Linker Histone H 1 Modulates Nucleosome Remodeling by Human SWI / SNF *

Chromatin, a combination of nucleosomes and linker histones, inhibits transcription by blocking polymerase movement and access of factors to DNA. ATP-dependent remodeling complexes such as SWI/SNF and RSC alter chromatin structure to increase or decrease this repression. To further our understanding of how human SWI/ SNF (hSWI/SNF) “remodels” chromatin we examined the octamer location, nature, and template specificity of hSWI/SNF-remodeled mononucleosomes when free or bound by linker histone H1. We find that, in the absence of H1, hSWI/SNF consistently moves nucleosomes to DNA ends, regardless of template sequence. On some sequences the repositioned histone octamer appears to be moved 45 bp off the DNA edge, whereas on others it appears to be normal, suggesting that the nature of the remodeled nucleosome can be influenced by DNA sequence. By contrast, in the presence of histone H1, hSWI/SNF slides octamers to more central positions and does not promote nucleosome movement off the ends of the DNA. Our results indicate that the nature and position of hSWI/SNF products may be influenced both by DNA sequence and linker histone, and shed light on the roles of H1 and hSWI/SNF in modulating chromatin structure.

The basic unit of chromatin is the nucleosome core particle, composed of two copies each of the four core histones H2A, H2B, H3, and H4, which wrap ϳ146 bp of DNA into a lefthanded superhelix. The presence of nucleosome cores on DNA blocks access to sequence-specific DNA binding factors by steric inhibition, with histone:DNA contacts occluding regulatory sites. It also inhibits movement of RNA polymerase II (1,2). In addition to core histones, the chromatin of metazoan somatic cells also contains linker histones, such as H1, which are present at a ratio of approximately one per nucleosome. H1 binds to the nucleosome core, protecting ϳ10 bp of both the entering and exiting DNA, to form a structure termed the chromatosome. H1 is primarily thought to be involved in transcriptional repression. For instance, in Xenopus development the expres-sion of H1 causes the repression of oocyte-specific 5S rDNA genes (3,4). However, H1 has also been shown to be involved in transcriptional activation at certain promoters (5)(6)(7). H1 binding is known to promote the folding of chromatin into more compact structures, decrease nucleosome mobility, influence nucleosome positions, reduce the binding of some sequencespecific factors, inhibit core histone tail acetylation, and inhibit certain activities of chromatin remodeling complexes (8 -10). It is unclear, however, exactly how each of these effects is related to transcriptional regulation.
One of the mechanisms cells use to modulate the repressive effects of chromatin, is to employ a family of ATP-dependent chromatin remodeling complexes. Each of these complexes has important functions in the activation and/or repression of different subsets of genes. Mutations in genes encoding remodeling complex components often result in alterations of the chromatin structure of target genes in vivo (11)(12)(13). One activity that appears to be shared by all remodeling complexes is the ability to translationally reposition histone octamers on DNA (14). Different complexes, however, can produce differentially positioned products (which is true even for some complexes that contain the same catalytic ATPase subunit) (15). Nucleosome repositioning away from or over regulatory sites in chromatin is likely to be an important aspect of remodeling complex function, and differences in repositioning specificity likely contributes to the unique functions of each complex. Several important questions about octamer repositioning by ATP-dependent remodeling complexes remain unanswered, however, such as the importance of DNA sequence and structure in determining remodeled octamer positions and the effect of linker histones on repositioning.
In addition to the translational repositioning of normal nucleosomes, many remodeling complexes (particularly members of the SWI/SNF subfamily) are capable of creating stable, structurally altered nucleosomal products (11,16). This is evidenced by a reduction of the number of nucleosome-constrained negative supercoils, without apparent nucleosome loss, on circular polynucleosome templates (11,17), and by the reduced stability of remodeled polynucleosomal arrays to surface deposition in atomic force microscopy studies (18). Specific altered forms of mononucleosomes have also been characterized. The first such product to be identified is a structurally altered non-covalent mononucleosome dimer (11,16). In addition, recent reports have shown that yeast and human SWI/ SNF complexes can create structurally altered mononucleosome monomers in which both entering and exiting DNA ends appear to be held by the same histone octamer to create a loop structure (19,20). Altered nucleosomal products could be important for SWI/SNF complex function, because altered dimers have been shown to be more accessible than normal nucleosomes to sequence specific transcription and recombination factors (21,22). The ability of certain complexes to form struc-turally altered products may be important for their specific functional roles in vivo. We do not know, however, how these altered products are related to other remodeling effects and whether DNA sequence or linker histones can modulate the products that are formed.
We are studying the evolutionarily conserved human SWI/ SNF (hSWI/SNF) 1 complex, which is essential for mammalian development and has important tumor suppressor functions (16,23). hSWI/SNF is important for transcriptional co-activation through steroid receptors and other activators, as well as co-repression through factors like the retinoblastoma protein, Rb (24). We wish to understand how hSWI/SNF regulates transcription of its target genes by examining the nature of its chromatin products. hSWI/SNF generates altered nucleosome dimers from mononucleosomes and changes the supercoiling of closed-circular plasmid chromatin (17,21,25). It also repositions histone octamers on mono-and polynucleosomal templates (18,26).
Here, we use a variety of mononucleosome templates to examine octamer repositioning by hSWI/SNF. We find that repositioning favors DNA ends, ignores canonical nucleosome positioning sequences, and is persistent enough to transfer the octamer from one end of the DNA template to the other. We also find evidence for the formation of altered nucleosome monomers by hSWI/SNF. In contrast to recent observations for yeast SWI/SNF (19), however, altered monomers are only formed on one end of our DNA template, with a normal nucleosome being formed on the other end, indicating that the structure of remodeled products might depend on the underlying DNA sequence. On nucleosomes containing the linker histone H1, we find that the pattern of repositioning by hSWI/SNF is altered, promoting octamer movement away from DNA ends and toward the middle of several template DNAs. To our knowledge this is the first demonstration that H1 can influence the nature and distribution of products formed by an ATP-dependent remodeling complex.

EXPERIMENTAL PROCEDURES
DNA Fragments-The 215-bp original (ϩ0), 215-bp ϩ100, 265-bp, and 315-bp DNA templates were all generated by PCR using the plasmid pXP10 (27) as the template. The 215-bp original fragment was amplified using primers at the EcoRI and DdeI sites, at positions Ϫ79 and ϩ137, respectively, relative to the start site of transcription of the 5S rRNA gene. The ϩ100 template, also 215 bp long, was derived from pXP10 using an upstream primer at Ϫ181 and a downstream primer at ϩ34. The 265-bp template was generated using an upstream primer at Ϫ128 and the downstream DdeI primer at ϩ137; the 315-bp fragment was amplified using the upstream Ϫ181 primer and the downstream DdeI (ϩ137) primer. The 215-bp G5E4 template was generated from the p5SG5E4 plasmid (28), using primers at positions Ϫ177 and ϩ36 relative to the E4 promoter start site. The DNA fragments were radiolabeled by 5Ј kinase end-labeling either the upstream or downstream primer before adding it to the PCR reaction. Full-length PCR products were purified from agarose gels using the QIAquick gel extraction kit (Qiagen), ethanol-precipitated, and resuspended in TE (10 mM Tris-HCl, pH 7.5, 1 mM EDTA).
Protein Purification-Histone H1 used in this study was purified from HeLa cells by hydroxylapatite chromatography, essentially as described (29). The H1 preparation appeared to be ϳ90% pure as determined by SDS-PAGE and Coomassie Blue staining. H1 concentration was determined by use of a Bio-Rad protein assay with BSA as a standard. H1 dilutions were made in BC100 buffer (100 mM KCl, 0.2 mM EDTA, 20 mM Tris-HCl, pH 7.9, 0.5 mM dithiothreitol (DTT), 0.2 mM phenylmethylsulfonyl fluoride (PMSF), 20% v/v glycerol) before being frozen and stored at Ϫ70°C. Polynucleosome cores used for nucleosome assembly by octamer transfer were purified from HeLa cells as de-scribed (30). hSWI/SNF was purified to ϳ70% homogeneity (as determined by silver staining) by affinity chromatography using the FLAGtagged Ini1 subunit of the complex, essentially as described previously (21).
Nucleosome Reconstitution-Mononucleosomes were assembled onto end-labeled DNA fragments by octamer transfer from polynucleosome cores and salt dialysis as described (29) except that Tris-HCl was substituted for HEPES in all buffers. The nucleosomes were dialyzed in 200 volumes of dialysis buffer (10 mM Tris-HCl, pH 7.5, 1 mM EDTA, 5% glycerol) for 4 h at 4°C and the different mononucleosome populations separated by EMSA in 5% acrylamide gels (29:1 acrylamide:bisacrylamide) at 200 V in 0.5 ϫ TBE. Bands corresponding to different mononucleosome positions were cut out and eluted overnight at 4°C in glycerol gradient buffer (GGB, 50 mM Tris-HCl, pH 7.5, 1 mM EDTA, 100 g/ml BSA, 5% glycerol) with rocking (31). For the 315-bp template, the nucleosome mixture was separated by glycerol gradient centrifugation to purify mononucleosomes from dinucleosomes (21). All nucleosomes were stored at 4°C, where they are stable for several weeks. For Fig. 4, 2 g of PCR-amplified probe DNA was incubated with 4 g of HeLa cell core histones (32) in 220 l of TE with 1.5 M NaCl, followed by dialysis for 1 h each into TE with 1.5, 0.93, 0.6, 0.4 M NaCl, and, finally, overnight into TE with 1 mM DTT and 0.2 mM PMSF. The assembly reaction was 55% nucleosomal, as determined by EMSA analysis.
Remodeling Assays-In general, remodeling reactions contained 0.08 ng of nucleosomes (22 pM) in 25-l reactions adjusted to 20 mM Tris, pH 7.5, 0.3 mM EDTA, 5% glycerol, 0.4 mg/ml BSA, 60 mM KCl, 5 mM MgCl 2 , 0.6 mM DTT, 0.12 mM PMSF, and, where indicated, 2 mM ATP plus 2 mM additional MgCl 2 (ATP chelates Mg 2ϩ , and the additional MgCl 2 maintains the free Mg 2ϩ concentration at 5 mM). Differences in these conditions are described in the figure legends (e.g. the nucleosome concentration in Fig. 4B was 34 nM). Note that we have found that inclusion of BSA in our nucleosome storage and hSWI/SNF reaction buffers prevents nucleosomes at low concentration from falling apart to bare DNA or changing positions (Ref. 33 and data not shown). Where indicated, H1 was added to reactions at the concentrations noted in the figure legends. H1 has been shown to bind efficiently to nucleosomes under similar conditions (34 -36). Note that, due to the purification methods for positioned nucleosomes, our probe concentrations were often near or below the apparent K d for H1 binding (estimated from titration experiments to be 100 pM or below). Thus, stoichiometric H1 binding often required H1:nucleosome ratios of greater than one (as indicated by gel shift and chromatosome stop appearance, Fig. 2). ϳ300 ng of hSWI/SNF was then added, and reactions were incubated for 1 h at 30°C and stopped by stripping hSWI/SNF and H1 from the template using competitor polynucleosomes (0.9 -1.8 g) and plasmid DNA (1-2 g). We found that identical results were seen whether H1 was preincubated with nucleosome cores before SWI/SNF addition or added at the same time as SWI/SNF. The reactions were resolved by EMSA in 5% acrylamide gels at 200 V in 0.5 ϫ TBE at 4°C for ϳ2.5 h after which the gel was dried and subjected to PhosphorImager analysis (AP Biotech). An alternative method for the formation of chromatosomes is to add H1 midway through the assembly process (at NaCl concentrations between 0.6 to 0.7 M). This method could not be used for studies of nucleosomes at specific positions, because 1) the gel mobility shift due to bound H1 prevented the separation of differentially positioned H1bound nucleosomes by EMSA and 2) treatment of any singly positioned nucleosome at 0.7 M NaCl caused delocalization of the octamer to the three positions favored during assembly (e.g. see Ref. 36 and data not shown). We do find, however, that chromatosomes formed by this method show hSWI/SNF-dependent repositioning similar to those formed by H1 addition in remodeling buffer. 2 For heat treatment analysis, nucleosomes were incubated for the indicated times at 60°C in the absence of hSWI/SNF or stop solution and resolved by EMSA as above. Modifications to these general methods are noted in the figure legends. We have found that there is little change in hSWI/SNF activities when NaCl replaces KCl, when monovalent salt is between 20 and 60 mM, and when 0.1% Nonidet P-40 is added (Refs. 21 and 37 and data not shown).
Micrococcal Nuclease Analyses-For Fig. 2C, the indicated concentrations of H1 were titrated into 25-l reactions containing 0.3 ng of purified, positioned mononucleosomes (83 pM) in standard remodeling conditions (above), except that the buffer contained 0.1% Nonidet P-40, and 30 mM KCl. The tubes were incubated at room temperature for 10 min after which CaCl 2 was added to a final concentration of 5 mM, followed immediately by the addition of 0.59 unit of MNase (Sigma, St. Louis, MO). The reaction was stopped after 5 min by the addition of 25 1 The abbreviations used are: hSWI/SNF, human SWI/SNF; BSA, bovine serum albumin; DTT, dithiothreitol; PMSF, phenylmethylsulfonyl fluoride; EMSA, electrophoretic mobility shift assay; Exo III, exonuclease III; ISWI, imitation switch; MNase, micrococcal nuclease. 2 A. Ramachandran and G. R. Schnitzler, unpublished observations. l of stop solution (30 mM EDTA, 0.4% SDS), followed by phenol extraction, addition of 10 g of glycogen, ethanol precipitation, and end labeling with T4 polynucleotide kinase (New England Biolabs) according to the manufacturer's instructions. The labeled DNA was separated from free [␥-32 P]ATP nucleotides by passing through Bio-Spin columns (Bio-Rad), resolved by electrophoresis at 250 V in 8% acrylamide gels, and visualized by PhosphorImager analysis. For Fig. 4A, 0.3 ng of each control and remodeled nucleosomal species was digested with 0.004 unit of MNase in remodeling reaction buffer supplemented with 3 mM CaCl 2 , at 37°C for 8 min, followed by DNA purification as above. Samples were then electrophoresed in a 7% acrylamide 0.5 ϫ TBE gel, transferred to Genescreenϩ nylon membrane (PerkinElmer Life Sciences), and hybridized with a 215ϩ0 probe labeled by the random primer method (30). Exonuclease III Analyses-Remodeling reactions (as described above) were scaled up 8-to 16-fold (with or without 290 pM H1) and resolved by EMSA, and isolated bands were treated with Exo III essentially as described previously (38). Briefly, the appropriate bands were cut from the gel and incubated in 0.5 ml of equilibration buffer (10 mM Tris-HCl, pH 8, 3 mM MgCl 2 , 5 mM ␤-mercaptoethanol) for 15 min while rocking, at room temperature. 50 units of Exo III (New England Biolabs) was then added, and the reaction was allowed to proceed for 30 min at room temperature with rocking, before being stopped by the addition of 20 l of 0.5 M EDTA. The gel slices were then crushed, the tubes were spun briefly in a microcentrifuge, and the supernatant was carefully removed, after which the DNA was phenol-extracted, ethanolprecipitated, and resolved by electrophoresis in an 8% sequencing gel at 65-watt constant power. The gel was dried down, and the bands were visualized by using PhosphorImager analysis. Alternatively, each isolated species was eluted from the gel (as described above) and digested with a titration of Exo III, essentially as per Ref. 39 except that remodeling reaction buffer (without ATP) was used. We found that 30 -60 units of Exo III for 10 min at 30°C gave digestion patterns comparable to the in situ Exo III method described above.

Human SWI/SNF Moves Nucleosomes to the Ends of the DNA
Template-Mononucleosome cores were assembled onto a 215-bp Xenopus somatic 5S ribosomal DNA template by histone octamer transfer from HeLa polynucleosome cores. Because a nucleosome constrains only ϳ146 bp of DNA, the 215-bp template provides room for it to assume different positions on the DNA. On EMSA gels, mononucleosomes with the histone octamer in the center migrate slowly, whereas those with the octamer on a DNA end migrate rapidly (40 -42). This allowed the separation of the three major positions of assembled mononucleosomes from each other, as well as from bare DNA and HeLa polynucleosomes, by elution from EMSA gels. The EMSA mobilities of these purified, positioned nucleosomes is shown in Fig. 1 (lanes 1-3, upper, middle, and lower bands). The individual bands were incubated with hSWI/SNF in the presence or absence of ATP, after which unlabeled competitor DNA and polynucleosomes were added to remove bound hSWI/ SNF and resolve mononucleosome products by EMSA. Without ATP, incubation with SWI/SNF followed by addition of competitor did not change the nucleosome mobilities ( Fig. 1, lanes 4 -6). However, irrespective of the initial position of the nucleosome, hSWI/SNF action in the presence of ATP always resulted in the formation of a band (RL, "Remodeled Lower") that comigrates with input "Lower," as well as another still lower band (Rl, "Remodeled lowest") ( Fig. 1, lanes 7-9). Altered dimers also appear to be formed on 215 5S mononucleosomes, as indicated by the appearance of products with similar relative mobility to dimers formed by yeast RSC (Remodels the Structure of Chromatin) and human SWI/SNF on a variety of other templates ( Fig. 1, "Di.," migrating about half as fast as mononucleosomes) (21,43,44). Addition of more hSWI/SNF reduced the levels of any remaining input band but did not otherwise change the remodeled pattern (data not shown). These results indicated that human SWI/SNF moves nucleosomes toward the DNA ends on linear mononucleosomes. Furthermore, the appearance of the Rl band indicates that the complex does not simply move nucleosomes to one of the three octamer positions favored by assembly.
Histone H1 Causes hSWI/SNF to Move Nucleosomes to Central Positions-Because H1-containing chromatin is likely to be the in vivo substrate for mammalian remodeling complexes, it is critical to understand how H1 affects the remodeling reaction. Previous studies have examined the effect of H1 on yeast and human SWI/SNF-mediated changes in the DNase pattern of mononucleosomes and enhancement of restriction enzyme accessibility on mononucleosomes and polynucleosomal arrays (9,10). To better understand how hSWI/SNF might modulate the structure of H1-containing chromatin in vivo, we wished to determine the effect of H1 on hSWI/SNF-facilitated nucleosome repositioning and altered dimer formation. Unlike the core histones, H1 can bind specifically to preformed mononucleosome cores at physiological salt concentrations (34 -36). This allowed us to study the effect of hSWI/SNF on singleposition nucleosomes bound by H1 (see "Experimental Procedures"). We titrated HeLa cell H1 into reactions containing purified, positioned nucleosomes to establish conditions for chromatosome formation. Stoichiometric binding of H1 to cores causes a distinct gel mobility shift (e.g. see Refs. 34 and 36), which we observe at between 50 and 85 pM H1 (Fig. 2B, compare Core band to Chromatosome (Chrom.) band). Similar results were seen for each of the input nucleosome positions U, M, and L (data not shown). Properly bound H1 protects ϳ10 bp of DNA on either side of the nucleosome core from digestion with MNase, resulting in a "chromatosome stop" band of ϳ166 bp compared with the "nucleosome core stop" band at ϳ146 bp (e.g. see Refs. 36 and 45). We observe a chromatosome stop band at between 43 and 85 pM H1 (Fig. 2C). At higher concentrations, multiple H1 proteins can bind to each nucleosome resulting in a greater mobility shift, observed as a smear toward the wells (asterisk in Fig. 2B) and protection of a larger DNA fragment from MNase (uncut band in Fig. 2C).
We find that H1 can dramatically alter the octamer positions of hSWI/SNF-remodeled mononucleosomes. Remodeling reactions were carried out in the presence of increasing concentrations of H1, followed by addition of competitor DNA and polynucleosomes (which removes both hSWI/SNF and H1, allowing the resolution of nucleosome core positions). Control reactions showed that incubation of nucleosomes with H1 alone, or in the presence of hSWI/SNF without ATP, followed by removal of H1 to competitor DNA resulted in no change in nucleosome positions ( Fig. 2A, compare lanes 1 and 2, and data not shown). In the presence of both hSWI/SNF and ATP, we found that reactions that contained H1 at 50 pM or above showed an increase in slow migrating bands (primarily U, but also some M), loss of the input low band (L), and no generation of fast migrating remodeled low (RL) and lowest (Rl) products ( Fig. 2A, compare lanes 3, 11, and 19 with lanes 6, 14, and 22). This suggested that H1 could alter nucleosome repositioning by hSWI/SNF. Similar results were obtained regardless of the position of the input nucleosome ( Fig. 2A, panels labeled Lower, Middle, and Upper). This effect appears to be due to the specific binding of H1 to nucleosome cores, because the alteration of hSWI/SNF products, appearance of the chromatosome stop, and H1 binding to cores by EMSA all occur at the same H1 concentrations (between 50 and 85 pM H1; Fig. 2 A-C). Similar effects on repositioning were also seen at H1 concentrations up to 350 pM, indicating that additional nonspecific interactions of H1 with template nucleosomes did not prevent the chromatosome-specific effect ( Fig. 2A). Much higher concentrations of H1 (1.2 nM and above) did inhibit all nucleosome repositioning by SWI/SNF. At these concentrations, however, H1 induces template aggregation under our reaction conditions, resulting in the formation of a pellet after a brief microcentrifugation ("pelleting assay" essentially as per Ref. 46, data not shown).
At concentrations of up to 120 pM, H1 did not significantly inhibit the formation of bands at the position of altered nucleosome dimers ( Fig. 2A, "Di."). Dimer mobility was somewhat less in the presence of H1, however, suggesting perhaps that the mobility of dimers can be influenced by octamer positions of constituent mononucleosomes. Although repositioning was not inhibited by H1 concentrations up to 350 pM, dimer formation was inhibited at H1 concentrations over 120 pM, suggesting that these two hSWI/SNF activities are differentially sensitive to nonspecific H1 binding ( Fig. 2A).
Previous studies have shown that some aspects of chromatin remodeling can be inhibited by H1, although they did not specifically examine the effect of H1 on nucleosome sliding or dimer formation (9,10). To determine if H1 inhibits either of these hSWI/SNF activities, we examined the time course of remodeling for L input nucleosomes at a moderately low H1 concentration of 51 pM, allowing us to observe the formation of H1-preferred U and M as well as non-H1-preferred Rl products in the same remodeling reaction. We found that there was relatively little octamer movement up to 10 min, at which time both H1-preferred (U and M) and non-H1 preferred (Rl) products begin to accumulate (Fig. 2D, lane 7). Rl also begins to accumulate at this time in reactions lacking H1 (Fig. 2D, lane  6). The rate of formation of putative altered dimers (Di.) was also not significantly inhibited by H1. Together, these results indicate that histone H1 binding to form a chromatosome does not inhibit the rate of repositioning and altered dimer formation by hSWI/SNF on mononucleosomes, but, instead, alters the preferred octamer position resulting from hSWI/SNF action.
Mapping of Nucleosome Positions Using Exonuclease III-Our results indicate that hSWI/SNF, in the absence of H1, causes mononucleosomes to move to the ends of the DNA. However, the EMSA assays used above cannot reveal which end the nucleosome is on. They also cannot reveal whether the products, in the presence or absence of H1, were structurally normal nucleosomes. To answer these questions, we used exonuclease III digestion to map the nucleosome boundaries of input nucleosomes and all major hSWI/SNF products. Exo III is a 3Ј to 5Ј exonuclease that requires a double-stranded DNA template. On bare DNA it will chew in from each 3Ј-end until no double-stranded DNA is left, producing a mixture of singlestranded fragments about half the length of the original DNA (see Fig. 3, A and B, lanes 1 and 2). The presence of a nucleosome will block Exo III digestion, and the size of the resulting protected fragment is indicative of nucleosome position. When the top strand of a double-stranded length of DNA is 5Ј-endlabeled, the length of the protected DNA will define the right edge of the nucleosome (see Fig. 3B, diagram). Conversely, 5Ј labeling the lower strand and digesting with Exo III will define the left edge of the nucleosome (see Fig. 3A, diagram). Theoretically, for a normal nucleosome the distance between these marks will be 146 bp. In practice, however, the distance between these marks is often 10 to 20 bp shorter, because the propensity of DNA at the edge of the nucleosome to unpeel from the octamer surface can facilitate Exo III overdigestion (39).
To map octamer positions before and after remodeling, nucleosomes were assembled onto the 215-bp 5S rDNA template with either the top (upstream/EcoRI end) or the bottom (downstream/DdeI end) strands 5Ј-labeled by PCR, using the respective end-labeled primers. Isolated upper, middle, and lower nucleosome bands from each assembly were subjected to hSWI/ SNF remodeling in the presence or absence of histone H1, after which hSWI/SNF and H1 were removed to competitor DNA, and the products separated by EMSA. Bands for each major hSWI/SNF product and controls were excised from each lane and subjected to Exo III digestion, and the resulting DNA fragments were resolved by denaturing urea-PAGE. For each nucleosome species, the location of principal cutting sites from the top-labeled and bottom-labeled templates were then used to determine the major translational position of the nucleosome on the template. For instance, for the control Lower band, the 145 bp between the top strand cut (at ϩ66) and the bottom strand cut (at Ϫ79) maps the primary octamer position of this EMSA band to the left edge of the nucleosome (see Fig. 3, A and  B, lane 5, and C, top right). Similarly, the primary positions of the control Middle and Upper bands map further from this edge, toward the middle of the fragment. Note that, some bands (notably ϩ66 for top strand-labeled templates and Ϫ6 for bottom strand-labeled templates, see " ‡" symbol) appeared to be favored sites for Exo III digestion of nucleosomes and were generated, to some degree, from all controls and products. These favored sites do not result from octamer repositioning during electrophoresis, because when each band is eluted from the gel its position in subsequent EMSA analysis is unchanged (data not shown). These sites were deemed to be real nucleosome boundaries only if they were the most intense Exo III product and were ϳ146 bp away from the primary cutting site on the other strand. In general, we found paired major top and bottom strand sites to be separated by 136 -145 bp, consistent with the presence of a normal nucleosome, and a moderate degree of Exo III overdigestion. Control Middle nucleosomes, however, appeared to have only 120 bp between major cutting sites. Additional experiments using titrations of Exo III indicate that the first major cut on the bottom strand is really at ϩ87 (dashed arrow in Fig. 3C; data not shown) but that this is followed by a rapid subsequent digestion to the Exo III-favored site at ϩ66. MNase digestion also indicates that control Middle nucleosomes protect the normal ϳ146 bp of DNA (see Fig. 4A below). The positions measured for control nucleosome are consistent with other reports on the location of the Xenopus somatic 5S ribosomal DNA nucleosome positioning sequences, which are expected to localize the nucleosome at or near the upstream edge of our template (26,47).
H1-favored hSWI/SNF Products Are Similar to Control Upper and Middle Nucleosomes-Incubation with histone H1 in the absence of hSWI/SNF, followed by removal of H1 to competitor DNA, did not alter the nucleosome positions for any of the control nucleosomes, as observed by EMSA and Exo III digestion analysis (Fig. 3, A and B, compare lanes 5 and 6, Fig.  2A, and data not shown). When the bands produced by hSWI/ SNF action in the presence of H1 (primarily U, as well as some M, Fig. 2A) were treated with Exo III, it revealed a cutting pattern very similar to control Upper and Middle nucleosomes, respectively (Fig. 3C, and compare lanes 3 and 9 in Fig. 3, A and  B, and data not shown). Combined with the EMSA experiments, these results indicate that hSWI/SNF and H1, together, tend to result in movement of the histone octamer to the more central of the positions favored by the 5S nucleosome positioning sequences, but do not otherwise alter the nucleosome.
hSWI/SNF Products Formed in the Absence of H1 Differ from Input Nucleosomes-Exo III mapping of hSWI/SNF products in the absence of H1 revealed two striking results. First, the remodeled Lower band ("RL," which comigrates with control lower, "L," by EMSA) had a histone octamer primarily positioned on the downstream end of the DNA, in contrast to the control Lower nucleosome where the octamer was positioned at the upstream end ( Fig. 3C; and Fig. 3, A and B, compare lanes  5 and 7). This indicates that hSWI/SNF does not simply move a nucleosome to the nearest DNA end but that it can move an octamer that was positioned on one end up to 70 bp to the opposite end. By contrast, the remodeled lowest band (Rl) gave full protection of the bottom strand (suggesting a nucleosome positioned on the upstream end) but only protected 100 bp of the top strand from Exo III (Fig. 3C, and Fig. 3, A and B, lane  8). This was not due to overdigestion by Exo III, because the same result was observed at several concentrations of enzyme (data not shown). These results indicate that the remodeled lowest band is an altered nucleosome in which the histone octamer has either been shifted off the DNA edge by ϳ40 to 45 bp or has been structurally altered to constrain less than the normal 146 bp of DNA. Essentially identical results were seen for hSWI/SNF products of upper, middle, or lower band input nucleosomes (data not shown). Significantly, the octamer positions of these products differ from all control nucleosome positions, suggesting that repositioning by hSWI/SNF in the absence of H1 largely ignores the 5S rDNA nucleosome positioning signals. Furthermore, the observation of an altered product only on the upstream end and a normal product on the downstream end of this template suggests that DNA sequence can play a role in determining the nature of hSWI/SNF products.
Rl Protects Only 100 bp of DNA from MNase Digestion-The histone:DNA contacts in normal nucleosomes protect ϳ146 bp of DNA from MNase digestion. Accordingly, when control Middle and Lower bands are digested with MNase the principle product is ϳ146 bp (Fig. 4A, lanes 2 and 3). The protected fragment from the hSWI/SNF-remodeled Lower product (RL) was also ϳ146 bp, indicating that the histone:DNA contacts in this product are essentially normal (Fig. 4A, lane 4). The purified remodeled lowest product (Rl), however, protected only ϳ100 bp of DNA from MNase digestion (Fig. 4A, lanes 4 and 5), consistent with the Exo III mapping results. The observation that 100 bp of DNA in Rl is protected from both exonuclease and endonuclease action suggests that this length of DNA is in relatively tight association with histones. The size of this protected region is also consistent with that observed in crosslinking studies of a similar product formed by yeast SWI/SNF (19).
All hSWI/SNF Products Appear to Contain the Full Complement of Core Histones-Tetramer nucleosomes containing only histones H3 and H4, and hexamers lacking one copy of H2A and H2B, run faster than octamer particles in EMSAs (48). To test whether the remodeled hSWI/SNF products (Rl and RL) were complete octamers, scaled-up 215-bp 5S rDNA nucleosome assemblies were treated with hSWI/SNF in the presence or absence of ATP, and nucleosomal core products resolved by EMSA (Fig. 4B, upper panels). Note that hSWI/SNF products of the same mobilities as previously observed were formed even though these reactions contained 34 nM nucleosomes that were in 2.4-fold excess over hSWI/SNF. We also see similar repositioning effects when nucleosomes (as HeLa polynucleosomes) are in ϳ20-fold excess over hSWI/SNF (data not shown). This indicates that hSWI/SNF acts catalytically to reposition nucleosomes to DNA ends and that this result is unaffected by the hSWI/SNF:nucleosome ratio. The region containing mononucleosome bands from each lane was then cut out, and the gel slice was rotated 90° (Fig. 4B, top panels), before being directly loaded onto a 17% SDS-polyacrylamide gel to resolve the core histones and DNA (Fig. 4B, bottom panels). The competitor polynucleosomes used to remove hSWI/SNF run at the top of the EMSA gel and are absent from the mononucleosome region isolated. The SDS-PAGE gel of the control reaction in the absence of ATP, where remodeling did not occur, showed the four core histone bands at the U, M, and L positions, but no bands in the remodeled lowest position (bottom left panel). Bands for all four core histones were also observed for the remodeled Lower band (RL) at similar relative intensities to controls (bottom right panel), showing that even though the histones in this product have been moved to the end of the template opposite the nucleosome positioning sequences, their stoichiometry is unchanged. The remodeled lowest band also appears to contain all four core histones. However, because of the relatively low abundance of Rl in this scaled-up reaction, we cannot rule out the possibility that this product is a histone hexamer, i.e. it has lost one H2A/H2B dimer. We think that that is unlikely, however, because re-addition of H2A and H2B dimers did not restore band Rl to the L gel shift position, as has been shown for other hexamer particles (data not shown (48)). This conclusion is also supported by the recent observation that a similar product can be formed by yeast SWI/SNF action even when the histones in octamers have been cross-linked to each other with dimethyl suberimidate (19). Together, these data suggest that the remodeled lowest band is a monomeric nucleosomal species that contains a complete histone octamer but has altered histone:DNA contacts and that the ability to create this type of altered product is evolutionarily conserved in the SWI/ SNF subfamily of remodeling complexes.
Remodeled Lowest Cannot Be Further Repositioned by hSWI/ SNF-The accumulation of the remodeled lowest product as a FIG. 4. Characterization of hSWI/SNF products by MNase digestion and silver stain. A, MNase digestion of input and hSWI/SNFremodeled nucleosome bands in the absence of H1. Lanes 2 and 3 show the digestion pattern using input Middle and Lower nucleosomes, respectively. An ϳ146-bp nucleosomal band (as well as an ϳ140-bp overdigestion product) are observed in both cases (arrow). Lanes 4 and 5 show MNase digestions using the "remodeled low" (RL) and "remodeled lowest" (Rl) bands formed after hSWI/SNF action. Although MNase digestion of RL generates the normal 146-bp band, MNase digestion of Rl primarily generates 100-to 110-bp sub-nucleosomal bands (asterisks). Lane 1 is a radiolabeled 20-bp ladder (MW; Bio-Rad). B, 2.5 g of hSWI/SNF was used to remodel 600 ng of 215ϩ0 mononucleosomes (assembled by salt dialysis from pure histones) in 125-l reactions either without (left panels) or with (right panels) ATP. The reactions were stopped after 90 min by addition of 2.5 g of plasmid DNA and 4.5 g of polynucleosomes before separation by EMSA. Regions corresponding to mononucleosome bands were excised from the gel, turned 90°(as indicated by the top panels) and placed in large 6% stacking gel wells of a 17% SDS-PAGE gel. Histones and DNA were then stained with a reducing silver stain, and the wet gel was photographed (lower panels), essentially as described (21). result of hSWI/SNF action might indicate either 1) that it is a terminal product that can no longer be acted upon by hSWI/ SNF or 2) that it is merely one of the most highly favored remodeled positions and thus accumulates at any given point in the ongoing remodeling reaction. To test these models, the hSWI/SNF-generated Rl nucleosome was purified and again subjected to hSWI/SNF action. In the absence of hSWI/SNF or ATP, no change in the mobility or percent of bare DNA in Rl fractions was observed, even after incubation at 30°C for up to 24 h (Fig. 5A, lane 1, and data not shown). Thus, despite protecting only ϳ100 bp of DNA from nucleases, Rl did not appear to be inherently unstable. Even when treated with hSWI/SNF and ATP, no change in Rl band mobility was observed, suggesting that hSWI/SNF was not capable of moving the histone octamer from this position (Fig. 5A, lanes 2 and 3). Interestingly, hSWI/SNF appears capable of forming altered dimer-like products from Rl (Fig. 5A, lane 3, "Di."), indicating that the absence of translational movement of Rl was not due to complete inhibition of hSWI/SNF enzyme activity. For control nucleosomes, H1 and hSWI/SNF together promote movement to the center of the template (Upper and Middle bands) and block the formation of Rl (Fig. 5B, lanes 1-3, and see Fig. 2A). Thus, if Rl was still an effective substrate for octamer repositioning by hSWI/SNF, it should also be converted to U in the presence of H1. This, however, did not occur (Fig. 5B, lanes  4 -6), suggesting that Rl is highly resistant to further octamer repositioning by hSWI/SNF. This cannot be explained by an inability of Rl to bind to H1, because Rl and control nucleosomes were found to bind similar concentrations of H1 in EMSA analysis (data not shown). However, we cannot rule out the possibility that H1 may bind differently to Rl, which could change the outcome of hSWI/SNF action. In addition, these results cannot be explained by an inhibition of hSWI/SNF by a contaminant in the purified Rl fractions, because when purified Rl is mixed with input Middle, the Middle band is repositioned normally by hSWI/SNF (data not shown).
hSWI/SNF-dependent Repositioning Is Similar on a Variety of Template DNAs-We wished to determine the extent to which the repositioning effects we observed were dependent on specific DNA sequences as opposed to sequence-independent structural properties of DNA ends. We reasoned that if octamer translocation ϳ45 bp off the upstream end of the template was dependent on DNA sequences, then deletion of 100 bp of DNA from the downstream end and addition of 100 bp to the upstream end would generate a template on which hSWI/SNF action would force the octamer to the middle of the template (the ϩ100 template, see Fig. 6A). If formation of the altered nucleosome Rl was also dependent on these sequences, then the centrally located nucleosome that was formed would be altered. Also, if the position to which hSWI/SNF moved the octamer in the presence of H1 was sequence-dependent, we would expect hSWI/SNF and H1 to now generate fast migrating nucleosomes with octamers bound to the downstream end of the ϩ100 template. Alternatively, if repositioning was primarily due to nonsequence-dependent structural properties of DNA ends, remodeling of the ϩ100 template would give similar results to remodeling of the original ϩ0 template.
We generated this ϩ100 template by PCR from the pXP10 plasmid (which contains additional upstream 5S rDNA sequence), assembled it into mononucleosomes and isolated each of the three major assembly products from EMSA bands. Each of these was then remodeled by hSWI/SNF in the presence or absence of H1 (Fig. 6B). Our results show that, in the absence of H1, hSWI/SNF generates primarily fast migrating bands on this template (ϩ100L), whereas in the presence of H1 hSWI/ SNF generates primarily slow moving bands (ϩ100U, ϩ100M). This suggests that, on short mononucleosome templates at least, DNA ends strongly influence the hSWI/SNF-favored position of nucleosome cores and chromatosomes. Remodeling of the ϩ100 template, however, did not generate any gel shift band migrating faster than control nucleosomes, suggesting that the formation of altered nucleosomes might depend on DNA sequence, as was also indicated by the observation that Rl only formed on the upstream end of the 215ϩ0 template.
To further examine the sequence specificity of hSWI/SNFdependent repositioning, we generated mononucleosomes from a 215-bp non-homologous sequence. Recently, we used atomic force microscopy and restriction enzyme accessibility assays to characterize nucleosome repositioning by hSWI/SNF on the polynucleosomal template 5SG5E4. Those studies indicated that a nucleosome core protecting the unique XbaI restriction site (within the 430-bp promoter/transcription cassette of the 2.5-kb construct) was moved away by hSWI/SNF, resulting in a stable increase in XbaI digestion (18). We used PCR to generate an end-labeled 215-bp template surrounding this XbaI site (G5E4) and assembled this into nucleosomes. The two major assembly products (G5E4 U and G5E4 L, Fig. 6C, lanes 1 and  4) were eluted from EMSA bands. Exo III mapping of the most abundant of these assembly products (U), revealed a centrally positioned octamer with the pseudodyad near the XbaI site, suggesting that the preferred octamer position on this sequence is the same for mononucleosomes and polynucleosomes (Ref. 18 and data not shown). The results of hSWI/SNF remodeling of this template show that, similar to the 5S 215ϩ0 template, treatment of either input band with hSWI/SNF in the absence of H1 results in an accumulation of the lower band (G5E4 RL) and also an even faster migrating lowest band (Fig. 6C, G5E4 Rl). The appearance of a "Remodeled lowest" band suggests that formation of altered products is not unique to the upstream edge of the 5S 215ϩ0 template. Treatment of either input band with both hSWI/SNF and H1 also had a similar effect on G5E4 as it did on the 5S rDNA template, causing hSWI/SNF to move nucleosomes to the slowly migrating band position, U. These results indicate that end positioning in the absence of H1 and more central positioning in the presence of H1 are sequence-independent properties of hSWI/SNF action.
End Positioning Still Occurs on Longer Templates-The tendency of hSWI/SNF to move nucleosomes to DNA ends might be expected to be weaker if the ends were further from the original octamer position. Thus, we reasoned that, on longer mononucleosome templates, preferred internal positions for remodeled octamers might be observed. We used PCR to extend the 5S rDNA 215-bp template 50 or 100 bp upstream without changing the downstream end (265 and 315 templates, see Fig.  6A) and assembled these templates into mononucleosomes. For 265, we isolated each of the three major assembly bands (U, M, and L) by EMSA. The 315 template is long enough to allow dinucleosome formation during assembly, and we used glycerol gradient ultracentrifugation to separate the mixture of 315 mononucleosomes from both dinucleosomes and bare DNA. As for the 215 template, treatment of these longer mononucleosomes with hSWI/SNF resulted primarily in the accumulation of fast migrating bands and the loss of slow migrating, centrally positioned, input bands (Fig. 7A, lanes 3, 6, 9, and 12, black arrows). Some of these remodeled bands migrated faster than any of the bands from assembly, suggesting the genera-FIG. 6. Effect of hSWI/SNF and H1 on other templates. A, schematic representation of DNA templates used to assemble nucleosomes for this study. B, effect of HI on remodeling of the ϩ100 nucleosome template by hSWI/SNF. Mononucleosomes were assembled on the 215-bp ϩ100 DNA fragment as described earlier. The three major nucleosome bands U, M, and L were isolated (lanes 1, 4, and 7, respectively) and subjected to remodeling by hSWI/SNF and ATP in the absence (lanes 2, 5, and 8) or presence of 340 pM H1 (lanes 3, 6, and 9). C, effect of HI on remodeling of the 215-bp G5E4 nucleosome template by hSWI/SNF. The two major nucleosome positions (U and L) obtained after nucleosome assembly were purified (lanes 1 and 4) and remodeled by hSWI/SNF and ATP in the presence (lanes 3 and 6) or absence (lanes 2 and 5) of 120 pM H1. tion of some altered products like the 215 Rl band. These results indicate that the tendency of hSWI/SNF to form end positioned nucleosomes is strong, even on templates long enough to accommodate several internal positions.
hSWI/SNF Does Not Move Nucleosomes to Thermally Favored Positions-Treatment of nucleosomes at elevated temperatures (from 37°C to 60°C) results in redistribution of the nucleosomes to heat treatment-preferred positions (42,49,50). The positions assumed after heat treatment generally correspond to those favored in nucleosome assemblies and are thought to reflect the most stable, lowest energy octamer positions. A recent report has shown that for the ISWI-containing NURF (Nucleosome Remodeling Factor) complex, which moves nucleosomes from central to distal positions on a template, ATP-dependent remodeling, and heat treatment have similar effects on mononucleosome positions (42). This suggests that some remodeling complexes may act by reducing the energetic barrier to nucleosome movement, allowing redistribution to the most thermodynamically favored positions. The mapping and EMSA results above indicate that this is not the case for hSWI/SNF. To further explore this, we compared the nucleosome positions after remodeling to those after incubation of the template at 60°C for 1 h. The preferred positions after hSWI/ SNF remodeling differed from those favored by heat treatment on the 265 and 315 templates (Fig. 7A, compare the arrows in lanes 3, 6, 9, and 12 to asterisks in lanes 2, 5, 8, and 11). This was also true for the original 215 5S template and the G5E4 template, where, unlike hSWI/SNF remodeling, heat treatment favored Middle band and Upper band nucleosomes, respectively (Fig. 7B, lanes 2 and 4). These results indicate that hSWI/SNF does not simply lower the activation barrier to thermally driven movement but instead follows a different set of rules for octamer repositioning. DISCUSSION

hSWI/SNF Moves Mononucleosomes Toward DNA Ends-
Our findings establish that the tendency of hSWI/SNF to move octamers to DNA ends is a general property of the complex, independent of initial octamer position, DNA sequence, and template length. Together with other studies, this suggests that the movement of nucleosomes toward DNA ends is a common property of all members of the SWI/SNF class of ATP-dependent remodeling complexes (18,19,26,51). hSWI/ SNF action may favor end-positioned octamers, because the complex recognizes ends and specifically moves nucleosomes toward them. Alternatively, ends may be favored because either 1) they are the lowest energy binding sites for remodeled nucleosomes or 2) end-positioned nucleosomes can no longer be productively repositioned by the complex. This latter possibility may well be the case for the remodeled lowest product, which appears to be resistant to further repositioning by hSWI/ SNF (Fig. 5).
Previous studies show that hSWI/SNF can also reposition central nucleosomes on polynucleosomal arrays (18) and repositioning in the absence of DNA ends is likely to be important for the transcriptional regulatory functions of hSWI/SNF in vivo. The observed tendency toward end positioning on mononucleosomes indicates that fully understanding the nature and specificity of repositioning by hSWI/SNF will require analysis of nucleosomes far from DNA ends, such as at the center of polynucleosomal arrays. The present work, however, does suggest some likely characteristics of hSWI/SNF-dependent nucleosome repositioning in vivo. For instance, the comparison of hSWI/SNF-remodeled nucleosomes to thermally repositioned nucleosomes indicates that repositioning by hSWI/SNF can ignore nucleosome positioning sequences. In this way, hSWI/ SNF action may oppose that of other remodeling complexes, like the ISWI-containing NURF (Nucleosome Remodeling Factor) complex, which has been shown to move nucleosomes to heat treatment favored sites (42). Repositioning away from naturally favored sequences could be important for the attenuation of transcription after removal of an activating signal, in that nucleosomes moved to non-natural sites might snap back to their original, preferred positions over time, re-establishing a repressive chromatin state.
Generation and Position of Altered Nucleosome Monomers-We find that some hSWI/SNF products appear to be complete histone octamers that have moved ϳ45 bp off the edge of the DNA, resulting in DNA:histone contacts that protect only 100 bp of DNA from exonuclease and endonuclease digestion. A very similar product was seen in site-specific cross-linking studies of the mononucleosomal products of yeast SWI/SNF (19) and in MNase digestion studies of hSWI/SNF-remodeled mononucleosome products (20). A less dramatically altered product that protects only ϳ125 bp of DNA from MNase digestion is also created by the ISWI ATPase (52). Intriguingly, previous studies with yeast and human SWI/SNF found that altered nucleosomes were formed on both ends of the template. Using a somewhat longer, 215-bp template, however, we find that hSWI/SNF generates normal nucleosomes on one end of the DNA and "off the edge" nucleosomes on the other. This indicates that, although a tendency toward end positioning is a sequence-independent property of hSWI/SNF, the nature of the repositioned products can be influenced by underlying DNA sequence or sequence-dependent structure. It has been proposed that the altered product formed by yeast and human SWI/SNF is stabilized by having exiting DNA wrap back onto vacated sites on the histone octamer surface, forming a loop of DNA between the normal sites for entering and exiting DNA. If this were the case for Rl, there would be no place for exonuclease III to enter the structure, and only full-length DNA would be recovered. Alternatively, the DNA that wraps back to form a loop might be in relatively weak or dynamic association with the histone octamer, allowing Exo III to push past this looped DNA to give the 100-bp protection that we observe. The formation of an intramolecular DNA loop may also help to explain why the Rl species is so stable, despite having only ϳ100 bp of Exo III-and MNase-resistant histone:DNA contacts.
H1 Alters hSWI/SNF-favored Nucleosome Positions-We find that addition of H1 can switch hSWI/SNF action from favoring end positions to favoring central positions. This effect is first observed at H1 concentrations that give a first gel shift band and a chromatosome stop band. Thus, we feel that it is likely to be due to the specific binding of H1 to form chromatosomes. In addition, if H1 were simply inhibiting repositioning by nonspecifically coating the DNA, this should result in inhibition of repositioning from all octamer locations, which we do not see (e.g. L is converted to M and U). Finally, we see similar results when chromatosomes are first formed by adding H1 at the 0.6 M NaCl step during nucleosome assembly, and nucleosome concentrations are kept well above the apparent K d for H1 binding (data not shown). This method of assembly was not ideal, however, because it gives a mixture of upper, middle, and lower nucleosomes, preventing the analysis of hSWI/SNF products from defined nucleosome positions (see "Experimental Procedures," under "Remodeling Assays").
One possible explanation for this effect of H1 might be that by binding to ϳ10 bp of both entering and exiting DNA (as suggested by structural studies (53)) H1 stabilizes nucleosomes that are at least 10 bp in from the edge of the DNA. In support of this, we find that when H1 is added during nucleosome assembly the percentage of low band nucleosomes is reduced (data not shown). The favorable free energy of proper H1 bind-ing, especially at nucleosome positioning sequences, might overwhelm the mechanistic or energetic considerations that promote end positioning by hSWI/SNF. Alternatively, H1 bound to entering and exiting DNA to form a chromatosome might inhibit further repositioning by SWI/SNF (e.g. by blocking access). This latter theory would predict that, in the presence of H1, hSWI/SNF would move an end-positioned nucleosome to the first internal site with enough flanking DNA to allow chromatosome formation. We find, however, that hSWI/ SNF action on input lower nucleosomes (at the left edge) generates some middle (ϳ35 bp from the left edge), but primarily upper nucleosomes (ϳ45 bp from the left edge), suggesting that nucleosome repositioning in the presence of H1 is not just a matter of incremental sliding until the octamer is far enough from the DNA ends. It is possible, however, that nucleosomes already in H1-preferred positions are refractory to further repositioning by SWI/SNF. This is suggested by the observation that even though hSWI/SNF treatment of the H1-bound lower band generates some middle as well as upper bands, the same treatment of upper generates very little middle band (see Fig. 2A).
We do not see a strong inhibitory effect of H1 on repositioning or altered dimer formation. However, the ability of H1 to alter repositioning specificity might help explain how H1 decreases the effects of SWI/SNF complexes in other assays. H1 was shown to inhibit the ability of hSWI/SNF to alter the DNase digestion pattern and increase the restriction enzyme accessibility of mononucleosomes by about 3-fold (9). If "off the edge" nucleosomes were partially responsible for these changes, the ability of H1 to block Rl formation could lead to decreased hSWI/SNF activity in these assays. H1 was also shown to greatly inhibit the ability of yeast and human SWI/ SNF to increase restriction enzyme accessibility at a single, central restriction site on polynucleosomes (10). Our results raise the possibility that H1 might alter hSWI/SNF repositioning specificity on polynucleosomes, which could favor octamers positioned over any given restriction site, resulting in decreased accessibility.
The results presented here are the first indication that a non-sequence-specific chromatin-associated factor, H1, can also influence the specificity of nucleosome repositioning. In contrast, a recent study showed that HMGB1, a "linker histone replacement" protein linked to transcriptional activation, increased the rate of repositioning by the ISWI ATPase or two complexes based on it, ACF (ATP-utilizing Chromatin Assembly and Remodeling Factor) and CHRAC (Chromatin Accessibility Complex), but did not change the resulting octamer location (54). Taken together, these observations suggest that different chromatin-associated proteins might regulate transcription by differentially modulating chromatin remodeling complex action. The observation that some remodeling complex activities are inhibited by H1 has suggested a model by which chromatin can be remodeled only if histone H1 is first modified or removed (9, 10). Our work suggests a novel, non-exclusive model for the regulatory interactions between linker histones and remodeling complexes; that H1 may alter the nucleosome structures and/or positions resulting from remodeling complex action. By causing hSWI/SNF (and potentially other chromatin remodelers) to shift nucleosomes to sites on chromatin that differ from those preferred in its absence, H1 could play a critical role in revealing or masking regulatory sequences in order to regulate transcription in vivo.