The Saccharomyces cerevisiae Sae2 Protein Promotes Resection and Bridging of Double Strand Break Ends*

When eukaryotic chromosomes undergo double strand breaks (DSBs), several evolutionarily conserved proteins, among which the MRX complex, are recruited to the break site, leading to checkpoint activation and DNA repair. The function of the Saccharomyces cerevisiae Sae2 protein, which is known to work together with the MRX complex in meiotic DSB processing and in specific mitotic DSB repair events, is only beginning to be elucidated. Here we provide new insights into the role of Sae2 in mitotic DSB repair. We show that repair by single strand annealing of a single DSB, which is generated by the HO endonuclease between direct repeats, is defective both in the absence of Sae2 and in the presence of the hypomorphic rad50s allele altering the Rad50 subunit of MRX. Moreover, SAE2 overexpression partially suppresses the rad50s single strand annealing repair defects, suggesting that the latter might arise from defective MRX-Sae2 interactions. Finally, SAE2 deletion slows down resection of an HO-induced DSB and impairs DSB end bridging. Thus, Sae2 participates in DSB single strand annealing repair by ensuring both resection and intrachromosomal association of the broken ends.

When eukaryotic chromosomes undergo double strand breaks (DSBs), several evolutionarily conserved proteins, among which the MRX complex, are recruited to the break site, leading to checkpoint activation and DNA repair. The function of the Saccharomyces cerevisiae Sae2 protein, which is known to work together with the MRX complex in meiotic DSB processing and in specific mitotic DSB repair events, is only beginning to be elucidated. Here we provide new insights into the role of Sae2 in mitotic DSB repair. We show that repair by single strand annealing of a single DSB, which is generated by the HO endonuclease between direct repeats, is defective both in the absence of Sae2 and in the presence of the hypomorphic rad50s allele altering the Rad50 subunit of MRX. Moreover, SAE2 overexpression partially suppresses the rad50s single strand annealing repair defects, suggesting that the latter might arise from defective MRX-Sae2 interactions. Finally, SAE2 deletion slows down resection of an HO-induced DSB and impairs DSB end bridging. Thus, Sae2 participates in DSB single strand annealing repair by ensuring both resection and intrachromosomal association of the broken ends.
DNA double strand breaks (DSB 3 (s)) are a particularly dangerous form of DNA damage, because failure to repair these lesions can lead to loss of genetic information by deletions, duplications, translocations, and missegregation of large chromosome fragments (1). DSBs can arise by failures in DNA replication and by exposure to environmental factors, such as ionizing radiations and genotoxic drugs. However, they also play an important role as intermediates in meiotic and mitotic crossing over, V(D)J recombination and yeast mating type switching.
When DSBs occur, many proteins are recruited to the break sites and serve both to promote a checkpoint response and to initiate DNA repair that can occur through non-homologous end joining or homologous recombination (HR) (2). Whereas non-homologous end joining implies recombination between sequences with little or no homology, HR involves exchange of genetic information between homologous DNA sequences and is the major DSB repair process in Saccharomyces cerevisiae.
HR initiates with a DSB (3,4), whose 5Ј-ends resection leaves 3Ј-ended single-stranded DNA (ssDNA) tails. Then, depending on the position of the homologous partner, on the initiation event and on the length of the homology region in the recombinant molecules, HR may occur by different mechanisms, including double strand break repair, synthesis-dependent strand annealing, and break-induced replication (2,5). Moreover, when a DSB occurs between direct repeats, its repair is primarily achieved by a particular kind of HR named single strand annealing (SSA). SSA requires DSB resection to generate long 3Ј-ended single-stranded tails that can anneal with each other when resection is sufficient to uncover the duplicated sequences. Single-stranded tails are then removed by nucleases, and the resulting gaps/nicks are filled in by DNA repair synthesis and ligation, resulting in deletion of one repeat and the intervening region (6). Mitotic HR involves several proteins, including Rad51, Rad52, Rad54, Rdh54/Tid1, Rad55, Rad57, Rad59, and the ssDNA binding complex replication protein A (2). Consistent with a role for Rad52 in providing the single strand annealing activity (7,8), Rad52 is essential for initiation of most HR events, including SSA, whereas Rad51, Rad54, Rad55, and Rad57 are dispensable for SSA-mediated DSB repair (9).
Different functions in DSB repair can be envisaged also for a protein complex known as MRN (Mre11-Rad50-Nbs1) in mammals and MRX (Mre11-Rad50-Xrs2) in S. cerevisiae, which has been shown to localize at meiotic and mitotic DSBs almost immediately after their formation (10,11). The S. cerevisiae mrx null mutants are severely impaired in meiotic recombination (4,12) and hairpin-capped DSB repair (13) and exhibit decreased frequencies of ionizing radiation-induced sister chromatid and interhomologue recombination (14). Moreover, they markedly retard the kinetics of DSB repair by SSA, although they do not prevent its completion (15).
The Mre11 subunit of MRX has both exonuclease and endonuclease activities (16 -18) that likely function in DSB repair. However, whereas point mutations in the Mre11 phosphoesterase domain result in accumulation of unresected meiotic DSBs and sporulation failure in vivo (19), most S. cerevisiae mre11 nuclease defective alleles cause only a mild sensitivity to ionizing radiations and do not seem to affect either HOinduced DSB resection or ionizing radiation-induced sister chromatid recombination (14, 19 -21). Moreover, budding yeast Mre11 associates with a DSB only transiently (11), whereas resection of an HO-induced DSB can proceed for several hours until repair occurs (22). This indicates that the functions of MRX in mitotic DSB repair might be partially independent of its nuclease activity, which is instead essential in the single strand endonucleolytic removal of Spo11 from meiotic DNA ends (12,19).
Consistent with this hypothesis, the MRX complex has been shown to cooperate with Rad52 in holding the two ends of the broken chromosome together independently of its nuclease activity (23,24). Moreover, it promotes the assembly of DSB-specific cohesion domains, thus contributing to maintain association between sister chromatids during DSB repair (25)(26)(27). According to these functions, MRX structure resembles that of the SMC (structural maintenance of chromosome) proteins that are required for sister chromatid cohesion (28). The MRX complex associates with DNA ends through the Mre11 protein and the globular domain of Rad50, with the coiled-coil arms of Rad50 extending outwards and interacting with an MRX complex on the other site of the break through Cys-Xaa-Xaa-Cys motifs (29).
The S. cerevisiae Sae2 protein is known to work together with the MRX complex at least in meiotic DSB processing and in the repair of mitotic hairpin-capped DSBs. In fact, sae2 null mutants show meiotic recombination defects that are undistinguishable from those of hypomorphic rad50s and mre11s mutants (30 -33). Moreover, similarly to rad50s and mre11s mutants, sae2⌬ cells are both hypersensitive to the alkylating agent methyl methanesulfonate and defective in mitotic hairpin-capped DSB repair (13,34).
Sae2 has neither known homologues in other species nor obvious motifs, and its function is only beginning to be elucidated. The lack of Sae2 does not affect the overall recombination rate when a DSB is induced between two inverted repeats (35). Nonetheless, SAE2 deletion delays Rad52 recruitment at sites of DNA damage after ␥-irradiation or I-SceI-induced DSBs (11) and increases the frequency of repair events that lead to the duplication of inverted repeats (35) and to palindromic gene amplification (36). Because similar duplications of inverted repeats have been observed in the repair of an HO-induced DSB with only one end homologous to the donor (37), Sae2 may be involved in ensuring that both DSB ends participate in HR.
To provide new insights into the Sae2 function in mitotic DSB repair, we have investigated whether the absence of Sae2 affects resection and SSA repair of a single DSB generated by the site-specific HO endonuclease. We show that Sae2 contributes to both resection and intrachromosomal association of DSB ends, thus ensuring efficient DSB repair.
SAE2 and MRE11 deletions were performed as described previously (34). To generate strains carrying three copies of the GAL-SAE2 fusion integrated at the URA3 locus, the YMV80 strain was transformed with ApaI-digested plasmid pML508, carrying the entire SAE2 coding region fused to the GAL1 promoter. The YMV80 derivative SAE2-HA and GAL-SAE2-HA strains, carrying the fully functional SAE2-HA allele at the SAE2 chromosomal locus, or the GAL-SAE2-HA allele at the URA3 chromosomal locus, respectively, were constructed as described previously (34). To generate rad50s mutants, YMV80 and yJK40.6 cells were transformed with MscI-digested plasmid pML533, carrying the first 666 bp of the RAD50 coding region with the K81I mutation. Accuracy of all disruptions was verified by PCR, whereas GAL-SAE2 and GAL-SAE2-HA integrations were verified by Southern blot analysis.
Plasmids-To give rise to plasmid pML508, a fragment spanning from position ϩ1 to position ϩ1162 from the SAE2 translation initiation codon was amplified by PCR using yeast genomic DNA as a template and oligonucleotides PRP559 (5Ј-CGC GGA TCC ATA TGG TGA CTG GTG AAG AAA ATG-3Ј) and PRP560 (5Ј-AAC TGC AGC TGG TAA GTT AGG TGT CAT TTG-3Ј) as primers and was then cloned together with a 700-bp EcoRI-BamHI fragment containing the GAL1 promoter, into the EcoRI-PstI sites of plasmid YIplac211 (40). To give rise to plasmid pML533, a 1051-bp fragment containing the promoter and the first 666 bp of the coding region of the rad50s K81I allele was amplified by PCR using oligonucleotides PRP706 (5Ј-CGG AAT TCC CGA TAG TAC TTC CAC TTA CAA TAC-3Ј) and PRP707 (5Ј-CGG AAT TCC ATT GCT TTC GAT CTG TCT TTA TCC-3Ј) as primers and yeast genomic DNA from MJL1699 strain (kindly provided by N. Kleckner, Harvard University, Cambridge, MA) as a template and was then cloned into the EcoRI site of plasmid YIplac128 (40). The presence of the mutation was verified by nucleotide sequence analysis.
Other Techniques-Synchronization experiments, protein extract preparation, and Western blot analysis were performed as previously described (41). DSB formation and repair were detected by Southern blot analysis using an Asp718-SalI fragment containing part of the LEU2 gene as a probe (38). DSB end resection at the LEU2 locus was analyzed on alkaline-agarose gels as described in Ref. 42, using a single-stranded probe complementary to the unresected DSB strand. This probe was obtained by in vitro transcription using Promega Riboprobe System-T7 and plasmid pML514 as a template. Plasmid pML514 was obtained by cloning the KpnI-ClaI fragment containing part of the LEU2 gene in the KpnI-ClaI sites of pGEM-7Zf plasmid (purchased from Promega).

RESULTS
The Lack of Sae2 Impairs DSB Repair by SSA-We examined the effects caused by lack or excess of Sae2 on SSA repair by using YMV80 derivative strains (38), carrying either a deletion of SAE2 or a GAL1-SAE2 fusion integrated at the URA3 locus. In these strains, a single DSB can be generated by expressing the site-specific HO endonuclease gene from a galactose-inducible GAL1 promoter. The normal HO recognition sites at MAT, HML, and HMR loci are deleted, whereas an HO target sequence on chromosome III is flanked on either side by homologous sequence repeats, consisting of the 3Ј-end of the LEU2 gene. One of these repeats is adjacent to the HO cut site (leu2::cs), whereas the other repeat (his4::leu2) is located 25 kb centromere distal from the break site (Ref. 38; see also Fig. 1A). Because of the presence of homologous sequence repeats on either side of the HO cut site, the DSB in the YMV80 derivative strains has been reported to be mainly repaired by a Rad52-dependent SSA pathway (38). In fact, the ends of the DSB are degraded by 5Ј-3Ј-exonucleases until complementary sequences flanking the break are exposed and can be annealed. Subsequent ligation of the two chromosomal fragments repairs the break, concomitantly abolishing the HO recognition sequence and one of the repeats (38).
Because the ability to repair the HO-induced DSB is influenced by the cell cycle stages (43), we arrested all cell cultures in mitosis and kept them blocked by nocodazole treatment during break induction by galactose addition. We then monitored the kinetics of HO-cut formation and its subsequent repair by Southern blot analysis. As depicted in Fig. 1A, hybridization with a LEU2 probe of KpnI-digested YMV80 genomic DNA run on a native agarose gel reveals 8-and 6-kb DNA fragments in the absence of HO cut. HO-induced DSB formation results in decrease of the amount of 6-kb species and concomitant appearance of 2.5-kb fragments. The break can be eventually repaired by SSA, leading to a final product of 3.5-kb and disappearance of both the 2.5-and 8-kb fragments. As shown in Fig. 1B, HO-cut induction was rapid and efficient in all cell cultures, because the leu2::cs sequence was cut in more than 90% of cells within 30 min after galactose addition. Then, the 3.5-kb SSA repair products started to accumulate in both wild type and SAE2 overexpressing cells ϳ5 h after HO induction, concomitantly with the decrease of both the 2.5-and 8-kb fragments. SSA products were instead detectable only at very low levels even 10 h after HO induction in sae2⌬ cells (Fig. 1B).
Consistent with defective DSB repair by SSA in the absence of Sae2, only 10% of sae2⌬ cells were able to form colonies on galactose-containing plates (HO expression on), whereas cell viability was unaffected in wild type and GAL-SAE2 cells under the same conditions (Fig. 1C). Thus, Sae2 is required for efficient repair of the HO-induced DSB by SSA, which appears to be unaffected by SAE2 overexpression, indicating that increasing Sae2 levels (Fig. 1D) is not sufficient to accelerate this process in wild type cells.
The lack of Sae2 severely reduced SSA-dependent recombination events also in YMV45 derivative strains (Fig. 2), where SSA repair of HO-induced DSBs occurs very rapidly (38). In fact, the distal leu2 sequence in these strains is located only 4.6 kb away from the HO recognition site at leu2::cs (Ref. 38, see also Fig. 2A). When cell cultures of these strains were arrested in mitosis and kept blocked by nocodazole treatment during break induction, SSA products accumulated in wild type cells already 90 min after HO induction (Fig. 2B). Moreover, they became detectable in sae2⌬ cells ϳ150 min after galactose addition (Fig.  2B), but their amount was much lower than in wild type, thus confirming a critical role for Sae2 in SSA repair. Consistent with the faster DSB repair by SSA in YMV45 sae2⌬ cells compared with YMV80 sae2⌬ cells, 50% of YMV45 sae2⌬ cells were able to form colonies on galactosecontaining plates (Fig. 2C).
SAE2 Overexpression Partially Suppresses the SSA Defects of rad50s Mutants-In contrast to the mrx null alleles, which are unable to generate meiotic DSBs, sae2⌬ cells display the same meiotic recombination defects as the hypomorphic rad50s and mre11s alleles (30 -33), allowing to hypothesize that the Rad50s and Mre11s mutant proteins might specifically impair MRX-Sae2 interaction. We then asked whether rad50s mutant cells were defective in SSA. After construction of a rad50s K81I mutant (12) in the YMV80 background, cell cultures were arrested in  (38). Hybridization with a Asp718-SalI LEU2 probe of KpnI-digested YMV80 genomic DNA run on a native agarose gel reveals two fragments, 8-and 6-kb long (uncut) in the absence of HO-cut, whereas the HO-induced DSB results in the disappearance of the 6-kb species and concomitant formation of 2.5-kb fragments (HO-cut) that can be eventually repaired by SSA, leading to a final product of 3.5-kb (SSA product). The vertical lines indicate the restriction sites recognized by KpnI (K). B, HO-cut formation and SSA in YMV80 derivative strains. Exponentially growing YEPϩraf cell cultures of wild type YMV80 and isogenic sae2⌬ and GAL-SAE2 strains, all carrying the HO-cut site as depicted in A and a galactose-inducible HO gene, were arrested in mitosis with nocodazole and then transferred to YEPϩrafϩgal in the presence of nocodazole at time zero. KpnI-digested genomic DNA prepared from cell samples collected at the indicated time points were subjected to the Southern blot analysis described in A. C, serial dilutions of YMV80 and isogenic sae2⌬ and GAL-SAE2 strains were spotted on YEPD and YEPϩrafϩgal plates that were incubated at 30°C for 3 days. D, cell cultures of YMV80 derivative strains carrying either the SAE2-HA allele at the SAE2 chromosomal locus or the GAL-SAE2-HA fusion at the URA3 locus, exponentially growing in YEPϩraf (raf), were shifted to YEPϩrafϩgal (gal) for 5 h. Protein extracts were subjected to Western blot analysis with anti-hemagglutinin antibodies. NOVEMBER 18, 2005 • VOLUME 280 • NUMBER 46 mitosis and kept blocked by nocodazole treatment during break induction by galactose addition. Southern blot analysis was then used to follow the kinetics of DSB repair by SSA. As shown in Fig. 3A, SSA products were barely detectable in rad50s cells even 8 h after HO induction, similarly to what observed in isogenic sae2⌬ cells (Fig. 1B). Accordingly, rad50s cell survival on galactose containing plates was reduced to the same extent as in sae2⌬ cells (Fig. 3B).

Sae2 and DSB Repair
If the rad50s SSA defects were because of altered MRX ability to interact with Sae2, Sae2 overproduction might be expected to suppress it. This was indeed the case, because SSA products in GAL-SAE2 rad50s cells treated as above became detectable and started to accumulate, although to a lower extent than in wild type, ϳ5 h after HO induction (Fig. 3A). In agreement with partial suppression of rad50s SSA defects by SAE2 overexpression, GAL-SAE2 rad50s cells showed increased cell survival after HO induction on galactose-containing plates compared with rad50s cells (Fig. 3B).
The Lack of Sae2 Slows Down Resection of an HO-induced DSB-Because the HO break can undergo repair by SSA only once 5Ј-3Ј-resection has uncovered both direct repeats, allowing their subsequent annealing and ligation, SSA repair defects in sae2⌬ cells might be because of impairments in the generation of 3Ј-ended ssDNA. We therefore used denaturing gel electrophoresis and Southern blot analysis with a single-stranded LEU2 probe to monitor both HO cut at leu2::cs and the extent of resection, as depicted in Fig. 4A. When these analyses were performed on the same cell cultures shown in Fig. 1, initiation of DSB resection turned out to be delayed in YMV80 sae2⌬ cells (Fig. 4B). In fact, 3Ј-ended resection products (named r1-r6) could be detected in these cells ϳ1 h later than in isogenic wild type cells.
The sae2⌬ DSB resection defects seemed to only partially account for the reduced SSA efficiency. In fact, the amount of SSA products was very low for at least 10 h after HO induction in YMV80 sae2⌬ cells (Fig.  1B), whereas the 3Ј-ended ssDNA products were detectable in the same cells already 2-3 h after galactose addition and remained stable for at least 10 h (Fig. 4B). Conversely, the amount of resection products decreased in the isogenic wild type cells (Fig. 4B) concomitantly with the appearance of SSA products (Fig. 1B). Moreover, DSB resection in the sae2⌬ YMV80 derivative strain seemed to proceed for at least 25 kb and to uncover the homologous LEU2 sequence (his4::leu2) located 25 kb away from the HO-cut site. In fact, the amount of the 8-kb his4::leu2 DNA fragment started to decrease ϳ6 -7 h after HO induction in both wild type and sae2⌬ YMV80 cells (Fig. 1B). Because SSA products were dramatically reduced in the same sae2⌬ cells compared with wild type, the disappearance of the 8-kb DNA fragment in the absence of Sae2 was likely because of DSB resection proceeding long enough to uncover the homologous his4::leu2 sequence.
We also monitored the extent of DSB ends resection in the YMV45 cell cultures shown in Fig. 2, where the distal leu2 sequence is located only 4.6 kb away from the HO-cut site ( Fig. 2A). As shown in Fig. 4C, the FIGURE 2. SSA repair of an HO-induced DSB in YMV45 sae2⌬ cells. A, map of the YMV45 chromosome III region where the HO-cut site is flanked by homologous LEU2 sequence repeats (black boxes) that are only 4.6 kb apart (38). After digestion of genomic DNA with KpnI, DNA separation on a native agarose gel and hybridization with the LEU2 probe, a 14.5-kb fragments (uncut) is detected in the absence of HO-cut, whereas HO-induced DSB formation results in generation of 12-and 2.5-kb fragments (HO-cut) that can be eventually repaired by SSA, generating a final product of 8.0-kb (SSA product). B, HO-cut formation and SSA repair in YMV45 derivative strains. Exponentially growing YEPϩraf cell cultures of wild type YMV45 and its sae2⌬ derivative strain, both carrying the HO-cut site as depicted in A and a galactose-inducible HO gene, were arrested in mitosis with nocodazole and transferred to YEPϩrafϩgal in the presence of nocodazole at time zero. KpnI-digested genomic DNA prepared from cell samples collected at the indicated time points were subjected to Southern blot analysis as described in A. C, serial dilutions of YMV45 and isogenic sae2⌬ strain were spotted on YEPD and YEPϩrafϩgal plates that were incubated at 30°C for 3 days.
ssDNA resection products (named r1-r4) at leu2::cs started to accumulate in YMV45 sae2⌬ cells ϳ30 min later than in isogenic wild type cells under the same conditions. As expected, the resection products started to decrease in wild type cells (Fig. 4C) concomitantly with accumulation of SSA products (Fig. 2B). Resection products remained instead stable until the end of the experiment in sae2⌬ YMV45 cells, which showed very low efficiency of SSA product formation, similarly to what observed in the sae2⌬ YMV80 strain. Thus, although the lack of Sae2 slows down DSB resection, its effects on DSB repair by SSA likely reflect impairment of other Sae2 functions besides those involved in the generation of 3Ј-ended tails.
Sae2 Contributes to the Maintenance of Intrachromosomal Association between the Ends of a Broken Chromosome-DSB repair needs the DNA ends flanking the DSB site to remain adjacent to each other, and this requires the MRX complex (23,24). We therefore asked whether the lack of Sae2 might affect repair of DSBs by impairing not only their resection but also the intrachromosomal association of their ends. We used strains where the DNA proximal to the HO break could be visualized by binding of a LacI-GFP (green fluorescent protein) fusion protein to multiple repeats of the LacI repressor binding site, LacO, integrated at a distance of 50 kb on either side of an HO cleavage site on chromosome VII (Ref. 23 and Fig. 5A). To determine the separation efficiency of two unlinked LacO arrays, we used also a strain where the arrays were located both 50 kb telomere-proximal to the HO site on chromosome VII and on chromosome III (Ref. 23 and Fig. 5C). Both strains also expressed galactose-inducible HO endonuclease and the LacI-GFP fusion protein, the binding of the latter to the LacO arrays allowing their visualization. To ensure that the broken chromosomal fragments did not attempt recombination, the endogenous HO sites at the MAT, HML, and HMR loci on chromosome III were deleted in both strains (23).
Cultures of the original wild type and isogenic sae2⌬ derivative strains were arrested in mitosis and kept blocked by nocodazole treatment during break induction by galactose addition. As shown in Fig. 5A, the majority of wild type cells containing two LacO arrays flanking the HO site showed a single LacI-GFP spot in both uninduced and break-induced conditions throughout the experiment, because of their ability to hold the broken ends together. In contrast, 33% of the otherwise isogenic sae2⌬ cells showed two LacI-GFP spots at both 3 and 6 h after HO induction (Fig. 5A), with an ϳ3-fold increase compared with uninduced conditions (Fig. 5A, time zero). To rule out the possibility that loss of SAE2 caused appearance of two LacI-GFP foci by disrupting chromosome structure or sister chromatid cohesion in nocodazole-arrested cells, we performed the same analysis also on isogenic strains carrying a single LacO array 50 kb telomere-proximal to the HO site. As shown in Fig. 5B, 90% of sae2⌬ cells with the unique LacO array showed a single LacI-GFP spot even 3 and 6 h after DSB induction. Thus, the appearance of two LacI-GFP spots in sae2⌬ cells carrying two LacO arrays flanking the HO site was primarily because of impairment in DSB ends association. In any case, the latter strain did not mimic the wild type and sae2⌬ control strains carrying two unlinked LacO arrays, which showed ϳ75% of cells with two LacI-GFP foci in both uninduced and break-induced conditions (Fig. 5C), suggesting that some bridging of DSB ends takes place also in the absence of Sae2.
As previously reported, also the MRX complex is involved in bridging DSB ends (Refs. 23 and 24 and Fig. 5). Interestingly, the amount of cells containing two LacO arrays flanking the HO site and showing two LacI-GFP spots after HO-induced DSBs was higher in the mre11⌬ strain than in the otherwise isogenic sae2⌬ strain or in the rad50s strain that might be specifically impaired in MRX-Sae2 interaction (Fig. 5A). However, although appearance of two LacI-GFP foci in a subset of sae2⌬ cells seemed to be primarily because of defective intrachromosomal DSB ends association (see above), defects other than those affecting bridging of DSB DNA ends appear to contribute to generate two LacI-GFP foci in FIGURE 3. The SSA defects of a rad50s mutant are partially suppressed by SAE2 overexpression. A, HO-cut formation and repair in YMV80 derivative strains. Exponentially growing YEPϩraf cell cultures of GAL-SAE2 YMV80 and isogenic rad50s and GAL-SAE2 rad50s strains, all carrying the HO-cut site as depicted in Fig. 1 and a galactose-inducible HO gene, were arrested in mitosis with nocodazole and then transferred to YEPϩrafϩgal in the presence of nocodazole at time zero. KpnI-digested genomic DNA prepared from cell samples collected at the indicated time points were subjected to the Southern blot analysis described in the legend to Fig. 1. * denotes the rad50s::LEU2 construct integrated at the RAD50 locus that is detected by the probe. B, serial dilutions of YMV80 and isogenic sae2⌬, rad50s, GAL-SAE2, and GAL-SAE2 rad50s strains were spotted on YEPD and YEPϩrafϩgal plates that were incubated at 30°C for 3 days. NOVEMBER 18, 2005 • VOLUME 280 • NUMBER 46 mre11⌬ cells. In fact, a significant amount of mre11⌬ cells with a single LacO array exhibited two LacI-GFP spots in both uninduced and breakinduced conditions (Fig. 5B), likely because of impairment in sister chromatid cohesion that has been reported to be defective in mre11⌬ cells (26,27).

Sae2 and DSB Repair
As shown in Fig. 5A, the frequency of sae2⌬ mre11⌬ cells containing two LacO arrays flanking the HO site and showing two LacI-GFP spots after HO-induced DSBs was similar to that observed in otherwise isogenic mre11⌬ cells. Thus, Sae2 and MRX seem to act in the same pathway with respect to intrachromosomal association maintenance.

DISCUSSION
Although Sae2 has been implicated together with the MRX complex in the repair of meiotic DSBs by allowing the endonucleolytic removal of Spo11 from the break sites (30), its function in the cellular response to DSBs during a mitotic cycle is only beginning to be elucidated. Interestingly, Sae2 loss of function has been shown to impair the repair of hairpin-capped DSBs (13) and to increase the percentage of DSB repair events that lead to inverted repeat duplication, without affecting the overall recombination rate (35,36), FIGURE 4. Effects of Sae2 lack on DSB ends resection. A, schematic representation of the method used to detect DSB formation and 5Ј-3Ј-resection in the region immediately centromere-proximal to the leu2::cs DSB site in YMV80 and YMV45 derivative strains. Southern blot analysis with a single-stranded probe specific to the LEU2 sequence (see "Experimental Procedures") of SspI-digested genomic DNA run on an alkaline-agarose gel reveals a 1.5-kb SspI fragment (uncut) in the absence of HO-cut. After HO-cut, the size of this fragment is reduced to 1.1 kb (HO-cut), whereas 5Ј-3Ј-resection past the SspI sites eliminates cutting at these sites and thus produces larger SspI fragments detected by the probe. Bands labeled r1 through r6 are the products of resection through SspI sites located 2.2, 5.0, 5.2, 8.3, 9.1, and 14.1 kb centromere-proximal to the HO-cut site. B, HO-cut and ssDNA formation in YMV80 derivative strains. Genomic DNA prepared from samples taken at the indicated times from the same cell cultures analyzed in Fig. 1 was digested with SspI, separated on alkaline-agarose gel and analyzed as described in A. * denotes the distal unresected his4::leu2 SspI fragment that is detected by the probe. C, HO-cut and ssDNA formation in YMV45 derivative strains. Genomic DNA prepared from samples taken at the indicated times from the same cell cultures analyzed in Fig. 2 was digested with SspI, separated on alkaline-agarose gel and analyzed as described in A. * denotes the distal unresected leu2 SspI fragment detected by the probe.
suggesting that Sae2 may be involved in ensuring that both DSB ends participate in HR.
Based on the notion that a DSB between two direct repeats stimulates recombination by SSA (2, 5), we have studied the effects of Sae2 loss of function on the repair of an HO-induced DSB occurring between two direct repeats. We show that Sae2 is required to ensure efficient DSB repair by SSA. In fact, SSA repair of HO-induced DSBs is severely impaired in sae2⌬ cells, although not completely abolished.
Because also the mrx⌬ mutants show reduced efficiency of DSB repair by SSA (15), Sae2 may influence the SSA mechanism by modulating MRX functions. In contrast to mrx⌬ mutants that are unable to generate meiotic DSBs (4,12), both sae2⌬ and hypomorphic rad50s mutants allow meiotic DSB formation but do not allow their resection (30 -33), suggesting that the Rad50s protein variant might specifically impair MRX-Sae2 interactions. In agreement with this hypothesis, we found that the SSA defects of sae2⌬ cells are undistinguishable from those of a rad50s mutant. Importantly, high levels of Sae2, which are not able to restore SSA in a rad50⌬ mutant, 4 partially suppress rad50s SSA defects and cell death after an HO-induced DSB. This suggests that Sae2 may play some roles in regulating MRX activity and that the Rad50s amino acid substitutions, which all map to a region on the surface of the protein that is different from both the catalytic and the DNA binding domains (44), may cause loss of Sae2 functions, possibly by disrupting Rad50-Sae2 interaction.
We also examined the molecular mechanisms underlying the reduced efficiency of SSA in the absence of Sae2. Because the HO break within direct repeats can undergo SSA repair only once the 5Ј-3Ј-resection has uncovered both homologous sequences, the lack of Sae2 may affect DSB resection and the subsequent repair through SSA. We found that the resection of the HO-induced DSB is slowed down in the absence of Sae2, indicating that Sae2 is involved in generating 3Ј-ended ssDNA.
Because mrx⌬ mutants slow down DSB resection (15), Sae2 may affect processing of the 5Ј-DNA ends by altering MRX functions. However, the lack of Sae2 does not seem to affect MRX complex formation (45), whereas several data suggest that it likely impairs MRX nuclease activity. In fact, both sae2⌬ cells and mre11 nuclease-defective mutants are impaired in repairing hairpin-capped DSBs (13) and in resecting meiotic DSBs (19, 30 -33). Moreover, both mutants show defective Mre11 dissociation from unprocessed meiotic DSBs (10) and from DNA lesions induced by ␥-rays (11), suggesting that Mre11 release from DSBs requires its nuclease activity and that the lack of Sae2 may freeze 4 M. Clerici, D. Mantiero, G. Lucchini, and M. P. Longhese, unpublished observations. FIGURE 5. Sae2 contributes to DSB end bridging. Wild type, sae2⌬, mre11⌬, rad50s, and mre11⌬ sae2⌬ strains expressing LacI-GFP and carrying its target sequence LacO (see text for details) and the GAL-HO construct were arrested with nocodazole (time zero) and transferred to YEPϩrafϩgal to induce HO expression in the presence of nocodazole to maintain the cell cycle arrest. As indicated on the left side of the figure, where two vertical lines connected by horizontal bars represent sister chromatids with black circles indicating the centromeres, strains carried either two LacO arrays (green boxes) at 50 kb on either side of the HO-cut site on chromosome VII (A), or a single LacO array 50 kb telomere-proximal to the HO-cut site (B), or one LacO array 50 kb telomere-proximal to the HO site on chromosome VII and a second unlinked LacO array on chromosome III (C) (23). Cell samples taken at the indicated times after HO induction were analyzed with a fluorescence microscope to determine the fraction of cells in each sample that contained LacI-GFP foci separated by Ͼ0.5 m (graphs). The separation distance between foci was measured for 500 cells/sample. Representative pictures of the corresponding wild type and sae2⌬ cells exhibiting GFP-LacI foci 3 h after HO induction are shown on the right side of each panel. Scale bar is equal to 1 m.
Mre11 at the DNA-damaged sites by altering its nuclease function. Because Rad50 has been proposed to undergo an ATP-driven conformational switch that is well suited to prepare DNA ends/hairpins for nucleolytic cleavage by Mre11 (44), lack of Sae2 may affect Mre11 nuclease activity either directly or by altering Rad50 conformational switch. However, some of the nuclease-defective mre11 mutants do not seem to be significantly affected in mitotic DSB resection (14, 19 -21), suggesting that Sae2 may impair this process also by altering MRX functions other than its nuclease activity.
In any case, although the resection products in sae2⌬ cells appear 30 -60 min later than in wild type, they accumulate and remain stable for several hours, suggesting that the slowing down of DSB resection cannot be the sole cause of SSA defects. Indeed, we show that Sae2 is also required for intrachromosomal bridging of DSB ends, which have been previously shown to remain associated until the break is repaired by the action of the MRX complex (23,24). In fact, SAE2 deletion causes dissociation of broken ends in a significant subset of cells by acting in the same pathway of the MRX complex (Fig. 5). Because some DNA repair options are lost when DNA ends do not remain adjacent to each other, this raises the possibility that the reduced bridging of the DSB ends observed in sae2⌬ cells may account for the poor efficiency of SSA in repairing HO-induced lesions. Interestingly, Sae2 does not appear to be involved in ensuring sister chromatid cohesion (Fig. 5), which instead requires the MRX complex (26,27). The MRX-dependent intrachromosomal association has been shown to occur independently of Mre11 nuclease activity but to require the Cys-Xaa-Xaa-Cys zinc-hook motifs located in the coiled-coil region of Rad50, which in turn does not seem to be involved in ensuring sister chromatid cohesion (46). Therefore, the lack of Sae2 might specifically affect the MRX function in DSB end bridging.
Altogether, our data indicate that Sae2 participates in maintaining genome stability by ensuring the efficiency of DSB repair, possibly by modulating a subset of MRX functions. Finding Sae2 functional homologues in other species will be a challenging issue for future research and will increase our overall knowledge of the mechanisms preserving genetic integrity in eukaryotic cells.