Interferon response and epigenetic modulation by SMARCA4 mutations drive ovarian tumor immunogenicity Authors: Melica Nourmoussavi Brodeur, Higinio Dopeso, Yingjie Zhu, Ana Leda F. Longhini, Andrea Gazzo, Siyu Sun, Richard P. Koche, Rui Qu, Laura Rosenberg, Pierre-Jacques Hamard, Yonina Bykov, Hunter Green, Laxmi Gusain, Katherine B. Chiappinelli, Melih Arda Ozsoy, M. Herman Chui, Thais Basili, Rui Gardner, Sven Walderich, Elisa DeStanchina, Benjamin Greenbaum, Mithat Gönen, Nicolas Vabret, Britta Weigelt, Dmitriy Zamarin Published: 2024 ABSTRACT Cell-intrinsic mechanisms of immunogenicity in ovarian cancer (OC) are not well understood. Damaging mutations in the SWI/SNF chromatin remodeling complex, such as SMARCA4 (BRG1), are associated with improved response to immune checkpoint blockade; however, the mechanism by which this occurs is unclear. We found that SMARCA4 loss in OC models resulted in increased cancer cell–intrinsic immunogenicity, characterized by up-regulation of long-terminal RNA repeats, increased expression of interferon-stimulated genes, and up-regulation of antigen presentation machinery. Notably, this response was dependent on STING, MAVS, and IRF3 signaling but was independent of the type I interferon receptor. Mouse ovarian and melanoma tumors with SMARCA4 loss demonstrated increased infiltration and activation of cytotoxic T cells, NK cells, and myeloid cells in the tumor microenvironment. These results were recapitulated in BRG1 inhibitor–treated SMARCA4-proficient tumor models, suggesting that modulation of chromatin remodeling through targeting SMARCA4 may serve as a strategy to overcome cancer immune evasion. INTRODUCTION Immune targeted therapies, such as immune checkpoint blockade (ICB), have shown marked success in multiple cancer types, including some gynecologic cancers like endometrial (1, 2) and cervical cancers (3). To date, ICB trials in patients with ovarian cancer (OC) have yielded disappointing response rates with no clear biomarker for response (4). Nonetheless, several studies have demonstrated that a subset of OCs is characterized by an immunogenic tumor microenvironment (TME) that could make them sensitive to ICB (4, 5). As OC is the most lethal gynecologic cancer (6), a better understanding of the tumor cell–intrinsic mechanisms mediating immune response and resistance to immune cell killing is needed. In several cancer types, epigenetic modifications have been linked to changes in the immune TME and increased clinical benefit of ICB (7–10). Many of these epigenetic changes are directly or indirectly linked to damaging mutations in genes encoding subunits of the SWItch/Sucrose Non-Fermentable (SWI/SNF) chromatin remodeling complex (11–13). This complex normally enables selective gene expression and DNA repair through modulation of DNA compaction and accessibility (14). With more than 20% of human cancers harboring mutations in SWI/SNF genes, its implication in cancer development through dysregulation of oncogenes and tumor suppressor genes has started to be uncovered (14). In particular, SMARCA4, the catalytic subunit of the chromatin remodeler, ARID1A, SMARCB1, and PBRM1 are examples of SWI/SNF genes that have been associated with cancer outcomes and response to therapy (14). For example, SMARCB1-mutated rhabdoid tumors have been shown to exhibit increased immune cell infiltration and immune cell killing function (11). Furthermore, a clinical study evaluating ICB (anti–PD-L1) response in lung cancers showed increased response in SMARCA4-mutated compared to that in SMARCA4–wild-type tumors (13). We and others have demonstrated similar immune changes in the TME of SWI/SNF-deficient OCs (5, 15); however, the mechanism by which this immune response occurs is still not well understood. In particular, a subset of aggressive rhabdoid-like OCs, called small cell carcinoma of hypercalcemic type, harbors inactivating mutations in SMARCA4, leading to loss of the protein product known as BRG1 (5). Despite a low tumor mutational burden (TMB), one of the known predictive biomarkers of ICB response (16), these SMARCA4-mutated tumors have prominent T cell infiltration and PD-L1–expressing tumor-associated macrophages, and case reports show clinical responses to ICB (5). The mechanisms underlying this immunogenicity remain unclear. Here, we sought to examine the alterations in tumor cell–intrinsic and extrinsic immune signaling caused by SMARCA4 loss using syngeneic OC murine models. We demonstrate that the underlying immunogenicity in SMARCA4-mutated OCs is driven by up-regulation of downstream type I interferon (IFN)–stimulated genes (ISGs) and major histocompatibility complex (MHC) class I expression. In animal models, this led to delayed tumor growth and was associated with increased tumor infiltration of dendritic cells and activated T cells. Treatment of SMARCA4–wild-type cells/tumors with BRG1-targeted inhibitors and degraders resulted in activation of type I IFN signaling in cancer cells, decreased tumor burden, and enhanced tumor immune infiltration. Our findings generate a rationale for exploration of alterations in SWI/SNF-mediated signaling as biomarkers of response to immunotherapy and for targeting of BRG1 as a potential strategy to enhance tumor cell immunogenicity as a therapeutic strategy. RESULTS Pan-cancer association of SMARCA4 expression with immune expression signatures To investigate whether levels of SMARCA4 gene expression are associated with differential immune signatures in cancer, we compared immune-related properties defined by Thorsson et al. (17) between SMARCA4-low– and SMARCA4-high–expressing cancers across 29 tumor types [PanCancerAtlas dataset; (18)]. SMARCA4-low (lowest quartile) and SMARCA4-high (highest quartile) cases were defined within each tumor type to avoid confounding factors (see Materials and Methods). Comparison between SMARCA4-low (n = ~1900) and SMARCA4-high cases (n = ~1900) expressing tumors across cancer types revealed a significant increase in the immune scores, IFN-γ response (P < 0.0001; q < 0.0001), as well as a nonsignificant increase in the lymphocyte infiltration signature score in SMARCA4-low tumors (Fig. 1A). Pro-inflammatory immune cell types, such as T helper 1 (TH1) cells (P < 0.0001; q < 0.0001) and TH17 cells, (P = 0.0009; q = 0.0019), were significantly more prominent in the SMARCA4-low– compared to those in the SMARCA4-high–expressing cancers (Fig. 1B). By contrast, immune cell populations associated with inhibition of antitumor immunity such as TH2 cells (P < 0.0001; q < 0.0001) and regulatory T cells (P < 0.0001; q < 0.0001) were decreased in the SMARCA4-low– compared to those in SMARCA4-high–expressing cancers (Fig. 1C). Pan-cancer association of SMARCA4 expression with immune expression signatures. Transcriptomic analysis of The Cancer Genome Atlas (TCGA) pan-cancer RNA-sequencing data. (A) Immune signatures [from Thorsson et al. (17) dataset] by SMARCA4 expression for all tumor types combined. (B) Signature scores for pro-inflammatory immune cell populations by SMARCA4 expression for all tumor types combined. (C) Signature scores and cell populations for immunomodulatory cell populations by SMARCA4 expression for all tumor types combined. (D) Heatmap of immune signatures in SMARCA4-low and SMARCA4-high tumors for each tumor type. Gene expression fold change of SMARCA4-low versus SMARCA4-high tumors is color coded according to the legend. (E) Immune signatures and cell populations by SMARCA4 expression derived from RNA-sequencing data for high-grade serous OCs obtained from TCGA. Statistical analysis was performed using Mann-Whitney U test, Fisher exact test, and Benjamini-Hochberg adjustment. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. ACC, adrenocortical carcinoma; BLCA, bladder urothelial cancer; BRCA, breast invasive carcinoma; CESC, cervical squamous cell carcinoma and endocervical adenocarcinoma; CHOL, cholangiocarcinoma; COAD, colon adenocarcinoma; GBM, glioblastoma multiforme; HNSC, head and neck squamous cell carcinoma; KIRC, kidney renal clear cell carcinoma; KIRP, kidney renal papillary cell carcinoma; LGG, brain low-grade glioma; LIHC, liver hepatocellular carcinoma; LUAD, lung adenocarcinoma; LUSC, lung squamous cell carcinoma; OC/OV, ovarian serous cystadenocarcinoma; PRAD, prostate adenocarcinoma; SKCM, skin cutaneous melanoma; THYM, thymoma; UCEC, uterine corpus endometrial carcinoma. Similar analyses by cancer type demonstrated associations of immune signatures and immune cell infiltration with SMARCA4 expression across several tumor types (Fig. 1D). In OC (n = 171), a higher lymphocyte infiltration signature score (P = 0.0497; q = 0.27) was also observed in the SMARCA4-low (n = 43) cohort, with prominence of TH1 cells (P < 0.0001; q = 0.0013) and M1-type macrophages (P = 0.0098; q = 0.14) (Fig. 1, D and E). To determine the cancer cell–intrinsic impact of SMARCA4 expression on innate immune signaling, we explored the Cancer Cell Line Encyclopedia gene expression dataset. Our results demonstrate that overall IFN-related genes negatively correlate with SMARCA4 expression (fig. S1). We also investigated the prevalence of various type of SMARCA4 mutations in our predefined SMARCA4-low and SMARCA4-high groups across cancer types. Truncating and splice-site mutations were significantly more frequent in the SMARCA4-low– than those in SMARCA4-high–expressing cancers (P < 0.0001; table S1). In contrast, cancers with missense SMARCA4 somatic mutations were similarly distributed between SMARCA4-low– than SMARCA4-high–expressing cancers (P = 0.0527; table S1). Overall, these data support the rationale for studying the impact of loss-of-function mutations in SMARCA4 on the immune response. Loss of function of SMARCA4 in OC cells leads to increased IFN response and antigen presentation gene activation We generated SMARCA4-knockout [single-guide SMARCA4 (sgSMARCA4)] murine (ID8 and UPK10) and human (OAW28) stable OC cell lines as well as stable non-targeted control for each (sgNTC) (fig. S2; see Materials and Methods). RNA sequencing of the generated cell lines was used to examine differentially expressed coding genes in sgSMARCA4 and sgNTC OC cells (fig. S3A). Hallmark and Gene Ontology enrichment analyses identified significant up-regulation of IFN-α, IFN-β, and IFN-γ signaling pathways as well as genes related to nuclear factor κB (NF-κB) signaling and antigen processing/presentation in sgSMARCA4 as compared to that in sgNTC cells (Fig. 2A and fig. S3B). When specifically focusing on type I IFN pathway–related genes, we observed significant increase in basal levels of ISGs including OASL2, ISG15, IRF7, IFI44L, IFI27, MX1, and IFI44 in sgSMARCA4 cells (Fig. 2B and fig. S3C). By contrast, type III IFN genes were not up-regulated in sgSMARCA4 cells (fig. S3D). These results were validated by quantitative reverse transcription polymerase chain reaction (qRT-PCR) (Fig. 2C and fig. S3E). Loss of function of SMARCA4 in OC cells leads to increased interferon response and antigen presentation gene activation. (A) Up-regulated Hallmark and Gene Ontology (GO) pathways in ID8 cell single-guide SMARCA4 (sgSMARCA4) compared to those in sgNTC. Ribodeplete RNA sequencing was performed. Statistical analysis was based on hypergeometric test and performed using ClusterProfiler. IL-6, interleukin-6; JAK, Janus kinase; STAT3, signal transducer and activator of transcription 3; TNFA, tumor necrosis factor–α; FDR, false discovery rate. (B) Gene expression heatmap of type I IFN pathway–related genes in ID8 cells. Reads per kilobase of transcript per million mapped reads values were scaled to z-score for visualization. Gene expression fold change of sgSMARCA4 versus sgNTC cells is color coded according to the legend. (C) Quantitative reverse transcription polymerase chain reaction (qRT-PCR) validation results for IFN genes in ID8 cells (sgNTC and four sgSMARCA4 clones). Expression levels were normalized to β-actin expression, and comparisons of mRNA expression levels were performed relative to control (sgNTC). n = 3 independent experiments. (D) MHC1 expression in ID8 cells with or without IFN-ɣ by flow cytometry. (E) PD-L1 expression in ID8 cells with or without IFN-ɣ by flow cytometry. MFI, median fluorescence intensity. Statistical analysis was performed using two-tailored unpaired t test [(C) to (E)]. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001; n.s., not significant. Error bars represent ± SEM. Samples in duplicates [(A) and (B)] and triplicates [(D) and (E)]. KO, knockout; KO_1, target exon 14 clone 1; KO_2, target exon 14 clone 2; KO_3, target exon 23 clone 3; KO_4, target exon 23 clone 4; NTC, non-target control; NTC_1, NTC clone 1. We next examined the effects of SMARCA4 loss on the downstream immune signaling, including antigen presentation and immune checkpoint expression. MHC1 expression assessed by flow cytometry was significantly increased in sgSMARCA4 compared to that in sgNTC OC cells, both at baseline and in the setting of IFN-ɣ stimulation (Fig. 2D and fig. S3F). Similarly, at baseline, sgSMARCA4 cells expressed higher levels of PD-L1 than sgNTC cells and exhibited further up-regulation of PD-L1 in the presence of IFN-ɣ stimulation (Fig. 2E). Overall, these results highlight that loss of SMARCA4 leads to alteration of tumor cell–intrinsic immune signaling, potentially affecting both innate (type I IFN) and downstream adaptive (antigen presentation) immune responses. SMARCA4 loss-of-function mutation results in perturbation of chromatin accessibility at key immune genes Given that SMARCA4 is a core member of the SWItch/Sucrose Non-Fermentable (SWI/SNF) chromatin remodeling complex, we hypothesized that SMARCA4 loss alters chromatin accessibility for IFN-related gene transcription. We used the assay for transposase-accessible chromatin sequencing (ATAC-seq) to assess and compare chromatin accessibility changes in sgSMARCA4 or sgNTC OC cells. Loss of SMARCA4 led to decrease in accessibility in some sites and increase in others (Fig. 3A). Immune- and chemokine-related pathways were found to be enriched in genes with increased accessibility (Fig. 3B). Furthermore, we found significant increase of chromatin accessibility at TH1-type chemokine loci in sgSMARCA4 cells, such as CXCL10 (q = 0.02) and CCL2 (q < 0.0001) genes, with strong peaks on their transcription start sites (TSSs) (Fig. 3C). This was consistent with up-regulation of these genes seen in RNA-sequencing data. We further performed motif enrichment analysis and identified that the activator protein 1 (AP1), NF-κB, and IFN regulatory factor (IRF) motifs, which are bound by transcription factors that regulate innate and ISGs, are enriched in the open chromosomal regions affected by SMARCA4 deficiency (Fig. 3D). Overall, these results suggest that loss of SMARCA4 leads to changes in chromatin accessibility associated with induction of IFN-response genes. SMARCA4 loss-of-function mutation results in perturbation of chromatin accessibility at key immune genes. (A) Changes in genomic site accessibility in sgSMARCA4 (KO1) versus sgNTC cells. Log2 fold change and FDR-adjusted P value (Wald test P values from DESeq2 with Benjamini-Hochberg correction). (B) Gene set enrichment analysis (GSEA) of immune pathways with increased accessibility in sgSMARCA4 versus sgNTC cells. Kolmogorov-Smirnov statistic with Benjamini-Hochberg correction. Exact q values indicated in each panel. (C) Changes in chromatin accessibility at transcription start sites (TSSs) of ISGs (CXCL10 and CCL2) in sgSMARCA4 versus sgNTC cells. (D) Motifs enriched in open chromosomal regions affected by SMARCA4 deficiency. P value and binomial test were performed using Homer2. Experiments performed in duplicates. GO, Gene Ontology; KO, knockout; KO_1, target exon 14 clone 1; NTC, non-target control; TF, transcription factor. Increase in ISGs is not dependent on signaling through the type I IFN receptor in sgSMARCA4 OC cells We initially hypothesized that up-regulation of ISGs was mediated through enhanced signaling through type I IFN receptor (IFNAR). While IFNAR blockade with anti-IFNAR antibody resulted in attenuation of ISG expression in response to type I IFN stimulation (fig. S4A), unexpectedly, blocking the IFNAR did not attenuate the ISG expression in ID8 sgSMARCA4 cells (Fig. 4A). In the human sgSMARCA4 OAW28 cell line, blocking IFNAR resulted in partial attenuation of ISG expression (fig. S4B). These findings suggest that, in sgSMARCA4 cells, transcription activation of ISGs is not fully dependent on activation of the IFNAR by type I IFN, although the discrepancy between the murine and human cells indicates that this dependency may be cell line or species specific. Increase in ISGs is not dependent on signaling through the type I IFN receptor in sgSMARCA4 OC cells. (A) IFNAR-1 neutralizing assay in ID8 single-guide NTC (sgNTC) compared to that in sgSMARCA4 cells (four sgSMARCA4 clones). Cells were treated with Mock or IFNAR-1 antibody for 48 hours at 10 μg/ml. (B) IRF3 expression levels in sgSMARCA4 versus sgNTC cells by qRT-PCR. (C) IRF3 expression in doxycycline-inducible short hairpin NTC (shNTC) and shIRF3-transfected sgSMARCA4 cells by qRT-PCR. (D) qRT-PCR quantification of ISGs in ID8 sgNTC and sgSMARCA4 cells. The latter were transfected with shIRF3 or shNTC (with and without POLYI:C). (E) Enriched IRF motif at TSSs of ISG locus from ATAC-seq analysis of ID8 sgNTC versus sgSMARCA4 cells. Experiments performed in duplicates. (F) Analysis of publicly available ChIP-Atlas data of IRF3 DNA binding sites on ISGs in murine immune cell lines. Statistical analysis was performed using two-tailored unpaired t test [(A) to (D)]. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. Error bars represent ± SEM. n = 3 independent experiments in (A) to (D). Expression levels were normalized to β-actin expression, and comparisons of mRNA expression levels were performed relative to control [sgNTC_1 in (A), (B), and (D) and sgSMARCA4 transfected with shNTC in (C), as indicated]. KO, knockout; KO_1, target exon 14 clone 1; KO_2, target exon 14 clone 2; KO_3, target exon 23 clone 3; KO_4, target exon 23 clone 4; LPS, lipopolysaccharide stimulated, NTC, non-target control; NTC_1 = NTC clone 1; POLYI:C, polyinosinic-polycytidylic acid stimulated. We proceeded to evaluate whether activation of ISGs is dependent on the IRF3, a key upstream regulator of type I IFN gene transcription. Loss of SMARCA4 was associated with increase in IRF3 RNA and protein expression (Fig. 4B and fig. S5A) and nuclear translocation (fig. S5B). Knocking down IRF3 (Fig. 4C and fig. S5, C and D) abrogated ISG up-regulation in sgSMARCA4 cells (Fig. 4D and fig. S5E), confirming that ISG activation is at least partially dependent on IRF3. We examined the IRF3 binding sites at the ISG promoters as a potential mechanism that could explain the IFNAR-independent ISG up-regulation that we observed with SMARCA4 loss. Using the ATAC-seq data, we found that IRF3 motifs are enriched at the TSS of ISG locus (OASL2), suggesting that IRF3 may bind at a response element locus at the promoter site of ISGs (Fig. 4E). Analysis of publicly available datasets from chromatin immunoprecipitation (ChIP)–Atlas revealed IRF3 binding sites at the promoters of key ISGs including MX1 and OASL2 in human cancer cell lines (Fig. 4F) as well as key ISGs such as OASL, MX1, and MX2 in murine immune cell lines with or without immunostimulants (fig. S5F). Together, these results provide evidence to suggest that, in the setting of SMARCA4 deficiency, up-regulation of ISGs is at least, in part, driven by IRF3, potentially through increased IRF3 expression or upstream activation. Increase in ISGs is mediated through the RNA-sensing pathway Next, we focused on the potential upstream signaling mechanisms, leading to IRF3 activation. We hypothesized that chromatin remodeling associated with loss of SMARCA4 could lead to reexpression of noncoding and transposable elements (TEs) and tandem repeats that carry a potential to engage innate receptors that have been shown to lead to type I IFN pathway activation in other settings (7, 11, 19). To address this, we assessed the differentially expressed TEs in the transcriptomes of sgSMARCA4 versus sgNTC cells. Loss of SMARCA4 in the murine ID8 cells induced a significant increase in long terminal repeat (LTR) type of TEs that include endogenous retroviruses (ERVs) (Fig. 5, A and B), which was confirmed in the UPK10 model using qRT-PCR of a panel of ERV genes (fig. S6A). The ERV increase was not observed in the human OAW28 sgSMARCA4 cells (fig. S6, B and C). In contrast, long and short interspersed nuclear element TEs were generally down-regulated in the sgSMARCA4 cells compared to those in sgNTC cells (Fig. 5A and fig. S6B). Tandem satellite repeats were up-regulated in OAW28 sgSMARCA4 cells compared to those in sgNTC cells (fig. S6D) but remained unchanged in ID8 sgSMARCA4 cells (Fig. 5A). Increase in ISGs is associated with production of TEs and up-regulation of dsRNA. (A) Volcano plots for differential expression of TEs in ID8 cells. (B) Heatmap of long terminal repeats (LTR) with adjusted P < 0.05. (C) Representative pictures (top) of double-stranded RNA (dsRNA) identification by immunofluorescence (IF) in sgSMARCA4 and sgNTC ID8 cells with quantification (bottom). RNAse, ribonuclease. (D) qRT-PCR quantification of ISGs in ID8 sgNTC cells and sgSMARCA4 cells transfected with shMAVS or shNTC. (E) MAVS expression in doxycycline-inducible shNTC and shMAVS-transfected sgSMARCA4 cells by qRT-PCR. Expression levels were normalized to β-actin expression, and comparisons of mRNA expression levels were performed relative to control [sgSMARCA4 transfected with shNTC in (C) and sgNTC_1 in (D), as indicated]. Statistical analysis was performed using two-tailored unpaired t test [(C) to (E)]. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. Error bars represent ± SEM. Samples in duplicates [for (A) and (B)]. n = 3 independent experiments [in (C) and (D)]. Red line in (A) represents a cutoff of an adjusted P value of <0.05. KO, knockout; KO_1, target exon 14 clone 1; KO_2, target exon 14 clone 2; KO_4, target exon 23 clone 4; NTC, non-target control; NTC_1, NTC clone 1. We observed the presence of cytosolic double-stranded RNA (dsRNA) in sgSMARCA4 compared to that in short hairpin NTC (shNTC) OC cells using immunofluorescence (IF) (Fig. 5C and fig. S6E), suggesting that the RNA recognition pathway could be driving the increased innate immune signaling. We thus proceeded to determine whether the RNA-sensing retinoic acid-inducible gene I (RIG-I)–like receptor (RLR) pathway could be potentially driving ISG up-regulation, possibly through detection of increased LTR expression in sgSMARCA4 cells. We performed short hairpin RNA (shRNA) knockdown of the mitochondrial antiviral signaling (MAVS) protein, a key downstream adapter mediating RIG-I and MDA5 signaling (fig. S6, F and G). Knockdown of MAVS resulted in down-regulation of expression of different ISGs in sgSMARCA4 ID8 cells (Fig. 5D). There is increasing recognition that the cytosolic DNA-sensing Stimulator of interferon genes (STING) signaling pathway also plays a central role in recognition of RNA viruses due to cross-talk with MAVS signaling (20). To determine the role of STING signaling in ISG regulation in the setting of SMARCA4 loss, we performed STING-knockout experiments in sgSMARCA4 ID8 cells. Silencing of STING resulted in down-regulation of expression of different ISGs in sgSMARCA4, including IRF3 (Fig. 5E). Together, these data suggest that ISG up-regulation induced by loss of function of SMARCA4 is dependent on MAVS and STING signaling, possibly through increased expression of repeat elements as has been previously demonstrated within the setting of hypomethylating agent therapy (7). Loss of SMARCA4 leads to increased tumor immunogenicity in vivo We proceeded to examine the impact of SMARCA4 loss on the immune response in vivo in the peritoneal ID8 OC model (Fig. 6A and figs. S7 and S8). While there was no significant change in the numbers of CD4+ and CD8+ T cells in the ascites of sgSMARCA4 OCs (fig. S9A), we observed evidence of enhanced CD4+ and CD8+ T cell activation, as demonstrated by increased expression of programmed cell death protein 1 (PD1) (P < 0.0001; Fig. 6B) and increased but not statistically significant Granzyme B expression (fig. S9B). There was no substantial change in regulatory T cells between the sgSMARCA4 and sgNTC ascites (fig. S9C). We also observed increased activation of natural killer (NK) cells as evidenced by Granzyme B expression (P = 0.03; Fig. 6C) although there was a relative decrease in the overall NK cell numbers in this model (fig. S9D). sgSMARCA4 leads to increased ovarian tumor immunogenicity in vivo. (A) Workflow for the OC tumor model. C57BL/6 (Cg)–Tyrc-2J/J mice were inoculated intraperitoneally with 10 million ID8 sgNTC or sgSMARCA4 (KO_4) tumor cells, and spectral flow cytometry was performed 21 days later on harvested ascites samples. ip, intraperitoneal. (B) Frequency of tumor PD1+CD4+ T cells and PD1+CD8+ T cells. (C) Frequency of NK1.1+ cells expressing Granzyme B+. (D) Frequency of tumor dendritic cells (of CD45+ cells) and MHCII-expressing dendritic cells. (E) Frequency of tumor macrophages (of CD45+ cells) and PD-L1–expressing macrophages. (F) In vivo bioluminescence imaging of tumor burden in sgSMARCA4 versus sgNTC tumors (unpaired t test of area under the curve). Statistical analysis was performed using two-tailored unpaired t test [(B) to (F)]. *P < 0.05, **P < 0.01 and ****P < 0.0001. Error bars represent ± SEM. n = 10 mice per group. KO, knockout; KO_4, target exon 23 clone 4; NTC, non-target control; NTC_1, NTC clone 1. To validate these findings in another tumor model, we generated B16-F10 melanoma cell lines where SMARCA4 was depleted either using single-guide RNA (sgRNA) or doxycycline-inducible knockdown (Fig. 7A and fig. S10A). Depletion of SMARCA4 with either strategy led to increased tumor immunogenicity in vivo, as observed by increased infiltration with CD3+, CD4+, and CD8+ T cells and NK cells as well as evidence of activation in these cell compartments (Fig. 7, B to D, and fig. S10, B to D). Loss of SMARCA4 up-regulates innate and adaptive immune response in melanoma. (A) Workflow for the B16-F10 tumor model. C57BL/6 (Cg)–Tyrc-2J/J mice were inoculated subcutaneously with 150,000 shNTC, sgEV, shSMARCA4_1, shSMARCA4_2, or sgSMARCA4 tumor cells, and spectral flow cytometry was performed 21 days later on harvested tumor samples. (B) Frequency of tumor CD8+ T cells and MFI of ICOS+-expressing CD8+ T cells in B16-F10 model. (C) Frequency of NK1.1+ cells in B16-F10 knockout and knockdown models (of CD45+ cells). (D) MFI of MHCII in macrophages in B16-F10 SMARCA4 knockout and knockdown models. (E) Tumor volume of shSMARCA4 versus shNTC B16 tumors (unpaired t test of final time points). n = 5 mice per group, three biological replicates. (F) Frequency of NK1.1+ cells in shSMARCA4 versus shNTC tumors (of live cells) in RAG2−/− mice. (G) MFI of MHCII-expressing macrophages in B16-F10 shSMARCA4 versus shNTC tumors in RAG2−/− mice. (H) Tumor volumes of shSMARCA4 versus shNTC B16 tumors in RAG2−/− mice (unpaired t test of final time points). n = 3 to 5 mice per group, three biological replicates. Statistical analysis was performed using two-tailored unpaired t test. *P < 0.05, **P < 0.01, ***P < 0.001. Error bars represent ± SEM. Dox, doxycycline; EV, empty vector; NTC, non-target control; SC, subcutaneous. In the myeloid cell compartment in the ID8 model, we observed significantly increased numbers of dendritic cells and MHCII up-regulation (P = 0.01; Fig. 6D). Furthermore, there was a significant decrease in the overall number of macrophages (P < 0.0001), although the remaining macrophages exhibited increased expression of PD-L1 (Fig. 6E). Similarly, in the B16 model, SMARCA4 depletion led to an increase in dendritic cells and myeloid cells with concomitant up-regulation of MHCII in these compartments consistent with activated phenotype (Fig. 7D and fig. S10, E and F). Last, while depletion of sgSMARCA4 resulted in comparable growth of cells in vitro in both tumor models (fig. S9E), there was a substantial delay in tumor growth in vivo in both ID8 (P = 0.002; Fig. 6F) and B16 models (Fig. 7E and fig. S10G). To determine whether the increased immunogenicity was dependent on the adaptive immune response, we evaluated the impact of SMARCA4 depletion on B16 TME and tumor growth in RAG2−/− mice. There was a notable increase in tumor-infiltrating NK cells, dendritic cells, and myeloid cells (Fig. 7F and fig. S11, A to C). The myeloid cells exhibited evidence of increased activation, as noted by increased CD86 and MHCII expression in different cell subsets (Fig. 7G and fig. S11, D and E). Despite the lack of adaptive immune response, tumor growth was also significantly inhibited by loss of SMARCA4 in the RAG2−/− mice with either SMARCA4 depletion strategy (Fig. 7H and fig. S11F). Overall, these findings highlight that cancer cell–intrinsic loss of SMARCA4 is associated with increased immunogenicity in vivo, as evidenced by a more active innate and adaptive immune TME and tumor growth delay. Therapeutic targeting of BRG1 recapitulates the immunogenic effects of sgSMARCA4 Our findings suggest that genetic loss of SMARCA4 is associated with increased tumor immunogenicity, which led us to hypothesize that targeting SMARCA4 therapeutically may also promote tumor immunogenicity. To mimic the loss of SMARCA4, a dual BRG1/BRM inhibitor (HY-119374) and BRG1/BRM proteolysis–targeting chimera (PRTD-0043) were tested in SMARCA4–wild-type murine OC cell lines (ID8 and UPK10). Treatment of cells with either compound led to up-regulation of key type I IFN–related genes (Fig. 8A and fig. S12, A and B). We proceeded to assess the impact of treatment with the HY-119374 compound on ID8 tumors in vivo (Fig. 8B). Akin to genetic SMARCA4 depletion, treatment with the inhibitor resulted in increased CD3+ cells with significant increase in CD8+ T cells (P = 0.02; Fig. 8C) and a similar trend for CD4+ T cells (fig. S12C). Both CD4+ and CD8+ T cells exhibited higher levels of activation, as evidenced by up-regulation of Inducible T-cell COStimulator (ICOS), PD1, and Granzyme B expression (Fig. 8, D and E). We similarly observed an increase in NK cell cytolytic activity (Fig. 8F) although the relative NK cell numbers remained unchanged (fig. S12D). In the myeloid compartment, we found a significant increase in macrophage MHCII expression (P = 0.0077; fig. S12E and Fig. 8G) and increased infiltration with dendritic cells and up-regulation of MHCII expression on their surface (Fig. 8H). Myeloid-derived suppressor cells (defined by F4/80intermediate/negativeLy6C−CD206+ phenotype) were overall decreased (fig. S12F), reflecting a more pro-inflammatory TME. Last, treatment with HY-11934 significantly decreased tumor burden (P = 0.015; Fig. 8I) and inhibited ascites accumulation (Fig. 8J and fig. S12G). Overall, these results demonstrate that therapeutic inhibition of BRG1 leads to immune activation in the TME, supporting the rationale for this therapeutic strategy as an approach to increase tumor immunogenicity. Therapeutic targeting of BRG1 recapitulates the immunogenic effects of sgSMARCA4. (A) qRT-PCR quantification of IFN gene expression using ID8 parental cell treated with 1 μM of BRG1/BRM inhibitor or vehicle [dimethyl sulfoxide (DMSO)]. Expression levels were normalized to β-actin expression, and comparisons of mRNA expression levels were performed relative to control (vehicle). (B) Workflow for the OC tumor model and treatment. C57BL/6 mice were inoculated intraperitoneally with 2 million ID8 parental cells and treated with oral BRG1/BRM inhibitor (20 mg/kg) or vehicle DMSO and Solutol HS in 10 mM citrate buffer (pH 6 in a ratio of 5:20:75, respectively) starting on day 5 for a total of eight doses (4 days on, 3 days off × 2 weeks). Flow cytometry was performed 3 weeks after inoculation on harvested ascites samples. (C) Frequency of tumor CD3+ cells and CD8+ T cells (of CD45+ cells). (D) Frequency of tumor ICOS+CD4+ T cells and PD1+ CD4+ T cells. (E) Frequency of tumor ICOS+CD8+ T cells, PD1+CD8+ T cells and Granzyme B–expressing CD8+ T cells. (F) MFI of activated (CD11b+) NK1.1+ cells. (G) MFI of MHCII-expressing tumor macrophages. (H) Frequency of tumor dendritic cells (of CD45+ cells) and MHCII-expressing dendritic cells. (I) In vivo bioluminescence imaging of tumor burden in vehicle versus BRG1/BRMi-treated tumors (unpaired t test of area under the curve). (J) Mouse weights in ID8 tumors treated with BRG1/BRM inhibitor or vehicle. Weights normalized to baseline weight (day 5). Statistical analysis was performed using two-tailored unpaired t test. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. Error bars represent ± SEM. n = 10 mice per group in (C) to (H) and (J) and n = pooled 20 mice per group from two independent experiments in (I). BRG1/BRMi, BRG1/BRM inhibitor. DISCUSSION Traditionally, tumor immunogenicity and response to ICB have been correlated with biomarkers such as TMB and PD-L1 expression (4), neither of which carries a strong predictive value in OC (21). Several studies have suggested that disruption in chromatin remodeling may serve as a predictor of immunogenic properties of cancers (5, 12, 22). Proficient chromatin remodeling has been associated with T cell exhaustion and poor tumor control (23), and, conversely, depletion of chromatin regulators correlates with increased ICB sensitization (9). Previous preclinical murine data also suggest that SWI/SNF-deficient tumors, including loss of function of SMARCB1, PBRM1, and ARID1A, respond better to ICB (11, 12, 24). Clinical results further support these preclinical findings including in SMARCA4-deficient lung cancers (13), PBRM1-mutated clear cell renal cell carcinoma (25), and clear cell OCs (26), of which more than 50% have SWI/SNF mutations (15). Overall, the impact of epigenetic dysregulation in OC on the immune TME remains relatively uncharacterized. Our study results provide a mechanistic understanding of observations made from previous published work, suggesting improved ICB response (5, 13) and immunogenic OC TME in SMARCA4-deficient cancers (5). We demonstrate that loss of SMARCA4 leads to activation of innate immune signaling, leading to IRF3-dependent induction of downstream ISGs, chemokines, and antigen presentation machinery. Our findings also highlight that up-regulation of ISGs across cancer models is not uniform and could be driven by both type I IFN–dependent and independent mechanisms. These discrepancies could be attributed to multiple factors including differential impact on SMARCA4 loss on chromatin remodeling, dependencies on different signaling pathways, and potential differences in expression of innate immune signaling components such as type I IFN receptor. We find that this immune activation may be, in part, triggered by activation of the RNA-sensing pathway. We also show that SMARCA4 loss leads to reexpression of normally silenced TEs such as LTRs and production of dsRNA. Although our study does not establish a direct link between TE expression and RLR activation, several studies in the past have demonstrated that derepression of repeat elements within the context of hypomethylating agent therapy can drive activation of RLR-dependent activation of type I IFN pathway (7, 27). In addition, although traditionally categorized as noncoding proteins, recent studies have suggested that induced TEs may be translated into proteins that can serve as neoantigens (9). Recent studies have also linked the integrity of the SWI/SNF complex to proficient DNA damage repair (28, 29). Increased R-loops, cytosolic single-stranded DNA, and RNA:DNA hybrids caused by dysfunction of the chromatin remodeling complex subunits, including BRG1 or ARID1A, may lead to accumulation of DNA breaks and activation of the type I IFN pathway through DNA sensing (28, 29). Given the dependency on STING identified in our study, it is thus possible that both DNA- and RNA-sensing mechanisms play a role in activation of innate immune signaling in the setting of SMARCA4 loss. Consistent with the increase in tumor cell–intrinsic immunogenicity in response to SMARCA4 loss, we observe increased immune activation in vivo, as evidenced by delayed tumor growth, increase in activated T and NK cells, and increased infiltration of activated antigen-presenting cells. The impact on the innate immune cell compartment and restricted tumor growth was also present in RAG2−/− mice, particularly affecting NK cell infiltration and activation. This leads us to conclude that cancer cell–intrinsic SMARCA4 depletion likely directly affects the innate immune cells in the TME. It is possible that the impact on the adaptive immune cells lies downstream of innate immune response or may also be indirectly activated through production of new antigens, possibly through translation of noncoding elements. Last, we demonstrate the potential therapeutic application of these findings, by showing that a BRG1/BRM inhibitor mimicked the changes seen in sgSMARCA4 tumors, resulting in increased infiltration with activated T and NK cells and antigen-presenting cells. Other epigenetic therapies have recently been shown to enhance tumor immunogenicity (7, 30, 31). DNA methyltransferase inhibitors with or without histone deacetylase inhibitors have been shown to induce type I IFN genes, leading to activation of innate and adaptive immune responses in the TME (7, 30). CARM1 (a protein that methylates and displaces BAF155, a subunit of the SWI/SNF complex) was found to be a negative regulator of tumor-infiltrating T cells, and, when ablated genetically or therapeutically, this led to increased tumor CD8+ T cell infiltration and activation as well as immune-related tumor suppression (31). Further analyses showed that these effects were mediated by type I IFN induction and a heightened response to IFN-α/γ (31). Overall, our findings demonstrate an immunomodulatory role of SMARCA4 and generate rationale for further evaluation of SMARCA4 loss-of-function mutations as predictors of response to immunotherapy and for use of SMARCA4/BRG1 as a therapeutic target to increase OC immunogenicity. MATERIALS AND METHODS Cell lines The ID8-Luciferase (ID8) cell line was provided by R. Brentjens’ lab. The UPK10–green fluorescent protein (GFP; UPK10) cell line was a gift from J. Conejo-Garcia’s lab. OAW28 cells were purchased commercially from the European Collection of Authenticated Cell Cultures (85101601-1VL, Sigma-Aldrich). B16-F10 cells were originally obtained from the American Type Culture Collection (ATCC; no. CRL-6475). All cell lines (table S2) underwent routine authentication using short tandem repeat (STR) profiling and were tested for mycoplasma using the PCR-based Universal Mycoplasma Detection kit (ATCC, no. 30-1012K). ID8 and UPK10 were cultured in RPMI medium supplemented with 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin. B16-F10 cells were maintained in RPMI medium supplemented with 7.5% FBS and 1% penicillin/streptomycin. OAW28 cell line was maintained in Dulbecco’s Modified Eagle media (DME), 2 mM glutamine, 1 mM sodium pyruvate, bovine insulin (20 IU/liter) with 10% FBS and 1% penicillin/streptomycin. Cells were maintained in an incubator set at 5% CO2 atmosphere at 37°C. Experiments were carried out with cells between passages 2 to 10 (at 80% confluence). Isogenic SMARCA4-knockout murine (ID8 and UPK10) OC cell lines were generated using CRISPR-Cas9 lentiviral transduction and puromycin selection (fig. S2A). Isogenic murine cell lines further underwent clonal selection using single-cell sorting (fig. S2B). A panel of five human OC cell lines was tested for SMARCA4 baseline expression to identify the cell line with highest levels (fig. S2C). Isogenic sgSMARCA4 human OAW28 cell lines were generated using CRISPR-Cas9 lentiviral transduction and puromycin selection (fig. S2D). SMARCA4-knockout cell line establishment Stable sgSMARCA4 cell lines were created with ID8, UPK10, and OAW28 models. Individual guide RNAs (gRNAs; table S3) against murine and human SMARCA4 were cloned into pLentivectorCRISPR v2 (52961, Addgene). Non-target control (NTC) was used as control. Lentiviral particles were produced at Memorial Sloan Kettering Cancer Center’s (MSK) Gene Editing and Screening Core Facility. Lentiviral titers [Transducing Units (TU) per milliliter] were first determined (see the “Viral titration” section). Each cell line was infected for 24 hours in the presence of polybrene (10 μg/ml; TR-1003-G, MilliporeSigma) and then selected for 14 days in puromycin (2 μg/ml for ID8 and 8 μg/ml for UPK10; Thermo Fisher Scientific). Transduction efficiency was confirmed by qRT-PCR and IF. To generate single-cell clones for ID8 and UPK10 lines, single-cell suspensions stained with 4′,6-diamidino-2-phenylindole (DAPI) were cell sorted (BD FACSAria II Cell Sorter, BD Biosciences) to distribute cells into the wells of a 96-well plate with 100 μl of culture medium. Each single-cell clone was expanded, then tested for SMARCA4 mutation using Sanger sequencing, and validated using IF. For generation of stable sgSMARCA4 B16-F10 cell line, individual gRNAs (table S3) against murine SMARCA4 encoded in the px459 vector were generated by the MSK Gene Editing and Screening Core Facility. Cas9-expression empty vector was used as a control. B16-F10 cells were transfected with all three gRNAs by electroporation using the Eppendorf Multiporator. Cells were allowed to recover for 48 hours and then selected using puromycin (3 μg/ml) for 14 days. Single-cell clones were generated by seeding a 96-well plate with a limiting dilution of the single-cell suspension. For ID8, UPK10, and B16-F10, generated single-cell clones were expanded, then tested for SMARCA4 mutation using Sanger sequencing, and validated using Western blot or IF. Sanger sequencing Total DNA was extracted from isogenic cell lines using the DNeasy Blood & Tissue Kit (69504, QIAGEN), according to the manufacturer’s instructions. DNA was amplified using AmpliTaq Gold 360 Master Mix (4398886, Thermo Fisher Scientific) and primers (table S4). PCR products were purified using ExoSAP-IT PCR product Cleanup (78200, Thermo Fisher Scientific) and subjected to Sanger sequencing (Genewiz) using the same primers. Sequencing traces were reviewed using SnapGene (www.snapgene.com) to assess the sgRNA editing. Lentiviral transduction for generation of doxycycline-inducible knockdowns MAVS- and IRF3-knockdown cell lines were created in the murine model ID8. SMARCA4-knockdown cell lines were created in the murine model B16. shRNAs were designed using the SplashRNA algorithm (32). A previously described optimized lentiviral miR-E expression backbone system was used for inducible LT3RENIR (red, Neomycin) expression (33) for ID8 and inducible LT3GEPIR expression in B16. Individual LT3RENIR/LT3GEPIR lentiviral miR-E–based expression vectors containing shRNA against MAVS, IRF3, SMARCA4, and NTC (table S5) were cloned, and lentiviral particles were produced in MSK’s Gene Editing and Screening Core Facility. Lentiviral titers (TU per milliliter) were first determined (see the “Viral titration” section). Cells were infected for 24 hours and then selected for 14 days in geneticin (50 mg/ml; 10131035, Thermo Fisher Scientific) for ID8 and in puromycin (3 μg/ml; Thermo Fisher Scientific) for B16 cells. shRNAs were induced with 6 days and 18 hours of doxycycline treatment, respectively, before experiments. Knockdown efficiency was confirmed by qRT-PCR and Western blot. The short hairpin sequences used are described in table S5. STING-knockdown cell line establishment STING-knockout cell lines were created in the murine model ID8. Knockouts of STING were generated using gene-specific CRISPR-Cas9–knockout plasmids for STING (no. SC428364, Santa Cruz Biotechnology) and for NTC (SC-421688, Santa Cruz Biotechnology). ID8 cells were seeded into a six-well plate 1 day before transfection of 1 μg of plasmid DNA using Lipofectamine 2000 (11668027, Thermo Fisher Scientific) and Plasmid Transfection Medium (SC-10862, Santa Cruz Biotechnology). After 48 hours, cells were sorted based on GFP expression using the BD FACSAria Cell Sorter. Cell lines were expanded for 1 week before use. To confirm knockout, cells were seeded into 24-well plates and transfected with Salmon DNA (1 μg/ml; 15632011, Thermo Fisher Scientific) or low–molecular weight poly(I:C) (30 ng/ml; tlrl-picw, InvivoGen) using Lipofectamine 2000 and Plasmid Transfection Medium as per the manufacturer’s instructions. Treatments with recombinant IFN-β (100 U/ml; 300-02 BC, PeproTech) were used as a control. Cells were collected after 24 hours for flow cytometry. To examine cell surface expression of Bst2, cells were harvested and incubated with viability stain (L34966, Thermo Fisher Scientific) and Bst2 (127016, BioLegend) and then acquired on the NxT Attune flow cytometer (Thermo Fisher Scientific). Data were analyzed using FlowJo software (BD Biosciences). Viral titration Cells were seeded in 24-well plates (25,000 to 50,000 cells per well). On the following day, the cells were transduced for 6 hours with fivefold serial dilutions (protocol 25–78125, Horizon Discovery) of lentivirus stock. Cells were incubated for 48 hours. Transduction efficiency was confirmed by the number of puromycin- or geneticin-selected colonies. RNA-sequencing analyses Total RNA extracted from ID8 and OAW28 cells with or without SMARCA4 deficiency using the RNeasy Mini Kit (74104, QIAGEN) was subjected to ribodeplete RNA sequencing at MSK’s Integrated Genomics Operation (IGO). For coding gene mapping and analysis, RNA-sequencing reads were aligned to the reference mouse (version mm10) and human (version hg38) genomes using STAR aligner (34), and the number of reads counted was mapped to each gene using GenomicAlignments package in Bioconductor (35, 36). Expression fold changes between sgSMARCA4 and sgNTC cells were calculated from DESeq2 (37), which was performed using raw read counts. P values obtained from DESeq2 were corrected for multiple testing using the Benjamini-Hochberg method (38). ClusterProfiler (39) was used for gene set enrichment analysis (GSEA) according to fold change and differentially expressed genes. Read counts were normalized by DESeq2 for visualization. For noncoding gene mapping, counts filtering and normalization and differential expression analysis, RNA-sequencing reads were 3′ trimmed for base quality less than 20 using Skewer (40). Reads shorter than 40 after trimming were removed. The quality score of the remaining bases were sorted, and the quality at the 20th percentile was computed. Reads with quality at the 20th percentile less than 15 were discarded. Only paired reads that both passed the quality check were mapped to the reference genome (mm10 or hg38) using STAR v2.7 with default parameters (34). Gene counts were assigned based on Gencode annotation using featureCounts (subread-1.4) with the external Ensembl annotation. Repeat counts per element family were primarily quantified against RepeatMasker using featureCounts (subread-1.4) and then adding the counts of the unassigned reads that mapped to Repbase consensus sequence. Repeat counts of a given family are the sum of mapped reads to RepeatMasker and unmapped reads against Repbase. Expression of repeats and coding genes was normalized across samples using trimmed mean of M values in edgeR. The size factor for each sample was calculated using calcNormFactors based on coding genes alone as described in a previous study (41). Low-count genes/repeats in each sample with counts per million (CPMs) smaller than 2 were removed before calculating the size factor. Log2-transformed CPMs were used for downstream visualization. Differential expressions of coding genes and repetitive elements were analyzed separately using DESeq2 (v1.33.4) (37). ATAC-seq and epigenome analyses Freshly harvested ID8 sgSMARCA4 and sgNTC cells were directly sent to MSK’s Epigenetics Research Innovation Lab. ATAC was performed as previously described (42) using the Tagment DNA TDE1 Enzyme (Illumina, 20034198). Sequencing libraries were prepared using the ThruPLEX DNA-Seq Kit (Takara Bio, R400676) and sent to the MSK IGO core facility for sequencing on a NovaSeq 6000. Raw sequencing reads were trimmed and filtered for quality (Q > 15) and adapter content using version 0.4.5 of TrimGalore (www.bioinformatics.babraham.ac.uk/projects/trim_galore) and running version 1.15 of cutadapt and version 0.11.5 of FastQC. Version 2.3.4.1 of bowtie2 (http://bowtie-bio.sourceforge.net/bowtie2/index.shtml) was used to align reads to mouse assembly mm10, and alignments were deduplicated using MarkDuplicates in Picard Tools v2.16.0. Enriched regions were discovered using MACS2 (https://github.com/taoliu/MACS) with a P value setting of 0.001, filtered for blacklisted regions (http://mitra.stanford.edu/kundaje/akundaje/release/blacklists/mm10-mouse/mm10.blacklist.bed.gz), and a peak atlas was created using ±250 base pairs around peak summits. The BEDTools suite (http://bedtools.readthedocs.io) was used to create normalized bigwig files. Version 1.6.1 of featureCounts (http://subread.sourceforge.net) was used to build a raw counts matrix, and DESeq2 was used to calculate differential enrichment for all pairwise contrasts. Peak-gene associations were created by assigning all intragenic peaks to that gene, while intergenic peaks were assigned using linear genomic distance to TSS. Pathway enrichment was calculated by assigning each gene a unique score based on the associated peak with the greatest magnitude change and running GSEA in pre-ranked mode. Motif signatures were obtained using Homer v4.5 (http://homer.ucsd.edu) on differentially enriched peak regions. IRF3-enriched regions were downloaded from the Peak Browser of the mm10 version of the ChIP-Atlas database (http://chip-atlas.org) and displayed using Integrative Genomics Viewer (IGV) (43). Transcript abundance (qRT-PCR) Total RNA (500 to 1000 ng) extracted from isogenic cell lines was used to generate cDNA with the Superscript VILO Master Mix (11755050, Thermo Fisher Scientific) according to the manufacturer’s instructions. In murine and human cells, qRT-PCR (table S6) was performed using TaqMan Universal Master Mix II, with UNG (4440044) for IFN genes and PowerUp SYBR Green (A25776, Thermo Fisher Scientific) for endogenous retroviral genes using StepOnePlus Real-Time PCR System (Applied Biosystems, Thermo Fisher Scientific). β-Actin was used as the reference gene. The threshold cycles (ΔΔCT) method was used to calculate relative expression levels, as previously described (44). All qRT-PCR assays were carried out in triplicate and then repeated with new cDNA synthesis. Reverse transcription–negative cDNA synthesis reactions were performed for each gene tested. Treatments For BRG1/BRM-targeted treatment, ID8 and UPK10 parental cells were seeded in a six-well plate at 50,000 cells per well. After 24 hours, cells were treated for 48 hours with 1 μM of the dual BRG1/BRM inhibitor (no. HY-119374, MedChemExpress) or dual BRG1/BRM proteolysis–targeting chimera (PRTD-0043, Prelude Therapeutics) or vehicle (dimethyl sulfoxide). Cells were then collected after 48 hours for RNA extraction and qRT-PCR of ISGs. To validate the efficiency of IFNAR blockade, ID8 and OAW28 NTC cells were seeded in a six-well plate at 50,000 to 200,000 cells per well. After 24 hours, cells were treated for 24 hours with or without recombinant murine IFN-α (12100-1, PBL Assay Science) or human IFN-α (11100-1, PBL Assay Science) at 2000 U/ml. After 24 hours, cells were treated for 48 hours with mock (medium) or antihuman IFNAR2 (21385-1, PBL Assay Science) or antimouse IFNAR-1 (BE0241, BioXCell) at 8 to 10 μg/ml. Cells were collected after 48 hours for RNA extraction and qRT-PCR of IFN genes. Protein analyses Standard Western blotting was conducted (45). Primary antibodies against MAVS (BS-4053R, Bioss; 1:500), IRF3 (MA5-32348, Invitrogen; 1:500), and BRG1 (49360, Cell Signaling Technology; 1:400) were used. α-Tubulin (2144S, Cell Signaling Technology; 1:1000) and vinculin (sc-73614, Santa Cruz Biotechnology; 1:5000) were used as reference protein. Conjugated secondary antirabbit antibodies were used and detected using the iBright Systems (FL1500, Invitrogen). For detection of BRG1 and vinculin, horseradish peroxidase–linked secondary antirabbit (7074P2, Cell Signaling Technology; 1:10,000) and antimouse (7076P2, Cell Signaling Technology; 1:10,000) antibodies were used respectively. For all blots, quantification and analysis were performed using iBright Analysis Software (V5.1.0). Experiments were performed in triplicate. For IF analysis, cells were grown on glass coverslips and fixed with 4% formaldehyde (28908, Thermo Fisher Scientific), blocked in 10% normal goat serum (S-1000, Vector Laboratories) and then incubated with primary antibody against BRG1 (3508, Cell Signaling Technology; 1:200) or IRF3 (MA5-32348, Invitrogen; 1:100). The samples were incubated overnight at 4°C. Conjugated secondary antibody Alexa Fluor 488 goat antirabbit (A-11034, Thermo Fisher Scientific; 1:500) was incubated for 2 hours. Slides were mounted using Gold ProLong antifade reagent containing DAPI (P36930, Thermo Fisher Scientific). Slides were scanned at MSK’s Molecular Cytology Core using a Pannoramic Scanner (3DHistech, Budapest, Hungary) with a 40×/0.95 numerical aperture air objective. No post-acquisition processing was performed besides minor adjustments of brightness and contrast, applied equally to all images, using the QuPath software (46) (version 0.3.2). Experiments for each condition were performed in duplicate. For IF protein analyses, ImageJ/Fiji software (version 1.53 t) was used to quantify the number of pixels per cell analyzing at least five representative images (×40 field) for each condition annotated on CaseViewer (version 2.4), as previously described (44). BRG1 and IRF3 images were analyzed using ImageJ/Fiji with a custom macro. The number of cells within each image was counted by segmenting the DAPI channels. After setting an appropriate threshold, nuclear intensity of BRG1 was measured for the image and then normalized to the total cell area. For subcellular IRF3 quantification, nuclear area of IRF3 was measured for the imaged normalized to the total cell area. For J2 IF analysis, ID8 cells were grown on coverslips (GG-12-1.5H-PDL, NeuVitro). High–molecular weight poly(I:C) (2 μg/ml) was transfected onto coverslips as a positive control using Lipofectamine 2000 and Plasmid Transfection Medium as per the manufacturer’s instructions. Four hours later, coverslips were washed with phosphate-buffered saline (PBS) before the addition of 2% paraformaldehyde for 20 min. Slides were washed with PBS before the addition of Alexa Fluor 555 Wheat Germ Agglutinin stain (W32464, Thermo Fisher Scientific; 1:200), which incubated for 10 min at 37°C. Cells were permeabilized and blocked in PBS containing 1% BSA (BP1605-100, Thermo Fisher Scientific) and 0.3% Triton X (X100, MilliporeSigma). For dsRNA depletion, slides were treated with 30 U of ribonuclease III (AM2290, Thermo Fisher Scientific) for 30 min at 37°C. Samples incubated overnight at 4°C with J2 anti-dsRNA antibody (RNT-SCI-10010200, Jena Bioscience; 1:1000). Conjugated secondary Alexa Fluor 488 antimouse immunoglobulin G2A antibody (407121, BioLegend, 1:1000) incubated for 2 hours. Coverslips were mounted onto slides using ProLong Gold Antifade Mountant and DAPI (P36941, Thermo Fisher Scientific). Slides cured for a minimum of 2 days at room temperature before storing at 4°C for long-term storage. Slides were imaged on the Zeiss LSM780 Confocal Microscope using 63× oil objective. dsRNA signal was quantified in ImageJ. The corrected total cell fluorescence was calculated as integrated density − (area of selected cell × mean fluorescence of background readings). In vivo studies Six- to 10-week-old female C57BL/6J (RRID:IMSR_JAX:000664, the Jackson Laboratory) and B6(Cg)-Tyrc-2J/J (RRID:IMSR_JAX:000058, the Jackson Laboratory) mice were used for in vivo studies and were cared for in accordance with the guidelines approved by the MSK Institutional Animal Care and Use Committee (IACUC) and Research Animal Resource Center (approved protocol no. 19-02-004). ID8 sgNTC and ID8 sgSMARCA4 tumors were generated by injecting 10 million cells with PBS intraperitoneally. ID8 tumors were generated by injecting 2 million cells with PBS IP. Tumors were measured weekly using IVIS Spectrum in vivo imaging, and tumor burden was calculated using the Live Imaging 4.0 software. Body weight and clinical signs were also assessed. At the end of the study, the ascites was collected for flow cytometry analysis. B16-F10 sgEV and sgSMARCA4 tumors were generated by injecting 150,000 cells in 100 μl of PBS intradermally in the flank. Doxycycline-inducible shNTC and shSMARCA4 tumors were similarly generated, and mice were provided with doxycycline in drinking water (2 mg/ml) starting with the day of tumor implantation. Tumors were measured every 48 hours using a caliper, and volume was calculated using the formula volume = length × width × height × 0.5. Animals were euthanized at the humane endpoint of 1000-mm3 tumor volume, weight loss of ≥ 20%, or acute physical distress in accordance with the IACUC guidelines. At the end of the study, the tumors were collected for flow cytometry analysis. Flow cytometry After ascites and tumor harvest, B16-F10 tumors were chopped and digested using deoxyribonuclease I (10104159001, Sigma-Aldrich) and Liberase (5401020001, Sigma-Aldrich) for 30 min at 37°C. Samples were strained through 70-μm filters and centrifuged. Percoll (40%) was used to remove fatty cells in B16-F10 tumors. ID8 ascites samples were centrifuged and treated with ACK lysis buffer (A1049201, Gibco) and then washed with PBS. Viability stain (table S7) was performed by incubating cells in PBS for 15 min. Cells were washed in PBS, incubated in Fc-blocking buffer 2.4G2 for 15 min, and stained with the extracellular antibodies (table S7) in Brilliant stain buffer (563794, BD Biosciences) for ID8 cells and FACS buffer for B16-F10 cells for 30 min on ice. Cells were fixed and washed using transcription factor staining buffer set (00-5523-00, Thermo Fisher Scientific) and permeabilization buffer according to the manufacturer’s protocol and then stained with the intracellular antibodies (table S7) in permeabilization buffer for 30 min on ice. Cells were resuspended in 200 μl of FACS buffer and acquired on the five-laser Cytek Aurora spectral flow cytometer (Cytek Biosciences). Data were analyzed using FlowJo software (BD Biosciences). To examine PD-L1 and MHC1 expression in ID8 and OAW28 sgNTC and sgSMARCA4 cell lines, cultured isogenic cells were harvested with trypsin and quenched with media. Cell pellets were washed in PBS and collected by centrifugation. Viability stain (table S7) was performed by incubating cells in PBS for 15 min. Cells were washed in PBS and incubated in Fc-blocking buffer 2.4G2 for 15 min. ID8 and OAW28 cells were stained with extracellular antibodies (table S7) for 30 min in blocking buffer. Cells were resuspended in 200 μl of FACS buffer and acquired on the five-laser Cytek Aurora spectral flow cytometer (Cytek Biosciences). Data were analyzed using FlowJo software (BD Biosciences). Transcriptomic analysis of TCGA data Normalized RNA-sequencing data from The Cancer Genome Atlas (TCGA) pan-cancer study (18) were obtained (https://gdc.cancer.gov/about-data/publications/pancanatlas). Immune-related scores were obtained from Thorsson et al. (17). Tumor types with cases in both datasets with a tumor purity ≥ 20% were included to assess SMARCA4 expression levels. The consensus purity estimates (47) were used to correlate with each parameter in our study. To avoid confounding factors related to tissue type, SMARCA4 groups in each cancer type were stratified: (i) 25% cases with lowest SMARCA4 expression levels (SMARCA4-low) and (ii) 25% cases with highest SMARCA4 expression levels (SMARCA4-high). Tumor type with < 3 cases in either group was excluded from our analyses. Then, SMARCA4-low–expressing cases were compared with SMARCA4-high–expressing cases across all cancer types and for each cancer type. ChIP-Atlas analysis Genome-wide transcription factor binding site information was obtained from ChIP-Atlas, an integrative database that covers public ChIP-seq data submitted to the National Center for Biotechnology Information Sequence Read Archive (48). Under PeakBrowser, IRF3 protein binding was visualized at a given genome locus TSS using IGV genome browser (43) using reference murine (version mm10) and human (version hg38) genome. Statistical analyses Data are presented as means ± SEM from at least three independent experiments. For TCGA RNA sequencing–based gene expression analyses, comparisons of immune-related scores between SMARCA4-low– and SMARCA4-high–expressing cancers were performed using the analysis of variance (ANOVA) test to account for purity as a confounding factor, followed by P-value adjustment using Benjamini-Hochberg method for multiple comparisons (38), where P < 0.05 was considered significant. For qRT-PCR and flow cytometry data analyses, comparisons between two groups (sgSMARCA4 versus sgNTC) were determined by unpaired t test (parametric or nonparametric) comparison test. P values of < 0.05 were considered significant. For RNA-sequencing data analyses for noncoding genes, statistical significance was considered if the adjusted P value (after Benjamini-Hochberg correction) was P < 0.05. For live imaging of tumor growth data, studies were analyzed by first reducing the entire data on a given animal to a single number, the area under the curve (AUC), and then comparing the groups using the AUCs as data points, via a Wilcoxon test. When two independent experiments were pooled, the statistical test was stratified by the experiment. For subcutaneous tumor growth data, studies were analyzed by performing unpaired t test on final endpoint of experiments and then comparing groups via log-rank test. P values of < 0.05 were considered significant. Statistical analyses were done using GraphPad Prism version 9 (GraphPad Software Inc., http://graphpad.com). Flow cytometry data were analyzed using FlowJo software (Tree Star Inc.).