Radiation-induced Assembly of Rad51 and Rad52 Recombination Complex Requires ATM and c-Abl*

Cells from individuals with the recessive cancer-prone disorder ataxia telangiectasia (A-T) are hypersensitive to ionizing radiation (I-R). ATM (mutated in A-T) is a protein kinase whose activity is stimulated by I-R. c-Abl, a nonreceptor tyrosine kinase, interacts with ATM and is activated by ATM following I-R. Rad51 is a homologue of bacterial RecA protein required for DNA recombination and repair. Here we demonstrate that there is an I-R-induced Rad51 tyrosine phosphorylation, and this induction is dependent on both ATM and c-Abl. ATM, c-Abl, and Rad51 can be co-immunoprecipitated from cell extracts. Consistent with the physical interaction, c-Abl phosphorylates Rad51 in vitro and in vivo. In assays using purified components, phosphorylation of Rad51 by c-Abl enhances complex formation between Rad51 and Rad52, which cooperates with Rad51 in recombination and repair. After I-R, an increase in association between Rad51 and Rad52 occurs in wild-type cells but not in cells with mutations that compromise ATM or c-Abl. Our data suggest signaling mediated through ATM, and c-Abl is required for the correct post-translational modification of Rad51, which is critical for the assembly of Rad51 repair protein complex following I-R.

Ataxia telangiectasia (A-T) 1 is an autosomal recessive genetic disease characterized by diverse clinical symptoms that include neuronal degeneration, immune deficiency, gonadal abnormalities, cancer predisposition, premature aging, and oculocutaneous telangiectasias (1,2). Cells from individuals with A-T are hypersensitive to ionizing radiation (I-R) and chemicals that cause DNA double-strand breaks (DSB). Upon I-R, A-T cells exhibit defects in multiple cell cycle checkpoint functions, including an aberrant p53 response and DNA DSB repair (1,3). The DSB repair defect in A-T cells is subtle, but nevertheless contributes significantly to the radiosensitivity and chromosomal instability phenotype of these cells (4 -6). At a time when rejoining of I-R-induced DSB is completed in normal mitotic cells, A-T cells still contain a significant number of DSBs, which may be converted into chromosomal breaks (7,8). Additionally, the DSB repair process in A-T cells is errorprone (9,10). A-T cells also possess an elevated level of spontaneous intrachromosomal recombination (11). In meiosis of A-T cells, the synaptonemal complexes fail to form at the pachytene stage, thus aborting gametogenesis (12)(13)(14). Little is known about the molecular basis for the meiotic and DNA DSB repair defects in A-T cells.
In eukaryotes, DNA DSB can be repaired by either homologous recombination or nonhomologous end-joining pathways (15). Homologous recombination is achieved through multiple enzyme-catalyzed steps that include processing and resection of the DSB, searching for and pairing of DNA homologs, strand invasion, DNA synthesis, and finally, resolution of recombination intermediates (16,17). Rad51, a eukaryotic homolog of the bacterial recombinase RecA, is required for recombinational repair of DSB. In in vitro recombination assays using purified components, Rad51 catalyzes strand exchange in an ATP-dependent reaction (16,17). Interactions between yeast Rad51 and other recombination factors including RPA, Rad52, Rad54, Rad55, and Rad57 have been described (18,19). Human Rad51 also binds Rad52 (20,21). Recent biochemical studies have demonstrated that the strand exchange activity of Rad51 is stimulated by Rad52, Rad54, and the Rad55⅐Rad57 complex, indicating that the protein-protein interactions are functionally significant (22)(23)(24)(25)(26). The results of studies with human Rad51 and Rad52 suggest that the mechanisms of DSB repair by homologous recombination are conserved in higher eukaryotes (27).
ATM, the gene mutated in A-T, encodes a 370 kDa protein that is predominantly nuclear (28 -31). It contains a phosphatidylinositol kinase domain that is also present in several other proteins known to function in cellular responses to DNA damage, cell cycle control, and telomere maintenance (32,33). Upon activation by DNA damage, the ATM kinase phosphorylates p53 on serine 15, which results in stabilization of p53 (34,35). In addition to modifying p53, ATM is also required for the activation of the nonreceptor tyrosine kinase c-Abl (36,37). ATM binds c-Abl constitutively and may phosphorylate c-Abl on serine 465, leading to enhanced kinase activity of c-Abl (37). Whether defective c-Abl activation affects DSB repair and thus * This work was supported by grants from Ataxia Telangiectasia Children's Project and National Institutes of Health Grant 1R01NS378381-01 (to E. L.). 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. § The first two authors contributed equally to this work. ¶ Recipient of National Institutes of Health Postdoctoral Fellowship. ¶ ¶ To whom correspondence should be addressed. Tel.: 210-567-7326; Fax: 210-567-7324; E-mail: Leee@uthscsa.edu. 1 The abbreviations used are: A-T, ataxia telangiectasia; I-R, ionizing radiation; ATM, ataxia telangiectasia mutated; DSB, double-strand break(s); GST, glutathione S-transferase; MEF, mouse embryonic fibroblast(s); PAGE, polyacrylamide gel electrophoresis; NTA, nitrilotriacetic acid; LZP, leucine zipper and proline-rich region of ATM.
contributes to the phenotypes of A-T cells remains to be addressed.
In this study, we tested the hypothesis that the aberrant DNA repair phenotype of A-T cells is due at least in part to defective modulation of protein components of the DNA DSB repair machinery by ATM kinase-mediated signals. We provide evidence that ATM and c-Abl kinase are required for correct post-translational modification and assembly of Rad51 protein complexes.

MATERIALS AND METHODS
Antibodies-The monoclonal anti-ATM antibody 3E8 was made using GST-ATM-LZP fusion protein as the immunogen. Polyclonal and monoclonal antibodies against human Rad51 and Rad52 were produced using HisRad51 and HisRad52 purified from bacteria. The anti-phosphotyrosine antibodies 4G10 and RC20, anti-␤-actin antibody, and anti-Myc antibody 9E10 were from Upstate Biotechnologies, Transduction Laboratories, Sigma, and Santa Cruz Biotechnologies, respectively. 8E9 and P6D are monoclonal antibodies that recognize human c-Abl.
Purification of Rad51, Rad52, and c-Abl-Human Rad51 and Rad52 were overexpressed in Escherichia coli. Consecutive chromatographic steps involving Q-Sepharose, Ni-NTA-agarose (for histidine-tagged Rad51 only), Affi-blue, hydroxyapatite, and Mono-Q were performed to purify Rad51, and chromatography on Ni-NTA and Mono-S columns was used to purify histidine-tagged Rad52. Myc-tagged c-Abl, overexpressed in COS-7 cells by transient transfection, was purified by affinity chromatography with the 9E10 antibody immobilized on agarose. The Myc epitope peptide was used to compete off the c-Abl bound to the column.
Analysis of Protein Interactions-Cells were lysed in EBC buffer (50 mM Tris, pH 7.6, 120 mM NaCl, 0.5% Nonidet P-40, 1 mM EDTA, 1 mM ␤-mercaptoethanol, 50 mM NaF, and 1 mM Na 3 VO 4 ) plus protease inhibitors. Immunoprecipitation was performed as described earlier (28), using secondary antibodies cross-linked to magnetic beads (Dynal Inc. Oslo, Norway). To study whether protein complex formation requires DNA, immunoprecipitation was performed in the presence of ethidium bromide (20 g/ml), and the immunoprecipitates were treated with 100 units of DNase I at 37°C for 1 h. To examine Rad51 tyrosine phosphorylation, immunoblots containing Rad51 were incubated with anti-phosphotyrosine antibody and then horseradish peroxidase-conjugated secondary antibody. After detection by enhanced chemiluminescence (Amersham Pharmacia Biotech), bound antibodies in the blot were stripped off by incubation in 0.1 M glycine, pH 3.0, for 30 min. Total Rad51 amounts in the blot were determined with anti-Rad51 antibody. In GST pull-down assays, 1 g of GST or GST fusion proteins that had been absorbed on glutathione beads was incubated with target proteins in EBC buffer for 2 h at 4°C.
To examine the effects of Rad51 phosphorylation by c-Abl on interaction between Rad51 and Rad52, 1 g of Rad51 was incubated with affinity purified wild-type or kinase inactive (K290R) c-Abl in kinase buffer (20 mM Tris, pH 7.5, 10 mM MgCl 2 , 1 mM MnCl 2 , and 1 mM dithiothreitol, 100 M ATP) at 30°C for 20 min. After addition of genistein to inactivate the c-Abl kinase, Rad51 in the kinase reactions was subjected to immunoprecipitation with a rabbit polyclonal anti-Rad51 antibody and the anti-phosphotyrosine antibody 4G10. Secondary antibodies conjugated with magnetic beads were used to capture the immune complex. 0.5 g of Rad52 was then added to the mixture containing free and antibody-bound Rad51 and incubated at 4°C for 45 min. Rad51 and Rad52 levels in antibody-bound and -unbound fractions were determined by immunoblotting with their respective antibodies.
Mapping the Residues of Rad51 Phosphorylated by c-Abl-His 6 -Rad51 cDNA was cloned into pCMV vector for expression in mammalian cells. Specific tyrosine residues of Rad51 were mutated to phenylalanine using Quick-change TM (Stratagene). 8 g of pSV40-Myc-c-Abl or -K290R-c-Abl, together with 2 g of pCMV-His 6 -Rad51, were used to transfect COS-7 cells in one 10-cm plate. Two days later, the transfected cells were lysed in 1% Triton X-100 in EBC buffer containing 1 mM Na 3 VO 4 and protease inhibitors. His 6 -Rad51 in the cell lysates was absorbed to Ni-NTA resin. Proteins bound to the Ni-NTA resin were eluted with 250 mM imidazole in EBC and were then subjected to immunoprecipitation with rabbit anti-Rad51 antibody. Tyrosine phosphorylation of the immunoprecipitated His 6 -Rad51 was detected by immunoblotting with the anti-phosphotyrosine antibody RC20.

I-R-induced Rad51
Tyrosine Phosphorylation Is Dependent on ATM and c-Abl-We used glutathione beads with glutathione S-transferase (GST)-ATM fusion proteins as the ligand to bind cellular proteins that may associate with ATM. Rad51 and c-Abl bound to a GST-ATM fusion protein encompassing the leucine zipper and proline-rich region (LZP) of ATM (residues 980 -1, 437, Fig. 1a, lane 3) but not to GST or a GST-ATM-N (a fusion protein containing residues 249 -523 of ATM) (Fig. 1a,  lanes 2 and 4). Immunoprecipitation was then carried out to further assess the association between ATM, c-Abl, and Rad51. Each of these three proteins was detected in immunoprecipitations with anti-ATM, anti-c-Abl or anti-Rad51 antibodies but not with preimmune IgG (Fig. 1b). Treatment of the immunoprecipitates with DNase I had no effect on the association between ATM, c-Abl, and Rad51, indicating that complex formation among these proteins is probably not mediated by DNA (Fig. 1b). The amount of co-immunoprecipitating proteins did not change significantly after I-R treatment (data not shown).
Previous studies have demonstrated an ATM-dependent ac- tivation of c-Abl tyrosine kinase following I-R (36,37). Prompted by the physical interaction between Rad51 and c-Abl, we examined if Rad51 is tyrosine phosphorylated before and after I-R. Extracts prepared from T24 cells before and after I-R treatment were immunoprecipitated with anti-Rad51 antibody followed by sequential immunoblotting with anti-phosphotyrosine and anti-Rad51 antibodies. Although the levels of Rad51 did not change, tyrosine phosphorylation of Rad51 increased more than 3-fold following I-R treatment (Fig. 2a). Tyrosine phosphorylation of Rad51 was also examined in embryonic fibroblasts established from wild-type mice, mice with nullizygous c-Abl mutation (38), and also from ATM knockout mice (14). Induction of Rad51 tyrosine phosphorylation after I-R was seen in wild-type mouse embryonic fibroblasts (MEF), but not in c-Abl Ϫ/Ϫ or ATM Ϫ/Ϫ MEF (Fig. 2b). Thus, the induced tyrosine phosphorylation of Rad51 is dependent upon both ATM and c-Abl. An elevated basal level of Rad51 tyrosine phosphorylation was detected in c-Abl Ϫ/Ϫ MEFs (Fig. 2b). The reason for this result is unknown, but it could be because of the presence of a compensatory mechanism that results in an elevated basal level of c-Abl-like activity in c-Abl Ϫ/Ϫ MEF, which also leads to increased tyrosine phosphorylation of other c-Abl substrates (37).
Whether Rad51 protein can serve as a direct substrate for c-Abl was tested using purified Rad51, Myc-tagged c-Abl and kinase-inactive c-Abl (39). Rad51 (Fig. 3a) was phosphorylated by active c-Abl but not by the kinase-inactive c-Abl. Immunoblotting with an anti-phosphotyrosine antibody indicates that phosphorylation of Rad51 by c-Abl occurred on tyrosine residues (Fig. 3a). To investigate possible Rad51 phosphorylation by c-Abl in vivo, Rad51 was expressed together with kinase inactive or active c-Abl in COS-7 cells. Co-expression of Rad51 with wild-type c-Abl, but not kinase inactive c-Abl, stimulated Rad51 tyrosine phosphorylation (Fig. 3b). Mutation of Rad51 Tyr 315 , but not Tyr 205 , Tyr 191 , or Tyr 54 to phenylalanine abolished Rad51 tyrosine phosphorylation by c-Abl (Fig. 3b). These results strongly suggest that c-Abl phosphorylates Rad51 Tyr 315 in vivo. Interestingly, Rad51 Tyr 315 is present in a YXXP context, a substrate sequence preferred by c-Abl kinase activity (40), and is a highly conserved residue in Rad51 in all known species of the animal kingdom.
To determine whether the elevated tyrosine phosphorylation of Rad51 in c-Abl Ϫ/Ϫ MEFs is because of phosphorylation on Tyr 315 , wild-type and Y315F Rad51 was expressed in these cells by transient transfection. The levels of tyrosine phosphorylation of wild-type and Y315F Rad51 were similar (Fig. 3c), suggesting that the majority of Rad51 tyrosine phosphorylation in c-Abl Ϫ/Ϫ MEF does not occur on Tyr 315 .
Tyrosine Phosphorylation of Rad51 by c-Abl Enhances the Interaction between Rad51 and Rad52-Genetic and biochemical studies indicate that Rad51 cooperates with Rad52 in recombinational DNA repair (18,(22)(23)(24)27). Whether phosphorylation of Rad51 by c-Abl modulates its interaction with human Rad52 was investigated in binding assays employing purified Rad51 and Rad52 (Fig. 4a). To avoid interference from possible un-phosphorylated Rad51, anti-phosphotyrosine antibody was used to isolate tyrosine phosphorylated Rad51 after kinase reaction with wild-type c-Abl (Fig. 4b). Approximately 4-fold more Rad52 was associated with tyrosine phosphorylated Rad51 than with Rad51 treated with K290R c-Abl (Fig.  4b), indicating that tyrosine phosphorylation of Rad51 by c-Abl enhances its interaction with Rad52 in vitro.
Because I-R induces Rad51 tyrosine phosphorylation in wildtype cells, but not in ATM Ϫ/Ϫ or c-Abl Ϫ/Ϫ MEF, we examined the association between Rad51 and Rad52 in cells with these distinct genetic backgrounds before and after I-R. Extracts prepared from mock-treated and irradiated T24, wild-type MEF, as well as c-Abl Ϫ/Ϫ and ATM Ϫ/Ϫ MEF were subjected to immunoprecipitation using anti-Rad51 antibody. The level of Rad52 in the Rad51 immunoprecipitates was quantitated by immunoblotting with anti-Rad52 antibody. Significantly more (Ͼ3-fold) Rad52 was co-precipitated after I-R treatment of T24 and wild-type MEF. In contrast, enhanced association of Rad51 and Rad52 was not seen in c-Abl Ϫ/Ϫ or ATM Ϫ/Ϫ MEF after I-R (Fig. 5a). I-R did not affect the overall levels of Rad51 and Rad52 proteins (Fig. 5b). Thus, the I-R-induced association between Rad51 and Rad52 appears to correlate with the I-R- induced Rad51 tyrosine phosphorylation. Interestingly, there was an elevated association of Rad51 and Rad52 in c-Abl Ϫ/Ϫ MEF before I-R treatment (Fig. 5a), perhaps because of the increased basal level of Rad51 tyrosine phosphorylation in these cells.

DISCUSSION
Because A-T cells arrested in the G 1 phase of the cell cycle are radiosensitive, it appears that defective checkpoint control is not the only cause of radiosensitivity of A-T cells (4 -6). Moreover, the aberrant mitotic recombination and elevated chromosomal breaks in A-T cells after I-R suggest that ATM plays a direct role in the repair of DNA DSB (1, 4, 7, 9 -11). In this study, we report that ATM and ATM-mediated signaling are required for both Rad51 phosphorylation and enhanced assembly of recombination complexes. Our data support the hypothesis that defective modulation of repair proteins may underlie the DNA repair deficiency in A-T cells.
Genetic and biochemical studies have assigned a central role for Rad51 in homologous recombination. The other recombination factors, Rad52, Rad54, and Rad55⅐Rad57 stimulate the strand exchange activity of Rad51 (22)(23)(24)(25)(26)(27). This stimulation likely requires direct interaction between the recombination proteins because mutant Rad52 lacking the Rad51 binding site failed to stimulate Rad51-mediated strand exchange (23). Our study has identified an I-R-induced phosphorylation of Rad51 on tyrosine residues and a concomitant increase in association between Rad51 and Rad52, which may lead to increased DNA repair proficiency. The enhanced association of Rad51 and Rad52 after I-R is likely because of Rad51 tyrosine phosphorylation mediated by ATM and c-Abl since the enhanced binding was also observed after phosphorylation of Rad51 by c-Abl in vitro. It is not yet known whether phosphorylation of Rad51 affects its binding with other members of the recombination complex.
While this study was in progress, Yuan et al. (41) reported that phosphorylation of yeast and human Rad51 by c-Abl inhibited the binding of Rad51 to DNA. Furthermore, the strand exchange activity of yeast Rad51 also decreased upon treatment with c-Abl (41). The authors concluded that c-Abl might have inhibitory effects on Rad51 activities in vivo. In contrast, our studies of Rad51⅐Rad52 complex formation in vitro and in vivo suggest that the ATM and c-Abl-mediated signaling is likely to promote repair given the biochemical evidence that Rad51 acts in concert with Rad52 in homologous recombination (22)(23)(24)27). Future biochemical studies are needed to address whether in the presence of Rad52, Rad51 would indeed exhibit higher recombinase activity upon phosphorylation by c-Abl.