BRG1/SMARCA4 inactivation promotes non-small cell lung cancer aggressiveness by altering chromatin organization

Abstract

SWI/SNF chromatin remodeling complexes regulate critical cellular processes, including cell-cycle control, programmed cell death, differentiation, genomic instability, and DNA repair. Inactivation of this class of chromatin remodeling complex has been associated with a variety of malignancies, including lung, ovarian, renal, liver, and pediatric cancers. In particular, approximately 10% of primary human lung non-small cell lung cancers (NSCLC) display attenuations in the BRG1 ATPase, a core factor in SWI/SNF complexes. To evaluate the role of BRG1 attenuation in NSCLC development, we examined the effect of BRG1 silencing in primary and established human NSCLC cells. BRG1 loss altered cellular morphology and increased tumorigenic potential. Gene expression analyses showed reduced expression of genes known to be associated with progression of human NSCLC. We demonstrated that BRG1 losses in NSCLC cells were associated with variations in chromatin structure, including differences in nucleosome positioning and occupancy surrounding transcriptional start sites of disease-relevant genes. Our results offer direct evidence that BRG1 attenuation contributes to NSCLC aggressiveness by altering nucleosome positioning at a wide range of genes, including key cancer-associated genes.

Figure 1. In vitro and in vivo growth properties of BRG1-deficient H358 cell lines

(A) Phase contrast photomicrographs of the H358 cell line and clonal cell lines with reduced BRG1 expression (Brg1i.1.25 and Brg1i.2.4). (B) Each cell line was inoculated intrathoracically into 4 Nu/Nu female mice and monitored for tumor development 3X weekly. (C) Paraffin-embedded sections of representative lung and surrounding tissues from mice after intrathoracic inoculation of parental or Brg1i.1 cells. Mice were sacrificed at 138 days (parental), 36 days (Brg1i.1) and 43 days (Brg1i.1). Sections were stained for histology by H&E by standard methods (top row). NL= normal lung; T= tumor; H= heart; Magnification=400X

Figure 2. Gene expression changes in response to BRG1 silencing

Heatmap visualization of gene expression values for 591 differentially expressed genes in which rows are genes and columns are cell line replicates. Expression levels are indicated by the shading in the heatmap. BRG1 knockdown cell lines (Brg1i.1 and Brg1i.2) exhibit highly concordant expression patterns that differ markedly from those seen in control cell lines (Control). Dendrograms show the results of hierarchical clustering of genes and cell line replicates, and all replicates cluster together.

Figure 3. Characterization of BRG1 target gene expression in BRG1 knockdown cell lines

mRNA expression of BRG1 and 6 putative target genes in BRG1 knockdown H358 and H441 cell lines was assessed by quantitative RT-PCR as described in the Material and Methods. The mRNA level of each gene was by qPCR and normalized for ß-actin expression. Values are the mean of at least two replicates of two independent experiments; bars, ± SD.

Figure 4. Expression of BRG1 target genes in lung adenocarcinoma

(A) Median centered expression values of wild type and mutant BRG1 in the TCGA lung adenocarcinoma cohort are plotted. Color coding indicates mutation status, and samples below the dashed line are classified as having low BRG1 expression. (B) – (E) Boxplots comparing expression values of BRG1 target genes for samples with low BRG1 expression vs. other in the TCGA lung adenocarcinoma cohort. Two-sided Wilcoxon rank sum p-values are shown.

Figure 5. BRG1 loss results in variation in nucleosome occupancy and positioning

MNase signal was normalized to a global average across the three experimental conditions. Signal ±1 kb (A) and ±10 kb (B) were plotted around the TSS at single base pair resolution (red = Control, blue = Brg1i.1, black = Brg1i.2). Nucleosome occupancy scores (C) and positional ambiguity scores (D) were assigned for all predicted nucleosomes in Control (n = 9,587,862) and Brg1i.2 (n = 10,344,643) cells (All) or for those predicted to be ±3 kb from TSS in Control (n = 498,437) and Brg1i.2 (n = 470,542) cells (TSS). Green lines mark the first and third quartile boundaries for all nucleosome in Control cells. A two-sided Wilcoxon rank sum test was used to determine significance and was p < 10⁻⁷ (**). The center line of each box indicates the median value. Outliers not shown.

Figure 6. Nucleosomes reposition into the NDR with BRG1 loss

(A) Heatmaps show normalized nucleosome signal ±1 kb of surrounding RefSeq TSS of downregulated, upregulated, and expression-matched unregulated genes. Each row reflects one gene. Lines plots show the kernel smoothed average normalized MNase signal at single base pair resolution for all regions depicted in the heatmap (red = Control, black = Brg1i.2). (B) UCSC genome browser tracks showing window-smoothed, normalized MNase signal around individual TSS. Red lines indicate the region used to define the NDR. (C) Signal difference between the −1 nucleosome and the NDR (−1nuc – NDR) for each gene in the set (purple = downregulated genes, orange = unregulated genes). The log transformed ratio of signal difference between Control and Brg1i.2 was plotted for each gene $lo{g}_2\left(\frac{{(-nuc-NDR)}_{control}}{{(-1nuc-NDR)}_{BRG1i.2}}\right)$. The unregulated gene was sampled to match the number of downregulated genes 1,000 times. The average value for each position from the sampling was plotted, and black dashed lines indicate the standard deviation from this sampling. Significance between the downregulated and unregulated gene sets was determined using a Kolmogorov–Smirnov test.