BRG1 Chromatin Remodeling Activity Is Required for Efficient Chromatin Binding by Repressor Element 1-silencing Transcription Factor (REST) and Facilitates REST-mediated Repression*

Chromatin remodeling enzymes such as SWI/SNF use the hydrolysis of ATP to power the movement of nucleosomes with respect to DNA. BRG1, one of the ATPases of the SWI/SNF complex, can be recruited by both activators and repressors, although the precise role of BRG1 in mechanisms of repression has thus far remained unclear. One transcription factor that recruits BRG1 as a corepressor is the repressor element 1-silencing transcription factor (REST). Here we address for the first time the mechanism of BRG1 activity in gene repression. We found that BRG1 enhanced REST-mediated repression at some REST target genes by increasing the interaction of REST with the local chromatin at its binding sites. Furthermore, REST-chromatin interactions, mediated by BRG1, were enhanced following an increase in histone acetylation in a manner dependent on the BRG1 bromodomain. Our data suggest that BRG1 facilitates REST repression by increasing the interaction between REST and chromatin. Such a mechanism may be applicable to other transcriptional repressors that utilize BRG1.

Repression of genes often involves the establishment and maintenance of chromatin in a structure that prevents transcription. ATP-dependent chromatin remodeling activities have more typically been associated with transcriptional activators, but more recent evidence has implied that they have a role in transcriptional repression. Many of these chromatinmodifying enzymes do not bind DNA specifically, and although it is possible for chromatin to be modified in a global manner (1), more often the enzymatic activity is targeted to appropriate genes by site-specific transcription factors via complexes containing multiple proteins.
The repressor element 1-silencing transcription factor (REST), 3 also known as neuron-restrictive silencing factor (NRSF), is a 116-kDa C 2 H 2 Krüppel-type transcription factor that acts as a repressor of numerous genes, many of which are involved in neuronal function (2,3). However, this protein has varied and complex functions that extend into neurogenesis (4 -6), cardiogenesis (7), and oncogenesis (8,9). REST binds to the 21-bp repressor element 1 (RE1), also known as neuron-restrictive silencing element (NRSE), present in the regulatory regions of its target genes. REST was first identified because of its binding to a negatively acting DNA regulatory element in the 5Ј regions of the genes encoding two proteins that are important for neuronal function, the voltage-dependent sodium channel type II (NaV 1.2 , encoded by SCN2A2) and the growthassociated protein, superior cervical ganglion 10 (SCG10, encoded by STMN2) (2,3). Recently, we have identified putative RE1s in more than 1000 loci within the human genome; these sequences are responsible for the negative regulation of genes that code for proteins involved in many cellular functions (10,11).
REST has been shown to interact with a number of proteins, many of which are required for its repressor function, including histone deacetylases HDAC1 and HDAC2 (12)(13)(14), SWI/SNF components BAF57, BAF170, and BRG1 (15), an H3-K4 histone demethylase, LSD1 (16), and the H3 K9 histone methyltransferase G9a (17). Even though many of these proteins have known repressor functions, BRG1 is more commonly associated with transcriptional activation. However, some studies have suggested that BRG1 is required for repression, for example, in Rb-E2F-(18) and p65-mediated repression (19) and in the repression of the c-fos (20), CAD (21), and TRPO (22) genes (18,21,22). BRG1 has also been found in corepressor complexes associated with the transcriptional repressor REST (15), and tumor cell lines lacking BRG1 activity show increased expression of some REST-repressed genes (23). Despite the evidence of a role for BRG1 in transcriptional repression, the mechanisms underlying this repression have not been identified. Here we show that the chromatin remodeling activity of BRG1 is required for efficient REST occupancy at some RE1 sites within chromatin, suggesting that BRG1 acts to facilitate REST binding to promoters. This BRG1-facilitated occupancy is dependent on chromatin remodeling activity and is enhanced by increased histone acetylation via the BRG1 bromodomain. This is the first description of a mechanism by which a chromatin remodeling enzyme facilitates repression by a transcription factor and is likely to be applicable to other transcriptional repressors that utilize BRG1.
Extraction of Nuclear Proteins and Western Blot-Nuclear protein extracts from HEK293 cells were prepared essentially as described by Andrews and Faller (24). Protein concentrations were determined by Bradford assay, and 20 g of each extract was boiled for 10 min in 2ϫ SDS loading buffer (100 mM Tris-HCl, pH 6.8, 4% SDS, 200 mM dithiothreitol, 20% glycerol, 0.2% bromphenol blue) and cooled on ice. Denatured protein samples were electrophoresed through an SDS-6% (w/v) polyacrylamide gel. Proteins were transferred to a Hybond C Extra membrane (Amersham Biosciences), and the membrane was screened with 1:1000 anti-BRG1 (25) in 5% (w/v) milk powder in PBS-Tween 20 0.1% (v/v) followed by 1:2000 secondary horseradish peroxidase-conjugated anti-rabbit antibody (Santa Cruz Biotechnology). Bands were visualized using an ECL Plus detection kit (Amersham Biosciences), and the membrane was exposed to x-ray film (Kodak). To control for equal protein loading, membranes were stripped in 100 mM 2-mercaptoethanol, 2% SDS, 62.5 mM Tris-HCl, pH 6.7, for 30 min at 50°C, washed in PBS-0.1% Tween 20, and reprobed with an antimouse TAF1 antibody (Upstate Biotechnology).
Co-immunoprecipitation-Cells were washed twice with PBS, resuspended in extraction buffer (10 mM Tris-HCl, pH 8.0, 0.15 M NaCl, 5 mM EDTA, 0.5% Triton X-100), and passed through a 23-gauge needle 10 times. Extracts were centrifuged for 5 min at 13,000 rpm at 4°C, and equal aliquots of the supernatant were incubated with a sham antibody (preimmune serum) or an anti-REST antibody (26) overnight at 4°C. Antibody complexes were precipitated with protein G-Sepharose and washed twice with extraction buffer. Proteins were eluted in SDS loading buffer along with 10% input and subjected to Western blotting with an anti-BRG1 antibody (Abcam).
Preparation of BRG1 DN Plasmid-the FLAG epitope-tagged BRG1 DN construct was excised from pBS.CehBRG1K798R (a kind gift from Dr. Tony Imbalzano, University of Massachusetts) using ClaI and SpeI and ligated into the pCS2 ϩ vector digested with ClaI and XbaI.
ChIP-Chromatin was immunoprecipitated from ten 10-cm dishes of HEK293 cells (ϳ1 ϫ 10 8 cells) as follows. Cells were washed in PBS and cross-linked with formaldehyde at a final concentration of 1% in PBS for 10 min at room temperature. Fixation was quenched by the addition of glycine to a final concentration of 125 mM, and the cells were harvested and centrifuged for 5 min at 13,000 rpm at 4°C. The chromatin was fragmented into ϳ500-bp lengths by sonication (Bandelin), with 8 ϫ 30-s pulses, on ice at 30% power, duty cycle 7. The chromatin was precleared with 5% bovine serum albumin-blocked protein G-Sepharose, and 270 l of input chromatin was removed at this point and stored at Ϫ20°C. Samples of chromatin were incubated at 4°C overnight with 5 g of primary antibodies or control IgG or with 10 l of rabbit serum antibodies or normal rabbit serum. Protein G-Sepharose beads were incubated with the chromatin-antibody complexes for 3 h, and the chromatinantibody-bead complexes were collected by centrifugation. The beads were washed twice in wash buffer 1 (20 mM Tris-HCl, pH 8.1, 50 mM NaCl, 2 mM EDTA, 1% Triton X-100, 0.1% SDS), once in wash buffer 2 (10 mM Tris-HCl, pH 8.1, 250 mM LiCl, 1 mM EDTA, 1% Nonidet P-40, 1% deoxycholic acid), and twice in 10 mM Tris-HCl, pH 8.0, and 1 mM EDTA before the chromatin-antibody complexes were eluted with 1% SDS in 100 mM NaHCO 3 . The immunoprecipitated samples, along with the inputs, were de-cross-linked by incubation at 65°C for 6 h. De-cross-linked samples were treated with 0.5 l of 10 mg/ml RNase (0.35 unit) and 9 l of 25 mg/ml proteinase K (6.75 units), and DNA was purified by phenol-chloroform extraction and resuspended in 100 l water.

RESULTS
BRG1 Is Recruited to RE1s by REST-To determine the role of BRG1 in REST repression, we used HEK293 cells, which express functional REST protein (28) and BRG1 (Fig. 1A, lane 1). First we examined the recruitment of BRG1 to known RE1 sites. Chromatin from HEK293 cells was precipitated with the anti-BRG1 antibody or IgG, and the presence of RE1 sites from five well characterized REST-regulated genes (2, 3, 10, 29, 30) was quantified by realtime PCR. BRG1 was enriched at each of the five RE1s under study but not at a control sequence in HEK293 cells (Fig. 1B). BRG1 and REST are present in a single complex, as shown by the co-immunoprecipitation of the two proteins in HEK293 cells (Fig. 1C). To determine whether the BRG1 recruitment to RE1s was mediated by REST, a dominant negative REST construct (REST DN) was transfected into HEK293. REST DN contains only the DNA-binding domain of REST, without the N-and C-terminal repression domains, and inhibits REST function (31,32). Expression of REST DN resulted in a specific reduction in the recruitment of BRG1 at each of the five RE1s under study but had no effect on the control sequence (Fig. 1D). Together, these data suggest that BRG1 is recruited to each of these RE1s and that this interaction occurs via the N or C terminus of REST (Fig. 1D). This observation is consistent with earlier studies that have reported interactions between BRG1 and mSin3A (21,33) and between BRG1 and CoREST (15).
Inhibition of BRG1 Activity Results in Reduced REST Binding to RE1s-Having shown that BRG1 is recruited to RE1 sites, we wanted to determine whether BRG1 is required for REST-mediated repression. To do this, we utilized a BRG1 dominant negative, which contains a point mutation in the ATPase domain (K798R) (34). Accordingly, we transfected HEK293 cells with REST DN or BRG1 DN and examined the effect on the expression of the five RE1-containing genes examined previously. Expression of either REST DN or BRG1 DN resulted in increased expression of each of the five transcripts (Fig. 2, A and  B). We considered the possibility that BRG1 DN may affect expression of RE1-containing genes indirectly by decreasing endogenous REST expression levels. Thus we examined the expression level of REST in response to BRG1 DN. REST levels were not changed in response to BRG1 DN (Fig. 2B). Taken together with the demonstration that BRG1 DN was recruited to each of the RE1 sequences (Fig. 2C), this suggests that BRG1 plays a direct role in facilitating REST repression.
Stable recruitment of MyoD to the myogenin promoter during muscle differentiation requires BRG1 chromatin remodeling activity (35). Thus we hypothesized that the chromatin remodeling activity of BRG1 may also be important for efficient occupancy of REST at RE1 sites within chromatin. To test whether BRG1 chromatin remodeling activity affects the ability of REST to bind RE1s, we examined the effect of BRG1 DN on REST recruitment to RE1 sites. Expression of BRG1 DN resulted in a significant decrease in REST recruitment at the highly enriched SNAP25 and L1CAM RE1 sequences (Fig. 2D). Recruitment at the moderately enriched CHRM4 RE1 was also reduced (although this reduction did not achieve statistical significance), whereas we could not detect any change in REST occupancy at the STMN2 and SCN2A2 RE1 sequences (Fig.  2D). Although the BRG1 DN appeared to have no effect on REST recruitment at the STMN2 and SCN2A2 RE1 sequences, both of these sites were only modestly enriched in control cells, and given the inherent variation with ChIP, we cannot be certain whether the BRG1 DN had any effect at these sites or whether an effect was lost within the experimental variation.
Nevertheless, our data showing that REST occupancy at the SNAP25 and L1CAM RE1 sequences is significantly reduced in the presence of BRG1 DN suggest that BRG1 chromatin remodeling activity is indeed important for REST recruitment to some, if not all, RE1s.
Increasing Histone Acetylation Enhances REST Binding to RE1s in a BRG1-dependent Process-In addition to containing an ATP-dependent nucleosome remodeling activity, BRG1 also contains a bromodomain (36). Bromodomains are present in many chromatin-associated proteins and have been shown to be acetyllysine-binding motifs (37,38). In Saccharomyces cerevisiae the bromodomain of Swi2/ Snf2, the yeast homologue of BRG1, binds chromatin following histone acetylation by SAGA or NuA4 (39). In fact, the bromodomain of Swi2/ Snf2 is important for anchoring the SWI/SNF complex to acetylated promoters (40), and increasing histone H4-K8 acetylation increases Swi2/Snf2 recruitment (41). If BRG1 remodeling activity is involved in the stable recruitment of REST to RE1 sites, then an increase in H4-K8 acetylation levels should result in increased REST recruitment. To test this idea we assessed REST occupancy consequent to inhibition of HDAC activity with TSA. These experiments showed that TSA treatment resulted in an increase in both acetylation of H4 and in REST occupancy at RE1 sequences (Fig. 3, A and B). Inhibition of HDAC activity with TSA results in global hyperacetylation of histones and a consequent increase in "open" chromatin. Accordingly, any increase in REST recruitment may be the result of such an open chromatin configuration rather than being due to an increase in BRG1 activity. To determine whether the increased REST recruitment was mediated by BRG1, we examined the effect of TSA on REST recruitment in cells expressing BRG1 DN. Expression of BRG1 DN resulted in reduced REST recruitment at the SNAP25 RE1 sequence (Fig. 2D), and although TSA had no effect on REST expression (Fig. 3C), it did result in increases in acetylation of H4 (Fig. 3A) and a 5-fold increase in the level of REST recruitment to the SNAP25 RE1 (Fig. 3B). However, in the presence of BRG1 DN, TSA had no effect on the levels of REST recruitment (Fig. 3D), indicating that the acetylation-dependent recruitment of REST requires BRG1 remodeling activity. Given that the bromodomain of Swi2/Snf2 has been shown to be important for binding acetylated histones, we wanted to determine whether the bromodomain of BRG1 was required for the acetylation-dependent increased REST occupancy. To do this, we utilized a BRG1 mutant lacking the bromodo-  BRG1 DN (B). Data were normalized to cyclophilin levels and expressed relative to expression of control. Shown are mean Ϯ S.E., n ϭ 3. C, quantification of RE1 sites from the SNAP25, L1CAM, CHRM4, STMN2, and SCN2A2 genes precipitated by anti-FLAG antibody normalized to precipitation using IgG from control (black bars) and BRG1 DN transfected (hatched bars) cells. D, quantification of the same RE1 sequences from C and a control region precipitated by anti-REST antibody from control (black bars) and BRG1 DN transfected (hatched bars) cells. Data are expressed relative to precipitation using IgG. Shown are mean Ϯ S.E., n ϭ 3. *, indicates p Ͻ 0.05. DECEMBER 22, 2006 • VOLUME 281 • NUMBER 51 main (BRG1 ⌬bromo) (27). As for the BRG1 ATPase point mutant, BRG1 ⌬bromo was expressed in HEK293 in the presence of TSA or a vehicle control and subjected to ChIP. Expression of BRG1 ⌬bromo abrogated the TSA-induced increase in REST recruitment, suggesting that the bromodomain of BRG1, as well as the ATPase domain, is crucial in mediating this effect (Fig. 3D). Together our data show that BRG1 chromatin remodeling activity facilitates REST-mediated repression by increasing REST occupancy. Furthermore this increased occupancy is influenced by the level of histone acetylation in a manner dependent on the BRG1 bromodomain. Thus acetylation levels at REST-regulated promoters will influence the level of REST repression.

DISCUSSION
BRG1 has previously been shown to affect transcription in both a positive and negative manner. Overexpression of BRG1 DN in B22 cells leads to increased and decreased gene expression (35), and in yeast and mammals, the SWI/SNF complex has been shown to interact with both positively acting factors, e.g. SAGA-Gcn5 (42) and p300/CBP (43), and negatively acting factors, e.g. HDAC1 and Sin3-Rpd3 (33). Battaglioli et al. (15) have shown that BRG1 and its associated factors BAF57 and BAF170 interact with the corepressor CoREST and that BRG1 is required for REST-mediated repression. Furthermore, decreased BRG1 function in cancer cell lines has been linked with an increase in expression of REST regulated genes (23). The results presented here show that BRG1 is recruited to RE1s by REST (Fig. 1, B and D) and that inhibition of BRG1 nucleosomal remodeling activity with the BRG1 DN results in a reduction in REST enrichment at some RE1s and a derepression of REST target genes (Fig. 2, B and D). These results suggest that BRG1 activity increases the stability of REST-RE1 interactions in chromatin. BRG1-dependent transcription factor binding has been reported for transcriptional activators such as MyoD, which can bind to the myogenin promoter with weak affinity, but recruitment of BRG1 and its chromatin remodeling activity by MyoD is required to permit the stable binding of MyoD to the promoter along with its cofactors (35). However, enhancement of repressor binding by BRG1 has not been demonstrated previously.
Many complexes that contain chromatin remodeling activity also contain histone acetylases and/or deacetylases, and it would seem that the two activities are linked both in the repression and activation of genes. BRG1 binds chromatin via acetyllysines (39 -41), and inhibition of HDAC activity with TSA resulted in an increase in both histone acetylation and recruitment of REST to RE1s (Fig. 3, A, B, and D). Given that DNA can be made inaccessible to binding proteins by nucleosomes, the role of BRG1 in REST-mediated repression may therefore be to allow REST to gain better access to its chromatin targets by remodeling the local chromatin environment. It may be the case that high levels of REST recruitment (e.g. SNAP25, L1CAM) are observed only at those RE1s that utilize BRG1 activity, whereas lower REST recruitment is observed at RE1s that recruit BRG1 even though BRG1 activity has little effect (e.g. STMN2, SCN2A2). Interestingly, despite the fact that only a small change in REST recruitment at the STMN2 RE1 was observed following expression of the BRG1 DN, it resulted in an 8-fold increase in STMN2 expression. Furthermore, expression of REST DN caused only a 2-fold increase in STMN2 expression. This could potentially be because of: 1) a role for BRG1 in repression via a mechanism that does not involve affecting REST recruitment; or 2) the presence of a second transcriptional repressor that also interacts with BRG1 or a secondary effect of BRG1 DN on the expression of another transcription factor that regulates STMN2. FIGURE 3. BRG1 facilitates efficient REST occupancy of RE1 sites. A and B, quantification of RE1 sites from SNAP25, L1CAM, CHRM4, STMN2, and SCN2A2 genes and a control region precipitated by anti-acetylated H4 antibody (AcH4) (A) or by anti-REST antibody (B). C, quantification of mRNA for REST and REST DN following incubation with 100 nM TSA compared with vehicle control. Data were measured by quantitative reverse transcription-PCR and normalized to cyclophilin levels. Shown are mean Ϯ S.E., n ϭ 3. D, quantification of SNAP25 RE1 precipitated by anti-REST antibody in mock-transfected HEK293 (Control) or cells transfected with the BRG1 ATPase mutant (BRG1 DN) or the BRG1 mutant lacking the bromodomain (BRG1 deltabromo). All data are expressed relative to precipitation using IgG; those enrichments obtained with TSA-treated cells were divided by those of vehicle-treated cells to yield -fold changes in REST recruitment with TSA. Shown are mean Ϯ S.E., n Ն 3.
The binding of REST to DNA is highly specific. Its only known binding site is the 21-bp RE1, and yet the presence of this sequence in a gene is not sufficient to result in REST binding in vivo. For example, REST was recruited to the RE1s of the expressed L1CAM and SNAP25 genes but not to the RE1s of the silent BDNF, STMN2, and SYN1 genes in the human glioma cell line U373 (10). The presence of acetylated lysines at particular genes may explain the cell-specific recruitment of REST at different RE1 sites. Similarly, this may explain the distinct profiles of REST recruitment observed in different cell types. Different complements of genes were regulated by REST in three different developmental stages in embryonic stem cells, neural stem cells, and hippocampal neurons (44). Additionally, both BRG1 null mice (45) and REST knock-out mice (31) are embryonic lethal. The fact that changes in both REST (46) and BRG1 activity affect neural stem cell differentiation and gliogenesis (47) is consistent with the notion that BRG1 activity is important in maintaining appropriate REST function.
Our data show that REST occupancy is also influenced by the level of histone acetylation in a manner dependent on BRG1. These data predict that the level of REST repression at any particular gene will be modified by local levels of histone acetylation, providing one potential explanation for the different levels of REST repression seen at different target genes (10,26,32). Despite the fact that BRG1 is thought to be involved in the repression of genes by different transcription factors (reviewed in Ref. 48), the precise role of BRG1 in this repression had not previously been identified. The mechanism proposed here, in which BRG1 remodels chromatin to allow increased interaction of repressors with their binding sites, is likely to be applicable to BRG1-mediated repression by other transcription factors.