest exhibited by rad1Δ mutants is dependent upon RAD9, presumably via Rad9-mediated activation of the DNA damage response (Harrison and Haber 2006). Elimination of the arrest had very little effect on the viability of rad1Δ mutants (Tables 2 and 3), pointing to an inability of the DNA damage response to promote repair.

Slx4 forms a complex with Rad1-Rad10 that is critical for 3′ nonhomologous tail removal during repair by single-strand annealing (Flott et al. 2007). As shown in Table 3, slx4Δ and rad1Δ slx4Δ mutants exhibited significantly shorter cell cycle delays than rad1Δ single mutants (1 vs. 2 hr for “two switched” and 2 vs. 3.5 hr for “one switched, one dead”). It is not surprising that the absence of Slx4 reduces the delay, since Slx4 is a known target of the Mec1 and Tel1 checkpoint kinases, requires checkpoint-dependent phosphorylation for Rad1-dependent SSA, and has been shown to regulate checkpoint-dependent processes (Flott and Rouse 2005; Roberts et al. 2006; Flott et al. 2007). The fact that slx4Δ mutants exhibit rad1Δ-like phenotypes, but with shorter cell cycle delays, is additional evidence that Slx4 provides a link between the 3′-end-processing machinery and the DNA damage checkpoint.

Several studies have suggested a link between the DNA damage response and the spindle checkpoint (Aylon and Kupiec 2003; Kim and Burke 2008). We hypothesized that the cell death in rad1Δ mutants was due to aberrant repair involving gross chromosomal changes that might activate the spindle checkpoint and thus tested whether the G₂/M arrest in these mutants required MAD2. As shown in Tables 3 and 4, mad2Δ mutants induced for mating-type switching had only slightly decreased viability and displayed cell cycle lengths similar to wild type. However, rad1Δ mad2Δ double mutants exhibited reduced cell cycle delays relative to rad1Δ mutants (Table 4). Cells in the “two dead” and “one switched, one dead” pedigree categories took ∼6.5 hr to divide in rad1Δ mad2Δ mutants, compared to ∼8 hr in rad1Δ single mutants (P < 0.015). Interestingly, rad1Δ mutants that formed two switched colonies exhibited arrests that appeared fully MAD2 dependent, unlike the partially MAD2-dependent arrests described above (Table 4). These results suggest that, although gene conversion occurs without loss of viability for cells in the “two switched” class, repair is inefficient and disruptive to the assembly of the mitotic spindle. The variety of arrest phenotypes in rad1Δ mutants further distinguishes the pedigree viability categories from each other and suggests that different defects or modes of repair operate in these subsets of cells.

Unique viability pattern in pedigree analysis is consistent with replication-mediated repair:

Approximately one-third of rad1Δ mutant cells divided to form both one switched and one dead colony in the pedigree analysis (Table 3). We hypothesized that these cells may complete gene conversion by replicating partially repaired intermediates containing one intact switched strand and one unrepaired strand (Figure 5; Kang and Symington 2000). To test such a model, we examined whether post-replicative lesion bypass repair pathways were involved in completing gene conversion during mating-type switching. We focused on MMS2- and MPH1-dependent repair pathways, mutations in which cause defects in the error-free bypass pathways involving fork reversal and recombinational replication restart, respectively (Torres-Ramos et al. 2002; Schürer et al. 2004; Watts 2006). Both mms2Δ and mph1Δ mutants displayed decreased viability in the wild-type background during mating-type switching. As shown in Table 3, the percentage of cells in the “two switched” category for pedigree analysis was reduced to 79% in mms2Δ and 69% in mph1Δ mutants compared to 96% in wild type. mms2Δ mutants also displayed a slight reduction in viability in the absence of the DSB (Table 3), but viability was further decreased for cells induced for switching. Both of these mutants had an increased proportion of cells in both the “one switched, one dead” and “two dead” categories, indicating that replicative lesion bypass pathways play a role in the completion of gene conversion during mating-type switching.

Figure 5.—

Model for mating-type switching facilitated by DNA replication. We propose that mating-type switching can be mediated by ongoing DNA replication. DSB formation, 5′–3′ resection, strand invasion, synthesis, and repair of the invading strand occur as shown in Figure 1 and as predicted by SDSA models. The partially repaired recombination intermediate shown at the top containing a single-stranded break can then be acted on by either the DNA replication machinery or the Rad1-Rad10-Slx4 complex. In the presence of Rad1-Rad10-Slx4, the 3′ Ya nonhomologous tail is removed efficiently prior to, during, or following DNA replication, and once DNA replication has been completed, the cell can divide to produce two viable cells of the switched mating type. In the absence of Rad1-Rad10-Slx4, mating-type switching products are produced largely by replication of the partially repaired recombination intermediate to yield either one switched and one dead daughter cell or, after the action of inefficient nucleases, two viable switched daughters.

To test whether the error-free lesion bypass pathways are responsible for repair in the absence of Rad1-Rad10-Slx4, we analyzed both rad1Δ mms2Δ and rad1Δ mph1Δ double mutants in pedigree experiments. rad1Δ mms2Δ double mutants exhibited a decrease in the percentage of “two switched” cells from 38 to 19%, and this decrease was directly correlated with an increase in the “one switched, one dead” category; however, no change in the percentage of “two dead” cells was seen for either rad1Δ mms2Δ or rad1Δ mph1Δ relative to rad1Δ single mutants, suggesting that death of these cells occurs by a separate mechanism. The mph1Δ mutation appeared roughly epistatic to rad1Δ in this assay, with very little decrease in viability relative to rad1Δ.

The mms2Δ mutation reduced the length of cell cycle delay in rad1Δ mutants from 2 hr to 1 hr for the “two switched” cells and from 3.5 to 2 hr for the “one switched, one dead” and “two dead” categories (Table 4). rad1Δ mph1Δ double mutants exhibited a similar decrease in length of arrest for “two switched” cells, but not for dying cells. Thus, it is tempting to speculate that checkpoint signaling in rad1Δ mutants might be initiated or enhanced by the collision of a replication fork with recombination intermediates (see discussion). Together, the above phenotypes suggest a role for post-replicative lesion bypass repair in the completion of gene conversion during mating-type switching.

DISCUSSION

In this study, we investigated the requirements for the Rad1-Rad10, Slx4, and Msh2-Msh3 factors in 3′ nonhomologous tail removal during gene conversion at the chromosomal MAT locus. Rad1-Rad10 and Msh2-Msh3 have been proposed to act during mating-type switching in steps involving the removal of a single 3′ nonhomologous tail on the non-invading strand, primarily on the basis of their roles in 3′ nonhomologous tail removal during single-strand annealing and in plasmid-based assays (Haber 1998; Pâques and Haber 1999). As described above, mating-type switching in rad1Δ mutants led to a checkpoint-dependent G₂/M cell cycle delay and decreased viability. In the absence of functional Rad1-Rad10-Slx4, cells displayed a unique viability profile consistent with a model in which gene conversion can be facilitated by replication of partially repaired recombination intermediates. Msh2-Msh3, however, played only a subtle role in such repair, in contrast to its critical role in DSB repair involving two 3′ nonhomologous tails.

Previous work in the Symington lab proposed that replication of partially repaired recombination intermediates might bypass the requirement for Rad1-Rad10-dependent 3′ nonhomologous tail removal in a plasmid retention assay (Kang and Symington 2000). We extend this model to explain the unique viability pattern observed in rad1Δ and slx4Δ mutants in pedigree experiments, where one-third of cells divide to produce both repaired (switched) and dead daughter cells (Table 3, Figure 5). In this model, mutants lacking Rad1-Rad10-Slx4 initiate repair normally, but encounter difficulty after annealing of the repaired invading strand back to the MAT locus. In the absence of 3′ nonhomologous tail removal activity, the remaining broken strand is unable to prime repair DNA synthesis to complete gene conversion. If, instead, DNA replication occurs prior to 3′ nonhomologous tail removal, template switching could produce both an intact chromosome of the switched mating type and a broken chromosome. Segregation of these chromosomes to daughter cells could then lead to the “one switched, one dead” phenotype (Table 3), whereas repair of the broken chromosome by an inefficient nuclease could yield two viable, switched cells as seen in wild type (Figure 5). In further support of replication-mediated repair, another study found that mating-type switching in G₁-arrested cells led to a much more severe reduction in product formation (37% product formation at 5 hr) in rad1Δ mutants than is seen in this study in cycling cells, with product formation in rad1Δ mutants reduced to only ∼90% of wild type at 4 hr (Holmes and Haber 1999b). Additional evidence for this model of replication-mediated recombination is discussed below.

We show that mating-type switching in mutants lacking Rad1-Rad10 or Slx4 induces a G₂/M cell cycle delay involving both the DNA damage and spindle checkpoints (Table 4). Interestingly, the arrest phenotypes correlated with the viability phenotypes observed by pedigree analysis. Those cells that produced two viable, switched daughter cells exhibited shorter cell cycle delays (∼2 hr) that were completely dependent upon the spindle checkpoint, whereas cells that produced two dead cell clusters or one switched colony and one dead cell cluster exhibited longer arrests (∼3.5 hr) and were only partially dependent on the spindle checkpoint (Table 4).

Several studies have indicated potential links between the DNA damage and spindle checkpoints (Garber and Rine 2002; Kim and Burke 2008). It is not surprising that DNA damage that triggers the damage checkpoint might also impede the correct attachment and formation of tension between the chromosomes and the mitotic spindle. It was recently demonstrated that the spindle assembly checkpoint arrests cells in response to MMS-induced DNA damage in a Mec1- and Tel1-dependent manner, independent of a functional kinetochore (Kim and Burke 2008). We show here that, in response to a single DSB at the MAT locus, the DNA damage response factor Rad9 appears to be required for both the DNA damage and spindle checkpoints, as rad1Δ rad9Δ mutants exhibit no G₂/M arrest and rad1Δ mad2Δ mutants exhibit shorter arrests than rad1Δ single mutants (1 vs. ∼3.5 hr; Table 4). In contrast to other studies, we do not see residual G₂/M arrest in rad9Δ mutants in these experiments (Aylon and Kupiec 2003; Kim and Burke 2008).

Previous work has shown that the length of G₂/M arrest in response to DNA damage correlates with the amount of single-stranded DNA present (Lee et al. 1998). Our results are consistent with this, as rad1Δ mutants exhibit shorter arrests relative to donorless strains; rad1Δ mutants are able to initiate strand invasion, whereas donorless mutants accumulate ssDNA because they cannot initiate repair (Lee et al. 1998). We also observed distinct adaptation phenotypes in the rad1Δ and donorless strains. donorless strains adapted for one cell cycle only and died at the next G₂/M transition, whereas rad1Δ mutants exhibited a classical break adaptation phenotype and died as eight-cell clusters (supplemental Figure 3). Since dying rad1Δ mutants exhibit a cell cycle delay followed by adaptation, it is possible that repair in this subset of the population occurs by BIR or by crossing over. BIR initiated from the MAT locus by strand invasion into HMLα would lead to loss of half of chromosome III, including the centromere, and crossing over would similarly create an intrachromosomal deletion. Such repair would be expected to be associated with delayed product formation as well as with a DNA damage checkpoint- and spindle checkpoint-dependent G₂/M arrest, as seen in our pedigree analysis (McEachern and Haber 2006).

The fact that Msh2 localization to the MAT locus is prolonged in rad1Δ and donorless mutants implies the presence of unrepaired recombination intermediates several hours after DSB formation. While this is expected in donorless mutants that lack homologous donor sequences, the fact that rad1Δ mutants exhibit donorless-like Msh2 localization highlights the fact that repair occurs aberrantly in these cells. The prolonged presence of Msh2 in both mutants is also consistent with the fact that these mutants have an activated DNA damage response and may indicate a role for Msh2 in this checkpoint.

In contrast to proposed models of mating-type switching and to gene conversion involving a 3′ nonhomologous tail on the invading strand, Rad1-Rad10-dependent 3′ nonhomologous tail removal on the second, non-invading strand appears to be independent of Msh2-Msh3. Viability was only slightly reduced in msh2Δ mutants undergoing mating-type switching, and msh2Δ mutants did not exhibit the unique viability pattern characteristic of rad1Δ and slx4Δ mutants in pedigree analysis. In this way, Rad1-Rad10-dependent 3′ nonhomologous tail removal during mating-type switching is analogous to its role in cleavage of 3′ DNA-bound Top1 lesions, which is also Msh2-Msh3 independent (Vance and Wilson 2002).

There are at least two separate error-free lesion bypass pathways in S. cerevisiae: one pathway involves the homologous recombination machinery and the Mph1 helicase and the second pathway is the Rad5-Mms2-Ubc13 branch of the Rad6-Rad18 pathway that is thought to regress replication forks and promote bypass of lesions by template switching (Torres-Ramos et al. 2002; Schürer et al. 2004; Watts 2006; Blastyák et al. 2007). We observed that both of these pathways contributed to the viability and cell cycle phenotypes of cells undergoing mating-type switching. Mph1 is a helicase that is known to be in the Rad52 epistasis group, but it is thought to function in recombinational restart of stalled replication forks (Schürer et al. 2004; Prakash et al. 2005). The fact that the rad1Δ and mph1Δ mutations were mostly epistatic suggests that the Mph1-dependent fork restart is hindered by the presence of the nonhomologous 3′-end that remains in rad1Δ mutants, although it is unclear why this might be. We cannot rule out that the role of the Mph1 helicase during gene conversion is separate from its role in replication fork restart.

Replicative lesion bypass pathway choice depends on whether the lesion (in this case, a 3′ nonhomologous tail followed by a significant single-stranded gap) is on the leading strand vs. on the lagging strand. Presumably, priming of the next Okazaki fragment on the lagging strand could bypass such a lesion and allow replication to proceed without employing specialized fork restart machinery, which may explain why the decreased viability in mms2Δ and mph1Δ mutants is relatively subtle. In addition, there is in vitro evidence using bacterial proteins that repriming of DNA synthesis can occur on the leading strand (Heller and Marians 2006). The nearest replication origin to the MAT locus is located on the centromere-proximal side, ∼2.5 kb from the Y region at MAT (http://www.oridb.org; http://www.yeastgenome.org), so it may be more likely that the 3′ Ya tail is replicated by lagging- rather than leading-strand synthesis.

Mating-type switching does not require progression through S-phase, since efficient gene conversion is detected in G₂-arrested cells, although MAT switching in G₁-arrested cells is severely reduced due to the absence of CDK1 (Cdc28) activation (Holmes and Haber 1999b; Ira et al. 2004; Wang et al. 2004). However, DNA replication may contribute to mating-type switching by priming DNA synthesis across the top strand of the partially repaired intermediate pictured in Figure 5, bypassing the need to use the cleaved 3′-end as a primer for repair synthesis and relaxing the dependence on Rad1-Rad10-Slx4. Indeed, mutations in the genes encoding polymerase α-primase or Rad27 were shown to greatly reduce mating-type switching in G₁-arrested cells (Holmes and Haber 1999b). While it was later shown that these lagging-strand synthesis factors were dispensable for mating-type switching in G₂-arrested cells (Wang et al. 2004), it is possible that cycling cells might utilize lagging-strand synthesis in addition to specialized lesion bypass pathways to promote efficient completion of gene conversion. Moreover, recent work has shown that endonuclease-induced DSBs formed during G₁ are recognized by the RPA subunit Rfa1 only after cells have entered S-phase and that formation of Rad52 foci following irradiation treatment required release of G₁-arrested cells into S-phase (Barlow et al. 2008). Further studies will be necessary to parse out the interplay between DNA replication and repair of DSBs by homologous recombination.

In summary, we conclude that gene conversion intermediates containing 3′ nonhomologous tails are principally processed by Rad1-Rad10-Slx4, even on the non-invading strand, and we propose that repair is aided by concurrent DNA replication and its associated post-replicative lesion bypass pathways.