Asymmetrical recognition and processing of double-strand breaks formed during DNA replication

Significance

DNA end resection is a critical early step in the repair of DNA double-strand breaks (DSB) by homologous recombination (HR). Studies using endonucleases that generate DSBs in all phases of the cell cycle have shown that end resection proceeds by the sequential action of the Mre11 complex and long-range resection nucleases, Exo1 or Dna2-Sgs1. In this study, we restricted DSB generation to S phase and found that the genetic requirements for end resection at more physiologically relevant DSBs are different to those at endonuclease-induced DSBs. The asymmetrical recognition and processing of replication-dependent DSBs may also explain the bias toward HR-mediated repair over nonhomologous end joining.

Abstract

DNA end resection to generate 3′ single-stranded DNA (ssDNA) overhangs is the first step in homology-directed mechanisms of double-strand break (DSB) repair. While end resection has been extensively studied in the repair of endonuclease-induced DSBs, little is known about how resection proceeds at DSBs generated during DNA replication. We previously established a system to generate replication-dependent double-ended DSBs at the sites of nicks induced by the Cas9^D10A nickase in the budding yeast genome. Here, we suggest that these DSB ends form in an asymmetric manner, with one being blunt or near blunt and the other bearing a 3′ ssDNA overhang of up to the size of an Okazaki fragment. We find that Mre11 preferentially binds blunt ends and is required to evict Ku from these DSB ends and to promote end resection. In contrast, the ends predicted to have 3′ overhangs have minimal Ku binding, and resection at these break ends can proceed in a mostly Mre11-independent manner through either the Exo1 or Dna2-Sgs1 long-range resection mechanisms. These findings indicate that resection proceeds differently at replication-dependent DSBs than at canonical DSBs and reveal that Ku selectively binds blunt ends, potentially explaining why replication-dependent DSBs are poorly repaired by nonhomologous end joining.

DNA double-strand breaks (DSBs) are cytotoxic DNA lesions that must be repaired to maintain genome integrity. DSBs can be generated directly using site-specific endonucleases, such as I-SceI, HO, or AsiSI (1, 2), or produced randomly in the genome, for example, by ionizing radiation or use of drugs that trap Topoisomerase II cleavage complexes (3, 4). Studies using endonucleases have provided a general framework for how DSBs are recognized and repaired (5, 6). These studies suggest that cells initially attempt repair by nonhomologous end joining (NHEJ), but if the DSB ends are not compatible for ligation, and cells are in the appropriate cell cycle phase, they undergo end resection to repair by homologous recombination (HR) (7–9). End resection is initiated by the Mre11-Rad50-Xrs2/Nbs1 (MRX in Saccharomyces cerevisiae, MRN in other eukaryotes) complex with its cofactor, phosphorylated Sae2, to form short 3′ overhangs (10). Overhangs generated by MRX are then extended by the long-range resection machinery, including Exo1 and the Dna2-Sgs1-Top3-Rmi1 ensemble (referred to as Dna2-Sgs1 hereafter) (10). In budding yeast, the long-range nucleases can directly initiate resection of DSBs lacking a covalent modification, bypassing the requirement for the Mre11 nuclease (10). Consequently, Mre11 nuclease-deficient mutants exhibit only modest sensitivity to DNA-damaging agents and a slight delay in resection initiation at endonuclease-induced DSBs but are completely deficient for processing Spo11-bound ends during meiosis (11, 12). The resulting single-stranded DNA (ssDNA) is initially bound by RPA and then by Rad51 to promote strand exchange with an intact donor sequence, followed by repair synthesis to replace sequences lost by end resection (13).

While studies with site-specific endonucleases have been invaluable in deciphering DSB repair pathways (5), it is increasingly recognized that endonuclease-induced DSBs do not accurately mimic DNA damage generated during DNA replication. Given that replication-dependent DNA damage is predicted to be quite common and a driver of genome instability (14, 15), further research is needed to fully understand DNA repair processes in the context of DNA replication. Single-stranded breaks (SSBs), also called nicks, are common intermediates in DNA metabolic processes and have the potential to be converted to DSBs by replication fork progression (14, 16). Indeed, some cancer therapeutics act by inducing or stabilizing SSBs resulting in the formation of cytotoxic DSBs during S phase (17–19). An in vitro study using Xenopus egg extracts and prenicked plasmid templates demonstrated formation of a single-ended DSBs (seDSBs) by replisome run off at the nick, regardless of whether the nick was on the leading or lagging-strand template (20). However, the end structure of the seDSB differed for leading and lagging-strand collapses. Replisome run off at a nick on the leading-strand template generated an end with a 3 nucleotide (nt) 5′ overhang, referred to here as a blunt end, whereas the seDSB produced by a lagging-strand collapse had a 3′ overhang of ~70 nt (20).

Recent work has demonstrated the utility of modified CRISPR/Cas9 enzymes for generation of strand-specific nicks that can be converted to DSBs during replication (21–26). Cas9 makes a DSB through the cooperative action of two endonuclease domains, whereas Cas9^D10A harbors a mutation in one of these two domains (27) rendering the enzyme able to cut only one strand of the DNA duplex. In human cells, nicks generated by Cas9^D10A on the leading-strand template were shown to convert to mostly seDSBs during S phase, while nicks on the lagging-strand template resulted in double-ended DSBs (deDSBs) (21). Formation of a deDSB was shown to result from continued translocation of the CMG replicative helicase on the leading-strand template, bypassing the nick on the lagging-strand template (21). An Okazaki fragment initiated after replisome bypass would be predicted to terminate at the nick, yielding a deDSB (SI Appendix, Fig. S1A). In budding and fission yeasts, nicks formed by Cas9^D10A, M13 gpII nickase, or a minimal HO cut site that is preferentially nicked, generate mostly deDSBs during replication, even for nicks generated on the leading-strand template (22, 23, 28). In the case of the Flp-nick system (29), there are conflicting reports on whether single end or deDSBs are formed (23, 30–32). The deDSB signal detected for leading-strand template nicks is thought to be due to convergence of two opposing forks at the nick. Such a scenario might be more likely in organisms such as budding yeast with a high density of active and dormant origins (33).

Genetic analysis in S. cerevisiae revealed that Cas9^D10A-induced replication-coupled DSBs are repaired by Rad51-dependent sister-chromatid recombination with no apparent contribution by NHEJ (22). This system therefore provides the opportunity to study sister-chromatid recombination in a replication context, potentially revealing mechanistic aspects of HR-mediated repair that have gone undetected in studies of canonical DSB repair using endonucleases such as HO or I-SceI. In this study, we address how the ends of replication-dependent DSBs are recognized and repaired. Since work using the Xenopus egg extract system showed different end structures for seDSBs produced at nicks on leading and lagging-strand templates (20), we investigated whether the ends of a replication-dependent deDSB are asymmetrical, consisting of one blunt end and the other with a 3′ overhang. We find that resection at blunt ends is dependent on Mre11. In contrast, resection at 3′ overhang ends can proceed in a mostly Mre11-independent manner by either Exo1 or Dna2-Sgs1-catalyzed long-range resection. We also find that the Ku complex is recruited to blunt ends of replication-dependent DSBs and that the removal of the Ku complex by MRX is required for repair of breaks. Taken together, these results paint a more nuanced picture of resection dynamics at breaks formed during replication and highlight the significance of the structural asymmetry of these breaks.

Results

MRX Nuclease Activity Is Dispensable for Repair of Replication-Dependent DSBs.

To study breaks derived from Cas9^D10A-initiated nicks, we used a set of four guide RNAs (gRNA) flanking ARS607, a highly efficient, early-firing origin on chromosome VI of S. cerevisiae (Fig. 1A) (22). gRNAs 1 and 20 generate nicks on the lagging-strand template, while gRNAs 6 and 14 generate nicks on the leading-strand template when replication originates from ARS607 (Fig. 1A) (22). In previous work, we found that targeting the nickase with gRNAs 1, 6, and 20 results in replication-dependent deDSBs, while targeting the nickase with gRNA14 generates a mostly seDSB signal (SI Appendix, Fig. S1A) (22). Since HR-deficient mutants were previously reported to show the highest sensitivity to gRNA1 and gRNA6 (22), we focused on these gRNAs for most assays in this study. We first verified that each member of the MRX complex is required for cell viability to nicks produced on leading and lagging-strand templates (Fig. 1B and SI Appendix, Fig. S1B). In contrast, the mre11-H125N allele, in which the MRX complex lacks nuclease activity (12), conferred only mild sensitivity to gRNA1 and no sensitivity to gRNA6 (Fig. 1C). We also tested the involvement of Sae2 and Tel1, which stimulate MRX resection initiation and signal the DNA damage checkpoint via MRX (34), respectively (Fig. 1B). However, neither of these mutants displayed detectable nickase sensitivity (Fig. 1C). These results demonstrate that MRX nuclease activity is not per se required for replication-dependent DSB repair, in agreement with past work in S. cerevisiae studying repair of DSB ends lacking covalently bound proteins (12).

Fig. 1.

MRX nuclease activity is redundant with Sgs1 for replication-dependent DSB repair. (A) Chromosomal location of the main gRNAs used to generate replication-dependent DSBs in this study. gRNA1 and gRNA20 generate nicks on the lagging-strand template, while gRNA6 and gRNA14 generate leading-strand template nicks with Cas9^D10A. Parental strands are depicted as black lines, nascent leading strands are shown in red, and nascent lagging strands are shown in blue. (B) Schematic of the structure and roles of the MRX complex highlighting the ability of the complex to initiate end resection in an Sae2-dependent manner and to promote Tel1 signaling (Top). Schematic for DNA end resection showing one side of a DSB (indicated by a lightning bolt). MRX, stimulated by phosphorylated Sae2 incises the 5′-terminated strand near a DSB to create the entry site for bidirectional resection using the Mre11 3′-5′ exonuclease activity and either Exo1 or Dna2-Sgs1 nucleases (Bottom). (C and D) 10-fold serial dilutions of the indicated strains expressing gRNA1 or gRNA6 were spotted on medium ± β-estradiol to induce Cas9^D10A expression.

To further define the mechanism for resection of replication-dependent DSBs, we generated double mutants with mre11-H125N and either exo1∆ or sgs1∆. Because the MRX complex is known to physically recruit Dna2-Sgs1 to DSBs in addition to facilitating short-range resection (35–37), we reasoned that the mre11-H125N sgs1∆ mutant would be highly sensitive to nickase induction. We found this to be the case for breaks generated by gRNAs 1, 6, 14, and 20 (Fig. 1D and SI Appendix, Fig. S1C), while the mre11-H125N exo1∆ mutant showed sensitivity comparable to the mre11-H125N single mutant (Fig. 1D and SI Appendix, Fig. S1C). Curiously, the mre11-H125N and mre11-H125N exo1∆ mutants both seem to be slightly more sensitive to nicks on the lagging-strand rather than the leading-strand template (Fig. 1D and SI Appendix, Fig. S1C). These genetic assays support a previous study reporting the requirement for Dna2-Sgs1-dependent resection in the absence of MRX nuclease activity for repair of ionizing radiation-induced DSBs (38). Additionally, we verified that the single deletions of exo1∆ or sgs1∆ do not confer nickase sensitivity, whereas loss of long-range end resection renders cells sensitive to nickase induction (Fig. 1D and SI Appendix, Fig. S1 B and C), as previously reported (22, 39, 40).

Replication-Dependent DSBs Are Likely Asymmetrical.

We next wished to determine whether the structure of replication-dependent DSBs is changed in the absence of Mre11. We expressed Cas9^D10A with gRNA6 or gRNA20 in asynchronous populations of wild-type (WT) or mre11∆ cells and analyzed genomic DNA by restriction endonuclease digestion followed by Southern blot analysis (Fig. 2A). As reported previously, these gRNAs generate a mostly deDSB signal in WT cells (22). Surprisingly, mre11∆ cells appear to generate seDSBs for both targeted sites when cut with Cas9^D10A, but with the opposite polarity (Fig. 2 B and C). Given that the Cas9^D10A nicks introduced with these gRNAs are only 7 nt apart on opposite strands, they should both be replicated by a replication fork initiated from ARS607. For both gRNAs, the cut fragment bearing an end predicted to be blunt is persistent in mre11∆ cells (Fig. 2B and SI Appendix, Fig. S1A). Notably, the cut fragments predicted to have 3′ overhang ends are present at similar levels in WT and mre11∆ cells (Fig. 2 B and C). The same trend of a predominantly seDSB signal but with opposite polarity was observed when targeting Cas9^D10A with gRNA1 and gRNA14 in mre11∆ cells (SI Appendix, Fig. S2 A–C). Since the role of MRX is to generate 3′ overhang ends for Exo1 and/or Dna2-Sgs1 to further resect, the most plausible explanation for this result is that resection proceeds in an MRX-independent manner at the end predicted to have a 3′ overhang (16, 20, 22). As a control, we expressed Cas9 with the same gRNAs in mre11∆ cells and observed the expected persistent two-ended DSB pattern using the same restriction endonuclease digests and hybridization probes (Fig. 2 B and C SI Appendix, Fig. S2 B and C).

Fig. 2.

Physical analysis of replication-dependent DSBs in WT and mre11∆ cells. (A) Schematic showing the location of ARS607, gRNAs, and the sizes of DNA fragments from EcoNI digestion of genomic DNA. The 6.2 kb fragment is generated by fork collapse at the nick and 3.7 kb fragment from fork progression beyond the nick or fork convergence from a downstream origin. The 1 kb hybridization probe, indicated by the purple line, was designed to hybridize equally with each break fragment. (B) Southern blots of EcoNI-digested DNA before induction and at different time points following expression of Cas9^D10A with gRNA20 or gRNA6 in WT and mre11∆ cells and following induction of Cas9 in mre11∆ cells. The 9.9 kb fragment corresponds to uncut DNA. (C) Quantification of the DSB fragments from the Southern blots in B.

MRX Activity Is Required at Blunt DSB Ends.

To directly test this hypothesis, we performed a qPCR-based resection assay measuring the ssDNA fraction in proximity to break sites (SI Appendix, Fig. S3A) (22, 41). For both gRNA1 and gRNA6, we assessed the ssDNA fraction about 1 kb centromeric or telomeric from each DSB end (labeled CEN and TEL). We strategically measured long-range resection to avoid detecting ssDNA close to the break sites that may result from replication-mediated 3′ overhangs. We first verified that Cas9^D10A cutting efficiency is comparable and highly efficient 4 h after nickase induction for all genotypes tested by using a qPCR assay with an amplicon spanning the cut site (SI Appendix, Fig. S3 B and C). To validate that the measured cut fraction was mostly the result of replication-dependent DSBs and not nicks, we arrested cells in G1 and then quantified Cas9^D10A-induced nicks by their sensitivity to S1 nuclease (SI Appendix, Fig. S3D) (21). We found that prior to S1 nuclease treatment to convert nicks to DSBs, the cut fraction was negligible 2 h after nickase induction (SI Appendix, Fig. S3F). In contrast, treating cells with S1 nuclease resulted in 17% and 24% cleavage products for the gRNA1 and gRNA6 sites, respectively (SI Appendix, Fig. S3F). This suggests that the cut fraction in cycling cells is mostly composed of replication-dependent breaks, since the cut fraction reached 59% at the gRNA1 locus and 86% at the gRNA6 locus 2 h after nickase induction in WT cycling cells (SI Appendix, Fig. S3 B and C).

As previously reported (21, 22, 40), rad51∆ cells display a hyperresection phenotype at replication-dependent DSBs and were therefore used as controls to detect the maximum ssDNA fraction at the tested loci. In contrast, exo1∆ sgs1∆ cells were used as a resection-deficient control, as they are expected to have very minimal resection at sites roughly 1 kb distal to a DSB (42, 43). WT cells have intermediate levels of ssDNA because resection intermediates are transient in repair-competent cells (21, 22, 40). We found that blunt ends are processed in an MRX and Exo1/Dna2-Sgs1-dependent manner (Fig. 3 A and C). In contrast, mre11∆ cells displayed a much smaller resection defect compared to exo1∆ sgs1∆ cells at ends predicted to bear 3′ overhangs (Fig. 3 A and C). While the ssDNA fraction in mre11∆ is lower than rad51∆ cells at 3′ overhangs, it is apparent that resection can proceed in a mostly MRX-independent manner. Since MRX is also known to physically recruit Dna2-Sgs1 and Exo1 to DSBs (37, 44), it is possible that the slight reduction in ssDNA in mre11∆ cells at ends with 3′ overhangs compared to rad51∆ levels is due to decreased activity of the long-range resection pathways. We also monitored resection at the gRNA6 locus after induction of a Cas9-induced DSB. As expected, Cas9 cutting at this locus is highly efficient (SI Appendix, Fig. S4). We found that Mre11 is essential for resection at both DSB ends, consistent with a requirement for MRX to process blunt or nearly blunt ends (SI Appendix, Fig. S4B). Together, these data further support the model that replication-dependent DSBs are formed asymmetrically because of replication fork dynamics and demonstrate that resection can occur in an MRX-independent fashion at 3′ overhang ends of these breaks.

Fig. 3.

Mre11 preferentially binds to and is required for resection at blunt ends. (A) Fraction of ssDNA generated at sites approximately 1 kb centromeric and telomeric to the gRNA1 cut site 4 h after Cas9^D10A induction assessed with a qPCR-based assay. Significance values are relative to the mre11∆ condition using an unpaired t test, ns P ≥ 0.05, *P ≤ 0.05, **P ≤ 0.01. (B) Schematic of break orientation at the gRNA1 locus (Top), enrichment of Mre11 flanking the gRNA1 locus (Bottom) determined by ChIP-qPCR. Amplicons to detect Mre11 binding are located 0.1 to 0.3 kb centromeric or telomeric to the gRNA1 cut site. Significance values compare Mre11 binding to the blunt or 3′ overhang end using an unpaired t test, ns P ≥ 0.05, *P ≤ 0.05. (C) Fraction of ssDNA generated at sites about 1 kb centromeric and telomeric to the gRNA6 cut site 4 h after Cas9^D10A induction assessed with a qPCR-based assay. Significance values are relative to the mre11∆ condition using an unpaired t test, ns P ≥ 0.05, *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001. (D) Schematic of break orientation at the gRNA6 locus (Top), enrichment of Mre11 flanking the gRNA6 locus (Bottom) determined by ChIP-qPCR. Amplicons to detect Mre11 binding are located 0.2 to 0.4 kb centromeric or telomeric to the gRNA6 cut site. Significance values compare Mre11 binding to the blunt or 3′ overhang end using an unpaired t test, *P ≤ 0.05, **P ≤ 0.01.

To further study the role of MRX in initiating resection, we measured Mre11 binding at replication-dependent DSBs by chromatin immunoprecipitation (ChIP)-qPCR in cycling cells following Cas9^D10A induction. We found Mre11 to be more strongly recruited to the ends that are predicted to be blunt than to 3′ overhang ends with gRNA1 and gRNA6 (Fig. 3 B and D). The higher retention of Mre11 at the gRNA6 site could be due to a delay between fork collapse at the leading-strand template nick and arrival of a converging fork to convert the seDSB into a deDSB (SI Appendix, Fig. S1A). In contrast, Mre11 is recruited equally to both sides of a Cas9-induced DSB (SI Appendix, Fig. S5A). We verified the asymmetrical binding of Mre11 at the gRNA20 locus targeting Cas9^D10A to the lagging-strand template (SI Appendix, Fig. S5B). We previously reported that a Cas9^D10A-induced nick is mostly converted to a seDSB using gRNA14 (22), and consistent with this finding, we observed enrichment of Mre11 only at the end produced by fork collapse (SI Appendix, Fig. S5C). Nonetheless, the data with the other three gRNAs that generate deDSBs support the notion MRX binding is biased to blunt ends, consistent with end resection being dependent on MRX at these ends.

Exo1 and Dna2-Sgs1 Can Process 3′ Overhangs in the Absence of MRX Activity.

Finally, we assessed resection dynamics at breaks formed by lagging and leading-strand template nicks in resection-deficient strains to dissect the relative role of the Exo1 and Dna2-Sgs1 resection pathways. At the blunt ends of gRNA1 and gRNA6-generated breaks, resection can proceed efficiently in the mre11-H125N strain (Fig. 4 A and B). Upon deletion of EXO1 in the mre11-H125N background, we observed a modest reduction in resection (Fig. 4 A and B). This suggests that the Exo1 pathway plays a minor role in the resection of these breaks, and resection is mostly dependent on Dna2-Sgs1 in the absence of MRX nuclease activity. Indeed, the mre11-H125N sgs1∆ mutant is almost as resection deficient as the mre11∆ strain (Fig. 4 A and B).

Fig. 4.

Exo1 and Dna2-Sgs1 can resect DSB ends with 3′ overhangs in the absence of MRX nuclease activity. (A) Fraction of ssDNA generated at sites approximately 1 kb centromeric and 1 kb telomeric to the gRNA1 cut site 4 h after Cas9^D10A induction assessed by a qPCR-based assay. (B) Fraction of ssDNA generated at sites approximately 1 kb centromeric and telomeric to the gRNA6 cut site 4 h after Cas9^D10A induction assessed by a qPCR-based assay. Significance values relative to the mre11∆ condition are shown with black asterisks using an unpaired t test, ns P ≥ 0.05, *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001. Significant differences between mre11-H125N derivatives are marked with blue asterisks.

In contrast to blunt ends, we observed either a slight increase in resection or an insignificant decrease in resection at 3′ overhang ends in the mre11-H125N exo1Δ and mre11-H125N sgs1Δ double mutants compared to the mre11∆ control (Fig. 4 A and B). The most resection-deficient mutant in this assay was mre11-H125N exo1∆ at the gRNA6 locus; however, the ssDNA fraction in this mutant was still more than twofold higher than in the exo1∆ sgs1∆ mutant (16.4% vs. 7.8%). The resection profiles from the 3′ overhang ends at both guide sites indicate that in the absence of MRX nuclease activity, either the Exo1 or Dna2-Sgs1 pathway can initiate resection. In contrast, Dna2-Sgs1 is required for the processing of blunt ends in the absence of MRX nuclease activity with only a minor contribution by Exo1.

MRX Activity Is Required for Removal of Ku at Blunt Ends of Replication-Dependent DSBs.

Previous studies have shown that removal of Ku partially suppresses the DNA damage sensitivity and end resection defect of Mre11-deficient cells in an Exo1-dependent manner (37, 38, 45, 46). Furthermore, overexpression of Exo1 suppresses the DNA damage sensitivity of the mre11∆ mutant (47–49). To increase Exo1 expression, we used a previously constructed high-copy number plasmid with EXO1 expressed from its native promoter (38). We found that overexpressing Exo1 rescued the sensitivity of mre11∆ cells to Cas9^D10A-induced nicking with gRNA1 by ~100-fold compared to the empty vector control (Fig. 5A). Surprisingly, Exo1 overexpression was less effective in suppressing the sensitivity mre11∆ cells to induction of Cas9^D10A with gRNA6 (Fig. 5A). This finding suggests that leading strand collapses are processed in a distinct way from DSBs resulting from a lagging-strand template nick. To confirm this finding, we also measured the viability of mre11∆ and mre11∆ cells ± Exo1 overexpression in response to gRNA14 or gRNA20 expression with Cas9^D10A. These guides confer less of a growth defect to mre11∆ cells than gRNAs 1 and gRNA6; however, we still found that the lagging-strand DSB showed less of a dependence on MRX for repair under Exo1 overexpression conditions when compared to the leading strand collapse (SI Appendix, Fig. S6A). Similarly to Exo1 overexpression, loss of Ku resulted in a stronger suppression of mre11∆ sensitivity to gRNA1 than to gRNA6 (Fig. 5B).

Fig. 5.

Lagging-strand breaks are more accessible to Exo1. (A) Fraction of mre11∆ and mre11∆ + Exo1 OE cells that survived expression of Cas9^D10A with gRNA1 (Left) or gRNA6 (Right). (B) Serial dilutions of the indicated strains expressing gRNA1 or gRNA6 were spotted on medium ± β-estradiol to induce Cas9^D10A expression. (C) Schematic of break orientation at the gRNA1 or gRNA6 locus (Top), enrichment of Yku70 at sequences flanking the gRNA1 or gRNA6 site in WT cells at 0, 2, or 4 h post-Cas9^D10A induction determined by ChIP-qPCR. (D) Enrichment of Yku70 at sequences flanking the gRNA1 or gRNA6 locus in mre11∆ cells. Amplicons to detect Yku70 binding are located 0.1 to 0.3 kb centromeric or telomeric to the gRNA1 cut site, or 0.2 to 0.4 kb centromeric or telomeric to gRNA6. All significance values calculated using an unpaired t test. ns P ≥ 0.05, *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001.

Since Exo1 functions redundantly with Dna2-Sgs1 in long-range resection, we also tested whether overexpressing Dna2 could rescue the growth defect of mre11∆ cells in response to expression of Cas9^D10A with gRNA1 or gRNA6. In contrast to Exo1 overexpression, we found that Dna2 expression compounded the repair defect of mre11∆ cells with both gRNAs (SI Appendix, Fig. S6B). We verified that Dna2 overexpression is neither toxic nor does it interfere with the repair of replication-dependent DSBs in WT cells (SI Appendix, Fig. S6B). Similarly, overexpressing the nuclease-dead exo1-D173A allele exacerbated the repair defect of mre11∆ cells, especially in the gRNA1 condition (SI Appendix, Fig. S6B), likely due to the nuclease-dead protein blocking DSB ends from potential resection by endogenous Exo1 or Dna2-Sgs1.

To test whether Ku prevents access to Exo1 at the leading strand collapse, we performed ChIP-qPCR for Yku70 in WT and mre11∆ cells targeted with gRNA1 or gRNA6. We found that Ku plays a relatively minimal role at the gRNA1 locus in WT cells with a peak enrichment of 1.9-fold, while Ku is strongly enriched at the blunt end of the leading strand collapse at the gRNA6 site with a peak enrichment of 4.5-fold (Fig. 5C). Notably, we observed minimal Ku enrichment at the 3′ overhang end flanking the gRNA1 and gRNA6 breaks (Fig. 5C). In the absence of Mre11, Ku enrichment at blunt ends increased by 10-fold at both leading and lagging-strand breaks, but there was still little Ku binding at 3′ overhang ends (Fig. 5D). We also measured Ku enrichment at a Cas9 DSB targeted with gRNA1. As expected, and in contrast to Cas9^D10A-induced breaks, we found strong, symmetrical Yku70 enrichment to both DSB ends (SI Appendix, Fig. S6C). The asymmetrical binding of Ku to replication-dependent DSBs likely contributes to the failure to repair these breaks by NHEJ (22).

Discussion

Here, we investigated resection dynamics at DSBs formed by replication fork collision with a nick on the leading or lagging-strand template. In brief, we find that the MRX complex is only required to initiate resection at deDSB ends that are predicted to be blunt, or nearly blunt, but not at ends predicted to have 3′ overhangs resulting from lagging-strand synthesis (Fig. 6). Consistently, we observe higher Mre11 binding to blunt ends of deDSBs than to 3′ overhang ends. We suggest that Exo1 and Dna2-Sgs1 can directly resect ends with long (~70 nt) 3′ overhangs, bypassing the need for resection initiation by MRX (Fig. 6). Although it has been shown that resection of ends produced by ionizing radiation or endonucleases is uncoordinated in the absence of the MRX complex (50, 51), this is likely due to the low efficiency of resection initiation at each end of a DSB, rather than to different end structures. Our findings have some parallels with telomeres, where it has been shown that resection can proceed in the absence of MRX at lagging-strand telomeres, but not at leading-strand telomeres (52, 53). The asymmetrical recognition and processing of the ends of replication-dependent deDSBs agrees with in vitro studies showing that replication fork collapse at nicks on leading and lagging-strand templates results in seDSBs with different end structures (20).

Fig. 6.

Replication-dependent DSB processing. Model of resection dynamics at both blunt and 3′ overhang DSB ends in wild-type (WT) cells (Left) and mre11∆ cells (Right). In WT cells, coordinated resection of both DSB ends promotes efficient sister-chromatid recombination. In mre11∆ cells, Ku accumulates at blunt ends, particularly at the leading strand collapse, and resection by Exo1 or Dna2-Sgs1 is inefficient. In contrast, Mre11 is dispensable for resection of ends with 3′ overhangs and resection proceeds by Dna2-Sgs1 or Exo1. Failure to process the blunt DSB end compromises repair in mre11∆ cells, leading to eventual cell death.

Although MRX is required to repair replication-dependent DSBs, the Mre11 nuclease activity is dispensable due to redundancy with Dna2-Sgs1. This finding is consistent with a role for MRX in recruiting Dna2 and Sgs1 to DSBs (35–37). When Mre11 nuclease is active, it creates a nick internal to DSB ends that acts as an entry site for Exo1-mediated degradation. In the absence of Mre11 nuclease, Exo1 has to compete with Ku for access to dsDNA ends. Thus, one possible reason for the failure of Exo1 to resect nCas9-induced DSBs formed in mre11Δ or mre11-H125N sgs1Δ cells is the inhibitory effect of Ku at blunt ends. This suggestion is consistent with the partial suppression of the mre11Δ repair defect by Exo1 overexpression or elimination of Ku (Fig. 5).

The finding that resection can proceed at one end of replication-dependent deDSBs in the absence of MRX raises the question of why the mre11Δ mutant shows such high sensitivity to nickase expression. It is possible that the 3′ overhang end can invade the sister chromatid to initiate repair synthesis but in the absence of timely second-end capture, the cells cannot complete repair. The other possible role for MRX is to maintain sister-chromatid cohesion during repair of replication-dependent DSBs. Previous studies have shown an increase in spontaneous and nick-induced interchromosomal recombination in the absence of the MRX complex in yeast, likely because of a sister-chromatid cohesion defect (22, 54). Nick-induced unequal sister chromatid recombination between closely linked repeats is also less impacted by loss of Mre11 (30, 39). How MRX achieves sister chromatid tethering is currently unknown. Several groups argue that the Rad50 subunit of MRX might tether sister chromatids at the site of a DSB through its long coiled-coil domains, analogous to cohesin (54–58). Others propose that through initiating end resection, MRX assists in de novo cohesin loading which indirectly stimulates sister chromatid tethering (59, 60). Resolving whether MRX plays a direct physical role maintaining sister chromatid tethering or whether MRX-initiated resection stimulates sister chromatid tethering would help better define how replication-dependent DSBs are repaired.

Additionally, we find that Ku binds blunt ends of replication-dependent DSBs, and that MRX is required to remove Ku from these ends (Fig. 6). This finding is consistent with studies showing delayed clearance of Ku from endonuclease-induced DSBs in the absence of MRX (61). An alternative explanation for this finding is that MRX physically competes with Ku for access to DNA ends. We did not detect Ku enrichment at 3′ overhang ends, even in the absence of MRX. Our results contrast with telomeres where Ku has been found to associate with both leading and lagging-strand telomeres (52). If the 3′ overhang ends are rapidly resected by Exo1 and/or Dna2-Sgs1 then the long tracts of ssDNA would be poor substrates for Ku binding. Furthermore, RPA bound to ends with ssDNA overhangs has been shown to prevent access to Ku (62). Since the 3′ overhang ends of replication-dependent DSBs would be bound by RPA, they would not be available for Ku to bind even prior to long-range resection.

Unexpectedly, we find higher enrichment of Ku at the site of a leading-strand collapse, as opposed to a blunt end generated by termination of an Okazaki fragment at the site of a nick formed on the lagging-strand template. The reason for this difference is not immediately obvious. It is possible that during Okazaki fragment termination, DNA polymerase δ (Pol δ) or an associated protein remains bound to the site of the DSB. In this scenario, Ku would have to compete with Pol δ for DSB binding. In contrast, leading-strand synthesis is carried out by Pol ε, which is coupled to the CMG helicase (63). Since in vitro work demonstrates that the CMG helicase dissociates from DNA when a nick is induced on a leading-strand template (20, 21), we propose that these ends might be less protected than blunt ends resulting from lagging-strand synthesis. This could result in higher Ku enrichment at a leading-strand collapse than to blunt ends of lagging-strand DSBs. In support of this idea, Exo1 overexpression more robustly rescues the growth defect of mre11∆ cells challenged with a lagging-strand DSB compared to a leading-strand collapse, suggesting that Ku plays less of an inhibitory role to resection at lagging-strand DSBs. Another possibility is that the blunt ends of breaks formed by nicks on the lagging-strand template have a different structure to the end formed by CMG run off at a nick on the leading-strand template. The in vitro study mapping ends generated at preformed nicks on plasmid templates detected only seDSBs (20); therefore, the structure of the end produced by polymerase run off at a nick on the lagging-strand template to form a deDSB is unknown. Since Exo1 preferentially degrades 5′ recessed ends, has reduced activity at blunt ends, and low activity at ends with 5′ overhangs (64), the 5′ protruding end produced by a leading-strand collapse would be a poor substrate for Exo1. If one end of the lagging-strand break were blunt, it would be a better substrate for Exo1.

The physiological relevance of Ku binding to one end of replication-dependent DSBs is unclear as these ends are not substrates for NHEJ repair (22, 23, 26, 30). It is possible that Ku protects one end of replication-dependent DSBs to minimize formation of potentially mutagenic ssDNA tracts. Consistent with this idea, work in Schizosaccharomyces pombe has demonstrated that the mutagenic break-induced replication pathway is more active in the absence of the Ku complex (23). Finally, since the Ku protein block has been shown to stimulate MRX endonuclease activity in vitro (8, 65), Ku may function at asymmetric DSB ends to ensure that resection proceeds in a mostly symmetric manner. One potential hazard of Ku binding to one end of replication-dependent DSBs is the possibility of joining the blunt ends of two independent replication-dependent DSBs, resulting in chromosome rearrangements. It would be of interest to determine whether the blunt ends produced at replication-dependent DSBs are competent for NHEJ by measuring Dnl4 binding.

In summary, this work details how resection proceeds at DSBs generated during DNA replication. While our findings are in line with some aspects of resection regulation at canonical DSBs, we also find key differences in repair factor binding and nuclease usage at each end of replication-dependent deDSBs. Related to our findings, BRCA1 is dispensable for resection of replication-dependent DSBs but is required for resection of endonuclease-induced DSBs (21, 30). In future studies, it will be of interest to determine whether other repair factors are specifically required for repair of replication-associated DSBs.

Materials and Methods

Media and Yeast Strains.

Complete yeast media contained 1% yeast extract, 2% peptone, 10 µg/mL adenine, and 2% glucose (YPAD) as a carbon source. Synthetic media contained 1× yeast nitrogen base, 1× amino acid dropout mix, and 2% glucose. For β-estradiol induction of Cas9^D10A/Cas9, a 10 mM stock of β-estradiol was diluted and added to plate or liquid media for a final concentration of 2 µM.

All yeast strains are in the RAD5-corrected W303 background (leu2-3,112 trp1-1 can1-100 ura3-1 ade2-1 his3-11,15) unless otherwise noted and are listed in SI Appendix, Table S1. Strains were constructed by standard genetic methods. Lithium acetate transformation was used to introduce plasmids or deletion cassettes containing a marker of choice and homology arms flanking the gene to be deleted. Other strains were made by genetic cross, followed by tetrad dissection and marker selection.

Plasmids.

Cas9 and gRNA plasmids.

Plasmids expressing FLAG-tagged Cas9 (pLS504) or Cas9^D10A (pLS503) were generated in the pRG203MX and integrated into the genome as previously described (22). For strains used for Yku70-FLAG ChIP, we derived plasmids without a FLAG tag fused to Cas9^D10A or Cas9 (pLS755 and pLS758, respectively) by amplifying the entire pLS503 or pLS504 plasmid except the 3xFLAG tag by PCR (see SI Appendix, Table S2 for primers) and ligating the resulting linear product with T4 DNA ligase after T4 polynucleotide kinase treatment. gRNAs were expressed in a pRG205MX (LEU2) integrating plasmid backbone (pLS505) as previously described (SI Appendix, Table S3) (22).

Episomal centromeric plasmids were previously constructed that express both Cas9^D10A and either gRNA1, gRNA6, gRNA14, or gRNA20 (pLS627, pLS633, pLS733, and pLS735, respectively) (22). A previously constructed episomal centromeric plasmid that expresses both Cas9 and gRNA6 (pLS715) was used to study canonical DSB end resection (22). Full plasmid sequences are available upon request.

Overexpression plasmids.

Overexpression plasmids were derived from the high-copy-number vector pRS426 (URA3). Full plasmid sequences are available upon request.

Spot Assays for Assessing Growth Sensitivity of Cas9^D10A Strains.

For all spot assays except those testing Exo1 overexpression, Cas9^D10A and the relevant gRNA were coexpressed from a centromeric plasmid. Single colonies were used to inoculate 2 mL SC-Leu media, and cultures were grown overnight at 30 °C while shaking. The following day, 10-fold serial dilutions were spotted on media with or without 2 µM β-estradiol. Plates were grown for 3 d at 30 °C, unless otherwise noted, after which plates were imaged with a desktop scanner.

Survival Assays for Overexpression Experiments.

All survival assays with Cas9^D10A were done in strains in which the relevant gRNA was integrated into the genome and grown on SC-URA media to maintain episomal overexpression vectors. Fresh single colonies from each strain were picked and grown for 4 h in SC-URA liquid media. Cells were then diluted and plated in technical replicate on SC-URA ± β-estradiol. Colonies were allowed to grow for 5 d at 30 °C, then counted. The survival frequency was calculated by the formula:

$$\frac{\#colonies on SC-URA+\beta -estradiol plates\times \mathrm{fold} \mathrm{dilution}}{\#colonies on SC-URA plates\times \mathrm{fold} \mathrm{dilution}}.$$

Physical Analysis of Cas9^D10A-Induced Breaks by Southern Blotting.

Induction of Cas9^D10A in asynchronous cultures.

Strains with integrated Cas9^D10A/Cas9 and gRNA cassettes were used for physical analysis of replication-dependent DSBs. For Cas9^D10A/Cas9 induction, single colonies of relevant strains were used to inoculate 5 mL YPAD cultures that were grown overnight at 30 °C. The following day, cultures were diluted to an OD₆₀₀ of 0.4 in YPAD and grown for 1 h. 1 × 10⁸ cells were collected before induction (t₀), and then, β-estradiol was added to the remaining culture at a final concentration of 2 µM to induce Cas9^D10A/Cas9. The same number of cells were collected at each timepoint, sodium azide was added to 0.1% final, cells were collected by centrifugation at 3,000 rpm for 5 min at 4 °C, cell pellets were then washed with dH₂O, centrifuged at 2,000× g for 2 min at 4 °C in a microcentrifuge and cell pellets were stored on ice prior to embedding in agarose plugs. Cells were collected at 0-, 1-, 2-, and 4-h points, and agarose plugs were prepared and processed as previously described (22, 66), followed by Southern blotting (22).

Flow Cytometry.

Two volumes of 95% ethanol were added to cultures containing 1 × 10⁷ cells, vortexed, and stored at −20 °C freezer overnight. The following day, cells were collected by centrifugation at 2,000× g for 3 min at room temperature, resuspended twice in 500 µL of 50 mM Tris-HCl pH7.5, incubated for 10 min at room temperature, and collected at 2,000× g for 3 min. 500 µL of RNAse A solution (1 mg/mL RNAse A in 50 mM Tris-HCl pH7.5, 7.5 mM NaCl) was added to each pellet, and incubated at 37 °C overnight. The following day, cells were collected, resuspended in 500 µL of 1 mg/mL Proteinase K in 50 mM Tris-HCl pH7.5, and incubated at 50 °C for 1 h. Cells were collected, resuspended in 500 µL of 200 mM Tris-HCl pH7.5, 200 mM NaCl, 80 mM MgCl₂, stained with 300 nM SYTOX Green, and sonicated briefly. Fluorescence was measured using an LSR-II (Becton Dickinson) and FACSDiva (Becton Dickinson) software using a 488 nm blue laser and 525/50 nm emission filter. Data analysis was performed using FlowJo_v10 (Becton Dickinson Biosciences) software.

qPCR-Based Resection assay.

Resection was measured using a qPCR-based assay as previously described (67) with the following modifications. ~10⁸ cycling cells were collected at the indicated time intervals after β-estradiol was added to a final concentration of 2 µM for Cas9^D10A/Cas9 induction. Per sample and amplicon, reactions were performed in technical triplicates containing 4.4 µL digested or undigested genomic DNA, 300 nM of each forward and reverse primer (listed in SI Appendix, Table S4), and 1× SsoAdvanced Universal SYBR Green Supermix (Bio-Rad) in a total volume of 10 μL and analyzed them on a CFX384 Real-Time System (Bio-Rad) with 10 min initial denaturation at 95 °C, followed by 40 cycles of 1 min denaturation at 95 °C and 1 min annealing and extension at 58 °C. The fraction of ssDNA, f_ssDNA, at the restriction enzyme recognition site was calculated as previously described (22) with the following formula:

$$\begin{array}{c}{f}_{\mathit{ssDNA}}=2/\left(\left(\right(E_{\mathit{RS}}\mathrm{\Delta }Cq\left(digest‐mock\right)/E_{\mathit{ADH1}}\mathrm{\Delta }Cq\left(digest‐mock\right))+1{)}^{*}{f}_{\mathit{DSB}}\right)\hfill \end{array}$$

where E_RS and E_ADH1 are the primer efficiencies for the restriction site (RS) and ADH1 amplicons, ΔC_q(digest-mock) is the difference between quantification cycles for the digested minus the mock-digested sample, and f_DSB is the fraction of genomes where Cas9^D10A/Cas9 cleaved the target locus. f_DSB for a locus cleaved by Cas9^D10A can be calculated as

$$f_{\mathit{DSB}}=\left(1-\frac{E_{nCas9cs}\left(\mathrm{\Delta }Cq\left(t0-t\right)\right)}{E_{ADH1}\left(\mathrm{\Delta }Cq\left(t0-t\right)\right)}\right)÷\frac{3}{4}.$$

f_DSB for a locus cleaved by Cas9 can be calculated as

$$f_{\mathit{DSB}}=1-\frac{E_{Cas9cs}\left(\mathrm{\Delta }Cq\left(t0-t\right)\right)}{E_{ADH1}\left(\mathrm{\Delta }Cq\left(t0-t\right)\right)},$$

where E_nCas9cs/E_Cas9cs and E_ADH1 are the primer efficiencies for the Cas9^D10A/Cas9-cut site and ADH1 amplicons and ΔC_q(t₀ − t) is the difference between the quantification cycles at the first time point (before β-estradiol induction of Cas9^D10A/Cas9) and the evaluated time point.

ChIP-qPCR to Measure Protein Recruitment to Replication-Dependent DSBs.

Protein recruitment to replication-dependent DSBs was assessed with a ChIP-qPCR assay. 25 mL cycling cells at a density of OD₆₀₀ = 0.4 were collected at the indicated time intervals after β-estradiol was added to a final concentration of 2 µM for Cas9^D10A/Cas9 induction. 37% formaldehyde was added to a final concentration of 1% and incubated at RT for 15 min with shaking. The crosslinking reaction was quenched with 2.5 M glycine, followed by washing with HEPES-buffered saline and ChIP lysis buffer (50 mM HEPES, 140 mM NaCl, 1 mM EDTA, 1% IGEPAL CA-630, and 0.1% sodium deoxycholate).

Cells were lysed in ChIP lysis buffer containing 1 mM PMSF using a FastPrep-24 (MP Biomedicals) protocol, run for 30 s at 6.5 m/s four times. Lysate was placed on ice between each cycle for 45 s. Chromatin was fragmented with a QSonica Q800R sonicator for 10 cycles with an amplitude of 65% for 30 s. After clearing the lysate, IP samples were incubated with the appropriate antibody preincubated with A/B Dynabeads in lysis buffer at a concentration of 1:20 for 3 h. Control input DNA samples were kept on ice during immunoprecipitation of the remaining DNA. Immunoprecipitated DNA was washed and reverse crosslinked at 65 °C for 16 h.

Input and IP DNA was purified using the Monarch PCR and DNA cleanup kit (NEB). qPCRs were performed in the same manner as described in the resection assay with the following modification to the CFX384 Real-Time System protocol. qPCRs were run with a 3 min initial denaturation at 95 °C, followed by 40 cycles of 15 s denaturation at 95 °C and 30 s annealing and extension at 60 °C. Fold enrichment is defined as 2^∆∆Cq, where ∆∆Cq is defined as

$$\mathrm{\Delta }\mathrm{\Delta }\mathrm{Cq}=\left(\mathrm{Cq}\left({\mathit{IP}}_{\mathit{control}}\right)-\mathrm{Cq}\left({\mathit{input}}_{\mathit{control}}\right)\right)-\left(\mathrm{Cq}\left({\mathit{IP}}_{\mathit{DSB}}\right)-\mathrm{Cq}\left({\mathit{input}}_{\mathit{DSB}}\right)\right).$$

Quantification and Statistical analysis.

Quantification and statistical analysis were done using PRISM (GraphPad). ChIP-qPCR experiments were performed at least two times, and statistical significance was determined by a two-tailed Student’s t test. For ChIP-qPCR and survival assays, error bars indicate SD. The two-tailed Student’s t test was also used for all survival assays, with the following symbols to indicate significance thresholds: ns = not significant, P > 0.05, *P < 0.05, ***P < 0.001, and ****P < 0.0001.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

Study data are included in the article and/or SI Appendix. Plasmids and yeast strains generated in this study are available from the corresponding author without restriction.