|Home | About | Journals | Submit | Contact Us | Français|
The Saccharomyces cerevisiae Swi2-like factors Rad54 and Rdh54 play multifaceted roles in homologous recombination via their DNA translocase activity. Aside from promoting Rad51-mediated DNA strand invasion of a partner chromatid, Rad54 and Rdh54 can remove Rad51 from duplex DNA for intracellular recycling. Although the in vitro properties of the two proteins are similar, differences between the phenotypes of the null allele mutants suggest that they play different roles in vivo. Through the isolation of a novel RAD51 allele encoding a protein with reduced affinity for DNA, we provide evidence that Rad54 and Rdh54 have different in vivo interactions with Rad51. The mutant Rad51 forms a complex on duplex DNA that is more susceptible to dissociation by Rdh54. This Rad51 variant distinguishes the in vivo functions of Rad54 and Rdh54, leading to the conclusion that two translocases remove Rad51 from different substrates in vivo. Additionally, we show that a third Swi2-like factor, Uls1, contributes toward Rad51 clearance from chromatin in the absence of Rad54 and Rdh54, and define a hierarchy of action of the Swi2-like translocases for chromosome damage repair.
Rad54 and Rdh54 are members of the Swi2 protein family. These evolutionarily conserved proteins possess dsDNA-dependent ATPase activity that fuels their translocation on dsDNA, resulting in DNA supercoiling and transient strand unwinding. Both proteins physically interact with the recombinase Rad51 and synergize with the Rad51–ssDNA nucleoprotein filament to promote D-loop formation, DNA branch migration and chromatin remodeling, all of which are essential steps in homologous recombination (HR) (1). Interestingly, Rad54 and Rdh54 both can remove Rad51 from dsDNA in vitro. The ability to dissociate the Rad51–dsDNA complex has been postulated to be important for releasing Rad51 from bulk chromatin, to ensure that a sufficient pool of free recombinase is available for repair and to prevent the accumulation of toxic Rad51–DNA intermediates. Moreover, removal of Rad51 by Rad54 and Rdh54 may be necessary to allow access of a DNA polymerase to the primer terminus in the newly made D-loop during HR. RAD54 and RDH54 likely serve distinct functions in mitotic and meiotic recombination, as mutants have distinct phenotypes (2–4). rad54Δ mutants are sensitive to DNA damaging agents and have significant reduction in mitotic recombination whereas rdh54Δ are only slightly sensitive to DNA damage and have a modest reduction in interchromosomal recombination, but are not affected in intrachromosomal recombination. rdh54Δ diploids have significant meiotic recombination defects and are delayed in the repair of meiotic double strand breaks whereas rad54Δ diploids does not show a delay in the repair of meiotic double strand breaks, although spore viability is reduced.
Even though both Rad54 and Rdh54 can dissociate Rad51 from dsDNA in vitro (5,6), whether these proteins remove Rad51 from chromatin in vivo and the functional significance and relative contributions of Rad54 and Rdh54 toward Rad51 clearance from chromatin remain unanswered. In vivo, RAD54 has a more important role than RDH54 in the recombinational repair of methyl methanesulfonate (MMS) damaged DNA and gaps that occur from replication across a damaged DNA template, as seen by the strong MMS sensitivity of rad54 mutants, while rdh54 mutants show only a modest sensitivity (3,4). In a recent study we have found that Rdh54 has a critical role in removing Rad51 from chromatin when Rad51 is expressed in excess, while Rad54 has only a minor role in this regard (7). In contrast, following irradiation damage, Rad51 foci persist in rad54 mutants, but not in rdh54 mutants (7). Likewise, we do not yet know whether Uls1, another Swi2 family member that was originally identified based on its two-hybrid interaction with the meiotic recombinase Dmc1, also plays a role in Rad51 clearance from chromatin (8). ULS1 has no known role in DNA repair, as deletion of the gene does not render cells sensitive to DNA damaging agents, although it appears to play a role in removing excess Rad51 from chromatin (7). Here, we show synthetic growth deficiency and DNA damage sensitivity of double and triple mutants of the aforementioned Swi2-like factors that can be efficiently suppressed by deleting RAD51. In congruence with this, a genomic suppressor of the rad54Δ uls1Δ mutant defects is shown to harbor a mutation, A265V, in RAD51. We provide evidence that the suppressor activity of rad51A265V stems from the combined effect of this mutation on the affinity of Rad51 for DNA and the accelerated removal of the mutant rad51 protein by Rdh54 from dsDNA.
Taken together, our results provide compelling evidence for a cytotoxic effect of gratuitous Rad51–dsDNA complexes and suggest that Rad54, Rdh54 and Uls1 all contribute toward clearance of these toxic nucleoprotein complexes in a hierarchical fashion. Additionally, our results suggest that Rad54 and Rdh54 recognize different Rad51–dsDNA complexes in vivo.
Yeast cultures were incubated overnight at 30°C in YPD medium. After determining cell density, the cultures were adjusted to 107 cells/ml and then serially diluted. Aliquots of 4µl from the serial dilutions were spotted onto SC or SC-containing MMS at the indicated concentration. SC plates containing MMS were made directly before use. The plates were then incubated at 30°C for 5–6 days.
The rad54Δ uls1Δ strain used for EMS mutagenesis is HKY1287-11B MATa rad54::LEU2 uls1::KANMX leu2-3, 112 his3-11, 15 ade2-1 ura3-1 trp1-1 can1-100 hom3-10 RAD5. For EMS mutagenesis, cells were grown overnight at 30°C, collected and washed twice with water and resuspended in an equal volume of 0.1M sodium phosphate buffer (pH 7). The exact cell density was determined with a hemacytometer and adjusted to 2×108 cells/ml. Two 1-ml aliquots were made, and 0.5µl EMS (Sigma) was added to one aliquot, while the other aliquot served as control. The tubes were vortexed vigorously before incubating for 1h at 30°C with agitation. After incubation, the cells were collected and washed three times with 8ml 5% sodium thiosulfate, and once with sterile distilled water. The cells were then resuspended in 5ml YPD, and incubated for 3h to allow cells to express the mutant proteins. After outgrowth, 106 EMS-treated and control cells were plated onto YPD plates containing 0.004% MMS. In addition, 104, 102 and 101 EMS-treated and control cells were plated onto YPD to estimate the percentage cell death resulting from the EMS mutagenesis protocol. Colonies growing on YPD were counted after 36h of incubation at 30°C, and percent survival was calculated to be ~25%.
Colonies growing on 0.004% MMS after 7 days of incubation at 30°C were streaked onto fresh 0.004% MMS plates to yield single colonies and compared to growth of rad54Δ, uls1Δ and rad54Δ uls1Δ strains. Mutants that were less sensitive to 0.004% MMS than the rad54Δ uls1Δ double mutant were screened for growth in 0.0025% MMS to confirm resistance to levels of MMS at which rad54Δ uls1Δ is not viable. Colonies with significant growth in 0.0025% MMS were chosen as putative suppressors of rad54Δ uls1Δ MMS sensitivity. Suppressor-containing strains were crossed several times to a wild-type strain to recover the suppressor mutation in a RAD54 ULS1 background and to eliminate any unlinked mutations that might have arisen during the mutagenesis.
Primers 5′CATATCCCACGACTAGGCCA3′ and 5′CATGGGTGACAGACAATACG3′ were used to amplify the RAD51 gene from yeast strains containing putative suppressors of rad54Δ uls1Δ. The PCR product was sequenced and a mutation at base 794 changing C to T or amino acid residue 265 changing alanine to valine was found in RAD51.
The rad51 A265V mutant allele was introduced into an unmutagenized wild-type strain to replace the endogenous RAD51 allele. A 1.7-kb DNA fragment including the rad51 A265V open reading frame was amplified using primers SSM93 (5′ CGGGGTACCCGGGGATCCCACGACTAGGCCACAC) and SSM94 (5′ GCCAAGCTTGCATGCCTGCAGGAGGAAGTAGTCATC) from genomic DNA of a mutant yeast strain. The PCR product was then introduced into vector YIplac211 as BamHI–PstI fragment to yield the integration construct pHK440. pHK440 was then cut with the SpeI restriction endonuclease and used to transform yeast cells to Ura+. Positive transformants were then grown on 5-FOA-containing medium and the resulting clones confirmed for the rad51 A265V mutation by DNA sequencing.
Recombination assays for haploid intragenic recombination and diploid intragenic heteroallelic recombination were performed as previously described (3). Recombination rates were determined by the method of the median (9).
Rad51 and rad51 A265V proteins were expressed in the rad51Δ strain LSY411 by the use of the PGK promoter in plasmid pMA91 (2µ, PGK, leu-2d) (10) and were purified to near homogeneity, following our published protocol (11). His6-tagged Rad54 or Rdh54 was expressed in Escherichia coli and purified to near homogeneity as described (5,12).
The ϕX replicative form I DNA was purchased from Gibco/BRL. Topologically relaxed ϕX DNA was prepared by treatment with calf thymus topoisomerase I (Invitrogen), as described previously (13). For DNA mobility shift assay, the 80-mer Oligo 1–5′ TTATATCCTTTACTTTGAATTCTATGTTTAACCTTTTACTTATTTTGTATTAGCCGGATCCTTATTTCAATTATGTTCAT-3′ was 5′-end labeled with [γ-32P] ATP (Amersham Bioscience) and polynucleotide kinase (Roche) and separated from the unincorporated ATP using a Spin 6 column (Bio-Rad). To generate radiolabeled dsDNA, radiolabeled Oligo 1 was annealed to its exact complement, Oligo 2. The resulting duplex was purified from a 10% polyacrylamide gel by overnight diffusion at 4°C into TE buffer (20mM Tris–HCl, pH 7.5, 0.5mM EDTA).
For the D-loop assay, the 90-mer oligonucleotide D1 (14), being complementary to pBluescript SK DNA from position 1932 to 2022, was 5′-end labeled and then purified using the MERmaid Spin Kit (Bio101). The 600 base pairs biotinylated dsDNA used for monitoring Rad51 removal was prepared by PCR amplification of pBluscript SK DNA using the 5′-biotinylated primer 1 and non-biotinylated primer 2, as described previously (5). The amplified DNA was deproteinized by phenol–chloroform extraction, ethanol precipitated and dissolved in TE. The biotinylated dsDNA was immobilized on streptavidin magnetic beads (Roche Molecular Biochemicals), as described previously, to give 50ng of the biotinylated DNA per microliter of suspended volume (5). The 83-mer 5′- TTTATATCCTTTACTTTATTTTCTATGTTTATTCATTTACTTATTTTGTATTATCCTTATACTTATTTACTTTATGTTCATTT-3′ ssDNA was used as the Rad51 trap.
The 32P-labeled 80-mer ssDNA (4.5µM nucleotides) or dsDNA (4.5µM base pairs) was incubated with the indicated amounts of Rad51 or rad51A265V for 5min at 37°C in 10µl of buffer B (35mM Tris–HCl at pH 7.5, 1mM DTT, 50mM KCl, 100µg/ml BSA) containing 2mM ATP and the indicated concentration of MgCl2. The reaction mixtures were resolved in 10% polyacrylamide gels in TA buffer (40mM Tris–acetate, pH 7.5) at 4°C. The gels were dried onto a sheet of DEAE paper and then subject to phosphorimaging analysis.
Topologically relaxed ϕX dsDNA (10µM based pairs) was incubated with the indicated amount of Rad51 or rad51A265V in 9.6µl buffer B containing 2mM ATP and the indicated concentration of MgCl2 for 5min, followed by the addition of 3 units of calf thymus topoisomerase I (Invitrogen) in 0.4µl. Reaction mixtures were incubated for 15min at 37°C and then deproteinized with 0.5% SDS and 0.5mg/ml proteinase K for 10min. Samples were resolved in 0.8% agarose gels run in TAE buffer (40mM Tris–acetate, pH 7.5, 0.5mM EDTA) at 23°C, and the DNA species were stained with ethidium bromide (2µg/ml in water) for 1h. After being destained in water at 4°C for 24h, the gels were analyzed in a gel documentation station (Bio-Rad).
The 32P-labeled 90-mer ssDNA (2.4µM nucleotides) was incubated with the indicated amounts of Rad51 or rad51 A265V in 10.5µl buffer B containing an ATP-regenerating system and 5mM MgCl2 for 5min at 37°C. Rad54 (140nM) or Rdh54 (300nM) was then added in 1µl, followed by a 1-min incubation at 30°C. The D-loop reaction was initiated by adding pBluescript replicative form I DNA (35µM base pairs) in 1µl. The reaction mixtures were incubated for 5min at 30°C, deproteinized and processed for electrophoresis in 0.9% agarose gels in TAE buffer (40mM Tris–acetate, pH 7.5 0.5mM EDTA). The gels were dried onto a sheet of DEAE paper and the radiolabeled D-loop was visualized and quantified in the phosphorimager.
Rad51 or rad51 A265V (4µM each) was incubated with 100µM ATP, 0.1µCi/µl [γ-32P] ATP at 37°C in a 10µl buffer containing 40mM Tris–HCl pH 7.5, 1mM DTT, 100ng/µl BSA, 50mM KCl, 2 or 5mM MgCl2 (as indicated) in the presence of ϕX 174 viral (+) strand (30µM nucleotides) or replicative form I DNA (30µM base pairs) or in the absence of DNA. Aliquots (2µl) were drawn at the indicated times and mixed with an equal volume of 500mM EDTA to stop the reaction. After thin layer chromatography in polyethyleneimine sheets (J.T. Baker Inc.), the level of ATP hydrolysis was determined by phosphorimaging analysis of the chromatography plate (14).
Rad51 or rad51 A265V (3.7µM) was incubated with magnetic beads containing biotinylated dsDNA (15µM base pairs) in 18µl buffer B containing 2mM ATP, 5mM MgCl2 and an ATP-regeneration system for 5min at 37°C. After the incorporation of the indicated amounts of Rad54 or Rdh54 in 1µl and a 3-min incubation at 23°C, the reaction was completed by adding 83-mer ssDNA (150µM nucleotides), as Rad51 trap, in 1µl. Following a 10-min incubation at 30°C, the beads were captured with the Magnetic Particle Separator (Boehringer Mannheim), and the supernatant was set aside. Bound proteins were eluted from the beads with 20µl of 2% SDS. The various supernatant and SDS eluate (8µl each) were analyzed by SDS–PAGE and Coomassie blue staining to determine their content of proteins.
Rad51 or rad51 A265V (5µg each) was incubated with His6-tagged Rad54 or Rdh54 (5µg each) in 30µl buffer D (20mM KH2PO4, pH 7.4, 75mM KCl, 10% glycerol, 0.5mM EDTA, 0.01% Igepal, 1mM 2-mercaptoethanol) for 30min on ice. The reaction mixture was incubated with 8µl of Ni2+-NTA agarose beads for 30min on ice, with gentle mixing every 30s. The beads were pelleted by centrifugation, and the supernatant was removed. After being washed twice with 30µl buffer D containing 10mM imidazole, the beads were treated with 20µl of 2% SDS to elute bound proteins. The supernatant (8µl), wash (8µl) and SDS eluate (8µl) were subject to SDS–PAGE and Coomassie blue staining to determine their protein contents.
A Sephacryl S-400 column (30×1cm) was used to analyze Rad51 and rad51 A265V with buffer C (25mM Tris, pH 7.5, 10% glycerol, 0.5mM EDTA, 1mM DTT, 0.01% Igepal and 150mM KCl) as the eluent and collecting 0.2-ml fractions. The indicated column fractions were subjected to SDS–PAGE and Coomassie blue staining. Thyroglobulin (669kDa), and catalase (232kDa) were used for calibrating the column.
Flowcells and DNA curtains were prepared essentially as described (15). The DNA substrate was a 23-kb segment of the human β-globin locus that was amplified with the Expand 20kBPLUS PCR system from human genomic DNA using forward and reverse primers that are covalently linked to biotin and digoxigenin, as described elsewhere (16). The DNA was labeled with quantum dots covalently conjugated to anti-digoxygenin Fab-fragments (Roche). Rad51 (700µL of 1µM, either wild-type or A265V mutant) was injected at 37°C in buffer containing 1mM ATP and either 10 or 2mM of MgCl2, as indicated, and reactions were chased with buffer lacking Rad51. Images were recorded for 15min at 1-s intervals. The DNA length was measured by tracking the location of the quantum dot (17). The data corresponding to filament disassembly are fit with a sigmoidal curve the slope of which is used to determine the disassembly rate, and all reported rates represent the mean obtained from three independent experiments, as described previously (16).
The ULS1 gene was identified in a two-hybrid screen for yeast proteins that interact with Dmc1, a meiosis-specific Rad51 homolog (8). Uls1 protein (also known as Ris1 or Tid4) is a member of the Swi2 family. Interestingly, the Uls1 protein possesses a RING finger domain suggestive of an ubiquitin ligase activity (18). The most closely related Saccharomyces cerevisiae proteins are Rad5 and Rad16, which are involved in post-replication repair (PRR) and nucleotide excision repair (NER), respectively.
Mutations in the homologous recombination factors Rad54 and Rdh54, which are also Swi2 family members, result in DNA damage sensitivity and reduced homologous recombination, although the rad54 mutant is much more affected than the rdh54 mutant. The rad54Δ rdh54Δ haploid double mutant grows at the same rate as the haploid rad54Δ strain, but diploid growth of rad54Δ rdh54Δ is greatly impaired, with clonal lethal sectors (3) (Figure 1A). Deletion of RAD51 in the rad54Δ rdh54Δ diploids restored normal growth (3) (Figure 1A), suggesting that Rad51 may form toxic intermediates in these diploids. To determine if additional SWI2 genes could be involved in avoidance of the Rad51 toxic intermediates, we focused on ULS1, based on its reported interaction with the Rad51-related recombinase protein Dmc1 (8). We found that a homozygous mutation of ULS1 in the rad54Δ rdh54Δ diploids further aggravates growth deficiency (Figure 1A), while the haploid triple mutant strain rad54Δ rdh54Δ uls1Δ grows slower than single or double mutants, but without clonal lethality (7). As expected, the poor diploid growth of the triple mutant can be completely overcome by deleting the RAD51 gene. Importantly, these results implicate Uls1 in the regulation of HR, possibly via the clearance of Rad51 from chromatin (see below). Indeed, we have found that rad54Δ rdh54Δ uls1Δ strains spontaneously accumulate Rad51 foci (7).
To further explore a possible overlap in function between Uls1 and either Rad54 or Rdh54, haploid double mutants rad54Δ uls1Δ and rdh54Δ uls1Δ were examined for sensitivity to DNA damage. uls1Δ did not show a significant increase in DNA damage sensitivity as a single mutant, nor did it enhance the DNA damage sensitivity of a rdh54Δ mutant. However, uls1Δ did slightly elevate the MMS sensitivity of rad54Δ at low MMS doses, providing further support for the idea that Uls1 can partially substitute for Rad54 in DNA damage repair and recombination (Figure 1B).
The above observations are reminiscent of the synthetic lethality that results from combining mutations in genes whose products act to remove Rad51 from DNA combined with other DNA repair factors. The Srs2 DNA helicase can disrupt Rad51 filaments on ssDNA, and mutations in this gene show synthetic lethality with a variety of other mutations, which arise because of the accumulation of toxic recombination intermediates, likely filaments of Rad51 on ssDNA (19). Since Rad54 and Rdh54 remove Rad51 from dsDNA (5,6), we suspected that the poor growth and DNA damage sensitivity of various double and triple mutants, seen in Figure 1, were the result of toxic intermediates generated by persistent association of Rad51 with DNA. We validated this hypothesis via the isolation and characterization of a suppressing mutation in RAD51.
The enhanced MMS sensitivity of the haploid rad54Δ uls1Δ double mutant at low MMS doses provided a means for isolating recessive suppressors, which we reasoned, could be hypomorphic loss-of-function alleles in HR factors. Complete loss-of-function mutations should not rescue the DNA damage sensitivity phenotype, as the HR factors are needed for repairing chromosome damage induced by MMS.
Following EMS mutagenesis of the haploid rad54Δ uls1Δ strain, variants of the reduced MMS sensitivity were sought. One such variant displayed enhanced MMS resistance that was retained in appropriate spore segregants from genetic crosses. Since the suppressor strain had a slight MMS sensitivity in an otherwise wild-type background, we reasoned that it harbors a hypomorphic mutation in a HR gene, possibly RAD51. Indeed, sequencing of the chromosomal RAD51 locus in the suppressor strain revealed the alteration of alanine 265 to valine (change of nucleotide C at position 794 to T). This mutation lies near the subunit interface between adjacent monomers in the Rad51 filament (Figure 2) (20).
To further study the rad51A265V mutation, we replaced the wild-type allele with this mutant version in an unmutagenized strain, and used this strain in crosses to generate all of the strain combinations with rad51A265V shown in Figure 1A, B and D. Importantly, the rad51A265V mutation overcomes the growth deficiency of the rad54Δ rdh54Δ uls1Δ diploid (Figure 1A). This suppression is semidominant, as suppression of the growth deficiency of the rad54Δ rdh54Δ uls1Δ diploid is seen in RAD51/rad51A265V, but the suppression is stronger when rad51A265V is homozygous. Thus, the rad51A265V mutation alleviates the DNA damage sensitivity of the rad54Δ uls1Δ double mutant and the growth deficiency of the rad54Δ rdh54Δ uls1Δ triple mutant. At low MMS doses, 0.0025 and 0.005%, the rad51A265V mutation is epistatic to the rad54Δ mutation and suppresses the rad54Δ mutation. At higher MMS doses of 0.01 and 0.015%, the rad51A265V mutation does not suppress the rad54Δ mutation (data not shown).
Genetic assays revealed that the rad51A265V strain is reduced in mitotic intrachromosomal gene conversion compared to wild type, although it is still 10-fold higher than the rad51Δ rate (Figure 1C). Mitotic interchromosomal gene conversion recombination between homologous chromosomes is not significantly reduced in the rad51A265V mutant, whereas it decreases >100-fold in the rad51Δ mutant. However, the rad51A265V mutation does not suppress the recombination gene conversion defect of rad54Δ strains, showing that the rad51A265V mutation does not bypass the need for the Rad54 protein in homologous recombination (Supplementary Figure S1).
uls1Δ and rdh54Δ show no genetic interaction in a RAD51 background (Figure 1D). However, the uls1Δ rad51A265V double mutant has an enhanced MMS sensitive phenotype. One explanation for this is that Rad54 and Rdh54 compete for the rad51 A265V protein and may prematurely remove the rad51 A265V protein from damage repair sites, thus leading to increased damage sensitivity. To test this, the triple mutant rdh54Δ uls1Δ rad51 A265V was examined. Loss of RDH54 restored MMS resistance to the uls1Δ rad51A265V strain. This is consistent with a model in which Rdh54 acts inappropriately on Rad51 dsDNA nucleofilaments to remove Rad51, preventing Rad54 from performing its function in DNA repair. To provide further support for this model, the biochemical properties of the rad51 A265V protein were studied.
To determine if rad51A265V encodes a hypomorphic mutant protein that either forms unstable filaments on DNA (hence its suppression of the poor diploid growth of the rad54Δ rdh54Δ uls1Δ triple mutant) and/or is more readily removed from dsDNA by Rdh54 (hence its suppression of the DNA damage sensitivity of the rad54Δ uls1Δ mutant), we expressed the rad51 A265V mutant protein in yeast cells and purified it (see ‘Materials and Methods’ section) to near homogeneity (Supplementary Figure S2A) for biochemical characterization (see below). During purification, the rad51 A265V protein exhibited the same chromatographic properties as the wild-type Rad51 protein, and a yield of the mutant protein very similar to that of the wild-type counterpart was obtained.
Even though the A265V mutation lies near the subunit interface between adjacent monomers in the Rad51 filament (Figure 2), gel filtration analysis showed that the rad51 A265V mutant protein has the same oligomeric structure as wild-type Rad51 in the absence of DNA and ATP (Supplementary Figure S2B). We next tested the purified rad51 A265V mutant alongside wild-type Rad51 for DNA binding, using radiolabeled ssDNA and dsDNA as substrates and mobility shift of these substrates in polyacrylamide gels as assay. Since Rad51 needs ATP to bind DNA, the experiments were conducted in the presence of 2mM ATP with either 2 or 5mM MgCl2. As shown in Figure 3A, at 2mM MgCl2, rad51 A265V is less capable than the wild-type protein in binding ssDNA and dsDNA. This DNA binding deficiency is alleviated to a significant degree upon increasing the MgCl2 concentration to 5mM (Figure 3A).
Duplex DNA becomes extended (by ~50% relative to B form DNA) when bound by Rad51 (1). The use of calf thymus topoisomerase I allows one to register the DNA extension as a topological change; the product of this topoiosmerase-linked reaction is a negatively supercoiled species, called Form UW (Supplementary Figure S3A). In congruence with the results from the DNA mobility shift assay, rad51 A265V was significantly less effective than Rad51 in generating Form UW DNA at 2mM MgCl2 (Supplementary Figure S3B), whereas a much less pronounced deficit was seen upon increasing the MgCl2 concentration to 5mM (Supplementary Figure S3C).
Collectively, the results from the DNA mobility shift and topoisomerase-linked assays (Figure 3A and Supplementary Figure S3), revealed a DNA binding deficiency in rad51 A265V. Additional evidence, presented below, further indicated that rad51 A265V dissociates at a faster rate from dsDNA.
In fulfilling its biological role, Rad51 needs to physically and functionally interact with several other factors, including Rad54 and Rdh54. We tested the rad51 A265V mutant for physical interaction with Rad54 and Rdh54 in a pulldown assay that made use of the (His)6 tag on the latter two proteins and nickel NTA agarose beads to capture protein complexes. As shown in Figure 2B, rad51 A265V was just as proficient in interacting with Rad54 or Rdh54 as wild-type Rad51. We further asked whether rad51 A265V retains the ability to functionally synergize with Rad54 or Rdh54 in the D-loop reaction. To do this, a 32P-labeled 90-mer ssDNA was incubated with Rad51 or rad51 A265V before adding Rad54 or Rdh54, followed by the incorporation of the negatively supercoiled homologous duplex target to complete the reaction. Pairing of the ssDNA and homologous duplex yields a 32P-labeled D-loop. Rad54 or Rdh54 greatly enhanced the ability of Rad51 or rad51 A265V to catalyze D-loop formation (Figure 3C).
The above results demonstrated that the A265V mutation has little or no negative impact on Rad51′s interaction with Rad54 or Rdh54 or on its ability to functionally synergize with these Swi2-like factors in catalyzing D-loop formation.
To ask whether the A265V mutation affects the stability of the Rad51–dsDNA filaments, total internal reflection fluorescence microscopy (TIRFM) was used to monitor the dissociation of the rad51 A265V mutant protein from dsDNA (Figure 4A) (16). To do this, nucleoprotein filaments of either Rad51 or rad51 A265V were assembled on single molecules of dsDNA within the context of a DNA curtain (17). As expected, Rad51 and rad51 A265V both extended the length of the DNA by ~68% upon nucleoprotein filament assembly (16,17), and, importantly, rad51 A265V extended the DNA to the same degree as the wild-type protein. This greater than expected increase in the apparent length of the DNA can be attributed to the behavior of the naked DNA versus the nucleoprotein complexes in shear flow; there is an increase in persistence length as Rad51 binds the DNA (21–23), which makes the Rad51-bound DNA easier to stretch at a given flow rate (16,17). These results are in complete agreement with the topological assays presented above (Supplementary Figure S3) confirming that rad51 A265V forms an elongated nucleoprotein filament on dsDNA. The presence of a fluorescent quantum dot at the end of the DNA molecule allowed us to accurately measure the DNA length and quantitate the filament dissociation rate using an automated particle-tracking algorithm, as previously described (24). As shown in Figure 4, the TIRFM analysis provided evidence for a reduced stability of the rad51 A265V nucleoprotein filaments compared to Rad51 filaments. Specifically, at 10mM MgCl2 wild-type Rad51 filaments disassembled at a rate of 5.6±0.8nm/s, whereas a more rapid disassembly rate of 7.0±1.1nm/s was seen for rad51 A265V under the same buffer conditions. At 2mM MgCl2, the difference in stability was even more pronounced, with an observed value of 5.8±1.1nm/s for wild type Rad51 compared to 9.9±1.1nm/s for the mutant protein. These results thus demonstrate that the rad51 A265V mutant is more prone to spontaneously dissociating from dsDNA than wild-type Rad51. Since the turnover of Rad51 from DNA is coupled to ATP hydrolysis, we also tested whether the rad51 A265V mutant has an enhanced ability to hydrolyze ATP. However, results from ATPase assays done in the absence or presence of DNA revealed that the rad51 mutant in fact possesses a slightly attenuated ATPase activity (Supplementary Figure S4).
The biochemical experiments described earlier (Figure 3) have provided evidence that rad51 A265V has a lower affinity for DNA, which, as indicated from TIRFM experiments, likely stems from a heightened tendency of the mutant protein to dissociate from DNA (Figure 4). Previous studies have found an ability of Rad54 and Rdh54 to remove Rad51 from dsDNA (5,6) and, as noted above, our genetic data have hinted at the possibility that rad51 A265V may be removed from dsDNA at a faster rate by Rdh54. To directly test this premise, we used a biochemical assay, devised previously (5) (Figure 5A), to monitor the dissociation of the Rad51–dsDNA nucleoprotein filament. Briefly, filaments of Rad51 or rad51 A265V were assembled on a biotinylated dsDNA fragment fixed to streptavidin magnetic beads, followed by the addition of Rad54 or Rdh54 together with a non-biotinylated ssDNA molecule to trap the Rad51 molecules that have been dislodged from dsDNA by either of the latter two proteins. As shown before (5,6) and reiterated here, there was a Rad54 or Rdh54 concentration-dependent transfer of Rad51 protein from the magnetic bead-bound DNA to the non-biotinylated ssDNA trap, indicative of dissociation of the Rad51–dsDNA nucleoprotein filament by these Swi2-like factors. Interestingly, while rad51 A265V was removed by Rad54 from the dsDNA at about the same rate as wild-type Rad51 (Figure 5B), a significantly accelerated rate of rad51 A265V dissociation was seen with Rdh54 (Figure 5C). In particular, at the lowest concentration (70nM) of Rdh54 used, there was a nearly 3-fold difference in susceptibility of the mutant rad51 and wild-type Rad51 filaments to the dissociative action of the former protein (Figure 5C). Taken together, these results suggest that compared to wild-type Rad51 protein, the mutant rad51 A265V protein is more prone to spontaneous dissociation from dsDNA and is also more readily removed from dsDNA by Rdh54.
With the use of S. cerevisiae mutants ablated for the Swi2-related factors Rad54, Rdh54 and Uls1, we have furnished additional evidence that the inability to properly regulate Rad51 nucleoprotein filaments can adversely affect the fitness of mitotic cells. Specifically, haploid and diploid strains deficient in these Swi2-like proteins are sensitive to the DNA damaging agent MMS or are growth impaired, phenotypes that can be overcome via RAD51 deletion. That the toxicity of Rad51 in the mutant cells stems from gratuitous Rad51–dsDNA complexes is supported by our isolation of the rad51A265V suppressor allele and the biochemical demonstration of a reduced stability of the rad51 A265V–dsDNA filament and its enhanced susceptibility to clearance by Rdh54.
Rad51–dsDNA complexes can form either during DSB repair or can arise from nonspecific binding of Rad51 to chromatin. During DSB repair, Rad51–dsDNA complexes are created after DNA strand exchange where the invading single strand coated with Rad51 protein pairs and forms a heteroduplex joint with the homologous duplex partner chromatid. Rad51 removal likely allows DNA polymerases access to the primer terminus for completion of the repair reaction (1,2). Similar to the meiotic recombinase Dmc1, Rad51 accumulates on chromatin and requires translocases to promote its removal from dsDNA, ensuring there is a pool of free protein available when required for DSB repair (7,25). Unlike the bacterial RecA protein, the yeast Rad51 protein displays little binding preference for ssDNA over dsDNA, arguing that recycling mechanisms must come into play if the protein is to be targeted to the correct locations during DSB repair (26–28).
Even though Rad54 and Rdh54 can both remove Rad51 from dsDNA (5,6), unlike Rdh54, Rad54 is no more capable of dissociating rad51 A265V from dsDNA than does wild-type Rad51. Whether this distinction owes to different epitopes on Rad51 being recognized by Rad54 and Rdh54, or to another reason, remains to be delineated, although we suggest that Rad51 bound to dsDNA at a homologous recombination repair intermediate may differ from Rad51 bound to chromatin (7). Regardless, the differential sensitivities of the rad51 A265V–dsDNA filament to Rad54 and Rdh54 provide evidence for important mechanistic differences in the way these Swi2-like factors act to prevent the non-specific association of Rad51 with bulk chromatin, which is further supported by studies of translocase mutant sensitivity to Rad51 overexpression (7). The damage sensitivity of the uls1 rad51A265V mutant may reflect an interference of DNA repair by Rdh54 stemming from the accelerated removal of rad51 A265V mutant protein from a recombination intermediate by this motor protein. An alternative explanation for suppression phenotypes observed in this study may be focused on the action of Rad54 after the formation of a recombination intermediate. Rad54 can promote dissociation of D-loops (29), and this function may be easier to perform with the rad51 A265V protein. In this manner, the rad51A265V mutation would prevent the accumulation of recombination intermediates.
We note that genetic studies have revealed other important differences between RAD54 and RDH54 as well. Specifically, RAD54 is more important for DSB repair, DNA extension after strand invasion (30) and haploid recombination than RDH54, while RDH54 makes a more prominent contribution to inter-homologue recombination than to intra-homologue recombination. This suggests that for mitotic DSB repair and recombination, RAD54 is the frontline translocase that is used in Rad51 removal from dsDNA. The observation that rad54 rdh54 diploids have a growth defect and are increased in DNA damage sensitivity shows that in diploid mitotic situations, RDH54 has a role in Rad51 removal, and suggests that it is secondary to RAD54 for damage repair. In meiosis, RAD54 and RDH54 also seem to fulfill distinct roles; Rdh54 interacts with Dmc1 and dissociates Dmc1 from bulk chromatin while Rad54′s action appears to be specific for Rad51 (25,31).
In the initial step of the recombination reaction, Rad51 polymerizes onto ssDNA to form the presynaptic filament. Biochemical and genetic studies have provided evidence that the Srs2 helicase regulates recombination outcome and DNA checkpoint signaling by dismantling the Rad51 presynaptic filament (19). Although we believe that the altered rad51A265V binding to dsDNA is responsible for the suppression phenotypes reported in this article, we found that rad51 A265V also has reduced binding to ssDNA. Consistent with this altered property, we have observed that the synthetic sickness/lethality of srs2Δ sgs1Δ (32) and srs2Δ rad54Δ (32) haploids is suppressed by the rad51A265V mutation (unpublished data). That Rad51 binding to ssDNA may cause some of the rad54 rdh54 uls1 growth problems remains a possibility, although we note neither Rad54 nor Rdh54 can remove Rad51 from ssDNA. However, it is possible that they might remove Rad51 from secondary structures in ssDNA that resemble dsDNA, and this problem is alleviated by the rad51 A265V mutant.
That diploid growth becomes severely affected upon introducing the uls1Δ deletion into rad54Δ rdh54Δ mutant cells suggests that Uls1 protein can function to minimize the toxic accumulation of Rad51 on dsDNA in the absence of Rad54 and Rdh54. Uls1 possesses both Swi2-like DNA translocase motifs and a RING finger domain suggestive of a ubiquitin ligase activity. Whether these activities are germane for Rad51′s removal from dsDNA represent important issues to address in future, although preliminary experiments with the ATPase null show that this mutant version behaves similar to the complete deletion mutant in vivo.
Additional rad51 mutant alleles have been isolated as suppressors of other genetic deficiencies. Some of these rad51 mutations affect DNA binding or at least lie in regions of Rad51 thought to be involved in DNA binding (33–35). It may be informative to examine these rad51 mutant proteins for their susceptibility to removal from dsDNA by Rad54 and Rdh54. It remains possible that the rad51 mutants that have been recovered from screens for suppressors for DNA damage sensitivity phenotypes and appear to have reduced DNA binding capacity really allow secondary translocases to remove Rad51 protein from DNA and thereby suppress DNA damage sensitivity in the absence of a major factor for DNA damage repair. Such mutant rad51 proteins would be expected to support some degree of DNA repair and recombination, and indeed rad51 A265V fulfills all of these expectations.
Given the high degree of evolutionary conservation of recombination genes (36), our findings could provide the requisite framework for dissecting the role of Rad54 and related proteins, such as Rad54B, in preventing the accumulation of toxic Rad51–dsDNA complexes in human cells. The human RAD54 gene has been found mutated in tumors (37), and it will be interesting to determine whether any of the tumor-associated hRad54 mutations affect the ability of this protein to remove Rad51 from dsDNA. Additionally, the closest human homolog to the yeast Uls1 protein is SMARCA3/HTLF. Gene silencing of SMARCA3/HTLF has been linked to gastric tumors (38–40), but any connection between hRad54 and SMARCA3/HTLF has not been explored yet. A synergistic enhancement of hRad54 mutations by a SMARCA3/HTLF deletion or knockdown would be indicative of some overlap in function.
Supplementary Data are available at NAR Online.
National Institutes of Health (ES007061 to P.S., GM053738 to H.L.K., GM074739 to E.C.G., GM057814 to P.S. and CA146940 to H.L.K, E.C.G. and P.S.). Funding for open access charge: NIH GM053738.
Conflict of interest statement. None declared.
The authors thank Lorraine Symington for the gift of plasmids, and the members of the Klein, Sung and Greene laboratories for helpful discussions.