Recurrent SMARCB1 Mutations Reveal a Nucleosome Acidic Patch Interaction Site That Potentiates mSWI/SNF Complex Chromatin Remodeling

Abstract

Mammalian switch/sucrose non-fermentable (mSWI/SNF) complexes are multi-component machines that remodel chromatin architecture. Dissection of the subunit- and domain-specific contributions to complex activities is needed to advance mechanistic understanding. Here, we examine the molecular, structural, and genome-wide regulatory consequences of recurrent, single-residue mutations in the putative coiled-coil C-terminal domain (CTD) of the SMARCB1 (BAF47) subunit, which cause the intellectual disability disorder Coffin-Siris syndrome (CSS), and are recurrently found in cancers. We find that the SMARCB1 CTD contains a basic α helix that binds directly to the nucleosome acidic patch and that all CSS-associated mutations disrupt this binding. Furthermore, these mutations abrogate mSWI/SNF-mediated nucleosome remodeling activity and enhancer DNA accessibility without changes in genome-wide complex localization. Finally, heterozygous CSS-associated SMARCB1 mutations result in dominant gene regulatory and morphologic changes during iPSC-neuronal differentiation. These studies unmask an evolutionarily conserved structural role for the SMARCB1 CTD that is perturbed in human disease.

Keywords: ATP-dependent chromatin remodeling; BAF complex; Coffin-Siris syndrome; SMARCB1 (BAF47); chromatin accessibility; intellectual disability; mammalian SWI/SNF complexes; nucleosome acidic patch; nucleosome remodeling; structure.

Figure 1.. CSS-associated mutations in the SMARCB1 CTD inhibit mSWI/SNF nucleosome remodeling and ATPase activity on nucleosomes.

A. Summary of missense mutations and in-frame deletions in SMARCB1-associated intellectual disability (ID) syndromes (Coffin-Siris syndrome, Kleefstra syndrome, and non-syndromic severe ID), and cancer (COSMIC). Legends are indicated. B. Immunoblot performed on total nuclear protein and anti-HA immunoprecipitations in SMARCB1-deficient HEK-293T cells. C. Proteomic mass spectrometry results of 293T-SMARCB1 knockout 293Ts expressing WT or mutant SMARCB1 constructs. D. HA-epitope purification of HA-tagged SMARCB1 WT and mutant variant-bound complexes from SMARCB1-deficient HEK-293T cells. Silver staining confirms capture of expected mSWI/SNF subunits and their stoichiometry (superfluous lane deletion separated by white line). E. Schematic for restriction enzyme accessibility assay (REAA) with ST601-GATC1 nucleosome core particle (NCP) harboring a DpnII restriction cut site used to assess nucleosome remodeling of purified mSWI/SNF complexes. F. Nucleosome remodeling (REAA) comparing SMARCB1 WT and mutant variant complexes visualized by Tapestation D1000 (200 ng purified complexes, 30 °C, 90 min). G. Summary of nucleosome remodeling assay using REAA comparing all SMARCB1 WT and mutant variants over time course (200 ng purified complexes, 30 °C, 0-60 minutes, mean ± S.D., n=2; AdjP-values determined by Dunnett’s multiple comparison test compared to WT at each time point). H. Endogenous ARID1A-IP of SMARCB1 WT- and mutant variant- bound complexes. I. ATPase assays performed on mSWI/SNF complexes via ARID1A IP (for canonical BAF complexes from (H)) in solution with 601 Widom DNA, recombinant tetra polynucleosomes, or HeLa polynucleosomes (30 °C, 90 min). Luminescence signal is plotted (mean ± S.D., n=2; AdjP-values determined by Dunnett’s multiple comparison test to WT for each substrate). J. REAA remodeling assay performed in parallel to (I) with recombinant mononucleosomes. See also Figure S1, Table S1, and Table S2.

Figure 2.. The SMARCB1 CTD binds directly to nucleosomes, mediated by a basic, alpha-helical amino acid cluster.

A. Top, conservation of minimal SNF5 homology putative c-terminal domains across species showing ConSurf conservation score, mean pI, sequence logo, and similarity. CSS-associated mutated residues are highlighted in gray. Bottom, N-terminally biotinylated SMARCB1-CTD peptide (aa 351-385) variants generated. B. H. sapiens SMARCB1 CTD WT and mutant intellectual disability-associated biotin-tagged peptide pull downs of mammalian mononucleosomes; immunoblot for histone H3 and histone H2B. C. Immunoblot of peptide pull down of mammalian mononucleosomes across SMARCB1-CTD homologues (H. sapiens SMARCB1 scramble and wild-type, D. melanogaster SNR1, C. elegans CeSNF5, or S. cerevisiae SNF5 and SFH1). D. Backbone assignment. 2D 15N-HSQC spectrum of 0.5 mM 15N-SMARCB1-CTD in PBS, pH 6.5 acquired at 15°C. The backbone NH peaks from SMARCB1-CC residues are assigned in red, and residues from N-terminal cloning tag are assigned in blue. E. Superposition of backbone traces of the 10 lowest-energy structures of the SMARCB1 c-terminal domain (aa 351-385). F. Barrel view cartoon diagram of a representative structure from the of SMARCB1 c-terminal alpha helix highlighting CSS mutated residues in dark blue and additional positive (Arg/Lys) residues in light blue (aa 357-378). G. Electrostatic surface potential of the SMARCB1-CTD, calculated using ABPS (Dolinsky et al., 2004), from −5.0 kTE^-1 (red) to +5.0 kTE^-1 (blue). 180 degree rotations are shown. See also Figure S2 and Table S3.

Figure 3.. The SMARCB1 CTD binds to the nucleosome acidic patch, which is disrupted by CSS-associated missense mutations.

A. Assay schematic for photocrosslinking-based assessment of SMARCB1 CTD binding sites with photocrosslinkable histone residues. B-C. SDS-PAGE immunoblots for biotin resolving Histones H2A/B and H4 as well as non-XL peptide across acidic patch residues for WT and mutant SMARCB1 CTD peptides. D. Summary of crosslinking results within the nucleosome acidic patch (PDB ID: 1ZLA). E. WT SMARCB1 CTD peptide pull-down of WT and acidic patch mutant recombinant mononucleosomes. F. (left) Electrostatic potential of nucleosome (PDB ID: 1KX5) with acidic patch highlighted, with 180 degree rotations; and (right) ZDOCK predicted docking region of SMARCB1-CTD (aa 358-377) on nucleosome overlaid in light blue (compiled across binding constraints) (Pierce et al., 2014). H2A-green, H2B-cyan, H3-maroon, H4-yellow. See also Figure S3.

Figure 4.. CSS-associated mutations in SMARCB1 disrupt genome-wide enhancer DNA accessibility without affecting mSWI/SNF complex targeting.

A. Introduction of C-terminal V5-tagged SMARCB1 WT and mutant variants in TTC1240 SMARCB1-deficient MRT cells. Immunoblot for BRG1, SMARCB1, and TBP are shown. B. Chromatin occupancy of mSWI/SNF complexes (marked by SMARCB1, SMARCC1, and SMARCA4) and H3K27Ac occupancy mapped over overlapped merged SMARCB1/SMARCC1 peaks. C. Heatmap of ATAC-seq genomic accessibility reads over residual and de novo sites. D. Summary metaplots reflecting accessibility at residual (top) and gained (bottom) sites for empty vector, SMARCB1 WT, and SMARCB1 CTD mutant conditions. E. Metaplot of MNase-seq over all SMARCA4 WT summits (top) and gained mSWI/SNF sites. F. Example ChIP-seq and ATAC-seq tracks over the CAPZB (top) and RTFN1 (bottom) loci. G. Principal component analysis (PCA) performed on ATAC-seq peaks overlapping SMARCB1 ChIP-seq sites; experimental replicates for empty vector, SMARCB1 WT, and SMARCB1 CTD mutant conditions. H. PCA performed on RNA-seq experimental replicates for empty vector, SMARCB1 WT, and SMARCB1 CTD mutant conditions (top 10% most variable genes). See also Figure S4.

Figure 5.. CSS-associated heterozygous SMARCB1 mutations in iPSCs block neuronal differentiation.

A. CRISPR-Cas9 mediated genome editing was used to obtain heterozygous SMARCB1 K364del and indel mutant iPSCs which underwent NGN2-mediated neuronal differentiation with RNA-seq collected along an 8 day timecourse for WT and K364del mutant cells. B. Heatmap of ChIP-seq for mSWI/SNF subunits (SMARCB1, SMARCC1, SMARCA4) and H3K27ac as well as ATAC-seq of SMARCB1 +/+ and K364del/+ iPSCs. C. Box plot of normalized difference between WT and K364del/+ mutant SMARCB1, SMARCC1, SMARCA4, H3K27ac ChIP-seq and ATAC-seq results. D. HOMER motif analysis of sites with reduced ATAC-seq accessibility (from B) in the K364del mutant versus WT iPSCs. E. Heatmap of gene cluster downregulated in K364del versus WT cells at Day 8 of NGN2-induced differentiation (Cluster 6). Select differentially regulated genes are highlighted. See also Figure S5I F. Venn diagram depicting overlap of Cluster 6 genes with intellectual disability- and NGN2-induced differentiation- associated genes. G. Bar graphs of intellectual-disability associated genes downregulated in the mutant versus wild-type cells along the differentiation time course. H. Immunoblot demonstrating lentiviral expression of full-length V5-tagged SMARCB1 in WT and mutant iPSCs. I. Neurite outgrowth and neuron count at Day 10 of NGN2 differentiation (mean ± S.E.M, n=12; AdjP-values determined by Dunnett’s multiple comparison test to WT(EV)). J. Imaging of DAPI (DNA), TUJ1, and NGN2 in SAH iPSCs at day 10 of differentiation. K. Model of genome-wide localization and activity of mSWI/SNF complexes assembled with WT or CSS-associated SMARCB1 mutant variants. See also Figure S5 and Table S4.

Figure 6.. Model of SMARCB1-mediated nucleosome engagement and remodeling.

A. Model of SMARCB1 C-terminal alpha helix bound to nucleosomes in complex with SNF2h bound (left) at the SHL2 position (PDB ID: 5X0Y) and (right) at the SHL6 nucleosomal position (PDB ID: 5X0X) generated using ZDOCK (H2AE91 binding constraint). B. Model of mammalian SWI/SNF complex based on Mashtalir et al. with WT or C-terminal mutant SMARCB1 subunit as part of core module. C. Model of genome-wide BAF complex occupancy (ChIP-seq), chromatin accessibility (ATAC-seq), nucleosome occupancy (MNase-seq) and gene expression between SMARCB1-null, WT, and C-terminal mutant conditions at SMARCB1-driven BAF complex sites.