INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Chromatin acts as a barrier to eukaryotic transcription by blocking transcription factor access and polymerase movement. To contend with this barrier, cells make use of a variety of evolutionarily conserved ATP-dependent chromatin remodeling complexes. These complexes have been shown to influence transcription and chromatin access in vivo and in vitro, but the extent of their roles and their mechanisms of action are only beginning to be understood (15, 17, 40, 43).

Human cells contain a family of SWI-SNF complexes that are closely related to the yeast (y) SWI-SNF and RSC complexes. Human (h) SWI-SNF has been implicated in transcriptional activation of several genes (1, 6, 10, 16, 20, 25), as well as in transcriptional repression and growth control through the p105Rb retinoblastoma protein (9, 26, 34, 38). To determine the function of SWI-SNF in gene regulation, we have examined its effects on chromatin in vitro. hSWI-SNF contains at least eight subunits and can be isolated in two forms, which contain either BRG1 or hBRM as a central ATPase subunit (18, 41). These complexes possess DNA- and nucleosome-stimulated ATPase activity and have an ATP-dependent ability to remodel mononucleosome core particles (referred to as nucleosomes or cores herein) (as assayed by changes in DNase I digestion patterns) and plasmid chromatin (as assayed by a reduction in nucleosome-constrained negative supercoils) (18). Each ATPase alone, when purified from insect cells, appears capable of these activities, although at reduced levels (28). Addition of three other conserved subunitsnamely BAF155, BAF170, and Ini1to the ATPase subunit generates a minimal hSWI-SNF complex that is almost as active as the intact complex.

While ATP is required for remodeling by intact hSWI-SNF, the remodeled state is stable in the absence of ATP (13) or SWI-SNF in several in vitro assays. These observations led to the discovery that SWI-SNF can act enzymatically to create a stable, altered dimer of mononucleosome cores (30). This product, which we refer to here as the remodeled dimer, could be separated from hSWI-SNF and was shown to have the proper ratio of the four core histones and DNA but an apparent molecular weight twice that of a normal nucleosome. It was determined to have distinct DNase and micrococcal nuclease digestion patterns and altered susceptibility to restriction enzyme digestion. Increased affinity of GAL4 for this product suggested that it might be more amenable to transcriptional activation. The hSWI-SNF complex also converts this product back to cores, and both the creation of the remodeled dimer and reconversion to cores requires ATP hydrolysis. In general, these activities and properties are shared by hSWI-SNF and the related yeast complexes. Notably, yeast RSC generates and reconverts a highly similar stable product (22), and the remodeling of mononucleosomes by ySWI-SNF is stable after ySWI-SNF removal (8).

Many different chromatin remodeling complexes have been identified to date. Some can be classified as SWI-SNF-like, by virtue of their subunit composition and the similarity of their ATPases to ySWI2-SNF2 (for a review, see reference 17). A second major family contains ISWI or a related protein as its central ATPase (for reviews, see references 11 and 40). The complexes from this ISWI-based family are unlike SWI-SNF-like complexes in that their ATPase activity is stimulated primarily by nucleosomes and not by DNA alone. Nevertheless, ISWI is similar to BRG1 and hBRM in that it can also perform at least some of the in vitro activities of the complex in which it is found (7, 19). A third family, which includes the nucleosome remodeling and deacetylase (NuRD) complexes, contains a CHD/Mi-2 protein(s) as the central ATPase(s) (for a review, see reference 2). All of these complexes have been shown to remodel chromatin in vitro in some of the many available assays. An exhaustive comparison across all assays has not been done, but current data indicate that each remodeling complex can perform only a subset of known remodeling activities (see, e.g., references 17 and 40 for reviews). These activities include those described above, as well as the abilities to disorder arrays of nucleosomes, space nucleosomes evenly along an array, reposition individual nucleosomes, increase restriction enzyme access to nucleosomes, increase transcriptional initiation or elongation, and transfer histone octamers to acceptor DNAs in trans. It is not known how these activities are related at a mechanistic level, nor is it known which components of the complexes are required for each specialized function.

To begin to address these questions, we have examined two activities of the hSWI-SNF complex: the ability to form a remodeled nucleosome dimer (30) and the ability to transfer histone octamers in trans (this report). Recently, the ySWI-SNF and RSC complexes have been shown to possess the ability to transfer histone octamers to acceptor DNAs in trans under certain conditions (23, 42). Here we show that hSWI-SNF has a similar activity. Analysis of this activity and comparison of it to the ability to form remodeled dimers suggest that both products may be formed through a shared mechanism. To address the protein subunit requirements for these reactions, we tested isolated BRG1 and hBRM. Each protein also possesses both activities, suggesting that the enzymatic capabilities of the complex are largely determined by its ATPase subunits.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Protein purification. hSWI-SNF was purified from a HeLa cell line overexpressing the FLAG-tagged Ini1 subunit of the complex, using an M2 anti-FLAG affinity resin (Sigma) as described previously (33). The purity of these fractions was ~70%, as estimated from silver-stained gels. Molarity estimations assumed this level of purity and a molecular mass of 2 MDa. There were trace amounts (<1%, as estimated from silver-stained gels) of histones evident in the SWI-SNF preparation, which might have been in the form of nucleosomes. The FLAG epitope-tagged hSWI-SNF ATPases BRG1 and hBRM were expressed and purified from Sf9 cells by using a baculovirus system as described previously (28). These proteins were ~80% pure as determined by silver staining. Molarity estimations assumed this level of purity and a monomeric molecular mass of 190 kDa. For the experiments presented in Fig. 3, BRG1 was further purified in monomeric form by glycerol gradient centrifugation as follows. To promote self-dissociation, immunopurified BRG1 was incubated on ice in BC buffer (20 mM HEPES [pH 7.9], 10% glycerol, 0.2 mM EDTA, 1 mM dithiothreitol [DTT], 0.2 mM phenylmethylsulfonyl fluoride [PMSF]) containing 1 M urea and approximately 1 M NaCl or KCl for 3.5 h; then it was diluted with BC buffer containing no KCl and no glycerol to reduce the density of the solution, immediately layered onto a 10 to 40% glycerol gradient (50 mM Tris [pH 8.0], 0.1 M urea, 0.1 M KCl, 0.1% NP-40, 1 mM DTT, 1 mM EDTA, 0.2 mM PMSF), and centrifuged for 17 h at 35,000 rpm (~150,000 × g) and 4°C in an SW55 rotor (Beckman). Twenty-two 6-drop fractions were collected from the bottom of the gradient by the use of a butterfly needle. For Fig. 3A, 50 µg of a BRG1 sample was incubated in 158 µl of BC buffer containing 1 M urea, 1 M NaCl, 95 mM KCl, and 3% glycerol. The incubation mixture was then diluted 1.9-fold, and 200 µl of this dilution (containing about 34 µg of BRG1) was layered onto the glycerol gradient. For Fig. 3B, approximately 7.5 µg of BRG1 was incubated in 50 µl of BC buffer containing 1 M urea, 930 mM NaCl, 65 mM KCl, and 13% glycerol. The incubation mixture was then diluted fourfold, and 190 µl of this was layered onto the glycerol gradient.

Nucleosome assembly and labeling. The 155-bp TPT MluI-EcoRI fragment (which contains two 20-bp nucleosome phasing sequences at the MluI end [32]) was used to assemble nucleosomes. It was purified, assembled into nucleosomes, and labeled on the MluI site end with [³²P]dCTP as described previously (30). For nonlabeled TPT nucleosomes, the full-length 155-bp TPT fragment was amplified by PCR and purified by 5% polyacrylamide gel electrophoresis (PAGE) followed by elution by soaking of the excised band and ethanol precipitation. Mononucleosomes were assembled as described above or by gradient dialysis (29), using histone/DNA ratios of 0.9:1 or 1.6:1 (Geeta Narlikar, G. R. Schnitzler, and R. E. Kingston, unpublished data). They were purified by centrifugation in 10 to 30% glycerol gradients prepared with GGB or 50 mM Tris (pH 7.5)-1 mM EDTA-100 mM KCl, and concentrations were determined as described above. Bulk HeLa cell mononucleosomes (0.25 µg) were labeled by using T4 polynucleotide kinase (10 U) and [-³²P]ATP for 30 min at 30°C and 15 min at 42°C in 26 µl of 10% gradient buffer with 7 mM MgCl₂. Unreacted label was removed with a G50 TE spin column. For random-sequence mononucleosome DNA, an identical reaction product was extracted with neutralized phenol before the spin column step.

Remodeling reactions. Experiments involving interconversion of mononucleosome cores and remodeled dimer species were performed as described previously (30) except where noted in the figure legends. In general, dimer formation and octamer transfer reactions were performed at 30°C in 25-µl reaction volumes containing in a solution containing 16 mM HEPES (pH 7.9), 10 mM Tris (pH 8.0), 60 mM KCl, and 5 mM MgCl₂, with or without 2 mM ATP-MgCl₂, 7 to 10% glycerol, 0.1% NP-40, 20 µg of bovine serum albumin (BSA)/ml, 0.16 mM EDTA, 0.8 mM DTT, and 0.16 mM PMSF. Significant variations on these conditions are noted in the figure legends. We have found that these reactions occur efficiently in the following solutes concentration ranges: 2 to 5 mM MgCl₂, 0.1 to 2 mM ATP-MgCl₂, 20 to 70 mM KCl or NaCl, and up to 0.1% NP-40 (13, 28, 30) (this report and data not shown). For gel shift analysis, plasmid DNA or KCl was added at the concentrations indicated in the figure legends to disrupt SWI-SNF interactions with the templates and allow products to be resolved by 5% polyacrylamide-0.5× Tris-borate-EDTA electrophoresis at 4°C (2 h, 200 V). For footprinting analysis, reaction mixtures were treated with DNase for 2 min at room temp and analyzed as described previously (13). Variations are detailed in the figure legends. Quantitation was performed with a PhosphorImager (Molecular Dynamics).

Purification and analysis of reaction products. Scaled-up remodeling or transfer reactions were performed with reduced proportions of glycerol (to allow layering on the gradient). These reactions were stopped by addition of plasmid DNA (which binds SWI-SNF and its ATPases, competing them off of test templates) and separated on glycerol-GGB gradients. The large size of the plasmid DNA caused it to pellet, removing SWI-SNF proteins from nucleosome-containing gradient fractions. Alternatively, reactions were stopped by addition of KCl to 230 mM (which disrupts SWI-SNF-nucleosome interactions [30]) and purified on glycerol-GGB gradients containing 180 mM KCl. Western blot analysis showed that there was some contamination of salt-stopped remodeled dimer fractions with SWI-SNF complexes. These complexes, however, were inactive for remodeling when ATP was added.

Restriction enzyme assay for remodeling activity. End-labeled nucleosomal array 5S-G5E4 (27) (a gift from J. L. Workman) was formed by salt dialysis as described previously (24), with the following modifications. DNA and purified HeLa cell core histones were mixed at about a 1:1 ratio at a final DNA concentration of ~0.1 mg/ml in a 100-µl volume containing 20 mM Tris (pH 7.8), 2 M NaCl, 10 mM DTT, 1 mM EDTA, 0.5 mM benzamidine-HCl, and 0.1 mg of BSA/ml. This mixture was transferred to dialysis tubing (SpectraPor) with a 6- to 8-kDa molecular mass cutoff and dialyzed against 200 ml of a solution containing 20 mM Tris (pH 7.7), 2 M NaCl, 1 mM DTT, 1 mM EDTA, and 0.5 mM benzamidine-HCl. The dialysis buffer was exchanged for a similar buffer containing 0.25 M NaCl at a rate of 12 to 13 ml/h over 48 h, and then the assembly was further dialyzed for 12 h against Tris-EDTA buffer. Assembly was verified by electrophoresis on a Tris-acetate-EDTA-1% polyacrylamide gel. To assay remodeling, 3.35 ng of this nucleosomal array was incubated in a total volume of 25 µl containing 15 µl of gradient buffer or BRG1-containing glycerol gradient fractions, 20 U of XbaI (under standard conditions but with 3.5 mM MgCl₂, 0.5 mM ATP, and 60 mM urea) at 30°C. Samples were removed at three time points, deproteinized, and resolved on a Tris-acetate-EDTA-1% agarose gel to determine the extent of XbaI digestion.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Creation of a stably remodeled mononucleosome dimer by the ATPase subunits. The hSWI-SNF complex, immunopurified from a HeLa cell line expressing a FLAG-tagged INI1 subunit, can generate a stable remodeled dimer from mononucleosome cores (30). To determine whether the individual ATPase subunits are also capable of creating this remodeled dimer, we examined the activity of FLAG-tagged versions of BRG1 and hBRM that had been overexpressed and immunopurified from baculovirus-infected Sf9 cells. Such preparations of BRG1 or hBRM have been shown to remodel both mononucleosomes and nucleosomal arrays (28).

View larger version (71K): [in this window] [in a new window]   FIG. 1.   BRG1 and hBRM are capable of interconverting mononucleosome cores between a rotationally phased monomeric state and a stably remodeled dimeric state. (A) Creation of remodeled dimer product by hSWI-SNF and isolated ATPases. SWI-SNF (hS/S; 1 nM), 4 nM BRG1, or 12 nM hBRM was incubated with 390 pM labeled TPT cores and 2 nM unlabeled HeLa cell mononucleosomes in 200-µl remodeling reactions under standard conditions (see Materials and Methods), but with 30 mM KCl, 3.5 mM MgCl2, and 0.5 mM ATP, at 30°C for 40 min. The reactions were stopped by addition of KCl to 240 mM, and the products were separated on a 10 to 30% glycerol gradient by centrifugation for 19.5 h. Samples (2 µl) of the reaction products loaded on the gradients (Load) were analyzed by the gel shift method (lanes 1, 4, and 7). Peak gradient fractions (5-µl volumes) containing mononucleosome cores (mono) (peak 1; lanes 2, 5, and 8) and the novel dimer species (peak 2; lanes 3, 6, and 9) were also analyzed by the gel shift method. Percentages indicate the distances of these fractions from the top of the gradient, which is indicative of relative S values. (B) Comparison of the DNase I digestion patterns of mononucleosome dimers formed by BRG1, hBRM, and the hSWI-SNF complex. Aliquots of peak gradient fractions from the gel shown in panel A (0.075 ng of labeled DNA) were incubated with 0.25 ng of unlabeled HeLa cell mononucleosomes/µl under standard conditions, but with 36 mM KCl, 3.5 mM MgCl2, and 0.5 mM ATP, at 30°C for 30 min; this was followed by equilibration to room temperature for 5 min and digestion with 80 U of DNase I/ml for 2 min. DNase digestion was stopped with 200 µl of phenol, and DNA was isolated and separated by denaturing PAGE (lanes 1 to 6). Positions where cutting is increased (closed arrows) or decreased (open arrows) in the dimer species (lanes 2, 4, and 6) are indicated. For comparison, the DNase I patterns of unseparated BRG1 and SWI-SNF remodeling reactions, performed in the presence (+) or absence () of ATP, are shown (lanes 7 to 10). (C) BRG1 reconverts remodeled dimer back to mononucleosome cores. Remodeled dimers (4 pM), generated by BRG1 and isolated by glycerol gradient centrifugation (Fig. 1A, lane 6), were incubated with 2.5 nM HeLa cell mononucleosomes and with 250 pM (lanes 1 and 2) or 500 pM (lanes 3 and 4) hSWI-SNF or 4 nM (lanes 5 and 6) or 8 nM (lanes 7 and 8) and BRG1 for 35 min at 30°C in 50-µl standard reaction volumes containing 44 mM KCl and 3.5 mM MgCl2, with (+) or without () 0.5 mM ATP. The reactions were stopped by increasing the KCl concentration to 240 mM, and the products were resolved by EMSA. The bare DNA in these lanes resulted from dissociation of the nucleosomes in the dimer-containing gradient fraction following storage at 4 to 8°C for several weeks.

Octamer transfer by the hSWI-SNF complex and its ATPase subunits. The yeast RSC and SWI-SNF complexes have been shown to be able to transfer histone octamers to DNA in trans or in cis around a barrier (23, 42). Little is known, however, about the mechanism involved or the requirement for different subunits in this reaction. To address these issues, we examined the ability of hSWI-SNF and its ATPases to carry out octamer transfer. Transfer from an excess of unlabeled HeLa cell-derived mononucleosome cores to a radiolabeled 155-bp nucleosomal template, TPT, was measured by the creation of a radiolabeled product with the same mobility as control mononucleosome cores on a native gel (Fig. 2A). In the absence of ATP or remodeler, a small amount of transfer appeared to occur (Fig. 2A, lanes 2 to 4, 6, and 8). Addition of ATP to reaction mixtures containing hSWI-SNF or either of the ATPases (BRG1 or hBRM) resulted in a significant increase in the amount of transfer (Fig. 2A, lanes 5, 7, and 9, and data not shown).

View larger version (88K): [in this window] [in a new window]   FIG. 2.   hSWI-SNF and ATPases BRG1 and hBRM can transfer histone octamers from nucleosomes to bare DNA. (A) hSWI-SNF, BRG1, and hBRM transfer histone octamers. Labeled bare TPT fragment (120 pM) was incubated for 35 min at 30°C with 24 nM bulk HeLa cell mononucleosomes (mono) in 12.5-µl standard reactions but with 30 mM KCl and 3.5 mM MgCl2 and with (+) or without () 0.5 mM ATP. The remodeler concentrations were 5.6 nM hSWI-SNF complex (lanes 4 and 5), 8 nM BRG1 (lanes 6 and 7), and 8 nM hBRM (lanes 8 and 9). Reactions were stopped by addition of 2 µg of plasmid DNA, and the products were separated by EMSA. The product marked with an asterisk is created by SWI-SNF and bare DNA in the absence of nucleosomes. Preliminary results suggest that its formation, which is stimulated by ATP (see, e.g., Fig. 5A), is due to protein binding (J. Guyon and R. E. Kingston, unpublished observations). However, since its formation is actually repressed by the presence of donor nucleosomes, this product was not further analyzed here. (B) The transfer product has the same gradient mobility as a nucleosome core. A scaled-up transfer reaction mixture contained 590 pM bare TPT, 280 ng of HeLa cell polynucleosomes (~11 nM in nucleosomes), and 4 nM SWI-SNF in a 200-µl volume, and the reaction was performed under standard reaction conditions but with 30 mM KCl, 3.5 mM MgCl2, 1 mM ATP, and no glycerol. After 1 h, the reaction was stopped by addition of 22 µg of a 8-kb plasmid (pKS-BRG1), and the products were purified on a 10 to 30% GGB-glycerol gradient. Fractions were then separated by EMSA. (C) DNase I accessibility of DNA before and after octamer transfer. Gradient-isolated cores (fraction no. 12, lane 4) or bare DNA (fraction no. 18, lane 2) and control cores (lane 3) or DNA (lane 1) (0.3 ng each) were brought up to a 100-µl volume by addition of 10% GGB, adjusted to 5 mM MgCl2 and 40-µg/ml BSA, and digested for 2 min at room temp with DNase (0.03 U for bare DNA; 0.3 U for nucleosomal DNA). The DNase reaction was stopped by addition of EDTA to 15 mM prior to DNA purification and denaturing PAGE. Arrows indicate the 10-bp periodicity of DNase cuts indicative of a rotationally phased nucleosome core.

Isolation of ATPase subunits under denaturing conditions. The results reported above, combined with previous results from studies of the enzymatic activities of individual hSWI-SNF subunits (28), indicate that the isolated BRG1 and hBRM ATPases are able to perform a variety of remodeling reactions. Despite the fact that these proteins were highly overexpressed in the Sf9 cells prior to purification, there remained the formal possibility that copurifying contaminants from insect cells were partially responsible for the observed activities. From previous work this appeared unlikely; BRG1 and hBRM fractions did not contain significant amounts of additional peptides, and preparations of non-ATPase hSWI-SNF subunits purified from Sf9 cells did not exhibit remodeling activity and thus did not copurify with Sf9 ATPases.

View larger version (74K): [in this window] [in a new window]   FIG. 3.   Remodeling activities coelute with monomeric BRG1. (A) Coelution of octamer transfer and remodeled-dimer activities with monomeric BRG1. BRG1 incubated in a solution containing 1 M urea, 1 M NaCl, and 95 mM KCl was centrifuged through a 10 to 40% glycerol gradient (see Materials and Methods). Fractions were analyzed by silver staining following sodium dodecyl sulfate-PAGE (top panel; 16.7 µl of each peak fraction). The relative position of myosin (200 kDa) centrifuged under identical conditions is indicated (below), as are the positions of molecular mass markers in the gel (left; in kilodaltons). The ~150-kDa band comigrates with a BRG1 breakdown product identified by Western blotting (data not shown). The BRG1-containing fractions were analyzed for their ability to convert mononucleosome cores to a dimeric form (middle panel; EMSA) and to transfer octamers from unlabeled genomic HeLa mononucleosomes to labeled TPT DNA (bottom panel; EMSA). These reactions were performed under standard conditions, except that the final reaction conditions were 33 mM KCl, 0.3% NP-40, and 11 mM (dimer formation) or 33 mM (octamer transfer) urea. Dimer formation reactions contained 2% gradient fraction and 240 pM TPT mononucleosomes, while octamer transfer reactions contained 20% gradient fraction, 120 pM bare TPT, and 24 nM HeLa cell mononucleosomes (mono). The concentration of BRG1 in these reactions was estimated, from the silver stained gel, to range from 0.5 to 3 nM (dimer formation) or from 5 to 25 nM (octamer transfer). (B) Remodeling activity correlates with BRG1 in a quantitative assay. BRG1 incubated in a solution containing 1 M urea, 930 mM NaCl, and 65 mM KCl was centrifuged through a 10 to 40% glycerol gradient (see Materials and Methods) and analyzed by silver staining following sodium dodecyl sulfate-PAGE (top panel; 20 µl of each fraction). The remodeling activity of these fractions was assayed by measuring changes in the rate of restriction enzyme cutting of a nucleosomal array (shown; see Materials and Methods for details). The rate of cutting in the presence of 15 µl of each gradient fraction (or a 1:1 mixture of fractions 19 and 20) in a total of 25 µl is shown in the bottom panel. The concentration of BRG1 in these reactions was estimated to range from 1 to 5 nM.

Role of the remodeled dimer and ATP hydrolysis in octamer transfer. It is conceivable that octamer transfer does not proceed directly from nucleosome cores but rather proceeds from the remodeled dimer generated by SWI-SNF. Either the dimer would be the substrate on which SWI-SNF acts to facilitate transfer, or it might spontaneously transfer octamers once formed. To test these possibilities, we compared the abilities of mononucleosome cores and the remodeled dimer to act as donors in the transfer reaction (in the presence or absence of SWI-SNF). hSWI-SNF-remodeled, unlabeled HeLa cell mononucleosome dimers were isolated by glycerol gradient centrifugation and tested for their ability to donate octamers. We did not detect significant octamer transfer from either mononucleosomes or remodeled dimers in the absence of hSWI-SNF (Fig. 4A, lanes 2 and 3). This indicated that transfer from the remodeled dimer did not proceed spontaneously. The inclusion of hSWI-SNF and ATP in the reaction mixture allowed transfer from the remodeled dimer to occur (Fig. 4A, lanes 4 to 6). The rate of this transfer was comparable to the rate of transfer from mononucleosomes (Fig. 4A, lanes 7 to 9). Under these conditions, the rate of dimer formation was not significantly higher than the rate of octamer transfer (data not shown); thus, mononucleosomes need not first be converted to dimers before acting as donors.

View larger version (58K): [in this window] [in a new window]   FIG. 4.   Octamer transfer requires an unstable intermediate of the hSWI-SNF remodeling reaction. (A) The remodeled dimer does not support transfer without SWI-SNF. Nonlabeled remodeled dimer was generated by incubating 100 nM HeLa cell mononucleosome cores with 3.9 nM SWI-SNF in a 200-µl volume under standard reaction conditions, but in the absence of glycerol and in the presence of 30 mM KCl and 3.5 mM MgCl2, for 2 h. The reaction was stopped by addition of KCl to 230 mM, and the products were separated on 180 mM-KCl-GGB-glycerol gradients as described in Materials and Methods. Octamer transfer reactions were performed with 80 pM labeled TPT bare DNA and 0.8 nM control mononucleosomes or gradient-purified remodeled dimers under standard conditions but with 0.1 mM ATP-MgCl2, 3 mM MgCl2, and 50 mM KCl. SWI-SNF was added at the onset of the reactions or either 10 or 2 min before the reactions were stopped (at 50 min) with 3 µg plasmid DNA. *, donor-independent nonnucleosomal band described in the legend to Fig. 2. (B) Transfer requires continuous ATP hydrolysis. Octamer transfer reactions with 4 nM HeLa cell cores and 120 pM labeled TPT bare DNA and 4.1 nM SWI-SNF were performed under standard conditions but with 2.5 mM MgCl2 and 0.1 mM ATP. Apyrase was added where indicated, and in the last two lanes the labeled TPT bare DNA was added after 30 min. All reactions were stopped with 3 µg of plasmid DNA at 55 min, and products were separated by EMSA. Percentages of cores formed, relative to total counts in the lane, are indicated.

Donor nucleosome and acceptor DNA requirements for octamer transfer. To characterize the generality and significance of octamer transfer, we further examined the donor and acceptor requirements for the reaction. We first examined the effect of using different DNA lengths and sequences as octamer donors and acceptors. hSWI-SNF was able to transfer octamers from HeLa cell mononucleosomes, dinucleosomes, and polynucleosomes to bare acceptor DNA, suggesting that the length of the donor is not important for transfer (Fig. 5A, lanes 6 to 9, and data not shown). In all cases, the percentage of labeled acceptor DNA converted to nucleosomes increased with increasing amounts of donor.

View larger version (51K): [in this window] [in a new window]   FIG. 5.   Effects of donor nucleosome and acceptor DNA type and concentration on transfer. (A) Donor and acceptor requirements for transfer. Transfer reactions (standard conditions but with 21 mM KCl) were performed with 0.12 nM labeled TPT DNA (left panel) or labeled bulk ~150-bp DNA purified from HeLa cell mononucleosome cores (right panel), 3.8 nM hSWI-SNF, 2 mM MgCl2 with (+) or without () ATP, and 0.6 or 3 nM bulk HeLa cell mononucleosomes (Mono), Dinucleosomes (Di.), or TPT mononucleosomes (TPT) where indicated. At 30 min, 3 µg of plasmid DNA was added, and the reaction products were separated by EMSA. *, see the legend to Fig. 2. (B) Inhibition of transfer by high concentrations of bare DNA. Transfer reactions (standard conditions but with 3 mM MgCl2 and 34 mM KCl) were performed with 0.24 nM TPT cores (labeled [left panel] or unlabeled [right panel]), 3.8 nM SWI-SNF (where indicated), and the indicated amount of bare TPT DNA (80 pM of which was labeled in the panel on the right). Reactions were stopped with 3 µg of plasmid at 40 min, and products were separated by EMSA. (C) Quantitation of the results in panel B. For transfer from labeled cores (light bars), lane 4 (no added bare DNA) was set to background. For transfer from unlabeled cores (dark bars), lane 11 (no SWI-SNF) was set to background. The line represents the amount of remodeled dimer formed as a percentage of the amount formed in the absence of bare DNA (lane 4).

Inhibition of remodeling and transfer by bare DNA. To further elucidate the relationship between formation of the stably remodeled dimer and transfer of histone octamers, we examined the effect of bare DNA concentration on both reactions. We found that while bare DNA is a required substrate in octamer transfer, high concentrations of bare DNA inhibited both reactions.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

We have shown that the highly purified hSWI-SNF ATPase subunits, BRG1 and hBRM, can interconvert a stable remodeled mononucleosome dimer with a standard nucleosome core in a reaction that is indistinguishable from that involving the entire hSWI-SNF complex. In addition, hSWI-SNF and the isolated ATPases are each capable of transferring histone octamers from donor nucleosomes to acceptor DNA in trans. The dependence of both reactions on continuous ATP hydrolysis suggests that they may proceed through an unstable intermediate formed by hSWI-SNF. To date, both the formation of a stably remodeled dimer and octamer transfer are activities that have been demonstrated only for members of SWI-SNF-related complexes. The observation that the isolated ATPases of hSWI-SNF can catalyze both of these reactions suggests that it is the ATPase subunit that defines these particular capabilities of the SWI-SNF family of complexes.

A model for nucleosome remodeling and octamer transfer. We have proposed a model for the ability of SWI-SNF to interconvert mononucleosome cores and the remodeled dimer in which SWI-SNF has two nucleosome binding sites with which to bind two mononucleosomes or one dimer (17, 30, 31) (Fig. 6A, top). The presence of two nucleosome binding sites per SWI-SNF complex is suggested by electron microscopic studies of ySWI-SNF (3) and the apparent molecular weight of hSWI-SNF (17). It is important to note, however, that the stoichiometries of the ATPase and other subunits in SWI-SNF complexes have not yet been determined. After substrate binding, the hydrolysis of ATP is hypothesized to result in the generation of an activated intermediate in which histone-DNA and perhaps interhistone interactions are dramatically loosened. Following the hydrolysis cycle, these interactions would reform stochastically to local energy minima, generating either individual mononucleosomes (Fig. 6A, reaction I) or the remodeled dimer (Fig. 6A, reaction II). In this way, hSWI-SNF might be acting analogously to ATP-dependent chaperones that unfold misfolded proteins and allow them to refold spontaneously, effectively reducing the activation barrier between folded states (for a review, see reference 12).

View larger version (46K): [in this window] [in a new window]   FIG. 6.   Model of hSWI-SNF mechanism. (A) When the SWI-SNF ATPase(s) (BRG1 or hBRM alone, or in the SWI-SNF complex) binds two mononucleosome cores (left side) or one remodeled dimer (right side), the hydrolysis of ATP is hypothesized to generate a high-energy SWI-SNF nucleosome intermediate. The loosened histone-DNA contacts in this intermediate facilitate remodeling, allowing these transiently disrupted nucleosomes to stochastically reform normal mononucleosomes (reconversion; reaction I) or remodeled dimers (remodeling; reaction II). In addition, the weakened contacts may allow the histones in this disrupted complex to be pulled away by exogenous bare DNA, resulting in octamer transfer (reaction III). (B) Alternatively, transfer may occur in a concerted reaction when one SWI-SNF site is bound by a nucleosome and the other is bound by bare DNA.

Donor and acceptor requirements for octamer transfer. Random-sequence poly-, di-, and mononucleosomes isolated from HeLa cell nuclei by MNase digestion, as well as in vitro-assembled rotationally phased TPT mononucleosomes, all function as octamer donors. Both TPT DNA and nucleosome-length HeLa cell genomic DNA function as acceptors. A moderate influence of the phasing sequence can be seen as an approximately fivefold decrease in transfer from phased- to random-sequence DNA. As is seen with methods of nucleosome assembly involving salt dialysis, the rotational orientation of nucleosomes formed by octamer transfer appears to be determined by the phasing sequences (see, e.g., Fig. 2C). This is presumably because having the phasing sequences out of alignment produces a nucleosome with higher free energy. The low efficiency of transfer in this and previous work means that it is possible that transfer requires a contaminant or a certain form of the nucleosome that is present at low levels in our donor samples. However, the protein content in these HeLa and TPT nucleosome preparations is greater than 95% core histones, and other contaminants are likely to differ among these samples. Also, the most likely candidate for contamination across these samples, free histones, cannot be used by SWI-SNF to assemble nucleosomes from bare DNA (G. R. Schnitzler and R. E. Kingston, unpublished observation).

Functions of isolated BRG1 and hBRM ATPases. In this article we have extended our earlier results to show that the central ATPases of human SWI-SNF complexes, BRG1 and hBRM, can individually perform all of the tested catalytic functions of the entire complex. Both can generate a remodeled dimer product indistinguishable from that of complete hSWI-SNF. They can also reconvert this product back to normal cores, establishing an equilibrium between the two nucleosomal states. As is true for the complete SWI-SNF complex, the altered DNase digestion pattern due to hBRM and BRG1 action can be largely explained as being a mixture of the DNase patterns for normal cores and remodeled dimers (Fig. 1B) (30). Furthermore, BRG1 and hBRM can catalyze the octamer transfer reaction. We believe that these activities are intrinsic to the ATPases, since theyand nucleosomal array remodeling activitycoelute with BRG1 under conditions in which BRG1 runs as a monomer.

Role of octamer transfer in SWI-SNF function. hSWI-SNF can stimulate RNA polymerase II elongation in vitro on nucleosomal templates activated by HSF or VP16 activators (5). The ability of SWI-SNF to transfer an octamer to bare DNA may be important for this effect and thus potentially for transcriptional elongation in vivo. As RNA polymerase II moves, downstream nucleosomes act as barriers and cause dramatic pauses (5, 14). The transfer of a nucleosome to upstream sequences as a polymerase passes has been observed in vitro for a bacteriophage polymerase and RNA polymerase III (35, 36). The ability of SWI-SNF to carry out octamer transfer may facilitate this process, accommodating both polymerase movement and maintenance of the nucleosomal content of the gene. An ability of SWI-SNF to reposition or alter nucleosomes to accommodate a passing polymerase is one possible explanation for yeast SWI-SNF being continuously required for transcription in vivo (4, 37). Remodeling complexes probably also facilitate transcription in vivo by helping transcription factors bind their cognate sites in promoter chromatin. An ability to transfer an octamer to surrounding DNA, either in cis or in trans, would open up DNA binding sites and allow factor access. Together with the creation of remodeled nucleosome dimers, which allows increased access of Gal4 to a site originally near the dyad of the unremodeled nucleosome (30), octamer transfer and/or mobilization may allow remodeling complexes to create highly dynamic and accessible chromatin.