BRCA1 function is necessary for maintenance of centrosome number in mammary cell lines. Evidence from many tumors reveals that cancer-causing mutations can occur in the carboxy terminus of BRCA1. For example, truncation of the last 11 amino acids can render BRCA1 nonfunctional as a tumor suppressor (
38), suggesting that this domain of BRCA1 is vital for its tumor suppressor function. In a previous study, we had developed a strategy to inhibit the function of the BRCA1 carboxy terminus by expressing a peptide that specifically binds to amino acids 1650 to 1800 of BRCA1. That peptide, termed BIF in this study, is a fragment of RNA helicase A (RHA), a protein that mediates the interaction of BRCA1 and RNA polymerase II holoenzyme (
1). BIF contains amino acid residues 89 to 344 of RHA and an epitope tag; it lacks other functional domains of RHA important for transcription. Expression of BIF in the mammary cell line MCF10A and in the osteosarcoma cell line U2OS was shown to interfere with BRCA1 localization to nuclear foci after DNA damage (
36). This observation supports the notion that BIF expression inhibits a known function of BRCA1. Interestingly, in the previous study, when the two cell lines analyzed were compared, expression of BIF caused a centrosome amplification phenotype in the mammary cell line but not in the osteosarcoma cell line. Since centrosome amplification was also observed in the BRCA1-mutant breast cancer cell line HCC1937 (
36), it was suggested that this centrosome amplification was in fact caused by disruption of BRCA1 function. In this study we tested whether other cell lines derived from breast tissue also show centrosome amplification after inhibition of BRCA1.
In order to express BIF in multiple cell lines, we utilized a recombinant adenovirus (adBIF). Cells were infected with adBIF and compared to cells that were infected at the same multiplicity of infection with recombinant adenoviruses expressing either GFP or LacZ. All nine cell lines derived from both breast and other cell types that were chosen had a low baseline of centrosome anomalies. Each cell line was infected with adBIF, and differences in BIF expression between cell types were normalized by assessing protein levels of the BIF polypeptide by Western blot analysis (data not shown). After 24 to 27 h, each infected cell line was fixed and stained with antibodies specific for the centrosome protein γ-tubulin. The number of cells with abnormal numbers of centrosomes (n > 2) was scored by fluorescence microscopy. The results in Fig. show that only the mammary tissue-derived cell lines expressing BIF exhibited a threefold increase in cells with amplified centrosome numbers (four out of five tested). Cell lines derived from other cell types—prostate (PC3), cervix (HeLa), colon (DLD-1), and osteosarcoma (U2OS)-did not have amplified centrosomes caused by expression of BIF (none of four tested). These results reveal a striking correlation: a BRCA1 function disrupted by BIF is important for maintenance of centrosome number in breast cells, but this BRCA1 function is not important in other cell types.
The cell lines used represented a spectrum of phenotypes regarding cancer and p53 status. The MCF10A and MTSV1-7 lines are nontumorigenic and are thought to reflect normal mammary epithelium. The ZR75 and T47D cells are from breast cancer lines, and they are estrogen receptor positive. The Hs578T breast cancer cell line is estrogen-receptor negative. Thus, the effects of BIF expression on centrosome number were obtained in all of these breast-derived cell lines, except for ZR75, regardless of cancer status. Interestingly, p53 status does not correlate with BIF-induced centrosome amplification, either. Among the mammary tissue-derived cell lines, only MCF10A is wild type for p53. The other cell lines, MTSV1-7, T47D, Hs578T, and ZR75, are all negative for p53. MTSV1-7 cells express simian virus 40 T antigen (
3), and the others harbor p53 mutations (
4,
19,
24,
30). Among the nonbreast cell lines, U2OS cells have wild-type p53, whereas HeLa, PC3, and DLD-1 cells either express proteins that degrade p53 or have mutant p53 (
2,
7,
26,
34,
35). Determining whether the p53 status of these cell lines correlates with the centrosome amplification phenotype is important, since p53 plays a role in centrosome number regulation, and p53 is known to interact with BRCA1 pathways (
13,
47). It should be noted that several cell lines with mutant p53 could not be used in this study because they had centrosome amplification independent of BIF expression.
In analysis of the centrosome number in cells by use of an antibody specific to γ-tubulin, it is possible that fragmentation, rather than replication, of centrosomes was scored as extra centrosomes. In order to test this possibility, the experiment for which results are shown in Fig. was repeated, but the cells were stained with an antibody specific to centrin. Centrin is located at each centriole, and by immunofluorescence each centrosome should have two pinpoints of centrin. Normal cells with one or two centrosomes would have two or four points of centrin in pairs. It was found that Hs578T cells expressing the control protein had no examples of cells with more than four centrin-positive spots in two pairs. By contrast, expression of BIF in these cells resulted in a rapid increase in the number of cells with more than four centrin-positive spots (9%), and these were frequently found not to be in pairs (Fig. ). These results suggest that both the proper regulation of centrosome replication and the proper maintenance of the centrosome structure require BRCA1 function in this cell line.
There is supporting evidence that the effects of BIF expression on centrosome number reflected an inhibition of BRCA1. BIF interferes with BRCA1 localization after DNA damage, and cells that lack wild-type BRCA1 (HCC1937) show a similar centrosome amplification phenotype (
36). Nevertheless, it is possible that BIF alone, and not its interaction with BRCA1, caused centrosome amplification. To use a third method to test whether BRCA1 is indeed necessary for proper centrosome maintenance in mammary cells, siRNA oligonucleotides were designed to decrease the expression of BRCA1. The Western blot in Fig. shows that the two different siRNA oligonucleotide pairs directed against BRCA1, but not a control siRNA, significantly reduced the amount of BRCA1 protein in transfected Hs578T and U2OS cells. To determine if a reduction in BRCA1 levels by siRNA caused amplified centrosomes, the number of cells with abnormal centrosome numbers was scored by immunofluorescence using a γ-tubulin-specific antibody. The results summarized in Fig. show that Hs578T mammary cells that have reduced BRCA1 levels have a 5- to 10-fold increase in the fraction of cells with more than two centrosomes. Figure shows examples of mitotic Hs578T cells with amplified centrosomes. While the expression of BRCA1 was reduced to the same level in U2OS cells, centrosomes from this cell line were not amplified. This suggests that the centrosome amplification phenotype caused by siRNA knockdown of BRCA1 followed the same mammary cell-specific trend observed for expression of BIF. This also suggests that the function of BRCA1 at the centrosome is inhibited by BIF binding.
HU abolishes centrosome amplification in cells expressing BIF. As an initial characterization of the centrosome amplification in these mammary epithelial cell lines expressing BIF, we asked whether passage through the cell cycle is important for the increase in centrosome number. It is known that centrosome duplication in mammalian cells is dependent on the activity of cyclin A (
27). Thus, passage through S phase of the cell cycle would be critical to duplicate the centrosomes if the BIF-dependent amplification was utilizing the normal cellular machinery for this duplication process. This was tested directly by treating cells with hydroxyurea (HU), which blocks cells early in S phase, and assaying centrosome amplification. MCF10A cells were infected at time zero with adBIF or a control, and HU was added after 3 h. The centrosome number was scored by immunofluorescence for γ-tubulin at 12, 18, and 24 h postinfection (hpi). In the absence of HU, amplification of centrosomes in these cells was apparent after 18 h. HU blocked the amplification of centrosomes when added to the medium after adBIF infection (Fig. ). These results suggest that treatment of cells with HU halted the cell cycle upstream of the point at which BRCA1 affected centrosome replication.
The BRCA1/BARD1 heterodimer ubiquitinates centrosome proteins in vitro. Our data clearly indicate that BRCA1 executes a major role in the maintenance of centrosome number in mammary cells (Fig. and ). BRCA1 has also been shown to localize to centrosomes during mitosis and to bind to γ-tubulin by coimmunoprecipitation (
16). BRCA1, together with BARD1, forms a heterodimer associating via the respective RING domains and adjacent α-helices of each protein, and the BRCA1/BARD1 heterodimer is an active ubiquitin ligase (
6,
15). We hypothesized that ubiquitination of centrosome-associated proteins by BRCA1 plays a role in centrosome regulation. To this end, centrosome fractions were prepared from MCF10A cells in order to identify targets of BRCA1/BARD1 ubiquitination in vitro. We tested whether BRCA1/BARD1 ubiquitinated any of the centrosome proteins by carrying out in vitro reactions that included copurified, full-length BRCA1/BARD1, purified E1, purified E2 UbcH5c, and biotinylated ubiquitin. Reaction products were resolved on protein gels and transferred to polyvinylidene difluoride membranes, and ubiquitinated species were identified by using streptavidin conjugated to HRP. In Fig. , lane 2, autoubiquitination of BRCA1 (>220 kDa) is apparent along with ubiquitination of a 110-kDa polypeptide, which may be BARD1. In the presence of centrosomes, E1, UbcH5c, and ubiquitin but no BRCA1/BARD1, no ubiquitinated species were apparent (lane 3). Addition of BRCA1/BARD1 to these reaction mixtures resulted in four or more new ubiquitinated products (lane 4). Comparison with the total polypeptide composition of the centrosome preparation (data not shown) reveals that only a subset of the centrosome-associated proteins was modified by ubiquitin.
One of the ubiquitinated polypeptides had a migration consistent with monoubiquitinated γ-tubulin (48 kDa unmodified plus 8 kDa of ubiquitin). To determine whether the ubiquitinated species at 56 kDa (Fig. , lane 4) was γ-tubulin, the same ubiquitination reaction, including the centrosome fraction, BRCA1/BARD1, E1, UbcH5c, and biotinylated ubiquitin, was analyzed for modification of γ-tubulin. Monoclonal antibodies to γ-tubulin were used to detect unmodified γ-tubulin at 48 kDa in lanes containing centrosomes (Fig. , lanes 3 and 4), and in the presence of BRCA1/BARD1, a shifted polypeptide consistent with monoubiquitinated γ-tubulin was apparent (lane 4). To test whether the reaction was specific for γ-tubulin, we also analyzed reaction products for ubiquitination of the related α- and
-tubulin polypeptides, and they were negative for BRCA1/BARD1-dependent ubiquitination under these conditions (Fig. ). These results suggest that γ-tubulin is a specific target of BRCA1/BARD1 for ubiquitination.
In order to define the specificity of the BRCA1/BARD1 ubiquitination reactions for γ-tubulin, we tested as substrates γ-tubulin and centrin-1, which had been expressed and purified from bacteria. Equal amounts of recombinant protein were included in reaction mixtures containing BRCA1/BARD1, E1, UbcH5c, and ubiquitin. Dependent on the presence of BRCA1/BARD1, the γ-tubulin band was shifted by 8 and 16 kDa, consistent with ubiquitination (Fig. ). By contrast, inclusion of BRCA1/BARD1 along with the other ubiquitination factors showed no modification of the centrin protein (Fig. ). This result implies that BRCA1/BARD1 specifically modifies γ-tubulin, either as a purified protein or in association with the centrosome. Interestingly, purified γ-tubulin was modified by one or two ubiquitin moieties, whereas centrosome-associated γ-tubulin was only modified by one ubiquitin molecule. It is possible that the recombinant protein has a second site that is monoubiquitinated, whereas the γ-tubulin in the centrosome would have the second site blocked.
In order to test whether the slower-migrating band in reactions using centrosome fractions was in fact a ubiquitin modification, the requirement for ubiquitination factors was tested. In the complete reaction, a band at 56 kDa was detected by using an antibody specific for γ-tubulin (Fig. , lane 1). Omission of any single ubiquitination protein from the reaction resulted in a loss of the 56-kDa γ-tubulin species (Fig. , lanes 2 to 5). Thus, the modification of γ-tubulin by BRCA1/BARD1 was also dependent on the presence of E1, UbcH5c, and ubiquitin.
To ensure that the 56-kDa band detected by Western blotting with anti-γ-tubulin was indeed γ-tubulin plus biotinylated ubiquitin, we utilized magnetic streptavidin beads to purify ubiquitinated proteins. Lanes 1 and 2 of Fig. show the products of the standard ubiquitination reaction before affinity purification by streptavidin beads. Following purification using streptavidin beads, the 48-kDa γ-tubulin did not bind to the matrix (lane 3). The 56-kDa band, dependent on BRCA1/BARD1 in the reaction, was purified via the biotin-ubiquitin modification (lane 4). Unmodified γ-tubulin (48 kDa) was also purified on the streptavidin beads in this reaction, although at a reduced level (lane 4). The presence of the 48-kDa γ-tubulin was anticipated because γ-tubulin is present as a multimer in large multisubunit complexes in centrosomes (γ-tubulin ring complex [γ-TuRC]). An inference from the presence of the unmodified γ-tubulin associated with the monoubiquitinated γ-tubulin is that the modification does not disrupt the γ-TuRC complex.
BRCA1/BARD1 monoubiquitinates γ-tubulin from centrosomes in vitro. Polymerization of ubiquitin into chains requires isopeptide linkages between the activated carboxy terminus of an incoming ubiquitin molecule and a lysine side chain of the protein substrate or the previously conjugated ubiquitin. Isopeptide linkages via lysine 48 of four or more ubiquitin molecules constitute a signal for proteasome-mediated degradation of the substrate protein. In order to facilitate easy detection of ubiquitinated proteins, the ubiquitin used in Fig. and Fig. was biotinylated via the amine group in the lysine side chains of ubiquitin. The ubiquitin, biotinylated via lysines, may inefficiently polymerize into multiubiquitin chains. In order to address this issue, we compared reactions that contained various ubiquitin molecules, including purified bovine ubiquitin, histidine-tagged-ubiquitin, and histidine-tagged-ubiquitin containing a mutation at lysine 48. In reactions in which any of the purified ubiquitin proteins was included, a 56-kDa band was observed, but ubiquitin polymers were not detected on γ-tubulin regardless of the type of ubiquitin used (Fig. ). This suggested that the detection of monoubiquitinated γ-tubulin was not due solely to a modification of ubiquitin. Increasing the amount of E2 factor or ubiquitin also resulted in monoubiquitination (data not shown), suggesting that the lack of a multiubiquitin chain was not due to limiting components in the reaction. It is possible that other factors not present in the reaction would cause the multiubiquitination of γ-tubulin. Since the proteasome recognizes ubiquitin chains of four units or longer, linked through lysine 48, these data suggest that γ-tubulin ubiquitination by BRCA1/BARD1 is not a signal for proteasome-mediated degradation. The pathway triggered by BRCA1 ubiquitination is currently unclear. In this study, centrosomal γ-tubulin is modified with a single ubiquitin molecule, and in other studies BRCA1/BARD1 polymerizes ubiquitin through a novel lysine 6 linkage, which has unknown consequences for the cell (
44). It is possible that proteins missing from the in vitro assays described here could facilitate polyubiquitination via lysine 6 of BRCA1-dependent target proteins such as γ-tubulin.
Function of the carboxy terminus of BRCA1 is required for ubiquitination of γ-tubulin. Although the amino-terminal 110 amino acid residues of BRCA1 contain the RING domain and are sufficient for ubiquitination (
15,
45), and although BRCA1 amino acid residues 504 to 803 bind γ-tubulin (
16), we noted that expression of the BIF peptide that binds to the BRCA1 carboxy terminus inhibited BRCA1-dependent regulation of centrosomes in cells. We tested whether BIF would inhibit the in vitro BRCA1/BARD1-dependent ubiquitination of γ-tubulin. The BIF peptide and a control peptide fragment from RHA (diagrammed in Fig. ) were included in the complete BRCA1/BARD1 ubiquitination reaction described above. While the control peptide had no effect on BRCA1/BARD1 ubiquitination of γ-tubulin, the BIF peptide powerfully inhibited the BRCA1/BARD1-dependent ubiquitination of γ-tubulin (Fig. ). This result shows that an unbound carboxy-terminal domain of BRCA1 is required for the ubiquitination reaction.
To test whether the carboxy terminus of BRCA1 is indeed necessary for ubiquitination of γ-tubulin in vitro, we purified mutants of BRCA1 truncated at the carboxy terminus as heterodimers associated with full-length BARD1. Protein preparations included full-length BRCA1 (amino acids 1 to 1863), a cancer-associated truncation of BRCA1, BRCA1(1-1852), and three other truncations, BRCA1(1-1527), BRCA1(1-1000), and BRCA1(1-500), each coexpressed and purified with full-length BARD1 protein. The BARD1 protein was untagged and was purified via association with the tagged BRCA1. These five BRCA1/BARD1 preparations were balanced for BARD1 content (Fig. ), and each was enzymatically active in a nonspecific ubiquitin polymerization assay (data not shown). Each of these heterodimers was included in a ubiquitination reaction with E1, UbcH5c, biotinylated ubiquitin, and the centrosome fraction. Only the full-length BRCA1/BARD1 complex robustly ubiquitinated γ-tubulin (Fig. , lane 2). Ubiquitination of γ-tubulin by the cancer-associated mutant that truncated the protein by only 11 amino acids, BRCA1(1-1852)/BARD1, was severely compromised (Fig. , lane 3). Trimming 336 or more amino acid residues from the carboxy terminus of BRCA1 completely abolished the ubiquitination of γ-tubulin by BRCA1/BARD1 (Fig. , lanes 4 to 6). The requirement for full-length BRCA1 protein was striking, since the domains required for nonspecific ubiquitination and for γ-tubulin binding reside in the amino-terminal 803 amino acids of BRCA1.
Expression in cells of γ-tubulin with a mutant lysine 48 resulted in centrosome amplification. The γ-tubulin residues modified by BRCA1-dependent ubiquitination were determined by using the bacterially expressed γ-tubulin (as in Fig. ) and mass spectroscopy of the ubiquitinated products. Lysines 48 and 344 of γ-tubulin were found to be modified (Fig. ). We mutated these lysines to arginine, singly and together, and expressed these γ-tubulin mutants by transient transfection in Hs578T mammary cells and in U2OS osteosarcoma cells. Western blotting of the epitope-tagged γ-tubulin mutants revealed that they were all expressed well (Fig. ). The γ-tubulin K344R mutant was found to have a slowed migration, even though the DNA sequence revealed only the lysine-to-arginine mutation. We tested whether transient transfection of these γ-tubulin mutants affected the centrosome number by cotransfecting GFP-centrin in order to analyze only cells that had been transfected. Remarkably, expression of these mutated γ-tubulin proteins caused centrosome amplification (Fig. ). The γ-tubulin K48R mutant had the most pronounced effect on centrosome number after 48 h, causing a 2.5-fold increase in the number of cells with more than two centrosomes. On the other hand, expression of the γ-tubulin K344R mutant caused only a modest increase in the number of cells with amplified centrosomes after 48 and 72 h. Expression of a mutant γ-tubulin containing both modified lysines caused no effect after 48 h but a greater-than-twofold increase in centrosome number at the 72-h time point. We interpret the slowed centrosome amplification by the γ-tubulin double mutant as a deficiency in localizing to the centrosome.
Expression of the mutant γ-tubulin in both Hs578T and U2OS cell lines resulted in centrosome amplification. In fact, the results from experiments with the two cell lines were nearly identical. This shows that the ubiquitination of γ-tubulin is important in both cell types but that the enzyme that mediates the ubiquitination may differ from cell type to cell type. Taking these findings together with the data from Fig. and , we conclude that the mammary cell lines tested depend on BRCA1 for this important ubiquitination event and that other cell lines, such as U2OS, which are unaffected by BRCA1 inhibition, rely on another factor to modify lysine 48 of γ-tubulin.
This article has been 
