Main

Individuals carrying a mutant allele of BRCA2 are genetically predisposed to the development of breast, ovarian and other cancers. About 80% of these individuals will develop cancers during their lifetime, and their tumours are characterized by a loss of heterozygosity1,2,3. The sequence of events that take place from the initial germline mutation to loss of the wild-type allele is not yet clear, but the consequences are catastrophic as indicated by the gross chromosomal instability phenotype associated with BRCA2-defective cells grown in culture4,5,6.

The BRCA2 gene encodes a large (384-kDa) protein containing 3,418 amino acids7,8. Although the sequence of BRCA2 reveals few clues to its cellular function, a role in DNA repair and replication fork maintenance is indicated by the hypersensitivity of BRCA2-defective cells to DNA-damaging agents4,9,10. Specifically, BRCA2 mutant cell lines are compromised in their ability to repair DNA double-strand breaks by homologous recombination11,12.

The function of BRCA2 in homologous recombination is likely to be mediated through interactions with RAD51, a protein that promotes the essential homologous-pairing and strand-exchange phases necessary for the recombinational repair of DNA damage13. Within exon 11 of BRCA2, a series of BRC repeats that bind RAD51 have been described14,15,16,17. An additional, but unrelated, RAD51 interaction domain has been mapped to exon 27 at the C terminus of BRCA2 (refs 9, 18).

In vitro, BRC peptides bind RAD51 monomers and block formation of the RAD51 nucleoprotein filaments necessary for recombinase activity19,20,21. Related interactions may occur in vivo, because ectopic expression of BRC fragments confers a dominant-negative (recombination-defective) phenotype on wild-type cells22,23. In response to DNA damage, however, BRCA2 and RAD51 colocalize to nuclear foci where DNA repair reactions are thought to take place24,25,26. Cell lines lacking BRCA2 do not elicit this repair response, and RAD51 localization to foci is reduced27,28,29. These observations led to the proposal that BRCA2 has a dual function in RAD51 control: by acting as a scaffold for RAD51 binding while enabling the rapid association of RAD51 to sites of DNA damage in response to cellular stress19. Although this is consistent with the recombinational repair defect and genome instability observed in BRCA2-deficient cells, precisely how the BRC repeats and the C-terminal region of BRCA2 bind and release RAD51 at the appropriate time and in response to a particular signal is unknown. Here we focus on the highly conserved BRCA2 C-terminal region that interacts with RAD51, and define its crucial function in homologous recombination.

Phosphorylation of the C terminus of BRCA2

A series of nine overlapping glutathione S-transferase (GST) fusion proteins designated B2-1 to B2-9 (ref. 30), spanning the entire coding region of BRCA2, were used to define the regions of BRCA2 that interact with RAD51 (Fig. 1a). Purified GST–BRCA2 fusion proteins were added to HeLa cell-free extracts, and GST–BRCA2–RAD51 complexes were isolated with glutathione beads. RAD51 was found to bind B2-3 (containing BRC repeats 1 and 2), B2-4 (containing BRC3, 4 and 5) and B2-9 (the C-terminal region of BRCA2) as detected by western blotting (Fig. 1b, middle panel, lanes 3, 4 and 9). Similar results were obtained with purified RAD51 (rRAD51; Fig. 1b, lower panel). We also determined whether any of the RAD51-interacting fragments were targets for phosphorylation. Incubation of the nine fusions with an extract prepared from asynchronous HeLa cells in the presence of [γ-32P]ATP revealed that B2-1, B2-2, B2-7 and B2-9 were phosphorylated (Fig. 1c), leading us to further explore the possibility that phosphorylation affects interactions between the BRCA2 C terminus (B2-9) and RAD51.

Figure 1: Phosphorylation of the C-terminal RAD51-interaction domain of BRCA2.
figure 1

a, Schematic diagram of the nine overlapping GST–BRCA2 fusion proteins, designated B2-1 to B2-9. The B2-9 region was further subdivided into the three overlapping GST fusions designated TR1, TR2 and TR3. The BRC repeats (blue), helical domain (grey), OB folds (black) and NLS sequences (red) are indicated. b, Interactions between the GST–BRCA2 fusions B2-1 to B2-9 with RAD51. GST fusion proteins attached to beads were incubated with HeLa extracts or purified recombinant RAD51 (rRAD51). After being washed, the beads were boiled and the GST fusions and bound RAD51 protein were detected by SDS–PAGE followed by western blotting with an anti-GST polyclonal antibody or the anti-RAD51 monoclonal antibody 14B4. c, Phosphorylation of B2-1, B2-2, B2-7 and B2-9 by HeLa extracts. GST fusion proteins attached to beads were incubated with an extract from asynchronous HeLa cells in the presence of [γ-32P]ATP. After being washed, 32P-labelled products were detected by SDS–PAGE followed by autoradiography. d, B2-9 (and TR2) phosphorylation activity co-purifies with the kinase that phosphorylates histone H1. Fractions from the butyl-Sepharose (BS) column are shown. e, Inhibition of B2-9 and TR2 phosphorylation by the CDK–cyclin A inhibitors roscovitine and p21. Butyl-Sepharose fraction 15 was used for the kinase assays. CDC6 was a control. f, Phosphorylation of the BRCA2 GST fusions by purified CDK2–cyclin A. Reactions were performed as described for c, except that purified CDK2–cyclin A (0.7 nM) replaced the HeLa extract. g, Dependence of B2-9 phosphorylation on the cyclin A (Cy) recognition sequence. Reactions were performed as described for f, except that they were supplemented with a Cy peptide inhibitor. Histone H1 was used as a control.

Full size image

First, the kinases(s) responsible for B2-9 phosphorylation were identified. When the extract was fractionated by precipitation with ammonium sulphate, followed by chromatography on phosphocellulose and butyl-sepharose (BS), we found that the activity migrated with histone H1 phosphorylation activity (Fig. 1d). Because histone H1 is a known substrate for cyclin-dependent kinases (CDKs) these results implicated CDKs. Confirmation was obtained when it was shown that B2-9 phosphorylation was blocked by roscovitine, a chemical inhibitor of CDK, and by p21, a cyclin A- and cyclin E-associated CDK inhibitor (Fig. 1e). Similar results were obtained with CDC6, a known protein target for CDK31. B2-9 could also be phosphorylated by purified CDK2–cyclin A (Fig. 1f) in a reaction that was dependent on a cyclin-binding sequence, known as a Cy or RXL motif32, as indicated by inhibition with a Cy peptide inhibitor (Fig. 1g). Control reactions showed that the Cy peptide did not inhibit histone H1 phosphorylation, which is consistent with the lack of a Cy motif in H1.

Division of B2-9 into three smaller GST fusions (designated TR1, TR2 and TR3; Fig. 1a) revealed that the kinase present in butyl-sepharose fraction 15 phosphorylated the overlapping fragments TR1 and TR2, whereas little activity was observed on TR3 (Fig. 1d, e, and data not shown). The phosphorylation of TR2 was particularly interesting because the RAD51 interaction domain maps exclusively to TR2, as determined by pull-down assays (Fig. 2b, lane 3). Sequence analysis of the TR2 region shows that it is highly conserved between mammalian species and contains five conserved S/TP potential CDK target motifs (Fig. 2a, red letters). Consistent with the data of Fig. 1g, the TR2 region also contains a Cy (RxL)-binding motif (amino acids 3,269-3,271; Fig. 2a, green box). Subsequent analyses were therefore limited to the RAD51 interaction domain TR2.

Figure 2: S3291 phosphorylation blocks interactions between the C terminus of BRCA2 and RAD51.
figure 2

a, Sequence alignment of the C-terminal (B2-9) region of BRCA2 from different species. The TR2 region (yellow shading) containing two NLS sequences (blue bars) is indicated. Conserved amino acids (blue) and CDK target sites (red letters) are highlighted. The green box indicates the cyclin A recognition motif (Cy or RxL); the red box indicates the CDK target site at S3291. b, Interaction of GST–TR2 with purified RAD51 protein. Reactions were performed as described in Fig. 1b. c, Glutamate-substituted B2-9 mutants in which one (or more) CDK consensus phosphorylation site was mutated. d, Glutamate substitution at S3291 blocks interaction with RAD51. All mutant GST fusion proteins (designated a to m as shown in c) were purified and mixed with recombinant RAD51; interactions were detected by pull-down assays. e, Synthetic TR2 phosphopeptides in which the various CDK target sites were phosphorylated (Pho). f, Phosphorylation of S3291 inhibits the interaction of TR2 with RAD51. Biotinylated (indicated by B in a circle) TR2 phosphopeptides were mixed with recombinant RAD51, and complexes were detected by streptavidin pull-downs. Biotinylated BRC4 peptide was used as a control.

Full size image

S3291 phosphorylation modulates RAD51 binding

To determine whether the five S/TP potential CDK target motifs in TR2 might be important for regulating interactions with RAD51, a series of glutamate-substituted mutants were constructed that mimic a negatively charged phosphate residue (Fig. 2c, constructs designated a to m). Strikingly, glutamate substitution at S3291 blocked interactions between B2-9 and RAD51 (Fig. 2d, lanes 3, 4 and 7–13). All other mutants interacted normally with RAD51. Most importantly, we found that phosphorylation of S3291 modulated the ability of TR2 to interact with RAD51. To show this, a series of biotinylated peptides were synthesized that were phosphorylated at defined positions (Fig. 2e). When incubated with RAD51, the three peptides phosphorylated at S3291 did not bind RAD51 (Fig. 2f, lanes 3, 5 and 6).

Dynamic regulation of S3291 phosphorylation

Because these data indicate that phosphorylation of S3291 by CDK might have a crucial function in modulating the ability of BRCA2 to interact with RAD51, we next analysed the in vivo phosphorylation status of S3291 with a phospho-specific antibody (S3291Ph). GST–B2-9, overexpressed in asynchronous HT1080 cells, was weakly recognized by S3291Ph (Fig. 3a, lane 3), whereas after nocodazole-induced mitotic arrest we observed a strong interaction with the phospho-specific antibody (Fig. 3a, lane 4). Similar results were obtained with endogenous full-length BRCA2 (Fig. 3b, compare lanes 3 and 4). After treatment with nocodazole, the CDK responsible for S3291 phosphorylation is likely to be CDK1–cyclin B, because only low concentrations of cyclins A and E were detectable in these extracts (Fig. 3c). We did not observe S3291Ph signals when extracts were treated with lambda phosphatase (data not shown).

Figure 3: Analysis of the phosphorylation status of S3291 in vivo.
figure 3

a, GST–B2-9 was overexpressed in HT1080 cells. The cells were treated with or without nocodazole (Noc), and after 24 h extracts were prepared and GST–B2-9 was pulled down with the GST tag. Samples analysed by SDS–PAGE were then probed with an anti-GST antibody or the phospho-specific S3291 antibody. b, HeLa cells were treated with or without nocodazole and extracts probed for full-length BRCA2 using a mouse monoclonal BRCA2 antibody (OP95) or the S3291 phospho-specific antibody. c, Whole extracts made from cells treated with or without nocodazole were blotted with the indicated monoclonal antibodies. d, HeLa cells were synchronized by using a double thymidine block. At various times after release samples were taken; extracts were prepared and blotted with the indicated antibodies. In the lower panel, cyclin A (red) or cyclin B (blue) immunoprecipitates were analysed for kinase activity with TR2 (lines) and histone H1 (histograms) as substrates. Maximum kinase activity was arbitrarily designated as 1.0.

Full size image

To determine the phosphorylation status of S3291 throughout the cell cycle, HeLa cells were synchronized at the G1–S boundary using a double thymidine block. Whole cell extracts were prepared at various times after release from cell cycle arrest and probed with the S3291 phospho-specific antibody as well as a variety of cyclin antibodies. In these experiments, the phospho-specific antibody MPM-2 was used to detect mitotic entry. S3291 phosphorylation occurred as the cells progressed from G2 to M phase after the expression of both cyclin A and cyclin B (Fig. 3d). Kinase activity measurements, determined after immunoprecipitation of cyclin A and cyclin B, and using TR2 and histone H1 as substrates, showed that both CDK–cyclin A and CDK1–cyclin B may phosphorylate the RAD51 interaction domain located at the C terminus of BRCA2. However, the overall phosphorylation status of S3291 is likely to be affected not only by these cyclin-associated kinase activities but also by changes to phosphatase activities and/or factors that modulate substrate recognition by CDKs throughout the cell cycle.

The S3291 phospho-specific antibody was also used to probe full-length endogenous BRCA2 in extracts made from cells at various times after treatment with ionizing radiation. Phosphorylation of S3291 decreased after ionizing radiation (Fig. 4a), a change that was coupled with a more than 50% decrease in kinase activity (Fig. 4b). A decrease in S3291 phosphorylation was not observed in ATM-deficient cells (data not shown). Inhibition of CDK activity in normal cells by treatment with the CDK inhibitor roscovitine promoted a similar effect to radiation, contrasting with the large increase in kinase activity observed after treatment with nocodazole. Using cells stably expressing GST–B2-9, we found that decreased S3291 phosphorylation caused by ionizing radiation was associated with a doubling in ability to bind RAD51 (Fig. 4c). Thus, S3291 phosphorylation, which is decreased by treatment with ionizing radiation or stimulated by a nocodazole-induced block to cell cycle progression, is inversely correlated with RAD51 binding ability. These results show that the phosphorylation status of S3291 is crucial for modulating interactions between the C terminus of BRCA2 and RAD51 protein.

Figure 4: Involvement of S3291 phosphorylation in mediating RAD51 interactions and efficient homologous recombination.
figure 4

a, HT1080 cells were treated with 4 Gy ionizing radiation and at the indicated times whole cell extracts were prepared and blotted with S3291 phospho-specific antibody or the BRCA2 monoclonal antibody OP95. Only the section of the blot showing full-size BRCA2 is shown. The ratio of BRCA2 phosphorylated at S3291 to unphosphorylated BRCA2 is indicated. b, Phosphorylation of TR2 after treatment with ionizing radiation (IR) or drugs. c, HT1080 cells expressing GST–B2-9 fusion protein were left untreated (NT) or treated with 4 Gy ionizing radiation, 0.13 µM nocodazole (Noc) or 10 µM roscovitine (Rosc). Irradiated cells were then grown for the indicated durations. Drug treatment was maintained for 24 h, at which time extracts were prepared and GST pull-down assays were used to determine the amounts of RAD51 associated with B2-9. d, Wild-type TR2, S3291A or S3291E, were expressed as GST fusions in 293T cells, and interactions with endogenous RAD51 were detected using GST pull-down assays. e, Schematic diagram indicating the double-strand break repair assay. 293DR-GFP cells have two GFP fragments integrated into genomic DNA, one of which has a unique I-SceI cleavage site. Expression of I-SceI endonuclease introduces a double-strand break that can be repaired by homologous recombination (HR) to produce a functional GFP gene (hatched). f, GST–TR2, GST–S3291A or GST–S3291E expression vectors were co-transfected with the I-SceI expression vector into 293DR-GFP cells. Recombinational repair of the induced double-strand break was analysed by fluorescence-activated cell sorting analysis. The error bars show standard deviations for three independent experiments (n = 3). A highly significant P value (equivalent to 0.000769) relating to I-SceI/GST and I-SceI/TR2 was determined by using a standard t-test.

Full size image

TR2 expression affects double-strand break repair

Analysis of the Breast Cancer Information Core database revealed that there are four independent examples of mutations in the S3291-P3292 CDK target sequence in individuals identified by increased cancer incidence. We therefore wished to compare the efficiency of recombinational repair in wild-type cells with those carrying an S3291 mutation in BRCA2. Initially, attempts were made to measure homologous recombination efficiency in BRCA2-defective human and hamster cells complemented with either wild-type or S3291 mutant derivatives of BRCA2. However, we found that BRCA2-defective cell lines grew very poorly and exhibited massive cell death after transfection (data not shown). As an alternative approach, we determined whether expression of GST–TR2, or a glutamate-substituted derivative (designated S3291E) that mimics a constitutively phosphorylated form of the protein, exerted different effects on the efficiency of recombinational repair. Preliminary studies showed that GST–TR2, but not the S3291E derivative, when expressed in vivo, interacted with RAD51 protein (Fig. 4d, lanes 2 and 4). We therefore used I-SceI endonuclease to induce chromosomal double-strand breaks in cells overexpressing TR2 or S3291E and determined the efficiency of double-strand break repair by the formation of green fluorescent protein (GFP)-positive cells22 (Fig. 4e).

Expression of TR2 reduced the efficiency of recombinational repair by 50%, whereas S3291E did not affect recombination (Fig. 4f). Control experiments, in which I-SceI cleavage was not induced, indicated that cell viability was unaffected by expression of the TR2 constructs (data not shown). These results show that overexpression of the non-phosphorylated RAD51-interaction domain of BRCA2 results in a dominant-negative recombination-deficient phenotype, whereas expression of the S3291E derivative had no affect on recombination efficiency. Additional experiments showed that expression of a construct carrying a serine → alanine mutation at S3291 also did not confer a dominant-negative phenotype (Fig. 4f). As with S3291E, the S3291A derivative did not bind RAD51 (Fig. 4c). These studies show that interactions between the C-terminal region of endogenous BRCA2 and RAD51 are necessary for efficient double-strand break repair by homologous recombination.

In control experiments, we found that a peptide 38 amino acids long (residues 3,265–3,302) interacts with RAD51, whereas an alanine or glutamate substitution at S3291 blocks the interaction. In contrast, conversion of Pro 3292 (P3292) to leucine did not affect RAD51 binding (data not shown). Because it is unlikely that the S3291A derivative would cause a conformational change in a small peptide but the P3292 derivative would not, these experiments exclude the possibility that the observed effects are due to gross conformational changes and emphasize that S3291 is a crucial residue necessary for interaction with RAD51.

A mechanism for RAD51 control

The work described here provides new insight in the involvement of BRCA2 in break repair by homologous recombination. Most importantly, the phosphorylation status of S3291 seems crucial for regulating the interaction of RAD51 with the C-terminal region of BRCA2. We found that radiation treatment resulted in reduced CDK-mediated phosphorylation of S3291, which stimulated the interaction between the BRCA2 C terminus and RAD51. Moreover, during the cell cycle, we observed that S3291 phosphorylation increased throughout G2/M and was reduced in G1. The low levels of phosphorylation observed at S phase are likely to be important for the repair of replication-induced breaks. Conversely, phosphorylation of S3291 in mitosis might be responsible for the inactivation of homologous recombination, and/or might contribute to the degradation of BRCA2 before BRCA2 synthesis occurs de novo in late G1 (ref. 33).

The results provide new insight into a mechanism by which checkpoint control and recombinational repair are coupled in mammalian cells. The repair pathway is thought to involve ATM, p53 and CHK2, all of which are related to breast cancer susceptibility and are involved in CDK inactivation and cell cycle arrest after DNA damage. ATM transduces the DNA damage signal to p53, leading to the transcriptional activation of p21 that inhibits CDK activity34. ATM also phosphorylates and activates CHK2, which in turn phosphorylates CDC25 phosphatase, leading to its degradation or cytoplasmic sequestration35. Because CDC25 is required for CDK activation, loss of CDC25 activity will help to maintain the phosphorylated/inactivated state of CDK36. Inactivation of CDK will reduce its ability to phosphorylate S3291 of BRCA2, thus stimulating interactions between RAD51 and the C-terminal region of BRCA2.

The consequences of loss of phosphorylation of S3291 after CDK inhibition lead us to suggest a model in which the BRC repeats and the C-terminal RAD51 interaction domain are functionally distinct. As suggested previously, under normal conditions interactions at the BRC repeats may serve to inactivate RAD51 by providing a scaffold for monomer binding19,20. It is expected that phosphorylation of S3291 by CDK will then block interactions between RAD51 and the C terminus of BRCA2. However, after ionizing radiation, and possibly other DNA-damaging treatments, decreased CDK activity leads to the loss of phosphorylation at S3291 and activation of the C-terminal RAD51 interaction domain.

Structural studies of the C-terminal region of BRCA2 complexed with DSS1 protein revealed the presence of three globular domains that form OB folds, which are also found in single-strand DNA-binding proteins such as RPA37. It is therefore tempting to speculate that the binding of RAD51 by the C-terminal interaction domain of BRCA2, which occurs after the loss of S3291 phosphorylation, might contribute to the loading of RAD51 to single-stranded DNA. The binding of RAD51 by this region might be an important component of the cellular DNA damage response by accepting RAD51 monomers from the BRC repeats and facilitating their transfer to DNA, where functional RAD51 nucleoprotein filaments are assembled. Indeed, it is possible that the C-terminal region of BRCA2 could fulfil a key role in the hand-over of RAD51 to the initiating DNA substrate.

The involvement of CDK in the recruitment of RAD51 to DNA break sites was recently shown in Saccharomyces cerevisiae38,39. These studies indicated that CDKs in yeast promote the activation of homologous recombination, in contrast to their role in mammalian cells. This difference might be due the fact that, in response to DNA damage, budding yeasts arrest their cell cycle in the mitotic phase with elevated CDK activity, rather than in G2 as in other organisms. Moreover, a functional BRCA2 homologue has not been identified in yeast.

Although it is unlikely that S3291 phosphorylation provides the only regulatory mechanism that controls recombinase activity, this model provides a simple explanation for the activation of RAD51 in response to DNA-damaging agents. It is also consistent with observations showing that the deletion of exon 27 results in recombination deficiency and radiation sensitivity12,40,41,42. Moreover, exon 27 deletion, or a more extensive deletion of the C-terminal half of BRCA2 that leaves several of the BRC repeats intact, results in a loss of ionizing-radiation-induced RAD51 focus formation40,41. The importance of such dynamic changes to the S3291 phosphorylation site is also highlighted by observations showing that it is a cancer-related mutation site, as indicated by the incidence of P3292L mutations in individuals affected with breast cancer.

Methods

Cell culture

Human 293T and HT1080 cells were cultured on plates in DMEM medium supplemented with 10% v/v fetal bovine serum and streptomycin/penicillin (100 units ml-1). For 293DR-GFP cells the plates were coated with 0.01% poly-(l-lysine). For the thymidine block, 25% confluent HeLa cells were cultured in the presence of 2 mM thymidine for 19 h, and released into fresh medium for 9 h. They were then incubated in the presence of 2 mM thymidine for 16 h before release into fresh medium.

Plasmids

The GST constructs B2-1 to B2-9 have been described30. GST–TR1, GST–TR2 and GST–TR3 were constructed by cloning by polymerase chain reaction (PCR) into pGEX-4T3 (Amersham) and the mammalian expression vectors pGST–B2-9, pGST–TR1, pGST–TR2 and pGST–TR3 were made by PCR cloning into the Gateway entry vector pDONR221 and transferred into pDEST27 (Invitrogen).

Proteins

Recombinant RAD51, CDK2–cyclin A and GST–p21N were prepared as described43,44,45. The GST–p53 peptide was purified from pGEX 2T p53(9-22), a gift from Dr M. Kastan. CDC6 (residues 1–110) was purified as a GST fusion protein. Histone H1 was purchased from Upstate. The Cy peptide PSACRNLFGPVD corresponds to residues 26–37 of p27. All peptides were prepared by solid-phase synthesis and purified by high-performance liquid chromatography; their sequences were confirmed by mass spectroscopy. Where indicated, peptides contained an N-terminal biotin group with an aminohexanoic acid spacer.

BRCA2 GST fusion plasmids were grown in BL21 RIL codon plus Escherichia coli (Stratagene) and protein was expressed by incubation overnight at 18 °C after the addition of 10 µM isopropyl β-d-thiogalactoside. Proteins were purified from soluble extracts with glutathione Sepharose 4 fast flow beads (Amersham).

Mammalian extracts

Mammalian cell pellets (about 0.6 g) were suspended in 3 ml HE buffer (20 mM Hepes-NaOH pH 7.6, 2 mM EGTA, 1.5 mM MgCl2, 50 mM NaF, 0.5% Nonidet P40, 10% glycerol, 1 mM Na3VO4, 20 mM β-glycerophosphate, 1 mM dithiothreitol, 1.5 µg ml-1 aprotinin and 1 µg ml-1 each of leupeptin, pepstatin A and chymostatin) containing 450 mM KCl, and placed on ice for 30 min. After centrifugation in a Microfuge for 10 min, the supernatant was diluted with 2 volumes of HE buffer (to 150 mM KCl), and then centrifuged at 35,000 r.p.m. for 60 min in a Beckman 70.1Ti rotor. Supernatants were used for protein interaction studies and kinase assays.

In vitro protein interactions

GST fusion proteins (about 1 µg) bound to glutathione beads were mixed with HeLa extract (1 ml; about 5 mg) and incubated for 2.5 h at 4 °C. Alternatively, the bead-bound fusions were mixed with BSA (5 mg ml-1) in HE buffer supplemented with 150 mM KCl, and after 30 min they were supplemented with RAD51 (50 ng). After a further 2 h, the beads were washed with HE buffer containing 150 mM KCl. Complexes were then boiled in SDS sample buffer and analysed by SDS–polyacrylamide-gel electrophoresis (SDS–PAGE) followed by western blotting.

Interactions between biotinylated peptides (100 ng) and RAD51 (80 ng) were performed similarly, and complexes were analysed by pull-down assays with streptavidin beads, followed by SDS–PAGE and western blotting.

Kinase assays

Protein substrates (2 µg in 20 µl) were phosphorylated in kinase buffer (10 mM Hepes-HCl pH 7.6, 50 mM β-glycerophosphate, 50 mM NaCl, 10 mM MgCl2, 10 mM MnCl2, 5 µM ATP, 1 mM dithiothreitol, 1 µCi [γ-32P]ATP) by the addition of 1 µl cell extract or fractionated extract. After 30 min at 30 °C, reactions were stopped by the addition of SDS sample buffer and boiled for 5 min. Proteins were analysed by SDS–PAGE and revealed by Coomassie blue staining and autoradiography.

Generation of 293DR-GFP reporter cells

About 5 × 106 293 Tet-Off cells (Clontech) were transfected with 50 µg SacI/KpnI-linearized hprtDR-GFP reporter vector46 by electroporation. After 24 h, puromycin selection (1 µg ml-1) was applied and a clone with an intact copy of the reporter (TOS A4) was identified by Southern blotting47.

I-SceI-induced double-strand break repair

To examine recombination induced by double-strand breakage, 293DR-GFP cells were co-transfected with the I-SceI expression vector pCBASce48 and pGST, pGST–TR2, pGST–TR2A or pGST–TR2E. For transfection, 0.6 ml of cells at 8 × 106 ml-1 were electroporated at 280 V and 960 µF with 30 µg pCBASce and 30 µg BRCA2-derived expression vectors. To determine transfection efficiency, 293DR-GFP cells were transfected under the same conditions with 3 µg GFP-expressing vector pNZE-CAG47 together with 27 µg mock plasmid DNA and the BRCA2-derived expression vector (30 µg). For each transfection a single-cell suspension was prepared 48 h after electroporation, and the percentage of GFP-positive cells was determined by flow cytometry on a Becton Dickinson FACScan. Dead cells were excluded from the analysis by staining with 7-amino-actinomycin D (Sigma), and fluorescence-activated cell sorting data were analysed with CellQuest software.