|Home | About | Journals | Submit | Contact Us | Français|
The precise placement of nucleosomes has large regulatory effects on gene expression. Recent work suggests that nucleosome placement is regulated in part by the affinity of the underlying DNA sequence for the histone octamer. Nucleosome locations are also regulated by several different ATP-dependent chromatin remodeling enzymes. This raises the question whether DNA sequence influences the activity of chromatin remodeling enzymes. DNA sequence could most simply regulate nucleosome remodeling through its effect on nucleosome stability. In such a model, unstable nucleosomes would be remodeled faster than stable nucleosomes. It is also possible that certain DNA elements could regulate remodeling by inhibiting the interaction of nucleosomes with the remodeling enzyme. A third possibility is that DNA sequence could regulate the outcome of remodeling by influencing how reaction intermediates collapse into a particular set of stable nucleosomal positions. Here we dissect the contribution from these potential mechanisms to the activities of yeast RSC and human ACF, which are representative members of two major classes of remodeling complexes. We find that varying the histone-DNA affinity over three orders of magnitude has negligible effects on the rates of nucleosome remodeling and ATP hydrolysis by these two enzymes. This suggests that the rate-limiting step for nucleosome remodeling may not involve the disruption of histone-DNA contacts. We further find that a specific curved DNA element previously hypothesized to inhibit ACF activity does not inhibit substrate binding or remodeling by ACF. The element however, does influence the distribution of nucleosome positions generated by ACF. Our data support a model in which remodeling enzymes move nucleosomes to new locations by a general sequence-independent mechanism. However, consequent to the rate-limiting remodeling step, the local DNA sequence promotes a collapse of remodeling intermediates into highly resolved positions that are dictated by thermodynamic differences between adjacent positions.
The eukaryotic genome is packaged by wrapping ~147 bp units of DNA around histone octamers to form chains of nucleosomes. The packaging of the DNA within a nucleosome reduces access of the DNA to most transcription factors and polymerases. In between nucleosomes, there are regions of more accessible DNA, called linker regions that vary from a few base pairs to several hundred base pairs.1 Thus within any given cell type, the precise partitioning of the genome into nucleosome-bound and nucleosome-free DNA regions can have large consequences on gene regulation and help define a particular cellular state.2
Recent studies suggest that the genome plays a large role in encoding its own packaging through differences in affinity of the underlying sequence for the histone octamer 3 (Fig. 1a). Genome wide nucleosome mapping studies have found that ~80% of all yeast nucleosomes adopt specific positions. 1, 4, 5 The nucleosome positions are often predicted from the primary DNA sequence based on the ability of the DNA sequence to bend in a manner required for nucleosome formation.6, 7 Other work has mapped local changes in nucleosome positions in promoter regions across the yeast genome upon environmental stress such as heat shock.2 These changes in nucleosome positions alter the accessibility of promoter regions to transcription factors. Since such changes in nucleosome positions are generally thought to accompany cellular differentiation and adaptation, it is of interest to understand to what extent and how DNA sequence influences transitions between different chromatin states.
While nucleosomes can assemble on any sequence, under physiological conditions, most nucleosomes do not move on their own,8 but remain kinetically trapped in their locations. In vivo, transitions between chromatin states are catalyzed by ATP-dependent chromatin remodeling machines that rapidly reorganize histone-DNA contacts.9–12 Thus, understanding whether DNA sequence affects the activity of remodeling enzymes is a key aspect of understanding how DNA sequence might influence transitions between different chromatin states.
One way in which DNA sequence can regulate remodeling activity is by affecting the stability of a nucleosome. This model draws on previous observations that nucleosomes assembled on strong positioning sequences are thermally repositioned more slowly than nucleosomes on weak positioning sequences.8, 13, 14 If the rate-limiting transition state for remodeling primarily involves loosened histone-DNA contacts, then remodeling enzymes, like heat, will also move unstable nucleosomes more readily than stable nucleosomes. In such a mechanism, the main function of a remodeling enzyme would be to catalyze rapid equilibration between different nucleosome positions (Fig. 1b). This model predicts that remodeling enzymes will move nucleosomes occupying high affinity DNA sequences more slowly than nucleosomes occupying lower affinity DNA sequences. Another recently proposed possibility is that certain DNA sequences adopt topological properties within nucleosomes that inhibit binding by chromatin remodeling complexes and thereby help stabilize nucleosomes in specific locations.15 While some biochemical studies have shown that ATP-dependent remodeling enzymes can move nucleosomes away from strong nucleosome positioning sequences16–19 other studies suggest that nucleosome positioning properties of the DNA do play an important role in determining the locations of remodeled nucleosomes.14, 20 These observations have suggested a third model in which DNA sequence does not directly affect the activity of remodeling enzymes but instead more locally regulates the outcome of remodeling by influencing the collapse of remodeling intermediates into a particular set of stable nucleosomal positions.14
To distinguish amongst the three models, we separately measured the effects of nucleosome stability and DNA structure on nucleosome remodeling rates. We tested the activities of two major classes of ATP-dependent remodeling complexes, the ISWI class and the SWI/SNF class on nucleosomes with different stabilities. These two classes of complexes are highly abundant, act broadly at a number of different loci, and generate substantially different products. We therefore reasoned that results with these two classes would lead to generalizable principles of how DNA sequence influences ATP-dependent nucleosome reorganization. We chose yeast RSC as a representative member of the SWI/SNF class and human ACF as a representative member of the ISWI class. Surprisingly, we find that the rates of nucleosome movement by these complexes are insensitive to 1000-fold changes in histone-DNA affinity. These and other results presented here support the model that DNA sequence regulates ATP-dependent remodeling by locally influencing the final nucleosome positions adopted by remodeling intermediates.
To determine how DNA sequence regulates the activity of remodeling complexes, we took two complementary approaches. In the first approach we measured whether changing the stability of nucleosomes by changing DNA sequence affected the remodeling rates of ACF and RSC. In the second approach we tested whether the introduction of a specific curved DNA element inhibits remodeling by ACF.
We selected 4 different DNA sequences as our experimental inputs designated as 601,21 TPT,22 5S,23 and ARB.18 601 is a synthetically selected positioning sequence. It has amongst the highest known affinities for the histone octamer and is commonly used as an experimental tool in nucleosome-based assays because of its strong ability to position nucleosomes in vitro. TPT is another previously used, synthetically derived positioning sequence. 5S is a naturally occurring nucleosome positioning sequence that has commonly been used to position nucleosomes in vitro. ARB is a sequence arbitrarily chosen from a bacterial plasmid that we have used previously.18
We used two approaches to measure differences in affinity amongst the test sequences for the histone octamer (Fig. 2). In the first approach, we used a competitive nucleosome assembly method developed previously.21, 24–26 In this method, a tracer amount of a specific sequence competes with a large amount of a background DNA sequence for a limiting amount of histone octamer. The tracer sequence is fluorescently labeled while the background sequence is unlabeled. Comparing the fraction of tracer DNA incorporated into nucleosomes for the different sequences then allows calculation of the relative free energies for nucleosome assembly, (see methods for more details). We visualized the tracer DNA by end-labeling it with Cy3. To ensure we were measuring the energetics of a specific nucleosome position, our competition experiment was done on 147 bp core segments of DNA. Using this approach we found that our sequences vary in affinity for the histone octamer over several hundred fold, representing a 3.0 kcal/mol range in the free energy differences at 4 °C (Table 1, ΔΔGARB). The free energy difference that we obtain between 601 and 5S agrees well with previous measurements.27 Furthermore, the range of free energy differences covered by our test sequences is comparable to the maximal range seen in vitro8 and substantially larger than the variability in affinity observed in vivo.26
In the second approach, we measured differences in the propensity of the DNA sequences to transiently unravel from the histone octamer. We used this additional approach for the following reasons. Nucleosomal DNA can spend as much as 10% of its time away from the octamer and exposed to solution and it has therefore been proposed that capturing unraveled nucleosomal DNA may be an early step in ATP-dependent chromatin remodeling.28–30 Further, the free energy required to unravel the same amount of DNA can vary by as much as ~7.3 kcal/mol depending on sequence, raising the possibility that sequence may regulate remodeling through effects on DNA unraveling.31 As described previously, the equilibrium between the unraveled and bound states of DNA can be obtained from the rate of restriction enzyme cutting at a restriction site within a nucleosome,31 (see Materials and Methods). We engineered a PstI site at the same location (~18 bp in from one end) within all four test sequences. Further, as with the competitive assembly experiments, we assembled nucleosomes on 147 bp sequences. By comparing the rates of cutting of PstI across the four sequences in the we found that the free energy required to expose the PstI site varied by ~1000-fold (~4 kcal/mol at 37 °C) (Fig. 2cd & Table 1).
Nucleosomes assembled on these sequences represent quantitatively different energy barriers for disruption of histone-DNA interactions and were consequently used as tools to investigate the reaction mechanisms of chromatin remodeling enzymes ACF and RSC.
To measure the rates of remodeling we used a previously developed FRET-based technique18 that allows detection of nucleosome movement in real time (Fig. 3a). We assembled nucleosomes on the four test sequences with 60 base pairs of flanking DNA. The DNA was end-labeled with the FRET donor Cy3 and histone H2A was labeled with Cy5 at residue 120. This design gives the maximal FRET value when a nucleosome is positioned on one end of the DNA and any FRET decrease accompanying translational movements of up to 15 base pairs away from this end position is easily detectable.18 Due to the strong positioning power of 601, nucleosomes assemble readily on the end of the 601 sequence with 60 flanking base pairs, corresponding to the maximal FRET position. Since weaker affinity sequences are also inherently weaker positioning sequences, for the rest of the sequences, we enriched the end positioned nucleosomal population through glycerol gradient purification. The 5S sequence with 60 flanking base pairs, however re-equilibrates away from the end position on the time scale of purification and gives rise to a higher variability of the FRET signal due to heterogeneity in the starting material. Therefore, as described below, for nucleosomes assembled on the 5S sequence we cross-validated our FRET-based results using independent gel-based assays.
To isolate effects of nucleosome stability on the maximal rates of remodeling, the remodeling reactions were carried out with excess and saturating remodeling enzyme. These conditions were also chosen to mimic the in vivo conditions where the effective concentration of the remodeler is increased by specific recruitment.
On short stretches of DNA, ACF moves nucleosomes from end positions to centered positions that contain approximately equal flanking DNA on either side. 32 Using changes in FRET to measure ACF catalyzed nucleosome movement we find that ACF remodels all 4 nucleosomes with comparable rate constants (Fig. 3c). Thus the ACF nucleosome remodeling rate is insensitive to large changes in nucleosome stability. As described above, the initial positions of 5S nucleosomes are more heterogeneous and therefore result in a more variable FRET signal. To control for this variability, we also compared the rates of remodeling using a gel-based assay that allows direct visualization of the final centered product. Centered nucleosomes migrate more slowly through a polyacrylamide gel and appear as an accumulation of a higher band in the course of an ACF remodeling reaction.18 Time courses of remodeling showed similar rates of movement to the centered position for all nucleosome constructs (Fig. 3f, see figure legend for quantification). This result also showed that the final remodeled state for all sequences was dominated by the nucleosome centering mechanism of ACF.
RSC remodels nucleosomes by generating a variety of nucleosomal products that include repositioned nucleosomes and nucleosomes with altered structures.9, 33 Many of these remodeled products involve displacement of the nucleosome away from its starting position on the DNA are therefore expected to result in reduced FRET. We used changes in FRET to measure RSC catalyzed nucleosome movement and found that RSC remodels three of the constructs with comparable rate constants (Fig. 3b). The heterogeneity in the initial positions of 5S nucleosomes combined with the heterogeneity of RSC products prevented us from using FRET to accurately measure remodeling for the 5S nucleosome, presumably because some products resulted in increased FRET, while others resulted in decreased FRET. We therefore instead assembled nucleosomes on core segments of the 5S and the 601 sequences to ensure a unique starting position, and compared RSC remodeling using a gel-based restriction enzyme accessibility assay. It has previously been shown that at low concentrations of the restriction enzyme PstI (0.2 U/μl) a nucleosomal PstI site is cut very slowly in the absence of remodeling, but is cut up to 1000-fold faster in the presence of remodeling enzyme and ATP.34 In these assays, the rate of cutting by PstI reflects the rate of remodeling by the remodeling complex. Using this approach we found that the rate of remodeling was the same for the high affinity 601 as the ~3 kcal/mol lower affinity 5S sequence (Fig. 3b,e).
In summary, by using two independent assays we found that both ACF and RSC remodel all tested nucleosomes at similar rates. Significantly, the minor differences in the rates show no correlation with the relative stabilities of the nucleosomes (Fig. 3 & Table 1).
The observed insensitivity of remodeling rates to DNA sequence can arise if the remodeling enzymes compensate by hydrolyzing more ATP to remodel the more stable nucleosomes. To test this possibility, we measured the initial rates of ATP hydrolysis by ACF and RSC with each of the four nucleosomes. The nucleosomes were assembled on core 147 bp DNA segments and we used excess nucleosomes over remodeling enzymes. The magnitude of the differences in ATPase rates is small (at most 3 fold) and shows no correlation with nucleosome stability (Table 1). This indicates that RSC and ACF are able to reposition nucleosomes assembled on all test sequences using similar amounts of ATP.
Together, the above results indicate that DNA sequence does not regulate the activity of ATP-dependent chromatin remodeling complexes by changing nucleosome stability. We next investigated if a particular curved DNA sequence can inhibit remodeling activity.
It has recently been suggested that the activity of ACF may be regulated in a sequence specific manner. It was proposed that a particular 40 base pair stretch of DNA with a high degree of intrinsic curvature, when positioned over the nucleosomal dyad adopts a conformation on the nucleosome that is incompatible with ACF binding.15 In this model, the ACF catalyzed remodeling of nucleosomes containing this sequence results in an accumulation of nucleosomes with the 40 base pair segment aligned with the dyad because these nucleosomes cannot be bound by ACF.
To dissect out the effect of the 40 bp element on ACF remodeling, we designed a construct in which the 40 base pair curved element replaced the central 40 bp of the 601 sequence (Fig. 4a). Insertion at this location is theoretically predicted to maintain the continuous curvature present in the 601 sequence, (Supplementary Figure 1). Despite this alignment, insertion of the 40 bp element slightly increased the proportion of alternative nucleosome positions (Fig. 4b). However, the majority of the nucleosomes (>75% relative to 601) were still positioned at one end of the DNA corresponding to the starting substrate for the reaction (Fig. 4b and see Methods.). We first tested if the presence of this sequence affected the rate of remodeling by ACF under multiple turnover conditions. In these experiments, a limiting amount of ACF was used to remodel excess and saturating amounts of nucleosomes. These conditions reflect an in vivo situation where DNA sequence can influence recruitment. The two types of nucleosomes were labeled with different fluorescent dyes. ACF remodels both types of end-positioned nucleosomes with very similar rates (Fig. 4b, top panel, Fig. 4c, left panel). We obtain a similar result when the 40 bp element is placed within centered nucleosomes (Supplementary Figure 3). Together these results indicating that the presence of the 40 bp element within nucleosomes does not slow down remodeling by ACF once ACF is fully bound.
To investigate if the 40 bp element reduced binding by ACF as previously hypothesized we used the same set of nucleosomes in a competitive remodeling experiment. Equal concentrations of nucleosomes with and without the curved element were mixed with a limiting amount of ACF. 35 The different fluorescent labels on each type of nucleosome allowed us to visualize their remodeling separately (Fig. 4b). Under these competitive conditions, ACF should partition between the two nucleosomes based on their relative binding affinities. The observed remodeling rates for the two types of nucleosomes then reflect both the binding preference and remodeling preference that ACF has for either type of nucleosome. As ACF does not show a remodeling preference, (Fig. 4b, top two panels and Fig. 4c, left panel) any observed difference in rates under competitive conditions would reflect a difference in the binding preference of ACF. We find that ACF remodels both types of nucleosomes with similar rates under these competitive conditions. This indicates that the 40 bp element does not inhibit binding by ACF (Fig. 4c).
The above results raise the strong possibility that DNA mainly influences remodeling by influencing the collapse of remodeling intermediates into specific nucleosomal locations. We tested this possibility in the context of ACF because unlike RSC, ACF generates only canonical nucleosomes. This allows us to more clearly visualize any differences in final nucleosome positions. We also used a longer nucleosome construct for two related reasons: (i) it has previously been reported that local positioning differences in the products are not easily visualized by native gel when the DNA fragment is on the order of 200 bp14 and (ii) on DNA constructs shorter than 250 bp, the DNA length sensing mechanism of ACF is expected to drive nucleosomes to a narrowly clustered set of centrally positioned nucleosomes, as seen in Figure 3f and described previously.18
We compared the outcome of ACF remodeling on two 347 bp constructs. Both constructs have a 601 sequence at one end to localize the starting nucleosome position. Unlike the experiment in Figure 4, the 40 bp sequence is not inserted within the 601 sequence. In this case, one construct contains the 40 bp curved DNA element in the center of the 347 base pair sequence (Fig. 5a). Using this longer nucleosomal construct, we observe many different remodeled products that migrate more slowly than the starting nucleosome. The presence of the 40 bp element significantly changes the final distribution of products (Fig. 5b&c). The differences in outcome are not due to an incomplete reaction, as the product distributions did not change with longer times (data not shown).
The data in Figure 5 is analogous to previous observations that the 40 bp element alters the final locations of the nucleosomes remodeled by ACF.15 Previously this ability of the 40 bp element to change the positions of remodeled nucleosomes was interpreted as arising from an inhibitory effect on ACF binding. 15 The results from Fig. 4 rule out an effect of the 40 bp element on ACF binding and together with the results from Fig. 5 and thermodynamic modeling of the resulting positions (Supplementary Figure 2) suggest an alternative model. By this alternative model (described in the Discussion), DNA sequence does not directly affect remodeling activity but rather helps direct the rotational setting adopted by a remodeled nucleosome once it is released by the remodeling enzyme.
ATP-dependent chromatin remodeling enzymes change the accessibility of DNA and alter transcriptional programs by directing nucleosomes to new positions. Here we analyze how the underlying DNA sequence influences ATP-driven transitions between different chromatin states. As it had already been shown that histone-DNA affinity can alter the rates of thermal repositioning,8, 13 we first considered the simple hypothesis that increasing the affinity between the histones and DNA would also slow down nucleosome repositioning by remodeling enzymes. By comparing the remodeling activities of two major classes of remodeling complexes on a series of nucleosomal templates controlled for DNA length, we found that variation of the affinity of histones for DNA did not affect the rates of remodeling. We next addressed whether the presence of a DNA sequence with strong intrinsic curvature properties could affect remodeling activity. It had previously been reported that a particular highly curved 40 bp sequence, when present within a nucleosome, inhibits binding by ACF and thereby helps stabilize nucleosomes in specific locations. By directly measuring ACF activity on nucleosomes containing this sequence, we found that the 40 bp sequence does not affect nucleosome remodeling or binding by ACF. The sequence however, does affect the preferred local positions adopted by the remodeled nucleosomes consistent with previous observations.15 We discuss the mechanistic and broader biological implications of these findings below.
Our observation that varying the strength of the binding interface between the histones and nucleosomal DNA does not affect the rates of remodeling most simply suggests that disrupting histone-DNA interactions is not rate-limiting during the overall remodeling reaction.49 Previous work has shown that both the ISWI class and SWI/SNF class of chromatin remodeling enzymes can translocate on DNA36, 37 and it has been proposed that this activity is critical for displacing the DNA from histones. It is therefore possible that the energetically most costly and consequently the slowest step in chromatin remodeling is translocation of the remodeling enzyme on nucleosomal DNA. It is also possible that a conformational rearrangement in the remodeling enzyme required for activation is the slowest step.
Alternatively, the slowest step may still involve the breaking of histone-DNA contacts, but this step may be directly coupled to ATP hydrolysis and therefore limited by the rate of ATP hydrolysis. In this case the energy of ATP hydrolysis would be used to lower the activation energy for breaking histone-DNA bonds. For example, hydrolysis of ATP may promote a conformational change in the remodeling enzyme that allows the enzyme to stably capture nucleosomal DNA that unpeels from the histone octamer. Hydrolysis of one ATP is sufficient to completely dissociate half of the DNA contacts on even the strongest affinity sequence used here. 30, 31 Furthermore, it has been speculated that the step size of ATP-dependent chromatin remodeling complexes is of the order of 10 bp, 18, 19, 38 in which case, the energy from one ATP hydrolysis event, if used efficiently, will be more than sufficient.
The data presented in this work leads to the question of how the activity of the enzyme can remain unaffected by changes in DNA sequence while the outcome of the remodeling reaction is altered.14 (Fig. 5) A sequence-dependent outcome in the context of sequence-independent activity can be explained if DNA sequence influences the collapse of a high-energy intermediate generated by enzyme action into highly resolved nucleosome positions (Fig. 6a). Our model conceptually divides the remodeling reaction in two components – local guidance based on sequence interaction with nucleosome versus global (or long-range) guidance by the remodeler. We first present a model based on the physical properties of the nucleosome for how DNA sequence can locally direct nucleosome locations (Fig. 6b–d) and then discuss implications for the regulation of the enzyme within an in vivo context.
When assembled into a nucleosome DNA adopts a high degree of curvature relative to the persistence length of DNA. The energetic cost of loop formation is balanced by the energy gained through forming 12 discrete contacts between histones and the DNA backbone.39 There are no base pair specific contacts between the histones and the DNA, so the base pair composition of a sequence is thought to influence affinity for the nucleosome only to the extent that it dictates its propensity and ability to curve around the histone octamer.6,7
Since the DNA molecule has a helical structure with a ~10 bp repeat, the groove facing the nucleosome alters every 5 bp (Fig. 6b). Correspondingly, it has been shown that sequences which contain 10 bp spaced repeats of AT and GC dinucleotides with 5 bp relative offsets preferentially bind nucleosomes. This is because every 10 bp AT/TT/TA dinucleotides provide a favorable local structure for compression of the minor groove, while a 5 bp offset of GC dinucleotides lowers the energetic cost of compression of the corresponding major groove when it faces the nucleosome core.21, 24, 25 Such an arrangement of dinucleotides confers a handedness to the bend of the DNA molecule such that the direction of curvature of the DNA is aligned with the curvature of the histone octamer.
Consider the effect of a 10 bp translation of such a sequence relative to the histone octamer on the overall free energy of the complex. Given the ~10 bp helical periodicity, such a translation would retain the alignment between the handedness of the DNA curvature and the octamer while maintaining the discrete points of contact. So, the change in overall affinity would be mostly attributable to the difference of the contributions from the departing and arriving 10 bp. In contrast, a translation of such a sequence by half a helical turn (5 bp) would disrupt the correspondence between the preferred helical handeness of the DNA and the curvature of the nucleosome, resulting in a highly unstable position.
These local features of an energy landscape are represented in Figure 6d for an idealized sequence of continuously distributed intrinsic curvature. This representation of the thermodynamic landscape is consistent with two properties of nucleosome positions in vivo; that many nucleosomes occupy highly resolved positions, and that nucleosomes tend to cluster at 10 bp offsets.4 A smooth energy landscape as in Figure 6c would correspond to a smooth probability distribution of nucleosome locations. In contrast, by analogy to the combination of positive and negative feedback to achieve higher precision in many biological control systems, the close interspersion of high and low affinity rotational registers (Fig. 6d) can achieve higher positioning precision. The distribution of bending properties may be periodic, as schematized in Figure 6d, or non-periodic. A non-periodic distribution of bending properties within a sequence can give rise to irregular thermodynamic landscapes, with varying well depths and barrier heights. This is suggested for the 40 bp element, the presence of which, introduces non-uniform changes in the predicted thermodynamic landscape of nucleosome positions (Supplementary Figure 2).
We hypothesize that the height of the peak that separates two wells representing adjacent rotational settings (Figure 6d) reflects the activation energy required for the nucleosome to switch between the two settings. This hypothesis is based on the assumption that switching requires transition through the intermediate rotational setting represented by the peak separating the two adjacent wells. When the energetic difference between the peak and wells is less than thermal energy (<kT~0.6 kcal/mol), this would allow for rapid equilibration between adjacent rotational settings and the relative depths of the wells will determine the distribution of rotational settings. In contrast, when difference between the peak and wells is kT, equilibration between adjacent rotational settings would be slow on a physiological time-scale, and nucleosomes would get kinetically trapped in certain rotational settings. Thus, in addition to deep wells, high peaks can also result in more sharply defined rotational positions and low peaks in combination with shallow wells will result in a wider distribution of rotational positions (Figure 6d). By this model, a strong rotational positioning sequence may be defined, not just by its absolute affinity, but also by the height of the energy barriers between neighboring rotational positions, (Fig. 6d).
Indeed, previous theoretical calculations suggest that the energetic barrier between neighboring nucleosome positions separated by 5 bp can span a large range depending on sequence. Calculations for K. lactis centromeric DNA sequences suggest that this local barrier can be as large as 15 kcal/mol41 while similar calculations for human telomeric DNA sequences suggest that the barrier is negligible compared to thermal energy.42
We have shown that the thermodynamic stability of nucleosomes does not regulate the efficiency of two key chromatin remodeling enzymes, RSC and ACF. Rather, our data with ACF suggest that the intrinsic remodeling mechanism dictates long-range translational shifts while the local nucleosome positioning landscape resolves the final position and rotational phase. Indeed, substantial previous work has shown that, on the same DNA sequence, different remodeling enzymes relocate nucleosomes to distinct translational positions as a result of intrinsic mechanistic differences. 9,11,14 We hypothesize that consequent to remodeling, the DNA sequence promotes the collapse of a high-energy remodeled intermediate into a particular set of stable nucleosomal positions. In vivo, the available positions will be further restricted by neighboring nucleosomes, other DNA bound factors, and the state of chromatin compaction.
The capacity of chromatin remodeling enzymes to remodel sequences of variable affinity at similar rates casts them as versatile molecular machines capable of preserving their specific function once recruited to any locus. This versatility with respect to the nucleosomal DNA sequence is consistent with substantial previous work suggesting that a large part of the specificity of remodeling enzymes arises from targeting via sequence-specific DNA binding factors and factors that recognize specific histone modifications.43,50 Significantly, ACF and RSC act globally at a diverse set of genomic loci dependent on cellular state, and participate in different functional pathways. Moreover, a key homolog of ACF in yeast, the Isw2 complex has been observed to directionally slide a nucleosome in vivo across low affinity dA-dT-rich sites shown to repel nucleosomes.44
Many nucleosomes occupy highly defined and reproducible positions in vivo, either at the level of single base pairs, or at the level of 10 bp rotational phases. While such 10 bp phasing could simply reflect the stochastic consequence of the dominant mode of curvature distribution within natural positioning sequences, recent work implies that the precise rotational position of the nucleosome has functional significance. This argument is based on observations that shifts in nucleosome positions by a few base pairs can have profound impact on transcription45 and by the observation that transcription factors show preferential alignments on the surface of the nucleosome with helical periodicity.4 Intriguingly, nucleosome bound sequences within regulatory regions exhibit selection against mutations that disrupt helical phase.51
Previous work has already revealed the potential of a sequence dependant nucleosome positioning code. Our study suggests a mechanism by which the remodeler and sequence can cooperate to switch nucleosomes between two different locally meaningful positions based on environmental stimuli. We hypothesize that after providing the energy to overcome local kinetic barriers between neighboring positions, the remodeler releases the nucleosome in the general vicinity of a regulatory region while the underlying DNA sequence directs the settling of the nucleosome into a functional rotational position. This collaboration between remodeling enzymes, which provide range, and DNA sequence, which provides precision, may allow nucleosomes to toggle between well-defined cellular state dependant positions like clicks of a switch.
RSC was purified from Saccharomyces cerevisiae via a TAP-tag attached to the Rsc2 subunit as described previously.46 The human ACF complex was purified assembled and purified from Sf9 cells via a FLAG tag attached to the SNF2h subunit as previously described.18
The 601, TPT, 5S and ARB positioning sequences were modified to contain a PstI site 18 bp in from one end. DNA constructs of different lengths were generated by PCR and gel-purified. Specifically, for the FRET kinetics measurements, ACF gel mobility shifts, and ATPase experiments, test sequences included 60 base pairs of flanking DNA to accommodate movement of the nucleosome. The experiment in Figure 5 required a longer construct with 200 flanking bp and other experiments (Fig. 2, ,3c)3c) were done on 147 bp sequences without any flanking DNA, as described in the text and figures. Cy3 or Cy5 labeled DNA was generated by PCR using end-labeled primers (IDT). (Primer sequences are available upon request.) DNA fragments were assembled into mononucleosomes with recombinant Xenopus histones using the gradient salt dialysis method.47 Nucleosomes used in the FRET kinetics experiment were assembled on octamers containing H2A labeled with Cy5 at residue 120, as described. 18 For 5S, TPT and ARB positioning sequence containing nucleosomes, we enriched for end positioned nucleosomes by glycerol-gradient purification.47
In this method developed by Shrader and Crothers24, 25 and further optimized by the Widom group,21, 26 labeled tracer DNA competes with a large excess of unlabeled competitor DNA for limiting amounts of histone octamer. The referenced method was used with slight modification. Tracer amounts of Cy3 labeled 147 bp core DNA fragments of each test sequence were mixed with 40 μg of ARB core DNA serving as background. Both were competing for 3 μg of octamer in a 60 μl reaction containing: 20 mM Tris-H Cl (pH 7.7), 10 mM DTT, 0.5 mM benzamidine. The reaction was gradually dialyzed at 4 °C from high salt 2M NaCl into TE buffer using an established assembly method shown to yield equilibrium measurements.21, 26 Reconstitution reactions were repeated multiple times on different days. Measurements were further validated by repetition with different background DNA (30 μg 5S, not shown.). The assembled reactions were run on a 0.5X TBE non- denaturing 5% (v/v) polyacrylamide gels. Tracer DNA was visualized by scanning on a Typhoon Variable Mode Imager (GE health care) with the Cy3 filter set, while total DNA was similarly visualized using SYBR gold staining (Invitrogen). ImageQuant (GE Healthcare) was used to quantify band intensities.
For a competitive assembly reaction with a given tracer sequence represented by “D”, an equilibrium constant (KeqD) was calculated from the molar ratio of nucleosomes to free DNA. A equilibrium constant was also obtained using the background sequence, ARB, as a tracer (KeqARB). The relative free energies were then obtained from Equation 1 for 4 °C.
For consistency the ΔΔG values calculated from the site exposure values are also reported relative to ARB.
Proper equilibration was indicated by the value of KeqARB. This value is obtained under conditions where the tracer and background sequences are identical and therefore should be equal to the molar ratio of octamer to background DNA. This ratio should also be equal for all reactions when total DNA is quantified. The variability of this ratio across all test conditions within a single dialysis experiment was used to calculate global experimental error (± 0.05 kcal/mol) and can be interpreted as the uncertainty of the ARB sequence, otherwise defined as 0, (see Table 1).
FRET-based remodeling reactions contained either 25 nM ACF or 40 nM RSC. All reactions contained 10 nM nucleosomes (assembled on various sequences as reported) in reaction buffer: 20 mM Tris-H Cl (pH 7.7), 60 mM K Cl, 10% glycerol and 3 mM free Mg2+. Reactions were initiated by the addition of 4 μM ATP at 30 °C. Both RSC and ACF activity under saturating ATP conditions is on a time scale too fast to be accurately measured by our FRET-based method. We therefore used a subsaturating ATP concentration of 4 μM to be able to follow the entire time course. FRET-based kinetics data was collected on an ISS K2 fluorimeter. 18 The data were fit to two exponentials using MATLAB (Mathworks). The rate constants reported in Table 1 are for the first fast phase. The second slower phase is thought to represent a small population of nucleosomes that get remodeled more slowly as explained previously. 18 Values reported in table are averaged across n>3 conditions, with error reported as SEM from a pool of individual runs with typical R values of >.99 to the fit.
All the ACF reactions analyzed by gel mobility shift were performed in the same basic buffer conditions as above with variations in ACF, nucleosome and ATP concentrations as described below. All reactions were stopped with 2X stop buffer (115 mM ADP, 0.8 mg/ml unrelated stop plasmid DNA to compete off the enzyme) and run on a 0.5X TBE non- denaturing 5% (v/v) polyacrylamide gel. The ACF reactants and products were visualized by scanning the gel on a Typhoon Variable Mode Imager (GE healthcare) after SYBR gold staining (Invitrogen). The reactions in Figure 3f were carried out with the same ACF and ATP concentrations as used in the FRET experiments. The reactions in Figure 5b contained 25 nM ACF, 10 nM nucleosomes and 2 mM ATP-Mg2+. We confirmed that these reactions had gone to completion by longer incubations and by controlling for time-dependent loss of enzyme activity and ATP depletion.
For limiting enzyme conditions used in the competition experiments (Fig. 4b) 15 nM ACF and a saturating concentration of 80 nM total nucleosome were used based on previous work 35 (and data not shown.) Three reactions were run in parallel. The individual reactions (Fig 4b, top two panels respectively) contained either (1) 80 nM of nucleosomes assembled on 601+60 sequence end-labeled with Cy3, or (2) 80 nM of nucleosomes assembled on 601+60 sequence containing the 40 bp curved insert in the center of 601 and end-labeled with Cy5 dye. The third condition (3) combined 40 nM of each nucleosome for a final concentration of 80 nM. The nucleosomes were visualized by scanning the gels on the Typhoon using the appropriate filters to visualize either the Cy3 or the Cy5 dyes. ImageQuant (GE Healthcare) was used to quantify band intensities, which were used to calculate the fraction of end positioned (i.e, unremodeled) nucleosomes relative to all positions as a function of time and these were then normalized to time zero to obtain the normalized fraction of unremodeled nucleosomes [fr(U)]. The data in Figure 4C represents [1-fr(U)] plotted vs. time. Reported results are representative across three experiments conducted on different days.
We adapted a previous restriction enzyme accessibility approach pioneered by the Widom laboratory.7, 31 We measured the initial rates of cutting at a PstI site located 18 bp in from the end of different core nucleosomes. Restriction enzymes can cut nucleosomal DNA when it transiently unravels from the histone octamer. The rate of cutting by the highest achievable restriction enzyme concentrations is significantly slower than the rate at which the unraveled DNA rebinds the histone octamer.48 As a result, cutting of nucleosomal restriction enzymes is slowed down relative to cutting of free DNA by a factor equal to the equilibrium constant for unraveling the amount of DNA required to expose the restriction site. As described previously, we obtained the equilibrium constants for unraveling nucleosomal DNA (Keqconf) by taking the ratio of the rate of cutting on a nucleosome to the rate of cutting on the corresponding free DNA sequence. The values in Table 1 are ΔΔG values calculated for 30 °C from differences between the given sequence and ARB for consistency and ease of comparison with the competition assembly values (Eq. 2).
Reactions were initiated by adding 10 U/μl PstI to 20 nM nucleosomes assembled on core segments of test sequences in a reaction buffer containing: 20 mM Tris-H Cl (pH 7.7), 60 mM K Cl, 5% glycerol and 3 mM free Mg2+ at 37°C Aliquots were removed at various times and quenched in 1.5 volumes of 10% glycerol, 70 mM EDTA, 20 mM Tris (pH 7.7), 2% SDS, 0.2 mg/ml xylene cyanole and bromophenol blue. Cut DNA was separated from uncut DNA on a 12% native polyacrylamide gel (1XTBE) after deproteinizing at 37°C for 1 hr with 1 mg/ml proteinase K. The DNA was visualized using a Typhoon after staining with SYBR gold. For each time point, the fraction of uncut versus cut DNA was quantified. Initial rates were obtained by fitting the first 10% of the reaction to a straight line.
Reactions containing 40 nM RSC, 10 nM nucleosomes, 3mM free Mg2+ at 30°C, and 0.2 U/μl PstI were initiated by the addition of 2 mM ATP-Mg2+. Reactions were stopped, separated, visualized and quantified as in the above section. Rate constants were obtained by applying a first order fit to the change in the fraction of uncut DNA as a function of time using MATLAB based on a previously established method.34
For reactions with ACF, 20 nM enzyme and 80 nM nucleosomes were mixed in 20 mM Tris-H Cl (pH 7.7), 60 mM K Cl, 10% glycerol and 3 mM free Mg2+. RSC reactions contained 1nM RSC and 40 nM nucleosome in the same buffer. Experiments were initiated by the addition of 4 μM ATP containing trace amounts of [g-32P]ATP. Reactions were quenched, processed and quantified as described.34
We thank J. Widom for the 601 plasmid, J.J. Hayes for the 5S plasmid, C. Rowe for purified RSC and L. Racki for purified ACF enzyme. We thank J. Widom, S. Lomvardas and members of the Narlikar laboratory for helpful comments on the manuscript.
FUNDING This work was supported by Grants from the NIH and Beckman Foundation to G.J.N. P.D.P. is supported by an NSF Graduate Research Fellowship.
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.