The non-homologous end-joining factor Nej1 inhibits resection mediated by Dna2–Sgs1 nuclease-helicase at DNA double strand breaks

Double strand breaks (DSBs) represent highly deleterious DNA damage and need to be accurately repaired. Homology-directed repair and non-homologous end joining (NHEJ) are the two major DSB repair pathways that are highly conserved from yeast to mammals. The choice between these pathways is largely based on 5′ to 3′ DNA resection, and NHEJ proceeds only if resection has not been initiated. In yeast, yKu70/80 rapidly localizes to the break, protecting DNA ends from nuclease accessibility, and recruits additional NHEJ factors, including Nej1 and Lif1. Cells harboring the nej1-V338A mutant exhibit NHEJ-mediated repair deficiencies and hyper-resection 0.15 kb from the DSB that was dependent on the nuclease activity of Dna2–Sgs1. The integrity of Nej1 is also important for inhibiting long-range resection, 4.8 kb from the break, and for preventing the formation of large genomic deletions at sizes >700 bp around the break. Nej1V338A localized to a DSB similarly to WT Nej1, indicating that the Nej1–Lif1 interaction becomes critical for blocking hyper-resection mainly after their recruitment to the DSB. This work highlights that Nej1 inhibits 5′ DNA hyper-resection mediated by Dna2–Sgs1, a function distinct from its previously reported role in supporting Dnl4 ligase activity, and has implications for repair pathway choice and resection regulation upon DSB formation.

DNA double strand breaks (DSBs) 2 are one of the most deleterious forms of DNA damage. Aberrant repair of these lesions results in mutations, genomic instability, and cell death. In Saccharomyces cerevisiae, DSBs are repaired predominately through homology-directed repair (HDR); however, non-homologous end joining (NHEJ), which requires little or no homology, is conserved and involves direct religation of the DNA ends. yKu70/80 (Ku), Mre11-Rad50 -Xrs2 (MRX), Lif1-Dnl4, and Nej1 are the core components of this pathway. It is becoming clear that the functionality of these factors impacts repair pathway choice (for a review, see Ref. 1). Ku and MRX localize first to the break site independently of one another (2). MRX tethers the DNA ends to prevent their separation (3,4), whereas Ku has high affinity for DSB ends and recruits other factors, including Lif1-Dnl4 and Nej1 (2,5,6). Dnl4 interacts with Lif1 and ligates DNA ends to complete repair (7)(8)(9)(10)(11). Nej1 has no known enzymatic activity, but it stimulates the ligase activity of Dnl4 -Lif1 by interacting with Lif1 and promoting Dnl4 deadenylation (8). Cells lacking NEJ1 are as defective in end joining as ku70⌬ and dnl4⌬ mutant cells, underscoring the importance of Nej1 in NHEJ (8,12,13). Nej1 and Lif1 physically interact via specific residues in the C-terminal region of Nej1 (12, 14 -16); however, the importance of their association in repair pathway choice or end resection has not been fully investigated.
The choice between HDR and NHEJ is largely dependent on the regulation of 5Ј resection at DSB ends (17). Nej1 and Lif1 enhance Ku stability at the unresected DNA ends (8); however, these observations stem from work in cells carrying full deletions of NEJ1 and LIF1, which could potentially impinge on the stable association of other factors at the break. In G 1 , NHEJ is the predominant pathway of repair as Mre11 endonuclease activity is inactive and Ku binds the ends of DNA, preventing resection (2, 18 -20). In S/G 2 , when a homologous template is present, Sae2 and Mre11 are activated to initiate resection, which exposes a short track of 3Ј ssDNA on each side of the DSB (21). MRX resects 100 -300 nucleotides of DNA (22)(23)(24)(25)(26). As Ku has reduced affinity for ssDNA ends, it dissociates. The exposed ends are substrates for resection primarily by Exonuclease 1 (Exo1) but also via Dna2-Sgs1 nuclease-helicase as these nucleases share functional redundancy to promote long-range 5Ј 3 3Ј resection (22)(23)(24)26). Once resection is initiated, NHEJ is no longer an option as the ssDNA generated is quickly bound by Rad51, which drives the search for homology and HDRmediated repair (27). Even in the absence of MRX or Sae2 function, DNA end protection by Ku is not indefinite as resection can be initiated via Exo1 and Dna2-Sgs1 but with delayed timing (26).
Alternative pathways can be utilized to promote survival if cells are unable to repair by NHEJ or HDR. One pathway, microhomology-mediated end joining (MMEJ), depends on a short region of homology (5-25 bp) for repair and is Rad52independent (1,28). MMEJ results in small insertions or dele-tions adjacent to the break site (29 -31). Alternatively, single strand annealing (SSA) can be used to repair DSBs when NHEJ and HDR are deficient or no homologous sister chromatid template is located for repair (28,32). In SSA, homologous ends recognize one another and repair the DSB but with a large deletion of genomic material between the regions of homology on adjacent strands. Yeast cells unable to repair through conventional NHEJ or HDR shift to these alternative pathways to survive, but this comes at the expense of genomic integrity. Thus, functional NHEJ and regulated resection for HDR are essential for proper pathway choice, genomic stability, and cell survival.
Here we characterize Nej1 V338A , a mutant we generated previously (16). Our results indicate that the binding between Nej1 and Lif1 subsequent to their localization at the DSB is critical for regulating nuclease function and 5Ј resection. We observed that the level of resection in nej1-V338A mutant cells was significantly above the level observed in wild type and similar to that of nej1⌬ and ku70⌬ mutants. Moreover, we found that Nej1 is important for inhibiting hyper-resection that is mediated by Dna2-Sgs1.

Nej1-Lif1 interactions at DSBs prevent 5 resection
We previously generated a Val 3 Ala mutation in Nej1 at amino acid 338 (Nej1 V338A ) that resulted in a marked reduction in its interaction with Lif1 (16); however, there was Ͼ90% efficiency in Dnl4-dependent end-joining activity compared with the ϳ2% repair rate in cells where NEJ1 was deleted. Surprisingly, even though nej1-V338A mutant cells were proficient for Dnl4 ligase activity, cell survival significantly decreased when a DSB in the genome was generated by the HO endonuclease (16). These observations suggested that the interaction between Nej1 and Lif1 contributes to repair through a mechanism distinct from DNA ligation (16).
Similar to ku70⌬ mutant cells, 5Ј resection at a DSB increases when NEJ1, LIF1, or DNL4 has been deleted, underscoring the importance of these factors in stabilizing Ku binding at the DSB (6,8,33,34). Determining the functionality of each individual protein, however, has been challenging because of the interdependency among these factors for their stable association with the DSB. To bring further insight into the role of Nej1 in NHEJ, we characterized the nej1-V338A point mutant at the DSB. We monitored 5Ј resection levels in cells blocked in G 1 by ␣-factor because HDR, the central repair pathway, is activated by Cdk1 only when cells go into S phase and is inhibited in G 1 -arrested cells (35). Moreover, DNA processing was measured in cells with hml⌬ hmr⌬, which lack a homologous template required for canonical HDR repair ( Fig. 1A and Ref. 36). Resection levels were determined by a quantitative PCR-based approach after HO-induced DSB formation and relied on two RsaI cut sites located 0.15 and 4.8 kb from the DSB (Fig. 1B and Ref. 37). If 5Ј resection has progressed beyond the RsaI recognition site, then the region would not be cleaved and would be amplified by PCR. The experimental design measures 5Ј resection at the DSB 6 h after HO induction. The initial kinetics of resection are highly dynamic, and our data do not capture these subtleties; however, the 6-h time point did allow for the accurate determi-nation of stabilized and consistent resection levels across all mutant combinations at both distances from the DSB. Consistent with previous reports and as a control for our experiments, hyper-resection was observed in cells lacking KU70 (Fig. 1, C and D, and Ref. 19). The level of resection in nej1-V338A mutants also increased at both distances from the DSB and was similar to the levels observed in nej1⌬ and lif1⌬ mutant cells (Fig. 1, C and D). The loss of LIF1 showed less of an impact at 4.8 kb compared with the loss of NEJ1; however, the reason(s) for this difference is currently not clear. Moreover, the increase in resection was not due to NHEJ mutant cells escaping into S phase as flow cytometry indicated all mutants arrested similarly to wild-type cells with Ͼ 90% in G 1 when resection was monitored (supplemental Table S1). Nej1 was previously shown to promote NHEJ by protecting DSB ends from resection (19), and our findings implicate the association of Nej1-Lif1 in end protection as well (6,8).
The strongest evidence that Nej1 contributes to Ku functionality comes from the increase in Rad51 levels at the DSB in nej1⌬ mutant cells (8). Previous reports showed that direct measurements of Ku were similar in wild type and nej1⌬ mutants by standard ChIP methods (8), and it was only when ChIP was performed under non-cross-linking conditions that a decrease in DNA-bound Ku was detected in cells lacking NEJ1. To address whether there is a Ku-independent role for Nej1 in resection, NEJ1 and KU70 were both disrupted. Resection was indistinguishable between ku70⌬ single and ku70⌬ nej1⌬ double mutant cells (Fig. 1, E and F), supporting a model whereby Nej1 does indeed prevent resection via a pathway epistatic with Ku.
To determine whether the mutant protein is properly recruited to the break, we performed ChIP at the DSB after HO induction. The level of Nej1 V338A -Myc recovery was similar to wild-type Nej1-Myc, suggesting that the resection defect with nej1-V338A is not attributable to failed recruitment but rather its functionality after localization ( Fig. 1G and supplemental Fig. S1A). In vitro work previously showed that Lif1 binding to Ku-bound DNA was not dependent on Nej1, but the interaction was enhanced with the addition of Nej1 (8). The importance of Nej1 for Lif1 recovery at a DSB in vivo has not been reported. Upon HO-induced DSB formation, there was no significant difference in the recovery of Lif1-HA in nej1⌬ mutants compared with wild-type cells, and there was no difference in Lif1-HA levels at the break in cells in which NEJ1 was deleted compared with the point mutant nej1-V338A ( Fig. 1H and supplemental Fig. S1B). However, compared with wild type, there was a ϳ50% decrease of Lif1-HA recovery in nej1-V338A mutant cells (Fig. 1H). We conclude that Nej1 is enriched at the DSB independently of its interaction with Lif1 but that through their association Nej1 may maximize Lif1 retention at the break.

MRX-dependent resection events and interactions between MRX and Nej1
In the absence of KU70, end protection is lost, and resection increases at both 0.15 and 4.8 kb from the DSB (6, 8) (Fig. 1, C  and D). The initial resection of 100 -300 nucleotides is regulated by MRX, and consistent with previous reports, we

Nej1 suppresses 5 hyper-resection at DSBs
observed that rad50⌬ mutant cells exhibited a decrease in resection that was restored back to wild-type levels in ku70⌬ rad50⌬ double mutants (Fig. 2, A and B, and Refs. 20, 22-25, 27, 35, and 36). However, when the deletion of RAD50 was combined with either nej1⌬ or nej1-V338A, the resection levels in double mutant cells remained indistinguishable from rad50⌬ (Fig. 2, A-D). A similar decrease in resection was also observed when rad50⌬ was combined with lif1⌬ ( Fig. 2, E and F). Consistent with previous literature, the level of resection in ku70⌬ rad50⌬ above that of rad50⌬ was Exo1-dependent (supplemental Fig. S2, A and B, and Ref. 20). Thus, there is a distinction when combining rad50⌬ with the loss of KU70 compared with the loss of factors that support Ku stability. The DNA ends of nej1⌬ rad50⌬ and nej1-V338A rad50⌬ mutant cells are not substrates for Exo1, suggesting that Ku still provides a level of DNA end protection when NEJ1 is deleted or when Nej1-Lif1 interactions are disrupted.
Ku and MRX are believed to function antagonistically in repair pathway choice (2,33). MRX promotes Ku removal from unrepaired DSBs (2,38,39), and although Ku stabilizes MRX at the break (6), Ku inhibits 5Ј end degradation by MRX. Interestingly, interactions between Lif1 and Xrs2 actually enhance the binding of core NHEJ factors to Ku-bound DNA (4,5,40). Previous work showed that the level of Rad50 at the break site decreased in cells lacking LIF1 (6). We wanted to investigate whether Nej1 influenced MRX binding at the DSB; therefore, we performed ChIP on Rad50-HA. In nej1-V338A mutant cells, Rad50-HA recovery was ϳ30% lower relative to wild type and nej1⌬, which were indistinguishable from one another ( Fig. 2G and supplemental Fig. S1B). In all, the presence of Nej1 V338A at the DSB results in a decrease in the association of Lif1 and MRX with the DSB that is statistically significant (Figs. 1, G and H, and 2G), but it is unclear why their recovery is decreased in nej1-V338A mutant cells but not in cells with the full NEJ1 deletion. . Error bars represent the standard error of three replicates. Significance was determined using one-tailed, unpaired Student's t test. All mutants were compared with wild-type cells (*, p Ͻ 0.05; **, p Ͻ 0.01). The enrichment of Nej1-13Myc (JC1687) and Nej1 V338A -13Myc (JC3160) (G) or Lif1-6HA in wild type (JC2665), nej1⌬ (JC2884), and nej1-V338A (JC3828) cells (H) was determined at 0.6 kb from the DSB. The -fold enrichment is normalized to recovery at the SMC2 locus. Error bars represent the standard deviation of three replicates. Significance was determined using one-tailed, unpaired Student's t test (*, p Ͻ 0.05). The non-tagged control wild type (JC727), nej1⌬ (JC1342), and nej1-V338A (JC2659) cells are shown in supplemental Fig. 1, A and B.

Nej1-Lif1 interactions inhibit Dna2-Sgs1-mediated resection at the DSB site
MRX activity is the central nuclease for initiating resection; however, two redundant nucleases, Exo1 and Dna2, also localize via distinct mechanisms to the DSB and function in longrange resection after MRX starts the process (25,26,41). Dna2 requires Sgs1 helicase; as such, the disruption of SGS1 has been used to characterize the role of Dna2-dependent 5Ј resection at the DSB because DNA2 is an essential gene involved in Okazaki fragment processing (26,(42)(43)(44). In line with previous reports and in contrast to rad50⌬ mutant cells, we observed no defects in the initiation of resection at the 0.15 kb site in sgs1⌬ or exo1⌬ single mutant cells (Fig. 3A).
We wanted to determine resection dynamics in RAD50 ϩ cells and the contribution of Exo1 and Dna2 nucleases to hyperresection at 0.15 kb in the nej1 mutants. Resection at 0.15 kb in exo1⌬ nej1-V338A double mutants was slightly higher than that measured in nej1-V338A single mutant cells (Fig. 3B). Similar trends were observed in double mutants when exo1⌬ was combined with nej1⌬ or ku70⌬ (Fig. 3C). In contrast, the hyper-resection levels at 0.15 kb in nej1-V338A, nej1⌬, and ku70⌬ single mutants were significantly reduced in all double mutants carrying sgs1⌬ and similar to the level measured in wild-type cells (Fig. 3, D and E). In all, these data suggest that Nej1 prevents unregulated resection 0.15 kb from the break that is mediated by Dna2-Sgs1 but not Exo1.

Long-range resection defects in nej1 mutants
We next wanted to understand whether the differences in short-range resection (0.15 kb) extended to long-range resection (4.8 kb). The levels of resection decreased at a distance 4.8 kb from the DSB in both sgs1⌬ and exo1⌬ single mutant cells, which is consistent with previous reports (Fig. 5A and Refs. 25, 26, and 41). Similar to resection levels at 0.15 kb, the loss of SGS1 partially reversed hyper-resection at 4.8 kb in nej1⌬ mutants to an intermediate level between nej1⌬ and sgs1⌬ single mutant cells and similar to wild-type cells (Fig. 5, B and C). The level of resection in nej1-V338A sgs1⌬ double mutant cells was indistinguishable from wild type and above the level measured in sgs1⌬ mutants (Fig. 5, B and C).
A different pattern emerged with the loss of EXO1 (Fig. 5,  D and E). The resection levels in nej1⌬ exo1⌬ and nej1-V338A exo1⌬ double mutant strains were reduced to less than 50% of wild-type levels and indistinguishable from exo1⌬ single mutants at the 4.8-kb site (Fig. 5, D and E). Long-range resection was decreased in all cells lacking EXO1 to below wild-type levels and in marked contrast to the levels Significance was determined using one-tailed, unpaired Student's t test where mutants were compared with wild-type cells for changes in resection levels (**, p Ͻ 0.01). G, ChIP assay as described in Fig. 1, E and F, was performed to measure the enrichment of Rad50-3HA at a DSB in wild type (JC3306), nej1⌬ (JC3307), and nej1-V338A cells (JC3347). Error bars represent the standard deviation of three replicates. Significance was determined using one-tailed, unpaired Student's t test. All mutants were compared with wild-type cells (*, p Ͻ 0.05).

Nej1 suppresses 5 hyper-resection at DSBs
of resection in these double mutants 0.15 kb from the DSB where hyper-resection was observed (Tables 1 and 2 and Fig.  3, B and C). Similar resection trends were observed when exo1⌬ or sgs1⌬ was combined with ku70⌬ (Fig. 5, C and E). In all, our data support Exo1 as the central nuclease for extended long-range resection.

Nej1-Lif1 prevents Sgs1-and Exo1-dependent large deletions during DSB repair
We next wanted to determine whether the interplay between Nej1 and the nucleases impacts genome stability specifically in the vicinity of a DSB where we had observed aberrant resection and in the same system where one DSB was generated by the HO endonuclease (29). Cells that survive continuous HO cutting (survivors) arise from an imprecise repair event where the HO recognition site becomes disrupted and no DSB is generated. The extent of genomic alterations during repair can be measured by determining the mating type of survivors. Two genes that regulate mating type, MAT␣1 and MAT␣2, are located adjacent to the HO-induced DSB (Figs. 1A and 6A). Their expression activates ␣-type genes and inhibits a-type   . Error bars represent standard error of three replicates. Significance was determined using one-tailed, paired Student's t test (ns, not significant; *, p Ͻ 0.05; **, p Ͻ 0.01); # denotes statistical significance as compared with wild type at p Ͻ 0.01.

Nej1 suppresses 5 hyper-resection at DSBs
genes. Conversely, large deletions (Ͼ700 bp) around the DSB where both ␣1 and ␣2 have been disrupted result in an a-like mating type (Fig. 6A). From three independent experiments, we determined the mating type of Ͼ300 independent survivors in each mutant background. The majority of repair events in wild-type survivors arise from a small 2-bp insertion that alters HO and expression of MAT␣1 (Fig. 6A), resulting in a yeast strain with a sterile mating type (29). Survivors with mating type ␣ result from mutations inactivating the HO endonuclease and not from mutations at the HO cut site (confirmed by sequencing the HO cut site of six nej1⌬ survivors) (29). We determine the mating types in all survivors ( Fig. 6B and supplemental Fig. S3, A-C) and were particularly interested in the level of large deletions (a-like survivors) in the various mutant backgrounds. Compared with wild type (1.3%), the percentage of NHEJ mutant survivors exhibiting an increase in large deletions (Ͼ700 bp) increased in ku70⌬ (9.0%), nej1⌬ (12.5%), and lif1⌬ (19.1%); however, this did not hold true for nej1-V338A (0.33%) (Fig.  6, B and C, and supplemental Fig. S3D). In rad50⌬ (5.1%), the percentage of survivors with large deletions was above that of wild type (Fig. 6D) but less than when other factors central to NHEJ were deleted. The level of large deletions in nej1-V338A rad50⌬ double mutant survivors was above the level of the rad50⌬ single mutant (Fig. 6D and Ref. (29) and similar to the level measured in cells lacking NEJ1 and nej1⌬ rad50⌬ ( Fig. 6D and supplemental Fig. S3A). In contrast, no large deletions were observed when either of the two longrange resection nucleases, SGS1 or EXO1, was disrupted (Fig. 6, E and F). Moreover, in all mutants containing sgs1⌬, we observed a decrease in the percentage of a-like survivors containing large genome deletions ( Fig. 6E and supplemental Fig. S3B). Similarly, all survivors in which EXO1 was further disrupted showed a decrease in the percentage of large deletions around the DSB (Fig. 6F and supplemental Fig.  S3C). Importantly, in nej1⌬ rad50⌬, yku70⌬ rad50⌬, and nej1-V338A rad50⌬ double mutants, the increase in large deletions was not overamplified as a result of reduced cell viability because the viability in triple mutants in which SGS1 or EXO1 was further disrupted still remained Ͻ2.5% of wild type even though the percentage of large deletion significantly decreased (Table 3 and Fig. 6, E and F). Taken together, our data support a model whereby the increase in large genomic deletions around a DSB is primarily a consequence of aberrant long-range nuclease activity.  and yku70⌬ sgs1⌬ (JC3850). Error bars represent standard error of three replicates. Significance was determined using one-tailed, unpaired Student's t test. Double mutants in B and D were compared with nej1 mutant cells, and double mutants in C and E were compared with sgs1⌬ or exo1⌬ single mutants, respectively (ns, not significant; *, p Ͻ 0.05; **, p Ͻ 0.01).

Nej1 suppresses 5 hyper-resection at DSBs
Discussion 5Ј-3Ј DNA end resection is a critical event in DSB repair. If resection is initiated, then NHEJ is no longer a repair pathway option. Sae2 binds the MRX complex, stimulating Mre11 to initiate resection, and Exo1 is the central nuclease in long-range resection. Dna2-Sgs1 serves as backup to both, and its regulation in DSB resection has remained somewhat obscure, particularly under conditions where MRX and Exo1 are present. Although Exo1 and Dna2-Sgs1 normally perform long-range resection, both can initiate resection at the break site in the absence of RAD50, albeit with delayed kinetics and low efficiency (26). Nucleases are under CDK1 and cell cycle control. In S phase, CDK1 activity increases, and Sae2 and MRX are phosphorylated to initiate resection. This regulatory step promotes HDR-mediated repair during a cell cycle stage when a homolog is present for recombination (35, 46 -48). In G 1 when CDK1 activity is low or when CDK1 is knocked down, resection is heavily reduced (35, 46 -48). It is under these conditions in G 1 when NHEJ is the major repair pathway that we observed the impact of Nej1 on 5Ј resection.

Nej1 suppresses 5 hyper-resection at DSBs
very high affinity for DNA ends not protected by Ku; however, the deletion of EXO1 did not restore resection to wild type like the loss of SGS1. In contrast to Exo1, Dna2-Sgs1 can initiate resection in the presence of Ku (19), and in nej1 mutants, our data suggest that a level of Ku remains bound to the DNA ends ( Fig. 2A). The physical presence of Nej1 and its integrity at the DSB are important to inhibit Dna2 activity, and this, by extension, is another way that Nej1 promotes NHEJ.
In RAD50 ϩ cells, deletion of either SGS1 or EXO1 reversed hyper-resection 4.8 kb from the break site in nej1-V338A mutant cells (Fig. 5, B-E, and Table 2). In all mutant combinations where EXO1 was deleted, there was a marked reduction in long-range resection to below the levels in wild-type cells (Fig.  5, D and E). In nej1-V338A sgs1⌬ and nej1⌬ sgs1⌬ double mutant cells, Exo1 is the functional nuclease 4.8 kb from the DSB, and resection was similar to levels in wild-type cells. Alternatively, in nej1-V338A exo1⌬ and nej1⌬ exo1⌬ double mutant cells where Dna2-Sgs1 is the only functional long-range nuclease available, resection levels were reduced to less than half that measured in wild-type cells (Fig. 5, D and E, and Table  2). These data demonstrate that Exo1 is the primary nuclease 4.8 kb from the break with Dna2-Sgs1 being less efficient. However, the level of resection at 4.8 kb was not merely an extension of resection dynamics closer to the break site in cells in which EXO1 was deleted. For example, in nej1-V338A exo1⌬ and nej1⌬ exo1⌬ double mutant cells, resection levels at 0.15 kb remained elevated above wild type and similar to levels in nej1-V338A and nej1⌬ single mutants (Tables 1 and 2). Taken together, our data support a model where in an otherwise wildtype background Dna2-Sgs1 has robust nuclease activity 0.15 kb from the DSB and must be actively inhibited by the core NHEJ factors, including the Nej1-Lif1 interaction, but Exo1 is less efficient and most active in the absence of KU and MRX.
Despite the marked similarities in resection between nej1⌬ and nej1-V338A (Figs. 2-4), differences were observed depending on the presence or absence of EXO1 together with mre11-3. For example, resection at both 0.15 and 4.8 kb in nej1⌬ mre11-3 double mutant cells was below wild type; however, in nej1-V338A mre11-3 mutants, hyper-resection was observed at both distances (Tables 1 and 2). In contrast, with the further loss of EXO1, nej1-V338A exo1⌬ mre11-3 triple mutant cells had a defect in resection at 0.15 and 4.8kb, and the nej1⌬ exo1⌬ mre11-3 triple mutant cells exhibited hyper-resection at 0.15 kb but a resection defect at 4.8 kb (Tables 1 and 2). One contributing factor might involve the decrease in Lif1 and Rad50 associated with the DSB in nej1-V338A but not nej1⌬ mutant cells (Figs. 1H and 2G). This might result in an overall "null-like" phenotype for MRX functionality at the break site. Indeed, nej1-V338A exo1⌬ mre11-3 and nej1-V338A exo1⌬ rad50⌬ triple mutant cells both show similar defects in resection at 0.15 kb (Table 1). However, the situation is likely more complex, involving the dynamic interplay of factors at the DSB, because in EXO1 ϩ cells hyper-resection was observed in nej1-V338A mre11-3 but not nej1-V338A rad50⌬ (Table 1).
MRX stabilizes Dna2-Sgs1 recruitment to the break through direct interactions (19,49). When Sgs1 was the sole nuclease present, resection at 0.15 kb was reduced in exo1⌬ mre11-3 and exo1⌬ rad50⌬ double mutants compared with wild-type cells (Table 1). When considering the disruption of NEJ1, the level of resection was restored to near wild-type levels in nej1⌬ exo1⌬ mre11-3 triple mutant cells but not nej1⌬ exo1⌬ rad50⌬ mutants, revealing that Nej1 integrity is important to inhibit Dna2-Sgs1 functionality at the break site. Furthermore, these data also confirm that, although its nuclease activity is nonfunctional, the physical presence of MRX at the DSB is impor-  (2,36). Ku has high affinity for DNA ends and recruits Nej1 and Lif1-Dnl4 to the DSB to promote NHEJ repair (2)(3)(4)(5)(6). Lif1 also interacts with Xrs2 of the MRX complex, stabilizing the complex at the DSB (4 -6, 42). MRX is also required for the recruitment of Dna2-Sgs1 to the DSB (21). Ku binds DNA ends to protect them from 5Ј DNA resection initiation mediated by the short-range nuclease Mre11 and subsequent resection elongation by the long-range nucleases Exo1 and Dna2-Sgs1, which have functionally redundant roles. MRX is the central short-range nuclease, and Exo1 is the central long-range nuclease with Dna2-Sgs1 performing a backup function. B, our data indicate that the interaction between Lif1 and Nej1 is critical to inhibit resection initiation by the Dna2 nuclease at DSBs. In nej1-V338A mutant cells, Dna2-Sgs1 is misregulated, and 5Ј hyper-resection ensues. The regulation of Dna2-Sgs1 at the DSB is dependent on Ku, but not solely through DNA end protection. All observations with nej1⌬ mutants were epistatic with yku70⌬ mutants. Our results provide insight to previous observations that Nej1 and Lif1 are believed to enhance Ku stability at unresected DNA ends and that direct measurements of Ku were similar in wild-type and nej1⌬ mutant cells, but there was an increase in Rad51 levels at the DSB in nej1⌬ cells (8).

Nej1 suppresses 5 hyper-resection at DSBs
tant to improve the efficiency of resection initiation by Dna2-Sgs1, likely through its stabilizing effects on Dna2-Sgs1 (19,49). As discussed above, the level of resection at 0.15 kb in nej1⌬ exo1⌬ mre11-3 triple mutant cells was not maintained 4.8 kb from the break ( Table 2), suggesting that the efficiency of Dna2-Sgs1 decreases between 0.15 and 4.8 kb. The regulation of Dna2-Sgs1 requires further characterization; however, these results underscore the importance of Exo1 in resection at a distance 4.8 kb from the break even when resection is initiated by Dna2-Sgs1. For cells to survive chronic HO induction (DSB formation), the break must be repaired imprecisely to prevent further cleavage. This can occur through imprecise NHEJ, MMEJ, or SSA, which requires long regions of homology and results in the loss of a large amount of genetic material during repair. Loss of NEJ1 resulted in an increase in a-like survivors, which exhibit repair with large deletions of Ͼ700 bp. The most common type of survivors in cells with the nej1-V338A mutation were similar to wild-type and were repaired with small insertions or deletions (Fig. 6B). This may be explained by the fact that Dnl4 ligase activity remains proficient in nej1-V338A cells (16); thus, the vast majority of survivors repair quickly via error-prone direct ligation through Dnl4 rather than an alternative pathway such as SSA. In addition to mating type determination, we also used PCR to confirm that the few nej1-V338A survivors with an a-like mating type contained large deletions (supplemental Fig.  S3D). Additionally, upon the further deletion of RAD50 in nej1-V338A mutant cells, an increase in a-like survivors (large deletions) was observed at levels found in nej1⌬ and nej1⌬ rad50⌬ mutant cells (Fig. 6D). The difference between nej1-V338A and nej1-V338A rad50⌬ could be attributed to DNA end tethering by MRX (3). One model is that the untethered DNA ends would be unable to repair efficiently via direct ligation, resulting in an increase in repair via SSA and large genomic loss from an overall increase in end resection by the time DSB ends rejoin (50,51). Lastly, we characterized whether Sgs1 and Exo1 activity contributed to the formation of large genomic deletions during repair. The number of a-like survivors was reduced in mutant cells when either nuclease was deleted (Fig. 6, E and F). Thus, the Nej1-Lif1 interaction and the stable association of NHEJ factors at the DSB promote genome stability and prevent large genomic rearrangements in part by inhibiting Dna2-Sgs1-dependent hyper-resection at the break site.

Yeast strains and media
All strains used in this study are listed in supplemental Table  S2. For experiments involving the induction of an HO DSB, cells were grown in YPLGg medium containing 1% yeast extract, 2% Bacto peptone, 2% lactic acid, 3% glycerol, and 0.05% glucose. For the mating type assays, the YPA plates (1% yeast extract, 2% Bacto peptone, and 0.0025% adenine) were supplemented with 2% glucose or 2% galactose (16).

Mating type assay
Mating type assays were performed as described previously (29). Surviving colonies from continuous growth on YPA ϩ 2% Gal for 3-4 days were restreaked onto YPA ϩ 2% Glc plates and incubated at 30°C for 1-2 days. Each plate was replica-plated onto fresh YPA ϩ 2% Glc plates with both JC158 and JC159 to test for ␣ and a-like survivors, respectively. These plates were incubated at 30°C overnight and then replica-plated onto minimal media (MIN) plates containing 2% glucose and 0.6% yeast extract and incubated at 30°C for 1 day. Colonies growing on the JC158 plate are "␣" survivors, colonies growing on the JC159 plate are "a-like" survivors, and colonies that do not grow on either plate are "sterile" survivors.

Chromatin immunoprecipitation
ChIPs were performed as described (16). Chromatin fractionation was performed by spinning the cell lysate at 13,200 rpm for 15 min after which the pellet was resuspended in lysis buffer and sonicated to yield DNA fragments ϳ500 bp in length. The sonicated lysate was then incubated with beads ϩ anti-HA (12CA5) or anti-Myc (9E10) antibody or unconjugated beads (control) for 2 h at 4°C. Quantitative PCR was performed using the Applied Biosystems QuantStudio 6 Flex machine. The reaction contained PerfeCTa qPCR Supermix (Quanta Biosciences Inc.), primers, and 5Ј FAM-labeled/3Ј TAMRAlabeled probes specific to 0.6 kb away from the cut site (HO2), the cut site to measure cutting efficiency (HO6), or the SMC2 control region. Enrichment was calculated relative to SMC2 and corrected for cut efficiency, which was determined from the loss of PCR product at HO6 after 3-h induction compared with time 0.

qPCR-based resection assay
The resection assay was performed as described (37). This method was previously verified by Southern blotting with ssDNA-binding probes on both sides of the DSB (37). Cells were grown overnight to 1 ϫ 10 7 cells/ml in YPLG. Cells were pelleted and resuspended in fresh YPLG Ϯ 2% galactose and ␣-factor to maintain G 1 arrest as verified by flow cytometry. At the t ϭ 0 and t ϭ 6-h time points, genomic DNA was purified using standard genomic preparation methods and resuspended in 100 of ml double distilled H 2 O. Genomic DNA was treated with 0.005 g/l RNase A (Sigma) for 45 min at 37°C. 2 l of DNA was added to tubes containing Cut Smart buffer with or without RsaI restriction enzyme (New England Biolabs) and incubated at 37°C for 2 h. Quantitative PCR was performed using the Applied Biosystems QuantStudio 6 Flex machine. PowerUp SYBR Green Master Mix (Applied Biosystems) was used to quantify resection 0.15 and 4.8 kb from the DSB, and Pre1 was used as a negative control. RsaI-cut DNA was normalized to uncut DNA as described previously to quantify the percentage of ssDNA/total DNA (37). Primers and probes are in supplemental Table S3. The primers are located at optimal sites where resection via Mre11 and the delayed effects of Dna2-

Nej1 suppresses 5 hyper-resection at DSBs
Sgs1 and Exo1 on close-range resection can be distinguished from their effects on long-range resection.
Author contributions-K. S. S. designed, performed, and analyzed the experiments and wrote the paper. B. L. M. constructed mutant strains and helped in the design of the experiments. S. P. L.-M. contributed to the interpretation of the data and the preparation of the manuscript. J. A. C. conceived and coordinated the study, analyzed the experiments, and wrote the paper. All authors reviewed the results and approved the final version of the manuscript.