|Home | About | Journals | Submit | Contact Us | Français|
Chromatin remodelers are multifunctional protein machines that use a conserved ATPase motor to slide nucleosomes along DNA. Nucleosome sliding has been proposed to occur through two mechanisms: twist diffusion and loop/bulge propagation. A central idea for both of these models is that a DNA distortion propagates over the surface of the nucleosome. Recent data from biochemical and single-molecule experiments have expanded our understanding of histone-DNA and remodeler-nucleosome interactions, and called into question some of the basic assumptions on which these models were originally based. Advantages and challenges of several nucleosome sliding models are discussed.
As the basic packaging units of eukaryotic genomes, nucleosomes provide both a rich medium for accumulation of epigenetic marks and a physical barrier that restricts access to DNA. These two nucleosomal features – presenting epigenetic signatures and blocking DNA access – are integrated by a diverse group of protein machines called chromatin remodelers. Chromatin remodelers can read epigenetic marks through auxiliary domains or associated subunits and can assemble, reposition, and disassemble nucleosomes in an ATP-dependent fashion [1,2]. The unifying feature of chromatin remodelers is a helicase-like ATPase motor required for nucleosome reorganization. A major challenge in the remodeling field is to understand how this ATPase motor engages with and manipulates nucleosomal DNA to effectively deposit, remove, and slide nucleosomes along DNA.
An early idea for how the central histone core might be shifted along the DNA is known as the twist diffusion model (Figure 1a) [3–5]. This model suggests that single base pairs can be transferred between linker DNA and the DNA wrapped around the histone core, such that segments of nucleosomal DNA twist or untwist to accommodate the loss or gain of one or more base pairs. Although such a loss or gain would generate a twist defect in a segment of nucleosomal DNA, the model predicts that a small twist defect could be tolerated by the nucleosome without disrupting histone-DNA contacts. If a twist defect from one or more base pairs were propagated all the way around the nucleosome from one DNA segment to the next, so-called twist diffusion, the histone octamer would effectively shift along the DNA by the size of the distortion. This model has been strengthened by the discovery of structural variability around the histone core, where certain regions of the nucleosome can support a gain or loss of one base pair [6–9]. However, due to the helical nature of DNA, shifting DNA past the histone core one base pair at a time would necessitate rotation of the DNA by ~35° with each base pair step, which would alter the rotational phasing of the DNA on the histone core. Notably, ATP-dependent nucleosome sliding is not inhibited by physical barriers such as DNA hairpins and biotin crosslinks that should prevent rotation of the DNA duplex during sliding [10,11]. In addition, chromatin remodelers have been reported to slide nucleosomes in increments of approximately 10 base pair steps, maintaining the rotational phasing of nucleosomal DNA [12””]. Thus, while twist diffusion remains an attractive mechanism for shifting nucleosomes along DNA in response to thermal fluctuations , chromatin remodelers likely utilize a different mechanism for ATP-dependent sliding of nucleosomes.
A distinct but related model for nucleosome sliding is known as the loop or bulge propagation model (Figure 1b) [13–15]. Similar to twist diffusion, the loop/bulge propagation model suggests that DNA from one linker transiently shifts onto the nucleosome, creating a loop or bulge of DNA that rapidly diffuses around the histone core to emerge on the other side. An important difference from twist diffusion is the idea that DNA is pushed onto the nucleosome in larger segments, thus locally disrupting one or more contacts with the histone core. The loss of histone-DNA contacts makes a DNA loop more energetically expensive than a twist defect, but unlike twist diffusion, loop propagation should allow DNA to maintain rotational phasing on the histone core .
Biochemical and single molecule experiments have brought into question several key assumptions that originally made loop propagation attractive, and two variations on the loop propagation model (discussed below) have emerged to address these new findings. The recent experimental results also allow for an alternative model for nucleosome sliding, originally dismissed on structural grounds, where the DNA slips past the histone core in a concerted motion. Such a shift would be equivalent to a rotation or swiveling of the histone core with respect to the outer wrapping of DNA, and effectively move the histone core along DNA without a change in rotational phasing.
A major advance in chromatin biology was achieved with the structure of the nucleosome  (Box 1). The pseudo-symmetric nature of the nucleosome structure led to the initial idea that the DNA is firmly held in place at each of the histone-minor groove contacts around the nucleosome, much like a chain engaged in the teeth of a gear . However, single molecule experiments designed to probe the energetics of histone-DNA contacts have demonstrated that the histone-DNA contacts vary greatly in strength between the central dyad and entry/exit sites. In particular, in experiments where the two ends of the DNA duplex were pulled in opposite directions, the outer wrap of DNA was found to be much more easily displaced from the histone core than the inner wrap [17,18]. These pulling experiments agree with FRET and AFM experiments, which demonstrated that the outer turn of DNA spontaneously unwraps and rewraps around the histone core [19”, 20–22].
(a) The crystal structure of the first nucleosome core particle (NCP) revealed 146 base pairs (bp) of duplex DNA wrapped around a central histone core in a left-handed superhelix (; pdb code 1AOI). The eight histone proteins (two each of H2A, H2B, H3, and H4) form two types of similarly organized heterodimers: H2A–H2B and H3-H4 . In the octamer, these four dimers are arranged about a two-fold symmetry axis, called the dyad, which also intersects with the middle of the DNA fragment. Two symmetry-related H3-H4 dimers define the center of the octamer, and contact the DNA around the dyad. (b) In the nucleosome, each histone dimer grips three consecutive minor grooves of DNA in a similar fashion, with the central contact primarily made by the N-terminal side chains and backbone of the α1 helix for each histone in the dimer (α1-α1), and the two outer contacts made by the loops preceding the second helix of one histone and the third helix of the other (L1–L2). The four histone dimers thus use just two basic motifs (α1-α1 and L1–L2) to coordinate 12 minor grooves of DNA. (c) Schematic diagram of histone-DNA organization. A simple and effective nomenclature for describing positions around the nucleosome is given by so-called Super Helical Locations, or SHLs, which reference the major and minor grooves of DNA as they alternately face the central core . The two-fold symmetry axis of the histone core (where the two H3–H4 dimers come together) aligns with the major groove at the middle of the ~146 bp DNA fragment, and this region at the dyad is referred to as SHL0. Traveling along the DNA away from this central position, each major groove facing the histone core is denoted SHL±1, SHL±2, etc (positive in one direction, negative in the other), whereas each minor groove facing the core, halfway between major grooves, are named SHL±0.5, SHL±1.5, etc. Thus, the three minor grooves coordinated by the histone fold of the H3-H4 dimers extend from either side of the dyad (SHL±0.5, SHL±1.5, and SHL±2.5), and the H2A-H2B dimers pick up the minor grooves for the flanking ~30 bp DNA segments (SHL±3.5, SHL±4.5, and SHL±5.5). One additional minor groove interaction is made on either side of the central ~120 base pair DNA segment by the N-terminal helix and tail for each copy of histone H3, which projects between the gyres or superhelical turns of DNA (blue cylinders). Note that for clarity, most of the DNA on one side of the dyad is not shown. The N- and C-termini that are not part of the canonical histone fold, typically referred to as histone tails, are indicated with colored paths. The histone tails extending away from the nucleosome core particle are believed to be largely disordered in solution, but may still help orient chromatin remodelers and other factors on the nucleosome.
Recently, an unprecedented, high-resolution view of histone-DNA energetics has been provided by the Wang group using a single molecule technique called DNA unzipping [23””,24”]. In DNA unzipping, a force-feedback laser tweezers apparatus is used to pull apart the two strands of a single DNA duplex, effectively unzipping the Watson-Crick basepairs. The distance of separation between the two attachment points can be measured with near base pair resolution, which allowed these researchers to map where on the DNA sequence a greater force is required for unzipping. Unexpectedly, the remarkably high resolution of this technique revealed that energetically distinct histone contacts occur at approximately 5 base pair intervals around the nucleosome (Figure 2). That is, the two strands of DNA at each minor groove contact site contribute independently to histone binding, rather than as a unit for each minor groove as previously thought. Another interesting finding was the striking variation in strength of each histone-DNA contact. In agreement with previous studies [17, 18, 19”, 20–22], unzipping revealed three strong regions of histone-DNA contacts: a strong central contact around the dyad and two lesser but energetically significant contacts about 50–60 base pairs away on either side of the dyad. The high resolution of this technique allowed for the identification of an energetically weak contact at SHL2.5 (see Box 1c). Interestingly, this contact flanks a segment of nucleosomal DNA at SHL2 that has been shown to bulge or stretch by ±1 basepair in the nucleosome crystal structure [6–9], and coincides with the location where chromatin remodelers engage nucleosomal DNA (see below).
Since the unzipping technique sequentially disrupts interactions along the DNA, the Wang group was uniquely able to probe how a nucleosome would respond to a DNA-translocating factor, such as a polymerase, moving along the nucleosomal DNA. Unzipping the DNA up to, but not past, the nucleosome dyad did not alter the position of the histone core on the DNA [23””]. This agrees with previous nucleosome unwrapping studies showing that nucleosome position is not shifted with unwrapping of the outer DNA segment [19”].
The concentration of energetically important histone-DNA contacts around the nucleosome dyad carries important implications for nucleosome sliding mechanisms. Is disruption of histone-DNA contacts around the dyad sufficient to allow for sliding? The answer appears to be yes. In searching for mutations that would suppress a functional loss of the SWI/SNF remodeler, genetic selections in yeast identified a class of so-called SIN mutations (for SWI/SNF independent) that localized to where histones H3 and H4 contact DNA at SHL±0.5 . In nucleosome positioning experiments using elevated temperature, SIN mutant histone octamers were found to shift more easily on DNA than wild type, a characteristic that explained how these mutations suppressed the loss of a nucleosome repositioning factor [26,27]. Structural analysis of several SIN mutants showed that despite differences in positional octamer stability on the DNA, the molecular interactions are essentially intact . Thus, in stark contrast with unwrapping (or unzipping) the outer segment of DNA from the nucleosome, even apparently small perturbations of these energetically important contact points at SHL±0.5 allow the histone core to more easily shift on DNA.
Chromatin remodelers comprise a diverse and highly specialized collection of protein machines, sharing only a conserved helicase-like ATPase motor necessary for powering the remodeling reactions [1,2]. This ATPase motor was shown to have 3’ to 5’ translocase activity on naked DNA and generate torsional strain in the presence of DNA and nucleosomes [28”, 29”,30”]. An important shift in thinking about chromatin remodeling was brought about by several elegant DNA-gap experiments, which suggested that the ATPase motor engaged with DNA at an internal site on the nucleosome [12””,31””, 32””, 33]. Surprisingly, the presence of DNA gaps – single-stranded regions interrupting the DNA duplex – had little effect on nucleosome sliding except when these gaps were located around SHL2 (Figure 3). This led to the idea, supported by crosslinking [34,35], that the ATPase motor primarily acted at SHL2, where it could disrupt histone-DNA contacts through the buildup of torsional strain in the nucleosomal DNA.
Beyond this region at SHL2, at present no atomic resolution details are known about how remodelers specifically engage with the nucleosome. Electron microscopic reconstructions of the larger remodelers, SWI/SNF and RSC, have revealed multilobed structures with a central chamber or trough where the nucleosome is believed to bind [35–37] (Figure 3b). For the SWI/SNF remodeler, which possesses 12 subunits, footprinting and crosslinking have shown significant interactions over much of the nucleosome, consistent with partial burial of the nucleosome within the remodeler complex . In contrast, the smaller Iswi-type remodelers, which typically have 2–4 subunits, have been shown by footprinting to make much more limited contacts with the nucleosome, but with significant interactions with linker DNA outside the nucleosome core [31””,38–40]. Nucleosome sliding by Iswi-type remodelers is highly dependent on linker DNA length, and the association of remodelers on either side of the nucleosome has been predicted to allow for centering the histone core on DNA fragments and in nucleosomal arrays [11,40–42].
New levels of histone-DNA interactions have recently been revealed by the Bartholomew lab, who used DNA footprinting and site-specific crosslinking in combination with DNA-gapped nucleosomes to investigate changes in Isw2-nucleosome interactions during the ATP hydrolysis cycle [43””]. Subsequent to the initial binding event, activation of the Isw2 ATPase was coupled to dramatically increased protection of nucleosomal DNA between SHL2 and the entry/exit site. Intriguingly, this larger footprint was observed even when DNA translocation around the histone core was blocked by gaps, which was interpreted as a channel-like organization of the protein around the nucleosomal DNA. Hydrolysis-dependent changes in remodeler-nucleosome interactions were also evidenced by the ATP-dependent crosslinking of the Isw2-specific subunit Dbp4, a histone-fold-like subunit, to SHL4. In addition, this work demonstrated that Iswi-type remodelers form template-committed complexes and slide nucleosomes processively. Template commitment depended on the presence of the histone H4 N-terminus, an epitope previously shown to increase nucleosome-specific interactions and influence remodeling efficiency [44–46].
How precisely the ATPase motors promote a shift in DNA placement is a subject of intense interest. The idea of a DNA loop propagating around the histone core has served as the basis for two distinct models for how nucleosome sliding might occur. In one model, the ATPase motor utilizes its DNA translocase ability to directly pull DNA in from the nearest entry/exit site in a continuous fashion [32””,47,48]. By continuously pumping DNA towards the dyad, the remodeler could create a bulge that would rapidly diffuse to the distal entry/exit site. This model is supported by single molecule studies of the RSC and SWI/SNF chromatin remodelers, which suggested that the remodeler continuously pulls DNA into loops [48,49]. Although the exact location of these loops relative to the nucleosome is not clear in single molecule studies, the previously defined site of ATPase action at SHL2 led to the conclusion that the DNA loops would be formed around the dyad region . This model is attractive in that it couples the known DNA translocation activity of remodelers directly to DNA movement. The idea of the remodeler maintaining its position around SHL2 and attempting to translocate away from the dyad explains why DNA gaps specifically block sliding when located around SHL2, and is in accordance with the observation that RSC and SWI/SNF remodelers can shift the histone core approximately 50 basepairs off the end of DNA [32””,50,51]. Continuous translocation on DNA also nicely explains processivity, as the ATPase motor would not need to detach from the nucleosomal DNA during nucleosome sliding. One difficulty with this model, however, is that continuous translocation along DNA would alter the rotational phasing of the nucleosomal DNA and significantly twist and supercoil the DNA, which, along with the twist diffusion model, goes against experimental evidence showing that the DNA does not appreciably rotate during ATP-dependent sliding [10,11,12””]. A more recent variation of this DNA pumping model suggests that each single base pair step is accompanied by an additional 10 base pair step made by a combination of remodeler domains, which would reduce rotation of the DNA duplex but still actively pump DNA towards the dyad .
In another model, a DNA loop is similarly proposed to form in the vicinity of SHL2 by a remodeler ATPase disrupting histone-DNA contacts and pulling DNA in from the linker [12””,43””]. In contrast to the DNA pumping model, however, the ATPase motor is proposed to remain fixed on the DNA as the loop is formed, such that the ATPase motor would shift towards the dyad along with the loop of DNA. In this model, the energy stored in the distorted DNA loop would be sufficient to promote loop propagation around the histone octamer, and some non-ATPase remodeler domains would play the important role of biasing propagation towards the dyad. The finding that the strongest histone-DNA contacts cluster around the dyad challenges the original idea of loop propagation [13–15], where a loop formed near the entry/exit site would have to travel a long distance and energetically uphill to cross the dyad. However, by acting around SHL2, the proximity of the remodeler ATPase motor to the dyad suggests that loop formation may occur close enough to the strong dyad contacts to directly assist propagation of a loop past this energetic barrier. A key aspect that remains to be sorted out is how the non-ATPase domains alternately bind and release nucleosomal DNA to help form and propagate DNA loops [12””,43””]. Clearly the non-ATPase domains must communicate with the ATPase motor, but whether this is a direct or indirect coupling awaits higher resolution information from structural and biochemical experiments.
With the marked variation in strength of histone-DNA contacts around the nucleosome, the original image of a sharp-toothed gear firmly gripping a chain on all sides is no longer applicable. Instead, the clustering of a few strong DNA contact sites around the dyad suggests a potentially simple strategy for releasing DNA from the grip of the histone core. Chromatin remodelers, with a helicase-like ATPase motor engaging with nucleosomal DNA around SHL2, are well positioned to alter histone-DNA contacts at SHL1.5 and perhaps SHL0.5. Since SIN mutants allow for an easier shifting of the histone core along DNA by weakening the histone-side of the contact, it is not difficult to imagine that chromatin remodelers could also perturb these energetically important interactions by transiently altering DNA structure.
How would such a disruption of key histone-DNA contacts allow for movement of DNA past the histone core? Instead of a local bulge or loop of DNA propagating around the core, the entire segment of DNA wrapped around the histone core could simply shift in concert. Since either the DNA or histone core can serve as a frame of reference, a concerted shift of the DNA is equivalent to a rotation of the histone core with respect to the DNA (Figure 4). Several observations lend support to the idea of a swivel-like shift in histone-DNA contacts. Electrostatic interactions between charged residues are by nature longer range and more delocalized than hydrogen bonds and van der Waals contacts, and the DNA-binding surface of the histone octamer is covered with a preponderance of basic residues (Figure 4b). The long reach of arginine and lysine side chains and dominant electrostatic nature of the interaction with DNA would allow for continued favorable interactions if the DNA minor grooves were shifted between SHL contacts. As shown by DNA unzipping [23””], the finding that the two DNA strands of the minor groove independently contribute to binding the histone core suggests that intermediate rotational positions of the histone core could achieve favorable interactions between DNA and α1-α1/L1–L2. A rotation of the histone core by ~18°, for example, would shift the DNA by the width of the minor groove, placing one half of the minor groove in the former location of the other half and thus potentially forming similar contacts with the histone backbone and side chains at α1-α1/L1–L2. Finally, the sharply bent organization of nucleosomal DNA may be maintained even with a transient loss of the canonical α1-α1/L1–L2 contacts. Creating or maintaining sharp bends in DNA is less energetically expensive than once thought , and DNA flexibility and condensation greatly increase in the presence of polyamines and positively charged surfaces [54–56]. For nucleosomes, DNA would be expected to be stabilized in a wrapped organization in large part due to the densely charged surface of the histone core.
A solid body of experiments has revealed key structural, biochemical, and energetic aspects of nucleosomes and nucleosome-remodeler interactions, providing a framework for understanding potential mechanisms underlying chromatin remodeling. Given the compositionally rich and dynamic nature of chromatin, distinct sliding mechanisms – twist diffusion, bulge propagation, and histone core swiveling – may coexist and are likely influenced by DNA sequence and associated chromatin factors. Many basic questions regarding the molecular mechanisms of nucleosome sliding remain: What DNA distortions accompany engagement of the remodeler ATPase with the nucleosome, and how do these changes encourage histone octamer repositioning? Why do the ATPase motors target SHL2, and how is the motor localized? How is processivity in nucleosome sliding achieved? How do other remodeler elements participate in the remodeling reaction? Beyond sliding, how do remodelers cooperate with histone chaperones and use chaperone-like subunits to promote nucleosome assembly and disassembly? Addressing these and other questions will require increasingly sophisticated experimental and computational approaches, and yet will assuredly bring us to new levels of understanding in the not too distant future.
I thank Ilana Nodelman, Jeff McKnight, and Ana Damjanovic for helpful discussions and comments on the manuscript. I gratefully acknowledge support from the National Institutes of Health (R01 GM084129).
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.