At Loose Ends: Resecting a Double-Strand Break

Abstract

Double-strand break (DSB) repair is critical for maintaining genomic integrity and requires the processing of the 5′ DSB ends. Recent studies have shed light on the mechanism and regulation of DNA end processing during DSB repair by homologous recombination.

The detection and repair of double-strand breaks (DSBs) is essential for cell survival, as unprocessed DSBs can lead to chromosomal rearrangements such as duplications, translocations, and deletions. All of these types of DNA rearrangements are precursors to genome instability and tumorigenesis. Therefore, cells have evolved different mechanisms to process DSBs depending on the type of DNA damage as well as the phase of the cell cycle in which the damage is detected. One of the most important steps in DSB repair after sensing the presence of damaged DNA is deciding which specific pathway to use to process the lesion. The commitment to a specific DNA-repair pathway can have pro-found consequences because some repair mechanisms are error prone. One conservative mechanism to repair a DSB is homologous recombination, which uses a homologous template to restore lost information at the break site, resulting in repair that is generally more accurate than with other mechanisms. On the other hand, processes such as non-homologous end joining and microhomology-mediated end joining directly religate DNA ends at the break and can be error prone if DNA bases are lost at the break site.

How do cells determine which pathway to use to repair DNA and how are the DNA break ends processed to initiate recombination? With the recent identification of key players in the initiation steps of DSB repair (Gravel et al., 2008; Hopkins and Paull, 2008; Huertas et al., 2008; Mimitou and Symington, 2008; Nimonkar et al., 2008; Zhu et al., 2008), we are moving closer to answering these questions. Given the importance of DSB repair, it is not surprising that multiple pathways exist for processing 5′ break ends during homologous recombination.

Sae2 and Cell-Cycle Control of DSB Processing

After a DSB occurs, the first group of proteins recruited to the ends of the break site is the MRX/MRN complex (Figure 1), which consists of Mre11-Rad50-Xrs2 (MRX) in budding yeast and Mre11-Rad50-Nbs1 (MRN) in fission yeast and mammals. The MRX/MRN complex does the initial processing of the DNA ends to enable cells to engage in the appropriate repair pathway. If cells are in the G1 phase of the cell cycle, the nonhomologous end joining pathway is preferentially used (Frank-Vaillant and Marcand, 2002). The Ku70-Ku80 protein heterodimer is loaded onto the DNA ends, stabilizing the ends and preventing resection. The ends can then be religated, a process that requires the MRX proteins and the DNA ligase activities of the Dnl4-Lif1/XRCC4 heterodimer and the protein Nej1/XLF. In mammalian cells, the MRN complex is not as important for nonhomologous end joining, but there are additional players such as Ligase 4, XRCC4, DNA-PKCS, and ARTEMIS proteins. If cells are in the S or G2 phase of the cell cycle when a DSB is detected, repair occurs by the homologous recombination pathway, which preferentially uses a homologous template from either the sister chromatid or the homologous chromosome to repair the damage (Figure 1).

Figure 1. DNA End Resection during Double-Strand Break Repair.

Recent studies characterize a two-step mechanism for the processing of double-strand breaks (DSBs) at the 5′ ends to expose the 3′ single-stranded DNA (ssDNA) overhangs. Depending upon the cell-cycle phase and the type of DNA lesion, a DSB is processed by either the nonhomologous end joining pathway or the homologous recombination pathway. The names of the human homologs of the yeast proteins depicted are indicated in magenta. Following a DSB, the MRX/MRN complex is loaded onto the DNA ends at the break site. If cells are in the G1 phase of the cell cycle, nonhomologous end joining is used to repair the break. If the cells are in S/G2 phases, phosphorylation of Sae2/CtIP by cyclin-dependent kinases (CDKs) favors homologous recombination-mediated repair. During the first step of the homologous recombination repair pathway, the initial resection of the DSB is promoted by MRX/MRN and Sae2/CtIP, resulting in 50–100 nucleotide ssDNA 3′ overhangs. In the second step, these fragments can serve as templates for long-range DNA end resection. This processive reaction can occur by two independent mechanisms: one that utilizes Sgs1/BLM and Dna2 (left) and the other using Exo1/hEXO1 (right). Following resection, the exposed ssDNA is coated by replication protein A (RPA), which recruits the Rad52 epistasis group of proteins (Rad52, Rad55, Rad57, Rad59, Rad54, Rdh54) to enable Rad51 filament formation. The DNA ends can be repaired using different mechanisms, such as single-strand annealing, break-induced replication, or gene conversion. DSB repair using the homologous chromosome is depicted here. After the search for homology, a joint DNA structure is formed. The resulting double Holliday junctions are then resolved and the lost bases are restored at the break site. The resolution of the junctions by the Sgs1-Top3-Rmi1/BLM-TOP3α-RMI1 (BLAP75) complex depicted here leads to a non-crossover product of DSB repair by homologous recombination.

In addition to the role of cell-cycle phase in selecting which repair pathway is used, the type of lesion also dictates different repair responses. For example, a single endonuclease-induced DSB that occurs in the G1 phase of the cell cycle is bound by the Ku70-Ku80 heterodimer and targeted to the nonhomologous end joining pathway for repair (Barlow et al., 2008; Zierhut and Diffley, 2008). In contrast, repair of DNA damage induced by ionizing radiation is preferentially targeted to the homologous recombination pathway regardless of the cell-cycle phase (Barlow et al., 2008). When cells are exposed to ionizing radiation during the G1 phase, they delay repair of the break until the S/G2 phases, presumably so that they can use the homologous template to restore missing bases.

How cell-cycle signals are integrated with different repair mechanisms has been puzzling. In addition to the type of DNA lesion having a role in dictating the cellular response, cyclin-dependent kinases (CDKs) also play an essential role (Ira et al., 2004). Previously, it was shown that the DNA endonuclease Sae2 functions with the MRX complex to process DSBs in a 5′ to 3′ direction (Clerici et al., 2005). Sae2 is important for recombination (Lengsfeld et al., 2007), is known to be phosphorylated, and contains CDK phosphorylation sites. Huertas and colleagues (2008) recently showed that one Sae2 phosphorylation site in particular, serine 267, is targeted by CDK activity to regulate Sae2 function. Interestingly, CtIP, the mammalian homolog of Sae2, also contains this CDK phosphorylation site (Huertas et al., 2008). Huertas et al. found that in yeast, expression of a sae2-S267A allele, which cannot be phosphorylated at the CDK site, renders cells as sensitive to DNA damage as cells lacking Sae2. Cells expressing sae2-S267A also accumulate less single-stranded DNA (ssDNA), indicating that resection is impaired at the induced DSB site (Huertas et al., 2008). Thus, CDK-dependent phosphorylation of Sae2 is necessary for efficient resection of the DNA ends of a DSB (Figure 1). Furthermore, cells either lacking Sae2 or expressing the sae2-S267A allele have increased rates of nonhomologous end joining, demonstrating that CDK-dependent phosphorylation of Sae2 is critical for promoting DNA repair at DSBs by homologous recombination (Huertas et al., 2008). Therefore, loss of Sae2 function compromises the delicate balance between using homologous recombination or nonhomologous end joining for the repair of DSBs.

Multiple Pathways for Initiating Resection

When a DSB occurs, the DNA ends can be resected to leave 3′-ended ssDNA tails (White and Haber, 1990). Although the ends are first processed by the MRX complex and the Sae2 protein (Figure 1), additional 5′ to 3′ DSB end resection occurs to trigger Rad51-mediated strand exchange, a key step in homologous recombination (Ira and Haber, 2002; Mimitou and Symington, 2008; Zhu et al., 2008). Therefore, additional proteins are required to expose long 3′ ssDNA ends and their identities remained a mystery. For example, in budding yeast, the Exo1 nuclease promotes DSB resection, but cells lacking Exo1 function still process DSBs and exhibit little sensitivity to DNA-damaging agents (Gravel et al., 2008; Mimitou and Symington, 2008). Similarly, yeast deficient in Sae2, the MRX complex, or Exo1 and the MRX complex still exhibit resection activity, suggesting the existence of additional resection pathways. Recently, a number of studies have clarified the steps in DSB end resection and identified some of the players able to promote the formation of 3′ ssDNA tails in both yeast and mammalian cells (Gravel et al., 2008; Huertas et al., 2008; Mimitou and Symington, 2008; Nimonkar et al., 2008; Zhu et al., 2008).

To monitor the formation of ssDNA, Mimitou and Symington (2008) and Zhu et al. (2008) use different assays to visualize resected DNA ends. Mimitou and Symington developed an assay in budding yeast where an inducible restriction enzyme cut site is inserted between two ade2 genes flanking a TRP1gene. The authors use this assay, in which a DSB is induced at the restriction site, to monitor yeast cells for the appearance of cells expressing Ade2 but not Trp1. They also examine the ssDNA adjacent to the cut site using DNA blots. Zhu et al. use HO endonuclease to induce a DSB at the HO-cleavage site in the MAT locus on chromosome III of budding yeast. This HO-inducible system enables resection to be monitored by DNA blots for up to 80 kb on each side of the DSB. Using these assays, both groups identify additional proteins that are needed for 5′ to 3′ end resection.

Both studies take cues from the bacterium Escherichia coli where the activity of RecBCD, a complex containing helicases and a nuclease, is sufficient to generate 3′ ssDNA tails. No corresponding helicase had been known to function in end resection in eukaryotes. Therefore, both Zhu et al. and Mimitou and Symington, using the assays described above, test yeast lacking several helicases important for DNA repair for defects in end resection. Interestingly, both groups report that Sgs1, a helicase that is a homolog of RecQ in E. coli, is important for end resection. Yeast cells lacking Sgs1 exhibit markedly slower ssDNA formation. Furthermore, the helicase function of the Sgs1 protein is necessary for end resection. Zhu et al. also show that disrupting the interacting partners of Sgs1—Rmi1 and Top3—also causes defects in end resection, suggesting that the entire Sgs1 complex is important in resection. The involvement of Sgs1 in this process is surprising because its mammalian homolog, BLM, has a well-established role with TOP3α in the resolution of double Holliday junctions in a later step in homologous recombination (Mankouri and Hickson, 2007). Although Sgs1 has previously been suggested to have functions during early steps in DSB repair (Cobb et al., 2003), the identity of these functions has been unclear.

Because Sgs1 lacks nuclease activity, it has been speculated that its helicase activity could unwind the double-stranded DNA to allow resection by a nuclease. To test this hypothesis, both groups compare resection phenotypes of cells lacking either Sgs1 or Exo1 to cells lacking both proteins. Surprisingly, cells lacking both Sgs1 and Exo1 process the 5′ strand more slowly and less efficiently than cells lacking only Sgs1 or Exo1 (Mimitou and Symington, 2008; Zhu et al., 2008). This finding suggests that Sgs1 and Exo1 function in different pathways that each contribute to the resection of DNA ends, indicating that Exo1 is not the nuclease that acts in concert with Sgs1. Zhu and colleagues identify the Sgs1-associated nuclease as Dna2, a surprising finding given that Dna2 is a helicase/nuclease known to function in Okazaki fragment processing during DNA replication (Bae et al., 2001). In addition to the single-strand processing defect, cells lacking both Sgs1 and Exo1 exhibit greater sensitivity to DNA-damaging agents in comparison to cells lacking only Sgs1 or Exo1. They also have increased gross chromosomal rearrangements and are unable to activate the DNA-damage checkpoint after DSB induction (Gravel et al., 2008). These defects may be due to insufficient amounts of exposed ssDNA at the break site to elicit a DNA-damage checkpoint response.

One puzzling observation in mutant yeast cells lacking both Sgs1 and Exo1 is the accumulation of DNA break ends that are 50–100 nucleotides shorter than the initial cut fragment (Mimitou and Symington, 2008; Zhu et al., 2008). Further analysis of the double mutant cells reveals that these DNA-repair intermediates contain regions of ssDNA formed by the MRX complex and Sae2 (Mimitou and Symington, 2008; Zhu et al., 2008). These results suggest a two-step model for end resection. In the first step, the DSB ends are trimmed by MRX and Sae2. In the second step, the ends are further processed by either Sgs1-Dna2 or Exo1 (Figure 1). This combination of resection activities exposes enough ssDNA to allow replication protein A (RPA) proteins to bind (Figure 1). RPA-coated ssDNA is a signal for activating the DNA-damage checkpoint response and is displaced during the break repair process by Rad52-recruited Rad51, which mediates strand exchange (Figure 1).

Although no diseases have yet been linked to mutations in human EXO1, Sgs1 has several RecQ homologs in humans (e.g., BLM, WRN, RTS) that when mutated result in increased tumorigenesis (Seki et al., 2008). Gravel and colleagues extended the yeast observations to human cells by monitoring cultured cells depleted for BLM or human EXO1. Interestingly, when compared to BLM or EXO1 single knockdowns, depletion of both BLM and EXO1 in cells exposed to the DNA-damaging agent camptothecin results in reduced phosphorylation of an RPA subunit, reduced activation of the Chk1 protein, and decreased cell survival, indicating reduced activation of the DNA-damage checkpoint (Gravel et al., 2008). Furthermore, the authors find that cells deficient in both BLM and EXO1 have fewer DSB sites that colocalize with RPA in comparison to cells depleted for only BLM or EXO1 (Gravel et al., 2008). These findings show that there is less ssDNA at DSBs in cells lacking BLM and hEXO1, suggesting that these human cells have end resection defects similar to those found in yeast.

Conservation of the End Resection Machinery in Archaea

Although the proteins needed for DSB processing are well conserved in eukaryotes, the enzymes used for these same end resection steps in archaea were unknown. Hopkins and Paull (2008) recently took a biochemical approach to address this question. They purified proteins from the archaeon Pyrococcus furiosus and identified two new proteins involved in 5′ end resection and strand exchange. In most thermophilic archaea, the Mre11 and Rad50 genes cluster in an operon with HerA and NurA genes (Constantinesco et al., 2004), similar to the clustering of genes encoding the RecBCD complex in eubacteria. HerA is an ATP-dependent DNA helicase and NurA is a 5′ to 3′ exonuclease (Constantinesco et al., 2004). Hopkins and Paull find that although purified archaeal Mre11 and Rad50 are not sufficient to degrade linear DNA substrates, they can act together with purified HerA and NurA to processively degrade double-stranded DNA in a 5′ to 3′ direction. These four archaeal proteins, together with the RadA recombinase, also catalyze ATP-dependent joint molecule formation from supercoiled DNA (Hopkins and Paull, 2008). Therefore, the cooperative activities of HerA, NurA, Mre11, and Rad50 efficiently process DSB ends, enabling RadA-initiated strand exchange. It is puzzling that despite the observed functional interactions between Mre11-Rad50 and HerA or NurA on DNA, Hopkins and Paull did not find any direct protein-protein interactions between Mre11-Rad50 and HerA or NurA (Hopkins and Paull, 2008). It remains unknown how Mre11 and Rad50 are stimulated by HerA and NurA, although it is possible that Mre11-Rad50 action creates an optimal substrate for HerA-NurA. Regardless, the helicase activity of HerA and the nuclease activity of NurA stimulate DNA end resection, which is reminiscent of the Sgs1 and Exo1 activities that are observed in eukaryotic cells. Hence, a common underlying mechanism for DNA end resection may be evolutionarily conserved from archaea to eukaryotes.

Human Exo1, BLM Helicase, and End Resection

Nimonkar and colleagues (2008) report progress in biochemically characterizing the human Exo1 nuclease and BLM helicase. They show that purified human Exo1 (hExo1B) and BLM proteins collaborate to resect double-stranded DNA templates in a 5′ to 3′ direction in the absence of ATP (Nimonkar et al., 2008). The resection activity is specific to BLM, as the other RecQ homologs (WRN, hRecQ1, hRecQ5β, hRecQ4) do not stimulate the resection reaction (Nimonkar et al., 2008). The hExo1-BLM resection activity is sufficient to promote strand exchange given that the ssDNA substrates created by hExo1-BLM stimulate hRad51-mediated joint molecule formation in vitro. However, one of the main differences between these findings in vitro and the observations made by Gravel et al. in mammalian cells is that in the in vitro experiments, BLM and hEXO1 physically interact and furthermore, BLM stimulates hEXO1 nuclease activity. In contrast, in vivo genetic analysis suggests that hExo1 and BLM function in different pathways, as the loss of both proteins results in more severe defects than the loss of either protein alone (Gravel et al., 2008). Perhaps, the activities of hEXO1 and BLM are regulated by cell-cycle phase or the type of DNA lesion encountered.

Future Directions

The studies highlighted here elucidate how induced DSBs are processed into ssDNA overhangs that are then used for strand invasion and break repair. However, the primary lesions that these proteins act on in the cell remain unknown. Myriad different types of cellular damage can cause DSBs, including uncapped telomeres, protein/DNA crosslinks, and excision of damaged base pairs. Resection activities are also likely important during the processing of collapsed replication, a process during which DSBs can be formed. A tantalizing possibility is that the Sgs1-Top3-Rmi1 complex specifically acts during S phase to promote resection of DSBs that occur opposite to incoming replication forks. In fact, the topological context that arises under these circumstances resembles converging replication forks and could create substrates for RecQ helicases and Top3. Furthermore, cell-cycle phase clearly has an important role in DNA-damage processing.

One of the major obstacles in studying DSB repair is that the disruption of central players in the process can lead to a pleiotropic phenotype that masks specific protein functions in different pathways. For example, the Sgs1 complex is important for end resection at the beginning of homologous recombination but also has functions later in the process during the resolution of DSB intermediates (Figure 1). Here, identification of mutations that specifically disrupt only one function will aid in the understanding of how these proteins act at different steps in repair. Indeed, a mutant allele of Sgs1 that separates the function of Sgs1 in the repair of replicative damage from its function in homologous recombination has been identified (Bernstein et al., 2009). Posttranslational modification (such as SUMOylation, ubiquitination, or phosphorylation) of these central repair proteins is likely to play an important role in regulating specific repair mechanisms, and their precise roles in these processes will be the focus of future studies.