Rad54 protein exerts diverse modes of ATPase activity on duplex DNA partially and fully covered with Rad51 protein.

Rad54 protein is a Snf2-like ATPase with a specialized function in the recombinational repair of DNA damage. Rad54 is thought to stimulate the search of homology via formation of a specific complex with the presynaptic Rad51 filament on single-stranded DNA. Herein, we address the interaction of Rad54 with Rad51 filaments on double-stranded (ds) DNA, an intermediate in DNA strand exchange with unclear functional significance. We show that Saccharomyces cerevisiae Rad54 exerts distinct modes of ATPase activity on partially and fully saturated filaments of Rad51 protein on dsDNA. The highest ATPase activity is observed on dsDNA containing short patches of yeast Rad51 filaments resulting in a 6-fold increase compared with protein-free DNA. This enhanced ATPase mode of yeast Rad54 can also be elicited by partial filaments of human Rad51 protein but to a lesser extent. In contrast, the interaction of Rad54 protein with duplex DNA fully covered with Rad51 is entirely species-specific. When yeast Rad51 fully covers dsDNA, Rad54 protein hydrolyzes ATP in a reduced mode at 60-80% of its rate on protein-free DNA. Instead, saturated filaments with human Rad51 fail to support the yeast Rad54 ATPase. We suggest that the interaction of Rad54 with dsDNA-Rad51 complexes is of functional importance in homologous recombination.

The key player of recombination is the Rad51 protein, the eukaryotic homologue of the bacterial RecA protein (reviewed in Ref. 2). With the aid of accessory factors, Rad52 protein and Rad55-Rad57 complex, Rad51 displaces RPA from ssDNA tails and assembles into a nucleoprotein filament. This filament, also named presynaptic complex, is the molecular device for homology search on partner duplex DNA molecules and subsequent DNA strand exchange. Further steps comprise hybrid DNA extension, priming of DNA synthesis, branch migration, and resolution of Holliday junctions, yielding DNA molecules repaired in an intrinsically error-free way. The detailed mechanism of these latter stages remains poorly understood.
The model presented above is supported by data on the physical association of the corresponding proteins and their recombinational activities in vitro (reviewed in Ref. 3). Inactivation of RAD51, RAD52, or RAD54 confers the strongest recombination/repair-deficient phenotypes. Rad51 protein can associate with itself, with Rad52, Rad54, and with the Rad55/ Rad57 heterodimer (4 -9). RPA is an essential component of the presynaptic phase required for proper loading of Rad51 on ssDNA and formation of functionally active filaments (10). RPA also associates with Rad52 (11)(12)(13). Rad52 is thought to enhance filament formation by favoring the displacement of RPA from ssDNA by Rad51 (14 -17). Likewise, Rad55/Rad57 heterodimer stimulates presynaptic filament formation in the presence of RPA, possibly by nucleating the ssDNA-Rad51 filament (18).
Biochemical and genetic evidence points to a specific function of Rad54 protein in eukaryotic homologous recombination. Rad54 protein belongs to the Swi2/Snf2 group of superfamily 2 of DNA-dependent/stimulated NTPases, which are known to act in transcriptional regulation of gene expression (e.g. Mot1 and the ATPase subunits of chromatin remodeling complexes), DNA repair (Rad5, Rad16, Rad26/ERCC6, and Rad54), and chromosome segregation (lodestar protein) (reviewed in Refs. 19 and 20). These proteins carry putative helicase motifs, but no helicase activity has been demonstrated for any of them in conventional strand displacement assays. Rad54 protein possesses potent dsDNA-dependent ATPase activity and DNA remodeling activity, which are dependent on ATP hydrolysis (21)(22)(23)(24)(25)(26). Mutations in the Rad54 ATP binding site (Walker A-box) confer repair-deficient phenotypes similar to that of a deletion mutant (23,27). This indicates that the cellular function of Rad54 protein is dependent on its ATPase activity.
Genetic studies have shown that deletion of the SRS2 gene is lethal in a rad54 mutant (28). Srs2 is a DNA helicase involved in the metabolism of ssDNA gaps, preparing the substrate for the RAD6 postreplicational repair pathway (29). In the absence of Srs2, these gaps are channeled into the homologous recombination pathway (29). Importantly, mutations in RAD52, RAD51, RAD55, and RAD57, genes that encode proteins acting in the presynaptic and synaptic phases of recombination, suppress this synthetic lethality (30). These observations, together with cytological data showing that Rad51 foci accumulate in a rad54 mutant (31), suggest that, in yeast, Rad54 protein acts in recombination after presynaptic filament formation. In vitro, Rad54 was found to stimulate Rad51-mediated D-loop formation and three-stranded DNA strand exchange via an interaction with the established presynaptic filament (21,26,32). Rad54 protein enhances, in an ATP-dependent way, both synaptic joint formation and further extension of heteroduplex DNA in the postsynaptic phase (32,33). This stimulation is accompanied by a concomitant increase of the Rad54 ATPase and DNA remodeling activities. It is thought that topological changes in the duplex DNA molecule made by Rad54 facilitate both homology search and DNA strand exchange (25,26). Rad54 protein can interact, via its NH 2 -terminal part, with Rad51 protein in solution (6,7), and this interaction is likely to have biological relevance at the level of the nucleoprotein complex (26,32). To achieve maximal stimulation of DNA strand exchange, Rad54 has to be added to the reaction mixture either right after presynaptic filament formation or as the last component after partner duplex DNA addition. It was proposed that the presynaptic filament first binds Rad54 protein and then targets the protein to duplex DNA. The activated Rad54 facilitates the homology search by Rad51 protein and further steps of the reaction (26,32).
It is known that Rad51 protein has affinity to both ssDNA and dsDNA (34 -38) and that Rad51 binding to dsDNA substrate inhibits initial stages of homologous pairing in vitro (39). On the other hand, the product of DNA strand exchange, heteroduplex DNA, is likely to be covered by Rad51 protein, as inferred from the mechanism of the RecA protein-mediated reaction. Thus, the dsDNA-Rad51 complex is an important intermediate or by-product of homologous recombination with as yet unknown functional significance. In addition, it is morphologically very similar to the presynaptic filament (35,39). We therefore decided to investigate the possible interaction of Rad54 protein with dsDNA-Rad51 complexes. We show that besides the basic ATPase activity on protein-free duplex DNA, Rad54 exerts distinct modes of activity on dsDNA-Rad51 complexes: an enhanced ATPase mode on partial Rad51 filament and a reduced ATPase mode on duplex DNA fully covered with Rad51 protein. Our findings point to a specific role of the Rad54 interaction with duplex DNA-Rad51 complexes in homologous recombination.

EXPERIMENTAL PROCEDURES
Proteins and DNA-S. cerevisiae glutathione S-transferase-tagged Rad54 (ScRad54) protein was purified as described (32). Plasmids for expression of recombinant S. cerevisiae Rad51 (ScRad51) and human Rad51 (hRad51) proteins were kindly provided by Dr. P. Sung. ScRad51 protein was purified as described (10,15). hRad51 protein was produced in Escherichia coli and purified essentially according to the protocol for the ScRad51 with an additional Mono S column step at the end. E. coli RecA protein was a kind gift of Dr. S. Kowalczykowski. Plasmid and phage DNA were purchased from Invitrogen and New England Biolabs. Rad51 and Rad54 concentrations are given in moles of monomers. Duplex DNA concentrations are given in moles of base pairs.
Charcoal-based ATPase Assay-Single-point analysis of ATP hydrolysis was performed using a charcoal-based assay as described (32). 50-l reaction mixtures containing either protein-free DNA or preformed DNA-Rad51 complexes at the indicated concentrations were incubated at 30°C with Rad54 protein in the ATPase reaction buffer containing 25 mM triethanolamine acetate, pH 7.5, 13 mM magnesium acetate, 1.8 mM dithiothreitol, 5 mM ATP, and 100 g/ml bovine serum albumin. Before the addition of a final component (usually Rad54 protein) the mixture was supplemented with 0.25 Ci of [␥-32 P]ATP (3,000 Ci/mmol). Reactions were terminated at specific time points by the addition of 10 volumes (500 l) of ice-cold charcoal mixture (2%, w/v, Norite TM charcoal (Sigma C-5260), 0.25 M HCl, 0.25 mM sodium pyrophosphate, 0.25 mM K 2 HPO 4 ). Samples were incubated on ice for 5 min, and then centrifuged for 5 min in an Eppendorf centrifuge at maximal speed; 400-l portions of supernatants were collected. Radioactivity of the supernatant reflecting the extent of ATP hydrolysis was counted in a Beckman scintillation counter.
ATP/NADH-coupled ATPase Assay-The assay was based on a reaction in which the regeneration of hydrolyzed ATP is coupled to the oxidation of NADH (40). Following each cycle of ATP hydrolysis, the regeneration system consisting of phosphoenolpyruvate and pyruvate kinase converts one molecule of phosphoenolpyruvate to pyruvate when the ADP is converted back to the ATP. The pyruvate is subsequently converted to lactate by L-lactate dehydrogenase resulting in the oxida-tion of one NADH molecule. The assay measures the rate of the NADH absorbance decrease at 340 nm, which is proportional to the rate of steady-state ATP hydrolysis. The constant regeneration of ATP allows monitoring the ATP hydrolysis rate over the entire course of the assay. A multicell holder permits the simultaneous analysis of up to 7 control and experimental samples. The assay was typically performed at 24°C, if not otherwise indicated, with 5 nM Rad54 protein and the indicated concentrations of DNA, Rad51 protein, or preformed DNA-Rad51 complexes in a 100-l volume of the ATPase reaction buffer supplemented with a regeneration system (3 mM phosphoenolpyruvate, 20 units/ml pyruvate kinase), 20 units/ml L-lactate dehydrogenase and NADH to give an A 340 of 1.6 -1.8 (about 250 -300 M). Extra NADH was added to resume those reactions where the NADH pool became exhausted. The absorbance data were collected using a Hewlett-Packard 8452A diode array spectrophotometer equipped with UV-visible ChemStation software. The rate of ATP hydrolysis was calculated from the following equation.
Calculated rates of ATP hydrolysis were plotted over corresponding time points where the absorbance readings were made. Comparison of different samples within the same kinetic time course was made over identical time intervals. DNA Unwinding Assay with Eukaryotic Topoisomerase I-The conditions of the DNA unwinding assay were adopted from a protocol originally used to measure DNA helicity in the region of homologous pairing promoted by RecA protein (41), with some modifications. Briefly, supercoiled pUC19 DNA at a concentration of 46 M was incubated with 2 M ScRad51 protein for 15 min at 30°C in 15 l of the ATPase reaction buffer. Then 2.5 units of wheat germ topoisomerase (Promega) were added and the reaction was continued for another 15 min. Subsequently, samples were deproteinized and analyzed on a 1.7% agarose gel in 0.5 ϫ TBE buffer containing chloroquine at 1 g/ml.
Electron Microscopy of Nucleoprotein Complexes-DsDNA-ScRad51 complexes were formed at 30°C for 15 min in the ATPase reaction buffer (omitting the bovine serum albumin and ATP regeneration system) and then fixed with 0.25% glutaraldehyde. Spreading and negative staining were performed as described (42). Specimens were subjected to transmission electron microscopy on CM-120 (Philips Electron Optics) at 80 kV tension. Electron micrographs were made at ϫ35,000 magnification using the low dose mode.

DsDNA-ScRad51 Complexes Enhance ScRad54
ATPase Activity-Binding of Rad51 protein to duplex DNA severely inhibits DNA strand exchange in vitro (39). The assembly of the presynaptic filament requires a stoichiometric DNA/protein ratio (3 bp/Rad51) under which an appreciable fraction of the Rad51 may evade binding ssDNA and may bind to dsDNA. This could explain, at least in part, the intrinsically slow rate of Rad51-mediated DNA strand exchange. We investigated whether Rad54 interacts with Rad51-covered duplex DNA and how this interaction may be related to the stimulatory effect of Rad54 protein on Rad51-mediated homologous recombination. Because the dsDNA-dependent ATPase activity is vital for the cellular function of the Rad54 protein and this activity is significantly increased upon Rad54 interaction with the presynaptic filament in vitro, we monitored the ScRad54 ATPase on dsDNA-ScRad51 complexes. Single-point measurements of hydrolyzed ATP were used to compare overall efficiencies of the Rad54 ATPase with various substrates and orders of addition. As observed previously (25,26), ssDNA-ScRad51 complexes stimulated the Rad54 ATPase (Fig. 1A). The addition of ScRad51 filaments preformed on short (100-mer) or long (X174) ssDNA at the saturating molar ratio of 3 nucleotides/ Rad51 monomer increased the Rad54 ATP hydrolysis to 140 -150% (Fig. 1A). Direct addition of subsaturating amounts of ScRad51 (300 bp/monomer) to dsDNA, when ssDNA was omitted from the reaction, resulted in an even greater increase of Rad54 ATPase activity (about 170%). With saturating ScRad51 amounts added to dsDNA (3 bp/monomer), partial inhibition of the Rad54 ATPase activity was observed, down to 75-80% of its activity on protein-free DNA.
To verify that the observed stimulatory effect on the Rad54 ATPase activity is not because of any particular DNA sequence, size, or topological state, the ATPase assays with the subsaturating bp/Rad51 ratio (300:1) were repeated using a set of duplex DNA substrates (Fig. 1B). The stimulation of the Rad54 ATPase was observed with all DNA substrates tested, including linear, nicked circular, and supercoiled duplex DNA. To rule out the possibility that the dsDNA substrates contained minute amounts of ssDNA, to which ScRad51 protein might preferentially bind and activate Rad54, supercoiled DNA was treated with E. coli exonuclease III followed by chromatography on benzoylated naphthoylated DEAE-cellulose. This treatment should quantitatively remove all potential ssDNA species including partial duplex molecules with single-stranded ends or gaps (43). No differences in the stimulation of the Rad54 ATPase using dsDNA substrates before and after this treatment were found (data not shown). Taken together, these data suggest that the activation of the Rad54 ATPase is a property common to dsDNA-ScRad51 complexes formed at a subsaturating ratio. Subsequently, this mode of Rad54 activity is referred to as the enhanced ATPase mode of Rad54 on partial dsDNA-Rad51 filament. We refer to the ATPase activity of Rad54 on protein-free dsDNA as the basic mode.
Previous studies indicated that the order of addition of individual components critically affects the progression of DNA strand exchange. The highest stimulation is achieved when Rad54 is added after ssDNA-ScRad51 filament formation, but prior to or soon after duplex DNA addition (26,32). To test whether the stimulatory effect of ScRad51 protein on the Rad54 ATPase is because of Rad54 interaction with the established dsDNA-ScRad51 filament, we performed a set of experiments with different orders of addition. The highest ATPase stimulation at the 45-min time point was achieved on preformed dsDNA-ScRad51 filaments. When protein-free duplex DNA was reacted with Rad54 prior to ScRad51 addition, the stimulatory effect was essentially absent ( Fig. 2A). As addressed below, we believe the absence of stimulation at this time point reflects a kinetic impediment. Incubation on ice of ScRad51 with Rad54 followed by dsDNA addition had an intermediary effect, whereas a similar incubation at 30°C resulted in a slight decrease of the Rad54 ATPase activity. This latter decrease is most likely because of the extreme temperature sensitivity of Rad54 in the absence of DNA (32). Taken together, these data suggest that to be activated to the enhanced ATP hydrolysis mode, the unbound Rad54 protein has to come from solution to the established dsDNA-Rad51 filament.
The single-point analysis of ATP hydrolysis (Figs. 1 and 2A) does not allow evaluating possible differences in the kinetics of the reaction, which may be particularly pertinent in the order of addition experiments. For a more detailed analysis of the Rad54 ATPase stimulation, the order of addition experiment depicted in Fig. 2A was repeated using the ATP/NADH-coupled assay, which allows monitoring the kinetics of ATP hydrolysis (see details under "Experimental Procedures"). First, we wanted to determine whether the stimulation of Rad54 protein by dsDNA-Rad51 complexes had a transient or continuing character. The Rad54 ATPase on protein-free dsDNA and on partial dsDNA-ScRad51 complexes was progressively inactivated with time (30 -40% loss of activity over 1800 s, Fig. 2B). To compensate for this inactivation effect, the rates of Rad54 ATPase in the presence of Rad51 protein were normalized to the Rad54 ATPase rates on protein-free DNA at the corresponding time points and expressed as percentage of Rad54 ATPase stimulation (Fig. 2C). With Rad54 protein added to the preformed subsaturated dsDNA-ScRad51 filament (optimal order of addition, triangles) the stimulation of the Rad54 ATPase was continuous and remained at maximal level. When Rad54 protein had been added to dsDNA prior to establishing the dsDNA-ScRad51 nucleoprotein filament (Rad54 added first to DNA or added together with ScRad51), a delay in the Rad51dependent stimulation of the Rad54 ATPase was observed. At limited amounts of ScRad51 protein (300 bp/Rad51), this kinetic delay probably reflects first, the time required for the formation of dsDNA-Rad51 complexes; and second, the time required for Rad54, which had bound protein-free DNA, to dissociate from it and find a Rad51 filament-containing DNA ATPase. Bar 1, control reaction: dsDNA (herring sperm) and Rad54 protein with no further additions. Bars 2 and 3, effect of short and long ssDNA-ScRad51 filaments: dsDNA was incubated for 15 min at 30°C with saturated filaments (3 nucleotides/Rad51 monomer) preformed on short (100-nucleotide oligonucleotide) and long (X174) ssDNA followed by the addition of Rad54 (equimolar amounts of Rad51 and Rad54 proteins). Bars 4 -6, effect of saturating and nonsaturating amounts of ScRad51 protein bound to dsDNA in the absence of ssDNA: bar 4, ATP hydrolysis by saturated dsDNA-ScRad51 complex (3 bp/Rad51 monomer), no Rad54 protein was added; bar 5, Rad54 was added to saturated dsDNA-ScRad51 complexes (3 bp/Rad51 monomer at a molar ratio of 100:1 Rad51:Rad54); bar 6, Rad54 was added to nonsaturated dsDNA-Rad51 complexes (300 bp/Rad51 monomer at a molar ratio of 1:1 Rad51: Rad54). All reactions contained 4.6 M dsDNA (herring sperm). The ScRad54 protein concentration was 15.3 nM; the ssDNA concentration was 46 nM. Complexes of ScRad51 protein with ssDNA and dsDNA were preformed for 15 min at 30°C. All reactions with Rad54 protein were carried out for 30 min at 30°C. Values of ATP hydrolysis were calculated from single-point measurements using the charcoal-based ATPase assay. B, effect of ScRad51 protein complexes with dsDNA of different lengths, sequence, and topology on the Rad54 ATPase. All reactions contained 4.6 M dsDNA, 15.3 nM ScRad51, and 15.3 nM ScRad54. DsDNA-ScRad51 complexes were preformed as in A. All reactions with Rad54 were carried out for 45 min at 30°C. The ATPase activity was determined as described in A. Open bars, protein-free DNA; filled bars, DNA-Rad51 complexes at 300 bp/Rad51 monomer stoichiometry. molecule. In similar experiments performed with longer incubation times, the Rad54 ATPase at nonoptimal order of addition gradually attained a level of the stimulation at optimal order of addition (data not shown).
Optima of the Enhanced Mode of ScRad54 ATPase on DsDNA-ScRad51 Complexes-To determine the optimal conditions for stimulating the Rad54 ATPase activity, we performed a set of titration experiments at a fixed dsDNA/Rad54 ratio using varying amounts of ScRad51 protein to form filaments of different saturation. The optimal DNA/Rad51 ratios fell within a narrow range of 20 -50 bp/Rad51 monomer that corresponds to filaments occupying 6 -15% of a dsDNA molecule (Fig. 3A).
The titration experiments performed with two different duplex DNA substrates (supercoiled pUC19 and dsX174) show very similar profiles.
It has recently been shown that the greatest stimulation of ScRad51 pairing activity by Rad54 protein in the D-loop assay between supercoiled dsDNA and short single-stranded oligonucleotides is achieved at equimolar ratio between the two proteins (26). To determine whether changes in the Rad51/Rad54 protein ratio would affect the enhancement of the Rad54 ATPase activity, we performed a titration experiment using a fixed amount of Rad54 protein and varying amounts of dsDNA-Rad51 filament with the optimized stoichiometry of 25 bp/ Rad51 monomer (see Fig. 3A). To normalize for the changes in the dsDNA/Rad54 ratio, a control titration was performed in the absence of Rad51 protein. The Rad54 ATPase activity on protein-free duplex DNA reached a saturation plateau within 75-150 bp per Rad54 protein (Fig. 3B, inset). A similar profile for the dsDNA-ScRad51 filament-stimulated reaction was observed, but with a 5-6-fold increase of ATPase activity at saturation within 150 -300 bp per Rad54. It appears that the limiting factors of the stimulation of Rad54 ATPase activity are the stoichiometry of the dsDNA-Rad51 complex (Fig. 3A) and the amount of dsDNA-Rad51 filament, but not the molar ratio between Rad51 and Rad54 proteins (Fig. 3C). At nearly equimolar ratios between Rad51 and Rad54 proteins, the Rad54 ATPase activity was far from its optimum (left most bars in Fig. 3C).
Partial DsDNA-ScRad51 Filaments Are the Active Species for the Enhanced Mode of ScRad54 ATPase-Unlike RecA protein, hRad51 protein exhibits low cooperativity of filament formation on dsDNA resulting in the prevalence of multiple short filaments over long filaments at a subsaturating DNA/protein ratio (44). To analyze the cooperativity of dsDNA binding of ScRad51 protein and to determine an average filament length responsible for the stimulation of the Rad54 ATPase, we employed a topological DNA unwinding assay and direct visualization by transmission electron microscopy. The DNA unwinding assay allows an estimate of the Rad51-covered portions of DNA molecules (41). Rad51 protein bound to the covalently closed duplex DNA partially unwinds it, introducing compensatory supercoils, which are removed upon addition of eukaryotic topoisomerase I. After deproteinization, the resulting linking number change of the unwound species can be determined using topological gels. At the 25 bp/Rad51 ratio, absence of the relaxed species and a broad distribution of slightly to moderately unwound species (Fig. 4A) indicates that virtually all DNA molecules contained patches of Rad51 filaments. No fastmigrating band of highly underwound DNA (so-called form X) corresponding to DNA species extensively covered with the Rad51 protein was detected. Transmission electron microscopy examination showed the predominance of short rather than long filaments or fully coated molecules at a 25 bp/Rad51 ratio (Fig. 4B). Similarly to DNA-RecA complexes, DNA-ScRad51 filaments exhibit a regular pattern seen by transmission electron microscopy as striations in rotary-shadowed specimens and a zigzag pattern in negatively stained specimens (45). By counting the number of zigzags and multiplying that number by 19, the number of bp per helical repeat in the Rad51 filament (46), the lengths of five representative filaments were determined (Fig. 4C). These measurements suggest that the filaments ranged from about 100 to 300 bp in size. This indicates a relatively low binding cooperativity of the ScRad51 protein. It appears that Rad54 protein acts in the enhanced ATP hydrolysis mode on duplex molecules containing single or multiple patches of relatively short Rad51 filaments.

ScRad54 Protein Exerts Different Modes of Interaction with Partial and Saturated DsDNA-ScRad51 Filaments-The
ATPase activity of Rad54 protein on protein-free duplex DNA reaches a plateau at a stoichiometry 75-150 bp/Rad54 (Fig.  3B). This points to the requirement that a Rad54 molecule needs at least about 100 bp of DNA for optimal ATPase activity. were reacted with fixed amounts of ScRad54 protein (15.3 nM). Relative values of ATP hydrolysis were calculated from single-point measurements using the charcoal-based ATPase assay after 45 min of incubation with Rad54 protein. All incubations were done at 30°C. The amount of complexes is expressed as bp/ Rad54 ratio. Diamonds, titration in the absence of Rad51 protein; circles, titration with dsDNA-Rad51 complexes at the 25 bp/Rad51 stoichiometry. Inset, data were re-plotted in the appropriate scales to show the plateau of ATPase activity in the absence of Rad51 more clearly (diamonds). C, filament titration data presented in B were re-plotted as a bar graph. Bar graph corresponds to the right y axis. Open bars, control titration of dsDNA/Rad54 ratio, in the absence of Rad51 protein; filled bars, dsDNA-Rad51 filament. Line, fold stimulation of Rad54 ATPase (left y axis). Data were normalized to the corresponding control titration with protein-free DNA.
Rad54 is known to form dimers or oligomers on dsDNA (23,47). Thus, the required minimum of a stoichiometry of 100 bp/ Rad54 corresponds to 30 or less Rad54 oligomers per plasmidsize DNA substrate. A similar DNA/Rad54 stoichiometry is required for Rad54 protein in the enhanced ATPase mode stimulated by partial dsDNA-Rad51 filaments (see Fig. 3B). At the same time, the greatest ATPase activation is achieved at subsaturating amounts of Rad51 protein when larger parts of the dsDNA molecule are protein-free. All of the above implies that the active species for the enhanced mode of the Rad54 ATPase are duplex DNA molecules nucleated with short patches of Rad51 filaments rather than the filaments themselves. The suboptimal stimulatory effect on Rad54 ATPase at very low Rad51/dsDNA ratios can be interpreted as a lack of such species because at these ratios only a fraction of the DNA molecules will contain filaments. The 300 bp/monomer ratio in Fig.  3A corresponds to an average of 1% of the DNA to be covered with 18 Rad51 protomers per DNA molecule. To verify whether the number of filament-containing dsDNA molecules is a limiting factor for the Rad54 ATPase in the enhanced mode, we performed a filament dose titration experiment, similar to that depicted in Fig. 3B, with filaments formed at 300 bp/ScRad51 monomer. We took advantage of the ATP/NADH-coupled assay measuring real-time kinetics of ATP hydrolysis and used this technique in all subsequent experiments. For the sake of improved Rad54 stability, we also lowered the incubation temperature for Rad54 protein to 24°C. The titration revealed a filament-dependent character of the Rad54 ATPase stimulation (Fig. 5A). These results also suggest a higher affinity of Rad54 protein to filament-containing molecules versus proteinfree dsDNA present in the mixture. Interestingly, in the two filament titration experiments performed at different Rad54 concentrations and dsDNA/Rad51 ratios (Figs. 3B and 5A) a molar excess of Rad51 over Rad54 was always required for the saturation of the enhanced mode of Rad54 ATPase activity (compare 147-298 bp/Rad54, 6.4 -12.8 Rad51/Rad54, at a fixed concentration of 15.6 nM Rad54 in Fig. 3B with 920 -1840 bp/Rad54, 3-6 Rad51/Rad54, at a fixed concentration of 5 nM Rad54 in Fig. 5A). In the latter case, the lower excess of Rad51 front of the relaxed species. DsDNA-Rad51 complexes at the 25 bp/ ScRad51 monomer ratio treated with eukaryotic topoisomerase I (right lane) appear as broad distribution of slightly to moderately unwound species; their average linking number change corresponds to the unwinding induced by short filament patches (less than 400 bp of total Rad51-coated DNA per plasmid), see B and C. Note that topoisomers corresponding to Rad51-free DNA (middle lane) or extensively covered species (form X) are virtually absent. B, transmission electron microscopy visualization of dsDNA-ScRad51 complexes formed at a stoichiometry of 25 bp/Rad51 monomer. Complexes were formed between supercoiled ds X174 DNA and ScRad51 protein in the buffer used for the ATP hydrolysis assay in which bovine serum albumin and the ATP regeneration system were omitted. Incubation was done at 30°C for 15 min followed by fixation in 0.25% glutaraldehyde. Specimens were prepared and subjected to transmission electron microscopy as described under "Experimental Procedures." Please note that negative staining conditions only provide a low contrast for visualizing DNA as  Duplex DNA is thought to be hidden inside the Rad51 fila-ment and thus protected from nucleases (33,36). Rad54 protein requires duplex DNA as a co-factor of its ATPase activity. Thus, it was of interest to test, if covering the dsDNA molecule with saturating amounts of Rad51 protein would lead to inhibition of the Rad54 ATPase activity. When dsDNA was covered with ScRad51 at a saturating ratio of 3 bp/Rad51 monomer, we observed the Rad54 ATP hydrolysis at 80% of the original rate on protein-free dsDNA (Fig. 1A, bar 5). We rationalized that the Rad54 ATPase activity observed with saturated dsDNA-ScRad51 complexes may be a composite of several effects. First, the dsDNA-ScRad51 complexes formed at saturating ratios may still contain a small number of partially covered dsDNA molecules. Because the Rad54 ATPase is greatly enhanced on partial filaments, all observed hydrolysis may be due to these minor species of partial filaments, whereas fully covered molecules do not support the Rad54 ATPase at all. Second, Rad54 protein may have a specific mode of interaction with completely Rad51-covered dsDNA having an inherent rate of ATP hydrolysis. Third, both processes may contribute simultaneously to the observed ATP hydrolysis. To distinguish between these possibilities and determine whether the Rad54 ATPase could be affected by increasing the number of minor species with partially covered duplex DNA, we performed a filament dose titration similar to that in Fig. 5A using saturated dsDNA-ScRad51 filaments. In contrast to the effect with the partial filaments at the 300 bp/Rad51 ratio (Fig. 5A), titration of saturated filaments formed at 2 bp/Rad51 revealed no dose-dependent stimulation of the Rad54 ATPase (Fig. 5C). Therefore, partially covered dsDNA species if any, are unlikely to contribute to the observed Rad54 ATPase activity. The stable kinetics of ATP hydrolysis on saturated dsDNA-ScRad51 filaments (Fig. 5B) is most consistent with an inherent mode of Rad54 activity on this substrate. We refer to the ATPase activity on dsDNA saturated with ScRad51 as the reduced mode of Rad54. hRad51 Can Activate ScRad54 to the Enhanced Mode on Partial Filaments, but Fails to Support ScRad54 ATPase on Fully Covered dsDNA-Among the proteins of the RAD52 group, Rad51 protein of S. cerevisiae has very high speciesspecificity, as mutants are not complemented by Rad51 homologs of other organisms (48). Rad54 protein, in contrast, has retained some functional conservation throughout evolution, in that certain rad54 phenotypes in S. cerevisiae (methyl methane sulfonate and UV, but not IR sensitivity) are partially complemented by the human Rad54 cDNA (49). It was worthwhile to determine whether partial and saturated filaments of hRad51 on duplex DNA could affect the ATPase activity of yeast Rad54. Partial filaments formed by hRad51 protein at the 30 bp/Rad51 ratio were able to stimulate ScRad54 protein, although to a lower extent than ScRad51 (Fig. 6A). However, unlike with the ScRad51 protein, the saturated filament of hRad51 protein was completely unable to support any ATPase activity of ScRad54 protein. The ATPase activity of hRad51 alone accounts for the hydrolysis above background in Fig. 6A, panel h51 (3:1).
Rad51 filaments on ssDNA and dsDNA are known to be structurally very similar to those formed by RecA protein (46). Therefore, we wished to test whether the dsDNA-RecA filament may activate the Rad54 ATPase. It is known that RecA forms stable complexes with duplex DNA only under specific conditions (e.g. in the presence of ATP␥S, a poorly hydrolyzable ATP analog). Thus, we prepared dsDNA-RecA-ATP␥S complexes at a ratio of 30 bp/RecA monomer and then titrated these complexes in reactions with Rad54 protein. To normalize for possible adverse effects of ATP␥S on Rad54 protein, control reactions with the same ATP␥S concentrations were performed in the absence of RecA. Under the highest ATP␥S concentration tested only a moderate (20%) decrease of the Rad54 ATPase rate was seen (compare bars 6 and 7 on Fig. 6B). We detected no sign of RecA filament-dependent Rad54 ATPase stimulation. In fact, RecA filaments slightly inhibited the Rad54 ATPase compared with protein-free duplex (Fig. 6B). Our conclusion is that Rad54 protein is activated to the enhanced mode on the partial dsDNA-Rad51 filament via specific protein-protein contacts common for various Rad51 proteins but not for RecA.

DISCUSSION
Rad54 Protein Acts on DNA in a Catalytic Fashion Resembling ATP-dependent Chromatin Remodeling Complexes-Rad54 protein is a Snf2-like ATPase with a specialized function in recombinational DNA repair. Structural and functional features of the Rad54 protein closely resemble other members of the Snf2-like family, in particular ATP-dependent chromatin remodeling complexes. Rad54 protein and Snf2, the ATPase subunit of the prototype chromatin remodeling complex Swi/ Snf, both have high affinity to dsDNA independent of nucleotide cofactor, but require the presence of duplex DNA for ATPase activity (21,22,50). Structural studies reveal that the seven conserved "helicase" motifs of Snf2-like proteins are clustered together in space forming the nucleotide-binding pocket and a portion of the DNA-binding site (51). The ATPase is essential for the cellular function of Snf2-like proteins, with the exception of Ercc6/CS-B, where an ATPase-deficient mutant exhibited only a partial defect (52). Current evidence indicates that most, if not all, Snf2-like enzymes use ATP hydrolysis to introduce unconstrained superhelicity in DNA (DNA torsion activity). It was suggested that this activity mirrors a fundamental feature of these proteins to translocate along DNA altering the interaction of other proteins with dsDNA (20,25,53).
Generally, Snf2-like proteins do not form extensive nucleoprotein complexes resembling ssDNA-RPA or ssDNA/dsDNA-Rad51 complexes, but act catalytically with an effective stoichiometry of one/several proteins per DNA molecule. For instance, one molecule of Swi/Snf complex is sufficient to remodel plasmid DNA containing many nucleosomes (54). The optimum for DNA torsion activity of the Swi/Snf complex is achieved at similar protein per DNA molecule ratios (55). In our context, a "catalytic" mode of action of a DNA-binding protein would involve multiple cycles of activity on a DNA substrate (e.g. ATP hydrolysis-driven DNA remodeling), as opposed to a "stoichiometric" mode where the protein acts solely by virtue of one-step binding.
Measured optima of the ATPase and DNA torsion activities indicate that Rad54 protein acts on dsDNA in a catalytic fashion with an effective ratio corresponding to about one or few protein mono-or oligomers per DNA molecule. The minimal optimal per-base pair stoichiometry of the Rad54 ATPase was shown to be 30 bp/Rad54 for short duplex oligonucleotides (63-mer (26)) and 75-150 bp/Rad54 for longer dsDNA (5.4 kbp; this study). Given that Rad54 di/oligomerizes on dsDNA (23,47), this corresponds to an effective per molecule stoichiometry of 1 Rad54 dimer per short oligo and less than 30 Rad54 dimers per long DNA molecule. Excess Rad54 resulted in decreased rates of the Rad54 ATPase (Ref. 26 and this work) and torsion activities (25) suggesting that part of the protein cannot form active complexes with DNA. For instance, transient ssDNA regions or some noncanonical DNA structures arising from excessive topological strain may sequester Rad54 protein via unproductive binding. It is unlikely that Rad54 protein has an additional "structural" role uncoupled from its ATPase activity, because DNA binding-proficient, but ATPase-deficient RAD54 mutants have cellular phenotypes similar to those of the deletion mutant (23,27) and retain essentially none of the known in vitro protein activities (21)(22)(23)(24)(25)(26). On the whole, the effective per molecule stoichiometry of dsDNA-Rad54 complexes sustaining optimal protein activities in vitro reflects the catalytic action of Rad54 in vivo where it acts in an ATP hydrolysisdriven force-generating process and not simply by one-step DNA binding.
Direct visualization of the Swi/Snf complex and Rad54 protein on DNA is in good agreement with the biochemical data. Bazett-Jones and colleagues (56) found DNA-bound Swi/Snf complexes with a typical morphology of one to three protein globules per plasmid DNA. These complexes frequently created loops on DNA that are indicative of torsion activity. Recently, where ATP␥S was omitted. Rad54 protein was at 5 nM concentration. ATPase rates were measured at 24°C using the NADH/ATP-coupled assay.
Ristic and colleagues (47) observed by scanning force microscopy DNA-bound complexes of human Rad54 protein formed at a ratio of 112 bp/protein. They found that each 1.8-kb DNA circle contained only one or sometimes two large complexes consisting of presumably 3 or more Rad54 monomers (47). Similarly to the Swi/Snf complex, Rad54 protein created loops in DNA.
Rad54 Protein Switches from a Basic to an Enhanced Mode of Activity on Partial dsDNA-Rad51 Filament-During the search of homology, the presynaptic ssDNA-Rad51 filament is thought to form a tertiary complex with Rad54 protein. This interaction dramatically increases the ATPase and torsion activities of Rad54 on dsDNA (26). However, it was noted that Rad54 activities on dsDNA were also stimulated by Rad51 protein when ssDNA-Rad51 filament was omitted (25). This stimulation could be attributed to an effect of dsDNA-Rad51 complex and/or to Rad51 filament formed on ssDNA, which may be present in dsDNA preparations. Here we demonstrate that Rad54 protein has an inherent activity on dsDNA partially covered with Rad51 and that the nucleoprotein complex, but not free Rad51 protein, is responsible for the stimulation of the Rad54 ATPase. It was previously shown that Rad51 and Rad54 proteins could interact (6,7). Our data, together with evidence of the tertiary complex of Rad54 protein and the presynaptic filament (21,26,32), suggest that productive interaction between the two proteins during DNA strand exchange occurs in the context of the nucleoprotein complex.
Interaction with DNA-binding proteins and alteration of DNA-protein contacts is a distinguishing feature of Swi2/Snf2like proteins. All known ATP-dependent chromatin-remodeling complexes possess at least one subunit capable of interacting directly with chromatin. The Swi/Snf complex destabilizes DNA-histone contacts in nucleosomes and is capable of shifting or even completely removing them from DNA. Swi/Snf can distinguish nucleosomal structures from nonspecific complexes of DNA with histones and is unable to remodel the latter (57). The enhanced mode of the Rad54 ATPase also shows specificity for Rad51 protein from eukaryotic organisms compared with bacterial RecA protein. We propose that Rad54 in the enhanced mode acts as a remodeling enzyme specific for complexes of Rad51 protein on duplex DNA. One possible outcome of the remodeling is removal of the Rad51 filament from duplex DNA coupled with translocation of Rad54 protein along filamentcontaining molecule. 2 The energy required to dissociate the dsDNA-Rad51 complex would account for the increased ATP hydrolysis and DNA torsion activities of Rad54 protein.
Saturated Nucleoprotein dsDNA-ScRad51 Filaments Sustain a Reduced Mode of ScRad54 ATPase-Because the availability of protein-free dsDNA is crucial for the Rad54 ATPase in the basic and enhanced modes, we anticipated that ATP hydrolysis by Rad54 on fully covered dsDNA-ScRad51 complexes would be severely inhibited. However, even at a saturating ratio (2 bp/Rad51 monomer) Rad54 protein retains over 50% of its activity of the basic mode. The observed effect was unlikely because of a preferential interaction of Rad54 protein with a minor fraction of partially ScRad51-covered DNA. These potential minor species are clearly ruled out in the case of dsDNA covered by saturating amounts of hRad51 protein (Fig.  6A) on which the Rad54 ATPase is severely inhibited. Thus, ATP hydrolysis on saturated dsDNA-ScRad51 filaments is likely to represent a distinct, reduced mode of Rad54 ATPase independent on protein-free duplex DNA.
Several lines of evidence suggest that this reduced mode may be of biological relevance and not an epiphenomenon of the in vitro reaction. The kinetics of Rad54 on saturated dsDNA-ScRad51 filament shows a continuous ATPase activity with a rate comparable with that of Rad54 on protein-free duplex DNA (60 -80%). The interaction of Rad54 protein with saturated dsDNA-ScRad51 filament also clearly differs from its interaction with the presynaptic filament because the latter does not support the ATPase activity of Rad54, and because of the absence of dsDNA (26). Moreover, the reduced mode appears to be species-specific, as ScRad54 protein is completely inactive on saturated filaments formed by hRad51 protein.
Recently, several potential sites of interaction of yeast Rad51 protein with Rad52, Rad54, and Rad55/57 proteins were identified (58). The majority of amino acid residues affecting the Rad51-Rad54 interaction is highly conserved between human and yeast Rad51, with the exception of the NH 2 terminus, which is absent in the hRad51 protein. This may suggest that yeast Rad51 and Rad54 interact through multiple contacts that are not all conserved in the human proteins. The speciesspecificity of the Rad54 ATPase activity in the reduced mode suggests that the amino-terminal domain of the ScRad51 protein is important for this mode. Because the Rad54 ATPase is inactive on ssDNA-Rad51 filaments (26), Rad54 protein probably has access to dsDNA in the nucleoprotein filament.
Stimulation of Homologous Pairing by Rad54 Protein Is Accomplished via Its Diverse Modes of Activity on Protein-free DNA and on DNA-Rad51 Complexes-The proper and timely assembly of DNA-protein complexes is essential for homologous recombination. In presynapsis, the mediator proteins Rad52 and Rad55/Rad57 coordinate the disassembly of the ssDNA-RPA complex and assembly of the presynaptic ssDNA-Rad51 filament. Evidence from genetic, cytological, and biochemical studies points to a later action of Rad54 protein in recombination, after the presynaptic phase (see Introduction). We hypothesize that Rad54 protein accomplishes its manifold action in DNA strand exchange via interactions with Rad51 complexes on ss-and dsDNA (Fig. 7).
Rad54 is unable to interact productively with protein-free ssDNA. However, a tertiary complex formed between the Rad54 and presynaptic ssDNA-Rad51 filament is of functional significance during the synaptic phase of recombination. It facilitates the search of homology by a mechanism that probably involves Rad54 translocating on the dsDNA lattice. Whereas Rad54 is presumed to remain in tight contact with the presynaptic filament until homologous pairing is established, the interaction of the tertiary complex with dsDNA during homology sampling is likely to proceed via multiple association/ dissociation cycles. On protein-free duplex DNA, Rad54 has basic ATPase and DNA torsion activities. The direct binding of Rad51 to dsDNA switches the Rad54 ATPase from the basic to the enhanced mode during which Rad54 may act as a remodeling/removal factor for partial dsDNA-Rad51 filament. Complete saturation of dsDNA-Rad51 filament switches Rad54 to the reduced mode of ATPase activity.
In vivo, human RAD54 cDNA partially complements some phenotypes of rad54⌬ mutants, suggesting that the human Rad54 homolog can productively act in concert with the yeast recombination machinery. In the heterologous in vitro system, ScRad54 protein fails to stimulate synaptic joint formation by hRad51 (25), 2 and is inactive on saturated dsDNA-hRad51 filament, but still retains the enhanced ATPase mode on partial dsDNA-hRad51 filament. Because the enhanced ATPase mode is not sufficient to sustain the synaptic phase of recombination, we envision two mutually nonexclusive possibilities. First, the synaptic phase may necessitate additional Rad54 activity(ies) (for instance, the ATPase on saturated dsDNA-Rad51 filament). Second, the enhanced mode may be required only later in the postsynaptic phase. In fact, during the presynaptic and early synaptic stages of recombination, both donor and acceptor DNA strands may or may not contain dsDNA-Rad51 filaments, but later it is highly likely that Rad51 is bound to the product heteroduplex DNA (by analogy to RecAmediated pairing reactions). The enhanced mode of Rad54 ATPase may be then important for recycling Rad51 on heteroduplex DNA and/or its repartition between ss-and dsDNA species. It is conceivable that Rad54 protein released from the presynaptic filament can be redirected to dsDNA-Rad51 complexes. Transition of modes of Rad54 ATPase activity following changes in the stoichiometry of dsDNA-Rad51 complexes pro-vides a further clue how the function of Rad54 protein during recombination may be coordinated.