|Home | About | Journals | Submit | Contact Us | Français|
The ATP-dependent chromatin remodeling complex SWI/SNF regulates transcription and has been implicated in promoter nucleosome eviction. Efficient nucleosome disassembly by SWI/SNF alone in biochemical assays has however not been directly observed. Employing a model system of dinucleosomes rather than mononucleosomes, we demonstrate that remodeling leads to ordered and efficient disassembly of one of the two nucleosomes. An H2A/H2B dimer is first rapidly displaced and then in a slower reaction an entire histone octamer is lost. Nucleosome disassembly by SWI/SNF did not require additional factors such as chaperones or acceptors of histones. Observations in single molecules as well as bulk measurement suggest that a key intermediate in this process is one in which a nucleosome is moved towards the adjacent nucleosome. SWI/SNF recruited by the transcriptional activator Gal4-VP16 preferentially mobilizes the proximal nucleosome and destabilizes the adjacent nucleosome.
ATP-dependent chromatin remodelers are known to be involved in reorganization of chromatin and to move nucleosomes along DNA. Activation of transcription often requires remodeling of chromatin in promoter regions for the transcription machinery to gain access to the DNA template (Boeger et al., 2003; Boeger et al., 2004; Korber et al., 2006; Reinke and Hörz, 2003). SWI/SNF was the first of these remodeling complexes to be discovered. It is required for transcription of a series of different genes and is well conserved from yeast to humans. Single molecule and bulk experiments have found that SWI/SNF and its paralog RSC can translocate nucleosomes along DNA (Lia et al., 2006; Shundrovsky et al., 2006; Zhang et al., 2006). High-resolution mapping has shown that SWI/SNF moves nucleosomes towards the ends of DNA; approximately 1/3 of the histone octamer surface becomes exposed and 2/3 retains contacts with ~100 bp DNA (Kassabov et al., 2003; Zofall et al., 2006). SWI/SNF remodeling does not require the loss of histones from nucleosomes as shown by SWI/SNF efficiently remodeling nucleosomes containing histones that had been covalently cross-linked to each other (Kassabov et al., 2003). Based on these and other data it has been thought that SWI/SNF makes nucleosomal DNA more accessible by simply moving nucleosomes or by generating transient DNA bulges on their surface without being disassembled (Fan et al., 2003; Narlikar et al., 2001).
Other studies found that RSC and human SWI/SNF transfer histone octamer from nucleosome to free DNA at a reduced efficiency, but this required an excess of donor nucleosomes to free DNA and long reaction times (Lorch et al., 1999; Phelan et al., 2000). SWI/SNF was also found to promote the exchange of H2A/H2B heterodimers from nucleosomes to acceptor tetrasomes (H3/H4 tetramer plus DNA), but this process was as inefficient as observed for octamer transfer (Bruno et al., 2003). A series of elegant experiments showed that the SWI/SNF-dependent remodeling and gene activation observed in vivo at the PHO5 promoter region was not caused by merely moving nucleosomes away from the promoter, but was due to the disassembly of nucleosomes (Boeger et al., 2003; Boeger et al., 2004). Some of the strongest evidence for disassembly comes from experiments where the excision of minicircles containing the PHO5 promoter was used to demonstrate that SWI/SNF mediates nucleosome loss (Boeger et al., 2003; Dhasarathy and Kladde, 2005; Reinke and Hörz, 2003). It has remained unclear how the SWI/SNF-induced nucleosome eviction observed in vivo at promoters can be reconciled with the activities of SWI/SNF observed in vitro.
It is possible that SWI/SNF alone does not have the ability to disassemble nucleosomes without an additional factor. Asf1 is a histone chaperone that interacts with histone H3 and H4 that was shown to promote the disassembly of nucleosomes in vivo at the PHO5 promoter (Adkins et al., 2004; Adkins et al., 2007; Korber et al., 2006). Asf1 could disassemble nucleosomes in conjunction with SWI/SNF (Gkikopoulos et al., 2009; Takahata et al., 2009). Another histone chaperone, Nap1, acting with RSC, was shown to promote the complete disassembly of the nucleosome, but the experimental conditions required and the fact that Nap1 interacts primarily with the histone H2A/H2B dimer cast doubt on the physiological relevance of this observation (Andrews et al., 2008; Lorch et al., 2006; Park and Luger, 2008).
Another reason that SWI/SNF alone may not have been found to disassemble nucleosomes in vitro is that the mononucleosomes used in earlier studies may not be an appropriate nucleosomal substrate. Although nucleosomal arrays consisting of 5S rDNA or other DNA repeats have been generally useful, the level of information obtained has been limited to either changes in accessibility at particular restriction endonuclease cut site(s) or the size of DNA protected by the nucleosome as determined by micrococcal nuclease digestion (Bazett-Jones et al., 1999; Côté et al., 1998; Logie and Peterson, 1997; Ulyanova and Schnitzler, 2005). We have addressed this problem by constructing more manageable short nucleosomal arrays of di- and trinucleosomes that contain two or three copies of strong nucleosome positioning sequences (NPS) (Lowary and Widom, 1998). The dinucleosome substrates used in this report allowed better monitoring of changes in histone content, translational position, and orientation of DNA relative to histone octamers than has been possible with larger arrays. A key finding of these studies is that the ability of SWI/SNF to displace H2A/H2B dimers and histone octamers from DNA has been significantly underestimated in earlier studies. Our data with recruitment of SWI/SNF by a transcription activator made it possible to discern the structural changes that occur as one nucleosome is moved by SWI/SNF towards an adjacent nucleosome. We find that one of the two nucleosomes in this case is preferentially mobilized by SWI/SNF.
Mono-, di- and trinucleosomes were assembled with high affinity NPS (Figure 1A), remodeled with varying amounts of SWI/SNF and ATP, then analyzed by 5% native polyacrylamide gel electrophoresis (PAGE) (Figure 1B). When mononucleosomes were remodeled by SWI/SNF a typical single faster-migrating species was observed (Figure 1B, compare lane 1 and lanes 2-6) typical of more compact unraveled nucleosomes (Kassabov et al., 2003). In contrast, remodeled di- and trinucleosomes migrated as multiple discrete faster-migrating products (Figure 1B, labeled #1-3; compare lane 7 to lanes 8-12 and lane 13 to lanes 14-18). The multiple remodeled species with di- and trinucleosomes species likely reflect multiple translational positions of the nucleosomes and possibly even histone loss. The remodeled trinucleosome had no discrete bands, but the remodeled species tended to localize as three faster-migrating populations (indicated by bars).
The length of linker DNA can vary extensively between adjacent nucleosomes and potentially affect the ability of SWI/SNF to gain access and mobilize nucleosomes, as previously observed for ISW2 and ISW1a (Zofall et al., 2004). Translational positions were found by mapping the contact of residue 53 of histone H2B with DNA by site-directed cross-linking and cleavage. Nucleosomes were positioned using the 601 NPS and the length of extranucleosomal DNA varied from 0 to 105 bp. SWI/SNF does not have any requirement for extranucleosomal DNA unlike ISWI-type ATP-dependent chromatin remodeling complexes (Kagalwala et al., 2004; Zofall et al., 2004). Instead SWI/SNF can move a core nucleosome with no extranucleosomal DNA until the DNA ends are bound ~20 bp from the pseudodyad axis (see Figure S1). These mapping studies show that nucleosomes without and with extranucleosomal DNA of 20 to 105 bp can be mobilized with similar efficiencies.
The effect of linker DNA length on SWI/SNF remodeling of dinucleosomes was examined using dinucleosomes with 6, 30, and 79 bp of linker DNA between the two nucleosomes. SWI/SNF was unable to change the translation position of dinucleosomes with 6 bp linker DNA (30N6N9) as efficiently as dinucleosomes with 30 (30N30N9) and 79 bp (30N79N9) linker DNA as shown by gel shift assay (Figure 2A, compare lanes 2-4 with lanes 6-8 and 10-12). Site-directed mapping of nucleosome position confirmed that nucleosomes separated by 6 bp of linker DNA were not shifted. By contrast, new nucleosome positions were observed after remodeling by SWI/SNF when there was 30 and 79 bp of linker DNA (data not shown). The inability of SWI/SNF to mobilize dinucleosomes with 6 bp of linker DNA is not because a minimal length of linker DNA is required (Figure S1), but most likely a close apposition of the two nucleosomes sterically impairs SWISNF binding (Dechassa et al., 2008). Binding titrations of SWI/SNF with the three dinucleosomes showed a strong preference for SWI/SNF binding to 30N30N9 and 30N79N9 dinucleosomes over 30N6N9 dinucleosomes (Figure 2B, compare lanes 1-4 with lanes 5-8 and 9-12). Only 9% of the 30N6N9 dinucleosomes with 6 bp of linker were bound at a SWI/SNF concentration sufficient to saturate the dinucleosomes with 30 and 79 bp of linker DNA (lanes 3, 7, and 11).
Potential histone loss from dinucleosomes was measured in the different complexes using fluorescent histones. Histones were tagged at cysteine 19 of histone H2A (H2A C19) or cysteine 80 of histone H3 (H3 C80) with a fluorescent label and DNA was radiolabeled at one 5′ end. The amount of labeled histone in the different dinucleosome species was found by determining the ratio of the fluorescent to radioisotope signal in each band after normalizing both signals (Figure 3). These experiments were done with no DNA present other than the PCR-generated DNA. Under these conditions free DNA is not available to promote histone exchange since the efficiency of nucleosome assembly is ≥95%. The gel shift pattern was similar to that observed in Figure 1 in which carrier DNA was present (Figure 3A). The remodeled dinucleosomes that migrate slightly faster than unremodeled nucleosomes (remodel #1) had lost ~22% of histone H2A or about 1/4 of their total histone H2A content (Figure 3B). The faster-migrating species (remodel #2) had approximately 44% reduction in H2A content which corresponds to displacement of 2 of 4 H2A/H2B dimers. No loss of histone H2A with SWI/SNF remodeling was observed with mononucleosomes under the same conditions (compare Figure 3B to 3D). Changes in the relative amount of histone H3 in the remodeled nucleosomes were noticeably different than that for H2A. The species migrating closest to the original dinucleosome (remodel #1) had nearly complete retention of H3 (94%) (Figure 3C). The faster-migrating remodeled dinucleosome products (remodel #2) however had ~50% depletion of histone H3 consistent with the loss of one H3/H4 tetramer from the original dinucleosome substrate. No H3 was lost when mononucleosomes were remodeled (Figure 3E). These data suggest that one reason for the significant change in electrophoretic mobility of dinucleosomes when remodeled by SWI/SNF is in part due to the loss of histones, a property that is not observed in remodeling of mononucleosomes. The change in mobility correlates well with the extent of histone loss. Remodeled species migrating most like the original dinucleosome seems to have only lost one H2A/H2B dimer, while the fastest-migrating species have lost a complete histone octamer. The loss of one H2A/H2B dimer or an entire histone octamer did not depend on the presence of free DNA as an acceptor and sets this observed activity apart from the less efficient reactions previously observed for dimer or octamer exchange (Bruno et al., 2003; Lorch et al., 1999; Phelan et al., 2000).
A potential order in SWI/SNF remodeling could be first displacing one H2A/H2B dimer followed by a complete loss of one octamer from the DNA. To examine this hypothesis SWI/SNF was prebound to 40N39N6 dinucleosome substrates at 30°C with no carrier DNA and the temperature lowered to 25°C before starting remodeling by the addition of 55 μM ATP. The remodeled dinucleosome product that appeared first was the one in which only one H2A/H2B dimer has been displaced (Figure 4A, compare lane 1 to lanes 2-6 and had an initial rate of 15.6 nM min-1 (Figure 4A, lanes 5-8 and Figure 4B). The faster-migrating remodeled species appeared 30 times slower at a rate of 0.45 nM min-1 demonstrating that SWI/SNF first displaces one histone H2A/H2B dimer followed by a slower step in which the rest of the octamer is displaced from DNA.
Nucleosome movement in dinucleosomes was tracked using the site-directed approach described earlier by mapping the contacts of histone H3 (residue 120) and H2B (residue 53) similar to that described for mononucleosomes in Figure S1 (Kassabov et al., 2002; Kassabov et al., 2003; Kassabov and Bartholomew, 2004). Residue 120 of histone H3 cross-links to DNA -2 and +6 bp from the pseudodyad axis of the nucleosome with a modest preference for the -2 bp site. Residue 53 of histone H2B preferentially contacts only one DNA strand which causes it to have one cleavage site for each nucleosome position. The two H2A/H2B dimers per nucleosome each cleave a different strand making it easier to assign contacts to a specific H2B molecule after cross-linking and strand cleavage. Quantification of site-directed mapping of H3 to the lower and upper DNA strands showed that the region of the N1 nucleosome was mostly devoid of nucleosomes after remodeling. Nucleosomes were instead clustered in several positions around the N2 NPS (Figures 5A and S2A-B). These data show that nucleosomes are primarily moved to one end of DNA and that one of the two nucleosomes is likely evicted during the process consistent with data in Figure 3.
SWI/SNF is known to be recruited to particular genomic sites through interactions with transcription factors bound at cognate promoter regions. Therefore the effect of SWI/SNF recruitment by Gal4-VP16 to one end of a dinucleosome was examined. These conditions reflect those in vivo in which SWI/SNF has to discriminate between numerous genomic sites and is actively recruited to specific target sites. This approach also allows changes in dinucleosome structure to be examined more thoroughly because SWI/SNF action is preferentially oriented in one direction. SWI/SNF binding and remodeling was Gal4-VP16 dependent using competitor DNA and nucleosome movement was tracked as before using H3- and H2B-modified nucleosomes (Figures 5B and S2B and data not shown). The dependence of SWI/SNF remodeling on Gal4-VP16 under these conditions is also shown below using single molecule mapping techniques. The dinucleosome was constructed with the 601b NPS, a slightly modified version of the original 601 NPS with three newly inserted M.SssI sites to increase probing resolution, and a second completely different NPS (603) (Figures 1A, ,5B,5B, and S5).
After remodeling, a proportion of nucleosomes at both original positions had been eliminated and new nucleosome positions appeared in a region near the central linker DNA (Figures 5B and S2C). Cleavage at both original translational positions was reduced, and new H3 and H2B cleavage sites were observed at bp 221 to 240 and 273 to 295, respectively for H3 and H2B (Figures 5B and S2C). Based on the H3 contact sites (-2 and +6 from the pseudodyad), the new nucleosome positions were with the pseudodyad axis ranging from bp 219 to 238. This new nucleosome position indicates that a centrally positioned nucleosome has encroached on either the original 601b or 603 nucleosomes as schematically represented (Figure 5B). As these mapped positions are a population average, it is also possible that there are two different subpopulations of remodeled dinucleosomes; one with a centrally mapped nucleosome and another with either the original 601b or 603 nucleosome. The nucleosome mobilized to the center has also retained its H2A/H2B dimer proximal to DNA from bp 273 to 295 (Figure S2C). The overlap of the central nucleosome and either of the original nucleosomes is 39-58 bp based on this mapping data. The extent to which these nucleosomes overlap would readily explain the displacement of one H2A/H2B dimer in this remodeling reaction. In contrast to SWI/SNF remodeling alone, recruitment of the remodeler by Gal4-VP16 reduces the movement of nucleosomes towards the DNA ends and instead preferentially moves nucleosome towards the DNA center (Figure 5).
The idea of overlapping nucleosomes does not seem to be consistent with data in Figure 2 showing that nucleosomes positioned too close together are not effectively bound or remodeled by SWI/SNF. The ability of SWI/SNF however to unwrap up to ~50 bp of DNA from the core nucleosome could provide a sufficient tether in terms of DNA length to bring two nucleosomes in that could appear as an overlapping dinucleosome-like structure. The suggestion is that the two nucleosomes do not directly contact each other, but instead the distance in terms of linear DNA position is reduced by partially unwrapping and moving one of the two nucleosomes along DNA.
Site-directed mapping of nucleosome position and other assays are only able to determine the average distribution of nucleosome positions. A single molecule approach referred to as MAPit (Methyltransferase Accessibility Protocol for Individual Templates) overcomes this limitation by probing with a DNA methyltransferase (Bouazoune et al., 2009; Fatemi et al., 2005; Frommer et al., 1992; Jessen et al., 2006; Kilgore et al., 2007; Pondugula and Kladde, 2008). DNA within nucleosomes is protected from DNA methylation and provides a clear footprint of the nucleosome (Kladde et al., 1996). DNA methylation protection unlike other DNA footprinting techniques can be done at the single molecule level by converting all unmethylated deoxycytidine residues to deoxyuridine with bisulfite and then sequencing clones of individual PCR-amplified DNA molecules.
SWI/SNF remodeling of the 601b-603 dinucleosome was first done without recruitment and control samples with no SWI/SNF or no ATP added were performed in parallel. In the absence of carrier DNA, dinucleosomes alone or incubated with SWI/SNF but without ATP showed two regions of strong protection against methylation by M.SssI DNA methyltransferase, corresponding to the 601b and 603 NPS, whereas flanking and linker DNA were nearly completely methylated (Figure S3B, panels b and c). Protection was due to formation of positioned 601b and 603 nucleosomes as free DNA was almost completely methylated under identical conditions (Figure S3B, panel a). Increase in accessibility of octamer-bound positions to M.SssI in remodeled compared to unremodeled dinucleosomes was highly significant (Figure 6A; 601b NPS, p = 1.5 × 10–19; 603 NPS, p = 1.0 × 10–14). In other molecules, the linker DNA became significantly protected upon SWI/SNF remodeling (Figure 6A; P = 3.7 × 10–10). The protection patterns were clustered into 6 groups of remodeled dinucleosomes as shown in Figure 6A along with a schematic of the protection pattern of each group. One of the largest groups observed showed protection between the two nucleosomes and in the flanking DNA region (Figure 6A, group 1). Group 1 does not seem to represent two nucleosomes being moved together because the footprint region is larger than ~300 bp in 7 of 18 molecules. SWI/SNF is not contributing to this large footprint since it was competed off with excess competitor DNA before probing with M.SssI. The data suggest that the histone-DNA interactions are altered such that a single nucleosome can occlude and make inaccessible more than 147 bp of DNA to the DNA methyltransferase.
In the majority of cases among the remaining groups of molecules it appears that only one nucleosome is retained and was positioned towards either DNA end (groups 2 and 3, 38%) or in a more central position (group 4, 13%). These observations are consistent with the nucleosome loss previously observed in Figures 3 and and5A.5A. Under these conditions there is no preference for which nucleosome was likely to be mobilized by SWI/SNF as approximately the same number of molecules had one nucleosome remaining at either DNA end. These data suggest that the binding of SWI/SNF to these dinucleosomes is stochastic and therefore determines the direction of nucleosome movement. The last group of DNA protection patterns indicate that a nucleosome bound at one end might bridge and contact the other end of DNA (groups 5 and 6), although each end could be bound by a different octamer (Kassabov et al., 2003).
Rather than having SWI/SNF bind stochastically to dinucleosomes, the second approach used Gal4-VP16 bound to one end of the DNA template to specifically recruit and tether SWI/SNF to a unique site on the dinucleosome. In this manner it is possible to observe the systematic movement of one nucleosome towards an adjacent nucleosome and to visualize SWI/SNF remodeling under conditions that reflect those in vivo. Similar controls as in Figure 6A were done demonstrating that no changes in DNA methylation of dinucleosomes (Figure S4B, group b) were observed when Gal4-VP16 alone (group d) or SWI/SNF and ATP were added in the absence of Gal4-VP16 (group c). The latter shows that dinucleosome remodeling was dependent on recruitment of SWI/SNF by Gal4-VP16 in the presence of competitor DNA. Compared to unremodeled dinucleosomes, highly significant changes in accessibility were observed when SWI/SNF, ATP, and Gal4-VP16 were added together (Figure 6A; 601b NPS, P = 4.6 × 10–9; 603 NPS, P = 2.9 × 10–34) and the overall methylation distribution was notably different than that of the non-recruited samples (compare Figures 6A and 6B). The protection patterns of the remodeled nucleosomes were clustered into two groups. Group I showed a clear pattern of loss of 603 NPS protection or progressive movement of the 603 nucleosome towards the 601b nucleosome position (Figure 6B, panel I). Trends in changes of nucleosome position were observed by determining the percent of methylation at each of the 40 CpG sites (Figure 6C-D). The 601b and 603 nucleosomal regions under conditions without recruitment became equally more accessible to M.SssI after SWI/SNF remodeling were consistent with no preferred direction for nucleosome movement (Figure 6C). When SWI/SNF was recruited by Gal4-VP16, 603 nucleosomes were preferentially moved in a graded fashion thus making the 603 NPS more accessible to DNA methylation than the 601b region (compare Figure 6C with 6D). The trend pattern in Figure 6B group I molecules is thus consistent with the observed preferential movement of the 603 nucleosome. The methylation protection patterns also revealed that the original linker DNA between the 601b and 603 nucleosomes in group I became occluded to DNA methylation (Figure 6A, group 1). The 603 nucleosome is eventually moved from its original site with only one nucleosome remaining on the DNA template as seen in the lower molecules of group I.
The remodeled dinucleosomes in groups I and II have in common the same progressive loss of the 603 nucleosome, but in group II the linker DNA remains accessible (Figure 6B). It is surprising that the linker DNA remains accessible even as the 603 nucleosome is moved towards the 601b nucleosome region. Increased accessibility to DNA methylation is however not just reserved to the linker DNA region, but was highly significant in the group II set of remodeled dinucleosomes versus group I molecules (P = 1.6 × 10–22). The spans of methylation protection observed in the group II set of remodeled dinucleosomes are substantially smaller than are typical for mononucleosomes, indicating a loss of canonical nucleosome structure (Figure S4C). The majority of footprints are smaller than 121 bp (~80%) and approximately 40% are even less than 76 bp. Increased accessibility within the 601b nucleosome could be due to loss of histones, formation of DNA loops, or partial loss of DNA interactions within the nucleosome. These data suggest that although recruitment of SWI/SNF by Gal4-VP16 bound to its cognate site mobilized the proximal 603 nucleosome, it is the distal 601b nucleosome that is being dramatically altered or destabilized.
The MAPit data clarifies that the central nucleosome in Figure 5B found by site-direct mapping of remodeled 601b-603 dinucleosomes is the mobilized 603 nucleosome. Site-directed mapping of cross-linked H2B shown in Figure S2 part C demonstrates that most of the H2B was retained at the leading edge of the mobilized 603 nucleosome that moved to within 35 bp of the original pseudodyad axis of the 601b nucleosome. Movement and retention of the H2B/H2A dimer is seen by the H2B cleavage site shifting from bp 188 to bp 273, 283, and 295. These data suggest that the H2A/H2B dimer that is lost initially upon remodeling is likely from the 601b nucleosome. The nucleosome positions from MAPit correlate well with those found by site-directed mapping. Both sets of data point to the 603 nucleosome being moved until SWI/SNF encounters the H3/H4 tetramer of the 601b nucleosome.
The single molecule approach of MAPit also clarifies that SWI/SNF remodeling of dinucleosomes indeed promoted loss of one of the two nucleosomes by showing that a significant percentage of remodeled dinucleosomes retained only one nucleosome. These data also provide evidence for eviction being linked to the movement of one nucleosome in the direction of a second nucleosome. Together the mapping of single molecules by MAPit and bulk measurement of detailed histone-DNA interactions in remodeled dinucleosomes shows that SWI/SNF has a strong intrinsic activity in the context of nucleosomal arrays to disassemble nucleosomes.
It has long been thought that SWI/SNF primarily moves nucleosomes along DNA, but is unable on its own to disassemble nucleosomes based on studies using mononucleosomes as substrates. There have been hints over the years that SWI/SNF may have the ability to partially disassemble nucleosomes as it can inefficiently transfer histone octamers to free DNA and exchange histone H2A/H2B dimers (Bruno et al., 2003; Lorch et al., 1999). The efficiency of these reactions however did not suggest that they are the primary activities of SWI/SNF. It has become a point of debate whether SWI/SNF or RSC have the intrinsic ability to disassemble nucleosomes alone or instead require histone chaperones such as Asf1 and Nap1 or free DNA as histone acceptors for efficient nucleosome disassembly (Adkins et al., 2004; Adkins et al., 2007; Boeger et al., 2003; Boeger et al., 2004; Korber et al., 2006; Reinke and Hörz, 2003).
We have shown that when SWI/SNF remodels a dinucleosome first one nucleosome is moved towards an adjoining nucleosome with one H2A/H2B dimer being displaced (Figure 7). In a subsequent and slower step, an H3/H4 tetramer is displaced resulting in the entire disassembly of one nucleosome and ultimately a lone nucleosome remaining on the DNA template. These data also demonstrate that nucleosome disassembly by SWI/SNF does not require free DNA or histone chaperones as histone acceptors, but instead can proceed efficiently in their absence. In vivo data had suggested that chromatin remodeling at the PHO5 promoter acts in this manner to create a nucleosome free region (Boeger et al., 2008). Our data point to efficient displacement of H2A/H2B and nucleosome disassembly by SWI/SNF requiring an adjacent nucleosome. Why then does SWI/SNF require two nucleosomes in order to displace one? A recent model was proposed for SWI/SNF and RSC colliding nucleosomes together to form overlapping dinucleosomes (Engeholm et al., 2009). This model predicts the loss of H2A/H2B dimers, but falls short of explaining how complete nucleosome disassembly might occur. Also our data shows that SWI/SNF is unable to move two nucleosomes together with ≤6 bp of linker DNA. It is expected that the narrow, enclosed nucleosome binding trough constituting the SWI/SNF active site (Dechassa et al., 2008) would preclude accommodating one octamer being actively moved along DNA and a second encountered nucleosome to form an overlapping dinucleosome. These same constraints are also probably true for the RSC complex (Chaban et al., 2008). Therefore it is more likely that SWI/SNF itself rather than the nucleosome on which it is translocating directly collides with and preferentially destabilizes an adjoining, distal nucleosome.
We propose a different model based on our data and the key observation of the ordered nucleosome disassembly of SWI/SNF resembling that observed by RNA polymerase II (Kireeva et al., 2005; Kulaeva et al., 2007; Studitsky et al., 2004). The nucleosome collision model assumes that the role of SWI/SNF is merely to mobilize nucleosomes and discounts its ability to translocate and track along DNA to directly affect an adjacent nucleosome. Both SWI/SNF and RNA polymerase II are large molecular machines that in translocating along DNA dissociate it from the histone octamer surface in nucleosomes they encounter and trap or prevent DNA from re-associating with the histone octamer. The observed sequential loss of an H2A/H2B dimer followed by the displacement of an H3/H4 tetramer is the expected outcome from this kind of mechanism. Evidence for SWI/SNF mobilizing a single nucleosome bound in its active site while invading a second adjacent nucleosome comes from mapping single dinucleosomes remodeled under recruitment conditions (Figure 6B). DNA protection shows that, as the nucleosome proximal to the Gal4 binding site moves towards the distal nucleosome, the distal nucleosome can become more accessible and generally lacks a canonical 147 bp size protection (Figure 6B, Group II). These changes in the nucleosome distal to the bound transcription activator all occur without the linker DNA separating the two nucleosomes being protected once SWI/SNF is removed. The model in Figure 7 shows SWI/SNF competing for DNA bound to the octamer in the distal nucleosome (D) while it pulls DNA into and mobilizes the proximal nucleosome (P).
There are two differences between SWI/SNF and RNA polymerase. First, while mobilizing one nucleosome deep in its active site trough, SWI/SNF itself rather than the nucleosome invades a second nucleosome by moving DNA from it into the SWI/SNF-bound nucleosome (Figure 7B-D). The other difference is that SWI/SNF does not transfer histone octamers to DNA immediately behind it as is observed for RNA polymerase. SWI/SNF can therefore start with a completely assembled nucleosomal array and remove some or most of the nucleosomes from the template, while RNA polymerase would require a sufficient stretch of nucleosome-free DNA to begin translocation and then only shuttles nucleosomes on DNA rather than evicting them.
Another observation that is consistent with the SWI/SNF collision with distal nucleosomes model rather than the two nucleosome collision is retention of a particular H2A/H2B dimer after remodeling. In the overlapping dinucleosome model, loss of an H2A/H2B dimer is expected to occur with the same frequency from either of the two colliding nucleosomes. We find instead that the H2A/H2B dimer at the leading edge of the mobilized proximal nucleosome is well retained and an H2A/H2B dimer is likely lost from the distal nucleosome. Retention of the H2A/H2B dimer in the nucleosome bound to SWI/SNF is consistent with this nucleosome being unchanged in histone content and agrees with findings that remodeled mononucleosomes do not lose an H2A/H2B dimer. It also suggests that the displacement of an H2A/H2B dimer (and subsequently an H3/H4 tetramer) is not due to the collision of nucleosomes, but rather the collision of SWI/SNF itself with an adjoining nucleosome.
Although in more than half of the single remodeled dinucleosomes mapped the linker DNA remains accessible as one nucleosome is moved towards the other, what about those that resemble overlapping dinucleosomes (Figure 6B, group I)? Dinucleosomes in both groups I and II are most plausibly due to the initial remodeled dinucleosomes collapsing into one of these two structural classes after removal of SWI/SNF. By our model, the proposed loop extruded by SWI/SNF (Figure 7B, DNA in red) is resolved by migration around the octamer away from the distal nucleosome, advancing the proximal towards the distal nucleosome and yielding group I. The two nucleosomes may also be brought closer together in group I molecules when DNA displaced from distal nucleosome (Figure 7B-C, DNA in green) is rewrapped onto the octamer surface. Overlapping dinucleosomes may only form after displacement of SWI/SNF and therefore not be an accurate representation of the true state of a remodeled dinucleosome.
Previously it had been suggested that the same Gal4-VP16 transcription factor can promote nucleosome disassembly by directing SWI/SNF to move nucleosomes towards the bound Gal4 site (Gutierrez et al., 2007). We found however that nucleosome movement in a dinucleosome context is away from the Gal4 binding site consistent with other published data (Dechassa et al., 2008; Zofall et al., 2006). DNA footprinting found that SWI/SNF recruited by an activator contacted the DNA gyre of the nucleosome opposite from the side of the Gal4 binding site (Dechassa et al., 2008). In this orientation the nucleosome should move away from the Gal4 binding site based on mapping the binding and translocation site of SWI/SNF with DNA gaps that block nucleosome movement (Zofall et al., 2006). The direction of nucleosome movement by SWI/SNF in regard to the transcription activator is important because if nucleosome movement is towards the transcription factor then SWI/SNF would cause efficient displacement of the transcription factor from DNA (Fletcher et al., 2002). SWI/SNF would be less likely to remain at this site once the transcription factor is released. If instead the direction is as observed in this study it could explain how SWI/SNF might be retained at the promoter site and its remodeling activity restricted to a region encompassing a few nucleosomes (Almer et al., 1986; Jessen et al., 2006). Targeting SWI/SNF tethered by transcription factors to proximal nucleosomes would be consistent with their preferential clearing as observed for the PHO5 promoter region (Almer et al., 1986; Jessen et al., 2006). Our in vitro data correlates well with that observed by Kornberg and colleagues when modeling the stochastic nature of the PHO5 promoter region in which one nucleosome is retained at the promoter (Boeger et al., 2008). The well-positioned dinucleosome substrate bound by Gal4 has proven to be an effective test system for examining the properties of SWI/SNF in unprecedented detail that appear to reflect and explain several of its in vivo properties.
Nucleosome arrays were constructed, labeled, and reconstituted as described in Supplemental Data. The SWI/SNF binding and remodeling assays were done as previously reported (Dechassa et al., 2008) with additional modifications described in Supplemental Data.
Site-directed mapping was conducted as previously described using nucleosomes with cysteine substituted at residue 53 of histone H2B or residue 120 of histone H3 that were coupled to p-azidophenacyl bromide (Fluka) (Kassabov et al., 2002; Kassabov and Bartholomew, 2004). Methods for monitoring histone eviction using dual labeled nucleosomes and methyltransferase accessibility are described in the Supplemental Data section.
We would like to thank members of the Bartholomew lab as well as Russ Darst and other members of the Kladde lab for critical discussions and for support from the National Institutes of Health (GM 48413 to BB and CA 95525 to MPK).
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.