p53 Interacts with hRAD51 and hRAD54, and Directly Modulates Homologous Recombination¹²

HR⁴ is important for the generation of genetic diversity and for chromosome segregation during meiosis, as well as the repair of complex forms of DNA damage, such as double-strand breaks, interstrand cross-links, or DNA adducts in close proximity on opposite strands. Such damage can be induced by exogenous agents, like ionizing radiation, or it can arise from endogenous processes, like replication fork stalling during attempted replication over a single-strand break (1).

Maintenance of genetic stability likely relies heavily on the high fidelity of HR-mediated repair. Cells with mutations in HR genes (e.g., BRCA1, BRCA2, RAD51, RAD54, XRCC2, or XRCC3) exhibit high levels of genetic instability and sensitivity to cross-linking agents and ionizing radiation (2). Conversely, inappropriate activity of the HR pathway also could contribute to genetic instability by causing improper recombination (e.g., between repeat sequences or mutant alleles), resulting in translocations, deletions, duplications, or loss of heterozygosity (3, 4, 5, 6, 7). hRAD51 overexpression is observed in several tumor-derived cells (8, 9, 10, 11, 12) and correlates with elevated HR frequencies (8, 13, 14), which could account for some of these observed instabilities. In addition, some cancer-derived mutations have been reported for RAD51, RAD54, and RAD54B (a Rad54 paralog), and there are reports indicating loss of heterozygosity at the chromosomal locations of RAD51, RAD51B (a Rad51 paralog), RAD52, and RAD54, as well as BRCA1 and BRCA2 (reviewed in Ref. 2). Direct links between hRAD51 overexpression or HR gene modifications and the underlying genome instability in tumorigenesis remain to be established. However, these early reports are consistent with the idea that HR-mediated repair activity must be tightly regulated to maintain genetic stability (15).

The initial steps of HR involve recognition and preprocessing of the DNA lesions to form single-stranded DNA tails. Similar to RecA and yeast Rad51, hRAD51 can form polymers along single-stranded DNA, forming nucleoprotein filaments in vitro (16). hRAD51 also associates with hRAD54 (17). RAD54 belongs to the SWI2/SNF2 family of helicase-like proteins (18), which contain seven evolutionarily conserved motifs, including a DNA-dependent ATPase motif. Although hRAD54 does not possess classic helicase activity, it can alter DNA conformation (19, 20, 21). There is evidence that RAD54 can bind to RAD51 stoichiometrically and stabilize the protein-DNA complex, thereby stimulating strand invasion (22, 23). RAD51 also possesses DNA-dependent ATPase activity through a conserved Walker motif, which presumably catalyzes the pivotal homologous pairing and strand exchange reactions between the nucleoprotein filaments and the repair template DNA (24, 25).

Expression of mammalian RAD51 and RAD54 is induced in late G₁, S, and G₂ (18, 23, 26, 27). RAD51 forms nuclear foci in S phase cells (26, 28). Several other proteins implicated in HR, including BRCA1, BRCA2, RAD54, BLM, and RPA, bind to RAD51 and/or colocalize with RAD51 foci (17, 19, 29, 30, 31, 32, 33, 34). These foci probably represent large complexes of proteins engaged in, or attempting, HR-mediated DNA repair, because they increase in number and intensity after DNA damage induced by agents such as ionizing radiation, cross-linkers, and UV radiation. HR may be favored in S phase cells because of the presence of sister chromatids as proximal repair templates.

Mutations in the p53 tumor suppressor gene are common in a variety of human cancers (35). p53 can be activated by a variety of stress signals, including DNA damage (36), ribonucleotide depletion (37), hyperploidy (38, 39), hypoxia (40), activated oncogenes (41), and loss of cell adhesion (42). On stimulation, p53 can initiate cell cycle arrest, premature senescence, or apoptosis by activating transcription of several downstream genes (43, 44, 45, 46). On the basis of data derived from a transactivation-deficient mouse model, it has been suggested that transcriptional transactivation is the determining tumor suppressor function of p53 (47).

However, there is also evidence that p53 can initiate apoptosis in a transcription-independent manner (48, 49, 50). We have shown previously that this can occur through physical interactions of p53 with members of the DExH-containing DNA helicase superfamily, including the repair factors BLM, XPB, XPD, and WRN (51, 52, 53, 54, 55, 56). This superfamily also includes members of the SWI2/SNF2 family, such as CSB and RAD54 (18), with no demonstrated classic helicase activity. We showed previously that p53 can bind to CSB (52).

p53 also may help maintain genetic stability by modulating HR activity. p53-deficient cells often exhibit elevated frequencies of gene amplifications and karyotypic abnormalities, such as aneuploidy, deletions, inversions, and translocations (e.g., 57, 58, 59). In addition, mammalian cells in which p53 has been inactivated exhibit elevated spontaneous and induced HR (60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70), as well as DNA damage-induced sister chromatid exchange (71, 72, 73), a process that can be mediated by HR. Furthermore, p53 can bind directly or indirectly to several proteins implicated in HR, including BLM, BRCA1, BRCA2, and RAD52 (54, 74, 75, 76, 77). We showed recently that p53 can modulate the helicase activities of BLM and WRN (56), which are likely involved in recombinational repair (78). In addition, p53 can bind to and inhibit the HR factors hRAD51, RecA, and RPA (79, 80, 81). Interestingly, the primary binding site on hRAD51 is the putative homo-oligomerization domain, which may be necessary to form functional hRAD51 nucleoprotein filaments (16, 82). Several studies indicate that p53 modulation of HR activity is independent of its transcriptional activation function (65, 67, 70, 83).

In this report, we characterize p53 interactions with the HR factors hRAD51 and hRAD54. We demonstrate both physical interactions and colocalization of the proteins in nuclear foci. As a functional consequence of these interactions, we demonstrate p53-dependent modulation of HR frequencies in vivo. These findings indicate that p53 has a transcription-independent effect on HR via direct interaction with hRAD51 and hRAD54.

NHF strain GM07532 was obtained from Coriell Cell Repositories (Camden, NJ). LFS041 cells are postcrisis Li-Fraumeni fibroblasts that are p53 null. All of the cells were maintained in DMEM supplemented with glutamine, penicillin, streptomycin, and 10% fetal bovine serum. HU was obtained from Sigma, and NCS was obtained from the Drug Synthesis Branch of the National Cancer Institute. For HU treatments, cells were grown to confluency to attain synchronization in G₁, then subcultured at a lower density for 24 h so that most of the cells were in S phase. The cells were then either left untreated or treated with 1 mm HU for 16 h. For NCS treatments, asynchronous cells were treated at a concentration of 100 ng/ml for the indicated times.

cDNAs coding for wild-type hRAD51, the p53 nonbinding mutant of hRAD51 (186P), and the Walker motif dominant-negative mutant of hRAD54 (189R) were subcloned into and expressed from the pCLXSN retroviral vector (84). Subsequent sequencing revealed that hRAD51/186P also contains the following mutations, which did not substantially affect its ability to enhance HR in the host cell reactivation assay: 47R, 59G, and 122V. cDNAs with incompatible restriction sites were first subcloned into pJS-1, a shuttle vector produced by changing the BssHII sites flanking the polylinker of pBluescript to NotI sites, and then subcloned into a version of pCLXSN in which the EcoRI site had been converted to a NotI site. In this study, no retrovirus was produced; rather, the plasmids were transfected and the genes expressed under the control of the CMV promoter. Wild-type p53 and hRAD54 were expressed from the pcDNA3 expression plasmid (Invitrogen, Carlsbad, CA). The tumor-derived p53 mutants 175H and 248W were expressed from the pCMV-Neo-Bam expression plasmid. Dominant-negative RAD51 (dnRAD51 and SMRAD51) is a chimera of Saccharomyces cerevisiae and murine Rad51 (85), which was expressed from the pcDNA3.1 expression plasmid (Invitrogen). Transcriptionally inactive p53 (22/23) was expressed from pRCp53 (22, 23, 86).

GST-p53 NC encodes GST fused to full-length human wild-type p53. GST-p53 N5, NTD, 2C, 3C, 25, and 35, kindly provided by Bruce Stillman (Cold Spring Harbor Laboratory, Cold Spring Harbor, NY), encode GST fused to wild-type p53 amino acid residues 2–293, 2–124, 94–393, 155–393, 94–293, and 155–293, respectively.

The antibodies used in this study were: mouse monoclonal anti-p53 PAb421 (Ab-1), PAb1801 (Ab-2), and DO-1 (Ab-6; Oncogene Research Products); rabbit polyclonal anti-hRAD51 Ab-1 and mouse monoclonal anti-hRAD51 Ab-2 (Oncogene Research Products); rabbit polyclonal anti-phospho-p53 (Ser15; Cell Signaling Technology, Beverly, MA); and goat polyclonal anti-hRAD54 d-18 and l-17 (Santa Cruz Biotechnology). Horseradish peroxidase-linked secondary antibodies were donkey antirabbit IgG and donkey antimouse IgG (Jackson ImmunoResearch Laboratories, Inc., West Grove, PA), and goat antimouse κ horseradish peroxidase, which does not recognize the heavy chain (SouthernBiotech, Birmingham, AL).

Cells were lysed with immunoprecipitation buffer [0.5% NP40, 50 mm Tris (pH 8.0), and 120 mm NaCl, or 1% sodium deoxycholate, 0.1% SDS, 1 mm Tris (pH 7.4), and 150 mm NaCl] containing protease inhibitors. Protein concentrations were determined, and the samples were incubated for 1–2 h at 4°C with the indicated primary antibody or an equivalent amount of the appropriate control (mouse IgG or rabbit serum). Protein complexes were then precipitated with agarose- or Sepharose-conjugated protein A and/or G. Development was carried out by enhanced chemiluminescence according to the manufacturer instructions (Amersham), and densitometry analysis was performed on a Macintosh computer using the public domain NIH Image program (developed at the NIH and available on the internet).⁵ Experiments were repeated at least twice, and representative blots are shown.

The cells were seeded onto coverslips or glass chamber slides (Nalge Nunc International) and treated as indicated at least 24 h later. Just before fixation, the cells were washed with PBS (pH 7.4). In some cases, the live cells were fixed immediately with 4% paraformaldehyde in PBS or ice-cold absolute methanol. Paraformaldehyde-fixed cells were permeabilized with 0.2–0.5% Triton X-100 in PBS for 20 min, and methanol-fixed cells were washed briefly with ice-cold acetone. Cells were then washed several times with PBS and blocked for 20 min with 2.0% BSA/0.15% glycine in PBS. In other cases, the live cells were treated for 8 min with a hypotonic lysis solution [0.5% NP40, 10 mm Tris (pH 7.4) 2.5 mm MgCl₂, and 1 mm phenylmethylsulfonyl fluoride) on ice. The cells were washed once with ice-cold PBS and then fixed for 4 min in ice-cold 100% ethanol. They were then washed three times with PBS containing 0.2% Tween 20 and blocked for 30 min with 10% normal chicken serum in PBS.

All of the subsequent antibody incubations were carried out at room temperature in a humidified, light-proof chamber with the antibody diluted in a blocking solution. Incubations with the indicated primary antibodies were conducted for either 1 h at room temperature or overnight at 4°C, and all of the secondary antibody incubations were carried out for ∼1 h at room temperature. Extensive washing with PBS or PBS containing 0.2% Tween 20 was carried out after both incubations. Secondary antibodies used were goat antimouse IgG-Alexa488, goat antirabbit IgG-Alexa568, and donkey antigoat-Alexa488 from Molecular Probes (Eugene, OR), as well as donkey antirabbit IgG-fluorescein-5-isothiocyanate and horse antigoat IgG-Texas Red (Jackson Immuno Laboratories). Coverslips were applied with Vectashield antifade mounting medium with or without 0.5 μg 4′,6-diamidino-2-phenylindole (Vector Laboratories, Burlingame, CA).

For conventional immunofluorescence microscopy, the cells were examined with a Zeiss Axioskop fluorescence microscope equipped with a Photometrics SenSys CCD camera. Images were captured, pseudocolored, and, in some cases, merged using IP Lab Spectrum v3.1 software (Scanalytics, Inc., Fairfax, VA). Alternatively, confocal fluorescent images were collected with a Bio-Rad MRC 1024 confocal scan head mounted on a Nikon Optiphot microscope with a 60× Planapochromat lens or a Zeiss LSM410. Z sections were collected at 0.5-μm intervals for each cell using LaserSharp software (Bio-Rad). Images were analyzed with Confocal Assistant software (Bio-Rad). For calculation of percentages of cells with foci, 50 cells were scored in three separate experiments. For quantitation of signal overlap, the Z sections of five random cells that were scored as containing foci were analyzed and averaged with the Bio-Rad LaserSharp software.

For the p53-hRAD54 assays, recombinant GST-p53 fusion proteins were produced in Escherichia coli and purified on glutathione-agarose beads according to the manufacturer (Sigma). Protein concentrations were determined by Coomassie blue staining of SDS-PAGE gels and comparison with molecular weight standards (Bio-Rad). hRAD54 was produced in vitro using the TNT-coupled rabbit reticulolysate system (Promega) and radiolabeled with [³⁵S]cysteine according to the manufacturer. Immediately after protein production, 5 μl of the hRAD54 lysate was incubated with 1 μg each of the recombinant GST-p53 proteins for the specified time period in immunoprecipitation buffer (120 mm NaCl, 50 mm Tris, and 0.5% NP40) on a rocker at room temperature. Then, the agarose beads were washed five times with at least 25 volumes of buffer. The precipitated proteins were denatured in loading buffer by heating at 95°C for 10 min. Samples were run on a 10% SDS-PAGE gel. The gel was then fixed in 50% methanol/10% acetic acid, enhanced in 1.5 m sodium salicylate, dried, and exposed overnight to autoradiographic film. Binding was measured as the amount of radiolabeled hRAD54 in the binding lanes compared with a control direct-translation product. Antibodies used in the competition studies were obtained from Oncogene Research Products (San Diego, CA).

All of the transfections were carried out for 6 h in serum-free medium with Lipofectamine Plus or Lipofectamine 2000 according to manufacturer instructions (Invitrogen). Complete medium was added, and cells were incubated for a total of 36 h for the host cell reactivation assay or 72 h for analysis of hRAD51 structures.

The host cell reactivation assay to assess HR capacity was carried out as described (69). Different combinations of expression plasmids were cotransfected with the pBHRF plasmid, encoding a truncated GFP and a full-length BFP. In the absence of HR, only BFP is expressed. However, HR between the BFP and truncated GFP can lead to the creation of a functional GFP. Green and blue fluorescence were examined simultaneously with a Becton-Dickinson FACS Vantage flow cytometer with a 488-nm argon laser (GFP) and an UV (350–360 nm) laser (BFP). Each experiment was carried out at least three times.

We showed previously an in vitro interaction between hRAD51 and p53 (80). We also showed an in vivo interaction between RAD51 and murine temperature-sensitive p53 in rat embryo fibroblasts, and between hRAD51 and mutant p53 (220C) in the hepatocarcinoma cell line, HuH7 (82). To investigate the association of hRAD51 with wild-type p53 at endogenous levels in normal cells, we conducted co-IP experiments in NHFs. hRAD51 co-IPed with p53 in untreated NHF (Fig. 1,A). On the basis of densitometry, the amount of hRAD51 that co-IPed with p53 increased modestly within 2 h of treatment with the double-strand break-inducing agent NCS, but then decreased at later time points. Although the amount of co-IPed hRAD51 had decreased significantly after 18 h of treatment, the decrease was comparable with that observed for the overall hRAD51 level (Fig. 1 A). A 16-h treatment with the ribonucleotide reductase inhibitor HU, which can collapse replication forks leading to DNA breaks in S phase, led to an increase in the amount of hRAD51 that co-IPed with p53 (87). Similar results were observed after γ-irradiation of the hepatoma cell line HepG2, which contains wild-type p53 (data not shown).

Because hRAD51 colocalizes in nuclear foci with other factors involved in HR-mediated DNA repair in S phase cells and/or DNA-damaged cells, we sought to determine whether p53 is also present in these foci. After standard paraformaldehyde fixation of untreated NHF, p53 staining with the NH₂-terminal antibody, DO-1, or a phosphoserine-15-specific antibody showed only faint staining with little evidence of nuclear foci (data not shown). However, within 2 h of NCS treatment, which induces ATM-dependent phosphorylation of p53 at serine 15, a focal staining pattern of endogenous p53 phosphoserine-15 was observed in 15% ± 3% (mean ± SD) of cells. These foci colocalized with hRAD51 (Fig. 1, B–E). The relatively low fraction of asynchronously growing cells with RAD51 foci is consistent with their formation specifically in S phase cells. Approximately 35–40% of the signal overlapped in these cells (Fig. 1 F). The remainder of the cells contained less than five foci of either protein, so they were not scored. A much higher percentage of NHF exhibited focal p53-hRAD51 colocalization after HU treatment, possibly because of an accumulation of cells in S phase (87). The focal staining was specific, because no staining was apparent in NCS-treated p53-null cells (data not shown). Although no colocalization of endogenous p53 and hRAD51 was evident in untreated NHF, p53 overexpression led to colocalization in a subset of cells when stained with a mixture of p53 NH₂- and COOH-terminal antibodies (data not shown).

p53 can physically interact with several DNA helicases, including BLM, XPB, XPD, and WRN (51, 52, 53, 54). We investigated whether p53 could also interact with hRAD54, a member of the same superfamily. HU treatment of NHF for 16 h also induced association, as evidenced by an increase in hRAD54 in p53 immunoprecipitates (Fig. 2,A). Treatment of NHF or normal human lymphoblasts with γ radiation also induced the p53 signal in hRAD54 immunoprecipitates (data not shown). Consistent with these findings, p53 also colocalized with hRAD54 after HU treatment, with approximately 40–60% signal overlap in cells with foci (Fig. 2, B--E). In cells overexpressing hRAD54, the hRAD54 antibodies revealed well-defined bands of the same and predicted molecular weight as that seen in the p53 immunoprecipitates (Supplementary Fig. 1). In addition, the hRAD54 focal staining was not present when the hRAD54 antibodies were preincubated with control peptides (data not shown). These data support the specificity of the hRAD54 antibodies.

On the basis of the data described above, binding of p53 to hRAD54 could be indirect through mutual interaction with hRAD51. However, hRAD54 is the human homologue of a yeast DNA helicase that shares structural motifs with other members of a superfamily that includes DNA helicases to which p53 can bind (BLM, CSB, XPB, XPD, and WRN). Therefore, we sought to determine whether p53 could bind directly to hRAD54. Radiolabeled, in vitro-translated hRAD54 was incubated with E. coli-produced, GST-tagged, wild-type p53, which had been conjugated to reduced glutathione-linked agarose beads. hRAD54 precipitated specifically with GST-p53, but not with GST alone or unconjugated beads (Fig. 3 A). These results show a clear, specific interaction between p53 and hRAD54 in vitro.

Given that p53 can bind directly to DNA and that hRAD54 is a homologue of a DNA helicase, it is possible that the interaction between these proteins could be through a nucleic acid intermediate (88). To test this possibility, binding of the in vitro-translated hRAD54 to GST-p53 was assayed after treatment of the lysate with DNase I, RNase A, or both. Nuclease treatment had no effect on the hRAD54-p53 interaction (Fig. 3 B), indicating that the association is direct.

p53 can be divided into three topographical domains based on function. The NH₂-terminal domain is active in transactivational activities; the central region is a specific DNA-binding domain, where most of the cancer-associated p53 point mutations are found; and the multifunctional COOH-terminal domain possesses oligomerization, nuclear localization, and apoptosis-inducing activities (reviewed in Ref. 89). A panel of p53 deletion mutants was used to determine which functional domain(s) of p53 were required for hRAD54 binding. Equal amounts of GST-tagged, full-length p53 or p53 deletion mutants were incubated for equal time periods with radiolabeled hRAD54. The three mutants with an intact COOH terminus showed strong binding, whereas the other two mutants showed weak binding (Fig. 4,A). A major reduction in binding was observed with the N5 p53 deletion mutant relative to full-length p53 (NC), showing that hRAD54 binds preferentially to the multifunctional COOH-terminal domain of p53 within residues 294–393 (gray boxes in Fig. 4 A).

To additionally verify the hRAD54 interaction with the p53 COOH-terminal domain, competition experiments were conducted on the p53 deletion mutants with the p53 antibody PAb421, which binds to residues 371–380 (Ref. 90; Fig. 4,B). Wild-type and deletion mutants of p53 were incubated with radiolabeled hRAD54 in either the presence or absence of saturating amounts of PAb421 (Fig. 4 C). PAb421 diminished the binding of hRAD54 to the forms of p53 containing the COOH terminus. Little binding occurred with the form of p53 lacking the COOH terminus (N5), either with or without the antibody.

To show that the competitive disruption of hRAD54-p53 binding by PAb421 was specific to its binding to the COOH terminus of p53, a panel of p53 antibodies that bind to various regions were used (Fig. 4,B). The antibodies were incubated with GST-tagged, full-length p53 and radiolabeled hRAD54 (Fig. 4 D). Consistent with the above results, PAb421 strongly inhibited binding, whereas TrpE (an antibody to a bacterial protein used as a negative control) did not decrease binding at all. The other p53 antibodies caused only minor reductions in binding.

Previous studies have shown focal colocalization between murine RAD51 and RAD54 (19), and between overexpressed hRAD51 and hRAD54 (91). To demonstrate colocalization of hRAD51 and hRAD54 at endogenous levels, we used a mixture of two goat polyclonal Rad54 antibodies. Focal nuclear staining of hRAD54 was observed in 5% ± 1% of untreated cells, and these foci colocalized with hRAD51 foci (data not shown). The percentage of cells with these hRAD54 foci increased to 15% ± 4% within 4 h of NCS treatment (Fig. 5, A–D). Similar to the results with hRAD51, the relatively low fraction of asynchronously growing cells with hRAD54 foci is consistent with their formation specifically in S phase cells. In addition, hRAD54 was present in hRAD51 immunoprecipitates in both untreated and HU-treated NHF (Fig. 5 E), which may reflect an HU-induced increase in the fraction of S phase cells.

When hRAD51 is overexpressed it can polymerize into high-order, filamentous nuclear structures (92). Although the physiological significance of these structures has not been established, they serve as a marker for RAD51 polymerization. Given that p53 can bind to the homo-oligomerization domain of hRAD51 (82), we determined whether p53 could affect hRAD51 structure formation by transfecting postcrisis LFS041 fibroblasts (p53 null) with combinations of wild-type or mutant p53, or hRAD51. About 60% of the cells expressing wild-type hRAD51 contained the structures (Fig. 6, A, B, and G), whereas the remainder exhibited diffuse nuclear staining. In contrast, only ∼18% of cells coexpressing wild-type hRAD51 and wild-type p53 contained the structures (Fig. 6,G), whereas the remainder exhibited diffuse (Fig. 6, C and D) or more focal (Fig. 6, E and F) nuclear staining. The p53/273H point mutant exhibits greatly reduced binding to hRAD51 (80). Consistent with this, coexpression of p53/273H with wild-type hRAD51 failed to significantly inhibit structure formation (Fig. 6,G). The hRAD51/186P mutant exhibits greatly reduced binding to p53 in vitro (Fig. 6,H). Although this mutant still formed hRAD51 structures in over half of the RAD51-positive cells, coexpression of wild-type p53 failed to significantly inhibit structure formation (Fig. 6 G). These data indicate that p53 can inhibit hRAD51 polymerization in vivo through direct binding to hRAD51.

Several studies have shown that recombination is elevated in p53-deficient cells (60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70). To help establish a mechanistic role for p53 in this process, a host cell reactivation assay (69) was conducted to assess HR. NHFs were transfected with p53, hRAD51, and/or hRAD54 mutants, along with a plasmid encoding a truncated GFP and a full-length BFP. In the absence of recombination, only BFP is expressed, but successful recombination between the BFP and truncated GFP can lead to the creation of functional GFP. Although the recombination occurs extrachromosomally in this assay, the intermediate DNA structures and mechanism are presumably the same as those that occur during chromosomal HR. Consistent with previous reports (62, 65, 70), expression of the tumor-derived p53/175H and p53/248W mutants led to an elevation in spontaneous HR (Fig. 7,A). In contrast, expression of a dominant-negative RAD51, SMRAD51 (85), or a dominant-negative RAD54, hRAD54/189R (19), led to a reduction in spontaneous HR. Interestingly, when either of the p53 mutants was coexpressed with either of the RAD dominant-negatives, the HR levels did not increase over the levels seen with the dominant-negatives alone (Fig. 7 A). These data validate the assay for the study of the RAD51 pathway and indicate that the elevation in HR observed after p53 inactivation is dependent on the RAD51/RAD54 pathway.

The RAD51 overexpression observed in many cancer cells may result in elevated HR. We next determined the effects of p53 on hRAD51-mediated HR using the host cell reactivation assay. Overexpression of wild-type hRAD51 or hRAD51/186P increased spontaneous HR approximately 2–3-fold over the basal level (Fig. 7,B). For this study, the transcriptionally inactive 22/23 mutant of p53 (86) was used to minimize the potential contribution of secondary, transcription-dependent effects of p53 on HR. Consistent with a recent report (93), p53 did not significantly affect the basal level of spontaneous HR in the plasmid substrate. However, when coexpressed with hRAD51, p53 prevented any significant increase in HR by hRAD51. In contrast, p53 did not significantly inhibit the elevated HR induced by overexpression of the p53 nonbinding mutant of hRAD51 (Fig. 7,B). Expression of the transfected proteins was verified by Western blot (Fig. 7, C and D). Although indirect effects of p53 on HR cannot be fully excluded, these data additionally support the hypothesis that p53 can repress HR through a direct interaction with hRAD51.

Previous studies have correlated p53 status with HR frequencies (60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70). Other studies have suggested a direct interaction between p53 and RAD51 (80, 81, 82), a central component of the HR pathway. In this study, co-IP experiments indicate that p53, hRAD51, and hRAD54 can form complexes at endogenous levels in normal cells. In addition, these proteins colocalize in nuclear foci, presumably during the S and/or G₂ phases of the cell cycle when the HR proteins are at their peak levels and sister chromatids are available as repair templates. The phosphoserine 15-specific p53 antibody used in this study resulted in focal staining, but similar results were achieved with antibody #588 (87), raised against the COOH terminus of p53 (94), so it is unclear whether any post-translational modifications to p53 are required for the interactions. The foci may form during the processing of replication-associated or exogenously induced DNA damage and, thus, represent sites of active repair. Another, more likely possibility is that the amount of protein at sites of active repair is too low to detect with conventional immunofluorescence, and that the visible foci represent accumulations of protein at persistent lesions that are irreparable or only slowly repaired (95). DNA breaks can stimulate recombination, and persistent breaks are prime candidates for misrepair. Thus, the accumulation of p53 in hRAD51 and hRAD54 foci may inhibit inappropriate recombination.

A host cell reactivation assay was used to demonstrate the direct role of p53 in modulating HR (Fig. 7 A). This study confirms previous reports that expression of the tumor-derived p53 mutants 175H or 248W leads to elevated recombination (62, 65, 70). This elevation is independent of the transactivation activity of p53 (65). We showed previously that p53/248W exhibits reduced binding to Holliday junctions (56). We also demonstrate that dominant-negative forms of either RAD51 or RAD54 reduce recombination in the assay. The dominant-negative RAD51 used was a chimera of yeast and murine RAD51, SMRAD51, which reportedly eliminates spontaneous recombination in a Chinese hamster ovary reporter line. The dominant-negative RAD54 used, hRAD54/189R, contains a mutation in the conserved DNA-dependent ATPase domain that may be require for RAD54-mediated DNA unwinding (19). Coexpression of the p53 mutants with either of these dominant-negatives failed to elevate recombination over the reduced level observed with the dominant-negatives alone. This indicates that p53 modulation of HR is dependent on hRAD51 and hRAD54.

Elevated levels of hRAD51 are observed in a variety of tumor cell lines, suggesting that hyper-recombination may play a role in tumorigenesis (8, 9, 10, 11, 12). We demonstrate here that hRAD51 overexpression leads to elevated HR in the host cell reactivation assay and that coexpression of transactivation-deficient p53 reduces the recombination to baseline levels. The binding site for p53 on hRAD51 between amino acids 125 and 220 is highly conserved from bacteria to human and includes the putative homo-oligomerization domain, which may be necessary to form functional hRAD51 nucleoprotein filaments (16, 80). Interestingly, expression of the p53 nonbinding mutant hRAD51/186P also elevated HR, but coexpression of p53 did not significantly inhibit this activity (Fig. 7,B). Parallel results were achieved in assays of the ability of hRAD51 to form higher-order structures in vivo using these same mutants (Fig. 6). These data indicate that p53 modulates HR through direct interaction and inhibition of hRAD51.

This study also shows for the first time that p53 can interact directly with hRAD54 in vitro and in vivo. Although it has been demonstrated that hRAD51 and hRAD54 can bind directly to each other (17), the data presented here and elsewhere (80) indicate that p53 can bind to both separately. hRAD54 binds mainly to the extreme COOH terminus of p53, which is consistent with the binding of p53 to several other members of the same superfamily. It is also possible that in vivo p53 binds to either hRAD51 or hRAD54 only indirectly through association with the other protein.

Several studies indicate an even more direct role of p53 in DNA repair, including HR-related functions. For example, p53 reportedly binds to single- and double-stranded DNA ends through its COOH terminus, and can catalyze DNA renaturation and strand transfer reactions (88, 96, 97, 98). Other studies show that p53 binds to DNA base mismatches through its COOH terminus, resulting in HR inhibition (83, 99). In addition, there is evidence that p53 binds to Holliday junctions and may catalyze their resolution or inhibit HR (81, 100). It is also of interest that the hRAD51 paralog, RAD51B, can phosphorylate p53 (101), and that p53 reportedly associates directly with a number of other proteins implicated in HR, including BLM, BRCA1, BRCA2, RAD52, and RPA (54, 74, 75, 76, 77, 79). The specific nature of these physical and functional interactions remains to be elucidated.

We also have shown that p53 colocalizes with BLM in nuclear foci that presumably form at sites of replication fork collapse, and that p53 localization in these foci is dependent on functional BLM (87). BLM is an antirecombinase that may catalyze reverse branch migration at these sites (102). In addition, we have shown that p53 can cooperate with BLM to inhibit RAD51 pathway-dependent sister chromatid exchanges (87). The data support a model in which BLM recruits p53 to certain sites at which recombination is favorable, but not necessarily desirable. Inhibitory interactions of p53 with hRAD51 and/or hRAD54 may assist in the down-regulation of inappropriate recombination.

Normally, HR serves to enhance genetic stability by mediating the accurate repair of double-strand breaks and other DNA lesions that cannot be handled by other repair processes. However, inappropriate activity of the HR machinery may lead to indiscriminate recombination, which could lead to genetic instability in the form of chromosomal aberrations. Although it has been suggested that transactivation is the determining function of p53 in tumor suppression (47), HU-induced replication fork stalling is known to induce transcriptionally inactive p53 in S phase (103). The relocalization of p53 to nuclear foci and formation of complexes with hRAD51 and hRAD54 are consistent with this finding. The results presented here, in combination with other studies showing that HR levels are elevated in the absence of functional p53 (60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70), support the hypothesis that p53 inhibits inappropriate recombination through inhibitory biochemical interactions with both hRAD51 and hRAD54. Thus, regulation of recombination may be another mechanism by which p53 maintains genetic stability.

We thank John Kolman for pJS-1, Robert Naviaux for pCLXSN, and Robbert Slebos for pBHRF. We also thank Jan Hoeijmakers and Roland Kanaar for additional reagents and advice, Jill Coursen for technical support, Ana Robles for critical review of the manuscript, Dorothea Dudek for editorial assistance, and Karen MacPherson for bibliographic assistance.