• We are sorry, but NCBI web applications do not support your browser and may not function properly. More information
Logo of molcellbPermissionsJournals.ASM.orgJournalMCB ArticleJournal InfoAuthorsReviewers
Mol Cell Biol. Dec 2005; 25(24): 11156–11170.
PMCID: PMC1316949

Human SWI/SNF Generates Abundant, Structurally Altered Dinucleosomes on Polynucleosomal Templates


Human SWI/SNF (hSWI/SNF) is an evolutionarily conserved ATP-dependent chromatin remodeling complex required for transcriptional regulation and cell cycle control. The regulatory functions of hSWI/SNF are correlated with its ability to create a stable, altered form of chromatin that constrains fewer negative supercoils than normal. Our current studies indicate that this change in supercoiling is due to the conversion of up to one-half of the nucleosomes on polynucleosomal arrays into asymmetric structures, termed “altosomes,” each composed of two histone octamers and bearing an asymmetrically located region of nuclease-accessible DNA. Altosomes can be formed on chromatin containing the abundant mammalian linker histone H1 and have a unique micrococcal nuclease digestion footprint that allows their position and abundance on any DNA sequence to be measured. Over time, altosomes spontaneously revert to structurally normal but improperly positioned nucleosomes, suggesting a novel mechanism for transcriptional attenuation as well as transcriptional memory following hSWI/SNF action.

Human SWI/SNF (hSWI/SNF) is a chromatin remodeling complex that has essential functions in gene regulation, hormonal signaling, and cell cycle control (for recent reviews, see references 25, 34, and 35). It is the human member of the SWI/SNF family of remodeling complexes, which is remarkably conserved from Saccharomyces cerevisiae to humans. hSWI/SNF functions as a transcriptional coactivator when recruited to target genes through interaction with steroid receptors, as well as MyoD, β-catenin, Sp1, p53, and others. It also acts as a corepressor through interactions with Rb, REST, and prohibitin. hSWI/SNF function is correlated with its in vitro chromatin remodeling activities, including the translational repositioning of normal nucleosomes (3, 38, 41) and the generation of structurally altered products from mononucleosomes (11, 22, 26, 38, 40). hSWI/SNF also generates some form of stable, structurally altered product on polynucleosomes, as evidenced by changes in the supercoiling of circular plasmid chromatin in vivo and in vitro. The left-handed wrapping of DNA around the histone octamer results in one negative supercoil per nucleosome. When treated with hSWI/SNF in vitro, that number is decreased to about half (22, 26). This change is stable but reverts back to normal on the timescale of several hours (15, 42). Critically, reversion occurs even in the presence of a vast excess of competitor DNA which would sequester any histones released by remodeling, indicating that the change in supercoiling is not due to histone loss but is instead due to some as yet unknown change in nucleosomal structure. A regulatory function for this altered structure is indicated by studies showing a correlation between SWI/SNF-dependent loss of supercoiling and transcription of genes on episomal plasmid DNAs (e.g., see references 24, 30, and 39). In addition, direct recruitment to GAL4 sites of the GAL4 DNA binding domain fused to the major hSWI/SNF catalytic ATPase, BRG1, caused a dramatic decrease in the linking number of episomal chromatin in Xenopus oocytes (21). This effect was dependent on the ATPase activity of BRG1 and mimicked the supercoiling changes observed during transcriptional activation driven by nuclear receptors, where SWI/SNF is a required coactivator (21).

What is the structural nature of hSWI/SNF-altered chromatin? While our previous studies and those of others examined the characteristics of SWI/SNF products formed from mononucleosomes (11, 23, 27, 32, 37, 38, 40), the relationship between these products and those formed on polynucleosomes, in the absence of DNA ends, has been unclear. Here, we find that the hSWI/SNF-dependent loss of nucleosome-constrained supercoils results from the formation of abundant structurally altered pairs of adjacent nucleosomes (termed “altosomes”). The properties of altosomes allow them to be readily detected on any given DNA sequence and also suggest new models for transcriptional regulation by SWI/SNF complexes.


Chromatin and protein purification.

Plasmid chromatin was reconstituted from linearized or supercoiled p5SG5E4 plasmid (36) and core histones (46), by salt dilution as in reference 41, followed by a final exchange of buffer to Tris-EDTA using centrifugation through a 10-kDa filter membrane (Millipore). Assembled and characterized 12 by 208 5S rRNA gene repeat polynucleosomal arrays with or without the major human H1 subtype, H1.0, or chicken H5 were generous gifts of J. Hansen and X. Lu (8, 9, 20, 33). Full saturation (12 octamers per 12 by 208 5S array), and lack of supersaturation, was confirmed by sedimentation velocity analysis (9). Proper stoichiometric binding of linker histone was confirmed by a characteristic shift in S value for the array (8, 33) and by the appearance of “chromatosome stop” bands after micrococcal nuclease (MNase) digestion (indicated in Fig. Fig.7).7). MCF7 polynucleosomes were purified as described in reference 5, except that 3- to 10-kb chromatin fragments were released by sonication rather than MNase digestion. hSWI/SNF was purified to ~80% homogeneity by affinity chromatography using the FLAG-tagged Ini1 subunit, as described previously (40). To demonstrate coelution of altered MNase footprinting activity with the intact hSWI/SNF complex, further purification to apparent homogeneity was achieved by ultracentrifugation on a 22 to 30% glycerol gradient as described in reference 44.

FIG. 7.
Linker histones do not block SWI/SNF remodeling. (a) Kinetics of MNase footprint changes. H1-lacking or H1-containing 12 by 208 5S rRNA gene polynucleosomal arrays were remodeled with SWI/SNF, with aliquots treated with ADP and EDTA to terminate remodeling ...

Remodeling assays.

In general, 100 ng of polynucleosomes (measured as DNA weight) was treated with 64 ng of SWI/SNF in a 25-μl reaction mixture containing 10 mM HEPES-HCl, 15 mM Tris-HCl, pH 7.5, 38 mM KCl, 1 mM MgCl2, 0.5 mM ATP, 0.05% NP-40, 1 mM dithiothreitol, 0.2 mM phenylmethylsulfonyl fluoride, and 6% glycerol. After 1 h at 30°C, reactions were stopped by adjusting them to 10 mM ADP or to 10 mM ADP plus 5 mM EDTA or by the addition of 1 unit of apyrase (as indicated in the figure legends). In control reactions, the stop reagent was added before hSWI/SNF. Control experiments showed that these stop conditions completely inhibited hSWI/SNF remodeling (data not shown). In some experiments (e.g., Fig. Fig.2,2, ,3a,3a, and and7c),7c), reversion of the altered state was stimulated by adjusting the reaction buffer to 150 mM KCl and/or incubation at 37°C after cessation of remodeling. Reversion without added salt (at 30 mM KCl) was considerably slower (e.g., Fig. Fig.6a).6a). Where noted, 20-fold excess competitor DNA was added, which stops the remodeling reaction and removes hSWI/SNF from the template (e.g., see references 15 and 38; also data not shown). Differences from these standard conditions are noted in the figure legends.

FIG. 2.
Altered footprints are not associated with histone loss. 12 by 208 5S polynucleosomal arrays were remodeled, and reactions were stopped by addition of ADP and EDTA, followed by incubation at 37°C in reaction buffer containing 150 mM KCl, in the ...
FIG. 3.
MNase footprint changes and supercoiling changes arise from the same products. Circular p5SG5E4 plasmid chromatin (5.7 kb) was treated with SWI/SNF in the presence of topoisomerase I for 1 h. After the reaction was stopped by 20× excess ADP and ...
FIG. 6.
Altosomes revert without reestablishing normal positions. 12 by 208 5S polynucleosomal arrays were remodeled, and the reaction was stopped by addition of ADP, followed by incubation at 37°C for 0 or 48 h in reaction buffer containing 30 mM KCl, ...

Supercoiling assays.

hSWI/SNF remodeling reactions on circular p5SG5E4 plasmid chromatin were as described above, with the inclusion of 120 units/ml of wheat germ topoisomerase I (Topo I; Promega). To ensure continuous topoisomerase activity, fresh enzyme was added every 2.5 h (100 U/ml). Topo I activity at each time point was confirmed by measuring the ability of an aliquot of the reaction to relax added supercoiled plasmid DNA. At each time point, 5-μl aliquots were transferred to 5 mM EDTA, 2% sodium dodecyl sulfate (SDS), and 20 μg proteinase K for 30 min at 37°C. DNA was then purified by phenol-chloroform-isoamylalcohol extraction (25:24:1) and precipitated with ethyl alcohol in the presence of 0.3 M sodium acetate and 1 μg d-glycogen as a carrier. Samples of DNA topoisomers (5 ng) were resolved by gel electrophoresis in 0.8% agarose 0.5× Tris-borate-EDTA (TBE) for 16 h at 2.3 V/cm (first dimension), soaked in running buffer containing 2 μg/ml chloroquine (an intercalating agent that adds positive supercoils to DNA) for 4 h, rotated 90°, and then run in the second dimension for 16 h at 2.3 V/cm. DNA was transferred to a positively charged nylon membrane (GeneScreen Plus; PerkinElmer) by capillary transfer and visualized by Southern blot hybridization with a random hexamer-labeled 196-bp EcoRI-to-EcoRI 5S sequence probe, and the relative intensity of each topoisomer was quantitated by phosphorimager analysis (Molecular Dynamics).

Micrococcal nuclease footprint size assays.

After SWI/SNF remodeling, reactions were stopped, and 5-μl aliquots containing 20 ng of polynucleosomes were transferred to 20 μl of reaction buffer that included 2.5 to 40 units/ml of MNase (Roche), 20 mM HEPES-HCl, pH 7.5, 30 mM KCl, 2.5 mM CaCl2, 0.05% NP-40, 1 mM dithiothreitol, 0.2 mM phenylmethylsulfonyl fluoride, and 3% glycerol. After 15 min at 30°C, digestion was stopped and DNA was purified as described for the supercoiling assay. MNase footprints were analyzed after separation on a 5% polyacrylamide gel (29:1 acryl-bis, 0.5× TBE), electrophoretic transfer to the GeneScreen Plus membrane, and Southern hybridization with the probes indicated in the figure legends. For Fig. Fig.1c,1c, 200-ng samples of remodeled and control polynucleosomes were loaded, and the gel was incubated in 1 μg/ml ethidium bromide for 20 min and then extensively washed for 1 h in water before photographing. We used the following equation to calculate the percentage of altered nucleosomes: % altered nucleosomes = 100 × (abnR/totalR − abnC/totalC)/(1 − abnC/totalC), where “abn” indicates signal from all abnormal band regions between 50 to 600 bp (subnucleosomal and internucleosomal bands; brackets to the right in figures), “total” is the total signal between 50 to 600 bp, R indicates signal from the remodeled reaction, and C indicates signal from the control reaction. Subtraction of abnC/totalC from both the top and bottom of the equation is to account for background signal in abnormal regions in the control lanes, giving a value of 0% altered for controls. For instance, for Fig. Fig.1a,1a, lanes 2 and 5, abnC/totalC equals 0.142, abnR/totalR equals 0.483, and % altered equals 40%.

FIG. 1.
Altered micrococcal nuclease footprint of SWI/SNF remodeled polynucleosomes. (a) Polynucleosomal arrays containing 12- by 208-bp sea urchin 5S rRNA gene repeat sequences were remodeled by hSWI/SNF (lanes 4 to 6) for 1 h. Remodeling was stopped with 20-fold ...

Measurement of MNase-resistant DNA.

Samples were deproteinated after digestion with 2, 4, 10, 20, or 40 units/ml MNase as described for the supercoiling assay, except that the phenol-chloroform-isoamylalcohol extraction step was omitted, to avoid DNA loss. Samples (1 μl) were dotted manually to the GeneScreen Plus membrane, followed by hybridization with a 196-bp 5S sequence probe. The signal from each remodeled dot was divided by the signal for the corresponding control dot (for a total of 70 pairs), and both the mean and standard error were determined.

Restriction enzyme analysis.

Ten nanograms of control, remodeled, or 48-h reverted 5S polynucleosomal arrays were digested with EcoRI or MspI (New England Biolabs) in 1× NEB2 buffer for 30 min at 30°C. DNA was purified and subjected to polyacrylamide gel electrophoresis (PAGE) and Southern hybridization as above. The average number of 5S repeats in the digested samples (L) was calculated by multiplying the fractional signal intensity for all bands corresponding to a given number of repeats by the corresponding repeat number. The percentage of sites cut (F) was then calculated according to the following equation: F = 100 × (12/L − 1)/11. For instance, if the only bands generated were ~400-bp bands containing two 5S repeats, the average repeat length would be 0.0 × 1 + 1.0 × 2 + 0.0 × 3… = 2, corresponding to 5 out of 11 of the internal sites in the 12-repeat array being cut, or a cutting percentage of 45.5%.

Fractionation of micrococcal nuclease products.

SWI/SNF remodeling reactions were scaled up to 2 to 5 μg of p5SG5E4 plasmid chromatin. Reactions were stopped by addition of 10 mM ADP plus 5 mM EDTA before (control) or 1 h after (remodeled) hSWI/SNF addition. The buffer was exchanged for reaction buffer lacking MgCl2, ATP, ADP, and EDTA and concentrated to 175 μl using 10-kDa filter membranes (Millipore). Polynucleosomes were then digested with MNase to primarily mononucleosome size (using 40 U/ml MNase) after addition of CaCl2 to 2.5 mM and MgCl2 to 1 mM. EDTA was then added to 15 mM to stop digestion, and 10-fold excess of competitor DNA was added to remove any template-bound hSWI/SNF. The reaction mixtures were then layered on top of linear 10 to 30% glycerol gradient (5 ml, 50 mM Tris-HCl, pH 7.5, 0.2 mM EDTA, 0.05% NP-40) and centrifuged for 16 h at 35,000 × g, and fractions were collected, as per references 22 and 40. For electrophoretic mobility shift assay (EMSA) analysis, fractions were directly loaded on 0.5× TBE 5% PAGE gels.

Analysis of histone content.

Approximately 300 μl of altered and control dinucleosomal gradient fractions were mixed with trace amounts of 32P-labeled 215-bp DNA (as a control for precipitation efficiency) and 100 μg deoxycholate, precipitated with 10% trichloroacetic acid, pelleted by microcentrifugation, and washed with acetone to remove trichloroacetic acid, before resolution by 15% SDS-PAGE gel and staining with colloidal blue protein staining kit (Invitrogen). The relative number of moles of control and remodeled dinucleosomes loaded on the gel was determined by dividing the dot blot signal from remodeled and control fractions by the average length of DNA in each fraction.


Altered MNase footprint of hSWI/SNF remodeled polynucleosomes.

We reasoned that structural changes associated with loss of supercoiling would alter the accessibility of DNA to the double-strand endonuclease, MNase. On normal polynucleosomal chromatin, MNase digests the linker DNA between nucleosomes to give an ~146-bp nucleosomal “footprint” for a single nucleosome or multiples of ~146 bp for multiple nucleosomes that are not separated by sufficient linker DNA (e.g., ~290 for dinucleosomes, ~430 for trinucleosomes, etc.; Fig. Fig.1a,1a, lane 2). This pattern of minimal MNase footprints was seen under all conditions where hSWI/SNF was absent (Fig. (Fig.1b,1b, lane 1) or where hSWI/SNF was not active due to lack of ATP (Fig. (Fig.1b,1b, lane 2), hydrolysis of ATP by apyrase before hSWI/SNF addition (Fig. (Fig.1b,1b, lane 5; “apyrase 1st”), the presence of 20-fold excess ADP as a competitive inhibitor (Fig. (Fig.1a,1a, lane 2), or the presence of both ADP and 5 mM EDTA (Fig. (Fig.2,2, lane 1). Note that, in these and later experiments, MNase concentrations were chosen to fully digest all accessible linker DNA (e.g., Fig. Fig.1a,1a, lane 2) and to minimize both underdigestion of linker DNA (which results in increased mono-, di-, and trinucleosome band lengths; e.g., Fig. Fig.1a,1a, lane 1) and overdigestion of octamer-associated DNA (which results in accumulation of ~120- and ~100-bp bands; lane 3).

The MNase footprint pattern changed markedly after remodeling in the presence of hSWI/SNF and ATP (Fig. (Fig.1b,1b, lane 3), such that each of the bands corresponding to normal mono-, di-, and trinucleosomes was greatly reduced while internucleosomal fragments were greatly increased (especially 170- to 260-bp fragments between mono- and dinucleosomes; brackets on the right). Subnucleosomal bands between 50 and 100 bp were also significantly increased (lower bracket). Importantly, this altered footprint pattern was still observed when hSWI/SNF remodeling was stopped by addition of ADP (Fig. (Fig.1a,1a, lane 5), ADP and EDTA (Fig. (Fig.2,2, lane 3), or apyrase (Fig. (Fig.1b,1b, lane 4) before MNase digestion, indicating that it was due to stable hSWI/SNF products. These footprint changes were not due to any nonspecific effect on MNase digestion rate, since control and remodeled reactions differed only in the timing of ADP or apyrase addition and were identical in composition during MNase digestion. In addition, similar effects were also observed at fourfold higher and lower MNase concentrations (Fig. (Fig.1a,1a, lanes 4 and 6). These footprint changes were not due to hSWI/SNF binding, since they persisted after removal of hSWI/SNF to competitor DNA (Fig. (Fig.2,2, compare lanes 3 and 7). This pattern was formed even when the ratio of hSWI/SNF to nucleosomes was low (e.g., 1 SWI/SNF to 23 nucleosomes in Fig. Fig.1a),1a), suggesting that it was a catalytic effect of the complex. In addition, this activity was seen to precisely comigrate with the peak of homogeneous intact hSWI/SNF complex fractionated by sedimentation velocity (data not shown). In summary, since this pattern cannot come from any arrangement of normal nucleosomes (which can only give footprints that are multiples of ~146 bp), it must be representative of stable, qualitatively-altered nucleosomal hSWI/SNF products.

The altered MNase footprint is a sequence-independent general property of hSWI/SNF action.

In the examples above, we only considered the MNase footprint pattern on the 5S rRNA gene repeat sequences present in our templates. To test whether hSWI/SNF generates altered MNase footprints on other DNA sequences, we blotted control and remodeled products and hybridized with probes to three different regions of the linear 5SG5E4 template (Fig. (Fig.1b,1b, lanes 4 to 9; see the diagram for locations of probes used). We found that MNase footprint changes occur to approximately the same degree at each location. Note that, the promoter and vector probes were each at least 1 kb from the DNA ends, proving that nucleosomes with altered MNase footprints do not require DNA ends for their formation. A similar change in MNase footprint pattern is seen when bulk, heterogeneous-sequence chromatin from MCF7 cells is remodeled by hSWI/SNF followed by ethidium bromide staining (Fig. (Fig.1c),1c), further indicating that altered nucleosomes can be created on a variety of DNA sequences. Note that, on fixed-sequence templates, fine-scale differences in sizes and intensities of the altered bands are observed (e.g., internucleosomal bands in Fig. Fig.1b,1b, lanes 4, 6, and 8, and 1c, lane 4), which probably reflect different MNase sensitivities of the underlying DNA. By contrast, the more even distribution of internucleosomal products from MCF7 cell chromatin may reflect the average distribution of hSWI/SNF product footprint lengths over all sequences (e.g., Fig. Fig.1c,1c, compare lanes 2 and 4).

Correlation between supercoiling changes and MNase changes.

Where the results from control reactions showed that all linker DNA was digested, all MNase footprints that differed from control mono-, di-, tri-, and tetranucleosome sizes were judged to be due to altered nucleosomal products. To estimate the fraction of altered nucleosomes, we calculated the ratio of altered subnucleosomal and internucleosomal bands (brackets) to total signal between 50 and 600 bp and normalized to this ratio from the control lane (see Materials and Methods). Applying this calculation to Fig. Fig.1a,1a, lanes 2 (control) and 5 (remodeled), for instance, indicated that 40% of the nucleosomes on the remodeled array had altered MNase footprints. In general, hSWI/SNF remodeling altered ~30 to 50% of the nucleosomal footprints from the polynucleosomes (see figures or legends for values from each experiment). Similar percentages of altered footprints were also seen when the analysis was limited to dinucleosomal and smaller fragments (between 50 and 300 bp; data not shown).

hSWI/SNF also has a similarly large effect on nucleosome-constrained negative supercoils (reduction of up to 50% [15, 26]). Furthermore, we find that reduced supercoiling and MNase footprint changes were generated by hSWI/SNF at approximately the same rate (data not shown). These observations suggest that both effects might arise from the same altered structures. If this were the case, then we would expect the MNase footprint change to revert to normal over time, similar to what has been observed for the supercoiling change (15). To test this, circular plasmid chromatin was treated with hSWI/SNF in the presence of Topo I to relax any unconstrained supercoils. After remodeling was stopped by addition of excess ADP and 5 mM EDTA, each sample was incubated for the indicated times at 37°C to allow reversion to occur. Half of each sample was digested with MNase (Fig. (Fig.3a)3a) and quantitated as described above. Undigested DNA from the other half was separated on two-dimensional gels to measure the abundance of each topoisomer, and average linking numbers were determined by comparison to relaxed DNA (Fig. (Fig.3b).3b). Quantitation and normalization of these results revealed a very similar reversion rate of ~50% in 1 h (Fig. (Fig.3c),3c), strongly supporting the hypothesis that the MNase footprint and supercoiling assays are measuring the same altered structure. Quantitation of the remodeled lanes before reversion showed that the percentage of nucleosomes with altered footprints was similar to the percentage loss of nucleosome-constrained supercoils (~27% versus ~34%; see Discussion).

Internucleosomal MNase footprints are associated with altered dinucleosomes.

To understand the nature of alterations behind the MNase footprint changes, we purified and characterized the nucleosomal species released by MNase digestion of remodeled and control polynucleosomal arrays. Briefly, 5SG5E4 plasmid chromatin was remodeled by hSWI/SNF or left unremodeled (ADP added before hSWI/SNF) and then digested with MNase. After addition of competitor DNA to remove hSWI/SNF binding, all nucleosomal products were separated by ultracentrifugation on linear glycerol gradients (which uses differences in S values to efficiently separate mixtures of mono-, di-, and trinucleosomes [e.g., see reference 40]). Gradient fractions were analyzed for the mobility of nucleosomal species by EMSA (Fig. (Fig.4a;4a; only peak fractions are shown), as well as for the length of purified, nucleosome-protected DNAs (Fig. (Fig.4b).4b). The relative gradient position of each peak fraction, with the top of the tube as 0.0 and the bottom as 1.0, is shown above each lane. These results show that the MNase digestion products that sediment as mononucleosomes, from both control and remodeled reactions, are associated with ~146-bp MNase-resistant DNA fragments typical of normal mononucleosomes (Fig. 4a and b, compare lanes 2 and 6). Strikingly, hSWI/SNF remodeled products that migrate as dinucleosomes (compare lanes 3 and 7) are associated primarily with DNA fragments that are significantly shorter (170 to 260 bp; average size, ~220 bp) than control dinucleosomes (~290 bp). This suggests that MNase footprints in the 170- to 260-bp range on hSWI/SNF remodeled arrays arise from altered dinucleosomes that contain two histone octamers but that protect ~30 to 120 bp less DNA from MNase digestion. Similar results were seen in each of three separate experiments and using probes to either the 5S repeat sequence or the whole plasmid. For ease of reference, we will refer to the remodeled dinucleosomal product, as it exists before MNase digestion, as the “altosome” (for altered dinucleosome). Note that DNA associated with remodeled tri- and tetranucleosomes was similarly reduced in length (Fig. 4a and b, lanes 4 and 8; also data not shown), suggesting that these products are structurally related to altosomes. An altered trinucleosome, for instance, could be composed of a normal nucleosome immediately adjacent to an altosome.

FIG. 4.
Fractionation of nucleosome products with altered MNase footprints. 5SG5E4 plasmid chromatin (2 μg) was remodeled with SWI/SNF and ATP as in Fig. Fig.1a,1a, stopped by addition of ADP and EDTA, and digested with 40 U/ml MNase. ADP and ...

Subnucleosomal DNA fragments are not strongly associated with histones.

In addition to the internucleosomal fragments associated with altered di- and trinucleosomes, hSWI/SNF products are also characterized by the presence of 50- to 100-bp subnucleosomal bands. In all cases where subnucleosomal fragments were well resolved, remodeling increased the signal in this range by 5 to 8% of the total signal in the lane (e.g., Fig. 1a and c, ,2,2, and 7a and c). This corresponds to ~20% of all altered footprint bands. After ultracentrifugation, subnucleosomal fragments (of 70-bp average length and accounting for ~8% of total DNA signal) were found in upper fractions of the remodeled gradient but were absent from the control gradient (Fig. 4a and b, compare lanes 5 and 1). The relative gradient mobility of these fragments is consistent with that of free DNA of this size (data not shown). Furthermore, these fragments had identical gel migration whether they were loaded on an EMSA gel directly (Fig. (Fig.4c,4c, lane 1) or were deproteinized first (lane 2), showing that subnucleosomal DNA was not bound by histones. This is in contrast to the effect of deproteinization on the small amount of mononucleosomes also present in these fractions. Since bare DNA templates are digested to completion under the MNase digestion conditions used here, subnucleosomal DNA fragments are likely to have been protected during MNase digestion by histone-DNA interactions that were subsequently lost upon addition of competitor DNA or during glycerol gradient purification (see Discussion).

Are hSWI/SNF-altered products more sensitive to MNase digestion?

When we calculated the average lengths of all histone-bound DNA fragments in the gradient fractions, we found an ~13% decrease in the lengths of DNAs associated with mononucleosomal or larger remodeled species (Fig. (Fig.4;4; also data not shown). This corresponds to a loss of ~70 bp of MNase-protected DNA per pair of remodeled nucleosomes. This decrease may be largely explained by the corresponding increase in subnucleosomal fragments (8% of the total). Alternatively, the DNA in remodeled arrays might be more accessible to nucleases than in control arrays. This latter possibility was tested directly by comparing the total amount of MNase-resistant DNA in control and remodeled arrays. Briefly, control and remodeled 12 by 208 5S rRNA gene polynucleosomes were treated with MNase, proteins were removed with proteinase K treatment, and samples were dotted onto a nylon membrane followed by hybridization and quantitation of the hybridization signal. Combining the results from 70 pairs of control and remodeled dots showed that remodeled chromatin protects 93% ± 2% as much DNA from MNase digestion as control chromatin. In addition, comparing total counts in 14 pairs of control versus remodeled MNase footprint lanes from reactions on linear array templates (including Fig. 1a and c, ,2,2, and 7a and c [−H1 lanes]) also showed a similar ratio of 93% ± 5%. Analysis of 26 pairs of lanes from reactions on circular plasmid chromatin, however, indicated that remodeled arrays protected 98% ± 3% as much DNA as controls. Given the magnitude of the standard errors for the ratios from linear versus circular arrays, this difference may not be significant. If real, however, it might be explained by the presence of altered structures that only form on the ends of the polynucleosomal templates. For instance, hSWI/SNF action on linear mononucleosomes results in apparent movement of the histone octamer ~50 bp off the edge of the DNA (38). If this occurred for every end-proximal nucleosome on remodeled arrays it could decrease MNase-resistant DNA by up to ~6% for 12 by 208 5S arrays. Note that, our atomic force microscopy studies on 5S arrays showed that slightly more than half of the DNA ends were nucleosome free after hSWI/SNF remodeling, arguing that this effect could reduce protection by at most 3% (41; also data not shown). The ratio of nucleosome-protected DNA could also be influenced by changes in footprint lengths after remodeling—especially generation of subnucleosomal fragments. This is because, in control experiments, the hybridization signals for short 50- or 100-bp fragments averaged only ~40% to ~60% of those for longer 200- to 500-bp fragments (N. P. Ulyanova, unpublished observation). MNase footprint size analysis indicated that an average of 30% of nucleosomes in the dot blots and 35% of those in the footprint lanes were altered (data not shown). Thus, the ~7% decrease in accessibility observed for linear arrays might arise from altered nucleosomes being ~20% more accessible than control nucleosomes. However, given possible end effects and weaker hybridization of small fragments, we cannot rule out the possibility that remodeling has less of an effect or even no effect on overall accessibility. Note also that reduced hybridization efficiency for small fragments may mean that the percentage of MNase footprint products that are subnucleosomal is greater than the measured ~5 to 8%, potentially allowing subnucleosomal fragments to fully compensate for the loss of DNA length in the gradient-isolated histone-bound remodeled products (see Discussion).

Alterations in MNase footprint are not due to histone loss.

The Saccharomyces cerevisiae SWI/SNF and RSC complexes were recently shown to be capable of transferring a subset of H2A/B dimers from one template to an excess of polynucleosomes, H3/H4 tetramer arrays, or competitor DNA (7, 47). This raised the possibility that H2A/B dimers or other core histones might be lost from hSWI/SNF-altered products. For instance, loss of one H2A/B dimer might shorten the dinucleosomal footprint to ~220 bp, while nonspecific binding of H2A/B dimers to linker DNA regions might give rise to subnucleosomal DNA fragments. Free H2A/B dimers are known to bind with moderately high affinity to double-stranded DNA under low to moderate salt conditions similar to those used here (4). In addition, bare competitor DNA appears to efficiently trap H2A/B dimers released as a result of yeast SWI/SNF action on mouse mammary tumor virus (MMTV) mononucleosomal templates (47). Accordingly, if H2A/B dimers are removed from octamers during remodeling, added competitor DNA would act as a sink for transferred histones and prevent reversion of altosomes to normal. However, we find that addition of competitor DNA does not prevent reversion of altered MNase footprints to normal (Fig. (Fig.2,2, compare lanes 3 and 4 to lanes 7 and 8), indicating that altosomes have not lost H2A/H2B dimers or other histones. This is consistent with previous studies which showed that the reduced supercoiling by hSWI/SNF-altered nucleosomes could revert to normal in the presence of competitor DNA (15).

We also compared the histone contents of gradient-purified remodeled dinucleosomes (Fig. (Fig.4d,4d, lane 4), control dinucleosomes (lane 3), and pure histone octamers (lanes 1 to 2). From quantitation of dot blots, we estimated that the molar ratio of control to remodeled dinucleosomes loaded on the gel was 1.2:1 ± 0.3. This fits with the ratio of 1.3:1 (259 ng in lane 3 versus 200 ng in lane 4) derived from quantitation of the combined Coomassie staining intensity for all four core histones (which was linear between 210 and 640 ng). The relative stoichiometry of H3-H2A-H2B-H4 between lane 2 (control octamers) and lane 4 (altered dinucleosome) was 1.0:1.1:1.1:1.1 (as normalized to H3; see the Fig. Fig.44 legend for details of quantitation). Together with the observed gradient and EMSA mobilities, these results indicate that the altered nucleosomal products containing ~220-bp DNAs contain two complete histone octamers (since loss of one H2A/B dimer per dinucleosome would result in a 1.0:0.75:0.75:1.0 ratio).

Altosomes may contain one normal and one altered nucleosome.

To further explore altosome structure, we redigested control and remodeled dinucleosome fractions with MNase and examined the resulting nucleosomal products and DNA lengths by EMSA and PAGE (Fig. 5a and b). The ~170- to 260-bp fragments in remodeled dinucleosomes largely resist redigestion (Fig. (Fig.5b,5b, lanes 6 to 8; bracket on the right), indicating that the internucleosomal footprint is an intrinsic property of the altered dinucleosomal species. By contrast, redigestion of mononucleosomes resulted in essentially identical patterns of ~146-, ~120-, and ~100-bp bands, indicating that mononucleosomal hSWI/SNF products are structurally normal with regard to MNase accessibility (data not shown). Redigestion of normal dinucleosomes resulted in release of some mononucleosomes, presumably because the short linker DNA between the two octamers was cut (Fig. (Fig.5a,5a, lane 4; 15% more of the total counts in the lane were mononucleosomes compared to the undigested control). Intriguingly, redigestion of remodeled dinucleosomes resulted in a comparable release of particles with the EMSA mobility of mononucleosomes (Fig. (Fig.5a,5a, lane 8; 16% increase). In both cases there was a similar increase in ~146-, ~120-, and ~100-bp bands characteristic of redigested normal mononucleosomes (Fig. (Fig.5b,5b, 23% in lane 4 and 26% in lane 8). Taken together, these results indicate that gradient-purified altered dinucleosomes contain a relatively normal mononucleosome linked by a partially MNase-resistant bridge to an altered nucleosome containing only 60 to 80 bp of DNA.

FIG. 5.
Redigestion of gradient fractions with MNase. Gradient-purified control (lanes 1 to 4) and remodeled (lanes 5 to 8) dinucleosomes were treated with 0, 2.5, 5, or 10 U/ml MNase (lanes 1 and 5, 2 and 6, 3 and 7, and 4 and 8, respectively), followed by addition ...

Relationship between nucleosome alteration and repositioning.

Our atomic force microscopy studies revealed that SWI/SNF repositions nucleosomes on 5S polynucleosomal arrays (41), increasing the frequency of nucleosomes without detectable linker DNA between them. This nucleosome movement is also evidenced by a decreased frequency of mononucleosomal products and an increased frequency of di- and trinucleosomal products released by MNase (Fig. (Fig.4).4). Since gradient-isolated mononucleosomes resulting from hSWI/SNF action have normal footprints, the percentage of total signal in the mononucleosomal range in any MNase footprint gel lane will closely approximate the percentage of normal, well-spaced mononucleosomes on the arrays before digestion. By this measure, the percentage of normal mononucleosomes surrounded by accessible linker DNA decreases greatly after hSWI/SNF remodeling (e.g., Fig. Fig.6a,6a, from 53% in lane 1 before remodeling to 15% in lane 2 after remodeling). Intriguingly, while a 48-h incubation resulted in 86% reversion of products with altered footprints to normal, the percentage of mononucleosomal MNase footprints remained low (Fig. (Fig.6a,6a, lane 3; 24% mononucleosomal, for 24% reversion). Note that in this experiment, reversion was done in 30 mM KCl buffer (as opposed to the 150 mM KCl buffer used for Fig. Fig.3),3), which resulted in a much slower rate of altosome reversion, consistent with previous observations for supercoiling reversion using similar low salt conditions (15; also data not shown). In each of seven separate experiments (including those shown in Fig. Fig.2,2, ,3a,3a, ,6a,6a, and and7c),7c), incubation after remodeling was approximately twice as effective at reverting altered bands to normal than in restoring the original percentage of well-spaced mononucleosomes (see legend to Fig. Fig.6d;6d; also data not shown). This indicates that the normal nucleosomes formed after altosome reversion do not resume their normal, unremodeled internucleosomal spacing.

On 5S polynucleosomal arrays, the positioning sequence in each 208-bp 5S repeat tends to place nucleosomes over the MspI restriction sites and away from the EcoRI sites (Fig. (Fig.6b;6b; ~50% of nucleosomes adopt a single position with these properties [9, 18]). On control arrays (where excess ADP prevents hSWI/SNF action), EcoRI sites are relatively accessible and the DNA fragments released upon EcoRI digestion of the array are small (Fig. (Fig.6c,6c, lane 1; average fragment length of 2.2 repeats equals 41% cutting under the conditions used here; see Materials and Methods for calculations). hSWI/SNF remodeling reduces the level of EcoRI digestion, increasing the abundance of large fragments (Fig. (Fig.6c,6c, lane 2; average fragment length of 2.8 repeats equals 29% cutting). By contrast, MspI sites, which are normally protected by positioned nucleosomes in the array (lane 4; 4.0 repeat average equals 18% cut), are more accessible after remodeling (lane 5; 3.2 repeat average equals 25% cut). Note that, before reversion, these changes could be due either to repositioning of normal nucleosomes or differential accessibility of altosomes. When altosomes have reverted to normal, however, restriction enzyme accessibility provides a measure of the new positions of the normal nucleosomes that are formed. Similar to what was observed for mononucleosome frequency, 86% altosome reversion did not greatly restore the original cutting percentages at EcoRI (34% cut) or MspI (23% cut). The degree to which each of these four properties of hSWI/SNF remodeled chromatin has reverted to normal after 48 h is summarized in Fig. Fig.6d.6d. Similar results were seen in three separate experiments (see the legend for Fig. Fig.6d).6d). Together, these results indicate that nucleosome repositioning is a relatively stable effect of hSWI/SNF that persists after altosome reversion.

Effect of linker histones on the formation of altered nucleosomes.

Linker histone H1 is a major component of metazoan chromatin, estimated to be bound to over three-quarters of chromosomal nucleosomes. To determine whether SWI/SNF can generate altosomes from more physiological linker histone-containing chromatin, we performed the MNase footprint size assay on 5S arrays containing stoichiometric H1, as confirmed by preparative ultracentrifugation (9, 33). Note that in H1-containing control arrays, the MNase pattern is shifted up ~20 bp, due to H1 binding to entering and exiting DNA (shifting the mononucleosome digestion product to ~166 bp, the “chromatosome stop”; e.g., Fig. Fig.7a,7a, lane 8, and Fig. Fig.7c,7c, lane 7). Despite this change in the footprints of control chromatin, both inter- and subnucleosomal products were still observed after hSWI/SNF remodeling (Fig. (Fig.7a,7a, lane 14, and Fig. Fig.7c,7c, lane 10; brackets). Similar results were also observed with arrays containing the chicken erythrocyte linker histone, H5 (data not shown). Control experiments showed that H1 was not in rapid exchange on and off the template under the conditions used here (since incubation with 20-fold excess competitor DNA before MNase digestion did not cause the loss of the chromatosome stop band; data not shown). Accordingly, these results indicate that hSWI/SNF can generate altosomes from nucleosomes containing stably bound H1. H1 did not greatly decrease the percentage of altered nucleosomes formed nor the rate of their formation (Fig. 7a and b). This is in contrast to a previous study showing that linker histone inhibits the ability of SWI/SNF complexes to increase restriction enzyme accessibility on polynucleosomes during continuous cycles of ATP hydrolysis (20). The differential effects of H1 in these two studies could arise from differences in assay conditions. Alternatively, they may indicate that H1 inhibits some hSWI/SNF activities but not others. H1 also did not change the reversion rate back to normal nucleosomes (Fig. 7c and d), suggesting that binding of linker histone to an altosome does not greatly alter its stability.


In this study we show that hSWI/SNF converts about half of the nucleosomes on polynucleosomal templates to altered dinucleosomal structures, termed altosomes, which are characterized by an unconventional MNase footprint pattern. Altosomes appear to be responsible for hSWI/SNF-driven changes in supercoiling, since supercoiling and MNase changes are of similarly high magnitude and revert to normal on the same timescale. Since SWI/SNF-driven supercoiling changes are correlated with transcriptional activation in vivo (15, 21, 24, 26, 30, 39), and since altosomes can also be formed in the presence of H1, altosome formation is likely to be an important aspect of SWI/SNF function. Importantly, while the supercoiling assay is limited to topologically closed circular templates and can only tell the average change in linking number over the whole template, the MNase footprint size assay can work on linear templates and can reveal the presence of altosomes at any chosen sequence. Accordingly, use of the MNase footprint assay is expected to allow a more detailed examination of the function of hSWI/SNF products in transcriptional regulation.

Intriguingly subnucleosomal and/or internucleosomal MNase digestion patterns suggestive of altosomes have been observed on the promoters of several hSWI/SNF-regulated genes upon transcriptional activation in vivo or in vitro (including HSP70 [29, 45], MMTV [6, 14], and promoters activated by the progesterone receptor [31] or the heterodimer of thyroid hormone receptor and retinoic acid receptor [TR-RXR] [21, 48]). For example, activation of the Xenopus TRA gene promoter by hormone-bound TR-RXR resulted in formation of ~100-bp subnucleosomal bands as well as an increased signal between mono- and dinucleosomal bands, which correlated with the loss of negative supercoils on the minichromosome (48, 49). These studies suggest that altosomes with both altered MNase footprints and altered supercoiling may be formed during transcriptional activation of SWI/SNF target genes.

Transcriptional activation is frequently associated with the apparent “disruption” of positioned promoter nucleosomes (as measured by MNase digestion followed by indirect end labeling or primer extension). For instance, on the MMTV viral and yeast Suc2 promoters, sequence-specific patterns of MNase cuts indicative of well-positioned repressive nucleosomes are lost upon promoter activation, in a SWI/SNF-dependent manner (14, 19). These and other observations of nucleosome “disruption” suggested that SWI/SNF complexes either removed these nucleosomes from the DNA or moved them so that they lost their sequence-specific positioning. Our findings now suggest an additional possible explanation, that these “disrupted” nucleosomes are altosomes whose positions cannot be readily mapped, since indirect end-labeling techniques require that each nucleosome protect ~146 bp of DNA to give meaningful results.

Models for altosome structure.

Three possible models for altosome structure are presented in Fig. Fig.8.8. While each model fits well with the bulk of our data (namely inter- and subnucleosomal footprints, and two intact octamers), each has its particular strengths and weaknesses. Model I simply postulates that in moving one nucleosome (lower nucleosome in Fig. Fig.8a),8a), hSWI/SNF can peel some of the DNA off of an adjacent nucleosome. This results in the formation of an altered dinucleosome (Fig. (Fig.8b)8b) containing one normal nucleosome connected by a very short linker to a nucleosome bearing only half the normal length of DNA (~70 to 80 bp). This would have an end-to-end footprint of ~220 bp, explaining the internucleosomal bands of altosomes. In addition, the length of DNA protected by each altered dinucleosome would be reduced by ~25%, consistent with the apparent ~20% decrease in nucleosome-protected DNA per nucleosome with altered footprints. If the linker DNA in this structure was cut it would give rise to normal ~146-bp mononucleosome-sized bands as well as the observed ~80-bp subnucleosomal bands (consistent with the redigestion results seen in Fig. Fig.5a).5a). Note however, that if all subnucleosomal bands arise from digestion of altosomes to give ~80- and ~146-bp products, we will have underestimated the percentage of altered nucleosomes (since the assumption used in our calculations, that ~146-bp bands only arise from normal mononucleosomes, will not hold). For instance, in Fig. Fig.2,2, lane 3, where remodeling-specific subnucleosomal bands account for 6.9% of the total signal, this effect would increase the percent remodeled from 41% to ~75%. If this were the case, then very few normal mononucleosomes might remain after hSWI/SNF action. Alternatively, ~80-bp subnucleosomal fragments might arise from a structurally distinct hSWI/SNF product, such as an altered mononucleosome. However, different altered mono- and dinucleosomal products might be expected to have different stabilities, and we saw no apparent differences in the rates of disappearance of internucleosomal versus subnucleosomal footprint products in reversion experiments.

FIG. 8.
Hypothetical models for altosome structure. (a) hSWI/SNF-driven movement of the lower nucleosome might pull DNA off the surface of the upper nucleosome. (b) This might create an altered structure containing an intact nucleosome as well as a partially ...

The major drawback to model I is that it cannot adequately account for the observed percentage loss of negative supercoils. In this model each pair of altered nucleosomes will go from constraining −2 supercoils to −1.5. If we assume that our calculation from MNase footprinting of percent altered is accurate (e.g., 27% in Fig. Fig.3),3), the theoretical reduction in supercoiling would be only 6.5% as opposed to the observed 34%. Even if we were to assume that 100% of nucleosomes are converted to altosomes, this would only produce an ~25% decrease in negative supercoiling (whereas we and others have observed as much as a 50% drop in negative supercoiling after hSWI/SNF action [15; also data not shown]).

In Fig. 8c to g, we present two alternative models which can better explain the supercoiling results. Model II stems from evidence suggesting that remodeling complexes move nucleosomes by generating a loop or bulge of DNA on one end of the nucleosome, which then travels across the octamer surface and is released on the other side (Fig. (Fig.8c;8c; for a review, see reference 28). It then postulates that when two nucleosomes have been moved together such that there is no linker DNA left between them, further hSWI/SNF action will cause a loop of DNA to be trapped on the surface of one of the nucleosomes (Fig. (Fig.8d).8d). In this model, MNase digestion at the loop and at entering and exiting DNA would yield both the ~220-bp internucleosomal and ~70-bp subnucleosomal fragments observed. Since a crossed loop does not require any broken histone-DNA contacts, the model allows for a negligible change in overall MNase accessibility. Alternatively, torsional stress in the loop as well as the effective presence of two new access points into the nucleosome (at the base of the loop) might significantly increase MNase digestion of remodeled chromatin (consistent with the apparent reduction of nucleosome-protected DNA after remodeling). Variability in loop position on the octamer surface could explain the observed heterogeneity of altosome MNase footprint sizes. In addition, DNA supercoiling would be reduced from two to one due to the positive crossing of the loop.

It has been argued (based on theoretical modeling studies) that a crossed loop is more energetically costly than an uncrossed bulge that spans a significant fraction of the octamer surface and breaks many histone-DNA contacts (13). Recent studies, however, indicate that the energetic cost of bending DNA is significantly lower than previously thought (less than 8 kcal/mol for a 94-bp circle, as opposed to over 13 [10]). Estimations based on these new values suggest that an ~70-bp crossed loop would have a similar energetic cost as a planar bulge which accommodates this same 70 bp of added DNA by breaking one-third of the histone DNA contacts (+10 versus +8 kcal/mol, considering that each 10.5 bp of broken histone-DNA contacts costs ~0.9 kcal/mol and that the cost to bend DNA is proportional to [degrees of arc/bp]2). Since the free energy of nucleosome formation is approximately −12 kcal/mol, a dinucleosomal structure (total free energy of −24 kcal/mol) might be able to accommodate a DNA loop with a cost of +10 kcal/mol without being greatly destabilized. This same loop would however destabilize a mononucleosome, which might explain why we found no evidence for stable, remodeled mononucleosomal structures. If such constrained loops exist on hSWI/SNF products, they are likely to be small, since our recent atomic force microscopy studies of hSWI/SNF remodeled polynucleosomes showed no evidence for loops protruding from the surface of remodeled structures (under imaging conditions sufficient to detect ~15 nm or ~90-bp loops) or for a reduction of overall polynucleosomal array length which might result from formation of constrained loops (41). Alternatively, loops may not have been observed in those studies because they were destabilized by interaction with the charged mica surface used for imaging (41).

Model II only reduces the number of negative supercoils by one per pair of altered nucleosomes. Thus, to get the 34% decrease in supercoiling observed in Fig. Fig.33 would require 68% of array nucleosomes to have altered footprints, instead of the measured 27% altered. This might be possible if we assumed that a large portion of apparently normal mononucleosomal bands actually come from digestion of altered dinucleosomes at the linker DNA between the two octamers, resulting in an underestimation of altered products. One final structural model, however, does not require this assumption, since it would predict an approximately equal change in footprinting and supercoiling (Fig. 8e to g, model III). If DNA enters near the pseudodyad of an octamer, wraps to the normal entry point, and then crosses over to the normal exit point before finally exiting near the pseudodyad, it would create a structure that constrains one positive rather than one negative supercoil (Fig. (Fig.8e).8e). This structure is unlikely to be stable as shown, due to the high cost of bending the DNA to make the bridge between entry and exit points. However, consider the case of two immediately adjacent nucleosomes. If action of a remodeling complex were to disrupt the histone-DNA contacts over about half of the leftmost nucleosome (as suggested for the bulge propagation model proposed in reference 13), it would result in a dinucleosome containing one partially unwrapped mononucleosome. If linker DNA from the far side of the rightmost nucleosome replaced these lost contacts, this would result in a stable altosome structure which is topologically similar to that in Fig. Fig.8e,8e, except that the bridge between entry and exit sites is occupied by a second nucleosome, with relatively little energetic penalty (Fig. (Fig.8g).8g). This structure constrains zero total supercoils and thus fits best with our measurements for supercoiling and MNase changes. On the other hand, it may not adequately explain the observed asymmetry of MNase footprint sizes, since there is no a priori reason to predict that one DNA bridge will be more accessible than the other. It also cannot readily explain the observed heterogeneity in MNase fragment lengths. Note that this model assumes no major changes in the underlying structure of the histone octamer itself. Another possibility, however, is suggested by studies showing that the H3/H4 tetramer can wrap DNA in either a positive or negative superhelix, facilitated by changes in the H3-H3 protein interface (16, 17). While stable histone octamer structures with positive supercoiling were not observed after salt dialysis assembly, hSWI/SNF action combined with interactions with a second nucleosome could potentially promote the formation and stabilization of such a structure. Ongoing biochemical studies are aimed at distinguishing between these possible models for altosome structure.

Similarities and differences between hSWI/SNF-altered mono- and polynucleosomal products.

Yeast or human SWI/SNF action on mononucleosomal templates results in altered mononucleosomes in which the histone octamer has been slid ~50 bp off the edge of the template or in which exiting DNA appears to form a loop that wraps back onto the normal entry site of the histone octamer. These altered mononucleosomes appear similar to altosomes in that they have subnucleosomal MNase footprints (11, 12, 23, 38). We find no evidence for stable mononucleosomes with altered footprints on remodeled arrays (using similar reaction conditions to those used in mononucleosome remodeling assays [38]), suggesting that the DNA ends on mononucleosome templates allow for the stabilization of remodeled products that are not stable on polynucleosomes. However, the ability of hSWI/SNF to create off-the-edge mononucleosomes or loop mononucleosomes may be related to an ability to create altosomes of the type shown in model I or in model III, respectively. hSWI/SNF action on mononucleosomal templates also results in altered noncovalent mononucleosome dimers, which are characterized by subnucleosomal MNase footprints, as well as greater protection of full-length DNA from both MNase and exonuclease III digestion, compared to control mononucleosomes (40, 41). In addition, we find that altered mononucleosome dimers revert to normal mononucleosomes on a timescale similar to that of altosome reversion (N. P. Ulyanova and G. R. Schnitzler, unpublished observations). Despite these similarities, the structures of dimers and altosomes are unlikely to be identical, because of the expected influence of DNA ends on dimer structure and because the internucleosomal footprint size we observe for altosomes is longer than the 155-bp mononucleosome templates that can support dimer formation.

Models for transcriptional regulation by altosomes.

How might altosomes regulate transcription? If the altosome were to cover less DNA than two normal nucleosomes (e.g., model I), this would increase the likelihood that any transcription factor binding site was in accessible linker DNA. Even if the amount of histone-associated DNA was unchanged (as allowed in models II and III), factor access to specific sites might be increased. This is because the accessibility of any nucleosomal DNA sequence to transcription factors decreases the further it is from the entry and exit points (since breaking the histone-DNA interactions at a central location effectively requires breaking all interactions between that position and the entry or exit point [1, 2]). A nucleosome which constrains either a crossed loop (model II, Fig. Fig.8d)8d) or a second nucleosome (model III, Fig. Fig.8g)8g) on its surface has two additional sites where DNA enters and exits (at the base of the loop or at the bridges), which could provide greatly increased accessibility of transcription factors to sequences near these sites. Finally, the tendency of altosomes to revert to normal over time, together with the observation that hSWI/SNF remodeled polynucleosomes are less stable to surface deposition than control polynucleosomes (41), suggests that altosomes are a higher energy nucleosomal state that might be less resistant to polymerase II passage during elongation or that might facilitate histone removal by chaperones or other factors.

In addition to forming altered nucleosomes, hSWI/SNF is also expected to regulate transcription factor access by changing nucleosome positions. Intriguingly, we find that altosomes revert to normal with a half-life of ~1 h at near-physiological salt concentrations but that nucleosome positions are restored to normal much more slowly. This suggests that altosome formation and nucleosome repositioning may have distinct, temporally staged regulatory effects. For instance, reversion of altosomes may help repress transcription soon after the loss of an activating signal, but repositioned nucleosomes may keep the promoter in a more readily activatable chromatin configuration for much longer. Future studies, using the MNase footprint size assay combined with nucleosome mapping techniques, are expected to help elucidate how specific, stable alterations in chromatin structure regulate hSWI/SNF target gene transcription.


This work was supported by grants to G.R.S. from The Medical Foundation, the National Cancer Institute (CA088835), and the American Cancer Society (RSG-04-188-01-GMC).

We thank Tony Imbalzano, Bob Kingston, and Aruna Ramachandran for critical comments on the manuscript. We also thank Jeff Hansen for 12 by 208 5S arrays with and without H1, Jerry Workman for the p5SG5E4 template, and the National Cell Culture Center for large-scale HeLa FLAG-Ini1 culture.


1. Anderson, J. D., A. Thastrom, and J. Widom. 2002. Spontaneous access of proteins to buried nucleosomal DNA target sites occurs via a mechanism that is distinct from nucleosome translocation. Mol. Cell. Biol. 22:7147-7157. [PMC free article] [PubMed]
2. Anderson, J. D., and J. Widom. 2000. Sequence and position-dependence of the equilibrium accessibility of nucleosomal DNA target sites. J. Mol. Biol. 296:979-987. [PubMed]
3. Aoyagi, S., and J. J. Hayes. 2002. hSWI/SNF-catalyzed nucleosome sliding does not occur solely via a twist-diffusion mechanism. Mol. Cell. Biol. 22:7484-7490. [PMC free article] [PubMed]
4. Aragay, A. M., P. Diaz, and J. R. Daban. 1988. Association of nucleosome core particle DNA with different histone oligomers—transfer of histones between DNA-(H2A,H2B) and DNA-(H3,H4) complexes. J. Mol. Biol. 204:141-154. [PubMed]
5. Ausubel, F. M., R. Brent, R. E. Kingston, D. D. Moore, J. G. Seidman, J. A. Smith, and K. Struhl (ed.). 2001. Current protocols in molecular biology, vol. 4. John Wiley & Sons, Inc., New York, N.Y.
6. Belikov, S., B. Gelius, G. Almouzni, and O. Wrange. 2000. Hormone activation induces nucleosome positioning in vivo. EMBO J. 19:1023-1033. [PMC free article] [PubMed]
7. Bruno, M., A. Flaus, C. Stockdale, C. Rencurel, H. Ferreira, and T. Owen-Hughes. 2003. Histone H2A/H2B dimer exchange by ATP-dependent chromatin remodeling activities. Mol. Cell 12:1599-1606. [PMC free article] [PubMed]
8. Carruthers, L. M., J. Bednar, C. L. Woodcock, and J. C. Hansen. 1998. Linker histones stabilize the intrinsic salt-dependent folding of nucleosomal arrays: mechanistic ramifications for higher-order chromatin folding. Biochemistry 37:14776-14787. [PubMed]
9. Carruthers, L. M., C. Tse, K. P. Walker III, and J. C. Hansen. 1999. Assembly of defined nucleosomal and chromatin arrays from pure components. Methods Enzymol. 304:19-35. [PubMed]
10. Cloutier, T. E., and J. Widom. 2004. Spontaneous sharp bending of double-stranded DNA. Mol. Cell 14:355-362. [PubMed]
11. Fan, H. Y., X. He, R. E. Kingston, and G. J. Narlikar. 2003. Distinct strategies to make nucleosomal DNA accessible. Mol. Cell 11:1311-1322. [PubMed]
12. Flaus, A., and T. Owen-Hughes. 2003. Dynamic properties of nucleosomes during thermal and ATP-driven mobilization. Mol. Cell. Biol. 23:7767-7779. [PMC free article] [PubMed]
13. Flaus, A., and T. Owen-Hughes. 2003. Mechanisms for nucleosome mobilization. Biopolymers 68:563-578. [PubMed]
14. Fryer, C. J., and T. K. Archer. 1998. Chromatin remodelling by the glucocorticoid receptor requires the BRG1 complex. Nature 393:88-91. [PubMed]
15. Guyon, J. R., G. J. Narlikar, E. K. Sullivan, and R. E. Kingston. 2001. Stability of a human SWI-SNF remodeled nucleosomal array. Mol. Cell. Biol. 21:1132-1144. [PMC free article] [PubMed]
16. Hamiche, A., V. Carot, M. Alilat, F. De Lucia, M.-F. O'Donohue, B. Revet, and A. Prunell. 1996. Interaction of the histone (H3-H4)2 tetramer of the nucleosome with positively supercoiled DNA minicircles: potential flipping of the protein from a left- to a right-handed superhelical form. Proc. Natl. Acad. Sci. USA 93:7588-7593. [PMC free article] [PubMed]
17. Hamiche, A., and H. Richard-Foy. 1998. The switch in the helical handedness of the histone (H3-H4)2 tetramer within a nucleoprotein particle requires a reorientation of the H3-H3 interface. J. Biol. Chem. 273:9261-9269. [PubMed]
18. Hansen, J. C., J. Ausio, V. H. Stanik, and K. E. van Holde. 1989. Homogeneous reconstituted oligonucleosomes, evidence for salt-dependent folding in the absence of histone H1. Biochemistry 28:9129-9136. [PubMed]
19. Hirschhorn, J. N., S. A. Brown, C. D. Clark, and F. Winston. 1992. Evidence that SNF2/SWI2 and SNF5 activate transcription in yeast by altering chromatin structure. Genes Dev. 6:2288-2298. [PubMed]
20. Horn, P. J., L. M. Carruthers, C. Logie, D. A. Hill, M. J. Solomon, P. A. Wade, A. N. Imbalzano, J. C. Hansen, and C. L. Peterson. 2002. Phosphorylation of linker histones regulates ATP-dependent chromatin remodeling enzymes. Nat. Struct. Biol. 9:263-267. [PubMed]
21. Huang, Z. Q., J. Li, L. M. Sachs, P. A. Cole, and J. Wong. 2003. A role for cofactor-cofactor and cofactor-histone interactions in targeting p300, SWI/SNF and mediator for transcription. EMBO J. 22:2146-2155. [PMC free article] [PubMed]
22. Imbalzano, A. N., G. R. Schnitzler, and R. E. Kingston. 1996. Nucleosome disruption by human SWI/SNF is maintained in the absence of continued ATP hydrolysis. J. Biol. Chem. 271:20726-20733. [PubMed]
23. Kassabov, S. R., B. Zhang, J. Persinger, and B. Bartholomew. 2003. SWI/SNF unwraps, slides, and rewraps the nucleosome. Mol. Cell 11:391-403. [PubMed]
24. Kim, Y., and D. J. Clark. 2002. SWI/SNF-dependent long-range remodeling of yeast HIS3 chromatin. Proc. Natl. Acad. Sci. USA 99:15381-15386. [PMC free article] [PubMed]
25. Klochendler-Yeivin, A., C. Muchardt, and M. Yaniv. 2002. SWI/SNF chromatin remodeling and cancer. Curr. Opin. Genet. Dev. 12:73-79. [PubMed]
26. Kwon, H., A. N. Imbalzano, P. A. Khavari, R. E. Kingston, and M. R. Green. 1994. Nucleosome disruption and enhancement of activator binding by a human SWI/SNF complex. Nature 370:477-481. [PubMed]
27. Kwon, J., K. B. Morshead, J. R. Guyon, R. E. Kingston, and M. A. Oettinger. 2000. Histone acetylation and hSWI/SNF remodeling act in concert to stimulate V(D)J cleavage of nucleosomal DNA. Mol. Cell 6:1037-1048. [PubMed]
28. Langst, G., and P. B. Becker. 2004. Nucleosome remodeling: one mechanism, many phenomena? Biochim. Biophys. Acta 1677:58-63. [PubMed]
29. Lee, M.-S., and W. T. Garrard. 1991. Transcription-induced nucleosome ‘splitting’: an underlying structure for DNase I sensitive chromatin. EMBO J. 10:607-615. [PMC free article] [PubMed]
30. Li, Q., A. Imhof, T. N. Collingwood, F. D. Urnov, and A. P. Wolffe. 1999. p300 stimulates transcription instigated by ligand-bound thyroid hormone receptor at a step subsequent to chromatin disruption. EMBO J. 18:5634-5652. [PMC free article] [PubMed]
31. Liu, Z., J. Wong, S. Y. Tsai, M. J. Tsai, and B. W. O'Malley. 1999. Steroid receptor coactivator-1 (SRC-1) enhances ligand-dependent and receptor-dependent cell-free transcription of chromatin. Proc. Natl. Acad. Sci. USA 96:9485-9490. [PMC free article] [PubMed]
32. Lorch, Y., B. R. Cairns, M. Zhang, and R. D. Kornberg. 1998. Activated RSC-nucleosome complex and persistently altered form of the nucleosome. Cell 94:29-34. [PubMed]
33. Lu, X., and J. C. Hansen. 2004. Identification of specific functional subdomains within the linker histone H10 C-terminal domain. J. Biol. Chem. 279:8701-8707. [PubMed]
34. Martens, J. A., and F. Winston. 2003. Recent advances in understanding chromatin remodeling by Swi/Snf complexes. Curr. Opin. Genet. Dev. 13:136-142. [PubMed]
35. Muller, C., and A. Leutz. 2001. Chromatin remodeling in development and differentiation. Curr. Opin. Genet. Dev. 11:167-174. [PubMed]
36. Neely, K. E., A. H. Hassan, A. E. Wallberg, D. J. Steger, B. R. Cairns, A. P. Wright, and J. L. Workman. 1999. Activation domain-mediated targeting of the SWI/SNF complex to promoters stimulates transcription from nucleosome arrays. Mol. Cell 4:649-655. [PubMed]
37. Patenge, N., S. K. Elkin, and M. A. Oettinger. 2004. ATP-dependent remodeling by SWI/SNF and ISWI proteins stimulates V(D)J cleavage of 5 S arrays. J. Biol. Chem. 279:35360-35367. [PubMed]
38. Ramachandran, A., M. Omar, P. Cheslock, and G. R. Schnitzler. 2003. Linker histone H1 modulates nucleosome remodeling by human SWI/SNF. J. Biol. Chem. 278:48590-48601. [PubMed]
39. Ryan, M. P., R. Jones, and R. H. Morse. 1998. SWI-SNF complex participation in transcriptional activation at a step subsequent to activator binding. Mol. Cell. Biol. 18:1774-1782. [PMC free article] [PubMed]
40. Schnitzler, G., S. Sif, and R. E. Kingston. 1998. Human SWI/SNF interconverts a nucleosome between its base state and a stable remodeled state. Cell 94:17-27. [PubMed]
41. Schnitzler, G. R., C. L. Cheung, J. H. Hafner, A. J. Saurin, R. E. Kingston, and C. M. Lieber. 2001. Direct imaging of human SWI/SNF-remodeled mono- and polynucleosomes by atomic force microscopy employing carbon nanotube tips. Mol. Cell. Biol. 21:8504-8511. [PMC free article] [PubMed]
42. Schnitzler, G. R., S. Sif, and R. E. Kingston. 1998. A model for chromatin remodeling by the SWI/SNF family. Cold Spring Harbor Symp. Quant. Biol. 63:535-543. [PubMed]
43. Schwarz, P. M., A. Felthauser, T. M. Fletcher, and J. C. Hansen. 1996. Reversible oligonucleosome self-association: dependence on divalent cations and core histone tail domains. Biochemistry 35:4009-4015. [PubMed]
44. Sif, S., A. J. Saurin, A. N. Imbalzano, and R. E. Kingston. 2001. Purification and characterization of mSin3A-containing Brg1 and hBrm chromatin remodeling complexes. Genes Dev. 15:603-618. [PMC free article] [PubMed]
45. Sullivan, E. K., C. S. Weirich, J. R. Guyon, S. Sif, and R. E. Kingston. 2001. Transcriptional activation domains of human heat shock factor 1 recruit human SWI/SNF. Mol. Cell. Biol. 21:5826-5837. [PMC free article] [PubMed]
46. Utley, R. T., T. A. Owen-Hughes, L.-J. Juan, J. Cote, C. C. Adams, and J. L. Workman. 1996. In vitro analysis of transcription factor binding to nucleosomes and nucleosome disruption/displacement. Methods Enzymol. 274:276-291. [PubMed]
47. Vicent, G. P., A. S. Nacht, C. L. Smith, C. L. Peterson, S. Dimitrov, and M. Beato. 2004. DNA instructed displacement of histones H2A and H2B at an inducible promoter. Mol. Cell 16:439-452. [PubMed]
48. Wong, J., D. Patterton, A. Imhof, D. Guschin, Y. B. Shi, and A. P. Wolffe. 1998. Distinct requirements for chromatin assembly in transcriptional repression by thyroid hormone receptor and histone deacetylase. EMBO J. 17:520-534. [PMC free article] [PubMed]
49. Wong, J., Y. B. Shi, and A. P. Wolffe. 1997. Determinants of chromatin disruption and transcriptional regulation instigated by the thyroid hormone receptor: hormone-regulated chromatin disruption is not sufficient for transcriptional activation. EMBO J. 16:3158-3171. [PMC free article] [PubMed]

Articles from Molecular and Cellular Biology are provided here courtesy of American Society for Microbiology (ASM)
PubReader format: click here to try


Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...


Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...