Nuclear targeting and cell cycle regulatory function of human BARD1.

The BARD1 gene is mutated in a subset of breast and ovarian cancers, implicating BARD1 as a potential tumor suppressor. BARD1 gains a ubiquitin E3 ligase activity when heterodimerized with BRCA1, but the only known BRCA1-independent BARD1 function is a p53-dependent proapoptotic activity stimulated by nuclear export to the cytoplasm. We described previously the nuclear-cytoplasmic shuttling of BARD1, and in this study, we identify the transport sequences that target BARD1 to the nucleus and show that they are essential for BARD1 regulation of the cell cycle. We used deletion mapping and mutagenesis to define two active nuclear localization signals (NLSs) present in human BARD1 that are not conserved in rodent BARD1. Site-directed mutagenesis of the primary bipartite NLS abolished BARD1 nuclear import and caused its cytoplasmic accumulation. Using flow cytometry and 5-bromo-2-deoxyuridine incorporation assays, we discovered that transiently expressed BARD1 can elicit a p53-independent cell cycle arrest in G1 phase, and that this was abrogated by mutation of the BARD1 NLS but not by mutation of the nuclear export signal. Thus, BARD1 regulation of the cell cycle is a nuclear event and may be linked to its induced expression during mitosis and its possible involvement in the DNA damage checkpoint.

The BARD1 gene is mutated in a subset of breast and ovarian cancers, implicating BARD1 as a potential tumor suppressor. BARD1 gains a ubiquitin E3 ligase activity when heterodimerized with BRCA1, but the only known BRCA1-independent BARD1 function is a p53-dependent proapoptotic activity stimulated by nuclear export to the cytoplasm. We described previously the nuclear-cytoplasmic shuttling of BARD1, and in this study, we identify the transport sequences that target BARD1 to the nucleus and show that they are essential for BARD1 regulation of the cell cycle. We used deletion mapping and mutagenesis to define two active nuclear localization signals (NLSs) present in human BARD1 that are not conserved in rodent BARD1. Site-directed mutagenesis of the primary bipartite NLS abolished BARD1 nuclear import and caused its cytoplasmic accumulation. Using flow cytometry and 5-bromo-2-deoxyuridine incorporation assays, we discovered that transiently expressed BARD1 can elicit a p53-independent cell cycle arrest in G 1 phase, and that this was abrogated by mutation of the BARD1 NLS but not by mutation of the nuclear export signal. Thus, BARD1 regulation of the cell cycle is a nuclear event and may be linked to its induced expression during mitosis and its possible involvement in the DNA damage checkpoint.
BARD1 is the major cellular binding partner of the breast and ovarian cancer susceptibility protein, BRCA1 (1). BARD1 itself was reported to be mutated in a subset of breast and ovarian cancer patients, and BARD1 germ line mutations were identified in breast cancer patients negative for BRCA1 or BRCA2 gene alterations, suggesting a possible tumor-suppressive function of BARD1 (2)(3)(4). This hypothesis is supported by findings that the repression of BARD1 expression in murine mammary epithelial cells by antisense RNA resulted in a premalignant phenotype (5) and the observation that BARD1 induced apoptosis via a pathway depending on p53 but not requiring BRCA1 (6).
BARD1 plays a key role in regulating BRCA1 stability, localization and function, and several lines of evidence indicate that the BARD1-BRCA1 heterodimer is the physiologically relevant form of BRCA1. The two proteins are coordinately expressed in a variety of different tissues and cell types during Xenopus laevis development, and they stabilize one another (7)(8)(9). Furthermore, the phenotypes of either BARD1 or BRCA1 null mice, as well as BARD1:BRCA1 double null mice, display striking similarities in their accumulation of chromosomal abnormalities and early embryonic death due to severe cell proliferation defects (9,10). BARD1 and BRCA1 are predominantly found in the cell nucleus, where they co-localize in S phase-specific dots (11). Following DNA damage, they redistribute within the nucleus into discrete foci that also co-stain with DNA repair-associated factors such as BRCA2 and Rad51 (12). BARD1 and BRCA1 also co-fractionate with several DNA repair-associated nuclear protein complexes, suggesting a role for the two proteins in DNA repair and replication (13). BRCA1 possesses E3 ubiquitin ligase activity, which is ablated by tumor-derived point mutations within the RING motif (14,15) but greatly enhanced by complex formation with BARD1 (8, 16 -18). Recently, it was shown for the first time in vivo that BARD1-BRCA1 can induce the formation of and colocalize with ubiquitin conjugates during replication and DNA repair (19).
Both BARD1 and BRCA1 contain nuclear export sequences (NES), 1 which enable the proteins to shuttle between nucleus and cytoplasm (20,21). We showed recently that BARD1 nuclear export stimulates its apoptotic activity (21), and a comparable finding was made by Jefford et al. (22), who observed a correlation between cytoplasmic staining of BARD1 and apoptosis. The co-expression of BRCA1 inhibited both BARD1 nuclear export and apoptotic activity (21). Whereas two nuclear localization signals (NLSs) have been identified in BRCA1 (23,24), no NLS has yet been defined for BARD1. In this study, we mapped three active nuclear localization signals in BARD1, of which two are not conserved in the rodent BARD1 protein. Furthermore, we show that BARD1 itself can elicit a G 1 phase cell cycle arrest when overexpressed and that this activity was dependent on its NLS-mediated nuclear localization. This discovery is reminiscent of previous findings for BRCA1 (25) and reveals that the two proteins not only bind and regulate one another but perhaps can substitute for each other in several functions, including cell cycle arrest (this study), apoptosis (6,21), genomic stability (9), and homologous DNA repair/recombination (26). Collectively, these findings support a role for BARD1 in tumor suppression.

MATERIALS AND METHODS
Cell Culture and Transfections-MCF-7 breast cancer cells and U2OS and Saos-2 osteosarcoma cells were grown under standard conditions in Dulbecco's modified Eagle's medium supplemented with 10% * This work was supported by a grant from the National Health and Medical Research Council (NHMRC) of Australia. 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  fetal calf serum. Transient transfections were done using Lipo-fectamine2000 (Invitrogen) according to the manufacturer's instructions. Briefly, 24 h after seeding, the cells were transfected at 50% confluence with either 2 g (6-well plate for immunofluorescence) or 4 g (T25 flask for Western blots) of DNA. 16 h post-transfection, the transfection mixture was replaced with Dulbecco's modified Eagle's medium containing 10% fetal calf serum, and the cells were processed 32-48 h post-transfection for immunofluorescence microscopy or immunoblot analysis.
Construction of Mammalian Expression Plasmids-Full-length BARD1 cDNAs coding for wild-type (Lys 153 ) or polymorphic (Glu 153 ) forms were kind gifts of Prof. Richard Baer (Columbia University). The FLAG-tagged as well as the yellow fluorescence protein (YFP)-tagged expression plasmids for BARD1-WT, -L107A, -(95-777), -(171-777), and -(1-188) have been described elsewhere (21). FLAG-BARD1(Lys 153 ) was constructed in the same way as described for FLAG-BARD1-WT (21). BARD1-(1-202) was a kind gift of Dr. Mary Moynahan (26). All of the other BARD1 truncation and deletion constructs were prepared using a PCR-based approach (details available upon request) and inserting the corresponding fragments into the pFLAG-CMV-2 vector (Sigma). YFP fusion constructs were obtained by amplifying the YFP gene from the pEYFP-C1 vector (Invitrogen) by PCR and inserting into the NotI site of the respective pFLAG-BARD1 plasmids. YFP-BRCA1 and YFP-BRCA1(⌬306 -1312) have been described (27). The pCMV-␤Gal-NLS constructs were obtained by inserting double-stranded oligonucleotides as NotI/HindIII fragments into the pCMV-␤Gal-FusB vector (a kind gift of Dr. Z. Ivics, Berlin, Germany). The BARD1 NLS2 point mutants were constructed by a PCR-based mutagenesis approach using standard methodology (details available upon request).
Immunofluorescence Microscopy and Imaging-Fixation and immunostaining of cells was performed as described (27). YFP-tagged proteins were detected directly following fixation, washing, and mounting. Ectopic, FLAG-tagged BARD1 was detected using the anti-FLAG mouse monoclonal antibody, M2 (Sigma), diluted 1:1200. ␤-Galactosidase was detected with a specific monoclonal antibody (Sigma; 1:1000). Bound antibodies were detected with either AlexaFluor 488-conjugated (1:1000; Molecular Probes, Inc., Eugene, OR) or biotin-conjugated (1: 500; DAKO) secondary antibodies. In the latter case, this was followed by incubation with avidin D-Texas Red (Vector Laboratories), diluted 1:600. Nuclei were counterstained with the DNA dye Hoechst 33285 (Sigma). The subcellular localization of each protein was determined by scoring cells with an Olympus BX40 epifluorescence microscope. Images were captured with a SPOT digital camera, and quantification of fluorescence was carried out using Image Pro software.
Whole Cell Extracts, Cell Fractionation, and Immunoblots-Cells were resuspended in protein extraction buffer (20 mM Tris⅐HCl, pH 8.0, 100 mM NaCl, 1 mM EDTA, 0.5% Nonidet P-40, supplemented with protease inhibitor mix (Roche Applied Science)) and shock-frozen in liquid nitrogen. After a quick thaw at 37°C, they were refrozen in liquid nitrogen, rethawed on ice for 20 min, and cleared of insoluble components by centrifugation at 13,000 rpm at 4°C for 15 min. Cell extracts were then mixed (2:1) with sample buffer (100 mM Tris⅐HCl, pH 6.8, 20% glycerol, 0.01% bromophenol blue, 10% ␤-mercaptoethanol, 5% SDS) and denatured at 95°C for 5 min. Cell separation into nuclear and cytoplasmic fractions was done with the NE-PER kit (Pierce) according to the manufacturer's instructions, and the proteins were denatured in sample buffer as described above. 30 g of whole cell protein or equal cellular amounts of fractionated protein (ϳ10 g of nuclear extract and ϳ30 g of cytoplasmic extract), respectively, were loaded per lane, separated on a 7.5% SDS-polyacrylamide gel, and transferred overnight at 4°C onto nitrocellulose membrane (Millipore Corp.). The membranes were incubated in blocking solution (5% skim milk powder in PBST (PBS containing 0.2% Tween 20)) for 60 min at room temperature followed by incubation with primary antibody diluted in PBST containing 2% skim milk powder for 2 h at room temperature. Incubation with secondary horseradish peroxidase-conjugated antibodies (1:10,000; Sigma) was followed by detection by ECL (Amersham Biosciences) reaction. ␤-Galactosidase was detected with a specific monoclonal antibody (1:1000; Sigma), BARD1 by the 59L antibody (a kind gift from Prof. Richard Baer), and topoisomerase II using a monoclonal antibody (Santa Cruz Biotechnology, Inc., Santa Cruz, CA).
Cell Cycle Analysis by Flow Cytometry-At 48 h post-transfection, cells were detached with trypsin and resuspended in complete medium. Cells were pelleted by centrifuging at 1500 rpm for 5 min and then washed in 1 ml of PBS and centrifuged again at 1500 rpm for 5 min. Cells were resuspended in 400 l of PBS and fixed in 4.5 ml of ice-cold 85% ethanol at 4°C for at least 1 h. Cells were pelleted by centrifuging at 1500 rpm for 5 min and resuspended in 600 l of PBS containing RNase A (1 mg/ml) and propidium iodide (2 mg/ml). YFP-expressing cells were gated, and the cell cycle profiles were determined using a BD Biosciences flow cytometer. Cell cycle profiles were analyzed using the ModFIT software.
Cell Synchronization and Analysis-U2OS cells at 50% confluence were treated with nocodazole (80 ng/ml) for 18 h. Mitotic cells were harvested by shake-off and released into nocodazole-free medium after washing three times with complete Dulbecco's modified Eagle's medium. At each time point, the cells were harvested by trypsinization. One half of the cell sample was then processed for flow cytometry as described above, and fractionated cell extracts were prepared from the other half as described above and analyzed by immunoblotting.
5-Bromo-2-deoxyuridine (BrdUrd) Incorporation Assay-At 48 h post-transfection, cells expressing YFP-tagged proteins were treated with 20 M BrdUrd for 1 h, fixed in 3.7% formalin/PBS for 15 min, washed three times with PBS, and treated with 50 mM hydrochloric acid for 20 min. Cells were washed with ice-cold 0.009% NaCl, 1% Tween 20 until the pH reached 7. Incorporated BrdUrd was detected using an anti-BrdUrd monoclonal antibody (Amersham Biosciences), a biotinconjugated secondary antibody (Santa Cruz Biotechnology), and Texas Red-avidin D (Vector Laboratories). At least 100 transfected cells were scored for BrdUrd incorporation in each experiment.

Human BARD1 Contains a Conserved Nuclear Localization Signal Whose Activity Is Impaired by a Naturally Occurring
Polymorphism-BARD1 has previously been implicated as a nuclear chaperone for the breast cancer gene product, BRCA1 (27), and is known to heterodimerize with BRCA1 in the nucleus to function in DNA repair (1,13,21). In order to identify the BARD1 nuclear localization signal(s), we first used a computer analysis program (available on the World Wide Web at psort.nibb.ac.jp/) to predict the probability of NLSs in human BARD1 and identified six putative signals (Fig. 1A). A comparison revealed that one of the predicted sequences, NLS2, was completely conserved among four different species, whereas human NLS6 exhibited high homology with mouse and rat but not Xenopus BARD1 sequences (Fig. 1B). The other four candidate NLSs showed very little evolutionary conservation (Fig.  1B). NLS2 is classified as a bipartite NLS characterized by the presence of two pairs of basic amino acid residues (lysine or arginine) separated by a spacer of 11 amino acids (28). Interestingly, a polymorphic form of BARD1 has been described that codes for glutamic acid instead of lysine at position 153 (K153E) (2). Lysine 153 is part of the conserved NLS2 sequence, and its mutation might therefore affect BARD1 subcellular targeting.
To address this hypothesis, we tested whether NLS2 indeed represented a functional nuclear import sequence by inserting it into a ␤-galactosidase expression vector and examining the subcellular localization of the ␤-galactosidase-NLS fusion protein in transfected MCF-7 cells and U2OS cells (Fig. 1C). Insertion of NLS2 induced a 2-fold increase in nuclear localization of the ␤-galactosidase reporter in MCF-7 cells and a 4-fold increase in nuclear staining in U2OS cells (Fig. 1C). This increase in nuclear import was abrogated when lysine 153 was substituted by glutamic acid. To determine the impact of this sequence on nuclear import of full-length BARD1, we compared the localization of the polymorphic K153E mutant with wild-type BARD1 (Fig. 1D). Interestingly, despite its clear effect on the NLS2 sequence in a reporter system, the mutation of Lys 153 had no significant influence on the subcellular distribution of full-length BARD1 in either MCF-7 or U2OS cells. This suggests that, despite its evolutionary conservation, NLS2 is not required for the nuclear import of BARD1. Since the NLS6 sequence, which resembles the SV40 T antigen-like signal, is also highly conserved in mammalian species, we tested its nuclear import activity in the ␤-galactosidase system. However, NLS6 did not exhibit any nuclear import activity (data not shown).

Deletion Mapping of Two Additional Regions Essential for
Nuclear Import of BARD1-In order to systematically define the region(s) of BARD1 necessary for nuclear import, we next performed a deletion/truncation analysis (Fig. 2). To avoid any conflicting influence of NLS2, we performed the subsequent mapping experiments using the polymorphic K153E form of BARD1. A series of BARD1 protein fragments were tested, and all sequences were fused to the YFP marker to reduce passive diffusion through the nuclear pore due to the small size of some BARD1 fragments and to facilitate unambiguous detection of the ectopic protein. The BARD1-YFP constructs were transfected into MCF-7 cells, and after 48 h cells were fixed and examined by fluorescence microscopy. The subcellular distribution of the different BARD1 peptides (in particular a comparison of aa 299 -593 and aa 355-525) suggested that a primary NLS was located between amino acids 299 and 355 (indicated by a dark gray box; see Fig. 2, A and B). In addition, careful comparison of the data further indicated a second minor nuclear localization activity immediately downstream between amino acids 355 and 425. To assess the relevance of these sequences, we prepared constructs in which amino acids 292-338 or 342-379 were deleted within full-length BARD1 (K153E mutant). The deletion of aa 292-338 resulted in an almost complete loss of BARD1 from the nucleus (Fig. 3). On the other hand, deletion of aa 342-379 elicited a consistent but more moderate impairment of nuclear import (Fig. 3).
Identification of Two Functional Nuclear Localization Signals That Are Specific to Human BARD1 and Are Not Well Conserved in Rodents-Interestingly, the predicted NLS3 and NLS4 sequences described in Fig. 1A are contained within the BARD1 regions 292-338 and 342-379, respectively. Therefore, we examined their nuclear import potential by inserting them into the ␤-galactosidase expression vector (Fig. 4). Both sequences efficiently supported nuclear import of the heterologous protein and were observed to be much more active than the conserved NLS2 sequence when tested in two different cell lines. NLS3 displayed a particularly strong activity, and its presence resulted in the complete accumulation of ␤-galactosidase in the nucleus, strongly suggesting that NLS3 is the major nuclear localization signal in BARD1. Although it is not well conserved between human and rodent species, particularly at the important basic arginine and lysine residues, the basic clusters are conserved in Xenopus BARD1, and NLS3 shows high homology to other well known NLSs such as those found in human retinoblastoma protein or estrogen receptor (28).
Site-directed Mutagenesis of the Bipartite NLS3 Sequence Prevents BARD1 Nuclear Localization-In order to characterize NLS3 in more detail, we introduced two sets of double point mutations into the ␤-galactosidase-NLS3 construct and examined their effect on the import activity by immunofluorescence microscopy in MCF-7 cells (Fig. 5, top). Substitution of the two pairs of lysine/arginine residues, by either aspartic acid/glutamic acid or by aspartic acid/aspartic acid, resulted in the abolition of nuclear import activity (see cell images and quantification of nuclear/cytoplasmic fluorescence in Fig. 5). We confirmed the results of the cell staining experiments by Western blot analysis of fractionated cell extracts from U2OS cells, where it was clear that NLS3 functions as a potent NLS and that its activity was inhibited by mutation of its basic residues (Fig. 5, bottom).
In order to confirm that NLS3 represents the dominant nuclear import sequence in BARD1, we introduced by site-directed mutagenesis the double amino acid substitutions described above and determined their effect on the subcellular localization of full-length BARD1(K153E) (Fig. 6). Scoring by immunofluorescence microscopy of transiently transfected MCF-7 cells revealed that both mutants strongly accumulated in the cytoplasm. However, whereas mutant 1 showed some residual nuclear localization, the mutation of the second lysine/arginine pair (K334N,R335N; NLS3mutant2) completely abolished nuclear import. A similar redistribution was observed in U2OS cells by cell staining (data not shown) and confirmed by Western blot analysis of fractionated extracts from transfected U2OS cells (Fig. 6, bottom). When considered together, these results identify NLS3 as the critical nuclear import signal of BARD1.
BARD1 Distribution Is Predominantly Nuclear throughout the Cell Cycle, and Its Expression Increases at the G 2 /M Transition-We next examined endogenous BARD1 subcellular localization and expression in U2OS cells after arresting them at G 2 /M phase following 18 h of nocodazole treatment. Cells were released in normal medium, and parallel cultures were fractionated and tested by Western blot, or their cell cycle distri-

FIG. 1. A conserved N-terminal NLS is not required for BARD1 nuclear localization.
A, putative NLSs within human BARD1. B, schematic representation of BARD1 and protein domains. The predicted NLSs are indicated by asterisks and numbered from 1 to 6. A sequence comparison between species of the six NLSs is shown below. C, MCF-7 and U2OS cells were transfected with ␤-galactosidase reporter constructs, either with no NLS (␤-galactosidase (␤-Gal) control) or containing the BARD1 wild-type NLS2 sequence (Lys 153 ( 153 K)) or BARD1 NLS2 with a polymorphism (Glu 153 ( 153 E)). At 48 h post-transfection, cells were stained with anti-␤-galactosidase antibody and analyzed by fluorescence microscopy. The ratio of nuclear to cytoplasmic fluorescence intensity (N/C) was determined, and the results are shown as mean Ϯ S.E., for a minimum of 50 cells from two independent experiments. D, the intracellular distribution of FLAG-tagged, transiently transfected BARD1 in MCF-7 cells or U2OS cells was scored by microscopy, and the data are graphed as mean Ϯ S.D. n, number of transfected cells scored. N, nuclear; NC, nuclear/cytoplasmic; C, cytoplasmic.
bution was determined by fluorescence-activated cell sorting analysis. As shown in Fig. 7, BARD1 protein levels decreased within 4 h after release from G 2 /M phase arrest and then progressively increased as the G 2 /M content increased 28 h later. BARD1 levels were lowest during G 1 phase. These findings are consistent with data recently published for a different cell line by Choudhury et al. (29) and confirm a link between BARD1 expression and mitosis. Our fractionation revealed that BARD1 was predominantly nuclear throughout the cell cycle (Fig. 7), consistent with the action of its multiple nuclear localization signals and indicating that nuclear export does not dominate at any specific stage of the cell cycle.  3. Deletion of two centrally located sequences reduces BARD1 nuclear localization. MCF-7 cells were transfected with expression vectors encoding wild-type BARD1 or two mutants in which specific sequences predicted to stimulate nuclear targeting (see Fig. 2) were deleted. After transient expression, the cells were stained with FLAG antibody, and the tagged BARD1 proteins were visualized and scored using fluorescence microscopy (see cell images). Quantification of the distribution of ectopic BARD1 (see graphs) indicated that deletion of the sequences aa 292-338 and aa 342-379 reduced nuclear localization. The most potent effect was observed after loss of aa 292-338. N, nuclear. C, cytoplasmic.

FIG. 4. Identification of two novel nuclear localization signals in human BARD1.
The ␤-galactosidase reporter system was used to assess the nuclear localization activity of two potential NLS sequences (named NLS3 and NLS4 in Fig. 1). The plasmids were transiently expressed in MCF-7 and U2OS cells and detected following staining with anti-␤-galactosidase antibody. Typical cell images are shown, and quantification of actual nuclear and cytoplasmic fluorescence in at least 60 cells resulted in assignment of nuclear/cytoplasmic (N/C) ratios for each construct, as shown (mean Ϯ S.E. from two experiments). is its proapoptotic activity (6,21,22). This was reduced by mutation of the BARD1 nuclear export signal (21) but not affected by NLS mutations (data not shown), implicating a role for cytoplasmic BARD1 in apoptosis. Since BARD1 levels changed during the cell cycle (Fig. 7), we tested the effect of BARD1 overexpression on cell cycle progression. We performed cell cycle analysis of YFP-BARD1-transfected U2OS cells and discovered that the transient overexpression of BARD1 caused a G 1 cell cycle arrest (Fig. 8A, top). When compared with cells transfected with YFP alone, YFP-BARD1 caused a 45% reduction in cells entering S phase, and a corresponding 55% increase in cells arrested in G 1 phase. This observation was confirmed by a cell proliferation assay that involves incorporation of the nucleotide analog, BrdUrd, wherein YFP-BARD1 reduced the proportion of BrdUrd-positive cells by ϳ37% compared with YFP alone (Fig. 8B). This degree of cell cycle arrest was comparable to that elicited by the overexpression of YFP-BRCA1 (Fig. 8B), a protein well known to arrest cells at the G 1 /S transition (25). The cell cycle arrest mediated by BARD1 was at least partly independent of p53 status, as indicated by the BARD1-induced G 1 arrest observed in the Saos-2 osteosarcoma cell line, which expresses mutant p53 (see Fig. 8A, lower panel).
We next determined whether the localization of BARD1 contributed to its regulation of the cell cycle, by comparing NES and NLS mutant forms of BARD1 in transfected cells. The nuclear localized NES mutant (L107A) of BARD1 was almost as effective as wild-type BARD1 in blocking the cell cycle. In contrast, the NLS mutation almost completely abolished the BARD1-dependent cell cycle arrest (see Fig. 8, A and B). The inability of the NLS mutant to cause a G 1 arrest was not due to a difference in expression of the ectopic BARD1 proteins or an alteration in p53 status, since equivalent expression of ectopic BARD1 and endogenous p53 was demonstrated by the Western blot in Fig. 8A. This is the first report describing BARD1-dependent G 1 phase cell cycle arrest, and our results indicate that this novel activity requires the NLS-mediated nuclear localization of BARD1. DISCUSSION The BARD1 gene is mutated in a subset of breast/ovarian cancer patients (2-4) and has therefore been implicated as a FIG. 5. Identifying critical active residues in BARD1 NLS3. To test the importance of the basic arginine (R) and lysine (K) residues in NLS3, double amino acid substitutions were introduced into the ␤-galactosidase-NLS3 expression construct, and their influence on subcellular distribution was determined by immunofluorescence microscopy of MCF-7 cells (top). The nuclear/cytoplasmic ratio of fluorescence was determined as indicated. In addition, we assessed the effect of an NLS3 mutation (Mut1) relative to wild-type NLS3 (with ␤-galactosidase (␤-Gal) alone as control), using immunoblotting of fractionated cell extracts from U2OS cells (bottom). Both approaches revealed that mutation of key basic amino acids in NLS3 abolishes its strong nuclear localization activity. Topoisomerase II staining was performed to ensure proper cell fractionation in nuclear and cytoplasmic extracts. N, nuclear. C, cytoplasmic.
FIG. 6. Site-directed mutagenesis of NLS3 prevents nuclear accumulation of BARD1. Site-directed mutagenesis was used to introduce the indicated double point mutations into NLS3 of full-length BARD1. The constructs were transiently expressed in MCF-7 cells, and the transfected cells were then analyzed for subcellular localization of ectopic BARD1. Representative immunofluorescence images and intracellular distribution of FLAG-tagged BARD1 (values are mean Ϯ S.D. scoring of 100 cells from each of three independent experiments) are shown. Similar data were observed in U2OS cells, and for those transfected cells, fractionated extracts were tested for BARD1 expression by Western blot. As shown in the bottom panel, the YFP-BARD1-NLS3mutant2 was more cytoplasmic than wild-type protein (WT). Staining for Topoisomerase II was included as a loading and fractionation control. N, nuclear. C, cytoplasmic.

FIG. 7. Cell cycle regulation of endogenous BARD1 expression and localization.
Expression and localization patterns of endogenous BARD1 were assessed in U2OS cells synchronized by a nocodazoleinduced G 2 /M phase cell cycle arrest (see "Materials and Methods") and then released into drug-free medium. Cell fractions were harvested at the indicated time points after removal of nocodazole. Parallel samples were collected for fractionation and Western blotting and for analysis of the cell cycle population by fluorescence-activated cell sorting (see percentage of cells in G 1 , S, or G 2 /M phase below Western blot). The level of BARD1 expression decreased soon after release from the G 2 /M phase arrest and then increased as the fraction of cells in G 2 /M phase increased. Staining for topoisomerase II was included as a fractionation control. N, nuclear. C, cytoplasmic. potential tumor suppressor protein. We previously showed that BARD1 can shuttle in and out of the nucleus (21) and that nuclear export of BARD1 stimulates its apoptotic function. Given that BARD1, in association with BRCA1, functions within the nucleus in genomic maintenance (1,9,13,27), in particular through the homology-directed DNA repair pathway (26), understanding how it is targeted to the nucleus may provide insights into its regulation and activity. We therefore mapped the nuclear localization signals of human BARD1. Three functional NLSs were defined; the weakest NLS is conserved in other species, but the two strongest NLSs are not conserved in mouse or rat BARD1, indicating a strong requirement for nuclear targeting of BARD1 in humans. Consistent with its potential tumor suppressor role, we showed for the first time that the overexpression of BARD1 induced a G 1 phase cell cycle arrest. Mutation of the dominant BARD1 NLS sequence impaired the ability of BARD1 to enter the nucleus and consequently reduced its growth arrest capability. Thus, whereas cytoplasmic (but not nuclear) localized BARD1 was previously shown to induce apoptosis (21,22), we now report that nuclear (but not cytoplasmic) localized BARD1 can induce cell cycle arrest (Fig. 8). We speculate that the regulation of BARD1 shuttling between nucleus and cytoplasm may dictate its influence on cell survival or proliferation.
BARD1 displays a striking similarity to BRCA1 in the positioning of its RING and BRCT protein interaction domains, and we recently identified a BARD1 nuclear export signal residing at the same relative position as the NES in BRCA1, within the dimerization region and flanking the RING domain (20,21). Here we identified two main NLSs in BARD1, and as for BRCA1 (23), these two NLSs are centrally located and do not overlap any known functional protein domain. Whereas the mutation of NLS4 only partially impaired BARD1 nuclear import, the integrity of NLS3 was absolutely essential for nuclear localization of human BARD1. Therefore, BRCA1 and BARD1 share a striking conservation in the position of not only their protein interaction domains (RING and BRCT) but also their nuclear import and export sequences. BARD1 is much smaller in size than BRCA1, and its distinguishing characteristic is the presence of an ankyrin repeat domain located in its C terminus, between the NLS and BRCT domains. We propose that the positioning of the functional NLS signals in both proteins is related to optimal accessibility for interaction with the importin receptors after formation of the BARD1/BRCA1 heterodimer. In this regard, the weaker but highly conserved NLS2 signal is located very close to the BRCA1-binding site of BARD1 and is therefore likely to become masked when these proteins interact.
We previously attributed a "nuclear chaperone" function to BARD1, based on its ability to stimulate nuclear entry and retention of BRCA1 (27,30). BARD1 provides a cellular mechanism for the nuclear import of frequently expressed BRCA1 exon 11 splice variants that lack a nuclear import signal of their own (27). Indeed, mutation of the BARD1 NLS reduced its ability to import NLS-deficient forms of BRCA1 into the nucleus (data not shown), supporting the role of BARD1 as a nuclear chaperone for BRCA1. Given the strong sequence conservation of BARD1 between different species (31,32), it is somewhat surprising that human BARD1 should differ from rodents in its NLS sequences. It will be of interest to determine which sequences are required for nuclear localization of murine, rat, or X. laevis BARD1. A recent report found that an N-terminally truncated murine BARD1 protein in which the conserved NLS2 was deleted accumulated in the nucleus (22), indicating that this conserved sequence might not function as an NLS in other species either. These findings suggest that the positioning of the major NLSs is not conserved in BARD1, a finding reminiscent of BRCA2, where nuclear local- . The transient expression of wild-type BARD1 caused a G 1 phase arrest, and this was blocked by the NLS mutation but not the NES mutation. Equivalent expression of the different YFP-BARD1 constructs was confirmed by Western blot, and BARD1 overexpression did not alter p53 protein levels. The YFP-BARD1 construct was also found to cause a modest but reproducible G 1 phase arrest in the p53-mutant cell line, Saos-2. B, to confirm the flow cytometry data, YFP-BARD1 (WT or NLS mutant) and YFP-BRCA1 were compared for their ability to arrest cell cycle in U2OS cells as assessed by BrdUrd incorporation. 48 h post-transfection, the proportion of BrdUrd-positive transfected cells was scored by immunofluorescence microscopy (see graph). Values shown are mean Ϯ S.E. from two independent experiments. YFP-BARD1 induced a similar block to S phase as achieved by YFP-BRCA1 (relative to YFP alone), and the BARD1 NLS mutation abrogated this effect.
ization of the human protein depends on C-terminal NLSs, whereas the murine protein requires an intact N terminus to enter the cell nucleus (33,34).
Our findings are not consistent with those of Westermark et al. (26), who reported that an N-terminal fragment of BARD1 encompassing the first 202 amino acids is localized in the nucleus of mouse embryonic stem cells. This contrasts with our data using the exact same peptide (kindly supplied by Dr. M. Moynahan), which strongly accumulated in the cytoplasm in immortalized human and mouse cells (Fig. 2 and data not shown). Our observation was highly reproducible and consistent with results seen using other short N-terminal BARD1 peptides (e.g. aa 1-267), which contained the nuclear export signal. A possible explanation for this difference in results could be that specific transport sequences are functional during different stages of development. Ultimately, it would also be interesting to determine whether there is any functional significance to changes in NLS positioning that may have occurred during evolution.
The majority of studies on BARD1 have been in the context of its interaction with BRCA1, revealing that this stable heterodimer acquires ubiquitin ligase activity (8, 16 -18), is trapped in the nucleus due to nuclear retention (27), localizes in nuclear DNA repair foci (12,27), and functions in DNA repair (26) and centrosome duplication (35). Interestingly, a comparison of BRCA1-null and BARD1-null mice revealed that loss of either protein or both caused embryonic lethality due largely to altered cell proliferation and reduced DNA repair function (9,26). Moreover, each protein individually can induce apoptosis (see references within Refs. 6 and 30) or cell cycle G 1 phase arrest (see Ref. 25 for BRCA1 and this study for BARD1) when overexpressed. In U2OS osteosarcoma cells, BARD1 induced a G 1 phase arrest and was just as effective as BRCA1 at blocking cell proliferation as assessed by a BrdUrd incorporation assay (Fig. 8). This duality of function suggests that either protein can substitute for the other, possibly reflecting the similarity in their protein interaction domains. Alternatively, the ankyrin repeats in BARD1 could be responsible for its cell cycle arrest activity, in that the p16 tumor suppressor, which is well known to arrest cells at the G 1 /S transition, also contains an ankyrin repeat sequence (reviewed in Ref. 36). Since the ankyrin repeat is a protein interaction domain, it will be interesting to determine whether BARD1 regulates the cell cycle via direct interaction with p16 or other ankyrin repeatcontaining proteins.
BARD1 resembles BRCA1 in its expression during the cell cycle. BRCA1 is known to become hyperphosphorylated and to increase in expression as cells enter S phase (37), although BRCA1 levels can continue to increase throughout the G 2 /M phase transition (11,29). We observed an increase in expression of human BARD1 during G 2 /M phase and a rapid loss of BARD1 expression at 4 h after nocodazole treatment, which is consistent with recent findings by Choudhury et al. (29). We have extended those findings, however, by employing subcellular fractionation at different time points after release from the G 2 /M phase arrest and found that endogenous BARD1 was predominantly localized in the nucleus throughout the cell cycle (keeping in mind a breakdown of the nuclear envelope during mitosis). This may be due to the driving influence of multiple nuclear localization signals as identified here. It is interesting to note that the BARD1/ BRCA1 heterodimer was reported to ubiquitinate a nucleolar protein, nucleophosmin/B23, and this protein co-localized with BARD1/BRCA1 at mitosis (38). B23 is implicated in regulation of centrosome duplication (39) and apoptosis inhibition, two processes associated with formation of the BARD1/BRCA1 heterodimer. The up-regulation of BRCA1 and BARD1 in mitotic cells may be linked to their role in the G 1 /S or G 2 /M phase DNA damage checkpoints (reviewed in Ref. 40). The implications of cell cycle regulation of BARD1 and the mechanism by which BARD1 elicits a cell cycle arrest will be important avenues of future investigation.