SMARCAL1 is a dual regulator of innate immune signaling and PD-L1 expression that promotes tumor immune evasion

SUMMARY

Genomic instability can trigger cancer-intrinsic innate immune responses that promote tumor rejection. However, cancer cells often evade these responses by overexpressing immune checkpoint regulators, such as PD-L1. Here, we identify the SNF2-family DNA translocase SMARCAL1 as a factor that favors tumor immune evasion by a dual mechanism involving both the suppression of innate immune signaling and the induction of PD-L1-mediated immune checkpoint responses. Mechanistically, SMARCAL1 limits endogenous DNA damage, thereby suppressing cGAS-STING-dependent signaling during cancer cell growth. Simultaneously, it cooperates with the AP-1 family member JUN to maintain chromatin accessibility at a PD-L1 transcriptional regulatory element, thereby promoting PD-L1 expression in cancer cells. SMARCAL1 loss hinders the ability of tumor cells to induce PD-L1 in response to genomic instability, enhances anti-tumor immune responses and sensitizes tumors to immune checkpoint blockade in a mouse melanoma model. Collectively, these studies uncover SMARCAL1 as a promising target for cancer immunotherapy.

In Brief

The DNA translocase SMARCAL1 favors tumor immune evasion by two distinct mechanisms: it suppresses the cGAS-STING pathway by limiting endogenous DNA damage and induces PD-L1 expression by modulating chromatin accessibility at a PD-L1 regulatory element.

Graphical Abstract

INTRODUCTION

Immune regulation plays a pivotal role in cancer biology^1,2. Studies of tumor-immune system interactions have yielded potent cancer treatments, such as immune checkpoint blockade (ICB) therapies targeting the cytotoxic T lymphocyte antigen 4 (CTLA-4)³ or the programmed cell death 1 (PD-1) receptor and its ligand (PD-L1)^4–6, which induce robust and durable responses against numerous cancer types⁷. Nonetheless, tumors often employ immunosuppressive mechanisms to attain ICB resistance⁸. Consequently, new strategies to enhance the efficacy of ICB therapies are of paramount importance⁹.

The DNA damage response (DDR) and chromatin regulation are both crucial for maintaining genome stability and controlling gene expression^10,11. Recent studies show that targeting certain DDR factors or chromatin regulators in cancer cells can enhance anti-tumor immune responses by inducing cancer-intrinsic innate immune signaling^12–19. This signaling cascade is initiated by pattern-recognition receptors (PRRs), such as the RIG-I-like receptors and cytosolic nucleic acid-sensing cGAMP synthase (cGAS), which recognize cytosolic RNA or DNA²⁰, eliciting the expression of type I interferons (e.g., IFN-α and IFN-β) by interferon regulatory factors (IRFs) and NF-κB, followed by JAK/STAT-dependent expression of cytokines and other interferon-stimulated genes (ISGs)²⁰. Consequently, the secretion of IFNs and ISGs into the tumor microenvironment (TME) by cancer cells facilitates tumor antigen presentation and promotes the recruitment and activation of cytotoxic T cells that mediate tumor rejection^20–23.

Tumor cells employ multiple immuno-suppressive mechanisms to counteract the activation of cancer-intrinsic innate immunity⁸. For example, in response to IFN signaling resulting from chemotherapy- and radiotherapy-induced DNA damage, DNA repair inhibition, or chromatin deregulation, cancer cells can evade immunity by overexpressing the immune checkpoint regulator PD-L1^{13–17,24,25}. Thus, clinical strategies that induce cancer-autonomous immune-stimulatory signals, while simultaneously limiting PD-L1-mediated immune evasion, should enhance anti-tumor immune responses.

In this study, we show that the DDR factor SMARCAL1 promotes cancer immune evasion via two complementary mechanisms. Specifically, SMARCAL1 deficiency induces cancer-autonomous immune signaling, while concomitantly downregulating PD-L1 levels, despite the IFN signaling that arises from the activation of the innate immune response. These findings identify SMARCAL1 as a regulator of anti-tumor immunity.

RESULTS

CRISPR-Cas9 screens identify cancer cell-autonomous immuno-regulatory factors

To identify DDR and chromatin factors that regulate both cancer-intrinsic immune signaling and PD-L1 levels, we devised a fluorescence-activated cell sorting (FACS)-based CRISPR-Cas9 screening strategy that uses as a readout dual staining for PD-L1 and the nuclear fraction of the IRF3 transcription factor (TF), a marker of cell-intrinsic innate immune activation²⁶. Our screens were conducted in the triple-negative breast cancer MDA-MB-436 cells, which display intact immune signaling (Figure S1A) and high basal levels of PD-L1 that can be further induced by type I IFN (Figure S1B)²⁵. MDA-MB-436 cells were transduced with a lentiviral sgRNA library targeting 609 genes (library 1), including DDR factors, chromatin regulators and cell cycle regulators, and were then collected as a pool (unsorted) or sorted into three populations based on nuclear IRF3 and surface PD-L1 levels (IRF3^High/PD-L1^Low, IRF3^High/PD-L1^High, IRF3^Low/PD-L1^Low) (Figures 1A–B and S1C–D). Multi-comparison analysis²⁷ of sgRNA representation in the isolated cell populations identified genes targeted by multiple sgRNAs that were depleted or enriched in the sorted compared to unsorted populations (Figure 1C–E). Among the most significantly depleted sgRNAs in cells with high nuclear IRF3 (IRF3^High/PD-L1^Low and IRF3^High/PD-L1^High), we identified sgRNAs targeting IRF3 itself (Figure 1C–D). In contrast, sgRNAs targeting CTCF, RAD21, and SMC3, negative modulators of PD-L1 and cell-intrinsic immune signatures²⁸, were enriched in cells expressing high levels of nuclear IRF3 and PD-L1 (IRF3^High/PD-L1^High) (Figure 1D), indicating that our screen can identify factors that simultaneously regulate IRF3-mediated immune-stimulatory signals and PD-L1 levels.

Figure 1. FACS-based CRISPR-Cas9 genetic screens in MDA-MB-436 triple negative breast cancer cells.

(A), Gene network targeted by sgRNA library 1. The size of nodes (genes) reflects the number of genetic and physical interactions (gray lines) between nodes, according to PrePPI¹¹⁸.

(B), Schematic of CRISPR screens to identify regulators of nuclear IRF3 and PD-L1. Following selection of MDA-MB-436 cells transduced with the lentiviral sgRNA library 1, cells were collected as a pool (unsorted), or sorted into three populations, as indicated. sgRNA abundance in distinct cell populations was determined by next-generation sequencing.

(C-E), Distribution of normalized β-scores for genes targeted by sgRNA library 1 and ranked according to the abundance of their sgRNAs in the examined groups. The normalized β-scores for genes in the IRF3^High/PD-L1^Low (C), IRF3^High/PD-L1^High (D), and IRF3^Low/PD-L1^Low (E) populations were calculated with MAGeCK MLE using the unsorted population as a reference. Red and blue dots represent genes associated with enriched and depleted sgRNAs (positive or negative β-score and p-value <0.05), respectively.

Loss of SMARCAL1 triggers a cell-intrinsic immune response and downregulates PD-L1

Our screen also identified twenty-two genes targeted by sgRNAs specifically enriched in IRF3^High/PD-L1^Low cells (Figure S1E–F and Table S1), including SMARCAL1, a DNA translocase mutated in Schimke immuno-osseous dysplasia (SIOD)²⁹ (Figure 1C and Table S1). While SMARCAL1 maintains genome stability in response to replication stress^30–34, it can also regulate gene expression^35–38. To examine the impact of SMARCAL1 loss on PD-L1 and IRF3 levels, we generated SMARCAL1-knockout (SMARCAL1-KO) MDA-MB-436 cells with CRISPR-Cas9 or depleted SMARCAL1 in MDA-MB-436 cells with siRNAs (Figure 2A). Consistent with our screen data, SMARCAL1-deficient cells exhibited reduced PD-L1 levels by immunoblotting and flow cytometry (Figures 2A and S2A). Furthermore, SMARCAL1-deficient cells displayed higher levels of IRF3 phosphorylation on serine 386 (S386), a known readout of IRF3 activation³⁹ (Figure 2B). Collectively, these findings demonstrate that SMARCAL1 loss simultaneously down-regulates PD-L1 and induces IRF3 activation.

Figure 2. Analysis of PD-L1 expression and cancer-intrinsic innate immunity upon SMARCAL1 deficiency.

(A), Immunoblot showing SMARCAL1, PD-L1, and vinculin levels in MDA-MB-436 cells treated with the indicated siRNAs, and MDA-MB-436 control cells or SMARCAL1-KO clones. S.e., short exposure; l.e., long exposure.

(B), Immunoblot showing SMARCAL1, phosphorylated IRF3 (pIRF3 S386), total IRF3, and tubulin or vinculin levels in the cell lines described in A.

(C), Pathway activity within the indicated Hallmark gene sets (MSigDB) exhibited by SMARCAL1-depleted cells (Query 1) and SMARCAL1-KO clones (Query 2–4) relative to control MDA-MB-436 cells, as determined by VIPER.

(D), Normalized enrichment scores (NES) for Hallmark gene sets (MSigDB) in SMARCAL1-KO clone #1 compared to control MDA-MB-436 cells.

(E), Gene set enrichment analysis for SMARCAL1-KO clone #1 vs control MDA-MB-436 cells for the indicated Hallmark gene sets (MSigDB).

(F), RT-qPCR analysis of mRNA levels of IFNA, IFNB1 or selected ISGs and pro-inflammatory cytokines in control and SMARCAL1-KO #1 MDA-MB-436 cells. Data represent the fold change of gene expression in SMARCAL1-KO #1 vs control cells. Columns represent the mean ± SEM of independent biological replicates (dots). P-values were calculated by multiple unpaired t test.

(G), Cytokine antibody arrays showing cytokines expressed in control and SMARCAL1-KO #1 MDA-MB-436 cells. Reference and background signals are shown.

(H), Cytokines levels normalized over reference signals in control and SMARCAL1-KO #1 MDA-MB-436 cells from the experiment in G.

(I), Immunoblot showing SMARCAL1, phosphorylated STAT1 (pSTAT1 Y701), total STAT1, and tubulin levels in control and SMARCAL1-KO #1 MDA-MB-436 cells.

(J), Volcano plot of RNA-seq analysis comparing SMARCAL1-KO #1 to control MDA-MB-436 cells (FDR <0.05). Red dots indicate genes with significantly increased expression in SMARCAL1-KO #1 vs control cells from the Hallmark gene sets in D. The CD274 (PD-L1) gene is indicated in blue.

(K), RT-qPCR analysis of PD-L1 mRNA levels in MDA-MB-436 cells treated with the indicated siRNAs, and MDA-MB-436 control cells and SMARCAL1-KO clones. Expression data were analyzed and represented as in F. P-values were calculated by one-way ANOVA.

To evaluate the functional consequences of SMARCAL1 loss, the mRNA expression patterns of SMARCAL1-depleted and SMARCAL1-KO cells were determined by PLATE-seq⁴⁰ (Figure S2B–C) and then converted into protein activity profiles using VIPER⁴¹. Enrichment analysis for Hallmark gene sets⁴² revealed downregulation of cell cycle control pathways and upregulation of IFN-dependent and cell-intrinsic immune signaling pathways in SMARCAL1-deficient cells (Figure 2C). Upregulation of these pathways in SMARCAL1-KO cells was confirmed by RNA-seq (Figure 2D–E and Table S2) and RT-PCR (RT-qPCR), including the elevated expression of type I IFNs, ISGs (e.g., SAMD9, ISG15) and NF-κB-responsive genes (e.g., IL6) (Figure 2F). Moreover, SMARCAL1-deficient MDA-MB-436 cells exhibited higher cytokine protein levels (Figure 2G–H) and enhanced STAT1 phosphorylation on tyrosine 701 (Y701) (Figure 2I), a marker of IFN activation. Increased expression of IFNs and ISGs was also observed in SMARCAL1-mutant fibroblasts from a SIOD patient (SD31)³¹ and in Smarcal1-deficient mouse embryonic fibroblasts (MEFs) (Figure S2D–G). Thus, SMARCAL1 deficiency induces cell-intrinsic innate immune signaling in both human and mouse cells.

Gene expression analysis by RNA-seq and RT-qPCR showed that the PD-L1 downregulation observed in SMARCAL1-deficient MDA-MB-436 cells resulted from reduced transcription of the CD274 (PD-L1) gene (Figure 2J–K). PD-L1 downregulation was also observed in SD31 cells (Figure S2J), as well as various SMARCAL1-depleted human breast cancer (SK-BR-3, BT-20, MDA-MB-231), prostate cancer (PC3), ovarian cancer (PEO1) and osteosarcoma (U2OS) cell lines (Figure S2H–I). Importantly, SMARCAL1 depletion had minimal effect on cell cycle progression in those cells (Figure S2K). Reduction of PD-L1 in SMARCAL1-deficient cells was also noted after treatments with known inducers of PD-L1 expression, including IFN-β, IFN-γ and epidermal growth factor (EGF)⁴³ (Figure S2L). Thus, SMARCAL1-dependent PD-L1 downregulation occurs across multiple human cell types in an IFN-independent manner.

Genomic instability activates cGAS-STING-dependent cell-intrinsic innate immunity in SMARCAL1-deficient cells

Given the role of SMARCAL1 in preserving genome integrity^30,31,44,45, we hypothesized that the upregulation of immune signaling in SMARCAL1-deficient cells may result from genomic instability. As expected^44–46, SMARCAL1-KO cells displayed increased levels of spontaneous DNA breaks, γH2AX foci, chromatin bridges and micronuclei (Figures 3A and S3A–C). Moreover, a higher proportion of SMARCAL1-KO cells harbored micronuclei positive for the cytoplasmic DNA sensor cGAS (Figures 3A and S3D), consistent with evidence that micronuclear DNA is sensed by cGAS^47,48. Loss of SMARCAL1 also enhanced the intracellular levels of 2’3’-cGAMP (cGAMP), a cyclic dinucleotide synthesized by cGAS that activates STING and IRF3-mediated transcription^49–54 (Figures 3B and S3E). Consistently, the increased IRF3 S386 phosphorylation observed in SMARCAL1-deficient cells (Figure 2B) was abolished by cGAS depletion (Figure 3C). Additionally, either cGAS or STING depletion reduced the expression of IFNB1 and other inflammatory genes (ISG15, IL6, and CXCL10) in SMARCAL1-deficient cells (Figure 3D). Micronuclei formation and ISG expression (IL6 and CXCL10) were also reduced upon treatment of SMARCAL1-KO cells with RO-3306 (Figure S3F–H), an inhibitor of the cell cycle kinase CDK1, in line with the known requirement of cell cycle progression for micronuclei formation, and induction of inflammatory genes⁴⁷. Collectively, these results indicate that SMARCAL1 loss causes cGAS-STING-dependent upregulation of cell-intrinsic immune signaling.

Figure 3. Analysis of cell-intrinsic immune signaling in SMARCAL1-deficient cells.

(A), Images of DAPI and cGAS staining in MDA-MB-436 control cells and SMARCAL1-KO clones (left). The percentage of cells with one or more micronuclei (middle) and with cGAS-positive micronuclei (right) is shown. Columns represent mean ± SEM of independent biological replicates (dots). P-values were determined by one-way ANOVA.

(B), Relative intracellular cGAMP levels in MDA-MB-436 control cells and SMARCAL1-KO clones. Graphical representation and statistical analysis were conducted as in A.

(C), Immunoblot showing SMARCAL1, cGAS, phosphorylated IRF3 (pIRF3 S386), total IRF3, and tubulin levels in MDA-MB-436 cells, control and SMARCAL1-KO #1 cells treated with the indicated siRNAs.

(D), RT-qPCR analysis of mRNA levels of selected ISGs and pro-inflammatory cytokines in control and SMARCAL1-KO #1 MDA-MB-436 cells treated with the indicated siRNAs. Data represent fold change of gene expression relative to control siRNA-treated cells. Graphical representation was conducted as in A. Statistical analysis was performed by multiple unpaired t test.

(E), Schematic of SMARCAL1 and its mutants used in this study.

(F), Percentage of cells with one or more micronuclei (left) or with cGAS-positive micronuclei (right) in SMARCAL1-deficient MDA-MB-436 cells reconstituted with SMARCAL1 WT or the indicated mutants. Graphical representation and statistical analysis were conducted as in A.

(G), Intracellular cGAMP levels in SMARCAL1-deficient MDA-MB-436 cells reconstituted with the indicated SMARCAL1 mutants relative to WT. Graphical representation and statistical analysis were conducted as in A.

To define the SMARCAL1 activities required to limit cGAS induction, SMARCAL1-deficient MDA-MB-436 cells were reconstituted with either wild-type (WT) or ATPase-defective (R764Q) forms of SMARCAL1³¹, or SMARCAL1 mutants lacking the N-terminal 115 amino acids (ΔN_1–115), which are necessary and sufficient for RPA binding³⁰, or the HARP domains (Δ220–398, ΔHARP), which mediate DNA binding^44,55 (Figures 3E and S3I). Interestingly, SMARCAL1-deficient cells reconstituted with the R764Q or ΔHARP, but not the ΔN_1–115, SMARCAL1 mutant displayed increased micronuclei, with and without cGAS, and enhanced 2’3’-cGAMP levels relative to cells expressing WT-SMARCAL1 (Figure 3F–G). Thus, the ATPase and HARP domains of SMARCAL1 are necessary to restrict cGAS-dependent signaling during unperturbed cell growth.

SMARCAL1 inactivation impairs PD-L1 expression through loss of chromatin accessibility

To understand how SMARCAL1 controls PD-L1, we first monitored PD-L1 levels following DNA damage treatment. As expected^{13,15–17,24,25}, DNA damage increased PD-L1 levels in control cells, although to a lesser extent in SMARCAL1-depleted cells (Figure S2M–O), indicating that SMARCAL1 loss downregulates PD-L1 independently of the response to DNA damaging agents.

Given SMARCAL1’s role in remodeling stalled replication forks^{30,31,33,44,56,57}, we examined whether other SMARCAL1-related fork remodelers, such as ZRANB3 and HLTF, also promote PD-L1 expression^{33,56,58–64}. Interestingly, unlike SMARCAL1 loss, depletion of either ZRANB3 or HLTF did not downregulate PD-L1 (Figure S2P). To then determine whether PD-L1 regulation by SMARCAL1 requires cell proliferation, SMARCAL1-proficient and -deficient RPE1 cells (RPE1-hTERT TP53^−/−) and U2OS-Fucci cells⁶⁵ were arrested in G0/G1 by serum-starvation (Figures 4A and S4A–B) and PD-L1 levels were monitored by immunoblotting or RT-qPCR. Consistent with previous findings⁶⁶, non-replicating cells displayed lower PD-L1 levels than replicating cells. Nonetheless, SMARCAL1 depletion further reduced PD-L1 protein and mRNA levels in serum-starved cells (Figures 4B–C and S4C), demonstrating that SMARCAL1 regulates PD-L1 in both cycling and non-cycling cells.

Figure 4. Regulation of PD-L1 expression by SMARCAL1.

(A), Schematic of the assay to evaluate PD-L1 levels in replicating and non-replicating RPE1-hTERT TP53^−/− or U2OS-Fucci cells treated with the indicated siRNAs.

(B), Immunoblot showing SMARCAL1, PD-L1, and vinculin levels at day 7 in RPE1-hTERT TP53^−/− cells cultured in either growth or starvation medium and treated with the indicated siRNAs at day 4, as in A.

(C), RT-qPCR analysis of PD-L1 mRNA levels in RPE1-hTERT TP53^−/− cells treated with the indicated siRNAs and cultured as in B. Data represent the fold change of PD-L1 expression relative to control cells (RPE1-hTERT TP53^−/− cells treated with control siRNA and cultured in growth medium). Columns represent the mean ± SEM of technical replicates (dots) for two independent biological replicates. P-values were calculated by one-way ANOVA.

(D), Heatmaps of IgG and SMARCAL1 signals aligned to H3K4me3 peaks identified by CUT&RUN in MDA-MB-436 cells. RPKM, reads per kilobase per million mapped reads.

(E), Heatmaps of IgG, SMARCAL1, H3K4me3 and H3K27me3 CUT&RUN signals aligned within −2 kb of the transcription start site (TSS) and +2 kb of the transcription end site (TES) of 23,245 protein-coding genes (normalized for gene length) and ranked by H3K4me3 signal in MDA-MB-436 cells.

(F), Number of common and unique ATAC-seq peaks between control and SMARCAL1-KO #1–2 MDA-MB-436 cells. Unique and common peaks represent peaks present in at least one biological replicate of a single dataset, or all datasets, respectively.

(G), Normalized ATAC-seq tracks (counts per million) for the CD274 (PD-L1) locus in control and SMARCAL1-KO #1 MDA-MB-436 cells. Each track represents the merge of three independent biological replicates. Identified peaks (P1, 5450318–5450905; P2, 5455316–5455499; P3, 5459366–5459591) are highlighted. The H3K27ac ChIP-seq track for the PD-L1 locus in MDA-MB-436 cells was obtained from a publicly available dataset (GSE85158).

(H), Schematic of CRISPRi assays targeting the P3 peak with dCas9-KRAB-MeCP2 (top). RT-qPCR analysis of PD-L1 and SMARCAL1 mRNA levels in MDA-MB-436 cells treated with the indicated siRNAs, and transfected with dCas9-KRAB-MeCP2 and the indicated sgRNAs (bottom). Data represent the fold change of gene expression in the indicated conditions relative to MDA-MB-436 cells treated with control siRNA and transfected with dCas9-KRAB-MeCP2 and a non-targeting sgRNA. Columns represent the mean ± SEM of independent biological replicates (dots). P-values were calculated by multiple unpaired t test.

(I), Immunoblot showing SMARCAL1, PD-L1, and tubulin levels in MDA-MB-436 cells expressing dCas9-KRAB-MeCP2 (MDA-MB-436-CRISPRi), the indicated sgRNAs and treated with the indicated siRNAs.

(J), ChIP-qPCR analysis of SMARCAL1-Flag and RNAPII (RNA polymerase II) occupancy in the genomic region of the P1 and P3 peaks or in the GAPDH promoter as control. Data are representative of three independent experiments and show the mean ± SEM of technical replicates. P-values were calculated by multiple unpaired t test.

(K), Immunoblot showing SMARCAL1, PD-L1, and tubulin levels in MDA-MB-436 cells treated with SMARCAL1 siRNA and complemented with either WT or mutant SMARCAL1 cDNAs or a vector control.

In light of these observations, we then sought to investigate the transcriptional functions of SMARCAL1^35–37,67. To this end, we examined SMARCAL1’s chromatin occupancy using CUT&RUN⁶⁸ (Figure S4D–E). These studies showed that SMARCAL1 largely occupies chromatin regions enriched for histone H3 lysine 4 tri-methylation (H3K4me3) (Figures 4D and S4F, Table S2), a mark of transcriptionally-active chromatin⁶⁹. Moreover, the SMARCAL1 signal overlapped with H3K4me3 around the transcription start sites (TSS) of protein-coding genes, including PD-L1 (Figures 4E and S4G), implying that SMARCAL1 controls PD-L1 levels by direct transcriptional regulation.

SMARCAL1 belongs to the SNF2 family of chromatin remodelers and DNA translocases⁷⁰. Therefore, we investigated whether it possesses chromatin remodeling activity in an in vitro nucleosome mobilization assay with and without SMARCA4, a catalytic component of SWI/SNF chromatin remodeling complexes known to bind SMARCAL1⁷¹. Unlike SMARCA4, SMARCAL1 did not exhibit in vitro nucleosome mobilization activity (Figure S4I), nor did it enhance nucleosome remodeling by SMARCA4 (Figure S4J). Moreover, its ATPase activity was stimulated by free dsDNA to a greater extent than nucleosomal dsDNA (Figure S4H). These observations suggest that SMARCAL1 does not directly mediate chromatin remodeling.

To determine whether SMARCAL1 influences chromatin accessibility in human cells, we conducted ATAC-seq⁷² studies in control and SMARCAL1-KO MDA-MB-436 cells. This work revealed a global loss of accessible chromatin regions in SMARCAL1-deficient cells (Figures 4F and S4K; Table S2). Analysis of the PD-L1 locus showed three distinct accessible peaks in SMARCAL1-proficient cells: one located near the PD-L1 promoter (P1) and two located ~4.8 kb (P2) and ~9 kb (P3) downstream of the PD-L1 TSS (Figure 4G). Each of these ATAC peaks overlapped with H3K27ac-rich regions, which mark active enhancers and promoters^73,74 (Figure 4G; Table S2). Notably, the P3 peak, which corresponds to a cis-regulatory element that promotes PD-L1 expression⁷⁵, was present in SMARCAL1-proficient cells, but not in two distinct SMARCAL1-KO clones (Figure 4G; Table S2). Repression of the P3 element by CRISPR interference (CRISPRi) using the dCas9-KRAB-MeCP2 fusion⁷⁶ (Figure 4H) decreased both PD-L1 mRNA and protein levels in control cells without altering PD-L1 expression in SMARCAL1-depleted cells (Figure 4H–I), suggesting that P3 might be inaccessible in SMARCAL1-deficient MDA-MB-436 cells. Interestingly, analysis of ATAC-seq and H3K27ac ChIP-seq datasets^74,77,78 showed that MCF10A cells, which do not downregulate PD-L1 levels upon SMARCAL1 loss (Figure S4L), do not display an open P3 region, further linking SMARCAL1-dependent PD-L1 regulation to P3 accessibility (Figure S4M). Chromatin immunoprecipitation studies in SMARCAL1-KO MDA-MB-436 cells reconstituted with Flag-tagged SMARCAL1 revealed the presence of both SMARCAL1 and RNA polymerase II at the P1 and P3 sites (Figure 4J), suggesting a direct role for SMARCAL1 in promoting PD-L1 expression. Notably, PD-L1 levels were restored in SMARCAL1-deficient MDA-MB-436 cells upon reconstitution with WT and ΔN_1–115, but not R764Q or ΔHARP, SMARCAL1 cDNAs, highlighting the role of SMARCAL1’s ATPase and HARP domains in PD-L1 regulation (Figure 4K). Collectively, these findings suggest that SMARCAL1 could promote PD-L1 expression by preserving chromatin accessibility of the P3 element through its ATPase activity.

SMARCAL1 cooperates with JUN to stimulate PD-L1 expression

ChromVAR analysis⁷⁹ and TOBIAS framework⁸⁰ on our ATAC-seq data identified 142 TFs (Figures S5A–C; Table S2 and S3), whose binding to chromatin was potentially affected by SMARCAL1 loss. To evaluate their ability to promote PD-L1 expression in a SMARCAL1-dependent manner, we conducted a FACS-based CRISPR-Cas9 screen in SMARCAL1-KO #1 and control MDA-MB-436 cells using a lentiviral sgRNA library targeting the 142 TFs (library 2). Transduced cells were collected as a pool (unsorted population) or sorted based on low PD-L1 staining (PD-L1^low population), and the sgRNA composition of these populations was then determined by sequencing (Figures 5A and S5D–E). These studies identified fifteen TFs whose sgRNAs were enriched in the PD-L1^low population of SMARCAL1-proficient but not -deficient cells, suggesting their potential involvement in SMARCAL1-dependent PD-L1 regulation (Figures 5B and S5F–G; Table S3). Validation by RT-qPCR confirmed that loss of ETV3, ELK4, SP4, MIXL1 or the AP-1 family members JUN and FOSL1 significantly decreased PD-L1 mRNA levels in control but not SMARCAL1-deficient cells (Figure 5C and Table S3). Collectively, these findings suggest that SMARCAL1 may cooperate with these TFs to induce PD-L1 expression in cancer cells.

Figure 5. Identification of transcription factors that cooperate with SMARCAL1 in promoting PD-L1 expression.

(A), Schematic of CRISPR screens to identify transcription factors (TFs) that regulate SMARCAL1-dependent PD-L1 expression. Following selection of control or SMARCAL1-KO #1 MDA-MB-436 cells transduced with the sgRNA library 2, cells were collected as a pool (unsorted) or sorted as shown. sgRNA abundance in distinct cell populations was determined by next-generation sequencing.

(B), Distribution of RRA scores for genes targeted by sgRNA library 2 ranked according to their positive selection score from the MAGeCK RRA output in the indicated comparisons. Pink dots represent genes associated with enriched sgRNAs (positive log fold change and p-value <0.05).

(C), Heatmap showing the mean of PD-L1 mRNA levels from three biological replicates measured by RT-qPCR in control and SMARCAL1-KO #1 MDA-MB-436 cells transduced with the indicated sgRNAs. Clustering analysis was performed using Euclidian distance and Ward’s method. See Table S3.

(D), Schematic of the BioID assay to identify SMARCAL1 interactors. Biotinylated proteins from MDA-MB-436 cells expressing BirA* alone or fused to WT or mutant SMARCAL1 proteins were captured by streptavidin pulldown and subjected to mass spectrometry (left). Numbers of proteins identified by mass spectrometry that are enriched in the indicated pulldowns relative to the BirA* control are shown (right). See Table S4.

(E), Normalized enrichment of the intensity of JUN peptides obtained by mass spectrometry upon pulldown of biotinylated proteins from MDA-MB-436 cells expressing the indicated BirA* fusions. Data are presented as fold change in the intensity of JUN peptides from the indicated pulldowns relative to a BirA* control.

(F), Number of PLA spots in MDA-MB-436 cells obtained using anti-SMARCAL1, anti-RPA and anti-JUN antibodies, either in combination or individually. Data were obtained from three independent biological replicates, with each dot representing PLA spots per cell. Median values are indicated.

(G), P-value distribution for all motifs identified by de novo discovery motif analysis, using SMARCAL1-bound chromatin regions as input peaks for the Homer algorithm. The top three motifs are shown (#1, NFY-like; #2, G-rich, #3 AP-1/JUN-like).

(H), Heatmaps of IgG and SMARCAL1 CUT&RUN signals aligned within −2 kb of the transcription start site (TSS) and +2 kb of the transcription end site (TES) of protein-coding genes in MDA-MB-436 cells, with or without JUN depletion (top heatmap panel). Heatmaps of IgG and SMARCAL1 signals aligned to H3K4me3 and JUN peaks in MDA-MB-436 cells, depleted or not of JUN, are also shown (bottom two panels).

(I), RT-qPCR analysis of PD-L1 mRNA levels in MDA-MB-436 cells expressing dCas9-KRAB-MeCP2 (MDA-MB-436-CRISPRi) and the indicated sgRNAs. Data represent the fold change of gene expression in the indicated conditions relative to MDA-MB-436 cells expressing non-targeting and AAVS1 sgRNAs. Columns represent the mean ± SEM of independent biological replicates (dots). P-values were calculated by one-way ANOVA.

(J), Immunoblot showing JUN, PD-L1, and tubulin levels in MDA-MB-436-CRISPRi cells expressing the indicated sgRNAs.

To investigate whether any of the above TFs interact with SMARCAL1, we performed proximity-dependent biotin labeling (BioID) experiments⁸¹ coupled to mass spectrometry in MDA-MB-436 cells expressing the biotin ligase BirA* fused to WT or mutant (ΔN_1–115 or ΔHARP) SMARCAL1 (Figure 5D). Enrichment analysis of BirA*-SMARCAL1-WT interactors for Reactome gene sets identified pathways associated with DNA repair and gene expression (Table S4). Notably, BirA*-SMARCAL1-WT captured an interaction with JUN (Table S4) that was lost in cells expressing the ΔHARP, but not the ΔN_1–115, mutant, indicating that this interaction relies on the HARP domains, but not the RPA-binding motif of SMARCAL1 (Figure 5E and Table S4). SMARCAL1’s proximity to JUN was confirmed by Proximity Ligation Assay (PLA) (Figures 5F and S5H) and CUT&RUN experiments, which showed an overlap of approximately half of the SMARCAL1-bound sites with JUN-bound regions in MDA-MB-436 cells (Figure S5I–J and Table S2). Accordingly, de novo motif discovery analysis identified the AP-1 JUN-like motif among the top three motifs associated with SMARCAL1-bound regions (Figure 5G).

AP-1 family members, including JUN, JUNB and JUND, can regulate gene transcription by recruiting SWI/SNF complexes that enhance chromatin accessibility^82–84. Therefore, we conducted CUT&RUN experiments to monitor the impact of JUN depletion on SMARCAL1’s chromatin occupancy. Notably, a significant decrease in SMARCAL1 signal was observed at H3K4me3- and JUN-bound genomic regions of JUN-depleted MDA-MB-436 cells (Figure 5H and S5L), especially around gene TSS (Figures 5H and Table S2), implicating JUN in localizing and/or stabilizing SMARCAL1 in the vicinity of transcriptionally-active regions. Of note, SMARCAL1 and JUN protein levels were unaffected by JUN and SMARCAL1 depletion, respectively (Figure S5K), suggesting that SMARCAL1 and JUN do not regulate each other’s protein abundance.

Given the importance of SMARCAL1 for regulating P3 accessibility (Figure 4G–I), we examined whether the P3 element is also controlled by JUN. Interestingly, similar to SMARCAL1 loss (Figure 4H–I), JUN depletion reduced PD-L1 levels in control dCas9-KRAB-MeCP2-expressing MDA-MB-436 cells, while it had a minimal impact on PD-L1 levels when dCas9-KRAB-MeCP2 was targeted to the P3 region (Figure 5I–J). These findings confirm that SMARCAL1 and JUN cooperatively regulate PD-L1 expression by preserving P3 accessibility.

SMARCAL1 loss enhances T cell-mediated anti-tumor immunity in a cGAS- and PD-L1dependent manner

To determine the biological consequences of SMARCAL1 deficiency on anti-tumor immune responses, we disrupted Smarcal1 in B16/F10 mouse melanoma cells (Smarcal1-KO), and then reconstituted them with either murine Smarcal1 cDNA (Smarcal1-KO + Smarcal1 cDNA) or an empty vector (Smarcal1-KO + EV) (Figure S6A). Smarcal1 re-expression in Smarcal1-KO cells diminished the fraction of γH2AX foci-positive cells and the percentage of micronuclei-positive cells without affecting cell cycle progression (Figures S6B–D), while concurrently decreasing the expression of Ifna1, Ifnb1 and ISGs (Figure S6E) and the intracellular levels of cGAMP (Figure S6F). Conversely, Smarcal1 did not regulate PD-L1 levels in B16/F10 mouse cells (Figure S6G–H), which lack an accessible P3 region (Figure S6I). Likewise, loss of Smarcal1 or Jun, either individually or in combination, did not alter PD-L1 levels in the murine colorectal and renal carcinoma cells MC38 and Renca (Figure S6J–K). Collectively, these findings suggest that Smarcal1 loss results in activation of cell-intrinsic immune signaling in murine cells without affecting PD-L1 levels.

To examine whether SMARCAL1 influences anti-tumor immunity, we inoculated C57BL/6 mice subcutaneously with B16/F10 Smarcal1-KO cells with or without Smarcal1 reconstitution. Remarkably, Smarcal1 deficiency significantly suppressed tumor growth, resulting in reduced tumor size and increased animal survival (Figure 6A–B). Importantly, this effect was due to enhanced anti-tumor immunity rather than impaired cell growth, since Smarcal1-deficient and -reconstituted B16/F10 cells grew with equal efficiency in vitro (Figure S7A) and upon inoculation into immunodeficient NOD.Cg-Prkdc^scid Il2rg^tm1Wjl/SzJ (NSG) mice (Figure S7B–C). Notably, disruption of Cgas in Smarcal1-deficient B16/F10 cells (Figure S7D) increased the growth of Smarcal1-KO tumors and decreased animal survival (Figure 6C–D), demonstrating that Smarcal1 deficiency elicits a Cgas-dependent anti-tumor immune response. Flow cytometry revealed a significant increase in intra-tumoral immune cells (CD45⁺), including CD8⁺ cytotoxic T cells and CD4⁺ T cells, in Smarcal1-deficient tumors (Figures 6E and S7E). Remarkably, depletion of CD8⁺ T cells fully rescued the growth of Smarcal1-deficient tumors (Figures 6F–G and S7F–G), indicating that CD8⁺ T cells are essential effectors of the observed anti-tumor immune response. These findings show that Smarcal1 inactivation leads to the induction of tumor-intrinsic cytosolic DNA sensing, thereby enhancing tumor immunogenicity and cytotoxic T cell-mediated tumor rejection.

Figure 6. Analyses of anti-tumor immune responses against SMARCAL1-deficient cancer cells.

(A), Analysis of tumor volume in C57BL/6 mice subcutaneously injected with B16/F10 Smarcal1-KO cells reconstituted with Smarcal1 cDNA or empty vector (EV). Graphs are representative of three independent experiments and show the mean ± SEM of individual mice per group. P-value was calculated by unpaired t test.

(B), Survival analysis conducted on the C57BL/6 mice described in A. P-value was calculated by log-rank test.

(C), Analysis of tumor volume in C57BL/6 mice subcutaneously injected with B16/F10 Smarcal1-KO cells reconstituted as in A with or without Cgas disruption. Graphs represents two independent experiments and show the mean ± SEM of individual mice per group. P-values were calculated as in A.

(D), Survival analysis conducted on the C57BL/6 mice described in C. Experiments were conducted and analyzed as in B.

(E), Number of tumor infiltrating lymphocytes in C57BL/6 mice injected with the B16/F10 Smarcal1-KO cells described in A. Each dot represents the number of infiltrating lymphocytes per gram in a single tumor. P-values were calculated by unpaired t test.

(F), Analysis of tumor volume in C57BL/6 mice treated with isotype control or anti-mouse CD8α antibody before injection of the B16/F10 Smarcal1-KO cells described in A. Graphical representation and statistical analysis were performed as in A. See Figure S7F–G.

(G), Survival analysis conducted on the C57BL/6 mice described in F. Experiments and statistical analysis were conducted as in B.

(H), Representative images from (NLS)-dsRed-MDA-MB-436 control and SMARCAL1-KO #1 cells transduced with PD-L1 cDNA (pHAGE_PD-L1) or empty vector (pHAGE_EV) and incubated for forty-eight hours (T2) with or without human CD8⁺ T cells. See Figure S8D.

(I), Quantification of the growth of the tumor cells described in H at T2 post-incubation with activated CD8⁺ T cells. Bar graphs represent the proliferation rate of (NLS)-dsRed-MDA-MB-436 control and SMARCAL1-KO #1 cells transduced with the indicated constructs in the presence of CD8⁺ T cells relative to their proliferation rate in the absence of CD8⁺ T cells. Columns represent the mean ± SD of three replicates. P-values were calculated by two-way ANOVA.

Given the lack of Smarcal1-mediated PD-L1 regulation in mouse B16/F10 tumors, the above experiments did not allow us to evaluate the contribution of SMARCAL1 deficiency to anti-tumor immunity via PD-L1 downregulation. To monitor the anti-tumor effects induced by Smarcal1 deficiency in B16/F10 tumors when combined with PD-L1 impairment, as observed in SMARCAL1-deficient human cancer cells (Figures 2A, S2A and S2H–I), C57BL/6 mice inoculated with Smarcal1-deficient or -reconstituted B16/F10 cells were treated with an antibody inhibiting murine PD-L1 (Figure S8A). Remarkably, PD-L1 blockade markedly reduced the growth of Smarcal1-deficient, but not Smarcal1-reconstituted, B16/F10 tumors and extended animal survival (Figures S8A–C). To test more directly whether the PD-L1 downregulation induced by SMARCAL1 loss contributes to anti-tumor immune responses, we co-cultured SMARCAL1-deficient MDA-MB-436 cells with activated human CD8⁺ T cells derived from a healthy donor with HLA class I and class II alleles matched to MDA-MB-436 tumor cells (Figure S8D). These studies revealed that SMARCAL1 loss enhanced the sensitivity of MDA-MB-436 cells to CD8⁺ T cell-mediated killing (Figures 6H–I and S8E). Notably, PD-L1 overexpression partially rescued the sensitivity of SMARCAL1-deficient cells, but not control cells, to activated CD8⁺ T cells (Figures 6I and S8E), suggesting that the PD-L1 downregulation induced by SMARCAL1 loss contributes to anti-tumor immunity.

Having established that Smarcal1 deficiency sensitizes cancer cells to treatment with anti-PD-L1 antibodies (Figures S8A–C), we next asked whether it might enhance the efficacy of other ICB therapies. Indeed, treatment of Smarcal1-deficient B16/F10 tumors with anti-CTLA-4 antibodies significantly reduced tumor growth and increased animal survival (Figures S8F–H). These effects were further enhanced by combined treatment of Smarcal1-deficient tumors with anti-PD-L1 and anti-CTLA-4 antibodies (Figures 7A–C). These results demonstrate that Smarcal1 deficiency synergizes with ICB therapy in reducing tumor growth.

Figure 7. Effects of immune checkpoint blockade in mice carrying Smarcal1-deficient tumors and analysis of the clinical outcome of cancer patients with low SMARCAL1 expression.

(A), Schematic of ICB treatments in mice injected with B16/F10 Smarcal1-KO cells reconstituted with Smarcal1 cDNA or empty vector (EV).

(B), Analysis of tumor volume in C57BL/6 mice subcutaneously injected with B16/F10 Smarcal1-KO cells reconstituted as in A, and subsequently treated with isotype control or anti-mouse CTLA-4 plus anti-mouse PD-L1 antibodies. Data represent the mean ± SEM of individual mice per group. P-values were calculated by multiple unpaired t test.

(C), Survival analysis conducted on the C57BL/6 mice described in B. P-values were calculated by log-rank test.

(D), Analysis of SMARCAL1 expression in tumors and their respective normal tissues from TCGA patients with the indicated cancer types. P-values were calculated by unequal variance t test. CPM, counts per million.

(E), Tumor-context specific differential analysis of SMARCAL1^Low vs SMARCAL1^High cancer patients for the indicated Hallmark gene sets (MSigDB), PD-L1 expression and leukocyte score ratio.

(F), Pearson correlation between the normalized enrichment score (NES) of inflammatory response pathway and the leukocyte score ratio for the indicated TCGA tumor types.

(G), SMARCAL1 levels prior to treatment with anti-PD-1 checkpoint inhibitors in non-responder and responder groups of patients from the referenced datasets. P-values were calculated by unequal variance t test. FPKM, fragments per kilobase of transcript per million mapped reads.

(H), Proposed model for the dual role of SMARCAL1 in regulating the response of tumors to host immunity and PD-L1-mediated immune checkpoint.

SMARCAL1 expression is associated with poor clinical outcome in cancer patient datasets

To determine the translational significance of our findings in human cancer patients, we first examined SMARCAL1 mutations and expression levels in the TCGA dataset. Although SMARCAL1 is rarely mutated, amplified, or deleted in human cancer (Figure S9A), it is overexpressed in most tumors (Figure 7D). Stratifying patients within each cancer type according to low (SMARCAL1^Low) or high (SMARCAL1^High) SMARCAL1 expression (Figure S9B) revealed a shorter progression-free survival of SMARCAL1^High patients for certain cancer types (Figure S9C). The SMARCAL1^Low tumors showed downregulation of pathways involved in cell proliferation, c-Myc activation, TGF-β signaling, and DNA repair, accompanied by upregulation of pathways related to the inflammatory response (Figure 7E). Moreover, PD-L1 expression was significantly downregulated in 85% of tumor types of the SMARCAL1^Low group (Figure 7E). Notably, the ratio of leukocyte score⁸⁵ between SMARCAL1^Low and SMARCAL1^High groups was greater than 1 in 53% of tumor types (Figure 7E) and correlated positively with the inflammatory response signature observed in the SMARCAL1^Low group (R= 0.68) (Figure 7F), suggesting that SMARCAL1 deficiency induces leukocyte infiltration.

To investigate the clinical relevance of SMARCAL1 deficiency for cancer immunotherapy, we evaluated SMARCAL1 expression in two datasets of cancer patients treated with anti-PD-1 therapy^86,87. Significantly, responders to immunotherapy expressed lower levels of SMARCAL1 relative to non-responders (Figure 7G), in line with our observation that SMARCAL1 deficiency enhances the response to ICB therapy (Figures 7A–C, S8A–C and S8F–H). These findings suggest that inhibiting SMARCAL1 may improve immuno-oncological therapies.

DISCUSSION

In this study, we investigate the links between innate immune signaling and immune checkpoint regulation in cancer cells. Using a CRISPR library targeting DDR and chromatin factors, we recovered known regulators of innate immunity and/or PD-L1 levels (Figure 1), such as CTCF and the cohesins SMC3 and RAD21^28,88. While a positive correlation between cell-intrinsic innate immunity and PD-L1 expression has been documented⁴³, our screens unexpectedly identified genes whose loss results in both innate immune stimulation and PD-L1 downregulation (Figure 1 and Table S1). Analysis of cancer patient datasets revealed that lower expression of these genes often correlates with enhanced inflammatory response and leukocyte infiltration, reduced PD-L1 expression, increased progression-free survival, and/or elevated response to immunotherapy (Figure S9D–F). Recognizing the potential benefits of combined innate immune stimulation and PD-L1 downregulation for anti-tumor responses, we sought to characterize the SNF2-family DNA translocase SMARCAL1, a top hit in both our CRISPR screen and patient dataset analyses. These studies demonstrated that SMARCAL1 limits endogenous DNA damage and suppresses cGAS-STING-dependent immune signaling, while simultaneously stimulating PD-L1 expression. As such, SMARCAL1 stands apart from previously studied DDR factors, whose loss/inhibition typically leads to an upregulation, rather than a reduction, of PD-L1 levels⁸⁹.

SMARCAL1 suppresses cGAS-STING activation through the maintenance of genome stability

Activation of the cGAS-STING pathway in cancer cells can occur in response to micronuclei formation or the accumulation of cytoplasmic DNA fragments^{47,48,90–93}. As previously observed⁴⁶, micronuclei are elevated in SMARCAL1-deficient cells (Figure 3A), and inhibition of their formation following cell cycle arrest suppresses the expression of inflammatory genes (Figure S3F–H). Micronuclei in SMARCAL1-deficient cells may arise from compromised remodeling of stalled forks^{30–33,44,57,94,95}, which could lead to aberrant fork cleavage by the endonuclease MUS81, resulting in DSBs⁴⁴ that may contribute to chromosome mis-segregation⁹⁶. SMARCAL1-deficient cells also exhibit breakage of telomeric sequences mediated by the MUS81-SLX4 complex, identifying telomere replication as a source of DNA damage induced by SMARCAL1 loss^45,46,97. Like telomeres, other difficult-to-replicate genomic regions (e.g., fragile sites, repetitive or G-rich DNA sequences) might also require SMARCAL1 for efficient replication, and thus contribute to micronuclei and cGAS-mediated immune signaling in SMARCAL1-deficient cells. Interestingly, micronuclei formation and cGAS activation depend on the ATPase but not the RPA-binding domain of SMARCAL1 (Figures 3F–G and S3I). These findings could reflect the ability of SMARCAL1 to maintain genome stability in an RPA-independent manner during normal cellular growth, as previously noted⁴⁵. Alternatively, they could be a consequence of supraphysiological expression of the RPA-binding mutant.

SMARCAL1 regulates PD-L1 by controlling chromatin accessibility at the PD-L1 locus

Recent studies have shown that genomic instability induces PD-L1 expression in a cGAS-STING-dependent manner^13,17,25. Surprisingly, despite causing genomic instability and activating cGAS-STING, SMARCAL1 loss downregulates PD-L1 (Figures 4, 7H, S2 and S4). Moreover, it does so in both replicating and non-replicating cells, as well as in cells treated with DNA damaging agents (Figures 4A–C, S2O and S4A–C). In contrast, PD-L1 is not downregulated upon loss of the SMARCAL1-related fork remodelers ZRANB3 and HLTF (Figure S2P), indicating that SMARCAL1 possesses distinct functions from ZRANB3 and HLTF in PD-L1 regulation. Collectively, these results suggest that PD-L1 regulation is independent of the DNA replication-associated functions of SMARCAL1.

Instead, we show that SMARCAL1 localizes to transcriptionally-active chromatin, including the TSS of certain protein-coding genes, such as PD-L1 (Figures 4D–E, J and S4G). Furthermore, SMARCAL1 regulates PD-L1 expression by maintaining the accessibility of the cis-regulatory P3 region of the PD-L1 locus⁷⁵ in a manner dependent on its ATPase and HARP domains (Figure 4K). While SMARCAL1 did not exhibit chromatin remodeling activity in an in vitro nucleosome mobilization assay⁹⁸ (Figure S4I), we cannot exclude that it might do so in vivo through accessory proteins and/or post-translational modifications. Alternatively, SMARCAL1 might employ its DNA annealing activity⁹⁹ to remodel DNA structures that influence chromatin accessibility^35–37. Notably, motif analysis of SMARCAL1-bound genomic sites revealed an enrichment for G-rich sequences (Figure 5G). These sequences can form G-quadruplexes, known regulators of chromatin accessibility and gene expression¹⁰⁰. Interestingly, binding sites of TFs, such as the AP-1 complex, are enriched in G-quadruplexes¹⁰¹, raising the possibility that SMARCAL1 operates at those structures to regulate gene expression.

SMARCAL1 cooperates with the AP-1 transcription factor to regulate PD-L1 expression

Our findings suggest that SMARCAL1 may cooperate with certain TFs (JUN, FOSL1, ETV3, SP4, MIXL1 and ELK4) to facilitate PD-L1 expression (Figure 5A–C). Notably, JUN and FOSL1, which are part of the AP-1 complex, were previously identified as positive regulators of PD-L1 in multiple cancer types^102–106 and AP-1 binding sites are found in intragenic PD-L1 regulatory regions, including the P3 element^103,106,107. The cooperation between SMARCAL1 and AP-1 is also supported by both genome-wide protein occupancy analyses and proximity-labeling experiments, and depends on SMARCAL1’s HARP domains (Figure 5). On a functional level, JUN facilitates the localization and/or retention of SMARCAL1 near TSS and at transcriptionally-active regions, including the PD-L1 locus (Figure 5H and Table S2). AlphaFold2¹⁰⁸ (AF2) analyses (Table S4) did not predict a physical interaction between SMARCAL1 and JUN, suggesting that their association is likely indirect. Of the seven high-confidence direct interactors of SMARCAL1 identified by AF2 among proteins detected in BirA*–SMARCAL1 pulldowns (Table S4), the DEAD box protein DDX21 was previously shown to interact with JUN and regulate its activity^109,110. Additionally, two other predicted SMARCAL1 direct interactors, the WD repeat domain protein WDR82 and the translation initiation factor EIF6, were identified by AF2 as potential direct interactors of FOS/FOSL1 and JUN, respectively (Table S4). Thus, DDX21, WDR82 and EIF6 may coordinate the activities of SMARCAL1 and AP-1 to modulate gene expression.

SMARCAL1 loss induces anti-tumor immunity and enhances the response to immune checkpoint blockade

SMARCAL1 is overexpressed in most types of human cancer (Figure 7D), and cancer patients with high SMARCAL1 levels display lower expression of inflammatory response genes, reduced leukocyte infiltration, and increased PD-L1 expression (Figure 7E–F). These observations suggest that SMARCAL1 overexpression may suppress cGAS-mediated tumor immunogenicity by reducing genomic instability, while simultaneously enhancing tumor immune evasion through PD-L1 induction. By examining the growth of Smarcal1-deficient cancer cells in vivo, we observed that Smarcal1 deficiency elicited potent Cgas-mediated anti-tumor activity (Figure 6A–D). Besides SMARCAL1 deficiency, inhibition of other DDR regulators, such as PARP1/2, also promotes cGAS activation in tumors and enhances anti-tumor immunity^16,17. However, in the case of SMARCAL1 deficiency, the observed PD-L1 downregulation also contributes to anti-tumor immunity by potentiating CD8⁺ T cell-mediated tumor cell killing (Figures 6H–I and S8D–E).

Smarcal1 deficiency elicits potent T cell-mediated anti-tumor immunity in poorly immunogenic tumors, such as mouse B16/F10 tumors (Figure 6), which do not respond effectively to PD-(L)1- or CTLA-4-dependent ICB^111,112. Notably, Smarcal1 deficiency synergizes with both anti-PD-L1 and anti-CTLA-4 treatments to reduce the growth of B16/F10 tumors (Figures 7A–C and S8A–C, F–H). Since Smarcal1 is dispensable for PD-L1 expression in mouse B16/F10 cells (Figure S6G–H, K), Smarcal1 loss enhances the response of B16/F10 tumors to ICB therapy primarily by augmenting innate immune signaling and T cell tumor infiltration. In human tumors, SMARCAL1 loss might potentiate the response to ICB therapy also by reducing PD-L1 levels. Indeed, increasing evidence indicates that anti-PD-1 or anti-CTLA-4 immunotherapy can be improved when combined to PD-L1 downregulation^113–117. Collectively, our findings demonstrate that SMARCAL1 loss in cancer cells stimulates anti-tumor immunity through a dual mechanism involving both the induction of cGAS-mediated innate immunity and the suppression of PD-L1-associated tumor immune evasion (Figure 7H), suggesting SMARCAL1 as a promising target for more effective immuno-oncological therapies.

Limitations of the study

As previously reported⁶⁶, PD-L1 levels are regulated in a cell cycle-dependent manner (Figures 4A–C and S4A–C). Further studies are needed to ascertain whether the hits identified in our screen (Figure 1A–B) control PD-L1 levels directly, like SMARCAL1, or whether they operate indirectly by altering cell cycle progression.

SMARCAL1-mediated PD-L1 regulation is not operative in mouse cells (Figure S6G–K). Consequently, pre-clinical studies of the impact of Smarcal1 depletion on anti-tumor immunity conducted in mice cannot fully model its anticipated impact in human patients. Further investigations are needed to assess whether the anti-cancer effects of SMARCAL1 disruption observed in mice are more pronounced in the context of human tumors, where induction of innate immune signaling and PD-L1 downregulation can synergistically contribute to enhance anti-tumor immune responses.

SMARCAL1 loss-of-function mutations in SIOD patients cause multisystemic phenotypes, including T cell deficiency and lymphopenia²⁹. While our study focuses on the consequences of SMARCAL1 deficiency in tumor cells, additional work is needed to evaluate the effects of SMARCAL1 inhibition on immune cells, particularly in the context of anti-tumor immune responses.

STAR Methods

RESOURCE AVAILABILITY

Lead Contact

Further information and request for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Alberto Ciccia (ac3685@cumc.columbia.edu).

Materials availability

Plasmids and cell lines generated in this study will be made available by the Lead Contact under a Material Transfer Agreement.

Data and Code availability

Deep sequencing data from this study are publicly available in the Gene Expression Omnibus database (GEO: GSE245447, GSE245448, GSE245449 and GSE245450). Sequencing data from previously published ATAC-seq and ChIP-seq studies^74,77,78 are publicly available in the GEO database (GEO: GSE89013, GSE114964 and GSE85158). The mass spectrometry data from this study have been deposited to the ProteomeXchange Consortium and can be accessed publicly through the PRIDE¹¹⁹ partner repository (PRIDE: PXD046384). Unprocessed blots and microscopy images are publicly available at Mendeley Data, V1, doi: 10.17632/2fmtcfwfv4.1. Any additional information required to reanalyze the data reported in this work paper is available from the Lead Contact upon request.

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Cell lines and cell culture

MCF7 cells were cultured in DMEM (#11965, Thermo Fisher Scientific) supplemented with 10% (v/v) fetal bovine growth serum (#F6765, Sigma-Aldrich), and 100 U/ml of penicillin, and 100 μg/ml of streptomycin (#15140, Thermo Fisher Scientific). HEK293T, MDA-MB-436, BT-474, MDA-MB-361, SK-BR-3, BT-20, MDA-MB-231, T-47D, PEO1, B16/F10, U2OS, RPE1-hTERT TP53^−/− and U2OS-Fucci cells were grown in DMEM (#11965, Thermo Fisher Scientific) supplemented with 10% (v/v) Fetalgro bovine growth serum (RMBIO), 100 U/ml of penicillin, and 100 μg/ml of streptomycin (#15140, Thermo Fisher Scientific). hTERT-immortalized SD31 fibroblast and mouse embryonic fibroblasts (MEFs) were grown in DMEM supplemented with 15% (v/v) Fetalgro bovine growth serum, 100 U/ml of penicillin, and 100 μg/ml of streptomycin. PC3 cells were grown in RPMI 1640 (#11875, Thermo Fisher Scientific) medium supplemented with 10% (v/v) Fetalgro bovine growth serum, 100 U/ml of penicillin, and 100 μg/ml of streptomycin. Human CD8⁺ T cells from a healthy donor with HLA class I and class II alleles matched to MDA-MB-436 tumor cells were purchased from Cellero (Donor 455, #1041) and grown in ImmunoCult-XF T Cell Expansion medium (#10981, STEMCELL Technologies) supplemented with 10 ng/ml hr-IL2 (#78036, STEMCELL Technologies). In the experiments performed under starvation conditions, U2OS-Fucci cells and RPE1-hTERT TP53^−/− cells were cultured in DMEM with low glucose/pyruvate without glutamine and phenol red supplemented with 0.01% (v/v) Fetalgro (starvation medium). All cells were cultured at 37°C and 5% CO₂. The cell lines used were either obtained from ATCC (https://www.atcc.org/en.aspx) or kindly provided by other laboratories. hTERT-immortalized SD31 fibroblasts reconstituted with SMARCAL1 were described in our previous studies³¹. Smarcal1-proficient and -deficient primary MEFs were kindly provided by Dieter Egli (Columbia University). PC3 cells were kindly provided by Cory Abate-Shen (Columbia University). PEO1 cells were kindly provided by Toshiyasu Taniguchi (Tokai University). B16/F10, MC38 and Renca cells were kindly provided by Charles Drake (Columbia University). RPE1-hTERT TP53^−/− cells were generated as previously described¹²⁰ and kindly provided by Daniel Durocher (University of Toronto). U2OS-Fucci cells were kindly provided by Stephen Elledge (Harvard Medical School).

Mice

Five-six weeks old female mice were purchased from Jackson Laboratories (#005557 and #000664) and maintained under pathogen-free conditions in 14 hr light/10 hr dark cycle with access to food and water. All animal handling and procedures were performed in accordance with a protocol approved by the Institutional Animal Care and Use Committee of Columbia University.

METHOD DETAILS

DNA clones and plasmids

To generate the pXPR206_Hygro lentiviral vector, the puromycin resistance gene downstream of the P2A site in pXPR206 (plasmid #96920, Addgene) was replaced by a modified hygromycin resistance gene with its internal BsmBI restriction site disrupted by a silent mutation. Briefly, the hygromycin resistance gene from pMSCV-Hygro was PCR-amplified with J1178/J1179, which introduced flanking 5’-HindIII and 3’-XbaI restriction sites, and the resulting amplicon cloned into pCR4Blunt-TOPO (#450031, Thermo Fisher Scientific). A silent point mutation was then introduced at the internal BsmBI restriction site within the hygromycin resistance gene by inverse PCR mutagenesis with J1357/J1358. Following HindIII/XbaI digestion, the BsmBI-mutated hygromycin resistance gene-containing fragment was subsequently ligated into the BspEI/XbaI sites of pXPR206, thereby replacing the puromycin resistance gene. sgRNAs targeting the genes shown in Figure 5C were cloned into the pXPR206_Hygro lentiviral vector. The eCas9-P2A-Puro plasmids containing sgRNAs targeting human SMARCAL1 were previously described³³. sgRNAs targeting mouse Smarcal1 were cloned into the pLenti-SpBsmBI plasmid (plasmid #62205, Addgene) or into the plentiCRISPRv2-Hygro (plasmid #98291, Addgene). sgRNA targeting the Rosa26 locus was also cloned into the plentiCRISPRv2-Hygro. sgRNAs targeting mouse Cgas or used in the CRISPRi experiments were cloned into the pSB700-H2B-EYFP-P2A-Puro plasmid kindly provided by Alejandro Chavez (Columbia University) or into LentiGuide-puro-NLS-GFP plasmid (plasmid #185473, Addgene). The dCas9-KRAB-MeCP2 plasmid (plasmid #110821, Addgene) or the dCas9-KRAB-MeCP2 lentiviral plasmid (plasmid #122205, Addgene) were used for CRISPRi-mediated transcriptional repression experiments. sgRNAs targeting human JUN for CRISPRi experiments¹²¹ were cloned into a pSAM110 plasmid, a modified version of the pLenti-SpBsmBI plasmid (plasmid #62205, Addgene). Site-directed mutagenesis by inverse polymerase chain reaction was used to mutate pDONR223-SMARCAL1³³ within SMARCAL1 sequences targeted by SMARCAL1 siRNA or sgRNAs expressed by the eCas9-P2A-Puro plasmids. pDONR223-SMARCAL1-ΔHARP was generated by PCR-amplification of pDONR223-SMARCAL1 using the DNA oligos SL1-ΔHARP-FW and SL1-ΔHARP-RV. The mutated pDONR223-SMARCAL1 was then used in LR clonase II assays (#11791020, Thermo Fisher Scientific) to recombine SMARCAL1 into the lentiviral expression vector pHAGE-Ct-Flag-HA-DEST-Puro, pHAGE-TREX-NtBioID-DEST, or pHAGE-Ct-Flag-HA-DEST-Puro after replacement of the puromycin resistance gene with the hygromycin resistance gene¹²². The pDONR223-CD274 (#13856, ORFeome) was mutated to introduce a STOP codon at the end of the ORF of the CD274 (PD-L1) gene. The sequence of the PD-L1 gene was confirmed by Sanger sequencing prior to LR clonase II-mediated recombination (#11791020, Thermo Fisher Scientific) into the lentiviral expression vector pHAGE-UbC-DEST (pHAGE-UBC-CD274) after replacement of the hygromycin resistance¹²² gene with the neomycin resistance gene. Briefly, the pgk-puro cassette-containing sequence from pHAGE-TREX-DEST-Puro, including upstream ApaI and downstream ClaI restriction sites, was PCR-amplified with J1147/J1148 and the resulting amplicon cloned into pCR4Blunt-TOPO (#450031, Thermo Fisher Scientific). The entire plasmid, excluding the puromycin resistance gene, was then amplified by inverse PCR with J1138/J1139, while the neomycin/G418 resistance gene was amplified from pCDNA 3.1 with J1232/J1233. The two amplicons were then seamlessly fused by Gibson assembly to generate pCR4Blunt-TOPO carrying the newly-formed pgk-neo cassette. Following ApaI/ClaI digestion, the pgk-neo-containing fragment was subsequently ligated into the ApaI/ClaI-sites of pHAGE-UbC-DEST-Hygro, thereby replacing the pgk-hygro cassette. Mouse Smarcal1 cDNA (NM_018817) amplified from MEFs was recombined into the Gateway entry vector pDONR223 with BP clonase II (#1178910, Thermo Fisher Scientific). The identity of the insert was confirmed by Sanger sequencing prior to LR clonase II-mediated recombination (#11791020, Thermo Fisher Scientific) into the retroviral expression vector pMSCV-Nt-HA-FLAG¹²³. The cDNA encoding murine Smarcal1 in the pMSCV expression vector (pMSCV-mSmarcal1) was confirmed by BsrGI digestion. DNA oligo sequences are available in Table S5.

Recombinant viral production and transduction

Recombinant lentiviruses and retroviruses were generated by co-transfecting helper packaging vectors together with lentiviral or retroviral vectors into HEK293T cells using the TransIT-293 transfection reagent (Mirus). Virus-containing supernatants were collected 72 hr after transfection and utilized at low MOI (<0.5) to infect target cells in the presence of 8 μg/ml of polybrene. 48 hr after viral addition, successfully infected cells were selected using puromycin (1 μg/ml), hygromycin (100 μg/ml) or blasticidin (10 μg/ml) for 3–5 days. Experiments with infected cells were performed three or more weeks after selection.

Cell transfection

RNA interference experiments were carried out by reverse transfection of the indicated siRNAs using RNAiMAX reagents (#13778075, Thermo Fisher Scientific) according to the manufacturer’s instructions. For transfections involving DNA plasmids, 500 ng or 1 μg of DNA were combined with lipofectamine 3000 (#L3000015, Thermo Fisher Scientific) according to the manufacturer’s instructions, and used for transfection of adherent cells at 50–60% confluency in 6-multiwell tissue culture plates. For CRISPRi experiments, dCas9-KRAB-MeCP2 and sgRNA-pSB700-H2B-EYFP-P2A-Puro plasmids at 1:1 ratio were transduced into cells 24 hr after transfection of the indicated siRNAs. For the experiments involving SMARCAL1 mutants, 500 ng of DNA was transduced into cells 48 hr before transfection of the indicated siRNAs.

Cell line generation

MDA-MB-436 SMARCAL1-KO, B16/F10 Smarcal1-KO and, B16/F10 Cgas-KO cells were obtained by CRISPR-Cas9 technology. In particular, MDA-MB-436 cells were transfected with SMARCAL1 sgRNA-containing eCas9-P2A-Puro plasmids and briefly selected with puromycin (1 μg/ml). Transfected cells were then re-plated at low density without puromycin selection for colony formation. Colonies were subsequently isolated and screened by western blot for loss of SMARCAL1 expression. MDA-MB-436 cells transfected with eCas9-P2A-Puro plasmids and selected with puromycin (1 μg/ml) were used as control cells. SMARCAL1-KO MDA-MB-436 cells were reconstituted with pHAGE-SMARCAL1-Ct-Flag-HA. Control and SMARCAL1-KO MDA-MB-436 cells expressing nuclear-localization signal (NLS)-dsRed were generated by lentiviral infection at low MOI (<0.5) and sorting for cells with similar levels of the fluorescent protein. PD-L1 overexpression in control and SMARCAL1-KO MDA-MB-436 cells expressing NLS-dsRed was obtained by lentiviral infection at low MOI (<0.5) with pHAGE-UBC-CD274 (pHAGE_PD-L1). To generate B16/F10 Smarcal1-KO cells, B16/F10 cells expressing a doxycycline-inducible SpCas9¹²⁴ were obtained by lentiviral infection at low MOI (<0.5) and blasticidin selection. Doxycycline-inducible SpCas9-expressing cells were further infected with lentiviral sgRNA constructs targeting Smarcal1 and selected with hygromycin (100 μg/ml). Three days after doxycycline-induced SpCas9 expression, cells were sorted as single cells in 96-multiwell plates by flow cytometry (BD Influx Cell Sorter) and cultured in the absence of drug selection and doxycycline. Approximately 180 single clones were subsequently screened by immunofluorescence for Smarcal1 expression using ImageXpress Nano Automated Imaging System microscope (Molecular Devices). Single clones with the lowest Smarcal1 signal were further validated by western blot for loss of Smarcal1 expression. B16/F10 Smarcal1-KO cells were subsequently reconstituted with pMSCV-mSmarcal1 or empty vector by infection at low MOI (<0.5) and puromycin (1 μg/ml) selection. To generate B16/F10 Smarcal1/Cgas-KO, B16/F10 Smarcal1-KO cells were treated with doxycycline to induce SpCas9 expression and transfected with Cgas sgRNA-containing pSB700-H2B-EYFP-P2A-Puro plasmid or an empty vector. Single clones were screened by western blot for loss of Cgas expression. B16/F10 Smarcal1-KO and B16/F10 Smarcal1/Cgas-KO were subsequently reconstituted with pMSCV-mSmarcal1 or an empty vector by infection at low MOI (<0.5) and puromycin (1 μg/ml) selection.

To generate Smarcal1-deficient MC-38, Renca and B16/F10 cells, sgRNAs targeting Smarcal1 were first cloned into the BsmBI/Esp3I sites of plentiCRISPRv2-Hygro. MC-38, Renca and B16/F10 cells were then infected at low MOI (<0.5) with Smarcal1 sgRNA-containing plentiCRISPRv2-Hygro followed by hygromycin selection (100 μg/ml). The MDA-MB-436-CRISPRi cell line was generated by lentiviral infection at low MOI (<0.5) with dCas9-KRAB-MeCP2 and blasticidin selection (1 μg/ml). To generate the MDA-MB-436-CRISPRi-AAVS1 or MDA-MB-436-CRISPRi-P3 cells, the MDA-MB-436-CRISPRi cell line was further infected at low MOI (<0.5) with AAVS1¹²⁵ or P3 sgRNA-containing LentiGuide-puro-NLS-GFP plasmid followed by puromycin selection (1 μg/ml). MDA-MB-436-CRISPRi-AAVS1 and MDA-MB-436-CRISPRi-P3 cell lines were further infected at low MOI (<0.5) with JUN sgRNA-containing pSAM110 plasmid and hygromycin selection (100 μg/ml). To immortalize primary MEFs, passage two (P2) MEFs were seeded in 6-cm plates at approximately 40% confluence and transfected with SV40 large T antigen using 5 μg of pMSSVLT plasmid kindly provided by Richard Baer (Columbia University)¹²⁶. After transfection, MEFs were cultured for approximately 10 passages until they became immortalized.

Library design and cloning

For the cell sorting-based CRISPR-Cas9 screens in Figures 1 and S1, we generated a 6-sgRNA-per-gene CRISPR/SaCas9 deletion lentiviral pooled library targeting 609 protein-coding human genes including DDR factors, chromatin, and cell cycle regulators. 97 non-targeting sgRNA with no binding sites in the genome were included as a control (library 1). The 609 genes that were targeted included some known essential genes and non-essential genes as controls¹²⁷. To select the genes, curated gene lists from the Reactome pathways “DNA repair” and “Chromatin organization” were used as seeds and expanded to include first neighbors from the BioGRID protein-protein interaction (PPI) network. These expanded gene sets were then trimmed using ClueGo gene ontology (GO) classifications to select genes annotated to either the DNA damage response or chromatin organization GO terms. Further gene set trimming was done using network features to prioritize genes with high network bottleneck statistics¹²⁸. A subset of potential unclassified DDR genes was added to this list based on their ability to localize to micro-irradiation laser stripes or associate with known DNA repair factors (data not shown). Non-targeting guides and early terminating guides were modified and added from the GeckoV2 library¹²⁹. To identify transcription factors (TFs) that promote PD-L1 expression specifically in SMARCAL1-proficient but not -deficient cells, we generated a 6-sgRNA-per-gene CRISPR/SaCas9 deletion lentiviral pooled library targeting 142 TF-coding human genes. The library contained two distinct classes of negative control gRNAs: 25 non-targeting control sgRNAs with no binding sites in the genome and 25 control sgRNAs targeting the AAVS1 locus (library 2). The 142 TF-coding human genes were identified by analyzing ATAC-seq data with ChromVar and TOBIAS framework^79,80. The sgRNAs for each gene included in library 1 or 2 were picked using the Genetic Perturbation Platform (GPP) sgRNA designer^130,121 (Tables S1 and S3). Two separate oligonucleotide pools (one per library) were ordered from Agilent Technologies. Oligonucleotides were designed including BsmBI restriction sites for cloning into the pXPR206 (plasmid #96920, Addgene) or pXPR206_Hygro lentiviral vector (library 1 and library 2, respectively) and primers targeting the flanking sequences of the oligonucleotides for library amplification. Final oligonucleotides for library 1 were designed for Gibson assembly-based cloning and oligonucleotides for library 2 were designed for Golden Gate-based cloning as sgRNA concatemers separated by Esp3I sites, allowing for dissociation of individual sgRNAs. Libraries were amplified as previously described¹³¹. The amplified library 1 was digested with Esp3I prior to cloning into the pXPR_206_Puro lentiviral vector using the Gibson assembly method. The amplified library 2 was cloned into a pXPR_206_Hygro lentiviral vector using a modified Golden Gate assembly protocol. Library DNA was precipitated with isopropanol and used to transform Endura electrocompetent cells (Lucigen). Plasmid DNA libraries were then purified using ZymoPure II Plasmid Maxiprep kit (Zymo Research) and sequenced (PE150) on an Illumina HiSeq sequencer to evaluate the representation of each sgRNA in the library.

Cell sorting-based CRISPR-Cas9 screens

All screens were performed in biological triplicates. MDA-MB-436 cells or control and SMARCAL1-KO #1 MDA-MB-436 cells were infected with the pooled lentiviral sgRNA library 1 or 2 at low MOI (<0.4) and selected for 48 hr with 1 μg/ml puromycin or 100 μg/ml hygromycin, respectively, commencing 24 hr after transduction. Transduced cells were then cultured for six days after selection, washed in PBS twice and collected by trypsinization and centrifugation (1500 g at 4°C for 5 min). Cells were collected as a pool (unsorted) or sorted into three different populations based on nuclear IRF3 staining and PD-L1 levels (IRF3^High/PD-L1^Low, IRF3^High/PD-L1^High, IRF3^Low/PD-L1^Low; screening performed with library 1), or into a PD-L1^Low population (30% of cells with lowest PD-L1 levels; screening performed with library 2). For sorting, cell pellets were washed in PBS twice and resuspended in flow cytometry staining buffer (10% FBS and 0.1% sodium azide PBS) before assessing PD-L1 levels on the cell surface via staining with BV421-conjugated anti-PD-L1 for 30 min. Cells were washed twice with flow cytometry staining buffer, fixed and permeabilized using Fixation/Permeabilization Diluent and Concentrate (#00–5223, #00–5123, eBioscience) and Permeabilization buffer (#00–5523, eBioscience) before intracellular staining with APC-conjugated anti-IRF3 and BV421-conjugated anti-PD-L1 for 30 min at room temperature (RT). Samples were then washed twice with Permeabilization buffer and PBS sequentially before sorting on the BD Influx sorter. A parallel batch of samples was stained with APC-conjugated and/or BV421-conjugated isotype controls (see antibody details in Table S5) and used as negative controls for the gating strategy. Importantly, Fixation/Permeabilization buffers used in the above procedure allow for the specific detection of nuclear proteins due to the insufficient retention of cytoplasmic proteins after the permeabilization procedure¹³². Genomic DNA from the sorted and unsorted pool populations was then extracted using the Quick-DNA Miniprep Plus kit (Zymo Research) using the protocol for fixed cells with minor modifications. sgRNA sequences were amplified by two rounds of PCR, with the second-round primers containing TruSeq Illumina dual indexes (UDI) for Illumina sequencing. The resulting libraries were sequenced (PE150) on an Illumina HiSeq sequencer.

CRISPR-Cas9 screens analysis

Pair-end reads were trimmed at the 5’ and 3’ using Cutadapt (v3.5) to remove constant sequences flanking the sgRNA sequence¹³³. After trimming, the number of reads per sgRNA was estimated using MAGeCK¹³⁴ (v0.5.9) -count function. Read counts were normalized using the total method. For the multicomparison analysis of the sorted populations (IRF3^High/PD-L1^Low, IRF3^High/PD-L1^High, IRF3^Low/PD-L1^Low) from the screens in Figures 1 and S1, we utilized MAGeCK maximum-likelihood estimation (MLE)²⁷. We used the unsorted population as a reference (see Matrix_design for MLE analysis in Table S1) and the normalized counts filtered for the sgRNAs with zero counts as input with the following parameters: --norm-method total --no-permutation-by-group –genes-varmodeling 610 --max-sgrnapergene-permutation 100 –permutation 1000. The resulting beta-scores from MAGeCK MLE analysis were converted into normalized beta-score using the NormalizeBeta function from the MAGeCKFlute R package (v1.10.0)²⁷. The single comparisons between PD-L1^Low population versus unsorted group, for each specific cell line or between SMARCAL1-proficient versus -deficient MDA-MB-436 cells with low PD-L1 levels (PD-L1^Low) from the screens in Figures 5 and S5 were performed by executing MAGeCK Robust Ranking Aggregation (RRA) with the following parameters: --gene-lfc-method median --gene-test-fdr-threshold 0.2 --paired. The paired function was used to compare sgRNAs between the PD-L1^Low and the unsorted populations of the same cell line.

Western blotting

Cells were seeded in 6-multiwell tissue culture plates and grown for two to three days until culture plates reached 80–90% confluency. In the case of the starvation experiments and drug treatment studies, cells were seeded in a 6-multiwell plate and after 24 hr treated for 24 hr with 20 ng/ml of recombinant human EGF (#AF100–15, PeproTech), 1000 IU/ml of recombinant human IFNB1 (#1160–05, Gold Biotechnology), 100 IU/ml of recombinant human IFNG (#1160–06, Gold Biotechnology), or PBS containing 0.1% BSA. Cells were then washed in PBS twice and collected by trypsinization and centrifugation (1500 × g at 4°C for 5 min). Cell pellets were washed in PBS twice, resuspended in lysis buffer (37.5 mM Tris-HCl pH 6.8, 1.25% SDS, 10% glycerol, 3% β-mercaptoethanol, 0.002% bromophenol blue), supplemented with protease inhibitor cocktail (#GB331–5, Gold Biotechnology), sonicated on ice and boiled for 5 min at 95°C. Sonication was performed on the Fisher Scientific Sonic Dismembrator Model 100 instrument (low power output, level 7, 10 sec ON/60 sec OFF double pulses). Protein lysates were resolved on a 4–15% Mini-Protean TGX precast polyacrylamide gels (#456–108, Bio-Rad) and transferred to 0.45 μm nitrocellulose membrane (#HATF00010, Millipore). Membranes were incubated for 1 hr in TBST (50 mM Tris/HCl pH 8, 150 mM NaCl, 0.1% Tween-20) containing 5% non-fat dry milk and then incubated for 1 hr at RT or overnight at 4°C with primary antibodies listed in Table S5. Detection was achieved using appropriate horseradish peroxidase (HRP)-conjugated secondary antibodies (Table S5).

Immunofluorescence

Cells were seeded in black 96-multiwell bottom-glass plates or grown on coverslips in 6-multiwell plates. When culture plates were 80–90% confluent (2–3 days of incubation), cells were simultaneously fixed and permeabilized (4% paraformaldehyde, 0.5% Triton X-100) for 10 min at RT. In the case of EdU staining, cells were pulse-labeled with 20 μM EdU for 2.5 hr, washed in PBS before fixation/permeabilization. EdU staining was performed using the Click-iT EdU assay kit (#C10337, Thermo Fisher Scientific) as per manufacturer’s recommendations. In the case of treatment with the RO-336 CDK1 inhibitor (#217721, Sigma-Aldrich), cells were treated for 72 hr before fixation and permeabilization. Cells were incubated in blocking solution (3% BSA in TBS-Tween 0.1%) for 1 hr and then incubated for 1.5 hr at RT or overnight at 4°C with the primary antibodies listed in Table S5. Cells were washed 3 times with PBS and incubated for 1 hr at RT with the appropriate secondary antibody, Alexa Fluor 488-labeled anti-mouse, Alexa Fluor 488-labeled anti-rabbit and Alexa Fluor 594-labeled anti-rabbit at 1:1000 dilution (A-11008 and A-11005, Thermo Fisher Scientific). After three washes in PBS, cells were incubated with DAPI for 5 min at RT to counterstain nuclei. Multi-color acquisitions were made using the ImageXpress Nano Automated Imaging System microscope or the Nikon Eclipse 50i microscope equipped with a 40× Plan Apo objective (0.95 numerical aperture). MetaXpress 6 or CellProfiler software was used for automated image analysis.

In situ PLA assay

MDA-MB-436 cells were seeded at a density of 1.5 × 10⁴ cells per well in a 96-well plate. The next day cells were washed with PBS, fixed with cold 4% formaldehyde in PBS for 10 min, and permeabilized with cold 0.5% Triton X-100 in PBS for 5 min. Following the fixation and permeabilization steps, cells were briefly rinsed with PBS. Next, cells were blocked in 5% bovine serum albumin (BSA) in PBS for 1 hr and stained overnight with the indicated primary antibodies at 4°C. Primary antibodies were diluted in PBS supplemented with 1% BSA and 0.1% saponin (Sigma-Aldrich) and cells incubated in blocking solution (3% BSA) overnight at 4°C with the primary antibodies listed in Table S5. After the incubation with primary antibodies, cells were washed (3 × 5 min) in PBS. Next, cells were incubated with PLA probes (Sigma-Aldrich, Duolink In Situ PLA Probe Anti-Rabbit PLUS, DUO92002; Duolink In Situ PLA Probe Anti-Mouse MINUS, DUO92004), followed by ligation of PLA probes and rolling-circle amplification reaction (Sigma-Aldrich, Duolink In Situ Detection Reagents Red, DUO92008) according to the manufacturer’s instructions. Staining with a single antibody was used as an internal negative control. Images were acquired using an ImageXpress Nano Automated Imaging System microscope (Molecular Devices) equipped with a 60x objective. MetaXpress High-Content Image Acquisition and Analysis Software was used to quantify the PLA signal from individual cells (over 500 cells per sample).

Alkaline comet assay

Endogenous DNA breakages were examined by alkaline comet assay (single-cell gel electrophoresis). MDA-MB-436 cells were plated in 12-multiwell plates (80,000 cells/well) and grown for 48 hr before collection. Collected cells were mixed with molten LMAgarose and pipetted onto CometSlide (double layer of 1% of NMAgarose). The slides were incubated with a lysis solution at pH 10 (25 mM NaCl, 100 mM EDTA, 10 mM Tris base) for 16 hr and with an alkaline solution at pH 13 (300 mM NaOH, 1 mM EDTA) for 20 min. The slides were then placed in horizontal chambers (FisherBiotech) and alkaline electrophoresis was performed at 25 V for 20 min. After incubation with a neutralization buffer (0.4 M Tris-HCl, pH 7.5) for 15 min, and ice-cold ethanol for 5 min, the slides were stained with GelRed (#41002, Biotium). Then, a minimum of 75 cells were analyzed per experimental point using a Nikon Eclipse 50i microscope equipped with a 40× Plan Apo objective (0.95 numerical aperture). To assess the amount of DNA damage, computer-generated tail moment values by CometScore Software Version 1.5 were used. Apoptotic cells (small comet head and very large comet tail) were excluded from the analysis.

Cell cycle analysis

Cell cycle analysis was performed as previously described with a few modifications¹³⁵. Briefly, exponentially growing cells were collected by trypsinization. After centrifugation (1500 × g at 4°C for 5 min), harvested cells were washed with PBS and 0.1% BSA twice, and fixed in 50% cold methanol. After an overnight incubation at −20°C, cells were centrifuged and washed twice with PBS and 0.1% BSA. Finally, pellets were resuspended, and cells were stained with PI-RNase solution (propidium iodide 20 μg/ml, RNase A 0.5 μg/ml) for 20 min at RT protected from light. Samples were analyzed by flow cytometry (BD LSRFortessa), and data analysis was conducted with the FlowJo software.

cGAMP assay

For intracellular 2’3’-cGAMP (cGAMP) quantification, 4.5 × 10⁵ MDA-MB-436 or 2 × 10⁶ B16/F10 cells were seeded in a 6-multiwell plate or 10 cm plate, respectively. When plates reached 80–90% confluency (2–3 days of incubation), cells were washed in PBS twice and collected by trypsinization and centrifugation (1500 × g at 4°C for 5 min). Cell pellets were subjected to three sequential washes in cold PBS to remove residual phenol red. All steps were performed on ice. To generate cell lysates, pellets were resuspended in Mammalian Protein Extraction Reagent buffer (#78501, M-PER, Thermo Fisher Scientific) and incubated for 5 min under gentle shaking. Whole-cell lysates were centrifugated at 14,000 × g for 15 min at 4°C and the supernatant was used for cGAMP quantification. cGAMP ELISA was performed according to the manufacturer’s instructions (#501700, Cayman Chemical). For sample comparison, protein concentration was determined by Bio-Rad protein assay (#5000006, Bio-Rad) and used for cGAMP level normalization across the samples.

Cytokine antibody array assay

Cytokine antibody assays were performed using a human cytokine array kit (R&D Systems, ARY005B) according to the manufacturer’s protocol. Briefly, 2.5 × 10⁶ MDA-MB-436 cells were seeded in a 10 cm plate. When plates reached 80–90% confluency (2–3 days of incubation), cells were washed in PBS twice and collected by trypsinization and centrifugation (1500 × g at 4°C for 5 min). Cell pellets were subjected to three sequential washes in cold PBS to remove residual phenol red and 1 × 10⁷ cells were resuspended in 1 ml of lysis buffer (#895943, Lysis buffer 17, R&D Systems) supplemented with Protease Inhibitor Cocktail I (#550, Tocris) and incubated for 30 minutes at 4°C. Whole-cell lysates were centrifugated at 14,000 × g for 15 min at 4°C and the supernatant was used for cytokines quantification. For sample comparison, protein concentration was determined by Bio-Rad protein assay (#5000006, Bio-Rad) and used for cGAMP level normalization across the samples.

Proximity-dependent labeling with BioID

To identify SMARCAL1 interactors by mass spectrometry, 10 × 10⁶ MDA-MB-436 cells were transfected with pHAGE-BioID control, pHAGE-BioID-SMARCAL1-WT, pHAGE-BioID-SMARCAL1-ΔN_1–115 or pHAGE-BioID-SMARCAL1-ΔHARP plasmids. Two days after DNA transfection, cells were treated with 50 μM biotin for 24 hr, washed 3 times and then harvested in PBS. Pulldown of biotinylated proteins was performed as previously described¹²².

Immunoprecipitation and on-beads digestion for mass spectrometry

Proteins bound to streptavidin were washed five times with 200 μl of 100 mM Tris-pH 8.0. Proteins were reduced with 10 mM TCEP and alkylated with 11 mM iodoacetamide (IAA), which was quenched with 5 mM DTT. Protein digestion was performed by adding 1 μg of trypsin/LysC mix overnight at 37°C and 1400 rpm. The next day, digested peptides were collected in a new microfuge tube, and digestion was stopped by the addition of 1% TFA (final v/v), followed by centrifugation at 14,000 × g for 10 min at room temperature. Cleared digested peptides were desalted on an SDB-RPS Stage-Tip¹³⁶ and then dried in a speed-vac. Peptides were then dissolved in 3% acetonitrile/0.1% formic acid.

Liquid chromatography with tandem mass spectrometry (LC-MS/MS)

Peptides were separated within 80 min at a flow rate of 400 nl/min on a reversed-phase C18 column with an integrated CaptiveSpray Emitter (25 cm × 75 μm, 1.6 μm, IonOpticks). Mobile phases A and B were with 0.1% formic acid in water and 0.1% formic acid in ACN. The fraction of B was linearly increased from 2 to 23% within 70 min, followed by an increase to 35% within 10 min and a further increase to 80% before re-equilibration. The timsTOF Pro was operated in PASEF mode¹³⁷ with the following settings: Mass Range 100 to 1700m/z, 1/K0 Start 0.6 Vs/cm⁻², End 1.6 Vs/cm⁻², Ramp time 100ms, Lock Duty Cycle to 100%, Capillary Voltage 1600V, Dry Gas 3 l/min, Dry Temp 200°C, PASEF settings: 10 MSMS Frames (1.16 seconds duty cycle), charge range 0–5, active exclusion for 0.4 min, Target intensity 20000, Intensity threshold 2500, CID collision energy 59eV. A polygon filter was applied to the m/z and ion mobility plane to select features most likely representing peptide precursors rather than singly charged background ions.

LC-MS/MS data analysis

Acquired PASEF raw files were analyzed using the MaxQuant environment v.2.3.1.0 and Andromeda for database searches at default settings with a few modifications¹³⁸. The default is used for first search tolerance and main search tolerance (20 ppm and 4.5 ppm, respectively). MaxQuant was set up to search with the reference human proteome database downloaded from UniProt. MaxQuant performed the search trypsin digestion with up to 2 missed cleavages. Peptide, site, and protein false discovery rates (FDR) were all set to 1% with a minimum of 2 peptides needed for identification. The following modifications were used for protein identification and quantification: Carbamidomethylation of cysteine residues (+57.021 Da) was set as static modifications, while the oxidation of methionine residues (+15.995 Da), and deamidation (+0.984) on asparagine were set as a variable modification. The protein group table obtained from MaxQuant was further used for data analysis. In particular, naturally biotinylated carboxylases were specifically excluded from the analysis of BioID experiments⁸¹. Moreover, proteins with less than two unique peptides were filtered out. Identified proteins were considered putative SMARCAL1 interactors if the peptide intensity was at least 2-fold enriched in BirA*–SMARCAL1 relative to BirA* alone.

AlphaFold analysis

Sequences of the genes detected as potential partners of SMARCAL1 from the BioID experiment were retrieved from the UniProt database and were submitted to three iterations of MMseqs2¹³⁹ against the uniref30_2202 database¹⁴⁰. The resulting alignments were filtered by hhfilter¹⁴¹ using parameters (‘id’=100, ‘qid’=25, ‘cov’=50) and the taxonomy was assigned to every sequence keeping only one sequence per species. Full-length sequences in the alignments were then retrieved and the sequences were realigned using MAFFT¹⁴² with the default FFT-NS-2 protocol. To build the so-called mixed co-alignments, sequences in the alignment of individual partners were paired according to their assigned species and left unpaired in case no common species were found¹⁴⁰. The concatenated MSAs were used as input to run 1 to 5 independent runs of the AlphaFold2 algorithm with 3 recycles each^108,143 using ColabFold v1.5.2¹⁴⁰ with the Multimer v2.3 model parameters¹⁴³ on NVidia A100 GPUs. Four scores were provided by AlphaFold2 to rate the quality of the models, the pLDDT, the pTMscore¹⁰⁸, the ipTMscore and the model confidence score (weighted combination of pTM and ipTM scores with a 20:80 ratio)¹⁴³. The scores obtained for all the generated models are reported Table S4.

ChIP-qPCR analysis

For ChIP-qPCR analysis, SMARCAL1-KO MDA-MB-436 cells reconstituted with pHAGE-SMARCAL1-Ct-Flag-HA were seeded into 10 cm plates. When culture plates reached 90–95% confluency (~10⁷ cells), cells were treated with 1% formaldehyde for 10 min at RT prior to the addition of glycine at the final concentration of 0.125 M for 5 min. Cross-linked cells were processed for ChIP according to the manufacturer’s instruction (#17–10086, EZ-Magna ChIP A/G kit Millipore) using the antibodies listed in Table S5. The immunoprecipitated DNA samples were then quantified by RT-qPCR. DNA oligo sequences used in this study are available in Table S5.

ADP-Glo ATPase assay

ATP to ADP conversion by recombinant full-length SMARCAL1 (purified as previously⁵⁶) was measured using the ADP-Glo kinase assay kit (Promega) as per manufacturer’s recommendations. In brief, SMARCAL1 (0 to 20 nM) was mixed with potential stimulant (20 nM 6-N-66 nucleosome [EpiCypher], 20 nM free DNA [217 bp as used to wrap the 6-N-66 nucleosome] or buffer control) and 1 mM ATP in remodeling buffer (20 mM Tris-HCl pH 7.5, 50 mM KCl, 3 mM MgCl₂, 0.01% Tween-20, 0.01% BSA). Remodeling reactions were prepared at 5 μl each in 384-well format and incubated at 23°C for 1 hr before ATP depletion and ADP detection. Results were read as luminescence signals using the EnVision Workstation (version 1.13.3009.1409). All reactions were performed in duplicate, and averages and standard error of mean (SEM) calculated.

EpiDyne-PicoGreen nucleosome remodeling assay

Nucleosome remodeling activities were measured with EpiDyne-PicoGreen, a restriction-enzyme accessibility (REA) assay optimized for increased throughput and sensitivity⁹⁸. In brief, recombinant ATPase (SMARCAL1 or SMARCA4 [#15–1014, EpiCypher]; 0 to 10 nM) was mixed with 10 nM EpiDyne nucleosome remodeling substrate (5’ biotinylated and a terminally (6-N-66) or centrally (50-N-66) located nucleosome positioned so as to protect a GATC restriction enzyme (RE) motif; EpiCypher) and 1 mM ATP in 20 μl remodeling buffer. Reactions were incubated at 23°C in 384-well format and, at indicated time points, mixed with an equal volume of streptavidin magnetic beads (New England Biolabs, NEB) pre-equilibrated in 2x quench buffer (20 mM Tris-HCl pH 7.5, 600 mM KCl and 0.01% Tween-20) and incubated for 1 hr at 23°C on a nutator. At reaction completion, potentially remodeled bead-immobilized substrates were collected by magnetic pulldown (384-well magnet; V&P scientific) and successively resuspended/recaptured in three changes of wash buffer (20 mM Tris-HCl pH 7.5, 300 mM KCl, 0.01% Tween-20) and one change of RE buffer (20 mM Tris-HCl, pH 7.5, 50 mM KCl, 3 mM MgCl₂, 0.01% Tween-20). Beads were then resuspended in 20 μl RE buffer with 50 units/ml DpnII (NEB), and incubated at 23°C for 30 min while on a nutator. After magnetic pulldown, RE-released nucleosomal DNA fragments were collected as supernatants, transferred to a new black 384-well plate, mixed with an equal volume of Quant-iT PicoGreen dsDNA reagent (Thermo Fisher Scientific; Component A) and 1 unit/ml thermolabile proteinase K (NEB) in TE, and incubated at 23°C for 1 hr. Fluorescence intensity (excitation at 480 nm / emission at 531 nm) was detected on an EnVision Workstation (version 1.13.3009.1409), and data expressed as relative fluorescence units (RFU). For cooperativity assays, SMARCA4 (1 nM) was mixed with SMARCAL1 (0–8 nM) before addition of 10 nM terminally or centrally positioned nucleosome substrates and 1 mM ATP in remodeling buffer. Time-course based nucleosome remodeling reactions and subsequent data acquisition were performed as above.

CUTANA CUT&RUN, Illumina sequencing, and data analysis

CUT&RUN was performed using CUTANA ChIC / CUT&RUN Kit User Manual Version 3.4, an optimized version of that previously described¹⁴⁴. For each CUT&RUN reaction, 500,000 permeabilized native cells (parental or JUN-depleted MDA-MB-436) were immobilized onto Concanavalin-A beads (#21–1401, Con-A; EpiCypher) and incubated overnight (4°C with gentle rocking) with 0.5 mg of antibodies listed in Table S5. All K-methyl PTM antibodies were validated to SNAP-CUTANA K-MetStat nucleosome standards panel (#19–1002, EpiCypher), as previously⁶⁹. pAG-MNase (#15–1016, EpiCypher) was added to the reaction and activated, and CUT&RUN enriched DNA was purified using the CUTANA ChIC/CUT&RUN Kit (#14–1048, EpiCypher). 5 ng DNA was used to prepare sequencing libraries with the CUTANA CUT&RUN Library Prep Kit (#14–1001/2, EpiCypher).

Libraries were sequenced on the Illumina NextSeq 2000 platform and paired-end fastq files aligned to the hg38 reference genome using the Bowtie2 algorithm¹⁴⁵. Uniquely aligned reads were retained, while duplicate reads and those in blacklist regions¹⁴⁶ were filtered out prior to subsequent analyses. Peaks were called using SICER2 (Spatial clustering approach for the Identification of ChIP-Enriched Regions)¹⁴⁷ and annotated by the function “annotatePeaks.pl” using Homer (4.11)¹⁴⁸. Venn diagrams were generated using intervene (0.6.5) and matplotlib_venn (0.11.7). The correlation plots and the heatmaps for genomic distribution of CUT&RUN signals were generated by the functions “multiBamSummary”/”plotCorrelation” (--corMethod pearson --skipZeros) and “computeMatrix”/”plotHeatmap” of deepTools (v3.5.1)¹⁴⁹. TSS input file was derived from the hg38 RefSeq annotation, filtered for protein-coding genes only and using one isoform per gene with the most 5’ TSS. The de novo motif discovery analysis was performed by the function “findMotifsGenome.pl” of Homer (4.11)¹⁴⁸ using SMARCAL1 peaks called in all three biological replicates (“common peaks”, see also Table S2).

ATAC-seq and ATAC-seq analysis

ATAC-seq was performed as previously described⁷² with the exception that 0.1% NP-40 was used for cell lysis instead of IGEPAL. ATAC-seq libraries were quantified with the KAPA Library Quantification kit (Roche), pooled in equal amounts and sequenced (PE150) on an Illumina HiSeq sequencer. Raw reads were trimmed with Cutadapt (v1.16)¹³³, with the following parameters: -m 50 -l 50 --trim-n. ATAC-seq reads were then aligned to the Homo sapiens (hg38) or Mus musculus (mm10) genome using HISAT2 (v2.1.0, parameter: -X 2000)¹⁵⁰. Potential PCR duplicates were then removed by the function “MarkDuplicates” (parameter: REMOVE_DUPLICATES=true) of Picard (v2.23.1) (http://broadinstitute.github.io/picard/) and mitochondrial DNA removed using the function “intersect” of BEDTools (v2.27.1)¹⁵¹ based on the blacklist from ENCODE¹⁴⁶. The correlation analysis for genomic distribution of ATAC-seq signals was performed by the functions “multiBigwigSummary” (parameter: --binSize 1000) and “plotCorrelation” (--corMethod pearson --skipZeros) of deepTools (v3.3.2)¹⁴⁹. Peaks of ATAC-seq data were defined using MACS2 (v2.1.2, parameters: -g hs -p 1e-5 -m 5 50)¹⁵². The peaks from biological replicates were merged using the function “merge” of BEDTools (v2.27.1) and the p-values of merged peaks were calculated as the average of the p-values of peaks from biological replicates. Merged peaks were annotated by the R package “ChIPseeker”¹⁵³. The reads number for each peak was measured by featureCounts (v1.6.1)¹⁵⁴. A peak was defined as “common” if one base pair or more overlaps with the peak in the dataset of comparison. The H3K27ac peaks from ChIP-seq experiments (GSE85158) were mapped to the ATAC-seq peaks by the function “map” of BEDTools (v2.27.1).

PLATE-seq assay and VIPER analysis

MDA-MB-436 SMARCAL1-KO clones, MDA-MB-436 control cells and MDA-MB-436 parental cells were seeded in 6-multiwell plates (450,000 cells per well). MDA-MB-436 parental cells were depleted of SMARCAL1 by using reverse transfection or treated with siRNA control. 24 hr after transfection, MDA-MB-436 cells were washed twice with PBS, collected by trypsinization and centrifugation (1500 × g at 4°C for 5 min), and re-seeded in duplicate in a white 96-multiwell plate at a starting density of 10,000 cells per well. After 48 hr, cells were lysed in Buffer TCL (Qiagen) supplemented with 2-mercaptoethanol and processed for high-throughput mRNA sequencing in the PLATE-seq assay⁴⁰. Reads from each well were mapped to the UCSC GRCh38 human genome assembly using STAR (v2.5.2b). Counts were generated using STAR parameter quantMode = “GeneCounts”. Count files were merged in a data matrix and processed together for downstream analysis. Next, genes were filtered retaining only those expressed at least in one sample. Variance stabilizing transformation from DESeq2 (v1.30.1) was used to normalize the counts¹⁵⁵. The biological replicates for each SMARCAL1-KO clone and SMARCAL1-depleted cells were pooled together and a linear model was fitted for each SMARCAL1-KO clone and SMARCAL1-depleted cell line against a set of control cells (see Queries in Figure 2C for the specific analyses). The limma package (v3.46.0) was then used to fit the linear model and compute p-values and moderated t-statistics for each gene¹⁵⁶. To generate a gene expression signature (GES) for SMARCAL1 deficiency, we used a vector of the above statistics to create a gene expression profile of z-score values representing the differential effect between SMARCAL1-deficient and -proficient cells (for each SMARCAL1-KO clone and SMARCAL1-depleted cell line, independently). We then used the ARACNe algorithm^157–159 to reverse-engineer a transcriptional regulatory network from Breast Cancer (BRCA) patient data as released by the TCGA consortium¹⁶⁰. To infer the activity status of regulatory proteins (RPs) resulting from SMARCAL1-deficient cells, we used the generated BRCA-specific network and the GES associated with SMARCAL1 deficiency (condition-specific) as input in the VIPER algorithm⁴¹. To limit biased effects due to excessively large or small regulons in the enrichment analysis, each RP was pruned by retaining the set of 50 targets with the highest likelihood scores, and RPs with less than 30 targets were removed from the analysis. This analysis led to a Protein Activity Signature (PAS) associated with each SMARCAL1-KO clone and SMARCAL1-depleted cell line. The generated PASs were used as input in a pathway enrichment analysis (pathway activity) performed using the Broad MSigDB Hallmark gene sets, as described below.

RNA extraction and RT-qPCR assay

For total RNA extraction, cells were seeded in 6-multiwell culture plates. When culture plates reached 80–90% confluency (2–3 days of incubation), cells were washed in PBS twice and collected by trypsinization and centrifugation (1500 × g at 4°C for 5 min). For total RNA extraction after drug treatment, cells were seeded in a 6-multiwell plate and after 24 hr exposed to the RO-336 CDK1 inhibitor (#217721, Sigma-Aldrich) or DMSO for 72 hr before collection. All reagents, buffers and containers involving RNA work were RNase-free grade. Total RNA was isolated using the RNeasy kit (#74104, Qiagen) according to the manufacturer’s instructions. For real-time quantitative PCR (RT-qPCR) assay, 1 μg of total RNA was reverse-transcribed into cDNA using random hexamer primers (#N8080127, Thermo Fisher Scientific) and MMLV High Performance Reverse Transcriptase kit (#RT80125K, Lucigen). An equal amount of cDNA from different samples was mixed with Power SYBR green PCR master mix (#4367659, Thermo Fisher Scientific). Gene-specific primers with sequences listed in Table S5 were used for RT-qPCR amplification performed on a QuantStudio 3 System instrument. Gene-specific mRNA level for each experimental condition was determined using the ΔΔCT method and normalized to TBP or Gapdh mRNA. RT-qPCR results were presented as fold change of gene expression in the test samples compared to the control.

RNA-seq library preparation and analysis

RNA quality was assessed on an Agilent Bioanalyzer instrument using Agilent RNA 6000 Pico Reagents. RNA-seq libraries were prepared with the Illumina TruSeq Stranded Total RNA Sample Preparation kit. Final libraries were quantified with the KAPA Library Quantification Kit (Roche), pooled in equal amounts and sequenced (PE150) on an Illumina HiSeq4K sequencer. Sequencing was done at the Next Generation Sequencing Platform of the Genome Institute of Singapore. Reads were then aligned to the Homo sapiens genome (hg38) using STAR aligner¹⁶¹ (V2.6.1d) with default settings and annotated using GENCODE (v29). Read counting per gene was performed by using HTseq software (v0.11.2). Raw counts were filtered retaining only genes expressed in at least one sample. The “upperquartile” method from edgeR (v3.32.1) was used to normalize the data. Sample biological replicates were used as blocking factor in the experimental design and a common negative binomial dispersion parameter was estimated for the whole dataset. Next, count data were fitted using a quasi-likelihood negative binomial generalized log-linear model. Gene-wise quasi-likelihood dispersions were estimated, and moderated using empirical Bayes. Contrasts were created between SMARCAL1-KO and control cells, and quasi-likelihood gene-wise F-tests were performed across the conditions to test for differential expression. To correct for multiple hypothesis testing, we adopted the Benjamini-Hochberg False Discovery Rate (FDR) procedure, selecting genes with FDR <0.05 as differentially expressed. To generate a GES encompassing SMARCAL1-KO transcriptional changes, and containing all the genes from the RNA-seq assay, we mapped the DGE data (FDR-adjusted p-values and log-Fold Change scores) to a normal distribution. We then used SMARCAL1-KO GES to interrogate the Broad MSigDB Hallmark gene sets⁴². The analytical Rank Enrichment Analysis (aREA) algorithm was used to perform pathway enrichment analysis⁴¹, reporting positive Normalized Enrichment Scores (NES) for pathway upregulation in SMARCAL1-KO cells, and negative NES for downregulation. To select leading-edge genes from enrichment analysis, we performed canonical Gene Set Enrichment Analysis (GSEA) using 1,000 permutations, and reported p-value and enrichment scores on figure panels for visualization purposes only.

Pan-cancer analysis of the TCGA dataset

Cancer patients data were obtained from the TCGA research network. Thirty-three TCGA cohort RNA-seq Level 3 and clinically annotated datasets were downloaded from the Broad GDAC firehose website (http://gdac.broadinstitute.org). mRNA scaled estimates were transformed in log₂ transcript per million (log2(TPM)). To perform a robust assessment of SMARCAL1 expression differences between tumor and normal samples, a Welch’s unequal variances t test with the normal samples as reference was performed. To assess whether SMARCAL1 expression is associated with survival outcome, we assigned patients to two equally-sized groups based on SMARCAL1 expression. Upper (SMARCAL1^High) and lower (SMARCAL1^Low) quartiles were utilized in our studies. Following recent recommendations from a TCGA pan-cancer study on survival analysis data quality¹⁶², we performed Progression-Free Survival (PFS) analysis using progression-free interval endpoints, which included death events from tumors but not from other causes. PFS survival analysis was performed with the survival R package (v2.41.3). Patient-specific GES were computed as z-score values between SMARCAL1^Low and SMARCAL1^High patient groups. The Stouffer’s method was used to integrate the scores and generate a cohort-specific GES to interrogate changes in transcriptional programs associated with reduced SMARCAL1 expression in patient tumors. Pathway enrichment analysis between SMARCAL1^Low and SMARCAL1^High patient groups was performed with the aREA function, as described above. For leukocyte infiltration analysis in TCGA datasets, patient-specific leukocyte infiltration values were downloaded from Taylor et al.⁸⁵. Then, a cohort-specific leukocyte score ratio was computed as the ratio of the average leukocyte score in SMARCAL1^Low over SMARCAL1^High patient groups. Cross-cohort Pearson’s correlation analysis was performed between the leukocyte score ratio and the NES scores (SMARCAL1^Low vs SMARCAL1^High patient groups) associated with the “inflammatory response” from the Hallmark gene sets. The panels in Figure S9D–E were generated as described above by using as input the indicated genes.

Gene expression data analysis in immunotherapy datasets

For SMARCAL1 expression analysis in non-responder and responder groups of patients treated with anti-PD-1 therapy, the gene expression data were downloaded from Hugo et al. (GSE78220) and Riaz et al. (GSE91061)^86,87 and used as released by the authors. To define the responders and non-responders, we used the RECIST criteria¹⁶³ defined in the original clinical trials (CR, complete response; PR, partial response; SD, stable disease; PD, progressive disease). For the Riaz et al. cohort that includes only one CR patient, CR and PR patients were considered as responders and SD and PD patients as non-responders. For the Hugo et al. cohort that includes multiple CR patients, we considered CR patients as responders. To assess SMARCAL1 expression differences between responders and non-responders, a Welch’s unequal variances t test was performed.

T cell-mediated tumor cell killing co-culture assay

Human CD8⁺ T cells were activated with 25 μl/ml of ImmunoCult Human CD3/CD28 T Cell Activator (#10971, STEMCELL Technologies) for two days. MDA-MB-436 target cells expressing NLS-dsRed were seeded (2 × 10⁴) in black 96-multiwell bottom-glass plates. After 24 hr, cells were imaged to obtain the baseline of cell seeding and then incubated or not with activated CD8⁺ T cells at the indicated CD8⁺ T cells/tumor cells ratio. The plates were imaged again 24, 48 and 72 hr after activated CD8⁺ T cells addition. Image acquisition was performed with an ImageXpress Nano Automated Imaging System microscope (Molecular Devices) with a 10x objective. Image analyses for segmentation and counting of RFP-labeled tumor cells were performed using the MetaXpress imaging software. Each experiment was performed in at least two biological replicates (independent rounds of seeding and CD8⁺ T cells activation) and each condition was assessed in technical triplicates. T cell-mediated tumor cell killing activity was reported as tumor cell proliferation in presence of CD8⁺ T cells normalized on the tumor cell proliferation in absence of CD8⁺ T cells of the same cell population.

In vitro tumor growth

B16/F10 cells were seeded into 12-multiwell plates at 10–15% density, and allowed to grow for up to five days (80–90% density), with fresh medium addition on day 2 without removing the old medium. Cells were fixed and stained with a solution containing 1% formaldehyde and 1% crystal violet in methanol at the indicated time points. After five days, when all the time points were collected, the absorbed dye was resolubilized with methanol containing 0.1% SDS, transferred into 96-multiwell plates and measured photometrically (595 nm) with a microplate reader. After background subtraction, tumor cell growth was reported in arbitrary units. Data are representative of three independent experiments.

Animal studies

For in vivo tumor growth and survival analysis, 7–8 weeks old female mice were anesthetized with isoflurane, shaved at the injection site, and then injected in the flank subcutaneously with 500,000 B16/F10 tumor cells (Smarcal1-KO + EV, Smarcal1-KO + Smarcal1 cDNA, Smarcal1-KO /Cgas-KO + EV and Smarcal1-KO/Cgas-KO + Smarcal1 cDNA). Tumor dimensions were measured with an electronic caliper every 2–3 days, as indicated. Tumor volume was measured using the formula (length × width²/2), where length represents the largest tumor diameter and width represents the perpendicular tumor diameter. Mice were sacrificed when tumors reached 1500 mm³ or upon ulceration/bleeding. For antibody treatments, mice were randomized and treated with 200 μg of anti-PD-L1 (clone 10F.9G2, Bio X Cell) and/or anti-CTLA-4 (clone 9H10, Bio X Cell) or rat IgG2a isotype control (#BE0090, Bio X Cell) via intraperitoneal injection (IP) at day 4, 7, 10, and 13 post tumor injection, as indicated. For CD8⁺ cells depletion, mice were given 300 μg of α-CD8 (clone 2.43, Bio X Cell) or rat IgG2a isotype control via IP starting on day −3 relative to tumor injection, and then every 3 days until completion of the experiment. All antibodies were diluted in the dilution buffer suitable for in vivo injections (#IP007, Bio X Cell).

Tumor-infiltrating leukocyte flow cytometry

Tumors were excised on day 11 post-injection when the tumor volumes were comparable between groups. Single cell suspensions were obtained via mechanical dissociation of B16/F10 tumors and subsequently filtered through a 70 μm filter. Cells were then stained with the fluorochrome-conjugated antibodies listed in Table S5. For intracellular staining, cells were fixed and permeabilized using Fixation/Permeabilization Diluent and Concentrate (#00–5223, #00–5123, eBioscience) and Permeabilization buffer (#00–5523, eBioscience) before staining with fluorochrome-conjugated FOXP3 antibody (see details in Table S5). Samples were analyzed by flow cytometry (BD LSRFortessa), and data analysis was conducted with the FlowJo software.

QUANTIFICATION AND STATISTICAL ANALYSIS

Statistical analysis

Statistical analysis was performed using GraphPad Prism software (v7.0), unless otherwise specified. Statistical analysis details for the different experiments are reported in figure legends or the methods section. In all cases: ns not significant; * p<0.05; ** p<0.01; *** p<0.001; **** p<0.0001.

Software

RStudio (v3.6), QuantStudio, MetaXpress (v6), FlowJo (v10.7), and GraphPad Prism (v7.0) were utilized in this study.

Key resources table.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Mouse monoclonal anti-SMARCAL1 (A2)	Santa Cruz Biotechnology	Cat# sc-376377, RRID:AB_10987841
Rat monoclonal anti-TUBULIN (YL1 / 2)	Abcam	Cat# ab6160, RRID:AB_305328
Mouse monoclonal anti-VINCULIN (hVIN-1)	Sigma-Aldrich	Cat# V9131, RRID:AB_477629
Rabbit monoclonal anti-phospho-STAT1 (Tyr701) (58D6)	Cell Signaling Technology	Cat# 9167, RRID:AB_561284
Mouse monoclonal anti-STAT1 (STAT1-79)	Thermo Fisher Scientific	Cat# AHO0832, RRID:AB_1502071
Rabbit monoclonal anti-phospho-IRF3 (Ser386)	Abcam	Cat# ab76493, RRID:AB_1523836
Mouse monoclonal anti-IRF3 (D9J5Q)	Cell Signaling Technology	Cat# 10949, RRID:AB_2797733
Rabbit monoclonal anti-cGAS (D1D3G)	Cell Signaling Technology	Cat# 15102, RRID:AB_2732795
Rabbit monoclonal anti-Cgas (D3O8O)	Cell Signaling Technology	Cat# 31659, RRID:AB_2799008
Rabbit polyclonal anti-PD-L1	GeneTex	Cat# GTX104763, RRID:AB_1240586
Mouse monoclonal anti-PD-L1 (MIH1)	eBioscience	Cat# 14-5983-82, RRID:AB_467784
Mouse monoclonal anti-PD-L1 (2096C)	R&D Systems	Cat# MAB90781, RRID:AB_2921258
Rabbit polyclonal anti-phospho-H2A.X (Ser139)	Bethyl Laboratories	Cat# A300-081A, RRID:AB_203288
Mouse monoclonal anti-phospho-H2A.X (Ser139)	BioLegend	Cat# 613402, RRID:AB_315795
Rabbit polyclonal anti-RPA32	Bethyl Laboratories	Cat# A300-244A, RRID:AB_185548)
Rabbit monoclonal anti-JUN	Cell Signaling Technology	Cat# 9165, RRID:AB_2130165
Rabbit isotype control IgG	EpiCypher	Cat# 13-0042, RRID:AB_2923178
Rabbit polyclonal anti-SMARCAL1	Abcam	Cat# ab154226
Rabbit mixed monoclonal anti-H3K4me3	EpiCypher	Cat# 13-0041
Rabbit monoclonal anti-H3K27me3	Thermo Fisher Scientific	Cat# MA5-11198
Rabbit polyclonal anti-JUN	EpiCypher	Cat# 13-2019
Mouse monoclonal anti-RNA Pol II	Millipore	Cat# 05-623
Mouse monoclonal anti-FLAG (M2)	Sigma-Aldrich	Cat# F1804, RRID:AB_262044
Mouse polyclonal anti-IgG	Millipore	Cat# 12-371B, RRID:AB_2617156
Rat monoclonal APC-Cy7 conjugated anti-CD45	BioLegend	Cat# 103115, RRID:AB_312980
Rat monoclonal BV786 conjugated anti-CD3	BD Horizon	Cat# 564010, RRID:AB_2738540
Rat monoclonal BV605 conjugated anti-CD8	BioLegend	Cat# 100744, RRID:AB_2562609
Rat monoclonal APC conjugated anti-CD4	BioLegend	Cat# 100516, RRID:AB_312719
Rat monoclonal FITC conjugated anti-FOXP3	Thermo Fisher Scientific	Cat# 11-5773-82, RRID:AB_465243
Mouse monoclonal APC conjugated IgG1, k isotype ctrl	BioLegend	Cat# 400119, RRID:AB_2888687
Mouse monoclonal BV421 IgG2b, k isotype ctrl	BioLegend	Cat# 400341, RRID:AB_10898160
Mouse monoclonal APC conjugated anti-human IRF3	BioLegend	Cat# 395906, RRID:AB_2888751
Mouse monoclonal BV421 conjugated anti-PD-L1	BioLegend	Cat# 329714, RRID: AB_2563852
Rat monoclonal InVivoPlus anti-PD-L1	Bio X Cell	Cat# BP0101, RRID:AB_2934050
Rat monoclonal InVivoPlus anti-CTLA-4	Bio X Cell	Cat# BP0131 RRID:AB_10950184
Rat monoclonal InVivoPlus anti-mouse CD8α	Bio X Cell	Cat# BP0061 RRID:AB_1125541
Mouse IgG, HRP-linked	Jackson ImmunoResearch	Cat# 115-035-003 RRID_AB_10015289
Rabbit IgG, HRP-linked	Millipore	Cat# AP156P, RRID:AB_91699
Rat IgG, HRP-linked	Thermo Fisher Scientific	Cat# 31470, RRID:AB_228356
Alexa Fluor 488 (anti-mouse)	Thermo Fisher Scientific	Cat# A-11001, RRID: AB_2534069
Alexa Fluor 488 (anti-rabbit)	Thermo Fisher Scientific	Cat# A-11008, RRID: AB_143165
Alexa Fluor 594 (anti-rabbit)	Thermo Fisher Scientific	Cat# A-11012, RRID: AB_2534079
PLA Probe Anti-Rabbit PLUS	Sigma-Aldrich	Cat# DUO92002
PLA Probe Anti-Mouse MINUS	Sigma-Aldrich	Cat# DUO92004
Bacterial and virus strains
Subcloning Efficiency DH5α	Thermo Fischer Scientific	Cat# 18265-017
Endura electrocompetent cells	Lucigen	Cat# 60242-2
Chemicals, peptides, and recombinant proteins
Fixable Viability Dye eFluor 780	eBioscience	Cat# 65-0865
Puromycin	Gold Biotechnology	Cat# P-600-100
Blasticidin	Gold Biotechnology	Cat# B-800-100
Hygromycin	Gold Biotechnology	Cat# H-270-1
Polybrene	Fisher Scientific	Cat# TR-1003-G
pCR4Blunt-TOPO	Thermo Fisher Scientific	Cat# 450031
Transfection reagent: TransIT-293	Mirus	Cat# MIR 2700
Transfection reagent: Lipofectamine RNAiMAX	Thermo Fisher Scientific	Cat#13778-075
Transfection reagent: Lipofectamine 3000	Thermo Fisher Scientific	Cat# L3000015
Mini-PROTEAN TGX 4–15% gels	Bio-Rad	Cat# 4561084DC
0.45 μm nitrocellulose membrane	Millipore	Cat# HATF00010
Crystal violet	Fisher Scientific	Cat# C581-25
Random hexamer primers	Thermo Fisher Scientific	Cat# N8080127
MMLV High Performance Reverse Transcriptase kit	Lucigen	Cat# RT80125K
Power SYBR green PCR master mix	Thermo Fisher Scientific	Cat# 4367659
RNaseOUT Recombinant Ribonuclease Inhibitor	Thermo Fisher Scientific	Cat# 10777019
Propidium iodide solution	Sigma-Aldrich	Cat# 4864
Cisplatin	Sigma-Aldrich	Cat# P4394-25MG
Olaparib	Selleck Chemicals	Cat# S1060
Camptothecin	Sigma-Aldrich	Cat# C9911
GelRed	Biotium	Cat# 41002
Human recombinant IL2	Stemcell Technologies	Cat# 78036
Human recombinant IFNB1	Gold Biotechnology	Cat# 1160-05
Human recombinant IFNG	Gold Biotechnology	Cat# 1160-06
Human recombinant EGF	PeproTech	Cat# AF100-15
EdU	Thermo Fisher Scientific	Cat# A10044
RO-336 CDK1 inhibitor	Sigma-Aldrich	Cat# 217721
M-PER	Thermo Fisher Scientific	Cat# 78501
Bio-Rad protein assay	Bio-Rad	Cat# 5000006
Lysis buffer 17	R&D Systems	Cat# 895943
Protease Inhibitor Cocktail I	Tocris	Cat# 550
D-Biotin	Caisson Labs	Cat# B-002-1GM
SMARCA4 Chromatin Remodeling Enzyme (Human BRG1)	EpiCypher	Cat# 15-1014
Con-A	EpiCypher	Cat# 21-1401
pAG-MNase	EpiCypher	Cat# 15-1016
SNAP-CUTANA K-MetStat Panel	EpiCypher	Cat# 19-1002
BP clonase II	Thermo Fisher Scientific	Cat# 1178910
LR clonase II	Thermo Fisher Scientific	Cat# 11791020
Fixation/Permeabilization Diluent	eBioscience	Cat# 00-5223
Fixation/Permeabilization Concentrate	eBioscience	Cat# 00-5123
Permeabilization Buffer	eBioscience	Cat# 00-5523
ImmunoCult Human CD3/CD28 T Cell Activator	STEMCELL Technologies	Cat# 10971
EpiDyne Nucleosome Remodeling Assay Substrate ST601-GATC1	EpiCypher	Cat# 16-4111
EpiDyne Remodeling Assay Substrate DNA, ST601-GATC1	EpiCypher	Cat# 16-4111
EpiDyne Nucleosome Remodeling Assay Substrate ST601-GATC1,50-N-66	EpiCypher	Cat# 16-4114
Buffer TCL	Qiagen	Cat# 1031576
ATP solution, Tris buffered	Thermo Fisher Scientific	Cat# R1441
Hydrophilic streptavidin magnetic beads	New England Biolabs	Cat# S1421S
DpnII	New England Biolabs	Cat# R0543M
Thermolabile Proteinase K	New England Biolabs	Cat# P8111S
Critical commercial assays
Click-iT EdU assay kit	Thermo Fisher Scientific	Cat# C10337
Duolink In Situ Detection Reagents	Sigma-Aldrich	Cat# DUO92008
Cytokine array kit	R&D Systems	Cat# ARY005B
cGAMP ELISA Kit	Cayman Chemical	Cat# 501700
EZ-Magna ChIP A/G kit	Millipore	Cat# 17-10086
ADP-Glo kinase assay kit	Promega	Cat# V9101
CUTANA ChIC/CUT&RUN Kit	EpiCypher	Cat# 14-1048
CUTANA CUT&RUN Library Prep Kit	EpiCypher	Cat# 14-1001/2
Quick-DNA Miniprep Plus kit	Zymo Research	Cat# 4203
Quant-iT PicoGreen Kit	Thermo Fisher Scientific	Cat# P11495
KAPA Library Quantification kit	Roche	Cat# KK4824
TruSeq Stranded Total RNA Sample Preparation kit	Illumina	Cat# 20020595
TDE1, Tagment DNA Enzyme and Buffer kit	Illumina	Cat# 15027865
Nextera Index Kit	Illumina	Cat# 15055289
Deposited data
Unprocessed microscopy images and blots	This study	Mendeley Data, V1, doi: 10.17632/2fmtcfwfv4.1
ATAC-seq data	This study	GEO: GSE245447
CUT&RUN data	This study	GEO: GSE245448
RNA-seq and Plate-seq data	This study	GEO: GSE245449
ATAC-seq, CUT&RUN, RNA-seq and Plate-seq data (SuperSeries)	This study	GEO: GSE245450
Mass spectrometry data	This study	PRIDE: PXD046384
Publicly available ATAC-seq data	Liu Y. et al, 2017; Osmanbeyoglu H.U. et al, 2019	GEO: GSE89013 and GSE114964
Publicly available ChIP-seq data	Franco H.L. et al, 2018	GEO: GSE85158
Experimental models: Cell lines
T-47D	ATCC	Cat# HTB-133; RRID:CVCL_0553
HEK293T	ATCC	Cat# CRL-11268; RRID: CVCL_1926
MDA-MB-436	ATCC	Cat# HTB-130; RRID: CVCL_0623
BT-474	ATCC	Cat# CRL-7913, RRID:CVCL_0179
MDA-MB-361	ATCC	Cat# HTB-27, RRID:CVCL_0620
SK-BR-3	ATCC	Cat# HTB-30, RRID:CVCL_0033
BT-20	ATCC	Cat# HTB-19, RRID:CVCL_0178
MDA-MB-231	ATCC	Cat# HTB-26; RRID: CVCL_0062
U2OS	ATCC	Cat# HTB-96, RRID:CVCL_0042
MCF7	ATCC	Cat#HTB-22; RRID:CVCL_0031
PEO1	Toshiyasu Taniguchi	RRID:CVCL_2686
PC-3	Cory Abate-Shen	RRID:CVCL_0035
RPE1-hTERT TP53−/−	Daniel Durocher	N/A
U2OS-Fucci	Stephen Elledge	N/A
SD31-hTERT	Ciccia A. et al, 2009	N/A
Human CD8+ T	Cellero	Cat# 1041
MEF Smarcal1-proficient	Dieter Egli	N/A
MEF Smarcal1-deficient	Dieter Egli	N/A
B16/F10	Charles Drake	RRID:CVCL_0159
Renca	Charles Drake	RRID:CVCL_2174
MC38	Charles Drake	N/A
MDA-MB-436 SMARCAL1-KO	This study	N/A
MDA-MB-436 SMARCAL1-KO_SMARCAL1_WT	This study	N/A
MDA-MB-436_Control_(NLS)-dsRed	This study	N/A
MDA-MB-436_Control (NLS)-dsRed_pHAGE_PD-L1	This study	N/A
MDA-MB-436 SMARCAL1-KO_(NLS)-dsRed	This study	N/A
MDA-MB-436 SMARCAL1-KO_(NLS)-dsRed_pHAGE_PD-L1	This study	N/A
B16/F10 Smarcal1-KO_EV	This study	N/A
B16/F10 Smarcal1-KO_mSmarcal1	This study	N/A
B16/F10 Smarcal1-KO_EV_mCgas-KO	This study	N/A
B16/F10 Smarcal1-KO_mSmarcal1_mCgas-KO	This study	N/A
MDA-MB-436-CRISPRi	This study	N/A
MDA-MB-436-CRISPRi-AAVS1	This study	N/A
MDA-MB-436-CRISPRi-P3	This study	N/A
MDA-MB-436-CRISPRi-AAVS1-gJUN	This study	N/A
MDA-MB-436-CRISPRi-P3-gJUN	This study	N/A
Experimental models: Organisms/strains
NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ	Jackson Laboratories	RRID:IMSR_JAX:005557
C57BL/6J	Jackson Laboratories	RRID:IMSR_JAX:000664
Oligonucleotides
List of oligonucleotides	This study, Table S5	N/A
Recombinant DNA
Plasmid: eCas9-P2A-Puro-gSMARCAL1ex3-1 (sgRNA sequence: GCTCAGAGAGTGTAACGCCC)	Taglialatela A. et al, 2017	N/A
Plasmid: eCas9-P2A-Puro-gSMARCAL1ex3-2 (sgRNA sequence: GTGAGAGCCATTTGACTACG)	Taglialatela A. et al, 2017	N/A
Plasmid: pXPR206	Addgene	RRID:Addgene_96920
Plasmid: pXPR206_Hygro	This study	N/A
Plasmid: pXPR206_ Hygro _gNT (sgRNA sequence: CACGAACTCACACCGCGCGA)	This study	N/A
Plasmid: pXPR206_ Hygro _gJUN (sgRNA sequence: GACCTCCTCACCTCGCCCGAC)	This study	N/A
Plasmid: pXPR206_ Hygro _gELF4 (sgRNA sequence: GAGCCCGATGCCCTGAACAGG)	This study	N/A
Plasmid: pXPR206_ Hygro _gELK4 (sgRNA sequence: AATCAATGTCTGTGTCGATGT)	This study	N/A
Plasmid: pXPR206_ Hygro _gETV3 (sgRNA sequence: CCGGCTGCACATCATTGGTTG)	This study	N/A
Plasmid: pXPR206_ Hygro _gFOSL1 (sgRNA sequence: TGGCACCAGGTGGAACTTCTA)	This study	N/A
Plasmid: pXPR206_ Hygro _gIRF1 (sgRNA sequence: CAGCTCAGCTGTGCGAGTGTA)	This study	N/A
Plasmid: pXPR206_ Hygro _gMIXL1 (sgRNA sequence: CGTACCGGGCCCCCCACGCCG)	This study	N/A
Plasmid: pXPR206_ Hygro _gSMAD5 (sgRNA sequence: CAGGCGTCCATCTAAAGATCT)	This study	N/A
Plasmid: pXPR206_ Hygro _gSNAI2 (sgRNA sequence: ATTATTTCCCCGTATCTCTAT)	This study	N/A
Plasmid: pXPR206_ Hygro _gSP4 (sgRNA sequence: CCAACTAGTAGTTCATCTCTA)	This study	N/A
Plasmid: pXPR206_ Hygro _gTCF4 (sgRNA sequence: GGACCTTCTCATAATGGAGCC)	This study	N/A
Plasmid: pXPR206_ Hygro _gTFDP2 (sgRNA sequence: CTGAGTAACCATTGCTGGTGC)	This study	N/A
Plasmid: pXPR206_ Hygro _gTGIF2LY (sgRNA sequence: GAAGACCAATTTGTCTTTGTT)	This study	N/A
Plasmid: pXPR206_ Hygro _gZBTB7A (sgRNA sequence: AAGATCCGAGCCAAGGCCTTC)	This study	N/A
Plasmid: pLenti-SpBsmBI	Addgene	RRID:Addgene_62205
Plasmid: pLenti-SpBsmBI-gSmarcal1_1 (sgRNA sequence: GAGAAGCGGACCTTTCCGGAG)	This study	N/A
Plasmid: pLenti-SpBsmBI-gSmarcal1_2 (sgRNA sequence: GTTAGGATAGGCACAGCTGCT)	This study	N/A
Plasmid: plentiCRISPRv2-Hygro	Addgene	RRID:Addgene_98291
Plasmid: plentiCRISPRv2-Hygro-gRosa26 (sgRNA sequence: GACTCCAGTCTTTCTAGAAGA)	This study	N/A
Plasmid: plentiCRISPRv2-Hygro-gSmarcal1_1 (sgRNA sequence: GAGAAGCGGACCTTTCCGGAG)	This study	N/A
Plasmid: plentiCRISPRv2-Hygro-gSmarcal1_3 (sgRNA sequence: GAGCAGCTGTGCCTATCCTAA)	This study	N/A
Plasmid: plentiCRISPRv2-Hygro-gSmarcal1_4 (sgRNA sequence: GTTTGGGTTATAAATCCAGCG)	This study	N/A
Plasmid: pSB700-H2B-EYFP-P2A-Puro	Alejandro Chavez	N/A
Plasmid: pSB700-H2B-EYFP-P2A-Puro-gCgas (sgRNA sequence: GTCGGGGCGCGCTTCGCGGA)	This study	N/A
Plasmid: pSB700-H2B-EYFP-P2A-Puro-gP3 (sgRNA sequence: CCACAGGAGTTTACTCACCA)	This study	N/A
Plasmid: pSB700-H2B-EYFP-P2A-Puro-gNT (sgRNA sequence: CACGAACTCACACCGCGCGA)	This study	N/A
Plasmid: LentiGuide-puro-NLS-GFP	Addgene	RRID:Addgene_185473
Plasmid: LentiGuide-puro-NLS-GFP-gP3 (sgRNA sequence: CCACAGGAGTTTACTCACCA)	This study	N/A
Plasmid: LentiGuide-puro-NLS-GFP-gAAVS1 (sgRNA sequence: GGGGCCACTAGGGACAGGAT)	This study	N/A
Plasmid: pdCas9-KRAB-MeCP2	Addgene	RRID:Addgene_110821
Plasmid: pLentidCas9-KRAB-MeCP2	Addgene	RRID:Addgene_122205
Plasmid: pSAM110	This study	N/A
Plasmid: pSAM110-gJUN (sgRNA sequence: TGAGTGTGGCTGAAGCAGCG)	This study	N/A
Plasmid: pXPR206_Hygro	This study	N/A
Plasmid: pDONR223-SMARCAL1_siRNAres_sgRNA_res	This study	N/A
Plasmid: pHAGE-Ct-Flag-HA-SMARCAL1_siRNAres_sgRNAres-hygro	This study	N/A
Plasmid: pDONR223-SMARCAL1_siRNAres	This study	N/A
Plasmid: pDONR223-SMARCAL1-ΔHARP_siRNAres	This study	N/A
Plasmid: pDONR223-SMARCAL1-ΔN1-115_siRNAres	This study	N/A
Plasmid: pDONR223-SMARCAL1-R764Q_ siRNAres	This study	N/A
Plasmid: pHAGE-Ct-Flag-HA-SMARCAL1_siRNAres-Puro	This study	N/A
Plasmid: pHAGE-Ct-Flag-HA-SMARCAL1-ΔHARP_ siRNAres-Puro	This study	N/A
Plasmid: pHAGE-Ct-Flag-HA-SMARCAL1-ΔN1-115_ siRNAres-Puro	This study	N/A
Plasmid: pHAGE-Ct-Flag-HA-SMARCAL1-R764Q_ siRNAres-Puro	This study	N/A
Plasmid: pHAGE-TREX-NtBioID-DEST	Huang J.W. et al, 2020	N/A
Plasmid: pHAGE-TREX-NtBioID-SMARCAL1_siRNAres	This study	N/A
Plasmid: pHAGE-TREX-NtBioID-SMARCAL1-ΔHARP_ siRNAres	This study	N/A
Plasmid: pHAGE-TREX-NtBioID-SMARCAL1-ΔN1-115_ siRNAres	This study	N/A
Plasmid: pDONR223-CD274	ORFeome	Cat# 13856
Plasmid: pHAGE-UbC-DEST	Huang J.W. et al, 2020	N/A
Plasmid: pHAGE-UbC-CD274	This study	N/A
Plasmid: pDONR223-mSmarcal1	This study	N/A
Plasmid: pMSCV-mSmarcal1	This study	N/A
Software and algorithms
Cutadapt	Martin M., 2011	v3.5 or v1.16
edgeR	https://bioconductor.org/packages/release/bioc/html/edgeR.html	v3.32.1
MAGeCK	Li W. et al, 2014	v0.5.9
MAGeCKFlute R package	Wang B. et al, 2019	v1.10.0
MetaXpress Analysis Software	https://mdc.custhelp.com/euf/assets/content/MetaXpress_4_Analysis_Guide.pdf	v6
MaxQuant environment	Cox J. et al, 2011	v2.3.1.0
ColabFold	Mirdita M. et al, 2022	v1.5.2
FlowJo software	https://www.flowjo.com	v10.7
Bowtie2	Langmead B. and Salzberg S.L., 2012	N/A
SICER2	Zang C. et al, 2009	v 1.0.3
deepTools	Ramirez F. et al, 2016	v3.5.1 or v3.3.2
Homer	http://homer.ucsd.edu/homer/	v4.11
HISAT2	Kim D. et al, 2019	v2.1.0
BEDTools	Quinlan A.R. and Hall IM., 2010	V2.27.1
STAR	Dobin A. et al, 2013	v2.5.2b
DESeq2	Love M.I. et al, 2014	v1.30.1
limma package	Ritchie M.I. et al, 2015	v3.46.0
ARACNe algorithm	Lachmann A.G. et al, 2016	N/A
VIPER algorithm	Alvarez M.J. et al, 2016	N/A
QuantStudio	Thermo Fisher Scientific	N/A
GraphPad Prism software	https://www.graphpad.com/features	v7.0
Other

Highlights.

• SMARCAL1 suppresses the cGAS-STING pathway by maintaining genome stability

• SMARCAL1 cooperates with JUN to promote PD-L1 expression

• SMARCAL1 regulates chromatin accessibility at a PD-L1 regulatory element

• SMARCAL1 loss enhances T-cell mediated anti-tumor immunity