Transcription factors and nuclear receptors interact with the SWI/SNF complex through the BAF60c subunit.

Transcriptional activity relies on coregulators that modify chromatin structure or serve as bridging factors between transcription factors and the basal transcription machinery. We identified a new coregulator of peroxisome proliferator-activated receptor gamma, BRG1/Brm-associated factor of 60 kDa, subunit c2 (BAF60c2), in a yeast two-hybrid screen of a human adipose tissue cDNA library. BAF60c2 represents a new isoform of BAF60c, a component of the SWI/SNF (mating type switching/sucrose non-fermenting) chromatin remodeling complex. This new isoform as well as the previously identified protein, renamed BAF60c1, is localized primarily in the cell nucleus and is expressed in a wide variety of tissues. Both BAF60c isoforms bind to several nuclear receptors and transcription factors of various families. BAF60c proteins interact in a ligand-independent manner with peroxisome proliferator-activated receptor gamma and enhance its transcriptional activity. Both isoforms are enriched in the central nervous system and also modulate the transcriptional activity of retinoic acid-related orphan receptor alpha1. In conclusion, BAF60c represents a new coregulator that constitutes an important anchoring point by which the SWI/SNF complex is recruited to nuclear receptors and other transcription factors.

In eukaryotic cells the structure of chromatin has been shown to repress gene activation, and both remodeling and repositioning of nucleosomes are necessary to allow binding of transcription factors and the formation of the transcriptional preinitiation complex (1). Chromatin modifying proteins and complexes have been classified in two main groups. The first one comprises factors that covalently modify histone tails (2)(3)(4)(5). The second group of chromatin modifiers comprises complexes that change the location or conformation of nucleosomes and remodel chromatin in an ATP-dependent manner (6 -8). These ATP-dependent chromatin remodeling complexes have been classified based on their catalytic core ATPase subunit (9). They are involved in transcriptional control as well as in DNA replication, DNA repair, and recombination. The large multiprotein SWI/SNF (mating type switching/sucrose non-fermenting) complexes are one subfamily of these remodeling factors (10,11). The SWI/SNF complexes are evolutionarily conserved from yeast to human, and all complexes contain a core set of conserved components, including BRG1/Brm-associated factors (BAFs) 1 (10) and a DNA-dependent SWI2/SNF2-like ATPase, which enables chromatin remodeling. In man, two SWI2/SNF2like ATPases have been described and called hBrm (or hSNF2␣) and BRG1 (or hSNF2␤). Complexes containing either hBrm or BRG1 can act as transcriptional activators, but there is also growing evidence for a role as repressors (12)(13)(14)(15).
The peroxisome proliferator-activated receptor ␥ (PPAR␥ or NR1C3) is one of the three PPARs that together constitute a distinct subfamily of the nuclear receptors. PPAR␥ has mostly been studied because of its key role in adipocyte differentiation, where it acts as a master controller of the "thrifty gene response" (16,17), but it has many additional functions (18). PPAR␥ heterodimerizes with the retinoid X receptors (RXRs) and is activated by naturally occurring fatty acids or fatty acid derivatives (19). In addition to these natural PPAR␥ ligands, several classes of synthetic PPAR␥ agonists have been described including the thiazolidinediones, which are potent insulin sensitizers used in the treatment of type 2 diabetes mellitus (20,21). To activate transcription, PPARs, like most transcription factors, rely on coregulators that modify chromatin (22,23). Coregulators partially determine the specificity of action of nuclear receptors and integrate their different activities to orchestrate a specific cellular response.
We characterized here a new coregulator isolated from adipose tissue using a yeast two-hybrid screen with the ligand binding domain of PPAR␥ as bait. This regulator represents a novel isoform of BAF60c (BRG1-associated factor of 60 kDa, subunit c) that we named BAF60c2. BAF60c2 as well as the previously identified protein, renamed BAF60c1, is part of a protein complex with BRG1 and is broadly expressed in the cell nucleus where it binds to multiple nuclear receptors in a ligand-independent manner and enhances its transcriptional activity. BAF60c1 and -c2 also bind to various other (but not all) transcription factors. BAF60c does not seem to affect adipocyte differentiation and cell proliferation.
Construction of the Yeast Two-hybrid Library and Screening-Adipose tissue was obtained from a female non-obese adult subject undergoing endoscopic cholecystectomy after informed consent was obtained. The local ethics committee of the CHU in Lille approved the project. cDNA library preparation, library screening, and prey construct purification were done according to the manufacturer (Stratagene). To create the "bait" vector the DE domains of PPAR␥2 (residues 179 -505) were cloned downstream of the DNA binding domain of Gal4. YRG-2 yeast (Stratagene) were sequentially transformed with the bait construct and with the library and grown on the appropriate selective medium in the presence of the PPAR␥ ligand rosiglitazone (1 M).
In Situ Hybridization Experiments-CD1 embryos from 11.5 to 16.5 days post coitum and C57B6 adult brains were directly embedded in cryomatrix (Shandon, Pittsburgh, PA). In situ hybridization were performed as described (47). A portion of mouse BAF60c cDNA in pBS was linearized by XhoI and EcoRI. Antisense mBAF60c mRNA was synthesized using T7 polymerase (Promega, Madison, WI). A sense RNA probe for mBAF60c (synthesized using T3 polymerase) was used as negative control (data not shown).
Cell Culture, Transfections, Adipocyte Differentiation, Retroviral Infection, RNA interference (RNAi)-Cell lines were maintained at 37°C, 5% CO 2 according to the supplier's instructions (ATCC, Manassas, VA). Cells were transfected with LipofectAMINE (Invitrogen) in 6-or 24-well plates. Empty expression vectors were used to maintain equivalent amounts of DNA in the transfections. Luciferase and ␤-galactosidase activities were measured as described (44). Graphs represent the means of three independent experiments in which each point was performed in triplicate. Differentiation of 3T3-L1 cells into adipocytes and oil-red-O were performed as described (48) Retroviral infection of 3T3-L1 or NIH3T3 cells was performed as described (49). Briefly, the recombinant retroviral vectors pLP-LNCX, pLP-LNCX-hBAF60c1, and pLP-LNCX-hBAF60c2 were transiently transfected in 293 cells stably expressing Moloney gag and pol (293gp). Viral supernatants were collected 48 h after transfection and added on 3T3-L1 or NIH3T3 cells for 5 h in the presence of Polybrene (4 g/ml). 48 h after infection cells were selected with neomycin.
The RNAi experiment was performed using the pSUPER RNAi system TM (OligoEngine, Seattle, WA). Briefly, a double stranded oligonucleotide targeting nucleotides 1388 -1406 (5Ј-gATCCCCggTgAT-gACAgATgTggCATTCAAgAgATgCCACATCTgTCATCACCTTTTTg-gAAA-3Ј) of mBAF60c mRNA was cloned in the BglII/HindIII site of the pSUPER vector. Stable transfection was carried out on 3T3-L1 or NIH3T3 cells using 6 g of the empty pSUPER or BAF60c RNAi vector, 0.6 g of pPur vector, and LipofectAMINE. Puromycin (1 g/ml) was then added in the medium for selection.
The GST fusion proteins were prepared as described (44). Pull-down assays were performed with in vitro 35 S-radiolabeled translated proteins (TNT Quick Rabbit Reticulocyte, Promega) or purified proteins (His-PPAR␥E) as described (44). Ligand was added when indicated. Immunoblotting was performed as described (44).
Nuclear extracts for immunoprecipitation were prepared from HeLa cells as described (50). Immunoprecipitations were performed with antibodies directed against various parts of BAF60c, PPAR␥, or BRG1 (2SNF-2E12). Mock immunoprecipitations were performed with preimmune serum.
Generation of Polyclonal Antibodies and Immunofluoresence-Peptides corresponding to aa 1-13 of hBAF60c1 (MTPGLQHPPTVVQ), 4 -20 of hBAF60c2 (DEVAGGARKATKSKLFE), and 293-310 of hBAF60c2 (KTNRLQDSHDKEYINGDKY, common to both proteins) were synthesized with an additional cysteine at their extremity. These peptides were coupled to activated ovalbumin following the manufacturer's protocol (Imject® maleimide-activated ovalbumin, Pierce) and were injected into rabbits in the presence of Freund's adjuvant to obtain rabbit sera containing antibodies against hBAF60c1, hBAF60c2, or hBAF60c C-terminal part.
For immunofluoresence cells were grown on cover slips and incubated with antibodies after fixation and permeabilization with methanol. Preparations were then incubated with Texas Red-conjugated antimouse IgG and fluorescein isothiocyanate-conjugated anti-rabbit IgG.
Cell Counting and 5-Bromo-2Ј-deoxyuridine (BrdUrd) Incorporation Assay-Cells (20000) were plated in 6-well plates and counted in triplicate each day for 4 days. For the BrdUrd incorporation assay cells were plated on cover slips. After 1 night cells were incubated 2 h with BrdUrd, and an immunofluoresence assay was performed using a mouse anti-BrdUrd antibody (DAKO A/S, Glostrup, Denmark). Nuclei were labeled by Hoechst staining. The percentage of cells incorporating BrdUrd was assessed by counting 5 different areas of about 100 cells. Counting was done in three independent experiments, and error bars indicate S.D.

Isolation of BAF60c2 and Genomic
Organization-PPAR␥ is highly expressed in adipose tissue, where its function has been well characterized (16). To isolate new PPAR␥ coregulators we performed a yeast two-hybrid screen of a human adipose tissue cDNA library with the DE domain of PPAR␥2 fused to the DNA binding domain of the yeast Gal4 activator as a bait. The C-terminal DE domain of PPAR␥ encompasses the ligand bind-ing domain and the ligand-dependent activation function AF-2.
One of the constructs isolated in our yeast two-hybrid screen (clone 7) was a cDNA containing a 1452-bp open reading frame encoding a polypeptide of 484 residues. This cDNA was identical to the one coding for the BAF60c protein except that the 39 bp, coding for the 13 first amino acids of this protein, were replaced by 78-bp coding for 26 residues (Fig. 1A). To determine whether the isolated clone could represent an alternatively spliced form of BAF60c, we first reconstituted the genomic structure of BAF60c by comparing its cDNA sequence to human genome sequences available in databases (AC005486 and NM_003078) and identified the intron/exon boundaries in that gene (Fig. 1C). The human BAF60c gene is located on chromosome 7 (7q35-36) and comprises 13 exons and 12 introns (Fig.  1B). The N-terminal domain of clone 7 is encoded by a so far unidentified exon located between exon 1 and 2 of the previously described hBaf60c gene (10) (Fig. 1B). Hence, we renamed the exon encoding the N-terminal part of the Baf60c gene described in the literature as "exon 1A" and the newly identified exon as "exon 1B." Sequencing of the cDNAs obtained by reverse transcription-PCR with two different primer sets confirmed the existence of two different mRNAs for BAF60c. Therefore, our yeast two-hybrid screen allowed us to isolate a splice variant of BAF60c, which we named BAF60c2, encoded by a cDNA of 2004 bp. The nucleotide sequence for the BAF60c2 gene has been deposited in the GenBank TM data base under GenBank TM accession number AY450430. The BAF60c protein previously identified (10) was renamed BAF60c1. Its nucleotide sequence has been deposited in the GenBank TM data base under GenBank TM accession number AY450431.
The new BAF60c2 isoform has not yet been described in human or mouse. A data base search revealed, however, the existence of a mouse clone (NM_025891) whose coding sequence is highly similar to hBAF60c2 (93% identity between DNA sequences, with a 100% identity for the first 78 bp; see Fig. 1A) and which might be the mouse counterpart of hBAF60c2. The genomic sequence of mouse chromosome 5 (NW_000225), carrying the mBAF60c gene, also contains a sequence with high homology to exon 1A, but the putative exon 1A in mouse contains 4 additional bases, disrupting the reading frame. The mBAF60c1 protein can, hence, not be produced in contrast to hBAF60c1 and mBAF60c2. The mouse genomic sequence furthermore demonstrates the perfect conservation of exon 1B coding for the N terminus of mBAF60c2 as compared with hBAF60c2 exon 1B.
BAF60c Is Abundantly Expressed and Enriched in the Central Nervous System-Both isoforms of BAF60c were broadly expressed, with the highest levels in brain, testis, and uterus ( Fig. 1, D-E). Strikingly, hBAF60c2 is the predominant isoform in adipose tissue (from which it was cloned), skeletal muscle, lung, heart, and thyroid. hBAF60c1 showed only a higher expression than hBAF60c2 in brain, spleen, and trachea. These expression data extend and refine the expression results previously reported for BAF60c1 (10).
We also performed in situ hybridization experiments to study the expression of BAF60c mRNA during mouse development (Fig. 2). In mouse embryos, BAF60c mRNA expression was detected by reverse transcription-PCR starting from 7 days post coitum (data not shown). In situ hybridization experiments indicated BAF60c mRNA expression in numerous organs in mouse embryos, such as brain (cortex, thalamus, cerebellum), spinal cord, muscle and heart, diaphragm, tongue, stomach epithelium, and olfactory epithelium ( Fig. 2A).
In adult mice (Fig. 2B) mBAF60c mRNA is highly enriched in the brain and cerebellum. In the cerebellum, which is involved in the coordination and control of motor functions, BAF60c is highly expressed in Purkinje cells and in deep cerebellar nuclei, i.e. medial, posterior interposed, and lateral (or dentate) nuclei. Purkinje cells are the principal cells of the cerebellum, which transmit their motor signals via the deep nuclei. Consistent with this, the deep nuclei have been shown to play a role in motor functions as well as sensory-motor learning and memory. BAF60c mRNA is also expressed in cerebellar pedoncule, more particularly, in the vestibular nuclei, such as spinal and medial vestibular nuclei (Fig. 2B), which are involved in equilibrium and motricity. BAF60c mRNA is broadly expressed in the primary and secondary motor cortex. Furthermore, BAF60c mRNA is localized in the hippocampus. A high expression of BAF60c mRNA is also observed in piriform cortex as well as olfactory tubercle (data not shown) and the adjacent anterior olfactory nuclei, i.e. medial and ventral (Fig. 2B). Altogether these data indicate that BAF60c could be involved in various functions in the central nervous system and, in particular, motor and olfactory activities.
Cellular Localization of BAF60c-The presence of a putative nuclear localization signal in BAF60c isoforms (hatched boxes, Fig. 1A) suggested that it functions as a nuclear protein. To evaluate the intracellular localization of BAF60c2, 293T cells were transfected with expression vectors for GFP, GFP-hBAF60c1, or GFP-hBAF60c2 fusion proteins and visualized by fluorescence microscopy (Fig. 3A). GFP-BAF60c1 and GFP-BAF60c2 proteins were detected exclusively in the nucleus of 293T cells (middle and bottom panels), whereas the GFP protein was evenly distributed between the nucleus and the cytoplasm (top panel). This nuclear localization was further supported by immunofluorescence studies in transfected cells. We raised polyclonal antibodies directed against specific domains of hBAF60c1 (aa 1-13) or hBAF60c2 (aa 4 -20) or against their common region (aa 292-310) (see the asterisks for localization of the epitope, Fig. 1A). COS-1 cells were transfected with expression vectors for hBAF60c1 or hBAF60c2. The three antibodies are highly specific for their respective targets. The overexpressed BAF60c proteins were invariably detected in the cell nucleus (Fig. 3B). Despite the fact that some endogenous BAF60c mRNA was present in these COS-1 cells (data not shown), no endogenous BAF60c protein could be detected with our antibodies. Altogether, these results suggest that both BAF60c isoforms are primarily nuclear proteins.
BAF60c Establishes Multiple Contacts with PPAR␥ and Anchors the SWI/SNF Complex to PPAR␥-To confirm the interaction between BAF60c and PPAR␥ detected in the yeast twohybrid screen, we produced GST fusion proteins consisting of full-length BAF60c1 or BAF60c2 or regions of BAF60c (Fig.  4A). The quality and quantity of the GST fusion proteins used for these experiments were first verified by Coomassie staining of these proteins separated on denaturating gels (data not shown). In vitro interactions between these various GST fusion proteins and a His-tagged version of the E domain of PPAR␥, His-PPAR␥E, were assayed in pull-down experiments. BAF60c1 and BAF60c2 full-length as well as the N-terminal part of BAF60c2 (BAF60c2Nt) interacted with His-PPAR␥E (Fig. 4B, top panel, lanes 3-8). The interaction occurred both in the absence as well as in the presence of 1 M rosiglitazone, which was in sharp contrast to the ligand-dependent interaction between the GST-p300 protein and His-PPAR␥E (Fig. 4B, top panel, lanes [11][12]. Both isoforms of BAF60c contain in their common C-terminal domain two LXXLL motifs (see Figs. 1A and 4A), which are consensus motifs for the ligand-dependent interaction of some coregulators with nuclear receptors. However, a fusion protein containing the C-terminal domain of BAF60c with only one of these LXXLL motifs (BAF60c2Ct) does not interact with His-PPAR␥E in our assay (Fig. 4B, top panel,  lanes 9 -10), whereas the N-terminal fragment of BAF60c that lacks these LXXLL motifs still interacts with His-PPAR␥E (lanes 7-8).
We then tested in a pull-down experiment the interaction between in vitro-translated full-length 35 S-labeled PPAR␥2 and BAF60c. When fused to GST both the full-length BAF60c proteins (Fig. 4B, bottom panel, lanes 3-6) as well as their N-terminal domains (Fig. 4B, bottom panel, lanes 7-8) interacted with PPAR␥ in a ligand-independent manner, confirming the results described above with the E domain. The GST-p300 protein interacted in a ligand-dependent manner with PPAR␥ (Fig. 4B, bottom panel, lanes 11-12), as described previously (44). It was noteworthy that, although the C-terminal domain of BAF60c did not interact with His-PPAR␥E (Fig. 4B, top  panel, lanes 9 -10), it showed a weak but consistent interaction with PPAR␥ (Fig. 4B, bottom panel, lanes 9 -10). This result suggests that several interaction domains exist between BAF60c and PPAR␥.
To further consolidate the interaction pattern between BAF60c and PPAR␥ we performed pull-down experiments with various GST-PPAR␥ fusion proteins (Fig. 4C) and in vitro translated BAF60c proteins. The N-terminal part of PPAR␥2 (bAB domain) strongly interacted with both isoforms of BAF60c (Fig. 4D, lane 2). The interaction of the PPAR␥DE fusion protein with BAF60c was again ligand-independent (Fig. 4D, lanes  3-4) and was weaker than the interaction with the bAB domain. This result confirms the existence of multiple contact domains between PPAR␥ and BAF60c.
BAF60c has been described as a subunit of the SWI/SNF complex whose core subunit is either BRG1 or hBrm. We performed co-immunoprecipitations from nuclear extracts to demonstrate the existence of endogenous complexes composed of PPAR␥, BAF60c, and BRG1 in HeLa cells. Nuclear extracts of HeLa cells were prepared, and immunoprecipitations with antibodies directed against BAF60c1, BAF60c2, the C terminus domain of BAF60c, and PPAR␥ were performed. In each case the endogenous BRG1 protein was detected in the immunoprecipitate by Western blot (Fig. 4E, bottom). Immunoprecipitations performed with preimmune sera did not reveal any significant presence of BRG1 (Fig. 4E, top). This confirms the in vivo association of BAF60c, BRG1, and PPAR␥ and indicates that BAF60c anchors the SWI/SNF complex to PPAR␥.
BAF60c Interacts with Other Nuclear Receptors and Transcription Factors-We next tested whether RXR␣ (or NR2B1), the heterodimeric partner of PPAR␥, also interacted with BAF60c. A GST-RXR␣ fusion protein was incubated with various parts of in vitro translated BAF60c (Fig. 5A, BAF60c1Nt, aa 1-728; BAF60c2Nt, aa 1-767) or with the N-terminal part of p300 (p300Nt). GST-RXR␣ binds p300Nt in a strictly liganddependent manner (Fig. 5B, panel e, lanes 2-3). In contrast, the binding between GST-RXR␣ and the BAF60c1 and -c2 fulllength proteins is again ligand-independent (Fig. 5B, panels a  and b, lanes 2-3). Similarly to PPAR␥, RXR␣ interacts also with the N-terminal part of BAF60c (Fig. 5B, panels c and d,  lanes 2-3). We also tested the interaction between BAF60c and various other nuclear receptors. ER␣ (or NR3A1), a nuclear receptor that binds DNA as a homodimer, and the bile acid receptor farnesoid X receptor (FXR or NR1H4) also interacted with GST-BAF60c full-length proteins in a ligand-independent manner (Fig. 5C). A GST-BAF60c fusion protein binds to the retinoic acid-related orphan receptor ␣1 (ROR␣1 or NR1F1), the liver receptor homolog 1 (LRH-1 or NR5A2), and the steroidogenic factor 1 (SF1 or NR5A1) (Fig. 5D, lanes 2-3), whereas the GST protein alone does not interact with these receptors (Fig. 5D, lane 1). To determine whether BAF60c only interacted with nuclear receptors or also with other transcription factors, we performed pull-down experiments with several transcription factors belonging to distinct transcription factor families. We tested the interaction between BAF60c isoforms and the sterol regulatory element-binding protein 1a (SREBP1a), a helix-loop-helix transcription factor, c-Jun, a bZIP (basic region and leucine zipper domain) transcription factor and two homeobox proteins, the pre-B cell leukemia transcription factor 1 (PBX1) and the pancreas/duodenum homeobox-1/insulin promoter factor 1 (PDX-1). In vitro-translated SREBP1a interacted specifically with the two GST-BAF60c fusion proteins but not with the GST protein alone (Fig. 5E,  panel a). This was also the case for the proto-oncogene c-Jun and the homeobox protein pre-B cell leukemia transcription factor 1 (Fig. 5E, panels b and c). However, we found no interaction between GST-BAF60c proteins and PDX-1, another homeobox protein (Fig. 5E, panel d). We hence conclude that BAF60c interacts promiscuously with various nuclear receptors and a diverse array of transcription factors. It is, however, important to note that although BAF60c binds to many transcription factors, it does not bind to all of them.
BAF60c Coactivates Nuclear Receptor-mediated Transcription-To test the possibility that BAF60c acts as a coactivator, CV-1 cells were co-transfected with the proliferative-responsive element driven-promoter construct pREP4-J3-TK-Luc together with expression vectors for PPAR␥2 and BAF60c1 or -c2 or the corresponding empty vector in the presence or absence of rosiglitazone (0.1 M) (Fig. 6A, top panel). PPAR␥ transcriptional activity is stimulated in the presence of rosiglitazone. BAF60c1 and BAF60c2 increase this activation of PPAR␥ in the presence of ligand by 3.4-and 3.9-fold, respectively (Fig. 6A).
To see the influence of BAF60c on different domains of PPAR␥, modified mammalian two-hybrid experiments were performed. CV-1 cells were co-transfected with expression vectors coding for hPPAR␥DE domain fused to the binding domain of the Gal4 yeast transcription factor (BDGal4-PPAR␥DE), increasing amounts of hBAF60c expressing vectors and a Gal4responsive reporter construct. In the absence of any coregula- GST-BAF60c1, GST-BAF60c2, GST-BAF60c2Nt, GST-BAF60cCt, or GST-p300Nt fusion proteins in the presence or absence of 10 Ϫ6 M rosiglitazone (Rosi) when indicated. His-PPAR␥E was detected by Western blot using an anti-PPAR␥ E antibody. C, scheme of the various PPAR␥ proteins used. D, GST-PPAR␥bAB or GST-PPAR␥DE fusion proteins were incubated with in vitro translated hBAF60c1 or -c2 full-length proteins. GST fusion proteins were used bound to glutathione-Q-Sepharose beads, and in vitro translated proteins were 35 Sradiolabeled. Beads were washed, and samples were separated on a 12% SDSpolyacrylamide gel. Radiolabeled proteins were detected by autoradiography. E, BAF60c, PPAR␥, and BRG1 are present in the same complex in vivo. Antibodies directed against PPAR␥, BRG1, or BAF60cCt or specific for BAF60c1 or BAF60c2 or the corresponding preimmune sera were used to immunoprecipitate (IP) the corresponding proteins in nuclear extracts prepared from HeLa cells. A Western blot (WB) was then performed with an anti-BRG1 antibody. BAF60c transactivates PPAR␥ and ROR␣1. A, CV-1 cells were cotransfected in 6-well plates with expression vectors for hBAF60c (200 ng/ well) and PPAR␥ (100 ng/well) and with the pREP4-J3-TK-Luc reporter (1 g/ well). Cells were then grown for 24 h in the presence or absence of 10 Ϫ7 M rosiglitazone. The numbers above the shaded bars indicate the fold induction of the normalized luciferase activity compared with control (Ctrl). RLU, relative luciferase units; ␤-gal, ␤-galactosidase activity; DMSO, dimethyl sulfoxide. The experimental design is schematized on the top of each panel. B, cells were cotransfected with different amounts of expression vectors for BAF60c (0 -1200 ng/well), an expression vector for BDGal4-PPAR␥DE (100 ng/well), and with the pGL3-(GAL 5 )-TK-Luc reporter (500 ng/well). Cells were then grown for 24 h in the presence or absence of 10 Ϫ6 M rosiglitazone. Results are presented as described under A. C, cells were cotransfected with different combinations of expression vectors for BAF60c (500 ng/well), an expression vector for BDGal4-PPAR␥DE (100 ng/well), an expression vector for p300 (500 ng/ well), and with the pGL3-(GAL 5 )-TK-Luc reporter construct (500 ng/well). Cells were then grown and analyzed as described under B. D, COS-1 cells were cotransfected in 24-well plates with expression vectors for BAF60c1 or -c2 (0 -40 ng/ well) and ROR␣1 (50 ng/well) and with the RORE(3)-TK-Luc reporter (200 ng/ well). The numbers above the shaded bars indicate the fold induction of the luciferase activity compared with control. tor, the fixation of BDGal4-PPAR␥DE to Gal4 response elements induces the transcription of the luciferase reporter gene in the presence of rosiglitazone (Fig. 6B). Increasing amounts of BAF60c induced the transcriptional activity of the chimeric BDGal4-PPAR␥DE protein in a BAF60c-dose-dependent manner (Fig. 6B). This observation supports the idea that the BAF60c protein coactivates PPAR␥ in vivo. The activation of the transcriptional activity of PPAR␥ by BAF60c was in the same range (2-3-fold) as the activation induced by p300, a known coactivator of PPAR␥ (Fig. 6C). No additional coactivation was observed when p300 was cotransfected with BAF60c proteins (Fig. 6C). Similar experiments were performed to study the influence of BAF60c on the N-terminal AF-1 function of PPAR␥ using the BDGal4-PPAR␥AB protein. BAF60c was, however, shown not to influence the AF-1 activation function contained within the bAB domain of PPAR␥ (data not shown).
We also verified whether BAF60c influenced ROR␣1 activity. COS-1 cells were co-transfected with a ROR-responsive element driven-promoter construct together with expression vectors for ROR␣1 and BAF60c or the corresponding empty vectors. BAF60c isoforms increase the transcriptional activity of ROR␣1 like that of PPAR␥ in a dose-dependent manner (Fig.  6D). A difference with PPAR␥ is the fact that BAF60c1 was a better coactivator than BAF60c2 for ROR␣.
BAF60c Does Not Influence Adipogenesis or Cell Proliferation-BAF60c2 was identified as a partner of PPAR␥ in adipose tissue. To study an eventual role of BAF60c in adipocyte dif-ferentiation and cell proliferation we changed the expression level of BAF60c using retroviral overexpression and the RNAi technique in the preadipocyte 3T3-L1 cell line. For overexpression of BAF60c we infected cells with retroviruses that encode hBAF60c2 or with an empty control retrovirus, whereas for down-modulation we used stable RNAi targeting BAF60c or the corresponding empty control vector. Stable cell lines were selected with neomycin (for the cells retrovirally infected) or puromycin (when BAF60c expression was down-modulated). Change in expression of the BAF60c mRNAs in the cell lines was confirmed by real-time PCR (Fig. 7A). Experiments were performed with a stable modulation of the expression level of BAF60c in view of the duration of the adipocyte differentiation process, which extends over a week. The different confluent 3T3-L1 stable cell lines were treated with the adipocyte differentiation mix for 2 days and then with insulin alone. After 6 days, all cells displayed a differentiated adipocyte-like phenotype, as evidenced by the lipid droplets in their cytoplasm revealed by oil-red-O staining (Fig. 7B). No significant difference could, however, be detected between the various cell lines (Fig. 7B, Control versus BAF60c). The mRNA expression of lipoprotein lipase (LPL) and aP2, two PPAR␥ target genes, analyzed by real-time PCR was also not significantly different between the different cell lines (Fig. 7C). Hence, stable modulation of the expression of BAF60c in 3T3-L1 fibroblasts does not seem to influence the differentiation of these cells into adipocytes. To evaluate eventual effects of BAF60c on cell proliferation and DNA synthesis, we also modified BAF60c expression level in NIH3T3 fibroblastic murine cells by using the same techniques as previously described for 3T3-L1 cells. No significant differences in cell proliferation were found between the cell lines with either increased or decreased levels of BAF60c expression when cell number was followed for 4 days (Fig. 7D). Levels of DNA synthesis in these NIH3T3 cell lines, as measured by BrdUrd incorporation for 2 h were again not different between the distinct cell lines (Fig. 7E).
In combination, these studies suggest that changes in BAF60c expression level in 3T3-L1 do not influence adipocyte differentiation. Furthermore, changes in BAF60c expression levels in NIH3T3 cells do not influence their proliferation rate. DISCUSSION In an attempt to isolate new coregulators of PPAR␥ activity, we identified, cloned, and characterized a so far unidentified isoform of the BAF60c subunit of the SWI/SNF complex (10) that we named BAF60c2. BAF60c protein is also known as SMARCD3 (SWI/SNF-related, matrix-associated, actindependent regulator of chromatin, subfamily d, member 3). We demonstrate that PPAR␥, BAF60c, and BRG1 are physically present in the same complex in vivo. The N-terminal part of BAF60c was shown to bind to the C-terminal part of PPAR␥, whereas the C-terminal part of BAF60c interacts with the N-terminal part of PPAR␥. Also other nuclear receptors and transcription factors interacted with the BAF60c subunit of the SWI/SNF-remodeling complex. This suggests that these transcription factors recruit the SWI/SNF complex and induce the remodeling of their target genes by contacting multiple SWI/ SNF subunits. Consistent with this, BAF60c was shown to influence the transcriptional activity of PPAR␥ and ROR␣ in a positive manner. Altogether, these results suggest that BAF60c1 and BAF60c2 are genuine coactivators involved in the modulation of the activity of several transcription factors.
The SWI/SNF complex refers to a large family of multisubunit complexes of varying composition (8,14,15). They all contain a core set of conserved components, including some BAF proteins and a DNA-dependent SWI2/SNF2-like ATPase (hBrm and BRG1), which accounts for their chromatin remodeling activity. Complexes containing both hBrm and BRG1 have been characterized as transcriptional activators as well as repressors of transcription. It was discovered more than 10 years ago that the SWI/SNF complex can cooperate with nuclear hormone receptors to activate transcription (24 -27). The mechanisms through which nuclear receptors recruit SWI/SNF complexes to their target promoters are multiple and include a direct interaction of hBrm (29), BRG1 (29,31), and BAF57 (30) with ER and an interaction between BAF250 (33) or BAF60a (41) and glucocorticoid receptor. Our current study adds BAF60c to the list of proteins of the SWI/SNF complex that interact with nuclear receptors.
PPAR␥ is one of the key factors triggering adipogenesis. Recent studies showed that PPAR␥ activity in adipocytes is highly influenced by a number of coregulators. More particularly, the coregulators cAMP-response element-binding protein (CREB)-binding protein (51), steroid receptor coactivator 1/transcriptional intermediary factor 2 (52), PPAR␥ coactivator 1 (53) and thyroid hormone receptor-associated protein (54) have been shown to impact adipocyte activity. Interestingly, a recent study suggests that recruitment of the SWI/SNF complex is important for adipogenesis (55). Hence, we hypothesized that the BAF60c protein, isolated in an adipose tissue DNA library, could act as a bridging factor between PPAR␥ and SWI/SNF and could represent a regulatory element in this process. This was also suggested to be relevant since it has been shown that one BRG1-containing complex called polybromo, BRG1-associated polybromo, BRG1-associated factors (PBAF) is necessary for PPAR␥ to induce transcription on chromatinized templates in vitro (39). However, 3T3-L1 cells in which BAF60c expression was altered differentiate in vitro at the same rate and to the same extent as cells where these coregulators are expressed at the normal level. This suggests either that the amount of endogenous BAF60c is adequate to achieve SWI/SNF recruitment by PPAR␥ or that other redundant proteins can anchor the SWI/SNF complex to PPAR␥ independent of the level of BAF60c in these cells. We also demonstrated that changes of BAF60c expression in NIH3T3 cells do not influence cell proliferation and DNA synthesis. It is possible that BAF60c assists PPAR␥ in other processes or other cell types but is not required for the processes of adipogenesis and cell proliferation. Although we were, hence, unable to identify a clear biological function for BAF60c using a well defined cell proliferation and differentiation system, an in vivo approach focusing on organs/tissues expressing high amounts of BAF60c might shed light on its action.
Expression studies on BAF60c suggest that its activity is not limited to coactivation of PPAR␥. In fact, multiple tissues that express no or little PPAR␥ are highly enriched in BAF60c. This is consistent with our studies that demonstrate that BAF60c interacts with multiple nuclear receptors (including RXR␣, ROR␣, ER␣, FXR, steroidogenic factor 1, and liver receptor homolog 1) and other transcription factors belonging to distinct families, such as certain homeobox, bZIP, and helix-loop-helix transcription factors. Especially interesting was the high level of BAF60c expression found in brain, skeletal muscle, and heart, suggesting that BAF60c may carry out a particular function in these tissues. It is tempting to speculate that BAF60c could in fact be relevant for higher brain functions coordinated by the cerebellum and cortex. In fact, in the adult cerebellum BAF60c is expressed in nuclei that are involved in motor functions as well as in sensory motor learning and memory. The cerebellar pedoncule coordinates equilibrium and motricity, whereas the hippocampus has a primary role in cognition (memory and learning). Therefore, future work should focus on identifying those signaling pathways and transcription factors that crucially depend on BAF60c (and the SWI/SNF complex) for activity. This is a major challenge that will require a systematic approach combining studies in cellular systems, but more important is an in vivo approach using well defined animal model systems.
In summary, we report a functional interaction between several transcription factors and BAF60c isoforms of the SWI/ SNF complex and demonstrate that BAF60c influences transcriptional activity of nuclear receptors, more particularly PPAR␥ and ROR␣, in a positive way. Our data also indicate that BAF60c isoforms do not influence adipocyte differentiation and cell proliferation.