Specific Interactions between the Human RAD51 and RAD52 Proteins*

Processing of DNA damage by the DNA double-strand break repair pathway in mammalian cells is accomplished by multiprotein complexes. However, the nature of these complexes and details of the molecular interactions are not fully understood. Interaction of the yeast RAD51 and RAD52 proteins plays a crucial role in yeast DNA homologous recombination and DNA doublestrand break repair. Here, specific interactions between human RAD51 and RAD52 proteins are demonstrated both in vivo, using the yeast two-hybrid system and immunoprecipitation of insect cells co-infected with RAD51 and RAD52 recombinant viruses, and in vitro, using affinity chromatography with purified recombinant proteins. These results suggest that RAD52 may modulate the catalytic activities of RAD51 protein such as homologous pairing and strand exchange through a direct physical interaction. In addition, the domain in RAD52 that mediates this interaction was determined in vitro and in vivo. The RAD51-interacting region (amino acids 291–330) of the human RAD52 protein shows no homology with the yeast RAD52 protein, indicating that the interaction between RAD51 and RAD52 is species-specific.

Processing of DNA damage by the DNA double-strand break repair pathway in mammalian cells is accomplished by multiprotein complexes. However, the nature of these complexes and details of the molecular interactions are not fully understood. Interaction of the yeast RAD51 and RAD52 proteins plays a crucial role in yeast DNA homologous recombination and DNA doublestrand break repair. Here, specific interactions between human RAD51 and RAD52 proteins are demonstrated both in vivo, using the yeast two-hybrid system and immunoprecipitation of insect cells co-infected with RAD51 and RAD52 recombinant viruses, and in vitro, using affinity chromatography with purified recombinant proteins. These results suggest that RAD52 may modulate the catalytic activities of RAD51 protein such as homologous pairing and strand exchange through a direct physical interaction. In addition, the domain in RAD52 that mediates this interaction was determined in vitro and in vivo. The RAD51-interacting region (amino acids 291-330) of the human RAD52 protein shows no homology with the yeast RAD52 protein, indicating that the interaction between RAD51 and RAD52 is species-specific.
The genome of living organisms is constantly damaged by exposure to exogenous and endogenous agents. The inherent stability of the DNA molecule provides the first level of protection from such damage. Moreover, organisms have developed intricate machineries of DNA repair to fix various types of damage. During the last 2 decades, investigators have identified and characterized different DNA repair pathways using various approaches such as yeast genetic studies. Three epistasis groups have been identified in Saccharomyces cerevisiae that control nucleotide excision repair (RAD3 epistasis group), error prone/post-replication repair (RAD6 epistasis group), and recombinational repair (RAD52 epistasis group). Both genetic and molecular studies on the yeast RAD52 epistasis group have clearly indicated close links between the mechanism of homologous recombination and DNA double-strand break (DSB) 1 repair. Mutants belonging to this group exhibit elevated sensitivity to ionizing radiation and methyl methanesulfonate. In addition, this group of mutants is defective in both meiotic and mitotic recombination processes (1)(2)(3)(4)(5).
Most biochemical processes in various mammalian DNA repair require multiprotein complexes. Efficient removal of photodamages or bulky adducts requires more than a dozen proteins in the nucleotide excision repair (6,7). This is also the case for mismatch repair machineries (8). Recently, a mammalian protein complex that repairs double-strand breaks by recombination has been identified (9). Genetic and physical evidence for the interaction of yeast RAD52 and RAD51 suggests that at least a subset of the RAD52 epistasis group proteins interact with each other (10 -12). Additional evidence for a multiprotein repair complex comes from the dominant negative phenotypes exhibited by mutant alleles of RAD52. Some rad52 dominant-negative alleles act via nonproductive interaction with RAD51, whereas others can act independently of RAD51, perhaps associating with other repair proteins. Therefore, RAD52 epistasis group-mediated DSB repair appears to require multiprotein complexes for damage processing.
The null mutant of RAD51 is partially defective in the formation of physical recombinants and accumulates doublestrand breaks at a meiotic recombination hot spot (10,13). RAD51 from yeast and other eukaryotes has considerable homology to bacterial RecA (10, 14 -21), which has been shown to play a major role in recombination-mediated DNA repair in bacteria. RecA promotes homologous pairing and DNA strand exchange by forming nucleofilaments in the presence of ATP (22)(23)(24). Recently, yeast RAD51 has been shown to form nucleofilaments (25) and possess biochemical characteristics similar to RecA (26). RAD51 can catalyze homologous pairing and strand exchange activities in vitro (26). The human RAD51 protein has also been shown to form nucleofilaments with DNA (27), and forms nuclear foci after ionizing radiation (28).
Most of the information regarding the functions of the RAD52 protein has come from studies in yeast. Some studies have suggested that the RAD52 protein is not required for the initiation of recombination, but is essential for the intermediate stage following the formation of DSBs but before the appearance of stable recombinants (10). Most recently, human RAD52 protein has been shown to confer resistance to ionizing radiation and induce homologous recombination in monkey cells (29).
Although interaction between yeast RAD52 and RAD51 proteins has been observed both in vitro (10) and in vivo (11,12), two observations prompted us to study the interaction between human RAD51 and RAD52 proteins. First, while the human RAD51 protein is highly homologous to the yeast RAD51 protein (83%) throughout the whole polypeptide, the human and yeast RAD52 proteins show significant homology only in the amino-terminal one-third. The carboxyl-terminal two-thirds of human RAD52 has little or no homology with the yeast RAD52 protein (20,30,31). However, it is the carboxyl terminus of yeast RAD52 that is involved in the interaction with RAD51 (11,12). This raises the question of whether human RAD52 and RAD51 interact at all. Second, neither human RAD51 cDNA (20) nor yeast Kluyveromyces lactis RAD51 complements the Rad51 defect of yeast S. cerevisiae (12). This suggests that the interaction between RAD51 and RAD52 via the carboxyl terminus of RAD52 may be species-specific.
In this study, we show that human RAD52 physically associates with human RAD51 in vitro and in vivo. Additionally, the region of the RAD52 protein that provides this interaction with RAD51 has been determined. This RAD52 region shows no homology with yeast S. cerevisiae RAD52. These results suggest that a species-specific interaction between human RAD51 and RAD52 may be important for homologous recombination and/or DSB repair in mammalian cells.
Two-hybrid Plasmid Constructs-To clone the RAD51 or RAD52 cDNA into the two-hybrid vector, PCR was used to generate the inserts. These inserts contained a BamHI site at the 5Ј ends and a stop codon and a SalI site at the 3Ј ends. The PCR products digested with BamHI and SalI were ligated into the two-hybrid vectors that had been predigested with the same enzymes. Final constructs were sequenced by automated sequencing (Applied Biosystems Inc.). The full-size RAD51 cDNA insert was PCR-amplified from a cDNA library, and its sequence was compared to published cDNA sequences (19,20). The full-length or truncated human RAD52 cDNAs that were fused with either the Gal4-DB (pGBT-52) or the Gal4-DA (pGAD-52) were constructed by ligating the RAD52 cDNA synthesized by PCR from a RAD52 cDNA clone (30).
Two-hybrid Protein-Protein Interaction Assay-To determine the potential interactions between RAD51 and RAD52 proteins, plasmids for the two fusion constructs (one fused with the Gal4-DB, the other fused with Gal4-DA) were co-transfected into the SFY526 using the polyethylene glycol/lithium acetate method (34). Transformed yeast cells were selected on Trp Ϫ /Leu Ϫ synthetic agar plates for 3 days at 30°C. Three independent transformants were then transferred to gridded filter paper and grown overnight on Trp Ϫ /Leu Ϫ plates. These clones were then frozen in liquid nitrogen for 5-10 s, placed on top of another filter which was presoaked in Z-buffer (60 mM Na 2 HPO 4 ⅐7H 2 0, 40 mM NaH 2 PO 4 ⅐H 2 O, 100 mM KCl, 1 mM MgSO 4 ⅐7H 2 O, pH 7.0) containing 0.27% (v/v) 2-mecapthoethanol and 0.334% (w/v) of 5-bromo-4-chloro-3indolyl-␤-D-galactoside, and incubated at 30°C. Either 5 ml or 1.5 ml of Z-buffer was used for a 150-mm or a 100-mm filter, respectively. Interaction between these two fusion protein was monitored by color development at 1 h, 2 h, 4 h, 8 h, and 24 h postincubation. Clones that developed visible blue color by 1 h were registered as ϩϩϩ, 2-4 h as ϩϩ, and 8 -16 h as ϩϪ. If no color developed after 16 h, clones were recorded as Ϫ.
Purification of RAD51 and RAD52 Proteins Overexpressed in Bacteria-Coding regions of both proteins were cloned into a T7 promoterbased bacterial expression vector (35), pET28, purchased from Novagene, Inc. (Madison, WI). The resulting plasmids were designated pET-hRAD51 and pET-hRAD52. BL21(DE3) strain of E. coli was transformed with either pET-hRAD51 or pET-hRAD52, induced with isopropyl-1-thio-␤-D-galactoside and harvested for protein purification. Proteins were purified by a combination of metal-chelate affinity and Mono Q ion exchange column chromatography. The biochemical activities of RAD51 including ATP hydrolysis and DNA binding were demonstrated. 2 Co-expression of RAD51 and RAD52 in Insect Cells-Spodoptera frugiperda cells (Sf9) were kindly provided by Glenn Godwin at Life Technologies, Inc. and grown in Sf-900 II SFM medium (Life Technologies, Inc.). Cells were maintained and infected either as monolayer or in suspension on a rotary shaker in a low temperature incubator set at 27°C. Cells were routinely infected with 10 plaque-forming units of recombinant virus per ml of culture at a density of 1 ϫ 10 6 cells (Ͼ95% viability) per ml. Linearized wild type baculovirus (BaculoGold TM ) and transfer vector pVL1393 were purchased from Pharmingen (San Diego, CA) and Invitrogen (San Diego, CA), respectively. The recombinant transfer vectors pVL1393-hRAD51 and pVL1393-hRAD52 were constructed from the transfer vector pVL1393 and the plasmid pET-hRAD51 or pET-hRAD52. The resulting constructs were then co-transfected with the linearized wild type baculovirus DNA (BaculoGold TM ) into Sf9 cells by the cationic liposome method, as described by the manufacturer's instruction (Invitrogen). Recombinant BaculoGold-hRAD51 or BaculoGold-hRAD52 baculoviruses were detected by visual inspection and were further screened by PCR. For the co-expression of RAD51 and RAD52 in single Sf9 cells, an equal number of each recombinant virus was used (multiplicity of infection ϭ 10). All methods pertaining to the growth, transfection, screening, and manipulation of the Sf9 cells and baculovirus have been described elsewhere (36,37) unless otherwise indicated. Antibody Production-Recombinant RAD52 proteins purified from Escherichia coli were used as antigens to immunize rabbits. Antibody was affinity-purified using protein A membranes (Amicon). Anti-yeast RAD51 was kindly provided by Dr. Patrick Sung (26) (University of Texas, Galveston).
SDS-Polyacrylamide Gel Electrophoresis and Immunoblotting-After electrophoresis in SDS (10 -20%) polyacrylamide gels, proteins were transferred onto nitrocellulose filters using a semi-dry electroblot apparatus. Nitrocellulose sheets were probed for 1 h with a 1:3000 dilution of affinity-purified anti-RAD51 or anti-RAD52 antibodies (stock of A 280 ϭ 3) and developed using the indirect alkaline phosphatase procedure (38).
RAD52 Affinity Column-RAD52 protein and BSA were coupled to AminoLink-Plus gel (Pierce) following the manufacturer's instructions. RAD52 protein (4 mg) was dialyzed against buffer D (0.1 M sodium phosphate, 0.15 M NaCl, pH 7.2) and mixed with 0.5 ml of AminoLink slurry equilibrated with buffer D. After incubation overnight at 4°C by end-over-end rocking, the slurry was washed with buffer D, and the buffer was replaced by buffer D with 0.1 M cyanoborohydride. The slurry was further incubated overnight at 4°C, quenched with 1 M Tris-HCl, pH 7.4, and washed with buffer E (25 mM Tris-HCl, pH 7.6, 1 mM EDTA, 5 mM MgCl 2 , 0.05% sodium azide). BSA (5 mg) was coupled to Amino-Link gel following the same procedure.
RAD51 protein was incubated in batch with the AminoLink resin coupled with either RAD52 or BSA. RAD51 was dialyzed against buffer E and concentrated by centrifugation through an Amicon Centricon 30 filter. The concentrated RAD51 (250 g in 250 l) was mixed with 0.25 ml (drained volume) of the AminoLink resin covalently linked to either RAD52 or BSA. After incubating the mixture for 2 h at 4°C with gentle mixing, the slurry was applied to a polyethylene column. The column was washed with buffer E. The RAD51 were eluted from the column with a step gradient of increasing NaCl concentration; each step (1 ml) contained buffer E plus either 50 mM, 0.15 M, 0.25 M, or 0.5 M NaCl. All steps were carried out at 4°C and a flow rate of 0.2 ml/min. Fractions of 200 l were collected, the absorbance at 280 nm was measured, and aliquots were used for immunoblot analysis using anti-RAD51 or anti-BSA antibodies to confirm the presence of relevant proteins in each 2 M. S. Park, manuscript in preparation.

fraction.
Generation of Deletion Mutants for in Vitro Assay-Deletion mutants expressing the truncated form of RAD52 were made by generating a series of nested deletions of pET-hRAD52 from the amino terminus of RAD52 protein using the ExoIII method (39). To use the ExoIII approach to generate deletions in RAD52 cDNA, the original pET-hRAD52 construct was modified by inserting a linker (sense strand: 5Ј-CATATGGGTACCACTAGTCGGGATCC-3Ј; antisense strand: 5Ј-GGATCCCGACTAGTGGTACCCATATG-3Ј) into the pET-hRAD52 that was restriction-digested with NdeI and BamHI. The modified pET-hRAD52 was restriction-digested with KpnI and BamHI and subjected to ExoIII/mung bean nuclease treatment and subsequent ligation. To further map the interacting domain, two mutants containing a part of the RAD52 protein (amino acids 260 -340 and 270 -330) were made by ligating the PCR products to pET28b that was restriction-digested with NcoI and XhoI. Mutant constructs were first transformed into an E. coli strain, XL2-Blue (Stratagene), and subsequently screened for clones with a proper RAD52 reading frame. Then, BL21(DE3) was transformed with deletion construct plasmids isolated from XL2-Blue, induced, harvested, and subjected to protein purification. These mutant constructs produced truncated RAD52 proteins with 6 histidine residues at the amino terminus of the protein. Mutant proteins were purified by metal-chelate affinity columns.
RAD51 Affinity Chromatography-Purified RAD51 protein was covalently linked to AminoLink resins as described for RAD52 protein.

Association of RAD51 and RAD52 in Vivo-
To determine whether the human RAD52 protein interacts with human RAD51 protein, the following four constructs that contained the full-length coding regions of either human RAD52 or RAD51 cDNA were made: pGBT52, pGAD52, pGBT51, and pGAD51. Plasmids pGAD424 and pGBT9 were used as negative control. Two plasmids, one derived from pGBT9 (first column in Table I) and the other from pGAD424 (second column in Table I) were co-transfected into SFY526 yeast cells by different combinations of these constructs. The enzymatic activity of ␤-galactosidase (lacZ) resulting from the association of these fusion proteins is shown (Table I). Table I demonstrates that 1) human RAD51 and RAD52 protein can each self-associate, 2) human RAD52 protein associates with the human RAD51 protein.
The self-association of yeast RAD51 protein and the region responsible for this activity have been reported (12). However, self-association of RAD52 protein has not been reported. A comprehensive analysis of the human RAD52 self-association will be reported elsewhere. 3 Interestingly, we consistently observed weaker interactions between pGAD52 and pGBT51 as compared to interaction between pGBT52 and pGAD51. Plasmid pVA3 that contains mouse p53 protein (amino acids 72-390)/Gal4-DB fusion, and pTD1 that contains SV40 large Tantigen (amino acids 64 -708)/Gal4-DA fusion were used as experiment controls. Although p53 interacts with large T-antigen, p53 (amino acids 72-390) and T-antigen (amino acids 64 -708) did not interact with human RAD51 or RAD52 in this yeast two-hybrid assay (Table I).
Although the two hybrid data indicated in vivo interaction of human RAD52 and RAD51 proteins in yeast cells, this interaction was relatively weak as compared to the self-association of either protein (Table I). Therefore, another in vivo experiment was conducted to confirm the association of RAD51 and RAD52. RAD51 and RAD52 were expressed independently or simultaneously in insect cells (Fig. 1A) using recombinant baculoviruses. The RAD51 and RAD52 complex was immunoprecipitated with antibodies specific to either RAD51 or RAD52 (Fig. 1B). The immunoprecipitates prepared by anti-RAD51 antibody contained a polypeptide that cross-reacted with the anti-RAD52 antibody. The immunoprecipitates prepared with anti-RAD52 contained a polypeptide that cross-reacted with the anti-RAD51 antibody. Furthermore, immunoprecipitation experiments with both antibodies on 35 S-labeled HeLa cell extracts also indicated physical association of RAD51 and RAD52 (data not shown). These results show that RAD51 and RAD52 interact with each other in vivo.
Association of RAD51 and RAD52 in Vitro-Although the physical association between RAD52 and RAD51 was demonstrated in vivo, it was not clear whether this interaction was a direct one or was mediated by another protein(s) in vivo. In order to distinguish between these possibilities, in vitro interaction studies using purified RAD51 and RAD52 proteins were conducted. For this purpose, RAD51 and RAD52 proteins were expressed in E. coli ( Fig. 2A) and purified to near-homogeneity (Fig. 2B). The corresponding antibodies reacted with the purified RAD51 and RAD52 proteins (Fig. 2C). A RAD52 affinity column was constructed by coupling purified RAD52 protein to AminoLink resin at a concentration of 4 mg of RAD52 protein per ml of resin. The resin (0.25 ml) was mixed with 0.25 mg of RAD51 and, after incubating for 2 h at 4°C, washed with increasing concentrations of NaCl. In the absence of NaCl, 3 Z. Shen, submitted for publication.
T-antigen (pTD1) ϩϩ a See text for the symbols of lacZ activity. b Full size RAD51 protein fused to Gal4-DA (pGAD51). c Full size RAD51 protein fused to Gal4-DB (pGBT51). d Full size RAD52 protein fused to Gal4-DA (pGAD52). e Full size RAD52 protein fused to Gal4-DB (pGBT52). greater than 90% of RAD51 bound to the resin (Fig. 3). Less than 10% of RAD51 was eluted with NaCl concentrations of 50 and 150 mM. However, upon addition of Ͼ250 mM NaCl, ϳ60% of the bound RAD51 eluted from the column (Fig. 3). The corresponding fractions contained RAD51 as determined by immunoblot (data not shown).
As a control, an identical experiment was carried out with RAD51 protein using AminoLink gel, that had been coupled with BSA instead of RAD52 protein (Fig. 3). In this case, most of the RAD51 protein eluted in the absence of NaCl, with only a low level (less than 5%) bound to the column. Thus, the affinity of RAD51 for the resin coupled to RAD52 protein is due to an interaction of RAD51 with RAD52 protein and not to any nonspecific interactions between the RAD51 and the resin. Therefore, this experiment shows that the interaction between human RAD52 and RAD51 is direct.
Identification of RAD52 Domain That Interacts with RAD51-To identify the region(s) in the RAD52 protein that mediates the demonstrated interaction, in vitro and in vivo experiments were performed. For the in vitro experiment, a series of deletion mutants of RAD52 was generated and purified to apparent homogeneity (Fig. 4A). Mutant RAD52 proteins were tested by affinity chromatography with the RAD51 column. But this time the concentration of NaCl was fixed to 0.5 M to elute RAD52 proteins that were bound to the RAD51 column. As shown (Fig. 4B), a small region in RAD52 (amino acids 291-330) was accountable for the interaction of RAD52 with the RAD51 column.
For the in vivo experiment, a series of truncated RAD52 cDNAs were fused to the Gal4-DB in vector pGBT9. These constructs were co-transfected with pGAD51 into SFY526 cells. lacZ activity resulting from the interaction of Gal4-DB/RAD52 fusion constructs with full size RAD51/Gal4-DA fusion protein was assayed. Table II indicates that the amino acids 287-333 are responsible for the interaction in yeast cells. Based on the in vivo and in vitro results taken together, we conclude that the RAD51 interaction domain of human RAD52 protein resides in the region of 291-330. DISCUSSION The data presented in this study clearly indicate that human RAD51 and RAD52 proteins physically associate both in vivo and in vitro. Similar results have been described for yeast RAD51 and RAD52 proteins (10 -12). The present study further mapped the human RAD52 region that interacts with the human RAD51 protein. A region of approximately 40 amino acids near the carboxyl terminus of RAD52 is required for interaction with RAD51 in vitro. Previous reports have shown that the human RAD51 protein cannot complement the Rad51 defect in yeast S. cerevisiae (10), and yeast K. lactis RAD51 has little ability to restore the Rad51 defect in yeast S. cerevisiae (12). Since the RAD51 interaction region (291-330) of human RAD52 does not share homology with the yeast RAD52 protein, the interaction between RAD51 and RAD52 appears to be species-specific. This specificity does not appear to be determined by the RAD51 protein, because RAD51 and its homologs FIG. 3. Human RAD52 protein affinity chromatography of human RAD51 protein. RAD51 protein was applied to AminoLink resin that had been coupled with either RAD52 or BSA as described under "Materials and Methods." RAD51 protein was eluted from the column using 1-ml step gradients containing 0 mM, 50 mM, 150 mM, 250 mM, and 500 mM NaCl (arrows). Two hundred-l fractions were collected, and the absorbances at 280 nm were determined.
Ϫ Ϫ a Numbers in parentheses denote the human RAD52 amino acids contained in the construct. are well conserved. The RAD52 self-association region has been mapped to its amino-terminal region, 3 which is independent of the RAD51 interaction region. Therefore, we have demonstrated that the human RAD52 protein has at least two domains responsible for protein-protein interactions.
Self-association of human RAD51 is consistent with the results seen in yeast (12). This self-association may be required for self-assembly of RAD51 on DNA to form nucleofilaments with a characteristic regularity that has been observed previously (27). The homotypic interaction domain of human RAD51 appears to be present near its amino terminus 4 as seen with yeast RAD51 protein (12).
The yeast RAD51 protein has been shown to exhibit the catalytic activities of strand exchange and homologous pairing (26). More recently, the human RAD51 has been shown to bind to DNA and also form RecA-like nucleofilaments on DNA (27), suggesting the possibility that the human RAD51 also possesses either or both homologous pairing and strand exchange activities (29). The yeast RAD52 protein can bind to both single-and double-stranded DNAs in the absence of ATP and carries out annealing of homologous single-stranded DNA (40). It can also promote the strand transfer reaction (40). However, the reaction was ATP-independent and had an efficiency only 5% of that of the RecA protein (40). Besides the heterotypic and homotypic interactions of RAD52, there is no direct information about the biochemical activities of the mammalian RAD52 protein to date. Although it can be speculated that the RAD52 protein might modulate the catalytic activities of RAD51 based upon the information presented in this study showing the direct association of RAD52 and RAD51 in vivo and in vitro. The catalytic activity of yeast RAD51 can be exhibited in the absence of RAD52 protein in vitro. 2 However, a finer tuning of these processes in vivo may require other proteins such as RAD52. Overexpression of the RAD52 protein in monkey cells provides a positive effect on homologous recombination and protection from ionizing radiation (29). These results may further support the notion of the RAD52 protein as a positive modulator of RAD51 activities for DSB repair in mammalian cells. However, more direct evidence for this hypothesis needs to be drawn from biochemical experiments in the presence of both proteins in vitro.