Mutation of SMARCA4 Induces Cancer Cell–Intrinsic Defects in the Enhancer Landscape and Resistance to Immunotherapy

Epigenetic reprogramming in SMARCA4-mutant cancer cells alters immune infiltration and limits immunotherapy efficacy by downregulating immunostimulatory gene expression, which could potentially be targeted to overcome immunotherapy resistance in SMARCA4-deficient tumors.

Abstract

Cancer genomic studies have identified frequent alterations in genes encoding components of the SWI/SNF chromatin remodeling complex, including SMARCA4 and ARID1A. Importantly, clinical reports indicate that SMARCA4-mutant lung cancers respond poorly to immunotherapy and have dismal prognosis. In this study, we corroborated the clinical findings by using immune-humanized, syngeneic, and genetically engineered mouse models of lung cancer harboring SMARCA4 deficiency. Specifically, models with SMARCA4 loss showed decreased response to anti-PD-1 immunotherapy associated with significantly reduced infiltration of dendritic cells and CD4⁺ T cells into the tumor microenvironment. SMARCA4 loss in tumor cells led to profound downregulation of STING1, IL1β, and other components of the innate immune system, as well as inflammatory cytokines that are required for efficient recruitment and activity of immune cells. The deregulation of gene expression was caused by cancer cell–intrinsic reprogramming of the enhancer landscape with marked loss of chromatin accessibility at enhancers of genes involved in innate immune response, such as STING1, IL1β, type I IFN, and inflammatory cytokines. Interestingly, the transcription factor NF-κB–binding motif was enriched in enhancers that lose accessibility upon SMARCA4 deficiency. Furthermore, SMARCA4 and NF-κB co-occupied the same genomic loci on enhancers associated with STING1 and IFNβ, indicating a functional interplay between SMARCA4 and NF-κB. Taken together, these findings provide the mechanistic basis for the poor response of SMARCA4-mutant tumors to immunotherapy and establish a functional link between SMARCA4 and NF-κB in innate immune and inflammatory gene expression regulation.

Significance: Epigenetic reprogramming in SMARCA4-mutant cancer cells alters immune infiltration and limits immunotherapy efficacy by downregulating immunostimulatory gene expression, which could potentially be targeted to overcome immunotherapy resistance in SMARCA4-deficient tumors.

Introduction

Lung cancer is a devastating disease that remains the top cause of cancer mortality (1, 2). Despite the advent of targeted therapy against oncogenic kinases and recent KRAS^G12C inhibitors, the majority of patients with lung cancer still lack effective therapeutics, underscoring the dire need for additional treatment option (3). Significant strides have been made in recent years, with the development of immune checkpoint inhibitors targeting CTLA4 and PD-1/PD-L1 improving the outcomes for some patients with lung cancer (4).

The SWItch/Sucrose Non-Fermenting (SWI/SNF) chromatin remodeling complex is a large multiprotein assembly that uses the energy derived from ATP hydrolysis to remodel nucleosomes and facilitate chromatin-dependent cellular processes such as DNA replication, repair, and transcription (5). Subunits of the SWI/SNF chromatin remodeling complex, including SMARCA4 and ARID1A, are frequently mutated in lung cancer [16% in early-stage disease as in The Cancer Genome Atlas (TCGA) and up to 33% in advanced stages; refs. 6, 7]. Several recent independent clinical studies have shown that SMARCA4-mutant lung cancers have one of the worst prognosis among genetically defined subtypes of lung cancer. Moreover, SMARCA4-mutant tumors show poor response to immunotherapy (4, 8). Bioinformatic analysis into the tumor microenvironment (TME) of SMARCA4-mutant lung cancer revealed these tumors have significantly lower levels of infiltrating cytotoxic T cells and show “immune-cold” features (4, 8), which is in line with the lack of clinical response to immunotherapy. In addition to primary resistance to immunotherapy, SMARCA4 mutation was recently identified as the second most common genetic alteration upon development of acquired resistance to immune checkpoint inhibitors in lung cancer (9), strongly suggesting a key role of SMARCA4 in modifying antitumor immunity. Importantly, SMARCA4-mutant tumors have high tumor mutational burden (TMB; refs. 7, 10); thus, the lack of immunotherapy response is particularly puzzling. To date, the molecular mechanisms of resistance of SMARCA4-mutant tumors to immunotherapy remain unknown.

The most commonly clinically utilized immunotherapies are T cell–centered (11, 12). However, the effector functions of T cells are nonautonomous, and studies have shown that the initiation and sustainability of T-cell response and the maintenance of T-cell memory depend on interactions with other immune cells such as dendritic cells (DC) and innate immunity (13, 14). Importantly, a growing body of work has recently shown critical role of DCs in antitumor immunity and shaping T-cell activity during immunotherapy (15–20). In these processes, IFNs and inducers of IFN response, such as STING1, play a key role (16, 21). Specifically, innate immune sensing of tumors largely occurs through the host STING1 pathway, which leads to type I IFN production, DC activation, cross-presentation of tumor-associated antigens to CD8⁺ T cells, and T-cell recruitment into the TME (16, 18). This has generated tremendous excitement, and efforts to reengage the innate immune system using STING1 agonists are underway (22, 23). Furthermore, therapies aimed at activating DCs such as by administration of Fms-like tyrosine kinase 3 ligand, a growth factor promoting DC development, in combination with αCD40 (to activate DCs), termed DC therapy, are being actively investigated (24).

In this study, we show that SMARCA4-mutant tumors respond poorly to anti-PD-1 immunotherapy by using several orthogonal models of lung cancer. We show that SMARCA4 loss leads to significantly reduced infiltration of DCs and CD4⁺ T cells into the TME. Mechanistically, we show that loss of SMARCA4 in cancer cells results in a profound epigenetic deregulation of enhancers, leading to downregulation of STING1, IL1β, and other components of the innate immune system that are required for efficient recruitment and activity of immune cells. We also show that SMARCA4 and NF-κB cooccupy the same genomic loci on enhancers associated with STING1 and IFNβ, indicating a functional interplay between SMARCA4 and NF-κB. Taken together, our findings provide a cancer cell–intrinsic mechanistic basis for the poor response of SMARCA4-mutant tumors to anti-PD-1 immunotherapy and establish a functional link between SMARCA4 and NF-κB on innate immune and inflammatory gene expression regulation.

Materials and Methods

Ethics statement

Animal studies were carried out according to protocols approved by the University of Texas MD Anderson Cancer Center (UTMDACC) Institutional Animal Care and Use Committee. Mice were fed commercial rodent diet (PicoLab Rodent diet 5053 from LabDiet) and water ad libitum. All mice were kept in a specific pathogen-free vivarium at the MDACC mouse facilities. Mice were kept in a 12-hour light/12-hour dark cycle as commonly used and housed at 18°C to 23°C with humidity of 50% to 60%.

Humanization mice

Mice humanization were generated described previously (25, 26). Female 3- to 4-week-old NOD.Cg-Prkdc^scidIl2rg^tm1Wjl/SzJ (NSG) mice were obtained from The Jackson Laboratory. Mice were housed in microisolator cages under specific pathogen-free conditions in a dedicated humanized mice room in the animal facility at UTMDACC. Mice were provided with autoclaved acidified water and fed a specialized diet known as the Uniprim diet.

As previously described, human umbilical cord blood units for research were sourced from MD Anderson Cord Blood Bank under an Institutional Review Board–approved protocol (Lab04-0249) in which we obtained written informed consent from the patients (26). The studies were conducted in accordance with recognized ethical guidelines, including the Declaration of Helsinki and Belmont Report. The cord blood bank acquires umbilical cord blood through voluntary donations from mothers following informed consent under the Institutional Review Board–approved protocol. Fresh cord blood units were promptly transported to the research laboratory within 24 hours of collection for HLA typing at MD Anderson HLA typing core facility. The cord blood was diluted with PBS at a 1:3 ratio, and mononuclear cells were isolated by using density-gradient centrifugation with Ficoll medium. The isolated mononuclear cells were then directly utilized for CD34⁺ enrichment procedures.

After mononuclear cells were separated from human umbilical cord blood, CD34⁺ hematopoietic stem cells (HSC) were isolated using CD34⁺ MicroBead Kit (Miltenyi Biotec, cat. #130-046-702). NSG mice, 3 to 4 weeks old, were irradiated with 200 cGy using a ¹³⁷Cs γ irradiator. More than 90% pure freshly isolated CD34⁺ HSCs were injected intravenously, 24 hours after irradiation, at a density of 1 to 2 × 10⁵ CD34⁺ cells/mouse. The engraftment levels of human CD45⁺ cells were determined in the peripheral blood at 6 weeks after CD34⁺ injection by flow cytometric quantification, as well as other human immune populations. Mice with more than 25% human CD45⁺ cells were considered humanized (hu-NSG mice). In-depth analysis of peripheral blood for human immune cell subpopulations, including CD45⁺, CD3⁺, CD4⁺, and CD8⁺ T cells, B cells, NK cells, and lineage-negative cells, was performed using a 10-color flow cytometry panel at weeks 6 after CD34⁺ engraftment. Hu-NSG mice from different cord blood donors with varying levels of engraftment were randomized into every treatment group in all the experiments. All hu-NSG mice were verified for humanization before tumor implantation.

Genetically engineered mouse model of lung adenocarcinoma

We have established novel inducible lung cancer mouse models, termed KPS, driven by loss of SMARCA4 in combination with p53 loss and Kras mutations (Kras^LSLG12D/WT, p53^fl/fl, Smarca4^fl/f; ref. 5). Briefly, Smarca4^flox/flox mice with loxp sites flanking exons 2 and 3 of mouse Smarca4 gene were crossed with previously described KP (K-ras^LSL-G12D/WT; p53^flox/flox) mice (27, 28). KP and KPS mice were backcrossed to the C57BL/6 background for 10 generations to achieve a pure genetic background. Ad5CC10-Cre virus (UI Viral Vector Core, cat. #VVC-Berns-1166) at a viral titer of 10¹² particles per mL (pt/mL) was purchased from The University of Iowa Viral Vector Core Facility. KP and KPS mice, 5 to 8 weeks, were intratracheally injected with 50 μL of 4 × 10¹⁰ pt/mL Ad5CC10-Cre virus. MicroCT scan was performed 3 months after virus injection to assess lung tumor development.

Micro-CT scans of mice

Bruker Skyscan 1276 (Bruker) was used for data acquisition in prone position under isoflurane inhalation anesthesia (tube voltage 70 kV, tube current 200 μA, and 30 μm effective pixel size) with retro-active respiratory gating (i.e., synchronization of acquisition of micro-CT projections with a timepoint in the respiratory cycle of the individual mouse). Scanning took approximately ∼3.5 minutes. Respiratory monitoring was performed using a camera and tape to monitor the respiratory sweep. Images were reconstructed and assessed at a constant window width/window level (0–0.025).

Anti-PD-1 treatment in hu-NSG xenograft and syngeneic tumor models

A measure of 1 × 10⁷ H2122 parental or SMARCA4 knockout (KO) cells in 200 μL PBS:Matrigel (R&D Systems, cat. #3432-001; 1:1) mix were injected subcutaneously into female hu-NSG mice 6 to 8 weeks after CD34 engraftment. When tumor size reached approximately 200 mm³, 200 μg anti-human PD-1 antibody (pembrolizumab, Keytruda, cat. #88078K) or corresponding isotype control antibody were delivered intraperitoneally in 100 μL PBS to each mouse twice a week for 2 to 3 weeks.

For the FM471 model, 1 × 10⁷ FM471 parental and Smarca4 KO cells or control and Nf-κb short-hairpin RNA (shRNA)-containing cells in 200 μL PBS:Matrigel (1:1) mix were injected subcutaneously into female 8-week-old C57BL/6 mice (RRID: IMSR_JAX:000664). Treatment with 200 μg rat anti-mouse PD-1 IgG2a antibody (clone: RMP1-14, Bio X Cell, cat. #BP0146, RRID: AB_10949053) or corresponding isotype control antibody (clone: 2A3, Bio X Cell, cat. #BP0089, RRID: AB_1107769) in 100 μL PBS was administered to each mouse 3 time a week. The endpoint was determined at 9 days after treatment, when the tumor size decreased to approximately 200 mm³.

The length and width of the tumor were measured using calipers. The tumor size was calculated according to the following formula: L × W²/2, in which L means length and W means width.

Cell culture

H2122 (RRID: CVCL_1531), HCC44 (RRID: CVCL_2060), H1792 (RRID: CVCL_1495), H1975 (RRID: CVCL_1511), H1693 (RRID: CVCL_1488), H1299 (RRID: CVCL_0060), and H322 were purchased from the ATCC. HCC44 was purchased from DSMZ. H322 was from Sigma (RRID: SCR_000488). These cells were cultured in RPMI-1640 medium supplemented with 10% heat-inactivated FBS and 1% penicillin–streptomycin at 37°C with 5% CO₂.

The FM471 (MDA-F471) cell line (29) was kindly provided by Dr. Humam Kadara (UTMDACC). These cells were derived from lung adenocarcinomas of C57BL/6 mice subjected to tobacco smoke and possess Kras and p53 mutations. The cells are cultured in DMEM/F-12 medium supplemented with 10% heat-inactivated FBS and 1% penicillin–streptomycin at 37°C in a humidified incubator with 5% CO₂. Routine Mycoplasma testing was performed using Mycoplasma Detection Kit. The cells used in this study were limited to less than 20 passages. The cells used in this study were limited to less than 20 passages.

cGAMP (10 μg/mL; InvivoGen, cat. #tlrl-nacga23) and poly(dA:dT; 5 μg/mL; InvivoGen, cat. #tlrl-patn-1) were transfected into cells using Lipofectamine 3000 (Life Technologies) diluted in Opti-MEM according to the manufacturer’s instructions. ACBI1 (1 μmol/L; Selleck Chemicals, cat. #S9612) was diluted in cell medium to treat cells. cGAMP and poly(dA:dT) were all reconstituted in ddH₂O. ACBI1 was dissolved in DMSO. Equivalent amounts of ddH₂O and DMSO were used for vehicle controls as appropriate.

The generation of KO cell lines

Human lung cancer cell lines, H2122 and HCC44, deficient in SMARCA4 were generated by CRISPR-Cas9. Single-guide RNAs (sgRNA) were designed with CRISPick (https://portals.broadinstitute.org/gppx/crispick/public).

The sgRNA was synthetized by Thermo Fisher Scientific (https://www.thermofisher.com/us/en/home/life-science/genome-editing/crispr-libraries/trueguide-grnas.html).

The sgRNA was transfected to cells with TrueGuide Synthetic gRNA and TrueCut Cas9 Protein v2 using the Neon Transfection System protocol (https://assets.thermofisher.com/TFS-Assets/LSG/manuals/MAN0017058_TrueGuide_Synthetic_gRNA_UG.pdf).

Mouse lung cancer cell line FM471 deficient in Smarca4 was generated by Advanced Cell Engineering and 3D Models Core at Baylor College of Medicine using CRISPR/Cas9.

KO cell lines were confirmed using immunoblotting and tested Mycoplasma-negative. The cells used in this study were limited to less than 20 passages.

Immunoblotting

Cells were pelleted and lysed using RIPA buffer. Cell lysates were centrifuged at 12,000 × g for 20 minutes at 4°C, and supernatants containing the soluble proteins were collected. Protein concentration was determined using Bio-Rad Protein Assay Kit II (Bio-Rad, cat. #5000002). Subsequently, 20 to 30 μg of total protein from each sample was loaded into individual lanes. Proteins were mixed with 4× LDS buffer, boiled at 95°C for 5 minutes, and separated by gradient SDS-PAGE (4%–15%), followed by transfer to nitrocellulose membranes. Membranes were blocked with 5% nonfat dried milk in PBS and incubated overnight at 4°C with primary antibodies specific to the target proteins of interest. After incubation, membranes were washed with PBS and incubated for 1 hour at room temperature with the appropriate horseradish peroxidase–conjugated secondary antibody, which facilitate enhanced chemiluminescence detection. Membranes were imaged using the LI-COR Odyssey software.

Coimmunoprecipitation analysis

For endogenous coimmunoprecipitation, cells were collected and resuspended in cold PBS. Nuclear extracts were isolated using NE-PER Nuclear and Cytoplasmic Extraction Kit (Thermo Fisher Scientific, cat. #78833). Nuclear extracts were incubated at 4°C overnight with 2 μg of the following antibodies: anti-SMARCA4, anti–NF-κB, and rabbit mAb IgG XP (Cell Signaling Technology, cat. #3900) isotype control. The samples were then incubated with Dynabeads protein A/G (Thermo Fisher Scientific, cat. #80104G) for 1 hour at 4°C. Beads were washed three times in IP buffer (Thermo Fisher Scientific, cat. #87787) and eluted with 1× sample buffer. The presence of NF-κB and SMARCA4 from the immunoprecipitated protein complexes was detected by immunoblotting.

RNA isolation and qRT-PCR

Total RNA isolation from cells were performed using Qiagen RNeasy Kit (Qiagen, cat. #74104) according to the manufacturer’s instructions. For tumor RNA isolation, tumors were homogenized by zirconium oxide beads. The homogenates were purified using Qiagen RNeasy Kit.

Reverse transcription from total RNA was performed from 2 μg of total RNA using the SuperScript IV VILO Master Mix (Invitrogen-Life Technologies, cat. #11766050) according to the manufacturer’s instructions. qRT-PCR was performed with SYBR Green dye using AriaMx Real-time PCR System (Agilent). PCR reactions were performed in triplicate, and the relative amount of cDNA was calculated by the comparative ΔΔCT method using GAPDH as an internal control.

Preparation of single-cell suspensions for flow cytometry

Erythrocytes in the peripheral blood were lysed with ACK lysis buffer (Thermo Fisher Scientific, cat. #A1049201). Single-cell suspensions were prepared from fresh primary tumors and spleen tissues using standard procedures (30, 31). The suspensions were generated by mechanical dissociation and incubation for 20 minutes at 37°C with dissociation solution (100 μg/mL collagenase IV and 20 μg/mL DNase I in RPMI-1640). The digestion was ended with 10 mmol/L EDTA at room temperature for 5 to 10 minutes, and the digested tissue suspension was filtered through a 100-μm nylon filter. The remaining red blood cells were lysed using 0.5 mL RBC lysis buffer by incubation 5 minutes at room temperature. Single cells were obtained in FACS buffer (1× PBS, 1 mmol/L EDTA, 2% FBS).

Flow cytometry

H2122 tumors and spleen of hu-NSG mice were stained with the following antibodies

CD45.2-AF700 (clone: 2D1, BioLegend, cat. #368514, RRID: AB_2566373), CD4 Pacific Blue (clone: OKT4, BioLegend, cat. #317429, RRID: AB_571953), CD8-APC-Cy7 (clone: HIT8a, BioLegend, cat. #300926, RRID: AB_10613636), CD19-FITC (clone: HIB19, BioLegend, cat. #302206, RRID: AB_2564143), CD3-FITC (clone: HIT3a, BioLegend, cat. #300306, RRID: AB_314041), CD11b-PE-Cy7 (clone: ICRF44, BioLegend, cat. #301322, RRID: AB_830643), CD103-SB600 (clone: B-Ly7, Thermo Fisher Scientific, cat. #63-1038-42, RRID: AB_2762548), and HLADR-PerCp-Cy5.5 (clone: L243, BioLegend, cat. #307630, RRID: AB_893567).

Humanized mice overall test FACS antibodies

Anti-mouse CD45.2-FITC (clone: 104, BioLegend, cat. #109806, RRID: AB_313442), anti-human CD45.2-AF700 (clone: 2D1, BioLegend, cat. #368514, RRID: AB_2566373), anti-human CD3-PerCp-Cy5.5 (clone: HIT3a, BioLegend, cat. #300328, RRID: AB_1575008), anti-human CD19-PE-Cy7 (clone: HIB19, BioLegend, cat. #302216, RRID: AB_314245), and anti-human CD56-PE (clone: HCD56, BioLegend, cat. #318306, RRID: AB_604101).

FM471 tumors and KP and KPS tumors of C57BL/6 mice were stained with the following antibodies

CD45.2-AF700 (clone: 104, BioLegend, cat. #109822, RRID: AB_493730), CD4-BV510 (clone: RM4.5, BD, cat. #563106, RRID: AB_2687550), CD8-FITC (clone: 53-6.7, BioLegend, cat. #100706, RRID: AB_312744), CD3-PerCP.Cy5.5 (clone: 17A2, BD, cat. #560527, RRID: AB_1727463), CD11b-BV510 (clone: M1/70, BioLegend, cat. #101263, RRID: AB_2561390), CD3-FITC (clone: 17A2, BD, cat. #555274, RRID: AB_395698), CD19-FITC (clone: 1D3, BD, cat. #553785, RRID: AB_396681), and CD11c-PE (clone: N418, BioLegend, cat. #117308, RRID: AB_313776). Dead cell staining was carried out with LIVE/DEAD Fixable Blue Dead Cell Stain (Invitrogen, cat. #L34961) at 1:1,000 dilution. For flow cytometry analysis, all events were acquired on a BD LSRFortessa X-20 analysis or Attune NxT flow cytometer (Thermo Fisher Scientific) and carried out on FlowJo v10 (RRID: SCR_008520).

Inducible ectopic expression of SMARCA4

eGFP as control or SMARCA4 were cloned into the pInducer20 doxycycline inducible lentiviral vector (RRID: Addgene_44012). Lentivirus was produced using standard virus production methods by cotransfecting target and packaging plasmids (psPAX2 – RRID: Addgene_12260, and pMD2.G- RRID: Addgene_12259) into HEK293T cells (RRID: CVCL_0063). H322 cells were transduced with 0.45 μmol/L filtered viral particles with polybrene (8 μg/mL). After 16 hours of transduction, media were replaced with fresh regular growth media. Two mg/mL G418 was selected for 48 hours. After selection, cells were termed stably transduced. GFP or SMARCA4 expression was induced with 48 hours 1 μg/mL doxycycline induction.

NF- κB shRNA knockdown

shRNA sequences targeting NF-κB were cloned into the LT3_EPIR (EGFP deleted) inducible vector system for human lung cancer cell lines (H2122 and H322) and S_EP (EGFP deleted) noninducible vector system for mouse lung cancer cell line (FM471). Lung cancer cell lines were infected with lentiviruses carrying NF-κB shRNAs, and the effect of NF-κB knockdown determined by immunoblot analysis.

RNA sequencing analysis

Bowtie2-build (32) was used to index the human reference genome (hg38). Raw sequencing reads were aligned to human reference genome and transcriptome gene annotation GENCODE V44 (33) using TopHat2 v2.1.1 (34). HTseq-count v0.11.0 (RRID: SCR_01186; ref. 35) was used to count the expression level of each gene.

Differentially expressed gene analysis was carried out with R package EdgeR v3.42.4 (36) using raw read count matrices with cutoff of FDR < 0.05 and absolute log₂-fold change >1.5. Genes were ranked by the log-transformed P values in differential expression analysis and set to negative/positive values for downregulation/upregulation, respectively. Preranked gene set enrichment analysis (GSEA; ref. 37) was performed on the ranked gene list to calculate the normalized enrichment score for hallmark gene sets. For heatmap generation showing differentially expressed genes, CPM values were transformed into z-score and ComplexHeatmap, an R package was used to draw heatmaps (38).

ATAC-seq analysis

Assay for transposase-accessible chromatin using sequencing (ATAC-seq) fastq reads were aligned to the human genome hg38 using Burrows–Wheeler Aligner (39). SAMtools v1.9 (RRID: SCR_002105; ref. 40) was used to convert SAM files into BAM format. PCR duplicates were remove using Picard tool (v.2.27.4) MarkDuplicates function with “validation_stringency = lenient remove_duplicates = true” options.

SAMtools v1.9 was used to remove reads aligning to the mitochondrial genome and incomplete assemblies and filter mapping quality using the parameter -q 30 -F 1804. The regions of ENCODE hg38 blacklist (https://www.encodeproject.org/annotations/ENCFF356LFX/) were filtered out.

Replicate coverage files were merged using bigWigMerge function from deepTools v3.5.4 (41) for visualization in Integrative Genome Viewer (IGV) browser.

ATAC-seq broad peaks were called using MACS2 v2.1.1 (42) software using parameters “–keep-dup all” and BAMPE option. The bigWig files were generated using deepTools for visualization in IGV (43).

Peaks identified in replicates were combined together using bedtools (44) to generate intersect peaks. Peaks were annotated using “annotatePeaks” function from HOMER v4.8 (45), and associated genes from each peak were identified. The lost peaks, gained peaks, and common peaks were identified using bedtools intersect function. The peak overlapping analysis, Venn diagram visualization, and peak profile barplot were generated using in-house R script (R v4.3.0).

For data normalization and visualization, the BAM files were converted to the bigWig format using bamCoverage with RPKM normalization. The heatmaps and average profiles were generated using the plotHeatmap and plotProfile scripts from deepTools v3.5.4 (41).

Genomic Regions Enrichment of Annotation Tool (46) was used to collect genes associated with lost peaks. The collected genes were used to find enriched pathways using online tool Enrichr (47). Motif analysis of genes associated with pathways was done using Homer (findMotifs.pl). Enhancer peaks were identified as the intersection of ATAC-seq peaks and H3K27acetylation (H3K27ac) peaks using bedtools.

CUT&RUN

CUT & RUN was performed using CUTANA CUT & RUN Kit v4 (EpiCypher, cat. #13-2002) according to the manufactures’ instruction.

Briefly, 0.5 × 10⁶ H1975 cells per antibody/condition were harvest and washed twice with wash buffer (20 mmol/L HEPES, pH 7.5, 150 mmol/L NaCl, and 0.5 mmol/L spermidine). Subsequently, the cells were incubated with activated concanavalin A–coated magnetic beads (EpiCypher, cat. #21-1401) overnight at 4°C on a nutator with primary antibody at a 1:50 dilution (wash buffer +0.05% of digitonin, 2 mmol/L EDTA). Following antibody incubation, pAG-MNase was introduced to each reaction, mixed, and allowed to incubate at room temperature for 10 minutes. A measure of 100 mmol/L CaCl₂ were gently added to each reaction while kept on ice, followed by a 2-hour incubation at 4°C. The reaction was quenched by adding stop buffer (340 mmol/L NaCl, 20 mmol/L EDTA, 4 mmol/L EGTA, 50 μg/mL RNase, and 50 μg/mL glycogen). E. coli Spike-in DNA served as an internal control. After 10 minutes of incubation at 37°C, cells and beads were pelleted by centrifugation and placed on magnet to collect the supernatant. The supernatant was transferred to a fresh tube, and DNA was purified using CUT&RUN DNA Purification Kit. The released DNA was quantified using Qubit (Thermo Fisher Scientific, cat. #Q32851). CUT&RUN libraries were constructed using NEBNext Ultra II DNA Library Prep Kit (New England Biolabs, cat. #E7645S) according to the manufacturer’s instructions. Fragment size analysis of the libraries was conducted using TapeStation. Equimolar libraries were pooled together into one tube and sequenced on NextSeq500 platform with pair-end reads (2 × 75 bp).

FASTQ files were quality checked using FastQC v0.11.8 (RRID: SCR_014583; https://www.bioinformatics.babraham.ac.uk/projects/fastqc/), and paired-end reads were aligned to the hg38 genome using Bowtie2 v2.4.5 (32). Picard v2.23.8 (RRID: SCR_006525; http://broadinstitute.github.io/picard/) was used to remove PCR duplicates using the “MarkDuplicates function.” SAMtools v1.9 was used to remove reads aligning to the mitochondrial chromosome, hg38 blacklisted regions, and incomplete assemblies and filter mapping quality using the parameter “-q 30 -F 1804.” The read coverage, bigwig files were created with deepTools bamCoverage (41). Replicate coverage files were merged using “bigWigMerge” function from deepTools v3.5.4 (41) for visualization in IGV browser. MACS2 v2.1.1 (42) was used to call narrow peaks using parameters “–keep-dup auto” and BAMPE option.

Peaks identified in replicates were combined together using bedtools (44) to generate intersect peaks. The lost peaks were identified using bedtools intersect function. GREAT tool (46) was used to annotate the peaks in the long-distance region.

To find overlapping peaks, peak regions, called by MACS2 for SMARCA4 and NF-κB, were uploaded in BED format into the R environment in the package ChIPpeakAnno (v3.34.1; ref. 48) and used as input for the command “findOverlapsOfPeaks” with the settings minoverlap = 100 and connectedPeaks = “keepAll”). The overlapping peaks were then plotted using the “makeVennDiagram” command, and a hypergeometric test was performed to test the significance of overlap between the peaksets.

The lost peaks were then plotted using the “makeVennDiagram” command from ChIPpeakAnno (48), a Bioconductor package. For data normalization and visualization, BAM files from Cut&Run were converted to bigWig format using bamCoverage with RPKM normalization. The heatmaps and average profiles were generated using the plotHeatmap and plotProfile scripts from deepTools v3.5.4 (46).

Expression analysis of TCGA data in patient dataset

TCGA lung adenocarcinoma (TCGA-LUAD) expression data were downloaded using the R package TCGAbiolinks (version 2.28.4; ref. 49) in order to find whether innate immune response gene expression has association with low SMARCA4 expression. A total of 530 samples were used for expression analysis. Mutation information for each sample was retrieved from the cBioPortal for Cancer Genomics. Of 530 samples, 17 were found to be having class 1 mutation (nonsense mutation), and the rest were treated as wild-type (WT). The highest value of SMARCA4 expression in SMARCA4 nonsense mutant samples was taken as threshold for low expression. Of 530 samples, 91 were considered as low SMARCA4–expressing samples and 439 were treated as high SMARCA4–expressing samples. Raw expression value of gene expression was converted into counts per million and log-transformed. A two-sided Wilcoxon rank-sum test was applied to see whether there is an association between low SMARCA4/BRG1 expression status and inflammatory gene expression. Differential gene expression between low SMARCA4 and high SMARCA4 was performed by using DESeq2 R package (version 1.40.2; RRID: SCR_000154; ref. 50).

GSEA was performed using GSEA v 4.3.2., with the number of permutations as 1,000. We examined hallmark gene sets and the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway. The ranked list of genes for GSEA was generated using DESeq2 based on scores calculated as -log₁₀(FDR) × sign[log₂(fold change)].

Data availability

The RNA sequencing (RNA-seq) data generated in this study were deposited in the Gene Expression Omnibus (GEO; RRID: SCR_005012) repository at GSE269551 for H2122 bulk tumors, GSE278868 for H2122 tumor cells in vitro, and GSE269618 and GSE269930 for H1975, HCC44, and H1792 tumor cells in vitro. The ATAC-seq datasets analyzed in this study were deposited in the GEO repository at GSE269686 for H1975 and HCC44 tumor cells in vitro. The CUT&RUN datasets of H1975 tumor cells in vitro in this study were deposited in the GEO repository at GSE278869. The chromatin immunoprecipitation sequencing (ChIP-seq) H3K27ac datasets of H1975 tumor cells in vitro in this study were deposited in the GEO repository at GSE288649. TCGA-LUAD expression data (7) were obtained from (https://www.cancer.gov/ccg/research/genome-sequencing/tcga). All other regent or resource information can be found at Supplementary Key resource table. All other raw data generated in this study are available upon request from the corresponding author.

Results

SMARCA4 deficiency leads to decreased response to anti-PD-1 immunotherapy and impaired immune cell infiltration into tumors

To study the role of SMARCA4 loss on antitumor immunity, we generated immune-humanized mouse models by reconstitution of umbilical cord–derived purified CD34⁺ HSCs into hu-NSG mice as described previously (25). Briefly, freshly isolated CD34⁺ HSCs from cord blood were injected intravenously into mice after 24 hours of irradiation (Fig. 1A). After 4 to 6 weeks, mice that had more than 25% human CD45⁺ cells in peripheral blood were considered humanized, hu-NSG mice. Moreover, human immune cell populations including CD3⁺ T cells, CD19⁺ B cells, and CD56⁺ NK cells (Supplementary Fig. S1A–S1D) were determined in peripheral blood and CD4⁺ and CD8⁺ T cells were determined in spleen of hu-NSG mice (Supplementary Fig. S1E).

Figure 1.

SMARCA4 deficiency leads to decreased response to anti-PD-1 immunotherapy and impaired immune cell infiltration into tumors in humanized xenograft models. A, Schematic representation detailing the development of the humanized mouse model. Parental and SMARCA4 KO H2122 tumors were introduced into humanized mice at 6 weeks after human CD34⁺ stem cells implantation, with a minimum of 25% huCD45⁺ cells observed in peripheral blood. Parental xenograft models were subjected to i.p. injections of vehicle or pembrolizumab (250 μg/mouse) every 3 days for 3 weeks, whereas SMARCA4 KO xenograft models were treated in the same manner but for 2 weeks. B, Immunoblots of parental and SMARCA4 KO H2122 cell lysates stained using anti-SMARCA4 and β-actin antibodies. Protein molecular weight markers are in kilodaltons (kDa). C and D, Growth curves of parental H2122 xenograft models (n = 6 animals per group; C) or SMARCA4 KO H2122 xenograft models (n = 5 animals per group; D) after treatment with isotype control antibodies or anti-PD-1 antibodies were generated. Data points represent the mean ± SEM. *, P < 0.05, two-sided unpaired t test. E–K, The cells isolated from dissociated subcutaneous parental and SMARCA4 KO H2122 xenograft models were analyzed by flow cytometry. E–G, Flow cytometry plot of HLA-DR+ DCs (E). The percentage of HLA-DR+ DCs out of CD11b⁺ cells (F) and CD103⁺ cells out of CD3⁻CD19⁻ cells (G). H–K, Flow cytometry plot of CD4⁺ T cells and CD8⁺ T cells (H and I). The percentage of CD4⁺ T cells (J) and CD8⁺ T cells (K) out of CD3⁺ T cells. Data points represent the mean ± SEM. *, P < 0.05; **, P < 0.01, one-tailed Mann–Whitney test. ns, not significant. (A, Created with BioRender.com.)

We next injected isogenic pair of parental and SMARCA4 KO H2122 human lung cancer cells subcutaneously into hu-NSG mice (Fig. 1A and B). H2122 cells have been reported to exhibit moderate response to anti-PD-1 treatment (51). When tumors reached around 200 mm³, hu-NSG mice were treated with vehicle or anti-PD-1 antibodies every 3 day for 2 to 3 weeks. Whereas parental tumors showed moderate but significant tumor growth inhibition, SMARCA4 KO tumors were fully resistant to anti-PD-1 treatment (Fig. 1C and D).

Next, we determined the repertoire of tumor infiltrating immune cells by using flow cytometry–based immune profiling of dissociated tumors. This revealed a significant reduction of total DCs, CD103⁺ cDC1 cells, and CD4⁺ T cells and trend toward reduced CD8⁺ T cells in SMARCA4 KO tumors (Fig. 1E–K).

To confirm and expand on these observations, we utilized an orthogonal experimental model with mouse lung cancer cells, FM471 (29). These cells were derived from lung adenocarcinomas of C57BL/6 mice subjected to tobacco smoke, have Kras and p53 mutations, and harbor high TMB and thus are an excellent model system to investigate the immune system in immunocompetent mice against a high TMB tumor type such as lung cancer. We injected parental and isogenic Smarca4 KO FM471 cells subcutaneously in C57/BL6 mice and treated the mice with isotype control or anti-PD-1 antibodies once daily for four times (Fig. 2A and B). Whereas parental tumors showed robust response to anti-PD-1 therapy with greater than 50% tumor growth inhibition, Smarca4 KO tumors had minimal response of about 20% reduction (Fig. 2C; Supplementary Fig. S2A). We then performed comprehensive FACS profiling of infiltrating immune cells. Similar to our previous observation, the most consistent and significant differences were profound reduction of total DCs, cDC1 cells (Fig. 2D–F; Supplementary Fig. S2B), and CD4⁺ T cells in Smarca4 KO tumors (Fig. 2G–I).

Figure 2.

Smarca4 deficiency leads to decreased response to anti-PD-1 immunotherapy and impaired immune cell infiltration into tumors in mouse syngeneic models. A, Schematic of immunotherapy experiments setup for FM471 tumor model. Parental and Smarca4 KO FM471 cancer cell lines were subcutaneously injected into 8-week-old C57BL/6 female mice (n = 20 animals per group). Isotype control antibodies or anti-PD-1 antibodies were treated after tumors were established at day 4 after implantation. The mice were given treatment once daily four times. B, Immunoblots of parental and Smarca4 KO FM471 cell lysates stained using anti-SMARCA4 and β-actin antibodies. Protein molecular weight markers are in kilodaltons (kDa). C, Analysis of the percent change in tumor volume of parental and Smarca4 KO FM471 syngeneic models after treatment with isotype control antibodies or anti-PD-1 antibodies. D–I, Cells were isolated from dissociated subcutaneous parental and Smarca4 KO FM471 syngeneic models. D–F, Flow cytometry plot of total CD11c⁺ cells (D). Percentage of DCs defined by total CD11c⁺ gating from CD11b⁺ cells (E) and cDC1 cells defined by inter medium CD11c⁺ and CD11b⁺ cells (F). G–I, Flow cytometry plot of CD4⁺ T cells and CD8⁺ T cells (G). The percentage of CD4⁺ T cells (H) and CD8⁺ T cells (I) out of CD3⁺ T cells. Data points represent the mean ± SEM. *, P < 0.05; **, P < 0.01; ***, P < 0.001, two-sided unpaired t test. ns, not significant. (A, Created with BioRender.com.)

To further confirm our conclusions in an orthotopic model of lung cancer, we utilized a genetically engineered mouse model that we have previously established termed KPS (5). Briefly, this model allows selective genetic manipulation of floxed alleles within the lung epithelium by intranasal inhalation of Adenovirus-encoding Cre under the CC10 promoter, resulting in inactivation of Smarca4 and p53 and activation of mutant Kras (Kras^LSLG12D/WT, p53^fl/fl, Smarca4^fl/fl; Supplementary Fig. S3A), following previously established protocols (27). As control, we used established KP mice (Kras^LSLG12D/WT, p53^fl/fl). Three months after Cre administration, mice were imaged with micro-CT that showed lung cancer development in the thoracic cavity of KP and KPS mice (Supplementary Fig. S3B). Importantly, we were able to corroborate our previous results in this orthotopic model and found that KPS tumors showed significant reduction of cDC1 cells and CD4⁺ T cells (Supplementary Fig. S3C).

Taken together, the three orthogonal model systems showed consistent and significant reduction in DCs and CD4⁺ T cells, whereas the frequency of other immune cells was not changed, minimally altered, or showed mixed alteration. Hence, we decided to focus our next investigations on how loss of SMARCA4 in cancer cells results in diminished DCs and CD4⁺ T cells within the TME.

To gain deeper insights into the mechanistic basis for the impaired response to immunotherapy and compromised recruitment of DCs and T cells, we performed RNA-seq on H2122 parental and SMARCA4 KO tumors. Differential gene expression analysis showed 485 genes were upregulated and 612 were downregulated in SMARCA4 KOs, with SMARCA4, as expected, being the most downregulated gene (Fig. 3A and B). GSEA of the most significantly downregulated genes in SMARCA4 KO tumors showed IFNα and IFNγ pathways as the top enriched pathways (Fig. 3C–E). As the above expression analysis was done on bulk tumors, to further confirm that the altered IFNα and IFNγ pathways are tumor cell–autonomous, we performed RNA-seq on H2122 parental and SMARCA4 KO tumor cells in vitro. Consistently, the IFNα and IFNγ pathways are downregulated in SMARCA4 KO tumor cells (Fig. 3F and G). IFNα belongs to the type I IFN family, which is downstream of the cGAS-STING pathway (52). The cGAS-STING pathway has emerged as the master mediator of inflammation in the settings of stress, tissue damage, infection, and cancer by sensing microbial and host-derived DNAs (collectively known as danger-associated molecules) and is an attractive therapeutic target against cancer (22, 53, 54). Notably, decreased STING1 expression in SMARCA4 KO tumors is confirmed by qRT-PCR analysis (Fig. 3H).

Figure 3.

Transcriptional analysis of xenograft models identifies SMARCA4-dependent inflammatory response. A–E, RNA-seq analysis of parental vs. SMARCA4 KO H2122 bulk tumors (n = 4 biological replicates). A, Volcano plot displaying differentially expressed genes assessed by RNA-seq in parental and SMARCA4 KO H2122 tumors with thresholds FDR < 0.05 and absolute log₂-fold change >1.5. Down, downregulation; up, upregulation. B, Heatmap presenting the top 50 upregulated and downregulated genes in parental tumors, with SMARCA4 being the most significantly upregulated gene. C, GSEA of hallmark pathways differentially enriched in parental vs. SMARCA4 KO tumors. D and E, Enrichment plots showing downregulation of IFNα (D) and IFNγ (E) pathways in SMARCA4 KO tumors. F and G, RNA-seq analysis of parental vs. SMARCA4 KO H2122 tumor cells in vitro (n = 3 biological replicates). Enrichment plots showing downregulation of IFNα (F) and IFNγ (G) pathways in SMARCA4 KO tumor cells. NES, normalized enrichment score. H, qRT-PCR analysis of STING1 expression in parental and SMARCA4 KO H2122 xenograft tumors after vehicle treatment, with statistical significance indicated. Data points represent the mean ± SEM. **, P < 0.01, two-sided unpaired t test.

SMARCA4 loss results in cancer cell–intrinsic defects in expression of innate immune and inflammatory genes

SMARCA4 is a key catalytic subunit of the SWI/SNF complex, which is required for proper chromatin accessibility and gene expression (55–57). Hence, we hypothesized that mutations or loss of SMARCA4 in cancer cells could result in downregulation of STING1 and other innate immune or inflammatory genes observed above. To conclusively determine whether there is a cancer cell–intrinsic dysregulation of gene expression, we performed qRT-PCR analysis of key genes such as STING1 in in vitro grown cancer cells. Strikingly, we noticed that STING1 mRNA expression was completely abolished in H2122 and HCC44 SMARCA4 KO cells (Fig. 4A–D). Next, parental and SMARCA4 KO H2122 and HCC44 cells were treated with STING1 activator cGAMP. STING1 expression is significantly increased in H2122 parental cells but not SMARCA4 KO cells (Fig. 4C and D). The cGAS-STING pathway can sense cytosolic DNA to activate innate immunity (52, 58) and initiate cytokines, such as IL1β and IL18 (59). In parental H2122 and HCC44 cells, double-stranded DNA-mimetic poly (dA:dT) markedly increased mRNA expression of STING1, IL1β, IRF3, and IL1β secretion in HCC44 (Fig. 4E–G). Consistently, IL1β-converting enzyme, CASPASE-1 mRNA expression is also profoundly increased by poly(dA:dT) treatment but not cGAMP treatment (Supplementary Fig. S4A–S4D). However, SMARCA4 KO completely abolished the expression of these genes in response to poly(dA:dT; Fig. 4E–G; Supplementary Fig. S4C and S4D). Next, we asked whether the upregulation of STING1 has actual functional consequences. One of the key downstream targets of the STING pathway that plays an active role in the recruitment of DCs is expression of IFNβ (22, 54). Indeed, cGAMP treatment of parental H2122 cells potently induced expression of IFNβ whereas this was completely abolished in SMARCA4 KOs (Fig. 4C). Similarly, stimulation with poly(dA:dT) massively upregulated IFNβ mRNA expression in SMARCA4 parental cells but was completely undetectable in SMARCA4 KOs (Fig. 4E). Additionally, the IFNα mRNA expression was highly induced by poly (dA:dT) treatment in H2122 and HCC44 parental cells but profoundly diminished in knocking out SMARCA4 KOs (Supplementary Fig. S4C and S4D). Importantly, protein levels of STING pathway components such as STING1, IFNβ, and phosphorylated IRF3 mirrored the results of mRNA levels (Fig. 4H; Supplementary Fig. S4E and S4F).

Figure 4.

SMARCA4 loss results in cancer cell–intrinsic defects in expression of innate immune and inflammatory genes. A and B, Immunoblots of parental and SMARCA4 KO H2122 (A) and HCC44 (B) cell lysates stained using anti-SMARCA4 and β-actin antibodies. C and D, Expression of STING1, IL1β, and IFNβ mRNA levels was evaluated in parental and SMARCA4 KO H2122 (C) and HCC44 (D) cells treated with control and 10 μg/mL 2′3′-cGAMP for 6 hours (n = 3 biological replicates) by qRT-PCR. E and F, STING1, IL1β, IFNβ, and IRF3-P mRNA expression levels were evaluated in parental and SMARCA4 KO H2122 (E) and HCC44 (F) cells treated with control and 5 μg/mL poly (dA:dT) overnight (n = 3 biological replicates) by qRT-PCR. G, HCC44 IL1β secretion was evaluated by ELISA. H, Immunoblot analysis of STING1, IL1β, IFNβ, IRF3, and IRF3-P in parental and SMARCA4 KO H2122 cells treated with control and 5 μg/mL poly (dA:dT) for 6 or 18 hours. I, Immunoblot showing inducible expression of SMARCA4 in SMARCA4-deficient human lung cancer cell line H322 by administration of doxycycline (DOX; 1 μg/mL) with GFP as control. J, mRNA expression of STING1, IL1β, and IFNβ by qRT-PCR in H322 control and SMARCA4 reconstituted cell line treated with control and 5 μg/mL poly (dA:dT) overnight (n = 3 biological replicates). Data points represent the mean ± SEM. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001; ns, not significant, two-sided unpaired t test. CON, control; Par, parental.

Type I IFNs have been widely demonstrated to be essential for cross-priming and generation of tumor-specific immunity acting through DCs, potentially explaining the lack of DCs and poor immune response in SMARCA4-mutant tumors (54, 60). To definitively show the requirement of SMARCA4, we next sought to determine whether reexpression of SMARCA4 can rescue lack of expression of STING1, IFNβ, and IL1β in SMARCA4-mutant cells. Hence, we inducibly reexpressed SMARCA4 in H322, a SMARCA4-mutant lung cancer cell line (Fig. 4I). Interestingly, reexpression of SMARCA4 was able to potently rescue STING1, IFNβ, and IL1β expression after poly (dA:dT) stimulation as compared with the parental cell line (Fig. 4J). CASPASE-1 and IFNα mRNA expression is not altered (Supplementary Fig. S4G). We recapitulated these observations also in our genetically engineered mouse model tumors whereby, compared with KP tumors, KPS tumors showed significantly downregulated expression of Smarca4 and Sting1 and trends toward reduced Ifnβ, Ifnα, Il1β, and Caspase-1 expression (Supplementary Fig. S5A–S5F).

Finally, we asked whether similar transcriptomic alterations exist in human patient samples. To answer this, we analyzed the TCGA lung cancer dataset as this dataset contains high-quality genomic and transcriptomic information (7). SMARCA4 is inactivated primarily by truncating or frameshift mutations that result in total/near total loss of activity of the protein. Additionally, there are cases in which SMARCA4 is epigenetically silenced that resulted in very low expression. In order to capture the relevance of both types of events, we have performed two different types of analysis. First, we have compared those samples with bona fide SMARCA4-inactivating mutations with WT tumors (Supplementary Fig. S6A–S6E, bottom). Next, we compared samples with low levels of SMARCA4 expression with the rest of the samples (Supplementary Fig. S6A–S6E, top). In both cases, we observed that samples with low SMARCA4 expression or mutant SMARCA4 have significantly reduced expression of STING1, NF-κB, and markers of DCs such as ITGAM (CD11b), ITGAX (CD11c), and ITGAE (CD103). Additionally, GSEA analysis revealed that SMARCA4-low tumors had significantly reduced immune-related pathways such as inflammatory response, IL2_STAT5 signaling, and IL6_JAK_STAT3_signaling (Supplementary Fig. S6F). These results strongly suggest that the observations in mouse models are broadly applicable to human patient tumors as well.

Next, we utilized a pharmacologic approach to inactivate SMARCA4 by using a potent small-molecule SMARCA4 degrader based on the PROTAC principle termed ACBI1 (61). Treatment of SMARCA4 WT cell lines (HCC44, H1792, and H1975) by ACBI1 completely degraded SMARCA4 protein levels compared with DMSO-treated cell lines. However, as expected, no obvious SMARCA4 protein levels were found in SMARCA4-mutant cells (Fig. 5A). GSEA revealed that top pathways downregulated in SMARCA4 WT (HCC44, H1792, and H1975) cells upon treatment with ACBI1 were inflammatory response, TNFα signaling via NF-κB, and IFNα response (Fig. 5B and C). Importantly, IL1β was the top downregulated gene by ACBI1 treatment in SMARCA4 WT cells (Fig. 5D). The results were confirmed by qRT-PCR in which IL1β expression was significantly reduced by SMARCA4 deficiency (Fig. 5E). Importantly, STING1 mRNA expression and protein level were also abolished by SMARCA4 degradation (Fig. 5F and G).

Figure 5.

SMARCA4 degradation abrogates cancer cell–intrinsic expression of innate immune and inflammatory genes. A, SMARCA4 WT cells (HCC44, H1792, and H1975) and SMARCA4 mutation cells (H322, H1693, and H1299) were treated with DMSO control or 1 μmol/L ACBI1 for 96 hours. Immunoblots of SMARCA4 and β-actin in SMARCA4 WT cells (left) and SMARCA4 mutation cells (right). B–D, RNA-seq analysis of SMARCA4 WT (HCC44, H1792, and H1975) cells treated with DMSO vs. 1 μmol/L ACBI1 for 96 hours. B, GSEA curves showing upregulation of the inflammatory response and TNFα_signaling_via_NF-κB pathway in DMSO treatment. NES, normalized enrichment score. C, GSEA of hallmark top enriched pathways in DMSO treatment. Inflammatory response and IFNα response are at the top positions. D, Heatmap presenting the top 50 upregulated and downregulated genes in DMSO treatment, with IL1β being the one of significantly upregulated genes. E, qRT-PCR analysis of IL1β expression from SMARCA4 WT cells (HCC44, H1792, and H1975) treated with DMSO vs. 1 μmol/L ACBI1 for 96 hours and SMARCA4 WT cells (HCC44, H1792, and H1975) vs. SMARCA4 mutation cells (H322, H1693, and H1299). F, STING1 mRNA expression from SMARCA4 WT cells (HCC44, H1792, and H1975) treated with DMSO vs. 1 μmol/L ACBI1 for 96 hours. G, Immunoblots of STING1 and VINCULIN in H1792 cells treated with 0.1 or 1 μmol/L ACBI1 for 96 hours. Data points represent the mean ± SEM. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001, two-sided unpaired t test. MUT, mutant.

Taken together, these results gained through genetic and chemical biology approaches firmly establish the requirement of SMARCA4 for basal as well as inducible expression of STING1, IFNβ, and IL1β and broadly for expression of innate immune genes in a cancer cell–intrinsic manner.

NF-κB transcription factor–binding motif is enriched at enhancers that lose chromatin accessibility in SMARCA4-deficient cells

Next, we asked how loss of SMARCA4 affected expression of the above-described genes. As SWI/SNF complexes actively remodel nucleosome DNA packaging, we performed ATAC-seq upon treatment with DMSO and ACBI1 in SMARCA4 WT cells, H1975 and HCC44. We used the chemical biology approach as we can degrade SMARCA4 rapidly and assess primary molecular events that will likely help us decipher the direct actions of SMARCA4. As expected, degradation of SMARCA4 resulted in genome-wide reduction in chromatin accessibility at more than 20,000 sites in H1975 and HCC44 (Fig. 6A and B). When we examined the distribution of lost chromatin peaks, we found the majority of the peaks to be mapped to intronic regions, followed by distal intergenic regions (Fig. 6A). Many of these sites coincided with enhancers (Fig. 6A–C). These changes at introns and distal intergenic by ACBI1 are consistent with SWI/SNF complexes localizing at enhancers to regulate gene expression (55, 62, 63). Hence, we next interrogated chromatin accessibility changes at putative enhancers by performing integrative analysis and identifying genomic loci having both ATAC-seq and H3K27ac ChIP-seq peaks. We then performed pathway analysis of the genes associated with those enhancers that lost chromatin accessibility upon SMARCA4 degradation. Pathway enrichment analysis from MsigDB revealed two prominent pathways, TNFα signaling via NF-κB and inflammatory response were associated with lost chromatin accessibility (Fig. 6D). Enhancers are associated with specific transcription factors. Hence, we asked what transcription factor–binding motifs were enriched in the genomic loci of enhancers in these two pathways. Interestingly, the NF-κB (also known as RelA or p65) motif was by far the most frequent and significantly observed sequence in the enhancers with lost accessibility (Fig. 6E). NF-κB transcription factor is a well-known and extensively studied regulator of expression of inflammatory genes (64–66). Thus, our data suggests that NF-κB and SMARCA4 could occupy the same regulatory enhancers and cooperate to drive expression of innate immune and inflammatory genes. To prove this, we next performed CUT&RUN using NF-κB and SMARCA4 antibodies and determined chromatin occupancy by these two proteins in the H1975 cell line treated with DMSO or ACBI1. As expected, acute SMARCA4 degradation by ACBI1 resulted in significant reduction in chromatin accessibility and chromatin occupancy by SMARCA4 on an enhancer associated with IFNβ (Fig. 6F) and STING1 (Supplementary Fig. S7A). Importantly, NF-κB is also colocalized to the same enhancer in DMSO-treated cells but profoundly lost upon SMARCA4 degradation (Fig. 6F; Supplementary Fig. S7A).

Figure 6.

Loss of SMARCA4 abrogates chromatin accessibility at enhancers associated with innate immune inflammation and are co-occupied by NF-κB. SMARCA4 WT cells (HCC44 and H1792) were treated with DMSO control or 1 μmol/L ACBI1 for 96 hours. A, Venn diagrams illustrating alterations in chromatin accessibility upon ACBI1 treatment along with annotation of ATAC peaks distribution across the genomic regions. B, ATAC-seq read density heatmaps for total peaks with DMSO and ACBI1 treatment in HCC44 and H1975. C, ATAC-seq read density heatmaps for enhancer peaks with DMSO and ACBI1 treatment in HCC44 and H1975. D, Pathway enrichment analysis for lost peaks in HCC44 and H1975 with ACBI1 treatment. E, Examination of transcription factor motifs on enhancers with lost accessibility with ACBI1 treatment as identified in D. F, Genome browser tracks of chromatin accessibility at IFNβ loci (n = 2 biological replicates) in H1975 cells, following with representative integrative genome viewer tracks of views of the SMARCA4 and NF-κB CUT&RUN peaks on the IFNβ genomic locus. H3K27ac peaks derived from ChIP-seq denote putative enhancers. The overlapping peaks with differential intensity in the ACBI1 treatment group are highlighted. G, Endogenous interactions of SMARCA4 and NF-κB were analyzed by coimmunoprecipitation.

In additional to looking at specific genomic loci, we also performed global analysis of the number of overlapping peaks between SMARCA4 and NF-κB. We found that 80% (854/1063) of NF-κB peaks overlapped with SMARCA4-binding sites on the genome, showing an extensive overlapping of NF-κB and SMARCA4 (Supplementary Fig. S7B). We also found that SMARCA4 depletion with the ACBI1 treatment reduces NF-κB binding by 78% across the genome (Supplementary Fig. S7C). This was further exemplified by the significantly reduced CUT&RUN peak intensity of NF-κB when SMARCA4 was depleted with ACBI1 (Supplementary Fig. S7D). Next, we determined whether NF-κB and SMARCA4 can physically interact. For this, we performed coimmunoprecipitation experiments in H2122 parental cell lines after stimulation with LPS for 4 hours by immunoprecipitating using SMARCA4 antibody followed by probing for NF-κB or immunoprecipitating by NF-κB antibody followed by probing for SMARCA4. In both cases, we see robust signals indicating SMARCA4 and NF-κB can physically interact or at least exist in an interacting complex (Fig. 6G).

NF-κB loss partially recapitulates the effects of SMARCA4 depletion on innate immune gene expression, immune cell infiltration, and response to anti-PD-1 treatment

Because interplay between SMARCA4 and NF-κB regulates innate immune and inflammatory gene expression, we asked whether NF-κB loss recapitulates the impact of SMARCA4 depletion on immune response. To this end, we engineered human lung cancer cell lines (SMARCA4 WT cell line, H2122, and SMARCA4 reexpressing cell line, H322) with doxycycline-inducible shRNA constructs targeting NF-κB and mouse lung cancer cell line FM471 with shRNA constructs targeting Nf-κb. In all three lung cancer cell lines, induction of NF-κB shRNAs produced highly efficient depletion of NF-κB protein (Fig. 7A–C). Then we treated control and NF-κB shRNA containing H2122 and H322 cells with cGAMP or poly (dA:dT). Importantly, we found that NF-κB depletion inhibited IFNβ and IL1β expression under stimulation (Fig. 7D–F). Finally, we subcutaneously implanted control and Nf-κb shRNA containing FM471 cells in C57BL/6 mice and treated mice with isotype control or anti-PD-1 antibodies once daily for four times (Fig. 7G). Similar to our previous observation, control tumors showed significant tumor growth inhibition response to anti-PD-1 treatment whereas Nf-κb knockdown tumors had minimal response (Fig. 7H). Importantly, FACS profiling showed that cDC1 cells and CD4⁺ T cells are significantly reduced in Nf-κb knockdown tumors, whereas total DC percentage was unchanged (Fig. 7I–M). CD8⁺ T-cell percentage is increased by Nf-κb knockdown in tumors (Fig. 7L and N).

Figure 7.

NF-κB loss abrogates cancer cell–intrinsic expression of innate immune and inflammatory genes. A–C, Immunoblots of control and NF-κB knock down H2122 (A), SMARCA4 reexpressing H322 (B), and FM471 cell lysates stained using anti–NF-κB and VINCULIN antibodies (C). DOX, doxycycline. D and E, Expression of STING1, IFNβ, and IL1β mRNA levels were evaluated in control and NF-κB shRNA containing H2122 cells treated with or without 10 μg/mL 2′3′-cGAMP (D) and 5 μg/mL poly (dA:dT) (n = 3 biological replicates; E) by qRT-PCR. F, STING1, IFNβ, and IL1β mRNA expression levels were evaluated in control and NF-κB shRNA containing SMARCA4 reexpressing H322 cells (n = 3 biological replicates) treated with with 5 μg/mL poly (dA:dT). G, Schematic of immunotherapy experiments setup for the FM471 tumor model. Control and Nf-κb shRNA–containing FM471 cancer cell lines were subcutaneously injected into 8-week-old C57BL/6 female mice (n = 10 animals per group). Isotype control antibodies or anti-PD-1 antibodies were treated after tumors were established at day 4 after implantation. The mice were given treatment once daily four times. H, Analysis of the percent change in tumor volume of control and Nf-κb shRNA–containing FM471 syngeneic model after treatment with isotype control antibodies or anti-PD-1 antibodies. I–N, Cells were isolated from dissociated subcutaneous control and Nf-κb shRNA–containing FM471 syngeneic models. I–K, Flow cytometry plot of total CD11c⁺ cells (I). Percentage of DCs defined by total CD11c⁺ gating from CD11b⁺ cells (J) and cDC1 cells defined by inter medium CD11c⁺ and CD11b⁺ cells (K). L–N, Flow cytometry plot of CD4⁺ and CD8⁺ T cells (L). The percentage of CD4⁺ T cells (M) and CD8⁺ T cells (N) out of CD3⁺ T cells. Data points represent the mean ± SEM. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ns, not significant, two-sided unpaired t test. CON, control. (G, Created with BioRender.com.)

In conclusion, our results demonstrate that SMARCA4 is an essential regulator of cancer cell–intrinsic expression of innate immune and inflammatory genes, immune cell infiltration, and response to anti-PD-1 immunotherapy and that NF-κB, at least partially, is a key conspirator in this process.

Discussion

SMARCA4-mutant lung cancer has emerged as a distinct clinicopathologic entity with dismal prognosis and poor response to immunotherapy (4, 8). In additional to primary resistance, SMARCA4 is now implicated in acquired resistance to immunotherapy (9), suggesting a strong cancer cell–intrinsic mechanism whereby SMARCA4 loss gives fitness advantage to tumors to escape antitumor immunity. In addition to lung cancer, SMARCA4 is frequently altered across most solid tumors, including endometrial carcinoma (10%), esophageal (6%), stomach (5%), and bladder (5%) cancers with a long tail of 1% to 2% mutation across many other solid tumors (cBioportal compilation; ref. 67). Due to the dismal prognosis and large patient populations affected, understanding mechanisms of resistance and targeting SMARCA4-mutant tumors is a large unmet medical need.

In this study, we provide a cancer cell–intrinsic mechanistic basis for poor response of SMARCA4-mutant tumors to immunotherapy. Specifically, we demonstrate a mechanism by which SMARCA4 loss leads to profound epigenetic dysregulation that primarily involves enhancers, resulting in transcriptional deregulation. Part of this dysregulation involves well-studied and validated modulators of the innate immune system with roles in antitumor immunity such as STING1 and type I IFNs.

We have utilized several orthogonal and complementary model systems to show robust and reproducible molecular events associated with SMARCA4 loss. Whereas model-specific cellular and molecular alterations are evident, such as increase in CD3⁺ T cells solely in the FM471 model, we have primarily focused our investigations on recurring themes that we observed across all model systems utilized that are more likely to be of functional relevance. In this instance, we have observed reduced response to anti-PD-1 treatment and decreased infiltration of DCs and CD4⁺ T cells in H2122 SMARCA4 KO xenograft tumors in immune-humanized as well as FM471 Smarca4 KO model systems.

Type I IFNs and inducers of IFN response such as STING1 are central regulators of innate immune sensing of tumors and activators of DCs (16, 21). Using genetic loss-of-function and gain-of-function approaches as well as pharmacologic tools, we demonstrate the requirement of SMARCA4 for proper expression of STING1, type I IFNs at basal and stimulated levels. Tumors that have higher levels of DC infiltration tend to show improved responses to immunotherapies. Moreover, signals originating from cancer cells are necessary for DC recruitment and immune response. There are multiple evidences that type I IFNs within tumors is critical for cDC1s activation (68, 69). However, the specific mechanism by which type I IFN contributes to activating DCs still remains unclear. Positive correlations between STING1 expression in tumors and immune cell infiltration have been observed in the cancer genome atlas program (TCGA) database (70). In our study, SMARCA4 deficiency decreased STING pathway activation and type I IFN secretion, likely negatively affecting DC recruitment into the TME and contributing to resistance to anti-PD-1 immunotherapy. Importantly, several markers of cDC1 (CD11b, CD11c, and CD103) as well as STING1 are also decreased in SMARCA4 mutants/low expressors in the TCGA-LUAD dataset, indicating our observations in mouse models can be applicable in human patients.

By performing epigenomic profiling, we identified deregulation of enhancers associated with innate immune genes and inflammatory genes in SMARCA4-deficient states. Our finding of enhancer deregulation is consistent with previous observations that showed loss of SWI/SNF activity mostly alters enhancer targeting (56, 71). Importantly, our integrative analysis strongly suggests that NF-κB transcription factor and SMARCA4 are colocalized on enhancers of genes such as STING1 and IFNβ in cancer cells and regulate STING pathway activation. These data are in agreement with multiple studies demonstrating that STING pathway senses cytosolic DNA and cGAMP to induce type I IFN and inflammatory cytokine gene expression via transcription factor NF-κB (52, 72). Taken together, we provide a working model whereby SMARCA4 deficiency leads to enhancer shut down at critical innate immune genes and inflammatory cytokines dampening the recruitment of DCs and T cells, culminating in an immune-cold phenotype and lack of response to immunotherapy (Fig. 8). Whereas our findings are highly interesting, a number of questions remain open for future investigations. The detailed functional interplay including direct physical interaction and dynamics of recruitment and cooperativity between SMARCA4 and NF-κB are active areas for future studies. Whereas we have focused our investigations mainly on cancer cell–intrinsic dysregulations, it is likely that dysfunctions exist within the immune cells in SMARCA4-mutant tumors and need to be systematically investigated. Additionally, SMARCA4-mutant tumors have distinct metabolic features, including heightened oxidative phosphorylation (5). To date, it is unknown whether and how these metabolic alterations affect antitumor immunity of SMARCA4-mutant tumors. Lastly, a series of recent reports have identified role of SWI/SNF in various aspects of immune regulation. The Hargreaves laboratory showed that inhibition of the SWI/SNF complex disrupts activation of gene expression in response to bacterial endotoxin in macrophages with implications in enhancer deregulation (71). The core epigenomic dysregulation observed in this report are similar to our observations suggesting conserved mechanisms. Interestingly, a recent publication demonstrated how loss of ARID1A, a subunit of a particular type of the SWI/SNF complex termed the canonical BAF complex, is associated with enhanced immunotherapy response and increased cGAS-STING pathway activation (73), which is in stark contrast to our findings of SMARCA4 mutation. It is critical to note that there are three distinct subtypes of the SWI/SNF complex, the canonical BAF (BRG-/BRM-associated factor) complex and polybromo-associated BAF or noncanonical BAF complexes with distinct functional roles (74). ARID1A exists only in the canonical BAF complex, whereas SMARCA4 is the catalytic subunit in all three subunits (74). Thus, the biochemical and epigenomic consequences of loss of ARID1A and SMARCA4 are expected to be divergent. Taken together, these reports from the Hargreaves and Adelman studies and our report suggest distinct roles for each subtype of the SWI/SNF complex, including potential antagonistic or competitive functions at various genomic loci. Importantly, another report has shown that ARID1A interacts with mismatch repair protein MSH2, and its loss is correlated to microsatellite instability and increased antitumor immunity (75), further highlighting its distinctiveness.

Figure 8.

Schematic summarizing working model. Left, in SMARCA4 WT cancer cells, cytosolic DNA sensing can be efficiently transmitted into transcriptional responses involving STING1, type I IFN, and inflammatory cytokines, leading to recruitment of DCs and T cells. Right, in contrast, in SMARCA4 mutant cells, defective transcriptional responses prevent efficient recruitment of immune cells into the TME. (Created with BioRender.com.)

In conclusion, our study provided the mechanistic basis for resistance of SMARCA4-mutant tumor to immune checkpoint blockade therapy. Based on our findings, therapeutic efforts to boost DCs or STING1 signaling might help overcome poor response of SMARCA4-mutant tumors to immunotherapy.

Supplementary Material

Supplementary Data

Supplementary results

Authors’ Disclosures

Y. Wang reports grants from the NIH during the conduct of the study. M. Qudratullah reports grants from the NIH during the conduct of the study. S. Kotagiri reports grants from the NIH during the conduct of the study. Y. Han reports grants from the NIH during the conduct of the study. J.A. Roth reports grants and personal fees from Genprex during the conduct of the study. Y. Lissanu reports grants from the NIH during the conduct of the study and has a patent for WO 2023/129506 A1 pending. No disclosures were reported by the other authors.

Authors’ Contributions

Y. Wang: Conceptualization, data curation, software, formal analysis, validation, investigation, methodology, writing–original draft, project administration, writing–review and editing. I.M. Meraz: Data curation, formal analysis. M. Qudratullah: Data curation, software. S. Kotagiri: Investigation. Y. Han: Investigation. Y. Xi: Software, formal analysis. J. Wang: Data curation, formal analysis. K.C. Akdemir: Data curation, software. J.A. Roth: Conceptualization, investigation. Y. Lissanu: Conceptualization, supervision, funding acquisition, project administration, writing–review and editing.