Complex formation by the human Rad51B and Rad51C DNA repair proteins and their activities in vitro

The human Rad51 protein is essential for DNA repair by homologous recombination. In addition to Rad51 protein, five paralogs have been identified: Rad51B/Rad51L1, Rad51C/Rad51L2, Rad51D/Rad51L3, XRCC2, and XRCC3. To further characterize a subset of these proteins, recombinant Rad51, Rad51B-(His)(6), and Rad51C proteins were individually expressed employing the baculovirus system, and each was purified from Sf9 insect cells. Evidence from nickel-nitrilotriacetic acid pull-down experiments demonstrates a highly stable Rad51B.Rad51C heterodimer, which interacts weakly with Rad51. Rad51B and Rad51C proteins were found to bind single- and double-stranded DNA and to preferentially bind 3'-end-tailed double-stranded DNA. The ability to bind DNA was elevated with mixed Rad51 and Rad51C, as well as with mixed Rad51B and Rad51C, compared with that of the individual protein. In addition, both Rad51B and Rad51C exhibit DNA-stimulated ATPase activity. Rad51C displays an ATP-independent apparent DNA strand exchange activity, whereas Rad51B shows no such activity; this apparent strand exchange ability results actually from a duplex DNA destabilization capability of Rad51C. By analogy to the yeast Rad55 and Rad57, our results suggest that Rad51B and Rad51C function through interactions with the human Rad51 recombinase and play a crucial role in the homologous recombinational repair pathway.


INTRODUCTION
Homologous recombinational repair (HRR) is an important pathway in repairing DNA double-strand breaks (DSBs) with high accuracy, which is indispensable for the maintenance of genome stability (for review see [1][2][3][4]. The human Rad51 protein, a structural and functional homolog of E. coli recombinase RecA (5)(6), plays a central role in HRR. In vitro studies have shown that Rad51 binds single-stranded and double-stranded DNA (ssDNA and dsDNA), exhibits DNA-dependent ATPase activity, and functions to catalyze homologous pairing and DNA strand exchange (6)(7)(8). As a prerequisite step for the action of Rad51 in HRR, Rad51 binds DNA to form highly ordered nucleoprotein filaments (9,10). It has also been shown that these filaments form preferentially on tailed duplex DNA substrates that mimic the 3'-overhanging ssDNA tails at the sites of DSBs (11). Thus far, Rad51 is the only mitotic protein that catalyzes the key reactions of homologous pairing and strand transfer.
These Rad51 paralogs share 20-30% sequence identity with Rad51 and with each other.
Interactions between these proteins have been shown by a yeast two-hybrid assay and by baculovirus co-expression experiments (24). Recent evidence from mutations in hamster and chicken DT40 cell lines (25)(26)(27), showing increased sensitivity to DNA damage and increased spontaneous chromosomal instability, suggests an important role for these paralogs in HRR.
With the aim of determining the function of these paralogs, recent efforts have been devoted to purifying the proteins and to revealing their biochemical properties. The Rad51D protein was reported to display ssDNA binding and DNA-stimulated ATPase activity, and to interact with XRCC2 in vivo (28). The co-purified XRCC3/Rad51C and XRCC2/Rad51D complexes bind to ssDNA and form protein-DNA networks, as visualized by electron microscopy (29)(30)(31). These heterodimers were also reported to exhibit homologous pairing activity (30,31).
Here we report several in vitro activities of the human Rad51B and Rad51C proteins, including DNA binding, ATP hydrolysis, and apparent DNA strand exchange. We also compared these activities to those of Rad51. The activities of mixed proteins (Rad51+Rad51C, Rad51+Rad51B, and Rad51B+Rad51C) were determined as well. Our results demonstrate the formation of multiprotein complexes comprised of Rad51B-Rad51C and Rad51-Rad51C-Rad51B, and show that Rad51C both separates duplex DNA and facilitates the binding of Rad51 to DNA.

EXPERIMENTAL PROCEDURES
Protein Expression -The human Rad51, Rad51B and Rad51C ORFs were cloned into the baculoviral vector and the recombinant proteins were expressed in Sf9 insect cells as described previously (24). The human Rad51 protein was purified by selective spermidine precipitation (32) with a modified step -an additional Q-Sepharose purification after spermidine precipitation to eliminate DNA.
The N-terminal 6x(His)-tagged Rad51B was initially constructed and expressed in insect cells using the baculovirus system. An attempt to purify this protein was unsuccessful because the protein did not bind to a Ni-NTA column. It is possible that the N-terminus of the expressed Rad51B protein is folded into the structure and thus inaccessible to bind a Ni-NTA column. Therefore, we switched to a C-terminal tagged construct. To do so, the C-terminal tagged RAD51B gene was PCR amplified from a human testis cDNA library with a hexahistidine encoded sequence in the reverse primer and cloned into a pFastBac1 baculoviral vector. The sequence of RAD51B was confirmed to be identical to those published previously. The Cterminal 6x(His)-tagged Rad51B protein was then expressed in Sf9 insect cells. To study the interaction between Rad51B and Rad51C, we chose to make an untagged Rad51C construct because the Rad51B was tagged with 6x(His). The RAD51C gene was also PCR amplified from a human testis cDNA library and cloned into a pFastBac1 vector. The sequence of RAD51C was confirmed and the recombinant protein was expressed in Sf9 cells.
Purification of Recombinant 6x(His)-tagged Rad51B -All purification procedures were carried out at 4°C. Frozen cells (from 2 L insect cell culture) were thawed on ice and resuspended in 120 ml of ice-cold lysis buffer containing 50 mM NaH 2 PO 4 (pH 8.0), 300 mM NaCl, 10% glycerol, 10 mM imidazole, and protease inhibitors. The mixture was incubated on ice for 30 min. PMSF was added to a final concentration of 0.5 mM prior to lysis by sonication.
The lysate was clarified by centrifugation at 14,000 rpm for 40 min. The resulting supernatant (soluble fraction) was then applied onto a Ni-NTA agarose column (5 ml, Qiagen), which had been previously equilibrated with start buffer (50 mM NaH 2 PO 4 , pH 8.0, 300 mM NaCl, 10% glycerol, and 10 mM imidazole). The column was then washed with 10 column volumes of start buffer, followed by 10 column volumes of wash buffer containing 50 mM NaH 2 PO 4 (pH 8.0), 300 mM NaCl, 10% glycerol, and 20 mM imidazole. Finally, elution buffer (50 mM NaH 2 PO 4 , pH 8.0, 300 mM NaCl, 10% glycerol, and 250 mM imidazole) was added, and the eluate was The pure Rad51B protein, as determined by SDS-PAGE, was finally dialyzed for 12 hr against storage buffer consisting of 25 mM Tris-HCl, pH 7.5, 100 mM NaCl, 0.1 mM EDTA, 1 mM DTT, and 10% glycerol, and was stored in aliquots at -70°C.
Purification of Recombinant Rad51C -Frozen cells (from 2 L insect cell culture) were thawed on ice and resuspended in 120 ml of ice-cold lysis buffer containing 100 mM Tris-HCl (pH 8.0), 300 mM NaCl, 0.1% Triton-X 100, 0.5 mM DTT, 2 mM EDTA, 5% glycerol, and protease inhibitors. The mixture was incubated on ice for 30 min and was sonicated briefly to reduce viscosity. Cell debris was removed by ultracentrifugation in a Beckman 70Ti rotor at 39,000 rpm for 1 hr. The clarified supernatant was dialyzed 2 times against 4 L of spermidine buffer (20 mM Tris-acetate, pH 7.5, 7 mM spermidine-NaOH and 0.1 mM DTT) for a total of 10 hr. The precipitate was recovered by centrifugation at 12,000 rpm for 30 min and resolubilized in 40 ml of P buffer (0.1 M KH 2 PO 4 , pH 6.8, 0.1 M KCl, 0.5 mM DTT, and 10% glycerol) by stirring slowly about 2 hr until the pellet had largely dissolved. Insoluble material was removed by centrifugation in a Beckman 70Ti rotor at 39,000 rpm for 30 min. The clarified supernatant was loaded directly onto a Hydroxyapatite column (20 ml, Bio-Rad) which had been preequilibrated with P buffer. After loading, the column was washed with 10 column volumes of P buffer, then eluted with a linear gradient of 0.1-0.8 M KH 2 PO 4 in P buffer. The eluate was collected in 2-ml fractions. An aliquot (10 µl) of each fraction was analyzed by SDS-PAGE, and peak fractions containing Rad51C, which eluted at approx. 0. Antibodies -Polyclonal antiserum against human Rad51B was raised in rabbits using a synthesized peptide as the immunogen. The 6x(His)-tagged bacterial recombinant protein of human Rad51C was purified and used for polyclonal antiserum production in rabbits. Both Rad51B and Rad51C antibodies were affinity-purified from the antiserum. The human Rad51 antibody was kindly provided by Dr. Akira Shinohara.

Ni-NTA Magnetic Agarose Beads Pull-down Experiments -The pull-down experiments
were carried out with Rad51, Rad51C and 6x(His)-tagged Rad51B using the procedures as previously described (24). of loading buffer (30% glycerol, 0.1% bromophenol blue). Samples were loaded onto a 10% polyacrylamide gel and analyzed by electrophoresis in TBE at 9 V/cm for 2 hr. Gels were dried on filter paper and [ 32 P]-labeled DNA-protein complex was detected by autoradiography. DNA Strand Exchange Assays -To form nucleoprotein complexes, Rad51C protein was incubated with ssDNA in the standard buffer containing 33 mM HEPES (pH 7.0), 2 mM DTT, 100 µg/ml BSA, 2 mM ATP, and 3 mM magnesium acetate at 37°C for 15 min. DNA strand exchange was initiated by the addition of dsDNA to the nucleoprotein complexes and incubated at 37°C. Aliquots were withdrawn from the reaction mixture, deproteinized by the addition of EDTA to 50 mM, SDS to 1%, and proteinase K to 700 µg/ml, followed by incubation for 15 min at 37°C, mixed with a 1/10 vol. of loading buffer (30% glycerol, 0.1% bromophenol blue), and loaded on a 10% polyacrylamide gel. Products of DNA strand exchange were quantified using a Storm 840 PhosphorImager (Molecular Dynamics).
Nuclease Assays -To control for nuclease activity, the Rad51C protein preparation was assayed at the standard conditions for DNA strand exchange. Rad51C nucleoprotein complexes were assembled on 63-mer ssDNA (#2). Duplex DNA (#1-2) was used as a substrate for DNA strand exchange. The dsDNA was 32 P-labeled at the 5'-end of the identical DNA strand (#2) using T4 polynucleotide kinase and [γ-32 P]-ATP. DNA strand exchange was carried out for 30 min at 37°C.
After deproteinization, the products of this reaction were separated using a 10% polyacrylamide gel run in TBE. The DNA strand that was displaced from dsDNA was eluted from the gel, and its integrity was confirmed by electrophoresis in a 15% polyacrylamide gel containing 8 M urea.

RESULTS
Purification of Rad51B protein -After cells were lysed, roughly 50% of the 6x(His)tagged Rad51B protein was found in the soluble fraction, and the protein bound to a Ni-NTA column. The Ni-NTA column provides a powerful purification step yielding 90% pure Rad51B.
The eluate from the Ni-NTA column was then run through a Q-Sepharose column to remove contaminating DNA from the protein preparation. Subsequently, Rad51B was further purified to near homogeneity by a MonoQ column (Fig. 1A). Finally, about 400 µg of homogeneous Rad51B protein was obtained from two liters of insect cell culture. The identity of the protein was confirmed by Western blotting analysis with α-Rad51B antibody (Fig. 1B).
Purification of Rad51C protein -The solubility of Rad51C protein was determined to be approximately 50%, and the soluble fraction was used for purification. In light of the selective precipitation of the E. coli RecA (33,34) and human Rad51 protein by spermidine (32,(35)(36)(37)(38), the effect of spermidine on Rad51C was examined. We found that Rad51C is indeed precipitated by spermidine in a concentration-dependent manner (data not shown). Soon after the soluble fraction of cell lysate was dialyzed against spermidine buffer, a white precipitate was observed. The amount of precipitate accumulated continuously up to 10 hr with a substantial increase following the second buffer change. Approximately 70% of the Rad51C was precipitated by spermidine. The Rad51C protein was subsequently recovered from the precipitate by slowly stirring the precipitate ~ 2 hr at 4°C, with about 70% of the protein being dissolved into the solution. The solution was then subjected to sequential column chromatography using Hydroxyapatite, Q-Sepharose, Heparin, and MonoQ columns. About 500 µg of homogeneous Rad51C protein was obtained from two liters of insect cell culture ( Fig. 2A). The identity of the protein was confirmed using α-Rad51C antibody (Fig. 2B).
Complex formation between Rad51, Rad51B, and Rad51C -The 6x(His)-tagged Rad51B and untagged Rad51 and Rad51C were co-expressed in Sf9 cells. Using Ni-NTA beads, we found that Rad51 could be pulled down from insect cell extracts by the 6x(His)-tagged Rad51B through its interaction with Rad51C (Fig. 3A). This result demonstrates a simultaneous interaction between Rad51, Rad51B and Rad51C, implying a heterotrimer formation between these three proteins. However, the amount of Rad51 in the complex is much less than the amount of Rad51C. This observation was found consistently in several experiments, and may indicate that the association between Rad51C and Rad51 is weak. In addition, to ensure that the observed protein associations were not caused by DNA, the Ni-NTA pull-down experiments were also carried out in the presence of an increasing amount of ethidium bromide. We found that up to 700 µg/ml of ethidium bromide did not affect the pull-down (data not shown). In the presence of ethidium bromide, a consistent amount of Rad51 was detected after pull-down, suggesting that the presence of Rad51 in the complex is due to protein-protein interactions rather than to DNA binding.
To further confirm this observation, individual purified proteins were mixed and used in similar Ni-NTA pull-down experiments. Fixed amounts of purified Rad51B and Rad51C (molar ratio = 1:1) were mixed with increasing amounts of Rad51 in the absence and presence of 500 µg/ml of ethidium bromide. The results showed that an increasing amount of Rad51 was pulled down in both conditions (Fig. 3B), confirming that the interaction is not due to DNA association. Importantly, no detectable Rad51B and Rad51C were found in the supernatant (data not shown), suggesting a strong interaction between Rad51B and Rad51C. However, the majority of Rad51 remained in the supernatant, and only ~ 5-10% of Rad51 was pulled down by Rad51B (data not shown), indicative of a weak interaction between Rad51B-Rad51C and Rad51.
Based on these results, we conclude that Rad51B and Rad51C form a stable complex with a strong association constant, and that Rad51, Rad51C, and Rad51B form a heterotrimer involving a weak interaction between Rad51 and Rad51C. To our knowledge, this is the first in vitro evidence showing the simultaneous interaction between Rad51B-Rad51C and Rad51.
Rad51B and Rad51C preferentially bind ssDNA-tailed dsDNA -Based on sequence alignments, we have found that four of the human Rad51 paralogs, including Rad51B, Rad51C, Rad51D, and XRCC3, contain a conserved helix-hairpin-helix motif (39,40) in their N-terminus (our unpublished data). The helix-hairpin-helix motif is a domain of around 20 amino acids widely present in prokaryotic and eukaryotic proteins involved in non-sequence-specific DNA binding, e.g., E. coli RadC, eukaryotic ERCC1/Rad10 family, yeast SW10 and human Rad51 (41). Therefore, we examined the DNA-binding ability of Rad51B and Rad51C using an electrophoretic mobility-shift assay. Both Rad51B and Rad51C were found to bind φX174 single-stranded DNA (data not shown). To further determine the substrate preference of these two proteins, oligonucleotide substrates were employed: ssDNA (63-mer), dsDNA (63/63-mer), and tailed dsDNA with 3'-tailed ssDNA (63/32-mer). Both Rad51B and Rad51C preferentially bind 3'-tailed dsDNA over ssDNA, and they have the lowest affinity for dsDNA (Fig. 4A). Two forms of Rad51C-DNA complexes were observed: an aggregate that was trapped in the loading well, and a protein-DNA complex that entered the gel (as indicated in Fig. 4A). However, for Rad51 (data not shown) and Rad51B, only the aggregates were observed. It is possible that Rad51C exists as multimers of limited size, some of which form a large aggregate of protein-DNA complexes. Our gel filtration results demonstrating tetramer formation by Rad51C (data not shown) indeed support this speculation. Since the repair of DNA double-stranded breaks (DSBs) by homologous recombination requires processing of the break to produce a 3'-overhanging ssDNA tail, our observations indicate that Rad51B and Rad51C proteins preferentially bind the 3'-overhanging ssDNA tail created during the repair process.
We also compared the DNA binding capabilities of Rad51B and Rad51C with Rad51 using the same three DNA substrates: 63-mer, 63/63-mer and 63/32-mer. We found that Rad51C possesses a higher affinity for all of three DNAs than does Rad51 (Fig. 4A and data not shown).
The relative affinity of these three proteins for DNA was found to be: Rad51C > Rad51 > Rad51B ( Fig. 4A and data not shown).
In addition, we investigated the DNA binding capacity of various mixed proteins. For the 63-mer ssDNA substrate, we observed enhanced DNA binding with mixed Rad51 and Rad51C ( Fig. 4B), as well as with mixed Rad51B and Rad51C (Fig. 4C). As is particularly evident from substrate use, comparison of the DNA binding ability of an individual protein (Rad51 or Rad51C) with the mixed proteins (Rad51 + Rad51C) shows that the DNA binding activity of protein mixture is elevated. Using 3'-tailed dsDNA (63/32-mer) as the substrate, DNA binding by the Rad51B/Rad51C complex was also enhanced over the individual proteins (data not shown). However, no stimulatory effect was found with mixed Rad51 and Rad51B compared to individual Rad51 or Rad51B (data not shown). It is likely that the direct interaction between Rad51 and Rad51C, as well as between Rad51B and Rad51C, is responsible for their increased ability to bind DNA.
Both Rad51B and Rad51C exhibit DNA-stimulated ATPase activity -Amino acid sequence analysis indicates that the five human Rad51 paralogs contain the conserved ATPbinding domains (3), the Walker A and B motifs, suggesting ATP binding and hydrolysis activities for these proteins. Like human Rad51 and Rad51D (7,28), both Rad51B and Rad51C were found to display ATPase activity, and this activity was stimulated by DNA (Fig. 5A).
Single-stranded DNA was found to be a better stimulator of the ATP hydrolysis activity than dsDNA for both proteins (Fig. 5A). The effect of various ssDNA species on the activity of each protein was also determined, including φX174 ssDNA, M13 ssDNA, poly[dT], and a 63-mer oligonucleotide (Fig. 5B). Each ssDNA stimulated the activity of Rad51B and Rad51C, indicating that the secondary structure of ssDNA has no significant effect on the stimulation. The ATPase activity of Rad51B was shown to be dependent on the presence of a divalent cation, Mg 2+ , and the activity is Mg 2+ concentration-dependent -the ATPase activity of Rad51B increased with higher concentration of Mg 2+ . In addition, Mn 2+ was able to partially substitute for Mg 2+ (Fig. 5C). Similar effects by a divalent cation were also found for Rad51C (data not shown) and were reported for Rad51D (28).
The ATP hydrolysis activity of Rad51B and Rad51C in the presence of ssDNA was found to be lower than that of Rad51 (data not shown). When the ATPase activity of mixed proteins was examined (Rad51+Rad51B, Rad51+Rad51C, and Rad51B+Rad51C), an additive effect on the ATP hydrolysis activity was observed, rather than a stimulatory effect (data not shown).
Rad51C promotes an apparent DNA strand exchange reaction -Using oligonucleotide DNA substrates, we tested the ability of Rad51B and Rad51C proteins to promote DNA strand exchange in vitro. For this purpose, Rad51C nucleoprotein complexes were formed with ssDNA 63-mer (#2) and these complexes were mixed with dsDNA of three different lengths: 32, 48, and 63 base pairs. To detect the products of DNA strand exchange, the strand in dsDNA that was identical in sequence to the ssDNA (#2) was labeled (i.e. the strand that was expected to be displaced from the dsDNA). After incubation, we observed the appearance of the displaced ssDNA, the product of DNA strand exchange (Fig. 6A). Rad51C generated ~ 20-30% displaced ssDNA product after 30 min of incubation. As shown in Fig. 6B, the reactions work equally well with the 32, 48, and 63 base-pair dsDNA substrates. The activity of Rad51C protein only slightly depended on DNA length in the tested range of dsDNA substrate, with the activity increasing slightly with increasing dsDNA length. Consistent DNA strand exchange products were observed using three independent Rad51C protein preparations. In contrast to Rad51C, Rad51B protein did not show any activity under the conditions examined (data not shown).
For comparison, the DNA strand exchange activity of Rad51 was examined using the same DNA concentrations and assay conditions as used for Rad51C, except that the optimal concentration of magnesium acetate (20 mM) was used and the concentration of Rad51 (0.6 µM) was the optimal 1:3 stoichiometrical ratio relative to ssDNA. Under these conditions, Rad51 catalyzed the formation of ~15-35% of ssDNA product after 30 min reaction (Fig. 6C).
However, Rad51 shows a preference for the shorter dsDNA substrates. Thus, we conclude that Rad51C displays a significant apparent DNA strand exchange activity, which is comparable to that of Rad51.
The apparent DNA strand exchange promoted by Rad51C is protein-concentration and magnesium ion-dependent but not ATP-dependent -To determine the reaction stoichiometry between Rad51C and ssDNA, the DNA strand exchange reaction was carried out with a fixed amount of ssDNA (1.5 µM) and various amounts of Rad51C. Using 63-mer ssDNA and dsDNA substrates, the reaction yield was found to increase with increasing concentrations of Rad51C, and saturated at ~5.5 µM (Fig. 7A). This amount greatly exceeds the classic 1:3 stoichiometry for Rad51 and DNA, and it indicates that the binding stoichiometry of Rad51C and DNA is different from that of the Rad51 protein.
Because both RecA and Rad51 proteins (7) require a nucleotide cofactor for DNA strand exchange, we tested whether this requirement applied to the Rad51C-mediated reactions. We found that, contrary to expectations based on the RecA/Rad51 family members, the reaction promoted by Rad51C protein is independent of a nucleotide cofactor (Fig. 7B). We also found that AMP-PNP had no stimulatory effect on the reactions (Fig. 7B).
Since the DNA strand exchange activity of Rad51 shows a strong dependence on magnesium ion concentration, we examined the effect of magnesium acetate concentration on the activity of Rad51C. A titration experiment (Fig. 7C) shows that the Rad51C-promoted reaction exhibits a dependence on magnesium ion concentration, with maximal activity at 3 mM of magnesium that remains approximately constant to 20 mM. Because ammonium sulfate was found to stimulate the DNA strand exchange activity of Rad51 (42), we also tested its effect on Rad51C. However, under our conditions, 100 mM ammonium sulfate instead inhibited the reaction promoted by Rad51C (Fig. 7C). A second possibility is that Rad51C is simply destabilizing and melting the dsDNA, resulting in the production of a free ssDNA strand that in a reaction does not require the homologous ssDNA as a strand exchange partner. Therefore, the homology dependence of the reaction was examined.
We found that no ssDNA was produced when the DNA substrates were heterologous; in comparison, product formation was ~30% when the partner DNA was homologous (Fig. 8B).
This result demonstrates that Rad51C protein-promoted DNA strand transfer is homologydependent. Finally, we were concerned that Rad51C might still be melting the dsDNA, but that the free ssDNA strands would quickly re-anneal after deproteinization. Hence, the reaction would appear homology-dependent, but only because the homologous ssDNA was acting to trap the strand-separated DNA duplex. To investigate this possibility, the standard homologydependent experiment was carried out, but in addition, an excess amount of homologous ssDNA was added to the stop buffer. This added homologous ssDNA, although complementary to the displaced [ 32 P]-labeled ssDNA product (#2), was of a different length (79-mer) than the original ssDNA (63-mer) present in DNA strand exchange reaction. Thus, pairing that might occur during the deproteinization step would be distinguished from DNA strand exchange with the intended partner as diagramed in Fig. 8C, panel I. We found that, in the absence of 79-mer ssDNA in the stop/quench mix, a displaced 63-mer ssDNA product was generated as observed before (Fig. 8C, panel II, lane 1). When the 79-mer ssDNA was present, the displaced 63-mer ssDNA quickly annealed with 79-mer ssDNA to form a 79/63-mer duplex (Fig. 8C, lane 2) Because Rad51C was found to associate with Rad51, we were concerned that the observed DNA strand exchange activity might originate from contaminating Rad51 protein. To exclude the possibility that the purified Rad51C (and Rad51B) preparation contains Rad51 from insect cells, the same amount of the purified Rad51B and Rad51C preparation used in all assays was examined by probing with Rad51 antibody. The α-Rad51 recognized the insect Rad51 in control samples from Sf9 cells, but no Rad51 was detected in either preparation (data not shown), indicating that the DNA binding, ATPase, and apparent DNA strand exchange activities are intrinsic to Rad51B and Rad51C proteins, rather than due to Rad51 contamination.

DISCUSSION
To establish a purification strategy for the Rad51 paralogs, we discovered that spermidine selectively precipitates overexpressed Rad51C from cell extracts, resulting in ~ 70% purity of the protein. This step greatly facilitated the purification of Rad51C protein. Based on our unpublished observations, Rad51B also favorably precipitates with spermidine. Thus, in addition to purification of RecA and Rad51 as described previously (32)(33)(34)(35)(36)(37)(38), spermidine precipitation is demonstrated here as a powerful strategy for purification of Rad51-related proteins. Our results suggest that this strategy might work for purifying other untagged Rad51 paralogs. However, since spermidine also precipitates DNA, it is important to include a Q-Sepharose column to eliminate DNA in the purification process.
In previous studies using a yeast two-hybrid assay and baculovirus co-expression with Ni-NTA pull-down analysis (24), we reported that Rad51B directly interacts with Rad51C, and that Rad51C weakly interacts with Rad51. Here we demonstrate a simultaneous interaction among these three proteins. This is the first in vitro evidence for Rad51-Rad51C-Rad51B complex formation. However, in the course of our study, it was reported that overexpressed 6x(His)-tagged Rad51C did not pull down Rad51 in human cells (43,44). Because we observed that the interaction between Rad51C and Rad51 is weak -only ~5-10% of the overexpressed Rad51 was pulled down by the Rad51B-Rad51C complex in insect cell extracts as well as by purified complex of Rad51B and Rad51C proteins -the failure to detect the Rad51C-Rad51 interaction in human cells (43,44) is likely due to the weak interaction between the Rad51B-Rad51C complex and Rad51.
To elucidate the mechanism of homologous recombination involving Rad51 and Rad51 paralogs and to understand the functions of each protein, the biochemical activities of these proteins were determined. Using established in vitro assays for Rad51, including DNA binding, ATPase and DNA strand exchange assays, the biochemical properties of Rad51B and Rad51C were determined and compared to those of Rad51.
To catalyze DNA strand transfer, Rad51 must first bind to ssDNA to form a nucleoprotein filament (9,10). Since the Rad51 paralogs play a role in HRR, a key question is whether the paralogs are involved in the DNA binding step. We found that four of the five Rad51 paralogs, including Rad51B, Rad51C, Rad51D and XRCC3, contain a non-sequence-specific DNA binding domain -a helix-hairpin-helix motif (39,40). The DNA binding ability of each paralog was assessed using an electrophoretic mobility-shift assay. In addition to the previously reported DNA binding activity of Rad51D (28), we demonstrated that both Rad51B and Rad51C proteins exhibit DNA binding capacity. This finding suggests a role for the Rad51 paralogs in the DNA binding step. Since after DSB formation the break sites of DNA are nucleolytically processed to produce ssDNA with a 3'-overhanging end for binding by RPA, Rad51 and other associated proteins, we examined the substrate binding preferences of Rad51B and Rad51C. Our data, showing that both Rad51B and Rad51C proteins bind the 3'-end tailed dsDNA substrates in preference to either ssDNA or dsDNA, support the notion that the Rad51 paralogs bind the damaged sites and play a role in the repair processes. The observation that the DNA binding activity of Rad51C is higher than that of Rad51 is somewhat interesting. It may suggest that Rad51C plays a crucial role in the DNA binding processes of the HRR-related proteins. Another significant finding is that the complex formed by Rad51 and Rad51C, as well as by Rad51B and Rad51C, displays a higher DNA binding activity than that of the individual proteins. This result suggests that the direct protein-protein interaction serves possibly to recruit Rad51, and perhaps other proteins as well, to the processed ssDNA.
Alignments of the amino acid sequences of Rad51 and five Rad51 paralogs revealed that all proteins contain the Walker A and B motifs (3), suggesting ATP binding and hydrolysis capability. The ATPase activity of each paralog was confirmed in vitro. We showed that purified Rad51B and Rad51C proteins display ATPase activity that is stimulated by DNA. Little activity was observed in the absence of DNA, implying that DNA is the key cofactor in activating ATP binding/hydrolysis by Rad51B and Rad51C. The role of ATP binding and hydrolysis by the Rad51 paralogs in HRR is still unclear. It was reported that the ATPase mutant of XRCC2 does not affect the formation of Rad51 foci (45), indicating that XRCC2 does not use ATP binding or hydrolysis to promote its function. Whether the ATPase mutants of Rad51B and Rad51C are altered in their DNA binding activity remains to be determined.
Thus far, Rad51 is the only protein found to catalyze the key reactions of homologous pairing and DNA strand transfer that are needed by HRR to repair DNA double-strand breaks in human mitotic cells. Current evidence supports a mediating role for Rad51 paralogs in HRR, where these proteins may function as accessory cofactors assisting the action of Rad51 (4).
During the course of our research, the Rad51B-Rad51C complex was purified in the Sung laboratory (46). They reported that the Rad51B-Rad51C complex acts to facilitate the replacement of RPA from the nucleofilaments by Rad51 and to promote the DNA strand exchange activity of Rad51 (46), and they suggest a mediator role for the Rad51B-Rad51C complex in HRR. We found, using oligonucleotide DNA substrates that Rad51C, but not Rad51B, promotes a DNA strand exchange reaction. After further investigation, we discovered that the observed DNA strand exchange products were generated by a DNA melting and strand separation activity of Rad51C. This result implies a distinct role for Rad51C in HRR and suggests that Rad51C may have a more direct function in addition to its mediator role. It is very possible that Rad51C recruits Rad51 to ssDNA, and also mediates the separation of duplex DNA to promote DNA strand exchange. This proposed function is based on several aspects of our work: i) Rad51 and Rad51C (with and without Rad51B) physically interact; ii) Rad51 and Rad51C show synergy in their DNA binding affinity; and iii) Rad51C promotes dsDNA separation for 32, 48, and 63 base-pair substrates equally well, but Rad51 acts preferentially on the shorter substrates, suggesting that Rad51C might assist in the melting of longer duplex regions. Together, these results demonstrate that Rad51C behaves very differently from Rad51, indicating the Rad51C may serve a different role from Rad51 in the DNA strand exchange stage of homologous recombination. It is most likely that Rad51C "helps" with the strand exchange by Rad51 in several ways: Rad51C may not only directly bind DNA at sites of DSB but also facilitate the DNA binding of Rad51 by complex formation, and subsequently catalyze the separation of duplex DNA for strand exchange.
Among the protein-protein interactions between the five Rad51 paralogs, Rad51C has been shown to be a central player that interacts directly with Rad51B, Rad51D, XRCC3, and weakly with Rad51 (24), and which is present in various multiprotein complexes in human cells, including Rad51B-Rad51C, Rad51C-XRCC3 and Rad51B-Rad51C-Rad51D-XRCC2 (43,44,(46)(47)(48). The evidence of multiple protein interactions supports the notion that Rad51C may play multiple roles in the recombinational repair processes. In addition, two hamster cell lines that are mutated in Rad51C, irs3 and CL-V4B, were recently identified, and these cells show a reduction in sister chromatid exchange and genomic stability (49,50), emphasizing a key role for Rad51C in HRR. Here we described several biochemical properties that help define the functions of Rad51C in this process, including complex formation, DNA binding, ATP hydrolysis and DNA melting/separation activities. These findings underpin the significant biological function of Rad51C protein in the DNA strand exchange events of homologous recombination.     Each value represents an average from two independent determinations. C, The ATPase activity of Rad51B is dependent on a divalent cation. Rad51B was incubated with φX174 ssDNA and three different amounts of Mg 2+ ( ) or 2.5 mM of Mn 2+ (▲) at 37°C for 2 hr and the ATP hydrolysis activity was determined.  Nucleoprotein complexes were formed by incubating 1.5 µM of ssDNA (63-mer, #2) with Rad51C protein at concentrations of 0, 1, 3, 4, 6, or 9 µM. To initiate DNA strand exchange, these complexes were mixed with 3 µM of 32 P-labeled 63 base-pair dsDNA (#1-2) and incubated for 30 min at 37°C. The products of DNA strand exchange were determined as described in "Experimental Procedures", and plotted. B, The reaction catalyzed by Rad51C protein is ATPindependent. The reaction conditions of DNA strand exchange were identical to those described in Fig. 7A, except that 6 µM of Rad51C was used and the reactions were assayed in the presence (▼) or absence (▲) of ATP, or ATP was replaced with AMP-PNP (◆), when indicated. The control reaction without protein (■) was carried out under identical conditions. The aliquots were withdrawn from the reaction mixtures at the indicated time points, deproteinized and analyzed by electrophoresis in a 10% polyacrylamide gel run with TBE buffer. C, The reaction catalyzed by Rad51C protein is dependent on the magnesium ion concentration. The reaction conditions for DNA strand exchange were identical to those described in Fig. 7A, except that the Rad51C protein concentration was 6 µM and assayed in the presence of 100 mM ammonium sulfate or increasing concentrations of magnesium acetate.