INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

The wrapping of DNA around histone octamers to form nucleosomes blocks access of DNA binding factors and/or advancing polymerases, resulting in inhibition of transcription, recombination, and replication. In order for these processes to occur, nucleosomes need to be either (i) modified to make them less inhibitory or (ii) moved away from regulatory sequences or advancing polymerases. Two distinct classes of complexes are believed to carry out these functions. Histone acetyltransferases covalently modify histone N termini but do not alter nucleosome positions. In contrast, an evolutionarily conserved family of ATP-dependent nucleosome-remodeling complexes can both noncovalently modify and reposition nucleosomes in chromatin (for reviews, see references 13 and 14).

The SWI/SNF subfamily of remodeling complexes is highly conserved between yeast and humans. Members appear to be functionally conserved in terms of their effects on nucleosomes: where two complexes have been carefully compared, they have almost always shown similar activities (for reviews, see references 13 and 35). Some of the remodeling effects introduced by SWI/SNF complexes are transient, requiring continuous ATP hydrolysis to be observed (18, 19), while others are stable (5, 12, 20, 21, 28, 29, 31, 34). One stable effect is the formation of a novel nucleosome structure that results from SWI/SNF remodeling of mononucleosomes. This structure appears to be a noncovalently bound dimer of nucleosomes, as judged by chromatographic and sedimentation size estimates and the stoichiometry of its components. The DNA in these dimers is thought to wrap around the histone octamers in an altered manner, as judged by changes in DNase, micrococcal nuclease, restriction enzyme, and Gal4 access (20, 31). Very little is known, however, about the altered dimer's gross structure and the mechanism by which it is formed.

SWI/SNF and related complexes also remodel polynucleosomal templates. Several aspects of this remodeling are stable after removal of ATP from the reaction mixture and/or SWI/SNF from the template (for a review, see reference 13). One stable change is seen in a supercoiling assay, in which human SWI/SNF (hSWI/SNF) or yeast SWI/SNF reduces the degree of negative supercoiling of plasmid chromatin without apparent nucleosome loss, suggesting the presence of nucleosomes around which DNA wraps in a nonstandard manner (7, 11, 12, 32). Other stable changes are indicated by alterations in endonuclease cleavage patterns by yeast SWI/SNF (12). Remodeling enhances endonuclease cutting at sites normally blocked by the presence of a nucleosome and diminishes cutting at sites normally free of a nucleosome. Furthermore, on an array of evenly spaced nucleosomes, micrococcal nuclease digestion results in evenly spaced cuts on the DNA, while on yeast SWI/SNF-treated arrays, the cutting appears random. These changes suggest that the nucleosome positions in polynucleosomes are altered by remodeling, consistent with the observation that several remodeling complexes have been shown to reposition mononucleosomes (16, 22, 28, 34). However, these assays cannot distinguish between normal and structurally altered nucleosomes and do not provide information about the distribution of nucleosomes throughout individual arrays. One transmission electron microscopic study reported that yeast SWI/SNF could bind arrays at two positions, forming a loop, and that nucleosomes within that constrained loop had altered properties. It is unclear, however, whether these changes would be stable upon the removal of yeast SWI/SNF (2).

Here we have further investigated the structure of hSWI/SNF-remodeled products by using atomic force microscopy (AFM). In AFM, samples are deposited on a flat mica substrate and their structure is imaged by a probe tip that is attached to a force-sensing cantilever. AFM allows direct visualization of individual biological macromolecules, making it ideal for studying the tertiary structure of large, irregular multicomponent biomolecules such as chromatin (1, 17, 30, 36). Standard silicon tips have variable apex diameters, which can change during use, and the resulting variability in resolution can complicate analysis of novel structures. Here we used carbon nanotube AFM tips, which are geometrically well defined, robust, and small in diameter (4, 9), to characterize hSWI/SNF and remodeled products. Our results indicate that hSWI/SNF can form dimers of mononucleosomes with weakened histone-DNA interactions and that it can dramatically alter the positions and stability of nucleosomes on polynucleosomal arrays.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Nucleosome and hSWI/SNF isolation. Mononucleosomes were isolated from HeLa cells by micrococcal nuclease digestion and glycerol gradient centrifugation (31) (gradient buffer [GGB] contains 20 mM HEPES [pH 7.9], 1 mM EDTA, 180 mM KCl, 0.1% NP-40, and 10 or 30% glycerol), followed by dialysis in TE (10 mM Tris [pH 7.5], 1 mM EDTA). These samples are >90% pure by sodium dodecyl sulfate-polyacrylamide gel electrophoresis and gel shift analysis (data not shown). Concentrations are given as the weight of DNA in the nucleosomes. hSWI/SNF was affinity purified from HeLa cells by virtue of a FLAG tag on its Ini1 subunit (33) and was >50% homogeneous, as estimated by silver staining. For imaging, hSWI/SNF was further purified by glycerol gradient centrifugation as for nucleosomes (see above), except that the gradient buffer contained 50 mM Tris (pH 7.5), 1 mM EDTA, bovine serum albumin (BSA) at 100 µg/ml, and 180 mM KCl with 22 or 30% glycerol.

Assembly of chromatin arrays. A nonradioactive or ³²P-end-labeled (Klenow fill-in) nucleosomal array, 5S-G5E4 (27), was formed by salt dialysis with HeLa core histones and dialyzed into TE as described previously (24) with the modifications noted (29). Assembly was verified by electrophoresis on a 1% Tris-acetate-EDTA gel and/or by EcoRI digestion of the template, which cuts between 208-bp 5S DNA sequences. Eighty to 90% of the 208-bp EcoRI fragments were nucleosomal, corresponding to an average of 9 to 11 nucleosomes per array. However, this is likely to be an underestimate of the nucleosome number since nucleosomes covering EcoRI sites prevent cutting and cannot be counted.

Mononucleosome remodeling reactions and separation of products. Remodeling reaction mixtures (200 µl) contained 1 µg (~1.3 nM) of hSWI/SNF fraction and 2 µg (~100 nM) of mononucleosomes in 34 mM KCl-20 mM HEPES (pH 7.9)-0.1 mM phenylmethylsulfonyl fluoride-0.5 mM dithiothreitol-0.1% NP-40-0.05 mM EDTA-2.9 mM MgCl₂ and (where indicated) 2 mM ATP/MgCl₂. After 2 h at 30°C, the KCl concentration was increased to 233 mM and the reaction products were separated by glycerol gradient centrifugation (10 to 30% GGB). Reactions yielded ~20 to 30% altered dimers and ~70 to 80% mononucleosomes, as measured by gel shift of input reaction mixtures and gradient fractions, followed by ethidium bromide fluorescence staining. For exonuclease III (ExoIII) analysis, dimers and mononucleosomes were labeled by T4 polynucleotide kinase (10 U) and [-³²P]ATP for 30 min at 37°C in 0.5× GGB with BSA at 100 µg/ml and 7 mM MgCl₂. Labeled products were then purified on a 5 to 30% glycerol gradient containing 50 mM Tris (pH 7.5), 1 mM EDTA, and BSA at 100 µg/ml. Bare DNA was prepared from labeled mononucleosomes by phenol extraction and ethanol precipitation. Peak fractions were adjusted to 60 mM KCl-0.1% NP-40-20 mM HEPES (pH 7.9)-5.6 mM MgCl₂ and digested with 6 U of ExoIII for 3 or 15 min before stopping with EDTA and purification of the DNA as previously described for DNase digestions (31).

Remodeling of 5S array templates. For analytical restriction assays (see Fig. 3A), 3.4 ng of labeled arrays was incubated at 30°C for 60 min in 25-µl standard hSWI/SNF reaction mixtures (with 4 mM MgCl₂, 60 mM KCl, 0.1% NP-40, and other conditions as described previously [31]). We added 200 ng of hSWI/SNF, 0.5 mM ATP/MgCl₂, and/or 1 U of apyrase where indicated. We added 20 U of SacI or XbaI and stopped samples at 10 and 50 min with sodium dodecyl sulfate stop buffer and proteinase K (31) before separating them by 1.5% agarose-Tris-borate-EDTA electrophoresis. The dried gel was quantitated with a Molecular Dynamics PhosphorImager. For AFM analysis, 200 ng (~2.5 nM) of nonradioactive arrays was remodeled by 250 ng of hSWI/SNF (2.5 nM) as described above but with 3 mM MgCl₂ and 2 mM ATP/MgCl₂ (where indicated) for 60 or 90 min. The reaction was stopped by addition of EDTA to 5 mM and dialyzed into TE at 4°C. To assay for stable changes (1 week later; see Fig. 3B), 10 µl of each dialyzed reaction mixture was adjusted to 50 mM KCl and 5 mM MgCl₂ and digested with 20 U of SacI for 20 min before electrophoresis as described above. Ethidium bromide signals from cut and uncut bands were quantitated on a digital camera adjusted to the linear range.

Preparation of samples for AFM, imaging, and analysis. hSWI/SNF, mononucleosomes, and products were fixed with 0.25% glutaraldehyde at 4°C for 6 h and dialyzed into TE at 4°C overnight with one change of buffer. Remodeled array reaction mixtures were dialyzed into TE and (where indicated) fixed for 1 to 2 h on ice by four- to sixfold dilution into 0.16% glutaraldehyde. A freshly cleaved mica surface was treated for 1 min with 1 mM spermidine or a solution of 0.1% poly-L-lysine, washed with several milliliters of water, and dried under N₂. Samples were deposited for 2 min, rinsed with water, and dried under N₂. The samples were imaged with a Multimode Nanoscope IIIa (Digital Instruments, Santa Barbara, Calif.) in tapping mode in air using carbon nanotube tips (4, 9) and/or silicon tips (for some polynucleosome images where fine-scale resolution of features was not critical) with scan sizes of 0.5 to 2 µm and scan rates of 1 to 2 Hz at a resolution of 512 by 512 pixels. Apparent full widths are overestimates due to the width of the tips, and heights tend to be underestimates due to deformation of the sample (3). Due to these considerations, we report average heights and widths (from measurements of >20 molecules) only as approximate values and compare values only for samples imaged with the same tip. Internucleosomal distances were measured from the positions of the nucleosome centers along the DNA. In all cases, we saw similar results with at least two spreads of each of two independent preparations of nucleosomal samples.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Structure of hSWI/SNF-remodeled mononucleosomes. AFM images of mononucleosomes and hSWI/SNF were recorded to provide well-defined points of comparison for the analysis of remodeled products. HeLa mononucleosomes appeared as roughly spherical structures with an average diameter of ~11 nm (measured as full width at half-maximal height) and an average height of ~4 nm (Fig. 1A). These dimensions are consistent with the crystal structure of similar intact mononucleosomes (23). Approximately half of the nucleosomes had a small tail of proper dimensions to be bare DNA extending <10 nm from the nucleosome. Note that the full widths (at zero height) of nucleosome images are often much greater than 11 nm due to broadening by the finite size of the tip (3). Our recent advances in nanotube tip fabrication have provided individual single-walled nanotube tips (8, 9) that image the mononucleosomes with a full width of only ~16.5 nm, thus demonstrating much higher resolution than the ~23-nm full widths previously reported with standard silicon tips (25). Images of gradient-purified, nearly homogeneous hSWI/SNF reveal that it is a large structure (~25-nm full width at half- maximal height by ~6-nm height) with at least four structural lobes. These measurements are in good agreement with the volume expected for a cylindrical protein with a 2-MDa molecular mass and a 1.3-g/ml density (2,900 nm³ compared to the 2,600 nm³ predicted; Fig. 1B), and the width is similar to that seen in transmission electron microscopy pictures of the yeast SWI/SNF complex (2).

View larger version (62K): [in this window] [in a new window]   FIG. 1.   AFM images of SWI/SNF and altered dimers. Samples were fixed, deposited, and imaged with a nanotube tip as described in Materials and Methods. (A) Mononucleosomes on spermidine-treated mica. DNA tails, where visible, are indicated by arrows. (B) Gradient-purified hSWI/SNF on spermidine-treated mica. Multiple lobes are indicated by arrows. Small molecules are BSA from the gradient buffer. (C) hSWI/SNF-remodeled dimers on poly-L-lysine-treated mica. DNA tails are indicated by arrows.

View larger version (71K): [in this window] [in a new window]   FIG. 2.   ExoIII digestion of hSWI/SNF products. 5'-end-labeled remodeled dimers (lanes 7 to 9) and control mononucleosomes (lanes 4 to 6) or bare DNA from mononucleosomes (lanes 1 to 3) were digested with ExoIII for 3 (lanes 2, 5, and 8) or 15 (lanes 3, 6, and 9) min before DNA purification and resolution by denaturing polyacrylamide gel electrophoresis. The values on the left are molecular sizes in base pairs.

5S Arrays are stably remodeled by hSWI/SNF. At its most basic level, cellular chromatin consists of arrays of nucleosomes on the DNA fiber. By comparison to mononucleosomes, little is known about the stable products of SWI/SNF on polynucleosomal arrays. To study this, we used salt dialysis to assemble nucleosomes onto the 5S-G5E4 DNA template, which contains five 5S-rRNA gene (rDNA) sequences (which each tend to position a single nucleosome in a preferred location) flanking each side of a transcriptional reporter DNA sequence that accommodates two nucleosomes (diagram in Fig. 3A) (27). The transcriptional reporter sequence contains a unique SacI site that is generally not covered by a nucleosome. Under control conditions, this can be seen as rapid digestion (in the first 10 min) of ~50% of the templates, in which nucleosome positions have left the SacI site bare, followed by much slower digestion of the templates in which the SacI site is covered by a nucleosome (Fig. 3A, left panel, triangles). By contrast, the XbaI site is generally nucleosomal, resulting in less than 20% cleavage of the templates in the first 10 min (Fig. 3A, right panel, triangles).

View larger version (50K): [in this window] [in a new window]   FIG. 3.   5S nucleosomal arrays are stably remodeled by hSWI/SNF. (A) Diagram of the 5S-G5E4 rDNA array used. White circles represent nucleosomes at preferred sites on 208-bp 5S rDNA sequences. Grey circles represent nucleosomes in the transcription template. Locations of SacI and XbaI sites relative to inferred unremodeled nucleosome positions are indicated. For the graphs, the array was treated with hSWI/SNF without ATP (triangles), with ATP for 30 min (squares), or with ATP for 30 min, followed by apyrase for 18 min (circles), and then cut with SacI (left) or XbaI (right) for the indicated times. The purified DNA was separated by agarose electrophoresis, and percent cutting was quantified. Similar results were obtained in three separate experiments. (B) Unlabeled arrays were incubated with both SWI/SNF (S/S) and ATP (lane 4), without ATP (lane 5), or without SWI/SNF (lane 6) for 90 min before dialysis into TE. Samples of these reaction mixtures and bare DNA (lane 2) or the untreated assembled array (lane 3) were digested with SacI, purified, and separated as described above, followed by ethidium bromide staining and quantitation. Lane 1 contained uncut DNA.

AFM of SWI/SNF-remodeled 5S arrays. AFM images of fixed, control polynucleosomal arrays (lacking either ATP or SWI/SNF) clearly show the nucleosomes evenly spaced over the DNA (Fig. 4A and data not shown). Analysis of the positions of all of the nucleosomes on 20 arrays yielded an average of 12.5 ± 0.3 nucleosomes per array with an average center-to-center distance of 33 ± 2 nm. The theoretical distance between two nucleosomes at the preferred DNA position on the 208-bp 5S rDNA nucleosome-positioning sequence, assuming 146 bp bound by a 10-nm nucleosome particle and 62 bp (~21 nm) of linker DNA, is 31 nm, in good agreement with our analysis. High-definition mapping studies have shown that nucleosomes occupy their preferred position on 5S rDNA sequences only ~60% of the time (6, 26), which would result in only ~36% of internucleosomal distances matching the theoretical value. This is consistent with the high standard deviation (±22 nm, as distinct from the standard error of the mean noted above) for center-to-center distances observed in our direct imaging experiments.

View larger version (77K): [in this window] [in a new window]   FIG. 4.   AFM images of fixed, hSWI/SNF (S/S)-remodeled arrays on spermidine-treated mica. (A) An array from the control reaction mixture lacking ATP (Fig. 3B, lane 5). (B) Arrays remodeled by hSWI/SNF and ATP for 90 min (Fig. 3B, lane 4). (C) Histogram of internucleosomal distances (peak to peak along the DNA contour) from two experiments. The data (23 arrays with ATP, 19 arrays without ATP) were grouped into three bins (14 nm, 14 to 60 nm, and >60 nm) and normalized to 100%. (D) An hSWI/SNF-bound array from the same control reaction mixture as in panel A. (E) An hSWI/SNF-bound remodeled array from the same reaction mixture as in panel B. The arrow shows potential DNA loops constrained by SWI/SNF. The height range in panels D and E is greater than in that in panels A and B to allow structural features of SWI/SNF to be evident.

Images of unfixed remodeled arrays reveal instability of remodeled nucleosomes. When arrays of nucleosomes are deposited on charged surfaces and imaged without fixation, they had properties similar to those of fixed arrays, with only a slight reduction in the nucleosome count (e.g., see reference 36). Consistent with those studies, we found that unfixed control arrays treated with SWI/SNF but not ATP looked similar to the fixed arrays (compare Fig. 5A to Fig. 4A) but had a significantly reduced nucleosome count of 10.7 ± 0.5 (n = 10). The spacing of nucleosome pairs on unfixed control arrays was similar to that on fixed control arrays (4% ± 2% 14 nm, 76% ± 9% 14 to 60 nm), except that the loss of nucleosomes from the unfixed arrays resulted in an increased frequency of greatly separated pairs (20% ± 4% >60 nm), with a corresponding increase in overall length (517 ± 14 nm). Surprisingly, unfixed remodeled arrays looked strikingly different from both fixed remodeled arrays and unfixed controls (compare Fig. 5B with Fig. 4B and 5A). Clear nucleosome particles were rare, and the DNA was frequently looped or kinked, making it nearly impossible to follow its path. Such a tangle might be expected if the DNA had been pulled from the histones onto the surface. This result indicates that, in addition to altered positions, the stability of the nucleosomes in hSWI/SNF-remodeled arrays is dramatically reduced.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

In this study, AFM imaging with high-resolution carbon nanotube probes revealed the product formed by SWI/SNF from mononucleosomes to be two closely joined particles equal in size to the input nucleosomes (Fig. 1), which is consistent with a previous biochemical analysis (31). The observation of long DNA tails on dimers (Fig. 1C), combined with the results of exonuclease III digestion (Fig. 2), suggests that 60 to 70 bp of DNA is more weakly associated with histones in dimers than in normal nucleosomes. This fraction of weakened histone-DNA contacts may explain why remodeled dimers are less resistant to disruption by salt than are normal nucleosomes (20; G.R.S., unpublished data). It has been argued that SWI/SNF-like complexes may release 60 to 80 bp of DNA from the surface of the nucleosome (22). Our results suggest, instead, that this length of DNA is associated with histones in an alternative, weaker conformation.

The hSWI/SNF complex (Fig. 1B) is seen as a multilobed structure with apparent twofold symmetry and often with a distinct saddle shape. Transmission electron microscopy pictures of the yeast SWI/SNF complex bound to polynucleosomes, while revealing less about the surface structure of the complex, did show two DNA-nucleosome binding sites per molecule, suggestive of a similar symmetry (2). The details visible in the images presented here suggest that carbon nanotube AFM might be an excellent method for studying the placement and function of subunits and conformational changes in the complex during substrate binding and catalysis.

SWI/SNF dramatically alters the positions of nucleosomes in polynucleosomes compared to those of control arrays (Fig. 4). The frequency of internucleosomal distances of greater than 60 nm are more than doubled, as is that of distances of less than 14 nm. At the same time, the frequency of internucleosomal distances around the 30-nm range, established by the strong 5S rDNA positioning sequences, decreases significantly. These observations provide an explanation for the stable changes in restriction enzyme access introduced by hSWI/SNF or yeast SWI/SNF into arrays of nucleosomes (12) (Fig. 3). These data do not tell us whether SWI/SNF only moves nucleosomes until further movement is blocked by adjacent nucleosomes or whether it can allow histone octamers to be transferred in cis past other nucleosomes or even in trans to other DNAs, as suggested by other studies (21, 29, 34). We did not observe a significantly broader distribution of the number of nucleosomes per DNA molecule, however, which might be expected if a large number of nucleosomes were removed from some DNAs and deposited on others in trans.

Despite alterations in nucleosome positions, the overall length of the arrays does not change, suggesting that the length of DNA associated with each remodeled nucleosome remains the same. This argues against the hypothesis that 60 to 80 bp of DNA is unwound from the ends of remodeled nucleosomes in arrays (22), since this would be predicted to increase array length by over 200 nm. Transmission electron microscopy studies of polynucleosomes remodeled by yeast SWI/SNF indicated an average loss of ~40 bp of DNA from each nucleosome within loops of chromatin physically constrained by SWI/SNF but not from those outside these loops (2). In this study, nucleosomes in arrays not bound by SWI/SNF also appear to have a normal DNA content. We cannot address the nature of nucleosomes in hSWI/SNF-constrained loops, since the potential DNA loops we observed (Fig. 4D and E) were too small.

The increase in closely abutting nucleosomes (14 or fewer nm apart) after hSWI/SNF action could result from either (i) nucleosomes simply being moved close together or (ii) the creation of pairs of altered nucleosomes similar to altered mononucleosome dimers. The analysis of remodeled dimers (Fig. 1C) showed that the center-to-center distance between the two nucleosome lobes was 14 nm or less (2 standard deviations above the average). This analysis, however, did not allow positive identification of similar products on arrays, since each lobe was indistinguishable from normal mononucleosomes in shape and dimensions. Note that, since each measurement on the array represents a pair of nucleosomes, a 10% increase in internucleosome distances of 14 nm or less corresponds to an ~20% increase in nucleosomes with a neighboring nucleosome (on either side) close enough to be part of an altered pair. This fits well with the distribution of SWI/SNF products formed from mononucleosomes, which is generally ~75% mononucleosomes and ~25% dimers at apparent equilibrium. Theoretically, dimers need not be formed between adjacent nucleosomes but might also form between a distant nucleosome pair in cis (creating loops) or nucleosomes on two arrays (creating linked arrays). We did not observe a significant increase in these types of structures with remodeled arrays, suggesting that, if dimers are formed, such events might be relatively rare or unstable.

The fact that unfixed remodeled arrays are much less stable than control arrays under our deposition conditions indicates that the nucleosomes on the array have been qualitatively altered and not just repositioned (Fig. 5). Dimers formed from mononucleosomes have been shown to have reduced resistance to dissociation by high salt concentrations (20; G.R.S., unpublished data) and appear to have weaker histone interactions with ~60 bp of associated DNA (Fig. 1C). Thus, an accumulation of dimers on the array might explain the reduced stability of remodeled arrays. It has also been proposed that ATP-dependent chromatin remodeling might involve the formation of distorted single nucleosomes that constrain loops or bubbles of DNA unbound by histones (10). Such changes might also reduce the resistance of the nucleosomes to deposition conditions. They might also lead naturally to the formation of dimers or higher-order multimers that might stabilize distorted nucleosomes against reversion to the normal structure.

View larger version (68K): [in this window] [in a new window]   FIG. 5.   AFM images of unfixed remodeled arrays suggest reduced nucleosome stability. (A) Control arrays (shown fixed in Fig. 4A) were deposited on spermidine-treated mica without fixation. (B) Remodeled arrays (shown fixed in Fig. 4B) were deposited on spermidine-treated mica without fixation. The small molecules in the background are BSA.

hSWI/SNF is involved in transcriptional activation through steroid receptors, heat shock factor, and globin gene regulators, as well as transcriptional repression through the retinoblastoma protein Rb (for reviews, see references 13 and 35). Our observations, summarized in the model proposed in Fig. 6, indicate that hSWI/SNF can accomplish a dramatic restructuring of arrays of nucleosomes to generate both long stretches of bare DNA and clumps of nucleosomes even on DNA harboring strong nucleosome positioning sequences. Thus, the complex may have great power to disrupt and reshuffle nucleosome organization over promoters and transcribed regions in vivo. If the original organization was repressive, this could result in transcriptional activation, either by moving nucleosomes away (e.g., Fig. 6, site B) or by creating distorted mononucleosomes or altered dimers (e.g., Fig. 6, site C), which can be more accessible than normal nucleosomes to transcription and recombination factors (15, 31). If the original organization was active, however, SWI/SNF could result in repression, as nucleosomes are moved over or near transcription factor binding sites (e.g., Fig. 6, site A). Note that in the continued presence of SWI/SNF, nucleosome positions and conformations would be fluid. These changes might be fixed either by the removal of SWI/SNF or by the binding of factors that act as boundaries to nucleosome movement or stabilize one form of the nucleosome (altered or normal) over the other.

View larger version (11K): [in this window] [in a new window]   FIG. 6.   Model for SWI/SNF remodeling of arrays. Nucleosomes in initial positions (top) block some transcription factor binding sites (B and C) while leaving others open. SWI/SNF alters nucleosome positions (bottom) to uncover some sites (B) and cover others (A). In this way, SWI/SNF remodeling can facilitate both activation and repression. Nucleosome dimers are also formed, which are more accessible to some factors (site C).

Clearly, many questions remain unanswered. How do our present observations relate to the effects of hSWI/SNF on chromatin in vivo, which, for instance, exists with linker histones in a more highly compacted form? Do the changes introduced by hSWI/SNF (dimer formation, repositioning, or both) revert back to normal, and at what rate? How might the complex be regulated, for instance, to promote an active conformation at one promoter and a repressive conformation at another? Members of the ISWI-based family of remodeling complexes aid the regular spacing of nucleosomes. Do these complexes work to counteract complexes like hSWI/SNF that appear adept at disorganization? In seeking answers to all of these questions, and others, we see great potential in a combination of biochemical and molecular imaging techniques.