Quality control of homologous recombination

Abstract

Exogenous and endogenous genotoxic agents, such as ionizing radiation and numerous chemical agents, cause DNA double-strand breaks (DSBs), which are highly toxic and lead to genomic instability or tumorigenesis if not repaired accurately and efficiently. Cells have over evolutionary time developed certain repair mechanisms in response to DSBs to maintain genomic integrity. Major DSB repair mechanisms include non-homologous end joining and homologous recombination (HR). Using sister homologues as templates, HR is a high-fidelity repair pathway that can rejoin DSBs without introducing mutations. However, HR execution without appropriate guarding may lead to more severe gross genome rearrangements. Here we review current knowledge regarding the factors and mechanisms required for accomplishment of accurate HR.

Introduction

During the life of any single cell, the genome is continually challenged by a plethora of endogenous and exogenous factors, including free radicals generated from cellular metabolism, ultraviolet light from the sun, and ionizing radiation (IR) [1, 2]. Arising from endogenous (e.g., replication-associated errors and T- and B cell development) and exogenous (e.g., IR and chemotherapeutic agents) sources, double-strand breaks (DSB) are one of the most dangerous of all types of DNA damage because it results in physical cleavage of the DNA backbone [3–7].

Unrepaired or misrepaired DNA damage can result in cell senescence, apoptosis or tumorigenesis. Consequently, cells have evolved numerous highly efficient DNA repair pathways to sense and repair the various types of DNA damage to maintain genomic integrity and stability [8–10]. Two major pathways responsible for DSB repair in eukaryotic cells are non-homologous end joining (NHEJ) pathway and homologous recombination (HR) pathway [11, 12]. Defects in either pathway cause genome instability and promote tumorigenesis [12, 13]. Although it is generally believed that NHEJ is an error-prone pathway and HR is an error-free pathway, mounting evidences implicated that this view is too simplistic [14–18]. Thus, selection of the appropriate pathway is essential and is highly regulated in mammalian cells to maintain genomic stability [19–22]. It is widely accepted that DSBs that occur in the S phase due to collapsed replication forks and in the G2 phase are repaired mainly by HR, while NHEJ functions throughout the cell cycle [23–26]. However, the precise mechanism by which cells select a specific pathway to fix DSBs is not yet known, although some proteins that function in the early steps of both pathways may play key roles in this decision [27, 28].

Briefly, HR uses a sequence similar or identical to the broken DNA as a template, which makes HR the most accurate repair pathway for DSBs. To ensure accuracy, multiple factors are involved [29, 30]. We will discuss those pivotal proteins and steps involved in HR to clarify how cells execute this tightly controlled DNA repair mechanism.

NHEJ or HR? Make a decision

Non-homologous end joining is a relatively simple and straightforward, DSB repair pathway that ligates two DNA ends without requiring an accessible homologous DNA segment [12, 31, 32]. Canonical NHEJ (C-NHEJ) is initiated by the recognition and binding of the broken DSB ends by the Ku70/Ku80 protein complex, followed by the recruitment of the DNA-dependent protein kinase catalytic subunit (DNA-PKcs) [33–35]. Once bound to the broken ends, DNA-PKcs is activated and then phosphorylates itself and other targets, such as Artemis, which exerts nuclease activity to trim the DNA ends [36–39]. Finally, the DNA ligase complex, including DNA ligase IV, X-ray-cross-complementation group 4 (XRCC4), and XRCC4 like factor (XLF)/Cernunnos, is recruited to seal the DNA break [40–49]. A special DNA ligase that is conserved in all eukaryotes but is not essential for DNA replication, DNA ligase IV, is the core C-NHEJ enzyme that catalyzes the final step in joining the ends of the two DNA strands [50]. In addition to the extensively studied C-NEHJ pathway, another alternative NHEJ (Alt-NHEJ) pathway has been uncovered recently [14, 51]. Studies have shown that Alt-NHEJ can ligate DNA ends in the absence of C-NHEJ factors [14, 51]. Interestingly, although the NHEJ pathway is considered to be an inherently mutagenic process, due to the lack of homologous templates, studies in different species suggested that the C-NHEJ is not an intrinsically error-prone pathway and the accuracy of the repair is dictated by the structure of DNA substrates, but not by the C-NHEJ machinery [14]. In contrast to C-NHEJ, Alt-NHEJ-mediated end-joining is an error-prone event and often results in increased risks of chromosome rearrangements [52, 53].

Compared with NHEJ, HR requires the generation of single-stranded DNA (ssDNA) intermediates, which are used for homology searching and pairing (Fig. 1) [11, 13]. The sister chromatid is typically used as the template in the S and G2 phases of the cell cycle after replication of the DNA [11]. Thus, HR is generally considered to be an error-free pathway in mammalian cells. However, HR is believed to be restricted to particular phases in higher eukaryotes to maintain accuracy, as spurious HR between paternal and maternal chromosomes leads to loss of heterozygosity [15–17, 25]. In general, HR is initiated by 5′-end resection, which generates an extended 3′ single-strand overhang (ssDNA) that then is recognized and subsequently coated by replication protein A (RPA) to remove the secondary structure and protect the ssDNA tail [11, 54]. After RPA binding, RAD51 displaces RPA to form a nucleoprotein filament (presynaptic filament) with the 3′ overhangs at a ratio of one monomer of RAD51 per three nucleotides [11, 54]. The filament then searches for and aligns with its homologous DNA sequence on the sister chromatid to form an intermediate, called a D-loop [11, 54]. After the strand invasion and alignment catalyzed by RAD51 and several other proteins, the D-loop expands and then captures the second 3′ ssDNA overhang generated by resection of the opposite end of the DSB, resulting in the formation of a double Holliday junction (dHJ) [11, 54]. The dHJ can be dissolved or cleaved to yield double-stranded DNA (dsDNA) products. In the process of cleavage, both non-crossover products and crossover products are generated [11, 54]. However, evidence demonstrates that DSB repair is rarely associated with crossovers in somatic cells, which suggests that the high-fidelity dissolution pathway is the major choice in HR repair [11, 54]. An alternative HR model, known as synthesis-dependent strand annealing (SDSA), does not involve Holliday junctions and results in only non-crossover products (Fig. 1) [55].

Fig. 1.

Models of double-strand break repair. Upon induction of DSBs, the broken DNA ends are either coated by Ku complex which initiates the canonical non-homologous end-joining (C-NHEJ) pathway or MRN/X complex which processes the DNA ends into short 3′-single-stranded DNA (ssDNA) tails. The first step of resection by MRN/X complex could trigger either the alternative end-joining (Alt-NHEJ) pathway or homologous recombination (HR) pathway. In HR pathway, the short ssDNA tails that are initially coated by the replication protein A (RPA) complex can be further resected into longer ssDNA tails. In a subsequent step, recombination mediator proteins such as BRCA2 and RAD51 paralogs catalyze the replacement of RPA with RAD51, resulting in the formation of ssDNA-RAD51 nucleoprotein filament. The ssDNA-RAD51 nucleoprotein filament then catalyzes strand invasion into homologous duplex DNA, leading to the formation of D-loop. After DNA synthesis primed by the invading strand, the repair can bifurcate into two alternative sub-pathways referred to as synthesis-dependent strand annealing (SDSA) and double-strand break repair (DSBR). In SDSA, the extended D-loop can be dissolved by specialized DNA helicases and the newly synthesized strand is annealed to the ssDNA tail on the other break end, which is followed by gap-filling DNA synthesis and ligation. The repair products from SDSA are always non-crossover. In DSBR, the second DSB end is captured to form an intermediate with two Holliday junctions, called double Holliday junction (dHJ). dHJ a central intermediate of HR that can be processed to yield crossover or non-crossover recombination products

Although there are differences in fidelity and template requirements, both DSB repair pathways play important roles in maintaining genomic stability in mammalian cells [19]. However, the exact mechanism by which the choice between NHEJ and HR is made remains unclear, although only the appropriate selection results in optimal repair [19, 56]. Non-homologous end joining has long been believed to be the predominant pathway for DSB repair, as it is not restricted in certain cell cycle phases. Observations in several species suggest that HR is active in mainly the late S and G2 phases [57]. This is consistent with the requirement for sister chromatids in HR, which exist only in the S/G2 phase, after DNA synthesis. Moreover, several HR proteins are regulated in a cell cycle-dependent manner. For example, the abundance of the C-terminal binding protein interacting protein (CtIP) is cell cycle regulated, with low CtIP protein in G1 phase and the highest amount observed during S and G2 phases [58]. In addition, CtIP/Sae2 activation is also regulated by the cyclin-dependent protein kinases (CDKs) [59–61]. Besides the cell cycle, the repair mechanism used (HR or NHEJ) also depends on substrate complexity [56]. A frank DSB could be repaired by either NHEJ or HR, while a collapsed replication fork is typically repaired by the HR machinery [56]. Mechanically, the choice of repair pathway may be controlled by the proteins involved in the early stages of both pathways, followed by recruitment of HR- or NHEJ-specific proteins to complete the repair process [56]. A series of studies published recently suggested that tumor suppressor p53-binding protein 1 (53BP1) and tumor suppressor breast cancer 1 (BRCA1) determine whether the HR or NHEJ pathway is used through DNA-end resection control; we will discuss the details in the subsequent section [25, 62].

DNA end resection: the initiation of homologous recombination

When DSBs occur in the S or G2 phase, a process called DNA end resection can be activated, which processes the DSB ends to generate 3′ overhang ssDNA tails [63, 64]. The generation of 3′ ssDNA by end resection not only provides a platform to recruit the proteins that participate in DSB repair but also inhibits NHEJ and triggers HR-mediated DSB repair [65]. The MRN (MRX in yeast) complex (comprised of MRE11, RAD50 and NBS1/Xrs2), CtIP (Sae2 in yeast), Exonuclease 1 (EXO1), nuclease/helicase DNA2, and multiple chromatin remodeling factors are involved in this process [66]. However, if cells are in the G1 phase and lack a homologous duplex DNA template, the presence of Ku70/80 and other proteins prevents resection and, together with the MRN complex, mediates NHEJ pathway initiation.

MRN/X complex

The MRE11, RAD50, and NBS1 (MRN) complex in mammals and fission yeast, and Mre11, Rad50, and Xrs2 (MRX) complex in budding yeast, are the first group of proteins to respond to DSBs [67–70]. When DSBs occur, the MRN complex, comprised of NBS1₂RAD50₂MRE11, is recruited to the DNA damage site and carries out at least three functions to initiate DNA repair [71, 72]. First, the MRN complex is capable of sensing the broken DNA end. Second, the MRN complex recruits and activates ataxia-telangiectasia mutated (ATM) and other pivotal proteins (i.e., tat-interactive protein 60 kDa [TIP60], acetyltransferase, and p53-binding protein 1 [53BP1]) to trigger the DNA damage response (DDR) [73–79]. The DDR includes several programs in different cellular contexts, including cell-cycle arrest, apoptosis, senescence, and DNA repair [75, 80, 81]. Third, the MRN complex is critical for DNA resection to generate free ssDNA tails, which are required for HR [68].

MRE11 is the core of the MRN complex; both the ATM activation and end resection functions of the MRN complex require MRE11 activity [75, 82]. MRE11 is a highly conserved protein comprised of an N-terminal phosphoesterase domain and two distinct C-terminal DNA-binding domains. In vitro experiments revealed that MRE11 exhibits a number of enzyme activities, including Mn²⁺-dependent dsDNA 3′-5′ exonuclease, ssDNA endonuclease, DNA annealing and unwinding [83, 84]. In DNA resection, MRE11 couples with RAD50 to bind to DNA and process the broken DNA end [71]. Structural studies revealed that MRE11 is dimerized through its conserved N-terminal domain [85, 86]. However, although DNA binding and protein dimerization can stabilize each other in this case, dimerization is not required for its nuclease activity, which suggests that the dimerization may not be essential for the end resection directly, but may contribute to other functions, such as assembly of the MRN complex [86]. The early embryonic lethality observed in MRE11-depleted mice implicates that it functions in an essential process [87]. In addition, cells containing nuclease-defective MRE11 suffer growth defects and chromosomal abnormalities and are sensitive to DNA-damaging agents [88]. Furthermore, in vivo experiments in mice, tissue culture, and structural analysis suggest that the nuclease activity of MRE11 is not required for ATM activation but is essential for proper end resection [89]. Interestingly, ATM activation depends on the presence of ssDNA oligonucleotides, although evidence suggests that these ssDNAs do not have to be produced by MRE11 itself. Paradoxically, DNA repair by HR requires exposure of a 3′ overhang ssDNA created by a 5′–3′ exonuclease, while the observed exonuclease activity of MRE11 is opposite and unlikely to perform this role [89, 90]. Notably, a two-step mechanism for MRE11′s role in DSB resection has been proposed recently [91]. In the first step, MRE11 makes the initial single-strand nick through its endonuclease activity to direct repair toward HR [91]. In a second step, MRE11 digests 3′–5′ toward the DNA end through its 3′–5′ exonuclease activity to generate 3′ overhangs [91].

NBS1 is considered to be a regulator in the MRN complex; the complete disruption of NBS1 in mice is lethal, while heterozygotes develop a wide variety of tumors that affect several organs [92–95]. Human NBS1 mutations are responsible for Nijmegen breakage syndrome (NBS), a rare autosomal recessive hereditary disorder that imparts an increased predisposition to the development of malignancies [96]. Cells derived from NBS patients share similar phenomena as ATM mutant cells, including hypersensitivity to IR, chromosomal fragility, and abnormal cell cycle checkpoint regulation [97, 98]. Although the function of NBS1 in HR and end resection is not fully understood, much information suggests that NBS1 plays a critical regulatory role in MRN complex function, even though it lacks DNA-binding and enzymatic activities [85]. NBS1 contains dual phosphopeptide-binding motifs, a Forkhead-associated domain (FHA domain), and a tandem BRCA C-terminal domain (BRCT domain). NBS1 binds to γ-H2AX after DSBs and then recruits MRE11 and RAD50 to the broken DNA [99, 100]. As a result, NBS1 depletion abolishes MRE11 and RAD50 nuclear translocation to DNA damage sites.

RAD50 is a member of the structural maintenance of chromosomes (SMC) protein family [94, 101]. Like other SMC members, RAD50 contains a long coiled-coil domain, and dimerization of two RAD50s through the coiled-coil domain is essential for MRN complex formation [102, 103]. Separated by the coiled-coil domain, two ATPase motifs are present in both RAD50 termini. Both the N- and C-terminal ATPase motifs have both ATPase and adenylated kinase activities and are required for HR. Briefly, by binding to the DNA duplex through its ATPase motifs and holding the broken ends together using its coiled-coil arms, RAD50 is critical for the DNA-binding and DSB-end-tethering function of the MRN complex [104–106].

CtIP/Sae2

C-terminal binding protein interacting protein was characterized initially as a transcriptional cofactor; it was subsequently shown to be a protein with multiple functions involved in various cellular processes, including cell cycle regulation, transcription regulation, and DSB end resection [107–110]. C-terminal binding protein interacting protein is the human homolog of yeast Sae2, another nuclease whose roles are redundant with those of Mre11 in yeast [111]. However, the observation that CtIP lacks nuclease activity makes MRE11 the only nuclease identified to date that initiates end resection in mammalian cells [111, 112]. Exclusively in S/G2 phase, CtIP localizes to DNA damage sites after DSBs occur and physically interacts with MRN [112–114]. Interestingly, CtIP recruitment to DSBs is delayed compared with the MRN complex [110]. Based on observations that CtIP is phosphorylated before recruitment by ATM, and the fact that ATM activation depends on the MRN complex, the MRN complex may indirectly facilitate CtIP recruitment. Furthermore, the MRN complex counterpart in yeast, the MRX complex, interacts with the chromatin remodeling complexes RSC and INO80 to overcome DNA end barriers for CtIP binding [115, 116]. Although CtIP recruitment requires the MRN complex, emerging evidence suggests that the function of MRN in end resection requires CtIP participation. In vitro experiments revealed that purified recombinant human CtIP protein could stimulate MRE11-RAD50 complex nuclease activity, suggesting that CtIP may redirect MRN function from DNA damage sensing to end resection. The depletion of CtIP results in attenuated recruitment of RPA and ATR to damage sites and reduced HR frequency. Double knockdown of CtIP and MRE11 reduces HR frequency to the same degree as CtIP depletion alone, which is consistent with the concept that CtIP and the MRN complex function in the same pathway.

EXO1/DNA2

As discussed above, MRE11 lacks the 5′-3′ exonuclease activity required to generate the long 3′ssDNA overhangs necessary for RPA binding [66]. Other 5′-3′ exonucleases (e.g., EXO1, DNA2) likely contribute to the extension of DNA end resection, and a two-step model has been proposed for DSB end resection [66]. The MRN/X complex and CtIP/Sae2 initiate DSB resection by removing the first 50–100 nucleotides from the DNA 5′ termini, followed by further resection to generate long 3′ss DNA tails, which are catalyzed by EXO1 exonuclease and DNA2 helicase/endonuclease [66, 117].

Exonuclease 1 is a conserved exonuclease in eukaryotes that was initially identified in Schizosaccharomyces pombe (S. pombe) [118–120]. Exonuclease 1 over-expression suppresses the DNA repair defect of MRX-complex-depleted yeast cells, suggesting that Exo1 can perform some functions of the MRX complex [121]. However, genetic experiments in yeast indicated that Exo1 is not the only activity shared with Mre11, which led to the identification of another nuclease involved in this process, DNA2 [122]. It is now believed that the second step of end resection is performed by two parallel pathways: one pathway depends on EXO1 and the other on DNA2, facilitated by the DNA helicase sgs1 in yeast and its human counterpart, Bloom syndrome protein (BLM) [90, 124–126].

Chromatin remodelers

Genomic DNA is wrapped around histone proteins to form highly condensed DNA-protein fibers that are covered by multiple factors [127]. To overcome the natural barriers that restrict access to DNA, multiple histone-modifying and chromatin-remodeling proteins are recruited to modify the structure of chromatin and hence facilitate DNA repair [128–130]. After development of DSBs, phosphorylation and subsequent acetylation events occur in histone H2A over a large region of the DNA break site to allow recruitment of chromatin unwinding and remodeling complexes [131–138]. Chromatin remodeling complexes use the energy from ATP hydrolysis to induce multiple chromatin changes, including disruption of DNA-histone contacts and repositioning of nucleosomes, which in turn allow repair protein access [115]. Several chromatin remodeling proteins, such as RSC, INO80, SWR1, and SWI/SNF, are reportedly involved in this process to overcome the barrier to repair proteins represented by the tight chromatin structure [139–144].

Recently, two groups independently reported a new ATP-dependent chromatin remodeler in budding yeast, Fun30 (SMARCAD1 in humans) [145, 146]. Fun30 is a member of a highly conserved Etl1 subfamily of Snf2 nucleosome remodeling factors and is implicated in gene silencing. Although Fun30 deletion renders cells hypersensitive to CPT, and its overexpression results in genomic instability, direct evidence of its function in DSB repair was lacking [145, 146]. Based on the observations that Fun30 localizes to DSBs and Fun30 depletion in yeast causes similar extended resection defects as Exo1/Sgs1/Dna2 mutants, but only mild defects in the initial resection step, Fun30 likely participates in DNA end resection extension [145–147]. Fun30-depleted mutants consistently fail to repair DSBs by SSA, an alternative repair pathway that requires extensive resection [145, 146].

This raised the question of whether Fun30 directly or indirectly affects end resection [145, 146]. Evidence supporting the hypothesis that Fun30 directly affects end resection includes: Fun30 is recruited and spreads at DSBs with the same kinetics as other resection extension factors; Fun30 co-immunoprecipitates with Exo1, Dna2, and RPA; and resection enzymes fail to spread further from the break site after recruitment [145, 146]. Meanwhile, evidence against the possibility that Fun30 is indirectly involved in long-range resection includes: no significant changes in transcript accumulation of end resection factors, and no nucleosome position changes are observed [145, 146].

Briefly, the hypothesis was raised that Fun30 functions with either Exo1 or Sgs1 pathways to overcome the resection barrier formed by Rad9-bound chromatin to allow the resection extension step in yeast (Fig. 1) [127, 145, 146]. Rad9 is a checkpoint adaptor protein and binds to methylated K79 residue of histone H3 through its Tudor domain. Histone-bound Rad9 inhibits DNA end resection at both DSBs and uncapped telomeres [148]. Consistently, Fun30 is less important for resection in the absence of Rad9 and either histone H3K79 methylation or γ-H2A, which are essential for Rad9 recruitment [145, 146]. Importantly, the ATPase activity of Fun30 is required for efficient resection [145, 146].

However, other chromatin remodelers, like the previously identified INO80, RSC, and SWR1, are also reportedly associated with resection, suggesting that their functions in end resection are partially redundant [149, 150]. Questions have been raised regarding why many chromatin remolding factors are involved in end resection, and how they collaborate to complete the resection efficiently. It is possible that more chromatin remodelers remain to be identified, and it will be interesting to determine if those chromatin remodelers or others are required for recovery after repairs are complete, for chromatin repacking, and their mechanisms of regulation in this dynamic process.

53BP1/Rif1/PTIP/BRCA1

53BP1 is rapidly localized to the DNA damage site after DSB occurs, and its function in checkpoint regulation was the first noticed [151–154]. Moreover, 53BP1 stimulates NHEJ in specific contexts, including BRCA1-deficient cells, class switch recombination, long-range V(D)J recombination, and fusion of dysfunctional telomeres [155–161]. The ability of 53BP1 to promote NHEJ is explained in part by its ability to block 5′ end resection at DSBs [31, 158, 162].

BRCA1, a tumor suppressor that plays multiple roles in DSB repair, positively regulates end resection by forming a complex with CtIP and MRN [58, 108, 163–166]. Double-strand breaks resection is consistently impaired in the absence of BRCA1, although this may not be due entirely to the lack of the BRCA1-CtIP-MRN complex (BRCA C complex), since cells containing the CtIP mutation, which are incapable of binding BRCA1, have normal resections at DSBs [165, 167–169]. This model explains why the two major mechanisms, HR and NHEJ, are restricted to different phases of the cell cycle. In G1 phase, 53BP1 negatively regulates resection in cells, while BRCA1 promotes 53BP1 removal in S phase to allow resection [22, 56, 158, 170].

Regarding the mechanisms underlying this process, two 53BP1-associated proteins, Rap1-interacting factor 1 (Rif1) and Pax2 transactivation domain interaction protein (PTIP), were recently reported to bind to different 53BP1 sites [171–175]. Rif1 was initially identified as part of the telomeric complex in budding yeast and acts downstream of 53BP1 to inhibit resection in mammalian cells [176]. Indeed, previous results suggested the possibility of Rif1 involvement in DSB responses and the association between Rif1 and 53BP1. In 2004, Silverman et al. reported that Rif1 foci colocalize with 53BP1 foci upon DNA damage, and Rif1 foci formation is dependent on the existence of 53BP1, not other DNA repair proteins [177]. Furthermore, Rif1 depletion in three tumor cell lines resulted in increased sensitivity to IR treatment [177]. BRCA1 is normally present but does not accumulate at DSB sites in G1 cells; if 53BP1 or Rif1 is deleted, BRCA1 forms foci at DSBs in G1 and hence promotes HR [175, 178, 179].

Pax2 transactivation domain interaction protein is known for its role in transcription initiation and was suggested to function in both HR and NHEJ [180–182]. Through its BRCT domains, PTIP interacts directly with phosphorylated 53BP1 upon DNA damage [183]. Similar to Rif1 deletion, PTIP loss also results in increased end resection [184]. In contrast to the only partial contribution of Rif1 to HR defects in BRCA1-deficient cells, PTIP ablation completely rescues HR in BRCA1-deficient cells [161]. Although the exact relationship among 53BP1, Rif1, and PTIP requires further investigation, all three are likely components of a complex that directs cells into the NHEJ pathway instead of the HR pathway [161]. Thus, emerging evidence suggests that DNA end resection is important in the selection between HR and NHEJ, and the on–off switch is mediated at least in part by the controversial roles of 53BP1/Rif1/PTIP or BRCA1.

Presynaptic filament assembly and D-loop formation: central events in homologous recombination

After DNA end resection, the exposed ssDNA overhang long chain is coated with heterotrimeric replication protein A (RPA) [185]. Replaced by RAD51 recombinase, the nucleoprotein filament invades and pairs with intact homologous donor DNA to form a D-loop structure [186].

RAD51

Two recombinases, RAD51 and DMC1, which mediate the process of pairing and strand exchange, exist in eukaryotes [186–189]. While RAD51 is required for both mitotic and meiotic HR, DMC1 is specifically expressed during, and functions in, meiosis [187, 190, 191]. Human RAD51 shares 30 % identity with its bacterial counterpart, RecA, which forms the critical nucleoprotein filament in the SOS response [192–194]. The core domains conserved among the RecA/RAD51 family proteins include two conserved nucleotide-binding motifs, Walker A and B, which are involved in ATP binding and hydrolysis activities [190, 192, 195]. In vitro experiments confirmed that mammalian RAD51 forms a helical nucleoprotein filament and further catalyzes homologous paring and strand exchange [196]. Its function in this step involves three stages. First, RAD51 displaces RPA to form the presynaptic filament [192, 197]. Second, it catalyzes strand invasion and D-loop formation [198, 199]. Finally, RAD51 dissociates from DNA to expose the 3′end required for DNA synthesis [11].

However, RAD51 loading onto ssDNA is a relatively slow process and must be facilitated by several partners to ensure efficiency and correction. First, DNA–RAD51 affinity is weaker than that of DNA–RPA, which means that RAD51 alone cannot displace bound RPA [200–202]. Second, RAD51 has a propensity to bind to dsDNA, which could inhibit presynaptic filament formation; however, this is important during strand invasion, in which RAD51-dsDNA filaments with both invading and donor DNA strands are formed [196, 203]. Increasing evidence suggests that certain recombination mediators, including the tumor suppressor breast cancer susceptibility gene 2 (BRCA2), RAD51 paralogs, and partner and localizer of BRCA2 (PALB2), are required to overcome the inhibitory effect of RPA and ensure RAD51-ssDNA filament formation and invasion [11, 204].

In addition to the recombinant mediators mentioned above that will be discussed in detail later, multiple post-translational modifications are required to regulate RAD51 activity, such as phosphorylation and SUMOylation [205, 206]. One observation suggested that depletion of SUMO E3 ligase MMS21 disrupts RAD51 foci formation, and another that RAD51 is phosphorylated at Ser-14 by casein kinase (CK2) in a DNA damage-responsive manner, which leads to its direct binding to NBS1 [207, 208].

BRCA2

As a tumor suppressor gene that is frequently mutated in breast, ovarian, and other cancers, the link between BRCA2 and HR was first recognized after its identification as a RAD51 interaction partner; subsequent overwhelming evidence suggested BRCA2 to be a mediator in HR [209–213]. Schematically, BRCA2 is a large protein that contains a DNA-binding domain (DBD) that binds to both ssDNA and dsDNA, and eight BRC repeats that bind to RAD51 [210, 214–217]. BRCA2-deficient cells are defective in RAD51 foci formation and homologous repair [212, 218–220]. Initial knowledge of BRCA2 was derived mostly from studies of its orthologs in other species, such as Brh2 in Ustilago maydis and BRC-2 in Caenorhabditis elegans; the full-length BRCA2 protein was purified recently [221–226]. In U. maydis, Brh2 was found to bind to DNA at the resected DSB ends, where both dsDNA and ssDNA exist [222, 227]. These findings, coupled with biochemical data using purified human BRCA2, which showed that BRCA2 could indeed catalyze many steps of the RPA to RAD51 transition, suggested that BRCA2 mediates RAD51 filament formation at the appropriate ssDNA sites and prevents it from binding to dsDNA [195, 226, 228, 229].

RAD51 paralogs

Five canonical RAD51 paralogs have been identified in mammalian cells: RAD51B, RAD51C, RAD51D, XRCC2, and XRCC3 [230, 231]. All five paralogs share 20–30 % amino acid identity and all are essential for cell viability and DSB repair by HR [230, 231]. They are believed to form several functional complexes in cells, including the RAD51B/C/D-XRCC2, RAD51C-XRCC2, RAD51B-RAD51C, RAD51C-XRCC3, RAD51C-RAD51D, and RAD51/D-XRCC2 complexes [230, 231].

The canonical RAD51 paralogs play essential roles in HR and the maintenance of genomic stability. Although the precise molecular mechanisms by which RAD51 paralogs regulate HR remain unclear, evidence indicates that they are involved in both the early and later stages of HR [232–235]. From the beginning, RAD51 paralogs were thought to assist RAD51 in HR initiation [231]. The yeast RAD51 paralogs, Rad55, and Rad57, form a heterodimer to augment Rad51 nucleoprotein filament stabilization and facilitate the strand invasion reaction [236, 237]. Further research demonstrated that on a RAD51 paralog-deficient background, RAD51 foci are abolished in response to IR [231]. As reported, RAD18 is recruited to IR-induced DSBs through recognition and binding to ubiquitinated proteins at the break site [231]. Through its interaction with RAD51C, one of the five RAD51 paralogs, RAD18, transmits DNA damage signals to initiate HR [238, 239]. The function of another RAD51 paralog, XRCC3, is not limited to HR initiation but is extended to later stages, such as HR intermediate formation and resolution, possibly together with RAD51C [240–242].

Interestingly, studies from our and other group have indicated that human Shu complex may represent a non-canonical “RAD51 paralog”, which functions to facilitate the action of RAD51 and may be required for other HR processes [243, 244]. The Shu complex, comprised of Shu1, Shu2, Psy3, and Csm2, was first identified in Saccharomyces cerevisiae (S. cerevisiae) in a genetic screen to identify top3 suppressors [244–248]. Further studies in S. pombe revealed that Rdl1 and Sws1 are putative Psy3 and Shu2 homologs, respectively, and that they associate with Rlp1 together to form the whole complex [244]. Mutations of any Shu complex members in either budding or fission yeast result in increased sensitivity to the DNA alkylating agent methyl metanesulfonate (MMS) and cross-linking agents [245–247]. Interestingly, mutations in all subunits of the Shu complex in yeast do not cause additive effects different from that of a single mutant, which indicates that these proteins function in the same pathway [245–247]. As a highly conserved protein in eukaryotes, hSWS1 was the first identified human Shu complex component, and hSWS1 ablation reduces the number of RAD51 foci [244]. The original hSWS1 study suggested that XRCC2-RAD51D acts together with hSWS1 to function as a human Shu complex [244]. However, in our attempt to isolate the whole human Shu complex in human cells, hSWS1-associated protein 1 (SWSAP1) was found in a tight complex with SWS1, despite the low sequence similarity between SWSAP1 and its yeast counterparts [243]. Evidence of a direct interaction between SWS1 and SWSAP1, in addition to SWS1 and SWSAP1 interdependence for stability, indicates that they are in a physiological protein complex [243]. Although both SWS1 and SWSAP1 bind to RAD51 paralogs, we questioned the notion of a SWS1-XRCC2-RAD51D complex based on TAP purification findings and false observations of the interdependence between hSWS1 and RAD51 or RAD51 paralogs [243]. Although the precise function of the Shu complex in HR remains unclear, several phenomena, such as SWS1 or SWSAP1 depletion abolishing RAD51 foci formation, suggest that the human Shu complex affects the efficiency and/or timing of HR repair.

PALB2/BRCA1

PALB2 was originally identified as a BRCA2-associated protein that is crucial for BRCA2 function [249, 250]. Interacting with ~ 50 % of cellular BRCA2 through its C-terminal domain, PALB2 is believed to be critical for BRCA2 recruitment after development of DSBs [249, 250]. Several BRCA2 mutations identified in breast cancer patients result in loss of PALB2-binding ability, and a mutation of PALB2 itself was also identified in breast cancer patients [250–255]. PALB2-knockdown cells phenocopy BRCA2 deficiency and exhibit reduced HR frequency, MMC sensitivity, and intra-S-phase checkpoint defects. Further evidence suggested that the binding between PALB2 and BRCA2 is essential for RAD51 loading onto RPA-bound ssDNA and for RAD51-coated nucleoprotein filament formation.

BRCA1 is another tumor suppressor gene that is frequently mutated in different cancers and functions in multiple DNA repair pathways, including HR, NHEJ, single-strand annealing (SSA), and checkpoint regulation. Although previous studies have shown that BRCA1 and BRCA2 coexist within the same biochemical complex, exactly how this complex is assembled remains unclear [218, 256, 257]. Interestingly, recent studies showed that PALB2 can serve as a molecular bridge between BRCA1 and BRCA2 [218, 256, 257]. Through the interaction with BRCA1 by its N-terminal domain and with BRCA2 by its C-terminal domain, PALB2 connects BRCA1 and BRCA2, and the BRCA1-PALB2 association seems to be a prerequisite for BRCA2 and RAD51 loading to the DNA damage site after DSB [256–258].

DSS1

Deleted in split-hand/split foot syndrome (DSS1) was originally identified as one of three candidate genes for an inherited congenital malformation syndrome and then found to interact with BRCA2 [259, 260]. Although the precise mechanism of DSS1 in HR is unclear, the following evidence supports its role in BRCA2 functioning: (1) approximately half of endogenous BRCA2 associates with DSS1, and the majority of DSS1 in cells interacts with BRCA2; (2) BRCA2 point mutants, which are incapable of DSS1 binding, are found in cancer; and (3) depletion of DSS1 results in BRCA2 destabilization [259, 261–263]. Studies performed in U. maydis also revealed that Dss1 interacts with Brh2 and regulates the Brh2-DNA interaction [264].

dHJ dissolution: ensuring accuracy

Besides the D-loop, double Holliday Junction (dHJ) is another key recombination intermediate in HR, which is a mobile junction between four strands of DNA [265, 266]. This structure can be cleaved to yield either crossover or non-crossover products or be dissolved to produce exclusively non-crossover products [266]. In mitotically proliferating cells, the primary mechanism for dHJ processing involves the BTR (BLM-TopoIIIα-RMI1/2) complex, which catalyzes nonnucleolytic dissolution of dHJ [267, 268]. In addition to the dissolution pathway, dHJ can also be nucleolytically processed by structure-selective endonucleases [269–271]. To date, three endonucleases in human cells have been implicated in this cleavage process, namely MUS81-EME1, GEN1, and SLX1 [272–283]. Notably, SLX4, a scaffolding protein, associates with both MUS81 and SLX1 and has been implicated in enhancing the activity of these two nucleases [278, 279, 284, 285].

BLM/BTR complex

Bloom syndrome protein is a member of the RecQ family of DNA helicases and probably the most extensively studied RecQ helicase in humans [286, 287]. BLM gene mutations cause a rare autosomal recessive genetic disorder, called Bloom syndrome (BS), first described in 1954, that is characterized by sun sensitivity, increased susceptibility to infections and diabetes, and a predisposition to a broad array of cancers [267, 288, 289]. Cells derived from BS patients display a marked increase in chromosomal abnormalities, defective response to replication stress, and a highly elevated frequency of sister-chromatid exchanges (SCEs) [267]. The frequent crossover in cells from BS patients suggested a function in ensuring HR accuracy [267]. Biochemically, BLM helicase is an ATP-dependent 3′-5′ DNA helicase capable of unwinding several DNA structures, including dsDNA substrates, G-quartet, D-loop, and dHJ DNA [286]. However, BLM alone is a poorly processive helicase and it forms large, functionally important protein complexes with other HR proteins [267].

The BTR complex, the most well-studied functional BLM complex, comprised of BLM, TopoIIIα, RMI1 (RecQ-mediated genome instability protein 1, BLAP75), and RMI2 (RecQ mediated genome instability protein 2, BLAP18), is a conserved protein complex that regulates HR in favor of non-crossover products via dHJ dissolution [290–293]. TopoIIIα is a type-I topoisomerase that functions to relieve the torsional stress that arises from DNA supercoiling by breaking the DNA backbone and then passing intact DNA through the opening before refilling the break [294, 295]. A linkage between TopoIIIα and DNA helicases was first observed in yeast and E. coli; an association between TopoIIIα and BLM was subsequently demonstrated in human cells [291, 296, 297]. The observation that deleting the TopoIIIα-interaction domain from BLM results in elevated SCEs indicates that the anti-crossover function of BLM requires TopoIIIα involvement [291, 298].

RMI1 and RMI2, are two other proteins identified in human cells that also associate with BLM [292]. Together with TopoIIIα, the BLM-TopoIIIα-RMI1-RMI2 (BTR) complex is highly organized and is referred to as the BLM dissolvasome [267, 299]. Evidence demonstrated that RMI1 could stabilize the whole BTR complex, since it interacts with both BLM and TopoIIIα, and RMI1 depletion affects their protein levels.

RMI2, first identified as a RMI1-binding partner in human cells, binds to RMI1 through the oligonucleotide/oligosaccharide-binding (OB) fold interaction [293]. The association between RMI2 with BLM and Topo IIIα is through RMI1, and RMI2 depletion in cells results in BTR complex destabilization and increased SCE [267].

SPIDR/FIGNL1

Upon DSBs, BLM localizes to the DNA damage site to form discrete foci and perform its dHJ dissolution function. Mounting evidence suggests that BLM may also be involved in other HR steps, such as end resection and D-loop formation [125, 300–304]. However, how BLM is recruited to DNA damage sites and how it collaborates with other proteins to mediate HR remain largely unexplored. Although it is known that DNA repair proteins, such as FANCM, RMI1 and TopoIIIα, contribute to BLM recruitment, we identified a new scaffolding protein involved in DNA repair (SPIDR) and demonstrated that it directly interacted with and mediated BLM foci formation [305, 306]. Multiple regulatory mechanisms may reveal the importance of BLM in DNA repair. Interestingly, SPIDR also binds to RAD51 and contributes to RAD51 recruitment after DSB, which allows it to act as a bridge between RAD51 and BLM, ensuring tight regulation between the early and late HR steps, and maximizing HR repair efficiency [306]. While other HR proteins, such as the BRCA proteins, also affect RAD51 foci formation, they have opposite effects on SCE frequency [307, 308]. Depletion of SPIDR results in an elevated SCE rate, whereas lack of BRCA1/2 leads to a reduction in the SCE frequency [305]. A simple explanation for this phenomenon is that although the overall HR efficiency is reduced in SPIDR-deficient cells, most residual HR intermediates resolve into crossover products, which leads in turn to elevated SCE frequency [305]. Almost simultaneously with our identification of SPIDR, Yuan et al. [308] found that a novel RAD51 binding partner, fidetin-like 1 (FIGNL1), is recruited to DSBs and is involved in homologous recombination repair. Although FIGNL1 interacts with RAD51 through a conserved RAD51-binding domain, FIGNL1 recruitment to DNA damage sites depends on H2AX rather than RAD51 [308]. Interestingly, the authors reported that FIGNL1 also interacts with SPIDR and suggested the existence of a new protein complex consisting of FIGNL1, SPIDR, and other uncharacterized components that accomplish unique functions in HR [308]. The exact underling mechanism of how SPIDR and FIGNL1 are involved in HR appears complex and needs further investigation.

Summary

Double-strand breaks arise from a number of endogenous and exogenous agents that cause interruptions in the continuity of the DNA double helix and are potentially lethal and highly genotoxic. NHEJ and HR are the two major mechanisms that safeguard genome integrity in eukaryotic cells upon occurrence of DSBs. Although it is generally believed that HR is an error-free repair pathway while NHEJ is error-prone, loss of precise HR regulation may lead to undesirable DNA rearrangements, since the required genetic information exchange between different DNA duplexes is potentially dangerous. For this reason, HR is highly regulated to ensure proper repair and protect against genome instability.