To trim or not to trim

Abstract

DNA double-strand breaks (DSBs) are the most cytotoxic form of DNA damage, since they can lead to genome instability and chromosome rearrangements, which are hallmarks of cancer cells. To face this kind of lesion, eukaryotic cells developed two alternative repair pathways, homologous recombination (HR) and non-homologous end joining (NHEJ). Repair pathway choice is influenced by the cell cycle phase and depends upon the 5′-3′ nucleolytic processing of the break ends, since the generation of ssDNA tails strongly stimulates HR and inhibits NHEJ. A large amount of work has elucidated the key components of the DSBs repair machinery and how this crucial process is finely regulated. The emerging view suggests that besides endo/exonucleases and helicases activities required for end resection, molecular barrier factors are specifically loaded in the proximity of the break, where they physically or functionally limit DNA degradation, preventing excessive accumulation of ssDNA, which could be threatening for cell survival.

Introduction

DNA double-strand breaks (DSBs) are the most dangerous form of DNA damage. They can result from exposure to genotoxic agents used in cancer therapy, such as ionizing radiation or chemotherapeutic compounds (camptothecin or etoposide), but they occur also spontaneously, for example during replication of a damaged template or particular secondary DNA structures, which leads to replication fork collapse. Nevertheless, DSBs are also the natural molecular intermediates of important physiological events, such as V(D)J and class switch recombination in immune system development, and homologous recombination during meiosis; moreover, telomeres, the ends of eukaryotic chromosomes, are structurally related to DSBs. The formation of DSBs poses a significant threat to the cells survival, because if the DNA ends of a DSB are left unrepaired, they can lead to genome instability and large-chromosome rearrangements, hallmarks of cancer cells.¹

Eukaryotic cells have evolved two mechanistically distinct pathways to face this kind of DNA lesion: homologous recombination (HR) and non-homologous end joining (NHEJ). HR is the most accurate: it takes advantage of sequences elsewhere in the genome, homologous to the broken ends (sister chromatids, homologous chromosomes or repeated region on the same or different chromosomes) to prime repair synthesis in an error-free manner (for a complete review, see ref. 2). NHEJ involves direct ligation of the two broken ends, even if there is no complementary base pairing at the junction, resulting frequently in loss of genetic information (for molecular details, see ref. 3). Because a sister chromatid is the favorite template for HR, this pathway is mainly active during the S and G₂ phases of the cell cycle, whereas NHEJ, although being active throughout the cell cycle, is relatively more important during G₁.^4,5 The choice between the two repair pathways is thereby influenced by the cell cycle stage and it is governed by cyclin-dependent protein kinases (CDKs). This control is exerted mainly at the initial processing step of HR. Indeed, resection of double-stranded DNA ends to generate ssDNA tails promotes HR at the expense of NHEJ, which can only repair minimally processed DNA ends.^4-7

In the last few years, remarkable progress has been made in understanding the mechanism and dynamics of DSB ends resection. The emerging picture shows that, besides endo/exonucleases and helicases required for the 5′-3′ processing of the ends, inhibitory factors are loaded onto chromatin to finely tune DNA degradation, which may otherwise threaten genome integrity and cell survival.

Here we will provide an outline of the end resection mechanism in mitotic DSBs and in telomere homeostasis, reviewing the recent advances in the identification of protein factors involved in the negative control of this crucial aspect of DNA metabolism.

DNA End Resection: Not Only Nucleases

Homologous recombination begins with processing of DSB ends, which are subjected to an extensive 5′-3′ nucleolytic trimming, called resection, generating 3′ single-stranded DNA tails, which are immediately coated by the ssDNA binding protein RPA. ssDNA-RPA represents the molecular intermediate that promotes the loading of DNA repair factors, and the recruitment and subsequent activation of the apical DNA damage checkpoint kinase (Mec1-Ddc2 in S. cerevisiae, ATR-ATRIP in metazoa), triggering a checkpoint signal transduction cascade.^2,8 Nucleolytic degradation also occurs at telomeric ends after DNA replication to regenerate 3′-G overhangs, necessary for telomerase activity, or after the inactivation of the telomeric cap protection.⁹

A large amount of work has led to the elucidation of the key components of the DSBs end resection machinery and how this crucial process is finely regulated. The current model envisages a two steps mechanism: the first one relies on the partnership between the MRX/MRN complex and Sae2/CtIP to generate a short 3′ overhang; this will be extensively processed, in the second step, by the synergistic action of the Exo1 nuclease and the Sgs1-Top3-Rmi1 helicase-topoisomerase complex, together with the Dna2 endonuclease.^10,11 (Fig. 1).

Figure 1. Two-step model for DNA end resection at DSB sites. When a mitotic DSB is formed (A), the MRX complex (Mre11-Rad50-Xrs2) and Sae2 rapidly bind DNA ends. Their cooperative action allows the initial nucleolytic clipping of the ends which is required to clean “dirty” ends and accelerate resection of “clean” ends (B). The intermediate formed by the action of MRX/Sae2 (C) is then extensively processed through two parallel pathways governed by Exo1 and STR complex (Sgs1-Top3-Rmi1) in concert with Dna2 (D). Extensive end resection is prevented by the action of checkpoint adaptor Rad9 bound to methylated H3-K79 and phosphorylated H2A-S129, that acts as a barrier against nucleases, thereby inhibiting and attenuating DNA processing.

The MRX/MRN complex (Mre11-Rad50-Xrs2 in yeast, and Mre11-Rad50-Nbs1 in metazoa) is one of the first factors recruited at DSBs,¹² where it binds DNA ends, and it is necessary for triggering the DNA damage checkpoint.¹³ S. cerevisiae mrx mutants are deficient in processing meiotic DSBs,^14-16 are sensitive to ionizing radiation¹⁷ and display delayed resection of HO endonuclease-induced DSBs, even if the processing is not completely abolished.^18,19 Thus, for a long time, it has been hypothesized that MRX/MRN was the only factor responsible for DNA ends resection. Indeed, Mre11 possesses both a single-stranded endonuclease and an exonuclease activity in vitro.^13,20-22 However, the exonuclease activity operates in the 3′-5′ direction, opposite to the polarity of the end processing in vivo. Moreover, the Mre11 catalytic activity is not required for resection of HO-induced DSBs, although it is essential for processing ends blocked by covalent adducts—such as Spo11 or hairpin-capped ends—and it is required for cell survival at high doses of ionizing radiations (IR).^23,24 These data argue against MRX/MRN being the main nuclease for resection, suggesting that this complex works in collaboration with other factors.

The ideal candidate was Sae2. Although in budding yeast MRX and Sae2 do not physically interact, a sae2Δ mutant shares many phenotypes with the nuclease-deficient mre11 and rad50S mutants. All these strains are defective in the endonucleolytic removal of Spo11 from meiotic DSBs^25,26 and delay Rad52 foci formation after γ-irradiation or I-SceI- induced DSBs,¹² consistent with a role for these proteins in the early steps of HR. Indeed, Sae2 and MRX complex play a unique role in processing mitotic DSBs with terminal hairpin structures, both in vivo²⁷ and in vitro,²⁸ and participate together to the resection of HO-induced DSB ends.^29,30 CtIP, the human counterpart of Sae2, promotes ATR activation and HR by interacting physically and functionally with MRN in mediating DSB resection,³¹ suggesting that MRX/MRN-Sae2/CtIP coupling is an evolutionarily conserved mechanism.

The activity of Sae2 in end processing requires its CDK1-dependent phosphorylation on Ser267; a sae2S267A mutation abolishes this phosphorylation and phenocopies a sae2Δ mutant. On the contrary, the phospho-mimicking mutant sae2S267E rescues these phenotypes and strongly, although not completely, bypasses the requirement for CDK activity in DSB ends processing.³² This role of CDK1 in the control of resection is conserved also in human cells. Indeed, mutation of the corresponding residue in hCtIP to alanine (T847A) causes hypersensitivity to camptothecin (CPT)³² and affects ssDNA generation, RPA recruitment and RPA phosphorylation in response to CPT, laser-induced DNA damage or IR.³³

A fine analysis of DSB resection in various mrx mutants and sae2Δ cells showed that inactivation of these proteins causes only a delay in end processing²⁹; it impairs resection at early time points and at sites close to the break, whereas it does not significantly affect sites distal to the break.^34,35 Moreover, cells lacking both Sgs1 or Dna2 and Exo1, miss the functions responsible for the second step of resection and exhibit only weak nucleolytic processing, removing 100–200 nt from the 5′ strand proximal to the DSB, which depends upon MRX and Sae2.^34,35

Overall it can be hypothesized that the requirement for MRX complex and Sae2 in end processing depends upon the nature of DNA ends. The two players seem to be necessary for the processing of “dirty” ends, such as those created by a Spo11 cut or by IR and IR-mimetic agents, where protein-DNA adducts or altered DSB ends structures have to be removed before further processing can proceed. Interestingly, Sae2 itself can process hairpin DNA in vitro thanks to its ssDNA nuclease activity, and it is stimulated by MRX complex.²⁸ Thus it seems that Sae2 regulates Mre11 nuclease activity, but it can also act as a nuclease itself. Conversely, resection of “clean” ends, such as those produced by endonucleases, does not absolutely require MRX and Sae2. Indeed, in cells lacking MRX, the initiation of resection is partly impaired, but once nucleolytic degradation starts, it proceeds at a rate comparable to a wild type (wt) strain.³⁴ In this perspective, MRX-Sae2 are not essential for the starting of resection at every break site, but rather they accelerate the rate of initiation, which occurs stochastically and represents the rate-limiting step in resection biochemistry.^36,37 Their role may be to catalyze the removal of a short oligonucleotide through endonucleolytic cleavage, generating a short 3′ ended ssDNA region that might be redirected to extensive resection.^34,35

The second step in DSB resection is based on two distinct pathways, involving, respectively, Exo1 and Sgs1/Dna2. Exo1 is a conserved 5′-3′ exonuclease involved in several DNA repair pathways, including mismatch repair, stalled replication fork and UV-damage processing and DSB repair (DSBR) (for a review, see ref. 38). Despite exo1Δ cells being not particularly sensitive to ionizing radiation, the analysis of resection kinetics at HO-induced DSBs revealed that the exo1Δ mutant displays reduced resection rate compared with a wt strain.^24,39 Further investigations revealed that in the absence of Exo1, the initiation of end processing is comparable to a wt strain, whereas long-range processing is impaired,^30,34,35 suggesting a role for Exo1 in processive resection of the break. Analogously, depletion of hEXO1 from human fibroblasts decreases the amount of ssDNA formation as well as RPA recruitment at laser-induced breaks.⁴⁰ sae2Δexo1Δ and mre11Δexo1Δ mutants show a greater impairment of end processing and a higher sensitivity to DNA damaging agents than the single mutants, indicating genetic interactions.^39,41 In vitro reconstitution of the resection machinery with purified MRX, Sae2 and Exo1 revealed that the catalytic activity of the latter is strongly stimulated by the first;^42,43 at high Exo1 concentration, the stimulatory effect of MRX and Sae2 is less evident, consistently with in vivo data showing that overexpression of Exo1 can bypass the requirement for a functional MRX complex in the resection of a HO-induced DSB.^39,41,44 At limiting Exo1 concentrations, MRX and Sae2 stimulate 5′ resection by Exo1 by 60–300-fold; this effect does not require Mre11 nuclease activity, and is not mediated by the direct recruitment of Exo1 to DNA ends by MRX-Sae2. Rather, it appears that MRX-Sae2 generates a specific DNA structure that recruits very efficiently Exo1.⁴²

Several studies in human and yeast demonstrated that Exo1 activity is tightly controlled via DNA damage-induced phosphorylation. After hydroxyurea treatment, ATR-dependent phosphorylation of human Exo1 tags it for proteasome-mediated degradation,⁴⁵ whereas its ATM-dependent phosphorylation limits its role in repair.⁴⁶ Analogously, budding yeast Exo1 is phosphorylated on four sites in a checkpoint-dependent manner in response to telomere uncapping, CPT and HU (see below).^47,48 Mutation of these residues indicates that Exo1 phosphorylation limits the accumulation of ssDNA at unprotected telomeres and participates in a negative feedback loop to limit DNA damage checkpoint activation.⁴⁷

exo1Δmre11Δ and exo1Δsae2Δ double mutant cells exhibit residual resection and HR,^30,34,35,49 indicating that additional pathways of resection must exist. Indeed, three independent reports showed that Sgs1, a 3′-5′ RecQ helicase, is responsible for the residual DSB end resection in yeast.^34,35,50SGS1 deletion does not compromise end processing in the proximity (5–7 kb) of an HO-induced DSB, whereas it greatly affects nucleolytic processing 25 kb away from the cut site.^34,35 Simultaneous inactivation of Sgs1 and Exo1 abolishes long-range end resection, rendering cells hypersensitive to DSB-inducing agents and defective in DNA damage checkpoint signaling.^34,35,50 Moreover, Top3 and Rmi1, the Sgs1 partners in the STR complex (Sgs1-Top3-Rmi1), are also required for efficient end processing, demonstrating that STR acts as a functional unit in resection. Interestingly, the impairment of Top3 catalytic activity does not affect resection;⁴³ Top3-Rmi1 likely enhance Sgs1 affinity for DNA, facilitating its recruitment on DSB ends,^43,51 suggesting that Top3 and Rmi1 promote the helicase activity of Sgs1 in a structural than in a catalytic mode.

The ssDNA generated by Sgs1-dependent unwinding is then degraded by Dna2, a nuclease/helicase conserved among eukaryotes that is involved in Okazaki fragment processing, in DSB and post-replication repair pathways.^52,53 In the absence of Dna2, initiation of HO-induced resection is comparable to a wt strain, whereas long range resection is strongly defective. Moreover exo1Δdna2Δ double mutant cells are completely impaired in DSB processing, similarly to the exo1Δsgs1Δ strain, indicating that Sgs1 functions together with Dna2 in a pathway parallel to Exo1. Recently it has been reported that CDK1 phosphorylation of Dna2 stimulates its recruitment at DSBs thus influencing resection.⁵⁴

The function of Exo1, Sgs1 and Dna2 orthologs is conserved in human cells. hEXO1 and BLM, the human counterparts of Exo1 and Sgs1, function in parallel to promote DSB resection, DSB signaling and resistance to DSB-inducing agents.^50,55 Simultaneous depletion of BLM and EXO1 in U2OS cells treated with CPT significantly impairs RPA phosphorylation and focal accumulation, as well as the consequent ATR-mediated signaling.⁵⁰ Human DNA2 facilitates repair of replication-associated DSBs by promoting ssDNA formation at stalled replication forks,^56,57 acting redundantly with hEXO1.^48,57 In vitro data confirm that end resection occurs through two pathways, suggesting that the relationship between BLM-DNA2 and hEXO1 is not simply complementary. In the first one, BLM and DNA2 physically interact and collaborate to promote 5′-3′ resection, in a process requiring both BLM helicase and DNA2 nuclease catalytic functions. This process is stimulated by MRN, which seems to promote BLM recruitment to DNA ends, and it requires RPA, which increases the activity of BLM and guarantees the right directionality of DNA2, similarly to what observed for yeast proteins.^43,51 In the second pathway, the nucleolytic activity of hEXO1 is stimulated by MRN, RPA and BLM through physical association. In particular, BLM seems to enhance hEXO1 affinity for DNA ends, whereas MRN increases the processivity of the nuclease.⁵⁵

Chromatin Remodelers and End Resection

DNA repair, as all DNA transactions, occurs within the context of chromatin. Therefore, it is not surprising that cells evolved a complex network of post-translational histone modifications and ATP-dependent chromatin remodeling reactions to modulate chromatin structure and its accessibility within each step of DNA repair (reviewed in refs. 58–61). Here, we will discuss how DSBs resection is influenced by modulators of chromatin structure.

RSC

RSC (remodels the structure of chromatin) is an ATP-dependent chromatin remodeler of the SWI/SNF-family, implicated in a multitude of processes, among which transcriptional control, kinetochore formation, SAC adaptation, meiotic sporulation and cohesin loading.^62,63 RSC is also involved in multiple steps of the DNA damage response, as indicated by the sensitivity of rsc mutants to DNA damaging agents (MMS, phleomycin, bleomycin, HU, UV and IR).^64,65 RSC is necessary for the establishment of a normal nucleosome pattern in unperturbed conditions, suggesting that at least a small amount of RSC is constitutively acting on chromatin.⁶⁶ On the other hand, different subunits of the RSC complex are recruited at HO-induced DSB sites a few minutes after HO induction, suggesting a role in the early steps of DSB processing.^65,67 Indeed, it was later shown that, after break induction, RSC is required for nucleosome repositioning around the HO-induced break site to open the chromatin for the repair machinery.^66,68 Consistently, ChIP analysis demonstrated that Mre11 and Ku70 recruitment to DSBs is reduced in rsc mutants, as are ssDNA formation and RPA loading. Moreover, in a rsc2Δ strain, the enrichment of Mec1 and Tel1 in the proximity of the break, as well as H2A phosphorylation, is strongly reduced, likely as a consequence of limited ssDNA-RPA accumulation.^66,68,69 The current model predicts that upon DSBs induction, a small amount of Mre11 could be loaded at the break in a RSC-independent manner, facilitating further recruitment of RSC. This leads to nucleosome repositioning around the break, promoting DNA access to additional Mre11 molecules, which start resection and further stimulate RSC recruitment.⁶³

Ino80

Ino80 is a Swr1-like ATP dependent chromatin remodeler composed of 15 subunits in budding yeast, seven of which are conserved in human INO80.⁶³ Similarly to RSC, Ino80 has been linked to DSB repair: deletion of non-essential Ino80 subunits makes cells sensitive to HU, MMS, UV and IR.^70-73 However, direct involvement of Ino80 in DSBR became clear with the finding that the complex localizes at HO-induced DSB sites through a specific interaction with phosphorylated H2A.^70-72,74 Moreover, ino80 mutants accumulate less ssDNA compared with a wt strain, consistently with a mild resection defect,^71,72 probably due to a reduced MRX loading at the break.⁷² How Ino80 modulates the recruitment of the resection machinery is still unknown; however, it has been proposed that, by catalyzing nucleosome sliding or histone exchange, it may increase accessibility to the break.⁷⁵ Surprisingly, Ino80 accumulation at a DSB site is evident only 1–2 h after break induction, whereas MRX is loaded onto the damaged chromatin very rapidly. This apparent discrepancy can be explained by feedback-loop recruitment, where immediately after the formation of the break Mre11 binds the DSB ends independently of Ino80.^70,71,74 This would start DSB processing and Mec1 activation, thus allowing H2A phosphorylation and the consequent loading of Ino80 on chromosomes. Ino80-dependent chromatin remodeling might then increase the accessibility of the DSB to more MRX molecules that further increase end resection and Ino80 recruitment.⁶³

Ino80 involvement in DSB processing seems to be conserved among eukaryotes. Indeed, after ionizing radiation treatment, hINO80 localizes to damaged chromatin and it is required for homology-directed repair of an I-SceI induced break.⁷⁶ Similarly to what observed for yeast, the role of mammalian INO80 on HR repair could be ascribable to its modulation of DSB ends resection, since INO80-depleted cells show reduced ssDNA and RPA foci formation.⁷⁶

Fun30/SMARCAD1

Fun30 (in S. cerevisiae)/ SMARCAD1 (in human cells) is an ATP-dependent chromatin remodeler belonging to the Etl1 Snf2 family. As other family members, Fun30 binds nucleosomes and, thanks to its ATPase activity, facilitates the exchange of H2A-H2B dimers and the sliding of nucleosomes in vitro.⁷⁷ The ATP-dependent chromatin remodeling activity is required in vivo to establish gene silencing in heterochromatic regions, such as the HMR locus, rDNA repeats and telomeres,⁷⁸ and for the formation of correct architecture at centromeres.⁷⁹ While the sensitivity to CPT of fun30Δ cells was reported a while ago,⁷⁸ a possible role for this protein in DNA damage response has been investigated only recently. In the last year, three groups reported a role for Fun30 in resection of DNA ends. Deletion of Fun30 mildly affects nucleolytic processing in close proximity to an HO-cut site, whereas it strongly impairs resection further from the break, suggesting an involvement of Fun30 in controlling long-range resection.^80-82 Consistently with a role in DSB processing, Fun30 is recruited to DSB sites and spreads along the chromatin in both directions, similarly to the resection machinery. Moreover, epistasis analysis revealed that Fun30 promotes both Exo1- and Sgs1/Dna2- dependent resection pathways.^80-82 Despite the fact that its ATPase activity is necessary for CPT resistance and proper DNA ends processing, Fun30 does not promote resection by modulating histone occupancy. Indeed, the pattern of nucleosome positioning at multiple loci around the HO cut site is comparable in wt and in fun30Δ cells,^80,81 but it may facilitate resection machinery access to DNA within damaged chromatin. This would be consistent with the observation that in fun30Δ mutant Exo1 and Dna2 are recruited at the DSB site, but fail to spread along the chromatin. Another way Fun30 seems to stimulate long range resection is by removing the barrier represented by nucleosome-bound Rad9 (see below for the discussion on the barrier). Indeed, in fun30Δ cells Rad9 accumulation at DSBs is increased, and in rad9Δ or dot1Δ cells, where the barrier is lost,⁸³ Fun30 requirement is greatly reduced.⁸⁰

Similarly, the SMARCAD1 human counterpart is recruited to laser- and FokI-induced DSBs, where it colocalizes with γ-H2AX, with kinetics similar to that of hEXO1, suggesting a role in end resection. Indeed, SMARCAD1 downregulation affects DSB ends processing and reduces formation of RPA and ssDNA foci after laser microirradiation-induced DSBs. In agreement with a resection defect, SMARCAD1-knockdown cells are defective in recombinational repair and are hypersensitive to CPT or ABT-888 (a PARP inhibitor) treatment.⁸¹

Resection in the Telomeric Context

While telomeres resemble one-half of a DNA DSB, their unique nucleoproteic structure distinguishes them from DSBs and makes them stable (reviewed in refs. 9 and 84). The basic structure of telomeres in eukaryotic cells consists of G-rich repetitive DNA sequences terminating with a 3′ ssDNA overhang, known as G-tail, caused by the protrusion of the G-strand over its complementary C-strand. The G-tail is of crucial importance for telomere homeostasis, since it serves as substrate for telomerase; it also represents the docking site for ssDNA binding proteins that, together with dsDNA binding proteins, protect chromosome ends from nuclease degradation and end-to-end fusions.

The requirement for the G-tail poses a problem following genome replication, since the leading-strand polymerase generates blunt ended chromatids. Different evidences indicate that the G-rich overhangs at telomeres are derived through a controlled 5′-3′ nucleolytic processing of the C-strand⁸⁵ that occurs only during the S and G₂ phases of the cell cycle, when CDK1 activity is high.^86,87 The activities responsible for telomere processing share common players with the machinery resecting DNA ends at DSBs. MRX/MRN is the major responsible for the generation of the G-tail at telomeres, where its activity is stimulated by the PIKK kinase Tel1.⁸⁸ Indeed, mrx mutants are defective in G-tail formation in a de novo telomere addition reaction^89,90; moreover, both mrx mutants and MRN-depleted cells exhibit shorter G-tails.^91,92 Further investigations indicated that MRX and Sae2 act together in the processing the C-strand, even though MRX may give a greater contribution when Sae2 is absent.⁹⁰ As for DSBs resection, Sae2 function in a telomeric context needs CDK1-dependent phosphorylation on Ser267; however, other unknown CDK1 targets might contribute to modulating telomere processing, in fact the phospho-mimicking sae2S267D allele does not bypass the CDK1 requirement for degradation of the C-rich strand.⁹⁰ Loss of Sae2 inhibits but does not abolish ssDNA telomeric generation; indeed, in sae2Δ cells residual resection depends upon the action of Exo1 and Sgs1 in cooperation with Dna2.⁹⁰ On the other hand, inactivation of Exo1, Sgs1 or Dna2 does not prevent telomeric resection, if MRX/Sae2 are functional, suggesting that Exo1 and Sgs1-Dna2 could represent a back-up mechanism for nucleolytic processing.^10,90 Since the MRX complex is present only on the leading-strand telomere,⁹³ another option is that MRX/Sae2 may process the leading-strand telomere, whereas Exo1 and Sgs1-Dna2 may extend the gap created by the removal of the farthest RNA primer on the lagging-strand telomere¹⁰ (Fig. 2A).

Figure 2. Resection in the telomeric context. (A) Resection at telomere ends. Replication of the lagging strand, after the removal of the farthest RNA primer, results in chromosomes with short a 3′ overhang. On the other hand, replication of the leading strand generates blunt end chromatids. In this case, the G-rich overhangs are formed through a controlled 5′-3′ nucleolytic processing of the C-strand that is performed primarily by the action of MRX in concert with Sae2, whereas Exo1 and Sgs1-Dna2 represent a back-up mechanism for resection. Alternatively, MRX/Sae2 may process the leading strand telomere, whereas Exo1 and Sgs1-Dna2 may extend the gap created by the removal of the farthest RNA primer. (B) Resection at uncapped telomeres. Chromosomal DNA ends are protected by nucleoproteic structures containing proteins that specifically bind the dsDNA (e.g., the shelterin -like complex composed by Rap1, Rif1, Rif2 and Yku) and ssDNA tracts (e.g., the CST complex, composed by Cdc13, Stn1, Ten1). The essential capping function is performed by the CST complex. If telomeres become uncapped, they undergo extensive resection with the accumulation of ssDNA and, ultimately, activate the DNA damage checkpoint. Exo1 is the major nuclease responsible for ssDNA accumulation in a cdc13-1 mutant background, whereas MRX complex together with Sae2 rather prevent end processing. The helicases Sgs1 and Pif1 play redundant roles along with Exo1, likely unwinding telomeric DNA and thus exposing ssDNA tracts further processed by other nucleases.

The telomere cap is a nucleoproteic structure in which the telomeric repeats are packed together with ssDNA and dsDNA binding proteins, ensuring that telomeric ends do not inappropriately activate a DNA damage response, thus behaving differently from DSBs. A large number of proteins contribute to this structure; in budding yeast, the telomere cap contains Rap1, Rif1, Rif2, Ku70, Ku80 and the CST complex (Cdc13, Stn1 and Ten1).⁸⁴ Inactivation of any of these factors leads to telomere “uncapping”, which causes resection of telomeric DNA, accumulation of ssDNA and, ultimately, DNA damage checkpoint activation, similarly to what happens at DSBs. The temperature-sensitive cdc13-1 allele has been extensively used to investigate the molecular determinants of resection at uncapped telomeres.^94-98 Although the factors involved are the same as in DSBs processing and in the resection of normal telomeres, the hierarchical relationships between these factors are somewhat different. The most striking difference is that, whereas MRX/Sae2 initiate DSB end resection, they do not contribute to nucleolytic processing of uncapped telomeres; rather, they protect telomeric DNA.^96,97 Differently from what happens during C-strand degradation at normal telomeres, the principal nuclease required for processing of uncapped chromosomes is Exo1.^95,99 Indeed, EXO1 deletion reduces both the accumulation of ssDNA and the temperature sensitivity of a cdc13-1 strain,^95,99 indicating that Exo1 is detrimental to cell growth in these mutants. In response to telomere uncapping, Exo1 is phosphorylated at various sites in a checkpoint-dependent manner.⁴⁷ Mutation of these phosphorylation sites to alanine negatively affects growth of cdc13-1 cells, whereas their conversion into the constitutive phospho-mimicking glutamic acid has the opposite effect, suggesting that damage-dependent Exo1 phosphorylation has an inhibitory role.⁴⁷

Concurrent elimination of Exo1 and Sgs1 has a negative synergistic effect on ssDNA accumulation, suggesting that Exo1 and Sgs1 are also partially redundant at uncapped telomeres.⁹⁷ Finally, a role for the 5′-3′ helicase Pif1 has recently emerged, given that deletion of EXO1 and PIF1 completely suppresses the cdc13-1 phenotypes.⁹⁸ The current model predicts that Pif1 unwinds uncapped telomeric DNA, exposing ssDNA that is processed by a still undefined endonuclease, which acts redundantly with Exo1^84,98 (Fig. 2B).

The different requirements for MRX at DSBs and uncapped telomeres may be explained considering the different nature of the DNA ends produced in these two contexts. Whereas after inactivation of the telomere cap, a 3′ssDNA overhang is naturally exposed, the HO-induced ends require an initial processing (generally MRX-Sae2-dependent) to generate a suitable substrate for Exo1.⁸⁴ This interpretation is further corroborated by findings obtained with purified proteins.⁴²

Regulation of Resection: Molecular Barriers

In budding yeast, resection is estimated to proceed at approximately 3.5–4 kb/h at different chromosomal locations.^36,100 The length of ssDNA regions generated varies significantly with the availability of homologous sequences: in meiotic cells, the average length is 850 nt, whereas after a mitotic DSB, the amount of ssDNA produced is around 2–10 kb, when a homologous partner is available,¹⁰¹ but it can extend much longer when the homologous donor sequences are absent.³⁴ End resection is needed to uncover homologous sequences, which will then be engaged in HR repair; moreover, the ssDNA regions generated by DSB processing represent the signal that triggers the Mec1/ATR-dependent response.^5,8 The extent of resection needs to be kept under tight control to generate enough substrate to allow efficient repair, avoiding excessive ssDNA formation. A number of recent studies showed that end resection is finely modulated not only at the level of nucleases and DNA processing enzymes, but also through the presence of functional or structural “barriers.”

Ku

As anticipated above, once a DSB occurs, the cell must choose between HR and NHEJ repair to fix the break. The choice is made at the level of end resection, since the generation of ssDNA strongly stimulates the HR pathway and inhibits end joining. HR repair, and thus end resection, is usually confined to the S and G₂ phase of the cell cycle, when DNA has been replicated and a sister chromatid is available as homologous repair template.^4,5 Conversely, in G₁, NHEJ is the predominant pathway of DSB repair. For this reason, HR and NHEJ are often considered to be in competition with each other, and this rivalry is mirrored also at a molecular level. Indeed, the major restrain to end resection is represented by binding to DNA ends of the Ku complex, the core component of NHEJ machinery. In S. cerevisiae, Ku is a heterodimer encoded by YKU70 and YKU80, which forms a ring that interacts with DNA ends. Once bound, it protects ends from degradation, and it mediates the recruitment of other components of the NHEJ machinery.¹⁰² The main role of Ku at DSB site seems to be the inhibition of resection, achieved by opposing Exo1 recruitment. Indeed, deletion of YKU70 increases Exo1 binding at a DSB site and accelerates end processing, which in yku70Δ cells is due to increased Exo1 activity.¹⁰³ Consistently, YKu inhibits 5′ strand degradation by Exo1 also in a reconstituted in vitro system.¹⁰³ Interestingly, elimination of YKu partially suppresses IR hypersensitivity in mre11Δ and sae2Δ mutants, but only when Exo1 (and, in the case of the sae2Δ mutant, also Sgs1) is present, in agreement with the idea that in the absence of MRX/Sae2, Ku blocks access to Exo1 and Sgs1.^103-105

In yku70Δ cells, nuclease-mediated end resection can occur in the G₁ phase of the cell cycle, when it is usually inhibited, although it is limited to the region proximal to the break site.^{36,103,106,107} This observation suggests that, in G₁, DNA processing is prevented by the YKu complex bound to the DSB ends. This inhibitory effect was observed also for the other components of NHEJ machinery, Dnl4 and Lif1, although to a lesser extent, suggesting that both end blocking and ligation functions, mediated by Ku and other NHEJ factors, contribute to end protection.^36,106,107 ChIP analysis revealed that in yku70Δ cells Mre11 association at DSBs is enhanced.^106,108 Altogether, these evidences indicate that YKu and MRX compete for end binding, and when YKu succeeds in occupying the ends, it impairs the loading and/or the activity of resection factors, thus limiting the formation of ssDNA. The competition between MRX and Ku works also in the opposite direction: MRX inhibits Ku binding at DNA ends. In fact, loss of MRX increases Ku binding to DNA,^103,108,109 and dissociation of Ku from DSB ends in vivo is dependent on MRX and correlates with bulk resection.¹⁰⁹

In G₂-arrested cells, loss of YKu does not further stimulate end resection at HO-induced DSBs; however, its overexpression reduces Mre11 recruitment and, consequently, delays nucleolytic processing.¹⁰⁶ As previously mentioned, CDK1 activity is necessary to achieve effective resection; indeed, the conditional overexpression of the Sic1 inhibitor leads to a significant reduction of 5′-3′ processing.^4,5 Interestingly, the CDK requirement is bypassed by deletion of YKu, allowing the accumulation of short-range resection products. This suggests that YKu is the principal rate-limiting factor for initiation of resection in G₁, and its action is blocked in G₂ by CDK-dependent phosphorylation events.¹⁰⁶ Consistently, NHEJ protein accumulation at HO break sites is more efficient in G₁ than in the G₂, and this recruitment is inhibited by CDK activity.¹¹⁰ It is has to be clarified, though, that the role of CDK in end resection is not limited to this aspect: both in wt and yku80Δ cells, CDK is required to promote extensive resection by modulating the action of several proteins, such as Exo1, Sgs1 and/or Dna2 (see ref. 106 and see above).

YKu inhibits resection also at telomeres, where it contributes to telomere capping together with the CST complex and the shelterin-like proteins Rif1, Rif2 and Rap1.⁸⁴ Indeed, at high temperatures, telomeres of cells lacking YKu become unprotected and undergo Exo1-dependent resection, which causes checkpoint-dependent cell cycle arrest, similarly to what is described for a cdc13-1 strain. However, resection at uncapped telomeres lacking Ku is less extensive than in cdc13-1 cells, and accumulation of ssDNA requires multiple cell cycles.⁹⁹ Recent evidence highlight that YKu prevents 5′ C-strand resection in the G₁ phase of the cell cycle, but not in G₂, at both native and HO-induced telomeres. The inhibitory effect of Ku is independent of its role in NHEJ, as Dnl4 is not required for end protection, contrary to what happens at a DSB site. Interestingly, the accumulation of ssDNA in YKu-deficient cells is again Exo1-dependent, consistently with a role of YKu in Exo1-inhibition.¹¹¹

Rad9

RAD9 was the first checkpoint gene to be identified in S. cerevisiae.¹¹² Later studies recognized it as an “adaptor” protein in the DNA damage checkpoint signal transduction cascade, linking the upstream kinase Mec1 to the activation of effector kinases Rad53 and Chk1 (for a complete review on DNA damage checkpoint in budding yeast, see ref. 113). Through its tandem Tudor domain, Rad9 physically interacts with the methylated K79 residue of histone H3,¹¹⁴ and this interaction allows its loading onto chromatin in unperturbed conditions.¹¹⁵ This constitutive binding to H3, together with a DNA damage-dependent interaction with phosphorylated H2A and with a cell-cycle regulated interaction with Dpb11 are necessary for efficient checkpoint arrest, as well as for cell survival after genotoxic treatments, and also for proper execution of the DDR at uncapped telomeres.^83,115-120 The function of Rad9, on the other hand, goes beyond its adaptor role in signal transduction. A seminal paper indicated that checkpoint factors affect both ssDNA production at uncapped telomeres and cell cycle arrest.¹²¹ In particular, whereas the checkpoint genes RAD17, RAD24 and MEC3 seem to promote degradation of dsDNA at and near telomere repeats, RAD9 prevents the accumulation of ssDNA at uncapped telomeres.¹²¹ This inhibitory function of Rad9 was later reported also at sites of damage.⁸³ Indeed, in rad9Δ cells, ssDNA is produced more rapidly at HO-induced breaks, and Mec1 kinase activation is anticipated compared with wt cells. These findings suggested that Rad9 may act as a physical or functional barrier toward nucleases responsible for end processing (Fig. 1). This “end protection” role of Rad9 is strictly dependent on its binding to methylated H3K79, since abrogation of this methylation or mutation of the Tudor domain recapitulate the phenotypes observed in the absence of Rad9.⁸³ In the same study it was shown that Rad9 binding to methylated H3K79 is also required to inhibit ssDNA accumulation at uncapped telomeres, revealing that this mechanism of resection control is conserved in different breaks contexts. Surprisingly, lack of Rad9 bypasses the requirement for CDK1 activity in HO-induced end resection, suggesting that Rad9 may restrict a CDK-independent nuclease, which seems to be MRX-dependent; indeed, loss of Rad9 does not affect resection in rad50∆ mutant cells⁸³ and suppresses the inhibitory effect of elevated levels of Cdc5 kinase on DSB processing.¹²² Consistently, it has been recently observed that the Fun30 chromatin remodeler, which promotes Exo1 and Sgs1/Dna2-dependent resection, is able to overcome the Rad9 barrier to resection. In fact, deletion of RAD9 suppresses the resection defects in fun30Δ cells.⁸⁰ On the other hand, at uncapped telomeres, where both MRX and Sae2 play a protective role rather than contributing to resection, Rad9 seems to inhibit an Sgs1-dependent nuclease, which is likely Dna2.⁹⁷ The molecular mechanism through which Rad9 prevents inappropriate end processing has not been thoroughly investigated in budding yeast, but recent studies in human and mouse cells helped to clarify how the barrier may work.

53BP1 and RIF1

The structural and functional vertebrate ortholog of Rad9 is represented by 53BP1 (p53 binding protein 1), a well-established DDR protein involved both in DNA damage signaling and repair.¹²³ Rad9 and 53BP1 share a similar domain organization at the C terminus, comprising a tandem Tudor domain followed by a tandem BRCT domain. 53BP1 also contains a homo-oligomerization domain, a GAR (glycine-arginine rich) domain and 28 S/TQ sites, potential targets of PIKK-mediated phosphorylation at its N terminus.¹²³ Similarly to Rad9, 53BP1 is constitutively bound to chromatin through the interaction between its Tudor domain and methylated H4K20. Following ionizing radiation, 53BP1 is rapidly relocalized in focal structures, which act as platforms for DNA repair and checkpoint factors recruitment. This process is dependent on PIKK-dependent phosphorylation of histone H2AX and on the consecutive chromatin ubiquitylation cascade orchestrated by the E3 ubiquitin ligases RNF8 and RNF168.^124,125

The DNA repair role of 53BP1 was initially highlighted by the observation that 53BP1 deficiency is associated with IR hypersensitivity, immunodeficiency and increased chromosome instability.^126-130 However, cells lacking 53BP1 do not display any defect in HR¹²⁷; rather, it seems that the protein promotes specific subsets of NHEJ events in different DNA break contexts, including DSB repair within heterochromatin,¹³¹ telomere uncapping,¹³² immunoglobulin class switch recombination (CSR) and V(D)J recombination.^127,133,134 Indeed, seminal studies demonstrated that 53BP1 deficiency in mice B cells results in the complete loss of long-range CSR with a consequent reduction in the concentration of switched serum IgH isotypes, caused by an increased nonproductive intra-switch region recombination within the IgH locus.^127,133,135 Likewise, V(D)J recombination between distal gene segments is impaired in 53BP1-defective lymphocytes, causing lymphopenia in 53BP1^−/− mice.¹³⁴ Finally, in MEFs, 53BP1 promotes trans-chromosomal fusions of unprotected telomeres lacking the protective TRF2 shelterin subunit.¹³²

53BP1 function in mediating NHEJ was initially ascribed to its capacity to facilitate tethering and synapsis of distant regions, thanks to its constitutive interaction with methylated histone H4. In this view, 53BP1 could increase the chance of damaged DNA ends to come together, thereby allowing their repair by NHEJ.^127,132-135 However, further investigations showed that 53BP1 facilitates NHEJ at the expense of HR, also through a mechanism which consists in the protection of DNA ends from inappropriate 5′ resection, similarly to what Rad9 does in budding yeast. Indeed, in MEFs, in the absence of 53BP1 the overhang signal of shelterin-lacking telomeres is enhanced, in agreement with a role of 53BP1 in thwarting end processing.¹³⁶ Moreover, the absence of BRCA1 leads to failure to dislodge 53BP1 and downregulates NHEJ in S phase, causing the formation of chromosomal abnormalities linked to replication-associated DNA breaks; consistently, loss of 53BP1 restores HR repair in Brca1^∆11/∆11cells¹³⁷. The authors suggested that this can be explained by the fact that, in the absence of Brca1, the lack of 53BP1 promotes end resection of the replication-associated breaks. Enhanced processing would generate ssDNA competent for HR-mediated DNA repair. An antagonistic role of BRCA1 and 53BP1, and the re-establishment of an active HR by the loss of 53BP1 may explain the rescue of phenotypes typical of Brca1 mutants, such as embryonic lethality, tumorigenesis, chromosomal aberrations and PARPi and CPT sensitivity.¹³⁷ Loss of 53BP1 correlates with extensive resection, which, in turn, promotes microhomology-based A-NHEJ (alternative-NHEJ) at the expense of classic NHEJ, also at the TCRα locus during V(D)J recombination in mice thymocytes,¹³⁴ and in B cells after I-SceI induced DSBs at IgH locus during CSR.¹³⁸ Consistently, a recent study reported that 53BP1^−/− IgkAID lymphomas present wide deletions in the IgH locus, and the TC-Seq (translocation capture sequencing) analysis of 53BP1^−/− B cells pointed out that in this background a higher number of resected chromosome breaks accumulates compared with wt cells.¹³⁹

Detailed analysis of the contribution of 53BP1 domains to the modulation of DSB repair pathways revealed that the Tudor domain, the oligomerization domain and the N terminus of 53BP1 are necessary to prevent nucleolytic processing after I-SceI-induced DSB at IgH locus, while BRCT domains are dispensable.¹⁴⁰ The same regions of 53BP1 are required to protect DNA ends from processing also in Brca1 mutant cells after treatment with PARP inhibitors.¹⁴⁰ However, a recent study highlighted that the requirement for 53BP1 in mediating end protection changes in different contexts (i.e., CSR, Brca1⁻ cells and at dysfunctional telomeres). Indeed, at telomeres of cells conditionally lacking TRF2, the 53BP1 oligomerization domain is dispensable to prevent CtIP-dependent resection, while the Tudor domain and the ATM phosphosites at the 53BP1 N terminus are critical.¹⁴¹ The disparity between different break contexts may be due to variations in the end joining mechanism and/or to cell-specific requirements for 53BP1.¹⁴¹

The importance of the ATM phosphorylation sites in 53BP1 for its role in controlling of DSB repair pathways suggested the involvement of an effector protein. In this perspective, the genome stability factor RIF1 (Rap1-interacting factor 1) was a good candidate, since its accumulation at DSB sites is ATM and 53BP1-dependent.^142,143 Recent reports demonstrate that 53BP1 and RIF1 physically interact in an ATM-, phosphorylation- and DNA damage-dependent manner.^143-145 Furthermore, the 53BP1^28A, the 53BP1^20AQ and the 53BP1^15AQ alleles, which carry alanine substitutions at specific S/TQ sites, prevent RIF1 accumulation at deprotected telomeres and at DSB sites.^143-146 Several recent results support the notion that Rif1 mediates the function of 53BP1 in modulating DSB repair pathway choice. RIF1 knockout leads to an increase of the telomeric overhang caused by loss of shelterin protection, which is dependent on CtIP, BLM and Exo1; this effect is similar to that detected in 53BP1 deficient cells.¹⁴⁶ Analogously, loss of RIF1 or 53BP1, results in an increase in zeocin-induced RPA foci and IR-induced RPA and Chk1 hyperphosphorylation, indicating that RIF1 inhibits end resection also at DSBs sites.^143,146 Interestingly it was shown that, when either 53BP1 or RIF1 are absent, RPA foci increase also in unperturbed cells, probably due to excessive processing of stalled replication-induced breaks.¹⁴⁶ The same 53BP1-RIF1 paradigm holds true also for immunoglobulin CSR in B cells, since the ablation of 53BP1 or RIF1 impairs CSR and results in the accumulation of ssDNA-binding protein RPA and of Rad51, both at the IgH locus and at AID (activation-induced cytidine deaminase)-damaged genes, as a result of increased DNA-end resection.^143-145,147

An intriguing recent paper shows that 53BP1 antagonizes the formation of IR-induced BRCA1 foci in G₁. This role of 53BP1 requires its binding to methylated histone and its phosphorylation by ATM. The critical effector is again Rif1, which, by binding to phosphorylated 53BP1, is recruited at the damage sites and prevent BRCA1 accumulation This function of 53BP1 and RIF1 is important to facilitate NHEJ in G₁ cells; indeed, genetic data show that RIF1 acts downstream of 53BP1 in the same pathway as Ku; moreover, loss of Rif1 causes a G₁-specific hypersensitivity to IR¹⁴⁴.

Altogether, these data lead to a model where 53BP1 and RIF1 antagonize BRCA1/CtIP, modulating DSB repair pathway choice (Fig. 3). In G₁ cells 53BP1 and RIF1 protect DSB ends from BRCA1-dependent resection, while in S-G₂, BRCA1-CtIP prevent RIF1 accumulation at DSBs, likely by promoting initiation of end resection. In this view, lack of BRCA1 would allow 53BP1 and RIF1 to persist at DSBs even in S-G₂, downregulating resection and HR. Accordingly, loss of RIF1 suppresses BRCA1 deficiency phenotypes.¹⁴⁴ Given the functional parallels between 53BP1 and Rad9, and between SMARCAD1 and Fun30, it will be very interesting to investigate the regulatory circuits that may allow SMARCAD1 to promote resection and HR by controlling 53BP1-RIF1 persistence at broken DNA ends. How does RIF1 execute its function is presently unclear. Last year several papers reported that, during S phase, RIF1 is important to preserve the proper timing of DNA replication, and it was suggested that RIF1 may have a negative effect on the firing of a subset of replication origins by modulating the binding of replication factors.^148-150 During DDR or at unprotected telomeres, RIF1 also seems to affect the recruitment of checkpoint factors to damaged chromatin, thus playing an anticheckpoint role.^151-153 In all these processes, RIF1 may be acting by directly controlling the recruitment of specific factors to DNA or by modulating chromatin accessibility to these factors. The identification of the molecular mechanism through which RIF1 antagonizes BRCA1 and modulates the choice between HR and NHEJ is of the utmost relevance, since it may unveil important targets for cancer therapy.

Figure 3. 53BP1 and RIF1 inhibit DNA end processing 53BP1 bound to methylated H4-K20 acts as a barrier against nucleases responsible for processing DSBs. This function is mediated by its interacting partner Rif1. In the presence of a DSB, ATM phosphorylates 53BP1 and creates a docking site for the recruitment of Rif1 the chromatin. The association of Rif1 and 53BP1 is crucial for inhibiting extensive end resection.

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.