NSD3-Short Is an Adaptor Protein that Couples BRD4 to the CHD8 Chromatin Remodeler

Abstract

The bromodomain and extraterminal (BET) protein BRD4 is a therapeutic target in acute myeloid leukemia (AML). Here, we demonstrate that the AML maintenance function of BRD4 requires its interaction with NSD3, which belongs to a subfamily of H3K36 methyltransferases. Unexpectedly, AML cells were found to only require a short isoform of NSD3 that lacks the methyltransferase domain. We show that NSD3-short is an adaptor protein that sustains leukemia by linking BRD4 to the CHD8 chromatin remodeler, by using a PWWP chromatin reader module, and by employing an acidic transactivation domain. Genetic targeting of NSD3 or CHD8 mimics the phenotypic and transcriptional effects of BRD4 inhibition. Furthermore, BRD4, NSD3, and CHD8 colocalize across the AML genome, and each is released from super-enhancer regions upon chemical inhibition of BET bromodomains. These findings suggest that BET inhibitors exert therapeutic effects in leukemia by evicting BRD4-NSD3-CHD8 complexes from chromatin to suppress transcription.

Figure 1. A short isoform of NSD3 lacking catalytic function is essential in AML and binds directly to the BRD4 extraterminal domain

(A) Domain architectures of human BRD4 and NSD3. BDI: bromodomain I, BDII: bromodomain II, ET: extraterminal domain, CTD: C-terminal domain. PWWP: Pro-Trp-Trp-Pro chromatin reader module. PHD: plant homeodomain finger. AWS: Associated with SET domain region, SET: Su(var)3–9, enhancer-of-zeste and trithorax domain which catalyzes H3K36 methylation. A region unique to NSD3-short is represented by a black rectangle. (B) Competition-based assay in RN2 cells evaluating effects of the indicated LMN shRNAs (which express GFP) on cell proliferation. Each horizontal bar represents the average fold-decrease in GFP percentage for an independent shRNA over 10 days in culture. n=3. (C) Immunoprecipitation followed by Western blotting performed with the indicated antibodies. The nuclear lysate was prepared from the human AML cell line NOMO-1. IP: immunoprecipitation, IgG: isotype control immunoglobulin. Note: a background band appears in the control IP at ∼70 kDa, which is near the NSD3-short band. Shown is a representative experiment of three independent biological replicates. (D) (top) Competition-based assay to evaluate the effect of the indicated LMN shRNAs on RN2 cell proliferation. GFP percentages were normalized to day 2 (d2) measurements. shNSD3.1653 targets both long and short NSD3 isoforms and shNSD3L.2963 targets exclusively the long isoform. (bottom) Western blotting analysis of whole cell lysates prepared from RN2 cells following doxycycline (dox)-induced knockdown of NSD3. RN2 cells were transduced with TRMPV-Neo shRNAs followed by dox treatment for 48 hours. A representative experiment of three biological replicates is shown. (E–F) Competition-based assay evaluating the effect of the human NSD3-short cDNA (which is not recognized by the murine shRNAs and is expressed with the PIG vector linked to GFP) on the proliferation arrest induced by NSD3 shRNAs (expressed using the LMN-mCherry vector). Results were normalized to the d2 percentage of GFP+mCherry+ cells. Results shown are the average of three biological replicates. (G) FLAG-BRD4 ET domain pulldown assays evaluating interactions with the indicated GST-NSD3 fragments. FLAG-BRD4 ET domain was expressed in HEK293T followed by immobilization on anti-FLAG agarose beads and extensive washing. ET immobilized beads were then incubated with purified GST-NSD3 fragments expressed in E. coli. A representative experiment of three biological replicates is shown. (H) Recombinant protein purity as assessed by SDS-PAGE and Coomassie blue staining. Strep₂SUMO-BRD4 ET and Strep₂SUMO-NSD3 100–263 proteins were individually expressed in Sf9 cells and purified by affinity, ion exchange, and size exclusion chromatography. The final purity of these proteins was assessed by SDS-PAGE. Molecular weight markers are shown for reference. (I) Binding affinity of the BRD4 ET-NSD3 interaction as determined by surface plasmon resonance using purified proteins. Steady-state response values from the association phase of each injection were plotted as a function of analyte concentration and fit to determine the K_D. Points indicate the average of three technical replicates with error bars representing the standard deviation. (J) Diagram of NSD3-short fragments evaluated in the BRD4 ET pulldown assay. All error bars shown except for (I) represent SEM. See also Figure S1.

Figure 2. NSD3-short is an adaptor protein that links BRD4 to the CHD8 chromatin remodeling enzyme

(A) Mass spectrometry analysis of proteins identified using anti-FLAG IP performed with nuclear lysates prepared from HEK293T cells transfected with FLAG-NSD3-short or empty vector (for mock IP). The list was ranked by the total number of matched peptides recovered. Complete results are provided in Table S1. (B) Competition-based assay in RN2 cells evaluating the effect of LMN shRNAs targeting the indicated proteins. Each bar represents the average fold-decrease in the percentage of GFP+ cells over 8 days for individual shRNAs. n=3. (C) CRISPR-scan of Chd8 with all possible sgRNAs. Deep sequencing based measurement of the impact of 907 Chd8 sgRNAs on the proliferation of Cas9-expressing RN2 cells. The location of each sgRNA relative to the CHD8 protein is indicated along the x axis. Shown is a representative experiment of two biological replicates. (D) IP-Western blotting performed with anti-FLAG antibodies and nuclear lysates prepared from RN2 cells stably expressing the indicated FLAG-NSD3 constructs or empty vector. (E) Endogenous IP-Western blotting performed with the indicated antibodies and nuclear lysates prepared from NOMO-1 cells. (F) The amino acid sequence of the human BRD4 ET domain indicating the surface residues that were subjected to mutagenesis. Combinations of mutations were used in some cases in an attempt to disrupt specific clusters of charged residues. (G) IP of the indicated FLAG-BRD4 ET domains expressed transiently in HEK293T cells followed by Western blotting with the indicated antibodies. (H) The molecular surface of the BRD4 ET domain (PDB: 2JNS) with hydrophobicity indicated in green (Lin et al., 2008). (I) (top) Luciferase reporter assay evaluating the activation function of the indicated GAL4-ET domain fusions on a minimal plasmid-based reporter harboring GAL4 recognition sequences. HEK293T cells were co-transfected with p9xGAL4-UAS-luciferase (firefly) reporter and the indicated GAL4 fusion expression plasmids expressing Renilla luciferase from a constitutive promoter. Plots indicate firefly luciferase activity normalized to the Renilla luciferase control. n=3. *p<0.05, two-tailed Student’s t-test. (bottom) Western blotting analysis of HEK293T cells transfected with the indicated plasmids shown in the top panel. All error bars represent SEM and all IP-Western and Western blotting experiments shown are a representative experiment of at least three independent biological replicates. See also Figure S2.

Figure 3. NSD3-short utilizes four distinct interaction surfaces to sustain AML cell proliferation

(A) Western blotting analysis of whole cell lysates prepared from RN2 cells transduced with the indicated PIG retroviral expression constructs. A representative experiment of three independent biological replicates is shown. (B–G) Competition-based assay tracking the abundance of GFP+mCherry+ cells during culturing of transduced RN2 cells. GFP is linked to the indicated cDNA and mCherry is linked to the indicated LMN shRNA. Plotted is the average of three independent biological replicates, normalized to d2. (H) Luciferase reporter assay with the indicated GAL4 fusions, performed as described in Figure 2I. n=3. ***p<0.001, two-tailed Student’s t-test. (I) Western blotting analysis of HEK293T cells transfected with the indicated plasmids. A representative experiment of three biological replicates is shown. (J) Diagram of the functionally important surfaces of NSD3-short. All error bars in this figure represent SEM. See also Figure S3.

Figure 4. Targeting of NSD3 or CHD8 leads to differentiation of AML cells and suppression of BRD4-dependent gene expression

(A–B) (top) Competition-based assays to evaluate the effect of NSD3 or CHD8 LMN shRNAs on RN2 cell proliferation. GFP percentages are normalized to d2 measurements. All error bars represent the SEM of three independent biological replicates, (bottom) Western blotting analysis of whole cell lysates prepared from RN2 cells transduced with the indicated TRMPV-Neo constructs following 48 hours of dox treatment. A representative experiment of three biological replicates is shown. (C–D) Flow cytometry analysis of c-kit and Mac-1 stained RN2 cells following TRMPV-Neo shRNA induction with dox for 96 hours. Gating was performed on dsRed+/shRNA+ cells. A representative experiment of three biological replicates is shown. (E–F) Light microscopy of May-Grünwald/Giemsa-stained RN2 cells expressing the indicated NSD3 shRNAs or Chd8 sgRNAs, in the presence or absence of ectopic c-Myc expression. For sgRNA experiments, an RN2 line stably expressing Cas9 was used. shRNA expression was induced using the TRMPV-Neo vector treated with dox for 4 days. For F, cells were imaged 6 days following transduction with the indicated LRG sgRNAs. Imaging was performed with a 40x objective. A representative image of three independent biological replicates is shown. (G–H) Gene Set Enrichment Analysis (GSEA) of RNA-seq data obtained from RN2 cells expressing NSD3 TRMPV-Neo shRNAs (induced with dox for 48 hours) or RN2-Cas9 cells expressing Chd8 LRG sgRNAs (4 days following transduction). Two independent NSD3 shRNAs or Chd8 sgRNAs were compared to a Ren.713 shRNA or Rosa26 sgRNA, respectively in this analysis. NES: Normalized enrichment score. For each of the indicated gene sets shown, the false discovery rate and nominal p-value were <0.01. See also Figure S4 and Table S3.

Figure 5. Genomewide colocalization of BRD4, NSD3, and CHD8 at active promoters and enhancers across the AML genome

(A) Density plot analysis comparing ChIP-seq datasets obtained using the indicated antibodies. Indicated is a 20 kilobase interval surrounding 1,950 BRD4-occupied promoters and 3,185 BRD4-occupied enhancers, identified previously as high-confidence BRD4 occupied sites (Roe et al., 2015). H3K27ac and H3K4me3 datasets from RN2 were described previously (Shi et al., 2013a). Each row represents a single peak. (B–E) ChIP-seq occupancy profiles with the indicated antibodies at various loci. The y-axis reflects the number of cumulative tag counts in the vicinity of each region. Validated transcript models from the mm9 genome assembly are depicted below. The asterisks indicate non-coding RNAs. See also Figure S5.

Figure 6. BRD4 recruits NSD3 and CHD8 to the Myc +1.7 Mb super-enhancer region

(A–C) ChIP-qPCR analysis at the E1-E5 Myc super-enhancer region evaluating the occupancy of the indicated FLAG-NSD3-short constructs using anti-FLAG antibody or control IgG. (D–F) ChIP-qPCR analysis with the indicated antibodies in RN2 cells treated with DMSO vehicle or 500 nM JQ1 for 6 hours. (G–L) ChIP-qPCR analysis with the indicated antibodies in RN2 cells transduced with the indicated TRMPV-Neo shRNA constructs and treated with dox for 48 hours. All error bars represent the SEM of three independent biological replicates. *p<0.05, **p<0.01, two-tailed Student’s t-test. See also Figure S6 and Table S4.