Nucleosome sliding induced by the xMi-2 complex does not occur exclusively via a simple twist-diffusion mechanism.

ATP-dependent chromatin remodeling complexes can induce the translocation (sliding) of nucleosomes in cis along DNA, but the mechanism by which sliding occurs is not well defined. We previously presented evidence that sliding induced by the human SWI/SNF complex does not occur solely via a proposed "twist-diffusion" mechanism whereby the DNA rotates about its helical axis without displacement from the surface of the nucleosome (Aoyagi, S., and Hayes, J. J. (2002) Mol. Cell. Biol. 22, 7484-7490). Here we examined whether the Xenopus Mi-2 nucleosome remodeling complex induces nucleosome sliding via a twist-diffusion mechanism with nucleosomes assembled onto DNA templates containing branched DNA structures expected to sterically hinder rotation of the DNA helix on the nucleosome surface. We find that the branched DNA-containing nucleosomes undergo xMi-2-catalyzed sliding at a rate and extent identical to that of nucleosomes assembled on native DNA fragments. These results indicate that both the hSWI/SNF and xMi-2 complexes induce nucleosome sliding via a mechanism(s) other than simple twist diffusion and are consistent with models in which the DNA largely maintains its rotational orientation with respect to the histone surface.

In the nucleus, the eukaryotic genome is packaged in the form of a highly condensed chromatin fiber through its interactions with histones and non-histone proteins (1,2). To allow efficient progression of nuclear processes, cells have developed several mechanisms to facilitate access of DNA target sites within chromatin by trans-acting factors (3)(4)(5). One strategy involves targeted post-translational modifications of histone proteins such as acetylation, methylation, and phosphorylation that may directly alter the biochemical properties of chromatin or signal binding and recruitment of ancillary factors (6 -10). A second strategy involves the activity of ATP-dependent chromatin remodeling complexes that couple the energy derived from ATP hydrolysis to alter chromatin structure, facilitating the activity of trans-acting factors (4,(11)(12)(13).
All ATP-dependent chromatin remodeling complexes contain a subunit that belongs to the SNF2 superfamily of ATPases (14). There are four main classes within this superfamily, dif-ferentiated by homology to the SWI2/SNF2, ISWI, INO80, and Mi-2 ATPase subunits within these complexes (11)(12)(13)15). The molecular mechanism of nucleosome remodeling by the ATPdependent chromatin remodeling complexes has been under intense study. Remodeling complexes such as the SWI/SNF complex have been shown to cause disruption of histone-DNA interactions as detected by electron energy loss microscopy and atomic force microscopy studies (16,17). The loss of histone-DNA interactions within the nucleosome is correlated with the increase in activities of various transcription factors, restriction enzymes, and DNase I on nucleosomal DNA (18 -21). Evidence indicates that Mi-2 complexes also cause such disruptions in histone-DNA interactions as detected by increase in accessibility to DNase I and restriction enzyme cleavage (22)(23)(24). 1 A number of chromatin remodeling complexes including Drosophila and Xenopus Mi-2 complexes have been shown to induce nucleosome translocation (sliding) along the DNA in cis for both mononucleosomes and nucleosome arrays (17,(25)(26)(27)(28)(29)(30)(31). Nucleosome sliding is thought be an important outcome of chromatin remodeling, which allows stable exposure of DNA target sites to trans-acting factors involved in chromatin directed activities in vivo (4,11,13). There are many possible ways in which chromatin remodeling could result in nucleosome sliding, and several models have been put forward (11,30,32). In one model nucleosome sliding is envisioned to occur by twisting of the DNA helix like a screw in a groove on the surface of the histone octamer, without significant displacement of the helix from the nucleosome surface. This model is supported by recent crystal structures of nucleosome core particles in which a turn of DNA within the core contains one less base pair compared with the symmetry-related position, causing an overwinding of the DNA at the former (33,34). This "twist defect" could stochastically diffuse throughout the nucleosome with the cumulative effect of many such events occurring in one direction resulting in the twisting of the DNA helix along the histone surface (twist diffusion). Nucleosome sliding also may occur via a DNA uncoiling-recapture mechanism, initiated by the unraveling of DNA from one end of the nucleosome surface that is then recaptured to form a loop. The loop would then be propagated through the rest of the nucleosome, leading to change in the position of the histone octamer along the DNA (30,32,35). Evidence for such a mechanism during nucleosome sliding induced by the yeast SWI/SNF complex recently has been reported (35). Finally, apparent nucleosome movement may occur upon relaxation of a structurally altered, remodeled nucleosome to a canonical structure (11).
In a previous study, we tested whether hSWI/SNF-depend-ent nucleosome sliding involves a twist-diffusion mechanism by using nucleosomes assembled on branched and nicked DNAs that would sterically hinder rotation of the DNA on the histone surface and inhibit retention of torsional stress within the DNA. Remodeling of these nucleosomes showed that SWI/SNFinduced nucleosome sliding does not occur principally via a simple twist-diffusion pathway or rotation of the DNA on the histone surface for this complex (31). In addition, we found that the presence of a nick, which could relieve torsional stress generated by the hSWI/SNF complex, did not affect the nucleosome sliding process (31) in agreement with results obtained with Drosophila ISWI (36).
In this study, we tested whether Xenopus Mi-2-induced nucleosome sliding occurs via a simple twist-diffusion mechanism using nucleosomes containing branched DNA structures expected to sterically hinder rotation of the DNA helix on the histone surface. Nucleosomes with very homogenous translational positions were used, allowing comparison of the rate of sliding between nucleosomes assembled on native and branched DNA templates. Our results indicate xMi-2 induces nucleosome sliding on both types of templates at identical rates and to identical extents, suggesting that, like the hSWI/SNF complex, xMi-2 induces nucleosome sliding via a mechanism other than simple twist diffusion or rotation of DNA on the histone surface.

MATERIALS AND METHODS
DNA Fragments-All three DNA templates are based on the 215-bp EcoRI-DdeI fragment containing a Xenopus borealis somatic 5S RNA gene derived from the plasmid pXP-10 (37). The native 215-bp DNA fragment encompassing positions from Ϫ78 to ϩ137 in the 5S sequence was radiolabeled at 5Ј end of the EcoRI site and purified on 6% native polyacrylamide gels as described (37). The related hairpin-containing template was generated as described previously (31) (see Fig. 1). The continuous hairpin substrate was prepared in two steps. First, a continuous bottom strand extending from positions Ϫ78 to ϩ137 in the 5S sequence was generated as described previously (31). The top strand of the continuous hairpin substrate was generated by ligating 450 pmol each of the primer (5S primer Ϫ78 to Ϫ12) AAT TCG AGC TCG CCC CGG GAT CCG GCT GGG CCC CCC CCA GA, (5S primer-12 to ϩ18) AGG CAG CAC AAG GGG AGG AAA AGT CGA GAC CAG GAG GGG TTT TTC CCC TCC TGG TCA GAG CCT TGT GCT CGC CTA CGG CCA TAC CAC C, (5S primer ϩ18 to ϩ75) CTG AAA GTG CCC GAT ATC GTC TGA TCT CGG AAG CCA AGC AGG GTC GGG CCT GGT TAG, and (5S primer ϩ75 to ϩ137) TAC TTG GAT GGG AGA CCG CCT GGG AAT ACC AGG TGT CGT AGG CTT TTG CAC TTT TGC CAT TC. The oligonucleotides were annealed with 900 pmol each of three bridging oligonucleotides, GGC GGT CTC CCA TCC AAG TAC TAA CCA GGC CCG ACC CTG C, ACG ATA TCG GGC AGT TTC AGG GTG GTA TGG CCG TAG GCG A, and TTC CTC CCC TTG TGC TGC CTT CTG GGG GGG GCC CAG CCG G, and then ligated with T4 DNA ligase (Invitrogen) as described previously (31). The 248-mer ligated product was isolated on preparative denaturing polyacrylamide gels (6%) after visualization of the bands by ethidium bromide staining. Finally, double-stranded continuous hairpin substrate was generated by annealing the 5Ј-radiolabeled ligated 247-mer strand with the 215-mer bottom strand prepared as described above. These oligonucleotides were mixed in annealing buffer (10 mM Tris-EDTA, pH 8.0, 50 mM NaCl), heated to 95°C for 10 min, and cooled slowly by turning off the heating block. The annealing reaction created a 215-bp double-stranded 5S DNA with a hairpin (12-bp stem and a 5-nucleotide loop) with 2 nucleotide hinges at the beginning and at the end of the stem at the Ϫ12 position (Fig. 1).
Nucleosome Reconstitution-Recombinant Xenopus H2A and the cysteine-substituted mutant H2BG26C were prepared as preformed dimers, and the latter was modified with 4-azidophenacyl bromide (APB, 2 Sigma) as described (37,38). Histone H3/H4 tetramers were prepared from chicken erythrocyte nuclei, and nucleosomes were reconstituted with either native H2A/H2B or H2A/H2BG26C-APB. Reconstitution with the 215-bp 5S DNA fragment yields a majority of nucleosomes in which the dyad axis of symmetry is located near position Ϫ3 with respect to ϩ1, the transcription start site of the 5S gene (39). To select for the 5Ј-most positions, reconstitutions were incubated with 125 units of BamHI restriction enzyme (New England Biolabs) in the presence of 10 mM HEPES, pH 7.5, 0.2 mM EDTA, and 5 mM MgCl 2 in a 200-l reaction. The BamHI reaction was stopped by adding EDTA stop solution to 10 mM final concentration and addition of excess calf thymus DNA (0.75 g) to dissociate the restriction enzyme from the nucleosomes. The reaction was then loaded onto 5-30% glycerol (in 10 mM Tris-Cl, pH 8.0) 10-ml gradients, and nucleosomes were sedimented at 34,000 rpm by an SW41Ti rotor for 18 h at 4°C. Fractions containing nucleosomes were identified by running a small portion of each on a 0.7% agarose nucleoprotein gel (39). In addition, portions of the fractions from the glycerol gradient were loaded on a 5% polyacrylamide "translational gels" (40) to monitor the purity of the BamHI-selected nucleosomes. Nucleosomes used for the cross-linking experiment were dialyzed for 3 h against a buffer containing 10 mM Tris-Cl, pH 8.0, prior to UV irradiation.
Mapping of Nucleosome Positions by Restriction Enzyme-Translational Gel Assay-Nucleosomes reconstituted on native DNA were digested with 20 units of BamHI, RsaI, or HhaI restriction enzyme in a 20-l reaction for 15 min in 10 mM HEPES, pH 7.5, 0.2 mM EDTA, 5 mM MgCl 2 . The reactions were then stopped with 10 mM EDTA, and excess calf thymus DNA (0.75 g) was added to compete the restriction enzymes away from the nucleosomes. The reactions were loaded on a polyacrylamide translational gel, and electrophoresis carried out as described above.
xMi-2 and hSWI/SNF Reactions and Translational Gel Assay-The Xenopus Mi-2 complex was prepared as described previously (41,42). In a 100-l reaction mixture, ϳ5 ng of nucleosomes was incubated with 800 ng of xMi-2 in 10 mM HEPES, pH 7.5, 0.2 mM EDTA, 5 mM MgCl 2 , 1 mM ATP, and 100 ng of bovine serum albumin/l (25). Reaction mixtures were incubated at 25°C for 30 min or for the times indicated in the figures. The hSWI/SNF complex was a generous gift from Dr. Robert Kingston. In a 100-l reaction, ϳ5 ng of nucleosomes were incubated with 245 ng of hSWI/SNF in 12 mM HEPES, pH 7.9, 60 mM KCl, 7 mM MgCl 2 , 0.6 mM dithiothreitol, 60 M EDTA, 100 ng of bovine serum albumin/l, and 4 mM ATP at 30°C for 15 min (31). The reactions were then loaded on a translational gel (5% polyacrylamide) as described previously (40).
FokI Nucleosome Sliding Assays-BamHI-selected and gradient-purified nucleosomes were incubated with Mi-2 as described above for the times indicated in the figure legends, and then aliquots (10 l) were removed and the Mi-2 remodeling reaction stopped by the addition of 0.75 g of calf thymus DNA. Aliquots were then digested with 4 units of FokI (New England Biolabs) for 15 min, and the digestions were stopped with 0.1% SDS, 10 mM EDTA stop solution, and samples were loaded onto a 6% native polyacrylamide gel. Note that for the FokI experiments, both the hairpin and continuous hairpin nucleosomes were mixed with an equal amount of native nucleosomes before the incubation with Mi-2 as an internal reference. To assess the effect of histone-DNA cross-linking on nucleosome sliding, nucleosomes were reconstituted with the native 5S DNA fragment and H2BG26C-APB as described above. The nucleosomes were irradiated with 365 nm UV light (VWR LM20E transilluminator) for 25 s prior to incubation with xMi-2, and samples were treated as described above except that crosslinked and uncross-linked DNAs were separated by SDS-PAGE (6% polyacrylamide), and gels were dried and analyzed by Phosphor-Imager (37).
Hydroxyl Radical Footprinting of Nucleosomes-Hydroxyl radical footprinting of reconstituted nucleosomes was performed by pipetting 20 l each of 1 mM Fe-EDTA and 20 mM sodium ascorbate to the side of a tube containing 140 l of nucleosomes in the presence of 10 mM HEPES, pH 7.5, 0.2 mM EDTA, 5 mM MgCl 2 . The reactions were initiated by pipetting 20 l of a 0.12% solution of H 2 O 2 into the Fe-EDTA/ascorbate drop and then quickly mixing the reagents with the nucleosomes. After 2 min, the reactions were stopped by the addition of 13.5 l of 50% glycerol. Then 6.5 l of 20 units/l BamHI was added to each tube, and the samples were incubated at 37°C for 15 min for the BamHI selection process. The digestion was stopped by addition of 20 l of 100 mM EDTA and then loaded onto a preparative 0.7% agarose gel (1/2ϫ Tris borate-EDTA) to isolate nucleosomes from naked DNA. DNA isolated from the preparative agarose gel was ethanol-precipitated, resuspended in formamide loading dye, and then loaded onto a 6% denaturing polyacrylamide gel, and footprints were analyzed as described (43).
Exonuclease III Assays-BamHI-selected nucleosomes were incubated with xMi-2 as described above, and then 0.75 g of sheared calf thymus DNA was added to compete Mi-2 complex away from the nu-cleosomes. Exonuclease III (Exo III, 0.5 units; New England Biolabs) was added, and the reactions were incubated for 15 min and then stopped by the addition of 0.1% SDS, 10 mM EDTA stop solution. DNA in the samples was then ethanol-precipitated, and cleavage products were analyzed as described (31). Note that for the Exo III digestions the nicked hairpin substrate was radiolabeled at the base of the hairpin (at Ϫ12) ( Fig. 1 and Fig. 6B), thus precluding BamHI selection of nucleosomes on this template.

RESULTS
Mi-2 complexes from Drosophila and Xenopus have been shown to catalyze nucleosome sliding in vitro (25,26). Previous work by Guschin et al. (25) demonstrated that the xMi-2 complex catalyzes the sliding of nucleosomes toward the center of a 250-bp DNA fragment from the Xenopus thyroid hormone receptor ␤A gene as detected by translational gel assays. To determine whether xMi-2-induced nucleosome sliding could be observed on another DNA sequence, nucleosomes were reconstituted onto a 215-bp 5S DNA fragment, and the distribution of translational positions was examined on native polyacrylamide gels (44). Prior to xMi2 remodeling, this method revealed at least 5 translational positions (Fig. 2, lane 1). The approximate location of each of these translational positions was determined by cleavage of the nucleosomes with either RsaI, HhaI, or BamHI, before separation on the translational gel (Fig. 2, lanes 2, 3, and 5, respectively). We find that ϳ70% of the nucleosomes occupy the upstream end of the 215-bp 5S DNA fragment (thick oval, Fig. 2). This results closely agrees with chemical and nuclease mapping of the Xenopus 5S nucleosome (38,40,45).
To assess the ability of xMi-2 to catalyze sliding of nucleosomes, glycerol gradient purified nucleosomes were incubated with Mi-2 in the presence or absence of ATP followed by addition of excess calf thymus DNA to promote dissociation of Mi-2 complexes from the nucleosomes. The reactions were then analyzed on translational gels to determine whether remodeling resulted in alterations in the distribution of nucleosome positions (Fig. 3). Incubation of nucleosomes with Mi-2 in the presence of ATP caused most of the major fast-migrating species (the darkest band in Fig. 3, lanes 2 and 3) to shift to a slower migrating species (compare Fig. 3, lanes 3 and 4). This indicates that xMi-2 remodeling results in the accumulation of nucleosomes positioned near the center of the DNA fragment, in agreement with the previous study (25). Interestingly, in contrast to nucleosomes remodeled by the Mi-2 complex, remodeling with the hSWI/SNF complex alters the original distribution of translational positions to faster migrating species (Fig. 3, lane 5). This is in agreement with previous work (27,31) showing the propensity of yeast and human SWI/SNF to catalyze nucleosome movement toward the edge of DNA fragments.
To determine whether Mi-2 induced nucleosome sliding occurs via a twist-diffusion mechanism, nucleosomes were assembled on templates containing DNA hairpin structures. The hairpins consist of a 12-bp stem and a 5-nucleotide loop placed near the center of the nucleosome positioning sequence within the 5 S DNA fragment. The location of the junction was chosen so that the hairpin will extend out away from the histone surface after reconstitution (Fig. 1). If xMi-2-induced nucleosome sliding occurs by a simple twist-diffusion mechanism, then the presence of a DNA hairpin is expected to result in severe steric clash during rotation of the DNA on the histone surface. Two different types of hairpin structures were used in this study. One contains a nick at the hairpin junction (Fig. 1,  hairpin), whereas the other construct does not contain a nick and forms a continuous top strand with a hairpin (Fig. 1,  continuous hairpin). The continuous and nicked hairpin substrates were designed to ascertain whether the generation of torsional stress during remodeling, as demonstrated previously (46) for a number of ATP-dependent chromatin remodeling complexes, affects nucleosome sliding efficiency.
To generate a population of uniformly translationally positioned nucleosomes prior to remodeling, nucleosomes assembled on native, hairpin, and continuous hairpin DNAs were subjected to BamHI digestion. BamHI digestion selects for nucleosomes positioned at the 5Ј end of the 5S DNA fragment because cleavage of these nucleosomes is inhibited, whereas nucleosomes occupying more downstream positions have BamHI sites exposed and are rapidly cleaved, resulting in the loss of the radiolabel (Fig. 2, lane 5). Purification of the BamHIselected nucleosomes by glycerol gradient from the free DNA and subnucleosomal species yields a population of homogeneously positioned nucleosomes for subsequent analysis (Fig.  4). The vast majority of these nucleosomes are located at the 5Ј end of the 5 S DNA templates (see below).
The BamHI-selected nucleosomes assembled with continuous hairpin and native DNA templates were analyzed by hydroxyl radical footprinting (Fig. 5). Nucleosomes reconstituted on continuous hairpin substrates have rotational positioning that is indistinguishable from those of nucleosomes reconstituted on native DNA (Fig. 5, compare lanes 3 and 4). The characteristic 10-bp hydroxyl radical cleavage patterns for both the hairpin and control nucleosomes align perfectly in the lower part of the gel (from ϳϪ43 to ϳϪ23 on the 5S sequence, Fig. 5). However, with the continuous hairpin nucleosome, the hydroxyl radical cleavage pattern is shifted up on the gel by 33 bp, the precise length of the hairpin sequence (Fig. 5). We note a lack of hydroxyl radical cleavage near position Ϫ12 in the continuous hairpin nucleosome pattern that is present in native nucleosomes (Fig. 5). This inhibition of hydroxyl radical cleavage is most likely due to the proximity of this position to the hairpin/template junction, as observed in previous studies of branched DNA structures (31,47). In addition, the hairpin structure itself is susceptible to hydroxyl radical cleavage, as expected (Fig. 5, lane 4), and a similar pattern can also be seen in the hydroxyl radical cleavage of the naked continuous hairpin DNA (Fig. 5, lane 5). The cleavage pattern of the hairpin structure (horizontal bars on the right-hand side of the gel in Fig. 5) shows robust cleavage for most of the stem region of the hairpin, but cleavage is inefficient within the single-stranded region of the hairpin (Fig. 5, compare S1, S2 to H-1, LP, and  H-2), as expected (47).
We next determined whether the presence of either of the branched DNAs within the nucleosome inhibited xMi-2-catalyzed sliding by Exo III analysis. Exo III digests doublestranded DNA in a 3Ј-5Ј direction until the progression of the enzyme is impeded by the edge of a nucleosome (29,48) and has been used to identify the downstream edge of the 5S nucleosome (31). BamHI-selected nucleosomes reconstituted onto both the native and continuous hairpin DNAs yield strong stops at approximately ϩ75, which places the nucleosomes at the upstream end of the DNA fragment, as expected (Fig. 6, A and C, ϪMi-2). Exo III digestion of the nicked hairpin nucleosomes resulted in a strong pause sites at approximately ϩ75 and ϩ125 (Fig. 6B, ϪMi-2), due to the inability to carry out BamHI selection with this template (see "Materials and Methods"). Exo III also detects a small fraction of nucleosomes with downstream edges at approximately ϩ125 in the continuous hairpin nucleosomes that was not completely removed by BamHI selection process, but this does not affect the analysis (Fig. 6B, ϪMi-2). After remodeling by the xMi-2 complex, the distribution of translational positions detected by Exo III digestion is clearly altered. In all three cases, stops near ϩ75 were severely diminished, and several new Exo III pause sites were detected near the downstream end of the DNA (Fig. 6, A-C, ϩMi-2), indicating movement of nucleosomes from the upstream edge toward the downstream edge of the DNA fragment upon remodeling by xMi-2. Importantly, Mi-2-induced nucleosome movement does not appear to be inhibited by the presence of the hairpin structures.
We also determined whether the hairpin structures inhibited xMi-2-induced nucleosome sliding by a FokI site protection assay. The FokI site is about 20 bp beyond the edge of main translational position selected by BamHI digestion (Fig. 1), and thus about 90% of the DNA in the native nucleosome sample was rapidly digested before remodeling with Mi-2. However, protection of this site was drastically increased upon Mi-2 remodeling such that only about 45% of the DNA was digested by FokI after 30 min of remodeling (data not shown; see below). This result is consistent with sliding of nucleosomes over the FokI cleavage site upon Mi-2 remodeling (27,31). Importantly, a similar increase in protection of the FokI site was observed with the continuous hairpin and nicked hairpin nucleosomes after 30 min of Mi-2 remodeling (data not shown; see below).
In order to verify these results and to eliminate the possibility that hairpin-containing nucleosomes undergo Mi-2-induced sliding via an alternative pathway, we examined the rate of sliding as determined by the FokI assay. BamHI-selected nucleosomes reconstituted on native, hairpin, and continuous hairpin substrates were subjected to Mi-2 remodeling for var- ious times, and then remodeling was quenched by addition of excess calf thymus DNA, which competes the complex away from the nucleosome. Each aliquot was then subjected to FokI digestion (Fig. 7A). When native nucleosomes were incubated with xMi-2 and ATP, protection of the FokI site increased with increasing Mi-2 incubation time (Fig. 7A, compare open and closed diamonds), consistent with xMi-2 inducing sliding of an increasing fraction of nucleosomes with time.
To substantiate that the increase in protection of the FokI site is due to nucleosome sliding, the histone octamer was first fixed in position by site-specific histone-DNA cross-linking (31) and then assayed for changes in accessibility at the FokI site upon Mi-2 remodeling (Fig. 7B). In this case only a much smaller increase in the protection at the FokI site was observed for nucleosomes containing a histone covalently cross-linked to nucleosomal DNA (Fig. 7B, ovals). The ϳ10% increase in protection at the FokI site for cross-linked nucleosomes may reflect some flexibility of the cross-link, allowing for short nucleosome movements along the DNA which could result in a small increase in protection. Also, it is possible that some rearrangement of the histone N-terminal tails may occur after Mi-2 remodeling leading to small increase in protection at the FokI site. Nonetheless, the increase in protection at the FokI site in cross-linked nucleosomes is minimal compared with the increase in uncross-linked nucleosomes (Fig. 7B,  compare ovals and diamonds), supporting the notion that the protection is due to nucleosome sliding.
We then determined whether the DNA hairpin structures affected the rate of xMi-2-catalyzed nucleosome sliding by the FokI assay. Both BamHI-selected nucleosomes reconstituted on nicked hairpin and continuous hairpin DNAs show increases in protection at the FokI site upon Mi-2 remodeling (Fig. 7, C and D, respectively, squares). Importantly, the rate of the increase in protection at this site is identical to that of nucleosomes reconstituted with native DNA. It is important to note that the native nucleosome controls were assayed in the same reaction tubes simultaneously along with the hairpin nucleosomes (Fig. 7, C and D, compare squares and diamonds). Thus xMi-2-catalyzed sliding of nucleosomes containing hairpin structures occurs to the same extent and at the same rate as nucleosomes containing native DNA. DISCUSSION Nucleosome movement along the DNA resulting from chromatin remodeling activities is likely to play an important role in exposing target DNA sites within chromatin for trans-acting factors, thereby facilitating various nuclear processes (11,13). Many of the ATP-dependent chromatin remodeling complexes have been shown to be able to promote nucleosome sliding along a DNA fragment in vitro, but the mechanism(s) by which sliding occurs are not well understood (17,(25)(26)(27)(28)(29).
In a previous study (31), we employed nucleosomes containing branched DNA structures to demonstrate that hSWI/SNFinduced nucleosome sliding does not occur solely via a twistdiffusion mechanism. We wished to assess if these observations could be extended to a different ATP-dependent remodeling complex, the xMi-2 complex, which also has robust nucleosome sliding activity (25). We used a nucleosome with hairpin structures placed near the dyad to determine whether xMi-2-dependent nucleosome sliding occurs by the twist-diffusion mechanism. If nucleosome sliding occurs by the twisting of the DNA helix 1 bp at a time on the surface of the histone octamer, the hairpin structure will sterically hinder any twisting motion of the DNA, greatly inhibiting sliding.
Our results indicate that xMi-2-catalyzed nucleosome sliding occurs to the same extent and at the same rate in the presence of a hairpin structure compared with native nucleosomes using two independent assays. These results indicate that similar to the hSWI/SNF complex, xMi-2-dependent nucleosome sliding does not proceed via a simple twist-diffusion mechanism. Moreover, our results support models in which the DNA maintains its rotational orientation with respect to the histone surface and/or partially dissociates from this surface. Importantly, the rate of nucleosome sliding catalyzed by Mi-2 was found to be identical when comparing the native and hairpin-containing nucleosomes (Fig. 7). This is important because if the primary sliding pathway is blocked by the hairpin structure and the nucleosomes undergo sliding via an alternative pathway, then the rate of sliding is expected to be slower due to the energetic cost of adopting the alternative pathway. However, in the case of xMi-2-induced sliding, this clearly is not the case. This implies that sliding of nucleosomes reconstituted on DNAs with hairpin structures occurred via a pathway identical to that of native nucleosomes. Thus it is highly unlikely that sliding induced by both the Mi-2 complex and the hSWI/SNF complex involves a simple twist-diffusion mechanism.
The actual mechanism of xMi-2-dependent nucleosome sliding may be very complicated, involving aspects of twist diffu- FIG. 7. xMi-2 induces nucleosome sliding at equivalent rates and to equivalent extents on native and branched DNA templates. Gradient-purified BamHI-selected nucleosomes were incubated with or without Mi-2 and ATP for varying amounts of time, and then the reaction was terminated by addition of excess competitor DNA, and the accessibility of the FokI site was determined as described under "Materials and Methods." A, xMi-2 remodeling activity induces protection of the FokI site. The percent native nucleosomal DNA remaining uncut by FokI after incubation in the absence or presence of Mi-2 ϩ ATP (closed and open diamonds, respectively) for the times indicated is plotted. B, histone-DNA cross-linking blocks xMi-2-induced protection of the FokI site. Nucleosomes containing H2BG26C-APB were UVirradiated to induce cross-linking in ϳ20% of the sample and then subjected to xMi-2 remodeling activity and FokI digestion as in A. The extent of FokI cleavage in the cross-linked and uncross-linked fraction (circles and diamonds, respectively) is plotted as a function of xMi2 remodeling time. C and D, xMi-2 remodeling induces protection of the FokI site at equivalent rates and extents for native, hairpin (C), and continuous hairpin (D) nucleosomes. The percent DNA remaining uncut after FokI digestion for native and hairpin nucleosome (diamonds and squares, respectively) versus time of incubation with Mi-2 is plotted. Data was normalized to the initial level of protection at the FokI site. Note that approximately equal amounts of native and hairpin nucleosomes were incubated together in the reactions.
FIG. 6. Exonuclease III analysis of nucleosome positioning before and after remodeling by xMi-2. Glycerol gradient purified BamHI isolated nucleosomes assembled on native (A), hairpin (B), or continuous hairpin (C) templates were incubated in the absence or presence of Mi-2 and ATP and then subjected to Exo III mapping. Each panel shows, from left to right, G-specific cleavage of native 5 S DNA as a marker, undigested nucleosomes, Exo III digestion of naked template DNA, and Exo III digestion of nucleosomes before and after Mi-2 remodeling. The positions of the downstream edge of the nucleosomes before and after Mi-2 remodeling are schematically represented on the left and right sides of the gels, respectively. Dark and light ovals represent major and minor nucleosome positions as determined from the Exo III mapping. sion as well as an uncoiling-and-recapture mechanism, and/or collapse of structurally altered remodeled nucleosomes to canonical structures (5,11). A recent single base pair resolution mapping of ySWI/SNF-remodeled nucleosomes by a site-specific cross-linking technique supports a model in which SWI/ SNF remodeling causes unraveling of DNA from the edge of the nucleosome, which then rebinds to new sites closer to the dyad axis, resulting in formation of a DNA loop. The resulting loop may propagate around the nucleosomes causing the nucleosome to slide up to 50 bp off of the end of the DNA and exposing DNA sites previously located within the interior of the nucleosome (35). Such a mechanism would be consistent with data obtained from this study. Identification of a looped intermediate during the remodeling reaction, perhaps via preferential chemical cleavage of the highly bent DNA within the loop, would directly support such a model. Also, in agreement with previous studies using hSWI/SNF and ISWI complexes (31,36), we find that the presence of a nick did not affect the rate of Mi-2-induced nucleosome sliding, suggesting that this process is not dependent highly on the propagation of torsional stress within the DNA.