All ATP-dependent chromatin remodelers have a DNA translocase domain that moves along double-stranded DNA when hydrolyzing ATP, which is the key action leading to DNA moving through nucleosomes. Recent structural and biochemical data from a variety of different chromatin remodelers have revealed that there are three basic ways in which these remodelers self-regulate their chromatin remodeling activity. In several instances, different domains within the catalytic subunit or accessory subunits through direct protein–protein interactions can modulate the ATPase and DNA translocation properties of the DNA translocase domain. These domains or subunits can stabilize conformations that either promote or interfere with the ability of the translocase domain to bind or retain DNA during translocation or alter the ability of the enzyme to hydrolyze ATP. Second, other domains or subunits are often necessary to anchor the remodeler to nucleosomes to couple DNA translocation and ATP hydrolysis to DNA movement around the histone octamer. These anchors provide a fixed point by which remodelers can generate sufficient torque to disrupt histone–DNA interactions and mobilize nucleosomes. The third type of self-regulation is in those chromatin remodelers that space nucleosomes or stop moving nucleosomes when a particular length of linker DNA has been reached. We refer to this third class as DNA sensors that can allosterically regulate nucleosome mobilization. In this review, we will show examples of these from primarily the INO80/SWR1, SWI/SNF and ISWI/CHD families of remodelers.
Chromatin remodelers physically rearrange nucleosomes using energy derived from ATP hydrolysis and their remodeling activity can be modulated in response to variations in chromatin such as histone post-translational modifications (i.e. particularly histone tails), histone variants such as H2A.Z or length of linker DNA separating nucleosomes. Several remodelers, like the ISWI and SWI/SNF families, have been shown to require histone tails for efficient remodeling or recruitment of the remodeler, and when histones are acetylated or otherwise covalently modified to either enhance or block nucleosome remodeling [1–7]. The interplay of histone modifications and chromatin remodelers is an active area of research and there are likely many of these interactions yet to be discovered that regulate the activity or recruitment of remodelers. A large number of the pivotal studies have been done with only the catalytic subunit or with subcomplexes of a limited number of accessory subunits present in the complex. The INO80, ISWI and CHD families of remodelers are important for regulating nucleosome spacing at promoters or coding regions of genes, making them crucial for transcription regulation [8–15]. These remodelers require not only a minimum length of linker DNA for mobilizing nucleosomes but also a sufficient length of linker DNA to be recruited to chromatin. Some chromatin remodelers target chromatin regions that are marked by the incorporation of histone variants that are less abundant in the cell . These same remodelers can also change the levels of histone variants incorporated at distinct regions in the genome. By understanding more fully the various ways in which chromatin remodelers are regulated, we will learn more about the mechanisms of chromatin remodeling and how these are disrupted in diseases through mutation of these complexes.
The brakes and accelerators of the ATPase domains
The importance of histone tails for regulating chromatin remodeling is probably best characterized in the ISWI family. For ISWI complexes, the H4 histone tail has been shown to be required for efficient remodeling and the H4 tail peptide alone with free DNA to stimulate the ATPase activity of ISWI [5–7,17]. Histone acetylation can modulate ISWI chromatin remodeling activity and acetylation of the H4 tail mitigates the stimulatory effect of the H4 tail with ISWI. Most of the studies with the ISWI remodeler have been done with only the catalytic subunit, even though these complexes generally have 2–4 subunits, and have found both negative and positive regulators of the DNA translocase domain. Deletion analysis of ISWI found two domains that negatively regulate ISWI remodeling called the AutoN and NegC domains [18–20]. Analysis of the crystal structure of ISWI with and without H4 peptide found the AutoN to bridge lobes 1 and 2 of the DNA translocation or ATPase domain, promoting the formation of an inactive ISWI conformation (Figure 1A), and H4 tail relieves inhibition from the AutoN domain by competitive binding to lobe 2 . The role of the C-terminal NegC domain is not as clear in the structure because it reaches out to an adjacent ATPase domain and binds to lobe2, which could be consistent with reports of ISWI forming dimers . These experiments were all performed with only the ISWI catalytic subunit, which was truncated and lacked the C-terminus containing the HAND, SANT and SLIDE domains in most experiments, and did not have any of its accessory subunits. From other experiments, it is evident that the accessory subunit does have a significant impact on the activity of the catalytic subunit and the accessory subunit can also bind the H4 tail and affect H4 tail sensing . CHD1 has similar structures to ISWI in regard to both the AutoN and NegC domains of ISWI acting similar to the chromodomains and C-terminal bridge of CHD1 (Figure 1B) . The AutoN and NegC domains are therefore examples of domains acting to brake the ATPase domain (Figure 2A).
Structural examples of brakes, accelerators and anchors from different chromatin remodeling families.
Modes of regulation for ATP-dependent chromatin remodelers.
Other domains and accessory subunits act as accelerators by positively regulating the activity of the ATPase domain (Figure 2A). For many of the chromatin remodelers, it is fairly clear that accessory subunits stimulate the overall activity of complexes. These effects could be due to enhancing substrate binding or recruitment and not be directly regulating the ATPase activity. One example from yeast SWI/SNF shows that the Snf5 accessory subunit directly contacts the ATPase domain by mapping intra- and inter-subunit interactions using chemical cross-linking and mass spectrometry (CX-MS). Snf5 in mammalian SWI/SNF is a tumor suppressor, and in yeast, SWI/SNF forms a submodule consisting of Snf5, Swp82 and Taf14 (a YEATS domain protein) and the entire submodule is lost from SWI/SNF when Snf5 is deleted. When this aberrant SWI/SNF complex is fully saturating nucleosomes, the ATPase and remodeling activities of SWI/SNF are reduced and indicates that the Snf5 submodule has a catalytic role in SWI/SNF remodeling . The Arp7 and 9 subunits bind to the Snf2 catalytic subunit through an N-terminal domain called the HSA (Helicase/SANT-Associated) domain and have been shown to genetically interact with the Protusion I region between the two lobes of the ATPase domain [24,25]. Arp7 and 9 in a minimal RSC complex lacking most of its accessory subunits stimulate its ATPase and DNA translocation activities, which, in turn, enhances nucleosome remodeling .
In the CX-MS study of SWI/SNF, domains within the Snf2 catalytic subunit such as the SnAC (Snf2 ATP Coupling) domain were also shown to contact the second lobe of the ATPase domain . Earlier studies found that the SnAC domain is conserved throughout all eukaryotes, is a SWI/SNF specific domain and activates the ATPase domain as well as serves as a crucial histone anchor needed to couple DNA translocation/ATPase activity to nucleosome movement [27,28]. The recent cryo-EM structure of the yeast Snf2 subunit confirmed that the SnAC domain interacts with lobe 2 of the Snf2 ATPase domain consistent with the earlier CX-MS data, even though much of the SnAC domain is not observed in this structure because of its inherent flexibility and missing parts of SnAC in the truncated Snf2 used in these studies (Figure 1C and ref. ). Another example comes from CHD1, which is one of the few remodelers in yeast that consists of only one subunit, the catalytic subunit and no other accessory subunits. The chromodomains of CHD1 appear to positively regulate the ATPase domain by swinging into a position juxtaposed to the ATPase domain on nucleosomal DNA, thereby causing closure of the ATPase domain and promoting its translocation along DNA (Figure 1B and ref. ). The CHD1 structure provides compelling evidence for the chromodomains positively regulating CHD1 through altering the conformation of the ATPase domain.
Torque and nucleosome anchors in chromatin remodeling
Next, we examine the roles of anchors within chromatin remodelers that bind to nucleosomes and work independent of regulating either the ATPase or DNA translocation activity of ATPase domains, but are nonetheless essential for nucleosome mobilization. This type of anchor is thought to be needed in order for the ATPase domain to translocate on DNA without DNA slipping and releasing the torque needed next to break histone–DNA contacts and move DNA through nucleosomes (Figure 2B). An early example of this type of anchor is the SnAC domain in SWI/SNF complexes which required the correct conditions such that the nucleosome anchor activity can be distinguished from the ATPase-stimulating activity of the SnAC domain. Experiments were done using a higher ATP concentration with SWI/SNF lacking the SnAC domain such that the rate of ATP hydrolysis with the mutant was equivalent or slightly higher than wild-type SWI/SNF. In these conditions, SWI/SNF lacking the SnAC domain was found to remodel nucleosomes nearly 200 times slower than wild type, even though the rate of ATP hydrolysis was slightly higher with the mutant complex . A protein footprinting technique in which Fe-EDTA was tethered to the surface of nucleosomes showed that the SnAC domain of yeast SWI/SNF was located proximal to the histone octamer surface. In tethered single-molecule experiments, the SnAC mutant SWI/SNF complex was found to translocate as efficiently along DNA as wild-type SWI/SNF, further showing the critical difference in the mutant complex is its histone-anchoring activity .
The other example is from the INO80 remodeling complex, which has 15 different subunits, mobilizes and spaces nucleosomes, and can potentially exchange H2A–H2B dimers [12,31]. Loss of the actin-related protein (Arp) 5 and Ies6 subunits of human INO80 uncouples the ATPase activity from the nucleosome-mobilizing activity of INO80 as shown using recombinant INO80. Omission of any one of the two subunits loses both subunits from the complex . In yeast, deletion of different regions of Arp5 causes the loss of Arp5 and Ies6 from the INO80 complex and these complexes, while the ATPase activity stimulated by the addition of nucleosome remained equivalent to wild-type INO80, the mutant INO80 was severely impaired for nucleosome mobilization . In these experiments, the addition of Arp5/Ies6 to the mutant INO80 complexes was able to restore the nucleosome-mobilizing activity of INO80, confirming the role of Arp5 and Ies6 for coupling ATPase activity to nucleosome mobilization . In the recent human and Chaetomium thermophilum INO80 (ctINO80) structures, Arp5 was found to directly contact nucleosomal DNA and the histone octamer close to the acidic pocket region, consistent with Arp5/Ies6 serving as a nucleosome anchor for INO80 with minimal interactions detected with the ATPase domain (Figure 1D and refs [34,35]). The Arp5 and Ies6 module is probably one of the more definitive examples of the important roles nucleosome anchors have in regulating the efficiency of nucleosome movement by the remodeler. Because of where Arp5 binds nucleosomes, Arp5 may be capable of distinguishing between H2A and H2A.Z where one of the primary differences between these two histones is the extended acidic patch found in H2A.Z.
DNA sensors for nucleosome spacing-type chromatin remodelers
At the C-terminus of the ISWI catalytic subunit are located three domains called the HAND, SANT and SLIDE domains. For the yeast ISW2 remodeling complex, the SLIDE domain was found to bind extranucleosomal DNA 19 bp from the edge of nucleosomes by site-directed DNA cross-linking and peptide mapping . Similarly, the SLIDE and SANT domains of ISW1a have been mapped by a different photocross-linking approach to ∼10 and 0 bp from the edge of nucleosomes when ISW1a binds dinucleosomes . The SLIDE domain is structurally similar to the SANT and MYB DNA-binding domain, except that it has more basic residues in the appropriate positions to promote DNA binding than the SANT domain . Based on the SLIDE domain's location on extranucleosomal DNA and ISW2 ability to move nucleosomes until the linker DNA has reached a minimum length of >20 bp , the SLIDE domain appeared to be a likely candidate in the ISW2 complex for sensing linker DNA length to regulate the ISW2 chromatin remodeling activity (Figure 2C). The DNA-binding interface of the SLIDE domain was targeted by mutating four basic residues suggested by molecular modeling to be critical for binding DNA, which was confirmed by DNA footprinting with hydroxyl-free radicals . The mutant SLIDE ISW2 complex did, however, bind normally everywhere else on nucleosomes like wild-type ISW2. Mutant SLIDE ISW2 hydrolyzes ATP and remodels nucleosomes 6–9 times slower than wild-type ISW2, and suggest that stable binding of the SLIDE domain to linker DNA is required for efficient ISW2 remodeling . Single-molecule FRET experiments showed that mutant SLIDE complexes were more prone to move DNA bi-directionally, whereas wild-type ISW2 had a strong preference to move DNA uni-directionally, thereby properly repositioning nucleosomes. These and other data point to the SLIDE domain being important for promoting DNA entering into the nucleosome as mediated by ISW2 and suggest that there are co-ordinated actions between the SLIDE and ATPase domains that are both required for efficient nucleosome movement by ISW2.
The INO80 complex also has DNA sensor activity in that it spaces nucleosomes 30 bp apart and required a minimum of ∼35–40 bp of extranucleosomal DNA with 70 bp being optimal for binding and mobilizing nucleosomes . In the recent cryo-EM structure, the part of INO80 binding to the extranucleosomal DNA was not resolved due to its inherent flexibility and does not provide much structural information or insights into the DNA sensor module of INO80 [34,35]. An alternative approach using site-directed DNA cross-linking was used instead to map the spatial arrangement of the INO80 subunits bound to linker DNA and by peptide mapping identified the protein domains or regions that directly bind linker DNA . The Arp8 and Arp4 subunits were found to bind close to each other 37–51 bp from the edge of nucleosomes and are at the correct location to properly sense linker DNA length for the INO80 complex. The non-conserved N-terminus of Arp8 and the actin fold region of Arp4 near its C-terminus were found to contact DNA. Deletion of the N-terminus of Arp8 eliminated binding of the Arp8–Arp4 module to linker DNA without perturbing the complex integrity of INO80 or its ability to bind nucleosomes or for nucleosomes to stimulate its ATPase activity. Chromatin remodeling was, however, reduced and shows that deletion of the N-terminus of Arp8 causes ATPase activity to be uncoupled from nucleosome movement . Additional experiments showed that loss of the N-terminus of Arp8 and its corresponding loss of binding to linker DNA causes an allosteric effect with the Ino80 catalytic subunit and its ATPase domain shifting from inside nucleosomes at super helical loop (SHL)-6 to the linker DNA region ∼30 bp away from its original position. There is also another allosteric effect in that Arp5's interaction with the acidic patch of nucleosomes is reduced, thus mirroring the effect when Arp5 is not present in the complex [32,33]. For INO80, there is evidence that the DNA sensor consisting of the Arp8–Arp4–actin module is an allosteric switch that disengages the critical histone anchor of INO80 required for coupling ATPase activity to nucleosome movement and causes repositioning of the DNA translocase domain that also leads to reduced nucleosome movement.
Yeast INO80 appears to have some redundancy in how it senses linker DNA as there is another protein module besides the Arp8–Arp4–actin module that binds to linker DNA. Nhp10, an HMG-like protein that associates with Ies1, 3 and 5, also binds close to the same region as Arp8–Arp4–actin and remains bound to this region in the absence of Arp8 and Arp4 binding to this region . In a separate study, deletion of Nhp10 causes the loss of an entire multi-subunit module from the INO80 complex and eliminates the DNA-sensing capability of INO80 altogether such that it mobilizes nucleosomes regardless of the length of linker DNA present . The Arp8 and Nhp10 modules appear to work in very distinct ways in that Nhp10 module negatively regulates INO80 remodeling such that it can only work with nucleosomes with particular linker DNA lengths, whereas binding of Arp8 is required for positively regulating INO80's chromatin remodeling activity.
ATP-dependent chromatin remodelers are key epigenetic factors involved in development and human diseases that have significant diversity of operation as well as compositional complexity, making it difficult at times to study. Much of the efforts in the field have focused on understanding how the fundamental motors of these complexes can be regulated, but there is a whole other level of regulation that we are just beginning to appreciate involving the other subunits within these complexes. The recent cryo-EM structures of the human and yeast INO80 complexes bound to nucleosomes provide important structural information and clues as to how accessory subunits can potentially regulate chromatin remodeling as critical anchors to the nucleosome core particle including both histones and nucleosomal DNA. Nevertheless, certain regions within the structure could not be resolved as exemplified by the cryo-EM structure of INO80, and additional information gained by other methods such as site-directed DNA and histone cross-linking, chemical CX-MS and other methods are needed. As the work moves forward, more attention should be given to the regulation of these large multi-subunit chromatin remodeling complexes not as individual subunits or small subcomplexes, but with the intact full complex to properly understand how these complexes are regulated and sometimes mis-regulated in various diseases. It will also be important to find the interplay and potential overlap in the ways of altering remodeling activity with brakes/accelerators, anchors and sensors. One domain may work in multiple ways such as the SnAC domain of SWI/SNF which enhances ATPase activity as well as functions as a histone anchor.
The Authors declare that there are no competing interests associated with the manuscript.