Breast Cancer Metastasis Suppressor 1 (BRMS1) Forms Complexes with Retinoblastoma-binding Protein 1 (RBP1) and the mSin3 Histone Deacetylase Complex and Represses Transcription*

suppressor 1 (BRMS1) suppresses metastasis of multiple human and murine cancer cells without inhibiting tumorigenicity. By yeast two-hybrid and co-immunoprecipitation, BRMS1 interacts with retinoblastoma binding protein 1 and at least seven members of the mSin3 histone deacetylase (HDAC) complex in human breast and melanoma cell lines. BRMS1 co-immunoprecipitates enzymatically active HDAC proteins and represses transcription when recruited to a Gal4 promoter in vivo . BRMS1 exists in large mSin3 complex(es) of (cid:1) 1.4–1.9 MDa, but also forms smaller complexes with HDAC1. Deletion analyses show that the carboxyl-terminal 42 amino acids of BRMS1 are not critical for interaction

The complex process of cancer cell dissemination and the establishment of secondary foci involves the acquisition of multiple abilities by metastatic cells. For example, blood-borne metastasis requires cells to invade from the primary tumor, enter the circulation, survive transport, arrest at a secondary site, recruit a blood supply, and proliferate at that site (1). The ability to accomplish all of these steps likely involves changes in, and coordinated expression of, a large assortment of genes. Consistent with this notion, several genes, proteins, and pathways have been associated with metastatic progression, including oncogenes, motility factors, and matrix metalloproteinases (1,2). In addition to metastasis-promoting genes, a new class of molecules called metastasis suppressors has been described (reviewed in Refs. [2][3][4][5]. By definition, metastasis suppressors inhibit metastasis without blocking primary tumor growth, presumably by inhibiting one or more steps necessary for metastasis. To date, 13 metastasis suppressor genes have been identified that reduce the metastatic ability of cancer cell line(s) in vivo without affecting tumorigenicity, namely breast cancer metastasis suppressor 1 (BRMS1), 1 CRSP3, DRG1, KAI1, KISS1, MKK4, NM23, RhoGDI2, RKIP, SSeCKs, VDUP1, E-cadherin, and TIMPs (reviewed in Refs. 4 and 5).
We identified BRMS1 using differential display to compare highly metastatic breast carcinoma cells with related but metastasis-suppressed cells (6). Enforced expression of BRMS1 suppressed metastasis in three animal models, namely human breast (6), murine mammary (7), and human melanoma cells (8). Additionally, BRMS1 mapped to loci in murine (7) and human (6) genomes that had previously been implicated in metastasis control (9). The BRMS1 protein localized to nuclei and restored gap junctional intercellular communication in both breast and melanoma tumor cell lines (8,10,11), but its molecular functions remain to be elucidated.
One approach to determine a mechanism of action involves identifying which proteins interact with BRMS1. In this report, we utilized yeast two-hybrid and co-immunoprecipitation (co-IP) to demonstrate that BRMS1 interacts with retinoblastomabinding protein 1 (RBP1). This association led to experiments to demonstrate that BRMS1 interacts with at least seven members of the mammalian Sin3 (mSin3) mSin3⅐histone deacetylase (HDAC) complexes, including HDAC1 and HDAC2.
Human HDACs exist in many large, multi-subunit protein complexes (12) that are recruited to specific regions by DNAbinding factors. As their name indicates, HDACs remove acetyl groups from lysine residues at the N-terminal tails of core histones (13)(14)(15). Histone deacetylation favors transcriptional repression, whereas acetylation (mediated by histone acetyltransferases) favors transcriptional activation. mSin3⅐HDAC complexes are named for the large mSin3A and mSin3B proteins, which are thought to serve as scaffolds for complex assembly (14). HDAC enzymatic activity in mSin3 complexes is mediated by a core subunit consisting of HDAC1, HDAC2, RbAp46, and RbAp48 (13). The core HDAC subunit is also found in at least one other HDAC complex, NuRD (nucleosomal remodeling and deacetylation (16).
Mammalian Sds3 (mSds3; suppressor of defective silencing 3) was recently reported to be an integral component of the mSin3 complex and acts to stabilize HDAC1 within the complex (17). BRMS1 shares homology with mSds3, suggesting that BRMS1 belongs to a protein family (17). mSin3-associated proteins, SAP18 and SAP30, which are believed to serve as adapter molecules, complete the core complex as currently understood (18 -20).

MATERIALS AND METHODS
Cell Lines, Cell Culture, and Transfections-MDA-MB-231 is a human estrogen receptor-and progesterone receptor-negative cell line derived from a pleural effusion from an infiltrating ductal breast carcinoma. C8161 is a metastatic, amelanotic human melanoma cell line derived from an abdominal wall metastasis. C8161.9 is a highly metastatic clone obtained by limiting dilution cloning of C8161 (21). 66cl4 is a murine mammary carcinoma cell line derived from a spontaneous carcinoma in BALB/cfC3H mice (22,23). All cell lines were cultured in a 1:1 mixture of Dulbecco's modified minimum essential medium and Ham's F12 medium supplemented with 5% fetal bovine serum (Atlanta Biologicals, Atlanta, Georgia), 1% non-essential amino acids, and 1 mM sodium pyruvate. Transfected cells also received 500 g/ml G418 (Geneticin; Invitrogen). All cells were maintained on 100-mm Corning tissue culture dishes at 37°C with 5% CO 2 in a humidified atmosphere. MDA-MB-231 cells were passaged at 80 -90% confluence using a solution of 0.125% trypsin and 2 mM EDTA in Ca 2ϩ /Mg 2ϩ -free Dulbecco's phosphate buffered saline (CMF-DPBS). C8161.9 and 66cl4 cells were passaged at 80 -90% confluence using 2 mM EDTA in CMF-DPBS. BRMS1 was cloned into the constitutive mammalian expression vector pcDNA3 (Invitrogen, San Diego, CA) under control of the cytomegalovirus promoter. No antibiotics or antimycotics were used. All cell lines were found to be negative for Mycoplasma spp. contamination using a PCR-based method (TaKaRa, Madison, WI).
To detect BRMS1 protein expression, a chimeric molecule was constructed with an N-terminal epitope tag (SV40T epitope 901) (24,25). Epitope-tagged full-length BRMS1 and deletion mutants were cloned into pcDNA3 before introduction into cells by electroporation (Bio-Rad model Gene Pulser; 220 V, 960 microfarads, ϱ ohms). Briefly, cells (0.8 ml; 1 ϫ 10 7 cells/ml) from 80% confluent plates were detached, plasmid DNA (10 -40 g) was added to the cells, and the mixture was placed onto ice for 5 min before electroporation, followed by 10 min on ice prior to plating on 100-mm cell culture dishes. Transfectants were selected using G418 (Geneticin; 500 g/ml). Single-cell clones were isolated by limiting dilution in 96-well plates. Stable transfectants were assessed for protein expression by immunoblotting.
Yeast Two-hybrid Screen-A yeast two-hybrid screen was performed to isolate cDNAs encoding BRMS1-interacting proteins essentially as described in the manufacturer's instructions (Clontech MATCHMAKER LexA). Full-length BRMS1 was cloned in-frame with the GAL4 DNA binding domain in the pDBTrp (Invitrogen) vector to obtain pDB-BRMS1. This GAL4DB-BRMS1 fusion (bait) construct was used to transform AH 109 (MATa, trp-901, leu2 -3, 112, ura3-52, his3-200, gal4⌬, gal80⌬, LYS2::GAL UAS -GAL1 TATA -HIS3, GAL2 UAS -GAL2 TATA -ADE2, URA3:: MEL1 UAS -MEL1 TATA -lacZ, MEL1). Human breast, prostate, and placenta cDNA libraries in pACT2 (MATCHMAKER, BD Biosciences Clontech) were screened in yeast drop-out minimal medium lacking histidine, tryptophan, and leucine. His ϩ colonies were tested for growth on minimal medium lacking adenine, tryptophan, leucine, and ␤-galactosidase activity as described previously (27). cDNA plasmids were isolated from each positive yeast clone using Zymoprep (Zymo Research, Orange, CA) and sequenced. The interaction phenotype was lost when either the bait or prey plasmid was lost from the cell. Re-introduction of missing partners restored growth on minimal medium lacking histidine, tryptophan, and leucine, growth on medium lacking adenine, tryptophan and leucine, and restoration of ␤-galactosidase activity. 35 S Protein Labeling-Cells were grown to 80 -90% confluence in 100-mm tissue culture plates. Media were removed and replaced with 3 ml of cysteine-methionine-free media (Invitrogen) containing 5% fetal bovine serum for 1 h. Media were removed and replaced with 3 ml of cysteine-and methionine-free media containing 5% fetal bovine serum and 100 Ci/ml 35 S-express protein labeling mix (PerkinElmer Life Sciences). Cells were incubated for 18 h before protein was collected for co-IP.
Co-immunoprecipitation-Cells (90 -95% confluence) were washed twice with ice-cold PBS and lysed with ice-cold lysis buffer (0.5% Igepal CA-630 (Sigma), 50 mM Tris, pH 8, 150 mM NaCl, and 2 mM EDTA) containing 1 mM phenylmethylsulfonyl fluoride, 2 g/ml aprotinin, 50 mM NaF, 0.2 mM Na 3 VO 4 , and 10 l/ml of a protease inhibitor mixture containing 4-(2-aminoethyl) benzensulfonylfluoride (AEBSF), pepstatin A, trans-epoxysuccinyl-L-leucylamido(4-guanido)butane (E-64), bestatin, leupeptin, and aprotinin (Sigma). Lysate was kept at 4°C during all subsequent steps. Lysate was passed through a 21-gauge needle several times, incubated on ice for 1 h, then centrifuged for 1 h at 12,000 ϫ g in a Sorvall MC 12V microcentrifuge with an F12/M.18 rotor to remove insoluble debris. Lysates were then rocked gently in the presence of antibody for 1 h, followed by the addition of 20 l of protein A/G PLUS agarose beads (Santa Cruz Biotechnology) and rocking overnight. Agarose beads were washed twice with ice-cold PBS, heated to 60°C in sample buffer, subjected to SDS-PAGE, and transferred to polyvinylidene difluoride membrane for immunoblotting. For 35 S-labeled samples, films were exposed directly to polyvinylidene difluoride membranes. In each experiment, blots were probed with antisera to the immunoprecipitated protein to verify the quality of the immunoprecipitation and assess the equal loading of lanes.
Size Exclusion Chromatography-Whole cell protein lysate (pooled from 10 100-mm plates using 1 ml of lysis buffer each) was applied to a Superose 6 HR 10/30 size exclusion column (Amersham Biosciences). The column was run using lysis buffer with 1 mM phenylmethylsulfonyl fluoride and 0.5 mM dithiothreitol at a flow rate of 0.2 ml/min. Fractions (500 l) were collected, and 420 l of each fraction were used for co-IP. The remaining 80 l was used for immunoblotting.
HDAC Activity Assay-Following co-IP, agarose beads were combined with 400 l of HDAC assay buffer (15 mM Tris, pH 7.9, 10 mM NH 4 Cl, 0.25 mM EDTA, 10% glycerol, and 10 mM ␤-mercaptoethanol) containing 1.5 g 3 H-labeled chicken reticulocyte core histones (28) with or without 250 mM sodium butyrate (an HDAC inhibitor). Samples were inverted continuously on a rotating wheel for 3 h at 30°C, and HDAC activity was measured as described previously (28). Briefly, the reaction was stopped by adding 100 l of 1 M HCl/0.4 M acetic acid and 0.8 ml ethyl acetate. Samples were vortexed for 30 s and centrifuged at 8,000 ϫ g for 5 min. An aliquot (0.6 ml) of the upper (organic) phase was then counted for radioactivity in a 5-ml scintillation mixture (Fisher).
Reporter Assays-BRMS1 cDNA was cloned in-frame with the Nterminal Gal4-DNA binding domain in pBIND (Promega). Subconfluent (80 -90%) COS7 cells were transfected using the FuGENE reagent (Roche Diagnostics) with the GAL4-BRMS1 fusion construct and a luciferase reporter plasmid containing four GAL4 binding sites upstream of the myelomonocytic growth factor minimal promoter, kindly provided by Dr. Ron Eisenman. pRLSV40 (Renilla luciferase) was used as a transfection control. Trichostatin A (50, 150, and 300 ng/ml, Sigma) was added for 24 h prior to lysis. Cells were lysed in Passive lysis buffer (Promega) 48 h post-transfection. Cell extracts were assayed for luciferase activity using the Dual luciferase reporter assay system (Promega) and an automated luminometer Monolight TM 3010 (Pharmingen). Transfection efficiencies were normalized using the Renilla luciferase control.

RBP1 and mSds3
Were Identified as BRMS1-interacting Proteins by Yeast Two-hybrid Screen-A yeast two-hybrid screen was performed using prey libraries from three human tissues, breast, placenta, and prostate. Breast was chosen because BRMS1 was first identified as a metastasis suppressor in breast cancer. Placenta and prostate were chosen because BRMS1 mRNA is highly expressed in these tissues (6). Fulllength BRMS1 was used as the "bait." RBP1 was present in the majority of positive clones from breast and placenta libraries, so it was chosen for further studies (Fig. 1, A and B).
The FLJ00052 expressed tag was present as two independent positive clones in a prostate library screen. During the completion of the work reported here, FLJ00052 was identified as the mammalian ortholog (mSds3, GenBank TM accession number XM_045014 mapping to human chromosome 12q24.23) of the yeast Sds3 protein. There are other related genes according to the LocusLink (www.ncbi.nlm.nih.gov/LocusLink/ list.cgi?Q ϭ FLJ00052&ORG ϭ &V ϭ 0), suggesting the existence of additional mSds3 orthologs. mSds3 is an integral component of the mSin3⅐HDAC co-repressor complex, modulates HDAC activity, and stabilizes the complex (17). Antibodies recognizing mSds3 are not available commercially; thus, we have not yet been able to test whether BRMS1 pulled down mSds3.

BRMS1 and RBP1 Are Reciprocally Co-immunoprecipitated in Human Breast and Melanoma Cancer Cells-MDA-MB-231
human breast carcinoma cells and C8161.9 human melanoma cells were transfected with 901 epitope-tagged BRMS1. Immunoprecipitation of BRMS1 followed by immunoblot with two RBP1-specific antibodies (clones LY11 and LY32) (Fig. 1, C and  D) showed that BRMS1 co-immunoprecipitates RBP1 (Fig. 1, C  and D). Negative controls (co-IP using anti-901 in vector-transfected cells or co-IP using an irrelevant antibody, anti-Lamin A/C) did not pull down RBP1 (Fig 1, C and D). Antibody directed against RBP1 co-immunoprecipitated BRMS1 in both breast carcinoma (Fig. 1E) and melanoma (Fig. 1F) cells.
To begin defining the binding domains of BRMS1 responsible for interactions with RBP1, three C-terminal deletion mutants of 901-tagged BRMS1 were generated by exonuclease III digestion, designated BRMS1(⌬204 -246), BRMS1(⌬164 -246), and BRMS1(⌬91-246) (Fig. 2C). Deletion constructs were transfected into both MDA-MB-231 and C8161.9. The latter expressing clones were experimentally more useful, because expression of all three deletion mutants was approximately equivalent to full-length protein (data not shown, but can be inferred from Fig. 2B). In MDA-MB-231, only BRMS1(⌬204 -246)-expressing clones had protein levels approximating fulllength BRMS1 (inferred from Fig. 2A). Anti-901 antibody was used to co-immunoprecipitate deletion mutants, and immunoblotting was used to detect RBP1 (Fig. 2, A and B). Loss of amino acids 204 -246 did not decrease binding to RBP1 in either cell line (Fig. 2, A and B). Loss of amino acids 164 -246 diminished binding (by ϳ90% by densitometry), and loss of amino acids 91-246 abrogated binding (Fig. 2B). Absence of binding by BRMS1 (⌬91-246) was controlled internally for nonspecific binding of RBP1 to the 901 epitope. Interestingly, in both MDA-MB-231 and C8161.9, BRMS1 (⌬204 -246) coimmunoprecipitated RBP1 more effectively (ϳ1.5-fold) than full-length BRMS1 (Fig. 2, A and B). BRMS1 Does Not Appear to Complex with Rb or p107 or to Modulate E2F-dependent Gene Expression-RBP1 binds Rb family members p105 (RB) and p107 (30 -32). Rb proteins, in turn, bind E2F and tether RBP1 to E2F-responsive gene promoters. In this way, RBP1 directly suppresses transcription. We tested the hypothesis that BRMS1 is part of an RBP1⅐Rb⅐E2F complex; however, BRMS1 did not co-immunoprecipitate p105 or p107 in MDA-MB-231 ( Fig. 2A) or C8161.9 cells (data not shown). Likewise, BRMS1 did not affect luciferase expression using an E2F-responsive promoter (data not shown). Taken together, these findings suggest that BRMS1 does not act as part of an RBP1⅐Rb⅐E2F complex and that BRMS1 might be part of a previously undescribed RBP1 complex that does not contain Rb.
BRMS1 Co-immunoprecipitates Several 35 S-labeled Proteins in MDA-MB-231-Anti-901 was used to co-immunoprecipitate BRMS1 from 35 S-labeled lysates from BRMS1-transfected MDA-MB-231. Vector-transfected cells were used as controls. In addition to BRMS1, several additional bands were evident, including prominent large proteins at Ն200 kDa, ϳ160 kDa, and ϳ65 kDa as well as less intense bands just below 50 kDa and another at ϳ30 kDa. (Fig. 3). Parallel experiments were performed using BRMS1-transfected C8161.9 and Brms1 (murine ortholog; Ref. 7)-transfected 66cl4. Similar 35 S-labeled proteins were co-immunoprecipitated by anti-901 (data not shown). The pattern was reminiscent of previously published Immunoblotting with ␣-901 was used as a positive control, and ␣-Lamin A/C was used as a negative control. F, anti-RBP1 co-immunoprecipitated BRMS1 in C8161.9 cells. Immunoblotting with ␣-901 was used as a positive control, and ␣-mSin3A and ␣-SAP18 were used as negative controls.
BRMS1 Is a Component of the mSin3⅐HDAC Complex in C8161.9 and MDA-MB-231-Immunoprecipitation of epitopetagged BRMS1 followed by immunoblotting showed that BRMS1 pulled down seven proteins shown previously to be part of mSin3⅐HDAC complexes, namely mSin3A, mSin3B, HDAC1, HDAC2, SAP30, RbAp46, and RbAp48 (Fig. 4). The same proteins were not precipitated in vector-transfected cells (Fig. 4, lane 1), nor were they pulled down using an antibody to the nuclear protein Lamin A/C (Fig. 4, lane 2). Western blots demonstrated that BRMS1-associated proteins were present at comparable levels in both vector-and BRMS1-transfected cell lysates (data not shown), ruling out the possibility that vectortransfected cells had lower levels of mSin3⅐HDAC complex components. Interactions between BRMS1 and mSin3⅐HDAC were relatively strong, because they persisted in 0.5 M NaCl. Antibodies recognizing mSin3B, HDAC1, HDAC2, and SAP30 "reverse" co-immunoprecipitated BRMS1 in C8161.9 cells as well (Fig. 6A). To determine relative protein expression, 50 g of protein lysate from each transfected construct was immunoblotted (lanes 4 -6) (the exposure for ␣-901 shown here was not long enough to show expression in lanes 4 and 5). BRMS1 did not co-immunoprecipitate Rb or p107. IP, immunoprecipitation; WB, Western blot. B, whole cell lysates (1 mg) were prepared from C8161.9 cells expressing BRMS1 and BRMS1 deletion mutants (see panel C) with protein levels comparable with the clone expressing full-length BRMS1. The deletion mutants exhibited varying abilities to co-immunoprecipitate the above-mentioned proteins (lanes 4 -6). Anti-901 (lane 1) and an irrelevant antibody (anti-Lamin A/C, lane 2) did not pull down RBP1 in vector-transfected cells. Ͼ indicates IgG light chain. C, schematic of BRMS1 deletion mutants. Equal loading of immunoprecipitate is inferred from the data, because equal intensity is observed by probing with anti-901. 35 S-labeled whole cell lysates. Using radiolabeled protein lysate from MDA-MB-231 cells, anti-901 was used to immunoprecipitate epitope-tagged BRMS1. Immunoprecipitation (IP) of BRMS1 revealed at least 12 co-immunoprecipitated proteins. Arrows with numbers indicate co-immunoprecipitated proteins and approximate molecular mass (MW) in kDa. Ͼ indicates IgG light chain.
BRMS1 Interacts with a Subset of mSin3⅐HDAC Complexes-Many proteins that bind HDAC complexes are responsible for recruiting complexes to specific promoters. However, BRMS1 does not have a predicted DNA-binding motif, suggesting that it might serve a different role as a member of subsets of mSin3⅐HDAC complexes.
As a first step to evaluate those potential roles, the ability of BRMS1 to co-immunoprecipitate selected HDAC complex components was tested. Mad and Max were the first proteins shown to recruit the mSin3⅐HDAC to a specific promoter (33)(34)(35), but BRMS1 did not co-immunoprecipitate Mad1 or Max (data not shown). The unliganded nuclear hormone co-receptors SMRT and NCoR have also been reported to recruit the mSin3 (36 -39), but there are contradictory data (40). In our system, BRMS1 did not co-immunoprecipitate SMRT or NCoR (data not shown). mSin3⅐HDAC interaction with MeCP2, a methyl CpG-binding protein, has also suggested that repression associated with DNA methylation may be mediated, in part, by deacetylation (41). Yet, BRMS1 did not co-immunoprecipitate MeCP2 (data not shown). Because the core HDAC subunit (HDAC1, HDAC2, RAp46, and RbAp48) is also present in the NuRD HDAC complex (16), we asked whether BRMS1 complexed with NuRD. BRMS1 did not co-immunoprecipitate Mi-2 or MTA1, two members of the NuRD complex (data not shown). HDAC3, which is related to HDAC1 and HDAC2 and can complex with RBP1 (32), did not co-immunoprecipitate with BRMS1 (data not shown).  6. HDAC1, HDAC2, SAP30, RBP1, mSin3B, and mSin3A co-immunoprecipitated BRMS1. A, in BRMS1-transfected C8161.9 cells, antibodies recognizing mSin3A, mSin3B, SAP30, HDAC1, HDAC2, and RBP1 co-immunoprecipitated BRMS1 from 1 mg of whole cell lysate. Antibodies directed against SAP18, RbAp46, RbAp48, and pRb did not co-immunoprecipitate BRMS1. Presence of the respective antigen recognized for co-immunoprecipitate was confirmed by re-staining the blots with the same antibody used for precipitation. *, in the bottom panel, increased exposure time was used to reveal co-immunoprecipitated BRMS1, causing a cross-reacting band of slower mobility to become visible. B, in BRMS1-transfected MDA-MB-231 cells, antibodies directed against mSin3B, SAP30, HDAC1, HDAC2, and RBP1 co-immunoprecipitated BRMS1 from 1 mg of whole cell lysate. Antibodies directed against mSin3A, SAP18, RbAp46, RbAp48, and pRb did not co-immunoprecipitate BRMS1. Anti-901 was used as a positive control. Presence of the precipitated antigen was verified by re-staining the blots with the antisera used for co-IP. IP, immunoprecipitation; WB, Western blot.
Taken together, these data suggest that BRMS1 exists in a specialized subset of mSin3⅐HDAC complexes rather than existing as an integral component of the complex. In other words, BRMS1 is not a ubiquitous member of mSin3⅐HDAC complexes.

BRMS1 Exists in Large (1.4 and 1.9 MDa) mSin3⅐HDAC Complexes as Well as Smaller Complexes
Containing HDAC1-To determine the size of BRMS1⅐mSin3⅐HDAC complex(es) and the distribution of these molecules in complexes of various sizes, whole cell protein lysates from C8161.9 were subjected to Superose 6 size exclusion chromatography. Fractions were separated by PAGE, transferred to polyvinylidene difluoride, and immunoblotted for 901-BRMS1, HDAC1, SAP30, and mSin3B. These four proteins were chosen because they are core members of the complex. BRMS1 eluted in multiple peaks from the column with complex sizes ranging from ϳ100 to 2,000 kDa. BRMS1 elution was most prominent in peaks 5 and 6 (ϳ1.7 MDa). HDAC1 also eluted in multiple peaks (fractions 4 -22) with the majority present in fractions 8 and 9 (ϳ 1.4 MDa, Fig. 8A). SAP30 was detected in two peaks, one from fractions 4 through 14 and another from fractions 19 to 24, suggesting the existence of at least two complexes, the first Ͼ1 MDa and the second Ͻ200 kDa (Fig. 8A). mSin3B is detected uniformly in fractions 3-17, indicating involvement in complexes ranging from ϳ2 MDa to hundreds of kDa (Fig. 8A).
BRMS1 Co-immunoprecipitates HDAC Activity-To determine whether BRMS1-associated HDAC1 and HDAC2 were enzymatically active, complexes were assessed for deacetylase activity in C8161.9. Full-length BRMS1 co-immunoprecipitated HDAC activity; BRMS1(⌬204 -246) pulled down less HDAC activity. BRMS1(⌬164 -246) co-immunoprecipitates still less HDAC activity, whereas BRMS1(⌬91-246) pulled down only background activity (Fig. 7). This pattern is reminiscent of the pattern of interaction with HDAC1 seen by immunoblot (Fig. 4). As a positive control, anti-HDAC1 antibodies were able to pull down HDAC activity (Fig. 7) proportionate to the amount of antibody used (i.e. when 2ϫ anti-HDAC1 was used, double the HDAC activity was precipitated). These results show that only a small portion of the HDAC1 FIG. 8. BRMS1 co-immunoprecipitated a large (ϳ1.6 MDa) complex containing HDAC1, SAP30, and mSin3B as well as smaller complexes containing HDAC1. A, elution profile of BRMS1, HDAC1, SAP30, and mSin3B in BRMS1-transfected C8161.9 cells. Whole cell lysate (3 mg total protein) was prepared and applied to a Superose 6 size exclusion column. Fractions (500 l) were collected, and 20 l of each fraction were subjected to SDS-PAGE and immunoblotting. MW, molecular mass. B, immunoprecipitation of BRMS1 within eluted fractions. Whole cell lysate (3 mg of total protein) was prepared from BRMS1-transfected C8161.9 cells and applied to a Superose 6 size exclusion column. Fractions (500 l) were collected, and anti-901 was used to immunoprecipitate BRMS1 from 420 l of each fraction. Immunoprecipitated complexes were subjected to PAGE and immunoblotting.
FIG. 7. BRMS1 pulls down HDAC activity. Whole cell lysate (6 mg of total protein) was prepared from BRMS1-transfected C8161.9 cells as well as from C8161.9 cells expressing BRMS1 deletion mutants (⌬204, ⌬164, and ⌬91) and vector-transfected (V) cells. Anti-901 was used to immunoprecipitate BRMS1 and BRMS1 deletion mutants from this lysate, and co-immunoprecipitated HDAC activity was measured. The HDAC inhibitor sodium butyrate (250 mM) was used to show that the release of 3 H-acetyl groups was due specifically to HDAC activity. Anti-HDAC1 was used as a positive control (*, 10 g of anti-HDAC1 was used; **, 5 g of anti-HDAC1 was used). Anti-Lamin A/C was used as a negative control. Bars with error bars represent mean Ϯ S.E. for two independent experiments. See Fig. 2C for a schematic of the BRMS1 deletion mutants. activity present in the protein lysate is being measured. Vectortransfected cells and co-IP with anti-Lamin A/C served as negative controls (Fig. 7).
BRMS1 Represses Transcription in Luciferase Reporter Assays-On the basis of its physical interactions with mSin3 and HDAC1, it was predicted that BRMS1 would repress transcription. To investigate this prediction, we measured the effect of BRMS1 on transcription using a luciferase reporter containing four GAL4 binding sites upstream of the myelomonocytic growth factor minimal promoter. BRMS1 strongly repressed (ϳ80%) basal transcription compared with the pBIND vector alone (Fig. 9).

DISCUSSION
Epigenetic regulation of the metastatic phenotype was proposed in 1889 when Sir Stephen Paget recognized that tumor cells colonize certain organs preferentially based, in part, upon how they respond to signals from the microenvironment (42). Trainer and co-workers later showed that treatment of murine melanoma cells with the DNA de-methylating agent 5-azacytidine resulted in reversible reduction of metastatic lung colonization (43). Recent studies have shown that treatment of cells with 5-azacytidine can induce expression of the metastasis suppressor genes Nm23 (44) and KAI1 (45). Links between metastasis and HDAC activity first became apparent when the breast cancer metastasis promoting gene, MTA1, was identified as a component of the NuRD⅐HDAC complex (46,47). MTA1 has subsequently been shown to repress estrogen receptor-dependent transcription in an HDAC-dependent manner (48). Likewise, loss of expression of heterochromatin protein 1 (HP1) has been associated with acquisition of metastatic potential in human breast cancer (49). Together, these findings support the hypothesis that regulation of the transcriptome by a variety of mechanisms is a critical determinant of cancer spread. The findings reported here represent the first direct evidence that a metastasis suppressor gene is a component of an HDAC complex. It is possible that specialized HDAC complexes may promote (as implied by MTA1) or inhibit (as implied by BRMS1) cancer metastasis. The data compel the hypothesis that metastasis is regulated, at least in part, by histone deacetylase activity, chromatin remodeling, and/or transcriptional repression.
Connections between HDAC activity and cancer have emerged in recent years, stemming from observations that HDAC inhibitors, such as trichostatin A and suberoylanilide hydroxamic acid (SAHA), can induce growth arrest, differentiation, and/or apoptosis in transformed cultured cells (50). In pre-clinical animal models, HDAC inhibitors have demonstrated impressive anti-tumor activity which, in turn, led to several ongoing HDAC inhibitor clinical trials (50 -53). The data presented here, along with data regarding MTA1 and HP1 cited above, are consistent with the hypothesis that HDAC inhibitors may influence not only primary tumors but also distant metastases.
Interestingly, BRMS1 appears to be part of a protein family in which all of the characterized members are components of the mSin3⅐HDAC complex. During the original yeast two-hybrid screen, two cDNA clones identified as FLJ00052 were identified in the prostate library. As studies were underway to follow-up RBP1, mSin3, and HDAC findings, FLJ00052 was re-designated by GenBank TM as mSds3, the mammalian ortholog of Saccharomyces cerevisiae Sds3. Sds3 has been implicated in gene silencing through a Sin3⅐Rpd3 pathway (Rpd3 in a yeast HDAC1 ortholog) and is an integral component of the yeast Sin3⅐Rpd3 complex that is required for histone deacetylase activity (17,54). BRMS1 shares 18% identity and 49% similarity with a large region of yeast Sds3 and 23% identity and 49% similarity with mSds3. mSds3, analogous to its yeast ortholog, is a component of the mSin3⅐HDAC complex, stabilizes HDAC1 within the complex, and augments HDAC activity (17). Another predicted mammalian protein of unknown function (designated MGC11296) is homologous to both Sds3 and BRMS1. Homology to BRMS1 is particularly strong (58% identity; 79% similarity for the C-terminal 196 amino acids of BRMS1 and the N-terminal 196 amino acids of MGC11296). The high level of sequence similarity between these molecules, combined with their associations with mSin3⅐HDAC complexes, suggests the existence of a BRMS1 family of proteins that may play a crucial role in altering metastasis by regulating the so-called histone code (29,49).
Although specific role(s) for BRMS1 within mSin3⅐HDAC complexes remain to be elucidated, the following lines of evidence suggest that the metastasis suppressor may be involved in recruiting and stabilizing HDAC1 and/or modulating HDAC activity. 1) BRMS1 forms small complexes (ϳ100 kDa and greater) with HDAC1 but forms only large complexes (ϳ1.4 to 1.9 MDa) with Sin3B and SAP30 (Fig. 8B). 2) BRMS1 has distinct binding site(s) for the HDAC1⅐RbAp46/48 core subunit as compared with the rest of the complex (mSin3A, mSin3B, SAP30, HDAC2, and RBP1) as demonstrated by BRMS1-(⌬204 -246) binding less effectively to HDAC1⅐RpAp46/48 than does full-length BRMS1; in contrast, BRMS1(⌬204 -246) binds the remaining complex components as effectively (Fig. 4). 3) The C-terminal 42 amino acids of BRMS1 appear to stabilize HDAC1⅐RbAp46/48 within the complex, as deletion of these residues specifically compromises binding to these three components (Fig. 4). 4) Both characterized BRMS1 family members (Sds3 and mSds3) are required for optimal HDAC activity, and mSds3 specifically stabilizes HDAC1 within the mSin3 complex.
Although remarkably similar in breast carcinoma and melanoma cell lines, BRMS1⅐mSin3⅐HDAC complexes were distinct. RbAp46 complexes with BRMS1 were not detected in MDA-MB-231 (Fig. 5), and the interaction with RbAp48 appeared less robust than in C8161.9 (compare Figs. 4 and 5). Differential binding of BRMS1(⌬204 -246) to the HDAC1-RbAp46/48 subunit in C8161.9 was not observed in MDA-MB-231 (compare Figs. 4 to 5). At this juncture, it is not possible to distinguish whether the differences are due to cell origin or presence of mutations that abrogate interactions of RbAp46 with BRMS1⅐mSin3a. BRMS1-transfected MDA-MB-231 cells are suppressed for metastasis less than C8161.9 (40 -90 versus 90 -100%). It is tempting to speculate that differences in metastasis suppression may be related to a differential interaction between BRMS1 and the HDAC1-RbAp46/48 subunit.
Preliminary data obtained with the BRMS1 deletion mu- FIG. 9. BRMS1 represses transcriptional activity in vivo. Using a luciferase reporter assay containing four GAL4 binding sites upstream of the myelomonocytic growth factor minimal promoter, BRMS1 strongly repressed (ϳ80%) basal transcription compared with the pBIND vector alone. tants reported here are consistent with a correlation between complexes involving BRMS1, mSin3, and HDAC and metastasis suppression. C8161.9.BRMS1(⌬164 -246) and (⌬91-246) clones (mSin3 interactions severely impaired or lost; Fig. 4) fail to suppress metastasis (data not shown). However, more refined BRMS1 mutants will be required to determine whether binding to the mSin3⅐HDAC complex is necessary for metastasis suppression. Systematic site-directed mutagenesis of BRMS1 coupled with metastasis assays are underway.
In summary, the metastasis suppressor BRMS1 is shown here to interact with enzymatically active mSin3⅐HDAC complexes. BRMS1 is also shown to form smaller complexes with HDAC1 and to repress transcription when recruited to a promoter region. Besides defining a milieu in which BRMS1 works within cells, the data presented here imply that specific downstream mediators, regulated in part by HDAC activity, are critical to controlling metastatic behavior. Indeed, preliminary cDNA microarray and proteomic studies have identified a limited number of BRMS1-regulated genes. 2 Understanding the role(s) of BRMS1⅐mSin3⅐HDAC complexes in the regulation of gene expression promises to provide insights into metastasis suppression, HDAC-mediated chromatin regulation, and BRMS1 physiology in noncancerous cells.