Yeast Rad54 Promotes Rad51-dependent Homologous DNA Pairing via ATP Hydrolysis-driven Change in DNA Double Helix Conformation*

Saccharomyces cerevisiae RAD54 gene functions in the formation of heteroduplex DNA, a key intermediate in recombination processes. Rad54 is monomeric in solution, but forms a dimer/oligomer on DNA. Rad54 dimer/oligomer alters the conformation of the DNA double helix in an ATP-dependent manner, as revealed by a change in the DNA linking number in a topoisomerase I-linked reaction. DNA conformational alteration does not occur in the presence of non-hydrolyzable ATP analogues, nor when mutant rad54 proteins defective in ATP hydrolysis replace Rad54. Accordingly, the Rad54 ATPase activity is shown to be required for biological functionin vivo and for promoting Rad51-mediated homologous DNA pairing in vitro. Taken together, the results are consistent with a model in which a Rad54 dimer/oligomer promotes nascent heteroduplex joint formation via a specific interaction with Rad51 protein and an ability to transiently unwind duplex DNA.

Saccharomyces cerevisiae genes of the RAD52 epistasis group, viz, RAD50, RAD51, RAD52, RAD54, RAD55, RAD57, RAD59, RDH54/TID1, MRE11, and XRS2, are required for genetic recombination and DNA double-strand break repair by recombination. Since genetic recombination is indispensable for the disjunction of homologous chromosomal pairs during meiosis I, mutational inactivation of the RAD52 group genes engenders severe meiotic defects, manifest as a failure to sporulate and spore inviability (1)(2)(3)(4)(5). The results from recent cloning studies have revealed a remarkable degree of conservation of the RAD52 group genes among eukaryotes, from yeast to humans.
A conceptual model concerning the mechanism of homologous recombination has been formulated based on genetic studies in S. cerevisiae (6). When S. cerevisiae cells enter meiosis, DNA double-strand breaks are formed at various chromosomal "hot spots" that exhibit a propensity to recombine. Subsequent to break formation, unidirectional nucleolytic end-processing of the break yields 3Ј ssDNA 1 tails of a considerable length (7,8).
It is believed that nucleation of various recombination factors onto the ssDNA tails gives rise to a nucleoprotein complex that has the ability to conduct a search to locate a DNA homolog and to invade the homolog to form heteroduplex DNA. Concurrent and subsequent events include DNA synthesis to replace the genetic information eliminated during double-strand break processing, resolution of the DNA intermediates, and DNA ligation to complete the recombination process. The repair of DNA double-strand breaks induced by ionizing radiation and radiomimetic chemicals very likely proceeds through the same mechanistic route, as the repair process shares the same requirement for the RAD52 epistasis group genes.
Extensive genetic evidence has indicated that the nucleolytic end-processing of DNA double-strand breaks during recombination processes is dependent on the RAD50, MRE11, and XRS2 genes. The Mre11 protein from both yeast (9 -11) and humans (12) and a protein complex (13) consisting of the human Rad50, Mre11, and the Xrs2 equivalent NBS1, product of the gene mutated in Nijmegen breakage syndrome (14,15), have ssDNA endonuclease and dsDNA specific 3Ј to 5Ј exonuclease activities. It has been suggested that the Mre11-associated nuclease complex functions with a DNA helicase to create the 3Ј ssDNA tails known to form during recombination processes (9 -11, 13).
Once the 3Ј single-strand (ss) tail is generated by the endprocessing reaction, it is believed that a number of recombination factors, including Rad51, Rad52, Rad54, Rad55, Rad57, Rdh54, the ssDNA binding factor RPA, and possibly Rad59, nucleate onto this ssDNA tail to form a nucleoprotein complex, which initiates a search to locate a DNA homolog. Invasion of the homolog by the nucleoprotein complex then leads to the formation of heteroduplex DNA. Rad51 protein is the eukaryotic equivalent of Escherichia coli RecA protein, which is central to recombination processes by virtue of its ability to mediate the homologous DNA pairing and strand exchange reaction that yields heteroduplex DNA (reviewed in Refs. 16 and 17). Evidence emerging during the past few years has indicated that Rad51 protein carries out the homologous DNA pairing and strand exchange reaction (18,19). Biochemical studies have revealed that Rad51 protein polymerizes on ssDNA to form a nucleoprotein filament in which the DNA is highly extended (20,21). The assembly of the Rad51-ssDNA nucleoprotein filament requires ATP binding and is stimulated by RPA (18,19,22). Rad52 protein (23)(24)(25) and a heterodimer of Rad55 and Rad57 proteins (26) have been shown to function as molecular mediators (25,27), enhancing the efficiency of Rad51-ssDNA nucleoprotein filament assembly when there is a necessity for Rad51 to compete with RPA for binding sites on the ssDNA. In its role as molecular mediators between Rad51 and RPA, Rad52 protein and the Rad55-Rad57 heterodimer resemble the complex of E. coli RecO and RecR proteins (28) and the bacteriophage T4 UvsY protein (29 -32), which function to enhance the nucleation of their cognate recombinases RecA and UvsX onto the ssDNA when the DNA substrate is coated with a single-strand DNA-binding protein.
The level of homologous DNA pairing and strand exchange that can be achieved by Rad51 protein, even under optimized reaction conditions (21) and in the presence of various ancillary factors including Rad52 protein and the Rad55-Rad57 heterodimer (23)(24)(25)(26), is still lower than that seen with its prokaryotic analogs RecA and UvsX proteins. These observations have suggested that perhaps another protein factor(s) functions to augment the homologous DNA pairing and strand exchange activities of Rad51 protein to achieve the desirable level of efficiency of heteroduplex DNA formation during recombination processes in vivo.
Among the RAD52 group proteins required for heteroduplex DNA formation, Rad54 is of particular interest, as it is the only member of this class for which there does not appear to be a structural or functional homolog in prokaryotes. Rad54, a member of the Swi2/Snf2 family of proteins which function in diverse chromosomal processes (reviewed in Refs. 33 and 34), has a DNA-dependent ATPase activity (35). Importantly, the addition of Rad54 to a homologous pairing reaction dramatically stimulates the pairing rate, elevating it to a level comparable to what can be achieved with the prokaryotic recombinases (35). These results indicate that efficient homologous DNA pairing requires the cooperation between Rad51 and Rad54 proteins, and they provide an explanation as to the requirement for Rad54 in heteroduplex DNA formation during recombination processes. However, the manner in which Rad54 protein promotes heteroduplex DNA formation has remained completely unknown. Delineating Rad54 function is clearly of paramount importance for understanding the mechanism of heteroduplex DNA formation. Since the Rad54 protein from other eukaryotes exhibits a high degree of functional and structural homology to the S. cerevisiae counterpart (36 -41), the results obtained with yeast Rad54 should be germane for de-lineating the role of Rad54 protein in recombination processes in other eukaryotes. Here we present results from our biochemical and genetic studies which address the function of Rad54 in heteroduplex DNA formation.
Yeast Strains-The strains used for protein purification and genetic experiments are listed in Table I. DNA Substrates-X 174 viral (ϩ) strand was purchased from New England Biolabs and the replicative form (about 90% supercoiled form and 10% nicked circular form) was from Life Technologies, Inc. The 83-mer oligonucleotides used in the strand exchange experiments were: oligo 1 with 16% GC content: 5Ј-AAATGAACATAAAGTAAATAAGTA-TAAGGATAATACAAAATAAGTAAATGAATAAACATAGAAAATAAA-GTAAAGGATATAAA; oligo 2: the exact complement of oligo 1; oligo 3 with 37% GC content: 5Ј-TTGATAAGAGGTCATTTTTGCGGATGGCT-TAGAGCTTAATTGCTGAATCTGGTGCTGTAGCTCAACATGTTTTA-AATATGCAA; oligo 4: the exact complement of oligo 3. Oligos 2 and 4 were labeled at the 5Ј end with [␥-32 P]ATP by T4 polynucleotide kinase, and then annealed to their complements, oligos 1 and 3. The resulting duplexes were purified from 10% polyacrylamide gels by overnight diffusion at 4°C into TAE buffer (40 mM Tris acetate, pH 7.5, 0.5 mM EDTA).
Purification of Rad54 Protein-BJ5464 harboring pR54.1 was grown in galactose containing medium as described (35). Cells were harvested by centrifugation and stored at Ϫ70°C until use. All the purification steps were carried out at 4°C. Extract was prepared from 400 g of yeast paste using a French Press in cell breakage buffer containing high salt (0.6 M KCl) and protease inhibitors, as described (42). The crude lysate (Fraction I) was clarified by ultracentrifugation (100,000 ϫ g for 120 min) and then treated with ammonium sulfate at 0.28 g/ml to precipi- Purification of rad54 K341A and rad54 K341R Mutant Proteins-For the purification of the rad54 K341A and rad54 K341R mutant proteins, 400 g of yeast strain BJ5464 harboring pR54.2 and pR54.3 were used. Extract preparation and the column chromatographic steps were carried out as described for wild type Rad54 protein above.
Purification of RPA-The RPA used in this study was purified either from a bacterial overexpression system (43) or from yeast strain BJ5464 harboring plasmids which overexpress the three subunits of RPA simultaneously; the latter yeast-based RPA overexpression system was a kind gift from Richard Kolodner. Chromatographic fractionation of bacterial and yeast extracts was done as described previously (26). We have noticed no difference between RPA preparations obtained from bacteria or from yeast in their ability to promote Rad51/Rad54-mediated homologous DNA pairing.
MMS Sensitivity-The sensitivity of various yeast strains to methyl methanesulfonate was examined as described previously (22). Briefly, the rad54⌬ yeast strain LSY403 harboring pTB326, pR54.4, pR54.5, or pR54.6 was grown in complete synthetic medium lacking tryptophan for 24 h to stationary phase, collected by centrifugation, washed once with 50 mM KH 2 PO 4 , pH 7.5, and then resuspended in the same buffer to the density of 1 ϫ 10 7 cells/ml. MMS (Aldrich; Ͼ99%) was added to the cell suspensions to 0.5% final concentration and after varying times at 25°C, aliquots of the cell suspensions were withdrawn, treated with an equal volume of 10% Na 2 S 2 O 3 to neutralize the MMS, and the cells were plated on complete synthetic medium lacking tryptophan after appropriate dilutions with distilled water. Colonies were counted after 5 days of incubation at 30°C.
Recombination Rate Determinations-Recombination rates were calculated according to the median method of Lea and Coulson (44) as described (45). Strains were streaked onto solid YEPD medium and grown at 30°C for 2-3 days. Nine colonies from each strain were used for each fluctuation test. For the haploid strains, three strains of each genotype were subject to fluctuation tests. Rate determinations in diploid strains were done on two crosses of each genotype using fresh zygotes. Three zygotes from each cross were subject to fluctuation tests.
ATPase Assay-The indicated amounts of Rad54 and mutant rad54 proteins were incubated at 37°C with X174 replicative form DNA (30 M base pairs; 90% supercoiled form and 10% nicked circular form) in 10 l of reaction buffer (30 mM Tris-HCl, pH 7.5, 100 g/ml BSA, 5 mM MgCl 2 , 1.5 mM [␥-32 P]ATP, 0.5 mM DTT, and 50 mM KCl), and the reaction was terminated by the addition of SDS to 1% final concentration. Reaction products were separated by thin layer chromatography on PEI cellulose (46) and quantified in the PhosphorImager.
Homologous DNA Pairing Reaction-The reaction was assembled as described previously (21,35). Briefly, 3.6 g of Rad51 protein (6.6 M) and 82 ng of X174 viral (ϩ) strand (20 M nucleotides) were mixed in 10 l of reaction buffer (35 mM potassium/MOPS, pH 7.2, 40 mM KCl, 2.5 mM ATP, 3 mM MgCl 2 , 1 mM DTT, and an ATP regenerating system consisting of 20 mM creatine phosphate and 28 g/ml creatine kinase). After a 5-min incubation at 37°C, 1.6 g of RPA (1.1 M) was added, followed by a 5-min incubation at 37°C, and then the indicated amounts of Rad54 protein or rad54 mutant proteins, 50 ng of 32 Plabeled ApaLI linearized X174 dsDNA (12.3 M nucleotides), and 1 l of 50 mM spermidine were incorporated. The complete reaction mixture (12.5 l) was incubated at 37°C, and 6-l portions were withdrawn at the indicated times and processed for agarose gel electrophoresis as described (35).
Oligonucleotide-based Homologous DNA Pairing System-Oligo 2 or oligo 4 (23 M nucleotides) and 3.3 g of Rad51 (7.7 M) were incubated in 8 l of reaction buffer (30 mM potassium/MOPS, pH 7.2, 1.5 mM ATP, 10 mM creatine phosphate, 28 g/ml creatine kinase, 3 mM MgCl 2 , 0.5 mM DTT) for 4 min at 37°C. Rad54 protein, 1 l of 50 mM spermidine, and the homologous 32 P-labeled duplex (46 M nucleotides) were incorporated to complete the reaction mixture (10 l). After incubation at 37°C for the indicated times, reaction mixtures were deproteinized and resolved in 12% polyacrylamide gels, which were dried onto a sheet of DEAE paper (DE81 from Whatman), and the reaction product was quantified in the PhosphorImager.
DNA Binding-The 83-mer duplex obtained by hybridizing oligo 1 to 5Ј 32 P-labeled oligo 2 was prepared as described above. The indicated amounts of Rad54 protein (18 to 108 nM) and the DNA substrate (7.6 M nucleotides) were incubated for 5 min at 25°C in 10 l of reaction buffer (30 mM potassium/HEPES, pH 7.2, 5 mM MgCl 2 , 0.5 mM DTT, 10 mM creatine phosphate, 28 g/ml creatine kinase, 50 g/ml BSA, and 50 mM KCl) and DNA mobility shift was analyzed in 12% polyacrylamide gels run in TAE buffer at 4°C. ATP was added to 1.5 mM final concentration as indicated.
Rad54 Protein Cross-linking-In the standard cross-linking reaction, wild type Rad54 or mutant rad54 protein, 2 g (2 M), was preincubated with or without 400 ng of DNA (61 M base pairs) for 5 min at 25°C in 10 l reaction buffer (20 mM HEPES, pH 7.2, 5 mM EDTA, 5 mM MgCl 2 , 50 mM NaCl), followed by the addition of bis-maleimidohexane (BMH; purchased from Pierce) to the final concentration of 200 M. After a 10-min incubation at 25°C, the reaction was quenched by the addition of 1 l of 2-mercaptoethanol. The samples were analyzed by SDSpolyacrylamide gel electrophoresis in 7.5% gels. For the sizing experiments, the protein cross-linking reactions were scaled up 10-fold to contain 20 g of Rad54 protein in a reaction volume of 100 l. After quenching the reaction with 2-mercaptoethanol, the volume of the mixtures was adjusted to 600 l with buffer K containing 500 mM KCl, 0.01% Nonidet P-40, and 1 mM 2-mercaptoethanol, before it was filtered through a Sephacryl S300 column as described below. The content of Rad54 protein in various column fractions was determined by immunoblot analysis. In the experiment in Fig. 7D, Rad54 protein, 30 ng (30 nM), was incubated with 5 ng to 2 g of DNA (0.75 to 300 M base pairs) in 10 l of buffer and BMH, quenched with mercaptoethanol, and then subject to immunoblot analysis.
Molecular Sizing-All the sizing experiments were carried out at 4°C. In the experiment described in Fig. 7B, the sample containing Rad54 protein was filtered through a Sephacryl S300 column (1 ϫ 44.5 cm; total 35 ml) at a flow rate of 0.1 ml/min, collecting 4-min fractions. Portions of the column fractions were subject to immunoblot analysis to determine their content of the Rad54 protein. The molecular size markers used for calibrating the column were thyroglobulin (669 kDa), aldolase (158 kDa), and bovine serum albumin (67 kDa). The column void (fractions 29 to 31) was identified by filtering blue dextran. Buffer K containing 500 mM KCl, 0.01% Nonidet P-40, and 1 mM 2-mercaptoethanol was used as the column buffer.
Topoisomerase I-linked Assay-Relaxed DNA was prepared by incubating 15 g of X 174 replicative form DNA with 40 units of calf thymus topoisomerase I (Life Technologies, Inc.) in 100 l of buffer (50 mM Tris-HCl, pH 7.5, 10 mM MgCl 2 , 50 g/ml BSA, and 1 mM DTT) for 2 h at 37°C, purified by phenol extraction and ethanol precipitation, and redissolved in TE to a concentration of 900 g/ml. The indicated amounts of Rad54 protein was incubated with 90 ng of relaxed DNA (27.3 M nucleotides) in 8.5 l of reaction buffer (30 mM Tris-HCl, pH 7.5, 50 g/ml BSA, 5 mM MgCl 2 , 0.5 mM DTT, 50 mM KCl, 20 mM creatine phosphate, and 28 g/ml creatine kinase) for 5 min at 25°C, followed by the addition of 1 l of 20 mM ATP and 5 units of calf thymus topoisomerase I in 0.5 l. The incubation was continued for 10 min at 37°C, and the reaction quenched by adding 1 l of 10% SDS and 0.5 l of proteinase K (10 mg/ml stock). Reaction mixtures were deproteinized at 37°C for 10 min, subject to electrophoresis in 0.8% agarose gels in TAE buffer, followed by ethidium bromide staining of the gels to visualize the DNA species.
In the experiment described in Fig. 6B, a 10 times scaled up reaction (2 M Rad54 and 27.3 M nucleotides) was deproteinized and then loaded onto an agarose gel. The agarose gel slice containing the faster migrating Form U DNA species was excised and the DNA was isolated using the GeneClean kit (BIO 101 Inc.). To obtain relaxed DNA marker, topologically relaxed DNA was purified from an agarose gel containing 20 M ethidium bromide the same way. Form U DNA, relaxed DNA, and supercoiled X DNA were analyzed in an 0.8% agarose gel or one that contained 20 M chloroquine diphosphate in TAE buffer, and then stained with ethidium bromide to reveal the DNA species.
Rad51 and Rad54 Complex Formation-For making affinity matrix, purified Rad51 protein was covalently conjugated to Affi-Gel 15 beads following the instructions of the vendor (Bio-Rad) to yield a matrix containing 2 mg/ml of Rad51 protein. As control matrix, bovine serum albumin was coupled to Affi-Gel 15 beads to 5 mg/ml. To examine binding of wild type Rad54 protein and mutant rad54 proteins to the Rad51 affinity beads, these proteins, 1 g each, were mixed gently with 5 l of Affi-Rad51 or Affi-BSA beads for 45 min at 25°C in 150 l of buffer (25 mM Tris-HCl, pH 7.5, 10% glycerol, 0.01% Nonidet P-40, and 0.5 mM DTT) containing 150 mM KCl and 10 g of BSA. The beads were washed once with 150 l of buffer with 300 mM KCl and then treated with 30 l of 3% SDS at 37°C for 10 min to elute the bound Rad54 or mutant rad54 protein. Portions of the starting material (10 l), the supernatant containing unbound Rad54 or mutant rad54 protein (10 l), and the SDS eluate (2.5 l) were subject to immunoblot analysis to determine their content of Rad54 or mutant rad54 protein.

Rad54
Protein Promotes Homologous DNA Pairing-In a previous study (35), we reported that the addition of Rad54 protein to a homologous DNA pairing and strand exchange reaction that employed X ssDNA and dsDNA as substrates increases the level of joint molecules markedly, demonstrating that Rad54 promotes the formation of the heteroduplex joint. Although Rad54 has a robust ATPase function that is dependent on dsDNA for its activation, it appears to be devoid of a DNA helicase activity. Consistent with the absence of a DNA helicase activity, no clear evidence for a branch migration function in Rad54 has been observed (35).
To further characterize the role of Rad54 in promoting heteroduplex joint formation, we examined the effect of Rad54 in an oligonucleotide-based homologous DNA pairing and strand exchange system. The reaction end product, a displaced radiolabeled ssDNA, is quantified by PhosphorImager analysis after electrophoresis in a polyacrylamide gel (Fig. 1A). Two sets of 83-mer DNA substrates were used, the first set had a relatively low GC content (16% GC) and the second set had a significantly higher GC content (37% GC) as described in Ref. 47. Given the short length of these oligonucleotides, one might expect that the search for DNA homology should not be as rate-limiting as when X DNA substrates are used, because the probability of productive collisions between the two oligonucleotide substrates resulting in homologous registry should be higher, and that extensive sliding of the substrates as part of the homology search process should be irrelevant in the oligonucleotide system. Furthermore, the resolution of the synapsed DNA substrates by strand exchange should also not be rate-limiting, especially when the low GC content substrate set was used (48). As shown in Fig. 1B, even with these short DNA substrates, the inclusion of Rad54 protein resulted in a marked stimulation of product formation. This is true whether the low GC content pair of substrates or the relatively high GC content pair was used in the reaction. Thus, the results from these experiments appear to support our overall premise that Rad54 functions primarily to promote the formation of the nascent heteroduplex joint.
ATP Hydrolysis Defective Mutant Rad54 Proteins-In our study on the mechanism of Rad54 action in heteroduplex joint formation, we sought to first establish the role of ATP binding and hydrolysis in Rad54 function, as it seems reasonable to assume that the dsDNA-dependent ATPase activity is important for the ability of Rad54 to stimulate Rad51-catalyzed homologous DNA pairing. To accomplish this goal, two specific mutant variants of Rad54, with substitutions of the highly conserved lysine residue (lysine 341) in the Walker type A nucleotide binding sequence by either alanine or arginine, were generated. Based on studies with equivalent mutations in other DNA-dependent ATPases (22, 49 -51), the rad54 K341A mutant protein was expected to be defective in interaction with ATP, whereas the rad54 K341R mutant was expected to retain the ability to bind ATP but be devoid of ATPase activity. For biochemical studies, wild type Rad54 protein and the two mutant rad54 proteins were overexpressed in yeast cells ( Fig. 2A) and purified to near homogeneity (Fig. 2B).
We examined the ATPase activities of the purified proteins using X dsDNA as cofactor. As shown in Fig. 2C, the rad54 K341A mutant was found to be completely defective in ATP hydrolysis; this conclusion was validated in ATPase reactions that used a vast excess of the mutant protein and over a wide range of pH values from 5.5 to 8.5 in increments of 0.5 (data not shown). As for the rad54 K341R mutant protein, we found that it retains a residual level of ATPase activity, at about 2 to 3% of the wild type level (Fig. 2C) over a wide range of pH and protein concentration (data not shown). These results are consistent with the overall premise that the rad54 K341A and rad54 K341R mutations specifically inactivate ATP binding and ATP hydrolysis, respectively.
DNA Binding Is Independent on ATP-Since the ATPase activity of Rad54 is dependent on dsDNA, it was of considerable interest to examine whether DNA binding itself is affected by ATP. We used DNA mobility shift in polyacrylamide gels as assay to address this point. Rad54 protein was incubated in the presence or absence of ATP with an 83-mer 32 P-labeled duplex, and the amount of Rad54-DNA nucleoprotein complex formed in both cases was compared. The results from this experiment showed no apparent dependence of the level of nucleoprotein complex formation on ATP (Fig. 3A). When we replaced ATP with either ADP or ATP␥S, we again could not detect any difference in dsDNA binding by Rad54 (data not shown). We also examined the rad54 K341A and rad54 K341R mutant proteins for the ability to interact with dsDNA, and as expected, found that they bind dsDNA with wild type affinity (Fig. 3B). Similar results were obtained when X dsDNA was used as substrate in DNA mobility shift assays in agarose gels (data not shown). Taken together, the results demonstrate that Rad54 binding to dsDNA does not require ATP, and that the two ATPase-defective rad54 mutant proteins are just as capable of binding dsDNA as wild type Rad54 protein.
Rad54 ATPase Is Required for Homologous DNA Pairing-To examine the requirement of the Rad54 ATPase activity in homologous DNA pairing, increasing amounts of Rad54, rad54 K341A, and rad54 K341R proteins were added to reactions that used the circular (ϩ) strand and 32 P-labeled linear double-stranded DNA from the bacteriophage X174 as DNA substrates (19). The reaction products, the joint molecule, and nicked circular duplex (Fig. 4A) were resolved in an agarose gel and then visualized by autoradiography of the dried gel. As shown in Fig. 4, B and C, whereas Rad54 protein exerts a strong stimulation on the homologous DNA pairing rate to yield joint molecules, as reported previously (35), neither the rad54 K341A nor the rad54 K341R mutant protein was active. We showed previously that Rad54 protein promotes the pairing between a linear ssDNA molecule and a supercoiled DNA molecule to yield a D-loop (35). We have found that neither of the two rad54 mutants has the ability to promote D-loop formation by Rad51 (data not shown).
Results from independent studies have demonstrated that Rad54 protein interacts specifically with the Rad51 protein (35,52,53). Other experiments have indicated that Rad54 has no effect on homologous DNA pairing mediated by the E. coli RecA protein and human Rad51 protein (data not shown), which is consistent with the premise that a specific interaction with Rad51 protein is required for stimulation of joint molecule formation by Rad54. To eliminate the possibility that the mutant rad54 proteins may be deficient in interaction with Rad51 protein, we examined the binding of the mutant rad54 proteins to Affi-Gel 15 beads containing covalently conjugated Rad51 protein (Affi-Rad51), using Affi-Gel beads containing bovine serum albumin (Affi-BSA) as control (Fig. 4D). The two mutant rad54 proteins are as proficient as wild type Rad54 in interacting with Rad51, as indicated by the retention of Ͼ90% of Rad54 and mutant rad54 proteins on the Affi-Rad51 beads (Fig. 4D).
Role of ATP Binding and Hydrolysis in Biological Function-Having established that the two mutant rad54 proteins are defective in ATP hydrolysis, we proceeded to test whether these mutants have any biological activity in vivo. For this purpose, we introduced plasmids expressing either Rad54 or the mutant rad54 proteins into the rad54 deletion strain LSY403, and examined the sensitivity of the plasmid bearing strains to MMS. As shown in Fig. 5, whereas expression of the wild type RAD54 gene restored MMS resistance, cells expressing the rad54 K341A and rad54 K341R mutant alleles remained highly sensitive to MMS. From this result, we infer that the FIG. 2. Overexpression and purification of Rad54, rad54 K341A, and rad54 K341R proteins. A, extracts from yeast strain BJ5464 harboring plasmids pR54.1 (2, GAL-PGK-RAD54), pR54.2 (2, GAL-PGK-rad54 K341A), and pR54.3 (2, GAL-PGK-rad54 K341R) were subject to immunoblot analysis with affinity purified anti-Rad54 antibodies, as described (35). The level of Rad54 protein in the extract from wild type BJ5464 cells without Rad54 overexpression is too low to detect under these conditions (Ref. 35; not shown). B, purity analysis. A 7.5% SDS-polyacrylamide gel containing 1 g each of purified Rad54, rad54 K341A, and rad54 K341R proteins (lanes 1, 2, and 3, respectively) and size markers was stained with Coomassie Blue. C, time course of dsDNA-activated ATP hydrolysis by Rad54 (30 nM, Ⅺ), rad54 K341R (100 nM, E), and rad54 K341A (100 nM, छ) proteins. dsDNA-dependent ATPase activity of Rad54 is indispensable for its DNA repair function. Since the rad54 K341R protein is expected to retain the ability to bind ATP, the genetic results further suggest that ATP binding alone is not sufficient for MMS repair. In diploid cells also, the homozygous rad54 K341A and rad54 K341R mutants and the heterozygous rad54 K341A/rad54⌬ and rad54 K341R/rad54⌬ mutants are as defective as the homozygous rad54⌬ mutant in MMS repair (data not shown).
To determine the effect of the rad54 mutations on recombination in vivo, we replaced the wild type chromosomal RAD54 allele with the rad54 K341R or rad54 K341A allele and determined spontaneous mitotic recombination rates. Using a heteroallelic LEU2 duplication, the rates of gene conversion events, scored as Leu ϩ Ura ϩ segregants (3), were determined. As shown in Table II, the rad54 null strain is 90-fold reduced in haploid intrachromosomal gene conversions. The rad54 K341A mutant shows a similar reduction in haploid recombination, with a 77-fold decrease compared with wild type. The rad54 K341R mutant is not as severely affected, but the 42-fold decrease in recombination is highly significant. These results show that the K341A and K341R mutations grossly affect the ability of the Rad54 protein to support mitotic gene conversion events.
We then measured interchromosomal gene conversion events in diploids, using the same LEU2 alleles. Interestingly, the rad54 null mutant is not as affected in diploid gene conversion as in haploid gene conversion, with only a 5-fold decrease in the Leu ϩ rate (Table II). Neither the rad54 K341A nor the rad54 K341R mutant was significantly decreased in diploid mitotic gene conversion. This is in contrast to the results with the haploid strains. One possible explanation is that the RAD54-related gene RDH54/TID1 supports most of the diploid interchromosomal gene conversions, while only a small fraction is effected through the RAD54 gene in diploids (3). The fact that little decrease in the Leu ϩ rate was observed in the rad54 ATPase mutants are devoid of the ability to promote homologous DNA pairing. Homologous DNA pairing reactions with Rad51 and RPA only (lane 2) or containing, in addition to Rad51 and RPA, wild type Rad54 protein (40,80, and 120 nM in lanes 3 to 5, respectively), rad54 K341A protein (40,80, and 120 nM in lanes 6 to 8, respectively), and rad54 K341R protein (40,80, and 120 nM in lanes 9 to 11, respectively). The reactions were incubated at 37°C for 8 min (panel I) and 20 min (panel II). In lane 1, the DNA substrates were incubated in buffer without any recombination factor. Shown are the autoradiograms of dried agarose gels. Symbols: ds, input 32 P-labeled linear duplex; jm, joint molecules formed between the viral (ϩ) strand and the linear duplex; nc, nicked circular duplex, formed as a result of complete strand exchange between the viral (ϩ) strand and the linear duplex. C, the gel shown in panel I of B was subject to PhosphorImager analysis to obtain data points for a graphical representation of the results. D, rad54 K341A and rad54 K341R mutant proteins interact with Rad51. Purified Rad54, rad54 K341A, and rad54 K341R proteins were mixed with Affi-Gel 15 beads bearing bovine serum albumin (Affi-BSA) or Rad51 protein (Affi-Rad51), which were collected by centrifugation, washed with 300 mM KCl, and then treated with 3% SDS to elute the bound protein. The input material (IP), the supernatant containing unbound Rad54 protein (S), and the SDS eluate (E) were subject to immunoblot analysis to determine their Rad54 content. All of the KCl washes were found to contain an insignificant amount (Ͻ4%) of the total Rad54 and rad54 proteins and were not included in this figure. Symbols in B, C, and D: rad54K/A, rad54 K341A protein; rad54K/R, rad54 K341R protein.
K341A and rad54 K341R mutants could mean that their encoded products possess the ability to promote diploid mitotic gene conversion on their own, or alternatively, Rad54 protein, but not its ATPase activity, is required for the integrity of a higher order complex important for gene conversion in diploid cells.
Although the two rad54 K341 mutants have only a negligible effect on diploid mitotic recombination, their phenotype in meiosis is similar to that of the null mutant. Both the rad54 K341A and rad54 K341R diploids show 40 -50% sporulation (compared with 90% in the wild type diploid) and 60 -68% spore viability (compared with 95% in the wild type diploid). These numbers are very similar to those obtained with the rad54 null mutant diploid (3).
Rad54 Induces ATP Hydrolysis-dependent Alteration in DNA Structure-Our results have strongly suggested that the ATPase activity of Rad54 is not used to fuel a progressive, vectorial DNA unwinding reaction characteristic of DNA helicases. We reasoned that perhaps ATP hydrolysis may mediate a change in the conformation of the bound DNA, in a fashion that would promote joint molecule formation during the homologous DNA pairing reaction. One such possibility is the localized unwinding of the double helix. To test this idea, we used a topoisomerase I-linked assay, in which topologically relaxed DNA was added to reaction mixtures containing increasing amounts of Rad54 protein, calf thymus topoisomerase I, and either in the presence or absence of ATP. Unwinding of the relaxed duplex by Rad54 was expected to result in the formation of positively supercoiled domains in the adjacent regions of the duplex, which would be removed by the topoisomerase I present in the reaction. Subsequent removal of the bound Rad54 protein and the topoisomerase by deproteinization treatment would then yield a negatively supercoiled species. Fig. 6A shows that in the presence of Rad54 protein and ATP, supercoiled DNA species were formed, which were collectively designated as Form U DNA. The formation of Form U DNA was completely dependent on the presence of ATP, as even at the highest concentration of Rad54 protein, there was no trace of Form U DNA when ATP was omitted from the reaction.
To establish that Form U DNA is a negatively supercoiled DNA species as we expected, it was gel purified from a scaled up reaction and then analyzed, along with topologically relaxed DNA and naturally negatively supercoiled DNA, in agarose gels with or without 20 M chloroquine. Chloroquine intercalates into DNA, which, in the case of a topologically relaxed or

TABLE II Recombination rates in rad54 mutants
Rates for the haploid strains were determined using the direct repeat leu2-ri::URA3::leu2-bst. Leuϩ Uraϩ rates reflect intrachromosomal gene conversion events. Interchromosomal gene conversion rates were determined in diploids using the heteroalleles lei2-ri and leu2-bst. Strains used to determine recombination rates are the HKY series listed in Table I  positively supercoiled species, will induce positive superhelical turns in the DNA and cause the DNA to migrate faster in an agarose gel. In contrast, for a negatively supercoiled species, intercalation of a low amount of chloroquine would render the DNA migrating more slowly in an agarose gel as it would remove some of the negative superhelical turns. As shown in Fig. 6B, Form U DNA assumed a retarded mobility in the presence of chloroquine, indicating that it is indeed a negatively supercoiled species.
To examine whether ATP binding by Rad54 protein alone is sufficient for mediating the change in DNA linking number, or hydrolysis of ATP by Rad54 protein is in fact necessary, Rad54 protein was incubated with the relaxed DNA as before (Fig.  6A), except that the non-hydrolyzable ATP analogue ATP␥S or ADP was used instead of ATP. The results from this experiment indicated that ATP␥S (Fig. 6C, lanes 4 and 5) and ADP (Fig. 6C, lanes 6 and 7) were ineffective for the formation of Form U DNA, an observation consistent with the suggestion that ATP hydrolysis is required for Rad54-induced structural changes. To further confirm this finding we substituted the wild type Rad54 protein with either the rad54 K341A or the rad54 K341R mutant protein (Fig. 3C). As can be seen in Fig.  6D, neither of the two mutant proteins was capable of producing Form U DNA (lanes 4 -7), thus further confirming that ATP hydrolysis is indeed required for mediating the DNA linking number change.
Rad54 Oligomerizes on DNA-Sizing experiments in Sephacryl S300 with protein standards have suggested that Rad54 behaves as a monomer in solution (see later). Since a number of protein factors which function in various processes of chromosomal metabolism have been shown to behave as a monomer in solution, but in the presence of DNA, to either dimerize or form a higher order multimeric structure (reviewed in Ref. 54), it was of considerable interest to examine whether Rad54 protein would assemble into an oligomer on DNA. To do this, we used BMH, which is a protein cross-linking agent specific for free sulfhydryl groups.
Rad54 protein migrates in a denaturing SDS gel with an apparent size of about 105 kDa, which is in good agreement with the predicted size (102 kDa). As shown in Fig. 7A, in the absence of DNA, trapping with BMH yielded a novel Rad54 doublet that migrates with an apparent size of about 145 kDa and an additional minor species about 210 kDa (Fig. 7A, lane  2). The Rad54 doublet above the unmodified Rad54 protein in Fig. 7A (labeled as I-monomer) most likely corresponded to intramolecularly cross-linked Rad54 monomer as it migrated precisely at the position of unmodified Rad54 protein in the S300 sizing column (Fig. 7B). We tentatively attribute the aberrant mobility of the intramolecularly cross-linked monomeric Rad54 species in SDS gels to incomplete unfolding of the intramolecularly linked polypeptide by SDS. The species that shows a size of 210 kDa in SDS gels (labeled dimer in Fig. 7A), on the other hand, emerged from the S300 sizing column at a position earlier than unmodified Rad54 (Fig. 7B), consistent of it being a cross-linked dimeric/oligomeric form of Rad54 protein. There were also traces of other species of Rad54 protein that migrate above the Rad54 dimer, which could correspond to dimeric Rad54 species consisting of unmodified Rad54 crosslinked to an intramolecularly cross-linked monomer and two molecules of intramolecularly linked monomer cross-linked to one another, and perhaps other higher order species as well. For the sake of simplicity, these other Rad54 species are not marked.
We next examined the effect of DNA on the distribution of unmodified Rad54 protein relative to other species. Interestingly, the addition of X174 duplex DNA greatly enhanced the formation of the dimeric species (Fig. 7A, lane 3), which was present in only a trace quantity in the absence of DNA (Fig. 7A,  lane 2). Since Rad54 is a dsDNA-dependent ATPase, it was of considerable interest to examine whether or not ATP would influence DNA-induced oligomerization of the Rad54 protein.
However, the addition of either ATP or the non-hydrolyzable analogue ATP␥S did not affect the distribution of the various species of Rad54 protein (Fig. 7A, lanes 4 to 9), suggesting that neither ATP binding nor its hydrolysis has a significant effect on Rad54 oligomerization. This suggestion is supported by the observation that both the rad54 K341A and rad54 K341R proteins display the same tendency as wild type Rad54 in forming the oligomeric species when DNA is included in the reaction (Fig. 7C).
Since a relatively high concentration of Rad54 compared with the DNA was used in the cross-linking experiments, it could be argued that the cross-linked species seen in earlier experiments were due to clustering of Rad54 molecules on the DNA substrate and did not reflect dimerization/oligomerization of the protein. To eliminate this caveat, a fixed quantity of Rad54 protein (30 nM) was used in cross-linking reactions with increasing concentrations of DNA, from a base pairs/protein monomer ratio of 25:10,000. The results from this experiment show that the amount of the cross-linked Rad54 species did not change significantly from the lowest to the highest level of DNA used, thus indicating an intrinsic ability of Rad54 protein to dimerize/oligomerize in the presence of DNA (Fig. 7D). As expected, oligomerization of the rad54 K341A and rad54 K341R mutant proteins occurs just as efficiently in the presence of high base pairs/monomer as when lower amounts of DNA was used (data not shown).

DISCUSSION
The rad54 K341A and rad54 K341R alleles are shown in this study to encode mutant rad54 proteins that are defective in ATPase activity. Clever et al. (52) reported previously that whereas overexpression of wild type RAD54 gene suppresses the ultraviolet and MMS sensitivities of a rad51⌬ mutant, overexpression of the rad54 K341R allele has no such effect. Here we extend the observations of Clever et al. (52) by showing that the rad54 K341R allele is also defective in the repair of DNA lesions induced by MMS, in intrachromosomal gene conversion in haploid cells, and in meiosis. In addition, we have documented that the rad54 K341A mutant gene behaves similarly to the rad54 K341R allele phenotypically. Taken together, it seems clear that the Rad54 ATPase function is indispensable for different types of mitotic and meiotic recombination.
The gel filtration and protein cross-linking results (Fig. 7) have suggested that Rad54 protein is a monomer in solution. In the presence of dsDNA, a dimeric and other higher order species of Rad54 protein can be trapped by cross-linking treatment, suggesting a propensity of the Rad54 protein to assemble into a dimer/oligomer. We have also demonstrated that DNAinduced dimerization/oligomerization of Rad54 does not require ATP, and that rad54 K341A and rad54 K341R mutant proteins also undergo DNA induced dimerization/oligomerization with or without ATP.
In speculating about how ATP hydrolysis may be utilized in promoting Rad54 function, a number of distinct possibilities were considered. First, we thought it was possible that Rad54 might possess a DNA helicase activity, which could be utilized to promote the extension of the heteroduplex DNA joint. However, such a branch migration function alone should primarily promote the conversion of the joint molecule (see Fig. 4A) to the final strand exchange product, nicked circular duplex, while it is not expected to have any significant effect on the rate of joint molecule formation. Our results have shown that Rad54 protein exerts a strong stimulatory effect on the formation of joint molecules but has no obvious effect on their conversion to the final products ( Fig. 4B; Ref. 35). Moreover, despite considerable efforts to look for a DNA helicase activity, we have thus far been unable to demonstrate such an activity in Rad54 (35). Further evidence for a role of Rad54 protein in promoting the formation of the heteroduplex joint has come from the observation that even with short 83-mer DNA substrates where branch migration is not expected to be a rate-limiting step, the product formation is still strongly stimulated by Rad54 protein (Fig. 1).
In the topoisomerase I-linked assay, we found that Rad54 protein induces a change in the DNA topology in a reaction that is strictly dependent on ATP. The negatively superhelical nature of the product suggests that Rad54 could mediate transient DNA strand separation in the bound duplex. It is also possible Rad54 could bend or untwist the DNA double helix in the bound regions without actual DNA strand separation. We have shown that this DNA conformational change requires ATP hydrolysis, as ATP␥S does not support this reaction and mutant rad54 K341A and rad54 K341R proteins are both ineffective. The requirement for ATP hydrolysis is not due to a dependence of DNA binding on ATP, because Rad54 does not need ATP to interact with dsDNA, and rad54 K341A and rad54 K341R mutant proteins appear to be just as proficient as Rad54 in dsDNA binding. While our manuscript was in preparation, a report appeared indicating that the human Rad54 protein also mediates an ATP hydrolysis-dependent change in the DNA linking number (41); these authors used nicked circular duplex as substrate and DNA ligation to trap the linking number change. Tan et al. (41) favor the idea that the change in the DNA linking number is due to unwinding of the DNA duplex by human Rad54. We do not yet have definitive evidence that the DNA linking number change induced by yeast Rad54 in the topoisomerase I-linked assay is due to DNA unwinding or to another type of structural change in the DNA. However, it seems reasonable to propose that the ability to alter DNA conformation in a manner that is dependent on ATP hydrolysis is a conserved property of eukaryotic Rad54 proteins, and that this property is germane for the function of eukaryotic Rad54 proteins in recombination.
How May Rad54 Work?-Based on previous biochemical studies, it is clear that homologous DNA pairing and strand exchange occur within the confines of the Rad51-ssDNA "presynaptic" filament. We propose that by virtue of its ability to interact with Rad51, Rad54 rests on the Rad51-ssDNA filament. At the expense of ATP hydrolysis, Rad54 alters the conformation of the incoming duplex molecule, and this conformational change is directly linked to the ability of Rad54 to stimulate the formation of heteroduplex joint. This structural alteration may render the separation of the DNA strands in the duplex more facile, or perhaps entail DNA strand separation. If one of the unwound DNA strands is homologous to the ssDNA strand bound within the presynaptic filament, then the formation of a nascent heteroduplex joint will ensue, an event that is expected to momentarily capture the duplex molecule. Subsequent to the capture of the incoming duplex, homologous alignment of the recombining DNA molecules and formation of a plectonemic linkage upon the location of a free end will follow. In addition, destabilization of the duplex DNA by Rad54 next to a region of pre-existing heteroduplex DNA may increase the length of the heteroduplex joint and may therefore stabilize the joint. Such a function is expected to be important in the early stages of plectonemic joint formation, as the nascent heteroduplex may be particularly prone to dissolution. Rad51/Rad54 interaction is critical in this scenario since the same reaction catalyzed by RecA or human Rad51 is refractory to yeast Rad54 protein (data not shown). Protein cross-linking studies described here suggest that the functional form of Rad54 protein is a dimer or an oligomer. Whether the Rad54 protein dimerizes/oligomerizes upon the incorporation of the duplex DNA into the Rad51-ssDNA protein filament or at a different time during the DNA homology search and pairing process are issues that remain to be determined.

RDH54, Haploid-and Diploid-specific Gene Conversion-
Genetic studies have indicated that intrachromosomal gene conversion in haploid cells is considerably more dependent on the RAD54 gene than is interchromosomal gene conversion in diploid cells (Ref . 3; Table II of this work). Interestingly, the rad54 K341A and rad54 K341R mutants are near wild type in interchromosomal gene conversion, as compared with a 5-fold drop in gene conversion rate in the rad54⌬ diploid (Table II). These observations suggest a structural role for Rad54 protein in mediating interchromosomal recombination in diploid cells that is independent of the Rad54 ATPase function. These results can be explained by proposing that Rad54 protein, but not its ATPase activity, is indispensable for the functional integrity of a higher order protein complex required for wild type levels of interchromosomal recombination.
The S. cerevisiae RDH54 encoded product shows 35% identity to the Rad54 protein. Although the rdh54⌬ mutation confers only slight sensitivity to MMS, it increases the MMS sensitivity of a rad54⌬ strain, and the rad54⌬,rdh54⌬ double mutant is much more impaired in meiosis than either single mutant (3,5). Interestingly, diploid yeast strains harboring homozygous deletions of RAD54 and RDH54 are growth impaired, and the impairment can be overcome by simultaneously deleting RAD51, indicating that the growth deficiency observed in the former is due to attempted, but incomplete recombination. Likewise, a diploid harboring homozygous deletions of RDH54 and SRS2 is inviable, and the inviability is rescued by simultaneously deleting RAD51 (3). In another study, Dresser et al. (55) have identified Rdh54 as a protein which interacts with the meiosis-specific Rad51 homolog Dmc1 in the twohybrid assay; these authors called the RDH54 gene TID1. Based on these and other data, it has been suggested that Rdh54 protein functions with Rad51 protein in gene conversion between homologs in mitotic cells (3) and with both Dmc1 and Rad51 in interhomolog recombination in meiotic cells (3,5). Like Rad54 protein, Rdh54 has Walker-type nucleotide binding motifs. Considering the involvement of Rdh54 in recombination and repair as well as the structural similarity of Rdh54 to Rad54, it is a distinct possibility that Rdh54 also possesses biochemical functions similar to what have been reported for Rad54 protein and affects heteroduplex DNA formation by a similar mechanism.
As alluded to above, the genetic results suggest that Rad54 may be important for the functional integrity of a protein complex that helps mediate diploid specific interhomolog gene conversion. Whether such a complex exists, and whether Rdh54 protein is also part of this complex, are issues to be addressed.