Location of BRCA1 in Human Breast and Ovarian Cancer Cells

present data suggesting thalt the BRC A1 g(yenc product i's excluded fromn the nucleuLs in spoI-adic breast aV-nd ovwarian cancer and in certain breast and ovalrian cancer cell lines (1). ImLinostaininrg with al single, nlonaffilnity pUrified p(olvc lonal antibody served as the critcrion for establishing, the subcellular localizaltion of the protein. Here, we describe resuLlts thalt differ substantially from thosc of Cihen et al. ( 1). We have dCve1()pope several BRCA I -specific antibodies. With thlC uISC of a BRCA1 carboxyl-terminal peptide (CQELDTYLIPQIPHSHY) as i'MMuLn0,o,geCn, eC raised a rabbit polyclonal antiserum. It was then affinity-purified igainst the aforeimentioned peptide immun"ogn affixed to agarose beads (Pierce, Rockford, Illinois). We also generated a panel of monoclonal antibodies (mAbs) to BRC2A 1 by imtimunizing mice with the saime peptide or with defined segmilCents of BRC-A 1 thilt ha(id been encoced by elemiients of cloned, 11Lman BRCA1 complementary DNA (cDNA) (2). The latter were synithesized as glutathione-S-transferase (GST) fLsion proteins in Escherichia coll. We validated the specificity of these antibodSies in three ways: (i) by immunoprecipitation (IP) of intact BRCA1 (_220 k )) present in `S-methionine-labeled whIolc-cell lysates; (ii) by performing a proteinl immunoblot o)f unlabeled whole-cell IysaJtes; and (iii) c by 11 Of BRCA1 from lalblleled whole-cell lysates, followed by a protein immllunoblot, with the uLsC of antibodies to distinct epitoecs of BRCAI to immun1lioprecipitate the protein (Fig. 1). Endogenols BRCA1, immunoprecipitated by the.sc antibodies, migrated as a doublet of about 220 kD) in 6% SDS-polyacrylamide gels, thc lowcr band of which co-migrated with in vitro-translated BRCA1 (Fig. 1). Immrntinostainingy of neuLtral paraformaldehyde-fixed cell lines witlh BRCAI antibodies gave rise to a nuLc:lea-r signal (Fig. 2). In particular, thc affinity-pUrified rabbit polyclonal antiserutmi raised against the carboxyl-terminal peptide of BRCA 1, and caclh of seven differ-ent mAbs, all produced a nulclear d(ot pattcrn. TwTo-col1or conIfocal immtUnofluorescence stuidies uissinig this affinity-prUrifiecd rabbit polyclonal antiserum in combination-uwith eachi of the seven mAbs demonstrated co-locllization of the nuLIclear cot pattern in each case (Fig. 2). Thus, all antibodies appeared to react with a commilon structuLre or strLctuLres. The BRC(A I 1nLuclear d.ot pattern was observed in all human ccll lines examined, regardlless of thC tisSuLe of origin, as well as in primary hunman diploid fibroblasts, primary human mammary epithelial cells (HMECs), and in all (six) breast and (six) ovarian cancer cell lines tested. Furthermiiore, both

g(yenc product i's excluded fromn the nucleuLs in spoI-adic breast aV-nd ovwarian cancer and in certain breast and ovalrian cancer cell lines (1). ImLinostaininrg with al single, nlonaffilnity pUrified p(olvc lonal antibody served as the critcrion for establishing, the subcellular localizaltion of the protein. Here, we describe resuLlts thalt differ substantially from thosc of Cihen et al. ( 1).
We validated the specificity of these an-tibodSies in three ways: (i) by immunoprecipitation (IP) of intact BRCA1 (_220 k )) present in`S-methionine-labeled whIolc-cell lysates; (ii) by performing a proteinl immunoblot o)f unlabeled whole-cell IysaJtes; and (iii) c by 11 Of BRCA1 from lalblleled whole-cell lysates, followed by a protein immllunoblot, with the uLsC of antibodies to distinct epitoecs of BRCAI to immun1lioprecipitate the protein (Fig. 1). Endogenols BRCA1, immunoprecipitated by the.sc antibodies, migrated as a doublet of about 220 kD) in 6% SDS-polyacrylamide gels, thc lowcr band of which co-migrated with in vitro-translated BRCA1 (Fig. 1).
Immrntinostainingy of neuLtral paraformaldehyde-fixed cell lines witlh BRCAI antibodies gave rise to a nuLc:lea-r signal (Fig. 2). In particular, thc affinity-pUrified rabbit polyclonal antiserutmi raised against the carboxyl-terminal peptide of BRCA 1, and caclh of seven differ-ent mAbs, all produced a nulclear d(ot pattcrn. TwTo-col1or conIfocal immtUnofluorescence stuidies uissinig this affinity-prUrifiecd rabbit polyclonal antiserum in combination-u with eachi of the seven mAbs demonstrated co-locllization of the nuLIclear cot pattern in each case (Fig. 2). Thus, all antibodies appeared to react with a commilon structuLre or strLctuLres.
The BRC(A I 1nLuclear d.ot pattern was observed in all human ccll lines examined, regardlless of thC tisSuLe of origin, as well as in primary hunman diploid fibroblasts, primary human mammary epithelial cells (HMECs), and in all (six) breast and (six) ovarian cancer cell lines tested. Furthermiiore, both In each case, a different monospecific antibody to BRCAl was used. Immunoprecipitates were dissolved, separated in a 6% SDS-polyacrylamide gel, and a protein immunoblot for BRCAl was performed (5). BRCAl migrated above the 200-kD molecular weight marker. Lanes in which an immunoprecipitation step was performed contain an intensely staining immunoglobulin heavy chain signal (resulting from cross-reaction with the secondary antibody) that migrated to the lower molecular weight regions of the gel. Lanes labeled "IVT" and "L" were not preceded by an immunoprecipitation step. (A) IP performed with mAbs to BRCAl. IPs in lanes 1 to 5 were performed with the addition of 5 p.l of polyclonal antiserum and mAh elicited the same co-localizing nuclear dot immnunofluorescence pattern in cells fixed with netutral paraformaldehyde, or methanol, or 70% ethanol. Thus, the nuclear dot distribution of BRCA1 is a general celltular characteristic and not the result of a fixation artifact.
In a further effort to learn whether the nuclear staining observed with the varioLus affinity-purified rabbit polyclonal antibody to mouse IgG (Cappel, West Chester, Pennsylvania). IVT, 15 pl of rabbit reticulocyte lysate containing in vitro-translated, clonal human BRCAl was analyzed, as a control. The 220-kD band seen in this lane was shown by autoradiography to be labeled by -Z9Smethionine. In vitro translation performed in the absence of clonal BRCAl did not synthesize 35S-labeled proteins migrating in this region of the gel; thus, the 220-kD band detected after in vitro translation of clonal BRCAl is presumed to be a BRCAl gene product. L, MCF7 whole-cell lysate (50 >g of total protein); lane 1, IP with mAb MS1 10 (2); lane 2, IP with mAb MS13 (2); lane 3, IP with mAb AP16 (2); lane 4, IP with mAb GG22 (negative control, raised against human E2F4); and lane 5, IP with SG1 1 (2). This gel is presented as a composite, as the in vitro translate (IVT) immunoblot signal was stronger than other lanes at equivalent exposures. (B) IP using Al 9, an affinity-purified rabbit polyclonal antiserum to the BRCAl carboxyl-terminal peptide CQELDTYLIPQIPHSHY. Lane 1, IP with Al 9; lane 2, IP with Al 9 that has been preincubated with a 20-fold molar excess of the immunizing peptide; L, MCF7 whole-cell lysate (50 kg of total protein). This gel is presented as a composite, as the signal from the whole-cell lysate immunoblot (L) was stronger than that of the Al 9 IP at equivalent exposures. Efficiency of immunoprecipitation by some BROAl antibodies appeared to be improved by the presence of small quantities of SDS and deoxycholate in the lysis buffer. Fig. 2. BRCAl localizes to a nuclear structure or structures. Cell lines were cultivated on glass cover slips, fixed in neutral paraformaldehyde, and then Triton-permeablized as described previously (3). Cover slips were stained with antibodies to BRCAl with Al 9 (rhodamine/red) and mAb MS13 [fluorescein isothiocyanate (FITC)/green], were counterstained for DNA (DAP/blue), and observed with confocal microscopy. Antibodies were incubated at 37°C. Image shows ovarian cancer cell line, SKOV-3. (A) Image showing Al 9 + 4,6-diamidino-2-phenylindole (DAPI) stains. Al 9 stain produces a nuclear dot pattern. (B) Image showing MS1 3 + DAPI stains. MS13 stain produces a nuclear dot pattern. (C) Conjoint image showing the additive effect of the Al 9 + MS1 3 stains. Yellow signal occurs where Al 9 (red) and MS13 (green) stains overlap, indicating the localization of the BRCA1 protein. Negative controls yielded no significant nuclear signal. Similar BRCA1 -specific nuclear dot patterns were seen in AXM 1, AT1 4, SW626, OVC-1, and CAOV (ovarian cancer cell lines); in MCF7, T47D, MDAMB435S, MDAMB415, MDAMB157, and MDAMB231 (breast cancer cell lines); and in U20S, HeLa, 293, DUl 45, SAOS-2, IMR90 (primary human diploid fibroblasts), and HMEC. Antibodies elicited a nuclear dot pattern in CV1 and COS cells, indicating extensive sharing of epitopes between human and monkey BRCAl.
Location of BRCA1 in Human Breast and Ovarian Cancer Cells antibodies depends on special cell fixation conditions, we tested the subcellular localization of BRCA1 by biochemical extraction analysis in unfixed cells. Three cancer cell lines (MCF7, SKOV-3, and U20S), each characterized by dotlike nuclear staining and the absence of cytoplasmic staining after neutral paraformaldehyde fixation, were analyzed by cell fractionation and immunoblotting for BRCA1 (Fig. 3). In all three of these lines, BRCA1 was concentrated in the nuclear fraction. The validity of the fractionation procedure was confirmed by assaying for p300 [a nuclear protein (3)], P-tubulin, and GDI-1 (cytosolic proteins). Moreover, there was minimal cross-contamination of nuclear and cytosolic fractions (Fig. 3). Hence, BRCA1 behaved as a nuclear protein in two different analytic tests, one performed with multiple, specific antibodies on cell lines derived from human breast and ovarian cancers. This conclusion differs from that drawn by Chen et al. (1), who concluded that, in such cell lines (some of which were also tested here) BRCA1 was cytoplasmic, being specifically excluded from the nucleus (1). We did detect a weak cytoplasmic signal with some of our BRCA1 mAbs in certain breast and ovarian carcinoma cell lines. In HMECs, but not in other cell lines, the affinity-purified polyclonal antiserum also gave a cytoplasmic signal. However, twocolor immunofluorescence and confocal microscopic analysis did not reveal co-localization of any two cytoplasmic staining sig-  IVT,15 .tl of in vitro-translated BRCA1; L, whole-cell lysate; C, cytosolic fraction; N, nuclear fraction; M, membrane fraction. Controls for the quality of fractionation were as follows: p300 (detected with mouse mAb RW1 28, a nuclear protein (3); 1-tubulin (antibody is from Boehringer-Mannheim); and GDI-1 (detected with an affinity-purified rabbit polyclonal antiserum). In the experiment shown, there was some widening of lanes in the lower molecular weight regions of the gel. nals generated with different monospecific antibodies, which suggests that the cytoplasmic signals represent nonspecific crossreactions.
An effort was also made to determine the subcellular localization of BRCA1 in tumor cells in alcoholic formalin-fixed, paraffin-embedded sections of primary invasive ductal breast carcinoma (4). A similar analysis was performed by Chen et al. (1). With the use of either of two different BRCA1 mAbs, a variety of different tumor cell staining patterns was noted in the 14 samples we analyzed. They ranged from predominantly nuclear to mainly cytoplasmic to both nuclear and cytoplasmic. By contrast, microwave heating of slides from the same tumor samples, performed with the intention of maximizing antibody access to the available BRCA1 epitope or epitopes before immunostaining, produced a predominantly cytoplasmic signal in 14 out of 14 samples. The discrepancy in the signal observed with and without microwave treatment raised a question as to which, if any, of the detected signals most accurately reflected the true intracellular distribution of BRCA1 in these tumors.
To pursue this question further, we again analyzed aliquots of MCF7 and SKOV-3 cells, where BRCA1 was re-C * _ D Fig. 4. BRCAl immunostaining in formalin-fixed, paraffin-embedded pellets of cancer cell lines. Cell lines SKOV-3 (ovarian cancer) and MCF7 (breast cancer) were grown and pelleted. Pellets were divided into two aliquots that were fixed in alcoholic formalin (AF) and neutral buffered formalin (NBF), respectively. Subsequent processing and immunoperoxidase staining methods were identical to those used on sections of primary invasive ductal breast carcinoma (4). For a given fixation-microwave combination, similar staining patterns were obtained in sections of pelleted MCF7 or pelleted SKOV-3 cells, with the use of mAb SG1 1 or mAb MS1 3 (2). The combination shown in this figure-SG1 1 staining of MCF7 cell pellet sections-is therefore typical of the other cell line-antibody combinations. (A) AF fixation, no microwave treatment before staining; (B) AF fixation, with microwave treatment before staining; (C) NBF fixation, no microwave treatment before staining; and (D) NBF fixation, with microwave treatment before staining. vealed to be nuclear by two independent criteria. Each cell line was pelleted and divided into two aliquots. One aliquot of each line was fixed in alcoholic formalin and the other in neutral buffered formalin. Cut sections of each pellet were processed with or without microwave treatment in the same manner as the above-noted tumor tissue. One of the two BRCA1 mAbs used in our earlier experiment on sections of primary invasive ductal breast carcinoma was used for immunoperoxidase staining of MCF7 cells (Fig. 4). Identical results were obtained when SKOV-3 cells were reacted with this antibody and when each of the two cells lines, fixed in the same manner, was reacted with the second mAb used in the earlier experiment. In alcoholic formalin-fixed cells, a strong cytoplasmic staining pattern was seen in all cases. By contrast, in cells fixed in neutral buffered formalin (different from neutral paraformaldehyde) and exposed to microwave heating before immunoperoxidase staining, the BRCA1 signal was predominantly nuclear. However, in neutral buffered formalin-fixed cells that had not undergone microwave treatment, the signal was strongly and exclusively cytoplasmic (Fig. 4).
Thus, cells known to contain exclusively nuclear BRCA1 (as shown by biochemical extraction and by immunostaining performed under certain conditions of fixation) revealed non-nuclear staining under other fixation conditions-those commonly used to analyze tumor sections. This result, along with the observation that differences in BRCA1 staining patterns of breast cancer sections can be linked to variation in fixation or staining conditions, raises questions about the biological significance of detecting largely cytoplasmic BRCA1 staining in any breast cancer section ( ). Taken together, the results reported here do not support the hypothesis (1) , ibid. 266, 66 (1994). We obtained the cDNA for BRCAl by probing a 293-cell cDNA library. Two partially overlapping fragments were reassembled to produce full-length BRCAl cDNA, which was then sequenced and found to contain no mutations. Antibodies were raised in mice against a BRCAl -GST fusion protein that contained residues 1 to 304 of BRCAl (amino terminal segment, MS series of mAbs) or residues 1313 to 1863 of BRCA1 (carboxyl-terminal segment, AP series of mAbs). The SG series of mAbs was raised against the BRCAl carboxyl-terminal peptide CQELDTYLIPQIPHSHY. Monoclonal antibody fusions were screened by immunoprecipitation of 35S-methionine-labeled, in vitro-translated BRCA1, followed by 6% SDS-PAGE and autoradiography. Positive wells were recloned and rescreened twice, the second subcloning being at limiting dilution, which allowed single-cell clones to be picked. 3. R. Eckner et al., Genes Dev. 8, 869 (1994). 4. Alcoholic formalin is composed of 10% buffered formalin (Anatech Ltd, Battle Creek, Ml) diluted into 70% ethanol. Immunoperoxidase staining was performed using a Ventana Automated Immunostainer (Ventana Medical Systems, Tucson, AZ). Monoclonal antibody SG11 was used at 1:10 dilution; mAb MS13 was used without dilution. Where microwave heating was used for antigen retrieval, sections were heated in the microwave in citrate buffer pH 6 for 10 minutes, with buffer replacement in between the heating periods. 3-3' diaminobenzadine was used as the chromogen, and the sample was lightly counterstained with methyl green.
Eckner; and rabbit polyclonal antiserum to GDI-1 (Fig. 3) was a gift from P. E. Bickel. This work was also funded by grants from the U.S. National Institutes of Health. 16 January 1996;accepted 20 February 1996 Response: Scully et al. state that they find BRCA1 protein exclusively in the nucleus of many types of human cells, including cells derived from breast and ovarian cancer cell lines. Using different reagents (1), we also found that BRCA1 was localized in the nucleus in many types of human cells. However, we also found that BRCA1 was localized almost exclusively in the cell cytoplasm of breast and ovarian cancer cell lines. We agree with Scully et al. that accurate localization of gene products by immunocytochemistry depends on antibody specificity, as well as on methods of fixation and staining. Let us consider in more detail some of the similarities and differences between the studies. It is important to determine whether the different antibodies used by each group are specific for BRCA1 alone, or whether they also recognize cross-reacting proteins that may profoundly influence the results of immunohistochemistry (IHC) and subcellular fractionation experiments. In our report, we used two different mouse polyclonal antibodies, raised against large and distinct regions encoded by BRCA1 exon 11, to characterize the BRCA1 protein.
One of our antibodies was raised against a GST-BRCA1 fusion protein corresponding to amino acids 762 through 1315, and the other, against a GST fusion protein corresponding to amino acids 341 through 758. Both antibodies were purified by preabsorption with GST beads. Both gave essentially identical results, but only one (antibody to BRCA1 762-1315) was emphasized in our report (1) because of space limitations. We carefully determined the specificity of these antibodies by immunoprecipitating 35S-labeled BRCA1 and reprecipitating it with either of the two polyclonal antibodies, to minimize contamination with cross-reacting and co-precipitating proteins. Likewise, in protein immunoblots, performed after we immunoprecipitated the protein from cellular lysates and probed with the same antibody, only a single protein migrating at 220 kD was visualized on a full-length blot (Fig. 1A, lanes 1  through 4). These data strongly suggest that our antibodies are specific for BRCA1, possess little if any cross-reactivity, and are appropriate for IHC studies.
We, like Scully et al., have made other polyclonal and monoclonal antibodies against various regions of the 220-kD BRCA1 protein. Most of our antibodies cross-react with other proteins, which suggests that truly specific reagents are difficult to obtain. With the use of our relatively nonspecific antibodies in immunoprecipitation or in straight protein immunoblots of cell lysates, cross-reacting proteins are often much more abundant (by a factor of ten) than is the 220-kD BRCA1 protein itself (Fig. 1, lanes 5 through 8). Furthermore, the cross-reacting antibodies usually show predominant nuclear immunostaining of the same breast cancer cell lines that also demonstrate predominantly cytoplasmic staining with our BRCA1-specific antibodies. Our only truly specific antibodies were purified from polyclonal sera from samples taken soon after immunization of the mice. This result suggests that repeated boosting may favor the more abundant cross-reacting substrates, again making it difficult to obtain specific antibodies indefinitely.
In an attempt to resolve the problems associated with potential antibody crossreactivity, we have recently created a tagged BRCA1 expression vector. BRCA1 protein is expressed from a plasmid (based on Invitrogen's pCEP4, San Diego, California) that contains the entire 5.5-kb coding sequence of the BRCA1 cDNA, in-frame with a Flag epitope-tag sequence at the N-terminal region. The tag permits detection of exogenous BRCA1 with the use of the specific antibody to Flag (M2, Kodak, Rochester, New York) (2) after the CEP4-BRCA1 construct is transfected into antiserum to BRCAl (amino acids 762 through 1315). Immunoprecipitates were separated by SDS-PAGE. The blot was developed by probing with the same antibody. Lanes 5 through 6: straight Western blotting of HBL1 00 cell lysates (5 x 106 cells, lane 6), using mAb 6B4; the thin horizontal line marks BRCA1, which migrates at about 220 kD, and the arrowhead marks an abundant, cross-reacting protein migrating at about 110 kD. Lanes 7 through 8: same experiment using mAb 24G1 1; two cross-reacting proteins are marked by arrowheads.