BAF60a Interacts with p53 to Recruit the SWI/SNF Complex*

To understand the tumor-suppressing mechanism of the SWI/SNF chromatin remodeling complex, we investigated its molecular relationship with p53. Using the pREP4-luc episomal reporter, we first demonstrated that p53 utilizes the chromatin remodeling activity of the SWI/SNF complex to initiate transcription from the chromatin-structured promoter. Among the components of the SWI/SNF complex, we identified BAF60a as a mediator of the interaction with p53 by the yeast two-hybrid assay. p53 directly interacted only with BAF60a, but not with other components of the SWI/SNF complex, such as BRG1, SRG3, SNF5, or BAF57. We found out that multiple residues at the amino acid 108–150 region of BAF60a were involved in the interaction with the tetramerization domain of p53. The N-terminal fragment of BAF60a containing the p53-interacting region as well as small interfering RNA for baf60a inhibited the SWI/SNF complex-mediated transcriptional activity of p53. The uncoupling of p53 with the SWI/SNF complex resulted in the repression of both p53-dependent apoptosis and cell cycle arrest by the regulation of target genes. These results suggest that the SWI/SNF chromatin remodeling complex is involved in the suppression of tumors by the interaction with p53.

The SWI/SNF complex is an ATP-dependent chromatin remodeling complex that alters the location or conformation of nucleosomes by using the energy of ATP hydrolysis (1). It is recruited to the promoter region of a target gene by the specific transcription factor and remodels the nucleosomes nearby to facilitate or repress transcription (2). The chromatin remodeling activity of the SWI/SNF complex is essential for developmental processes, hormone responses, and tumor suppression (3)(4)(5). The SWI/SNF complex is a multisubunit complex that consists of 9 -12 subunits (6,7). Among the components, BRG1 and BRM are the catalytic subunits with the ATPase domain, which alters the DNA-histone interaction. SRG3, the mouse homologue of human BAF155, is a scaffold protein that controls the stability of the complex by interaction with other components (8). SNF5 and BAF57 have a mediating function of recruiting transcription factors to the SWI/SNF complex (9 -11). BAF60 has three isoforms, BAF60a, BAF60b, and BAF60c. They show different expression patterns among tissues and interact with different combinations of nuclear receptors as a mediator (6,12,13). In the case of BAF60a, not only various nuclear receptors but also nonreceptor nuclear proteins, such as AP-1 and FEZ1, have been shown to associate with BAF60a to utilize the SWI/SNF complex (12, 14 -16). Since BAF60a has been revealed as a mediator with the versatility to interact with various forms of transcription factors and the N terminus has been mapped to be the interface for the interaction, the mechanism of the versatility of BAF60a is becoming an interesting question (12-14, 16, 17).
The chromatin-remodeling mechanism of the SWI/SNF complex has been best described with the transcriptional activity of nuclear receptors. Since chromatin structures play a repressive role in the control of transcriptional regulation (18), nuclear receptors utilize the SWI/SNF complex to unravel and activate the chromatin-structured promoters upon hormone stimulation (19). The requirement of the remodeling activity of the SWI/SNF complex on the transactivity of glucocorticoid receptor (GR) 3 has been demonstrated with the mouse mammary tumor virus (MMTV) long terminal repeat (LTR) promoter (19). The hormone response elements (HRE) in LTR, which are responsible for the transcriptional response to hormones, such as glucocorticoid and testosterone, are concealed in the nucleosomal array when MMTV LTR is stably integrated. GR initiates transcription from the chromatin-structured MMTV LTR by using the SWI/SNF complex on the hormone stimulation. Thus, when the wild-type BRG1 is substituted by the chromatin remodeling-deficient BRG1 mutants or the interaction between GR and the SWI/SNF complex is disrupted, the chromatin-structured MMTV LTR is not remodeled and therefore not activated by hormone stimulation (12,20). Besides this remodeling function of the SWI/SNF complex, general cofactor activity of the complex is also suggested. For example, substitution of BRG1 by the chromatin remodeling-deficient BRG1 mutants, which are not able to induce tran-scription from the stably integrated MMTV LTR, can activate the transiently transfected MMTV-LTR promoter as efficiently as wild-type BRG1 (20). Furthermore, overexpression of SRG3 alone is also capable of enhancing the transcriptional activity of androgen receptor (AR) even in a BRG1-null cell line (21). Thus, a refined system is needed to study the mechanism of the SWI/SNF complex-mediated activation precisely.
p53 is a tumor suppressor protein that plays a critical role in mediating cellular response to stress (22). It is well known that p53 exerts the tumor suppressor effect mainly by its transcriptional activity. Recruitment of many cofactors, such as p300/ CBP, HAT, HDAC, CARM1, and PRMT1, that modify histones is crucial for p53 to regulate the expression of target genes (23)(24)(25)(26). Recently, Lee et al. have shown that the components of the SWI/SNF complex can interact with p53 (27). They also reported enhancement of the p53 activity by overexpression of BRG1 and SNF5 of the SWI/SNF complex using the transiently transfected plasmid reporter. However, it is unclear whether the chromatin remodeling activity exerted by the SWI/SNF complex is actually required, since the transiently transfected plasmid reporter rarely forms a chromatin structure. It is also possible that the cofactor activity of BRG1 or SNF5 to recruit the general transcriptional machinery may enhance the transcriptional activity of p53 (20,21). Therefore, the role of the SWI/SNF complex in the p53-mediated transcription remains to be carefully delineated.
Here, we characterized the interaction and cooperation between the SWI/SNF chromatin remodeling complex and p53 to understand the tumor-suppressing activity of the SWI/SNF complex. By using an episomal pREP4-luc reporter, we found that p53 utilizes the chromatin remodeling activity of the SWI/ SNF complex. Although the core components of the SWI/SNF complex were co-immunoprecipitated with p53, only BAF60a was shown to interact directly with p53, whereas the other components did not in the yeast two-hybrid assays, suggesting BAF60a as a linker between the SWI/SNF complex and p53. After mapping the p53-interacting region on BAF60a in detail, we designed a dominant negative form of it (BAF60a160), which specifically disrupted the SWI/SNF complex-mediated transcriptional activity of p53. The dominant negative form of BAF60a as well as siRNA for baf60a attenuated the p53-dependent apoptosis and cell cycle arrest by inhibiting bax and p21 expression. Collectively, our results demonstrate the cooperative role of the SWI/SNF complex for the p53-mediated anti-tumor functions.

EXPERIMENTAL PROCEDURES
Plasmids and Site-directed Mutagenesis-For cloning of the yeast two-hybrid vector, flag-brg1, myc-srg3, flag-baf60a, ha-snf5, and ha-baf57 were PCR-amplified with primers containing N-terminal tag sequences and inserted into pGBKT7 vector in fusion with the GAL4-DNA binding domain. flag-p53 and myc-srg3 were cloned in frame to pACT2 vector, which has the GAL4 activation domain. Several oligonucleotides were designed for PCR amplification of deletion mutants of baf60a and p53 based on NCBI sequences. The PCR-amplified fragments were inserted in frame into pGBKT7 and pACT2 vector, respectively. For mammalian expression plasmids, flag-p53, ha-baf60a, and ha-baf60a deletion mutants, including ha-baf60a160 and ha-baf60a⌬108 -150, were cloned into pcDNA3 vector, and flag-brg1 and ha-gr were inserted into pCAGGS vector. pcDNA3-AR was a generous gift from Dr. K. Lee (Hormone Research Center, Chonnam National University, Gwangju, Korea). baf60a160 was inserted to MigRI retroviral vector for viral transduction. For the interference of baf60a mRNA, pSilencer1.0-U6 siRNA expression vector was used. 5Ј-AATCGAAATCACAATGCAA-3Ј is the target sense sequence designed for si-baf60a. pTK-Hyg vector was used for selection of cells that are resistant to hygromycin B. pREP4-p53RE-luc and pREP4-MMTV-luc reporter were constructed by inserting p53-responsive elements and the LTR of MMTV, respectively into pREP4-luc episomal vector, which forms chromatin structure when transiently transfected in cells (from Dr. Keji Zhao, NHLBI, National Institutes of Health, Bethesda, MD). baf60a internal deletion mutants were generated using the QuikChange site-directed mutagenesis kit (Stratagene, La Jolla, CA) according to the manufacturer's instructions. All of the genes used in this study were from mouse cDNA. The sequence correctness, orientation, and frame correctness of each plasmid were confirmed by automated sequencing.
Cell Culture and Establishment of Stable Cell Lines-SaOS-2, SW-13, COS-1, 293T, and Phoenix cell lines were cultured in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum. NIH3T3 was cultured in Dulbecco's modified Eagle's medium supplemented with 10% BCS, and the LNCaP cell line was cultured in Iscove's modified Eagle's medium supplemented with 10% fetal bovine serum. For the establishment of NIH3T3-U6 and NIH3T3-U6-si-BAF60a cells that have stably integrated pSilencer-U6 and pSilencer-U6-si-BAF60a vectors, respectively, NIH3T3 cells were split into two 60-mm dishes, 18 -24 h before transfection. Then 1 g of pTK-Hyg vector was co-transfected with 10 g of pSilencer1.0-U6 or pSilencer1.0-U6-si-BAF60a, respectively. 48 h after the transfection, the cells were treated with hygromycin B (200 g/ml). The medium was replaced with fresh medium containing 200 g/ml hygromycin B every 3-4 days, and this selection process lasted for 3 weeks to generate pooled stable cells.
Transfection, Retroviral Transduction, and Luciferase Assay-Transfection experiments were performed by the CaPO 4 method except for the establishment of stable cells and the transfection of the LNCaP cell line. Those cell lines were transfected with Lipofectamine2000 (Invitrogen). Retroviral transduction was carried out as described previously (44). Appropriate mock DNA, an empty vector, was supplemented to adjust the total DNA amount in all groups. For luciferase assay, the luciferase activity was assayed according to the manufacturer's protocol (Promega, Madison, WI). CMV-LacZ plasmid was cotransfected, and ␤-galactosidase activity was measured to normalize the transfection efficiency.
Antibodies, Immunoprecipitation, and Immunoblotting-For immunoprecipitation experiments, whole cell extracts or nuclear extracts were prepared as previously described (28). 500 g of extracts were diluted into 500 l with Buffer X and precleared once with 40 l of 50% protein A-or G-Sepharose (Amersham Biosciences) at 4°C for 1 h (29). The designated antibody was incubated with the supernatant for 3 h at 4°C and then with protein A-or G-Sepharose for 1 h. The immunoprecipitated complexes were washed three times with 500 l of HEGNDT buffer (10 mM HEPES, pH 8.0, 1 mM EDTA, 10% glycerol, 50 mM NaCl, 2 mM dithiothreitol, 0.1% Triton X-100) and once with HEGND buffer (HEGNDT buffer without Triton X-100). The washed pellets were eluted with SDS-PAGE loading buffer and boiled for 4 min. For immunoblot analysis, proteins were resolved by SDS-PAGE and transferred to Immobilon-P membrane (Millipore). Anti-FLAG (M2; Sigma), anti-HA (HA-7; Sigma), anti-BAF60a (Transduction Laboratories, Lexington, KY), anti-␤-actin (AC-15; Sigma), anti-p53 (FL393 and DO-1; Santa Cruz Biotechnology, Inc., Santa Cruz, CA), anti-p21 (BD Biosciences), and annexin V-PE (BD Biosciences) antibodies were purchased commercially, and antisera against SRG3 and BRG1 were raised from rabbits in our laboratory.
Yeast Two-hybrid Assay and Yeast Extract Preparation-After overnight growing on YPDA, a rich plate, at 30°C, the same amount of AH109 yeasts was co-transformed with 1 g of each set of plasmids in pGBKT7 and pACT2. Since this strain cannot synthesize tryptophan, leucine, adenine, and histidine by itself, all of these four nutrients were supplemented exogenously to yield colonies. Tryptophan and leucine were supplemented by the presence of the pGBKT7 and pACT2 vectors, respectively, and adenine and histidine were supplemented through the separate GAL1 and GAL2 reporter when prey and bait interacted. Co-transformed yeast was grown on a selection plate without tryptophan, leucine, adenine, and histidine and incubated at 30°C for a week. The number of colonies that appeared was counted to measure interaction strength. For preparation of yeast extract, we used the urea/SDS method as per the manufacturer's instructions (Clontech). The extracts were analyzed by Western blotting.
Annexin V and Propidium Iodide (PI) Staining and Flow Cytometry-For annexin V staining, cells were washed twice with phosphate-buffered saline and stained with annexin V-PE antibody according to the manufacturer's instruction. For PI staining, cells were washed twice with phosphate-buffered saline, fixed with 70% ethanol for 30 min at 4°C, and then harvested to stain with PI solution for 30 min at room temperature. Flow cytometry was carried out using a FACSCaliber (BD Biosciences), and data were analyzed by CellQuest software (BD Biosciences).

RESULTS
p53 Utilizes the Chromatin Remodeling Activity of the SWI/ SNF Complex-We first tested whether the transcriptional activity of p53 is dependent on the chromatin remodeling activity of the SWI/SNF complex. We examined whether p53 and the SWI/SNF complex interact in the nucleus. SaOS-2 cells, which lack p53 but express the SWI/SNF complex, were transfected with FLAG-p53, and nuclear extracts were subjected to an immunoprecipitation assay with antiserum against SRG3 and BRG1. FLAG-p53 specifically co-immunoprecipitated with both SRG3 and BRG1, the core components of the SWI/SNF complex (Fig. 1, A and B). Next, we investigated interaction between the endogenous proteins. When an immunoprecipitation assay was performed with anti-p53 antibody, both BRG1 FIGURE 1. p53 utilizes the chromatin remodeling activity of the SWI/SNF complex. A, SaOS-2 cells were transfected with expression vector for FLAG-p53. The nuclear extracts (500 g) were immunoprecipitated (IP) without (No Ab) or with (␣-SRG3) anti-SRG3 antiserum and subjected to immunoblot analysis (WB) with antiserum for BRG1 (top) and SRG3 (middle) and anti-FLAG antibody (bottom). B, immunoprecipitation and immunoblot were performed as in A, but the immunoprecipitation assay was carried out with anti-BRG1 antiserum. C, 293T cells were treated with 1 g/ml doxorubicin for 24 h to activate endogenous p53. Subsequently, nuclear extract (500 g) was prepared and subjected to immunoprecipitation with anti-p53 antibody (DO-1). The Sepharose used in preclearing (Pre) was resolved in parallel with 50 g of nuclear extracts (Input) and immunoprecipitants (IP) by SDS-PAGE. Anti-BRG1 antiserum (top), anti-SRG3 antiserum (middle), and anti-p53 antibody (bottom) were used for immunoblotting as indicated. D, the pREP4-p53RE-luc episomal reporter was co-transfected with expression vectors for FLAG-BRG1 (9 g) and FLAG-p53 (3 g) in SW-13 cells as indicated. The luciferase activity was measured, and the relative activities of three independent experiments are presented with S.D. values (lane 1 set as 1-fold). E and F, FLAG-BRG1 (9 g) and FLAG-p53 (3 g) were transfected as indicated below the panels. Cells were harvested, and each sample was split into two separate tubes. Whole cell extracts of one set were Western blotted with anti-BRG1 antiserum (top), anti-FLAG antibody (middle), and anti-␤-actin antibody (bottom) (E). Total RNAs of the other set were extracted and were subjected to RT-PCR with primers for p21 after standardization with ␤-actin primer (F). The numbers in parenthesis mean relative value of the mRNA level of p21. PCR cycles are indicated at the right.

BAF60a Links the SWI/SNF Complex to p53
and SRG3 were co-purified with p53, which was endogenously induced by the treatment of doxorubicin, the DNA-damaging agent (Fig. 1C). These results suggest the nuclear association between the SWI/SNF complex and p53.
Previously, the transcriptional activity of p53 was reported to be enhanced by overexpression of the components of the SWI/ SNF complex (27). However, this enhancement could be due to either the general cofactor activity of the components or the chromatin remodeling activity of the complex. To distinguish between these two possibilities, we employed a pREP4-luc reporter. This episomal reporter has previously been shown to form a nucleosomal structure and stringently require the SWI/ SNF chromatin remodeling activity to be remodeled and activated (32)(33)(34). We constructed the pREP4-p53RE-luc reporter, which contains p53-responsive elements in the promoter. The reporter assay was performed in SW-13 cells, because this cell line has a nonfunctional SWI/SNF complex due to the lack of the BRG1 subunit but is capable to restore functional SWI/SNF complex upon BRG1 expression (35). Transfection of neither p53 (4-fold) nor BRG1 expression vector (2-fold) alone induced a significant amount of luciferase activity from the pREP4-p53RE-luc reporter (Fig. 1D, lanes 2 and 3 versus lane 1). However, when BRG1 was co-expressed with p53, the reporter gene activity was increased 28-fold without increasing the expression level of p53 ( Fig. 1, D (lane 4) and E). This was not due to a p53-independent activation of the reporter by the SWI/SNF complex, because the expression of BRG1 alone did not induce luciferase activity (Fig. 1D, lane 3). Subsequently, we examined if p53-mediated transcription of the target gene actually utilizes the SWI/SNF chromatin remodeling complex. Similar to the result obtained from the pREP4-luc reporter assay, expression of p21, the target gene of p53, was enhanced in the presence of both BRG1 and p53 but not with either p53 or BRG1 alone in the SW-13 cell line (Fig. 1F). These results indicate that the chromatin remodeling activity of the SWI/SNF complex is required for p53 to initiate transcription from the chromatinstructured promoter.
BAF60a Mediates the Interaction between the SWI/SNF Complex and p53-To identify the component(s) of the SWI/ SNF complex interacting directly with p53, we performed the yeast two-hybrid assays between subunits of the SWI/SNF complex and p53. We cloned p53 in the GAL4 activation domain (GAL4-AD)-containing vector (pACT2) and brg1, srg3, baf60a, snf5, and baf57 in the GAL4-DNA binding domain (GAL4-DBD)-containing vector (pGBKT7) in frame. Each cloned plasmid was co-transformed with a matching mock vector, resulting in no false-positive colonies. Then each GAL4-DBD-SWI/SNF component was co-transformed with GAL4-AD-p53. On average, about 1000 colonies appeared only in baf60a construct, but no colonies were observed in any other subunit constructs. Surprisingly, BRG1 and SNF5, which were previously reported to interact with p53 by the glutathione S-transferase pull-down assay (27), did not interact directly with p53 in our yeast two-hybrid assays. In order to exclude the possibility that the transformed brg1 and snf5 constructs were neither expressed nor functional, we carried out the Western blot analysis with brg1-and snf5-transformed yeast extracts. Both GAL4-DBD-BRG1 and GAL4-DBD-SNF5 plasmids were found to be expressed well in the transformed yeasts (data not shown). Furthermore, when we co-transformed each of them with the GAL4-AD-SRG3 which is known to interact directly with both of them (8), colonies were observed as reported, suggesting that they are functional in yeast for the assay (data not shown). These results suggest that the interaction between the SWI/SNF complex and p53 is indirect, except for BAF60a, although other components of the SWI/SNF complex, such as BRG1, SNF5, and SRG3, have also been co-immunoprecipitated with p53 (27) (Fig. 1, A-C). To confirm the interaction between p53 and BAF60a also in mammalian systems, we cotransfected mammalian expression plasmids of HA-BAF60a and FLAG-p53 in the COS-1 cell line and performed reciprocal immunoprecipitation assays with both anti-FLAG and anti-HA antibodies. As shown in Fig. 2A, the interaction between exogenous p53 and BAF60a was confirmed by reciprocal co-immunoprecipitation in mammalian cells. Next, we examined the nuclear association between the endogenous BAF60a and p53. After the treatment of doxorubicin to induce endogenous p53 in the 293T cell line, the nuclear extracts were immunoprecipitated with anti-BAF60a antibody. As shown in Fig. 2B, endogenous p53 as well as BRG1 and SRG3 co-immunoprecipitated with endogenous BAF60a. These results suggest that the interaction between BAF60a and p53 is formed not only in yeast but also in the nucleus of mammalian cells.
Multiple Residues in Amino Acids 108 -150 of BAF60a Are Implicated in the Interaction with p53-We mapped the p53interacting region on BAF60a using the yeast two-hybrid assay. There is the SWIB/MDM2 domain on BAF60a, which is known to fold like the p53-interacting region of MDM2, the master regulator of p53 (36). Thus, we considered this domain as a putative candidate for the p53-interacting region and constructed C-terminal serial deletion mutants of baf60a into the GAL4-DBD vector to perform the yeast two-hybrid assays with GAL4-AD-p53. Until deletion to amino acid 126 from the C terminus, interaction strengths of BAF60a 1-251, BAF60a 1-160, and BAF60a 1-126 with p53 were similar to that of wild-type BAF60a (Fig. 3A). BAF60a 1-113 interacted, but only weakly, with p53, whereas none of the BAF60a 1-107 and fur- ther deletion mutants interacted with p53 ( Fig. 3A) (data not shown). Thus, the amino acid (aa) 108 -126 region of BAF60a appeared to be implicated in the interaction with p53. We next cloned N-terminal serial deletion mutants of baf60a in the GAL4-DBD vector and performed the yeast two-hybrid assays again with GAL4-AD-p53 in order to exclude the possibility that the C-terminal region of BAF60a, including the SWIB/ MDM2 domain, is involved in the interaction with p53. BAF60a 114 -475 without aa 108 -113 even interacted with p53 as efficiently as wild-type BAF60a. Strikingly, despite the lack of aa 108 -126, BAF60a 127-475 and BAF60a 142-475 were able to interact with p53, although weakly, suggesting the presence of an additional interacting region(s) (Fig. 3B). However, BAF60a 151-475 and further deletion mutants did not interact with p53 ( Fig. 3B) (data not shown). Taken together, the aa 108 -150 region of BAF60a, but not the SWIB/MDM2 domain, seems to be involved in the interaction with p53. To rule out the misfolding or insolubilization problem of the terminal deletion mutants, we constructed several internal deletion mutants of baf60a within aa 108 -150 into the GAL4-DBD vector and performed the yeast two-hybrid assays with GAL4-AD-p53. As shown in Fig. 3C, BAF60a⌬108 -150, in which the p53-interacting region was completely deleted, did not interact with p53 at all, whereas none of the partial deletion mutants in this region completely abrogated the interaction with p53. These results correlate well with the results of the N-and C-terminal deletion studies. In addition, BAF60a 84 -160, which contains the entire aa 108 -150 region, was able to interact with p53 as strongly as wild-type despite being the shortest deletion mutant (Fig. 3C). aa 108 -150 region-dependent interaction of BAF60a with p53 was also observed in a mammalian system. The mammalian expression vector for FLAG-p53 was co-transfected with mock, HA-BAF60a, or HA-BAF60a⌬108 -150 expression vector into the COS-1 cell line, and immunoprecipitation assays were carried out with anti-HA antibody. p53 did specifically immunoprecipitate with BAF60a as shown in Fig. 2A, whereas p53 did not co-precipitate with BAF60a⌬108 -150 (Fig. 3D). Taken together, multiple residues within aa 108 -150 of BAF60a are implicated in the interaction with p53.
Down-regulation of BAF60a Reduces the SWI/SNF Complexmediated Transcriptional Activity of p53-To examine the function of BAF60a in mediating the SWI/SNF complex-mediated transcriptional activity of p53, we attempted to down-regulate endogenous BAF60a in NIH3T3 cells that express the functional SWI/SNF complex, including BAF60a. For significant down-regulation of BAF60a in protein level, we established NIH3T3-U6-si-BAF60a cells that stably express si-baf60a and NIH3T3-U6 cells that contain mock vector. NIH3T3-U6-si-BAF60a cells showed decreased BAF60a in pro- FIGURE 3. Mapping of the p53-interacting region on BAF60a. A, the interactions between various deletion mutants of BAF60a and p53 were analyzed by the yeast two-hybrid assay. The schematic drawing of baf60a deletion mutants fused to GAD4-DBD is depicted. The diagonally striped box represents the SWIB/MDM2 domain, and the numbers below the bars represent amino acid residues of each end. The dotted line indicates the border of the p53-interacting regions identified by each set of experiments. Serial C-terminal deletion mutants of baf60a in pGBKT7 vector were co-transformed with pACT2-p53 for the yeast two-hybrid assays. Binding strengths of the deletion mutants of BAF60a for p53 were measured by the colony number relative to wild-type BAF60a and are represented by minus or plus signs on the right under Binding for p53; Ϫ, no colony; ϩ, Ͻ50% of colonies with wild-type BAF60a; ϩϩ, Ͼ80% of colonies with wild-type BAF60a. B, serial N-terminal deletion mutants of baf60a in pGBKT7 vector were co-transformed with pACT2-p53 for the yeast two-hybrid assay, and the binding strength was measured and presented as above. C, internal deletion mutants of baf60a and baf60a aa 84 -160 (BAF60a84 -160) in pGBKT7 vector were co-transformed with pACT2-p53 for the yeast two-hybrid assay. The interactions were measured and presented as above. D, empty, HA-BAF60a, or HA-BAF60a⌬108 -150 expression vector was co-transfected with expression vector for FLAG-p53 into COS-1 cells. At 36 h post-transfection, whole cell extracts (500 g) were prepared and subjected to an immunoprecipitation assay with anti-HA antibody. Immunoprecipitants (IP) and 50 g of input (Input) were immunoblotted (WB) with anti-HA (top) and anti-FLAG (middle and bottom) antibodies.

BAF60a Links the SWI/SNF Complex to p53
tein level compared with NIH3T3-U6 cells (Fig. 5A, top, lane 2 versus lane 1). With these cells, we analyzed the effect of the reduced BAF60a on the SWI/SNF complex-mediated transcriptional activity of p53. The pREP4-p53RE-luc reporter was co-transfected with or without p53 in NIH3T3-U6 and NIH3T3-U6-si-BAF60a cell lines. Although the basal reporter activity was equal in both cells (Fig. 5B, lane 1 versus lane 3), the transcriptional activity of ectopic p53 was significantly repressed in NIH3T3-U6-si-BAF60a cells (Fig. 5B, lane 2 versus  lane 4). To examine the knockdown effect of BAF60a on p53 target gene expression, we treated cells with doxorubicin to activate p53 and measured changes in the induction of p21. In the immunoblot analysis, p21 protein levels were reduced in NIH3T3-U6-si-BAF60a cells, whereas p53 protein induction was not different (Fig. 5A). Because prolonged exposure to doxorubicin induces apoptosis in the NIH3T3 cell line, we investigated if endogenous BAF60a is also required for p53mediated apoptosis. We measured the percentage of annexin V-positive population by flow cytometry in both cells. Doxorubicin-induced apoptosis was repressed in NIH3T3-U6-si-BAF60a cells relative to NIH3T3-U6 cells at the 24 and 48 h time points (Fig. 5C). Because p53-mediated apoptosis was affected by the knockdown of BAF60a, we next examined changes in the expression of proapoptotic p53 target gene, bax (45). When assessed by RT-PCR, enhanced expression of both p21 and bax (Fig. 5D, lane 3 versus lane 1) was clearly suppressed by knockdown of endogenous BAF60a (Fig. 5D,  lane 4 versus lane 3). Taken together, these results suggest that knockdown of BAF60a may repress the transcriptional activity of p53 and thereby inhibit the p53-mediated tumor suppression function.
BAF60a160 Disrupts the SWI/SNF Complex-mediated Transcriptional Activity of p53-To further evaluate the importance of the p53-SWI/SNF complex interaction in the activation of p53-mediated transcription, we attempted to generate a dominant negative mutant of BAF60a. Based on the C-terminal deletion mutant study of Fig. 3A, BAF60a 1-160 (BAF60a160) was tested for its dominant negative activity, because it contains the p53-interacting region and has the potential to compete with the BAF60a in the SWI/SNF complex for binding to p53 (Fig.  3A, BAF60a and BAF60a1-160). This form of BAF60a mutant is known to disrupt the interaction between GR and the SWI/ SNF complex (12). First, we examined whether BAF60a160 inhibits the p53-mediated chromatin remodeling activity of the SWI/SNF complex in the SW-13 cell line. In agreement with the results of Fig. 1D, the transcriptional activity of p53 on the pREP4-p53RE-luc reporter was enhanced about 10-fold by the co-expression of BRG1 in SW-13 cells (Fig. 6A, lane 2 versus lane 1). However, the SWI/SNF complex-mediated transcriptional activity of p53 was repressed by about half when BAF60a160 was co-expressed with BRG1 and p53 (Fig. 6A, lane  3 versus lane 2). Therefore, BAF60a160 acted in a dominant negative fashion and disrupted the SWI/SNF complex-mediated transcriptional activity of p53 in the SW-13 cell line. Next, we examined whether BAF60a160 suppresses the transcriptional activity of p53 in a dose-dependent manner. If BAF60a160 inhibits by competition with the SWI/SNF complex for p53, the transcriptional activity of p53 will be repressed further as the expression of BAF60a160 increases. In NIH3T3 cells, which have the functional SWI/SNF complex, co-expression of a fixed amount of p53 and increasing amount of BAF60a160 resulted in progressive down-regulation of the transcriptional activity of p53 (Fig. 6B, lanes 3-5). This was not due to the reduction of p53 expression, because the p53 expression was not repressed by co-expression of BAF60a160 (Fig.  6C). Similar results were also obtained in the SaOS-2 cell line (Fig. 6D). Therefore, these results indicate that BAF60a160 is a   MAY 2, 2008 • VOLUME 283 • NUMBER 18 dominant negative mutant that suppresses the SWI/SNF complex-mediated transcriptional activity of p53.

BAF60a Links the SWI/SNF Complex to p53
BAF60a160 Does Not Disrupt the Functionality of the SWI/ SNF Complex-Although the deletion studies support the notion that the dominant negative effect of BAF60a160 is probably due to disruption of the linker function of BAF60a between p53 and the SWI/SNF complex, disruption of the intact SWI/ SNF complex by BAF60a160 still remains a possibility. Therefore, we investigated whether the functionality of the SWI/SNF complex is maintained in the presence of BAF60a160. The SWI/SNF complex-mediated transcriptional activity of well characterized GR and AR was examined in the presence or absence of BAF60a160. GR and the SWI/SNF complex are linked by BAF60a, whereas AR and the SWI/SNF complex are mediated by BAF57 (10,12). If BAF60a160 disrupts the functionality of the SWI/SNF complex, the SWI/SNF complex-mediated transcriptional activity of both AR and GR will be disrupted. On the other hand, if BAF60a160 specifically inhibits the linker activity of BAF60a without disrupting the SWI/SNF complex, AR-mediated transactivity will be maintained, whereas GR-and p53-mediated activities are repressed. To test this hypothesis, we first cloned the pREP4-MMTV-luc reporter in which the LTR sequence of MMTV is inserted as a promoter. The LTR sequence of MMTV has response elements for both GR and AR. The chromatin-structured pREP4-MMTV-luc reporter was activated by the SWI/SNF complex-mediated transcriptional activity of GR and AR when dexamethasone (Dex) and dihydrotestosterone (DHT) were treated as ligands, respectively (Fig. 7, A (lane 4 versus lane 3) and B (lane 4 versus lane 3)). However, the transcriptional activity of GR on the pREP4-MMTV-luc was significantly inhibited by co-expression of BAF60a160, whereas that of AR was not (Fig. 7, A (lane  6 versus lane 4) and B (lane 6 versus lane 4)). In AR-positive and the SWI/SNF complex-positive prostatic adenocarcinoma LNCaP cell line, endogenous AR-mediated transcription on the pREP4-MMTV-luc reporter was enhanced by the treatment of DHT (Fig. 7C, lane 2 versus lane 1). However, the transcriptional activity of AR was not affected despite the increasing amount of BAF60a160 expression vector transfected (Fig. 7C,  lanes 4 and 6 versus lane 2). In conclusion, BAF60a160 specifically inhibited the BAF60a-mediated recruitment of the SWI/ SNF complex without disrupting its functionality. These results indicate that BAF60a160 may interfere with the docking of transcription factors to BAF60a of the SWI/SNF complex.
p53-induced Apoptosis and Cell Cycle Arrest Are Also Alleviated by BAF60a160-We next examined if the interaction between p53 and the SWI/SNF complex is also crucial for p53 to exert its tumor-suppressing activity. Introduction of p53 is known to induce apoptosis in p53-deficient SaOS-2 cell line (37). Thus, we analyzed the effect of BAF60a160 on p53-dependent apoptosis in SaOS-2 cell line, in which the transcriptional activity of p53 had also been significantly down-regulated by BAF60a160 (Fig. 6D, lane 3 versus lane 2). The percentages FIGURE 6. BAF60a160 represses the transcriptional activity of p53. A, the pREP4-p53RE-luc episomal reporter was transfected with expression vector for p53 (1 g) in combination with BRG1 (3 g) and BAF60a160 (10 g) in SW-13 cells. Relative luciferase activities due to p53 are presented as a percentage of the SWI/SNF complex-induced activity (lane 2 set as 100%). B, the pREP4-p53RE-luc was transfected with plasmid for mock or p53 in NIH3T3 cells. The effect of BAF60a160 was tested by co-transfecting increasing amounts of BAF60a160 expression vector (2, 5, and 10 g for lanes 3, 4, and 5, respectively) with a fixed amount of p53 expression vector (1 g). C, samples from B were Western blotted (WB) with antibodies for FLAG (top) and BAF60a (bottom) after luciferase assay in the same order as in B, lanes 1-5. D, the pREP4-p53RE-luc reporter was transfected with empty vector or expression vector for p53 in SaOS-2 cells. BAF60a160 expression vector (10 g) was cotransfected with p53 expression vector (2 g) to evaluate the effect of BAF60a160. Relative luciferase activities due to p53 are presented as a percentage of the SWI/SNF complex-induced activity (lane 2 set as 100%). Columns in A, B, and D represent the mean Ϯ S.D. (n ϭ 3). of sub-G 1 population, which represents the apoptotic bodies, were similar at ϳ10% in both mock-and BAF60a160-transfected cells (Fig. 8A, panels 1 and 2). Although the expression of exogenous p53 resulted in enhanced sub-G 1 population (Fig.  8A, panel 3 versus panel 1), the addition of BAF60a160 partially reversed the enhancement (Fig. 8A, panel 4 versus panel). Statistical analysis of three independent experiments yielded a 27% reduction in sub-G 1 population by BAF60a160 (Fig. 8B). p53 protein level was not affected by the co-expression of BAF60a160 (Fig. 8C, lane 3 versus lane 4). We next examined the effect of BAF60a160 on cell cycle progression, since SaOS-2 cells transfected with wild-type p53 are unable to progress into S phase and arrest in G 1 phase (38). Cell cycle distribution of transfected SaOS-2 cells as in Fig. 8A was assessed by PI staining and flow cytometry. As shown in Fig. 8D, mock-and BAF60a160-transfected cells gave similar cell cycle distribution patterns (lanes 1 and 2). However, p53-transfected cells displayed a significant increase in G 1 phase and a decrease in S phase (Fig. 8D, lane 3  versus lane 1), which was partially reversed by co-expression of BAF60a160 (Fig. 7D, lane 4 versus  lane 3). Because such results are probably due to changes in p53-mediated transcription of cell cycle and apoptotic regulators, we next examined the effect of BAF60a160 on the transcription of few known regulators. Expression of p53 alone in SaOS-2 cells resulted in enhanced expression of p21 and bax at the transcript levels (Fig. 8E, lane 3 versus lane 1), but co-expression of p53 and BAF60a160 yielded significantly reduced expression (Fig. 8E, lane 4 versus lane 3). p53-mediated target gene expression was also repressed by BAF60a160.
To investigate the effect of BAF60a160 on p53-mediated apoptosis in a more physiological setting, we introduced BAF60a160 expression vector by the retroviral transduction and induced endogenous apoptosis by treatment of doxorubicin. Since the retroviral construct contained internal ribosome entry site sequence followed by green fluorescence protein (GFP), GFPpositive cells are those transduced with the retroviral vector. When we compared apoptotic induction between the GFP-positive population of vector-and BAF60a160transduced cells at the 24 and 48 h time points, induction of apoptosis was inhibited in BAF60a160transduced cells at both time points (Fig. 9B). However, the degree of apoptotic induction was similar in GFP-negative populations (Fig. 9C).
To further evaluate the physiological relevance of BAF60a in mediating the endogenously induced apoptosis, we established NIH3T3 cells that express low and high levels of BAF60a160 and analyzed the dose-dependent effect of BAF60a160 on doxorubicin-mediated apoptosis (Fig. 9D). When we treated doxorubicin and measured the percentage of annexin V-positive population at the 24 and 48 h time points, doxorubicininduced apoptosis was repressed in correspondence to the expression level of BAF60a160 (Fig. 9E).
Taken together, BAF60a160 acts dominant negatively on p53-mediated apoptosis and cell cycle arrest by attenuating the SWI/SNF complex-aided transcriptional activity of p53 on target regulatory genes. Note that the value of the left column is set as 1. **, significance was p Ͻ 0.01 for t test. C, whole cell extracts were subjected to immunoblotting (WB) with the indicated antibodies. ␤-Actin was used as a loading control (bottom). This is a representative of three similar results. D, only the live cells were gated to determine the percentage of cells in the G 1 , S, and G 2 /M phases of the cell cycle. Note that the sub-G 1 population, which corresponded to M1 gates of A, was excluded in gatings and subsequent calculations. Results shown are representative of similar results. E, total RNAs of A were subjected to RT-PCR with primers specific for the genes shown on the left of each panel. ␤-Actin was used as a loading control. Relative mRNA amounts are represented in parenthesis, and PCR cycles are indicated at the right.

DISCUSSION
Our results demonstrate that p53 requires the interaction with the SWI/SNF complex to induce transcription from the nucleosomal promoter. p53-responsive elements packaged in the nucleosomal structure were hardly activated by p53 until the functional SWI/SNF complex was reconstituted by the restoration of BRG1 in the SW-13 cell line. Notably, this enhancement of transcriptional activity of p53 was dependent on its protein-protein interaction with the SWI/SNF complex. The disruption of the interaction by the down-regulation or by the dominant negative form of the linker significantly reduced the SWI/SNF complex-mediated transcriptional activity of p53 and the p53-dependent effects as a result.
Among p53 target genes, we strongly suggest p21 as a p53 target gene that requires the chromatin remodeling activity of the SWI/SNF complex for p53-mediated transcription. In SW-13 cell line, we investigated few p53 target genes but found no other genes that were dramatically changed like p21. It was previously reported that the reconstitution of the SWI/SNF complex in the SW-13 cell line induces p21 expression independently of p53 (33,40). Although transfection of either p53 or BRG1 yielded minor induction of p21 expression, co-transfection of BRG1 and p53 gave substantial induction of p21 expression (Fig.  1F). Thus, although expression of p21 could also be induced by the SWI/SNF complex independently of p53 in some situations, cooperation between the SWI/SNF complex and p53 seems to be the major mechanism of inducing p21 in response to stresses.
BAF60a, among the core components of the SWI/SNF complex, is shown here as the mediator of interaction between the SWI/SNF complex and p53. We were not able to find direct interaction between p53 and the other core subunits of the SWI/SNF complex, including BRG1 and SNF5, although they have previously been reported to interact with p53 by immunoprecipitation and the GST pull-down assay (27). Our results suggest BRG1 and SNF5 to be co-purified with p53 through indirect interaction mediated by BAF60a in the cell extracts or the reticulocyte lysates (39). Although the involvement of the unexamined components of the SWI/SNF complex cannot be completely excluded, the effect of si-baf60a and the dominant negative effect of BAF60a160 strongly support the bridging role of BAF60a between p53 and the SWI/SNF complex. Unlike the effect of si-baf60a and BAF60a160 on the transcriptional activity of p53, overexpression of wild-type BAF60a did not enhance the transcriptional activity in either the transient reporter or the nucleosomal reporter in various cell lines (data not shown). This result suggests that BAF60a does not have either a cofactor function on p53 by itself or an additive role for the chromatin remodeling activity of the SWI/SNF complex. Taken together, BAF60a seems to be the mediator between p53 and the SWI/SNF complex.
The N-terminal aa 108 -150 region of BAF60a distinct from the SWIB/MDM2 domain was mapped to be involved in the interaction with p53. BAF60a⌬108 -150, which lacks the p53interacting region, did not interact with p53 either in the yeast two-hybrid or immunoprecipitation assays. These results coin- cide with earlier studies that report the N terminus of BAF60a as the interacting region with various transcription factors (12-14, 16, 17). Although a substantial amount of studies report BAF60a as a mediator, no study has mapped the interacting region at the amino acid level. To unveil the mechanism of versatile interaction of BAF60a, we tried to narrow down the p53-interacting region to the amino acid level. Although both the aa 108 -113 and aa 142-150 regions of BAF60a were shown to contain the interacting residue(s) for p53, either deletion of the interacting region or the partial deletion of the interacting region did not abrogate the interaction with p53. Thus, multiple residues in aa 108 -150 of BAF60a seem to be engaged in the interaction with p53.
There is no known domain in the aa 108 -150 region of BAF60a; however, the region is rich with charged residues (40% of aa among aa 108 -150 are charged aa) and is predicted to form a structure that consists of helixes and strands (analyzed with PSIPRED, Protein Structure Prediction Server). Thus, we suggest that other sets of multiple charged residues among aa 108 -150 may also be engaged in the interaction with other transcription factors. Therefore, we can further suggest that BAF60a160 may disrupt the interaction between the SWI/SNF complex and other transcription factors, as well as GR and p53, which are known to be mediated by BAF60a (12). Notably, BAF60a160 inhibited the SWI/SNF complex-mediated activation linked by BAF60a without disabling the SWI/SNF complex. The transcriptional activity of p53 and GR was repressed by BAF60a160, whereas that of AR was not. Thus, the N-terminal fragment of BAF60a seems to be the dominant negative inhibitor that specifically inhibits the interactions toward BAF60a.
Many studies have reported the relationship between cancer and mutations in components of the SWI/SNF complex (5). Lack of certain components within the SWI/SNF complex increases the cancer susceptibility in mice, and mutations in components are being identified in various cancer cell lines (3,(41)(42)(43). This tumor-suppressing activity of the SWI/SNF complex has been elucidated in relation to RB, BRCA1, and c-MYC, but these have not been enough to fully explain the prognosis of mutations of the SWI/SNF components (5,9,30,31). Here, we show that p53 was not able to fully exert its anti-tumor function, induction of apoptosis and arrest of cell cycle, when the interaction with the SWI/SNF complex was disrupted by the dominant negative form of BAF60a or suppressed by baf60aspecific siRNA. Although the relationship between the occurrence of cancer and mutations in BAF60a is not reported yet, our study suggests that mutation in BAF60a may be implicated in the tumorization, particularly in cancers with wild-type p53. In conclusion, our study not only demonstrates BAF60a-mediated interaction between the SWI/SNF complex and p53 but also suggests a tumor-suppressing mechanism of the SWI/SNF complex in relation to p53.