Direct association between the yeast Rad51 and Rad54 recombination proteins.

The RAD54 and RAD51 genes are involved in genetic recombination and double-strand break repair in the yeast Saccharomyces cerevisiae. The Rad51 protein is thought to be a yeast analogue of the Eschericia coli recA gene product and catalyzes strand exchange between homologous single- and double-stranded DNAs in vitro. RAD54 exhibits homologies to several known ATPases and is a member of the SWI2/MOT1 family. We show here that the Rad54 protein interacts with the Rad51 protein in vivo and in vitro and that the NH2-terminal 115 residues of the Rad54 protein are necessary for this interaction. Combined with previously reported results, these data imply that the Rad54 protein is part of a multiprotein yeast recombination complex.

The RAD54 and RAD51 genes are involved in genetic recombination and double-strand break repair in the yeast Saccharomyces cerevisiae. The Rad51 protein is thought to be a yeast analogue of the Eschericia coli recA gene product and catalyzes strand exchange between homologous single-and double-stranded DNAs in vitro. RAD54 exhibits homologies to several known ATPases and is a member of the SWI2/MOT1 family. We show here that the Rad54 protein interacts with the Rad51 protein in vivo and in vitro and that the NH 2terminal 115 residues of the Rad54 protein are necessary for this interaction. Combined with previously reported results, these data imply that the Rad54 protein is part of a multiprotein yeast recombination complex.
The RAD52 epistasis group includes genes involved in homologous recombination in the yeast Saccharomyces cerevisiae. Mutations in these genes result in phenotypes that include an inability to repair double-stranded breaks, as well as defects in mitotic and meiotic recombination (1)(2)(3)(4). The Rad51, Rad55, and Rad57 proteins show considerable homology to the Eschericia coli RecA protein, the paradigmatic prokaryotic strand transferase. Indeed, Rad51 protein has been shown to mediate strand exchange in vitro between homologous single-and double-stranded DNAs in the presence of replication protein A (RPA), 1 the yeast single-stranded DNA-binding protein (5).
A number of results suggest that the Rad51 protein functions as part of a multiprotein complex in vivo. For example, the Rad51 and Rad52 proteins have been shown to bind one another both in vivo (6) and in vitro (7), and there is genetic evidence that Rfa1 is associated with the putative recombination complex (8,9). Furthermore, experiments from the Berg (Stanford University School of Medicine) and Symington (Columbia College of Physicians and Surgeons) laboratories demonstrated that Rad51 protein binds to the Rad55 protein in vivo, which in turn interacts with Rad57 protein (10,11).
In this report, we present both in vivo and in vitro evidence for a direct association between the Rad51 protein and the RAD54 gene product, another member of the RAD52 epistasis group (12,13). The discovery of a Rad54-Rad51 protein inter-action provides further support for the existence of a "protein machine" for mitotic recombination in yeast and raises the possibility that the Rad54 protein could play a direct role in strand exchange.
RAD54 (A) was cloned into pKM260, to generate pHis6RAD54(A) for E. coli expression. pKM260 (kindly provided by Drs. Stephen Johnston and Karsten Melcher, University of Texas Southwestern Medical Center) was constructed by inserting a 330-base pair BglII/EcoRV fragment derived from pET-14b and including the promoter through transcriptional termination sequences, into BamHI/EcoRV-cut pBR322. The pET-14b fragment was modified such that a TEV protease cleavage site located between NheI and NcoI sites replaced that of thrombin. The final construct incorporated an NH 2 -terminal 6-histidine tag followed by the cleavage site for the highly specific TEV protease. All the clones were sequenced at both the NH 2 and COOH terminus of the gene.
The RAD51 gene was amplified from genomic S. cerevisiae DNA (strain S288C) by PCR using partially complementary primers with the unique restriction sites NcoI and BamHI. The resultant DNA fragment was cleaved with NcoI and BamHI and ligated into NcoI/BamHI-cut pKM260 to give pKHHis6-51. The entire gene was sequenced and compared to the published sequence to ensure that no mutations had occurred during amplification. S10 epitope-tagged derivatives of His6Rad51 and His6Rad54 proteins were constructed by inserting an oligonucleotides encoding the 11-amino acid S10 epitope (Met-Ala-Ser-Met-Gly-Gly-Gln-Gln-Met-Gly) into NheI/NcoI-cleaved pKHhis6Rad51 and pHis6Rad54. This replaced the TEV site with the epitope tag. The coding sequence of these genes were then amplified using the PCR and inserted into HindIII/BamHIcleaved pYES2.0 to provide pYESS10Rad51 and pYESS10Rad54, respectively.
The S10-tagged Rpa1-encoding plasmid was made by ligating into NheI/NcoI-cut pKM260 an S10-encoding oligonucleotide. This product * This research was supported in part by grants from the Welch Foundation (to T. K.) and the National Institutes of Health (to R. R.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
All S10-tagged genes were expressed in yeast from the GAL1/10 promoter of the parent pYES 2.0 plasmid. The S10-tagged derivatives of Rad51 and Rad54 complemented the radiation-sensitivity of the corresponding mutant strain (Fig. 3), although at very high doses, the strain expressing S10-Rad54 was somewhat more sensitive than wild-type yeast. The ability of S10-Rpa1 to complement was not tested, since the RPA1 gene is essential.
The procedure employed for the liquid assay was as follows. Colonies were grown in 10 ml of SC-Leu-Trp medium containing 3% glycerol and 2% lactic acid as the carbon source until the OD 600 reached 0.6 -0.8. Cells were then centrifuged at 5,000 rpm for 10 min. The pellet was resuspended in 200 l of cold extraction buffer (20 mM HEPES, pH 7.5, 1 mM dithiothreitol), and vortexed with 200 l of cold 0.4-mm glass beads for 1 min. The resulting crude extract was centrifuged at 13,000 rpm for 20 min. The supernatant from the spin was assayed for ␤-galactosidase activities by the O-nitrophenyl-␤-D-galactopyranoside method (16). The activities shown in Table I were based on the average of three colonies.
Purification of His6Rad51 Protein from E. coli-pKHHis651 was transformed into E. coli BL21(DE3) pLysS cells. A 10-liter culture was grown in L-broth media supplemented with 100 mg/ml ampicillin and 20 mg/ml choramphenicol at 37°C to an OD 600 of 0.5, then supplemented with glucose to 0.1% and induced with isopropyl-1-thio-␤-Dgalactopyranoside (1 mM final concentration). The cells were grown for an additional 4 h at 30°C, then harvested by centrifugation and stored at Ϫ70°C. 35 g of cell paste was resuspended in 60 ml of CS buffer (40 mM Tris-Cl, pH 8.0, 200 mM NaCl, 10% glycerol) supplemented with 5 mM imidazole, 10 mM MgCl 2 , 1 mM phenylmethylsulfonyl fluoride, 1 mM benzamidine, 0.5% Triton X-100, and 34 g/ml DNase I. Cells were incubated on ice for 30 min, then homogenized with a glass/Teflon homogenizer. All subsequent steps were carried out at 4°C. The lysate was cleared by centrifugation at 35,000 rpm for 1 h in a Beckman L8 -70M microcentrifuge using a Ti 50.2 rotor.
The supernatant was loaded onto a 30-ml chelating Sepharose Fast Flow (Pharmacia) column pre-equilibrated with NiSO 4 according to the manufacturer's instructions. The column was washed with 90 ml of CS buffer supplemented with 5 mM imidazole, 300 ml of CS supplemented with 60 mM imidazole, then eluted with a 150-ml linear gradient of 60 -500 mM imidazole in CS buffer. His6Rad51 protein-containing fractions were pooled based on purity (determined by SDS-polyacryamide gel electrophoresis) and dialyzed into DA buffer (20 mM Tris acetate, pH 8.0, 1 mM dithiothreitol, 10% glycerol) containing 90 mM potassium acetate, pH 7.0.
This fraction was loaded onto a 45-ml single-stranded DNA cellulose column with a capacity of 1 mg/ml single-stranded DNA. The column was washed with 500 ml of DA 90 buffer, then eluted with 250 ml of linear gradient of DA 90 to DA 500 . ATP hydrolysis assays were performed across the protein peak and fractions were pooled based both on ATP hydrolysis activity and SDS-polyacrylamide gel electrophoresis. The pool was dialyzed into P buffer (200 mM NaCl, 10% glycerol, 1 mM dithiothreitol) containing 20 mM potassium phosphate, pH 7.1.
This fraction was loaded onto a 6-ml hydroxylapatite (Bio-Rad):CF11 (Whatman) (1:1) column and washed with 100 ml of P 20 buffer. The protein was eluted with a 30-ml linear gradient of P 20 to P 300 . ATP hydrolysis assays were performed and fractions pooled accordingly. The pool was dialyzed into storage buffer (20 mM Tris-Cl, pH 8.0, 1 mM EDTA, pH 8.0, 1 mM dithiothreitol, 100 mM NaCl, and 10% glycerol). The final fraction was Ͼ95% homogeneous as judged by Coomassie Brilliant Blue staining. In overloaded lanes, small amounts of proteolytic fragments were the only visible contaminants. No single-stranded DNA-or double-stranded DNA-degrading nuclease activities were detected in the final fraction. NH 2 -terminal amino acid sequencing confirmed the identity of the purified protein. Approximately 10 mg of His6Rad51 protein/liter of liquid culture was obtained. The metalbinding tag was cleaved from His6Rad51 protein to give GAMG-Rad51 by incubation of a 5-fold molar excess of the protein with the TEV protease for 5 min as described (17).
Purification of His6Rad51 Protein from Yeast-S. cerevisiae reg1-501 cells containing pKHYesHis651 were grown on SC (1.7 g of yeast nitrogen base (Life Technologies, Inc.), 5 g of ammonium sulfate, and 20 g of Bacto-agar (Life Technologies, Inc.) per liter) Ura-Leu drop-out plates at 30°C using glucose as the carbon source. A single colony was picked and grown in SC ura-leu drop-out media overnight. The culture was diluted 1:500 in SC ura-leu drop-out media and grown until the OD 600 was between 0.8 and 1.0. The cells were induced by the addition of galactose to 2%.
The cell pellet was resuspended in 1 volume of CS buffer supplemented with 5 mM imidazole, 10 mM MgCl 2 , 1 mM phenylmethylsulfonyl fluoride, 1 mM benzamidine, 25 g/ml bestatin, 2 g/ml leupeptin, and 1 g/ml pepstatin A. The suspension was lysed by French Press (SLM Aminco). Cell debris was removed by centrifugation in a Beckman JA-17 rotor run an 14,000 rpm for 1 h. The protein was then purified using the procedure described above for His6Rad51 protein expressed in E. coli, except that a DE-52 column was substituted for the hydroxylapatite column. The DNA-cellulose pool was loaded in a buffer containing 100 mM NaCl and, after thorough washing, eluted with a buffer containing 400 mM NaCl. Approximately 150 g of purified His6Rad51 protein per liter of liquid culture was obtained.
Purification of Rad54⌬BamHI-The RAD54 coding sequence was amplified by PCR and cloned into pKM260 to give pKHHis6RAD54. Efforts to express full-length Rad54 protein in E. coli from this vector were unsuccessful. To produce a derivative of Rad54 protein, pKHHis6RAD54 was cleaved with BamHI, and after removal of the internal fragment, religated to give pKHHis6RAD54(⌬BamHI). Deletion of the BamHI fragment deletes the region of RAD54 encoding residues 115-750, but retains the reading frame. The protein, His6Rad54(⌬BamHI), was expressed at high levels in E. coli containing the expression plasmid, albeit in insoluble form. The polypeptide was purified to apparent homogeneity under denaturing conditions by chelating nickel-Sepharose chromatography and denaturing electrophoresis using a Bio-Rad Prep Cell. NH 2 -terminal sequencing confirmed the identity of the polypeptide. The concentrations of each protein was determined by absorbance at 280 nm in denaturing solutions as described (18).
Far-Western Blot Analysis-The protocol employed was slightly modified from that employed by Horiuchi et al. (19). Proteins were electrophoresed through a denaturing polyacrylamide gel and subsequently stained with Coomassie Blue or transferred to polyvinylidene difluoride membrane (Millipore). The membranes were then incubated in 8 M urea containing 1% 2-mercaptoethanol (v/v). Renaturation of the proteins was achieved through stepwise dilutions of the urea concentration in FW buffer (20 mM Tris-HCl, pH 7.5, 60 mM NaCl, 0.1 mM EDTA, 10 mM MgCl 2 , 5% (w/v) glycerol, 0.02% Nonidet P-40) until the final urea concentration was below 10 mM. The membranes were blocked in FW buffer containing 5% non-fat dry milk, and then incubated in FW buffer (containing 2% non-fat milk) with TEV-cleaved His6Rad51 protein (purified from yeast) (2 mg/ml) or, as a control, in the same buffer lacking the Rad51 protein. The bands which retained the Rad51 protein were detected by probing the membrane with mouse polyclonal antibodies raised against TEV protease-cleaved His6Rad51 protein (1 h at room temperature) followed by a goat anti-mouse antibody conjugated to horseradish peroxidase (1 h) and finally, chemiluminescence reagent (DuPont) (1 min at room temperature). The membrane was washed with FW buffer between each incubation.
Preparation of the Rad51 Protein Affinity Column-Affi-Gel 10 beads (Bio-Rad) were washed with 5 volumes of double-distilled H 2 O, then twice with 5 volumes of FIX buffer (100 mM potassium phosphate, pH 6.5, 10% (w/w) glycerol). His6Rad51 protein was dialyzed into the same buffer. The beads were mixed with the protein and agitated gently on a rotorary table at 4°C. Typically, 5 ml of a 6 mg/ml protein solution was mixed with 5 ml of packed beads. The progress of the reaction was checked periodically by monitoring the A 280 of the supernatant. When this had decreased 15%, the reaction was terminated by the addition of a large excess of ethanolamine, pH 7.4. The same method was employed to couple bovine serum albumin (BSA) to Affi-Gel 10. The beads were washed thoroughly with Bind 500 buffer (see below) and stored at Ϫ20°C.
Protein Affinity Chromatography-Columns were constructed from 1-ml syringes plugged with glass wool. The Rad51 protein beads (0.9 mg of protein/packed ml) were added to the column to give a packed volume of 0.3 ml. A 0.3-ml BSA control column (6 mg of protein/packed ml) was also constructed. The columns were then washed extensively with Bind 2000 buffer (Bind buffer: 40 mM Tris-HCl, pH 7.4, 1 mM Na 2 EDTA, 1 mM dithiothreitol, 2 mM MgCl 2 , 10% glycerol, 1 mM phenylmethylsulfonyl fluoride, 1 mM benzamidine hydrochloride, supplemented with NaCl as indicated in the subscript in millimolar), then equilibrated with Bind 50 buffer.
Lysates from cells expressing S10 epitope-tagged derivatives of the Rad51, Rad54, and Rpa1 proteins were made according to the procedure of Melcher and Johnston (20). The lysates were cleared by centrifugation at 13,000 rpm in a benchtop microcentrifuge and passage through a 0.44-micron filter (Millipore), mixed together (ratio of lysates containing S10-Rad54, S10-RPA, S10-Rad51 ϭ 10:3:1) in Bind 50 buffer, and applied simultaneously to the Rad51 and BSA columns at a flow rate of 1 column volume/h. Each column was then washed with 10 column volumes of Bind 50 buffer, then eluted with 1 column volume of Bind 500 buffer. The beads were then removed from the column and boiled in SDS-containing gel loading buffer to release bound proteins not eluted by the 500 mM salt wash. Each fraction was made 50 g/ml in lysozyme, then the total protein was precipitated by the addition of sodium deoxycholate to 0.5 mg/ml and trichloroacetic acid to 20%, followed by incubation on ice for 1 h and centrifugation at 13,000 rpm for 30 min in a benchtop microcentrifuge at 4°C. The precipitate was washed with acetone and diethyl ether then boiled in sample loading buffer for 3 min and applied to a 7.5% denaturing polyacrylamide gel. The proteins were analyzed by Western blot using monoclonal antibodies that recognize the S10 epitope (Novagen).
Complementation of X-ray Sensitivity-Cells were irradiated on solid selection media (galactose as the carbon source) with x-rays from a 60 Co source at a dose rate of 115 rad/s. After 3-4 days growth, cell numbers were counted and compared to the number of cells growing on plates that had not been irradiated.

RESULTS
The Rad54 Protein Binds the Rad51 Protein in Vivo-We employed the yeast two-hybrid system (14,21) to probe for protein-protein interactions between some of the known RAD gene products. Plasmids were constructed in which the Gal4 AD or DBD was fused to the amino terminus of various RAD genes. These constructs were transformed into a tester strain carrying a Gal4-responsive lacZ reporter gene. Control experiments in which only the DBD fusion, but not the AD fusion, was introduced into the cell showed that none of the DBDrecombination protein fusions alone acted as an activator. The results of these two-hybrid experiments are shown in Table I.
The data indicate that the Rad51 protein binds to the Rad52 protein and to itself, as reported previously by others (6). The levels of reporter gene expression recorded in cell lysates containing these fusions (47 units for Rad51-Rad51 and 20 units for Rad51-Rad52) were well above background (0.2 units).
A novel result from these experiments was the discovery of an apparent interaction between the Rad51 and Rad54 proteins, as evidenced by the high level of ␤-galactosidase (␤-Gal) activity detected in cells bearing DBD-Rad54 and AD-Rad51 fusions. Rad54 is a large protein of unknown function with some homology to DNA-dependent ATPases (see "Discussion"). In the reciprocal experiment in which the Rad54 protein was fused to the AD and the Rad51 protein to the DBD a much lower level of ␤-Gal activity was observed. A similar result was obtained for the well characterized Rad51-Rad52 interaction. These observations suggest that fusion of the Rad51 protein to the DBD interferes with its interactions with other recombination proteins, although apparently not with self-association.
No physical interaction between the RPA1 gene product, the yeast single-stranded DNA-binding protein, and any of the Rad proteins tested was indicated by the two-hybrid data, although Rpa strongly stimulates Rad51-mediated strand exchange in vitro (5). However, this negative result should be interpreted with caution. It might be that stoichiometric amounts of the other two components of heterotrimeric Rpa (22) are required for a stable interaction, or that fusion of Rpa1 protein to the GAL4 DBD interferes with its binding to other factors.
The Rad51 Protein-binding Domain Maps to the NH 2 Terminus of the Rad54 Protein-Many eukaryotic proteins contain independent functional domains. To investigate whether Rad51 protein binding activity could be mapped to a particular region of the Rad54 protein, fusion constructs were made in which the GAL4 DBD was linked to various fragments of Rad54 (Fig. 1). They were then employed in two-hybrid experiments along with the AD-Rad51 fusion. None of the Rad54 fragment fusions were themselves able to stimulate lacZ transcription. Rad54(A) (residues 1-327) and Rad51 protein provided a strong positive signal, while fragments lacking the NH 2 -terminal region did not. These data show that residues in the NH 2 -terminal region of Rad54 protein are important in binding Rad51 protein while the COOH terminus is not essential for this interaction. However, we cannot exclude the possibility the COOH-terminal region of Rad54 protein plays a secondary role in Rad54-Rad51 interactions, since we were unable to demonstrate that these fusions were stably expressed. Quantitation of the in vivo expression levels of these various constructs was attempted by Western blotting using antibodies raised against a fragment of Rad54 protein (see below), but no signal was detected in any case, including that of the full-length protein. It may be that the fusions were expressed at very low levels, that they were proteolytically unstable upon cell lysis, or that fusion with the Gal4 DBD interfered with antibody binding.
The Rad51 Proteins Binds to the NH 2 -terminal 115 Residues of Rad54 Protein in Vitro-While the two-hybrid assay is useful as an indicator of protein-protein interactions, it is important to check results obtained in these experiments by an independent method. Furthermore, it is conceivable that GAL4 activity could be reconstituted by an indirect Rad51-Rad54 protein interaction bridged one or more endogenous factors. Therefore, in vitro experiments were employed to probe Rad54-Rad51 protein-protein interactions.
To facilitate these studies, large quantities of a Rad51 protein derivative containing a 6-histidine tag and a specific protease (TEV protease) cleavage site (17) was expressed in both yeast and E. coli and purified. The fusion protein contains 15 non-native residues between the NH 2 -terminal methionine and the Rad51 amino acids (Fig. 2). To ensure that these amino acids did not interfere with the function of the Rad51 protein, we expressed this fusion (henceforth called the His6Rad51 protein) in a rad51⌬ yeast strain and asked if the fusion was able to complement the severe x-ray sensitivity of this strain. As shown in Fig. 3, this was the case, demonstrating that the His6Rad51 protein was active in vivo. The presence of the 6-histidine tag facilitates purification and allows large quantities of the protein to be obtained easily (see Fig. 2 and "Materials and Methods"). The rate of DNA-dependent ATP hydrolysis catalyzed by His6Rad51 protein was measured (data not shown) and found to be essentially identical to that reported for native Rad51 protein (5,23), again arguing that the NH 2terminal fusion does not compromise the activity of the protein.
His6Rad51 protein was cleaved efficiently by purified TEV protease (17) in vitro, resulting in a polypeptide lacking the metal-binding tag and containing only four non-native residues, GAMG (Fig. 2). Unfortunately, many attempts to express large quantities of Rad54 protein in E. coli, S. cerevisiae, and Pichia were unsuccessful. 2 We therefore turned to the expression of fragments of Rad54 protein. Two polypeptides were expressed, the 327-residue NH 2 -terminal fragment, Rad54(A), and Rad54⌬BamHI, a fusion containing a large deletion of the central region of the protein. The corresponding gene was constructed by eliminating the internal BamHI fragment in the RAD54 gene and religating (this retains the reading frame). This derivative lacks the ATPase homology regions, but retains the NH 2 -terminal 115 residues and the COOH-terminal 182 amino acids.
Expression of each Rad54 protein fragment as 6-histidine fusions in E. coli BL21DE3 cells under the control of a T7 promoter resulted in the production of large amounts of protein, most of which was insoluble. Each was purified under denaturing conditions, the Rad54⌬BamHI protein to near homogeneity and Rad54(A) protein to approximately 90% purity (see Fig. 4). Finally, mouse polyclonal antibodies were raised against the TEV protease-cleaved His6Rad51 protein and His6Rad54⌬-BamHI to facilitate their detection in crude lysates (see below).
With these tools in hand, a Far-Western blotting experiment (19) was employed to look for direct binding between the Rad51 protein and the Rad54 protein fragments. The results are shown in Fig. 4. Part A shows a Coomassie Blue-stained denaturing gel of the purified Rad54 fragments (lanes 1 and 3) and the crude lysates from which the purified proteins were derived (lanes 2 and 4). A control lysate lacking any Rad54-derived proteins is shown in lane 5. This gel was blotted onto nitrocellulose and the proteins renatured. The blots were then probed with either TEV-cleaved His6Rad51 protein (Fig. 4B) or with buffer as a control (data not shown). Mouse anti-Rad51 antibody, followed by goat anti-mouse antibodies conjugated with horseradish peroxidase were then used to probe the blot to detect the position of the bands that bound Rad51 protein in the first step. As shown in lanes 2 and 4 of Fig. 4B, Rad51 protein bound to both the purified Rad54(A) and Rad54⌬-BamHI polypeptides, respectively. More importantly, the Rad51 protein also recognized these Rad54 fragments selectively in the lysate lanes. While some E. coli proteins in these lanes "light up" (see also lane 5), they are also observed in the blot where the Rad51 protein probing step was omitted (data not shown). The background bands are therefore due to a low level of cross-reactivity of some bacterial proteins with the polyclonal anti-Rad51 antibody and/or the secondary antibody. These data strongly suggest that only the NH 2 -terminal 115 amino acids of Rad54 are absolutely required for binding Rad51 protein in vitro.
Detection of Rad54-Rad51 Binding by Protein Affinity Chromatography-Since the Far-Western experiment described above employed fragments of the Rad54 protein which had been denatured and renatured prior to exposure to Rad51, it remained an important goal to demonstrate in vitro interactions between the intact proteins. Protein affinity chromatography using His6Rad51 protein immobilized on Affi-Gel beads was employed for this purpose. Since native Rad54 protein is present at exceedingly low levels in yeast (50 molecules/cell or less) 3 and is difficult to detect with the polyclonal antibodies in hand, a vector was constructed that expressed an epitopetagged derivative of the protein. This species was clearly detectable by Western blotting using a commercially available monoclonal antibody directed against the S10 epitope. Several other plasmids expressing S10-tagged derivatives of yeast recombination proteins were also constructed, including Rpa1 and Rad51. S10-Rad51 and S10-Rad54 proteins were shown to complement the x-ray sensitivity of yeast strains lacking the wild-type proteins, although at high doses, cells expressing S10-Rad54 protein were somewhat more sensitive than wildtype yeast (Fig. 3).
Lysates from yeast expressing S10-tagged Rpa1, Rad51, and Rad54 proteins were mixed to provide approximately equal amounts of the three proteins and the mixture was passed over the Rad51 protein column as well as a BSA control column. After extensive rinsing, the columns were eluted with a buffer containing 500 mM sodium chloride. Finally, the column matrix was then boiled in SDS-containing buffer to release bound proteins not eluted by the high salt wash. Each fraction was concentrated then electrophoresed through a denaturing polyacrylamide gel. The fate of each S10-tagged protein was assessed by Western blotting.
As shown in Fig. 5, significant quantities of the S10-tagged Rad54 protein bound to the Rad51 column, but not the BSA column, indicating a specific association between the two proteins. All of the bound Rad54 protein was eluted in the 500 mM salt wash. Much lower levels of S10-tagged Rad51 and Rpa1 proteins were also detected in this fraction. Finally, large amounts of the S10-Rad51 protein were detected in the bead-denatured fraction, indicating a salt-stable His6Rad51-  3. Complementation of x-ray sensitivity by S10-tagged Rad51 and Rad54 proteins and His6Rad51 protein.

S10Rad51 association.
A slightly greater amount of S10-tagged Rpa1 was present in the 500 mM salt fraction of the Rad51 column, than the BSA control column, but the level of S10-Rpa1 retained was much lower than that observed for S10-Rad54. It is not clear if this represents a weak association of Rpa with Rad51 protein. It should be noted that the two-hybrid data do not suggest such an association in vivo.

DISCUSSION
Several genetic and biochemical techniques, including the two-hybrid system, protein affinity chromatography, and Far-Western blotting, have been employed to demonstrate a direct interaction between the yeast Rad51 and Rad54 recombination proteins. There is growing evidence that mitotic recombination in yeast is catalyzed by a multiprotein complex containing several RAD gene products, including Rad51, Rad52, Rad55, Rad57, and probably many other factors as well. Given the results reported here, and the fact that the RAD51 and RAD54 genes have been shown to be epistatic (24), it seems very likely that the Rad54 protein is also an integral part of this putative "recombinosome." The function of the Rad54 protein in recombination is unknown and the protein has yet to be purified and characterized biochemically. It is homologous to a number of ATPases and appears to be a member of a family that includes Swi2/Snf2 (25)(26)(27)(28) and MOT1 (29 -31), ATPases involved in transcriptional regulation. The Swi-Snf complex is thought to remodel chromatin structure and may influence transcription factor binding (32,33). MOT1 protein has been shown to be a TATAbinding protein-associated factor that disrupts TATA-binding protein-DNA complexes in an ATP-dependent fashion. This raises the exciting possibility that Rad54 may be involved in allowing the recombination machinery to access the DNA in a histone-occluded substrate. Indeed, Haber and colleagues (34) have shown that the Rad54 protein is required for an early step in mitotic recombination in vivo when one of the DNA partners is part of a relatively inaccessible chromatin structure (34). However, Rad54 protein is not required when the DNA is carried on a plasmid and is therefore part of a more accessible chromatin structure.
FIG. 5. Epitope-tagged Rad54 protein binds to a Rad51 protein affinity column. Lysates prepared from yeast strains expressing S10tagged Rad54, Rad51, and Rpa1 proteins, respectively, were mixed to provide approximately equal levels of the three proteins. The mixture was passed over a Rad51 protein affinity column and a BSA control column. After washing, bound proteins were eluted with a buffer containing 500 mM sodium chloride. Salt-stable interactions were detected by boiling the beads in a denaturing buffer. Considerable quantities of S10-tagged Rad54 and Rad51 proteins were retained by the Rad51 protein column, but not by the BSA column.