|Home | About | Journals | Submit | Contact Us | Français|
Eukaryotic genomes harbor a large number of homologous repeat sequences that are capable of recombining at high frequency. Their potential to disrupt genome stability highlights the need to understand how homologous recombination processes are coordinated. The S. cerevisiae Rad1-Rad10 endonuclease performs an essential role in recombination between repeated sequences by processing 3′ single-stranded intermediates formed during single-strand annealing and gene conversion events. Several recent studies have focused on factors involved in Rad1-Rad10-dependent removal of 3′ nonhomologous tails during homologous recombination, including Msh2-Msh3, Slx4, and the newly identified Saw1 protein (1–4). In addition, these studies suggest mechanisms for how DNA repair is coordinated by the DNA damage checkpoint machinery (1). This review aims to integrate these new findings with previous work to create a comprehensive model for how DNA repair and checkpoint factors act in concert to process 3′ nonhomologous tail intermediates that arise during homologous recombination.
Homologous stretches of DNA sequences scattered throughout the genome are a threat to genome stability because of their potential to recombine. It is estimated that nearly half of the human genome consists of repetitive DNA (5, 6), and genome rearrangements caused by recombination between repeats are known to contribute to a variety of human diseases, including many cancers (7–9). Repetitive sequences such as Alu elements are particularly susceptible to non-conservative homologous recombination via single-strand annealing (SSA), which results in the deletion of sequences located between the repeats (10–12). The abundance of such sequences in the human genome underscores the need for a comprehensive understanding of the homologous recombination processes that act on them.
SSA is a major recombination pathway for repairing spontaneous and induced double-strand breaks (DSBs) that arise between repeated sequences (10, 13, 14). During SSA in S. cerevisiae, DSBs are resected 5′ to 3′ to reveal single-stranded homologous sequences (Figure 1). Following Rad52- and Rad59-dependent annealing of the homologous sequences, the 3′ single-stranded DNA ends are nonhomologous to the new flanking regions, and must be cleaved in order to complete repair of the broken strands. The Rad1-Rad10 endonuclease and Msh2-Msh3 mismatch recognition complex are required for cleaving 3′ single-stranded nonhomologous tails on either side of the annealed region (15–19). Once both 3′ tails have been removed, SSA is completed by DNA synthesis initiated off of the newly cleaved 3′ ends followed by ligation (reviewed in 10).
Rad1-Rad10 is a structure-specific endonuclease that functions in both nucleotide excision repair (NER) and homologous recombination (10, 11, 19, 20). The significance of the role of Rad1-Rad10 in recombination is demonstrated by the severe developmental abnormalities in mice lacking the mammalian RAD10 homolog ERCC1 which, unlike NER mutant phenotypes, include severe runting, reduced liver function, and death before weaning (21, 22). A recently described human patient with ERCC1 deficiency also exhibited severe fetal development defects that are clearly distinct from NER-related phenotypes (23). In yeast, the absence of Rad1-Rad10 leads to cell death or plasmid loss (depending on the assay) during recombination by SSA due to lack of repair, since 3′ nonhomologous tail removal is an essential step in SSA (15, 17). Several recent papers have highlighted factors involved in Rad1-Rad10-dependent 3′ nonhomologous tail removal during homologous recombination in S. cerevisiae (1–4), and it is these non-NER functions of Rad1-Rad10 that are reviewed here. The role of Rad1-Rad10 in nucleotide excision repair has been reviewed elsewhere (19, 20).
Homologous recombination by gene conversion also involves the removal of 3′ nonhomologous tails. Most mitotic gene conversion events are thought to occur by a synthesis-dependent strand annealing mechanism (10, 11, 24). During such gene conversion events, the DSB is resected 5′ to 3′, and one of the 3′ ends undergoes Rad51-mediated strand invasion into a duplex region of DNA containing a homologous sequence (Figure 2A). DNA synthesis initiating from the 3′ invading strand allows for copying of DNA sequence from the donor template, and unwinding of the invading strand from the donor template allows it to anneal back to its native locus. The non-invading strand is then able to be repaired using the invading strand as a template (reviewed in 10, 11).
Rad1-Rad10-dependent nonhomologous tail removal during gene conversion can occur during the strand invasion step as well as after annealing, depending on whether one or both 3′ ends contain nonhomology with respect to the donor locus. If both sides of a DSB are nonhomologous to the donor (Figure 2B), the invading strand contains a 3′ nonhomologous tail that must be removed in order to prime repair synthesis off of the donor. When nonhomologous sequence resides on only one side of a DSB (Figure 2A), the 3′ end of the break that shares homology with the donor sequence performs the strand invasion step, and there is thus no barrier to initiate new DNA synthesis on the invading strand.
The requirement for Rad1-Rad10 during gene conversion depends on both the number and length of nonhomologous tails. Rad1-Rad10 is critical for gene conversion when both ends of a DSB contain 30 or more nucleotides of nonhomologous sequence (3, 17, 25, 26), but DSB repair is more subtly reduced in rad1Δ mutants when only one nonhomologous tail is present (3, 25, 27). The differential requirement for Rad1-Rad10 during gene conversion when one or two ends of a DSB contain nonhomology has been ascribed to the structural nature of the DNA junctions. The initial invasion of 3′ single-stranded DNA into a homologous duplex is proposed to create an unstable paranemic joint, which might be a better substrate for Rad1-Rad10 (17). In contrast, when nonhomologous sequence is only on one side of a DSB, the homologous 3′ end can perform strand invasion, leaving the nonhomologous tail on the second, non-invading end (Figure 2A). The sequence adjacent to this nonhomologous 3′ end would likely form a stable plectonemic joint, since the rest of the strand can fully base pair. It is possible that Rad1-Rad10 is only minimally required when only one nonhomologous tail is present because plectonemic joints are not ideal substrates for Rad1-Rad10, and/or Rad1-Rad10 is only one of a host of other factors that process these types of structures.
When 3′ nonhomologous tails are only 10 nucleotides in length, gene conversion remains efficient in the absence of RAD1, MSH2, or MSH3, and the short 3′ tails are removed by the proofreading 3′ to 5′ exonuclease activity of Polymerase δ (26). A second Rad1-Rad10- and Msh2-Msh3-independent pathway of 3′ nonhomologous tail removal is proposed to exist (3, 16, 26, 27), though the inefficiency of this proposed pathway suggests that Rad1-Rad10-dependent end processing is preferred. There is no evidence that known factors play a role in this backup pathway of nonhomologous tail removal, as neither Mus81-Mms4 nor the proofreading activities of Pol δ and Pol ε appear to contribute (3, 26). Replication of partially repaired recombination intermediates might also bypass the requirement for 3′ nonhomologous tail removal (3, 28).
The Msh2-Msh3 DNA mismatch recognition complex functions in Rad1-Rad10-dependent 3′ end processing during homologous recombination (Figures 1, ,2B;2B; 17, 25, 26, 29, 30). Msh2-Msh3 specifically recognizes insertion/deletion loops of up to 17 base pairs in DNA mismatch repair (31–33). During homologous recombination, Msh2-Msh3 is proposed to act in the recognition and stabilization of 3′ tails at the junction of double-stranded and single-stranded DNA to aid in either the recruitment or cleavage activity of Rad1-Rad10 (17, 34–36). A similar mechanism involving Msh2-Msh3 and Rad1-Rad10 also functions to remove large loops during meiosis (32, 37, 38). Consistent with its role in 3′ tail removal, an in vitro DNA binding study showed that purified Msh2-Msh3 binds specifically to branched DNA substrates containing 3′ single-stranded ends, with an affinity comparable to that of its binding to +8 mismatch loops (4). Msh2-Msh3 appears to bind asymmetrically around double-strand/single-strand junctions, and binding opens the conformation of the junction slightly, potentially to facilitate Rad1-Rad10-dependent cleavage of 3′ tails (4).
Msh2 physically interacts with both Rad1 and Rad10 independently of other mismatch repair factors (35), and no other mismatch repair factors are required for Rad1-Rad10-dependent 3′ end processing besides Msh3 (17, 29, 36). The role of Msh2-Msh3 in nonhomologous tail removal during recombination can be distinguished from its role in DNA mismatch repair, since mutations have been isolated in MSH2 that disrupt mismatch repair but are functional for recombination (34). Msh2 localizes rapidly to DSBs flanked by nonhomologous sequence on chromosomal and plasmid substrates (3, 30, 39), and Msh2 and Msh3 have been reported to physically interact with subunits of the single-strand binding protein RPA (40–42). Together these data support a very early role for the Msh2-Msh3 complex in 3′ nonhomologous tail removal that might aid in 3′ tail recognition.
Rad1-Rad10-dependent 3′ end processing does not always require Msh2-Msh3. During SSA, the requirement for Msh2-Msh3 depends upon the length of the annealed region. Strains lacking MSH2 or MSH3 are defective in SSA when the annealed region is only 205 bp, but show only a small reduction in repair relative to wild-type when the annealed region is more than 1 kb (17). Decreased dependence on Msh2-Msh3 is also observed with larger loop sizes during Rad1-Rad10-dependent meiotic loop repair (38). Despite a predicted role and clear localization of Msh2 to DSBs, Msh2-Msh3 is dispensable for gene conversion during mating type switching, where Rad1-Rad10 plays a significant role (3). Only when nonhomologous sequence is inserted on the invading strand does the role of Msh2-Msh3 in Rad1-Rad10-dependent 3′ nonhomologous tail removal become apparent (3). Altogether, these results support the idea that Msh2-Msh3 plays a role in stabilizing 3′ tail intermediates in preparation for Rad1-Rad10-dependent cleavage.
Recent work by Flott et al. (1) identified Slx4 as a essential component of the Rad1-Rad10 3′ nonhomologous tail removal pathway. Slx4 was initially characterized as a subunit of the Slx1-Slx4 endonuclease, deletion of which is synthetically lethal with sgs1Δ (43). Slx1-Slx4 is a 5′ flap endonuclease, of which Slx1 is thought to be the catalytic subunit (44). In addition to its function as a heterodimer with Slx1, Slx4 appears to have at least two other separate functions, one involving Rad1-Rad10, and another that is independent of both Slx1 and Rad1-Rad10 and promotes cellular resistance to MMS (1, 2, 45, 46). A screen for mutants defective in SSA recently found that slx4Δ mutants are blocked at the 3′ tail removal step of SSA (2), and Slx4 was found to play an important role during mating type switching in the same pathway as the Rad1-Rad10 complex (3). Additionally, the Flott et al. study found that at least three residues on Slx4 are directly phosphorylated by the Mec1 and Tel1 checkpoint kinases, and that this phosphorylation is essential for the SSA functions of Slx4 but not for resistance to MMS or viability in sgs1Δ mutants (1, 47).
These new findings support a model in which Slx4 is acted upon directly by the checkpoint machinery in response to DSBs to recruit or activate Rad1-Rad10 and promote 3′ nonhomologous tail removal. The DNA damage response is thought to be activated during SSA because of the relatively slow kinetics of repair and the extensive resection required. Since phosphorylation of Slx4 is absolutely essential for 3′ nonhomologous tail removal during SSA, but is dispensable for its other functions (1), this phosphorylation is likely to provide the specificity to channel Slx4 to its recombination function.
While gene conversion involving a single 3′ nonhomologous tail does not require checkpoint activation, recent work has shown that rad1Δ and slx4Δ mutants exhibit RAD9- and MAD2-dependent checkpoint activation during mating type switching, a single-nonhomology gene conversion event (3). It is possible that the recruitment or activity of Rad1-Rad10-Slx4 is required to turn off the DNA damage checkpoint by signaling that repair is proceeding normally.
Slx4 is not required for checkpoint activation, but provides a crucial link between DNA damage sensing and activation of DNA repair. Slx4, but not its endonuclease partner Slx1, is required for the repair of alkylation damage (47). In addition to SSA, Slx4 is phosphorylated in a Mec1/Tel1-dependent manner in response to a variety of DNA-damaging agents, including MMS, camptothecin, hydroxyurea, ionizing radiation, and the UV mimetic 4-NQO, and is required for efficient DNA repair throughout the cell cycle (47). Thus it appears that Slx4 plays a critical role in the response to many types of DNA damage, and the strict requirement for its checkpoint-dependent phosphorylation in 3′ nonhomologous tail removal provides a beautiful example of how DNA damage sensed by the checkpoint machinery directly promotes DNA repair.
The SAW1 (YAL027W) gene was recently identified in a microarray-based screen for mutants defective in SSA (2). Saw1, for single-strand annealing weakened 1, physically interacts with Rad1-Rad10, Msh2-Msh3, and Rad52 (2, 40, 42), all of which function in SSA, and like slx4Δ mutants, saw1Δ mutants are defective specifically in 3′ nonhomologous tail removal (2). Li et al. (2) found that the Rad1 protein fails to localize to SSA intermediates in saw1Δ mutants, and saw1Δ 18–24 mutants, whose mutant Saw1 protein fails interact with Rad1 but still interacts with Msh2 and Rad52, are completely defective in SSA. These new findings provide strong evidence that Saw1 recruits Rad1-Rad10 to recombination intermediates containing Rad52.
Rad14 is thought to target Rad1-Rad10 to its substrates during NER (20, 48, 49), but it has been unclear how Rad1-Rad10 is targeted to 3′ tails during recombination, especially since Msh2-Msh3 is dispensable for some gene conversion and SSA events. In human cell-free extracts, Rad52 and the Rad1-Rad10 homolog XPF-ERCC1 stably associate through interaction of XPF with the N-terminus of Rad52 (50). This physical interaction stimulates the endonuclease activity of XPF-ERCC1 and attenuates the strand annealing activity of Rad52, both of which promote the removal of 3′ tails from recombination intermediates (50). Though an equivalent interaction between Rad1-Rad10 and Rad52 has not been observed, it appears that Saw1 provides this physical interaction in S. cerevisiae (2). These new findings reveal how Saw1, and potentially Slx4, may recruit and regulate Rad1-Rad10 and 3′ nonhomologous tail removal during homologous recombination.
The new findings reviewed here provide a more comprehensive understanding of how Rad1-Rad10-dependent 3′ end processing is coordinated during homologous recombination. As shown in Figure 3, resection of DSBs during SSA creates 3′ single-stranded tails that are coated by the RPA single-stranded DNA binding protein. The extensive single-stranded DNA activates the DNA damage response, which promotes phosphorylation of Slx4 by the Mec1 and Tel1 checkpoint kinases. Once resection has revealed regions of homology, the Rad52 strand annealing protein anneals the homologous sequences, creating 3′ nonhomologous tails on either side of the intermediate. When the length of homology is limited and creates an unstable paranemic intermediate, Msh2-Msh3 stabilizes the junction and slightly opens it to create a more suitable substrate for Rad1-Rad10 cleavage. Saw1 recruits Rad1-Rad10 to Rad52-containing annealed sequences at double-strand/single-strand junctions bound by Msh2-Msh3. Phosphorylated Slx4 binds to Rad1-Rad10, but does not appear to interact with Rad52, Saw1, or Msh2-Msh3. The result of these physical interactions is the positioning of Rad1-Rad10 at the double-strand/single-strand junction between Saw1 and Slx4. Rad1-Rad10 cleaves the strand near the junction, removing the 3′ nonhomologous tail and providing a free 3′ OH for extension by DNA polymerases.
It is not clear whether these physically interacting factors arrive at recombination intermediates in a sequential fashion, or whether a stable complex of Saw1-Msh2-Msh3-Rad1-Rad10-Slx4 exists in vivo and is simply recruited to Rad52-containing DNA, perhaps in response to Slx4 phosphorylation by Mec1/Tel1. It is striking that saw1Δ and slx4Δ mutants exhibit opposing phenotypes with regard to rDNA stability (2), so additional work is needed to understand the distinct functions of Slx4 and Saw1 in recombination and how they relate to coordination of Rad1-Rad10 3′ tail cleavage. It will be interesting to identify the critical function of Slx4 in this process, whether it is primarily to transmit a repair signal from the DNA damage checkpoint or whether there are additional roles in recruiting, positioning, or activating the endonuclease activity of Rad1-Rad10. While the findings reviewed here provide a clearer picture of how Rad1-Rad10 identifies and cleaves its substrates during recombination, many questions still remain. What is the physical activity of Saw1? How does Slx4 arrive at junctions containing 3′ tails? What is the function of phosphorylated Slx4? Why is Msh2-Msh3 essential in some cases and dispensable for others? Addressing these questions will edge us closer to understanding the link between DNA damage recognition, checkpoint signaling, and efficient DNA repair.
We thank K. T. Nishant, Aaron Plys, Justin Sibert, and Sarah Zanders for helpful comments on the manuscript. A.M.L and E.A. are supported by National Institutes of Health grant GM53085.