|Home | About | Journals | Submit | Contact Us | Français|
Homologous recombination (HR) performs crucial functions including DNA repair, segregation of homologous chromosomes, propagation of genetic diversity, and maintenance of telomeres. HR is responsible for the repair of DNA double-strand breaks and DNA interstrand cross-links. The process of HR is initiated at the site of DNA breaks and gaps and involves a search for homologous sequences promoted by Rad51 and auxiliary proteins followed by the subsequent invasion of broken DNA ends into the homologous duplex DNA that then serves as a template for repair. The invasion produces a cross-stranded structure, known as the Holliday junction. Here, we describe the properties of Rad54, an important and versatile HR protein that is evolutionarily conserved in eukaryotes. Rad54 is a motor protein that translocates along dsDNA and performs several important functions in HR. The current review focuses on the recently identified Rad54 activities which contribute to the late phase of HR, especially the branch migration of Holliday junctions.
HR is critical for maintaining genome stability in all living organisms [1,2]. HR is responsible for the repair of DNA double-strand breaks, the most harmful type of DNA damage, and for accurate chromosome segregation [3–6]. Malfunction of HR causes genome instability leading to cancer and various chromosomal abnormalities such as Down’s and other syndromes [7–10]. The salient feature of the HR mechanism is the use of homologous DNA sequences as a template to achieve high fidelity repair of DNA double-strand breaks. The process of HR involves enzymatic processing of the broken dsDNA into a resected DNA duplex with protruding 3’-ssDNA tails (Fig. 1) [11–14]. Following resection, a recombinase, Rad51 (or its prokaryotic homolog RecA), is loaded onto the ssDNA to form a contiguous helical nucleoprotein filament, which searches for an intact homologous dsDNA template [15–19]. Once the homologous sequence is found, the recombinase protein promotes the exchange of DNA strands that leads to formation of joint molecules (D-loops) [20–22]. Further steps along this pathway involve the recovery of lost information by DNA polymerase using intact DNA as a template, capture of the second end of broken chromosome, formation and branch migration of Holliday junctions, and resolution of the recombination intermediates to complete the process of DNA double-strand break repair and genetic exchange between homologous DNA molecules .
Rad54 protein belongs to the core of the enzymatic machinery that carries out HR in eukaryotes . Rad54 is a motor protein that translocates along dsDNA in an ATP hydrolysis dependent manner [24,25]. It promotes chromatin remodeling and protein displacement from dsDNA. Rad54 interacts physically and functionally with Rad51 protein [26,27] and strongly stimulates the Rad51 DNA strand exchange activity . Rad54 binds Holliday junctions and drives their branch migration . It interacts with Mus81-Eme1(Mms4), a structure-specific endonuclease, stimulating its DNA cleavage activity [30–32]. Thus, it has become increasingly clear that Rad54 is not merely a multifunctional protein; by acting through virtually all steps of HR and interacting with different protein partners Rad54 helps to link the entire HR process together. Several excellent review articles covering various activities and functions of Rad54 have been published recently [2,33–35]. The current paper places an emphasis on the recently identified Rad54 activities, especially the branch migration of Holliday junctions which contributes to the late phase of HR.
The RAD54 gene was discovered in genetic screens for Saccharomyces cerevisiae mutants that would confer strong ionizing radiation (IR) sensitivity, but only moderate ultra violet (UV) light sensitivity [36–38]. Based primarily on complementation analysis, the RAD54 gene was classified as a member of the RAD52 epistasis group that constitutes the core of the enzymatic machinery responsible for HR and DNA double-strand break repair . Together with rad51 and rad52, rad54 is one of the three most IR-sensitive single mutants in S. cerevisiae [39,40].
After Rad51, Rad54 is the second best conserved protein of the Rad52 group, a generally well conserved group in eukaryotes. For instance, S. cerevisiae Rad54 (ScRad54) shares 66% similarity and 48% identity with its human homologue [24,41,42]. Rad54 homologues have been identified in a wide range of eukaryotes including S. pombe , Arabidopsis , Drosophila , chicken , zebrafish , mouse, humans , and, at least, in one archaean genus, Sulfolobus [47,48].
In spite of the structural conservation, functions of the Rad52 group members in recombination and DNA repair across different organisms are not fully conserved. In S.cerevisiae and S. pombe, rad51 and rad54 mutants (rhp51 and rhp54, in S. pombe) display severe and almost indistinguishable phenotypes with respect to their damage sensitivities (e.g., IR, UV-light), recombination, and chromosome loss in mitotic cells [43,49]. In contrast, the phenotypes of RAD51−/− and RAD54−/− mutants are markedly different in mice, as well as in chicken cells. Disruption of the RAD51 gene causes early embryonic lethality of animals; Rad51 knockouts fail to proliferate even in cell culture [50–52]. On the other hand, disruption of the RAD54 gene results in viable mice. Still there is heightened sensitivity of mouse embryonic stem cells and chicken DT-40 cells to IR and methyl methanesulfonate (MMS) [46,53,54].
In mammals, the RAD54 gene plays the most important role during early developmental stages. Whereas mouse RAD54−/− embryonic stem cells are sensitive to IR, the adult RAD54−/− mice are no more IR-sensitive than the wild type animals, because their IR-sensitivity is rescued by the non-homologous end-joining (NHEJ) repair pathway . However, regardless of the developmental stage, RAD54−/− mice are hypersensitive to DNA damage caused by interstrand cross-linking agents (ICL), e.g., mitomycin C, which cannot be repaired by the NHEJ .
Rad54 plays an important role in reducing the rate of spontaneous chromosome loss. Rad54 inactivation causes increased rates of spontaneous chromosome loss in S. cerevisiae diploid cells , spontaneous minichrosome loss in S. pombe cells , and various spontaneous chromosome aberrations in DT-40 cells . Consistent with the role of Rad54 in the maintenance of genome stability, several missense mutations of the human RAD54 gene were linked to breast, colon and lymphoma cancers .
In addition to recombinational DNA repair, Rad54 also plays an important, but less critical, role in other classes of HR events. In S. cerevisiae, rad54 mutations cause a relatively mild defect in mating-type switching, whereas rad51 and rad52 mutations completely abolish this recombination event . Also, in S. cerevisiae, Rad54 plays a relatively minor role in meiotic recombination, due to the presence of a meiosis-specific homologue, Rdh54/Tid1 . Consequently, the rad54 mutant forms viable spores, albeit at reduced frequency (25–65%), while the rad51 mutant is fully defective in viable spore formation [61,62]. However, in the rad54 rdh54 double mutant, spore viability is reduced to the level of the rad51 mutant, indicating that the activities collectively possessed by these two homologous proteins are important for meiotic recombination . In mice, Rad54 plays a minor role in meiotic recombination as well, since RAD54−/− mutants are fertile . In higher eukaryotes, meiosis-specific Rad54 homologues remain to be identified. Mammalian Rad54B, which was initially thought to be the Rad54 meiosis-specific homolog, does not appear to have an important function in meiosis; as both RAD54B−/− and RAD54−/− RAD54B−/− mice are fertile .
In vivo, Rad54 is a moderately abundant nuclear protein. It was estimated that there are 7× 103 and 2.4 × 105 Rad54 molecules per cell in exponentially growing diploid yeast cells and in an unsynchronized population of mouse embryonic stem (ES) cells, respectively [65,66]. In comparison, there are 6.7 × 105 Rad51 molecules in mouse ES cells . In both yeast and mammals, RAD54 expression shows a cell cycle dependence; RAD54 transcription increases during late G1 phase [66–69]. The increased level of Rad54 and other recombination proteins in the S and G2 phases corresponds to the important role of HR during DNA replication and may account for the increased resistance of cells to DSB-inducing agents in G2 . In mammals, RAD54 gene expression is controlled by E2F family transcription factors; during Go and early G1 RAD54 gene expression is repressed by E2F4, and subsequently activated in G1/S phase by E2F1 . Recently, it was found that the level of Rad54 protein during the cell cycle can also be regulated post-translationally. In S. pombe, Rhp54 (a Rad54 homologue) is ubiquitylated and degraded in G1 phase . In S. cerevisiae, expression of RAD54 and RAD51, but not RAD52, is damage inducible [61,73]. In contrast, mammalian cells show no increase in the Rad54 protein level in response to DNA damage. However, treatment of mammalian cells with IR causes Rad54 and other proteins of the Rad52 group to accumulate in foci at sites of DNA damage in the nucleus [74,75].
The S. cerevisiae RAD54 gene was first cloned and sequenced in Mortimer’s laboratory [76,77]. Sequencing identified Rad54 as a member of the group of proteins involved in translocation along nucleic acids, which are broadly defined as helicases [78,79]. The common feature of these proteins is that they couple nucleotide triphosphate hydrolysis to directional movement along nucleic acids, which may or may not result in strand separation of nucleic acids. Members of this group have a role in almost every cellular process from DNA replication and repair, transcription, translation, splicing, and nuclear transport .
Of the six helicase superfamilies , Rad54 is a member of Superfamily (SF) 2 . Like all other SF2 and SF1 (Supefamily 1 helicases) proteins, Rad54 possesses seven classical “signature” motifs: I, Ia, II, III, IV, V, and VI  plus several more recently identified motifs, like TxGx (Fig. 2A). These motifs define the two tandem RecA-like folds (RecA-like Lobes 1 and 2), which constitute the “core” or the minimal translocation motor that is responsible for conversion of chemical energy of ATP into protein conformational changes and, consequently, for the mechanical motion .
The conservation of these signature motifs can clearly be seen by comparing the structures of Rad54 and the SF1 DNA helicase protein PcrA [24,81]. Based on a series of structures solved with the PcrA protein, a mechanism has been proposed to explain how ATP hydrolysis by helicases is linked to DNA translocation. In the apoenzyme, the lysine residue of motif I occupies the magnesium ion binding site. Binding of the ATP-Mg2+ complex displaces the lysine, which in turn, forms a hydrogen bond with the ATP β-phosphate. The threonine of motif I and aspartate of motif II coordinate the magnesium ion. This alignment allows the glutamate of motif II to activate a water molecule that hydrolyzes the bond between the β and γ-phosphates, which is coordinated with a glutamate in motif III, and arginines in motifs IV and VI . The arginine of motif VI, also known as the arginine finger, senses the hydrolysis and transmits the hydrolysis-induced conformation changes from the ATPase active site to the DNA binding site, resulting in DNA translocation. Mutational analysis of PcrA found that mutation of the arginine finger has the most deleterious effect on the helicase activity of any single mutation on this protein . All of these amino acid residues are conserved between PcrA and Rad54, with the only exception being the PcrA arginine of motif IV, which is instead found in motif VI of Rad54 .
SF1 and SF2 proteins were segregated based on the sequence divergence of motifs III, IV, V, and VI and spacing between the seven conserved motifs [79,83,84]. Moreover, motif III is located in different regions of SF1 and SF2 proteins.
Specifically, Rad54 belongs to the Snf2 (also known as SWI2/SNF2) protein family of dsDNA-dependent ATPases. In contrast to canonical DNA helicases, none of the Snf2 family members was shown so far to promote strand separation of duplex DNA. Instead, the Snf2 proteins are capable of chromatin remodeling, DNA topology alterations, and displacement of proteins from DNA . All of these activities are presumed to require translocation of the proteins on DNA . Several structural characteristics distinguish the Snf2 proteins from proteins of other SF2 families, e.g., DEAD-box RNA helicases, the RecQ helicase family, the DEAH helicase family, and others . Each RecA-like lobe contains one of the Snf2-specific insertions HD1 and HD2 (helical domains 1 and 2) (Fig. 2B). In the primary sequence, these protrusions are inserted between motifs III and IV, making the spacing between these two motifs much larger in Snf2 proteins as compared to the rest of the SF2 proteins; there are more than 160 residues in this region in Snf2 proteins compared to 38 and 78 for the typical SF2 helicases NS3 and RecG, respectively . Besides these protrusions, the Snf2 family is defined by 15 conserved blocks designated Snf-A to N (block A is also known as the TxGx motif) (Fig. 2B) . The N-terminal domain (NTD) is unique to Rad54 orthologs (Fig. 2C). Proteolysis studies have shown the Rad54 NTD is relatively unstructured , and has been shown to contain the Rad51 binding site [26,27,89–91].
X-ray crystallographic structures of large regions of two Rad54 orthologs, from zebrafish and Sulfolobus solfataricus, have been solved [24,47,92]. The zebrafish Rad54 structure includes a part of the N-terminal domain, the two RecA-like α/β-domains (lobe 1 and 2) found in all SF1 and SF2 helicases, and the C-terminal domain (CTD) . Each RecA-like lobe contains one insertion HD1 and HD2, respectively, which appear as protrusions [24,47]. The CTD appears to be the only element unique to Rad54 that is present in the truncated structures. The CTD contains a tentative Zn-coordinating motif that may stabilize its entire assembly (Fig. 2D).
A crystal structure of the catalytic domain of the S. solfataricus Rad54 homolog bound to a dsDNA substrate has given an insight into the mechanism of Snf2 ATPases translocation along dsDNA. The crystal structure indicates that the Rad54 RecA-like bi-lobal core requires at least 12–15 base pairs to grab DNA. S. solfataricus Rad54 binds both strands of the dsDNA through mainly phosphodiester backbone contacts, along the DNA minor groove. However the most extensive protein:DNA contacts are made with one strand of the duplex, indicating the particular importance of one DNA strand in the translocation mechanism. Rad54 binding does not cause strand separation; and the bound dsDNA retains a B-form conformation . Overall, the Rad54 structural data are consistent with the inchworm translocation mechanism, similar to that described for other helicases , however important details of this mechanism remain to be clarified.
The biochemical properties of Rad54, especially the yeast and human orthologues, have been extensively characterized. As expected for an Snf2 protein, Rad54 was found to be an ATPase with a strict dependence on dsDNA [28,93,94]. Both yeast and human Rad54 are robust ATPases that hydrolyze ATP with a kcat ~ 3000–6000 min−1 ; our unpublished observations); earlier kcat estimates gave lower kcat values (1000–2000 min−1) probably due to suboptimal assay conditions. The KM ATP of the HsRad54 ATPase activity is approximately 400–500 µM (; our unpublished observations).
Rad54 binds ssDNA and dsDNA with similar affinities, although binding to ssDNA does not stimulate ATP hydrolysis [28,94,96]. Compared to linear ssDNA or dsDNA, Rad54 shows significantly higher binding affinity for branched DNA structures, with the strongest binding preference for the partial Holliday junction, in which one of the four arms is represented by ssDNA [29,95]. The affinity of HsRad54 for the PX junction is, approximately, 200-fold higher than that for ssDNA or dsDNA fragments of the same length .
Like other members of the Snf2 family, but unlike canonical DNA helicases, Rad54 does not display DNA strand separation activity, at least, in conventional DNA helicase assays . However, like canonical helicases, Rad54 can translocate on DNA. DNA translocation of Rad54 was first inferred from biochemical data. It was found that Rad54 produces topological changes in DNA structure, introducing equivalent positive and negative supercoiled domains into closed circular dsDNA in an ATP hydrolysis dependent manner [75,93,97,98]. It is known that DNA translocating proteins generate positive supercoils in the DNA ahead and negative supercoils behind the protein [99,100]. Although on covalently closed circular plasmid the positively and negatively supercoiled regions annihilate each other, their mutual annihilation may be prevented by the presence of two (or more) oppositely oriented translocation complexes (Fig. 3). Consistent with Rad54’s ability to translocate on DNA, the ATP hydrolysis rate of Rad54 shows an increase with the increase of dsDNA length , a property shared with other DNA translocating ATPases [102,103]. Also, as expected for DNA translocating protein, Rad54 can dissociate a DNA triple-helix  in the assay that was originally developed to follow the translocation of a type I restriction endonuclease along DNA . Finally, translocation of ScRad54 and ScRdh54 (Tid1), a Rad54 meiosis-specific homologue, along single molecules of dsDNA was visualized directly, demonstrating rapid and highly processive movement at ~300 bp/s for ScRad54 and 80–120 bp/s for ScRdh54 [25,86,106].
The precise structure of the active Rad54 complex that tranlocates on DNA remains unknown. Generation of supercoils in covalently closed DNA by a translocating protein requires a very large resistance to the rotational motion of the protein around the DNA, which can be achieved, if a translocating enzyme interacts with two or more sites on DNA, e.g., within a multimeric complex formed by the enzyme [97,99]. Indeed, while both yeast and human Rad54 protein exists as a monomer in solution  (Mazina and Mazin, unpublished results), Rad54 binding to DNA causes oligomerization of the protein ; the cross-linking experiments indicated formation of a dimer, as the principal oligomeric species, and larger oligomers [93,95]. Formation of oligomers by HsRad54 and ScRad54 in complexes with DNA was also detected using atomic force microscopy  and electron microscopy .
Like other members of the Snf2 family, Rad54 promotes chromatin remodeling, i.e., nucleosome redistribution along DNA, in an ATP hydrolysis dependent manner [104,108–112]. This Rad54 activity likely depends on its dsDNA translocation activity. Chromatin remodeling may also involve specific interactions of Rad54 with histones, as the N-terminus of ScRad54 interacts specifically with the N-terminal tail of histone H3 . It was suggested that chromatin remodeling activity of Rad54 may serve at least two functions in HR. First, Rad54 can start to clear nucleosomes and other chromatin-bound proteins at the break site prior to end-processing. This hypothesis is supported by in vivo studies showing an effect of Rad54 on the accessibility of the HML donor site for the HO endonuclease . Second, Rad54 might also facilitate the search for homology during synapsis by contending with chromatin structure of the target DNA. Indeed, in vitro studies show that Rad54 stimulates DNA strand exchange of Rad51 on chromatin-loaded DNA substrates to a greater extent than on naked DNA substrates, . More work, however, is needed to fully appreciate the role of the Rad54 chromatin remodeling activity in vivo.
Rad54 performs its functions in an association with Rad51, a protein which promotes DNA strand exchange, a basic step of HR. Interactions between these two proteins are extensive and critical to the function of HR in eukaryotes [1,26,113]. It was found that Rad54 and Rad51 interact functionally and physically. The molecular mechanisms of these interactions were the subject of extensive biochemical studies over last decade. In S. cerevisiae, over-expression of Rad54 can suppress certain repair phenotypes of rad51 mutants [26,114]; and the rate and extent of Rad51 recruitment to the HO-induced DSB is significantly reduced in the absence of Rad54 . In mouse ES cells, IR-induced Rad54 foci co-localize with Rad51 foci . Moreover, Rad51 foci formation shows dependence on Rad54. Physical interactions between Rad51 and Rad54 proteins are species-specific and conserved from archea to humans [26,48,90,91,114]. Experiments with truncated Rad54 mutants show that the Rad54 N-terminal domain is primarily responsible for interactions with Rad51 [27,89]. Importantly, Rad54 physically interacts both with free Rad51 protein and with Rad51, when it is a component of the Rad51-nucleoprotein filament, which is the active species in DNA strand exchange .
The role of Rad54 in HR was significantly clarified by the key discovery made by Galina Petukhova and Patrick Sung, who found that S. cerevisiae Rad54 strongly stimulates the DNA strand exchange activity of Rad51 .This stimulation is evolutionarily conserved: arheal, Drosophila, and human Rad54 orthologs stimulate DNA strand exchange activity of their cognate Rad51 [48,101,108,117]. It was found that the stimulation depends on the ATPase activity of Rad54 [28,93], indicating that Rad54 translocation on dsDNA plays a role in stimulation of DNA strand exchange activity. The mechanism of DNA strand exchange stimulation involves formation of a complex between Rad54 and the Rad51 nucleoprotein filament [98,118,119]. In this complex, translocation of potential target DNA by Rad54 is linked to the DNA homology search process promoted by the Rad51-nucleoprotein filament. It was proposed that the translocation activity of Rad54 may both provide a more efficient delivery of dsDNA to the site of the homology search within the filament and cause transient disruption of dsDNA base pairs as a consequence of DNA translocation. Such disruption of normally stable base pairs may make them more available for interaction with the ssDNA bound within the Rad51 nucleoprotein filament.
Formation of a tripartite complex of Rad51-Rad54-ssDNA has significant and synergistic effect on the activities of both Rad51 and Rad54. While Rad54 stimulates the DNA strand exchange activity of Rad51, in its turn, Rad51 stimulates the biochemical activities of Rad54, the dsDNA-dependent ATP hydrolysis and DNA topology modification activity (generation of positive and negative supercoils) [98,117,118]. Also, Rad51 increases the processivity of Rad54 DNA translocation along DNA , though the effect of Rad51 on the rate of Rad54 translocation remains to be determined. Finally, Rad51 stimulates chromatin remodeling [104,108,109,111,112] and DNA branch migration activity of Rad54 protein .
While ATPase-dependent dsDNA translocation by Rad54 may play a critical role in stimulation of DNA strand exchange, Rad54 employs additional mechanisms of Rad51 stimulation. By forming a complex with the Rad51 nucleoprotein filament, Rad54 stabilizes the filament; Rad54 was shown to increase i) the filament resistance against dissociation at elevated salt concentrations, ii) the Rad51-dependent protection of dsDNA against cleavages by restriction endonucleases, and iii) the Rad51 ability to compete with RPA for ssDNA binding [115,116]. Thus, Rad54 may play a role of a mediator protein that specifically stimulates Rad51 by increasing its competiveness with RPA for ssDNA binding during the filament formation . The filament stabilization function of Rad54 does not depend on its ATPase activity, the Rad54 ATPase-deficient mutant is still proficient in Rad51 filament stabilization . The fact, that the ATPase activity of Rad54 is required for its biological function [65,93] indicates that the ATPase activity of Rad54 is essential for the steps of HR which occur after filament formation, but it does not rule out the biological significance of the Rad51-DNA filament stabilization by Rad54. Indeed, it was shown that the S. cerevisiae strain expressing the ATPase-deficient rad54K341R, but not rad54Δ allele, is proficient in recruitment of Rad51 to the site of DNA double-strand breaks, consistent with the stabilization of the Rad51-DNA filament by Rad54 .
While Rad54 can stabilize the Rad51 filament formed on both ssDNA and dsDNA, paradoxically, Rad54 was also shown to disrupt the Rad51-dsDNA filament [122,123]. The active form of Rad51 is the filament formed on ssDNA. However, by analogy with E. coli RecA protein , it is likely that Rad51 remains associated with the dsDNA heteroduplex, the product of DNA strand exchange. Therefore, dissociation from dsDNA would help recycling Rad51 for new rounds of recombination.
Co-existence of two apparently opposing Rad54 activities, filament stabilization and dissociation, may seem difficult to reconcile. However, as it was discussed above, Rad54 has two distinct binding sites, the N-terminal Rad51 binding site, and the motor core dsDNA binding site. It is therefore likely that Rad51 displacement from dsDNA and stabilization of the Rad51-DNA filament is mediated by interaction with two different Rad54 sites. The displacement of Rad51 protein from dsDNA occurs at the Rad54 dsDNA binding site during protein translocation along dsDNA, whereas Rad51 filament stabilization is mediated by interaction of Rad51 protein monomers with the Rad54 specialized binding site (Fig. 4). This model agrees with previous observations that Rad54 can stabilize Rad51 filaments on both ssDNA and dsDNA, but can displace Rad51 only from dsDNA [116,122]. It also agrees with the observation that Rad54 can disrupt only partially saturated Rad51-dsDNA complexes, in which protein-free dsDNA regions are available for initiation of Rad54 translocation [122,123]. Finally, the model is supported by the fact that both displacement of Rad51 from dsDNA and DNA translocation require ATP hydrolysis by Rad54 (and by Rad51) , whereas stabilization of the Rad51-ssDNA filament is ATPase independent [110,116].
Although Rad54 protein is likely to participate in HR at different stages [2,34], genetic data indicate the significance of the post-synaptic function of Rad54, downstream of Rad51 [1,35,113]. Critical evidence for the post-synaptic function of Rad54 was obtained from the analysis of rad54 srs2 double mutants. Thus, while the rad51 srs2 double mutant is viable, the rad54 srs2 double mutant is not . Importantly, the rad51 mutation suppresses the synthetic lethality of rad54 srs2 , indicating that Rad51-generated recombination intermediates are normally resolved by Rad54 and Srs2, but became lethal in the absence of these two proteins.
Post-synapsis comprises the steps after D-loop formation and includes 1) heteroduplex extension and formation of Holliday junctions, 2) priming of DNA synthesis from the invading 3’-OH end, 3) branch migration of Holliday junctions, 4) resolution of Holliday junction intermediates, and 5) sealing of the strands to restore two intact and contiguous duplex DNAs (Fig. 1). Current data indicate that Rad54, a truly versatile protein, may contribute to several post-synaptic steps. In vitro, it was shown that ScRad54 stimulates heteroduplex extension during DNA exchange promoted by ScRad51 . In vitro, both ScRad54 and its meiosis-specific homologue ScRdh54 catalyze ScRad51 removal from dsDNA in an ATP-dependent fashion [122,128], presumably to provide an access of DNA polymerases to the invading 3’-OH end for DNA synthesis . Chromatin immunoprecipitation (ChIP) analysis that monitored the recruitment of ScRad54 protein to an HO endonuclease-induced DSB demonstrated that ScRad54 is required for the creation of a mature strand invasion structure after synapsis. This result is consistent with a role of Rad54 in stimulation of DNA synthesis [110,130]. Finally, it was recently found that Rad54 binds efficiently to Holliday junctions and promote their branch migration with high efficiency in an ATP hydrolysis dependent manner [29,95,120]. Also, Rad54 stimulates the cleavage activity of Mus81-Eme1(Mms4), a structure-specific endonuclease, that cleaves Holliday junction-like structures [31,32]. In this section we focus on the DNA branch migration activity of Rad54.
The structure formed in the process of HR, when the strands are exchanged between two different DNA molecules is known as a Holliday junction (also called X-junction or 4-way junction) (Fig. 1). Formation of this structure was first predicted by Robin Holliday , whose model for HR provided molecular basis for both gene conversion and crossingover. The remarkable feature of the Holliday junction is the ability to branch migrate along the DNA axis. Branch migration of a Holliday junction is a process in which one DNA strand is progressively exchanged for another (Fig. 5). Branch migration of Holliday junctions may serve several important functions. It may extend or shorten the heteroduplex DNA, formed after DNA strand invasion, affecting the length of conversion tracks and thereby the amount of genetic information transferred between the two DNA molecules . Branch migration may cause dissociation of recombination intermediates and thereby affect the choice between a crossover and non-crossover pathway by which recombination will proceed . An increasing number of genetic and biochemical data indicate that formation and branch migration of Holliday junctions may also play a critical role in re-starting stalled replication forks during cell recovery after DNA damage[133–137].
The structure of the Holliday junction was analyzed by different methods including comparative gel electrophoresis, molecular modeling, fluorescence resonance energy transfer (FRET) assay, NMR and by X-ray studies . Surprising discoveries on the Holliday junction conformation have been made pertinent to the branch migration mechanism. Thus, early models of HR assumed a parallel orientation of DNA strands of the Holliday junction, because it was expected that the homologous duplexes would be aligned side by side during DNA strand exchange [138,139]. However, the antiparallel conformation of the Holliday junction has been firmly established by X-ray studies [140–142] (Fig. 6A). These results mean that at least one homologous chromosome must locally rotate 180° to keep the global side by side chromosome orientation and to maintain the antiparallel conformation of the Holliday junction at the same time.
The Holliday junction may exist in the stacked or unstacked (extended) conformation. In the stacked conformation, pairs of helical arms stack coaxially to form nearly continuous criss-crossed duplexes (Fig. 6B). In the unstacked one, the Holliday junction forms a four-fold symmetric nearly planar structure (Fig. 6C). X-ray studies show that the conformation of both duplexes composing the Holliday junction is indistinguishable from standard B-form DNA, with the exception of the crossovers at the junction; the Holliday junction has all standard Watson-Crick base pairs [140–142] (Fig. 6B). The compactness of the junction results in a sterical clash of four phosphate groups (within 6.5 Å of each other) in the DNA backbone at the point of the crossover. In the presence of metal ions such as magnesium or calcium (at 100 µM and greater) that neutralize electrostatic repulsions caused by the phosphates, Holliday junctions exist in the stacked conformation (Fig. 6B), whereas in the absence of metal ions, the DNA junction is fully extended (Fig. 6C). The switch between the extended and stacked conformations is important for proteins acting on Holliday junctions. Some of them, such as the RuvA protein that promotes loading of the RuvB branch migration protein on DNA, shows a binding preference to the Holliday junctions in the extended conformation , whereas other proteins, such as the hMSH4–hMSH5 complex, preferentially bind the Holliday junction in the stacked conformation .
Branch migration occurs while the Holliday junction is in an extended state, and it stops when the molecule adopts a stacked conformation (Fig. 6D) . A single step of branch migration requires the breakage of two base pairs at the point of strand exchange located on diametrically opposite arms, rotation around, and formation of the terminal base pairs on the remaining two arms. Rotation of DNA strands during branch migration promoted by RuvAB was observed in real time using single-molecule techniques . In the absence of bivalent ions, when the Holliday junction exists in the open conformation, preferable for branch migration (Fig. 6D), the rate of spontaneous branch migration is extremely fast with a step time of 300–400 µs , which corresponds to a migration rate of 2500–3300 bp/sec. In the presence of magnesium (>100 µM), branch migration is quite slow with an estimated rate of about 3.3–3.7 bp/sec . Because spontaneous branch migration proceeds bi-directionally as a random walk process, under physiological conditions spontaneous branch migration would be inefficient and therefore requires the assistance of specialized enzymes, which increase the overall rate of branch migration by providing directionality and processivity to the process.
The Holliday junction is a substrate for structure-specific nucleases which are responsible for its cleavage and formation of crossover products, and for proteins which promote its branch migration [148,149]. In prokaryotes, RuvAB and RecG enzymes promote branch migration of Holliday junctions [148,150–153]. More recently, several eukaryotic branch migration proteins have been discovered. Known branch migration enzymes belong to the ATPases associated with diverse cellular activities (AAA+) family (RuvB) and the SF2 helicase superfamily (all other proteins) (Table 1). These SF2 proteins include (1) RecG, a member of the RecG family; (2) several members of the RecQ helicase family; (3) FANCM, a member of the DEAH helicase family; and (4) Rad54, a member of the Snf2 family of DNA translocases. In humans, members of the evolutionarily conserved RecQ family include BLM, WRN, RECQ1, and RECQ5 [154–158]. The RecA family recombinases can also promote branch migration of Holliday junctions [159,160], however because of their unique nature and specific branch migration mechanisms these proteins need to be discussed separately.
Despite structural diversity among SF2 and AAA+ branch migration proteins, all of these proteins i) are capable of translocation on DNA in an ATPase-dependent manner, and ii) show high binding affinity to the Holliday junction (or other cruciform DNA structures that are capable to branch migrate). In the RuvAB protein complex, these two functions are segregated between RuvA and RuvB proteins, which function as the structure-specific DNA-binding protein and the motor protein, respectively . In other branch migration proteins these two functions are performed by the same polypeptide. In RecQ helicases, the structure-specific DNA-binding region is likely associated with the HRDC domain located in the C-terminus of these proteins, separately from the motor domain . In Rad54 and FANCM, the domains responsible for binding to the Holliday junction remain to be identified. In addition to branch migration activity, all branch migration enzymes, except Rad54, possess DNA helicase activity, albeit weak, with either the 3’→5’ or the 5’→3’ (RuvAB only) polarity.
Like other branch migration proteins, Rad54 can translocate on dsDNA in an ATP hydrolysis dependent manner  and bind efficiently to synthetic Holliday junctions [29,95]. HsRad54 shows an approximately 30-fold and 200-fold higher affinity for the X-junction and the partial X-junction (PX-junction), respectively, comparing with dsDNA. The PX-junction structurally resembles one end of a D-loop, the product of DNA strand invasion – an appropriate substrate for DNA branch migration proteins. Despite extensive biochemical studies, the branch migration activity of Rad54 eluded detection for a long time. Conventional DNA substrates (Fig. 7A) that were developed previously for enzymes that combine branch migration and helicase activity are unsuitable for Rad54, which does not have canonical helicase activity . To demonstrate the branch migration activity of Rad54 a fully movable X-junction has been constructed (Fig. 7B) [29,162]. This substrate can branch migrate until complete DNA strand separation without a need for a helicase activity. In addition to 4-stranded branch migration, Rad54 catalyzes 3-stranded branch migration of PX-junction (Fig. 5B). As expected, the branch migration activity of Rad54 depends on ATP hydrolysis . The branch migration activity of Rad54 was also demonstrated by using plasmid-size DNA substrates. It was found that HsRad54 protein efficiently promotes branch migration Holliday junctions through DNA regions of several thousand base pairs (Fig. 8A) . On similar DNA substrates, the rate of branch migration promoted by HsRad54 is comparable with that of RuvAB, a classical branch migration protein from E.coli (Fig. 8). Although HsRad54 does not have canonical helicase activity, it can promote branch migration of Holliday junctions through regions heterology up to six bases on short synthetic DNA substrates, or even longer stretches of heterology on plasmid DNA (M. Rossi and A. Mazin, unpublished data) (Table 1).
As described above, Rad54 stimulates the DNA pairing activity of Rad51. Interestingly, it was shown that ScRad54 can specifically stimulate the heteroduplex extension step in the 3-strand DNA exchange reaction promoted by ScRad51 . The mechanism of this stimulation remains controversial, primarily because the mechanism of heteroduplex extension promoted by Rad51 is not itself well understood. For instance, it was suggested that heteroduplex extension could be viewed as a succession of individual DNA strand invasion steps promoted by the Rad51 nucleoprotein filament . In this case, stimulation of heteroduplex extension may be mechanistically similar to stimulation of Rad51 DNA strand exchange activity, which involves formation of a Rad54-Rad51-DNA complex (see the previous section). Alternatively, Rad54 can stimulate heteroduplex extension due to its inherent DNA 3-strand branch migration activity . The former mechanism is supported by the observation that ScRad54 shows species-specificity in stimulation of the heteroduplex extension, i.e., it stimulates heteroduplex extension promoted by ScRad51, but not by EcRecA, indicating an important role of Rad54-Rad51 protein-protein interactions . Conversely, it was shown that the 3-strand branch migration activity of HsRad54 is inhibited on substrates that are fully covered by HsRad51 , arguing against the latter mechnaism. Further work is needed to determine whether the observed stimulation is due to an ability of Rad54 to enhance Rad51 DNA pairing through protein-protein interactions or due to the Rad54 DNA branch migration activity. It is also possible that both mechanisms contribute to the stimulation of heteroduplex extension to a certain degree.
The smallest branched DNA substrate that efficiently stimulates ATP hydrolysis consists of three DNA arms, two short dsDNA arms, 15 bp each, and an ssDNA arm of 45 nt . This result is consistent with the crystallographic data showing that the S. solfataricus Rad54 catalytic core domain can accommodate a 15 bp dsDNA fragment . A stiochiometric titration of HsRad54 with this “minimal” DNA substrate revealed that only two HsRad54 monomers were needed to achieve the maximal rate of ATP hydrolysis . In contrast, a significantly higher HsRad54 stoichiometry, 10±2 Rad54 monomers per junction, was required to achieve the maximal rate of branch migration. Formation of these larger complexes probably involves Rad54 monomer-monomer interactions, because the ATPase activity that depends on DNA binding did not increase in these complexes, compared to dimers. These data indicate that the HsRad54 dimer is responsible for initial specific recognition of the Holliday junction, whereas a large10±2 complex is required for branch migration per se. Previously, it was shown that different multimeric forms of RECQ1 are also associated with different enzymatic activities of the protein .
At the initial steps of HR, Rad51 protein promotes formation of joint molecules (D-loops) (Fig. 1). Following D-loop formation, the 3’ end of the invading ssDNA is extended by DNA polymerase, retrieving the information lost at the site of the break. Afterward, the extended joint molecules formed during the initial steps of HR may continue down one of two pathways (Fig. 1) [165,166]. Either they dissociate, leading to rejoining of the broken chromosome through synthesis-dependent strand annealing (SDSA) (Fig. 1A) , or they proceed through capture of the second processed DNA end, producing single or double Holliday junctions (Fig. 1B) [167,168], which are later resolved by structure-specific endonucleases . Dissociation of joint molecules (D-loops) after the completion of template-dependent DNA synthesis represents an essential step of DSB repair via the SDSA mechanism . Recent data show that HsRad54 can indeed promote dissociation of D-loops formed by HsRad51 protein owing to its ATP-dependent branch-migration activity (Fig. 9A). Importantly, HsRad54 can dissociate native (non-deproteinized) D-loops containing HsRad51 protein, which mimic in vivo recombination intermediates. HsRad54 can also dissociate double D-loops , which can be generated by Rad52 through annealing of the second end of broken DNA to the single D-loops [132,169–171]. It was demonstrated that dissociation of double D-loops by HsRad54 occurs through a mechanism, in which D-loop dissociation is directly associated to rejoining of the broken DNA ends without a reannealing step (Fig. 9B) . Double D-loops can lead to formation of double Holliday junctions and then crossovers, which in somatic cells might lead to detrimental loss of heterozygosity and lead to genome instability. Therefore, double D-loop dissociation by Rad54 may play an important role by promoting HR via non-crossover pathways (Fig. 1); see discussion in .
Genetic data support the hypothesis that its D-loop dissociation activity may contribute to the function of Rad54 during the late stages of HR. Specifically, the ability of Rad54 to dissociate D-loops in vitro agrees with the reduced length of gene conversion tracts in yeast strains that overexpress Rad54 protein .
In vitro, DNA intermediates containing a double Holliday junction can also be dissolved by BLM in a complex with topoisomerase IIIα and BLAP75 [174,175]. This activity implies a possible backup role of BLM during HR in somatic cells. Failure of Rad54 to dissociate D-loops, particularly double D-loops, may lead to generation of double Holliday junctions that can still be resolved by BLM-topoisomerase IIIα-BLAP75. Consistent with this view, RAD54−/− BLM−/− double knockouts in chicken DT-40 cells increase chromosomal instability to a level that is much greater than that observed for single knockouts of each gene .
Thus, DNA branch migration activity of HsRad54 protein may play a part in the double-stranded break repair mechanism postulated by the SDSA model  (Fig. 1A). According to this model, the invading ssDNA strand, after its extension by DNA polymerase, is displaced from the template DNA strand and then re-anneals with the second end of the broken DNA molecule. Further steps of double-stranded break repair via the SDSA mechanism lead predominantly to the formation of non-crossover recombinants (Fig. 1A).
Since interaction between Rad51 and Rad54 plays such a significant role in HR, some of the factors that affect this interaction may have an important regulatory effect on HR. In vitro, HsRad51 requires Ca2+ to form an active, stabile nucleoprotein filament capable of performing the strand exchange reaction . Under these conditions, HsRad54 stimulates the DNA strand exchange activity of HsRad51 . However, when Ca2+ is depleted, e.g., by adding EGTA, the HsRad51-DNA filament adopts an inactive conformation, and HsRad54 then dissociates D-loops structures containing HsRad51 . Thus, the conformational state of the HsRad51-ssDNA filament may regulate the outcome of the filament’s interaction with HsRad54.
It was recently discovered that S. cerevisiae Hed1 protein may affect interactions between Rad51 and Rad54 . Hed1 specifically disrupts interactions between ScRad51 and ScRad54, thereby downregulating ScRad51 activity. At the same time, Hed1 leaves the interaction between ScRad54 and ScDmc1 intact. It was proposed that preferential association of ScRad54 with ScDmc1 facilitates formation of crossovers, which are required for correct segregation of homologous chromosomes during meiosis.
Rad54 protein was shown to interact physically with Mus81, a structure-specific endonuclease; and this physical interaction between Rad54 and Mus81 proteins is evolutionarily conserved in eukaryotes [30–32,178,179]. Mus81 is a member of the XPF/MUS81 family of nucleases, which share a highly conserved motif (V/IERKX3D) that constitutes an integral part of the endonuclease catalytic site. Mus81 functions as a heterodimer with a non-catalytic partner protein known as Eme1 in fission yeast and humans or as Mms4 in budding yeast and Drosophila . Mus81 was first proposed to have a role in the resolution of Holliday junctions in studies on S. pombe, where Mus81 was identified as a critical factor for the production of viable spores, for survival under conditions that lead to stalling of replication fork progression, and for viability in the absence of RecQ helicase, Rqh1 [181,182]. All of the defects in S. pombe mus81 mutants were rescued by expression of RusA, a bacteriophage resolvase that is highly specific for Holliday junctions [181,183]. Mus81 protein is widely conserved among eukaryotes, including S. cerevisiae , S. pombe , Arabidopsis thaliana , mice [178,185], and humans . The mus81 and eme1 (mms4) mutants show meiotic defects in S. cerevisiae and mice, although not as severe as in S. pombe [187 ,188,189]. The role of Mus81-Eme1 as a component of an authentic nuclear Holliday junction resolvase remains controversial, because in vitro ScMus81 and HsMus81 exhibit very little cleavage activity on Holliday junctions, while delivering robust cleavage of a variety of other branched DNA structures, including D-loops, nicked Holliday junctions, forks and 3’-flaps [190–195]. Recently, another eukaryotic nuclear Holliday junction resolvase Gen1/Yen1, which has a strong preference for intact Holliday junctions, was identified . The specific contributions made by Mus81 and Gen1/Yen1 in resolving Holliday junctions remain to be elucidated.
The functional significance of Rad54 interaction with Mus81-Eme1(Mms4) was confirmed in the experiments demonstrating that HsRad54 and ScRad54 stimulate DNA cleavage activity of HsMus81-Eme1 and ScMus81-Mms, respectively [31,32]. Stimulation of DNA cleavage activity of Mus81-Eme1(Mms4) by Rad54 has important biological implications because it shows a functional link between Rad54, a branch migration protein, and Mus81–Eme1(Mms4), a structure-specific endonuclease. Interestingly, Bloom’s syndrome helicase (BLM), another human protein that promotes branch migration of Holliday junctions, can also stimulate the DNA cleavage activity of HsMus81-Eme1 . However, stimulation of HsMus81-Eme1 cleavage activity is specific for HsRad54 and BLM, and is not a generic property of all branch migration proteins, as ScRad54 and RECQ1 do not stimulate HsMus81-Eme1 .
The fact that Rad54 shows a high affinity for branched DNA and that Mus81-Eme1(Mms4) stimulation requires species-specific interactions with the cognate Rad54 are consistent with a model in which Rad54 targets Mus81 to DNA junctions via protein-protein interactions [31,32]. This model is also supported by the observation that the stimulation of HsMus81-Eme1(Mms4) activity occurs only if Rad54 is added to DNA-junctions first or, at least, concurrently with Mus81-Eme1(Mms4), indicating that stimulation occurs in cis and that formation of Rad54 complexes on DNA-junctions is important for Mus81-Eme1 stimulation. No stimulation of the HsMus81-Eme1 cleavage activity by HsRad54 was observed in the absence of a nucleotide cofactor; ADP and ATPγS support stimulation although at a smaller degree than ATP . The process of branch migration promoted by HsRad54 did not appear to be essential for stimulation of the HsMus81-Eme1 cleavage activity. Stoichiometric titration data indicate that formation of a multimeric hRad54 complex is required for stimulation of HsMus81-Eme1 . Although Rad54 binding to DNA and multimerization occurs in the absence of nucleotide cofactors , HsRad54 forms multimers of different size in the presence and in the absence of ATP . It therefore seems possible that in the presence of nucleotide cofactors, HsRad54 may form nucleoprotein complexes that either more efficiently stimulate loading of HsMus81 on DNA junctions or facilitate HsRad54 dissociation from DNA junctions, thereby providing HsMus81 with an access for DNA junction. Surprisingly, the stimulation of ScMus81-Mms4 by ScRad54 does not require a nucleotide cofactor . This result may indicate important differences between the mechanisms of Mus81 stimulation by human and by yeast Rad54 orthologs, which remain to be further explored.
Stimulation of Mus81-Eme1 activity by Rad54 in vitro is consistent with genetic data, indicating that Rad54 and Mus81 act in the same DNA repair pathway in response to UV-damage, to replication-blocking chemicals (camptothecin and hydroxyurea) in S. cerevisiae [30,32] and in response to ICL-inducing agents (mitomycin C and cisplatin) in mouse ES cells . However, Mus81 and Rad54 are not obligatory functional partners: whereas Mus81−/− ES cells are hypersensitive to hydroxyurea, Rad54−/− ES cells are not . Also, in S. pombe Mus81 and Rad54 likely act in parallel but distinct pathways of DNA repair, as mus81rad54 double mutants have severely reduced viability . However, poor growth and hypersensitivity of S. pombe mus81 rad54 double mutants to genotoxic agents is suppressed by loss of upstream proteins, such as Rad51 and Rad55 , demonstrating a link between the function of these two proteins and HR. Thus, genetic data together with the results on Rad54 stimulation of Mus81 endonuclease activity suggest that Rad54 and Mus81 may cooperate in processing of Holliday junctions intermediates which are formed during the repair of DSB or stalled replication forks.
Rad54 is one of the key proteins of HR that plays a crucial role in maintaining genome integrity. Rad54 has an important function in DNA repair and recombination (Fig. 10). It is a motor protein that translocates on duplex DNA in an ATPase-dependent manner. Through the network of interacting protein partners and unique set of biochemical activities, Rad54 helps to link together early and late steps of HR. At early steps, Rad54 may facilitate HR events by helping Rad51 filament formation on ssDNA through specific protein-protein interactions with Rad51. It may promote HR by strongly stimulating the DNA strand exchange activity of Rad51 protein and catalyzing chromatin remodeling. At later steps, Rad54 may promote an ATP dependent branch migration of Holliday junctions. This is a remarkably complex activity, in which Rad54 binds specifically to the Holliday junction, forms a multimeric active complex, and drives branch migration by coupling chemical energy consumption with mechanical motion. The transition from the Rad54 early function in stimulating Rad51 to branch migration of Holliday junctions remains to be understood. Rad51 stimulates Rad54 branch migration activity likely by promoting loading of Rad54 on the Holliday junction. However, Rad54 branch migration does not require the presence of Rad51, indicating that different types of Rad54-Rad51 nucleoprotein complexes are responsible for DNA strand exchange and branch migration of Holliday junctions. Also, Rad54 physically interacts and stimulates the DNA cleavage activity of the structure-specific endonuclease Mus81-Eme1(Mms4), which may play an important role in resolution of HR intermediates. The mechanism of this stimulation likely involves recruitment of the Mus81 endunuclease to the Holliday junction-like structure by Rad54. The wealth of knowledge accumulated on the Rad54 functions and activities through extensive genetic, biochemical, and structural studies provides a strong momentum for the outgoing research, which will undoubtedly deepen our understanding of the mechanisms of the Rad54 activities and their regulation and will help to better define the function of Rad54 in HR.
This work was supported by the NIH Grant CA100839, and the Leukemia and Lymphoma Society Scholar Award 1054-09 (to AVM) and NIH Grant F31 AG033484-01 (to MJR).
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.