Functional characterization of SMARCA4 variants identified by targeted exome-sequencing of 131,668 cancer patients

Abstract

Genomic studies performed in cancer patients and tumor-derived cell lines have identified a high frequency of alterations in components of the mammalian switch/sucrose non-fermentable (mSWI/SNF or BAF) chromatin remodeling complex, including its core catalytic subunit, SMARCA4. Cells exhibiting loss of SMARCA4 rely on its paralog, SMARCA2, making SMARCA2 an attractive therapeutic target. Here we report the genomic profiling of solid tumors from 131,668 cancer patients, identifying 9434 patients with one or more SMARCA4 gene alterations. Homozygous SMARCA4 mutations were highly prevalent in certain tumor types, notably non-small cell lung cancer (NSCLC), and associated with reduced survival. The large sample size revealed previously uncharacterized hotspot missense mutations within the SMARCA4 helicase domain. Functional characterization of these mutations demonstrated markedly reduced remodeling activity. Surprisingly, a few SMARCA4 missense variants partially or fully rescued paralog dependency, underscoring that careful selection criteria must be employed to identify patients with inactivating, homozygous SMARCA4 missense mutations who may benefit from SMARCA2-targeted therapy.

Fig. 1. SMARCA4 mutational spectrum in the FoundationCORE® patient cohort.

a SMARCA4 alteration frequency separated by disease ontology (n = 131,668 patients). b Distribution of SMARCA4 mutation types (n = 10,562 variants). c Zygosity of SMARCA4 truncating and nontruncating variants in patients where zygosity was determined (n = 5699 patients). d Mutual exclusivity of SMARCA4 mutations with the aggregate of other BAF genes profiled in the FoundationOne^® panel (ARID1A/B, ARID2, PBRM1, SMARCB1, SMARCD1).

Fig. 2. Homozygous SMARCA4 mutations in NSCLC.

a Co-occurrence of SMARCA4 mutations with common oncogenic drivers in lung cancer patients excluding variants of unknown significance (n = 19,848 patients). b Summary table of odds ratio (OR) and P values using a two-tailed Fisher’s exact test (not adjusted for multiple testing) for patients in a. c Overall survival of patients with SMARCA4 mutations. d Overall survival of patients with SMARCA4 mutations that have received cancer immunotherapy at any time during their treatment. The log-rank test was used to compare the overall survival of groups and resulting P values are unadjusted (c–d).

Fig. 3. Hotspot missense mutations in the SMARCA4 ATPase helicase domain.

a Lollipop plots of SMARCA4 missense mutations in TCGA (top, n = 379 missense variants) and the FoundationCORE^® (bottom, n = 6289 missense variants) cohorts. b Locations of frequently mutated residues on a SMARCA4 homology model derived from yeast Snf2-nucleosome complexes (PDB: 5X0X and PDB:5Z3U). c Amino acid changes of frequently mutated SMARCA4 residues.

Fig. 4. SMARCA4 missense mutants are deficient in opening chromatin and inducing gene expression.

a FRET nucleosome remodeling assays were performed with immunopurified SMARCA4 WT and mutants from 293T cells transduced with SMARCA4 WT or mutants. Cy3/Cy5 ratios are represented in a 60 min kinetic assay, each construct is normalized to its no ATP control (n = 2 biologically independent samples, lines represent the mean). b Significantly open and closed sites as measured by ATAC-seq in NCI-H1944 cells expressing SMARCA4 WT or mutants relative to the LACZ control (n = 2 per construct). Significance was tested with a moderated t-statistic (two-sided) and P values were adjusted for multiple testing with the Benjamini–Hochberg procedure. c Heatmap of ATAC-seq changes relative to LACZ control (log₂ fold-change) in the union of sites opened and closed from b (n = 2 per construct). d Representative IGV track of SMARCA2/SMARCA4 ChIP-seq and ATAC-seq changes in cells from b (overlay of 2 replicates per construct). e Heatmap of chromatin accessibility and SMARCA2 and SMARCA4 occupancy at sites from c in NCI-H1944 cells transduced with the LACZ control or SMARCA4 WT (n = 2 per construct). Data are shown as normalized peak counts per million genomic DNA fragments in a 2 kb window around the peak center. Rows are rank ordered by ATAC-seq peaks. R, replicate. f Heatmap of qRT-PCR analysis of a subset of SMARCA4 WT-induced genes in NCI-H1944 cells transduced with SMARCA4 WT or mutants (mean of n = 3 biologically independent samples). g SMARCA4 qChIP at target genes and a negative control region on chr14 in cells from f (each dot represents a technical replicate, n = 2; representative of 3 independent experiments). Source data are provided as a Source Data file.

Fig. 5. Differential effects of SMARCA4 mutants to rescue cell growth and chromatin accessibility loss after SMARCA2 knockdown.

a Long term clonogenic growth of NCI-H1944 cells transduced with SMARCA4 WT or mutants after SMARCA2 knockdown. Representative of at least 3 replicates. b Immunoblot of cells from a (representative of at least 3 replicates). c Heatmap of ATAC-seq changes at sites after SMARCA2 knockdown in cells from a (n = 2 per construct). Values represent log₂ fold-change relative to LACZ control after SMARCA2 knockdown. d Heatmap of SMARCA2 and SMARCA4 occupancy at regions with lower accessibility after SMARCA2 knockdown (sites from c) (n = 2 per construct). SMARCA2 ChIP-seq was performed in NCI-H1944 cells expressing LACZ. SMARCA4 ChIP-seq was performed in NCI-H1944 cells expressing SMARCA4 WT and doxycycline (DOX)-inducible expression of SMARCA2-targeting shRNA. Data are shown as normalized peak counts per million genomic DNA fragments in a 2 kb window around the peak center. Rows are rank ordered by SMARCA2 enrichment. R, replicate; INP, input. e Number of sites closed (left axis, blue bar, n = 2 per construct) and mean percent cell death (right axis, red dot, mean of 3 replicates) after SMARCA2 knockdown in cells from a. f Heatmap of genes downregulated after SMARCA2 knockdown in NCI-H1944 cells transduced with LACZ, WT or K785R mutant (n = 3 per construct). Data are shown as mean-centered normalized reads per kb of transcript per million mapped reads (nRPKM). g Long term clonogenic growth of CAL-12T, NCI-H1435 and HCC1897, which all harbor homozygous SMARCA4 missense mutations, after knockdown of SMARCA2 or SMARCA4 (left). Immunoblot confirming SMARCA2/SMARCA4 protein depletion (right). Data are representative of at least 2 replicates. Source data are provided as a Source Data file.