Molecular architecture of nucleosome remodeling and deacetylase sub-complexes by integrative structure determination

Abstract

The nucleosome remodeling and deacetylase (NuRD) complex is a chromatin-modifying assembly that regulates gene expression and DNA damage repair. Despite its importance, limited structural information describing the complete NuRD complex is available and a detailed understanding of its mechanism is therefore lacking. Drawing on information from SEC-MALLS, DIA-MS, XLMS, negative-stain EM, X-ray crystallography, NMR spectroscopy, secondary structure predictions, and homology models, we applied Bayesian integrative structure determination to investigate the molecular architecture of three NuRD sub-complexes: MTA1-HDAC1-RBBP4, MTA1^N -HDAC1-MBD3^GATAD2CC , and MTA1-HDAC1-RBBP4-MBD3-GATAD2A [nucleosome deacetylase (NuDe)]. The integrative structures were corroborated by examining independent crosslinks, cryo-EM maps, biochemical assays, known cancer-associated mutations, and structure predictions from AlphaFold. The robustness of the models was assessed by jack-knifing. Localization of the full-length MBD3, which connects the deacetylase and chromatin remodeling modules in NuRD, has not previously been possible; our models indicate two different locations for MBD3, suggesting a mechanism by which MBD3 in the presence of GATAD2A asymmetrically bridges the two modules in NuRD. Further, our models uncovered three previously unrecognized subunit interfaces in NuDe: HDAC1^C -MTA1^BAH , MTA1^BAH -MBD3^MBD , and HDAC1^60-100 -MBD3^MBD . Our approach also allowed us to localize regions of unknown structure, such as HDAC1^C and MBD3^IDR , thereby resulting in the most complete and robustly cross-validated structural characterization of these NuRD sub-complexes so far.

Keywords: Bayesian integrative structure determination; XLMS; chromatin remodeling complexes; cryo-EM; histone modification; integrative modeling; nucleosome remodeling and deacetylase complex.

FIGURE 1

Subunits in nucleosome remodeling and deacetylase (NuRD) sub‐complexes. (a) Sequences and isoforms of NuRD subunits modeled in this study are shown and domains are labeled. Only one paralogue of each subunit is shown. Domains are shown in progressively dark shades along the sequence for MTA1, HDAC1, and MBD3. Regions for which atomic structures exist, or can be predicted, are represented by rectangles whereas regions without known structure are represented by beads. PDB IDs are shown for existing subunit structures and templates of homology models. Orange and gray dashed rectangles represent MTA1^N and MTA1^mid respectively. Numbering is for the human proteins. (b) Stoichiometries of modeled sub‐complexes and the endogenous NuRD complex. (c) Previously published atomic structures that were used for modeling. PDB codes are given in (a). (d) Crosslinks used in this study, represented as a CIRCOS (http://cx‐circos.net/) plot. The gray, blue, and green lines represent all the BS3/DSS, ADH, and DMTMM crosslinks respectively.

FIGURE 2

Integrative structure determination of nucleosome remodeling and deacetylase (NuRD) sub‐complexes. Schematic describing the workflow for integrative structure determination of NuRD sub‐complexes. The first row describes the input information. The second‐row details how data are used to encode spatial restraints. The third row describes the sampling method, and the last two rows illustrate the analysis and validation protocol. The background colors of the input information show the stage of modeling in which the information is used, as shown in the legend at the top.

FIGURE 3

Integrative model of the MTA1‐HDAC1‐RBBP4 (MHR) complex. (a) Representative bead model from the most populated cluster of integrative models for the MHR complex, shown with the MHR EM map. The model is colored by subunit. For MTA1, the two copies are shown in different colors (brown and orange) in panels (a) and (c), to illustrate the crossover. The HDAC1 active site is shown in red. (b) Localization probability density maps showing the position of different domains/subunits in the cluster. The map specifies the probability of any volume element being occupied by a domain in the ensemble of superposed models from the cluster. The domain densities are colored according to Figure 1. These maps are contoured at ~10% of their respective maximum voxel values. (c) Representative bead model from panel (a) with regions of known structure shown in ribbon representation. (d) CX‐CIRCOS (http://cx‐circos.net/) plot for crosslink satisfaction on the ensemble of MHR models from the major cluster. Gray (red) lines indicate satisfied (violated) crosslinks in panels (c) and (d). (e) Schematic representation of the integrative model of the MHR complex. See also Figure 1 and Figures S2 and S5

FIGURE 4

Integrative model of the MTA1^N‐HDAC1‐MBD3^GATAD2CC (MHM) complex. (a) Representative bead model from the major cluster of analyzed integrative models for the MHM complex, with the corresponding EM map (EMD‐21382), colored by subunit. The domains of the two MBD3s are shown in shades of pink and green, respectively. (b) Localization probability density maps showing the position of different domains in the ensemble of models from the cluster. The domain densities are colored according to Figure 1. (c) The same density maps as (b) (front view), showing the two MBDs in pink and green, respectively, and illustrating that they localize differently on the MTA1‐HDAC1 dimer. The density maps of MTA1^mid and GATAD2^cc were omitted for clarity. (d) The density maps of the two MBD3^IDR domains on the MTA1‐HDAC1 dimer. Most of the maps are contoured at around 20% of their respective maximum voxel values (except MTA1^165–333 at 10% and GATAD2^cc at 27%). (c) Representative bead model from panel (a) with regions of known structure shown in ribbon representation. (d) CX‐CIRCOS (http://cx‐circos.net/) plot for crosslinks satisfaction on the ensemble of MHM models from the major cluster. Gray (red) lines indicate satisfied (violated) crosslinks in panels (c) and (d). (e) Schematic representation of the integrative model of the MHM complex. Note that MTA1^mid in this model corresponds to MTA1^334–431. See also Figure 1 and Figures S3 and S6

FIGURE 5

Integrative model of the nucleosome deacetylase (NuDe) complex. (a) Representative bead model from the dominant cluster of integrative models for the NuDe complex, with the corresponding EM map (EMD‐22904), colored by subunit. (b) Localization probability density maps showing the position of different domains in the ensemble of models from the cluster. The domain densities are colored according to Figure 1. Maps are contoured at ~10% of their respective maximum voxel values (except GATAD2^CC at 20%). (c) Representative bead model from panel (a) with regions of known structure shown in ribbon representation. (d) CX‐CIRCOS (http://cx‐circos.net/) plot for crosslinks satisfaction on the ensemble of NuDe models from the major cluster. Gray (red) lines indicate satisfied (violated) crosslinks in panels (c) and (d). (e) Schematic representation of the integrative model of the NuDe complex. See also Figure 1 and Figures S4, S7, and S10

FIGURE 6

COSMIC mutations mapped onto the NuDe integrative model. Somatic pathogenic point mutations from the COSMIC database mapped onto the representative bead model of the NuDe complex (Figure 5a). (a) Mutations of residues that map to previously undescribed protein–protein interfaces within our model. (b) Mutations on residues that map to exposed binding sites between modeled proteins and known binding partners. A bead is colored according to the maximum number of mutations on any residue in the bead, according to the legend. Representative mutations are labeled in both (a) and (b). See also Table S1 and Figure S12

FIGURE 7

Integrative model of NuDe complex with the nucleosome. The CHD4‐nucleosome structure is placed in the cleft of the NuDe integrative model. The regions with known atomic structure are shown in the NuDe integrative model from Figure 5a. CHD4, histones, DNA and the corresponding NuDe subunit residues they are proposed to bind to, are depicted in the same color, as given by the legend.

FIGURE 8

Model of MBD3 binding in nucleosome remodeling and deacetylase (NuRD) The figure shows two states of MBD3 in NuRD. (a) In the first state, the MTA1 dimerization interface is accessible for MBD3^IDR to bind. (b) In the second state, upon binding, MBD3 recruits GATAD2A and the chromatin remodeling module and shifts to one end of the MTA1‐HDAC1 dimer. GATAD2A localizes near MTA1dimer, precluding a second MBD3 from binding to it.