Distinct Roles of BARD1 Isoforms in Mitosis: Full-Length BARD1 Mediates Aurora B Degradation, Cancer-Associated BARD1β Scaffolds Aurora B and BRCA2 Free

Breast and ovarian cancers with mutations in BRCA1 and BRCA2 show a severe genomic instability phenotype (1). Genomic instability and a premalignant phenotype are also features of BRCA1-associated ring domain protein 1 (BARD1)–deficient cells (2), causing early embryonic death of BARD1 knockout mice (3). BARD1 and BRCA1 form a stable heterodimer via their respective NH₂ terminal RING finger domains (4), and either Bard1 or Brca1 deletion induces breast cancer in tissue-specific conditional knockout mice (5). BARD1 is the major protein-binding partner of BRCA1, with whom it forms a stable heterodimer and acts in tumor suppressor functions but has also BRCA1-independent functions in apoptosis (6, 7). The BARD1-BRCA1 heterodimer is an E3 ubiquitin ligase implicated in DNA repair and homologous recombination (8, 9), centrosome duplication (10), and mitotic spindle assembly (11), which are essential functions for maintaining genomic stability (10, 12).

During S-phase, BRCA1, BRCA2, and BARD1 are partially colocalized to distinct nuclear dots (13, 14). During mitosis, BRCA1 localizes to spindle poles (15), and BRCA2 localizes to the midbody during telophase and cytokinesis, wherein it is involved in contractile ring and midbody formation and completion of cytokinesis (16). Depletion of BRCA2 impedes abscission of the midbody and cell separation (16), providing an explanation for the genetic instability observed in cancers associated with mutations in BRCA2. Midbody formation and abscission are also controlled by the microtubule-binding protein transforming acidic coiled coil–containing protein 1 (TACC1) and the mitotic kinase Aurora B (17). TACC1 was reported to interact with BARD1 in Caenorhabditis elegans (18). BARD1 expression is up-regulated during mitosis (19), presumably due to phosphorylation by cyclin-dependent kinase/cyclin complexes (20), but its subcellular localization during mitosis has not been investigated.

The mitotic functions of BRCA1 and BARD1 (11) could explain why normal proliferating cells are not viable without BRCA1 or BARD1. However, mouse trophoblast giant cells, which undergo endomitosis, a nuclear division not followed by cell division, are not affected in the Bard1 knockout mouse (3), suggesting that the lethal phenotype of BARD1 depletion is linked to a function at exit of mitosis, similar to functions attributed to BRCA2 (16).

A genetic link between BRCA2 and BARD1 was found with the BRCA2 mutation 999del5, which, when combined with the BARD1 variant Cys557Ser, results in 100% probability for developing breast and/or ovarian cancer for carriers of the BRCA2 999del5/BARD1 Cys557Ser double mutation (21). Thus, BRCA2 and BARD1 might act in a common pathway.

Whereas mutations in BARD1 are rare in cancers (7), aberrant up-regulation of isoforms is associated with poor prognosis in breast and ovarian cancer (22, 23). We found aberrantly expressed, differentially spliced BARD1 isoforms, which lack the BRCA1-interacting RING finger, in breast, ovarian, and endometrial cancer cells (23, 24) and human cytotrophoblasts (25). Repression of these isoforms in cancer cells that lack full-length (FL) BARD1 led to growth arrest (23), which suggests that BRCA1-independent proproliferative, presumably mitotic, functions are retained in cancer-associated BARD1 isoforms.

We therefore designed this study to understand the different roles of FL BARD1 and cancer-associated isoforms in mitosis. We show that FL BARD1 and the RING finger–deficient p82 isoform BARD1β have different functions and protein-protein interaction properties. FL BARD1 interacts with BRCA1 and is involved in Aurora B ubiquitination and degradation during mitosis. In contrast, BARD1β stabilizes Aurora B and forms a complex with BRCA2 and Aurora B, proteins known to be involved in midbody formation (16, 17). These findings establish an unexpected link between BARD1 and Aurora B and BRCA2, which are aberrantly overexpressed or mutated in cancer. We suggest a molecular pathway that explains the function of FL BARD1 as tumor suppressor together with BRCA1 and loss of FL BARD1, but up-regulated expression of RING finger–deficient isoforms with proproliferative functions, in cancer.

Plasmid constructs. BRCA1 and BARD1 small interfering RNA (siRNA) constructs were generated by annealing complementary oligonucleotides and inserting them into the pSuperScript vector as BglII/HindIII fragment: human BRCA1-siRNA (si-56), human BARD1-siRNA 34 (si-34), and human BARD1-siRNA 78 (si-78) forward and reverse oligonucleotides have been described previously (23, 26); hBARD1-siRNA 423 (si-423) forward, 5′-G ATC CCC GTG CTC AGC AAG ACT CAT ATT CAA GAG ATA TGA GTC TTG CTG AGC ACT TTT TGG AAA-3′, and reverse, 5′-A GCT TTT CCA AAA AGT GCT CAG CAA GAC TCA TAT CTC TTG AAT ATG AGT CTT GCT GAG CAC GGG-3′. The pSuper fragments ClaI/EcoRI containing siRNA sequences were then subcloned in the lentiviral vector pLVTHM (27). pLVTHM and pLV-tTRKRAB plasmids were kindly provided by D. Trono. pLVTHM-SIGN (si-SIGN), kindly provided by V. Piguet, is used as a negative control (28).

Human FL BARD1 and splice variant BARD1β cDNAs were amplified from human primary fibroblasts and HeLa cells, respectively. Flag epitope M2 and c-myc tags were inserted at the 5′ and 3′ ends of the cDNA, and the resulting construct recombined into the 2K7_bsd lentiviral vector (29). Si-78–resistant FL BARD1 (FLm78) construct was generated by site-directed mutagenesis using the FL BARD1 construct as template and the following primers Mut78-F 5′-AAG TGT ATG CTC GGA ATA CTC AAT GGA TG-3′ and Mut78-R 5′-AGC ATC CAT TGA GTA TTC CGA GCA TAC AC-3′.

The expression plasmid HA-tagged ubiquitin was kindly provided by G. Courtois and M. Huber.

Cell culture and transduction with lentivirus. The 293T and HeLa cell lines were cultured in DMEM supplemented with 10% fetal calf medium. All recombinant lentiviruses were produced and purified according to standard protocols (30). 293T cells were cotransfected with the lentivirus vector, the packaging vector psPAX2, and the envelope vector pMD2G by calcium phosphate precipitation. After 16 h, medium was changed, and recombinant lentiviruses were harvested 24 h later. For lentivirus transduction, HeLa cells were plated on six-well plates (10⁴ cells per well) for 24 h and then incubated with the medium containing recombinant lentivirus vectors for 48 h. To control the expression of the siRNA, the cells were cotransduced with lentivirus derived from the vectors pLVTHM and pLV-tTRKRAB (27). Coexpressed tTR-KRAB locks onto the tet-on promoter. Upon doxycyclin exposure, KRAB is derepressed and allows transcription of siRNAs.

Generation of growth curves. HeLa cells were trypsinized, centrifuged, and resuspended in 5 mL of medium. The cells were counted using a hemocytometer, then diluted to have a 20 to 30% confluency in the Petri dish. Triplicate plates were counted every 48 to 72. The results were plotted on a logarithmic scale after determining the cell number along the exponential phase and calculating the mean population-doubling time.

Immunofluorescence microscopy. Cells were fixed with 2% paraformaldehyde for 15 min at room temperature (RT) or with methanol for 6 min at −20°C and rinsed in acetone for 30 s. Paraformaldehyde-fixed cells were permeabilized in 1% Triton/PBS for 15 min at RT and then blocked in 1% serum/PBS for 30 min. Coverslides were incubated with appropriate antibodies for 1 h at RT in 1% FCS/PBS, washed, and stained with 4′,6-diamidino-2-phenylindole for 3 min. Coverslips were mounted using fluorogard solution and analyzed under a Nikon epifluorescence microscope, and images were captured with a 3.3-megapixel CCD camera. Images were processed with Metamorph software (Visitron). Primary antibodies used were BRCA1 (Ab-1; Calbiochem), BARD1 (H-300; Santa Cruz), BARD1 (BL518; Bethyl Laboratories), Aurora B (AIM-1; BD Biosciences). PVC was reported previously (2). Antibody p25 was produced in rabbits against peptide sequence MVAVPGTVAPRC encoded in alternative open reading frame (ORF) of exon 1. Antibody specificity was probed in inversed ELISAs with different peptides and on human cancer cell lines (data not shown).

Immunoprecipitation and Western blot. Protein lysates used for immunoprecipitation were prepared from HeLa cells, unsynchronized and synchronized in G₁-S or G₂-M. At 24 h after plating, cells were arrested in S phase using 2 mmol/L thymidine for 18 h, released from the arrest for 9 h, arrested a second time using thymidine (2 mmol/L) for 18 h (G₁-S), and released from the arrest for 5 h (G₂-M).

Cells were lysed in radioimmunoprecipitation assay (RIPA) buffer [50 mmol/L Tris (pH 8), 150 mmol/L NaCl, 2 mmol/L EDTA, 0.5% NP40, 10% glycerol] supplemented with protease inhibitors (complete EDTA-free, Roche Applied Science). Protein concentrations were determined by the Bradford procedure (Bio-Rad). Lysates containing 400 μg of protein were precleared by stirring with 100 μL of protein G-Sepharose for 1 h at 4°C. After centrifugation, immunoprecipitation was performed with the precleared lysate and 1 μg of antibody for 3 h at 4°C. Thirty microliters of protein G-Sepharose were added and incubated for 30 min at 4°C. After centrifugation, beads were washed twice with lysis buffer, and proteins were eluted by heating at 95°C for 5 min in SDS loading buffer and 100 mmol/L DTT.

Proteins were subjected to SDS-PAGE and blotted onto polyvinylidene difluoride membranes (Immobilon-P, Millipore) using standard methods. The following primary antibodies were used: BRCA1 (Ab-1; Calbiochem), BARD1 (H-300; Santa Cruz), BARD1 (BL518; Bethyl Laboratories), Aurora B (AIM-1; BD Biosciences), BRCA2 (Ab-1; Calbiochem), TACC1 (Upstate Biotechnology Euromedex), γ-tubulin (C11; Santa Cruz), β-tubulin (D10; Santa Cruz), cdc-2 (POH-1; Cell Signaling), P-cdc2-Tyr 15 (Cell Signaling), actin (C-2; Santa Cruz), cyclin A (BF-683; Santa Cruz), anti-HA (HA.11; Covance). Antibody used against Aurora A was obtained from C. Prigent (31). Horseradish peroxidase–conjugated secondary antibodies (Amersham Biosciences) were used for detection of immunoreactive proteins by enhanced chemiluminescence (Amersham Biosciences).

Protein stability assay. To determine the effect of FL BARD1 and spliced variant BARD1-β on the stability of Aurora B, cells growing in six-well plates were transfected with 1 μg of 2K7 vector (control) or 2K7 vector expressing human FL BARD1 or spliced variant BARD1β and 3 μL of Fugene-6 (Roche Applied Biosciences). Forty-four hours after transfection, 100 μg/mL of cycloheximide (Sigma) was added to cells to inhibit protein synthesis, and cells were harvested in RIPA buffer at various time intervals, as indicated in the figure.

In vivo ubiquitination assay. At 30% confluence, cells growing in 10-cm-diameter dish were cotransfected with 1 μg of vector encoding HA-tagged ubiquitin and 2 μg of 2K7 vector (control) or 2K7 vector expressing human FL BARD1 or spliced variant BARD1β and 9 μL of Fugene-6. Thirty-six hours after transfection, cells were treated with 10 μmol/L of the proteasome inhibitor MG132 (Calbiochem) for 12 h. After cell lysis, immunoprecipitations were performed with 800 μg of total cellular proteins and 2 μg of anti-HA antibody. The immunocomplexes were pulled down, as described above, and analyzed by SDS-PAGE on 10% gels and probed for the presence of polyubiquitinated-Aurora B using anti-Aurora B antibody.

BARD1, but not BRCA1, colocalizes with Aurora B at the midbody during cytokinesis. To investigate the function of BARD1 during mitosis, we studied its intracellular localization at different stages of mitosis by immunofluorescence microscopy. Double labeling with anti-BARD1 and anti–β-tubulin antibodies revealed that BARD1 staining was partially localized along mitotic spindle microtubules in metaphase and early anaphase cells (Fig. 1A). This localization was also observed in several other cell lines, namely human HeLa, PC-3, and mouse TAC-2 cells (data not shown), with BARD1 H300 antibody, directed against the NH₂ terminal half of the BARD1 protein. During telophase and cytokinesis, colocalization of BARD1 and microtubules was also observed, specifically at the midbody, whereas staining during anaphase and early telophase was not restricted to spindle microtubules.

To investigate whether BARD1 localized to the midbody independently of BRCA1, we performed costaining experiments with anti-BARD1 and anti-BRCA1 antibodies in mitotic cells. BARD1 and BRCA1 showed colocalization during metaphase and anaphase but localized separately during telophase and cytokinesis (Fig. 1B). In early telophase, BARD1 staining was observed around the chromatin but increased at the site of cleavage furrow ingression. BRCA1 staining was around the chromatin but less staining was observed in regions of the presumed contractile ring. During telophase and cytokinesis, BARD1 staining was localized at the midbody but no BRCA1 staining was observed at the midbody (Fig. 1B).

The observation of BARD1 staining at spindle poles during metaphase and at the midbody during cytokinesis suggested that BARD1 interacts with BRCA1, as reported (10, 11), at early stages of mitosis, but with proteins other than BRCA1 during telophase and cytokinesis.

The specific subcellular localization of BARD1 during mitosis was reminiscent of the localization of regulators of cell division, known to translocate from the midplate to the midbody as part of the chromosomal passenger complex (CPC; ref. 32). One such protein, the mitotic kinase Aurora B, is of particular interest because it is a member of CPC (33–35), is required for midbody formation (17), and is up-regulated in breast cancer (36). Double-labeling for BARD1 and Aurora B was performed to determine a possible colocalization. Aurora B was only weakly expressed during S phase and localized to distinct spots in the nucleus in HeLa cells. During metaphase, Aurora B staining was increased and located at the midplate. In anaphase, Aurora B staining formed a tight line at the midplate and colocalized with BARD1 (Fig. 1C). During telophase and cytokinesis, when Aurora B is gradually degraded to form two distinct spots at the midbody adjacent to the microtubule abscission point, BARD1 staining was partially overlapping with Aurora B staining. Further inspection showed that BARD1 and Aurora B colocalize already in metaphase and gradually dissociate during cytokinesis (data not shown).

To investigate whether colocalization with Aurora B is also a property of BRCA1, we performed double labeling experiments with anti-BRCA1 and anti-Aurora B antibodies. In anaphase, BRCA1 staining was observed at the midplate, but only weakly colocalized with Aurora B compared with BARD1 colocalization with Aurora B. In telophase, no BRCA1 staining at the midbody and no colocalization with Aurora B were observed (Supplementary Fig. S1).

These data show BARD1 association with BRCA1 at the spindle poles during metaphase but costaining for BARD1 and Aurora B during anaphase, telophase, and cytokinesis. This suggests that BARD1, but not BRCA1, colocalizes with Aurora B at the midbody.

BARD1 interacts with Aurora B, TACC1, and BRCA2 during mitosis. Because BARD1 colocalizes with Aurora B during mitosis, we investigated whether BARD1 is part of an Aurora B protein complex during mitosis. Aurora B forms a complex with TACC1, which, like Aurora B, localizes at the metaphase plate during anaphase and the midbody during cytokinesis (17). TACC1 was also identified as binding partner of BARD1 in C. elegans cells (18). Like Aurora B, TACC1 is up-regulated and acts as an oncogene in breast and ovarian cancer (37).

To show that BARD1, TACC1, and Aurora B physically interact during mitosis, we performed coimmunoprecipitation experiments using anti-BARD1 H300 antibodies, which recognize the NH₂ terminal 300 amino acids of the BARD1 protein. Immunoprecipitation with anti-BARD1 antibodies confirmed TACC1 interaction with BARD1 (Fig. 2A).

We then probed anti-BARD1 immunoprecipitations with antibodies against Aurora B and Aurora A, the mitotic kinase that interacts with and phosphorylates BRCA1 (33–35). Both Aurora A and Aurora B levels were increased in G₂-M. However, coimmunoprecipitation with BARD1 was only observed for Aurora B and only in G₂-M cell extracts (Fig. 2A). Aurora A levels were very low, and a binding to BARD1 cannot be excluded. Together, these data indicate that BARD1 forms a complex with TACC1 and Aurora B during mitosis.

TACC1 and Aurora B are proteins required for midbody formation (17), a function also involving BRCA2 (16). Because BRCA2 also localizes to the midbody (16) and a genetic interaction of BARD1 and BRCA2 suggests that they act in a common pathway (21), we tested whether BARD1 interacts with BRCA2. We probed BARD1 coimmunoprecipitations with BRCA1, as positive control, and BRCA2 antibodies. BARD1 and BRCA1 coimmunoprecipitated in G₁-S and G₂-M cell extracts, and binding was not increased in mitotic extracts, although total BRCA1 and BARD1 levels increased in G₂-M cells (Fig. 2A).

Importantly, BRCA2 coimmunoprecipitated with BARD1 but exclusively in mitotic cell extracts. These data suggest that BARD1, TACC1, Aurora B, and BRCA2 act in a common pathway at the exit of mitosis.

To confirm BARD1 interaction with TACC1, Aurora B, and BRCA2, we performed coimmunoprecipitations with anti-TACC1, anti-BRCA2, and anti-Aurora B antibodies (Fig. 2B). As a positive immunoprecipitation control, we used anti-BRCA1 antibodies. BRCA1 immunoprecipitation, when probed on Western blots with the anti-BARD1 antibody BL518 directed against an epitope encoded on exon 4, showed 97-kDa FL BARD1 interaction with BRCA1. Negative immunoprecipitation control showed no precipitation. Anti-TACC1 antibodies also coprecipitated FL BARD1 and coprecipitated slightly better in extracts of mitotic cells. However, immunoprecipitations with anti-BRCA2 and anti-Aurora B showed coprecipitation of a 82-kDa protein, which we previously identified as BARD1β, but not of p97 FL BARD1. BARD1β lacks the region comprising the BRCA1-interacting RING domain (23, 25), hence cannot interact with BRCA1 in anti-BRCA1 immunoprecipitation. Thus, BARD1β binds to BRCA2 and Aurora B, and the interaction is strongest with Aurora B and restricted to mitotic cells. These results also show that FL BARD1 does not bind to BRCA2 and Aurora B, suggesting that the NH₂ terminal RING finger does not promote but rather hinders BARD1 interaction with BRCA2 and Aurora B.

To investigate cell cycle stage-specific expression, cell extracts used for immunoprecipitations were prepared from cells, either nonsynchronized or synchronized, and enriched for G₁-S or G₂-M cell stages. In parallel to immunoprecipitations, cell extracts were monitored for expression of mitotic marker proteins, such as phosphorylated CDC2 and cyclin A (Fig. 2C).

Distinct repression of FL BARD1 or FL BARD1β. To distinguish localization and function of FL BARD1 or the RING finger–deficient isoform BARD1β, we performed repression experiments with siRNAs targeting different regions or exons of BARD1 in HeLa cells (Fig. 3A). To determine the phenotype of BARD1 depletion, we generated stable inducible cell lines by cotransducing siRNAs in lentiviral vectors with the tTRKRAB repressor operated through the Tet-on system (27, 30).

The efficiency of BARD1 and BRCA1 siRNAs was analyzed on Western blots probed with anti-BARD1 BL518 (Fig. 3B) or H300 antibodies (Supplementary Fig. S2). Expression of si-34, directed against exon 2 of BARD1, led to repression of 97-kDa FL BARD1, but not BARD1β (Fig. 3B). FL BARD1 levels were also decreased after repression of BRCA1 (si-56) due to mutual stabilization of the two proteins. BARD1β was repressed by si-423, directed against exon 4 and si-78, directed against exon 9 but not by si-34, consistent with the structure of this isoform (Fig. 3A). The BARD1β isoform seemed more stable than FL BARD1, and its repression was observed later than repression of FL BARD1. BARD1β protein levels exceeded those of FL BARD1, which was also observed in different other cell lines (ref. 23; Supplementary Fig. S2).

BARD1β, but not FL BARD1, localizes to the midbody. Because BARD1, but not BRCA1, staining was observed at the midbody during telophase and cytokinesis, we wondered whether the observed staining was due to the presence of FL BARD1 or BARD1β isoform. To distinguish localization and possible functions of FL BARD1 and BARD1β, we performed immunofluorescence staining in cells that are deficient of isoforms but express FL BARD1, using the anti-BARD1 BL518 antibody, which recognizes both FL BARD1 and BARD1β, for detection (Fig. 3C).

To generate these cells, BARD1-FLm78 (FL BARD1 expression clone carrying a silent mutation at the si-78 target region) was constitutively expressed in si-78 expressing cells. Western blots probed with anti-BARD1 BL518 confirmed the identity of FL BARD1 and showed that levels of exogenous BARD1-FLm78 were not affected by si-78–induced repression (Fig. 3D).

Using anti-BARD1 BL518 in S-phase cells, BARD1 staining was found in the nucleus and cytoplasm. In cells that expressed exogenous BARD1-FLm78, nuclear staining was increased, indicating that FL BARD1 localized to the nucleus, whereas cytoplasmic staining might reflect BARD1β localization (Fig. 3C). This localization was reported previously for NH₂ terminal truncated forms of BARD1 (19). In anaphase, BARD1 staining was observed at the spindle poles and the midplate, and in telophase, an intense staining was observed at the midbody. BARD1-FLm78 expression in cells depleted of FL BARD1 and BARD1β by si-78 expression did not change the pattern of BARD1 staining during S phase, metaphase, or anaphase (data not shown). However, staining to the midbody in telophase was missing. These data indicate that the BARD1β isoform, but not FL BARD1, is localized to the midbody during telophase and cytokinesis.

We further confirmed that BARD1β, and not FL BARD1, localizes to the midbody by using antibodies that selectively recognize the two isoforms. Antibody BARD1 H300 (directed against the 300 NH₂ terminal amino acids) shows staining at midbody in cytokinesis, but antibody BARD1 PVC (directed against exon 3, which is deleted in BARD1β) does not show staining at the midbody (Supplementary Fig. S3A and B).

We also generated an antibody, p25, directed against an epitope encoded in the alternative ORF of exon 1, hence unique to BARD1β (Supplementary Fig. S3A). This p25 antibody specifically recognizes BARD1β by immunoprecipitation and Western blot (Supplementary Fig. S3C). Importantly, p25 shows staining at the midbody in cytokinesis (Supplementary Fig. S3B).

Thus, immunofluorescence staining of cells depleted of FL BARD1 and BARD1β and complementation with mutated FL BARD1, use of FL BARD1-specific antibody PVC, and use of BARD1β-specific p25 show that BARD1β, but not FL BARD1, is localized to the midbody.

BARD1β is required for cell growth. To analyze the phenotype of BARD1 depletion, we first monitored cell growth over a period of 10 days. Repression of BARD1 with different siRNAs led to distinct results (Fig. 4A and B). siRNA si-34, which affects only FL BARD1 (Fig. 3A), had little effect on cell growth, similar to repression of BRCA1 (si-56). However, expression of siRNA si-78, targeting both FL BARD1 and BARD1β, led to growth arrest. When compared with various control cells, growth rates of si-78 expressing cells were zero after 2 to 3 days of doxycyclin addition (Fig. 4C). si-423 expression, targeting exon 4 common to FL BARD1 and BARD1β, resulted in growth arrest but only after 5 to 6 days (Fig. 4B). This difference in growth arrest induced by si-78 or si-423 expression was correlated with the depletion of BARD1β, but not FL BARD1, proteins (Fig. 3B). These data indicate that BARD1β, but not FL BARD1 or BRCA1, retain the minimal protein functions required for cell proliferation.

The depletion of BARD1 induced various mitotic defects and genetic instability. We monitored the effects of BARD1 depletion by si-34 or si-78 and of BRCA1 depletion by si-56 in HeLa cells by time-lapse video imaging, immunofluorescence microscopy, and cytogenetic analysis (Supplementary Video S1–S4; Supplementary Table S1; Supplementary Figs. S4–S7). BARD1-depleted cells had difficulties in forming bipolar mitotic spindles and in abscission of the midbody. Although cell cycle times were only slightly elevated for cells expressing si-34 or si-56, cell cycle times of si-78 expressing cells could not be determined, because only few cells completed mitosis. Presumably, due to these defects, high levels of chromosomal instability were observed, consistent with previous observations (2, 3).

BARD1 controls protein expression levels of Aurora B. The phenotype of cytokinesis failure of BARD1-depleted cells is similar to the phenotype resulting from BRCA2 depletion (16) and, together with the observed interaction of BRCA2 with BARD1β (Fig. 2), supports the hypothesis that BARD1β, Aurora B, and BRCA2 act in a common pathway in cytokinesis.

To determine whether Aurora B expression depends on BARD1, we analyzed Aurora B and BARD1 expression in HeLa cells that were either untreated, depleted of BARD1 by si-RNA, or overexpressed FL BARD1.

We monitored protein levels of Aurora B after BARD1 repression or overexpression of FL BARD1 on Western blots. FL BARD1, BARD1β, and Aurora B expression was analyzed at various time points after siRNA induction on Western blots (Fig. 5A). Anti–β-tubulin staining was performed as loading control. Interestingly, Aurora B levels increased with time of induction of si-34 and paralleled the depletion of FL BARD1 but not of BARD1β. In si-423 cells, wherein a rapid depletion of FL BARD1 but slow decrease of BARD1β was observed, Aurora B levels first increased and then declined in parallel with the repression of BARD1β. Consistently, in si-78 cells, wherein FL BARD1 and BARD1β were rapidly degraded, no significant change of Aurora B expression is observed. The overexpression of FL BARD1, on the contrary, was associated with reduced Aurora B levels (Fig. 5A).

Seemingly, FL BARD1 and BARD1β have opposing effects on Aurora B stability. To determine how they affected Aurora B stability, cells were transfected with vectors expressing either FL BARD1 or BARD1β or empty vector as control and then treated with the protein translation inhibitor cycloheximide. We monitored Aurora B degradation during a time course of 8 hours. In cells overexpressing FL BARD1, Aurora B was degraded more efficiently than in control cells, whereas in cells overexpressing BARD1β, Aurora B remained nearly stable (Fig. 5B).

We then asked whether the observed Aurora B degradation was associated with Aurora B ubiquitination and whether ubiquitination of Aurora B was related to FL BARD1. Cells were cotranfected with expression plasmids encoding either FL BARD1 or BARD1β, and HA-tagged ubiquitin expression vector, and treated with the proteasome inhibitor MG132 to allow accumulation of ubiquitinated proteins. Immunoprecipitation of HA-tagged ubiquitinated Aurora B was then visualized by Western blotting with anti-Aurora B antibodies. The ubiquitinated form of Aurora B was only observed in cells overexpressing FL BARD1 and not in cells overexpressing BARD1β or control cells (Fig. 5C). The size of ubiquitinated Aurora B was consistent with previously found polyubiquitinated forms of Aurora B (38). These experiments show that FL BARD1, but not BARD1β, is involved in Aurora B degradation and suggest that FL BARD1 acts on Aurora B degradation via ubiquitination of Aurora B. BARD1β, on the contrary, is not associated with Aurora B ubiquitination and its overexpression inhibits Aurora B degradation.

To investigate Aurora B expression in cells as a function of BARD1 expression levels, we also performed immunofluorescence staining in wild-type HeLa cells and HeLa cells with exogenous overexpression or si-34 repression of FL BARD1. In untreated HeLa cells, Aurora B staining was confined to the nucleus, colocalized with chromatin during S phase, prophase, and metaphase, and aligned at the midplate during anaphase. Small dots of Aurora B staining were observed at the midbody during telophase and cytokinesis (Supplementary Fig. S8). In cells overexpressing FL BARD1, Aurora B staining was decreased at all stages compared with control cells and, most importantly, during anaphase. In contrast, Aurora B staining in cells depleted of FL BARD1 by si-34 expression was increased. This was most pronounced in cells arrested in cytokinesis. Typically, si-34 repressed cells were connected with elongated midbodies, as was also observed by anti–β-tubulin staining (Supplementary Fig. S6). The Aurora B dots on microtubule bridges connecting sister cells were larger than in control cells, and their localization was not restricted to the abscission point. Aurora B was accumulated in the cytoplasm and not restricted to the midbody.

Our results indicate that (a) FL BARD1 overexpression results in polyubiquitination of Aurora B and degradation and (b) FL BARD1 depletion leads to increased Aurora B accumulation, suggesting that FL BARD1 plays an important role in Aurora B degradation and turnover. However, BARD1β counteracts Aurora B degradation. Whereas BARD1β interaction with Aurora B is strong, FL BARD1 interaction with Aurora B is not observed and might be transient. One possible explanation for these observations might be that BARD1β protects Aurora B from degradation by forming a BARD1β-Aurora B-BRCA2 complex at the midbody.

A role for BARD1β in BRCA2-Aurora B interaction and midbody formation. FL BARD1 and BARD1β belong to different protein complexes and have distinct intracellular localization during mitosis. Both might be required for the regulated gradual degradation of Aurora B during mitosis, which is disturbed in many cancer cells.

Because a common pathway for TACC1 and Aurora B in cytokinesis has been described (17) and a function of BRCA2 in cytokinesis was also reported (16), we asked whether BRCA2 interacts with Aurora B. Immunoprecipitation with anti-BRCA2 antibodies showed coprecipitation of Aurora B (Fig. 6A). Therefore, Aurora B and BRCA2 form a complex in vivo. Immunoprecipitation of BRCA2 and TACC1 did not show interaction (data not shown).

When the same experiment was carried out with cells depleted of FL BARD1 after siRNA induction, Aurora B expression levels were increased and the proportion of Aurora B interacting with BRCA2 was also increased. These experiments suggest that BARD1β promotes the formation of a BRCA2-Aurora B complex. These data suggest that BARD1β acts with BRCA2 and Aurora B in midbody formation and abscission (Fig. 6B).

BARD1 has multiple functions in association with BRCA1 (7), including mitotic spindle formation (11). We show novel BARD1 functions during the late phases of mitosis, which involve interactions with TACC1, Aurora B, and BRCA2. Our data support the view that, during mitosis, FL BARD1 and the BARD1β isoform, which is expressed in many cancer cells (23), sequentially interact with BRCA1, TACC1, Aurora B, and BRCA2. BARD1-BRCA1 heterodimers are formed during metaphase and early anaphase at the centrosome, consistent with previous reports of BRCA1 localization to the spindle poles (11, 15) and its centrosome-related functions (39), whereas BARD1β interacts with Aurora B and BRCA2 during anaphase and cytokinesis.

BRCA1-independent and BRCA2-related functions of BARD1 during the late stages of mitosis. During the late stages of mitosis, in particular cytokinesis, BARD1β acts independently of BRCA1. The phenotypes induced by selective repression of FL BARD1 (si-34) or BRCA1 (si-56) were less pronounced than the phenotypes induced by si-78 and si-423, two siRNAs that target FL BARD1 and the BARD1β isoform, which lacks the BRCA1 interaction domain. BARD1β is also more abundant than FL BARD1 in many cancer cell lines (ref. 23; Supplementary Fig. S2C). During anaphase, telophase, and cytokinesis, BARD1, but not BRCA1, staining is observed at the midbody (Fig. 1). Staining at the midbody was not reconstituted by expression of exogenous FL BARD1 in cells depleted of all BARD1 isoforms (Fig. 3C), was not observed with an antibody specific for FL BARD1, but was observed with an antibody directed against the epitope encoded by an alternative ORF translated in BARD1β. These data suggest that BARD1β, and not FL BARD1, is implicated in functions at late stages of mitosis. Midbody localization was also reported for BRCA2 (16). There is strong evidence that both proteins act in a common pathway: (a) the phenotype of BARD1 depletion is similar to BRCA2 depletion (Supplementary Fig. S4), (b) BRCA2 coprecipitates with BARD1β (Fig. 2), and (c) strong genetic interaction of BRCA2 and BARD1 is observed in double mutation carriers (21).

Opposing roles of BARD1 and BARD1β in Aurora B degradation. The CPC protein Aurora B acts in a common pathway with TACC1 (17) in centrosome formation and cytokinesis and cancer progression (37, 40). TACC1 also interacts with BARD1, as was shown in C. elegans (18). We show here that human BARD1 interacts with TACC1 and Aurora B. Furthermore, BARD1 expression levels affect Aurora B stability and function: depletion of FL BARD1 results in Aurora B accumulation (Fig. 5) and hyperstabilization of the midbody, reminiscent to Aurora B up-regulation in cancer (41). BARD1 depletion also results in chromosome missegregation and aneuploidy (Supplementary Figs. S5–S7), as is observed in cancers with Aurora B up-regulation. Thus, mitotic and cytogenetic defects, induced by loss of FL BARD1 repression, could result from Aurora B up-regulation.

Interestingly, although FL BARD1 repression affects Aurora B ubiquitination and accumulation, FL BARD1 and Aurora B binding was not observed in coimmunoprecipitations, which might be due to a transient interaction between Aurora B and the BARD1-BRCA1 complex. In contrast to FL BARD1, BARD1β showed strong interaction with Aurora B in coimmunoprecipitations, which suggests that it is BARD1β that colocalized with Aurora B at the midbody during telophase and cytokinesis (Fig. 1C). Aurora B degradation seems to be under tight spatial and temporal control. Aurora B degradation, involving ubiquitination by APC, has been reported (42). Although the mechanism of inhibition of Aurora B degradation in FL BARD1 depleted cells is not clear at the moment, it is clear that the expression levels of FL BARD1 and BARD1β are critical (Fig. 5).

Our findings suggest further that BARD1β acts together with Aurora B in midbody formation and abscission and provide an explanation for the proproliferative action of BARD1 isoforms in cancer cells (23).

BARD1β links Aurora B and BRCA2 functions in cytokinesis. The BARD1 deficiency phenotype is similar to the phenotype induced by BRCA2 depletion and the Capan-1 mutation (16), namely cells with elongated midbodies (Supplementary Fig. S6). We show that BRCA2 interacts with Aurora B and BARD1β promotes this interaction, suggesting that the BARD1β-Aurora B-BRCA2 complex is required for completion of cytokinesis and deregulation of any of the components induces aneuploidy.

BRCA2 depletion affects myosin II organization at the cleavage furrow (16). Indeed, BARD1 protein concentrations (BARD1 overexpression and repression) influence ingression of the cleavage furrow and abscission,⁵

Our observations imply that (a) FL BARD1 versus BARD1β expression levels are important for degradation of Aurora B during mitosis, because FL BARD1 promotes and BARD1β hampers degradation (Fig. 5) and (b) BARD1β is promoting BRCA2-Aurora B interaction, which is likely to be crucial for midbody formation and abscission (Fig. 6).

In conclusion, we have shown that BARD1 and its cancer-associated isoform BARD1β act as functional link between early and late phases of mitosis and a few key regulators, BRCA1, Aurora B, and BRCA2. Thus, the function of FL BARD1, the tumor suppressor, in Aurora B degradation and the opposing function of BARD1β in scaffolding Aurora B and BRCA2 links the hitherto unrelated cancer-promoting pathways, in which Aurora B and BRCA2 are involved.

No potential conflicts of interest were disclosed.

Grant support: Swiss National Science Foundation grant 3100-068222 (I. Irminger-Finger), Inserm Ligue Nationale Contre le Cancer and Cancéropôle grants (D. Birnbaum), General Secretariat of Research of the Greek Ministry of Development grant 05NON-EU-449 (S. Gagos), and Fondation Pour La Recherche and Cancer and Solidarité (I. Irminger-Finger and S. Ryser).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank D. Suter for sharing precious constructs before publication and A. Caillon, L. Kcjancic Curty, S. Arnaudeau, and M. Hiourea for excellent technical assistance.