Hyperphosphorylation of the BARD1 tumor suppressor in mitotic cells.

Although the BRCA1 tumor suppressor has been implicated in a number of cellular processes, it plays an especially important role in the DNA damage response as a regulator of cell cycle checkpoints and DNA repair pathways. In vivo, BRCA1 exists as a heterodimer with the BARD1 protein, and many of its biological functions are mediated by the BRCA1-BARD1 complex. Here, we show that BARD1 is phosphorylated in a cell cycle-dependent manner and that the hyperphosphorylated forms of BARD1 predominate during M phase. By mobility shift analysis and mass spectrometry, we have identified seven sites of mitotic phosphorylation within BARD1. All sites exist within either an SP or TP sequence, and two sites resemble the consensus motif recognized by cyclin-dependent kinases. To examine the functional consequences of BARD1 phosphorylation, we used a gene targeting knock-in approach to generate isogenic cell lines that express either wild-type or mutant forms of the BARD1 polypeptide. Analysis of these lines in clonogenic survival assays revealed that cells bearing phosphorylation site mutations are hypersensitive to mitomycin C, a genotoxic agent that induces interstrand DNA cross-links. These results implicate BARD1 phosphorylation in the cellular response to DNA damage.

Although the BRCA1 tumor suppressor has been implicated in a number of cellular processes, it plays an especially important role in the DNA damage response as a regulator of cell cycle checkpoints and DNA repair pathways. In vivo, BRCA1 exists as a heterodimer with the BARD1 protein, and many of its biological functions are mediated by the BRCA1-BARD1 complex. Here, we show that BARD1 is phosphorylated in a cell cycle-dependent manner and that the hyperphosphorylated forms of BARD1 predominate during M phase. By mobility shift analysis and mass spectrometry, we have identified seven sites of mitotic phosphorylation within BARD1. All sites exist within either an SP or TP sequence, and two sites resemble the consensus motif recognized by cyclin-dependent kinases. To examine the functional consequences of BARD1 phosphorylation, we used a gene targeting knock-in approach to generate isogenic cell lines that express either wild-type or mutant forms of the BARD1 polypeptide. Analysis of these lines in clonogenic survival assays revealed that cells bearing phosphorylation site mutations are hypersensitive to mitomycin C, a genotoxic agent that induces interstrand DNA cross-links. These results implicate BARD1 phosphorylation in the cellular response to DNA damage.
Although hereditary breast cancer can often be traced to germ line mutations in the BRCA1 tumor suppressor gene, it is still unclear how BRCA1 lesions promote mammary carcinogenesis (1). The major isoform of BRCA1 is a large polypeptide that contains a RING domain at its N terminus and two tandem copies of the BRCT domain at the C terminus (2). In vivo, BRCA1 exists primarily as a heterodimer with BARD1, a distinct but related protein that also harbors an N-terminal RING motif and two C-terminal BRCT domains (3). Both proteins colocalize within the same nuclear foci of S phase cells (4,5), and their interaction is mediated by sequences encompassing their respective RING motifs (3). Co-immunoprecipitation analysis indicates that most of the cellular pool of endogenous BRCA1 exists in the form of a heterodimer with BARD1 (6,7), and structural studies have shown that dimerization occurs through a four-helix bundle formed by ␣-helices flanking the RING motifs of both proteins (8). Because the phenotypes of mice null for either Brca1 or Bard1 are essentially indistinguishable (9), the major functions of both proteins are likely to be mediated by the BRCA1/BARD1 heterodimer. This notion is supported by the fact that BRCA1 and BARD1 together form a potent enzymatic complex that can catalyze ubiquitin polymerization in vitro (10 -17). Recent studies have also shown that BARD1 is required for nuclear retention of BRCA1 (18) and that, together with BRCA1, it modulates mRNA processing during the DNA damage response (19,20) and promotes homology-directed repair of chromosomal breaks (21).
It is now apparent that phosphorylation exerts important controls over BRCA1 function. Early studies established that BRCA1 is hyperphosphorylated in several cellular settings. During normal cell cycle progression, hyperphosphorylation of BRCA1 occurs at the G 1 /S transition, and various phosphorylated forms of BRCA1 are apparent throughout the subsequent S, G 2 , and M phases (5,(22)(23)(24)(25). In addition, some BRCA1 phosphorylation events are induced in response to growth factor signaling (26,27). The phosphorylation status of BRCA1 also changes dramatically when cells are subjected to genotoxic stress (5,24). For example, in cells exposed to ionizing radiation (IR), 1 the ATM kinase phosphorylates BRCA1 at a series of sites that includes Ser 1387 , Ser 1423 , and Ser 1524 (28,29), whereas Chk2 phosphorylates Ser 988 (30). Indeed, several of these modifications have now been implicated in certain downstream functions of BRCA1. Thus, BRCA1 phosphorylation at Ser 1387 is required for the IR-induced intra-S (but not G 2 /M) checkpoint (31), whereas phosphorylation at Ser 1423 is necessary for the IR-induced G 2 /M (but not intra-S) checkpoint (32,33). In addition, phosphorylation at Ser 988 (but not Ser 1387 or Ser 1423 ) is required for BRCA1 modulation of homologous and non-homologous DNA recombination (34). These results indicate that the downstream functions of BRCA1 can be influenced by specific phosphorylation events.
By most measures, the biological functions of BRCA1 appear to be mediated through the heterodimeric BRCA1/BARD1 complex (35). Here, we provide evidence that BARD1 is also subject to post-translational phosphorylation and that hyperphosphorylated forms of BARD1 arise primarily during mitosis. Each of the seven identified BARD1 phosphorylation sites resides within an SP or TP sequence motif, indicating the involvement of proline-directed kinases. Significantly, cells that express mutant BARD1 polypeptides that lack these phosphorylation sites are hypersensitive to genotoxic stress, suggesting a role for BARD1 phosphorylation in the cellular response to DNA damage.

EXPERIMENTAL PROCEDURES
Cell Culture-The HCT116, HeLa, and H1299 cell lines were obtained from American Type Tissue Culture Collection. HCT116 cells were cultured in McCoy's 5A medium, and HeLa and H1299 cells were cultured in Dulbecco's modified Eagle's medium; both media were supplemented with 100 g/ml penicillin/streptomycin, 2 mM L-glutamine, and 10% fetal bovine serum. For cell cycle analysis, HeLa cells were synchronized in prometaphase by a thymidine/nocodazole block and subjected to fluorescence-activated cell sorting (FACS) and Western analysis as described previously (36).
Antibodies-The BARD1-specific polyclonal antiserum (3) and monoclonal antibody (36) have been described. Other antibodies used included anti-␣-tubulin (Ab-1, Oncogene Science Inc.); anti-BRCA1 (C-20), anti-Cdc27 (H-300), and anti-hemagglutinin (HA) (Y-11, all from Santa Cruz Biotechnology, Inc.); and anti-FLAG (M2, Sigma). A phospho-specific antiserum that recognizes Thr 299 -phosphorylated BARD1 was generated and affinity-purified at Zymed Laboratories Inc. (South San Francisco, CA) from rabbits immunized with the phosphothreonine-containing peptide CSRNEVVpTPEK conjugated to keyhole limpet hemocyanin. For detection of exogenous BARD1 polypeptides with the phospho-Thr 299 -specific antiserum (e.g. Fig. 5B), direct Western analysis was performed as described (36), except that 5% bovine serum albumin was used (instead of milk) in the TBS-T (20 mM Tris-HCl, pH 7.6, 137 mM NaCl, 0.1% Tween 20) buffer for blocking and for primary antibody incubation. For detection of endogenous BARD1 polypeptides with the phospho-specific antiserum (see Fig. 5C), cell lysates in 500 l of low salt Nonidet P-40 buffer (supplemented as described previously (36), except using 2ϫ phosphatase inhibitors: 10 mM NaF and 0.2 mM Na 3 VO 4 ) were precleared with 50 l of protein G-Sepharose beads (20% slurry; Amersham Biosciences) by rotating at 4°C for 1 h. After high speed centrifugation for 2 min, the supernatants were incubated with 5 l of anti-BARD1 monoclonal antibody for 2 h at 4°C. Protein G-Sepharose beads (50 l, 20% slurry) were added, and the mixtures were then incubated overnight at 4°C. The beads were washed once with low salt Nonidet P-40 buffer, twice with high salt (1 M NaCl) Nonidet P-40 buffer, and once again with low salt Nonidet P-40 buffer. The bound proteins were eluted from the beads by denaturing at 100°C in 2ϫ protein loading dye. The beads were filtered through compact reaction columns (U. S. Biochemical Corp.), and the eluted proteins were fractionated by SDS-PAGE for Western analysis with the phospho-Thr 299specific antiserum. To examine the ability of hyperphosphorylated BARD1 species to interact with BRCA1 (see Fig. 1C), cell lysates were immunoprecipitated with the anti-BARD1 monoclonal antibody (36) or an anti-BRCA1 monoclonal antibody (1 mg/ml; Ab-1, Oncogene Science Inc.) prior to immunoblotting with the BARD1-specific polyclonal antiserum.
Mapping BARD1 Phosphorylation Sites by Mutation Analysis-Expression plasmids encoding human BARD1 fragments were generated by inserting PCR-generated segments of human cDNA into the pCMV-FL3-Not vector, a derivative of the p3XFLAG-CMV-7 plasmid (Sigma). Plasmids encoding BARD1 mutants were prepared using the Stratagene QuikChange site-directed mutagenesis kit. To test for electrophoretic mobility shifts of BARD1 fragments during mitosis, HeLa cells were transfected with a BARD1 expression plasmid using Lipofectamine (Invitrogen). 10 -12 h after transfection, the cells were split into two plates: one to continue asynchronous cell growth and the other to be subjected to a thymidine/nocodazole block starting 12 h later (36). The asynchronous cells were collected 36 h after transfection, and the arrested cells were collected by shake-off 48 h after transfection. The cells were lysed in low salt Nonidet P-40 buffer supplemented with protease and phosphatase inhibitors (36). The lysates were then resolved by SDS-PAGE on 6, 10, or 12% Trisglycine gels (Invitrogen) and subjected to Western analysis with anti-FLAG monoclonal antibody M2.
Purification of Phosphorylated BARD1-To generate cell lines that stably express an exogenous BARD1 polypeptide, H1299 cells were transfected with a mammalian expression plasmid encoding human full-length BARD1 with an N-terminal FLAG epitope (NH 2 -MDYKD-DDDKGGS) and a C-terminal HA epitope (GGYPYDVPDYA-COOH) in the pIRESpuro vector (Clontech). H1299 cell transformants were then selected in 2 g/ml puromycin and maintained in 1 g/ml puromycin. To generate FLAG-BARD1-HA polypeptides for mass spectrometry, H1299 cells expressing FLAG-BARD1-HA (subclone 7) were seeded on 16 150-mm plates; and at Ͻ50% confluence, the cells were treated with 100 ng/ml nocodazole for 16 h. The cells were then collected by scraping to yield a cell pellet of ϳ1 ml. All lysis buffers were supplemented with 0.1 mM dithiothreitol, protease inhibitors (0.5 mM phenylmethylsulfonyl fluoride and Roche Protease Inhibitor Cocktail), and phosphatase inhibitors (5 mM NaF and 0.1 mM Na 3 VO 4 ). The cell pellet was resuspended in 25 ml of Buffer A (10 mM HEPES, pH 7.9, 10 mM KCl, 0.1 mM EDTA) for 15 min, followed by vigorous shaking for 15 s upon addition of 0.5 ml of 10% Nonidet P-40. After centrifugation at 3000 rpm for 5 min, the pellet was lysed in 2.5 ml of Buffer C (20 mM HEPES, pH 7.9, 400 mM NaCl, 1 mM EDTA) by vortexing for 15 min at 4°C. After centrifugation at 3000 rpm for 5 min, the supernatant was then added to 2.5 ml of Buffer D (20 mM HEPES, pH 7.9, 1 mM EDTA) and 100 l of 10% Nonidet P-40. The lysate was clarified by centrifugation at 25,000 ϫ g for 30 min and filtration through a 0.45-m filter. The lysate was then incubated with 250 l of anti-FLAG antibody M2 beads (20% slurry) overnight at 4°C. The beads were washed three times with BC200 buffer (20 mM Tris-HCl, pH 7.9, 200 mM NaCl, 0.2% Nonidet P-40, and 20% glycerol) and eluted twice for 3 h each with 500 g/ml FLAG peptide in 100 l of BC200 buffer. The eluted protein was then trichloroacetic acid-precipitated. For mass spectrometric analysis, the trichloroacetic acid-precipitated pellet was resuspended in 20 l of 1ϫ protein loading dye, denatured at 70°C for 10 min, and alkylated by incubation with 2 l of 125 mM iodoacetamide for 1 h in the dark (37). The sample was separated by SDS-PAGE on a 4 -12% Tris-glycine gel, which was then fixed and stained with GelCode Blue stain reagent (Pierce) according to the manufacturer's protocol.
Mass Spectrometric Analysis of BARD1 Phosphorylation Sites-BARD1 in-gel enzymatic digestion and extraction from gel bands were performed essentially as described (38, 39). Briefly, two Coomassie Blue-stained BARD1 protein bands containing a total protein amount of 1-2 g were excised and divided into four portions. Each portion was digested with 0.2 g of modified trypsin, endoproteinase Lys-C, endoproteinase Asp-N, or endoproteinase Glu-C (Roche Applied Science). Each digestion was terminated by adding 30 l of 7% formic and 0.1% trifluoroacetic acid containing 0.5 l (dry volume) of a 1:1 mixture of POROS R2 and R3 beads (Applied Biosystems, Foster City, CA) directly into the solution containing the gel pieces. The peptides were extracted in the presence of the POROS beads for 3-12 h. The solution containing the POROS beads was aspirated into an Eppendorf Geloader tip pinched at its tip. The column formed at the pipette tip was washed with 10 l of 0.1% trifluoroacetic acid in water, and the peptides were directly eluted onto the MALDI plate with 1 l of matrix solution (2ϫ dilution of a saturated solution of 2,5-dihydroxybenzoic acid in 60% methanol and 5% acetic acid). BARD1 phosphorylation sites were analyzed by MALDI-QqTOF mass spectrometry and tandem mass spectrometry on a QSTAR XL mass spectrometer with an oMALDI ion source (Applied Biosystems) operated in a positive ion mode using a nitrogen laser operating at 20 Hz. MALDI mass spectra were collected at m/z 700 -7000 for 2 min. MALDI tandem mass spectra were acquired for a period of time ranging from 1 to 6 min with the binning factor set at 8 (maximum number available in Analyst QS SP7) for sensitive fragment ion detection.
BARD1 Knock-out Targeting Vector (see Fig. 6A)-The targeting constructs were generated using the pBluescript II-SK plasmid (Stratagene) as a backbone. The homology arms were derived from bacteriophage clones containing human genomic BARD1 DNA (40). The 5Ј-homology arm spans a 3.6-kb region that includes the first 23 bp of exon 2 and sequences from the preceding intron. The 3Ј-homology arm spans a 3.6-kb region immediately downstream of exon 3. The arms flank a cassette that includes 1) the hygromycin resistance gene flanked by two loxP sites, 2) the herpes simplex virus thymidine kinase polyadenylation signal, and 3) the primer sequence GTGACAAGGAAT-GTCCTGTGGTGC to facilitate PCR screening of transformants. The diphtheria toxin gene was inserted 5Ј of the left homology arm for negative selection of random integration events.
BARD1 Knock-in Targeting Vector (see Fig. 6B)-The 5Ј-arm was derived from bacteriophage clones containing human genomic BARD1 DNA (40) and spans a 3.8-kb region that includes the first 23 bp of exon 2 and sequences from the preceding intron. The 3Ј-homology arm was generated by PCR amplification of genomic DNA from HCT116 cells with Platinum Taq High Fidelity polymerase (Invitrogen); it corresponds to a 3.5-kb region of genomic DNA including exon 3 and sequences from the preceding intron. The arms flank a cassette that includes 1) cDNA sequences encoding residues 61-777 of human BARD1; 2) an internal ribosome entry site followed by the neomycin resistance gene, with both flanked by loxP sites; 3) the herpes simplex virus thymidine kinase polyadenylation signal; and 4) the primer sequence ACTTGTCCAAGTTACAAAGCTTAGTAGACAGC to facilitate PCR screening of transformants. The diphtheria toxin gene was inserted 3Ј of the right homology arm for negative selection of random integration events.
Transfection and Selection of HCT116 Cells-The targeting vectors were linearized by restriction digestion with NotI (Promega) and purified by phenol/chloroform extraction and precipitation in 70% ethanol. Cells (6 ϫ 10 6 ) were seeded on each of three 100-mm plates; after 20 h, the cells on each plate (at ϳ70% confluence) were transfected with 10 g of the linearized targeting construct and 50 l of Lipofectamine in serum-free Opti-MEM I according to the protocol of Invitrogen. 24 h after transfection, the cells from each plate were trypsinized and seeded on twenty 96-well plates in medium containing 100 g/ml hygromycin B (Roche Applied Science) for knock-out targeting or 300 -325 g/ml G418 (Invitrogen) for knock-in targeting. Starting after ϳ19 days, cells from wells containing a single colony were transferred to fresh 96-well plates in medium containing 50 g/ml hygromycin B or 100 g/ml G418. After screening, cells were cultured in medium without the drug used for selection.
Screening HCT116 Transformants-Cells from each well of a 96-well plate were trypsinized and collected, leaving ϳ20% to repopulate the well; and genomic DNA was purified using the QIAamp DNA blood mini kit (Qiagen Inc.). Transformants were screened by PCR using Platinum Taq High Fidelity polymerase according to the protocol of Invitrogen with primers GTGACAAGGAATGTCCTGTGGTGC and GTCCATGATTTGA-CAAGTTCTTCCTAATTGC for knock-out targeting and with primers AC-TTGTCCAAGTTACAAAGCTTAGTAGACAGC and CCACCGGGTGTG-TCACATGC for knock-in targeting. The genomic DNAs from six clones were pooled for screening, after which individual clones in the positive pools were screened. For Southern analysis of HCT116 transformants, genomic DNA was isolated using the QIAamp DNA blood mini kit, and 10-g aliquots of BglII-digested DNA were subjected to Southern hybridization with the radiolabeled BARD1 5Ј-probe (a 1.3-kb segment of genomic BARD1 located between exon 1 and the region corresponding to the 5Ј-homology arm of the targeting constructs).
Western Analysis of HCT116 Transformants-Cells were arrested by a thymidine/nocodazole block as described above and collected by scrap-ing. The cell pellets were lysed in low salt Nonidet P-40 buffer supplemented with 2ϫ phosphatase inhibitors as described above, and the lysates were subjected to direct Western analysis or to immunoprecipitation with the anti-BARD1 monoclonal antibody followed by Western analysis as described above.
Clonogenic Survival Assays-Approximately 2000 cells were seeded on 100-mm plates. After 24 h, the cells were treated with the indicated concentrations of mitomycin C (Sigma) for 4 h. The cells were then washed twice with phosphate-buffered saline and incubated in complete medium, which was changed every 3-4 days. After 10 days, the cells were washed twice with phosphate-buffered saline, stained with 10% Giemsa solution (Invitrogen) in phosphate-buffered saline for 15 min, and washed twice more with phosphate-buffered saline, and drugresistant colonies were counted.

BARD1 Is
Hyperphosphorylated during Mitosis-Our previous study showed that BARD1 expression varies during cell cycle progression in a manner reminiscent of BRCA1 (36). Thus, the steady-state levels of BARD1 polypeptides are low in resting (G 0 ) cells and G 1 cycling cells, but increase markedly as cycling cells enter S phase and remain high throughout G 2 phase and mitosis. In the course of this study, we also noticed that the electrophoretic mobility of BARD1 from nocodazoletreated cells is retarded on SDS-polyacrylamide gels relative to that of BARD1 from unsynchronized cells. To investigate this phenomenon further, HeLa cells were arrested in prometaphase by a thymidine/nocodazole block, released into nocodazole-free medium, and collected at hourly time points. The BARD1 polypeptides from these cells were then visualized by Western analysis. As shown in Fig. 1A, the electrophoretic mobilities of the mitotic BARD1 polypeptides were retarded relative to that of the major BARD1 isoform from asynchronous cells (compare first and second lanes). The mobility shifts of mitotic BARD1 are due to phosphorylation because -phosphatase treatment converted these isoforms to a hypermobile band ( Fig. 1B, compare lanes 3 and 4). Hyperphosphorylated BARD1 was absent, however, in cells that had progressed into G 1 phase after being released from the nocodazole block for Ͼ2 h (Fig.  1A). In this respect, the pattern of BARD1 phosphorylation parallels that of Cdc27, a component of the anaphase-promoting complex known to be specifically phosphorylated during mitosis (Fig. 1A). Thus, the hyperphosphorylated BARD1 polypeptides of mitotic cells are either degraded or dephosphorylated at the M/G 1 transition.
To ascertain whether the hyperphosphorylated BARD1 species of mitotic cells retain the capacity to interact with BRCA1, synchronized HeLa cell lysates were immunoprecipitated with either the anti-BARD1 or anti-BRCA1 monoclonal antibody, and the immunoprecipitates were examined by immunoblotting with the BARD1-specific antiserum. As shown in Fig. 1C, equivalent proportions of hypo-and hyperphosphorylated BARD1 were recovered in the BARD1 (upper panel) and BRCA1 (lower panel) immunoprecipitates. Thus, mitotic phosphorylation of BARD1 does not appreciably alter its in vivo association with BRCA1.
Mapping BARD1 Phosphorylation Sites by Mutation Analysis-To identify the mitotic phosphorylation sites of BARD1, we initially tested whether individual fragments of the BARD1 polypeptide display an electrophoretic mobility shift when ex-pressed exogenously in nocodazole-arrested cells. Thus, HeLa cells were transfected with a series of expression vectors encoding different BARD1 fragments with a common N-terminal tag containing three tandem FLAG epitopes. Lysates were then prepared from 1) cells grown asynchronously and 2) cells arrested in prometaphase with nocodazole. Western analysis of the cell lysates with the anti-FLAG antibody revealed that fragments containing BARD1 amino acids 129 -228, 189 -287, and 280 -388 were visibly shifted in mitotic cells compared with asynchronously cycling cells (Fig. 1D, lanes 4, 6, and 8). The remaining BARD1 fragments (residues 1-120, 385-488, 464 -574, 563-681, and 661-777) did not display an obvious mobility shift in mitotic cells.
The consensus sequence for phosphorylation by the CDK1cyclin B complex, the major cyclin-dependent kinase of mitotic cells, is (S/T)PX 1-2 (R/K) (41). BARD1 fragment 280 -388 contains several sequences of this type, two of which are conserved in the mouse: Ser 288 and Thr 299 . As shown in Fig. 2A, the mitotic mobility shift of this fragment was abolished by mutation of Thr 299 to Ala (lane 6), but not mutation of Ser 288 to Ala (lane 4). Thus, Thr 299 is a potential mitotic phosphorylation site of BARD1.
Because there are no obvious phosphorylation motifs within BARD1 fragment 189 -287, we used site-directed mutagenesis In addition, a careful comparison of lanes 7 and 8 suggests that fragment 146 -240 still retains a modest mitotic shift even when it harbors both the S184A and S186A mutations. This shift was not observed, however, in a fragment bearing simultaneous mutations of Ser 148 , Ser 151 , Ser 184 , and Ser 186 (Fig. 2D, lanes 7-9). Thus, it is likely that Ser 148 and/or Ser 151 represents an additional site of BARD1 phosphorylation during mitosis. Phosphorylation at Ser 148 (but not Ser 151 ) was confirmed by mass spectrometric analysis of BARD1 polypeptides purified from mitotic cells (see below).
Although the T299A mutation abolished the mitotic shift of fragment 280 -388 ( Fig. 2A), the shift of BARD1 fragment 280 -435 was only partly reduced by this mutation (Fig. 2E,  lane 2). Thus, additional phosphorylation sites exist between residues 388 and 435 of BARD1. Because each of the five phosphorylation sites identified above resides within an SP or TP sequence, we evaluated the effect of mutating the three remaining SP/TP sequences (Ser 391 , Ser 394 , and Ser 410 ) in fragment 280 -435. Indeed, simultaneous alanine substitution of Thr 299 , Ser 391 , and Thr 394 abolished the mitotic shift of this fragment (lane 5). Because smaller mitotic shifts were also apparent in fragments bearing either the T299A and S394A mutations (lane 4) or the T299A and S391A mutations (lane 3), Thr 394 and Ser 391 both represent potential mitotic BARD1 phosphorylation sites.
Thus, mutation analysis identified seven potential phosphorylation sites responsible for the visible mitotic shift of exogenous BARD1 fragments during nocodazole block (Fig. 2). All seven sites are present in SP/TP motifs, and two of these, Ser 148 and Thr 299 , resemble the cyclin-dependent kinase consensus sequence. As shown in Fig. 3, a full- 4, 6, and 8), generated a detectable mitotic shift. Thus, phosphorylation of some (but not all) of the identified sites yields a visible mitotic shift in the context of full-length BARD1.
Mapping BARD1 Phosphorylation Sites by Mass Spectrometry-Our mutation analysis identified seven potential mitotic phosphorylation sites in BARD1. To confirm these results, we performed mass spectrometric analysis of BARD1 polypeptides purified from mitotic cells. To obtain a source of purified cellular BARD1 polypeptides, we stably transformed human lung carcinoma H1299 cells with a puromycin-resistant expression vector encoding full-length BARD1 with an N-terminal FLAG epitope (MDYKDDDDKGGS) and a C-terminal HA epitope (GGYPYDVPDYA). Cell extracts from the puromycin-resistant subclones were immunoprecipitated with the anti-FLAG antibody, and the presence of FLAG-BARD1-HA in each immunoprecipitate was then assessed by immunoblotting with the BARD1-specific antiserum. As shown in Fig. 4A (upper panel), FLAG-BARD1-HA polypeptides were readily detected in immunoprecipitates from H1299 subclones 2, 5, 7, and 8. When the extracts were examined by direct immunoblotting with the BARD1-specific antiserum, each of these four subclones displayed a closely migrating doublet that included both endogenous BARD1 and the exogenous FLAG-BARD1-HA polypeptide (lower panel). The expression levels of exogenous FLAG-BARD1-HA varied in the different subclones, ranging from ϳ2to 10-fold higher than that of endogenous BARD1.
For purification of FLAG-BARD1-HA, cultured cells of H1299 subclone 7 were arrested in mitosis with nocodazole. A lysate of these cells was immunoprecipitated with anti-FLAG antibody-conjugated agarose (Sigma), and the immunoprecipitated protein was eluted with the FLAG peptide. The eluate was denatured, alkylated with iodoacetamide, and separated by SDS-PAGE (Fig. 4B). The band corresponding to FLAG-BARD1-HA was then excised from the gel and digested with a panel of four different proteases, and the resulting peptides were subjected to MALDI-QqTOF tandem mass spectrometric analysis (38, 39,42). This analysis confirmed phosphorylation of BARD1 at each of the seven residues identified by mutation analysis (Table I, "Supplemental Data," and Supplemental Figs. 10 -14).
Cell Cycle Specificity of Thr 299 Phosphorylation-By combining mutation and mass spectrometric analyses, we identified seven phosphorylation sites in BARD1, all of which were specifically modified in cells treated with nocodazole, a microtubule-disrupting agent that can arrest cells in prometaphase by inducing the spindle assembly checkpoint. To confirm that phosphorylation of endogenous BARD1 at these residues occurs during normal cell cycle progression, phospho-specific antibodies that recognize the individual phosphorylation events will be necessary. Initially, we generated antibodies specific for Thr 299 -phosphorylated BARD1 by immunizing rabbits with a peptide encompassing this phosphothreonine residue. To confirm the phospho-specificity of the affinity-purified antiserum, exogenous FLAG-BARD1-HA polypeptides were immunoprecipitated from cell lysates of H1299 subclone 7 (Fig. 4A) with the anti-FLAG monoclonal antibody; the immunoprecipitates were fractionated by SDS-PAGE; and phosphorylated BARD1 was detected by immunoblotting with the phospho-Thr 299 -specific antiserum. As shown in Fig. 5A, Thr 299 -phosphorylated FLAG-BARD1-HA was observed in untreated immunoprecipitates (first lane), but not in immunoprecipitates treated with -phosphatase (second lane). To evaluate Thr 299 phosphorylation with respect to cell cycle progression, the FLAG-BARD1-HA-expressing H1299 cells were arrested in prometaphase with nocodazole, released into nocodazole-free medium, and collected at various time points. The BARD1 polypeptides from these cells were then visualized by Western analysis. As shown in Fig. 5B, the exogenous FLAG-BARD1-HA polypeptide was expressed at all time points after nocodazole release, consistent with the fact that its transcription is driven by the cytomega-lovirus promoter of the expression plasmid. In contrast, Thr 299phosphorylated BARD1, although prominent in nocodazolearrested cells, was absent in cells released from the nocodazole block for 9 -15 h. However, it reappeared in cells released for 17 h, a time when the synchronized cell population began to enter the subsequent M phase. Thus, Thr 299 phosphorylation occurs in a cell cycle-dependent manner. To ascertain whether Thr 299 phosphorylation of endogenous BARD1 polypeptides is also regulated with respect to the cell cycle, HeLa cells were arrested in prometaphase with nocodazole, released into nocodazole-free medium, and collected at various time points. As shown in Fig. 5C, BARD1 polypeptides were observed at time points corresponding to late G 1 phase (9 h) and S phase (13 and 16 h). However, Thr 299 -phosphorylated BARD1 did not reappear until later time points, when the synchronized cells had entered the subsequent M phase (20 and 23 h).
Generating Stable Cell Lines That Express Only Mutant Forms of BARD1-Functional analysis of BRCA1 phosphorylation has benefited from the use of HCC1937, a human cell line derived from a familial breast tumor that expresses truncated (but not full-length) BRCA1 (43). Indeed, specific functions have been ascribed to individual BRCA1 phosphorylation sites by comparing the phenotypes of HCC1937 derivatives reconstituted with either full-length wild-type BRCA1 or full-length mutant BRCA1 bearing defined phosphorylation site substitutions (31)(32)(33)(34). However, because mutant BARD1 cell lines are not currently available, this approach is not applicable to the analysis of BARD1 phosphorylation. Instead, we sought to generate cells expressing mutant BARD1 by applying a combined knock-out/knock-in strategy in HCT116, a line of human colorectal carcinoma cells that has been used for somatic cell knock-out of numerous genes (44), including p53, p21, SMAD4, securin, 14-3-3, and DNMT1 (45)(46)(47)(48)(49)(50). For this purpose, we constructed a hygromycin-based promoterless targeting vector to knock-out the BARD1 gene by homologous recombination (Fig. 6A). The coding sequences of human BARD1 are derived from 11 exons that are distributed over Ͼ65 kb of genomic DNA (40). As illustrated in Fig. 6A, homologous recombination with the knock-out vector would replace a 4.9-kb fragment of genomic BARD1 DNA encompassing coding exon 3 and part of coding exon 2 with a hygromycin cassette that is in-frame with the remaining exon 2 coding sequence. As a result, the "knockout" allele would encode a hybrid protein that includes the N-terminal 60 residues of BARD1 fused to the hygromycin gene product. Because most of the RING domain is encoded by the deleted sequences of exons 2 and 3, the knock-out allele cannot encode an enzymatically active BARD1 polypeptide. Upon transfection of HCT116 cells with this vector, we screened hygromycin-resistant clones for BARD1 targeting using a rapid PCR strategy (49), and properly targeted clones were confirmed by Southern analysis (data not shown). With this approach, we obtained five independent HCT116 clones that were successfully targeted on one of the two BARD1 alleles.
targeting the remaining wild-type allele of BARD1 ϩ/Ϫ cells with a neomycin-based derivative of the knock-out vector. However, these experiments repeatedly generated neomycin-resistant subclones in which the inactive (but not the wild-type) BARD1 allele was disrupted, suggesting that BARD1 expression is essential for the survival of HCT116 cells in culture. 2 This result is consistent with a previous study in which consecutive gene targeting failed to generate viable clones of Bard1-null mouse embryonic stem cells (9).
Having rendered one allele of BARD1 inactive, we used a "knock-in" strategy to replace the remaining germ line allele of BARD1 ϩ/Ϫ HCT116 cells with cDNA sequences encoding either wild-type or mutant BARD1. In this way, we sought to generate a series of isogenic HCT116 clones that express only either a wild-type or mutant form of BARD1. We initially replaced the functional BARD1 allele of BARD1 ϩ/Ϫ HCT116 cells with a cDNA sequence encoding wild-type BARD1. To do so, the knock-in targeting vector illustrated in Fig. 6B was constructed. Upon homologous recombination, this vector would replace a 1.3-kb fragment of genomic BARD1 DNA encompassing the 3Ј-region of coding exon 2 with a cassette that includes both the wild-type cDNA sequence for residues 61-777 of BARD1 and a neomycin coding sequence (Fig. 6B). The vector was designed such that the incorporated BARD1 cDNA sequence of this cassette is in-frame with the retained segment of BARD1 coding exon 2, thereby recreating a full-length BARD1 coding sequence whose expression is driven by the endogenous BARD1 promoter. Expression of the neomycin sequence is likewise driven by the BARD1 promoter, but requires an internal ribosome entry site for protein translation. BARD1 ϩ/Ϫ HCT116 cells were then transfected with this vector, and Ͼ1000 neomycin-resistant clones were screened in pools of six for BARD1 targeting events by long-range PCR. Oligonucleotide primers were designed such that PCR amplification would yield DNA fragments of different sizes from the wild-type (4.85 kb) and knock-in (3.8 kb) alleles, but would not amplify a fragment from the knock-out allele (which lacks a priming site for one of the two primers). As shown in Fig. 7A, PCR amplification of genomic DNA identified rare pools of neomycin-resistant clones that harbored the PCR fragment expected of a properly targeted knock-in BARD1 allele. The relevant clones from these pools were then identified, and their genotypes were confirmed by both PCR amplification (Fig. 7B) and Southern analysis (Fig. 7C). As illustrated in Fig. 7C, this experiment yielded seven subclones of BARD1 ϩ/Ϫ HCT116 cells in which the germ line BARD1 allele was successfully replaced with the wild-type knock-in BARD1 cassette. The genotype of these subclones was designated BARD1 wt/Ϫ (Fig. 6C).
To evaluate the function of the BARD1 phosphorylation sites, two mutant derivatives of the knock-in targeting vector were constructed. One derivative, BARD1 mutant A (mutA), harbored alanine substitutions at the two phosphorylation sites that resemble the cyclin-dependent kinase consensus sequence: Ser 148 and Thr 299 . The other derivative, BARD1 mutant B (mutB), carried alanine substitutions at the five remaining phosphorylation sites: Ser 184 , Ser 186 , Ser 251 , Ser 391 , and Thr 394 . With these targeting vectors, the same knock-in strategy was used to replace the remaining germ line allele of BARD1 ϩ/Ϫ HCT116 cells with cDNA sequences encoding mutant BARD1. In this manner, we generated five independent subclones of BARD1 ϩ/Ϫ HCT116 in which the germ line BARD1 allele was replaced with the knock-in BARD1 mutA cassette (Fig. 7C) and 13 in which it was replaced with the mutB cassette (Fig. 7D). The genotypes of these subclones were designated BARD1 mutA/Ϫ and BARD1 mutB/Ϫ , respectively (Fig. 6C). Close scrutiny of the Southern blots revealed that some subclones displayed unequal ratios of the knock-out and knock-in alleles (i.e. BARD1 mutA/Ϫ subclone 2 and BARD1 mutB/Ϫ subclones 9, 11, and 13), suggesting amplification of either the knock-out or knock-in chromosome; these subclones were not used in subsequent studies.
To confirm proper expression of the knock-in BARD1 alleles, extracts were prepared from the parental BARD1 ϩ/Ϫ HCT116 cells and from multiple independent subclones representing each of the three knock-in genotypes (BARD1 wt/Ϫ , BARD1 mutA/Ϫ , and BARD1 mutB/Ϫ ). The extracts were then im-2 H. Xu and R. Baer, unpublished data. were synchronized in prometaphase by a thymidine/nocodazole block (N). The mitotic cells were then collected by shake-off; and at time 0, the cells were replated at low density to induce cell cycle progression from prometaphase. Cells harvested at various times (0 -23 h) after induction were subjected to FACS analyses to determine the cell cycle distribution, and cell lysates were analyzed by immunoblotting with the anti-HA1 antibody (upper panel), the phospho-Thr 299 -specific antiserum (middle panel), and the anti-␣-tubulin antibody (lower panel). C, mitotic phosphorylation at Thr 299 on endogenous BARD1 polypeptides. HeLa cells were synchronized by a thymidine/nocodazole block. The mitotic cells were then collected and replated at low density to induce cell cycle progression from prometaphase. Cells harvested at various times (0 -23 h) after induction were subjected to FACS analyses to determine the cell cycle distribution, and cell lysates were analyzed either by direct immunoblotting with the anti-BARD1 antibody (upper panel) and the anti-␣-tubulin monoclonal antibody (lower panel) or by immunoprecipitation with the anti-BARD1 monoclonal antibody followed by immunoblotting of the resulting immunoprecipitates with the phospho-Thr 299 -specific antiserum (middle panel). munoprecipitated with the anti-BARD1 monoclonal antibody, and the immunoprecipitates analyzed by immunoblotting with either the BARD1-specific antiserum or the phospho-Thr 299specific antiserum. As shown in Fig. 8, the levels of steadystate BARD1 polypeptides in the knock-in subclones, including those encoding wild-type BARD1 (BARD1 wt/Ϫ ), varied over an ϳ4-fold range that corresponded to 20 -100% of the levels seen in the parental BARD1 ϩ/Ϫ cells. As expected, Thr 299phosphorylated species of BARD1 were readily detected in the BARD1 wt/Ϫ and BARD1 mutB/Ϫ subclones, but not the BARD1 mutA/Ϫ subclones. The levels of Thr 299 -phosphorylated BARD1 appeared to be reduced consistently in the BARD1 mutB/Ϫ subclones relative to the BARD1 wt/Ϫ subclones, suggesting that Thr 299 phosphorylation is partly enhanced by phosphorylation at one or more of the sites altered on the mutB allele (Ser 184 , Ser 186 , Ser 251 , Ser 391 , and Thr 394 ).
Cells Expressing BARD1 Phosphorylation Mutants Are Hypersensitive to Mitomycin C-To explore the functions of mi-FIG. 6. Knock-out/knock-in strategy for gene targeting in the BARD1 locus of HCT116 cells. A, knocking out the BARD1 locus. The map of the germ line BARD1 locus illustrates the first three coding exons of BARD1; the AUG initiation codon is also indicated. The knockout targeting vector is illustrated below the germ line BARD1 locus: the 5Ј-homology ("left") arm has 3.6 kb of genomic BARD1 DNA that includes the 5Ј-segment of coding exon 2, and the 3Ј-homology ("right") arm has 3.6 kb of genomic DNA. The two arms encompass a hygromycin cassette arranged such that the hygromycin coding sequence (hyg; denoted by a rectangle) is in-frame with the BARD1 coding sequences of exon 2. Another cassette containing the diphtheria toxin gene (DT; denoted by a rectangle) driven by the human ␤-actin gene promoter was included in the vector to select for recovery of homologous recombination events. Upon homologous recombination, a 4.9-kb fragment of genomic BARD1 DNA encompassing coding exon 3 and the 3Ј-segment of coding exon 2 is replaced with the hygromycin cassette. On the resultant knock-out allele, transcription of the hygromycin gene is driven by the BARD1 promoter and employs the herpes simplex virus thymidine kinase polyadenylation signal at the 3Ј-end of the hygromycin cassette. The positions of the PCR primers used to screen for gene targeting events are indicated with arrows. The triangles flanking the hygromycin coding sequence represent lox recombination signals not relevant to the current experiment. B, knocking in the BARD1 cDNA sequence. The knock-in targeting vector is illustrated below the germ line BARD1 locus: the left arm has 3.8 kb of genomic BARD1 DNA that includes the 5Ј-segment of coding exon 2, and the right arm has 3.5 kb of genomic DNA. The two arms encompass a cassette containing BARD1 cDNA sequences and a neomycin gene (each denoted by a rectangle). The cDNA sequences, which encode BARD1 residues 61-777, are arranged in-frame with the 5Ј-segment of coding exon 2 so as to recreate the full-length BARD1 coding sequence. Another cassette containing the diphtheria toxin gene (denoted by a rectangle) driven by the human ␤-actin gene promoter was included in the vector to select for recovery of homologous recombination events. Upon proper recombination, a 1.3-kb fragment of genomic BARD1 DNA containing the 3Ј-segment of coding exon 2 is replaced with the BARD1-neomycin cassette. On the resultant knock-in allele, transcription of the BARD1neomycin cassette is driven by the endogenous BARD1 promoter and employs the herpes simplex virus thymidine kinase polyadenylation signal. Translation of the neomycin promoter is facilitated by an inter- One of the six subclones shows the presence of the 3.8-kb PCR product representing the knock-in BARD1 allele and the absence of the 4.85-kb PCR product representing the germ line BARD1 allele. C and D, Southern analysis of genomic DNA from HCT116 cells (ϩ/ϩ genotype), BARD1 ϩ/Ϫ HCT116 cells, seven knock-in BARD1 wt/Ϫ subclones, five knock-in BARD1 mutA/Ϫ subclones, and 13 knock-in BARD1 mutB/Ϫ subclones. The genomic DNA was digested with BglII, fractionated by agarose gel electrophoresis, and analyzed by Southern hybridization with an upstream genomic BARD1 DNA probe.
nal ribosome entry site. The positions of the PCR primers used to screen for gene targeting events are indicated with arrows. The triangles flanking the internal ribosome entry site (IRES)-neomycin sequence are lox recombination signals not relevant to the current experiment. C, structures of the BARD1 alleles represented in the four HCT116 genotypes. All four genotypes contain one knock-out BARD1 allele. BARD1 ϩ/Ϫ cells also contain a germ line BARD1 allele. In BARD1 wt/Ϫ cells, the remaining germ line BARD1 allele is replaced with a knock-in allele encoding wild-type BARD1. In BARD1 mutA/Ϫ and BARD1 mutB/Ϫ cells, the germ line BARD1 allele is replaced with a knock-in allele encoding BARD1 mutA (S148A and T299A) or BARD1 mutB (S184A, S186A, S251A, S391A, and T394A), respectively. totic BARD1 phosphorylation, we compared the behavior of isogenic HCT116 subclones representing the three relevant genotypes: BARD1 wt/Ϫ , BARD1 mutA/Ϫ , and BARD1 mutB/Ϫ . Cell cycle progression into and out of mitosis appeared to be unaffected by the phosphorylation site mutants, as we did not uncover reproducible differences in the mitotic indices of asynchronous HCT116 cultures (Supplemental Fig. 15A) or in the kinetics of M/G 1 progression after release from nocodazole arrest (Supplemental Fig. 15B).
Expression and phosphorylation of BRCA1 are required for activation of the IR-induced transient G 2 /M checkpoint (32,33). To evaluate the role of BARD1 phosphorylation in this checkpoint, the mitotic indices of HCT116 subclones were measured by flow cytometric analysis of histone H3 phosphorylation at a short interval (90 min) after IR treatment. As expected, the proportion of mitotic BARD1 ϩ/Ϫ cells was greatly reduced by IR exposure (Supplemental Fig. 16). However, similar reductions were also obtained upon IR treatment of the BARD1 mutA/Ϫ and BARD1 mutB/Ϫ subclones. Thus, BARD1 phosphorylation does not appear to influence activation of the IR-induced transient G 2 /M checkpoint.
Defects in the DNA damage response are often manifested by cellular hypersensitivity to genotoxic stress. Because cells with hypomorphic lesions of BRCA1 are especially sensitive to DNA cross-linking agents (51,52), we examined the effect of mitomycin C exposure on survival of the isogenic HCT116 subclones. Clonogenic survival assays using increasing doses of mitomycin C were performed on four subclones representing each of the three HCT116 genotypes (BARD1 wt/Ϫ , BARD1 mutA/Ϫ , and BARD1 mutB/Ϫ ). As shown in Fig. 9, the sensitivity of the four BARD1 wt/Ϫ subclones (red lines) did not differ significantly from that of the parental BARD1 ϩ/Ϫ cells (black lines). In contrast, the BARD1 mutA/Ϫ (green lines) and BARD1 mutB/Ϫ (blue lines) subclones showed hypersensitivity at all doses of mitomycin C relative to both the BARD1 wt/Ϫ subclones and parental BARD1 ϩ/Ϫ cells. In some experiments, the increased sensitivity reached statistical significance for all mutant subclones in the low dose range (e.g. 50 -150 ng/ml) (Fig.  9A), whereas in other experiments, it was achieved at higher doses (e.g. 150 -200 ng/ml) (Fig. 9B). Nonetheless, the reproducible hypersensitivity of the mutant subclones suggests a role for BARD1 phosphorylation in the cellular response to DNA damage. In particular, the behavior of the BARD1 mutA/Ϫ subclones implicates phosphorylation at Ser 148 and/or Thr 299 in the cellular response to mitomycin C, whereas the behavior of the BARD1 mutB/Ϫ subclones indicates a role for one or more of the other sites (Ser 184 , Ser 186 , Ser 251 , Ser 391 , and Thr 394 ).

DISCUSSION
Phosphorylation of the BRCA1 tumor suppressor occurs with cell cycle progression and during the cellular response to DNA damage (5,(22)(23)(24)(25)(26)(27)(28)(29)(30), and some of these phosphorylation events have been shown to impact specific BRCA1 functions (31)(32)(33)(34). Because the downstream effects of BRCA1 are mediated primarily through the BRCA1/BARD1 complex, post-translational modifications of the BARD1 polypeptide are also likely to influence BRCA1-mediated tumor suppression. Here, we have shown that BARD1 is phosphorylated in a cell cycle-dependent manner and that hyperphosphorylated BARD1 polypeptides accumulate primarily during mitosis. This differs from the major pattern observed with BRCA1, in which hyperphosphorylated BRCA1 species become prominent at the G 1 /S transition and remain so throughout the subsequent G 2 and M phases. Nonetheless, a recent study has shown that at least one site within BRCA1 is phosphorylated at the G 2 /M transition (53). This event, which is mediated by the mitotic kinase Aurora A (53), should be contemporaneous with the cell cycle-dependent phosphorylation of BARD1 reported here. The fact that both BRCA1 and BARD1 are subjected to specific posttranslational modifications during M phase is consistent with growing evidence that BRCA1 has important functions during mitosis (36,(53)(54)(55)(56)(57)(58)(59)(60) and that disruption of these functions may be partly responsible for the genomic instability observed in Brca1-deficient cells (61,62).
In this study, we identified seven residues of BARD1 that are phosphorylated specifically in cells treated with nocodazole, a microtubule poison that induces prometaphase arrest by activating the spindle assembly checkpoint. As such, in unstressed cells, these sites are likely to be phosphorylated prior to or during mitosis. It is possible, however, that some of these modifications do not occur in unperturbed cells, but instead represent specific responses to nocodazole treatment. Using a phospho-specific antiserum, we have shown that mitotic phosphorylation of BARD1 does indeed occur at Thr 299 independently of nocodazole treatment. Formal proof that the other six sites are also modified in unperturbed cells will require production of the cognate phospho-specific antibody reagents. Interestingly, all seven BARD1 phosphorylation sites occur within either an SP or TP sequence, suggesting the involvement of proline-directed kinases. Two sites (Ser 148 and Thr 299 ) resemble the consensus sequence ((S/T)PX 1-2 (R/K)) recognized by cyclin-dependent kinases (41) and as such could be substrates for the mitotic CDK1-cyclin B1 complex (60). In contrast, none of the other sites bears a clear resemblance to current consensus motifs for the two other major mitotic kinases, PLK1 and Aurora A (63,64).
In a recent study, Hayami et al. (60) also observed mitotic phosphorylation of BARD1 and showed that an N-terminal fragment of BARD1 (residues 1-320) is an efficient substrate for in vitro phosphorylation with either CDK2-cyclin E1 or CDK1-cyclin A1. They also demonstrated that this fragment (but not a corresponding fragment bearing simultaneous alanine substitutions at Ser 148 , Ser 251 , Ser 288 , and Thr 299 ) is hyperphosphorylated upon cotransfection with either CDK2-cyclin E1/A1 or CDK1-cyclin B1. Although we did not obtain evidence for Ser 288 phosphorylation in our study, we did observe phosphorylation at Ser 148 , Ser 251 , and Thr 299 by both mutation analysis and mass spectrometry. Remarkably, Hayami et al. (60) found that cotransfection with CDK2-cyclin E1/A1 (but not CDK1-cyclin B1) reduced the enzymatic activity of the BRCA1/BARD1 complex by inducing nuclear export of BRCA1. However, this behavior could not be attributed to BARD1 phosphorylation because BARD1 polypeptides harboring substitutions at the four sites (Ser 148 , Ser 251 , Ser 288 , and FIG. 8. Expression of BARD1 polypeptides in knock-in HCT116 subclones. Whole cell extracts from the parental BARD1 ϩ/Ϫ HCT116 cells and from multiple independent subclones representing each of the three knock-in genotypes (BARD1 wt/Ϫ , BARD1 mutA/Ϫ , and BARD1 mutB/Ϫ ) were immunoprecipitated (IP) with the anti-BARD1 monoclonal antibody. The immunoprecipitates were then analyzed by Western blotting (WB) with either the anti-BARD1 polyclonal antibody (upper panel) or the phospho-Thr 299 -specific antiserum (␣ pT299; middle panel). The extracts were also evaluated by direct immunoblotting with the anti-␣-tubulin antibody (lower panel).
To evaluate the functional consequences of BARD1 phosphorylation, we devised a knock-in strategy to generate isogenic cell lines that express only wild-type BARD1 or defined BARD1 mutants. This entailed knocking out one allele of BARD1 in HCT116 colorectal carcinoma cells and replacing the remaining allele with a knock-in BARD1 cDNA construct. In this manner, we generated knock-in HCT116 subclones that encode 1) wildtype BARD1, 2) BARD1 mutA (harboring the S148A and T299A mutations), or BARD1 mutB (harboring the S184A, S186A, S251A, S391A, and T394A mutations). In contrast, we were unable to derive BARD1-null subclones of HCT116 by knocking out both BARD1 alleles in succession. This is consistent with a previous futile attempt to generate Bard1-null embryonic stem cells (9) and suggests that BARD1 function is essential for the viability of both embryonic stem cells and HCT116 cells. However, the fact that we successfully obtained subclones expressing the mutA and mutB forms of BARD1 indicates that these are hypomorphic mutations and that BARD1 phosphorylation, at least at the seven sites ablated in these mutants, is not required for cell viability.
Although the BRCA1 tumor suppressor is involved in a diverse spectrum of cellular processes, it plays an especially important role in the cellular response to genotoxic stress (65)(66)(67). In particular, BRCA1 is required for several of the cell cycle checkpoints activated by DNA damage and has also been implicated in certain repair pathways, such as homology-directed repair of double-stranded DNA breaks (66). As such, cells bearing hypomorphic lesions of BRCA1 show enhanced sensitivity to DNA-damaging agents, especially those that induce interstrand DNA cross-links, such as mitomycin C (51,52). Recent studies have shown that inactivation of BARD1 also impairs homology-directed repair and promotes chromosome instability (9,21). In this study, we observed that cells expressing BARD1 mutA (which lacks the Ser 148 and Thr 299 phosphorylation sites) or BARD1 mutB (which lacks the Ser 184 , Ser 186 , Ser 251 , Ser 391 , and Thr 394 sites) are hypersensitive to the DNA-cross-linking reagent mitomycin C. Although the precise molecular mechanism by which these cells are rendered hypersensitive is not known, these data implicate BARD1 phosphorylation in the cellular response to genotoxic stress. FIG. 9. Hypersensitivity to mitomycin C of the mutant BARD1 subclones. The parental BARD1 ϩ/Ϫ HCT116 cells and cells from four independent subclones representing each of the three knock-in genotypes (BARD1 wt/Ϫ , BARD1 mutA/Ϫ , and BARD1 mutB/Ϫ ) were examined for mitomycin C sensitivity in clonogenic survival assays. The cells were seeded on 100-mm plates so as to yield ϳ500 colonies in the untreated plates. The next day, the cells were treated with 0, 50, 100, 150, or 200 ng/ml mitomycin C (MMC) for 4 h. After 10 days, the surviving colonies were stained with Giemsa and counted. Survival was calculated as a percentage of the colonies on the mocktreated plates. Each subclone was tested in triplicate, and the error bars represent S.D. of survival for each subclone.