|Home | About | Journals | Submit | Contact Us | Français|
The repair of DNA double-strand breaks (DSBs) by homologous recombination (HR) is initiated by nucleolytic degradation of the 5′ terminated strands in a process termed end resection. End resection generates 3′ single-stranded DNA tails, substrates for Rad51 to catalyze homologous pairing and exchange of DNA strands, and for activation of the DNA damage checkpoint. The commonly accepted view is that end resection occurs by a two-step mechanism. In the first step, Sae2/CtIP activates the Mre11-Rad50-Xrs2/Nbs1 (MRX/N) complex to endonucleolytically cleave the 5′-terminated DNA strands close to break ends, and in the second step Exo1 and/or Dna2 nucleases extend the resected tracts to produce long 3′-ssDNA-tailed intermediates. Initiation of resection commits a cell to repair a DSB by HR because long ssDNA overhangs are poor substrates for non-homologous end joining (NHEJ). Thus, the initiation of end resection has emerged as a critical control point for repair pathway choice. Here, I review recent studies on the mechanism of end resection and how this process is regulated to ensure the most appropriate repair outcome.
DNA double-strand breaks (DSBs) can form as a consequence of normal cell metabolism or by action of exogenous agents, such as ionizing radiation (IR) or chemotherapeutic drugs. In addition, DSBs are intermediates in several programmed recombination events. Failure to repair a chromosomal DSB leads to loss of genetic information or even cell death, and inaccurate repair can cause mutagenesis or chromosome rearrangements. Indeed, defects in DSB repair are associated with developmental, immunological and neurological disorders, and can accelerate tumorigenesis (O’Driscoll, 2012).
Typically, DSBs are repaired by an end joining mechanism or by homologous recombination (HR). Non-homologous end joining (NHEJ) involves the direct ligation of DNA ends with little or no homology and is generally considered to be an error-prone process because nucleotides can be lost or gained at the ends prior to ligation (Chiruvella, Liang, and Wilson, 2013). Furthermore, in cells with multiple DSBs on different chromosomes, end joining can result in chromosome translocations. Classical NHEJ requires the Ku70-Ku80 heterodimer (hereafter referred to as Ku), which protects ends from degradation, and the DNA ligase IV (Dnl4/Lig4) complex to ligate ends (Chiruvella, Liang, and Wilson, 2013). NHEJ predominates in the G1 phase of the cell cycle when end resection is less active (Figure 1). Alternative Ku and Dnl4/Lig4 independent joining mechanisms (alt-NHEJ) can repair DSBs at a lower frequency than NHEJ. One form of alt-NHEJ involves alignment of ends through very short sequence homologies (generally <18 nt) internal to the ends that are revealed by end resection and is termed microhomology-mediated end joining (MMEJ) (Sfeir and Symington, 2015). If long (>200 bp) direct repeats flank the DSB, then repair can occur by annealing of the complementary 3′-ssDNA tails formed by end resection in a mechanism referred to as single-strand annealing (SSA) (Paques and Haber, 1999). In several respects SSA and MMEJ are similar, they differ by the extent of sequence homology and requirement for a dedicated annealing protein for SSA (Rad52) resulting in a higher efficiency of repair than MMEJ. SSA and MMEJ are mutagenic repair processes because they are associated with deletions and chromosome translocations (Symington and Gautier, 2011). Since SSA requires sufficient resection to expose long repeats it has frequently been used as read out for extensive resection (Karanja et al., 2012; Mimitou and Symington, 2008; Zhu et al., 2008).
HR requires a homologous DNA duplex as a template for repair of the broken chromosome, and consequently HR is restricted to the S and G2 phases of the cell cycle when a sister chromatid is present and kept in close proximity by Cohesin (Figure 1) (Symington, Rothstein, and Lisby, 2014). A chromosome homolog or ectopic sequence can be used as repair template, but there is a strong bias to use the sister chromatid (Kadyk and Hartwell, 1992). HR initiates by nucleolytic degradation of the 5′-terminated strands to generate 3′ single-stranded DNA (ssDNA) tails, a process termed end resection (Paques and Haber, 1999; Mimitou and Symington, 2009). The 3′ ssDNA tails are initially bound by the ubiquitous and abundant ssDNA binding protein, Replication Protein A (RPA) (Wang and Haber, 2004; Lisby et al., 2004). The recombination mediators, yeast Rad52 or human BRCA2, replace RPA with the conserved Rad51 recombinase (San Filippo, Sung, and Klein, 2008). Once Rad51 nucleates on ssDNA it can cooperatively bind to form a right-handed helical filament, displacing RPA. The Rad51 nucleoprotein filament is essential for the homology search and pairing of the ssDNA bound by Rad51 with the complementary strand of the donor duplex (San Filippo, Sung, and Klein, 2008). The 3′ end of the invading strand can then be extended by DNA synthesis, followed by resolution of branched DNA structures, DNA synthesis and ligation. The reader is referred to several excellent reviews of the late steps of HR that are not described here (West et al., 2015; Bizard and Hickson, 2014).
The nucleolytic activities responsible for end resection in eukaryotes were identified, and have been best characterized, in Saccharomyces cerevisiae (Mimitou and Symington, 2009). Here, I review these studies and refer to other model systems where findings have been confirmed or differ from budding yeast. Physical analysis of resection intermediates generated at naturally occurring DSBs during meiosis and at an HO endonuclease-induced DSB in somatic cells demonstrated that ends are processed to generate 3′ ssDNA tails (Sun, Treco, and Szostak, 1991; White and Haber, 1990). In meiosis, a dimer of the topoisomerase II-like protein Spo11 generates DSBs by covalent attachment to the 5′ ends at recombination hotspots (Keeney, Giroux, and Kleckner, 1997; Bergerat et al., 1997). Spo11 is then removed along with a short oligonucleotide by an endonucleolytic mechanism requiring the Mre11-Rad50-Xrs2 (MRX) complex and Sae2 (Figure 2) (Mimitou and Symington, 2009; Neale, Pan, and Keeney, 2005; Garcia et al., 2011). The size of the oligonucleotide attached to Spo11 is shorter than the average length of 3′ ssDNA tails in meiosis (~850 nt) suggesting MRX and Sae2 are responsible for the initial clipping to remove Spo11 prior to more extensive resection (Neale, Pan, and Keeney, 2005; Zakharyevich et al., 2010). The initiation of resection at an HO endonuclease-induced DSB is greatly delayed in mre11Δ, rad50Δ and xrs2Δ mutants, but processing eventually occurs indicating that other nucleases can act directly on ends that lack chemical modifications (Mimitou and Symington, 2009). Once resection initiates in mre11Δ or rad50Δ mutants, it proceeds at around 4 kb/h, the same rate as wild-type cells (Zhu et al., 2008). Exo1 and Dna2 were identified as redundant nucleases required for extensive resection of HO-induced DSBs (Zhu et al., 2008). Dna2 is an endonuclease that requires Sgs1 helicase and the Sgs1-interacting proteins, Top3 and Rmi1 (STR complex), for resection of dsDNA (Zhu et al., 2008). In the exo1Δ sgs1Δ mutant, resection initiates but the ssDNA tracts are short (~100–600 nt) (Mimitou and Symington, 2008; Zhu et al., 2008; Gravel et al., 2008). Elimination of MRX or Sae2 in the exo1Δ sgs1Δ background blocks resection and is cell lethal, highlighting the importance of DNA end processing for chromosome maintenance (Mimitou and Symington, 2008).
MRE11, RAD50 and XRS2 were identified by their roles in resistance to ionizing radiation (IR) and meiotic recombination (Mimitou and Symington, 2009). Null mutation of any of the genes confers a similar phenotype and the mutations are epistatic, consistent with them functioning in the same pathway. Subsequent showed the proteins form a stable complex (Carney et al., 1998; Usui et al., 1998). In addition to its important role in the initiation of HR, MRX/N is essential for NHEJ in budding yeast, and is partially required for NHEJ in mammals (Dinkelmann et al., 2009; Boulton and Jackson, 1998; Rass et al., 2009; Xie, Kwok, and Scully, 2009). Thus, MRX/N plays a pivotal role in repair pathway selection.
Mre11 and Rad50 are conserved in all phylogenetic kingdoms and even in some bacteriophages. Mre11 has phosphodiesterase motifs in the N-terminal region of the protein, and displays Mn2+-dependent 3′-5′ dsDNA exonuclease and ssDNA endonuclease activities in vitro (Usui et al., 1998; Furuse et al., 1998; Moreau, Ferguson, and Symington, 1999; Paull and Gellert, 1998). Mutation of conserved residues within the phosphodiesterase motifs eliminates nuclease activity in vitro and confers a separation of function in vivo (Usui et al., 1998; Furuse et al., 1998; Moreau, Ferguson, and Symington, 1999; Bressan et al., 1998). Mre11 nuclease-defective mutants (termed mre11-nd here) are unable to repair meiotic DSBs because Spo11 is retained at the 5′ ends; however, HO-induced DSBs are processed with close to normal kinetics and mre11-nd cells exhibit intermediate sensitivity to IR, relative to mre11Δ and wild-type cells (Moreau, Ferguson, and Symington, 1999; Krogh et al., 2005; Llorente and Symington, 2004). Lobachev and colleagues showed that closely spaced inverted Alu repeats stimulate mitotic recombination and this stimulation requires the Mre11 nuclease, Sae2 and other components of the MRX complex (Lobachev, Gordenin, and Resnick, 2002). The inverted repeats are thought to extrude to form a hairpin or cruciform structure that is cleaved at the base by a resolvase, or to form a fold back structure when a nearby DSB is resected. The tip of the foldback or hairpin structure would then be cleaved by MRX creating a DSB to be repaired by HR (Figure 2). The MR complex and E. coli SbcCD (SbcC is the homolog of Rad50, and SbcD of Mre11) cleave hairpin DNA structures in vitro consistent with their role in hairpin cleavage in cells (Connelly, Kirkham, and Leach, 1998; Paull and Gellert, 1999; Trujillo and Sung, 2001). While the MRX complex is required for NHEJ, DNA damage checkpoint signaling and telomere length homeostasis in yeast, the mre11-nd mutants are normal for these functions (Moreau, Ferguson, and Symington, 1999). In Schizosaccharomyces pombe and mouse cells, mre11-nd mutations confer greater defects in resection, HR and resistance to DNA damaging agents than is observed in budding yeast (Buis et al., 2008; Langerak et al., 2011; Williams et al., 2008). The reason for this difference is unknown, but could be due to lower compensating activity by Dna2 and/or Exo1.
Most of the mre11-nd mutations eliminate both exo- and endonuclease activities in vitro; however, the Pyrococcus furiosus mre11-H52S mutation eliminates only the exonuclease activity (Williams et al., 2008). The separation of activities is less distinct for the S. cerevisiae mre11-H59S protein (equivalent to Pfmre11-H52S), which exhibits slightly impaired endonuclease activity in addition to reduced exonuclease activity (Garcia et al., 2011). The mre11-H59S mutation confers an interesting phenotype during meiosis: Spo11-oligonucleotides are released but show increased size heterogeneity as compared with wild-type cells (Garcia et al., 2011). This phenotype could stem from MRX cleaving further from ends than originally proposed with the short Spol11-oligonucleotides resulting from Mre11 3′-5′ exonuclease activity initiated at the nick (Figure 2), an appealing model because it provides a rationale for the conserved Mre11 3′-5′ exonuclease activity.
Rad50 is a member of the SMC family of proteins, characterized by nucleotide-binding motifs at the N- and C-termini of the protein separated by long coiled-coil domains (Alani, Subbiah, and Kleckner, 1989). The coiled coils fold on themselves to create a nucleotide-binding domain (NBD) at one end of the polypeptide and the other end has a Zn2+-hook dimerization motif (Hopfner et al., 2002). A dimer of Mre11 interacts with the NBDs of a Rad50 dimer to form a DNA-binding head domain (Hopfner, 2014). ATP binding and hydrolysis induce large conformational changes to Rad50 that influence the Mre11 nuclease activity (Lammens et al., 2011; Lim et al., 2011; Williams et al., 2011). The Rad50 NBDs bind ATP at their interface and engage the dsDNA-binding cleft of Mre11, partially occluding the nuclease active site. In this conformation the Mre11 3′-5 exonuclease is inhibited while the endonuclease activity is unaffected. Conserved residues in the Rad50 NBD are essential for Rad50 function in vivo and for activation of the Mre11 endonuclease activity in vitro (Alani, Padmore, and Kleckner, 1990; Cannavo and Cejka, 2014). A class of rad50 mutants (rad50s) is deficient in Spo11 removal and hairpin cleavage in vivo, but shows much higher resistance to DNA damaging agents than the rad50Δ mutant, a similar phenotype to mre11-nd mutants (Lobachev, Gordenin, and Resnick, 2002; Alani, Padmore, and Kleckner, 1990). The rad50s mutations are close to the NBD on the surface of the protein suggesting they might be important for protein-protein interactions.
Xrs2/Nbs1 is less conserved than Mre11 and Rad50 and has only been identified in eukaryotes (Stracker and Petrini, 2011). Hypomorphic mutations of NBS1 are found in Nijmegen breakage syndrome (NBS), which is characterized by cellular radiosensitivity, immune deficiency and cancer proneness, similar to individuals with Ataxia telangiectasia (A-T) (Carney et al., 1998; Stracker and Petrini, 2011; McKinnon and Caldecott, 2007). Similarly, hypomorphic alleles of human RAD50 and MRE11 are associated with NBS-like and AT-like disorders (Stewart et al., 1999; Waltes et al., 2009). Xrs2/Nbs1 is required for nuclear localization of Mre11 and has a number of protein-protein interaction motifs, suggesting it functions as a chaperone and scaffold (Carney et al., 1998; Stracker and Petrini, 2011; Nakada, Matsumoto, and Sugimoto, 2003; Tsukamoto et al., 2005). The C-terminal region of Xrs2/Nbs1 contains Mre11 and Tel1/ATM binding motifs (Nakada, Matsumoto, and Sugimoto, 2003; Tsukamoto et al., 2005; Falck, Coates, and Jackson, 2005; You et al., 2005). Mutations of residues within the Mre11 interaction motif that prevent binding to Mre11 confer a phenotype indistinguishable from xrs2Δ and mre11Δ null mutations (Tsukamoto et al., 2005). Interestingly, expression of a 35 amino acid peptide encompassing the Mre11 interaction motif fused to a nuclear localization sequence in the xrs2Δ mutant restores partial resistance to DNA damaging agents, suggesting some of the repair functions of MR are independent of Xrs2 (Tsukamoto et al., 2005). S. pombe Nbs1 binds to sequences in the N-terminal region of Mre11, including an insertion loop between nuclease motifs II and III that is absent from the bacterial and archaeal proteins (Schiller et al., 2012). Mutations of contact residues within the loop result in reduced Mre11 nuclear localization and are associated with NBS-like disorder in humans. Deletion of the Tel1 interaction motif results in a phenotype similar to tel1Δ mutants, including a defect in DNA damage signaling and short telomeres (Nakada, Matsumoto, and Sugimoto, 2003). The N-terminal region of Nbs1 contains motifs associated with binding to phosphorylated proteins: a forkhead associated (FHA) domain and a tandem breast cancer associated C-terminal (BRCT) domain; Xrs2 has only the FHA domain. Several phosphorylated DNA repair and checkpoint proteins bind to the Nbs1/Xrs2 FHA domain, including CtIP/Ctp1/Sae2, Lif1, and Mcd1 (Chapman and Jackson, 2008; Lloyd et al., 2009; Williams et al., 2009; Palmbos et al., 2008; Wang et al., 2013).
Like Xrs2, Sae2 is poorly conserved and is only present in eukaryotes. The S. pombe Ctp1 and vertebrate CtIP (CtIP is also known as RBBP8) proteins are thought to be the functional orthologs of Sae2 (Limbo et al., 2007; Sartori et al., 2007; Akamatsu et al., 2008). These proteins have an N-terminal coiled-coil multimerization domain that is essential for function, and are regulated by cyclin-dependent kinase (CDK)- and DNA damage checkpoint kinase-dependent phosphorylation (Andres et al., 2015; Davies et al., 2015; Baroni et al., 2004; Huertas et al., 2008; Huertas and Jackson, 2009; Wang et al., 2012; Kim et al., 2008). The sae2Δ mutant exhibits a similar phenotype to rad50s and mre11-nd mutants, including defects in Spo11 removal and hairpin opening in vivo, consistent with Sae2 activating the Mre11 endonuclease (Lobachev, Gordenin, and Resnick, 2002; McKee and Kleckner, 1997; Prinz, Amon, and Klein, 1997; Rattray et al., 2005). The role of Sae2 in hairpin cleavage is puzzling because MR or MRX/N can cleave hairpin structures in vitro without Sae2 (Paull and Gellert, 1998; Paull and Gellert, 1999; Trujillo and Sung, 2001). Sae2 is likely to have additional cellular functions because sae2Δ mutants exhibits higher sensitivity to DNA damaging agents than mre11-nd mutants, Mre11 persists at ends for longer in the absence of Sae2 than observed when the Mre11 nuclease is eliminated and the DNA damage checkpoint is hyper-activated in sae2Δ mutants (Lisby et al., 2004; Clerici et al., 2006; Mimitou and Symington, 2010). The prolonged Mre11 retention at ends and checkpoint deactivation defects of sae2Δ can be rescued by mutations in the N-terminal region of Mre11 that decrease DNA binding without affecting end resection, indicating that these defects are separable (Chen et al., 2015; Puddu et al., 2015). Although Sae2 and CtIP were reported to exhibit endonuclease activity in vitro (Lengsfeld et al., 2007; Makharashvili et al., 2014; Wang et al., 2014), the proteins lack homology to known nucleases and other groups have not found nuclease activity associated with highly purified Sae2 or Ctp1 (Cannavo and Cejka, 2014; Andres et al., 2015). Further investigation of these conflicting results is needed to clarify the role of Sae2/Ctp1/CtIP.
Because the polarity of the Mre11 exonuclease is opposite to that predicted for processing DSBs, interest has centered on the endonuclease activity of Mre11 and how it is activated to cleave dsDNA. MRN exhibits limited ATP-dependent melting activity in vitro and this could potentially generate the substrate for Mre11 cleavage (Cannon et al., 2013). In efforts to model the substrate for resection in meiotic cells, Cannavo and Cejka generated short linear duplexes with biotin/streptavidin blocks at 5′ and/or 3′ ends and were able to demonstrate MRX and Sae2 dependent cleavage of the 5′ strand 15–20 nucleotides (nt) from the end (Cannavo and Cejka, 2014). The reaction was dependent on ATP, the Mre11 nuclease activity and phosphorylated Sae2. Variants of Rad50 deficient in ATP binding and/or hydrolysis were unable to support Sae2-stimulated MRX endonucleolytic cleavage. Furthermore, the MRX complex with the rad50s mutation (K81I) was also completely deficient for the endonuclease activity, consistent with in vivo studies. Over-expression of Sae2 has been shown to partially suppress the rad50s phenotype suggesting that Sae2 activates the Mre11 endonuclease via Rad50 (Clerici et al., 2005). Using plasmid-sized DNA substrates, a low level of MRX endonuclease activity was found in the absence of Sae2, in agreement with an earlier report of weak endonucleolytic activity for the archaeal MR complex (Cannavo and Cejka, 2014; Hopkins and Paull, 2008). How Sae2 activates MRX, and how the 5′ strand is selected for cleavage are unanswered questions.
The MRN complex and CtIP are essential for early embryonic development and cell proliferation in mammalian cells (Chen et al., 2005; Luo et al., 1999; Xiao and Weaver, 1997; Zhu et al., 2001). In cells expressing conditional knock-out alleles or in which MRN components or CtIP have been depleted using short interfering RNAs (siRNA), DSB-induced RPA and Rad51 foci are greatly reduced (Buis et al., 2008; Sartori et al., 2007). Furthermore, a physical method to detect ssDNA formed by resection of site-specific DSBs was recently developed in human cells and verified the important roles of Mre11 and CtIP in resection (Zhou et al., 2014).
Although Mre11 nuclease and Sae2 are essential for processing meiosis-specific DSBs and hairpins, resection of DSBs produced by HO and I-SceI endonucleases still occurs in their absence (Llorente and Symington, 2004; Clerici et al., 2005). Exo1 is member of the Rad2/XPG family of nucleases and was first identified as a meiosis induced 5′-3′ dsDNA exonuclease (Szankasi and Smith, 1992; Szankasi and Smith, 1995). The exo1Δ mutant shows only mild sensitivity to DNA damaging agents and normal kinetics of resection initiation of an HO-induced DSB, but exhibits a partial defect in extensive resection (Mimitou and Symington, 2008; Llorente and Symington, 2004). Resection of endonuclease-induced DSBs is further reduced in the exo1Δ mre11Δ mutant compared to mre11Δ, and EXO1 overexpression partially suppresses the mre11Δ resection initiation defect (Lee et al., 2002; Moreau, Morgan, and Symington, 2001; Tsubouchi and Ogawa, 2000). Elimination of Ku suppresses the resection initiation defect of the mre11Δ, mre11-nd and sae2Δ/ctp1Δ mutants in an Exo1-dependent fashion suggesting Ku prevents access of Exo1 to DNA ends (Figure 3) (Langerak et al., 2011; Williams et al., 2008; Limbo et al., 2007; Mimitou and Symington, 2010; Shim et al., 2010; Foster, Balestrini, and Petrini, 2011). Biochemical studies with purified Ku and Exo1 support this idea (Shim et al., 2010; Krasner et al., 2015). Less Exo1 is localized to DNA ends in the absence of MRX as a result of increased Ku binding and possibly by a minor role for MRX in recruiting Exo1 (Shim et al., 2010; Clerici et al., 2008).
In the absence of Exo1 nuclease, meiotic DSBs are processed to form ssDNA tails of ~270 nt and the length of the ssDNA tails is unaffected by sgs1Δ mutation (Zakharyevich et al., 2010). These data are consistent with the model that MRX initiates resection by cleaving the 5′ strand ~270 nt from the end and further processing from the 5′ end is carried out by Exo1 (Figure 2) (Garcia et al., 2011; Zakharyevich et al., 2010).
In vitro, purified Exo1 preferentially degrades dsDNA with a 5′ recessed end (Cannavo, Cejka, and Kowalczykowski, 2013). Exo1 can also degrade from a nick in line with its role in MRX-initiated resection and mismatch repair (Szankasi and Smith, 1992; Modrich, 2006). MRX/N stimulates resection by Exo1 in vitro, even though the genetic studies indicate that Exo1 can function without MRX as long as Ku is absent (Mimitou and Symington, 2010; Cannavo, Cejka, and Kowalczykowski, 2013; Nicolette et al., 2010; Nimonkar et al., 2011). CtIP also physically interacts with EXO1 and negatively regulates its activity in vitro (Eid et al., 2010). The functional interaction between Exo1 and RPA is complex. Exo1 is recruited to DSBs in the absence of RPA and forms brighter foci; however, RPA is required for extensive resection by Exo1 in vivo (Chen, Lisby, and Symington, 2013; Myler et al., 2016). Independent in vitro studies using reconstituted proteins have reported that RPA could both stimulate and inhibit yeast Exo1 (Cannavo, Cejka, and Kowalczykowski, 2013; Nicolette et al., 2010), conflicting data have also been obtained with the human proteins (Nimonkar et al., 2011; Yang et al., 2013; Genschel and Modrich, 2003). The stimulatory role of RPA was proposed to occur by preventing non-specific interaction of Exo1 with ssDNA (Cannavo, Cejka, and Kowalczykowski, 2013). Single-molecule fluorescence imaging has shown that yeast and human Exo1 are both processive nucleases that are rapidly stripped from DNA by RPA attenuating resection (Myler et al., 2016). Multiple cycles of binding and release would be required for extensive resection. Mammalian cells have another ssDNA binding complex, SOSS1, which contains a subunit related to E. coli SSB, hSSB1, shown to stimulate the resection activity of EXO1 (Yang et al., 2013). Human BLM helicase (the ortholog of yeast Sgs1) interacts with EXO1 and stimulates its nuclease activity in vitro in a helicase-independent manner (Nimonkar et al., 2008). To date, there is no evidence to suggest a similar interaction between yeast Exo1 and Sgs1.
Substantial resection of an HO-induced DSB occurs in the absence of Exo1 and Sae2 or the Mre11 nuclease indicating that at least one other nuclease is active. The sgs1Δ and/or dna2 mutations were shown to synergize with exo1Δ to prevent extensive resection of endonuclease-induced DSBs (Mimitou and Symington, 2008; Zhu et al., 2008; Gravel et al., 2008). DNA2 is essential for viability due to its role in Okazaki fragment processing (Bae et al., 2001), consequently, the sgs1Δ mutation is more often used to study end resection. Dna2 has helicase and endonuclease activities, but only the nuclease is required for resection (Cejka et al., 2010; Niu et al., 2010). Instead, Dna2 relies on the Sgs1/BLM helicase to generate ssDNA for cleavage (Nimonkar et al., 2011; Cejka et al., 2010; Niu et al., 2010). The Sgs1 interacting partners, Top3 and Rmi1 are essential for Dna2-Sgs1 catalyzed resection in cells, and exhibit a stimulatory effect on the reaction in vitro by increasing Sgs1 recruitment to DNA ends (Zhu et al., 2008; Cejka et al., 2010; Niu et al., 2010). Unlike the decatenation activity of Top3, which is important for removing late recombination intermediates, resection does not require Top3 catalytic activity. MRX/N interacts with Sgs1/BLM and promotes its association with DNA ends resulting in more efficient DNA unwinding (Nimonkar et al., 2011; Cejka et al., 2010; Niu et al., 2010). In Xenopus egg extracts, DNA2 partners with WRN, another member of the RecQ family of helicases, to catalyze end resection (Liao, Toczylowski, and Yan, 2008; Toczylowski and Yan, 2006; Yan et al., 2011). In vivo studies support roles for both WRN and BLM in DNA2-catalyzed resection in mammalian cells (Gravel et al., 2008; Sturzenegger et al., 2014; Thangavel et al., 2015). Dna2 can cleave 5′ or 3′ flaps in vitro, but in the presence of RPA cleavage is directed to the 5′ strand and the 3′ strand is protected (Bae et al., 2001; Cejka et al., 2010; Niu et al., 2010). Dna2 fails to be recruited to DSBs when RPA is depleted from cells resulting in a block to extensive resection (Chen, Lisby, and Symington, 2013). RPA is also required to stabilize ssDNA generated by the Sgs1/BLM and WRN helicases, and is considered an obligate factor for resection by Dna2-STR (Nimonkar et al., 2011; Cejka et al., 2010; Niu et al., 2010; Yan et al., 2011).
Genetic studies suggest that Dna2-STR provides an alternative mechanism to initiate resection when Sae2 or the Mre11 nuclease is absent (Figure 3). The sae2Δ mutation is lethal when combined with sgs1Δ, and the lethality is suppressed if Ku is eliminated or if Exo1 is over expressed (Mimitou and Symington, 2010). Mutation of MRE11 or RAD50 does not cause lethality in the sgs1Δ background, although cells grow very slowly, suggesting the sae2Δ sgs1Δ lethality is not due solely to the end resection defect (Bernstein et al., 2013). Moreover, the mre11Δ sae2Δ sgs1Δ triple mutant is viable suggesting it is retention of MRX at ends that causes lethality in sae2Δ sgs1Δ cells (Hardy et al., 2014). The mre11-nd sgs1Δ double mutant is viable, but grows slowly and early resection is greatly reduced (Mimitou and Symington, 2010; Shim et al., 2010; Budd and Campbell, 2009). In yeast, MRX is required for Dna2 and Sgs1 localization to DSBs potentially explaining the greater resection defect observed for mre11Δ compared with mre11-nd cells (Shim et al., 2010). Although Dna2 recruitment to DSBs is unaffected by sae2Δ, a recent study showed that DNA2 localization to DSBs in vertebrate cells requires CtIP (Hoa et al., 2015). Furthermore, CtIP and Dna2 cooperate to promote end resection in Xenopus extracts (Peterson et al., 2013). The recruitment function could be by direct physical interaction between the MRN or CtIP and DNA2, or potentially by RPA bound to the short tracts of ssDNA formed by resection initiation or MRN-mediated duplex DNA melting. Thus, Dna2-STR is able to initiate end resection in the absence of MRX-Sae2 clipping and appears to be less inhibited by Ku binding to ends. By contrast, Exo1 initiates resection from ends poorly, except when Ku is absent. Notably, Dna2 requires a free 5′ end for endonucleolytic cleavage explaining why it is unable to remove Spo11, but can function at an HO-induced DSB (Balakrishnan et al., 2010).
In the absence of a homologous template to direct repair, an endonuclease-induced DSB undergoes extensive resection, up to 50 kb from each end over 12 hours (Zhu et al., 2008). The physiological relevance of extensive resection by Exo1 and/or Dna2-STR is not obvious because the short ssDNA tails generated by MRX-Sae2 are sufficient for meiotic recombination (Zakharyevich et al., 2010). In mitotic cells, sister-chromatid repair of IR-induced DSBs is also unaffected by exo1Δ and sgs1Δ mutations, but the efficiency of ectopic recombination is reduced by ~30% (Zhu et al., 2008; Westmoreland and Resnick, 2016). The DNA damage checkpoint is defective in the exo1Δ sgs1Δ mutant due to insufficient ssDNA to activate the Mec1/ATR kinase (Zhu et al., 2008; Gravel et al., 2008). Thus one possible explanation for the ectopic recombination defect is that pairing is slower from a non-sister template necessitating cell cycle arrest to allow more time for repair before chromosome segregation. There are pathological consequences of extensive resection: 3′ ends become unstable over time and clustered mutations are generated due to the inherent lability of ssDNA compared with dsDNA (Zierhut and Diffley, 2008; Roberts et al., 2012). The 3′ ssDNA tails would normally be bound and protected by RPA, but when RPA is depleted from cells the ssDNA forms secondary structures that can be attacked by structure-selective endonucleases, including MRX-Sae2 (Chen, Lisby, and Symington, 2013). The failure of MRX-Sae2 to remove foldback structures formed when RPA is dysfunctional can lead to palindromic gene amplification and more complex chromosome rearrangements (Deng et al., 2015).
The DNA damage checkpoint responds to DSBs, and to ssDNA formed by end resection or by replicative stress (Ciccia and Elledge, 2010). The sentinel kinases for the response are Tel1/ATM and Mec1/ATR, members of the phosphoinositide 3-kinase-related protein kinase (PIKK) family (Gobbini et al., 2013). The MRX/N complex functions as a sensor for DSBs and, after end binding, recruits and activates Tel1/ATM (Ciccia and Elledge, 2010). The RPA-coated ssDNA formed by end resection recruits Mec1-Ddc2/ATR-ATRIP (Zou and Elledge, 2003); in turn, the DNA damage checkpoint diminishes resection to prevent excessive ssDNA formation (Symington and Gautier, 2011; You et al., 2009; Shiotani and Zou, 2009). In yeast, end resection is rapid and the primary checkpoint signal in response to DSBs is from the Mec1; however, if resection initiation is delayed, for example, in the sae2Δ mutant, MRX is retained at ends for longer, Tel1 is hyper-activated and the mec1Δ checkpoint defect is partially bypassed (Usui, Ogawa, and Petrini, 2001). The substrates for Mec1 and Tel1 are largely overlapping and include yeast H2A (H2AX in metazoans), the downstream effector kinases Rad53/CHK2 and Chk1, Rad9/53BP1, Sae2/CtIP, Dna2 and RPA (Baroni et al., 2004; Ciccia and Elledge, 2010; Gobbini et al., 2013; Brush et al., 1996; Chen et al., 2011). Phosphorylated H2A/H2AX (γH2A/γH2AX) extends over large chromosomal domains and plays an important role in recruitment of other checkpoint proteins (Ciccia and Elledge, 2010).
The DNA damage checkpoint acts in both positive and negative ways to regulate end resection. The substitution with alanine of five of the Sae2 S/TQ sites, which conform to the consensus phosphorylation motif of Mec1 and Tel1, prevents DNA damage-dependent phosphorylation and the sae2-5A mutant is phenotypically similar to sae2Δ (Baroni et al., 2004). Phosphorylation of Thr-90 of Sae2 mediates interaction with the FHA domains of Xrs2, Rad53 and Dun1, whereas Rad53 and Dun1 additionally interact with phosphorylated Thr-279 (Liang, Suhandynata, and Zhou, 2015). The sae2-2AQ mutant (T90A, T279A) is more resistant to DNA damaging agents than sae2Δ and sae2-5A, but exhibits prolonged Rad53 activation, similar to the sae2Δ mutant. Thr-859 of human CtIP (CtIP-T818 of Xenopus), equivalent to Sae2 Thr-279, is phosphorylated in response to DSBs and is required for CtIP association with chromatin, DNA end resection and HR repair (Wang et al., 2013; Peterson et al., 2013). Phosphorylation of S. pombe Ctp1 by casein kinase 2 is required for its interaction with the FHA domain of Nbs1 (Lloyd et al., 2009; Williams et al., 2009). Phosphorylation of Sae2 is suggested to alter the oligomeric state and solubility of Sae2 in response to DNA damage (Fu et al., 2014). Dna2 is also phosphorylated in response to DNA damage but the functional consequences of this modification are currently unknown (Chen et al., 2011).
Although the DNA damage checkpoint is important for resection initiation, it acts in a negative fashion to limit accumulation of long tracts of ssDNA in the genome (Figure 4). Rad9 is an adaptor protein that is phosphorylated by Mec1 and/or Tel1 and is required for activation of the Rad53 kinase (Harrison and Haber, 2006). Rad9 interacts with dimethylated histone H3 (H3-K79) and γH2A in the vicinity of DSB sites forming a barrier to end resection by Exo1 and Dna2-STR (Chen et al., 2012; Lazzaro et al., 2008; Ngo and Lydall, 2015). Both chromatin modifications contribute to Rad9 recruitment and inhibition of end resection (Chen et al., 2012; Lazzaro et al., 2008; Eapen et al., 2012). Rad9 also prevents extensive resection by Exo1 and Dna2-STR at uncapped telomeres (Ngo et al., 2014; Ngo and Lydall, 2010). Elimination of Rad9 can suppress the sae2Δ resection initiation defect by allowing more efficient recruitment of Dna2-STR to DSBs (Bonetti et al., 2015; Ferrari et al., 2015). Similarly, partial loss of function alleles of TEL1 and RAD53 that reduce Rad9 accumulation at DSBs can also suppress the sae2Δ resection defect (Gobbini et al., 2015). Elimination of Tel1 causes a modest delay in resection initiation while the mec1Δ mutant exhibits increased end resection due to reduced Rad9 recruitment (Clerici et al., 2014; Mantiero et al., 2007). The mechanism by which Rad9 inhibits resection is not fully understood, but is dependent on chromatin association and oligomerization of Rad9 suggesting it forms a physical barrier to Exo1 and Dna2-STR (Ferrari et al., 2015). Activation of Rad53 could also contribute to the Rad9 inhibition of end resection (see below).
The 9-1-1 PCNA-like damage clamp binds to the junction of ssDNA and dsDNA at recessed 5′ ends and contributes to Mec1 activation, and thereby Rad9 recruitment to DSBs. When the 9-1-1 complex is removed (e.g., by mec3Δ mutation), extensive resection at DSBs is increased but not to the same extent as seen for the rad9Δ mutation (Ngo and Lydall, 2015). The increase in long-range resection observed in the mec3Δ mutant is SGS1, but not EXO1 dependent, suggesting 9-1-1 facilitates resection by Exo1. At DSBs and uncapped telomeres, 9-1-1 contributes to recruitment and activity of Exo1 and Dna2-STR; however Dna2-STR is more inhibited by Rad9 at DSBs than at telomeres (Ngo and Lydall, 2015; Ngo et al., 2014). Consequently, extensive resection at uncapped telomeres shows greater dependence on 9-1-1, whereas at DSBs the stimulatory function of 9-1-1 is mainly through Exo1 (Figure 4). Moreover, the purified human 9-1-1 complex stimulates end resection by EXO1 and DNA2 in vitro (Ngo et al., 2014). PCNA also facilitates recruitment of human EXO1 to damaged sites and promotes processive degradation by EXO1 in vitro (Chen et al., 2013).
53BP1, the metazoan ortholog of Rad9, acts in a similar fashion to inhibit end resection at DSBs and at telomeres (Zimmermann and de Lange, 2014; Sfeir and de Lange, 2012). The increased resection observed in the absence of 53BP1 is dependent on CtIP, but this could be through the role of CtIP in recruiting DNA2 (Hoa et al., 2015). 53BP1 inhibits end resection via three effector proteins: RIF1, PTIP and REV7/MAD2L2 that interact with ATM-phosphorylated 53BP1 (Chapman et al., 2013; Di Virgilio et al., 2013; Escribano-Diaz et al., 2013; Feng et al., 2013; Zimmermann et al., 2013; Callen et al., 2013; Boersma et al., 2015; Xu et al., 2015).
Another mechanism to prevent long-range resection is through inhibition of Exo1 independently of Rad9/53BP1. In budding yeast, Rad53 phosphorylates Exo1 and this phosphorylation is proposed to inhibit its activity (Morin et al., 2008). ATR-dependent phosphorylation of EXO1 in mammalian cells triggers its ubiquitylation and subsequent degradation following fork stalling (El-Shemerly et al., 2005); it is not known whether a similar mechanism operates to regulate EXO1 activity at DSBs. The HELB 5′-3′ ssDNA translocase was recently identified as a 53BP1-independent inhibitor of end resection in mammalian cells (Tkac et al., 2016). HELB is recruited to ssDNA by interacting with RPA and inhibits extensive resection by EXO1 and DNA2. Interestingly, HELB1 is exported from the nucleus during S phase, concomitant with the up-regulation of end resection. Expression of a variant of HELB lacking the nuclear export signal potently inhibited end resection, even in 53BP1−/− cells. Unlike 53BP1, HELB has no role in promoting NHEJ consistent with its function of inhibiting extensive resection and not resection initiation.
As noted in the Introduction, resection is reduced in G1 phase cells and is activated at the G1 to S transition to coordinate with DNA replication (Zierhut and Diffley, 2008; Ira et al., 2004; Aylon, Liefshitz, and Kupiec, 2004; Barlow, Lisby, and Rothstein, 2008). In yeast, low resection in G1 is due to both reduced CDK activity and Ku binding to DNA ends (Clerici et al., 2008; Ira et al., 2004; Aylon, Liefshitz, and Kupiec, 2004). Inhibition of CDK in G2/M phase-arrested cells inhibits end resection whereas activation of CDK in G1 phase promotes resection (Clerici et al., 2008; Ira et al., 2004; Aylon, Liefshitz, and Kupiec, 2004). Elimination of Ku in G1-arrested cells causes increased Mre11 and RPA association with DSBs and greater Mre11-dependent resection initiation, but long-range resection is not fully restored (Clerici et al., 2008; Barlow, Lisby, and Rothstein, 2008). By contrast, resection in G2-M phase arrested cells is unaffected by Ku. Inhibition of CDK in yku80Δ (Ku is encoded by YKU70 and YKU80 in yeast) G1 or G2-M arrested cells prevents long-range resection, but not short-range resection suggesting Ku controls short-range resection by modulating MRX and Exo1 access to ends, whereas high CDK activity is required for resection initiation when Ku is present as well as for extensive end processing (Clerici et al., 2008). In contrast to yku70Δ, the rad9Δ mutant does not exhibit increased resection in G1-arrested cells; however, in the yku70Δ rad9Δ double mutant both initiation and extensive resection can occur in G1-arrested cells (Trovesi et al., 2011). Since Dna2 is not active in G1 cells (see below), resection under these circumstances is most likely due to Exo1 activity.
Some resection factors are constitutively expressed throughout the cell cycle while others are regulated transcriptionally and/or post-transcriptionally. Expression of S. pombe Ctp1 is induced in S-phase and in response to DNA damage, whereas Sae2 is constitutively expressed (Limbo et al., 2007; Akamatsu et al., 2008; Baroni et al., 2004). In budding yeast and human cells, CDK-dependent phosphorylation of Sae2/CtIP residue S267/T847 is required for end resection (Huertas et al., 2008; Huertas and Jackson, 2009; Manfrini et al., 2010). Expression of the phosphomimetic S267E/T847E mutant of Sae2/CtIP is able to partially rescue the block to resection in G1 phase cells consistent with this modification playing a major role in the cell cycle control of resection (Huertas et al., 2008; Huertas and Jackson, 2009). The CtIPT847A/T847A homozygous mutation causes mouse embryonic lethality, suggesting that end resection is the essential function for CtIP in mammalian cells (Polato et al., 2014). CDK-dependent phosphorylation of CtIP is required for subsequent damage-induced phosphorylation by ATM and/or ATR ensuring activation of CtIP at the correct cell-cycle stage (Wang et al., 2013). Similarly, Nbs1-dependent DNA damage-induced phosphorylation of Ctp1 is eliminated by mutation of Ctp1 consensus casein kinase 2 sites (Lloyd et al., 2009). CDK-dependent phosphorylation of CtIP serine 327 is required for interaction with BRCA1, but this modification is not required for end resection or HR, and homozygous CtIPS327A/S327A mice develop normally (Polato et al., 2014; Reczek et al., 2013; Nakamura et al., 2010). CDK also phosphorylates yeast Dna2 to promote nuclear entry, association with DSBs and extensive resection (Chen et al., 2011). Although there is no evidence for CDK-mediated regulation of Exo1 in yeast, human EXO1 is phosphorylated by CDK, and substitution of the target serine residues with alanine reduces RPA recruitment to DSBs and HR repair (Tomimatsu et al., 2014). Transcription of EXO1 is elevated during meiosis in several organisms, consistent with its role in meiotic DSB processing (Zakharyevich et al., 2010; Szankasi and Smith, 1992; Digilio et al., 1996; Tishkoff et al., 1998).
NHEJ can occur throughout the cell cycle, but is favored in G1 cells when end resection activity is low; conversely, activation of end resection in S phase coordinates with HR. Although HR is considered to be a mostly accurate repair mechanism, this only applies to recombination between identical sister chromatids. Gene conversion between homologs carries the risk of local loss of heterozygosity (LOH), and if associated with a crossover can cause LOH from the point of exchange to the telomere. Furthermore, a crossover between repeats on different chromosomes results in translocation or more complex rearrangements. Thus, NHEJ might be considered a safer option than HR between non-sister chromatids, particularly in organisms with many repeated sequences.
Early evidence for competition between DSB repair pathways was obtained in mammalian cells where it was shown that elimination of Ku results in an increased frequency of DSB-induced HR between direct repeats (Pierce et al., 2001). Removal of Lig4 also increased HR, but not to the same extent as Ku, in agreement with the dual role of Ku in suppressing end resection and promoting NHEJ (Pierce et al., 2001). Even in G1-arrested yeast cells, Rad51-dependent gene conversion between ectopic repeats can be restored in the absence of Ku (Trovesi et al., 2011). Interestingly, crossovers were not detected in G1 yku70Δ cells, indicating an additional layer of regulation by CDK beyond end resection. Elimination of Sae2 or the Mre11 nuclease results in a much higher frequency of NHEJ (Huertas et al., 2008; Deng et al., 2014; Lee and Lee, 2007). Similarly, a small molecule analog of mirin, which specifically inhibits the MRE11 endonuclease activity and prevents formation of IR-induced RPA foci, promotes NHEJ in human cells (Shibata et al., 2014; Dupre et al., 2008). On the other hand, elimination of Rad51 or Rad52 in yeast does not increase NHEJ (Deng et al., 2014; Lee and Lee, 2007), consistent with the view that once resection initiates cells are committed to HR and NHEJ is not an option.
Resection initiation also creates substrates for repair by mutagenic MMEJ. While this mechanism is inefficient in yeast, unless microhomologies of >14 nt are close to the ends, it is quite frequent in metazoan cells (Sfeir and Symington, 2015; Deng et al., 2014; Villarreal et al., 2012). In one study that directly compared different repair outcomes in mammalian cells, MMEJ was shown to occur at 10–20% of the frequency of gene conversion (Truong et al., 2013). Consistent with the requirement for resection initiation, MMEJ is dependent on Mre11 and CtIP, whereas EXO1 and BLM are dispensable (Truong et al., 2013; Bennardo et al., 2008). Resection initiation by Sae2 and Mre11 nuclease is not limiting for MMEJ in yeast, instead RPA inhibition of annealing between microhomologies accounts for the low frequency of MMEJ (Deng et al., 2014).
In yeast, removal of Ku from DNA ends is thought to result primarily from MRX/N and Sae2/Ctp1 clipping activity (Langerak et al., 2011; Mimitou and Symington, 2010). A recent study provided evidence that ubiquitylation of Ku is an additional mechanism to eject Ku from DSBs in mammalian S/G2 phase cells. RNF138, a ubiquitin E3 ligase, was identified in a screen for positive regulators of HR (Ismail et al., 2015). RNF138 is recruited to DNA damage sites where it stimulates end resection and promotes HR repair thereby contributing to cellular resistance to DNA damaging agents. Elimination of RNF138 results in decreased ubiquitylation of Ku80 and retention of Ku at DSBs in S/G2 phase cells resulting in an increased frequency of NHEJ repair.
The tumor suppressors 53BP1 and BRCA1 play antagonistic roles in resection and repair pathway choice in mammalian cells. 53BP1 is enriched in the vicinity of DSBs in G1 phase cells via interaction with dimethylated histone H4 (H4-K20) and histone H2A ubiquitylated on K15. 53BP1 plays an important role in promoting NHEJ by preventing end resection and increasing chromosome mobility (Zimmermann and de Lange, 2014). As cells progress to S phase, 53BP1 is replaced by BRCA1, which interacts directly with MRN and CtIP (Chapman et al., 2012; Chen et al., 2008). Exactly how BRCA1 influences end resection is not well defined, but activation of end resection is concomitant with replacement of 53BP1 by BRCA1 on chromatin. Interestingly, the cell lethality, HR defects and chromosome instability associated with loss of BRCA1 are abrogated in the absence of 53BP1 by restoration of CtIP and ATM-dependent end resection (Bouwman et al., 2010; Bunting et al., 2010). Indeed, removal of 53BP1 alleviates the embryonic lethality and tumorigenesis associated with BRCA1 deficiency. By contrast, removal of Ku does not reverse the cell and embryonic lethality of BRCA1−/ − cells even though it does partially restore Rad51 foci and decrease radial chromosome formation (Bunting et al., 2010; Bunting et al., 2012).
ATM sites in the N-terminal region of 53BP1 are essential for function suggesting 53BP1 promotion of NHEJ and inhibition of BRCA1 recruitment requires phosphorylation-dependent protein interactions (Chapman et al., 2012). RIF1 and PTIP were identified as downstream effectors of 53BP1 that are recruited to different subsets of ATM-dependent phosphorylation sites (Callen et al., 2013). RIF1 and PTIP both repress end resection, but promote NHEJ in different contexts. RIF1 is more important for physiological NHEJ (immunoglobulin class switch recombination [CSR]) and less effective in promoting pathological NHEJ that occurs in the absence of BRCA1 or at unprotected telomeres. On the other hand, PTIP is dispensable for CSR and appears to promote the toxic NHEJ events that result from BRCA1 deficiency. PTIP interacts with Artemis, a nuclease required for hairpin opening and end processing during NHEJ (Wang et al., 2014). Similar to PTIP, Artemis depletion partially rescued the PARP inhibitor sensitivity of BRCA1-deficent cells by restoring Rad51 foci. REV7 is an additional factor that acts downstream of 53BP1 to repress end resection and promote NHEJ. Like loss of 53BP1, REV7 deficiency restores HR to BRCA1−/− cells (Boersma et al., 2015; Xu et al., 2015). The mechanism by which REV7 is recruited to DSBs is currently unknown, but does require the N-terminal ATM phosphorylation sites of 53BP1. Whether RIF1, PTIP and REV7 interact simultaneously with 53BP1 or form distinct sub-complexes with different activities remains to be determined.
BRCA1 interacts with BRCA2 and PALB2 to mediate RAD51 assembly at resected DSBs during S and G2 phases of the cell cycle. Elimination of 53BP1 allows BRCA1 recruitment in G1 phase cells, but not its binding partners PALB2 and BRCA2 (Escribano-Diaz et al., 2013; Orthwein et al., 2015). A recent study showed that PALB2 is ubiquitylated in G1 cells and this prevents its interaction with BRCA1 (Orthwein et al., 2015). Removal of the E3 ubiquitin ligase component, KEAP1, restores BRCA1-PALB2 interaction; however, KEAP1 is not cell cycle regulated raising the question of how the interaction between BRCA1 and PALB2 is recovered in S-G2. The USP11 deubiquitylase, which had previously been identified as a positive regulator of HR (Wiltshire et al., 2010), is rapidly turned over in G1 phase cells and the HR defect caused by its deletion is reversed by removal of KEAP1 (Orthwein et al., 2015). Although BRCA1-PALB2 interaction was restored in G1 phase 53BP1−/ − KEAP1−/− cells, radiation-induced RAD51 foci were not detected suggesting there is insufficient ssDNA formed for productive RAD51 nucleoprotein filament formation. Indeed, expression of the phosphomimetic T847E mutant of CtIP in these cells resulted in formation of damage-induced RAD51 foci and partial restoration of HR during G1 (Orthwein et al., 2015).
As described above, the chromatin surrounding DSBs undergoes extensive modification and several different chromatin remodelers are recruited to DSBs. ATP-dependent chromatin remodelers translocate on dsDNA disrupting histone-DNA contacts by nucleosome sliding, eviction or by histone exchange. Genetic studies in budding yeast demonstrated an important role for the RSC complex in recruitment of the MRX complex to DSBs and early resection, while Fun30 is important for extensive resection (Chen et al., 2012; Eapen et al., 2012; Costelloe et al., 2012; Shim et al., 2007). A role for the INO80 complex in early resection is only apparent in the absence of Fun30 and the RSC complex (Chen et al., 2012). While Fun30 facilitates both extensive resection mechanisms, the phenotype of fun30Δ is more similar to exo1Δ and over-expression of Exo1 suppresses the DNA damage sensitivity of the fun30Δ mutant (Chen et al., 2012; Costelloe et al., 2012). Additionally, SMARCAD1, the human ortholog of Fun30, is required for RPA localization to laser-induced DNA damage, similar to the role of EXO1 (Costelloe et al., 2012; Tomimatsu et al., 2012). Interestingly, the extensive resection defect of the fun30Δ mutant is completely suppressed by elimination of Rad9 suggesting Fun30 helps to overcome the resection barrier formed by Rad9 (Chen et al., 2012; Eapen et al., 2012). Nucleosomes assembled on a linear dsDNA template impede resection by Exo1 in vitro, but the inhibitory effect is less for Sgs1-Dna2, particularly if a nucleosome-free gap is adjacent to the DNA ends (Adkins et al., 2013). It will be of interest to examine end resection with nucleosomal substrates to determine whether the chromatin remodelers are required to move nucleosome prior to resection or allow resection in the context of nucleosomes.
A number of different cell types, including germ cells and developing lymphocytes, create DSBs in their genomes to initiate high frequency recombination. In meiosis, recombination intermediates physically link chromosome homologs to ensure their accurate segregation at the reductional division; recombination also contributes to diversity of haploid gametes (Kerr, Sarkar, and Arumugam, 2012). DSBs are introduced after S-phase when resection activity is high and are repaired solely by HR (Murakami and Keeney, 2014). The presence of Spo11 at DNA ends is a barrier to NHEJ and the coupling of Spo11 removal with creation of ssDNA tails ensures substrates are channeled to HR. In budding yeast, the removal of Spo11 oligos and Exo1-mediated resection are tightly coupled to DSB formation and consequently intermediates in the process are not detected, unless mutants are used (Zakharyevich et al., 2010). By contrast, the initiation of resection at accidental DSBs is the rate-limiting step and plays a central role in repair pathway choice.
DSB-induced genome rearrangements are essential for diversification of antibody and T-cell receptor genes in B and T lymphocytes, respectively (Schatz and Swanson, 2011). RAG1 and RAG2 (RAG) proteins induce DSBs at recombination signal sequences flanking V, D and J segments in G1-phase lymphocytes when resection activity is low and are repaired exclusively by NHEJ (Schatz and Swanson, 2011). RAG creates hairpin intermediates at coding ends that require processing by DNA-PK and Artemis in order for end joining to proceed (Ma et al., 2002). The processing of hairpin ends and addition of untemplated nucleotides by terminal deoxynucleotidyltransferase contribute to variable region diversification. In the absence of classical NHEJ factors, MMEJ or HR fail to repair RAG-induced DSBs due to RAG association with ends. However, in certain RAG2 mutants, repair can be channeled to alternative pathways indicating that the RAG proteins collaborate with NHEJ factors to ensure the most appropriate repair outcome (Corneo et al., 2007). The examples of budding yeast meiosis and lymphocyte recombination highlight the importance of coupling between the enzymes that create programmed DSBs and the repair pathway utilized to repair them.
A great deal of progress has been made in identifying the nucleases responsible for end resection in eukaryotic cells. Reconstitution of the early and late steps of resection has significantly advanced our understanding of this critical process but some key questions remain, including how Sae2 activates the Mre11 endonuclease, the role of BRCA1 in end resection and how human EXO1 overcomes the inhibitory effect of RPA. Furthermore, reconstitution of resection with nucleosomal templates incorporating modified histones is needed to fully understand how chromatin-remodeling complexes facilitate end resection. Although the basic machinery catalyzing end resection is conserved, it has become apparent in recent years that the control of this process is more tightly regulated in mammalian cells than in yeast, presumably because of the hazard of inappropriate HR in organisms with vast amounts of repetitive DNA in their genomes. Understanding the mechanisms by which RIF1, PTIP and REV7 inhibit end resection is an important area of future investigation.
I thank P. Ruff and W. Holloman for comments on the manuscript.
DECLARATION OF INTEREST
Research in the Symington laboratory is supported by grants from the NIH (R01 GM041784, R01 GM94386 and P01 CA174653). The author reports no declaration of interest.