The chromodomains of the Chd1 chromatin remodeler regulate DNA access to the ATPase motor

Abstract

Chromatin remodelers are ATP-driven machines that assemble, slide, and remove nucleosomes from DNA, but how the ATPase motors of remodelers are regulated is poorly understood. Here we show that the double chromodomain unit of the Chd1 remodeler blocks DNA binding and activation of the ATPase motor in the absence of nucleosome substrates. The Chd1 crystal structure reveals that an acidic helix joining the chromodomains can pack against a DNA-binding surface of the ATPase motor. Disruption of the chromodomain-ATPase interface prevents discrimination between nucleosomes and naked DNA and reduces the reliance on the histone H4 tail for nucleosome sliding. We propose that the chromodomains allow Chd1 to distinguish between nucleosomes and naked DNA by physically gating access to the ATPase motor, and we hypothesize that related ATPase motors may employ a similar strategy to discriminate among DNA-containing substrates.

Figure 1. Overview of the S. cerevisiae Chd1 remodeler structure

(A) Schematic of Chd1 domain organization and residue boundaries for the crystallization construct (S. cerevisiae numbering). (B) Two views of the remodeler domain organization. The left view shows how the two ATPase lobes are flanked by the double chromodomains (yellow) and the extended C-terminal bridge (green). The right view emphasizes the meandering path of the C-terminal bridge from the second ATPase lobe back to the first ATPase lobe. This and all subsequent molecular images were made using PyMOL (http://www.pymol.org/). (C) Alignment of the C-terminal bridge segment for several Chd1 and Iswi orthologs. This and other sequence alignments were produced using ClustalX (Larkin et al., 2007) and formatted with TEXshade (Beitz, 2000). See also Figure S1 and Movie S2.

Figure 2. Comparison with the SF2 helicase Vasa reveals that the ATPase cleft of Chd1 is opened and not properly organized for ATP hydrolysis

(A) Crystal structure of the Vasa RNA helicase (Sengoku et al., 2006); PDB code 2DB3). The ATP analog AMP-PNP in the ATPase cleft is shown as gray spheres. Helicase motif VI on the second ATPase lobe is colored green, with two residues considered to serve as arginine fingers, R579 and R582, shown as sticks and magenta spheres. (B) The ATPase motor of Chd1 (this study; PDB code 3MWY). The coloring is similar to (A), with the ATP analog ATPγS shown as gray spheres, the region corresponding to helicase motif VI (residues 798–809) colored green, and the Cα positions of two arginine residues corresponding to the Vasa arginine fingers shown as magenta spheres. Swi2/Snf2-specific inserts on the second ATPase lobe are colored gray. (C) A schematic diagram illustrating the more opened ATPase cleft of Chd1 compared to Vasa. Transformation of the Chd1 motor to match the tightly packed organization of Vasa would require a closure of the ATPase cleft by 52°. Unlike the closed, hydrolysis competent structure observed for Vasa, this opened cleft of Chd1 does not permit the two arginines of motif VI to directly contact the phosphate tail of the bound nucleotide. See also Figure S2.

Figure 3. An acidic helix on the double chromodomains contacts the ATPase motor at a predicted DNA-binding site

(A, B, and C) Electrostatic surface representations of the double chromodomains (A), the ATPase motor (B), and a close-up of the contact between the helical linker of the chromodomains and the second ATPase lobe, with the Cα atoms of acidic residues shown as red spheres (C). The electrostatic surface potentials were calculated using APBS (Baker et al., 2001), and shown in the range of ±5.0 k_BT/e with the negative and positive electrostatic potentials shown as red and blue surfaces, respectively. (D) Sequence alignment of the acidic helix of the chromo-wedge that contacts the second ATPase lobe in the crystal structure. (E) A view of the predicted DNA binding surface on the second ATPase lobe. The structural cores of three other SF2 ATPases bound to their nucleic acid substrates (Vasa:RNA, PDB code 2DB3, magenta (Sengoku et al., 2006); Hel308:DNA, PDB code 2P6R, green (Büttner et al., 2007); NS3:DNA, PDB code 2F55, blue (Kim et al., 1998)) were superimposed on the second ATPase lobe of Chd1. Only the nucleic acid substrates are shown. The surface of the second ATPase lobe that is within 5 Å of the acidic chromodomain helix is shown as an orange footprint.

Figure 4. The wild-type chromodomain-ATPase interface is required for substrate discrimination

(A) Schematic of the S. cerevisiae crystal structure, highlighting residues targeted for mutagenesis. (B) Schematic representations of S. cerevisiae constructs used for biochemical analysis. (C and D) ATPase activities in the presence of buffer alone, naked DNA, or mononucleosome substrates, measured using an NADH-coupled assay (Kiianitsa et al., 2003). A 206 bp DNA fragment containing the core 601 nucleosome positioning sequence at one end was used for the DNA alone and mononucleosome substrates. All experiments were performed three or more times and shown as means with standard errors. See also Figure S3.

Figure 5. Disruption of the chromodomain-ATPase interface enhances binding of DNA to the ATPase motor

Comparison of DNA binding abilities of wildtype and variant Chd1_142–939 proteins (lacking the DNA-binding domain) with a FAM-labeled 16 bp DNA duplex using native PAGE. Protein concentrations were 1.7, 7, 28, 110 µM, and labeled DNA was 25 nM. See also Figure S4.

Figure 6. The chromodomain-ATPase interface both positively and negatively influences nucleosome sliding activity

Mononucleosome sliding assays for Chd1 proteins with wildtype and mutated chromodomain-ATPase interfacial residues. Starting with end-positioned (0-601-60) mononucleosomes, this assay reports on nucleosome centering as an up-shift of the nucleosome bands. (A) End-positioned mononucleosomes (12.5 nM) were incubated with Chd1_ΔN and Chd1-Δchromo proteins for 60 minutes, or Chd1_142–939 proteins for 180 minutes and resolved by native PAGE. Protein concentrations of Chd1_ΔN and Chd1-Δchromo proteins were 0.1, 1, 10, and 100 nM, and concentrations of Chd1_142–939 were 0.01, 0.1, 1, and 10 µM. Data are representative of experiments performed three or more times. (B –E) Disruption of the chromodomain-ATPase interface partially relieves requirement for the histone H4 tail. Both wildtype (B) and H4Δtail (D) recombinant S. cerevisiae histone octamers assembled on differently labeled 0-601-60 DNA fragments (25 nM each) were mixed and incubated with 5 nM Chd1 proteins for the indicated times. Quantification of nucleosome sliding shown in (B) and (D) is shown in (C) and (E), respectively. The percent shifted was calculated as the loss in intensity of the bottom band (end positioned nucleosome) relative to all nucleosome bands in the lane. Data are representative of experiments performed four or more times. (F and G) Chd1-Δchromo is sensitive to the absence of the H4 tail. Chd1-Δchromo protein (100 nM) was incubated with wildtype and H4Δtail nucleosomes as (B) and (D) above. Quantification of nucleosome sliding (G) was performed as (C) and (E) above. The 40 minute time points show averages of two measurements, and all other time points show the averages and standard deviations of three measurements.

Figure 7. A model for the regulation of chromatin remodeling by the Chd1 chromodomains

In this model, the chromodomains can occupy an inhibitory (gated) position that prevents activation of the ATPase motor. Interaction with a nucleosome relieves this inhibition by stabilizing the chromodomains in an ungated state that allows the ATPase motor to achieve a closed, hydrolysis competent conformation. Subsequent ATP hydrolysis by the motor promotes nucleosome sliding.