Loss of Snf5 Induces Formation of an Aberrant SWI/SNF Complex

Abstract

The SWI/SNF chromatin remodeling complex is highly conserved from yeast to human, and aberrant SWI/SNF complexes contribute to human disease. The Snf5/SMARCB1/INI1 subunit of SWI/SNF is a tumor suppressor frequently lost in pediatric rhabdoid cancers. We examined the effects of Snf5 loss on the composition, nucleosome binding, recruitment, and remodeling activities of yeast SWI/SNF. The Snf5 subunit is shown by crosslinking-mass spectrometry (CX-MS) and subunit deletion analysis to interact with the ATPase domain of Snf2 and to form a submodule consisting of Snf5, Swp82, and Taf14. Snf5 promotes binding of the Snf2 ATPase domain to nucleosomal DNA and enhances the catalytic and nucleosome remodeling activities of SWI/SNF. Snf5 is also required for SWI/SNF recruitment by acidic transcription factors. RNA-seq analysis suggests that both the recruitment and remodeling functions of Snf5 are required in vivo for SWI/SNF regulation of gene expression. Thus, loss of SNF5 alters the structure and function of SWI/SNF.

Keywords: BAF47; Chromatin remodeling; INI1; SMARCB1; SWI/SNF; Snf5.

Figure 1. Interactions of the Snf5 module in SWI/SNF

(A) The interactions of Snf5 with other subunits in the SWI/SNF complex were determined by crosslinking with the homo-bifunctional, lysine-specific crosslinker BS3 and subsequent analysis by mass spectrometry. The SWI/SNF subunits shown are Snf2 (brown), Swi1 (purple), Swi3 (orange), Swp82 (blue), Snf6 (magenta) and Taf14 (green). Lysine residues are indicated by green circles and the labeled boxes are the conserved domains. Only those interactions in which 2 or more crosslinks were detected in a region spanning ~30 amino acids are shown and the number of crosslinks detected are indicated by the circled numbers. The thickness of the black line is also indicative of the number of crosslinks detected in that region. (B) SWI/SNF complexes were purified by immuno-affinity chromatography with either Snf5 or Swp82 deleted and analyzed on a 4–20% SDS polyacrylamide gel. Wild type SWI/SNF (lane 3) contains two FLAG tags at the C-terminus of Snf2. Snf6 was tagged with a double FLAG tag at its C terminus in the Δsnf5 (lane 2) and Δswp82 (lane 1) SWI/SNF complexes. (C) Recruitment of wild type and mutant SWI/SNF by the acidic transcription activator Gal4-VP16 was tracked by gel shift analysis. Nucleosomes alone (N), nucleosomes bound by Gal4-VP-16 (N-Gal4) and nucleosomes bound by SWI/SNF with or without Gal4-VP16 (N-Gal4-SWI) migrated at the indicated positions. Gal4-VP16 was added to lanes 2, 4, 7, 8, 11, and 12. Wild type, Δsnf5 and Δswp82 SWI/SNF were add in lanes 3–4, 5–8, and 9–12, respectively. (D) Purified SWI/SNF complexes from wild type (WT) and mutant yeast strains (as indicated) were added to nucleosomes reconstituted with radiolabeled 601 DNA and Xenopus laevis histone octamers and resolved on a 4% native polyacrylamide gels. Free nucleosomes are labeled N and nucleosomes bound by SWI/SNF as N-SWI. (E) The amount of bound nucleosomes versus SWI/SNF was determined from gel shift assays as shown in (B) and plotted. The estimated K_D for wild type, Δsnf5 and Δswp82 SWI/SNF were 102 ± 19, 101 ± 15 and 82 ± 21 nM (standard deviation from mean), respectively, and were from three technical replicates.

Figure 2. The highly conserved core domain of the Snf5 subunit contacts nucleosomes near the exposed surface of the histones H2A-H2B

(A) SWI/SNF was recruited to nucleosomes by Gal4-VP16 in presence of competitor DNA to ensure specific SWI/SNF interaction with nucleosomes. After photocrosslinking the labeled Snf5 was isolated and cleaved at Asn-Gly positions with hydroxylamine. The cleaved products were resolved by SDS-PAGE and visualized by phosphorimaging to determine the crosslinked sites using Snf5 marker polypeptides described in section. (B) Four different regions of Snf5 were prepared by in vitro coupled transcription-translation in presence of S-35 methionine. The numbers on the top show the regions of Snf5. (C) The domain organization of Snf5 and the locations of the Asn-Gly cleavage sites are shown. The expected hydroxylamine partial and complete cleavage products are shown.

Figure 3. The catalytic subunit does not make stable contact with nucleosomal DNA in the absence of the Snf5 module, but does with the histone octamer surface

(A) The interactions of wild type (WT) and Δsnf5 SWI/SNF with nucleosomes were mapped by histone photocrosslinking with photoreactive radio-iodinated N-((2-pyridyldithio)ethyl)-4-azidosalicylamide (PEAS) attached to specific histone positions as indicated by histone type and residue position. (B) The relative crosslinking efficiency of Snf2, Snf5, Swi3 and in some instances Swi1 was plotted for each of the five histone positions and for WT and Δsnf5 SWI/SNF. Signals were normalized relative to crosslinking in WT SWI/SNF to residue 80 of histone H3 (H380). (C–D) Purified WT SWI/SNF complex was bound to nucleosomes reconstituted with radiolabeled DNA probes containing photoreactive groups at the positions indicated (“0” corresponding to the dyad). Reactions contained either wild type (lanes 1–3) or Δsnf5 (lanes 4–6) complexes. In some instances 150 μM γ-thio-ATP (lanes 2 and 5) or 13 μg/ml competitor DNA (lanes 3 and 6) were added. (E) The efficiency of Snf2 crosslinking to DNA was plotted for four different positions within nucleosomal DNA and were normalized to nt-17. These experiments were done in triplicates and the standard deviation (SD) is shown.

Figure 4. Snf5 contacts the ATPase domain of the catalytic subunit

(A) The interactions of Snf2 with other SWI/SNF subunits were examined in the same manner as described for Snf5 in Figure 3. Besides the known interactions of the Arp7, Arp9 and Rtt102 module there also are interactions of the Snf5, Swp82 and Taf14 module with Snf5. Only those crosslinks in which 2 or more crosslinks were detected in a region spanning ~30 amino acids are shown. The total number of crosslinks detected is indicated by the circled number. (B) The ATPase domain of Snf2 bound to DNA was modeled based on the sequence similarity with the ATPase domain of RAD54 (Dechassa et al., 2012). The lysine residues crosslinked to the Snf5 subunit are indicated as blue space filling residues and are primarily in the HELIC lobe of the ATPase domain. Domains are color coded as indicated in the schematic at the bottom of the figure.

Figure 5. Snf5 regulates the catalytic activity of the ATPase domain of SWI/SNF

The rate of ATP hydrolysis (V) was determined at different concentrations of ATP and plotted as a function of ATP concentration. K_M and k_cat values were obtained by fitting data to the Michaelis-Menten equation using GraphPad (PRISM Version 6.0b). Concentrations of substrates and enzymes were (A) 8 nM nucleosomal array (12 repeats of 200bp), 1.6 nM enzyme (WT/Δsnf5) and 2–600 μM ATP; (B) 8 nM 29N59 nucleosome, 1.6 nM enzyme, and 2–300 μM ATP; and (C) 50 nM 100 bp DNA, 10 nM enzyme, and 10–300 μM ATP. Error bars and reported values are mean ± SE from three technical replicates.

Figure 6. Loss of the Snf5 subunit adversely affects the intrinsic nucleosome mobilizing activity of SWI/SNF

(A–B) Nucleosomes modified at residue 53 of histone H2B were used for monitoring movement of DNA on the octamer surface. DNA cleavage products were resolved on a denaturing 6% polyacrylamide gel and visualized by phosphorimaging. Numbers on the right side of the gel image refer to number of nucleotides (nt) moved from the starting cleavage position (0). Nucleosomes were remodeled with WT (A), and Δsnf5 SWI/SNF (B) for 0, 5, 10, 20, 30, 40, 50, 60, 70, 80 and 160 s using 4.4 μM ATP. (C) The amount of DNA cleaved at starting position (0) was plotted versus time for WT SWI/SNF and Δsnf5 SWI/SNF. Rate constants (k) obtained by fitting data to single exponential function were 0.044± 0.004 s⁻¹ (WT) and 0.016 ± 0.005 s⁻¹ (Δsnf5). Bars and reported values are the mean ± SE from two replicates. (D) ATPase assays were performed in the same conditions as (A and B) and the released radioactive phosphate (Pi) was separated from non-hydrolyzed ATP by thin layer chromatography (TLC). The percentage Pi released is plotted as a function of time. ATP hydrolysis rates obtained were 0.9 μM s⁻¹ (WT) and 0.7 μM s⁻¹ (Δsnf5).

Figure 7. In vivo targets affected by loss of Snf5 both overlap and are distinct from that of Snf2

(A) RNA-seq heat map depicts those genes with fold change >= 2 in expression (506) and satisfy the parameters of a p < 0.05. Genes were grouped into 6 clusters using k-Means clustering. A total of 5,935 genes were analyzed.