The Nucleosome Remodeling and Deacetylation Complex Modulates Chromatin Structure at Sites of Active Transcription to Fine-Tune Gene Expression

Abstract

Chromatin remodeling complexes play essential roles in metazoan development through widespread control of gene expression, but the precise molecular mechanisms by which they do this in vivo remain ill defined. Using an inducible system with fine temporal resolution, we show that the nucleosome remodeling and deacetylation (NuRD) complex controls chromatin architecture and the protein binding repertoire at regulatory regions during cell state transitions. This is primarily exerted through its nucleosome remodeling activity while deacetylation at H3K27 follows changes in gene expression. Additionally, NuRD activity influences association of RNA polymerase II at transcription start sites and subsequent nascent transcript production, thereby guiding the establishment of lineage-appropriate transcriptional programs. These findings provide a detailed molecular picture of genome-wide modulation of lineage-specific transcription by an essential chromatin remodeling complex as well as insight into the orchestration of molecular events involved in transcriptional transitions in vivo. VIDEO ABSTRACT.

Keywords: Mediator; NuRD; RNA polymerase II; chromatin; embryonic stem cells; enhancer; transcription; transcription factor.

Graphical abstract

Figure 1

Mbd3 Induction Restores NuRD Activity to Mbd3-Null ESCs (A) Model of the induction system: Mbd3-null ESCs (left) contain ER-Mbd3b-ER (green diamonds, ER-M3b-ER) in the cytoplasm. Upon addition of tamoxifen, ER-M3b-ER enters the nucleus. (B) Nuclear extracts from Mbd3-inducible ESCs at different times after tamoxifen addition or from wild-type ESCs (WT) were probed with antibodies indicated at right. In the α-Mbd3 panel, the location of the ER-Mbd3b-ER transgene is indicated with an arrow, as are the locations of endogenous Mbd3 isoforms in WT cells. Sin3A and Pcna act as loading controls. Protein sizes are shown at left in kilodaltons. (C) Chd4 was immunoprecipitated from nuclear protein extracts across tamoxifen induction time course and probed with antibodies indicated at right. IgG, immunoglobulin G (IgG) control; Input, 10% input; IP, Chd4 immunoprecipitation. Protein sizes are shown at left in kilodaltons. (D) ChIP-qPCR for Chd4 (blue lines), Mbd3 (black lines), and IgG (gray lines) across the promoter and an enhancer of NuRD target gene Ppp2r2c at 0 and 24 hr of tamoxifen treatment. x axes show locations relative to the annotated transcription start site of Ppp2r2c. N ≥ 3 biological replicates. See also Figure S2C. (E) ChIP-qPCR at the peak of Mbd3 ChIP signal in (D) plotted across a time course of tamoxifen addition. Significant enrichment of Mbd3 relative to no tamoxifen occurs from 15 min onward (^∗∗∗∗p < 0.0001, ^∗∗p < 0.01, and ^∗p < 0.05 using a two-tailed t test). N ≥ 3 biological replicates. (F) Input-normalized ChIP signal across WT Mbd3 peaks plotted for ER-ChIP in uninduced cells (left) or after 48 hr of tamoxifen treatment (middle) and for Mbd3-ChIP in WT ESCs (right). Mean signal is plotted across the top. (G) qRT-PCR using nascent RNA of indicated genes over 48 hr of the time course of tamoxifen addition (mean of relative expression ± SEM; plotted relative to time 0). Tamoxifen was added to either the Mbd3 inducible line (in red) or to the parental, Mbd3-null line as a control (in black). N ≥ 3 biological replicates. (H) Unsupervised clustering of nascent RNA-seq during the Mbd3 induction time course. Significant changes (|FC| > 2; p < 0.05) in transcript level compared to time 0 are shown by gray bars in the left-hand panels. Triplicate samples were prepared and sequenced for each time point. See also Figures S1, S2, and S3.

Figure 2

H3K27Ac Levels Do Not Immediately Correlate with NuRD-Induced Gene Expression Changes (A) H3K27Ac ChIP-seq data normalized to H3 ChIP-seq at indicated time points of tamoxifen addition are plotted across a metagene for genes showing significant reduction in mRNA levels within 8 hr (“repressed early”) or only after 24–48 hr (“repressed late”). Data are plotted as mean ± 95% confidence intervals (N = 3 biological replicates). (B) As in (A) but for genes showing significant increases early (≤8 hr) or late (24–48 hr). (C) ChIP-qPCR for H3K27Ac plotted relative to H3 ChIP across the Ppp2r2c and Htra1 transcription start sites for 0, 24, and 48 hr following tamoxifen addition. Mean ± SEM; N ≥ 3 biological replicates. (Below) Data at the peak of enrichment from the panels above are displayed across the time course of tamoxifen exposure. Mean ± SEM (^∗∗p < 0.01 and ^∗p < 0.05 relative to 0 hr). N = 2–5 biological replicates. (D) ChIP-seq for H3K27Ac normalized to H3 ChIP for each time point of Mbd3 induction centered at peaks of NuRD binding ± 2 Kb, classified into active and inactive enhancers, promoters, or other sequences. Mean signal for each category is plotted across the top.

Figure 3

NuRD Controls Nucleosome Density at Regulatory Sequences (A and B) MNase-seq data collected after 0 or 30 min or 24 hr of tamoxifen treatment are plotted (A) across peaks of enrichment for indicated proteins defined in WT cells or (B) across indicated features (n = number of peaks/feature). Data are plotted as mean ± 95% confidence intervals. p values were calculated for each time point relative to time 0 for the center of the feature indicated with a dashed line. ^∗∗∗p < 0.001. N = 3 biological replicates. (C and D) MNase-qPCR data (mean ± SEM) for the 0- and 24-hr time points plotted across an enhancer and the TSS for Ppp2r2c (C) and for Bmp4 (D). The x axis indicates Kb relative to the annotated TSS for each gene. Overlaid in red are ChIP-qPCR data for Mbd3-ER at 24 hr post-tamoxifen addition. The blue arrows indicate the positions further analyzed in panels at right, where the horizontal line shows the mean. A schematic of each gene is shown above the ChIP-qPCR panels. Filled and open boxes represent coding and non-coding exons, respectively, and the red box below the line indicates the position of the relevant enhancer. N ≥ 6 biological replicates. ^∗∗∗∗p < 0.0001, ^∗∗∗p < 0.001, ^∗∗p < 0.01, and ^∗p < 0.05 relative to 0 hr using a two-tailed t test.

Figure 4

Nucleosome Remodeling Is Chd4 Dependent (A) Timeline of the experiment. DOX was added to induce expression of the WT or mutant Chd4 cDNA 18 hr prior to tamoxifen addition, which induced nuclear translocation of Mbd3-ER. (B) Nuclear extracts from ESCs expressing either the WT or ATPase mutant (Mut) Chd4 transgene (Chd4 TG) with and without DOX treatment were probed with an anti-Chd4 (left) or an anti-FLAG antibody (right). “No TG,” ESCs with no Chd4-transgene. Lamin B1 serves as a loading control. Protein sizes are given at left in kilodaltons. (C) MNase qPCR as in Figures 3C and 3D for indicated locations in the Ppp2r2c enhancer and Bmp4 promoter in ESCs expressing the WT Chd4 transgene (left), the ATPase mutant Chd4 transgene (middle), and in cells in which expression of the ATPase mutant Chd4 was not induced with DOX (right). ^∗∗∗∗p < 0.0001, ^∗∗∗p < 0.001, ^∗∗p < 0.01, and ^∗p < 0.05 relative to 0 hr using a two-tailed t test. N = 4 or 5 biological replicates. (D) Expression data of nascent RNA measured by qRT-PCR for Ppp2r2c or Bmp4 across the tamoxifen induction time course in ESCs expressing the WT (Chd4 WT) or ATPase mutant (Chd4 ATPase mutant) CHD4 are plotted over time. N = 6–9 biological replicates. See also Figure S4.

Figure 5

NuRD Controls Transcription Factor Access to Chromatin (A) Protein binding at indicated time points across ChIP-seq peaks defined in WT cells (Nanog and Klf4) or at the most significant peaks identified at the 24-hr time point in the in the Mbd3 inducible cell line (Med12). Data are plotted as mean ± 95% confidence intervals. The later time points were compared with 0 hr, and significant differences in mean are indicated as ^∗∗p < 0.01 or ^∗∗∗p < 0.001 using a two-tailed t test. (B) Protein binding or MNase protection are plotted across sites defined as having significantly increased binding in Mbd3-null ESCs versus WT cells (Nanog and Klf4) or were called as peaks at 0 hr, but not at 24 hr, in the Mbd3 induction time course (Med12; higher in KO) or the opposite (higher in WT) for each protein. The later time points were compared with 0 hr, and significant differences in mean are indicated as ^∗p < 0.05 or ^∗∗∗p < 0.001 using a two-tailed t test. (C) ChIP-qPCR for indicated proteins at the peak of binding for each feature (Figure S5B) at the indicated Ppp2r2c or Bmp4 enhancers across the time course of tamoxifen exposure. A schematic of each gene is shown above the ChIP-qPCR panels. Time after tamoxifen addition in hours is indicated along the x axis. Mean ± SEM is plotted for all points. ^∗∗p < 0.01 and ^∗p < 0.05 relative to 0 hr using a two-tailed t test. See also Figure S5.

Figure 6

NuRD Induction Results in Loss of RNA Polymerase II from Transcription Start Sites and a Transient Reduction in Nascent RNAs (A) MNase-seq reads plotted as in Figure 3A across all TSS or just those showing reduced (down genes) or increased (up genes) expression after 48 hr of tamoxifen treatment. p values were calculated as in Figure 3A. ^∗∗∗p < 0.001. (B) ChIP-seq signal for RNA polymerase II (S5P) plotted as in (A). (C) Schematic of traveling ratio calculation. The blue line represents S5P ChIP-seq signal across a paused gene (TR = 0.75), whereas the red line shows signal across an unpaused gene (TR = 0.5). TSS, transcription start site; TTS, “transcription termination site” (defined as polyA addition site). (D) Cumulative traveling ratio calculated from RNA polymerase II (S5P) ChIP-seq data at time 0 (dark purple), 30 min (0.5H; light purple), 4 hr (green), and 24 hr (blue). CDF, cumulative density function. p values are given for each time point compared to time 0. (E) Nascent RNA-seq plotted as in (A). Only data for the + strand are plotted as confidence intervals. N = 3 biological replicates. See also Figure S6.

Figure 7

Inappropriate Enhancer Chromatin Remodeling and Protein Binding in Mbd3-Null ESCs during Lineage Commitment (A) Expression of Klf4, Zfp42, and Otx2 over a differentiation time course relative to that in WT cells in 2iL conditions. Mean ± SEM; N ≥ 3 biological replicates. (B) ChIP-qPCR for Mbd3-FLAG in WT (black) and Mbd3-null (magenta) ESCs across indicated enhancer sequences in 2iL (self-renewing conditions [SR]) or after 24 hr in differentiation conditions (24h Diff). Mean ± SEM; N ≥ 4 biological replicates. (C) MNase-qPCR profiles across enhancers associated with Klf4, Zfp42, and Otx2 are plotted for WT (black) or Mbd3-null ESCs (blue) in SR and after 24h Diff. Mean ± SEM; N ≥ 3 biological replicates. (D) ChIP-qPCR for Med12 and Med1 in WT (black) or Mbd3-null ESCs (magenta) before and after 24 hr in differentiation media. Data are plotted relative to levels in WT at time 0. Mean ± SEM; N ≥ 3 biological replicates. (E) ChIP-qPCR for either total (NTD) or serine 5 phosphorylated (S5P) RNA polymerase II as in (D). Data are plotted relative to time 0. Mean ± SEM; N ≥ 3 biological replicates. (F) Model of how NuRD controls transcription. In the absence of Mbd3 (top left), NuRD does not form and regulatory sequences for a hypothetical gene adopt a specific nucleosome structure (gray spheres), are bound by transcription factors (TFs), and associate loosely with Mediator and with RNA polymerase II at the start of a gene (gray arrow). Adding back Mbd3 results in rapid NuRD reformation, an increase in nucleosome density, and eviction of TFs, Mediator, and RNA polymerase I from the regulatory sequences in the first 30 min (top middle). By 24 hr, the NuRD-dictated nucleosome structure has adopted a new transcription factor repertoire and is more stably associated with Mediator. This is the WT situation (top right). When cells are induced to differentiate, NuRD holds the nucleosome structure in place while Mediator is excluded in preparation for a change in transcriptional output (bottom right). In the absence of Mbd3 (bottom left), the differentiation signal results in a shift in nucleosome structure that does not exclude Mediator, making it less likely that the required change in transcription will occur.