SMARCA4 loss promotes late-stage tumor progression in non-neuroendocrine small-cell lung cancer

Abstract

Role of SMARCA4, a core component of the SWI/SNF chromatin remodeling complex, remains unclear in small-cell lung cancer (SCLC). Using genetically engineered mouse models, we found that Smarca4 deletion markedly reduced tumor formation and decreased expression of ASCL1, a key neuroendocrine lineage factor. However, Smarca4-deficient tumors, though smaller, exhibited aggressive features, including variant histology and loss of neuroendocrine differentiation. In established tumor cells, Smarca4 knockdown did not affect proliferation in vitro but unexpectedly promoted tumor growth in vivo, accompanied by reduced expression and cell-surface display of PVR, a ligand critical for T and NK cell activation. Although the contribution of these immune cells to SMARCA4 tumor-suppressor activity remains unknown, these findings suggest distinct roles of SMARCA4 in promoting early tumor development but restraining progression of late-stage, neuroendocrine-low tumors.

Introduction

SCLC accounts for 13% of all lung cancers and is an aggressive disease with a 5-year survival rate of 7% in patients (1). SCLC generally lacks actionable driver mutations such as mutant KRAS and EGFR found in NSCLC (2). Instead, it harbors intractable loss-of-function mutations in RB1 and TP53 and other tumor suppressor genes (2), presenting significant challenges in developing targeted therapies. Functional understanding of these alterations is necessary for gaining important insights into the biology of SCLC and to uncover novel preventive and therapeutic strategies.

Emerging evidence points to the SWI/SNF chromatin remodeling complex as a central player in cancer biology (3). The SWI/SNF complex controls gene accessibility to transcriptional machinery and its subunits, including the catalytic component and ATP-dependent helicase (SMARCA4), are mutated in around 20% of all cancers (4) and at a similar rate in a cohort of SCLC patients (Supplementary Figure S1A). SMARCA4 has been shown to act as a tumor suppressor in mouse models of lung adenocarcinoma (5,6), while its role in SCLC development remains largely unknown. Recent studies showed that pharmacological inhibition of SMARCA4 reduces the proliferation of SCLC cell lines and patient-derived cells (7–9) and knockdown of SMARCA4 causes vulnerability within cell lines deficient for MAX, a member of MYC family proteins (10). While indicating SMARCA4 dependency in SCLC cells, these studies relied on established tumor cell lines and have not directly addressed its role in SCLC development. It also remains unclear whether the inhibitory effect on some SCLC cells results from the dual-inhibition of SMARCA4 and SMARCA2, which may cause a synthetic lethality among the components of the SWI/SNF complex (11). While another study suggested that Smarca4 knockout decreases tumor incidence in Rb1/Trp53-mutant GEMM of SCLC (12), the exact role of Smarca4 in SCLC development remains unclear due to the scarce description of phenotypes in the GEMM and absence of mechanistic insight underlying the decreased tumor number. Whilst SMARCA4 mutations have been observed in a “real world” SCLC patient cohort (13), caution should be exercised as inactivating mutations in SMARCA4 are associated with SMARCA4-deficient undifferentiated tumors (SMARCA4-UT), a newly described tumor that is a mimic of SCLC (14). These observations together with the recent reclassification of YAP1-positive SCLC human cell lines as SMARCA4-UT (15) contribute to the confusion around a role of SMARCA4 in SCLC. Despite the growing interest in exploring the SWI/SNF complex components as prognostic biomarkers and therapeutic targets, critical gaps exist in understanding the roles they play in SCLC.

To address these unresolved questions, we used GEMMs and complementary in vitro models to dissect the functional role of SMARCA4 across stages of SCLC development. Our findings reveal that SMARCA4 supports early tumor formation but may restrain progression later in tumor development, highlighting a temporally distinct and dual role in SCLC biology. More broadly, this study underscores the complex interplay between chromatin remodeling, lineage plasticity, and immune evasion in the evolution of aggressive cancers.

Materials and Methods

Mouse strains, Adenovirus Cre infection, and allografts

The mouse strains employed Rb1^lox/lox; Trp53^lox/lox; Rbl2^lox/lox (RPR2), Rb1^lox/lox;Trp53^lox/lox (RP), and Rb1^lox/lox;Trp53^lox/lox; H11^{lox-stop-lox-MYCT58A} (RPM) and Smarca4^lox/lox were kindly provided by Drs. Anton Berns, Tyler Jacks, Julien Sage, Robert Wechsler-Reya, Trudy Oliver and Pierre Chambon and have been previously described (16–18). 6- to 8-week-old RP mice were intranasally administered with the adenovirus-CMV-Cre. For the RPR2 and RPM mice, adenovirus-Cre driven by either the CMV or CGRP promoter was intratracheally instilled in 10-week-old mice, including both male and female mice, for tumor induction. Ad-Cre viruses were purchased from the University of Iowa Gene Transfer Vector Core. Mice were aged 23 weeks for the RPR2 model and 10 weeks for the RPM model. All mice were genotyped before and after experiments by polymerase chain reaction (PCR) of tail DNA which was purified using lysis buffer containing proteinase K (Thermo Fisher Scientific, BP100–100). Mice were maintained according to practices prescribed by the National Institutes of Health and all animal protocols were approved by the Institutional Animal Care and Use Committee (IACUC) at the University of Virginia. For the RP mice, all animal experiments were conducted in accordance with the regulatory standards approved by the Animal Ethics Committee of the Walter and Eliza Hall Institute of Medical Research (WEHI). For allograft experiments, 5.0 × 10⁵ murine cells were injected in the flanks of B6.129S F1 mice (the Jackson Laboratory). Following implantation, mice were fed a doxycycline diet (625 mg/kg, Envigo). Perpendicular tumor diameters were measured using calipers. Volume was calculated using the formula L × W² × 0.52, where L is the longest dimension and W is the perpendicular dimension. The injected mice were maintained and observed for palpable tumors according to the procedures approved by the IACUC and euthanized when tumor size reached 1.5 cm in diameter, the endpoint of allograft study under the guideline of the institutional animal policy.

Histology, Immunostaining, and Immunoblots

At endpoint, the lungs were isolated and perfused with and incubated in 4% paraformaldehyde (PFA)/phosphate-buffered saline (PBS) for several days before paraffin embedding. Five micrometer-thick sections were stained with hematoxylin and eosin (H&E) staining. Additional slides were used for immunostaining. To dewax and hydrate the slides, they underwent sequential incubation in xylene, 100% ethanol, 95% ethanol, 80% ethanol, 70% ethanol, and PBS followed by antigen retrieval with EDTA. Antibodies used and their respective concentrations are listed in Supplementary Table S1. Macroscopic images of the lung sections were taken with the Olympus MVX10. Microscopic images of slides were acquired by the Nikon Eclipse Ni-U microscope. Image analysis and automated quantification of all H&E and immunostained slides completed using NIS-Elements Basic Research (Nikon). Quantification of immunostaining was performed on ≥5 randomly selected ROIs per tumor section. Thresholds were applied uniformly across samples. Scoring was conducted in a blinded fashion. Immunoblots were performed with protein extracted from mouse tissue acquired from both tumor tissue and culture cells using RIPA buffer (50mmol/L Tris-HCl pH7.4, 150mmol/L NaCl, 2mmol/L EDTA, 1% NP-40, 0.1% SDS). Generally, 10 μg protein lysate was loaded into various percentages of SDS-polyacrylamide gels for electrophoresis in PAGE running buffer (25 mM Tris, 192 mM glycine, 0.1% SDS, pH8.3). Proteins were then transferred to 0.2 or 0.45 μm PVDF membrane in transfer buffer (25 mM Tris, 192 mM glycine, 20% methanol). Membranes were blocked for 1 hour at room temperature using TBST buffer (150 mM NaCl, 10 mM Tris pH8.0, 0.1% Tween20) supplemented with 5% non-fat milk, then blotted with primary antibodies overnight at 4°C on an orbital shaker. After washing 5 times with TBST, membranes were incubated with secondary antibodies for 2 hours at room temperature and washed another 5 times with TBST. Membranes were then incubated with ECL Western Blotting Detection Reagent (Thermo Fisher Scientific, 32106) for 5 minutes, and signals were detected using a ChemiDoc machine (Bio Rad). A list of all primary and secondary antibodies along with the concentrations utilized is listed in Supplementary Table S1.

Multispectral Imaging

Methods for staining of paraffin-embedded lung sections for multispectral imaging has previously been described (19) and were implemented (20). Antigen retrieval was performed using AR9 buffer (PerkinElmer) according to the manufacturer instructions. OPAL Multiplex IHC Staining (PerkinElmer) was performed according to the manufacturer instructions. Antibodies used and their respective concentrations are listed in Supplementary Table S1. Slides were mounted using prolong diamond antifade (Life Technologies) and scanned at low magnification using the Vectra 3.0 system and software (PerkinElmer). Regions of interest were identified in Phenochart software, and high-powered magnification images were acquired with the Vectra 3.0 system (Akoya Biosciences). These images were spectrally unmixed using single stain positive controls and analyzed using the InForm software (PerkinElmer). Immune cells were quantified using HALO software (Indica Labs, Albuquerque, NM).

Flow Cytometry

Cells were stained with a fixable live/dead cell stain (BioLegend Cat# 423105) at 1:1000 dilution for 15 min at room temperature, followed by staining with fluorophore-conjugated antibodies, listed in Supplementary Table 1, for 20 min on ice and in the dark. Cells were washed twice with a FACS buffer (1% BSA, 1mM EDTA in PBS) after every staining step and analyzed using an Aurora Borealis 5 laser Spectral Flow Cytometer at the University of Virginia Flow Cytometry Core, RRID: SCR_017829. Compensation was completed if samples were stained with multiple antibodies. The data were analyzed using FlowJo software. Example gating strategy is provided in Supplementary Figure S4.

Quantitative RT-PCR

For quantitative RT-PCR (RT-qPCR), total RNA was isolated from cells or tumors using the RNeasy Mini Kit (Qiagen, 74106) according to the manufacturer’s protocol. cDNA was generated by the ProtoScript II First-Strand cDNA Synthesis Kit (New England Biolabs, E6560). RT-qPCR was performed using SYBR Green (Thermo Fisher Scientific, 4367659) with Applied Biosystems 7900 or StepOnePlus following the manufacturer’s protocol. All qPCR primer sequences were retrieved from the online Universal Probe Library at Roche and included in Supplementary Table S1.

Chemicals, vectors, and virus production

A dual SMARCA2/SMARCA4 inhibitor, FHD-286, was obtained from MedChemExpress. Tet-pLKO-puro plasmid was obtained from Addgene (a gift from Dmitri Wiederschain). The shRNA oligo nucleotide sequences are listed in Supplementary Table S1. Lentivirus particles were produced by co-transfecting the Tet-PLKO-puro plasmid with packaging plasmids psPAX2 and pMD2.G (gifts from David Rekosh and Lou Hammarskjold), into HEK 293T/17 cells using polyethylenimine (Sigma-Aldrich, 408727). Forty-eight hours post-transfection, viral supernatants were harvested, centrifuged to remove cell debris, and stored at −80°C. Viral titers were determined by quantitative PCR (qPCR). Cells were transduced with lentiviral supernatant in the presence of DEAE-Dextran to enhance transduction efficiency. Puromycin selection (Thermo Fisher Scientific, A1113803) was applied 48 hours after transduction to enrich for successfully transduced cells, based on puromycin resistance conferred by the Tet-pLKO-puro vector. For inducible knockdown, puromycin-selected cells were treated with 1 uM doxycycline for 48 hours to induce shRNA expression. Knockdown efficiency was confirmed by RT-qPCR.

Cells and soft agar assays

Murine SCLC cells were isolated from lung tumors developed in the RPR2 and RPM GEMMs. Precancerous cells (preSCs) were derived from early-stage neuroendocrine lesions in the RP GEMM (21). All SCLC cells were cultured in RPMI-1640 media supplemented with 10% BGS and 1% penicillin-streptomycin-glutamine. 293T cells were cultured in DMEM supplemented with 10% BGS and 1% penicillin-streptomycin. These cells were tested negative for mycoplasma using the PlasmoTest-Mycoplasma Detection Kit (InvivoGen, rep-pt1). For soft agar assays, 1×10⁴ cells per well in 0.5mL of growth medium containing 0.35% low-melting point agarose (Invitrogen, 16520–100_ and seeded on top of a 0.5mL base layer of medium containing 0.5% agar. After 3 weeks, colonies were fixed with 10% MeOH and 10% acetic acid and then stained with 1% MeOH, 1% Formaldehyde, and 0.05% crystal violet. Images of the wells were acquired using a Chemidoc and counted using the NIS-Elements Basic Research (Nikon) imaging software. All of the cell culture experiments were performed in triplicates and repeated for a minimum of two biological replicates.

Gazdar Small Cell Lung Cancer Neuroendocrine Explorer

The heatmap of RNA-seq Expression (log2 PRKM + 1) of SMARCA4, ASCL1, and MYC from 19 SCLC PDX samples (22) generated using the Gazdar Small Cell Lung Cancer Neuroendocrine Explorer (https://lccl.shinyapps.io/GSNE/) (23).

Statistical analysis

GraphPad Prism was utilized for all statistical analysis as the mean ± standard deviation and evaluated using an unpaired two-tailed Student’s t-test when comparing between wild-type and Smarca4-targeted samples. A two-way ANOVA with multiple comparisons was used for analysis between genetic GEMM backgrounds. p<0.05 was considered statistically significant. Kaplan-Meier curves were used for plotting patient survival from (2,24) with HR as a continuous variable and log-rank test was used to determine significance. Plots were generated and statistics were calculated using the SurvSig webtool (https://survsig.hcemm.eu/) (25).

Data availability

The RNA-seq data used to assess transcript levels of SMARCA4 and NE markers in RPM cells cultured at varying timepoints have been published and are available under GEO accession number GSE149180 (26). The NE score was calculated using Zhang et al. signature (27). The ChIP-seq data to assess SMARCA4 binding have been published and are available under GEO accession number GSE256346 (8).

Results

Smarca4 loss inhibits tumor development in RPR2 mice

To determine the impact of losing SMARCA4 on SCLC development, we utilized a GEMM known as RPR2 mice. In this model, tumors are initiated by adenovirus Cre-mediated deletion of Rb1, Trp53, and Rbl2 floxed alleles in the airway epithelium and recapitulate ASCL1-dominant human SCLC (17). Following genetic crosses, we generated RPR2 mice and their littermates carrying additional homozygous floxed alleles of Smarca4 (hereafter RPR2S) and infected them with adenovirus Cre via intratracheal instillation (Fig. 1A). Six months after infection, lung tumor burden (tumor area/total lung area) was significantly lower in RPR2S mice compared to RPR2 littermates (Fig. 1B). RPR2S tumors had a significant decrease in the levels of proliferation marker phosphorylated histone H3 (pHH3) and concomitant increase in the levels of apoptotic marker cleaved Caspase 3 (CASP3) compared to controls (Fig. 1C; Supplementary Figure S1B). SMARCA4 loss was confirmed by IHC (Supplementary Figure S1C). Soft agar assays showed primary cells isolated from RPR2S tumors formed significantly fewer colonies than RPR2 cells (Fig. 1D), indicating the inhibitory effect of losing SMARCA4 on cell expansion in vitro. To determine the role of SMARCA4 in the initial stages of tumor development, we inhibited the protein in precancerous cells (preSCs), immortalized Rb1/Trp53-mutant NE cells derived from early lesions developed in GEMMs (21). Treatment of a dual SMARCA4/2 inhibitor, FHD-286, resulted in significantly fewer colonies than the vehicle treated cells in soft agar assays (Fig. 1E). Together, these results suggest that SMARCA4 is necessary for SCLC development and its functional loss hinders the proliferative capacity of RPR2 tumor cells and preneoplasia.

Figure 1. Smarca4 loss inhibits tumor development in RPR2 mice.

A. Schematic of tumor induction in RPR2 GEMM. B. Representative images of H&E-stained lung sections (scale bars: 5mm) and quantification of tumor burden in the GEMMs normalized to the control RPR2 lungs (n=6 per group). C. Quantification of pHH3 foci and cleaved CASP3 foci within region of interest (ROI) in the lungs of RPR2 and RPR2S (n=30 ROIs). D. Representative images of soft agar containing cells isolated from RPR2 or RPR2S tumors and quantification of colony number. E. Representative images of soft agar containing preSCs treated with either DMSO or FHD-286 and quantification of colony number (n=3). F. Representative H&E images of the sections of lungs in GEMMs as indicated. (scale bars: 50μm) G. CellProfiler quantification of average number of nuclei and average size of nuclei per ROI from H&E-stained slides (F) (n=24 ROIs). H. Plots of percentage of low-grade and high-grade tumors present in the lungs of RPR2 (n=7) and RPR2S (n=6) mice. I. Representative images of CGRP-stained lung sections (scale bars: 50μm) and quantification of CGRP stain (n=10 per group). AR: alveolar region. J. RT-qPCR data showing Ascl1 transcript level, normalized to levels of Gapdh, in RPR2 and RPR2S cells (n=3 per group). K. Immunoblots for ASCL1 and VINCULIN in various cells as indicated. Bottom, ratio of ASCL1 levels relative to VINCULIN levels, measured by densitometry.

To elucidate the mechanism by which Smarca4 loss inhibits tumor development, we examined the histological features of Smarca4-deficient tumors. Unlike the classical NE morphology observed in RPR2 tumors and Rb1/Trp53-deficient tumors (henceforth RP)(16) (Fig. 1F) – characterized by cells with scant cytoplasm and arranged in tight nests or occasional rosettes – RPR2S and Smarca4-deficient RP (RPS) tumors displayed variant histology, including increased cytoplasmic volume and distinct cell boundaries (Fig. 1F; Supplementary Figure S1D). These differences were measured using CellProfiler to confirm an increase in nuclei size and decrease in nuclei number per region of interest (ROI) in the Smarca4-deficient tumors (Fig. 1G). Classification by an independent pathologist revealed that there was an increase in the proportion non-nested tumors with diffuse growth and parenchymal invasion in the RPR2S mice compared to controls (Fig. 1H). This variant morphology of RPR2S tumors was associated with markedly lower levels of CGRP, a well-described NE marker (Fig. 1I). To determine if the reduction in CGRP resulted from loss of NE differentiation or if RPR2S tumors were initiated in a non-NE cell type, RPR2 and RPR2S mice were infected with adenovirus Cre driven by the Cgrp promoter (Ad-CGRP-Cre) to induce tumorigenesis specifically in NE cells (28). After infection with Ad-CGRP-Cre, RPR2S tumors showed a significant reduction in levels of CGRP, confirming the loss of NE features is due to Smarca4 loss (Supplementary Figure S1E).

NE differentiation is tightly controlled by ASCL1, a transcription factor necessary for tumor development in RPR2 mice(29). Because the SWI/SNF complex and ASCL1 functionally interact during neural development(30), we questioned whether SMARCA4 loss influences ASCL1 expression in tumors. ASCL1 transcript and protein levels were significantly decreased in RPR2S tumors compared to controls (Fig. 1J, K). A causal relationship between the proteins in preSCs was shown as inhibition of SMARCA4 using FHD-286 significantly reduced ASCL1 expression in RT-qPCR and immunoblots (Supplementary Figure S1F). Taken together, these findings suggest that SMARCA4 loss results in inhibition of tumor development and loss of NE differentiation via ASCL1 loss, uncovering the role of the SMARCA4-ASCL1 axis early in SCLC development.

SMARCA4 is necessary for tumor development in RPM mice

A recent study suggests that the spontaneous loss of NE differentiation occurs in a SCLC GEMM in which tumor development is initiated by deletion of Rb1 and Trp53 and driven by a stable form of MYC (henceforth RPM) (18,26). We tested if Smarca4 deletion differentially affects tumor development in RPM mice (henceforth RPMS) (Fig. 2A). Ten weeks after Ad-CGRP-Cre infection, the tumor burden was significantly lower in the lungs of RPMS mice compared to RPM littermates (Fig. 2B). Intriguingly, pHH3 levels were similar between the RPM and RPMS tumors, while cleaved CASP3 levels were significantly increased in RPMS tumors (Fig. 2C). Notably, SMARCA4 protein was detected in some of the tumors in RPMS mice although its level was markedly lower than preSCs and almost undetectable (Fig. 2D). Similar to the observations in RPR2 mice, the RPMS tumor had decreased expression of both ASCL1 and CGRP compared to RPM controls (Fig. 2E, F). These findings along with the observations in the RPR2 model (Fig. 1I–J) clearly indicate a crucial role of SMARCA4 in regulating ASCL1 and the tumor initiation capacity of SCLC. These results combined with the established necessity of ASCL1 for tumor formation in both the RPR2 and RPM models (29,31) strongly suggest that the reduced tumor burden observed with loss of Smarca4 is due, in part, to the disruption of the SMARCA4-ASCL1 axis. This relationship along with the retention of SMARCA4 in some RPMS tumors indicate its necessity for tumor development.

Figure 2. SMARCA4 is necessary for tumor development in RPM mice.

A. Schematic of tumor induction in the RPM GEMM. B. Representative images of H&E-stained lungs of RPM (control) and RPMS (Smarca4-knockout) GEMMs, 10 weeks after Ad-CGRP-Cre infection, and enlarged images of highlighted regions depicting nodular tumors. (scale bars: top, 5mm; bottom, 1mm) Right, quantification of tumor burden in the GEMMs normalized to the control RPM lungs (n=7 RPM and n=5 RPMS). C. Representative images of pHH3-and cleaved CASP3-stained RPM and RPMS lung sections (scale bars: 50 μm) and quantification of pHH3 foci and cleaved CASP3 foci within region of interest (ROI) in the lungs of RPR2 and RPR2S (n=20 ROIs). D. Immunoblot for SMARCA4 expression in RPMS cells compared to preSC and RPM cells. RPMS1/2 represent individual tumors isolated from different RPMS mice. Ratio of SMARCA4 levels relative to VINCULIN levels, measured by densitometry, shown on the bottom. E. Left, RT-qPCR data showing Ascl1 transcript level, normalized to levels of Gapdh, in RPM and RPMS cells (D) (n=3 per group). Right, Immunoblot for ASCL1 expression in RPMS2 cells (utilized in D) compared to RPM cells. Ratio of ASCL1 levels relative to VINCULIN levels, measured by densitometry, shown on the bottom. F. Representative images of CGRP-stained RPM and RPMS slides (scale bars: 50 μm) and quantification of CGRP levels (n=10 ROIs per group). AR: alveolar region. G. Representative images of multiplex IHC in RPM and RPMS slides stained for CD4, CD8, GZMB. (scale bars: 50 μm). H. Quantification of immune cells (CD4+ T cells, CD8+ T cells, GZMB+ cells, GZMB/CD8 Dual-positive cells) per tumor area (cells/mm²) in the lungs of RPM and RPMS mice. Each dot represents a nodular tumor (n=3 mice per group).

The decreased tumor burden in RPR2S mice may be attributed to the decreased proliferation and increased apoptosis observed in the tumors (Fig. 1C). However, the RPMS tumors only displayed increased levels of apoptosis (Fig. 2C). This led us to explore whether cell-extrinsic factors contributed to the slower tumor development in RPMS mice. The tumor microenvironment is complex and made up of various factors that can influence tumor growth. This includes immune cells that can hinder tumor growth and induce apoptosis. We performed multiplex IHC using the Vectra platform to visualize immune cell infiltration within RPM and RPMS tumors. SCLC is largely considered to be “immune-excluded” with immune cells aggregating around the tumor edge and rare infiltration into the core of patient tumors (32). We observed a similar localization pattern of immune cells at the periphery of the tumors. The RPMS tumors and adjacent periphery had a significant increase in the number of CD4+ and CD8+ T cells (Fig. 2G–H). CD8+ T cells can act directly on tumor cells through the release of Granzyme B (GZMB), a serine protease that can trigger apoptosis in target cells (33). The RPMS tumors had a significant increase in the number of CD8/GZMB dual-positive T cells (Fig. 2G–H). These findings indicate an increase in both the number and activity of immune cells within the RPMS tumors (Fig. 2G–H). We propose that both cell-intrinsic factors, including the disruption of the SMARCA4-ASCL1 axis, combined with the cell-extrinsic increased immune presence immediately surrounding the RPMS tumors contribute to the increased levels of apoptosis and ultimately the reduced tumor burden observed upon loss of Smarca4. Given these observed differences in immune cells in the RPM background, we measured immune cell presence across the RP and RPR2 backgrounds (Supplementary Figure S2). Immune cell presence was generally low across these models with no significant differences based on Smarca4 status. These observations led us to speculate that the impact of Smarca4 loss on the microenvironment is specific to MYC-driven SCLC, modeled by the RPM GEMM. While there was no significant difference between the wild-type and Smarca4-targeted RP and RPR2 mice, we noted intriguing differences in immune cell presence across the different GEMM backgrounds. The RPM GEMMs had the highest number of immune cell presence with the lowest abundance in the RP background. These observations echo recent studies showing an inverse relationship between NE status and immune cell presence in SCLC tumors (32).

Acute SMARCA4 knockdown in tumor cells increases the tumorigenic potential

The most intriguing observation in Smarca4-deficient RPR2 and RPM tumors is that the tumors, albeit smaller, display more aggressive histology and molecular features of late-stage tumors. This aggressive phenotype is marked by a poorly differentiated morphology that coincided with loss of NE differentiation. These apparently conflicting phenotypes may be attributed to the shortcomings of the GEMMs in which Rb1, Trp53 and Smarca4 are simultaneously deleted - an event unlikely to occur in patient tumors. Additionally, immunoblots revealed a spontaneous loss of SMARCA4 expression in cells from more malignant models of SCLC progression, particularly from the RPM model, compared to preSCs or RPR2 cells (Fig. 2D; Supplementary Figure S3A). In a study showing gradual loss of NE features, defined by applying an established NE signature (27), in RPM cells during a 21 day period in culture (Supplementary Figure S3B) (26), suggesting that the loss of SMARCA4 may occur during the later stages of tumor development. Analyzing the transcriptome data of SCLC PDXs, we noted that low SMARCA4 expression correlated with higher levels of MYC (Fig. 3A) which was interesting given the decreased tumor burden in RPMS mice (Fig. 2B). These observations led us to develop models that better recapitulated the temporal order of alterations in patient tumors. To determine the functional consequences of SMARCA4 loss in established tumor cells, we transduced RPM cells, that had maintained expression of SMARCA4 and NE status, with lentivirus containing either Smarca4 or non-targeting shRNA (Fig. 3B) and validated knockdown using immunoblots and RT-qPCR (Fig. 3C; Supplementary Figure S4A). Doxycycline-induced Smarca4 knockdown significantly reduced levels of Ascl1 transcript and ASCL1 protein as assessed by RT-qPCR and immunoblots (Fig. 3C; Supplementary Figure S4B). Smarca4 knockdown did not significantly affect the colony-forming capacity of the RPM cells in soft agar assays (Fig. 3D). This lack of impact is notable, compared to the inhibitory effect of SMARCA4 loss on the proliferation of preSCs and Smarca4-deficient RPR2 tumor cells (Fig. 1), and yet it does not support the role of SMARCA4 as a tumor suppressor. We reasoned that if not affecting cell-intrinsic changes in vitro, Smarca4 knockdown could influence the tumorigenic capacity of tumor cells in vivo. To test this idea, we implanted 5×10⁵ cells in the flanks of athymic nude mice and measured tumor growth. Smarca4 knockdown tumors grew significantly faster than controls (Fig. 3E; Supplementary Figure S4C). Intriguingly, whereas levels of pHH3 was not significantly different between the groups, levels of cleaved CASP3 were lower in Smarca4 knockdown tumors (Supplementary Figure S4D). We postulated that the enhanced tumor growth in vivo is due to altered tumor-host interactions, especially innate immune cells, such as NK cells, that remain functionally intact in athymic nude mice. Flow cytometry for NK cell interacting ligands on tumor cells showed that polio virus receptor (PVR/CD155) levels on pre-injected Smarca4 knockdown cells were significantly reduced compared to controls (Fig. 3F; Supplementary Figure S5A) while levels of other NK cell interacting ligands (PVRL2, ULBP1, RAE1, and CD276) were generally low and not significantly different between the groups (Fig. 3F; Supplementary Figure S5B). RT-qPCR and immunoblots showed that PVR expression was significantly downregulated in Smarca4 knockdown cells compared to controls (Supplementary Fig. S5C). Additionally, analysis of chromatin immunoprecipitation sequencing (ChIP-seq) of four SCLC PDXs from a recent study revealed that SMARCA4 occupies regions of the PVR loci (Fig. 3G)(8). We transduced RPM cells with a second Smarca4-targeting shRNA and observed similar expression patterns showing concomitant decrease in ASCL1 and PVR expression in Smarca4-knockdown cells (Supplementary Figure S5D).

Figure 3. Acute SMARCA4 knockdown in tumor cells increased the tumorigenic potential.

A. Heatmap showing expression of SMARCA4, ASCL1, MYC (rows) in 19 patient-derived xenografts (columns) arranged by NE score (22). Plot generated with the Gazdar Small Cell Lung Cancer Neuroendocrine Explorer (23). B. Schematic of targeting SMARCA4 in cells isolated from RPM tumors. C. Immunoblots for SMARCA4 and ASCL1 from non-targeted (shNT) and shSmarca4-1 cells treated with 1 uM doxycycline for 48 hrs. Ratio of SMARCA4 levels to VINCULIN levels or ASCL1 levels to VINCULIN levels, measured by densitometry, shown on the bottom. D. Representative images of soft agar containing shNT and shSmarca4-1 RPM cells treated with 1uM doxycycline for the duration of the experiment and quantification of colony number. E. Measurements of tumor volume of shNT and shSmarca4-1 subcutaneous allografts in athymic nude mice (n=3 per group). F. Representative plots of flow cytometry quantification of PVR-positive cell population in shNT or shSmarca4-1 RPM cells treated with doxycycline for 48 hours. Right, flow cytometry quantification of positive PVR (n=3), PVRL2 (n=3), ULBP1 (n=2), RAE1 (n=2), and CD276 (n=2) cell populations in shNT and shSmarca4-1 RPM cells treated with doxycycline for 48 hours. G. ChIP-Seq for SMARCA4 at the PVR loci in 4 SCLC PDXs (LX95, LX276, LX761, LX891) adapted from (7). H. Model depicting the temporal-specific role of SMARCA4 in SCLC progression. I. Kaplan-Meier curve of SCLC patients from the highest and lowest quartile after stratification based on SMARCA4 expression (n=31 per group) (2,24,25). Hazard ratio (HR) and p values by log-rank test are indicated.

Since PVR on tumor cells interacts with DNAX accessory molecule-1 (DNAM-1) on NK cell surface and can activate NK cell mediated cytotoxicity (34), the decreased PVR display may render Smarca4 knockdown cells a selective advantage to avoid NK cell mediated killing in subcutaneous allografts. To examine this relationship in the GEMM models used in Figs. 1 and 2, we used flow cytometry to measure the levels of PVR between RPR2 and RPR2S cells and between RPM and RPMS cells. PVR expression was generally absent in the RPR2 background. In the RPM background, Smarca4-knockout cells had a significant decrease in PVR expression compared to controls (Supplementary Figure S6), providing yet another example for a uniquely immunogenic profile of the RPM GEMM.

To determine if NK cell infiltration was different between control and Smarca4 knockdown tumors, we used IHC to stain for the NK cells which were generally absent from the tumor core but remained at the tumor periphery (Supplementary Figure S7A). We did not observe any differences in NK cell recruitment between the two tumor groups (Supplementary Figure S7B). This lack of difference may be because tumors were isolated at different timepoints once each individual tumor grew to 1.5cm in length (average time to endpoint: 21 days for Smarca4 knockdown tumors and 37 days for controls). Therefore, we speculate that the levels of NK cells may differ in the earlier stages of tumor development and future studies should focus on a fixed timepoint consistent between groups. Additionally, we wanted to determine NK cell presence in the GEMM models. NK cells were completely absent from lung tumor periphery in both the wild-type and Smarca4-deficient RPR2 and RPM lungs (Supplementary Figure S7C). The differences in NK cell presence between the lung and flanks suggest that NK cell control of SCLC tumors is more important outside of the lungs, in line with recent studies implicating NK cells in SCLC metastasis (35,36). Although NK cells were generally absent from lung tumors, we must acknowledge that PVR can interact with DNAM-1 also found on T cells (37), which were present in lung tumors, especially from the RPM background (Fig. 2G–H). Therefore, functional studies should be conducted to determine the role of PVR in the interactions and activation of both T cells and NK cells in driving the tumor suppressive effect of SMARCA4 loss.

Our findings provide novel insight into the temporal specific role of SMARCA4 in SCLC development. The SMARCA4 dependency in tumor development was discovered thanks to the approach of deleting Smarca4 simultaneously with the tumor-initiating alterations in GEMMs. The expression pattern of SMARCA4 spontaneously decreasing from preneoplastic cells to tumor cells led to the idea that while required in preneoplastic lesions progressing towards tumor, it acts as a tumor suppressor in fully developed tumors (Fig. 3H). The tumor suppressive effect of SMARCA4 is validated in part by our finding that acute Smarca4 knockdown enhanced the tumorigenic capacity of tumor cells. To understand the impact of SMARCA4 levels in SCLC patients, patient tumors from two datasets (2,24) were stratified based on their SMARCA4 expression levels (either upper or lower quartile), revealed that low SMARCA4 expression correlated with worse overall survival (Fig. 3I) (25). This result is consistent with the tumor-promoting effect of low SMARCA4 levels observed in mouse models. We wanted to determine how expression of key NE and immune markers differed between the tumors separated between the upper and lower quartiles of SMARCA4 expression (Supplementary Figure S8). We did not observe any clear patterns of differential expression in the markers assessed between these groups but believe that the lack of differences may be due to the small sample size. Future work should explore expression of SMARCA4 in relation to these markers in a larger cohort of SCLC samples.

Discussion

The concept of differential impacts of SMARCA4 with respect to the early versus late period of SCLC development is significant in that it addresses a fundamental question how the timing of tumor suppressor gene alterations shapes tumor evolution. However, the potential tumor suppressor role of SMARCA4 appears contradictory to its dependency in SCLC cells as suggested in recent studies where targeting SMARCA4 selectively inhibits the proliferation of POU2F3-positive SCLC cell lines representing a non-NE subtype of SCLC (7). The fact that the Smarca4 knockdown RPM cells used in our study do not model the POU2F3-positive SCLC cells may explain the difference in SMARCA4 dependency. Along this line, another study showed that dual inhibition of SMARCA4/2 inhibited proliferation in patient, NE+ SCLC cell lines that are seemingly stuck in their differentiation state and are less likely to undergo loss of NE differentiation status, unlike the RPM cells used to knockdown Smarca4 (8). While these recent studies showed the inhibitory effect of pharmacological dual-inhibition of SMARCA4 and SMARCA2 on SCLC cells, the effect could indicate a synthetic lethality among the components of SWI/SNF complex. It is also worth noting that Smarca4-knockdown enhanced the tumorigenic capacity of cells via a mechanism independent of cell proliferation. Our findings suggest a novel mechanism involving altered interactions in the tumor microenvironment. Specifically, the down-regulation of PVR on Smarca4-knockdown cells reduces its binding to DNAM-1 to activate T cells and NK cell mediated cytotoxicity (37). This relationship is notable and requires further studies for functional validation, which may include in vitro co-culture cytotoxicity assays or in vivo depletion of either CD8+ T cells or NK cells. This SMARCA4-PVR axis may provide novel insight into the emerging role of NK cells in SCLC growth and progression (35,36) and suggest a need to evade NK cell surveillance in late-stage SCLC. In conclusion, the differential impact of losing SMARCA4 at the various stages of SCLC development revealed novel insights into the molecular patterns that drive the transformation of these tumors and reveal the varying vulnerabilities that may arise. SMARCA4 may prove to be an attractive target for SCLC prevention.

Our results provide intriguing insights into the temporal-specific role of SMARCA4 in SCLC. However, these conclusions were drawn from use of separate, independent models. To understand how loss of SMARCA4 impacts spontaneous tumor formation in vivo, we propose use of a recombinant lentiviral vector containing the Cre-recombinase and a conditional shRNA targeting Smarca4 in established SCLC GEMMs. In this model, the Cre will initiate tumor formation while the conditional shRNA will target Smarca4 once tumors have started to grow. Induction of Smarca4 loss can be controlled by administration of a doxycycline diet. This experimental design will provide insights into the impact of targeting Smarca4 in a continuous manner and eliminate the experimental variation that may have been introduced in our models described in Fig. 3. Although our results show a crucial role of SMARCA4 regulating NE differentiation markers and PVR, any additional conclusions about the broader transcriptional impact are limited. Future work should use both RNA-seq and ATAC-seq to understand global transcriptional changes occurring when SMARCA4 is lost across the various stages of tumor development.

Implications:

This study underscores the importance of tumor context and timing in understanding and targeting SMARCA4 and other chromatin regulators in SCLC.