ISWI proteins form the catalytic core of a range of chromatin-remodeling activities that play important roles in gene regulation. In this study, we have used a range of different approaches to define the interaction between the ISWI ATPase and nucleosomal substrates. As reported previously, we found that ISWI forms more stable complexes with nucleosomes that contain additional linker DNA (6
). However, we also found that ISWI binds to DNA fragments in a length-dependent fashion. Binding to a 23-bp fragment was weak and undetectable on a 15-bp fragment. ISWI bound to 32- and 41-bp fragments with greater affinity, and binding to these longer fragments was observed to result in the formation of more discrete complexes. We next investigated the dependence of ISWI ATPase activity on nucleosomal linker DNA length and found that this also displays length dependence. Although we found that ISWI ATPase activity is stimulated to a lesser extent in the presence of free DNA as reported previously, we also found that this stimulation showed a dependence on DNA length that peaked at approximately 40 bp. The ATPase activities of both SWI/SNF and RSC have previously been reported to be stimulated by both double- and single-stranded DNAs (9
). Here, we report that the ATPase activity of ISWI is also stimulated by single-stranded DNA. In addition, we found that the stimulation of ISWI ATPase activity by single-stranded DNA shows a length dependence that correlates well with that observed using duplexes. The DNA length dependence that ISWI displayed in these assays raises the possibility that ISWI functions as an ATP-dependent DNA translocase.
In order to test this further, we adapted a triplex displacement assay to study the action of ISWI on nucleosomal templates. ISWI was able to displace TFOs most efficiently from sites close to the edge of the nucleosomes. However, this activity reduced to undetectable levels 50 bp from the edge of a nucleosome. This would be consistent with ISWI, exhibiting a DNA translocase activity oriented away from the edge of nucleosomes with limited processivity. Further evidence supporting the orientation of the translocase activity was obtained through the observation that the introduction of gaps between the nucleosome and the triplex could prevent triplex displacement in a strand-polarity-dependent manner.
While this work was in preparation, two reports were published presenting evidence that chromatin-remodeling activities translocate along DNA (26
). Fyodorov and Kadonaga used a template commitment assay to show that the ISWI-containing ACF complex remains associated with, and preferentially functions on, one template following exposure to an equivalent amount of a competing template. This commitment to the template to which ACF is exposed first depends upon ongoing chromatin assembly and may offer valuable new insight into the way in which ACF functions. Fyodorov and Kadonaga argue that the phenomenon of template commitment provides evidence that ACF functions as an ATP-dependent DNA-translocating motor. However, template commitment has been observed previously for many DNA and chromatin binding proteins that are not ATP-dependent DNA translocases (3
). In order to explain template commitment, it has been proposed that, during transient dissociation from a site on one template, a DNA binding protein will be more likely to reassociate with the same template, as the local concentration of the molecule to which it was originally bound will be far higher than that of other molecules in the solution (55
). It is difficult to rule out the possibility that similar processes explain the observations made by Fyodorov and Kadonaga.
The study by Saha et al. investigated the DNA length dependence of ATP hydrolysis by the yeast RSC complex and its catalytic subunit Sth1p. Both of these activities are more potent in ATP hydrolysis than ISWI. This has enabled a more thorough investigation of the relationship between the DNA length and ATPase activity to be performed than has been possible with ISWI. Importantly, for both Sth1 and RSC, it has been possible to establish that DNA length affects ATP hydrolysis by affecting the maximal rate of hydrolysis, Vmax
, rather than Km
. This is consistent with the increased length of the template DNAs, allowing more rapid ATP hydrolysis before the translocase reaches the end of the DNA fragment. This form of length dependence is similar to that observed for bona fide helicases such as T4 gene 41 (68
). In addition, Saha et al. performed an elegant experiment in which they showed that RSC and Sth1 have similar activities on short circular DNA molecules to longer linear ones. All of these approaches are consistent with these activities functioning as ATP-dependent DNA translocases. However, it should be noted that ATP hydrolysis need not necessarily provide a direct measure of DNA translocation.
In order to obtain additional evidence for DNA translocation, Saha et al. also used a triplex oligonucleotide displacement assay. They found that both Sth1p and RSC are capable of displacing TFOs in an ATP-dependent reaction. Although triplex displacement has been used previously to monitor the action of ATP-dependent DNA translocases and helicases (23
), it is conceivable that TFOs could be displaced by other means. As a control for this, Saha et al. showed that triplex displacement is reduced from a triplex DNA that lacks double-stranded extensions and is still possible on one that contains a nick at the duplex-to-triplex interface.
Our studies of ISWI are entirely consistent with those described by Saha et al. and provide additional support for the model they present. We obtained two lines of evidence that support the involvement of a processive activity in triplex displacement. Firstly, triplex displacement decays with distance from the nucleosome. As an epitope within the H4 tail has been defined as playing a key role in the stimulation of ISWI ATPase activity, it is likely that at least a proportion of the ISWI protein functionally associates with this part of the histone octamer (13
). The fact that the TFO displacement decays with distance from this epitope supports a model in which translocation is orientated away from the nucleosome. Additional evidence for this orientation is provided by the observation that the insertion of 5- or 10-bp gaps in the 3′-5′ strand between the nucleosome and the triplex prevents triplex displacement.
In contrast to the effect of gaps in the 3′-5′ strand, the introduction of 5- or 10-bp gaps in the 5′-3′ strand has no effect on the ability of ISWI to function in triplex displacement assays. These observations are most readily explained if ISWI exhibits a strand-specific translocase activity. Strand specificity is a property of bona fide DNA helicases, which can be separated into two groups, 5′-3′ or 3′-5′, reflecting the polarity in which DNA transport occurs (10
). Many members of the SF2 group, to which the ATP-dependent chromatin-remodeling activities are most closely related, display a 3′-5′-strand specificity. Our observation that 3′-5′, but not 5′-3′, gaps have an effect is consistent with this. Further evidence supporting the ability of remodeling activities to function on a single DNA strand stems from the observations that the ATPase activity of RSC (9
), Sth1 (51
), and ISWI (Fig. ) is stimulated by single- as well as double-stranded DNA.
It is interesting to note that the ability of ISWI to function in the displacement of triplexes located up to 40 bp from the edge of a nucleosome bears similarities to studies of the yeast Mot1 protein. This protein has homology to the ATP-dependent remodeling activities and functions to displace the TATA binding protein from DNA in an ATP-dependent reaction. Mot1-driven TATA binding protein displacement requires contact with a DNA “handle” adjacent to the TATA box, like ISWI, and the introduction of single-stranded DNA gaps does not prevent Mot1 functioning (16
Although the observations of Saha et al. together with the evidence presented here suggest to us that ISWI, RSC, and Sth1 are likely to function as ATP-dependent DNA translocases, it is important to note that it has not yet been possible to directly monitor the motion of any remodeling activity along DNA or determine the translocation rate as has been possible with Eco
). Until this has been achieved, it remains a formal possibility that ATP-dependent remodeling activities distort DNA in some way that does not involve translocation. This will be technically difficult, as it appears that RSC and Sth1 have a processivity limited to 80 bp (51
), while in the case of ISWI, our own assays suggest that ISWI processivity is limited to 40 bp. This is a relatively short distance compared to previously characterized ATP-dependent nucleic acid translocases such as gp14 (68
) and Eco
). In addition, this processivity limit is typically reached as a result of a stochastic dissociation event occurring at distances distributed around a mean length. This is expected to give rise to an exponential decay in processivity that fits to an equation hyperbolic for cofactor length. However, the work presented here and much of the data obtained by Saha et al. suggest that the length dependence displayed by these chromatin-remodeling enzymes is more abrupt than would be expected for an exponential decay process. This, combined with the fact that highly processive motion over long distances may not be required for the local distortion of chromatin structure, raises the possibility that a more precise mechanism regulates the processivity of remodeling enzymes. Other helicases, such as the RecBCD enzyme, have been observed to translocate in a stepwise fashion (4
). Our observations of DNA binding, double- and single-stranded ATPase activity, and triplex displacement are all consistent with a processivity limit for ISWI of approximately 40 bp. However, unlike RecBCD, our observations would suggest that ISWI is able to dissociate from the template after each cycle of translocation.
Interestingly, Sth1 and ISWI appear to have different processivity limits. Whereas RSC and Sth1 appear to be capable of translocating for 80 bp, for ISWI, both the ATPase assays and triplex displacement assays suggest that the processivity of ISWI is limited to approximately 40 bp. This could explain differences in the activities of the two enzymes. While RSC is able to disrupt nucleosomes in addition to moving them along DNA, ISWI appears to be more specialized toward nucleosome mobilization. The distortion of less DNA by ISWI may suit it for a role in nucleosome mobilization without nucleosome disruption. It is also interesting that the 40-bp limit for ISWI is compatible with the average nucleosome spacing in many higher eukaryotes.
How might a DNA translocase activity stimulate nucleosome mobilization? Our data were generated by using chromatin templates, which provide a means of orientating the distortion of DNA driven by an ATP-dependent remodeling activity, in this case, ISWI, relative to that of a nucleosome. As ISWI requires residues 16 to 19 of the H4 tail for maximal ATPase activity, it is likely that the ISWI retains contact with at least this part of the nucleosome during ATP hydrolysis. Translocation of the ISWI protein along DNA, while it remains associated with the histone octamer, could cause DNA to be injected onto the surface of nucleosomes. This would result in the generation of a loop or bulge of DNA on the surface of the nucleosome, which might then be able to transit around the surface of the octamer in a fashion similar to that proposed for the translocation of SP6 RNA polymerase around nucleosomes (56
Since our results suggest that ISWI has strand specificity, translocation would be expected to follow one DNA strand around the helical DNA backbone. A key consequence of this would be the generation of superhelical torsion between the translocase and the site at which DNA contacts the nucleosome surface. Consistent with this suggestion, the generation of superhelical torsion on chromatin templates by ISWI was previously detected (35
). This provides a second means by which DNA might be moved over the surface of the nucleosome. Local defects to the superhelical twist of DNA on the nucleosome might be transmitted incrementally over the surface of the nucleosome, as proposed previously (62
). This is also supported by the distortion of DNA in high-resolution structures of the nucleosome (17
). Where rotation of the DNA on the surface of the nucleosome is not favored, for example, in the presence of strong rotational phasing, a buildup of torsion might be capable of impeding nucleosome mobilization. Consistent with this, it has been observed that the ySWI/SNF complex is inefficient in remodeling nucleosomes on small minicircles that cannot readily accommodate changes in twist (28
). It has also recently been observed that the presence of nicks in a DNA template located at sites that could relieve torsion as it accumulates as a consequence of ISWI translocation stimulates the ability of ISWI to function in nucleosome mobilization (40
). This is also consistent with a role for torsion in remodeling.
Neither of these mechanisms are necessarily mutually exclusive, and it is even possible to imagine a combination of the two. It is possible that the structural characteristics of the DNA, especially its flexibility and inherent curvature, play an important role in determining what combination of the two mechanisms is involved. Whatever mechanism is involved, it remains unclear why ISWI appears unable to move nucleosomes from the ends of some DNA fragments such as the one used in this study. It appears to be a general property of the movement of nucleosomes on short DNA fragments that they tend to accumulate toward DNA ends (25
). One explanation for this has been that there is an energetic penalty associated with the crossover of DNA duplexes entering and/or exiting the nucleosome in close proximity that is absent for nucleosomes at the ends of DNA fragments (52
). However, this phenomenon has been characterized on only very few DNA fragments to date and we have observed cases where nucleosomes positioned at the ends of some DNA fragments can be relocated to other positions by ISWI or thermal incubation (A. Flaus, unpublished data). Thus, it is possible that the stability of end-positioned nucleosomes is a result of the combined energetics of DNA-sequence-dependent nucleosome positioning and the absence DNA duplex crossover at the entry and/or exit to the nucleosome such that some nucleosomes, like the one used here, strongly resist movement.
Although the ISWI protein provides an excellent model system to study mechanisms for ATP-dependent chromatin remodeling due to the fact it can be expressed as a recombinant protein and has ATPase activity that is targeted to a specific site on the histone octamer, it is likely that the means by which other ATP-dependent remodeling activities function may differ, at least to some extent. Indeed, the ISWI protein itself is normally found to be associated with other proteins that alter the way in which it moves mononucleosomes (19
). Thus, it will also be important to determine how the assembly of ISWI into native complexes alters its action in nucleosome mobilization.