Identification of a γ-Tubulin-binding Domain in BRCA1¹

The breast and ovarian cancer susceptibility gene BRCA1 encodes a nuclear phosphoprotein of 1863 amino acids (M_r ∼220,000). Mutations of the BRCA1 gene are associated with more than 50% of familial breast and ovarian cancer and most BRCA1 mutations generate truncated proteins. BRCA1 has pleiotropic biological functions, possibly playing a role in transcriptional regulation, chromatin remodeling, DNA damage repair, cell cycle regulation, and checkpoint control. An NH₂-terminal ring-finger domain interacts with the proteins BARD1 and BAP1, and the COOH-terminal BRCT exhibits transcriptional activation. BRCA1 also interacts with proteins involved in transcriptional regulation, including p53, myc, pRB, RNA polymerase II, HDAC1 and 2, and CtIP. Overexpression of BRCA1 inhibits estrogen receptor-mediated transcription, which may provide an explanation for why BRCA1 mutations mainly affect breast, ovarian, and prostate cancers. BRCA1 protein interacts with Rad51, BRCA2, and Rad50, which suggests a functional role in DNA repair. Indeed, it has been reported that Brca1-deficient mouse embryonic stem cells display a defect in transcription-coupled repair of oxidative DNA damage. In addition, BRCA1 has been identified as a substrate of ATM, hCds1, and the ATM-related kinase, ATR (ataxia telangiectasia). These are serine/threonine kinases participating in DNA-damage checkpoint control, which indicates that BRCA1 might be activated by phosphorylation after DNA damage to allow cells to undergo DNA repair before proceeding to mitosis (reviewed in Refs. 1, 2, 3).

The centrosome controls the assembly of microtubules, a process that is important for determining cell structure, polarity, and movement during interphase, and for orchestrating formation of the bipolar spindle during mitosis (4). We have shown that BRCA1 protein is associated with the centrosome during mitosis and that a hypophosphorylated isoform of BRCA1 is associated with an essential component of the centrosome, γ-tubulin, which is responsible for microtubule nucleation and mitotic spindle formation (5). These observations suggest that BRCA1 might play a functional role with the centrosome. Truncation of Brca1 exon 11 results in defective cell proliferation, centrosome amplification, and a defective G₂-M checkpoint and eventually leads to the accumulation of extensive chromosome abnormalities in mouse embryonic fibroblast cells (6). As described in this report, we have identified a γ-tubulin-binding domain in BRCA1 protein, which is encoded by part of exon 11. Overexpression of this γ-tubulin-binding domain results in phenotypic changes similar to those observed in Brca1 exon 11 isoform-deficient cells (6).

A human γ-tubulin cDNA clone (lafmid γ-tubulin) was purchased from American Type Culture Collection (ATCC 365467) and sequenced. Compared with GenBank human γ-tubulin cDNA (locus HUMGTUB, GI:183702), there are four base changes that are either silent (no amino acid changes) or result in conserved amino acid changes and are, therefore, likely to be polymorphisms. In fact, the predicted amino acid sequences are identical to that of tubulin γ1, clone MGC:1593 (GI:12653672). This cDNA clone was then used as a template for generating γ-tubulin cDNA by PCR. γ-Tubulin cDNA, generated using primers 5′-ggg gaa ttc atg ccg agg gaa atc atc acc-3′ and 5′-aaa gcg gcc gca ctg ctc ctg ggt gcc cca-3′, was inserted into pCITE-4a(+) (Novagen, Madison, WI). The resulting pCITE-4a(+)-γ-tubulin expresses a γ-tubulin protein with an NH₂-terminal S-tag. γ-TubulinGFP³ was constructed by inserting the PCR product from primers 5′-tat aag ctt ctg ctc ctg ggt gcc cca gga-3′ and 5′-tat ctc gag gcc acc atg ccg agg gaa atc atc acc-3′ into a modified pEGFP-N1 (Clontech), pEGFP-N1-KS(−), in which the Kozak sequence and ATG start codon of the GFP gene were deleted to prevent leaky expression of GFP.

A full-length BRCA1 cDNA clone was kindly provided by Dr. J. Holt. A hemagglutinin tag was inserted to generate pScriptHABRCA1. pScriptHABRCA1mutant, a COOH-terminal truncation mutant lacking the COOH-terminal 11 amino acid residues was constructed by replacing a NcoI-HindIII fragment of pScriptHABRCA1 with NcoI- and HindIII-digested PCR product from primers 5′-ctt caa cag aaa ggg tca ac-3′ and 5′-tgc caa gct ttc agg tgt cca gct cct ggc act-3′. GST-BF3del1 and GST-BF3del2 were constructed using the vector pGEX-2TK (Pharmacia). Inserts were generated by PCR using primer pairs 5′-ggg gga tcc tca ggc ctt cat cct gag gat-3′, 5′-ggg aat tcc act cac aca ttt att tgg ttc-3′ and 5′-ggg gga tcc gcg ctt gaa cta gta gtc agt-3′, 5′-ggg aat tcc act cac aca ttt att tgg ttc-3′, respectively. pcDNABF3myc was constructed by inserting the PCR product from primers 5′-ggg gaa ttc gcc acc atg cgt aaa agg aga cct aca tca-3′ and 5′-ggg aag ctt act cac aca ttt att tgg ttc-3′ into pcDNA3.1(−)/Myc-His A (pcDNAmyc; Invitrogen, San Diego, CA). pBRCA1GFP was constructed by subcloning full-length BRCA1 cDNA without the stop codon into pEGFP-N1-KS(−). All of the inserts generated by PCR were validated by DNA sequencing.

BRCA1, mutant BRCA1, and S-tagged γ-tubulin proteins were obtained by TNT reactions (Promega, WI) using pScriptHABRCA1, pScriptHABRCA1mutant, and pCITE-4a(+)-γ-tubulin, respectively, as the templates. A TNT reaction from pCITE-4a(+) was used as a negative control for pCITE-4a(+)-γ-tubulin. Ten μl of in vitro translated wild-type or mutant BRCA1 was incubated with 10 μl of in vitro translated S-tag control or S-tagged γ-tubulin for 30 min at room temperature. One hundred μl of binding buffer [20 mm Tris-HCl (pH 7.5), 150 mm NaCl, and 0.1% Triton X-100) with protease inhibitors (20 μg/ml aprotinin, 10 μg/ml leupeptin, 1 μg/ml pepstatin A, and 1 mm phenylmethylsulfonyl fluoride) and 30 μl of 50% S-protein agarose (Novagen) were added. The mixture was incubated for 2 h at 4°C and then was washed three times with the binding buffer. Samples were subjected to SDS-PAGE, followed by Western analysis (5) using BRCA1 antibody MS110 (1:100 dilution; Calbiochem) or γ-tubulin antibody (1:10,000, GTU-88; Sigma Chemical Co.).

The GST-BRCA1 constructs GST-BRCA1 1 to 6 were kindly provided By Drs. J. Chen and D. Livingston (7, 8). These vectors were introduced into Escherichia coli DH5α or BL21 (Novagen). GST-fusion proteins were purified using GSH-Sepharose beads (Pharmacia) according to the manufacturer’s instructions. Approximately 5 μg of GST or GST-fusion protein and 5 μl of in vitro translated S-tagged γ-tubulin were diluted to 90 μl with L buffer (PBS with 0.1% NP40 and 0.1% Triton X-100) containing protease inhibitors (EDTA-free protease inhibitor cocktail; Boehringer Mannheim). Ten μl of 10× binding buffer [250 mm Tris-HCl (pH 7.5), 500 mm NaCl, 10 mm DTT, 1% NP40, and 1 mg/ml BSA] and 20 μl of GSH beads were added. The mixture was incubated for 2 h at 4°C, washed three times with L buffer, and subjected to SDS-PAGE. Western analysis was performed using γ-tubulin antibody (GTU-88; Sigma Chemical Co.) to detect γ-tubulin interacted with GST-BRCA1 fusion proteins.

COS-7 cells were cultured in low-glucose DMEM supplemented with 10% fetal bovine serum. Transfection was performed using the FuGENE 6 Transfection reagent (Boehringer Mannheim) or SuperFect Transfection reagent (Qiagen). Twenty-four to 48 h after transfection, cells were harvested for IP and Western analysis. Stable transformants of pcDNABF3myc were obtained by selection with 500 μg/ml G418.

Transfected cells were harvested by trypsinization, lysed in L buffer with protease inhibitors (Boehringer Mannheim), and sonicated briefly at 4°C. Cell lysates cleared by centrifugation were incubated with 7.5 μl of anti-myc antibody (clone 9E10; Santa Cruz Biotechnology), 15 μl BRCA1 D-20 antibody (Santa Cruz Biotechnology), or 2 μl of γ-tubulin antibody (Sigma Chemical Co.) for 1 h on ice. Protein G or protein A was added and the incubation was continued for 2 h at 4°C. Samples were washed four times with L buffer and separated by SDS-PAGE. Western hybridization was performed as described previously (5). Antibodies used for Western analysis were myc (1:250; Santa Cruz Biotechnology), γ-tubulin (1:10,000; Sigma Chemical Co.), GFP (1:1000; Clontech), BRCA1 MS110 (1:100; Calbiochem), BRCA1 Ab-2 (1:250; NeoMarkers), Caspase-3 (1:1000; PharMingen), and p21 (1:100; Calbiochem). Signals were detected by enhanced chemiluminescence. Quantitation was performed using the LumiImager software (Roche, NJ).

Cells were plated at 2.5 × 10⁴ per 35-mm dish in triplicates and fed every 3 days. At different time periods, cells were trypsinized and counted using a Coulter counter.

Exponentially growing cells were spun to glass slides using a cytospin centrifuge. TUNEL assay was performed according to the manufacturer’s instructions (in situ cell death detection kit, fluorescein; Boehringer Mannheim). Apoptotic cells were labeled with green fluorescence.

Exponentially growing cells were spun to glass slides using a cytospin centrifuge at 1 × 10⁵ cells per spin. Cells on slides were fixed in 2% neutral paraformaldehyde or cold methanol, followed by immunofluorescence staining as described previously (5). Primary antibodies used were mouse monoclonal α-tubulin (ICN) and rabbit polyclonal pericentrin (Babco) antibodies at 1:1000 dilution each. Secondary antibodies were FITC-conjugated goat anti-mouse IgG (Southern Biotechnology Associates) and Texas Red-conjugated goat anti-rabbit IgG (Accurate Scientific), diluted 1:200 each. Mitotic cells with two centrosomes (normal) or more than two centrosomes (abnormal) were scored. To ensure unbiased cell counting, slides were coded before examination by L. C. H. and T. P. D. One hundred to 500 mitotic cells were counted each time.

Exponentially growing COS-7 cells were treated for 24 h with 0–200 nm OA, an inhibitor of serine/threonine PPs types 1 and 2A. Cells were harvested by trypsinization for PP assay, Western analysis, and immunofluorescence staining. PP assay was performed according to the manufacturer’s instructions (Life Technologies, Inc.). Briefly, cells were lysed in a buffer containing 50 mm Tris-HCl (pH 7.0), 0.1% β-mercaptoethanol, 0.1 mm EGTA, 0.1 mm EDTA, 0.5% Triton X-100, and protease inhibitors. Assays were carried out in a final volume of 30 μl with 0.5 μg of cell lysate and 30 μg of [³²P]phosphorylase a in PP assay buffer containing 0.1 mm EDTA, 1 mg/ml BSA, 20 mm imidizole-HCl (pH 7.63), and 0.1% β-meracptoethanol. PP activity was measured by the release of ³²Pi from [³²P]phosphorylase a. Assays were performed in duplicates. PP 1 activity was defined as the activity insensitive to 5 nm OA. PP 2A and PP2A-like activities were defined as the activity inhibited by 5 nm OA. An aliquot of cells was lysed in RIPA buffer with protease inhibitors (5), and 100 μg of protein were fractionated on 4–12% SDS-PAGE, followed by Western blotting. BRCA1 protein was detected by BRCA1 MS110 antibody. For immunostaining, cells were spun to glass slides by cytospin, fixed in methanol for 5 min at −20°C and costained with mouse monoclonal α-tubulin and rabbit polyclonal pericentrin antibodies or mouse monoclonal γ-tubulin and rabbit polyclonal BRCA1 C-20 (Santa Cruz Biotechnology) antibodies.

We have shown that BRCA1 is associated with mitotic centrosomes by using immunofluorescence microscopy and analysis of isolated centrosomes. Coimmunoprecipitation also indicates an interaction between a hypophosphorylated isoform of BRCA1 and γ-tubulin, an essential component of the centrosome (5). To look for an in vitro interaction between BRCA1 and γ-tubulin, in vitro translated full-length BRCA1 protein and S-tagged γ-tubulin were assayed by S-protein pull-down assay. S-tagged γ-tubulin as well as the full-length BRCA1 was brought down by S-protein agarose (Fig. 1 A), which confirmed an interaction between hypophosphorylated BRCA1 (the isoform generated by TNT) and γ-tubulin.

Many BRCA1 mutations result in COOH-terminal truncation, the smallest of which is a deletion of the COOH-terminal 11 amino acid residues (9). This mutant BRCA1 protein loses transcriptional activation activity (10, 11) and the ability to associate with RNA polymerase II holoenzyme (12) but continues to interact with BRCA2 (8). The mutant protein generated by TNT retained the ability to bind γ-tubulin (Fig. 1 A), which indicated that the COOH-terminal region, although required for the transcriptional activation activity of BRCA1, is not required for association with γ-tubulin.

We then used GST pull-down assay to identify the γ-tubulin-binding domain in BRCA1. GST-BRCA1 fragments 1–6 (GST-BF1–6) are overlapping BRCA1 fragments that cover the entire BRCA1 open reading frame (Fig. 1,B; Refs. 7, 8). As shown in Fig. 1,C, only GST-BF3, which contains amino acids 504–803 of BRCA1 protein, interacted with γ-tubulin, which demonstrated that BF3 contains a γ-tubulin-binding domain. It has been reported that Rad50 (13) and ZBRK1 (14) interact with a BRCA1 fragment containing amino acids 341–748, which is partly overlapping with BF3. However, neither protein binds to a fragment containing amino acids 513–914 of BRCA1 (13, 14). Therefore, it is unlikely that BF3 contains a binding site for Rad50 or ZBRK1. BF3 does contain two NLSs: NLS1 (amino acids 503–508) and NLS2 (amino acids 606–615). NLS1 has been shown to be essential for nuclear localization of BRCA1 (15, 16). GST-BF3del1 (amino acids 510–803), which does not contain NLS1, still interacted with γ-tubulin (Fig. 1,D). However, deletion of a region covering both NLS1 and NLS2 (GST-BF3del2; amino acids 622–803) abolished the interaction with γ-tubulin (Fig. 1 D). These results suggest that the region between amino acids 510 and 622 is required for the interaction with γ-tubulin, and nuclear localization might not be important for this interaction.

To determine whether there was an association between BF3 and γ-tubulin in vivo, COS-7 cells were cotransfected with pcDNABF3myc and γ-tubulin GFP, followed by IP with an anti-myc antibody. Western analysis with an anti-γ-tubulin antibody detected both γ-tubulin-GFP fusion protein and endogenous γ-tubulin. Anti-myc or BRCA1 Ab-2 antibody (a rabbit polyclonal antibody raised against amino acids 768–793 of BRCA1; NeoMarkers) detected myc-tagged BF3. The γ-tubulin-GFP fusion protein (M_r ∼75,000; also detected by GFP antibody) coimmunoprecipitated with myc-tagged BF3 (M_r ∼40,000). In addition, endogenous γ-tubulin, which migrated very close to the mouse immunoglobulin, also coimmunoprecipitated with myc-tagged BF3 but to a lesser extent (Fig. 2 A). These results indicate that BF3 also interacts with γ-tubulin in vivo.

We next tested whether overexpression of BF3 in COS-7 cells could interfere with the interaction between full-length BRCA1 and γ-tubulin. BRCA1 IP of lysate from cells transfected with γ-tubulinGFP and pcDNAmyc brought down BRCA1 and γ-tubulin-GFP fusion protein, as well as endogenous γ-tubulin (Fig. 2,B, Lane 1). In contrast, γ-tubulin-GFP fusion protein and γ-tubulin coimmunoprecipitated by BRCA1 antibody was substantially decreased in a lysate from cells transfected with γ-tubulinGFP and pcDNABF3myc (Fig. 2,B, Lane 2). γ-tubulin IP was performed using lysate of cells transfected with pBRCA1GFP, γ-tubulinGFP, and pcDNAmyc (Fig. 2,B, Lane 3) or pBRCA1GFP, γ-tubulinGFP, and pcDNABF3myc (Fig. 2,B, Lane 4). BRCA1-GFP fusion protein coimmunoprecipitated by γ-tubulin antibody (Fig. 2,B, Lane 3) was also diminished in the presence of BF3 (Fig. 2 B, Lane 4). The expression of BF3 is indicated in Western analysis of WCL. These results demonstrate that BF3 interferes with the association between BRCA1 and γ-tubulin and may have a dominant-negative effect on biological function(s) of BRCA1.

When transiently transfected with the BF3 expression vector, most cells expressing high levels of BF3 were arrested in interphase. To determine the effect of BF3 expression in mitotic cells, we selected stable transformants, which might express the protein at a lower level and thus continue through the cell cycle. pcDNABF3myc was introduced into COS-7 cells, and G418-resistant cell clones were selected. Four COS-7 cell clones expressing BF3 at various levels as detected by Western analysis were characterized further. Clone 3 did not express detectable BF3, although it was selected from cells transfected with pcDNABF3myc. Clone 10 expressed a low level of BF3. Clones 5 and 11 expressed high levels of BF3 (by ∼17-fold that of clone 10; Fig. 3 A), although the expression levels of BF3 in clones 5 and 11 were still considerably lower than those in transiently transfected cells.

Knockout experiments have revealed that Brca1 is required for cell proliferation and that Brca1-deficient cells exhibit excessive cell death (reviewed in Refs. 1, 2, 3). Fig. 3,B shows the growth curves of clones 3, 5, 10, and 11. Clones 5 and 11 exhibited lower proliferation rates compared with clone 3 and 10. This result suggested that the higher levels of BF3 in clones 5 and 11 might interfere with cell proliferation. Furthermore, clones 5 and 11 had more cells undergoing apoptosis. An example of TUNEL assay of clones 3, 10, and 11 is shown in Fig. 3,C. There were more apoptotic cells in clone 11. This result was confirmed by Western analysis of caspase-3 (Fig. 3,D). In apoptotic cells, the M_r 32,000 pro-caspase-3 is cleaved into smaller M_r 17,000 and M_r 11,000 subunits of the active caspase-3. The caspase-3 antibody (PharMingen) recognizes both the M_r 32,000 and M_r 17,000, but not the M_r 11,000 protein. Western analysis indicated that the level of M_r 17,000 caspase-3 subunit in clone 11 was 2-fold higher than that detected in clones 3 and 10. In addition, Western analysis of these clones also indicated that clone 11 expressed a higher level of p21 protein, which suggested that growth suppression and the induction of apoptosis in COS-7 cells expressing high levels of BF3 might be p21 dependent (Fig. 3 D).

If BRCA1 through its binding to γ-tubulin plays a role in the regulation of centrosome duplication, mitotic spindle formation, and proper segregation of chromosomes during mitosis, the overexpression of BF3 might decrease the association between BRCA1 and γ-tubulin as shown in Fig. 2,B and might induce aberrant centrosome duplication and mitotic abnormalities. We stained the four BF3 cell clones with α-tubulin and pericentrin antibodies, which reveal microtubules and centrosomes respectively, and scored the percentage of mitotic cells with more than two centrosomes and abnormal mitotic spindles. Fig. 3,E shows examples of normal bipolar mitotic cells (left panel) and mitotic cells with multiple centrosomes and abnormal mitotic spindles (middle and right panels). Cells with multiple centrosomes may result in abnormal chromosome segregation and accumulation of aneuploid cells. As shown in the right panel of Fig. 3,E, a telophase cell was divided into three aneuploid daughter cells. Clones 3 and 10, which did not express BF3 or expressed low levels of BF3, showed ∼20% and 22.4% mitotic abnormalities, similar to the parental COS-7 cells. Clones 5 and 11, which expressed BF3 at high levels, showed 36.1% and 39.8% mitotic abnormalities, respectively (Fig. 3,F). The data of clone 5 might not be as significant as that of clone 11 because of a large SD of the percentage of abnormal mitotic cells (Fig. 3 F). However, both clones 5 and 11 showed the same trend. Thus, overexpression of BF3 doubled the proportion of mitotic cells with multiple centrosomes and abnormal spindles in COS-7 cells.

The expression and phosphorylation of BRCA1 protein is cell-cycle regulated. BRCA1 becomes hyperphosphorylated after DNA damage and may play a role in DNA repair (1, 2, 3). γ-Tubulin interacts preferentially with hypophosphorylated BRCA1 at mitotic centrosomes (5). To further investigate the connection between the phosphorylation state of BRCA1 and its function, we tested whether OA, an inhibitor of PP1 and PP2A, interfered with the association of BRCA1 with mitotic centrosomes. PP assay indicated that OA in the range of 2–200 nm inhibited PP2A or PP2A-like activity, and might also inhibit PP1 at high dose, although to a lesser extent (Fig. 4,A). As shown in Fig. 4,B, with the increase of OA from 20 to 200 nm, the mobility of BRCA1 became slower, an indication of hyperphosphorylation. A dramatic decrease of BRCA1 protein levels was observed at 100–200 nm OA, whereas γ-tubulin levels showed only a slight decrease that might be attributable to loading variations (Fig. 4,B), which indicated that treatment with high doses of OA leads to a loss of BRCA1 but not to a general loss of protein. Immunofluorescence staining showed that OA treatment induced a decrease of both BRCA1 and γ-tubulin at mitotic centrosomes at concentrations ≥20 nm (Fig. 4,C) and an accumulation of multipolar mitotic cells as reported previously (Ref. 17; Fig. 4,D). OA impacts many proteins that are involved with the mitotic spindle and induces aberrant mitotic cells. Interestingly, 20 nm OA was the starting concentration that induced hyperphosphorylation of BRCA1 (Fig. 4,B), a decrease of both BRCA1 and γ-tubulin associated with mitotic centrosomes (Fig. 4,C), and accumulation of abnormal mitotic cells with multipolar spindles (Fig. 4,D). OA inhibited the dephosphorylation of BRCA1 protein, which resulted in an accumulation of the hyperphosphorylated isoform and reduced the association of BRCA1 and γ-tubulin with mitotic centrosomes, and thus may have contributed to the induction of abnormal mitotic cells. This is consistent with the results shown above that overexpression of BF3 interfered with the interaction of BRCA1 and γ-tubulin (Fig. 2 B).

Accumulation of chromosome aberrations is one of the hallmarks of cancer cells. It has been reported that breast tumors carrying BRCA1 mutations accumulate chromosomal abnormalities (18). Centrosome defects have been observed in several cancer types including breast cancer (19). In addition, centrosomes of high-grade human breast tumors usually display abnormal structure and function (20). Proper DNA repair after damage and the maintenance of fidelity during cell division may equally contribute to the prevention of cancer development. Numerous reports have demonstrated that BRCA1 plays an important role in DNA repair (1, 2, 3). We reported previously that endogenous BRCA1 protein was associated with the centrosome during mitosis (5). Here we have identified the γ-tubulin-binding domain in BRCA1 (BF3; amino acids 504–803) and have demonstrated that BF3 coimmunoprecipitates with γ-tubulin and competes with BRCA1 for the interaction with γ-tubulin. Overexpression of this domain results in slower growth, an increase in apoptosis, and accumulation of mitotic cells with multiple centrosomes and abnormal mitotic spindles.

When pcDNABF3myc was introduced into 184A1, a normal human breast epithelial cell line, BF3 expression was detected in transiently transfected cells. However, no stable transformants selected by G418 expressed BF3 as examined by either Western analysis or immunofluorescence staining (data not shown), which indicated that overexpression of BF3 might be lethal to 184A1 cells. Deficiency of p53 or p21 partially rescues the lethal phenotype observed in the Brca1 knockout mouse embryo (1, 2, 3). BF3 might have a dominant-negative effect and might be lethal to cells carrying the wild-type p53, such as 184A1 cells (21). This may explain why we were not able to select stable transformants from 184A1 cells. COS-7 cells, on the other hand, express SV40 T antigen, which inactivates p53, and, therefore, survive in the presence of BF3. However, cells that expressed high levels of BF3 tended to grow slower and undergo apoptosis, which might be p21 dependent.

OA affects the phosphorylation state of many proteins that might contribute to the induction of aberrant mitotic cells. We found a good correlation between the induction of hyperphosphorylation of BRCA1 (Fig. 4,B) and a decrease of both BRCA1 and γ-tubulin associated with mitotic centrosomes (Fig. 4,C) by OA treatment. Because γ-tubulin interacts with hypophosphorylated BRCA1, OA treatment reduces the binding of BRCA1 to γ-tubulin and their association with mitotic centrosomes, and might consequently contribute to the formation of abnormal spindles (Fig. 4 D). This is consistent with the hypothesis that overexpression of BF3 might cause the accumulation of abnormal mitotic cells by competing with BRCA1 for the binding to γ-tubulin and by decreasing BRCA1 association with mitotic centrosomes.

Interestingly, mouse embryonic fibroblasts with a homozygous deletion of Brca1 exon 11 exhibited defective cell proliferation, centrosome duplication, and G₂-M checkpoint (6). BF3 is encoded by exon 11, which suggests that the absence of BF3 might contribute to phenotypic changes observed in the Brca1 exon 11-deleted cells (6).

BF3 does not contain any identified domain interacting with proteins involved in DNA repair or transcriptional regulation except STAT1. Ouchi et al. has reported that BRCA1 interacts with STAT1 through BF3 and has a specific synergistic effect on transcriptional induction of p21 by IFN-γ (22). Overexpression of BF3 might, therefore, have a dominant-negative effect and interfere with transcriptional activation of p21 and the growth inhibition induced by IFN-γ and STAT1. The phenotypic changes attributable to overexpression of BF3, such as growth suppression and induction of apoptosis, which correlates with the induction of p21 (Fig. 3), however, seem not easily understood under this hypothesis.

It has been demonstrated that BRCA1 binds DNA by using a gel mobility shift assay (23). Amino acids 452-1079 of BRCA1 showed the highest binding affinity to DNA. Amino acids 504–802 (corresponding to BF3) also bound DNA, but to a lesser extent. How the DNA binding activity, without DNA sequence specificity detected in vitro, fits into DNA repair and gene transcription in vivo is still an open question. At this point, we cannot exclude the possibility that the phenotypic changes, which we observed in COS-7 cell clones expressing BF3, is partly caused by interfering with the DNA binding activity of BRCA1. Nevertheless, our data strongly support the observation that reducing the interaction between BRCA1 and γ-tubulin and decreasing their association with mitotic centrosomes might be sufficient to induce significant phenotypic changes and might result in apoptosis or accumulation of aneuploid cell population.

Taken together, results presented here suggest that BRCA1 plays a functional role at mitotic centrosomes, perhaps through the regulation of centrosome duplication, mitotic spindle formation, and proper segregation of chromosomes during mitosis, and helps to maintain the fidelity of cell division and to preserve genomic stability.

We thank Dr. Jeff Holt at Vanderbilt University (Nashville, TN) for providing the BRCA1 cDNA, Drs. Junjie Chen and David Livingston at Dana-Farber Cancer Institute (Boston, MA) for providing GST-BRCA1 constructs, and Drs. David Virshup and Don Ayer for useful discussion and comments. We also thank Dr. Jun Liu for the GST pull-down assay protocol, and Diana Lim for help with the graphics.