CHD4 slides nucleosomes by decoupling entry- and exit-side DNA translocation

Abstract

Chromatin remodellers hydrolyse ATP to move nucleosomal DNA against histone octamers. The mechanism, however, is only partially resolved, and it is unclear if it is conserved among the four remodeller families. Here we use single-molecule assays to examine the mechanism of action of CHD4, which is part of the least well understood family. We demonstrate that the binding energy for CHD4-nucleosome complex formation-even in the absence of nucleotide-triggers significant conformational changes in DNA at the entry side, effectively priming the system for remodelling. During remodelling, flanking DNA enters the nucleosome in a continuous, gradual manner but exits in concerted 4-6 base-pair steps. This decoupling of entry- and exit-side translocation suggests that ATP-driven movement of entry-side DNA builds up strain inside the nucleosome that is subsequently released at the exit side by DNA expulsion. Based on our work and previous studies, we propose a mechanism for nucleosome sliding.

Fig. 1. CHD1/CHD4 domain topology and nucleosome repositioning assays.

a Schematic showing the 15 superhelical location (SHL) sites, the most outward-facing positions of the minor groove, in the 601 nucleosome positioning sequence (labelled +7 to –7). The histone–DNA interface consists mainly of inward-facing sections of the DNA minor groove, which are defined as superhelical locations (SHL) ± 0.5–6.5. Orange bands indicate four phased TpA dinucleotides that are spaced 10 bp apart. b Domain architectures of yeast CHD1 and human CHD4 with residues at domain boundaries indicated (NTD: N-terminal domain, PHD: plant homeodomain, CHCT: CHD1 C-terminal domain). c Gel-based nucleosome repositioning assays carried out with the indicated nucleosomes and remodellers. Fluorescently labelled nucleosomes were treated with the indicated remodeller for 60 min, the reaction was stopped by adding dsDNA (33 µg/mL) and then the samples were run on 5% native polyacrylamide gels. Symmetrically positioned nucleosomes are retarded relative to asymmetrically positioned species, as indicated. d Gel-based nucleosome repositioning assays carried out as described in (c), except that an incubation time-course was carried out at a single CHD4 concentration (5 nM). Assays using 0w60 (upper panel) or 30w30 (lower panel) substrates are shown; a 0w60 control lane was included in the lower panel. Source data are provided as a Source Data file. e, f Nucleosome sliding assays carried out as described in (c), using the indicated nucleosome substrates and 5 nM CHD4. Remodelled products (0w30 and 30w0) are indicated by red arrows, and the possible hexosome band is indicated by black arrows. Source data are provided as a Source Data file.

Fig. 2. The affinity of CHD4 for DNA is not strongly dependent on the presence of flanking extra-nucleosomal DNA.

a Pulldowns in which FLAG-CHD4 is immobilized on FLAG-Sepharose beads and incubated with the indicated nucleosomes, followed by elution with FLAG peptide. Nucleosome input, unbound nucleosome/flow through (FT), final wash (FW) and elutions of each nucleosome construct were analysed by SDS-PAGE, along with a size indicator, Mark12 Standard (M). The negative control contains no nucleosome in the input. b EMSAs showing the binding of CHD4 to the indicated nucleosomes. Nucleosome concentration was 90 nM (0w0) or 60 nM (0w30 and 0w60). Open and filled triangles indicate complexes that probably contain one or two CHD4 molecules, respectively.

Fig. 3. A single-molecule FRET (smFRET) assay for nucleosome sliding shows that CHD4 drives multi-base-pair movements of nucleosomal DNA at the exit side.

a Diagram depicting the setup for the smFRET assay and nucleosome substrate before and after remodelling by CHD4. A 0^AF555w60^Bio nucleosome containing either proximal or distal or both AF647-labelled H2A is assembled onto a PEGylated coverslip via biotin at the longer end of the flanking DNA (AF555 and AF647 are represented by blue and red circles, respectively). FRET is then monitored as a function of time under different conditions. b Pre-reaction distribution of nucleosomal FRET states for 0^AF555w60^Bio nucleosomes. Low, mid- and high-FRET states correspond to particles containing H2A in the distal, proximal or both positions (relative to the DNA-bound AF555), respectively. c FRET vs time trace of 0^AF555w60^Bio nucleosome bearing a proximal H2A label. Donor AF555 fluorescence (green), acceptor AF647 fluorescence (magenta) and FRET (blue) are shown. d FRET vs time traces of 0^AF555w60^Bio nucleosome bearing a proximal H2A label in the presence of 2 nM CHD4 (top) or both 2 nM CHD4 and 1 mM ATP (bottom). Two clear drops in FRET are observed in the latter. We define the pause time t_pause as the duration between two FRET changes. e Calibration of FRET values for n^AF555w60^Bio nucleosomes. The FRET of proximally labelled particles was measured as a function of the number of base pairs (n) added to linker DNA at the exit site and mid-FRET peak values for each construct were obtained by fitting to a Gaussian distribution. Plotting the change in mid-FRET value as a function of n yielded a slope of –0.051 ± 0.002. Error bars represent standard deviation of the fit from at least two independent measurements. f Distribution of the 1st and 2nd step sizes for 40 molecules undergoing remodelling in presence of 2 nM CHD4 and 1 mM ATP. The histograms are fitted to a Gaussian distribution.

Fig. 4. CHD4-mediated nucleosome sliding is processive and depends on the binding of multiple ATP molecules.

a Distribution of t_pause times for remodelling of 0^AF555w60^Bio at the indicated CHD4 concentrations and 20 mM ATP. Error bars represent standard deviation (of 45, 74 and 56 independent single particles at 0.2, 2 and 20 nM CHD4, respectively). b Typical traces from smFRET assays carried out with 2 nM CHD4 and at the indicated ATP concentrations. c Histograms showing t_pause distributions of 70–82 particles from the experiments shown in (b). The histograms are fitted to a Gaussian distribution. d Pause time histograms for experiments carried out at 10 μM ATP (top) or 5 mM ATP (bottom) are overlaid with gamma distributions depicting different numbers of fundamental reaction steps (N = 1–5). e A 1:1 binding isotherm fit of the mean t_pause time as a function of ATP concentration, with data taken from the assays in (c). Error bars represent standard deviation.

Fig. 5. CHD4 binding induces changes in extra-nucleosomal DNA at the TA-poor side.

a Schematic showing the labelling scheme for 0w(9^AF555)60^Bio and nucleosomal conformations before and after CHD4 remodelling. b FRET vs time traces of 0w(9^AF555)60^Bio alone with a proximal (top) and distal (bottom) AF647-labelled H2A. c FRET vs time trace of 0w(9^AF555)60^Bio bearing a distal AF647 label, showing a gradual increase upon the addition of 2 nM CHD4. This newly established structure can be relatively stable (left) or can be transient, dropping back to the initial state (right). The proportions of each scenario (from 74 molecules) are illustrated as a pie chart. d FRET vs time trace of 0w(9^AF555)60^Bio bearing a distal AF647 label, showing an increase upon the addition of 2 nM CHD4 and 1 mM AMP-PNP. e FRET vs time traces of 0w(9^AF555)60^Bio bearing a distal AF647 label showed a greater increase during CHD4 remodelling in presence of ATP, comparing to the changes induced by CHD4 alone.

Fig. 6. CHD4 induces similar FRET changes in extra-nucleosomal DNA at the TA-rich side.

a Schematic showing the labelling scheme for 9^AF555w60^Bio and the two possible remodelling directions after treating with CHD4, depending on whether the ‘9’ end acts as an entry or an exit side. b FRET vs time traces for 9^AF555w60^Bio alone with a proximal (top) or distal (bottom) H2A AF647 label. c FRET vs time trace for 9^AF555w60^Bio bearing a distal AF647 label, showing an increase upon the addition of 2 nM CHD4. This newly established structure can be relatively stable (left) or can be transient, dropping back to the initial state (right). The proportions of each scenario (from 78 particles) are illustrated as a pie chart. d FRET vs time trace for distally labelled 9^AF555w60^Bio in presence of both 2 nM CHD4 and 1 mM ATP, showing an increase to a similar level to that of CHD4 alone. e FRET vs time trace for proximally labelled 9^AF555w60^Bio, showing a stepwise drop upon treatment with 2 nM CHD4 and 1 mM ATP. In this case, the AF555 tag at the ‘9’ end is moving away from the histone octamer, in contrast to the direction of movement observed in (d).

Fig. 7. A model for CHD4-driven nucleosome sliding.

a Schematic of one-half of a nucleosomal DNA sequence. SHL positions are indicated. b Binding of the two-lobed ATPase domain of CHD4 (not drawn in scale) induces a 1-bp shift in the so-called tracking strand of the DNA, creating a distortion that reaches from the SHL2 site all the way back to the 5′ end of the tracking strand. The DNA segment moved by CHD4-induced remodelling is coloured in red in this and the following figures. c Binding of ATP induces a conformational change in CHD4, ‘closing’ the two lobes around the ATP. This change results in the guide strand ‘catching up’ to the tracking strand and then the movement of both strands by ~1 bp past CHD4, so that a small bulge (or other irregularity) forms in the region of the dyad. d Hydrolysis of ATP drives a return of CHD4 to the open conformation, inducing a second 1-bp movement of the tracking strand, analogous to that in part b, and initiating a second cycle of the same process. e Following four further cycles of ATP binding and hydrolysis, a large irregularity is built up near the dyad. f The strain induced by this irregularity causes a concerted rearrangement of the DNA such that 5 bp are expelled from the exit side. This whole cycle can in principle be repeated many times.