Vital Roles of the Second DNA-binding Site of Rad52 Protein in Yeast Homologous Recombination*

RecA/Rad51 proteins are essential in homologous DNA recombination and catalyze the ATP-dependent formation of D-loops from a single-stranded DNA and an internal homologous sequence in a double-stranded DNA. RecA and Rad51 require a “recombination mediator” to overcome the interference imposed by the prior binding of single-stranded binding protein/replication protein A to the single-stranded DNA. Rad52 is the prototype of recombination mediators, and the human Rad52 protein has two distinct DNA-binding sites: the first site binds to single-stranded DNA, and the second site binds to either double- or single-stranded DNA. We previously showed that yeast Rad52 extensively stimulates Rad51-catalyzed D-loop formation even in the absence of replication protein A, by forming a 2:1 stoichiometric complex with Rad51. However, the precise roles of Rad52 and Rad51 within the complex are unknown. In the present study, we constructed yeast Rad52 mutants in which the amino acid residues corresponding to the second DNA-binding site of the human Rad52 protein were replaced with either alanine or aspartic acid. We found that the second DNA-binding site is important for the yeast Rad52 function in vivo. Rad51-Rad52 complexes consisting of these Rad52 mutants were defective in promoting the formation of D-loops, and the ability of the complex to associate with double-stranded DNA was specifically impaired. Our studies suggest that Rad52 within the complex associates with double-stranded DNA to assist Rad51-mediated homologous pairing.

diates of homologous recombination. D-loop formation is the base pairing of an invading single-stranded DNA (ssDNA) 3 tail, derived from one end of the DNA double-stranded break, with its complementary sequence in an internal region of a homologous double-stranded DNA (dsDNA), which serves as the template to repair the break. RecA of Escherichia coli is the prototype of homologous pairing proteins that catalyze ATPdependent D-loop formation (5,6), and Rad51 is an orthologue of RecA in eukaryotes (7)(8)(9). Following the D-loop formation, RecA/Rad51 promotes the extension of a core heteroduplex joint by displacing a strand of the parental dsDNA with the invading ssDNA (branch migration) in an ATP hydrolysis-dependent reaction (5,10,11). D-loop formation is observed in all domains of life. RecA catalyzes this reaction in prokaryotes (5,6), RadA catalyzes the reaction in Archaea (12), and Rad51 and Dmc1 catalyze the reaction in eukaryotes (8,(13)(14)(15)(16)(17)(18).
When an ssDNA region is formed after double-stranded breakage, the ssDNA is covered by a single-stranded binding protein (SSB) in bacteria and replication protein A (RPA) in eukaryotes. The binding of SSB or RPA to ssDNA stimulates, or is required for, D-loop formation by RecA or Rad51 to unfold the secondary structures of the ssDNA, especially when the ssDNA is long enough to fold tightly (8,19,20). However, the prior binding of SSB or RPA to ssDNA inhibits the initial binding of RecA or Rad51. The cells have a family of proteins called recombination mediators, which load RecA or Rad51 onto ssDNA by overcoming the inhibitory effects of SSB or RPA (21,22). Rad52, which is essential for homologous recombination in Saccharomyces cerevisiae, is the prototype of the mediators (21) and is ubiquitous in various eukaryotes, except for plants and flies. The yeast Rad52 physically interacts with RPA and Rad51 and facilitates the displacement of RPA by Rad51 (23,24). In addition to its mediator function, Rad52 also possesses potent ssDNA annealing activity in vitro (25). This activity is believed to be important for capturing the second end of the doublestranded break. Precisely, after the 3Ј ssDNA tail derived from the first end of the double-stranded break has been incorporated into a D-loop and repair DNA synthesis initiated at the 3Ј termini enlarges the D-loop, the second end is annealed with the displaced strand of the D-loop. Thus, Rad52 is considered to primarily function on ssDNA.
Although Rad52 has been shown to bind to dsDNA, the functional relevance of this activity is unknown. A crystallographic study combined with mutational analyses of human Rad52 revealed two distinct DNA-binding sites (26,27). The first site, located inside the groove that runs around the ring structure, is a binding site for ssDNA. The second site is located at the entrance of the groove, and is a binding site for dsDNA or ssDNA. In the first site, Arg-55 is the key residue for ssDNA binding. Mutating this residue greatly impairs the ssDNA binding activity of human Rad52. The corresponding residue in yeast Rad52 is Arg-70, and in vivo studies revealed homologous recombination deficiencies in yeast when this residue was mutated (28,29). Subsequent studies suggested that Arg-70 is important for the annealing activity of Rad52, which is required for the second end capture of the double-stranded break (30 -33). On the other hand, the second DNA-binding site of human Rad52 is composed of Lys-102 and Lys-133, and in vitro studies revealed that these residues are important for the ssDNA annealing activity of Rad52. However, no clear in vivo defects have been demonstrated yet for this DNA-binding site.
We previously reported that the yeast Rad51 and Rad52 proteins form a stoichiometric 2:1 complex and that the complex promotes D-loop formation efficiently (34). The D-loop formation promoted by the complex does not appear to involve the mediator and ssDNA annealing activities of Rad52. This observation led us to investigate the possible role of the putative second DNA-binding site of yeast Rad52 in the D-loop formation activity of the complex. In the current study, we found that yeast Rad52 also harbors the second DNA-binding site identified in human Rad52. The DNA-binding site was important for homologous recombination in yeast cells and was indispensable for D-loop formation by the Rad51-Rad52 complex. Based on these results, we discuss the possible roles of the second DNA-binding site of Rad52 in yeast.

MATERIALS AND METHODS
Enzymes and Reagents-Taq DNA polymerase, T4 polynucleotide kinase, calf intestine alkaline phosphatase, and all of the restriction endonucleases were purchased from Takara Shuzo Co., Ltd. Pfu polymerase was obtained from Promega Co. Hydroxyapatite (Bio-Gel HTP) was purchased from Bio-Rad. The micro Spin S-400 HR column, the Probe Quant G-50 Micro Column, and the Q-Sepharose FF, Mono Q HR5/5, and Sephacryl S-300 HR matrices were obtained from GE Healthcare Biosciences.
DNA-DNAs used in these experiments were prepared as described previously (34). The DNA concentrations are expressed as the concentrations of nucleotide residues in the DNA. The following are the details of each DNA species.
Negatively Supercoiled dsDNA (Form I DNA)-We prepared pUC119 and pNS11 form I DNA by the previously described methods (35), which included the disruption of cells with lysozyme and sarcosyl treatments, phenol extraction, and purification by sucrose density gradient centrifugation. Note that, to avoid the induction of non-B form structures in negatively supercoiled DNA and to minimize contamination and damage, the purification procedure for the form I DNA was devoid of denaturation, gel electrophoresis, and contact with intercalators, such as ethidium bromide. The pNS11 plasmid is a pUC119 derivative containing ARG4 and part of the DED81 gene of S. cerevisiae (34).
pUC119 Circular ssDNA-We prepared pUC119 circular DNA by the previously described methods (35).
Preparation of rad52K117A/R148A and rad52K117D/ R148D ORF-Both the Lys-117 and Arg-148 residues (numbering from the first ATG) of the wild type Rad52 on pNS145 were replaced by alanine (K117A/R148A) or aspartic acid (K117D/ R148D), using a QuikChange site-directed mutagenesis kit (Stratagene). The replacement of both the AAG (Lys-117) and AGA (Arg-148) codons with the GCG (alanine) or GAC (aspartic acid) codon was confirmed by DNA sequence analyses.
Construction of the Plasmid for Rad52 Expression in Yeast Cells-The expression vector for the mutant and wild type Rad52 in S. cerevisiae, containing the ADH1 promoter and terminatory, ARSH4-CEN6 (a yeast centromere sequence and autonomously replicating sequence) (37) and the LEU2 marker (named pNS31), was constructed by inserting the BamHI fragment containing the ADH1 promoter and the terminator of pAUR123 (Takara Shuzo Co., Ltd.) into the PvuII site of pRS415 (Stratagene). The BamHI fragment was blunt-ended by a treatment with Klenow fragment before insertion. The open reading frame of wild type RAD52 (pNS145), rad52K117A/R148A (pNS178), or rad52K117D/R148D (pNS179) on pET3a for expression in E. coli (see below) was removed by EcoT14I digestion and partial digestion with NdeI and then blunt-ended by a treatment with Klenow fragment. The fragment containing the open reading frame of rad52 was inserted in the SmaI site of pNS31. The cells of the rad52⌬ strain (XS560-1C-1D2) were transformed with the plasmids.

dsDNA Binding to Rad52 in Homologous Recombination
Spot Test for Methyl Methanesulfonate (MMS) Sensitivity of rad52⌬ Transformants Expressing Mutant rad52-The SD liquid medium (2% glucose and 0.67% yeast nitrogen base without amino acid (Difco)) for the spot test was supplemented with 2 g/ml uracil and 2 g/ml histidine. Cells from an overnight culture (10 ml) were concentrated to 1 ml by centrifugation at 200 ϫ g for 3 min. The number of cells was counted under a microscope using an improved Neubauer hemocytometer and was adjusted to ϳ10 5 cells/l. Aliquots (10 l) of a 10-fold dilution series of each transformant were spotted onto SD plates containing 2 g/ml uracil and 2 g/ml histidine with 0.25, 0.59, or 1.18 mM MMS. The plates were sealed with Parafilm and incubated at 30°C for 5 days.
Quantitative Tests for MMS Sensitivity of rad52⌬ Transformants Expressing Mutant rad52-The SD liquid medium for quantitative analysis was supplemented with 4 g/ml adenine sulfate, 2 g/ml uracil, 2 g/ml histidine, 4 g/ml lysine, and 4 g/ml tryptophan (and 6 g/ml leucine for YPH499 strain without the plasmid)). Cells from an overnight culture (10 ml) were concentrated to 1 ml by centrifugation at 200 ϫ g for 3 min. After the dilution of the cell cultures, the cells were spread on freshly prepared SD plates containing the appropriate supplements, with MMS at 0, 0.13 (0.001%), 0.25 (0.002%), 0.59 (0.005%), or 1.18 mM (0.010%). The plates were sealed with Parafilm and incubated at 30°C for 7 days.
Purification of Mutant and Wild Type Rad52-The wild type Rad52 and the mutant proteins, rad52K191A/R148A and rad52K191D/R148D, were expressed from the third ATG codon of the cloned RAD52 on pET3a (pNS145) in E. coli BLR (DE3). Rad52 was purified as described previously (34), i.e. disruption by lysozyme treatment of the expressing cells, Brij58 treatment, ammonium sulfate precipitation, and fractionation by a series of column chromatography steps using SP-Sepharose FF, Sephacryl S300HR, and Mono Q. The purified Rad52 preparations were dialyzed against storage buffer (20 mM Tris-HCl, pH 7.5, 0.1 mM EDTA, 100 mM KCl, 10% glycerol, 1 mM DTT, and 20 M PMSF), and then 20-l aliquots were frozen by dipping into liquid nitrogen and stored at Ϫ80°C (final concentration, 162 M wild type Rad52; 131 M rad52K191A/R148A; and 101 M rad52K191D/R148D). The molar concentration of Rad52 was determined from the extinction coefficient, 2.42 ϫ 10 4 M Ϫ1 cm Ϫ1 at 280 nm, reported by New et al. (38). The mutant and wild type Rad52 does not have tags of any kinds.
Purification of Rad51-The wild type Rad51 (cloned in pET3a) proteins were prepared by a procedure consisting of expression in E. coli BLR (DE3), cell disruption by lysozyme and Brij58 treatments, polymin P precipitation, ammonium sulfate fractionation, and a series of column chromatography steps on Q-Sepharose FF, Hydroxyapatite, Sephacryl S300HR, and Mono Q, as described previously (34). The purified Rad52 preparation was dialyzed against storage buffer, and then 20 l-aliquots (final concentration 170 M) were stored at Ϫ80°C. The molar concentration of the Rad51 was determined from the extinction coefficient (1.29 ϫ 10 4 M Ϫ1 cm Ϫ1 at 280 nm) reported by Sugiyama et al. (39). The Rad51 does not have tags of any kind.
Purification of RPA-The natural form of RPA was purified from S. cerevisiae YPH501 (a/␣, ura3-52, lys2-801 amber, ade2-101 ochre, trp1-⌬63, his3-⌬200, leu2-⌬1) as described by Alani et al. (40), except the DEAE 650M step was replaced by a Mono Q step. The purified RPA was dialyzed against storage buffer containing 0.02% IGEPAL CA-630. The concentration of RPA was determined based on the extinction coefficient (8.8 ϫ 10 4 M Ϫ1 cm Ϫ1 at 280 nm) reported by Sugiyama et al. (39). We confirmed that the purified RPA stimulated Rad51-promoted strand exchange between circular ssDNA and linear dsDNA (data not shown). It is noted that the purified RPA is the authentic yeast RPA produced in yeast cells.
DNA Binding Assay-DNA (0.8 M Cy5-labeled pNS11 ssDNA 259-mer or 15 M pUC119 form I DNA) was incubated with Rad52 in the standard reaction buffer (10 l), except that it contained 20 mM KCl and lacked both MgCl 2 and ATP at 37°C for 10 min. The free and bound DNA were separated by 1% SeaKem GTG agarose (LONZA Group Ltd.) gel electrophoresis with TAE buffer and analyzed by Typhoon 9410 image analyzer for the Cy5-labeled pNS11 ssDNA 259-mer or a Gel Doc XR system (Bio-Rad) for the form I DNA, after ethidium bromide staining.
ATPase Assay-[␣-32 P]ATP (0.1 mM) was incubated with 0.88 M RPA and 5.0 M pUC119 circular ssDNA in the standard reaction buffer (10 l) but containing 7 mM MgCl 2 at 37°C for 10 min. After the incubation, 1.0 M Rad51 and the indicated concentrations of Rad52 were added to the reaction mixture, which was further incubated for 30 min. The reaction was terminated by adding 10 l of stop solution (25 mM EDTA, 3 mM ATP, 3 mM ADP, 3 mM AMP). Aliquots (15 l) of the sample were spotted on polyethyleneimine plastic film (POLYGRAM CELL 300 PEI for TLC; MACHERY-NAGEL) and developed in a mixture of 0.5 M lithium chloride and 1 M formic acid (41). After the plastic film was exposed to a Phosphor Screen (GE Healthcare Bioscience) for ϳ1 h, the radioactivity was analyzed by a Typhoon 9410 Variable Image Analyzer.
Immunoprecipitation with an Anti-Rad51 Antibody-The purified anti-Rad51 antibody (90 g) (34) was mixed with 330 l of protein A-agarose bead suspension (Calbiochem) in PBS with 0.05% IGEPAL CA-630 (total, 1.8 ml) and gently shaken at 4°C for 16 h. The beads were washed twice with 1 ml of 200 mM triethanolamine (pH 8.2), and were suspended in 500 l of the same buffer containing 20 mM dimethyl pimelimidate (Pierce), followed by an incubation at room temperature for 30 min to form a covalent bond between the antibody and the protein A. To stop the reaction, the beads were collected by centrifugation at 3,000 rpm for 1 min at room temperature with a microcentrifuge, resuspended in 1 ml of 50 mM Tris-HCl (pH 7.5) containing 150 mM NaCl, and incubated at room temperature for 15 min. After washing three times with 1 ml of TBS (20 mM Tris-HCl, pH 7.5, 150 mM NaCl, and 0.05% IGEPAL CA-630), the beads were resuspended in 1 ml of TBS.
Rad52 (62.5 pmol) and Rad51 (25 pmol) were mixed in the storage buffer (total, 5 l) using a Protein LoBind Eppendorf tube and then incubated on ice for 30 min for complex formation. After 100 l of TBS and 100 l of the anti-Rad51 antibody protein A-agarose beads were added to the Rad52 and Rad51 mixture, the mixture was incubated at room temperature for 1 h with mixing. The beads were gently washed twice with 200 l of TBS, and the proteins captured by the anti-Rad51 antibody were eluted by suspending the beads in 12 l of sample buffer (3% SDS, 100 mM DTT, 10% glycerol, 0.02% bromphenol blue, and 65 mM Tris-HCl, pH 6.8) and heating for 1 min at 90 -95°C. Aliquots (10 l) were fractionated by polyacrylamide gel electrophoresis in the presence of SDS, and the proteins were stained with Coomassie Brilliant Blue. The optical densities of the protein bands were measured by the Gel Doc XR system.
CD Spectra-CD spectra of wild type and mutant Rad52 were measured at 25°C in the storage buffer and 10 M wild type or mutant Rad52 by use of the Jasco J-720 spectropolarimeter. The light path length was 1 mm. [] is the observed molar ellipticity in mdeg.
D-loop Assay-D-loops were analyzed by detecting the radioactive signals from ssDNA that was paired with homologous dsDNA after agarose gel electrophoresis, as described previously (34). Unless otherwise stated, the reaction mixture for l of 10% SDS and deproteinized by incubating the mixture at 37°C for 15 min after the addition of 2 l of 2.6 mg/ml proteinase K (Sigma). Then 2 l of 0.1% bromphenol blue and 50% (v/v) glycerol were added to the reaction products, which were separated by electrophoresis for 30 min at 5 V/cm on a 1.0% agarose gel with TAE buffer (40 mM Tris acetate, pH 8.0, 1 mM EDTA). The gel was dried on Gel Bond film (BMA Inc.) and was exposed to a Phosphor Screen (GE Healthcare Biosciences) for ϳ16 h. The distribution of radioactivity was analyzed with a Typhoon 9410 variable image analyzer (GE Healthcare). The amounts of D-loops formed were determined by the shift of the radioactivity derived from the [ 32 P]pNS11 ssDNA 259-mer to the position of pNS11 form I DNA.
Recovery of Protein-ssDNA Complexes with Magnet Beads-pNS11 ssDNA 120-mer bound to Dynabeads was prepared by the incubation of 5Ј-biotinylated pNS11 ssDNA 120-mer (0.4 M) and 2 l of a Dynabead M-280 streptavidin suspension (streptavidin-associated magnetic beads; Dynal), in 100 l of the standard reaction buffer containing 0.05% IGEPAL CA-630, for 20 min at room temperature. Previously mixed Rad52 (1.0 M) and Rad51 (0.4 M) were incubated with the pNS11 ssDNA 120-mer bound to the Dynabeads in the standard reaction buffer (final, 100 l) at 37°C for 10 min in a Protein LoBind Eppendorf tube. The pNS11 ssDNA 120-mer ssDNA bound to the Dynabeads was recovered from the mixture by the use of a magnet. Sample buffer (10 l) was added to the recovered Dynabeads and heated for 1 min at 90 -95°C. After applying the magnet again, the proteins in the supernatant were fractionated by polyacrylamide (10%) gel electrophoresis in the presence of SDS and stained with Coomassie Brilliant Blue. The optical densities of the protein-bands were measured by the Gel Doc XR system.
Analyses of Form I DNA and Proteins in the Rad52-Rad51-ssDNA-Form I DNA Complex (Ternary Complex)-pNS11 ssDNA 120 mer -Dynabeads were prepared by the method described above, with the following modifications. A magnet was applied to 140 l of a Dynabead M-280 streptavidin suspension (Dynal), and the beads were recovered. To the beads, a solution (280 l) consisting of 2 M 5Ј-biotinylated pNS11 ssDNA 120-mer , 10 mM Tris-HCl (pH 7.5), 0.1 mM EDTA, 50 mM KCl, and 0.1% IGEPAL CA-630 was added, and the mixture was incubated at room temperature for at least 20 min. The pNS11 ssDNA 120-mer -Dynabead suspension (20 l) was added to 75 l of the standard reaction buffer, containing 0.1% IGEPAL CA-630 and previously mixed Rad52 (final, 1.0 M) and Rad51 (final, 0.4 M), in a Protein LoBind Eppendorf tube and incubated at 37°C for 10 min. After 5 l of 220 M pUC119 form I DNA was added to the mixture, the incubation was continued for 10 min. After the incubation, the Dynabeads were immediately recovered by applying a magnet. The Dynabeads were suspended in 15 l of stop solution (1.5% SDS, 30 mM EDTA) and incubated at room temperature for at least 5 min. After applying the magnet again, aliquots of the supernatant (3 and 10 l) were used for the analysis of the form I DNA and the proteins (Rad52 and Rad51), respectively. For the analysis of the form I DNA, 6 l of 200 g/ml proteinase K were added to 3 l of the supernatants and incubated at 37°C for 15 min. After the addition of 1 l of orange G-dye (0.5% orange G and 50% glycerol), the supernatant was loaded onto a 1% SeaKem GTG agarose gel, and the DNA molecules were separated by electrophoresis.
The amounts of the form I DNA and form II DNA, which were contained in the form I DNA preparation, were measured by the Typhoon 9410 imager after staining for 10 min with 1.0 g/ml ethidium bromide and destaining with distilled water for 1-2 h. For the analysis of the Rad52 and Rad51 proteins, 2 l of a dye-DTT solution (0.1% bromphenol blue, 25% glycerol, 100 mM dithiothreitol) were added to 10 l of the supernatants and heated for 2 min at 90 -95°C, and 10 l of the supernatants were analyzed by SDS-PAGE (10% acrylamide). After staining by Coomassie Brilliant Blue R-250, the optical densities of the protein bands were measured by the Gel Doc XR system.

Amino Acid Replacements K117A/R148A and K117D/ R148D in Yeast Rad52 Cause Deficiencies in Homologous
Recombination Repair in Vivo-The conserved, N-terminal domain of human Rad52 has two distinct DNA-binding sites along the outside of the ring. The first site is located at the bottom of the groove formed around the stem of the ring (26,27) and appears to be specific for ssDNA. The second DNAbinding site accommodates either dsDNA or ssDNA and is located at the opening of the groove. Lys-102 and Lys-133 are key residues located within the second DNA-binding site and have been shown to directly interact with dsDNA (27). The Lys-102 and Lys-133 residues of human Rad52 correspond to the Lys-117 and Arg-148 residues, respectively, of S. cerevisiae Rad52 (Fig. 1A) (26). These basic amino acid residues are located within the stem region of the ␤-␤-␤-␣-fold made up by a highly conserved amino acid sequence among Rad52 of human, yeast, and other eukaryotes and yeast Rad59 (Fig. 1A) dsDNA Binding to Rad52 in Homologous Recombination (26). Note that the amino acids for yeast Rad52 are counted from the first ATG codon closest to the native RAD52 promoter of S. cerevisiae (29) and that the third ATG codon is the translation start site of the native Rad52 (42,43). To understand the roles of these basic residues in in vivo functions of Rad52, we introduced mutations in the yeast RAD52 that replaced both Lys-117 and Arg-148 with either alanine or aspartic acid (K117A/R148A or K117D/R148D, respectively). We separately expressed the two mutant and wild type Rad52 proteins in rad52-defective yeast cells, using a vector containing a yeast centromere sequence (CEN6) and an autonomously replicating sequence. Although the expression of the wild type Rad52 fully restored the resistance of the rad52 null cells to the MMS, the expression of either rad52 mutant only slightly restored the MMS resistance (Fig. 1, B and C). Thus, the cells expressing the rad52 mutants were much more sensitive to MMS, as compared with the cells expressing the wild type Rad52. The K117D/R148D replacement caused a more severe phenotype than the K117A/R148A replacement (Fig. 1, B and C). These results suggest that the region in yeast Rad52 corresponding to the second DNA-binding site of human Rad52 is important for homologous recombination repair in yeast.
Lys-117 and Arg-148 in Yeast Rad52 Are Important for dsDNA Binding-We next purified the two mutant rad52 proteins containing the dual amino acid replacements (K117A/ R148A and K117D/R148D) that were overexpressed in E. coli cells ( Fig. 2A). We note that wild type and mutant Rad52, as well as Rad51, were expressed without purification tags and were purified to near homogeneity. Gel mobility shift assays revealed that the K117A/R148A and K117D/R148D mutants were

. Effects of the amino acid replacements in the second DNA-binding site of Rad52 on the repair of MMS-induced DNA damage in vivo.
A, alignment of human Rad52 and S. cerevisiae Rad52. The amino acid sequences of the N-terminal conserved domain of human and yeast Rad52 are aligned with the secondary structure identified by an x-ray crystallographic analysis of human Rad52 (26). Identical and similar amino acid residues are enclosed by orange and yellow boxes, respectively. B and C, wild type and mutant Rad52 proteins (rad52K117A/R148A and rad52K117D/R148D) were expressed under the ADH1 (alcohol dehydrogenase 1) promoter on a single-copy plasmid in rad52⌬ haploid transformants. B, spot tests for MMS sensitivity of cells grown on solid medium containing MMS are shown. WT, wild type Rad52; K117A/R148A, rad52K117A/R148A; K117D/R148D, rad52K117D/R148D; vector, a control transformant containing a vector without the Rad52 clone. C, quantitative tests of the recovery from MMS-induced damage are shown. We repeated these experiments at least three times for the quantitative representation. F, wild type Rad52; OE, rad52K117A/R148A; f, rad52K117D/R148D, E, control transformants without cloned Rad52; ᭛, wild type RAD52 cells (YPH499) containing empty vector; ϫ, wild type RAD52 cells without any plasmid. MAY 20, 2011 • VOLUME 286 • NUMBER 20

JOURNAL OF BIOLOGICAL CHEMISTRY 17611
slightly defective in ssDNA binding (Fig. 2C) and significantly defective in dsDNA binding (Fig. 2D). The gel profiles of ssDNA and dsDNA binding by the mutant rad52 proteins did not significantly change at KCl concentrations between 20 and 100 mM (data not shown).
Previously, human Rad52 was shown to form ternary complexes with ssDNA and dsDNA (26,27). In the complex, the dsDNA was shown to bind to the second DNA-binding site. We therefore investigated whether the corresponding location in yeast Rad52 also exhibited DNA binding properties similar to those of human Rad52. To do so, the rad52 mutants were first incubated with biotinylated pNS11 ssDNA 120-mer . Afterward, dsDNA (plasmid DNA) lacking sequence homology with the ssDNA was added to the Rad52-ssDNA complex. The biotinylated ssDNA was captured with streptavidin-conjugated, magnetic beads (Dynabeads), and the presence of dsDNA in the captured fraction was confirmed by agarose gel electrophoresis (Fig. 2E). The rad52 mutants were significantly defective in dsDNA binding, as compared with the wild type protein (Fig. 2,  F and G). By contrast, the amounts of Rad52 bound to the biotinylated ssDNA were nearly the same between the wild type and mutant proteins (see Fig. 5B, lanes 3-5), suggesting that Lys-117 and Arg-148 are not essential for ssDNA binding. This is consistent with the studies of human Rad52 (27). These observations indicate that Lys-117 and R148 are important for dsDNA binding, and as in human Rad52, they apparently compose a second DNA-binding site in yeast Rad52.
Mutations in Lys-117 and Arg-148 Do Not Affect Protein Folding of Rad52-To examine whether the defects in DNA binding displayed by the two Rad52 mutants are not the result of misfolded proteins, we measured CD spectra of wild type and the two mutant Rad52 (Fig. 3A). We confirmed that there were no significant changes in the spectra of K117A/ R148A and K117D/R148D mutants. Thus, the defects in dsDNA binding of rad52 K117A/R148A and rad52K117D/ R148D proteins are likely to be direct effects of the amino acid replacements, rather than the secondary effects of the global change in the protein folding.
Lys-117 and R148 Are Not Important for Either the Recombination Mediator Function or the Stoichiometric Complex Formation with Rad51-We next tested whether Lys-117 and Arg-148 of Rad52 participate in the mediator function, which is the loading of Rad51 onto RPA-coated ssDNA (Fig. 3B). When Rad51 binds to ssDNA and ATP, it displays ATPase activity. By measuring the ATPase activity of Rad51 in the presence of RPA-bound ssDNA and Rad52, it is possible to indirectly monitor the displacement of RPA by Rad51. The ATP hydrolysis activity of Rad51 was observed with substoichiometric amounts of the wild type Rad52 (Fig. 3C), which is consistent with the previous report (44). The rad52 mutants were nearly proficient FIGURE 3. The effects of the K117A/R148A and K117D/R148D replacements on Rad52 structure, mediator activity, and stoichiometric complex formation with Rad51. A, circular dichroism spectra of mutant and wild type Rad52. ⅜, wild type; ‚, K117A/R148A; Ⅺ, K117D/R148D. B, the recombination mediator functions of the mutant rad52 were assessed by the ability of Rad51 to gain access to RPA-coated ssDNA and hydrolyze ATP in the presence of the mutant rad52. C, ATPase activity of Rad51 is shown. F, the complete system containing wild type Rad52; OE, the complete system containing rad52K117A/R148A; f, the complete system containing rad52K117D/R148D; E, Rad51 alone (without RPA and Rad52); छ, RPA alone; ࡗ, wild type Rad52 alone; ϫ, storage buffer without RPA, Rad51, and Rad52. D, the ability to form complexes of Rad52 and Rad51 was examined by immunoprecipitation using anti-Rad51 antibody-protein A agarose. The precipitated proteins were analyzed by SDS-PAGE followed by Coomassie Brilliant Blue staining. W, wild type Rad52; A, rad52K117A/R148A; D, rad52K117D/R148D. Lane 1, mixture of wild type Rad51 (5 pmol) and wild type Rad52 (5 pmol) directly loaded on gel. E, the ratio of Rad52 to Rad51 was calculated and shown. The lane numbers coincide with the lane numbers in D. We repeated these experiments three times for quantitative representation. MAY 20, 2011 • VOLUME 286 • NUMBER 20 in this assay (Fig. 3C). These results indicate that the ability of Rad52 to load Rad51 onto RPA-coated ssDNA is independent of Lys-117 and Arg-148.

dsDNA Binding to Rad52 in Homologous Recombination
We previously reported the formation of a stoichiometric complex between yeast Rad51 and Rad52, and the capability of this complex to efficiently promote the formation of D-loops under reaction conditions that are optimal for the bacterial RecA protein (34). We examined whether Lys-117 and Arg-148 affect the stoichiometric complex formation between Rad52 and Rad51. Immunoprecipitation experiments using an anti-Rad51 antibody in the absence of DNA showed that the mutations of Lys-117 and Arg-148 did not affect the complex formation (Fig. 3, D and E). Thus, the stoichiometric complex formation by Rad51 and Rad52 does not require Lys-117 and Arg-148.
Lys-117 and Arg-148 of Rad52 Are Important for the D-loop Formation Catalyzed by the Rad51-Rad52 Complex-To further investigate the roles of Lys-117 and Arg-148, we next examined whether the mutations in Lys-117 and Arg-148 affect the functions of the Rad51-Rad52 complex. As shown in our previous study, the complex promoted the formation of D-loops under reaction conditions optimal for RecA (34). The reaction was free of Ca 2ϩ ions and contained 12 mM Mg 2ϩ ions (standard conditions for D-loop formation by RecA: 45, 46), which are nonoptimal conditions for either Rad51 or Rad52 to promote the formation of D-loops alone. Rad51 alone required Ca 2ϩ ions to efficiently promote the reaction, and the D-loop formation activity of Rad52 alone was sensitive to high concentrations of Mg 2ϩ ions (supplemental Fig. S1) (47). Furthermore, RPA was absent from the reaction. Thus, in the present reaction conditions, Rad52 does not function as a recombination mediator or independently catalyze the formation of D-loops.
To measure the D-loop formation activity of the Rad51-Rad52 complex, we used 32 P-labeled ssDNA 259-mer and negatively (right-handed) supercoiled, closed circular dsDNA (Form I DNA) containing a homologous sequence to the ssDNA (Fig. 4A). Rad52 was clearly required for the formation of D-loops (Ref. 34 and Fig. 4B, lanes C1-C3 and C6), and mutant Rad52 did not form D-loops by themselves (Fig. 4B,  lanes C4 and C5). When the wild type Rad52 was substituted with either of the rad52 mutants, the amount of D-loop formation significantly decreased, even after a prolonged incubation of the reaction mixture (Fig. 4, B-E). These results suggest that the second DNA-binding site in yeast Rad52 is important for the Rad51-Rad52 complex to promote D-loop formation.
Mutations in Lys-117 and Arg-148 Do Not Affect the ssDNA Binding Activity of the Stoichiometric Rad51-Rad52 Complex-We previously showed that, for the Rad51-Rad52 complex to promote D-loop formation efficiently, it must bind to the ssDNA first, before binding to dsDNA (34). We examined the ability of the stoichiometric complex containing the rad52 mutants to bind to biotinylated pNS11 ssDNA 120-mer . The ssDNA binding activity of the complex was judged by detecting the amount of protein bound to the ssDNA, which was captured with streptavidin beads (Fig. 5A). The experiment was performed using the same buffer conditions as for the D-loop formation assay. We found that the stoichiometric complexes containing rad52 mutants bound to ssDNA with efficiencies similar to that of the wild type Rad52 (Fig. 5, B and C, lanes 6 -8). The amounts of Rad51 in the complex did not significantly change (Fig. 5, B and D). In addition, the stoichiometry of Rad51 and Rad52 was nearly a 2:1 ratio and was independent of the mutations in Rad52 (Fig. 5E). Thus, we concluded that the mutations affect neither the binding of the complex to ssDNA nor the formation of the stoichiometric Rad51-Rad52 complex on ssDNA.
Lys-117 and Arg-148 Are Important for the dsDNA Binding Activity of the Rad51-Rad52 Complex-We next addressed the question of whether the second DNA-binding site of Rad52 is important for the Rad51-Rad52 complex to bind to dsDNA, after binding to ssDNA. In the D-loop formation reaction catalyzed by the prototypical bacterial RecA, dsDNA binds to the recombinase after the complex formation between RecA and ssDNA. The resulting ternary complex is important for aligning the homologous sequences between the two DNA molecules (48). To examine the effects of the mutations in the second DNA-binding site on the ternary complex formation by the Rad51-Rad52 complex, ssDNA, and dsDNA, we reconstituted the complex using a biotinylated pNS11 ssDNA 120-mer , Rad51, Rad52, and a heterologous dsDNA (Fig. 6A). The complex was then captured with the streptavidin beads, and the bound proteins and DNA were fractionated through polyacrylamide and agarose gels, respectively. The efficiency of the complex formation was judged from the amounts of Rad51, Rad52, and dsDNA detected. In the absence of Rad52, the amount of Rad51 that was captured decreased significantly (Fig. 6, B and D), probably because of the competition between ssDNA and dsDNA for Rad51 binding. In the presence of both Rad51 and Rad52, the amounts of these proteins that were captured were nearly the same between the wild type and mutant Rad52 (Fig. 6, B and C). By contrast, relatively little dsDNA was captured when either the K117A/R148A or the K117D/R148D mutant was complexed with Rad51 (Fig. 6, F and G). These results indicate that the second DNA-binding site of Rad52 is required for the dsDNA binding activity of the Rad51-Rad52 complex.

DISCUSSION
In the present study, we demonstrated that the second DNAbinding site, previously identified in human Rad52, is also conserved in yeast Rad52, and mutations in the site severely affect homologous recombination in yeast cells. Contrary to the general consensus that the primary role of Rad52 is the mediator function, we found that the second DNA-binding site is not essential for the mediator activity (Fig. 3, B and C). This result suggests that Rad52 has other important roles in yeast. Previous studies by others have shown that Rad52 plays an important role in capturing the second end of the double-stranded break. This step involves annealing between the second end of the double-stranded break and the displaced strand in the D-loop. Studies of the second DNA-binding site in human Rad52 have revealed that this site is important for ssDNA annealing (27). Thus, one possible role for the second DNA-binding site in FIGURE 5. ssDNA binding of the Rad51-Rad52 complex containing the rad52 mutants. A, reaction scheme. Rad52-Rad51 complex were incubated at 37°C for 10 min with biotinylated pNS11 ssDNA 120-mer connected to streptavidin-attached magnetic beads. The beads were collected by using a magnet. B, the corecovered proteins were analyzed by SDS-PAGE and Coomassie Brilliant Blue staining. Lanes 2-8, proteins were incubated with biotinylated pNS11 ssDNA 120-mer ; lanes 9 -11, proteins were incubated with pNS11 ssDNA 120-mer without biotin. Lane 1, a mixture of Rad51 (5 pmol) and wild type Rad52 (5 pmol) was directly loaded. W, wild type Rad52; A, rad52K117A/R148A; D, rad52K117D/R148D. C and D, the amounts of Rad52 (C) and Rad51 (D) bound to ssDNA are represented as the relative values to the amount (indicated as 1.0) of each protein in the complex of wild type Rad52 and Rad51 (lane 6 in B). We repeated these experiments at least twice for the quantitative representation. E, ratios of Rad52 to Rad51 were calculated. The column numbers coincide with the lane numbers in B. MAY 20, 2011 • VOLUME 286 • NUMBER 20 yeast Rad52 could be to capture the second end of the doublestranded break.

dsDNA Binding to Rad52 in Homologous Recombination
We examined the possible role of the second DNA-binding site of Rad52 in the D-loop formation promoted by the Rad51-Rad52 complex. Previously, we showed that Rad52 and Rad51 form a stoichiometric 2:1 complex, in both the absence and presence of ssDNA (34). This complex is capable of promoting the formation of D-loops under in vitro conditions that are unfavorable for Rad51 to function alone. Furthermore, a previous study showed that in the presence of ATP, Rad51 prevents Rad52-mediated annealing of ssDNA (49). This suggests that the ssDNA annealing activity of Rad52 is also absent in the D-loop formation reaction promoted by the Rad51-Rad52 complex. The current study shows that the second DNA-binding site of Rad52 is indispensable for the complex to promote D-loop formation (Fig. 4). Therefore, our present study reveals another possible role of the second DNA-binding site of Rad52 in homologous recombination in yeast.
We previously suggested that the homologous alignment of ssDNA and dsDNA is achieved within a stoichiometric Rad51-Rad52 complex (34), as in the case of the D-loop formation catalyzed by bacterial RecA (48,50,51). For the Rad51-Rad52 complex to promote D-loop formation efficiently, the complex must bind to ssDNA first, prior to interacting with dsDNA (34). In the present study, we found that the second DNA-binding site of Rad52 was not important for the ssDNA binding activity of the stoichiometric complex (Fig. 5). On the other hand, the second DNA-binding site was important for the dsDNA binding and D-loop formation activities of the complex bound to ssDNA (Figs. 4 and 6). These and previous observations (34) on the D-loop formation activity of the stoichiometric Rad51-Rad52 complex suggest that (i) Rad51 is the primary catalyst in the complex, because of the ATP-dependence of the formation; (ii) like bacterial RecA, the complex forms a ternary complex with ssDNA and dsDNA; and (iii) the association with dsDNA is dependent on the second DNA-binding site of Rad52. Thus, FIGURE 6. Ternary complex formation by the Rad51-Rad52 complex. A, reaction scheme. Previously mixed Rad51 and Rad52 were incubated with biotinylated pNS11 ssDNA 120-mer bound to streptavidin-conjugated magnetic beads at 37°C for 10 min. Heterologous form I DNA (pUC119) was then added to the reaction mixture (final, 100 l), and incubated for 10 min. B and F, after the magnetic beads were recovered by the magnet, the corecovered proteins (Rad52 and Rad51) and form I DNA were analyzed by SDS-PAGE (B) and agarose gel electrophoresis (F), respectively. W, wild type Rad52; A, rad52K117A/R148A; D, rad52K117D/R148D; M, a mixture of wild type Rad51 (10 pmol) and wild type Rad52 (10 pmol). C and D, the amounts of Rad52 (C) and Rad51 (D) recovered with the precipitated ssDNA are represented as the relative values to the amount (indicated as 1.0) of each protein in the complex of wild type Rad52 and wild type Rad51 (lane 3 in B). E, ratios of Rad52 to Rad51 were calculated. G, the amounts of the recovered dsDNA are represented as relative values to the amount of input pUC119 form I DNA (lane 11 in F). The column numbers coincide with the lane numbers in gel profiles. We repeated these each series of experiments three times for the quantitative representation.
we propose that in the D-loop formation reaction promoted by the Rad51-Rad52 complex, Rad52 functions as a dsDNA chaperone for Rad51. This mechanism is clearly different from the mediator activity of Rad52, which requires only substoichiometric amounts of Rad52 relative to Rad51 (44) but is independent of the second DNA-binding site of Rad52 (Fig. 3C) and from the second end capture, which involves ssDNA annealing.
This study revealed that the binding of dsDNA to the second DNA-binding site of Rad52 is important for D-loop formation by the Rad51-Rad52 complex, and this function may be important in homologous recombination in vivo, as discussed above. It would be interesting to determine why two types of D-loop forming proteins, one ATP-independent (Rad52, Ustilago maydis BRCA2 (Brh2 (52) and RecO (53)) and the other ATP-dependent (Rad51 and RecA), are both required for the various homologous recombination systems in eukaryotes and prokaryotes.